article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
the nature of the non - baryonic dark matter , which dominates the visible matter by about 4:1 , is perhaps the most interesting experimental challenge for contemporary particle astrophysics .
a hint for a solution has been found in particle physics where the wimps arise naturally in supersymmetric extensions of the standard model ( e.g. , * ? ? ?
* ) , among other possibilities .
the wimp is typically defined as a stable , electrically neutral , massive particle . assuming that non - baryonic dark matter is dominated by wimps , the pair annihilation cross - section is related to the observed relic density @xcite .
a pair of wimps can annihilate producing ordinary particles and @xmath0-rays .
wimps are expected to form high density clumps according to n - body simulations of test particles with only gravitational interactions @xcite .
the highest density `` free space '' dark matter regions occur for dark matter particles captured within the gravitational potential of adiabatically grown smbhs @xcite .
higher dark matter densities are possible for dark matter particles captured inside of stars or planets .
any star close enough to a smbh can capture a large number of wimps during a short period of time .
annihilation of captured wimps may lead to considerable energy release in stellar cores thus affecting the evolution and appearance of such stars .
such an idea has been first proposed by @xcite and further developed by @xcite who applied it to main - sequence stars .
the model led to the conclusion of suppression of stellar core convection , thus predicting a concentration of stars in the galactic center masquerading as cold red giants .
an order - of - magnitude estimate of the wimp capture rates for stars of various masses and evolution stages @xcite lead us to the conclusion that wds , fully burned stars without their own energy supply , are the most promising candidates to look for . in this paper
we calculate the wimp capture by wds located in a high density dark matter region , and discuss their observational features .
we use current limits on wimp - nucleus interaction and wimp annihilation cross sections , as well as recent estimates of wimp energy density near an adiabatically grown smbh .
in a steady state the wimp capture rate @xmath1 is balanced by the annihilation rate @xcite @xmath2where @xmath3@xmath4 is the velocity averaged wimp pair annihilation cross - section , the effective radius @xmath5is determined by matching the star core temperature @xmath6 with the gravitational potential energy ( assuming thermal equilibrium ) , @xmath7 is the speed of light , @xmath8 is the gravitational constant , @xmath9 is the star core density , and @xmath10 is the wimp mass .
the total number of wimps captured by a star is @xmath11where @xmath12 is the star s age , and the equilibrium time scale is given by @xmath13 the number density distribution of wimps can be estimated as @xcite : @xmath14where @xmath15 is the central wimp number density . in thermal equilibrium ,
the effective radius @xmath16 is determined by the core temperature @xmath6 and density @xmath9 @xmath17where @xmath7 is the speed of light , @xmath8 is the gravitational constant , and @xmath10 is the wimp mass .
limits from direct detection of dark matter on the wimp - nucleon cross - section imply that only a fraction of the wimps crossing the star will scatter and be captured .
the capture rate for a maxwellian wimp velocity distribution ( in the observer s frame ) by a star moving with an arbitrary velocity @xmath18 relative to the observer is given by @xcite : @xmath19where @xmath20 \right.\nonumber\\ & + & \left .
\frac12a_+e^{-a_-^2 } -\frac12a_- e^{-a_+^2}-\eta e^{-\eta^2 } \right\},\nonumber\\ a^2&= & \frac{3v^2(r)\mu}{2\bar{v}^2\mu_-^2},\nonumber\\ a_\pm&=&a\pm\eta,\nonumber\\ \eta&=&\frac{3v_*^2}{2\bar{v}^2 } , \nonumber\\ \chi(a , b)&=&\int_a^b dy\ , e^{-y^2}= \frac{\sqrt\pi}{2}[{\rm erf}(b)-{\rm erf}(a ) ] , \nonumber\end{aligned}\]]@xmath21 is the ambient wimp energy density , @xmath22 is the atomic number of the star s nuclei , @xmath23 is the nucleus mass , @xmath24 is the wimp velocity dispersion , and @xmath25 , @xmath26 .
the escape velocity at a given radius @xmath27 _ inside _ of a star is given by @xmath28^{1/2 } = \left [ \frac{gm_*}{r _ * } \left ( 3 -\frac{r^2}{r_*^2}\right ) \right]^{1/2 } , \label{vesc}\]]where we assumed the same mass density @xmath29 and the same chemical composition over the entire scattering volume @xmath30 .
this is a reasonable assumption for a degenerate electron core . near a smbh , where orbital motion around a single mass dominates , the test particle ( wimp or star ) velocities are keplerian @xmath31 ; in this case @xmath32 , although the exact value does not significantly change the result .
the value of the spin - independent wimp - nucleon scattering cross - section @xmath33 is limited by direct detection experiments , i.e. less than @xmath34 @xmath35 @xcite . if the star is composed of nuclei with atomic number @xmath22 , the cross section increases by a coherent factor of @xmath36 .
if a wd is heavy ( @xmath37 ) and/or @xmath38 , almost all wimps crossing the star will be captured . in this case , the wimp capture rate is determined by the geometrical limit @xmath39 rather than the total interaction cross section @xmath40 .
we thus use a modified interaction cross section @xmath41 defined as @xmath42 figure [ c_rate ] shows the capture rate by oxygen wds ( @xmath43 ) vs. wimp velocity dispersion for several masses of wds , assuming keplerian orbits around the smbh , @xmath44 gev , @xmath45 @xmath35 , @xmath46 gev @xmath47 which corresponds to the maximal central particle dark matter density allowed by the age of the smbh @xmath48 gyr and our selected value @xmath49 @xmath50 s@xmath51 @xcite .
the left panel corresponds to wd effective temperature of @xmath52 k ( without an envelope ) , where the masses and radii used in the calculation are ( fig . 5 in * ? ? ?
* ) : @xmath53 = 0.6 , 0.8 , 1.0 , 1.2 , 1.4 , and @xmath54 = 0.02 , 0.012 , 0.0085 , 0.006 , 0.0045 , correspondingly .
the right panel corresponds to the zero - temperature approximation @xcite , where the mass - radius relation has been obtained by fitting the numerical results for carbon wds with a function @xmath55)$ ] in the interval @xmath56 .
the solid lines show the capture rate calculated using the modified interaction cross section ( eq . [ [ geom_limit ] ] ) , and the dashed lines are calculated for @xmath57 . in the case of oxygen
wds the geometrical limit is reached for @xmath58 ; the larger mass wds have smaller radii and therefore smaller capture rates . for @xmath59 km s@xmath51 our geometrical limit calculations agree well with the results of @xcite , their equation ( 26 ) .
for @xmath60 km s@xmath51 the @xcite formula , derived under the assumption that each wimp crossing the star is captured , gives a systematically larger capture rate , up to a factor of 10 for @xmath61 km s@xmath51 .
this can be treated as an upper limit , whereas our approximation to the geometrically limited case can be considered as a lower limit .
it can be seen ( figure [ c_rate ] ) that cooler wds have a capture rate ( eqs .
[ [ gould2.27]],[[gould2.24 ] ] ) larger than hot ones of the same mass because of the larger escape velocity ( eq . [ [ vesc ] ] ) .
the latter is the result of a smaller radius @xmath62 and consequently stronger gravity .
this effect may be explained in terms of the `` focusing factor '' @xcite or simply because wimps can be captured from the larger volume of the maxwellian velocity phase space .
a larger capture rate by a cooler wd will lead to accelerated heating until the wd radius increases due to increased temperature .
a hotter wd will be less efficient for wimp capture and cool down .
this mechanism will thus lead to fast self - regulation of the wd temperature and capture rate . the capture rate for a different ( @xmath22 ) composition wd can be estimated from scaling the oxygen wd curves by a factor of @xmath63 , e.g.
, the curves for a carbon wd can be obtained from scaling the oxygen wd curves down by a factor of @xmath64 .
wds with heavier nuclei ( up to iron ) may exist @xcite ; in this case , the capture rate is restricted mostly by the geometrical limit .
figure [ c_profile ] shows the capture rate for oxygen ( left panel ) and iron ( right panel ) wds vs. distance from the central black hole with @xmath65 @xcite ; this includes effects of the radial dependence of the wimp velocity dispersion and the wd orbital ( keplerian ) velocity .
following @xcite and @xcite , the wimp mass density is normalized as @xmath66 pc@xmath67 . for the central spike
we assume a power - law profile with indices ( top to bottom ) : 7/3 , 3/2 , 4/3 ; these profiles are predicted for different scenarios of the black hole growth , adiabatic @xcite , quasi - equilibrium @xcite , and instanteneous @xcite , correspondingly . here
we use the same estimate for @xmath68 as for figure [ c_rate ] . the mass - radius relation for oxygen and iron wds of @xmath52 k
is taken from @xcite .
the capture rate scales linearly with the wimp density , so that the largest capture rate is reached with the adiabatic profile .
the capture rate increases toward the smbh until the maximal wimp mass density @xmath68 is reached ; then the capture rate decreases due to increases in the wimp velocity dispersion and the orbital velocity of the star . for the quasi - equilibrium profile , @xmath68 is reached only at @xmath69@xmath70 pc , while the instanteneous profile is even flatter . as can be seen from figure [ c_profile ] and a simple inspection of the capture rate formulae , in the geometrically limited case ( eq . [ [ geom_limit ] ] )
the capture rate becomes essentially independent of the wimp - nucleon scattering cross - section and degenerate core parameters .
observationally , the brightest wimp burners may be the geometrically limited ones .
the main uncertainty in the geometrically limited capture rate is the dark matter density ; thus it may be possible to perform largely model independent " measurements of the dark matter density profile by measuring the luminosity of different wimp burners orbiting within a particular dark matter spike .
a smaller annihilation cross section @xmath71 @xmath50 s@xmath51 would allow for larger ambient wimp densities near the smbh .
this would lead to a larger capture rate and consequently larger burning rate at the innermost radii .
the energy release due to wimp annihilation in the stellar core is @xmath72 erg s@xmath51 which is actually independent of the wimp mass @xmath10 .
where does the energy released during the wimp annihilation go ?
table 1 in @xcite shows that the effective radius of the thermal distribution of wimps in the stellar core is much smaller than the radius of the star @xmath73 , therefore , the products of wimp annihilation can not propagate to the stellar surface and are converted into thermal energy and neutrino emission .
a wd , a star without its own energy supply consisting of carbon and oxygen , may emit up to @xmath74 erg s@xmath51 , i.e. @xmath6910 times the luminosity of the sun , burning wimps only and this energy source will last forever ! ( note that this estimate is based on our _ approximation _ to the geometrically limited case and larger luminosities are possible even for the given set of parameters . ) at such a luminosity , the surface temperature of the wd would be close to @xmath69140,000 k , assuming @xmath75 , @xmath76 .
the maximum of the black body emission falls into the uv band making such stars strong thermal uv emitters concentrated in the inner @xmath690.01 pc . a smaller annihilation cross section @xmath71 @xmath50 s@xmath51 and/or larger wimp density normalization
@xmath77 pc@xmath67 would allow for a larger ambient wimp density near the smbh , thus increasing the capture and burning rates further .
the energy transport in the interiors of wds is dominated by degenerate electrons and is very efficient ( see * ? ? ?
* for a recent review ) ; therefore , the large number of captured wimps and their annihilation in the core would not change the internal structure of wds .
a recently published catalog of spectroscopically confirmed wds from the sloan digital sky survey ( sdss ) @xcite contains several hot wds with surface temperature in the range of 100,000 k , thus providing observational evidence that high temperature does not change the appearance of wds .
a bare wd with an effective temperature as high as 170,000200,000 k has also been observed ( h1504 + 65 , * ? ? ?
the number of very hot wds in the sdss catalog is small , just a handful out of 9316 .
this means that observation of a concentration of very hot wds at the galactic center would be extremely unlikely unless they are `` dark matter burners . ''
the spectra of confirmed hot wds can serve as templates for spectroscopic analysis of wds at the galactic center where only a limited part of the near - ir band can be used .
an independent determination of the @xmath78 ratio is possible using the gravitational redshift that has to be equivalent to a radial velocity of about 50 km s@xmath51 @xcite . a bare wd with a highly eccentric orbit around the central black hole may exhibit variations in brightness correlated with the orbital phase ( figure [ c_profile ] ) . to have this working , the orbital period should exceed the equilibrium time scale @xmath79 ( eq . [ [ taueq ] ] ) .
carbon burning stars have @xmath80 yr , and it is even shorter @xmath690.5 yr in case of a wd @xcite .
if a wd appears in a high - density wimp region , the wimp density in its material would quickly reach equilibrium ; thus , the surrounding wimp density variation as the wd orbits would result in variations of brightness .
this makes bare wds ideal objects to test the wimp density in the environment in which they are orbiting . since @xmath81 , a population of wds , bare or with envelopes , located at different distances from the smbh would exhibit a luminosity correlated with the radial wimp density profile .
geometrically limited wimp burners have the highest luminosities and therefore will be the easiest to observe .
their luminosity is largely independent of wimp - nucleon scattering cross - section , wimp pair annihilation cross - section , and degenerate core parameters .
advances in near - ir instrumentation have made possible observations of stars in the inner parsec of the galaxy @xcite .
the apparent k - band brightness of these stars is 1417 mag .
the observed absorption line widths imply high temperatures and lead to a `` paradox of youth : '' apparently young stars in the region whose current conditions seem to be inhospitable to star formation
. one of the possibilities is that they are old stars masquerading as youths . assuming a central spike with index 7/3 , the k - band brightness for oxygen wds with @xmath82 k and @xmath83 is about 2223 mag not including extinction , which may be as large as 3.3 mag @xcite .
it is , therefore , unlikely that the currently observed stars in the k - band are wds burning wimps ; however , stars with degenerate electron cores plus envelopes can not be ruled out .
we thank r. blandford , j. edsj , j. faulkner , s. kahn , j. primack , and r. romani for interesting discussions and the anonymous referee for useful comments .
l. l. w. would like to thank s. nagataki for interesting discussions on massive stars .
i. v. m. acknowledges partial support from nasa astronomy and physics research and analysis program ( apra ) grant .
a part of this work was done at stanford linear accelerator center , stanford university , and supported by department of energy contract de - ac03 - 768sf00515 .
|
we show that a star orbiting close enough to an adiabatically grown supermassive black hole ( smbh ) can capture weakly interacting massive particles ( wimps ) at an extremely high rate .
the stellar luminosity due to annihilation of captured wimps in the stellar core may be comparable to or even exceed the luminosity of the star due to thermonuclear burning .
the model thus predicts the existence of unusual stars , essentially wimp burners , in the vicinity of a smbh .
we find that the most efficient wimp burners are stars with degenerate electron cores , e.g. white dwarfs ( wds ) ; such wds may have a very high surface temperature . if found , such stars would provide evidence for the existence of particle dark matter and can possibly be used to establish its density profile . on the other hand ,
the lack of such unusual stars may provide constraints on the wimp density near the smbh , as well as the wimp - nucleus scattering and pair annihilation cross - sections .
|
rank rigidity was first proved in the higher euclidean rank setting by ballman @xcite and , using different methods , by burns and spatzier @xcite .
a manifold is said to have higher euclidean rank if a parallel , normal jacobi field can be found along every geodesic .
ballman and burns - spatzier proved that if an irreducible , compact , nonpositively curved manifold has higher euclidean rank , then it is locally symmetric .
ballman s proof works for finite volume as well and the most general version of this theorem is due to eberlein and heber , who prove it under only a dynamical condition on the isometry group of @xmath0 s universal cover @xcite .
hamenstdt showed that a compact manifold with curvature bounded above by -1 is locally symmetric if along every geodesic there is a jacobi field making curvature -1 with the geodesic direction @xcite .
she called this situation higher hyperbolic rank .
shankar , spatzier and wilking extended rank rigidity into positive curvature by defining spherical rank .
a manifold with curvature bounded above by 1 is said to have higher spherical rank if every geodesic has a conjugate point at @xmath5 , or equivalently , a parallel vector field making curvature 1 with the geodesic direction .
they proved that a complete manifold with higher spherical rank is a compact , rank one locally symmetric space @xcite .
these results settle many rank rigidity questions , but leave questions about other curvature settings open ( see @xcite for an excellent overview ) . in this paper
we prove the following theorem , which can be applied to various settings in nonpositive curvature .
let @xmath0 be a compact , ( euclidean ) rank 1 , nonpositively curved manifold .
suppose that along every geodesic in @xmath0 there exists a parallel vector field making sectional curvature @xmath1 with the geodesic direction . if @xmath0 is odd dimensional , or if @xmath0 is even dimensional and satisfies the sectional curvature pinching condition @xmath2 with @xmath3 then @xmath0 has constant negative curvature equal to @xmath1 . in strict negative curvature ,
the proof of this theorem is considerably simpler .
in fact , for negative curvature the full frame flow is ergodic under the conditions of theorem 1 in all dimensions but 7 and 8 ( @xcite for odd dimensions , @xcite for even dimensions ) .
then the proof of theorem 1 is immediate by considering a frame with dense orbit . for dimensions 7 and 8
ergodicity of the @xmath6-frame flow holds under very strong curvature pinching ( see @xcite ) but under the curvature restrictions of theorem 1 one only has ergodicity of the 2-frame flow .
note that ergodicity of this flow alone does not establish theorem 1 since the set of 2-frames giving the distinguished sectional curvature @xmath1 may , a priori , have zero measure .
however , the ergodic theory of these types of flows , developed by brin , proceeds via explicit geometric descriptions of the ergodic components , and this allows theorem 1 to be obtained from the 2-frame flow dynamics alone .
the proof proceeding via 2-frame flow gives the result in the exceptional dimensions 7 and 8 in negative curvature .
in addition , it suggests an adaptation to the rank 1 nonpositive curvature setting , where the ergodic theory of frame flows has not been developed .
the simplifications possible in the strictly negative curvature setting will be noted throughout the paper , but here we observe that although obtaining this paper s result for nonpositively curved rank 1 spaces necessitates a more technical proof , the resulting theorem forms a better complement to the rank rigidity theorem of ballmann and burns - spatzier .
note that , unlike previous rank rigidity results , theorem 1 allows for situations where the distinguished curvature @xmath1 is not extremal .
however , in the cases where @xmath1 is extremal the hypotheses of our theorem can be weakened , as demonstrated in section [ sec : parallel ] of this paper .
the following two results are then easy corollaries of theorem 1 : let @xmath0 be a compact , rank 1 manifold with sectional curvature @xmath7 .
suppose that along every geodesic in @xmath0 there exists a jacobi field making sectional curvature @xmath8 with the geodesic direction . if @xmath0 is odd dimensional , or if @xmath0 is even dimensional and satisfies the sectional curvature pinching condition @xmath9 then @xmath0 is hyperbolic . _
( compare with hamenstdt @xcite ) _ let @xmath0 be a compact manifold with sectional curvature bounded above by @xmath8 .
suppose that along every geodesic in @xmath0 there exists a jacobi field making sectional curvature @xmath8 with the geodesic direction .
if @xmath0 is odd dimensional , or if @xmath0 is even dimensional and satisfies the sectional curvature pinching condition @xmath10 then @xmath0 is hyperbolic . in corollary 1
we obtain a new rank rigidity result analogous to those described above .
this is the first positive result for lower rank , i.e. when the distinguished curvature value is the lower curvature bound ; @xcite provides a discussion of counterexamples to lower spherical and euclidean rank rigidity . in corollary 2
we obtain a shorter proof of hamenstdt s result , under an added pinching constraint in even dimension . in @xcite
, connell showed that rank rigidity results can be obtained using only a dynamical assumption on the geodesic flow , namely an assumption on the lyapunov exponents at a full measure set of unit tangent vectors .
his paper deals with the upper rank situations treated by ballman , burns - spatzier and hamenstdt .
he proves that having the minimal lyapunov exponent allowed by the curvature restrictions attained at a full measure set of unit tangent vectors is sufficient to apply the results of ballman and burns - spatzier or hamenstdt . in the lower rank
setting of this paper , this viewpoint translates into let @xmath0 be a compact , rank 1 manifold with sectional curvature @xmath11 , where @xmath12 .
endow @xmath13 with a fully supported ergodic measure ; one can take the measure of maximal entropy or , if the curvature is known to be negative , the liouville measure .
suppose that for a full measure set of unit tangent vectors @xmath14 on @xmath0 the maximal lyapunov exponent at @xmath14 is @xmath15 , the maximum allowed by the curvature restriction . if @xmath0 is odd dimensional , or if @xmath0 is even dimensional and satisfies the sectional curvature pinching condition @xmath16 with @xmath17 then @xmath0 is of constant curvature @xmath1 .
the adaptation of connell s arguments for this setting is discussed in section [ sec : lyap ] .
the proof of theorem 1 relies on dynamical properties of the geodesic and frame flows on nonpositively curved manifolds .
the work of brin and others is the starting point for the proof ; the results needed are summarized in section [ sec : background ] ( see also @xcite for a survey of the area ) .
although none of his work is undertaken for rank 1 nonpositively curved manifolds , the ideas used in this paper to deal with that situation are clearly inspired by brin s work .
the proof will proceed as follows .
we utilize the transitivity group @xmath18 , defined for any vector @xmath14 in the unit tangent bundle of @xmath0 , which acts on @xmath19 . essentially
, elements of @xmath18 correspond to parallel translations around ideal polygons in @xmath0 s universal cover . in negative curvature
, brin shows that this group is the structure group for the ergodic components of the frame flow ( see e.g @xcite or @xcite ) . for the rank 1 nonpositive curvature case the definition of this group
must be adjusted and we use only that it is the structure group for a subbundle of the frame bundle . the considerations for the rank 1 case are discussed in section [ sec : rank1 ] . in section [ sec : trans ] we show that @xmath18 preserves the parallel fields that make curvature @xmath1 with the geodesic defined by @xmath14 . finally , we apply results of brin - gromov ( adapted to the rank 1 case ) and brin - karcher on the 2-frame flow which imply that @xmath18 acts transitively on @xmath20 and conclude that the curvature of @xmath0 is constant .
i would like to thank chris connell for discussions helpful with the arguments in section [ sec : parallel ] of this paper , jeffrey rauch for the proof of lemma [ rauchlemma ] , and ben schmidt for helpful comments on this paper .
in particular , special thanks are due to my advisor , ralf spatzier , for suggesting this problem , for help with several pieces of the argument and for helpful comments on this paper .
let us begin by fixing the following notation : * @xmath0 : a compact , rank 1 riemannian manifold with nonpositive sectional curvature , @xmath21 its universal cover , @xmath22 the boundary at infinity .
* @xmath13 and @xmath23 : the unit tangent bundles to @xmath0 and @xmath21 , respectively .
* @xmath24 : the @xmath25-frame bundle of ordered , orthonormal @xmath25-frames on @xmath0 .
* @xmath26 : the geodesic flow on @xmath13 or @xmath23 . * @xmath27 : the frame flow on @xmath24 ; when clear , @xmath25 will not be referenced . * @xmath28 and @xmath29 : the foliations of @xmath23 given by inward and outward pointing normal vectors to horospheres .
* @xmath30 : the bowen - margulis measure of maximal entropy on @xmath13 .
* @xmath31 : the geodesic in @xmath0 or @xmath21 with velocity @xmath14 at time 0 .
* @xmath32 : a parallel normal vector field along @xmath31 making the distinguished curvature @xmath1 with @xmath33 . * @xmath34 : the sectional curvature operator .
note that @xmath35 mapping a frame to its first vector is a fiber bundle with structure group @xmath36 acting on the right ; @xmath37 is a principal bundle .
the measure @xmath30 is used in place of the standard liouville measure as it has better known dynamical properties for rank 1 , nonpositively curved spaces ( see section [ sec : rank1 ] ) . in the negative curvature setting liouville measure can be used .
unless otherwise specified , @xmath30 and its product with the standard measure on the fibers of @xmath24 inherited from the haar measure on @xmath38 will be the measures used in all that follows . in negative curvature , @xmath28 and @xmath29
are the stable and unstable foliations for the geodesic flow . in negative curvature
, brin develops the ergodic theory of frame flows as summarized below . in section [ sec : rank1 ] we will discuss how suitable portions of this setup can be generalized to the rank 1 setting .
first , the frame flow also gives rise to stable and unstable foliations @xmath39 and @xmath40 of @xmath24 as shown by brin ( see @xcite prop .
brin notes that the existence of these foliations can be established in two ways , either by applying the work of brin and pesin on partially hyperbolic systems or by utilizing the exponential approach of asymptotic geodesics . in the second approach the leaves of the foliation are constructed explicitly - they sit above the stable / unstable leaves for the geodesic flow , and @xmath41 and @xmath42 are in the same leaf if the distance between @xmath43 and @xmath44 goes to zero as @xmath45 for the stable leaves , or @xmath46 for the unstable leaves . here
we present a proposition that makes possible this definition of @xmath47 by establishing that frames asymptotic to @xmath41 exist and are unique .
the proof follows the sketch given by brin in @xcite .
[ prop : leaf ] let @xmath14 be a unit tangent vector and let @xmath41 be a @xmath25-frame with first vector @xmath14 .
let @xmath48 ( respectively @xmath49 ) so that the distance between @xmath50 and @xmath51 goes to zero exponentially fast as @xmath45 ( resp .
@xmath46 ) . then there exists a unique @xmath25-frame @xmath42 with first vector @xmath52 such that the distance between @xmath43 and @xmath44 goes to zero as @xmath45 ( resp .
@xmath46 ) .
note that in a compact , negatively curved any two asymptotic vectors approach each other exponentially fast so this proposition allows us to define all leaves of the foliation . in rank 1
spaces this may no longer be the case ; thus we have added exponential approach as a hypothesis to the proposition as it will be used for the rank 1 case later in the paper .
assume @xmath48 ; the unstable case is analogous .
uniqueness of the limit is simple since it is clear that two different frames can not both approach @xmath41 .
we thus have only to show existence .
for @xmath53 large enough , @xmath50 and @xmath51 are very close to each other , and then for every frame @xmath54 with first vector @xmath50 there exists a unique frame , call it @xmath55 , which minimizes the distance from @xmath54 among frames with first vector @xmath51 .
thus to approximate the unique frame @xmath42 we are looking for , consider the frames @xmath56 .
we want to show that the @xmath57 have a limit as @xmath45 ; this limit will clearly be our @xmath42 .
consider the sequence @xmath58 for @xmath59 .
since the frame flow is smooth , by choosing large enough @xmath60 the difference between the frame flow along the segments @xmath61 $ ] and @xmath62 $ ] can be made arbitrarily small , and thus if the @xmath58 have a limit , it must be a limit for the @xmath57 . again
, since the frame flow is smooth and since the distance between the geodesics decreases exponentially fast , the distance @xmath63 goes to zero exponentially fast as well .
note then that @xmath64 .
as the summands go to zero exponentially fast the last sum shown converges and given @xmath65 one can pick @xmath66 so large that this tail sum is less than @xmath67 . then we see that the @xmath58 form a cauchy sequence so they have a limit as desired .
let @xmath68 be the map from the fiber of @xmath24 over @xmath14 to the fiber over @xmath52 that takes each @xmath41 to @xmath69 .
note that @xmath68 corresponds to a unique isometry between @xmath70 and @xmath71 and commutes with the right action of @xmath36 . for most of this paper
we will consider the maps @xmath68 acting on 2-frames .
one can think of @xmath72 as the result of parallel transporting @xmath41 along @xmath31 out to the boundary at infinity of @xmath21 and then back to @xmath52 along @xmath73 . if @xmath52 and @xmath14 belong to the same leaf of @xmath29 there is similarly an isometry corresponding to parallel translation to the boundary at infinity along @xmath74 and back along @xmath75 .
we will also denote this isometry by @xmath68 . in the spirit of brin ( see @xcite defn .
4.4 ) we define the transitivity group at @xmath14 as follows : given any sequence @xmath76 with @xmath77 such that each pair @xmath78 lies on the same leaf of @xmath28 or @xmath29 we have an isomorphism of @xmath70 given by @xmath79 the closure of the group generated by all such isometries is denoted by @xmath18 and is called the transitivity group .
the idea of the transitivity group is that it is generated by isometries coming from parallel translation around ideal polygons in @xmath21 with an even number of sides , such as the one shown in figure [ fig : rectangle ] . ]
note that this definition differs slightly from that in brin s work .
brin requires that @xmath80 and thus there is no @xmath81 term in his formula for @xmath82 brin proves that his group describes the ergodic components of the frame flow .
he shows in @xcite that the ergodic components are subbundles of @xmath24 with structure group a closed subgroup of @xmath38 , now acting from the left ( see also @xcite section 5 for an overview ) .
in addition , his proof demonstrates that the structure group for the ergodic component is the transitivity group ( see @xcite remark 2 or @xcite proposition 2 ) .
note that the action of @xmath18 can be taken to be a left action as it commutes with the @xmath36 action of the structure group .
this can be seen from noting that @xmath83 for any @xmath84 in the structure group , and that these maps define the transitivity group .
the proof that this group gives the ergodic components follows the hopf argument for ergodicity , showing that the ergodic component is preserved under motion along stable and unstable leaves , and using the birkhoff ergodic theorem to show that switching from stable to unstable also preserves the component .
the transitivity group as defined here is certainly at least as large as that defined by brin . on the other hand , the addition of the @xmath81 term here certainly preserves the ergodic components so this group still describes ergodic components and therefore is , in the end , the same as brin s .
the advantage to this formulation of the definition is that it allows all ideal polygons , not just those that are ` equilateral ' in the sense that they can be traversed only by following leaves of the foliations .
the explicit geometric description of the ergodic components given here is the central inspiration for our proof .
we use two results on the ergodicity of the 2-frame flow in our proof .
[ thm : bg ] _
( brin - gromov @xcite proposition 4.3 ) _ if @xmath0 has negative sectional curvature and odd dimension then the 2-frame flow is ergodic .
[ thm : bk ] _ ( brin - karcher @xcite ) _ if @xmath0 has sectional curvature satisfying @xmath2 with @xmath85 then the 2-frame flow is ergodic .
theorem [ thm : bk ] is not directly stated as above in @xcite , rather it follows from remarks made in section 2 of that paper together with proposition 2.9 and the extensive estimates carried out in the later sections .
the following dynamical lemma is one of the main tools in our proof .
it will be used in the proof of lemma [ lemma : existence ] and in the arguments of section [ sec : trans ] .
[ recurrencelemma ] suppose @xmath86 is a recurrent geodesic in @xmath0 with a parallel normal field @xmath87 along it such that @xmath88 as @xmath89 .
then @xmath90 for all @xmath53 .
since @xmath86 is recurrent we can take an increasing sequence @xmath91 tending to infinity such that @xmath92 approaches @xmath93 .
since the parallel field @xmath87 has constant norm and the set of vectors in @xmath94 with this norm is compact , we can , by passing to a subsequence , assume that @xmath95 has a limit @xmath96 . extend @xmath96 to a parallel vector field @xmath97 along @xmath86 . by construction , @xmath98 .
in addition , for any real number @xmath60 , the recurrence @xmath99 implies recurrence @xmath100 . by continuity of the frame flow
, we get that @xmath101 for the vector field @xmath102 defined above .
thus @xmath97 makes curvature @xmath103 with @xmath104 for any time @xmath53 .
we can repeat the same argument as above , letting @xmath97 recur along the same sequence of times to produce @xmath105 , and likewise @xmath106 recur to produce @xmath107 , forming a sequence of fields all making curvature identically @xmath103 with the geodesic direction .
now , observe that @xmath108 for some @xmath109 .
note here that @xmath84 is not well - defined by looking at @xmath110 and @xmath102 alone , but will be well defined if we consider @xmath66-frame orbits with second vector @xmath110 recurring to @xmath66-frames with second vector @xmath96 ; this is the @xmath84 we utilize . by construction and the fact that the @xmath38 action commutes with parallel translation , @xmath111 .
@xmath38 is compact , so the @xmath112 have convergent subsequences .
in addition , since the terms of this sequence are all iterates of a single element , we can , by adjusting terms of such a subsequence by suitable negative powers of @xmath84 , have the subsequence converge to the identity .
choose a subsequence @xmath113 such that @xmath114 as @xmath115 .
these @xmath116 approach our original field @xmath87 showing that @xmath110 makes constant curvature @xmath103 with @xmath117 as well .
in this section we discuss some details of the extension to rank 1 , nonpositively curved spaces , and note how the necessary results on the dynamics of the frame flow can be appropriated to this situation .
first we discuss the measure of maximal entropy @xmath30 .
this measure was developed for rank 1 spaces by knieper @xcite and is constructed there as follows .
let @xmath118 be the patterson - sullivan measures on @xmath119 .
fix any @xmath120 .
let @xmath121 be the set of pairs @xmath122 in @xmath119 that can be connected by a geodesic .
then @xmath123 defines a measure on @xmath121 ; @xmath124 is a positive function which can be chosen to make the measure invariant under @xmath125 .
let @xmath126 be the projection @xmath127 .
we then get a @xmath26 and @xmath128 invariant measure @xmath30 on @xmath129 by setting , for any borel set @xmath130 of @xmath129 @xmath131 where here @xmath132 is the standard projection and @xmath133 is the volume element on the submanifolds @xmath134 .
we need three key facts about this measure .
first , @xmath30 is ergodic for the geodesic flow ( see @xcite theorem 4.4 ) .
second , @xmath30 has full support .
this follows from the facts that @xmath30 is supported on the rank 1 vectors ( see @xcite again ) and that the rank 1 vectors are dense in @xmath13 ( see e.g. @xcite ) .
third , @xmath30 is absolutely continuous for the foliations @xmath28 and @xmath29 .
absolute continuity of a measure for a foliation is a way of asking that a fubini - like property hold for the foliation when integrating with respect to the measure . in our situation
, we can justify fubini s theorem for this measure and the foliations @xmath28 and @xmath29 directly .
up to multiplication by the positive function @xmath124 , @xmath30 is locally a product measure .
variation of the set @xmath130 in the @xmath28 and @xmath29 directions is measured by the @xmath135 measures as well as the part of @xmath133 that measures variation normal to the geodesic flow direction when @xmath134 is a flat of dimension greater than one .
the rest of the measure @xmath133 measures variation of the set @xmath130 in the geodesic flow direction . combining these measures together as a product measure with the scaling by @xmath124 yields @xmath30 , demonstrating that fubini - style arguments for this measure hold .
next , we want to extend the definition of the transitivity group to rank 1 , nonpositively curved spaces spaces .
the central difficulty here is that the distance between asymptotic geodesics may approach a nonzero constant .
thus it is no longer clear whether foliations like @xmath39 and @xmath40 can be defined .
we deal with this difficulty by avoiding defining foliations for the frame flow , but still defining maps @xmath68 used to produce a transitivity group . in section [ sec : trans ] we will show that the transitivity group preserves the distinguished parallel fields .
the technical points involved in the definitions of the @xmath136 maps and the transitivity group are necessary to make that proof work .
first , let @xmath137 be the set of endpoints @xmath138 of recurrent , rank 1 geodesics such that almost all geodesics ending at @xmath138 are recurrent .
since recurrence and rank 1 are full measure conditions for the ergodic measure @xmath30 and @xmath30 is absolutely continuous for @xmath26 this is a full measure set of @xmath21 for any patterson - sullivan measure on @xmath21 . in particular , it is dense as the patterson - sullivan measures have full support .
the first two conditions are placed on @xmath139 to allow the proof of the following lemma ; the third will be needed for our work in section [ sec : trans ] .
[ lemma : strictlyasymptotic ] there exists a full measure subset @xmath140 such that if @xmath141 then any two geodesics @xmath142 and @xmath143 with @xmath144 are exponentially strictly asymptotic , that is , the distance between them goes to zero exponentially fast as they head towards @xmath138 .
first we show that the distance between @xmath142 and @xmath143 goes to zero ( see also @xcite prop .
4.1 ) . as @xmath145
it is the end point of a rank 1 recurrent geodesic , call this geodesic @xmath146 .
suppose @xmath142 and @xmath143 are not strictly asymptotic .
then @xmath146 is not strictly asymptotic to one of these geodesics , without loss of generality , say @xmath142 . since @xmath146 is recurrent there
exists a sequence @xmath147 of isometries of @xmath148 and a sequence of real numbers @xmath149 tending to infinity such that @xmath150 as @xmath151 .
consider the sequences of geodesics @xmath152 and @xmath153 .
by choice of the @xmath154 the first sequence converges to @xmath146 .
also , since @xmath142 is asymptotic to @xmath146 , after perhaps passing to a subsequence , the second sequence converges to a geodesic , call it @xmath155 .
as @xmath146 and @xmath142 are not _ strictly _ asymptotic , @xmath156 , but since they are asymptotic , @xmath157 and @xmath158
. then the flat strip theorem ( see @xcite ) implies that @xmath146 and @xmath155 bound a totally geodesically embedded flat strip , contradicting the fact that @xmath146 is rank 1 .
now we show that this convergence is exponential for almost all @xmath138 .
since the manifold is rank 1 and the geodesic flow is ergodic for @xmath30 , @xmath30 is a hyperbolic measure for @xmath26 ( see the supplement on pesin theory to @xcite ) .
then pesin theory tells us that there exists a geodesic @xmath159 whose central foliation consists solely of @xmath160 . then every stable jacobi field
@xmath161 along @xmath159 eventually decreases in magnitude .
in fact , if @xmath162 is a basis of the stable jacobi fields , we can pick @xmath163 large enough that for any @xmath164 , @xmath165 for some real number @xmath15 a positive fraction of the time @xmath166 $ ] .
as this is true for the basis , it is true for any stable jacobi field .
then by continuity of the geodesic flow and its derivatives the jacobi fields , we can chose a small @xmath65 and consider the set of all geodesics @xmath167 with @xmath168 for @xmath169 $ ] so that we have for all stable jacobi fields @xmath170 along @xmath167 that @xmath171 a positive fraction of the time @xmath172 $ ] . consider pairs of asymptotic geodesics @xmath173 and @xmath174 in this set .
we claim that the distance between the geodesics decreases over the interval @xmath172 $ ] at least by a fraction bounded away from 0 which is independent of the choice of @xmath173 and @xmath174 .
to see this , take @xmath175 to be the shortest path in @xmath176 with @xmath177 and @xmath178 .
the geodesics @xmath179 give a variation of asymptotic geodesics connecting @xmath173 and @xmath174 .
now , at any @xmath180 $ ] the distance between @xmath181 and @xmath182 along @xmath176 is @xmath183 now @xmath184 for a fixed @xmath185 is a stable jacobi field along the geodesic @xmath186 . by our choice of the @xmath67-neighborhood in which @xmath187 lies , we know that we have @xmath188 a positive fraction of the time , independently of our choice of @xmath173 and @xmath174 . since this is true for all the jacobi fields in the integrand for equation [ eqn : d ] we see that @xmath189 will decrease over the interval @xmath166 $ ] by at least by some positive fraction independent of the choice of @xmath173 and @xmath174 .
if this is true for the distance @xmath189 then it is true for the actual distance in @xmath0 between @xmath181 and @xmath182 so this distance also decreases by a positive fraction .
now , let @xmath190 be a small enough neighborhood of @xmath191 in @xmath13 that all @xmath192 remain within @xmath193 of @xmath194)$ ] for the next @xmath60 units of time .
by the birkhoff ergodic theorem applied to the geodesic flow , there exists a full measure set of @xmath192 such that @xmath50 returns to @xmath190 a positive fraction of the time
. let @xmath140 be the endpoints at infinity of the geodesics generated by such vectors .
we now show that this is the set we are looking for .
let @xmath167 be a geodesic asymptotic to @xmath146 for @xmath195 .
we show that it approaches @xmath146 exponentially , and hence any two geodesics asymptotic to @xmath146 approach each other exponentially .
we know from the first part of this proof that the distance between @xmath167 and @xmath146 goes to zero , so after some finite amount of time the distance between them is less than @xmath193 . then for a positive fraction of the time after that
we know @xmath196 is in @xmath190 and hence both @xmath146 and @xmath167 will spend a positive fraction of the time in the @xmath67-neighborhood of @xmath194)$ ] discussed above . as shown in the previous paragraph , during each visit to this neighborhood of @xmath159
the distance between @xmath167 and @xmath146 decreases at least by a set fraction .
as this happens a positive fraction of the time as @xmath53 goes to infinity , we see that @xmath167 approaches @xmath146 exponentially fast , proving the lemma .
let @xmath48 ( the case @xmath197 proceeds in a similar manner ) .
we make definitions of @xmath68 in two cases .
+ * case i : * @xmath198 .
lemma [ lemma : strictlyasymptotic ] tells us that we have exponential convergence of the geodesics in question
. then proposition [ prop : leaf ] allows us to define @xmath68 mapping frames over @xmath14 to frames over @xmath52 as in the negative curvature case .
+ * case ii : * @xmath199 .
we define a family of maps @xmath200 in the following manner .
let @xmath201 .
consider all sequences of points @xmath202 in @xmath203 that approach @xmath138 in the sphere topology on @xmath119 .
as noted , @xmath203 is dense in @xmath119 , so we can find such sequences approaching any @xmath138 .
let @xmath204 and @xmath205 be the geodesics connecting the footpoints of @xmath14 and @xmath52 to @xmath206 such that @xmath207 is the footpoint of @xmath14 and @xmath208 ( see figure [ fig : defn ] ) .
the maps @xmath209 are defined under case i. as @xmath66 tends to infinity , @xmath210 and @xmath211 so limit points of the maps @xmath212 will give maps from frames over @xmath14 to frames over @xmath52 . let us restrict the allowed sequences @xmath202 to only those for which @xmath212 has a unique limit ; call that limit @xmath213
. these will be the allowed maps for the second case of the definition .
+ as before , the transitivity group will be defined as a composition of the @xmath136 and @xmath214 maps corresponding to translations around ideal polygons .
again , nonpositive curvature necessitates some technical considerations .
first , as noted in @xcite 1.11 , not all pairs of points on @xmath22 can be connected by geodesics .
however , we note in the following lemma , that a given point can be connected to almost all other points on @xmath22 .
[ lemma : connect ] let @xmath138 be an element of @xmath22 and let @xmath215 be the subset of @xmath22 consisting of points that can be connected to @xmath138 by geodesics
. then @xmath215 contains an open , dense set .
this lemma simply works out some consequences of @xcite .
let @xmath216 be the set of all points at infinity which can be connected to @xmath138 by a geodesic that does not bound a flat half plane .
ballmann s theorem 2.2 ( iii ) tells us that @xmath217 contains all endpoints of periodic geodesics that do not bound a flat half plane .
together with his theorem 2.13 , this implies that the set @xmath217 is dense .
in addition , ballmann s lemma 2.1 implies that @xmath217 is open , proving the lemma .
we also introduce the following technical criterion : + _ * composition criterion . * _ let @xmath14 be in @xmath23 and let @xmath218 be the endpoints at infinity of @xmath146 .
let @xmath219 and @xmath220 be sequences in @xmath203 .
we will say the pair @xmath221 satisfies the _ composition criterion for @xmath14 _ if @xmath222 for all @xmath66 and @xmath223 as @xmath224 , where @xmath225 is the geodesic connecting @xmath206 and @xmath226 .
+ this criterion will be required of pairs @xmath221 if we are to compose the maps @xmath214 and @xmath227 in forming elements of the transitivity group
. it will be important in the proof of proposition [ prop : case2 ] .
we must first , however , establish that , given @xmath202 and @xmath14 , there exist sequences @xmath228 that satisfy the composition criterion for @xmath14 with @xmath202 . without this fact the definition we will make of the transitivity group could be vacuous .
[ lemma : existence ] given @xmath229 with @xmath230 and a sequence @xmath231 in @xmath203 , there exists a sequence @xmath232 in @xmath203 such that @xmath202 and @xmath228 satisfy the composition criterion for @xmath14 . for
any @xmath233 let @xmath234 be the projection defined by setting @xmath235 equal to @xmath236 where @xmath237 is the geodesic with @xmath238 and @xmath239 .
let @xmath240 be the footpoint of @xmath14 and @xmath241 be the ball of radius @xmath242 around @xmath240 in @xmath21 .
lemma 3.5 in @xcite tells us that given @xmath240 , there exists an @xmath243 such that @xmath244 contains an open set @xmath190 in @xmath22 .
examining knieper s proof one sees that @xmath242 can be taken to be any number greater than the distance from @xmath240 to some rank 1 geodesic @xmath159 which has an endpoint at @xmath206 .
consider the geodesic joining @xmath240 and @xmath206 . since @xmath245 , it is the endpoint of a rank 1 , recurrent geodesic , call it @xmath237 .
lemma [ lemma : strictlyasymptotic ] then implies that the geodesic @xmath204 from @xmath240 to @xmath206 is strictly asymptotic to @xmath237 and then lemma [ recurrencelemma ] can be applied with @xmath246 to show that @xmath204 must be rank 1 as @xmath237 is .
thus the rank 1 geodesic @xmath159 needed for knieper can be taken to be @xmath204 .
it is distance 0 from @xmath240 , therefore @xmath242 can be taken to be arbitrarily small ; in particular , take @xmath247 for @xmath206 . for each @xmath248 , the open set @xmath249 given to us by knieper
s lemma contains elements of the set @xmath250 since @xmath251 is open and dense and @xmath203 has full measure .
pick @xmath252 to form our sequence @xmath228 .
then the geodesics @xmath225 connecting @xmath206 and @xmath226 enter @xmath253 for all @xmath66 , and as @xmath254 and @xmath255 we must have @xmath256 .
thus @xmath221 satisfies the composition criterion for @xmath14 as desired .
the transitivity group is defined via the following two definitions .
we start by defining its action on the frames above one particular vector @xmath14 .
[ defn : rk1transgpv ] let @xmath192 be such that @xmath257 and @xmath258 are in @xmath203 . consider any sequence @xmath76 with @xmath259 for some real @xmath60 , such that each pair @xmath78 lies on the same leaf of @xmath28 or @xmath29 .
furthermore , take for each pair @xmath78 falling under case ii a choice of a sequence @xmath260 as described above .
we require that @xmath261 satisfies the composition criterion for @xmath262 .
then we have an isomorphism of @xmath70 given by @xmath263 here @xmath264 when @xmath78 falls under case i and @xmath265 when @xmath78 falls under case ii .
the closure of the group generated by all such isometries is denoted by @xmath18 .
we extend the action of this group to any @xmath266 by connecting @xmath14 to @xmath267 by a segment of an ideal polygon . to do
so we simply need a point @xmath138 in @xmath268 . by lemma [ lemma :
connect ] this set is open and dense , so in fact we can choose @xmath269 .
let @xmath84 be the isometry from @xmath70 to @xmath270 given by frame flow along the segment connecting @xmath14 and @xmath267 via @xmath138 .
more specifically , let @xmath271 lie on the geodesic connecting @xmath257 and @xmath138 such that @xmath272 , let @xmath273 lie on the geodesic connecting @xmath138 and @xmath274 such that @xmath275 , and let @xmath163 be such that @xmath276
. then let @xmath277 in the negative curvature case , it is clear that @xmath278 .
thus we complete the definition of the transitivity group by making the following [ defn : rk1transgp ] let @xmath279 note that the choices of @xmath14 and @xmath138 only affect the group @xmath280 up to multiplication by an element of @xmath18 , so the specific choices are not relevant .
in addition , attempting to define elements of @xmath280 for vectors
@xmath267 for which neither endpoint is in @xmath203 by ideal polygons based at @xmath281 is problematic as , due to the composition criterion , the composition of such elements may not be in the group .
hence we have define such @xmath280 via @xmath18 where no such issues arise . in the end
we have a well defined action of an abstract group @xmath282 isomorphic to @xmath18 on the frame bundle , which in the negatively curved case essentially reduces to brin s definition .
again , as the @xmath283 maps constructed here are invariant under elements of the structure group @xmath36 , the action of @xmath282 commutes with the action of @xmath36 and thus takes the form of a left action .
we now construct a subbundle of @xmath24 for any @xmath284 with an action of @xmath282 on it .
[ defn : subbundle ] given a @xmath25-frame @xmath41 based above a vector @xmath195 let @xmath285 be the smallest set containing @xmath41 and closed under all @xmath286 , @xmath27 for all @xmath53 and all isometries @xmath84 as in equation [ eqn : g ] .
[ prop : subbundle ] @xmath287 is a subbundle of @xmath24 .
since for any @xmath266 we have found an isometry @xmath84 as in definition [ defn : rk1transgp ] , we see that @xmath288 .
let @xmath289 be an extension of the @xmath25-frame @xmath41 to an @xmath66-frame with first @xmath25 vectors given by @xmath41 .
we first show that @xmath290 is a subbundle . by construction
, @xmath290 admits an action of @xmath282 , an abstract group isomorphic to @xmath18 .
it is clear that @xmath291 is the @xmath280 orbit of @xmath292 for any @xmath266 , where @xmath84 is as in equation [ eqn : g ] .
furthermore , @xmath282 acts freely on @xmath37 so all orbit types of this action are the same .
thus theorem 5.8 from @xcite applies and @xmath293 is a fiber bundle with structure group @xmath282 as desired
. for @xmath294 , embed @xmath295 into @xmath38 , the structure group for @xmath37 so that it acts on the last @xmath296 vectors in a given frame .
the map @xmath297 is the subbundle of @xmath25-frames . to produce @xmath287
we would like to apply the same process to @xmath290 but must proceed carefully . let @xmath298 where @xmath295 acts on @xmath299 via the same embedding .
this is the stabilizer of the first @xmath25 vectors in @xmath299 ( that is , @xmath41 ) in the subgroup of the structure group that preserves the @xmath18-orbit of @xmath299 .
we now examine this stabilizer for any other frame @xmath300 in @xmath290 .
any such @xmath300 takes the form @xmath301 for some @xmath302 and @xmath84 as in eqn .
[ eqn : g ] .
we then compute @xmath303 but @xmath304 if and only if @xmath305 , and @xmath306 is an element of @xmath18 so we see that @xmath307 for all @xmath308 .
we can refer to this group simply as @xmath309 , and we note that @xmath310 by @xmath311 .
thus , we can obtain @xmath312 as @xmath313 .
the fibers of this map are of the form @xmath314 everywhere so again we can apply @xcite to see that we have a fibration , as desired . we have proved that @xmath287 is a topological sub - fiber bundle - nothing has been claimed about smoothness .
@xmath315-smoothness of @xmath287 in the negative curvature case is proven by brin and is key to his proof that @xmath287 is the ergodic component containing @xmath41 ; here , however , we need only the topological result to appropriate the needed results from brin - gromov .
the transitivity group @xmath282 acts transitively on the fiber of 2-frames over any @xmath192 .
first , note that when @xmath66 is even we are restricted to strict negative curvature and this result is theorem [ thm : bk ] due to brin and karcher .
when @xmath66 is odd the result is theorem [ thm : bg ] due to brin and gromov and found in section 4 of @xcite .
they discuss the proof only in the strict negative curvature case , but it works perfectly well in nonpositive curvature . we have included it here for completeness . our work in proposition [ prop : subbundle ] produced a subfibration @xmath312 with fiber @xmath314 of the fibration @xmath316 with fiber @xmath317 .
let us now restrict our attention to 2-frames , and specifically to @xmath318 , the restriction of the 2-frame bundle to those frames based at a point @xmath136 of @xmath0 .
we get bundles @xmath319 & q(\alpha)|_p / k \ar[dd]^{\pi } & & s^{n-2 } \ar[r ] & st_2m|_p \ar[dd]^{\pi } \\ & & \ar@{^{(}->}[r]^{i } & & \\ & s^{n-1 } & & & s^{n-1 } } \ ] ] where @xmath320 and @xmath321 is the unit tangent sphere above @xmath136 .
take @xmath322 and @xmath323 .
these fibrations , together with the inclusion map @xmath324 give the following commutative diagram , which connects the homotopy long exact sequences for the fibrations by the induced inclusion map @xmath325 ( see @xcite theorem 4.41 ) : @xmath326^{i _ * } \ar[r]^{\pi _ * } & \pi_{n-1}(s^{n-1 } , b_0 ) \ar[d]^{\cong } \ar[r]^{\bar\partial } & \pi_{n-2}(h / k , x_0)\ar@{^{(}->}[d]^{i _ * } \\
\pi_{n-1}(st_2m|_p , x_0 ) \ar[r]^{\pi _ * } & \pi_{n-1}(s^{n-1 } , b_0 ) \ar[r]^{\partial } & \pi_{n-2}(s^{n-2 } , x_0 ) } \ ] ] note that @xmath327 .
now suppose @xmath282 does not act transitively on the fiber of two frames over some @xmath229 .
then @xmath328 so @xmath329 on @xmath330 and thus @xmath331 on @xmath332 .
this implies that the map @xmath5 admits a section , thus giving a nonvanishing vector field on @xmath321 .
if @xmath66 is odd this is a contradiction .
as noted in the introduction , the transitivity group is crucial to this paper s proof . in this section
we show that certain distinguished vector fields @xmath333 along @xmath334 are preserved under the action of the transitivity group and use this result to prove theorem 1 . throughout
we will utilize our dynamical lemma , lemma [ recurrencelemma ] with @xmath335 .
consider , for example , the ideal rectangle defined by @xmath14 , @xmath271 , @xmath273 and @xmath336 as pictured in figure [ fig : rectangle ] . if @xmath337 and @xmath338 are positively recurrent and if @xmath146 and @xmath339 are negatively recurrent , lemma [ recurrencelemma ] implies that the element of @xmath282 corresponding to this ideal polygon preserves the distinguished fields .
the following arguments show how this idea can be worked out for _ all _ ideal polygons , first in the negative curvature case and then in the general case . in the negative curvature case
the argument of this section is considerably simpler , so we discuss it first .
consider the situation depicted in figure [ fig : rectangle ] .
lemma [ recurrencelemma ] shows that , when @xmath337 is recurrent in forward time , the map @xmath340 preserves the distinguished vector fields in the sense that it sends a vector from one such field , @xmath341 , to a vector from another such field along @xmath337 .
thus , if in figure 1 we have that @xmath337 and @xmath338 are recurrent in positive time and @xmath146 and @xmath339 are recurrent in negative time , then the element of @xmath18 given by parallel translation around this ideal rectangle will map @xmath341 to another element of @xmath70 which is in a parallel field along @xmath146 making curvature @xmath1 . if these sort of recurrence properties held for all ` equilateral ' ideal polygons based at @xmath14 we would have that the transitivity group preserves the distinguished vector fields .
we can not assure that these recurrence properties are always present , but ergodicity of the geodesic flow on @xmath0 indicates that they will be present almost all the time .
furthermore , the fact that we have defined elements of the transitivity group using the continuous foliations provided by brin can be used to argue that elements of the transitivity group depend continuously on the choice of ideal polygon .
thus the transitivity group will preserve the distinguished vector fields . for the general case
we need arguments to deal with the problem of pairs of geodesics that are asymptotic but not strictly asymptotic , and the fact that we know longer know we have a continuous foliation . by assumption
, the frame flow preserves the distinguished vector fields .
the only question in terms of how they behave under the action of elements from the transitivity group is how they behave when they are transferred across corners of the ideal polygons . as lemma [ recurrencelemma ] shows , when the geodesics involved are strictly asymptotic and the second geodesic is recurrent , the fields are transferred as desired .
thus , there are two problems to deal with : when the second geodesic is not recurrent , and when the geodesics are not strictly asymptotic .
the new definition of the transitivity group provides a way to deal with both of these issues .
first , note that under the case ii of definition [ defn : rk1transgp ] , we have defined the maps @xmath342 as limits of the maps from the first case of the definition .
thus , to show that a distinguished field @xmath343 is preserved by some @xmath342 we need to realize @xmath343 as a limit of distinguished fields along the geodesics @xmath204 used to define the map @xmath342 ( see figure [ fig : defn ] ) . in view of this fact
we will work with distinguished fields that can be realized as limits , and ensure that this property of arising as a limit is also preserved by the @xmath344 maps . in particular , we will consider distinguished fields @xmath345 that arise as limits of distinguished fields @xmath346 along geodesics @xmath204 as depicted in figure [ fig : corner ] and show that such a field is transferred by a map @xmath344 to a field @xmath347 arising as a limit of fields @xmath348 along geodesics @xmath349 which connect @xmath350 to @xmath351 .
if the next corner to be traversed falls under case ii , the sequence @xmath228 is determined by the map @xmath352 ; otherwise we are free to take any sequence .
the arguments are slightly different in the two cases so we address them separately : + * case i. * suppose @xmath192 falls under case i , and that @xmath345 is the limit of @xmath346 .
then for any @xmath48 ( respectively @xmath49 ) , @xmath353 is a distinguished field along @xmath354 arising as a limit of distinguished fields @xmath348 .
assume @xmath48 ; the proof for the unstable case is essentially the same .
since we are under case i , @xmath354 is limit of recurrent geodesics @xmath355 for @xmath356 .
the maps @xmath357 preserve distinguished fields as demonstrated in lemma [ recurrencelemma ] and @xmath358 , so @xmath347 will be a distinguished field as well . now we need to demonstrate @xmath359 as a limit in the proper way .
construct geodesics @xmath349 connecting @xmath360 and @xmath226 and @xmath361 connecting @xmath257 and @xmath226 as in figure [ fig : corner ] . the field @xmath362 will be a distinguished field by the argument of the previous paragraph . likewise , the fields @xmath363 will be distinguished fields as @xmath364 falls under case i. since the @xmath365 it is clear that the @xmath366 and we are done
. * case ii .
* [ prop : case2 ] suppose @xmath192 falls under case ii , and that @xmath345 is the limit of @xmath346 .
then for any @xmath48 ( respectively @xmath49 ) , @xmath367 is a distinguished field along @xmath354 arising as a limit of distinguished fields @xmath348 .
again , assume @xmath48 .
since the maps @xmath209 are under case i , they preserve distinguished fields .
thus , @xmath347 , which is defined as the limit of @xmath368 , will be a distinguished field . immediately , we have that @xmath347 arises as a limit of distinguished fields along the geodesics @xmath205 . as in figure [ fig : corner ] , let @xmath225 be the geodesic connecting @xmath206 and @xmath226 and let @xmath349 be the geodesic connecting the footpoint of @xmath52 and @xmath226 .
if @xmath369 is not in @xmath203 the map @xmath352 supplies the sequence @xmath228 . in this case
we have required that @xmath221 satisfies the composition criterion for @xmath52 , so @xmath205 , @xmath349 and @xmath225 all approach each other ( and @xmath354 ) as @xmath370 .
if @xmath369 is in @xmath203 it is easy to see that these geodesics still all converge as otherwise we would find a flat strip along @xmath354 . using @xmath136 maps under case i
, the fields @xmath371 can be transfered to distinguished fields @xmath372 along @xmath225 and subsequently to distinguished fields @xmath348 along @xmath349 .
it is then clear that @xmath373 and @xmath348 all limit on @xmath347 ; specifically , @xmath374 shows that @xmath347 arises as a limit in the desired manner .
this work proves [ prop : preserves ] the transitivity group preserves distinguished vector fields that arise as limits of distinguished fields in the correct manner .
we can now apply the results of brin - karcher and of brin - gromov as adapted to the rank 1 situation and prove theorem 1 easily .
let @xmath0 be a compact , rank 1 , nonpositively curved manifold .
suppose that along every geodesic in @xmath0 there exists a parallel vector field making sectional curvature @xmath1 with the geodesic direction . if @xmath0 is odd dimensional , or if @xmath0 is even dimensional and satisfies the sectional curvature pinching condition @xmath2 with @xmath3 then @xmath0 has constant negative curvature equal to @xmath1 .
we showed in proposition [ prop : preserves ] that the sectional curvature @xmath1 fields that arise in the desired way as limits are preserved by the transitivity group . in the setting of the theorem ,
the adapted results of brin - gromov and brin - karcher tell us that the transitivity group acts transitively on @xmath375 .
in particular , by considering the orbit of a distinguished field that arises correctly as a limit we see that @xmath376 is identically @xmath1 , and the theorem is proved .
in @xcite a distinction is made between ` weak ' and ` strong ' rank .
the existence of _ parallel _ fields making extremal curvature is called strong rank ; the existence only of _ jacobi _ fields making extremal curvature is called weak rank .
a parallel field scaled by a solution to the real variable version of the jacobi equation ( where the standard derivative replaces the covariant derivative ) produces a jacobi field .
thus , a proof under the less stringent hypothesis of weak rank implies a proof for strong rank .
hamenstdt s is the sole result prior to this paper for weak rank .
she states her main theorem for parallel fields only , but she shows in lemma 2.1 that in negative curvature a jacobi field making maximal curvature is a parallel field scaled by a function @xcite .
essentially , she shows that jacobi fields making maximal curvature grow at precisely the rate one finds for the constant curvature case .
connell accomplishes the same in @xcite lemma 2.3 .
this , together with some of the arguments below , shows that these jacobi fields are in fact parallel fields scaled by an appropriate function .
therefore , corollary 2 is a weak rank result , needing only the jacobi field hypothesis . in this section
we show that jacobi fields making _ minimal _ curvature with the geodesic direction are also scaled parallel fields .
this will justify the phrasing of corollary 1 as a weak rank result . in this section
, @xmath377 will denote the riemannian inner product and @xmath378 the curvature tensor .
first , note that we need only consider non - vanishing jacobi fields ; hence it will be enough to prove that stable and unstable jacobi fields are scaled parallel fields .
stable jacobi fields are those which have norm approaching zero as @xmath45 ; unstable jacobi fields have the same property in the negative time direction .
suppose @xmath379 is a stable jacobi field along the geodesic @xmath86 making curvature @xmath1 with the geodesic ( take @xmath12 now ) , where @xmath1 is the curvature minimum for the manifold ( the modifications of what follows for unstable jacobi fields are straightforward ) .
the rauch comparison theorem ( see @xcite chapt 10 , theorem 2.3 ) can be used to show that @xmath380 ) . write @xmath381 where @xmath382 and @xmath383 is a unit vector field .
then the jacobi equation for @xmath161 reads : @xmath384 where @xmath385 denotes the standard derivative and @xmath386 denotes covariant derivative . taking the inner product of ( [ eq : jacobi ] ) with @xmath190 and noting that @xmath387 we obtain @xmath388 we now know the following about the magnitude of @xmath161 : @xmath389 by definition , @xmath390 since @xmath161 is a stable jacobi field , and @xmath391 by ( [ eq : simpjacobi ] ) .
these allow the following conclusion ; its proof was shown to the author by jeffrey rauch : [ rauchlemma ] let @xmath392 be a non - negative , real valued function satifsying @xmath391 and @xmath390
. then @xmath393 .
we have that @xmath394 on the interval @xmath395 for @xmath396 define @xmath397 by @xmath398 , @xmath399 and @xmath400 .
note that as @xmath401 , @xmath402 .
we claim that @xmath403 ; the lemma follows in the limit .
this claim is essentially the maximum principle .
first , @xmath403 holds at @xmath404 and @xmath242 .
now suppose @xmath405 has a positive maximum at @xmath406 .
then @xmath407 . however , we know @xmath408 , so a positive value of @xmath405 at @xmath159 together with a negative value of @xmath409 yields a contradiction .
therefore @xmath403 holds on all of @xmath410 $ ] as desired .
this lemma , together with equation ( [ eq : geq ] ) , tells us that @xmath411 . examining equation ( [ eq : simpjacobi ] ) we see that having the growth rate @xmath412 , as in the constant curvature @xmath1 case , implies that @xmath413 ,
that is , @xmath161 is a scaled parallel field , as desired .
in this section we discuss how the results of connell in @xcite can be adapted to prove theorem 2 as a simple consequence of corollary 1 .
the necessary changes are for the most part cosmetic ; the discussion here is included for completeness , but the author does not claim to have added anything of substance to connell s work .
the notation below that has not already been assigned follows connell s for ease of reference .
recall that lyapunov exponents are a tool for measuring long - term asympotic growth rates in dynamical systems ( see katok and mendoza s supplement to @xcite section s.2 for an exposition ) . in the setting of the geodesic flow
they can be defined as follows .
let @xmath229 and @xmath414 .
let @xmath415 be the unstable jacobi field along @xmath146 with initial condition @xmath416 .
then , the _ positive lyapunov exponent at @xmath14 in the @xmath417-direction _ is @xmath418 define @xmath419 this is the maximal lyapunov exponent at @xmath14 ; the curvature bound @xmath420 ( again , take @xmath421 ) implies that @xmath422 .
let @xmath423 we can now rephrase theorem 2 more succinctly . let @xmath0 be a compact manifold with sectional curvature @xmath11 .
suppose that @xmath203 has full measure with respect to a geodesic flow - invariant measure @xmath30 with full support . if @xmath0 is odd dimensional , or if @xmath0 is even dimensional and satisfies the sectional curvature pinching condition @xmath16 with @xmath17 then @xmath0 is of constant curvature @xmath1 .
connell shows in the upper rank case that the dynamical assumption implies the geometric one , that is , that the manifold in fact has higher rank , allowing the application of an appropriate rank rigidity theorem .
he first shows ( @xcite proposition 2.4 ) that along a closed geodesic @xmath424 implies the existence of an unstable jacobi field making curvature @xmath1 with the geodesic direction .
essentially , if the jacobi field giving rise to the lyapunov exponent does not have this curvature , it will continually see non - extremal curvature a positive fraction of the time as it moves around the closed geodesic .
this contradicts the supposed value of the lyapunov exponent .
the lower curvature bound version of the argument is exactly the same as that presented by connell , with the proper inequalities reversed ; also note that the work in section [ sec : parallel ] of this paper gives the results analogous to connell s lemma 2.3 necessary for the argument .
it is clear that if a dense set of geodesics have the distinguished jacobi fields , then all geodesics will .
since the velocity vectors for closed geodesics are dense in @xmath13 , connell finishes his proof in section 3 of @xcite by showing that these vectors are all in @xmath203 and using the argument of the previous paragraph . adapted to the setting of theorem 2 the argument runs as follows .
if @xmath425 is tangent to a closed geodesic and @xmath426 the previous paragraph implies that any unstable jacobi field along @xmath281 must make curvature strictly greater than @xmath1 a positive fraction of the time . by continuity , this will also be true of any unstable jacobi field along a geodesic @xmath146 in a sufficiently small neighborhood of @xmath281 ( in the sasaki metric on @xmath13 ) .
the ergodic theorem implies that for a full measure set of @xmath229 , @xmath146 will spend a positive fraction of its life in this small neighborhood of the periodic geodesic @xmath281 ; the positivity follows from the fact that @xmath30 has full support .
the intersection of this full measure set with the full measure set @xmath203 thus contains vectors @xmath14 which have @xmath424 but spend a positive fraction of their life so close to @xmath281 that no jacobi fields along them can make the minimal curvature @xmath1 with the geodesic direction during this fraction of the time .
in fact , since @xmath281 is compact , so is the closure of this small neighborhood and therefore the curvature between these jacobi fields and the geodesics , when in this neighborhood , can be bounded away from @xmath1 , i.e. @xmath427 for a fixed @xmath159 . having this curvature bound a positive fraction of the time contradicts @xmath424 ; therefore all closed geodesics must lie in @xmath203 and the argument is complete .
again , this version of the argument , relevant for the lower curvature bound situation , is the same as that presented by connell with the proper inequalities reversed .
thus , the dynamical assumption implies the geometric assumption of corollary 1 and theorem 2 follows .
note that for these arguments the extremality of the distinguished curvature is essential and we do not obtain a result that parallels theorem 1 in allowing non - extremal distinguished curvature .
we conclude with a few remarks on possible extension of this work .
note that in even dimension a result directly parallel to our odd dimensional result can not be hoped for .
since parallel translation preserves the complex structure on a khler manifold the 2-frame flow will not be ergodic ( see @xcite for some results on unitary frame bundles ) .
these known counterexamples to ergodic frame flow are excluded by requiring @xmath428 , leading brin to conjecture that strict 1/4-pinching implies that the frame flow is ergodic ( @xcite conjecture 2.6 ) . a positive answer to this conjecture , or any
extended results for ergodicity of the 2-frame flow in negative curvature would extend the results on rank rigidity presented here correspondingly , using the same proof as presented above .
one still hopes that lower hyperbolic rank rigidity ( in the sense that higher rank implies the space is locally symmetric ) could be true without any curvature pinching in even dimensions , and the result here as well as the extensive analogous results for the other rank rigidity theorems seem to make such a theorem more likely .
however , such a result would call for a significantly different method of proof from that presented here .
|
this paper presents hyperbolic rank rigidity results for rank 1 , nonpositively curved spaces .
let @xmath0 be a compact , rank 1 manifold with nonpositive sectional curvature and suppose that along every geodesic in @xmath0 there is a parallel vector field making curvature @xmath1 with the geodesic direction .
we prove that @xmath0 has constant curvature equal to @xmath1 if @xmath0 is odd dimensional , or if @xmath0 is even dimensional and has sectional curvature pinched as follows : @xmath2 where @xmath3 .
when @xmath1 is the upper curvature bound this gives a shorter proof of the hyperbolic rank rigidity theorem of hamenstdt , subject to the pinching condition in even dimension ; in all other cases it is a new result .
we also present a rigidity result using only an assumption on maximal lyapunov exponents in direct analogy with work done by connell .
the proof of the main theorem is simplified considerably by assuming strict negative curvature ; in fact , in all dimensions but 7 and 8 it is then an immediate consequence of ergodicity of the @xmath4-frame flow . in these exceptional dimensions ,
recourse to the dynamics of the 2-frame flow must be made and the scheme of proof developed there can be generalized to deal with rank 1 , nonpositively curved spaces .
[ section ] [ thm]definition [ thm]lemma [ thm]proposition [ thm]remark
|
the classical risk measure for insurance companies is the probability of ruin .
this quantity is intensively studied , since in 1903 @xcite introduced his model for the surplus of an insurance company . however , in many models a trajectory can only either lead to ruin or tend to infinity .
hence , @xcite introduced expected discounted future dividend payments as a valuation principle for a homogeneous insurance portfolio , which builds an alternative risk measure .
this concept originally comes from corporate finance , where the firm value is often determined by the accumulated future dividend payments .
our aim is to solve the valuation problem of an insurance company in these terms . since its introduction
, the dividend maximization problem has been solved in various setups , for example by @xcite , @xcite , @xcite , and @xcite in diffusion models .
@xcite study a similar problem with random intervention times .
overviews can be found in @xcite , or @xcite .
for an introduction to optimization problems in insurance in general we refer to @xcite and @xcite .
+ however , all of these contributions assume a constant economic environment . the dividend maximization problem is considered over a potentially long time horizon , which makes the assumption that the economy does not change a strong one . since a changing economic environment structurally influences the insurance market , we would like to incorporate such changes into our model . in setups which allow for a change in the economic environment
the dividend maximization problem has been studied for example by @xcite , @xcite , @xcite , @xcite , and @xcite .
@xcite , @xcite , and @xcite consider a diffusion model for the surplus , the parameters of which are driven by an underlying markov chain .
such a setup is called _ regime switching model_. in a cramr - lundberg type model @xcite allow the economy to become worse once before ruin , and @xcite allow for a finite number of shifts of the economy . in all these models ,
the driving markov chain is assumed to be _ observable_. this means that they assume full information and therefore their models differ from the model studied herein .
+ we use the dividend maximization approach to solve the valuation problem of an insurance company in a _ hidden _ markov model .
more precisely , we model the surplus process of the insurance company as a brownian motion with drift .
the drift is assumed to be driven by an underlying markov chain , the current state of which is _ unobservable _ under the available information .
in contrast to @xcite and @xcite , where the setup was bayesian , i.e. , the underlying markov chain was not allowed to change its state , we allow for regime shifts .
this gives the model a different interpretation .
here , the different states of the markov chain represent different phases of the economy .
it is of certain interest on the one hand to allow for changes in the economic environment and on the other hand it is realistic that the current state of the economy is not exactly known instantaneously , but only over time , and also over time the drift of a diffusion can not be estimated satisfactorily , see ( * ? ? ?
* chapter 4.2 ) even more if shifts are allowed .
thus allowing for uncertainty extends @xcite to a more practically relevant direction .
+ in the literature of mathematical finance hidden markov models have already been used intensively for studying investment problems , e.g. , in @xcite , @xcite , @xcite , or @xcite .
also dividend problems were solved in the mathematical finance literature , see @xcite and @xcite , who seek to maximize the expected accumulated utility of dividends , and the expected utility of accumulated dividends , respectively . however , in the insurance context related results concerning hidden markov models are scarce .
we refer to @xcite as an example , who models the value of a single insurance policy as a brownian motion with unobservable drift describing uncertainty about the quality of a risk .
another , more recent example is the paper by @xcite , who study optimal reinsurance and investment under unobservable claim size and intensity . in @xcite
a valuation problem similar to the dividend maximization problem is studied in a rather specific model from the point of view of corporate finance . + this paper
is organized as follows . in section [ sec : setup ]
we define our model and show how to overcome uncertainty by applying a result from stochastic filtering theory , and thus transform the setup into one under full information .
the stochastic optimization problem under study is presented in section [ sec : opti ] .
for solving the stochastic optimization problem we derive the hamilton - jacobi - bellman ( hjb ) equation and characterize the optimal value function as a viscosity solution to this hjb equation .
we also prove uniqueness of the viscosity solution even though there is a lack of boundary conditions . in section [ sec : num ]
we treat the problem numerically .
first of all we examine the filter dynamics .
then , we solve the hjb equation numerically .
for this we need to introduce a correction term to ensure positivity of the scheme , but we can show that the approximate solution converges .
we present a multitude of numerical examples .
furthermore , we are able prove admissibility of the candidate optimal strategies , which means showing that the underlying system of stochastic differential equations with discontinuous drift and degenerate diffusion has a solution .
section [ sec : concl ] concludes the paper . +
the main contribution of this paper is the analytical characterization of the solution to the dividend maximization problem in a hidden markov switching model , including a non - standard uniqueness proof of the generalized solution to the associated hamilton - jacobi - bellman ( hjb ) equation , and an extensive numerical study of the outcomes of this model .
the intention behind this study is to impart good comprehension of the model and of how to optimally pay out dividends . assuming only partial information makes the model more natural and realistic .
all stochastic variables introduced in the following are assumed to be defined on the filtered probability space @xmath0 . + the surplus process is given by @xmath1 with initial capital @xmath2 , where @xmath3 is the unobservable drift process
, @xmath4 is the constant and known volatility , and @xmath5 is a standard brownian motion . the accumulated dividend process @xmath6 is given by @xmath7 with @xmath8 and density @xmath9 $ ] for all @xmath10 that serves as the control variable in our optimization problem .
note that the surplus process @xmath11 is always associated to a dividend strategy , however , for notational reasons , we will not make that explicit .
+ furthermore , let @xmath12 , where @xmath13 is an @xmath14-state markov chain with known generator matrix @xmath15 .
let @xmath16 , where @xmath17 , if @xmath18 , and @xmath19 is the @xmath14-dimensional unit vector the @xmath20-th component of which is @xmath21 .
without loss of generality let @xmath22 .
we assume that the current state of the markov chain is unobservable under the observation filtration , but we know its initial distribution @xmath23 with @xmath24 for all @xmath25 and @xmath26 .
+ note that it is crucial that the volatility is assumed not to be driven by the markov chain , as then it would be possible to estimate the current state from the quadratic variation .
+ the uncontrolled surplus process @xmath27 is given by @xmath28 as the dividend payments have to be adapted to the uncontrolled process , the observation filtration is given by @xmath29 , which is the augmentation of the filtration generated by @xmath30 .
+ as @xmath31 is not @xmath32-measurable , we are in a situation of partial information . to overcome this uncertainty ,
we apply a result from stochastic filtering theory .
this means we replace the unobservable parameter @xmath31 by an estimator , which potentially uses all the information generated by @xmath32 , but not more .
we refer the interested reader to @xcite for more information about hidden markov models and their filtering , and to @xcite for stochastic filtering in general . @xcite
suggest using the wonham filter ( see @xcite ) for the case where the unobservable variable is driven by a markov chain .
+ from ( * ? ? ?
* theorem 9.1 ) we know the following proposition .
denote the conditional probability that the markov chain is in state @xmath20 at time @xmath33 ( and hence @xmath17 ) as @xmath34 for @xmath35 , and the estimator for the drift as @xmath36 then @xmath37 solves the following system of stochastic differential equations @xmath38 for @xmath39 , with the innovation process @xmath40 furthermore , @xmath41 is an @xmath42 - brownian motion . in particular , we have that the estimator @xmath43 is adapted to the observation filtration .
+ please note that @xmath44 for all @xmath45 , which implies that the correct state space of the filter is the simplex @xmath46^m : \sum_{i=1}^m \pi_i = 1\}$ ] , the interior of which is denoted by @xmath47 . for later use ,
we define @xmath48 and @xmath49 . from now on we consider the following @xmath14-dimensional system of sdes : @xmath50 where @xmath51 since there is only one source of uncertainty , which is adapted to the observation filtration , in system , , we are now in a situation of full information , but at the cost of @xmath52 additional dimensions .
in this section we at first define the stochastic optimization problem under study .
then we derive the associated hjb equation . finally , we present the main result of this section , which is the characterization of the solution of the optimization problem as the unique viscosity solution to the hjb equation .
+ we would like to find the optimal value function @xmath53 , which is the supremum over all dividend policies of the discounted dividend payments up to the time of ruin @xmath54 , @xmath55 where @xmath56 is the discount rate , @xmath57 denotes the set of admissible controls , and @xmath58 denotes the expectation under the conditions @xmath59 and @xmath60 .
admissible controls are @xmath61-progressively measurable , @xmath62$]-valued , and fulfill @xmath63 for @xmath64 .
+ note that the underlying system of stochastic processes , describes autonomous state dynamics in the sense of ( * ? ? ?
* section iv.5 ) .
furthermore , we will consider an infinite time horizon .
therefore , the optimal control will be markovian . [
lem : x - infinity ] the optimal value function @xmath53 is continuous .
we have @xmath65 , @xmath53 increases in @xmath66 , and @xmath67 uniformly in @xmath68 . from ( * ? ? ?
* chapter 3 , theorem 5 ) we know that the optimal value function @xmath53 is continuous .
+ the monotonicity of @xmath53 with respect to @xmath66 follows by an argument from ( * ? ? ?
* chapter 2.5.1 , p. 97 ) .
+ clearly , the optimal value function is bounded by @xmath69 , and it is easy to check that in the limit it converges to @xmath70 , cf .
* chapter 2.5.1 , p. 97 ) . for deriving the hjb equation
we need a version of the dynamic programming principle , or bellman principle , see ( * ? ? ?
* chapter 3 , theorem 6 ) .
[ prop : bellman ] for every bounded stopping time @xmath71 we have @xmath72 now , assuming @xmath73 , one can derive the associated hjb equation from the bellman principle : @xmath74}(u(1-v_x))=0\,,\end{aligned}\ ] ] where @xmath75 and @xmath76 is given by .
the hjb - equation is a second order degenerate - elliptic pde , since there is only one brownian motion driving the @xmath14-dimensional process @xmath77 .
the supremum in is attained at @xmath78 as boundary conditions we have for @xmath79 and @xmath80 @xmath81 for @xmath82 , @xmath83 , we have no boundary conditions available . however , as the filter @xmath68 never reaches the boundary , boundary conditions in these particular directions are not required for the solution being well - defined .
the reason for this is that we still get uniqueness in the interior and on the relevant part of the boundary , see corollary [ cor : comparison ] .
+ it should be mentioned at this place that the solution to the problem with complete information , i.e. , for observable @xmath84 , does not serve as boundary condition .
this is because even if we knew that we started in a certain state , then a moment later we would again not be able to observe the state , whereas in the model suggested by @xcite , one is then still able to observe it .
now we come to the analytic characterization of the optimal value function . in the markov switching setup where the current state of the markov chain is observable ( see @xcite ) , the hjb equation can be solved explicitly and the solution is smooth . in our case
the hjb equation is a degenerate - elliptic pde , which makes the existence of a smooth solution questionable .
thus , we deal with a weaker concept of solutions , namely viscosity solutions . the only required smoothness for this is continuity , however , the concept is still strong enough to prove uniqueness .
furthermore , it is also useful for numerical treatment , see @xcite , or ( * ? ? ? * chapter ix ) .
these are two important strengths of the concept of viscosity solutions and make it very beneficial to problems like ours .
therefore , we are going to characterize the optimal value function @xmath53 as the unique viscosity solution of .
+ denote by @xmath85 the state space of the first @xmath52 dimensions of the filter with closure @xmath86^{m-1 } : \sum_{i=1}^{m-1 } { p}_i \le 1\}$ ] and boundary @xmath87 .
further denote @xmath88 , @xmath89 and let @xmath90 be its boundary .
furthermore , let @xmath91 .
then @xmath92 denotes the so - called _ relevant part _ of the boundary .
[ def : viscosity ] ( viscosity solution ) 1 .
a function @xmath93 is a _ viscosity subsolution _ to , if @xmath94}(u(1-\phi_x(\bar{x},\bar{{p}})))\ge 0\ ] ] for all @xmath95 and for all @xmath96 such that @xmath97 attains a maximum at @xmath98 with @xmath99 .
2 . a function @xmath93 is a _ viscosity supersolution _ to , if @xmath100}(u(1-\psi_x(\bar{x},\bar{{p}})))\le 0\ ] ] for all @xmath95 and for all @xmath101 such that @xmath102 attains a minimum at @xmath98 with @xmath103 .
3 . @xmath93 is a _ viscosity solution _ to , if it is both a viscosity sub- and supersolution .
the basic idea of viscosity solutions is to estimate the function from below and from above by smooth test functions . for details about viscosity solutions ,
see , e.g. , @xcite , or @xcite .
+ the following theorem shows the connection between the solution of the optimization problem and the weak solution of the hjb equation .
[ thm : viscosity ] the optimal value function @xmath53 is a viscosity solution of with boundary conditions and .
we have to show that the optimal value function @xmath53 is a viscosity sub- and supersolution , cf .
* proof of theorem 5.1 ) .
+ let @xmath101 , @xmath104 and @xmath98 such that @xmath105 .
let @xmath106 .
+ applying the dynamic programming principle we get @xmath107 for any fixed @xmath108 $ ] .
+ now we apply it s formula to @xmath109 , note that the stochastic integrals are martingales , divide by @xmath71 and let @xmath110 .
this yields @xmath111 as @xmath112 was arbitrary , @xmath113 } \left(u ( 1-\psi_x ( \bar{x},\bar{{p}}))\right)\,.\ ] ] thus , @xmath53 is a viscosity supersolution .
+ let @xmath114 , @xmath115 and @xmath116 such that @xmath117 . for @xmath118 let @xmath119 and @xmath120 be an @xmath121-optimal dividend policy for the first part of the dynamic programming principle , and denote the surplus coming from @xmath120 as @xmath122 .
then @xmath123 where @xmath124 is continuous in @xmath33 , has values in @xmath62 $ ] , and approximates @xmath125 in @xmath126 .
furthermore , we applied it s formula and used that the stochastic integrals are martingales . rearranging the inequality and dividing by @xmath71 yields @xmath127 now we apply the mean value theorem : @xmath128 and let @xmath110 along a sequence : @xmath129 fatou s lemma gives @xmath130 where @xmath131 . as @xmath132 was arbitrary , @xmath133 since @xmath134 } u ( 1- \phi_x(\bar{x } , \bar{{p}}))$ ] , we get @xmath135 } u ( 1- \phi_x(\bar{x } , \bar{{p } } ) ) \geq 0\,.\ ] ] thus , v is also a viscosity subsolution .
+ altogether , @xmath53 is a viscosity solution .
now it remains to prove uniqueness . for this , one has to prove a weak maximum principle , which in standard proofs results in the statement that if two viscosity solutions are equal on the boundary , they are also equal in the interior of the domain . however , as mentioned above , we have no boundary conditions available in the @xmath68 directions .
but @xcite shows that if the underlying stochastic process does not reach some parts of the boundary of the domain with a positive probability , then as these parts are not reached anyway , the study can be restricted to the interior and the relevant part of the boundary .
[ thm : comparison ] let @xmath136 and @xmath137 be bounded and continuous viscosity solutions of .
+ if @xmath138 on @xmath139 and @xmath140 uniformly in @xmath68 , then @xmath138 on @xmath141 .
define @xmath142 .
we need to check whether @xmath143 . from (
* corollary 2.2 ) we know that the wonham filter never reaches the boundary . therefore , the above probability is indeed zero . hence , we may apply ( * ? ? ?
* corollary ii.1 ) , which proves the statement .
uniqueness of the solution of now follows as a corollary .
[ cor : comparison ] the optimal value function @xmath53 is the unique bounded viscosity solution of on @xmath144 with boundary conditions and .
the following theorem shows that our analytic characterization includes smooth solutions to the hjb equation .
furthermore , it can be concluded that if there is a dividend policy leading to a smooth value function that solves the hjb equation in the viscosity sense , then this policy is optimal .
[ thm : verification ] let @xmath145 be a viscosity supersolution of with boundary conditions and , and @xmath146 almost everywhere .
then @xmath147 .
the proof runs along the same lines as ( * ? ? ?
* proof of theorem 5.3 ) .
we begin with convoluting @xmath145 with a gauss weierstrass kernel function . due to notational ambiguities , we remark that here the area of a circle with radius @xmath21 is denoted by @xmath148 .
let @xmath149 and @xmath150 for @xmath151 , where @xmath152 .
clearly , as @xmath153 , @xmath154 and @xmath155 , see @xcite .
+ for an admissible strategy @xmath156 and @xmath157 , @xmath158\,dt + m_t\,,\end{aligned}\ ] ] where @xmath159 is a martingale .
therefore , @xmath160\,dt\right)\,.\end{aligned}\ ] ] note that @xmath145 fulfills @xmath161 thus , for @xmath132 we can choose @xmath162 large enough such that @xmath163 and hence @xmath164 therefore , @xmath165\,dt\right)\\ & = \varphi^n(x,{p } ) - { { \mathbb{e}}}_{x,{p } } \left ( \int_0^{t \wedge \tau } e^{-\delta t } u_t \,dt - \varepsilon \int_0^{t \wedge \tau } e^{-\delta t } \,dt\right)\,.\end{aligned}\ ] ] by dominated convergence we have for @xmath166 @xmath167 as @xmath168 was arbitrary , @xmath169 and hence @xmath170 since @xmath145 is bounded we have that @xmath171 .
thus , by bounded convergence , @xmath172 since for each @xmath112 , @xmath145 dominates the optimal value function , by taking the supremum over all @xmath108 $ ] in the derivation , we get @xmath173 .
if there is a strategy @xmath174 such that @xmath175 is a viscosity supersolution with @xmath176 almost everywhere , then by theorem [ thm : verification ] , @xmath177 is the classical solution to the problem , and @xmath178 is the optimal policy .
in this section we first simulate a path of the markov chain @xmath84 and a path of the wonham filter to get an idea of its behaviour .
then we describe a numerical procedure for computing approximations to @xmath53 and the optimal dividend policy .
we will restrict our numerical analysis to the case @xmath179 .
for a better understanding of the numerical results we transform the state process and consider @xmath180 , where @xmath181 , and the corresponding transformed hjb equation , instead of considering @xmath182 . + for simulating paths of the wonham filter we need to express @xmath183 in terms of @xmath30 .
the representation follows from : @xmath184 . with this and equations
and we get @xmath185 now we simulate the increments of the brownian motion @xmath186 as @xmath187 , where @xmath188 . furthermore , we simulate a path @xmath189 of the markov chain @xmath84 , and calculate @xmath190 . with this and equation
we are ready to calculate the path of the estimator @xmath191 by applying the euler - maruyama scheme .
figure [ fig : wonhamplot ] shows a path of the drift of the uncontrolled process governed by the underlying markov chain , and its estimator .
we see that the estimator always needs some time to notice the change in the drift and only adapts to it slowly , but this clearly depends on the choice of @xmath192 .
furthermore , we see that the estimator indeed does not reach the boundary .
+ now we are going to solve the hjb equation numerically .
the transformed hjb equation is @xmath193}(u(1-v_x))=0\,,\end{aligned}\ ] ] where @xmath194 to solve this pde numerically , we first introduce an approximation based on an idea from @xcite to the hjb equation to ensure convergence of the scheme .
define @xmath195 and @xmath196 .
@xmath197 is a brownian motion independent of @xmath183 , and @xmath198 is a smooth function that is bounded , has bounded derivatives , and is vanishing at @xmath199 and @xmath200 on @xmath201 for some small @xmath202 .
furthermore , denote by @xmath203 the value function and the optimal value function associated to the approximate underlying system .
this introduces an additional second order term in the approximate hjb equation @xmath204}(u(1-v^\epsilon_x))=0\,.\end{aligned}\ ] ] the difference to the approximation in @xcite is that here we do not regularize our hjb equation with the additional term , but just use it to ensure positivity of the scheme in the interior of the computation domain .
note however that the analytic characterization of the optimal value function herein does not require regularization .
+ the following theorem states that the solution still converges to the optimal value function .
let @xmath205 be the solution to . then @xmath206 .
we show that the result holds for every value function @xmath207 . from this
we may conclude that the result also holds for @xmath53 , see ( * ? ? ?
* corollary 7.4 ) .
+ let @xmath208 denote the corresponding value functions stopped at time @xmath209 .
we have @xmath210 where we skipped the arguments in the calculations and used that all terms in the last but one row are bounded since the optimal value function is bounded due to lemma [ lem : x - infinity ] .
now we show that all terms tend to @xmath211 . @xmath212 and using that the last term is bounded we get @xmath213 analogously we obtain @xmath214 . for the last term
we get @xmath215 where we used boundedness of the term in the last but one row .
now it remains to show that @xmath216 .
proving that @xmath217 and @xmath218 runs along the same lines as in ( * ? ? ?
* proof of lemma 7.2 ) . since @xmath219 u.c.p .
we have that along a subsequence @xmath220 , @xmath221 u.c.a.s . and hence also @xmath222 a.s .
thus @xmath223 , which closes the proof .
now we are ready to solve numerically .
we first of all have to restrict our computation domain and therefore choose a sufficiently large number @xmath224 to approximate the domain of the value function by @xmath225\times [ \mu_2,\mu_1]$ ] . +
to compute the approximate outcomes of our problem , we use policy iteration .
initially , we use a dividend policy of threshold type , since such strategies solve the problem with complete information and thus are good candidates also in our situation . as the initial threshold level we use the convex combination of the threshold levels which are outcomes to the problem with complete information from @xcite , as we expect it to be close to the correct solution to our problem : @xmath226 , where @xmath227 denote the threshold levels in the case with complete information for states @xmath21 and @xmath228 , respectively .
the initial strategy is given by @xmath229 .
note that the mesh we use for our computation is generated such that there are more grid points available where they are needed the most between @xmath230 and @xmath231 . for more details about the mesh generation
we refer to @xcite .
now we iteratively apply the following procedure : * for a given strategy @xmath232 calculate @xmath233 by solving @xmath234 where @xmath235 is the operator @xmath236 with differentiation operators replaced by finite differences , @xmath237 is a finite difference approximation to differentiation with respect to @xmath66 , and @xmath238 is the finite difference operator replacing the second derivative w.r.t .
* the next iterate @xmath240 is chosen to maximize @xmath241 .
thus @xmath242 . in our experiments
the iteration stops after @xmath243 steps , since @xmath244 .
+ the idea behind the construction of the finite difference method is based on the fact that in the discretized setting the diffusion is approximated by markov chains which locally preserve properties of the original process ( cf .
the additional term @xmath198 is required to guarantee positivity of the scheme and hence to obtain its markov chain interpretation .
corresponding convergence results can be found in @xcite and ( * ? ? ?
* chapter ix ) .
in @xcite it is noted that the policy iteration converges , yielding value function and associated policy .
we computed both value function and dividend policy for the parameter sets @xmath245 , @xmath246 , @xmath247 , @xmath248 , @xmath249 , @xmath250 , @xmath251 , and @xmath252 .
the resulting strategies turn out to be threshold strategies with threshold levels depending on the estimate of @xmath253 .
figure [ fig : thresholdplot ] shows the resulting threshold levels and compares them to the threshold levels which are outcomes to the dividend maximization problem with constant and unobservable drift , i.e. , @xmath254 , which is studied in @xcite .
interestingly , while in the case studied herein the threshold level is falling in @xmath239 for @xmath255 , it increases in the bayesian case . but note that for this parameter choice both curves are rather flat .
the intuition behind the fact that the threshold level grows for some parameter sets , whereas it falls for others is the following .
usually , the level falls in @xmath239 since in the better state the company can pay out dividends earlier as it will recover from it due to the higher drift .
however , for low values of @xmath239 the volatility comes more into effect and hence if @xmath256 is small anyway , then it becomes better to pay dividends even for low values of @xmath66 , because volatility might lead into early ruin . then , as @xmath239 grows it becomes more effective than the volatility and hence soon ruin is not expected anymore and the strategy is designed for a longer living company for higher values of @xmath239 .
it is interesting that for all of the four parameter sets in the state with the lower drift dividends are paid out more cautiously in the markov switching case than in the bayesian case , and in the state with the higher drift it is the other way round .
an explanation for this is that since in the markov switching case there is a chance that the economy gets better , it is the better strategy to wait , if @xmath239 is small and pay out dividends at higher values of @xmath239 .
hence the state with the lower drift is the state of saving , whereas the other state is the state of spending . in the bayesian case
the drift does not change and therefore the situation is more balanced . + figure [ fig : valueplot ] shows the value function corresponding to @xmath257 ( but they all look rather similar ) . .
] figure [ fig : valueplot ] suggests that the optimal value function is smooth , however proving smoothness is due to the degeneracy of the hjb equation , which is highly non - standard out of the scope of this paper .
+ figure [ fig : valuecomp ] compares the resulting value function to the one from the bayesian case .
we observe that for a smaller estimate of the drift the value is higher in the case with switching , and for a high estimate of the drift the value is higher in the bayesian case .
this is due to the fact that if a high drift is expected in the bayesian case , it is more probable that the drift is in fact high , whereas in the case studied in this paper the drift might change to the worse . for a small estimate of the drift
it is exactly the other way round .
+ ( blue ) compared to the resulting value function from the bayesian case ( dashed red).,title="fig : " ] ( blue ) compared to the resulting value function from the bayesian case ( dashed red).,title="fig : " ] one more interesting case to study is that of a high dividend bound @xmath256 .
figure [ fig : kgrows ] suggests that for growing parameter @xmath256 the dividend policy converges to what we expect to be the optimal barrier level in the case of unbounded dividend payments . in future research , it would be of interest to study the singular control problem with unbounded dividend rates . .
] for the markov switching model under full information @xcite find threshold strategies to be optimal . for each state @xmath20 of the underlying markov chain they get a constant threshold level @xmath258 such that no dividends are paid below this level and dividends are paid at the maximum rate @xmath256 , if the surplus process exceeds the level .
remember however that in their setup the current state is observable , and therefore it can always be decided which threshold level has to be chosen .
+ in our case , as the current state is estimated in a continuous way , we have only one threshold level .
we see that the numerical approximation of the optimal dividend policy is of threshold type with a threshold level @xmath259 depending on the estimate of the drift , and hence is of the form @xmath260 therefore , it is of considerable importance to study the admissibility of this type of strategies .
the question is whether the system @xmath261 with @xmath76 as in , has a solution . as the drift coefficient of this sde is discontinuous , classical results from the sde literature as (
* theorem 2.2 ) can not be applied . however , for threshold levels @xmath259 which satisfy the assumptions of ( * ? ? ?
* theorem 3.20 ) , we get existence and uniqueness of a unique global strong solution to system , .
we have presented a diffusion model for the surplus process of an insurance company , where the drift coefficient changes in response to a change of the economic environment .
+ the change of the economic environment has been modeled by a markov chain , and uncertainty has been introduced by not allowing to observe the current state of the markov chain .
we have shown how to overcome uncertainty in this situation by applying a result from stochastic filtering theory .
then we have stated the dividend maximization problem and we have derived the associated hjb equation .
+ we have been able to characterize the solution to the stochastic optimization problem as the unique viscosity solution to this hjb equation .
+ finally , we have presented an extensive numerical study for the solution to the optimization problem , which suggests that the optimal dividend policy is of threshold type .
we have shown that such strategies are indeed admissible using a non - standard result on stochastic differential equations .
+ the main contribution of the current paper is the improvement of the regime switching models that have been studied in the literature by not assuming full information any more .
furthermore , emphasis was put on the numerical study to impart a deeper understanding of the behaviour of the optimal dividend policy for different parameter sets on the one hand , and in comparison to the bayesian case on the other hand .
the author thanks gunther leobacher ( johannes kepler university linz ) , stefan thonhauser ( graz university of technology ) and ralf wunderlich ( btu cottbus - senftenberg ) for fruitful discussions and helpful advice that improved this paper .
+ furthermore , the author thanks two anonymous referees for their suggestions .
+ m. szlgyenyi is supported by the vienna science and technology fund ( wwtf ) : project ma14 - 031 . + the main part of this paper
was written while m. szlgyenyi was member of the department of financial mathematics and applied number theory , johannes kepler university linz , 4040 linz , austria .
+ during this time , m. szlgyenyi was supported by the austrian science fund ( fwf ) : project f5508-n26 , which is part of the special research program `` quasi - monte carlo methods : theory and applications '' .
h. albrecher and s. thonhauser .
_ on optimal dividend strategies in insurance with a random time horizon _ , pages 157180 .
stochastic processes , finance and control .
festschrift for robert j. elliott .
world scientific , singapore , 2012 .
m. szlgyenyi . .
in m. vanmaele , g. deelstra , a. de schepper , j. dhaene , w. schoutens , s. vanduffel , and d. vyncke , editors , _ handelingen contactforum actuarial and financial mathematics conference , interplay between finance and insurance , february 7 - 8 , 2013 _ , pages 7782 .
koninklijke vlaamse academie van belgi voor wetenschappen en kunsten , brussel , 2013 .
isbn 9789065691231 .
|
in this paper we study the valuation problem of an insurance company by maximizing the expected discounted future dividend payments in a model with partial information that allows for a changing economic environment .
the surplus process is modeled as a brownian motion with drift .
this drift depends on an underlying markov chain the current state of which is assumed to be unobservable .
the different states of the markov chain thereby represent different phases of the economy .
+ we apply results from filtering theory to overcome uncertainty and then we give an analytic characterization of the optimal value function .
finally , we present a numerical study covering various scenarios to get a clear picture of how dividends should be paid out .
_ keywords : _ dividend maximization , hidden markov model , filtering theory , stochastic optimal control , viscosity solutions + mathematics subject classification ( 2010 ) : 91b30 , 91b70 , 93e20 + m. szlgyenyi + institute of statistics and mathematics , vienna university of economics and business , 1020 vienna , austria + [email protected]
|
understanding the nature of mesons and baryons has always been one of the most challenging topics in hadron physics .
the new observations @xcite have challenged the conventional wisdom that mesons are made of quark - antiquark pairs and baryons are composed of three quarks . in this respect , it is not surprising that both the @xmath0 state and the @xmath1 have generated a lot of interests in their true nature .
the @xmath0 has been suggested of being either a @xmath17 , a tetraquark state , a meson - meson molecule , a glueball , or a dynamically generated state @xcite .
similarly , the @xmath1 has been found to be consistent with both a naive three quark picture and a molecular picture dynamically generated from @xmath9 and @xmath18 interactions . in the chiral unitary approach ,
the @xmath0 state was shown to be dynamically generated from the interaction of @xmath19 and @xmath20 treated as coupled channels in isospin @xmath21 @xcite .
it decays into @xmath22 in @xmath23-wave with a total decay width around @xmath24 mev .
the pole of the @xmath0 is much tied to the coupled channels and it disappears if either the @xmath22 or the @xmath14 channel is discarded .
the closeness of its pole to the @xmath14 threshold and its strong coupling to @xmath14 have led to the suggestion that the @xmath0 state might be a cusp effect @xcite . as a result
, the @xmath0 total decay width increases very fast as its mass increases .
for instance , ref .
@xcite has claimed a large width for the @xmath0 around @xmath25 mev , where the @xmath0 state was studied within a realistic meson - exchange model for the @xmath26 and @xmath22 interactions . on the other hand , in ref .
@xcite , it was claimed that both the @xmath0 and the @xmath27 resonance emerge from a single @xmath28 seed state .
this state interacts with other mesons , giving rise to meson - meson ( @xmath29 ) loop contributions to the corresponding mass .
these contributions shift the pole of the seed state to higher energies , turning into the @xmath27 . for the @xmath0 ,
the meson cloud eats up the original seed , becoming the largest component @xcite ( see also refs .
by now it is commonly accepted that the @xmath0 is not a standard @xmath30 state but an extraordinary state @xcite . for the @xmath1 hyperon resonance , based on the analysis of the available high precision data on @xmath31 reaction ,
it was argued to be a three - quark state @xcite .
such a conclusion is also supported in the study of the @xmath32 reaction at low energies within the chiral quark model @xcite .
however , in the chiral unitary approach , the @xmath1 resonance can be dynamically generated from the @xmath23-wave meson baryon interaction in the strangeness @xmath33
sector @xcite .
experimentally , the @xmath1 resonance is known to have a strong coupling to the @xmath34 channel @xcite .
hence , it is expected that the @xmath4 production is dominated by formation of the intermediate @xmath1 resonance . in recent years
, it has been shown that the selection of particular modes is possible in the nonleptonic weak decays of heavy hadrons @xcite .
in particular , the nonleptonic weak decays of charmed baryons can be useful tools to study hadronic resonances , some of which are subjects of intense debate about their nature @xcite .
in addition , those weak decays are also helpful to investigate final state interactions and hence have the potential to bring further light into the nature of some puzzling hadrons @xcite .
for instance , the @xmath35 decay was studied in refs .
@xcite , where the @xmath36 state is generated in the final state interaction of the ground state pseudoscalar mesons and octet baryons ( @xmath37 ) . in ref .
@xcite , the work of refs .
@xcite on the @xmath38 weak decays was extended to the @xmath39 weak decays .
it is shown there that these weak decays might be ideal processes to study the @xmath36 and @xmath1 resonances , because they are dominated by the @xmath40 contribution . in ref .
@xcite , the @xmath41 mass distribution was studied in the @xmath42 decays with the aim of extracting the @xmath41 scattering lengths . in a recent work @xcite the role of the exclusive @xmath43 decays into a neutron in testing the flavor symmetry and
final state interaction was investigated .
it was shown that the three body nonleptonic decays are of great interest to explore the final state interaction in @xmath44 decays . along this line , in the present work
, we revisit the @xmath45 decay taking into account not only the @xmath34 final - state - interaction ( fsi ) , but also the fsi of @xmath46 , which gives the line shape of the @xmath0 state .
the pure @xmath15 nature of the @xmath3 channel is particularly beneficial to the study of the @xmath0 state .
this article is organized as follows . in sec .
[ sec : formalism ] , we present the theoretical formalism of the weak @xmath2 decay , explaining in detail the hadronization and final state interactions of the @xmath9 and @xmath47 pairs . numerical results and discussions are presented in sec .
[ sec : results ] , followed by a short summary in the last section .
as shown in ref .
@xcite , a way for the @xmath2 to proceed is the following : in the first step the charmed quark in @xmath43 turns into a strange quark with a @xmath48 pair by the weak decay shown in fig .
[ fig : feynd ] .
the next step consists in introducing a new @xmath49 pair with the quantum numbers of the vacuum , @xmath50 , to form a meson ( baryon ) @xmath51 from the @xmath23 quark ( @xmath52 diquark ) or @xmath29 from the @xmath48 pair with @xmath12 from the @xmath7 cluster .
finally , the final - state interactions ( fsis ) of the @xmath29 or @xmath37 will lead to dynamical generation of the @xmath0 and @xmath1 . in the following ,
we discuss the @xmath4 and @xmath3 fsis separately .
decaying into a @xmath48 pair and a @xmath53 cluster . the solid lines and the wiggly line stand the quarks and the @xmath54 boson , respectively . ]
cluster into one meson and one baryon for the @xmath55 decay with the @xmath11 emission from the @xmath48 pair . ]
we first discuss the decay of @xmath43 to produce the @xmath11 from the @xmath48 pair and the @xmath7 cluster hadronization into a meson - baryon pair . to create the @xmath37 final state
, we must proceed to hadronize the @xmath7 cluster by creating an extra @xmath49 pair as depicted in fig .
[ fig : mbproduction ] . in this process ,
the @xmath52 diquark in @xmath43 is the spectator , and the @xmath7 cluster is combined into a pure @xmath56 state @xmath57 as in refs .
@xcite , one can straightforwardly obtain the meson - baryon states after the @xmath49 pair production as @xmath58 where the flavor states of the baryons are as follows : @xmath59 after the production of a meson - baryon pair , the final - state interaction between the meson and the baryon takes place , which can be parameterized by the re - scattering shown in fig . [
fig : mbfsi ] at the hadronic level for the@xmath60 decay . in fig .
[ fig : mbfsi ] , we also show the tree level diagram for the @xmath60 decay . decay . ] according to eq .
, we can write down the @xmath2 decay amplitude of fig . [ fig : mbfsi ] as , @xmath61 where @xmath62 expresses the weak and hadronization strength , which is assumed to be independent of the final state interaction . in the above equation
, @xmath63 denotes the one - meson - one - baryon loop function , which depends on the invariant mass of the final @xmath34 system , @xmath64 .
the meson - baryon scattering amplitudes @xmath65 are those obtained in the chiral unitary approach , which depend also on @xmath64 .
details can be found in refs .
@xcite .
next , we discuss the decay of @xmath43 to produce the @xmath12 from the @xmath7 cluster and the @xmath48 pair undergoes hadronization to form a meson - meson pair .
the hadronization is now realized combining an extra @xmath49 pair from the vacuum with the @xmath48 pair , as shown in fig .
[ fig : mmproduction ] . in this process
, the @xmath40 @xmath7 cluster will form the @xmath12 state as @xmath66 in the following , we explain in detail which mesons are produced in the hadronization of the @xmath48 pair . we first introduce the @xmath17 matrix @xmath67 which has the followng property @xmath68 next , we rewrite the @xmath49 matrix @xmath69 in terms of meson components , and as a result @xmath69 can be identified with the matrix @xmath70 @xcite @xmath71 which incorporates the standard @xmath72 and @xmath73 mixing @xcite . then , in terms of mesons , the hadronized @xmath48 pair is given by @xmath74 where we have omitted the @xmath73 term because of its large mass .
taking into account the @xmath12 production of fig .
[ fig : mmfsi ] and the weight of eq .
, we obtain the meson - meson production with a @xmath12 baryon as @xmath75 where the re - scattering of @xmath3 and @xmath76 will give the line shape of the @xmath0 state . decay . ]
more specifically the transition amplitude is of the following form : @xmath77 where @xmath78 is the loop function of the two intermediate meson propagators @xcite and @xmath79 are the meson - meson scattering amplitudes obtained in ref .
@xcite , which depend on the invariant mass @xmath80 of the @xmath3 system . in general , the factor @xmath81 should be different from the factor @xmath62 in eq . . with all the ingredients obtained in the previous section , one can write down the invariant mass distributions for the @xmath82 decay as @xcite : @xmath83 where @xmath84 is the total decay amplitude for the decay of @xmath45 depending on the invariant @xmath3 and @xmath4 masses
the @xmath44 decay is a three body decay and the invariant mass distribution with respect to any of the two invariant masses is evaluated by integrating over the other invariant mass . for a given value of @xmath80 , the range of @xmath64 is fixed as , @xmath85 where @xmath86 and @xmath87 are the energies of @xmath72 and @xmath12 in the @xmath3 rest frame .
similarly , for a given value of @xmath64 , we can easily obtain the range of invariant masses allowed for the @xmath47 , namely @xmath80 .
taking @xmath88 mev , we obtain , @xmath89 which indicates that the meson - meson final state interaction to form the @xmath0 state only contributes to @xmath4 invariant masses beyond the peak around @xmath90 mev .
noting the fact that the obtained peak of the @xmath1 resonance is narrow @xcite , we expect that the meson - meson final state interaction will not affect much the peak structure of the @xmath1 resonance in the @xmath4 invariant mass distribution .
similarly , if we take @xmath91 mev , we obtain , @xmath92 which seems to imply that the meson - baryon final state interaction to form the @xmath1 state contributes to the @xmath3 invariant masses beyond the peak / cusp in the @xmath3 mass around @xmath93 mev of the @xmath0 state .
however , we will see that the meson - baryon final state interaction does affect the @xmath0 structure in the invariant @xmath3 mass distribution .
because the factors @xmath62 and @xmath81 are unknown in our model , and the relative phase between @xmath94 and @xmath95 is unknown either , we will study three models : model a takes into account only the meson - baryon final state interaction ; model b includes only the meson - meson final state interaction ; model c considers both meson - baryon and meson - meson final state interactions . because the relative strong phase , @xmath96 , between these two decay mechanisms is unknown , in model c , we take @xmath96 as a free parameter . for these three models ,
we can write the total decay amplitude @xmath84 as , @xmath97 where the relative strong phase @xmath96 is defined as , @xmath98 on the other hand , since the values of @xmath62 and @xmath81 are unknown , we impose a constraint on the values of @xmath62 and @xmath81 such that @xmath99 in this respect , we assume that the strengths of the two decay mechanisms shown in figs . [
fig : mbproduction ] and [ fig : mmproduction ] are equal .
this is a reasonable assumption , since any proposed mechanism should first explain the experimental decay width , and then the corresponding invariant mass distributions will allow one to distinguish between different decay mechanisms .
at present , neither the decay width nor the invariant mass distribution of the @xmath100 is known experimentally . from eq . , we get and @xmath95 into account , we discard the other solution , @xmath101 . furthermore , in the following calculations ,
we assume @xmath62 to be constant and take @xmath102 mev@xmath103 . ]
in this section , we show the numerical results for the @xmath45 decay using the formalism described in the previous section .
first we show @xmath105 in fig .
[ fig : dgdm - etalambda ] .
the dashed and dotted curves represent the numerical results obtained with model a and b , respectively , while the black - solid , red - solid , and green - solid lines stand for the results obtained with model c and the relative strong phase @xmath106 , @xmath107 , and @xmath108 , respectively .
the meson - baryon amplitudes are taken from ref .
@xcite and the meson - meson amplitudes from ref .
@xcite . in fig .
[ fig : dgdm - etalambda ] , a peak corresponding to the @xmath1 resonance can be clearly seen as in ref .
@xcite , regardless of the value of @xmath96 .
the interference between @xmath94 and @xmath95 is instructive and destructive with @xmath109 and @xmath110 , respectively . from fig .
[ fig : dgdm - etalambda ] , it is clear that , for the @xmath4 invariant mass distribution particularly regarding the @xmath1 , the effect from the meson - meson final state interaction is small and can be safely neglected , supporting the assumption made in ref .
we expect that the @xmath1 resonance can be seen and studied from the weak decay of @xmath2 in the future experiments .
mass distribution for the @xmath2 decay .
the dashed , dotted , and solid lines represent the results obtained with model a , b , and c , respectively . ]
next we turn to the @xmath3 distribution shown in fig .
[ fig : dgdm - pioneta ] . from the numerical results of model
b , we see a clear peak / cusp around @xmath111 mev which corresponds to the @xmath0 state .
in addition , the effect of the meson - baryon final state interaction broadens the line shape of the @xmath3 mass distribution . nevertheless , the peak / cusp structure of the @xmath0 state is still visable .
hence , additional experimental information on @xmath60 decay can be used to investigate the nature of the @xmath0 state .
it should be noted that the visibility of the @xmath0 in model c is tied to the assumption we made for the interference of the two hadronization amplitudes and their relative coupling strengths .
an experimental measurement of the line shape will ultimately allow one to extract such information . mass distribution for the @xmath2 decay .
the dashed , dotted , and solid lines represent the results obtained with model a , b , and c , respectively . ]
in the present work we have studied the @xmath3 and @xmath34 invariant mass distributions in the @xmath112 decay to understand better the @xmath0 and @xmath1 resonances . the weak interaction part is dominated by the @xmath5 quark decay process : @xmath6 , while the hadronization part can take place in two different mechanisms . in the first one , the @xmath7 cluster picks up a @xmath8 pair from the vacuum and hadronizes into a meson - baryon pair , while the @xmath10 pair from the weak decay turns into a @xmath11 . in the second mechanism ,
the @xmath7 cluster hadronizes into a @xmath12 , while the @xmath10 pair from the weak process hadronizes into a meson - meson pair together with a @xmath8 pair with the quantum numbers of the vacuum .
the following final state interactions of the meson - meson and baryon - baryon pairs are described in the chiral unitary model that dynamically generates the @xmath0 and @xmath1 states . from the line shapes of the invariant mass distributions , the @xmath0 and @xmath1 states
are clearly seen . on the experimental side , the decay mode
@xmath60 has been observed @xcite and the branching ratio @xmath113 is determined to be @xmath114 , which is one of the dominant decay modes of the @xmath43 state . for the decay of @xmath2 , the final @xmath3 and @xmath4 states are in pure isospin @xmath15 and @xmath16 combinations , respectively .
hence , the @xmath60 decay can be an ideal process to study the @xmath0 and @xmath1 resonances .
future experimental measurements of the invariant mass distributions studied in the present work will be very helpful in illuminating the nature of the @xmath0 and @xmath1 resonances .
for example , a corresponding experimental measurement could in principle be done at besiii @xcite .
we would like to thank e. oset and f. k. guo for useful discussions .
this work is partly supported by the national natural science foundation of china under grant nos .
11475227 , 11375024 , 11522539 , 11505158 , and 11475015 .
it is also supported by the youth innovation promotion association cas ( no . 2016367 ) and the open project program of state key laboratory of theoretical physics , institute of theoretical physics , chinese academy of sciences , china ( no.y5kf151cj1 ) .
s. k. choi _ et al . _
[ belle collaboration ] , `` observation of a narrow charmonium - like state in exclusive @xmath115 decays , '' phys .
lett . * 91 * , 262001 ( 2003 ) [ hep - ex/0309032 ] .
d. acosta _ et al .
_ [ cdf collaboration ] , `` observation of the narrow state @xmath116 in @xmath117 collisions at @xmath118 tev , '' phys . rev .
lett . * 93 * , 072001 ( 2004 ) [ hep - ex/0312021 ] . v.
m. abazov _ et al . _
[ d0 collaboration ] , `` observation and properties of the @xmath119 decaying to @xmath120 in @xmath121 collisions at @xmath118 tev , '' phys .
lett . * 93 * , 162002 ( 2004 ) [ hep - ex/0405004 ] .
m. ablikim _ et al .
_ [ besiii collaboration ] , `` observation of a charged charmoniumlike structure in @xmath122 at @xmath123 = 4.26 gev , '' phys . rev .
lett . * 110 * , 252001 ( 2013 ) [ arxiv:1303.5949 [ hep - ex ] ]
. z. q. liu _ et al . _ [ belle collaboration ] , `` study of @xmath122 and observation of a charged charmoniumlike state at belle , '' phys .
lett . * 110 * , 252002 ( 2013 ) [ arxiv:1304.0121 [ hep - ex ] ] .
r. aaij _ et al . _
[ lhcb collaboration ] , `` observation of @xmath124 resonances consistent with pentaquark states in @xmath125 decays , '' phys .
lett . * 115 * , 072001 ( 2015 ) [ arxiv:1507.03414 [ hep - ex ] ] . v.
m. abazov _ et al . _
[ d0 collaboration ] , `` observation of a new @xmath126 state , '' [ arxiv:1602.07588 [ hep - ex ] ] .
e. klempt and a. zaitsev , `` glueballs , hybrids , multiquarks .
experimental facts versus qcd inspired concepts , '' phys .
* , 1 ( 2007 ) [ arxiv:0708.4016 [ hep - ph ] ] . j. a. oller and e. oset , `` chiral symmetry amplitudes in the s wave isoscalar and isovector channels and the @xmath127 , @xmath128 , @xmath0 scalar mesons , '' nucl . phys .
a * 620 * , 438 ( 1997 ) [ nucl .
a * 652 * , 407 ( 1999 ) ] [ hep - ph/9702314 ] .
j. nieves and e. ruiz arriola , `` bethe - salpeter approach for meson meson scattering in chiral perturbation theory , '' phys .
b * 455 * , 30 ( 1999 ) [ nucl - th/9807035 ] .
g. janssen , b. c. pearce , k. holinde and j. speth , `` on the structure of the scalar mesons @xmath129 and @xmath0 , '' phys .
d * 52 * , 2690 ( 1995 ) [ nucl - th/9411021 ] .
t. wolkanowski , f. giacosa and d. h. rischke , `` @xmath130 revisited , '' phys .
d * 93 * , 014002 ( 2016 ) [ arxiv:1508.00372 [ hep - ph ] ] .
e. van beveren , t. a. rijken , k. metzger , c. dullemond , g. rupp and j. e. ribeiro , `` a low lying scalar meson nonet in a unitarized meson model , '' z. phys .
c * 30 * , 615 ( 1986 ) [ arxiv:0710.4067 [ hep - ph ] ] .
n. a. tornqvist and m. roos , `` resurrection of the sigma meson , '' phys .
lett . * 76 * , 1575 ( 1996 ) [ hep - ph/9511210 ] .
a. h. fariborz , r. jora and j. schechter , `` global aspects of the scalar meson puzzle , '' phys .
d * 79 * , 074014 ( 2009 ) [ arxiv:0902.2825 [ hep - ph ] ] .
a. h. fariborz , n. w. park , j. schechter and m. naeem shahid , `` gauged linear sigma model and pion - pion scattering , '' phys .
d * 80 * , 113001 ( 2009 ) [ arxiv:0907.0482 [ hep - ph ] ] .
jaffe talk at the hadron conference 2015 , jefferson lab , https://www.jlab.org/conferences/hadron2015 + /talks
/ monday/ a. starostin _ et al . _ [ crystal ball collaboration ] , `` measurement of @xmath31 near threshold , '' phys
. rev .
c * 64 * , 055205 ( 2001 ) .
x. h. zhong and q. zhao , `` the @xmath131 reaction at low energies in a chiral quark model , '' phys .
c * 79 * , 045202 ( 2009 ) [ arxiv:0811.4212 [ nucl - th ] ] .
e. oset , a. ramos and c. bennhold , `` low lying @xmath132 excited baryons and chiral symmetry , '' phys .
b * 527 * , 99 ( 2002 ) [ phys . lett .
b * 530 * , 260 ( 2002 ) ] [ nucl - th/0109006 ] .
e. oset _ et al .
_ , `` weak decays of heavy hadrons into dynamically generated resonances , '' int .
j. mod .
e * 25 * , 1630001 ( 2016 ) v. crede and c. a. meyer , `` the experimental status of glueballs , '' prog . part .
* 63 * , 74 ( 2009 ) h. x. chen , w. chen , x. liu and s. l. zhu , `` the hidden - charm pentaquark and tetraquark states , '' arxiv:1601.02092 [ hep - ph ] .
l. roca , m. mai , e. oset and u. g. meiner , `` predictions for the @xmath133 decay , '' eur .
j. c * 75 * , 218 ( 2015 ) [ arxiv:1503.02936 [ hep - ph ] ] .
a. feijoo , v. k. magas , a. ramos and e. oset , `` @xmath134 decay and the higher order chiral terms of the meson baryon interaction , '' phys .
d * 92 * , 076015 ( 2015 ) [ arxiv:1507.04640 [ hep - ph ] ] .
k. miyahara , t. hyodo and e. oset , `` weak decay of @xmath135 for the study of @xmath12(1405 ) and @xmath12(1670 ) , '' phys .
c * 92 * , 055204 ( 2015 ) [ arxiv:1508.04882 [ nucl - th ] ] .
t. hyodo and m. oka , `` determination of the @xmath41 scattering lengths from the weak decays of @xmath136 , '' phys .
c * 84 * , 035201 ( 2011 ) [ arxiv:1105.5494 [ nucl - th ] ] . c. d. l , w. wang and f. s. yu , `` test flavor su(3 ) symmetry in exclusive @xmath136 decays , '' phys .
d * 93 * , 056008 ( 2016 ) [ arxiv:1601.04241 [ hep - ph ] ] .
j. a. oller and u. g. meissner , `` chiral dynamics in the presence of bound states : kaon nucleon interactions revisited , '' phys .
b * 500 * , 263 ( 2001 ) [ hep - ph/0011146 ] .
a. bramon , a. grau and g. pancheri , `` intermediate vector meson contributions to @xmath137 decays , '' phys .
b * 283 * , 416 ( 1992 ) .
l. roca , j. e. palomar and e. oset , `` decay of axial vector mesons into vp and p gamma , '' phys . rev .
d * 70 * , 094006 ( 2004 ) [ hep - ph/0306188 ] .
d. gamermann , e. oset and b. s. zou , `` the radiative decay of @xmath138 into the predicted scalar state @xmath139 , '' eur .
phys . j. a * 41 * , 85 ( 2009 ) [ arxiv:0805.0499 [ hep - ph ] ] .
r. ammar _ et al . _
[ cleo collaboration ] , `` new decay modes of the @xmath44 charm baryon , '' phys .
lett . * 74 * , 3534 ( 1995 ) .
m. ablikim _ et al .
_ [ besiii collaboration ] , `` measurements of absolute hadronic branching fractions of @xmath140 baryon , '' phys . rev .
lett . * 116 * , 052001 ( 2016 ) , [ arxiv:1511.08380 [ hep - ex ] ] .
|
we propose to study the @xmath0 and the @xmath1 resonances in the @xmath2 decay via the final state interactions of the @xmath3 and @xmath4 pairs .
the weak interaction part proceeds through the @xmath5 quark decay process : @xmath6 , while the hadronization part takes place in two different mechanisms . in the first mechanism
, the @xmath7 cluster picks up a @xmath8 pair from the vacuum to form the @xmath9 meson - baryon pair while the @xmath10 pair from the weak decay hadronizes into a @xmath11 . in the second
, the @xmath7 cluster turns into a @xmath12 , while the @xmath10 pair from the @xmath5 decay picks up a @xmath8 pair and hadronizes into a meson - meson pair ( @xmath13 or @xmath14 ) .
because the final @xmath3 and @xmath4 states are in pure isospin @xmath15 and @xmath16 combinations , the @xmath2 decay can be an ideal process to study the @xmath0 and @xmath1 resonances . describing the final state interaction in the chiral unitary approach
, we find that the @xmath3 and @xmath4 invariant mass distributions , up to an arbitrary normalization , show clear cusp and peak structures , which can be associated with the @xmath0 and @xmath1 resonances , respectively .
the proposed mechanism can provide valuable information on the nature of these resonances and can in principle be test by facilities such as bepcii .
|
anomaly - free effective models of loop quantum gravity , derived for spherically symmetric configurations @xcite and cosmological perturbations at high density @xcite , have revealed an unexpected phenomenon : at large curvature , signature change appears to be a generic feature of quantum space - time geometry as provided by this theory @xcite .
not only the general phenomenon but also the specific form of signature change seems to be universal in these models , giving further support of the genericness of the effect . for the typical form of `` holonomy modifications '' used widely in models of loop quantum gravity , the speed of physical modes differs from the classical speed of light by a factor of @xmath6 where @xmath7 is a function of the spatial geometry ( the metric @xmath8 ) , @xmath9 is a quantization parameter often assumed to be related to the planck length , and @xmath10 is a measure for extrinsic curvature ( or the hubble parameter in cosmological models ) . for large curvature @xmath11 ,
the speed is negative and commonly hyperbolic mode equations turn elliptic , which in these models are of the form @xmath12\ ] ] with source and lower - derivative terms @xmath13 $ ] .
all modes @xmath14 , gravitational as well as matter , are affected in the same way .
the overall picture shows some relationships with other approaches to quantum cosmology , mainly the no - boundary proposal of hartle and hawking @xcite , and with other physical phenomena such as transonic flow or phases of nano - wires .
related mathematical questions have been studied in the mathematical literature since the 1930s , following seminal work by tricomi @xcite .
nevertheless , as a basis of effective space - time models , equations of the form ( [ wave ] ) and the phenomenon of signature change show several new and surprising features . in this article , we give a general presentation of implications in cosmology .
section [ s : def ] gives a definition of signature change in the absence of a classical space - time metric , and contains a brief comparison with classical models of signature change .
( a review of the formal origin and gauge independence of equation ( [ wave ] ) and signature change can be found in app .
[ s : spacetime ] . )
section [ s : effects ] provides several related examples in different areas of physics together with the appropriate mathematical formulation in terms of well - posed partial differential equations .
section [ s : cosmo ] applies these results to gravitational questions , compares the pictures based on effective equations with wave - function methods , draws cosmological implications , and ends with cautious notes on global issues .
despite first appearance , the wave equation ( [ wave ] ) is covariant under a deformed algebra replacing classical coordinate or poincar transformations .
the corresponding quantum space - time structure is not riemannian , but has a well - defined canonical formulation using hypersurface deformations , as reviewed briefly in the appendix .
a riemannian manifold has a covariant metric tensor which transforms by @xmath15 when coordinates are changed on the manifold .
accordingly , the line element @xmath16 is invariant .
covariance under coordinate transformations ( of solutions to the field equations ) is canonically represented as gauge - invariance under hypersurface deformations in space - time @xcite . since the latter , along with poincar transformations ,
are modified in the effective space - time structures we are considering , they can no longer correspond to coordinate transformations . as a consequence ,
the effective `` metric '' does not give rise to an invariant line element . instead of using metric - space notions such as geodesics and invariant scalar products of 4-vectors , in order to extract predictions
there is only the possibility of computing canonical observables invariant under the modified gauge transformations . in this way ,
one still has access to all observable information .
however , the lack of a metric structure implies that many of the convenient and well - known techniques of evaluating solutions of general relativity are no longer available .
although there is no standard notion of geodesics and light rays or null lines , it remains straightforward to associate a causal structure to the modified space - time structure underlying ( [ wave ] ) , as long as @xmath17 . instead of computing null lines for a metric
, we just use characteristics of the wave equation ( [ wave ] ) , for instance for gravitational waves to be specific .
( in the presence of inverse - triad corrections , scalar modes may propagate at speeds different from tensor modes @xcite . )
a characteristic is then a hypersurface which at any point is normal to a vector @xmath18 satisfying latexmath:[$(k^t)^2-\beta can define light cones , a causal structure , and derived notions such as penrose diagrams .
characteristics exist as long as @xmath17 . for holonomy modifications ,
@xmath20 changes sign at high density , and the characteristic equation no longer has non - trivial solutions @xmath21 . as a consequence , ( [ wave ] )
does not provide a causal structure in such a regime .
since the same condition of @xmath22 makes the mode equation ( [ wave ] ) turn elliptic , we call this regime `` euclidean , '' while for @xmath17 we call it `` lorentzian . '' in this way , we generalize the two standard discrete choices of signature to a continuous range of values taken by @xmath20 as it varies from the classical limit @xmath23 at low curvature to a negative value at high curvature . unless @xmath24 , the canonical fields from which ( [ wave ] ) is derived do not provide a classical notion of lorentzian space - time or 4-dimensional euclidean space .
but the most important properties regarding physical consequences , including the existence of a causal structure and the type of initial or boundary value problem required for reliable solutions , only depend on the sign of @xmath20 rather than its precise value .
for this reason , we still speak of euclidean or lorentzian signature even if we have neither euclidean nor lorentzian space(-time ) .
signature change has been studied in quite some detail in classical general relativity ( @xmath23 ) ; see for instance @xcite . however , such models are crucially different from what is considered here , in that they have @xmath25 ( or @xmath26 ) changing discontinuously . as a consequence , in classical models
the transition from euclidean to lorentzian signature is always singular , which is not the case in our effective models .
( note that for this reason signature change in the model of @xcite is _ not _ analogous to deformed space - time structures . )
the subtleties and controversies related to the distributional nature of solutions with classical signature change , discussed for instance in @xcite , do not play a role in our context .
( the models we study could be considered as a version of dynamical signature change as anticipated in the conclusions of @xcite . )
it is necessary to consider inhomogeneity in order to see modified space - time structures in which modes obey ( [ wave ] ) .
in particular the phenomenon of signature change , which is the main topic of this article and is realized because @xmath20 may change sign , can only be seen in inhomogeneous models : it is manifested by the relative sign in space and time derivatives in field equations . however , signature change ( or a modified space - time structure ) is not a consequence of inhomogeneity , which in the perturbative context would be dubious @xcite .
the origin of signature change lies in the modification that one makes in the classical dynamics if one quantizes the theory following loop quantum gravity , which provides operators for holonomies ( exponentiated and integrated connections ) instead of ordinary connection components .
this modification appears already in the background dynamics of a homogeneous minisuperspace model , but in this context one does not notice signature change because all spatial derivatives vanish .
perturbative inhomogeneity then makes the effect visible .
still , inhomogeneity or perturbation theory is not the origin of signature change , which is easy to see by the following two arguments : first , signature change persists in a perturbative field equation no matter how small the inhomogeneity is , as long as it is non - zero .
the coefficients of space and time derivatives in ( [ wave ] ) depend on background quantities , and for @xmath22 they have the same sign for all values of inhomogeneity .
secondly , the same form of signature change appears in spherically symmetric models in which inhomogeneity need not be treated perturbatively .
in the explicit form as provided by models of loop quantum gravity , signature change is a new effect .
but it has precursors in physics as well as mathematics .
besides the examples discussed in details below , the case of helically symmetric binary systems @xcite is worth mentioning .
the configuration is described by partial differential equations which change signature along a spacelike direction ( far from the center ) , rather than a timelike one as in our cosmological models . a lorentzian space - time
can not be closed off at a finite time without a boundary . as proposed in @xcite , however , one may postulate that quantum gravity gives rise to a modified space - time structure that allows a transition to euclidean 4-dimensional space .
a euclidean cap can then be attached to a space - time manifold which becomes lorentzian at low curvature . in @xcite and
the literature based on it , this scenario has been used mainly to specify the high - density asymptotics of wave functions satisfying the wheeler dewitt equation of homogeneous models . the scenario suggested by effective constraints in models of loop quantum gravity
can be seen as a concrete realization of the quantum - gravity effects that may give rise to signature change .
the condition that extrinsic curvature vanish at the interface between euclidean and lorentzian parts of semiclassical solutions @xcite is then replaced by @xmath27 .
nevertheless , it is not guaranteed that the same consequences are implied for wave functions , not the least because the possibility of signature change relies on a quantum - geometry effect ( based on the use of holonomies in loop quantum gravity ) which turns the wheeler
dewitt equation into a difference equation @xcite .
the minisuperspace discreteness implied by this difference equation is relevant especially at high density , that is in the euclidean phase where asymptotic properties are discussed according to @xcite .
qualitatively different conclusions for wave functions could then be reached , so that it is not clear whether the close relationship in the space - time picture of @xcite with the models discussed here implies a similar relationship in predictions .
specific details of wave functions might well be closer to @xcite than to @xcite .
we leave this question open in the present article , in which we are concerned mainly with effective equations .
in addition to related cosmological models in terms of space - time properties , there are rather different physical phenomena which give rise to mathematical descriptions similar to ( [ wave ] ) .
the main examples known to us are transonic flow and hyperbolic metamaterials @xcite .
especially the latter phenomenon provides an interesting analog picture of the cosmological effects .
an example of a hyperbolic metamaterial is given by an array of parallel conducting nano - wires immersed in a dielectric medium .
such a material behaves as a conductor in the direction parallel to the nano - wires while in the normal one it exhibits dielectric properties @xcite .
choosing the @xmath28-axis to be parallel to the nano - wires , the wave equation for the component @xmath29 of the electromagnetic field can be written as @xcite @xmath30 for the `` wired '' metamaterials the electric permittivity is @xmath31 , as for the standard dielectric medium . however , because the metamaterial is conducting in the @xmath28-direction , @xmath32 will be negative in a sufficiently low frequency range . and magnetic permeability @xmath33 in some frequency band
have been constructed .
an interesting property of such metamaterials is that the refractive index is negative @xmath34 , leading to very interesting and sometimes counterintuitive behavior @xcite .
the relevance of this phenomenon in the context of analog models of quantum gravity is an open issue . ] in particular , for the drude theory of a conducting medium @xmath35 where @xmath36 is the plasma frequency dependent on the geometry and composition of the metamaterial . at low frequencies ( @xmath37 ) , @xmath38 and
the spatial part of equation ( [ metamaterialeq ] ) is elliptic .
however , for @xmath39 , we have @xmath40 and the sign in front of the second derivative with respect to @xmath28 changes .
the spatial laplace operator becomes hyperbolic , with the @xmath28-component playing a role of the second time variable .
the dispersion relation @xmath41 where @xmath42 , for fixed @xmath43 is in this regime no longer represented by ellipsoids but hyperboloids . in consequence , the @xmath44-space undergoes a topology change .
passing between the elliptic and hyperbolic regions is not only a matter of frequency dependence . for a fixed ( sufficiently low ) frequency
, the transition between the different signs of @xmath32 can be induced by a rearrangement in the structural form of the wired " metamaterial . in particular , melting of the nano - wires to the liquid phase ( associated with a first - order phase transition ) has been observed experimentally . during the process the value of @xmath32 smoothly changes its sign ( from negative to positive ) with increasing temperature .
alternatively one can imagine the transition to be of the second order .
this idea in the context of signature change due to holonomy corrections has been discussed in @xcite .
let us imagine that the nanowires are immersed in a dielectric fluid , having an ability to rotate .
furthermore , let us introduce magnetic dipole - dipole type interactions between the wires by attributing magnetic moments to them .
then , in the high temperature unordered phase , with all nano - wires pointing in random directions , there is hardly any electric conductivity .
but when the phase changes to an ordered one ( lowering the temperature ) , a significant electric current starts flowing , just as time starts flowing in our universe models when the density is small enough to trigger signature change to lorentzian space - time . in the gravitational case
, the process can be modelled by a bi - metric gravity theory , where the effective metric experienced by the fields is @xmath45 the metric may be viewed as a 4-dimensional analog of the inverse of the electric permittivity tensor @xmath46 .
the @xmath47 is an effective mean field having the interpretation of an order parameter , defined such that in the unordered phase @xmath48 ( here @xmath49 ) . in this case
the fields experience the euclidean metric @xmath50 . in the fully ordered phase @xmath51 ,
the lorentzian signature emerges in a spontaneously chosen direction in the 4-dimensional space . in order to describe the transition from the unordered to the ordered state more quantitatively
let us postulate the following form of the free energy for a model with a test field @xmath14 : @xmath52}_{v(\chi^{\mu},\rho)}\ , , \nonumber\end{aligned}\ ] ] where @xmath53 is a constant .
because of the @xmath54 contribution to the kinetic term , @xmath55 is not so(4)-invariant in the internal indices of @xmath56 .
the explicit symmetry - breaking term can be treated perturbatively .
then , equilibrium corresponds to a minimum of the so(4)-invariant potential @xmath57 . in order to model signature change in loop quantum cosmology with maximum energy density @xmath58 ,
the parameters of the potential are fixed such that at energy densities @xmath59 the vacuum state maintains the so(4 ) symmetry the order parameter is @xmath48 . for @xmath60 ,
we choose the potential so that its minimum is located at @xmath61 , with some spontaneously chosen direction in the 4-dimensional configuration space of the order parameter . without loss of generality , we may assume @xmath62 to point in the @xmath63-direction ( @xmath64 and @xmath65 for @xmath66 ) ; thus , @xmath67 and @xmath68 .
the corresponding wave equation for the test field is @xmath69 which is of the form ( [ wave ] ) with @xmath70 as in models of loop quantum cosmology . at @xmath48 ( @xmath23 )
the equation exhibits the @xmath71 symmetry while in the fully ordered state @xmath51 ( @xmath72 ) the @xmath73 poincar symmetry is satisfied . it is worth stressing that the presented model relating signature change with the spontaneous symmetry braking of @xmath74 has no support from more fundamental considerations at the moment .
therefore , its physical validity requires further studies .
in particular , a possible relation of the order parameter @xmath47 with the elementary degrees of freedom of loop quantum gravity remains unclear .
nevertheless , the model presents an interesting example because of its formal resemblance to signature change in loop quantum cosmology .
there is , however , an important conceptual difference : while analog models of signature change are fully well - defined and even have some observable properties , quantum - cosmology effects on space - time itself ( rather than matter fields in space - time ) may be subject to several global problems on which we will comment in sec . [
s : global ] .
partial differential equations of mixed type have been studied in the mathematics literature since the 1930s , beginning with the tricomi problem for the differential equation @xmath75 in the @xmath76-plane .
this type of equation can be seen as an approximation to our equation ( [ wave ] ) around one of the finite boundaries of the elliptic regime , where @xmath77 . here , we recall relevant features of a suitable characteristic problem @xcite .
let us suppose that we would like to solve equation ( [ tricomieq ] ) in a domain @xmath78 , where @xmath79 and @xmath80 are elliptic and hyperbolic domains , respectively .
for the tricomi problem , the sub - domain @xmath80 can not be chosen in an arbitrary manner .
namely , the boundary @xmath81 is determined by characteristics of ( [ tricomieq ] ) attached to the endpoints of @xmath82 at the interface between the hyperbolic and elliptic regions ; see fig .
[ tricomi1 ] .
the characteristics in the hyperbolic region are determined by the characteristic equation to ( [ tricomieq ] ) : @xmath83 .
this equation has two families of solutions with a different sign , enabling us to introduce the characteristic coordinates @xmath84 ( we have @xmath85 in the hyperbolic region . ) ) ( hyperbolic for @xmath85 and elliptic for @xmath86 ) exist uniquely and stably in the shaded region @xmath87 , provided boundary data are imposed on one characteristic @xmath53 of the hyperbolic region together with a curve @xmath88 in the elliptic region .
characteristic coordinates ( [ charac ] ) are indicated by @xmath89 and @xmath90.,width=453 ] solutions to the tricomi equation ( [ tricomieq ] ) are unique and stable if adequate boundary conditions are imposed .
applying the notation from fig .
[ tricomi1 ] , they are as follows : in region @xmath79 , the value of @xmath91 has to be specified at the boundary curve @xmath88 , while in region @xmath80 the value of @xmath91 at the characteristic curve @xmath53 has to be given .
the value of @xmath92 is a result of fixing the boundary condition @xmath93 .
the reason why the boundary condition in the hyperbolic region is imposed on the characteristic curve becomes clear by analyzing the case of the standard wave equation with one spatial dimension : @xmath94 the corresponding characteristic equation @xmath95 has solutions @xmath96 , which allow us to introduce the characteristic ( light cone ) variables @xmath97 in terms of these variables the wave equation ( [ waveeq ] ) reduces to the canonical form @xmath98 with general solution @xmath99 .
an important observation is that an initial - value problem ( cauchy problem ) in the variables @xmath76 translates into the boundary - value problem in the characteristic variables @xmath100 . in particular
, the cauchy - problem initial condition @xmath101 and @xmath102 translates into the dirichlet boundary condition at the light cone @xmath103 and @xmath104 . using characteristics
, it is therefore easier to combine hyperbolic and elliptic regimes . in the case of the tricomi equation
, the boundary conditions in the elliptic domain have to be smoothly extended into the hyperbolic sector .
this requires the introduction of characteristic variables in the hyperbolic sector , imposing boundary conditions at either @xmath105 constant or @xmath106 constant ( but not both ) .
signature change plays an important role in cosmological scenarios . in practical terms ,
one of the most important consequences is the presence of instabilities , as indicated formally by an imaginary speed of sound in mode equations .
the same kind of instability plays a role in the mathematical discussion of equations of the form ( [ wave ] ) : one of the conditions for a well - posed initial / boundary - value problem of a partial differential equation is the stable dependence of solutions on the chosen data , in addition to the conditions that a solution exist and be unique for given data . stability is easily violated when one attempts to use an initial - value problem for an elliptic equation .
for instance , fourier modes of solutions to the standard wave equation would not oscillate as per @xmath107 but change exponentially according to @xmath108 .
the exponentially growing mode implies instability in the sense that solutions , if they exist and are perhaps unique , depend sensitively on the initial values . choosing a boundary - value problem in @xmath109 ( as well as spatial directions ) , on the other hand , ties down the solution at both ends of the @xmath109-range , so that the growing branch is sufficiently restricted for solutions to be stable .
signature change implies instability of initial - value problems , but it is stronger in two respects .
first , signature change affects all modes ( of matter or gravity ) equally , which all become unstable at the same `` time . ''
it is therefore a space - time effect , rather than an exotic matter phenomenon .
secondly , it can be seen in space - time symmetries such as hypersurface deformations or the poincar algebra .
none of these effects happen in known cases of instabilities in matter theories or higher - curvature gravitational actions , such as @xcite .
signature change is therefore more fundamental than other phenomena hat might imply instability .
an important property is the fact that the theory does not provide any causal structure whatsoever when the signature turns euclidean .
when used in cosmological model - building , signature change implies a new and interesting mixture of linear and cyclic models .
one of its main consequences can be described as a finite beginning of the universe . in the simplest versions of existing models of loop quantum cosmology
, effective equations do not imply divergences at high density but instead extend solutions to a new low - density regime @xcite .
the main example is a modified friedmann equation @xmath110 postulated for spatially flat isotropic models sourced by a free , massless scalar @xmath14 .
the energy density @xmath111 is therefore of the form @xmath112 with the momentum @xmath113 of @xmath14 and the scale factor @xmath114 . in ( [ modfried ] ) , @xmath9 is a parameter which approaches @xmath115 in the classical limit , and whose precise value remains undetermined .
( as a parameter motivated by quantum - gravity effects , it is often assumed to be close to the planck length . ) moreover , @xmath116 is a modified version of the hubble parameter .
it is straightforward to see that the modified equation ( [ modfried ] ) implies that the energy density , for fixed @xmath9 , is always bounded , @xmath117 .
the parameter @xmath58 is the maximum kinetic energy density realized at the bounce .
there is then a good chance that cosmological singularities may be resolved .
indeed , the quantum model with a free , massless scalar can be solved completely @xcite , implying that the scale factor @xmath118 never becomes zero .
( see also @xcite . )
equation ( [ modfried ] ) , although it remains unclear how generically it is realized for the background evolution @xcite of general homogeneous models , has suggested a picture in which our expanding universe descends from a preceding collapse phase which produced large but not infinite density at the big bang .
however , when modes on such a bouncing background are subject to equations of the form ( [ wave ] ) , as required for a consistent system of both background variables @xmath119 and inhomogeneous modes in models of loop quantum gravity , the transition is not deterministic owing to the lack of causal structure at high density . in the cosmological case ,
the role of @xmath10 in ( [ wave ] ) is played by the modified hubble parameter @xmath120 . with @xmath121
, one can easily see that signature change happens when the energy density is half the maximum it can achieve ( or half the bounce density of a homogeneous model ) .
given the solution ( [ at ] ) , the mode equation ( [ wave ] ) is elliptic for @xmath122 ( for @xmath58 close to the planck density , the @xmath109-range in which the equation is elliptic is close to the planck time , but it can be significantly longer in models with a scalar potential or strong quantum fluctuations at high density @xcite . )
the high - density regime can not be accessed causally , and there is no deterministic transition from collapse to expansion .
for practical purposes , such a scenario therefore shares some features with the traditional singular big - bang model , in which one must pose initial values just after the singularity .
similarly , one must pose initial values `` after '' the euclidean phase of a signature - change scenario .
however , the relationship between singular and signature - change models is subtle , as shown by the detailed discussion of suitable data for mixed - type differential equations on which we now embark . to be more specific ,
let us focus now on the case of gravitational waves with holonomy corrections . in this case ,
the equation of motion for each component of polarization of the gravitational waves is @xcite @xmath123 where @xmath124 is the hubble parameter and the deformation factor is @xmath125 depending on the value of the parameter @xmath20 , equation ( [ eomphi ] ) can be either hyperbolic ( @xmath126 ) , elliptic ( @xmath127 ) or parabolic ( @xmath128 ) .
the transition between elliptic and hyperbolic type ( associated with signature change ) takes place at @xmath129 . at this moment
, the square of the hubble factor @xmath130 reaches its maximal value @xmath131 . for the model with a free scalar field introduced in the previous section , this takes place at @xmath132 . equation ( [ eomphi ] )
is precisely of the form ( [ wave ] ) with the source term @xmath133=-\left(3h-\frac{\dot{\beta}}{\beta}\right)\frac{\partial \phi}{\partial t}\ , .\ ] ] because generally @xmath134 at @xmath77 , signature change is associated with a divergence of the source .
the divergence , however , does not lead to pathological behavior at the level of solutions to equation ( [ eomphi ] ) because , as we will see later , it is due to a regular pole , which does not disturb regularity of the solution . in the vicinity of @xmath77 , equation ( [ eomphi ] ) can be approximated by @xmath135 which after double integration leads to the solution @xmath136 where @xmath137 and @xmath138 are some functions of the spatial variables . because @xmath20 occurs in the numerator of the integrand ,
the approximate solution is indeed regular .
let us now analyze the behavior of equation ( [ eomphi ] ) in the vicinity of the instance of signature change for the model with a free scalar field .
without loss of generality , we perform an expansion around @xmath139 , for which @xmath140 ( in the first step we have used ( [ at ] ) , ( [ t ] ) and ( [ betarho ] ) in order to write @xmath141 and @xmath142 . ) because @xmath143 , the pole at @xmath144 is of the regular type .
furthermore @xmath145 where we fixed @xmath146 . applying the above expansions to the leading order and reducing to the ( 1 + 1)-dimensional case and @xmath28 directions , leading to @xmath147 . ] , we obtain @xmath148 redefining the time variable to @xmath149 equation ( [ reducedeom1 ] ) can be rewritten as @xmath150 the left - hand side of this equation is identical to tricomi s expression in ( [ tricomieq ] ) while the right - hand side can be treated as a source term . applying the characteristic tricomi problem to obtain solutions to the equation ( [ alatricomi ] )
is therefore justified .
it is worth noting at this point that equation ( [ alatricomi ] ) in the vicinity of @xmath151 reduces to @xmath152 with solution @xmath153 ( in agreement with ( [ approxsol1 ] ) ) .
the solution is perfectly regular across the signature change , in contrast to classical signature change . in particular , signature change resulting from the line element @xmath154 has been widely studied in the literature @xcite . in this case , the analog of equation ( [ alatricomireduced ] ) is @xmath155 , with the metric signature change at @xmath151 .
while both forms of the equation look very similar , the extra factor of `` 2 '' plays a significant role leading to the non - differentiability of solution at the signature change . for discussion of controversies around this issue as well as some proposals of dealing with the problem we refer to ref .
@xcite .
coming back to expression ( [ alatricomi ] ) , the corresponding equation for the fourier component @xmath156 is @xmath157 where @xmath158 has been introduced to absorb the wave number @xmath44 . in order to facilitate solving this equation
, we observe that by differentiating the airy equation @xmath159 and subsequently substituting for @xmath91 using again ( [ airy ] ) , we find an equation @xmath160 identical to ( [ eomphik ] ) if @xmath161 for a suitable ( @xmath44-dependent ) @xmath91 solving ( [ airy ] ) .
solutions to ( [ eomphik ] ) are therefore @xmath162 where @xmath163 and @xmath164 are the airy prime functions ( derivatives of the airy functions ) and @xmath165 and @xmath166 are @xmath44-dependent constants of integration . because of the reality condition for the field @xmath14 the following relations have to be fulfilled : @xmath167 and @xmath168 . as an example
, we will now apply the tricomi boundary conditions to the resulting 2-dimensional solution @xmath169\ , .
\label{alatricomisolution}\ ] ] first , let us choose the contour @xmath170 and fix @xmath171 .
this condition reduces ( [ alatricomisolution ] ) to @xmath172\ , , \label{alatricomisolutionreduced1}\ ] ] with a parameter @xmath173 .
secondly , the form of @xmath88 implies that : @xmath174 finally , by choosing @xmath175 the constant @xmath176 is fixed to equal zero , which leads to @xmath177 the solution together with the boundary @xmath178 is shown in fig .
[ solution ] . for all other values of @xmath176 ,
the contribution from @xmath179 in ( [ alatricomisolutionreduced1 ] ) does not vanish , and therefore the solution is characterized by an exponential growth in the @xmath86 region .
nevertheless , in the domain enclosed by @xmath180 the solution remains regular .
it is transparent from fig .
[ solution ] that the monotonic solution in the elliptic region ( @xmath86 ) transforms smoothly into the oscillatory solution in the hyperbolic domain ( @xmath85 ) , as expected intuitively . ) together with the boundary @xmath178 .
this special solution is exponentially decreasing for @xmath86 , as can be seen by the fading - away of oscillations.,width=453 ] we are now in a position to adapt the tricomi problem to effective space - time models of loop quantum gravity .
we begin with a qualitative discussion which also provides heuristic reasons for why the mixed - type initial - boundary value problem must be used . in a non - singular universe model in which collapse
is joined by expansion , one can start with initial values in the collapsing lorentzian phase in the past and evolve all the way to the boundary with the euclidean phase ( at @xmath139 in the solvable model ) . for a second - order hyperbolic differential equation in this regime
, we need to fix initial values @xmath14 and @xmath181 , where we denote by @xmath181 the derivative of @xmath14 normal to constant-@xmath109 hypersurfaces . in this regime , @xmath181
may therefore be considered as the time derivative of @xmath14 . when we reach the euclidean phase , time no longer exists in a causal sense and our coordinate @xmath109 is one of the four spatial ones .
we will continue denoting the @xmath109-derivative as @xmath181 , now strictly in the sense of a normal derivative replacing the time derivative .
the field and its first normal derivative @xmath181 in the past lorentzian phase as well as the field @xmath14 ( but not its normal derivative ) at the beginning of the expanding phase must be prescribed .
the other data then follow from a solution to the mixed - type differential equation : @xmath14 and @xmath181 at the left boundary of the euclidean region by an initial - value problem in the past lorentzian phase ( left arrow ) ; @xmath181 at the right boundary of the euclidean phase as the limiting normal derivative of the solution in the euclidean phase based on the evolved @xmath14 on the left boundary and the prescribed boundary data @xmath14 on the right boundary ( center circle ) ; and finally the future field @xmath14 and @xmath181 in the expanding phase evolved from initial values at the right boundary of the euclidean phase ( right arrow ) .
this initial / boundary - value formulation is made more precise in the tricomi problem , shown in fig .
[ fig : tricomi ] .
[ fig : ib],width=453 ] once we reach the euclidean phase and enter the range between @xmath144 and @xmath182 in the model , we must switch to a boundary - value problem for well - posedness .
we can use our field @xmath14 evolved up to the part of the euclidean boundary bordering the collapse phase , at @xmath182 .
but we must complete it to a set of boundary values on a hypersurface enclosing the euclidean region in which we are looking for a solution .
for a solution in all of space(-time ) , this complete boundary includes asymptotic infinity in directions other than @xmath109 as well as the boundary at @xmath182 where the euclidean region borders on the expanding phase .
we may take for granted that asymptotic fall - off conditions for our fields are imposed at spatial infinity , but we must choose a new and arbitrary function on the border with the expanding phase . when this function is fixed , together with the evolved past initial data , we obtain a solution in the euclidean phase
. we may then compute normal derivatives on all constant-@xmath109 hypersurfaces in this region and take the limit toward the border with the expanding phase .
we obtain suitable data for an initial - value problem in the latter . in summary , as illustrated in fig .
[ fig : ib ] , in addition to past initial data we must specify one additional function for every mode at the beginning of the expansion phase .
past initial data in the collapse phase do not determine a unique solution in the expansion phase .
there is no deterministic evolution across the euclidean high - density regime .
the intuitive description just given provides the correct picture of free and determined data , but it is not the most suitable one for the problem .
one question one may pose is about the smoothness of solutions : a solution to ( [ wave ] ) should be required to be smooth ( or at least twice differentiable ) at the borders between lorentzian and euclidean regions . in the picture of fig .
[ fig : ib ] , it is not clear whether this requirement is satisfied .
we have to match two solutions , one in the past lorentzian phase and the one in the euclidean phase , so that the normal derivative @xmath181 obtained from these solutions changes differentiably across the border .
since the field @xmath14 on the border with the expanding phase is free and affects the solution in the euclidean region but not in the collapse phase , differentiability seems in danger unless one severely restricts the boundary data left free around the euclidean region : if we keep past initial data fixed but vary the boundary field @xmath14 at the border @xmath182 with the expansion phase , the limit of @xmath181 at @xmath144 will change for solutions of the elliptic equation . only a narrow range of boundary fields @xmath14 would seem to provide a @xmath181 matching the evolved initial data in the collapse phase . if this is so , one would essentially be led back to an initial - value problem across the whole domain , because @xmath14 and @xmath181 everywhere would be uniquely determined by initial values , a conclusion which can not be correct if there is an elliptic regime . by these considerations , for a smooth solution we would have to give up some of the initial values in the collapse phase as free data , but using just @xmath14 and letting @xmath181 be determined by smoothness can not be correct either , because we would end up with a boundary - value problem which is not well - posed in the lorentzian phase . of the equation in the lorentzian phase and on an arc @xmath88 in
the euclidean phase connecting the endpoints of @xmath53 and of a second characteristic arc starting at the same point as @xmath53 ( dashed ) .
the data on the second characteristic arc are determined for solutions of ( [ wave ] ) and can not be chosen freely .
[ fig : tricomi],width=453 ] in order to address differentiability , a tricomi - style characteristic problem is more appropriate . in the lorentzian phase ,
one specifies data not on initial - value surfaces but on characteristics of the differential equation as long as it is hyperbolic . far away from the border with the euclidean region , these characteristics are standard light rays , but they approach the border normal to constant-@xmath109 surfaces and end there .
one example is shown as curve @xmath53 in fig .
[ fig : tricomi ] . considering just one spatial dimension , a second characteristic starting from the same point as @xmath53 completes the first one to a characteristic triangle with the two characteristics as well as the enclosed border with the euclidean region as its sides .
the problem is well - posed @xcite if one specifies the field @xmath14 on one of the two characteristics and on an arc @xmath88 which starts and ends at the border - endpoints of the two characteristics but stays in the euclidean region in between . if these data are smooth on the curve obtained as the union of @xmath53 and @xmath88 one obtains a unique , stable and smooth solution in the interior of the union of the characteristic triangle and the region enclosed by the curve @xmath88 .
smoothness is guaranteed in this way , with data on the curve @xmath88 being completely free except that they must extend the data on @xmath53 smoothly . the indeterministic behavior ( requiring future data ) is therefore confirmed .
if one lets @xmath88 approach the other border of the euclidean region , one can proceed as in the intuitive description based on fig .
[ fig : ib ] .
an initial - value problem starting at the beginning of the expanding phase is then well - posed toward the future . in four dimensions ,
the characteristic triangle is extended to the interior of the future light cone starting somewhere in the collapse phase , and the region enclosed by a curve @xmath88 is extended to a dome enclosed by a 3-dimensional hypersurface in the euclidean region .
although we are not aware of a complete mathematical proof , one may expect the characteristic problem to be well - posed with data given by the mode on half the light cone and on the hypersurface in the euclidean region .
the smoothness of solutions shows that there is still a well - defined manifold which combines the two lorentzian phases and the euclidean phase , but this manifold is no longer riemannian .
standard obstacles to signature change in general relativity therefore do not apply .
the use of a finite region in the characteristic formulation does not cause a problem in cosmology because we always have observational access to a finite part of our universe .
the characteristic triangle can always be chosen large enough to include the whole observational range . besides the boundary , initial - value and the characteristic problems discussed in the previous subsection we would like to present one more possibility of posing the initial - value problem for the evolution in the lorentzian domain .
let us start by finding the solution in the euclidean domain . in this domain ( which we call @xmath183 ) the equations of motion are of the elliptic type .
therefore , for finding solutions in a subset @xmath184 the proper boundary conditions have to be imposed . specifying @xmath185 )
is sufficient to determine a solution @xmath186 to the elliptic equation in the domain @xmath187 uniquely .
an interesting situation corresponds to the case when the boundary @xmath188 is expanded so that it encloses the whole euclidean domain : @xmath189 .
such a boundary can be decomposed as follows @xmath190 .
the @xmath191 are boundaries between the euclidean and lorentzian domains at @xmath192 , respectively , while @xmath193 encloses the euclidean domain at spatial infinity ( assuming that the space part is unbounded ) . the solution @xmath186 at @xmath183 obtained by imposing boundary ( dirichlet or von neumann ) conditions at @xmath194 can now be used to determine initial ( or final ) conditions for the lorentzian sectors @xmath195 , where @xmath196 corresponds to @xmath197 and @xmath198 to @xmath199 .
the @xmath192 ( as previously ) are the times at which signature change is taking place .
let us decide to impose the dirichlet boundary condition @xmath200 , leading to a solution @xmath186 .
then , the cauchy initial conditions at @xmath192 for the solutions @xmath201 in the lorentzian domains @xmath202 respectively are : @xmath203 in this way by solving the equation of motion in the whole euclidean domain , the initial ( or final ) data for evolution in the lorentzian sector can be recovered .
it is worth stressing that , as already discussed in the previous subsection , the continuity of solutions up to at least second time derivatives is required .
( the equations are of second order . )
the additional conditions @xmath204 therefore have to be satisfied . in general
, one can expect that only a subset of all possible boundary values @xmath205 obey this condition .
such a possibility might be attractive because it may put restrictions on the form of the allowed boundary conditions , based on the requirement of mathematical consistency . for the typical equation of motion under consideration , eq .
( [ wave ] ) , the requirement ( [ secondordercon ] ) leads to the following condition of continuity of the source term @xmath206= \lim_{t\rightarrow \pm { t_{\rm max}}^{\mp } } s[\phi_e]\ , .\ ] ] in order to illustrate some features of the method discussed above , let us study the equation of motion @xmath207 with the step - like @xmath20-function @xmath208 in this ( 1 + 1)-dimensional example the variation of @xmath20 is not continuous across the signature change , therefore condition ( [ secondordercon ] ) will have no chance of being fulfilled .
nevertheless , let us choose the following boundary conditions : @xmath209 where @xmath210 are constants .
( the boundary condition at @xmath193 will not play any role . ) with boundary values specified , a solution to eq .
( [ stepfunctioneq ] ) in the euclidean domain @xmath211 can be found : @xmath212 using the solution we can find that @xmath213 which will be used to impose cauchy initial conditions in the lorentzian domains . taking this into account , solutions to eq .
( [ stepfunctioneq ] ) in the domains @xmath214 can be determined : @xmath215 the solutions @xmath186 and @xmath216 are presented in fig .
[ euclidsol1 ] in the form of the density plot for the sample values of @xmath217 and @xmath218 . and @xmath201 to the equation ( [ stepfunctioneq ] ) .
the observed asymmetry is due to the different boundary conditions at @xmath218 and @xmath219 : @xmath220 .
the values @xmath221 , @xmath222 and @xmath223 have beed used.,width=453 ] the solutions obtained in this way are regular in the whole domain @xmath224 .
however , in agreement with the earlier expectations , the second derivatives are discontinuous across the signature change : @xmath225 mixed - type differential equations for modes are a consequence of properties of effective equations in cosmological models of loop quantum gravity .
these properties follow from a general analysis of consistent and anomaly - free realizations of field equations , so that physical predictions made in these models are independent of coordinate or gauge choices .
the same modifications to space - time structure can be seen from commutator algebras at the quantum level , but it is then more difficult to derive dynamical effects .
it is nevertheless interesting to discuss how signature change or its accompanying instabilities could be seen in the evolution of wave functions . as long as one has unitary evolution , as can easily be guaranteed in deparameterized models which provide the majority of explicit wave - function evolutions in quantum cosmology
, the wave function can not be subject to instabilities .
( however , we note that at a fundamental level it may be difficult to have unitary evolution under all circumstances , owing to the problem of time . )
its norm is preserved even throughout high density .
accordingly , the schrdinger or klein gordon type equation which the wave function is subject to does not change its form to become elliptic ; the relevant coefficients in the wave equation do not change sign . after all
, what changes is the signature of space(-time ) , not the signature of ( mini)superspace on which the wave function is defined .
( see also @xcite .
in fact , depending on which representation one chooses for the wave function , the equation it is subject to may be elliptic or hyperbolic even for standard space - time . )
signature change implies instabilities in an initial - value problem for the modes , solving effective equations .
the fields in effective equations are expectation values of mode operators in a state obeying the equation for the wave function . at the level of wave - function evolution ,
signature change would therefore be recognized in an exponential change of expectation values of mode operators .
( if wave packets are used , their peak position would change exponentially . ) for these expectation values ( or peak positions ) , the same sensitivity to initial values as observed in solutions of effective equations would occur , even if they are computed from an evolved wave function .
these observable quantities therefore evolve in an unstable way , causing the same problem as seen in an initial - value formulation for effective equations .
instability problems are just more hidden in schemes that first evolve wave functions and then compute expectation values , but they are just as pressing as in effective equations . in this respect , the situation is reminiscent of quantum chaos , which , compared with the classical phenomenon , is more difficult to define and discuss , but not absent , for wave functions subject to linear differential equations . in both cases , the sensitivity of evolved expectation values to initial choices is relevant .
it has been discussed in detail in the context of quantum chaos .
following @xcite , we can use the same arguments for states in the presence of signature change : unitary quantum evolution ensures that the overlap @xmath226 of two different initial states @xmath227 and @xmath228 is conserved in time , and therefore does not indicate any sensitivity to choices of initial values . however , quantum evolution of classically chaotic systems is very sensitive to perturbations of the hamiltonian operator , in the sense that the overlap @xmath226 changes rapidly if @xmath229 and @xmath230 are now defined as states evolved from the same initial wave function @xmath231 but with different hamiltonians @xmath232 and @xmath233 with some small @xmath234 and a perturbation potential @xmath235 .
similarly , the same definition of the overlap in euclidean signature leads to rapid ( exponential ) change because the perturbed evolution by @xmath236 contains a factor of @xmath237 . in discussions of quantum chaos
, @xmath238 is usually thought of as coming from unknown interactions with an environment hard to control . the same source of @xmath238 exists in quantum cosmology because the precise degrees of freedom and interactions at high density are not well known .
in addition to this , loop quantum gravity is subject to a large set of quantization ambiguities , so that the dynamics is not precisely determined and a second source of @xmath239 results .
signature change in effective equations , in all existing models , is implied by the requirement of anomaly freedom , so that predictions are guaranteed to be independent of gauge choices .
most models in which one evolves wave functions are based on deparameterization , in which one formulates `` evolution '' of a wave packet in terms of a so - called internal time , which is not a coordinate but one of the dynamical fields .
the conditions on well - defined evolution after deparameterization are less severe than the general covariance conditions when one does not select a specific time , be it a coordinate or internal .
if one were to fix the gauge and work in one specific set of coordinates , one could avoid having to use effective equations subject to signature change . similarly
, the possibility of formulating stable wave - function dynamics in one chosen internal time does not mean that wave functions in general evolve in a stable manner .
one would have to show , first , that predictions of one s model do not depend on which field is chosen as internal time ( a task which has not been performed in deparameterized models of quantum cosmology ; see @xcite ) .
after this , one could reliably analyze evolution and test whether it is always meaningful or has to be stopped when a euclidean regime is reached .
performing this task turns out to be much more complicated than analyzing anomaly freedom of effective equations .
but interestingly , even in the absence of such an analysis there are hints that wave - function evolution becomes unstable at high density , in the sense that expectation values of modes change exponentially @xcite .
a well - posed treatment of initial and boundary values in loop quantum cosmology implies significant departures from the scenarios commonly made in this setting .
for instance , there has been some interest in a `` super - inflationary '' phase at high density , around a bounce , during which the hubble parameter grows quickly @xcite .
although the rapid change happens for background variables and is therefore not the same as the instabilities of mode equations such as ( [ wave ] ) , one can easily check in explicit models that the super - inflationary phase falls within the elliptic range of the partial differential equations for inhomogeneity @xcite .
the rapid growth of background variables can not be used for cosmological effects because their values at the border of the euclidean phase must be prescribed as part of the boundary - value problem . obtaining a large parameter with an ill - posed initial - value problem
does not have physical significance .
a related effect has been seen in a combination of loop - modified background equations with inflationary models : at high density , the modified equation for the inflaton has an anti - friction term which can easily push the inflaton up to a high value in its potential @xcite . given the right potential for suitable inflation , the correct initial conditions can therefore be provided . also this effect is subject to the verdict of being based on ill - posed data .
an alteration may , however , still be realized if one can show that for a given background value @xmath240 at the beginning of the expansion phase there must generically be a large @xmath241 according to the well - posed characteristic formulation .
values of @xmath240 far from the minimum of a potential could then be achieved afterwards , using well - posed evolution with the predicted initial values in the expansion phase . with the discussion of initial and boundary values in sec .
[ s : char ] , we can make our expectations for cosmological scenarios more precise .
some data must be chosen at the beginning of the expansion phase , even if the energy density or other coefficients of the differential equation never become infinite .
this feature is shared with the singular big - bang model , in which the main conceptual problem is caused not so much by divergences but rather by the presence of unrestricted initial data .
( in general relativity , it is possible to extend solutions across singularities in a distributional sense , but the extension is not unique @xcite . ) if there would be a unique way of deriving initial values at the beginning of the expansion phase , the singularity of the traditional big - bang model would not be so much of a problem .
one could make clear predictions about the initial state and further evolution of the universe , as attempted in inflationary scenarios with their ( somewhat controversial ) initial conditions for the inflaton . in practical terms , the singularity problem is therefore one of indeterminedness , which is implied but not necessarily equivalent to the occurrence of divergences .
an example for the latter part of the preceding statement is given by signature - change cosmology .
there are no divergences in these models , and yet an important part of initial values for the expansion phase remains free .
the resulting scenario , which we call a finite beginning to emphasize the absence of divergences together with the requirement of choosing new initial values , is rather close to the standard big - bang model . in particular , for practical purposes the need for new initial values once the density is sufficiently low to trigger signature change back to a lorentzian structure makes the euclidean region appear as a singularity .
we therefore speak of a finite beginning instead of a non - singular one , keeping in mind that the main problem of a singularity is the indeterminedness of initial data . there may yet be a crucial difference with the standard big - bang model : what we need to prescribe at the finite beginning , according to fig .
[ fig : ib ] , is only the field @xmath14 , not its normal derivative @xmath181 if we start evolution in the collapse phase .
if one has some means to know @xmath14 in a neighborhood of the border between the euclidean region and the expansion phase , for instance from some hypothetical observations , one can derive @xmath181 and draw conclusions about the field in the collapse phase .
there is therefore some connection between collapse and expansion , although it is not causal and not fully deterministic .
a thorough cosmological analysis of scalar and tensor power spectra in signature - change models is required before one can tell how much about the collapse phase could be deciphered in this indirect way .
while effective methods for constrained systems provide reliable local equations , as confirmed by the results of this paper , there may seem to be several global problems related to the new type of partial differential equations which imply the disappearance of time or causality at high density .
one example has been discussed in @xcite in the context of black holes , where the non - singular beginning in cosmology as detailed here is replaced by a naked singularity ( again in the sense of indeterminism ) with a cauchy horizon .
even if local field equations are regular , the lack of a causal structure in some regions may lead to unacceptable indeterministic behavior at a global level .
a different kind of global problem is indicated by one feature of solutions to tricomi s problem which we have not mentioned so far .
again , locally the equation and its solutions are well - defined .
but generic solutions turn out to have a pole at one point @xmath242 in fig .
[ fig : tricomi ] , where the euclidean arc ends @xcite : although solutions are smooth in the interior of the characteristic region ( for smooth boundary data ) most of them have a pole at one endpoint of the boundary .
( this result explains the ubiquity of sonic booms in analogous acoustic models . )
the solution may remain finite , but not all of its derivatives do .
derived quantities such as the energy density in a matter field could therefore diverge .
if this happens , it is not likely that cosmological perturbation theory gives reliable results , even though the local perturbative equations are regular .
the acoustic analog illustrates a further point : one could think that signature change in cosmology ( or black - hole physics ) is harmless because analogous effects can appear in well - known systems such as transonic motion ( or the model discussed in sec .
[ s : analog ] ) .
the difference is that transonic motion leads to signature change only for excitations in the fluid , while all other propagating degrees of freedom including the bulk fluid motion and space - time are still governed by deterministic equations .
( the speed of light is an absolute limit for any causal motion , very much unlike the speed of sound in a fluid . if @xmath22 in ( [ wave ] ) , all motion is eliminated . )
if the bulk of the fluid is forced to move faster than its own speed of sound , it overtakes any sound wave generated in it , so that its density profile provides future conditions for the wave .
mathematically , this is represented by tricomi s future data .
however , the bulk fluid itself moves in a deterministic ( but not wave - like ) way even if it is faster than its own speed of sound , and no causality issues appear .
the situation is very different when signature change happens for space - time physics , in which case no reference time remains to define a causal structure and no mode can evolve deterministically .
for this reason , signature change in effective space - time models of loop quantum gravity has much more radical consequences than analogous effects in matter systems , as discussed in the present paper as well as @xcite .
we have described several fundamental properties of a new scenario , based on signature change , that has emerged from spherically symmetric and cosmological models of loop quantum gravity in recent years .
we emphasize that this scenario can not be seen in the minisuperspace models traditionally studied in loop quantum cosmology @xcite .
in fact , the possibility of signature change casts significant doubt on the viability of minisuperspace models of loop quantum cosmlogy because in such models one would not see any sign of a disappearing causal structure at high density .
minisuperspace models of loop quantum cosmology may be used for some estimates of background properties , but they can no longer be considered as reliable sources of stand - alone cosmological scenarios .
one always has to go beyond homogeneity to make sure that there is a well - defined space - time structure and to check whether mode equations remain hyperbolic , or to exclude other exotic effects .
signature - change cosmology is therefore a new scenario which crucially relies on inhomogeneous features of loop quantum gravity ( even though , needless to say , it has not yet been derived from a full quantization of gravity ) .
its details , regarding for instance power spectra , still have to be worked out , but we believe that the more mathematical and conceptual properties discussed in this article already show a large number of interesting features . the main consequence in practical terms is the occurrence of instabilities , related to the sensitive dependence of solutions on initial data . in some examples , instabilities may be physical effects which imply rapid change but no inconsistencies . in quantum gravity , however , instabilities of the kind encountered in the presence of signature change are fatal : the theory remains subject to a large number of quantization ambiguities in its equations , and not much is known about suitable initial states for quantum space ( and not just quantum matter ) . with this inherent vagueness
, one can not afford any instabilities that would magnify theoretical uncertainties in a short amount of time , even if this time may be as small as the planck time .
no predictions would be possible .
we are therefore sympathetic with the verdict `` in light of the fact that even this well - behaved signature change system predicts its own downfall , it may be prudent to reassess the inclusion of signature changing metrics in quantum gravity theories . ''
@xcite obtained after a detailed analysis of initial - value problems in a model of classical signature change .
our model is provided by effective equations of loop quantum gravity in which signature change appears to be generic ( and is not included by choice ) .
the verdict of @xcite can therefore be applied to loop quantum gravity at least to the extent that extreme caution is called for when one considers evolution through high density . in the presence of signature change , instabilities can be avoided only if one switches to a 4-dimensional boundary - value problem at high density , giving up causality .
although ambiguities remain in the theory , there are several qualitative effects with interesting implications .
as one of the main observations in this article , the mixed - type partial differential equations for modes in this context strike a nice balance between deterministic cyclic models and singular big - bang models .
there are no divergences , and yet initial data in the infinite past do not uniquely determine all of space(-time ) .
for every mode , one must specify one function at the beginning of the expansion phase even if one has already chosen initial values for the contraction phase .
still , the normal derivative of the field is not free and may carry subtle but interesting information about the pre - big bang .
this work was supported in part by nsf grant phy-1307408 to mb .
jm is supported by the grant dec-2014/13/d
/ st2/01895 of the national centre of science .
we present here a somewhat technical discussion of fundamental properties of space - times underlying ( [ wave ] ) .
symmetries and gauge transformations are especially important in this context , as well as properties of the canonical formulation of gravity . in our following exposition
, we also provide more details on the derivation and reliability of equations such as ( [ wave ] ) in effective models of loop quantum gravity . in spite of its appearance , equation ( [ wave ] )
is covariant in a generalized sense , according to an effective ( and canonically defined ) non - riemannian structure of quantum space - time .
the usual lorentz and poincar symmetries , under which the classical version of ( [ wave ] ) with @xmath23 ( and constant @xmath114 ) would be invariant , are realized in canonical gravity as a subalgebra of the infinite - dimensional algebra ( or rather , algebroid @xcite ) of deformations of 3-dimensional spacelike hypersurfaces in space - time .
( see for instance @xcite . )
these hypersurface deformations are gauge transformations of any generally covariant theory , including gravity .
hypersurface deformations @xcite are more suitable than poincar transformations in situations in which no background space - time metric is assumed .
they provide the proper framework for a discussion of generalized space - time structures as the may be implied by canonical quantum gravity .
hypersurface deformations have an infinite set of generators , spatial ones given by @xmath243 $ ] with spatial vector fields @xmath244 tangential to the hypersurface , and normal ones @xmath245 $ ] with functions @xmath246 on the hypersurface so that the deformation is by an amount @xmath247 along the unit normal @xmath248 .
the classical space - time geometry @xcite ( as well as the canonical form of general relativity @xcite ) imply that these generators have commutators ( or classical poisson brackets ) @xmath249,d[n_2^b]\ } & = & d[n_1^a{\rm d}_bn_2^b - n_2^a{\rm d}_an_1^b]\\ \{h[n],d[n^a]\ } & = & h[n^a{\rm d}_an]\\ \{h[n_1],h[n_2]\ } & = & -d[h^{ab}(n_1{\rm d}_bn_2-n_2{\rm d}_bn_1 ) ] \label{hh}\end{aligned}\ ] ] with the induced metric @xmath250 and covariant derivative @xmath251 on a hypersurface .
when one tries to quantize the theory canonically , one should turn the gauge generators into operators , so that they still have closed brackets given by commutators .
otherwise , the classical gauge transformations would be broken and the quantum theory would not be consistent ; it would have gauge anomalies .
the anomaly problem of canonical quantum gravity is important , but also very difficult and unresolved so far .
( one reason is the presence of the metric in ( [ hh ] ) , which at the quantum level would be an operator and give rise to complicated ordering questions with a quantized @xmath243 $ ] . ) nevertheless , there have been several independent indications in recent years which give some hope that the problem can be solved . at the operator level ,
especially @xmath252-dimensional models have been analyzed in quite some detail , paying attention to the full algebra of gauge generators .
consistent versions have been found in different ways @xcite , including also spherically symmetric models @xcite . independently , effective calculations start from the observation that the quantum operation of commutators together with a closed algebra of operators @xmath253 , @xmath254=\hat{f}^k_{ij}\hat{c}_k$ ] with structure constants ( or functions / operators ) @xmath255 , implies a closed algebra under poisson brackets of effective constraints @xmath256 .
these effective constraints are defined as expectation values @xmath256 of the constraint operators in an arbitrary state @xcite .
effective constraints are therefore functions on the quantum state space , which can conveniently be parameterized by expectation values and moments with respect to a set of basic operators . in addition to these effective constraints , obtained as direct expectation values of constraint operators , there is an infinite set of independent ones , derived from the same constraint operators as @xmath257 for all polynomials in basic operators .
all these functions on the space of states would be zero on the subspace annihilated by the constraints @xmath253 , imposed following dirac s prescription .
the need for an infinite set of effective constraints for every single constraint operator follows from the requirement that a whole tower of infinitely many moments must be constrained together with every constrained expectation value .
a poisson bracket on the quantum state space of expectation values and moments can be defined by @xmath258\rangle / i\hbar$ ] , extended by the leibniz rule to products of expectation values ( as they appear in quantum fluctuations and higher moments ) . with these definitions
, it follows that effective constraints form a closed algebra under poisson brackets if the constraint operators form a closed algebra under commutators .
practically , it is easier to evaluate poisson brackets than commutators , an observation on which the idea of canonical effective equations @xcite and constraints @xcite is based .
also , there are useful approximation schemes , such as a semiclassical one in which one would do calculations order by order in the moments , which are easier to implement for effective constraints than for constraint operators and allow one to handle the unwieldy space of all states more efficiently . for finite orders in the moments
, there is a finite number of independent effective constraints , and the task of computing their algebra becomes feasible .
if we have a closed algebra @xmath254 $ ] of operators , the corresponding effective constraints truncated at some moment order form a closed poisson - bracket algebra up to this order , irrespective of the states used .
this formulation of effective theories is therefore much more general than the usual idea of effective actions @xcite .
( the latter are often combined with further approximations such as derivative expansions , or with restrictions of states such as near - vacuum states for the low - energy effective action . ) in our case , we need not make any assumption on the class of states in order to obtain a closed algebra of effective constraints .
one may just have to include higher orders in the moments for a better approximation to solutions corresponding to strong quantum states ; but even at lower orders , the algebra must be closed ( to within the same order ) .
effective constraints therefore provide a good test of possible anomaly - free quantum constraint algebras . in particular ,
if one can show that no closed effective constraint algebra exists to within some order in moments , using a large - enough parameterization of quantum corrections , there can not be a closed algebra of constraint operators . and
if closure of effective constraints can be achieved only if the classical constraint algebra is modified , the full quantum constraint algebra must be subject to quantum corrections which change the form of gauge transformations .
signature change is just one of the consequences found in this way : instead of ( [ hh ] ) , we then have @xmath259,h[n_2]\ } = -d[\beta h^{ab}(n_1{\rm d}_bn_2-n_2{\rm d}_bn_1)]\ ] ] with a phase - space function @xmath20 which may turn negative ( while the other brackets involving @xmath243 $ ] remain unchanged ) .
canonical field equations consistent with this modified bracket are of the form ( [ wave ] ) , as shown by @xcite using the methods of @xcite .
( the same modification of derivative terms is realized to higher orders in derivatives @xcite . )
so far , calculations in models of loop quantum gravity have been performed only to zeroth order in the moments .
but they still show interesting effects because , in addition to moment terms , the theory implies further quantum corrections . at high density ,
holonomy modifications are relevant , which are implied by the basic assumption of loop quantum gravity @xcite that the gravitational connection can be represented as an operator only when it is integrated and exponentiated to a holonomy . a second effect , inverse - triad corrections @xcite , is more indirect but also related to the basic assumption .
it implies corrections relevant at lower curvature @xcite .
effective constraint algebras have been computed in both cases and found to allow for consistent versions : @xcite as examples for inverse - triad corrections , @xcite for holonomy corrections , and @xcite for combinations of both .
whenever holonomy modifications are present , all generic consistent effective constraints found so far imply mode equations of the form ( [ wave ] ) .
the derivation shows that covariance is not broken because all classical gauge generators have a valid analog as a constraint operator @xmath253 or as an effective constraint . for every classical gauge transformation
, there is a corresponding quantum or effective gauge transformation .
no transformations are violated , including the lorentz and poincar ones that one obtains as a special case of hypersurface deformations @xcite .
nevertheless , it is possible for ( [ wave ] ) to differ from standard covariant wave equations because quantum gauge generators , while they must not be broken for an anomaly - free theory , may be subject to quantum corrections .
these corrections can be seen in the structure operators or functions @xmath255 of the quantum or effective constraint algebras . for holonomy and inverse - triad corrections ,
the classical structure functions are , as in ( [ hhbeta ] ) , multiplied with a phase - space function @xmath20 , which determines a new consistent form of quantum space - time covariance .
the same function appears in mode equations ( [ wave ] ) derived from the effective constraints . even though the speed in ( [ wave ] ) provided by quantum geometry depends on the spatial metric and extrinsic curvature
quantities that are not covariant in classical space - time the effect is frame independent in a subtle way .
the models in which such modified speeds have been derived are anomaly free , so that the classical set of gauge transformations ( given by coordinate changes ) is not violated
. however , not just the dynamics but also the structure of space - time receives quantum corrections . there is a new set of gauge transformations under which the quantum - corrected field equations are invariant , and which has the full set of standard coordinate changes as its classical limit .
the effective theories , including modified speeds they predict , are covariant under these deformed gauge transformations .
accordingly , the space - time structure is no longer classical or riemannian , but it remains well - defined in canonical terms . put differently , one could worry that models in which waves propagate with speeds depending on the spatial metric and extrinsic curvature of constant - time hypersurfaces violate the slicing independence of the classical theory .
classically , for any given space - time , one can , depending on one s set of coordinates , choose equal - time slices with large or small extrinsic curvature , even if the covariant curvature of space - time vanishes .
a speed depending on extrinsic curvature would then suggest that predictions depend on the slicing or the choice of an initial - value surface , which would be unacceptable .
this worry is unjustified because the two initial - value surfaces , one with small and one with large extrinsic curvature , would give rise to different physical solutions of the effective theory even though they would present the same classical solution .
predictions , for instance of the speed , then differ simply because one would consider physically different solutions .
transformations between different slicings are , just like coordinate changes , gauge transformations which are modified in the effective theory .
slicings of one and the same classical space - time , which are related by classical gauge transformations , are not gauge related in the quantum - corrected setting .
m. bojowald and a. skirzewski , quantum gravity and higher curvature actions , 4 ( 2007 ) 2552 , [ hep - th/0606232 ] , proceedings of `` current mathematical topics in gravitation and cosmology '' ( 42nd karpacz winter school of theoretical physics ) , ed .
borowiec , a. and francaviglia , m.
|
signature change at high density has been obtained as a possible consequence of deformed space - time structures in models of loop quantum gravity .
this article provides a conceptual discussion of implications for cosmological scenarios , based on an application of mathematical results for mixed - type partial differential equations ( the tricomi problem ) .
while the effective equations from which signature change has been derived are shown to be locally regular and therefore reliable , the underlying theory of loop quantum gravity may face several global problems in its semiclassical solutions .
some implications of signature - change + in cosmological models of loop quantum gravity + martin bojowald@xmath0 and jakub mielczarek@xmath1 + @xmath2institute for gravitation and the cosmos , the pennsylvania state university , + 104 davey lab , university park , pa 16802 , usa + @xmath3laboratoire de physique subatomique et de cosmologie , ujf , inpg , cnrs , in2p3 53 , avenue des martyrs , 38026 grenoble cedex , france + @xmath4institute of physics , jagiellonian university , ojasiewicza 11 , 30 - 348 cracow , poland + @xmath5department of fundamental research , national centre for nuclear research , + hoa 69 , 00 - 681 warsaw , poland
|
* strange transport * has been the object of intense studies in recent years ( a very recent qualitative review is @xcite ) .
( we use the terminology `` strange transport '' @xcite rather than `` anomalous transport '' which is customary in dynamical systems theory , but has a different meaning in plasma transport theory ) .
the concept first appeared in the theory of stochastic processes , especially the theory of continuous time random walks ( * ctrw ) * @xcite , @xcite .
consider a _ disordered _ system ( _ e.g. _ a turbulent fluid or plasma ) .
the position @xmath0 at time @xmath1 of one of its particles is determined by its interactions with the other particles and/or with external sources . in a strongly disordered system
these interactions are modelled by a random field .
the statistical description of the system thus involves the definition of an ensemble of realizations .
one of the important quantities describing the system is the _
mean square deviation _ ( _ msd _ ) of the random variable @xmath2 . in many cases ,
this is represented asymptotically ( @xmath3 ) by a simple increasing function of time : @xmath4 where @xmath5 the value of the " _ diffusion exponent " _
@xmath6 characterizes the diffusion regime of the system , but they will not be considered here . ] : @xmath7 the subdiffusive and the superdiffusive regimes are the ones called `` strange '' .
it will be seen in the forthcoming sections that the concept of `` strange transport '' involves much more than a simple statement about the behaviour of the msd .
all present theories of strange transport are of stochastic nature . from the very voluminous literature
we only cite here a few of the more comprehensive , physically oriented review papers and books : @xcite - @xcite , where additional references will be found .
as stated in @xcite , transport theories ( e.g. , for fluids or plasmas ) can be constructed on three levels .
@xmath8 ) a purely statistical mechanical theory would be based on the kinetic equation for a set of interacting charged particles , combined with maxwell s equations .
this would be the most fundamental description , but becomes impossibly complicated in practice .
@xmath9 ) the next level of description would be a compromise , based on a `` semi - dynamical '' model of particles moving according to the laws of mechanics ( newton s equation ) , but under the action of a _ fluctuating field _
, representing the action of the turbulent environment .
this leads to stochastic ordinary differential equations of the langevin * * * * type .
more specifically , we consider * v - langevin equations * @xcite , * * * * _ _ i.e . _ _
equations for the position of a `` particle '' : @xmath10,\ \ \ \ \ \ \ \mathbf{x% } ( 0)=0 \label{eeq.3}\ ] ] where the right hand side represents a fluctuating velocity field , defined by its statistical properties .
we always consider a _
velocity field : @xmath11 associated with ( [ eeq.3 ] ) is a kinetic - type first - order partial differential equation for the fluctuating particle distribution function @xmath12 ( whose average is the _ density profile _ ) , called a * hybrid kinetic equation ( hke ) * @xcite . by definition ,
the characteristics of this equation are precisely the v - langevin equations ( eeq.3 ) : @xmath13 due to the property ( [ eeq.4 ] ) this equation is equivalent to the following : @xmath14 = 0 .
\label{eeq.6}\ ] ] this equation ensures the conservation of the number of particles , and is rightly interpreted as a kinetic equation .
@xmath15 ) the last level of description is the level of * continuous time random walk * * ( ctrw ) * theories and * fractional diffusion equations ( fde ) * , in which the deterministic dynamical laws are completely given up , and replaced by a purely random process .
all these concepts will be defined and used in the forthcoming sections .
these three levels are , of course , interrelated : each one of them should be justifiable as an approximation of the more fundamental one .
the strange transport theories associated with ctrw and with fde have been very successful in recent years in modelling many peculiar behaviors observed , among other fields , in plasma and fusion physics .
there is , however , still a serious gap in justifying these purely stochastic models on the basis of a molecular theory , _
i.e. _ , theories of the type @xmath16 ) , or at least @xmath9 ) .
in @xcite we treated several models exhibiting strange transport , showing that a stochastic ctrw can be associated with a semi - dynamical v - langevin equation under certain limiting conditions .
all these models ( diffusion in a fluctuating electrostatic field or in a fluctuating magnetic field ) were either diffusive , or _ subdiffusive_. one of the purposes of the present work is to try to determine whether there exist v - langevin equations leading to _ temporal superdiffusive behavior _ and associating them with a ctrw or with a fractional diffusion equation .
a first work following the same philosophy was done by west et al .
west - grigo . in that paper a specially simplified model of a one - dimensional
, two - state ctrw was used .
this means that a `` particle '' performs a ctrw in one dimension , moving with a velocity that is constant in absolute value : @xmath17 .
the particle moves during a given ( random ) time , after which it suddenly jumps and reverses its velocity .
the velocity autocorrelation function is given as an algebraically decaying function of time .
in a first step of their work , the jumps are supposed to be instantaneous events .
this model leads already to possible superdiffusion of time - fractional type .
the authors are , however , looking for a superdiffusive regime of lvy type ( _ space - fractional diffusion)_. they then modify their model .
the jump pdf and the waiting time pdf of the ctrw are now _ linked _ by assuming that the particles move during their `` jump '' with final velocity @xmath17 : thus longer jumps are `` penalized '' by longer durations ( `` _ _ lvy walk _ _ '' rather than `` lvy flight '' , @xcite ) .
it appears to us that the limitation to two states makes the model unadapted to describing the motion of a physical particle ; it is , however , necessary for the definition of the lvy walk in their model . in the next section
we show that the assumption about the two states can be dropped , and the velocity can be considered as a general random function of time : @xmath18 [ secs .
[ v(t ) ] - [ algeb correl ] ] .
the model can be further extended by considering a _ velocity field _ : @xmath19 ( secs .
[ v - field ] and following ) .
but this case leads to additional difficulties , because it requires the determination of the _ lagrangian velocity correlation _ , a problem that is well known as a basic difficulty of turbulence theory . here
we shall consider only the crudest approximation for the latter quantity , but the application of more precise methods that were recently developed in this field can be envisaged .
in the forthcoming text , the abbreviation `` _ _ pdf _ _ '' denotes a `` probability distribution function '' ; the symbol may equivalently be translated as `` probability density function '' . in the _ continuous time random walk _
model one considers a particle which at time zero makes an instantaneous jump from position @xmath20 to @xmath21 , then waits at the new position during time @xmath22 , then jumps again to a new position , where it waits , etc .
both @xmath23 and @xmath1 are random quantities . in all forthcoming calculations ,
we assume that the random processes are _ symmetrical ( centred ) _ , _ i.e. _ , @xmath24 , @xmath25 ; hence [ see eq .
( [ eeq.1 ] ) ] , @xmath26 in its standard form , a ctrw is defined by two functions : @xmath27 : the pdf of a jump defined by the vector @xmath28 ; fourier transform : @xmath29 @xmath30 : the waiting time pdf ; laplace transform : @xmath31 we only consider in the present work `` classical '' ctrw s , for which @xmath32 and @xmath30 are independent quantities .
this means that @xmath32 is independent of @xmath1 , and @xmath30 is independent of @xmath33 .
processes such as lvy walks @xcite or the vmsc model @xcite are not discussed here . a derived function that plays an important role is the _ memory kernel _ @xmath34 ( in laplace representation ) : @xmath35 this standard ctrw has been solved exactly by @xcite in 1965.let @xmath36 be the pdf of finding the walker at position @xmath23 at time @xmath1 , given that it started certainly at the origin @xmath37 at time @xmath38 [ _ i.e. _ , @xmath39 and @xmath40 as @xmath41 . clearly , from this definition , @xmath36 is the _ propagator _ ( or _ green s function _ ) of the ctrw process .
it is given , in fourier - laplace representation , by @xmath42 this is the celebrated _ montroll - weiss equation_. an equivalent form is : @xmath43 } \label{eeq.9}\ ] ] the latter equation is easily transformed into : @xmath44 ~\widehat{\widetilde{n}}(\mathbf{k},s ) \label{eeq.10}\ ] ] this is the fourier - laplace transform of the integro - differential equation obeyed by the propagator : @xmath45 \label{eeq.11}\ ] ] this _ montroll - shlesinger ( ms ) master equation _ is * non - local * , both in space and in time .
it shows that the rate of change at point @xmath33 and time @xmath1 is determined both by the past history , whose influence is measured by @xmath46 , and by the spatial environment characterized by @xmath27 .
although eqs .
( [ eeq.8 ] ) and ( [ eeq.9 ] ) are explicit solutions for the propagator ( or the _ resolvent _ , as the fourier - laplace transform of the propagator is usually called ) , the expression of the _ inverse _ fourier - laplace transform in terms of known functions of @xmath28 and @xmath1 is , in general , difficult or impossible . it has been shown in a series of important recent works that this problem can be solved under rather general conditions in the so - called * fluid limit , * which is actually an * asymptotic * limit of large distances and times [ @xcite , @xcite , @xcite , @xcite , @xcite ] .
( it should be noted that , in many interesting cases , the notion of `` long '' is ambiguous when there is no intrinsic time scale ! ) the fluid limit is defined as follows ( using dimensionless variables @xmath47 ) [ @xmath48 is the dimensionality of the system ] : @xmath49 in the same limit , the memory function is : @xmath50 this is the same as eq .
( 17.12 ) of @xcite , describing subdiffusion ( @xmath51 ) or normal diffusion ( @xmath52 ) [ see sec .
[ classification ] ] . later in the present work we will need an extension to the range @xmath53 .
this extension is not trivial and requires a closer discussion , which is developed in sec .
[ v(t ) ] .
thus , in the fluid limit , _ the ctrw is entirely defined by the two exponents _
@xmath8 _ _ and _ _ @xmath9 .
the ( f - l ) equation of evolution takes the form : @xmath54 its solution , which is the resolvent of the process , is : @xmath55 eq .
( [ eeq.15 ] ) is , in f - l form , a * fractional differential equation ( fde ) * , corresponding to : @xmath56 a primer on fractional calculus , including the definition of these objects is given in the appendix a. thus , eq . ( [ eeq.15 ] ) can be rewritten as : @xmath57 with the initial condition : @xmath58 thus , @xmath36 is the * propagator * ( or _ green s function _ , or fundamental solution ) * of the fde * ( [ eeq.19 ] ) . although the fourier - laplace inversion of eq .
( [ eeq.16 ] ) is difficult , an important property , _
i.e. _ , the _ scaling of the solution in _ @xmath59 _ _ space _ _ can be derived without performing explicitly the latter operation .
the following argument is due to mainardi @xcite , and is much more compact than the one presented in @xcite , bouch - g and @xcite .
it is based on the well - known properties of the fourier- and laplace transformations : @xmath60 where the double arrow connects a pair of fourier ( laplace ) transforms . applying this property to eq .
( [ eeq.16 ] ) , we find , after a short calculation : @xmath61 applying the relations ( [ eeq.21 ] ) to the last form we find : @xmath62 thus , finally : @xmath63
let us assume now that the density profile has the scaling form : @xmath64 under what condition is this form compatible with eq .
( [ eeq.23 ] ) ?
the left hand side of the latter yields : @xmath65 and the right hand side yields : @xmath66 these two expressions are equal , provided that @xmath67 .
we thus conclude : @xmath68 this is the important _ * scaling relation * _ obeyed by the density profile .
it tells us , in particular , that it depends on space only through the combination @xmath69 , called the _ * similarity variable*. _ from this result we immediately deduce the scaling of the msd : @xmath70 with the substitution @xmath71 we find : @xmath72 identical to eq .
( [ eeq.1 ] ) , with the following definition of the constant : @xmath73 eq .
( [ eeq.26 ] ) provides us with a criterion of `` strangeness '' expressed in terms of the two parameters @xmath74 : @xmath75 the dimensional argument leading to eq .
( [ eeq.26 ] ) does not tell us anything about the constant @xmath76 : we do not even know wheter it is finite or infinite ! the determination of @xmath76 requires the explicit form of the solution through the function @xmath77 .
we proceed here in a more direct way @xcite .
the inverse laplace transform of eq .
( [ eeq.16 ] ) is known @xcite , @xcite : @xmath78 where @xmath79 is the _ mittag - leffler function _
@xcite , a natural ( fractional ! ) generalization of the exponential : @xmath80 the first terms of the expansion of the fourier - density profile are thus : @xmath81 the msd can be determined directly from the fourier - density profile by the well - known formula : @xmath82 when applied to the mittag - leffler function , this yields ( in @xmath48 dimensions ) : @xmath83 two cases are to be distinguished : \a ) @xmath84 . in this case
, the first term has a finite positive value in @xmath85 ; all the other terms in the series vanish in this point .
hence : @xmath86 this completes eq .
( [ eeq.26 ] ) [ note : @xmath87 .
\b ) @xmath88 . the first term in the right hand side of eq . ( [ eeq.34 ] )
contains the factor @xmath89 which diverges in @xmath85 .
the remaining terms are either @xmath90 or @xmath91 in this range of @xmath8 .
thus , the constant in eq .
( [ eeq.26 ] ) is : @xmath92 this means that the density profiles of the ctrw in the present range of @xmath93 have long tails , whatever the value of @xmath94restricting the concept of superdiffusion to the sole discussion of eq .
( [ eeq.26 ] ) would put all the processes with @xmath95 on the same rank .
the distinction among them requires a deeper discussion of the density profile .
these questions will be further discussed in the next section .
the very clear papers of mainardi and coworkers @xcite , main 2001 , @xcite contain an exhaustive study of the green s function of the fde ( [ eeq.19 ] ) in one dimension , @xmath96 . in the forthcoming work ,
we take full advantage of the fact that the general solution @xmath97 ( for all @xmath98 ) of eq .
( [ eeq.19 ] ) has been obtained by these authors .. this solution was also obtained by @xcite in a different form ( in terms of fox functions ) .
we shall first discuss some particular cases ( for @xmath96 ) .
@xmath99 this is the classical case of * normal diffusion*. eq .
( [ eeq.15 ] ) becomes in this case : @xmath100 which is simply the f - l transform of the ordinary diffusion equation ( with a diffusion coefficient @xmath101 ) : @xmath102 it is well known that in this case the msd is : @xmath103 in agreement with eq .
( [ eeq.26 ] ) .
the solution of eq .
( [ eeq.37 ] ) is : @xmath104 the _ propagator _ associated with eq .
( [ eeq.38 ] ) is the _
gaussian wave packet _ : @xmath105 note that eq .
( [ eeq.41 ] ) is a special case of the _ scaling relation _ ( [ eeq.25 ] ) : @xmath106 the gaussian scaling function @xmath107 in the present case is a consequence of the central limit theorem of probability theory .
`` most '' physical systems behave in this normal diffusive way .
but it is the _ strange behavior _ , which deviates from this gaussian form , that will be the main object of our study . [ cols="^ " , ] this is the generic , mixed spatial and temporal fractional diffusion .
in this case there is a competition between super- and subdiffusion , expressed by the ratio @xmath108 [ see eq .
( [ eeq.26 ] ) ] . in the work of del - castillo negrete
et al @xcite , a model with @xmath109 , @xmath110 has been successfully compared with numerical simulations of a set of plasmadynamical equations . in a recent paper @xcite ,
additional values of @xmath111 were explored . in the present work we do not consider the case of space - fractional ( lvy ) superdiffusion .
we thus fix in all subsequent sections : @xmath112
we consider a particle moving in one dimension ; its position @xmath113 at time @xmath114 is determined `` semi - dynamically '' by a v - langevin equation ( @xcite , ch .
11 ) : denote _ dimensional quantities . _
an exception is made for the wave vector @xmath115 , because the symbol @xmath116 will be used later for the kubo number , eq .
( 91 ) . ]
@xmath117 where the random velocity @xmath118 ( _ depending solely on time ) _ is defined statistically by specifying the first two moments of the ensemble of its realizations , supposed to be gaussian , homogeneous and stationary : @xmath119 the function @xmath120 need not yet be specified explicitly , except by requiring @xmath121 .
@xmath122 , the initial velocity rms is supposed to be a given parameter .
let @xmath123 be the ( fluctuating ) distribution function of the particle .
it is represented as usual as : @xmath124 where the ensemble - average @xmath125 is called the _ density profile , _ and @xmath126 is the fluctuation .
the distribution function obeys the _ hybrid kinetic equation ( hke ) _ : @xmath127 the characteristic equation of this first - order partial differential equation coincides with the v - langevin equation ( [ eeq.48 ] ) .
this equation is split , as usual @xcite , into two components by the averaging operation : @xmath128 @xmath129 the latter equation is solved by using the propagator : @xmath130 \label{eeq.54}\ ] ] substituting its solution into ( [ eeq.52 ] ) , we find : @xmath131 \right\rangle \notag \\ & & + \partial _ { x}~\left\langle v(t)~\delta
f\left [ x-% \int_{0}^{t}dt_{1}v(t_{1}),\ 0\right ] \right\rangle \label{eeq.55}\end{aligned}\ ] ] this is the exact solution of the problem .
it is expressed by an integro - differential equation that is non - local both in time ( non - markovian ) and in space ( non - local ) .
we do not repeat here the arguments justifying the * * local approximation * * , by which the fluctuating displacements @xmath132 @xmath133 in the arguments of @xmath134 and of @xmath135 are neglected , and by which the second term of the right hand side , containing the initial fluctuation , is also neglected @xcite .
the result is an ( approximate ) _ local , but non - markovian _ diffusion equation : @xmath136 using eq .
( [ eeq.49 ] ) , this becomes : @xmath137 consider , on the other hand , a particle performing a one - dimensional ctrw characterized , in the _ fluid limit _ , by a gaussian jump pdf : @xmath138 where @xmath139 is a characteristic length and @xmath115 is the dimensional wave vector .
the form of the waiting time pdf is not specified at this stage . the density profile in other words , @xmath140 is the _ propagator _ of the equation .
we shall continue , however , to call this function the `` _ _ density profile _ _ '' , which has a more physical connotation . ]
( in f - l representation ) obeys eq .
( [ eeq.10 ] ) which , written with a dimensional laplace variable @xmath141 , combined with ( [ eeq.58 ] ) reduces to : @xmath142 its inverse f - l transform is : @xmath143 this equation constitutes the main result of the first part of our work.comparing it with eq .
( [ eeq.57 ] ) , we see that the density profile produced ( in the local approximation ) by the semi - dynamical v - langevin equation with a time - dependent @xmath144 , and the density profile produced by a ctrw with @xmath145 , - or , more specifically , _ by the fractional diffusion equation ( [ eeq.59 ] ) _ - obey formally identical equations of evolution .
this statement will be made more precise in the discussion of sec .
[ algeb correl ] .
the quantitative correspondence is obtained by _ identifying the velocity correlation function of the v - langevin process with the memory function : _
@xmath146 i.e. , in f - l representation @xmath147 we note that eqs .
( [ eeq.57 ] ) and ( [ eeq.59 ] ) were also obtained in ref .
@xcite in a quite different manner , which is only valid for the 2-state ctrw .
the relation ( [ eeq.62 ] ) is much more general and was obtained in a simpler direct way .
we now make the problem explicit , by adopting the same form of the velocity correlation as in ref .
@xcite : @xmath148 where @xmath149 is a quantity having dimension [ length]@xmath150[time]@xmath151 .
we then associate with the v - langevin equation ( for @xmath96 ) ( [ eeq.48 ] ) [ or the hybrid kinetic equation ( [ eeq.51 ] ) ] a ctrw with memory function given , in the fluid limit , by eq .
( [ eeq.61 ] ) , leading to the fractional diffusion equation * ( fde ) * ( [ eeq.59 ] ) : @xmath152 this relation agrees with eq . (
33 ) of @xcite .
it is trivially equivalent to : @xmath153 the f - l density profile is obtained from eq .
( [ eeq.64 ] ) in the form : @xmath154 comparing this result with eq .
( [ eeq.16 ] ) , we immediately see that the two expressions become identical by taking the following value for the waiting time exponent @xmath9 : @xmath155 thus , the process corresponds to _ * time - fractional superdiffusion * _ as defined in sec .
[ classification ] : @xmath156 at this point there appears a subtle difficulty ( the following argument results from a remark of j. klafter @xcite ) .
the fde ( eeq.64 ) is _ formally _ derived as the fluid limit of a ctrw with @xmath157 , and @xmath9 in the range ( [ eeq.66 ] ) : this is an _ extension _ of the range defined in eq .
( [ eeq.13 ] ) .
the knowledge of the laplace transform @xmath158 allows us to calculate the _ average waiting time _ between two jumps in the ctrw , by using a well - known formula :
@xmath159 for @xmath160 the average waiting time of the ctrw is infinite .
this denotes a waiting time pdf with a long tail , leading to strange transport , _
i.e. _ , _ subdiffusion _ ( sec .
[ classification ] and @xcite ) . for @xmath52
the average waiting time is finite , corresponding to normal diffusion . for @xmath53 ,
t__he average waiting time is zero _ _ , thus leading to a paradoxical situation .
recalling that @xmath161 is a semi - definite positive variable , it follows that @xmath162 .
thus , any random walker would remain trapped forever at its starting point , in clear contradiction with _
superdiffusion _ , as predicted by the diffusion exponent @xmath163 ( sec . [ classification ] )
. the only other possibility of obtaining a zero average waiting time would be to have @xmath164 negative in part of the range @xmath165 , but then @xmath166 does not qualify for a probability distribution function .
the conclusion of this argument is the following . _
the function _
@xmath167 _ defined by eq .
( [ eeq.65 ] ) is , indeed , the resolvent of the * fractional differential equation ( fde ) * ( [ eeq.64 ] ) , but the latter can not be derived as the fluid limit of a * ctrw*. _ we are now faced with a new subtlety .
the function @xmath168 is the solution of the linear algebraic equation ( [ eeq.64a ] ) .
but , _ is the latter equation the f - l transform of a fractional diffusion equation ? _ one is tempted to associate eq .
( [ eeq.64a ] ) with the fde in @xmath169 representation : @xmath170 with the initial condition : @xmath171 this fractional equation is an `` intermediate '' between a diffusion equation ( @xmath172 ) and a wave equation ( @xmath173 ) @xcite , gor - main 1997 , @xcite .
but , using the general formula for the laplace transform of the caputo fractional derivative @xcite , main 2003 and the appendix of the present paper , the f - l transform of eq .
( [ eeq.69 ] ) is : @xmath174 it thus appears that the general solution of the cauchy problem for eq .
( eeq.69 ) requires _ two initial conditions _ ( like the wave equation ) : @xmath175 and @xmath176_{t=0}$].as a result , the general solution is expressed in terms of _ two _ propagators ; indeed , the solution of eq .
( [ eeq.71 ] ) is : @xmath177 . we refer to ref .
@xcite for a complete solution and discussion of this problem . here
we shall only consider the simpler special case ( also considered in @xcite ) : @xmath178 @xmath179 in this case , eq .
( [ eeq.71 ] ) reduces to eq .
( [ eeq.64a ] ) , and eq .
( [ eeq.65 ] ) represents the f - l propagator of eq .
( [ eeq.69 ] ) equipped with the initial conditions ( [ eeq.73 ] ) . the explicit solution for the propagator of ( [ eeq.69 ] ) ( i.e. , the f - l inversion of eq .
( [ eeq.65 ] ) was obtained analytically by mainardi @xcite , @xcite ( and also by metzler and klafter @xcite in terms of fox h - functions ) : @xmath180 this has precisely the _ scaling form _ ( [ eeq.25 ] ) , with the similarity variable : @xmath181 the special function @xmath182 is called the _ mainardi function _ : it is defined for any order @xmath183 as : @xmath184},\ \ \
0<\nu < 1 \label{eeq.77}\ ] ] this is a special case of a _ wright function _ ,
defined as follows ( @xcite ) : @xmath185 the series ( [ eeq.77 ] ) is convergent , but the convergence is very slow . in practice , for the numerical calculations displayed below , we take @xmath186 terms in the series .
mainardi et al 2001 @xcite also showed that the _ asymptotic behavior _ of the propagator is given by a _
stretched exponential _ , which for the value @xmath187 is : @xmath188 with : @xmath189 ^{-1/2 } , \notag \\ a & = & \frac{1-\gamma } { \gamma } , \notag \\ b & = & \gamma \left [ \frac{(2-\gamma ) ^{2-\gamma } } { 4}\right ] ^{1/\gamma } \notag \\ c & = & \frac{2}{\gamma } , \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ 0<\gamma < 1 \label{eeq.80}\end{aligned}\ ] ] the important point here is that the present density profile decays asymptotically according to a stretched exponential law , with exponent @xmath190 .
this implies that _ the propagator decays asymptotically * faster * than the gaussian ( contrary to the subdiffusive case _
@xmath191 discussed in sec@xmath192[classification]_. _ mainardi calls this feature a `` _ _ thin tail _ _ '' .
it thus appears clearly that _ superdiffusion is not necessarily associated with a long ( `` fat '' ) tail of the propagator . _ another important property of the propagator ( [ eeq.75 ] ) [ _ i.e. _ , of the mainardi function @xmath193 for @xmath194 _ i.e. _
, @xmath195 is its possessing for @xmath196 two symmetrical maxima that move away from the origin , while becoming wider .
as pointed out above , the fractional equation ( [ eeq.69 ] ) , written in a simpler notation ( @xmath197 ) as : @xmath198 is a kind of interpolation between a diffusion equation ( @xmath52 ) and a wave equation ( @xmath199 ) .
thus , the solution represents a wave propagating in both directions away from the origin , and damping out on the way .
we illustrate these properties by the plot of fig .
[ lfde 3 ] , corresponding to @xmath200 , hence @xmath201 . for @xmath202
, the propagator tends toward @xmath203 , as it should ( [ eeq.70 ] ) . at four different times .
@xmath204,height=305 ] we show in fig .
[ lfde 4 ] the tail of the propagator ( for @xmath205 ) compared to the tail of the corresponding gaussian .
this situation is to be contrasted with figs .
[ lfde 1 ] and [ lfde 2 ] : both the time - fractional _ subdiffusive _ propagator and the space - fractional _ superdiffusive _ ( lvy ) propagator have _ `` fat '' tails_. , compared to the corresponding gaussian packet .
@xmath206,height=283 ] figs .
[ lfde 3 ] and [ lfde 4 ] allow us to understand how superdiffusion is possible in spite of the thin exponential tail .
the dispersion of the matter is produced not only by the broadening of the initial distribution ( as in ordinary diffusion ) but also by the symmetric wave - like outward motion of the maximum .
the results of the present section are not really new : all these results were also obtained in ref .
@xcite , using , however , the 2-state model .
we showed here that the results are easily extended to an arbitrary fluctuating velocity depending solely on time .
our main result can be summarized as follows . _
the motion of a particle in one dimension , with a random velocity depending solely on time , obeying a semi - dynamical v - langevin equation ( v - le ) ( [ eeq.48 ] ) , can be equivalently described by a time - fractional differential equation ( fde ) .
there is , however , no ctrw whose fluid limit is the fde ( [ eeq.74 ] ) . _ the v - le
is characterized by a velocity correlation depending ( for all positive times ! ) algebraically on time , @xmath207 , with an exponent @xmath208 .
the equivalent fde is characterized by a spatial exponent @xmath145 and a waiting time exponent @xmath209 thus @xmath53 .
this fde leads to _ time - fractional superdiffusion _ , with diffusion exponent @xmath210 .
the density profile has two maxima that are moving symmetrically away from each other , while broadening with increasing time .
the asymptotic decay is exponential , decreasing faster than a gaussian in space ( `` thin tail '' ) . the model v - le considered here , though allowing an exact analytical solution , has a physical drawback .
the algebraic law of the velocity correlation @xmath211 can not be valid for short times.for an obvious physical reason , @xmath212 [ as required by eq .
( [ eeq.49 ] ) ] , instead of diverging as in the present model .
the model should certainly be improved by using a correlation function that satisfies this constraint .
we now generalize our model by considering , instead of a velocity depending on a single time variable , a _ velocity field in two dimensions , for which the velocity depends on both time and space _ @xmath213 .
we now return to eq .
( [ eeq.3 ] ) , and consider a true velocity field , _ i.e. _ , a function both of space and time .
we immediately note that this generalization introduces nothing new in one dimension .
we recall the important condition of zero divergence ( [ eeq.3 ] ) which ensures the correct physical status of the hybrid kinetic equation , _
i.e. _ , the equivalence of eqs .
( [ eeq.4 ] ) and ( [ eeq.5 ] ) .
clearly , the constraint ( [ eeq.3 ] ) reduces , in one dimension , to @xmath214 , hence the velocity can only depend on time , and we are brought back to the situation of sec .
[ v(t ) ] . in order to study a non - trivial problem
, we must consider @xmath215 .
specifically , we shall consider in the rest of the present work _ two - dimensional systems _
, @xmath216 .
as will be shown below , such systems are actually of much greater physical interest than one - dimensional ones .
we thus consider a particle whose position is specified by the vector @xmath217 and the velocity @xmath218 .
the corresponding ( vectorial ) v - langevin equation is : @xmath219,\ \ \ \ \ \ \ \
\mathbf{x}(0)=% \mathbf{0}. \label{eeq.82}\ ] ] as a result of the previous discussion , we may associate with this v - langevin equation a hybrid kinetic equation , _ provided that the velocity field has zero divergence : _ @xmath220 contrary to the one - dimensional case , this condition can be realized with a non - trivial velocity _
field_. the associated hybrid kinetic equation possesses both required properties : it conserves the normalization of @xmath221 and its characteristic equations are eqs .
( [ eeq.82 ] ) .
thus , the two following forms are equivalent : @xmath222 = 0 , \label{eeq.84}\ ] ] @xmath223 we note that the condition of zero divergence of a two - dimensional velocity field is realized , in particular , by the intensely studied model of a charged particle moving perpendicularly to a strong constant magnetic field @xmath224 , in presence of a fluctuating electrostatic field ( see , _ e.g. _
@xcite , where many other references are given ] .
the coordinates of its guiding centre are animated by the well - known electrostatic drift velocity ( in gaussian units ) : @xmath225}{% \partial y}\right\vert _ { \mathbf{x = x(}t ) } , \notag \\ \frac{dy(t)}{dt } & = & \frac{c}{b}~\left .
\frac{\partial \phi \lbrack x , y , t]}{% \partial x}\right\vert _ { \mathbf{x = x(}t ) } , \label{eeq.86}\end{aligned}\ ] ] where @xmath226 is the ( fluctuating ) scalar potential .
the statistical properties of the velocity field are thus derived from the statistical definition of the potential .
we assume that the latter is a gaussian centred , homogeneous , isotropic and stationary random process , defined by a vanishing first moment and by the following factorized form of the second moment : @xmath227 the second relation defines the * eulerian potential correlation * , _ i.e. _ , the correlation between the values of the potential at the origin and at a _ fixed _ point @xmath228 at time @xmath161 . in the present work
we consider only the case where the spatial part of the correlation has a gaussian form : @xmath229 where @xmath230 is the so - called _
correlation length_. the temporal part @xmath120 will not yet be specified explicitly .
it should , however , satisfy the condition : @xmath121 , and is supposed to depend only on the absolute value of @xmath161 .
it is convenient to introduce from here on dimensionless quantities ( denoted by lower case letters ) , defined as follows : @xmath231 here @xmath232 is a measure of the intensity of the potential fluctuations , _
e.g. _ , the root mean square value of the potential fluctuations .
@xmath233 is an intrinsic characteristic time , to be defined later [ sec .
[ algeb gen ] ] .
the v - langevin equations of motion ( eeq.86 ) reduce to : @xmath234 , \notag \\
v_{x}(\mathbf{x},t)=-\frac{\partial \phi ( \mathbf{x},t)}{\partial y},\ \
\ v_{y}(\mathbf{x},t)=\frac{\partial \phi ( \mathbf{x},t)}{\partial x}. \label{eeq.90}\end{gathered}\ ] ] the dimensionless constant @xmath116 is called the _ kubo number : _
@xmath235 given the form of eqs .
( [ eeq.90 ] ) it is even more convenient to use a further reduction of the time variable : @xmath236 the v - langevin equations then reduce to : @xmath237,\ \ \ \
\mathbf{x}(0)=\mathbf{0}. \label{eeq.93}\ ] ] the relevant dimensionless eulerian correlations are now defined as follows : @xmath238 with ( [ eeq.88 ] ) : @xmath239 a first , well - known general result is obtained from the v - langevin equation ( [ eeq.93 ] ) : the _ mean square deviation ( msd ) _ in the @xmath240-direction is expressed as follows : @xmath241 here @xmath242 is the * lagrangian velocity correlation * , calculated along the trajectory @xmath243 of the particle , _ i.e. _ , using the solution of eq .
( [ eeq.93 ] ) : @xmath244\right\rangle .
\label{eeq.97}\ ] ] the msd is related to the dimensionless _ running diffusion coefficient _ @xmath245 in the @xmath240-direction by the well - known einstein relation : in eq .
( [ eeq.98 ] ) is to be replaced by @xmath246 . in the present situation , although the system is two - dimensional , we are looking for the diffusion in the single direction @xmath240 , hence we must take @xmath96 .
if we calculated @xmath247 we should take @xmath216 . ]
@xmath248 we now consider the evolution of the density profile @xmath249 .
the distribution function is governed by the hybrid kinetic equation ( [ eeq.84 ] ) , or equivalently , ( [ eeq.85 ] ) .
the latter is treated by a straightforward generalization of the reasoning leading from eq .
( [ eeq.51 ] ) to ( eeq.56 ) ; it will not be repeated here .
the final result differs from the latter equation in that the simple one - dimensional velocity correlation @xmath250 is replaced by the _
lagrangian velocity correlation tensor _
@xmath251~v_{k}[\mathbf{x}(\theta ) , \theta ] \right\rangle = % \mathcal{l}_{jk}(\theta -\theta _ { 1})$ ] : @xmath252 clearly , the relation ( [ eeq.96 ] ) can be obtained directly from eq .
( eeq.99 ) .
the study of transport is now reduced to the ( difficult ) determination of the lagrangian velocity correlation .
more precisely , we want to find a relation between the ( given ) eulerian velocity correlation and the lagrangian one . in the present work
we will use only the simplest approximation method for this problem : the _ corrsin approximation _
@xcite , @xcite .
we rewrite the exact lagrangian velocity correlation as follows : @xmath253\right\rangle \notag \\
= \int d\mathbf{x}~\left\langle v_{j}(\mathbf{0},0)~v_{n}(\mathbf{x},\theta ) ~\delta \lbrack \mathbf{x}-\mathbf{x}(\theta ) ] \right\rangle \label{eeq.100}\end{gathered}\ ] ] in the corrsin approximation , the average in the integrand is assumed to be factorized as follows : @xmath254\right\rangle \label{eeq.101}\ ] ] the first factor in the integrand is recognized as the eulerian correlation tensor , assumed to be of the form ( [ eeq.87 ] ) , thus : @xmath255\right\rangle .
\label{eeq.102}\ ] ] the last factor is evaluated as follows , in the second cumulant approximation : @xmath256\right\rangle & = & \int d\mathbf{k}~e^{i\mathbf{k\cdot x}}~\left\langle e^{-i\mathbf{k\cdot x% } ( \theta ) } \right\rangle \notag \\ & \approx & \int d\mathbf{k}~e^{i\mathbf{k\cdot x}}~\exp \left [ -~\tfrac{1}{2}% ~k^{2}\left\langle x^{2}(\theta ) \right\rangle \right ] \notag \\ & = & \frac{1}{2\pi ~\left\langle x^{2}(\theta ) \right\rangle } ~\exp \left [ -~% \frac{x^{2}+y^{2}}{2~\left\langle x^{2}(\theta ) \right\rangle } \right ] .
\label{eeq.103}\end{aligned}\ ] ] thus : @xmath257 .
\label{eeq.104}\ ] ] we now use the explicit forms ( [ eeq.95 ] ) for the eulerian correlation.it is immediately seen by symmetry that the non - diagonal components of the lagrangian tensor vanish , and the two diagonal ones are equal to each other
. @xmath258 with : @xmath259 .\ ] ] the integration over @xmath240 and @xmath260 is elementary . the result is combined with eq .
( [ eeq.96 ] ) to yield : @xmath261 ^{2 } } \label{eeq.106}\ ] ] we thus ended with an _ integral equation _ for the lagrangian velocity correlation in the corrsin approximation @xcite , @xcite .
this equation will be solved in secs .
[ determ lagr 1 ] and [ determ lagran 2 ] . the diagonal character of the lagrangian velocity correlation introduces a significant simplification in the equation for the density profile , which reduces to the following non - markovian diffusion equation : @xmath262 the diffusion coefficients in the @xmath240- and @xmath260-directions are equal to each other .
we may therefore drop the subscripts in eq .
( [ eeq.98 ] ) and write simply : @xmath263 whenever this integral converges , its limit for @xmath264 defines the ordinary diffusion constant : @xmath265 . whenever @xmath266 is finite and positive , the regime is normal diffusive .
we now make the problem more explicit .
we consider a class of v - langevin equations ( and associated hybrid kinetic equations ) for which the temporal part of the eulerian velocity correlation , @xmath267 , is a long - tailed function that decays algebraically for long times .
it is of the general form ( [ eeq.63 ] ) , but we take now some additional precautions to ensure a correct behavior at short times
. we must , in particular , avoid the divergence at @xmath202 .
physically , any correct ( dimensional ) correlation function should have the property : @xmath268 .
we thus define a characteristic cut - off time @xmath233 as the value of @xmath161 at which the right hand side of eq .
( [ eeq.63 ] ) equals @xmath269 . for all times
@xmath270 we put @xmath271 .
note that a correlation of the form ( [ eeq.63 ] ) has no intrinsic time scale ; but the adoption of a cut - off introduces a characteristic time @xmath233 .
the latter allows us to complete the definitions of the dimensionless time @xmath272 ( [ eeq.89 ] ) , of the kubo number ( [ eeq.91 ] ) and of the dimensionless variable @xmath273 ( eeq.92 ) .
we thus adopt the final form of the dimensionless temporal eulerian velocity correlation as : @xmath274 for convenience , @xmath15 will be called the * eulerian exponent * ; by definition it will be limited to the range ( @xmath275 ) . anticipating the results of the next section
, it will be shown that the lagrangian correlation function @xmath276 has the same general algebraic form , with different values of the exponent and of the coefficients : @xmath277 in order to ensure a correct connection of the two parts , we must take : @xmath278 @xmath279 will be called the * lagrangian exponent * : its value depends on the eulerian exponent @xmath15 and on @xmath116 .
it will be determined in the next sections . before going into specific calculations , we wish to derive an important property relating the nature of the macroscopic transport to the value of the lagrangian exponent . combining the definition of the dimensionless running diffusion coefficient ( [ eeq.108 ] ) with the form ( [ eeq.110 ] ) we obtain : @xmath280 the integrals
are easily evaluated : @xmath281 the more interesting long - time part is rewritten as the sum of a constant and of a time - dependent term ; using also eq .
( [ eeq.111 ] ) we find : @xmath282 we now see the existence of two sharply separated ranges of the lagrangian correlation exponent @xmath283 \a ) @xmath284 . in this case
the second term in the right hand side of ( [ eeq.114 ] ) is a monotonically _ increasing function of time .
_ hence , for sufficiently long times this term strongly dominates the first constant term , and the running diffusion coefficient behaves asymptotically as : @xmath285 the regime is thus ( time - fractional ) * superdiffusive*. \b ) @xmath286 eq .
( [ eeq.113 ] ) is no longer valid , but the integral in ( [ eeq.112 ] ) can be evaluated as : @xmath287 the regime is `` * * weakly superdiffusive * * '' .
\c ) @xmath288 .
the behavior is radically different .
the second term in the right hand side of eq . ( [ eeq.114 ] ) is now a monotonically _ decreasing function of time . _ for long times this term decreases to zero .
hence the running diffusion coefficient tends toward a _ positive constant _ : @xmath289 the regime is now * normal diffusive * for all values of @xmath290.the asymptotic @xmath291 is the ordinary diffusion constant .
clearly , the value @xmath292 is a very special one : as @xmath279 reaches @xmath269 from below , the regime changes abruptly from a superdiffusive to a normal diffusive one , and remains so for all values of @xmath290 .
this phenomenon is illustrated in fig .
[ lfde 5 ] . obtained from the kinetic equation , for various values of @xmath293 and @xmath290.,height=208 ] the difference between the normal diffusive regime obtained from the kinetic equation and the subdiffusive regime predicted by the ctrw for the same value of @xmath290 appears clearly in fig .
[ lfde 6 ] .
, as obtained from the kinetic equation ( solid red ) and as predicted from the ctrw ( dotted blue).,height=254 ] in order to obtain a clearer picture of what happens at this point , we study the density profile . more specifically
, we investigate whether its evolution can be described by an equivalent fde , as was found in sec .
[ ctrw fde ] . as shown in sec .
[ v(t ) ] , there exists a _
formal _ equivalence between , on one hand , the non - markovian diffusion equation ( [ eeq.107 ] ) obtained from the hybrid kinetic equation and on the other hand , the montroll - shlesinger equation for a ctrw in the fluid limit .
it requires the identification ( [ eeq.61 ] ) of the lagrangian velocity correlation with the memory function of the ctrw ( in dimensionless variables ) : @xmath294 or , in laplace representation : @xmath295 the laplace transform of the lagrangian correlation ( [ eeq.110 ] ) is , for long times : @xmath296 comparing with eq .
( [ eeq.14 ] ) we find that the temporal exponent of the `` formally equivalent '' fde is : @xmath297 substituting this value into eq .
( [ eeq.16 ] ) we find a result similar to ( [ eeq.65 ] ) , with @xmath279 replacing @xmath298 the resolvent given by the montroll - weiss equation is : @xmath299 the difference with ( [ eeq.65 ] ) is that we should not limit consideration to @xmath293 .
we consider , instead the various cases discussed above .
\a ) @xmath284 .
the situation is the same as in sec .
[ algeb correl ] .
there is indeed a complete identity between the kinetic result and the fde with @xmath300 , _
i.e. , _ @xmath53 .
as shown , however , in sec .
[ algeb correl ] , there is no equivalent ctrw in this case .
the regime is one of _ time - fractional superdiffusion_. the density profile is provided explicitly by the mainardi function ( [ eeq.75 ] ) ) is really relevant here . ] .
it has two moving maxima , as represented in fig .
[ lfde 3 ] and decays asymptotically as a stretched exponential
. the diffusion exponent @xmath301 .
all this is in agreement with the kinetic result .
\b ) @xmath302 .
the resolvent ( [ eeq.122 ] ) of the fde reduces to : @xmath303 this is the fourier - laplace transform of a _ gaussian packet _ , characteristic of a _ normal diffusive regime_. on the other hand , the result ( [ eeq.116 ] ) shows that the running diffusion coefficient obtained from the kinetic equation grows , slowly but indefinitely in time .
the fde is no longer equivalent to kinetic theory : @xmath292 appears as a _
bifurcation point_.
\c ) @xmath304 .
the propagator ( [ eeq.122 ] ) corresponds now to a fde with @xmath305 .
it describes a _ time - fractional subdiffusion , _ which can be derived as the fluid limit of a ctrw .
the corresponding propagator has a single maximum at the origin , and decays according to a stretched exponential [ see eq .
( [ eeq.43 ] ) ] . this regime has been studied in detail in @xcite . on the other hand ,
the kinetic theory predicts a _ normal diffusive _ regime throughout this range of @xmath279 [ eqs .
( [ eeq.114 ] ) , ( [ eeq.117 ] ) ] .
there appears now a _ divorce between ctrw , fde and kinetic theory .
_ @xmath292 has indeed the property of a bifurcation point : as the parameter @xmath279 increases from @xmath90 ; the propagator first follows the `` fde branch '' . at @xmath306
it leaves this branch and follows the completely different `` diffusive branch '' .
the kinetic equation for the propagator in this regime is obtained from eqs .
( [ eeq.99 ] ) and ( [ eeq.110 ] ) : @xmath307 for long enough time , the equation can be markovianized by neglecting the retardation in the right hand side . this approximation is justified by our knowledge that a constant diffusion coefficient exists .
the final asymptotic equation is an ordinary diffusion equation : @xmath308 with @xmath291 given by eq .
( [ eeq.117 ] ) .
nevertheless , for any finite time this markovianization is not reliable , because of the non - exponential lagrangian correlation , hence the density profile is in general non - gaussian .
\d ) @xmath309 in this range
the divorce between kinetic theory and fde is complete .
the latter looses its meaning .
indeed , it would correspond to @xmath310 , which would describe a waiting time distribution that is an increasing function of time .
alternatively , its laplace transform diverges : @xmath311 as @xmath312 . on the other hand ,
the markovianization of the kinetic equation ( [ eeq.124 ] ) is more and more justified as @xmath276 decreases faster for large @xmath273 .
we now determine the lagrangian velocity correlation as a function of time and of the parameters @xmath15 and @xmath116 of the input eulerian velocity correlation , by solving the integral equation ( [ eeq.106 ] ) .
the procedure used is one of successive numerical iterations @xmath313 , starting with the natural eulerian trial function @xmath314 .
we checked the convergence of the procedure .
as can be seen in fig .
[ lfde 7 ] , the second iteration provides already an excellent approximation . , @xmath315,height=248 ] the first very important qualitative result is the following .
an input eulerian correlation ( [ eeq.109 ] ) with an algebraic tail produces a lagrangian correlation having the same type of algebraic tail .
this is clearly seen in a log - log representation of the result , as shown in fig .
we thus confirm the anticipated form ( [ eeq.111 ] ) .
our main task will be the determination of the lagrangian exponent @xmath279 , of the corresponding `` intensity '' @xmath316 and of the cut - off time @xmath317 as functions of the input eulerian exponent @xmath15 and of the kubo number @xmath116 .
, in log - log representation.,height=248 ] a first glance at fig .
[ lfde 7 ] shows that , even for relatively small values of @xmath116 , the lagrangian correlation deviates very quickly from the eulerian one : this is clearly seen in fig .
[ lfde 8 ] .
thus , the nonlinearity , measured by @xmath318 is far from producing just a small perturbation . , ( solid red )
compared to the eulerian correlation @xmath319 ( dotted blue ) at short times .
@xmath320,height=266 ] the first iteration is not quantitatively accurate ( fig .
[ lfde 7 ] ) , but it has the advantage that it can be evaluated analytically .
it is therefore interesting to study explicitly some of its properties which provide a global qualitative picture .
the explicit form is ( [ eeq.106 ] ) : @xmath321 ^{2}}% .
\label{eeq.126}\ ] ] the integral in the denominator , combined with the definition ( [ eeq.109 ] ) , can be written in terms of two other integrals as follows : @xmath322 where : @xmath323 @xmath324 combining these partial results , we finally obtain : @xmath325 @xmath326 ^{2}},\ \ \
\theta > k. \notag \\ & & \label{eeq.131}\end{aligned}\ ] ] the behavior of this function for _ short times _ is very simple : starting from @xmath269 , it decreases parabolically : @xmath327 as @xmath273 reaches @xmath116 , we find , of course , that the two branches ( eeq.130 ) and ( [ eeq.131 ] ) join each other at the same point .
but the derivative at @xmath328 is discontinuous .
indeed , consider the two integrals ( [ eeq.128 ] ) and ( [ eeq.129 ] ) ; one easily finds : @xmath329 thus , @xmath330 exhibits a kink at @xmath328 , as can be seen in fig .
[ lfde 9 ] . the _ asymptotic behavior for large _
@xmath273 is more interesting.here we must distinguish two ranges of the eulerian exponent @xmath15 .
\a ) @xmath331 . in this case
the term proportional to @xmath332 dominates all others in the denominator of eq .
( [ eeq.131 ] ) for large @xmath273 .
we thus find : @xmath333 we thus find , indeed , for very small @xmath15 , an algebraic tail of the form ( [ eeq.110 ] ) : @xmath334 with : @xmath335 ^{2}k^{-\gamma } .
\label{eeq.134}\ ] ] next , we note that the lagrangian exponent in this approximation is : @xmath336 although this value is not accurate ( as can be seen from fig .
[ lfde 8 ] ) , it has the important property ( which will be confirmed ) that _ it only depends on _ @xmath15 , _ not on _ @xmath116 ( for very small @xmath337 .
\b ) @xmath338 . in this case
the dominant term at large @xmath273 in the denominator of eq .
( [ eeq.131 ] ) is the term proportional to @xmath339 .
the asymptotic behavior is thus different : @xmath340 hence : @xmath341 where @xmath342 is found from this equation .
the important point is that the lagrangian exponent takes a different dependence on @xmath15 for larger values of this exponent : @xmath343 thus , @xmath344 , which was a decreasing function , starts increasing with @xmath15 beyond @xmath345 .
we may calculate analytically the exact dependence of @xmath344 by evaluating the slope of the log - log graph of @xmath330 ( see fig .
[ lfde 8 ] ) : @xmath346-\log [ % \mathcal{l}^{(1)}(1000,k,\gamma ) ] \label{eeq.138}\ ] ] the result is shown in fig .
[ lfde 10 ] .
comparing this type of figures for different values of @xmath116 , one finds a very weak dependence of @xmath279 on @xmath116 for @xmath15 close to @xmath269 .
the left ( right ) dotted straight lines correspond to eqs .
( eeq.135 ) [ resp .
( [ eeq.137])].,height=254 ] the important consequence of this new `` bifurcation '' is the fact that , for all values of @xmath15 we find : @xmath348 the first iteration thus predicts , in accordance with the discussion at the end of sec .
[ algeb gen ] , that the first iterate of the lagrangian correlation leads to a * diffusive regime * _ for all values of _ @xmath349 .
more than that : the lagrangian exponent is in the range d ) , where there is no equivalence between kinetic theory and ctrw or fde .
this statement , however , has to be confirmed by the final solution .
as shown in the previous section , the second iterate of eq .
( [ eeq.106 ] ) yields already an excellent approximation , while not requiring long numerical calculations ( we therefore omit the superscript @xmath350.the short - time ( @xmath351 ) part can even be calculated analytically .
we thus have : @xmath352 @xmath353 ^{2}},\ \ \ \
\theta > k. \label{eeq.141}\ ] ] eq .
( [ eeq.131 ] ) is substituted in the integral , and the expression is evaluated numerically . for short times , the lagrangian correlation behaves as : @xmath354 _ i.e. _ , exactly as in the first approximation ( [ eeq.132 ] ) . as can be seen in fig .
[ lfde 11 ] , the kink which appeared at @xmath328 in the first approximation is still present in the final expression . , ( solid red )
compared to the first iterate @xmath355 ( dashed green ) and to the eulerian correlation @xmath356 ( dotted blue ) at short times .
@xmath357,height=261 ] the long - time behavior is , of course , more interesting . the first important point , that is seen in fig .
[ lfde 8 ] , is that the lagrangian velocity correlation behaves asymptotically as a _ power law , _ ( [ eeq.110 ] ) , thus confirming our previous anticipation .
we now study its parameters .
the _ lagrangian exponent _
@xmath279 is in all cases smaller than the one given by the first iterate , but much larger than the eulerian exponent @xmath349 , as seen in fig .
[ lfde 8 ] . in fig .
[ lfde 12 ] we plotted @xmath358 vs @xmath359 for given @xmath360 and three values of @xmath361 .
for all these relatively small values of @xmath362 the exponent @xmath279 is strictly the same : @xmath363 @xmath364 .
, in log - log representation .
the curves correspond to @xmath365 ( solid red ) , @xmath347 ( dotted blue ) and @xmath366 ( dashed green ) .
@xmath315,height=243 ] next , we measured the dependence of @xmath279 on the eulerian exponent.the result is a very simple linear dependence , shown in fig .
[ lfde 13 ] .
this dependence is very accurately fitted by the formula : @xmath367 _ vs_. eulerian exponent @xmath368 .
the curves for @xmath369 are all superposed.,height=231 ] we note already here the following very important point : _ the lagrangian exponent _ @xmath370 . hence _ the process described by the v - langevin equation ( [ eeq.93 ] ) , with the eulerian velocity correlation ( [ eeq.110 ] ) in the corrsin approximation , leads to a _ * * diffusive regime * * _ _ for all values of _ _ @xmath371 _ _ and of _ _ @xmath116 .
this point will be further discussed below .
we now consider the numerator @xmath316 of eq .
( [ eeq.110 ] ) , determined from the logarithmic representation of @xmath276 ( fig .
[ lfde 12 ] ) .
we first note ( fig .
[ lfde 14 ] ) that , for given @xmath116 , it is a very slowly varying function of @xmath15 : we therefore consider it approximately independent of @xmath15 .
_ vs_. the eulerian exponent @xmath368 , for three values of @xmath116 : @xmath365 ( solid red ) , @xmath372 ( dotted blue ) , @xmath347 ( dashed green).,height=208 ] the dependence on @xmath116 appears to be quadratic , for relatively small values of @xmath116 .
it is quite well fitted by the following empiric formula , valid for @xmath373 ( fig .
[ lfde 15 ] ) : @xmath374 as a function of the kubo number @xmath116 .
red dashed : @xmath375 , blue dotted : @xmath376 .
the solid black curve is the empirical fit ( [ eeq.146]).,height=234 ] finally , the value of the cut - off time is obtained by setting @xmath377 .
@xmath378 to sum up , we found a reasonable analytical representation for the lagrangian velocity correlation in the corrsin approximation . a simple feature of this representation is the fact that ( see fig . 13 and fig .
14 ) each of the two output parameters , @xmath379 is a function of a single input parameter @xmath15 or @xmath116 : @xmath380 combined with eqs .
( [ eeq.144 ] ) and ( [ eeq.145 ] ) .
this analytic approximation is compared to the `` exact '' ( numerical ) lagrangian velocity correlation in fig .
[ lfde 16 ] . )
( dashed black).,height=282 ]
the running diffusion coefficient for an input algebraic eulerian correlation is obtained from eq .
( [ eeq.108 ] ) . using the algebraic form ( [ eeq.111 ] ) this integral
was calculated analytically in eqs .
( eeq.110 ) and ( [ eeq.114 ] ) . using also the forms ( [ eeq.143 ] ) - ( eeq.145 )
we obtain the result plotted in fig .
[ lfde 17 ] for fixed @xmath381 and three values of @xmath382 .
as we know that the regime is always diffusive , the running diffusion coefficient tends toward a finite positive diffusion constant for sufficiently large times ( in practice , @xmath383 ) .
one sees that the asymptotic diffusion coefficient is a weakly decreasing function of the eulerian exponent . for three values of @xmath375 ( solid red ) , @xmath384 (
dotted blue ) , @xmath385 ( dashed black ) .
@xmath386,height=253 ] fig .
[ lfde 18 ] shows the dependence of the diffusion coefficient on the kubo number for fixed @xmath360 . the asymptotic diffusion coefficient is a rather strongly increasing function of @xmath116 . for three values of @xmath365 ( solid red ) , @xmath387 ( dotted blue ) , @xmath347 ( dashed black ) .
@xmath315,height=239 ] in most existing works on the langevin equation the eulerian velocity time - correlation is assumed to decay exponentially ( @xmath52 ) .
this produces a lagrangian correlation that also has an exponential asymptotic behavior , and thus automatically ensures the convergence of the integral ( [ eeq.108 ] ) defining the diffusion constant .
it is interesting to compare the diffusion coefficients obtained in the algebraic case and in the exponential case . for the latter we use a eulerian correlation that is tailored in such a way as to resemble the algebraic one
: we thus start the exponential decay at the cut - off time of the algebraic one ( fig .
[ lfde 19 ] ) : @xmath388 , \ \ \ \ \ \theta > k% \end{array}% \right . . \label{eeq.147}\ ] ] ) ; dotted blue : algebraic , eq . ( [ eeq.109 ] ) .
@xmath389,height=246 ] the lagrangian velocity correlation is calculated in the corrsin approximation up to the second iteration , by following exactly the same steps as in secs .
[ determ lagr 1 ] and [ determ lagran 2 ] . the result is shown in fig .
[ lfde 20 ] .
( note that @xmath390 , hence it is not necessary to introduce a cut - off time @xmath317 as for the algebraic case ) .
as expected , @xmath391 decays much more quickly : it has no long - time tail as the algebraic one . ] and of an exponential ( dotted blue ) eulerian correlation with the same @xmath116.,height=241 ]
the running diffusion coefficients corresponding to the same parameters are compared in fig .
[ lfde 21 ] .
the algebraic correlation produces a running diffusion coefficient that saturates at a longer time . as a result , the asymptotic diffusion constant @xmath392 this results from the fact that the area under the lagrangian correlation is larger in the algebraic case ( fig .
[ lfde 20 ] ) , due to the slow decay of the correlation . ] and of an exponential ( dotted blue ) eulerian correlation with the same @xmath116.,height=241 ]
our main purpose in the present work was to study the relation between a `` semi - dynamical '' description of a system of particles , based on the v - langevin equation and its associated hybrid kinetic equation , and on the other hand on a purely stochastic description in terms of a continuous time random walk and its limiting form as a fractional diffusion equation .
we wanted , in particular , to find out under what conditions on the velocity correlation function a v - langevin equation describes a `` strange transport regime '' , which could be equivalently represented by a ctrw or a fde .
it was known from previous work @xcite that a _ subdiffusive regime _ can appear in some limiting cases .
this occurs , _
e.g. _ , for a collisional plasma in a strong magnetic field , in the limit of infinite correlation length perpendicular to @xmath393 @xcite , @xcite or for electrostatic drift wave turbulence in the limit of infinite kubo number @xcite . in these cases ,
subdiffusion is produced by sticking of the particles to the magnetic field lines , or by their trapping in the turbulent field .
it might be stressed that in the latter case , a diffusive or a subdiffusive ( @xmath394 regime was found even when the spatial correlation is long - range ( lorentzian ) , but the time correlation is exponential @xcite .
a _ superdiffusive regime _ is , however , much more elusive . to the best of our knowledge
, a v - langevin equation producing lvy - type superdiffusion was only found for a one - dimensional two - state model ( velocity = @xmath17 ) @xcite .
the restriction to a velocity constant in absolute value was necessary in order to replace the lvy flight ctrw by a lvy walk . in the present work we showed that , for one - dimensional systems , the results of @xcite concerning the _ time - fractional superdiffusion _ can be extended to the more physical case of a general random time - dependent velocity function equipped with an algebraic time - correlation function ( but with a gaussian space - correlation ) .
we confirm that in this case we find a process that is an `` intermediate '' between diffusion and wave propagation.this type of superdiffusive processes , distinct from the lvy - type , has been studied in detail in @xcite and @xcite .
we then generalized our investigation to the study of a random velocity _
field _ in two dimensions . in this case one
is faced with the difficult problem of determining the _ lagrangian velocity correlation _ , a problem that was treated within the corrsin approximation .
it was first shown that if the eulerian correlation decays algebraically as @xmath395 for long times ( and its spatial part is gaussian ) , the lagrangian correlation has the same property , with a different exponent : @xmath396 .
we then established a general criterion for the existence of a superdiffusive regime , based on the value of the lagrangian exponent @xmath397 .
it was shown in sec .
[ algeb gen ] that whenever @xmath398 , the regime is superdiffusive and can be equivalently described by a time - fractional fde .
when @xmath290 , however , the semi - dynamical description predicts a diffusive regime for all @xmath279 , whereas the formally associated fde predicts subdiffusion , or even looses meaning , for @xmath399 .
thus , the purely stochastic ctrw or fde descriptions of the turbulence are severely limited in this case by the existence of a bifurcation at @xmath292 . in the final stage ( sec .
[ determ lagran 2 ] ) the lagrangian exponent @xmath397 was determined numerically .
the result shows that @xmath400 , for all values of @xmath15 and of @xmath116 .
thus , for the present model and with the approximations used in its treatment : [ local approximation of the non - markovian diffusion equation ( [ eeq.55 ] ) , corrsin approximation of the lagrangian velocity correlation ( [ eeq.101 ] ) ] : * _ the regime is always _ * * normal diffusive , * * _ _ for all values of the input eulerian parameters _
_ @xmath15 and @xmath116 .
there never appears any superdiffusion , in spite of the long algebraic tail of the temporal velocity correlation .
this is to be contrasted with the one - dimensional case of a purely time - dependent velocity , where the regime is superdiffusive for all values @xmath401 . * _ the `` formally associated '' fractional differential equation , has no meaning . _
the evolution can not be described by a purely stochastic process . * _ although the regime is diffusive , _ _ the density profile is not a gaussian packet .
_ indeed , although the lagrangian correlation decays sufficiently fast in order to ensure the existence of a finite asymptotic diffusion constant , the non - markovian character of the equation ( [ eeq.107 ] ) can not be neglected for finite time .
the results obtained here may appear somewhat disappointing , in the sense that no time - fractional superdiffusion was found for two - dimensional motion .
nevertheless , we believe that the result of the exhaustive analysis performed here is interesting in exhibiting the limits of a purely stochastic description .
it shows that a long - tailed algebraic temporal correlation is not sufficient for producing superdiffusion ( for dimensionality @xmath402 ) . in order to obtain a lvy - type superdiffusive regime , we must presumably act on the spatial part of the eulerian velocity correlation .
this problem is , however , highly non - trivial . on one hand , in the corrsin approximation ,
( [ eeq.106 ] ) is invalid , because it is based on the gaussian form of the spatial eulerian correlation .
but one should go back even further in the chain of approximations . the local approximation leading from eq .
( [ eeq.55 ] ) to ( [ eeq.56 ] ) ( in two dimensions ) is questionable in the presence of long - range spatial correlations .
we hope to come back to this problem in forthcoming work .
fractional calculus has been known to mathematicians for a very long time.the problem appears to have been first raised by lhospital and by leibniz in 1695 .
the first systematic construction appears in the early 19-th century and is due to liouville ( 1832 ) ( well known to every physicist from the first lecture in statistical mechanics ) , who used this concept in potential theory .
it was further developed by b. riemann in 1847 . for a long time
the concept was studied only by mathematicians .
it entered physics only in the last half of the 20-th century , through the theory of stochastic equations associated with the problems of random walks ( sec .
ctrw fde ) .
the search for a natural generalization of the ordinary derivative @xmath403 , with @xmath134 : an integer , is clearly attractive for a mathematician .
it is as natural as the passage from the integers to rational , and later to real numbers .
we must find a way to interpolate in a consistent way the concept of differentiation between two successive integers , @xmath134 and @xmath404 .
the way of doing this is however not obvious.indeed , several different approaches have been devised by mathematicians.we shall only discuss here the one which is most widely used in the literature ( for a discussion of other approaches , see @xcite , butzer and @xcite )
. one will find detailed expositions of the subject in @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite.excellent tutorials are the papers by @xcite and especially main 2001 , @xcite .
a brief , but very clear summary is given in @xcite .
the names of _ fractional _ integral , _ fractional _ derivative , _ fractional _ calculus are actually rather improper .
the object of these concepts is to extend the notions of integration and of differentiation from integer order @xmath134 to _ arbitrary real order _
@xmath405 , not necessarily a fraction .
the name has , however , acquired general practice .
* riemann - liouville fractional integral * consider first the _ multiple integral of order _
@xmath134 _ _ ( _ _ @xmath134 _ _ : integer ) for integer indices , and greek letters for real , non - integer indices : @xmath406 _ _ of a real - valued function @xmath407 of the real variable @xmath240 : @xmath408 the two lower subscripts relate to the limits of integration ( @xmath409 is a fixed real number ) ; the superscript @xmath134 denotes the order ( multiplicity ) .
this equation can be written in the following equivalent form , in terms of a _ single _ integral : @xmath410 the equivalence between the two forms is easily demonstrated by partial integration .
we may put @xmath411 @xcite . for @xmath412 ,
the formula reduces to the single integration : @xmath413 let us check @xmath414 : @xmath415 in the form ( [ eeq.149 ] ) the way toward a generalization of the operation of integration ( [ eeq.149 ] ) is obvious : @xmath416 this is the definition of the * riemann - liouville ( rl ) fractional integral of * @xmath407 * of arbitrary real order * @xmath8 . some of the more important properties are given here : \a ) _ rl fractional integral of a monomial _
@xmath417 \b ) _ commutative composition rule : _
@xmath418 proof : _ _ _ _
@xmath419 substitute : @xmath420 @xmath421^{-1+\nu } ~[(1+\zeta ) ( x-\tau ) ] ^{-1+\mu } \\ & = & \frac{1}{\gamma ( \mu ) \gamma ( \nu ) } \int_{0}^{x}d\tau ~f(\tau ) ~(x-\tau ) ^{-1+\mu + \nu } ~\int_{0}^{1}d\zeta ~\zeta ^{-1+\nu } ~(1+\zeta ) ^{-1+\mu } \end{aligned}\ ] ] the @xmath422-integral is well - known is expressed in terms of the well - known @xmath423-function : @xmath424 hence : @xmath425 * riemann - liouville fractional derivative * the fractional derivative is constructed by combining the operations of the usual ( integer - order ) derivative with the fractional integral .
we define the _ fractional derivative _ as follows : @xmath426 , \ \ \
\alpha > 0,\ \ \ \
m-1<\alpha \leq m. \label{eeq.154}\ ] ] where @xmath8 is any positive real number , not an integer , and @xmath427 is the smallest integer larger than @xmath8 . in order to `` understand intuitively '' this rather unusual definition , we may think , very casually , and without any rigour , as follows .
we have a solid reference point , _
viz_. a rigorously defined fractional object , the rl integral .
we then start with a rl integral of order @xmath428 , where @xmath429 is the smallest integer which makes @xmath430 , thus ensuring that @xmath431 is a true fractional integral ( [ eeq.150 ] ) .
next , we `` differentiate away '' the integer @xmath427 , by taking the @xmath427-th ordinary derivative of this object .
this looks like `` lowering the order '' of @xmath432 thus `` transforming '' it into @xmath433 , which could be viewed as a fractional integral of negative order , @xmath434 , to be interpreted as a fractional derivative of order @xmath435 .
but , again , this paragraph is just cavalier talk ; in particular , viewing @xmath436 as an `` inverse '' to @xmath437 is an ambiguous statement , as will be shown below .
we now go back to business , and consider eq .
( eeq.154 ) as a given definition , and study some of its consequences .
a basic requirement is that the definition ( [ eeq.154 ] ) be compatible with the ordinary ( integer - order ) derivative .
when @xmath8 is an integer , @xmath438 , we have @xmath439 thus @xmath440 , and recalling eq .
( eeq.149a ) , we have : @xmath441 = % \frac{d^{n}}{dx^{n}}~\frac{d}{dx}~\int_{a}^{x}dy~f(y)=\frac{d^{n}}{dx^{n}}% ~f(x);\ ] ] the operation thus reduces to the ordinary differentiation , as it should . combining eqs .
( [ eeq.150 ] ) and ( [ eeq.154 ] ) we find : @xmath442 this is the definition of the * riemann - liouville ( rl ) left - fractional derivative of order * @xmath8 .
we may also define a * riemann - liouville ( rl ) right - fractional derivative * in which it is the upper integration limit @xmath443 that is fixed : @xmath444 we now explore a few properties of the fractional derivatives : we shall meet some surprises
. these are all due to a fundamental fact . _ the fractional derivative ( for any of the forms defined above ) of a function _
@xmath445 _ is expressed as an _ * integral * _ of _ @xmath407 _ _ over a given range of _ _
@xmath240 it is therefore an intrinsically * non - local operator*. in other words , _ the value of the fractional derivative ( for non - integer _
@xmath8 ) @xmath446 _ at the point _ @xmath240
_ depends on the values of _ @xmath221 _ _ taken in the whole range of integration defining this quantity .
_ _ it is quite surprising that , upon varying continuously the order @xmath8 , this non - local character disappears _ suddenly _ whenever @xmath8 reaches any integer value !
what is not surprising , therefore , is that fractional calculus entered physics when researchers began to be interested in evolutionary processes endowed with memory ( `` non - markovian processes '' ) or in the influence of the spatial neighborhood in the phenomena at a given point @xmath447 .
\a ) _ linearity : _ @xmath448=\lambda ~_{0}d_{x}^{\alpha } ~f(x)+\mu ~_{0}d_{x}^{\alpha } g(x ) \label{eeq.156a}\ ] ] the proof follows trivially from the definition of the rlderivative .
\b ) _ rl fractional derivative of a monomial : _
@xmath449 _ _ proof .
_ _ using the definition ( [ eeq.155 ] ) : @xmath450 let us do this calculation in some detail . substituting @xmath451
we find @xmath452 the integral appearing here is expressed in terms of the @xmath423-function , ( [ eeq.152a ] ) , hence : @xmath453 next , we calculate ( setting @xmath454 @xmath455 noting the identity : @xmath456 we combine all these partial results to obtain the final formula ( eeq.157 ) .
the limitation to @xmath457 in eq . ( [ eeq.157 ] ) is due to the factor @xmath458 which has a pole in @xmath459 .
note that for any non - integer value of @xmath8 the rl derivative @xmath460 has a single zero ( for any @xmath461 in the allowed region , namely for @xmath462 ( see below ) .
indeed , any zero of the right hand side of eq .
( [ eeq.157 ] ) originates from the factor @xmath463 which vanishes whenever its argument is zero or a negative integer , see fig .
lfde 22a ; a single one of them is larger than @xmath464 .
@xmath465 of the rl derivative of @xmath466 as a function of @xmath467,height=292 ] the formula ( [ eeq.157 ] ) is deceptively simple , and would correspond with intuition .
just as the ordinary integral - order derivative ( @xmath468 lowers the power of @xmath240 by @xmath134 , the fractional derivative transforms @xmath469 into @xmath470 , with a prefactor .
but a further reflection shows us that this result is not as innocent as it appears .
b1 ) _ rl fractional derivative of a constant : _ consider , indeed , the case @xmath471 , i.e. , the @xmath472 i.e. , @xmath473 thus , the _ rl fractional derivative of non - integer order _ @xmath474 _ _ of a constant is non - zero ! _ _ is this result consistent with the ordinary derivative ?
the answer comes from the properties of the gamma function . indeed , whenever @xmath475 , where @xmath134 is a non - zero integer , @xmath476 the argument of the @xmath477 function is a _ negative integer , or zero_. but we know that @xmath478 hence , for any integer value of @xmath8 , eq .
( [ eeq.158 ] ) reduces , as it should , to : @xmath479 another feature of eq .
( [ eeq.157 ] ) seems to contradict our previous statement : @xmath480 is a perfectly _ local _ expression , depending only on the value of @xmath469 at @xmath240 . in order to understand what happens here
, we consider the rl derivative of @xmath481 for a lower integration limit @xmath482 instead of zero . starting the calculation as above
, we find after the first step : @xmath483 there are two important differences with eq .
( [ eeq.157a ] ) .
first , the integral depends on @xmath240 , through its lower limit ; hence , the @xmath427-th derivative acts on the whole product , thus yielding a much more complicated @xmath484-dependence .
next , the integral with a non - zero lower limit is no longer a @xmath423-function , but rather a very complicated hypergeometric function . in order to illustrate our case , we consider the simplest non - trivial case that leads to simple calculations : @xmath485 , @xmath486 , hence @xmath487 .
( [ eeq.160 ] ) then reduces to : @xmath488 the integral is tabulated ; a simple calculation then leads to the result : @xmath489 , \ \ \ x > a .
\label{eeq.162}\ ] ] the result reduces to eq .
( [ eeq.157 ] ) when @xmath490 : @xmath491 the non - local character of the fractional derivative is now luminously illustrated . the complicated function appearing in the right hand side of eq .
( [ eeq.162 ] ) is determined by the choice of the function @xmath492 in eq .
( [ eeq.160 ] ) , hence by its values taken over the whole integration range .
in particular , the fractional derivative depends explicitly on the lower integration limit @xmath409 .
a striking fact is the _ singularity of the rl fractional derivative _ when @xmath493 .
this singularity is present for any @xmath409 , arbitrarily close to zero , but disappears when @xmath494 .
when @xmath495 , @xmath496 .
all this clearly shows that the value @xmath494 is not a generic one .
the situation is illustrated in fig .
[ lfde 22 ] .
: @xmath497 , for @xmath498.,height=324 ] c)_inversion of the rl fractional derivative .
_ many authors state rather loosely that `` the rl fractional derivative is the inverse of the rl fractional integral '' .
they even adopt the notation : @xmath499 @xcite .
it will be shown now that this statement is , for the least , ambiguous .
hence podlubny s notation should preferably be avoided in order to avoid misunderstandings ( mainardi ) .
c1 ) _ rl fractional derivative of a fractional integral of same order . _
@xmath500=f(x).% \end{array } \label{eeq.164}\ ] ] _ proof _ : for integer @xmath501 : @xmath502 & = & \frac{d^{n}}{dx^{n}}~\frac{1}{(n-1)!}% ~\int_{0}^{x}dy~(x - y)^{n-1}f(y ) \\ & = & \frac{d}{dx}~\int_{0}^{x}dy~f(y)=f(x)\end{aligned}\ ] ] for non - integer @xmath8 , we enclose this number again between its nearest neighbor integers , defining the integer @xmath503 as : @xmath504 using now the composition rule ( [ eeq.152 ] ) for the fractional integrals , we note the identity : @xmath505=~_{0}j_{x}^{k}~f(x)\ ] ] using now the definition ( [ eeq.154 ] ) we have : @xmath506=\frac{d^{k}}{dx^{k}}~\left\ { _ { 0}j_{x}^{k-\alpha } ~[_{0}j_{x}^{\alpha } ~f(x)]\right\ } = \frac{d^{k}}{dx^{k}% } ~_{0}j_{x}^{k}~f(x)=f(x).\ ] ] ( the result follows from the fact that the last step only involves operations of integer order @xmath503 , i.e. , ordinary derivative and integral . ) from the result ( [ eeq.164 ] ) one is tempted to conclude that , indeed , `` the rl fractional derivative is the inverse of the rl fractional integral '' : the fractional rl derivative annuls the action of the fractional integral .
the situation is , however , not so clear - cut .
it appears from the following property that the double operation @xmath507 is _ not commutative _ !
indeed : c2 ) _ fractional integral of a rl fractional derivative of the same order .
_ @xmath508=f(x)-\sum\limits_{j=1}^{k}~ \left [ _ { 0}d_{x}^{\alpha -j}~f(x)\right ] _ { x=0}~\dfrac{x^{\alpha -j}}{\gamma ( \alpha -j+1)},\ \ \ \ k-1\leq \alpha < k.% \end{array } \label{eeq.165}\ ] ] _ proof .
_ @xmath509 = ~\frac{1}{\gamma ( \alpha ) } ~\int_{0}^{x}dy~(x - y)^{\alpha -1}~_{0}d_{y}^{\alpha } ~f(y ) \\ = \frac{1}{\gamma ( \alpha ) } ~\int_{0}^{x}dy~\frac{1}{\alpha } ~\left [ \frac{d}{% dx}(x - y)^{\alpha } \right ] ~_{0}d_{y}^{\alpha } ~f(y ) \\ = \frac{d}{dx}~\left\ { \frac{1}{\gamma ( \alpha + 1)}~\int_{0}^{x}dy~(x - y)^{% \alpha } ~_{0}d_{y}^{\alpha } ~f(y)\right\ } \\ = \frac{d}{dx}~\left\ { \frac{1}{\gamma ( \alpha + 1)}~\int_{0}^{x}dy~(x - y)^{% \alpha } ~\frac{d^{k}}{dy^{k}}~_{0}j_{y}^{k-\alpha } ~f(y)\right\ } .~\end{gathered}\ ] ] a partial integration yields : @xmath510 _ { x=0}\right\ } .~\end{aligned}\ ] ] continuing the partial integrations , and using eqs .
( eeq.150 ) and ( [ eeq.152 ] ) , we end up with : @xmath511 _ { x=0}~\frac{x^{\alpha -j+1}}{\gamma ( \alpha -j+2)}\right\ } \\
= \frac{d}{dx}\left\ { ~_{0}j_{x}^{\alpha -k+1}\cdot ~_{0}j_{x}^{k-\alpha } f(x)\right .
\\ \left . -\sum\limits_{j=1}^{k}\left
[ \frac{d^{k - j}}{dx^{k - j}}% ~_{0}j_{x}^{k-\alpha } f(x)\right ] _ { x=0}~\frac{x^{\alpha -j+1}}{\gamma ( \alpha -j+2)}\right\ } \\
= ~\frac{d}{dx}\left\ { _ { 0}j_{x}^{1}f(x)-\sum\limits_{j=1}^{k}\left [ \frac{% d^{k - j}}{dx^{k - j}}~_{0}j_{x}^{k-\alpha } f(x)\right ] _ { x=0}~\frac{x^{\alpha -j+1}}{\gamma ( \alpha -j+2)}\right\ }
\\ = f(x)-\sum\limits_{j=1}^{k}\left [ \frac{d^{k - j}}{dx^{k - j}}% ~_{0}j_{x}^{k-\alpha } f(x)\right ] _ { x=0}~\frac{x^{\alpha -j}}{\gamma ( \alpha -j+1)},\ \ \ \ qed\end{gathered}\ ] ]
it is precisely because of this non - commutativity of integration and differentiation that one can not state unambiguously that one operation is the inverse of the other .
\d ) _ general composition rules . _ in all subsequent formulae we take the lower limit @xmath494 .
we simplify the notations as follows : @xmath512 @xmath513 the proofs of the following results are very similar to the previous ones @xcite .
d1 ) _ composition of two fractional integrals .
_ [ see ( [ eeq.152 ] ) ] @xmath514 d2 ) _ rl derivative of fractional integral .
_ @xmath515 d3 ) _ fractional integral of rl derivative .
_ @xmath516 _ { x=0}~\dfrac{x^{\alpha -j}}{\gamma ( \alpha -j+1)},\ \\ 0<\alpha < \beta , \ \ \
\ n-1\leq \beta < n% \end{array } \label{eeq.168}\ ] ] @xmath517 _ { x=0}~\dfrac{x^{\alpha -j}}{\gamma ( \alpha -j+1)},\ \\
0<\beta < \alpha , \ \ \
n-1\leq \beta < n% \end{array } \label{eeq.168a}\ ] ] d4 ) _ composition of rl derivatives
. _ @xmath518 _ { x=0}~% \dfrac{x^{-\alpha -j}}{\gamma ( -\alpha -j+1 ) } , \\ _ { 0}d_{x}^{\beta } \cdot ~_{0}d_{x}^{\alpha } ~f(x)=~_{0}d_{x}^{\alpha + \beta } ~f(x)-\sum\limits_{j=1}^{m}\left [ ~_{0}d_{x}^{\alpha -j}f(x)\right ] _ { x=0}~% \dfrac{x^{-\beta -j}}{\gamma ( -\beta -j+1 ) } \\ \ m-1\leq \alpha < m,\ \ \ \ n-1\leq \beta < n% \end{array}% \ \
\ \end{array } \label{eeq.169}\ ] ] * integral transforms of fractional operators .
* * a ) fourier transform * the fourier transform involves integrals ranging from @xmath519 to @xmath520 .
this implies a number of technical assumptions about the regularity of the functions appearing in the theory : we do not discuss them here , assuming that all conditions are satisfied for the existence of the objects of the forthcoming equations .
we begin with the following result : @xmath521 _ proof : _ @xmath522 where we set @xmath523 .
we now note that , from the definition of @xmath427 , we have : @xmath524 .
we may thus use the tabulated integral : @xmath525 hence : @xmath526 this expression is not very convenient in applications .
on one hand , the appearance of @xmath527 is somewhat awkward .
but more important , if we wish to use this result in relation with the fourier transform , we need to cover the whole range ( @xmath528 , and not only the left range ( @xmath529 ) .
we therefore consider also the right rl fractional derivative ( [ eeq.156 ] ) : @xmath530 _ proof : _
@xmath531 setting @xmath532 , we have : @xmath533 thus : @xmath534 in order to adapt the previous results to the fourier transform , in particular to cover the whole range @xmath535 , we define a new type of fractional derivative , which is totally symmeric .
it is justified by the linearity property of the rl derivative ( [ eeq.156a ] ) : @xmath536 ~f(x)% \end{array } \label{eeq.172}\ ] ] this is called the * riesz fractional derivative .
* the usefulness of this concept appears obviously in the following result , expressing the _ fourier transform of the riesz derivative _ : @xmath537 where @xmath538 denotes the fourier transform of @xmath407 .
_ proof : _
@xmath539 the result ( [ eeq.173 ] ) ends our quest of the fourier transform of the ( riesz ) fractional derivative of a function , expressed in terms of the fourier transform of that function . *
b ) laplace transform * the ( direct ) laplace transform of a function involves an integral over the variable from @xmath90 to @xmath540 it is therefore particularly well adapted to the study of functions of _ time _
, when the latter is restricted to positive values .
the two properties of the laplace transform that are most useful in applications ( especially for the solution of differential equations are : _ the convolution theorem : _ for any two functions _ _ _ _ @xmath541 , @xmath542 of time , defined in the range ( @xmath543 ) , their convolution is defined as : @xmath544 _ laplace transform of the ( ordinary ) derivative : _ @xmath545 where @xmath546 is the @xmath134-th time derivative of @xmath541 , and @xmath547 is the laplace transform of @xmath541 .
the explicit appearance of the functions @xmath548 in this formula makes the laplace transform a particularly useful tool for the solution of initial value problems .
we may now note that the value @xmath549 in the sum of the right hand side is `` special '' : it corresponds to @xmath550 , whereas all other terms in the sum contain true derivatives : @xmath551 we now consider the laplace transforms of the various objects of fractional calculus . _
a ) laplace transform of the fractional integral .
_ this is easily obtained by noting that the rl integral can be viewed as a convolution : @xmath552 hence , noting the well - known result : @xmath553 we find immediately , using the convolution theorem ( [ eeq.174 ] ) : @xmath554 _ b ) laplace transform of the rl derivative . _ we write the rl derivative ( [ eeq.154 ] ) in the form : @xmath555 with : @xmath556 using eq .
( [ eeq.175 ] ) for the @xmath427-th ( ordinary ) derivative , we have : @xmath557 next , we calculate : @xmath558 using eq .
( [ eeq.167 ] ) , we find : @xmath559 where : @xmath560 from the two constraints : @xmath561 and @xmath562 , we deduce : * when @xmath563 , we have @xmath564 , hence : @xmath565 , * when @xmath566 , we have @xmath567 , hence : @xmath568 note also , from eq .
( [ eeq.178 ] ) : @xmath569 we thus obtain the final form of the laplace transform of the rl fractional derivative : @xmath570 _ { t=0}-\sum% \limits_{j=0}^{m-2}s^{j}~\left [ ~_{0}d_{t}^{\beta -j-1}~f(t)\right ] _ { t=0}.% \end{array } \label{eeq.185}\ ] ] the analogy with eq .
( [ eeq.175a ] ) is striking .
the first term has the same form . in the second term
the initial value of the function @xmath571 is replaced , for non - integer @xmath572 by the initial value of the fractional _ integral _
@xmath573 , whereas in the terms under the sum , the initial values of the ordinary derivatives are replaced by the initial values of the corresponding rl fractional derivatives . in spite of this analogy , the rl time - derivative is not easily adapted to the physical applications .
the laplace transform is well known to be a valuable tool for the solution of _ initial value problems of differential equations _ , because it incorporates these initial values in eq .
( [ eeq.175a ] ) . if we now consider _ initial value problems of fractional differential equations involving * rl * fractional derivatives with respect to time _ , we are faced with a serious difficulty if we try to use a laplace transformation .
the specification of ( physical ) values of the unknown function and of a sufficient number of its derivatives at @xmath38 , this does not allow us to determine the correponding initial valued of the fractional integrals and derivatives appearing in eq . ( [ eeq.185]_)_. this difficulty has been overcome by m. caputo in @xcite , @xcite by introducing an alternative definition of the fractional derivative , that is particularly useful for functions of a variable restricted to a finite range , such as time @xmath574 . * the caputo fractional derivative . *
the caputo fractional derivative is defined as follows : @xmath575 thus , the only difference with the rl derivative eq.(156 ) is the commutation of the integration and the differentiation . the caputo derivative results from a composition of a fractional integral with an ordinary derivative : @xmath576 \label{eeq.186a}\ ] ]
let us introduce a simplified notation : @xmath577 the caputo derivative is related to the riemann - liouville one as follows @xcite : @xmath578 we give here the proof for the case @xmath160 , which implies @xmath579 .
the proof of the general case is similar , but longer .
@xmath580 \right\vert _
{ \tau = t}+\int_{0}^{t}d\tau ~\left [ \frac{d}{dt}~(t-\tau ) ^{-\beta } \right ] ~f(\tau ) \\ & = & \left . \left
[ \frac{f(\tau ) } { ( t-\tau ) } \right ] \right\vert _ { \tau = t}-\left . \left
[ \frac{f(\tau ) } { ( t-\tau ) } \right ] \right\vert _ { \tau = 0}^{\tau = t}+\int_{0}^{t}d\tau ~\frac{1}{(t-\tau ) ^{\beta } } ~\frac{d}{d\tau } f(\tau ) \\ & = & t^{-\beta } ~f(0)+\int_{0}^{t}d\tau ~\frac{1}{(t-\tau ) ^{\beta } } ~\frac{d}{% d\tau } f(\tau ) ; \end{aligned}\ ] ] hence , finally : @xmath581 which agrees with eq . ( [ eeq.188 ] ) . an important property , which is obvious from eq .
( [ eeq.186 ] ) , and also from ( [ eeq.189 ] ) , combined with ( [ eeq.158 ] ) , is : @xmath582 which makes the caputo derivative closer to the ordinary derivative .
we now consider the composition of the caputo derivative @xmath583 with the fractional integral of the same order @xmath584 .
a first relation is immediately obtained from the combination of eqs .
( [ eeq.164 ] ) and ( [ eeq.188 ] ) : @xmath585 _ { t=0}~\dfrac{t^{j-\beta } } { \gamma ( j-\beta + 1)}.% \end{array } \label{eeq.191}\ ] ] commuting the integral and the derivative and using successively eqs .
( eeq.188 ) , ( [ eeq.165 ] ) and ( [ eeq.151 ] ) we obtain : @xmath586 _ { t=0}~j^{\beta } ~t^{j-\beta } \\ = f(t)-\sum\limits_{k=1}^{n}~\left [ d^{\beta -k}~f(t)\right ] _ { t=0}~\frac{% t^{\beta -k}}{\gamma ( \beta -k+1)}-\sum\limits_{j=0}^{n-1}~\left [ \frac{% d^{j}}{dt^{j}}~f(t)\right ] _ { t=0}~\frac{t^{j}}{\gamma ( j+1)}\end{gathered}\ ] ] and finally ( with @xmath587 ) : @xmath588 _ { t=0}~\dfrac{t^{\beta -j-1}}{\gamma ( \beta -j)}+f^{(j)}(0)~\dfrac{t^{j}}{\gamma ( j+1)}\right\}% \end{array } \label{eeq.192}\ ] ] eqs . ( [ eeq.191 ] ) and ( [ eeq.192 ] )
show that the caputo derivative is * not * an inverse of the fractional integral , either to the left or to the right .
general composition rules of the type of eqs .
( [ eeq.166 ] ) - ( [ eeq.169 ] ) can be derived for the caputo derivative , using the previous results ; they are rather complicated , and will not be written down explicitly .
we now consider the most important property of the caputo derivative .
_ laplace transform of the caputo derivative .
_ the result is easily obtained by combining the definition ( [ eeq.186a ] ) with eqs .
( [ eeq.178 ] ) and ( [ eeq.175a ] ) : @xmath589 & = & \mathcal{l}~\left [ ~_{0}j_{t}^{n-\beta } \left ( \frac{d^{n}f(t)}{dt^{n}}\right ) \right ] = s^{\beta -n}~\mathcal{l}~\left ( \frac{d^{n}f(t)}{dt^{n}}\right ) \\ & = & s^{\beta -n}~\left [ s^{n}~\widetilde{f}(s)-s^{n-1}~f(0)-\sum% \limits_{k=0}^{n-2}~s^{j}~f^{(n - k-1)}(0)\right ] , \end{aligned}\ ] ] and finally , setting @xmath590 : @xmath591 = s^{\beta } ~% \widetilde{f}(s)-s^{\beta -1}~f(0)-\sum\limits_{j=1}^{n-1}~s^{\beta -j-1}~f^{(j)}(0),\ \ \ \ \
n-1<\beta < n.% \end{array } \label{eeq.193}\ ] ] this result has exactly the same form as the laplace transform of an ordinary derivative , eq .
( [ eeq.175a ] ) , with @xmath592 .
contrary to the laplace transform of the rl derivative ( [ eeq.185 ] ) , the laplace transform of the caputo derivative only involves the initial values of the function @xmath541 and of its ordinary derivatives , instead of the initial values of the fractional integral and of the rl fractional derivatives . in any physical differential equation , the former are the given input .
this property makes the caputo derivative an invaluable tool for solving initial value problems of fractional differential equations .
we gratefully acknowledge extremely fruitful discussions with profs .
j.klafter , f. mainardi , r. gorenflo , p. grigolini and r. sanchez . e.w .
montroll and m.f .
shlesinger , on the wonderful world of random walks , in : _ studies of statistical mechanics _
( edited by j.l .
lebowitz and e.w .
montroll ) , vol . * 11 * , p.5 - 121 , north holland , amsterdam , 1984 .
r. metzler and j. klafter , the restaurant at the end of the random walk : recent developments in the description of anomalous transport by fractional dynamics , j. phys .
a : math .
* 37 * , r161-r208 , 2004 .
r. gorenflo and f. mainardi , fractional calculus , integral and differential equations of fractional order , in : _ fractals and fractional calculus in continuum mechanics _
( edited by a.carpinteri and f. mainardi ) , p. 223 - 276 , springer , wien , 1997 f. mainardi , yu . luchko and g. pagnini , the fundamental solution of the space - time fractional diffusion equation , fract .
analysis , * 4 * , 153 - 192 , 2001 .
available from _ http://www.fracalmo.org _
|
the following question is addressed : under what conditions can a strange diffusive process , defined by a semi - dynamical v - langevin equation or its associated hybrid kinetic equation ( hke ) , be described by an equivalent purely stochastic process , defined by a continuous time random walk ( ctrw ) or by a fractional differential equation ( fde ) ? more specifically , does there exist a class of v - langevin equations with long - range ( algebraic ) velocity temporal correlation , that leads to a time - fractional superdiffusive process ?
the answer is always affirmative in one dimension .
it is always negative in two dimensions : any algebraically decaying temporal velocity correlation ( with a gaussian spatial correlation ) produces a normal diffusive process .
general conditions relating the diffusive nature of the process to the temporal exponent of the lagrangian velocity correlation ( in corrsin approximation ) are derived .
|
guided by experiment , e.g. in @xmath1c @xcite and @xmath2o @xcite it is a long standing opinion that the spin - orbit splitting of the @xmath0 single particle levels in @xmath0 hypernuclei should be very small or almost zero .
the most prominent theoretical explanation for this effect has been given through the quark - spin substructure of the @xmath0 leading to an additional @xmath0@xmath3 tensor coupling that almost exactly cancels the more conventional scalar - vector spin - orbit force of the @xmath0 single particle states in finite systems @xcite .
a reanalysis of older @xmath4c@xmath5 @xcite and @xmath6o@xmath5 @xcite proton emitter emulsion data by dalitz _
_ , however , lead already to a remarkably bigger splitting , i.e. @xmath70.8 mev for @xmath1c and [email protected] mev for @xmath2o though the statistics was not so good .
new experiments at kek , measuring high resolution @xmath0 single - particle spectra for medium and heavy hypernuclei , show a spin - orbit splitting for the @xmath0 levels in the range of 12 mev @xcite .
this is quite sizeable but still smaller by a factor of about 2 than the splitting to be expected from the general quenching of @xmath0n interactions . in this letter
we will show that the reduced splitting in medium and heavy nuclei can be attributed to the strong delocalization of the @xmath0 wave functions in the nucleus due to the shallow @xmath0 potential ( sect .
[ sec : deloc ] ) . in light nuclei with
weakly bound @xmath0 states particle threshold effects are superimposed on the general delocalization .
the very small splitting in light nuclei is due to these continuum threshold effects which lead to a squeezing of the single particle levels close to threshold and thereby almost completely dilutes generic spin - orbit potential effects in this mass region ( sect .
[ sec : thresh ] ) . according to our investigations , the extremely small splitting of the spin - orbit doublets in light @xmath0 hypernuclei as well as the reduced splitting of those in medium to heavy hypernuclei can be completely explained in the framework of the density dependent relativistic hadron ( ddrh ) field theory @xcite .
y with ddrh calculations.,width=226 ] .comparison of the measurements by nagae et al .
@xcite on @xmath9v with ddrh calculations .
[ cols= " < , > , > " , ] a suitable measure for the reduced overlap of the @xmath0 wave function with the nuclear interior is to define a localization coefficient by @xmath10 describing the fraction of the probability density which is localized within the volume @xmath11 . here
, @xmath12 is the upper component of the respective dirac wave function and @xmath13 is chosen such that @xmath14 for @xmath15 . in table
[ tab : deloc ] values @xmath16 for @xmath17 for @xmath0 and neutron orbitals in a number of nuclei are shown .
the overlap of the @xmath0 wave functions with the bulk of the nuclear mass distribution is seen to be strongly reduced compared to corresponding neutron states .
the delocalization increases rapidly with decreasing binding energy with a particular strong enhancement for @xmath0 states .
the results show clearly that a sizeable part of the @xmath0 wave function lies `` outside '' the nucleus , a much larger fraction than in the neutron case . a comparison with experiment , see figure [ fig : expy ] and table [ tab : expv ] , shows that this mechanism is able to explain the fine structure of @xmath0 single particle spectra .
experimentally observed single particle spectra in low mass @xmath0 hypernuclei show a much more drastic reduction of the spin - orbit splitting .
this led to the conjecture that the spin - orbit splitting in hypernuclei should be in general very small and could be reasonably explained by assuming that the spin of the @xmath0 is carried solely by the strange quark .
the level spacing in light nuclei is , however , strongly affected by the continuum threshold which is in these small systems especially for the rather weakly bound @xmath0s quasi omnipresent .
as is known from neutron rich nuclei , e.g. @xcite , the single particle level spacing becomes compressed for states approaching the threshold .
o when the doublet s centroid is pushed towards the continuum threshold .
the dot marks the actual theoretical value for @xmath6o , the data point is taken from @xcite.,width=294 ] the threshold compression effect is visible already in the ca region but is especially important around and below @xmath6o . the typical feature in these nuclei is that the centroid energy of a spin - orbit doublet is rather close to threshold , mainly because of the shallowness of the binding potential . under such conditions
the spectrum of valence levels becomes compressed inducing an apparent reduction of spin - orbit strength because the less bound j=@xmath18 - 1/2 member of a spin - orbit doublet tries to avoid the crossing in the unbound region .
this `` hindrance '' is related to the general phenomenon of `` avoided level crossing '' in quantal systems because passing over into the continuum means that for example a @xmath19-level would have to cross the region of low energy s - wave continuum levels before re - appearing as a single particle resonance .
we illustrate these special effects for the 1p spin - orbit doublet in @xmath2o . approaching the continuum threshold
is enforced by modifying artificially the @xmath0 potential depth by variation of the @xmath0@xmath3 coupling constant .
figure [ fig : thresh ] shows the dependence of the splitting strength of the 1p doublet on the centroid energy @xmath20 .
even strongly bound states ( e@xmath21 - 6mev ) are still considerably affected by the threshold .
the effect only saturates if a state of another parity from the continuum is lowered into the bound region , in this example a @xmath22 orbit . the same behavior is also found in the 1p - shell doublet in @xmath4c , often referred to as the `` standard hypernucleus '' .
the continuum threshold effect reflects the peculiar dependence of matrix elements on wave function overlaps for weakly bound states .
in such cases - as observed in recent studies of pure isospin dripline nuclei - generic properties of interactions are masked behind wave function effects .
this has important consequences for extracting the desired information on the underlying interaction from data : the threshold effect leads to local variations in energy splittings not following the systematics derived from well bound systems .
obviously , this will strongly inhibit a model independent derivation of @xmath0 interaction strengths from measurements , especially in light nuclei . around and beyond a@xmath2340
, however , the threshold effect rapidly vanishes . since the new kek data @xcite for @xmath24 and @xmath25
were obtained in mass region well outside the threshold regime they can be expected to provide suitable information on the generic @xmath0 spin - orbit strength , indicating a larger value than assumed before .
the larger than expected spin - orbit splitting observed in recent kek experiments on medium mass nuclei are well reproduced by relativistic ddrh mean - field calculations .
universal parameter sets , derived previously from dirac - brueckner theory , were used .
the results indicate that the medium dependence of hyperon - coupling constants is essential for the good description .
the calculations explain consistently also the apparent suppression of spin - orbit strength for weakly bound @xmath0 orbitals in light nuclei by the compression of spectra close to the continuum threshold due to avoided level crossing .
it is worthwhile to emphasize strongly the importance of high - resolution data , like those of ref .
@xcite , for hypernuclear structure theory and the understanding of hyperon interactions .
this work was supported by dfg grant le439/4 - 3 and the european graduate school `` complex systems of hadrons and nuclei , gieen copenhagen '' .
m. may _ et al .
_ , phys .
* 47 * ( 1981 ) 1106 .
w. brckner _
[ heidelberg - saclay - strasbourg collaboration ] , phys .
* b79 * ( 1978 ) 157 .
j. mares and b. k. jennings , phys .
* c49 * ( 1994 ) 2472 .
m. chiapparini , a. o. gattone and b. k. jennings , nucl .
* a529 * ( 1991 ) 589 .
k. tsushima , k. saito , j. haidenbauer and a. w. thomas , nucl .
* a630 * ( 1998 ) 691 [ nucl - th/9707022 ] .
r. h. dalitz , d. h. davis , d. n. tovee and t. motoba , nucl .
phys . a * 625 * ( 1997 ) 71 .
r. h. dalitz , d. h. davis and d. n. tovee , nucl .
phys . a * 450 * ( 1986 ) 311c .
h. hotchi , doctor thesis , dept . of physics ,
university of tokyo ( 2000 ) , * 200024003 * kek report 2000 - 3 , april 2000 , h ; t. nagae , t. nagae , nucl .
a * 670 * ( 2000 ) 269c c. m. keil , f. hofmann and h. lenske , phys .
* c61 * ( 2000 ) 064309 [ nucl - th/9911014 ] . c. fuchs , h. lenske and h. h. wolter , phys .
* c52 * ( 1995 ) 3043 [ nucl - th/9507044 ] .
h. j. pirner , phys .
lett . * b85 * ( 1979 ) 190 .
d. cortina - gil , t. baumann , h. geissel , h. lenske et al .
, eurphys.j a ( 2000 ) ( in press ) .
h. lenske , j.phys .
g : nucl . part . phys .
24 ( 1998 ) 1429 .
t. baumann , h. geissel , h. lenske et al .
, phys.lett .
b 439 ( 1998 ) 256 . w. schwab , h. geissel , h. lenske et al .
a350 ( 1995 ) 283 .
|
new experimental data on medium to heavy single @xmath0 hypernuclei revealed a much larger spin - orbit splitting than observed in older measurements of light hypernuclei . taking into account particle threshold effects and the density - dependence of in - medium coupling constants the apparent suppression of spin - orbit strength in light hypernuclei as well as the spin - orbit structure observed in medium to heavy nuclei
are explained in a unified manner within the density dependent relativistic hadron field theory .
it is concluded that the most valuable information on the @xmath0 spin - orbit dynamics in finite nuclei has to be extracted from medium to heavy mass nuclei .
hypernuclei , spin - orbit splitting 21.80.+a
|
the motion of small co - orbital satellites can be studied using the @xmath0 body problem .
this is a restricted version of the @xmath5 problem of one central mass with @xmath6 other masses in a ring surrounding it , which was first studied by @xcite as a model of the motion of the rings of saturn . in the restricted version ,
the @xmath6 orbiting masses @xmath7 interact with each other , but not the central mass @xmath8 i.e. the @xmath7 are infinitesimally small and @xmath9 tends to zero .
the relative equilibria ( stationary configurations in a frame rotating with the co - orbiting satellites ) of both problems have been extensively studied in the context of central configurations : a configuration where the total ( newtonian ) acceleration of each body is proportional to the position vector of the body relative the centre of mass of the configuration . the simplest version of the restricted @xmath0 body problem is planar with the co - orbiting satellites in circular orbits with the same mean motion and same mass .
in this case the equilibrium configurations lie on a circle centred on the primary mass @xcite . for @xmath10
there are several different equilibrium configurations possible ( see sec .
[ sec : eq ] ) @xcite while for @xmath11 no configurations other than a regular 1+@xmath6-gon have been found , where the satellites are equally spaced in a circle about the central mass @xcite .
numerical evidence has been provided for this ( e.g. @xcite ) and analytic proofs for the cases @xmath4 @xcite and @xmath12 @xcite , but it remains an open problem @xcite . in the unrestricted @xmath5 body problem this @xmath6-gon configuration has been shown to be linearly stable for @xmath13 for large mass ratios between the central mass and orbiting satellites @xcite , while in the circular restricted @xmath0 body problem numerical investigations have shown that at least one linearly stable configuration exists for @xmath14 @xcite .
the linear stability results in combination with the lyapunov centre theorem indicate that natural families of periodic orbits of the co - orbital satellites exist about these equilibrium configurations for each value of @xmath6 . in the simplest case with @xmath2
the problem is superintegrable and there are two well known stationary configurations with the satellites separated by either @xmath15 ( linearly stable ) or @xmath16 ( linearly unstable ) .
the trajectories about the first are the familiar tadpole orbits ending on a homoclinic connection to the second equilibrium configuration . beyond this connection the trajectories are horseshoe orbits .
such a co - orbital configuration is observed in the solar system in the dynamics of the satellites janus and epimetheus around saturn e.g. @xcite . for systems with three or more satellites the dynamics are more complex , for example @xcite look numerically at the dynamics about some of the equilibrium configurations for small values of @xmath6 and find both periodic motion and chaotic motion occurs .
however , little appears to be known in general about the families of periodic orbits associated with these configurations for @xmath17 . since such periodic motion
is observed in the solar system for @xmath2 the problem is worthy of further study .
in addition to natural satellites , there are two other applications of such orbits in astrophysics : the first is the potential of such orbits to offer solutions for stable ` parking ' orbits of large numbers of small objects , such as captured asteroids around the moon ( for example the easily retrievable asteroids identified by @xcite ) or the moons of mars or larger main belt asteroids , for future in - situ resource use .
knowledge of the primary families of periodic orbits associated with the problem , and their stability , form a basis of engineering such solutions .
the second concerns exoplanetary systems .
there has in recent years been significant interest in the idea of co - orbital or trojan planets : that is , planetary systems where two or more planets orbit at the same mean distance from their star .
this idea was first explored in detail by @xcite , and has subsequently been the subject of numerous other works investigating not only the dynamics , but detectability and formation as well . almost all concentrate on two co - orbital planets , either both of similar mass or one of jupiter mass and another of earth - mass .
although such configurations can remain ` stable ' for long times ( e.g. @xcite ) and numerical models show plausible formation mechanisms ( e.g. @xcite ) , analysis of photometric and radial velocity data has yet to reveal any such system @xcite . much of the work on the dynamics of exoplanetary co - orbital planets considers either two co - orbital eccentric gas giant planets in the general three - body problem or terrestrial planets at the classical triangular lagrangian points in the eccentric restricted three - body problem .
the @xmath0 body problem on the other hand is suitable for non - negligible but low mass ratios and small eccentricities and so is more suited to similar mass terrestrial planets in roughly circular orbits .
such systems have been less discussed in the literature than those with giant planets . in this work
we look at the primary families of periodic orbits emanating from the equilibrium configurations for @xmath18 equal mass satellites , as well as some secondary connecting families , using the boundary - value problem methods implemented in the numerical continuation software auto @xcite .
a summary of the known equilibrium configurations is presented in section [ sec : eq ] and details of the periodic orbits associated with them in section [ sec : po ] .
the equations of motion of the circular restricted @xmath0 body problem have been derived by several authors , including @xcite . here
we follow that of @xcite .
all masses are assumed to be point masses , and the @xmath6 co - orbital satellites to be in a circular and planar orbit with the same average mean motion @xmath19 and radius @xmath20 . in a frame
centred on the central mass rotating with the average mean motion the position of a satellite is described by its angular position @xmath21 relative to an arbitrary reference direction and the relative displacement from the mean radius @xmath22 .
units can be chosen such that @xmath20 , @xmath19 and the mass of the central object @xmath8 are all unity .
the equations of motion are then @xmath23 where @xmath24 and @xmath25
this system also assumes that there are no close approaches between the satellites , otherwise the motion will not remain co - orbital .
the problem is also hamiltonian and the @xmath26 are related to the conjugate momenta to the co - ordinates @xmath21 .
there are three integrals of motion for this system : @xmath27 the first two correspond to the invariance of the system under rotations and hence the conservation of angular momentum and the last is the energy integral , or jacobi constant .
the system [ eq : eom1 ] is a first order dynamical system in @xmath28 .
however , the integrals of motion given by [ eq : int1 ] and [ eq : int2 ] can be used to reduce the system to @xmath29 by eliminating a pair of the @xmath30 and @xmath26 coordinates . for convenience
a new angular coordinate @xmath30 can be defined as @xmath31 for @xmath32 .
the angle @xmath30 now represents the angular separation of satellites @xmath33 from satellite 1 .
given @xmath34 the position of satellite 1 is then constrained through eq .
( [ eq : int2 ] ) .
similarly eq .
( [ eq : int1 ] ) allows the radial displacement @xmath35 of satellite 1 to be eliminated from the equations of motion as well with no need for the introduction of any other new co - ordinates .
the new equations of motion are for example in the case @xmath4 with equal masses @xmath36 : @xmath37 although this is not a canonical transform the numerical continuation does not require the system to be in canonical co - ordinates and this form is convenient to work with .
the reduction also does not affect the linear stability : one pair of eigenvalues in the original system will represent the invariance under rotation and it is these that are not present in the reduced system .
the equations of motion possess a number of symmetries which are relevant to the equilibria and periodic orbits .
firstly , there is a reversing symmetry @xmath38 .
a second less obvious reversing symmetry is due to the nature of the function @xmath39 , and is a symmetry about a line through the equilibrium configuration .
@xcite look at the @xmath40 body problem and prove that all relative equilibria are symmetric about a symmetry line , and that either two satellites lie on the symmetry line or none do ( one of the configuration in fact possess several symmetry lines , and is of both types ) .
@xcite discuss symmetries of the @xmath0 body problem , noting similarly that there are two cases that occur for equal masses : satellites on the symmetry line ( one if @xmath6 is odd or two if @xmath6 is even ) or no satellites on the symmetry line .
in addition , the case of equal masses has an additional spatial symmetry that is less obvious : @xmath41 in the coordinates used here .
we can write @xmath42 without need for multiplication by @xmath7 , implying @xmath43 ( as the mean radius is assumed to be 1 ) .
thus @xmath44 from eq .
[ eq : eom1 ] and @xmath45 .
however , @xmath46 is a sum of angles and thus @xmath47 also .
the dynamics depend only on the relative angular separation of each satellite rather than the location relative to the reference direction , so for any give value of @xmath46 a rotation of the entire configuration by @xmath48 also exists .
this additional symmetry is seen by @xcite for equilibria in the @xmath49 body problem when the masses are equal .
the consequence of the symmetry are seen in the connections between the families of periodic orbits emanating from the equilibria , and is discussed further in section [ sec : po ] .
the requirement that @xmath50 means that @xmath51 for all equilibrium configurations so that the satellites are all located on the mean circle with angular separations are given by solving @xmath52 . @xcite
determine the stationary configurations for @xmath2 to 9 for equal mass satellites , using numerical methods for @xmath6 greater than 3 .
the configurations for @xmath53 are shown in table [ tab : eqcf ] for a fixed value of the angular momentum constant @xmath46 .
there are three distinct types of equilibrium configuration , which @xcite label as type i ( arranged in a ` line ' ) , type ii ( arranged in an @xmath6-gon ) and type iii ( arrange in a line with one satellite on the opposite side ) with subdivisions a and b for the type i and iii cases ( as for some higher values of @xmath6 two variants of these configurations exist ) . as mentioned , for @xmath11 the only configuration known to exist is type ii , and it appears to be always linearly stable .
in addition to the numerical evidence provided by numerous authors , @xcite provides a more general analytic proof that places a bound on @xmath6 as greater than approximately @xmath54 for this to be the case .
the three types can be defined as follows : type ii is the simple case of equal spacing , and is linearly stable for @xmath55 at least .
the type ii configuration is unaffected by a rotation by @xmath48 up to an interchange of satellites , and has several symmetry lines .
type i has the satellites in an arc with the spacing increasing from the central pairs outwards .
it is symmetric with respect to a symmetry line through the configuration .
type iii has @xmath56 satellites in a similar arc and the remaining one on the opposite side , and has similar symmetry .
for the type i and type iii configurations @xmath56 other alignments of the configuration other than that shown in the table also exist .
a numerical search for equilibria for @xmath57 was also performed as part of this work and no other configurations were found in agreement with the results of @xcite .
the problem is hamiltonian and thus the eigenvalues of the linearized system occur in reciprocal ( and complex conjugate ) pairs . @xcite
provide the linear stability of each configuration , and the spectrum for each point is shown in the table , excluding the pair of eigenvalues which are zero in the full system .
note that the eigenvalues for the type ii configuration occur in degenerate pairs .
lcrcrcrcr @xmath6 & 2 & & 3 & & 4 & + & & & & & & + & & & & & + ia & & @xmath58 & & @xmath59 & & @xmath60 + & & @xmath61 & & @xmath62 & & @xmath63 + & & & & @xmath64 & & @xmath65 + & & & & & & @xmath66 + & & & & & & + + & & & & & + ii & & @xmath67 & & @xmath59 & & @xmath68 + & & @xmath69 & & @xmath70 & & @xmath71 + & & & & @xmath72 & & @xmath73 + & & & & & & @xmath74 + & & & & & & + + & & & & & + iiia & & & & @xmath59 & & @xmath68 + & & & & @xmath75 & & @xmath76 + & & & & @xmath77 & & @xmath73 + & & & & & & @xmath78 + & & & & & & + +
it is well known that a hamiltonian system with a first integral has natural families of periodic orbits defined by the value of this integral .
the @xmath0 body problem has both angular momentum and energy integrals , however the natural families in the angular momentum integral are trivial : they correspond to the rotation of the periodic orbit about the central mass .
the natural families corresponding to the energy integral @xmath79 are not trivial , and it is these which we consider .
further , the lyapunov centre theorem states that : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ let a hamiltonian system in @xmath80 degrees of freedom have an equilibrium point @xmath81 with characteristic exponents @xmath82 with @xmath83 .
if one of the exponents , say @xmath84 , is purely imaginary and the ratio @xmath85 is not an integer for all @xmath86 then there is a one - parameter family of periodic orbits emanating from the equilibrium point , and its period tends to @xmath87 at @xmath81 .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ thus in the reduced system there are one - parameter natural families of periodic orbits emanating from equilibrium configurations that possess pairs of ( non - resonant ) purely imaginary eigenvalues . from the linear stability results of @xcite ( summarized in table [ tab : eqcf ] ) we can see that there will be one family associated with the ia configuration for @xmath2 , whereas for @xmath3 two families emanate from the ia configuration and one from the iiia configuration . for
@xmath4 three families will emanate from the type ia configuration , one from the type ii configuration and two from the type iiia configuraiton .
( note that there may be other families in these systems not associated with the equilibria : for example the well known horseshoe orbits in the @xmath2 case . ) natural families of periodic orbits can be continued in the boundary - value problem ( bvp ) numerical continuation software auto @xcite .
auto uses polynomial collocation in association with pseudo - arclength continuation to solve a bvp formulation of the system .
briefly , the problem of finding a periodic orbit @xmath88 with unknown period @xmath89 of the first order system @xmath90 with parameter p can be written as the bvp : @xmath91 where time has been rescaled so that the periodic orbit has a period of 1 .
the first equation is the equations of motion , the second the periodicity constraint and the third a phase condition which fixes the phase of the periodic orbit @xmath88 relative to a reference solution @xmath92 ( usually the last periodic orbit along the branch ) .
the final equation is the pseudo - arclength constraint , which permits the family to be continued in the parameter @xmath93 . here
@xmath94 is the step - size and @xmath95 the tangent to the reference solution .
the use of pseudo - arclength continuation in particular allows families to be continued around folds relative the continuation parameter @xmath93 , ensuring the whole family can be continued .
the continuation of a family can be started either by branching directly from an equilibrium point or a branch point in a previously generated family , or by providing a known reference solution .
for further details see for example @xcite or @xcite .
the natural families of periodic orbits to be considered here are parameterised by the energy level @xmath79 , which does not appear in the equations of motion as formulated . however , a simple solution to this is to introduce an unfolding term into the equations of motion to act as the parameter @xmath93 .
@xcite show that a suitable term for systems with a first integral @xmath96 is @xmath97 , where @xmath98 is an unfolding parameter . for this system
the first integral we are interested in is the energy constant @xmath79 . the unfolding term
is easily added to the equations of motion , and periodic orbits can be generated directly from an initial equilibrium configuration . if @xmath99 then the system implemented in auto is @xmath100 where @xmath101 are the equations of motion for the reduced system . this bvp formulation and unfolding technique
has been applied numerous times in the context of celestial mechanics , for example to the crtbp by @xcite and @xcite and to choreographies of the @xmath6-body problem by @xcite . given a solution of the reduced system
the full system can be calculated by fixing a value of the constant @xmath46 and using equations [ eq : int1 ] and [ eq : int2 ] to determine the coordinates @xmath102 .
although the system is in @xmath103 in practice the angles do not start to circulate for the periodic orbits presented here and there is no issue with the numerical continuation .
the dynamics of the system scale with the relative masses of the satellites @xmath104 , and a transform @xmath105 , @xmath106 removes the dependence completely from equations [ eq : eom1 ] . in this work
we use a very high mass ratio of @xmath107 to clearly illustrate the nature of the orbits in the spatial plane only , although realistic mass ratios in the context of the solar system are much lower , and the dynamics scale accordingly .
the periodic orbits are otherwise shown in normalized co - ordinates in all other plots of the motion presented in the following sections .
the linear stability of a periodic orbit can be determined from its floquet multipliers .
these are the eigenvalues of the monodromy matrix ( the state transition matrix evaluated around the orbit ) or equivalently the eigenvalues of the linearized poincar map .
the floquet multipliers of each periodic orbit are calculated by auto , so information about the linear stability is provided for each continued family . as this is a hamiltonian system the floquet multipliers occur in inverse and complex conjugate pairs , with one pair always be equal to unity for a periodic orbit . hence true linear stability ( all multipliers within the unit circle ) is not possible in hamiltonian systems . instead
the linear stability can be described as different orders of instability , depending on how many pairs of multipliers are on the unit circle . for for @xmath108
the only possibilities are : * order-0 instability ( neutral linear stability ) : all floquet multipliers on the unit circle ( possible for @xmath109 ) * order-1 instability : one pair of floquet multipliers on the real axis ( possible for @xmath110 ) * order-2 instability : two pairs of floquet multipliers on the real axis , or in the complex plane but off the unit circle ( possible for @xmath111 ) . for @xmath17 the floquet multipliers change along a family of periodic orbits , and thus the linear stability of the family can change as well . for @xmath3 or 4
such changes generically occur through either collisions of one pair of multipliers at @xmath112 ( tangent bifurcation : fold or branch point to another family ) or @xmath113 ( period - doubling bifurcation ) . for @xmath4
the collision of two pairs of multipliers on the unit circle ( krein collision ) or real axis ( inverse krein collision ) is also possible . as stated in section [ sec : po ] the lyapunov centre theorem can be used to determine the existence of natural families emanating from those configurations with ( non - resonant ) centre eigenvalues . for @xmath114
there are no resonances between any of the purely imaginary eigenvalues associated with each equilibrium configuration .
thus for a given configuration , each centre frequency ( purely imaginary eigenvalues ) has a family of periodic orbits associated with it .
the linear stability of each configuration is given in table [ tab : eqcf ] .
so for example , the @xmath3 case has three families that emanate from the equilibria : two associated with the type ia equilibrium ( linear stability centre - centre ) and one associated with the type iiia equilibrium ( linear stability centre - saddle ) .
similarly , the @xmath4 case has six such families : three associated with the centre - centre - centre configuration ia , one associated with the centre - saddle - saddle configuration ii and two associated with the centre - centre - saddle configuration iiia .
note that the ii configuration has no centre frequencies for @xmath2 and @xmath3 and thus there are no families emanating from it .
this does not preclude the existence of families not associated with the equilibria , for example horseshoe orbits in the case @xmath2 .
for a given equilibrium configuration the families associated with each configuration can be labelled as @xmath115 , @xmath116 , @xmath117 and so on , in order of increasing frequency ( and thus decreasing period ) .
for example , the ia equilibrium for @xmath4 has three centre frequencies , i.e. it has eigenvalues of the form @xmath118 , with @xmath119 and @xmath120 .
the @xmath115 family is associated with the @xmath121 frequency and has a period that tends to @xmath122 as the family tends to the equilibrium , the @xmath116 family is associated with the @xmath123 frequency has a period @xmath124 at the equilibrium and the @xmath117 family a period @xmath125 at the equilibrium .
the label for the equilibrium configuration can be added to the end of the designator , as well as a subscript to distinguish the value of @xmath6 .
so the family @xmath126 is the family associated with the lowest centre frequency , emanating from the equilibrium configuration ia for @xmath3 satellites .
a summary of the families considered in this work is shown in table [ tab : pos ] .
.[tab : pos ] families of periodic orbits associated with the equilibrium configurations from @xmath2 to @xmath4 [ cols="<,^,^,^,^,^",options="header " , ]
in the case @xmath2 the system has one degree of freedom and the periodic orbits are well known .
they are presented here to give context to the @xmath3 and @xmath4 systems .
the families of periodic orbits for this case can be calculated using either the analytic theory of @xcite ( see also @xcite ) or by numerical continuation in auto .
the tadpole orbits ( the @xmath127 family ) emanate from the type ia equilibrium and can be generated in auto by branching from the equilibrium configuration .
the horseshoe orbits on the other hand require an initial orbit to be provided ( as they do not emanate from an equilibria ) , which can be easily calculated from the analytic theory .
both continuation methods were used to generate the families of periodic orbits , and found to agree fully .
the two families of periodic orbits , from the auto continuation , are shown in figure [ fig : ia2 ] .
the @xmath128 equilibrium configuration has one family @xmath129 associated with it , corresponding to tadpole orbits .
this family is entirely stable until it terminates on a homoclinic orbit onto type ii configuration . beyond
this homoclinic connection is a family of horseshoe orbits .
there are two alignments of the @xmath128 equilibrium possible for a given value of @xmath46 : the one shown in the figure and another with the satellites at @xmath130 ( assuming the reference direction is along the positive @xmath131 axis ) .
thus the full picture of the dynamics includes the rotation about the origin by @xmath16 of the family shown in the figure , and horseshoe orbits about the type ii configuration , as shown in figure [ fig : ia2 ] .
this is comparable to the ` north ' and ` south ' branches of families of periodic orbits that occur in the crtbp .
+ for three satellites there are three possible families associated with the equilibria : two emanating from the ia configuration and one from the iiia configuration . the two associated with ia ( @xmath126 and @xmath132 ) are shown in fig . [
fig : ia3 ] .
the middle satellite in the @xmath132 family remains stationary ( @xmath133 ) for the entirety of the family .
both families initially have order-0 instability , with @xmath126 switching to order-1 after a fold in the energy constant and @xmath132 switching to order-1 after a period - doubling bifurcation , and the stable portion is highlighted in fig .
[ fig : ia3 ] . both families terminate on a homoclinic orbit to the @xmath134 equilibrium .
plots of each family in the period - energy plane are shown in fig .
[ fig : n3ep ] , scaled to eliminate the mass ratio .
the family @xmath135 associated with the type iiia equilibrium is shown in figure [ fig : ia3 ] as well .
as for @xmath132 one satellite remains stationary .
the family has order-1 instability until it ends on a homoclinic orbit to the @xmath136 equilibrium .
similar to the @xmath2 case , the two equilibrium configurations ia and iiia exist in three orientations for a given value of the constant @xmath46 as they are not symmetric under rotation by @xmath137 .
thus the families associated with them have two additional symmetric branches .
this illustrates that the homoclinic orbit that family @xmath126 is also a terminating orbit of another branch of the @xmath135 family : in the spatial plane these tadpole orbits are interior to the homoclinic orbit .
all three families of periodic orbits are symmetric about the line through the centre of the configuration , as would be expected from the symmetric nature of their associated equilibrium configuration .
they also possess the reversing symmetry @xmath138 , @xmath139 .
+ the three families for @xmath3 plotted in the energy - period plane in units that are scaled to remove the dependence on the mass ratio @xmath140 and such that @xmath141 .
the order-0 part of the ia families is shown in purple , and the order-1 part in gray .
the iiia family is shown in black .
the families start on the left and all continue out to infinite period ] for four satellites there are now three , two and one family associated with the type ia , ii and iiia equilibrium configurations respectively .
in addition , as before the type ia and iiia configurations can be rotated by @xmath142 for a given value of @xmath46 .
thus there are nine possible equilibrium configurations for any given value of @xmath46 , as the ia and iiia equilibria can be rotated three times by @xmath142 . as for the previous systems with @xmath2 and @xmath3
this affects the connections between families of periodic orbits .
there are three families associated with the ia equilibrium for this value of @xmath6 .
the first family @xmath143 connects to the other alignments of the @xmath144 configurations via a branch point that is also a connection to the higher symmetry family @xmath145 associated with the @xmath146 configuration .
further , this is a degenerate branch point , as the family @xmath147 also connects to @xmath145 at the same point .
one branch of the @xmath144 family is shown in fig .
[ fig : n4families ] , with the branching orbit plotted in blue .
the connection of @xmath143 and @xmath147 to the @xmath145 family is similar to the symmetry - breaking pitchfork bifurcations that occur in the crtbp ( e.g. the north and south branches of the halo family from the planar lyapunov family ) .
the four branches could be considered to be the north , east , south and west branches of the @xmath143 family . similarly in the @xmath2 case the @xmath128 family has two branches , while in the @xmath3 case the @xmath148 and @xmath149 families have three branches associated with the @xmath150 symmetry . the family @xmath151 is also shown fig .
[ fig : n4families ] .
it ends in a connection to @xmath143 via a reverse period - doubling bifurcation .
an additional family ( labelled @xmath152 ) branches off @xmath151 and reconnects further along the family .
this family branches off in a symmetry - breaking pitchfork bifurcation , the new family losing the reversing symmetry @xmath138 . the family @xmath153 appears to connect to @xmath143 twice : once as a reverse - period tripling bifurcation , and finally ending at a reverse period - quadrupling bifurcation , behaviour very similar to the planar families associated with the l4 and l5 points in the crtbp .
it also has a number of other branch points to four other families , that can be labelled @xmath154 , @xmath155 , @xmath156 , @xmath157 in the order they branch from the original family . a schematic bifurcation diagram summarizing the connections between the primary families emanating from the ia equilibrium configuration is shown in fig .
[ fig : bfdiag ] . the linear stability and the location of various bifurcations
are also shown .
in the diagram each primary family is represented as a single line emanating from a particular equilibrium configuration , as labelled .
( note that this diagram shows only one branch of each family ) .
the various bifurcations along each family are indicated by various symbols on this line ( as defined in the key in fig .
[ fig : bfdiag3 ] ) .
further , the linear stability of the family is indicated through dashing or dotting of the line representing the family .
for example , the @xmath158 family is represented as the horizontal line labelled @xmath115 , starting at the equilibrium labelled @xmath159 on the left of the diagram ( represented as a large square ) and ending at the degenerate branch point on the vertical line which represents the @xmath160 family .
the various bifurcations along the family are shown as indicated , and the linear stability can for example be seen to change from order-0 initially to order-1 after a fold ( gray square ) , and so on .
[ fig : bfdiag1 ] shows the connections between the primary families from the type ia , ii and iiia configurations ( for a description of the families associated with the ii and iiia configurations see the next two subsections ) .
the @xmath161 and @xmath162 families are complex and their linear stability and bifurcations are not represented fully on this first schematic diagram .
instead only the points at which they connect to the @xmath163 family are shown . the full bifurcation diagram for these two families , including the secondary families connecting to the branch points in the ia families
are shown separately in the schematic bifurcation diagram in fig .
[ fig : bfdiag2 ] .
note that whereas these diagrams show all branch points , folds and period - doubling bifurcations , the only period tripling and quadrupling bifurcations that have been detected are those between the three primary families associated with the @xmath159 configuration .
the secondary families have been labelled in the order they branch off the primary family .
there is one family branching from @xmath151 ( @xmath152 ) and four from @xmath153 ( @xmath164 , @xmath165 , @xmath166 , @xmath167 ) as discussed above , as well as one family labelled @xmath168 that connects between @xmath155 and @xmath156 .
all these secondary families are shown in the appendix , and it is worth noting that several of them also have some order-0 instability , for example the relatively complex asymmetric family labelled @xmath168 is completely neutrally linearly stable .
there is one family associated with the @xmath146 configuration : a family of symmetric horseshoe - type orbits @xmath145 . as for its associated equilibrium point ,
this family of periodic orbits has degenerate floquet multipliers ( the two non - trivial pairs are equal ) , as evidenced by the double branch point to the @xmath143 and @xmath147 families .
the @xmath145 family only has one branch ( unlike the four branches of the ia and iiia families ) , as it is invariant under rotation by @xmath142 up to the interchange of satellites .
interestingly , some of the horseshoe orbits in the @xmath145 family have order-0 instability .
this occurs just after the double branch point shown in fig .
[ fig:8d ] , but is ended shortly later along the family by a period - doubling bifurcation .
as for the @xmath144 configuration there are four possible alignments of the @xmath169 configuration , and thus four rotated branches of the two associated families .
the first , @xmath147 , behaves similarly to the @xmath143 family and connects to the branches from the other alignments of the configuration at the degenerate branch point in the @xmath145 family .
the second , @xmath151 , is a tadpole - like family , and ends on a homoclinic orbit from the @xmath146 equilibrium . unlike the @xmath143 and @xmath145 families
there are no regions of order-0 stability along either of the @xmath169 families .
both are initially order-1 , with the @xmath147 family transitioning to order-2 after a fold .
as the @xmath170 family evolves towards the homoclinic orbit it undergoes an infinite series of pairs of branch points and pairs of period - doubling bifurcations .
the pair of floquet multipliers on the real axis tend to zero and infinity , while the pair on the unit circle circulate with increasing frequency , briefly moving off onto the real axis at each bifurcation .
the first two families branching off connect to the @xmath147 family through reverse period quadrupling bifurcations .
this behaviour is similar to that of the the planar lyapunov families around the triangular equilibrium points in the crbtp ( see e.g. @xcite ) .
all six primary families for @xmath4 are shown in fig .
[ fig : n4families ] in the spatial plan .
in addition , the families are shown in the energy - period plane in fig .
[ fig : stabind ] .
the order-0 portion of each family is highlighted in this plot .
note that the connection of the @xmath143 and @xmath147 families to the @xmath145 family can be clearly seen in this plane .
the six primary families are shown plotted in the energy - period plane in fig .
[ fig : stabind ] . the order-0 portion of each family is again highlighted .
the connection of two of the ia families to the ii family is clearly shown in this plane .
the primary families @xmath143 , @xmath151 , @xmath153 , @xmath147 and @xmath170 emanating from the ia and iiia equilibria are symmetric about the symmetry line of the associated equilibrium configuration .
they also possess the reversing symmetry @xmath138 , @xmath139 .
the family @xmath145 emanating from the type ii equilibrium is symmetric about ( both ) symmetry lines of the equilibrium configuration and is also symmetric under @xmath171 .
it too has the reversing symmetry @xmath138 , @xmath139 .
the family @xmath152 branching from @xmath151 is not symmetric about the symmetry line of the equilibrium , but still possess the reversing symmetry @xmath138 , @xmath139 .
the families @xmath154 and @xmath156 branching from @xmath153 do not have the possess the reversing symmetry @xmath138 , @xmath139 but are symmetric about the symmetry line of the equilibrium , while the families @xmath155 and @xmath157 are not symmetric about the symmetry line of the equilibrium but do possess the @xmath172 reversing symmetry .
the family @xmath168 that connects between branch points in the @xmath155 and @xmath156 families possesses no symmetries .
it should be noted that those families that do not possess one of the reversing symmetries have two symmetric branches , similar to for example the axial families in the crbtp . because it lacks both reversing symmetries there are four branches of the @xmath168 family , connecting between the two branches of each of the @xmath155 and @xmath156 families .
the six primary families emanating from the equilibrium configurations for @xmath4 , in the energy - period plane .
the ia familes are shown in red , the ii family in green and the iiia families in blue .
the energy level of each equilibrium is also shown , and note that the families are labelled @xmath115 , @xmath116 etc in order of increasing initial period .
solid lines indicated order-0 stability , dotted lines order-1 or order-2 stability ]
the families associated with the equilibrium configurations of the planar restricted @xmath0-body problem are complex and exhibit a variety of interesting dynamical phenomena , such as homoclinic connections and degenerate symmetry - breaking bifurcations . for @xmath53
the families associated with the type ii equilibria resemble classical horseshoe orbits , while some associated with the type i and type iii equilibria resemble tadpole orbits . in addition
, there are many other periodic orbits , both symmetric and asymmetric exhibiting different linear stability properties ( although note that vertical stability of the orbits is not considered here ) .
while exactly equal mass satellites and asteroids in the solar system are unlikely to occur , the equilibrium configurations are known to exist in such cases ( e.g. @xcite ) and as such the periodic orbits presented here provide a starting point for exploring such systems .
preliminary investigation of unequal mass satellites have shown that many families of periodic still exist .
while this work has only considered up to four equal mass satellites the linear stability of the known equilibria for @xmath173 indicate a large number of families of periodic orbits exist in these cases as well .
for example , for @xmath174 there are type ia , ii and iiia equilibria as for the @xmath4 case , and it is found that there are two families emanating from the type ii configuration , one of which connects to the type ia equilibrium point and one of which connects to the type iiia equilibrium point .
the families remain of similar types to those seen in this work : largely tadpole or horseshoe type .
it should be noted here that the @xmath0-body problem is a restricted problem , and for the periodic orbits to be reproducible in the unrestricted system the mass ratios must be low . while this is not a problem for asteroids around planets or large moons in the solar system it can become an issue when considering trojan exoplanetary systems .
however , direct numerical integration of the full @xmath6-body equations of motion of an example order-0 periodic orbit from the @xmath168 family shows for example four earth - mass co - orbital planets at 1 au from a solar - mass star are well represented by this model .
we also note that such a periodic orbit of four earth - mass planets would have a period of approximately 250 years , making detection of the co - orbital dynamics difficult .
however , the existence of neutrally linearly stable periodic orbits of four trojans still presents an interesting possibility for exoplanetary systems .
the question remains whether such orbits could exist ( or be engineered ) in the solar system or exoplanetary systems .
for co - orbital asteroids , the easily retrievable asteroids suitable for capture into the earth - moon system identified by @xcite are at most diameter 65 m , with those that are retrievable in the near - term generally below 4 m diameter , which corresponds to @xmath175 to @xmath176 ( assuming an average density of 2 g/@xmath177 ) . for these extremely low mass ratios the co - orbital dynamics
will be negligible compared to other perturbations .
( by comparison , an asteroid with a higher mass ratio of @xmath178 would correspond to an spherical object roughly of diameter 40 km . )
this is in some sense a self - solving problem : if the asteroids do not significantly perturb each others orbits , there is no need to consider placing them in a stable parking configuration .
it should also be noted that the @xmath0 body problem is not a good approximation of asteroid orbits about the moon regardless of mass ratio , as the perturbations from the earth can not be ignored .
the families of periodic orbits found here however provide a starting point for more advanced models of the dynamics .
the mechanism by which asteroids could be transferred into such orbits is as well is another topic , but knowledge of the possible orbital configurations will also assist such endeavours .
this work was funded through the european research council , advanced investigator grant 227571 visionspace .
we thank the anonymous referee for helpful comments . 99 albouy , a. , cabral , h. e. , santos , a. a. : some problems on the classical n - body problem , celestial mechanics and dynamical astronomy , 113 , 369 ( 2012 ) albouy , a. , fu , y. : relative equilibria of four identical satellites , royal society of london proceedings series a , 465 , 2633 ( 2009 ) beaug , c. , sndor , z. , rdi , b. , sli , .
2007 , a & a , 463 , 359 calleja , r. c. , doedel , e. j. , humphries , a. r. , lemus - rodrguez , a. , oldeman , e. b. : boundary - value problem formulations for computing invariant manifolds and connecting orbits in the circular restricted three body problem , celestial mechanics and dynamical astronomy , 114 , 77 ( 2012 ) casasayas , llibre & nunes 1994 , _ central configurations of the planar 1+n body problem _
, cemda 60 , 274 corbera , cors & llibre 2011 , _ on the central configurations of the planar 1 + 3 body problem _ , cemda 109 , 27 cors , llibre & oll 2004 , _ central configurations of the planar coorbital satellite problem _ , cemda , 89 , 319 doedel , e. j. et al .
: auto-07p : continuation and bifurcation software for ordinary differential equations ( 2011 ) doedel , e. j. , romanov , v. a. , paffenroth , r. c. , keller , h. b. , dichmann , d. j. , galn - vioque , j. , vanderbauwhede , a. : elemental periodic orbits associated with the libration points in the circular restricted 3-body problem , int . j. bifurc
. chaos , 17 , 26252678 ( 2007 ) doedel , e. j. , govaerts , w. , kuznetsov , y. a. , dhooge , a. : numerical continuation of branch points of equilibria and periodic orbits , int .
j. bifurc .
chaos , 15 , 841860 ( 2005 ) doedel , e. j. , paffenroth , r. c. , keller , h. b. , dichmann , d. j. , galn - vioque , j. , vanderbauwhede , a. : computation of periodic solutions of conservative systems with application to the 3-body problem , int .
j. bifurc .
chaos , 13 , 1353 ( 2003 ) doedel , e. j. , keller , h. b. , kernvez , j. p. : numerical analysis and control of bifurcation problems ( i ) : bifurcation in finite dimensions , int . j. bifurc
. chaos , 1 , 493520 ( 1991 ) garca yrnoz , d. , snchez , j. p. , mcinnes , c. : easily retrievable objects among the neo population , celestial mechanics and dynamical astronomy , 116 , 367 ( 2013 ) giuppone , c. a. , bentez - llambay , p. , & beaug , c. 2012 , mnras , 421 , 356 laughlin , g. , & chambers , j. e. 2002 , aj , 124 , 592 madhusudhan , n. , & winn , j. n. 2009 , apj , 693 , 784 maxwell , c. j. : on the stability of the motion of saturn s rings , in scientific papers of james clerk maxwell , cambridge university press moeckel 1994 , _ linear stability of relative equilibria with a dominant mass _ , j. dynamics & diff .
eq . , 6 , 37 muoz - almaraz , f. j. , freire , e. , galan - vioque , j. , vanderbauwhede , a. : continuation of normal doubly symmetric orbits in conservative reversible systems , celestial mechanics and dynamical astronomy , 97 , 17 ( 2007 ) munz - almaraz , f. j. , freire , e. , galn , j. , doedel , e. , vanderbauwhede , a. : continuation of periodic orbits in conservative and hamiltonian systems , phys .
d nonlinear phenom . , 181 , 138 ( 2003 )
murray , c. d. , dermott , s. f. : solar system dynamics , cambridge university press , cambridge ( 1999 ) press w. h. , flannery b. p. , teukolsky s. , vetterling w. t. : numerical recipes in c , cambridge university press , cambridge ( 1989 ) renner , s. , & sicardy , b. 2004 , _ stationary configurations for co - orbital satellites with small arbitrary masses _ , celestial mechanics and dynamical astronomy , 88 , 397 salo & yoder 1988 , _ the dynamics of coorbital satellite systems _ ,
a & a , 205 , 309 schwarz , r. , sli , . ,
dvorak , r. , & pilat - lohinger , e. 2009 , celestial mechanics and dynamical astronomy , 104 , 69 schwarz , r. , dvorak , r. , sli , . , & rdi , b. 2007 , a & a , 474 , 1023 yoder , c. f. , colombo , g. , synnott , s. p. , yoder , k. a. : theory of motion of saturn s coorbiting satellites , icarus , 53 , 431 ( 1983 )
this section provides example orbits for the families associated with the @xmath4 equilibria .
the primary families @xmath143 , @xmath151 , @xmath153 , @xmath145 , @xmath147 and @xmath170 are shown in figs . [
fig : a1 ] to [ fig : a3 ] .
the secondary connecting families are shown in figs .
[ fig : b1 ] to [ fig : c1 ] .
as discussed in the text these families are shown for high mass ratio of @xmath179 in order to exaggerate the radial extent of the orbit .
|
we investigate the natural families of periodic orbits associated with the equilibrium configurations of the the planar restricted @xmath0 body problem for the case @xmath1 equal mass satellites .
such periodic orbits can be used to model both trojan exoplanetary systems and parking orbits for captured asteroids within the solar system . for @xmath2
there are two families of periodic orbits associated with the equilibria of the system : the well known horseshoe and tadpole orbits . for @xmath3
there are three families that emanate from the equilibrium configurations of the satellites , while for @xmath4 there are six such families as well as numerous additional connecting families .
the families of periodic orbits are all of the horseshoe or tadpole type , and several have regions of neutral linear stability .
[ firstpage ] celestial mechanics planets and satellites : dynamical evolution and stability minor planets , asteroids : general
|
current and planned cosmological surveys in the optical and infrared have fundamental limitations .
multi - fibre [ e.g. 2df @xcite and sdss @xcite ] and multi - slit [ e.g. deimos @xcite and vimos @xcite ] surveys have or will amass large catalogues of galaxies in order to deduce their global properties .
however these suffer from biases introduced when a fixed angular size aperture such as a single fibre is used to observe galaxies , irrespective of the size , distance or morphology ( see for example , fig.8 in * ? ? ?
* ) . on the other hand , integral field units ( ifus ) like teifu @xcite and gmos @xcite or image slicers such as sinfoni @xcite and nifs @xcite spatially sample the spectra , giving morphological and dynamical information
however , current ifus are restricted in the number of objects that can be observed .
the ideal is to combine multi - object spectroscopy ( mos ) positioning technology with ifus .
hexabundles can do exactly that .
hexabundles have up to many hundreds of closely - packed fibres to allow spatially resolved spectroscopy .
they have several practical advantages .
firstly , they are not as sensitive as single - fibre devices to seeing and positioning errors and are unaffected by seeing losses .
secondly , they can be used with ao - corrected or natural seeing and in the optical or infrared .
thirdly , there is no need for microlens arrays , and the plate scale can be changed simply with a single macro lens .
fourthly , they can be used with conventional fibre positioning technology as imaging bundles to obtain spatial information in a survey of many thousands of galaxies .
the resulting scientific gains over existing large galaxy surveys would then include the ability to investigate agn triggering and feedback including outflows , galaxy merger rates and merger - induced processes , the substructure in gravitational lenses , stellar populations and abundance gradients , as well as tracing the build up of dark matter , stellar mass and angular momentum in galaxies .
decomposition of bulge and disk components would be possible for thousands of galaxies .
hexabundles are particularly effective for galaxies with asymmetries , multiple components , mergers and substructures including high redshift galaxies that are not yet formed into spheroids or disks but consist of clumpy merging components which would be poorly sampled with a single fibre spectrum .
differential binning of the fibres can also be applied , particularly for more distant objects where the brightness decreases rapidly away from the nucleus .
a 100,000 galaxy survey out to a redshift of 0.2 will be possible with hexabundles in the proposed fireball instrument on esos very large telescope ( vlt ) .
currently the flames instrument uses the ozpoz positioner to place 132 single fibres across a @xmath1 field , which then feed into the giraffe spectrograph .
the proposed fireball upgrade would involve fifty hexabundles , with around 100-cores each sampling @xmath2 .
hexabundles are not limited in the number of cores in each bundle , beyond the physical size of the bundle holder .
so far we have made bundles up to 61 cores , but larger devices of several hundred cores would be suitable for multi - object positioners . in anticipated applications ,
the core will have @xmath3 per core . then
61-core and 367-core bundles would have imaging areas @xmath4 and @xmath5 across respectively .
the main trade - off with the design of hexabundles is to have the largest fill - fraction ( ratio of core area to total bundle area ) possible without compromising the optical performance .
the fill - fraction is affected by both the cladding thickness and how the fibres are packed together .
fibres that are fully - fused together have smaller gaps between cores giving a fill - fraction of over 90% , but are significantly distorted from a circular shape . on the other hand ,
lightly - fused fibres remain circular but have larger gaps between them , resulting in fill - fractions less than 87% .
this paper describes the performance tests of our first fully - fused and lightly - fused hexabundles with an aim to assess the trade - offs of fill - fraction versus optical performance .
section 2 describes the hexabundle devices , while the experimental method and data reduction are in section 3 .
sections 4 and 5 have the results for the fully - fused 61-core hexabundles and the lightly - fused 7-core hexabundles respectively .
final conclusions on the comparison of the different types of hexabundles is in section 7 .
six new hexabundle devices have been characterized . in each device
, one end has multimode fibres with reduced cladding thickness , fused into a single element , which is the hexabundle .
the fibres at the other end are loose and have normal cladding thickness .
the first of these hexabundles consists of 61 fully - fused fibres without interstitial holes ( fig .
[ 61bundle_pic ] ) giving a fill - fraction of @xmath6% .
however , all but the central fibre are significantly distorted from a circular shape .
the fibres are 4.5 m long and each fibre had core and cladding diameters of @xmath7 m and @xmath8 m respectively , before the cladding was etched away over a @xmath9 cm length where the fibres were then fused ( further details of the manufacturing process can be found in * ? ? ?
the resulting hexabundle has a diameter of @xmath10 m . on output ,
the 61 fibres are lined up in v - grooves between glass plates with three rows of 15 and one row of 16 fibres .
the fibres in adjacent grooves in the same glass plate are separated by @xmath11 m .
the remaining five hexabundles have seven lightly - fused fibres ( with small interstitial holes ) with cladding thicknesses of 1 , 2 , 4 , 6 and @xmath12 m .
the main difference to the previous bundle is that the fibres are circular rather than distorted in shape ( see fig . [ lightly - fusedbundle_pic ] ) , resulting in lower fill - fractions . in these bundles ,
the output fibres are loose and were mounted into 35 parallel v - grooves on a plate with careful attention not to apply pressure to the fibres .
we aim to investigate the effect of circular cores on the optical performance and consider the trade - off with fill - fraction .
the different cladding thicknesses are used to assess the ideal balance between minimising cross - talk while maximising fill - fraction .
each of the bundles were tested for cross - talk , relative throughput and mode - dependent losses .
it was crucial for the cross - talk tests in particular , to ensure that no stray light was getting into any core other than the intended target .
originally laser light was focussed into the individual fibre ( output ) end of the fibres while the hexabundle end was imaged with the camera .
this has the advantage that the fibres could be separated so that no stray light from the laser could go into adjacent fibres .
however , the light coming out of the multimode fibres at the hexabundle has a speckle pattern that is larger than the size of the core .
this is because the different modes focus in slightly different planes outside of the polished bundle face so that all of the light out of the fibre can not be focussed at the same time , resulting in an imaged spot size that has an apparent diameter of approximately 125% of the core diameter .
an example of this effect is shown in fig .
[ speckle ] . in any adjacent fibre
it is impossible to distinguish light from this oversize speckle pattern from light leaked into the fibre through cross - talk . the final method adopted , as shown in fig .
[ setup_pic ] , used an oriel led light source which was focussed into a @xmath13 m core fibre with a @xmath14 m diameter including cladding .
this fibre was then butt - coupled to the @xmath0 m hexabundle cores ( reversing the direction of the light through the fibres compared to the previous method - this is the direction the hexabundles will be used in astronomical applications ) .
we mounted the butt - couple ( input ) fibre in a v - groove on a narrow plate that was attached to a tilt / rotation stage off to the side of the hexabundle mount .
this minimised obstruction of the area around the hexabundle face so that the fibre could be adjusted to align perpendicular to the hexabundle face .
the narrow plate could then slot in from the side between the hexabundle and a microscope with the fibre bending abruptly away to the side .
this allowed the microscope to be positioned close enough to image the hexabundle and input fibre at high enough magnification to be sure the input fibre was centred on a chosen fibre core within the hexabundle .
the input fibre needed to be less than @xmath0 m from the hexabundle face to prevent light straying into adjacent fibres .
the small outer diameter of the butt - coupled fibre was essential in order to not obscure the view of the hexabundle cores .
two other microscopes mounted above and side - on were used to ensure that the fibre was perpendicular to the hexabundle face in both axes . the maximum error contribution to the na from the input alignment was @xmath15 .
the output fibres were each imaged through a pair of large camera lenses onto an sbig camera . a micrometer stage allowed the distance from the focus to the backfocus position of the camera to be accurately measured .
for all the measurements , the led source light level was kept constant .
each measurement was done twice , once with a bessel blue filter ( centred on @xmath16 m ) then with a bessel red filter ( centred on @xmath17 m ) between the led source and the focussing assembly .
time variation in the input light intensity plus variations in the sbig camera response , were quantified with repeated images through one core of the bundle at time periods ranging from second to hours .
the maximum variation in resulting integrated counts was @xmath18% with typical values of @xmath19% .
the relative throughput of every fibre in each bundle was measured by centring the input light on each fibre . for cross - talk measurements ,
the light was centred on a core while the surrounding fibres were imaged
. fits images from the sbig camera were processed using the iraf astronomical data reduction software @xcite .
firstly each image was flat fielded .
the background was fitted and subtracted in iraf radprof using an annulus with inner radius approximately 3 times the maximum radius of the fibre image .
then the total integrated counts were measured in iraf radprof by fitting the barycentre of the fibre image and selecting a fixed aperture size for all fibres that was large enough to include all the output from the fibre .
the target fibre and the surrounding fibres were measured through the same size aperture .
we also aimed to compare the performance of different fibres within the same bundle in terms of focal ratio degradation ( frd ) or mode - dependent losses .
frd ( e.g. carrasco and parry 1994 ) is a loss of optical entropy or non - conservation of _ ntendue _ due to mode mixing as light propagates down the fibre .
this modal mixing causes the opening angle of the output beam to be larger than that of the input beam resulting in a wider encircled energy distribution ( see fig . [ naeediag ] ) .
the modal coupling means that lower - order modes will couple " to higher - order modes which have a larger angle at the output beam , hence a larger na . in order to compare the consistency of the performance of different fibres within the same bundle
, it was therefore important to ensure that all modes were filled on input . when the cores are distorted , under - filling the na on input may mean the light couples into different modes at the input for different fibres , making it invalid to compare the distribution on output .
therefore we have filled the na of the fibres to excite all possible modes , so that the output energy distribution across the output image in the back - focal plane can be directly compared for each fibre .
the encircled energy profile at back - focus , will therefore be broader for higher modal coupling .
the camera was shifted to a known back - focal distance ; selected to give sufficient pixels across the output fibre image to accurately fit an encircled energy profile . using the known pixel size of the camera
, an na value was calculated from sin@xmath20 of the cone angle .
this gives the effective output na at each point across the image of the fibre .
this na value is used in the graphs to follow . after flat fielding
, iraf radprof was used to fit the centre position along with the background and peak counts in each image profile .
these initial values were then input into iraf pprofile , and the radially - collapsed encircled energy profile was fitted interactively .
an iterative procedure was used to determine the background value to subtract and the peak counts in the profile until the background was flat and the encircled energy levelled at 100% .
the encircled energy values could then be matched to the na calculated for each radial pixel position in the image .
the resulting plots of na vs encircled energy will therefore be shifted to the right when there are worse mode - dependent losses ( larger mode - dependent losses will give higher frd ) .
the encircled energy profile of each of the fibres in the 61-core hexabundle , was measured against the output na .
fig.[avee ] shows the encircled energy profiles through both the bessell blue ( @xmath21 m ) and red ( @xmath22 m ) filters .
the profiles through the red filter are at a lower na than those through the blue filter because the shorter wavelengths have more modes and therefore a higher na .
while the average profiles of 16 cores get worse as the range of core numbers gets further from the bundle centre , there is a very large variation in encircled energy profiles from fibre - to - fibre , which will give inconsistent performance across the bundle for imaging and spectroscopy . in order to quantitatively
compare the encircled energy profiles , the drop in encircled energy compared to the central fibre was calculated for a nominal na of 0.20 .
this na is chosen because it is less than the total size of the output cone ( output na at 100% encircled energy ) .
the maximum na is truncated by mode - dependent losses out of the fibre , however the distribution of encircled energy within that cone will also be broadened by modal coupling , and more modal coupling is indicative of more frd .
therefore , by selecting the encircled energy loss at an na of 0.2 , we are measuring the broadening of the profile due to modal coupling .
the resultant drop in encircled energy ranges from less than 1% up to more than 25% for different fibres in the bundle . at 90% encircled energy ,
the up - conversion of na is 0.10 for the worst fibre in both the blue and red filters . to see
if the deformation of each fibre affects the encircled energy profiles , the deformation was measured by fitting circles to the image shown in fig .
[ 61bundle_pic ] . the largest circle that does not include any cladding , and the smallest circle that includes all
the core were fit to each core .
the ratio of the diameters of these circles was used to define a ` deformation ratio ' . while this measure is subjective , in fig .
[ eedrop ] , the plot for deformation ratio against encircled energy drop shows a correlation with spearman rank correlation coefficient of 0.34 ( 95% probability ) .
the fibres with the highest deformation ratio tend to have the largest encircled energy drop , indicating the highest modal coupling .
cross - talk was measured for 14 random fibres in the bundle by measuring the integrated counts in each adjacent fibre while the light was centred on the target fibre .
the loss was not equally distributed into the adjacent fibres , with most target fibres showing the majority of the cross - talk into one or two of the adjacent fibres while in some cases the counts were more than 100 times lower in other adjacent fibres .
this was due partly to significant differences in the contact area and cladding width between different fibres .
such differences arise from the distortion of the fibres in the fully - fused bundle .
the total loss into adjacent fibres was predominantly in the range of @xmath23% ( 0.005 - 0.06db ) with one fibre at 4.2% with the blue filter , and one at 4.4% ( 0.2db loss ) in the red filter .
the total cross - talk could not be evaluated in the outer row of the hexabundle because we have shown that the loss is unevenly distributed to adjacent fibres and therefore could not be assessed where there were no adjacent fibres on the outer edge . as the outer row of fibres include some of the most distorted profiles , the total cross - talk is likely to be above 0.2db for some of those fibres .
for example , the cross - talk from fibre 56 into just 55 and 57 is already a total of 0.17db alone and therefore would be much higher if the losses could be measured on all sides . in fig .
[ eedrop ] there is an anti - correlation ( with spearman - rank test correlation coefficient of -0.65 , and 99.9% probability ) between the drop in encircled energy and the throughput .
throughput losses are a combination of higher cross - talk in the more distorted fibres and modal stripping ( indicative of worse frd )
. the encircled energy drop is not measured at the edge of the output light cone but at na@xmath24 .
frd causes lower - order modes to be coupled to higher orders , broadening the beam to give this encircled energy loss . at the same time , the higher - order modes can be lost entirely and therefore the throughput is affected by worse modal stripping .
the performance of different fibres within the same bundle is very inconsistent , making it ineffective as an imaging device . with the exception of one fibre with very high cross - talk , that performed differently to the others in every test
, the encircled energy drop was also correlated with cross - talk with a spearman rank correlation coefficient of 0.60 ( with @xmath25% probability ) .
it can also be seen that the throughput varied by a factor of five between each of the 61 cores which would have an impact on the accuracy of photometry in extended astronomical sources .
it is clear that the more distorted fibres have lower throughput , worse cross - talk and higher encircled energy drop .
it was also noticed that when the input light was directed at the cladding between fibre cores , that light was then dispersed through the entire bundle . in an astronomical application ,
the component of an extended source that hit the cladding would result in all positional information being lost and the spectrum of that component of the object being added to every fibre across the field . by centring the light on the cladding between fibres 11 and 12 , then measuring the counts in 6 fibres that were not adjacent , the counts were @xmath26% of the counts in fibre 12 when the light was directed into fibre 12 .
fibre 12 was chosen because the throughput counts in that fibre are average compared to the counts in each of the other 61 fibres .
based on an estimate of the cladding area covered by the light in this test compared to the total cladding area , up to 25% of the light in any one fibre could be scattered from cladding elsewhere in the bundle .
this fraction would lead to unacceptable confusion in the spectroscopic imaging of an extended astronomical source .
the original aim of fully - fusing the bundle was to increase the fill - fraction to @xmath6% .
however , based on this performance , the gain in coverage across an object being imaged , is outweighed by the confusion in the spatial origin of the light at the spectrograph due to cross - talk and cladding scatter . also , inaccuracies in photometry will result from the extreme differences in throughput across an image .
the next step was to consider the improvement in this performance if the hexabundle was lightly - fused so that the fibres remain circular at the expense of a slightly smaller fill - fraction .
we had five hexabundles made with cores that are lightly - fused .
these have the advantage that the fibres are not distorted and therefore is expected to perform more consistently . in order to compare cross - talk for a given fill - fraction
, they were made with cladding thicknesses of 1 , 2 , 4 , 6 and @xmath12 m ( see * ? ? ?
* for a discussion on the issues surrounding the manufacture of bundles with these cladding thicknesses ) .
the aim was to test how thin the cladding could be before the cross - talk became too large .
[ eelightly - fused ] shows the encircled energy profiles for the lightly - fused bundles .
the profiles of each of the 7 cores in the 2 , 4 and @xmath12 m clad bundles agree within errors , indicating consistent modal coupling performance across the bundle .
this is in contrast to the large variation between cores in the fully - fused bundle ( see fig . [ avee ] ) .
each of the profiles in fig .
[ eelightly - fused ] have lower na than ( sit to the left of ) the best performing fibre ( central fibre ) of the fully - fused bundle , indicating lower modal coupling ( and therefore lower frd ) in the lightly - fused bundles .
the @xmath27 m cladding bundle has a distribution of encircled energy profiles for the 7 cores that is slightly larger than errors , which may be due to mode stripping induced by the higher cross - talk . in this case
the central core has worse na than the other cores as would be expected from mode stripping based on the symmetry of the bundle . for each of the lightly - fused hexabundles , the relative throughput was measured for each of the 7 fibres , and
the cross - talk was measured by summing the losses into the surrounding 6 fibres when light was directed down the central core .
table [ ffvsct ] shows the cross - talk performance through the @xmath28 and @xmath29band filters compared to the fill - fraction . while the hexabundles with 4 , 6 and @xmath12 m clad fibres had no cross - talk down to the limit we could measure in our images
, they also have a lower fill - fraction .
the cross - talk was worse for the @xmath27 m clad bundle than for the @xmath30 m clad bundle , as expected . in the anticipated uses of these hexabundles ( e.g. fireball )
the fibres would be @xmath31/core - comparable to the best seeing at an excellent telescope site of @xmath32 , or average seeing at a good telescope site of @xmath33 .
if the core is the size of the fwhm of a gaussian seeing profile , then 50% of the encircled energy will be in the central core and @xmath34% in the adjacent the surrounding cores .
the @xmath30 m clad bundle has a total cross - talk of @xmath35% ( at @xmath17 m ) , which is @xmath36% of the star light in the surrounding cores , but for the @xmath27 m clad bundle ( with up to @xmath37% cross - talk ) , approximately a tenth of the light in the surrounding cores will be cross - talk contamination .
therefore , at optical wavelengths , the @xmath30 m clad bundle is suitable for applications where the core size is comparable to the fwhm , and the @xmath27 m clad bundle may also be suitable if there are several cores within the fwhm of the seeing profile . in the infrared
, longer wavelengths will cause the cross - talk to be worse and thicker cladding would be required .
when the light was directed at the cladding , there was no detectable transmission of that light into all the cores as was seen with the cladding scatter " in the fully - fused bundle ( see section [ cladscat ] ) .
.total cross - talk from the central fibre into all of the 6 surrounding fibres is listed as both a percentage of the integrated counts in the central fibre and as a db loss .
values are shown for the blue @xmath28band and the red @xmath29band bessel filters .
the first column lists the cladding thickness , then the fill - fraction ( ratio of core area to total bundle area ) is given for each cladding thickness .
limits listed are set by the image detection thresholds in cases where no emission was detected . [
cols="^,^,^,^,^,^ " , ] for the two bundles for which the cross - talk was measurable above the noise limit , the variation in the cross - talk values for the 6 surrounding fibres was at worst a factor of 4 .
this is significantly lower than the factor of @xmath38 we found for the 61-core fully - fused hexabundle .
we attribute this to the uniformly circular fibres in the lightly - fused bundles which have equal contact areas with adjacent cores .
the even , circular - nature of the fibres in the lightly - fused bundles has also resulted in a much more uniform throughput from fibre - to - fibre . in fig .
[ eedrop ] the variation in throughput between different fibres in the fully - fused bundle was up to @xmath39% .
however , the lightly - fused bundles showed at worst 24% variation between fibres in any bundle for a given filter .
typically the variation was a lot less than this , as shown in fig .
[ thrunum ] .
we have compared the performance of a 61-core fully - fused hexabundle to that of 5 lightly - fused bundles with varying cladding thickness .
the distortion of the cores in the fully - fused bundle was found to significantly increase cross - talk , reduce throughput and increase mode - dependent losses .
the throughput varied by a factor of 5 between different cores in the bundle , while the cross - talk reached over 4% in the worst core .
the up - conversion in na from the best ( central ) core to the worst was 0.1 at 90% encircled energy .
the performance of the lightly - fused cores was better in every respect , and we attribute that to the uniformly circular cores .
the throughput varied by at most 24% and the modal coupling results were consistent between the 7 cores in each bundle within errors . after comparing the cross - talk to the cladding thickness
, we found that @xmath30 m cladding bundle had @xmath40% cross - talk in the @xmath29 and @xmath28bands while still having a fill - fraction of 84% .
while the fill - fraction of the fully - fused bundle is @xmath6% , the superior optical properties of the lightly - fused bundles outweighed the gain in fill - fraction .
we would like to thank roger haynes for valuable discussions and input .
jjb and jbh are supported by arc grant ff00776384 which supports the astrophotonics programme at the university of sydney .
allington - smith j. r. , content r. , haynes r. , lewis i.j . , 1997 ,
spie 2871 , 1284 bland - hawthorn j. et al .
, 2011 , optics express , 19 , 2649 colless m. m. et al . , 2001 , mnras , 328 , 1039 eisenhauer f. et al . , 2003 , proc .
spie 4841 , 1548 ellis s. c. , driver s. p. , allen p. d. , liske j. , bland - hawthorn j. , de propris r. , 2005 , mnras , 363 , 1257 faber s. m. et al . , 2003 ,
spie 4841 , 1657 le fevre o. , et al .
, 2003 , proc .
spie 4841 , 1670 mcgregor p. j. , et al .
, 2003 , proc .
spie 4841 , 1581 murray g. j. , et al . , 2000 , proc .
spie 4008 , 611 tody d. , 1986 , proc .
spie 627 , 733 york d. g. , et al . , 2000 ,
aj , 120 , 1579
|
new multi - core imaging fibre bundles hexabundles being developed at the university of sydney will provide simultaneous integral field spectroscopy for hundreds of celestial sources across a wide angular field .
these are a natural progression from the use of single fibres in existing galaxy surveys .
hexabundles will allow us to address fundamental questions in astronomy without the biases introduced by a fixed entrance aperture .
we have begun to consider instrument concepts that exploit hundreds of hexabundles over the widest possible field of view . to this end , we have compared the performance of a 61-core fully - fused hexabundle and 5 lightly - fused bundles with 7 cores each .
all fibres in the bundles have @xmath0 m cores . in the fully - fused bundle ,
the cores are distorted from a circular shape in order to achieve a higher fill fraction .
the lightly - fused bundles have circular cores and five different cladding thicknesses which affect the fill fraction .
we compare the optical performance of all 6 bundles and find that the advantage of smaller interstitial holes ( higher fill fraction ) is outweighed by the increase in modal coupling , cross - talk and the poor optical performance caused by the deformation of the fibre cores .
uniformly high throughput and low cross - talk are essential for imaging faint astronomical targets with sufficient resolution to disentangle the dynamical structure
. devices already under development will have between 100 and 200 lightly - fused cores , although larger formats are feasible .
the light - weight packaging of hexabundles is sufficiently flexible to allow existing robotic positioners to make use of them .
[ firstpage ] instrumentation : miscellaneous : hexabundles techniques : miscellaneous methods : observational instrumentation : spectrographs techniques : imaging spectroscopy .
|
using the input - output formalism @xcite , we define @xmath18 the operator describing the incident @xmath14-polarized field , @xmath25 and @xmath26 the operators describing the @xmath13- and @xmath14-polarized output fields .
we also define the dimensionless intracavity fields @xmath98 and @xmath99 for both polarizations .
the continuity relations between these operators write : @xmath100 with @xmath76 the cavity intensity damping rate through the top mirror . + from these operators , one can define the output operators @xmath101 and @xmath102 in the diagonal / antidiagonal basis , @xmath103 and @xmath104 in the left - handed / right - handed circular basis : @xmath105 with these notations the intensity @xmath106 , i.e. the number of output photons per unit time in polarization @xmath107 ( with @xmath108 ) , is given by @xmath109 .
this intensity is generally higher than @xmath110 , which only describes the coherent part of the @xmath107-polarized output field .
we schematize in fig .
s1(a ) the three - level system of the neutral exciton confined in an anisotropic qd : the system eigenstates are the ground state @xmath19 and two excited states spectrally separated by the qd fine structure splitting @xmath111 .
these eigenstates @xmath71 and @xmath72 can be optically addressed with @xmath22-polarized and @xmath23-polarized light , respectively , @xmath22 and @xmath23 being the eigenaxes of the qd anisotropy . in our experiment ,
however , the important axes are given by the cavity anisotropy , corresponding to vertical ( @xmath14 ) and horizontal ( @xmath13 ) polarizations : the @xmath14 and @xmath13 eigenaxes generally differ from the qd eigenaxes @xmath22 and @xmath23 by an angle @xmath24 , as illustrated in fig .
s1(b ) @xcite .
[ fig_s1 ] fig . s1 . * ( a ) * quantum dot three level system . *
( b ) * relative orientation of cavity and qd eigenaxes . to describe the system we thus use the excitonic states @xmath21 and
@xmath20 that can be optically addressed with @xmath14- and @xmath13-polarized light : @xmath112 the frequency of the qd state @xmath113 , with @xmath114 , we find that : @xmath115 this allows writing the hamiltonian @xmath116 of the qd system in the @xmath117 , @xmath118 basis , in the frame rotating at the laser frequency @xmath119 : @xmath120 with @xmath121 the lowering operator @xmath122 and @xmath123 the corresponding raising operator .
+ we now turn to the hamiltonian @xmath124 of the cavity system . denoting @xmath125 the frequency of the @xmath107-polarized cavity mode , with @xmath126 , we find : @xmath127 denoting @xmath128 , we also find the hamiltonian @xmath129 describing the cw resonant excitation by the @xmath14-polarized input field : @xmath130 finally , denoting @xmath131 the light - matter coupling strength , the light - matter interaction hamiltonian @xmath132 writes : @xmath133
we model the irreversible interaction between the cavity - qed system and its external environment by a number of dissipative processes .
each process is described by a collapse operator @xmath134 , which participates to the evolution of the system density matrix @xmath135 via a limbladian super - operator @xmath136 @xcite : @xmath137=\frac{1}{2}\left(2\hat{c}\rho\hat{c}^{\dagger}-\hat{c}^{\dagger}\hat{c}\rho-\rho\hat{c}^{\dagger}\hat{c}\right)\ ] ] + the collapse operators associated to our dissipative processes are the following : * _ cavity damping _ , associated to the cavity optical losses , is described by the collapse operators @xmath138 , with @xmath126 and @xmath77 the total cavity damping rate .
* _ quantum dot spontaneous emission in the leaky modes _ , associated to direct spontaneous emission of a photon into the external environment , is described by the collapse operators @xmath139 , with @xmath126 and @xmath84 the qd spontaneous emission rate . *
_ excitonic pure dephasing _ , associated to the decoherence processes which preserve the population of the excited state @xmath140 and @xmath141 , is described by a collapse operator @xmath142 , with @xmath143 the pure dephasing rate .
these operators allow defining the corresponding limbladian super - operators @xmath144 , @xmath145 , @xmath146 , @xmath147 and @xmath148 . using the total system hamiltonian @xmath149
, we deduce the complete master equation describing the system evolution : @xmath150 \:+\:\hat{l}_h^{\textrm{cav}}\left[\rho\right]\:+\:\hat{l}_v^{\textrm{cav}}\left[\rho\right]\nonumber \\ & & \ : + \:\hat{l}_h^{\textrm{sp}}\left[\rho\right]\:+\:\hat{l}_v^{\textrm{sp}}\left[\rho\right]\:+\:\hat{l}^{\textrm{deph}}\left[\rho\right]\end{aligned}\ ] ] + finally , when describing cw experiments this master equation is numerically solved to deduce the stationary state density matrix , @xmath151 , from which average values such as @xmath152 can be deduced .
the master equation can also be used to describe second - order correlations such as @xmath153 , as was used for the numerical prediction of @xmath154 in fig .
2b ( see main text ) .
analyzing the reflected output from a cavity - qed device requires distinguishing between the light coupled into the cavity mode and the residual uncoupled light .
we define @xmath155 the input - coupling efficiency , i.e. the overlap between the spatial profiles of the incoming beam and of the fundamental cavity mode . for any polarization
@xmath107 the total input field intensity @xmath156 can then be divided in two contributions , namely @xmath157 and @xmath158 , respectively associated to coupled and uncoupled light . similarly , the output intensity in a given polarization @xmath107 is the sum of @xmath159 , denoting the contribution from light coupled to the mode , and @xmath160 for uncoupled light @xcite .
these two contributions can be computed separately : * _ coupled light_. the intensity @xmath159 in any polarization @xmath107 is described by @xmath161 , directly deduced from the stationary density matrix @xmath151 , taking into account a coupled input intensity @xmath162 . *
_ uncoupled light_. the intensity @xmath160 in any polarization @xmath107 is described by @xmath163 , i.e. the total amount of uncoupled light .
indeed , in our experimental configuration , the spatial profile of the incoming beam is slightly smaller than the micropillar top surface , allowing all the uncoupled light to be fully reflected by the optically flat surface .
furthermore , this reflection on the micropillar surface does not rotate the polarization of light , allowing @xmath160 to be proportional to the input intensity @xmath156 for any polarization @xmath107 . in our experiment
, the residual uncoupled light stays @xmath14-polarized and thus plays a role in the tomography measurement described in fig .
3a ; it also adds a constant contribution to the intensity @xmath41 described in fig .
the numerical fit presented in these figures takes this residual effect into account , with a best estimated value of @xmath164 of uncoupled light .
4ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty https://books.google.fr/books?id=liwsc3nlf0kc[__ ] ( , ) \doibase doi:10.1038/ncomms11986 [ * * , ( ) ] @noop _ _ , vol .
( , ) link:\doibase 10.1103/physrevlett.109.166806 [ * * , ( ) ]
|
we introduce a tomography approach to describe the optical response of a cavity quantum electrodynamics device , beyond the semiclassical image of polarization rotation , by analyzing the polarization density matrix of the reflected photons in the poincar sphere . applying this approach to an electrically - controlled quantum dot - cavity device
, we show that a single resonantly - excited quantum dot induces a large optical polarization rotation by 20@xmath0 in latitude and longitude in the poincar sphere , with a polarization purity remaining above 84@xmath1 .
the quantum dot resonance fluorescence is shown to contribute to the polarization rotation via its coherent part , whereas its incoherent part contributes to degrading the polarization purity . in the development of quantum photonic networks ,
a crucial challenge is to demonstrate highly - efficient interfaces allowing the coherent transfer of quantum information between a stationary qubit and a flying one @xcite . in this context
, it has been shown that a single natural or artificial atom in a cavity quantum electrodynamics ( cqed ) system can induce giant phase shifts on incoming photons @xcite : this allowed the recent implementation of atom - photon gates @xcite and , subsequently , photon - photon gates @xcite .
most of these achievements are based on polarization - encoding protocols @xcite , in which the optical phase shift is used to rotate the polarization state of the outgoing photons .
polarization rotation is also at the heart of numerous quantum computation proposals based on spin - photon logic gates , where a stationary spin qubit is used as a quantum memory with a long coherence time @xcite .
by essence , these concepts of optical phase shift and polarization rotation rely on a semiclassical image , considering that the optical response of the atom - based device is entirely coherent . yet
, the output photonic field includes a contribution from the resonance fluorescence emitted by the natural or artificial atom , which can be partially incoherent with respect to the incoming laser @xcite . in quantum optics this is described by an output field operator @xmath2 which can not be reduced to its expectation value @xmath3 .
the coherent contribution is governed by the average field @xmath4 , which keeps a defined phase with respect to the incoming field ; the incoherent component , however , arises from quantum fluctuations around the average field , and has no specific phase @xcite .
a direct consequence of this is that the optical response of a device can not , in general , be solely described by a reflection coefficient with a well - defined amplitude and phase .
similarly , a general output polarization is not necessarily a pure polarization state .
the distinction between coherent and incoherent responses is especially important in solid - state quantum technologies , where artificial atoms , such as semiconductor quantum dots ( qds ) , interact with a fluctuating environment .
as was experimentally demonstrated using spectrally - resolved @xcite or interferometry - based @xcite measurements , the fluorescence emitted by a resonantly - excited qd has its coherent fraction at best equal to @xmath5 , with @xmath6 the lifetime of the photon wavepacket and @xmath7 its coherence time @xcite .
qd - based cqed structures have also been used to demonstrate qd - induced optical phase shifts up to a few tens of degrees @xcite , and a spin - dependent polarization rotation of up to @xmath8 @xcite .
none of the reported works , however , have gone beyond the semiclassical image to describe the global optical response of a cqed device , where the response of the cavity itself is superposed to the extracted resonance fluorescence . here
we introduce a polarization tomography approach to investigate the polarization rotation of coherent light interacting with an electrically - controlled qd - cavity device @xcite .
we analyze the polarization density matrix in the poincar sphere @xcite , which allows distinguishing between a general mixed polarization and a pure polarization state .
we show that the superposition of emitted single photons ( h - polarized ) with reflected photons ( v - polarized ) leads to a large rotation of the output polarization , by 20@xmath0 both in latitude and longitude in the poincar sphere @xcite , with a polarization purity above 84 @xmath1 .
we implement a cqed model with which we analyze the complete information set provided by the polarization tomography .
we demonstrate that the coherent part of the qd emission contributes to the polarization rotation , while its incoherent part contributes to degrading the polarization purity . .
] we study a deterministically - coupled , electrically controlled qd - cavity system .
the sample , grown by molecular beam epitaxy , consists in a @xmath9-gaas cavity embedding ingaas qds , surrounded by two distributed bragg reflectors ( gaas and al@xmath10ga@xmath11as , with 30 and 20 pairs for the bottom and top mirrors ) .
in - situ lithography was used to deterministically fabricate a micropillar cavity around a single ingaas qd @xcite .
the 3 @xmath12 diameter micropillar is connected by four ridges to a circular frame where metallic ohmic contacts are defined ( fig .
[ fig : fig1](a ) ) .
this cavity presents a small anisotropy leading to linearly polarized modes considered as @xmath13 ( horizontal ) and @xmath14 ( vertical ) polarizations , with a @xmath15 @xmath16ev splitting .
the metallic contacts allow the electrical tuning of the qd transition with respect to these cavity modes @xcite . the experimental setup is sketched in fig .
[ fig : fig1](b ) : the device is kept inside a helium exchange gas cryostat at @xmath17k , and a tunable continuous - wave laser with 1mhz linewidth excites the cavity from the top mirror . the reflected beam
can be separated in two orthogonally - polarized components in various polarization bases , using calibrated waveplates and a wollaston polarizing prism .
the input and output field intensities can then be measured with avalanche photodiodes ( apds ) , as displayed in fig .
[ fig : fig1](b ) , or analyzed via spectral or autocorrelation measurements ( not shown ) . as sketched in fig .
[ fig : fig1](c ) , the experiment we perform consists in exciting the device with a @xmath14-polarized input field ( denoted @xmath18 ) , corresponding to one of the cavity eigenaxes . in the case of a fully - detuned qd
this would imply a @xmath14-polarized reflected beam ; yet , the interaction with a qd optical transition leads to a more complex output .
indeed , the neutral qd can be described as a three - level system with a ground state @xmath19 and two excitonic states @xmath20 and @xmath21 , that can respectively be excited with @xmath14- and @xmath13-polarized light .
however , @xmath21 and @xmath20 are not the system eigenstates , as the qd displays an anisotropy along two axes @xmath22 and @xmath23 , differing from @xmath13 and @xmath14 by an angle @xmath24 @xcite . as a consequence
, an initially excited state @xmath20 coherently oscillates between @xmath21 and @xmath20 ( fig .
[ fig : fig1](c ) ) . by exciting the system with
@xmath14 polarized light one thus populates the @xmath20 state , which rotates into @xmath21 : this leads to resonance fluorescence emitted both in @xmath14 and @xmath13 polarizations . as sketched in fig .
[ fig : fig1](c ) , polarization tomography provides a global analysis of the @xmath14- and @xmath13-polarized output fields , denoted @xmath25 and @xmath26 .
in particular , if both the @xmath13- and @xmath14 polarized output fields are coherent with the incoming laser , they superpose with a well - defined phase , resulting in a pure polarization state .
we first analyze the device in the cavity eigenbasis @xmath27 . setting the polarization analyzer
to separate the @xmath13 and @xmath14 polarizations , we measure the spectrum of the @xmath13-polarized output field with a spectrometer coupled to a ccd camera , filtering out the @xmath14-polarized light .
fig .
[ fig : fig2](a ) displays the @xmath13-polarized signal as a function of the bias voltage applied to the device and of the emission wavelength , for a fixed laser wavelength @xmath28 nm ( in resonance with the bias - independent @xmath14-polarized cavity mode ) and a fixed incoming power @xmath29 pw ( in the low - power regime where no saturation of the qd excitonic transition is observed ) .
when the bias is tuned to @xmath30v , which tunes the qd transition wavelength @xmath31 in resonance with @xmath32 , the emitted intensity is strongly increased as expected for a resonance fluorescence process . the logarithmic scale used in fig .
[ fig : fig2](a ) also allows observing two residual features at biases above @xmath33v : one corresponding to residual light at the laser wavelength , and one corresponding to raman - assisted qd emission .
the @xmath13-polarized signal arises from cross - polarized resonance fluorescence , i.e. from single photons emitted by the resonantly - driven qd .
this is evidenced by the measurement of the second - order autocorrelation of the @xmath13-polarized field , using a standard hanbury - brown - twiss experiment with two single - photon avalanche diodes @xcite , with @xmath34 both fixed .
fig .
[ fig : fig2](b ) displays the corresponding histogram of @xmath35 , with @xmath36 the delay between photon detection events .
the raw value of @xmath37 decreases down to @xmath38 , which is compatible with a single - photon emission by the @xmath21 state if one takes into account the finite time response of the single - photon diodes . as is also displayed in fig .
[ fig : fig2](b ) , a good agreement is found between the experimental data and our numerical prediction , obtained with cqed parameters that will be discussed later on .
-polarized resonance fluorescence as function of the bias voltage and emission energy , measured for @xmath28 nm . *
( b ) * autocorrelation measurements of the @xmath13-polarized signal , measured for @xmath39 nm , and corresponding numerical prediction taking into account the finite detector response time * ( c ) * normalized intensities in @xmath13 and @xmath14 polarizations , measured for @xmath40 nm and a varying @xmath32 , together with numerical predictions . ]
we then measure the intensities of the @xmath14- and @xmath13-polarized components of the output field , denoted @xmath41 and @xmath42 . in the following we work at a fixed bias of @xmath43 v , fixing @xmath44 to be in resonance with the @xmath14-polarized cavity mode , while scanning the laser wavelength . fig .
[ fig : fig2](c ) displays the evolution of the normalized intensities @xmath45 and @xmath46 , with @xmath47 the input field intensity , as a function of @xmath32 .
a peak in @xmath46 is observed at @xmath34 , indicating that up to 7@xmath1 of the incident light has been reconverted by the quantum dot into @xmath13-polarized resonance fluorescence .
the @xmath14-polarized output intensity also presents a peak at @xmath31 , though superposed to the broader reflectivity dip of the @xmath14-polarized cavity mode .
this behavior is understood by considering that the @xmath14-polarized light arises from the superposition of two fields : light directly reflected upon the top - mirror surface , and light extracted from the cavity via the top - mirror @xcite .
the high qd - induced peak is thus related to the @xmath14-polarized resonance fluorescence , emitted by the @xmath20 exciton , which is large enough to strongly change the amount of intracavity light , and thus the resulting output field . to perform a complete polarization tomography we now analyse our device in the other two bases . by adjusting the waveplates of the polarization analyzer introduced in fig .
[ fig : fig1](b ) , we measure the intensities @xmath48 and @xmath49 in the diagonal / anti - diagonal polarization basis , and @xmath50 and @xmath51 in the right - handed / left - handed circular polarization basis . for a given set of intensities @xmath52 , we define the corresponding stokes component as @xmath53 .
this allows measuring the density matrix of the polarization state , and representing it in the poincar sphere as a vector whose coordinates are @xmath54 , @xmath55 and @xmath56 , ranging between @xmath57 and @xmath58 .
the purity of the polarization density matrix is given by the norm of the poincar vector , @xmath59 .
this norm is equal to 1 only for a pure polarization state , as would be given by a coherent superposition of @xmath13- and @xmath14-polarized electromagnetic fields . , using the experimental conditions of fig.[fig : fig2](c ) .
markers : experimental data .
solid lines : numerical fit ( see text ) .
( b ) representation of the polarization state in the poincar sphere for both experimental data ( circles ) and numerical simulations ( continuous line ) .
each point is colored according to its corresponding polarization purity . ] in fig .
[ fig : fig3](a ) the three stokes components @xmath54 , @xmath55 and @xmath56 , as well as the polarization purity @xmath59 , are displayed as a function of @xmath32 .
the same set of data is also illustrated in fig . [
fig : fig3](b ) , where the stokes components are used as three dimensional cartesian coordinates on the poincar sphere , for the various values of @xmath32 .
far from the qd resonance we obtain @xmath60 together with @xmath61 and @xmath62 .
this corresponds to a @xmath14-polarized reflected field , as also illustrated by the accumulation of experimental points at the @xmath14 polarization in the poincare sphere . at resonance a maximal value of @xmath63
is obtained , and at the same time the stokes component @xmath55 and @xmath56 become non - zero in the region of the qd resonance . as seen in the poincar sphere
this corresponds to a rotation between 0 and more than 20 degrees , both in longitude and latitude , when @xmath32 is tuned around @xmath40 nm .
these results , translated in terms of angles in the polarization ellipse , correspond to a rotation of the orientation and ellipticity angles of more than 10 degrees .
they do not mean , however , that the output polarization is a pure polarization state : indeed a degradation of the polarization purity is observed around the qd resonance , though the purity remains above 84@xmath1 ( fig .
[ fig : fig3](a ) , top panel ) .
this is also illustrated in the poincar sphere of fig .
[ fig : fig3](b ) , where each experimental point is colored following a scale indicating the corresponding polarization purity , i.e. the distance between the data point and the center of the sphere .
we now analyze our experimental results based on the model sketched in fig .
[ fig : fig1](c ) , using the input - output formalism @xcite with a single input operator @xmath18 and two output operators @xmath25 and @xmath26 @xcite .
this allows computing the intensities @xmath64 and @xmath65 in any polarization basis , which gives access to the three stokes components and to the expected polarization purity .
fitting simultaneously the experimental curves in figures [ fig : fig2](c ) and [ fig : fig3](a ) allows determining all the parameters at play , more precisely than what was previously possible with the analysis of optical nonlinearities in qd - cavity devices @xcite .
we first extract a qd - cavity light - matter coupling @xmath66 @xmath16ev , a total cavity damping rate @xmath67 @xmath16ev for both @xmath13 and @xmath14 modes , and a total qd decoherence rate @xmath68 @xmath16ev ( using @xmath69 units ) . the qd anisotropy is also found to be characterized by a fine structure splitting @xmath70 @xmath16ev between the excitonic eigenstates @xmath71 and @xmath72 , with the @xmath73 axes rotated by an angle @xmath74 with respect to the @xmath13/@xmath14 cavity axes .
the top - mirror output - coupling @xmath75 , which is the ratio between the top - mirror damping rate @xmath76 and the total damping rate @xmath77 , is best fitted at @xmath78 .
if the qd resonance fluorescence field were entirely coherent with respect to the incoming laser , the output light would be completely described by only two expectation values @xmath79 and @xmath80 , each having a well defined amplitude and phase : this would lead to a coherent , classical superposition of @xmath13- and @xmath14- polarized fields , i.e. a pure polarization state . yet
, partial incoherence has to be considered to interpret the reduced polarization purity : in the numerical fits of fig . [
fig : fig3](a ) , this partial incoherence arises from the residual pure dephasing experienced by the exciton , whose rate is estimated at @xmath81 @xmath16ev .
we attribute this dephasing to fluctuations of the qd transition energy , which could be induced by a residual current at the bias of @xmath43v , and/or by fluctuations of the helium flow in our exchange gas cryostat . taking into account that the total qd decoherence rate @xmath82 is given by @xmath83 , with @xmath84 the rate of spontaneous emission in the external ( leaky ) optical modes ,
we determine that @xmath85 @xmath16ev , consistent with a 1 ns leak emission time .
we also note that the same set of parameters has been used to provide the numerical predictions in fig .
2(b ) , without any additional fitting , in agreement with the measured autocorrelation function . .
same parameters as those used in fig .
[ fig : fig3 ] except @xmath86 , @xmath87 and @xmath88 .
each panel corresponds to a different intracavity photon number @xmath89 , where @xmath89 is the intracavity photon number that is obtained at resonance in the empty - cavity regime , deduced from @xmath90 , and contains two curves corresponding to the exact calculation ( ex ) , and to the semiclassical approximation ( sc ) .
each point is colored according to its corresponding polarization purity . ]
pure dephasing , however , is not the only source of fluctuations leading to a partially incoherent output : quantum fluctuations around the expectation values @xmath91 and @xmath92 are also obtained in the intermediate - power regime , i.e. when the excitation results in a population of the excitonic transition @xcite .
we theoretically study , in fig .
[ fig : fig4 ] , the polarization tomography of an optimized device , with a pure dephasing rate @xmath86 , a top - mirror output - coupling @xmath88 and a qd - cavity eigenaxis angle @xmath87 ( the other parameters being unchanged ) .
the calculation is performed for various incoming powers @xmath931 , 100 , 500 pw and 2 nw , corresponding to different values of the intracavity photon number @xmath89 indicated in figs [ fig : fig4](a - d ) . in each panel
two numerical calculations are compared : the exact one considering that the intensities @xmath64 and @xmath65 , in an arbitrary basis , are given by @xmath94 and @xmath95 , and the semiclassical one neglecting the incoherent part of the output fields , i.e. considering that @xmath96 and @xmath97 . at low incoming power , in fig .
[ fig : fig4](a ) , the semiclassical approximation gives the same result as the exact calculation .
maximal polarization rotations are obtained and the qd transition even allows converting the @xmath14-polarized incoming light into a fully @xmath13-polarized output , as required for a photon - photon gate @xcite . in this situation
the output polarization vectors stay at the surface of the sphere , as expected for fully coherent fields . at higher incoming powers
, however , the incoherent resonance fluorescence emitted by the quantum dot modifies this situation .
as displayed in figs [ fig : fig4](b - d ) , the exact calculation predicts a significant degradation of the polarization purity at increasing powers . at higher powers
the quantum dot transition becomes saturated and the polarization purity is restored , as would be the case with an empty cavity . in figs .
[ fig : fig4](b - d ) we note the strong discrepancy between the exact calculation and the semiclassical approximation , which only predicts coherent outputs and , thus , pure polarization states .
the semiclassical image should thus , by no means , be considered as a complete description of the optical response for a cqed device ; it can be fruitfully used , however , in the absence of pure dephasing and for experiments in the low - excitation regime . in summary
, polarization tomography allows probing the optical response of a cqed device beyond the semiclassical approximation .
we have shown that the superposition of @xmath13-polarized single photons with a @xmath14-polarized reflected field can be partially coherent , leading to a polarization rotation whose degree of purity can be decreased by dephasing processes . the tomography approach will provide a comprehensive measurement tool for polarization - encoded protocols : in particular , a single electron spin described in the bloch sphere can be monitored by , or entangled with , a photon polarization qubit described in the poincar sphere .
this would constitute the heart of future quantum networks , where cqed devices interact by exchanging polarization - encoded photonic qubits @xcite .
this work was partially supported by the erc starting grant no .
277885 qd - cqed , the french agence nationale pour la recherche ( qdom : anr-12-bs10 - 0010 and spiqe : anr-14-ce32 - 0012 ) , the labex nanosaclay , and the renatech network .
37ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty https://books.google.fr/books?id=qj8wcgaaqbaj[__ ] ( , ) link:\doibase 10.1103/physrevlett.75.4710 [ * * , ( ) ]
link:\doibase 10.1088/1464 - 4266/5/3/304 [ * * , ( ) ] link:\doibase 10.1126/science.1154643 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.112.043601 [ * * , ( ) ] @noop * * , ( ) http://dx.doi.org/10.1038/nature13177 [ * * , ( ) ] http://www.nature.com/nnano/journal/v11/n6/abs/nnano.2015.334.html [ * * , ( ) ] link:\doibase 10.1126/science.1254699 [ * * , ( ) ] link:\doibase 10.1038/nature18592 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.92.127902 [ * * , ( ) ] link:\doibase 10.1103/physrevb.73.075312 [ * * , ( ) ] link:\doibase 10.1103/physrevb.78.125318 [ * * , ( ) ] link:\doibase 10.1103/physrevb.78.085307 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.104.160503 [ * * , ( ) ] link:\doibase 10.1103/physrevb.92.115305 [ * * , ( ) ] https://books.google.fr/books?id=tcp-ngeacaaj[__ ] , a wiley - interscience publication ( , ) link:\doibase 10.1103/physrevlett.108.093602 [ * * , ( ) ] link:\doibase 10.1126/sciadv.1501256 [ * * , ( ) ] link:\doibase s [ * * ( ) , s ] http://www.nature.com/nphoton/journal/v10/n3/abs/nphoton.2015.276.html [ * * , ( ) ] link:\doibase 10.1103/physreva.84.011803 [ * * , ( ) ] link:\doibase 10.1103/physrevb.91.241305 [ * * , ( ) ] http://dx.doi.org/10.1038/ncomms7236 [ * * , ( ) ] http://dx.doi.org/10.1038/nphoton.2016.23 [ * * , ( ) ] \doibase doi:10.1038/ncomms11986 [ * * , ( ) ] @noop ( ) , @noop _ _ ( ) @noop link:\doibase 10.1103/physrevlett.101.267404 [ * * , ( ) ] link:\doibase 10.1038/ncomms4240 [ * * , ( ) ] @noop link:\doibase 10.1126/science.290.5500.2282 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.109.166806 [ * * , ( ) ] https://books.google.fr/books?id=liwsc3nlf0kc[__ ] ( , ) link:\doibase 10.1103/physreva.75.053823 [ * * , ( ) ] link:\doibase 10.1364/ol.40.003173 [ * * , ( ) ] * supplemental materials : tomography of optical polarization rotation induced by a quantum dot - cavity device *
|
the structure of the vacuum plays a key role in understanding physical phenomena .
for example , qed and the asymptotic freedom phase of non - abelian gauge theory are based on the trivial vacuum @xmath0 and the physical vector fields , the transverse photon / gluon , are clearly exposed by using the coulomb gauge .
confinement , on the other hand , is believed to be a non - perturbative vacuum phenomenon that is not yet clearly understood .
presently , there are two confinement mechanisms discussed extensively in the literature .
these are the magnetic monopoles@xcite and vortices@xcite , which are related to the maximal abelian gauge@xcite and maximal central gauge@xcite , respectively .
although each of the two mechanism has its own set of successes and failures , they share an important shortcoming : they are disjointed from short - distance physics .
as a quark is pulled from inside a hadron , it goes through the asymptotic phase ( short - distance , where transverse gluons are exchanged ) to the confinement phase ( large - distance , with mass gap and linearly rising interaction ) eventually resulting in hadronization . ideally then
, the asymptotic phase and the confinement phase should continuously interpolate in a way where the effective degrees of freedom in each distance regimes are transparent . on the lattice
, this problem is dealt with using a decimation procedure.@xcite for a number of years now , this author has been claiming that this is achieved in the continuum by using the non - linear gauge .
the non - linear gauge@xcite , which is given by @xmath1 was proposed by the author eleven years ago because of the observation that field configurations that satisfy @xmath2 and equation ( [ eq:1 ] ) can not be gauge - transformed to the coulomb gauge .
this follows from the fact that the zero mode of @xmath3 , which is @xmath4 , is also the source in the equation that solves for the gauge parameter . and
conversely , field configurations on the coulomb surface can not be gauge transformed to the non - linear regime of equation ( [ eq:1 ] ) , i.e. , there is no @xmath5 such that @xmath6 , @xmath7 , and @xmath8 .
these mean that the two regimes of equation ( [ eq:1 ] ) , which are ( i ) the linear regime with @xmath9 , and ( ii ) the non - linear regime , which effectively is the gribov horizon of the surface @xmath10 , do not mix . thus , the coulomb gauge is an incomplete gauge - fixing for non - abelian theory , although there are claims to the contrary.@xcite using the running of the coupling constant , the author argued that the non - linear gauge interpolates between short - distance ( perturbative ) physics and large - distance ( non - perturbative ) physics . in short - distance regime , where the coulomb gauge is in effect , the transverse gluons are the physical degrees of freedom .
however , in the large - distance regime , the vector field @xmath11 and the scalar field @xmath12 , which satisfy two sets of constraints , are the effective degrees of freedom.@xcite the full quantum dynamics of the scalars lead to a parisi - sourlas mechanism@xcite with an @xmath13 symmetry@xcite , which unfortunately has a wrong sign for the kinetic term of the scalars .
thus , although we have shown equivalence to a 2d o(1,3 ) non - linear sigma model , the proof of confinement remains formal because aside from the wrong sign of the scalar kinetic term , the mechanism is not identified and quarks and `` gluons '' are not involved . to identify the mechanism
, the author proposes that we look for ( a ) classical solution(s ) of the pure @xmath12 dynamics which is ( are ) vacuum configuration(s ) .
we then expand on this ( these ) configuration(s ) and see the physics of quarks , `` gluons '' and scalar fluctuations around the background .
this program was started by the author about five years ago and the following results had been derived .
first , that all spherically symmetric functions @xmath14 , with @xmath15 , in @xmath16 are classical solutions with zero field strength.@xcite this shows that all the @xmath14 are vacuum configurations with zero action , which means the configuration space has a very broad minimum .
second , since we do not have just one classical configuration but a whole class of solutions , the author proposed to treat them as a stochastic background with a white - noise distribution .
this led to an area law behaviour for the wilson loop.@xcite thus , we have identified a possible mechanism for confinement .
third , when we consider the classical dynamics of the `` gluons '' in the large distance regime , we find the gluons acquiring a mass.@xcite this verifies the existence of a mass gap .
furthermore , the `` gluons '' also lose their self - interaction .
this result was also arrived at by kondo by a suitable redefinition of fields.@xcite in a forthcoming paper@xcite , which will provide the details of the results presented here and discuss quantum field theory in a stochastic background , the author will show that the scalars @xmath17 , which are fluctuations off the classical stochastic background @xmath14 , @xmath18 are non - propagating , i.e. , they do not have a kinetic term . as a matter of fact , we will show that @xmath19 is independent of the scalars @xmath20 and only dependent on @xmath11 .
this resolves the problem of scalars with a wrong sign for the kinetic term encountered in the derivation of the parisi - sourlas mechanism , where the full quantum theory of the scalars was shown to be equivalent to a 2d o(1,3 ) non - linear @xmath21 model . in this paper
, we will present the quantum dynamics of quarks and `` gluons '' in a stochastic background , resulting mainly in the derivation of a non - local four - fermi interaction with a linear potential .
in this section , we will provide the background needed to derive the main result which will be discussed in section 3 .
consider @xmath22 theory with action @xmath23=\int\mathcal{l}d^{4}x\ ] ] the gauge - fixing given by equation([eq:1 ] ) leads to the resolution of the potential in the non - linear regime as given by @xmath24 the scalars @xmath12 and the `` gluon '' @xmath11 satisfy the following constraints , which guarantee the same number of degrees of freedom , @xmath25\partial_{\mu}f^{b}t^{c}_{\mu}=0.\label{eq:6}\end{aligned}\ ] ] substituting equation ( [ eq:4 ] ) in the field strength , we find @xmath26,\label{eq:11}\\ r^{ab}&=&\frac{1}{(1+\vec{f}\cdot\vec{f})}(\delta^{ab}+\epsilon^{abc}f^{c}+f^{a}f^{b}),\label{eq:12}\\ y^{abc}&=&\frac{1}{(1+\vec{f}\cdot\vec{f})^{2}}[-(\vec{f}\cdot\vec{f})\epsilon^{abc}+(1+\vec{f}\cdot\vec{f})f^{a}\delta^{bc}-(1-\vec{f}\cdot\vec{f})\delta^{ac}f^{b}\nonumber\\ & + & 3\epsilon^{cad}f^{d}f^{b}-2f^{a}f^{b}f^{c}+\epsilon^{abd}f^{d}f^{c}+f^{a}\epsilon^{bcd}f^{d}],\label{eq:13}\\ t^{abc}&=&\frac{1}{(1+\vec{f}\cdot\vec{f})}[\epsilon^{abc}+(1+\vec{f}\cdot\vec{f})f^{b}\delta^{ac}-(1+\vec{f}\cdot\vec{f})\delta^{ab}f^{c}\nonumber\\ & + & \epsilon^{abd}f^{d}f^{c}+f^{a}\epsilon^{bcd}f^{d}+\epsilon^{adc}f^{d}f^{b}]\label{eq:14}.\end{aligned}\ ] ] the pure @xmath12 dynamics given by the action @xmath27 hints of non - perturbative physics because ( i ) of the @xmath28 factor that goes with @xmath29 , ( ii ) the action is infinitely non - linear , and ( iii ) @xmath30 .
for these reasons , we look for a classical solution to @xmath31 from equation ( [ eq:8 ] ) and ( [ eq:11 ] ) , we find that all spherically @xmath14 , with @xmath15 are classical solutions with zero field strength @xmath32 .
this follows from @xmath33 and the anti - symmetry of @xmath34 with respect to b and c. thus , we are dealing with a whole class of classical solutions with zero field strength .
this shows that the action has a very broad bottom ( note @xmath29 is positive semi - definite ) . since we do not have just one solution but a whole class of solutions , it was proposed in reference ( 7 ) to treat @xmath14 as a stochastic background with a white - noise distribution @xmath35=\mathcal{n}exp.(-\frac{1}{\ell}\int^{\infty}_{0}\tilde{f}^{a}(s)\tilde{f}^{a}(s)ds).\ ] ] we identify @xmath36 as the length scale where non - perturbative physics becomes important .
from the running of the coupling constant , we deduce @xmath37 . substituting the background decomposition given by equation ( [ eq:2 ] ) in @xmath38
, we will get a rather involved expression for the action of the `` gluon '' @xmath11 and the scalars @xmath20 in the presence of the background @xmath39 .
however , as we show in reference ( 18 ) , all the stochastic averages drop out except the following simple result @xmath40 this result shows that the scalars @xmath20 are non - propagating and the `` gluons '' @xmath11 acquired a mass and lost their self - interactions . the stochastic average of the fermionic term , on the other hand , yields @xmath41\psi(x).\ ] ] we can remove the scalars from equation ( [ eq:19 ] ) by using the stochastic average of the constraints given by equations ( [ eq:5 ] ) and ( [ eq:6 ] ) , which are @xmath42 substituting equation ( [ eq:20 ] ) in equation ( [ eq:19 ] ) , we find that stochastically averaged effective dynamics yield the following for fermi interaction @xmath43 the greens function @xmath44 , which is the propagator of the `` gluons '' as given by @xmath45 where @xmath46 and @xmath47 satisfies instead of @xmath48 in field equation . furthermore ,
since the scalar fluctuations @xmath20 where neglected , the gluons satisfied the constraint @xmath49 , instead of equation [ eq:20 ] , thus resulting in an inconsistent massive but transverse gluons in reference @xcite .
this is now corrected in this paper . ]
@xmath50 the solution to equation ( [ eq:24 ] ) for large @xmath51 is@xcite @xmath52 substituting equations ( [ eq:25 ] ) , ( [ eq:23 ] ) in equation ( [ eq:22 ] ) , we find that the effective four - fermi term from the `` gluon '' exchange does not confine because the exponential decay behaviour dominates .
the confining non - local four - fermi term must come from elsewhere , which must involve the vacuum configuration at the outset .
the path - integral in the non - linear regime of the non - linear gauge is given by@xcite @xmath53 where @xmath54 is given by equation ( [ eq:6 ] ) and @xmath55 is the dimension 4 `` fadeev - popov '' operator@xcite given by @xmath56 since we will do a background decomposition as given by equation ( [ eq:2 ] ) to find the vacuum to vacuum functional in the presence of @xmath14 and then do a stochastic averaging , we will change the delta functionals by @xmath57 where the powers @xmath58 and @xmath59 can be freely chosen so that stochastic averages related to the gauge - fixing and the `` fadeev - popov '' determinant vanish . to see how this happens ,
let us write the path - integral as @xmath60 where @xmath61}\theta^{ab}u^{b}\label{eq:32}.\end{aligned}\ ] ] the ghosts @xmath62 , @xmath63 are introduced to express the determinants .
next , we introduce the background decomposition given by equation ( [ eq:2 ] ) and equation ( [ eq:29 ] ) will yield @xmath64=\int(dt^{a}_{\mu})(d\psi)(d\bar{\psi})(d\phi^{a})(d\bar{u}^{a})(du^{a})exp .
\{-s'(t,\phi,\psi,\bar{\psi},u^{a},\bar{u}^{a};\tilde{f})\}\ ] ] next we do the stochastic averaging given by @xmath65\rangle_{\tilde{f}^{a}}=\int(dt^{a}_{\mu})(d\psi)(d\bar{\psi})(d\phi^{a})(du^{a})(d\bar{u}^{a})\langle exp.\{-s'\}\rangle_{\tilde{f}}\ ] ] we evaluate the stochastic average by expanding @xmath66 .
the following averages are needed : @xmath67 in equation ( [ eq:35 ] ) , we make use of equations ( [ eq:18 ] ) and ( [ eq:19 ] ) . in equations ( [ eq:36],[eq:37 ] ) and ( [ eq:38 ] ) , the symbols @xmath68 , etc . ,
represent the correlated points , which arise because the derivative of the white - noise @xmath69 is `` smoothed out '' via @xmath70 note that all odd correlations , such as @xmath71 vanish because of the white - noise character of @xmath14 .
the stochastic averages involve the following integral @xmath72 using equation ( [ eq:40 ] ) and because of the @xmath73 factors in the denominators of @xmath74 and @xmath75 , we find that when we do a background decomposition given by equation ( [ eq:2 ] ) , the stochastic averages of each of the expansion terms vanish .
thus , we find @xmath76 giving @xmath77 , which is given by equations ( [ eq:18 ] ) and ( [ eq:19 ] ) .
furthermore , the correlated terms in equations ( [ eq:36 ] ) , ( [ eq:37 ] ) and ( [ eq:38 ] ) that involve @xmath78 and @xmath79 also vanish for the same reason . taking everything so far into account
, we find @xmath80 -\frac{1}{3!}[\langle s\rangle^{3}\nonumber\\ & + & 3[(\int d^{4}xd^{4}y)(\langle\underbrace{\mathcal{l}(x)\mathcal{l}(y)}\rangle_{\tilde{f}})\langle s\rangle]\nonumber\\ & + & \frac{1}{4!}[\langle s\rangle^{4}+6(\int d^{4}xd^{4}y\langle\underbrace{\mathcal{l}(x)\mathcal{l}(y)}\rangle_{\tilde{f}})\langle s\rangle^{2}\nonumber\\ & + & \int d^{4}xd^{4}yd^{4}zd^{4}r\langle \underbrace{\mathcal{l}(x)\mathcal{l}(y)\mathcal{l}(z)\mathcal{l}(r)}\rangle_{\tilde{f}}]\nonumber\\ & + & .... \end{aligned}\ ] ] summing the series gives @xmath81 where @xmath82 since equation ( [ eq:44 ] ) is ghost independent , we simply drop the ghost path - integral and lump it with the normalization factors needed to make sense of the path - integral . at this point , where is the non - local four - fermi ( nlff ) interaction in equation ( [ eq:44 ] ) ?
it is found in @xmath83 we will evaluate the stochastic average by making use of @xmath84 from equation ( [ eq:46 ] ) , we must have @xmath85 where @xmath15 and @xmath86 . equation ( [ eq:47 ] )
implies that @xmath87 the equivalence follows from the fact that for a spherically symmetric function , @xmath88 and @xmath89 .
substituting equation ( [ eq:47 ] ) in nlff , we find @xmath90 where @xmath91
i.e. , @xmath92 represents the unit vectors in 4d spherical coordinates .
if the fermion field is spherically symmetric , the angular integration does not vanish only when @xmath93 , i.e. , the 4d vectors are collinear . using @xmath94 and @xmath95 , we find that we can write the equation ( [ eq:49 ] ) as @xmath96 with @xmath97 and @xmath98 collinear this hints of flux tube geometry as shown in figure [ fig:1 ] : , @xmath98 collinear hinting of a flux tube geometry ] this result may be a bit surprising because we started with an expansion about a classical background @xmath14 , which is spherically symmetric . but considering that the r potential is no less or no more spherically symmetric than the @xmath99 term , this result is not unreasonable at all .
when we determined the correlation @xmath100 to arrive at the linear term ( note there are extra terms ) , we made use of components of the vector field involving @xmath101 in a background @xmath14 .
all these components are important for the following reasons : ( 1 ) as was shown in section ii , just considering @xmath11 alone after stochastic averaging , does not lead to a confining term .
( 2 ) as for the @xmath20 alone , it has no kinetic term and if we integrate it out yields a constraint on the fermion fields , which under the assumption of spherical symmetry says the fermion field must vanish beyond @xmath36 ( see reference ( 18 ) ) .
( 3 ) considering @xmath102 alone will not do either because @xmath103 , if we neglect @xmath20 and @xmath11 , and we needed the @xmath104 at the right - hand side of equation ( [ eq:46 ] ) to derive equation ( [ eq:48 ] ) .
thus , a certain combination of @xmath101 in @xmath14 is key in deriving the confining non - local four - fermi term .
determining which components these are is difficult from a direct evaluation because there are many terms that contribute to it .
the linear term appears but an unusual``renormalization '' , to get rid of a @xmath105 term with @xmath106 , must be carried out .
thus , the derivation made use of equation ( [ eq:46 ] ) where , the divergence does not show up .
in this paper , we showed confinement even with dynamical quarks .
the confinement mechanism is the stochastic treatment of the scalar classical configurations @xmath14 , which arise in the non - linear regime of the non - linear gauge .
this research was supported in part by the natural sciences research institute of the university of the philippines .
|
we derive a non - local four - fermi term with a linear potential from yang - mills theory in a stochastic background .
the stochastic background is a class of classical configuration derived from the non - linear gauge .
|
nitrogen - vacancy ( nv ) centers in diamond have attracted much attention in research related to quantum computation @xcite due to their key advantages , such as high stability and long spin coherence times @xcite up to room temperature and beyond @xcite .
the spin coherence time can be increased further by isotopic engineering @xcite since only the @xmath4 carbon atoms have non - zero nuclear spin , thus contributing to spin decoherence due to hyperfine coupling . under resonant optical excitation
the @xmath5 center exhibits a strong and highly stable zero phonon line at @xmath6 ev @xcite with an excited state lifetime of about @xmath7 @xcite .
electron spin resonance analysis of the center has shown that both ground state and excited state are spin triplets , which implies that there is an even number of active electrons involved .
the ground state levels with spin projection @xmath2 and @xmath3 along the nv axis become degenerate in a magnetic field of about @xmath8 g. optical pumping causes a spin polarization of the ground state @xcite that can be attributed to a spin - orbit induced intersystem crossing with an intermediate singlet state@xcite .
when the zero field splitting is larger than the optical linewidth , repeated optical excitation leads to a spin selective steady state population in the lowest @xmath2 level of the ground state , generating a non - boltzmann steady state spin alignment and mixing of spin states @xcite , so the spin of the ground state can be both initialized and read out optically @xcite .
the standard procedure for spin manipulation in the ground state triplet involves an oscillatory ( radio - frequency ) magnetic field that gives rise to electron spin resonance . in this paper , we describe an alternative method for full spin control without rf - fields , based entirely on optical transitions .
all - optical spin manipulation of nv centers could allow for fast operations with high spatial resolution . in semiconductor quantum dots , picosecond optical control of single electron spins
has been achieved @xcite .
optically induced spin rotations in a single nv center in diamond have been demonstrated using off - resonant laser excitation @xcite .
this type of spin control relies on the optical stark effect , i.e. , the shift in energy levels induced by an applied optical field . here
, we describe an extension of this scheme which also allows for transitions between the three ground state levels , similar to existing schemes for coherent population trapping @xcite . -hybridized dangling bonds ( blue ) @xmath9 ( @xmath10 ) from the surrounding carbon atoms ( black ) and @xmath11 from the substitutional nitrogen atom ( orange ) overlapping within the vacancy ( center ) .
the @xmath12-axis ( long red arrow ) defines the @xmath13-axis of the coordinate system and , due to the zero field splitting determines the quantization axis of the electron spin . the magnetic field @xmath14 is applied along the nv - axis .
[ nv],scaledwidth=52.0% ]
to model optical spin rotations in an individual nv - center in diamond , we start from the commonly used description that fundamentally involves a total number of six electrons , but can be reduced to an effective two - electron @xcite or , equivalently , two - hole model @xcite .
the four relevant single - electron orbitals @xmath15 , @xmath16 , @xmath17 , and @xmath18 can be obtained by projecting the @xmath19-hybridized dangling bonds @xmath20 of the carbon atoms and @xmath11 of the nitrogen atom ( see fig .
[ nv ] ) onto the irreducible representations of the @xmath21 symmetry group of the nv center @xcite .
the electron configurations for the ground and first excited state are obtained as follows : in the ground state ( @xmath22 ) configuration , the lower - energy @xmath23 orbitals are completely filled with two electrons each , while the @xmath17 and @xmath18 contain one electron each .
the two - fold degenerate excited state configuration ( @xmath24 ) is obtained by promoting another electron from @xmath15 to @xmath17 or @xmath18 . due to the coulomb interaction between the two electrons ,
the spin triplet lies lowest in energy and forms the ground state @xmath25 , transforming according to the representation @xmath26 of @xmath21 @xcite .
electric dipole transitions connect this triplet to the excited state spin triplet @xmath27 , transforming according to the @xmath28 representation .
the two - fold orbital and three - fold spin degeneracies give rise to a total of six states in @xmath27 , compared to three states in @xmath25 .
the spin singlet states will not be of direct importance for our discussion , and are left out of our model .
the entire state space for our model is thus nine - dimensional .
the hamiltonian of the ground state spin triplet in the basis @xmath29 is @xmath30 where @xmath14 denotes an external magnetic field aligned with the @xmath12-axis , @xmath31 the land g - factor , and @xmath32 the bohr magneton . around the ground - state level anticrossing ( lac )
, we can split the zeeman energy into a term that compensates the ground state zero field splitting and an additional variation , @xmath33 . the zero field
splitting @xmath34 ghz @xcite is caused by the reduction of the symmetry in spin space to @xmath21 due to the crystal field .
the absence of orbital degeneracy in the ground state triplet implies that strain and spin - orbit interaction have very little effect on the ground state . here
, we have neglected the effect of the hyperfine coupling to the nuclear spins of the intrinsic nitrogen atom and surrounding @xmath35c atoms .
if necessary , the nuclear spin state could be prepared optically . at low temperatures
, the excited state fine structure can be understood to a large extent from strain and spin - spin interactions . in the basis of spin - orbit states with full symmetry , described by the @xmath21 double group including spin @xcite , @xmath36 the excited - state hamiltonian matrix is @xmath37 where @xmath38 is the axial spin - orbit splitting , and @xmath39 and @xmath40 are the well - known spin - spin interactions @xcite .
experimentally , it was found that @xmath41 ghz , @xmath42 ghz and @xmath43 ghz @xcite .
the land factors of ground and excited state were found to be equal @xcite , @xmath44 .
the energy gap @xmath45 is defined as the difference between the @xmath46 excited states and the @xmath47 ground states at @xmath48 .
the dependence of the ground- and excited state levels on @xmath14 is shown in fig .
[ pumping ] .
since electric dipole transitions are spin - conserving , our all - optical spin control scheme requires a spin non - conserving mechanism in the excited state .
the longitudinal spin - orbit interaction term @xmath38 only leads to an additional energy splitting between states with different spin projections and can not flip the spin .
it was speculated that the transversal part of the spin - orbit interaction can lead to spin flips @xcite , but it has recently turned out that it can only connect orbital states belonging to different irreducible representations@xcite .
however , the transversal component @xmath49 of the spin - spin interaction allows for the non - spin - conserving transitions between the @xmath28 and @xmath50 states , explaining the experimentally observed transitions@xcite .
the components @xmath51 and @xmath52 of the non - axial strain can be written in polar coordinates , @xmath53 and @xmath54 . here ,
@xmath55 was defined such that it corresponds to the angle between the symmetry axis of strain eigenstates @xmath56 and the symmetry axis of the unperturbed @xmath17-orbital .
we assume the system to be optically driven with a radiation field at fixed frequency @xmath57 near @xmath45 , therefore it is convenient to describe the excited states in a corotating frame , while keeping the ground states fixed .
we then work in the rotating wave approximation where counter - rotating terms with frequency @xmath58 are neglected .
this is justified as long as @xmath59 .
optical transitions between the ground and excited state are described with the electric dipole operator for two electrons , @xmath60 , where @xmath61 is the single - particle electric dipole operator , where @xmath62 denotes the electron position operator . here
, we assumed the incident light to be linearly polarized perpendicular to the @xmath12-axis ( which defines the @xmath13-axis of the coordinate system ) with polarization angle @xmath63 , with @xmath64 for polarization parallel to the symmetry axis of the @xmath17-orbital .
taking matrix elements with the ground and excited state basis states yields the hamiltonian @xmath65 where the detuning @xmath66 is defined with respect to the lowest - lying excited state energy levels @xmath46 ( neglecting strain and spin mixing @xmath67 ) , and @xmath68 .
the transition matrix is given as @xmath69 where , for linear polarization of the excitation field , @xmath70 and ( @xmath71 ) @xmath72 with the reduced matrix element of the position operator defined as @xmath73 we are interested in linearly polarized optical fields , where @xmath74 is real .
the magnitude of the dipole matrix elements can be estimated from the observed rabi oscillation period @xcite and the linear stark shift @xcite , typically @xmath75 .
since the energy levels of the ground state @xmath76 and the excited state @xmath77 are widely seperated by @xmath78 ev @xmath79 thz and coupled by small perturbations @xmath80 , we can use a schrieffer - wolff transformation @xcite of the hamiltonian ( block - matrix ) as a valid approach to determine the effective dynamics of the driven system up to second order in the perturbation .
the schrieffer - wolff transformation is defined as follows , @xmath81 + \frac{1}{2}\left[s,\left[s , h]\right]\right ] + \mathcal{o}\left(s^3\right),\ ] ] with the anti - hermitian transformation matrix @xmath82 the aim of the transformation is to remove the coupling @xmath83 in first order , which can be achieved if @xmath84= -v$ ] . in terms of the submatrices for the two seperated systems @xmath76 and @xmath77 and their coupling @xmath83
this condition reduces to @xmath85 which also implies that in a perturbation series in @xmath86 , the matrix @xmath87 will be first order , @xmath88 .
this particular choice of transformation secures that the first order terms in eq .
( [ heff ] ) cancel and we are left with an effective ground - state hamiltonian @xmath89 note that in the absence of strain @xmath90 ghz is the seperation between the two closest - lying energy levels @xmath91 and @xmath92 , and thus for resonant excitation between those two levels , the optical driving field strength has to be much smaller than @xmath93 mhz .
we focus on the transition between the @xmath3 and @xmath2 ground state spin levels .
this two - level system can be split off from the @xmath94 state @xcite , in the regime @xmath95 . in this case
the dynamics is described by the @xmath96 hamiltonian @xmath97 where the effective ( pseudo- ) magnetic field has components , @xmath98 and @xmath99 .
the dynamics within this two - dimensional subspace can be visualized using a bloch sphere picture , as shown in fig .
[ blochsphere ] .
optical driving results in a rotation of the state vector about an axis @xmath100 with polar angle ( axial orientation ) @xmath101 and azimuthal angle ( non - axial orientation ) @xmath102 .
the transversal part @xmath103 of the effective field provides for effective spin - flip transitions while @xmath104 accounts for the effective ( ac stark ) splitting between the two levels ( around the lac ) .
the term @xmath105 in eq .
( [ eq : qubith ] ) can be omitted since it merely leads to a global phase .
the precession frequency is given by @xmath106 spin flips can be implemented as rotation about an axis within the equatorial plane of the bloch sphere ( fig .
[ blochsphere ] ) which corresponds to @xmath107 and thus @xmath108 .
in the case of vanishing strain ( @xmath109 ) we obtain a simple analytical result for the transversal component of the qubit rotation axis , with magnitude , @xmath110 which is proportional to the intensity @xmath111 of the optical driving field and the transversal spin - spin coupling @xmath67 in the excited state .
the azimuthal angle of the rotation axis is determined by the optical polarization angle @xmath63 , @xmath112 where the factor of 2 reflects the double group character of spin representation .
the polar angle @xmath113 of the rotation axis is independent of @xmath63 and for small @xmath114 even independent of the driving field strength @xmath74 .
the residual zeeman splitting is limited by the hyperfine lac of about @xmath115 mhz@xcite , and therefore for an optical coupling @xmath116 we can always find pairs of parameters @xmath117 that fullfill the condition @xmath118 . in the limit of large detuning ,
i.e. , when @xmath119 dominates all other energies in the denominators of eq .
( [ inplane - axis ] ) , we can approximate the transverse component of the effective field as @xmath120 we now include the effect of strain in the diamond crystal into our discussion . for moderate strain @xmath121 ghz
, the @xmath122 levels of the excited state are largely seperated from the @xmath46 levels , and thus the strain - induced mixing of @xmath122 states and @xmath46 states can be neglected in lowest order . and detuning @xmath66 at fixed dipole coupling strength @xmath123 for different values of transversal strain @xmath124 and @xmath125 ghz ( from top to bottom ) to fullfill the condition @xmath107 for precession around a rotation axis within the equatorial plane of the bloch sphere for @xmath126 .
[ deltabrootstr],scaledwidth=43.5% ] for different values of transversal strain @xmath124 and @xmath125 ghz ( from left to right ) for polarization parallel ( @xmath127 ) and perpendicular ( @xmath128 ) to strain with dipole coupling @xmath129 mhz.[btrootstr],scaledwidth=43.5% ] the main effect of moderate strain is thus a shift of the resonances in eq .
( [ inplane - axis ] ) by @xmath130 , lifting the degeneracy of @xmath131 and @xmath132 levels .
though strain does not directly mix states with different spin projections , this shift reduces the energetic seperation between coupled @xmath28 and @xmath50 levels and therefore strongly enhances the efficiency of spin - flip - transitions . and transversal strain @xmath133 that match the condition @xmath107 for precession around a rotation axis within the equatorial plane of the bloch sphere for different dipole coupling @xmath134 and @xmath135 mhz ( from bottom to top ) for @xmath126.,scaledwidth=43.5% ] in fig .
[ deltabrootstr ] , we plot suitable pairs of parameter values for the detuning @xmath66 ( near resonant driving ) and zeeman splitting @xmath114 with varying strain @xmath133 that fulfill the condition @xmath107 for an in - plane rotation axis .
this defines an implicit function @xmath136 given @xmath63 and @xmath55 , e.g. in the case of @xmath118 for @xmath126 . to find the strength of spin - flip transitions , we substitute @xmath137 into the precession frequency for an in - plane rotation axis , @xmath138 and plot it in fig .
[ btrootstr ] as a function of the optical frequency detuning @xmath119 for different values of transversal strain .
we find that the precession frequency @xmath138 indeed increases with strain .
varying @xmath74 numerically shows that the precession frequency is still proportional to the intensity of the optical driving field @xmath139 .
note that the perturbative approach breaks down as detuning approaches strain ( divergence of @xmath138 in fig .
[ btrootstr ] ) , restricted by the validity condition @xmath140 for the schrieffer - wolff transformation in eq .
( [ ham ] ) .
we also investigate the dependence on the direction of strain and polarization ( see fig .
[ btbeta ] ) .
expectedly we get the highest efficiency for collinear strain and polarization and a minimal efficiency for perpendicular relative orientation with an overall sinusoidal form of twofold symmetry .
changing the optical polarization angle @xmath63 only leads to a uniform and continuous shift of this function ( this has been checked for a variety of different values , but for simplicity we only show it for @xmath64 and @xmath141 ) .
thus , for weak strain , the resulting effective field only depends on the relative angle @xmath142 between the strain and polarization angles . however , as the transversal strain @xmath133 increases beyond about @xmath143 and @xmath144 ghz , we start observing a modulation of the field with higher harmonics of @xmath142 . of precession axis for polarization parallel ( @xmath64 ) and perpendicular ( @xmath141 ) to @xmath17-orbitals symmetry axis in respect to strain orientation for different strengths of transversal strain .
[ phibeta],scaledwidth=40.0% ] numerical evaluation also reveals that the azimuthal angle ( for in - plane orientation of the precession axis ) is generally independent of optical coupling strength and magnetic field , @xmath145 from fig .
[ phibeta ] we see that for @xmath146-polarized light ( _ i.e. _ @xmath147 ) , the angle @xmath102 is well approximated by @xmath148 ( at low strain ) , with an amplitude proportional to the intensity of the strain , @xmath149 in fig . [ phialpha ] , we show that the sinusoidal shape of @xmath102 as a function of @xmath63 is slightly distorted for polarization angles @xmath150 , @xmath151 ; we also find that this distortion grows with increasing strain . of precession axis for different polarization angles
@xmath63 in respect to strain orientation at transversal strain @xmath152 .
[ phialpha],scaledwidth=40.0% ] in the high strain regime , the electronic states of the nv - center are energetically split into two orbital branches with largely seperated energies @xmath131 and @xmath132 corresponding to a specific choice of coordinate axes , that fixes the orientation angle @xmath55 . in this limit @xmath138 and @xmath104 ( and thus the polar angle @xmath113 ) become independent of the orientation angles of both strain and polarization . expanding in @xmath153 yields @xmath154 and @xmath155 for the transversal part of the pseudo - field , and @xmath156 for the longitudinal part .
the higher orders contain higher harmonics , such as terms @xmath157 for @xmath158 and @xmath159 for @xmath160 in @xmath161 , indicating strain induced third order transitions mediated by the @xmath122 levels .
we have shown that the effective precession axis and frequency of the ground state spin of the @xmath5-center can be fully controlled by off - resonant optical excitation , by adjusting the frequency detuning @xmath66 and linear polarization angle @xmath63 of the optical driving field for a given intensity @xmath74 ( optical dipole coupling ) and magnetic field @xmath14 .
the orientation of the precession axis is determined by two angles @xmath113 and @xmath102 , where the first is depending on all parameters ( including strain and polarization ) and the latter is independent of magnetic field and optical coupling strength and basically controlled by polarization and strain .
the strain effects can be compensated by external bias voltage @xcite . since any unitary qubit operation ( rotation around axis @xmath162 by angle @xmath163 )
can be composed by successive rotations around two orthogonal axes on the bloch sphere , a complete set of single - qubit operations can be generated optically in this way . from a purely geometric point of view ,
spin rotation about an axis within the equatorial plane of the bloch sphere ( where @xmath164 ) is most effective for flipping the spin , although any axis other than the @xmath13-axis would do ( the smaller the polar angle @xmath113 , the more pulses are be required ) .
a full spin - flip is obtained by a @xmath165-rotation around an axis within the equatorial @xmath166-plane of the bloch sphere , providing an estimate for the gate switching ( optical pumping ) time @xmath167 in the limit of large detuning and weak strain , @xmath168 the switching time for spin - flip transitions is limited by the spin mixing term @xmath67 , since the ( above ) condition implies via the off - resonant condition @xmath169 the following lower limit for the spin - flip , @xmath170 where for that latter estimate we assumed @xmath171 ghz .
we thank b. buckley for discussions and f. fehse for generating the graphics in fig .
[ nv ] and the konstanz center for applied photonics ( cap ) and bmbf quhlrep for funding .
|
we describe an all - optical scheme for spin manipulation in the ground - state triplet of the negatively charged nitrogen - vacancy ( nv ) center in diamond .
virtual optical excitation from the @xmath0 ground state into the @xmath1 excited state allows for spin rotations by virtue of the spin - spin interaction in the two - fold orbitally degenerate excited state .
we derive an effective hamiltonian for optically induced spin - flip transitions within the ground state spin triplet due to off - resonant optical pumping .
furthermore , we investigate the spin qubit formed by the zeeman sub - levels with spin projection @xmath2 and @xmath3 along the nv axis around the ground state level anticrossing with regard to full optical control of the electron spin .
|
in pyrochlore materials , with the general formula a@xmath2b@xmath2o@xmath3 , the a and b site ions form an interpenetrating network of corner - sharing tetrahedra .
@xcite when populated by magnetic ions with nearest neighbor antiferromagnetic ( afm ) interactions these materials display anomalous frustrated magnetism .
@xcite for classical spins and nearest neighbor interactions the ground - state is a manifold characterized by zero magnetization on every tetrahedron .
@xcite correspondingly pyrochlore magnets remain paramagnetic to much lower temperatures than their curie - weiss temperature ( @xmath8 ) where the collective properties are determined by longer range or anisotropic interactions and thermal and/or quantum fluctuations .
the resulting low temperature phases include spin - glasses , spin - liquids , and magneto - elastically induced nel order .
@xcite while the majority of pyrochlore magnets are insulators , an interplay between magnetism and strong electron correlations can occur when the b - site is a @xmath9- or @xmath10-ion resulting in a metal - to - insulator transitions ( mit ) , heavy fermion behavior and even superconductivity . @xcite the ruthenium pyrochlores for example display a variety of ground states near a correlation induced mit .
@xcite their electronic bandwidths are strongly influenced by the ru - o - ru bond angle , which in turn is controlled by the ionic radius of the a@xmath5 ion .
@xcite thus r@xmath2ru@xmath2o@xmath3 ( r = y , rare earths ) are insulating and show long - range magnetic order while bi@xmath2ru@xmath2o@xmath3 is a pauli paramagnet .
tl@xmath2ru@xmath2o@xmath3 is a metal at room temperature and a spin - singlet insulator below 120 k. bulk measurements show antiferromagnetic ordering of the ru sublattice for the insulating compounds with a critical temperature , t@xmath6 , that decreases monotonically from 160 k for pr to 81 k for yb ; consistent with the lanthanide contraction .
@xcite the temperature and energy scale of the magnetic interactions on the rare earth sublattice is an order of magnitude lower than for the transition metal b - site .
@xcite in this paper we examine the structure and magnetism of pr@xmath0bi@xmath1ru@xmath2o@xmath3 solid solutions.@xcite bulk measurements show substituting bi@xmath5 for pr@xmath5 drives the system from an antiferromagnetic insulator ( @xmath11 ) to a pauli paramagnetic metal ( @xmath12 ) .
while this transition has been observed in other ru pyrochlores , in pr@xmath0bi@xmath1ru@xmath2o@xmath3 the low-@xmath13 specific heat is greatly enhanced , reminiscent of what is observed in non - fermi - liquid and heavy fermion systems . in both pr@xmath0bi@xmath1ru@xmath2o@xmath3 and pr@xmath2ir@xmath2o@xmath3 , we find a quasi - doublet ground - state for pr and a singlet excited state with analogous wave functions . in previous inelastic neutron scattering experiments we showed the enhanced low-@xmath13 specific heat and heavy fermion - like properties
are actually a consequence of a static inhomogeneous splitting of the non - kramers pr@xmath5 ground - state doublet . here
we show that even without bi substitution in @xmath14 , the ground state degeneracy anticipated for non - kramers praseodymium in the pyrochlore lattice a - site is lifted .
this is evidence of a local structural distortion that breaks the three fold rotation axis . through exafs ,
we then provide direct structural evidence for a distribution of coordinating environments for praseodymium in @xmath14 .
while the b - site ( ru ) environment remains well ordered throughout the series , the a - site becomes progressively disordered with increasing @xmath15 , primarily near bismuth . for @xmath11 , below t@xmath6 ,
the ru sublattice orders in a similar arrangement as for y@xmath2ru@xmath2o@xmath3 so this order does not appear to be influenced by the pr rare earth anisotropy . @xcite for the pr sublattice however , no order is detected by diffraction down to 1.5 k for any @xmath15 .
the specific heat has a schottky - like anomaly centered at 3 k but no further anomalies associated with magnetic ordering at least down to 0.1 k. probed by inelastic neutron scattering , the low energy magnetic excitation spectrum of pr@xmath0bi@xmath1ru@xmath2o@xmath3 shows the corresponding mode of excitation . in the temperature dependence of the excitation spectrum we provide evidence for collective effects from pr - pr interactions for @xmath14 ,
these are shown to vanish for @xmath16 where the temperature dependence of the inelastic scattering can be described by inhomogenous single ion physics .
powdered samples of pr@xmath0bi@xmath1o@xmath3 ( @xmath4 , @xmath7 and @xmath12 ) and pr@xmath2ir@xmath2o@xmath3 were synthesized using the solid state reaction method . for the ru containing samples , mixtures of pr@xmath2o@xmath17 , bi@xmath2o@xmath17 and ruo@xmath2 in proper molar ratios
were pre - reacted at 850 @xmath18c for 15 h in air and then ground and pressed into pellets . in the case of pr@xmath2ir@xmath2o@xmath3 mixtures of pr@xmath19o@xmath20 and iro@xmath17 in proper molar ratios were pressed into pellets .
the pellets were subsequently sintered at 1000 - 1200 @xmath18c in air with intermediate grindings .
all samples were characterized by powder x - ray diffraction .
these measurements showed the samples all adopt the cubic pyrochlore structure and are single phase , except for the @xmath4 sample which contained 3.38(5 ) wt % of unreacted ruo@xmath2 .
detailed bulk measurements on these samples have been reported elsewhere .
@xcite for the heat capacity measurement @xmath14 powder was thoroughly mixed with silver powder 50% by weight and cold pressed into a solid pellet to achieve adequate thermal conductivity for thermal equilibration .
data were collected down to 90 mk with the adiabatic relaxation method using a commercial physical property measurement system ( ppms ) dilution refrigerator .
the specific heat capacity of @xmath14 was obtained by subtracting the measured specific heat capacity of silver from the measured total heat capacity .
@xcite exafs studies were carried out at the stanford synchrotron radiation lightsource ( ssrl ) for all the metal edges in @xmath16 and the pure end compounds @xmath4 and @xmath12 .
transmission mode exafs data were collected for the pr l@xmath21-edge ( 5964 ev ) , bi l@xmath21-edge ( 13419 ev ) , and ru k - edge ( 22,117 ev ) . we used a si ( 220 ) double monochromator for the ru edge and si ( 111 ) crystals for the bi and pr l@xmath21 edges .
the slit height was 0.5 mm , giving energy resolutions of 1 ev for the pr l@xmath21-edge , and @xmath222.7 ev for the bi l@xmath21- and the ru k - edge .
the monochromator was detuned 50% for the pr and bi l@xmath21 edges and 30% for the ru k - edge to minimize harmonics .
exafs samples were prepared by first brushing fine powder ( @xmath23 5 @xmath24 m ) onto scotch tape ; two pieces of tape were then pressed together ( double layer ) to encapsulate the powder . for the ru edge we used 7 , 9 , and 15 double layers for @xmath12 , @xmath7 and @xmath4 respectively .
3 double layers were used for the bi l@xmath21 edges , and 2 double layers for the pr l@xmath21-edges .
standard procedures were used to reduce the exafs data .
@xcite first a pre - edge subtraction was done to remove absorption from other atoms ; this yields the absorption edge of interest , @xmath25 .
then a spline was fit through the data above the edge to obtain an estimate of the absorption , @xmath26 , with no photoelectron backscattering .
next the exafs oscillations , @xmath27 , were obtained from @xmath25 = @xmath26(1+@xmath27 ) , and @xmath27 converted to @xmath28 using @xmath29 @xmath30/2 m = @xmath31 , where @xmath32 is the absorption edge energy . finally @xmath33(k )
( usually n = 1 - 3 ) was fourier transformed ( ft ) into r - space , where peaks in the ft correspond to various neighboring shells about the absorbing atom .
powder neutron diffraction data were collected on the @xmath4 sample at the national institute of standards and technology in gaithersburg , maryland ( bt1 ) and the isis facility , rutherford appleton laboratory , uk ( hrpd ) . for the bt1 experiment a 10 g sample was sealed in a vanadium container with length 50 mm and diameter 10.8 mm and temperature was controlled in a he cryostat . a ge ( 311 )
monochromator with a 90@xmath18 take - off angle ( @xmath34= 2.079 ) and 15 minutes of arc in - pile collimation were used .
data sets were collected for temperatures between 1.5 k and 180 k and scattering angle @xmath35 from 3 - 168@xmath18 with a step size of 0.05@xmath18 . for the hrpd experiment 10 g of sample
was placed in a vanadium container 15 mm@xmath3620 mm@xmath3610 mm ( h@xmath36w@xmath36d ) within a he cryostat .
data sets were collected for temperatures between 2 k and 300 k using 10 ms to 110 ms chopper settings .
rietveld analysis of the neutron powder diffraction patterns was performed using the fullprof software package .
@xcite inelastic neutron scattering experiments were carried out at the isis facility , rutherford appleton laboratory , uk .
high energy data were collected on all samples using the time - of - flight ( tof ) spectrometer het .
additional low energy data were collected on the @xmath4 sample using the iris spectrometer .
@xcite for the het experiment the samples were loaded in an al sachet and the total mass of sample in the beam was 19.2 g for @xmath4 , 24.9 g for @xmath7 , 22.55 g for @xmath12 and 14.086 g for pr@xmath2ir@xmath2o@xmath3 .
the samples were loaded into a top loading closed cycle he refrigerator .
incident energies of @xmath37= 35 and 160 mev were used for this experiment , the full width at half maximum ( fwhm ) energy resolution at the elastic line was 1.4 mev and 7 mev respectively .
data were collected at 5 k and 200 k. more details on normalization and the correction for the phonon contribution to the scattering data will be given below .
crystal field ( cf ) analysis of the data was performed using the focus program .
@xcite the iris experiment was carried out on 15 g of @xmath14 .
the sample was held in a sealed 2 mm double walled al can with diameter 23 mm and height 53 mm and loaded into a he cryostat .
bandwidth disk choppers selected an incident spectrum from 1.35 mev to 4.6 mev above @xmath38 mev pulsed at 25 hz and a backscattering pyrolytic graphite analyzer bank with a 25 k be filter selected the final energy , @xmath39 .
the fwhm elastic energy resolution was 17.5 @xmath24ev .
data were collected over a temperature range from 1.5 k to 200 k. the magnetic neutron scattering cross - section of a powder sample at wave vector transfer @xmath40 and energy transfer @xmath41 can be written as @xcite @xmath42 where @xmath43= -1.913 and @xmath44 are the spectroscopic g - factors of the neutron and the magnetic ion respectively , @xmath45 fm is the classical electron radius , and @xmath46 is the magnetic form factor .
@xcite the dynamic spin correlation function @xmath47 is given by @xmath48 here @xmath49 is the number of formula units encompassed in each of the double summations .
@xmath47 can be related to the generalized susceptibility through the fluctuation - dissipation theorem @xcite @xmath50 where @xmath51 and @xmath52 denotes the imaginary part of the generalized susceptibility . in a system that contains magnetic rare earth ions where the free ion ground - state @xmath53 multiplet is split due to the effect of a cf
, the generalized susceptibility can be calculated from the eigenfunctions and energies of the cf hamiltonian . for a cubic material where inter - site interactions are treated in the random phase approximation
, the generalized susceptibility per formula unit can be expressed as @xmath54 where @xmath55 is the number of magnetic ions per formula unit , @xmath56 is the exchange interaction and @xmath57 is the single site susceptibility , which can be expressed as follows @xcite @xmath58.\end{aligned}\ ] ] here @xmath59 indicates a carterian component ( @xmath60 ) of the angular momentum operator , @xmath61 and @xmath62 are the eigenfunctions and energies of the crystal field hamiltonian @xmath63 , and @xmath64 is the thermal population factor . for pr@xmath5 ions in the pyrochlore structure
we expect the 9-fold degenerate free ion ground - state @xmath53 multiplet @xmath65 to split into 3 doublets and 3 singlets under the effect of the @xmath66 symmetric @xmath63 .
choosing [ 111 ] as the quantization axis , the single - ion crystal field hamiltonian takes the form : @xmath67 where @xmath68 are the cf parameters and @xmath69 are stevens operator equivalents of the cf tensor operators as discussed by hutchings .
@xcite here the cf interaction in the ls coupling scheme is treated as a perturbation within the ground - state @xmath53 multiplet only .
the following corrections were applied to the neutron counts in the time histograms collected on both het and iris . first a time - independent background measured for @xmath70 -@xmath39 was subtracted .
then the data were scaled to the relevant count rate in a pre - sample monitor and finally converted into @xmath41 histograms .
this procedure gives @xmath41-dependent data @xmath71 , which are related to the scattering cross - section through convolution with a resolution function , as follows : @xmath72 where @xmath49 is the number of formula units in the sample , and @xmath73 is the spectrometer constant .
the instrumental resolution function @xmath74 is assumed to be unity normalised : @xmath75 the normalized intensity @xmath76 is related to the measured intensity as follows @xmath77 thus @xmath76 is the resolution smeared partial differential scattering cross section per formula unit which we express in absolute units of mbarn sr@xmath78 mev@xmath78 f.u.@xmath78 . for the het experiment ,
@xmath79 was determine by measuring the incoherent scattering from a standard flat vanadium slab sample for each of the chosen incident energies .
for the iris experiment , @xmath79 was determined from bragg scattering through a method that has been described elsewhere .
@xcite these procedures yield absolute measurements of @xmath76 to an overall scale accuracy of 20 @xmath80 .
to investigate possible distortions in the local structure we carried out exafs measurements . here
we present the main results of the exafs analysis . in figure [ r - space - pure]a ,
b we first show the experimental @xmath81-space plots ( ft@xmath82(k ) ) at t=4 k ( solid squares ) for the a - site atoms ( pr and bi ) in the end compounds @xmath4 and @xmath12 . the first peak in each scan is the metal - o peak , near 2.0 - 2.5 ; a sum of pr - o2 ( 2.243 ) and pr - o1 ( 2.546 ) for the pr l@xmath21 edge , and a sum of bi - o2 ( 2.228 ) and bi - o1 ( 2.538 ) peaks for the bi l@xmath21 edge .
the next peak ( near 3.3 ) is a combination of metal - metal peaks , i.e. for @xmath4 at the pr edge it would be a sum of pr - ru and pr - pr peaks .
note the peaks in the exafs spectra are shifted to lower @xmath81 compared to the actual distances , by a well known phase factor .
for example the two pr - o peaks in the exafs spectra of fig .
[ r - space - pure](a ) are located at @xmath22 1.8 and 2.1 if plotted separately - a shift of roughly -0.45 . for these peaks ,
the @xmath81-space phase ( real part of the transform ) of the peak for the shortest pr - o2 distance ( 2 neighbours ) is nearly out of phase with that for the longer distance pr - o1 peak ( 6 neighbours ) , leading to a dip in the spectra at 1.8 .
figure [ r - space - pure]c , d shows the corresponding ru k - edge data for the two pure samples ( solid squares ) ; c ) @xmath4 , d ) @xmath12 .
the two data sets are very similar . for ru ,
there is only one nearby o neighbor ( o1 ) , while the second peak is a sum of ru - ru and either ru - pr ( fig .
[ r - space - pure](c ) ) or ru - bi ( fig .
[ r - space - pure](d ) ) .
a quick evaluation of the data is obtained by simulating the exafs @xmath81-space data using the program feff8.2 , the known ordered pyrochlore structure , and a global broadening ( pair distribution width , @xmath83 ) for all peaks .
@xcite these simulations are shown as solid lines in figure [ r - space - pure ] , with @xmath83 = 0.07 and 0.08 for the pr and bi edges respectively , and 0.0725 and 0.05 for the ru k - edge data for @xmath4 and @xmath12 .
note that the only parameter adjusted here was the global broadening .
this modeling shows the environment about pr is relatively well ordered ( some disorder for pr - o2 and pr - pr is discussed later ) whereas that about bi is highly disordered the bi - o1 peak which should occur near 2.1 in figure [ r - space - pure](b ) , is strongly suppressed .
the weak peak near 1.7 - 1.8 is consistent with the small bi - o2 peak ( two o neighbors located inside the bi / pr tetrahedra ) . for the ru edge data ,
the ru - o1 peak near 1.6 is also well ordered , surprisingly , even for the bi sample which has significant disorder of the bi - o peak .
consequently since the ru - o1 peak has little disorder while the bi - o1 peak is disordered , bi must be displaced from the usual a - site position , in a direction perpendicular to the bi - o2 axis .
this leads to a small distortion of the bi - o2 peak .
next we consider disorder in the mixed sample , @xmath7 . in figure
[ r - space - mixed ] we show the @xmath81-space plots for the three edges ( pr and bi l@xmath21- , and ru k - edge ) at 4k . for each plot
the first peak corresponds to the nearest neighbor o shell ; both the combined pr - o1/pr - o2 peak in fig .
[ r - space - mixed]a and the ru - o1 peak in fig .
[ r - space - mixed]c have a large amplitude indicating relatively little disorder .
in contrast , the bi - o peak ( bi - o1 and bi - o2 ) in panel b ) is suppressed , mostly at the position for bi - o1 , similar to the results for the pure end compound @xmath12 .
this indicates that bi is also displaced perpendicular to the bi - o2 axis in the mixed compound and by a comparable amount .
we find similar results at higher temperatures , the ru - o and most pr - o peaks are generally well ordered while the bi - o peak is strongly suppressed , indicating significant disorder .
we also find that the pr - ru , ru - pr and ru - ru second neighbor peaks are reasonably ordered , but when bi is present , peaks that include bi second neighbors ( e.g. pr - bi or ru - bi ) also have disorder . to determine the crystal field level scheme and the relevant low energy spin degrees of freedom we carried out high energy inelastic neutron scattering measurements on het .
three methods were used to determine and then subtract the nonmagnetic phonon contribution to the data .
for pr@xmath0bi@xmath1o@xmath3 the scaling and direct subtraction methods were used , while for pr@xmath2ir@xmath2o@xmath3 the discus package was used .
@xcite in all three methods the nonmagnetic contribution to the neutron scattering at low angles was subtracted by scaling the spectrum measured at high wave vectors where the magnetic response is negligible . in the so - called scaling method ,
the nonmagnetic contribution at low angles ( @xmath84 19@xmath18 ) is estimated from the measured scattering at high scattering angles ( @xmath84 135@xmath18 ) by using an energy dependent scaling factor determined from direct measurements on the pauli paramagnetic compound @xmath12 .
this procedure is based on two assumptions .
first , the magnetic scattering intensity , which is proportional to the square of the pr@xmath5 magnetic form factor , is negligible in the high angle scattering data .
second , the energy dependent ratio between phonon scattering at low and high scattering angles is the same for all four compounds .
the second method employs a direct subtraction method to estimate phonon contributions using the @xmath85 compound as the phonon blank material , after accounting for the difference in the total scattering cross section @xmath83(@xmath16)@xmath22 0.9@xmath83(@xmath85 ) and @xmath83(@xmath11)@xmath22 0.8@xmath83(@xmath85 ) .
both methods produced a very similar magnetic response though with less statistical error for the first scaling method .
as a result all the data collected on pr@xmath0bi@xmath1o@xmath3 were analyzed using the first method . in the case of pr@xmath2ir@xmath2o@xmath3 the discus package uses a monte carlo method to calculate a wave - vector - dependent scaling factor , which showed no significant energy dependence .
figure [ phsub ] shows the total spectra for pr@xmath0bi@xmath86ru@xmath2o@xmath3 with @xmath85 , @xmath16 and @xmath11 measured at 5 k with incident energies @xmath37= 35 mev and 160 mev for low ( @xmath35 @xmath8719@xmath18 ) and high ( @xmath35 @xmath87135@xmath18 ) scattering angles .
for @xmath16 and @xmath11 the high angle spectra have been scaled down using the energy dependent scaling factor determined from the @xmath85 data ( solid lines in figs .
[ phsub](a ) and [ phsub](b ) ) . at low scattering angles , corresponding to q@xmath88= 1.36 @xmath78 and
2.91 @xmath78 respectively , the spectra from @xmath89 and @xmath11 contain both magnetic and phonon contributions . in the high scattering angles
spectra , corresponding to q@xmath88= 7.62 @xmath78 and 16.30 @xmath78 respectively however , the magnetic contributions are small due to the very small form factor for pr 4__f _ _ electrons at such large q values . as shown in figs .
[ phsub](a ) and [ phsub](b ) , the inelastic response of @xmath85 shows three clear peaks due to one - phonon scattering at 30 mev , 45 mev and 75 mev .
these features are reproduced in the scaled high angle scattering data of @xmath16 and @xmath11 , indicating that the phonon scattering is indeed similar for all three compounds .
this justifies use of the scaling method to estimate the phonon contribution to the low scattering angle spectra for the @xmath11 and @xmath16 samples .
figures [ mr_35mev ] and [ mr_160mev ] show magnetic scattering from pr@xmath0bi@xmath86ru@xmath2o@xmath3 and pr@xmath2ir@xmath2o@xmath3 , after subtracting the phonon and elastic scattering , at 5 k and 200 k. the magnetic neutron scattering cross section for @xmath14 at 5 k shows at least five magnetic excitations centered near 10 mev , 50 mev , 85 mev , 105 mev , and 116 mev energy transfer ( fig .
[ mr_35mev](b ) and [ mr_160mev](b ) ) .
closer examination of the 5 k data shows the excitations near 10 mev and 50 mev are broadened or split .
the remaining three high energy excitations take the form of isolated resolution limited peaks . comparing the spectrum at 200 k with that at 5
k the following changes are observed upon warming ; the strongest peak near 10 mev is broadened , decreases in intensity and shifts upward to 12 mev . a new broad double peak structure that resembles the broad 50 mev peak appears near 40 mev .
the three peaks near 100 mev remain in place but loose intensity on warming . in the nominal @xmath66 point group symmetry of the pyrochlore lattice , pr@xmath5 has five cf excitations .
the broadening and splitting of the two lowest energy cf excitation may indicate an inhomogenous environment for praseodymium , something we also find evidence for in high resolution measurements that will be described subsequently .
thermal expansion as well as magneto - striction and dipole fields from ru@xmath90 ordering at @xmath91 k may be responsible for the modifications in the lowest energy cf excitations near upon heating to 200 k. anomalous changes in crystal field excitations resulting from ruthenium spin ordering were for example previously document in ho@xmath2ru@xmath2o@xmath3.@xcite thermal population of the 10 mev cf level for @xmath92 k enables excitations from that level to higher energy cf levels to which dipole transitions are allowed from the excited state .
thus heating can produce extra versions of higher excitations downshifted by @xmath93 mev , which is the energy of the first excited cf state at 200 k. we interpret the heating induced peak near 40 mev as resulting from this mechanism .
this implies a finite dipole matrix element between the 12 mev and 50 mev cf levels . on the other hand
the loss of intensity for the three upper cf transitions indicates the dipole matrix elements between the first excited state state and these three levels is small or even zero .
the corresponding 5 k data for the @xmath16 sample also shows 5 excitations ( fig .
[ mr_35mev](a ) and [ mr_160mev](a ) ) .
we associate all of these magnetic peaks with pr@xmath5 cf excitations .
much as for @xmath14 , there are four relatively sharp features centered at 9 mev , 83 mev , 103 mev and 116 mev and a broad maximum near 50 mev .
the fwhm of these excitations is however , a factor 3 larger than for @xmath11 , an effect we may ascribe to alloying induced disorder in the electrostatic conditions for pr@xmath5 .
@xcite the effects of heating to 200 k are very similar to observations in @xmath14 .
as the @xmath16 sample has no magnetic phase transition down to 2 k , the similarity of the @xmath4 and @xmath7 data suggests ruthenium magnetic ordering does not have a significant effect on praseodymium here .
the 5 k and 200 k cf spectra for pr@xmath2ir@xmath2o@xmath3 are quite similar to the ruthenium based pyrochlores ( fig .
[ mr_35mev](c ) and [ mr_160mev](c ) ) .
there are again five energy levels ; here lying at 14 mev , 58 mev , 86 mev , 104 mev , and 120 mev .
the small maximum at the very top of the spectrum in fig .
[ mr_160mev](c ) may be a result of incomplete background subtraction for this strongly absorbing sample , which also is seen to impact the top of spectrum for @xmath94 mev in fig .
[ mr_35mev](c ) . in comparison to the other samples , the shifts are largest for the lower energy levels , which generally appear to respond more to the crystalline environment .
the thermal effects in this sample , which as @xmath95 has no magnetic order on the transition metal site , are qualitatively similar to both @xmath14 and @xmath95 . to determine the magnetic ordering and potential structural distortions in @xmath14 , we carried out neutron diffraction studies on bt1 at nist and hrpd at isis .
figure [ d_hrpd_fit ] shows the rietveld fits to the 300 k ( t @xmath96 t@xmath6 ) and 100 k ( t @xmath97 t@xmath6 ) data sets as collected on hrpd .
these fits show @xmath14 adopts the cubic pyrochlore structure and that the sample contained 3.46 wt % of unreacted ruo@xmath2 .
fits of the crystal structure to the data collected below t@xmath6 revealed no evidence , within the accuracy of the experiment , of a structural distortion associated with the magnetic phase transition .
they do however show that below t@xmath6 , certain low angle reflections gain intensity that can not be accounted for by the nuclear contributions alone ( figure [ d_nucl_fit ] and table [ tab : nucl_fit ] ) .
.refined structural parameters from fits to powder neutron diffraction profiles of the @xmath11 sample collected at 180 k , 60 k and 1.5 k on bt1 .
the crystal structure is cubic with space group @xmath98 with pr located on @xmath99 sites @xmath100 , ru located on @xmath101 sites and o located on @xmath102 ( o1 ) and @xmath103 ( o2 ) sites @xmath104 and @xmath105 respectively . [ cols="<,^,^,^",options="header " , ] in summary our analysis has shown that describing the ordering of the ru moments in @xmath11 either by the model proposed for y@xmath2ru@xmath2o@xmath3 or by the ir @xmath106 of space group @xmath98 gives identical fits to the data ( figs .
[ d_y2ru27_fit ] and [ irrep_fits ] ) . closer examination of both models reveals that the two associated bvs of @xmath106 are special cases of the more general description used for y@xmath2ru@xmath2o@xmath3 and these models are therefore identical ( fig .
[ magn_struct ] ) .
@xcite unlike what is found for the other magnetic rare earth containing ru - pyrochlores this suggests that the the ordering of the ru moments in @xmath4 is not influenced by the pr rare earth anisotropy .
@xcite this in turn implies that the pr ground - state is non - magnetic and can explain why no ordering associated with the pr sublattice is observed down to 100 mk .
interestingly the ordering observed for @xmath4 ( and y@xmath2ru@xmath2o@xmath3 ) is the same as found for er@xmath2ti@xmath2o@xmath3 , there the rare earth anisotropy selects co - planar ordering ( @xmath107 ) .
@xcite , @xmath107 and y@xmath2ru@xmath2o@xmath3 , within a single single tetrahedra .
@xcite , width=302 ] figure [ tdep_om ] shows the temperature dependence of the ordered ru moment , which was obtained by fitting the neutron power diffraction data using @xmath108 to describe the ru ordering .
this clearly shows the onset of an ordered ru moment below 170 k confirming that the anomaly observed in both the specific heat and magnetization measurements at 165 k is associated with the ordering of the ru - sublattice .
@xcite as the temperature is further decreased the size of the ordered moment increases , to level off below 100 k to a value of 1.5 @xmath109 .
this is in line with what is observed in the other ru - pyrochlores .
@xcite comparing the low energy magnetic response of the pure ( @xmath11 ) with that of the previously measured bi - doped material ( @xmath16 ) , figure .
[ iris_cont ] , one observes the following . in both samples the wave vector dependence of the dispersion indicates that it is associated with pr@xmath5 single ion physics and must arise from splitting of the doublet ground state .
however , while for @xmath11 the peak in the dispersion is rather sharp ( but not resolution limited ) and temperature dependent , for @xmath16 it is broad and temperature independent . the presence of lone pairs on the bi in @xmath16 ( non - magnetic and close to the insulator to metal transition ) result in a low density of extended defects or a density wave which generates a continuum of local pr environments that is temperature independent .
the exafs results show that the environment about bi is highly disordered for both the @xmath7 and @xmath12 samples .
this disorder on those a sites occupied by bi will modify the cf experienced by neighboring pr ions .
since pr is a non - kramers ion ( and the determined doublet ground state relies on @xmath66 point group symmetry ) this distribution of local distortions will lead to a broadened or split ground state . in the pure material this scenario with bi is not possible , but the exafs data indicate a small , but significant , disorder of the pr atoms along the pr - o2 axis which may also spit the ground state .
another possible scenario that might explain the splitting of the pr doublet ground state in the @xmath4 material is that it results from the ru - sublattice ordering .
it has already been observed for y@xmath2ru@xmath2o@xmath3 that magnetoelastic effects play an important role in allowing the ordering of the ru - sublattice to occur .
@xcite while no optical data is available for @xmath11 the ordering observed is identical to that of y@xmath2ru@xmath2o@xmath3 , as such it is very likely that the observed spin - phonon interactions that allow for the ordering in y@xmath2ru@xmath2o@xmath3 are also present in @xmath11 .
this combined with the lowering of the symmetry due to the ru ordering can result in a distribution of internal fields ( exchange and/or strain ) which in turn leads to the observed splitting of the doubled ground state .
we have performed exafs , elastic and both high and low energy inelastic neutron scattering measurements on pr@xmath0bi@xmath1ru@xmath2o@xmath3 to understand the magnetic properties of this materials and complete those already reported by us .
@xcite the exafs measurements reveal that the ru environment ( b - site ) remains well ordered throughout the series . in the case of the a - site
, the pr environment has some small , but significant , intrinsic disorder along the pr - o2 axis which likely contributes to the splitting of the kramers doublet .
in contrast , the environment about bi is highly disordered , and is attributed to the @xmath110 lone pairs on bi@xmath5 , which result in an off - center displacement . in agreement with previously reported diffraction studies
the bi appears to move midway between two o1 atoms .
our cf measurements reveal that the pr ions have a doublet ground state and singlet first excited state ( @xmath111 and latexmath:[$\psi_1= |3 > + for pr@xmath2ir@xmath2o@xmath3 .
the high energy inelastic neutron scattering data also suggest that strong cf - phonon coupling is present in both systems .
fits to the diffraction data of the @xmath4 end member reveal no evidence ( within our experimental resolution ) of a structural distortion associated with the ordering of the ru moments below t@xmath6 .
the magnetic ordering of the ru sublattice is similar to that of y@xmath2ru@xmath2o@xmath3 ( or by ir @xmath106 of space group @xmath113 ) .
@xcite the ordering of the ru moments is not influenced by the pr rare earth anisotropy and no ordering of the pr sublattice was observed down to 1.5 k. the low energy magnetic response of pr@xmath0bi@xmath1ru@xmath2o@xmath3 shows the presence of of a ( broad ) dispersion associated with the splitting of the pr doublet ground state .
for @xmath4 it is found to be temperature dependent , while for @xmath7 it is not .
the nature of the splitting of the ( non - kramers ) doublet ground state changes upon doping , going from being intrinsic and/or magnetoelastically induced in @xmath4 to result from bi induced a - site disorder in @xmath7 .
these measurements show that the pr ground - state can be very sensitive to local perturbations ( be they ex- or intrinsic ) , something that needs to be taken into account when studying materials containing this and other non - kramers rare earth ions , e.g. tb@xmath114ti@xmath2o@xmath3 , hg@xmath2ti@xmath2o@xmath3 or liho@xmath1y@xmath115f@xmath116 . @xcite the work at jhu and isis was supported by the us department of energy , office of basic energy sciences , division of material sciences and engineering under de - fg02- 02er45983 through 2008 and beyond that on de - fg02 - 08er46544 .
the work at rutgers was supported by nsf through grant no .
dmr-0103858 and by the doe under grant no .
de - fg02 - 07er46382 .
work done at the uclm was supported by the ramn y cajal program through grant no .
ryc-2005 - 001064 and the consejera de educacin y ciencia of the junta de comunidades de castilla - la mancha through grant no .
pii1i09 - 0083 - 2105 .
k. h. k. was partially supported by kosef through cscmr .
the exafs experiments were performed at ssrl , operated by the doe , division of chemical sciences .
99 j. e. greedan , j. mater . chem . * 11 * , 37 ( 2001 ) .
p. w. anderson , phys . rev . *
102 * , 1008 ( 1956 ) .
r. moessner and j. t. chalker , phys .
* 80 * , 2929 ( 1998 ) .
j. s. gardner , b. d. gaulin , s .- h .
lee , c. broholm , n. p. raju and j. e. greedan , phys . rev
. lett . * 83 * , 211 ( 1999 ) .
j. s. gardner , a. keren , g. ehlers , c. stock , e. segal , j. m. roper , b. fak , m. b. stone , p. r. hammar , d. h. reich and b. d. gaulin , phys .
b * 68 * 180401(r ) ( 2003 ) .
y. machida , s. nakatsuji , y. maeno , t. tayama , t. sakakibara and s. onoda , phys .
lett . * 98 * 057203 ( 2007 ) .
o. tchernyshyov , r. moessner and s. l. sondhi , phys .
b * 66 * , 064403 ( 2002 ) .
a. w. sleight , j. l. gillson , j. f. weiher and w. bindloss , solid state commun . * 14 * , 357 ( 1974 ) .
t. takeda , m. nagata , h. kobayashi , r. kanno , y. kawamoto , m. takano , t .
kamiyama , f. izumi and a. w. sleight , j. solid state chem .
* 140 * , 182 ( 1998 ) .
m. tachibana , y. kohama , t. shimoyama , a. harada , t. taniyama , m. itoh , h. kawaji and t. atake , phys .
b * 73 * , 193107 ( 2006 )
. h. sakai , k. yoshimura , h. ohno , h. kato , s. kambe , r. e walstedt , t. d. matsuda , y. haga and y. onuki , j. phys .
: condens . matter * 13 * l785 ( 2001 ) .
n. tiara , m. wakeshima and y. hinatsu , j. phys .
: condens .
matter * 11 * , 6983 ( 1999 ) .
n. tiara , m. wakeshima and y. hinatsu , j. matter .
* 12 * 1475 ( 2002 ) .
m. ito , y. yasui , m. kanada , h. harashina , s. yoshii , k. murata , m. sato , h. okumura and k. kakurai , j. phys .
japan * 69 * , 888 ( 2000 ) ; m. ito , y. yasui , m. kanadaa , h. harashinaa , s. yoshii , k. murata , m. sato , h. okumura and k. kakurai , j. phys .
solids * 62 * , 337 ( 2001 ) .
m. rams , a. zarzycki , a. pikul , k. tomala , j. magn .
magn . mater . *
323 * , 1490 ( 2011 ) . c. n. r. wiebe , j. s. gardner , s. -j .
kim , g. m. luke , a. s.wills , b. d. gaulin , j. e. greedan , i. swainson , y. qiu , and c.y .
jones , phys .
* 93 * , 076403 ( 2004 ) .
n. taira , m. wakeshima , y. hinatsu , a tobo and k. ohoyama , j. solid state chem . * 176 * , 165 ( 2003 ) .
m. k. haas , r. j. cava , m. avdeev and j. d. jorgensen , phys .
b * 66 * , 094429 ( 2002 ) . s. lee , j. g. park , d. t. adroja , d. khomskii , s. streltsov , k. a. mcewen , h. sakai , k. yoshimura , v. i. anisimov , d. mori , r. kanno , and r. ibberson , nat .
mater . * 5 * , 471 ( 2006 ) .
k. s. lee , d. k. seo and m. h. whangbo , j. solid state chem .
* 131 * , 405 ( 1997 ) .
j. van duijn , k. h. kim , n. hur , d. adroja , m. a. adams , q. z. huang , m. jaime , s. -w .
cheong , c. broholm and t. g. perring , phys .
94 * , 177201 ( 2005 ) .
k. c. rule and p. bonville , j. phys .
: conf . ser .
* 145 * , 01202 ( 2009 ) .
y. machidaa , s. nakatsuji , h. tonomura , t. tayama , t. sakakibara , j. van duijn , c. broholm , and y. maeno , j. phys . chem .
solids * 66 * , 1435 ( 2005 ) .
d. l. martin , phys .
b * 8 * , 5357 , ( 1973 ) . see http://lise.lbl.gov/rspak . j. rodriguez - carvajal , physica b * 192 * , 55 ( 1993 ) . c. j. carlile and m. a. adams , physica b * 182 * , 431 ( 1992 ) . v. garcia sakai , m. a. adams , w. s. howells , m. t. f. telling , f. demmel and f. fernandez - alonso , _ ral technical reports _ ral - tr-2011 - 004 ( 2011 )
. s. w. lovesey , _ theory of neutron scattering from condensed matter _
( clarendon press , oxford , 1984 ) .
p. j. brown , in _ international tabled for crystallography _ , vol .
c , edited by a.j.c . wilson ( kluwer academic , london , 1995 ) .
j. jensen , j. magn .
magn . mat .
* 29 * , 47 ( 1982 ) .
hutchings , solid state phys . *
16 * , 227 ( 1964 ) .
p. fabi , _ isis technical report no .
_ ral - tr-95 - 023 ( 1995 )
. s .- h . lee , c. broholm , m. f. collins , l. heller , a. p. ramirez , ch .
kloc , e. bucher , r. w. erwin , and n. lacevic , phys .
b * 56 * , 8091 ( 1997 ) .
a. l. ankudinov , b. ravel , j. j. rehr and s. d. conradson , phys .
b * 58 * , 7565 ( 1998 ) .
m. avdeev , m. k. haas , j. d. jorgensen and r. j. cava , j. solid state chem .
* 169 * , 24 ( 2002 ) .
d. p. shoemaker , r. seshadri , m. tachibana and a. l. hector , phys .
064117 ( 2011 ) . l. downward , c. h. booth , w. w. lukens and f. bridges , aip conference proceedings * 882 * , 129 ( 2007 ) .
m. w. johnson , _ harwell report no .
_ aere - r7682 ( 1974 ) ( unpublished ) .
p. c. hohenberg and w. f. brinkman , phys .
b * 10 * , 128 ( 1974 ) . s. nakatsuji , y. machida , y. maeno , t. tayama , t. sakakibara , j. van duijn , l. balicas , j. n. millican , r. t. macaluso , and j. y. chan , phys . rev
. lett . * 96 * , 087204 ( 2006 ) .
m. loewenhaupt , m. prager , e. gratza and b. frick , j. magn .
magn . mater . *
76 - 77 * , 415 ( 1988 ) .
r. schedler , u. witte , m. loewenhaupt and j. kudla , physica b * 335 * , 41 ( 2003 ) .
j. c. nipko , m. grimsditch , c. -k .
loong , s. kern , m. m. abraham and l. a. boatner , phys . rev .
b * 53 * , 2286 ( 1996 ) . c. -k .
loong , m. loewenhaupt , j. c. nipko , m. braden and l. a. boatner , phys .
b * 60 * , r12549 , ( 1999 ) .
d. t. adroja , a. del moral , c. de la fuente , a. fraile , e. a. goremychkin , j. w. taylor , a. d. hillier , and f. fernandez - alonso , phys .
lett . * 108 * , 216402 ( 2012 ) .
a. s. wills , physica b * 276 * , 680 ( 2000 ) .
k. momma and f. izumi , j. appl . crystallogr . * 44 * , 1272 ( 2011 ) .
a. poole , a. s. wills and e. lelivre - berna , j. phys .
matter * 19 * , 452201 ( 2007 ) .
n. taira , m. wakeshima and y. hinatsu , j. phys . : condens .
matter * 11 * , 6983 ( 1999 ) .
n. taira , m. wakeshima and y. hinatsu , j. solid state chem . * 152 * , 441 ( 2000 ) .
j. s. lee , t. w. noh , j. s. bae , in - sang yang , t. takeda and r. kanno , phys .
b * 69 * , 214428 ( 2004 ) .
j. van duijn , n. hur , j. w. taylor , y. qiu , q. z. huang , s .- w .
cheong , c. broholm , and t. g. perring , phys .
b * 77 * , 020405 ( 2008 ) j. s. gardner , s. r. dunsiger , b. d. gaulin , m. j. p. gingras , j. e. greedan , r. f. kiefl , m. d. lumsden , w. a. macfarlane , n. p. raju , j. e. sonier , i. swainson , and z. tun , phys . rev . lett . * 82 * , 1012 ( 1999 ) .
m. j. harris , s. t. bramwell , d. f. mcmorrow , t. zeiske , and k. w. godfrey , phys .
lett . * 79 * , 2554 ( 1997 ) .
j. brooke , d. bitko , t. f. rosenbaum , g. aeppli , science * 284 * , 779 ( 1999 ) .
|
the complex structure and magnetism of pr@xmath0bi@xmath1ru@xmath2o@xmath3 was investigated by neutron scattering and exafs .
pr has an approximate doublet ground - state and the first excited state is a singlet .
this overall crystal field level scheme is similar to metallic pr@xmath2ir@xmath2o@xmath3 , which is also reported here .
while the b - site ( ru ) is well ordered throughout , this is not the case for the a - site ( pr / bi ) .
a distribution of the pr - o2 bond length indicates the pr environment is not uniform even for @xmath4 .
the bi environment is highly disordered ostensibly due to the 6s lone pairs on bi@xmath5 .
correspondingly we find the non - kramers doublet ground state degeneracy otherwise anticipated for pr in the pyrochlore structure is lifted so as to produce a quadrupolar singlet ground state with a spatially varying energy gap . for @xmath4 , below t@xmath6 , the ru sublattice orders antiferromagnetically , with propagation vector * * k**= ( 0,0,0 ) , as for y@xmath2ru@xmath2o@xmath3 . no ordering associated with the pr sublattice is observed down to 100 mk
. the low energy magnetic response of pr@xmath0bi@xmath1ru@xmath2o@xmath3 features a broad spectrum of magnetic excitations associated with inhomogeneous splitting of the pr quasi - doublet ground state . for @xmath4 ( @xmath7 )
the spectrum is temperature dependent ( independent ) .
it appears disorder associated with bi alloying enhances the inhomogeneous pr crystal field level splitting so that inter - site interactions become irrelevant for @xmath7 .
the structural complexity for the a - site may be reflected in the hysteretic uniform magnetization of b - site ruthenium in the nel phase .
|
magnetic equiatomic alloys ( meas ) of general formula mm@xmath6x ( m , m@xmath6 = transition metals , x = si , ge , sn etc . ) and their derivatives have attracted renewed attentions for the observation of fascinating magneto - functional and physical properties such as , large magneticaloric effect ( mce ) , exchange bias effect ( ebe ) , spin glass ( sg ) like ground state etc .
@xcite among various members of meas , mnnige is one of the potential candidates which undergoes a first - order diffusionless structural phase transition , known as martensitic phase transition ( mpt ) , at 470 k during cooling and orders spiral antiferromagnetically below 346 k. @xcite physical and chemical pressures are the two most influencing parameters that can affect the physical properties of these materials simply by modifying structural parameters like lattice volume , bond angle etc .
@xcite among these two , chemical pressure approach ( different doping studies ) is much popular among the researchers due to easy sample preparation and measurement options .
however , the purest form of perturbation can only be given by the physical pressure approach . since mea s discovery , several doping studies ( chemical pressure effect )
have been performed to increase the magneto - functional properties by reducing the structural transition temperature below its magnetic transition .
@xcite but very little efforts have been made to tune the magneto - functional properties by applying physical pressure on the mnnige system .
@xcite recent investigation on the fe - doped mnnige alloy ( fe doping both in mn and ni sites of the alloy ) by liu _
et al . _ indicates the presence of ferromagnetic ( fm ) and sg - like ground state in the system .
@xcite fe doping at the mn site of the mnnige layered compound not only breaks the spiral antiferromagnetic ( afm ) mn - ge - mn inter - layer interaction but also induces fm interaction between mn atoms of the same layer by reducing lattice volume of the alloy .
@xcite in addition , a reasonable decrease in mpt temperature of the alloy has been observed with fe - doping .
initial increase in fe - concentration ( up to 16% doping ) results enhancement of fm strength and hence the overall magnetic moment of the alloy .
fm strength again starts to decrease with further increase in fe concentration and a sg like ground state was observed .
@xcite the direct exchange nature of the mn - mn intra - layer interaction plays the pivotal role towards the observation of non - monotonic effect of fe - doping . in the present work ,
our aim is to address the effect of clean perturbation by applying external hydrostatic pressure on the structural and magnetic transition of a fe - doped mnnige alloy with nominal composition mn@xmath0fe@xmath1nige .
the polycrystalline sample of nominal composition mn@xmath0fe@xmath1nige was prepared by argon arc - melting the constituent elements ( purity @xmath7 99.9% ) . to make the alloy homogeneous ,
the ingot was turned over and re - melted for four times .
further the ingot was sealed in an evacuated quartz tube and annealed at 800@xmath8c for 100 h followed by a rapid quenching in ice water .
the sample was characterized by room temperature x - ray powder diffraction ( xrd ) measurement recorded in a bruker axs diffractometer ( d8-advance ) using cu k@xmath9 radiation .
the collected powder diffraction pattern was used for the rietveld refinement analysis by maud software to obtain different structural parameters .
@xcite rietveld refinement along with the xrd patterns of the studied sample are shown in the main panel of fig .
xrd analysis confirms that at room temperature , the sample crystalizes with hexagonal ni@xmath10in - type structure ( space group @xmath11 ) .
no noticeable contribution of any impurities has been detected in this pattern .
the lattice parameters of the hexagonal phase are found to be @xmath12 = 4.081 , and @xmath13 = 5.327 .
the important fitted parameters are @xmath14 = 2.4 , @xmath15 = 4.8% , and @xmath16 = 3.6 % . the inset of fig .
[ fig1 ] shows the layered structure of hexagonal mn@xmath0fe@xmath1nige alloy at room temperature obtained by using vesta software package .
the mn / fe , ni and ge atoms occupy 2a , 2d and 2c sites respectively .
@xcite the dc magnetization ( @xmath17 ) was measured using a quantum design squid magnetometer ( mpms 7 , evercool model ) .
high pressure measurements were performed using a mpms high pressure capsule cell 8.5 .
daphne was used as pressure transferring medium .
temperature ( @xmath18 ) variation of dc magnetization , measured at an applied magnetic field ( @xmath4 ) of 10 koe , both in field cooling ( fc ) and field - cooled heating ( fch ) protocols are depicted in the main panel of fig .
[ fig2 ] . during cooling , a sharp increase in the @xmath19 data around 200 k in association with a large thermal hysteresis between fc and fch data has been observed and consequently signify the occurrence of magneto - structural transition in the alloy .
presence of thermal hysteresis confirms the first order nature of the transition .
further decrease in temperature results a sluggish drop in @xmath17 value after reaching maximum just below the structural transition .
the decreasing nature of @xmath19 data indicates the presence of a short range afm phase in the otherwise fm alloy . to shed more light on the magnetic nature of the sample
, we recorded isothermal @xmath17 data as a function of @xmath4 at selected temperatures ( see inset of fig . [ fig2 ] ) .
all the isothermal @xmath20 data were recorded during heating at different constant temperatures in a thermally demagnetized state .
paramagnetic nature of the high temperature austenite phase is clear from the linear @xmath20 data at 250 k. however , closer view of the @xmath20 data reveals that , below and around mpt , there is a clear signature of change in slope in the @xmath20 isotherms around the low field region ( marked by @xmath21 ) .
the isotherms at 200 k and 150 k , when the sample is well in the martensitic state , are associated with field induced afm to fm transition , which starts at @xmath21 and found to be completed at @xmath22 ( marked in the inset of fig .
[ fig2 ] ) and the alloy becomes totally fm .
similar behavior has also been observed at 5 k. both @xmath21 and @xmath22 are found to increase with decreasing temperature .
this type of field induced transition from a spiral afm to canted fm state has already been reported for mnnige alloy .
@xcite magnetic moment at 50 koe of applied @xmath4 is also found to increase with decreasing temperature of measurement .
now , let us examine the effect of external hydrostatic pressure ( @xmath5 ) on the magnetic and magneto - functional properties of the studied alloy .
the main panel of fig .
[ fig3](a ) depicts the @xmath18 variation of fc and fch magnetization data at 10 koe of @xmath4 in presence of different external @xmath5 .
application of external @xmath5 has a drastic effect on the magnetic and magneto - structural properties of the alloy .
mpt becomes broad and is found to be shifted towards lower @xmath18 side with increasing external @xmath5 .
in addition , a closer look reveals that the mpt becomes a two step process with initial increase in @xmath5 up to 5 kbar .
this may be due to the decoupling of magnetic and structural transitions in the alloy .
further increase in @xmath5 results the merging of these two transitions again .
this de - coupling and coupling process is very clearly visible in the @xmath23 vs. @xmath18 data depicted in the inset of fig .
[ fig3](a ) .
this observation is may be due to the faster response of structural transition than the magnetic transition .
initial application of @xmath5 results a sharp decrease in structural transition temperature and gets saturated around 5 kbar of @xmath5 . on the other hand , magnetic transition temperature decreases slowly with increasing @xmath5 up to 12.5 kbar .
monotonic decrease in magnetic transition temperature helped to couple these two transitions again .
it is also clear from the @xmath19 data that the low-@xmath18 martensitic state is more sensitive to the external @xmath5 than the high-@xmath18 austenite phase .
the magnetic moment in the low-@xmath18 martensitic state decreases with initial increase in @xmath5 , whilst @xmath17 starts to increase with further increase in @xmath5 .
this anomalous behavior may be related to different fm and afm phase fractions present in various applied @xmath5 .
this anomalous behavior was further investigated by recording isothermal @xmath20 data at 5 k in presence of different @xmath5 ( see fig .
[ fig3](b ) ) .
a number of features are associated with these @xmath20 isotherms .
firstly , the saturation magnetic moment at 50 koe of applied @xmath4 is found to decrease with increasing @xmath5 .
about 25 emu / g decrease in @xmath17 has been observed for the application of 12.5 kbar of external @xmath5 .
closer look at the low field region reveals that the hydrostatic pressure has a very significant effect on the field induced afm to fm transition present in the martensitic phase of the system .
the low-@xmath4 region of the @xmath20 data indicates that the starting field @xmath21 of the field induced transition remains almost constant with increasing @xmath5 .
on the other hand , the saturation field @xmath22 , where this field induced transition finishes , decreases rapidly with increasing @xmath5 .
the nature of the @xmath20 curves suggest that the afm phase fraction decreases with increasing @xmath5 .
reduced saturation moment with enhancement of @xmath5 is observed only due to the decrease in the in - plane mn - mn direct exchange interaction ( by decreasing mn - mn intra layer distance ) and hence fm strength . similar decrease in saturation moment at 5 k
has also been observed with the increase of chemical pressure ( fe doping in the present case ) beyond certain level .
@xcite up to 16% of fe doping , a clear increase in the saturation moment has been observed . whilst , further increase in fe concentration results a decrease in saturation moment .
the presently studied alloy is a 15% fe doped alloy , which is very near to the composition with maximum moment , and shows similar behavior with external hydrostatic pressure as that of the chemical pressure .
tempted by the observation of strong hydrostatic pressure effect on the magnetic properties of the presently studied alloy , we have investigated different magneto - functional properties in presence of different external @xmath5 .
the mm@xmath24x alloys are popular for their large magneto - caloric effect ( mce ) .
change in entropy ( @xmath2 ) on application of external @xmath4 is a measure of mce , and it is often calculated from the magnetization data using maxwell s relation , @xmath25 where @xmath26 denotes the entropy change for the change in @xmath4 from @xmath27 to @xmath28 . @xcite it is important to be very much careful while calculating @xmath2 around first order phase transition by using maxwell s relation .
discontinuity of @xmath17 around the ideal first order phase transition may give rise to infinite value of @xmath29 and hence unphysical value of @xmath2 .
@xcite no such discontinuity has been observed for the present alloy around or below the transition [ resulting finite value of @xmath29 ] and hence allows us to use the maxwell s relation for the present alloy . for @xmath2 calculation ,
different @xmath20 isotherms were recorded during heating between 50 - 300 k with 5 k interval in thermally demagnetized state ( data with 1 k interval were recorded around the transition temperature ) .
this specific state was obtained by cooling the sample from room temperature to 5 k and then heated back to the respective temperature of measurements in zero magnetic field .
these @xmath20 data were then convoluted to obtain @xmath19 data at different constant @xmath4 . finally , the change in entropy due to the application of @xmath4 has been calculated ( by using equation 1 ) and depicted in fig . [ fig4](a ) .
the magnetic entropy change @xmath2 is found to be -21 j / kg - k for @xmath4 changing from 0 to 50 koe .
negative value of @xmath2 indicates that the observed mce for the present alloy is of conventional type .
the above process was repeated for different external @xmath5 to examine its effect on mce .
application of external @xmath5 results decrease in mce value with a clear shift of the peak position towards lower @xmath18 .
about 18 j / kg - k decrease in @xmath2 has been observed on application of 12.5 kbar of external @xmath5 .
in addition , a clear split in @xmath2 vs. @xmath18 curve has been observed with initial application of @xmath5 up to 5 kbar .
the mce curve again becomes a single peak structure on further increase in @xmath5 .
this behavior also supports the de - coupling and coupling of magnetic and structural transition observed in @xmath19 data ( hence @xmath23 vs. @xmath18 data ) .
for the development of mce based refrigerator , beside the peak value of @xmath2 one has to look for it s refrigeration capacity ( rc ) also .
rc for a sample is defined as rc = @xmath31 , where @xmath32 and @xmath33 are the source and sink temperatures respectively .
@xcite we have calculated rc for the present alloy in ambient condition as well as in presence of external @xmath5 . in ambient condition
rc value is 218 j / kg for @xmath4 changing from 0 to 50 koe . on the other hand , in presence of 12.5 kbar of external @xmath5 rc
is found to be 173 j / kg .
the rc values observed for the present alloy both in ambient and high pressure conditions are reasonably high and can be used for practical applications .
as the effect of external @xmath5 is similar to that of chemical doping ( chemical pressure ) observed from the magnetization data , the afm and fm interaction present in the alloy may become comparable with increasing @xmath5 which often gives rise to exchange bias effect ( ebe ) as observed in a 20% fe doped alloy .
@xcite ebe is defined as the shift in the center of the @xmath20 isotherm while cooled in presence of external @xmath4 .
@xcite ebe is particularly important because of its application in various technological fields , such as permanent magnets , magnetic recording media , sensors , and read heads .
@xcite different magnetic interfaces like , fm / afm , fm / sg and fm / ferrimagnet based multilayers play the pivotal role towards the observation of ebe .
@xcite we have investigated the ebe of the presently studied alloy by recording isothermal @xmath20 data at 5 k after cooling the sample from 300 k in presence of 500 oe cooling field ( @xmath34 ) at different @xmath5 as depicted in the main panel and upper inset of fig .
[ fig4](b ) ( only @xmath5 = 0 , 5 and 12.5 kbar data are plotted here for clarity ) .
all the @xmath35 isotherms were recorded between @xmath3650 koe , which is well above the technical saturation field of the sample and rules out the possibility of minor loop effect towards the observation of ebe . however
, a restricted range of @xmath362.5 koe is shown in fig .
[ fig4](b ) for proper visualization .
the zero field cooled @xmath35 isotherms are found to be perfectly symmetric with respect to the origin on both horizontal ( @xmath4 ) and vertical ( @xmath17 ) axes ( see fig . [ fig3](b ) ) .
the field cooled @xmath20 isotherm remains symmetrical at ambient condition ( see upper inset of fig .
[ fig4](b ) ) .
this indicates the absence of any ebe in the alloy . on the other hand ,
a clear shift in the centre of the hysteresis loop , both in @xmath4 and @xmath17 axis , has been observed in presence of external @xmath5 .
variation of shift in the field axis @xmath37 as function of external @xmath5 has been plotted in the lower inset of fig .
[ fig4](b ) .
a monotonic increase in the @xmath37 value has been observed with increasing @xmath5 value .
this pressure induced ebe is in line with the fact that the application of external pressure results decrease in intra layer mn - mn fm interaction and makes it comparable with the inter layer mn - mn afm interaction .
our magnetic studies of the mn@xmath0fe@xmath1nige alloy under different applied hydrostatic pressure ( clean perturbation ) bring out several fascinating physical and functional properties of the system .
initial application of @xmath5 results broadening and splitting of magneto - structural transition along with a clear decrease in the transition temperature .
further increase in external @xmath5 results merging of magnetic and structural transitions .
this de - coupling and coupling of magnetic and structural transitions is one of the unique features observed in the presently studied alloy .
unequal response of magnetic and structural transitions towards the external @xmath5 plays the pivotal role towards this observation . on the other hand , application of @xmath5 affects mn - mn intra - layer direct exchange interaction and hence considerable decrease in the saturation moment at 5 k has been observed with increasing @xmath5 . in addition , a clear decrease in technical saturation field has also been observed with increasing @xmath5 . apart from these physical properties , external hydrostatic pressure also has a strong effect on the magneto - functional properties of the alloy .
the large conventional mce observed for the present alloy in ambient condition decreases and the peak temperature found to be shifted towards the lower temperature with increasing @xmath5 .
the most important observation is the occurrence of ebe in presence of external @xmath5 .
the strength of the coexisting afm and fm interactions in the present alloy becomes comparable with increasing @xmath5 and plays the pivotal role towards the observation of ebe .
the present alloy is a good candidate for memory applications under high pressure .
authors would like to thank department of science and technology , india for low temperature high magnetic field facilities at ugc - dae consortium for scientific research , kolkata centre .
sc would also like to thank prof .
s. majumdar of iacs , kolkata for valuable discussions .
26ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( )
|
magnetic and magneto - functional behavior of a fe - doped mnnige alloy with nominal composition mn@xmath0fe@xmath1nige have been investigated in ambient as well as in high pressure condition .
the alloy undergoes first order martensitic phase transition ( mpt ) around 200 k and also shows large conventional magnetocaloric effect ( mce ) ( @xmath2 @xmath3 -21 j / kg - k for magnetic field ( @xmath4 ) changing from 0 - 50 koe ) around the transition in ambient condition .
application of external hydrostatic pressure ( @xmath5 ) results a shift in mpt towards the lower temperature and a clear decrease in the saturation moment of the alloy at 5 k. the peak value of mce is also found to decrease with increasing external @xmath5 ( @xmath3 18 j / kg - k decrease in @xmath2 has been observed for @xmath5 = 12.5 kbar ) .
the most interesting observation is the occurance of exchange bias effect ( ebe ) on application of external @xmath5 .
the competing ferromagnetic and antiferromagnetic interaction in presence of external @xmath5 plays the pivotal role towards the observation of @xmath5 induced ebe .
|
the reduction of critical materials present in consumer and industrial products in general , and in rare earth ( re ) -based permanent magnets in particular , is a pressing concern due to increasing demand and a decline and/or uncertainty in the supply chain of these materials .
dysprosium , judged to be the most critical of these elements @xcite , is substituted onto the nd site in @xmath6 ( 2 - 14 - 1 ) permanent magnets to increase the value of their coercive field and maximum operating temperatures @xcite .
depending on the application , the amount of dy added can be large , in the range of 1.4 - 8.7 wt% @xcite . the highest temperature applications , those that rely on the greatest quantities of dysprosium , are required by the rapidly emerging technologies of wind turbines , magnetically levitated transport , and traction motors for hybrid and electric cars @xcite .
though dy is indeed considered the most critical element , by many criteria nd is not that far behind on the list @xcite .
it is therefore imperative that new and innovative solutions are applied to the synthesis of permanent magnets that contain more abundant elements . to this end , recent work by pathak @xmath8 @xmath9 .
@xcite has shown that partial substitution of ce for nd and co for fe ( @xmath10 @xmath11 ) results in a permanent magnet with properties superior to those of dy - substituted 2 - 14 - 1 magnets for temperatures above 450 k. since 1 ) ce is by far the most abundant of the rare earth elements and 2 ) these new magnets simultaneously eliminate the need for dy and reduce the amount of nd used , they represent an exciting opportunity for the development of lower - cost high performance permanent magnets .
the crystal structure of @xmath6 was established as tetragonal ( space group @xmath12 , no .
136 ) as early as 1984 by the authors of refs . @xcite and @xcite .
there are two separate and inequivalent re sites ; one is a @xmath3 site and the other a @xmath4 site .
also present are six inequivalent fe sites and one unique b site .
past investigations as to the effects of chemical substitution of different re elements onto the nd site are numerous , see for example ref . @xcite and the references therein .
however , detailed , quantitative experimental studies on the effects of ce substitution of varying amounts in pure single crystal samples have not previously been performed .
abache and osterreicher @xcite studied aligned powders to determine the crystal structure , magnetic anisotropy , and spin reorientation temperature in @xmath0 for @xmath13 and @xmath14 along with a multitude of other compositional variations of @xmath6 achieved by varying the degree and type of substitutions on the re and transition metal ( tm ) sites .
however , no detailed study was made concerning the magnetic properties as a function of ce concentration in this work .
they also do not make any claims whatsoever as to the stability of the @xmath0 compound for @xmath15 .
more recently , density functional theory calculations of alam @xmath8 @xmath9 .
@xcite actually predict a phase segregation for @xmath16 in @xmath0 .
pathak @xmath8 @xmath9 .
@xcite have experimentally observed phase segregation under the conditions required for the fabrication of melt - spun ribbons .
information also of interest in the @xmath0 system is the site preference of the ce as there are two structurally distinct re sites .
previous theoretical predictions about ce site preference exist @xcite , but direct experimental evidence to determine whether ce in @xmath0 is preferentially associated with the larger @xmath3 site or the smaller @xmath4 site has previously been lacking .
[ table1 ] the @xmath0 system has recently come under much attention .
experimental investigations by pathak @xmath8 @xmath9 . discovered an anomaly evident in the plot of the @xmath17 lattice parameter vs. ce concentration in melt - spun @xmath0 ribbons for @xmath18 @xcite .
similar anomalies were evident in plots of coercivity ( @xmath19 ) , maximum energy product ( @xmath20 ) , and remnant magnetization ( @xmath21 ) as functions of ce concentration , along with the aforementioned phase segregation .
the same investigation yielded an extraordinary result : when co is substituted for fe by two atoms per formula unit in @xmath22 where @xmath23 , the curie temperature ( @xmath24 ) increased by 150 k while coercivity dropped only 22% as compared to an equivalent sample with no co doping , @xmath25 @xmath26 . the anisotropy field ( @xmath27 ) , saturation moment ( @xmath28 ) , and @xmath21 remained more or less unchanged compared to the @xmath29 case .
the possible applications of such excellent dy - free magnets spawns an exciting line of research focused on linking the microscopic magnetic properties in pure crystalline samples with the macroscopic magnetic properties of bulk permanent magnets . in the current work , we have employed single crystal growth techniques to synthesize single crystals of ce - doped 2 - 14 - 1 compounds .
we show here observations contrary to those of the melt - spun ribbons of ref .
@xcite for our samples grown under the slow crystallizing conditions of flux growth ; we are able to stabilize uniform single crystals of @xmath0 for values @xmath30 .
furthermore , we have used these single crystal specimens to experimentally determine that ce has a slight site preference in the 2 - 14 - 1 structure for the larger re @xmath3 site @xcite .
next , we have also measured important magnetic properties such as @xmath28 , @xmath24 , @xmath27 etc . for @xmath0 for the composition range @xmath30 .
finally , we conclude this report with first principles density functional theory calculations which help explain the causes of the changes in @xmath28 , @xmath24 , and @xmath27 with ce substitution .
single crystals of both @xmath6 and @xmath31 type materials were grown from nd - fe flux following the techniques originally reported by canfield @xmath8 @xmath9 . @xcite and further refined by saparov @xcite .
the starting materials were cuttings of high purity metals : ce ( ames laboratory , 99.99% ) , nd ( ames laboratory , 99.99% ) , fe ( alfa aesar , 99.98% ) , co ( alfa aesar , 99.95% ) and isotopically pure @xmath32 ( @xmath32 , ornl , 99.99+% ) .
the use of @xmath32 was necessary for neutron diffraction experiments as @xmath33 is exceptionally good at neutron capture .
appropriate stoichiometries of these elements ( see table [ table1 ] ) were loaded into ta crucibles ( 1.25 cm diameter , 7 cm length ) and sealed under 0.5 atm ar using an arc - melter .
a ta frit was placed above the starting materials to act as a filter during the centrifugation process .
the ta crucibles were subsequently sealed in quartz ampoules under @xmath341/3 atm ar .
the sealed ampoules were placed into a large box furnace and heated to 1190 @xmath35 over 12 h and held at that temperature for 24 h. the furnace was then cooled to 800 @xmath35 over 390 h , after which the samples were removed and the flux was decanted using a centrifuge .
large , single crystal specimens ( some on the order of @xmath341 g ) of the @xmath6 ( 2 - 14 - 1 ) phase ( fig .
[ figure1 ] , inset ) and smaller elongated prisms ( @xmath34 2 mm @xmath36 2 mm @xmath36 10 mm ) of the @xmath31 ( 1 - 4 - 4 ) phase were extracted from the crucibles .
the ultimate sizes of the 2 - 14 - 1 crystals were not influenced by the compositions of the fluxes with the exception of the largest ce concentration ( in this case , large crystals of the laves phase compound @xmath37 formed rather than the 2 - 14 - 1 phase ) .
the crystallographic faces of all phases present after decanting were well - defined . in fig .
[ figure1 ] we present xrd reflections resulting from an ( 0 0 1 ) face of the undoped @xmath6 sample .
analyses of elemental compositions were performed using a hitachi tm-3000 electron microscope equipped with a bruker quantax 70 energy dispersive ( eds ) x - ray system ( @xmath38 table [ table1 ] ) .
for all compositions small crystals were ground into powder and used to produce xrd patterns using a panalytical xpert pro diffractometer ( cu @xmath39 , 1.54056 @xmath40 ) .
the resultant peak reflections were analyzed for lattice parameters using the lebail fitting function available with the onboard highscore plus software package .
single crystal diffraction was performed using a rigaku single crystal x - ray diffractometer with dectris pilatus 200k detector ( mo @xmath39 , 0.71073 @xmath40 ) ; the resultant diffraction patterns were analyzed using the onboard rigaku software with absorption corrections performed using a spherical approximation .
the structure was refined using shelxl ( ref .
@xcite ) with wingx .
single crystal neutron diffraction was measured using the hb-3a four circle diffractometer at the high flux isotope reactor ( hfir ) at the oak ridge national laboratory .
the @xmath34300 mg sample was measured at 4 k and 300 k by neutrons with a wavelength of 1.003 @xmath40 from a bent si-331 monochromator @xcite .
the structure was refined using the fullprof software package @xcite .
x - ray photoelectron spectroscopy ( xps ) data were collected with a phi 3056 xps spectrometer with an al @xmath39 source ( 1.4866 kev ) in a cryo - pumped ultra - high vacuum chamber with a pressure of @xmath41 torr .
fresh surfaces of the crystal were created by grinding a single crystal in an ar - filled glove box and transferring to the xps chamber using a vacuum transfer system .
synchrotron data was taken at the advanced light source ( als ) at the lawrence berkeley national laboratory ( berkeley , usa ) using beamline 9.3.1 .
this beam - line is a bent magnet beam - line with an energy range of [email protected] kev .
the minimal spot size at the beam - line is 0.7 mm ( v ) @xmath36 1.0 mm ( h ) @xcite .
magnetic properties were measured using 1 ) a quantum design magnetic property measurement system ( mpms ) and 2 ) the dc - extraction capability of the ac measurement system ( acms ) option of a 14 t quantum design physical property measurement system ( ppms ) . to measure the magnetization along the easy direction as a function of magnetic field 1 )
a @xmath17 axis face was found on a single crystal using the x - ray diffractometer , 2 ) the crystal was polished into a parallelepiped with a long axis aligned with either the @xmath43 or @xmath17 direction ( for hard and easy axis measurements , respectively ) and 3 ) the crystal segments were placed into the mpms or ppms under magnetic fields of up to 13 t. due to the tendency for the easy axis to align with the applied field direction , single crystal specimens were unable to quantitatively yield correct hard axis magnetization measurements . to obtain these measurements , small crystals of each composition
were ground into powder and placed into a gel cap with epoxy ; the epoxy was allowed to set under an applied magnetic field so as to align the powders .
measurements were taken with @xmath44 oriented perpendicular to the easy direction at 5 k , 300k , 350 k , and 400 k. to obtain a correct value for the saturation magnetization a measurement of the same gel cap was taken with @xmath44 oriented parallel to the easy direction ; these data were then normalized to the values obtained from the single crystal easy axis measurements .
the easy axis measurements of the aligned powder samples and the single crystal specimens were in good agreement .
the @xmath45 properties of the 2 - 14 - 1 crystals were also measured using the quantum design mpms .
the magnetization was measured as a function of temperature using an applied magnetic field of @xmath46 oe ; temperature ranges were 2 k @xmath47 750 k. further confirmation of the curie temperatures of the grown ferromagnetic crystals was obtained through thermal analysis using a perkin elmer pyris diamond thermo - gravimetric analyzer/ differential thermal analyzer ( tga / dta ) with an applied magnetic field ( @xmath48 0.1 t ) .
the first principles calculations are carried out within density functional theory ( dft ) .
we use the generalized gradient approximation @xcite within the projector augmented wave method @xcite as implemented in the vienna ab initio simulation package @xcite .
an energy cutoff of 400 ev and @xmath49 space sampling on a @xmath50 grid are employed .
the ce substitution on the nd site is handled by the virtual crystal approximation ( vca ) .
[ table2 ] [ table3 ]
the as - grown crystals were roughly cubic in shape and displayed well - defined faceting ( fig .
[ figure1 ] , inset ) .
the sizes of the 2 - 14 - 1 crystals were typically @xmath51 mm@xmath52 with some reaching @xmath53 mm@xmath52 in volume . except for the growth with a nominal flux composition of 18.1% nd and 32.9% ce where a laves phase compound formed rather than the 2 - 14 - 1 phase , all resulting single crystals were of similar size .
the 1 - 4 - 4 phases grew as needles @xmath54 mm@xmath52 and will be the subject of a forthcoming publication . through eds analysis
, we were able to measure the relative concentrations of the rare earths ( re ) and transition metals ( tm ) .
the boron content could not be reliably estimated from the eds measurements .
a variety of growths were used to maximize the ce concentrations within the crystals .
results suggest that @xmath5 is the maximum attainable by this synthesis route .
the room temperature powder xrd patterns were taken of the 2 - 14 - 1 phases using the powder from ground crystals .
lebail fitting was used to extract the lattice parameters , which are plotted in fig .
[ figure2 ] as functions of ce concentration .
the values of both the @xmath43 and @xmath17 lattice parameters decrease monotonically with increasing quantities of ce ( @xmath38 fig .
[ figure2 ] ) . at @xmath1 in @xmath0
we see that the value of a is 0.14% smaller than that of the undoped sample .
similarly , for the same composition , @xmath17 is 0.22% smaller .
no anomalies are present in the lattice parameters as a function of ce content , in contrast with the sintered magnets of the same composition synthesized by pathak @xmath8 @xmath9 .
@xcite where a two - phase region was observed for @xmath55 in @xmath0 .
however , it should be noted that there is a wide spacing between our data points which could potentially belie small anomalies in the lattice parameters as a function of ce content .
our single crystal specimens containing the two highest compositions of ce ( @xmath56 and @xmath1 ) fall within the two phase region suggested by the above authors , indicating that the range of solid solution in this system is sensitive to processing conditions such as temperature and solidification rate .
single crystal x - ray diffraction on the sample with composition @xmath57 was performed at 250 k and 110 k ; these temperatures bracket the spin reorientation temperature of @xmath34140 k in @xmath6 @xcite .
structural refinements were made using shelx @xcite from 1068 and 966 unique reflections , respectively , and are presented in tables [ table2 ] and [ table3 ] .
the space group matched that of previous reports and the powder diffraction data , @xmath12 .
the values from the single crystal data match fairly well with those obtained from powder diffraction .
no discernible structural difference was seen between the data collected above the spin reorientation temperature and that collected below it . for the single crystal analyses no attempts were made to refine the re position in terms of nd /
ce occupancy as the scattering factors of the two re elements were too similar .
later refinements using the occupancy results elucidated from the neutron diffraction data ( below ) yielded no change in the refinement .
the structure derived from the 250 k single crystal refinement is presented in figs .
[ figure3]a@xmath42[figure3]c and matches well with those published in the literature @xcite . :
a ) isometric view showing the positions of the two different re sites : @xmath3 and @xmath4 ( the larger @xmath3 site is denoted by a red label ) ; b ) view along a axis ; c ) view along c axis.,scaledwidth=45.0% ] single crystal neutron diffraction was performed on a @xmath34300 mg sample of @xmath57 at temperatures of 300 k and 4 k. we collected @xmath34630 reflections at each temperature .
the magnetic signals were found at both 4 k and 300 k with a propagation vector of @xmath58 , i.e. , they coincide with the nuclear peaks .
therefore , in the data refinements , the nuclear and magnetic structure parameters were refined together .
the results are presented in tables [ table4 ] and [ table5 ] .
given the fact that the easy axis of this material is along the direction , the magnetic vectors at 300 k were constrained as such to reduce the total number of refined parameters . at 300
k the total refined magnetic moment is @xmath59 @xmath7/f.u . , in reasonable accord with the mpms - determined value of 32.3 @xmath7/f.u .
the fe sites have moments ranging from 1.3(1)@xmath423.2(2 ) @xmath7 .
this variability is likely caused by the different steric environments experienced by each of the different fe sites . with the singular exception of the @xmath60 fe site the size of the moments
are correlated to the degree of fe coordination around the fe atoms .
for example , two of the largest moments are seen on the @xmath61 and @xmath62 sites which have fe coordinations of 12 and 10 , respectively .
the smallest moment is observed at the fe @xmath63 site where the highest re coordination is present , with 4 of the 12 nearest neighbors being nd / ce .
the two re sites have moments of 2.372 and 1.605 @xmath7 for @xmath3 and @xmath4 , respectively .
the @xmath3 site is , in terms of size , the larger of the two and is therefore predicted to be preferred for ce , assuming the ce is in the + 3 oxidation state @xcite .
the ce@xmath64 ion ( 115 pm diameter ) is larger than the ce@xmath65 ion ( 101 pm ) and is therefore likely to prefer this larger site .
when requiring that the thermal displacement parameters for the two different rare earth sites be equal , our 300k refinements presented in table [ table4 ] show that this preference is not absolute in that ce site preference for the @xmath3 site is 70% . if , however , we let the atomic displacement parameters for each of the re sites refine independently , the preference of ce for the @xmath3 site becomes even more absolute at 86% .
though the degree of ce preference for the @xmath3 site and the atomic displacement parameters are highly correlated in these refinements , the fact that ce does prefer the @xmath3 site does not change .
these observations concur with the preliminary xps data taken at the advanced light source ( als ) at the lawrence berkeley national laboratory whereby the xps spectrum for the ce @xmath66 electrons most closely resembles that of @xmath67 , indicating that the majority of ce is in the + 3 oxidation state for this composition .
further experimentation is necessary to confirm this initial experimental result .
for the 4 k refinement , we fixed the ce occupancy to the values found at 300 k ( 70% on the re @xmath3 site and 30% on the re @xmath4 site ) and let the magnetic vectors vary in direction , as below @xmath34140 k the easy axis is no longer the direction . at 4 k , the total magnetic moment for the sample is refined to be @xmath68 @xmath7 in reasonable agreement with the experimentally determined value of 34.9 @xmath7/f.u . for this composition ( below ) .
the fe sites have moments that refine in the range @xmath69 .
again , the moments associated with the fe sites largely scale with their degrees of coordination by other fe atoms .
the refinement also shows that at 4 k the moments are tilted by @xmath3439.0@xmath70 , consistent with the well - known phenomena of spin reorientation in the 2 - 14 - 1 compounds at @xmath34140 k @xcite whereby the magnetic moments orient themselves along the direction rather than the direction seen at room temperatures @xcite .
the angle between these two directions is 49.6@xmath70 , varying slightly from the experimentally determined value of 39.0@xmath70 from our refinement .
however , to date no work has been done to quantify the effect of ce substitution on this reorientation with respect to changes in the easy direction of magnetization . using single crystal neutron diffraction ,
the ( 0 0 6 ) reflection of @xmath0 was measured upon warming ( fig.[figure4 ] ) . referring to fig .
[ figure4 ] , we see at 130.5 k a noticeable increase in the ( 0 0 6 ) peak intensity is evident , representing the spin reorientation transition whereby the magnetization easy axis shifts from the direction at high temperatures to something near the direction at low temperatures .
previous investigations of undoped @xmath6 by abache and oesterreicher @xcite attributed this re - orientation to competing anisotropies of the @xmath3 and @xmath4 re sites . in their argument , the @xmath4 site ( the smaller of the two ) is more susceptible to planar moment alignment . at temperatures above the spin re - orientation temperature ( @xmath71 ) the axial anisotropies of the re and tm sites force the @xmath4 site to align axially in a ferromagnetic structure
however , below 140 k the preference of the @xmath4 site to align its moment in a planar orientation is much stronger compared to the competing influences of the fe sites , resulting in spin re - orientation and a new easy axis for the system about 49@xmath70 canted away from the @xmath17 axis @xcite .
the spin re - orientation is therefore dependent on the @xmath4 site .
the related compound @xmath72 exhibits no spin re - orientation @xcite implying that ce has no inclination for planar orientation of its moment . substituting ce on the @xmath4 site
is expected to affect the spin - reorientation . .
, scaledwidth=45.0% ] in fig .
[ figure5]a we plot the magnetization of undoped @xmath6 as a function of temperature and see increases in magnetization at 586 k and 143.5 k associated with @xmath24 and @xmath71 , respectively . in fig .
[ figure5]b we plot the evolution of @xmath71 as a function of ce doping . from the neutron diffraction refinements we know that 18% of the re @xmath4 site is substituted by ce for the sample with the composition @xmath57 .
presumably similar levels of ce are found at this site for the other ce compositions investigated concomitant with the level of ce introduced .
we see from fig .
[ figure5]b that @xmath71 is sensitive to the amount of ce present , indicating that although ce does prefer the @xmath3 site a not insignificant amount of ce is still substituting onto the @xmath4 site , assuming that the only way to modify the spin re - orientation temperature is to alter the occupancy of the @xmath4 site .
this contrasts with the prediction of alam @xmath8 @xmath9 .
whereby ce was predicted to sit on only the @xmath3 site @xcite .
relevant magnetic properties are summarized in table [ table4 ] .
the curie temperatures of the ce - substituted 2 - 14 - 1 samples were measured via mpms and dta / tga . a permanent magnet affixed to the exterior of the sample chamber
were used to produce magnetic field gradient at the sample , so that a change in magnetic susceptibility produced a change in the apparent sample mass .
the data from these analyses show a steady monotonic decrease in @xmath24 with increasing ce concentration .
the undoped @xmath6 yields a @xmath24 of @xmath73 584 k. the composition @xmath74 , on the other hand , shows a @xmath24 of @xmath73 544 k , a total decrease of 1.1% ce substituted .
this result is less than the rate one would expect if a vegard - like relation existed between the @xmath6 and @xmath72 ( @xmath24 @xmath73 430 k @xcite ) parent compounds . in figs .
[ figure6]a and [ figure6]b we show the @xmath75 curves for the four ce - substituted samples at 300 k and 400 k , respectively .
with 38% ce doping , the saturation magnetization , ms , decreases 9.2% at 300 k ( 30.5 @xmath7/f.u . for @xmath76 vs. 33.6 @xmath7/f.u . for the undoped @xmath6 ) . comparing these values to the 300 k saturation magnetizations of the parent compounds ( 23.9 @xmath7/f.u . for @xmath72 @xcite ) we find that the value of @xmath28 seems to linearly scale with composition .
the effects at higher temperatures are similar . at 400 k , for the same composition @xmath28
is decreased by 7.4% from the undoped value .
concomitant with the decreases in @xmath28 with increasing ce concentration are decreases in @xmath27 as determined from extrapolating the hard axis magnetization curves in figs .
[ figure6]a and [ figure6]b to the saturation magnetization of the easy axis magnetization curves . at 300
k the value of @xmath27 drops from an undoped value of 8.1 t to a value of 6.2 t for the highest achievable ce concentration . at 400
k @xmath27 drops from 5.5 t to 4.7 t for crystals with large amounts of ce present .
the @xmath77 class of materials ( where r = nd , gd , y etc . ) has been the subject of much study @xcite . with respect to @xmath6 ,
first principles studies have been used to understand the effect of partial substitution of the nd site by various rare earth elements ( e.g. , y , la , dy , tb etc .
@xcite ) as well as the fe site by si , ge , sn etc.@xcite . in this report
we calculate the effects of ce substitution for nd on the magnetic properties of @xmath0 , focusing on the parameters relevant for a high performance magnetic material : the saturation magnetization , curie temperature and magnetocrystalline anisotropy .
@xmath6 crystallizes in a tetragonal structure with four formula units per unit cell . in the atomic valence configuration
the four electrons in the nd-@xmath4 orbitals and the six in the fe-@xmath66 orbitals make these ions strongly magnetic .
the density of states ( dos ) for @xmath6 in the ferromagnetic state is shown in the blue curves in fig .
[ figure7 ] . comparing the total dos with partial dos in the lower panels
, we can see that the states near the fermi level predominantly consist of fe-@xmath79 and nd-@xmath80 states .
there is an exchange splitting of @xmath342 ev in the fe-@xmath79 states with a nearly fully occupied spin - up channel .
the spin down channel has close to two electrons , suggesting a spin magnetic moment of @xmath343 @xmath7 at the fe sites , which is significantly above the value for bcc fe ( @xmath342.2 @xmath7/fe ) . for comparison
, the experimental neutron diffraction data collected at 4 k shows fe having moments between 2.370 and 3.253 @xmath7 , within the range of these extrema .
( middle panel ) and nd / ce-@xmath80 ( bottom panel ) states.,scaledwidth=45.0% ] at the nd sites , the @xmath4 states are empty in the spin - up channel and are partially occupied in the spin - down channel , showing that in this calculation the nd moments point in the opposite direction compared to the fe moments .
the exchange splitting at the nd-@xmath80 dos is similar to that of fe-@xmath79 orbitals , even though the@xmath81 dos is substantially narrower .
counting the states in the nd-@xmath82 channel up to the fermi level gives 3.1 electrons , which suggest a spin magnetic moment of -3.1 @xmath7 at the nd sites . from our calculations
, we also find an orbital moment of 2.8 @xmath7 at the nd-@xmath80 states ( opposite to the spin moment , as expected from hund s rules ) , while the orbital moments at the fe sites are @xmath340.06 @xmath7 .
the much larger orbital moment on the rare - earth atom suggests that the magneto - crystalline anisotropy is primarily a result of the rare - earth atom .
spin magnetic moments calculated by integrating charge density around the atom sites agree well with the above estimates . integrating the spin - polarized charge density around the ions , we find average moments of ( in @xmath7 ) -3.4 , 2.5 and -0.17 for nd , fe and b sites respectively .
the fe moment value is larger than that for pure fe , contributing to the excellent magnetic properties of this compound . for comparison , the experimental neutron diffraction data collected at 4 k shows fe having net moments between 2.370 and 3.253 @xmath7 , in good agreement with calculated value .
consistent with the experimental neutron diffraction data , we find that the fe ions bonded with b atoms were found to have a reduced moment of 1.98 @xmath7 .
the net spin+orbital ( j = s+l ) magnetic moment of this system is 30.3 @xmath7 per unit cell , in good agreement with both the neutron diffraction refinements and the magnetization data .
to understand the effect of ce doping at the nd sites on the properties of @xmath6 , we replaced 10% of the nd atoms by ce using the virtual crystal approximation ( vca ) .
the total and partial dos for the doped system are presented as red curves in fig .
[ figure7 ] .
as we can see , the only significant change in the dos is at the re-@xmath83 states , which are shifted to lower energies by @xmath340.5 ev .
thus , we see that the electronic properties of the nd - fe - b magnets are not significantly affected by ce doping , except a small reduction in the magnetic moments at the nd / ce site since ce has two fewer electrons in its valence shells ( @xmath84 ) than nd ( @xmath85 ) . thus , the number of electrons in the 68 atom ( 8 rare earth atoms ) unit cell is reduced by 1.6 electrons by 10% ce substitution which lead to a subsequent reduction in the average magnetic moment at these sites to 3.2 @xmath7 from 3.4 @xmath7 .
we find from these spin - polarized calculations that the fe moments are mostly unaffected by the substitution . an important consequence of the reduced exchange splitting is a reduction in curie temperature , @xmath24 .
generally , for a ferromagnetic system characterized by large local moments such as the 2 - 14 - 1 materials , the curie temperature is controlled by the difference in energy between the ferromagnetic ground state and an essentially antiferromagnetic structure constructed so that most of the neighbors of fe atoms are antiparallel .
this in turn is a function of the exchange parameters @xmath86 connecting nearest neighbors @xmath87 and @xmath88 . in the simplest mean - field approximation one may estimate the curie temperature as a third of this energy difference , measured on a per fe basis .
we calculated the energy difference between the above - described ferromagnetic structure and an approximate antiferromagnetic structure obtained by flipping most of the nearest neighbor moments .
the calculated energy difference is plotted in fig .
[ figure8 ] as a function of the ce concentration@xmath2 .
the @xmath89 value of -2.71 ev per formula unit corresponds in the mean - field approximation to a @xmath24 of 748 k , which is to be compared to the experimental value of 585 k. the smaller experimental value suggests that effects beyond our simple mean - field approach may be relevant .
a similar calculation overestimate was also found in ref .
@xcite . upon substituting ce atoms via vca ,
the magnitude of the energy difference decreases by about 4 percent for a 10 percent ce substitution .
this suggests that the mean field @xmath90 decreases in @xmath6 as @xmath2 is increased which would in turn decrease the curie temperature .
this is consistent with our experimental results , which also find a decrease in curie temperature with ce alloying . as function of doping parameter @xmath2 .
the energy difference reduces upon ce doping , suggesting that the exchange interaction @xmath90 and consequently @xmath24 will reduce as a function of ce concentration . , scaledwidth=45.0% ] finally , we calculated the magnetic anisotropy energy ( mae ) of @xmath0 for ce concentrations @xmath91 and @xmath92 to understand the effect of doping on this important parameter .
we used the experimental structure at room temperature ( i.e. , in the ferromagnetic state ) to perform these calculations @xcite .
the total energy is then calculated in the presence of spin orbit coupling with the moments pointing along @xmath43 and @xmath17 directions .
the mae is defined as @xmath93 .
we find that for @xmath94 , @xmath95 = 12.3 mev / f.u .
( @xmath348 mj / m@xmath52 ) .
this is somewhat larger than the experimental value of 4.3 mj / m@xmath52 , but we note that temperature effects are known to be significant in this compound @xcite .
we find that the anisotropy energy remains more or less same upon doping ; we get @xmath95 = 13.2 mev / f.u .
when @xmath92 .
we do not consider the increase to be significant , but expect that larger levels of ce alloying would tend to show a reduction in the value of the mae .
in short , we find that that the high magnetic anisotropy energy of the nd - fe - b system is not significantly affected by the ce doping , which confirms experimental observations .
large single crystals of @xmath6 were grown from nd - fe flux .
these crystals were a maximum of @xmath340.5 g in mass and displayed sharp faceting .
cerium substitution was attempted for nd ; the maximum degree of substitution we were able to achieve was the composition @xmath76 .
we found that ce and nd formed a continuous solid solution in contrast to the melt - spun ribbons of similar compositions fabricated by pathak @xmath8 @xmath9 .
@xcite where a two phase region was observed for @xmath96 in @xmath0 .
single crystal neutron diffraction showed that the ce has a slight preference for the larger re @xmath3 site , favoring this location by a factor of @xmath347:3 over the re @xmath4 site .
the presence of ce serves to deleteriously affect magnetic properties such as @xmath24 , @xmath28 , and @xmath27 .
the values of these characteristics largely fall in line with empirical predictions based on the application of simple alloying rules between the two parent compounds ( @xmath6 and @xmath72 ) .
we have also used x - ray photoelectron spectroscopy and synchrotron emissions to determine the oxidation states and binding energies of all species in ce - substituted @xmath6 ; preliminary data shows that the xps spectrum for the ce 3@xmath79 electrons most closely resembles that of @xmath67 . with the help of first - principles calculations
, we studied the electronic and magnetic properties of ce doped nd - fe - b magnets .
we find that for up to @xmath97 doping of ce , the electronic structure around the fermi level is not significantly affected .
the exchange splitting and the magnetic moments at the nd sites are diminished as a consequence of ce doping , which in turn reduces the curie temperature .
finally , we find that the high magnetic anisotropy energy of the nd - fe - b system is not significantly altered by the ce doping , confirming the experimental observations .
llccccc & & & & & & + & & & & + & & & & & & + + @xmath98 & b1 ( @xmath3 ) & & & & & + @xmath99 & fe1 ( @xmath62 ) & & & & & + @xmath100 & fe2 ( @xmath101 ) & & & & & + & fe3 ( @xmath102 ) & & & & & + @xmath12 & fe4 ( @xmath61 ) & & & & & + & fe5 ( @xmath60 ) & & & & & + & fe6 ( @xmath63 ) & & & & & + & nd1 ( @xmath3 ) & & & & & + & ce1 ( @xmath3 ) & & & & & + & nd2 ( @xmath4 ) & & & & & + & ce2 ( @xmath4 ) & & & & & + & & & & & & + & & & & + & & r ( @xmath7 ) & _ @xmath103 _ & _ @xmath104 _ & & + + * total moment * & fe1 ( @xmath62 ) & 2.869(137 ) & 0 & 0 & + @xmath105 & fe2 ( @xmath101 ) & 2.247(135 ) & 0 & 0 & + & fe3 ( @xmath102 ) & 1.614(169 ) & 0 & 0 & + & fe4 ( @xmath61 ) & 2.268(182 ) & 0 & 0 & + & fe5 ( @xmath60 ) & 3.247(159 ) & 0 & 0 & + & fe6 ( @xmath63 ) & 1.305(137 ) & 0 & 0 & + & nd1 ( @xmath3 ) & 2.372(119 ) & 0 & 0 & + & ce1 ( @xmath3 ) & 2.372(119 ) & 0 & 0 & + & nd2 ( @xmath4 ) & 1.605(118 ) & 0 & 0 & + & ce2 ( @xmath4 ) & 1.605(118 ) & 0 & 0 & + & & & & & & + + * reliability factors * & & & & & & + & @xmath106 & @xmath107 & @xmath108 & @xmath109 & & + & 0.0473 & 0.0775 & 0.0461 & 4.06 & & + & & & & & & + llccccc & & & & & & + & & & & + & & & & & & + + @xmath110 & b1 ( @xmath3 ) & & & & & + @xmath99 & fe1 ( @xmath62 ) & & & & & + @xmath100 & fe2 ( @xmath101 ) & & & & & + & fe3 ( @xmath102 ) & & & & & + @xmath12 & fe4 ( @xmath61 ) & & & & & + & fe5 ( @xmath60 ) & & & & & + & fe6 ( @xmath63 ) & & & & & + & nd1 ( @xmath3 ) & & & & & + & ce1 ( @xmath3 ) & & & & & + & nd2 ( @xmath4 ) & & & & & + & ce2 ( @xmath4 ) & & & & & + & & & & & & + & & & & + & & r ( @xmath7 ) & _ @xmath103 _ & _ @xmath104 _ & & + + * total moment * & fe1 ( @xmath62 ) & 3.288(76 ) & 226.09(569 ) & 39.00(180 ) & + @xmath111 & fe2 ( @xmath101 ) & 2.878(76 ) & 199.04(510 ) & 39.00(180 ) & + & fe3 ( @xmath102 ) & 3.378(115 ) & 156.65(527 ) & 39.00(180 ) & + & fe4 ( @xmath61 ) & 3.528(92 ) & 191.39(471 ) & 39.00(180 ) & + & fe5 ( @xmath60 ) & 3.302(103 ) & 240.22(706 ) & 39.00(180 ) & + & fe6 ( @xmath63 ) & 2.819(131 ) & 210.84(711 ) & 39.00(180 ) & + & nd1 ( @xmath3 ) & 2.921(70 ) & 255.63(732 ) & 39.00(180 ) & + & ce1 ( @xmath3 ) & 2.921(70 ) & 255.63(732 ) & 39.00(180 ) & + & nd2 ( @xmath4 ) & 3.644(96 ) & 185.29(536 ) & 39.00(180 ) & + & ce2 ( @xmath4 ) & 3.644(96 ) & 185.29(536 ) & 39.00(180 ) & + & & & & & & + + * reliability factors * & & & & & & + & @xmath106 & @xmath107 & @xmath108 & @xmath109 & & + & 0.0459 & 0.0948 & 0.0648 & 0.0738 & & + a. k. pathak , m. khan , . a. gschneidner , jr . , r. w. mccallum , l. zhou , k. sun , k. w. dennis , c. zhou , f. e. pinkerton , m. j. kramer , and v. k. pecharsky .
cerium : an unlikely replacement of dysprosium in high performance nd - fe - b permanent magnets .
, page ahead of print , 2015 .
s. axnanda , e. j. crumlin , s. rani , z. hussain , b. mao , r. chang , p. g. karlsson , marten o. m. e. , m. lundqvist , r. moberg , p. ross , and z. liu . using `` tender '' x - ray ambient pressure x - ray photoelectron spectroscopy as a direct probe of solid - liquid interface .
, 5:9788 , 2015 .
|
single crystals of @xmath0 are grown out of fe-(nd , ce ) flux .
chemical and structural analysis of the crystals indicates that @xmath0 forms a solid solution until at least @xmath1 with a vegard - like variation of the lattice constants with @xmath2 .
refinements of single crystal neutron diffraction data indicate that ce has a slight site preference ( 7:3 ) for the @xmath3 rare earth site over the @xmath4 site .
magnetization measurements show that for @xmath5 the saturation magnetization at 400 k , a temperature important to applications , falls from 29.8 for the parent @xmath6 to 27.6 @xmath7/f.u . , the anisotropy field decreases from 5.5 t to 4.7 t , and the curie temperature decreases from 586 to 543 k. first principles calculations carried out within density functional theory are used to explain the decrease in magnetic properties due to ce substitution . though the presence of the lower - cost and more abundant ce slightly affects these important magnetic characteristics , this decrease is not large enough to affect a multitude of applications .
ce - substituted @xmath6 is therefore a potential high - performance permanent magnet material with substantially reduced nd content .
|
neutrinos have long been anticipated to help answering some fundamental questions in astrophysics like the mystery of the source of the cosmic rays ( for a general discussion see @xcite ) .
for neutrinos in the tev range , prime source candidates are galactic supernova remnants @xcite .
neutrinos in the pev range and above are suspected to be produced by active galactic nuclei ( agn ) and gamma ray bursts ( grb ) with many agn models predicting a significant neutrino flux @xcite .
recently , the icecube collaboration has reported the very first observation of a cosmic diffuse neutrino flux which lies in the 100 tev to pev range @xcite .
individual sources , however , could not be identified up to now .
while many astrophysical sources of origin have been suggested @xcite , there is yet not enough information to narrow down the possibilities to any particular source . due to the low interaction probability of neutrinos
, a large amount of matter is needed in order to detect them .
one of the detection techniques is based on the observation of inclined extensive air showers ( eas ) induced by taus from tau neutrino interactions deep in the atmosphere . as these showers are initiated close to the surface of the earth , they are still very young when reaching the detector and hence have a significant electromagnetic component leading to a broad time structure of the detected signal .
in contrast , showers from cosmic - ray nuclei are induced in the upper atmosphere and therefore have a strongly reduced electromagnetic component when reaching the detector . however , because of its low density , neutrino interactions are not very likely to happen inside the atmosphere .
a solution to this problem is to look for so - called earth skimming ( up - going ) tau neutrinos @xcite which interact within the earth or a mountain and produce a tau .
for neutrino energies of about a eev , the charged leptons have a range of a few kilometers and hence may emerge from the earth or mountain , decay shortly above the ground and produce eas detectable by a surface detector . in some cases , two consecutive eass might be observable , one coming from a tau neutrino interaction close to the surface and one from the decay of the resulting tau lepton .
these two showers , coming from the same direction in a time interval corresponding to the tau decay time , can generate a unique signature in the detector called a double - bang event @xcite .
the detection of such a double - bang event would be very important both from the astrophysical and the particle physics point of view , as it would be an unambiguous sign for an ultra - high energy ( uhe ) tau - neutrino .
up to now , there has been no clear identification of tau neutrinos at high energies .
the detection of pev tau neutrinos through optical signals also seems possible .
a combination of fluorescence and cherenkov light detectors in the shadow of steep cliffs could achieve this goal * @xcite*. recently , it has also been shown by the all - sky survey high resolution air - shower detector ( ashra ) experiment , that such kind of experiments could be sensitive to tau neutrinos from fast transient objects such as nearby grbs @xcite.we note , that the recent icecube results do not show any neutrino events at or above the glashow resonance at 6.3 pev @xcite .
this likely means that there is either a cutoff in the astrophysical neutrino flux below @xmath0 pev or the neutrino spectrum is steeper than the usually assumed @xmath1 spectrum . in principle also existing imaging air cherenkov telescopes ( iacts ) such as magic @xcite , veritas @xcite and h.e.s.s .
@xcite have the capability to detect pev tau neutrinos by searching for very inclined showers @xcite . in order to do that , the cherenkov telescopes need to be pointed in the direction of the taus escaping from the earth crust , i.e. at or a few degree below the horizon .
this is because the trajectory of the tau lepton has to be parallel to the pointing direction of the telescope within a few degrees as the cherenkov light is very much beamed in the forward direction .
for example , the magic telescope is placed on top of a mountain on la palma at an altitude of about 2200 m a.s.l .
thus , it can look down to the sea and monitor a large volume within its field of view ( fov ) . in @xcite , the effective area for up - going tau
neutrino observations with the magic telescope was calculated analytically with the maximum sensitivity in the range from 100tev to @xmath2 1eev . however , the calculated sensitivity for diffuse neutrinos was very low because of the limited fov , the short observation time and the low expected neutrino flux .
on the other hand , if flaring or disrupting point sources such as grbs are observed , one can expect an observable number of events even from a single grb if close by . in the case of magic , however , the topographic conditions allow only for a small window of about 1 degree width in zenith and azimuth to point the telescope downhill . in case of other iact sites with different topographic conditions , the acceptance for up - going tau neutrinos
will be increased by the presence of mountains .
mountains can work as an additional target and will lead to an enhancement in the flux of emerging tau leptons .
a target mountain can also shield against cosmic rays and star light . for cherenkov telescope sites , very often nights with high clouds
prevent the observation of gamma - ray sources . in such conditions , pointing the telescopes to the horizon could significantly increase the observation time and the acceptance for up - going tau neutrinos .
next - generation cherenkov telescopes , i.e. the cherenkov telescope array ( cta ) @xcite , can in addition exploit their much larger fov ( in extended observation mode ) , a higher effective area and a lower energy threshold . in this work ,
we present an update of the work in @xcite , where a detailed monte carlo simulation of event rates induced by earth skimming tau neutrinos was performed for an ideal cherenkov detector .
neutrino and lepton propagation was simulated taking into account the local topographic conditions at the magic site and at four possible locations of cherenkov instruments : two in argentina ( san antonio , el leoncito ) , one in namibia ( kuibis ) and one on the canary islands ( tenerife ) . in this work
, similar simulations have been performed for the location of the veritas telescopes and for two sites located close to veritas : meteor crater and yavapai ranch .
these two sites were recently also considered as possible locations for cta .
results are shown for a few representative neutrino fluxes expected from giant agn flares .
we would like to stress that in this work we are exploring the effect of different topographic conditions rather than providing a comprehensive survey of potential sites .
the propagation of a given neutrino flux through the earth and the atmosphere is simulated using an extended version of the anis code @xcite . for fixed neutrino energies ,
@xmath3 events are generated on top of the atmosphere with zenith angles ( @xmath4 ) in the range @xmath5@xmath6 ( up - going showers ) and with azimuth angles in the range @xmath7@xmath8 .
neutrinos are propagated along their trajectories of length @xmath9 from the generation point on top of the atmosphere to the interaction volume , defined as the volume which can contribute to the expected event rate , in steps of @xmath10/1000 ( @xmath11 km ) . at each step of propagation
, the @xmath12nucleon interaction probability is calculated according to parametrization of its cross section based on the chosen parton distribution function ( pdf ) .
in particular , the propagation of tau leptons through the earth is simulated .
all computations are done using digital elevation maps ( dem ) @xcite to model the surrounding mass distribution of each site under consideration .
the flux of the leptons emerging from the ground as well as their energy and the decay vertex positions are calculated inside an interaction volume , modeled by a cylinder with radius of 35 km and height 10 km . the detector acceptance for an initial neutrino energy @xmath13 is given by : @xmath14 where @xmath15 is the number of generated neutrino events .
@xmath16 is the number of @xmath17 leptons with energies @xmath18 larger than the threshold energy @xmath19pev and a decay vertex position inside the interaction volume .
@xmath20 is the probability that a neutrino with energy @xmath13 and zenith angle @xmath4 produces a lepton with energy @xmath18 ( this probability was used as `` weight '' of the event ) .
@xmath21 is the physical cross - section of the interaction volume seen by the neutrino and @xmath22 is the solid angle .
@xmath23 is the trigger efficiency for tau - lepton induced showers with the decay vertex position at ( @xmath24 , @xmath25 ) and height @xmath26 above the ground .
the trigger efficiency depends on the response of a given detector and is usually estimated based on monte - carlo simulations . in this work
, we used an average trigger efficiency extracted from @xcite , namely @xmath27% , which is comparable to what was calculated for up - going tau neutrino showers studied in @xcite .
this is a qualitative estimation and as such it is the major source of uncertainty on the results presented hereafter . a sample of representative neutrino fluxes from photo - hadronic interactions in agns .
see text for more details .
flux-1 and flux-2 are calculations for @xmath28-ray flare of 3c 279 @xcite . flux-3 and flux-4 represent predictions for pks 2155 - 304 @xcite .
flux-5 corresponds to a prediction for 3c 279 calculated in @xcite . ]
equation ( [ aperture ] ) gives the acceptance for diffuse neutrinos .
the acceptance for a point source can be estimated as the ratio between the diffuse acceptance , defined in eq .
( [ aperture ] ) , and the solid angle covered by the diffuse analysis , multiplied by the fraction of time the source is visible @xmath29 .
this fraction depends on the source declination ( @xmath30 ) and the latitude of the observation site ( @xmath31 ) . in this work
, the point source acceptance is calculated as @xmath32 . for energies above 1pev , neutrinos no longer efficiently penetrate the earth and are preferentially observed near the horizon where they traverse a reduced chord of the earth . thus the main contribution to the acceptance comes from a few degrees below horizon . in this work , we simulated propagation of tau neutrinos with zenith angles in the range @xmath33 corresponding to @xmath34sr .
[ cols="^,^,^,^,^,^,^,^",options="header " , ] definitive conclusions on the capability of cherenkov telescopes for detecting tau neutrinos , however , require accurate neutrino - trigger and background simulations which go beyond the scope of this paper .
the acceptance / sensitivity to point sources also depends on the source declination ( @xmath35 ) , which defines the fraction of a day for which a source is visible in the sky at zenith angles between @xmath36 and @xmath37 .
figure [ fig::visibility ] shows this fraction as a function of source declination for the different detector locations . as an example , 3c 279 ( @xmath38 ) and pks 2155 - 305 ( @xmath39 ) are observed for @xmath40% of a sidereal day ( for about 2.4 hours per day ) , while for point sources with declination @xmath41 the observation time is at least two times longer .
the plot also shows that for the different detector locations the visibility has a quite similar behavior . in conclusion , even though the sensitivity window of the earth - skimming analysis in zenith angle is small ( @xmath42 ) , it permits a point source survey in a broad range in declination angles spanning more than @xmath43 in the sky .
fraction of a sidereal day in percent having a point - like source at declination @xmath44 detectable with zenith angle @xmath45 i.e. for the simulated zenith range of earth - skimming ( up - going ) neutrinos .
the fraction of time is plotted as a function of source declination for a few selected sites . ]
in this paper , detailed monte carlo simulations have been employed to evaluate the sensitivity of air cherenkov telescopes to high - energy cosmic tau neutrinos in the pev to eev energy range coming from the horizon .
the simulations use recent predictions for neutrino fluxes in agn flares to calculate expected event rates at the veritas site and two future locations of cherenkov instruments : meteor crater and yavapai ranch , taking into account the local topographic conditions .
the calculated neutrino rates are comparable to what has been estimated for the icecube neutrino telescope assuming realistic observation times for cherenkov telescopes of a few hours . for models with significant neutrino fluxes at energies above @xmath46ev , the sensitivity of cherenkov telescopes actually surpasses that of icecube . for the sites considered , the expected event rates are up to factor three higher compared to what is expected for the la palma site ( magic ) . 201 t.k gaisser and t.stanev , astropart .
39 - 40 ( 2012 ) 120 .
k. koyama et al . ,
nature 378 ( 1995 ) 255 .
a.m. atoyan , c.dermer , new astron .
48 ( 2004 ) 381 .
a.neronov , m.ribordy , phys .
d 80 ( 2009 ) 083008 .
a.mucke , et al .
, astropart .
. 18 ( 6 ) ( 2003 ) 593 .
aartsen et al.,science 342 ( 2013 ) 1242856 . i. cholis and d. hooper jcap 06 ( 2013 ) 030 ; w. winter , phys . rev .
d 88 , ( 2013 ) 083007 ; o. kalashev , et al .
lett . 111 ( 2013 ) 041103 ; e. roulet et al . ; jcap 01 ( 2013 ) 028 ; f.w .
stecker , phys .
d 88 ( 2013 ) 047301 ; h.n ha et al . , phys . rev .
d 87 ( 2013 ) 063011 ; k. murase et al .
d 88 ( 2013 ) 121301 ; l. anchordoqui et al . ,
d 89 ( 2014).083003 ; m.c .
gonzalez - garcia et al . , astropar .
57 ( 2014 ) 39 ; s. razzaque , phys .
d , 88 ( 2013 ) 081302 .
d. fargion , astro - ph/9704205 ; astro - ph/0002453 ; the astrophysical journal 570 ( 2002 ) 909925 .
d. fargion et al . , report no he.6.1.10 ; icrc ( 1999),he 6.1.09 , p.396 - 398.1999 .
kieda , m. salamon , b. dingus(usa ) ; d. fargion et al . , the astrophysical journal 613 ( 2004 ) 1285 - 1301 .
h. athar , g. parente , e. zas , phys .
d62 ( 2000 ) 093010 .
a. letessier - selvon , aip conf .
( 2001 ) 157171 ; x. bertou et al .
, astropart .
17 ( 2002 ) 183 .
feng et al .
88 ( 2002 ) 161102 .
tseng et al .
d 68 ( 2003 ) 063003 . c. aramo et al . , astropart . phys . 23
( 2005 ) 65 - 77 .
j.g . learned et al .
, astropart.phys . 3 ( 1995 ) 267 - 274 .
p.yeh , et al . ,
a 19 ( 13n16 ) ( 2004 ) 1117 ; z.cao , et al . ,
j. of phys .
g : nucl . and part .
31 ( 7 ) ( 2005 ) 571 ; j.liu , et al .
, j. phys .
g36 ( 2009 ) 075201 .
y.asaoka , m.sasaki , astropart .
41 ( 2013 ) 7 .
hess collaboration : http://www.mpi-hd.mpg.de/hfm/hess/pages/about/telescopes d.fargion et .
al . , j. phys .
110:062008 , 2008 ; d. fargion et al .
a588 ( 2008 ) 146 - 150 arxiv:0710.3805 [ astro - ph ]
. m. gaug , c. hsu , j.k .
becker , et al .
, tau neutrino search with the magic telescope , in : international cosmic ray conference , volume 3 of international cosmic ray conference , pp .
1273 - 1276 .
j.k.becker , et al .
instr . and meth .
sect . a : 630 ( 1 ) ( 2011 ) 269 .
a.atoyan , c.d.dermer , phys .
87 ( 2001 ) 221102 .
j.a.aguilar , astro - ph 1301.6504v1 .
j. beacom et al.,phys .
d68 ( 2003 ) 093005 ; h. athar et al.,phys .
d62 ( 2000 ) 103007 .
r. abbasi et .
al , phys.rev .
d86 ( 2012 ) 022005 .
aartsen , phys .
d88 ( 2013 ) 112008 .
|
cherenkov telescopes have the capability of detecting high energy tau neutrinos in the energy range of 11000pev by searching for very inclined showers .
if a tau lepton , produced by a tau neutrino , escapes from the earth or a mountain , it will decay and initiate a shower in the air which can be detected by an air shower fluorescence or cherenkov telescope . in this paper , we present detailed monte carlo simulations of corresponding event rates for the veritas and two proposed cherenkov telescope array sites : meteor crater and yavapai ranch , which use representative agn neutrino flux models and take into account topographic conditions of the detector sites .
the calculated neutrino sensitivities depend on the observation time and the shape of the energy spectrum , but in some cases are comparable or even better than corresponding neutrino sensitivities of the icecube detector . for veritas and the considered cherenkov telescope array sites the expected neutrino sensitivities are up to factor 3 higher than for the magic site because of the presence of surrounding mountains .
|
chromium spinels provide unrivalled possibilities for studying magnetic interactions in solids . in these compounds with a general formula @xmath0cr@xmath1 , where @xmath0 is a divalent nonmagnetic cation ( mg , zn , cd , or hg ) and @xmath2 is a divalent anion ( o , s , or se ) , a cr@xmath7 ion is in the 3@xmath8 configuration .
its three 3@xmath9 electrons occupy the majority - spin states of a completely spin polarised @xmath4 sub - shell leading to the total spin @xmath10=3/2 .
although charge and orbital degrees of freedom in the @xmath0cr@xmath1spinels are frozen , these compounds show wide variety of magnetic properties ranging from those of a strongly frustrated antiferromagnet to a heisenberg ferromagnet . depending on the chemical composition their effective curie - weiss temperature ( @xmath11 ) varies from @xmath12400 k in oxides to 200 k in selenides,@xcite which indicates that the sign of the dominant exchange interaction changes from antiferromagnetic ( afm ) to a ferromagnetic ( fm ) one . in the @xmath0cr@xmath13o@xmath14 spinels afm
nearest - neighbor interactions between cr spins residing on a pyrochlore lattice are geometrically frustrated .
the magnetic ground state of a frustrated antiferromagnet is highly degenerate which leads to unusual low - temperature properties .
@xcite cr oxide spinels remain paramagnetic with the curie - weiss form of the magnetic susceptibility down to temperatures well below @xmath15 of 398 , 71 , and 32 k for @xmath0=zn , cd , and hg .
zncr@xmath13o@xmath14 ( @xmath16=12.5 k ) and cdcr@xmath13o@xmath14 ( @xmath16=7.8 k ) undergo a first order phase transition of the spin - peierls type into a magnetically ordered nel state at temperatures much lower than the characteristic strength @xmath15 of the interaction between cr spins.@xcite the transitions are accompanied by cubic to tetragonal structural distortions , however , the sign of the distortions and the magnetic order below @xmath16 are different . in zncr@xmath13o@xmath14 the lattice contracts along the @xmath17 axis ( @xmath18 ) and the nel state has a complex commensurate spin structure with 4 characteristic wave vectors . in contrast , the lattice of cdcr@xmath13o@xmath14 expands below @xmath16 ( @xmath19 ) and its ordered state is incommensurate with a wave vector @xmath20=(0,@xmath21,1 ) with @xmath220.09.@xcite hgcr@xmath13o@xmath14 also undergoes a transition to a magnetically ordered state at @xmath16=5.8 k but the symmetry of the lattice lowers to orthorhombic .
@xcite recently , a metamagnetic transition and a wide magnetisation plateau with the magnetic moment equal to one - half of the full cr moment have been observed in cdcr@xmath13o@xmath14 and hgcr@xmath13o@xmath14 .
@xcite in @xmath0cr@xmath13s(e)@xmath14 spinels dominant ferromagnetic interactions are not geometrically frustrated .
nevertheless , cr spins in zncr@xmath13s@xmath14 and zncr@xmath13se@xmath14 form helical spin structures below 15.5 and 18 k , respectively .
@xcite in zncr@xmath13s@xmath14 the helical structure coexists at low temperatures with a collinear afm one .
the transitions into helically ordered state are supposed to occur because of competing fm nearest - neighbor interaction and afm interactions between more distant cr neighbors .
recently , it has been shown that cdcr@xmath13s@xmath14 and hgcr@xmath13s@xmath14 exhibit ferroelectric behavior with strong increase of the dielectric constant below the temperature of magnetic ordering .
@xcite the diversity of magnetic properties of the @xmath0cr@xmath1 spinels can hardly be explained without understanding the mechanism of exchange interactions between cr spins , their range , and relative strengths .
so far theoretical analyses of the effective exchange interactions in cr spinels were mostly based on the goodenough - kanamori rules of superexchange .
@xcite j. goodenough in ref . explained the fm sign of the nearest - neighbor coupling @xmath3 in @xmath0cr@xmath13s(e)@xmath14 by indirect hopping between half - filled cr @xmath4and empty @xmath5 states via @xmath6 states of @xmath2 anions .
k. dwight and n. menyuk ( refs . ) analyzed various superexchange paths for interactions between up to 6-th cr neighbors and concluded that afm coupling constants @xmath23 between 3-rd neighbors and even weaker fm @xmath24 and @xmath25 may be relevant alongside @xmath3 .
then , the estimated @xmath26 were used to examine the stability of different spiral ground states in zncr@xmath13se@xmath14 .
to our knowledge the only attempt to obtain the values of @xmath26 from _ ab initio _ band structure calculations was made in ref . where the coupling constants @xmath3@xmath23 were calculated by comparing the total energies of several simple spin configurations .
the aim of the present work is to compare electronic band structures of @xmath0cr@xmath1 ( @xmath0=zn , cd , hg and @xmath2=o , s , se ) spinels calculated within the local spin density ( lsda ) as well as lsda+@xmath27 approximations and to estimate exchange coupling constants between cr spins by fitting the calculated energy of spin spirals to a classical heisenberg model .
the paper is organized as follows .
the spinel crystal structure is shortly described in sec .
[ sec : structure ] .
some details of the calculational procedure are given in sec .
[ sec : details ] . in sec .
[ sec : bandstr ] the band structures of the @xmath0cr@xmath1 spinels calculated using lsda and lsda+@xmath27 are compared and their dependence on the chemical composition is analyzed .
the results on the exchange coupling constants @xmath26 are presented in sec.[sec : exch ] .
the comparison of calculated @xmath26 to experimental data and the discussion of their origins are given in sec.[sec : discus ] .
finally , the results are summarised in sec.[sec : summ ] .
@xmath0cr@xmath1 compounds considered here belong to a large family of @xmath28 spinels which crystallise to a cubic @xmath29 ( n227 ) structure ; with @xmath0 , @xmath30 , and @xmath2 ions occupying @xmath31 ( 1/8,1/8,1/8 ) , @xmath32 ( 1/2,1/2,1/2 ) , and @xmath33 ( @xmath34,@xmath34,@xmath34 ) wyckoff positions , respectively .
the experimental values of the lattice constant ( @xmath35 ) and fractional coordinates ( @xmath34 ) are collected in table [ tab : str ] .
the spinel crystal structure plotted in fig .
[ fig : acrx_str ] can be considered as built of distorted cr@xmath36 cubes which share a cr site .
cr and @xmath2 ions that belong to the same cube form two regular tetrahedra with a common center , which coincides with the center of the cube .
each cube is linked via @xmath2 ions to four regular @xmath37 tetrahedra .
the centers of the cr@xmath36 cubes and @xmath37 tetrahedra form two diamond lattices shifted by a vector ( 1/4,1/4,1/4 ) .
finally , cr ions are arranged along chains running in @xmath38 directions and form the so - called pyrochlore lattice which consists of corner - sharing regular tetrahedra .
( color online ) the crystal structure of @xmath0cr@xmath1spinels with an @xmath0 ion shifted to the origin .
distorted cr@xmath36 cubes and @xmath37 tetrahedra are plotted by thick ( red ) and thin ( green ) lines , respectively.,scaledwidth=44.0% ] as the ionic radius of a @xmath39 cation increases in the row zn@xmath40cd@xmath40hg@xmath41 , @xmath2 ions are pushed further away from the center of a @xmath37 tetrahedron .
this leads to lattice expansion and corresponding increase of cr cr distances .
the distortion of cr@xmath36 cubes also increases , which allows to avoid too strong elongation of cr@xmath2 bonds and manifests itself in the increase of the @xmath34 parameter . when o@xmath42 ions are replaced by s@xmath42 or se@xmath42 , which have significantly larger ionic radii , the length of both @xmath0@xmath2 and cr@xmath2 bonds increases .
this results in even stronger increase of the lattice constant .
the nearest cr cr ( @xmath43 ) , cr@xmath2 ( @xmath44 ) , and @xmath0@xmath2 ( @xmath45 ) distances are summarized in table [ tab : str ] .
cd .. 2.3d .. 1.4d ..
1.3d .. 1.3d .. 1.3r compound & & & & & & @xmath11 + zncr@xmath13o@xmath14 & 8.327 & 0.2616 & 2.944 & 1.990 & 1.970 & -398 + cdcr@xmath13o@xmath14 & 8.600 & 0.2682 & 3.041 & 2.006 & 2.133 & -71 + hgcr@xmath13o@xmath14 & 8.661 & 0.2706 & 3.062 & 2.003 & 2.184 & -32 + zncr@xmath13s@xmath14 & 9.982 & 0.2619 & 3.529 & 2.383 & 2.367 & 7.9 + cdcr@xmath13s@xmath14&10.240 & 0.2647 & 3.620 & 2.419 & 2.478 & 90 + hgcr@xmath13s@xmath14&10.256 & 0.267 & 3.626 & 2.402 & 2.523 & 140 + zncr@xmath13se@xmath14&10.484 & 0.2599 & 3.707 & 2.522 & 2.450 & 155 + cdcr@xmath13se@xmath14&10.735 & 0.2642 & 3.795 & 2.540 & 2.588 & 184 + hgcr@xmath13se@xmath14&10.737 & 0.264 & 3.796 & 2.543 & 2.585 & 200 + each cr site is surrounded by a trigonally distorted @xmath46 octahedron ; with all cr@xmath2 distances in the octahedron being equal .
the degree of the trigonal distortion is determined by the value of the fractional coordinate @xmath34 . for @xmath470.25 the octahedron
is expanded along one of the @xmath48 directions and becomes regular for @xmath34=1/4 .
the local symmetry of a cr site is @xmath49 .
it is worth noting that since six cr ions nearest to a cr site form a trigonal antiprism , the symmetry remains trigonal even when @xmath34 is equal to the ideal value of 1/4 .
the calculations of the electronic band structure of the @xmath0cr@xmath1 spinels were performed for the experimentally observed lattice constant and @xmath2 fractional coordinates ( table [ tab : str ] ) using the linear muffin - tin orbital ( lmto ) method @xcite with the combined correction terms taken into account . in order to decrease overlap between atomic spheres 3 sets of empty spheres ( @xmath50 )
were added at @xmath51 ( 0,0,0 ) , @xmath52 ( 3/8,3/8,3/8 ) , and @xmath53 ( @xmath54,1/8,1/8 ) wyckoff positions .
muffin - tin orbitals with angular momentum @xmath55 for @xmath0 , cr , and @xmath2 spheres and @xmath56 for empty spheres were included into the basis set .
the perdew wang parameterisation @xcite for the exchange - correlation potential in the local spin - density approximation was used .
brillouin zone ( bz ) integrations were performed using the improved tetrahedron method .
@xcite calculations for spiral spin structures based on the generalized bloch theorem @xcite were performed under the assumption that the direction of the magnetisation @xmath57 is constant within each atomic sphere . inside a sphere at @xmath58 , where @xmath59 defines its position in a unit cell and
@xmath60 is a lattice vector , the magnetisation direction is determined by an angle @xmath61 between the spin moment and @xmath62 axis , the wave vector @xmath63 of a spiral , and a phase @xmath64 : @xmath65 with @xmath66 . when performing lmto calculations for spin spirals it is convenient to split the lsda exchange - correlation potential into a spin- and @xmath63-independent part @xmath67 and an effective exchange field @xmath68 , where @xmath69 and @xmath70 are exchange - correlation potentials for majority- and minority - spin electrons defined in the site - dependent local spin frame , in which the spin - density matrix for a given site is diagonal . in the present calculations _
spin - independent _ lmto basis functions were constructed starting from the solution of the kohn - sham equation with only the spin - independent part @xmath71 of the exchange - correlation potential included to the lsda one electron potential .
matrix elements of the spin - dependent part of the lmto hamiltonian @xmath72 where @xmath73 is a vector of pauli matrices , were included at a variational step .
radial matrix elements of @xmath74 between the solution of the kohn - sham equation inside a sphere @xmath75 and/or its energy derivative @xmath76 were calculated by numerical integration .
the diagonal in spin indices matrix elements of @xmath77 calculated between two bloch wave functions with wave vectors @xmath78 and @xmath79 do not depend on @xmath63 and , as usual , are non - zero only if @xmath80 . in the absence of the spin - orbit coupling the only off - diagonal in spin indices terms of the hamiltonian are @xmath81 and @xmath82 given by eq .
( [ eq : hb ] ) , which couple the states with @xmath83 .
then , the lmto hamiltonian matrix can be written in the following block form : @xmath84 thus , for a spin spiral with an arbitrary @xmath63 one needs to diagonalise one @xmath85 matrix instead of two @xmath86 matrices , one for each spin channel , in conventional spin - polarised calculations .
@xcite if , however , the spin - orbit coupling term is included into the hamiltonian it additionally couples majority- and minority - spin states with the same @xmath87 and the hamiltonian matrix becomes infinite .
most calculations reported in sec . [ sec : exch ] were performed for planar spin spirals with all @xmath88 .
then , for a given @xmath63 the magnetisation direction inside a sphere is defined solely by the phase @xmath64 .
the phases at cr sites were fixed by requiring that @xmath89 . at other atomic and empty sites they were determined self - consistently . at each iteration of a self - consistency loop a rotation to the local spin frame in which the spin - density matrix becomes diagonal was found and the corresponding rotation angles were used to determine new magnetisation direction for the next iteration .
@xcite iterations were repeated until self - consistency in the electron spin density as well as in the magnetisation direction in each atomic sphere was achieved . finally , in order to account for correlation effects in the cr 3@xmath9 shell we adopted the lsda+@xmath27 method @xcite in the rotationally invariant representation .
@xcite the so - called atomic limit @xcite was used for the double counting term . the effective screened coulomb repulsion @xmath27
between 3@xmath9 electrons was considered as a parameter of the model and varied from 2 to 4 ev .
for the on - site exchange integral @xmath90 the value of 0.9 ev estimated from the lsda calculations was used .
the effect of the chemical composition on the electronic structure of @xmath0cr@xmath1spinels can be analyzed by comparing the densities of cr @xmath9 and @xmath2 @xmath6 states obtained from spin - restricted lda calculations for @xmath0cr@xmath1 ( fig .
[ fig : acrx_dos ] ) .
since site - resolved densities of states ( dos ) for the @xmath0cr@xmath13s@xmath14 spinels show the same trends as those for @xmath0cr@xmath13se@xmath14 , only dos calculated for @xmath0cr@xmath13o@xmath14 and @xmath0cr@xmath13se@xmath14 compounds are presented in fig .
[ fig : acrx_dos ] .
( color online ) densities of cr @xmath9 ( red solid lines ) , @xmath2 @xmath6 ( blue dashed lines ) , and @xmath0 @xmath9 ( green dotted lines ) states in @xmath0cr@xmath13o@xmath14 and @xmath0cr@xmath13se@xmath14 .
zero energy is chosen at the fermi level.,scaledwidth=45.0% ] in all the @xmath0cr@xmath1 spinels considered in the present work occupied bands in the energy range down to @xmath1210 ev below the fermi level ( @xmath91 ) originate from @xmath0 @xmath9 , cr @xmath9 , and @xmath2 @xmath6 states .
@xmath9 states in the sulfides and selenides as well as cd 4@xmath9 states in cdcr@xmath13o@xmath14 form a narrow group of bands below the bottom of @xmath2 @xmath6 states .
zn @xmath92 and hg @xmath93 bands in the corresponding oxides cross the bottom of o @xmath6 states and hybridise strongly with the latter which results in the appearance of wide peaks of @xmath9 dos between @xmath94 and @xmath95 ev .
the @xmath0 @xmath9 states are completely occupied and lie well below @xmath91 so that they have no effect on the magnetic properties of the @xmath0cr@xmath1 spinels .
o @xmath6-derived bands in @xmath0cr@xmath13o@xmath14 spread over the energy range from @xmath128 to @xmath96 ev and are separated by a gap of @xmath97 ev from cr 3@xmath9 states which give prevailing contribution to the bands crossing the fermi level and a prominent dos peak at @xmath91 . because of weaker electronegativity of s@xmath42 and se@xmath42 ions as compared to o@xmath42 , @xmath2 @xmath6 states in @xmath0cr@xmath13s(e)@xmath14spinels move closer to @xmath91 and form bands between @xmath98 and @xmath99 ev . in contrast to oxides the top of @xmath2 @xmath6 bands overlaps with the bottom of cr @xmath9 ones and the gap between @xmath2 @xmath6 and cr @xmath9 states closes .
cr @xmath9 states are split by the cubic component of the crystal field at a cr site into a triplet @xmath4 ( @xmath100 , @xmath101 , and @xmath102 ) and a doublet @xmath5 ( @xmath103and @xmath104 ) . in a cr@xmath46 octahedron
the @xmath4 and @xmath5 states form relatively weak @xmath105- and much stronger @xmath106-type bonds with the @xmath2 @xmath6 states , respectively .
bonding cr @xmath9
@xmath2 @xmath6 combinations participate in the formation of the @xmath2 @xmath6-derived bands which is evidenced by the rather high density of cr @xmath9 states in this energy range ( see fig .
[ fig : acrx_dos ] ) .
these states are completely filled and stabilise the cr@xmath46 octahedron . in all the @xmath0cr@xmath1 spinels considered here ,
the partially occupied bands crossing @xmath91 are formed by antibonding combinations of the cr @xmath9 @xmath4 and @xmath2 @xmath6 states with the dominant contribution of the former .
these states , filled with 3 electrons , play the crucial role in the formation of cr magnetic moments and effective exchange interactions between them .
the cr @xmath9 states of the @xmath5 symmetry which form antibonding @xmath106 combinations with the @xmath2 @xmath6 states are shifted to higher energies and separated by an energy gap from the antibonding cr @xmath4 @xmath2 @xmath6 states . for the sake of brevity in the following these bands
are referred to simply as cr @xmath9 @xmath4 and @xmath5 bands .
( color online ) symmetry - resolved densities of the cr @xmath9 @xmath107 ( red solid lines ) , @xmath108 ( blue dashed lines ) , and @xmath109(green dotted lines ) states in @xmath0cr@xmath13o@xmath14 and @xmath0cr@xmath13se@xmath14 .
zero energy is chosen at the fermi level.,scaledwidth=44.0% ] as the local symmetry of a cr site is lower than cubic the @xmath4 states are additionally split into a singlet and a doublet which transform according to @xmath107 and @xmath5 representations of the @xmath49 group .
we denote this doublet as @xmath108 in order to distinguish it from the doublet formed by @xmath103 and @xmath104 orbitals ( @xmath109 ) which transforms according to the same @xmath5 representation .
the densities of the cr @xmath9 states of @xmath107 , @xmath108 , and @xmath109 symmetries are shown in fig .
[ fig : cr_d_dos ] .
the trigonal splitting between the @xmath107 and @xmath108 states is much smaller than their bandwidths ; with the center of gravity of the former being about 0.05 ev lower . the corresponding dos curves in @xmath0cr@xmath13o@xmath14 are , however , remarkably different .
the total width of the cr @xmath4 sub - band is determined by the @xmath108 states .
their dos curve is very asymmetric with low dos at the bottom and a huge dos peak at the top of the @xmath4 bands .
the @xmath107 states are significantly narrower .
they are responsible for the dos peak at @xmath91but contribute also to the high - energy @xmath108 dos peak .
this strong suppression of the width of the @xmath107 dos is a characteristic feature of @xmath92 transition metal oxides with the spinel structure and a small ionic radius of an @xmath0 ion .
recently , two scenarios based on a strong coulomb interaction within v @xmath9 shell and the localized behavior of v @xmath107 states have been proposed to explain heavy - fermion - like properties in liv@xmath13o@xmath14 .
@xcite a simplified analysis using slater
koster integrals @xcite shows that the narrowing of @xmath107 dos is caused by a delicate balance between direct @xmath9@xmath9 hopping matrix elements @xmath110 between the cr @xmath4 states and an effective indirect hopping @xmath111 via o @xmath6 states . under the assumption that @xmath112 , @xmath113 , and @xmath114 the strongest hopping between @xmath4 states of nearest cr ions is a negative @xmath115-type hopping between those orbitals which have their lobes directed along one of cr chains , e.g. a @xmath100@xmath100 hopping along the [ 110 ] direction .
hoppings of @xmath116 type between the other two @xmath4orbitals ( @xmath102 and @xmath101 ) along the same chain are positive and about two times weaker , whereas @xmath100@xmath117 hoppings are zero . after making transformation to the symmetrised combinations of @xmath4 orbitals it turns out that the matrix elements between @xmath107 and @xmath108 orbitals on neighboring cr sites are much smaller than the matrix elements between the states of the same symmetry , with the @xmath107@xmath107 hopping being stronger than the @xmath108@xmath108 one .
this results in almost decoupled @xmath107 and @xmath108 bands , with the bandwidth of the former being _ larger _ than of the latter .
the contribution of the hopping via o @xmath6 states to the lowest order in @xmath118 , where @xmath119 is the cr @xmath9o @xmath6 matrix element and @xmath120 and @xmath121 are the energies of the cr @xmath9 and o @xmath6 states , can be easily estimated using the lwdin partitioning technique .
@xcite in the ideal ( @xmath122 ) spinel structure with regular cr@xmath14o@xmath14 cubes the indirect hopping contributes only to @xmath102@xmath101 and @xmath101@xmath102hybridisations via o @xmath123 states .
the corresponding matrix element @xmath124 is negative , so that the direct and indirect terms compensate each other .
however , as it was shown for liv@xmath13o@xmath14 in ref . and verified for zncr@xmath13o@xmath14 in the present work , the densities of @xmath107 and @xmath108 states calculated for a hypothetical spinel structure with @xmath122 are of the same width .
thus , the reduction of the @xmath102@xmath101 hopping is not sufficient to explain the narrowing of the @xmath107states .
when the distortion of cr@xmath36 cubes is taken into account , non - zero indirect hoppings proportional to the degree of the distortion @xmath125 appear between all pairs of @xmath4 orbitals .
however , the most significant one is an indirect @xmath100@xmath100 matrix element between the orbitals pointing along cr chains .
already for @xmath34=0.26 it becomes as large as 0.6@xmath124 and , being of the opposite sign , strongly suppresses the direct @xmath100@xmath100 hopping .
the net effect of the competition between the direct and indirect contributions to the effective @xmath9@xmath9 hopping is the reduction of the @xmath107 bandwidth relative to the width of @xmath108states .
if , however , the direct hopping becomes weak , the width of the @xmath107and @xmath108 states is governed by the indirect hopping via @xmath2 @xmath6 states and depends strongly on the deviation of the fractional coordinate @xmath34 from the ideal value of 1/4 . finishing the discussion of the @xmath9@xmath9 hybridisation we have to mention strong hopping matrix elements between @xmath4 and @xmath109 states which appear as a result of either direct hopping between , for instance , @xmath100 and @xmath103orbitals in the [ 110 ] direction or indirect one via @xmath126 and @xmath127 states of two @xmath2 ions closest to the cr
the indirect hopping is allowed already in an undistorted cr@xmath36 cube and is stronger than @xmath124 for one of the involved @xmath6@xmath9 bonds is of @xmath106 type . however , the direct and indirect contributions are of opposite signs which may lead to strong suppression of effective @xmath4@xmath109 hopping matrix elements . finally , an indirect contribution to hopping between @xmath109 states of nearest cr ions is proportional to @xmath21 and is of the same sign as the corresponding direct hopping .
let us come back to the results of band structure calculations shown in figs .
[ fig : acrx_dos ] and [ fig : cr_d_dos ] . as zn is replaced by a larger cd or hg ion ,
the width of the cr @xmath4 states decreases from 1.8 ev in zncr@xmath13o@xmath14 to 1.6 ev in hgcr@xmath13o@xmath14 .
the narrowing of the bands can be explained by the reduction of the strength of both @xmath9@xmath9 and cr @xmath9o @xmath6 hopping matrix elements caused by the lattice expansion ( see table [ tab : str ] ) .
the shape of @xmath107 and @xmath108 dos , however , also varies considerably .
this indicates the change of relative strengths of the direct @xmath9@xmath9 and indirect @xmath9o @xmath6@xmath9 hoppings between cr @xmath4 states .
cr @xmath109 states are affected much stronger by the change of the chemical composition because of their interaction with @xmath128 bands of an @xmath0 ion . in zncr@xmath13o@xmath14
@xmath109 bands cross just the bottom of a zn 4@xmath128 band near the bz center .
cd 5@xmath128 and , especially , hg 6@xmath128 bands shift to lower energies and overlap with cr @xmath109 bands . in cdcr@xmath13o@xmath14 the hybridisation with a cd 5@xmath128 band is responsible for a low energy tail of cr @xmath109 dos . in hgcr@xmath13o@xmath14
the bottom of a hg 6@xmath128 band comes so close to cr @xmath4 bands that it starts to hybridise via o @xmath6 states with cr @xmath107 bands , which is evidenced by the appearance of noticeable density of @xmath107 and @xmath109 states in the energy gap between @xmath4 and @xmath109-derived bands .
symmetry resolved densities of cr @xmath9 states in selenides , presented on the right hand side of fig .
[ fig : cr_d_dos ] , are strikingly different from the corresponding dos curves in oxides .
the densities of @xmath107 and @xmath108states do not differ so much as in oxides ; with the top of the @xmath108 bands being shifted to somewhat higher energies .
the huge dos peak found at the very top of the @xmath4 states in zncr@xmath13o@xmath14 and slightly lower in the two other oxides shifts in selenides closer to the center of the @xmath4 bands .
the energy difference between the cr @xmath109 and @xmath4 states is somewhat smaller than in oxides .
the @xmath109 dos curves become wider and more symmetric . another important distinction of the @xmath0cr@xmath13s(e)@xmath14 electronic structure
is the enhanced weight of the @xmath109 states in the bloch wave functions of cr @xmath4bands , which is revealed by noticeable density of the @xmath109 states in the corresponding energy range .
( color online ) `` fat '' bands calculated along some symmetry lines in the fcc bz for zncr@xmath13o@xmath14 ( lower panel ) and zncr@xmath13se@xmath14 ( upper panel ) .
the size of filled ( red ) circles and ( blue ) squares is proportional to the partial weight of cr @xmath107 and @xmath109 states in the bloch wave function , respectively .
zero energy is chosen at the fermi level.,scaledwidth=44.0% ] the difference between the dispersions of cr @xmath9 bands in the oxides and selenides is illustrated in fig .
[ fig : fb_zn ] which shows `` fat '' bands calculated for zncr@xmath13o@xmath14 and zncr@xmath13se@xmath14 ; with the size of filled circles and squares being proportional to the partial weight of cr @xmath107 and @xmath109 states , respectively , in the bloch wave function at a given @xmath78-point .
the abovementioned differences in the band dispersions and dos suggest that the effective @xmath9@xmath9 hopping matrix elements should also be significantly different , which is not surprising taking into account the increase of the lattice constants in @xmath0cr@xmath13s(e)@xmath14 as compared to @xmath0cr@xmath13o@xmath14 ( table [ tab : str ] ) .
the increase of cr cr distances leads to strong reduction of the direct @xmath9@xmath9 hopping matrix elements which fall off as @xmath129 .
it should be noted , however , that a tight binding analysis of the band structure of cdcr@xmath13s@xmath14 performed in ref . showed that the direct hopping between cr @xmath9 states is not negligible even in sulfides and selenides . the contribution of the indirect hopping via @xmath2 @xmath6 states to the effective @xmath9@xmath9 hoppings , which is proportional to @xmath130 , does not change much because i ) s 3@xmath6 and se 4@xmath6 states
are significantly more spatially extended than o 2@xmath6 states and , although @xmath44 also increases in @xmath0cr@xmath13s(e)@xmath14 , cr @xmath9@xmath2 @xmath6 hoppings @xmath119 remain approximately of the same strength as in @xmath0cr@xmath13o@xmath14 ; ii ) the energy difference @xmath131 in sulfides and selenides is smaller than in oxides . similar to oxides
the cr @xmath4 states in selenides become narrower with the increase of the @xmath0 ionic radius .
the hybridisation of cr @xmath109 states with hg 6@xmath128 bands leads to sharpening of a dos peak near the bottom of the @xmath109 sub - band . when in spin - polarized lsda calculations cr @xmath9 states with two spin projections are allowed to have different occupations , the strong on - site exchange interaction splits the half - filled cr @xmath4 shell into occupied majority - spin @xmath132 and unoccupied minority - spin @xmath133 .
the densities of cr @xmath9 and @xmath2 @xmath6 states in @xmath0cr@xmath13o@xmath14 and @xmath0cr@xmath13se@xmath14 obtained from spin - polarized calculations with the ferromagnetic ( fm ) alignment of cr moments are shown in fig . [ fig : acrx_fm ] .
( color online ) cr @xmath9 ( red solid lines ) and @xmath2 @xmath6 ( blue dashed lines ) dos in @xmath0cr@xmath13o@xmath14 and @xmath0cr@xmath13se@xmath14 obtained from spin - polarized calculations with fm alignment of cr magnetic moments .
zero energy is at the fermi level.,scaledwidth=45.0% ] in all the spinels studied here the exchange splitting of cr @xmath9 states is about 2.7 ev which gives an estimate of 0.9 ev for the on - site hund s exchange coupling @xmath90 .
the exchange and crystal field splittings are of comparable strengths , so that cr @xmath133 states are found at the same energy as @xmath134 . in the oxides @xmath132 states remain separated by an energy gap from o @xmath6 bands . in zncr@xmath13o@xmath14 and cdcr@xmath13o@xmath14 the fm solution is insulating with completely occupied @xmath132 and empty @xmath133 states . in hgcr@xmath13o@xmath14
the bottom of a minority - spin hg @xmath128 band hybridised with cr @xmath135 states crosses @xmath132bands leading to a metallic solution .
in contrast to the oxides , in the @xmath0cr@xmath13s(e)@xmath14 spinels @xmath132 states move below the top of the @xmath2 @xmath6 states . as a result
the highest occupied majority - spin bands are formed mainly by @xmath2 @xmath6 states strongly hybridised with cr @xmath136 . in the zn and hg spinels they overlap with the very bottom of mostly unoccupied @xmath133 bands while for cdcr@xmath13s@xmath14 and cdcr@xmath13se@xmath14 insulating solutions with a tiny gap
were obtained .
the lsda band structures and dos calculated for cdcr@xmath13s@xmath14 and cdcr@xmath13se@xmath14 are in good agreement with the results of previous calculations .
@xcite in the present work , however , the fm solution for cdcr@xmath13s@xmath14 is insulating whereas in ref .
metallic fm solutions were obtained for both compounds .
a possible cause for the difference is that in ref . structural data with the s fractional coordinate @xmath34=1/4 were used .
a cr spin magnetic moment defined as a volume integral of the spin density over a cr atomic sphere varies from 2.86 @xmath137 in zncr@xmath13o@xmath14 to 3.15 @xmath137 in cdcr@xmath13se@xmath14 , which agrees with a good accuracy with the value of 3 @xmath137 expected for the completely polarized @xmath4 shell filled with 3 electrons .
the hybridisation with cr @xmath9 states induces weak negative spin - polarisation of @xmath2 @xmath6 states .
it is , however , partially compensated by a positive contribution coming from the `` tails '' of cr @xmath9 states that have the @xmath9 symmetry inside an @xmath2 atomic sphere .
the net spin magnetic moment @xmath138 induced on @xmath2 ions is small and negative with @xmath139 @xmath137 , @xmath140 @xmath137 , and @xmath141 @xmath137 .
the moment induced on @xmath0 ions is less than 0.05 @xmath137 .
the self - consistent fm solution discussed above can be obtained for any of the 9 spinels but it does not necessarily have the lowest total energy . if the coupling between cr spins is antiferromagnetic ( afm ) then , because of geometrical frustrations intrinsic to the pyrochlore lattice formed by cr ions , any configuration of classical spins such that the sum of 4 spin vectors sitting in the corners of each tetrahedron is equal to zero is the ground state of the heisenberg hamiltonian with nearest neighbor interactions .
@xcite the requirement of zero total spin in each tetrahedron is satisfied , for example , if cr spins are aligned ferromagnetically along the [ 1@xmath14210 ] chains and antiferromagnetically along [ 01@xmath1421 ] and [ 10@xmath1421 ] ones or , in other words , if cr spins in each ( 001 ) plane are parallel but the spins in consecutive planes along the [ 001 ] direction are antiparallel to each other . in the following this spin configuration
will be denoted as afz .
it is easy to check that in this case only 4 of 6 cr
cr bonds in each tetrahedron are afm while the other two remain fm and the magnetic energy can not be minimised for all 6 bonds simultaneously if the coupling between cr spins is antiferromagnetic . another spin configuration which gives zero total spin in each tetrahedron and does not break the cubic symmetry of the lattice is a non - collinear one with cr spins directed along the lines passing through the center of a tetrahedron , i.e. , along one of the @xmath48 directions ; with all four cr spins pointing either to or away from the center .
self - consistent band structure calculations performed with this configuration of cr spins gave for all the @xmath0cr@xmath1 spinels insulating solutions with zero net magnetic moment , which in the following is referred to as a zm solution .
cr magnetic moments calculated in the local spin frame are about 0.1 @xmath137 smaller than for the corresponding fm solution but still close to 3 @xmath137 . depending on the magnetisation directions of four cr ions in a cr@xmath36 cube the magnetisations in four @xmath2 atomic spheres are parallel or antiparallel to the vectors pointing to the center of the cube so that small @xmath2 magnetic moments also cancel each other . with this arrangement of cr and @xmath2 moments
the spin moments of @xmath0 ions are equal to zero .
it is worth noting that for all the @xmath0cr@xmath1 compounds the lsda total energy difference between the afz and zm solutions does not exceed 3 mev per formula unit .
the comparison of the lsda total energies of fm and zm solutions shows that in zncr@xmath13o@xmath14 the solution with zero net moment is significantly more favourable ; with the energy difference per formula unit ( f.u . ) being 171 mev . in cdcr@xmath13o@xmath14
this difference decreases to 19 mev , whereas in hgcr@xmath13o@xmath14 the two solutions are almost degenerate ; the fm one being 2 mev lower . in the sulfides and selenides the fm solution becomes more preferable . as zn is replaced by a heavier ion the energy difference between the two solution increases from 43 to 118 mev / f.u . in sulfides and from 80 to 134 mev
/ f.u.in selenides .
these results indicate that the change of the sign of dominant exchange interactions from afm in oxides to fm in sulfides and selenides is captured already by lsda band structure calculations .
moreover , all the compounds show a clear tendency to ferromagnetism as the lattice expands due to the increase of the radius of @xmath0 ions . in @xmath0cr@xmath1 , as well as in many other 3@xmath9 compounds
, lsda underestimates correlation effects in the rather localised cr 3@xmath9 shell .
the strong electronic correlations can be accounted for , at least on a mean - field level , by using the lsda+@xmath27 approach . since cr @xmath4
states are split by the on - site exchange interaction into occupied majority- and unoccupied minority - spin states , charge and orbital degrees of freedom are frozen already in lsda . as a consequence , lsda+@xmath27 band structures of the @xmath0cr@xmath1spinels are qualitatively similar to the lsda results .
when non - spherical terms of the coulomb repulsion are neglected , the orbital dependent lsda+@xmath27 potential @xmath143 can be approximated as @xmath144 , where @xmath145 is the occupation of @xmath146-th localised orbital , @xmath147 , and @xmath27 is the effective screened coulomb repulsion between 3@xmath9 electrons .
taking into account that the orbital occupations @xmath145 of cr @xmath132 and @xmath133 states are close to 1 and 0 , respectively , the main effect of the lsda+@xmath27 potential is to shift the @xmath132 states to lower and @xmath133 to higher energies by @xmath148 .
thus , the splitting between the @xmath133 and @xmath132 states increases from 3@xmath90 in lsda calculations to @xmath149 in lsda+@xmath27 ones . in oxides , because of the downward shift of the @xmath132 states , they start to overlap with o @xmath6 bands . in @xmath0cr@xmath13s(e)@xmath14
the @xmath132states move further below the top of @xmath2 @xmath6 bands and their contribution to the highest occupied majority spin bands decreases . due to the increase of the @xmath133@xmath132 splitting , ferromagnetic lsda+@xmath27 solutions for all 9 compounds considered in the present work become insulating starting from @xmath1503 ev .
because of the strong cr @xmath109@xmath2 @xmath6 hybridisation discussed in sec.[sec : lda ] the orbital occupations @xmath145 calculated for the formally unoccupied cr @xmath109 states are in the range 0.280.48 for the majority- and about 0.220.25 for the minority - spin orbitals . as a result the shift of the cr @xmath109 states to higher energies caused by the coulomb repulsion @xmath27 is somewhat smaller than for the @xmath133 states for which @xmath145 does not exceed 0.07 and the splitting between @xmath151 and @xmath133 states decreases with the increase of @xmath27 .
effective exchange coupling constants @xmath26 between cr spins were determined by performing band structure calculations for spiral spin structures with a varying wave vector @xmath63 .
then , the @xmath63-dependence of the energy of the spin spirals was mapped onto a classical heisenberg model . in order to get reliable estimates for @xmath26
one needs to perform calculations for a sufficiently large number of @xmath63 .
since self - consistent calculations for spin spirals are much more time - consuming than conventional collinear spin - polarised calculations , the @xmath63-dependence of their energy was calculated using the so - called local force theorem ( lft ) .
@xcite according to this approach the total energy difference between two spin configurations can be approximated by the difference of their band energies , provided that the calculations are performed starting from the same electron spin densities and only the magnetisation direction varies . strictly speaking this approximation
is justified only for small deviations of the magnetisation direction from some collinear spin arrangement , i.e. , for small @xmath152 , and before using it for short wavelengths we have numerically checked its accuracy for the case of the @xmath0cr@xmath1 compounds . first , self - consistent calculations were performed for two sets of planar spin spirals with the magnetisation direction inside a cr sphere fixed by polar angles @xmath153 and @xmath154 , where @xmath59 is the position of a cr site in the unit cell and @xmath60 is a lattice vector .
the wave vectors @xmath63=(0,0,@xmath155 ) and @xmath63=(@xmath155,@xmath155,0 ) of the spirals varied along the @xmath156@xmath2 and @xmath156@xmath157 high symmetry directions , respectively , in the range @xmath158 in @xmath159 units . obviously , the ( 0,0,0 ) spiral is the collinear fm structure with the magnetisation directed along the @xmath34 axis , whereas ( 2,2,0 ) and ( 0,0,2 ) spirals are equivalent to the collinear afz spin structure with zero net magnetisation discussed in sec.[sec : lsda ] .
the magnetisation directions in other atomic spheres were determined self - consistently .
( color online ) the dependence of the total @xmath160 and band @xmath161 energies calculated within the lsda on the wave vector of ( 0,0,@xmath155 ) and ( @xmath155,@xmath155,0 ) spin spirals .
the energies are plotted relative to the energy of the fm states with @xmath155=0.,scaledwidth=44.0% ] then , band energies for the same spin spirals were calculated starting from self - consistent fm electron densities for @xmath0cr@xmath13s(e)@xmath14 and from the densities for the zm solution in the case of oxides .
when using lft for spinels one needs to specify the magnetisation direction not only for magnetic cr ions but also in @xmath0 , @xmath2 , and @xmath50 spheres in which small but finite magnetic moments are induced . in a general case this direction is not uniquely determined by @xmath63 but instead depends in a nontrivial way on the orientation of cr magnetic moments .
we found that good agreement between the @xmath63-dependencies of the total @xmath160 and band @xmath161 energies can be obtained when the small exchange splitting in @xmath0 , @xmath2 , and @xmath50 spheres is completely neglected .
@xmath161 curves calculated in this way for some of the cr spinels are compared to @xmath160 in fig .
[ fig : eft2 t ] .
both energies are plotted relative to the energy of the fm solution with @xmath155=0 .
the difference between @xmath161 and @xmath160 does not exceed 5 mev and is much smaller than the variation of the total energy with @xmath155 even for cdcr@xmath13o@xmath14 .
somewhat larger differences of about 8 mev are calculated for sulfides ( not shown in fig . [
fig : eft2 t ] ) . if the spin - polarisation inside non - cr spheres was not switched - off and the magnetisation directions obtained from self - consistent calculations were used in lft calculations , slightly larger differences between the total and band energies were obtained .
below , when discussing exchange coupling constants between cr spins , we will drop the subscript @xmath162 and denote the band energy calculated with the spin - polarisation in @xmath0 , @xmath2 , and @xmath50 spheres set to zero simply as @xmath163 .
the calculations for spin spirals confirm the conclusion , drawn in sec.[sec : lsda ] on the base of comparison of the fm and zm total energies , that the energy of the zero net moment solution ( @xmath155=2 ) is lower than the fm one ( @xmath155=0 ) in oxides whereas in @xmath0cr@xmath13s(e)@xmath14 the fm solution becomes more stable .
however , the shift of the minimum of @xmath163 curves to intermediate values of @xmath155 , most clearly seen for the @xmath156@xmath157 direction , witnesses that alongside geometrical frustrations there is also a competition between the nearest - neighbor exchange interaction and more distant ones . only in cdcr@xmath13se@xmath14 and hgcr@xmath13se@xmath14
the minimum is at @xmath155=0 and the fm solution has the lowest energy .
( color online ) exchange coupling constants @xmath26 up to the 4-th cr neighbors.,scaledwidth=30.0% ] quantitative estimates for @xmath26 up to the 4-th shell of cr neighbors were obtained by mapping the energy of spin spirals onto a classical heisenberg model @xmath164 where @xmath146 numbers cr sites in the unit cell and @xmath165 runs over @xmath166 neighbors in the @xmath167-th shell around the site @xmath146 .
positive @xmath26 correspond to afm coupling between cr spins .
note that an additional factor of 1/2 appears in eq .
( [ eq : emodel ] ) because the magnetic energy is given per formula unit and there are two formula units in the unit cell .
the length of the cr spins @xmath168 was fixed to 3/2 .
then , @xmath169depends only on the angle between a pair of cr spins which is uniquely determined by the wave vector @xmath63 of a spiral provided that for 4 cr ions at positions @xmath170 in the unit cell the phases are fixed by @xmath171 .
some pairs of cr ions coupled by @xmath3@xmath24 are shown in fig.[fig : jnn ] .
each cr site has 6 first ( @xmath3 ) and 12 second ( @xmath172 ) neighbors at distances @xmath173 and @xmath174 , respectively .
the 3-rd shell consists of 12 cr sites at @xmath175 which are split into two inequivalent sextets .
the corresponding coupling constants are denoted as @xmath176 and @xmath177 .
@xmath176 couples cr sites which lie on one of @xmath178 chains and are actually the 2-nd cr neighbors along the chain
. cr sites coupled by @xmath177 belong to parallel cr chains . for
the spin spirals considered here terms proportional to @xmath176 and @xmath177 in eq .
( [ eq : emodel ] ) have the same @xmath63-dependence so that it was not possible to separate their contributions to the magnetic energy and only their average @xmath179 could be determined from the fit .
we will return to the discussion of @xmath176 and @xmath177 later .
finally , in the 4-th shell there are 12 cr sites at the distance @xmath180 ( @xmath24 ) which lie at the same chains as the 2-nd cr neighbors .
( color online ) the comparison of calculated band energies ( @xmath181 ) and the results of the fit using eq.([eq : emodel ] ) ( @xmath182 ) for @xmath63=(0,0,@xmath155 ) ( left panels ) and @xmath63=(@xmath155,@xmath155,0 ) ( right panels ) .
individual contributions @xmath183 of the terms proportional to @xmath26 are also plotted.,scaledwidth=45.0% ] in order to get reliable values of @xmath26 , additional lft calculations for spirals with wave vectors ( @xmath155,@xmath155,@xmath155 ) , ( 1,@xmath155,0 ) , and ( 1,1,@xmath155 ) were performed .
for the ( 1,@xmath155,0 ) and ( 1,1,@xmath155 ) spirals the contributions to @xmath184 proportional to @xmath23 and @xmath3 , respectively , do not depend on @xmath155 , which allows to determine these coupling constants with higher accuracy .
explicit expressions for @xmath185 for the above mentioned spin spirals are given in appendix .
the results of a simultaneous least - squares fit of @xmath169 given by eq.([eq : emodel ] ) to @xmath163 calculated within lsda along the five @xmath63directions are shown in fig .
[ fig : jnn_fit ] together with contributions to @xmath169 coming from different nearest neighbors shells ( @xmath183 ) .
the values of @xmath186 obtained from the fit are collected in table [ tab : jnn_lsda ] .
the dominant contribution to @xmath169 is provided by the nearest neighbor coupling @xmath3 which is afm in @xmath0cr@xmath13o@xmath14 and fm in @xmath0cr@xmath13s(e)@xmath14 . in oxides , however , the strength of afm @xmath3 decreases dramatically with the increase of the @xmath0 ionic radius : in cdcr@xmath13o@xmath14 @xmath3 is more than 5 times weaker than in zncr@xmath13o@xmath14 , while in hgcr@xmath13o@xmath14 it becomes almost zero . in @xmath0cr@xmath13s(e)@xmath14
the strength of fm @xmath3 tends to increase in the row zn@xmath187cd@xmath187hg but the changes are not as strong as in the oxides ; with the values of @xmath3 calculated for the cd and hg compounds being comparable .
another significant contribution to @xmath169 comes from the term proportional to @xmath23 which is afm in all the @xmath0cr@xmath1 spinels considered in the present work .
it is the competition between the @xmath3 and @xmath23 terms which shifts the minimum of the ( 0,0,@xmath155 ) and ( @xmath155,@xmath155,0 ) curves to incommensurate @xmath63vectors . for @xmath63=(0,0,@xmath155 ) @xmath169 has the same @xmath155-dependence as a linear chain of classical spins with competing nearest @xmath188 and next - nearest @xmath189 neighbor interactions .
the ground state of such a chain becomes incommensurate if the ratio @xmath190 is larger than the critical value of 0.25 . since @xmath23 is not very sensitive to the size of @xmath0 ion , the ratio @xmath191 , which is less than 0.1 in zncr@xmath13o@xmath14 , increases to 0.45 in cdcr@xmath13o@xmath14 , while in the hg oxide @xmath23 becomes the dominant magnetic interaction .
the values of @xmath23 calulated for @xmath0cr@xmath13s(e)@xmath14 are more than 2 times larger than in the oxides .
the largest @xmath192 ratios of 0.75 and 0.48 are found for zncr@xmath13s@xmath14 and zncr@xmath13se@xmath14 , in which the fm nearest neighbor interaction is the weakest . in other two
selenides @xmath193 is only slighly larger than the critical value of 0.25 .
according to the results of the the least - square fit the exchange interactions between the 2-nd and 4-th cr neighbors are much weaker than @xmath3 and @xmath23 .
while the sign of @xmath172 varies form one compound to another , @xmath24 is always ferromagnetic and somewhat larger in sulfides and selenides compared to oxides .
the reliability of the fitted @xmath26 was checked by comparing the values obtained from the fit to the band energies calculated for all 5 spin spirals ( @xmath194 ) to those fitted to ( 0,0,@xmath155 ) and ( @xmath155,@xmath155,0 ) results only ( @xmath195 ) . in the case of @xmath3 , the relative uncertainty of the determination of @xmath26 , defined as @xmath196 , is less than 1% , except for hgcr@xmath13o@xmath14 where @xmath3 itself is very small .
the uncertainty in @xmath23 values is about 1% in @xmath0cr@xmath13s(e)@xmath14 and 10% in oxides .
tha values of @xmath197% and @xmath198% are also smaller in @xmath0cr@xmath13s(e)@xmath14 than in the oxides , in which they are about 30% .
the largest uncertainties in the lsda values of @xmath172@xmath24 were found in the case of zncr@xmath13o@xmath14 .
a plausible reason is that the nearest - neighbors terms of the order of @xmath199 in the expansion of the magnetic energy @xcite may become comparable to the contribution of weak couplings between distant cr neighbors .
neither the fitted values of @xmath26 nor the quality of the fit were noticeably affected if exchange couplings between 5-th cr neighbors were included into the fit .
accounting for more distant 6-th neighbors allowed to decrease the mean deviation between the calculated and fitted energies for @xmath0cr@xmath13s(e)@xmath14 but the obtained values of @xmath25 were even smaller than @xmath172 and @xmath24 and could not be determined reliably .
finally , significantly smaller values of @xmath200 were obtained when fitting @xmath169 to lsda+@xmath27 band energies .
.[tab : jnn_lsda]exchange coupling constants @xmath201 ( k ) and @xmath11 ( k ) obtained from the least - squares fit of @xmath169 given by eq.([eq : emodel ] ) to the energy of spin spirals calculated within lsda .
[ cols="^ , > , > , > , > , > " , ] in order to separate @xmath176 and @xmath177 contributions to @xmath169 we performed calculations for non - collinear spin superstructures in which cr moments only in every second ( 001 ) plane , the one that contains [ 110 ] cr chains , are oriented in the @xmath202 plane ; with their directions being determined as before by @xmath153 and @xmath154 .
cr moments in other planes , which contain [ 1@xmath2030 ] chains , are aligned parallel to the @xmath17 axis ( @xmath204 ) so that their directions do not depend on @xmath205 and , consequently , on @xmath63 . in this case
the 3-rd cr neighbors in the @xmath202 plane that lie on the same [ 110 ] chain ( @xmath176 ) and on parallel chains ( @xmath177 ) give different contributions to the @xmath155-dependence of the energy of ( @xmath155,@xmath155,0 ) and ( @xmath155,@xmath206,0 ) spin spirals , which allows to determine @xmath176 and @xmath177 separately provided that other @xmath26 are known .
calculations performed for @xmath0cr@xmath13o@xmath14 and @xmath0cr@xmath13se@xmath14 show that @xmath176 and @xmath177 are of comparable strengths .
the calculated @xmath207 ratio is 0.6 , 0.5 , and 1.1 in zn , cd , and hg oxides , respectively , while for the selenides the values of 0.8 , 0.8 , and 1.0 were obtained .
it seems plausible that the contribution of hg 6@xmath128 states to cr @xmath9@xmath2 @xmath6cr @xmath9 hybridisation is responsible for somewhat larger values of @xmath177 in the hg compounds .
it should be mentioned that the values of @xmath26 presented in table [ tab : jnn_u ] are somewhat different from the preliminary results published in ref . for the following reasons : i ) for cdcr@xmath13o@xmath14 old structural data from ref . with the o fractional
coordinate @xmath34=0.260 , which seems to be too small , were used ; ii ) the calculations for oxides in ref .
were performed with o @xmath128 and @xmath6 states only included in the lmto basis set .
this gives smaller values of @xmath3 compared to the present calculations in which o @xmath9 are also included into the basis ; iii ) the calculations in ref .
were performed for ( @xmath155,@xmath155,0 ) and ( 0,0,@xmath155 ) spirals only and the coupling constants between 3-rd cr neighbors along the @xmath208 chains were included into the fit instead of @xmath24 .
later it was verified that accounting for true @xmath24 to 4-th cr neighbors improves the quality of the fit and seems more physical .
we have to stress , however , that despite the differences in calculated values of @xmath26 the main conclusions of ref . concerning the relative strengths of various exchange interactions and their origins remain unaffected .
the effective curie - weiss temperatures estimated from the calculatated @xmath26 according to @xmath209 with @xmath10=3/2 are given in the last column of tables [ tab : jnn_lsda ] and [ tab : jnn_u ] .
the comparison of the estimated @xmath11 to experimental values from ref .
( table [ tab : str ] ) shows that the lsda calculations reproduce the opposite signs of @xmath11 in @xmath0cr@xmath13o@xmath14 and @xmath0cr@xmath13s(e)@xmath14but strongly overestimate its magnitude , which is not surprising as one can expect that lsda underestimates correlation effects in the rather localised cr 3@xmath9 shell .
since the absolute values of @xmath11 calculated for oxides rapidly decrease with the increase of @xmath27 in the lsda+@xmath27 calculations , it is possible to adjust the value of @xmath27 so as to reproduce the experimental @xmath11 .
this , however , would require to use different values of @xmath27 for the zn , cd , and hg oxospinels . in @xmath0cr@xmath13s(e)@xmath14
the estimated values of @xmath11 , which are too high already in lsda , further increase when the lsda+@xmath27 approach is used . as the dominant contribution to @xmath11
is provided by @xmath3 , it seems that the tendency to ferromagnetic coupling between nearest cr neighbors is overestimated in the lsda+@xmath27 calculations .
the comparison of the calculated @xmath26 to experimental estimates for the exchange coupling constants in zncr@xmath13s@xmath14 ( @xmath3=2.66 k , @[email protected] k , @xmath23=0.29 k , @xmath24=0.13 k ) @xcite and cdcr@xmath13s@xmath14 ( @xmath3=13.25 k , @[email protected] k ) @xcite also shows that the calculations overestimate the strength of the exchange interactions between cr spins .
note that the experimental values of @xmath26 should be multiplied by a factor of 2 because of the different definition of the magnetic energy [ see eq .
( 2 ) in ref . ] , while theirs signs should be inverted since in refs .
the positive sign of @xmath26 corresponds to fm coupling .
also it is worth mentioning that @xmath26 are not directly measurable quantities and in order to determine them from experimental data additional assumption should be made .
the experimental wave vector @xmath210(0,0,0.79 ) of the helimagnetic structure in zncr@xmath13s@xmath14is very close to @xmath211(0,0,0.82 ) at which the lsda energy of the ( 0,0,@xmath155 ) spin spiral has a minimum . under assumption
that @xmath172 and @xmath24 are much smaller than @xmath3 and @xmath23 this gives the @xmath212 ratio of 0.39 and 0.37 in the experiment and theory , respectively . for zncr@xmath13se@xmath14
lsda gives the minimum of @xmath213 along the @xmath156@xmath2 at @xmath211(0,0,0.65 ) which is larger than the experimental wave vector @xmath210(0,0,0.466 ) .
@xcite in the lsda+@xmath27 calculations , however , the minimum shifts towards smaller @xmath155 values .
in order to understand better the dependence of the calculated exchange coupling constants on the chemical composition and on the value of the parameter @xmath27 in lsda+@xmath27 calculations , let us for simplicity consider a pair of cr@xmath7 ions with occupied @xmath132 states at zero energy , unoccupied @xmath133 states at the energy @xmath214 , and an effective @xmath4@xmath4 hopping matrix element @xmath215 .
the splitting @xmath216 is close to @xmath217 in the lsda and @xmath218 in the lsda+@xmath27 calculations .
cr @xmath109 states with both spin projections are unoccupied and also split by @xmath219 which in lsda is equal to 3@xmath90 .
the lsda energy difference @xmath220 between @xmath109 and @xmath4states with the same spin projections is equal to the crystal field splitting @xmath221 which is of the same order as @xmath219 as is evidenced by the results presented in sec . [
sec : lsda ] .
( color online ) a sketch of bonding and antibonding states between cr @xmath4 state for fm ( left ) and afm ( right ) alignments of cr magnetisations .
the fermi level is plotted by a horizontal dotted line.,scaledwidth=44.0% ] hybridisation , either direct or indirect via @xmath2 @xmath6 states , between the half - filled cr @xmath4 states results in afm exchange coupling provided that the @xmath4 bandwidth is less than @xmath214 .
indeed , if cr moments are aligned ferromagnetically , the hybridisation between the @xmath4 states gives no gain in the kinetic energy as in the majority - spin chanel both bonding and antibonding combinations with energies @xmath222 and @xmath223 are occupied , whereas in the minority - spin chanel both of them ( @xmath224 ) are empty ( see fig.[fig : t2gt2 g ] ) .
if the magnetisations of two cr ions are antiparallel , an occupied @xmath132 state can hybridise only with the unoccupied @xmath4 state with the same spin projection of another ion lying at the energy @xmath214 . in this case only their bonding combination with the energy @xmath225 is occupied . since the energy of the bonding state with the opposite spin projection is exactly the same , this gives @xmath226 for the energy difference of the fm and afm configurations .
cr @xmath5 states with both spin projections are unoccupied and @xmath4@xmath109hybridisation ( @xmath227 ) lowers the energy of the fm as well as afm configuration because in both cases only the bonding combination is occupied ( see fig .
[ fig : t2geg ] ) .
the energy of the bonding state @xmath228 is , however , lower if the magnetisations of two cr ions are parallel . in this case
the @xmath109 partner with the same spin as the occupied @xmath4 state has lower energy @xmath229 vs.@xmath230 in the afm case , so that @xmath231 and the effective coupling @xmath232 is ferromagnetic .
it should be mentioned that these estimates for the effective exchange couplings are in agreement with the well known goodenough - kanamori rules .
@xcite ( color online ) a sketch of bonding and antibonding states between an occupied cr @xmath132 state and an unoccupied @xmath109 state with the same spin projection for fm ( left ) and afm ( right ) alignments of cr magnetisations .
the fermi level is plotted by a horizontal dotted line.,scaledwidth=44.0% ] thus , the strength and even the sign of @xmath3 in @xmath0cr@xmath1 are determined by the competition of two contributions of opposite signs , @xmath233 and @xmath232 , which depend on the strength of the @xmath4@xmath4 and @xmath4@xmath5hybridisation , respectively . in zncr@xmath13o@xmath14 @xmath233 dominates because of the strong direct hopping of @xmath115 type between the cr @xmath4 states .
this conclusion has been confirmed numerically by the results of test lsda calculations in which the hybridisation between cr @xmath9 states was switched off .
@xmath3 derived from a fit to band energies calculated in this way is about 1 k , whereas for other @xmath26 the values comparable to those given in table [ tab : jnn_lsda ] were obtained . with the increase of the lattice constant in the row
zn@xmath187cd@xmath187hg the strength of the direct @xmath9@xmath9 hopping and , as a consequence , of @xmath233 rapidly decreases so that in hgcr@xmath13o@xmath14 according to the lsda results @xmath233 and @xmath232 almost exactly compensate each other .
the dependence of @xmath3 on the value of the parameter @xmath27 in the lsda+@xmath27 calculations can be explained by weakening of @xmath233 caused by the increase of the denominator in eq.([eq : jafm ] ) which in this case is given by @xmath234 . in @xmath0cr@xmath13s(e)@xmath14
spinels the distance between cr nearest neighbors ( see table [ tab : str ] ) is much larger than in @xmath0cr@xmath13o@xmath14 and the contribution of the direct @xmath4@xmath4 hopping to @xmath3 becomes less significant . for zncr@xmath13se@xmath14 , for instance , calculations with switched - off cr @xmath9@xmath9 hybridisation give the value of fm @xmath3 which is only 5% larger than the lsda one .
thus , the indirect hybridisation via @xmath2 @xmath6 states between the half - filled @xmath4 and empty @xmath5 states split by the on - site exchange interaction provides the dominant contribution to fm @xmath3 as was proposed by j. goodenough in ref .. a direct numerical proof is provided by test calculations in which an external orbital - dependent potential @xmath235 was added to cr @xmath134 states ; with its strength @xmath236 being equal to the lsda exchange splitting of the @xmath109 states .
then , the energies of cr @xmath109 states with both spin projections are equal so that @xmath237 and the nominator in the expression ( [ eq : jfm ] ) for @xmath232 becomes zero .
@xmath3=15 k estimated from such calculations for zncr@xmath13se@xmath14 changes its sign to afm and its absolute value is significantly smaller than @xmath238=54 k obtained from the lsda calculations .
this indicates that the indirect @xmath4@xmath2 @xmath6@xmath4 hybridisation also contribute to @xmath3 but it is much weaker than @xmath232 .
similar calculations for zncr@xmath13o@xmath14 result in substantial increase of the strength of afm @xmath3 which means the fm contribution is present also in oxides but @xmath233 wins due to direct @xmath9@xmath9 hybridisation . in order to explain the weird behavior of @xmath3 in lsda+@xmath27 calculations one needs to analyze in more details various parameters entering eq.([eq : jfm ] ) . recalling that the orbital dependent part of the lsda+@xmath27potential can be approximated as @xmath239 with @xmath147 one gets @xmath240 where @xmath241 and @xmath242 are the occupation numbers of @xmath134 and @xmath151 states , respectively .
here we neglected the small ( @xmath2430.05 ) deviation of the occupation numbers of the minority - spin @xmath4states from unity .
then , eq . ( [ eq : jfm ] ) can be written as @xmath244 } { { \ensuremath{\delta_{cf}}}+u'(1-n_{\uparrow } ) } { \nonumber}\\ & \times & \frac{1}{{\ensuremath{\delta_{cf}}}+u'(1-n_{\downarrow})+3{\ensuremath{j_{\text{h } } } } } \ , .
\label{eq : jfmu}\end{aligned}\ ] ] if the occupations of the cr @xmath109 were equal to zero eq .
( [ eq : jfmu ] ) would become @xmath245 which would lead to fast decrease of @xmath232 with the increase of @xmath246 .
however , since @xmath247 and @xmath248 were calculated for @xmath0cr@xmath13s(e)@xmath14 the appearance of the term proportional to @xmath249 in the nominator of eq .
( [ eq : jfmu ] ) slows down the decrease of @xmath232 .
also the terms proportional to @xmath241 and @xmath242 in the dominator of eq.([eq : jfmuion ] ) effectively decrease @xmath246 , thus leading to the increase of @xmath232 as compared to purely ionic model with zero occupations of the @xmath109 states .
this is one of possible reasons why the fm contribution to @xmath3 in overestimated in the lsda+@xmath27 calculations .
the calculated behavior of @xmath172 , @xmath23 , and @xmath24 follows the predictions made by k. dwight and n. menyuk in refs . on the base of the goodenough - kanamori rules of superexchange.@xcite hybridisation paths
cr @xmath9@xmath2 @xmath6@xmath2 @xmath6cr @xmath9 responsible for these interactions include an intermediate hopping between @xmath6 states of two @xmath2 ions .
this reduces the strength of the superexchange interactions but does not change the signs of @xmath4@xmath4 and @xmath4-@xmath5 contributions to @xmath26 , which remain antiferro- and ferromagnetic , respectively . indirect @xmath4@xmath4 and @xmath4@xmath5 hoppings give comparable contributions to @xmath172 .
being of opposite signs , these contributions compensate each other .
this explains the smallness of @xmath172 and the variation of its sign in the @xmath0cr@xmath1 spinels .
similar hybridisation paths between @xmath9 states of 3-rd neighbors , with the dominant contribution of the indirect @xmath4@xmath4 hopping , lead to the appearance of afm @xmath176 and @xmath177 . in the real structure
the distance between the pair of @xmath2 ions along the @xmath176 path is smaller than along the @xmath177 one ; the difference being proportional to the deviation of the positional parameter @xmath34 from 1/4 .
this may explain why the calculated values of @xmath176 are somewhat larger than @xmath177 .
on the other hand , since the @xmath177 pair of @xmath2 ions belongs to the same @xmath37 tetrahedron , @xmath177 is more sensitive to @xmath0 @xmath128@xmath2 @xmath6 hybridisation , which may be responsible for the increase of the @xmath207 ratio in the hg - containing spinels .
finally , indirect hybridisation between @xmath4 and @xmath5 states gives dominant contribution to the superexchange between 4-th neighbors @xcite which leads to fm @xmath24 .
the lsda(@xmath250 ) band structure calculations performed for some @xmath0cr@xmath1 spinels reproduce the change of the sign of the nearest - neighbor exchange interaction between cr spins from afm in @xmath0cr@xmath13o@xmath14 compounds to fm in @xmath0cr@xmath13s(e)@xmath14ones .
it has been verified that the strength and the sign of @xmath3 depend on relative strengths of two contributions of opposite signs , afm and fm , which are due to cr @xmath4@xmath4 and @xmath4@xmath5 hybridisations , respectively . in zncr@xmath13o@xmath14 @xmath233 dominates because of the strong direct hopping between the cr @xmath4 states .
as zn is replaced by cd or hg , the strength of the direct hopping rapidly decreases with the increase of the cr cr separation . as a result ,
@xmath3 in the corresponding oxides is much weaker than in zncr@xmath13o@xmath14 . in the @xmath0cr@xmath13s(e)@xmath14
spinels the fm contribution originating from the indirect hopping between half - filled @xmath4 and empty @xmath5 states , split by the on - site exchange interaction , becomes dominant
. however , the magnitute of the fm contribution to @xmath3 seems to be overestimated by the calculations .
the strongest among more distant exchange couplings are @xmath176 and @xmath177 to two sextets of 3-rd cr neighbors .
these interactions are always afm and of comparable strengths .
the couplings to 2-nd and 4-th cr neighbors are significantly weaker than @xmath3 and @xmath23 .
the obtained results are in accordance with the analysis of the exchange couplings in cr spinels based on the goodenough - kanamori rules of superexchange and may be helpful for understanding the magnetic properties not only of cr spinels but also ti ( mgti@xmath13o@xmath14 ) and v ( znv@xmath13o@xmath14 , cdv@xmath13o@xmath14 ) ones , in which orbital degrees of freedom come into play .
@xcite author is grateful to p. fulde , n. shannon , k. penc , and v. antonov for many helpful discussions and to h. ueda for providing recent experimental data on the crystal structures of cdcr@xmath13o@xmath14 and hgcr@xmath13o@xmath14 spinels .
below explicite expressions for the difference @xmath251 , where @xmath184 is given by eq .
( [ eq : emodel ] ) and @xmath252 are written as a functions of the angle @xmath253 for some directions of the wave vector @xmath63 of spin spirals : + @xmath63=(0,0,@xmath155 ) ( @xmath156@xmath2 ) : @xmath254 @xmath63=(@xmath155,@xmath155,0 ) ( @xmath156@xmath157 ) : @xmath255 @xmath63=(@xmath155,@xmath155,@xmath155 ) ( @xmath156@xmath256 ) : @xmath257 @xmath63=(1,@xmath155,0 ) : @xmath258 @xmath63=(1,1,@xmath155 ) : @xmath259 here @xmath260 is the average of the exchange coupling constants for two inequivalent groups of 3-rd cr neighbors .
|
we present the results of band structure calculations for @xmath0cr@xmath1 ( @xmath0=zn , cd , hg and @xmath2=o , s , se ) spinels .
effective exchange coupling constants between cr spins are determined by fitting the energy of spin spirals to a classical heisenberg model .
the calculations reproduce the change of the sign of the dominant nearest - neighbor exchange interaction @xmath3 from antiferromagnetic in oxides to ferromagnetic in sulfides and selenides .
it is verified that the ferromagnetic contribution to @xmath3 is due to indirect hopping between cr @xmath4 and @xmath5 states via @xmath2 @xmath6 states .
antiferromagnetic coupling between 3-rd cr neighbors is found to be important in all the @xmath0cr@xmath1 spinels studied , whereas other interactions are much weaker .
the results are compared to predictions based on the goodenough - kanamori rules of superexchange .
|
the ferrites materials have been widely used as various electronic devices such as inductors , transformers , and electromagnetic wave absorbers in the relatively high - frequency region up to a few hundreds of mhz .
the electromagnetic theory can be used to describe the macroscopic properties of matter .
the electromagnetic fields may be characterized by four vectors : electric field * e * , magnetic flux density * b * , electric flux density * d * , and magnetic field * h * , which at ordinary points satisfy maxwell s equations
. the ferrite media under study can be considerer as linear , homogeneous , and isotropic .
the relation between the vectors * b * and * h * can be expressed as : * b * = @xmath0*h * . where @xmath1 is the magnetic permeability of the material .
another important parameter for magnetic materials is magnetic susceptibility @xmath2 which relates the magnetization vector m to the magnetic field vector * h * by the relationship : * m * = @xmath3 * h * .
magnetic permeability @xmath1 and magnetic susceptibility @xmath2 are related by the formula : @xmath4 .
magnetic materials in sinusoidal fields have , in fact , magnetic losses and this can be expressed taking @xmath1 as a complex parameter:@xmath5 @xcite in the frequency range from rf to microwaves , the complex permeability spectra of the ferrites can be characterized by two different magnetization mechanisms : domain wall motion and gyromagnetic spin rotation .
domain wall motion contribution to susceptibility can be studied through an equation of motion in which pressure is proportional to the magnetic field @xcite .
assuming that the magnetic field has harmonic excitation @xmath6 , the contribution of domain wall to the susceptibility @xmath7 is : @xmath8 here , @xmath9 is the magnetic susceptibility for domain wall , @xmath10 is the resonance frequency of domain wall contribution , @xmath11 is the static magnetic susceceptibility , @xmath12 is the damping factor and @xmath13 is the frequency of the external magnetic field .
gyromagnetic spin contribution to magnetic susceptibility can be studied through a magnetodynamic equation @xcite@xcite .
the magnetic susceptibility @xmath14 can be expressed as : @xmath15 here , @xmath14 is the magnetic susceptibility for gyromagnetic spin , @xmath16 is the resonance frequency of spin contribution , @xmath17 is the static magnetic susceptibility , and @xmath18 is the damping factor and @xmath13 is the frequency of the external magnetic field .
the total magnetic permeability results @xcite : @xmath19 separating the real and the imaginary parts of equation ( [ eq : modelo ] ) we get : @xmath20 @xmath21 magnetic losses , represented by the imaginary part of the magnetic permeability , can be extremely small ; however , they are always present unless we consider vacuum @xcite . from a physics point of view , the existing relationship between @xmath22 and @xmath23 reflects that the mechanisms of energy storage and dissipation are two aspects of the same phenomenon @xcite .
genetic algorithms ( ga ) are probabilistic search procedures which generate solutions to optimization problems using techniques inspired by natural evolution , such as inheritance , mutation , selection , and crossover .
a ga allows a population composed of many individuals evolve according to selection rules designed to maximize `` fitness '' or minimize a `` cost function '' .
a path through the components of ag is shown as a flowchart in figure ( [ fig : diagramaflujo ] ) a cost function generates an output from a set of input variables ( a chromosome ) .
the cost function s object is to modify the output in some desirable fashion by finding the appropriate values for the input variables .
the cost function in this work is the difference between the experimental value of the permeability and calculated using the parameters obtained by the genetic algorithm . to begin the ag
is randomly generated an initial population of chromosomes .
this population is represented by a matrix in which each row is a chromosome that contains the variables to optimize , in this work , the parameters of permeability model .
@xcite survival of the fittest translates into discarding the chromosomes with the highest cost .
first , the costs and associated chromosomes are ranked from lowest cost to highest cost .
then , only the best are selected to continue , while the rest are deleted .
the selection rate , is the fraction of chromosomes that survives for the next step of mating .
now two chromosomes are selected from the set surviving to produce two new offspring which contain traits from each parent .
chromosomes with lower cost are more likely to be selected from the chromosomes that survive natural selection .
offsprings are born to replace the discarded chromosomes the simplest methods choose one or more points in the chromosome to mark as the crossover points .
then the variables between these points are merely swapped between the two parents .
crossover points are randomly selected .
if care is not taken , the ga can converge too quickly into one region of a local minimum of the cost function rather than a global minimum . to avoid this problem of overly fast convergence
, we force the routine to explore other areas of the cost surface by randomly introducing changes , or mutations , in some of the variables . the process described is iterated until an acceptable solution is found .
the individuals of the new generation ( selected , crossed and mutated ) repeat the whole process until it reaches a termination criterion . in this case
, we consider a maximum number of iterations or a predefined acceptable solution ( whichever comes first )
@xmath24 samples were prepared via sol - gel method with x=0.01 , 0.02 , and 0.05 .
the complex permeability of the samples was measured in a material analyzer hp4251 in the range of 1mhz to 1 ghz.@xcite .
the experimental data of magnetic permeability have been used for fitting the parameters of the model @xcite .
firstly , we fitted the magnetic losses based in equation ( [ muimag ] ) by the genetic algoritm method and we obtained the six fitting parameters .
we substituted , then , these six parameters into equation ( [ mureal ] ) to calculate the real part of permeability .
where @xmath13 is the frequency of the external magnetic field , y @xmath9 , @xmath14 , @xmath25 , @xmath16 , @xmath12 @xmath18 are unknown parameters , the problem is to estimate these from a set of pairs of experimental:@xmath27 .
the cost function was the mistake made in calculating @xmath28 with the expression ( [ muimag ] ) using the parameters obtained from the genetic algorithm and the experimental value of @xmath29 for each frequency .
2000 iterations were performed , with a population of 300 chromosomes ( each with 6 variables ) .
the fraction of the population that was replaced by children in each iteration was 0.5 and the fraction of mutations was 0.25 .
the figure [ fig : mincost ] shows the evolution of the error in successive iterations , we graphed the minimum value of function cost in the population for each iteration .
it show that the error converges at minimum value quickly , and then this value is stable .
although in the equations ( [ muimag ] ) and ( [ mureal ] ) , @xmath10 and @xmath16 must be in hz , we calculate in mhz and then multiplied in the equations by @xmath31 , the same treatment we performed with the parameter beta , we calculate its value in the range between 1 and 2000 , although in equations we multiplied by @xmath32 . this treatment was necessary for that
the ag values to variables within a limited range .
table [ tab : tabla1 ] shows the parameters of the model calculated for the three nizn ferrite samples doped with different amounts of yttrium .
figure [ fig : permeajuste ] graphs ( a ) , ( b ) and ( c ) shows the permeability spectra for the three ferrites .
solid lines represent the curves constructed from the adjusted parameters while dotted and dashed lines represent the curves from experimental data .
99 haupt r.l , haupt s.e `` practical genetic algorithms.''wiley - interscience publication ( 1998 ) a.von hippel .
`` dielectrics and waves '' . j. wiley sons .
r. f. sohoo .
`` theory and application of ferrites '' , prentice hall , nj , usa ( 1960 ) trainotti v. and fano w. `` ingeniera electromagnetica '' .
nueva libreria , 2004 .
landau and lifchitz .
`` electrodinamica de los medios continuos '' .
revert , 1981 .
w.g.fano , s.boggi , a.c.razzitte , `` causality study and numerical response of the magnetic permeability as a function of the frequency of ferrites using kramers kronig relations '' physica b , 403 , 526 - 530 , ( 2008 ) greiner .
`` clasical electrodynamics '' .
cap.16 , springer ( 1998 ) tsutaoka , t. `` frequency dispersin of complex permeability in mn - zn and ni - zn spinel ferrites and their composite materials''j ournal of applied physics , volumen 93 ( 2003 ) wohlfarth e. `` ferromagnetics materials '' , volumen 2 .
north holland , 1980 .
s.e.jacobo , s duhalde , h.r.bertorello , journal of magnetism and magnetic materials 272 - 276 ( 2004 ) 2253 - 2254 .
silvina boggi , adrin c. razzitte and water g.fano , _ non - equilibrium thermodynamics and entropy production spectra : a tool for the characterization of ferrimagnetic materials _ , journal of non - equilibrium thermodynamics .
volume 38 , issue 2 p.175 - 183 ( 2013 ) .
t. tsutaoka , t._frequency dispersin of complex permeability in mn - zn and ni - zn spinel ferrites and their composite materials _ , journal of applied physics , v 93 ( 2003 )
|
the magnetic permeability of a ferrite is an important factor in designing devices such as inductors , transformers , and microwave absorbing materials among others . due to this
, it is advisable to study the magnetic permeability of a ferrite as a function of frequency .
when an excitation that corresponds to a harmonic magnetic field * h * is applied to the system , this system responds with a magnetic flux density * b * ; the relation between these two vectors can be expressed as * b * = @xmath0 * h * . where @xmath1 is the magnetic permeability . in this paper ,
ferrites were considered linear , homogeneous , and isotropic materials .
a magnetic permeability model was applied to nizn ferrites doped with yttrium .
the parameters of the model were adjusted using the genetic algorithm . in the computer science field of artificial intelligence , genetic algorithms and machine learning
does rely upon nature s bounty for both inspiration nature s and mechanisms .
genetic algorithms are probabilistic search procedures which generate solutions to optimization problems using techniques inspired by natural evolution , such as inheritance , mutation , selection , and crossover . for the numerical fitting
usually is used a nonlinear least square method , this algorithm is based on calculus by starting from an initial set of variable values .
this approach is mathematically elegant compared to the exhaustive or random searches but tends easily to get stuck in local minima .
on the other hand , random methods use some probabilistic calculations to find variable sets .
they tend to be slower but have greater success at finding the global minimum regardless of the initial values of the variables
|
the interpretation of the pomeron in the framework of qcd is not fully understood .
it is expected to be generated at least by two gluons exchange @xcite . however , the exchange of two perturbative gluons can not reproduce the experimental results on diffractive scattering . in order to circumvent such difficulty , landshoff and nachtmann
@xcite(ln ) proposed a model where the pomeron is described by the exchange of two non - perturbative gluons , whose properties are dictated by the expected structure of the qcd vacuum .
these non - perturbative gluons should not propagate over long distances , i.e. there is a finite correlation length for the gluon field in the vacuum , which should be determined from first principles , and can be understood in terms of gluon condensates @xcite .
the ln model describes successfully the experimental results of diffractive phenomena using a quite simple ansatz for the non - perturbative gluon propagator @xcite .
recently we improved the model using a qcd motivated non - perturbative gluon propagator , where the correlation length is provided by a dynamically generated gluon mass @xcite .
the remarkable success of this pomeron model might provide physical insight in the properties of the gluon propagator in the infrared region .
it is , therefore , natural to speculate on the ability of such a model to describe other low - energy qcd phenomena , in particular , dynamical chiral symmetry breaking ( dcsb ) . within the spirit of the ln model
, we study the compatibility of both the pomeron and dcsb phenomenologies .
the quark self - energy is computed through the numerical calculation of the schwinger - dyson equation ( sde ) for the quark propagator , employing the same non - perturbative gluon propagators as the ones used in the pomeron phenomenology . according to the traditional approach to the dcsb @xcite
, these self - energies are used to obtain the experimental values of the quantities that characterize the chiral breaking , such as the quark condensate and the pion decay constant .
it is worth mentioning that the existence of a finite correlation length for the gluon field was already proposed ten years before the ln model in a completely different context @xcite , and it gained strong support by recent lattice simulation of the gluon propagator @xcite .
therefore , a study of its phenomenological implications is quite compelling , although there are few places where this phenomenon can be effectively tested . using only qcd motivated gluon propagators , _
i.e. _ propagators determined from the solutions of schwinger - dyson equations or obtained from lattice simulation , which satisfy the ln conditions and are constrained by the diffractive scattering data , we show that they do not lead to satisfactory dcsb parameters . in section
ii we discuss some of the non - perturbative propagators found in the literature , and show how they are constrained by the diffractive scattering data according to the ln model . in section
iii we use the non - perturbative gluon propagators , with the parameters found in the previous section , to compute the relevant quantities for dcsb . in section
iv we discuss the incompatibility found in the previous section and point out the possibles origins of such a failure .
there are several experimental features of diffractive scattering that can be computed through the two - gluon exchange model of the pomeron according to the ln prescription , and here we will choose two simple quantities which can be related to hadron - hadron total cross sections , and show how they constrain the gluon propagators .
the first one will be the quark - pomeron coupling and the second one , the pion - proton total cross section .
the strength of the pomeron coupling to quarks @xmath0 at leading order is given by @xcite @xmath1 ^ 2 , \label{e1}\ ] ] where @xmath2 is the running quark - gluon coupling , and the value of @xmath3 gev@xmath4 is extracted from the proton - proton cross section .
notice that we will be working with euclidean momenta @xmath5 .
the amplitude of meson - nucleon scattering is @xcite @xmath6 [ \alpha d((2{\bf q}-{\bf k})^2 ) ] \nonumber \\ & & \times 2 [ f_m(q^2)-f_m(({\bf q } - { \bf k})^2 ) ] \nonumber \\ & & \times 3 \left [ f_n(q^2 ) - f_n \left ( q^2 - \frac{3}{2 } { \bf q.k } + \frac{3}{4 } k^2
\right ) \right ] , \label{e2}\end{aligned}\ ] ] where @xmath7 is the square of the center - of - mass energy , to the couplings @xmath8 we associate the same momentum of its multiplying propagator , @xmath9 and @xmath10 are respectively the meson and nucleon form factors .
the total cross section is related to this amplitude by @xmath11 and eq.([e2 ] ) is computed with the form factors in the pole approximation @xmath12 we will calculate the pion - proton total cross section .
actually , the two gluon exchange model can not describe the full cross section growth with energy @xmath13 , which appears in a recent fit for the @xmath14 total cross section @xcite @xmath15 the model can only accommodate the value @xmath16 in the above expression .
we used the following mean squared radii for the proton and pion , respectively @xcite : @xmath17 @xmath18 and @xmath19 @xmath18 .
inspection of eq.([e1 ] ) and eq.([e2 ] ) show that they are quite dependent on the gluon propagator expression at @xmath20 and , therefore , well suited to study its infrared behavior .
we will now consider the propagators that satisfy some of the basic assumptions of the ln model : ( a ) the propagator has a finite correlation length and ( b ) it is finite at @xmath21
. we will only deal with qcd motivated propagators , _
i. e. _ those obtained as solutions of the sde for the gluon polarization tensor , as well as obtained in a lattice simulation .
we start with the propagator determined by cornwall @xcite , which has already been used by one of us to describe some diffractive scattering processes @xcite , and has the following expression @xmath22 bg^2 \ln \left [ \frac{k^2 + 4m^2(k^2 ) } { \lambda^2 } \right ] , \label{e6}\ ] ] where @xmath23 is a momentum - dependent dynamical mass @xmath24^{-12/11}. \label{e7}\ ] ] in the above equation @xmath25 is the effective gluon mass , and @xmath26 is the leading order coefficient of the @xmath27 function of the renormalization group equation , with @xmath28 for qcd , and where @xmath29 is the number of flavors taken to be 3 . in eq.([e6 ] ) , the quark - gluon coupling strength @xmath30 is such that @xmath31 , which was determined from a fit of eq([e6 ] ) to the numerical solution of the sde for the gluon propagator . for the running coupling constant @xmath32
we assume a functional form which interpolates between a constant and the renormalization group asymptotic behavior @xmath33 where @xmath34 , and we chose @xmath35 .
such freeze - out of the coupling constant in the infrared region is consistent with the cornwall and papavassiliou @xcite study of the trilinear gluon vertex . a similar form for the coupling constant has also been used in many phenomenological applications ( see , for instance , ref .
@xcite and references therein ) .
it has also been pointed out that the dynamical gluon mass may have a faster decrease with the momentum @xcite , according to the operator product expansion ( ope ) determination of the ultraviolet behavior of the gluon polarization tensor @xcite @xmath36 where @xmath37 is the gluon condensate .
therefore , to be consistent with the massive cornwall propagator and with ope , we will consider a gluon self - energy that interpolates between the constant infrared behavior of eq.([e7 ] ) and the ultraviolet one of eq.([e9 ] ) which will be given by @xmath38 where the scale @xmath39 will be limited by eq.([e1 ] ) and eq.([e2 ] ) .
another infrared finite propagator has been found by stingl and collaborators as a solution of the sde for the gluon polarization tensor @xcite .
its form agrees with that derived by zwanziger based on considerations related to the gribov horizon @xcite , and is given by @xmath40 where @xmath41 is a mass scale not determined in ref .
the term @xmath42 is exactly what is expected by ope analysis whenever a mass scale for the gluon is introduced , and the full solution found in ref .
@xcite did contain a mass term , although its consequences were not pursued .
it is also interesting to notice that the bernard _
_ lattice result for the gluon propagator @xcite can be fitted by eq.([e6 ] ) as well as eq.([e11 ] ) .
finally , marenzoni _ et al . _
@xcite also performed a lattice study of the gluon propagator in the landau gauge , obtaining for its infrared behavior the following fit @xmath43 where @xmath44 , @xmath45 and @xmath46 are constants determined by the numerical simulation .
@xmath44 is of @xmath47 , @xmath48 and @xmath49 what is slightly different from the previous propagators .
the results of bernard _
_ also show the behavior @xmath50 , but with a smaller value for @xmath46 . in this case
all the propagator parameters are determined , and we simply have to see if they are consistent with diffractive scattering and dcsb . all the above propagators where obtained in landau gauge , except cornwall s one , whose massive solution was shown to be gauge invariant . performing the calculation of the integrals in eqs .
( [ e1 ] ) and ( [ e2 ] ) for each one of the above propagators , we obtained the curves for @xmath51 displayed in fig.(1 ) , and the pion - proton total cross section shown in fig.(2 ) as functions of the gluon mass . from these figures and the experimental values of @xmath52 , and the pion - proton total cross section , it was possible to establish an effective gluon mass for each propagator .
our results are shown in table [ table1 ] .
it is clear from this table that almost all the gluon mass scales that fit the experimental value of @xmath52 give reasonable ( within @xmath53 ) values for the total pion - proton cross section , except in the case of the lattice propagator , eq.([e12 ] ) , where the comparison with the experimental values yields effective gluon masses differing by @xmath54 .
it is evident that not only the mass scale plays a role in this phenomenology , but also the functional form of the propagator introduces differences in the calculation , and this is why this problem becomes more interesting , because we can speculate about the infrared behavior of the gluon propagator . as we shall see in the next section , we will fail to obtain dcsb with the above propagators for the gluon mass scales presented in table [ table1 ] .
we discuss dynamical chiral symmetry breaking following the traditional approach @xcite .
this consists in solving the schwinger - dyson equation ( sde ) for the quark propagator and look for a mass term dynamically generated in this propagator .
the schwinger - dyson equation for the quark propagator is @xmath55 where we write the gluon propagator in the form @xmath56 the propagator has been written in the landau gauge , which will be used throughout our work . in the above equations @xmath57 is the vertex function , and
@xmath58 is the qcd running coupling constant given by eq.([e8 ] ) . to proceed
further we also need to introduce an ansatz for the quark - gluon vertex @xmath59 , which must satisfy a slavnov - taylor identity that , when we neglect ghosts , reads @xmath60 this identity constrains the longitudinal part of the vertex , and if we write @xmath61 in terms of scalar functions @xmath62 we find the solution @xcite @xmath63 \not\!{q } - [ b(p^2)-b(q^2 ) ] \right ) \nonumber \\ & & + a(p^2)\gamma^\mu + \,\ , transverse \ ,
part , \label{e17}\end{aligned}\ ] ] which is a much better approximation than the use of the bare vertex @xcite .
assuming that the transverse vertex part vanishes in the landau gauge we obtain @xmath64 and arrive at the approximate schwinger - dyson equation @xmath65\not\!{p}-b(p^2 ) = \imath\frac{4}{3 } \int \frac{d^4q}{(2 \pi ) ^4 } g^2 d^{\mu \nu}(p - q)\gamma_\mu \frac{a(q^2)}{a(q^2)\not\!{q}-b(q^2 ) } \gamma_\nu .
\label{e19}\ ] ] going to euclidean space , we will be working with the following nonlinear coupled integral equations for the quark wave - function renormalization and self - energy @xmath66p^2 & = & \frac{16\pi}{3 } \int_{}^ { } \frac{d^4q}{(2\pi)^4 } \frac{\alpha ( ( p - q)^2/\lambda^2)}{\phi[(p - q)^2 ] } \nonumber \\ & & \times \left ( p.q + 2 \frac{p.(p - q)q.(p - q)}{(p - q)^2 }
\right ) \nonumber \\ & & \times \frac{a^2(q^2)}{a^2(q^2 ) q^2+b^2(q^2 ) } , \label{e20}\end{aligned}\ ] ] @xmath67 } \frac{a(q^2)b(q^2)}{a^2(q^2)q^2+b^2(q^2 ) } , \label{e21}\ ] ] where @xmath68 and @xmath69 , and we introduced a function @xmath70 $ ] which , in the case of the perturbative propagator , is simply @xmath70=(p - q)^2 $ ] , for a massive bare gluon it will have the form @xmath70=(p - q)^2+m_g^2 $ ] , and will be more complex expression according to the propagators we discussed in the previous section .
the numerical code we used to solve the above equations is the same of ref .
@xcite , and for each one of the propagators in section ii we substitute : ( a ) @xmath71 , ( b ) @xmath72 , ( c ) @xmath73 and ( d ) @xmath74 . after obtaining for each case the function @xmath75 and @xmath76
, we can compute the quark condensate , which is expressed as @xmath77 where @xmath78 and @xmath79 , and we will also calculate the pion decay constant , which in terms of the functions @xmath80 and @xmath81 of eq.([e20 ] ) and eq.([e21 ] ) is given by @xmath82 .
\label{e23}\end{aligned}\ ] ] for all the propagators discussed here , with the respective gluon mass scales given in table [ table1 ] , we have not obtained dynamical quark mass generation !
our results for @xmath83 and @xmath84 are identically zero ( at least ten orders of magnitude below the scale @xmath85 )
. the gluon mass scales of table [ table1 ] are too large , in the sense that they cause a too strong screening of the force necessary to generate the symmetry breaking . to illustrate this incompatibility we present in fig.(3 ) the curve of the dynamically generated quark mass @xmath86 against the gluon mass , in the case of the cornwall propagator ( eq.([e6 ] ) ) , which gives the largest signal of mass generation that we obtained among all the propagators described above .
notice that in fig.(3 ) to obtain dynamical quark masses of @xmath87 , we would need a gluon mass almost half the value necessary to satisfy the constraints from diffractive scattering .
the inconsistency between diffractive scattering and the chiral symmetry breaking phenomenologies is not only a matter of adjustment of the infrared gluon mass scale , this scale really plays a different role in both cases . we need large gluon masses @xmath88 to soften the t - channel singularity in diffractive scattering , but smaller gluon masses if we do not want to erase the dynamical quark mass generation .
it is obvious that we have an inconsistency , although it is far from clear which is the solution .
nevertheless , this comparison of phenomenologies gives a powerful tool to constrain the detailed behavior of the infrared gluon propagator .
the ln pomeron model is able to explain a large amount of experimental data on diffractive scattering @xcite , making use of an infrared finite gluon propagator . as we have seen in section 2 , we need only a unique gluon mass scale to fit the pomeron - quark coupling and the pion - proton total cross section . on the other hand we have an extensive and successful phenomenology of dcsb , which was performed mostly with a `` perturbative '' @xmath89 gluon propagator . in section 3
we followed the standard procedures to compute the dynamical quark mass , quark condensate and pion decay constant , with the same non - perturbative propagators prescribed by the ln model .
both phenomenologies are inconsistent .
we need large gluon masses for diffractive scattering , and small ones to generate appropriate chiral symmetry breaking parameters ! there is a clear difference between our calculation of dcsb with the previous ones .
the use of `` massive '' propagators clearly screens the force necessary to generate the chiral symmetry breaking , and it is probably in this direction that we may look for troubles , another possibility would imply that the ln model is incorrect , or that the standard phenomelogy of dcsb has nothing to do with a finite correlation length for gluons .
let us discuss about each one of these possibilities .
the first common point to both phenomenologies is the existence of a mass scale for the gluon propagator , however , there is another claimed form for the gluon propagator in the infrared region , namely , @xmath90 ( see @xcite and references therein ) . in view of the theoretical arguments of ref .
@xcite about gluon mass scales , and the numerical simulations on the lattice @xcite , we were compelled to assume that the gluon field definitively has a finite correlation length in the vacuum , discarding the @xmath90 solution .
the gluon propagator being infrared finite , there is no comparison with the experimental data which would force us to abandon the ln model , and it remains to see if the dcsb phenomenology with such propagator is still consistent .
another difference to standard dcsb calculations is the use of the qcd running coupling constant given by eq.([e8 ] ) , which , as we said previously , is consistent with a study of the trilinear gluon vertex , and naturally embodies the freezing of the coupling at the relevant gluon mass scale @xcite .
both phenomenologies , diffractive scattering and dcsb , depend strongly on the behavior of this coupling in the infrared , however , its effect works basically in the same direction , i.e. increasing the coupling in the infrared we increase the diffractive scattering parameters as well as the dynamical quark mass .
therefore , there will not be much room to changes in this direction .
finally , if we assume that the ln pomeron model is correct , the gluon field has a finite correlation length in the vacuum , and the coupling constant has the freezing - out referred to above , we are forced to say that the dynamical quark mass is not generated by single - gluon forces , and the usual calculation of dynamical quark masses should be modified when the gluon mass is taken into account . actually , this view is not new , studying the coupled fermion gap and vertex equations for dcsb in qcd it was also found an incompatibility @xcite , and it was argued that the standard technique of using one - gluon exchange in the sde is not suitable when we consider the effect of massive gluons , and our work , in a different way , corroborates their result .
we finalize stressing that we are still far from knowing all the subtleties of qcd , and we have formalized the incompatibility between a qcd pomeron model and dcsb .
we believe that the confront of these different phenomenologies may provide strong constraints on the infrared behavior of the gluon propagator .
this research was partially supported by the conselho nacional de desenvolvimento cientifico e tecnologico ( cnpq)(aan ) , and fundao de amparo a pesquisa do estado de so paulo ( fapesp)(psrs ) .
we would like to thank jean - rene cudell and francis halzen for their comments on a previous manuscript , and also g. krein for discussions , and for helping us with the numerical code of ref.@xcite .
we are grateful for the kind hospitality at the institute for elementary particle physics research , university of wisconsin - madison ( aan ) , and at the international centre for theoretical physics ( ictp ) ( psrs ) , where part of this work was done . 99 f. e. low , phys .
* d12 * , ( 1975 ) 163 ; s. nussinov , phys .
lett . * 34 * , ( 1975 ) 1268 .
p. v. landshoff and o. nachtmann , z. phys .
* c35 * , ( 1987 ) 405 .
a. donnachie and p. v. landshoff , phys
b185 * , ( 1987 ) 403 ; nucl . phys . *
b311 * , ( 1988/89 ) 509 .
j. r. cudell , nucl
b336 * , ( 1991 ) 1 .
f. halzen , g. krein and a. a. natale , phys . rev . *
d47 * , ( 1993 ) 295 ; m. b. gay ducati , f. halzen and a. a. natale , phys .
* d48 * , ( 1993 ) 2324 . c. d. roberts and a.g .
williams , _ dyson - schwinger equations and their applications to hadron physics _ , to appear in progress in particle and nuclear physics , edited by a. faler ( pergamon press , oxford 1994 ) .
j. m. cornwall , phys . rev . * d26 * , ( 1982 ) 1453 ; in _ deeper pathways in high - energy physics _ , edited by b. kursunoglu ,
a. perlmutter and l. scott ( plenum , new york , 1977 ) , p. 683 ; nucl .
b157 * , ( 1979 ) 392 . c. bernard , c. parrinello and a. soni , phys
* d49 * , ( 1994 ) 1585 .
p. marenzoni et al .
, preprint hep - ph 9410355 ; p. marenzoni , g. martinelli , n. stella and m. testa , phys .
b318 * , ( 1993 ) 511 ; p. marenzoni , g. martinelli , and n. stella , nucl .
* b455 * , ( 1995 ) 339 .
j. f. gunion and d. e. soper , phys .
* d15 * , ( 1977 ) 2617 ; e. m. levin and m. g. ryskin , yad .
* 41 * , ( 1985 ) 1622 [ sov
. phys . * 41 * , ( 1985 ) 1027 ] ; j. dolejsi and j. hufner , z. phys .
* c54 * , ( 1992 ) 489 .
a. donnachie and p. v. landshoff , phys .
b296 * , ( 1992 ) 227 .
b. povh and j. hufner , phys .
lett . * 58 * , ( 1987 ) 1612 ; phys . lett . *
b245 * , ( 1990 ) 653 .
j. m. cornwall and j. papavassiliou , phys . rev .
* d40 * , ( 1989 ) 3474 . c. d. roberts and b. h. j. mckellar , phys.rev .
* d41 * , ( 1990 ) 672 .
j. m. cornwall and wei - shu hou , phys . rev . *
d34 * , ( 1986 ) 585 .
m. lavelle , phys .
* d44 * , ( 1991 ) r26 .
m. stingl , phys . rev . *
d34 * , ( 1986 ) 3863 ; u. habel et al . , z. phys .
* a336 * , ( 1990 ) 423 - 433 .
d. zwanziger , nucl . phys . * b378 * , ( 1992 ) 525 .
g. krein , p. tang and a. g. williams , phys .
* b215 * , ( 1988 ) 145 .
d. c. curtis , m. r. pennington and d. a. walsh , phys .
* b249 * , ( 1990 ) 528 .
a. g. williams , g. krein and c. d. roberts , ann .
( ny ) * 210 * , ( 1991 ) 464 .
k. buttner and m. r. pennington , phys .
* d52 * , ( 1995 ) 5220 ; phys . lett .
* b356 * , ( 1995 ) 354 .
j. papavassiliou and j. m. cornwall , phys . rev . *
d44 * , ( 1991 ) 1285 .
* table 1 * gluon masses ( in mev ) obtained fitting @xmath52 and @xmath91 for the several propagators discussed above .
we denote by @xmath92 the gluon propagator given by eq.([e6 ] ) with the dynamical gluon mass given by eq.([e10 ] ) . @xmath93 @xmath93 * fig . 1 * curves for @xmath51 as a function of the gluon mass .
we constrain the gluon mass by fixing @xmath94gev@xmath4 .
each gluon propagator is represented by : dashed line ( @xmath95 ) , dash - dotted line ( @xmath96 ) , dotted line ( @xmath97 ) , solid line ( @xmath98 ) . @xmath93 @xmath93 * fig . 2 * curves for pion - proton total cross section as a function of the gluon mass .
the constraint on the gluon mass comes from fixing the value @xmath99 mb .
the gluon propagators are represented by : dashed line ( @xmath95 ) , dash - dotted line ( @xmath96 ) , dotted line ( @xmath98 ) , solid line ( @xmath100 ) . @xmath93 @xmath93 * fig . 3 * dynamically generated quark mass for the cornwall s gluon propagator as a function of the gluon mass .
|
the phenomenology of a qcd - pomeron model based on the exchange of a pair of non - perturbative gluons , _
i.e. _ gluon fields with a finite correlation length in the vacuum , is studied in comparison with the phenomenology of qcd chiral symmetry breaking , based on non - perturbative solutions of schwinger - dyson equations for the quark propagator including these non - perturbative gluon effects .
we show that these models are incompatible , and point out some possibles origins of this problem .
= 10000 epsf * compatibility of a model for the qcd - pomeron and chiral - symmetry breaking phenomenologies * + a. a. natale and p. s. rodrigues da silva + instituto de fsica terica , universidade estadual paulista + rua pamplona , 145 , 01405 - 900 , so paulo , sp - brazil
|
ceres is the largest and most massive body in the asteroid main belt .
while many other large bodies of similar composition such as pallas , hygiea , euphrosyne have asteroid families , no such group has been so far identified for ceres @xcite . yet
collisional models suggest that about 10 craters larger than 10 km in diameter should have formed over 4.55 gyr of collisional evolution in the main belt @xcite . also
, observational data from the dawn probe show that at least two @xmath0280-km sized craters were formed in the last @xmath1 gyr on ceres surface , and larger impacts may have happened in the past @xcite .
the absence of a ceres family is therefore a major mystery in asteroid dynamics @xcite .
previous works have suggested the possibility that the outer shell of ceres could have been rich in ice , and that family members could have sublimated or been eroded by collisional evolution on timescales of hundreds of millions of years @xcite .
alternatively , it has also been hypothesized that km - sized fragments ejected at ceres escape velocity would likely disintegrate @xcite .
results from the dawn mission set however an upper limit of 40% in the ice content of ceres outer shell @xcite , which seems too low to explain the lack of a family . in this work we argue that the unique nature of ceres as a dwarf planet may have indeed produced families not easily detectable using methods calibrated for smaller , less massive bodies .
close encounters and linear secular resonances with ceres are expected to have significantly depleted the orbital region in the near proximity of this body , so reducing the number of the closest ceres neighbors .
more importantly , initial ejection velocities should have been significantly larger than those observed for any other parent body in the main belt , including vesta , spreading the collision fragments in a much larger area .
members of the ceres family would be significantly more distant among themselves than the typical distances between objects formed in collisions from smaller bodies , making an identification of the ceres family quite challenging . rather than looking for members of a ceres group in the central main belt ,
the region of the main belt with the highest concentration of asteroids , and where two other large ch families , dora and chloris , @xcite , would make the taxonomical identification of former members of a ceres family or paleo - family ( an old dispersed family with an age between 2.7 and 3.8 gyr , @xcite ) difficult , in this work we suggest to search for ch - type objects in the pristine region of the main belt , between the 5j:-2a and 7j:-3a mean - motion resonances ( or between 2.825 and 2.960 au in proper semi - major axis ) .
this region of the asteroid belt was depleted of asteroids during the late heavy bombardment ( lhb hereafter ) phase by sweeping mean - motion resonances , and the two mean - motion resonances with jupiter have since limited the influx of outside material from other areas of the asteroid belt @xcite
. the lower density of asteroids and the lack of other major c - type families at eccentricities and inclinations comparable to those of ceres makes the identification of possible members of the ceres family an easier task in this region .
identifying concentrations of c - type objects at values of inclinations comparable with that of ceres itself could therefore be a strong circumstantial evidence about the formation of a ceres family in the past .
ceres is the only dwarf planet in the asteroid belt .
while its mass is not large enough to clear its orbital neighborhood , significant dynamical perturbations are expected from this body , either as a consequence of close encounters @xcite or because of linear secular resonances involving the precession frequency of ceres node @xmath2 or pericenter @xmath3 , and those of a given asteroid @xcite . to investigate the importance of ceres as a perturber
, we obtain proper elements for 1400 particles in the plane of proper semi - major axis and inclination @xmath4 , with the approach described in @xcite , based on the method for obtaining synthetic proper elements of @xcite .
we integrated the particles over 20 myr under the gravitation influence of i ) all planets and ii ) all planets plus ceres as a massive body kg , as determined by the dawn spacecraft @xcite . ] with @xmath5 , the symplectic integrator based on @xmath6 from the _ swift _ package of @xcite , and modified by @xcite to include online filtering of osculating elements .
the initial osculating elements of the particles went from 2.696 to 2.832 au in @xmath7 and from @xmath8 to @xmath9 in @xmath10 .
we used 35 intervals in @xmath7 and 40 in @xmath10 .
the other orbital elements of the test particles were set equal to those of ceres at the modified julian date of 57200 .
[ fig : map_central ] displays our results , as two dynamical maps .
each blue dot is associated with a particle proper @xmath4 value . for the case without ceres as a massive body , displayed in panel a , the orbital region near ceres is relatively stable .
the only important non - linear secular resonance in the region is the @xmath11 ( or @xmath12 , in terms of linear secular resonances ) .
objects whose pericenter frequency is within @xmath13 arcsec / yr from @xmath14 arcsec / yr , likely resonators in the terminology of @xcite , are shown as red full dots in this figure .
this is a diffusive secular resonance , and asteroids captured into this resonance may drift away , but will not be destabilize @xcite .
the situation is much different and more interesting for the map with ceres as a massive body . as observed in fig .
[ fig : map_central ] , panel b , accounting for ceres causes the appearance of a 1:1 mean - motion resonance with this dwarf planet .
most importantly , linear secular resonances of nodes @xmath15 and pericenter @xmath16 , first detected by @xcite significantly destabilize the orbits in the proximity of ceres . combined with the long - term effect of close encounters with ceres , this has interesting consequences for the survival of members of the ceres family in the central main belt .
not many family members are expected to survive near ceres , and this would cause significant difficulties in using standard dynamical family identification techniques , since they are based on looking for pairs of neighbors in proper element domains ( hierarchical clustering method , hcm hereafter , @xcite ) .
since the close neighbors of ceres would have been removed on a short timescale , only objects whose distance from ceres is higher than the average distance between pairs of asteroids in the central main belt would have survived .
also , the area at higher inclination than that associated with the linear secular resonances with ceres should be significantly depleted of asteroids , and this is actually observed ( see for instance fig . 9 in @xcite ) .
finally , the dora family , a large ( @xmath17 members , @xcite ) group of the same spectral type of ceres , ch , is located at lower inclinations than that of ceres . considering that the large ch chloris family is also encountered in the central main belt near ceres , discriminating between a ch asteroid from ceres and one from dora , chloris , or other local sources would be a very daunting task , also considering that there are currently no identified members of a ceres family that could serve as a basis for a taxonomical analysis . to further investigate the efficiency of ceres in perturbing its most immediate neighbors
, we also performed numerical simulations of a fictitious ceres family with the @xmath18 integrator ( swift@xmath19yarkovsky@xmath19stochastic yorp@xmath19close encounters ) of @xcite , modified to also account for past changes in the values of the solar luminosity .
this integrators accounts for both the diurnal and seasonal version of the yarkovsky effect , the stochastic version of the yorp effect , and close encounters of massless particles with massive bodies .
the numerical set - up of our simulations was similar to what discussed in @xcite : we used the optimal values of the yarkovsky parameters discussed in @xcite for c - type asteroids , the initial spin obliquity was random , and normal reorientation timescales due to possible collisions as described in @xcite were considered for all runs .
we integrated our test particles under the influence of all planets ( case a ) , and all planets plus ceres ( case b ) , and obtained synthetic proper elements with the approach described in @xcite .
the fictitious ceres family was generated with the approach described in @xcite : we assumed that the initial ejection velocity field follows an isotropical gaussian distribution of zero mean and standard deviation given by : @xmath20 where @xmath21 is the body diameter in km , and @xmath22 is a parameter describing the width of the velocity field .
we used the relatively small value of @xmath23 m / s for our simulated family . in the next section
we will discuss why it is quite likely that any possible ceres family would have a much larger value of @xmath22 .
since here , however , we were interested in the dynamics in the proximity of ceres itself , our choice of @xmath22 was , in our opinion , justified .
we generated 642 particles with size - frequency distributions ( sfd ) with an exponent @xmath24 that best - fitted the cumulative distribution equal to 3.6 , a fairly typical value @xcite , and with diameters in the range from 2.0 to 12.0 km .
[ fig : ceres_scatter ] displays our results in the proper @xmath4 plane for the simulation without ceres as a massive body ( panel a ) and with a massive ceres ( panel b ) .
black full dots display the position of simulated ceres family members at the beginning of the simulation , after 1.2 my , while the red full dots show the orbital location at the end of the integration , for @xmath25 my . for the case without ceres one can notice that the family does not disperse much , and that most of the scattering is caused by the 3j:-1s:1a three - body resonance .
results including ceres as a perturber are quite more interesting : already after just 1.2 my secular resonances with ceres cleared a gap near the orbital location of the dwarf planet , identified by a full blue dot .
the situation is even more dramatic at the end of the simulation , where we observe that ceres scattered material at lower and higher inclinations with respect to the case in which ceres had no mass , and opened a significant gap at the family center , as also observed in the dynamical map of fig .
[ fig : map_central ] , panel b. as a function of time of the simulated family members for the case without ( blue line ) and with ( red line ) ceres as a massive body .
the horizontal black line displays the maximum nominal distance cutoff level @xmath26 for the whole simulated ceres family.,scaledwidth=45.0% ] how difficult would be to identify our simulated ceres family using hcm ? to answer this question , we computed the nominal distance velocity cutoff @xmath26 for which two nearby asteroids are considered to be related using the approach of @xcite , that defines this quantity as the average minimum distance between all possible asteroid pairs , as a function of time , for i ) all paired asteroids in the simulated ceres family ( @xmath26 ) , and ii ) just with respect to ceres itself ( @xmath27 ) .
this latter quantity provides an estimate of the minimum value of distance cutoff needed to identify a ceres family using the standard hcm .
the maximum value of the nominal distance velocity cutoff for the simulated ceres family was equal to 17.7 m / s . when ceres was not considered as a massive body , values of @xmath27 were of the order of 5 m / s , at all times below the nominal distance velocity cutoff ( see fig .
[ fig : dist_ceres ] , blue line ) .
once ceres was included as a massive body , however , values of @xmath27 quickly surpassed the nominal distance velocity cutoff @xmath26 ( see red line in fig .
[ fig : dist_ceres ] ) : after 200 my , except of short episodes in which @xmath28 , for most of the time @xmath27 was larger than @xmath26 .
after 380 my , this trend was consolidated and @xmath27 was always larger than @xmath26 .
the identification of a ceres family using hcm in the domain of proper elements of the simulated family and ceres as the first body would be quite challenging after 380 my .
so , where to look for hypothetical surviving members of a ceres family ?
the next section will be dealing with estimating the extent of the initial orbital dispersion of a putative family , and in assessing the regions where locating its members could be easier .
the escape velocity @xmath29 from ceres is considerably larger than that of other , less massive bodies : @xmath30 m / s @xcite versus the 360 m / s for vesta @xcite , the largest body in the main belt with a uncontroversial recognized asteroid family .
the vesta family is one of the largest in the main belt , and one would expect a putative ceres family to be even larger .
recently , @xcite investigated the shape of the ejection velocity field of 49 asteroid families .
typical observed values of @xmath31 , with @xmath22 given by eq .
[ eq : std_vej ] , are between 0.5 and 1.5 ( see also @xcite , sect . 5 ) and
argument of pericenter @xmath32 , and on the impact geometry , family members may have not reached the pristine region . for simplicity ,
in this work we concentrated our analysis on large basins forming events .
however , since collisional models predict that about 10 400 km or larger craters could have formed on ceres @xcite , we expect that , at least for a fraction of these events , some of the ejected material was likely to have reached the pristine region . ] . while for typical asteroid families this implies that @xmath22 is generally lower than 100 m / s , even a conservative choice of @xmath33 for ceres would produce a field with @xmath22 of the order of 240 m / s . this is more than a factor of 2 larger than the higher values of @xmath22 observed for some of the oldest asteroid families @xcite , but not inconsistent with the observed presence of crater ray systems around large craters on ceres , such as occator @xcite , that cover a large portion of the ceres surface . since the circular velocity needed to form such rays
is of the order of half the escape velocity from ceres , this suggests that values of @xmath34 near 0.5 for the @xmath22 parameter of a possible ceres family are not unreasonable .
following the approach of @xcite , see also @xcite , we generated fictitious ceres asteroids members with diameters 1 , 3 , and 7 km with the minimum and maximum value of @xmath34 from @xcite .
[ fig : ceres_ej ] displays our results for @xmath35 ( panel a ) and @xmath36 ( panel b ) . even in the most conservative case of @xmath35 ,
a value lower than those of any of the families studied in @xcite , we still found that a significative fraction and @xmath37 km populations , respectively .
] of the multi - km ( and potentially observable ) fragments from ceres could have been injected in the pristine region of the main belt .
we define this region as the area between the 5j:-2a and 7j:-3a mean - motion resonances with jupiter . because of these two dynamical barriers , very little material from external regions can reach this area @xcite . as a consequence ,
the local density of asteroids is much inferior to that of other regions of the asteroid belt .
also , only small c - type families , such those of naema ( 301 members ) , terpsichore ( 138 members ) , and terentia ( 79 members ) @xcite are found in the region , and at eccentricity values quite different than that of ceres , so easily distinguishable from putative ceres fragments .
if any large ( @xmath38 km ) , not easily movable by non - gravitation effects , members of a hypothetical ceres family would have been injected in this region after the lhb , it would be reasonable to expect that they could have remained there and still be visible today .
we selected a region in the pristine zone where most of the @xmath39 km objects could be found , according to our results for the most conservative case with @xmath35 ( black box in fig .
[ fig : ceres_ej ] , panels a , this also corresponds to the region in which most of the @xmath40 km objects would be located in the most optimistic scenario of @xmath36 , see black box in fig .
[ fig : ceres_ej ] , panels b. in the next section we will investigate if any such object could be identified in the pristine region .
results of a dynamical map with ceres as a massive body , not shown for the sake of brevity , demonstrate that the pristine region at inclination close to that of ceres is dynamically stable , with no linear secular resonances with ceres and only minor diffusive secular resonances with the giant planets . in the black region of fig .
[ fig : ceres_ej ] there are four ch asteroids , according to data from the three major photometric / spectroscopic surveys ( ecas ( eight - color asteroid analysis , @xcite ) , smass ( small main belt spectroscopic survey , @xcite ) , and s3os2 ( small solar system objects spectroscopic survey , @xcite ) : 195 eurykleia , 238 hypatia , 910 annelise , and 1189 terentia . the wise survey @xcite estimates that these objects have diameters of 85.7 , 148.5 .
47.1 , and 55.9 km , respectively . with the possible and very doubtful exception of annelise and terentia ,
the other objects are too big to be realistically assumed to be members of a ceres family . apart from a small family of 79 members around terentia @xcite , no major family has been identified around the largest objects , that appear to be isolated .
to increase statistics , we turn our attention to photometric data from the sloan digital sky survey - moving object catalog data , fourth release ( sdss - moc4 hereafter , @xcite ) . using the classification method of @xcite in the @xmath41 slope and @xmath42 colors domain , we identified 23 c - type photometric candidates in the pristine region .
[ fig : ceres_sdss_wise ] , panel a shows their orbital location in the @xmath4 plane .
vertical red lines display the orbital location of the main mean - motion resonances in the region , including the 2j:1s:-1a three - body resonance that divides the pristine area in two parts .
14 of the 23 sdss c - type asteroids , 60.9% of the total , can be found in the black region where we would expect ceres family members to be located .
by contrast , only 6 objects can be found at lower inclinations , and only 3 at higher values .
overall , c - type photometric candidates in the pristine region tend to cluster at inclination values close to that of ceres .
to check the statistical significance of this result , we checked what would be the probability that these 14 asteroids were produced by a fluctuation of a poisson distribution ( it has been suggested that the orbital distribution of background asteroids follows a poisson distribution @xcite ) . following the approach of @xcite , the probability that the expected number of @xmath43 occurrences in a given interval is produced by the poissonian statistics
is given by @xmath44 where @xmath45 is a positive real number , equal to the expected number of occurrences in the given interval .
here we used for @xmath45 the number of sdss candidates in the black region ( @xmath46 ) , weighted to account for the volume occupied by the region ( @xmath47 ) with respect to the total volume considered in our analysis ( @xmath48 , we considered asteroids with @xmath49 ) .
this is given by : @xmath50 for our considered black region , this yield @xmath51 , that corresponds to a probability to observe 14 bodies of just 0.13% , quite less than the null hypothesis level of 1.0% of the data being drawn from the poisson distribution .
similar analysis performed for uniform and gaussian distributions ( see @xcite for details on the procedures ) also failed to fulfill the null hypothesis , which suggests that the observed concentration of asteroids at inclinations close to those of ceres should be statistically significant . to increase the number of possible ceres members , we also look for objects in the wise database whose geometric albedo @xmath52 is between 0.08 and 0.10 , the range of albedo values observed for more than 90% of the surface material at ceres by the dawn spacecraft @xcite . while c - type objects with these values of wise albedo could be associated with a ceres family , other spectral types such as x , d , l , k , and even some s also have albedo values in this range , so this data must be considered with some caution .
nevertheless , it provides useful preliminary information . after eliminating objects belonging to spectral types other than c ( 13 bodies ) , members of the terpsichore and terentia families ( 7 asteroids ) , and of the x - type group charis ( 5 objects ) ,
we were left with a sample of 130 asteroids in that range of albedos .
the fact that only 4 c - type candidates were members of the terpsichore family , and 3 of terentia , 3.2% of the total , seems to indicate that local sources may not explain all the local population of c - type candidates .
since the other c - type objects in the region have diameters less than that of terpsichore , even if they produced families in the past ( of which there is no trace today ) , we would expect that to account at most for 10% of the observed population of c - type candidates .
[ fig : ceres_sdss_wise ] , panel b , displays the @xmath4 distribution of these objects .
again , 67 asteroids , 51.5% of the total , are found in the black region , while 39 , ( 30.0% ) are found at lower inclinations and 24 ( 18.5% ) at higher @xmath10 .
22 of the objects at low inclinations cluster near the x - type family charis , and could quite possibly be associated with its halo .
if we eliminate these objects , then the percentage of albedo candidates in the black region increase to 56.0% , a result similar to that from the sdss - moc4 data . to reduce possible contaminations from the local large families of charis , eos and koronis , which could have small populations of objects with wise albedos in the studied range , we eliminated objects i ) with @xmath7 beyond the 2j:1s:-1a mean - motion resonance that could potentially be associated with the charis and eos halo @xcite , and are less likely to have been produced by collisions on ceres with low values of @xmath34 , and ii ) those objects between the 5j:-2a and the 2j:1s:-1a mean - motion resonances with proper eccentricities lower than 0.08 and sine of proper inclinations lower than 0.06 , that could potentially be members of the koronis halo @xcite .
[ fig : ceres_histpv ] shows a histogram of all the 133 c - type albedo candidates in the pristine region ( panel a ) , and of the 45 asteroids that satisfied our criteria ( panel b ) . in the first panel
one can notice a peak at @xmath53 associated with the local charis x - type family , that is not visible once objects in the charis halo are removed ( panel b ) .
remarkably , there is a peak in the @xmath54 distribution that nearly coincides with the current value of the inclination of ceres . to check for its statistical significance , and in particular
if it could be the product of fluctuations in a poisson probability distribution , we used the approach described in @xcite , sect .
4.1 . for the @xmath55 values of the number of asteroids per size
bin in the histogram shown in fig .
[ fig : ceres_histpv ] , panel b , we best - fitted the value of @xmath45 in eq .
[ eq : poisson ] using the @xmath56 routine ( where @xmath56 stands for matrix laboratory ) _
poissfit_. we then generated 1000 random populations of asteroids following the poisson distribution with the routine _ poissrnd _ , and then performed a pearson @xmath57 test by computing the @xmath57-like variable defined as : @xmath58 where @xmath59 is the number of intervals used in the histogram in fig .
[ fig : ceres_histpv ] , panel b with a number of objects larger than 0 ( 9 ) , @xmath60 is the number of real objects in the @xmath61 interval , and @xmath62 is the number of simulated asteroids in the same @xmath54 range . assuming that the @xmath57-like variable follows an incomplete gamma function probability distribution , the probability of the two distributions being compatible can be computed considering the minimum value of @xmath57 ( 17.46 , in our case ) and the number of degrees of freedom ( 9 , @xcite ) .
this yields a probability of 0.8% of the two distributions being compatible , below the null hypothesis , which implies that the observed peak should be statistically significant .
similar results were obtained for the uniform and gaussian distributions . while not a conclusive proof of the possible existence of a ceres family , this interesting result represents , in our opinion , a circumstantial evidence in favor of this possibility .
a list of the photometric and albedo ceres members candidates with diameters less than 20 km in the pristine region identified in this work is provided in table [ table : ceres_members ] .
our results could be summarized as follows : * we studied the effect that the local dynamics may have had on a possible ceres family in the central main belt .
resonances with ceres would have depleted the population of members at higher inclinations than ceres itself , and close encounters with the dwarf planet would also caused further spread of the family , making the identification of a ceres dynamical group impossible after timescales of 200 - 400 my .
members of the ceres family at low inclinations would not be easily distinguishable from members of other local large ch families such as dora and chloris .
* we investigated the original spread that a ceres family may have had in the @xmath63 proper element domain . assuming that values of the initial ejection velocity parameter @xmath64 , would be in the range observed for other families
, we expect that a potentially observable population of @xmath65 km fragments , likely the results of the formation of craters on ceres larger than 300 - 400 km , should have been injected into the pristine region of the main belt .
collisional models predict that about 10 such craters could have formed on ceres @xcite . while the largest well - defined basins observed by dawn on ceres surface are smaller than 300 km , there is topographical evidence that larger impact structures formed in the ancient past @xcite .
the pristine region is characterized by a lower number density of asteroids , when compared to the central main belt .
we argue that ceres family members would be easier to recognize in this region , rather than near ceres itself . *
we analyzed which objects in the pristine region have taxonomical , photometric , or albedo data comparable to that of ceres .
remarkably , c - type candidates in this region cluster at inclinations that would be compatible with the range reached by possible ceres fragments . among albedo c - type candidates , we observe a statistically significant peak in the @xmath54 distribution at values close to that of ceres itself .
the number of c - type candidates at lower and higher inclinations is significantly lower , and no significant local sources can produce the observed distribution of c - type candidates . while we believe that we have provided a circumstantial evidence in favor of a possible ceres family in this work , we are aware that we have not proven its existence beyond any reasonable doubt . since ceres surface is characterized by a distinct @xmath66 absorption band @xcite , observing this feature in some of the proposed ceres family candidates could be a conclusive proof that such family exist , or existed in the past .
should the existence of a ceres family be finally proved , future lines of research could investigate the possibility of some main belt comets @xcite being escaped members of this long - lost asteroid group .
we are grateful to the reviewer of this paper , dr .
bojan novakovi , for comments and suggestions that improved the quality of this work .
we would like to thank the so paulo state science foundation ( fapesp ) that supported this work via the grant 14/06762 - 2 , and the brazilian national research council ( cnpq , grant 305453/2011 - 4 ) .
dn and sm acknowledge support from the nasa ssw and sservi programs , respectively .
the first author was a visiting scientist at the southwest research institute in boulder , co , usa , when this article was written .
this publication makes use of data products from the wide - field infrared survey explorer ( wise ) and of neowise , which are a joint project of the university of california , los angeles , and the jet propulsion laboratory / california institute of technology , funded by the national aeronautics and space administration .
+ identification & semi - major axis [ au ] & eccentricity & sine of inclination & absolute magnitude & diameter [ km ] & type + _ continued from previous page _ + identification & semi - major axis [ au ] & eccentricity & sine of inclination & absolute magnitude & diameter [ km ] & type + 2082 & 2.921907 & 0.193664 & 0.032701 & 13.30 & 17.12 & sdss + 5994 @xmath67 & 2.851295 & 0.184111 & 0.230810 & 11.98 & 17.80 & sdss + 6671 & 2.894602 & 0.135151 & 0.259412 & 12.54 & 13.75 & sdss + 6721 & 2.928060 & 0.160947 & 0.290183 & 12.40 & 14.93 & wise + 8560 @xmath67 & 2.956013 & 0.055937 & 0.234792 & 12.30 & 18.95 & sdss + * 8706 * @xmath67 & 2.847802 & 0.081223 & 0.142328 & 13.30 & 9.59 & wise + 11771 @xmath67 & 2.916868 & 0.182122 & 0.123918 & 13.90 & 7.60 & wise + 12356 & 2.935904 & 0.069843 & 0.035674 & 13.10 & 6.01 & sdss + 12986 @xmath67 & 2.930536 & 0.071421 & 0.192957 & 13.90 & 7.77 & wise + * 14037 * @xmath67 & 2.846048 & 0.113655 & 0.173687 & 12.80 & 12.86 & wise + 14338 @xmath67 & 2.944557 & 0.079173 & 0.217839 & 12.40 & 8.99 & sdss + 14531 @xmath67 & 2.833541 & 0.136888 & 0.098603 & 13.80 & 8.12 & wise + 16885 @xmath67 & 2.890294 & 0.089698 & 0.218012 & 12.90 & 15.54 & sdss + 18198 @xmath67 & 2.928039 & 0.165372 & 0.216076 & 13.60 & 9.67 & sdss + * 18463 * & 2.856107 & 0.121585 & 0.027381 & 14.40 & 5.89 & wise + * 20094 * @xmath67 & 2.861264 & 0.126932 & 0.185285 & 14.40 & 6.16 & wise + 20095 & 2.867089 & 0.044780 & 0.036435 & 14.44 & 5.73 & sdss + 20965 @xmath67 & 2.929609 & 0.106764 & 0.250502 & 13.20 & 10.62 & wise + 21714 @xmath67 & 2.896649 & 0.056614 & 0.220559 & 12.50 & 8.27 & sdss + 22540 & 2.874836 & 0.064495 & 0.036984 & 14.00 & 3.60 & sdss + 23000 @xmath67 & 2.931225 & 0.042298 & 0.221314 & 13.90 & 14.06 & sdss + 23136 & 2.927916 & 0.115123 & 0.259867 & 13.50 & 8.05 & sdss + 23432 & 2.910407 & 0.099806 & 0.060429 & 15.00 & 4.32 & wise + 26085 & 2.847349 & 0.071938 & 0.052672 & 14.20 & 6.12 & wise + 28297 @xmath67 & 2.933505 & 0.149849 & 0.150057 & 13.60 & 9.36 & sdss + 29514 & 2.936161 & 0.193732 & 0.290663 & 13.20 & 10.40 & wise + * 29906 * @xmath67 & 2.884096 & 0.043687 & 0.210437 & 14.00 & 6.84 & wise + 31232 @xmath67 & 2.857460 & 0.055909 & 0.194055 & 14.40 & 7.27 & sdss + 32421 @xmath67 & 2.915570 & 0.131466 & 0.183926 & 13.70 & 8.45 & wise + * 32586 * @xmath67 & 2.905601 & 0.142687 & 0.168171 & 12.80 & 11.24 & wise + 33202 & 2.931046 & 0.058748 & 0.009448 & 14.46 & 5.68 & sdss + 34129 & 2.868182 & 0.042556 & 0.036113 & 15.00 & 4.11 & sdss + 35533 @xmath67 & 2.902357 & 0.098794 & 0.100204 & 14.50 & 5.91 & wise + 38466 & 2.860576 & 0.046388 & 0.037292 & 15.10 & 4.09 & wise + 42544 & 2.914965 & 0.087749 & 0.092480 & 14.70 & 4.85 & wise + 42710 @xmath67 & 2.939916 & 0.070241 & 0.191939 & 13.90 & 7.49 & wise + * 43235 * @xmath67 & 2.835135 & 0.069649 & 0.229597 & 14.00 & 6.72 & wise + 44357 & 2.938663 & 0.176620 & 0.068078 & 14.50 & 5.65 & wise + 46645 & 2.944500 & 0.159099 & 0.091087 & 14.00 & 6.97 & wise + * 47651 * @xmath67 & 2.836642 & 0.070337 & 0.213856 & 14.60 & 5.36 & wise + * 47862 * @xmath67 & 2.887313 & 0.087727 & 0.202313 & 14.30 & 6.29 & wise + * 51050 * @xmath67 & 2.876180 & 0.169556 & 0.197835 & 14.00 & 6.87 & wise + 51338 @xmath67 & 2.947534 & 0.178943 & 0.161967 & 13.10 & 11.14 & wise + 51707 @xmath67 & 2.893943 & 0.194367 & 0.117033 & 14.50 & 5.82 & wise + * 53875 * @xmath67 & 2.895506 & 0.046369 & 0.211182 & 13.70 & 8.32 & wise + 54065 @xmath67 & 2.948372 & 0.120630 & 0.187491 & 15.00 & 4.25 & wise + 54930 @xmath67 & 2.915907 & 0.102511 & 0.199583 & 14.90 & 4.42 & wise + 54966 @xmath67 & 2.928465 & 0.101667 & 0.098231 & 14.50 & 5.30 & wise + 55903 & 2.941553 & 0.124014 & 0.300482 & 14.50 & 5.68 & wise + 57676 & 2.840613 & 0.081003 & 0.094847 & 15.40 & 3.51 & wise + 57695 & 2.953239 & 0.067685 & 0.026310 & 15.20 & 4.28 & wise + * 61674 * @xmath67 & 2.856602 & 0.145630 & 0.134328 & 13.70 & 8.21 & wise + 65774 & 2.876079 & 0.084274 & 0.023299 & 15.40 & 3.89 & wise + 66046 & 2.923553 & 0.054957 & 0.284802 & 13.80 & 7.83 & wise + * 66648 * @xmath67 & 2.857988 & 0.126132 & 0.189854 & 15.50 & 3.57 & wise + 67333 @xmath67 & 2.884682 & 0.186828 & 0.115222 & 14.00 & 7.41 & wise + * 71263 * @xmath67 & 2.851656 & 0.166470 & 0.153571 & 13.90 & 7.29 & wise + 72691 @xmath67 & 2.954464 & 0.092571 & 0.199377 & 14.40 & 5.88 & wise + 74398 & 2.926309 & 0.113313 & 0.313344 & 13.20 & 9.78 & wise + * 76814 * & 2.854806 & 0.079048 & 0.288523 & 13.90 & 7.56 & wise + 77346 & 2.925627 & 0.114702 & 0.090742 & 15.30 & 3.93 & wise + 77873 @xmath67 & 2.943052 & 0.183550 & 0.103473 & 14.10 & 6.63 & wise + * 82715 * @xmath67 & 2.831806 & 0.085409 & 0.194841 & 13.90 & 6.99 & wise + * 82897 * @xmath67 & 2.878531 & 0.056788 & 0.210677 & 14.50 & 5.69 & wise + 83337 @xmath67 & 2.831666 & 0.098578 & 0.099738 & 14.30 & 6.24 & wise + 83707 & 2.830936 & 0.085110 & 0.092492 & 14.00 & 6.80 & wise + * 85114 * & 2.835897 & 0.132118 & 0.277561 & 15.10 & 4.20 & wise + 88573 @xmath67 & 2.919644 & 0.132783 & 0.162074 & 14.40 & 5.92 & wise + 88723 & 2.834080 & 0.088113 & 0.088653 & 14.90 & 4.69 & wise + 89727 & 2.945443 & 0.101456 & 0.017127 & 15.40 & 3.75 & wise + *
91870 * @xmath67 & 2.845707 & 0.116806 & 0.164315 & 14.70 & 5.11 & wise + 91997 @xmath67 & 2.921480 & 0.078832 & 0.113695 & 14.80 & 5.09 & wise + * 97020 * & 2.901140 & 0.058409 & 0.274274 & 14.10 & 7.04 & wise + 103529 & 2.920771 & 0.052655 & 0.088095 & 15.50 & 3.36 & wise + 103923 @xmath67&2.933675 & 0.076817 & 0.172426 & 15.40 & 3.61 & wise + 104028 @xmath67&2.923278 & 0.054426 & 0.189286 & 14.60 & 5.39 & wise + * 105129 * @xmath67&2.862443 & 0.199475 & 0.134897 & 15.00 & 4.70 & wise + 108386 @xmath67&2.956425 & 0.080106 & 0.173829 & 15.20 & 4.10 & wise + 111331 & 2.917046 & 0.091109 & 0.083911 & 15.20 & 4.21 & wise + 111394 @xmath67&2.946302 & 0.090717 & 0.124884 & 14.50 & 5.73 & wise + 112976 @xmath67&2.878561 & 0.086200 & 0.121019 & 14.90 & 4.56 & wise + 113564 & 2.851523 & 0.085304 & 0.087892 & 14.80 & 4.78 & wise + 119722 & 2.852195 & 0.072637 & 0.046523 & 15.30 & 3.92 & wise + 120043 @xmath67&2.927114 & 0.071038 & 0.193386 & 14.60 & 5.40 & wise + 121281 & 2.844024 & 0.150695 & 0.095594 & 15.00 & 4.65 & wise + 125857 & 2.896361 & 0.102890 & 0.090084 & 15.70 & 3.28 & wise + * 128714 * & 2.868729 & 0.041501 & 0.290199 & 14.40 & 5.81 & wise + * 129048 * @xmath67&2.848874 & 0.132366 & 0.220818 & 14.90 & 4.71 & wise + * 130023 * @xmath67&2.880666 & 0.139291 & 0.156271 & 14.80 & 5.14 & wise + * 131267 * @xmath67&2.848184 & 0.130562 & 0.184916 & 14.90 & 4.57 & wise + 133841 & 2.932969 & 0.065715 & 0.086641 & 14.90 & 4.47 & wise + 139184 @xmath67&2.915685 & 0.191200 & 0.137376 & 15.60 & 3.20 & wise + 139793 & 2.953216 & 0.076844 & 0.095644 & 15.30 & 3.72 & wise + 140007 & 2.841540 & 0.052047 & 0.048892 & 14.60 & 5.45 & wise + * 140769 * @xmath67&2.860013 & 0.068603 & 0.209332 & 15.20 & 4.21 & wise + * 143166 * & 2.869961 & 0.101913 & 0.049494 & 15.30 & 3.81 & wise + 144712 @xmath67&2.915620 & 0.171867 & 0.019665 & 15.90 & 2.85 & wise + * 144715 * @xmath67&2.872016 & 0.110133 & 0.171309 & 15.40 & 3.73 & wise + * 145325 * @xmath67&2.849511 & 0.122737 & 0.125786 & 15.20 & 4.14 & wise + * 145392 * & 2.902675 & 0.059978 & 0.305121 & 14.80 & 5.07 & wise + 155547 & 2.935295 & 0.076924 & 0.086005 & 15.30 & 3.88 & wise + * 159263 * & 2.851540 & 0.121195 & 0.256403 & 15.40 & 3.66 & wise + 166760 & 2.945129 & 0.099172 & 0.082949 & 15.00 & 4.22 & wise + * 169161 * & 2.889445 & 0.116778 & 0.062533 & 15.90 & 3.04 & wise + * 172377 * @xmath67&2.902666 & 0.179680 & 0.154193 & 14.80 & 4.62 & wise + 174489 & 2.929807 & 0.132278 & 0.293032 & 14.90 & 4.48 &
wise + 176347 @xmath67&2.918230 & 0.086535 & 0.126204 & 14.80 & 4.77 & wise + 178669 & 2.956492 & 0.110930 & 0.083905 & 15.50 & 3.57 & wise + 181605 @xmath67&2.87433 & 0 0.158614&0.102414 & 15.30 & 4.01 & wise + 182687 & 2.910644 & 0.119178 & 0.095378 & 15.50 & 3.38 & wise + 184356 & 2.908767 & 0.107959 & 0.091759 & 15.50 & 3.43 & wise + 184487 & 2.945952 & 0.095575 & 0.087735 & 16.10 & 2.61 & wise + 189495 & 2.944911 & 0.133490 & 0.293474 & 14.30 & 6.32 & wise + 191486 & 2.947722 & 0.145902 & 0.256494 & 14.60 & 5.36 & wise + 192044 & 2.940902 & 0.074702 & 0.252544 & 14.70 & 5.04 & wise + 197144 & 2.925675 & 0.105006 & 0.088102 & 15.00 & 4.34 & wise + 198403 @xmath67&2.906537 & 0.088144 & 0.185480 & 15.50 & 3.37 & wise + 198808 @xmath67&2.924283 & 0.073838 & 0.238708 & 14.60 & 5.27 & wise + * 201824 * & 2.876024 & 0.082627 & 0.248968 & 14.90 & 4.54 & wise + 214826 & 2.949845 & 0.113408 & 0.084030 & 15.80 & 3.03 & wise + 218273 & 2.887695 & 0.078069 & 0.087647 & 15.50 & 3.70 & wise + 218776 @xmath67&2.924260 & 0.057623 & 0.134816 & 15.60 & 3.38 & wise + * 222080 * @xmath67&2.866610 & 0.149922 & 0.190036 & 16.10 & 2.68 & wise + 227004 @xmath67&2.913158 & 0.058757 & 0.135286 & 15.20 & 3.93 & wise + * 231791 * @xmath67&2.889728 & 0.157418 & 0.214961 & 14.90 & 4.84 & wise + 232724 & 2.903389 & 0.120023 & 0.091075 & 15.50 & 3.66 & wise + 235975 & 2.920486 & 0.101497 & 0.095416 & 15.30 & 3.77 & wise + 236004 @xmath67&2.943159 & 0.076640 & 0.183347 & 15.30 & 3.67 & wise + 238734 @xmath67&2.957799 & 0.056591 & 0.189721 & 15.60 & 3.26 & wise + * 239518 * & 2.859584 & 0.120313 & 0.258105 & 15.40 & 3.50 & wise + 239993 @xmath67&2.955953 & 0.053980 & 0.206900 & 15.20 & 3.98 & wise + * 240384 * & 2.834916 & 0.123127 & 0.261325 & 16.30 & 2.45 & wise + 243283 & 2.890796 & 0.122268 & 0.073358 & 16.40 & 2.45 & wise + 243302 @xmath67&2.938414 & 0.125304 & 0.208466 & 15.90 & 3.06 & wise + 243774 & 2.906543 & 0.130365 & 0.295275 & 14.90 & 4.43 & wise + 243854 @xmath67&2.947074 & 0.198642 & 0.126632 & 15.70 & 3.30 & wise + * 244955 * @xmath67&2.899970 & 0.057671 & 0.168345 & 15.50 & 3.50 & wise + * 245480 * & 2.858417 & 0.093258 & 0.262532 & 15.60 & 3.50 & wise + 246409 & 2.929076 & 0.145512 & 0.055515 & 15.90 & 2.84 & wise + 246466 & 2.830511 & 0.084131 & 0.085150 & 16.10 & 2.64 & wise + 246720 & 2.911652 & 0.076910 & 0.092850 & 16.10 & 2.82 & wise + * 247195 * & 2.847924 & 0.120740 & 0.261754 & 15.90 & 2.98 & wise + 248035 @xmath67&2.929195 & 0.139209 & 0.122659 & 15.80 & 3.00 & wise + 248581 & 2.920784 & 0.100786 & 0.076717 & 15.70 & 3.20 & wise + 249201 & 2.916190 & 0.077005 & 0.087743 & 15.70 & 3.08 & wise + * 249283 * @xmath67&2.894856 & 0.105884 & 0.214428 & 16.10 & 2.72 & wise + 250010 @xmath67&2.896574 & 0.126868 & 0.098989 & 16.20 & 2.58 & wise + 260891 @xmath67&2.945898 & 0.111076 & 0.172986 & 15.80 & 3.22 & wise + * 261489 * & 2.883625 & 0.146659 & 0.302045 & 15.60 & 3.21 & wise + * 268622 * @xmath67&2.858875 & 0.075933 & 0.127763 & 16.20 & 2.58 & wise +
|
ceres is the largest and most massive body in the asteroid main belt .
observational data from the dawn spacecraft reveal the presence of at least two impact craters about 280 km in diameter on the ceres surface , that could have expelled a significant number of fragments . yet
, standard techniques for identifying dynamical asteroid families have not detected any ceres family . in this work
, we argue that linear secular resonances with ceres deplete the population of objects near ceres .
also , because of the high escape velocity from ceres , family members are expected to be very dispersed , with a considerable fraction of km - sized fragments that should be able to reach the pristine region of the main belt , the area between the 5j:-2a and 7j:-3a mean - motion resonances , where the observed number of asteroids is low . rather than looking for possible ceres family members near ceres ,
here we propose to search in the pristine region .
we identified 156 asteroids whose taxonomy , colors , albedo could be compatible with being fragments from ceres . remarkably , most of these objects have inclinations near that of ceres itself . [ firstpage ] minor planets , asteroids : general minor planets , asteroids : individual : ceres celestial mechanics .
|
the question of dark energy and the accelerating universe has been the focus of a large amount of activities in recent years . this expansion has directly been measured from the light - curves of several hundred type ia supernovae @xcite and independently from observations of the cosmic microwave background ( cmb ) by the wmap satellite @xcite and other cmb experiments @xcite .
all these results strongly suggest that the universe is permeated smoothly by dark energy which has a negative pressure and violates the strong energy condition .
dark energy and the accelerating universe have been discussed extensively from various point of views over the past few years @xcite . in principle , a natural candidate for dark energy could be a small cosmological constant .
one approach in this direction is to employ what is known as modified gravity where an arbitrary function of the ricci scalar is added to the einstein - hilbert action .
it has been shown that such a modification may account for the late time acceleration and the initial inflationary period in the evolution of the universe @xcite .
one of the first proposals in this regard was suggested in @xcite where a term of the form @xmath1 was added to the usual einstein - hilbert action .
however , several models of @xmath0 gravity , proposed in @xcite , have been shown to suffer from unwanted instabilities @xcite or violate solar system constraints @xcite .
these instabilities arise because of an extra propagating scalar degree of freedom , and hence any viable @xmath0 model must be constructed in such a way as to render the dynamics of this scalar field carefully controllable .
this is in principle possible , and indeed @xmath0 theories can be made to satisfy solar system and laboratory tests by invoking the so - called chameleon mechanism @xcite .
recently , the authors in @xcite have tried to resolve the singularity problem arising in the strong gravity regime of the otherwise viable @xmath0 theories by imposing appropriate fine - tuning on their models . in @xmath0 gravity , einstein equations
posses extra terms induced from geometry which , when moved to the right hand side , may be interpreted as a matter source represented by the energy - momentum tensor @xmath2 , see equation ( [ eq03 ] ) . in a similar fashion ,
the space - time - matter ( stm ) theory , discussed below , results in einstein equations in @xmath3 with some extra geometrical terms which may be interpret as induced matter .
it therefore seems plausible to make a correspondence between the geometrical terms in stm and @xmath2 resulting in @xmath0 gravity .
we shall explore this idea to show that different choices of the parameter @xmath4 in stm may correspond to different choices of @xmath0 in curvature quintessence models in modified gravity .
the correspondence discussed above is based on the idea of extra dimensions .
the idea that our world may have more than four dimensions is due to kaluza @xcite , who unified einstein s theory of general relativity with maxwell s theory of electromagnetism in a @xmath5 manifold .
since then , higher dimensional or kaluza - klein theories of gravity have been studied extensively @xcite from different angles .
notable amongst them is the stm theory mentioned above , proposed by wesson and his collaborators , which is designed to explain the origin of matter in terms of the geometry of the bulk space in which our @xmath3 world is embedded , for reviews see @xcite .
more precisely , in stm theory , our world is a hypersurface embedded in a five - dimensional ricci - flat ( @xmath6 ) manifold where all the matter in our world can be thought of as being manifestations of the geometrical properties of the higher dimensional space .
the fact that such an embedding can be done is supported by campbell s theorem @xcite which states that any analytical solution of the einstein field equations in @xmath7 dimensions can be locally embedded in a ricci - flat manifold in @xmath8 dimensions .
since the matter is induced from the extra dimension , this theory is also called the induced matter theory .
applications of the idea of induced matter or induced geometry can also be found in other situations @xcite .
the stm theory allows for the metric components to be dependent on the extra dimension and does not require the extra dimension to be compact .
the sort of cosmologies stemming from stm theory is studied in @xcite .
in this paper we consider the correspondence between @xmath0 gravity and stm theory . in section 2
we present a short review of @xmath3 dark energy models in the framework of @xmath0 gravity . in section 3
the field equations are solved in stm theory by fixing a suitable metric and the resulting geometric terms are interpreted as dark energy .
the cosmological evolution in stm are considered in section 4 .
section 5 deals with an example for a special form of @xmath0 .
conclusions are drawn in the last section .
to begin with , let us start by a brief review of @xmath0 gravity theory .
we start by writing the four dimensional action as @xmath9 where @xmath0 is a function of the ricci scalar @xmath10 and @xmath11 is the action for the matter fields .
we use units such that @xmath12 .
the field equations are @xmath13 which can be recast into the more expressive form @xmath14 where an stress - energy tensor has been defined for the curvature contribution @xmath15 + f'(r)^{;\alpha\beta}(g_{\alpha\mu}g_{\beta\nu}-g_{\alpha\beta}g_{\mu\nu } ) \right\},\ ] ] and @xmath16 is the stress - energy tensor of the matter . to derive the field equations we consider the robertson - walker metric for the evolution of the cosmos @xmath17 where @xmath18 is the curvature of the space , namely , @xmath19 for the flat , closed and open universes respectively . substituting the above metric in equation ( [ eq03 ] )
we obtain the @xmath3 , spatially flat friedmann equations as follows @xmath20 and @xmath21,\ ] ] where a dot represents derivation with respect to time .
such a universe is dominated by a barotropic perfect fluid with the equation of state ( eos ) given by @xmath22 ( @xmath23 for pressureless cold dark matter and @xmath24 for radiation ) and a spatially homogenous curvature quintessence .
the energy density and pressure of the curvature quintessence are @xmath25 \right\},\ ] ] and @xmath26 -3\left(\frac{\dot{a}}{a}\right)\dot{r}f''(r ) \right\},\ ] ] respectively .
the equation of state of the curvature quintessence is @xmath27 recently , cosmological observations have indicate that our universe is undergoing an accelerated expanding phase .
this could be due to the vacuum energy or dark energy which dominates our universe against other forms of matter such as dark matter and baryonic matter .
we thus concentrate on the vacuum sector _ i.e. _ @xmath28 , from which the evolution equation of curvature quintessence becomes @xmath29 yielding @xmath30\nonumber\\ & \equiv&\rho_{0_{\rm curv } } e(z),\end{aligned}\ ] ] where , @xmath31 is the redshift and the subscript @xmath32 denotes the current value . in terms of the redshift , the first friedmann equation can be written as @xmath33 where @xmath34 and @xmath35 are the current values of the dimensionless density parameter and hubble parameter , respectively .
equation ( [ eq14 ] ) is the friedmann equation in terms of the redshift , @xmath36 , which is suitable for cosmological observations .
in fact , equations ( [ eq14 ] ) and ( [ eq26 ] ) , obtained in section 4 , are the cosmological connections between @xmath0 gravity and stm theory .
in the context of stm theory , a class of exact @xmath5 cosmological solutions has been investigated and discussed in @xcite .
this solution was further pursued in @xcite where it was shown to describe a cosmological model with a big bounce as opposed to the ubiquitous big bang .
the @xmath5 metric of this solution reads @xmath37 where @xmath38 and @xmath39 here @xmath40 and @xmath41 are two arbitrary functions of @xmath42 , @xmath18 is the @xmath43 curvature index @xmath44 , and @xmath45 is a constant .
this solution satisfies the @xmath5 vacuum equation @xmath6 .
the kretschmann curvature scalar @xmath46 shows that @xmath45 determines the curvature of the @xmath5 manifold .
such a solution was considered in @xcite with a different notation . using the @xmath3 part of the @xmath5 metric ( [ eq15 ] ) to calculate the @xmath3 einstein tensor
, we obtain @xmath47 as was mentioned earlier , since the recent observations show that the universe is currently going through an accelerated expanding phase , we assume that the induced matter contains only dark energy with @xmath48 , _
i.e. _ @xmath49 .
we then have @xmath50 @xmath51 from equations ( [ eq19 ] ) and ( [ eq20 ] ) , one obtains the eos of dark energy @xmath52 the hubble and deceleration parameters are given in @xcite and can be written as @xmath53 and @xmath54 from which we see that @xmath55 represents an accelerating universe while @xmath56 represents a decelerating one .
the function @xmath4 therefore plays a crucial role in defining the properties of the universe at late times .
in this section we will concentrate on the predictions of the cosmological evolution in the spatially flat case ( @xmath57 ) . to avoid having to specify the form of the function @xmath58
, we change the parameter @xmath42 to @xmath36 and use @xmath59 and define @xmath60 , noting that @xmath61 .
we then find that equations ( [ eq21])-([eq23 ] ) reduce to @xmath62 and @xmath63 there is an arbitrary function @xmath4 in the present @xmath5 model .
different choices of @xmath4 may correspond to different choices of @xmath0 in curvature quintessence models in modified gravity .
various choices of @xmath4 correspond to the choices of @xmath64 .
this enables us to look for the desired properties of the universe via equations ( [ eq24 ] ) and ( [ eq25 ] ) . using these definitions ,
the friedmann equation becomes @xmath65 this would allow us to use the supernovae observational data to constrain the parameters contained in the model or the function @xmath64 . by comparing equation ( [ eq26 ] ) with equation ( [ eq14 ] ) , we find that there exists a correspondence between the functions @xmath0 and @xmath64 .
we thus take @xmath64 as @xmath66^{-1}.\label{fz}\ ] ] according to ( [ eq13 ] ) , it is easy to see that the function @xmath67 is determined by the particular choice of @xmath0 which , in turn , determines the function @xmath64 through equation ( [ eq27 ] ) .
the evolution of the density components and eos of dark energy may now be derived . in the next section
we will consider the correspondence between modified gravity and the ricci - flat cosmology for a generic @xmath0 .
to continue , we must determine the functional form of @xmath0 . thus , for example , we choose @xmath0 as a generic power law of the scalar curvature and assume for the scale factor a power law solution in @xmath3 , investigated in the first reference in @xcite .
therefore @xmath68 the interesting cases are for the values of @xmath69 satisfying @xmath70 which would lead to an accelerated expansion of our universe .
let us now concentrate on the case @xmath49 .
inserting equation ( [ eq28 ] ) into the dynamical system ( [ eq07 ] ) and ( [ eq08 ] ) , for a spatially flat space - time we obtain an algebraic system for parameters @xmath71 and @xmath69 @xmath72=0 , \\ \\
\alpha\left[n^{2}-n+1+\alpha(n-2)\right]=n(n-1)(2n-1),\\ \end{array } \right.\ ] ] from which the allowed solutions are @xmath73 the solutions with @xmath74 are not interesting since they provide static cosmologies with a non - evolving scale factor .
note that this result matches the standard general relativity result , @xmath75 , in the absence of matter . on the other hand ,
the cases with generic @xmath69 and @xmath71 furnish an entire family of significant cosmological models .
using equations ( [ eq09 ] ) and ( [ eq10 ] ) we can also deduce the equation of state for the family of solutions @xmath76 as @xmath77 where @xmath78 as @xmath79 .
this shows that an _ infinite _
@xmath71 is compatible with recovering an _
infinite _ cosmological constant .
thus , using equation ( [ eq31 ] ) , the functions @xmath67 and @xmath64 are given by @xmath80},\end{aligned}\ ] ] @xmath81}\right]^{-1}.\end{aligned}\ ] ] now , using the above equations , we see that equations ( [ eq24 ] ) and ( [ eq25 ] ) can be written as @xmath82 and @xmath83 therefore , within the context of the present investigation , the accelerating , dark energy dominated universe can be obtained by using the correspondence between @xmath64 and @xmath0 in modified gravity theories .
we observe that in stm theory , @xmath5 dark energy cosmological models correspond to @xmath3 curvature quintessence models .
this result is consistent with the correspondence between exact solutions in kaluza - klein gravity and scalar tensor theory @xcite .
we also note that , as is well known , with a suitable conformal transformation , @xmath0 gravity reduces to scalar tensor theory . from equations ( [ eq32 ] ) and ( [ eq33 ] ) , we can rewrite equation ( [ eq26 ] ) as @xmath84},\ ] ] where @xmath85 and the contribution of ordinary matter has been neglected .
figure 1 shows the behavior of @xmath86 as a function of @xmath71 for @xmath87 and @xmath88 .
as can be seen , for @xmath89 and @xmath90 we have @xmath91 , that is , the universe finally approaches the curvature dominant state , thus undergoing an accelerated expanding phase .
figure 2 shows the behavior of @xmath92 as a function of @xmath36 for @xmath93 and @xmath94 .
we see that for small @xmath36 , @xmath95 .
thus , we have obtained late - time accelerating solutions only by using the correspondence between @xmath0 gravity and stm theory . here , we have interpreted the properties of @xmath5 ricci - flat cosmologies by dark energy models in modified gravity .
in this paper we have investigated the present accelerated expanding phase of the universe using a general class of @xmath5 ricci - flat cosmological models .
such an exact solution contains two arbitrary functions , @xmath4 and @xmath58 which are analogous to different forms of @xmath0 in curvature quintessence models .
also , once the forms of the arbitrary functions are specified , the characteristic parameters determining the evolution of our universe are specified .
we have noted that the correspondence between the functions @xmath64 and @xmath0 plays a crucial role and defines the form of the function @xmath64 . finally , by taking a specific form for @xmath0 we obtained solutions that are consistent with late - time acceleration of the universe .
this work has been financially supported by the research institute for astronomy and astrophysics of maragha .
9 a. g. riess _ et .
[ high - redshift supernova search team , ( hzt ) ] , astron .
j. 116 ( 1998 ) 1006 , [ astro - ph/9805201 ] .
s. perlmutter _ et .
j. 517 ( 1999 ) 565 , [ astro - ph/9812133 ] ; + d. n. spergel _ et .
j. suppl . 148
( 2003 ) 175 , [ astro - ph/0302209 ] . c. l. bennett _ et .
j. suppl . 148
( 2003 ) 1 , [ astro - ph/0302207 ] . c. b. netterfield_et .
j. 571 ( 2002 ) 604 , [ astro - ph/0104460 ] .
n. w. halverson _ et .
j. 568 ( 2002 ) 38 , [ astro - ph/0104489 ] .
i. zlatev , l. wang and p. j. steinhardt , phys .
82 ( 1999 ) 896 , [ astro - ph/9807002 ] ; + p. j. steinhardt , l. wang , i. zlatev , phys .
d 59 ( 1999 ) 123504 , [ astro - ph/9812313 ] ; + m. s. turner , int .
a 17s1 ( 2002 ) 180 , [ astro - ph/0202008 ] ; + v. sahni , class .
( 2002 ) 3435 , [ astro - ph/0202076 ] .
r. r. caldwell , m. kamionkowski , n. n. weinberg , phys .
91 ( 2003 ) 071301 , [ astro - ph/0302506 ] ; + r. r. caldwell , phys
b 545 ( 2002 ) 23 , [ astro - ph/9908168 ] ; + p. singh , m. sami , n. dadhich , phys . rev .
d 68 ( 2003 ) 023522 , [ hep - th/0305110 ] ; + j.g .
hao , x.z .
li , phys .
d 67 ( 2003 ) 107303 , [ gr - qc/0302100 ] .
armendriz - picn , t. damour , v. mukhanov , phy .
b 458 ( 1999 ) 209 ; + m. malquarti _ et . al .
d 67 ( 2003 ) 123503 ; + t. chiba , phys .
d 66 ( 2002 ) 063514 , [ astro - ph/0206298 ] .
s. capozziello , int . j. mod
d 11 ( 2002 ) 483 , [ astro - ph/0201033 ] ; + s. capozziello _ et .
al . _ , int .
d 12 ( 2003 ) 1969 , [ astro - ph/0307018 ] ; + s. capozziello _ et .
b 639 ( 2006 ) 135 ; + s. nojiri and s. d. odintsov , j. phys .
66 ( 2007 ) 012005 ; + s. nojiri and s. d. odintsov , j. phys . a 40 ( 2007 ) 6725 ;
+ f. briscese _ et .
b 646 ( 2007 ) 105 ; + s. nojiri , s.d .
odintsov and h. stefanic , phys .
d 74 , ( 2006 ) 086009 ; + m. e. soussa and r. p. woodard , gen .
36 ( 2004 ) 855 ; + r. dick , gen . rel . grav . 36 ( 2004 ) 217 ; + a. e. dominguez and d. e. barraco , phys . rev .
d 70 , ( 2004 ) 043505 ; + v. faraoni , phys . rev .
d 75 ( 2007 ) 067302 ; + j. c. c. de souza and v. faraoni , class .
( 2007 ) 3637 ; + d. a. easson , int .
a 19 ( 2004 ) 5343 ; + g. j. olmo , phys .
95 ( 2005 ) 261102 ; + g. allemandi _ et .
rel . grav .
37 ( 2005 ) 1891 ; + s. capozziello and a. troisi , phys .
d 72 ( 2005 ) 044022 ; + t. clifton and j.d .
barrow , phys .
d 72 ( 2005 ) 103005 ; + t. p. sotiriou , gen .
38 ( 2006 ) 1407 ; + s. capozziello , a. stabile and a. troisi , mod .
lett . a 21 ( 2006 ) 2291 ; + a. dolgov and d. n. pelliccia , nucl . phys .
b 734 ( 2006 ) 208 ; + k. atazadeh and h.r .
sepangi , int . j. mod
d 16 ( 2007 ) 687 , [ gr - qc/0602028 ] ; + s. nojiri and s. d. odintsov , int .
phys . 4 ( 2007 ) 115 ,
[ hep - th/ 0601213 ] .
s. m. carroll _ et .
phys . rev .
d 70 ( 2004 ) 043528 .
a. d. dolgov and m. kawasaki , phys .
b 573 ( 2003 ) 1 , [ astro - ph/0307285 ] . t. chiba , phys .
b 575 ( 2003 ) 1 , [ astro - ph/0307338 ] ; + a. l. erickcek , t. l. smith and m. kamionkowski , phys . rev .
d 74 ( 2006 ) 121501 , [ astro - ph/0610483 ] ; + t. chiba , t. l. smith and a. l. erickcek , phys . rev .
d 75 ( 2007 ) 124014 , [ astro - ph/0611867 ] .
j. khoury and a. weltman , phys .
93 ( 2004 ) 171104 , [ astro - ph/0309300 ] ; + j. khoury and a. weltman , phys .
d 69 ( 2004 ) 044026 , [ astro - ph/0309411 ] ; +
d. f. mota and j. d. barrow , mon . not .
( 2004 ) 291 , [ astro - ph/0309273 ] ; + t. clifton , d. f. mota and j. d. barrow , mon . not .
358 ( 2005 ) 601 , [ gr - qc/0406001 ] ; + d. f. mota and d. j. shaw , phys .
97 ( 2006 ) 151102 , [ hep - ph/0606204 ] ; + d. f. mota and d. j. shaw , phys
d 75 ( 2007 ) 063501 , [ arxiv : hep - ph/0608078 ] .
t. kobayashi and k. maeda , [ arxiv:0810.5664 ] .
t. kaluza , _ on the problem of unity in physics _ , sitzungsber . preuss
berlin math .
kl ( 1921 ) 966 . o. klein , _ quantum theory and five - dimensional relativity _ , z. phys . * 37 * 895 ( 1926 ) [ surveys high energy phys . 5 ( 1986 )
p. s. wesson , _ space - time - matter _
( singapore : world scientific ) 1999 ; + j. m. overduin and p. s. wesson , phys . rept . 283
( 1997 ) 303 , [ gr - qc/9805018 ] .
j. e. campbell , _ a course of differential geometry _ , ( clarendon oxford , 1926 ) ; + s. rippl , r. romero , r. tavakol , class .
12 ( 1995 ) 2411 ; + c. romero and r. tavakol and r. zalaletdinov , gen .
28 ( 1996 ) 365 ; + j. e. lidsey _ et .
( 1997 ) 865 ; + s. s. seahra and p. s. wesson , class .
( 2003 ) 1321 , [ gr - qc/0302015 ] .
v. frolov , m. snajdr and d. stojkovic , phys .
d 68 ( 2003 ) 044002
. h. y. liu and p. s. wesson , astrophys .
j. 562 ( 2001 ) 1 , [ gr - qc/0107093 ] .
t. liko and p. s. wesson , int
a 20 ( 2005 ) 2037 , [ gr - qc/0310067 ] ; + s. s. seahra and p. s. wesson , j. math . phys .
44 ( 2003 ) 5664 ; + j. ponce de leon , gen .
grav . 20 ( 1988 ) 539 ; + l. x. xu , h. y. liu , b. l. wang , chin . phys .
( 2003 ) 995 , [ gr - qc/0304049 ] ; + h. y. liu , phys .
b 560 ( 2003 ) 149 , [ hep - th/0206198 ] .
b. l. wang , h. y. liu and l.x .
a 19 ( 2004 ) 449 , [ gr - qc/0304093 ] ; + l. x. xu and h. y. liu , int .
d 14 ( 2005 ) 883 , [ astro - ph/0412241 ] .
h. y. liu and b. mashhoon , ann .
phys . 4 ( 1995 ) 565 .
|
we study the correspondence between two theoretical frameworks for describing dark energy , @xmath0 gravity and higher - dimensional space - time - matter ( stm ) or induced - matter theory .
we show that the hubble expansion parameter in @xmath0 gravity can be associated with a combination of metric functions in stm theory , and consider a specific example whose properties are consistent with late - time acceleration .
-2.5 cm
|
top quark physics is sensitive to new physics , which may exist near the electro weak scale , due to its large mass .
experimental and theoretical research about cp violation in top sector is one way to reveal new physics . to study the top quark cp
odd effects has its own advantage that the uncertainties coming from hardron matrix elements can be avoided .
if @xmath4 is near to @xmath5 , the cp assymmetry effects in top quark decays can be induced by new particles . until now
a lot of works about cp
asymmetry effects in the top quark decays@xcite have been done within the supersymmetric model .
most of those works have made the assumption that the mass of the light top squark is much smaller than that of the top quark . b.grzadkowski and w.y.keung calculated the cp violating effects induced by @xmath6-loop@xcite .
this contribution requires the condition @xmath7 , which has already been excluded .
e.christova and m.fabbrichesi computed the effects induced by @xmath8-loop@xcite which requires @xmath9 . s. bar shalom _
et al _ gave the cp asymmetry in top quark decay induced by @xmath10 .
it is at best 0.3% when light stop mass is between @xmath11 and @xmath12@xcite .
however , if the light top squark mass @xmath13 is approximately as heavy as the top quark , cp asymmetry effect in top quark two body decay induced by supersymmetric cp odd phase will not be observable because the top squark can not run on shell to produce the necessary absorptive cut . in this work we considered this case under the assumption that the light chargino is much lighter than @xmath14 . under this condition
, the @xmath3 ( which is always assumed to be the lsp ) and @xmath2 can provide necessary absorptive cut in top quark three body decays , such as in the process @xmath15 considered in the work .
these two particles can be on shell in the top quark three body decay loop diagrams when the invariant mass of the lepton pair is sufficiently large .
to our knowledge , a study on cp asymmetry in the top quark three body decays is missed in literature . in the present work the mass of the light top squark
is assumed to be above 140gev .
taking into account the direct experimental limit on super particles@xcite and the indirect limit coming from neutron edm limit@xcite , we take @xmath16 to be real and , the lightest neutralino to be above 30gev and the lighter chargino to be above 65gev .
the large mass of stop leads to relatively small cp violating effects . nevertheless , the cp
odd effects can reach up to @xmath17 in the most favorable parameter space .
the paper is organized as following : in sec .
2 we analyze the possible new cp violating sources in mssm and present our simplifying assumptions in calculation . in sec .
3 we sketch the main steps of our calculations . in sec .
4 we present our numerical results and in sec . 5 and sec .
6 we discuss and summarize our results .
the mass matrices for charginos , neutralinos and squarks are given in appendix a. in appendix b we give the relevant pieces of the lagrangian for our calculations and some analytic results are presented in appendix c.
the most general form of the low energy lagrangian of mssm@xcite , which is @xmath18 gauge invariant and does not violate the sm conservation laws , can be written as @xmath19 the superpotential @xmath20 is given by @xmath21 where @xmath22 = -1 .
the hat ` ^ ' indicates that the corresponding letter represents a superfield .
the capital indices i , j denote generations .
i , j refer to the components of a su(2 ) doublet .
the l , d , u are the yukawa coupling matrices .
the soft breaking terms can be divided into three pieces , @xmath23 these are the scalar particle mass terms , gaugino mass terms and the trilinear soft breaking terms respectively .
they are given by @xmath24 @xmath25 @xmath26 where @xmath27,@xmath28,@xmath29 are u(1 ) , su(2 ) and su(3 ) gauginos masses respectively .
fields in @xmath30 and @xmath31 are scalar components of the corresponding superfields
. in general , all the coupling parameters in the above expressions except those of the diagonal terms in @xmath30 may be complex which may be the cp violating sources
. however , not all of them are physical and , even the physical parameters are too many to be disposed of . in actual calculations , simplifying assumptions must be made .
we get the physical cp violating phases by the following steps .
first , we take the gut assumption that the @xmath32s are universal at the gut scale and can be set real by a phase rotation@xcite . thus , the @xmath33s are real at any scale .
second , we adjust the global phase between the two higgs superfields so that @xmath34 is real .
this adjustment makes the two vacuum expectation values of the neutral higgs fields @xmath35 , @xmath36 real@xcite . after the adjustment
the phases of the two higgs superfields are fixed and @xmath16 is complex in general .
third , @xmath37 , @xmath38 , and @xmath39 in the superpotential are diagonalized and the unphysical phases are absorbed by quark superfields similar to that done in the standard model .
this leaves a cp violating phases @xmath40 in the after the superfield are redefined .
fourth , to suppress the fcnc process in the susy extension of sm , as an approximation we require that all the matrices in @xmath41 , @xmath42 , @xmath43 , @xmath44 , @xmath45 , @xmath46 , @xmath47 , @xmath48 and @xmath49 are flavor diagonal in the basis where @xmath37 , @xmath38 and @xmath39 are diagonal ( flavor alignment@xcite ) . then the hermitian matrices in @xmath30 are now all real .
the phases of all squarks are fixed after the third step .
therefore , @xmath48 , @xmath47 are generally complex . in conclusion ,
@xmath40 , @xmath50 , @xmath51s and @xmath52s are the cp violating phases in the low energy supersymmtric model under our assumptions .
@xmath50 can not be larger than the order @xmath53 by the constraint from the experimental limit of the neutron edm@xcite . in our calculation
, we always take @xmath16 to be real and thus no cp violating effects are induced by @xmath16 .
the cp violating effects induced by @xmath54 are greatly suppressed because they are proportional to the masses of the corresponding quarks which can be neglected compared with the squark mass parameters in @xmath30 except that induced by @xmath55 which is associated with the top quark(see the form of squark mass matrices in ( a.8 ) and ( a.9 ) in appendix a ) .
thus , @xmath56 is the only new cp violating source in our calculation .
after the interaction terms in the potential are diagonalized , the mssm lagrangian will be expressed by mass eigenstates instead of gauge eigenstates .
the cp violating phases are then transferred to the gauge interaction vertices ( see appendix b for related lagrangian pieces ) .
this is reflected by mixing matrices in the interaction vertices .
the mixing matrices @xmath57,@xmath58,@xmath59 which diagonalize charginos and neutralinos are real if @xmath16 is taken real .
the mixing matrices @xmath60 for the top squark is in general complex due to the complexity of @xmath55 .
this implies that the cp violating effects comes from @xmath0 .
the mixing matrices will be discussed in detail in the appendix a.
we now discuss the cp violating effects in the process @xmath61 , which corresponds to diagrams in fig.1 , within the mssm framework .
first denote the invariant mass of @xmath62 and @xmath63 as @xmath64 , where @xmath65 .
@xmath66 and @xmath67 are the four momenta of @xmath68 and @xmath62 .
we calculated the cp asymmetry when @xmath69 .
this condition opens a new window so that @xmath2 and @xmath3 cut may give an absorptive part to the amplitudes for loop diagrams as depicted in fig.2 .
several points should be indicated at the moment .
there should be a minus sign in front of the amplitudes for box diagrams relative to that for triangle diagrams . and @xmath62 at two ends of the @xmath70 propagator and simultaneously changing @xmath70 to w boson in the wick contraction procedure .
the interchange of two fermion operators gives the minus sign .
an analogous case on qed can be found in @xcite . ] + ( 2 ) .
there are two kinds of box diagrams .
one of them , fig.2a and 2b , is fermion number nonconservative@xcite .
we have assumed that the absorptive part of the amplitude for the loop diagrams are induced by @xmath3 @xmath71 cut@xcite .
@xmath72 or @xmath70 can also be on shell and give new contribution to absorptive part under the condition @xmath73 or @xmath74 , respectively .
we excluded these two cases for simplicity for the following reasons . under our assumption about the stop mass
, @xmath73 can be satisfied only in very narrow susy parameter region which is simply excluded in our calculation .
@xmath74 means the cp
odd effect will appear as @xmath75 ( we take @xmath76 ) which has too small branch ratio and can be ignored .
two quantities are defined to represent the cp asymmetry effects , @xmath77 @xmath78 @xmath79 where @xmath80 and @xmath81 are differential widths of top quark and top anti
@xmath82 reflects the theoretical cp
odd effect appearing when the invariant mass @xmath64 is above the threshold .
@xmath83 reflects the experimental cp
asymmetry effect when we measure the decay events with @xmath84 , which is fixed to be @xmath85 in the work .
we have only considered the contribution to the denominator in eq .
[ def ] from the tree
level diagrams which gives @xmath86 .
the nominator comes from the interference of the one loop diagrams with the tree
level diagram , as @xmath87 the three body final state `` phase space '' is @xmath88 the above expression multiplied by a @xmath89 gives the phase space for fixed @xmath90 .
@xmath91 , @xmath92 are the amplitudes for the process @xmath61 and its cp conjugate process @xmath93 , respectively .
@xmath91 can be expressed as @xmath94 where the two terms come from tree level and one loop diagrams respectively .
a , @xmath95 contain the cp violating phases both from km matrix element @xmath96 and from stop mixing matrix elements @xmath97 .
@xmath98 develops an absorptive part for @xmath90 beyond the threshold .
@xmath92 can be expressed as @xmath99 so , we have @xmath100 @xmath101 is given by cutkosky rule , @xmath102 where @xmath103 are the phase space of @xmath3,@xmath2 as they are on shell .
@xmath104 , @xmath105 are the four momenta of @xmath3 and @xmath2 , respectively . after summing up all spins of external particles we get , e.g. , the interference term of fig.2a and the tree
level graph of the form @xmath106 where we have introduced @xmath107s to represent quantities like @xmath108 in eq .
[ aa ] arising from the susy couplings and the corresponding @xmath109s to represent quantities like @xmath110 which are lorentz invariant functions of the four momenta of @xmath3 , @xmath111 , @xmath62 and @xmath63 .
@xmath112 and @xmath113 are denominators of the two boson propagators in the loop graphs .
note that the sm phase from @xmath96 is cancelled in the interference of the tree diagram and the one loop diagrams .
we get the analytic expressions for @xmath114 by integrating the phase space of @xmath3 and @xmath2 in @xmath115 system .
the analytic results are given in appendix c. by expressing the formulae in lorentz invariant form , we translate the formulae to the top quark rest system . in this system
the final state three body phase space integration is implemented numerically .
all the @xmath116s for each graphs are proportional to @xmath117 ( for detail expressions of @xmath116s , see appendix c ) , so , we can write @xmath118 where @xmath119 @xmath120 and @xmath121 are given in ( a.13 ) and ( a.9 ) of appendix a. in terms of parameters in the top squark mass matrix we get that @xmath122 from the expression of @xmath123 in ( a.13 ) of appendix a we can see that @xmath124 can be as large as @xmath125 when the following conditions are satisfied at the same time , i.e. , @xmath126 , @xmath127 and @xmath55 is purely imaginary .
this is certainly difficult to reach .
we now turn to our main numerical results .
the calculation is based on the low energy mssm scenario whose parameter freedom has been greatly reduced as described in section 2 . another simplifying assumption taken in our calculation is the universal relationship between the gaugino masses , i.e. , @xmath128 , where @xmath129 is the weak mixing angle@xcite .
we write the parameters @xmath130 , @xmath131 in the top squark mass matrix as @xmath132 where @xmath133 is an arbitrary mass scale for scalar particles . neglecting the masses of @xmath134 and quarks except top quark
we are left with ten susy parameters , i.e. , @xmath16 , @xmath135 , @xmath136 , c , @xmath133 , @xmath137 , @xmath0 , @xmath138 , @xmath139 and @xmath140 .
we take @xmath141 to which the results are insensitive , and always take @xmath142 , @xmath143 .
the other free susy parameters are restricted by the experimental limits on the masses of super particles @xcite and our assumption @xmath144 .
in particular , we take that , @xmath145 , the mass of the lightest neutralino , is above 30gev and @xmath146 , the mass of the light chargino , is above 65gev .
another limit for the parameter space is adopted for simplicity that we require @xmath147 and @xmath148 .
the sm parameters are taken as @xmath149 , @xmath150 , @xmath151 , @xmath152 , @xmath153 and @xmath154 . a consequence of the above scenario ,
especially that @xmath50=0 and @xmath155 is that figs.2c , 2d and 2f do not acquire any cp violating phase .
so only figs.2a , 2b and 2e are considered .
it is found numerically that more than 90% of contributions to @xmath156 come from the triangle diagram fig.2e .
because of this , the results are not sensitive to the values of @xmath157 and @xmath140 .
we have studied the cp asymmetry , @xmath156 , as a function of susy parameters @xmath0 , @xmath16 , @xmath136 , @xmath135 and @xmath13 . in all the figures
there are two curves for the same values of the fixed parameters of which the curve giving larger @xmath156 represents @xmath82 while that giving a smaller value represents @xmath83 . in fig.3
we show the @xmath156 as a function of @xmath0 .
we can see that @xmath156 is approximately a sine function of @xmath0 as can be seen in eq .
[ rati ] and eq . 19 .
@xmath158 defined in eq .
[ rati ] is plotted in fig.4 .
we see that @xmath158 is just like a parabola . as a result @xmath156 does not reach its maximum when @xmath55 is purely imaginary , rather , it is maximal at @xmath159 .
@xmath158 depends on @xmath0 through the top squark mass ( see appendix a ) . in figs.57
, we plotted @xmath156 as a function of the higgs mass parameter @xmath16 for @xmath160 respectively , for different values of @xmath135 .
the global feature of the three figures is that @xmath156 decrease dramatically as @xmath161 increase .
notice that , in the high @xmath136 scenario @xmath156 becomes quite insensitive to the sign of @xmath16 .
@xmath156 is almost symmetric about @xmath161 for @xmath162 .
however , as @xmath163 @xmath156 is not small only for minus @xmath16 .
another feature of the figures is that for fixed @xmath136 and @xmath16 , @xmath83 decrease whereas @xmath82 increase as @xmath135 become larger .
the reason is evidently that the threshold varies with @xmath135 .
the dependence of @xmath156 on @xmath136 is plotted in fig.8 and fig.9 , for several values of @xmath135 and for @xmath164 and @xmath165 , respectively .
an interesting feature of these two figures is that @xmath166 decreases as @xmath136 increases whereas @xmath167 increases .
this is because low @xmath136 gives strong yukawa coupling for top quark so that we get large @xmath166 .
however , large @xmath136 makes the threshold lower and thus elevates @xmath167 .
so , @xmath167 prefers bigger @xmath136 in contrast to the situation in the top quark two body decays .
as @xmath168 , we can see from fig.8 that @xmath82 is almost insensitive to @xmath136 .
the dependence of @xmath156 on @xmath135 is plotted in fig.10 , for several values of @xmath136 .
we can see that the major part of the curve @xmath83 falls into the region between 0.1%0.2% . for @xmath169 , @xmath156 rises as @xmath135 increases whereas for large @xmath136 @xmath156 slightly drops as @xmath135 increases .
finally , we give the dependence of @xmath156 on the top squark mass in fig.11 for @xmath165 .
it is found that @xmath156 depends essentially only on @xmath13 , not separately on @xmath133 and c. @xmath156 decreases with @xmath170 , just as what has been expected .
when @xmath13 is around 140 gev @xmath156 can reach up to 0.5% . in summary
, @xmath167 is around 0.1% level in the major part of the parameter space that we have discussed . in a very narrow region of the parameter space
, @xmath167 can reach 0.5% level .
we would like to point out two points in this section .
the branch ratio for top quark decay drops rapidly when w boson is off shell . for @xmath171
the branch ratio for @xmath172 final state is only about @xmath173 according to our calculation .
however , as we have mentioned that the contributions to @xmath156 comes mainly from the triangle diagrams in fig.2 , @xmath156 for different final states have the same sign and approximately the same size .
so we can make a combining analysis for the data of the three body cp asymmetries for different three fermion final states ( @xmath174 ) , ( @xmath175 ) , ( @xmath176 ) , and all allowed three quark decays .
the total branch ratio for top quark three body decay for @xmath171 can reach up to about @xmath177
. however , there are cp asymmetries of the order @xmath178 for three quark decays coming from gluon and gluino corrections , which may be of the same order as considered here .
this needs to be studied further if one attempts to quantitatively compare the theory and experimental cp asymmetries for three quark decays .
the total width of top and anti
top quark are equal @xmath179 due to cpt theorem .
the following relation holds @xmath180 where @xmath181 represent all the w boson decay products mentioned above and the rhs of eq .
[ eq ] denotes the cp asymmetry from the diagrams with @xmath181 as intermediate states .
the relation can be easily seen from the feynman diagrams representing these two processes in fig.1 , 2 and fig.12 .
the four interference terms contributing to the rhs of eq .
[ eq ] between the two triangle diagrams and the two tree
level diagrams in fig.12 correspond to the interference terms between the four box diagrams in fig.2 and the tree
level diagram in fig.1 .
the other two interference terms of the fermion loop diagram with the two tree level graphs in fig.12 correspond to those of the triangle diagrams in fig.2 with the tree
level diagrams in fig.1 .
the relative minus sign on the rhs of eq .
[ eq ] can be explained as following .
we must add a minus sign to the amplitude for fig.12e due to the closed fermion loop .
such a minus sign is absent in the corresponding amplitude in fig.2 . on the contrary
, there is a minus sign in front of the box diagram in fig.2 .
no such minus sign is present in fig12.c , d. thus , the cpt relation eq .
[ eq ] manifests itself through the cutkosky rule in eq .
in this work we considered the cp asymmetries induced by mssm new phase in the semi
leptonic three body decays of the top quark under such an assumption that the top squark is so heavy that no cp
odd effects are observable in top quark two body decays in one loop level . in our calculation @xmath156 can reach up to 0.55% in the most favorable case .
considering the small total branch ratio for @xmath182 which is about @xmath177 for all three body decays it is really hard to detect such small effects experimentally .
however , several constraint in our calculations can be relaxed .
for example , the gaugino mass parameters @xmath183 and @xmath135 can be complex and give new cp violating sources if we do not make the universal assumption about the gauginos masses in gut scale .
another possible cp source is from the higgs mass parameter @xmath16 .
according to recent studies on the neutron edm , cancellation among different contributions can take place so that @xmath16 can have large imaginary part even when super particles are in @xmath184 level@xcite .
if this is the case , complex @xmath16 can introduce additional cp asymmetries in top quark decays . and
, if we relax the constraint on the @xmath13 to be slightly above 100gev , @xmath13 and @xmath185 cut can also give contribution to @xmath156 in three body decays . another improvement may come from branch ratio enhancement .
the direct experimental constraint on @xmath186 is even below @xmath187 today @xcite .
as @xmath64 decrease the branch ratio for top quark three body decay increase rapidly ( e.g. , the branch ratio for all three body decays is about @xmath188 for @xmath189 ) .
so , if @xmath186 is not much heavier than @xmath187 , the @xmath156 in the three body decay is hopefully detectable in the lhc which may be able to produce @xmath190 @xmath191 pairs . on the contrary , if @xmath192 and @xmath170 are both heavier than , e.g. , 140gev , all the windows for cp asymmetries induced by super particles will be shut up .
cp odd effects in top quark decays induced by mssm particles can only exist beyond one loop order .
this will be beyond the experiment ability in the near future .
* appendix a * + there are different conventions for super particles mass matrices adopted in literature so that it is very easy to make sign errors . to avoid sign errors we rederived the mssm lagrangian .
for most part we adopted the same conventions described by j. rosiek@xcite .
the original mssm lagrangian is given by eq .
[ 1]eq .
[ 10 ] of which all fields are gauge eigenstates .
we should point out that the @xmath193 convention will give an extra minus sign to parameter @xmath16 compared to those adopted the convention @xmath194(if universal relation for gauginos is taken only relative sign between the gauginos and @xmath16 is significant ) . to get the physical spectrum of particles
one should carry out the standard procedure of gauge symmetry breaking .
after ssb super particles will mix and form different mass eigenstates .
the charged higgsinos and charged winos mix and give two mass eigenstates named charginos .
the mass matrix of charginos is @xmath195 \ \ .\ ] ] the mixing matrices satisfy @xmath196 and is defined by @xmath197 @xmath198 in the above equations @xmath199 where @xmath200 are the first and second components of wino as in eq .
@xmath201 is the first ( or up ) fermion component of the second higgs super field doublet .
the fields on the left hand side of the above equations are gauge eigenstates and the fields on the right hand side are mass eigenstates .
the four component dirac spinor charginos are defined by @xmath202 $ ] .
the mass term which will appear in the final form of lagrangian is @xmath203 .
the third component of wino , photino and neutral higgsinos combine to give four majarana neutralinos .
the mass matrix for neutralinos is @xmath204 \ , \ \\ ] ] which is diagonalized by @xmath205 @xmath59 is defined by @xmath206 where @xmath207 is photino .
all fields on the left hand side of the above equation are gauge eigenstate and those on the right hand side are mass eigenstates
. the four component dirac spinor form for neutralino is @xmath208 $ ] .
the mass term in the final form of lagrangian of neutralino is @xmath209 .
ignoring generation mixing , the mass eigenstates of squarks are obtained by mixing the left handed and right handed eigenstates of squarks .
its mass matrix is @xmath210\\ % \end{equation } % \begin{eqnarray } l_f&=&m^2_f + \cos 2 \beta(t_{3f } - q_f \sin^2 \theta_w)m_z^2 + m^2_{\tilde{f}_l}\nonumber\\ r_f&=&m^2_f+\cos 2\beta q_f \sin^2 \theta_wm_z^2+m^2_{\tilde{f}_r}\nonumber\\ c_f&=&- m_f(r_f\mu + a_f^*)=|c_f|e^{i\phi_f}\end{aligned}\ ] ] where @xmath211 is @xmath125 for up squark and @xmath212 for down squark .
@xmath213 is the charge of the sparticle and @xmath214 is @xmath215 for up squark and @xmath136 for down squark .
@xmath216 , @xmath217 , @xmath218 are the corresponding diagonal elements of @xmath44 , @xmath45(@xmath46 ) and @xmath47(@xmath48 ) in @xmath41 , respectively .
the mixing matrix satisfies @xmath219 they are defined , e.g. , for up squark , by @xmath220 where @xmath221 is the up component of left
handed up squark doublet , @xmath222 is the right handed up squark .
@xmath223 are the two mass eigenstates of up squark . the conjugate of @xmath222 in eq .
[ dd ] comes from that we adopt the charge conjugate of left hand fermion to represent its right hand component in the original mssm lagrangian .
the final form of squark mass term in lagrangian is @xmath224 .
in particular , the mixing matrix for top squark is given by @xmath225\ \ , \end{aligned}\ ] ] where @xmath121 is defined in ( a.9 ) and @xmath226 the formulae for down squark are similar to up squark .
note the definition of mixing matrix for down squark given by @xcite is the complex conjugate of the mixing matrix given here .
* appendix b * + in this appendix we list the relevant pieces of the susy lagrangian in terms of the mass eigenstates@xcite .
@xmath227t\ + \ h.c.\\ { \cal l}_{b\tilde{t}\chi}&=&g\tilde{t}_{i}\bar{b}[c^{ij}p_{l}+d^{ij}p^{r } ] v_{tb}^{*}\chi^{c}_{j}+h.c.\\ { \cal l}_{b\tilde{b}\chi^0}&=&g\tilde{b}_i^{*}\bar{\chi^0_j}[e^{ij}p_{l}+f^{ij } p_{r}]b+h.c.\\ { \cal l}_{t\tilde{b}\chi}&=&g\tilde{b}_i^{*}\bar{\chi_j}[g^{ij}p_{l}+h^{ij}p_{r } ] v_{tb}^{*}t+h.c.\\ { \cal l}_{\tau\tilde{\tau}\chi^0}&=&g\tilde{\tau}^{*}_{i}\bar{\chi^0_j } [ m^{ij}p_{l}+n^{ij}p_{r}]\tau+h.c.\\ { \cal l}_{\nu\tilde{\tau}\chi}&=&gu^{ij}\tilde{\tau}_{i}^{*}\bar{\chi_j}p_{l } \nu_{\tau}+h.c.\\ { \cal l}_{\tau\tilde{\nu}\chi}&=&-g\tilde{\nu}_{\tau}^{*}\bar{\chi^c_i } [ z^+_{1i}p_{l}+l^{\tau}{z^{-}_{2i}}^{*}p_r]+h.c.\\ { \cal l}_{\nu\tilde{\nu}\chi^0}&=&gw_{i}\tilde{\nu}_{\tau}^ { * } \bar{\chi^0_i}p_{l}\nu_{\tau}+h.c.\\ \label{kkw } { \cal l}_{\chi\chi^{0}w}&=&g\bar{\chi_i}\nu^{\mu}[o^{ij}p_{l}+v^{ij}p_{r } ] \chi^{0}w^+_{\mu}+h.c.\end{aligned}\ ] ] where @xmath228 @xmath229 , @xmath230 and @xmath231 are the yukawa coupling parameters for top , bottom and @xmath134 respectively .
they are given by @xmath232 the vertex in eq .
[ kkw ] can be expressed as @xmath233\chi^0_{i}w_{\mu}^{-}+h.c.\ ] ] which can be used to read off the feynman rule directly for the fig.2e . * appendix c * + 2 x_a^2 in this section we give the formulae for the quantity @xmath234 in eq .
[ main ] for fig .
2c and 2e in detail . in the following formulae
the @xmath116s are not separated out explicitly as in eq .
[ main ] for convenience .
all the formulae and variables are given in the @xmath115 system .
the imaginary part of the amplitude for fig2.c is @xmath235\ [ { ( p_{\nu}-k)}^{2}-m^2_{\tilde{\nu } } ] } \end{aligned}\ ] ] where @xmath236 is the four - momenta of @xmath3 and the @xmath237s are interaction vertex factors . the quantity @xmath114 for fig.2c is @xmath238 where @xmath239 in above formulae , @xmath240 is the energy of @xmath241 in the @xmath115 system .
@xmath242 is the angle between the three momenta of top quark and @xmath63 in the @xmath115 system .
@xmath90 is the invariant mass square of the final state lepton pair .
other quantities are defined as : @xmath243 the non lorentz invariant four vector component are expressed as @xmath244 @xmath245 are given by @xmath246 the elements of the mixing matrices in the above equation are taken according to that only @xmath247 and @xmath248 are considered in our calculations .
the corresponding formulae for fig2.e are @xmath249 } \end{aligned}\ ] ] the quantity @xmath114 for fig.2e is @xmath250 where @xmath251|\vec{k}|^2\nonumber\\ y&=&\omega - s|\vec{p_t}||\vec{k}|+t\en^{'}\cos\alpha\nonumber\\ z&=&\sigma+see_t+t^{'}e\en^{'}+h\nonumber\\ s&=&(2\et \pb \cdot \pg -\eff m_{\chi^0}m_{t } ) \pg \cdot \ptau \nonumber\\ t&=&2\es m_{\chi}m_{t}\pb \cdot \pg -\eff m_{\chi^0}m_{t } \pb \cdot \pt-2\et q\cdot \pt \pb \cdot \pg \nonumber\\ t^{'}&=&2\es m_{\chi}m_{t}\pb \cdot \pg - \eff m_{\chi^0}m_{t}(\pb \cdot \pg -\pb \cdot \ptau ) -\nonumber\\ & & 2\et \pb \cdot \pg ( \pt\cdot \pg -\pt\cdot \ptau ) \nonumber\\ \omega&=&2\et \
{ -\pb \cdot \ptau e\en^{'}|\vec{k}|(|\vec{p_t}|+e_{t}\cos\alpha)+ \pt\cdot \pg ee^{'}_{\tau}|\vec{k}|(e_{t}\cos\alpha-|\vec{p_t}|)\nonumber\\ & & + ( 2ee_{t}-k\cdot q)|\vec{p_t}||\vec{k}|\pg \cdot \ptau - k\cdot q \en^{'}|\vec{k}|\cos\alpha_{0}\cdot \pg \}\nonumber\\ % \sigma&=&2\et \ { e^{2}e_{t}\en^{'}(\pb \cdot \ptau + \pt\cdot \pg ) - ee^{'}_{\tau}k\cdot q \pt\cdot \pg -\nonumber\\ & & ( ee_t - k\cdot q)ee_{t}\pg \cdot \ptau - \en^{'}(e^{2}-\frac{1}{2}|\vec{k}|^{2})\pt\cdot \pb \}\nonumber\\ h&=&\eff m_{\chi^0}m_{t}\pg \cdot \ptau k\cdot q + 2(\ef m_{\chi^0}m_{\chi}\pt\cdot \ptau + \eff m_{\chi^0}m_{t}\pg \cdot \ptau ) \pb \cdot \pg \nonumber\\ & & -\et m^{2}_{\chi^0}(\pb \cdot \pg \pt\cdot \ptau -\pb \cdot \pt\pg \cdot \ptau + \pt\cdot \pg \pb \cdot \ptau ) \end{aligned}\ ] ] the parameters @xmath252s are given by @xmath253 from the above expressions and the expressions for a , d in eq .
[ b10 ] we can see that the @xmath254s are proportional to @xmath255 as pointed out in section 3 .
99 s. bar shalom and g. eilam , hep ph/9810234 and
references therein .
b. grzadkowski and w. y. keung , phys .
b * 319 * , 526 ( 1993 ) .
e. christova and m. fabbrichesi , phys .
b * 320 * , 299 ( 1994 ) .
s. bar shalom , d. atwood and a. soni , phys .
d * 57 * , 1495 ( 1998 ) .
particle data group , eur .
j. * c3 * , 1 ( 1998 ) .
w. fischler , s. paban and s. thomas , phys .
lett * b289 * , 373(1992 ) ; + s. m. barr , int . j. mod
a8 * , 209 , ( 1993 ) ; + t. farlk , nncl . phys .
. suppl . * 52 * , 78 ( 1997 ) .
h. e. haber and g. l. kane , phys .
rep . * 117 * , 75 ( 1985 ) ; + h. p. nilles , phys
. rep . *
110 * , 1 ( 1984 ) .
j. rosiek , phys .
d * 41 * , 3464 ( 1990 ) .
s. dimoloulos and s. thomas , nucl .
b*465 * , 23 ( 1996 ) y. nir and n. seiberg , phys . lett .
* b309 * , 337 ( 1993 ) . j. d. brjoken and s. d. drell , _ relativistic quantum mechanics _
, + chapter 8 , mcgraw hill , new york , 1964 .
h. simma , g. eilam and d. wyler , nucl .
b * 352 * , 367 ( 1991 ) v. berger and p. ohmann , phys .
d * 49 * , 4908 ( 1994 ) .
michal brhlik , gerald j. good and g. l. kane , hep
|
we calculate the cp violating effects in the top quark semi
leptonic three body decays induced by the supersymmetric cp odd phase of the top squark trilinear soft breaking term @xmath0 .
the light top squark mass is assumed to be close to the top quark mass @xmath1 .
the cp conserving phase is provided by the @xmath2 and @xmath3 cut .
we find that the partial rate asymmetry is in the 0.1% level . in the most favorable parameter region
the decay rate asymmetry can reach up to 0.55% . * cp violation in semi
leptonic decays of the top quark within mssm * xiao - jun bi and yuan - ben dai +
|
the hydrodynamic ( hydro ) model has been applied extensively to heavy ion collisions at the super proton synchrotron ( sps ) and rhic as part of a search for formation of a quark - gluon plasma ( qgp ) @xcite .
hydro is intended to describe collision evolution in terms of flowing hot and dense matter , possibly a qgp @xcite .
validity of the hydro description could support inference of parton thermalization @xcite and direct comparisons with lattice qcd @xcite .
the hydro model appears to be successful in representing some aspects of particle data .
hadron spectra have been described in terms of radial flow combined with the statistical model @xcite .
some azimuth correlations have been described in terms of elliptic flow @xcite .
based on elliptic - flow systematics more - central rhic au - au collisions have been characterized in terms of a strongly - coupled qgp ( sqgp ) with small viscosity a `` perfect liquid '' @xcite . according to the hydro model
almost all particle production over much of a - a momentum space reflects rapid thermalization and development of a flow field in response to initial pressure gradients .
mass dependence of the hadron momentum distribution should reflect the underlying flow system @xcite .
however , differential spectrum and correlation analysis reveals structures whose variation with a - a centrality and collision energy contradict hydro expectations @xcite . interpretation of measured @xmath0 as the property of a flowing bulk medium can be strongly questioned @xcite . in this paper
i examine assumptions and procedures which seem to support a hydro description at rhic and assess their validity in the context of fiducial data .
i contrast the hydro description with a two - component model of elementary n - n collisions and the glauber linear superposition ( gls ) reference model of a - a collisions .
the emerging two - component phenomenology is consistent with qcd systematics , does not require the hydro model and conflicts with hydro in several ways .
the paper is organized as follows : after summarizing two rhic paradigms ( hydro / hard - probes vs quadrupole / minijets ) i review arguments for and against the hydro model .
i summarize analysis methods and plotting formats which play a central role in shaping interpretation of rhic data .
i review the two - component model of nuclear collisions and the systematics of minijets ( minimum - bias jets , @xmath1 gev ) .
i consider so - called `` triggered '' jet analysis and related systematic errors .
i review assumed hydro initial conditions compared to results from elementary and nuclear collisions .
lastly , i review hydro applications to single - particle spectra ( blast - wave model vs fragmentation ) and azimuth correlations ( elliptic flow vs jet systematics ) .
two paradigms compete to describe rhic data .
the conventional rhic paradigm emerged from the bevalac heavy ion program , where nuclear collisions were dominated by semiclassical molecular dynamics , some degree of thermalization was achieved and thermodynamic state variables described collisions . hydrodynamics plays a dominant role in the paradigm .
anticipated novel aspects of rhic collisions at much higher energies include possible formation of a thermalized partonic medium or qgp at larger energy densities and modification of parton hard scattering and jet formation by the medium .
an alternative paradigm is based on fragmentation processes in elementary @xmath2-@xmath3 ( e - e ) and n - n collisions as a qcd reference system .
n - n collisions above @xmath4 gev are described near mid - rapidity by a two - component model of longitudinal and transverse fragmentation .
linear superposition of n - n binary collisions provides a reference model for collisions .
deviations from the reference may reflect novel qcd physics in a - a collisions . in the conventional picture of rhic collisions
a thermalized `` bulk partonic medium '' results from copious rescattering of @xmath51 gev partons ( gluons ) @xcite .
resulting pressure gradients drive hydrodynamic expansion and formation of a velocity field ( elliptic , radial and longitudinal flows ) .
the medium expands , cools and `` freezes out '' to form hadrons which may continue to rescatter and expand collectively .
more - energetic partons ( hard probes ) are multiply scattered in the partonic medium and lose energy by gluon bremsstrahlung , possibly stopping in the medium ( parton thermalization ) .
the hadronic final state is described by hydrodynamic models , statistical - model hadrochemistry and perturbative qcd ( pqcd ) jet quenching @xcite .
phenomenologically , hadron single - particle ( sp ) transverse momentum @xmath6 spectra ( cf .
[ fdaa ] ) are divided into a low-@xmath6 region @xmath7 gev / c , nominally representing the thermalized bulk medium and described by blast - wave ( bw ) and statistical models , and a high-@xmath6 region @xmath8 gev / c dominated by parton scattering and fragmentation and described by pqcd .
the intermediate-@xmath6 region ( @xmath52 - 5 gev / c ) manifests novel structure possibly described by hybrid models ( e.g. , recombination or coalescence of `` constituent quarks '' ) denoted by reco @xcite . in recent years differential analysis of spectrum and correlation data from p - p collisions has provided a detailed _ phenomenological reference _ for nuclear collisions .
accurate determination of a - a centrality extending to n - n collisions was also achieved .
the combination established a _
glauber n - n linear superposition _ or gls reference : what if nothing new happens in a - a collisions ?
spectrum and correlation analysis of p - p collisions revealed a significant contribution from `` minijets '' ( fragments from the _ minimum - bias _ parton spectrum dominated by @xmath53 gev partons , cf .
[ minijets1 ] ) contributing a significant fraction of the p - p hadron spectrum down to 0.3 gev / c .
minijet phenomenology is fully consistent with jet systematics at larger energy scales@xmath2-@xmath3 and fragmentation functions @xcite .
minijet correlations demonstrate that observable consequences of qcd quanta at small energy scales persist in a - a collision data and even dominate the hadronic final state .
the 2d angular autocorrelation method developed to study minijet correlations also determines ( directly ) the _ azimuth quadrupole moment _
( model - independent terminology @xcite ) conventionally identified with `` elliptic flow . '' the autocorrelation technique accurately resolves the azimuth quadrupole ( `` flow '' ) from minijet structure ( `` nonflow '' ) .
quadrupole energy and centrality systematics appear to be inconsistent with hydro expectations or an equation of state ( eos ) , instead are suggestive of qcd multipole radiation @xcite . in the context of the gls reference minijet and quadrupole variations with a - a centrality and energy challenge assumptions supporting the hydro model .
minijets serve as _ brownian probes _ of hypothetical `` bulk matter '' @xcite and test conclusions from the hydro model : is there a collective flowing medium ?
is the medium opaque to jets ?
is viscosity relevant ?
this paper compares arguments supporting the two paradigms and challenges the validity of a hydro description of rhic collisions .
application of hydrodynamics to rhic collisions is inherited from the bevalac program which characterized nuclear collisions in terms of the molecular dynamics of nucleon clusters and nucleons and demonstrated collective flow phenomena .
similar results were obtained at the ags for systems of nucleons and nucleon resonances at the higher energy .
hydro arguments extended to the context of rhic collisions are summarized below .
[ [ initial - conditions ] ] initial conditions + + + + + + + + + + + + + + + + + + the pqcd energy spectrum for scattered partons is assigned a low cutoff energy ( @xmath9 gev)justified by saturation - scale arguments which implies a large parton ( gluon ) phase - space density .
[ [ parton - thermalization ] ] parton thermalization + + + + + + + + + + + + + + + + + + + + + the resulting large parton density ( 10 - 30 times larger than expected from glauber extrapolation of p - p collisions ) implies thermalization through parton multiple scattering . [ [ equation - of - state ] ] equation of state + + + + + + + + + + + + + + + + + assumed thermalization ( @xmath10 reversibility , detailed balance ? ) justifies system description with state variables related by an eos .
[ [ hydrodynamic - evolution ] ] hydrodynamic evolution + + + + + + + + + + + + + + + + + + + + + + the eos relates a large initial energy density to pressure gradients converted via hydrodynamic evolution to a bulk - matter velocity field [ [ parton - energy - loss ] ] parton energy loss + + + + + + + + + + + + + + + + + + `` hard probes '' of the thermalized medium exhibit parton energy loss and modified fragmentation . a `` well - understood '' pqcd phenomenon is interpreted to reveal bulk - medium properties [ [ late - hadronization ] ] late hadronization + + + + + + + + + + + + + + + + + + the thermalized , flowing partonic bulk medium hadronizes by constituent - quark coalescence , producing nearly all final - state hadrons . [
[ hadron - rescattering ] ] hadron rescattering + + + + + + + + + + + + + + + + + + + isentropic transverse expansion of the resulting hadron resonance gas leads to different chemical and kinetic decoupling temperatures .
[ [ final - state - flows ] ] final - state flows + + + + + + + + + + + + + + + + + evidence is sought in final - state hadronic spectra and correlations e.g . ,
mass dependence of @xmath11 , radial and elliptic flows for a flowing partonic medium established prior to hadronization .
[ [ parton - spectrum ] ] parton spectrum + + + + + + + + + + + + + + + a parton spectrum cutoff near 3 gev is established by minijet correlations and spectrum hard components from p - p and au - au collisions .
[ [ thermalization ] ] thermalization + + + + + + + + + + + + + + three separate components [ soft , hard ( minijets ) and quadrupole ] have distinct characteristics , different mechanisms and remain largely independent throughout the collision for all a - a centralities . [ [ state - variables ] ] state variables + + + + + + + + + + + + + + + there is no evidence for a thermodynamic state ( homogeneity , reversibility or detailed balance ) and therefore no support for state variables .
[ [ parton - multiple - scattering ] ] parton multiple scattering + + + + + + + + + + + + + + + + + + + + + + + + + + minijet correlations indicate that essentially all initial - state scattered partons appear as correlated structures in the hadronic final state , even in central au - au collisions .
there is negligible parton rescattering
minijet azimuth widths grow _ smaller _ , not larger with increasing a - a centrality . [ [ flows ] ] flows + + + + + no radial flow is observed in @xmath6 spectra .
sp spectrum structure is described by two fragmentation components , is inconsistent with the blast - wave model .
@xmath12 data can be interpreted in terms of qcd multipole radiation . in the context of the two - component model ,
arguments which support hydro implicitly invoke a competition between soft ( nucleon fragmentation ) and hard ( scattered - parton fragmentation ) components for final - state hadron production near mid - rapidity .
the spectrum soft component , the @xmath6 spectrum from projectile nucleon fragmentation , represents the majority of hadron production for all a - a centralities .
it is a universal feature of fragmentation observed also in e - e collisions .
the spectrum hard component , the @xmath6 or @xmath13 distribution of fragments from minimum - bias large - angle - scattered partons , represents the main transport mechanism from longitudinal to transverse parton phase space .
arguments supporting hydro assume that most of the hard component is thermalized , and the resulting medium dominates hadron production in more - central collisions .
collisions are in some sense `` opaque '' to scattered partons which undergo many rescatterings .
the n - n soft component likewise dissolves into the medium in more - central au - au collisions .
hadron production is dominated by hadronization of the thermalized medium . a small jet - correlated fraction results from fragmentation of rare high-@xmath6 partons ( possibly a `` surface bias '' ) .
observed strong minijet correlations and independent quadrupole systematics contradict those assumptions .
very different interpretations can emerge from the same data depending on analysis methods and plotting formats .
preferred analysis methods extract all information from spectra and correlations in directly - interpretable form .
_ differential _ analysis incorporates well - understood and comprehensive reference models of nucleon and parton scattering and fragmentation , a - a centrality dependence and two - particle correlations . [ [ glauber - model ] ] glauber model + + + + + + + + + + + + + a - a collision geometry parameters @xmath14 , @xmath15 , @xmath16 and @xmath17 related to observables such as @xmath18 at mid - rapidity ; the preferred centrality measure is mean participant pathlength @xmath19 @xcite .
[ [ parton - scattering - and - fragmentation ] ] parton scattering and fragmentation + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + complete parametrization of parton scattering and fragmentation for e - e and p - p collisions ; parton power - law spectrum inferred from p - p data ; calculated fragment distributions derived from a pqcd folding integral @xcite .
[ [ two - component - model ] ] two - component model + + + + + + + + + + + + + + + + + + + soft+hard spectrum and correlation components representing non - single diffractive ( nsd ) n - n soft scattering and ( longitudinal ) fragmentation and pqcd semihard parton scattering and ( transverse ) fragmentation @xcite . [ [ glauber - linear - superposition - gls ] ] glauber linear superposition ( gls ) + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + essential reference for centrality dependence of a - a collisions over the complete centrality range ; spectrum and correlation measures normalized by participant - pair number @xmath20 , reference soft component is invariant , reference hard component increases proportional to @xmath21 ; physics unique to a - a deviates from the gls reference . in high - energy collisions the choice of plotting variables@xmath13 _ vs _ @xmath6 , @xmath22 _ vs _
@xmath23 or @xmath24 , @xmath25 _ vs _ @xmath14 , @xmath0 _ vs _ @xmath26$]strongly affects the visibility of important structure and its physical interpretation .
conventional momentum - space variables are @xmath27 for longitudinal momentum and @xmath6 for transverse momentum
. a typical mid - rapidity detector acceptance at rhic ( star tpc ) corresponds ( for pions ) to @xmath28 gev / c ( @xmath29 ) and @xmath30 gev / c ( @xmath31 ) .
a more sensible choice might be @xmath27 and @xmath13 or even @xmath32 and @xmath13 .
linear @xmath6 over - emphasizes the `` high-@xmath6 '' region where physical changes are modest at the expense of the small-@xmath6 region where substantial new qcd physics emerges at rhic .
transverse rapidity @xmath33 more - clearly displays systematic qcd trends .
two examples of variable choices and consequences are presented , one from parton fragmentation , the other from differential spectrum analysis with ratios .
[ [ parton - fragmentation ] ] parton fragmentation + + + + + + + + + + + + + + + + + + + + the most important kinematic region for parton fragmentation in nuclear collisions includes small parton energies and small fragment momenta where most fragments are produced .
fragmentation functions ( ffs ) are conventionally represented on linear momentum fraction @xmath34 or @xmath35 , whereas the format most relevant to nuclear collisions is on rapidity @xmath36 or normalized rapidity @xmath37 , with @xmath38 and @xmath39 for @xmath2-@xmath3 collisions @xcite .
left panel : beta distribution ( solid ) @xcite and kkp ff ( dashed ) @xcite curves compared to opal 91 gev data points ( open circles ) @xcite on linear momentum variable @xmath23 .
right panel : the same curves and data transformed to normalized rapidity @xmath40 .
the vertical dotted lines correspond to @xmath41 .
, width=316,height=168 ] fig .
[ kkp ] illustrates the consequences .
fiducial ff data from opal ( @xmath42 gev ) @xcite are plotted on linear momentum fraction @xmath23 on the left , and on normalized rapidity @xmath40 on the right . less than 10% of the distribution falls to the right of the dotted line in each panel .
the dashed curves represent a 15-parameter pqcd - theory formulation on @xmath23 @xcite .
the solid curves describe a _
two_-parameter beta distribution on @xmath40 @xcite .
while more - recent theory describes ff data down to @xmath43 @xcite this figure illustrates visual consequences of plotting formats . in the left panel pqcd ( kkp ) and
beta parameterizations appear to describe the data equally well for @xmath44 where pqcd dglap evolution is emphasized @xcite .
however , the relation between models and data for 90% of the fragments , to the left of the dotted line ( the region which dominates nuclear collisions ) , is not resolved visually . in the right panel
the large discrepancy between theory and data becomes clear , as does the accuracy of the beta parametrization which describes all fragments over all energy scales relevant to nuclear collisions ( @xmath45 gev / c , @xmath46 gev ) @xcite .
ff modes occur in the interval 1 - 2 gev / c for all parton energies up to several hundred gev . in nuclear collisions at least 50% of scattered - parton fragments fall below 1 gev / c .
[ [ spectrum - ratios ] ] spectrum ratios + + + + + + + + + + + + + + + spectrum ratios are differential measures of spectrum centrality variation .
two ratio measures have been defined [ raaeq ] r_aa & = & = + r_aa & = & , where @xmath47 is the participant - nucleon mean pathlength , @xmath48 and @xmath49 are reference per - participant - pair soft and hard spectrum components respectively , and @xmath50 is the hard component extracted from data @xcite .
[ raa ] ( left panel ) shows ratio @xmath51 for five centralities of au - au collisions ( five bold curves of different line styles ) plotted on transverse rapidity @xmath13 @xcite .
also shown are five reference curves ( light solid curves ) with @xmath50 replaced by reference @xmath49 in eq .
( [ raaeq ] ) ( first line ) , and the solid points with @xmath50 replaced by @xmath52 data from p - p collisions @xcite corresponding to @xmath53 which fall along the reference line at 1 . left panel : spectrum ratio @xmath51 for five centralities of 200 gev au - au collisions ( bold curves ) and 200 gev collisions ( points ) plotted on transverse rapidity @xmath13 @xcite .
the thin solid curves are glauber linear - superposition ( gls ) references .
@xmath25 is a centrality measure .
right panel : alternative spectrum ratio @xmath54 for the same data .
the gls reference for all cases is unity .
the structure at smaller @xmath13 is newly revealed . , title="fig:",width=158,height=158 ] left panel : spectrum ratio @xmath51 for five centralities of 200 gev au - au collisions ( bold curves ) and 200 gev collisions ( points ) plotted on transverse rapidity @xmath13 @xcite .
the thin solid curves are glauber linear - superposition ( gls ) references .
@xmath25 is a centrality measure .
right panel : alternative spectrum ratio @xmath54 for the same data .
the gls reference for all cases is unity .
the structure at smaller @xmath13 is newly revealed .
, title="fig:",width=158,height=158 ] fig .
[ raa ] ( right panel ) shows alternative ratio @xmath54 defined in terms of spectrum hard components from the same data as in the left panel @xcite .
soft component @xmath48 is eliminated from the ratio , and full access to hard - component centrality trends is thus achieved .
the correct reference is unity for all @xmath13 , making deviations from the reference unambiguous .
more importantly , _ sensitivity _ to deviations is uniform over all @xmath13 .
the 200 gev p - p and 60 - 80% central au - au data clearly agree with the reference .
the large excursions at smaller @xmath13 for more - central au - au collisions , apparent for the first time with @xmath54 , represent most of the 30% increase in per - participant multiplicity for more - central collisions .
[ raa ] and eq .
( [ raademo ] ) demonstrate that @xmath51 can be quite misleading .
the soft component included in @xmath51 strongly suppresses hard - component structure below 4 gev / c , and the true reference curves are not acknowledged ( the reference is usually assumed to be unity ) . at @xmath55 ( pion @xmath56 gev / c ) [ raademo ] r_aa & & + r_aa , where @xmath57 for pions , and @xmath54 contains all information on spectrum centrality evolution .
hard - component centrality dependence below 2 gev / c ( @xmath58 ) is therefore strongly suppressed visually , and the relation to the correct reference is obscured .
plotting @xmath51 on conventional @xmath6 further distracts from relevant structure at smaller @xmath13 , emphasizing high-@xmath6 `` hard probes '' to the exclusion of important new fragmentation structure in a @xmath6 interval nominally assigned to hydro .
the upper - most ( solid ) data curve ( 60 - 80% central ) in the left panel could be described as `` suppressed , '' but is actually consistent with the gls reference ( uppermost thin solid curve ) over the entire @xmath13 acceptance .
the @xmath54 trend ( right panel ) for more - central collisions is still `` suppression '' above @xmath59 ( @xmath60 gev / c ) , but important new large - amplitude structure appears at the left end of the spectra .
centrality dependence near 10 gev / c ( @xmath61 , suppression ) is closely related to centrality dependence below 0.5 gev / c ( @xmath62 , enhancement ) .
no spectrum structure in this differential analysis corresponds to radial flow @xcite ( and cf .
[ protonrad ] ) .
figure [ v2comp ] illustrates variable choices and plotting formats critical to hydro interpretations .
the same @xmath12 data for three hadron species from minimum - bias au - au collisions at 200 gev appear in each panel @xcite .
in the left panel is the conventional @xmath12 vs @xmath6 format . in the right panel
the data have been processed so as to reveal quadrupole spectra on transverse rapidity @xmath13 for each hadron type .
@xmath12 is proportional to the ratio of two spectra : _ quadrupole _ spectrum @xmath63 in the numerator and the single - particle or sp spectrum in the denominator .
the quadrupole spectrum can be inferred from measured quantities by @xcite [ stuff ] ( y_t ) & = & \ { } \ { } + & & f(y_t;y_t0,y_t2)y_t2 _
2(y_t;y_t0,t_2 ) .
@xmath64 is the sp spectrum appearing in the denominator of @xmath65 .
monopole boost @xmath66 ( inferred from data ) is the transverse boost of a common source for all hadrons associated with the quadrupole .
the quantities in curly brackets are determined by @xmath66 [ except @xmath67 which is inferred from the shape of @xmath68 .
function @xmath69 is unity for smaller @xmath13 but increases smoothly to about 1.2 at @xmath70 for the plotted data @xcite .
left panel : @xmath6-differential elliptic flow measure @xmath12 for three hadron species from minimum - bias 200 gev au - au plotted in a conventional format @xcite .
hydro theory curves a and b are described in the text .
right panel : the same @xmath12 data plotted in an alternative format using proper rapidity @xmath13 for each mass species , which reveals azimuth quadrupole spectra with lvy form and common transverse boost @xmath71 . , title="fig:",width=158,height=168 ] left panel : @xmath6-differential elliptic flow measure @xmath12 for three hadron species from minimum - bias 200 gev au - au plotted in a conventional format @xcite . hydro theory curves
a and b are described in the text .
right panel : the same @xmath12 data plotted in an alternative format using proper rapidity @xmath13 for each mass species , which reveals azimuth quadrupole spectra with lvy form and common transverse boost @xmath71 . ,
title="fig:",width=158,height=165 ] fig .
[ v2comp ] ( left panel ) shows the conventional plotting format for @xmath12 .
the mass ordering at smaller @xmath6 is interpreted to imply hydro expansion .
`` saturation '' at larger @xmath6 might indicate jets correlated with the reaction plane .
[ v2pt ] presents a more - detailed discussion of @xmath12 .
the full sp @xmath6 spectrum in the @xmath0 denominator contains contributions from parton scattering and fragmentation ( hard component ) which dominate @xmath12 at larger @xmath6 .
theory curves a and b are described in sec . [ v2pt ] .
[ v2comp ] ( right panel ) based on eq . ( [ stuff ] )
removes from @xmath12 the sp spectrum in the denominator as well as an extraneous kinematic factor @xmath6 to reveal quadrupole spectra .
plotting data on rapidity proper to each hadron species reveals a common boosted source ( aligned left edges ) and source boost @xmath66 .
the universal lvy form of the spectra ( solid curves ) is also apparent , permitting direct interpretation of @xmath12 data without the intermediary of the theory to be tested .
support for the arguments in this paper falls into several categories .
basic analysis methods include centrality determination @xcite , 2d angular autocorrelations and inversion of fluctuation scale dependence @xcite .
determination of the two - component structure of single - particle ( sp ) spectra is described in @xcite .
fragmentation in elementary and a - a collisions is discussed in @xcite .
methods for two - particle correlations in p - p collisions are presented in @xcite .
correlation methods for a - a collisions are presented in @xcite .
analysis of triggered azimuth correlations is discussed in @xcite .
analysis methods related to the azimuth quadrupole ( elliptic flow ) are presented in @xcite .
the two - component ( soft+hard ) model of hadron production is an essential reference for spectra and correlations ( setting aside the azimuth quadrupole``elliptic flow''as an independent and relatively small third component ) .
the components represent 1 ) non - single diffractive ( nsd ) soft scattering and ( longitudinal ) fragmentation and 2 ) large - angle pqcd semihard parton scattering and ( transverse ) fragmentation respectively . coupled to glauber linear superposition the two - component model provides a comprehensive reference for a - a collisions .
the two - component spectrum model for p - p collisions sorted according to detected multiplicity @xmath72 is [ 2comprho ] & = & s_0(y_t ) + h_0(y_t ) , with @xmath73 and @xmath74 .
unit - normal model functions @xmath75 and @xmath76 are extracted from data in @xcite .
spectrum components @xmath77 and @xmath78 ( @xmath79-integrated 2d densities ) describe per - participant - pair particle densities ( @xmath80 are defined in one unit of @xmath81 ) .
soft component @xmath82 is derived from spectrum data as the physical - model - independent limit [ s0lim ] s_0(y_t ) .
hard component @xmath76 is defined by the average of n_chh_0(y_t ) - s_0(y_t ) , with @xmath83 .
figure [ ppspec ] ( left panel ) shows @xmath6 spectra for ten multiplicity classes from nsd p - p collisions at @xmath84 gev @xcite .
the spectra are normalized to soft - component multiplicity @xmath85 determined by an iterative process .
@xmath75 is clearly the limit of measured spectra as @xmath86 , per eq .
( [ s0lim ] ) . left panel : single - particle ( sp ) spectra from 200 gev p - p collisions for ten multiplicity classes normalized by soft - component multiplicity @xmath85 @xcite .
unit - normal lvy distribution @xmath75 is the soft - component model function common to all spectra .
right panel : two - component model for spectrum data in the left panel .
unit - normal hard component @xmath76 is a gaussian , with exponential tail representing a pqcd power law on @xmath6 . , title="fig:",width=158,height=158 ] left panel : single - particle ( sp ) spectra from 200 gev p - p collisions for ten multiplicity classes normalized by soft - component multiplicity @xmath85 @xcite .
unit - normal lvy distribution @xmath75 is the soft - component model function common to all spectra .
right panel : two - component model for spectrum data in the left panel .
unit - normal hard component @xmath76 is a gaussian , with exponential tail representing a pqcd power law on @xmath6 .
, title="fig:",width=158,height=158 ] fig .
[ ppspec ] ( right panel ) shows the two - component model of eq .
( [ 2comprho ] ) with @xmath87 and @xmath75 and @xmath76 from @xcite .
the @xmath18-independent model functions represent all significant structure in the spectrum data .
the two - component model is a nontrivial decomposition of p - p @xmath6 spectra .
the glauber model of a - a collisions relates mean participant - pair number @xmath88 and mean number of binary n - n collisions @xmath15 to the fraction of total cross section @xmath89 or impact parameter @xmath16 @xcite .
the fractional cross section in turn relates the glauber parameters to an observable such as multiplicity @xmath18 .
mean participant pathlength @xmath19 is the prefered centrality measure for the gls reference .
the two - component model for a - a spectra is then formulated by analogy with p - p @xmath90 dependence [ specaa ] & = & s_nn(y_t ) + h_aa(y_t , ) , where the soft component is by hypothesis independent of centrality , and all centrality dependence deviating from the gls is absorbed into hard component @xmath91 @xcite .
the glauber linear - superposition ( gls ) model corresponds to @xmath92 in eq .
( [ specaa ] ) .
[ spsspec ] ( left panel ) shows a pion spectrum ( @xmath93 ) on @xmath94 ( solid points ) for sps fixed - target 0 - 2% central collisions at 200a gev ( @xmath95 gev ) @xcite .
a hydro model applied to those data was used to infer radial flow with mean transverse speed @xmath96 @xcite .
the curve labeled @xmath48 is the soft component for @xmath97 200 gev nsd p - p collisions ( from @xmath75 in fig .
[ ppspec ] ) transformed to @xmath94 .
the line labeled m - b is the maxwell - boltzmann limiting case of @xmath48 . for comparison , a @xmath98 spectrum from p - p collisions at 158a gev ( @xmath99 gev )
is included ( open circles ) @xcite .
small differences at larger @xmath94 may be due to a combination of initial - state @xmath100 effects and semihard parton scattering @xcite .
is there significant radial flow in p - p collisions , or is the hydro model not relevant to s - s data ?
left panel : @xmath94 spectra from 0 - 2% central s - s collisions at 200a gev ( solid points , @xmath95 gev ) @xcite and from p - p collisions at @xmath101 gev ( open circles ) @xcite .
the dashed curve is the lvy soft component from 200 gev p - p collisions .
the dash - dotted curve is the maxwell - boltzmann limiting case of @xmath48 .
right panel : @xmath94 spectrum ( points ) from @xmath2-@xmath3 collisions at @xmath42 gev @xcite .
curves are duplicated from the left panel .
the data are normalized as indicated in the axis label to match the @xmath48 curve at larger @xmath94 . , title="fig:",width=158,height=158 ] left panel : @xmath94 spectra from 0 - 2% central s - s collisions at 200a gev ( solid points , @xmath95 gev ) @xcite and from p - p collisions at @xmath101 gev ( open circles ) @xcite .
the dashed curve is the lvy soft component from 200 gev p - p collisions .
the dash - dotted curve is the maxwell - boltzmann limiting case of @xmath48 .
right panel : @xmath94 spectrum ( points ) from @xmath2-@xmath3 collisions at @xmath42 gev @xcite .
curves are duplicated from the left panel .
the data are normalized as indicated in the axis label to match the @xmath48 curve at larger @xmath94 .
, title="fig:",width=158,height=158 ] fig .
[ spsspec ] ( right panel ) shows an @xmath94 spectrum ( solid dots ) from lep @xmath2-@xmath3 ( e - e ) collisions at @xmath42 gev @xcite .
the spectrum is derived from a sphericity analysis of @xmath102-@xmath103 dijets and estimates the fragment momentum distribution transverse to the @xmath102-@xmath103 axis .
lep @xmath104 data have been normalized as indicated to match p - p soft component @xmath48 ( dashed curve ) , and hence the sps s - s spectrum .
the shape of the @xmath94 spectrum from 91 gev lep ffs is thus consistent with the soft component of p - p spectra at 200 gev and with the full spectrum from central s - s collisions at 19 gev .
the commonality of the soft - component spectrum shape ( lvy distribution ) across energies and collision systems suggests that the soft component is a universal feature of fragmentation whatever the leading particle parton or hadron .
it is unlikely that hydro expansion plays a role in lep @xmath2-@xmath3 collisions .
thus , inference of a radial flow velocity from the same ( lvy ) spectrum shape in a - a collisions is questionable .
figure [ pphard ] ( left panel ) shows spectrum hard - component data from 200 gev nsd p - p collisions in the form @xmath105\ , h_{pp}$ ] for ten multiplicity @xmath90 classes @xcite .
the normalization corresponds to nsd value @xmath106 in one unit of @xmath81 . the common shape , well - described by a gaussian with exponential tail ( dash - dotted curve ) , is plotted as @xmath76 in fig .
[ ppspec ] ( right panel ) .
the exponential on @xmath13 represents qcd power law @xmath107 @xcite .
left panel : hard components ( solid curves and points ) from ten multiplicity classes of 200 gev p - p collisions @xcite .
the data are normalized to non - single diffractive ( nsd ) p - p collisions ( observed @xmath108 in one unit of @xmath81 ) .
the dash - dotted curve is gaussian+exponential tail reference @xmath76 .
right panel : p - p hard component ( points ) averaged over several @xmath90 classes and parton fragment distribution ( fd , solid curve ) obtained from a pqcd folding integral .
dotted curves correspond to @xmath109% shifts in the @xmath110 gev parton spectrum endpoint . , title="fig:",width=158,height=158 ] left panel : hard components ( solid curves and points ) from ten multiplicity classes of 200 gev p - p collisions @xcite .
the data are normalized to non - single diffractive ( nsd ) p - p collisions ( observed @xmath108 in one unit of @xmath81 ) .
the dash - dotted curve is gaussian+exponential tail reference @xmath76 .
right panel : p - p hard component ( points ) averaged over several @xmath90 classes and parton fragment distribution ( fd , solid curve ) obtained from a pqcd folding integral .
dotted curves correspond to @xmath109% shifts in the @xmath110 gev parton spectrum endpoint .
, title="fig:",width=158,height=158 ] figure [ pphard ] ( right panel ) shows p - p hard - component data averaged over several multiplicity classes ( solid points ) compared to a pqcd _ fragment distribution _ ( fd , solid curve ) and the gaussian - plus - tail reference ( dash - dotted curve ) .
the fd was calculated by combining p - p fragmentation functions @xcite with a power - law parton spectrum ( fig .
[ fds ] , left panel , solid curve ) in pqcd folding integral eq .
( [ fold ] ) @xcite .
this comparison confirms that the spectrum hard component is a minimum - bias fragment distribution ( minijets ) .
the hadron spectrum data determine the parton spectrum lower - cutoff energy to 5% . in a - a collisions hard component @xmath50 evolves dramatically , revealing suppression at larger @xmath6 and much larger enhancement at smaller @xmath6 @xcite ( and cf .
[ fdaa ] ) .
the soft+hard two - component decomposition applies also to two - particle correlations on angular difference variables @xmath111 ( e.g. , @xmath112 ) and transverse rapidity @xmath113 @xcite . in p - p collisions the soft component
is confined to unlike - sign pairs peaked at the origin on @xmath114 and nearly back - to - back on azimuth , the structure being consistent with longitudinal ( projectile nucleon ) fragmentation to charged - hadron pairs .
the hard - component correlation structure has the properties expected for ( mini)jets : a same - side ( ss , @xmath115 ) 2d peak at the angular origin and an away - side ( as , @xmath116 ) ridge corresponding to back - to - back ( @xmath117 ) parton scattering . on @xmath113 the hard - component peak mode at ( 2.7,2.7 ) ( @xmath118 gev / c ) is consistent with the single - particle spectrum hard component in fig .
[ pphard ] .
p - p structure agrees semi - quantitatively with pythia correlations @xcite .
in au - au collisions the correlation soft component decreases rapidly to zero with increasing collision centrality .
hard - component correlations change dramatically : both angular and @xmath113 correlations follow the gls reference up to a particular centrality , then develop large deviations from gls ( increases ) for more - central au - au collisions @xcite which correspond to a sharp transition at the same centrality for the spectrum hard component @xcite .
the increased jet structure contradicts triggered dihadron correlations which imply strong jet suppression ( parton thermalization ) ( cf .
[ trigger ] ) @xcite .
the term `` minijet '' refers experimentally to structure in hadron spectra and correlations produced by fragmentation from the _ minimum - bias _ energy spectrum of small-@xmath119 partons ( mainly gluons ) scattering to large angles near mid - rapidity .
no @xmath6 condition is imposed on detected hadron fragments .
the inferred parton spectrum is dominated by partons near 3 gev ( @xmath120 gev ) , each parton on average fragmenting to two charged hadrons with @xmath1211 gev / c .
parton spectrum structure and corresponding _ fragment distributions _ ( fds ) in hadron spectra and correlations are described in @xcite .
the minijet concept originated with an analysis of the spectrum of event - wise @xmath122 clusters down to 5 gev in calorimeter data from 200 gev p - p collisions @xcite .
pqcd analysis of the ua1 ( mini)jet spectrum assumed a parton spectrum cutoff near 3 gev to obtain a jet total cross section of 2 - 3 mb .
. a lower estimate of the cutoff for rhic a - a collisions ( 1 gev ) has been based on saturation - scale - model ( ssm ) arguments @xcite .
minijets were assumed by some theorists to be unresolvable in the a - a final state due to thermalization , but should contribute a substantial fraction of hadron and @xmath123 production in more - central au - au collisions @xcite . in a hydro theory context thermalized minijets are expected to provide the initial energy density required to drive hydrodynamic expansion @xcite . in effect , almost all particle , @xmath6 and @xmath122 production in the final state of central au - au collisions is attributed to thermalized minijets .
other descriptions of minijets were intended to distinguish true qgp manifestations in rhic collisions from pqcd processes .
minijets were modeled with the hijing monte carlo @xcite .
the default hijing parton spectrum cutoff is @xmath124 gev / c .
hijing without `` jet quenching '' is a glauber linear - superposition ( gls ) reference which provides a semi - quantitative description of observed minijet phenomenology in p - p and peripheral au - au collisions down to @xmath121 0.1 gev / c hadron momentum .
systematic studies of minijets in p - p and au - au collisions at rhic @xcite are inconsistent with theoretical assumptions of minijet thermalization .
spectra and correlations provide substantial evidence that back - to - back minijets are actually resolved in the final state of a - a collisions , are not `` thermalized '' even in central au - au collisions , although parton scattering and fragmentation are strongly modified .
novel techniques leading to that conclusion include accurate centrality measurement from n - n to @xmath125 au - au @xcite , differential two - component analysis of p - p and a - a spectra @xcite and 2d angular autocorrelations @xcite .
application of the two - component spectrum model to au - au collisions reveals soft+hard structure similar to p - p collisions @xcite .
the hard - component abundance relative to participant pairs increases at least as fast as @xmath126 ( mean participant path length ) representing the n - n binary - collision scaling expected for parton scattering and fragmentation .
hard - component centrality dependence near 0.5 gev / c ( @xmath55 ) ( large increase ) closely corresponds to that near 10 gev / c ( @xmath127 ) ( smaller decrease ) associated with parton energy loss or `` jet quenching , '' consistent with the hard component being a minimum - bias parton fragment distribution extending down as far as 0.1 gev / c @xcite .
angular correlations identified as minijets increase dramatically in more - central au - au collisions , in contrast to strong `` jet quenching '' inferred from @xmath51 and parton thermalization inferred from triggered dihadron correlations .
the close correspondence of trends for minijet correlations and spectrum hard component is reasonable since sp spectra include the marginal projections of jet correlations . observed minijets reveal the underlying parton spectrum and the `` initial state '' of a - a collisions .
comparison of spectrum hard components with calculated fds indicates that the parton spectrum cutoff is much higher ( 3 gev ) than what is inferred from saturation - scale arguments that accommodate hydro ( 1 gev ) .
survival of almost all minijets to hadronic decoupling contradicts hydro expectations . the gls reference and ideal hydro in an opaque medium are limiting cases of a - a collisions which define a metric between perfect transparency and complete opacity to partons .
where do a - a collisions of given centrality and energy lie on that axis ?
what is the _ quantitative _ deviation from gls transparency ?
minijets act as `` brownian probes '' to answer such questions @xcite .
as an alternative to event - wise jet reconstruction in the high - multiplicity a - a environment trigger / associated @xmath6 cuts are imposed on azimuth correlations to visualize jets combinatorially .
the cuts bias the underlying parton energy spectrum , the sp spectrum of hadron fragments and their angular correlations .
analysis and interpretation of triggered hadron correlations is challenging .
conventional trigger techniques result in substantial underestimation of jet yields @xcite and inference of a dense medium formed in more - central collisions @xcite , with parton multiple scattering , energy loss and production of mach shocks in the medium @xcite .
figure [ zyam1 ] illustrates basic issues for triggered - jet data analysis .
pairs of particles are combined from `` trigger '' and `` associated '' @xmath6 bins , the trigger @xmath6 bin being higher and usually disjoint from the associated bin .
the number of pairs is normalized by the number of trigger particles .
the basic pair distribution then contains a large fraction of uncorrelated combinatoric pairs and a small fraction of correlated pairs from two primary sources : azimuth quadrupole ( `` elliptic flow '' measured by @xmath0 ) and jets .
left panel : simulated `` raw '' ( unsubtracted ) dihadron correlations ( points ) , azimuth quadrupole sinusoid with amplitude @xmath128 ( light dotted curve ) , and corresponding zyam subtracted background ( bold dotted curve ) .
the zyam - estimated background offset is the dotted line .
right panel : result of zyam background subtraction in the left panel ( points and bold curve ) .
dash - dotted and dashed curves are same - side ( ss ) and away - side ( as ) jet peaks input to the simulation .
the minimum at @xmath129 in the as ( away - side ) peak is notable . ,
width=316,height=158 ] conventional analysis combines independent @xmath0 measurements with the zyam ( zero yield at minimum ) convention to separate combinatoric background from jet correlations . in fig .
[ zyam1 ] ( left panel ) the light dotted curve illustrates the azimuth quadrupole amplitude inferred by fitting the entire distribution with the form @xmath130 , with @xmath131 defining @xmath132 . for more - central collisions where the _ non - jet _ quadrupole amplitude is small @xmath133 , where @xmath134 is the amplitude of the same - side ( gaussian ) jet peak @xcite .
jet correlations then dominate @xmath132 .
the combination of @xmath132 and zyam defines the bold dotted curve subtracted from the `` raw '' pair distribution to estimate jet structure .
[ zyam1 ] ( right panel ) shows typical results ( solid curve and points ) for more - central a - a collisions .
the dash - dotted and dashed curves show the ss and as jet peaks input to the simulation .
the dotted sinusoid is the extraneous quadrupole component ( jet - induced @xmath132 ) imposed in the zyam subtraction .
the dotted line indicates the correct offset . from this simulation
it is clear that the conventional background subtraction procedure can reduce the true jet yield by a large fraction .
the as peak , distorted by @xmath0 oversubtraction , acquires a minimum at @xmath129 interpreted in terms of mach shocks or cones .
[ zyam2 ] ( left panel ) shows data from fig . 1 ( upper left ) of @xcite .
zyam - subtracted data from 200 gev central au - au collisions for ( trigger@xmath135associated ) 4 - [email protected] - 4 gev / c @xmath6 cuts are shown as solid points .
corresponding p - p data are shown as open circles .
the bold solid curve is a free fit to the au - au data with offset ( p1 ) , ss gaussian ( amplitude p2 , width p3 ) , as dipole ( p4 ) and quadrupole ( p5 ) .
the resulting offset is the solid line at @xmath136 , the ss gaussian is the dash - dotted curve , the dipole is the dashed curve , and the ( negative ) quadrupole is the dotted curve . the impression can be formed that jet yields in central au - au collisions are comparable to those in p - p collisions because of strong jet quenching in a dense medium .
left panel : zyam - subtracted angular correlations for central 200 gev au - au collisions ( solid points ) @xcite with free fit ( bold solid curve ) of ss gaussian ( dash - dotted curve ) , as dipole ( dashed curve ) and quadrupole ( dotted sinusoid ) .
open points are p - p data relative to zyam zero .
right panel : zyam subtraction reversed , true zero level recovered from free fits to data ( solid points ) , compared to p - p data ( open symbols ) . , title="fig:",width=158 ] left panel : zyam - subtracted angular correlations for central 200 gev au - au collisions ( solid points ) @xcite with free fit ( bold solid curve ) of ss gaussian ( dash - dotted curve ) , as dipole ( dashed curve ) and quadrupole ( dotted sinusoid ) .
open points are p - p data relative to zyam zero .
right panel : zyam subtraction reversed , true zero level recovered from free fits to data ( solid points ) , compared to p - p data ( open symbols ) .
, title="fig:",width=158 ] fig .
[ zyam2 ] ( right panel ) shows a reconstruction of the original ( `` raw '' ) pair distribution prior to zyam subtraction based on @xmath136 and @xmath137 from the left panel .
the relation of p - p to central au - au data is quite different .
both ss and as peaks _ increase by a factor six _ from p - p to central au - au , a large increase in the jet yield which is not apparent in the left panel determined by conventional zyam subtraction . strong suppression of sp hadron spectra at larger @xmath6 is apparently more than compensated by enhancement of jet yields at smaller @xmath6 .
calculations of hydrodynamic evolution in nuclear collisions must specify the `` initial conditions''the thermodynamic state at some initial time , including the energy density , velocity and matter density distributions @xcite . for hydrodynamics to be a valid description of nuclear collisions a large number of initial - state scattered partons must thermalize to produce the energy density required to drive subsequent hydro evolution .
parton spectra extending down to 1 gev and copious rescattering are required .
pqcd can predict the shape of the initial parton energy spectrum , but the _ effective _ lower limit or cutoff of the spectrum , and the extent of subsequent parton thermalization , must be inferred from the hadronic final state and/or supplementary theoretical arguments .
event - wise analysis of ua1 @xmath122 distributions from 200 gev p - p collisions revealed minijet structure down to 5 gev jet energy , later interpreted from calorimeter background estimates as due to partons down to 3 gev @xcite .
pqcd analyses which assumed a minijet spectrum cutoff near 3 gev estimated a jet cross section @xmath138 mb .
a 3-gev cutoff is consistent with the observed p - p @xmath6 spectrum hard - component at 200 gev @xcite .
estimates of the spectrum cutoff for rhic a - a collisions have been based on saturation - scale model ( ssm ) arguments @xcite . given a power - law form of the parton spectrum nearly all scattered partons emerge at the cutoff energy .
parton - related physics is then dominated by that ( saturation - scale ) energy .
the ssm argument maintains that parton scattering is limited only by the _ initial flux of partons _ from nucleon / nucleus projectiles .
the parton spectrum cutoff is then determined by ( gluon ) saturation in the parton distribution functions ( pdfs ) of the colliding nuclei ( or nucleons ) .
ssm arguments lead to cutoff energy 1 gev at @xmath139 gev @xcite .
the ssm - based cutoff implies large parton phase - space densities which could be consistent with secondary parton scattering and partial thermalization @xcite .
but the predicted hadron multiplicity and @xmath6 systematics are inconsistent with measured hadron spectra and correlations . and ,
the ssm argument that only initial - state saturation limits parton scattering may be incomplete .
hadronization ( hadron density of final states ) may be the determining factor in the parton spectrum cutoff , as observed in p - p collisions @xcite .
figure [ fds ] ( left panel ) shows parton power - law spectra with cutoff near 3 gev ( @xmath140 ) derived from comparisons with p - p and au - au spectra ( solid and dash - dotted curves respectively ) @xcite .
the bold dotted curve is an _ ab initio _ pqcd calculation extending down to 1 gev @xcite .
the absolute magnitudes agree well near 3 gev , but large differences in parton and hadron yields arise from the different cutoffs .
left : dijet ( parton - pair ) transverse energy spectra on rapidity @xmath141 .
the solid curve is determined by a measured p - p spectrum hard component .
the dash - dotted curve illustrates reduction of the cutoff energy inferred for central au - au collisions . the bold - dotted theory curve labeled pqcd was derived from @xcite . the light dotted extrapolation to 1
gev corresponds to a saturation - scale cutoff estimate @xcite .
right : ( color online ) argument of the pqcd folding integral eq .
( [ fold ] ) on @xmath142 based on p - p fragmentation functions .
, title="fig:",width=158,height=168 ] left : dijet ( parton - pair ) transverse energy spectra on rapidity @xmath141 .
the solid curve is determined by a measured p - p spectrum hard component .
the dash - dotted curve illustrates reduction of the cutoff energy inferred for central au - au collisions . the bold - dotted theory curve labeled pqcd was derived from @xcite . the light dotted extrapolation to 1 gev corresponds to a saturation - scale cutoff estimate @xcite .
right : ( color online ) argument of the pqcd folding integral eq .
( [ fold ] ) on @xmath142 based on p - p fragmentation functions . , title="fig:",width=158,height=170 ] an alternative estimation of initial conditions can be obtained by direct comparison of calculated _ fragment distributions _ ( fds ) with spectrum hard components .
the fd is defined by the pqcd folding integral @xcite [ fold ] & = & _ 0^ dy_max d_(y , y_max ) , where @xmath143 is the fd describing a spectrum hard component , @xmath144 is the ff ensemble from collision system xx , @xmath145 is the efficiency for detecting jet fragments in @xmath81 acceptance @xmath146 relative to @xmath147 acceptance @xmath148 , and @xmath149 is the dijet spectrum on parton rapidity .
figure [ fds ] ( right panel ) shows the integrand of the folding integral for p - p collisions .
the p - p ffs are based on a parameterization of @xmath2-@xmath3 ffs with lower limit @xmath150 @xcite , but measured ffs from p - p collisions @xcite require a higher cutoff at @xmath1511.5 represented by the horizontal dotted line .
the parton spectrum is the solid curve in the left panel .
the lhs of eq
. [ fold ] is the solid curve in fig .
[ pphard ] ( right panel ) .
the comparison with p - p hard component ( solid points ) determines the 3 gev parton spectrum cutoff to about 5% .
the dotted curves represent @xmath109% variations in the cutoff energy .
( color online ) measured spectrum hard components @xmath50 for five centralities from 200
gev au - au collisions ( bold curves of several colors ) @xcite and 200 gev nsd p - p collisions ( points ) @xcite compared to calculated fds for several conditions ( vacuum , medium , @xmath2-@xmath3 , p - p ) .
the hatched region at upper left estimates the uncertainty due to the soft - component @xmath48 subtraction common to all centralities .
, width=316,height=316 ] figure [ fdaa ] illustrates extension of the fragment distribution comparison to a - a collisions .
the data are spectrum hard components from five centralities of 200 gev au - au collisions @xcite . the bold dashed curve through p - p ( points ) and peripheral au - au ( 60 - 80% ) data is the solid curve from fig . [ pphard ] ( right panel ) corresponding to a parton spectrum cutoff of @xmath152 gev and @xmath153 .
bold dotted curve @xmath154 is a reference corresponding to _ medium - modified _
@xmath2-@xmath3 ffs with @xmath155 reduced to 0.35 and parton spectrum cutoff reduced to 2.7 gev .
the dashed curve labeled 5 passing through central ( 0 - 12% ) data corresponds to the same conditions but with @xmath1561.2 .
the spectrum hard component , even in more - central au - au collisions , reveals the parton spectrum cutoff unambiguously .
the cutoff for central au - au collisions is reduced by about 10% compared to the 3 gev observed for p - p collisions , resulting in a 50% increase in the dijet cross section because of the steep power - law spectrum @xcite .
the relation between final - state hadron and @xmath123 production and the initial partonic state falls between two extremes : 1 ) a two - component system with fragmenting free partons and parton spectrum with cutoff at 3 gev ( gls ) and 2 ) a single component of thermalized partons from a spectrum with cutoff near 1 gev ( ssm ) transitioning to hadrons . the total particle , @xmath6 or @xmath122 production density is less informative than what is produced _ per participant nucleon pair _ in one unit of @xmath81
. how are n - n collisions in a - a different from isolated p - p collisions ?
\1 ) gls in 200 gev nsd p - p collisions the charged - hadron yield is @xmath157 2.5 , with hard - component contribution 0.02 derived from the @xmath90 dependence @xcite .
the corresponding @xmath158 gev / c . in central 200 gev au - au collisions
the _ per - participant - pair _ hard - component yield increases relative to n - n collisions by a factor 54 = 6 ( mean number of n - n collisions per projectile nucleon ) @xmath135 3 ( increase in dijet multiplicity ) @xmath135 1.5 ( increase in minijet cross section ) @xmath135 2 ( increase in jet efficiency for many jets per collision ) @xcite .
the charged multiplicity per participant pair is then @xmath159 ( consistent with data from au - au central collisions ) , and the total-@xmath6 density is @xmath160 gev / c since @xmath11 for the hard component is reduced from 1 to @xmath161 gev / c in central au - au collisions due to medium modification of ffs @xcite .
the estimates assume that the soft component remains unchanged with a - a centrality .
there is no evidence to the contrary .
\2 ) ssm an example of the single - component approach estimates @xmath162 in central au - au collisions assuming a parton spectrum cutoff at 1 gev @xcite .
that parton ( jet ) density implies @xmath163 gev / c assuming 2/3 of jet fragments are charged ( pions ) and there are 191 participant pairs in central au - au collisions .
the @xmath6 density is twice what is observed in data , and the integrated dijet cross section is greater than the n - n total cross section . in another approach
nlo first moment @xmath164 in one unit of rapidity for central au - au collisions is estimated to be 8 and 120 mb - gev for parton spectrum cutoffs @xmath165 and 1 gev respectively @xcite . for @xmath166
@xcite we obtain @xmath167 and 21 gev for @xmath165 and 1 gev respectively .
a significant ( tens of percent ) reduction of those estimates would result from excluding the @xmath168 region .
lo estimates are @xmath52 times smaller .
such estimates are typically compared with the entire final state , not a separate ( semi)hard component explicitly identified with large - angle parton scattering .
data are consistent with @xmath169 gev / c for charged particles from observed minijets . aside from the excess @xmath123 compared with data the single - component approach ignores soft - component contributions to the final state in more - central a - a collisions without explaining what process causes the change .
peripheral au - au collisions follow the gls reference , including contributions from soft and hard components @xcite . at what point
does the dominant soft component disappear and its final - state contribution revert to scattered partons ?
equivalently , what happens to the pdfs of projectile nucleons ?
a hydro description of nuclear collisions requires that radial flow be observed in @xmath6 spectra as a manifestation of large thermalized energy density and corresponding radial pressure gradients @xcite .
evidence for radial flow in the final state is sought via blast wave ( bw ) fits to sp spectra @xcite . figure [ ppbw ] illustrates attempts to describe p - p @xmath6 spectra with the blast - wave model ( assuming hydro expansion in elementary collisions ) @xcite . in the left panel
a simple procedure is illustrated to generate a bw model spectrum .
a maxwell - boltzmann ( mb ) spectrum @xmath170 with slope parameter @xmath171 gev transformed to @xmath13 with @xmath172 is boosted ( blue shifted ) by varying amounts @xmath173 according to hubble expansion ( boost proportional to radius ) , with maximum boost @xmath174 .
the resulting bw spectrum corresponds to mean transverse speed @xmath96 sometimes attributed to p - p collisions @xcite .
left panel : maxwell - boltzmann ( mb , exponential on @xmath94 ) distributions on transverse rapidity @xmath13 for a linear distribution of boosts @xmath175 representing the blast - wave ( bw ) model , with @xmath176 and @xmath171 gev .
right panel : two - component model of the 200 gev nsd p - p @xmath13 spectrum ( dash - dotted curve ) , soft component @xmath48 ( dashed curve ) and hard component @xmath49 ( light solid curve ) .
the bold solid curve is the bw model obtained by averaging the left panel on @xmath175 .
the dotted curve is the mb limit of the bw for @xmath177 . , title="fig:",width=158,height=158 ] left panel : maxwell - boltzmann ( mb , exponential on @xmath94 ) distributions on transverse rapidity @xmath13 for a linear distribution of boosts @xmath175 representing the blast - wave ( bw ) model , with @xmath176 and @xmath171 gev .
right panel : two - component model of the 200 gev nsd p - p @xmath13 spectrum ( dash - dotted curve ) , soft component @xmath48 ( dashed curve ) and hard component @xmath49 ( light solid curve ) .
the bold solid curve is the bw model obtained by averaging the left panel on @xmath175 .
the dotted curve is the mb limit of the bw for @xmath177 .
, title="fig:",width=158,height=158 ] figure [ ppbw ] ( right panel ) compares the resulting bw model ( bold solid curve ) with the two - component representation of 200 gev p - p data ( dash - dotted curve ) described in sec .
[ ppspecmod ] , with soft component @xmath48 ( dashed curve ) and hard component @xmath178 ( light solid curve ) .
bw fits are typically restricted to a `` hydro '' @xmath13 interval bounded above by @xmath179 gev / c ( @xmath58 , vertical dashed line ) .
bw parameter @xmath180 is the same two - component value 0.145 gev determined in @xcite .
@xmath181 is the mean `` radial flow '' value obtained from bw fits to p - p spectra @xcite .
although the bw model appears to describe p - p spectra _ within a restricted @xmath182 interval _ it can not describe a more - extended @xmath13 interval because the curvature is much larger than typical data trends .
there is also no absolute - yield constraint to bw fits .
in contrast , the two - component spectrum model includes a gls reference which specifies expected absolute magnitudes for p - p @xmath18 dependence and all a - a centralities in the event of n - n linear superposition and binary - collision scaling .
this comparison indicates that for p - p spectra the bw model accommodates the lvy form of soft component @xmath183 which describes the _ zero - density limit _ of p - p collisions where radial flow would be least likely .
an early application of the bw model to s - s spectra @xcite is discussed in sec .
[ ssspectra ] .
the s - s result is identical to the 200 gev p - p case because the spectrum is well - described by @xmath48 for p - p collisions .
s - s collisions at @xmath95 gev appear to be well - described by a 200 gev gls reference with negligible hard component . the bw model has been applied more recently to rhic au - au spectra to infer systematic variation of transverse speed @xmath184 and kinetic decoupling temperature @xmath185 @xcite . left panel : similar to fig .
[ ppbw ] ( right panel ) but for 200 gev central au - au collisions with @xmath186 .
the bw parameters for this case are @xmath187 and @xmath188 gev .
right panel : @xmath189 data for 62 and 200 gev au - au collisions @xcite vs centrality measure @xmath25 .
construction of lines and curves is described in the text .
the hatched region locates a sharp transition in jet correlation characteristics in 200 gev au - au collisions . to the left of the transition jet correlations and yields
follow binary - collision scaling of p - p collisions . to the right jet yields
increase dramatically @xcite . , title="fig:",width=158,height=158 ] left panel : similar to fig .
[ ppbw ] ( right panel ) but for 200 gev central au - au collisions with @xmath186 .
the bw parameters for this case are @xmath187 and @xmath188 gev .
right panel : @xmath189 data for 62 and 200 gev au - au collisions @xcite vs centrality measure @xmath25 .
construction of lines and curves is described in the text .
the hatched region locates a sharp transition in jet correlation characteristics in 200 gev au - au collisions . to the left of the transition jet correlations and yields
follow binary - collision scaling of p - p collisions . to the right jet yields
increase dramatically @xcite .
, title="fig:",width=158,height=161 ] figure [ aabw ] ( left panel ) shows the two - component reference ( @xmath186 , dash - dotted curve ) for central au - au collisions at 200 gev and corresponding fixed soft component @xmath48 .
although spectrum data deviate substantially from the two - component reference at larger @xmath13 comparisons with hydro near @xmath190 gev / c ( @xmath70 ) are meaningful .
the bold solid curve is the bw model with @xmath191 and @xmath192 gev , values corresponding to typical evolution of fitted bw parameters with hi collision centrality as shown in fig .
[ aabw ] ( right panel ) . within the restricted @xmath13 interval conventionally assigned to hydro
( left of the dashed line ) the bw model again seems to describe the data well , but fails elsewhere .
the bw is determined for central au - au collisions primarily by the spectrum hard component ( light solid curve ) attributed to jets .
figure [ aabw ] ( right panel ) shows published @xmath189 ( points ) from bw fits to identified - particle spectra for collisions at 200 gev @xcite and au - au collisions at @xmath193 and 200 gev @xcite plotted on mean participant pathlength @xmath25 @xcite .
the solid curve for 200 gev is constructed as follows .
the dashed line is a fit to the right - most six points .
the dash - dotted line passing through the p - p point @xmath194 represents a glauber linear superposition ( gls ) reference for minijet production .
the solid curve then interpolates the lines .
the hatched region denotes a sharp transition observed in 200 gev _
characteristics from correlation data @xcite and in the au - au spectrum hard component for pions and protons @xcite , visible as the nonuniform trend with gap at large @xmath13 in fig .
[ raa ] ( right panel ) .
the 62 gev dotted curve is similarly constructed and also corresponds to jet correlations .
the bw model appears to describe spectra in the restricted interval @xmath195 $ ] gev / c ( @xmath196 $ ] ) .
comparison with the two - component model shows that evolution of bw parameters @xmath197 from ( 0.25,0.15 gev ) to ( 0.6,0.09 gev ) corresponds to _ at least six - fold increase _ of the hard component with centrality ( cf .
[ aabw ] ) .
bw parameter variations are seductive because they suggests isentropic expansion of a thermalized hadron fluid , with increasing separation between `` chemical '' and `` kinetic '' decoupling temperatures @xcite .
however , the bw competes with qcd processes already identified in elementary collisions and expected to follow binary - collision scaling in a - a collisions .
contrast the bw description of a - a spectra with details of hard - component evolution revealed by @xmath54 in fig .
[ raa ] ( right panel ) .
the `` temperature '' parameter of soft component @xmath48 ( equivalent to @xmath198 ) shows no change with a - a centrality .
this reinterpretation of bw trends in terms of fragmentation systematics ( jet correlations ) illustrates the importance of the two - component fragmentation reference for nuclear collisions . as a rule , qcd mechanisms
should be given first option to describe data before hydro conjectures are imposed .
figure [ pyt ] shows proton spectra from au - au collisions for five centralities at @xmath199 gev ( five solid curves ) @xcite .
the 3d density has the form @xmath200 .
soft- and hard - component reference functions @xmath48 and @xmath49 are defined as asymptotic limits of au - au proton spectrum @xmath25 dependence consistent with a similar analysis of p - p @xmath18 dependence @xcite .
the dashed curves represent gls reference @xmath201 for limiting cases @xmath53 and 6 .
`` jet quenching '' suppression relative to the reference is apparent above @xmath202 ( @xmath203 gev / c ) .
the excess proton yield near @xmath204 ( @xmath205 gev / c ) corresponds to the baryon / meson `` puzzle '' at rhic . because of
their greater mass protons ( baryons ) should be more sensitive to a boosted hadron source ( radial flow ) .
( color online ) proton @xmath13 spectra for five au - au centralities ( solid curves ) @xcite .
the dashed curves are two - component model functions for n - n ( @xmath53 ) and @xmath206 ( @xmath186 ) au - au collisions .
the dash - dotted curve is the n - n hard - component reference function .
the dotted curve labeled @xmath48 is the soft - component model for protons from au - au collisions .
, width=316 ] arrows and labels at the bottom of the figure indicate conventional division of @xmath6 spectra into hydro , reco and pqcd intervals .
the corresponding @xmath6 values are indicated at the upper margin . as noted ,
hydro phenomena are expected to dominate for @xmath207 gev / c ( @xmath208 ) .
however , proton spectra below @xmath70 ( @xmath190 gev / c ) closely follow the gls reference .
near @xmath209 hard component @xmath50 dominates the proton spectrum and follows _ binary - collision scaling _ to the error limits of data .
that trend is not expected if radial flow drives au - au spectra as assumed in the bw model .
if radial flow played a significant role @xmath13 spectra should be boosted increasingly to the right ( positive @xmath66 ) with increasing centrality ( and radial pressure ) , which trend is not observed . in this direct confrontation with hydro ,
proton data reveal that the majority of final - state hadrons does not manifest radial flow in au - au collisions at rhic .
instead , a large fraction of protons emerges from parton fragmentation .
corresponding pion spectra provide similar conclusions @xcite .
just as evidence of radial flow is required for a hydro interpretation of heavy ion collisions , measurement of elliptic flow in non - central collisions ( reflecting the eccentric interaction region ) is also essential for hydro .
the conventional elliptic flow measure is @xmath210\rangle$ ] measured relative to the reaction plane defined by the a - a impact parameter and collision axis @xcite . because the reaction plane is not observable @xmath0 analysis relies on variants of multi - particle azimuth correlations denoted by @xmath211 @xcite . in a hydro description
the transverse - rapidity @xmath13 distribution of a thermalized expanding medium would describe a common boosted source for all hadron species .
radial boost distribution @xmath212 is then equivalent to flow field @xmath213 [ @xmath214 . if the final - state hadron distribution is described by density @xmath215 @xmath6-differential @xmath12 is given by [ v2eq ] v_2(p_t , b ) = , with @xmath79 defined relative to the reaction plane .
@xmath6-integral @xmath216 is obtained by integrating numerator and denominator over @xmath6 .
@xmath216 is compared to eccentricity @xmath217 of the initial a - a configuration space as a test of hydro models .
although the @xmath0 concept appears simple , implementation and interpretation of @xmath0 analysis is not .
strong minijet contributions to azimuth correlations ( called `` nonflow '' ) dominate @xmath0 data obtained with conventional methods over a substantial part of the centrality interval @xcite , making interpretation of @xmath0 data within the hydro context problematic .
the model - neutral term `` azimuth quadrupole '' can be applied to @xmath0 as a measure of azimuth correlations .
quadrupole correlations are indeed observed in rhic data , but although @xmath0 is conventionally interpreted as elliptic flow @xcite physical interpretation of the quadrupole is an open question @xcite .
absence of measurable radial flow in sp spectra contradicts the concept of elliptic flow as a modulation of radial flow in non - central collisions .
the centrality and energy systematics of ratio @xmath220 have been interpreted to indicate monotonic increase with a ( knudsen ) number of secondary collisions ( low - density limit or ldl model ) suggesting approach to thermalization @xcite .
saturation of @xmath221 could indicate a hydro limit .
claims for achievement of the hydro limit @xcite and so - called constituent - quark scaling @xcite have been interpreted as evidence for a thermalized qgp at rhic .
[ ptint ] ( left panel ) shows _ per - particle _ azimuth quadrupole measure @xmath26 / \sqrt{\rho_\text{ref}}$ ] on centrality measure @xmath25 @xcite .
the quadrupole amplitude can be obtained from fits to 2d angular autocorrelations @xcite , and is related to @xmath0 by & & [ v_2\{2d}]^2 defining @xmath222 @xcite .
the definition is compatible with minijet angular correlation measurements and the glauber linear superposition ( gls ) reference @xcite .
left panel : @xmath6-integrated @xmath216 data ( solid points , open triangles ) for two analysis methods from 200 gev au - au collisions vs centrality measure @xmath25 @xcite .
the open squares are @xmath223 data from 17 gev pb - pb collisions @xcite .
the bold solid curve is derived from the line from the right panel .
the dotted curve includes jet correlations ( same - side jet peak ) .
the light solid and dashed curves apply to the 17-gev data .
right panel : data and curves from the left panel divided by @xmath224 .
the solid line is defined by eq .
( [ magic ] ) .
the dash - dotted curve indicates the trend if @xmath225 were used instead .
, title="fig:",width=158,height=158 ] left panel : @xmath6-integrated @xmath216 data ( solid points , open triangles ) for two analysis methods from 200 gev au - au collisions vs centrality measure @xmath25 @xcite .
the open squares are @xmath223 data from 17 gev pb - pb collisions @xcite .
the bold solid curve is derived from the line from the right panel .
the dotted curve includes jet correlations ( same - side jet peak ) .
the light solid and dashed curves apply to the 17-gev data .
right panel : data and curves from the left panel divided by @xmath224 .
the solid line is defined by eq .
( [ magic ] ) .
the dash - dotted curve indicates the trend if @xmath225 were used instead .
, title="fig:",width=158,height=158 ] fig .
[ ptint ] shows data ( points ) for conventional elliptic - flow measures @xmath132 and @xmath226 ( two- and four - particle correlations respectively ) @xcite . the solid curve ( transformed from the straight line in the right panel )
is defined by [ magic ] / & = & 0.0045 r ( ) _ opt^2 n_bin .
@xmath227 is consistent with the energy dependence of @xmath6 correlations attributed to minijets @xcite .
the dotted curve passing through @xmath132 data is obtained by adding to eq .
( [ magic ] ) `` nonflow '' term @xmath228 representing the @xmath229 fourier component of the measured same - side jet peak on azimuth @xcite ( cf .
sec . [ 2compcorr ] ) . for 17-gev pb - pb @xmath223 ( event - plane ) data ( open squares ) @xcite the same fractional relation between `` flow '' ( thin solid curve ) and `` nonflow '' ( dashed curve , with
@xmath230 added ) is observed .
thus , eq . ( [ magic ] ) with nonflow parameter @xmath230 obtained from ( mini)jet measurements describes _ all _ au - au ( or pb - pb ) @xmath0 data above @xmath231 gev .
[ ptint ] ( right panel ) shows the same data in a different format .
the _ optical _ eccentricity @xmath232 leads to a simple , universal linear relation .
( [ magic ] ) defines the solid line .
an alternative monte carlo eccentricity would produce the dash - dotted curve .
arguments supporting the optical eccentricity are given in @xcite .
deviations of the dotted and dashed curves represent the effect of jets on @xmath223 and @xmath132 measurements .
a combination of @xmath0 data , minijet correlations and the optical - glauber description of collision eccentricity reveals that over a large range of energies and centralities @xmath0 data are described accurately and completely by the initial state @xmath233 @xcite .
there is no apparent dependence on subsequent collision evolution , equation of state or degree of thermalization .
there is no correspondence with the recently - observed sharp transition in minijet angular correlations @xcite , calling into question any relation among the azimuth quadrupole , conjectured bulk medium and hydrodynamics .
[ ptdif1 ] ( left panel ) shows @xmath12 data for three hadron species @xcite .
the mass ordering below 2 gev / c is interpreted to imply a hydro phenomenon .
baryon vs meson trends at larger @xmath6 are interpreted to conclude that _ constituent quarks _ coalesce to form hadrons which exhibit elliptic flow , implying a thermalized flowing partonic medium prior to decoupling ( cf .
[ cqscale ] ) . dotted theory curves a and b for pions are from @xcite and @xcite respectively .
the curves passing through data are from @xcite ( and cf .
[ ptdif2 ] ) . left panel : @xmath12 data for three hadron species plotted in the usual format @xcite .
right panel : the same @xmath12 data divided by @xmath6 in the lab frame suggest universality on proper transverse rapidity for each hadron species in particular the correspondence of data near @xmath234 . dotted curves a and b in each panel are viscous hydro predictions from @xcite and @xcite respectively . the three curves through data are derived from this analysis . , title="fig:",width=158,height=158 ] left panel : @xmath12 data for three hadron species plotted in the usual format @xcite .
right panel : the same @xmath12 data divided by @xmath6 in the lab frame suggest universality on proper transverse rapidity for each hadron species in particular the correspondence of data near @xmath234 . dotted curves a and b in each panel are viscous hydro predictions from @xcite and @xcite respectively .
the three curves through data are derived from this analysis . , title="fig:",width=158,height=158 ] fig .
[ ptdif1 ] ( right panel ) shows the same data plotted as ratio @xmath235 vs @xmath236 denoting transverse rapidity ( a velocity measure ) computed with the correct mass for each hadron species .
motivation for the @xmath235 ratio is described in sec .
[ v2ptanalysis ] .
the three hadron species follow a common trend at smaller @xmath13 expected for hadrons emitted from a common boosted source .
the mean boost @xmath237 is approximately the source radial speed , since @xmath238 .
although a hydro mechanism might produce such a boost , these data to not _
require _ a hydro boost mechanism .
theory curves a and b are zero - viscosity limits of two viscous - hydro calculations @xcite .
the differences are most notable in the plot on the right . at small @xmath13 the theory boost distributions , a central element of the hydro model ,
are very different from data .
the hydro calculations are consistent with radial hubble expansion of a fluid , whereas the data are consistent with an expanding cylindrical shell @xcite .
[ ptdif2 ] shows quadrupole spectra ( points ) reconstructed as described in eq .
( [ stuff ] ) and @xcite from the @xmath12 data plotted in fig .
[ ptdif1 ] .
the additional factor @xmath239 relative to fig .
[ ptdif1 ] ( right panel ) is obtained from parametrizations of single - particle identified - hadron spectra in @xcite .
the sp spectrum factor eliminates the extraneous denominator in eq .
( [ v2eq ] ) to reveal the quadrupole spectrum implicit in the numerator .
all aspects of minimum - bias @xmath12 structure are simply and accurately described .
data from fig .
[ ptdif1 ] transformed to proper @xmath13 for each hadron species @xcite .
the dotted curves are soft components from respective single - particle spectra for comparison .
the prominent feature is the common edge at @xmath240 , implying that the quadrupole components for three hadron species originate from a common boosted source .
hadron abundances and quadrupole spectrum shapes are similar to the single - particle spectrum soft components .
, width=316 ] the quadrupole spectra are shifted by common monopole boost @xmath71 ( cf .
the left edges ) .
the spectrum shapes ( solid curves ) have the same lvy distribution _ form _ on @xmath94 that describes the sp soft components ( @xmath241 , dotted curves ) , with the same relative hadron abundances but substantially different parameters .
the `` temperatures '' of the quadrupole spectra are systematically reduced by at least 30% , e.g. to @xmath242 gev from the sp soft - component @xmath243 gev for pions .
the spectrum widths on @xmath13 of quadrupole spectra vary with ratio @xmath244 just as for sp spectra .
the solid curves , transformed to other plotting formats in figs . [ ptdif1 ] and [ consq ] , describe quadrupole data within errors .
quadrupole spectra are reduced in amplitude relative to sp spectra by a common factor @xmath245 .
the reduction factor includes the product of the true momentum eccentricity represented by quadrupole boost @xmath246 and the quadrupole spectrum absolute yield .
arguments in @xcite suggest that @xmath247 and the actual quadrupole yield is a small fraction ( @xmath248% ) of the total multiplicity .
the information derived from @xmath12 data is thus a boost distribution ( narrow relative to the mean , inconsistent with hubble expansion ) , a quadrupole source `` temperature , '' a relative abundance , and distinction between sp spectrum soft component and quadrupole component .
the combination of properties is inconsistent with a thermalized fluid medium as the source for most hadrons from a collision .
hadronization appears to follow the same process for soft and quadrupole spectrum components , but the hadron sources are distinct . the sequence of three panels in figs .
[ ptdif1 ] and [ ptdif2 ] reveals that @xmath12 as defined exaggerates the role of data at larger @xmath6 due to the presence of the single - particle spectrum in its denominator and a kinematic factor @xmath6 in its numerator @xcite .
the most important @xmath0 data from the standpoint of the hydro model are at smaller @xmath6 , especially for low - mass hadrons , as is apparent from fig .
[ ptdif2 ] .
those details are suppressed by the conventional plotting format of fig .
[ ptdif1 ] ( left panel ) .
the same @xmath12 data can be replotted in a format used to demonstrate so - called `` constituent quark scaling '' @xcite .
[ consq ] ( left panel ) shows @xmath12 plotted vs transverse mass @xmath249 .
offsets due to the common source boost at smaller @xmath6 are reduced on @xmath94 , and there may be vertical segregation into mesons and baryons at larger @xmath94 . left panel : @xmath12 data from fig .
[ ptdif1 ] are plotted _ vs _ kinetic energy @xmath250 .
the intercept spacing at small @xmath6 is reduced by a factor 3@xmath135 .
right panel : the left panel , but with constituent quark scaling in the form @xmath251 so that meson trends remain unchanged .
the shift for baryons is indicated by the arrows . , title="fig:",width=158,height=158 ] left panel : @xmath12 data from fig .
[ ptdif1 ] are plotted _ vs _ kinetic energy @xmath250 .
the intercept spacing at small @xmath6 is reduced by a factor 3@xmath135 .
right panel : the left panel , but with constituent quark scaling in the form @xmath251 so that meson trends remain unchanged .
the shift for baryons is indicated by the arrows .
, title="fig:",width=158,height=158 ] fig .
[ consq ] ( right panel ) is said to demonstrate constituent - quark scaling .
the number of valence quarks @xmath252 for hadrons ( 2 or 3 ) is applied to data .
scaling implies that data for all hadrons should lie on a common curve , which seems to be approximately the case below 2 gev / c@xmath253 .
additional factors 2 in this plot are included so that only baryon data shift from left to right panel .
the curves through data from @xcite and fig .
[ ptdif2 ] are transformed appropriately .
the baryon data seem closer to meson data in the right panel within a limited @xmath94 interval .
do those results support constituent - quark scaling ?
the boosted intercept points on @xmath6 in fig . [ ptdif1 ] ( left panel ) fall at @xmath254 , with @xmath255 .
the intercepts scale with hadron mass , and the common boost can thus be read off the plot directly . the corresponding intercepts in fig .
[ consq ] ( left panel ) are @xmath256 \sim m_0\,(\delta y_{t0})^2/2 $ ] .
sensitivity to the common boost has been reduced by a factor @xmath257 , _ minimizing the hydro phenomenon in question _ relative to data errors .
the curves from @xcite , which describe data well , diverge from one another above 2 gev / c@xmath253 and still retain significant spacings at small @xmath94 due to the common source boost . in fig .
[ consq ] ( right panel ) additional factors @xmath258 have proportions 1:1:1.5 for the three hadron species , whereas @xmath12 in all its aspects depends on hadron masses in approximate proportions 1:3.5:7 .
the boost manifestation on momentum , the quadrupole spectrum shapes ( widths ) and the fractional abundances are all determined by hadron masses , most evidently in fig .
[ ptdif2 ] .
since hadron masses do nt scale as the number of valence quarks the constituent - quark scaling hypothesis is falsified by @xmath12 data , despite the appearance of fig .
[ consq ] ( right panel ) .
[ ptdif1 ] ( right panel ) provides the strongest challenge to constituent - quark scaling .
the measured quantities appear as ratios@xmath235 and ratios within @xmath259for which constituent - quark numbers should cancel , and where quark coalescence in a common boosted frame should be most naturally described .
there is no scaling .
the large differences between hadron species are simply explained by mass - dependent effects in the quadrupole spectra of fig .
[ ptdif2 ] .
results from hydro - motivated analysis methods are conventionally interpreted to favor hydro mechanisms and a dense partonic medium .
is formation of an sqgp with anomalously small viscosity _ required _ by such results ?
do other results contradict such conclusions ?
thermalized minijets are expected to provide the large energy densities required to drive hydro expansion in more - central rhic collisions @xcite .
evidence for _ partonic _ hydro expansion would in turn confirm formation of a thermalized qcd medium or qgp arising from parton ( minijet ) multiple scattering .
high-@xmath6 jet tomography would map the resulting `` opaque core '' and determine its properties @xcite .
the opaque core @xcite is said to reduce the sensitivity of ratio measure @xmath51 due to the predominance of `` surface '' jets emitted from a thin `` corona '' layer @xcite . at most 1% of hadrons with up to 14 gev / c momentum from `` quenched jets
'' would survive according to @xcite .
those expectations are strongly contradicted by some data , as demonstrated in this analysis .
the most important challenge to hydro at rhic is the survival of essentially all minijets to the final state in central 200 gev au - au collisions @xcite .
minijet analysis of spectra and correlations reveals that even central au - au collisions are nearly transparent to all scattered partons ( not just `` surface '' partons ) .
_ back - to - back jet correlations are not diminished_. there is no reduction of correlations by parton or hadron rescattering .
hadron fragments down to 0.1 gev / c retain their angular correlations relative to parent partons .
however , parton fragmentation is strongly modified in more - central collisions @xcite .
some arguments for parton thermalization have been based on a conjectured saturation - scale parton spectrum cutoff near 1 gev which would increase the scattered - parton yield and phase - space density 10 - 30 fold ( compared to a 3-gev cutoff ) .
such large densities could insure thermalization by parton multiple scattering , but they seem to be inconsistent with the observed final state . a quantitative relation has been established among parton cross sections , ffs and hadron spectra / correlations @xcite .
comparison of pqcd elements to p - p hadron spectra establishes a parton spectrum cutoff near 3 gev .
peripheral au - au collisions follow a two - component gls reference up to a transition point on centrality . beyond that point spectra and correlations
are still described by parton ffs and a parton spectrum terminating near 3 gev , although there are significant changes to fragmentation .
the relation between soft and hard spectrum components follows pqcd expectations
. in central au - au collisions the soft component ( projectile nucleon fragmentation ) represents at least two - thirds of hadron production .
the other third is accounted for by _ additional _ parton scattering and fragmentation compared to the gls reference , with modified ffs .
@xmath123 production is similarly well described .
essentially all semihard scattered partons appear as correlated hadrons in the final state , contradicting claims of parton thermalization and opaque medium .
radial flow should be the primary manifestation of thermalized initial - state energy densities and pressure gradients .
radial flow inferred from hydro - inspired blast - wave fits to hadron spectra seems to confirm required large energy densities .
but conventional bw models assume that any deviation from a maxwell - boltzmann distribution below 2 - 3 gev / c is caused by radial flow .
the bw spectrum model must compete with the two - component model and contributions from minimum - bias parton fragmentation .
the two - component model coupled with glauber linear superposition describe data features associated with known qcd processes which should _ a priori _ appear in a - a collisions : projectile nucleon fragmentation and large - angle parton scattering and fragmentation consistent with elementary collisions .
the bw model approximates spectra with limited accuracy over a _ limited _
@xmath6 interval and fails qualitatively elsewhere . in effect ,
parton fragmentation processes are misidentified by bw fits and injected into the hydro context .
@xmath0 interpreted as elliptic flow is seen as the most convincing support for hydrodynamics at rhic .
the @xmath0 definition assumes a monolithic flowing medium common to most final - state hadrons .
but quadrupole spectra extracted from published @xmath12 data are quite different from corresponding sp spectra
. only a small fraction of the hadronic final state may actually `` carry '' the azimuth quadrupole @xcite .
sharp transitions in minijet characteristics at specific a - a centralities @xcite could be interpreted to indicate medium modification of parton fragmentation .
data show no evidence of such transitions , in conflict with their interpretation in terms of elliptic flow of a homogeneous thermalized medium .
the concept that elliptic flow results from rescattering ( of partons or hadrons ) in a medium is contradicted by minijet correlation data .
such rescattering would decouple partons from their initial - state scattering partners ( especially at 3 gev ) and destroy hadronic correlations resulting from parton fragmentation .
strong back - to - back parton correlations are inferred from strong hadron jet correlations , even in central au - au collisions .
differential analysis of single - particle spectra and two - particle correlations from rhic au - au collisions reveals that copious parton fragment yields extend down to small transverse momentum even in central collisions . according to pqcd and elementary collisions at least half of all parton fragments in nuclear collisions
should appear below 1 gev / c .
measured hard components of spectra and correlations at rhic are consistent with that expectation .
hard components are identified as fragment distributions from _ minimum - bias _ parton fragmentation or `` minijets '' quantitatively described by pqcd .
observable minijets play a central role in rhic collisions . in the hydro context single - particle spectra
are divided into three regions , with the interval @xmath207 gev / c assigned to hydro phenomena .
but that interval contains the peak of the spectrum hard component and most parton fragments .
the blast - wave model fitted to sp spectra returns an estimate of radial flow .
but the spectrum structure responsible is just the hard component associated with minijets .
spectrum ratio @xmath51 used to estimate jet suppression at larger @xmath6 strongly suppresses spectrum structure below @xmath260 gev / c , including most of the jet fragment yield in the spectrum hard component . conventional @xmath0 analysis , especially in more - peripheral and more - central a - a collisions , misidentifies the @xmath261 fourier component of jet azimuth structure ( `` nonflow '' ) as elliptic flow and can greatly overestimate @xmath0 in such cases .
ratio @xmath12 contains the single - particle spectrum in its denominator , and is thus strongly affected by the spectrum hard component ( minijets ) at larger @xmath6 , in addition to nonflow contributions from jet azimuth correlations in its numerator .
so - called `` triggered '' analysis of jet azimuth correlations suppresses true jet yields because of incorrect estimation of the combinatoric - background offset ( zyam estimate ) and @xmath0 component .
the conventional zyam procedure underestimates jet yields by as much as a factor 10 , and the away - side jet is distorted by the @xmath0 oversubtraction ( minimum at @xmath129 ) so as to suggest the presence of mach shocks .
the combination leads to inference of an `` opaque medium '' and parton thermalization .
the numerator of @xmath12 contains quadrupole spectrum @xmath262 which could represent a hydro phenomenon if that were relevant .
however , the quadrupole component is insensitive to dramatic changes in jet properties with increasing centrality that could be attributed to a common medium .
the quadrupole component seems to be carried by an isolated and small part of the final - state system . in conclusion
, the combination of two fragmentation components ( projectile nucleons and semihard - scattered partons ) plus an isolated third ( quadrupole ) component accurately describes all rhic spectrum and correlation data over a large @xmath6 interval .
no evidence for radial flow is observed .
jet yields increase according to binary - collision scaling in more - peripheral collisions , and even more rapidly for more - central au - au collisions .
fragmentation is strongly modified in the latter case .
there is no evidence for an opaque medium . as a rule , above @xmath263 gev qcd should be given first opportunity to describe nuclear collisions .
hydro should not be invoked until qcd has been shown to fail .
a hydro description of selected aspects of data within restricted kinematic intervals does not establish the necessity of the model . in some cases hydro
is contradicted .
the relevance of hydrodynamics to rhic collisions can therefore be questioned .
k. j. eskola , k. kajantie , p. v. ruuskanen and k. tuominen , nucl . phys .
b * 570 * , 379 ( 2000 ) .
g. c. nayak , a. dumitru , l. d. mclerran and w. greiner , nucl .
a * 687 * , 457 ( 2001 ) . c. r. allton _ et al . _ ,
d * 66 * , 074507 ( 2002 ) .
e. schnedermann , j. sollfrank and u. heinz , phys .
c * 48 * , 2462 ( 1993 ) .
f. becattini , j. manninen and m. gazdzicki , phys .
c * 73 * , 044905 ( 2006 ) .
a. m. poskanzer and s. a. voloshin , phys .
c * 58 * , 1671 ( 1998 ) .
m. gyulassy and l. mclerran , nucl .
a * 750 * , 30 ( 2005 ) .
b. mller , acta phys .
b * 38 * , 3705 ( 2007 ) ; arxiv:0710.3366 .
t. a. trainor , phys .
c * 78 * , 064908 ( 2008 ) . f. cooper , e. mottola and g. c. nayak , phys .
b * 555 * , 181 ( 2003 ) .
r. c. hwa and c. b. yang , phys .
c * 70 * , 024905 ( 2004 ) .
r. j. fries , b. muller , c. nonaka and s. a. bass , phys .
c * 68 * , 044902 ( 2003 ) .
v. greco , c. m. ko and p. levai , phys .
lett . * 90 * , 202302 ( 2003 ) s. albino , b. a. kniehl and g. kramer , nucl .
b * 803 * , 42 ( 2008 ) . c. adler _ et al . _
( star collaboration ) , phys .
c * 66 * , 034904 ( 2002 ) .
j. adams _ et al . _
( star collaboration ) , phys .
lett . * 92 * , 052302 ( 2004 ) .
t. a. trainor , r. j. porter and d. j. prindle , j. phys .
g * 31 * , 809 ( 2005 ) .
j. adams _ et al . _
( star collaboration ) , phys .
lett . * 95 * , 152301 ( 2005 ) .
s. s. adler _ et al .
( phenix collaboration ) , phys .
lett . * 97 * , 052301 ( 2006 ) .
a. adare _ et al .
_ ( phenix collaboration ) , phys .
c * 78 * , 014901 ( 2008 ) . c. albajar
_ et al . _
( ua1 collaboration ) , nucl .
b * 309 * , 405 ( 1988 ) .
t. hirano and m. gyulassy , nucl .
a * 769 * , 71 ( 2006 ) .
k. werner , phys .
lett . * 98 * , 152301 ( 2007 ) .
b. i. abelev _ et al .
_ ( star collaboration ) , phys . rev .
* 102 * , 052302 ( 2009 ) .
j. casalderrey - solana , e. v. shuryak and d. teaney , j. phys .
* 27 * , 22 ( 2005 ) ; nucl .
a * 774 * , 577 ( 2006 ) ; hep - ph/0411315 .
f. retiere and m. a. lisa , phys .
c * 70 * , 044907 ( 2004 ) .
j. adams _ et al . _
( star collaboration ) , phys .
lett . * 92 * , 112301 ( 2004 ) .
b. i. abelev _ et al . _
( star collaboration ) , phys .
c * 79 * , 034909 ( 2009 ) .
s. a. voloshin and a. m. poskanzer , phys .
b * 474 * , 27 ( 2000 ) .
r. a. lacey and a. taranenko , pos c * frnc2006 * , 021 ( 2006 ) .
b. i. abelev _ et al .
_ ( star collaboration ) , phys .
c * 77 * , 054901 ( 2008 ) .
j. adams _ et al . _
( star collaboration ) , phys .
c * 72 * , 014904 ( 2005 ) .
k. j. eskola , h. honkanen , c. a. salgado and u. a. wiedemann , nucl .
a * 747 * , 511 ( 2005 ) .
s. a. bass , c. gale , a. majumder , c. nonaka , g. y. qin , t. renk and j. ruppert , phys .
c * 79 * , 024901 ( 2009 ) .
|
the hydrodynamic ( hydro ) model applied to heavy ion data from the relativistic heavy ion collider ( rhic ) in the form of single - particle spectra and correlations seems to indicate that a dense qcd medium nearly opaque to partons , a strongly - coupled quark - gluon plasma ( sqgp ) , is formed in more - central au - au collisions , and that the sqgp may have a very small viscosity ( `` perfect liquid '' ) .
measurements of radial and elliptic flows , with possible coalescence of `` constituent quarks '' to form hadrons , seem to support the conclusion . however , other measurements provide contradictory evidence .
unbiased angular correlations indicate that a large number of back - to - back jets from initial - state scattered partons with energies as low as 3 gev survive as `` minijet '' hadron correlations even in central au - au collisions , suggesting near transparency .
two - component analysis of single - particle hadron spectra reveals a corresponding spectrum hard component ( parton fragment distribution described by pqcd ) which can masquerade as `` radial flow '' in some spectrum analysis .
reinterpretation of `` elliptic flow '' as a qcd scattering process resulting in fragmentation is also possible . in this paper
i review analysis methods and results in the context of two paradigms : the conventional hydrodynamics / hard - probes paradigm and an alternative quadrupole / minijets paradigm .
based on re - interpretation of fiducial data i argue that hydrodynamics may not be relevant to rhic collisions .
collision evolution may be dominated by parton scattering and fragmentation , albeit the fragmentation process is strongly modified in more - central a - a collisions .
|
aspects of few - body phenomena arise in many areas of physics . for example
, experiments using cold atomic gases can study few - body loss phenomena within a many - body background @xcite , directly trap small clusters in a single microtrap @xcite , see the transition to the mott insulating phase of an optical lattice @xcite , and ions can be held and studied for long periods of time @xcite .
another area includes the formation of exotic `` molecules '' from a combination of matter and antimatter .
more specifically , polyelectronic clusters have been experimentally verified such as ps @xcite , ps@xmath7 @xcite , and ps@xmath8 @xcite . on the other hand
, a recent study suggests that tripositronium is not bound @xcite . since wheeler s prediction of dipositronium ps@xmath8 in 1946 @xcite
, there has remained active theoretical interest in electron - positron clusters @xcite .
this interest remains partly due to the possibility of a bose - einstein condensate of ps @xcite .
a deeper understanding , however , requires the ability to study ever larger clusters .
one could simply use ever increasingly available computational power , or find breakthroughs in improving older techniques .
one such technique is the stochastic variational method using correlated gaussians , both the traditional approach @xcite , and that carried out at fixed hyperradius @xcite .
the latter method involves a two - step process to diagonalize the full hamiltonian .
first , after expressing the schrdinger equation using a single length , the hyperradius @xmath0 , and the remaining degrees of freedom as hyperangles using hyperspherical coordinates @xcite , the hyperangular schrdinger equation is solved parametrically in @xmath0 .
this leads to an infinite set of coupled `` born - oppenheimer '' potentials .
second , the set of one - dimensional differential equations in @xmath0 is solved .
once the fixed-@xmath0 hyperspherical potential curves and nonadiabatic couplings have been determined for any few - body system , it becomes relatively straightforward to compute complex scattering processes such as rearrangement and n - body recombination @xcite .
earlier studies involved 3- and 4-body systems with vanishing angular momentum @xmath3 and positive parity @xmath4 @xcite , and recent developments have extended the method to systems of finite angular momentum and different parities @xcite .
the present study further extends the correlated gaussian hyperspherical technique , developing a new approach for the calculation of matrix elements at fixed hyperradius .
we find that the matrix elements of the noninteracting system in an even number of dimensions can be evaluated in closed analytical form .
for odd dimensions , or for general interaction potentials in any dimension , the matrix elements reduce to a single numerical integration regardless of system size .
though this work focuses on states with @xmath9 symmetry , the technique presented here is straightforward to apply to other symmetry states . to demonstrate the technique ,
the adiabatic potential curves are calculated for particles interacting via @xmath10 potentials . in three dimensions
, this corresponds to polyelectron systems .
the remainder of the paper is organized as follows .
section [ sec_se ] introduces the hyperradial schrdinger equation and the problem to be solved .
section [ sec_matrix_elements ] describes the technique to calculate matrix elements with a fixed hyperradius @xmath0 using a basis of correlated gaussians . as an example
, the overlap between two basis functions is calculated in detail .
section [ sec_results ] uses the technique described in sec .
[ sec_matrix_elements ] to calculate the adiabatic potential curves for equal - mass few - body systems interacting via the @xmath10 power law potential .
discussion of future work and conclusions is given in sec .
[ sec_conclusion ] . for completeness ,
the appendices detail the calculation of the hyperangular kinetic energy and central potential matrix elements using the spherical gaussian basis functions .
consider the @xmath1-particle hamiltonian @xmath11 where each particle has @xmath2 degrees of freedom .
using @xmath12 mass - scaled jacobi vectors @xmath13 , @xmath14 , the center of mass @xmath15 and relative @xmath16 parts of the hamiltonian separate , @xmath17 our interest centers on the relative hamiltonian , @xmath18 where @xmath19 is the kinetic energy of the @xmath20 relative jacobi vectors and @xmath21 contains the interparticle interactions .
all jacobi vectors are scaled such that they are analogous to @xmath20 equal - mass `` particles '' of mass @xmath22 , @xmath23 where @xmath24 is the mass of the @xmath25 particle .
our choice of @xmath22 ensures that the coordinate transformation is unitary .
the relative hamiltonian @xmath16 is recast in hyperspherical coordinates in terms of @xmath26 hyperangles denoted by @xmath27 and a single length , the hyperradius @xmath0 .
the relative hamiltonian is then a sum of the hyperradial kinetic energy @xmath28 , the hyperangular kinetic energy @xmath29 , and the interaction potential , @xmath30 where @xmath31 the exact form of the hyperangular kinetic energy @xmath29 depends on the choice of jacobi vectors , but for this work the exact form is not needed . the solution @xmath32 to eq .
is expanded in terms of the radial functions @xmath33 and the channel functions @xmath34 , @xmath35 the channel functions at a fixed hyperradius @xmath0 form a complete orthonormal set over the hyperangles , @xmath36 and are the solutions to the adiabatic hamiltonian @xmath37 , @xmath38 where @xmath39 after applying eq . on the expansion eq . and projecting from the left onto the channel functions , the schrdinger equation reads @xmath40 the hyperspherical schrdinger equation eq
is solved in a two step procedure .
first , @xmath37 is solved parametrically in @xmath0 for the adiabatic potential curves @xmath41 . in a second step , the coupled set of one - dimensional equations in @xmath0
are solved . in eq .
, @xmath42 represents the coupling between channels , @xmath43 where @xmath44 and @xmath45 the brackets indicate that the integrals are taken only over the hyperangle @xmath27 with the hyperradius @xmath0 held fixed .
this paper concentrates on the first step , that is , on the solutions to the adiabatic hamiltonian , eq . .
the rest of this paper discusses how to calculate the adiabatic potential curves @xmath41 by expanding the channel functions @xmath34 using a basis of correlated gaussians .
the eigenfunctions @xmath34 of @xmath37 are expanded using a non - orthogonal basis of correlated gaussians @xcite , @xmath46 is a symmetrization operator that permutes identical particles .
equation could also include a spinor [ see sec .
[ sec_results ] ] , but for simplicity , this section focuses on the spatial part of the unsymmetrized basis functions .
the spherical gaussian part @xmath47 of the basis functions with @xmath9 symmetry are @xmath48 is an array of jacobi vectors , @xmath49 .
all jacobi vectors exist in @xmath2 dimensions , such that the @xmath25 jacobi vector reads @xmath50 .
@xmath51 is an @xmath52 symmetric positive definite coefficient matrix that describe the correlations .
the matrix @xmath51 contains @xmath53 independent variational parameters .
the following works out in detail the overlap matrix element between two generic unsymmetrized basis functions , dropping the superscript @xmath54 in favor of using @xmath55 and @xmath56 to describe two distinct basis functions .
this illustrates our main method applicable in general to other matrix elements e.g. the hyperangular kinetic energy and central potentials ( see appendices [ app_ke ] and [ app_central ] ) .
the overlap matrix element @xmath57 is given by @xmath58 \bm{x}\right ) 2r \times \nonumber \\ & \delta^{(1 ) } \ !
\left ( x_1 ^ 2+x_2 ^ 2+\ldots+x_n^2 - r^2 \right ) d^{nd } \bm{x}.\end{aligned}\ ] ] the hyperradius @xmath0 is fixed by introducing a dirac delta function in the square of the hyperradius where @xmath59 ( recall that the square of the hyperradius is given by the sum of scalar products of the jacobi vectors ) .
equation is an integral over @xmath60 degrees of freedom .
the factor of @xmath61 comes from the fact that if integrating over all coordinates , then then the volume element would include the additional factor @xmath62 ; the factor can also be seen due to dimensional analysis .
the dirac delta function is recast in terms of a complex exponential , @xmath63 .
this yields @xmath64 \bm{x}\right ) \times \nonumber \\ & \exp\left ( -\imath \omega r^2 \right ) d^{nd } \bm{x } d\omega,\end{aligned}\ ] ] where @xmath65 is the unit matrix and the integration over @xmath66 is over all @xmath66-space .
fixing the hyperradius shifts the diagonal elements of @xmath67 by @xmath68 , where @xmath69 is the imaginary number and @xmath66 is the conjugate variable to @xmath70 .
although introducing the dirac delta function leads to an additional auxiliary integration , it avoids switching explicitly to hyperspherical coordinates .
this is key since the integrals without a fixed hyperradius are known to be analytical @xcite .
a unitary coordinate transformation @xmath71 facilitates simplifying eq . . here
, @xmath72 is the transformation matrix that diagonalizes @xmath67 .
in particular , @xmath73 , where @xmath74 is a diagonal matrix with diagonal elements @xmath75 .
. becomes @xmath76 \bm{y}\right ) \times \nonumber \\ & \exp\left ( -\imath \omega r^2 \right ) d^{nd } \bm{y } d\omega.\end{aligned}\ ] ] note that the shift of the diagonal remains unchanged . performing the integration over all space @xmath77 yields @xmath78 thus
, the overlap integration is reduced to a one - dimensional fourier transform , regardless of the number of dimensions or the number of jacobi vectors , i.e. , the number of particles . here
we list some properties of eq . .
first , the integral is guaranteed to be real .
this can be seen from the fact that each factor of @xmath66 is paired with the imaginary number @xmath69 .
thus , the negative @xmath66-axis is the complex conjugate of the positive @xmath66-axis and the integration over all @xmath66-space is equivalent to taking twice the real part of the result .
second , in the special case where the number of dimensions @xmath2 is even , the overlap integral with fixed hyperradius eq .
reduces to an inverse fourier transform with simple poles along the negative imaginary axis .
the integral can be straightforwardly carried out using the method of residues .
if integrating over @xmath79 and additionally @xmath80 , then the result agrees with the overlap matrix element result of suzuki and varga @xcite .
third , the form of eq . is the inverse fourier transform of a simple product of factors like @xmath81 . by the convolution theorem
@xcite , if the inverse fourier transform of one of these factors is known , then the result is the convolution of the untransformed functions .
the inverse fourier transform @xmath82 of one factor is @xmath83 where the variable of importance here @xmath84 is the square of the hyperradius and not @xmath0 alone . by the convolution theorem , eq .
reduces to @xmath85\!(t),\end{aligned}\ ] ] where the convolutions between the square brackets can be done in any order , leaving a function of the variable @xmath86 .
the convolution of two functions , @xmath87\!(t)$ ] , is given by @xmath88\!(t ) = & \frac{1}{\sqrt{2 \pi } } \int_{0}^{t } f_{\gamma_1}^{k_1}(s)f_{\gamma_2}^{k_2}(t - s ) ds .
\end{aligned}\ ] ] the range of integration in eq .
is reduced from @xmath89 to @xmath90 since the functions @xmath82 , @xmath91 , are defined only for positive argument . the resulting function after performing any single convolution has this restriction as well .
for example , performing a third convolution with eq .
yields @xmath92\!(t ) = & \frac{1}{\sqrt{2 \pi } } \int_{0}^{t } \left [ f_{\gamma_1}^{k_1 } \ast f_{\gamma_2}^{k_2 } \right]\!(s ) f_{\gamma_3}^{k_3}(t - s ) ds . \end{aligned}\ ] ] for three particles the overlap integral is fully analytical for any number of dimensions . if @xmath93 , then eq .
reduces to @xmath94 t \right ) } { 2^{d/2 } \left(\gamma_1 - \gamma_2 \right)^{(d-1)/2 } \gamma\left(d/2 \right ) } i_{\frac{d-1}{2}}\left(\frac{1}{4}[\gamma_1-\gamma_2]t \right),\end{aligned}\ ] ] where @xmath95 is the modified bessel function .
this agrees with the results for @xmath96 of ref .
there is a simple physical interpretation here : in keeping the hyperradius @xmath0 fixed , one must convolve functions of the squared lengths of all jacobi vectors .
unfortunately , eq . is difficult to carry out for more than two jacobi vectors since after performing the first convolution analytically , one is left with the convolution of @xmath97 with a hypergeometric function .
this can be carried out , but the result is a kamp de frier function , which is a generalization of the hypergeometric functions to two variables . if convolving yet one more function , then to our best knowledge there is no analytical result .
if the number of dimensions @xmath2 is 2 , however , then eq . reduces to a gaussian .
the convolution of two gaussians is a sum of gaussians , hence eq .
can be carried out analytically .
this approach agrees with the method of residues in evaluating eq . .
in practice , our numerical integrations utilize adaptive gauss - kronrod quadrature to compute integrals of the form of eq . .
the decay of the integrand at large distance can be accelerated through a coordinate transformation @xmath98 .
we use complex arithmetic to calculate the integrand .
the real part of the integrand is also an even function , so we take twice the real part of integrating from zero to infinity .
this section examines the lowest adiabatic potential curve for various few - body equal - mass systems of spin half fermions with @xmath9 symmetry .
the following uses atomic units , where @xmath99 is the bohr radius and @xmath100 is the hartree unit of energy , that is , the particles are assumed to have mass equal to the electron mass @xmath101 .
the number of particles @xmath1 and the number of dimensions @xmath2 are varied , but the pairwise interaction potential is fixed at @xmath102 , @xmath103 being the magnitude of the interparticle distance vector between any two particles of either like or opposite charge . though this potential only corresponds to the coulomb potential for @xmath96 , the particles are labeled by @xmath104 or @xmath105 regardless of dimension to indicate which particles are taken to be identical .
for example , the four - body system with two positive and two negative charges corresponds to @xmath106 , all having spin @xmath107 .
the results presented here use a basis that includes a spinor @xmath108 , @xmath109 where @xmath110 and @xmath111 are the spinors for the positive and negative charges , respectively .
the operator @xmath112 explicitly antisymmetrizes the basis function . the spin quantum numbers @xmath5 and @xmath6 label the different systems under examination .
for example , @xmath113 labels the system of charges with the total spin of both charged subsystem in the singlet spin configuration . in practice
, the spin projection quantum number is chosen to equal @xmath114 .
figure [ fig_dimensions_anti ] for three equal - mass particles @xmath115 with @xmath9 symmetry and @xmath116 in @xmath2 dimensions interacting via @xmath117 potentials .
( a ) dashed , dash - dotted , dash - dot - dotted , and dotted lines are for @xmath118 and @xmath119 , respectively . for @xmath96 , the solid line indicates the position of the ps@xmath7 bound state at @xmath120 @xcite .
( b ) the dashed line is for @xmath121 .
the solid lines on the right indicate the large @xmath0 limit . , width=264 ] shows the lowest adiabatic potential curve for the @xmath122 system in varying dimensions @xmath2 where @xmath116 .
dashed , dash - dotted , dash - dot - dotted , and dotted lines in fig .
[ fig_dimensions_anti](a ) are for @xmath118 and @xmath119 , respectively . the dashed line in fig . [ fig_dimensions_anti](b ) is for @xmath121 . for @xmath121 through 6 , the basis sizes used are 60 , 50 , 18 , 17 , and 20 , respectively .
the large @xmath0 limiting behavior , shown as solid lines , is known from the dimensional scaling work of herschbach @xcite .
this limit corresponds to the break up into a dimer and a free particle . in three dimensions
[ see the dashed lines in fig .
[ fig_dimensions_anti](a ) and fig .
[ fig_dimensions_para](a ) ] , this corresponds to the break up into a ps `` atom '' and a free electron .
modified here for equal - mass particles , the large @xmath0 limits are given by the formula @xmath123 .
figure [ fig_dimensions_anti ] shows that the three - body bound state is most deeply bound for @xmath121 , where the depth of the well is roughly a factor of 10 deeper than that for @xmath96 . within a single - channel approximation
, the bound state energy is estimated using a laguerre basis in the discrete variable representation in the hyperradius @xmath0 , extrapolated to an infinite basis size .
neglecting the @xmath124 coupling gives a strict lower bound estimate of the energy .
solving for this lower bound estimate , we find that bound states only exist for @xmath125 . in particular
, the lower bound @xmath126 is @xmath127 , @xmath128 , and @xmath129 , for @xmath121 , 3 , and 4 , respectively . for @xmath130 ,
the lower bound converges within error bars to the break up threshold value .
this behavior is straightforward to understand .
since each degree of freedom increases the kinetic energy , pushing out the inner barrier to larger hyperradius @xmath0 , the fixed @xmath117 potentials can not `` keep up '' and maintain a bound state as @xmath2 increases , losing out to the increase in kinetic energy beyond @xmath131 .
figure [ fig_dimensions_para ] , except with @xmath132 .
the curves follow the same labeling as in fig .
[ fig_dimensions_anti ] .
note the different vertical scales between fig .
[ fig_dimensions_anti](b ) and panel ( b ) of this figure . , width=264 ] shows the corresponding system to that shown in fig .
[ fig_dimensions_anti ] , that is , @xmath122 and @xmath132 .
dashed , dash - dotted , dash - dot - dotted , and dotted lines in fig .
[ fig_dimensions_para](a ) are for @xmath118 and @xmath119 , respectively . the dashed line in fig . [ fig_dimensions_para](b )
is for @xmath121 , where the potential minimum is much shallower than that of the alternate spin symmetry [ note the different vertical scales between figs . [ fig_dimensions_anti](b ) and figs .
[ fig_dimensions_para](b ) ] .
the large @xmath0 limiting behavior , shown as solid lines , is the same as in fig .
[ fig_dimensions_anti ] . for @xmath121 through 6 , the basis sizes used are 68 , 50 , 27 , 18 , and 10 , respectively . estimating the energies in a single - channel approximation , we find that no bound states exist for any dimension @xmath133 .
the reason for this is simple : the symmetric spin configuration leads to an antisymmetric spatial wave function between identical fermions .
the identical fermions spend more time apart due to particle statistics , and they have more kinetic energy owing to the presence of an additional nodal surface , so even though there is attraction between opposite charges , a tightly bound trimer is less likely to form .
this manifests as a short - range repulsive wall extending to a larger hyperradius @xmath0 than that for the antisymmetric spin configuration [ see fig .
[ fig_dimensions_anti ] ] . fig .
[ fig_4body ] presents results obtained with the dimension fixed at @xmath96 .
symmetry for the @xmath106 system in three dimensions .
this system is of fundamental interest because it has been proposed that a bose - einstein condensate could be formed with positronium atoms at temperatures as high as 20 - 30k @xcite dashed and dotted lines are for @xmath113 and @xmath134 , respectively .
the solid line indicates the position of the @xmath135 bound state at @xmath136 @xcite .
the solid line in the inset is for @xmath137 .
, width=264 ] in particular , fig .
[ fig_4body ] shows the lowest adiabatic potential curves for the @xmath106 system .
dashed and dotted lines are for @xmath113 and @xmath134 , respectively , using respective basis set sizes of 230 and 247 . in the asymptotic ( large @xmath0 ) limit , considering only the spatial part of the wave function , the ground state potential curve dissociates into two ground state ps dimers .
the inner region potential curve depends strongly on the spin configuration , where the singlet - singlet configuration leads to the ps@xmath8 `` molecule '' at @xmath136 @xcite , indicated by the solid line , while the triplet - triplet configuration has a potential minimum that is too shallow to support a bound state . in the single - channel approximation
, we estimate a lower bound for the singlet - singlet configuration of @xmath138 .
the @xmath137 configuration , whose ground state adiabatic potential curve is shown as a solid line in the inset ( using 200 basis functions ) , is symmetry forbidden for @xmath139 to break up into two ground state ps dimers . instead ,
one of the dimers is in its first excited state such that the total asymptotic threshold is at @xmath140 .
the effect of adding a third electron to make this a 5-body system is shown in fig .
[ fig_5body ] .
symmetry for the @xmath141 system in three dimensions .
dashed , dash - dotted , dash - dot - dotted , and dotted lines are for @xmath142 , @xmath143 , @xmath144 , and @xmath145 , respectively . from top to bottom , the solid lines indicate the asymptotic limits of break up into @xmath146 , @xmath147 , and @xmath148 , respectively . , width=264 ] here , fig .
[ fig_5body ] shows the lowest adiabatic potential curves for the @xmath141 system in three dimensions .
dashed , dash - dotted , dash - dot - dotted , and dotted lines show the four possible spin configurations @xmath142 , @xmath143 , @xmath144 , and @xmath145 , respectively . in the same order ,
the four configurations use basis sizes 600 , 600 , 550 , and 450 . from top to bottom , the solid lines show the possible asymptotic limits of dissociation into two positronium atoms and a free electron [ @xmath149 , a positronium and positronium ion [ @xmath150 ) ] , and a dipositronium ps@xmath8 plus a free electron [ @xmath151 @xcite . the system is more repulsive when the three electrons are in the symmetric spin configuration @xmath152 ( see dash - dot - dotted and dotted lines of fig . [ fig_5body ] ) . for fixed @xmath6 value ,
the two positrons in the singlet spin configuration ( @xmath153 ) leads to a more repulsive system compared to the triplet case ( @xmath154 ) , analogous to hund s rules . the @xmath155 configuration ( see the dashed and dash - dotted lines of fig .
[ fig_5body ] ) leads to potential minima at small @xmath0 . within the single - channel approximation , the @xmath143 symmetry ( see the dashed - dotted line of fig . [ fig_5body ] )
does not have a bound state .
we estimate a lower bound to the energy for the @xmath156 symmetry ( see the dashed line of fig .
[ fig_5body ] ) at @xmath157 , which is below the break up threshold , but without the @xmath124 coupling term an accurate prediction to whether a bound state exists can not be made .
nevertheless , a true five body bound state is unlikely since our lower bound estimate lies in the continuum of the @xmath143 symmetry ( dashed line of fig .
[ fig_5body ] ) , that is , above the break up threshold into ps@xmath8 and a free electron .
most likely this state would decay either by radiating a photon or via a spin flip .
this paper considers the correlated gaussian hyperspherical method .
the method is improved by introducing a new technique to calculate matrix elements with a fixed hyperradius @xmath0 in the basis of spherical gaussians . using this technique ,
the matrix elements of the overlap , hyperangular kinetic energy , and central potentials are derived . to show the strength of this technique ,
the lowest adiabatic potential curves for @xmath158 and @xmath159 for the @xmath10 potential are calculated . from lower - bound estimates ,
we confirm the ps@xmath7 and ps@xmath8 bound states and suggest the absence of a true five - body bound state . though we restrict ourselves to the simplest gaussian basis functions , the technique presented is straightforward to generalize to finite angular momentum and parity .
the techniques developed in this study thus provide an efficient extension of the standard correlated gaussian method to the cghs method ( fixed hyperradius ) , which carries advantages for the future description of scattering and rearrangement processes involving several particles .
these topics will be explored in future publications in related work , it will be particularly interesting to study two - dimensional systems due to the fully analytical noninteracting matrix elements . moreover , because the even - dimensional noninteracting system matrix elements are evaluated in fully analytical expressions , it might be possible to find fully analytical odd - dimensional versions of the matrix elements with additional effort .
more immediately , with converged potential curves , properties of the bound and scattering states can be calculated .
support by the nsf through grant phy-1306905 is gratefully acknowledged .
we also acknowledge fruitful discussions with javier von stecher , doerte blume , and jose dincao about the cghs technique .
the hyperangular kinetic energy @xmath29 is proportional to the square of the grand angular momentum operator @xmath160 , @xmath161 the exact form of the grand angular momentum operator @xmath160 depends on the choice of hyperangles , but for our purposes the exact form is not needed and @xmath162 is expressed as the difference between the relative and hyperradial kinetic energies . @xmath163 the symmetrized form of the kinetic energy operator , namely , @xmath164 is calculated using the same technique of fixing the hyperradius via the auxiliary integral over the dirac delta function and integrating over all jacobi coordinates .
the hyperangular kinetic energy integrand takes the form @xmath165\bm{x } \right),\end{aligned}\ ] ] where @xmath166 and @xmath167\bm{x})^2\right].\end{aligned}\ ] ] here , tr is the trace operator .
there are three types of integrations to be performed with fixed hyperradius .
the first type comes from the term in eq . , which leads to an integral that is proportional to @xmath168 .
the second type comes from the terms in eq . , which are of the form @xmath169 .
the third type comes from the terms in eq . , which are of the form @xmath170 . here ,
@xmath171 and @xmath172 are generic symmetric matrices . in the basis that diagonalizes @xmath67 , the second integral type , after integrating over all space , yields @xmath173 where @xmath174 represents an inverse fourier transform akin to eq . .
it implies that all factors are included from eq . with the additional factors @xmath175 in the integrand .
integration over all space of the third integral type yields @xmath176 equation .
is analogous to the third entry of table 7.1 from ref . @xcite . for eq .
there is no such entry , but can be derived using derivative methods described in the appendices of ref .
@xcite . in the basis that diagonalizes @xmath67
, we define @xmath177 and @xmath178 .
equation reduces to @xmath179 where @xmath180 and @xmath181 \nonumber \\ - \left [ 2 \delta_{jk } \gamma_j \gamma_k+ d \gamma_j\gamma_k \right ] \big ) .
\end{aligned}\ ] ] again , if @xmath2 is even , then the kinetic energy matrix element eq .
can be analytically carried out by the method of residues .
this section simplifies the fixed-@xmath0 matrix element @xmath182 and is analogous to the derivation starting from eq .
( 7.6 ) of ref .
the choice of jacobi coordinates is such that the first jacobi vector @xmath183 represents the relative distance vector between the two particles of interest .
the integrand @xmath184 is @xmath185\bm{x}\right).\end{aligned}\ ] ] a unitary coordinate transformation transforms @xmath186 to the basis that diagonalizes @xmath67 and a dirac delta function shifts the argument of the potential , yielding @xmath187 where @xmath188 is the array of coefficients of the transformation of @xmath183 to a linear combination of @xmath189 .
the dirac delta function here is not the same that is used to fix the hyperradius .
the hyperradius @xmath0 is fixed using the fourier transform , yielding @xmath190\bm{y}\right ) \frac{r}{\pi } \exp\left(-\imath \omega r^2 \right ) d^{nd}\bm{y } d^d\bm{r } d\omega.\end{aligned}\ ] ] performing the integration over @xmath77 yields @xmath191 where @xmath192 if @xmath193 is a power law potential @xmath194 , then @xmath195 where @xmath196 is the angular integration over the @xmath2-sphere .
the integration over @xmath196 introduces a factor @xmath197 . performing the integration over @xmath198 yields
@xmath199/2)}{\gamma(d/2 ) } \langle c(\omega)^{k/2 } \rangle _ { \bm{\omega}}.\end{aligned}\ ] ] note that the inverse fourier transform of eq .
involves the @xmath66-dependent factor @xmath200 , yet the dimensional dependence of @xmath200 has dropped out .
|
the solution of the hyperangular schrdinger equation for few - body systems using a basis of explicitly correlated gaussians remains numerically challenging .
this is in part due to the number of basis functions needed as the system size grows , but also due to the fact that the number of numerical integrations increases with the number of hyperangular degrees of freedom .
this paper shows that the latter challenge is no more . using a delta function to fix the hyperradius @xmath0 ,
all matrix element calculations are reduced to a single numerical integration regardless of system size @xmath1 or number of dimensions @xmath2 . in the special case of @xmath2
an even number , the matrix elements of the noninteracting system are fully analytical .
we demonstrate the use of the new matrix elements for the 3- , 4- , and 5-body electron - positron systems with zero total angular momentum @xmath3 , positive parity @xmath4 , and varied spins @xmath5 and @xmath6 .
|
in the past few years , high - order harmonic generation ( hhg ) and above - threshold ionization ( ati ) from aligned molecules in strong laser fields of femtosecond duration have proven to be a powerful tool for resolving , or even controlling , processes in the subfemtosecond and subangstrom scale .
for instance , one may employ hhg and ati in the tomographic reconstruction of molecular orbitals @xcite , and in the attosecond probing of dynamic changes in molecules @xcite .
this is possible due to the fact that the physical mechanisms governing both phenomena take place in a fraction of the laser period , i.e. , within hundreds of attoseconds @xcite , and involve the recombination or the elastic scattering of an electron with its parent molecule @xcite .
thereby , high - order harmonics or high - energy photoelectrons , respectively , are generated .
thus , the spectral features are highly dependent on the spatial configuration of the ions with which the electron rescatters or recombines , and yield patterns which are characteristic of the molecule
. furthermore , they also depend on the alignment angle of the molecule , with respect to the laser - field polarization . due to their simplicity , in particular diatomic molecules have been investigated , and minima and maxima have been encountered in their hhg and ati spectra .
such patterns have been observed both theoretically tdse2005,doubleslit , kb2005,kbk98,mbbf00,madsen , usach2006,usachenko , henrikdipl , kansas , pracl2006,morecl , dm2006,hbf2007 and experimentally @xcite , and have been attributed to the interference between high- order harmonics or photoelectrons generated at different centers in the molecule .
they are , in a sense , the microscopic counterpart of those obtained in a double - slit experiment .
furthermore , the energy positions of the maxima and minima depend on the alignment angle and on the internuclear distance and , additionally , reflect the bonding or antibonding nature of the highest occupied molecular orbitals in question @xcite .
such features have been studied either numerically , by solving the time - dependent schrdinger equation ( tdse ) @xcite , or semi - analytically , by employing the strong - field approximation ( sfa ) kbk98,mbbf00,madsen , usach2006,usachenko , henrikdipl , kansas , pracl2006,morecl , dm2006,hbf2007 . in particular , in @xcite , the contributions to the yield from each molecular center have been singled out within a tdse computation .
therein , it has been explicitly shown that the maxima and minima in the spectra are obtained due to the interference between contributions from different centers , in agreement with the double - slit model in @xcite . specifically in the sfa framework
, the transition amplitude can be written as a multiple integral , with a semiclassical action and slowly - varying prefactors .
the structure of the molecule ( and thus its double - slit character ) can be either incorporated in the prefactors or in the action .
the latter approach also takes into account processes in which an electron rescatters or recombines with a center in the molecule different from the site of its release @xcite
. an open question is , however , which types of electron orbits are responsible for specific interference features .
for instance , are the dips and maxima originated by the interference between orbits in which the electron leaves and returns to the same center ( regardless of which ) , or between those in which the electron leaves one atom and recombines with the other ? on the other hand , it could also be that the interference patterns result from the combined effect of all such orbits , and one is not able to attribute them to specific sets .
since the strong - field approximation is a semi - analytical method , and allows an immediate association with the classical orbits of an electron returning to its parent molecule , it appears to be an ideal tool for tackling this problem .
this approximation , however , possesses several drawbacks .
first , the sfa is gauge dependent , which leads to different pre - factors and action , depending on whether the velocity gauge or the length gauge is taken .
second , even for a specific gauge , the precise expressions for the pre - factors are not agreed upon , and some of them lead to interference patterns which are in disagreement with the experiments , and with results from other methods @xcite . the main objective of this work is to make a detailed assessment of the contributions and relevance of the different types of recombination scenarios to the above - mentioned interference patterns , within the strong - field approximation , for a diatomic molecule . throughout the paper , we approximate the strong , infra - red field by a linearly polarized monochromatic wave @xmath0 , and consider a linear combination of atomic orbitals ( lcao approximation ) . in general ,
we consider that the electron reaches the continuum by tunneling ionization . as an exception
, however , we also take an attosecond - pulse train superposed to the monochromatic wave atthhg2004,attati2005,fsvl2006,fs2006 .
the attosecond - pulse train provides an additional pathway for the electron to reach the continuum , and , recently , has proven to be a powerful tool in order to control high - order harmonic generation @xcite and above - threshold ionization @xcite . in the context of the present work , it is a convenient way to avoid problems related to spurious potential - energy shifts .
these shifts are present in the length gauge and artificially modify the potential barrier through which the electron tunnels .
hence , they may lead to a questionable physical interpretation , what the relevance of the different sets of orbits to the patterns concerns .
this paper is organized as follows . in sec .
[ transampl ] , we discuss the strong - field approximation transition amplitude for high - order harmonic generation , starting from the general expressions ( sec .
[ general ] ) , and , subsequently , addressing the specific situation of a diatomic molecule in the lcao approximation ( sec .
[ diatomic ] ) .
thereby , we consider the situation for which the structure of the molecule is either incorporated in the pre - factor , or in the semiclassical action , in the presence and absence of the attosecond - pulse train .
when discussing the former case , we emphasize the role of the overlap integrals , in which the dipole moment and the atomic wave functions are localized at different centers in the molecule . in the latter case , we follow the model in ref .
@xcite , for a purely monochromatic driving field , and our previous work @xcite when the attosecond pulses are present , very closely .
subsequently ( sec . [ results ] ) , we investigate the interference patterns . finally , in sec .
[ concl ] we summarize the paper and state our main conclusions .
in general , the strong - field approximation ( sfa ) consists in neglecting the influence of the laser field when the electron is bound , the atomic or molecular binding potential when the electron is in the continuum and the internal structure of the system in question , i.e. , contributions from its excited bound states .
the sfa transition amplitude for high - order harmonic generation reads , in the specific formulation of ref .
@xcite and in atomic units , @xmath1+c.c .
, \label{amplhhg}\end{aligned}\]]with the action @xmath2^{2}d\tau -i_{p}(t - t^{\prime } ) + \omega t \label{actionhhg}\]]and the prefactors @xmath3 and @xmath4 in the above equations , @xmath5 , @xmath6 , @xmath7 @xmath8 and @xmath9 denote the dipole operator , the laser - polarization vector , the interaction with the field , the ionization potential , and the harmonic frequency , respectively . the explicit expressions for @xmath10 are gauge dependent , and will be given below .
the above - stated equation describes a physical process in which an electron , initially in a field - free bound - state @xmath11 , is coupled to a volkov state @xmath12 by the interaction @xmath13 of the system with the field .
subsequently , it propagates in the continuum and is driven back towards its parent ion , with which it recombines at a time @xmath14 , emitting high - harmonic radiation of frequency @xmath15 in eq .
( amplhhg ) , additionally to the above - mentioned assumptions , the further approximation of considering only transitions from a bound state to a volkov state in the dipole moment has been made . in the single atom case , this has been justified by the fact that the remaining contributions , from the so - called continuum - to - continuum " transitions , were very small . for a discussion of the various formulations of the sfa see , e.g. , @xcite and in particular @xcite .
due to the fact that it can be carried out almost entirely analytically , the sfa is a very powerful approach .
it possesses , however , the main drawback of being gauge - dependent ( for general discussions see , e.g. , ref .
@xcite , and for the specific case of molecules , ref .
@xcite ) . apart from the obvious fact that the interaction hamiltonians @xmath16 , which are present in @xmath17 , are different in the length and velocity gauges footngauge , in most computations where the sfa is employed , _ field - free _ bound states are taken , which are not gauge equivalent .
indeed , a field - free bound state @xmath18 in the length gauge would be gauge equivalent to the field - dressed state @xmath19 in the velocity gauge , with @xmath20.$ ] such a phase shift causes a translation @xmath21 on a momentum eigenstate @xmath22 hence , for field - free bound states in both gauges it leads to different dipole matrix elements @xmath23 and @xmath24 .
explicitly , in the length gauge @xmath25 , while in the velocity gauge @xmath26 for computations involving a single atom , the latter artifact can be avoided by placing the system at the origin of the coordinate system . for systems composed of several centers , such as molecules ,
however , this ambiguity will always be present .
indeed , in the literature , different results have been reported for molecular sfa computations in the velocity and in the length gauge @xcite . in recent , modified versions of the length - gauge sfa
, this problem has been eliminated for ati by considering the initial bound state @xmath27\left\vert \psi _ { 0}^{(l)}\right\rangle .$ ]
such a state is gauge - equivalent to a field - free bound state in the velocity gauge , and , physically , may be interpreted as a field - dressed state , in which the laser - field polarization is included @xcite . for hhg
, one may proceed in a similar way , with the main difference that the dressing should also be included in the final state , with which the electron recombines . in the dressedmodified sfa , @xmath28 .
we will now apply the sfa to a diatomic molecule . for this purpose
, we will consider the simplest scenario , namely a one - electron system and frozen nuclei .
furthermore , we will assume that the molecular orbital from which the electron is released and with which it may recombine is a linear combination of atomic orbitals ( lcao approximation ) .
explicitly , the molecular bound - state wave function reads @xmath29where @xmath30 and @xmath31 denote the positions of the centers @xmath32 and @xmath33 respectively , and @xmath34 is a normalization constant . for homonuclear molecules , which we will consider here , @xmath35 .
the positive and negative signs for @xmath36 correspond to bonding and antibonding orbitals , respectively . within this context ,
there exist two main approaches for computing high - order harmonic spectra , which will be discussed next .
the simplest and most widely used mbbf00,madsen , usachenko , morecl , kb2005,dm2006,jmocl2006 approach is to incorporate the structure of the molecules in the prefactors @xmath37 and @xmath38 and to employ the same action @xmath39 as in the single - atom case .
the multiple integral in the transition amplitude ( amplhhg ) can then either be solved numerically , or using saddle point methods @xcite .
the latter procedure can be applied for high enough intensities and low enough frequencies , and consists in approximating ( [ amplhhg ] ) by its asymptotic expansion around the coordinates @xmath40 for which @xmath41 is stationary .
this implies that @xmath42 and @xmath43 in this paper , we employ the specific saddle - point approximations discussed in ref .
@xcite .
for a single atom placed at the origin of the coordinate system , this leads to the equations @xmath44 ^{2}=-2i_{p } , \label{saddle1}\]]@xmath45 = 0 , \label{saddle3}\]]and @xmath46 ^{2}. \label{saddle2}\]]eq .
( [ saddle1 ] ) expresses the conservation of energy at the time @xmath47 at which the electron reaches the continuum by tunneling ionization .
this equation possesses no real solution , which reflects the fact that tunneling has no classical counterpart . in the limit @xmath48 ,
one obtains such a condition for a classical particle reaching the continuum with vanishing drift velocity .
( [ saddle3 ] ) constrains the intermediate momentum of the electron , so that it returns to its parent ion , and , finally , eq .
( [ saddle2 ] ) describes the conservation of energy at a later time @xmath14 , when the electron recombines with its parent ion and a high - frequency photon of frequency @xmath9 is generated .
the matrix element @xmath49 then reads @xmath50 , \label{diplcao}\]]where @xmath51\varphi _ { 0}(\mathbf{r}_{j})d^{3}r_{j } , \label{prefint1}\]]and @xmath52 is the dipole moment . in the length gauge , which we are mostly adopting in this paper , @xmath53 unless strictly necessary , in order to simplify the notation , we do not include the time dependence in @xmath54 the dipole moment can be written in several forms . if one considers the length form a natural choice is @xmath55 other possibilities are to consider the operator @xmath56 in its velocity and acceleration forms jmocl2006,ehrenfest .
inserting @xmath57 in eq .
( [ prefint1 ] ) yields @xmath58where @xmath59\varphi _ { 0}(% \mathbf{r}_{j})d^{3}r_{j}\]]and @xmath60\varphi _ { 0}(\mathbf{r}_{j})d^{3}r_{j},\nu \neq j.\ \label{inte2}\]]specifically , if the dipole is in the length form , the above - stated integrals read @xmath61and @xmath62where @xmath63\varphi _ { 0}(\mathbf{r}_{j})d^{3}r_{j}. \label{pspacephi}\ ] ] in @xmath64 , @xmath65 for @xmath66 and @xmath67 for @xmath68 .
( [ diplcao ] ) is then explicitly written as @xmath69 , \label{prefb}\]]for bonding molecular orbitals ( i.e. , @xmath70 or @xmath71 , \label{prefa}\]]in the antibonding case ( i.e. , @xmath72 with @xmath73 in eqs . (
[ prefb ] ) and ( [ prefa ] ) , the terms with a purely trigonometric dependence on the internuclear distance yield the double - slit condition in @xcite .
the maxima and minima in the spectra which are caused by this condition are expected to occur for @xmath74respectively , for bonding molecular orbitals ( i.e. , @xmath75 for antibonding orbitals , the conditions are reversed , i.e. , the maxima occur for the odd multiples of @xmath76 and the minima for the even multiples .
if the velocity gauge is taken , the above stated conditions hold for @xmath77 , instead of @xmath78 .
this is due to the fact that the initial and final free - field sttes are not gauge equivalent , as discussed in sec .
the remaining terms grow linearly with the projection @xmath79 of the internuclear distance along the direction of the laser - field polarization , and may lead to unphysical results @xcite . for this reason ,
they are sometimes neglected in the integrals @xmath80 @xcite .
there exists , however , no rigorous justification for such a procedure .
indeed , only recently , it has been shown that such terms can be eliminated by considering an additional interaction which depends on the nuclear coordinate .
this interaction is present in a modified molecular sfa , in its dressed and undressed versions @xcite , and leads to contributions which cancel out the linear term in @xmath79 . in the length gauge , if the length form of @xmath5 is taken , @xmath81 , with @xmath82 while in the velocity gauge , @xmath83^{2}}{(2\pi ) ^{3/2}}\cos ( \vartheta ) \phi ( \mathbf{\tilde{p}}),\]]or@xmath84^{2}}{(2\pi ) ^{3/2}}\sin ( \vartheta ) \phi ( \mathbf{\tilde{p}}),\]]with @xmath85 for bonding and antibonding molecular orbitals , respectively .
physically , if one employs the pre - factors ( [ prefb ] ) and ( [ prefa ] ) , this means that one is not modifying the saddle - point equations ( saddle1)-([saddle2 ] ) .
therefore , the orbits along which the electron is moving in the continuum remain the same as in the single - center case .
this approach is questionable in several ways . from the technical viewpoint
, there is no guarantee that such pre - factors are slowly varying , as compared to the semiclassical action , especially for large internuclear distances @xcite .
furthermore , in general , they do not incorporate processes in which the electron leaves one center of the molecule and recombines with the other , which , physically , are expected to be present in molecular hhg @xcite .
a slightly more sophisticated approach is to exponentialize the pre - factors obtained in the two - center case and incorporate the terms in @xmath86 in the action .
this procedure has been adopted in pracl2006 and will be closely followed in this work . for the sake of simplicity , we will consider the modified prefactor @xmath87 , \label{modifieddip}\]]for which the second term in eq .
( [ crossed ] ) is absent .
( modifieddip ) is also very similar to the dipole matrix element in the velocity form , apart from the fact that , in the latter case , @xmath88 is replaced by @xmath89 @xcite . in the expression for the antibonding case , the cosine term in ( [ modifieddip ] ) should be replaced by @xmath90 .
this leads to the sum @xmath91of the transition amplitudes @xmath92,\ \label{amplitudes}\end{aligned}\]]with @xmath93 ^{\ast } \partial _ { \tilde{p}_{x}}\phi ( \mathbf{% \tilde{p}(}t^{\prime } ) ) .$ ] the terms @xmath94 correspond to a modified action , which incorporates the structure of the molecule .
explicitly , for the undressed length - gauge sfa , @xmath95and @xmath96where @xmath97\cdot \mathbf{r}/2 $ ] and @xmath98\cdot \mathbf{r}/2.$ ] eq .
( [ ssame ] ) and ( [ sdiff ] ) may be directly related to physical processes involving one or two centers in the molecule , respectively , as it will be discussed next . for this purpose
, we will solve the multiple integrals in ( [ amplitudes ] ) employing saddle - point methods .
the conditions @xmath99 and @xmath100 upon the derivative of the action yield saddle - point equations , which , as in the previous section , can be related to the orbits of an electron recombining with the molecule .
explicitly , for the modified action @xmath101 [ eq .
( [ ssame ] ) ] , the saddle - point equations read @xmath102^{2}}{2}=-i_{p}-\mathbf{e}% ( t^{\prime } ) \cdot \mathbf{r}/2 , \label{tunnelc1}\]]@xmath103^{2}}{2}=\omega -i_{p}+\mathbf{e}% ( t)\cdot \mathbf{r}/2 , \label{recc1}\]]for @xmath104 or @xmath102^{2}}{2}=-i_{p}+\mathbf{e}% ( t^{\prime } ) \cdot \mathbf{r}/2 , \label{tunnelc2}\]]@xmath103^{2}}{2}=\omega -i_{p}-\mathbf{e}% ( t)\cdot \mathbf{r}/2 , \label{recc2}\]]for @xmath105 the saddle point equations ( [ tunnelc1 ] ) and ( [ tunnelc2 ] ) correspond to the tunnel ionization process from center @xmath32 and @xmath106 , respectively .
curiously , both equations contain extra terms , if compared to equation ( [ saddle1 ] ) for the single - atom case .
such terms are dependent on the internuclear distance @xmath107 and the external laser field @xmath108 at the time the electron is freed , and may be interpreted as potential - energy shifts in the barrier through which the electron tunnels out .
similar terms are also observed in eqs .
( [ recc1 ] ) and ( [ recc2 ] ) for the energy conservation at the time the electron recombines , as compared to the single - atom expression ( [ saddle2 ] ) .
the remaining saddle point equation is given by the same expression as in the single - atom case , i.e. , eq .
( [ saddle3 ] ) , and means that , for @xmath109 and @xmath110 , the electron is ejected and returns to the same center . if on the other hand , we consider the modified action ( [ sdiff ] ) , this yields @xmath111ds\pm \mathbf{r}=0 , \label{returndiff}\]]which , physically , mean that the electron is leaving from one center and recombining with the other .
the negative and positive signs refer to the transition amplitudes @xmath112 ( center @xmath32 to center @xmath113 and @xmath114 ( center @xmath106 to center @xmath32 ) , respectively . in the former case ,
the remaining saddle points are given by ( [ tunnelc1 ] ) and ( recc2 ) , i.e. , the electron tunnels from @xmath32 and recombines with @xmath115 whereas in the latter case they are given by ( [ tunnelc2 ] ) and ( [ recc1 ] ) , which physically , expresses the fact that the electron is ejected at @xmath106 and recombines with @xmath116 the energy shifts @xmath117 @xmath118 in ( [ tunnelc1])-([recc2 ] ) are absent in the velocity gauge sfa @xcite , and in a modified length - gauge sfa , in which the electric field polarization is incorporated in the initial and final electronic bound states . in both cases , the action @xmath119 , associated to orbits involving only one center , is given by the single - atom expression ( [ actionhhg ] ) , which leads to the saddle - point equations ( [ saddle1])-([saddle2 ] ) .
the action @xmath94 related to two - center orbits reads @xmath120the above - stated expression leads to the single - atom equations ( [ saddle1 ] ) , ( [ saddle2 ] ) for tunneling and rescattering , together with the two - center return condition ( [ returndiff ] ) .
the prefactors , however , are different in both cases . in the dressed modified length - gauge sfa , @xmath121 ^{\ast }
\partial _ { p_{x}}\phi ( \mathbf{p}),$ ] while in the velocity gauge @xmath122 ^{\ast } \phi ( \mathbf{p})[\mathbf{p}+\mathbf{a}(t^{\prime } ) ] ^{2}/2.$ ] one should note that , within the specific model employed here , the physical process in which the electron moves directly from one center to the other , without reaching the continuum , is not being considered .
such a process leads to a strong set of harmonics in the low - energy range of the spectra .
since , however , we are focusing on the plateau harmonics , the contributions from this extremely short set of orbits are not of interest to the present discussion . for a detailed study of this case , see , e.g. , kbk98,usach2006 .
the role of the energy shifts observed in ( [ tunnelc1])-([recc2])is not well understood . a way of eliminating such terms is to modify
the length - gauge sfa and include the influence of the laser field in the initial and final states .
however , even without such modifications , it is possible to provide an additional pathway for the electron to reach the continuum , so that , at least in the context of tunneling ionization , these shifts can be avoided .
for instance , if the electron is ejected by a high - frequency photon , it does not have to tunnel through potential barriers with energy shifts whose physical meaning is not clear .
such a pathway can be provided by a time - delayed attosecond - pulse train @xmath123 superposed to a strong , near infra - red field @xmath124 .
indeed , it has been recently shown that such pulses can be used to control the electron ejection in the continuum , and thus high - harmonic generation and above - threshold ionization atthhg2004,attati2005,fsvl2006 .
in @xcite , we employed such a scheme to control quantum - interference effects for high - harmonic generation and above - threshold ionization for the single - atom case , within the sfa framework .
our previous findings suggest that the probability of the electron reaching the continuum , in case it is ejected by the attosecond pulses , is roughly the same for all sets of orbits .
indeed , it appears that the sole , or at least main factor determining the intensities in the spectra is the excursion time of the electron in the continuum . in the specific case studied in @xcite
, there was a set of very short orbits , which led to particularly strong harmonics .
therefore , an attosecond pulse train superposed to a strong laser field is an ideal setup to avoid any artifacts due to modified tunneling conditions . in @xcite , we have approximated the attosecond pulse train by a sum of dirac - delta functions in the time domain .
this yields @xmath125where @xmath126 , @xmath127 , and @xmath128 denote the laser field frequency , the attosecond - pulse strength and the train temporal envelope , respectively .
this approximation is the limiting case for a train composed of an infinite set of harmonics , and has the main advantage of allowing an analytic treatment of the transition amplitudes involved up to at most one numerical integration .
furthermore , it is a reasonable asymptotic limit for pulses composed by a large high - harmonic set .
we consider here @xmath129 which , physically , corresponds to an infinitely long attosecond - pulse train .
clearly , the total field is given by @xmath130 we will now assume that the attosecond pulses are the only cause of ionization and that the subsequent propagation of the electron in the continuum is determined only by the monochromatic field . hence , @xmath131 and @xmath132 in eqs .
( [ amplhhg ] ) and ( [ amplitudes ] ) .
this eliminates the integral over the ionization time in the transition amplitudes .
hence , the values for which the single atom - action @xmath133 or the modified action @xmath94 is stationary must be determined only with respect to the variables @xmath14 and @xmath134 .
physically this means that the recombination and return conditions remain the same , with regard to the purely monochromatic case , but that there is no longer a saddle - point equation constraining the initial electron momentum .
in fact , the electron is being ejected in the continuum with any of the energies @xmath135 , since all harmonics composing the train are equivalent . for the other extreme limit , namely a high - frequency monochromatic wave ,
we refer to @xcite , where we provide a detailed discussion within the sfa . explicitly ,
if the action is not modified , these assumptions lead to the transition amplitude @xmath136 \notag \\ & & \times d_{\mathrm{rec}}^{\ast } ( \mathbf{\tilde{p}}(t))d_{\mathrm{ion}}(% \mathbf{\tilde{p}}(t^{\prime } ) ) .
\label{hhgatto}\end{aligned}\]]in case one considers the transition amplitudes ( [ amplitudes ] ) , this yields@xmath137.\\end{aligned}\]]in both equations , @xmath138 the saddle - point equations ( [ saddle3 ] ) and ( [ saddle2 ] ) for the single atom case can then be combined as @xmath139 \left ( \cos \omega t\mp \sqrt{\frac{\omega -i_{p}}{2u_{p}}}\right ) , \label{saddelt}\end{aligned}\]]which will give the return times @xmath140 if , on the other hand , these equations are modified , it is possible to distinguish four main scenarios .
specifically , for the processes in which the electron leaves and returns to the same center , the saddle - point equations differ from ( [ saddelt ] ) only in a shift @xmath141 the negative and positive signs correspond to @xmath109 and @xmath142 respectively .
for the scenarios involving two centers , the saddle - point equations read @xmath143 \left ( \cos \omega t\mp \sqrt{\frac{\omega -\tilde{i}_{p}}{2u_{p}}}\right ) , \end{aligned}\]]with @xmath144 the case @xmath145 and @xmath146 corresponds to @xmath112 and @xmath147 respectively .
we will now present high - harmonic spectra , in the presence and absence of the attosecond pulses .
we restrict the electron ionization times to the first half cycle of the driving field .
we also consider the six shortest pairs of orbits for the returning electron .
due to wave - packet spreading , the contributions from the remaining pairs are negligible spreading . for simplicity ,
we employ a bonding combination of 1s states , for which @xmath148^{2 } } , \label{dip1s}\]]and assume that the molecule is aligned parallel to the laser - field polarization , so that @xmath149 we will commence by considering the single - center action ( [ actionhhg ] ) and the prefactors discussed in the previous section . in fig . [ contour ] , we depict high - harmonic spectra computed for a wide range of internuclear distances , employing the prefactor ( [ prefb ] ) or the modified expression ( [ modifieddip ] ) , for which the linear term in @xmath79 is absent [ upper and lower panel , respectively ] .
the figure illustrates how the linear term masks the interference patterns .
in fact , for a.u . @xmath150 a.u . , it counterbalances the influence of the purely trigonometric term , and no clear minima and maxima are observed .
as the internuclear distance increases , this term starts to play a dominant role , displacing the maxima and minima away from the double - slit condition ( [ maxmin ] ) .
furthermore , it seems that the patterns become more distinct with increasing harmonic order . ) , linear combination of atomic orbitals , for internuclear distances @xmath151
the upper and lower panels correspond to the dipole prefactors ( [ prefb ] ) and ( modifieddip ) , respectively .
the atomic system was chosen as @xmath153 , which was approximated by the linear combination of @xmath154 atomic orbitals with @xmath155 a.u .. we took the the driving field intensity and frequency to be @xmath156 , and @xmath157 a.u . , respectively . , width=340 ] ) of the dipole operator , which excludes the term with a linear dependence on @xmath79 .
the atomic system was chosen as @xmath153 , which was approximated by the linear combination of @xmath154 atomic orbitals with @xmath155 a.u .. the internuclear distance and the alignment angle are @xmath158 a.u . , and @xmath159 respectively .
the driving field intensity and frequency are given by @xmath156 , and @xmath157 a.u .
, respectively .
the interference minima are indicated by the arrows in the figure.,width=340 ] a rough estimate of the influence of each term in ( [ prefb ] ) on the spectra agrees with fig . 1 . since the trigonometric functions in ( [ prefb ] ) are bounded , the ratio between the maxima caused by each term will be @xmath160 , where @xmath161 is the alignment angle and @xmath162 is given by eq .
( [ pspacephi ] ) .
if @xmath163 , the maxima will possess the same order of magnitude and there will be no noticeable modulation , while if @xmath164 the double - slit physical picture may still be reproduced .
for @xmath165 however , one expects that the linear term in @xmath166 will prevail .
hence , the critical value for the internuclear distance is @xmath167 .
this expression depends on the bound states with which the electron recombines , and also on the harmonic energy .
for instance , specifically for @xmath154 states , @xmath168 above the ionization threshold , according to eq .
( [ saddle2 ] ) , @xmath169 .
hence , one expects the linear term to be more prominent as the harmonic energy increases , leading to clearer , though incorrect , patterns . in order to avoid such problems
, we will employ the prefactor ( modifieddip ) throughout . in fig .
[ interfe1 ] , we compare spectra computed using either the pre - factor ( [ modifieddip ] ) or modified saddle - point equations .
both spectra are very similar , with maxima and minima at harmonic frequencies @xmath170 , as expected from the double - slit condition .
this similarity holds not only for the gross features , but , additionally , for the substructure caused by other types of quantum interference .
close to the minima , however , the yield from the latter case is larger .
nevertheless , the very good overall agreement shows that , in fact , the patterns obtained can also be interpreted as the result of the quantum interference between different types of electron orbits . in order , however , to identify which sets of orbits cause the dips and the maxima , we will analyze the interference between their individual contributions .
such results are shown in fig .
[ orbits1 ] . in the upper panels , we present the spectra computed from topologically similar sets of orbits , i.e. , from processes involving only one , or two centers [ panel [ orbits1].(a ) ] . in this case , the main interference patterns are absent .
this strongly suggests that they are due to the quantum interference of topologically different sets of orbits : the orbits along which an electron leaves and returns to the same center , and those along which it reaches the continuum at one center and recombines with the other .
physically , this could be attributed to the fact that , in this case , there would be an appreciable phase difference between the two sets of orbits , since the latter are much longer than the former .
this phase difference would cause the overall modulation .
hence , there are two remaining possibilities . concretely
, the modulation can be due to the quantum interference either between processes in which the electron leaves from different centers and recombines with the same center ( i.e. , between the orbits which start at @xmath171 and end at @xmath172 , with @xmath173 and those starting and ending at @xmath172 , and @xmath174 ) , or between those in which an electron starts at the same center and recombines with different centers ( i.e. , the electron is ejected at @xmath171 and recombines at @xmath172 , @xmath173 or it is freed and recombines at @xmath175 , with @xmath174 ) . in panel
[ orbits1].(b ) , we consider the former processes , whereas in panel [ orbits1].(c ) we depict the latter .
interestingly , only in case the electron leaves from the same center , the interference patterns are present .
furthermore , there is a difference in roughly four orders of magnitude between the two types of contributions .
such a difference is absent in the other cases .
additionally , the full contributions to the yield are practically indistinguishable from the transition probability @xmath176 from the orbits starting at @xmath177 a.u .
the remaining parameters are the same as in fig .
3 . panel ( a ) : orbits involving similar scattering scenarios , i. e. , @xmath178 and @xmath179 and @xmath180 for comparison , the full contributions @xmath181 are displayed as the light gray circles in the picture.,width=340 ] in order to understand this better , one must have a closer look at the individual contributions from different sets of orbits , and , in particular their orders of magnitude . such contributions , depicted in panels ( a ) and ( b ) of fig .
[ orbits2 ] , show that the transition probabilities latexmath:[$% orbits starting from the center @xmath33 are roughly four orders of magnitude larger than latexmath:[$% orbits starting at @xmath32 . therefore , it is not surprising that the yield is dominated by @xmath184 $ ] of the start times provides additional insight into this problem . due to the fact that tunneling ionization is a process which has no classical counterpart ,
this quantity is always non - vanishing , even if the energy range in question is lower than the maximal harmonic energy .
the larger @xmath185 $ ] is , the less probable it will be that tunneling ionization takes place . such an interpretation has been successfully employed in @xcite in order to determine the dominant pairs of orbits , in the context of nonsequential double ionization with few - cycle laser pulses , and will be also considered in this work . for that purpose , we will take the shortest pairs of orbits utilized in the computation of the transition probabilities in fig .
orbits2 and , for each case , display @xmath185 $ ] .
these are the dominant pairs of orbits .
the longer pairs have a less significant influence on the spectra , due to the spreading of the electronic wave packet @xcite .
such results are depicted in fig .
[ orbits3 ] .
clearly , @xmath186 $ ] is around four times larger for the orbits starting from the center @xmath32 , as compared to those starting from @xmath187 this means that , in order to reach the continuum , the electron must overcome a larger barrier if it comes from @xmath32 .
since , roughly speaking , the ionization probability per unit time decreases exponentially with @xmath186 $ ] , one expects the contributions from the orbits starting from @xmath188 to be around four orders of magnitude smaller than those from the orbits starting at @xmath106 .
an inspection of fig .
[ orbits2 ] shows that this is indeed the case . the above - stated effect could , however , be an artifact of the strong - field approximation in the length gauge .
indeed , the terms @xmath189 in the saddle - point equations ( [ tunnelc1 ] ) and ( [ tunnelc2 ] ) can be interpreted as potential energy shifts , due to the fact that the electron is displaced from the origin pracl2006,dm2006 .
such terms increase or sink the potential barrier for @xmath188 or @xmath33 respectively , and , consequently , change the orders of magnitude in the contributions starting from different centers . even though , as a whole , the results match those obtained by other means , their physical interpretation is controversial . of the shortest pairs of orbits , which contribute to the matrix elements @xmath190 and @xmath191 [ panel ( a ) ] , and to @xmath147 and @xmath109 [ panel ( b ) ] , for the same field and atomic parameters as in the previous figure . , width=302 ] one may , however , avoid this problem by providing an additional pathway for the electron to reach the continuum .
for that purpose , we shall superpose a time - delayed attosecond pulse train to the strong laser field , employing the model discussed in sec .
[ attopulses ] and in our previous work fsvl2006,fs2006 .
the maximal harmonic energies for this specific model are strongly dependent on the time delay @xmath192 between the attosecond - pulse train the infra - red field , extending from the ionization potential , for @xmath193 to @xmath194 , for @xmath195 .
furthermore , there exist many intermediate delays , for which a double plateau is present .
this substructure may be detrimental to the identification and physical interpretation of the interference patterns . in order to avoid such problems
, we will consider here vanishing time delay , i.e. , @xmath196 a.u , for an attosecond pulse train superposed to a strong , near - infrared field , and exhibiting vanishing time - delay ( @xmath197 ) with respect to it .
we consider a bonding combination of @xmath154 atomic orbitals , and take the intensity of the attosecond - pulse train to be @xmath198 .
the remaining parameters are the same as in the previous figures .
panel ( a ) : spectra computed employing the modified version of the two - center prefactors in the length form [ eq .
( [ modifieddip ] ) ] and single - atom saddle - point equations , compared to that obtained using modified saddle - point equations .
panel ( b ) : contributions from individual scattering scenarios , i.e. , from @xmath199 , with @xmath174 and @xmath200 .
the interference minima are indicated by arrows in the figure.,width=302 ] .,width=302 ] in fig .
[ atto1 ] , we depict the spectra obtained for a diatomic molecule subjected to such a field , assuming either a two - center prefactor and the single - center saddle - point equation ( [ saddelt ] ) , or the modified saddle point equations ( [ moddelt ] ) [ fig . [ atto1].(a ) ] . both computations exhibit a minimum near the harmonic frequency @xmath201 , in agreement with eq .
( [ maxmin ] ) .
if only the contributions @xmath202 from the individual scattering scenarios are taken , such a minimum is absent [ fig .
[ atto1].(b ) ] .
therefore , it is due to interference effects between different sets of orbits .
one should note that , in contrast to the purely monochromatic case , all contributions exhibit the same order of magnitude .
this is due to the fact that , if the attosecond pulses are present , the electron is being ejected in the continuum with roughly the same probability , regardless of the center it left from .
the precise role of the various recombination scenarios is illustrated in fig .
[ atto2 ] . for clarity
, we concentrate on the plateau region around the interference minimum .
the main difference observed , with regard to the purely monochromatic case , is that the overall shape of the spectrum , and consequently its minimum , is due to the collective interference of several types of orbits .
this is in contrast to the previous results , for which they were caused by the processes starting at a center @xmath171 and ending at different centers , i.e. , @xmath203 , with @xmath204 and @xmath205 @xmath206 .
indeed , a modulation is even present for the probability @xmath207.(a ) , and contradicts the previously made assumption that such features are due to the interference between topologically different sets of orbits .
in fact , it seems that the absence of overall maxima and minima for the one - center contributions , in the purely monochromatic case [ fig . [ orbits1].(a ) ] , is due to the different orders of magnitude for @xmath208 and @xmath110 [ c.f .
[ orbits2].(a ) ] .
furthermore , one needs several different processes in order to obtain the correct position of the minimum .
for instance , in fig .
[ atto2].(a ) , the overall spectrum closely follows @xmath209 in the low - energy region . in the vicinity of the minimum and for higher energies
, however , it follows neither such contrbutions nor @xmath210 from two - center processes . in fig .
[ atto2].(b ) , where we display the contributions @xmath211 @xmath212 from the orbits starting from the same center , the spectrum closely follows @xmath213 after the minimum . the remaining panel [ fig .
[ atto2].(c ) ] depicts the contributions from @xmath211 @xmath206 , which give the orbits _ finishing _ at the same center . in this case , the interference minimum is absent .
this is a strong evidence that the relevant condition for the presence of such features is that the orbits taken into account end at different centers , instead of being topologically different ( which is the case for both fig .
[ atto2].(b ) and fig .
[ atto2].(c ) ) .
therefore , these results agree with the double - slit picture , which has been put across in @xcite .
the additional attosecond pulses have the advantage of not introducing changes in the standard length - gauge sfa formulation .
they modify , however , the physics of the problem , since they provide a different mechanism for the electron to reach the continuum .
clearly , there is also the possibility of eliminating the spurious potential energy shifts , by considering a different version of the sfa , such as the field - dressed length - gauge formulation proposed in @xcite , or the velocity - gauge formulation . , compared to its field - undressed velocity - gauge counterpart [ panel ( b ) ] .
the arrows in the figure indicate the harmonic order for which a minimum is observed , and the vertical line marks the rough estimate for such a minimum .
the solid and dotted lines correspond to the prefactor ( 19 ) and to the situation for which @xmath214/2 has been set to one , respectively.,width=340 ] in fig . [
dressed ] , we display high - order harmonic spectra for the field - dressed sfa in the length gauge , and for the field - undressed sfa in the velocity gauge [ panels ( a ) and ( b ) , respectively ] . for simplicity ,
we exhibit the results obtained with the modified prefactor ( [ modifieddip ] ) , instead of a modified action .
we also provide curves for which only the cosine term has been set to one , in order to facilitate the identification of interference effects . as an overall feature
, we do not observe the interference patterns exhibited in the previous figures . indeed , the curves with and without the cosine term are very similar .
there is , however , a minimum near @xmath215 for the former case . for the dressed length - gauge sfa , the above - stated features can be attributed to the fact that the condition for maxima and minima is now given by ( [ maxmin ] ) , with @xmath216 instead of @xmath217 the harmonic frequencies for which they occur can be easily obtained from condition ( [ saddle2 ] ) , and are given by @xmath218 /2 .
\label{omegadress}\]]an upper bound for @xmath9 can be estimated as follows . at the electron return times , the vector potential is roughly @xmath219 this yields , for the parameters in fig . [
dressed ] , @xmath220 , which is slightly larger than the minimum encountered .
a breakdown of the interference patterns also occurs in the velocity gauge , for the very same reasons .
indeed , the interference condition for the sfa in the velocity gauge and for the field - dressed sfa in the length gauge are identical .
this is a direct consequence of the fact that field - free initial and final electron states in the velocity gauge are gauge - equivalent to the field - dressed states considered in this paper .
this gauge equivalence will lead to identical recombination form factors @xmath221 . since the interference conditions ( [ maxmin ] ) are mainly determined by @xmath222
, they will be the same .
hence , the harmonic orders for which the maxima and minima occur are given by eq .
( [ omegadress ] ) , and therefore are unrealistically high .
the discrepancies between both yields stem from the form factors @xmath223 , which are gauge - dependent .
one may , however , consider field - dressed states in the velocity gauge , which are gauge - equivalent to field - free states in the length gauge .
this is achieved by applying the transformation @xmath224 $ ] in the initial and final length - gauge electronic bound states .
this leads to a shift @xmath225 on a momentum eigenstate @xmath226 , which is exactly the opposite shift induced in the field - dressed length - gauge sfa .
this shift has the main consequence that the interference condition ( maxmin ) now holds for @xmath227 , even in the velocity gauge .
the results obtained employing such dressed states in the velocity - gauge sfa , depicted in fig .
9 , are indeed very similar to those obtained using its field undressed , length gauge formulation .
in fact , we have observed mainly quantitative differences , due to different prefactors @xmath228 @xcite .
one should note , however , that the transformation @xmath229 introduces the same additional potential energy shifts in the modified action as in the undressed length - gauge case .
thus , the price one pays for recovering the correct interference conditions is the loss of a direct connection to simple classical models .
the present results support the viewpoint that the maxima and minima in the high - order harmonic spectra of diatomic molecules are due to the interference of electron orbits finishing at different centers in the molecule .
this seems to hold regardless of whether the electron has been released in one center at the molecule and recombines with a different center , or whether it is ejected and returns to the same center .
such conclusions have been reached by employing modified saddle - point equations , within the strong - field approximation .
these modifications lead to orbits involving different centers , and are a slightly more refined approach than the standard procedure of considering single - center saddle - point equations and modified prefactors . in this framework
, we considered that the electron has been ejected by tunneling ionization and by an additional attosecond - pulse train , and compared the similarities and differences from both physical situations . in the former case ,
depending from which center the electron is leaving , it must overcome unequal potential barriers to reach the continuum , whereas in the latter case it is ejected with roughly equal probability . especially in the purely monochromatic case
, we observed that the contributions from orbits starting at one of the centers , namely @xmath230 , were much larger than those from @xmath231 , due to a narrower potential barrier .
in particular , there exist potential - energy shifts which are proportional to the electric field at the ionization time and the internuclear distance , which cause such differences .
it is however noteworthy that the electron excursion times have been confined to the first half - cycle of the laser field .
if the other half - cycle had been taken , the potential barrier would reverse and the contributions starting from @xmath231 would be more prominent . the above - mentioned potential - energy shifts , however , do not possess a clear - cut physical interpretation .
in fact , they are only present in the standard length - gauge formulation of the strong - field approximation , i.e. , if field - free bound states are taken when the electron is ejected and recombines , and are the source of several problems . for instance
, they reflect the fact that the sfa is translation - dependent .
moreover , due to their existence , it is difficult to establish an immediate connection between this approach and the classical equations of motion of an electron in the laser field .
on the other hand , it seems that such shifts are necessary in order to obtain the correct energy position of the maxima and minima in the hhg spectra .
in fact , an improved formulation of the sfa , in which the influence of the laser field is included in the electron bound states , restores its translation invariance , provides an unproblematic classical limit dm2006 , but yields incorrect energy positions for the interference patterns .
this discrepancy is related to the fact that the field dressing alters the double - slit interference conditions ( [ maxmin ] ) .
a similar absence of interference features has been reported very recently in ref .
@xcite , for hhg computations using a field - dressed version of the sfa in the length gauge .
finally , when employing different gauges , from our results it is clear that the dressing of the initial and the final states plays a far more important role than the different interaction hamiltonians @xmath232 .
if the dressing is applied consistently so that the the electronic bound states are gauge - equivalent , the interference patterns will remain the same .
this is due to the fact that the interference condition ( maxmin ) will then remain invariant .
for instance , the spectra obtained in the velocity - gauge sfa with undressed states are very similar to those computed in the length gauge with field dressed states .
the same holds for the undressed length - gauge , and the dressed velocity - gauge sfa spectra . in the two former cases ,
there is a breakdown of the interference patterns , as compared to the field - undressed length gauge sfa .
however , such patterns can be restored in the velocity gauge , by dressing the electronic bound states appropriately .
we would like to thank l. e. chipperfield , r. torres , j. p. marangos , and h. schomerus for useful discussions , and w. becker for calling ref .
@xcite to our attention .
we are also grateful to the imperial college and to the university of stellenbosch for their kind hospitality .
this work has been financed by the uk epsrc ( advanced fellowship , grant no .
ep / d07309x/1 ) .
h. niikura , f. lgar , r. hasbani , a. d. bandrauk , m. yu .
ivanov , d. m. villeneuve and p. b. corkum , nature * 417 * , 917 ( 2002 ) ; h. niikura , f. lgar , r. hasbani , m. yu .
ivanov , d. m. villeneuve and p. b. corkum , nature * 421 * , 826 ( 2003 ) ; s. baker , j. s. robinson , c. a. haworth , h. teng , r. a. smith , c. c. chiril , m. lein , j. w. g. tisch , j. p. marangos , science * 312 * , 424 ( 2006 ) .
j. muth - bhm , a. becker , and f. h. m. faisal , phys .
* 85 * , 2280 ( 2000 ) ; a. jarn - becker , a. becker , and f. h. m. faisal , phys .
a * 69 * , 023410 ( 2004 ) ; a. requate , a. becker and f. h. m. faisal , phys . rev .
a * 73 * , 033406 ( 2006 ) .
t. k. kjeldsen and l. b. madsen , j. phys .
b * 37 * , 2033 ( 2004 ) ; phys . rev .
a * 71 * , 023411 ( 2005 ) ; phys .
* 95 * , 073004 ( 2005 ) ; t. k. kjeldsen , c. z. bisgaard , l. b. madsen , h. stapelfeld , phys .
a * 71 * , 013418 ( 2005 ) ; c. b. madsen and l. b. madsen , phys .
a * 74 * , 023403 ( 2006 ) .
m. lein , n. hay , r. velotta , j. p. marangos , and p. l. knight , phys .
rev . lett . *
88 * , 183903 ( 2002 ) ; phys .
a * 66 * , 023805 ( 2002 ) ; m. spanner , o. smirnova , p. b. corkum and m. y. ivanov , j. phys .
b * * 37 * * , l243 ( 2004 ) .
b. shan , x. m. tong , z. zhao , z. chang , and c. d. lin , phys .
a * 66 * , 061401(r ) ( 2002 ) ; f. grasbon , g. g. paulus , s. l. chin , h. walther , j. muth - bhm , a. becker and f. h. m. faisal , phys . rev . a * 63 * , 041402(r)(2001 ) ; c. altucci , r. velotta , j. p. marangos , e. heesel , e. springate , m. pascolini , l. poletto , p. villoresi , c. vozzi , g. sansone , m. anscombe , j. p. caumes , s. stagira , and m. nisoli , phys .
a * 71 * , 013409 ( 2005 ) ; t. kanai , s. minemoto and h. sakai , nature * 435 * , 470 ( 2005 ) . k. j. schafer , m. b. gaarde , a. heinrich , j. biegert , and u. keller , phys . rev
. lett . * 92 * , 023003 ( 2004 ) ; m. b. gaarde , k. j. schafer , a. heinrich , j. biegert , and u. keller , phys . rev .
a * 72 * , 013411 ( 2005 ) ; j. biegert , a. heinrich , c. p. hauri , w. kornelis , p. schlup , m. p. anscombe , m. b. gaarde , k. j. schafer and u. keller , j. mod . opt . * 53 * , 87 ( 2006 ) ; j. biegert , a. heinrich , c. p. hauri , w. kornelis , p. schlup , m. p. anscombe , k. j. schafer , m. b. gaarde and u. keller , laser physics * 15 * , 899 ( 2005 ) .
p. johnsson , r. lpez - martens , s. kazamias , j. mauritsson , c. valentin , t. remetter , k. varj , m. b. gaarde , y. mairesse , h. wabnitz , p. salires , ph .
balcou , k. j. schafer , and a. lhuillier , phys .
lett . * 95 * , 013001 ( 2005 ) ; p. johnsson , k. varj , t. remetter , e. gustafsson , j. mauritsson , r. lopez - martens , s. kazamias , c. valentin , ph .
balcou , m. b. gaarde , k. j. schafer , and a. lhuillier , j. mod . opt . * 53 * , 233 ( 2006 ) . in these sfa formulations ,
for exponentially decaying states , @xmath235 and @xmath236 exhibit a singularity , which can be eliminated by being incorporated in the action .
this has not been included in our model , and may lead to further differences between both yields .
we expect , however , these discrepancies to be minimal , as this additional term is slowly varying . for a discussion in the length - gauge context , c.f . , c. figueira de morisson faria and m. lewenstein , j. phys .
b * 38 * , 3251 ( 2005 ) .
|
we perform a detailed analysis of high - order harmonic generation in diatomic molecules within the strong - field approximation , with emphasis on quantum - interference effects .
specifically , we investigate how the different types of electron orbits , involving one or two centers , affect the interference patterns in the spectra .
we also briefly address the influence of the choice of gauge , and of the initial and final electronic bound states on such patterns .
for the length - gauge sfa and undressed bound states , there exist additional terms , which can be interpreted as potential energy shifts .
if , on the one hand , such shifts alter the potential barriers through which the electron initially tunnels , and may lead to a questionable physical interpretation of the features encountered , on the other hand they seem to be necessary in order to reproduce the overall maxima and minima in the spectra .
indeed , for dressed electronic bound states in the length gauge , or undressed bound states in the velocity gauge , for which such shifts are absent , there is a breakdown of the interference patterns . in order to avoid such a problem
, we provide an alternative pathway for the electron to reach the continuum , by means of an additional attosecond - pulse train .
a comparison of the purely monochromatic case with the situation for which the attosecond pulses are present suggests that the patterns are due to the interference between the electron orbits which finish at different centers , regardless of whether one or two centers are involved .
|
in recent years , there is much interest in qgp phase diagram in @xmath5 plane , @xmath6 the temperature and @xmath7 the baryonic chemical potential @xcite,@xcite,@xcite,@xcite .
presently , only one point in the phase diagram is known from lattice qcd simulations , at @xmath80 , the qcd transition is a cross over @xcite at ( pseudo ) critical temperature @xmath9157 ( 3)(3 ) mev @xcite . from theoretical considerations ,
qcd phase transition is expected to be 1st order in baryon dense matter .
one then expects a qcd critical end point ( cep ) , where the transition changes from 1st order to cross - over . to access the qcd phase diagram and cep ,
systematic study of nuclear collisions , as a function of collision energy , has been planned at fair @xcite in the energy range @xmath10=5 - 40 gev . at finite baryon density , fermion determinant is complex and standard technique of monte - carlo importance sampling fails .
several techniques have been suggested to circumvent the problem , ( i ) reweighting @xcite,@xcite , ( ii ) analytical continuation of imaginary chemical potential @xcite , @xcite and ( iii)taylor expansion @xcite,@xcite .
these methods has been used to locate the phase boundary in @xmath5 plane .
the calculations suggest that the curvature parameter ( @xmath3 ) in the expansion , @xmath11 is small @xcite . as an example , in fig.[f1 ] , qcd phase diagram obtained in the analytical continuation method @xcite ( the filled circles ) and in taylor expansion @xcite of ( chiral ) are shown .
both the methods gives nearly identical phase diagram for @xmath12 , curvature parameter is small , @xmath13 . at larger @xmath7 ,
they differ marginally . for comparison , in fig.[f1 ] , chemical freeze - out curve @xcite,@xcite , obtained in statistical model analysis of particle ratios are shown ( the red line ) .
curvature of the chemical freeze - out curve is factor of 4 larger than the curvature in the qcd phase diagram .
small curvature of the qcd critical line , compared to the chemical freeze - out is interesting .
experimental signal of critical end point will get diluted as the deconfined medium produced at the critical end point will evolve longer to reach chemical freeze - out .
fluid will have more time to washout any signature of cep . in the present letter
, we obtain the curvature parameter @xmath3 by analysing experimental data on @xmath0 suppression .
existing data in @xmath14=17.3 gev pb+pb and @xmath14=200 gev au+au collisions , constrain the curvature parameter within a narrow range , @xmath15 .
it is factor of @xmath4 3 larger than in the lattice qcd calculations . to our knowledge , this is the first attempt to obtain qcd critical line from experiment data .
in a deconfined medium ( qgp ) , inter - quark ( color ) potential is screened .
the screening increases with temperature and above a certain temperature , the potential is too weak to bind charm and anti - charm quarks into a bound state .
charmonium states are then suppressed . matsui and
satz @xcite , first predicted the phenomena . over the years , in several experiments , @xmath0 yield in heavy ion collisions has been measured .
measurements at sps ( @xmath1=17.3 gev ) and rhic ( @xmath1=200 gev ) energy validate matsui and satz s predictions . to mimic the onset of deconfining phase transition above a critical energy density and subsequent melting of @xmath0 s ,
qgp motivated threshold model was proposed in @xcite . in the threshold model , @xmath0 suppression
is linked with the local transverse density . if the local transverse density , at the point where @xmath0 is formed , exceed a critical or threshold value , @xmath0 s are melted . in the present paper ,
we use a variant of the threshold model , if the local temperature of the fluid , exceed a threshold value @xmath16 , @xmath0 s are melted .
approximately 30 - 40% of observed @xmath0 s are from feed - down decay of the higher states , @xmath17 and @xmath18 @xcite .
higher states , @xmath17 and @xmath18 are expected to melt at temperature @xmath19,@xmath20 @xmath21 .
lattice qcd calculations indicate that @xmath0 s can survive up to @xmath22 @xcite , higher states survive up to @xmath23 .
for sequential melting of charmonium states , if the melting temperatures @xmath16 and @xmath24 are substantially different , @xmath0 survival probability should show at least two step distribution ( see for example fig.4 in @xcite ) .
no such two step distribution is seen either in @xmath1=200 gev au+au or in @xmath1=17.3 gev pb+pb collisions .
absence of two step structure in @xmath0 survival probability indicate that the melting temperatures of @xmath0 s and higher states are not substantially different . in the following
we assume direct and feed - down @xmath0 s are melted at the common temperature @xmath25 .
in a baryon rich medium , melting temperatures @xmath16 depend on the qgp phase diagram through the critical temperature @xmath26 as in eq.[eq1 ] .
threshold model require the local temperature ( @xmath27 ) and ( baryonic ) chemical potential ( @xmath28 ) of the fluid produced in nuclear collisions .
we propose to obtain them from experimentally measured quantities , e.g. charged particles rapidity density @xmath29 and net baryon rapidity density @xmath30 .
experimentally measured values of @xmath31 @xcite and @xmath32 @xcite,@xcite , in 0 - 5% @xmath1=17.3 and 200 gev collisions , are listed in table.[table1 ] .
the initial entropy density is then determined from the well known relation @xcite , @xmath33 in eq.[eq2 ] , @xmath34 is the initial time , for which we have used the canonical value @xmath34=1.0 fm .
we do note that eq.[eq2 ] is obtained under the assumption of isentropic , longitudinal expansion .
similarly , initial baryon density is obtained as , @xmath35 @xmath36 in eq.[eq2]-[eq3 ] indicate that the entropy density and baryon density in eqs.[eq2],[eq3 ] must be understood as spatially averaged over entropy / baryon density . to introduce spatial dependence ,
we assume , that in an impact parameter @xmath37 collision , initial entropy density and baryon density in the transverse plane is distributed as @xcite , .[table1 ] rapidity density @xmath29 , net baryon density @xmath30 in 0 - 5% @xmath1=17.3 pb+pb and @xmath1=200 gev au+au collisions .
corresponding average temperature ( @xmath27 ) and @xmath28 are also listed . [ cols="^,^,^,^,^ " , ] @xmath38 , \label{eq4 } \\
n_i(x , y ) & = & n_0 \left [ ( 1-f ) \frac{n_{part}(x , y)}{2}+ f n_{coll}(x , y)\right ] \label{eq5}\end{aligned}\ ] ] @xmath39 and @xmath40 is fixed to reproduce the average entropy density @xmath41 and baryon density @xmath42 , as determined from experimental values . @xmath43 and @xmath44 in eq.[eq4],[eq5 ] are the transverse profile of the participant density and binary collision number , obtained in a glauber model .
@xmath45 is the fraction of hard scattering .
it is approximately constant in the energy range @xmath1=17 - 200 gev , @xmath460.12 .
local entropy density ( @xmath47 ) and baryon density ( @xmath48 ) can be converted into local temperature and baryonic chemical potential using an equation of state . in @xcite , a parametric form of qgp equation of state
was given , which matches with the lattice qcd simulation at zero chemical potential and to the known properties of nuclear matter at zero temperature .
entropy density ( @xmath49 ) and baryon density ( @xmath50 ) in parameterised form are , @xmath51 @xmath52 , @xmath53 , @xmath54 . in table.[table1 ] , the average initial temperature and baryonic chemical potential in @xmath1=17.3gev pb+pb and @xmath1=200 gev au+au collisions are listed . from sps to rhic ,
chemical potential decreases by a factor of @xmath415 .
temperature however is increased by @xmath425% . to obtain @xmath0 survival probability ( @xmath55 )
, we randomly distribute @xmath0 s in the reaction plane and count the number of @xmath0 s that survive the threshold condition , @xmath56 .
we note that apart from suppression in deconfined matter , @xmath0 s are suppressed in cold nuclear matter also ( cnm effect ) .
it is important to eliminate the suppression due to cnm effect .
if cold nuclear matter effects are accounted for , the survival probability of @xmath0 s can be obtained as , @xmath57 where @xmath58 is the survival probability in cold nuclear matter and @xmath59 is the survival probability in qgp . in @xmath14=200 gev d+au collisions , @xmath0 production is consistent with cold nuclear matter effect quantified in a glauber model of nuclear absorption with @xmath60 mb @xcite .
@xmath61 @xmath62 mb @xcite in @xmath1=17.3 gev pb+pb collisions .
the present model has two parameter , the threshold temperature for @xmath0 melting , @xmath63 and the curvature parameter @xmath3 of the qcd critical line .
the critical temperature is fixed to @xmath64=160 mev .
the free parameters are obtained by fitting experimental data on @xmath0 production in @xmath1=17.3 gev pb+pb @xcite and @xmath1=200 gev au+au @xcite collisions . in @xmath1=200 gev
au+au collisions , the fluid is essentially baryon free and model predictions do not depend sensitively on the curvature parameter .
the dependence on melting temperature however is strong . in fig.[f2 ] , for threshold temperature @xmath65=1.5 , 1.55,1.6 , 1.65 , and 1.7 , model predictions are compared with the phenix data @xcite for nuclear modification factor in @xmath1=200 gev au+au collisions .
cnm effect is quantified in glauber model of nuclear absorption with @xmath66=3 mb .
experimental data are well explained for the threshold temperature @xmath67 .
the melting temperature @xmath68 is consistent with lattice qcd calculation @xcite .
the blue lines in fig.[f2 ] are model predictions with @xmath3=0.0 and 0.1 , @xmath69
. they can not be distinguished .
curvature parameter is constrained when @xmath0 s are absorbed in baryon rich plasma . in 158 agev pb+pb collisions
fluid is baryon rich and @xmath0 suppression depend sensitively on the curvature parameter @xmath3 . in fig.[f3 ] , centrality dependence of @xmath0 suppression in @xmath1=17.3 gev pb+pb collisions is shown @xcite,@xcite .
the data are shown as the ratio of measured @xmath0 s over the expected @xmath0 s ( in a glauber model of nuclear absorption ) .
upto @xmath70=200 , the ratio is consistent with unity .
suppression in addition to the cnm effect is required in @xmath71 collisions only . in fig.[f3 ] , the solid lines are threshold model predictions with @xmath72 and @xmath3=0.014 , 0.016 , 0.018 , 0.020 , 0.022 and 0.024 respectively .
unlike in @xmath1=200 gev au+au collisions , in @xmath1=17.3 gev pb+pb collisions , @xmath0 survival probability shows sensitive dependence on @xmath3 . for @xmath73 , @xmath0 s
are not suppressed in qgp medium .
best description to the data is obtained with @xmath740.02 ( @xmath75 ) .
description to the data deteriorate for lower or higher curvature .
if uncertainty in the melting temperature is accounted for , pb+pb data fix the curvature parameter within a narrow range , @xmath76 , for @xmath77 . in fig.[f1 ] , qcd critical line obtained with the curvature parameter @xmath76 is shown ( the blue circles with error bars ) .
curvature @xmath78 , obtained from @xmath0 data is @xmath4 3 larger than in lattice qcd simulations @xmath79 ) .
it is also closer to the chemical freeze - out curve .
the result is encouraging with respect to detection of qcd critical end point .
if qcd critical line is close to the chemical freeze - out , possibility of distortion of signals associated with cep , e.g. charge fluctuations etc .
, will be lessened . before we summarise the results some limitations of the model are noted .
experimental information of particle density and baryon density is converted to local temperature and baryonic chemical potential using the eos ( see eq.[eq7],[eq8 ] ) .
the eos is for massless quarks and gluons , adjusted to reproduce zero chemical potential lattice qcd simulations .
how far it represent the real world is uncertain .
the model is also a static model .
it neglect the possibility that @xmath0 s , initially produced in a cold region and survived , at a later time can drift into a hotter region and get melted .
such possibilities , if accounted for would reduce the survival probability . consequently , the present model give a lower limit of curvature parameter @xmath3 .
to summarise , we have shown that in a threshold model , @xmath0 suppression is sensitive to the curvature parameter of the qcd phase diagram . in threshold model ,
@xmath0 suppression depend on the local temperature and chemical potential .
a model is proposed to obtain local temperature and chemical potential , from experimentally determined quantities e.g. rapidity density , net baryon rapidity density .
qcd critical line with curvature parameter @xmath3=0.018 - 0.023 is consistent with experimental centrality dependence of @xmath0 suppression in @xmath1=17.3 gev pb+pb and @xmath1=200 gev au+au collisions .
@xmath3=0.018 - 0.023 is approximately 3 times larger than that obtained in lattice qcd simulations .
k. fukushima and t. hatsuda , rept .
phys . *
74 * , 014001 ( 2011 ) [ arxiv:1005.4814 [ hep - ph ] ] .
k. fukushima , phys .
d * 77 * , 114028 ( 2008 ) [ erratum - ibid .
d * 78 * , 039902 ( 2008 ) ] [ arxiv:0803.3318 [ hep - ph ] ] .
m. a. stephanov , pos lat * 2006 * , 024 ( 2006 ) [ hep - lat/0701002 ] .
m. asakawa and k. yazaki , nucl .
a * 504 * , 668 ( 1989 ) .
p. senger , nucl .
phys . a * 862 - 863 * , 139 ( 2011 ) .
z. fodor and s. d. katz , jhep * 0203 * , 014 ( 2002 ) [ hep - lat/0106002 ] .
z. fodor and s. d. katz , jhep * 0404 * , 050 ( 2004 ) [ hep - lat/0402006 ] .
p. de forcrand and o. philipsen , nucl .
b * 642 * , 290 ( 2002 ) [ hep - lat/0205016 ] .
m. delia and m. -p .
lombardo , phys .
d * 67 * , 014505 ( 2003 ) [ hep - lat/0209146 ] . c. r. allton , s. ejiri , s. j. hands , o. kaczmarek , f. karsch , e. laermann and c. schmidt , phys .
d * 68 * , 014507 ( 2003 ) [ hep - lat/0305007 ] .
r. v. gavai and s. gupta , phys .
d * 68 * , 034506 ( 2003 ) [ hep - lat/0303013 ] .
o. philipsen , prog .
* 174 * , 206 ( 2008 ) [ arxiv:0808.0672 [ hep - ph ] ] .
o. kaczmarek , f. karsch , e. laermann , c. miao , s. mukherjee , p. petreczky , c. schmidt and w. soeldner _
et al . _ ,
d * 83 * , 014504 ( 2011 ) [ arxiv:1011.3130 [ hep - lat ] ]
. f. becattini , j. manninen and m. gazdzicki , phys .
c * 73 * , 044905 ( 2006 ) [ hep - ph/0511092 ] .
j. cleymans , h. oeschler , k. redlich and s. wheaton , j. phys .
g * 32 * , s165 ( 2006 ) [ arxiv : hep - ph/0607164 ] .
t. matsui and h. satz , phys .
b * 178 * , 416 ( 1986 ) .
j. p. blaizot , m. dinh and j. y. ollitrault , phys .
lett . * 85 * , 4012 ( 2000 ) [ arxiv : nucl - th/0007020 ] . j. p. blaizot and j. y. ollitrault , j / psi suppression in pb pb collisions : a hint of quark - gluon plasma phys . rev
. lett .
* 77 * , 1703 ( 1996 ) [ arxiv : hep - ph/9606289 ] .
h. satz , j. phys .
g * 32 * , r25 ( 2006 ) [ arxiv : hep - ph/0512217 ] .
s. s. adler _ et al .
_ [ phenix collaboration ] , phys .
c * 71 * , 034908 ( 2005 ) [ erratum - ibid .
c * 71 * , 049901 ( 2005 ) ] [ nucl - ex/0409015 ] .
h. appelshauser _ et al . _
[ na49 collaboration ] , phys .
* 82 * , 2471 ( 1999 ) [ arxiv : nucl - ex/9810014 ] . i. g. bearden _ et al . _
[ brahms collaboration ] , phys . rev .
lett . * 93 * , 102301 ( 2004 ) [ arxiv : nucl - ex/0312023 ] .
r. c. hwa and k. kajantie , phys .
d * 32 * , 1109 ( 1985 ) .
p. f. kolb and u. w. heinz , in * hwa , r.c .
et al . : quark gluon plasma * 634 - 714 [ nucl - th/0305084 ] . j. i. kapusta , phys .
c * 81 * , 055201 ( 2010 ) [ arxiv:1001.3650 [ nucl - th ] ] .
r. vogt , acta phys . hung .
a * 25 * , 97 ( 2006 ) [ arxiv : nucl - th/0507027 ] .
r. arnaldi [ na60 collaboration ] , nucl .
a * 830 * , 345c ( 2009 ) [ arxiv:0907.5004 [ nucl - ex ] ] .
n. brambilla _ et al .
_ , eur . phys .
j. c * 71 * , 1534 ( 2011 ) [ arxiv:1010.5827 [ hep - ph ] ] .
a. adare [ phenix collaboration ] , arxiv : nucl - ex/0611020 .
s. digal , p. petreczky and h. satz , arxiv : hep - ph/0110406 .
|
qcd phase diagram is obtained by analysing centrality dependence of @xmath0 suppression in @xmath1=17.3 pb+pb and @xmath1=200 gev au+au collisions .
s produced in initial interactions are assumed to dissolve if local temperature exceeds a threshold temperature .
the threshold temperature depends on the ( local ) fluid temperature and baryonic chemical potential , which are obtained , under certain assumptions , from experimentally determined quantities e.g. rapidity density , net baryon density .
qcd critical line with curvature parameter @xmath2 is consistent with experimental @xmath0 data .
@xmath3 is factor of @xmath43 larger than in lattice qcd calculations .
|
thomas - fermi theory @xcite has proven to be a valuable tool for the qualitative understanding of atoms and molecules . in semiconductor quantum dots @xcite , which are a kind of artificial thomson atoms with many possibilities for fundamental research and technical applications
, thomas - fermi theory was shown to agree qualitatively and even quantitatively with a more ellaborated approach like density functional theory @xcite , being asymptotically exact in the limit of large electron numbers @xcite . from the computational point of view ,
thomas - fermi theory with minor corrections is able to reproduce the ground - state energy of electrons in a quadratic potential @xcite at the same level of accuracy of other semiclassical or semianalytic approaches like large-@xmath8 expansions @xcite or two - point pad approximants @xcite . in the present paper
, we would like to stress on a less studied aspect of thomas - fermi theory : the highly non - trivial scaling relations following from it .
we show that the number of electrons , @xmath0 , and the coupling constant , @xmath2 , enter the ground - state energy in a scaled form .
we perform extensive configuration interaction calculations for quantum dots with @xmath9 in order to verify this scaling .
in addition , on the basis of the numerical results , we show that similar scaling relations are valid for the number of excited states in intraband and interband excitations . in this way
, a universal parametrization of the density of energy levels in quantum dots is provided .
we start with the hamiltonian of a two - dimensional parabolic quantum dot charged with @xmath0 electrons . in oscillator units
, the hamiltonian can be written as : @xmath10 the only approximations made in writting eq .
( [ eq1 ] ) are the effective - mass description of electrons , the inclusion of an effective low - frequency dielectric constant , @xmath11 , to model the medium , and the description of confinement by means of a harmonic - oscillator potential .
these approximations are very common and well sustained @xcite .
the coupling constant @xmath12 is the ratio of coulomb and harmonic - oscillator characteristic energies .
the fact that the number of electrons may enter the energy in a scaled combination with @xmath2 is , however , not trivial .
let us write the thomas - fermi energy functional @xcite for the present problem : @xmath13 where @xmath14 is the ( surface ) density at point @xmath15 , and @xmath16 is a numerical constant .
the above functional should be extremized under the constraint @xmath17 now , it is easy to realize that we can scale @xmath15 and @xmath18 in such a way that the l.h.s . of eq .
( [ constraint ] ) becomes one , and a factor @xmath19 is extracted from the r.h.s . of eq .
( [ eq2 ] ) . as a result
, we get the following relation for the ground - state energy in the thomas - fermi approximation : @xmath20 notice that the scaled thomas - fermi equations depend on a single parameter , @xmath21 , which combines in a particular way the coupling constant and the number of electrons .
we first provide an analytical expression for @xmath4 based on two - point pad approximants @xcite in the large-@xmath0 limit .
it shows that the scaling predicted by thomas - fermi theory is quite general and compatible with true quantum effects .
let us recall the definition of the @xmath22 pad approximant for the ground - state energy , given in ref . ,
which interpolates between the @xmath23 ( perturbation theory ) and @xmath24 ( wigner `` crystal '' ) expansions : @xmath25 we use the large-@xmath0 asymptotic expressions for the coefficients @xcite , which lead to the following estimation for the ground - state energy : @xmath26 in order to verify the universal relation ( [ eq5 ] ) we performed extensive configuration interaction calculations for charged quantum dots . in these calculations
, we follow standard procedures of quantum chemistry @xcite or nuclear physics @xcite .
( color online ) different contributions to the ground - state wave function entering the configuration interaction calculation . ]
the starting point is the hartree - fock solution of the problem .
then a basis of functions made up from ( i ) the hartree - fock state , @xmath27 , ( ii ) one - particle one - hole ( 1p1h ) excitations , that is @xmath28 , and ( iii ) two - particle two - hole ( 2p2h ) excitations , i.e. @xmath29 , is used in order to diagonalize the hamiltonian .
notice that @xmath30 are single - particle states above the fermi level , and @xmath31 are states below the fermi level . a schematic representation is given in fig .
[ fig1 ] . in the hilbert subspace with the same quantum numbers of the hartree - fock state
, the electronic hamiltonian takes the form : @xmath32 where @xmath33 is the hartree - fock energy , @xmath34 is the tamm - dankoff matrix , @xmath35 , @xmath36 , and @xmath37 .
@xmath38 and @xmath39 are , respectively , the transposes of matrices @xmath8 and @xmath40 .
explicit matrix elements are given in appendix [ appa ] for completeness . in sectors with quantum numbers others than the hartree - fock state , the first row and column of matrix ( [ matrixh ] ) should be dropped .
an energy cutoff of 3 @xmath3 in the excitation energy is used to control the dimension of the hamiltonian matrix .
the estimated error in the ground - state energy is below 0.2 % .
( color online ) scaling of the ground - state energy in medium - sized dots . the large-@xmath0 expression for the pad estimate , eq .
( [ eq5 ] ) , is shown as a solid line . ]
we computed the ground - state energy of dots with @xmath41 , 30 , 42 , 56 , 72 and 90 , and confinement strengths @xmath42 , 12 and 18 mev .
notice that these are closed shell quantum dots with ground - state angular momentum and spin quantum numbers @xmath43 .
gaas parameters , @xmath44 and @xmath45 , were used .
we performed the calculations for three - dimensional dots in which the confinement along the symmetry axis ( the @xmath46 axis ) is modelled by a rigid - wall well of width , @xmath47 nm .
the constant @xmath48 , where @xmath49 , was removed from the ground - state energy .
the results are depicted in fig .
[ fig2 ] ( dots ) along with the large-@xmath0 pad estimate given by eq .
( [ eq5 ] ) ( solid line ) .
scaling of the ground - state energy is apparent .
the maximum deviations with respect to the pad estimate are below 10 % for the smallest dots with @xmath41 .
notice that , for the parameters used in the calculations , the scaled variable @xmath50 takes values around 1 , i.e. in the transition interval from weak to strong coupling @xcite . in order to test the whole interval , we use additional control cases : one of them deep in the strong coupling regime ( @xmath51 , @xmath52 mev ) , and the other in the weak coupling region ( @xmath41 , @xmath53 mev ) .
they also fit the pad estimate .
we now turn to the intraband excitations . for simplicity , we consider the excited states of the closed - shell quantum dots studied above .
we restrict the analysis to sectors with the same quantum numbers as the ground state , @xmath43 , in such a way that the ground and excited states come out from the same calculation . a sample of the results is shown in fig .
[ fig3 ] ( a ) for the 42-electron dot with confinement @xmath42 mev .
first , we notice that the excitation gap , which is @xmath54 in the noninteracting @xmath23 limit , is renormalized by coulomb interactions to around 6 mev , that is @xmath55 . in the opposite ,
@xmath24 , limit , the excitation spectrum is that of a big ( wigner ) molecule , whose small - oscillation frequencies are independent of @xmath2.@xcite the lowest of these frequencies , i.e. that one determining the gap , should go to zero for large @xmath0 in order to meet the acoustic phonon of the wigner lattice
. then , we can look for a simple interpolation formula in order to fit the numerical data for the excitation gap : @xmath56 the parameter @xmath57 appears to be very close to 1/4 , thus we fixed it to 1/4 and fit again the data in order to obtain @xmath58 and @xmath59 .
the results are shown in fig .
[ fig3 ] ( b ) as a function of @xmath21 .
we stress that this is only a useful representation because @xmath60 does not scales with @xmath46 , even though fig .
[ fig3 ] ( b ) shows an approximate scaling for intermediate couplings . for the parameters @xmath58 and @xmath59 , we get : @xmath61 , @xmath62 .
the result of the fit is excellent , with maximum deviations below 10 % , the same as for the ground - state energy .
notice also that @xmath63 goes to a universal value , @xmath64 , in the strong - coupling limit , @xmath24 .
expressions similar to eq .
( [ eq6 ] ) for the gap to the first excited state should be valid in other angular momentum and spin sectors , and also for the energy of collective states ( spin- and charge - density excitations ) .
( color online ) ( a ) the number of excited states in the 42-electron quantum dot as a function of the excitation energy .
the confinement strength is @xmath42 mev .
( b ) the excitation gap to the first excited state as a function of the scaled variable @xmath21 .
( c ) the temperature parameter , @xmath65 , in scaled variables .
fits in ( b ) and ( c ) correspond to eqs .
( [ eq6 ] ) and ( [ eq7 ] ) . ] the second point to notice in fig .
[ fig3 ] ( a ) is the exponential growth of the number of states for excitation energies above 6 mev .
this simple exponential dependence on excitation energy is known in nuclear physics as the constant temperature approximation ( cta ) : @xcite @xmath66 it seems to be a quite general property of the excitation spectrum of quantum systems .
we verified it , for instance , in the energy spectrum of small quantum dots in strong magnetic fields @xcite .
we fit the numerical data corresponding to the first 25 excited states of the quantum dots mentioned above in order to extract the `` temperature '' parameter , @xmath65 , in eq .
( [ eq3 ] ) .
we took the first excited state as the reference of energy .
the next 24 states are only a few ( 1 to 4 ) mev above the first excited state . in order to deduce the universal properties of @xmath65
let us recall the @xmath23 and @xmath24 asymptotic regimes . in the @xmath24 limit
, we expect for @xmath65 a behaviour similar to @xmath60 , that is @xmath67 should take a universal value .
on the other hand , in the @xmath23 limit the excitation energies ( with respect to the first excited state ) are proportional to @xmath2 , thus we may write a simple interpolation formula for the temperature parameter : @xmath68 where @xmath69 , @xmath70 .
the quality of the fit is also very good as can be seen in fig .
[ fig3 ] ( c ) .
( color online ) ( a ) the interband ( excitonic ) excitations in the quantum dot with 42 electrons and @xmath71 mev . in the @xmath72 axis
the reference energy is the first excitonic state .
( b ) scaling of the temperature parameter for the lowest - energy excitonic states . ]
next , we study the interband excitonic excitations of dots with @xmath41 , 30 and 42 , and @xmath42 , 12 and 18 mev . the two control cases in the strong and weak coupling regimes are also included .
a basis for excitonic states in these dots is build up in the following way : ( i ) states with one additional electron above the fermi level in the conduction band and a hole in the valence band , @xmath73 , ( ii ) states with two additional electrons above the fermi level in the conduction band , a hole in the conduction band , and a hole in the valence band , @xmath74 .
details of the computational scheme in the present case can be found in ref . .
the hartree - fock single - particle states for holes are obtained from the kohn - luttinger hamiltonian in the presence of the electronic background . in our model calculations ,
the oscillator lengths for electrons , heavy holes and light holes are equal .
kohn - luttinger parameters for gaas are used @xcite . with a cutoff in the excitation energy of 3 @xmath3 , the basis dimension
is reduced to around 5000 .
we show in fig .
[ fig4 ] ( a ) a typical spectrum of excitonic excitations , corresponding to a dot with @xmath51 and @xmath71 mev .
the states are characterized by the total angular momentum @xmath75 , and total electronic spin projections , @xmath76 .
@xmath77 and @xmath78 are orbital angular momenta of electrons and holes , respectively , and @xmath79 is the band momentum of holes along the @xmath46 axis . in fig .
[ fig4 ] ( a ) , the @xmath72 axis excitation energies are measured with respect to the first excitonic state .
the lowest 40 states shown in the figure follow two different cta fits , corresponding to @xmath80 mev and @xmath81 mev .
the discontinuity in the slope seems to be a quite general fact @xcite related to different mechanisms of formation of the states .
we use the lowest 10 states in order to find a temperature parameter in the studied dots , and a law like eq .
( [ eq7 ] ) to fit the data .
the found parameters are : @xmath82 , @xmath83 .
it is remarkable that the fit performs very good , as shown in fig .
[ fig4 ] ( b ) , signaling that the electronic background determines global properties of the excitonic excitation spectrum . unlike the intraband excitations
, however , we expect the parameters @xmath84 and @xmath85 to depend weakly on the dot material ( gaas in this case ) because of the kohn - luttinger hamiltonian entering the calculations .
we shall test in the future to what extent this happens .
( color online ) structure of the biexcitonic wave function in the configuration interaction calculations . ]
finally , let us consider the interband biexcitonic excitations in our medium - sized dots .
the basis functions for the configuration interaction calculations contains two additional electrons above the fermi level in the conduction band , and two holes in the valence band , @xmath86 , with @xmath87 and @xmath88 .
a schematic representation is given in fig .
[ fig5 ] . in appendix
[ appb ] , we give explicit expressions for the hamiltonian matrix elements .
( color online ) ( a ) the interband ( biexcitonic ) excitations in the quantum dot with 42 electrons and @xmath42 mev . in the @xmath72 axis
the reference energy is the first biexcitonic state .
( b ) scaling of the temperature parameter for the biexcitonic excitations . ] with a cutoff in the excitation energy of @xmath54 , the hamiltonian matrix has dimension around 3000 . we draw in fig . [ fig6 ] ( a ) the spectrum of biexcitonic excitations in the dot with 42 electrons and @xmath42 mev .
the quantum numbers of the states shown in the figure are : @xmath89 , @xmath90 .
we see that in a single cta fit we may comprise the first 35 states .
these first states are to be used in the determination of the temperature parameter .
the scaling of @xmath91 is shown in the lower panel of fig .
notice the power of @xmath0 , which is now 1/2 instead of 1/4 .
we verified that , by taking this power , the dispersion of points is reduced notably .
thus , we fit the data with the formula : @xmath92 where @xmath93 , @xmath94 .
the quality of the fit is very good .
the same comment about the dependence of the parameters on the dot material , made above for the excitonic states , applies in the present situation .
in conclusion , we have performed extensive configuration interaction calculations for medium - sized quantum dots in order to verify universal relations for the ground - state energy and the intraband and interband ( excitonic and biexcitonic ) excitation spectrum .
the coefficients in the r.h.s . of eqs .
( [ eq5],[eq6],[eq7 ] ) do not depend even on the material the dots are made of . on the other hand , the coefficients @xmath84 , @xmath85 , @xmath95 , and @xmath96 , we believe , are specific for gaas , but independent of @xmath0 and @xmath3 .
the work can be extended in many directions .
we may try to parametrize in a universal way the correlation energy , the excitonic and biexcitonic binding energies , the excitation gaps to different angular momentum and spin sectors , the energy of collective ( plasmonic ) excitations , etc . on the other hand ,
more efforts towards the undestanding of the empirical relations obtained for the @xmath65 parameters are needed .
research along these lines is in progress .
part of this work was performed using the computing facilities of the abdus salam ictp , trieste , italy .
the authors acknowledge support by the caribbean network for quantum mechanics , particles and fields ( ictp ) and by the programa nacional de ciencias bsicas ( cuba ) .
in eq . ( [ matrixh ] ) , @xmath97 is the hartree - fock total energy @xcite : @xmath98 where @xmath99 is the fermi level , @xmath100 is the hartree - fock energy of the electron state @xmath101 , @xmath102 is the energy of 2d oscillator states , characterized by the quantum numbers @xmath103 ( radial number ) , @xmath104 ( angular momentum ) , and @xmath105 ( spin projection ) .
that is : @xmath106 the state @xmath101 is expanded in oscillator states as follows : @xmath107 in the studied closed - shell dots , @xmath104 and @xmath105 are good quantum numbers of @xmath108 , and the above sum run only over @xmath103 . on the other hand , in eq .
( [ matrixh ] ) @xmath109 is the tamm - dankoff matrix@xcite : @xmath110 where the antisymmetrized coulomb matrix elements are defined as : @xmath111 coulomb matrix elements @xmath112 are computed in terms of matrix elements among oscillator states by using the expansion ( [ expansion ] ) . finally , matrices @xmath8 , @xmath40 and @xmath113 are explicitly written as : @xmath114 @xmath115 @xmath116
in the biexcitonic sector , the hamiltonian matrix elements take the form : the total hamiltonian , @xmath118 , in addition to the terms of eq .
( [ eq1 ] ) , now includes the single - particle energy of holes , electron - hole , and hole - hole interactions .
the hartree - fock electron and hole states should be expanded in oscillator functions when coulomb matrix elements are to be computed . for hole states , coming from a kohn - luttinger hartree - fock problem
, we have the expansion : where @xmath103 and @xmath104 are oscillator quantum numbers , @xmath120 is hole ( band ) angular momentum projection , and @xmath121 labels sub - band states in the well .
the relatively large number of terms entering the expansion ( [ expansionh ] ) makes the calculation of coulomb matrix elements for holes lengthy .
thomas , proc .
23 , 542 ( 1927 ) ; e. fermi , rend .
accad . naz .
lincei 6 , 602 ( 1927 ) .
kirzhnits , yu.e .
lozovik , and g.v .
shpatakovskaya , sov .
18 , 649 ( 1975 ) .
lieb , rev .
53 , 603 ( 1981 ) .
l. spruch , rev .
63 , 1512 ( 1991 ) .
l. kouwenhoven and c. marcus , phys .
world 11 ( 6 ) , 35 ( 1998 ) .
l. jacak , p. hawrylak , and a. wojs , _ quantum dots _
( springer - verlag , berlin , 1998 ) . ll . serra and a. puente ,
j. d 14 , 77 ( 2001 ) .
d. ullmo , h. jiang , w. yang , and h.u .
baranger , phys . rev .
b 70 , 205309 ( 2004 ) .
lieb , j.p .
solovej , and j. yngvason , phys . rev .
b 51 , 10646 ( 1995 ) .
a.gonzalez and r. gonzalez , sov .
phys . - lebedev inst .
reports , 3 , 36 ( 1990 ) .
a. belov , yu.e .
lozovik , and a. gonzalez , phys .
lett . a 142 , 389 ( 1989 ) .
a. gonzalez , b. partoens and f.m .
peeters , phys . rev .
b 56 , 15740 ( 1997 ) .
cramer , _ essentials of computational chemistry : theories and models _ ( wiley , chichester , 2006 ) .
p. ring and p. schuck , _ the nuclear many - body problem _ ( springer - verlag , new - york , 1980 )
. t. ericson , adv .
phys . 9 , 425 ( 1960 ) ; a. gilbert , a.g.w .
cameron , can .
43 , 1446 ( 1965 ) .
a. gonzalez and r. capote , phys .
b 66 , 113311 ( 2002 ) .
a. odriazola , a. delgado , and a. gonzalez , phys . rev .
b 78 , 035329 ( 2008 ) .
landolt - bornstein , _ numerical data and functional relationship in science and technology _ , group iii , volume 17 ( springer - verlag , berlin , 1982 ) .
|
in a two - dimensional parabolic quantum dot charged with @xmath0 electrons , thomas - fermi theory states that the ground - state energy satisfies the following non - trivial relation : @xmath1 , where the coupling constant , @xmath2 , is the ratio between coulomb and oscillator ( @xmath3 ) characteristic energies , and @xmath4 is a universal function .
we perform extensive configuration interaction calculations in order to verify that the exact energies of relatively large quantum dots approximately satisfy the above relation .
in addition , we show that the number of energy levels for intraband and interband ( excitonic and biexcitonic ) excitations of the dot follows a simple exponential dependence on the excitation energy , whose exponent , @xmath5 , satisfies also an approximate scaling relation _ a la _ thomas - fermi , @xmath6 .
we provide an analytic expression for @xmath4 , based on two - point pad approximants , and two - parameter fits for the @xmath7 functions .
|
data has been accumulating over the last several years on an increasing number of galaxies in terms of measurements of the bulge velocity dispersion @xmath0 , mass of the central supermassive blackhole ( smbh ) @xmath1 , and the circular velocity , @xmath2 , of the host galaxy ( mcconnell & ma 2013 ; hu 2009 ; graham 2008 ; hu 2008 ; ho 2007 ; courteau et al .
2007 ; pizella et al . 2005 ; baes et al . 2003
; ferrarese 2002 ; tremaine et al .
2002 ; onken et al 2004 ; peterson et al . 2004 ; nelson et al . 2004 ; merrit and ferrarese 2001 ; bedregal et al .
2006 ; bernardi et al . 2002 ; kronawitter et al .
2000 ; palunas et al .
2000 ; prugniel et al 2001 ; verheijen et al .
2001 ) . with the increase in the number of the galaxies for which @xmath0 , @xmath3 , and @xmath1 have been measured , it is becoming possible to ask , and attempt to answer , questions about the processes that govern the formation of galaxies and their central smbhs , and the symbiotic relationship between an smbh and its host galaxy .
one way to approach these issues is to compare observed correlations , or lack thereof , between @xmath2 , @xmath0 , and @xmath1 with theoretical predictions obtained from models of galaxy / smbh formation ( e.g. , di matteo et al . 2008 ; di matteo et al . 2003 ; haehnelt & kauffmann 2000 ; kauffmann & haehnelt 2000 ) .
the observed @xmath4 relation is now on a firm basis ( beifiori et al .
2012 ; gultenkin et al .
2009 ; hu 2009 ; graham 2008 ; hu 2008 ; tremaine et al .
2002 ; ferrarese & merritt 2000 ; gebhardt et al .
2000 ) and it points to a common history between the smbh and the spheroidal component of galaxies .
moreover , it was recently found that @xmath2 and @xmath0 are correlated ( courteau et al .
2007 ; baes et al .
2003 , ferrarese 2002 ) .
however , ho ( 2007 ) by a using a sample of 792 galaxies have shown that the @xmath5 relation has a very large scatter , and it depends on galaxy morphology .
ferrares ( 2002 ) investigated the possibility of the masses of the supermassive blackhole being correlated with the total gravitational mass of the host galaxy or the mass of the dark matter halo .
the author found a tight correlation between @xmath3 and @xmath0 for a sample of 36 galaxy , and by using the well - known @xmath6 relation , and connecting the mass of the dark matter halo with the circular velocity of the galaxy , report a correlation between the mass of the supermassive blackhole and the circular velocity of the host galaxy .
baes et al .
( 2003 ) reported a relation between @xmath1 and @xmath2 .
the rationale behind this correlation is that both quantities depend on the mass of the dark matter halo .
a heavier halo will result in a deeper potential well that will give rise to a higher @xmath1 , and at the same time lead to a higher @xmath2 .
zasov et al .
( 2005 ) , on the other hand , found a very weak correlation between @xmath1 and @xmath3 .
however , the @xmath1 values in both baez et al .
( 2003 ) and zasov et al .
( 2005 ) , as well as in ferrarese ( 2002 ) , were obtained from the @xmath4 relation .
ho et al . (
2008 ) report a correlation between @xmath1 and @xmath3 for a sample of active galaxies , where the smbh masses were measured using the virial method ( kaspi et al .
2000 ; greene & ho 2005 ) .
more recently , kormendy et al .
( 2011 ) used dynamical @xmath1 measurements to find that @xmath7 relation is very weak in a sample of 25 spiral galaxies .
the same result is confirmed in beifiori et al .
( 2012 ) who used a large sample , 105 galaxies , but the used @xmath1 s are a mix of dynamical measurements from gultekin et al .
( 2009a ) and upper limits from beifiori et al .
( 2009 ) , with the upper limits taken as surrogate of black mass .
* more recently , sun et al .
( 2013 ) analyzed the @xmath7 relation for 22 galaxies with @xmath2 determined from observations , and came up with a broad relation with large instrinsic scatter .
* given the controversy surrounding the @xmath7 relation and its strength , using relations to find blackholes masses will only introduce additional scatter and makes it more difficult to reach firm conclusions for or against a genuine relation .
most of the previous studies that have investigated the relation between @xmath1 and the circular velocity did so indirectly since they `` measured '' the smbh masses through the @xmath4 relation for galaxies with measured @xmath2 and @xmath0 .
the danger in the indirect approach is that one is at the risk of masking , or leading to , a correlation between @xmath7 through the @xmath8 and @xmath4 , and it necessarily ignores bulge - less spirals with smbhs , further masking the true connection between the smbh and the dark matter halo .
the aim of this paper is to present a direct study of the nature of the @xmath9 relation for * 53 * galaxies with dynamically measured @xmath1 .
this is the largest sample to date of galaxies that have these two properties available .
we also present , as a by - product of our study , the @xmath4 relation based on * 89 * galaxies .
we are revisiting the question of the @xmath9 relation with a slightly `` cleaner '' sample of blackhole masses .
it is our hope here that by using dynamical @xmath1 s we will be able to probe directly the intrinsic relations between a smbh and the dark matter halo of the host galaxy .
our main objective is to build a sample of galaxies that have dynamical @xmath1 and @xmath3 measurements . in order to make sure that we are not missing any objects we collected a large sample of 376 galaxies of all types for which measurements of @xmath1 , @xmath2 , and/or @xmath0 exist .
this large sample was culled from papers that present measurements of smbh masses ( mcconnell & ma 2013 ; gultenkin et al .
2009a ; pastorini et al 2007 ; de fransesco et al .
2006 ; ferrarese & ford 2005 ; onken et al . 2004 ; peterson et al . 2004 ; nelson et al . 2004 ; gebhardt et al .
2000 ; kaspi et al .
2000 ; richstone 1998 ) , papers that study the @xmath4 relation ( hu 2009 ; gultekin et al . 2009a , graham 2008 ; hu 2008 ; tremaine et al .
2002 ; gebhardt et al .
2000 ; ferrarese & merritt 2000 ) , and papers that deal with the @xmath10 and/or @xmath7 relationships ( kormendy et al . 2011 ; courteau et al .
2007 ; ho 2007 ; pizella et al . 2005 ; baes et al . 2003 ; ferrarese 2002 ) .
we found 342 galaxies with @xmath0 measurements , 269 galaxies with @xmath2 , and 125 galaxies with @xmath1 .
we reject 35 out of the 125 blackhole masses since they are determined via reverberation mapping , which is caliberated to the @xmath4 relation ( onken et al . 2004
; peterson et al . 2004 ) .
the intersections between these subsets resulted in 53 galaxies that have dynamical @xmath1 and @xmath3 , 89 galaxies with both dynamical @xmath1 and @xmath0 , and 251 galaxies that have both @xmath0 and @xmath2 .
when error bars on a quantity for a particular galaxy are not available we assign an error by multiplying that quantity by the average percent error derived for that quantity from all galaxies that have errors . out of 342 galaxies that have @xmath0 values , 332 have errors listed , and 251 out of 269 galaxies have errors on @xmath3 .
the average percent error on @xmath0 is @xmath11% and on @xmath2 is @xmath12% .
all available smbh dynamical mass measurements have error bars quoted in the literature .
the errors bars on @xmath1 are not symmetric .
we calculate the average of the interval and assign this error to the data point when peforming the fits .
we list in table 1 our final sample of 89 galaxies with dynamically - determined smbh masses , stellar velocity dispersions , and circular velocities ( 53 out of 89 ) .
all of the @xmath0 values are from hyperleda@xmath13 ( paturel et a. 2003 ) except for ic 2560 ( garcia - rissman et al .
2005 ) and cygnus a ( graham 2008 ) , and mw ( baes et al .
2003 ; ferrarese 2002 ) .
most of the blackhole masses are from hu ( 2009 ) and mcconnell & ma ( 2013 ) except for ngc 1300 ; ngc 2748 ; ngc 2778 ; ngc 4342 ; ngc 4374 ; ngc 4945 ; ngc 7582 ( graham 2008 ) , ngc 4303 ( pastorini 2007 ) , ngc 4486 kormendy et al .
( 1996 ) , ngc 4594 ( kormendy et al .
the distances in table 1 are from the above - mentioned papers .
we include in table 1 references to the original papers and the methods used to measure the blackhole masses .
we use circular velocities from optical rotation curves whenever available : ngc 2787 , ngc 1023 , ngc 3115 , ngc 3384 , and ngc 4649 ( neistein et al .
1999 ) ; ngc 1399 , ngc 3379 , ngc 4374 , ngc 4486 , and ngc 4486b ( kronawitter et al .
2000 ) , and ngc 5846 ( pizzella et al .
we use circular velocities derived from line widths as listed in courteau et al .
( 2007 ) for : mw , ngc 224 , and ngc 4258 ( ferrarese et al . 2002 ) ; ngc 2974 ( pizzella et al .
2005 ) ; and ngc 3031 , ngc 3227 , ngc 4303 , and ngc 4594 ( prugniel et al .
we use at face value the circular velocities in hyperleda for the remaining 34 galaxies .
they are also based on line widths , which brings the number of galaxies in our sample with @xmath3 derived from line widths to 42 .
we note here that the results that we arrive at below do not depend on the origin of the circular velocities .
studying the correlations involves fitting straight lines in log - log space : @xmath14 .
the results of the fitting have been known to depend in part on the details of the linear regression analysis ( tremaine et al .
for the sake of definiteness , we limit ourselves to the prescription of press et al .
( 1992 ) for data with symmetric error bars on both variables . *
the errors on @xmath15 and @xmath16 are @xmath17 fitting errors when @xmath18 differs by unity from its mininum value . *
a shortcoming of this approach is that it does not take into account the presence of intrinsic dispersion in the data .
we follow the quick fix offered by tremaine et al .
( 2002 ) : we add in quadrature a parameter to the error of the y - coordinates ( gebhardt et al .
this parameter is a measure of the intrinsic dispersion .
its value is adjusted by hand to lead a reduced chi - squared of unity . *
this prescription ( tremaine et al . 2002 ) is essentially implemented in fitexy and , its more advanced variant mpfitexy ( markwardt 2009 ; williams et al . 2010 ) , which is what we use in this paper .
* we plot in fig .
1 @xmath19 against @xmath20 for 53 galaxies , made up of 13 ellipticals , 16 lenticulars , and 24 spirals .
( table 1 ) .
it is clear that 3 galaxies whose @xmath21 km s@xmath22 do nt fall in the region where the other data points cluster .
these 3 galaxies are ic 1459 ( e3 ) , ngc 5252 ( s0 ) , and ngc 3608 ( e2 ) .
we do not include them in our subsequent analysis when we fit : @xmath23to the data points in fig . 1 . a casual inspection of fig .
1 shows that the correlation between the @xmath1 and @xmath3 , if there is any , is very weak .
taken separately , galaxies belonging to the same morphological type form essentially scatter plots and at best barely show a hint of a correlation . a visual inspection in fig .
1 that does not distinguish between the galaxy types can be misleading .
an apparent correlation will appear to be present if one is not careful .
the spirals , represented as blue crosses , have on average lower blackhole masses and velocities than the ellipticals and the lenticulars .
placed on the same plot , this offset between spiral and e / s0 galaxies makes one to believe that there is a correlation . to be more quantitative ,
we use the procedure outlined above to obtain : @xmath24 with an intrinsic scatter of 0.75 dex .
the intrinsic scatter is very large and so is the error on the slope .
this leads to conclude that there is no correlation between the mass of supermassive blackhole and the circular velocity of the host galaxy .
* this result is broadly consistent with those of sun et al .
( 2013 ) , beifiori et al .
( 2012 ) , and kormendy et al .
2011*. previous authors have studied the relation between @xmath1 and @xmath3 and their conclusions are varied .
baes et al .
( 2003 ) reported a slope of @xmath25 for a sample of 40 galaxies .
however , baes et al . ( 2003 ) determined the masses of the blackholes indirectly by using the well - established @xmath4 relation of tremaine et al .
they combined their @xmath5 relation with the @xmath4 relation of tremaine et al .
( 2002 ) to derive a @xmath9 relation .
ho ( 2007 ) using a sample of 792 galaxies has shown that the @xmath5 relation has appreciable variation depending on host galaxy properties
. ho ( 2007 ) cautions against replacing the bulge , i.e. stellar velocity @xmath0 , with the halo , @xmath3 , in attempts to use @xmath3 to derive blackhole masses based on a @xmath9 relation .
the slope arrived at by baes et al .
( 2003 ) is in agreement with the predictions of cosmological simulations ( di matteo et al .
2003 ) , but it is an artefact of combining the @xmath4 relation with the less accurate @xmath5 relation .
zasov et al .
( 2005 ) found significant scatter in the @xmath9 plots and concluded that the relation between @xmath9 is very weak .
their sample of 41 galaxies suffered contamination from galaxies , 20 in number , whose blackhole masses were determined indirectly from the @xmath4 relation .
sabra et al .
( 2008 ) reported a slope of @xmath26 based on a small sample of only 16 galaxies .
bandara et al . ( 2009 ) reported a correlation between @xmath1 and @xmath27 , the total gravitational mass of the host galaxy .
the total mass was determined from gravitational lensing observations , and their blackhole masses were determined indirectly through the @xmath4 relation ( gultenkin al .
however , when they calculate @xmath1 through the @xmath28 relation ( graham et al . 2001 ; 2007 ) , where @xmath29 is the sersic index , they find no signficant correlation between @xmath1 and @xmath27 .
we derive here the expected @xmath9 relation from the bandara et al .
( 2009 ) @xmath30 relation : @xmath31we use eq .
( 5 ) in ferrarse ( 2002 ) , in which the author assumes that @xmath3 is equal to the virial velocity and uses relations from bullock et al .
( 2001 ) between the mass of the dark matter halo and the circular velocity , we further assume that the total gravitational mass is equal to the mass of the dark matter halo to get : @xmath32combining eqs . ( 3 ) and ( 4 ) above we obtain the following : @xmath33 eq . ( 5 ) is semi - observational in the sense that it depends on an observational relation eq .
( 3 ) that connects the mass of the blakchole to the total gravitational mass of the galaxy , and on the theoretical relation eq . ( 4 ) that connects the mass of the dark matter halo to the circular velocity of the galaxy .
we also derive a theoretical counter - part to eq .
( 5 ) by using @xmath34reported in di matteo et al .
( 2003 ) from cosmological simulations , combined with the theoretical prescription of ferrarese ( 2002 ) derived from bullock et al .
( 2001 ) , eq .
( 4 ) above , to obtain : @xmath35 we overplot eqs . (
5 ) and ( 7 ) in fig . 1 .
most of the 46 galaxies fall between these two relations .
the 6 galaxies , the three left - most of which we do not include in the fits , that are to the left are all early - types .
the slopes in eqs .
( 5 ) and ( 7 ) are similar , but the intercepts are different
. this difference could be due to the assumptions made in deriving the two equations . in deriving equation ( 5 ) we assumed that mass of the dark matter halo is equal to the total gravitational mass of the galaxy .
this assumption affects the intercept of the resulting relation .
for example , if the mass of the dark matter halo is 90% of the total gravitiatonal mass , a reasonable assumption to make , then the intercept in eq .
( 5 ) would increase by 0.07 , not enough to make the intercepts comparable . on the other hand
, the reason of the difference could be even more fundamental and has to do with the nature of the simulations from which eq .
( 6 ) is derived ( di matteo et al .
2003 ) we plot in fig .
2 @xmath19 versus @xmath37 for the 89 galaxies from table 1 .
the data can be fit by a straight line : @xmath38we estimate the intrinsic dispersion to be @xmath39 dex .
the slope and normalization derived here agree , within the errors , with those of gultenkin et al .
( 2009 ) , hu ( 2009 ) , graham ( 2008 ) , tremaine et al .
( 2002 ) , and merritt & ferrarese ( 2001 ) , however , it is shallower than that reported by beifiori et al .
( 2012 ) and mcconnell & ma ( 2013 ) .
the relation is also in agreement with theoretical models ( di matteo et al .
2003 ; di matteo et al .
2007 ) .
the amount of intrinsic scatter in our @xmath4 is less than that for the @xmath9 above .
this is in contrast to the values reported in volonteri et al .
( 2011 ) in which the authors argue that the comparable intrinsic scatter in the @xmath4 and the @xmath9 relations indicates that there is a trend of some sort between @xmath1 and @xmath3 .
our sample , consisting of 89 galaxies , is larger than the one used in gultekin et al .
( 2009a ) , 49 galaxies , and volonteri et al .
( 2011 ) , 25 galaxies used by kormendy et al .
in contrast , to fig .
1 , galaxies in fig . 2 of all morphological types are spread out over roughly the same region . the large scatter in the @xmath9 plot ( fig .
1 ) is contrary to what was found in some earlier studies ( ferrarese 2002 ; baes et al .
baes et al .
( 2003 ) have calculated the blackhole masses indirectly by relying on the @xmath4 relation and then plotted @xmath1 against @xmath3 .
they used the @xmath4 relation of tremaine et al .
( 2002 ) with their @xmath5 relation to derive a @xmath9 relation .
it is important to keep in mind that baes et al .
( 2003 ) did _ not _ fit @xmath1 and @xmath3 .
our result here corroborates recent work by beifiori et al .
( 2012 ) and kormendy et al .
( 2011 ) , but using a larger sample of dynamical blackhole masses .
to circumvent biases that could affect analysis relying on small samples , we use the very large sample of ho ( 2007 ) to show that there will be a significant scatter in the @xmath9 plots , even when using a @xmath4 relation , which by itself could introduce a relation which is not necessarily there ( cf . our discussion above of ferrarese ( 2002 ) and baes et al .
ho ( 2007 ) presents @xmath0 and @xmath3 , together with other information , on 792 galaxies to study the @xmath5 relation .
given the @xmath0 s in ho ( 2007 ) , we use our @xmath4 relation , eq .
( 8) , to calculate the corresponding blackhole masses for all the 792 galaxies . the errors on @xmath0 and the uncertainties on the slope and normalization of the @xmath4 relation are propagated in quadrature together with the intrinsic scatter in @xmath1 at constant @xmath0 to calculate the uncertainties on the derived @xmath1 s .
we then plot the calculated masses against the corresponding @xmath3 for the 792 objects ( fig .
3 ) . the green circles are `` kinematically normal '' galaxies ( ho 2007 ) with @xmath40 .
we also overplot eq .
( 5 ) , which is based on the observed correlation between @xmath1 and @xmath41 ( bandara et al .
2009 ) , and eq .
( 7 ) , which is based on the theoretical correlation between @xmath1 and @xmath41 ( di matteo et al .
2003 ) . as a visual aid
, the horizontal line marks the blackhole mass obtained from eq .
( 8) for @xmath42 km s@xmath22 .
there is a significant amount of scatter with many galaxies occupying a wide swath with @xmath43 km s@xmath22 and @xmath44 m@xmath45 ; roughly the same region bounded by eqs .
( 5 ) and ( 7 ) .
most of the galaxies in this region are `` kinematically normal '' ( ho 2007 ) .
the majority of galaxies in this region are spirals ( sa to sc , fig , 3d - e ) . moreover ,
for any given morphological type , the spirals have the highest percentage of occupying this region . however , the region has quite a large scatter , roughly an order of magnitude , in @xmath1 for a given @xmath3 .
this makes using it as a kind of a relation between the mass of central blackhole and the circular velocity of the host galaxy problematic , or at least not as beneficial as the @xmath4 relation , which is used as probe of the co - evolution , in terms of redshift and morphology , of galaxy spheroids and the supermassive blackholes .
apart from showing that masses of the supermassive blackholes are poorly correlated with the circular velocities of their host galaxies , fig .
3 uncovers a few interesting points .
there are virtually no galaxies that have high circular velocities but low blackhole masses ( empty lower right corner ) , whereas low circular velocity apparently places no constraints on the mass of the blackhole ( the vertical scatter with @xmath46 km s@xmath22 ) .
the first part is expected : a high circular velocity is related to a more massive halo , and a more massive halo leads to a deeper potential well , and hence more infalling material that eventually accretes onto the central supermassive blackhole .
we address the second part through studying the @xmath9 plot , fig .
3a , by morphological type ( fig .
3b to 3e ) .
many elliptical and lenticular galaxies have high blackhole masses , @xmath47 m@xmath45 , but with circular velocities as low as @xmath48 km s@xmath22 and as high as @xmath49 km s@xmath22 .
they form the horizontal scatter in figs .
these galaxies have @xmath50 km s@xmath22 .
the horizontal scatter shows that the blackhole mass and the circular velocity of the host galaxies are not correlated in any way .
however , one must be careful in the interpretation .
the circular velocities used here are based on integrated velocity profiles , which are less than velocities obtained from optical rotation curves for early type galaxies ( see ho 2007 ) .
the discrepancy is almost a factor of 2 .
this can explain the spread of early type galaxies ( morphological index @xmath51 ) to the left in figs .
3b & 3c , and also in fig .
3d ( sa & sb spirals ) .
one the other hand , the points in the lower part of fig .
3f are very late type , bulgeless galaxies whose stellar velocity dispersion is @xmath52 km s@xmath22 and most of it is contributed by the rotation of the disk .
it seems that the only galaxies that follow some sort of a correlation are the kinematically normal galaxies .
however , the reasoning here seems circular .
this could be another manifestation of the `` cosmic conspiracy '' discussed in kormendy et al .
the kinematically normal galaxies by definition have @xmath53 with an average @xmath54 .
this fact , taken with the @xmath4 relation ( eq .
8) , means that @xmath55 . if we fit the 616 kinematically normal galaxies in fig .
3a , we get : @xmath56with no intrinsic dispersion , since it was already included in the calculation of the error on the derived @xmath1 . if this intrinsic scatter is not included , then the intrinsic dispersion in eq .
( 9 ) is @xmath57 dex .
the slope is consistent with that expected from a @xmath9 relation via a @xmath4 : the expected slope is @xmath58 .
the intrinsic scatter is significantly less than when fitting using dynamical @xmath1 as in eq .
the intrinsic scatter would disappear if we include the intrinsic dispersion of our @xmath4 relation in the errors of the calculated black hole masses .
this is expected intrinsic scatter in this case is build in , and it behaves like another source of error since it dominates the measurement errors .
this paper highlights the importance of the need to use direct values of the quantities being used to study the relation .
we attempted here to use dynamical measurements of blackhole masses since these measurements are `` clean '' in the sense that they do not depend on the properties of the host galaxy .
another issue that deserves similar attention is the getting similarily `` clean '' measurements of the circular velocities .
circular velocities depend on the where they are being measured in the galaxy , and the conversion between the circular velocity of the galaxy and the circular velocity of the halo is model dependent . in this study to took the pragmatic approach used by previous authors ( e.g. , * sun et al .
2013 * ; beifiori et al . 2012 ; kormendy et al .
2011 ) and implicitly assumed that all circular velocities are those of the dark matter halo . using this assumption , we found that the correlation between the masses of the supermassive blackholes and the circular velocities is very weak . by _ extrapolation _ , we conclude that the correlation between the blackholes and the dark matter halo is also weak in the present - day universe .
volonteri et al .
( 2011 ) argued the these two components of galaxies could have been more correlated in the past .
blackholes and dark matter halos are bound to be correlated somehow .
the deep potential wells of massive dark matter halos are needed for the material to accumulated in the center and form a blackhole .
the existence of bulgeless spiral galaxies with supermassive blackholes presents a challenge to the no @xmath7 correlation . on the other hand
, a @xmath9 correlation presisting to the present would imply the existence of very massive blackholes in the cores of dark matter halos of galaxy clusters ( kormendy et al .
2011 ) .
atkinson , j. w. , et al . 2005 , , 359 , 504 baes , m. , buyle , p. , hau , g. k. t. , & dejonghe , h. 2003 , , 341 , l44 bandara , k. , crampton , d. , & simard , l. 2009 , , 704 , 1135 barth , a. , et al .
2001 , , 555 , 685 bedregal , a. g. , salamanca - aragon , a. , & merrifield , m. r. 2006 , , 373 , 1125 beifiori , a. , et al .
2009 , , 692 , 856 beifiori , a. , et al .
2012 , , 419 , 2497 bender , r. , et al .
2005 , , 631 , 280 bernardi , f. et al .
2002 , , 123 , 2990 bower , g. a. , et al .
2001 , , 550 , 75 bullock , j. s. , et al .
2001 , , 321 , 559 capetti , a , et al .
2005 , , 431 , 465 cappellari , m. , et al .
2002 , , 578 , 787 cappellari , m. , et al .
2007 , arxiv:0709.2861v1 copin , y. , cretton , n. , & emsellem , e. 2004 , , 415 , 889 courteau , s. , mcdonald , m. , widrow , l. m. , & holtzman , j. 2007 , , 655 , l21 cretton , n. , & van den bosch , f. c. 1999 , , 414 , 704 dalla bonta , e. , et al .
2009 , , 690 , 537 davis , t. a. , et al .
2013 , , 429 , 534 davies , r. l. , et al .
2006 , , 646 , 754 de francesco , g. , capetti , a. , & marconi , a. 2006 , , 460 , 439 de francesco , g. , & capetti , a. 2008 , , 479 , 355 devereux , n. , et al .
2003 , , 125 , 1226 di matteo , t. , croft , r. a. c. , springel , v. , & hernquist , l. 2003 , , 593 , 69 di matteo , t. , et al .
2008 , , 676 , 33 ferrarese , l. , ford , h. c. , & jaffe , w. 1998 , iaus , 184 , 445 ferrarese , l. , & ford , h. c. 1999 , , 515 , 583 ferrarese , l. , & merritt , d. 2000 , , 539 , l9 ferrarese , l. 2002 , , 578 , 90 ferrarese , l. , & ford , h. c. 2005 , ssrv , 116 , 523 garcia - rissman , a. , et al . 2005 , , 359 , 765 gebhardt , k. , et al .
2000 , , 539 , l13 gebhardt , k. , et al .
2001 , , 122 , 2469 gebhardt , k. , et al .
2003 , , 583 , 92 ghez , a. m. , et al .
2008 , , 689 , 1044 gould , a. 2013 , arxiv1303.0834 g graham , a. , et al .
2001 , , 563 , 11 graham , a. , & driver , s. 2007 , , 655 , 77 graham , a. 2008 , pasa , 25 , 167 greene , j. e. , & ho , l. c. 2005 , , 630 , 122 greenhill , l. j. 2003 , , 590 , l62 greenhill , l. j. , moran , j. m. , & herrnstein , j. r. 1997 , , 481 , 23 gultekin , et al . 2009a , , 698 , 198 gultekin , et al .
2009b , , 695 , 1577 haehnelt , m. , & kauffmann , g. 2000 , , 318 , l35 harms , r. j. , et al .
1994 , , 435 , 35 hernstein , j. r. , et al .
1999 , nature , 400 , 539 hicks , e. k. s. , & malkan , m. s. 2008 , , 174 , 31 ho , l. c. 2007 , , 668 , 94 ho , l. c. , darling , j. , & greene , j. e. 2008 , , 681 , 128 hopkins , p. f. , et al .
2007 , , 669 , 5 houghton , r. c. w. , et al . 2006 , , 367 , 2 hu , j. 2008 , , 386 , 2242 hu , j. 2009 , arxiv:0908.2028v1 kaspi , s. et al .
2000 , , 533 , 631 kauffmann , g. & haehnelt , m. 2000 , , 311 , 567 kondratko , p. t. , greenhill , l. j. , & moran , j. m. 2008 , , 678 , 87 kormendy , j. 1988 , , 335 , 40 kormendy , j. , et al .
1996 , aas , 28 , 1423 kormendy , j. , & bender , r. 2011 , nature , 469 , 377 krajnovic , d. , et al .
2009 , , 399 , 1839 kronawitter , a. , saglia , r. p. , gerhard , o. , & bender , r. 2000 , , 144 , 53 kuo , c. y. , et al .
2011 , , 727 , 20 lodato , g. , & bertin , g. et al .
2003 , , 408 , 1015 maciejewski , w. , & binney , j. 2001 , , 323 , 831 macchetto , f. , et al .
1997 , , 489 , 579 mcconnell , n. j. , et al .
2011 , , 756 , 179 mcconnell , n. j. , & ma , c. 2013 , , 764 , 184 merritt , d. , & ferrarese , l. 2001 , , 547 , 140 nelson , c. h. , et al .
2004 , , 615 , 652 neistein , e. , et al .
1999 , , 117 , 2666 neumayer , n. , et al .
2007 , , 671 .
1329 nowak , n. 2007 , , 379 , 909 nowak , n. 2008 , , 391 , 1629 onken , c. a. , et al .
2004 , , 615 , 645 onken , c. a. , et al .
2007 , , 670 , 1050 palunas , p. , & williams , t. b. 2000 , , 120 , 2884 pastorini , g. , et al .
2007 , , 469 , 405 paturel , g. , et al . 2003 , , 412 , 57 peterson , b. m. , et al .
2004 , , 613 , 682 pizzella , a. , et al .
2005 , , 631 , 785 press , w. h. , et al .
1992 , numerical recipes in fortran : the art of scientific computing 2nd ed . ; cambridge : cambridge university press prugniel , p. , maubon , g. , & simien , f. 2001 , , 366 , 68 richstone , d. , et al .
1998 , nature , 395 , 1 richstone , d. , et al .
2004 , arxiv : astro - ph/0403257v1 rusli , s. p. , et al .
2013 , , 410 , 1223 sabra , b. m. , et al .
2008 , iaus , 245 , 257 sarzi , m. , et al .
2001 , , 550 , 65 shapiro , k. l. , et al .
2006 , , 370 , 559 tadhunter , c. , et al .
2003 , , 342 , 861 tilak , a. , et al .
2008 , , 678 , 701 tremaine , s. , et al .
2002 , , 574 , 740 wold , m. , et al .
2006 , , 460 , 449 valluri , m. , merritt , d. , & emsellem , e. , 2004 , , 602 , 66 van den bosch , r. c. e. , et al .
2012 , nature , 491 , 729 van der marel , r. p. , & van den bosch , f. c. 1998 , , 116 , 2220 verheijen , m. a. w. , & sancisi , r. , 2001 , , 370 , 765 verolme , e. k. , et al . 2002 , , 335 , 517 volonteri , m. , natarjan , p. , & gultekin , k. 2011 , , 737 , 50 yamauchi , a. , et al . 2004 , pasj , 56 , 605 zasov , a. v. , petrochenko , l. n. , & cherepshchuk , a. m. 2005 , arep , 49 , 362 cccccccc galaxy & morph & d ( mpc ) & a & @xmath59 & @xmath60 & @xmath61 & rf(@xmath1 ) + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) & ( 8) + circinus & sb & 2.8 & y & @xmath62 & @xmath63 & @xmath64 & 1-m + ic 1459 & e3 & 29.2 & y & @xmath65 & @xmath66 & @xmath67 & 2-s + ic 2560 & sbb , p & 41.4 & y & @xmath68 & @xmath69 & @xmath70 & 3-m + mw & sbbc & 0.008 & y & @xmath71 & @xmath72 & @xmath73 & 4-p + ngc 0224 & sb & 0.76 & y & @xmath74 & @xmath75 & @xmath76 & 5-s + ngc 1023 & sb0 & 11.4 & n & @xmath77 & @xmath78 & @xmath79 & 6-s + ngc 1068 & sb & 15.3 & y & @xmath80 & @xmath81 & @xmath82 & 7-m + ngc 1194 & s0 & 55.5 & y & @xmath83 & @xmath84 & @xmath85 & 47-m + ngc 1300 & sbbc & 20.7 & n & @xmath86 & @xmath87 & @xmath88 & 8-g + ngc 1332 & s0 & 22.7 & n & @xmath89 & @xmath90 & @xmath91 & 49-s + ngc 1399 & e & 21.1 & n & @xmath92 & @xmath93 & @xmath94 & 9-s + ngc 2273 & sba & 26.8 & y & @xmath95 & @xmath96 & @xmath97 & 47-m + ngc 2748 & sc & 24.9 & n & @xmath98 & @xmath99 & @xmath100 & 8-g + ngc 2787 & sb0/a & 7.5 & y & @xmath101 & @xmath102 & @xmath103 & 10-g + ngc 2960 & sa & 75.3 & y & @xmath104 & @xmath105 & @xmath106 & 47-m + ngc 2974 & e4 & 21.5 & y & @xmath107 & @xmath108 & @xmath109 & 11-s + ngc 3031 & sb & 3.9 & y & @xmath110 & @xmath111 & @xmath112 & 12-g + ngc 3079 & sb(s)c & 19.1 & y & @xmath113 & @xmath114 & @xmath115 & 13-m + ngc 3115 & s0 & 9.7 & n & @xmath116 & @xmath117 & @xmath118 & 14-s + ngc 3227 & sba & 17.5 & y & @xmath119 & @xmath120 & @xmath121 & 15-g , s + ngc 3245 & s0 & 20.9 & n & @xmath122 & @xmath123 & @xmath124 & 40-g + ngc 3379 & e1 & 10.3 &
n & @xmath125 & @xmath126 & @xmath127 & 16-g , s + ngc 3384 & s0 & 11.6 & n & @xmath128 & @xmath129 & @xmath130 & 17-s + ngc 3393 & sba , p & 51.8 & y & @xmath131 & @xmath132 & @xmath133 & 18-m + ngc 3414 & sb0 & 25.2 & n & @xmath134 & @xmath135 & @xmath136 & 11-s + ngc 3585 & s0 & 21.2 & n & @xmath137 & @xmath138 & @xmath139 & 30-s + ngc 3608 & e2 & 22.9 & n & @xmath140 & @xmath141 & @xmath142 & 17-s + ngc 3998 & s0 & 18.3 & y & @xmath143 & @xmath144 & @xmath145 & 19-g + ngc 4026 & s0 & 15.6 & n & @xmath146 & @xmath147 & @xmath148 & 30-s + ngc 4151 & sab & 13.9 & y & @xmath149 & @xmath150 & @xmath151 & 20-g , s + ngc 4258 & sabbc & 7.2 & y & @xmath152 & @xmath153 & @xmath154 & 21-m + ngc 4303 & sbbc & 16.1 & y & @xmath155 & @xmath156 & @xmath157 & 22-g + ngc 4342 & s0 & 17 & n & @xmath158 & @xmath159 & @xmath160 & 42-s + ngc 4374 & e & 18.4 & y & @xmath161 & @xmath162 & @xmath163 & 23-g + ngc 4388 & sb & 19.8 & y & @xmath164 & @xmath165 & @xmath166 & 47-m + ngc 4486 & e & 17.2 & y & @xmath167 & @xmath168 & @xmath169 & 24-g + ngc 4486b & e & 15.3 & n & @xmath170 & @xmath171 & @xmath172 & 25-s + ngc 4526 & s0 & 16.5 & n & @xmath173 & @xmath174 & @xmath175 & 50-g + ngc 4564 & e3 & 15.9 & n & @xmath176 & @xmath177 & @xmath178 & 17-s + ngc 4594 & sa & 9.8 & y & @xmath179 & @xmath180 & @xmath181 & 26-s + ngc 4596 & sb0 & 16.7 & n & @xmath182 & @xmath183 & @xmath184 & 10-g + ngc 4649 & e1 & 17.3 & n & @xmath185 & @xmath186 & @xmath187 & 17-s + ngc 4736 & sab & 4.9 & y & @xmath188 & @xmath189 & @xmath190 & 52-s + ngc 4826 & sab & 6.4 & y & @xmath191 & @xmath192 & @xmath193 & 52-s + ngc 4945 & sb & 3.8 & y & @xmath194 & @xmath195 & @xmath196 & 27-m + ngc 5128 & s0 & 3.5 & y & @xmath197 & @xmath198 & @xmath199 & 28-g + ngc 5252 & s0 & 96.8 & y & @xmath200 & @xmath201 & @xmath202 & 29-g + ngc 5576 & e3 & 27.1 & n & @xmath203 & @xmath204 & @xmath205 & 30-s + ngc 5846 & e0 & 24.9 & n & @xmath206 & @xmath207 & @xmath208 & 11-s + cccccccc galaxy & morph & d ( mpc ) & a & @xmath59 & @xmath60 & @xmath61 & rf(@xmath1 ) + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) & ( 6 ) & ( 7 ) & ( 8) + ngc 7052 & e4 & 71.4 & n & @xmath209 & @xmath210 & @xmath211 & 31-g + ngc 7457 & s0 & 13.2 & n & @xmath212 & @xmath213 & @xmath214 & 17-s + ngc 7582 & sbab & 22 & y & @xmath215 & @xmath216 & @xmath217 & 32-g + ugc 3789 & sa & 48.4 & n & @xmath218 & @xmath219 & @xmath220 & 47-m + & & & & galaxies with no @xmath3 & + a1836 & e & 157.5 & n & @xmath221 & & @xmath222 & 34-g + a3565 & e & 54.4 & n & @xmath223 & & @xmath224 & 34-g + cygnus a & e & 240 & y & @xmath225 & & @xmath226 & 33-g + ic 4296 & cd & 50.8 & y & @xmath227 & & @xmath228 & 34-s , g + ngc 0221 & e2 & 0.81 & n & @xmath229 & & @xmath230 & 35-s + ngc 0524 & s0,c & 23.3 & n &
@xmath231 & & @xmath232 & 36-s + ngc 0821 & e4 & 24.1 & n & @xmath233 & & @xmath234 & 37-s + ngc 1277 & s0 & 73 & n & @xmath235 & & @xmath236 & 48-s + ngc 1316 & sab & 20 & n & @xmath237 & & @xmath238 & 38-s + ngc 1407 & e & 29 & n & @xmath239 & & @xmath240 & 49-s + ngc 1550 & e & 53 & n & @xmath241 & & @xmath242 & 49-s + ngc 2549 & s0 & 12.3 & n & @xmath243 & & @xmath244 & 36-s + ngc 2778 & e2 & 22.9 & n & @xmath245 & & @xmath246 & 17-s + ngc 3091 & e & 52.7 & n & @xmath247 & & @xmath248 & 49-s + ngc 3377 & e5 & 11.2 & n & @xmath249 & & @xmath250 & 39-s + ngc 3607 & sa & 19.9 & y & @xmath251 & & @xmath252 & 30-s + ngc 4261 & e2 & 31.6 & y & @xmath253 & & @xmath254 & 41-g + ngc 4291 & e2 & 26.2 & n & @xmath255 & & @xmath256 & 17-s + ngc 4459 & s0 & 16.1 & n & @xmath257 & & @xmath258 & 10-g + ngc 4473 & e5 & 15.3 & n & @xmath259 & & @xmath260 & 17-s + ngc 4486a & e2 & 18.3 & n & @xmath261 & & @xmath262 & 43-s + ngc 4552 & e & 15.9 & y & @xmath263 & & @xmath264 & 11-s + ngc 4621 & e5 & 18.3 & n & @xmath265 & & @xmath266 & 11-s + ngc 4697 & e4 & 11.7 & n & @xmath267 & & @xmath268 & 17-s + ngc 4742 & e4 & 15.5 & n & @xmath269 & & @xmath270 & 44-s + ngc 4751 & e / s0 & 27.7 & n & @xmath271 & & @xmath272 & 49-s + ngc 5077 & e3,c & 40.2 & y & @xmath273 & & @xmath274 & 45-g + ngc 5328 & e & 65.9 & n & @xmath275 & & @xmath236 & 49-s + ngc 5516 & e & 60.1 & n & @xmath276 & & @xmath277 & 49-s + ngc 5813 & e1 & 32.2 & n & @xmath278 & & @xmath279 & 11-s + ngc 5845 & e3 & 25.9 & n & @xmath280 & & @xmath281 & 17-s + ngc 6086 & e & 139.1 & n & @xmath282 & & @xmath283 & 51-s + ngc 6251 & e2 & 107 & y & @xmath284 & & @xmath285 & 46-g + ngc 6264 & s & 145.4 & y & @xmath286 & & @xmath287 & 47-m + ngc 6323 & sab & 110.5 & n & @xmath288 & & @xmath289 & 47-m + ngc 6861 & e / s0 & 28.9 & n & @xmath290 & & @xmath291 & 49-s + notes .
( 2 ) : morphological classification ( nasa / ipac extragalactic database .
( 3 ) : distances ( see col . (
5 ) for references ) .
( 4 ) : active galaxy .
( 5 ) : blackhole masses from hu ( 2009 ) mcconnell & ma ( 2013 ) except for ngc 1300 , ngc 2778 ; ngc 4342 ; ngc 4374 ; and ngc 4945 ( graham 2008 ) , ngc 4303 ( pastorini 2007 ) , ngc 4486 kormendy et al .
( 1996 ) , and ngc 4594 ( kormendy et al .
( 6 ) : galaxy circular velocity ( see text for references ) .
( 7 ) : central velocity dispersion from hyperleda . col .
( 8) : original references for blackholes masses .
methods for derivation of blackhole masses in original references : g = gas dynamics ; s = star dynamics ; p = stellar proper motion ; m = maser .
original references for blackhole masses : ( 1 ) greenhill et al .
( 2003 ) ; ( 2 ) cappellari et al . ( 2002 ) ; ( 3 ) tilak et al .
( 2008 ) ; ( 4 ) ghez et al .
( 2008 ) ; + \(5 ) bender et al .
( 2005 ) ; ( 6 ) bower et al .
( 2001 ) ; ( 7 ) lodato et al .
( 2003 ) ; ( 8) atkinson et al .
( 2005 ) ; ( 9 ) houghton et al .
( 2006 ) ; ( 10 ) sarzi et al .
( 2001 ) ; ( 11 ) cappellari et al .
( 2007 ) ; ( 12 ) devereux et al .
( 2003 ) ; ( 13 ) yamauchi et al .
( 2004 ) ; ( 14 ) kormendy et al .
( 1996 ) ; ( 15 ) davis et al .
( 2006 ) and hicks et al .
( 2008 ) ; ( 16 ) gebhardt et al .
( 2001 ) and shapiro et al .
( 2006 ) ; ( 17 ) gebhardt et al .
( 2003 ) ; ( 18 ) kondratko et al .
( 2008 ) ; ( 19 ) de francesco et al .
( 2006 ) ; ( 20 ) onken et al .
( 2007 ) and hicks et al .
( 2008 ) ; ( 21 ) hernstein et al .
( 1999 ) ; ( 22 ) pastorini ( 2007 ) ; ( 23 ) maciejewski et al .
( 2001 ) ; ( 24 ) harms et al .
( 1994 ) and macchetto et al .
( 1997 ) ; ( 25 ) kormendy et al .
( 1996 ) ; ( 26 ) kormendy et al .
( 1988 ) ; ( 27 ) greenhill et al .
( 1997 ) ; ( 28 ) neumayer et al .
( 2007 ) ; ( 29 ) capetti et al ( 2005 ) ; ( 30 ) gultekin et al .
( 2009b ) ; ( 31 ) van der marel ( 1998 ) ; ( 32 ) wold et al .
( 2006 ) ; ( 33 ) tadhunter et al .
( 2003 ) ; ( 34 ) dalla bonta et al .
( 2009);(35 ) verolme et al .
( 2002 ) ; ( 36 ) krajnovic et al .
( 2009 ) ; ( 37 ) richstone et al .
( 2004 ) ; ( 38 ) nowak et al .
( 2008 ) ; ( 39 ) copin et al .
( 2004 ) ; ( 40 ) barth et al .
( 2001 ) ; ( 41 ) ferrarese et el .
( 1998 ) ; ( 43 ) nowak et al .
( 2007 ) ; ( 42 ) cretton et al .
( 1999 ) and valluri et al . ( 2004 ) ; ( 44 ) tremaine et al .
( 2002 ) ; ( 45 ) de fransesco et al ( 2008 ) ; ( 46 ) ferrarese et al .
( 1999 ) ; ( 47 ) kuo et al ( 2011 ) ; ( 48 ) van den bosch ( 2012 ) ; ( 49 ) rusli et al .
( 2011 ) ; ( 50 ) davies et al .
( 2013 ) and gould ( 2013 ) ; ( 51 ) mcconnell et al .
( 52 ) kormendy & bender ( 2011 ) .
|
we explore the connection between the central supermassive blackholes ( smbh ) in galaxies and the dark matter halo through the relation between the masses of the smbhs and the maximum circular velocities of the host galaxies , as well as the relationship between stellar velocity dispersion of the spheroidal component and the circular velocity .
our assumption here is that the circular velocity is a proxy for the mass of the dark matter halo .
we rely on a heterogeneous sample containing galaxies of all types .
the only requirement is that the galaxy has a direct measurement of the mass of its smbh and a direct measurement of its circular velocity and its velocity dispersion .
previous studies have analyzed the connection between the smbh and dark matter halo through the relationship between the circular velocity and the bulge velocity dispersion , with the assumption that the bulge velocity dispersion stands in for the mass of the smbh , via the well - established smbh mass - bulge velocity dispersion relation . using intermediate relations may be misleading when one is studying them to decipher the active ingredients of galaxy formation and evolution .
we believe that our approach will provide a more direct probe of the smbh and the dark matter halo connection .
we find that the correlation between the mass of supermassive blackholes and the circular velocities of the host galaxies is extremely weak , leading us to state the dark matter halo may not play a major role in regulating the blackhole growth in the present universe .
|
understanding the nature of the milky way halo
its shape , extent , density distribution , kinematics , abundance distribution and origin has long been a central topic in astronomy .
the importance of this endeavor has increased substantially with the growing , pervasive connections to a number of other astronomical enterprises bearing on such wide ranging astrophysical problems as , for example , the magnitude and distribution of dark matter and the frequency of microlensing events , the origin of the second parameter problem of horizontal branch morphology in globular clusters , the nature of high velocity hi clouds , the interaction of galaxies with their environment , and , of course , the origin and evolution of the milky way . in spite of the pressing need for a detailed picture of the halo , at present we still have only the most rudimentary prescriptions of , for example , the phase space distribution of halo stars .
unfortunately , for many applications we no longer can be satisfied with elementary analytical models of the halo . indeed , the very suitability of such simple descriptions may now be questioned .
a number of new lines of evidence indicate that the halo of the milky way has not achieved a dynamically relaxed state .
confirmation of this rather old idea is a long time in coming .
studies of candidate `` moving groups '' of metal - poor stars with halo kinematics in the solar neighborhood were made long ago by eggen and collaborators @xcite , and the idea of a halo with significant structure in the form of intermingling `` tube - like swarms '' was a viable theoretical description at least 35 years ago @xcite .
however , apart from consistent attention to the subject of halo `` moving groups '' by @xcite ( @xcite , and references therein ) , until recently the subject of halo substructure has received little other interest , in deference to more prosaic descriptions of the galactic halo those easily described by relatively simple analytical prescriptions .
this state of affairs was influenced perhaps , in part , by a growing emphasis on computer models of galactic structure that relied on analytical density laws ( * ? ? ?
* ; * ? ? ?
* _ et seq . _ ) , coupled with the lack of _ systematic _ observational efforts toward uncovering even first order global laws ( e.g. , density relations , amount of flattening , kinematical and chemical trends with position ) along more than a small number of lines of sight , let alone _ deviations _ from these simple global descriptions .
moreover , the rather smoother phase space distributions of the `` flattened halo '' and intermediate population ii ( which may or may not be parts of one continuous population ; see @xcite ) , locally dominate the more extended halo component what @xcite refers to as the `` pure races of the halo population ii '' .
indeed , that these flattened galactic populations contain stars with chemodynamical properties ( `` high velocity '' and low metallicity ) that are typically considered `` halo - like '' has long vexed understanding of the true `` mixture ratios '' and individual chemodynamical properties of the various overlapping stellar populations ( see , for example * ? ? ?
the more relaxed dynamical state of the locally more dominant , flattened metal - poor components of the galaxy may have long distracted attention from an unrelaxed halo population filled with _
substructure_. with the latter point as preface , it is worthwhile , therefore , to clarify our own working conception of the `` halo '' that galactic component we aim to explore in the present survey since the very definition of `` halo '' has taken on such a diversity of connotations .
presently fashionable `` dual halo '' models of the milky way those containing both flattened , prograde rotating and spherical , slow to non- ( or even retrograde ) rotating metal - poor components @xcite bear a resemblance to the commonly accepted galactic descriptions at the 1957 vatican conference @xcite , when one appreciates that the properties of the intermediate population ii ( ipii ) discussed there parallel properties assigned to the `` new '' flattened , contracted halo / thick disk components in today s models . in the present survey , we are concerned with oort s `` pure race '' of population ii : the extended , more or less spherical distribution of stars with the most extreme kinematics in the galaxy , where growing theoretical and observational work suggests evidence of past galactic accretion events may be fossilized .
the intermediate population ii , `` low '' or `` flattened '' halo which may all be the same thing @xcite we address more fully in our parallel astrometric survey ( e.g. , * ? ? ?
* ; * ? ? ?
* _ et seq . _ ) . in the last decade or so , a number of developments have spawned intense interest in the interaction of milky way - like galaxies with each other and with their satellites and clusters : e.g. , ( 1 ) the popularity of cold dark matter models of the universe , with the concomitant `` bottom up '' growth of structure from the collection of smaller subunits ( e.g. , * ? ? ?
* ) , ( 2 ) the realization of the prevalence of gravitationally interacting galactic systems in the nearby universe ( demonstrated of course by @xcite ; and explored more recently by , for example , @xcite ) and at high redshifts by @xcite , ( 3 ) the discovery of the formation of globular clusters @xcite and dwarf satellites @xcite in the course of such gravitational interactions , ( 4 ) the discovery of new galactic satellites @xcite , possibly associated galactic neighbors @xcite , and distant globular clusters ( e.g. , * ? ? ?
* ; * ? ? ?
* ) and the fact that some of these systems have complex star formation histories ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , and ( 5 ) the recognition of the complexities of the dynamics of the local group and the implications for the size of galactic dark matter components and the cosmological density , @xmath9 @xcite .
new computer studies of galaxy interactions ( @xcite , but all predated by @xcite ) confirm long - held notions ( @xcite ; see also the related results of @xcite ) that satellite galaxies , when experiencing tidal disruption by a larger galactic potential , can leave behind long - lived structures
oort s `` tube - like swarms '' along the satellite orbit .
the idea that these _ satellites themselves _ may not be dynamically relaxed systems @xcite has also played an important role in the debate on the existence of substantial dark matter halos in these systems ( cf .
discussion in * ? ? ?
* hereafter paper ii ) .
hints of substructure in the galactic halo that may be the `` swarms '' that are the hallmark of accretion of smaller stellar systems have been suggested in several surveys of halo stars undertaken since the concentrated effort of eggen to find halo ( and other ) `` moving groups '' among stars in the solar neighborhood .
this more recent evidence has most often been found in surveys of the more distant , `` pure '' halo , and most typically as clusterings of stars in distance and radial velocity
e.g. , the clumps of distant blue horizontal branch stars in @xcite , @xcite , and @xcite .
one very notable radial velocity - distance clump in the survey of distant stars by @xcite is now recognized as the structural paradigm of the type of tidal events sought here an extremely elongated , tidal structure forming through the destruction of the sagittarius dwarf galaxy .
indeed , the sgr galaxy was first identified through the clustering of radial velocities in distant _ giant stars _ , a technique we intend to exploit in the present survey . earlier evidence for possible tidal debris recognized as families of associated outer halo clusters and dwarf spheroidals was discussed by @xcite , @xcite and @xcite , and has received an increase in interest more recently @xcite .
one `` family '' of clusters ( including arp 2 , terzan 7 and 8 , m54 , and pal 12 ) is apparently associated with the paradigm tidal event , sagittarius @xcite .
phase space clumpings of more nearby , dwarf stars have also been identified by @xcite , @xcite , @xcite ( @xcite ) , and @xcite , with the clumps in the latter three works delineated in all three dimensions of motion very clearly being identified as halo members on the basis of both kinematics and abundance .
a surprising implication of @xcite is that the more distant halo appears to be _ dominated _ by phase space structure i.e. , very little `` random '' , dynamically relaxed halo field population exists outside of the ipii .
this empirical result would appear to echo theoretical suggestions that a large fraction of the halo might contain substructure @xcite .
detailed surveys of the halo including both precision proper motions and radial velocities , like the @xcite analysis , may be needed to understand the full spectrum of structure in halo phase space , and that particular survey continues to pursue that question .
however , to cover any significant amount of area in this detailed manner presents a formidable and improbable task at present .
moreover , to carry out this kind of work to distances as large as those of the present retinue of galactic satellites possibly major contributors to the overall halo field star population will need to await new generations of microarcsecond astrometry instruments , such as the space interferometry mission . in the meantime , questions regarding the existence of halo substructure , its extent , filling factor , size spectrum , etc . may be approached even when full kinematical data may not be obtainable .
the survey we describe here is aimed at divining and defining physical substructure in the halo when that substructure is coherent and has reasonable contrast above any well - mixed background in the occupied volume .
a primary aim of the present survey is to probe to large galactic distances with ease , and so provide a complement to the more detailed , but more confined , @xcite proper motion - radial velocity survey of the halo and ipii .
the greater distances probed in the giant survey described here lend certain advantages over a survey of more nearby stars , even one replete with full , 3-d kinematical information .
while perhaps decreasing sensitivity to subtle , more diffuse , structures requiring a full analysis of phase space , exploring at great distances from the galactic midplane unfetters the data and analysis from contamination by stars in the same survey volume having similar chemodynamical properties to , but coming from stellar populations ( e.g. , ipii , or the `` lower halo '' ) other than , the `` pure '' halo . from the standpoint of uncovering detailed information on the history of the galaxy from phase space substructure , surveys of distant halo stars
also have the added benefit that remote tidal structures are longer lived in the softer gradient of the galactic potential at large radii .
debris closer to the galactic center becomes phase - mixed rather quickly , and loses spatial coherence that would otherwise be easily identifiable ( although _ velocity _ coherence is _ increased _
distance also increases the contrast of strongly coherent structures against the background , because the linear width of a given structure subtends a smaller area on the sky , and the magnitude spread in the color - magnitude diagram from depth effects is decreased ( see @xcite ) , and because any smooth background of stars should have a rather steep density fall - off with galactocentric radius ( going as , say , @xmath10 ) .
recent work by @xcite and @xcite makes it evident that kinematic substructure in the halo would be difficult to see with simple starcounting techniques .
the densities of `` star swarms '' from dissociated galactic satellites are simply too low to be detectable above the foreground curtain of disk and ipii dwarfs a problem especially confounded when the swarms are at distances where the _ much smaller _ volume density of their luminous , evolved stars is the only signal likely to be observable . to make tidal streams in the outer halo more evident , it clearly would be beneficial to be able to reduce , or completely filter out , foreground disk contamination , to which main sequence stars make the greatest contribution redward of the field main sequence turnoff .
the signal - to - noise of a systematic search could be increased substantially if one were able _ a priori _ to key on stars of a set absolute luminosity because then disk and ipii stars with this luminosity would be easily differentiated from outer halo stars simply by their substantially brighter apparent magnitudes .
for example , halo rr lyrae stars and horizontal branch stars , which have a relatively limited range of absolute magnitudes , are rather easy to identify on the basis of variability or color and have been used as tracers of halo structure ( e.g. , * ? ? ? * ; * ? ? ?
on the other hand , horizontal branch stars are sufficiently rare that , while they can point to the _ presence _ of halo substructure , they may not be able to trace subtle or more _ tenuous _ halo substructure convincingly and/or thoroughly , as the rather small , candidate `` moving groups '' found by @xcite , @xcite , and @xcite would suggest .
unevolved , main sequence stars would , of course , make up the bulk of any stellar stream , but it is too difficult at present to explore dwarf stars to very great distances over large areas , though this technique has been used to study deep , pencil beam surveys with large telescopes in small numbers of strategically placed directions ( see , e.g. , * ? ? ?
* ; * ? ? ?
* ) . for exploring halo substructure
, giant stars may provide a reasonable compromise between the problems associated with the readily identifiable , but less numerous horizontal branch stars and the more common , but intrinsically faint dwarf stars .
giant stars are generally a few times more populous than horizontal branch stars in a given stellar population and provide the distinct advantage that they are bright enough to be imaged to great distances even with small telescopes . with giant stars large volumes of the outer galaxy may be explored efficiently with small telescopes ( where larger blocks of observing time are easier to come by ) if a reasonable means can be found by which to pick out the giants from the foreground dwarfs .
there have been several k giant studies at high galactic latitude @xcite from which we can infer expected densities of population ii giants .
@xcite found @xmath11 1 giant deg@xmath12 in a survey complete for the magnitude range @xmath13 covering @xmath14 deg@xmath15 at the sgp . in several high galactic latitude fields covering a total area of @xmath16 deg@xmath15
, @xcite found a mean density for population ii giants of order 2 deg@xmath12 in the magnitude range @xmath17 . with an expected mean radial fall - off of the halo with galactocentric radius as @xmath18
we obtain a flat differential count of giants with magnitude ( ) and we may extrapolate that to @xmath19 ( distances of @xmath20 kpc , for typical @xmath21 halo metallicity giants ) we should expect approximate mean densities of halo giants of @xmath22 deg@xmath12 .
thus , it is clear that a survey for giant stars must cover large areas of the sky in order to garner reasonably large statistical samples .
the requirement for large sky coverage drives the criteria to make our survey as efficient as possible at identifying giant stars .
it has long been known that the strength of the mgh + mgb triplet feature at 5100 is strongly dependent on surface gravity ( see figure 1 ; * ? ? ?
* ; * ? ? ?
* ) , and it is a common technique to identify giants by their weak absorption in this part of the spectrum @xcite .
@xcite ( @xcite ) showed that discrimination of k giants and dwarfs by this feature was possible with low resolution ( 20 ) objective prism plates from schmidt telescopes , while @xcite and @xcite had already devised a technique by which to identify giants _ photometrically _ with a pair of intermediate band filters , centered at 4880 ( @xmath23 `` continuum '' ) and 5150 ( @xmath0 `` mg '' ) .
this filter system has been applied to the giant surveys of @xcite .
later , @xcite ( @xcite , _ et seq . _ ) showed that good photometric luminosity classification was still possible with the @xmath0 filter ( see figure 1 ) when the continuum was measured with an appropriate broad band filter , in this case the @xmath6 filter of the washington system @xcite , at considerable savings in telescope time .
note that @xcite actually proposed the use of just such an intermediate band filter located around 5000 to allow luminosity classification .
the washington system itself was designed as a more efficient broadband system than @xmath24 for the study of the temperatures and abundances of g and k giants ( see * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
geisler s work has demonstrated the efficacy of separating g - k dwarfs and giants in the ( @xmath1 , @xmath4 ) color - color diagram ( where @xmath6 , @xmath3 , and @xmath25 are in the washington system ) for a sample of stars spanning a broad range of metallicity .
geisler pointed out additional features of his technique that are beneficial to a survey as described here : ( 1 ) the @xmath4 index is insensitive to surface gravity variations among g giants , ( 2 ) the reddening vector in the ( @xmath1 , @xmath4 ) diagram is such that more reddened , more distant giants will be even _ more _ separated from less - reddened foreground dwarfs ( see figure 4 ) , and ( 3 ) the metallicity sensitivity of the @xmath4 index is a second order effect , and in the sense that metal - poor giants have even _ smaller _ mg absorption .
the latter feature makes geisler s system even more useful , since there is additional discriminating power for our typically expected situation of selecting metal - poor population ii giants from among foreground metal - rich dwarfs .
* pb94 hereafter ) demonstrated the gravity discrimination of geisler s system using a grid of synthetic spectra over a range of surface temperatures , gravities and abundances appropriate to both population i and ii stars , and including realistic sequences of atmospheres for red giant branch isochrones .
their work provides a useful demonstration of the effects of gravity and abundance on the ( @xmath4 ) index : gravity dominates the color but there is a secondary sensitivity to abundance .
this is illustrated in figure 2 where the @xcite synthetic colors are translated to loci in washington / ddo51 color - color planes for dwarfs and giants of different [ fe / h ] .
we note here one possible shortcoming of the @xcite curves : as noted by @xcite , @xcite used filter passbands which differ slightly from the adopted standard washington filters , and furthermore , they have used an unpublished grid of model spectra , which are now somewhat out of date .
this probably gives rise to the blueward translation and small rotation in the color - color plane that we find necessary in 3.6 ( figure 12b and table 2 ) .
@xcite themselves point out that their isochrones are systematically redder than observed globular cluster giant branches .
unfortunately , @xcite did not include the @xmath0 filter when they produced their own ( more accurate ) synthetic colors .
this leaves @xcite as the only source for synthetic photometry with which we can compare our own photometry .
we emphasize that any problems with the @xcite colors appear to be quite small ( again , see 3.6 ) , in as much as they relate to our data .
two important and relevant effects fall out of this interplay of gravity and abundance as illustrated in figure 2 .
the first is that weakened absorption lines in metal - poor dwarfs that mimic the suppression of mg absorption from low gravity in giants becomes a problem only for subdwarfs with [ fe / h]@xmath26 .
main sequence stars more metal - rich than this , presumably all of those in the disk components and the majority in the halo as well , are not , in general , confused with giants , even giants as metal - rich as the sun .
the second relevant effect is that , for stars on the giant branch , the ( @xmath4 ) color provides a reasonably good abundance indicator [ @xmath27})\sim 0.13 $ ] at ( @xmath28 ) ] from solar [ fe / h ] down to [ fe / h]@xmath29
. both of these effects are critical to our enterprise here .
first , the metallicity sensitivity of ( @xmath4 ) we expect to be useful for an initial sorting of faint giant stars we encounter .
for example , the appearance of an apparent excess of giant stars of a single ( @xmath4)-based abundance would be an expected signal for a tidal tail from a mono - metallic parent object ( a commonly expected paradigm ) .
such an excess is clearly seen , for example , in our study of extratidal stars near the carina dwarf spheroidal galaxy in ( see , e.g. , figures 79 in that paper ) . in the case of very metal - poor subdwarfs
, we should not encounter an overwhelming number of contaminants in a sample of giants selected by ( @xmath4 ) techniques , based on the very small fraction of galactic stars with metallicities this poor . from the interim model of @xcite ,
the number of halo intermediate population ii / thick disk stars expected in the @xmath30 color range down to @xmath31 is about 200 deg@xmath12 at a galactic pole .
approximately half of these stars would be dwarfs and only about 8% of _ these _ would be expected to have metallicities [ fe / h ] @xmath32 , according to @xcite and @xcite .
this leaves an expected level of contamination of @xmath33 metal - poor subdwarfs deg@xmath12 , comparable to an expected density of giants of @xmath34 deg@xmath12 down to @xmath31 .
this is , of course , in the ideal situation of high galactic latitude . on the other hand , because of the rapid decline in the metallicity distribution function below [ fe / h]@xmath29 , it is possible , with only slightly more conservative giant selection in the color - color plane , to reduce the subdwarf contamination to practically nothing .
for example , if one were interested in selecting for [ fe / h]@xmath35 giants , one would only be concerned with about two [ fe / h]@xmath36 subdwarfs per square degree ( see figure 2 ; @xcite ) .
we believe this level of contamination in our initial photometric catalogues to be tolerable . in any case
, when such stars are encountered , they will be identifiable by followup spectra ( or , for the brighter stars , by their proper motion from such catalogues as the nltt ; ( * ? ? ?
* _ et seq . _ ) ) .
moreover , these metal - poor stars are interesting in their own right , and worthy of discovery for further exploration of the galaxy s evolution . for our survey , we have adopted a variant of the geisler technique that balances the goals of covering large areas as efficiently as possible with the desire to obtain as much information about identified giant stars as possible .
we have adopted a _ three filter _ photometric system that provides dwarf / giant separation capability , a surface temperature indicator , and a rough gauge of stellar abundance in giant stars . in @xcite ,
the @xmath1 color serves as a surface temperature index , while the @xmath4 is used as the luminosity index .
however , as we shall show ( 2.1 ) , the @xmath2 color is monotonically correlated to the @xmath1 color .
thus , @xmath2 can serve as a suitable temperature index ( as has also been demonstrated by @xcite ) , with the advantage that one less filter is needed in the observations with no loss in information a useful improvement in efficiency . in this paper
, we explore and calibrate the ( @xmath2 , @xmath4 ) diagram for discriminating dwarf and giant stars for a range of metallicities ( 2 ) .
we note that the washington @xmath37 filter is designed specifically for photometric abundance measurement in giant stars , and use of this filter would provide a much more accurate [ fe / h ] for our survey giants ( particularly metal - poor ones ) than relying on the secondary dependence of ( @xmath4 ) on abundance .
however , observations in the @xmath37 filter are rather expensive ( requiring 3 times the exposure time of the @xmath6 filter to reach an equivalent depth
since we already require a large investment in observing time for the @xmath0 filter , and our primary imaging goal is to identify giants with as great efficiency as possible , we have dispensed with using the @xmath37 filter in most of what we do .
our rationale is two - fold : ( 1 ) we believe that the coarse abundances afforded by the ( @xmath4 ) color are sufficient for tidal tail / halo substructure searches in our photometry , and ( 2 ) it is our intention to back up our photometrically identified giant candidates with spectroscopy as much as possible .
spectroscopy is needed not only as a check on our giant candidates , but also as a means to obtain dynamical information from their radial velocities .
much better abundances may be obtained from moderate resolution spectra ( sufficient for a radial velocity measure ) than from @xmath37 filter photometry .
typical spectra of the type we are using for this followup work are shown in figure 1 , where several spectroscopic indicators of surface gravity , including the mgb , mgh , and nad features , may be seen .
we will discuss our spectroscopic work for this survey further in future contributions .
the selection of fields for our survey reflects two distinct , but complementary , strategies .
the first is predicated on the paradigm afforded by the example of the sgr dwarf galaxy as well as by dynamical models of tidal effects on satellite galaxies , both which suggest that a non - negligible mass loss rate in the form of stripped stars may be discernible around presently known satellite galaxies .
thus , it makes sense to start a search for tidal tails in the galactic halo at the most obvious potential sites for their creation .
we have therefore begun a systematic search for giant stars beyond the tidal radii of the galactic dwarf satellite galaxies , as well as a sample of globular clusters .
we have already reported successful searches for extended , coherent stellar structures around our first two targets , the magellanic clouds @xcite and the carina dwarf spheroidal ( ) .
as the example of the sgr galaxy illustrates , the tidal debris of satellite galaxies may stretch to substantial lengths ( @xcite ) , and perhaps completely encircle the sky @xcite .
tracing substantially lengthy tails continuously outward from the parent could be an extremely time - consuming enterprise , and should be weighed against the potential for tracking the path of the debris with more disparate , strategically placed , pencil beam probes around the sky .
another problem that may be addressed with a series of probes is the existence of debris from objects that no longer exist with any recognizable , gravitationally intact core . in terms of the formation history of the milky way halo ,
it is obviously of great interest to assess the net contribution of disrupted bodies to the halo .
finally , in order to understand the distribution of any `` smooth background '' of halo stars the magnitude of which affects the contrast of any superimposed substructure @xcite it makes sense to perform a more systematic survey that can address the global distribution of giant stars . for all of these reasons , and
others , we have also embarked on the grid giant star survey ( ggss ) .
the ggss is a program to observe 1303 isotropically spaced ( mean field spacing @xmath38 ) galaxy pencil - beam probes , each of area 0.40.7 deg@xmath15 , to find giant and hb stars to explore galaxy structure .
the details and results of this systematic survey will be presented elsewhere .
the `` all - galaxy '' ggss has similar goals and strategy to the `` spaghetti survey '' described by @xcite .
this survey also adopts a ( @xmath39 ) photometric search phase to identify giants and hb stars for spectroscopic followup , and has presented initial findings and discussion of strategy in @xcite .
together , the spaghetti and ggss surveys should provide a wealth of new information on the outer galaxy . in 2 we determine transformations from our ( @xmath6 , @xmath25 ) system to both the ( @xmath6 , @xmath3 , @xmath25 ) and ( @xmath7 , @xmath40 ) systems . in
3 we calibrate the ( @xmath2 , @xmath4 ) color - color diagram for discriminating dwarfs and giants , and explore the metallicity trends for giant stars in this diagram .
we obtain good agreement in our empirical calibration of this two - color plane with that given by synthetic spectra .
finally , by way of @xmath5 centauri giants as templates , we determine a rough calibration of giant star absolute magnitudes as a function of photometric abundance ( i.e. , position in the two - color diagram ) , which can be used for photometric parallaxes .
the washington system was originally designed to obtain temperatures , metal abundances and cn indices ( which relate well to [ fe / h ] for population i stars , @xcite , but vary independently of abundance for [ fe / h]@xmath41 , @xcite ) for g and k giants _ photometrically _ @xcite .
the primary goal of the photometric part of our survey is simpler to _ identify _ potential giant stars .
thus , whenever possible , candidate giants will be subjected to spectroscopic follow - up to verify surface gravity ( weeding out k subdwarfs ) , obtain a radial velocity and determine a _ spectroscopic _ metallicity ) color . ] .
thus , we can hope to reduce the number of washington filter measurements to the minimum necessary to achieve our survey goals .
for example , at present we are not emphasizing measurement of carbon abundances , so the cn and g band - sensitive @xmath37 filter is not essential to our aims . on the other hand ,
the @xmath6 filter is needed to serve as a `` continuum '' complement to the @xmath0 filter , a la @xcite , and must be retained .
because the sensitivity of the @xmath4 index to surface gravity is a function of surface temperature , and because the derivation of spectroscopic abundances and gravities requires some measure of the effective temperature , we need an additional photometric measure sensitive to @xmath42 . in the standard washington system
, the @xmath1 parameter serves the role of a temperature index , but for our purposes greater efficiency could be obtained if a single additional filter could be combined with the already required @xmath6 filter to provide a suitable temperature index .
unfortunately , the @xmath6 filter is subjected to more line blanketing in metal - rich stellar atmospheres than the @xmath3 filter .
indeed , in the washington system @xcite , the combination of @xmath43 with @xmath1 is intended to provide a metallicity index , @xmath44 ( measured from the solar abundance locus similar to the definition of @xmath45 ) . however , according to figure 10 of @xcite , the range of @xmath44 , is at most about 0.15 magnitudes over @xmath46 [ fe / h ] @xmath47 for giants with a wide range of @xmath1 ; and for giants with [ fe / h ] @xmath48 typical of the expected metallicities of distant giants we might hope to find in our survey the range of @xmath44 is only about 0.05 magnitudes .
thus , the @xmath2 color may provide an adequate temperature index for our purposes .
indeed , @xcite conclude that `` @xmath2 is better than @xmath1 for temperature determinations , although it is slightly sensitive to surface gravity . '' in figure 3
we justify our exclusion of @xmath3 imaging in the present survey as a means to reduce our photometric survey to the simple @xmath49 filter system .
figure 3 shows all stars from tables v and vii in ( * ? ? ?
* _ open squares _ ) , table vi of ( * ? ?
* _ solid squares _ ) , table iii of ( * ? ?
* symbols as in that paper : _ open triangles _ for luminosity class i - ii , _ solid triangles _ for luminosity class iii - iv , and _ solid circles _ for luminosity class v - vi ) , table 2 of ( * ? ? ?
* _ open circles _ ) , and solar abundance field giants from table 1 of ( * ? ?
* _ solid triangles _ ) .
note , we exclude data for stars from @xcite when they were reobserved and updated in @xcite , but we include both of the separate measures for the few stars repeated in @xcite and @xcite , since these are separate data sets .
figure 3 , which includes dwarfs and giants of a wide range of metallicities , and which has not been corrected for reddening , already demonstrates the relative tightness of the monotonic and almost ( over most colors ) linear @xmath2 versus @xmath1 correlation . the upper dereddening vector in this diagram is from @xcite , while the lower one is based on the reddening ratios derived in 2.3 below .
the solid line indicates a 4th order fit through the points , ignoring the two reddest @xmath2 stars from @xcite , which are in highly reddened , galactic plane fields .
the root - mean - squared ( rms ) scatter around the fit to this assortment of stars , _ without _ accounting for differing metallicities and reddening , is only 0.046 ( or 0.036 when iteratively excluding greater than 3-@xmath50 points ) .
we conclude that while @xmath1 has been proven an excellent @xmath42 index for g and k stars @xcite , the @xmath2 color also provides an adequate color index suitably correlated to @xmath42 and with only slight sensitivity to abundances . that we can retain excellent dwarf - giant discrimination in the _ three filter _
@xmath51 system is illustrated by the synthetic photometry of @xcite in figure 2b . in the remainder of this paper
we provide an _ empirical _ formulation for this methodology .
it is useful to determine the transformation of our washington ( @xmath6 , @xmath25 ) filter system into the more commonly used cousins ( @xmath7 , @xmath40 ) system . in this way
we may compare our field star and cluster color - magnitude data to the extensive work done in the cousins system in the literature . to determine the transformation functions we take advantage of the tabulation of photometry by @xcite , who have measured @xmath7 magnitudes along with washington photometry in the establishment of their standard star system .
we also make the assumption that the @xmath25 filter is identical to the cousins @xmath52 filter .
this assumption is justified by the great similarity of the @xmath25 bandpass described by ( * ? ? ?
* see his figure 1 ) and that of the cousins @xmath52 band as illustrated by ( * ? ? ?
* see his figure 1 ) .
indeed , both bands are often produced by the same combination of schott rg-9 filter with dry - ice cooled ga - as photomultiplier @xcite , and similar effective filter central wavelengths , @xmath53 , are determined for giant star colors for the @xmath25 and @xmath52 filters by these two authors . however , as discussed by bessell , the standard stars of @xcite were established with a photomultiplier kept @xmath54c warmer , which gives rise to a 130 redward shift in the @xmath55 in the cousins standards .
we ignore here any possible affect this might have on any @xmath25 to @xmath52 transformation .
presumably , the @xmath25 magnitudes of @xcite would be identical to the `` natural gaas '' system @xmath52 band of ( * ? ? ?
* see his section iii ) , from which @xcite notes a slight non - unity ( a slope of 1.036 ) in conversion to the standard @xcite system in @xmath56 color , all attributed to the @xmath40 band shift with photomultiplier temperature .
thus , we might expect a systematic error in the slope of the ( @xmath2 ) to @xmath57 color conversion derived below of up to 3.6% .
also to be kept in mind are any possible extensions of the redward side of the @xmath52 or @xmath25 response function for any combination of , say , the rg-9 glass ( a `` cut - on '' filter ) and ccd detectors that may have significant response up to the substrate bandgap cutoff in the near - infrared ; this problem would affect the reddest stars .
for the 79 stars in @xcite that span @xmath58 , we obtain a good linear color transformation ( rms residual = 0.014 ) as follows : @xmath59 a significant contribution to the rms residual about this fit was contributed by the two reddest stars , each having @xmath60 , in the sample
. a cubic fit to the data yields no significant improvement . with the assumption @xmath61
, we find from the above relation that @xmath62 may be determined by @xmath63 because the objects we use to calibrate our ( @xmath2 , @xmath4 ) diagram in 3 are typically at low galactic latitudes , we must correct for extinction by dust .
we adopt the average interstellar extinction curve for @xmath64 of @xcite . for the @xmath6 ( @xmath65 , @xmath66 , @xcite ) and @xmath25 ( @xmath67 , @xmath68 ) filters we obtain @xmath69 and @xmath70 .
thus @xmath71 compared with @xmath72 from @xcite .
using older absorption curves , @xcite derived @xmath73 .
the effective wavelength of the @xmath0 filter yields @xmath74 , so that @xmath75 .
it follows that @xmath76 we also find @xmath77
in this section , we demonstrate how we discriminate between giants and dwarfs in the @xmath4 versus @xmath2 plane .
our demonstration utilizes both previous data in the literature , as well as new photometric data collected on the swope 1-m telescope at las campanas observatory .
we also investigate the secondary dependence of the ( @xmath4 ) filter on [ fe / h ] . in figure 4
we show the @xmath4 versus @xmath2 plane with data from @xcite and @xcite ( for solar metallicity field giants ) , where photometric data for stars with known luminosity class and reddening are provided .
figure 4 here is analogous to geisler s ( 1984 ) figure 3 . as can be seen , while there is little ability to discriminate between supergiants and giants , in general
there is excellent discrimination between dwarfs and evolved stars of all luminosity classes i - iii , as was suggested by the synthetic photometry presented in figure 2b . for @xmath78 ,
subgiants and dwarfs are also well discriminated , but , as might be expected , the separation begins to break down for bluer colors , near the main sequence turn off . with the new data presented here , we sample this important color regime with a large number of stars , with the aim of defining a useful guide for dwarf / giant demarcation in surveys of field stars .
data for all of the star clusters presented here , except for melotte 66 , were obtained with the site # 1 2048@xmath15 ccd attached to the swope 1-m on the nights of ut 15 - 17 march 1997 .
all but the very end of the last night of this run was photometric .
the melotte 66 data were obtained with the same filters , telescope and ccd on the photometric night of ut dec .
15 , 1997 .
we used the `` carnegie 3-inch washington filter set '' to which we added a @xmath0 filter .
we believe that the washington filters were purchased from omega optical and that the filters are based on the following prescription : * @xmath0 : * the @xmath0 filter was purchased from omega optical in 1993 , as filter 515bp12 , lot # 9349 .
this filter is similar to the one described by @xcite , but with a slightly larger peak transmission ( 88% versus geisler s 75% ) and a very slightly narrower passband ( half transmission at about 5095 and 5200 ) .
our cluster observations were calibrated with geisler s ( 1990 ) sa98 , sa110 , and ngc 3680 standard star fields , accounting for color , airmass , and , in the case of the @xmath6 band , ( airmass ) @xmath79 ( color ) terms , along the lines of the procedure followed in @xcite . for the @xmath6 and @xmath25 solutions ,
the ( @xmath2 ) color was used , whereas , for the @xmath0 solution , an ( @xmath4 ) color was used . for the first two nights of the march run , during which standard stars were measured , the solution was allowed to determine individual night zero - points , and these agreed to better than 0.013 in all cases . the extinction was assumed to be constant over all three march 1997 nights . in each of the @xmath6 , @xmath25 and @xmath0 solutions , there were 88 , 69 and 64 useful standard stars , yielding solutions with rms errors of 0.0036 , 0.0051 and 0.0073 magnitudes , respectively .
our photometric catalogue for the intermediate aged @xcite cluster ngc 3680 employed ten @xmath6 , eight @xmath25 and nine @xmath0 frames having integration times in the ranges of 5 - 8 seconds , 5 - 8 seconds and 50 - 90 seconds respectively .
the images included geisler s ( 1984 ) calibration sequence in the washington system .
these ccd frames were reduced using daophot ii @xcite , calibrated and checked against geisler s sequence , and combined into a single catalogue .
our ccd frames cover @xmath80 per side , at @xmath81 per pixel . in figure 5 we present the ( @xmath6 , @xmath2 ) color - magnitude diagram ( cmd ) for 1815 stars in our ccd field of ngc 3680 .
random errors in the photometry are presented in the right hand panels of figure 5 .
we have adopted @xmath82 ( @xcite , and consistent with ( * ? ? ?
* naa97 hereafter ) ) , and @xmath83 , or @xmath84 , @xmath85 , and @xmath86 .
@xcite determined the abundance of ngc 3680 to be [ fe / h]@xmath87 , but an analysis of the cmd based on members cleaned of binary systems by naa97 gives [ fe / h ] = + 0.11 ( similar to the value of + 0.09 @xmath88 of @xcite ) and an age of 1.45 @xmath89 0.3 gyr .
a significant amount of field star contamination is evident in the ngc 3680 cmd , especially around @xmath90 , which is near the main sequence turn - off of old disk field stars ( @xmath91 ) .
the latter overlap significantly the ngc 3680 main sequence stars of interest .
we have cross - referenced our photometric catalogue to both the proper motion catalogue of ( * ? ? ?
* k95 hereafter ) and to the radial velocity catalogue of naa97 .
this allows us to separate true ngc 3680 members from the numerous field star contaminants , and also eliminate binary stars . in figure 6a
we show the cmd for only those stars having measured proper motions by k95 and with determined joint proper motion - spatial membership probability @xmath92 . this low value was selected to mimic k95 s own selection ( see figure 8 in k95 ) and intended to preserve some fainter stars in the cluster cmd ( faint stars tend to have lower @xmath93 in the k95 survey ) .
however , as k95 point out , this liberal cut probably results in contamination of the lower main sequence by field stars ( e.g. , the proper motion errors in k95 begin to grow substantially for @xmath94 ) .
therefore , with larger symbols we denote a more conservative cut with the k95 estimator cutoff at @xmath95 . in figure
6b we show the resultant ( @xmath2)@xmath96-(@xmath4)@xmath96 diagram for the proper motion selected samples .
use of the much more refined membership analysis of naa97 , which employed precision radial velocities , clarifies the dwarf / giant discrimination by colors , albeit only for a bright subsample ( figures 6c and 6d ) .
we include the high quality membership data of figure 6d in figure 14 , below .
figure 6c looks qualitatively similar to naa97 s ( @xmath97 ) cmd for ngc 3680 .
it can be seen , from the general agreement of the points with the @xcite loci for [ fe / h]=0 stars , that the likely ngc 3680 giants , near @xmath98 , as well as the likely main sequence stars more or less fall in the appropriate places in the color - color diagram .
the rich open cluster ngc 2477 was observed in order to get a large sample of likely lower main sequence stars .
unfortunately , extending the cluster main sequence as faint as possible resulted in the saturation of our frames for a majority of the suspected cluster giants , which are lost from our analysis .
we obtained two @xmath6 , three @xmath25 and two @xmath0 ccd exposures , for integrations of 60 - 80 seconds , 60 seconds , and 600 seconds each , respectively . again
daophot ii was used to create a calibrated photometric catalogue . in figure 7
we present the ( @xmath6 , @xmath2)@xmath96 cmd for 11,300 stars in the field of ngc 2477 .
previous studies suggest the cluster shows substantial differential reddening , from @xmath99 to 0.4 @xcite ; following @xcite and @xcite , we have adopted a mean @xmath100 and @xmath101 .
this translates to @xmath102 , @xmath103 and @xmath104 .
the resultant cmd in figure 7 is qualitatively identical with the @xmath105 cmd of @xcite , apart from the lack of a concentration of giants in our cmd .
the broadening of the upper main sequence has been attributed to the differential reddening , and this is supported by the fact that the lower main sequence , which more nearly parallels the reddening vector , is thinner , particularly in the ( @xmath106 ) cmd @xcite .
cluster parameters for ngc 2477 have most recently been determined by isochrone fitting in @xcite .
this latter study finds best fits to [ fe / h ] = @xmath107 and an age of @xmath108 gyr .
earlier age estimates were slightly higher , e.g. , 1.5 @xmath89 0.2 gyr @xcite and 1.2 @xmath89 0.3 gyr @xcite , but @xcite , who used the same theoretical isochrones as @xcite but on the @xcite data , obtained 0.6 @xmath109 gyr .
the white dwarf cooling curve age of the cluster tends to support the higher ages , around 1 gyr @xcite .
the best fitting @xcite abundance for ngc 2477 is slightly lower , but not significantly so , from the [ fe / h]=0.04 value of @xcite . in spite of the amount of attention ngc 2477
has received lately by way of deep studies @xcite , no membership study , even at the bright end of the cmd , exists for this cluster . in an attempt to improve our chances of selecting true ngc 2477 main sequence stars so that we might better delineate the dwarf locus in the ( @xmath110 , @xmath111)@xmath96 diagram , we reduce the field sample by limiting the cmd to stars in two specific radii
@xmath112 ( figures 8a - b ) and @xmath113 ( figures 8c - d ) about the nominal cluster center ( which was judged by finding the peak of the x and y marginal distribution histogram of starcounts for @xmath114 ) .
by eye we selected stars along the apparently least - reddened ngc 2477 main sequence and show their position in the two - color diagram in figures 8b and 8d ( _ large dots _ ) .
the small dots show a large population of apparent metal - rich field giants along the predicted @xcite [ fe / h]=0 giant locus , while the predicted trend for the selected dwarf sequence is clearly borne out by the data .
the spreading of the locus at redder ( @xmath2)@xmath96 colors is a consequence both of the increased photometric error for the fainter stars in the sequence as well the contribution of differential reddening , which increases as the dwarf sequence curves ever more perpendicular to the reddening vector in the lower part of the diagram .
note the much larger spread in the unselected stars in figures 8b and 8d .
this spread about the locus is partly due to large photometric errors at the faint limit of the catalogue ( as high as 0.4 magnitudes in the colors ) .
however , the larger bulk of stars _ above _ the dwarf locus hints at a substantial population of field giants in this low latitude ( @xmath115 ) field .
melotte 66 has long been recognized as an archetypal old open cluster ( e.g. , * ? ? ?
recent studies estimate an age for the cluster of @xmath116 gyr @xcite and @xmath117 gyr @xcite .
@xcite adopt an abundance of [ fe / h]=@xmath118 , based on @xcite , which is similar to the adopted abundance ( [ fe / h]=@xmath119 ) of @xcite .
however , an intrinsic metallicity spread among the cluster giants has also been detected @xcite .
the reddening to the cluster does not seem well established , although @xcite settle on @xmath120 as the reddening that gives the most consistent ultraviolet excesses between the cluster dwarfs and giants .
we adopt this value , so that @xmath121 , @xmath122 and @xmath123 .
we obtained single @xmath6 , @xmath25 and @xmath0 exposures of melotte 66 with integration times of 12 , 12 and 100 seconds , respectively , for the dec 1997 run .
these were calibrated with 67 , 59 and 60 useful standard stars in the @xmath6 , @xmath25 and @xmath0 bands , respectively , yielding solutions with random errors of 0.0094 , 0.0074 and 0.0109 magnitudes , respectively . with these exposure times
, we were able to get reliable photometry for all but the very brightest part of the red giant branch .
we present in figure 9 the cmd in our filter system for the field of melotte 66 .
there is substantial field contamination , but the main sequence turn off region and the red clump of the cluster can be discerned easily . while it is in somewhat better shape than either ngc 3680 or ngc 2477 , melotte 66 is still in need of a better membership census .
some radial velocity memberships exist and in figures 10a and 10b we show those stars listed as members by @xcite , @xcite , @xcite , @xcite and from the @xcite analysis listed in @xcite .
the result shows the prominent double cmd sequence caused by the substantial binary fraction in the cluster .
this duplication causes both the double subgiant branches as well as a fattening of the msto region ( see figures 9 and 10c ) .
we show the @xmath124 diagram corresponding to the identified members in figure 10b . here
we delineate two groups of radial velocity member giants / subgiants : those falling more or less in the expected region of the color - color diagram ( _ open circles _ ) and those in more peculiar locations , more like expectations for dwarf stars ( _ closed circles _ ) .
it can be seen that the peculiarly - placed stars in figure 10b seem to be predominantly associated with the upper subgiant branch , that of the binary stars , although a few stars , near @xmath125 , are found below this subgiant branch .
it may be that the wide scatter in @xmath126 color near @xmath127 in figure 10b ( and figure 10d below ) , which gives rise to `` problem giant stars '' in the dwarf region of the two - color diagram , may be due to problems with the binarity .
it is interesting also to note that many of these `` problem stars '' have abnormally large daophot @xmath128 parameters compared to other stars at similar magnitudes .
perhaps this is some indication of a slight resolution of close binaries that is detectable by daophot . in figures 10c and 10d ,
we attempt to elucidate the true nature of the red giant branch for the cluster by limiting our field to a radius of @xmath129 . in figure 10c we trace likely cluster members using the known radial velocity members ( shown in figure 10c whether or not they fall within the @xmath129 radius ) and both the single and binary isochrones in figure 9d of @xcite as guides .
once again , we use _ open _ and _ closed _ circles to delineate stars with `` problematical '' colors from the standpoint of dwarf / giant separation .
we trace down the red giant branch to the red clump and down the subgiant branch to the point where both the binaries and the single stars join their respective main sequence turnoffs .
all other stars within @xmath129 of the cluster center are left as small symbols in figures 10c and 10d . in figure 10d
the giants above the red clump follow the @xcite locus for solar metallicity giants .
as expected ( from , for example , figure 4 ) , the subgiant locus in figure 10d merges with , and becomes less discriminated from , the locus of main sequence turnoff stars ( small points ) at bluer colors . the latter sweep lower than the main locus of giant stars in figure 10d , as expected .
finally , we note again a predilection for the `` giants '' with problematical locations in the two - color diagram to be associated with the subgiant branch for binary members .
because of its proximity and size , @xmath5 centauri is a convenient target that provides numerous bright giant stars for use as calibrators .
moreover , with an abundance ranging from solar to [ fe / h]@xmath130 @xcite , this one cluster allows the opportunity to explore the abundance sensitivity of the @xmath4 index , as well as to calibrate the giant star color - magnitude relation as a function of metallicity for old stellar populations , without introduction of relative systematic errors produced by differential distance errors . to explore the giant distribution in the ( @xmath2 , @xmath4 ) and the ( @xmath2 , @xmath6 ) planes , we have made an extensive imaging survey of the globular cluster @xmath5 centauri .
our @xmath5 cen data consists of eight separate pointings , offset by @xmath131 in each the north , south , east , west , northeast , northwest , southeast , and southwest directions .
our resulting catalogue contains over 100,000 stars photometered , in general , two or three times each , with daophot .
the data were matched into catalogues using daomaster .
the washington cmd for @xmath5 centauri is presented in figure 11 .
we have adopted @xmath132 as in @xcite ; this translates to @xmath133 , @xmath134 , and @xmath135 .
two points are worth mentioning about figure 11 .
first , as it is meant only for illustrative purposes , this figure has not yet been cleaned of multiple detections on overlapping frames , so that some stars appearing in overlap regions are represented two to three times .
second , because of saturation at the bright end , we have lost some of the very brightest , reddest @xmath5 cen giants in this truncated cmd .
an interesting aspect of figure 11 is the appearance of several distinct rgbs , an aspect we explore further elsewhere @xcite .
we have cross - referenced our catalogue with the spectroscopic sample of @xcite .
the latter is a careful analysis of abundance patterns in @xmath5 cen giants , and also provides a membership list of 343 radial velocity members , stars that were selected in the first place as likely members based on proper motions .
the stars studied by @xcite were in two distinct magnitude ranges , one on the giant branch and one along the subgiant branch . in figure 12a
, we show the distribution of the @xcite giants in the ( @xmath2 , @xmath6)@xmath96 diagram , and in figure 12b we show their distribution in the ( @xmath2 , @xmath4)@xmath96 plane .
giants with different metallicity ranges are indicated by different symbols .
the [ fe / h ] ranges were selected so that the mean abundance in each metallicity bin corresponded to more or less convenient [ fe / h ] values ( -1.74 , -1.60 , -1.40 , -1.20 , -0.95 ) on the @xcite abundance scale ; the most metal - rich bin contains a single star at [ fe / h ] = -0.62 .
the distribution of points in figure 12b demonstrates that while the @xmath4 index is less sensitive to broad metallicity changes than it is to luminosity class differences ( note the compressed range of @xmath4 in figure 12b ) , the ( @xmath2 , @xmath4)@xmath96 plane is still able to discriminate roughly between metal - poor and metal - rich giants .
the @xmath5 cen red giant branch is relatively well defined in figure 11 .
a well defined rgb will be useful for our future studies , e.g. , as a means to estimate rough photometric parallaxes for giants with washington photometry .
for each abundance , we fit the center of the rgb sequence analytically ( using the iraftask curfit ) to equations of the form especially at the extreme ends of the @xmath5 cen abundance range , however , we found these fits to be unsatisfactorily constrained due to the small sample sizes , poor distribution in @xmath138 color , and the contribution of photometric errors , which become larger than the mean separation of @xmath139 colors for the metal - poor loci .
the bimodal @xmath138 distribution of the data gave rise to unlikely `` bowing '' between nodes fixed by the two concentration of points .
therefore , to give a more realistic and natural fit to the data , we have taken advantage of the synthetic curves derived for cluster cmds by @xcite , which we `` calibrate '' to our data for @xmath5 cen .
this forging of the synthetic loci to the actual data has the added advantage of extending the usefulness of the @xmath5 cen example to abundance ranges outside those encompassed by the cluster itself , if we assume the _ relative _ placement of the synthetic loci is accurate . after trying various schemes to marry the synthetic and empirical loci , we found that we could obtain reasonable matches of the @xcite curves to the empirical distributions by metallicity with a simple 0.25 mag blueward translation ( in @xmath2 ) and a slight ( @xmath140 ) rotation of the @xcite curves in the two - color plane . thus we find @xmath141 note these relations are defined for the @xcite rgb 14 gyr isochrones , and may not be applicable to their other models , and for clarity they were not considered in the previous discussion of the open clusters .
as pointed out in 1.2 , the need to transform the @xcite relations is likely due to the nonstandard passbands adopted by @xcite @xcite , and it confirms the finding by @xcite that their isochrones are systematically redder than actual globular cluster red giant branches .
table 2 includes the original empirical fits to the metallicity bins in table 1 .
we include only the [ fe / h]@xmath143 , @xmath144 and @xmath145 groups , which guided our transformation , in figure 12b ( dashed lines ) , along with the revised @xcite rgb loci ( solid lines ) , and illustrates the reasonable matches of the latter to the @xmath5 cen data . for future analytical ease the rotated @xcite loci
have been fit by equations ( 7 ) .
the coefficients for these new relations are given in table 2 for the various metallicities .
the difference in the mean abundances for the listed loci in table 2 compared to table 1 is a result of our adopting the @xcite rgb curves ( and therefore their abundance selections ) for the latter . to obtain the absolute magnitude , @xmath146 for a star using the relations in table 1
, one must correct the calculated apparent magnitude for the @xmath5 cen distance modulus , which may be taken as ( m - m)@xmath147 @xcite , as well as the extinction in the @xmath6 band ( @xmath135 for @xmath132 ) .
it should be pointed out that since we have not been able to discriminate between first ascent red giant branch and asymptotic giant branch stars , the relations in table 1 may tend to give systematically overluminous magnitudes if applied to individual first ascent giant branch stars .
the ratio of first ascent to asymptotic giant branch stars in the brighter magnitude range in figure 12a ( i.e. above the horizontal branch ) may be something like four to one @xcite .
on the other hand , if the halo of the galaxy evolved roughly similarly to @xmath5 cen , with a similar age - metallicity relation and initial mass function , then we might expect a similar first ascent to asymptotic giant branch ratio in the field , and application of the table 1 relations to finding distances for halo field stars should provide results that are at least statistically correct for an ensemble of giant stars . in a series of papers over more than a decade doug geisler and collaborators ( @xcite , @xcite , @xcite ) , have continued to refine calibration of the washington photometry system through extensive observations of giant stars in open and globular clusters of all metallicities . in many cases ,
the washington data presented have been supplemented with observations in the @xmath0 filter as a means to discriminate cluster giants from foreground dwarfs . in figure
13b we compile these data in the two - color diagram of interest to our analysis ( similar to figure 12b ) .
many of the stars were observed less frequently in @xmath0 than in the other filters , and this has resulted in larger errors in the @xmath4 color . because of the sensitivity of the technique on relatively reliable @xmath4 colors , we have excluded all stars which have fewer than three measures in @xmath0 . in figure 13a
we show the @xmath148 color - absolute magnitude diagram for those stars with available magnitude data in addition to colors .
table 3 summarizes the data employed in the construction of figure 13 . in both table 3 and
figure 13 the clusters are grouped according to [ fe / h ] ( with groups defined by the curves in figure 13b ) , where the [ fe / h ] and other data are from @xcite for the open clusters and from @xcite for the globular clusters , with two exceptions : the distance moduli for ngc 2360 and ngc 6809 are taken from @xcite and @xcite , respectively . for these particular clusters
the values listed in the databases of @xcite and @xcite differ significantly from those in the literature ( and we suspect an error in the databases ) .
figure 13 demonstrates that the general trends with [ fe / h ] seen in the @xmath5 cen giants are exhibited in the cluster giant population as a whole
namely , that the giants in the most metal - poor clusters show almost no variation of @xmath139 color with @xmath138 , while more metal - rich cluster giants have progressively smaller @xmath139 , particularly at redder @xmath138 , as shown in figure 12b .
we have plotted the five color - color curves calibrated by the various metallicity populations in @xmath5 cen ( table 2 ) , with metallicity increasing towards the curves with smaller @xmath139 .
we also include in figure 13b the rough dwarf / giant demarcation ( _ straight solid lines _ ) recommended in the next section .
it may be seen that , apart from the very reddest and the very bluest stars , the majority of the giants lie above this demarcation line .
the red upturn in the demarcating line is necessitated by the overlap with red dwarfs at @xmath149 ( see figure 14 ) .
for a given abundance range the cluster data in table 3 , however , show much larger scatter in both figure 13a and 13b compared to the @xmath5 cen giants in figures 12a and 12b .
this is partly due to the heterogeneity in the sources and techniques employed to measure the fundamental parameters of distance , abundance and reddening for the table 3 data .
of course , there is also a considerable spread in ages in the objects in table 3 , particularly with the inclusion of the open clusters , which contributes to the spreading of the distributions in figure 13 .
in contrast , the @xcite isochrones we have used were derived from models with ages ( 14 gyr ) appropriate for globular cluster giant branches . because of the various problems with the heterogeneity of the data , we do not use figure 13 to calculate general formulae for the conversion of color into absolute magnitude , as we did in tables 1 and 2 for the @xmath5 cen giants .
we provide the figure 13 data here to lend ancillary support to our assertion that rough abundance discrimination is possible in the @xmath150 diagram .
while it would be nice to extend our table 1 calibrations to higher abundances through inclusion of the metal - rich open clusters , for our present purposes there is no urgency for this since at the magnitudes and galactic latitudes of our first applications of the technique we do not expect to encounter many giants with the metallicities and ages in the range of the metal - rich open clusters .
we aim to return to this problem if and as needed .
we reiterate that much better photometric abundances can be had with inclusion of the washington @xmath37 filter , as the numerous references listed in table 3 can attest .
however , our goal here is ( 1 ) to identify the minimal amount of photometric data necessary to _ find _ field giants that we have already begun to observe spectroscopically for radial velocities and , of course , abundances , and ( 2 ) to point out the additional leverage that our three color filter system has in characterizing the relative abundance of our giant star candidates . in figure 14
, we combine the two - color data of our open cluster main sequences from ngc 3680 ( both single and binary radial velocity members in figure 6d ) and ngc 2477 ( those stars selected as main sequence in figure 8d ) , our red giants from ngc 3680 ( both single and binary radial velocity members in figure 6d ) and @xmath5 cen ( figure 12b ) , the red giant and subgiant branch of melotte 66 ( radial velocity members from the single star locus only , figure 10b ) , the solar abundance field giants from @xcite , as well as the data on field dwarfs and giants from @xcite , and giants from @xcite ( figure 4 ) and the clusters listed in table 3 .
all evolved stars , from geisler s single luminosity class `` iv - v '' to class i , are lumped into one group ( `` non - dwarfs '' ) , as are all of the stars we show from the melotte 66 red giant / subgiant branch .
note that @xcite specifically investigates the sensitivity of the @xmath139 color to giants of different surface gravities and he finds little separation for giants of similar temperatures and metallicities but different log@xmath151 .
we see that the discrimination of evolved stars and dwarfs is rather good over a broad range of @xmath2 , breaking down only at the bluish colors ( @xmath152 ) of subgiant stars just evolved from the main sequence , and at the red end , starting around the colors of late k stars ( @xmath153 ) , where overlap with dwarfs occurs again .
the _ straight solid lines _ in the diagram roughly demarcate the region above which one might expect to find predominantly subgiant and giant stars .
the `` non - dwarf '' objects that fall below these lines in figure 14 are generally geisler stars classified as subgiants , the lowest luminosity melotte 66 subgiants , as well as the peculiar group of melotte 66 stars ( with @xmath154 ) discussed above .
the rather clear separation in figure 14 motivates our survey for distant giant stars . with a 1-m class telescope and ccd imaging
it is easy to obtain the photometric precision necessary to pick out rather complete samples of giant stars to distances of hundreds of kiloparsecs . by adopting this technique ,
we intend to explore the halo and local environment of the milky way and other galaxies in the local group with modest sized telescopes .
we greatly appreciate the assistance of wojtek krzemiski and beata mazur who collected the march 1997 ccd data presented in this paper while wek worked at another telescope .
we thank the referee , gordon drukier , for comments that helped to improve the text .
we gratefully acknowledge many helpful discussions with and comments from doug geisler , and thank him for providing some of his data in machine readable form .
we also appreciate helpful discussions about ( and catalogued data for ) @xmath5 centauri with noella dcruz , robert oconnell and robert rood .
jco thanks ata sarajedini for helpful discussions and mentoring during jco s 1997 research experience for undergraduates stay at kitt peak national observatories , when part of this work was done .
jco and srm acknowledge several grants in aid of undergraduate research from the dean of the college of arts & sciences at the university of virginia .
srm also acknowledges partial support from an national science foundation career award grant , ast-9702521 , a fellowship from the david and lucile packard foundation , and from a cottrell scholarship from the research corporation .
majewski , s. r. , ostheimer , j. c. , kunkel , w. e. , johnston , k. v. , patterson , r. j. & palma , c. 1999a , in iau symp .
190 , _ new views of the magellanic clouds
_ , eds . y .-
h . chu , j. hesser & n. suntzeff , ( san francisco : asp ) , 171 cccccccc @xmath1550.80 to @xmath1551.05 & @xmath1550.95 & 10 & 1.14 - 2.05 & 25.3408 & @xmath15512.2239 & 2.75723 & 0.242 ( 10 ) + @xmath1551.06 to @xmath1551.29 & @xmath1551.20 & 18 & 1.04 - 1.95 & 25.9265 & @xmath15513.9932 & 3.46823 & 0.236 ( 17 ) + @xmath1551.30 to @xmath1551.47 & @xmath1551.40 & 31 & 1.01 - 1.69 & 27.6675 & @xmath15516.5382 & 4.20810 & 0.196 ( 31 ) + @xmath1551.48 to @xmath1551.67 & @xmath1551.60 & 76 & 0.98 - 1.85 & 29.5291 & @xmath15519.3737 & 5.18975 & 0.231 ( 76 ) + @xmath1551.68 to @xmath1551.93 & @xmath1551.74 & 91 & 0.85 - 1.94 & 31.7516 & @xmath15522.9986 & 6.52318 & 0.206 ( 88 ) + @xmath1551.20 & @xmath1550.02598 & 0.12815 & @xmath1550.07806 + @xmath1551.40 & @xmath1550.14206 & 0.31176 & @xmath1550.13879 + @xmath1551.74 & 0.03193 & 0.02258 & @xmath1550.01605 + @xmath1550.47 & @xmath1550.12560 & 0.35652 & @xmath1550.23447 + @xmath1550.79 & @xmath1550.08477 & 0.27546 & @xmath1550.17658 + @xmath1551.25 & @xmath1550.05076 & 0.18657 & @xmath1550.10393 + @xmath1551.77 & 0.01888 & 0.04394 & @xmath1550.02363 + @xmath1552.03 & 0.02907 & 0.02549 & @xmath1550.01225 + @xmath1552.23 & 0.02336 & 0.03606 & @xmath1550.01415 + ngc 7099 & globular&&@xmath156 & 0.03 & 14.57 & & 14(6 ) & g88,gmc92 + & & & & & & & & + ngc 6809 & globular&&@xmath157 & 0.07 & 14.10 & & 15(6 ) & gmc92 + & & & & & & & & + ngc 6752 & globular&&@xmath158 & 0.04 & 13.08 & & 22(3 ) & gcm97 + ngc 1904 & globular&&@xmath159 & 0.01 & 15.53 & & 11(1 ) & gcm97 + ngc 1261 & globular&&@xmath160 & 0.01 & 16.05 & & 7(5 ) & g88 + ngc 6121 & globular&&@xmath161 & 0.36 & 12.78 & & 23(12 ) & g86 + ngc 1851 & globular&&@xmath162 & 0.02 & 15.49 & & 9(6 ) & gcm97 + & & & & & & & & + ngc 362 & globular&&@xmath163 & 0.05 & 14.75 & & 10(10 ) & g86 + ngc 6723 & globular&&@xmath164 & 0.05 & 14.82 & & 13(13 ) & g86 + & & & & & & & & + ngc 6637 & globular&&@xmath165 & 0.17 & 15.11 & & 9(7 ) & g86 + ngc 6352 & globular&&@xmath166 & 0.21 & 14.39 & & 19(1 ) & g86 + & & & & & & & & + ngc 2243 & open&&@xmath167 & 0.03 & 13.17 & & 6(6 ) & g87 + ngc 2204 & open&&@xmath168 & 0.08 & 13.44 & & 8(8 ) & g87 + melotte 71 & open&&@xmath169 & 0.01 & 12.53 & & 9(7 ) & gcm92 + ngc 2360 & open&&@xmath170 & 0.09 & 10.30 & & 6(6 ) & gcm92 + ngc 6940 & open&&@xmath171 & 0.23 & 10.83 & & 10(10 ) & c86 + ngc 6705 & open&&@xmath172 & 0.42 & 12.61 & & 22(6 ) & c86 + ngc 6791 & open&&@xmath173 & 0.18 & 13.93 & & 12(2 ) & c86 +
|
we have begun a survey of the structure of the milky way halo , as well as the halos of other local group galaxies , as traced by their constituent giant stars .
these giant stars are identified via large area , ccd photometric campaigns . here
we present the basis for our photometric search method , which relies on the gravity sensitivity of the mg i triplet + mgh features near 5150 in f - k stars , and which is sensed by the flux in the intermediate band @xmath0 filter .
our technique is a simplified variant of the combined washington / ddo51 four filter technique described by geisler [ 1984 , pasp , 96 , 723 ] , which we modify for the specific purpose of efficiently identifying distant giant stars for follow - up spectroscopic study : we show here that for most stars the washington @xmath1 color is correlated monotonically with the washington @xmath2 color with relatively low scatter ; for the purposes of our survey , this correlation obviates the need to image in the @xmath3 filter , as originally proposed by geisler . to calibrate our ( @xmath2 , @xmath4 ) diagram as a means to discriminate field giant stars from nearby dwarfs , we utilize new photometry of the main sequences of the open clusters ngc 3680 and ngc 2477 and the red giant branches of the clusters ngc 3680 , melotte 66 and @xmath5 centauri , supplemented with data on field stars , globular clusters and open clusters by doug geisler and collaborators . by combining the data on stars from different clusters , and by taking advantage of the wide abundance spread within @xmath5 centauri
, we verify the primary dependence of the @xmath4 color on luminosity , and demonstrate the secondary sensitivity to metallicity among giant stars .
our empirical results are found to be generally consistent with those from analysis of synthetic spectra by paltoglou & bell [ 1994 , mnras , 268 , 793 ] .
finally , we provide conversion formulae from the ( @xmath6 , @xmath2 ) system to the ( @xmath7 , @xmath8 ) system , corresponding reddening laws , as well as empirical red giant branch curves from @xmath5 centauri stars for use in deriving photometric parallaxes for giant stars of various metallicities ( but equivalent ages ) to those of @xmath5 centauri giants .
|
this work was supported by the jsps core - to - core program
non - equilibrium dynamics of soft matter and information " and jsps kakenhi grant number 24340100 .
|
during polymer translocation driven by e.g. voltage drop across a nanopore , the segments in the cis - side is incessantly pulled into the pore , which are then pushed out of it into the trans - side .
this pulling and pushing polymer segments are described in the continuum level by nonlinear transport processes known , respectively , as fast and slow diffusions . by matching solutions of both sides through the mass conservation across the pore
, we provide a physical basis for the cis and trans dynamical asymmetry , a feature repeatedly reported in recent numerical simulations .
we then predict how the total driving force is dynamically allocated between cis ( pulling ) and trans ( pushing ) sides , demonstrating that the trans - side event adds a finite - chain length effect to the dynamical scaling , which may become substantial for weak force and/or high pore friction cases .
# 1 the dynamics of translocation , i.e. , the polymer passage through a narrow pore , has been actively studied more than a decade @xcite .
in addition to its relevance to cellular biological processes , i.e. , biopolymer transports in cells , the phenomenon has found promising applications in genome sequencing and related technology as a nanopore sensor @xcite . being a unique mode of the molecular transport inherent in long flexible polymers ,
there have been numerous attempts to characterize the process and to uncover the underlying physics behind it @xcite .
such efforts have led to the consensus that the _ tension _ is a key physical quantity , which is created in the polymer by the spontaneous segment motions and/or the action of the driving force @xcite .
now we have a fairly good description of the translocation process , which may be categorized according to the magnitude @xmath0 of the bias as follows .
( i ) unbiased regime : the tension imbalance across the pore arising from the segment exchange between cis and trans sides creates a long term power - law decaying memory @xcite .
this retards the process , leading to the anomalous ( sub- ) diffusion of the so - called translocation coordinate @xcite .
( ii ) weakly driven regime : with weak enough force , the polymer conformation is essentially in equilibrium , and the same physics as that in the unbiased regime applies .
the linear response theory then leads to the anomalous drift of the translocation coordinate @xcite .
( iii ) driven regime : when the rate of the driving operation is faster in comparison with the terminal time of the polymer , the genuine nonequilibrium dynamics shows up @xcite .
the action of the driving force is transmitted along the chain in the cis - side , which induces the sequential conformational deformation . here to make the connection to the regime ( ii ) clearer , it is convenient to introduce the length scale @xmath1 and the corresponding time scale @xmath2 @xcite with @xmath3 , @xmath4 being the segment size , thermal energy , respectively , and @xmath5 is a dynamical exponent .
the shortest ( segment scale ) time scale is denoted as @xmath6 , where @xmath7 is the solution viscosity .
initially , the process takes place locally in the close proximity to the pore , and the equilibrium treatment is valid up to @xmath8 , during which the tension is transmitted up to @xmath9-th monomer according to the mechanism in regime ( ii ) .
the subsequent larger scale process is governed by the driving force , thus , described by the nonequilibrium regime ( iii ) , which eventually dominates the scaling limit .
this consideration provides the border between the regimes ( ii ) and ( iii ) as @xmath10 , the threshold force being rather weak for long chains , where @xmath11 is the polymerization index of the chain . for strong enough force ( @xmath12 )
, one only has the regime ( iii ) in the entire process .
we remark here an important distinction between regimes ( ii ) and ( iii ) .
while in equilibrium regime ( ii ) , one can treat the cis and trans sides on even ground , a qualitative difference should unclose as a characteristic feature in the nonequilibrium driven regime ( iii ) . here ,
compared to the sequential stretching of the cis - side chain owing to the propagating tension , the chain portion , which is pushed out of the pore faster than it relaxes in the trans - side , is compressed and resides dynamically in the crowded state . while the proper description of the cis - side dynamics constitutes a basis for our current understanding of the dynamical scaling in the driven translocation @xcite , the trans - side event has been mostly overlooked so far .
although the occurrence of the dynamical crowding has repeatedly been addressed in recent literature e.g. through the observation of simulation snapshots @xcite or the numerical results @xcite , it is poorly understood , and one is left with the question as to how the process in the trans - side can be described and when it becomes important .
below , we attempt to provide the answer to it .
the driving force @xmath0 sucks the cis - side polymer segments into the pore , then pushes them into the trans - side .
@xmath13 the polymer is stretched in the cis - side by the force @xmath14 , while it is dynamically compressed in the trans - side by @xmath15 , and the pore plays a role of the sink and the source in the cis- and trans - sides , respectively .
there is also a frictional force @xmath16 associated with the pore . to handle such a dynamically asymmetric situation on an equal footing ,
we introduce a length scale @xmath17 ; in the cis - side , this length is a tensile blob inversely proportional to the local chain tension , while it corresponds to the concentration blob in the trans - side , from which we can identify the thermodynamic force and the transport coefficient in the respective domains . below the length scale @xmath18 ,
the chain conformation is essentially in equilibrium , i.e. , letting @xmath19 be the segment number constituting the blob , @xmath20 @xcite . in larger scale , the asymmetry shows up in such a way that the problem is essentially one - dimensional in the cis - side , while it is three - dimensional in the trans - side .
keeping such a difference in the effective dimensionality @xmath21 in mind , let us introduce the dimensionless segment density @xmath22 , where @xmath23 and @xmath24 for the cis- and trans - sides , respectively , and @xmath25 .
given the free energy density @xmath26 and the transporting coefficient @xmath27 evaluated as a stokes friction for each blob ( non - draining case ) @xcite , the constitutive relation for the segment flux @xmath28 and the thermodynamic force can be standardized as @xmath29 where @xmath30 is a @xmath21-dimensional nabla operator , and @xmath31 is the segmental diffusion coefficient , and we introduce an exponent @xmath32 , which enables us to treat the effect of hydrodynamic interactions collectively , i.e. , setting @xmath33 reduces to the above non - draining case , while @xmath34 represents the free - draining ( rouse ) dynamics @xcite . with the continuity equation of segments
, this leads to the nonlinear diffusion equation of the type of porous media equation @xcite .
note that the exponent @xmath25 has different signs in the cis- and trans - sides @xcite . from eq .
( [ flux ] ) , the case with plus ( minus ) sign indicates the enhanced @xcite ( suppressed @xcite ) flux in less dense region .
such processes are called fast ( slow ) diffusions , which turns to be crucial in the following .
it is also useful to introduce the velocity potential @xmath35 the gradient of which leads to the segment velocity @xmath36 , hence its naming .
take @xmath37 coordinate perpendicular to a thin wall with a small hole , which is located at the origin ( fig . 1 ) .
we set the cis - side in @xmath38 region , where the polymer is initially placed taking an equilibrium conformation . at @xmath39 ,
the translocation process begins , when one of the chain ends finds the pore .
the segments are sequentially labeled from the first arriving end . at time @xmath40 , @xmath41-th segment is located at the pore , which is traditionally called a translocation coordinate in literature @xcite , while @xmath42-th segment is at the rear end of the moving domain , which represents the dynamics of the tension propagation in the cis - side ( see below and fig . 1 ) . _ cis - side dynamics ( @xmath23 ) : fast diffusion_ our dynamical equation in @xmath23 dimension describes the stretching process @xcite .
the essential nonequilibrium feature here can be captured by what we call the two - phase picture , in which the entire cis - side polymer is divided into the moving domain with a characteristic velocity @xmath43 and the yet quiescent domain ( fig . 1 ) .
integration of the velocity potential in the moving domain range , we have @xmath44 , \nonumber\\\end{aligned}\ ] ] where @xmath45 is the location of the tension propagation front ( see fig . 1 ) , and the positive numerical constant @xmath46 of order unity is absorbed in @xmath47 .
given the nonuniform stretched conformation @xmath48 , one can obtain the force - velocity relation for the cis - side polymer @xmath49 where the boundary conditions @xmath50 ( see ref . @xcite and @xcite arxiv ) and @xmath51 are used ( see fig . 2 ) .
in addition , from the definition of the tension front , we have the relation @xmath52 which traces back to the initial equilibrium conformation at @xmath39 . _ trans - side dynamics ( @xmath24 ) : slow diffusion_ the trans - side decompression process can be described by our dynamical equation in @xmath24 dimension @xcite . taking the spherical symmetry into account
, the integration of the velocity potential leads to @xmath53,\end{aligned}\ ] ] where @xmath54 is the radial position of the decompressed front of the trans - side chain ( see fig . 1 ) , and
the positive numerical constant @xmath55 of order unity is absorbed in @xmath47 .
to proceed , we assume that the velocity field rapidly decreases away from the pore so that the above integral can be dominated by the pore vicinity .
we then find @xmath56 where the boundary conditions @xmath57 and @xmath58 are used .
equation ( [ v_t ] ) states that the force - velocity relation in the trans - side is regularized locally as the condition for the blob closest to the pore ( see fig . 2 )
this may be justified by the physical observation that the injection rate is controlled by the local condition in the pore vicinity , but not aware of the position of the decompression front @xmath59 .
the contrast to eq .
( [ v_c ] ) , which includes the global information @xmath60 , reflects a qualitative difference between the pulling and pushing operations .
indeed , this first blob is pushed by @xmath61 , which is balanced by the drag force against it , i.e. , @xmath62 .
putting @xmath63 into it leads to eq .
( [ v_t ] ) . in eq .
( [ decoupled_a ] ) .
the flory and dynamical exponents are adopted as follows ; ideal chain ( i d ) @xmath64 , excluded volume chain ( ev ) @xmath65 , free - draining ( fd ) @xmath66 and non - draining ( nd ) @xmath33 . ]
_ mass conservation through pore_ we now patch the above described cis- and trans - sides dynamics by requiring the conservation of mass across the pore . the rate of segments sucked from the cis - side into pore is @xmath67 v_1(0 ^ -)$ ] , which is equal to the rate of segments pushed into the trans - side @xmath68 v_3(0^+ ) \xi(0^+)^2 $ ] . using the boundary conditions at the pore
, it can be rewritten as @xmath69 eliminating @xmath43 and @xmath70 using eqs .
( [ v_c ] ) and ( [ v_t ] ) , we obtain the expression of the tension front @xmath71 , hence @xmath42 via eq .
( [ init ] ) , in terms of the force allocation .
@xmath72 where , as already defined , @xmath73 is the number of segments in the blob immediate vicinity of the pore in the cis - side . to get a clear - cut time dependence of the fraction of the cis - side force @xmath74 , where @xmath75 , one can rewrite eq .
( [ fc_ft ] ) as @xmath76 with @xmath77 and the function @xmath78 which is monotone increasing given the domain @xmath79 ( in practice @xmath80 , however .
see below . ) in fig . 3 , we plot @xmath81 as a function of @xmath42 . at @xmath82 , which serves as the initial condition of the nonequilibrium driven regime ( iii ) , the chain in the cis - side forms an initial tensed blob .
this is connected to our remark in the introduction that , until this moment , cis- and trans - sides can be treated on even ground , thus , @xmath83 .
along with translocation process advanced , the tension propagates in cis - side , and @xmath14 monotonically increases and eventually dominates in magnitude over @xmath15 .
since the normalization factor of @xmath42 is @xmath84 , the growth of @xmath81 is unconcerned with the total chain length @xmath11 , thereby , the trans - side effect is irrelevant in the dynamical scaling scenario with the identification @xmath85 in the long chain limit @xcite .
it also predicts the smaller the driving force and/or the larger the pore friction , the larger the finite - chain length effect , since both of these make @xmath86 larger .
the translocation time in the asymptotic limit can be derived as follows . in this limit
, the tension propagation dynamics is described from eqs .
( [ v_c ] ) , ( [ init ] ) and the first equality in eq .
( [ flux_pore ] ) @xmath87 this equation can be closed by an additional relation @xmath88 valid in the driven regime ( iii ) ( see ref .
@xcite and @xcite arxiv ) .
then , the tension propagation time @xmath89 is identified as @xmath90 ; @xmath91 which coincides in leading order to the translocation time . as is evident from fig .
3 , the approach to the asymptote @xmath92 is rather slow , which may cause the subtlety and difficulty to measure the force exponent in eq .
( [ tau ] ) .
finally , note @xmath93 in the limit of the weak force @xmath94 , thus , we have a crossover to the weakly driven regime ( ii ) at this force , where the cis- and trans - sides can be treated on even ground during the whole translocation process . in summary
, we have provided a lucid description on the cis and trans dynamical asymmetry in the driven translocation .
the imbalance is further promoted under higher driving force ( @xmath95 ) , which leads to the almost full stretching in the cis - side . in such situations ,
we only need to modify exponents in the transport equation in cis - side as @xmath96 and @xmath97 @xcite . for our purpose
, this requires replacing the counterparts in eq .
( [ v_c ] ) and the first equation of eqs .
( [ flux_pore ] ) .
we can then show a similar result with @xmath98 and @xmath99 .
note that @xmath86 no longer has the meaning of the initial tensed blob in this strong force regime ; @xmath100 at @xmath101 and getting smaller with the force , which indicates the faster approach to the asymptote @xmath102 . a basic equation ( eq .
( [ flux ] ) ) can also be applied to the polymer transport in different space dimensions with appropriate exponents @xmath103 and @xmath5 .
therefore , the same asymmetric features would be verified in the translocation process , for instance , conducted in the slit geometry .
|
transverse oscillations of coronal loops were first identified by aschwanden et al .
( 1999 ) and nakariakov et al .
( 1999 ) using the observations of transition region and coronal explorer ( trace ) .
nakariakov et al .
( 1999 ) reported the detection of spatial oscillations in five coronal loops with periods ranging from 258 to 320 s. the decay time was @xmath4 min for an oscillation of @xmath5 mhz .
also wang and solanki ( 2004 ) described a loop oscillation observed on 17 april 2002 by trace in 195 @xmath6 .
they interpreted the observed loop motion as a vertical oscillation , with a period of 3.9 min and a decay time of 11.9 min .
nakariakov et al .
( 1999 ) obtained that the amplitude of the oscillations decreases by more than 50 percent in several oscillation periods . to estimate the energy flux of the euv kink oscillation observed by nakariakov et al .
( 1999 ) , we use the maximum kinetic energy flux of an oscillating loop given by @xmath7 . where @xmath8 , @xmath9 , @xmath10 , @xmath11 and @xmath12 are the mass density , the peak loop velocity , the loop radius , the loop length , and the damping time of oscillation , respectively . taking @xmath13 km , @xmath14 gr @xmath15 for a typical coronal loop in active region and using @xmath16 km s@xmath17 and @xmath18 min given by nakariakov et al .
( 1999 ) , we obtain the energy flux as @xmath19 erg @xmath20 s@xmath17 which is three orders of magnitude smaller than the radiated energy flux @xmath21 erg @xmath20 s@xmath17 in active regions ( see klimchuk 2006 ) .
therefore , the energy flux in the euv kink oscillations is unlikely to be sufficient to cover heating of coronal loops . since the discovery of the coronal green line during the 1869 eclipse which identified as fe xiv spectral line by grotrian ( 1939 )
, different theories of coronal heating have been put forward and debated ( for reviews see , e.g. , hollweg 1991 : narain and ulmschneider 1996 ; walsh and ireland 2003 ; erdlyi 2004 , 2005 ; klimchuk 2006 ; erdlyi and ballai 2007 ; taroyan and erdlyi 2009 ) .
ionson ( 1978 ) was first to suggest that the resonant absorption of mhd waves in coronal plasmas could be a primary mechanism in coronal heating .
since then , much analytical and numerical work has been done on the subject .
rae and roberts ( 1982 ) investigated both eikonal and differential equation approaches for the propagation of mhd waves in inhomogeneous plasmas .
hollweg ( 1987a , b ) considered a dissipative layer of planar geometry to study the resonant absorption of coronal loops .
davila ( 1987 ) derived the resonant heating rate in the low - beta compressible resistive mhd approximation with the shear viscosity and found that the heating does not depend explicitly on the dissipation coefficients .
poedts , goossens , and kerner ( 1989 , 1990 ) developed a finite element code to elaborate on the resonant absorption of alfvn waves in circular cylinders .
sakurai , goossens , and hollweg ( 1991a , b ) , goossens , hollweg , and sakurai ( 1992 ) , steinolfson and davila ( 1993 ) did much work on resonant absorption .
ofman , davila , and steinolfson ( 1994 ) , and erdlyi and goossens ( 1994 ) included viscous and resistive dissipations in their analysis and concluded that the heating rate due to shear viscosity is comparable in magnitude to the resistive resonant heating .
also , they concluded that the heating caused by compressive viscosity is negligible .
goossens , ruderman , and hollweg ( 1995 ) extended the analysis of sakurai et al .
( 1991b ) on resonant alfvn waves in nonuniform magnetic flux tubes for a static mhd equilibrium .
they showed that the conservation law found in ideal mhd is valid in dissipative mhd .
they derived analytical solutions to the dissipative mhd equations for the lagrangian displacement and for the total pressure perturbation .
erdlyi and goossens ( 1995 ) studied the heating of solar coronal loops by resonant absorption of alfvn waves in visco - resistive mhd .
they pointed out that under solar coronal conditions the two dissipative mechanisms are both operational and that electrical resistivity appears to be slightly more important than viscosity .
erdlyi and goossens ( 1996 ) studied the effect of an equilibrium flow on resonant absorption of linear mhd waves in compressible viscous cylindrical magnetic flux tubes .
they showed that the presence of the equilibrium flow is very determinant for resonant absorption and significantly affect the resonance absorption rate suffered by the incoming driving waves .
goossens , andries , and aschwanden ( 2002 ) used the trace data of ofman and aschwanden ( 2002 ) to infer the width of the inhomogeneous layer for 11 coronal loops .
ruderman and roberts ( 2002 ) solved the initial value problem for a zero - beta loop driven by a kink perturbation .
they pointed out that the observed damping rate of the coronal oscillations is due to the resonant damping of the quasi - mode kink oscillations .
they estimated the inhomogeneity length scale ( thickness of the boundary layer ) of the loop with the data of nakariakov et al .
note that nakariakov et al .
( 1999 ) showed that the damping rate of the global kink oscillation observed by the trace can be justified by the energy absorption due to anomalous viscosity . whereas ruderman and roberts ( 2002 ) cleared that for large values of the viscous reynolds number , the damping rate is independent of the viscous reynolds number and proportional to the ratio of the thickness of the inhomogeneous layer and the radius of the tube .
van doorsselaere et al .
( 2004a ) used the leda code ( large - scale eigenvalue solver for the dissipative alfvn spectrum ) to study the resistive absorption of the kink modes of cylindrical models .
they concluded that , when the width of the nonuniform layer was increased , their numerical results differed by as much as 25 per cent from those obtained with the analytical approximation .
safari et al .
( 2006 ) studied the resonant absorption of mhd waves in magnetized flux tubes with a radial density inhomogeneity . using the approximation that resistive and viscous processes
are operative in thin layers surrounding the singularities of the mhd equations , they concluded that as the longitudinal wave number increases , the maximum amplitude of the body eigenmodes shifts away from the resonant layer and causes a decrease in damping rates .
an additional features of the flux tube is that of twist .
there is observational evidence for the existence of twist in the solar atmosphere .
the trace 171 @xmath6 observations of a postflare loop system on 1999 confirmed the existence of the twisted loops in the corona ( see aschwanden 2005 ) .
rotational movement along a loop observed by chae et al .
( 2000 ) indicates the existence of twist .
aschwanden ( 2005 ) discusses observed noncoplanarity of loops which implies kinked field lines which , in turn , are also intimately related to twist . besides
, it is unfeasible that every flux tube within the solar atmosphere is entirely twist free in spite of the random continuous motions observed at the footpoints ( see erdlyi and carter 2006 ) . following aschwanden ( 2005 )
, the magnetic twist parameter @xmath22 can be estimated as @xmath23 . here
@xmath24 and @xmath25 are the azimuthal and axial components of the magnetic field , respectively .
also @xmath26 is the shear angle between the untwisted and the twisted field line .
the geometric shear angle @xmath26 , can observationally be measured in twisted coronal loops .
bennett , roberts , and narain ( 1999 ) examined the influence of magnetic twist on the modes of oscillations of a magnetic flux tube .
they found that twist introduces an infinite band of body modes .
erdlyi and fedun ( 2006 ) studied the wave propagation in a twisted cylindrical magnetic flux tube embedded in an incompressible but also magnetically twisted plasma .
they found that increasing the external magnetic twist from 0 to 0.3 caused an increase in the normalized periods of sausage mhd waves approximately by 1@xmath272@xmath28 .
erdlyi and fedun ( 2007 ) extended the work of erdlyi and fedun ( 2006 ) to the case of compressible plasma .
they found that increasing the internal twist may result in about up to @xmath29 changes in periods for sausage modes .
erdlyi and carter ( 2006 ) studied the propagation of mhd waves in a fully magnetically twisted flux tube consisting of a core , annulus and external region .
they investigated their analysis by considering magnetic twist just in the annulus , the internal and external regions having straight magnetic field .
two modes of oscillations occur in this configuration ; surface and hybrid modes .
they found that when the magnetic twist is increased the hybrid modes cover a wide range of phase speeds , centered around the annulus , longitudinal alfvn speed for the sausage modes .
carter and erdlyi ( 2007 ) investigated the oscillations of a magnetic flux tube configuration consisting of a core , annulus and external region each with straight distinct magnetic field in an incompressible medium .
they found that there are two surface modes arising for both the sausage and kink modes for the annulus - core model where the monolithic tube has solely one surface mode for the incompressible case .
also they showed that the existence and width of an annulus layer has an effect on the phase speeds and periods .
carter and erdlyi ( 2008 ) used the model introduced by erdlyi and carter ( 2006 ) to include the kink modes .
they found for the set of kink body modes , the twist increases the phase speeds of the modes .
also they showed that there are two surface modes for the twisted shell configuration , one due to each surface , where one mode is trapped by the inner tube , the other by the annulus itself .
mikhalyaev and solovev ( 2005 ) investigated the mhd waves in a double magnetic flux tube embedded in a uniform external magnetic field .
the tube consists of a dense hot cylindrical core surrounded by a co - axial shell .
they found two slow and two fast magnetosonic modes can exist in the thin double tube .
verwichte et al . ( 2004 ) , using the observations of trace , detected multimode oscillations for the first time .
they found that two loops are oscillating in both the fundamental and the first - overtone standing kink modes . according to the theory of mhd waves , for uniform loops
the ratio of the period of the fundamental to the period of the first overtone is exactly 2 , but the ratios found by verwichte et al .
( 2004 ) are [email protected] and [email protected] .
however , these values were corrected with the improvement of the observational error bars to [email protected] and [email protected] , respectively , by van doorsselaere , nakariakov , and verwichte ( 2007 ) .
also verth , erdlyi , and jess ( 2008 ) added some further corrections by considering the very important effects of loop expansion and estimated a period ratio of 1.54 .
all these values clearly differ from 2 .
this may be caused by different factors such as the effects of curvature ( see e.g. van doorsselaere et al .
2004b ) , leakage ( see de pontieu , martens , and hudson 2001 ) , density stratification in the loops ( see e.g. andries et al . 2005 ; erdlyi and verth 2007 ; karami and asvar 2007 ; safari , nasiri , and sobouti 2007 ; karami , nasiri , and amiri 2009 ) , magnetic field expansion ( see verth and erdlyi 2008 ; ruderman , verth , and erdlyi 2008 ; verth et al .
2008 ) and magnetic twist ( see e.g. erdlyi and fedun 2006 , 2007 ; erdlyi and carter 2006 ; karami and barin 2009 ) .
karami and barin ( 2009 ) studied both the oscillations and damping of standing mhd surface and hybrid waves in coronal loops in presence of twisted magnetic field .
they considered a straight cylindrical incompressible flux tube with magnetic twist just in the annulus and straight magnetic field in the internal and external regions .
they showed that both the frequencies and damping rates of both the kink and fluting modes increase when the twist parameter increases .
they obtained that the period ratio @xmath2 of the fundamental and first - overtone for both the kink and fluting surface modes are lower than 2 ( for untwisted loop ) in presence of the twisted magnetic field .
ruderman ( 2007 ) studied nonaxisymmetric oscillations of thin twisted magnetic tubes in a zero - beta plasma by taking into account the longitudinal density stratification . using the asymptotic analysis
, he showed that the eigenmodes and eigenfrequencies of the kink and fluting oscillations are described by a classical sturm - liouville problem for a second - order ordinary differential equation .
he also concluded that the results concerning nonaxisymmetric waves in twisted magnetic tubes obtained by bennett et al .
( 1999 ) for incompressible plasmas can be applied to global nonaxisymmetric waves , i.e. kink and fluting modes , in coronal loops despite that the coronal plasma is a low - beta plasma .
this conclusion was also in a good agreement with erdlyi and fedun ( 2007 ) .
in the present work , our aim is to investigate the effect of twisted magnetic field on the resonant absorption of standing mhd waves in the coronal loops to justify the rapid damping of oscillations and deviation of the period ratio @xmath2 from 2 observed by trace .
this paper is organized as follows . in section 2
we derive the equations of motion , introduce the relevant connection formulae and obtain the dispersion relation . in section 3
we give numerical results .
section 4 is devoted to conclusions .
the linearized mhd equations for an incompressible plasma are @xmath31 @xmath32 @xmath33 where @xmath34 , @xmath35 and @xmath36 are the eulerian perturbations in the velocity , magnetic field and thermal pressure , respectively ; @xmath8 , @xmath37 , @xmath38 and @xmath39 are the mass density , the electrical conductivity , the viscosity and the speed of light , respectively . in the momentum
( [ mhd1 ] ) like ruderman and roberts ( 2002 ) we write the viscous force in a simplified form @xmath40 instead of the classical braginskii s expression for the viscosity tensor in a magnetized plasma ( braginskii 1965 ) . because according to ruderman and roberts ( 2002 ) , under typical coronal conditions , the coefficient of the shear viscosity is at least 10 orders of magnitude smaller than that of the compressional viscosity .
however , in the problem of oscillations of coronal loops , dissipation is only important in an alfvnic dissipative layer embracing an ideal resonant magnetic surface .
numerical studies by ofman et al .
( 1994 ) , and erdlyi and goossens ( 1994 , 1995 ) have shown that in alfvnic dissipative layers only the shear viscosity is significant , all other terms in braginskii s tensorial expression being neglected .
equation ( [ mhd3 ] ) satisfies the incompressibility condition .
the assumption of incompressibility reduces the direct full applicability to coronal loops except for kink modes observed by the trace as they are highly incompressible perturbations ( see carter and erdlyi 2007 ) .
recently goossens et al .
( 2009 ) showed that in the thin tube approximation neglecting contributions proportional to @xmath41 then the frequency of the kink wave and its damping due to resonant absorption are the same in the three cases including a compressible pressureless plasma , an incompressible plasma and a compressible plasma which allows for mhd radiation .
this is expected as kink modes are linearly incompressible .
note that here the energy equation is absent .
because for incompressible plasmas when sound speed ( or adiabatic gas index ) goes to infinity then the energy equation is decoupled from the system of linearized mhd equations .
the simplifying assumptions are as follows . * the background magnetic field
is assumed to be @xmath42 where @xmath43 , @xmath44 are constant and the magnetic field is uniformly twisted inside and outside the tube ( see fig .
[ rho_va_bphi_p_ptot ] ) .
this is unphysical as @xmath45 , and caution has to be exercised .
following ruderman ( 2007 ) , to satisfy the shafranov - kruskal stability criterion we assume that the azimuthal component of magnetic field is smaller than the axial component .
this assumption is also compatible with observed weakly twisted coronal loops ( see erdlyi and fedun 2006 ) .
the stability of a twisted magnetic flux tube has been studied in some details by bennett et al .
( 1999 ) , ruderman ( 2007 ) , and carter and erdlyi ( 2008 ) .
note that the choice of @xmath46 implies that there is a constant longitudinal current density along the flux tube . * from safari et al .
( 2006 ) and karami et al .
( 2009 ) , the density profile is assumed to be @xmath47(r - r)+\rho_{\rm e},&(r_1<r < r),&\\ \rho_{{\rm e}},&(r > r),&\\ \end{array}\right.\ ] ] where @xmath10 denotes the radius of the tube and @xmath48 is the radius at which the resonant absorption occurs .
also @xmath49 and @xmath50 are the interior and exterior constant densities of the tube , respectively ( see fig .
[ rho_va_bphi_p_ptot ] ) .
note that for resonant absorption a number of density profiles have been considered .
for instance , ofman et al . ( 1994 , 1995 ) have taken @xmath51 with @xmath52 . here
since we use the thin boundary approximation , we assume that the inhomogeneous layer and the resonance layer coincide ( see goossens et al .
therefore any well behavior function of the density like that selected by ofman et al .
( 1994 , 1995 ) in the resonance layer , i.e. the region @xmath53 where the singularity occurs , can be approximated with a linearly varying profile . *
the equilibrium condition , i.e. radial force balance equation @xmath54}=-\frac{b^2_{\phi}(r)}{4\pi r}$ ] , gives the equilibrium plasma pressure @xmath55 as @xmath56 where @xmath57 is the plasma pressure at the center of the tube .
since we have no discontinuity in the background magnetic field , then the thermal pressure remains continuous across @xmath58 ( see fig .
[ rho_va_bphi_p_ptot ] ) .
* tube geometry is a circular with cylindrical coordinates , ( @xmath59 ) .
* there is no initial steady flow over the tube .
* viscous and resistive coefficients , @xmath38 and @xmath37 respectively , are constant .
* @xmath60- , @xmath61- and @xmath62- dependence for any of the components @xmath63 and @xmath64 is @xmath65 .
here @xmath66 , @xmath11 is the length of the tube , and @xmath67 , @xmath68 are the longitudinal and azimuthal mode numbers , respectively . in the remainder of this section the following steps are taken : a ) in @xmath69 and @xmath70 , dissipative terms are neglected .
solutions of eqs .
( 1 ) and ( 2 ) are obtained as in bennett et al .
( 1999 ) , erdlyi and carter ( 2006 ) , and karami and barin ( 2009 ) ; b ) in @xmath71 , within which the resonant layer resides , we use the thin boundary ( tb ) approximation in which we assume that the inhomogeneous layer and the dissipative layer coincide ( see goossens et al .
this enables us to avoid solving the non - ideal mhd equations in the inhomogeneous layer .
the jumps across the resonant layer are found by the prescriptions of sakurai et al .
( 1991a ) ; c ) substituting the solutions ( a ) in jump conditions ( b ) gives an analytical expression for a dispersion relation to be solved for the frequencies and the damping rates .
following bennett et al .
( 1999 ) , erdlyi and carter ( 2006 ) , and karami and barin ( 2009 ) , in the absence of dissipations , taking time derivative of eq .
( [ mhd2 ] ) and substituting for @xmath72 from eq .
( [ mhd1 ] ) , the resulting equation yields to bessel s equation for the eulerian perturbation in total pressure @xmath73 as @xmath74{\delta p_{\rm t}}=0,\label{mhd4}\end{aligned}\ ] ] where @xmath75 and @xmath76,\label{mhd6}\end{aligned}\ ] ] @xmath77 note that eq .
( [ mhd4 ] ) is valid only outside the inhomogeneous layer .
equation ( [ mhd4 ] ) is same as the result exactly derived by bennett et al .
( 1999 ) , and erdlyi and carter ( 2006 ) .
( [ mhd4 ] ) in its general form for a compressible plasma was derived by erdlyi and fedun ( 2007 ) .
note that subscripts @xmath78 ( which correspond to the internal region ) are replaced by @xmath79 corresponding to the external region .
solutions of eq .
( [ mhd4 ] ) for the interior region ( @xmath80 ) are : @xmath81 where @xmath82 and @xmath83 are bessel and modified bessel functions of the first kind , respectively . in the exterior region ( @xmath84 )
, the waves should be evanescent .
solutions are @xmath85 where @xmath86 is the modified bessel function of the second kind .
one must note that from eq .
( [ mhd6 ] ) , solutions ( [ mhd9 ] ) are satisfied only for the frequencies @xmath87 . our numerical results in sect .
3 confirms that this condition is hold for the obtained frequencies . according to the connection formulae given by sakurai et al .
( 1991a ) , the jump across the boundary ( resonance layer ) for @xmath88 and @xmath73 is @xmath89 = - { \rm i}\pi \frac{1}{|\delta| } \frac{{g_b } } { { \rho b^2 } } c_{\rm a},\label{mhd10}\end{aligned}\ ] ] @xmath90 = - { \rm i}\pi \frac{1}{|\delta| } \frac{{ab_z f}}{{2\pi \rho b^2}}c_{\rm a},\label{mhd11}\end{aligned}\ ] ] where @xmath91 note that @xmath92 is the thickness of the inhomogeneous layer and @xmath93 is the radius at which the singularity occurs .
davila ( 1987 ) showed that in the resonance absorption , however , the damping rate is independent of the dissipation coefficient values .
but the resonance layer width scales as @xmath94^{1/3}$ ] .
estimates for typical coronal values suggest that resonant layers have thicknesses from 0.3 km to 250 km ( davila 1987 ) . according to ofman et al .
( 1994 ) , and erdlyi and goossens ( 1994 , 1995 ) for resonant absorption the main contribution comes not from bulk viscosity but from shear viscosity . if we use the reynolds @xmath95 and lundquist @xmath96 numbers given by ofman et al .
( 1994 ) , and taking @xmath97 km , @xmath98 , @xmath99 , @xmath100 , and interior alfvn velocity @xmath101 km s@xmath17 for a typical coronal loop , then one can get @xmath102 km which is very close to the thickness of the inhomogeneous layer @xmath103 km .
therefore we can use the thin boundary approximation which assumes that the thickness of the resonance layer is the same as the inhomogeneous layer width ( see goossens et al .
2009 ) . substituting the solutions of eqs .
( [ mhd8])-([mhd9 ] ) in jump conditions gives the dispersion relation as @xmath104 where @xmath105 , @xmath106 is damping rate and @xmath107 @xmath108 with @xmath109 where ( j ) stands for ( i ) or ( e ) .
also a prime on @xmath83 and @xmath86 indicates a derivative with respect to their appropriate arguments .
the results for the body waves are the same as eqs .
( [ mhd14])-([mhd15 ] ) , except that @xmath83 is replaced by @xmath82 everywhere .
equations ( [ mhd10])-([mhd11 ] ) show that when the twist is absent , i.e. @xmath110 , the total eulerian pressure will be continuous and only @xmath111 jumps across the boundary which is in agreement with sakurai et al .
( 1991a ) . for the surface modes , solving eq .
( [ mhd13 ] ) yields one mode whose frequency exists between @xmath112 and @xmath113 . but for the body modes , there is an infinite band of body modes with the frequencies centered around @xmath112 .
the existence of an infinite set of body waves in the presence of twist has been already introduced by bennett et al .
( 1999 ) .
note that the distinction between surface and body waves is not so important when we study the kink waves in slender magnetic tubes .
recently goossens et al .
( 2009 ) concluded in the thin tube approximation , the kink mhd waves do not care about propagating ( body wave ) or evanescent ( surface wave ) behavior in the internal part of the flux tube .
although under coronal condition they are formally the body waves , they have typical properties of the surface waves .
for example , the radial displacement takes its maximum value at the tube boundary .
this is why ruderman and roberts ( 2002 ) suggested to call these waves the global kink waves .
also erdlyi and carter ( 2006 ) showed that under coronal conditions , for longer wavelength values the sausage surface mode only exists for smaller amounts of twist in the incompressible plasma .
the numerical solution of the dispersion relation ( [ mhd13 ] ) yields to frequencies @xmath114 and damping rates @xmath115 , which are characterized by a trio of wavenumbers ( @xmath116 , @xmath117 and @xmath118 ) that actually count the number of nodes or antinodes along @xmath119- , @xmath61- and @xmath62-directions , respectively .
as typical parameters for a coronal loop , we assume @xmath120 km , @xmath98 , @xmath99 , @xmath100 , @xmath121 gr @xmath15 , @xmath122 g. for such a loop one finds @xmath101 km s@xmath17 , @xmath123 rad s@xmath17 .
the effects of twisted magnetic field on both the frequencies @xmath124 and damping rates @xmath106 are calculated by numerical solution of the dispersion relation , i.e. eq . ( [ mhd13 ] ) .
the results are displayed in figs .
[ m1l1-surface - wa ] to [ bandwith ] .
figures [ m1l1-surface - wa ] to [ m132-surface - wa ] show the frequencies , damping rates and also the ratio of the oscillation frequency to the damping rate of the fundamental and first - overtone @xmath125 kink @xmath126 and fluting @xmath127 surface modes versus the twist parameter , @xmath128 for relative inhomogeneous layer width @xmath99 . according to carter and erdlyi ( 2008 ) , the kink @xmath126 speed is independent of the sound speed . hence , the kink modes are highly incompressible , and this makes the limit of incompressibility be of great interest , not just for wave studies in the deeper part of the solar atmosphere ( e.g. where the plasma - beta is high ) but can also be directly applicable from the lower solar atmosphere to the corona . for all other modes , @xmath129
, compressibility may be necessary .
figures [ m1l1-surface - wa ] to [ m132-surface - wa ] reveal that : i ) the frequencies , the damping rates and the ratio @xmath130 increase when the twist parameter increases .
the result of @xmath124 is in good agreement with that obtained by carter and erdlyi ( 2008 ) .
also the behavior of @xmath124 and @xmath106 versus the twist parameter are in good concord with that obtained by karami and barin ( 2009 ) .
but , the numerical values of our damping rates for @xmath1311 , 2 and 3 are six , seven and eight orders of magnitude , respectively , greater than those obtained by karami and barin ( 2009 ) .
this is due to existence of the resonant absorption which is absent in their work .
note that the dissipative processes in karami and barin ( 2009 ) depend on the reynolds @xmath132 and lundquist @xmath133 numbers .
they obtained the damping rates with @xmath134 and @xmath135 given by ofman et al .
but in the resonance absorption , the damping rate is independent of the dissipation coefficient values and only the resonance layer width depends on @xmath132 and @xmath133 ( see davila 1987 ) .
ii ) for @xmath0 with increasing @xmath22 , the frequencies , damping rates and their ratio , for instance @xmath136 , @xmath137 and @xmath138 , increase @xmath1398 - 66 , @xmath1396 - 50 and @xmath1392 - 11 percent , respectively , compared with an untwisted loop .
iii ) for a given @xmath117 and @xmath22 , when the longitudinal mode number , @xmath118 , increases , both the frequencies and damping rates increase . but the ratio @xmath130 : for @xmath0 , for the twists lower than 0.003 increases and then decreases ; for @xmath140 , always decreases ; but for @xmath141 , for the twists lower than 0.115 decreases and then increases .
iv ) for a given @xmath118 and @xmath22 , when the azimuthal mode number , @xmath117 , increases , the frequencies and damping rates increase but the ratio @xmath130 decreases . here in our calculations , the sausage modes ( @xmath142 )
are absent . because following edwin and roberts ( 1983 ) , and roberts , edwin , and benz ( 1984 ) , the sausage modes have a cut - off longitudinal wave number @xmath143 , and they are only expected in fat and dense loops . according to aschwanden ( 2005 ) the longitudinal wave number cut - off @xmath143 for sausage - mode oscillations
is expressed as a requirement of the loop length - to - width ratio @xmath144 as a function of the density contrast @xmath145 between the external and internal loop densities .
for instance , for a typical active region , loops which have a density contrast of the order of @xmath146 would be required to have length - to - width ratios of @xmath147 .
the period ratio @xmath2 of the fundamental and first - overtone , @xmath125 modes of both the kink ( @xmath0 ) , and fluting ( @xmath1 ) surface waves versus the twist parameter is plotted in fig .
[ surface - p1p2 ] .
figure shows that for both kink and fluting modes , the period ratio @xmath2 decreases when the twist parameter increases .
for instance , @xmath2 decreases from 2 ( for untwisted loop ) and approaches below 1.6 , 1.2 and 1.1 for @xmath117 = ( 1 , 2 and 3 ) , respectively , with increasing the twist parameter .
note that when the twist is zero , the diagrams of @xmath2 do not start exactly from 2 .
this may be caused by the radial structuring @xmath148 .
but for the selected thin tube with @xmath98 , this departure is very small , @xmath149 for kink ( @xmath0 ) and @xmath150 for fluting ( @xmath151 ) modes , and does not show itself in the diagrams ( see mcewan et al .
figure [ surface - p1p2 ] clears that for kink modes @xmath152 , the ratio @xmath2 is 1.821 when @xmath22=0.0065 , and is 1.584 for @xmath22=0.0212 .
these are in good agreement with the period ratios observed by van doorsselaere et al .
( 2007 ) , [email protected] and [email protected] , respectively , deduced from the observations of trace .
van doorsselaere et al .
( 2007 ) considered only the effect of density stratification and neglected the effect of loop expansion .
it has been shown by verth and erdlyi ( 2008 ) that failure to take account of the flux tube expansion causes the density scale height to be overestimated by a factor of approximately 2 .
verth et al .
( 2008 ) considered both effects of the longitudinal density stratification and magnetic field expansion and estimated the correct coronal density scale height .
they reported a period ratio of 1.54 deduced from the observations of trace for the kink modes . here in our numerical calculations , if we take @xmath22=0.0255 and @xmath153 then we can obtain the period ratio @xmath2=1.539 for the kink modes which is in good agreement with that reported by verth et al .
this shows that the magnetic twist can achieve deviations from 2 of the same order of magnitude as in the model of verth et al .
( 2008 ) containing the longitudinal density stratification and magnetic field expansion .
note that for the kink modes , the only period ratio obtained by karami and barin ( 2009 ) was 1.882 with @xmath22=0.0065 . therefore considering the magnetic twist in presence of the resonant absorption not only improved the period ratio estimated by karami and barin ( 2009 ) but also justified another period ratio , 1.54 , reported by verth et al .
( 2008 ) .
figure [ bandwith ] displays the frequency band width , @xmath154 , including infinite set of the fundamental kink @xmath152 body modes versus the twist parameter .
figure [ bandwith ] presents that @xmath154 increases when the twist parameter increases .
this is in good agreement with the result obtained by carter and erdlyi ( 2008 ) , and karami and barin ( 2009 ) .
resonant absorption of standing mhd surface and body waves in coronal loops in the presence of the twisted magnetic field is studied . to do this ,
a typical coronal loop is considered as a straight cylindrical incompressible flux tube with magnetic twist that undergoes a radial density structuring . under the thin tube approximation ,
the thickness of the inhomogeneous layer is same as the resonance layer width .
hence , the radial structuring is assumed to be a linearly varying density profile in the inhomogeneous layer . using the relevant connection formulae ,
the dispersion relation is obtained and solved numerically for obtaining both the frequencies and damping rates of the fundamental and first - overtone kink and fluting modes .
our numerical results show that \i ) the frequencies and damping rates as well as the ratio of the oscillation frequency to the damping rate of both the kink @xmath126 and fluting @xmath127 surface waves increase when the twist parameter increases . for @xmath0 with increasing @xmath22 , the ratio @xmath138 changes from 39.3 to 43.5 which approximately one order of magnitude is greater than the ratio reported by nakariakov et al .
( 1999 ) , wang and solanki ( 2004 ) , and verwichte et al .
( 2004 ) deduced from the trace data ; \ii ) the period ratio @xmath2 , for both the kink ( @xmath0 ) and fluting ( @xmath1 ) surface modes is lower than 2 ( for an untwisted loop ) in the presence of the twisted magnetic field .
the result of @xmath155 for kink modes is in agreement with the trace observations ; \iii ) frequency bandwidth of the fundamental kink ( @xmath0 ) body modes increase when the twist parameter increases .
note that the main differences between the present work and the work of karami and barin ( 2009 ) can be summarized as follows : i ) in our model , the magnetic field was assumed to be uniformly twisted inside and outside the tube . whereas they considered only a uniformly twisted magnetic annulus inside the loop .
ii ) they considered the weak damping due to viscous and resistive dissipations which can not justify the rapid damping of the coronal loops observed by nakariakov et al .
( 1999 ) , and wang and solanki ( 2004 ) . in the case one includes the resonant absorption , the rapid damping can be justified .
iii ) our model could justify the period ratios of kink modes reported by van doorsselaere et al .
( 2007 ) , i.e. @xmath1561.82 , 1.58 , and verth et al .
( 2008 ) , i.e. @xmath1561.54 , whereas karami and barin ( 2009 ) only justified @xmath1561.82 .
the authors thank the unknown referee for very valuable comments .
the authors also thank dr . h. safari for a number of useful discussions .
this work was supported by the department of physics , university of kurdistan , sanandaj , iran .
|
the standing quasi modes in a cylindrical incompressible flux tube with magnetic twist that undergoes a radial density structuring is considered in ideal magnetohydrodynamics ( mhd ) .
the radial structuring is assumed to be a linearly varying density profile . using the relevant connection formulae , the dispersion relation for the mhd waves
is derived and solved numerically to obtain both the frequencies and damping rates of the fundamental and first - overtone modes of both the kink ( @xmath0 ) and fluting ( @xmath1 ) waves .
it was found that a magnetic twist will increase the frequencies , damping rates and the ratio of the oscillation frequency to the damping rate of these modes .
the period ratio @xmath2 of the fundamental and its first - overtone surface waves for kink ( @xmath0 ) and fluting ( @xmath1 ) modes is lower than 2 ( the value for an untwisted loop ) in the presence of twisted magnetic field . for the kink modes , particularly , the magnetic twists @xmath30.0065 and 0.0255 can achieve deviations from 2 of the same order of magnitude as in the observations . furthermore , for the fundamental kink body waves , the frequency bandwidth increases with increasing the magnetic twist .
= 16truecm = 23truecm = -1.3truecm = -2truecm = 1.20 cm = 1.60 cm * keywords : * sun : corona sun : magnetic fields sun : oscillations
|
the most general interaction hamiltonian for nuclear @xmath0-decay which includes all possible interaction types consistent with lorentz - invariance @xcite contains 5 different terms , so - called scalar ( s ) , vector ( v ) , tensor ( t ) , axial - vector ( a ) and pseudoscalar ( p ) contributions .
the standard model ( sm ) of the weak interaction excepts only v and a interactions which leads to the well - known @xmath1 structure of the weak interaction .
however , the presence of scalar and tensor types of weak interaction is today ruled out only to the level of about 8% of the v- and a - interactions @xcite .
a possible admixture of a scalar or tensor type weak interaction in @xmath0-decay can be studied by determining the @xmath2 angular correlation .
this correlation for unpolarized nuclei can be characterized by the @xmath2 angular correlation coefficient _ a _
@xcite which is for instance for pure fermi transitions given by : @xmath3 since the standard model assumes only v- and a- interactions @xmath4 ( @xmath5 for pure gamow - teller ) .
any admixture of s- to v- ( t to a ) interaction in a pure fermi ( gamow - teller ) decay would result in @xmath6 ( @xmath7 ) .
it can be shown that the two leptons in @xmath0-decay will be emitted preferably into the same direction for a _ _ _ _ v ( t ) interaction and into opposite directions for an s ( a ) interaction leading to a relatively large energy of the recoil ion for a v ( t ) interaction and a relatively small recoil energy for an s ( a ) interaction ( fig .
[ fig:1-recoils - v - s ] ) . the witch experiment as a primary goal aims to measure the shape of the recoil energy spectrum in nuclear @xmath0-decay with high precision in order to deduce the parameter _ a _ which will give information on a possible scalar or tensor type interaction .
similar experiments were recently performed at triumf @xcite and at berkley @xcite , while another experiment is ongoing at ganil @xcite .
in order to fulfill the goal of the experiment the witch set - up was built and installed in isolde ( cern ) .
the main feature of this set - up is a combination of a double penning trap structure to store the @xmath0-decaying radioactive ions and a retardation spectrometer to probe the energy of the daughter recoil ion ( fig .
[ fig:2-setup ] ) .
the radioactive ions are produced by the isolde facility @xcite and are then first trapped , cooled and bunched by rextrap @xcite .
the ion bunches ejected from rextrap have 60 kev ( optionally 30 kev ) energy and are guided into the witch set - up .
they are decelerated from 60 kev to about 80 ev in the vertical beamline by means of the pulsed drift cavity @xcite and thus can be trapped in the first penning trap , the cooler trap .
this trap serves to cool the ions further down in energy by buffer gas collisions and to prepare the ion cloud to be injected into the second penning trap , the decay trap .
both traps are placed in a 9 t magnetic field .
after @xmath0-decay ions leave the decay trap and those emitted in the direction of the spectrometer ( upper half hemisphere ) are probed for their energy by an electrostatic retardation potential .
the magnetic field in the region of the retardation analysis plane is 0.1 t. the working principle of the retardation spectrometer is based on the principle of adiabatic invariance of the magnetic flux and similar to the @xmath0-spectrometers used for the neutrino rest - mass measurements .
the ions that pass the retardation plane are re - accelerated to @xmath810 kev and focused onto the micro - channel plate ( mcp ) detector which is equipped with delay line anodes for position sensitivity .
changing the retardation potential over the necessary range allows to measure a recoil ion spectrum with high precision .
simulations performed show that for a reasonable measurement time one can reach a precision for the @xmath2 angular correlation parameter @xmath9 which corresponds to @xmath10 at c.l.=95% .
further improvements in order to push the limit down to @xmath11 , or @xmath12 at c.l.=95% , should be possible .
more details about the witch set - up can be found in @xcite .
tests performed to check the beam transport , to confirm the functionality of pulsing down the ions and to verify the basic operation of the penning traps are described in @xcite . here
we present only the most recent results .
the main purpose of the cooler trap is to prepare the ion cloud for the measurement .
this involves buffer gas cooling , cleaning from possible contaminates , better centering and cooling of the magnetron motion to prevent the cloud from expanding .
the last three tasks can be achieved by _ sideband cooling _
this technique was tested by measuring the signal on the mcp detector installed in the middle of the retardation spectrometer as a function of the applied frequency near the true cyclotron frequency .
the isotope used is @xmath13cs . in a first step
all ions in the cooler trap were driven out of the center with dipole excitation and then the quadrupole excitation was applied ( fig .
[ fig:3-omega - c ] , square data points ) .
only ions which can be re - centered can reach the detector because they have to pass the 3 mm diameter pumping diaphragm ( see fig .
[ fig:6-traps - diaph ] ) . as can be seen from fig .
[ fig:3-omega - c ] a nice resonance curve is obtained with @xmath14(fwhm)=346 hz corresponding to a mass resolving power @xmath15(fwhm)@xmath162000 .
this is enough to completely remove isotopic contaminations which can be present in the incoming beam .
a pure quadrupole excitation without prior dipole excitation was applied as well .
as one can see from fig .
[ fig:3-omega - c ] ( triangle data points ) it has a similar effect as sideband cooling meaning that the initial injection into the cooler trap was not ideal , e.g. ions were injected off center and/or they had a too large magnetron component .
however , using the quadrupole excitation still allowed to correct for this .
it has to be noted here that the magnetic field homogeneity allows to improve the mass resolving power up to @xmath17 .
in june 2006 the witch experiment observed first recoil ions following the @xmath0-decay of trapped @xmath18 in ( fig .
[ fig:4-retard ] ) .
the incoming beam was a mixture of @xmath19 in ( @xmath20=3.11 s ) and @xmath21 in ( @xmath20=3.7 s ) .
as a first step a simple on - off measurement was performed : for 1.4 s no retardation was applied thus measuring the total signal and then for 0.4 s 400 v retardation was used stopping all recoil ions and thus providing the background signal ( fig .
[ fig:4-retard ] , left ) .
as expected the count rate drops down when the retardation is on .
the data were fitted and a half - life of 3.3(3 ) s was found which is in a good agreement with the incoming mixed beam of @xmath22 in .
the latter confirms that the ions trapped in the decay trap are @xmath22 in . as a next step different retardation schemes were used .
however , strong discharge in the re - acceleration part of the set - up ( see sect . [ sec:1 ] ) was observed .
thus the acceleration voltages were significantly reduced which , however , caused a focusing effect , i.e. recoil ions were focused differently depending on their energy and could miss the detector .
nevertheless , a correct general behavior of the spectrometer was observed ( fig .
[ fig:4-retard ] , right ) : increasing the retardation leads to lower count rate on the detector while decrease of the retardation corresponds to higher count rate .
recently a position sensitive mcp of 4 cm diameter was installed as the main detector in the system .
this mcp allows to detect a larger beam spot and to study a number of systematic effects like a possible dependence of the beam size on the ion energy , the focusing problem which already played a role during the beam time ( see sect .
[ subsec : recoils ] ) and the effect of the @xmath0-background .
this mcp is of the same type as described in @xcite .
position sensitivity is realized by delay line anodes , i.e. the one - dimensional position of the ion hit is deduced from the difference of the propagation times to both ends of the corresponding wire .
the position sensitivity read - out was tested by installing a special mask in front of the detector .
reading back the timing signals and reconstructing the hit positions allowed to reproduce the shape of the mask , thus confirming the functionality of the system .
presently achieved position resolution is of the order of 1 mm .
this mcp is located in the magnetic field of about 30 g , but no dependency of the mcp resolution on b - field was studied yet .
as for technical improvements we currently investigate the possibility to change part of the pumping diaphragm to a @xmath0-detector ( optionally to a scintillator detector)(fig .
[ fig:6-traps - diaph ] ) .
this will allow to have an additional normalization between different trap loads , to have a start signal for tof measurements and to evaluate the @xmath0-background on the main mcp detector on - line .
the witch physics program is not limited to the search for exotic scalar or tensor currents . if one assumes that the standard model is correct measuring the @xmath2 correlation parameter _ a _ yields information about the fermi to gamow - teller mixing ratio .
due to the properties of the retardation spectrometer different charge states will appear at different positions in the spectrum thus allowing to study the charge state distribution after nuclear @xmath0-decay .
also , ions from @xmath0-decay and from electron capture ( ec ) can be separated in the witch spectrometer .
this opens the possibility to determine the ec/@xmath23-branching ratio .
however , for this one has to include the charge state distribution as a parameter and has to measure the whole recoil spectrum .
one can also study the possibility to search for heavy neutrinos .
the presence of these neutrinos can be indicated by subtle kinks in the recoil spectrum .
the witch experiment has as the primary goal to measure exotic scalar or tensor currents in the weak interaction .
the set - up is installed in isolde ( cern ) .
all major components of the set - up including beamline transport , double penning trap system and the retardation spectrometer are checked for their functionality and show good or acceptable efficiency ( table [ tab:1-eff ] ) .
the overall efficiency of the set - up was improved by up to a factor of 30 in comparison with 2004 ( see @xcite ) .
this allowed to detect first recoil ions in the witch set - up .
a number of systematic effects still has to be studied and certain improvements have to be made in order to further increase the efficiency and to perform precise measurement of the recoil ion spectrum .
beamline transfer + pulse down & 50% & @xmath850%injection into b - field & 100% & 10%cooler trap efficiency & 100% & @xmath860%transfer between traps & 100% & @xmath880%storage in the decay trap & 100% & 100%fraction of ions leaving the decay trap & 40% & not studiedshake - off for charge state n=1 & 10% & not studiedtransmission through the spectrometer & 100% & @xmath850%mcp efficiency & 60% & 52.3(3)% @xcite this work is supported by the european union grants fmrx - ct97 - 0144 ( the eurotraps tmr network ) and hpri - ct-2001 - 50034 ( the nipnet rtd network ) , by the flemish fund for scientific research fwo and by the projects goa 99 - 02 and goa 2004/03 of the k.u.leuven .
lee and c.n .
yang , phys .
, 104 , 254 ( 1956 ) .
jackson , s.b .
treiman and h.w .
wyld , phys .
, 106 , 517 ( 1957 ) .
n. severijns , m. beck and o. naviliat - cuncic , rev .
phys . , 78 , 991 ( 2006 ) , nucl - ex/0605029 .
jackson , s.b .
treiman and h.w .
wyld , nucl .
phys . , 4 , 206 ( 1957 ) .
a. gorelov et al . , phys .
lett . , 94 , 142501 ( 2005 ) .
n. scielzo et al . , phys .
rev . a , phys .
lett . , 93 , 102501 ( 2004 ) .
d. rodriguez et al . , eur .
j. a , 25 ( suppl.1 ) , 705 ( 2005 ) .
e.g. adelberger et al . ,
lett . , 83 , 1299 ( 1999 ) .
e. kugler , hyp .
, 129 , 23 ( 2000 ) .
f. ames et al .
, nucl . instr . and meth .
a , 538 , 17 ( 2005 ) . s. coeck et al . , nucl .
instr . and meth .
a , submitted . m. beck et al . , nucl.instr . and meth .
a , 503 , 567 ( 2003 ) . v. kozlov et al . ,
int . j. mass spec .
, 251 , 159 ( 2006 ) . m. knig et al .
j. mass spec .
, 142 , 95 ( 1995 ) .
e. linard et al .
instr . and meth .
a , 551 , 375 ( 2005 ) .
|
primary goal of the witch experiment is to test the standard model for a possible admixture of a scalar or tensor type interaction in @xmath0-decay .
this information will be inferred from the shape of the recoil energy spectrum .
the experimental set - up was completed and is under intensive commissioning at isolde ( cern ) .
it combines a penning trap to store the ions and a retardation spectrometer to probe the recoil ion energy . a brief overview of the witch set - up and the results of commissioning tests performed until now are presented .
finally , perspectives of the physics program are reviewed .
|
inhomogeneity plays an important role in many highly correlated materials@xcite . for example
stripe phases which break translational symmetry of real space in one direction have been observed in high temperature superconductors@xcite .
similarly , a wigner crystal ( wc ) phase@xcite which has only discrete translational symmetry of triangular lattices and six - fold rotational symmetry has been observed in electrons on the surface of liquid helium@xcite and in ultra - clean two - dimensional hole gases@xcite . for interactions which decay as the power law @xmath3
, a classical argument shows that the potential energy scales with density as @xmath4 in two dimensions ( 2d ) and the kinetic energy as @xmath5 .
the ratio of the potential to the kinetic energy scales as @xmath6 which becomes @xmath7 , for coulomb interactions , a constant for @xmath8 and @xmath9 for @xmath1 potentials .
a stable wc crystal phase occurs when the interaction energy is greater than the kinetic energy@xcite .
hence , a wc is expected in the high density regime or low density regime depending on the power law decay of interactions . for coulomb forces
this suggests that there is a phase transition at decreasing densities from a homogeneous fermi liquid ( fl ) phase , which is conducting , to a wc phase , which is insulating . in this regime ,
perturbative methods fail and one must resort to numerical studies @xcite . in reality , many important systems do not have pure power law interactions . for two - dimensional electron gases ( 2degs ) in semiconductor inversion layers or quantum wells , the presence of a nearby gate modifies the coulomb interaction . image charges
emerge at the gate and screen the coulomb ( @xmath0 ) interaction , which become @xmath1 at large distances beyond a length scale given by the distance to the gate .
the effects of such long - ranged forces changes the energetics of the 2deg@xcite .
recently , the effects of long - ranged forces on superconductivity have been explored too@xcite .
a similar experimental scenario occurs in ionic liquid transistors@xcite , where an electrolyte is used as a dielectric in a standard field effect transistor configuration .
positive and negative charges accumulate at opposite ends of the electrolyte .
finally , recent progress in cooling techniques have allowed the study of degenerate dipolar gases@xcite including wigner crystallization of dipoles with @xmath1 potentials@xcite and collective modes@xcite .
such objects interact as @xmath1 at large separations . at short distances the interaction potential
is modified by the interactions between the electronic clouds .
the common feature of these systems is that there is a length scale in the interactions .
they are composed of charges of opposite signs which are located in layers separated by a dielectric .
as such , the interparticle potential crosses over between two different power law regimes .
in this work , we study the ground state properties and collective modes of polarized electrons in 2d with screened coulomb interactions .
such system is realized in 2deg in the presence of a nearby gate and a magnetic field parallel to the surface to avoid significant orbital effects .
we also describe the related problem of a gas of fermionic dipoles which have a finite size . to be concrete
, we consider an interaction interpolating smoothly between a @xmath0 ( @xmath1 ) potential at short ( long ) distances @xmath10 where @xmath2 represents the screening length .
while three dimensional effects have been investigated in a variety of contexts@xcite , here , we consider a 2d model .
our phase diagram agrees with studies of 2degs with a nearby gate@xcite , and of electrons on the surface of liquid helium@xcite . in our analysis
, we consider a many - body wave function in the form of a slater determinants and variational single - particle wave functions for the fl and wc phases . in the wc phase ,
the variational parameter is the spatial extent of the single - particle wave functions . optimizing the total energy at zero temperature we explicitly calculate and compare the ground state energies of the wc and fl phases .
we note that the effects of screened interactions on the capacitance of 2degs and ionic liquids have also been explored@xcite .
the corresponding zero - temperature @xmath11 phase diagram is presented in fig .
[ fig : zero - temperature - phase - diagram ] as a function of the length scale @xmath2 and density @xmath5 parametrized by @xmath12 , where @xmath13 is the bohr radius in the presence of the dielectric constant @xmath14 .
the fl and wc phases are indicated in the phase diagram and are labeled according to the dominant interaction regimes . for fixed @xmath15
there is a regime of densities where the wigner crystal is energetically more favorable than a uniform fl .
we also show the asymptotic forms of the phase boundaries in dashed ( red ) lines . in the regime dominated by @xmath1
, the asymptotic phase boundary corresponds to @xmath16 , while in the coulomb case it is @xmath17 constant @xmath18 .
intuitively , for @xmath1 potentials ( left of the @xmath19 line ) the kinetic ( @xmath20 ) and potential @xmath21 energies are comparable when @xmath16 , whereas for the @xmath0 regime , the potential energy scales as @xmath22 and these energies are comparable when @xmath23 is a constant ( @xmath24 ) .
notice that for @xmath25 there is no wc phase .
the existence of the maximum is a general feature that also follows from the physical argument described above ; if @xmath2 is too small the potential energy can not be of the same order as the kinetic energy .
the limit @xmath26 ( coulomb regime ) is not strictly accessible within our method .
however , by extrapolating , we find that the wc phase more favorable than the wc phase for @xmath27 . for @xmath28 and increasing density from zero to finite values , we find a first order phase transition from a dipolar fl to a wc . this phase transition has been studied numerically in the context of dipole gases with pure @xmath1 forces@xcite . with additional increases of the density beyond the @xmath19 line ,
the wc crosses over to a coulomb wc . upon further increase of the density
, the wc melts into a fl .
this reentrant behavior for the fl phase is a remarkable characteristic of any 2d system interacting via an effective potential which crosses over from long to short range .
[ fig : zero - temperature - phase - diagram ] constitutes one of our main results . in what follows
we show how this phase diagram is obtained . in section [ sec : model ] , we present the hamiltonian and in section [ sec : fl ] , we compute the energy of the homogeneous fl and discuss its collective modes , as they are modified by the presence of @xmath2 . in section [ sec :
wc ] , we obtain the energy in the wc phase with a variational wave function . in section [ sec : disscusion ] , we discuss in detail the physical significance of our results , and we compare the energies of these two states in different density regimes in order to determine the ground state of the system .
in addition , we discuss the scaling of energy of the wc with respect to density , give a schematic finite temperature phase diagram , and we present our concluding remarks .
we start from the microscopic hamiltonian @xmath29 where the first term corresponds to the kinetic energy and the second term is given in eq .
( [ eqn : two - body - int ] ) , which describes the interactions between spinless fermions .
the variable @xmath2 is the ` screening ' length and we treat it as a phenomenological parameter .
it could represent a number of physical quantities , e.g. , twice the distance between a polarized 2deg and a nearby gate or the classical effective size of the dipoles . here , @xmath14 is the dielectric constant of the medium .
note that for a fixed density and in the limit @xmath30 , the interaction becomes @xmath31 .
whereas , in the limit @xmath32 , it becomes @xmath33 to lowest non - vanishing order in @xmath2 . in section [ sec : disscusion ] , we also discuss polarized fermionic dipoles where there is a factor of two in the two - body interaction . for both fl and wc phases , we consider the many - body ground state wave function to be @xmath34 , \label{eqn : many - body - wave function}\end{aligned}\ ] ] where @xmath35 is the number of particles , and @xmath36 is an operator that describes the anti - symmetrization necessary for indistinguishable fermions .
results for indistinguishable bosons and for classical distinguishable particles are straight - forward extensions . for simplicity ,
we assume that the spin degrees of freedom are frozen , such that the wave function describing individual particles @xmath37 are either plane waves in the fl phase , or gaussian centered at individual sites of a triangular lattice in the wigner crystal phase .
to begin , let us consider the homogeneous fl phase .
our system has two length scales .
the first scale is @xmath2 , while the second is the interparticle distance @xmath38 set by the density @xmath39 , where @xmath40 is the fermi momentum of the fl with fermi energy @xmath41 .
the relative size of these length scales determines the range of the potential . using the relation between @xmath5 , @xmath40 and @xmath42
, we see that for @xmath43 the fermions interact with a @xmath1 potential , whereas for @xmath44 they interact with the coulomb @xmath0 potential ( see fig .
[ fig : zero - temperature - phase - diagram ] ) .
the energy per particle of a uniform and polarized fl is @xmath45 , \label{eqn : fermi - liquid - energy}\end{aligned}\ ] ] where @xmath46 is scaled by the bohr radius @xmath47 , and the energy @xmath48 is scaled by the bohr energy @xmath49 .
the functions @xmath50 are the @xmath51 order modified bessel function of the 1st kind and @xmath52 are the @xmath51 order struve functions . in obtaining eq .
( [ eqn : fermi - liquid - energy ] ) , we used the fact that the two - body interactions shown in eq .
( [ eqn : two - body - int ] ) have fourier transform @xmath53 , where @xmath54 is the magnitude of the two dimensional momentum vector . the first term in eq .
( [ eqn : fermi - liquid - energy ] ) is the kinetic energy and scales as the density @xmath5 . the second term in eq .
( [ eqn : fermi - liquid - energy ] ) is the hartree contribution , which also scales as @xmath5 . in the limit @xmath55 , @xmath56 is formally divergent as interaction potential is long - ranged , that is , it behaves as @xmath0 . in a 2deg
the positive background charge cancels the hartree term and , in this case , @xmath57 , in agreement with known results@xcite .
the last term within the square brackets is the fock contribution , which becomes @xmath58 as @xmath59 and leads to @xmath60 in this @xmath1 regime . to @xmath61
the fock term cancels the hartree term for pure @xmath1 potentials .
physically , the system behaves as a charge neutral fl of dipoles .
we note that including a background to enforce charge neutrality , as in 2degs with @xmath0 potentials , leads to negative energies and a self - bound system . however , in the present case the energies are positive and the system has positive pressure .
the energy of the uniform fl is shown in fig .
[ fig : efl_esf_e_wc ] for @xmath62 .
the length scale @xmath2 in the potential introduces important modifications to the collective excitations in the fl phase .
the fourier transform of eq .
( [ eqn : two - body - int ] ) at zero momentum is well - defined , @xmath63 . for these kinds of forces
, it is very important to keep the hartree and fock terms in any approximation to the self - energy@xcite as this provides a conserving approximation@xcite .
the collective modes are given by the zeros of the dielectric function @xmath64 .
a non - conserving rpa calculation gives , _ q = v_f ( |/2)^1/2 q , for @xmath65 and @xmath66~q$ ] for @xmath67 .
hence , we expect the system to support zero sound modes .
a detailed calculation likely renormalizes the zero sound velocity , but does not alter the qualitative physical phenomenon ; see , for example , the analysis performed in the context of polarized dipole gases@xcite .
the speed of the zero sound diverges when @xmath30 , since the power expansion in @xmath68 breaks down in this limit .
indeed , the plasmon dispersion relation in 2degs has been extensively studied in gaas@xcite semiconductor quantum wells .
the fourier transform of the pure coulomb potential @xmath69 diverges at zero momentum and leads to gapless @xmath70 mode dispersion . according to the variational principle , the energy obtained using the many - body wave function described in eq .
( [ eqn : many - body - wave function ] ) is a rigorous upper bound for the true ground state energy of the system . in the following analysis
, we perform a variational calculation of the ground state energy of the wc phase and compare it with that of the fl phase obtained in eq .
( [ eqn : fermi - liquid - energy ] ) .
in the wigner crystal phase , the single - particle wave functions are localized at sites @xmath71 of a 2d triangular lattice , @xmath72\exp\left[-({\ensuremath{\mathbf{r}}}-{\ensuremath{\mathbf{r}}}_i)^2/(2 s^2)\right],$ ] where @xmath73 is the site position and @xmath74 parametrizes the 2d size of the wave function .
these single - particle wave functions are approximately orthonormal , since the overlap at different sites is exponentially small , @xmath75 = \exp \left [ -r_{ij}^2/(4 s^2 ) \right ] , $ ] provided that the separation @xmath76 between sites @xmath71 and @xmath77 is much larger than the extent @xmath74 of the single - particle wave function .
we denote the lattice spacing by @xmath78 and consider the regime of weakly overlapping single - particle wave functions , where @xmath79 .
the localized nature of the many - body wave function is reasonable only in this regime .
explicitly , the energy per particle in the wc phase is , @xmath80 where the functions @xmath81 are lattice sums given by @xmath82 appearing in the second term of eq .
( [ eqn : wigner - crystal - energy ] ) , and @xmath83 appearing in the third term of eq .
( [ eqn : wigner - crystal - energy ] ) . here , we defined the dimensionless variables @xmath84 , @xmath85 , @xmath86 and @xmath87 is the @xmath51 order bessel function of the first kind .
the first term is the zero point motion due to the localization of the particles at the lattice sites .
this term favors large single - particle wave functions .
the second and third terms are the hartree - fock ( hf ) contribution ( respectively ) .
the hf contribution is non - monotonic in @xmath74 .
the extension to particles with bosonic statistics results in a change of the sign of the fock term whereas for classical particles only the hartree term is present .
we comment on the limiting behavior of eq .
( [ eqn : wigner - crystal - energy ] ) .
fermions interact via a coulomb potential when @xmath30 . in this limit ,
the ground state energy is & = & + _ ij e^-|r_ij^2/4|s^2 i_0(|r_ij^2/4|s^2 ) + & & - _ ij e^-|r_ij^2/2|s^2 .
[ eqn : coloumbwcenergy ] in the regime of non - overlapping wave functions @xmath88 , the asymptotic form of the bessel function is @xmath89 .
as we see , the series in the second ( hartree ) term diverges when @xmath90 as expected , since we have not included a neutralizing background charge . in this case , taking a finite @xmath35 gives a well defined energy which is non - monotonic in @xmath74 .
we find that this behavior extends to the regime of @xmath91 and produces a wc state energy @xmath92 that does not have a minimum as a function of @xmath74 , when the density is either too small or too large . on the other hand , in the limit of @xmath93
, we obtain that the @xmath61 contribution from the fock term cancels the hartree term ( as in the fl phase ) and the first non - vanishing term is @xmath94 . since the resulting expression is not very illuminating , we omit it here .
for @xmath91 , the integral appearing in eq .
( [ eqn : wigner - crystal - energy ] ) needs to be calculated numerically .
we computed the energy per particle as a function of density @xmath95 in the wc phase for a 2d triangular array of @xmath96 and @xmath97 particles . in fig .
[ fig : zero - temperature - phase - diagram ] , we show the resulting phase diagram for 55 particles , but we would like to emphasize that the qualitative behavior found is essentially unchanged for higher number of particles .
the calculation of phase boundaries for the full range of parameters @xmath2 and @xmath42 is very intensive when it involves a large number of particles , however , we do not expect any major qualitative changes in behavior .
we establish the phase boundary between solid and liquid phases by minimizing the energy in the solid phase with respect to the variational parameter @xmath74 . a typical behavior of the energy as a function @xmath74
is shown in fig .
[ fig : efl_esf_e_wc ] .
when there is a local minimum of @xmath98 located at @xmath99 , we compare the energy @xmath100 with the energy of the screened fl with the same density .
we find that the energy of the wc is always lower than the fl energy in the regime of tested values of @xmath2 .
however , a local minimum of @xmath98 does not exist for densities below a minimum @xmath101 and for densities above a maximum @xmath102 , where the wc phase is unstable .
this establishes that the wc phase is unstable for densities @xmath5 satisfying the condition @xmath103 and @xmath104 , in which case , the fl phase is the stable phase . a phase transition from a ferromagnetic fl to a ferromagnetic wc phase in a 2deg with @xmath0 interactions is expected to occur at @xmath105 according to recent monte - carlo simulations@xcite . in the @xmath30 limit , we find such a phase transition at @xmath106 , with finite - sized samples .
we note that our results do not rigorously apply in this regime since interactions are of pure coulomb character and a neutralizing background charge must be explicitly considered . here
, we are interested in the regime of @xmath91 , where our results are expected to be qualitatively correct .
we note that the optimized energy values in the wc can be fitted with a power law e_(s_0 ) ~n^4/3 in the low density regime , i.e. , @xmath107 where interactions are of @xmath1 character .
this is shown explicitly in fig .
[ fig : ewcvsn ] for @xmath108 .
we tested the scaling for @xmath109 to within @xmath110 error , but we can not completely rule out a crossover behavior@xcite . a similar @xmath111 scaling behavior has been studied in 2degs with a @xmath1 potential and spin - orbit interactions@xcite . in the high density regime ( @xmath112 )
the energy is a linear power law of the density @xmath113 , where interaction are of @xmath0 character .
this scaling persists in the same tested range and within the same error . as mentioned earlier , the potential shown in equation eq .
( [ eqn : two - body - int ] ) describes the interactions between polarized electrons in a clean 2deg with a nearby gate . as is well known , by applying a magnetic field parallel to the surface one avoids significant orbital effects and a one component 2deg is obtained . however , disorder is always present and makes it difficult to reach the low density regime where the reentrant fl is predicted .
in contrast , another system where disorder does not play a role , corresponds to degenerate polarized dipolar fermi gases@xcite .
we extended our results to describe a system of dipoles in 2d whose centers move in the plane perpendicular to the polarization axis@xcite . in this case , @xmath2 could parametrize a hard core radius ( or size of the molecule ) below which interactions are no longer of the @xmath1 . in this new situation , the interaction potential of eq .
( [ eqn : two - body - int ] ) acquires a prefactor of two , but the calculations are entirely analogous . in fig .
[ fig : zero - temperature - phase - diagram2 ] we show the phase diagram for 43 dipoles .
we observe the same features as for screened 2degs .
for @xmath114 there is no stable wc phase . in the coulomb limit
the wc is stable for @xmath115 . in the expressions for the energy of the fl and wc a prefactor of 2
is needed in the interaction energy .
the collective modes dispersion is @xmath116 for @xmath117 and @xmath118 q$ ] for @xmath67 .
our results suggest that realistic dipole gases with size @xmath119 , are fermi liquids at low _ and _ very high densities .
we provided estimates for the critical densities as a function of @xmath2 in a simplified model .
if only @xmath1 interactions are considered@xcite one expects a wc instability with increasing densities near the dashed red line on the left in fig .
[ fig : zero - temperature - phase - diagram2 ] , where the interaction energy is of the order of the kinetic energy . however this red dashed line is strongly modified ( blue line ) , when the hard core is considered .
next , we discuss the finite temperature phase diagram of the 2deg with screened interactions .
the assumption is that the interactions can be modeled with the simple form eq .
( [ eqn : two - body - int ] ) . at finite temperatures
, we expect the melting of the wc to a liquid - like state@xcite . in fig .
[ fig : phase_diagram_t ] , we show the phase diagram in the ( @xmath120 ) parameter space . in the @xmath121 plane ,
the thermodynamics of the classical 2deg is determined by the quantity @xmath122 , which is the ratio of the interaction energy to the kinetic energy per particle . for a classical 2deg ,
this ratio is @xmath123 . for @xmath124 ,
the kinetic energy dominates and the system behaves as a classical liquid . for @xmath125 , the coulomb interaction dominates and we expect a classical wc .
the phase boundary is given by the criterion@xcite @xmath126 .
for temperatures below @xmath127 , we obtain a quantum wc . at high densities the melting temperature is eventually driven down with increasing density as the pauli principle prevents strong correlations .
there is a quantum phase transition in the regime with pure coulomb interactions at @xmath128 .
this transition can be seen along the @xmath23 line in fig .
[ fig : phase_diagram_t ] , where @xmath129 in the hartree - fock approximation .
this defines a critical density above which @xmath130 and a quantum fluid is recovered .
the full melting line and the quantum / classical crossover in the plane @xmath121 is sketched in fig .
[ fig : phase_diagram_t ] , see also @xcite . connecting the wc phases of the @xmath121 and @xmath131 planes
is a dome , where a wc is stable . within
this dome there exists at least four regions corresponding to quantum vs classical and @xmath0 vs @xmath1 regimes . outside the dome
there are only quantum and classical liquid - like phases .
+ in conclusion , we have obtained the phase diagram of fermions with interactions which interpolate smoothly between short - ranged and long - ranged regimes in 2d .
the phase diagram obtained is generic to any system with such crossover , e.g , clean 2degs in semiconductor inversion layers or quantum wells in the presence of a screening gate , and polarized dipolar gases .
|
we study fermions in two dimensions interacting via a long - ranged @xmath0 potential for small particle separations and a short - ranged @xmath1 potential for larger separations in comparison to a length scale @xmath2 .
we compute the energy of the wigner crystal and of the homogeneous fermi liquid phases using a variational approach , and determined the phase diagram as a function of density and @xmath2 at zero temperature .
we discuss the collective modes in the fermi liquid phase , finite temperature effects on the phase diagram , and possible experimental realizations of this model .
|
in the standard model ( sm ) the phase in the cabibbo - kobayashi - maskawa ( ckm ) matrix is the only source of cp violation . the small amount of cp violation in the sm , however , is not sufficient to explain the baryon - antibaryon asymmetry of the universe @xcite .
the lagrangian of the minimal supersymmetric standard model ( mssm ) contains several complex parameters , which can give rise to new cp - violating phenomena @xcite . in the neutralino sector of the mssm two complex parameters
appear , which lead to cp - violating effects in neutralino production and decay .
these parameters are the @xmath11 gaugino mass parameter @xmath0 and the higgsino mass parameter @xmath1 .
the phase of the @xmath12 gaugino mass parameter @xmath13 can be eliminated by an @xmath11-symmetry of the model .
one of the main goals of a future @xmath6 linear collider will be a careful study of the properties of supersymmetric ( susy ) particles @xcite .
the neutralinos @xmath14 , @xmath15 , will be particularly interesting , because the lightest neutralino @xmath16 is expected to be the lightest susy particle ( lsp ) , which is stable if @xmath17-parity is conserved .
the second lightest neutralino , @xmath18 , will presumably be among the lightest visible susy particles .
therefore , the study of production and decay of the neutralinos @xmath14 @xcite and a precise determination of the underlying supersymmetric parameters @xmath0 , @xmath13 , @xmath1 and @xmath19 including the phases @xmath20 and @xmath21 of @xmath0 and @xmath1 will play an important role at future linear colliders . the phases of @xmath0 and @xmath1 cause cp - violating effects already at tree level . therefore these effects could be large and thus be measurable at a high luminosity @xmath6 linear collider . methods to determine these parameters in neutralino and chargino production have been presented in @xcite . in particular the method of @xcite
is based on the analysis of the masses and production cross sections of only the light neutralinos @xmath16 and @xmath18 and the light chargino @xmath22 and will allow the determination of the parameters @xmath23 , @xmath13 , @xmath24 and the phases @xmath20 and @xmath21 .
all these methods involve only cp - even quantities , in which the signs of the phases @xmath20 and @xmath21 can not be determined . for an unambiguous determination of @xmath20 and @xmath21 including their signs one has to rely on cp - sensitive observables .
the phases of the complex parameters are constrained or correlated by the experimental upper limits on the electric dipole moments of electron , neutron and the atoms @xmath25hg and @xmath26tl @xcite . in a constrained mssm
the restrictions on the phases can be rather severe
. however , there may be cancellations between the contributions of different complex parameters , which allow larger values for the phases @xcite .
for example , in a constrained mssm and if substantial cancellations are present , @xmath21 is restricted to @xmath27 , whereas @xmath20 and @xmath28 , the phase of the trilinear scalar coupling parameter , turn out to be essentially unconstrained , but correlated with @xmath21 @xcite . moreover , the restrictions are very model dependent . for example , when also lepton flavour violating terms are included , then the restriction on @xmath21 may disappear @xcite .
therefore it is necessary to determine the phases in an unambiguous way by measurements of cp - odd observables . a useful tool to study these cp - violating effects are t - odd observables , based on triple products of momenta or spin vectors of the particles involved @xcite .
in this paper we study a t - odd asymmetry in neutralino production @xmath29 with subsequent leptonic three - body decay @xmath30 with @xmath5 . using the triple product of the initial electron momentum @xmath31 and the two final lepton momenta @xmath32 and @xmath33 , @xmath34
we define a t - odd asymmetry @xmath35 where @xmath36 is proportional to the cross section @xmath37 .
it is essential to include the spin correlations between production and decay , because otherwise the numerator and hence the asymmetry @xmath38 would vanish .
this observable has the advantage that it is not necessary to reconstruct the momentum of the decaying neutralino .
the asymmetry @xmath38 , eq .
( [ asy ] ) , is odd under the naive time - reversal operation . by cpt
it is a cp - sensitive asymmetry , if final - state interactions and finite - widths effects are unimportant .
we will neglect these effects , because they are of higher order . by its definition ,
the asymmetry @xmath38 is the difference of the number of events with the final lepton @xmath39 above and below the plane spanned by @xmath40 , normalised by the sum of these events . in this paper
we will first present the complete analytic formulae at tree level for the cross section and for the asymmetry ( [ asy ] ) of the processes ( [ process1 ] ) and ( [ process2 ] ) , taking into account the full spin correlations between production and decay of the neutralino @xmath18 @xcite .
we will study the dependence of the asymmetry and the cross section on the phases @xmath21 and @xmath20 and present detailed numerical results for an @xmath41 linear collider with @xmath42 gev and polarised beams .
analogous cp asymmetries in neutralino production and subsequent two - body decays have been studied in @xcite .
cp asymmetries based on triple products in decays of scalar fermions have been discussed in @xcite .
cp - odd observables involving the polarisation of the outgoing @xmath43 leptons from neutralinos decaying via two - body decays @xmath44 have been analysed in @xcite .
a systematic study of the impact of the supersymmetric phases on neutralino , chargino and selectron production at linear colliders has been performed in @xcite .
a monte carlo study of the asymmetry ( [ asy ] ) for @xmath45 , @xmath46 , including sm backgrounds and detector effects has been given in @xcite , however , no analytic formulae for the cross section and the asymmetries have been given there .
in @xcite the impact of the phases @xmath47 , @xmath48 , @xmath49 , @xmath21 and @xmath20 on the two - body decays of the third generation sfermions has been analysed in detail .
the influence of these phases on the polarisation of the outgoing fermions in third generation sfermion decays has been studied in @xcite . in section 2
we shortly present the basics of the spin density matrix formalism . in section 3
we give the analytic expressions for the asymmetry @xmath38 and the cross sections .
section 4 contains numerical results and discussions . in section 5
we present our conclusions .
appendices a and b contain some details of the formalism and of the analytic expressions .
the production process @xmath50 , @xmath15 , proceeds via @xmath51 and @xmath52 exchange ( fig .
[ fig : feynprod ] ) . in the decay process
@xmath53 , @xmath54 , @xmath51 and @xmath55 exchanges contribute ( fig .
[ fig : feyndecay ] ) .
( 5,9 ) ( -3,-10 ) ( 1.2,0.4)@xmath56 ( 10.8,0.4)@xmath57 ( 1.2,4.9)@xmath58 ( 10.8,4.9)@xmath59 ( 6.5,3.3)@xmath51 ( 2.5,5 ) ( -4,-10 ) ( 2.1,-1.5)@xmath56 ( 2.1,6.8)@xmath58 ( 7.8,-1.5)@xmath57 ( 7.8,6.8)@xmath59 ( 4.4,2.8)@xmath52 ( 2.5,5 ) ( -4.5,-10 ) ( 1.7,-1.5)@xmath56 ( 1.7,6.8)@xmath58 ( 7.4,-1.5)@xmath60 ( 7.4,6.8)@xmath61 ( 3.9,2.8)@xmath52 ( 5,7 ) ( -5.2,-5.2 ) ( 4.7,1.7)@xmath62 ( 0.3,5.5)@xmath63 ( 4.7,5.5)@xmath64 ( 3.7,7.1)@xmath65 ( 2.2,3.8)@xmath51 ( 5,7 ) ( -5.2,-6 ) ( 4.7,5.5)@xmath62 ( 3.7,7.1)@xmath64 ( 4.7,1.5)@xmath65 ( 0.3,5.5)@xmath63 ( 2.1,3.9)@xmath55 ( 5,7 ) ( -5.2,-6 ) ( 3.7,7.1)@xmath62 ( 4.7,5.5)@xmath64 ( 0.3,5.5)@xmath63 ( 4.7,1.5)@xmath65 ( 2.1,3.9)@xmath55 the interaction lagrangians for these processes are @xcite ( in our notation and conventions we follow @xcite ) @xmath66\ell,\ ] ] @xmath67{\tilde{\chi}}^0_n,\ ] ] @xmath68 with @xmath69 and the couplings @xmath70\,,\ ] ] @xmath71\,,\ ] ] @xmath72 @xmath73 @xmath74 where @xmath75 , @xmath76 is the weak coupling constant ( @xmath77 , @xmath78 ) . @xmath79 and @xmath80 are the charge ( in units of @xmath81 ) and the third component of the weak isospin of the fermion @xmath82 .
@xmath83 is the weak mixing angle and @xmath84 is the ratio of the vacuum expectation values of the higgs fields .
the unitary @xmath85-matrix @xmath86 which diagonalises the complex symmetric neutralino mass matrix is given in the basis @xmath87 @xcite . from these interaction
lagrangians the helicity amplitudes @xmath88 and @xmath89 for production and decay can be calculated . here
@xmath90 denotes the production or decay channel and the respective exchanged particle . @xmath91 and @xmath92 are the helicities of the neutralinos @xmath14 and @xmath93 . in @xmath89
we suppress the label of the helicity of the final neutralino @xmath94 .
the amplitudes @xmath88 for the production process @xmath95 ( see fig . [
fig : feynprod ] ) are @xmath96 where @xmath97 & & \mbox{\hspace*{-.4 cm } }
\phantom{t_p^{\lambda_i\lambda_j}(s , z)=\frac{g^2}{\cos^2\theta_w}\delta^s(z ) } \bar{u}(p_4,\lambda_j)\gamma_{\mu } ( o^{''l}_{ji } p_l+o^{''r}_{ji } p_r)v(p_3,\lambda_i),\label{eq_a1}\\[-0.2 cm ] & & \mbox{\hspace*{-.4 cm } } t_p^{\lambda_i\lambda_j}(t,\tilde{e}_{l } ) = -g^2 f^l_{e i}f^{l*}_{e j } \delta^t(\tilde{e}_l)\bar{v}(p_2 ) p_r v(p_3,\lambda_i ) \bar{u}(p_4,\lambda_j)p_l u(p_1),\label{eq_a2a}\\ & & \mbox{\hspace*{-.4 cm } } t_p^{\lambda_i\lambda_j}(t,\tilde{e}_{r } ) = -g^2 f^r_{e i}f^{r*}_{e j } \delta^t(\tilde{e}_r)\bar{v}(p_2 ) p_l v(p_3,\lambda_i ) \bar{u}(p_4,\lambda_j)p_r u(p_1),\label{eq_a2}\\ & & \mbox{\hspace*{-.4 cm } } t_p^{\lambda_i\lambda_j}(u,\tilde{e}_l ) = g^2 f^{l*}_{e i } f^l_{e j } \delta^u(\tilde{e}_l)\bar{v}(p_2 ) p_r v(p_4,\lambda_j ) \bar{u}(p_3,\lambda_i)p_l u(p_1),\label{eq_a3a}\\ & & \mbox{\hspace*{-.4 cm } } t_p^{\lambda_i\lambda_j}(u,\tilde{e}_r ) = g^2
f^{r*}_{e i } f^r_{e j } \delta^u(\tilde{e}_r)\bar{v}(p_2 ) p_l v(p_4,\lambda_j ) \bar{u}(p_3,\lambda_i)p_r u(p_1 ) \label{eq_a3}\end{aligned}\ ] ] with @xmath98 @xmath99 and the mandelstam variables @xmath100 @xmath101 , @xmath102 , @xmath103 , @xmath104 , @xmath105 , @xmath106 denote the corresponding propagator , mass and width of the exchanged particle .
the amplitudes @xmath89 for the decay @xmath107 ( see fig . [
fig : feyndecay ] ) read @xmath108 with @xmath109 & & \mbox{\hspace*{-.4 cm } } \phantom{t_{d , \lambda_i}(s_i , z)=-\frac{g^2}{\cos^2\theta_w}\delta^{s_i}(z ) } \bar{u}(p_5 ) \gamma_{\mu } ( o^{''l}_{ki } p_l+o^{''r}_{ki } p_r ) u(p_3 , \lambda_i ) , \label{eq_a4}\\[-0.2 cm ] & & \mbox{\hspace*{-.4 cm } } t_{d , \lambda_i}(t_i,\tilde{\ell}_l ) = - g^2 f^l_{\ell k } f^{l*}_{\ell
i } \delta^{t_i}(\tilde{\ell}_l)\bar{u}(p_7 ) p_r v(p_5 ) \bar{v}(p_3 , \lambda_i ) p_l v(p_6),\label{eq_a5a}\\ & & \mbox{\hspace*{-.4 cm } } t_{d , \lambda_i}(t_i,\tilde{\ell}_r ) = -g^2 f^r_{\ell k } f^{r*}_{\ell i } \delta^{t_i}(\tilde{\ell}_r)\bar{u}(p_7 ) p_l v(p_5 ) \bar{v}(p_3 , \lambda_i ) p_r v(p_6 ) , \label{eq_a5}\\ & & \mbox{\hspace*{-.4 cm } } t_{d , \lambda_i}(u_i,\tilde{\ell}_l ) = + g^2 f^l_{\ell i } f^{l*}_{\ell k } \delta^{u_i}(\tilde{\ell}_l)\bar{u}(p_7 ) p_r u(p_3 , \lambda_i ) \bar{u}(p_5 ) p_l v(p_6),\label{eq_a6a}\\ & & \mbox{\hspace*{-.4 cm } } t_{d , \lambda_i}(u_i,\tilde{\ell}_r ) = + g^2 f^r_{\ell
i } f^{r*}_{\ell k } \delta^{u_i}(\tilde{\ell}_r)\bar{u}(p_7 ) p_l u(p_3 , \lambda_i ) \bar{u}(p_5 ) p_r v(p_6 ) \label{eq_a6}\end{aligned}\ ] ] and the kinematic variables @xmath110 the amplitude for the whole process @xmath111 , with @xmath14 decaying , can be written as @xmath112 where @xmath113 $ ] , @xmath114 , @xmath115 are the propagator , mass and width of the decaying neutralino @xmath14 . following the formalism of @xcite , the amplitude squared @xmath116 of the combined processes of production @xmath117 and decay @xmath118 of the neutralino @xmath14 , with neutralino @xmath93 being unobserved , can be written as @xmath119 where @xmath120 refers to the polarisation state of the neutralino @xmath14 , which is described by the polarisation vectors @xmath121 given in appendix a. @xmath122 and @xmath123 are the terms of production and decay independent of the polarisation of the decaying neutralino , whereas @xmath124 and @xmath125 are the terms containing the spin correlations between production and decay . according to our choice of the polarisation vectors @xmath121 , eqs .
( [ eq_21 ] ) ( [ eq_23 ] ) , @xmath126 is the longitudinal polarisation , @xmath127 is the transverse polarisation in the production plane and @xmath128 is the polarisation perpendicular to the production plane of the neutralino @xmath14 .
the explicit expressions for these contributions are given in appendix b. the complete expressions for the amplitude squared including spin - spin correlations for the case when both neutralinos are decaying and for longitudinally polarised @xmath129 beams are given in @xcite .
the differential cross section in the laboratory system is @xmath130 where @xmath131 is the beam energy and @xmath132 is the lorentz invariant phase space element .
integrating over all angles gives the cross section for the combined process of production and decay @xmath133
in this section we will derive the analytic formulae for the cross section @xmath134 and for the asymmetry @xmath38 , eq .
( [ asy ] ) .
as can be seen from the numerator of @xmath38 , for this purpose we have to identify those terms in @xmath116 , eq .
( [ tsquared ] ) , which contain a triple product of the form eq .
( [ tpdef ] ) .
triple products follow from expressions @xmath135 in the terms @xmath124 and @xmath125 , where @xmath136 , @xmath137 , @xmath138 , @xmath139 are 4-momenta and spins of the particles involved ( see appendix b , eqs .
( [ eq_sumsp1 ] ) ( [ eq_kinsp4 ] ) and ( [ eq_dssum ] ) ( [ eq_dssub6 ] ) ) . the expressions @xmath135 are imaginary and when multiplied by the imaginary parts of the respective couplings they yield the terms which contribute to the numerator of @xmath38 , eq . ( [ asy ] ) .
hence only the second term of eq .
( [ tsquared ] ) , which is due to the spin correlations , contributes to the numerator of the asymmetry @xmath38 .
it is convenient to split @xmath140 and @xmath141 into t - odd terms @xmath142 and @xmath143 containing the respective triple product , and t - even terms @xmath144 and @xmath145 without triple products : @xmath146 then the terms of @xmath116 , eq . ( [ tsquared ] ) , which contribute to the numerator of @xmath38 are @xmath147,\ ] ] where the first ( second ) term is sensitive to the cp phases in the production ( decay ) of the neutralino @xmath14 . in the following
we derive explicitly the t - odd contributions to the spin density matrices of production and decay . as can be seen from eq .
( [ eq_kinsp4 ] ) in appendix b , in @xmath148 the expressions @xmath149 vanish for @xmath150 , because they do not contain three linearly independent vectors , hence @xmath151 . only for @xmath152
we get a t - odd contribution to the production density matrix @xmath153 , which is related to the polarisation of the neutralino @xmath14 perpendicular to the production plane ( see the expression for @xmath154 , eq .
( [ eq_22 ] ) in appendix a ) .
we obtain from eqs .
( [ eq_spzz ] ) ( [ eq_spsubs ] ) for @xmath152 @xmath155 with @xmath156 , \label{eq_spzz_to}\\ \sigma_p^{2 , \mathrm{o}}(z\tilde{e}_l ) & = & \frac{g^4}{2 \cos^2\theta_w } l_{\ell } ( 1-p_{e^-})(1+p_{e^+ } ) \nonumber\\ & & \times re\big\{\delta^s(z ) \big[f^{l*}_{\ell i}f^{l}_{\ell j } o^{''l*}_{ij } \delta^{t*}(\tilde{e}_l ) -f^{l}_{\ell i}f^{l*}_{\ell j } o^{''l}_{ij } \delta^{u*}(\tilde{e}_l ) \big]f_4^{2}\big\ } , \label{eq_spzel_to}\\ \sigma_p^{2 , \mathrm{o}}(\tilde{e}_l \tilde{e}_l ) & = & \frac{g^4}{4 } ( 1-p_{e^-})(1+p_{e^+ } ) re\big\ { \delta^{u}(\tilde{e}_l)\delta^{t*}(\tilde{e}_l ) ( f^{l*}_{\ell i})^2 ( f^{l}_{\ell j})^2 f_4^{2}\big\ } , \label{eq_spelel_to}\\ \sigma_p^{2 , \mathrm{o}}(z\tilde{e}_r ) & = & \frac{g^4}{2 \cos^2\theta_w } r_{\ell } ( 1+p_{e^-})(1-p_{e^+ } ) \nonumber\\ & & \times re\big\{\delta^s(z ) \big[f^{r*}_{\ell i}f^{r}_{\ell j } o^{''r*}_{ij } \delta^{t*}(\tilde{e}_r ) -f^{r}_{\ell i}f^{r*}_{\ell j } o^{''r}_{ij } \delta^{u*}(\tilde{e}_r ) \big]f_4^{2}\big\ } , \label{eq_spzel_tor}\\ \sigma_p^{2 , \mathrm{o}}(\tilde{e}_r \tilde{e}_r ) & = & \frac{g^4}{4 } ( 1+p_{e^-})(1-p_{e^+ } ) re\big\ { \delta^{u}(\tilde{e}_r)\delta^{t*}(\tilde{e}_r ) ( f^{r*}_{\ell i})^2 ( f^{r}_{\ell j})^2 f_4^{2}\big\ } , \label{eq_spelel_tor}\end{aligned}\ ] ] where @xmath157 and @xmath158 denotes the degree of longitudinal polarisation of the electron beam and positron beam , respectively .
the function @xmath159 , eq .
( [ eq_kinsp4 ] ) , for @xmath152 reads @xmath160 note that @xmath161 is purely imaginary . when inserted , for example , into the expression for @xmath162 , eq . ( [ eq_spzel_to ] ) ,
@xmath161 is multiplied by the factor @xmath163 , which is non - vanishing only if the couplings are complex .
hence it gives a cp - sensitive contribution to the asymmetry @xmath38 , which depends on the phases @xmath21 and @xmath20 .
analogous contributions come from the other terms in @xmath164 , eqs .
( [ eq_spzz_to]),([eq_spelel_to ] ) ( [ eq_spelel_tor ] ) .
we have to multiply @xmath164 in eq .
( [ cpint ] ) by @xmath165 , for which we obtain from eqs .
( [ eq_dssum ] ) ( [ eq_dssub3 ] ) , appendix b , @xmath166 with @xmath167g^{2}_3 + |o^{''l}_{ki}|^2(g^{2}_1-g^{2}_2)\big ] , \label{eq_dszz_te}\\ \sigma_d^{2 , \mathrm{e}}(z \tilde{\ell}_l ) & = & \frac{4 g^4}{\cos^2\theta_w}l_{\ell } re\big\{\delta^{s_i}(z ) \big [ f^l_{\ell i } f^{l*}_{\ell k } \delta^{t_i*}(\tilde{\ell}_l ) \big(-2 o^{''l}_{ki } g^{2}_1 + o^{''l*}_{ki } g^{2}_3\big)\nonumber\\ & & \phantom{\frac{4 g^4}{\cos^2\theta_w}l_{\ell } re\big\{\delta^{s_i}(z ) } + f^{l*}_{\ell i } f^{l}_{\ell k } \delta^{u_i*}(\tilde{\ell}_l ) \big(2 o^{''l*}_{ki } g^{2}_2 + o^{''l}_{ki } g^{2}_3\big ) \big]\big\},\label{eq_dszel_te}\\ \sigma_d^{2 , \mathrm{e}}(\tilde{\ell}_l \tilde{\ell}_l)&= & 2 g^4 \big [ |f^{l}_{\ell i}|^2 |f^l_{\ell
k}|^2 [ |\delta^{u_i}(\tilde{\ell}_l)|^2 g_2^{2 } -|\delta^{t_i}(\tilde{\ell}_l)|^2 g_1^{2}]\nonumber\\ & & \phantom{2 g^4 \big [ } + re\big\ { ( f^{l*}_{\ell i})^2 ( f^l_{\ell k})^2 \delta^{t_i}(\tilde{\ell}_l ) \delta^{u_i*}(\tilde{\ell}_l)g_3^{2}\big\}\big],\label{eq_dselel_te}\\ \sigma_d^{2 , \mathrm{e}}(z \tilde{\ell}_r)&= & \frac{4 g^4}{\cos^2\theta_w}r_{\ell } re\big\{\delta^{s_i}(z ) \big [ f^r_{\ell i } f^{r*}_{\ell k } \delta^{t_i*}(\tilde{\ell}_r ) \big(2 o^{''r}_{ki } g^{2}_1 -o^{''r*}_{ki } g^{2}_3\big)\nonumber\\ & & \phantom{\frac{4 g^4}{\cos^2\theta_w}r_{\ell}\delta^{s_i}(z ) } + f^{r*}_{\ell i } f^{r}_{\ell k } \delta^{u_i*}(\tilde{\ell}_r ) \big(-2 o^{''r*}_{ki } g^{2}_2 -o^{''r}_{ki } g^{2}_3\big ) \big]\big\},\label{eq_dszel_ter}\\ \sigma_d^{2 , \mathrm{e}}(\tilde{\ell}_r \tilde{\ell}_r)&= & 2 g^4 \big [ |f^{r}_{\ell i}|^2 |f^r_{\ell k}|^2 [ -|\delta^{u_i}(\tilde{\ell}_r)|^2 g_2^{2 } + |\delta^{t_i}(\tilde{\ell}_r)|^2 g_1^{2}]\nonumber\\ & & \phantom{2 g^4 \big [ } - re\big\ { ( f^{r*}_{\ell i})^2 ( f^r_{\ell k})^2 \delta^{t_i}(\tilde{\ell}_r ) \delta^{u_i*}(\tilde{\ell}_r)g_3^{2}\big\}\big],\label{eq_dselel_ter}\end{aligned}\ ] ] where @xmath168 .
\label{eq_g3_a=2}\end{aligned}\ ] ] the kinematic functions @xmath169 , @xmath170 , @xmath171 are real . when multiplied by the purely imaginary @xmath161 , eq .
( [ eq_f4_a=2 ] ) , this leads to triple products sensitive to the cp phases in the production process , which in the laboratory system read : @xmath172 as outlined above , these expressions will be multiplied in eqs .
( [ eq_spzz_to ] ) ( [ eq_spelel_tor ] ) by the factors @xmath163 etc . and
contribute to the first term of eq .
( [ cpint ] ) and , hence , to the numerator of the asymmetry @xmath38 , eq .
( [ asy ] ) . the second term in eq .
( [ cpint ] ) is sensitive to cp violation in the neutralino decay process ( [ process2 ] ) .
the expressions @xmath173 , eq .
( [ eq_dssub6 ] ) , in @xmath174 contain three linearly independent vectors , however , @xmath175 ( eq . ( [ eq_sumdecay_te ] ) ) vanishes because @xmath154 is perpendicular to the scattering plane .
therefore , only the terms @xmath176 for @xmath177 have to be taken into account .
the cp - sensitive terms of the decay density matrix following from eqs .
( [ eq_dssum ] ) ( [ eq_dssub3 ] ) for @xmath177 , are @xmath178 with @xmath179 , \label{eq_dszz_to}\\ \sigma_d^{a , \mathrm{o}}(z \tilde{\ell}_l ) & = & \frac{4 g^4}{\cos^2\theta_w}l_{\ell } re\big\{\delta^{s_i}(z ) \big[-f^l_{\ell i } f^{l*}_{\ell k } o^{''l*}_{ki } \delta^{t_i*}(\tilde{\ell}_l)\nonumber\\ & & \phantom{\frac{4 g^4}{\cos^2\theta_w}l_{\ell}re\big\{\delta^{s_i}(z)\big [ } + f^{l*}_{\ell i } f^{l}_{\ell k } o^{''l}_{ki } \delta^{u_i*}(\tilde{\ell}_l ) \big]g_4^a\big\},\label{eq_dszel_to}\\ \sigma_d^{a , \mathrm{o}}(\tilde{\ell}_l \tilde{\ell}_l)&= & 2 g^4 re\big\ { ( f^{l*}_{\ell i})^2 ( f^l_{\ell k})^2 \delta^{t_i}(\tilde{\ell}_l ) \delta^{u_i*}(\tilde{\ell}_l ) g_4^a\big\},\label{eq_dselel_to}\\ \sigma_d^{a , \mathrm{o}}(z \tilde{\ell}_r ) & = & \frac{4 g^4}{\cos^2\theta_w } r_{\ell } re\big\{\delta^{s_i}(z ) \big[-f^r_{\ell
i } f^{r*}_{\ell k } o^{''r*}_{ki } \delta^{t_i*}(\tilde{\ell}_r)\nonumber\\ & & \phantom{\frac{4 g^4}{\cos^2\theta_w}r_{\ell}re\big\{\delta^{s_i}(z)\big [ } + f^{r*}_{\ell i } f^{r}_{\ell k } o^{''l}_{ki } \delta^{u_i*}(\tilde{\ell}_r ) \big]g_4^a\big\},\label{eq_dszel_tor}\\ \sigma_d^{a , \mathrm{o}}(\tilde{\ell}_r \tilde{\ell}_r ) & = & 2 g^4 re\big\ { ( f^{r*}_{\ell i})^2 ( f^r_{\ell k})^2 \delta^{t_i}(\tilde{\ell}_r ) \delta^{u_i*}(\tilde{\ell}_r ) g_4^a\big\},\label{eq_dselel_tor}\end{aligned}\ ] ] where for @xmath177 we have @xmath180 now @xmath181 , @xmath150 , is purely imaginary . when inserted , for example , in eq .
( [ eq_dszel_to ] ) it is multiplied by the factor @xmath182 , which depends on the phases @xmath21 and @xmath20 and contributes to @xmath183 .
analogous contributions follow from eqs .
( [ eq_dszz_to]),([eq_dselel_to ] ) ( [ eq_dselel_tor ] ) .
the corresponding t - even terms of the production density matrix also entering in eq .
( [ cpint ] ) are obtained from eqs .
( [ eq_spzz ] ) ( [ eq_spsubs ] ) , @xmath184 where @xmath185 f_3^a \big ] , \label{eq_spzz_te}\\ \sigma_p^{a , \mathrm{e}}(z\tilde{e}_l ) & = & \frac{g^4}{2 \cos^2\theta_w } l_{\ell } ( 1-p_{e^-})(1+p_{e^+ } ) \nonumber\\ & & \times re\big\{\delta^s(z ) \big[2 f^{l}_{\ell
i}f^{l*}_{\ell j } o^{''l*}_{ij } \delta^{u*}(\tilde{e}_l)f_1^a -2 f^{l*}_{\ell i}f^{l}_{\ell j } o^{''l}_{ij } \delta^{t*}(\tilde{e}_l)f_2^a \nonumber\\ & & \phantom{\times re\big\{\delta^s(z ) } + \big[f^{l}_{\ell i}f^{l*}_{\ell j } o^{''l}_{ij } \delta^{u*}(\tilde{e}_l ) + f^{l*}_{\ell i}f^{l}_{\ell j } o^{''l*}_{ij } \delta^{t*}(\tilde{e}_l)\big ] f_3^a\big]\big\ } , \label{eq_spzel_te}\\ \sigma_p^{a , \mathrm{e}}(\tilde{e}_l \tilde{e}_l ) & = & \frac{g^4}{4 } ( 1-p_{e^-})(1+p_{e^+ } ) \big [ |f^l_{\ell i}|^2 |f^l_{\ell j}|^2 \big ( |\delta^u ( \tilde{e}_l)|^2 f_1^a-|\delta^t ( \tilde{e}_l)|^2 f_2^a \big ) \nonumber\\ & & \phantom{\frac{g^4}{4 } ( 1-p_{e^-})(1+p_{e^+ } ) } + re\ { ( f^{l*}_{\ell i})^2 ( f^{l}_{\ell j})^2 \delta^{u}(\tilde{e}_l)\delta^{t*}(\tilde{e}_l ) f_3^a\}\big ] , \label{eq_spelel_te}\\ \sigma_p^{a , \mathrm{e}}(z\tilde{e}_r ) & = & \frac{g^4}{2 \cos^2\theta_w } r_{\ell } ( 1+p_{e^-})(1-p_{e^+ } ) \nonumber\\ & & \times re\big\{\delta^s(z ) \big[-2 f^{r}_{\ell
i}f^{r*}_{\ell j } o^{''r*}_{ij } \delta^{u*}(\tilde{e}_r)f_1^a + 2 f^{r*}_{\ell i}f^{r}_{\ell j } o^{''r}_{ij } \delta^{t*}(\tilde{e}_r)f_2^a \nonumber\\ & & \phantom{re\big\{\delta^s(z ) } - \big[f^{r}_{\ell i}f^{r*}_{\ell j } o^{''r}_{ij } \delta^{u*}(\tilde{e}_r ) + f^{r*}_{\ell i}f^{r}_{\ell j } o^{''r*}_{ij } \delta^{t*}(\tilde{e}_r)\big ] f_3^a\big]\big\ } , \label{eq_spzel_ter}\\ \sigma_p^{a , \mathrm{e}}(\tilde{e}_r \tilde{e}_r ) & = & \frac{g^4}{4 } ( 1+p_{e^-})(1-p_{e^+ } ) \big [ |f^r_{\ell i}|^2 |f^r_{\ell j}|^2 \big(-|\delta^u ( \tilde{e}_r)|^2 f_1^a+|\delta^t ( \tilde{e}_r)|^2 f_2^a \big ) \nonumber\\ & & \phantom{\frac{g^4}{4 } ( 1+p_{e^-})(1-p_{e^+ } ) } - re\ { ( f^{r*}_{\ell i})^2 ( f^{r}_{\ell j})^2 \delta^{u}(\tilde{e}_r)\delta^{t*}(\tilde{e}_r ) f_3^a\}\big ] , \label{eq_spelel_ter}\end{aligned}\ ] ] where @xmath186 .
\label{eq_kinsp3-te}\end{aligned}\ ] ] the triple products sensitive to the cp phases in the decay read in the laboratory system : @xmath187 & & \phantom{i m_i m_k ( p_2 p_4)\big\ { } + e_6 \vec{p}_5 ( \vec{p}_7\times \vec{p}_1 ) + e_5 \vec{p}_1 ( \vec{p}_7\times \vec{p}_6 ) \big\ } , \label{eq_te1}\\[2 mm ] \sum_{a=1,3 } f^a_2 \cdot g^a_4 & = & i m_i m_k ( p_1 p_4)\big\{-e_b \vec{p}_5 ( \vec{p}_7\times \vec{p}_6 ) + e_7 \vec{p}_5 ( \vec{p}_6\times \vec{p}_1 ) \nonumber\\[-3 mm ] & & \phantom{i m_i m_k ( p_2 p_4)\big\ { } - e_6 \vec{p}_5 ( \vec{p}_7\times \vec{p}_1 ) - e_5 \vec{p}_1 ( \vec{p}_7\times \vec{p}_6 ) \big\ } , \label{eq_te2}\\[2 mm ] \sum_{a=1,3 } f^a_3 \cdot g^a_4 & = & i m_j m_k \big\{[(p_2 p_3)-(p_1 p_3 ) ] e_b \vec{p}_5 ( \vec{p}_7\times \vec{p}_6)\nonumber\\[-3 mm ] & & \phantom{i m_j m_k \big\ { } + [ ( p_2 p_3)+(p_1 p_3)]\nonumber\\ & & \phantom{i m_j m_k \big\ { + \ ; } \big[e_7 \vec{p}_5 ( \vec{p}_6\times \vec{p}_1 ) -e_6 \vec{p}_5 ( \vec{p}_7\times \vec{p}_1 ) -e_5 \vec{p}_1 ( \vec{p}_7\times
\vec{p}_6 ) \big ] \big\ } .
\label{eq_te3}\end{aligned}\ ] ] as already mentioned , these quantities will be multiplied in eqs .
( [ eq_dszz_to ] ) ( [ eq_dselel_tor ] ) by the factor @xmath182 etc . and
contribute to the second term of eq .
( [ cpint ] ) .
we take into account only the contributions to the t - odd asymmetry @xmath38 which stem from the complex couplings and neglect the contributions from the widths of @xmath188 , @xmath189 and @xmath190 in the propagators eqs .
( [ eq_zprop ] ) , ( [ eq_selprop ] ) , because they are of higher order .
in this section we analyse numerically the cp asymmetry @xmath38 , eq .
( [ asy ] ) , and the cross section for reactions ( [ process1 ] ) and ( [ process2 ] ) , at an @xmath6 linear collider with @xmath7 gev and longitudinally polarised @xmath191 beams .
in particular , we investigate the dependence on the phases of the complex parameters @xmath192 and @xmath193 . for our numerical analysis
we choose two scenarios a and b , such that @xmath194 and @xmath195 to prevent two - body decays of @xmath18 .
the susy parameters and the masses of @xmath14 , @xmath196 , and @xmath189 , @xmath190 are given in table 1 .
the decay widths of the neutralinos have been computed with help of the program spheno @xcite .
.[scentab ] input parameters @xmath23 , @xmath13 , @xmath24 , @xmath19 , @xmath197 and @xmath198 and the resulting masses @xmath199 , @xmath196 , for @xmath200 and @xmath201 . the parameters @xmath13 , @xmath24 and @xmath19 in scenario b are chosen according to the scenario sps1a in @xcite .
all masses are given in gev . [ cols="^,^,^,^,^",options="header " , ] + in scenario a the neutralinos @xmath16 , @xmath18 , @xmath202 are mainly @xmath203-higgsino , higgsino-@xmath203-@xmath204 , higgsino-@xmath203 mixtures , respectively , where the @xmath203 component of the @xmath18 has a relatively strong @xmath20 dependence . in fig . [
fig : gamchi2 ] we show the total decay width @xmath205 and the branching ratio @xmath206 , @xmath207 or @xmath1 in scenario a as a function of @xmath20 for @xmath208 .
as can be seen , the cp - even quantities @xmath205 and @xmath206 depend quite strongly on @xmath20 .
@xmath205 varies between 34 kev for @xmath209 , @xmath210 and 4.2 kev for @xmath211 , and @xmath206 between 6.4% for @xmath212 and 2.6% for @xmath211 .
the parameters @xmath13 , @xmath24 and @xmath19 in scenario b are chosen like in the scenario sps1a in @xcite , with @xmath16 , @xmath18 , @xmath202 being mainly @xmath203- , @xmath204- , higgsino - like , respectively . in this scenario @xmath205
varies between 80 kev for @xmath213 , @xmath214 and 9.9 kev for @xmath211 , and @xmath215 between 1.9% for @xmath216 , @xmath217 and 2.9% for @xmath211 . in both scenarios we take for the slepton masses @xmath218 gev , @xmath219 gev and for the squark masses @xmath220 gev , @xmath221 gev and @xmath222 gev , @xmath223 gev for the first and second generation , and @xmath224 gev , @xmath225 gev ( @xmath226 gev ) for the bottom squarks . in most of the examples we take values for the phases @xmath21 and @xmath20 which are in agreement with the constraints from the electron and neutron edms . in order to show the full phase dependence of the asymmetry and the cross section
, we also give some examples where we vary @xmath21 and @xmath20 in the whole range , relaxing the constraints from the edms .
( 16,5.9 ) ( 0.1,5.5)@xmath227kev ( 8.4,5.5)@xmath206 in % ( 14.85,0.1)@xmath228 ( 6.7,0.1)@xmath228 ( 0.8,4.8)(a ) ( 9.3,4.8)(b ) ( 0,0.4 ) ( 8.35,0.4 ) first we discuss neutralino production @xmath230 with subsequent leptonic three - body decay @xmath231 . in this case
one obtains cp - violating contributions from both production and decay . in fig .
[ fig : at12 ] ( a ) we show the cp asymmetry @xmath38 as a function of the phase @xmath20 for scenario a defined in table [ scentab ] for @xmath232 and two centre of mass energies @xmath233 gev and @xmath7 gev .
the beam polarisations are fixed at @xmath234 and @xmath235 .
the cp asymmetry attains the largest values of about 11% ( @xmath236 ) for @xmath7 gev at @xmath237 ( @xmath238 ) . for @xmath233 gev , i.e. closer to threshold of the production , @xmath38 reaches values of @xmath239 at @xmath240 .
choosing @xmath241 instead of @xmath242 leaves the asymmetry @xmath38 almost unchanged .
note that @xmath38 does not have its largest values at @xmath243 ( @xmath244 ) , where the imaginary parts of the respective couplings are maximal , but at smaller ( larger ) phases .
the reason is an interplay between the @xmath20 dependence of the cross section , shown in fig .
[ fig : at12 ] ( b ) , and the @xmath20 dependence of the numerator of @xmath38 , eq .
( [ cpint ] ) .
this numerator is a sum of products of cp - odd and cp - even factors , which essentially have a @xmath245 and @xmath246 behaviour , respectively ( for @xmath208 ) .
we plot in fig .
[ fig : at12 ] ( b ) the cross section @xmath247 for the production and subsequent decay process @xmath248 ( eq . ( [ crosss ] ) ) , summed over @xmath5 .
note that also the cross section has a rather strong @xmath20 dependence . at @xmath211
the production cross section @xmath249 has a maximum , whereas the branching ratio @xmath250 has a minimum ( see fig .
[ fig : gamchi2 ] ( b ) ) resulting in the dip of the cross section for the combined process of production and decay at @xmath211 .
( 16,5.8 ) ( 0.3,5.4)@xmath38 in % ( 8.5,5.4)@xmath251fb ( 14.85,0.1)@xmath228 ( 6.7,0.1)@xmath228 ( 3.5,4.7)(a ) ( 9.3,4.7)(b ) ( 0,0.4 ) ( 8.35,0.4 ) in fig .
[ fig : at12phim1phimu ] we show the contour lines of the cp asymmetry @xmath38 in the @xmath20-@xmath21 plane in scenario a ( table [ scentab ] ) for @xmath42 gev and two sets of beam polarisations @xmath234 , @xmath235 and @xmath252 , @xmath253 . in the scenario
considered the @xmath20 dependence is stronger than that on @xmath21 . in the case of @xmath234 ,
@xmath235 the largest asymmetries @xmath254 are reached for @xmath255 , rather independent of @xmath21 ( fig .
[ fig : at12phim1phimu ] ( a ) ) . for @xmath252 , @xmath253 ( fig .
[ fig : at12phim1phimu ] ( b ) ) the maximal values of @xmath256 are attained near @xmath257 and @xmath258 .
the sign flip of @xmath38 between fig .
[ fig : at12phim1phimu ] ( a ) and ( b ) and the weak @xmath21 dependence follows from the structure of the @xmath259 couplings and the facts that @xmath260 and that the magnitude of the @xmath203 component of @xmath18 depends rather strongly on @xmath20 . for right - handed ( left - handed )
electrons the @xmath203 component ( @xmath203 and @xmath204 components ) of @xmath16 and @xmath18 contribute , which leads to a more pronounced @xmath20 dependence than that on @xmath21 .
( 16,8.6 ) ( 0.2,8.3)@xmath228 ( 8.4,8.3)@xmath228 ( 6.9,0.1)@xmath261 ( 15.1,0.1)@xmath261 ( 1,6.2)(a ) ( 9.2,6.9)(b ) ( 4.1,7.75)0 ( 4.15,7.41)@xmath262 ( 4.05,7.1)@xmath263 ( 4.15,5.95)@xmath262 ( 4.3,4.3)0 ( 4.3,2.85)5 ( 4.2,1.7)10 ( 4.3,1.33)5 ( 4.3,1.1)0 ( 0,0.4 ) ( 12.3,7.75)0 ( 12.5,7.3)5 ( 13.4,6.95)7.5 ( 12.5,6.45)5 ( 12.5,4.3)0 ( 12.3,2.35)@xmath262 ( 11,1.85)@xmath264 ( 12.3,1.55)@xmath262 ( 12.5,1.1)0 ( 8.2,0.4 ) in fig .
[ fig : at12m2mu ] ( a ) ( d ) we show the contour lines of the cp asymmetry @xmath38 and of the cross section for @xmath229 , @xmath265 , summed over @xmath5 , in the @xmath13-@xmath24 plane for @xmath266 , @xmath267 and @xmath242 , @xmath268 , @xmath269 gev , @xmath270 gev , at @xmath42 gev and two sets of beam polarisations , @xmath234 , @xmath235 and @xmath252 , @xmath253 . for both polarisation configurations
@xmath38 is largest in the region @xmath271 gev and @xmath272 gev with maximal values @xmath273 for @xmath234 , @xmath235 ( fig .
[ fig : at12m2mu ] ( a ) ) and @xmath274 for @xmath252 , @xmath253 ( fig . [ fig : at12m2mu ] ( b ) ) . for
@xmath275 ( @xmath276 ) the main contributions to the asymmetry @xmath38 come from the @xmath188-@xmath189 ( @xmath188-@xmath190 ) interference terms , therefore large asymmetries are attained if both neutralinos have significant higgsino and gaugino components . in fig .
[ fig : at12m2mu ] ( c ) and ( d ) we show the cross section @xmath277 , summed over @xmath5 . for @xmath234 , @xmath235 ( fig .
[ fig : at12m2mu ] ( c ) ) the cross section is larger than about 10 fb in the region with maximal @xmath38 ( @xmath271 gev , @xmath272 gev ) . for @xmath278 gev and @xmath279 gev the cross section reaches values larger than 40 fb , the asymmetry , however , is rather small , @xmath280 . for @xmath252 , @xmath253 ( fig . [
fig : at12m2mu ] ( d ) ) the cross section has values between 1 and 10 fb in the region where @xmath38 is maximal ( @xmath271 gev , @xmath272 gev ) . ( 16,17 ) ( 2.6,7.3)(c ) ( 10.7,7.3)(d ) ( 0.1,8.1)@xmath281 ( 8.3,8.1)@xmath281 ( 6.2,0.1)@xmath282 ( 14.4,0.1)@xmath282 ( 3.9,8.05)@xmath251fb ( 12,8.05)@xmath251fb ( 7.16,2.7)40 ( 6.4,2.7)30 ( 5.85,2.7)20 ( 5.26,2.7)10 ( 5.13,3.3)5 ( 3.9,3.7)2.5 ( 3.7,4.3)5 ( 3.25,5.15)10 ( 2.85,6.55)20 ( 3.65,5)a ( 5.37,3.6)b ( 0,0.4 ) ( 10.75,6.7)15 ( 11,4)10 ( 11.28,3.6)5 ( 11.7,3.2)1 ( 13.9,2.8)1 ( 11.75,5.13)a ( 13.85,3.55)b ( 8.2,0.4 ) ( 0.1,16.6)@xmath281 ( 8.3,16.6)@xmath281 ( 6.2,8.6)@xmath282 ( 14.4,8.6)@xmath282 ( 3.6,16.55)@xmath38 in % ( 11.8,16.55)@xmath38 in % ( 2.6,15.8)(a ) ( 10.8,15.8)(b ) ( 3.75,14.25)7.5 ( 3.2,12.6)5 ( 2.9,12)2.5(4.2,11.6)2.5 ( 3.7,11.8)1(5.3,12)1(2.61,13)1 ( 5.6,11.5)@xmath283 ( 6.6,11.2)@xmath284 ( 3.65,13.5)a ( 5.65,12.05)b ( 0,8.9 ) ( 11.9,14.1)@xmath263 ( 11.9,13.6)@xmath264 ( 11.7,13.2)@xmath262 ( 11.2,12.6)@xmath284 ( 11.1,12.1)@xmath283 ( 12.35,11.9)1 ( 11.7,13.63)a ( 13.85,12.05)b ( 8.2,8.9 ) ( 4.5,6.5)2-body decays ( 12.7,6.5)2-body decays ( 4.5,15)2-body decays ( 12.7,15)2-body decays in principle , also for the case of hadronic neutralino decays into heavy quarks , @xmath285 and @xmath286 , it will be possible to measure the asymmetry @xmath38 .
we have calculated @xmath38 in the @xmath13-@xmath24 plane and found values of roughly the same order of magnitude as those shown in fig . [
fig : at12m2mu ] for the leptonic decay .
the cross sections are larger by roughly a factor 5 10 in the main part of the parameter region of fig .
[ fig : at12m2mu ] and approximately of the same order of magnitude as in the leptonic case for @xmath278 gev . the experimental errors
, however , are expected to be larger than for the leptonic @xmath18 decays , because the distinction of the @xmath287 and @xmath288 or @xmath289 and @xmath290 charges and the resolution of the jet momenta are worse than in the leptonic case @xcite .
we also studied the @xmath23 dependence of the asymmetry @xmath38 and the cross section , relaxing the gut relation for @xmath23 with all other parameters as in scenario b of table [ scentab ] . in fig .
[ fig : atsps1a12 ] ( a ) we show the t - odd asymmetry @xmath38 as a function of @xmath20 for four values of @xmath291 gev , 95 gev , 100 gev and 105 gev .
here maximal asymmetries @xmath292 are reached . for @xmath293
gev @xmath38 shows nearly two complete oscillations , whereas for @xmath294 gev there is only one oscillation with extrema at @xmath295 and @xmath296 . in both cases
this behaviour is caused by cancellations between the contributions @xmath297 from production and @xmath298 from decay in eq .
( [ cpint ] ) , which have opposite sign . for @xmath294
gev the contributions from the decay are dominant , whereas for @xmath293 gev both contributions are of similar magnitude , resulting in two oscillations .
the corresponding cross sections are shown in fig .
[ fig : atsps1a12 ] ( c ) . in fig .
[ fig : atsps1a12 ] ( b ) we show @xmath299 as a function of @xmath21 for @xmath300 and three values of @xmath301 gev , 100 gev and 105 gev .
( we do not discuss @xmath294 gev because in this case @xmath302 for @xmath303 . ) for @xmath304 gev @xmath38 exhibits the strongest variation with @xmath21 with maximum @xmath305 for @xmath306 and minimum @xmath307 for @xmath308 .
the corresponding cross sections are shown in fig .
[ fig : atsps1a12 ] ( d ) .
( 16,11.5 ) ( 0.2,11.2)@xmath38 in % ( 8.4,11.2)@xmath38 in % ( 15.1,5.8)@xmath261 ( 6.7,5.8)@xmath228 ( 0.9,10.45)(a ) ( 9.1,10.45)(b ) ( 0,6.1 ) ( 8.15,6.1 ) ( 0.2,5.45)@xmath251fb ( 8.4,5.45)@xmath251fb ( 15.1,0.1)@xmath261 ( 6.7,0.1)@xmath228 ( 0.9,4.8)(c ) ( 9.1,4.8)(d ) ( 0.05,0.4 ) ( 8.2,0.4 ) furthermore we have analysed the dependence of the asymmetry @xmath38 on the masses of the right and left sleptons , assuming @xmath198 and @xmath197 are free parameters , for @xmath267 and @xmath242 with the other parameters as in scenario a of table [ scentab ] . for @xmath42 gev and beam polarisations @xmath234 and @xmath235 we find that the asymmetry @xmath38 is larger than 6% for @xmath309 gev and @xmath310 gev . for @xmath311 gev and @xmath312 gev
the asymmetry is smaller than @xmath313 .
this is due to the fact , that for these beam polarisations the contributions from @xmath189 exchange to the cp asymmetry are dominant . outside of these narrow regions around @xmath309
gev or @xmath311 gev the asymmetry is very small . for
opposite beam polarisations @xmath252 , @xmath253 the @xmath190 contributions to @xmath38 are dominant and the sign of the asymmetry changes : for @xmath314 gev and @xmath315 gev it is @xmath316 and for @xmath311 gev and @xmath317 gev it is @xmath318 . in the reaction @xmath319
there are only t - odd contributions from the decay , i.e.from the second term in eq .
( [ cpint ] ) .
there is no t - odd contribution from the production , because the amplitudes eqs .
( [ eq_a1 ] ) ( [ eq_a3 ] ) are real and the first term in eq .
( [ cpint ] ) vanishes
. we will give numerical results for the case where one neutralino decays leptonically and the other decays hadronically .
if both neutralinos would decay into lepton pairs of the same flavour , it may be experimentally difficult to distinguish the two lepton pairs coming from different neutralinos . in fig .
[ fig : at22m2mu350 ] we show the contour lines in the @xmath13-@xmath24 plane of the cp asymmetry @xmath38 and of the cross section with one neutralino decaying leptonically , @xmath320 , summed over @xmath5 , and the other neutralino decaying hadronically , @xmath321 , summed over @xmath322 , for @xmath323 gev and @xmath234 , @xmath235 . the other parameters are @xmath268 , @xmath269 gev , @xmath270 gev , @xmath266 , @xmath267 and @xmath242 . for this process
@xmath38 is considerably smaller than for @xmath324 , because for production of a pair of equal neutralinos only the decay amplitudes contribute to the asymmetry .
the largest asymmetries are obtained for @xmath278 gev and @xmath325 gev , where @xmath16 and @xmath18 have gaugino character .
the maximum values are @xmath326 .
the cross section , however , reaches values up to 10 fb in this region .
the hadronic branching ratio @xmath327 is between 50% and 95% in the main part of the parameter region of fig .
[ fig : at22m2mu350 ] and between 30% and 50% for @xmath278 gev . for @xmath42 gev the asymmetries are smaller with maximum of 0.7% for @xmath278 gev and @xmath325 gev .
we want to remark that it may also be possible to distinguish the two lepton pairs , if one neutralino decays to an @xmath6 pair , the other one to a @xmath328 pair . similarly it may be possible to distinguish the two quark pairs , if one neutralino decays to a @xmath329 pair and the other one to a @xmath330 pair @xcite or pair of light quarks .
however , in the case of both neutralinos decaying into lepton pairs the event rate is expected to be very low , whereas in the case of the decay into quark pairs of different flavours the experimental errors are expected to be much larger .
( 16,8.5 ) ( 2.6,7.3)(a ) ( 10.8,7.3)(b ) ( 0.1,8.1)@xmath281 ( 8.3,8.1)@xmath281 ( 6.2,0.1)@xmath282 ( 14.4,0.1)@xmath282 ( 3.6,8.05)@xmath38 in % ( 2.65,5.3)@xmath331 ( 3.78,3.45)@xmath332(3.03,4.45)@xmath332 ( 3.35,3.9)@xmath333 ( 5,2.8)@xmath333 ( 4.58,2.8)@xmath332 ( 4.02,3)@xmath331 ( 6.48,3.07)@xmath334 ( 0,0.4 ) ( 12,8.05)@xmath335 in fb ( 12.3,3.2)0.5 ( 11.85,3.1)1 ( 12.65,2.9)1 ( 13.17,3.05)2.5 ( 13.3,2.77)5 ( 13.7,2.7)10 ( 14.4,2.6)15 ( 8.2,0.4 ) ( 4.5,6.5)@xmath336 ( 12.7,6.5)@xmath336 in fig .
[ fig : at32phim1phimu ] we show the contour lines of the cp asymmetry @xmath38 in the @xmath20-@xmath21 plane in scenario a ( table [ scentab ] ) for @xmath42 gev and two sets of beam polarisations @xmath234 , @xmath235 and @xmath252 , @xmath253 . as in the case of @xmath338 production ( fig .
[ fig : at12phim1phimu ] ) the @xmath20 dependence is stronger than that on @xmath21 in this scenario , since the @xmath203 component of the decaying particle @xmath18 depends stronger on @xmath20 .
the largest asymmetries of @xmath339 are reached in the case of @xmath234 , @xmath235 ( fig .
[ fig : at32phim1phimu ] ( a ) ) for @xmath340 and @xmath341 . in the case of @xmath252 , @xmath253 ( fig .
[ fig : at32phim1phimu ] ( b ) ) the sign of @xmath38 flips and the largest asymmetries @xmath342 appear for @xmath340 and @xmath343 .
( 16,8.6 ) ( 0.2,8.3)@xmath228 ( 8.4,8.3)@xmath228 ( 6.9,0.1)@xmath261 ( 15.1,0.1)@xmath261 ( 0.9,7)(a ) ( 9.1,7)(b ) ( 4.05,7.75)0 ( 4.15,7.44)@xmath262 ( 4.05,7.1)@xmath344 ( 4.15,5.75)@xmath262 ( 4.3,4.3)0 ( 4.3,3.05)5 ( 4.3,1.7)9 ( 4.3,1.38)5 ( 4.3,1.1)0 ( 0,0.4 ) ( 12.3,7.75)0 ( 14.5,7.2)3 ( 13.75,6.95)5 ( 12.95,6.7)7 ( 12.5,5.37)3 ( 12.5,4.3)0 ( 12.3,3.5)@xmath345 ( 11.7,2.15)@xmath346 ( 10.95,1.85)@xmath262 ( 10.2,1.65)@xmath345 ( 12.5,1.1)0 ( 8.2,0.4 ) in fig .
[ fig : at32m2mu ] we show the contour lines of the cp asymmetry @xmath38 and of the cross section for the process @xmath347 with the subsequent decay @xmath348 , summed over @xmath349 , in the @xmath13-@xmath24 plane for @xmath268 , @xmath269 gev , @xmath270 gev , @xmath266 , @xmath267 and @xmath242 for @xmath42 gev and two sets of beam polarisations @xmath234 , @xmath235 and @xmath252 , @xmath253 .
@xmath350 has two maxima in the regions @xmath271 gev , @xmath351 gev and @xmath352 gev , @xmath353 gev , respectively , for both polarisation configurations .
the maximum values are @xmath354 for @xmath234 , @xmath235 ( fig .
[ fig : at32m2mu ] ( a ) ) and @xmath355 for @xmath234 , @xmath235 ( fig . [ fig : at32m2mu ] ( b ) ) .
the cross section reaches values larger than 10 fb ( 6 fb ) in the region with maximal asymmetry around @xmath271 gev , @xmath351 gev for @xmath234 , @xmath235 , fig .
[ fig : at32m2mu ] ( c ) ( @xmath252 , @xmath253 , fig .
[ fig : at32m2mu ] ( d ) ) . in the region around @xmath352
gev , @xmath353 gev the cross section is @xmath356 fb .
( 16,17 ) ( 2.83,7.3)(c ) ( 10.8,7.3)(d ) ( 0.1,8.1)@xmath281 ( 8.3,8.1)@xmath281 ( 6.2,0.1)@xmath282 ( 14.4,0.1)@xmath282 ( 3.9,8.05)@xmath251fb ( 12,8.05)@xmath251fb ( 4.9,2.9)1 ( 2.6,6.85)1(4.5,3.8)1 ( 3.4,3.1)2(2.65,5.95)2 ( 3.1,3.4)5(2.68,4.8)5 ( 3,4.28)10 ( 3.75,5.55)15 ( 3.35,5)a ( 0,0.4 ) ( 11.58,3.22)1(10.8,5.85)1 ( 11.27,3.45)2(10.82,5.28)2 ( 11,3.95)4(10.87,4.72)4 ( 12.05,5.45)6 ( 11.55,5)a ( 8.2,0.4 ) ( 0.1,16.6)@xmath281 ( 8.3,16.6)@xmath281 ( 6.2,8.6)@xmath282 ( 14.4,8.6)@xmath282 ( 3.6,16.55)@xmath38 in % ( 11.8,16.55)@xmath38 in % ( 2.6,15.8)(a ) ( 10.8,15.8)(b ) ( 3.4,16.17)1(4.25,11.65)1 ( 4.63,12.3)1 ( 3.57,16.17)2.5(4,11.9)2.5 ( 4.58,11.9)2.5 ( 2.85,12.2)5(3.47,11.5)5 ( 4.55,11.3)5 ( 3.3,12.2)7.5(4.78,11.27)7.5 ( 3.95,13.15)10(5.4,11.22)10 ( 3.4,13.5)a ( 0,8.9 ) ( 11.73,15.5)@xmath283 ( 11.9,14.6)@xmath345(11.68,11.6)@xmath345 ( 11.25,11.8)@xmath357 ( 12,12.9)@xmath358 ( 12.3,11.7)@xmath283(12.65,12.15)@xmath283 ( 12.9,11.8)@xmath345 ( 13,11.25)@xmath357 ( 13.48,11.45)@xmath358 ( 11.6,13.5)a ( 8.2,8.9 ) ( 4.3,6.7)@xmath359 ( 4.3,6)or 2-body decays ( 12.5,6.7)@xmath359 ( 12.5,6)or 2-body decays ( 4.3,14.5)or 2-body decays ( 4.3,15.2)@xmath359 ( 12.5,14.5)or 2-body decays ( 12.5,15.2)@xmath359 we have also calculated the branching ratio @xmath360 , @xmath207 or @xmath1 , of @xmath202 . in parameter regions , where @xmath202 decays via a three - body decay and where it is especially difficult to distinguish the two lepton pairs coming from different neutralinos , @xmath360 is always smaller than about 7% . furthermore we have calculated the cp asymmetry @xmath38 for the the process @xmath361 with the subsequent decay @xmath4 , @xmath349 . for the associated production of @xmath18 and @xmath362 the accessible parameter space at a linear collider with @xmath42
gev is rather small and is approximately constrained by @xmath363 gev . for the same parameters as in fig .
[ fig : at32m2mu ] we find that the cp asymmetry has maximum values @xmath364 .
however , in parameter regions with @xmath365 the cross section @xmath366 , summed over @xmath54 , is always smaller than 1 fb .
in this paper we have studied cp violation in neutralino production and subsequent three - body decay processes . as cp - sensitive
observable we have chosen a t - odd asymmetry based on triple product correlations between the incoming and outgoing particles .
a non - vanishing value for the asymmetry indicates directly cp - violating effects .
we have provided compact analytical formulae for the cross section of the whole process as well as for the t - odd asymmetry for longitudinal beam polarisations .
it is necessary to include the full spin correlations between production and decay .
the asymmetry can be directly measured in the experiment without reconstruction of the momentum of the decaying neutralino or further final - state analyses . in our numerical results
we have chosen representative scenarios and have shown that this t - odd asymmetry can reach values up to 13% in some parts of the parameter space of the unconstrained mssm , even for small cp - violating phases as may be indicated by the edm constraints .
this observable will therefore be an important tool in the search for cp - violating effects in the neutralino sector and the determination of the phases of the complex susy parameters .
we thank w. porod for providing us a version of spheno for complex susy parameters , prior to its release , which has been used to compute the neutralino decay widths .
we thank t. kernreiter , o. kittel and w. majerotto for useful discussions .
this work is supported by the ` fonds zur frderung der wissenschaftlichen forschung ' of austria , fwf project no .
p16592-n02 , by the european community s human potential programme under contract hprn - ct-2000 - 00149 and by the deutsche forschungsgemeinschaft ( dfg ) under contract no ..
we choose a coordinate frame in the laboratory system , where the momenta are @xmath367 @xmath368 is the scattering angle between the incoming @xmath56 beam and the outgoing neutralino @xmath57 , the azimuth @xmath369 can be chosen equal to zero .
the energy and the momenta of the neutralinos @xmath57 and @xmath59 are @xmath370 where @xmath371 are the masses of the neutralinos and @xmath372 is the kinematical triangle function @xmath373 .
the polarisation vectors @xmath374 and @xmath375 ( @xmath376 ) of the neutralinos in the laboratory system are @xmath377 where @xmath378 describes the longitudinal polarisation , @xmath379 the transverse polarisation in the scattering plane , and @xmath380 the transverse polarisation perpendicular to the scattering plane .
here we give the analytic expressions for the production spin density matrix .
the terms @xmath381 are independent of the polarisation of the neutralinos @xcite , @xmath382f_3 \right\ } , \label{eq_pzz}\\ p(z\tilde{e}_l)&= & \frac{g^4}{2\cos^2\theta_w}l_{\ell } c_l \nonumber\\ & & \times re\big\{\delta^s(z)\big[-(\delta^{t*}(\tilde{e}_l)f^{l*}_{\ell i}f^l_{\ell j } o^{''l*}_{ij}+ \delta^{u*}(\tilde{e}_l)f^{l}_{\ell i}f^{l*}_{\ell j } o^{''l}_{ij } ) f_3\nonumber\\ & & \phantom{\times re\big\{}+ 2 \delta^{t*}(\tilde{e}_l ) f^{l*}_{\ell i } f^l_{\ell j } o^{''l}_{ij } f_1 + 2 \delta^{u*}(\tilde{e}_l ) f^{l}_{\ell i } f^{l*}_{\ell j } o^{''l*}_{ij } f_2\big]\big\ } , \label{eq_pzel}\\ p(\tilde{e}_l\tilde{e}_l)&=&\frac{g^4}{4 } c_l \big\{|f^l_{\ell i}|^2 |f^l_{\ell j}|^2 ( |\delta^t(\tilde{e}_l)|^2 f_1+|\delta^u(\tilde{e}_l)|^2 f_2 ) \nonumber\\ & & \phantom{\frac{g^4}{4 } c_l \big\ { } -re\ { ( f^{l*}_{\ell i})^2 ( f^{l}_{\ell j})^2 \delta^u(\tilde{e}_l ) \delta^{t*}(\tilde{e}_l ) f_3\ } \big\}. \label{eq_pelel}\end{aligned}\ ] ] @xmath383 , @xmath384 are obtained by the following substitutions in eqs .
( [ eq_pzel ] ) , ( [ eq_pelel ] ) : @xmath385 the beam polarisation weighting factors are given by @xmath386 and the kinematical factors in invariant form are @xmath387 the terms @xmath388 which depend on the polarisation of the neutralino @xmath14 and which are relevant for the calculation of the cp asymmetries , are given by @xmath389 f_3^a + 2 ( re o^{''l}_{ij } ) ( i m o^{''l}_{ij } ) i f_4^a\big\ } , \label{eq_spzz}\\ \sigma_p^a(z\tilde{e}_l ) & = & \frac{g^4}{2 \cos^2\theta_w } l_{\ell } c_{l } \nonumber\\ & & \times re\big\{\delta^s(z ) \big[2 f^{l}_{\ell i}f^{l*}_{\ell j } o^{''l*}_{ij } \delta^{u*}(\tilde{e}_l)f_1^a -2 f^{l*}_{\ell i}f^{l}_{\ell j } o^{''l}_{ij } \delta^{t*}(\tilde{e}_l)f_2^a \nonumber\\ & & \phantom{\times re\big\{\delta^s(z ) \big [ } + \big[f^{l}_{\ell i}f^{l*}_{\ell j } o^{''l}_{ij } \delta^{u*}(\tilde{e}_l ) + f^{l*}_{\ell i}f^{l}_{\ell j } o^{''l*}_{ij } \delta^{t*}(\tilde{e}_l)\big ] f_3^a\nonumber\\ & & \phantom{\times re\big\{\delta^s(z ) \big [ } + \big[f^{l*}_{\ell i}f^{l}_{\ell j }
o^{''l*}_{ij } \delta^{t*}(\tilde{e}_l ) -f^{l}_{\ell i}f^{l*}_{\ell j } o^{''l}_{ij } \delta^{u*}(\tilde{e}_l)\big ] f_4^a\big]\big\},\label{eq_spzel}\\ \sigma_p^a(\tilde{e}_l \tilde{e}_l ) & = & \frac{g^4}{4 } c_{l}\big [ |f^l_{\ell i}|^2 |f^l_{\ell j}|^2 \big ( |\delta^u ( \tilde{e}_l)|^2 f_1^a-|\delta^t ( \tilde{e}_l)|^2 f_2^a \big ) \nonumber\\ & & \phantom{\frac{g^4}{4 } c_{l}\big [ } + re\ { ( f^{l*}_{\ell i})^2 ( f^{l}_{\ell j})^2 \delta^{u}(\tilde{e}_l)\delta^{t*}(\tilde{e}_l ) ( f_3^a+f_4^a)\}\big ] .
\label{eq_spelel}\end{aligned}\ ] ] @xmath390 are obtained by the substitutions eq.([eq_psub ] ) and @xmath391 in eqs .
( [ eq_spzel ] ) , ( [ eq_spelel ] ) , where the kinematical factors in invariant form are given by @xmath392 , \label{eq_kinsp3}\\ f^a_4 & = & i m_j \epsilon_{\mu \nu \rho \sigma } p_2^{\mu } p_1^{\nu } s^{a \rho } p_3^{\sigma}.\label{eq_kinsp4}\end{aligned}\ ] ] the polarisation vectors @xmath121 of the neutralino @xmath14 are given in the laboratory system by eqs .
( [ eq_21 ] ) ( [ eq_23 ] ) . here
we give the analytical expressions for the different contributions to the decay density matrix for the three - body decay @xmath393 , where we sum over the spins of the final - state particles @xcite .
the contributions independent of the polarisation of the neutralino @xmath14 are @xmath394 with @xmath395 g_3 \big ] , \label{eq_dzz}\\ d(z \tilde{\ell}_l)&=&4 \frac{g^4}{\cos^2\theta_w } l_{\ell } re\big\{\delta^{s_i}(z ) \big[f^l_{\ell i } f^{l*}_{\ell k } \delta^{t_i*}(\tilde{\ell}_l ) ( 2o^{''l}_{ki } g_1 + o^{''l*}_{ki } g_3 ) \nonumber\\ & & \phantom{4 \frac{g^4}{\cos^2\theta_w } l_{\ell } re\big\{\delta^{s_i}(z ) \big [ } + f^{l*}_{\ell i } f^{l}_{\ell k } \delta^{u_i*}(\tilde{\ell}_l ) ( 2o^{''l*}_{ki } g_2 + o^{''l}_{ki } g_3 ) \big]\big\},\label{eq_dzel}\\ d(\tilde{\ell}_l \tilde{\ell}_l)&= & 2 g^4 \big [ |f^{l}_{\ell i}|^2 |f^l_{\ell k}|^2 \big(|\delta^{t_i}(\tilde{\ell}_l)|^2 g_1 + |\delta^{u_i}(\tilde{\ell}_l)|^2 g_2\big)\nonumber\\ & & \phantom{2 g^4 \big [ } + re\big\{(f^{l*}_{\ell i})^2 ( f^l_{\ell k})^2 \delta^{t_i}(\tilde{\ell}_l ) \delta^{u_i*}(\tilde{\ell}_l)\big\ } g_3 \big ] . \label{eq_delel}\end{aligned}\ ] ] the quantities @xmath396 can be derived from eqs .
( [ eq_dzel ] ) , ( [ eq_delel ] ) by the substitutions @xmath397 the kinematical factors are @xmath398 the contributions which depend on the polarisation of the decaying neutralino @xmath14 are @xmath399 with ) and ( [ eq_dszel ] ) misprints in eqs . ( 79 ) and ( 80 ) of @xcite are corrected . ]
@xmath400g^a_3 + |o^{''l}_{ki}|^2(g^a_1-g^a_2 ) \nonumber\\ & & \phantom{\times\big [ } - 2 re(o^{''l}_{ki } ) im(o^{''l}_{ki } ) i g_4^a \big],\label{eq_dszz}\\ \sigma_d^a(z \tilde{\ell}_l)&= & \frac{4 g^4}{\cos^2\theta_w}l_{\ell } re\big\{\delta^{s_i}(z ) \big [ f^l_{\ell i } f^{l*}_{\ell k } \delta^{t_i*}(\tilde{\ell}_l ) \big(-2 o^{''l}_{ki } g^a_1 + o^{''l*}_{ki } ( g^a_3-g^a_4)\big)\nonumber\\ & & \phantom{\frac{4 g^4}{\cos^2\theta_w}l_{\ell } re\big\ { } + f^{l*}_{\ell i } f^{l}_{\ell k }
\delta^{u_i*}(\tilde{\ell}_l ) \big(2 o^{''l*}_{ki } g^a_2 + o^{''l}_{ki } ( g^a_3+g^a_4)\big ) \big]\big\},\label{eq_dszel}\\ \sigma_d^a(\tilde{\ell}_l \tilde{\ell}_l ) & = & 2 g^4 \big [ |f^{l}_{\ell i}|^2 |f^l_{\ell k}|^2 [ |\delta^{u_i}(\tilde{\ell}_l)|^2 g_2^a -|\delta^{t_i}(\tilde{\ell}_l)|^2 g_1^a]\nonumber\\ & & \phantom{2 g^4 \big [ } + re\big\ { ( f^{l*}_{\ell i})^2 ( f^l_{\ell k})^2
\delta^{t_i}(\tilde{\ell}_l ) \delta^{u_i*}(\tilde{\ell}_l)(g_3^a+g_4^a)\big\}\big].\label{eq_dselel}\end{aligned}\ ] ] the contributions @xmath401 are obtained from eqs .
( [ eq_dszel ] ) , ( [ eq_dselel ] ) by applying the substitutions in eq .
( [ eq_substdecayp ] ) and in addition @xmath402 the kinematical factors are @xmath403 , \label{eq_dssub5}\\ g^a_4 & = & i m_k \epsilon_{\mu \nu \rho \sigma } s^{a \mu } p_3^{\nu } p_7^{\rho } p_6^{\sigma}. \label{eq_dssub6}\end{aligned}\ ] ] f. csikor , z. fodor and j. heitger , phys .
lett . * 82 * , 21 ( 1999 ) [ arxiv : hep - ph/9809291 ] ; + m. dine and a. kusenko , rev .
phys . * 76 * , 1 ( 2004 ) [ arxiv : hep - ph/0303065 ] .
m. dugan , b. grinstein and l. j. hall , nucl .
b * 255 * ( 1985 ) 413 ; + a. masiero and o. vives , new j. phys . * 4 * ( 2002 ) 4 .
e. accomando _ et al .
_ [ ecfa / desy lc physics working group collaboration ] , phys .
* 299 * ( 1998 ) 1 [ arxiv : hep - ph/9705442 ] ; + t. abe _ et al . _
[ american linear collider working group collaboration ] , in _ proc . of the aps / dpf / dpb summer study on the future of particle physics ( snowmass 2001 ) _ ed .
n. graf , arxiv : hep - ex/0106056 ; + j. a. aguilar - saavedra _ et al .
_ [ ecfa / desy lc physics working group collaboration ] , arxiv : hep - ph/0106315 ; + k. abe _ et al . _
[ acfa linear collider working group collaboration ] , arxiv : hep - ph/0109166 .
a. bartl , h. fraas and w. majerotto , nucl .
b * 278 * ( 1986 ) 1 .
g. j. gounaris and c. le moul , phys .
d * 66 * ( 2002 ) 055007 [ arxiv : hep - ph/0204152 ] .
j. l. kneur and g. moultaka , phys .
d * 61 * ( 2000 ) 095003 [ arxiv : hep - ph/9907360 ] .
v. barger , t. han , t. j. li and t. plehn , phys .
b * 475 * ( 2000 ) 342 [ arxiv : hep - ph/9907425 ] .
s. y. choi , a. djouadi , h. s. song and p. m. zerwas , eur .
j. c * 8 * ( 1999 ) 669 [ arxiv : hep - ph/9812236 ] ; + s. y. choi , m. guchait , j. kalinowski and p. m. zerwas , phys .
b * 479 * ( 2000 ) 235 [ arxiv : hep - ph/0001175 ] ; + s. y. choi , a. djouadi , m. guchait , j. kalinowski , h. s. song and p. m. zerwas , eur .
j. c * 14 * ( 2000 ) 535 [ arxiv : hep - ph/0002033 ] .
s. y. choi , j. kalinowski , g. moortgat - pick and p. m. zerwas , eur .
j. c * 22 * ( 2001 ) 563 [ arxiv : hep - ph/0108117 ] ; addendum , eur .
j. c * 23 * ( 2002 ) 769 [ arxiv : hep - ph/0202039 ] .
g. harris _ et al .
_ , phys .
* 82 * ( 1999 ) 904 ; + m. v. romalis , w. c. griffith and e. n. fortson , phys . rev .
* 86 * ( 2001 ) 2505 [ arxiv : hep - ex/0012001 ] ; + b. c. regan , e. d. commins , c. j. schmidt and d. demille , phys . rev .
* 88 * ( 2002 ) 071805 . t. falk and k. a. olive , phys .
b * 375 * ( 1996 ) 196 [ arxiv : hep - ph/9602299 ] ; phys .
b * 439 * ( 1998 ) 71 [ arxiv : hep - ph/9806236 ] ; + t. ibrahim and p. nath , phys .
b * 418 * ( 1998 ) 98 [ arxiv : hep - ph/9707409 ] ; phys . rev .
d * 57 * ( 1998 ) 478 [ arxiv : hep - ph/9708456 ] ; phys .
d * 58 * ( 1998 ) 111301 [ arxiv : hep - ph/9807501 ] ; phys .
d * 61 * ( 2000 ) 093004 [ arxiv : hep - ph/9910553 ] ; talk at 9th international conference on supersymmetry and unification of fundamental interactions , 1117 june 2001 , dubna , russia , arxiv : hep - ph/0107325 ; + a. bartl , t. gajdosik , w. porod , p. stockinger and h. stremnitzer , phys .
d * 60 * ( 1999 ) 073003 [ arxiv : hep - ph/9903402 ] ; + e. accomando , r. arnowitt and b. dutta , phys .
d * 61 * ( 2000 ) 115003 [ arxiv : hep - ph/9907446 ] ; + v. barger , t. falk , t. han , j. jiang , t. li and t. plehn , phys .
d * 64 * ( 2001 ) 056007 [ arxiv : hep - ph/0101106 ] ; + s. abel , s. khalil and o. lebedev , nucl .
b * 606 * ( 2001 ) 151 [ arxiv : hep - ph/0103320 ] ; + a. bartl , t. gajdosik , e. lunghi , a. masiero , w. porod , h. stremnitzer and o. vives , phys .
d * 64 * ( 2001 ) 076009 [ arxiv : hep - ph/0103324 ] .
s. y. choi , m. drees and b. gaissmaier , arxiv : hep - ph/0403054 .
a. bartl , w. majerotto , w. porod and d. wyler , phys .
d * 68 * ( 2003 ) 053005 [ arxiv : hep - ph/0306050 ] .
s. y. choi , h. s. song and w. y. song , phys .
d * 61 * ( 2000 ) 075004 [ arxiv : hep - ph/9907474 ] .
j. f. donoghue , phys .
d * 18 * ( 1978 ) 1632 ; + y. kizukuri and n. oshimo , phys .
b * 249 * ( 1990 ) 449 ; + g. valencia , arxiv : hep - ph/9411441 .
g. moortgat - pick , h. fraas , a. bartl and w. majerotto , eur .
j. c * 9 * , 521 ( 1999 ) [ erratum - ibid .
c * 9 * , 549 ( 1999 ) ] [ arxiv : hep - ph/9903220 ] .
a. bartl , h. fraas , o. kittel and w. majerotto , phys .
d * 69 * ( 2004 ) 035007 [ arxiv : hep - ph/0308141 ] ; + a. bartl , h. fraas , o. kittel and w. majerotto , arxiv : hep - ph/0402016 .
a. bartl , h. fraas , t. kernreiter and o. kittel , eur .
j. c * 33 * ( 2004 ) 433 [ arxiv : hep - ph/0306304 ] .
a. bartl , t. kernreiter and o. kittel , phys .
b * 578 * ( 2004 ) 341 [ arxiv : hep - ph/0309340 ] ; + s. y. choi , m. drees , b. gaissmaier and j. song , phys .
d * 69 * ( 2004 ) 035008 [ arxiv : hep - ph/0310284 ] .
j. a. aguilar - saavedra , arxiv : hep - ph/0404104 .
a. bartl , k. hidaka , t. kernreiter and w. porod , phys .
b * 538 * ( 2002 ) 137 [ arxiv : hep - ph/0204071 ] ; phys .
d * 66 * ( 2002 ) 115009 [ arxiv : hep - ph/0207186 ] ; + a. bartl , s. hesselbach , k. hidaka , t. kernreiter and w. porod , arxiv : hep - ph/0306281 ; phys .
b * 573 * ( 2003 ) 153 [ arxiv : hep - ph/0307317 ] ; arxiv : hep - ph/0311338 .
a. bartl , t. kernreiter and w. porod , phys .
b * 538 * ( 2002 ) 59 [ arxiv : hep - ph/0202198 ] ; + t. gajdosik , r. m. godbole and s. kraml , arxiv : hep - ph/0405167 .
h. e. haber and g. l. kane , phys
. rep .
* 117 * ( 1985 ) 75 .
h. e. haber , proceedings of the 21st slac summer institute on particle physics , eds .
l. deprocel , ch .
dunwoodie , stanford 1993 , p. 231
, arxiv : hep - ph/9405376 . w. porod , comput .
commun .
* 153 * ( 2003 ) 275 [ arxiv : hep - ph/0301101 ] . n. ghodbane and h. u. martyn , arxiv : hep - ph/0201233 ; + b. c. allanach _ et al . _ ,
j. c * 25 * ( 2002 ) 113 [ arxiv : hep - ph/0202233 ] .
c. j. damerell and d. j. jackson , _ prepared for 1996 dpf / dpb summer study on new directions for high - energy physics ( snowmass 96 ) , snowmass , colorado , 25 jun - 12 jul 1996 _ ; + k. abe _ et al . _ [ sld collaboration ] , phys .
* 88 * ( 2002 ) 151801 [ arxiv : hep - ex/0111035 ] ; + j. abdallah _ et al .
_ [ delphi collaboration ] , eur .
j. c * 32 * 185 ( 2004 ) 185 [ arxiv : hep - ex/0311003 ] ; + s. m. xella - hansen , m. wing , d. j. jackson , n. de groot , c. j. s. damerell , linear collider note lc - phsm-2003 - 061 .
|
we study cp - violating effects in neutralino production and subsequent decay within the minimal supersymmetric standard model with complex parameters @xmath0 and @xmath1 . the observable we propose is a t - odd asymmetry based on a triple product in neutralino production @xmath2 , @xmath3 , with subsequent leptonic three - body decay @xmath4 , @xmath5 , at an @xmath6 linear collider with @xmath7 gev and polarised beams .
we provide compact analytical formulae for the cross section and the t - odd asymmetry taking into account the complete spin correlations between production and decay .
we give numerical predictions for the cross section and the t - odd asymmetry .
the asymmetry can go up to 10% .
dcpt-04 - 30 + ippp-04 - 15 + uwthph-2004 - 11 + wue - itp-2004 - 012 + hep - ph/0406190 + a. bartl@xmath8 , h. fraas@xmath9 , s. hesselbach@xmath8 , k. hohenwarter - sodek@xmath8 , g. moortgat - pick@xmath10 @xmath8_institut fr theoretische physik , universitt wien , a-1090 vienna , austria _
+ @xmath9_institut fr theoretische physik und astrophysik , universitt wrzburg , _ + _ am hubland , d-97074 wrzburg , germany _ + @xmath10_ippp , university of durham , durham dh1 3le , u.k . _
|
the @xmath0 wznw model is a key example of the sigma models with supergroup targets that appear in the supersymmetric description of non interacting disordered systems in low dimensional statistical mechanics .
the first occurrence of this model probably arose via a supersymmetrization of the path integral for two copies of the two dimensional critical ising model .
it was shown in @xcite how a @xmath5 system ( with central charge @xmath6 ) could be introduced to cancel out the pair of free majorana fermions ( regrouped for convenience into a dirac fermion ) path integrals @xmath7\ , \exp[s_{0}+\delta s]\ =
\ 1 \label{ising}\ ] ] where @xmath8 and @xmath9 the theory without random mass @xmath10 is obviously a free osp(2@xmath112 ) theory , which can be considered as a @xmath0 wznw model at level @xmath4 .. in part of the literature , the level is defined as @xmath12 , so the free system in ( [ ising ] ) has @xmath2 there . ]
averaging over disorder produces a marginally irrelevant current current perturbation of this wznw model .
this is crucial to understanding the ( logarithmic ) corrections to pure ising model scaling .
the deep infrared ( ir ) behavior however is not changed by the disorder , which corresponds to the fact that ( [ isingfree ] ) is a simple free theory , with pure fermionic correlators identical to those of the usual ising model .
the second occurrence of the @xmath0 model is more involved .
it arises in the study of @xmath13 dimensional spin - full electrons in the presence of a random ( non abelian ) gauge potential .
the supersymmetrization of the path integral for two copies of the dirac fermions produces a free osp(4@xmath114 ) theory which has been argued to flow to the @xmath0 wznw model at level one under the action of the disorder @xcite .
the nature of the spectrum and correlation functions play an important role in the description of the electronic wave functions at that fixed point .
previous works on the @xmath0 model have focused on some correlation functions @xcite,@xcite and on the construction of some characters @xcite , but a complete picture of the theory has been missing . indeed , the analysis of wznw on supergroups is notoriously difficult , even for the simplest case of @xmath14 @xcite . in a recent paper @xcite ,
we have shown how a careful study of the particle limit ( in particular , of the simultaneous left and right invariant actions on the space of functions on the group ) could provide considerable insight into this problem .
combining this insight with some additional input from the representation theory of the current algebras allowed us to formulate a complete proposal for the state space of the theory in the case of @xmath14 .
the latter involves a rather intricate mixing of left and right movers that is intimately related to the representation theory of lie superalgebras , in particular to the importance of indecomposable representations .
we were then able to check this proposal through an exact construction of the theory in the continuum formulation .
the aim of this work is to extend the lessons we have learned in @xcite to a non - abelian setup , using @xmath0 as the simplest non - trivial example .
in which the abelian , time - like circle is replaced by the real line .
we shall comment on this in much more detail in section 4 .
] once more , the analysis of the particle limit ( section 2 ) along with some input from the representation theory of the @xmath15 current algebra ( section 3 ) shall provide all the necessary ingredients for the construction of the field theory state space ( section 4 ) , in close analogy to our previous investigation of the @xmath14 model . in the present case
we shall not attempt to verify the structure of the state space through calculations of correlators , though this would be possible as well ( see @xcite ) .
instead we shall use results on an integrable @xmath15 spin chain to test our continuum constructions .
such a spin chain was first investigated in @xcite as a discrete version of the @xmath0 wznw model .
we shall see that both approaches are consistent .
the comparison , however , is a bit subtle , mainly due to the fact that the supergroup @xmath0 has an indefinite metric . while this poses no problem for the ( algebraic ) conformal field theory analysis , the computation of the partition function on the lattice suffers from divergencies which need to be regularized .
we shall do this through some appropriate analytic continuation . in this sense
, our analysis also supports a particular prescription for extracting information from spin chains with an indefinite metric .
the aim of this section is to decompose the space of functions on the supergroup @xmath0 into ( generalized ) eigenfunctions of the quadratic casimir element in the regular representations .
since the casimir commutes with the generator , the eigenspaces may be decomposed into representation of the lie superalgebra @xmath15 .
it is therefore useful to have some background on the representation theory of @xmath15 .
we shall review a few known facts below before addressing the harmonic analysis .
more details can be found e.g. in @xcite . in this subsection
we provide a short overview on finite dimensional representations of @xmath15 . rather than reproducing a complete list of such representations we shall focus on those that are relevant below , namely on kac modules and the projective covers of atypicals .
the even part @xmath16 gl(1 ) @xmath17 sl(2 ) of the lie superalgebra @xmath18 @xmath15 is generated by four bosonic elements @xmath19 , @xmath20 and @xmath21 which obey the commutation relations @xmath22&\ = \
\pm e^\pm\quad , & [ e^+,e^-]&\ = \ 2h\quad , & [ b , e^\pm]&\ = \ [ b , h]\ = \ 0\ \ .\end{aligned}\ ] ] in addition , there exist two fermionic multiplets @xmath23 and @xmath24 which generate the odd part @xmath25 .
they transform as @xmath26 with respect to the even subalgebra , i.e.@xmath27&\ = \ \pm\frac{1}{2}f^\pm & [ h,\bar{f}^\pm]&\ = \ \pm\frac{1}{2}\bar{f}^\pm\nn\\[2 mm ] [ e^\pm , f^\pm]&\ = \ [ e^\pm,\bar{f}^\pm]\ = \ 0 & [ e^\pm , f^\mp]&\ = \ -f^\pm & [ e^\pm,\bar{f}^\mp]&\ = \ \bar{f}^\pm\\[2 mm ] [ b , f^\pm]&\ = \ \frac{1}{2}f^\pm & [ b,\bar{f}^\pm]&\ = \ -\frac{1}{2}\bar{f}^\pm\ \ .\nn\end{aligned}\ ] ] finally , the fermionic elements possess the following simple anti - commutation relations @xmath28 among each other
. formulas to provide a complete list of relations in the lie superalgebra @xmath15 .
kac modules @xcite are the basic tool in the construction of irreducible representations . in the case of @xmath29 @xmath15 ,
these form a 2-parameter family @xmath30 of @xmath31-dimensional representations .
we may induce them from the @xmath32-dimensional representations @xmath33 of the bosonic subalgebra @xmath34 by applying the pair @xmath35 of fermionic elements .
our label @xmath36 denotes a gl(1)-charge and spins of sl(2 ) are labeled by @xmath37 . the dual construction which promotes the fermionic generators @xmath38 to creation operators , yields anti - kac modules @xmath39 ( @xmath40 and @xmath41 take the same values as above ) .
the bosonic content of ( anti-)kac modules may be read off rather easily from their construction , @xmath42 for generic values of @xmath40 and @xmath41 , the modules @xmath30 and @xmath39 are irreducible and isomorphic . at the points @xmath43 , however , they degenerate , i.e. the representations are indecomposable and no longer isomorphic . in fact ,
kac and anti - kac modules are then easily seen to possess different invariant subspaces . to be more precise
the ( anti-)kac modules @xmath44 and @xmath45 are built from two atypical representations such that @xmath46 \{\overline{\pm j , j}\}:&\qquad\{j-{{\scriptstyle\frac{1}{2}}}\}_\pm\ \longrightarrow\ \{j\}_\pm\ \ .
\end{split}\ ] ] the atypical irreducible representations @xmath47 that appear in these small diagrams are @xmath48 dimensional . with respect to the even subalgebra
they decompose according to @xmath49 ( -j , j)\oplus\bigl(-(j+{{\scriptstyle\frac{1}{2}}}),j-{{\scriptstyle\frac{1}{2}}}\bigr ) & , \text { for } -\text { and } j={{\scriptstyle\frac{1}{2}}},1,\dots\ \ \end{cases}\ ] ] for @xmath50 , only the trivial representation @xmath51 occurs .
it is also useful to introduce the characters of these representations . by definition ,
these are obtained as @xmath52 where the super - trace extends over all states in the representation @xmath53 of @xmath15 . for kac modules
the character is rather simple .
in fact , it factorizes @xmath54 with a fermionic contribution @xmath55 that is independent of the kac module under consideration , @xmath56 the characters of atypical representations can be obtained easily form their decomposition formulas ( [ atypdec ] ) .
we would like to pursue a rather different route here that uses the decomposition ( [ kmdec ] ) of kac modules into atypicals .
the first formula implies that @xmath57 we can solve these equations for the characters of atypical representations by the following infinite sums @xmath58 one may check by explicit computation that the contributions from all but two bosonic multiplets cancel each other in the infinite sum through a mechanism that is visualized in figure 1 .
the remaining two terms certainly agree with the decomposition formulas ( [ atypdec ] ) .
our derivation here may seem like a rather complicated path for such a simple result , but we shall see later that the same trick works for characters of atypical affine representations which are otherwise difficult to obtain
. by definition , the projective cover of a representation @xmath47 is the largest indecomposable representation @xmath59 which has @xmath47 as a subrepresentation ( its socle ) .
we do not want to construct these representations explicitly here .
instead , we shall display how they are composed from atypicals .
the projective cover of the trivial representation is an @xmath60-dimensional module of the form @xmath61 for the other atypical representations @xmath47 with @xmath37 one finds the following diagram , @xmath62 these representation spaces are @xmath63-dimensional .
let us agree to absorb the superscript @xmath64 on @xmath65 into the argument , i.e. @xmath66 , wherever this is convenient .
now we are prepared to analyze the space of functions on the supergroup @xmath0 .
for this purpose , let us introduce coordinates through the following explicit decomposition of elements @xmath67 , @xmath68 here , the bosonic base su(2)@xmath69 is parametrized by an element @xmath70 su(2 ) @xmath71 along with the time - like variable @xmath72 .
in these coordinates , the generators of the right regular action read @xmath73 r_h = r_h^0 + \frac12 \eta_- \partial_- - \frac12 \eta_+ \partial_+\ \ & , & \ \
r_b \ = \ -i\partial_z - \frac12 \eta_-\partial_- - \frac12 \eta_+ \partial_+ \ \ , \\[2 mm ] & & \hspace*{-6 cm } r_{\bar f^{\pm } } \ = \
i e^{-iz/2 } \ , d^{1/2}_{\nu ( \pm 1/2 ) } ( g ) \ , \bar\partial_{-\nu } + i \eta_\pm ( r^0_{e^\pm } + \eta_\mp \partial_\pm ) - i \eta_\mp ( i \partial_z \mp r^0_h ) \end{aligned}\ ] ] where @xmath74 are the generators of the right regular representation of su(2 ) .
they act on the matrix elements @xmath75 according to @xmath76 r_{e^-}^0 \ , d^j_{a(b+1)}(g ) & = & \sqrt{(j+b+1 ) ( j - b ) } \
d^j_{ab}(g ) \ \ .\end{aligned}\ ] ] matrix elements with @xmath77 appear as coefficients in the differential operators @xmath78 and their behavior under the action of @xmath74 plays an important role in checking that the above generators of the right regular representation obey the defining relations of @xmath15 .
formulas for the left regular representation may be obtained similarly , @xmath79 l_h \ = \ l_h^0 + \frac12 \bar \eta_- \bar \partial_- - \frac12 \bar \eta_+ \bar \partial_+\ \ & , & \ \
i\partial_z + \frac12 \bar \eta_-\bar \partial_- + \frac12 \bar \eta_+ \bar \partial_+ \ \ , \\[2 mm ] & & \hspace*{-6 cm } l_{f^{\pm } } \ = \
i e^{-iz/2 } d^j_{(\pm 1/2)\nu}(g ) \
\partial_{-\nu } + i \bar \eta_\pm ( l^0_{e^\pm } - \bar \eta_\mp \bar \partial_\pm ) + i \bar
\eta_\mp ( i\partial_z \mp l^0_h ) \end{aligned}\ ] ] it is probably not necessary to stress that left and right generators ( anti-)commute with respect to each other . by construction ( see however @xcite ) , the generators of the left and right regular representation act on the space of all grassmann valued functions with square integrable coefficients on the bosonic base , i.e. on the space @xmath80 where @xmath81 denotes the grassmann algebra that is generated by our four fermionic coordinates @xmath82 and @xmath83 . with respect to the left regular action , the space of square integrable functions can be shown to decompose as follows , @xmath84 & & \hspace*{1 cm } \oplus \ \sum_j \ ( 2j+1)\ , \left(\p^+_j \oplus \p^-_j\right ) \oplus 2j \left(\p^+_j \oplus \p^-_j\right ) ' \ \ . \nonumber\end{aligned}\ ] ] here , the summation runs over @xmath85 , @xmath30 denotes the typical representations of @xmath15 and @xmath86 are the projective covers of the atypical representations @xmath47 .
most of our conventions can be found e.g. in @xcite .
a prime @xmath87 on a representation means that the degree is inverted , i.e. that fermionic vectors become bosonic and vice versa .
the result is a special case of the general observation made in @xcite and it generalizes a similar decomposition we described in @xcite for the left regular action of @xmath88 .
the interested reader can find an explicit proof in appendix a. let us comment that the decomposition of the left regular action displays the same violation of the peter - weyl theorem as in the case of @xmath14 .
in particular , since the quadratic casimir is not diagonalizable in the projective covers @xmath89 , the laplacian on the supergroup @xmath0 can only be brought into jordan normal form .
the blocks can reach a rank up to three .
the functions on our supergroup carry another ( anti-)commuting action of the lie superalgebra @xmath90 by left derivations .
there is a corresponding decomposition which is certainly identical to the decomposition above .
a more interesting problem is to decompose the space of functions with respect to the graded product @xmath91 in which the first factor acts through the left regular action while for the second factor we use the right regular action . in the typical sector ,
the @xmath92-dimensional multiplicity spaces in the first line of eq .
( [ ldec ] ) get promoted to typical representations of the right regular action , i.e.@xmath93 where @xmath94 is a single indecomposable , containing all the atypical building blocks .
its structure may be summarized by the following picture @xmath95 \{{{\scriptstyle\frac{1}{2}}}\}_-\times \{{{\scriptstyle\frac{1}{2}}}\}_+ \ar[dr]\ar[drrr ] & & & & \ar[dl ] \ar[dlll ] \{0\}\times \{0\ } \ar[dr ] & \\
\cdot\cdot \{1\}_- \times \{{{\scriptstyle\frac{1}{2}}}\}_+ \ar[dr ] & & \{0\}\times \{{{\scriptstyle\frac{1}{2}}}\}_+ \ar[dl]\ar[drrr ] & \ \ & \{{{\scriptstyle\frac{1}{2}}}\}_- \times \{0\ } \ar[dr ] \ar[dlll ] & & \{{{\scriptstyle\frac{1}{2}}}\}_+\times \{0\ } \ar[dl ] \cdot\cdot \\ & \{{{\scriptstyle\frac{1}{2}}}\}_-\times \{{{\scriptstyle\frac{1}{2}}}\}_+ & & & & \{0\}\times \{0\ } & } \ ] ] this diagram is the natural extension of the corresponding picture for @xmath14 .
it extends to infinity in both directions and combines all the atypical sectors into a single indecomposable representation .
note that by construction , each projective cover in the decomposition of the right regular representation appears with the correct multiplicity .
we shall see below how this picture is modified in the full quantum theory .
the previous analysis of the particle limit applies to all sigma models on @xmath0 , but the information it provides is usually not sufficient in order to reconstruct the entire field theory from it .
this is very different for the wznw model in which the entire spectrum can be generated from particle wave functions through current algebra symmetries .
we need some facts on the representation theory of the @xmath15 current algebra and shall provide them in the following section .
all the results we collect here are well known from @xcite .
their derivation , however , is somewhat original . in particular , we shall use a simple , but highly efficient prescription to construct characters of atypical representations of @xmath1 through infinite sums over typicals .
this extends the formula ( [ atypchar ] ) we have discussed in section 2 to an infinite dimensional setting , thereby generalizing a trick that has first been proposed in the context of the @xmath88 current algebra @xcite .
irreducible representations of the affine @xmath15 algebra can be built over the irreducible typical representations @xmath30 with @xmath96 as well as over the atypicals @xmath47 with @xmath97 .
ground states in the former set of representations possess conformal dimension @xmath98 while the conformal dimension for ground states in the latter set vanishes . following the work @xcite of bowcock et al .
we shall divide these representations into three different classes .
the generic class i representations occur for @xmath99 with @xmath100 where we defined @xmath101 class ii representations include those erected over @xmath30 with @xmath102 along with the sectors generated from atypicals @xmath103 . the vacuum representation that is generated from
the atypical @xmath104 is the only member of the final class , which we denote as class iv for historical reasons .
our aim is to describe the singular vectors in the corresponding verma modules and to provide the associated formulas for the super - characters @xmath105 of irreducible representations .
the results we describe have first appeared in @xcite .
before we start our discussion of characters let us quickly recall that it is possible to construct @xmath15 currents in terms of decoupled bosonic and fermionic variables . to be more precise , we introduce a set of bosonic currents @xmath106 and assume them to satisfy the operator product expansions of an affine sl(2 ) algebra at level @xmath107 .
in addition , let us introduce two sets of fermionic fields @xmath108 and @xmath109 obeying the canonical relations @xmath110 then we can construct an @xmath15 current algebra at level @xmath111 through the following prescription , @xmath112 e^-(z ) & = \
e^-(z ) + : \theta^1 p^2:(z ) \ \ , \ \
b(z ) \ = \
b(z ) - \frac12\ , : \left(\theta^1p^1 + \theta^2p^2\right):(z ) \ , \ \ \ \ \\[3 mm ] f^+(z ) & = \
p^2(z ) \ \ \ \ \ , \ \ \ \ \ \ \bar{f}^{+ } \ = \
\theta^2 e^+(z ) + \theta^1(b + h)(z)\ , - : \theta^1 \theta^2 p^2:(z)\ , \\[3 mm ] f^-(z ) & = \
p^1(z ) \ \ \ \ \ , \ \ \ \ \ \ \ \bar{f}^{- } \ = \
\theta^1 e^-(z ) + \theta^2(b - h)(z)\ , + : \theta^1 \theta^2 p^2:(z)\ . \end{split}\ ] ] since the fermionic fields @xmath109 and @xmath108 are supposed to commute with the bosonic fields @xmath113 and @xmath114 , the characters of typical representations factorize with the factors @xmath115 arising from the fermionic pairs .
the shift @xmath116 in the bosonic contribution may be traced back to a similar shift in the labeling of typical @xmath15 representations , see eq .
( [ typdec ] ) .
the generic class i representations have no singular vectors except from the ones that arise through the representations of a bosonic su(2 ) current algebra at level @xmath107 . in this sense
, they may be considered the typical representations of the affine @xmath15 algebra .
the statement implies a precise expression for the characters of class i representations @xmath117 & & \hspace*{-3 cm } \mbox { where } \ \
\theta_1(y , q ) \ = \
-i y^{1/2 } q^{1/8}\ \prod_{n=1}^\infty \ , ( 1-q^n ) ( 1-yq^n ) ( 1-y^{-1 } q^{n-1})\end{aligned}\ ] ] and @xmath118 and @xmath119 .
we also recall that the su(2 ) characters are given by @xmath120 we shall use the symbol @xmath121 for these irreducible representations of the affine algebra .
the formulas are easy to understand : they follow directly from the representation ( [ bfdec ] ) of the @xmath15 current algebra .
in fact , each pair of fermionic fields contributes a factor @xmath122 while the bosonic sl(2 ) and u(1 ) current algebras are responsible for the characters @xmath123 and an additional factor @xmath124 , respectively .
nothing prevents us from evaluating the previous character formulas at the points @xmath125 .
but the resulting functions turn out to be the characters of indecomposable representations @xmath126 which contain one fermionic singular multiplet . in order to state this
more precisely , let us consider in more detail the set of atypical labels , @xmath127 the set @xmath128 is visualized in figure 2 .
our picture shows clearly that the projection to the @xmath40-coordinate of each element in @xmath128 is injective and hence it can be used to enumerate our atypical labels .
note , however , that values @xmath129 are omitted .
this motivates to introduce an improved enumeration map @xmath130 from @xmath128 to non - zero half - integers which is defined by @xmath131 \hat\nu ( \{j^-_m , j\ } ) \ = \
j^-_m + m \ \ \mbox { for } \ \
m\ \leq 0 \ \ & , & \ \
\hat\nu ( \{j^+_m , j\ } ) \
= \ j^+_m + m - 1/2 \ \
\mbox { for } \ \ m < 0\ \ .\end{aligned}\ ] ] by construction , @xmath130 is not only an injection but its image now also consists of all nonzero half - integers
. we may view @xmath132 as an affine version of the enumeration map @xmath133 for representations of @xmath15 . at first sight , the enumeration of atypical labels for our @xmath15 current algebra may seem like a rather technical device .
but there is more to it .
we recall that atypical labels @xmath44 can also be enumerated by non - zero half - integers , i.e. @xmath134 .
our claim now is that the atypical class i representation with label @xmath135 behaves very similarly to its finite dimensional counterpart @xmath136 in the sense that @xmath137 this formula is a rather central result for the representation theory of our current algebra .
let us stress that it is the affine version of a corresponding equality ( [ chardiff ] ) between characters of @xmath15 representations .
as in the finite dimensional setup , eq .
( [ classi / ii ] ) emerges from the existence of fermionic singular vectors in atypical class i representations . in the case
@xmath138 , it claims that the only such singular vectors are those that appear in the atypical kac - module spanned by the ground states .
when @xmath139 , however , the ground states form a typical representation and the singular vectors appear only on the @xmath140 level of the class i module . at first it may seem a bit surprising that affine representations @xmath141 and @xmath142 behave so similarly to the kac modules of @xmath15 .
in the next subsection we shall understand this behavior in terms of spectral flow symmetries in the representation theory of the current algebra . before any study of spectral flow automorphisms
, it might be useful to illustrate the similarity between atypical representations of the current algebra and their finite dimensional counterpart more explicitly , at least for one example .
to this end , let us focus on the representation @xmath143 which we claim to be a close cousin of the @xmath15 representation @xmath144 . by construction ,
the ground states of the current algebra representation transform in the typical multiplet @xmath145 and they possess conformal weight @xmath146 . from these vectors
we can generate states with vanishing conformal weight with the help of modes in the current algebra .
such modes transform in the 8-dimensional adjoint representation @xmath147 of @xmath15 . through decomposition of the tensor product between @xmath145 and @xmath147 one
finds that the verma module over @xmath148 contains an atypical @xmath15 multiplet with conformal weight @xmath149 .
in fact , the results of @xcite imply that the latter transforms according to the projective cover @xmath150 , see eq .
( [ p ] ) .
not all of these states survive when we descend from the verma module to the class i representation .
this step involves removing bosonic singular vectors and a moment of reflections shows that such vectors with @xmath149 exist and that they transform in the submodule @xmath151 of @xmath150 .
hence , the states with @xmath149 in our class i representation decompose into the kac - module @xmath148 plus a bunch of typical representations .
the fermionic singular vectors that are responsible for eq .
( [ classi / ii ] ) transform in the subrepresentation @xmath152 of @xmath148 , giving rise to the identity ( [ classi / ii ] ) with @xmath153 and @xmath154 . before we draw some conclusions from eq .
( [ classi / ii ] ) , let us quickly comment on our notations .
note that for @xmath77 and @xmath155 the above formula involves the character @xmath156 of a representation which has a somewhat special status .
in fact , it can not be obtained as quotient of one of the indecomposable representations @xmath126 , unlike the representations with characters @xmath157 .
instead , it arises as a submodule of the representations @xmath158 .
our discussion suggest that @xmath156 must be the character of the vacuum representation . in the terminology of bowcock et al .
the latter is a class iv representation .
thus , we shall also write @xmath159 for this quantity . even though equation ( [ classi / ii ] ) is not a closed formula for the characters of class ii representations
, we can now use the same trick as in section 2.1.2 and write characters of class ii representations as an infinite sum of class i characters , @xmath160 for @xmath161 .
note that the map @xmath130 is invertible on all non - zero half - integers and it furnishes the label of the kac module that sits at the bottom of the corresponding class i representation . by inserting our explicit formulas for class i characters
we find @xmath162 & & \hspace*{1 cm } \times \ \sum_{a \in \mathbb{z } } \
q^{(k+1)a^2 \mp 2aj } \ , \xi^{\pm j}\ , \left(\frac{z^{-a(k+1 ) \pm j}}{1+q^a z^{-1/2}\xi^{-1/2 } } - \frac{z^{a(k+1 ) \mp j } } { 1 + q^a z^{1/2}\xi^{-1/2 } } \right ) \ \ .
\nonumber\end{aligned}\ ] ] character formulas of this type have to be used with some care : before the denominators are expanded , one should spit the summation over @xmath163 into two parts .
the one arising from positive values of @xmath163 can be converted into a power series right away . in all terms with non - negative @xmath163 ,
however , one must first reduce the fraction by @xmath164 such that the subsequent expansion contains only non - negative powers of @xmath165 . in the end
, we recover the known results on the representation theory of the affine @xmath15 algebra @xcite .
our derivation was based on three ingredients : the decoupling formulas ( [ bfdec ] ) for bosonic and fermionic generators , the structure ( [ kmdec ] ) of atypical kac modules for @xmath15 and the fact that atypical class i representations with @xmath142 decompose in the same way as in the case of @xmath138 .
we shall argue now that the last ingredient emerges from spectral flow symmetries in the representation theory of affine @xmath15 .
the affine @xmath15 algebra admits several interesting automorphisms .
we shall be mainly concerned with two such spectral flow automorphisms @xmath166 . by construction , @xmath166
are defined on the entire current algebra , but for our purposes it is sufficient to know how they act on the generators @xmath167 , @xmath168 from these formulas we may infer how ( super-)characters behave under the action of @xmath166 and this in turn is sufficient to determine how spectral flow automorphisms map representations of the current algebra onto each other . along with @xmath166
we shall also be interested in the composite automorphism @xmath169 which acts as @xmath170 any automorphisms of the current algebra gives rise to a map between representations and hence to a map between characters . from the action on the zero modes @xmath171 and @xmath172 we can easily read off that @xmath173 for all characters @xmath174 of the @xmath15 current algebra .
if @xmath53 is any representation of @xmath1 and @xmath175 is its character , then the image @xmath176 of @xmath53 under an automorphism @xmath177 obeys @xmath178 given the character @xmath175 of some representation @xmath53 , we can use eqs .
( [ gchi ] ) to compute its image under the above automorphisms @xmath179 .
this in turn allows us to recover uniquely the representations @xmath180 and @xmath181 . in the following
we shall spell out the action of our spectral flow automorphisms on the class i and ii representations we have studied above .
our rather compact notations allow us to summarize the results for the spectral flow automorphisms @xmath166 in a single line @xmath182 to verify our assertions , the reader is invited to convert them into identities between super - characters and to check these identities by direct computation .
the formulas become somewhat more explicit if we label irreducible representations according to the representation their ground states transform in , @xmath183 \ { b , j\ } & \stackrel{\gamma_\pm}{\longrightarrow } \ \{k/2 + { { \scriptstyle\frac{1}{2}}}- j \}_\mp \hspace*{2.5cm}\ \
\ \mbox { for } \ \ b = \pm(j - k - 1)\\[2 mm ] \{j\}_\pm & \stackrel{\gamma_\pm}{\longrightarrow } \ \{\pm(j + k/2 + { { \scriptstyle\frac{1}{2}}}),k/2 + { { \scriptstyle\frac{1}{2}}}-j\ } \ \ \ \mbox { for } \ \
j \neq 0 \\[2 mm ] \{j\}_\mp & \stackrel{\gamma_\pm}{\longrightarrow } \
\{k/2 -j\}_\pm \ \ \ \ \ \ , \ \ \ \ \ \ \{0\ } \ \stackrel{\gamma_\pm}{\longrightarrow } \ \ { k/2\}_\pm \ \ .\end{split}\ ] ] the third line , for example , tells us that the image of the irreducible representation generated from the atypical representation @xmath184 under the action of @xmath185 is an irreducible representation whose ground states transform in the typical representation @xmath186 .
the latter may be obtained from the corresponding verma module by removing singular vectors on some excited levels .
we also want to spell out analogous formulas for the automorphism @xmath187 . in the compact notation
, its action is given by @xmath188 note that the symmetry @xmath189 maps sectors whose ground states transform in an atypical representation @xmath47 of the lie superalgebra @xmath15 into sectors with typical spaces of ground states according to the following rules , @xmath190 \{j\}_\pm & \stackrel{\gamma^m}{\longrightarrow } & \{j_m^\pm \pm 1/2,j + 1/2\ } \ \ \ \mbox { for } \ \ \ \pm m
\ < \ 0 \ \ .\end{aligned}\ ] ] hence , the existence of the spectral flow symmetries explains why the representations @xmath141 behave like atypical representations of the affine @xmath15 algebra : they are simply related to the sectors erected over atypical @xmath15 representations by an automorphism .
we would like to conclude this section on the representation theory of the @xmath15 current algebra with a few comments on modular properties of the characters . in the following
we shall consider the characters as functions of @xmath191 and @xmath192 .
they are related to the variables we used above through @xmath193 and @xmath194 , as usual . from our explicit formula ( [ typei ] ) for characters of class i representations it is easy to infer the auxiliary formula @xmath195 note that the right hand side contains an explicit @xmath196 dependence which , if we demand that the modular transform be interpreted in a conventional sense and closes onto characters , suggests the contribution of a continuous spectrum of exponents .
the need is confirmed by the modular transformation of the character for @xmath30 representations with @xmath197 , which require an integral representation of @xmath198 etc .
after some gaussian integration , we find @xmath199 where we formally evaluated integrals of the type @xmath200 [ of course , the integrals are naively divergent as @xmath201 . ] modular transformations of the type ii and iv characters are a bit more cumbersome to work out .
it can be attacked rather efficiently using our representations ( [ classiichar ] ) as infinite sums of class i characters . here
we shall content ourselves with the example of the class iv representation at @xmath2 .
if we also set @xmath202 we find that @xmath203 where the contour has to avoid the poles .
rotating into the purely imaginary direction gives a contribution from poles which is easily identified with @xmath159 .
the remaining integral can be factored in terms of @xmath204 , @xmath205 we thus recover by this very elementary means the results of @xcite obtained through use of the mordell integral @xcite .
the construction of modular invariants using these transformation formulas is a complex problem , which we shall address later in the case @xmath2 .
our aim now is to formulate a proposal for the states space of the @xmath15 wznw model .
we shall then verify our claim in the special case @xmath2 through a free field representation of the model .
the third subsection is devoted to the partition function of the theory .
the latter forgets all information about the complicated way in which irreducible blocks are glued together to build @xmath206 .
we then specialize once more to @xmath2 and comment on the global topology of the target space .
it is now rather straightforward to come up with a proposal for the state space of the wznw model on @xmath0 .
in fact , we can simply depart from formula ( [ lrdec ] ) and make it symmetric with respect to the action of our spectral flow symmetry .
the invariance under the action of @xmath189 should be considered as an additional input . in principle , the spectral flow symmetry of the @xmath15 current algebra could be broken by the physical couplings of the theory . since this did not happen for the @xmath14 wznw model , it seems natural to propose @xmath207 where @xmath208 is a single indecomposable representation of the two ( anti-)commuting super current algebras that contains all the atypical contributions .
it is composed from the atypical building blocks @xmath209 in the same way as in the minisuperspace theory . to obtain the corresponding diagram one simply has to replace @xmath210 with @xmath211 . by construction ,
all the @xmath15 currents act on the state space ( [ hcft ] ) and they obey periodic boundary conditions .
this applies in particular to the fermionic fields .
one can find a second , closely related theory in which only bosonic fields are periodic . in order to construct its state space ,
we need to revisit our discussion of spectral flow symmetries .
as we have mentioned above , the automorphisms we have investigated in the previous section all extend to the entire current algebra .
in particular , they map fermionic modes with integer mode numbers onto each other , i.e. they respect periodic boundary conditions on the fermionic currents .
there exists yet another important isomorphism that intertwines between integer and half - integer mode numbers for the fermionic generators .
it can be considered as the square root of the automorphism @xmath189 . on the bosonic zero modes , the new isomorphism @xmath212 is given by @xmath213 @xmath212 extends to the full current algebra such that it acts trivially on the bosonic sl(2 ) currents and it shifts modes of the fermionic currents by @xmath214 , as usual .
our isomorphism @xmath212 induces a map between representations of the current algebra with integer fermionic modes and a new type of representations in which fermionic generators have half integer mode numbers . according to the usual terminology , the former class of representations form the r sector while the latter belong to the ns sector of the theory
the theory with state space ( [ hcft ] ) includes exclusively r sector representations in which all currents obey periodic boundary conditions .
another option is to consider a theory that encompasses both r and ns sector with the state space given by @xmath215 note that the ns sector has exactly the same intricate structure as the r sector since the former is the image of the latter under the action of @xmath216 . in the following
we shall refer to both models as wznw model on @xmath0 . even though it seems natural to include the ns sector , it is not required by all applications .
so far , the main motivation for our proposal ( [ hcft ] ) came from the harmonic analysis on @xmath0 . by construction ,
we are guaranteed to recover the correct state space of the particle limit when we send the level @xmath111 to infinity .
our formula ( [ hcft ] ) applies to finite @xmath111 and it suggest that field theory effects would merely truncate the spin @xmath41 to an value @xmath217 and then make the whole theory symmetric with respect to spectral flow .
we are now going to test this proposal in the extreme quantum case , namely at @xmath2 . at this point ,
the wznw model admits a free field representation that we are going to spell out momentarily .
it involves a pair of symplectic fermions @xmath218 , and a pair of bosons @xmath219 .
while the boson @xmath220 comes with the usual metric , @xmath221 is assumed to be time - like . for their propagators
this means @xmath222 note that the central charge of this free field theory is @xmath223 and hence it agrees with the central charge of @xmath0 wznw models .
we shall begin our discussion of the wznw model with explicit formulas for the currents . in order to construct the four bosonic currents , we need to split the space - like bosonic field @xmath224 into its chiral components .
our bosonic currents then read , @xmath225 h ( z)\ = \ \frac{1}{\sqrt{2}}\ , i\partial\phi(z ) \ \ \ & , & \ \ \
b ( z)\ = \ -\frac{1}{\sqrt{2}}\ , i\partial\phi'(z)\ \ . \end{aligned}\ ] ] the necessity to split @xmath220 into its chiral components means that the boson @xmath220 is compactified to the so - called self - dual radius , as usual in the free field representation of the su(2 ) wznw model at level @xmath2 .
in addition , the following expressions for the four fermionic currents also involve the chiral components @xmath226 and @xmath227 of the time - like bosonic field @xmath228 , @xmath229 similarly , one may spell out the anti - holomorphic generators of the @xmath15 current algebra .
it is rather easy to check that the above expressions give rise to fields with the correct operator product expansions .
let us note that the free field representation we consider in this section has to be distinguished clearly from the kac - wakimoto type construction ( [ bfdec ] ) we have used earlier to construct the characters at integer levels @xmath111 .
we shall comment on this a bit more later on .
it is possible to check that fields of dimension zero can be organized exactly as it is suggested by our diagram ( [ diagram ] ) .
we shall just sketch the relevant arguments because a full proof is rather laborious to write down .
let us consider the left part of the diagram only and identify the fields that make up the various blocks of the composition series .
clearly , the @xmath230 representation at the top corresponds to the field @xmath231 .
from here we can act with the fermionic currents @xmath232 and arrive at expressions for the two blocks on the intermediate level of the diagram , @xmath233 \{0\}\times\{1/2\}_{+}\ : \ \ & & e^{\pm i\bar{\varphi}/\sqrt{2}}\ , e^{i\bar{\varphi}'/\sqrt{2}}\ , \eta_{1}\ \ \ , \ \ \
e^{i\sqrt{2}\bar{\varphi}'}\ , \eta_{1}\bar{\partial}\eta_{1 } \ \ .
\nonumber \end{aligned}\ ] ] from the previous formulas we can read off the fields that make up the topmost representation @xmath234 in our diagram , @xmath235 e^{-i\sqrt{2}\varphi'}\ , \eta_{2}\partial\eta_{2}\end{array } \right .
\ \times \ \left\{\begin{array}{c } e^{\pm i\bar{\varphi}/\sqrt{2}}\ , e^{i\bar{\varphi}'/\sqrt{2}}\ , \eta_{1}\\[1 mm ] e^{i\sqrt{2}\bar{\varphi}'}\ , \eta_{1}\bar{\partial}\eta_{1 } \ \ . \end{array}\right .
\label{bigie}\ ] ] acting with the holomorphic fermionic currents @xmath236 we arrive at the following formulas for fields that belong to the multiplet @xmath237 \eta_{2}\ , e^{\pm i\sqrt{2}\varphi}\ , e^{-i\sqrt{2}\varphi'}\\[1 mm ] \eta_{2}\partial\phi \ , e^{-i\sqrt{2}\varphi ' } \end{array } \right.\ \times\ \left\{\begin{array}{c } e^{\pm i\bar{\varphi}/\sqrt{2}}\ , e^{i\bar{\varphi}'/\sqrt{2}}\ , \\[1 mm ] e^{i\sqrt{2}\bar{\varphi}'}\ , \bar{\partial}\eta_{1 } \ \ .. \end{array}\right.\ ] ] on the intermediate level of the diagram .
our notation means that every product of the three holomorphic fields on the left hand side with the three anti - holomorphic fields on the right hand side is part of this 9-dimensional block .
similarly , we can now descend to the bottom of the diagram , @xmath238 \eta_{2}\partial \phi'\ , e^{-i\sqrt{2}\varphi ' } \end{array}\right.\ \times\ \left\{\begin{array}{c } e^{\pm i\bar{\varphi}/\sqrt{2}}\ , e^{i\bar{\varphi}'/\sqrt{2}}\ , \\[1 mm ] e^{i\sqrt{2}\bar{\varphi}'}\ , \bar{\partial}\eta_{1 } \end{array}\right.\ ] ] finally , the representation @xmath230 in center bottom position is represented by the identity field .
it is easy but laborious to check that the different representations are connected by the action of the left and right generators as indicated in the diagram . in checking this , notice that @xmath239 up to a total derivative .
there are a number of interesting further comments and observations that we would like to make .
let us begin with a brief comment on the relation with kac - wakimoto like representations of the form ( [ bfdec ] ) . as discussed in @xcite , a naive evaluation of the action of the latter on vertex operators leads to a much simpler picture
in which the atypical sector is a smooth deformation of the typical part .
in particular , there is no mixing between left- and right movers as in the case of the representation @xmath208 . in order to see the latter , the screening charge of the kac - wakimoto representation
must be taken into account ( see @xcite for details ) .
the free field representation we have employed in this subsection is much simpler to use , but it is restricted to @xmath2 .
the free field representation also allows us to illustrate very explicitly how atypical fields of dimension @xmath149 are embedded into sectors with ground states in typical multiplets once their spin exceeds @xmath240 .
take , for instance , the field @xmath241 from the @xmath242 representation and observe that @xmath243 thus @xmath244 is not a highest weight of the current algebra , and applying a current operator can give rise to a field of dimension @xmath245 , namely the field @xmath246 .
this field belongs to the typical representation @xmath247 in the sector @xmath248 with affine highest weight @xmath249 .
one may easily generalize these observations to all representations in the complex , hence confirming our analysis based on spectral flow .
let us finally come to the most important point , which concerns the possible construction of consistent theories that are realized on a subspace of our state space ( [ hcft ] ) .
note that @xmath250 does contain a large number of fermionic singular vectors that we decided not do decouple , partly because the minisuperspace analysis suggested that it was unnatural to do so .
but to a certain extend one should consider ( [ hcft ] ) as some kind of maximal choice from which other models can be obtained by consistent decoupling of singular vectors . in general
there can be several such reduced theories .
once more we may use our free field representation for the @xmath2 model to illustrate nicely how this works .
note that the expressions ( [ fcurrents ] ) for the currents only contain derivatives of the fermionic fields .
hence , we do not spoil the @xmath15 current algebra symmetry if we decide to work with a model in which the fundamental fields are e.g.@xmath251 and @xmath252 .
since @xmath253 is not part of this model , some of the sectors we discussed above do no longer appear .
this concerns the sectors @xmath254 and @xmath255 on the top floor of @xmath208 and the sector @xmath256 on the intermediate floor . in the resulting model
, there is still a single atypical sector that comprises all the irreducible atypicals , but it is reduced to two floors and has the shape of a saw blade . obviously , a similar analysis applies to the theory that contains @xmath257 instead of @xmath258 .
but we can even go one step further and drop both @xmath253 and @xmath258 so that only fermionic derivatives remain . what results is a model whose atypical sector decomposes into an infinite sum of irreducibles .
the latter are the sectors that appear on the bottom floor of @xmath208 , i.e. @xmath259 and @xmath260 from our list above .
all others need either @xmath253 or @xmath258 .
similar phenomena are possible at other levels .
we shall see another explicit example in section 5.2 .
let us stress , however , that the free field construction at @xmath2 does not include the `` defining '' @xmath261 field of the wznw model . a careful study of the knizhnik - zamolodchikov equations @xcite shows that consistency in the presence of the @xmath261 sector requires the identity field to be embedded into an indecomposable sector . in this sense , the fully truncated atypical sector we have just described can not be embedded into the @xmath15 wznw model .
we now go back to the full theory based on our proposal ( [ hcft ] ) .
we would like to compute the partition function of the model , with and without inclusion of the ns sector .
since partition functions are obtained by taking the trace over the state space , the details of the action of fermionic generators in the atypical sector @xmath208 do not show up in the result .
in other words , the contribution from the indecomposable @xmath208 is the same as if we would take the trace over a sum of its irreducible components .
the latter can be resumed as follows , @xmath262 & = & \sum_{\nu \in \mathbb{z}/2 } \ , \left ( \chi^{ii}_{\{\nu\}}(\bar q ) - \chi^{ii}_{\{\nu-1/2\}}(\bar q)\right ) \left ( \chi^{ii}_{\{-\nu\}}(q ) - \chi^{ii}_{\{-\nu+1/2\}}(q)\right ) \\[2 mm ] & = & \sum_{\nu = 1/2,1,\dots } \ , \left ( \chi^{ii}_{\{\nu\}}(\bar q ) - \chi^{ii}_{\{\nu-1/2\}}(\bar q)\right ) \left ( \chi^{ii}_{\{-\nu\}}(q ) - \chi^{ii}_{\{-\nu+1/2\}}(q)\right ) \\[2 mm ] & & + \ \sum_{\nu = 1/2,1,\dots } \ , \left ( \chi^{ii}_{\{-\nu\}}(\bar q ) - \chi^{ii}_{\{-\nu+1/2\}}(\bar q)\right ) \left ( \chi^{ii}_{\{\nu\}}(q ) - \chi^{ii}_{\{\nu-1/2\}}(q)\right ) \\[2 mm ] & = & \sum_\pm \sum_{m \in \mathbb{z } } \sum_{j=1/2}^{k/2 } \ , \chi^i_{\{j^\pm_m , j\}}(\bar q)\ , \chi^i_{\{-j^\pm_m , j\}}(q)\ \ . \end{aligned}\ ] ] in the last step we have inserted the relation ( [ classi / ii ] ) between characters of class i and class ii representations and we used the isomorphism @xmath263 to convert the sum over non - zero half - integers into a sum over @xmath264 and @xmath41 .
our result shows that the contribution from the atypical representations agrees exactly with the part that we omitted from the typical sector of the theory .
hence , the full partition function becomes @xmath265 let us also briefly discuss how the partition functions is modified when we want to include the ns sector . in that case , the trace extends over both @xmath266 and its image under the spectral flow @xmath212 .
the modular invariant partition function @xmath267 of this theory contains four different terms , two in which @xmath268 is inserted and two in which it is not .
it is customary to label the corresponding contributions with @xmath269 , @xmath270 , @xmath271 and @xmath272 where the small @xmath273 signals the insertion of @xmath268 . with these notations ,
the standard super - characters we have discussed throughout the previous section should all carry a superscript @xmath270 .
it is easy to find explicit formulas for the other three sets of ( super-)characters using the relation @xmath274 to convert @xmath270 characters into @xmath272 super - characters .
the same prescription is used when we construct @xmath271 characters from the r sector , only that we have to replace the @xmath269 super - characters @xmath174 by ordinary characters .
the partition function @xmath275 , finally , has the same form as eq .
( [ pfh ] ) with an additional summation over all four types of terms .
let us illustrate the previous results in the case of @xmath2 again .
the ( @xmath270 ) characters of this theory take a particularly simple form , as was first observed by bowcock et al . in @xcite , @xmath276\ ] ] where @xmath277 and @xmath278 are @xmath279 level one characters for spin @xmath280 and @xmath281 respectively .
this expression allows us to determine the modular invariant physical partition function @xmath282 involving periodic or antiperiodic boundary conditions for the fermions along both periods of the torus . the doubly periodic sector ( @xmath283 ) gives a vanishing contribution for characters @xmath284 since the super - dimension of the horizontal kac modules vanishes .
we are left with three contributions , which read respectively @xmath285\\ \chi^{sns}_{\{b,1/2\}}(q)&=&\frac{q^{-b^{2}/2 } } { \eta(q)}\ \left[\chi_{0}^{2}(q)-\chi_{1/2}^{2}(q)\right]\ \ .
\end{aligned}\ ] ] their modular transformations are easy to obtain for @xmath286 @xmath287 obviously , the @xmath196 dependence of these formulas originates in the @xmath288 term in the characters , and has to be compensated by a similar factor coming from the @xmath40 sum in order to obtain a modular invariant quantity . the question we want to ask now is what kind of sum over @xmath40 one should consider
it seems at first sight that , since we are dealing with @xmath0 , the @xmath40 number should be discrete , in agreement with the imaginary exponential appearing in section 2.2 ( note that this is also compatible with invariance under action of the spectral flow ) . it is likely that such a theory would make sense in genus zero .
difficulties arise , however , when we try to establish modular invariance of the partition function , i.e.consider the theory in genus one . indeed , even if @xmath40 is discretized , there is no reason to truncate its range , and thus the naive spectrum of conformal weights is unbounded from below , and exhibits arbitrarily large negative dimensions .
this should not come as a surprise since the metric on the group is not positive definite , and thus the naive functional integral in the wznw model is divergent .
what is required to obtain a physical partition function on the torus - one that could be compared with the spectrum of some lattice hamiltonian say - is some sort of analytic continuation .
this raises some interesting questions on which we would like to digress briefly . for a compactified time - like boson
, partition functions would involve sums of the form @xmath289 this sum is obviously divergent , but one could be tempted to give it a meaning by analytical continuation from a similar sum with @xmath290 .
an equivalent problem arises when we try to continue a theta function such as @xmath291 into the lower half plane @xmath292 .
it is known that this continuation is not possible , since the @xmath216 function has singularities which are _ dense _ on the real axis ( a quick proof is obtained by first observing that @xmath216 is singular for @xmath196 an even integer , and then using modular transformations ) .
theta functions have a natural boundary , and are simple examples of lacunary functions , i.e. `` almost all '' their fourier coefficients are zero @xcite .
the partition function of a compactified time - like boson is thus a formal object from which it is hard to extract physical meaning . on the other hand , without compactification
, the partition function can easily be analytically continued . indeed , replacing the discrete sum by an integral we have @xmath293 which can be continued in the lower half plane since it has a single cut along the negative imaginary axis .
we thus restrict to the theory with continuous spectrum of @xmath40 , and propose the simplest partition function @xmath294 which we interpret through analytic continuation , up to an irrelevant phase , as @xmath295 & = & \left(|\chi_{0}|^{2}+|\chi_{1/2}|^{2}\right)^{2}\times \frac{1 } { \eta\bar{\eta}}\int_{-\infty}^{\infty}d\alpha ( q\bar{q})^{\alpha^{2}/2}\label{conj}\ \ .\end{aligned}\ ] ] this object is obviously modular invariant since , from a direct calculation of the integrals , @xmath296 hence compensating the factors coming from the @xmath297 functions in the characters .
we note that the spectrum of conformal weights in the periodic sector is a continuum starting at @xmath298 . the field with @xmath149
does not appear . in conclusion ,
the requirement for our theory to possess a `` physical partition function '' has forced us to let @xmath40 be continuous .
geometrically , this amounts to a decompactification of the time - like circle .
hence we are led to consider the universal cover of @xmath0 so that we can perform an analytical continuation on the number @xmath40 .
one may interpret the prescription that leads to the physical partition function as an effective change of the target space along the lines advocated in @xcite .
let us also stress that , while our arguments were based on the @xmath2 theory , it is clear that a similar reasoning can be carried out for other levels .
the argument leading to @xmath40 being continuous also seems to exclude the smaller theories where part of the complex @xmath94 is dropped , at least for general values of @xmath111 .
we shall see an exception in the case @xmath4 later .
this following section contains some selected applications of our general analysis . in the first subsection we shall compare our results with studies of the continuum limit of the integrable @xmath299 spin chain @xcite
the agreement we find supports a new interpretation of lattice results .
the second subsection is devoted to the @xmath4 theory which was not included above , but we shall see that it shares many of the structures we uncovered throughout the last few sections . in @xcite
an integrable @xmath15 invariant super - spin chain was studied using both analytical and numerical techniques .
its hamiltonian acts on the tensor product @xmath301 where @xmath302 and @xmath303 stand for the representations @xmath304 in our previous terminology .
it was argued that in the continuum limit this chain flows to a @xmath0 wznw model at level @xmath2 . at the time ,
the wznw model on the supergroup @xmath0 had not been constructed and it seems instructive to revisit the issue now on the basis of our improved understanding of the continuum field theory .
we shall see that the suggested identification with the continuum limit of the spin chain can be maintained , but some of the lattice results receive an interesting reinterpretation .
let us begin by reviewing briefly some results on the spectrum of the lattice model .
in @xcite it was found analytically that this spectrum exhibits a unique ground state at @xmath305 , which lies in the single `` true singlet '' of the model , i.e. it is a @xmath15 invariant state that is not part of a larger indecomposable representation .
this ground state corresponds to an extremely degenerate solution of the bethe ansatz equations where all roots collapse to the origin . besides the ground state
, many excited states were also found .
the lowest lying state above the ground state corresponds to a filled sea of some ( non complex conjugate ) string complexes .
the rest of the spectrum is given by excitations obeying the usual pattern of holes and shifts of the sea .
the scaled energies of these excitations over the ground state were found analytically to be @xmath306 & & \ \ + \
c_n(l)\ , \left(\delta n_{+}-\delta n_{-}\right)^{2 } + c_d(l)\ , \left(d_{+}-d_{-}\right)^{2 } \label{gaps}\\[3 mm ] \mbox{where } & & \ \
\rightarrow\ 0 \ \ , \ \
c_d(l)\ \rightarrow \ \infty \ \ \mbox{for } \ \ \ \ \ l\rightarrow\infty\nonumber\end{aligned}\ ] ] here , @xmath307 are quantum numbers characterizing the bethe ansatz solution . in the continuum limit , the quantity ( [ gaps ] ) is expected to converge to @xmath308 ( all weights in the @xmath309 theory ) , as usual .
formula ( [ gaps ] ) indicates an infinite degeneracy of the level @xmath310 ( obtained with @xmath311 say ) in the limit @xmath312 .
numerical studies confirm this behavior : indeed , they show that that an infinite number of levels converges to @xmath310 as @xmath313 increases .
this was already interpreted in @xcite as indicating the existence of a continuum of conformal weights starting at @xmath310 in the thermodynamic limit .
although an analytical study of the asymptotic corrections to ( [ gaps ] ) seems still out of reach , numerical studies in a closely related model suggest that the leading contributions to @xmath314 and @xmath315 can be well fitted by the formulas @xmath316 for large number @xmath317 of lattice sites .
when these leading terms are plugged back into the formula ( [ gaps ] ) for the spectrum of the lattice model , we see that the second line resembles very much the spectrum of a free boson which has been compactified to a circle with radius square of the order @xmath318 . in other words , if we assume that eqs .
( [ gapsfit ] ) are correct , the contribution from the `` antisymmetric sector '' ( i.e. form excitations for which @xmath319 or @xmath320 are non zero ) to the partition function can be estimated as @xmath321 & = & \frac{r } { \sqrt{2}}\frac{1 } { \sqrt{\hbox{im } \tau}\ \eta\bar{\eta}}\ \sum_{m , m ' } \ \exp\left(-\frac{\pi r^{2}|m\tau - m'|^{2 } } { 2\hbox{im } \tau}\right ) \nonumber\\[2 mm ] & \approx & \frac{r } { \sqrt{2 } } \frac{1 } { \sqrt{\hbox{im } \tau}\eta\bar{\eta } } \end{aligned}\ ] ] where @xmath322 is dedekind s eta function , as before .
the divergence is proportional to @xmath269 , i.e. to the size of the target space , as expected .
we conclude that in the lattice model , the contribution from the antisymmetric sector to the partition function multiplies the contribution from the symmetric sector in eq .
( [ gaps ] ) by a term of the order of @xmath323 .
the ground state meanwhile , being a very degenerate bethe ansatz solution , does not come with such an extra factor .
the generating function of levels in the periodic sector will therefore have the form @xmath324\ ] ] where the dots represent excitations from the symmetric sector in ( [ gaps ] ) .
the first conclusion we draw is that the contribution of the continuum completely overrides the one from the discrete state ( as would be the case in any quantum mechanics problem with discrete states and a continuum with delta function normalizable states ) , and that a properly normalized partition function does not see the singlet with @xmath305 .
the resulting object is in good agreement with our conjectured partition function ( [ conj ] ) .
reference @xcite contained various failed attempts to build a conformal field theory containing both the continuum of representations @xmath325 and a single identity field associated with the representation @xmath104 .
given our new insight into the continuum model , the problems to incorporate the singlet state may not come as a complete surprise . although the free field construction at @xmath2 suggests the possibility of smaller theories , the study of modular invariants ( as well as of four point functions , as we mentioned above ) seems to preclude the appearance of the singlet representation on its own - i.e. without being part of a big indecomposable with vanishing super - dimension .
in addition all the states we found in the continuum approach were non normalizable .
both observations lead us to speculate that eq .
( [ conj ] ) represents the full operator content of the continuum limit , and that there is no discrete state associated with a true singlet .
put differently , the new investigation suggests that the true singlet observed on the lattice is an artifact of the regularization and does not belong to the continuum limit .
our new interpretation of the lattice results receives additional support from the very singular nature of the bethe ansatz solution that corresponds to the singlet state .
it would be interesting to check further the decoupling of the true singlet by studying the scaling behavior of matrix elements of lattice regularized current algebra generators .
there is one more potential objection one might raise .
note that in our continuum theory fermionic and bosonic states are perfectly paired so that the witten index of the @xmath0 wznw model is guaranteed to vanish .
meanwhile , for our lattice spin chain on the space @xmath301 one finds an excess by one for the number of bosonic states over the number of fermionic ones .
hence , the witten index is non - zero on the lattice and one would naively expect the same to be true for the continuum limit , in conflict with what we have proposed above . in order to resolve this issue
, we suggest that there exist different spin chains which give rise to the same continuum limit while possessing an excess of fermionic states over bosonic . more concretely , while we do not understand the whole structure yet , we have found that the ground states of integrable chains of the type @xmath326 and @xmath327 scale to conformal weight @xmath149 as well ( in fact the ground state energy is given exactly by @xmath328 where @xmath329 has no finite size correction and is the same for all chains ) , but this time they come in the representation @xmath302 ( resp .
@xmath330 ) .
once we sum over the various lattice models , the balance between bosonic and fermionic states may be restored even before taking the continuum limit .
our investigation above was restricted to integer level @xmath111 .
but as we have mentioned before , these are some fractional values of @xmath111 , in particular @xmath4 , which play an important role for applications .
while we are not prepared to give a systematic account on fractional level theories , we would like to discuss briefly a model with @xmath4 .
our analysis will lead to the remarkable conclusion that the basic structure of this model is essentially the same as for integer @xmath111 , only that there exist several components within the atypical sector , each of them being modeled after @xmath94 . in this case
@xmath4 , the relevant representation theory of the @xmath15 current algebra is particularly simple .
in fact , all relevant representations can be obtained from the vacuum sector @xmath331 through application of spectral flow symmetries .
it is not difficult to show that at @xmath4 the automorphism @xmath332 is inner , i.e. @xmath333 .
this means that application of @xmath332 does not lead to any new representations .
the remaining nontrivial automorphisms are of the form @xmath334 with @xmath335 and @xmath336
. we shall denote the corresponding irreducible representations of the @xmath15 current algebra by @xmath337 by construction , this set closes under fusion .
in fact , the fusion product simply amounts to a composition of the associated automorphism . with the exception of the sectors labeled by @xmath338 , the representations @xmath339 do not contain a highest or lowest weight .
the representation @xmath340 is to be identified with the vacuum representation .
@xmath341 is the only other admissible representation at @xmath4 .
it is generated from the 4-dimensional typical multiplet @xmath261 of ground states with conformal weight @xmath342 .
in addition , there are four more highest / lowest weight weight representations which are erected over the atypical discrete series representations @xmath343 and @xmath344 corresponding to a negative spin @xmath345 .
the choice of the sign in the first argument of the bracket determines on whether the representation is highest @xmath346 or lowest @xmath347 weight .
the subscript , on the other hand corresponds to the two different choices of the parameter @xmath40 that make these representations atypical .
all four representations possess ground states of conformal weight @xmath149 . in all other representations
@xmath339 with @xmath348 , the conformal weight is unbounded from below .
since we can generate every representations from @xmath349 , is suffices to display the character of the vacuum representation , @xmath350 \ \ .\ ] ] we shall explain the origin of this formula in a moment .
characters of all the other representations are obtained from the vacuum character @xmath351 through @xmath352 to derive the above character formula and for the subsequent discussion we note that the @xmath15 current algebra at level @xmath4 possesses a free field representation which employs the same free fields as in the case of the @xmath2 theory , i.e. two free bosonic fields @xmath220 and @xmath221 with space - like and time - like signature , respectively , and a pair of symplectic fermions @xmath353 .
the bosonic @xmath15 currents read @xmath354 e^{-}(z)\ = \ \frac12\ , e^{2i\varphi'(z)}\ , \partial^{2}\eta_{2}(z)\ , \partial\eta_{2}(z ) \ \ \ & , & \ \ \ \ b(z ) \ = \ \frac{i}{2}\ , \partial \phi(z)\
\ \ .\end{aligned}\ ] ] note that , unlike in the case of @xmath2 , the bosonic currents involve the symplectic fermions and the time - like free boson .
for the fermionic currents one finds @xmath355 w^{-}(z ) \ = \
\frac{1}{\sqrt{2}}\ , e^{i(\varphi(z)+\varphi'(z))}\ , \partial\eta_{2}(z ) \ \ \ \ & , & \ \ \ \ w^{+}(z)\ = \ \frac{1}{\sqrt{2}}\ , e^{i(\varphi(z)-\varphi'(z))}\ , \partial\eta_{1}(z ) \ \ . \nonumber\end{aligned}\ ] ] as in the case of the @xmath2 theory , the free field construction determines a consistent model with a @xmath15 current algebra symmetry .
if we do not include the symplectic fermions ( note that once more the currents only involve derivatives ) , but only their derivatives then the state space reads @xmath356 since the spectral flow automorphisms @xmath166 and @xmath189 correspond to multiplication with the fields @xmath357 is is fairly easy to write down at least one field in each sector of the model , @xmath358 the space @xmath359 contains r sector representations only , but it is certainly possible to include the ns sector by adding the image under the spectral flow @xmath212 .
since this works just in the same way as above , we shall not repeat the discussion here .
even though all the representations we are working with are atypical , the state space decomposes into irreducible building blocks .
this is quite different from the structure of the atypical sectors @xmath208 we described above . on the other hand ,
is is very similar to one of the consistent theories with @xmath2 that we described at the end of section 4.2 . in the @xmath2 theory , the singular vectors of the indecomposable block @xmath208 were decoupled by restricting to a theory that contained only derivatives of the fermionic fields .
conversely , the experience from @xmath2 suggest that in the @xmath4 case we may be able to construct a theory with a more complicated atypical sector by including one or both of the symplectic fermion fields @xmath253 and @xmath258 @xcite . we claim that in case we include both fermionic zero modes we end up with an atypical structure that decomposes into four different blocks , each of them being built in the same way as our sector @xmath208 .
we shall present the analysis only for the block that contains the vacuum sector @xmath360 .
the other three sectors are obtained by acting with @xmath361 and @xmath362 .
let us start our discussion with the field @xmath363 .
any action with @xmath364 and @xmath20 will remove one of the two fermionic zero modes and hence @xmath365 sits at the top of a sector @xmath360 .
the action of @xmath366 and @xmath367 takes us from here into a set of fields which all contain a factor @xmath253 .
these fields can be shown to belong to a sector that is isomorphic to @xmath368 .
further application of @xmath369 and @xmath370 bring us to a set of fields that contain only derivatives of fermions .
these form a subrepresentation @xmath331 at the bottom of our atypical representation .
a similar analysis applies if we act with @xmath369 and @xmath370 first .
this time , we descend to @xmath331 via the sector @xmath371 .
continuing along this line of thoughts , one can see that the sectors @xmath372 form the composition series for an indecomposable representation @xmath208 with @xmath373 in place of @xmath374 .
the state space of the maximal theory therefore decomposes into four indecomposable blocks .
once more , there are two intermediate theories , each of which has four saw - blade shaped atypical sectors .
they are obtained if we omit either @xmath253 or @xmath258 ( but not their derivatives , of course ) from the above maximal theory . even though we are not prepared to analyze wznw models for generic fractional levels , it is remarkable that the structure we have first uncovered in our minisuperspace limit , re - appears even for @xmath375 .
it seems very likely that the same is true for a generic choice of the level .
an obvious conclusion of our study is that wznw models on supergroups are interesting examples of logarithmic cfts , much richer than it has been anticipated in earlier works .
gurarie ( see @xcite and references therein ) , for instance , argued that super wznw models with @xmath309 could be considered as made of two `` decoupled '' component theories with opposite values of the central charge , an observation justified in part by the fact that in the gl(1@xmath111 ) wznw model , the stress energy tensor belongs naturally to a four dimensional gl(1@xmath111 ) multiplet in which @xmath376 is diagonalizable , and hence @xmath377 has no `` non trivial logarithmic partner '' .
the su(2@xmath111 ) wznw clearly does not obey any such decoupling .
in fact , restricting to the right moving current algebra as in @xcite we see that the identity field belongs to a projective representation of the zero mode algebra on which the casimir - and hence @xmath376 - is not diagonalizable . applying @xmath378 to this representation
produces a virasoro jordan cell at level @xmath379 and a non trivial logarithmic partner of the stress energy tensor .
this can be seen quite explicitely in the case @xmath2 where , within the free field representation ( and similarly to the case of symplectic fermions ) , the field @xmath380 is a logarithmic partner of @xmath381 note that the whole structure of indecomposables is in fact much more complicated than envisioned in @xcite when the interplay of left and right current algebras is taken into account .
even though the structure of the state space is rather difficult when analyzed with respect to the combined left and right action , it is surprisingly simple once we restrict to either the left or the right action alone .
note that the lie superalgebra @xmath15 has a large number of indecomposables ( see e.g. @xcite ) from which only a very distinguished sub - class does actually occur within the state space of our model . in fact ,
we have seen above that all states ( both in the minisuperspace theory and the full field theory ) transform according to the so - called projective representations of @xmath15 , i.e. either in typicals and projective covers of atypicals .
this is not to say , however , that other representations of @xmath15 have no relevance for sigma models on supergroups .
in addition to the left and right regular representation there is yet one more important symmetry that arises from the adjoint action of @xmath15 on the state space . with respect to the latter
, states can transform in other indecomposables .
the underlying mathematical structure turns out to be quite intriguing and will be described elsewhere .
since the adjoint action is left unbroken by maximally symmetric boundary conditions , the resulting decomposition should have applications , in particular to the study of boundary conditions for sigma models on supergroups . as a final comment
let us point out one generic feature we have encountered in both @xmath14 and @xmath0 , namely that the contribution of the indecomposable sector @xmath94 simply makes up for the subtractions in the atypical sectors of the theory , so that the partition function sees only contributions from kac modules , and has a simple factorized form .
this behavior is sufficient for a modular invariant partition function but it is not necessary .
the potential existence of different versions of the theory where only parts of the complex @xmath94 appear , requires more study .
we note that some hints in this direction are provided by the study of four point functions . in the case
@xmath2 , the four point function of the fields in the @xmath261 representation has been studied in detail .
it turns out that the kz equations factorize , and that it is possible to decouple one conformal block .
two blocks remain , leading to logarithmic dependence , and indicating that the identity field remains part of an indecomposable representation .
this suggests that the smallest theory , where the complex @xmath94 is reduced to an infinite sum of irreducibles , can not appear in the @xmath15 wznw model . in the case
@xmath4 meanwhile it is possible to decouple two conformal blocks , leaving only the identity field , and indicating that the smallest theory does make sense this time - a feature consistent with the free field representation and the modular invariant .
the @xmath15 wznw model at fractional level and the explicit construction of consistent theories with a truncated atypical sector certainly deserve a more systematic investigation .
* acknowledgments : * we thank fabian essler , gerhard gtz , thomas quella and anne taormina for interesting conversations . v.s . would like to thank the spht for the warm hospitality during several stays .
this work was partially supported by the eu research training network grants `` euclid '' , contract number hprn - ct-2002 - 00325 and `` forcesuniverse '' , contract number mrtn - ct-2004 - 005104 .
in this appendix we would like to prove the decomposition formula for the right regular representation .
we shall use the same notations that were introduced in section 2.2 . in order to analyze the decomposition of the space of functions under the right regular action of @xmath15
, we shall first study its restriction to the lie sub - superalgebra @xmath88 .
more precisely , we shall make use of the following embedding @xmath382 the main technical lemma of this section implies that under the action of @xmath88 , the space @xmath250 of functions on the supergroup @xmath0 decomposes into projectives only .
* lemma : * _ under the action of @xmath383 of the generators @xmath384 @xmath88 , the space @xmath250 of functions of @xmath0 decomposes according to @xmath385 here , @xmath386 is a direct sum of typical @xmath88 representations and @xmath387 denotes the projective cover of the atypical irreducible @xmath388 . _ before we prove this statement , let us formulate two consequences for the right regular representation of @xmath15 . to begin with , let us recall from @xcite that an @xmath15 representation @xmath389 descends on a projective representation of the embedded @xmath88 algebra if and only if @xmath389 is projective .
our lemma claims that the @xmath88 action on h contains only projectives .
hence , the same must be true for the right regular action of @xmath15 .
proof of lemma : for the proof it will be useful to introduce the following odd functions @xmath390 it is not difficult to see that the space @xmath250 is spanned by functions of the form @xmath391 where @xmath392 is an arbitrary element in the algebra generated by the arguments .
it is very easy to describe explicitly the space of functions which are organized in atypicals of @xmath88 .
the latter is characterized by the vanishing of @xmath393 , @xmath394 we can easily solve the equation for @xmath395 and describe the space @xmath396 explicitly .
in fact , it is spanned by the functions @xmath397 on the subspace @xmath398 the other generators of @xmath88 simplify to @xmath399 r_{\psi^+ } \phi \ = \ -i \partial_+ \phi\ \ & , & \ \
r_{\psi^- } \phi \ = \ i e^{-iz/2 } \ , d^{1/2}_{\nu ( - 1/2 ) } ( g)\ \bar \partial_{-\nu } - i \eta_- r^0_{e^- } \ .\end{aligned}\ ] ] for all @xmath400 .
the representation of @xmath88 can be restricted to the space @xmath401 of all elements @xmath400 such that @xmath402
. a short look on the action of the @xmath88 generators reveals that @xmath403 similarly , we see that @xmath404 since all representations are projective we conclude that @xmath405 this concludes the proof of our lemma .
d. bernard , _ ( perturbed )
conformal field theory applied to 2d disordered systems : an introduction _ , http://arxiv.org/abs/hep-th/9509137[hep-th/9509137 ] .
m. j. bhaseen , j. s. caux , i. i. kogan and a. m. tsvelik , _ disordered dirac fermions : the marriage of three different approaches _ , nucl .
* b618 * ( 2001 ) 465499 [ http://arxiv.org/abs/cond-mat/0012240 [ cond - mat/0012240 ] ] .
z. maassarani and d. serban , _ non - unitary conformal field theory and logarithmic operators for disordered systems _ , nucl . phys .
* b489 * ( 1997 ) 603625 [ http://arxiv.org/abs/hep-th/9605062 [ hep - th/9605062 ] ] .
a. w. w. ludwig , _ a free field representation of the @xmath406 current algebra at level @xmath407 , and dirac fermions in a random @xmath279 gauge potential _ , http://arxiv.org/abs/cond-mat/0012189 [ cond - mat/0012189 ] .
p. bowcock , m. hayes and a. taormina , _ characters of admissible representations of the affine superalgebra @xmath408 _ , nucl . phys .
* b510 * ( 1998 ) 739764 [ http://arxiv.org/abs/hep-th/9705234 [ hep - th/9705234 ] ] .
l. rozansky and h. saleur , _ quantum field theory for the multivariable alexander - conway polynomial _ , nucl
* b376 * ( 1992 ) 461509 .
v. schomerus and h. saleur , _ the @xmath409 wzw model : from supergeometry to logarithmic cft _ , http://arxiv.org/abs/hep-th/0510032[hep-th/0510032 ] .
g. gotz , t. quella and v. schomerus , _ the wznw model on psu(1,1@xmath112 ) _ , http://arxiv.org/abs/hep-th/0610070[hep-th/0610070 ] .
f. h. l. essler , h. frahm and h. saleur , _ continuum limit of the integrable @xmath408 @xmath300 superspin chain _ , nucl .
* b712 * ( 2005 ) 513572 [ http://arxiv.org/abs/cond-mat/0501197 [ cond - mat/0501197 ] ] .
l. frappat , p. sorba and a. sciarrino , dictionary on lie algebras and superalgebras .
academic press inc .
, san diego , ca , 2000 . extended and corrected version of the e - print [ hep - th/9607161 ] .
g. gtz , t. quella and v. schomerus , _ representation theory of @xmath408 _ , http://arxiv.org/abs/hep-th/0504234 [ hep - th/0504234 ]
. v. g. kac , _ lie superalgebras _ , adv . math .
* 26 * ( 1977 ) 896 .
a. hffmann , _ on representations of supercoalgebras _
, j. phys . *
a27 * ( 1994 ) 64216432 [ http://arxiv.org/abs/hep-th/9403100 [ hep - th/9403100 ] ] .
p. bowcock and a. taormina , _ representation theory of the affine lie superalgebra @xmath410 at fractional level _
, commun . math .
* 185 * ( 1997 ) 467493 [ http://arxiv.org/abs/hep-th/9605220 [ hep - th/9605220 ] ] .
m. hayes and a. taormina , _
admissible @xmath411 characters and parafermions _ , nucl .
phys . * b529 * ( 1998 ) 588610 [ http://arxiv.org/abs/hep-th/9803022[hep-th/9803022 ] ] .
a. m. semikhatov and a. taormina , _ twists and singular vectors in @xmath412 representations _ , theor .
* 128 * ( 2001 ) 12361251 [ http://arxiv.org/abs/hep-th/0311166 [ hep - th/0311166 ] ] .
a. m. semikhatov , a. taormina and i. y. tipunin , _ higher level appell functions , modular transformations , and characters _ , comm . math .
* 255 * ( 2005 ) 469512 [ http://arxiv.org/abs/math.qa/0311314 [ math.qa/0311314 ] ] .
l. rozansky and h. saleur , _ s and t matrices for the @xmath413 wzw model : application to surgery and three manifolds invariants based on the alexander - conway polynomial _ , nucl . phys . *
b389 * ( 1993 ) 365423 [ http://arxiv.org/abs/hep-th/9203069[hep-th/9203069 ] ] .
|
motivated by a careful analysis of the laplacian on the supergroup @xmath0 we formulate a proposal for the state space of the @xmath0 wznw model .
we then use properties of @xmath1 characters to compute the partition function of the theory . in the special case of level @xmath2 the latter
is found to agree with the properly regularized partition function for the continuum limit of the integrable @xmath3 super - spin chain .
some general conclusions applicable to other wznw models ( in particular the case @xmath4 ) are also drawn .
|
g.h.b gratefully acknowledges funding from the royal society . n.s.b acknowledges trinity college , cambridge for funding
this work has been supported by the epsrc under grant no .
ep / j003867/1 .
39ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , ) pp . @noop
* * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( )
|
we present an approach to the calculation of arbitrary spectral , thermal and excited state properties within the full configuration interaction quantum monte carlo framework .
this is achieved via an unbiased projection of the hamiltonian eigenvalue problem into a space of stochastically sampled krylov vectors , thus enabling the calculation of real - frequency spectral and thermal properties and avoiding explicit analytic continuation .
we use this approach to calculate temperature - dependent properties and one- and two - body spectral functions for various hubbard models , as well as isolated excited states in _ ab initio _ systems .
quantum monte carlo ( qmc ) in its various guises , is undoubtedly one of the most important approaches for accurate elucidation of properties for correlated systems@xcite .
however , these successes have focused primarily on the ground state energy and observables which commute with the hamiltonian .
critical importance for a deeper understanding of correlated systems comes from _ dynamic _ correlation functions and spectral quantities .
these mirror how we perceive our environment , namely by perturbing a system and measuring its response the basis of nearly all spectroscopic and experimental approaches .
this gives us direct insight into optical , magnetic and other beyond - ground - state properties , and allow for direct comparison to experimental results .
direct access to dynamic properties is a persistent difficulty for qmc approaches in general . while in the absence of a sign problem
, unbiased imaginary - time spectra can be obtained@xcite , the analytic continuation to physical , real - frequency functions is notoriously ill - conditioned and can lead to artefacts and smoothing of features.@xcite for more general fermionic systems , higher temperatures must be simulated to alleviate the sign problem@xcite , while nodal constraints bias towards a particular solution and are difficult to extend to spectra@xcite .
alternatively , projections into effective hamiltonians have been able to obtain a few low - energy states , but again these are isolated states rather than practical approaches for thermal or spectral quantities@xcite , while a modification of the propagator can lead to debilitating timestep issues@xcite . here
, we present a new qmc approach for computing dynamic correlation functions , temperature - dependent quantities and isolated excited states for correlated quantum systems , even in the presence of a sign - problem .
these correlation functions are unbiased in the limit of large averaging , and exact in the limit of large walker number .
this is achieved by extending the recently developed full configuration interaction quantum monte carlo ( fciqmc ) method@xcite , by combining it with ideas from the dynamical and finite - temperature lanczos ( ftlm ) methods.@xcite the key advantage of the approach is that it avoids any explicit storage over the full hilbert space , instead only storing occupied states in the discretized wavefunction at each snapshot .
this allows for sparsity in the wavefunction to be exploited to minimize memory bottlenecks , which are a primary limitation in conventional approaches which require explicit storage over the space@xcite .
the result is a qmc method which although weakly exponentially scaling , in common with the ground state fciqmc approach , can allow for systems to be treated well outside that possible by conventional means , and retains many of the important features of the parent method@xcite .
these include a cancellation algorithm to ameliorate the sign problem , an absence of time - step error and large - scale parallelism .
an arbitrary dynamic correlation function is defined as @xmath0 where @xmath1 represents the hamiltonian of the system , @xmath2 is the ground state wavefunction and energy , @xmath3 is a small broadening parameter , and @xmath4 and @xmath5 are arbitrary operators which define the perturbation and observed quantity in the correlation function . in the case of these operators being single annihilation and creation operators
one obtains the single - particle green function , defining the system bandstructure and density of states .
the aim of our method is to stochastically obtain a projection of the hamiltonian from the complete hilbert space to an effective , reduced dimensionality space , such that it spans the degrees of freedom required to accurately describe the desired spectral or thermal quantity . in this work
, we use a set of stochastically sampled wavefunctions from a fciqmc calculation to define this transformation .
if the initial state of the calculation is a stochastic representation of the wavefunction @xmath6 , then propagation from this state@xcite to the ground state will in principle span all states required to represent the expression in eq .
[ eq : correlation_function ] , equivalent to the space of ground state and all imaginary - time response vectors .
once the hamiltonian is projected into this space , it can be exactly diagonalized , and the desired correlation function of eq .
[ eq : correlation_function ] directly constructed in this eigenbasis the lehmann representation . for thermal quantities
the approach is analogous , with the initial vector taken from the infinite - temperature distribution .
similar themes have been explored within continuum qmc , but applied to accelerate convergence for ground state properties@xcite . _
method:- _ an fciqmc iteration consists of stochastically applying a projection operator , @xmath7 , to a walker distribution , denoted at iteration @xmath8 by @xmath9 , such that exact projection is achieved on average , whose distributions we denote as @xmath10@xcite .
the aim is to stochastically sample the krylov subspace @xmath11 . in projector qmc
approaches one samples from the large @xmath12 limit of this subspace , which converges to the ground state .
however , to obtain finite - temperature and dynamic quantities , the aim is now to stochastically project the hamiltonian into the whole sampled krylov subspace , which represents an efficient span of all states of interest , provided that @xmath13 is chosen appropriately . by averaging the fciqmc walker amplitudes ,
the results of an exact propagation are rigorously approached for expectation values which depend linearly on the wavefunction@xcite . in this work ,
quadratic quantities are required , but now @xmath14 \neq \boldsymbol{\psi}_i^{\dagger } \boldsymbol{\psi}_j$ ] , due to correlations between walker amplitudes , where @xmath15 $ ] denotes the expectation value .
to compute these , two independent replica sets of walkers are propagated simultaneously ( indexed via superscripts ) , such that the amplitudes are uncorrelated between them@xcite , allowing for unbiased estimates of @xmath16 as @xmath17 $ ] or @xmath18 $ ] .
this approach for static correlation functions has been found to scale without difficulty within fciqmc@xcite . at
selected iterations in an fciqmc calculation the walker distribution is stored@xcite , and the overlap ( @xmath19 ) and hamiltonian ( @xmath20 ) matrices between these subspace vectors calculated as @xmath21 whilst the overlap matrix estimate is trivial , calculating the @xmath20 matrix exactly is expensive , and so instead it is stochastically sampled in the same manner as spawning steps in fciqmc@xcite .
thus , a simulation provides an estimate of the overlap matrix and the projected hamiltonian in the basis of krylov vectors chosen , and so we denote the method _ krylov projected _ ( kp)-fciqmc .
averaging these quantities over independent simulations can reduce errors in an unbiased manner , resulting in a generalised eigenvalue problem for the projected hamiltonian .
this can be solved by standard techniques ( see supplementary material )
. many of the eigenvalues of @xmath19 will be very small ( or even negative within stochastic errors ) , since the sampled space becomes increasingly linearly dependent with continued propagation .
we therefore discard these vectors of @xmath19 without substantial loss of information .
we refer to the eigenvectors which are kept as vectors .
we note that although the estimates of @xmath20 and @xmath19 are unbiased , the final eigenvalues will not be because eigenvalues are non - linear functions of these matrices .
however , this bias can be systematically reduced with further averaging of @xmath20 and @xmath19 . for exact propagation with @xmath22 , our approach will yield results identical to the lanczos method . however , because the method exploits sparsity via a stochastic representation of the wavefunctions , large calculations can often use significantly less memory than an equivalent lanczos calculation , as has been the case for ground - state fciqmc .
although our approach is in theory systematically improvable to exactness for the entire frequency range , in practice this becomes increasingly difficult for higher energy excitations .
this is because high - energy excitations have a small component in the krylov vectors , which decreases exponentially with imaginary time .
this renders them particularly difficult to sample and susceptible to stochastic error in the sampled matrices . despite this limitation
, the approach can nevertheless be expected to obtain near - exact spectra for low - energy excitations in systems out of reach of traditional dynamical lanczos approaches .
_ finite - temperature:- _ we assess the method with the half - filled hubbard model ( defined in supplementary information)@xcite . within the ftlm ,
thermal expectation values are computed via @xmath23 where @xmath24 labels a state in the @xmath25-dimensional hilbert space , and @xmath8 labels the @xmath26 states of an eigensystem @xmath27 resulting from a lanczos subspace with initial state @xmath24 .
thus , by performing @xmath25 lanczos calculations consisting of @xmath28 applications of @xmath1 each , one can obtain thermal quantities which are correct to order @xmath29 .
@xmath25 can be very large for systems of interest and so in practice one starts from a much smaller number of states , @xmath30 , chosen as a random linear combination of all basis states , @xmath31 .
this turns out to converge quickly with @xmath32 , particularly at high temperatures@xcite . in our stochastic approach
the initial random vectors are created by distributing a given number of walkers randomly throughout the hilbert space with coefficients @xmath33 .
these initial states represent stochastic snapshots of the high - temperature limit which is exactly reproduced in the limit of large @xmath32 . as an initial test , figure [ fig : finite_t_12 ] presents the temperature - dependent energy , @xmath34 , in the one - dimensional 12-site hubbard model at @xmath35 . including all symmetry sectors the hilbert space dimension is @xmath36 , with the largest symmetry sector containing @xmath37 determinants . however , the system is significantly undersampled with only @xmath38 walkers used throughout , with the projected hamiltonian and overlap matrices averaged over @xmath39 calculations for each initial vector , @xmath40 .
all symmetry sectors were obtained in one calculation , rather than symmetry - blocking eq .
[ eq : thermal_approx ] , resulting in a choice of @xmath41 , while the number of krylov vectors used was @xmath42 , with @xmath43 vectors kept to form the final space .
the results were found not to change significantly by including more vectors . at high temperatures results
are calculated with great accuracy .
this is easily understood because the quantity calculated at @xmath44 , @xmath45 , is rigorously equal to @xmath46 , and therefore the quality is mainly dependent on the sampling of the initial vectors ( and not on the error of individual eigenvalue estimates ) . at low temperatures
the results are dominated by the ground state , which has a large component in the sampled krylov vectors and so is accurately calculated by kp - fciqmc . however , at intermediate temperatures the errors are larger .
the most significant source of error is in replacing an exact trace over @xmath47 by an approximate one over @xmath48 in eq .
[ eq : thermal_approx ] . for the 12-site 1d hubbard model at @xmath35 sampled with @xmath49 walkers , with comparison ftlm .
error bars show standard deviation ( note not standard error ) over @xmath39 independent calculations to demonstrate the spread of results .
high and low temperature results are almost exact , whilst at intermediate temperatures , the variance in the stochastic sampling as well as systematic errors ( such as from the non - linear diagonalization step , and finite @xmath32 ) increases the variation between runs .
simulation parameters were @xmath50 , @xmath51 , and a deterministic space of double excitations@xcite . ] in figure [ fig : finite_t_18 ] , @xmath34 for the two - dimensional 18-site hubbard model at @xmath35 is presented .
including all symmetry sectors the hilbert space dimension is @xmath52 , with the largest symmetry sector containing @xmath53 determinants .
again , the space was undersampled , with @xmath54 walkers used throughout , with @xmath55 and @xmath42 , of which @xmath56 vectors are kept . since ftlm was unfeasible ,
also plotted is a highly - accurate ground - state fciqmc energy for comparison .
a complete calculation took around @xmath57 core hours .
we find again that the high - temperature results have only a small variation between repeated calculations and we have a high degree of confidence here . at lower temperatures the confidence in the results
is reduced , with possible systematic errors including initiator error , bias in the eigenvalue estimates and an insufficient choice of @xmath32 .
for the 18-site 2d hubbard model at @xmath35 , with ground - state fciqmc energy for comparison .
10 independent simulations were used to create the standard deviations shown as error bars .
simulation parameters were @xmath50 , @xmath51 , and a deterministic space of double excitations . ]
_ dynamical correlation functions:- _ to demonstrate the ability of kp - fciqmc to calculate dynamical quantities , we first consider the following zero - temperature @xmath58-resolved single particle green function , defined from eq .
[ eq : correlation_function ] with @xmath59 .
the corresponding spectral function , @xmath60 $ ] , defines the bandstructure of the material . the initial walker distribution is given by the perturbed ground state , @xmath61 , where @xmath62 is obtained from a prior ground - state fciqmc calculation .
this starting wavefunction ensures that on average the component of a particular eigenstate in any imaginary - time snapshot is proportional to its transition amplitude in the correlation function .
this approach works particularly well for spectra dominated by a small number of states with large transition amplitudes . because the transformation to the basis introduces large errors if many states are kept ( due to small overlap eigenvalues )
, we typically limit the number of vectors to between @xmath39 and @xmath63 , which limits the resolution of the spectrum .
furthermore , high - energy states die away rapidly in the krylov vectors and so there tends to be significant stochastic errors associated with the calculation of such states .
although this limits the accuracy of kp - fciqmc over a large energy range , we find that the method is capable of producing accurate spectra in the critical low - energy region and can accurately capture important features such as bandgaps .
figure [ fig:1_particle_spectra](a ) presents @xmath64 for the 14-site hubbard model at @xmath65 with @xmath66 walkers , with @xmath19 and @xmath20 averaged over @xmath39 repeats .
@xmath67 krylov vectors were sampled and @xmath39 vectors were retained .
a complete calculation for a given @xmath58-sector typically took only @xmath68 core hours .
the results are compared to highly - accurate dynamical lanczos results , using @xmath69 lanczos vectors .
figure [ fig:1_particle_spectra](b ) presents the local density of states , computed from the results in ( a ) via @xmath70 .
the kp - fciqmc results give high accuracy for low - energy features , with sum rules and causality conditions exactly fulfilled .
errors on individual poles can be estimated by repeating results . by comparing eigenvalue estimates from @xmath39 independent calculations ,
the bandgap was estimated as @xmath71 compared to the exact value of @xmath72 . from @xmath73 ( bottom ) to @xmath74 ( top ) for the 1d 14-site hubbard model at @xmath65 , compared to dynamical lanczos .
poles coming from the ground state or low - lying excited states with large transition amplitudes are captured accurately .
( b ) the local density of states .
the low - energy results are reproduced accurately by kp - fciqmc while the qualitative behaviour is captured at high energies .
simulation parameters were @xmath50 , @xmath75 , and a deterministic space of @xmath76 determinants@xcite . ]
we also consider the @xmath77-wave pair - pair dynamic correlation function , a two - body response property of significant relevance in the detection of superconducting quasiparticles .
@xmath4 is defined by the singlet pairing operator , @xmath78 , with @xmath79 . in fig .
[ fig:2_particle_spectra ] we present results for this pairing spectrum ( @xmath80 ) for the 10-site hubbard model at @xmath35 , by computing all @xmath58-space contributions .
the number of walkers was typically between @xmath81 and @xmath82 .
the initiator adaptation was not applied because the walker population is above the plateau@xcite height for this system .
no averaging of @xmath20 or @xmath19 over repeated calculations was performed .
once again , it is found that low - energy features are calculated accurately , but the quality decreases for higher energy regions of the spectrum .
calculated for the 10-site hubbard model with @xmath35 , and compared to near - exact dynamical lanczos .
inset shows integrated weight , @xmath83 .
simulation parameters were @xmath50 , with a deterministic space of double excitations . ] _ isolated excited states:- _ as a further application to larger - scale _ ab initio _
systems , we consider the all - electron ground and first excited state of neon , in aug - cc - pvdz and aug - cc - pvtz basis sets@xcite .
we work in spaces where @xmath84 is constrained to be zero , but the total spin , @xmath85 , is not .
the @xmath86 and @xmath87 states are therefore both contained within the same symmetry sector , and the spin - gap can be directly targeted with kp - fciqmc .
the determinant space sizes with these two basis sets are @xmath88 and @xmath89 , respectively . in order to ensure large components of the desired states in the sampled krylov vectors ,
the initial wavefunction was created from a linear combination of trial estimates of the ground and first excited states at the inexpensive cisd level of theory .
kp - fciqmc results are presented in table [ tab : ne_results ] , with density matrix renormalisation group ( dmrg ) results for comparison .
dmrg is a highly accurate algorithm , which can also be extended to thermal and spectral quantities , and so is a suitable choice for comparison@xcite .
kp - fciqmc results and errors were estimated by averaging over @xmath39 independent calculations .
for the aug - cc - pvdz results , @xmath90 walkers were used , while @xmath91 walkers were used for the aug - cc - pvtz basis , with each calculation taking around @xmath69 core hours for this larger basis .
calculations used 35 krylov vectors , with 10 vectors retained , providing excellent agreement with dmrg .
.results for the ground ( @xmath86 ) and first excited ( @xmath87 ) states of the ne atom ( @xmath92 ) , comparing kp - fciqmc with dmrg ( using m=500 spin - adapted renormalised states for the larger basis)@xcite .
@xmath93 , @xmath94 and a deterministic space of single and double excitations .
[ cols="^,^,^,^,^ " , ] _ conclusion:- _ we have presented a novel approach to the calculation of excited state , spectral and thermal properties within the fciqmc framework . in this approach the full hamiltonian eigenvalue problem is projected into a stochastically sampled krylov subspace , thus allowing finite - temperature and dynamical quantities to be calculated .
since the method exploits sparsity in the sampled wavefunctions , the stochastic dynamic avoids storing krylov vectors in their entirety , rendering the approach scalable to systems sizes outside the range of the lanczos method , although in practice this is likely to be restricted if attempting to probe high frequency spectral features .
|
the aharonov - bohm effect @xcite is usually investigated in terms of abelian gauge theories , _ e.g. _ electromagnetism formulated via maxwell s equations .
further , the electromagnetic fields considered in the canonical aharonov - bohm effect are static fields . for the vector / magnetic aharonov - bohm effect
, this means a static vector potential , @xmath0 , which then translates to a static magnetic field via @xmath1 . in this article , we wish to consider the aharonov - bohm effect in the presence of _ time - dependent , non - abelian _ gauge fields .
there has been some prior work on the aharonov - bohm effect in the presence of time - independent , non - abelian fields @xcite . unlike the abelian case of electromagnetism
, it may not be possible to observe the aharonov - bohm effect for static , non - abelian fields .
for the strong interaction , with the non - abelian @xmath2 gauge group , the theory is thought to exhibit confinement .
thus , it is not clear that one could arrange a non - abelian flux tube that one could control , as is the case with electromagnetism .
further , since the color charges are always confined , one can not send isolated , unconfined color charges around hypothetical non - abelian magnetic flux tubes , unlike the abelian case of electromagnetism , where one can send isolated , unconfined electric charges around abelian magnetic flux tubes . despite these experimental obstacles , in this paper , we study the _ time - dependent _ aharonov - bohm effect for non - abelian fields .
the first reason is that the aharonov - bohm effect is an important consequence of combining gauge theories with quantum mechanics , and so , it is of interest to see how replacing an abelian gauge theory by a non - abelian gauge theory changes ( if at all ) the aharonov - bohm effect .
second , the time - dependent aharonov - bohm effect has not been investigated to any great degree , even for abelian gauge theories . in the two papers @xcite , the time - dependent aharonov - bohm effect for abelian fields was investigated and a cancellation was found between the usual magnetic aharonov - bohm phase shift and the additional phase shift coming from the electric field , which inevitably occurs for time varying magnetic fields . in this paper , we want to see if a similar cancellation occurs between the non - abelian magnetic and electric fields . for our time - dependent non - abelian field configurations , we take the non - abelian plane wave solutions of coleman @xcite and the time - dependent wu - yang monopole solution @xcite .
both solutions satisfy the yang - mills field equations for non - abelian gauge fields of the form @xmath3 where @xmath4 is the coupling constant and @xmath5 are the group structure constants .
@xmath6 is the non - abelian vector potential and the field strength tensor , @xmath7 is given by @xmath8 at this point in this paper , we will set @xmath9 .
for the coleman non - abelian plane wave solutions , we find the same cancellation between the non - abelian magnetic and electric phase shifts that occur in the abelian case .
we also find the same cancellation for the time - dependent wu - yang monopole solution .
we conclude by giving some remarks as to the similarities between these two time - dependent non - abelian solutions and the time - dependent abelian case .
we further postulate that the cancellation between the non - abelian magnetic and electric phase shifts found for the two specific solutions investigated here may be a feature of more general time - dependent non - abelian solutions .
we begin by reviewing the properties of the two coleman plane wave solutions . the non - abelian vector potential for the first/@xmath10 solution is @xmath11 where @xmath12 , in light front coordinates , _ i.e. _ @xmath13 ( the speed of light will be set to unity , @xmath14 ) .
the @xmath10 in the superscript labels this as the light front form of the solution traveling in the negative @xmath15 direction .
the second solution gives waves traveling in the positive @xmath15 direction .
the second solution is only a function of the light front coordinate @xmath16 , @xmath17 again , the superscript @xmath18 indicates this is the light front form of the solution traveling in the positive @xmath15 direction .
the ansatz functions , @xmath19 and @xmath20 , are functions of @xmath21 but are otherwise arbitrary .
first , plugging @xmath22 from into , the field strength tensor for the light front form of the @xmath10 solution becomes @xmath23 the non - zero components here are @xmath24 and @xmath25 .
next plugging @xmath26 from into , the field strength tensor for the light front form of the @xmath18 solution becomes @xmath27 the non - zero components here are @xmath28 and @xmath29 .
it is interesting to note that neither of the forms of @xmath30 depend on the ansatz function @xmath31 . also , the non - abelian term , @xmath32 , is always zero for both the @xmath10 and @xmath18 solutions .
thus , the solutions are very weakly " non - abelian since this prototypical non - abelian / non - linear term is absent .
coleman noted that the @xmath10 solution given by , in terms of the vector potential , and by , in terms of the fields , provides an example of the wu - yang ambiguity @xcite that in non - abelian theories , @xmath30 does not contain all the gauge invariant information about a particular solution , as is the case in abelian gauge theories .
specifically , in terms of the vector potential , one has the quantity @xmath33 ~,\ ] ] where @xmath34 indicates path ordering and the @xmath35 are the lie algebra elements .
the expression in is identified as the non - abelian aharonov - bohm phase factor in terms of the potentials .
as well , the expression in is the wilson loop for gauge theories @xcite .
we now consider a unit loop in the @xmath36 plane ( _ i.e. _ @xmath37 plane ) and starting from @xmath38 and @xmath39 and going in the direction given in figure [ fig1 ] . for this loop
the integral in the exponent in becomes @xmath40 d
\zeta ^+ ~.\end{aligned}\ ] ] the second and fourth integrals ( _ i.e. _ @xmath41 and @xmath42 ) are zero since @xmath43 for the @xmath10 solution in . for the third
integral , we get @xmath44 d \zeta ^+$ ] , since for this leg , @xmath45 and @xmath46 and one is going backward along the @xmath47 direction , while for the first integral , we get @xmath48 since @xmath49 .
note that @xmath48 and @xmath50 do not cancel , due to the path ordering in .
taking the path ordering into account and combining and one gets @xmath51 d \zeta
^+ \right ) \right ] ~.\end{aligned}\ ] ] from it is evident that there is no cancellation of the @xmath52 functions due to the non - trivial commutation relationship of @xmath35 and @xmath53 .
the above result is equivalent to the result given in equation ( 8) of @xcite .
now , the field strength version of ( and the field strength version of the aharonov - bohm phase for non - abelian theories @xcite ) is @xmath54 ~,\ ] ] where @xmath55 is the area and @xmath56 means area " ordering @xcite . for the unit area spanning the unit loop in the @xmath37 plane , the differential
area " is given by @xmath57 .
the reason that @xmath58 has @xmath59 rather than @xmath60 is that the area " vector should be perpendicular to the surface spanned by @xmath37 , and it is @xmath61 , not @xmath62 , which is perpendicular to @xmath58 , as seen in figure [ fig1 ] .
similarly , @xmath47 is perpendicular to @xmath63 .
note that for this unit square the area ordering denoted by @xmath56 is simple since there is only one area vector . with only a single area vector the issue of ordering
does not arise . as a result of the above discussion
, the integral in the exponential in for the unit loop in the @xmath36 plane becomes @xmath64 this integral is zero since @xmath65 , which is due to the @xmath47 functional dependence of @xmath66 , but there is a sign difference between the @xmath67 integration and @xmath68 integration . for a unit loop in the @xmath69 plane ( _ i.e. _ @xmath70 plane ) , we would find a similar result as in , except for the replacement @xmath71 .
since the area for this loop is @xmath72 , we again get @xmath73 , which then cancels because of the @xmath74 in the unit area element .
the same result also holds for the @xmath18 solution from . in this case , the unit loop is in the @xmath75 plane ( _ i.e. _ @xmath76 plane ) or @xmath77 plane ( _ i.e. _ @xmath78 plane ) .
the perpendicular areas in this case will be @xmath79 and @xmath80 .
now , the relevant integrals will be @xmath81 and @xmath82 , so that one again gets zero for the area integrals like @xmath83 and @xmath84 .
this vanishing of the area " integral of the non - abelian field strengths , @xmath85 , occurs for both of these time - dependent , non - abelian solutions we examined .
although we do not have a general proof , we conjecture that this cancellation will occur generally for time - dependent , non - abelian solutions . an important point to note
is that it is only the time - dependent part of @xmath86 which is conjectured to vanish the static parts of the fields still give the usual non - abelian aharonov - bohm phase . in the light front coordinates ,
the split between the non - abelian electric and magnetic field components is not so straight forward .
just as the @xmath87 coordinates are mixtures of space and time coordinates , so too @xmath88 are mixtures of the non - abelian electric and magnetic field components .
for this reason , we now perform the above analysis of @xmath89 and @xmath90 in cartesian coordinates @xmath91 .
we do this for the @xmath10 solution of but the same analysis applies to the @xmath18 solution of .
the @xmath92 components of the @xmath10 solution are the same in light front and cartesian coordinates . the time and @xmath15 components
are obtained via @xmath93 and @xmath94 , which is the same way one transforms between @xmath95 and @xmath96 . also , note that the superscript @xmath10 labels the solution while the subscripts label the @xmath97 components of this solution .
thus , in cartesian coordinates , the @xmath10 solution is @xmath98 where now @xmath99 rather than @xmath100 .
the cartesian field strength tensor following from is @xmath101 where again the indices @xmath102 , @xmath103 are given by @xmath104 rather than @xmath105 . as a check , it is easy to verify that and satisfy the yang - mills field equations . in the form
, the split between the non - abelian electric and magnetic components is obvious
the first row and column are the electric components and the @xmath106 sub matrix below and to the right of the first row and first column are the magnetic components .
thus , in cartesian coordinates , the loop integral in the exponent in becomes @xmath107 ( dt ~ { \rm or } ~ dz ) ~.\end{aligned}\ ] ] again , the paths @xmath108 and @xmath109 do not contribute since the @xmath110-component of the @xmath10 solution in cartesian coordinates is zero , @xmath111 , as was the case for the solution in light front coordinates .
paths @xmath112 and @xmath113 have equal components in the @xmath114 and @xmath15 directions and thus , pick out the gauge field components @xmath115 and @xmath116 . along path @xmath112 ,
@xmath49 , so only @xmath52 appears ; while for path @xmath113 , @xmath45 and @xmath46 , so now both @xmath52 and @xmath66 appear . because of the dependence of @xmath117 on @xmath118 , the @xmath67 and @xmath68 integrals of these functions are the same .
the end result is that the loop integral of the gauge field gives the same results in light front coordinates , , and in cartesian coordinates , .
thus the phase shift calulated by the loop integral of the gauge fields equation is the same in both cartesian and light front coordinates .
next , we calculate the surface area integral , @xmath86 , for the unit area from figure [ fig1 ] .
taking into account the anti - symmetry of @xmath119 and @xmath55 under exchange of the indices , the @xmath10 solution , in cartesian coordinates , and with the surface in the @xmath120 plane from figure [ fig1 ] , we find that the surface area integral of the field strength tensor is @xmath121 there are two integrals @xmath122 and @xmath123 which are absent from since for the unit area in the @xmath120 plane , @xmath124 .
if we take a surface in the @xmath125 plane , then these two integrals would appear and instead , @xmath126 and @xmath127 from would vanish .
the two integrals in cancel since @xmath65 , due to the fact that @xmath66 depends on @xmath118 .
similarly , the two integrals for the @xmath70 plane cancel since @xmath73 , due to the fact that @xmath128 depends on @xmath118 .
thus , in both the light front and cartesian form , the @xmath10 solution has @xmath129 . in
, it is more apparent that the cancellation is between the non - abelian electric pieces ( the @xmath130 integrals ) and the corresponding non - abelian magnetic pieces ( the @xmath131 integrals ) .
as coleman already noted , the expression @xmath89 in does not in general agree with @xmath132 in for the light front solutions in and .
we have shown that the same is true ( as it should be ) in cartesian coordinate version of the solution .
this is taken by coleman @xcite as an illustration of the wu - yang ambiguity @xcite in non - abelian gauge theories , not all the gauge invariant information is contained in the field strength tensor .
we now examine the issue of the time - dependent , non - abelian aharonov - bohm phase for the time - dependent @xmath133 wu - yang monopole solution , where @xmath134 .
this solution is different in two respects from the non - abelian plane - waves of the previous section .
first , in contrast to the non - abelian plane wave solution of the previous section , the time - dependent @xmath133 wu - yang monopole solution is more strongly non - abelian since the prototypical non - abelian term ( @xmath135 ) is non - zero .
second , in the light front coordinates of the previous section the non - abelian electric and magnetic components of the field strength tensor are mixed up ( combined ) in non - zero components like @xmath136 or @xmath137 . for the time - dependent @xmath133 wu - yang monopole solution of this section ,
the field strength tensor is split into separate non - abelian electric ( _ e.g. _ @xmath138 ) and magnetic ( _ e.g. _ @xmath139 ) pieces .
the vector potential for the time - dependent wu - yang solution is given by @xcite @xmath140 ~,\ ] ] where @xmath141 is a radial and time dependent function and @xmath142 is the su(2 ) levi - civita structure constant of the group .
for this time - dependent solution , one works specifically with the su(2 ) group , whereas for the non - abelian plane - wave solution of the last section , the lie group was arbitrary .
as pointed out in the previous section , the non - abelian part of the field strength tensor did not play a big role in the coleman solutions .
for this reason , the exact nature of the non - abelian group was not so crucial for the coleman plane wave solutions . from
, one can immediately find the associated field strength tensor as @xmath143 where the dot denotes a time derivative and the prime denotes a radial derivative .
the first two terms in @xmath131 are the abelian / pure curl part of the non - abelian magnetic field ( _ i.e. _ @xmath144 _ k$ ] ) . also the non - abelian electric field , @xmath145 , is abelian in character since the prototypical , non - abelian piece , @xmath146 , is not present .
one can thus show , generally , that there is a cancellation between the abelian part of the non - abelian electric field and the abelian part of the non - abelian magnetic field , @xmath147 _ i$ ] .
for this abelian part of the fields from , the area " integral that appears in the field strength version of the aharonov - bohm phase is @xmath148 where for the first , electric area " integral , we have done the time integration to get @xmath149 ; for the second , magnetic area integral , we have used stokes theorem to get @xmath150 , which then cancels the first , electric term .
note , for the first , electric term , one leg of the area " is a time piece .
however , the last , non - abelian term in the magnetic field
_ i.e. _ the term @xmath151 in could give a non - zero contribution to the calculation , so at first sight one would not , in general , expect the same kind of cancellation between the non - abelian electric and magnetic parts that occurred in for the coleman solution . to this end , we will look at the full non - abelian fields of for a specific contour bounding a specific area , and we will see that there is still a complete cancellation between the electric and magnetic parts , as there was for the coleman solutions of the previous section . the specific contour we take is a hoop of radius @xmath152 in the @xmath153 plane going in the counterclockwise direction .
the area spanning this contour is a circle with area @xmath154 , _
i.e. _ pointing in the @xmath155 direction
. the surface and contour are shown in figure [ fig2 ] . in evaluating the electric and magnetic parts of
, we will consider infinitesimal path lengths , @xmath156 , and infinitesimal areas , @xmath157 see figure [ fig3 ] .
we begin by looking at the non - abelian electric contribution @xmath158 . for the @xmath130 from ,
one finds @xmath159^a \int \dot f(r , t ) dt = \delta \varphi \delta f \delta^{a3 } ~,\ ] ] where we have used the fact that @xmath160^a = \delta \varphi r^2 \delta^{a3}$ ] , _ i.e. _ this integral gives twice the area of an infinitesimal wedge from the surface in figure [ fig2 ] ( see also figure [ fig3 ] ) .
the direction of the area is in the @xmath113 or @xmath161 direction . from the @xmath162 symbol in , with @xmath163 ,
the color index is fixed as @xmath164 . to make @xmath165 explicit ,
we have inserted @xmath166 into the final expression in .
the time integration has been done for an infinitesimal interval @xmath167 , so @xmath168 .
now we calculate the non - abelian magnetic contribution to the phase .
we do this in two separate pieces : the abelian part , @xmath169 , and the non - abelian part , @xmath170 .
first , for the abelian part , we have @xmath171 for our contour from figure [ fig2 ] , the first term in has @xmath172 since we are in the @xmath153 plane . but
as well , for the summation over the @xmath173 index , we also need @xmath174 or @xmath175 . for @xmath176 , we would have @xmath177 , but @xmath178 for the contour and surface we are using , so for @xmath176 , the first term in is zero .
taking all this into account , if we look at @xmath179 and @xmath180 , we find that the indexed part of the first term in is @xmath181 since in the @xmath153 plane @xmath182 .
note also that the color index , @xmath183 , is forced to be @xmath165 . with this , the first term of becomes @xmath184 next , since our surface and contour from figure [ fig2 ] are in the @xmath153 plane , this means @xmath185 , thus , for the last term in , this implies that @xmath165 , and we find that this term becomes @xmath186 where in the @xmath153 plane @xmath187 become @xmath188 .
the second term in will cancel the term in .
the first term in will have a @xmath189 from the infinitesimal @xmath190 integration . then using infinitesimal notation for the @xmath191 integration ( _ i.e. _ @xmath192 )
, we arrive at @xmath193 which then cancels the electric contribution .
thus , at this point we have confirmed , with specific contours and areas , the cancellation between the electric and abelian magnetic " parts of the non - abelian aharonov - bohm phase , which was shown generally in .
the final piece we need to deal with is the prototypical non - abelian piece of the magnetic contribution , namely @xmath194 this piece is seen to vanish since @xmath163 due to the contour / area from figure [ fig2 ] lying in the @xmath153 plane .
this forces @xmath176 so that @xmath195 , but since we are in the @xmath153 plane @xmath178 , so this prototypical non - abelian contribution vanishes .
thus , as for the coleman plane wave solution of the previous section , we find a cancellation between the electric and magnetic parts of the non - abelian aharonov - bohm phase .
although here we have shown this cancellation for only two types of time - dependent non - abelian solutions and with specific contours , we nevertheless advance the hypothesis that this cancellation is a general feature of both abelian and non - abelian aharonov - bohm phases for time - dependent fields .
we conclude this section by noting that , like the coleman solutions , the time - dependent wu - yang monopole solution shows a non - zero phase when calculated using the potential @xmath196 where in the last step we are considering an infinitesimal path length as in figure [ fig3 ] , in conjunction with an infinitesimal change in the potential @xmath197 . using the form for @xmath198 from , we find @xmath199^a \right)~.\ ] ] in the last step , we have canceled two minus signs but have switched the @xmath200 and @xmath201 index which then gives an additional minus sign . also , we have used @xmath202 .
finally , we again use @xmath203^a = \delta \varphi r^2 \delta^{a3}$ ] and find that @xmath204 this produces only the non - abelian magnetic phase contribution from the fields calculation . in the next section , we will discuss this non - equivalence between the time - dependent aharonov - bohm phase shift , calculated using the potentials versus the field strengths , by comparing with the time - dependent _ abelian _ aharonov - bohm phase case .
in the previous two sections , we discussed the time - dependent aharonov - bohm phase shift for non - abelian fields for two specific solutions ( the coleman plane wave solutions and the time - dependent wu - yang monopole ) and two specific contours ( a unit square in the @xmath37 plane and a ring of radius @xmath152 in the @xmath153 plane ) . for the coleman solutions , the field strength tensors had no contribution from the non - abelian term , @xmath205 , and the functional form of the field strengths was essentially the same as for abelian plane wave solutions .
one distinction between the coleman non - abelian plane waves and abelian planes waves is that superposition does not apply for the coleman plane waves . for the time - dependent wu - yang monopole solution ,
the electric field strength terms had no contribution from the non - abelian part of the solution , but the magnetic field did .
however , this non - abelian part for the magnetic field was found not to contribute to the aharonov - bohm phase for the specific paths and surfaces we used , which are shown in figure [ fig2 ] ( see also figure [ fig3 ] ) .
we now review the abelian , time - dependent aharonov - bohm effect and draw parallels with the non - abelian case .
the time - independent abelian aharonov - bohm effect has been well studied theoretically and also been confirmed experimentally @xcite ( see @xcite for a recent experimental tests using tunneling ) .
in contrast , the _ time - dependent _ abelian aharonov - bohm effect has received much less attention .
some of the papers dealing with the time - dependent abelian aharonov - bohm effect are @xcite @xcite @xcite @xcite . the paper @xcite predicts a time - shifting interference pattern for the time - dependent abelian aharonov - bohm effect , while @xcite @xcite find a cancellation of the time dependent electric and magnetic contributions and thus a non - shifting interference pattern .
the few experiments done on the time - dependent abelian aharonov - bohm effect ( see @xcite @xcite and also @xcite for an accidental " test of the time - dependent abelian aharonov - bohm effect ) confirm that predictions of @xcite @xcite , but each experiment had problems .
thus , a definitive test of the time - dependent abelian aharonov - bohm effect , to distinguish between the predictions of @xcite and @xcite @xcite , still needs to be done .
the vector potential for a time - dependent aharonov - bohm solenoid is given by @xcite @xcite ( we use cylindrical coordinates @xmath206 and the magnetic flux tube has a radius @xmath152 ) @xmath207 to begin with , we have taken the scalar potential , @xmath208 , as zero .
we return to this point later since there are non - single valued gauges where there is a non - zero _ and _ non - single valued @xmath208 .
the possibility of a non - single valued @xmath208 leads to something similar to the wu - yang ambiguity but for time - dependent abelian fields .
the magnetic and electric fields coming from are @xmath209 and @xmath210 evaluating the aharonov - bohm phase using the fields , for the infinitesimal path and associated area in figure [ fig3 ] gives @xmath211 by expanding @xmath212 and identifying the infinitesimal path , @xmath213 and the area , @xmath214 , equation becomes @xmath215 the time - dependent parts of the phase shift cancel each other , while the static aharonov - bohm phase shift , due to @xmath216 , remains . comparing the above calculations and discussion with the previous non - abelian results
, we see that in both abelian and non - abelian theories there is a cancellation between the time - dependent electric and magnetic contributions to the aharonov - bohm phase shift . for the non - abelian case we have only shown this cancellation for two specific , time - dependent solutions and for special contours and surfaces .
although we have not shown this cancellation for the non - abelian fields in general , we nevertheless conjecture that this is a feature of more general time - dependent non - abelian field configurations .
in this paper we have investigated the aharonov - bohm effect for _ time - dependent _ non - abelian fields .
in contrast to the abelian aharonov - bohm effect , much less work has been done on even the _ time - independent _ non - abelian aharonov - bohm effect
reference @xcite is one of the few works to deal with the time - independent non - abelian aharonov - bohm effect .
the reason for this most likely lies in the difficulty in experimentally setting up and controlling non - abelian field configurations . in comparison ,
abelian fields can be much more easily manipulated _
e.g. _ setting up the magnetic flux tube , which is used in the canonical abelian , aharonov - bohm setup . in this work , we studied ( for the first time as far as we could determine from the literature ) the aharonov - bohm effect for time - dependent non - abelian fields .
we did this using two specific , known time - dependent solutions ( the coleman plane wave solutions @xcite and the time - dependent wu - yang monopole of @xcite ) and using specific contours and associated areas ( see figures [ fig1 ] and [ fig2 ] ) .
there were two common results of this investigation : ( i ) the non - abelian aharonov - bohm phase calculated via the fields ( _ i.e. _ @xmath132 see ) did not agree in general with the aharonov - bohm phase calculated via the potentials ( _ i.e. _ @xmath89 see ) .
this point was already remarked on by coleman @xcite as an example of the wu - yang ambiguity @xcite for non - abelian fields .
( ii ) there was a cancellation of the time - dependent contribution to the aharonov - bohm phase shift coming from the electric and magnetic non - abelian fields .
for the coleman @xmath10 solution , this cancellation is given in equation and for the time - dependent wu - yang monopole , this cancellation is given in equations , and .
in section iv , we carried out a review of the time - dependent abelian aharonov - bohm effect and made comparison to the results from the time - dependent non - abelian aharonov - bohm effect from sections i - iii .
for both abelian and non - abelian fields , we found a cancellation of the time - dependent part of the phase shift .
for the abelian case , this was shown to occur generally see equations and .
this led to the conjecture that this cancellation was also a feature of more general time - dependent non - abelian fields and more general contours / area .
m. b. was supported through a csu fresno college of science and mathematics fssr grant and a norma t. craig grant .
|
in this article , we study the _ time - dependent _ aharonov - bohm effect for non - abelian gauge fields .
we use two well known time - dependent solutions to the yang - mills field equations to investigate the aharonov - bohm phase shift . for both of the solutions
, we find a cancellation between the phase shift coming from the non - abelian `` magnetic '' field and the phase shift coming from the non - abelian `` electric '' field , which inevitably arises in time - dependent cases .
we compare and contrast this cancellation for the time - dependent non - abelian case to a similar cancellation which occurs in the time - dependent abelian case .
we postulate that this cancellation occurs generally in time - dependent situations for both abelian and non - abelian fields .
|
in this paper , we recall that second - order logic is a convenient language to express non - termination of while loops , modeled as rules . such rules
are commonly used abstractions of real program pieces , see , e.g. , @xcite for the java programming language .
our main contribution is the definition of two new classes of rules , the termination of which is decidable , by restricting the language to known decidable cases , namely s1s and s2s .
we also show and illustrate how decision procedures for their weak versions ws1s and ws2s can help proving termination / non - termination .
we organize the paper as follows .
section [ sec : preliminaries ] presents the main concepts we need while section [ sec : sof - nt ] gives the theoretical results of the paper .
section [ sec : examples ] illustrates the results by means of examples and section [ sec : related - works - conclusion ] concludes .
we give a quick description of s1s and s2s , see @xcite for a more detailed presentation .
s1s is the monadic second - order theory of 1 successor .
interpretations correspond to finite or infinite words over a given finite alphabet @xmath0 .
terms are constructed from the constant 0 and first - order variables @xmath1 , @xmath2 , by application of the successor function @xmath3 , which is left - associative .
we abbreviate @xmath4 successive applications of @xmath3 starting from @xmath5 ( @xmath6 ) to @xmath4 .
atomic formul are constructed from terms , second - order variables @xmath7 , @xmath8 , and predicates of the form @xmath9 where @xmath10 .
they have the form @xmath11 , @xmath12 , @xmath13 , @xmath14 where @xmath15 and @xmath16 are terms .
formul are constructed from atomic formul , the usual boolean connectives ( @xmath17 , @xmath18 , ) and quantification ( @xmath19 and @xmath20 ) over first and second - order variables .
first - order variables are interpreted as elements of @xmath21 representing positions in words and second - order variables as subsets of @xmath21 .
constant @xmath5 is interpreted as the first position in a word and function @xmath3 as the next position .
the formula @xmath14 is true in a word @xmath22 if at position @xmath15 of @xmath22 there is character @xmath23 .
ws1s ( weak s1s ) is a restriction of s1s where second - order variables are interpreted as _
finite _ sets only .
s2s is the monadic second - order theory of 2 successors .
interpretations correspond to finite or infinite labelled binary trees over a given finite alphabet @xmath0 .
terms and formul are constructed as in s1s except that constant 0 is replaced with @xmath24 and the successor function @xmath3 is replaced with functions @xmath25 and @xmath26 , which are left - associative .
we abbreviate successive applications of these functions , for instance @xmath27 stands for @xmath28 , which corresponds to @xmath29 , and @xmath30 stands for @xmath31 .
first - order variables are interpreted as elements of @xmath32 representing positions in binary trees and second - order variables as subsets of @xmath32 .
constant @xmath24 is interpreted as the root position of a binary tree , @xmath25 as the left successor , @xmath26 as the right successor and @xmath33 as the proper - prefix relation ( for instance @xmath34 but @xmath35 ) .
ws2s ( weak s2s ) is a restriction of s2s where second - order variables are interpreted as _
finite _ sets only .
a _ rule _ has the form @xmath36 where @xmath37 is the identifier of the rule , @xmath38 is a binary relation and @xmath39 and @xmath40 are tuples of distinct first - order variables ranging over a given domain . if @xmath37 is a monadic rule of the form @xmath41 and @xmath42 is a monadic second - order formula of s1s ( s2s ) with @xmath1 and @xmath2 as free variables , we call @xmath37 a _ monadic _ s1s ( respectively , s2s ) rule .
some examples can be found in section [ sec : examples ] .
we define an operational semantics as follows .
starting from a concrete tuple @xmath43 of elements of the domain , we first check whether there exists a concrete tuple @xmath44 such that @xmath45 .
if no such tuple exists , the computation stops .
otherwise , we choose any such tuple @xmath44 and reiterate .
the rule @xmath37 _ loops _ if we can find a concrete tuple @xmath43 starting an infinite computation .
if no such tuple exists , @xmath37 _ terminates_.
we consider the following second - order formulation of non - termination .
let @xmath36 be a rule .
[ def : recurrence - set ] we let @xmath46 denote the second - order formula @xmath47 \forall \tilde{x } \exists \tilde{y}\ ( \tilde{x}\in x \rightarrow [ \psi(\tilde{x},\tilde{y } ) \land \tilde{y}\in x])\ ; & ( 2 ) \end{array}\right.\ ] ] a _ recurrence set _ for @xmath37 is a set @xmath7 satisfying @xmath46 . condition ( 1 ) of definition [ def : recurrence - set ] simply states that the recurrence set @xmath7 is not empty .
condition ( 2 ) ensures that for any element @xmath1 of @xmath7 , there is an element @xmath2 of @xmath7 which satisfies the formula @xmath48 defining the rule @xmath37 .
the existence of a recurrence set is equivalent to non - termination .
[ theorem : nonterm - cns ] @xmath46 is true if and only if @xmath37 loops .
we prove both implications .
( @xmath49 ) .
as @xmath46 is true , we can start by selecting any arbitrary @xmath50 .
we know that there exists @xmath51 s.t . @xmath52 . by iterating this process
, we construct an infinite computation
. hence @xmath37 loops .
( @xmath53 ) . as there exists @xmath43 such that @xmath37 loops , let us consider an infinite computation starting at @xmath43 : @xmath54 let @xmath55 .
@xmath7 is a non - empty set verifying @xmath56)$ ] .
hence @xmath46 holds . the notion of _ closed recurrence set _
is introduced in @xcite .
[ def : closed - recurrence - set ] we let @xmath57 be the second - order formula @xmath47 \forall \tilde{x } \exists \tilde{y}\ ( \tilde{x}\in x \rightarrow \psi(\tilde{x},\tilde{y } ) ) \land \phantom{x } & ( 2 ) \\[1ex ] \forall \tilde{x}\forall\tilde{y}\ ( [ \tilde{x}\in x \land \psi(\tilde{x},\tilde{y } ) ] \rightarrow \tilde{y}\in x ) & ( 3 ) \end{array}\right.\ ] ] a _ closed recurrence set _ for @xmath37 is a set @xmath7 satisfying @xmath57 .
let @xmath58 denote the set of successors of @xmath39 with @xmath38 , the set of @xmath40 such that @xmath48 holds .
definition [ def : closed - recurrence - set ] imposes that @xmath59 for any @xmath39 in @xmath7 .
hence , we _ must _ stay in @xmath7 when applying @xmath38 to @xmath39 .
in contrast , definition [ def : recurrence - set ] requires that @xmath60 , we _ can _ stay in @xmath7 when applying @xmath38 to @xmath39 . therefore , closed recurrence sets are recurrence sets but we also have that recurrence sets always contain closed recurrence sets . [
theorem : open - iff - closed ] if there is a recurrence set @xmath7 for @xmath37 then there exists a rule @xmath61 with @xmath62 and @xmath63 such that @xmath64 is a closed recurrence set for @xmath65 .
the second - order formula @xmath46 is a necessary and sufficient condition for non - termination of at least one of the computations @xmath37 can generate .
symmetrically , @xmath66 is true if and only if for every value @xmath43 , any computation starting at @xmath43 halts .
as such a problem is in general undecidable ( see , e.g. , @xcite ) , it follows that @xmath46 is not computable . however , when the second - order logic is restricted to decidable cases , we obtain classes of rules for which the termination / non - termination property is decidable . termination of a _ monadic _ s1s or s2s rule is decidable .
the monadic second - order logics s1s and s2s are decidable @xcite and so is @xmath46 for a monadic s1s or s2s rule @xmath37 . if @xmath46 is true then @xmath37 loops else @xmath37 terminates .
weak versions of these logics , where second - order variables range over _ finite _ sets , are also decidable and decision procedures have been implemented ( see , e.g. , mona @xcite ) .
let @xmath37 be a monadic s1s or s2s rule .
[ dec - proc - w - computable - sufficient - conditions ] decision procedures for ws1s and ws2s provide computable sufficient conditions for proving non - termination of @xmath37 in the corresponding structure .
if such a decision procedure states that @xmath46 is true , then we know that there exists a non - empty finite set @xmath7 such that @xmath46 holds
. hence @xmath37 loops . note that if the decision procedure states that @xmath46 is false , then there is no finite set @xmath7 satisfying @xmath46 but an infinite set @xmath7 satisfying @xmath46 may exist .
hence we can not conclude , except in the following case .
[ dec - proc - w - dec - term ] when we know that the set of points which can start a computation from @xmath37 is finite , decision procedures for ws1s and ws2s also decide termination of @xmath37 in the corresponding structure .
if a decision procedure states that @xmath46 is true , then by corollary [ dec - proc - w - computable - sufficient - conditions ] @xmath37 loops .
else it states that @xmath46 is false .
so there does not exist a finite set @xmath7 satisfying @xmath46 .
as @xmath7 can not be infinite by hypothesis , it means that there does not exist a set @xmath7 such that @xmath46 holds .
hence @xmath37 terminates .
note that the condition of corollary [ dec - proc - w - dec - term ] can be decided in ws1s as it can be stated as @xmath67 .
however example [ corollary - does - not - decide - term ] shows that it does not decide termination .
consider @xmath68 where @xmath69 the set of points which can start a computation from @xmath37 is finite : @xmath70 .
mona tells us that @xmath46 is false . by corollary [ dec - proc - w - dec - term ]
, @xmath37 terminates .
consider @xmath68 where @xmath71 mona reports that @xmath46 is true , with a computed satisfying @xmath72 . indeed for any @xmath73 , there is a @xmath2 in @xmath7 such that @xmath42 holds : if @xmath74 , take @xmath75 and if @xmath76 , @xmath77 .
note that the set @xmath7 is not unique , as @xmath46 is also true for , e.g. , @xmath78 . by corollary [ dec - proc - w - computable - sufficient - conditions ] ,
@xmath37 loops .
[ corollary - does - not - decide - term ] consider @xmath68 where @xmath79 although mona tells us that there is no finite @xmath7 satisfying @xmath46 , as the set of points which can start a computation is infinite , we can not apply corollary [ dec - proc - w - dec - term ] . indeed , taking @xmath80 shows that @xmath46 is true . hence by theorem [ theorem : nonterm - cns ] , @xmath37 loops .
note that any decision procedure for s1s will prove that @xmath46 is true .
consider @xmath68 where @xmath81 mona reports that there is no finite @xmath7 satisfying @xmath46 .
as the set of points which can start a computation is infinite , we can not apply corollary [ dec - proc - w - dec - term ] .
assume that @xmath46 is true .
so there is a non - empty @xmath82 satisfying @xmath46 .
let @xmath83 be its least element .
condition ( 2 ) of definition [ def : recurrence - set ] states that there exists @xmath84 in @xmath7 such that @xmath85 , which contradicts that @xmath83 is the least element of @xmath7 .
hence @xmath46 is false , as should be shown by any decision procedure for s1s . by theorem [ theorem : nonterm - cns ] , @xmath37 terminates .
consider @xmath68 where @xmath86 with @xmath87 we have @xmath88 is true if and only if @xmath7 is closed by application of the successor function @xmath3 .
so , @xmath42 is true if and only if @xmath89 .
mona reports that @xmath90 is true , with a computed satisfying @xmath91 . by corollary [ dec - proc - w - computable - sufficient - conditions ] , @xmath37 loops .
consider @xmath68 where @xmath92 the set @xmath93 is not empty and for any @xmath1 in @xmath7 there is a @xmath2 in @xmath7 such that @xmath42 holds .
so @xmath46 is true ( also shown by mona ) . by corollary [ dec - proc - w - computable - sufficient - conditions ] ,
@xmath37 loops .
note that @xmath7 is a recurrence set for @xmath37 which is not closed .
indeed , @xmath94 and @xmath95 holds with @xmath96 .
hence , condition ( 3 ) of definition [ def : closed - recurrence - set ] does not hold . by theorem [ theorem : open - iff - closed ]
, there must exist @xmath97 with @xmath62 and @xmath63 such that @xmath64 is a closed recurrence set for @xmath65 .
for @xmath98 we have that @xmath62 , @xmath63 and @xmath64 is a closed recurrence set for @xmath65 .
consider @xmath68 where @xmath99 the set @xmath100 is not empty and for any @xmath1 in @xmath7 there is a @xmath2 in @xmath7 such that @xmath42 holds .
hence @xmath46 is true ( also shown by mona ) , so @xmath37 loops .
note that @xmath7 is a closed recurrence set for @xmath37 .
consider @xmath68 where @xmath101 the infinite set @xmath102 is not empty and for any @xmath1 in @xmath7 there is a @xmath2 in @xmath7 such that @xmath42 holds .
hence @xmath46 is true , as should be shown by any decision procedure for s2s .
so @xmath37 loops .
note that @xmath7 is a closed recurrence set for @xmath37 .
consider @xmath68 where @xmath86 with @xmath103 we have @xmath88 is true if and only if @xmath7 is closed by application of the successor functions @xmath25 and @xmath26 .
so , @xmath42 is true if and only if @xmath89 .
mona reports that @xmath46 is true , with a computed satisfying @xmath104 . by corollary [ dec - proc - w - computable - sufficient - conditions ] , @xmath37 loops .
recurrence sets were first introduced in @xcite where @xmath38 denotes any binary relation .
two symbolic analyses are presented in this paper for constructing such sets : a bitwise analysis , which assumes that the state space is finite and encoded using boolean variables , and a linear arithmetic analysis , which assumes that the program transitions can be represented as rational linear constraints . in contrast to our work , no second - order formulation is considered in this paper .
let us now focus on termination - decidable classes of rules . in @xcite ,
the authors present a decision procedure for an arbitrary rule @xmath105 where @xmath48 is a conjunction of equality constraints over rational trees . in @xcite ,
one finds variations of a decision procedure for finite sets of rules @xmath105 where @xmath48 is a conjunction of constraints @xmath106 or @xmath107 over a well - founded domain ( such as the natural numbers ) or the integers .
generalizing @xcite , termination of an arbitrary deterministic linear loop is shown decidable in @xcite over the integers , the rationals , and the reals . to the best of our knowledge ,
termination of a non - deterministic linear loop remains an open problem .
summarizing the paper , we have seen that second - order logic is a convenient language to express non - termination as a necessary and sufficient condition .
such a condition is in general undecidable . by restricting the language to the decidable cases s1s and s2s
, we have defined two new classes of rules , the termination of which is decidable .
finally , we have shown that the weak versions of these logics provide sufficient conditions for termination and non - termination of such rules .
a. r. bradley , z. manna , and h. b. sipma .
termination of polynomial programs . in r.
cousot , editor , _ proc . of the 6th international conference on verification , model checking and abstract interpretation ( vmcai05 ) _ , volume 3385 of _ lecture notes in computer science _ , pages 113129 .
springer , 2005 .
m. braverman .
termination of integer linear programs . in t. ball and r. b. jones , editors , _
proc . of the 18th international conference on computer
aided verification ( cav06 ) _ , volume 4144 of _ lecture notes in computer science _ ,
pages 372385 .
springer , 2006 . j. r. bchi . on a decision method in restricted second - order arithmetic . in e.
nagel , p. suppes , and a. tarski , editors , _ proc . of the 1960 international congress on logic , methodology and philosophy of science ( lmps60 ) _ , pages 111 .
stanford university press , june 1962 .
chen , b. cook , c. fuhs , k. nimkar , and p. ohearn .
proving nontermination via safety . in e.
brahm and k. havelund , editors , _ proc . of the 20th international conference on tools and algorithms for the construction and analysis of systems ( tacas14 ) _ , volume 8413 of _ lecture notes in computer science _ ,
pages 156171 .
springer , 2014 .
m. codish , v. lagoon , and p. stuckey .
testing for termination with monotonicity constraints . in m.
gabbrielli and g. gupta , editors , _ proc .
of the 21st international conference on logic programming ( iclp05 ) _ , volume 3668 of _ lecture notes in computer science _ , pages 326340 .
springer , 2005 .
a. gupta , thomas a. henzinger , r. majumdar , a. rybalchenko , and r .-
xu . proving non - termination . in g. c. necula and p. wadler , editors , _ proc .
of the 35th acm sigplan - sigact symposium on principles of programming languages ( popl08 ) _ , pages 147158 .
acm press , 2008 . c. s. lee , n. d. jones , and a. m. ben - amram .
the size - change principle for program termination . in c. hankin and d. schmidt , editors , _ proc . of the 28th acm sigplan - sigact symposium on principles of programming languages ( popl01 ) _ , pages 8192 .
acm press , 2001 .
a. tiwari .
termination of linear programs . in r.
alur and d. peled , editors , _ proc . of the 16th international conference on computer aided verification ( cav04 ) _ , volume 3114 of _ lecture notes in computer science _
, pages 7082 .
springer , 2004 .
|
we consider the termination / non - termination property of a class of loops .
such loops are commonly used abstractions of real program pieces .
second - order logic is a convenient language to express non - termination .
of course , such property is generally undecidable . however , by restricting the language to known decidable cases , we exhibit new classes of loops , the non - termination of which is decidable .
we present a bunch of examples .
termination , non - termination , monadic second - order logic .
|
over the past two decades , _ fair resource allocation _ problems have received considerable attention in many application areas , including internet congestion control @xcite , rate control in software defined networks @xcite , scheduling in wireless networks @xcite , multi - resource allocation and scheduling in datacenters @xcite , and a variety of applications in operations research , economics , and game theory @xcite . in
most of these applications , positive linear ( packing ) constraints arise as a natural model of the allowable allocations . in this paper
, we focus on the problem of finding an _ @xmath0-fair _ vector on the set determined by packing constraints @xmath17 where all @xmath18 . is commonly assumed to be a 0 - 1 matrix @xcite , important applications ( such as , e.g. , multi - resource allocation in datacenters ) are modeled by a more general constraint matrix @xmath9 with arbitrary non - negative elements @xcite . ] we refer to this problem as @xmath13fair packing . for a vector of positive weights @xmath19 and @xmath20 , an allocation vector @xmath21 of size @xmath22 is weighted @xmath0-fair , if for any alternative feasible vector @xmath23 : @xmath24 @xcite . for a compact and convex feasible region
, @xmath21 can be equivalently defined as a vector that solves the problem of maximizing @xmath25 @xcite , where : @xmath26 @xmath0-fairness provides a trade - off between efficiency ( sum of allocated resources ) and fairness ( minimum allocated resource ) as a function of @xmath0 : the higher the @xmath0 , the better the fairness guarantees and the lower the efficiency @xcite .
important special cases are proportional fairness ( @xmath5 ) and max - min fairness ( @xmath16 ) . when @xmath27 , we have the unfair " case of linear optimization
. distributed algorithms for @xmath13fair packing are of particular interest , as many applications are inherently distributed ( such as , e.g. , network congestion control ) , while in others parallelization is highly desirable due to the large problem size ( as in , e.g. , resource allocation in datacenters ) .
we adopt the model of distributed computation commonly used in the design of packing linear programming ( lp ) algorithms @xcite and which generalizes the model from network congestion control @xcite . in this model
, an agent @xmath28 controls the variable @xmath29 and has information about : ( i ) the @xmath30 column of the @xmath31 constraint matrix @xmath9 , ( ii ) the weight @xmath8 , ( iii ) upper bounds on the global problem parameters @xmath32 , and @xmath33 , where @xmath34 , and @xmath35 , and ( iv ) in each round , the relative slack of each constraint @xmath36 in which @xmath29 takes part .
distributed algorithms for @xmath13fair resource allocations have been most widely studied in the network congestion control literature , using a control - theoretic approach @xcite .
such an approach yields continuous - time algorithms that converge after `` finite '' time ; however , the convergence time of these algorithms as a function of the input size is poorly understood .
some other distributed pseudo - polynomial - time approximation algorithms that can address @xmath0-fair packing are described in table [ table : prev ] .
these algorithms all have convergence times that are at least linear in the parameters describing the problem .
no previous work has given truly fast ( poly - log iterations ) distributed algorithms for the general case of @xmath0-fair packing . only for the unfair @xmath37 case ( packing lps ) ,
are such algorithms known @xcite .
[ [ our - results . ] ] our results .
+ + + + + + + + + + + + _ we provide the first efficient , distributed , and stateless algorithm for weighted @xmath0-fair packing _ , namely , for the problem @xmath38 , where distributed agents update the values of @xmath29 s asynchronously and react only to the current state of the constraints .
we assume that all non - zero entries @xmath39 of matrix @xmath9 satisfy @xmath40 .
considering such a normalized form of the problem is without loss of generality ( see appendix [ appendix : scaling ] in the full version of the paper ) .
the approximation provided by the algorithm , to which we refer as the @xmath12-approximation , is ( i ) @xmath41-multiplicative for @xmath3 , and ( ii ) @xmath42-additive can not be avoided here , as additive approximation is not invariant to the scaling of the objective . ] for @xmath5 , where @xmath43 .
the main results are summarized in the following theorem , where , to unify the statement of the results , we treat @xmath0 as a constant that is either equal to 1 or bounded away from 0 and 1 , and we also loosen the bound in terms of @xmath44 and @xmath33 . for a more detailed statement ,
see theorems [ thm : convergence - alpha<1 ] [ thm : convergence - alpha>1 ] .
( main result ) for a given weighted @xmath0-fair packing problem @xmath45 , where @xmath46 is given by ( [ eq : f - alpha ] ) , there exists a stateless and distributed algorithm ( @xmath0-fairpsolver ) that computes an @xmath12-approximate solution in @xmath47 rounds .
_ to the best of our knowledge , for any constant approximation parameter @xmath12 , our algorithm is the first distributed algorithm for weighted @xmath0-fair packing problems with a poly - logarithmic convergence time . _
the algorithm is _
stateless _ according to the definition given by awerbuch and khandekar @xcite
: it starts from any initial state , the agents update the variables @xmath29 in a cooperative but uncoordinated manner , reacting only to the current state of the constraints that they observe , and without access to a global clock .
statelessness implies various desirable properties of a distributed algorithm , such as : asynchronous updates , self - stabilization , and incremental and local adjustments @xcite .
we also obtain the following structural results that characterize @xmath13fair packing allocations as a function of the value of @xmath0 : * we derive a lower bound on the minimum coordinate of the @xmath13fair packing allocation as a function of @xmath0 and the problem parameters ( lemma [ lemma : lower - bound ] ) .
this bound deepens our understanding of how the fairness ( a minimum allocated value ) changes with @xmath0 .
* we prove that for @xmath48 , @xmath13fair packing can be @xmath49approximated by any @xmath11approximation packing lp solver ( lemma [ lemma : lp - close - to - small - alpha - fair ] ) .
this result allows us to focus on the @xmath50 cases .
* we show that for @xmath51 , @xmath13fair allocation is @xmath11approximated by a @xmath52fair allocation returned by our algorithm ( lemmas [ lemma : alpha - close - to-1-below ] and [ lemma : alpha - close - to-1-above ] ) . *
we show that for @xmath53 , the @xmath13fair packing allocation @xmath21 and the max - min fair allocation @xmath54 are @xmath12-close to each other : @xmath55 element - wise .
this result is especially interesting as ( i ) max - min fair packing is not a convex problem , but rather a multi - objective problem ( see , e.g. , @xcite ) and ( ii ) the result yields the first convex relaxation of max - min fair allocation problems with a @xmath56 gap .
we now overview some of the main technical details of @xmath0-fairpsolver . in doing so , we point out connections to the two main bodies of previous work , from packing lps@xcite and network congestion control @xcite .
we also outline the new algorithmic ideas and proofs that were needed to obtain the results .
[ [ the - algorithm - and - kkt - conditions . ] ] the algorithm and kkt conditions .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the algorithm maintains primal and dual feasible solutions and updates each primal variable @xmath29 whenever a karush - kuhn - tucker ( kkt ) condition @xmath57 is not _
approximately _ satisfied . in previous work ,
relevant update rules include : @xcite ( for @xmath58 ) , where the update of each variable @xmath29 is proportional to the difference @xmath59 , and @xcite ( for @xmath37 ) , where each @xmath29 is updated by a multiplicative factor @xmath60 , whenever @xmath57 is not approximately satisfied . for our techniques
( addressing a general @xmath0 ) such rules do not suffice and we introduce the following modifications : ( i ) in the @xmath61 case we use multiplicative updates by factors @xmath62 and @xmath63 , where @xmath64 and ( ii ) we use additional threshold values @xmath65 to make sure that @xmath29 s do not become too small .
these thresholds guarantee that we maintain a feasible solution , but they significantly complicate ( compared to the linear case ) the argument that each step makes a significant progress . [
[ dual - variables . ] ] dual variables .
+ + + + + + + + + + + + + + + in @xmath0-fairpsolver , a dual variable @xmath66 is an exponential function of the @xmath67 constraint s relative slack : @xmath68 , where @xmath69 and @xmath70 are functions of global input parameters @xmath71 and @xmath33 . packing lp algorithms @xcite use similar dual variables with @xmath72 .
our work requires choosing @xmath69 to be a function of @xmath73 rather than a constant .
[ [ convergence - argument .
] ] convergence argument .
+ + + + + + + + + + + + + + + + + + + + + the convergence analysis of @xmath0-fairpsolver relies on the appropriately chosen concave potential function that is bounded below and above for @xmath74 $ ] , @xmath75 , and that increases with every primal update .
the algorithm can also be interpreted as a gradient ascent on a regularized objective function ( the potential function ) , using a generalized entropy regularizer ( see @xcite ) .
a similar potential function was used in many works on packing and covering linear programs , such as , e.g. , in @xcite and ( implicitly ) in @xcite . the lyapunov function from @xcite is also equivalent to this potential function when @xmath68 , @xmath76 .
as in these works , the main idea in the analysis is to show that whenever a solution @xmath23 is not close " to the optimal one , the potential function increases substantially .
however , our work requires several new ideas in the convergence proofs , the most notable being _
stationary rounds_. a stationary round is roughly a time when the variables @xmath29 do not change much and are close to the optimum .
poly - logarithmic convergence time is then obtained by showing that : ( i ) there is at most a poly - logarithmic number of non - stationary rounds where the potential function increases additively and the increase is `` large enough '' , and ( ii ) in all the remaining non - stationary rounds , the potential function increases multiplicatively .
our use of stationary rounds is new , as is the use of lagrangian duality and all the arguments that follow .
.comparison among distributed algorithms for @xmath13fair packing . [ cols="^,^,^,^",options="header " , ] [ table : prev ] [ [ relationship - to - previous - work . ] ] relationship to previous work .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + very little progress has been made in the design of efficient distributed algorithms for the general class of @xmath0-fair objectives .
classical work on distributed rate control algorithms in the networking literature uses a control - theoretic approach to optimize @xmath0-fair objectives .
while such an approach has been extensively studied and applied to various network settings @xcite , it has never been proven to have polynomial convergence time ( and it is unclear whether such a result can be established ) .
since @xmath0-fair objectives are concave , their optimization over a region determined by linear constraints is solvable in polynomial time in a centralized setting through convex programming ( see , e.g. , @xcite ) .
distributed gradient methods for network utility maximization problems , such as e.g. , @xcite summarized in table [ table : prev ] , can be employed to address the problem of @xmath0-fair packing . however , the convergence times of these algorithms depend on the dual gradient s lipschitz constant to produce good approximations .
while @xcite provide a better dependence on the accuracy @xmath12 than our work , the dependence on the dual gradient s lipschitz constant , in general , leads to at least linear convergence time as a function of @xmath22 , @xmath77 , and @xmath33 . as mentioned before , some special cases have been addressed , particularly for max - min fairness ( @xmath16 ) and for packing lps ( @xmath27 ) .
relevant work on max - min fairness includes @xcite , but none of these works have poly - logarithmic convergence time .
there is a long history of interesting work on packing lps in both centralized and distributed settings , e.g. , @xcite .
only a few of these works are stateless , including the packing lp algorithm of awerbuch and khandekar @xcite , flow control algorithm of garg and young @xcite , and the algorithm of awerbuch , azar , and khandekar @xcite for the special case of load balancing in bipartite graphs .
additionally , the packing lp algorithm of allen - zhu and orecchia @xcite is `` semi - stateless '' ; the lacking property to make it stateless is that it requires synchronous updates .
the @xmath58 case of @xmath0-fair packing problems is equivalent to the problem of finding an equilibrium allocation in eisenberg - gale markets with leontief utilities ( see @xcite ) .
similar to the aforementioned algorithms , the algorithm from @xcite converges in time linear in @xmath78 but also ( at least ) linear in the input size ( see table 1 ) . in terms of the techniques , closest to our work
is the work by awerbuch and khandekar @xcite and we now highlight the differences compared to this work . _ some preliminaries _ of the convergence proof follow closely those from @xcite : mainly , lemmas [ lemma : feasibility ] , [ lemma : approx - comp - slack ] , and [ lemma : potential - increase ] use similar arguments as corresponding lemmas in @xcite .
some parts of the lemmas lower - bounding the potential increase in @xmath61 , @xmath5 , and @xmath79 cases ( lemmas [ lemma : potential - increase - alpha<1 ] , [ lemma : potential - increase - proportional ] , and [ lemma : potential - increase - alpha>1 ] ) use similar arguments as @xcite , however , even those parts require additional results due to the existence of lower thresholds @xmath65 .
the similarity ends here , as the main convergence arguments are different than those used in @xcite .
in particular , the convergence argument from @xcite relying on stationary intervals can not be applied in the setting of @xmath13fair objectives .
more details about why this argument can not be applied and where it fails are provided in section [ section : convergence ] .
as already mentioned , we rely on the appropriately chosen definition of a stationary round . to show that in a stationary round a solution @xmath23 is @xmath11approximate , we use lagrangian duality and bound the duality gap through an intricate case analysis .
we remark that such an argument could not have been used in @xcite , since in the packing lp case there is no guarantee that the solution @xmath80 is dual - feasible .
[ [ organization - of - the - paper . ] ] organization of the paper .
+ + + + + + + + + + + + + + + + + + + + + + + + + + the rest of the paper is organized as follows .
section [ section : prelims ] provides the background .
section [ section : algorithm ] describes the algorithm , and section [ section : convergence ] provides illustrates the convergence analysis and structural results .
section [ section : conclusion ] concludes the paper .
we conclude in section [ section : conclusion ] .
the omitted technical details can be found in the full version of the paper .
[ [ weighted - alpha - fair - packing . ] ] weighted @xmath0-fair packing .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + consider the following optimization problem with positive linear ( packing ) constraints : @xmath81 where @xmath46 is given by ( [ eq : f - alpha ] ) , @xmath82 is the vector of variables , @xmath9 is an @xmath31 matrix with non - negative elements , and @xmath83 is a vector with strictly positive , @xmath84 , then trivially @xmath85 , for all @xmath28 such that @xmath86 . ] elements .
we refer to @xmath87 as the weighted @xmath0-fair packing .
the following definition and lemma introduced by mo and walrand @xcite characterize weighted @xmath0-fair allocations .
in the rest of the paper , we will use the terms weighted @xmath0-fair and @xmath0-fair interchangeably . _
let @xmath88 be a vector with positive entries and @xmath89 .
a vector @xmath90 is weighted @xmath0-fair , if it is feasible and for any other feasible vector @xmath23 : @xmath91 _ @xcite _ a vector @xmath21 solves @xmath87 for functions @xmath92 if and only if it is weighted @xmath0-fair .
notice in @xmath87 that since @xmath93 , @xmath76 , and the partial derivative of the objective with respect to any of the variables @xmath29 goes to @xmath94 as @xmath95 , the optimal solution must lie in the positive orthant .
moreover , since the objective is strictly concave and maximized over a convex region , the optimal solution is unique and @xmath87 satisfies strong duality ( see , e.g. , @xcite ) .
the same observations are true for the scaled version of the problem denoted by @xmath96 and introduced in the following subsection .
[ [ normalized - form . ] ] normalized form .
+ + + + + + + + + + + + + + + + we consider the weighted @xmath0-fair packing problem in the normalized form : @xmath97 where @xmath98 , @xmath99 is defined by ( [ eq : f - alpha ] ) , @xmath100 is a vector of positive weights , @xmath82 is the vector of variables , @xmath9 is an @xmath31 matrix with non - negative entries , and @xmath101 is a size-@xmath77 vector of 1 s .
we let @xmath33 denote the maximum element of the constraint matrix @xmath9 , and assume that every entry @xmath39 of @xmath9 is non - negative , and moreover , that @xmath40 whenever @xmath86 .
the maximum weight is denoted by @xmath102 and the minimum weight is denoted by @xmath103 .
the sum of the weights is denoted by @xmath104 and the ratio @xmath105 by @xmath106 .
we remark that considering problem @xmath87 in the normalized form @xmath96 is without loss of generality : any problem @xmath87 can be scaled to this form by ( i ) dividing both sides of each inequality @xmath36 by @xmath107 and ( ii ) working with scaled variables @xmath108 , where @xmath109 .
moreover , such scaling preserves the approximation ( [ appendix : scaling ] ) .
[ [ section : lower - bound - duality - gap ] ] kkt conditions and duality gap + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we will denote the lagrange multipliers for @xmath96 as @xmath110 and refer to them as `` dual variables '' .
the kkt conditions for @xmath96 are ( see appendix [ appendix : primal - dual - duality - gap ] ) : @xmath111 y_i \geq 0 , \quad\forall i\in\{1, ...
,m\ } \quad & \text{(dual feasibility)}\tag{k2}\label{eq : k2}\\[-3pt ] y_i\cdot\big({\mathop{\textstyle\sum}}_{j=1}^m a_{ij } x_j - 1\big ) = 0 , \quad\forall i\in\{1, ...
,m\}\quad & \text{(complementary slackness)}\tag{k3}\label{eq : k3}\\[-4pt ] { x_j}^\alpha{\mathop{\textstyle\sum}}_{i=1}^m y_i a_{ij } = w_j , \quad \forall j\in \{1, ... ,m\}\quad & \text{(gradient conditions)}\tag{k4}\label{eq : k4}\end{aligned}\ ] ] the duality gap for @xmath3 is ( see appendix [ appendix : primal - dual - duality - gap ] ) : @xmath112 where @xmath113 , while for @xmath114 : @xmath115 [ [ model - of - distributed - computation ] ] model of distributed computation + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we adopt the same model of distributed computation as @xcite , described as follows .
we assume that for each @xmath116 , there is an agent controlling the variable @xmath29 .
agent @xmath28 is assumed to have information about the following problem parameters : ( i ) the @xmath30 column of @xmath9 , ( ii ) the weight @xmath8 , and ( iii ) ( an upper bound on ) @xmath32 , and @xmath33 . in each round ,
agent @xmath28 collects the relative slack , agent @xmath28 would need to have information about @xmath117 . ]
@xmath118 of all constraints @xmath36 for which @xmath119 .
we remark that this model of distributed computation is a generalization of the model considered in network congestion control problems @xcite where a variable @xmath29 corresponds to the rate of node @xmath28 , @xmath9 is a 0 - 1 routing matrix , such that @xmath120 if and only if a node @xmath28 sends flow over link @xmath36 , and @xmath10 is the vector of link capacities . under this model ,
the knowledge about the relative slack of each constraint corresponds to each node collecting ( a function of ) congestion on each link that it utilizes .
such a model was used in network utility maximization problems with @xmath0-fair objectives @xcite and general strongly - concave objectives @xcite .
the pseudocode for the @xmath0-fairpsolver algorithm that is run at each node @xmath28 is provided in fig 1 .
the basic intuition is that the algorithm keeps kkt conditions ( [ eq : k1 ] ) and ( [ eq : k2 ] ) satisfied and works towards ( approximately ) satisfying the remaining two kkt conditions ( [ eq : k3 ] ) and ( [ eq : k4 ] ) to minimize the duality gap .
the algorithm can run in the distributed setting described in section [ section : prelims ] . in each round
, an agent @xmath28 updates the value of @xmath29 based on the relative slack of all the constraints in which @xmath28 takes part , as long as the kkt condition ( [ eq : k4 ] ) of agent @xmath28 is not approximately satisfied .
the updates need not be synchronous : we will require that all agents make updates at the same speed , but without access to a global clock . to allow for self - stabilization and dynamic changes
, the algorithm runs forever at all the agents , which is a standard requirement for self - stabilizing algorithms ( see , e.g. , @xcite ) .
the convergence of the algorithm is measured as the number of rounds between the round in which the algorithm starts from some initial solution and the round in which it reaches an @xmath11approximate solution , assuming that there are no hard reset events or node / constraint insertions / deletions in between . without loss of generality , we assume that the input parameter @xmath12 that determines the approximation quality satisfies @xmath121 for any @xmath0 , and @xmath122 for @xmath61
. the parameters @xmath123 , @xmath124 , and @xmath125 are set as follows .
for technical reasons ( mainly due to reinforcing dominant multiplicative updates of the variables @xmath29 ) , we set the values of the lower thresholds @xmath65 below the actual lower bound of the optimal solution that we derive in lemma [ lemma : lower - bound ] : @xmath126 we denote @xmath127 , @xmath128 . the constant @xmath69 that multiplies the exponent in the dual variables @xmath66
is chosen as @xmath129 .
because @xmath65 only depends on @xmath8 and on global parameters , we also have @xmath130 , @xmath75 . the parameter @xmath70 that appears in the exponent of the @xmath66 s
is chosen as @xmath131 .
the `` absolute error '' of ( [ eq : k4 ] ) @xmath132 is set to @xmath133 . for @xmath134 ,
we set @xmath135 , where the choice of @xmath136 is described below . for @xmath137 , we set @xmath138 , @xmath139 .
similar to @xcite , we choose the value of @xmath136 so that if we set @xmath140 , in any round the value of each @xmath141 changes by a multiplicative factor of at most @xmath142 .
since the maximum increase over any @xmath29 in each iteration is by a factor @xmath143 , and @xmath23 is feasible in each round ( see lemma [ lemma : feasibility ] ) , we have that @xmath144 , and therefore , the maximum increase in each @xmath66 is by a factor of @xmath145 .
a similar argument holds for the maximum decrease .
hence , we choose @xmath136 so that : @xmath146 and it suffices to set : @xmath147 * remark : * in the @xmath61 cases , since @xmath148 , the maximum decrease in @xmath149 is by a factor @xmath150 , @xmath75 .
in this section , we analyze the convergence time of @xmath0-fairpsolver .
we first state our main theorems and provide some general results that hold for all @xmath2 .
we show that starting from an arbitrary solution , the algorithm reaches a feasible solution within poly - logarithmic ( in the input size ) number of rounds , and maintains a feasible solution forever after . similar to @xcite
, we use a concave potential function that , for feasible @xmath23 , is bounded below and above and increases with any algorithm update .
then , we analyze the convergence time separately for three cases : @xmath61 , @xmath5 , and @xmath79 . with an appropriate definition of a _ stationary round _ for each of the three cases ,
we show that in every stationary round , @xmath23 approximates `` well '' the optimal solution by bounding the duality gap . on the other hand , for any non - stationary round
, we show that the potential increases substantially .
this large increase in the potential then leads to the conclusion that there can not be too many non - stationary rounds , thus bounding the overall convergence time .
then , we sketch the proof of theorem [ thm : convergence - alpha>1 ] ( @xmath79 ) , while we defer the full proofs of the three theorems to the full version of this paper . the main proof idea in all the cases
is as follows . with an appropriate definition of a _ stationary round _ for each of the three cases @xmath61 , @xmath5 , and @xmath79 , we show that in every stationary round , @xmath23 approximates `` well '' the optimal solution by bounding the duality gap . on the other hand , for any non - stationary round
, we show that the potential increases substantially .
this large increase in the potential leads to the conclusion that there can not be too many non - stationary rounds , thus bounding the overall convergence time .
we make a few remarks here .
first , we require that @xmath0 be bounded away from zero .
this requirement is without loss of generality because we show that when @xmath48 , any @xmath11approximation lp provides a @xmath151approximate solution to @xmath96 ( lemma [ lemma : lp - close - to - small - alpha - fair ] ) .
thus , when @xmath48 we can switch to the algorithm of @xcite , and when @xmath50 , the convergence time remains poly - logarithmic in the input size and polynomial in @xmath78 .
second , the assumption that @xmath122 in the @xmath61 case is also without loss of generality , because we show that when @xmath0 is close to 1 ( roughly , @xmath152 ) , we can approximate @xmath96 by switching to the @xmath58 case of the algorithm ( lemma [ lemma : alpha - close - to-1-below ] ) . finally , when @xmath153 , the algorithm achieves an @xmath11approximation in time @xmath154 .
we believe that a polynomial dependence on @xmath0 is difficult to avoid in this setting , because by increasing @xmath0 , the gradient of the @xmath0-fair utilities @xmath155 blows up on the interval @xmath156 : as @xmath0 increases , @xmath157 quickly starts approaching a step function that is equal to @xmath158 on the interval @xmath159 $ ] and equal to 0 on the interval @xmath160 $ ] . to characterize the behavior of @xmath13fair allocations as @xmath0 becomes large ,
we show that when @xmath161 , all the coordinates of the @xmath13fair vector are within a @xmath56 multiplicative factor of the corresponding coordinates of the max - min fair vector ( lemma [ lemma : mmf - alpha - fair ] ) .
finally , we note that the main convergence argument from @xcite that uses an appropriate definition of _ stationary intervals _ does not extend to our setting . the proof from @xcite `` breaks '' in the part that shows that the solution is @xmath11approximate throughout any stationary interval , stated as lemma 3.7 in @xcite .
the proof of lemma 3.7 in @xcite is by contradiction : assuming that the solution is not @xmath11approximate , the proof proceeds by showing that at least one of the variables would increase in each round of the stationary interval , thus eventually making the solution infeasible and contradicting one of the preliminary lemmas . for @xmath162 , unlike the linear objective in @xcite , @xmath0-fair objectives are negative , and the assumption that the solution is not @xmath11approximate does not lead to any conclusive information . for @xmath61 , adapting the proof of lemma 3.7 from @xcite leads to the conclusion that for at least one @xmath28 , in each round @xmath163 of the stationary interval @xmath164 , where @xmath21 is the optimal solution , and @xmath165 is the solution at round @xmath163 . in @xcite , where @xmath37 , this implies that @xmath29 increases in each round of the stationary interval , while in our setting ( @xmath89 ) it is not possible to draw such a conclusion . [ [ main - results . ] ] main results .
+ + + + + + + + + + + + + our main results are summarized in the following three theorems .
the objective is denoted by @xmath166 , @xmath165 denotes the solution at the beginning of round @xmath163 , and @xmath21 denotes the optimal solution .
[ thm : convergence - alpha<1 ] ( convergence for @xmath137 ) @xmath0-fairpsolver solves @xmath96 approximately for @xmath137 in time that is polynomial in @xmath167 .
in particular , after at most @xmath168 rounds , there exists at least one round @xmath163 such that @xmath169 .
moreover , the total number of rounds @xmath170 in which @xmath171 is also bounded by ( [ eq : thm - alpha<1-convergence - time ] ) .
[ thm : convergence - alpha=1 ] ( convergence for @xmath5 ) @xmath0-fairpsolver solves @xmath172 approximately in time that is polynomial in @xmath173 .
in particular , after at most @xmath174 rounds , there exists at least one round @xmath163 such that @xmath175 .
moreover , the total number of rounds @xmath170 in which @xmath176 is also bounded by ( [ eq : alpha=1-conv - time - bound ] ) .
[ thm : convergence - alpha>1 ] ( convergence for @xmath79 ) @xmath0-fairpsolver solves @xmath96 approximately for @xmath153 in time that is polynomial in @xmath177 .
in particular , after at most : @xmath178 rounds , there exists at least one round @xmath163 such that @xmath179 .
moreover , the total number of rounds @xmath170 in which @xmath180 is also bounded by ( [ eq : thm - alpha>1-convergence - time ] ) .
proofs of theorem [ thm : convergence - alpha<1 ] and theorem [ thm : convergence - alpha=1 ] are provided in the full version of the paper .
we sketch the proof of theorem [ thm : convergence - alpha>1 ] in section [ section : alpha>1 ] .
[ [ feasibility - and - approximate - complementary - slackness . ] ] feasibility and approximate complementary slackness .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the following three lemmas are preliminaries for the convergence time analysis .
lemma [ lemma : feasibility ] shows that starting from a feasible solution , the algorithm always maintains a feasible solution .
lemma [ lemma : self - stabilization ] shows that any violated constraint becomes feasible within poly - logarithmic number of rounds , and remains feasible forever after .
combined with lemma [ lemma : feasibility ] , lemma [ lemma : self - stabilization ] allows us to focus only on the rounds with feasible solutions @xmath23 .
lemma [ lemma : approx - comp - slack ] shows that after a poly - logarithmic number of rounds , approximate complementary slackness ( kkt condition ( [ eq : k3 ] ) ) holds in an aggregate sense : @xmath181 .
[ lemma : feasibility ] if the algorithm starts from a feasible solution , then the algorithm maintains a feasible solution @xmath23 : @xmath182 , @xmath75 and @xmath144 , @xmath76 , in each round . by the statement of the lemma ,
the solution is feasible initially . from the way that the algorithm makes updates to the variables
@xmath29 , it is always true that @xmath183 , @xmath75 .
now assume that @xmath23 becomes infeasible in some round , and let @xmath184 denote the ( feasible ) solution before that round , @xmath185 denote the ( infeasible ) solution after the round .
we have : @xmath186 for this to be true , @xmath23 must have increased over at least one coordinate @xmath28 such that @xmath187 .
for such a change to be triggered by the algorithm , it must also be true that : @xmath188 since , by the choice of @xmath138 , this term can increase by a factor of at most @xmath189 , it follows that : @xmath190 this further implies : @xmath191 and since whenever @xmath187 we also have @xmath192 , we get : @xmath193 on the other hand , since @xmath194 , @xmath195 , and @xmath196 : @xmath197 which contradicts ( [ eq : feasibility - less - that - wj ] ) .
[ lemma : self - stabilization ] if for any @xmath36 : @xmath198 , then after at most @xmath199 rounds , it is always true that @xmath200 .
suppose that @xmath198 for some @xmath36 .
then @xmath201 , and for every @xmath29 with @xmath86 : @xmath202 and therefore , none of the variables that appear in @xmath36 increases .
since @xmath198 , there exists at least one @xmath203 with @xmath204 such that @xmath205 . for each such @xmath203 , since @xmath206 : @xmath207 and therefore , @xmath203 decreases ( by a factor @xmath63 ) . as @xmath208 , after at most @xmath209 rounds in which @xmath198 , we must have @xmath210 , and therefore , @xmath200 . using the same arguments as in the proof of lemma [ lemma : feasibility ] ,
the constraint @xmath36 never gets violated again .
[ lemma : approx - comp - slack ] if the algorithm starts from a feasible solution , then after at most @xmath211 rounds , it is always true that : 1 .
there exists at least one approximately tight constraint : @xmath212 , 2 .
@xmath213 , and 3 .
@xmath214 .
suppose that @xmath215 .
then for each @xmath66 we have : @xmath216 due to lemma [ lemma : feasibility ] , we have that @xmath23 is feasible in every round , which implies that @xmath217 @xmath75 .
this further gives : @xmath218 and , therefore , all variables @xmath29 increase by a factor @xmath143 . from lemma [ lemma : feasibility ] , since the solution always remains feasible , none of the variables can increase to a value larger than 1 .
therefore , after at most @xmath219 rounds , there must exist at least one @xmath36 such that @xmath220 .
if in any round @xmath221 decreases , it can decrease by at most @xmath222 .
therefore , in every subsequent round @xmath223 for the second part of the lemma , let @xmath224 be the set of constraints that are at least @xmath225-looser " than the tightest constraint .
then for @xmath226 we have @xmath227 this further gives : @xmath228 moreover , for each @xmath229 we have @xmath230 , since for @xmath229 : @xmath231 therefore : @xmath232 interchanging the order of summation in the last line , we reach the desired inequality .
the proof of the last part of the lemma follows from feasibility : @xmath233 , @xmath76 ( lemma [ lemma : feasibility ] ) , and from @xmath234 .
lemmas analogous to [ lemma : feasibility ] and [ lemma : approx - comp - slack ] also appear in @xcite .
however , the proofs of lemmas [ lemma : feasibility ] and [ lemma : approx - comp - slack ] require new ideas compared to the proofs of the corresponding lemmas in @xcite .
we need to be much more careful in our choice of lower thresholds @xmath65 and constant @xmath69 in the dual variables , particularly by choosing @xmath69 as a function of several variables , rather than as a constant .
the choice of @xmath65 s is also sensitive as smaller @xmath65 s would make the potential function range too large , while larger @xmath65 s would cause more frequent decrease of `` small '' variables . in either case
, the convergence time would increase .
[ [ decrease - of - small - variables . ] ] decrease of small variables .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + the following lemma is also needed for the convergence analysis .
it shows that if some variable @xmath29 decreases by less than a multiplicative factor @xmath63 , i.e. , @xmath235 and @xmath29 decreases , then @xmath29 must be part of at least one approximately tight constraint .
this lemma will be used later to show that in any round the increase in the potential due to the decrease of `` small '' variables is dominated by the decrease of `` large '' variables ( i.e. , the variables that decrease by a multiplicative factor @xmath63 ) .
[ lemma : small - x - tight - yi ] consider the rounds that happen after the initial @xmath199 rounds . if in some round there is a variable @xmath235 that decreases
, then in the same round for some @xmath36 with @xmath119 it holds that : @xmath236 and @xmath237 .
suppose that some @xmath235 triggers a decrease over the @xmath30 coordinate .
the first part of the lemma is easy to show , simply by using the argument that at least one term of a summation must be higher than the average , i.e. , there exists at least one @xmath36 with @xmath86 such that : @xmath238 for the second part , as @xmath239 , we have that : @xmath240 since @xmath29 decreases , we have that @xmath241 , and therefore @xmath242 moreover , as @xmath243 , and @xmath244 , it follows that : @xmath245 observe that for @xmath246 : @xmath247 while for @xmath153 , since @xmath248 : @xmath249 where we have used the generalized bernoulli s inequality for @xmath250 @xcite , and then @xmath251 . recalling
that @xmath252 , and combining ( [ eq : y - bound-1 ] ) with ( [ eq : eps-8 - 1 ] ) and ( [ eq : eps-8 - 1 - 2 ] ) : @xmath253 finally , as @xmath254 , it follows that @xmath255 which gives : @xmath256 [ [ potential . ] ] potential .
+ + + + + + + + + + we use the following potential function to analyze the convergence time : @xmath257 where @xmath258 and @xmath99 is defined by ( [ eq : f - alpha ] ) .
the potential function is strictly concave and its partial derivative with respect to any variable @xmath29 is : @xmath259 the following fact ( given in a similar form in @xcite ) , which follows directly from the taylor series representation of concave functions , will be useful for the potential increase analysis : [ fact : taylor ] for a differentiable concave function @xmath260 and any two points @xmath261 : @xmath262 using fact [ fact : taylor ] and ( [ eq : potential - derivatives ] ) , we show the following lemma : [ lemma : potential - increase ] starting with a feasible solution and throughout the course of the algorithm , the potential function @xmath263 never decreases . letting @xmath184 and @xmath185 denote the values of @xmath23 before and after a round update , respectively ,
the potential function increase is lower - bounded as : @xmath264 since @xmath265 is concave , using fact [ fact : taylor ] and ( [ eq : potential - derivatives ] ) it follows that : @xmath266 if @xmath267 , then the term in the summation ( [ eq : potential - increase - before - bounds ] ) corresponding to the change in @xmath29 is equal to zero , and @xmath29 has no contribution to the sum in ( [ eq : potential - increase - before - bounds ] ) . if @xmath268 , then , as @xmath29 increases over the observed round , it must be @xmath269 . by the choice of the parameters , @xmath270 , and
therefore @xmath271 it follows that @xmath272 , and therefore @xmath273 finally , if @xmath274 , then it must be @xmath275 . by the choice of the parameters , @xmath276 , implying @xmath277 we get that @xmath278 , and therefore @xmath279 completing the proof . the outline of the proof is as follows .
we first derive a lower bound on the potential increase ( lemma [ lemma : potential - increase - alpha<1 ] ) , which will motivate the definition of a stationary round .
then , for the appropriate definition of a stationary round we will first show that in any stationary round , solution is @xmath49approximate .
then , to complete the proof , we will show in any non - stationary round there is a sufficiently large increase in the potential function , which , combined with the bounds on the potential value will yield the result .
the following lemma lower - bounds the increase in the potential function in any round of the algorithm .
[ lemma : potential - increase - alpha<1 ] if @xmath61 and @xmath280 , @xmath184 , @xmath281 and @xmath282 , @xmath185 , @xmath283 denote the values of @xmath265 , @xmath23 , and @xmath80 before and after a round , respectively , and @xmath284 , then if @xmath184 is feasible : 1 . @xmath285
@xmath286 ; 3 .
@xmath287 .
@xmath288 + * proof of 1 .
* observe that for @xmath289 , @xmath290 . from the proof of lemma [ lemma : potential - increase ]
, we have that : @xmath291 the proof that @xmath292 is implied by the proof of part 3 of this lemma ( see below ) . for each @xmath289
, we have that : @xmath293 therefore : @xmath294 * proof of 2 .
* let @xmath295 denote the set of @xmath28 s such that @xmath29 increases in the current round . then , recalling that for @xmath296 @xmath297 and that from the choice of parameters @xmath298 : @xmath299 since @xmath296 , @xmath300 , it follows that @xmath301 observing that for any @xmath302 we have that @xmath303 , we get : @xmath304 * proof of 3 .
* let @xmath305 denote the set of @xmath28 s such that @xmath29 decreases in the current round . in this case
not all the @xmath29 s with @xmath289 decrease by a multiplicative factor @xmath63 , since for @xmath289 : @xmath306 .
we will first lower - bound the potential increase over @xmath29 s that decrease multiplicatively : @xmath307 , so that @xmath308 .
recall that for @xmath289 : @xmath309 and @xmath310 .
it follows that : @xmath311 next , we prove that the potential increase due to decrease of @xmath29 such that @xmath312 is dominated by the potential increase due to @xmath203 s that decrease multiplicatively by the factor @xmath63
. choose any @xmath29 such that @xmath312 , and let @xmath313 .
from lemma [ lemma : small - x - tight - yi ] , there exists at least one @xmath36 with @xmath86 , such that : @xmath314 @xmath315 from ( [ eq : x - tight - constraint ] ) , there exists at least one @xmath316 such that @xmath317 and @xmath318 since @xmath319 $ ] and @xmath320 , using ( [ eq : large - xp ] ) , we have that @xmath321 . recalling ( [ eq : yi - lower - bound - zj ] ) : @xmath322 recalling that @xmath323 , it further follows that : @xmath324 because @xmath325 and @xmath326 , it follows that @xmath327 .
therefore : @xmath328 as @xmath61 , we have that @xmath329 , and @xmath330 .
similar to ( [ eq : yi - lower - bound - zj ] ) , we can lower - bound @xmath66 as : @xmath331 then , recalling @xmath332 , and using ( [ eq : yi - lower - bound - zj-2 ] ) , it is simple to show that : @xmath333 as @xmath334 and @xmath335 , it immediately follows from ( [ eq : alpha<1small - var - inc - cond1 ] ) that @xmath336 decreases by a factor @xmath63 . in the rest of the proof
we show that ( [ eq : alpha<1small - var - inc - cond1 ] ) and ( [ eq : alpha<1small - var - inc - cond2 ] ) imply that the increase in the potential due to the decrease of variable @xmath336 dominates the increase in the potential due to the decrease of variable @xmath29 by at least a factor @xmath22 .
this result then further implies that the increase in the potential due to the decrease of variable @xmath336 dominates the increase in the potential due to the decrease of _ all _ small @xmath203 s that appear in the constraint @xmath36 ( @xmath203 s are such that @xmath204 , @xmath337 , and @xmath338 )
. consider the following two cases : @xmath339 and @xmath340 .
* case 1 : @xmath339*. then , using ( [ eq : alpha<1small - var - inc - cond1 ] ) : @xmath341 * case 2 : @xmath340*. then , using ( [ eq : alpha<1small - var - inc - cond2 ] ) : @xmath342 combining ( [ eq : alpha<1-dominant - inc-1 ] ) and ( [ eq : alpha<1-dominant - inc-2 ] ) with ( [ eq : pot - increase - bound - multiplicative ] ) , it follows that : @xmath343 finally , since for @xmath344 : @xmath345 : @xmath346 completing the proof . parts 2 and 3 of lemma @xcite appear in a somewhat similar form in @xcite
. however , part 3 requires significant additional results for bounding the potential change due to decrease of small @xmath29 s ( i.e. , @xmath29 s that are smaller than @xmath347 ) that were not needed in @xcite .
the rest of the results in this paper are new .
consider the following definition of a stationary round : [ def : alpha<1-stationary - round ] ( stationary round . )
let @xmath284 .
a round is stationary if it happens after the initial @xmath348 rounds , where @xmath349 and @xmath350 , and both of the following two conditions hold : 1 .
@xmath351 , and 2 .
@xmath352 . in the rest of the proof
, we first show that in any stationary round , we have an @xmath49approximate solution , while in any non - stationary round , the potential function increases substantially .
we first prove the following lemma , which we will then be used in bounding the duality gap .
[ lemma : alpha<1-lower - bound - xi - j ] after the initial @xmath348 rounds , where @xmath349 and @xmath350 , in each round of the algorithm : @xmath353 , @xmath75 .
suppose without loss of generality that the algorithm starts with a feasible solution .
this assumption is w.l.o.g . because , from lemma [ lemma : self - stabilization ] , after at most @xmath354 rounds the algorithm reaches a feasible solution , and from lemma [ lemma : feasibility ] , once the algorithm reaches a feasible solution , it always maintains a feasible solution .
choose any @xmath28 .
using the same argument as in the proof of lemma [ lemma : feasibility ] , after at most @xmath355 rounds , there exists at least one round in which @xmath356 ( otherwise @xmath357 , which is a contradiction ) .
observe that in any round for which @xmath358 , @xmath29 increases by a factor @xmath359 .
therefore , the maximum number of consecutive rounds in which @xmath360 is at most @xmath355 , otherwise @xmath29 would increase to a value larger than 1 , making @xmath23 infeasible , which is a contradiction due to lemma [ lemma : feasibility ] .
the maximum amount by which @xmath361 can decrease in any round is bounded by a factor @xmath362 .
therefore , using the generalized bernoulli s inequality , it follows that in any round : @xmath363 a simple corollary of lemma [ lemma : alpha<1-lower - bound - xi - j ] is that : [ cor : alpha<1-lower - comp ] after the initial @xmath348 rounds , where @xmath349 and @xmath350 , in each round of the algorithm : @xmath364 . from lemma [ lemma : alpha<1-lower - bound - xi - j ] , after the initial @xmath348 rounds , it always holds @xmath365 , @xmath75 . multiplying both sides of the inequality by @xmath366 , @xmath75 and summing over @xmath28 ,
the result follows .
recall that @xmath367 denotes the primal objective .
the following lemma states that any stationary round holds an @xmath368-approximate solution .
[ lemma : alpha<1-stationary - round ] in any stationary round : @xmath369 , where @xmath21 is the optimal solution to @xmath370 .
since , by definition , a stationary round can only happen after the initial @xmath348 rounds , we have that @xmath23 in that round is feasible , and also from lemma [ lemma : approx - comp - slack ] : @xmath371 .
therefore , recalling eq .
( [ eq : duality - gap - alpha ] ) for the duality gap and denoting @xmath372 , we have that : @xmath373 from lemma [ lemma : alpha<1-lower - bound - xi - j ] , @xmath374 , @xmath75 .
partition the indices of all the variables as follows : @xmath375 then , using ( [ eq : alpha<1-duality - gap - bound ] ) : @xmath376 where : @xmath377 and @xmath378 the rest of the proof follows by upper - bounding @xmath379 and @xmath380 .
* bounding @xmath379 . * observing that @xmath75 : @xmath381 , we can write @xmath379 as : @xmath382 denote @xmath383 . it is simple to verify that @xmath384 is a convex function . since @xmath385 , @xmath386
, it follows that @xmath387 . now : @xmath388 if @xmath389 , then as @xmath390 , it follows that @xmath391 .
therefore : @xmath392 if @xmath393 , then ( using generalized bernoulli s inequality and @xmath394 ) : @xmath395 on the other hand : @xmath396 combining ( [ eq : alpha<1-r-1])([eq : alpha<1-r-3 ] ) with ( [ eq : alpha<1-g-1 ] ) : @xmath397 * bounding @xmath380 .
* because the round is stationary and @xmath398 , we have that : @xmath399 . using the second part of the stationary round definition and that @xmath400 ( follows from lemma [ lemma : alpha<1-lower - bound - xi - j ] ) : @xmath401 above
, first inequality follows from @xmath402 ( part 2 of the stationary round definition ) and corollary [ cor : alpha<1-lower - comp ] .
second inequality follows by breaking the left summation into two summations : those with @xmath403 and those with @xmath404 .
the third inequality follows from @xmath405 and part 1 of the stationary round definition .
observe that as @xmath406 , we have that @xmath407 . using ( [ eq : alpha<1-xjy - bound ] )
, it follows that : @xmath408 finally , combining ( [ eq : alpha<1-g-1 - 1 ] ) and ( [ eq : alpha<1-g-2 ] ) : @xmath409 theorem [ thm : convergence - alpha<1 ] from lemma [ lemma : alpha<1-stationary - round ] , in any stationary round : @xmath410 .
therefore , to prove the theorem , it suffices to show that there are at most @xmath411 non - stationary rounds in total , where @xmath412 , because we can always run the algorithm for @xmath413 to get an @xmath11approximation , and this would only affect the constant in the convergence time . to bound the number of non - stationary rounds
, we will show that the potential increases by a `` large enough '' multiplicative value in all the non - stationary rounds in which the potential is not too `` small '' .
for the non - stationary rounds in which the value of the potential is `` small '' , we show that the potential increases by a large enough value so that there can be only few such rounds . in the rest of the proof
, we assume that the initial @xmath414 rounds have passed , so that @xmath23 is feasible , and the statement of lemma [ lemma : approx - comp - slack ] holds . this does not affect the overall bound on the convergence time , as @xmath415 to bound the minimum and the maximum values of the potential @xmath265 , we will bound @xmath416 and @xmath417 . recall that
@xmath418 .
since @xmath419 , @xmath23 is always feasible , and @xmath217 , @xmath75 , we have that : @xmath420 and @xmath421 thus , we have : @xmath422 and @xmath423 recall from lemma [ lemma : potential - increase ] that the potential never decreases .
we consider the following three cases for the value of the potential : * case 1 : @xmath424*. since in this case @xmath425 , we have that @xmath426 . from lemma [ lemma : approx - comp - slack ] , @xmath427 , thus implying : @xmath428 as @xmath429 and @xmath325 . combining part 3 of lemma [ lemma : potential - increase - alpha<1 ] and ( [ eq : alpha<1-neg - phi - inc ] ) , the potential increases by at least
: @xmath430 since the potential never decreases , there can be at most @xmath431 case 1 rounds .
* case 2 : @xmath432 . * from lemma [ lemma : approx - comp - slack ]
, there exists at least one @xmath36 such that @xmath433 . since @xmath434 @xmath435 , it is also true that @xmath436 , and as @xmath437 and @xmath429 , it follows that @xmath438 . from ( [ eq :
alpha<1-p - alpha - bound ] ) , we also have @xmath439 . therefore : @xmath440 if @xmath441 , then @xmath442 from lemma [ lemma : approx - comp - slack ] , @xmath443 from the third part of lemma [ lemma : potential - increase - alpha<1 ] , the potential increases additively by at least @xmath444 and , therefore , @xmath445 after at most @xmath446 rounds .
* case 3 : @xmath447 . * in this case , @xmath448 . if the round is stationary , then from lemma [ lemma : alpha<1-stationary - round ] , @xmath369 .
if the round is not stationary , then from definition [ def : alpha<1-stationary - round ] , either : 1 .
@xmath449 , or 2 .
@xmath450 .
if the former is true , then using the first part of lemma [ lemma : potential - increase - alpha<1 ] , the potential increases by at least @xmath451 . if the latter is true , from the third part of lemma [ lemma : potential - increase - alpha<1 ] , the potential increases by at least @xmath451 .
it follows that there are at most @xmath452 non - stationary case 3 rounds . combining the three cases with the bound on @xmath453 ( [ eq : tau1+tau0-bound ] )
, the total convergence time is at most : @xmath454 rounds , as claimed .
the proof outline for the convergence of @xmath0-fairpsolver in the @xmath5 case is as follows .
first , we show that in any round it can not be the case that only `` small '' @xmath29 s ( i.e. , @xmath29 s that are smaller than @xmath347 ) decrease .
in fact , we show that the increase in the potential due to updates of `` small '' variables is dominated by the increase in the potential due to those variables that decrease multiplicatively by a factor @xmath455 ( lemmas [ lemma : mul - increase - prop ] and [ lemma : potential - increase - proportional ] ) .
we then define a stationary round and show that : ( i ) in any non - stationary round the potential increases significantly , and ( ii ) in any stationary round , the solution @xmath23 at the beginning of the round provides an additive @xmath456approximation to the optimum objective value .
[ lemma : mul - increase - prop ] starting with a feasible solution , in any round of the algorithm : 1 .
2 . @xmath458 .
fix any round , and let @xmath459 and @xmath460 denote the values of @xmath461 at the beginning and at the end of the round , respectively . if for all @xmath289 @xmath462 , there is nothing to prove .
suppose that there exists some @xmath463 that decreases .
then from lemma [ lemma : small - x - tight - yi ] there exists at least one @xmath464 such that @xmath86 , and : * @xmath465 , and * @xmath466 . since @xmath465 , there exists at least one @xmath316 such that @xmath332 . recalling that @xmath467 : @xmath468
since @xmath29 decreases , it must be @xmath469 . using ( [ eq : alpha=1-n - times - inc ] ) : @xmath470 and , therefore , @xmath336 decreases as well .
moreover , since ( [ eq : alpha=1-n - times - inc ] ) implies @xmath471 the proof of the first part of the lemma follows .
the second part follows from ( [ eq : alpha=1-n - times - inc ] ) as well , since : @xmath472 which , given that @xmath29 was chosen arbitrarily , implies : @xmath473 [ lemma : potential - increase - proportional ] let @xmath459 and @xmath460 denote the values of @xmath461 at the beginning and at the end of any fixed round , respectively . if @xmath184 is feasible , then the potential increase in the round is at least : 1 .
[ item : prop-1 ] @xmath474 ; 2 .
[ item : prop-2 ] @xmath475 .
[ item : prop-3 ] @xmath476 .
@xmath288 + * proof of [ item : prop-1 ] : * recall that : @xmath477 let @xmath478 , @xmath479 . if @xmath296 , then @xmath300 and @xmath480 .
since from the choice of parameters @xmath481 increases by at most a factor of @xmath189 , it follows that : @xmath482 , which gives @xmath483 .
therefore : @xmath484 * proof of [ item : prop-2 ] : * the proof is equivalent to the proof of the second part of lemma [ lemma : potential - increase - alpha<1 ] and is omitted .
* proof of [ item : prop-3 ] : * using that for @xmath289 we have that @xmath485 and @xmath486 , we can lower bound the increase in the potential as : @xmath487 now consider @xmath488 such that @xmath489 . from the proof of lemma [ lemma : mul - increase - prop ] , for each such @xmath203 there exists a constraint @xmath36 and a variable @xmath490 with @xmath491 such that @xmath204 , @xmath492 , @xmath493 , and @xmath494 . if @xmath495 then @xmath496 on the other hand , if @xmath497 , then : @xmath498 it follows from ( [ eq : delta - phi - j - in - s- ] ) that : @xmath499 finally , since for @xmath344 we have that @xmath500 : @xmath501 consider the following definition of a stationary round : [ def : alpha=1-stat - round ] a round is stationary if it happens after the initial @xmath348 rounds , where @xmath502 , @xmath503 and if both of the following conditions hold : * @xmath504 ; * @xmath505 .
we first show that in any non - stationary round there is a sufficient progress towards the @xmath11approximate solution .
[ lemma : non - stat - round - alpha=1 ] in any non - stationary round the potential function increases by at least @xmath506 .
a round is non - stationary if either of the two conditions from definition [ def : alpha=1-stat - round ] does not hold . if the first condition does not hold , then from the first part of lemma [ lemma : potential - increase - proportional ] , the potential increases by @xmath506 . if the second condition does not hold , then from either the second or the third part of lemma [ lemma : potential - increase - proportional ] the potential increases by at least @xmath507 . before proving that in every non - stationary round
, the solution is @xmath49approximate , we will need the following intermediary lemma .
[ lemma : cond - lower - bound ] starting with a feasible solution and after at most @xmath211 rounds , in any round of the algorithm : @xmath508 first , we claim that after the algorithm reaches a feasible solution it takes at most @xmath509 additional rounds for each agent @xmath28 to reach a round in which @xmath510 .
suppose not , and pick any agent @xmath511 for which in each of the @xmath509 rounds following the first round that holds a feasible solution : @xmath512 .
then @xmath203 increases in each of the rounds and after @xmath513 rounds we have @xmath514 .
therefore , after at most @xmath515 rounds the solution becomes infeasible , which is a contradiction ( due to lemma [ lemma : feasibility ] ) .
now choose any @xmath29 and observe @xmath516 over the rounds that happen after the first @xmath509 rounds .
the maximum number of consecutive rounds for which @xmath517 is @xmath518 , otherwise we would have @xmath357 , a contradiction .
since in any round , due to the choice of the algorithm parameters , @xmath481 decreases by at most a factor of @xmath519 , the minimum value that @xmath481 can take is at least @xmath520 , thus completing the proof .
now we are ready to prove that a solution in a stationary round is @xmath49approximate .
[ lemma : alpha=1-stat - round ] in any stationary round : @xmath521 , where @xmath21 is the optimal solution . since
, due to definition [ def : alpha=1-stat - round ] , a stationary round can only happen after the initial @xmath348 rounds , we have that in any stationary round the solution is feasible ( lemmas [ lemma : feasibility ] and [ lemma : self - stabilization ] ) and approximate complementary slackness ( lemma [ lemma : approx - comp - slack ] ) holds . recall the expression ( [ eq : duality - gap - proportional])for the duality gap : @xmath522 from the second part of lemma [ lemma : approx - comp - slack ] : @xmath523 therefore : @xmath524 since the round is stationary , we have that @xmath525 , which gives : @xmath526 let @xmath527 .
the remaining part of the proof is to bound @xmath528 . for @xmath529
, we have that @xmath530 . to bound the remaining terms
, we will use lemma [ lemma : cond - lower - bound ] and the bound of the sum of the weights @xmath8 for which @xmath531 ( that is , @xmath532 ) .
it follows that : @xmath533 combining ( [ eq : alpha=1-duality - gap - bound ] ) and ( [ eq : bound - on - ln - terms ] ) , and recalling that @xmath534 , the result follows .
theorem [ thm : convergence - alpha=1 ] consider the values of the potential in the rounds following the initial @xmath348 rounds , where @xmath502 , @xmath503 ( so that the solution @xmath23 is feasible in each round and the approximate complementary slackness holds ) .
observe that @xmath535 .
we start by bounding the minimum and the maximum values that the potential can take .
recall ( from lemma [ lemma : potential - increase ] ) that the potential never decreases . due to lemma
[ lemma : feasibility ] , @xmath536 $ ] , @xmath75 , and therefore we can bound the two summations in the potential as : @xmath537 @xmath538 @xmath539 and @xmath540 from ( [ eq : alpha=1-pot - lbound-1 ] ) and ( [ eq : alpha=1-pot - lbound-2 ] ) : @xmath541 on the other hand , from ( [ eq : alpha=1-pot - ubound-1 ] ) and ( [ eq : alpha=1-pot - ubound-2 ] ) : @xmath542 consider the following two cases : * case 1 : @xmath543 . * then @xmath544 and @xmath545 . from the third part of lemma [ lemma : approx - comp - slack ] , we have that @xmath546 . thus using the part [ item : prop-2 ] of lemma [ lemma : potential - increase - proportional ]
, we get that the potential increases by @xmath547 finally , since @xmath548 , there can be at most @xmath549 case 1 rounds .
* case 2 : @xmath550 . * then @xmath551 . from lemma [ lemma : alpha=1-stat - round ] ,
if a round is stationary , then @xmath552 .
if a round is non - stationary , from lemma [ lemma : non - stat - round - alpha=1 ] , the potential increases ( additively ) by at least @xmath506
. therefore , the maximum number of non - stationary rounds is at most : @xmath553 combining the results for the case 1 and case 2 , the theorem follows by invoking @xmath0-fairpsolver for the approximation parameter @xmath554 . the outline of the proof of theorem [ thm : convergence - alpha>1 ] is as follows .
first , we show that in any round of the algorithm the variables that decrease by a multiplicative factor @xmath63 dominate the potential increase due to _ all the variables _ that decrease ( lemma [ lemma : alpha>1-mul - increase - over - s- ] ) .
this result is then used in lemma [ lemma : potential - increase - alpha>1 ] to show the appropriate lower bound on the potential increase .
observe that for @xmath153 the objective function @xmath166 , and , consequently , the potential function @xmath263 is negative for any feasible @xmath23 . to yield a poly - logarithmic convergence time in @xmath555 , and @xmath33 , the idea is to show that the negative potential @xmath556 decreases by some multiplicative factor whenever @xmath23 is not a `` good '' approximation to @xmath21 the optimal solution to @xmath96 .
this idea , combined with the fact that the potential never decreases ( and therefore @xmath556 never increases ) and with upper and lower bounds on the potential then leads to the desired convergence time .
[ lemma : alpha>1-mul - increase - over - s- ] in any round of the algorithm in which the solution @xmath184 at the beginning of the round is feasible : @xmath557 and @xmath558 if @xmath462 , @xmath75 , there is nothing to prove , so assume that there exists at least one @xmath28 with @xmath559 .
the proof proceeds as follows .
first , we show that for each @xmath28 for which @xmath29 decreases by a factor less than @xmath560 there exists at least one @xmath336 that appears in at least one constraint @xmath36 in which @xmath29 appears and decreases by a factor @xmath560 .
we then proceed to show that @xmath336 is in fact such that @xmath561 and @xmath562 this will then imply that the terms @xmath563 and @xmath564 dominate _ the sum _ of all the terms corresponding to @xmath29 s with @xmath86 and @xmath565 , thus completing the proof .
from lemma [ lemma : small - x - tight - yi ] , for each @xmath289 with @xmath566 there exists at least one constraint @xmath36 such that : * @xmath567 , and * @xmath568
. therefore , there exists at least one @xmath336 with @xmath317 such that @xmath569 , which further gives @xmath570 , where the last inequality follows from @xmath571 and @xmath79 . combining the inequality for @xmath572 with the inequality for @xmath573 above : @xmath574 using the generalized bernoulli s inequality : @xmath575 and @xmath576 @xcite , and recalling that @xmath577 , @xmath578 , we further get : @xmath579 which further implies : @xmath580 as @xmath581 .
since @xmath29 decreases , @xmath582 , and therefore @xmath336 decreases as well . using similar arguments , as @xmath569 and recalling that @xmath583 : @xmath584 as @xmath585 and @xmath586 ( since @xmath587 $ ] ) .
from ( [ eq : slackness - n - dominance ] ) , it follows that @xmath588 which further implies the first part of the lemma .
for the second part , consider the following two cases : * case 1 : * @xmath589 .
then : @xmath590 implying the second part of the lemma . * case 2 : * @xmath591 .
then : @xmath592 thus implying the second part of the lemma and completing the proof .
the following lemma lower - bounds the increase in the potential , in each round .
[ lemma : potential - increase - alpha>1 ] let @xmath184 and @xmath185 denote the values of @xmath23 before and after any fixed round , respectively , and let @xmath593 , @xmath594 .
the potential increase in the round is lower bounded as : 1 .
@xmath595 ; 2 .
@xmath596 ; 3 .
@xmath597 .
@xmath288 + * proof of 1 . * from lemma [ lemma : potential - increase ] : @xmath598 let @xmath599 . from the proof of lemma [ lemma : potential - increase ] ,
if @xmath268 , then @xmath600 , as @xmath601 . if @xmath602 , then @xmath603 , which implies @xmath604 , and thus @xmath605 .
therefore : @xmath606 , which further gives : @xmath607 if @xmath296 , then @xmath300 , and therefore @xmath608 . similarly ,
if @xmath289 and @xmath462 , then @xmath609 , and therefore @xmath610 .
using part 1 of lemma [ lemma : alpha>1-mul - increase - over - s- ] : @xmath611 * proof of 2 : * consider @xmath289 such that @xmath462 .
then @xmath609 , @xmath612 , and using lemma [ lemma : potential - increase ] : @xmath613 using the second part of lemma [ lemma : alpha>1-mul - increase - over - s- ] and the fact that for @xmath614 : @xmath615 , we get the desired result : @xmath616 * proof of 3 : * the proof is equivalent to the proof of lemma [ lemma : potential - increase - alpha<1 ] , part 2 , and is omitted for brevity .
consider the following definition of a stationary round : [ def : stationary - round ] ( stationary round . )
a round is stationary , if both : 1 .
@xmath617 , and 2 .
@xmath618 hold , where @xmath593 , @xmath594 . otherwise , the round is non - stationary .
the following two technical propositions are used in lemma [ lemma : alpha>1-stationary - near - opt ] for bounding the duality gap in stationary rounds .
[ prop : alpha>1-duality - gap ] after the initial the initial @xmath348 rounds , where @xmath502 , @xmath503 , it is always true that @xmath619 , where @xmath620 .
recall from ( [ eq : duality - gap - alpha ] ) that the duality gap for @xmath461 in @xmath96 is given as : @xmath621 from lemma [ lemma : approx - comp - slack ] , after at most initial @xmath622 rounds : @xmath623 and letting @xmath113 , we get : @xmath624 [ prop : alpha>1-r - bound ] let @xmath625 , where @xmath113 . if @xmath153 and @xmath626 @xmath627 , then @xmath628 .
observe the first and the second derivative of @xmath629 : @xmath630 as @xmath631 , @xmath384 is convex for @xmath79 , and therefore : @xmath632 we have that : @xmath633 on the other hand : @xmath634 completing the proof .
the following lemma states that in any stationary round current solution is an @xmath635-approximate solution .
[ lemma : alpha>1-stationary - near - opt ] in any stationary round that happens after the initial the initial @xmath348 rounds , where @xmath502 , @xmath503 , we have that @xmath636 , where @xmath21 is the optimal solution to @xmath96 and @xmath23 is the solution at the beginning of the round .
observe that for any @xmath637 ( by the definition of @xmath295 and @xmath305 ) we have that @xmath638 , which is equivalent to : @xmath639 using stationarity and ( [ eq : tight - stationary - xk ] ) : @xmath640 since @xmath641 , using ( [ eq : bound - on - all - wj - xj ] ) : @xmath642 and therefore : @xmath643 as @xmath644 and @xmath325 . as @xmath645 , from proposition [ prop : alpha>1-duality - gap ] : @xmath646 from proposition [ prop : alpha>1-r - bound ] : @xmath647 observe @xmath648 .
since @xmath79 , each @xmath649 , and therefore : @xmath650 now , from stationarity @xmath651 and using ( [ eq : tight - active - objective - part ] ) we get : @xmath652 finally , combining ( [ eq : approx - for - passive - x ] ) and ( [ eq : approx - for - active - x ] ) : @xmath653 the following two lemmas are used for lower - bounding the potential increase in non - stationary rounds .
[ lemma : alpha>1-large - sum - yi ] consider any non - stationary round that happens after the initial @xmath348 rounds , where @xmath502 , @xmath503 .
let @xmath184 and @xmath185 denote the values of @xmath23 before and after the round update .
if @xmath654 , then @xmath655 . observe that as @xmath654 , @xmath656 where the last inequality follows from lemma [ lemma : approx - comp - slack ] . since the round is not stationary , we have that either : 1 .
@xmath657 , or 2 .
@xmath658 . * case 1 : @xmath657 . *
if : @xmath659 then @xmath660 and , from the first part of lemma [ lemma : potential - increase - alpha>1 ] , the potential increase is lower bounded as : @xmath661 on the other hand , if : @xmath662 then , from the second part of lemma [ lemma : potential - increase - alpha>1 ] : @xmath663 * case 2 : @xmath658 . * then , using the third part of lemma [ lemma : potential - increase - alpha>1 ] : @xmath664 where in the second line we have used that @xmath665 .
this can be shown using the generalized bernoulli s inequality and @xmath666 as follows : @xmath667 [ lemma : alpha>1-mul - inc - non - stat ] consider any non - stationary round that happens after the initial @xmath348 rounds , where @xmath502 , @xmath503 .
let @xmath184 and @xmath185 denote the values of @xmath23 before and after the round update .
if @xmath668 , then @xmath669 . observe that as @xmath668 , @xmath670 from the definition of a stationary round , we have either of the following two cases : * case 1 : * @xmath671 . from the first part of lemma [ lemma : potential - increase - alpha>1 ] ,
the increase in the potential is : @xmath672 . as @xmath673 ,
the increase in the potential is at least : @xmath674 * case 2 : * @xmath675 . using part 3 of lemma [ lemma : potential - increase - alpha>1 ] ,
the increase in the potential is then @xmath676 .
therefore , using that @xmath665 as in the proof of lemma [ lemma : alpha>1-large - sum - yi ] : @xmath677 theorem [ thm : convergence - alpha>1 ] we will bound the total number of non - stationary rounds that happen after the initial @xmath348 rounds , where @xmath502 , @xmath503 .
the total convergence time is then at most the sum of @xmath622 rounds and the number of non - stationary rounds that happen after the initial @xmath348 rounds , since , from lemma [ lemma : alpha>1-stationary - near - opt ] , in any stationary round : @xmath678 .
consider the non - stationary rounds that happen after the initial @xmath348 rounds .
as @xmath74 $ ] , @xmath75 , it is simple to show that : @xmath679 and @xmath680 recall that @xmath681 and that the potential @xmath263 never decreases
. there can be two cases of non - stationary rounds : those in which @xmath682 dominates in the absolute value of the potential , and those in which @xmath683 dominates in the absolute value of the potential .
we bound the total number of the non - stationary rounds in such cases as follows .
* case 1 : @xmath684 . * from ( [ eq : alpha>1-palpha - bounds ] ) and ( [ eq : alpha>1-sum - yi - bounds ] ) , in any such round , the negative potential is bounded as : @xmath685 moreover , from lemma [ lemma : alpha>1-large - sum - yi ] , in each case 1 non - stationary round , the potential increases by at least @xmath686 .
it immediately follows that there can be at most : @xmath687 case 1 non - stationary rounds , as @xmath688 .
* case 2 : @xmath689 . * from ( [ eq : alpha>1-palpha - bounds ] ) and ( [ eq : alpha>1-sum - yi - bounds ] ) , in any such round , the negative potential is bounded as : @xmath690 moreover , from lemma [ lemma : alpha>1-mul - increase - over - s- ] , in each such non - stationary round the potential increases by at least @xmath691
. therefore , there can be at most : @xmath692 case 2 non - stationary rounds .
the total number of initial @xmath622 rounds can be bounded as : @xmath693 combining ( [ eq : alpha>1-conv - bound-1 ] ) , ( [ eq : alpha>1-conv - bound-2 ] ) , and ( [ eq : alpha>1-conv - bound-3 ] ) , the total convergence time is at most : @xmath694 finally , running @xmath0-fairpsolver for the approximation parameter @xmath695 , we get that in any stationary round @xmath696 , while the total number of non - stationary rounds is at most : @xmath697 in this section , we outline the main ideas of the proof of theorem [ thm : convergence - alpha>1 ] , while the technical details are omitted and are instead provided in the full version of the paper .
first , we show that in any round of the algorithm the variables that decrease by a multiplicative factor @xmath63 dominate the potential increase due to _ all the variables _ that decrease ( see lemma 4.21 in the full paper ) .
this result is then used in lemma [ lemma : potential - increase - alpha>1 ] to show the following lower bound on the potential increase : [ lemma : potential - increase - alpha>1 ] let @xmath184 and @xmath185 denote the values of @xmath23 before and after any fixed round , respectively , and let @xmath593 , @xmath594 . the potential increase in the round
is lower bounded as : 1 .
@xmath595 ; 2 .
@xmath596 ; 3 .
@xmath597 .
observe that for @xmath153 the objective function @xmath166 , and , consequently , the potential function @xmath263 is negative for any feasible @xmath23 . to yield a poly - logarithmic convergence time in @xmath555 , and @xmath33 ,
the idea is to show that the negative potential @xmath556 decreases by some multiplicative factor whenever @xmath23 is not a `` good '' approximation to @xmath21 the optimal solution to @xmath96 .
this idea , combined with the fact that the potential never decreases ( and therefore @xmath556 never increases ) and with upper and lower bounds on the potential then leads to the desired convergence time . consider the following definition of a stationary round : [ def : stationary - round ] ( stationary round . )
a round is stationary , if both : 1 .
@xmath698 , and 2 .
@xmath618 hold , where @xmath593 , @xmath594 .
otherwise , the round is non - stationary .
recall the expression for the negative potential : @xmath699 . then , using lemma [ lemma : potential - increase - alpha>1 ] , it suffices to show that in a non - stationary round the decrease in the negative potential @xmath556 is a multiplicative factor of the larger of the two terms @xmath700 and @xmath417 . the last part of the proof is to show that the solution @xmath23 that corresponds to any stationary round is close to the optimal solution .
this part is done by appropriately upper - bounding the duality gap . denoting by @xmath701 the set of coordinates @xmath28 for which @xmath29 either increases or decreases in the observed stationary round and using definition [ def : stationary - round ] ,
we show that the terms @xmath702 contribute to the duality gap by no more than @xmath703 .
the terms corresponding to @xmath704 are bounded recalling ( from @xmath0-fairpsolver ) that for such terms @xmath705 . [
[ lower - bound - on - the - minimum - allocated - value . ] ] lower bound on the minimum allocated value .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + recall ( from section [ section : prelims ] ) that the optimal solution @xmath21 to @xmath96 must lie in the positive orthant .
we show in lemma [ lemma : lower - bound ] that not only does @xmath21 lie in the positive orthant , but the minimum element of @xmath21 can be bounded below as a function of the problem parameters .
this lemma motivates the choice of parameters @xmath65 in @xmath0-fairpsolver ( section [ section : algorithm ] ) .
[ lemma : lower - bound ] let @xmath706 be the optimal solution to @xmath96 .
then @xmath627 : * @xmath707 , if @xmath708 , * @xmath709 , if @xmath79 , where @xmath710 is the indicator function of the expression @xmath711 , i.e. , 1 if @xmath711 holds , and 0 otherwise . ] is the number of non - zero elements in the @xmath712 row of the constraint matrix @xmath9 , and @xmath713 .
fix @xmath0 .
let : @xmath714 for the purpose of contradiction , suppose that @xmath706 is the optimal solution to @xmath96 , and @xmath715 for some fixed @xmath116 . to establish the desired result , we will need to introduce additional notation .
we first break the set of ( the indices of ) constraints of the form @xmath716 in which variable @xmath29 appears with a non - zero coefficient into two sets , @xmath717 and @xmath718 : * let @xmath717 denote the set of the constraints from @xmath96 that are not tight at the given optimal solution @xmath21 , and are such that @xmath719 for @xmath720 .
let @xmath721 denote the slack of the constraint @xmath720 .
* let @xmath718 denote the set of tight constraints from @xmath96 that are such that @xmath722 for @xmath723 .
observe that since @xmath21 is assumed to be optimal , @xmath724 .
let @xmath725 . notice that by increasing @xmath29 to @xmath726 none of the constraints from @xmath717 can be violated ( although all the constraints in @xmath718 will ; we deal with these violations in what follows ) . in each constraint @xmath723 , there must exist at least one variable @xmath203 such that @xmath727 , because @xmath728 , as each @xmath723 is tight , and @xmath729 select one such @xmath203 in each constraint @xmath723 , and denote by @xmath730 the set of indices of selected variables . observe that @xmath731 ( @xmath732 ) , since an @xmath203 can appear in more than one constraint . for each @xmath733 ,
let @xmath734 denote the constraints in which @xmath203 is selected , and let @xmath735 if we increase @xmath29 by @xmath736 and decrease @xmath203 by @xmath737 @xmath738 , each of the constraints @xmath723 will be satisfied since , from ( [ eq : epsilon - choice ] ) and from the fact that only one @xmath203 gets selected per constraint @xmath723 , @xmath739 .
therefore , to construct an alternative feasible solution @xmath740 , we set @xmath741 , @xmath742 for @xmath733 , and @xmath743 for all the remaining coordinates @xmath744 .
since @xmath28 is the only coordinate over which @xmath23 gets increased in @xmath740 , all the constraints @xmath745 are satisfied . for @xmath740 to be feasible , we must have in addition that @xmath746 for @xmath733 . we show that @xmath747 as follows : @xmath748 where the second line follows from @xmath749 , and the last line follows from the choice of @xmath203 . the last part of the proof is to show that @xmath750 , which contradicts the initial assumption that @xmath21 is optimal , by the definition of @xmath0-fairness from section [ section : prelims ] .
we have that : @xmath751 consider one term from the summation ( [ eq : sum - ind - terms ] ) . from the choice of @xmath737 s , we know that for each @xmath737 there exist @xmath723 such that @xmath752 , and at the same time ( by the choice of @xmath203 ) we have @xmath753 , so that @xmath754 * case 1 .
* suppose first that @xmath755 .
then @xmath756 , as @xmath757 .
plugging into ( [ eq : xk - general ] ) , we have : @xmath758 by the initial assumption , @xmath759 , and therefore @xmath760 since it must be @xmath761 ( @xmath762 ) . from ( [ eq : xk ] ) and ( [ eq : xj ] )
, we get that every term in the summation ( [ eq : sum - ind - terms ] ) is strictly positive , which implies : @xmath763 and therefore @xmath21 is not optimal . * case 2 .
* now suppose that @xmath79 . then @xmath764 therefore : @xmath765 as @xmath761 , and @xmath766 ( since for any @xmath767 : @xmath768 ) .
finally , from ( [ eq : xk - general ] ) and ( [ eq : xj - alpha1 ] ) we get that every term in the summation ( [ eq : sum - ind - terms ] ) is positive , which yields a contradiction . [
[ asymptotics - of - alpha - fair - allocations ] ] asymptotics of @xmath13fair allocations + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the following lemma states that for sufficiently small ( but not too small ) @xmath0 , the values of the linear and the @xmath13fair objectives at their respective optimal solutions are approximately the same .
this statement will then lead to a conclusion that to @xmath11approximately solve an @xmath13fair packing problem for a very small @xmath0 , one can always use an @xmath11approximation packing lp algorithm .
[ lemma : lp - close - to - small - alpha - fair ] let @xmath370 be an @xmath13fair packing problem with optimal solution @xmath21 , and @xmath769 be the lp with the same constraints and the same weights @xmath19 as @xmath370 and an optimal solution @xmath54 . then if @xmath48 , we have that @xmath770 , where @xmath771 $ ] .
the proof outline is as follows .
first , we show that the @xmath13fair objective @xmath772 can be upper - bounded by a linear objective as @xmath773 .
then , to complete the proof , we use the optimality of @xmath54 for the lp : @xmath774 ( @xmath775 from the first part of the proof ) .
let @xmath776 .
consider the case when @xmath777 .
solving @xmath777 for @xmath29 , we get that it should be @xmath778 choose @xmath0 so that @xmath779 , which is equivalent to @xmath780
. then to have @xmath777 , it suffices to have @xmath781 , because ( i ) @xmath782 $ ] for @xmath783 $ ] , where @xmath711 is the base of the natural logarithm , and ( ii ) @xmath779 by the choice of @xmath0 .
now , as @xmath780 , summing over @xmath28 such that @xmath784 we have : @xmath785 now we bound the rest of the terms in @xmath772 , i.e. , we consider @xmath786 .
observe that since @xmath787 for @xmath788 is a feasible solution to @xmath370 and @xmath21 is the optimal solution to @xmath370 , we have that @xmath789 , which gives : @xmath790 therefore : @xmath791 combining ( [ eq : large - coord - approx ] ) and ( [ eq : small - coord - approx ] ) , we now get : @xmath792 finally , since @xmath54 optimally solves @xmath769 ( which has the same constraints and weights as @xmath370 ) , we have that @xmath21 is feasible for @xmath769 , and using ( [ eq : alpha - lp ] ) and optimality of @xmath54 , it follows that : @xmath793 as claimed . observing that for any @xmath794 , @xmath795 ( since
, due to the scaling , @xmath796 $ ] ) , a simple corollary of lemma [ lemma : lp - close - to - small - alpha - fair ] is that an @xmath11approximation @xmath797 to @xmath769 ( @xmath798 ) is also an @xmath49approximation to @xmath370 , for @xmath48 .
thus , to find an @xmath11approximate solution for @xmath48 , the packing lp algorithm of @xcite can be run , which means that there is a stateless distributed algorithm that converges in poly(@xmath799 ) time for @xmath0 arbitrarily close to zero .
the following two lemmas show that when @xmath0 is sufficiently close to 1 , @xmath370 can be @xmath11approximated by @xmath11approximately solving @xmath172 with the same constraints and weights .
[ lemma : alpha - close - to-1-below ] let @xmath23 be an @xmath11approximate solution to a 1-fair packing problem @xmath172 returned by @xmath0-fairpsolver .
then , for any @xmath800 , where @xmath349 , @xmath23 is also a @xmath801approximate solution to @xmath96 , where the only difference between @xmath172 and @xmath96 is in the value of @xmath0 in the objective .
suppose that @xmath23 is a solution in some stationary round , provided by @xmath0-fairpsolver run for @xmath5 .
fix that round .
it is clear that if @xmath23 is feasible in @xmath172 , it is also feasible in @xmath96 , since all the constraints in @xmath172 and @xmath96 are the same by the initial assumption .
all that is required for a dual solution @xmath80 to be feasible is that @xmath802 , for all @xmath36 , and therefore @xmath803 is a feasible dual solution for @xmath96 .
the rest of the proof follows by bounding the duality gap @xmath804 .
recall from ( [ eq : duality - gap - alpha ] ) that : @xmath805 since @xmath23 is a solution from a stationary round , from the second part of the definition of a stationary round ( definition [ def : alpha=1-stat - round ] ) , we have that : @xmath806 further , from lemma [ lemma : approx - comp - slack ] : @xmath807 next , we show that : @xmath808 rearranging the terms and taking logarithms of both sides in ( [ eq : alpha - below=1-all - x - large ] ) , we obtain the equivalent inequality @xmath809 .
recall from @xmath0-fairpsolver that in every ( except for , maybe , the first ) round @xmath810 .
as @xmath811 , it therefore suffices to show that @xmath812 .
but from the statement of the lemma , @xmath813 , completing the proof of ( [ eq : alpha - below=1-all - x - large ] ) . combining ( [ eq : alpha - below-1-acs ] ) and ( [ eq : alpha - below=1-all - x - large ] ) , we get that : @xmath814 where the second inequality follows from @xmath815 , @xmath816 . using ( [ eq : alpha - below-1-bnd-1 ] ) , we can bound the duality gap ( eq .
( [ eq : duality - gap - alpha - recap ] ) ) as : @xmath817 to complete the proof , recall from lemma [ lemma : cond - lower - bound ] that in any round of the algorithm , for all @xmath28 : @xmath818 . as @xmath137 and
@xmath819 $ ] , @xmath75 , it holds that @xmath820 , @xmath75 , and therefore : @xmath821 finally , recalling that @xmath822 , and combining ( [ eq : alpha - below-1-bnd-2 ] ) with ( [ eq : duality - gap - alpha - betterbound ] ) , we get : @xmath823 where the third inequality follows from @xmath824 , and the fourth inequality follows from @xmath825 and @xmath826 .
[ lemma : alpha - close - to-1-above ] let @xmath23 be an @xmath11approximate solution to a 1-fair packing problem @xmath172 returned by @xmath0-fairpsolver .
then , for any @xmath827 $ ] , where @xmath349 , @xmath23 is also a @xmath801approximate solution to @xmath96 , where the only difference between @xmath172 and @xmath96 is in the value of @xmath0 in the objective .
similar to the proof of lemma [ lemma : alpha - close - to-1-below ] , we will fix an @xmath23 from some stationary round of @xmath0-fairpsolver run on @xmath172 , and argue that the same @xmath23 @xmath801approximates @xmath96 by bounding the duality gap @xmath828 , although we will need to use a different set of inequalities since now @xmath153 .
similar to the proof of lemma [ lemma : alpha - close - to-1-below ] , as @xmath23 is ( primal-)feasible for @xmath172 , @xmath23 and @xmath803 are primal- and dual - feasible for @xmath96 . by the same token as in the proof of lemma [ lemma : alpha - close - to-1-below ] : @xmath829 as @xmath153 and @xmath830 $ ] , @xmath75 , we have that @xmath831 , @xmath75 , and therefore : @xmath832 therefore , we can write for the duality gap : @xmath833 notice that , as @xmath834 , the objective for @xmath96 , @xmath835 , is now negative . using
the same arguments as in the proof of lemma [ lemma : alpha - close - to-1-below ] , it is straightforward to show that @xmath836 , @xmath75 . from lemma [ lemma :
cond - lower - bound ] , we have that @xmath837 , @xmath75 , and therefore : @xmath838 recalling that @xmath839 ( by the statement of the lemma ) and using ( [ eq : alpha - above-1-bnd-2 ] ) , we have : @xmath840 finally , plugging ( [ eq : alpha - above-1-bnd-3 ] ) into ( [ eq : alpha - above-1-duality - gap ] ) , we have : @xmath841 where the equality follows from @xmath644 , and the last inequality follows from @xmath842 . finally , we consider the asymptotics of @xmath13fair allocations , as @xmath0 becomes large .
this result complements the result from @xcite that states that @xmath13fair allocations approach the max - min fair one as @xmath843 by showing how fast the max - min fair allocation is reached as a function of @xmath844 , and @xmath33 .
first , for completeness , we provide the definition of max - min fairness .
( max - min fairness @xcite . )
[ def : max - min - fairness ] let @xmath845 be a compact and convex set .
a vector @xmath846 is max - min fair on @xmath847 if for any vector @xmath848 it holds that : if for some @xmath116 @xmath849 , then there exists @xmath850 such that @xmath851 and @xmath852 . on a compact and convex set @xmath853 ,
the max - min fair vector is unique ( see , e.g. , @xcite ) .
the following lemma shows that for @xmath854 , the @xmath13fair vector and the max - min fair vector are @xmath11close to each other .
notice that because of a very large gradient of @xmath166 as @xmath0 becomes large , the max - min fair solution provides only an @xmath855approximation to @xmath96 .
[ lemma : mmf - alpha - fair ] let @xmath21 be the optimal solution to @xmath856 , @xmath54 be the max - min fair solution for the convex and compact set determined by the constraints from @xmath96 . then if @xmath857 , we have that : 1 .
@xmath858 , i.e. , @xmath54 is an @xmath859approximate solution to @xmath96 , and 2 .
@xmath860 , for all @xmath861 .
suppose that , starting with @xmath54 , we want to construct a solution @xmath797 that is feasible in @xmath96 and is such that @xmath862 .
then we need to increase at least one coordinate @xmath28 of @xmath54 .
suppose that we increase a coordinate @xmath28 by a factor @xmath863 , so that @xmath864 .
since @xmath54 is the max - min fair vector , to keep @xmath797 feasible , the increase over the @xmath865 coordinate must be at the expense of decreasing some other coordinates @xmath511 that satisfy @xmath866 .
we will assume that whenever we decrease the coordinates to keep the solution feasible , we keep the solution pareto optimal ( i.e. , we decrease the selected coordinates by a minimum amount ) .
using fact [ fact : taylor ] , we have : @xmath867 now , suppose that we want to further increase the @xmath865 coordinate by some small @xmath868 .
call that new solution @xmath869 . then
, the total amount by which other coordinates must decrease to keep the solution feasible is at least @xmath870 , since the feasible region is determined by packing constraints and it must be @xmath871 , where @xmath872 , @xmath435 .
moreover , since @xmath54 is max - min fair , each coordinate @xmath511 that gets decreased must satisfy @xmath866 .
it follows that : @xmath873 the last inequality can be verified by solving the inequality @xmath874 for @xmath0 , and verifying that it is implied by the initial assumption that @xmath854 .
therefore , the maximum amount by which any coordinate of @xmath54 can be increased to improve the value of the objective @xmath875 is by a multiplicative factor of at most @xmath41 .
since we can construct @xmath21 , the optimal solution to @xmath96 , starting with @xmath54 and by choosing a set of coordinates @xmath28 that we want to increase and by only decreasing coordinates @xmath511 such that @xmath866 whenever coordinate @xmath28 is increased , it follows that @xmath876 , @xmath75 .
moreover , from ( [ eq : mmf - part-1 ] ) and ( [ eq : mmf - part-2 ] ) : @xmath877 and we can conclude that : @xmath878 which means that @xmath54 is an @xmath859approximate solution to @xmath96 .
now consider the coordinates we need to decrease when we construct a solution @xmath797 from @xmath54 , such that @xmath862 .
suppose that to increase some other coordinates , a coordinate @xmath511 is decreased by a factor @xmath879 : @xmath880 . as @xmath54 is max - min fair , only coordinates larger than @xmath881 can increase at the expense of decreasing @xmath881 .
suppose now that we decrease the @xmath882 coordinate further by some small @xmath868 .
call that solution @xmath869
. then the maximum number of other coordinates @xmath28 that can further increase is @xmath883 .
moreover , each coordinate @xmath28 that gets increased satisfies @xmath884 , and can be increased by at most @xmath885 .
using fact [ fact : taylor ] , it follows that : @xmath886 where the last inequality follows from @xmath887 , which is implied by the initial assumption that @xmath161 .
therefore , using ( [ eq : mmf - part-5 ] ) , the @xmath882 coordinate can decrease by at most a multiplicative factor @xmath879 . using similar arguments as for increasing the coordinates
, it follows that @xmath888 , @xmath75 .
we presented an efficient stateless distributed algorithm for the class of @xmath0-fair packing problems . to the best of our knowledge ,
this is the first algorithm with poly - logarithmic convergence time in the input size . additionally , we obtained results that characterize the fairness and asymptotic behavior of allocations in weighted @xmath13fair packing problems that may be of independent interest .
an interesting open problem is to determine the class of objective functions for which the presented techniques yield fast and stateless distributed algorithms , together with a unified convergence analysis .
this problem is especially important in light of the fact that @xmath0-fair objectives are not lipschitz continuous , do not have a lipschitz gradient , and their dual gradient s lipschitz constant scales at least linearly with @xmath22 and @xmath33 .
therefore , the properties typically used in fast first - order methods are lacking @xcite . finally , for applications of @xmath0-fair packing that do not require uncoordinated updates
, it seems plausible that the dependence on @xmath78 in the convergence bound can be improved from @xmath889 to @xmath890 by relaxing the requirement for asynchronous updates , similarly as was done in @xcite over @xcite .
we thank nikhil devanur for pointing out the equivalence of the @xmath0-fair packing for @xmath5 and the problem of finding an equilibrium allocation in eisenberg - gale markets with leontief utilities .
let the @xmath0-fair allocation problem be given in the form : @xmath891 @xmath19 is an @xmath892length vector of positive weights , @xmath23 is the vector of variables , @xmath9 is an @xmath893 constraint matrix , and @xmath10 is an @xmath894length vector with positive entries .
denote @xmath895 .
it is not hard to see that the assumption @xmath896 @xmath76 is without loss of generality , since for @xmath897 we can always divide both sides of the inequality by @xmath107 and obtain 1 on the right - hand side , since for ( non - trivial ) packing problems @xmath898 .
therefore , we can assume that the input problem has constraints of the form @xmath899 , although it may not necessarily be the case that @xmath40 @xmath900 .
the remaining transformation that is performed on the input problem is : @xmath901 where @xmath902 the problem @xmath87 after the scaling becomes : @xmath903 @xmath904 @xmath905 + as @xmath906 is a positive constant .
recall that @xmath0-fairpsolver returns an approximate solution to @xmath907 , and observe that @xmath23 is feasible for @xmath87 if and only if @xmath908 is feasible for @xmath96 .
choose the dual variables ( lagrange multipliers ) for the original problem @xmath87 as : @xmath909 and notice that @xmath910 it is clear that @xmath66 s are feasible dual solutions , since the only requirement for the duals is non - negativity .
recall ( from ( [ eq : duality - gap - alpha ] ) ) that the duality gap for a given primal- and dual - feasible @xmath23 and @xmath80 is given as : @xmath911 since @xmath58 , we have that @xmath912 for all @xmath36 , and using ( [ eq : condition - equality ] ) , it follows that @xmath913 since we demonstrate an additive approximation for the proportional fairness via the duality gap : @xmath914 , the same additive approximation follows for the original ( non - scaled ) problem . for @xmath915 , we show that the algorithm achieves a multiplicative approximation for the scaled problem .
in particular , we show that after the algorithm converges we have that : @xmath916 , where @xmath917 is the optimal solution , @xmath908 is the solution returned by the algorithm , and @xmath918 is a constant .
observe that since @xmath919 , we have that @xmath920 and @xmath921 . therefore : @xmath922
the lagrangian for this problem can be written as : @xmath925 where @xmath926 are lagrange multipliers , and @xmath927 are slack variables . the dual to this problem is : @xmath928 where @xmath929 . to maximize @xmath930 ,
we first differentiate with respect to @xmath29 , @xmath116 : @xmath931 which gives : @xmath932 plugging this back into the expression for @xmath930 , and noticing that , since @xmath933 @xmath934 , @xmath930 is maximized for @xmath935 , we get that : @xmath936 since @xmath937 @xmath938 , and @xmath939 .
let @xmath940 denote the primal objective .
the duality gap for any pair of primal - feasible @xmath23 and dual - feasible ( nonnegative ) @xmath80 is given by : @xmath941 since the primal problem maximizes a concave function over a polytope , the strong duality holds @xcite , and therefore @xmath942 for any pair of primal - feasible @xmath23 and dual - feasible @xmath80 , with equality if and only if @xmath23 and @xmath80 are primal- and dual- optimal , respectively .
|
over the past two decades , fair resource allocation problems have received considerable attention in a variety of application areas . however , _ little progress has been made in the design of distributed algorithms with convergence guarantees for general and commonly used @xmath0-fair allocations_. in this paper , we study weighted @xmath0-fair packing problems , that is , the problems of maximizing the objective functions ( i ) @xmath1 when @xmath2 , @xmath3 and ( ii ) @xmath4 when @xmath5 , over linear constraints @xmath6 , @xmath7 , where @xmath8 are positive weights and @xmath9 and @xmath10 are non - negative .
we consider the distributed computation model that was used for packing linear programs and network utility maximization problems . under this model ,
_ we provide a distributed algorithm for general @xmath0 _ that converges to an @xmath11approximate solution in time ( number of distributed iterations ) that has an inverse polynomial dependence on the approximation parameter @xmath12 and poly - logarithmic dependence on the problem size .
_ this is the first distributed algorithm for weighted @xmath13fair packing with poly - logarithmic convergence in the input size . _
the algorithm uses simple local update rules and is stateless ( namely , it allows asynchronous updates , is self - stabilizing , and allows incremental and local adjustments ) .
we also obtain a number of structural results that characterize @xmath13fair allocations as the value of @xmath0 is varied .
these results deepen our understanding of fairness guarantees in @xmath13fair packing allocations , and also provide insight into the behavior of @xmath13fair allocations in the asymptotic cases @xmath14 , @xmath15 , and @xmath16 .
|
one of the methods used for meteor detection is the radio meteor observation @xcite .
the principle is based on the electromagnetic waves reflection from the meteor s trail in the atmosphere , which is called the meteor scattering .
when a meteor enters the atmosphere , it ionizes the electrons in atmospheric e - layer ( 90 - 150 km , ionosphere layer ) and leaves a plasma trail which , for a short time , works like a mirror for radio waves .
although the effect of the scattering is the strongest for 20 - 50 mhz frequencies , it is possible to receive reflected signals from distant media radio stations operating in higher frequency range ( 90mhz and above ) .
this makes the phenomenon observable even by amateurs with a relatively low - cost equipment .
the phenomenon of scattering is utilized extensively for meteor burst communication , which is a radio propagation mode allowing to establish short communication windows between distant stations . while the service of local media stations ( like tv or fm ) extends only slightly the optical horizon , reaching tens of kilometers , the maximum range the meteor scattering allows reception of signals from stations up to 2000 km away @xcite .
the phenomenon is visualized in fig .
[ f1 ] . the number of meteoroids entering the atmosphere per hour changes periodically during the day .
while in the morning we can observe the meteors approaching the earth and the ones which are slower than our planet , in the evening only the meteoroids faster than the earth enter the atmosphere .
this fact causes characteristic diurnal variations of the meteor rate .
the principle of this phenomenon is depicted in fig .
[ f2 ] .
in contrast to visual techniques , like video registration , which can not be carried out during the day , full moon or cloudy weather , the radio observations of meteors are of great interest as they can be performed continuously .
the obtained count histogram is scientifically valuable and provides a number of meteor characteristics which require the analysis of the underdense and the overdense meteor distributions @xcite .
we would like to note that the overdense and the underdense meteors are the results of respectively heavy and light particles entering the atmosphere .
the main coefficient - the mass index - provides information about the composition of meteoroids and their origins .
while the high index is observed for showers from remnants of comets , the low index suggests the planetoid origins . the brief review of tasks , which may be accomplished employing the count histogram , is given below .
1 . the time occurrences of maximum of meteor showers are compared with the predictions to obtain the time - shifts whose origins have to be investigated .
new meteor showers , observed in the form of abnormal activity , may be discovered .
the overdense meteor mass distribution may be obtained by plotting logarithm t ( the duration of phenomena ) against the logarithm of the number of meteors with duration exceeding t. only the phenomena longer than 0.5 seconds have to be included here .
the slope should be @xmath0 , where @xmath1 is the mass index of overdense meteors ( see the left plot in fig [ lf3 ] ) .
the mean mass index , which includes information from small and large particles , may be derived from the dependency between the logarithm of phenomenon amplitude and the number of detections above the given amplitude .
the method of calculation of mean mass index @xmath2 is explained graphically on the right plot of fig .
one can obtain required data in two ways : directly , by investigating detected signal amplitude , or indirectly , by utilizing parallel detectors set to various sensitivities .
it has to be mentioned that by using a more sophisticated transmitter - receiver pair ( active radar ) some further properties of the meteoroids can be obtained .
the array of receivers and high - speed signal correlation computations can provide information about meteoroid velocity vector @xcite .
moreover , some observations of the ablation and the fragmentation of the meteoroids are possible with a specially designed radar system @xcite . in our project , we concentrate on a passive radio detection circuit for polish fireball network ( pfn ) research activities .
the pfn is the organization which continuously monitors meteor showers and fireballs over poland @xcite .
the network consists of several base stations located all over the country , equipped with high - speed camera systems , recording and analyzing images in real time .
they also use passive radio detection methods to enable observations during bad weather conditions or during the day @xcite . in fig .
[ f4 ] we show an exemplary monthly radio count histogram provided by pfn , which includes quadrantids meteor shower .
for this purpose they used a pc computer , specially developed software ( metan ) and a realistic dx-394 receiver tuned to distant , east european radio stations .
the pfn pc computers running 24 hours a day and detecting meteor radio signals are neither energy efficient nor portable solutions .
it is to be noted that while a single station is still able to provide useful information about meteors , a system of multiple detectors , located at different sites and directed to various distant stations , may significantly improve the measurement efficiency and extend the analysis capabilities .
these two main issues are discussed in detail below . 1 .
firstly , we should note that the detector robustness depends on local electromagnetic environment .
the sites far from strong radio stations are generally preferred since there are fewer false detections resulting from interfering frequencies .
however , the best sites usually suffer from lack of access to the power , thus ideally the detection module should be self - sufficient e.g. powered by solar cells . 2 .
the reception of reflected signal depends on the relative position of the receiver , meteor trail and remote radio station . by directing the detector toward different stations we would gain additional information about the brief position of trail relatively to the detector
moreover , a distributed system of detectors , spread over hundreds of kilometers , should provide far better meteor localization capabilities and would work like a radar .
we plan to investigate this issue in the future . in the article
we introduce a dedicated detection electronic circuit , easy connectible to any radio receiver , which can work as a module in a remote , battery powered , meteor scatter detector . for our task
we selected a field programmable analog array ( fpaa ) manufactured by anadigm .
we also intend to show that for such a purpose an analogue array is an interesting alternative to commonly used digital signal processing units .
the circuit is dedicated to planned distributed system of remote meteor detectors .
we also consider the amateur application of our module . with a growing number of astro - amateurs and dedicated self - sufficient equipment , it may be possible to explore the meteor phenomena as it is performed for variable stars in the american association of variable star observers ( aavso ) @xcite .
the article is organized as follows . in chapter 2
we introduce the internal structure of the anadigm fpaa circuit and its principle of operation .
our signal chain implemented for meteor detection is explained in detail in chapter 3 .
finally , in chapter 4 , we describe our verification methodology based on the comparison with the results obtained by metan software , which is currently used for meteor scatter detection by the polish fireball network .
the results of the tests are presented and analyzed in chapter5 .
we summarize the results of our research in chapter 6 .
the an231e04 device represents the third generation of the dynamically reconfigurable fpaa anadigmapex offered by anadigm @xcite .
it is powered by 3.3 v supply voltage .
the previous fpaa family , called anadigmvortex , was powered by 5v , but its dc performance was much worse .
the fpaa circuit consists of four configurable analog blocks ( cab ) built on the switched - capacitance principle , surrounded by a network of programmable connections .
there are also seven programmable input / output cells on the silicon chip .
three of them can have access to a specialized chopper amplifier resource , which allows accurate amplification of very low energy input signals .
all internal signal paths of the block are differential .
io cells contain both passive and active circuitry , which allows direct signal input and output , building active filters , sample and hold circuits , digital inputs , and digital outputs .
the frequency response of continuous time input and output filters is determined by a combination of internal programming and external components .
the block diagram of the an231e04 device is shown in fig
. [ f5 ] , where one can see its basic elements : configurable analog blocks ( cab ) , input and output cells , look up table ( lut ) , clocking circuits , communication interfaces and reference voltage sources .
the new an231e04 devices can be reconfigured dynamically ( on the run ) , which enables e.g. auto ranging , auto calibration and multimode work ( allows to build many functions in a single ic ) .
the microcontroller connected to the an231e04 device can load the new device configuration data to the device still working in its previous configuration .
the configuration data can then be swept in a single clock cycle .
this new and prospective feature of the dynamic reconfiguration ability of the an231e04 devices enables the creation of a modern analog system , which can be ( completely or in part ) rebuilt in real time .
we chose this particular fpaa because of its proved very good capability of audio signal processing @xcite . by using different configurations and changing individual parameters we were able to create a circuit well suited for our purpose .
such designing flexibility makes fpaa a very attractive prototyping device .
moreover , the fpaa devices have been recently used as hardware detectors of qrs complexes in the ecg signal @xcite , which is an application similar to our audio meteor detector . in comparison with popular field programmable gate arrays ( fpga ) , which also might be utilized as a meteor detector , fpaas
have several advantages .
below , we enumerate the most crucial ones for our project . 1 .
fpaa devices introduce much lower electromagnetic disturbance , which should be considered when analyzing weak analogue signals @xcite .
the implementation process is very simple and efficient as the construction of signal path may be accomplished with an easy - to - use , drag - and - drop , diagram based environment - _ anadigmdesigner _
( fpgas have to be programmed using dedicated vhdl or verilog register transfer languages ) .
nearly all modifications of the existing signal chain may be tested by changing single values in user - friendly environment ( the properties are clear and simple , like filter bandwidth or gain ) .
4 . the analogue filters implemented in fpaa are more reliable and efficient in contrast to their digital versions ( the infinite response iir digital filters suffer from low stability while the finite response fir implementations , due to the high number of parallel calculations , usually require specialized high - efficiency units , like digital signal processors dsp )
the fpaas do not require any analogue - to - digital conversions since they operate in the analogue domain , therefore the number of elements and the total cost of the circuit is lower .
our signal chain consists of several blocks ( see fig.[f6 ] ) : primary band - pass filter 10hz - 600hz ( bpf ) , full - wave rectifier ( rect ) , secondary filters ( lpf1 , lpf2+rclpf ) , primary comparator ( comp1 ) and analogue realization of the time off - delay function ( toff ) .
we used nearly all available analogue cells of anadigm fpaa .
as the analyzed signal is in the audio frequency range , we decided to choose a primary band - pass filter 10 - 600 hz .
it is due to the fact that most of the spectral components are within this range . as a proof
, we present a plot of a part of registered radio signals in the time - frequency domain in fig .
the corresponding audio signals are the representative examples obtained by the pfn radio meteor station in our experiment .
we depicted various phenomena , with long and short duration , high and low amplitudes .
we checked that all investigated audio signals in our recording were well within the assumed frequency range .
the secondary filters ( lpf2 with rc lpf ) provide information about the mean level of the signal power .
we constantly compare the output of 20 hz filter lpf1 with the biased ( tr - threshold ) output of the 0.1 hz filter lpf2 .
the biasing prevents the false detections induced by temporal noise fluctuations .
both filters outputs represent the level of signal power : the first one detects rapid level changes , the second one shows a long - term trend .
the first comparator ( comp1 ) is set high when the lpf1 output signal exceeds the lpf2 output .
comp1 output is our first - stage detector ( raw out ) .
however , we observed that during a single radio station signal appearance the raw out output had several positive detections , although there should be only one meteor phenomenon detected . a solution to this problem is provided by an analogue version of a time off - delay ( toff ) circuit .
it is realized with a filtered peak detector ( peak det ) , a reference signal and a comparator .
when the positive edge of the first comparator ( comp1 ) appears , the output of the peak detector ( peak det ) and the output of the whole device ( toff out ) immediately rise and are held as long as the output of the first comparator ( comp1 ) remains high .
when it switches to the low state , the output of the peak detector ( peak det ) slowly falls down as an rc circuit in the peak detector discharges .
as long as the reference ( ref ) is not reached , the second comparator ( comp2 ) remains high .
when the reference level is reached , the second comparator ( comp2 ) changes the output state to low , but if there is another detection in the meantime , the device output is not affected and remains high .
adjusting the reference ( ref ) or the rc time constant in the peak detector can change the time delay .
we decided to choose a 3 second interval , which is also our estimation of the minimal interval between the subsequent meteors .
the disadvantage of this method - the time delay of the detection end - can be easily overcome by reducing measured length of the phenomena by a corresponding and constant time interval .
we would like to note that the detector will produce an additional detection if there is a longer than 3-second silence in the radio program .
it may be easily mitigated by extending the delay ; however , it also reduces the detector ability to distinguish close events .
the principle of operation is presented in fig .
we used two meteor scatter audio signals - a long one ( on the left ) and a short one ( on the right ) - and showed the most important waveforms obtained inside the designed fpaa circuit .
the figures present the screenshots registered by the tectronix tds2014b digital oscilloscope .
the detections start when the signal from lpf1 output exceeds lpf2 output .
the operation of the analogue time off - delay circuit is clearly visible on the peak det out and the toff out waveforms .
we used two - hour recording provided by the pfn .
their equipment for our tests included a 5-element log - periodic antenna ( fig .
[ f10 ] ) , 9dbi gain , provided by vpa - system .
the antenna was directed toward a distant belorussian transmitter .
we provide all necessary transmitter s information in tab .
[ tab1 ] . an audio signal from the realistic pro and dx-394 receiver
tuned to 70.10 mhz was captured and saved on a pc computer as a _ wav _ audio file .
since the polish broadcast fm radio stations use the frequency band of 88 - 108 mhz , none of them should affect the measuring meteor scatter audio signals and cause false detections .
p5cm|p5 cm location & mahilyow / rtps polykovici + longitude & 30@xmath3 19 36.41 +
latitude & 53@xmath3 59 25.92 + frequency & 70.10 mhz + program & br radio mahilyow + power & 4.00 kw + azimuth from receiver & 67@xmath3 + distance from receiver & 667 km + as the detection algorithms implemented in the metan software and in our detector are highly dependent on a threshold level , we decided to repeat the detection procedure for several values of the threshold .
we did it for both the metan software and our hardware detector . for this purpose we used several parts of
previously mentioned recording , to reduce the measurement time for each sensitivity level .
the constructed , well defined , test signal included 40 clearly audible signals extracted from 2-hour recording .
they were combined with the proceeding and following 10 seconds , thus our test recording was about 15 minutes long ( instead of 2 hours ) .
the resulting dependencies of the number of detections and the total detections duration are presented in fig .
[ f1112 ] .
it is visible that each resulting curve shows a characteristic break related to the rising number of false detections due to the noise level comparable to the threshold .
we decided to set both detectors to their maximum and medium sensitivities , which are represented by the vertical solid and dashed lines in both corresponding figures .
firstly , we selected the maximum sensitivity thresholds , for which the number of obtained detections was as close as possible to the number of given phenomena ( 40 ) .
it also let us estimate the thresholds that should not be exceeded to avoid a rapidly rising number of false detections .
the medium sensitivity thresholds were chosen for the number of detections equal to about 30 .
it corresponds to the situation , where the detectors have a slightly higher detection threshold and consequently they miss about 10 weakest signals .
we would like to note , that the total detections duration for metan software is significantly larger because we did not know the additional time interval added after each detection . in the case of our hardware detector
the delay of detection end was known - exactly 3 seconds - thus we were able to remove this additional offset .
finally , the main recording ( 2 hours long ) was given at the fpaa input and at the line - in input of a pc computer which was running the metan software detector .
we observed simultaneously the output of both detectors and the input signal .
the detections for both medium and maximum sensitivities were compared .
all events detected by both detectors were carefully examined using headphones to check if the audio signal was audible . in this way
we classified detections as the false and the true ones .
the results are presented in fig .
[ f1314 ] . each event time can be computed by adding timea and timeb .
the total duration of our recording ( 2 hours ) was reduced to 6500 seconds in fig .
[ f1314 ] as the last part did not include any detection .
we chose full circles and squares to identify false detections of respectively fpaa and metan detector , empty circle markers for positive fpaa detections and crosses for metan positive detections .
the results for both sensitivities show that many of the detections overlap .
almost all detections of the metan software were also detected by fpaa ( except for 4 and 2 detections respectively for maximum and medium sensitivities ) , which confirms that fpaa covers the domain of metan .
however , we would like to indicate that our detector found a lot of additional events in the full recording , which were not recognized by the software detector ( red circles without blue crosses inside , respectively 22 for maximum and 8 for medium sensitivity ) .
moreover , the number of false positive detections is higher for the metan analyzer , thus our detector is more resistant to noise fluctuations , which may influence the detector s decision .
we summarize the results in table [ tab2 ] .
the quality of measurement , hence the quality of obtained results , depends on both the number of true detections ( the signal ) and the number of false detections ( the noise floor ) .
ideally , the detector should maximize the signal level while keeping the low noise floor ( i.e. maximize the signal - to - noise ratio ) .
the results of our experiment for both sensitivities prove that this done much better in our solution than in the competitive detector .
p2.7cm|p1.5cmp1.5cmp1.5cm|p1.5cmp1.5cmp1.5 cm & & + & fpaa&metan&common&fpaa&metan&common + true positive&29&23&21&55&37&33 + false positive&0&0&0&1&3&0 +
in the article we presented a hardware meteor radio detector which can be attached to any radio device to detect the meteor scatter phenomena . the signal processing circuit of our detector was implemented in anadigm field programmable analog array ( fpaa ) using nearly all its available resources .
a signal path contains a rectifier , a peak detector , several comparators , low - pass and band - pass filters .
the output signals of the filter provide information about signal power in the audio frequency range . additionally , the implemented time off - delay analogue circuit prevents multiple detections during long scatter phenomena .
such an fpaa device enabled flexible signal flow designing to fulfill the detector requirements .
we compared the efficiency of detection of our implementation with the current solution used by the polish fireball network ( pfn ) - the metan software dedicated to a pc computer .
we presented the results of the analysis of 2-hour recording received from a pfn meteor station receiving belarusian radio signals . by setting both detectors to medium and maximum sensitivities we compared the effectiveness of both solutions for different detection threshold levels .
the results proved a significant difference between the detectors .
the fpaa detector proved its ability to pick up noticeably more audio signals for both thresholds . when set to the maximum sensitivity , it detected even hardly audible scatter events while maintaining its low false detection rate ( only 1 false positive per 57 detections ) . in comparison ,
current software detector found significantly fewer events and had lower true detection percentage .
our experiment confirmed that the measurements employing fpaa detector are of higher efficiency , thus the outcomes have a better signal to noise ratio .
currently , we are developing a ready - to - go electronic board with peripheries incorporating the presented fpaa - based detector , recording on an sd - card and an optional screen presenting counts histograms .
such a device could be easily handled not only by professional meteor researchers , but also by amateur astronomers who wish to contribute to additional radio observations . with its efficiency and very little power consumption ( comparing to a pc computer ) ,
our detector might replace current software - based detecting methods .
we also wish to set a network of remote detectors which will provide complex data and enable further investigation of its radiolocation capabilities .
we are grateful to the polish fireball network team ( pfn ) , who shared with us their specialist radio equipment used during the signal acquisition .
we would like to thank anonymous referee for all helpful and constructive comments and suggestions .
williams t. r. , saladyga m. , advancing variable star astronomy : the centennial history of the american association of variable star observers , cambridge university press , 2011 mocha j. , woznica t. , immunity of field programmed analog arrays fpaa to electromagnetic disturbances in the case of fpaa application to medical equipment , electrical review , vol .
88 , no 2 , 2012 zloczewski k. , szewc k. , first results of radio observations by cmw : the 2003 perseids , wgn , journal of the international meteor organization , vol .
21 - 22 , 2004 fietkiewicz k. , wisniewski m. , zoladek p. , cieslikowski z. , radio observations of draconids 2011 outburst , european planetary science congress 2012 , madrid , spain , sep .
23 - 28 , 2012 , pp .
725 wisniewski m. et al . , polish fireball network , european planetary science congress 2012 , madrid , spain , sep .
23 - 28 , 2012 , pp .
497 trigo - rodriguez j. m. , llorca j. , rietmeijer f. j. m. , janches , d. , advances in meteoroid and meteor science , springer science+business media b.v .
, new york , 2008 g. r. sugar , radio propagation by reflection from meteor trails , proceedings of the ieee , vol .
116 - 136 , 1964 j.m .
wislez , forward scattering of radio waves off meteor trails , proceedings of the international meteor conference , brandenburg , germany , sep .
14 - 17 , 1995 , pp
. 99 - 117 mckinley d.w.r
. meteor science and engineering , mcgraw - hill co. , 1961 j.a .
weitzen , w.t .
ralston , meteor scatter : an overview , ieee transactions on antennas and propagation , vol .
36 , no 12 , pp .
1813 - 1819 , 1988 pfn polish fireball network online resource www.pkim.org/files/2012_01_pfn28www.png accessed 6 march 2013 j. kero et al .
, a meteor head echo analysis algorithm for the lower vhf band , annales geophysicae , vol .
30 , no 4 , pp .
639 - 659 , 2012 a. malhotra , j.d .
mathews , a statistical study of meteoroid fragmentation and differential ablation using the resolute bay incoherent scatter radar , journal of geophysical research - space physics , vol .
a4 , 2011 resource document 1 .
an231e04 datasheet rev 1.1 .
www.anadigm.com/_doc/ds231000-u001.pdf accessed 6 march 2013 resource document 2 .
anadigmdesigner2 eda software .
www.anadigm.com/anadigmdesigner2.asp accessed 6 march 2013 p. falkowski , a. malcher , dynamically programmable analog arrays in acoustic frequency range signal processing , metrology and measurement systems , vol .
18 , no 1 , 2011 , pp .
77 - 89 falkowski p. , malcher a. , audio signal processing based on dynamically programmable analog arrays .
proceedings of international conference on signals and electronic systems icses 2010 , gliwice , poland , sep . 7 - 10 , 2010 , pp .
29 - 32 malcher a. , pietraszek s. , przybyla a. , hybrid qrs detection circuit based on dynamic reconfigurable field programmable analog array , proceedings of ifac workshop on programmable devices and embedded systems pdes 2010 , pszczyna , poland , oct .
6 - 7 , 2010 , pp .
69 - 74 malcher a. , pietraszek s. , przybyla t. , kidon z. , real time qrs detector based on field programmable analog array , journal of medical informatics & technologies , vol . 13 , 2009 , pp .
|
in the article we present a hardware meteor detector . the detection principle is based on the electromagnetic wave reflection from the ionized meteor trail in the atmosphere .
the detector uses the anadigm field programmable analogue array ( fpaa ) , which is an attractive alternative for a typically used detecting equipment - a pc computer with dedicated software .
we implement an analog signal path using most of available fpaa resources to obtain precise audio signal detection .
our new detector was verified in collaboration with the polish fireball network - the organization which monitors meteor activity in poland . when compared with currently used signal processing pc software employing real radio meteor scatter signals ,
our low - cost detector proved to be more precise and reliable . due to its cost and efficiency superiority over the current solution
, the presented module is going to be implemented in the planned distributed detectors system .
|
as a central problem of non - equilibrium thermodynamics , heat conduction has been of vital interest since the beginning of the 19th century . during his extensive investigations on heat transport in solids
fourier discovered around 1807 @xcite a proportionality between the temperature gradient @xmath0 and the heat respectively energy current density in the material , @xmath1 where @xmath2 is some material constant the heat conductivity .
concentrating on one - dimensional chains of interacting subsystems the temperature gradient can be discretized as @xmath3 , where @xmath4 is the temperature difference between two adjacent subsystems and @xmath5 their distance . introducing the current @xmath6 through some area @xmath7 , fourier s law ( [ eq:1 ] )
can be rewritten as @xmath8 with @xmath9 .
instead of approaching a global equilibrium state with some temperature @xmath10 , a solid , under some external perturbation by a temperature gradient ( introduced by heat baths with different temperatures ) , enters a local equilibrium state only : the temperature may vary from one macroscopically small but microscopically large part of the system to the other . the heat transport itself
is called _ normal _ if the conductivity @xmath11 remains finite , whereas a diverging @xmath11 indicates a _ ballistic _ transport .
having established a normal diffusive behavior , all further efforts on the topic of heat conduction in the last decades have been directed towards a microscopic foundation of the above phenomenological law .
one of the first concepts addressing the occurrence of regular heat conduction from a microscopic basis is the peierls - boltzmann theory @xcite .
essentially , peierls proposed a modified boltzmann equation replacing classical particles by quantized quasi particles , the phonons .
he was able to show that in order to obtain normal heat conduction an anharmonicity in the interaction potential was necessary , associated with the well known umklapp processes of phonon scattering in solids .
since quantized normal modes are treated here as classical particles , i.e. , as being always well localized in configuration as well as momentum space , this approach contains some conceptual shortcomings , though .
another microscopic approach to heat conduction is the famous green - kubo formula , derived on the basis of linear response theory @xcite .
thereby , the current is taken as the response of the material to an external perturbative potential , originally , the electromagnetic potential . by a direct mapping of these ideas on thermal transport phenomena ( perturbations due to thermal gradients @xcite )
serious questions remain open , since in this case it is not possible to define a proper potential in the hamiltonian of the system resulting in a temperature gradient . to account for the problem of defining a proper potential for a thermal perturbation
, there have been efforts to model the system coupled to several heat baths in terms of an open quantum system @xcite .
these methods reveal a normal transport behavior even in very small quantum systems , but are numerically challenging , especially for larger systems , since the investigations are done in the liouville space instead of the hilbert space .
all of the above microscopic approaches to heat conduction assume that the system under consideration is perturbed by some external environment ( e.g. a scenario , where a solid is heated at one side cooled at the other ) .
as already predicted by einstein in his considerations on brownian motion @xcite , the diffusion constant appearing in a local equilibrium steady state scenario of heat conduction should equal the decay constant for the relaxation from a non - equilibrium state to the global equilibrium in the same system . in the latter case
there is no further external perturbation involved . in the same spirit we will introduce yet another approach to heat transport within quantum systems .
recently , the relaxation behavior of small quantum systems has been considered within the theory of quantum thermodynamics @xcite , calling for a direct solution of the schrdinger equation .
based on the hilbert space average method ( ham ) @xcite , the complicated schrdinger dynamics of a multi level system may be reduced to a set of simple rate equations . in the present paper
we will discuss the application of ham on heat conducting model systems .
furthermore , we will discuss the connection of heat diffusive behavior with energy diffusion and present a comparison of the results of ham with the full numerical solution of the schrdinger equation .
the goal of the present investigation is to show how heat respectively energy diffusive behavior may emerge from first principles here the schrdinger equation @xcite .
the class of systems we are going to analyze first is depicted in fig .
[ fig:1 ] , and described by the hamiltonian @xmath12 coupled subunits with ground level and band of @xmath13 equally distributed levels ( @xmath14 ) .
black dots indicate the initial state used.,width=264 ] here @xmath15 identical subunits according to the same local hamiltonian @xmath16 are assumed to have a non - degenerate ground state and a band of @xmath13 exited states each , equally distributed over some band width @xmath17 such that the band width is small compared to the local energy gap @xmath18 ( see fig .
[ fig:1 ] , @xmath19 , @xmath17 in units of @xmath18 ) .
these subunits are coupled by an energy exchanging next neighbor interaction @xmath20 , chosen to be a ( normalized ) random hermitian matrix allowing for any possible transition such as to avoid any bias .
our results will turn out to be independent of the exact form of this matrix .
we choose the next neighbor coupling to be weak compared to the local gap ( @xmath21 , @xmath22 in units of @xmath18 ) .
this way the full energy is approximately given by the sum of the local energies and those are approximately given by @xmath23 , where @xmath24 is the probability to find the @xmath25-th subsystem in its excited state .
one of the big advantages of the present model is the clear partition of the hamiltonian in a local and an interaction part , respectively .
this partition allows for a unique definition of both a local energy or even a local temperature as well as a proper current .
however , note that the presented method itself does not need such a special kind of model system but is in principle also applicable to more general situations .
we now restrict ourselves to the single excitation subspace , i.e. , all states with one subsystem excited ( irrespective of which state within the band ) , all others in the ground state .
let us introduce a set of projection operators @xmath26 in this subspace , projecting out the set of states corresponding to the @xmath25-th subsystem occupying its excitation band , rather than its ground state ( these operators feature the unit operator of dimension @xmath13 for the excited band of the @xmath25-th system else zeros ) .
thus , the total system state is given by @xmath27 where none of the @xmath28 is normalized individually .
furthermore , one may introduce also some standard transition operators @xmath29 ( unit operator at position @xmath30 ) . according to the properties of transition operators we find @xmath31 and hence @xmath32 both , the local and the interaction part of the hamiltonian ( [ eq:3 ] ) in our reduced hamiltonian model
can now be written in terms of these new operators as @xmath33 with a local diagonal hamiltonian matrix @xmath34 and a hermitian interaction matrix @xmath35 .
the dynamics of the model system defined by the hamiltonian ( [ eq:3 ] ) respectively ( [ eq:6 ] ) and ( [ eq:7 ] ) is governed by the schrdinger equation . here
we will mainly be interested in a coarse - grained description suppressing the information about the state index within the respective bands , concentrating , e.g. , on the total probability of occupying a band or not , only .
after transforming the hamiltonian to the interaction picture we perform a second order dyson expansion of the time evolution for the short time step @xmath36 @xmath37 with the two time evolution operators @xmath38 a pertinent quantity in the context of heat conduction will be the change of local energy in one subunit @xmath25 in the time interval @xmath36 given by the probability @xmath39 using the short time evolution operator ( [ eq:8 ] ) , one gets @xmath40 for simplicity we suppress in the following the explicit notation of the time dependence of the operators , states and probabilities . according to the definition of @xmath41 we find in second order @xmath42 (
note that @xmath43 is hermitian due to the hermiticity of the interaction ) . due to the additive representation of the interaction ( [ eq:7 ] ) and the definition of the first oder time evolution operator ( [ eq:9 ] ) one may find the analogous off - diagonal block form for the first oder time - evolution operator @xmath44 the time dependent coefficients @xmath45 contain the integration of the interaction matrix @xmath46 over @xmath47 ( cf .
( [ eq:7 ] ) ) .
unfortunately , the second oder time evolution operator ( [ eq:10 ] ) has a more complicated structure in terms of the transition operators .
introducing the interaction ( [ eq:7 ] ) into ( [ eq:10 ] ) ( using the property ( [ eq:5 ] ) ) one finds @xmath48 this operator decomposes into a diagonal and an off - diagonal part @xmath49 with @xmath50 and a suitable definition of the coefficients @xmath51 . by using ( [ eq:5 ] )
one may convince oneself of the operator identities @xmath52 by means of the expansion of the time - evolution operators in terms of projectors and transition operators and those operator identities one is able to reformulate ( [ eq:13 ] ) by exchanging time evolution operators and projectors as @xmath53 since @xmath54 is a probability we get , furthermore , the normalization condition @xmath55 .
this condition is already fulfilled in the zeroth order of all probabilities @xmath56 . due to this fact the sum over all other orders has to vanish .
all first order terms together cancel out each other directly .
remaining second oder terms lead to the additional conditions @xmath57
according to ( [ eq:11 ] ) the expectation value of the operator ( [ eq:20 ] ) defines the time dependence of the probability @xmath24 .
we approximate this expectation value by an average over some appropriate hilbert space compartment say the mean of the expectation value over all reachable states called the hilbert space average @xmath58 . the hilbert space average method ( ham )
is , in essence , a technique to guess the value of some quantity defined as a function of the full system state @xmath59 if @xmath59 itself is not known in full detail .
here it is used to produce a guess for the expectation values @xmath60 if the only information about @xmath59 is the given set of initial expectation values @xmath61 .
such a statement naturally has to be a guess , since there are , in general , many different @xmath59 that are in accord with the given set of @xmath62 but possibly produce different values for @xmath60 .
thus , the key question for the reliability of this guess is whether the distribution of the @xmath60 produced by the respective set of @xmath59 s is broad or whether almost all those @xmath60 are approximately equal .
it can be shown that if the spectral width of @xmath63 is not too large and @xmath63 is high - dimensional almost any @xmath59 yields an expectation value close to the mean of the distribution of the @xmath60 and thus the ham guess will be reliable @xcite . to find that mean one has to average with respect to the @xmath59 s .
this is called a hilbert space average and denoted as @xmath64 such averages are well known and derived in all detail in @xcite .
for any hermitian operator @xmath65 one finds @xmath66 with the partial traces in the respective subspace @xmath25 with dimension @xmath67 ( note @xmath68 ) . plugging in the operator @xmath63 from ( [ eq:20 ] ) into the probability ( [ eq:12 ] ) , approximating the expectation values by the hilbert space average and furthermore using the operator identities ( [ eq:21 ] ) , one finds for the probability @xmath69 furthermore , using ( [ eq:23 ] ) with the dimension @xmath70 in each subspace
, we end up with the probability @xmath71 a first order approximation of these traces is straightforward and done like in the derivation of fermi s golden rule ( cf .
@xcite ) getting here @xmath72 defining the decay constant @xmath73 and remembering that @xmath74 one eventually finds @xmath75 the above approximation for the trace of the square of the time - evolution operator ( [ eq:26 ] ) is only valid within a small time interval like for fermi s golden rule implying the two conditions ( see @xcite ) @xmath76 as long as these conditions are fulfilled the given second order time evolution approximation is valid and one may iterate ( [ eq:28 ] ) in the limit of @xmath36 being extremely small to find the _ rate equation _ for the system . since the system is a finite chain of @xmath15 subsystems we get a slightly different equation for the first and the last system in the chain .
the complete system of differential equations , thus , reads @xmath77 @xmath78 under the conditions ( [ eq:29 ] ) the system should follow this rate equation thus establishing a statistical behavior from pure schrdingerian dynamics ! from ( [ eq:31 ] ) one finds that the change of the probability in the @xmath25th subsystem only depends on its left and right neighbor . therefore ,
if the above criteria ( [ eq:29 ] ) are fulfilled , so that the system indeed follows the respective rate equation , we may concentrate on the minimal model of two subunits only , without loss of generality . because of the special next neighbor relationship of the above dynamical equation the results of the minimal model can be generalized to the behavior of a chain of a large number of such subsystems .
that such a behavior as described in the last sections is , indeed , feasible can be demonstrated by a comparison of the solution of the above rate equation with the exact solution of the full system schrdinger equation .
for this reason we consider a model consisting of three subunits with @xmath79 levels each in the exited band of width @xmath80 coupled by a random energy exchanging next neighbor coupling @xmath81 .
for these parameters the two criteria ( [ eq:29 ] ) are fulfilled .
initially we start with an excitation in the first system , all other systems being in their ground state , observing the redistribution of energy . the comparison for @xmath82 in fig .
[ fig:2 ] shows a very good accordance between the solutions of these two different dynamical descriptions both the solution of the full schrdinger equation as well as the rate equation ( [ eq:30])-([eq:32 ] ) .
comparison of the ham prediction ( lines ) and the exact schrdinger solution ( dots ) .
( @xmath82 , @xmath79 , @xmath81 , @xmath80),width=302 ] furthermore , fig .
[ fig:2a ] shows the same comparison for a chain of increased length @xmath83 .
again , one finds a very good accordance .
comparison of the ham prediction ( lines ) and the exact schrdinger solution ( dots ) .
( @xmath83 , @xmath79 , @xmath81 , @xmath80),width=302 ] least squares ) for 15000 equally distributed initial states ( @xmath84 , @xmath79).,width=302 ] to investigate the accuracy of the ham for , e.g. , the probability in the first system @xmath85 we introduce @xmath86 , as the time - averaged quadratic deviation of the exact ( schrdinger ) result for @xmath85 from the ham prediction @xmath87 to check the validity of ham with respect to different pure initial states we have computed @xmath86 for 15000 equally distributed initial states , all corresponding to a two subunit system with one subunit in the upper band and the other subunit in its ground state .
the results are condensed into the histogram shown in fig .
[ fig:3 ] . obviously , almost all initial states lead to approximately the same small deviation .
thus a good agreement of ham with the exact result may safely be expected regardless of ham being a best guess only .
( in fact this is the essence of thermodynamical statements : they hold typically , but not always . )
furthermore , to analyze how big this `` typical deviation '' is , we have computed @xmath86 for a @xmath82 system with the first subunit initially in an arbitrary excited state , for different numbers of states @xmath13 in the bands .
least squares ) : dependence of @xmath86 on @xmath13 for @xmath82 .
inset : dependence of @xmath86 on @xmath15 for @xmath79.,width=302 ] as shown in fig .
[ fig:3a ] , the deviation scales like @xmath88 with the band size , i.e. , vanishes in the limit of high dimensional subunits .
this behavior does not come as a complete surprise since it has theoretically been conjectured and numerically verified in the context of equilibrium fluctuations of non - markovian systems @xcite .
the inset of fig .
[ fig:3a ] shows that @xmath89 goes down also with increasing number of subunits @xmath15 but then levels off .
altogether the new method ham appears to be applicable even down to moderately sized subsystems , i.e. , not too few levels @xmath13 in the upper band .
thus , as follows from fig .
[ fig:3 ] , a system with only a single excited level would lead to large deviations from the rate equation behavior . using instead of pure initial states mixed states ( which are typical in the context of thermodynamical phenomena )
a drastic further reduction of @xmath90 can be expected .
this is due to the fact that pure state fluctuations may cancel partially if added together .
we concentrate in the following on a model ( cf . fig . [ fig:1 ] ) for which the criteria ( [ eq:29 ] ) are fulfilled .
in essence , the used subunits should be weakly coupled to fulfill the given criteria , i.e. , the interaction between the subunits has to be small compared to the local energy .
the dynamical behavior is then approximately governed by the rate equation ( [ eq:30])-([eq:32 ] ) . as argued in the last paragraph of sect .
[ sec:4 ] , it is enough to consider a minimal model of @xmath84 subunits , without loss of generality .
this minimal system behaves according to the rate equation ( [ eq:30 ] ) with the decay constant ( [ eq:27 ] ) . the respective equation for the second system is simply obtained by an exchange of the indices in ( [ eq:30 ] ) .
in order to compute the energy diffusion constant in the given model system , we have to consider the energy current inside the system .
the energy current is defined by the change of the internal energy @xmath91 in the two subunits @xmath92 the total internal energy in such a subunit is given by the probability to be in the excited band of the subunit times the width of the energy gap , @xmath93 .
thus , we may reformulate the current due to the change of the probability being in the excited band as @xmath94 the change of the probabilities in time are given by the rate equation ( [ eq:30 ] ) so that @xmath95 for a larger system with @xmath96 the consideration is quite similar and the result in principle the same .
of course , the current definition is a little bit more complicated , since the energy relaxation in a central system leads to a current to both sides .
but by properly splitting up the result into a current to the left respectively right system leads to the same dependence of the current between subunit @xmath25 and @xmath97 on the respective probability difference .
according to ( [ eq:36 ] ) the current is a linear function of the probability difference ( gradient ) respectively _ energy difference ( gradient ) _ inside the system .
this refers to a statistical behavior since energy diffuses through the system only due to the energy distribution in the system .
therefore we may extract the energy diffusion constant from the above equation , identifying the parameter @xmath98 with the transport coefficient given in ( [ eq:27 ] ) .
this energy diffusion constant only depends on the coupling strength of the subunits @xmath22 and on the state density of the excited band @xmath99 .
( note that in the model at hand the energy gradient is discretized as an energy difference , because of the discrete structure of the model , as already stated in sect .
[ sec:1 ] [ cf .
( [ eq:2 ] ) ] . )
up to now we have restricted ourselves to the energy transport within the system on the route from a non - equilibrium initial state to a global equilibrium final state . of course
this could be translated into heat transport insofar as heat is nothing but thermal energy .
however , the states far away from equilibrium preclude a straight - forward interpretation : in the example of fig .
[ fig:2 ] we have prepared the system initially in a state with the first system completely in its exited band , all others being in their ground states and , indeed , the system behaves statistically . by a simple translation to temperatures by fitting to a boltzmann distribution
this initial state would show a really strange temperature profile : one finds the first system at negative temperature ( completely excited , i.e. , inversion and therefore negative , infinite temperature ) , all others at zero temperature .
nobody would expect a normal heat conduction behavior in such a far from equilibrium situation .
nevertheless , even in this case we have confirmed a statistical relaxation of energy ( see fig .
[ fig:2 ] ) . to investigate a situation more appropriate for heat transport we should use an initial state already near the final global equilibrium state , with only a small temperature gradient .
we first start with a pure state observing the decay of such a state by solving the schrdinger equation only . finally , one may sum up over many pure state trajectories finding the time evolution for an initially mixed state .
but with a temperature diffusive behavior already for pure states we are sure to find it also for an initially mixed state , even with reduced fluctuations .
let us restrict ourselves in the following to only two subunits the proper minimal model as argued before .
however , let us immediately generalize the model to a more complicated structure with more than one excitation band per subunit .
the results of ham from the above considerations is valid also for this more general situation .
the model system is shown in fig .
[ fig:4 ] . equally distributed levels ( @xmath84 subunits).,width=302 ] the band index @xmath100 is reserved for the left system whereas @xmath101 belongs to the right system .
we take a product state @xmath102 as the respective initial state for all further investigations . for the single subsystem states we choose a `` quasi '' thermal state with temperature @xmath103 , i.e. , a superposition of states according to the boltzmann distribution . here , we use the term `` quasi '' to label this special group of pure states , even if it is _ not _ a mixed state as usual for thermal states .
nevertheless , the occupation probability of the energy eigen states refers to the boltzmann distribution !
thus , a mixture of many of such `` quasi '' thermal pure states may lead to the usual thermal state .
furthermore , we treat the bands as `` quasi '' degenerate since the band width is much smaller than the gap @xmath19 .
thus , the levels inside each band are equally occupied according to the temperature @xmath103 . the probability distribution of the first subsystem depending on the band energy @xmath104 and the appropriate number of states @xmath105 in the band is therefore given by @xmath106 with the partition function @xmath107 .
the second subsystem distribution is obtained simply by exchanging @xmath100 by @xmath101 and the respective quantities of the second subsystem .
the important quantity for all further considerations is the probability distribution of the whole system .
since the total state is a product state the probability distribution of the complete system state @xmath108 is simply the product of the single probabilities @xmath109 however , it is convenient here to use a new set of coordinates for the combined spectrum in terms of the mean energy @xmath110 of a band and some energy displacement @xmath111 @xmath112 , @xmath113 axes respectively ; transformation to the new parameters @xmath114 ; black bands indicate the complete spectrum .
examples : @xmath115 are states , where the first system is in its first band , the second one in its ground state , @xmath116 are states , where both systems are in the first excited band etc.,width=302 ] ( all energies , also @xmath110 and @xmath111 , are measured in units of @xmath18 . ) to support intuition we show both sets of coordinates in fig . [ fig:5 ] : the full spectrum ( black bands ) can either be characterized in terms of the single subsystem spectra @xmath117 or by @xmath114 .
the latter coordinates appear as a @xmath118-rotation of the original ones .
introducing a small temperature difference @xmath4 around the mean temperature @xmath10 so that @xmath119 and transforming to the new parameters ( [ eq:39 ] ) one finds the probability distribution @xmath120 with the number of states @xmath121 in the band @xmath114 .
furthermore , in order to get a near equilibrium situation the temperature gradient @xmath4 has to be very small in comparison to the mean temperature @xmath122 .
therefore we expand the exponential function in ( [ eq:40 ] ) in terms of @xmath4 finding in first order @xmath123 where also the product of partition functions has been approximated in first order of the temperature gradient @xmath4 by @xmath124 .
again , the important quantity is the energy current between the two subsystems . the total energy in the first subunit is now given by @xmath125 , where @xmath126 is the time dependent probability distribution in the first system .
to get the respective first system probability distribution @xmath126 from the time dependent full system probability distribution @xmath108 we have to trace over the second subsystem by summing over @xmath101 , finding @xmath127 . according to this total energy in subsystem 1
the current along the chain reads @xmath128 or , switching to the new parameters ( cf .
( [ eq:39 ] ) ) @xmath129 for a single band model the sum over @xmath110 would break down and we recover the result ( [ eq:35 ] ) .
the change of the probability @xmath130 is again described by the rate equation according to the ham .
since the complete model decomposes into several energy subspaces , the derivation is essentially just as given in sect .
[ sec:4 ] .
the respective energy transport coefficient results as a more complicated form depending on the band it belongs to ( @xmath131 ) . however , we require that @xmath98 is symmetric under the exchange of @xmath111 and @xmath132 , since the energy transport from system 1 to system 2 should be as good as in the reverse direction
. therefore we end up with the ham rate equation @xmath133 the fact that @xmath98 depends on the respective band is already obvious from ( [ eq:27 ] ) , which depends on the state density of the band . plugging in the rate equation ( [ eq:44 ] ) into the current ( [ eq:43 ] )
we find @xmath134 again , the current is simply a function of the actual state . for a short time
step we may use the first order approximation of the `` thermal '' initial state ( [ eq:41 ] ) . since we are summing over @xmath111 and @xmath132 the zeroth order terms according to ( [ eq:41 ] )
will cancel .
the remaining first order part @xmath135 gives a heat current proportional to the temperature difference @xmath4 . by comparing this current with fourier s law ( [ eq:2 ] )
, one can extract the heat conductivity coefficient as @xmath136 a quantity depending only on parameters of the respective model system and on temperature .
this is a very general formula for the heat conductivity derived from ham , valid for the class of model systems depicted in fig .
[ fig:4 ] with an arbitrary number of bands .
let us return in the following to the single band model as shown in fig .
[ fig:1 ] , with the partition function of a single @xmath137-level subsystem @xmath138 . in this case ham leads to the unique energy diffusion constant @xmath139 ( see ( [ eq:27 ] ) ) . furthermore
, in this single band model only the energy band @xmath140 contributes to the conductivity @xmath11 , since for all other bands all possible @xmath111 parameters are zero ( cf .
[ fig:5 ] ) .
there are two possible values for the band @xmath141 , namely @xmath142 and the number of levels in these two sub - bands are equal @xmath143 , due to the construction of the model .
one thus gets for the conductivity ( [ eq:47 ] ) @xmath144 the dependency of the transport coefficient @xmath11 on temperature is displayed in fig .
[ fig:6 ] . ) over temperature for a system with @xmath79 , @xmath145 and @xmath146.,width=302 ] it roughly looks as expected for typical isolating solids .
there is , however , no @xmath147 behavior for small temperatures , but note that this is only to be expected for three dimensional crystal lattices with a linear phonon dispersion relation for small momenta . the exponential behavior for small temperatures @xmath148 results from the gap in the system and is difficult to treat for small temperatures since it is an essential singularity ( typical for systems with an energy gap ) .
furthermore we find a @xmath149 behavior rather than the usual @xmath150 behavior in the high temperature limit .
again , the usual @xmath150 is no rigorous general result , peierls only states that the exponent should be between minus one and minus two . by introducing the heat capacity
, the above result for the heat conduction coefficient can further be simplified . starting with the canonical partition function of a single subunit @xmath151 and the definition of the heat capacity in terms of the partition function @xmath152 one finds for the heat capacity of one subunit @xmath153 from a comparison with ( [ eq:48 ] )
one derives @xmath154 i.e. , the heat conductivity is given by the energy diffusion constant @xmath98 multiplied by the heat capacity @xmath155 of a single subunit .
rather than considering heat conduction for a system coupled to two heat baths with different temperature in a stationary local equilibrium state , we have studied the time evolution of a closed one - dimensional chain of weakly coupled many level systems . by an application of the new hilbert space average method ( ham )
, we have been able to directly address a coarse - grained description . on this level of description
the effective dynamics can be shown to be diffusive allowing us to derive a theoretical expression for the _ energy diffusion coefficient _
@xmath98 from first principles ( schrdinger equation ) .
furthermore , two conditions have been established which could help to classify whether a given model should show a statistical diffusive behavior or not . the approximation involved ( mainly the replacement of exact expectation values by a hilbert space average ) has been tested by a numerical comparison of the derived rate equation and the exact schrdinger equation .
surprisingly , we have found for the present model system that energy diffusion occurs right from the start and already far from equilibrium . for an initial quasi thermal state featuring a small but finite temperature gradient
we have been able to find a normal heat conducting behavior in terms of a finite _ heat conductivity _
@xmath11 . even for a more complicated model system ( several bands per subunit ) it was possible to derive a theoretical expression for the transport coefficient in terms of the parameters of the hamiltonian
finally , a reformulation of the transport coefficient has led to a connection with the energy transport coefficient and the heat capacity .
what impact do those results for our `` design model '' have on real physical systems ?
for an application of ham one has to organize the system as a `` grain - structure '' of weakly coupled subunits .
then , for the emergence of normal transport , the local bands and their couplings that mediate the transport should obey the criterion ( [ eq:29 ] ) individually .
if this can be established ham predicts that energy will diffuse from subunit to subunit .
it has long since been appreciated that coarse - graining ( i.e. , tracing out irrelevant degrees of freedom ) may change the effective dynamical laws even qualitatively . an obvious example is encountered when some accelerated electron in a solid transfers energy and momentum to vibrational degrees of freedom thus giving rise to particle transport . for the origin of heat transport in insulators
a correspondingly simple picture seems to be missing though .
here we have introduced a specific model class for which such a concept of information reduction can easily be incorporated .
we thus have been able to clearly specify the conditions under which the resulting effective dynamics can be characterized as normal energy diffusion .
irreversibility and thermal resistance are thus traced back to a `` reduced description '' : `` in reality '' the full dynamics is still controlled by the schrdinger equation , as has explicitly been checked here for still small enough total quantum systems . the desired structure may be achieved by coarse - graining periodic systems like spin chains , crystals , etc . into entities containing many elementary cells and taking the mesoscopic entities as the subunits proper and the effective interactions between the entities as the couplings .
increasing the `` grain size '' will result in higher state densities within the subunits and relatively weaker couplings such that above a certain minimum grain size the criteria ( [ eq:29 ] ) will eventually be fulfilled . in this way
a qualitatively new type of dynamical behavior can be seen to evolve as the observational level becomes more and more coarse - grained .
of course , the resulting subunits can not generally be expected to feature the same gapped spectral structure as our design model .
but ham should be applicable also to more realistic systems , and work in that direction is under way .
we thank h .-
breuer , m. hartmann , m. henrich , ch .
kostoglou , h. michel , h. schmidt , m. stollsteimer and f. tonner for fruitful discussions .
financial support by the deutsche forschungsgesellschaft is gratefully acknowledged .
|
we analyze closed one - dimensional chains of weakly coupled many level systems , by means of the so - called hilbert space average method ( ham ) .
subject to some concrete conditions on the hamiltonian of the system , our theory predicts energy diffusion with respect to a coarse - grained description for almost all initial states .
close to the respective equilibrium we investigate this behavior in terms of heat transport and derive the heat conduction coefficient .
thus , we are able to show that both heat ( energy ) diffusive behavior as well as fourier s law follows from and is compatible with a reversible schrdinger dynamics on the complete level of description .
|
over the last decades , amorphous silicates containing several wt.@xmath0 of water have attracted the interest of many experimental and theoretical research groups due to its importance in science and technology . e.g. these systems are supposed to play an important role in magmatic flow in the earth crust and have therefore been investigated extensively by many geologists ( for reviews see @xcite ) . in silicate melts ( lava ) , water can be released if temperature and pressure conditions are altered and in that case , the released water molecules assemble in bubbles .
these bubbles can set up pressure for explosive volcanism @xcite .
the microscopic origin of such bubbles is still very poorly understood @xcite and one of the main motivating aspects for our study is to find indicators for such bubble formation . the possible use of water containing silica in fuel cells @xcite and the use of water as defect passivant in semiconductor devices ( and reladed problems ) @xcite and its disturbing presence in optic fibers @xcite underline the technological importance .
+ over the past years pure amorphous silica has been well understood regarding its electronic and structural properties and experimental and simulational data seem to agree well @xcite . in recent studies these investigations
have been extended to the technologically highly important class of sodium silicates @xcite .
new fundamental ideas of the structural properties and the diffusion mechanism of sodium atoms in these melts have been obtained .
+ the impact of water on the structure and dynamics of silica is supposed to be at least as important as the addition of sodium @xcite .
various methods have been used to investigate water speciation and diffusion in a sio@xmath1 network .
nmr , infrared and raman spectroscopy as well as neutron scattering experiments suggest a chemical reaction of molecular water to sioh groups in the melt for low initial water concentrations and a saturation of this reaction if more water is added to the liquid @xcite . in chemical terms it represents a balance of the form @xmath2 that follows the chatelier principle @xcite and which can be shifted to any side by the variation of external conditions like temperature or concentration of one species . the energetics is supposed to be such that a reaction to the right is endothermal and hence at high temperatures sioh is the dominant species @xcite . on the other hand ,
the mentioned experiments have difficulties to provide reliable quantitative data .
the different fabrication processes of the samples and the overlapping of hydroxyl and water signals in the mentioned spectra are the main reasons for this @xcite .
in - situ measurements in the geologically relevant liquid state are hardly accessible due to elevated temperatures and pressures .
+ computer simulations do usually not encounter these problems since , in principle , they can be done at high temperature and pressure .
nevertheless problems for computer simulations that use inter atomic potentials arise directly from the above reaction : potentials that describe appropriately all possible types of oxygen ( bridging oxygens in the tetrahedral network , oxygens in hydroxyl groups , molecular water oxygens ) have been developed with some success but do not describe the dynamics very well @xcite . here
the great advantage of _ ab initio _
simulations comes into play as the potential and the speciation are found and updated from accurate electronic structure calculations at each molecular dynamics time step .
especially _ ab initio _ car - parrinello simulations turned out to be a powerful tool for investigations of disordered silica and sodium silicate systems @xcite . for a review of previous _ ab initio _ simulated amorphous systems
see @xcite .
+ in this paper we present an equilibrated liquid of water - containing sio@xmath1 and its structural and dynamical properties . these properties will be compared to that of a silica and a sodium silicate melt when possible .
the paper is organized as follows : in section [ sec : simdet ] we discuss the used simulation techniques and in section [ sec : res ] we present the structure of the liquid and draw some conclusion on diffusion mechanisms for hydrogen .
conclusions are summarized in the last section .
the molecular dynamics simulations were performed with the car - parrinello code cpmd @xcite at 3000 k and 3500 k. the electronic subsystem was treated with a density functional ( dft ) approach in a generalized gradient approximation using the pbe functional @xcite .
the core electrons of silicon and oxygen atoms were treated in a pseudo potential model using troullier - martins type pseudo potentials @xcite . a plane wave
@xmath3 point expansion with an energy cutoff of 50 ry turned out to be sufficient for an appropriate description of the inter atomic forces .
the 50 ry cutoff with the pbe functional was first tested on the h@xmath1o dimer and on @xmath4-quartz .
these tests showed that e.g. the two sio distances of @xmath4-quartz are equal to 1.624 and to 1.628 independent of the energy cutoff between 50 ry and 100 ry . on the other hand
, we found for the h@xmath1o dimer that the o - o distance , the quantity which is the most sensitive to a change of the cutoff , shows only a variation from 2.925 to 2.950 if the cutoff is decreased from 90 ry to 50 ry .
+ for the sio@xmath1-h@xmath1o liquid we used a simulation box containing 102 atoms ( 30 silicon , 64 oxygen , 8 hydrogen ) corresponding to a concentration of water of 3.84 wt.% .
the simulation box was prepared by starting from an equilibrated liquid silica configuration ( obtained with the effective potential proposed by van beest _
et al _ @xcite ) with 32 sio@xmath1 units .
two randomly chosen silicon atoms were then substituted by four hydrogen atoms each so that the chemical stoichiometry was preserved .
these hydrogen atoms were attached to the oxygen atoms which were formerly bound to the removed silicon atoms .
using cpmd the liquid was subsequently equilibrated at 5000 k for some time and finally cooled down to 3500 k and 3000 k and annealed for roughly 5 ps before recording the trajectories for the present work . since , to our knowledge , densities of water - containing silica systems ( especially in the liquid state ) have not been measured , the volume of the box had to be chosen such that the calculated internal stress of the system had a mean value of zero .
this was the case for a box size of 11.50 and hence a density of 2.05 g/@xmath5 .
comparisons of structural quantities like pair distribution functions , mean square displacements or angle distributions extracted from molecular dynamics simulations performed at 3500 k with a cutoff of 50 ry and of 80 ry did not show differences within the statistical error bars . thus we concluded that the 50 ry cutoff was sufficiently high .
for equilibration , the masses of the ions were all set to 28 a.u .
( the mass of a silicon atom ) .
note that a change of the ionic masses does not affect the structure of the liquid since at equilibrium all structural quantities are independent of the mass .
on the other hand , the increase of the ionic masses ( from 1 to 28 for hydrogen and from 16 to 28 for oxygen ) should increase the energy gap in the electronic structure and allow an increase of the car - parrinello electronic mass and hence the use of a larger time step which thus leads to a faster equilibration .
the equilibration of the system was performed at the two ionic temperatures of 3000 k and 3500 k employing nose - hoover thermostats and an electronic mass of 600 a.u .
( energy @xmath6 time@xmath7 ) at a time step of 4.5 a.u .
( 0.1088 fs ) . at high temperature ,
the electronic gap is too small compared to @xmath8 to ensure the decoupling of the ionic and the electronic degrees of freedom , which is needed to perform car - parrinello dynamics .
the use of thermostats is therefore compulsory . to speed up the equilibration and to perform an efficient canonical sampling ,
one separate nose - hoover thermostat chain for each ionic degree of freedom was used ( known as `` massive '' thermostating @xcite ) .
the electrons were controlled with one single thermostat chain @xcite .
unfortunately , due to the use of thermostats the direct access to dynamical properties is no longer available .
+ mean square displacements of the si , o and h atoms at 3500 k
( solid lines ) and 3000 k ( dashed lines).,width=226 ] the system was equilibrated at two temperatures ( 3500 k and 3000 k ) until in a log - log plot the averaged mean square displacements ( msd ) of the si atoms showed at long times a slope close to unity .
this is illustrated in fig .
[ fig : msd ] where the mean square displacements ( msd ) of the different atomic species are shown for the 3000 k ( dashed lines ) and the 3500 k ( bold lines ) liquids .
usually msds of viscous liquids are composed of three regions : the ballistic one in which the atoms move without noticing their neighbors and hence a msd that is proportional to @xmath9 .
this ballistic region is followed by a region where the atoms are temporarily confined in a cage made of their nearest neighbors . in this regime ,
the atoms rattle around in the cage without significant displacement , leading to a msd that increases only slowly .
finally the atoms leave this cage and start to show a diffusion motion , i.e. a msd that is proportional to @xmath10 .
the choice of the masses and the thermostats affect also the msd . however , the height of the plateau and the displacement at the onset of the diffusional regime should be independent of the thermostat .
hence , we consider the system to be equilibrated once the diffusional regime is reached which was the case after 4.4 ps at 3500 k and 10.9 ps at 3000 k. in order to check that the liquids were indeed well equilibrated and that there were no aging effects , the trajectories were cut into three equal parts .
the averaged mean square displacements were then calculated for each part separately and compared to each other . since the three different averaged msd did not show any drift , aging effects can be excluded and equilibration was indeed obtained after the above mentioned times .
all structural quantities were extracted from the beginning of the diffusional part up to the total length of the trajectory ( 4.4 to 12.5 ps at 3500 k and 10.9 to 22.5 ps at 3000 k ) . from fig .
[ fig : msd ] we note that at the end of the 3500 k trajectory the msd is 29.2 @xmath7 , 8.1 @xmath7 , and 5.0 @xmath7 for the h , o and si atoms respectively . at 3000
k the msds were 16.4 @xmath7 , 3.5 @xmath7 and 1.73 @xmath7 . according to the above mentioned equilibration times and the time step of 4.5 a.u .
( 0.1088 fs ) , the numbers of computed time steps were 114900 at 3500 k and 206800 at 3000 k. using a hitachi sr8000 computer , one time step takes 56 s on a single processor .
in this section we present the structural analysis of the liquids at 3000 k and 3500 k. commonly considered quantities like pair distribution functions , bond angle distributions , coordination numbers , @xmath11-species distributions and bridging to non - bridging oxygen ratio will be discussed and , as far as possible , be compared to the data extracted from _ ab initio _ simulated silica and sodium silicate ( ns4 ) melts @xcite .
in particular the comparison to a recently investigated sodium tetrasilicate melt @xcite is highly interesting since the valence shell configuration of the sodium and the hydrogen atoms are equivalent . + in order to analyze the structure with particular attention to the formation and the rupture of the silica network , water molecules and hydroxyl groups , it turns out to be useful to distinguish several types of oxygen atoms : + [ cols= " < , < " , ] the average number of o - h bonds in the sio@xmath1-h@xmath1o melts is found to be equal to 8.4 at 3500 k and to 8.1 at 3000 k and is therefore larger than the total number of hydrogen atoms in the system . hence we conclude that intermediate states with two oxygen atoms close to a single hydrogen exist and that they must serve as intermediate states for hydrogen exchange reactions ( process number * 1 . * ) .
+ by counting the number of o - h ruptures along the trajectories and the type of reactants / decay products associated to these ruptures , we are able to show the existence of hydrogen diffusion processes of types * 1 . * and * 2.*. for the following we assumed that an o - h bond was formed if the o - h interatomic distance was larger than the o - h cutoff at step @xmath10 and smaller than the o - h cutoff at step @xmath12 ( @xmath13 being the length of a time step of 0.1088 fs ) .
the nature of the reactant is found by counting , at time @xmath10 , all silicon and hydrogen neighbors ( again in terms of the nearest neighbor cutoff ) of the later o - h oxygen .
a rupture of an o - h bond was assumed to have happened if one h and one o atom had an interatomic distance smaller than the o - h cutoff at time @xmath10 and if their interatomic distance became larger than the o - h cutoff at step @xmath12 , irrespective of whether subsequently the same bond was formed again or not .
the decay product of an o - h rupture is the h donator unit ( without the h atom ) and is found by counting , at step @xmath12 , all silicon and hydrogen neighbors ( again in terms of the nearest neighbor cutoff ) of the former o - h oxygen .
+ at 3500 k the total number of o - h formations and ruptures was 97 and 94 respectively , during the equilibrated part of the trajectory ( 8.1 ps ) . at 3000 k
we found 68 o - h formations and 69 ruptures during the equilibrated part of the trajectory ( 11.6 ps ) .
however , in order to take into account the o - h formations and ruptures that serve for the hydrogen diffusion , we also have to distinguish between the formations and ruptures that give rise to _ hydrogen transfers _ or _
recombinations_. in order to perform this separation between transfers and recombinations it is necessary to relate each decay of an oh bond to a formation of an oh bond .
a hydrogen transfer requires the rupture of one o - h bond and the formation of a _ different _ one , whereas a recombination implies only one bond .
note that the rupture of the first bond will occcur before the formation of the new bond if a transient free hydrogen is formed , or , as in almost every case , after the formation of a new bond which implies the formation of the intermediate state . making this distinction between transfers and recombinations
, we find 28 transfers at 3500 k ( corresponding to a ratio of @xmath14 transfers and 71.2 @xmath15 recombinations ) and 8 transfers at 3000 k ( corresponding to a ratio of @xmath16 transfers and @xmath17 recombinations ) .
the ratio between the number of ruptures involved in a transfer and the total number of ruptures gives the recombination rate .
[ fig : decohtrans]a and b give the percentage of the reactants and decay products for the o - h formations and ruptures including recombinations and figs .
[ fig : decohtrans]c and d the related transfer reactions ( i.e. without recombinations ) . for the transfers , due to the small numbers , the error bars are quite large . at both temperatures ,
the contributions of the si - o dangling bonds are higher for the transfer related ruptures than for the overall ruptures , i.e. transfers and recombinations .
furthermore we recognize from figs .
[ fig : decohtrans]c and d that the si - o dangling bonds and the bos are the main acceptors and rupture products , with almost equal probability .
concerning transfers , the si - o dangling bonds react with a probability of 40 % at 3500 k and 60 % at 3000 k into an sioh group and the sioh group ( o@xmath18 ) decays with a probability of 40 % at 3500 k and 50 % at 3000 k into a si - o dangling bond
. however the corresponding probability that includes recombinations is for both the formation and the decay around 30 % .
this result indicates that the si - o dangling bonds are more involved in the formation and rupture of the o - h bonds when the transfers are considered and the recombinations are taken out .
furthermore it also demonstrates the existence of large vibrations of the h atoms around the o atoms of the si - o dangling bonds and hence the existence of the `` weak '' hydrogen bonds described in sec .
[ sec : rdf ] .
+ the bos serve , with a probability of 50 % at 3500 k and of 40 % at 3000 k , as reactants for a hydrogen transfer leading to a new o - h group ( 60 % including recombinations at both temperatures ) . on the other side ,
the bos serve as products of a hydrogen transfer with a probability of about 50 % , and therefore the participation of the bos to the formation and rupture of the o - h bonds is slightly decreased when transfers are compared to the reactions including recombinations .
+ finally , the o@xmath18 contribute much less to the o - h group formation and decay than the other species , with a probability of about 10 % ( 12 % including recombinations ) at 3500 k and 0 % ( 8 % including recombinations ) at 3000 k. + from fig .
[ fig : decohtrans ] , we deduce that hydrogen transfers involving si - o dangling bonds or bos as reactants as well as decay products dominate with @xmath19 90 % . since these units are associated to o@xmath18 or o3h as initial or final states , they correspond to the hydrogen diffusion process number * 1 . * which means that among the processes * 1 . * and * 2 . * , the former dominates with @xmath19 90 % .
let us now consider the third considered hydrogen diffusion mechanism .
it is associated to the motion of a o - h unit as a whole ( without any o - h rupture ) which can be obtained via the formation and decay of an o3h cluster as sketched in fig .
[ fig : diffreact3 ] . typical hydrogen diffusion reaction of process three involving a o3h .
a o - h group is released from an si - o - h ( o@xmath18 ) to a o3h cluster decaying subsequently in another o@xmath18 .
an oversaturated and an unsaturated silicon atoms are subsequent products . , width=529 ] note that this process requires the presence of `` weak '' si - o bonds , the existence of which was inferred from fig .
[ fig : specosi ] and leads to the formation of threefold and fivefold coordinated silicon atoms . as for the o - h bonds
, we have analyzed the probabilities to find the different reactants and decay products for the o3h clusters formation and rupture . along the equilibrated part of the trajectories ,
we have found 142 formations and 139 ruptures at 3500 k and 104 formations and 103 ruptures at 3000 k. in the case of o3h clusters , the possible reactants and decay products are the oxygen atoms involved in a si - o - h unit ( o@xmath18 ) or a bridging oxygen ( bo ) .
the latter case involves an o - h rupture event whereas the first case does not .
this is the reason why a distinction between _ recombinations _ and _ transfers _ is not easily feasible for these species since the transfered object can either be a hydrogen atom ( as in tab .
[ tab : decay ] ) or an entire o - h group ( as in fig .
[ fig : diffreact3 ] ) .
we hence display the distribution for o3h formations and ruptures including recombinations in fig .
[ fig : deco3h ] .
there are only very small differences between formation and decay . at both temperatures ,
the o3h clusters form and decay from and into bo and o@xmath18 , with almost the same probabilities .
however at 3000 k , the probability from and into o@xmath18 is slightly preferred with a value of 56 % for o@xmath18 vs. 44 % for bo .
this indicates that the hydrogen diffusion process involving the motion of a o - h group without any o - h bond rupture ( process number * 3 . * ) occurs in the liquids as well . to conclude the section on the diffusion process
, we analyzed the activation energy for a hydrogen transfer corresponding to the rupture of an o - h bond ( without recombination ) .
we can calculate the activation energy for a hyrogen transfer reaction assuming an arrhenius behavior @xmath20 where the reaction rate @xmath21 is given by a constant @xmath22 and the activation energy @xmath23 .
using the two numbers of o - h ruptures leading to transfers of 28 for 3500 k and 8 for 3000 k we obtain 280 kj / mol ( 2.91 ev ) for the activation energy ( the different lengths of the equilibrated trajectories already taken into account ) .
unfortunately the constant @xmath22 depends explicitly on the transfers per time and can hence not be determined due to the above mentioned problem of the thermostats .
the activation energies for the generation and dissociation of various si@xmath24o@xmath25h@xmath26 molecules in the gas phase have been determined by zachariah _
@xcite with the gaussian _ ab initio _
package @xcite .
et al _ found a value for the activation energy of the formation of osioh from sio@xmath1 and h of 285 kj / mol which compares very well to the above given one of 280 kj / mol for a hydrogen transfer .
we have equilibrated a liquid of silica containing 3.84 wt .
% h@xmath1o at 3500 k and 3000 k using first - principles molecular - dynamics simulations .
we observed that the water is mostly attached to the silica network in the form of si - oh groups .
water molecules or free o - h groups occur only at the highest temperature but are not stable and decay rapidly .
the sio@xmath27 tetrahedra are still the basic network forming units , as in pure silica .
the short range correlation ( i.e. the radial distribution functions ) suggest that the structure of the matrix is as much changed by the addition of water than as by the addition of the same amount of sodium oxide to the liquid .
in contrast to this , the way the modifier cation itself is attached to the matrix seems to be quite different .
the sodium atoms tend to form bonds of ionic character with oxygen , whereas the hydrogen atoms are attached by covalent bonds .
this difference in the bonding character of o - h and o - na could be the reason for the slower diffusion of the hydrogen atoms in liquid silica compared to that of the sodium atoms .
+ the neutron scattering structure factor shows a pronounced prepeak at a wave vector of about 1@xmath28 , i.e. we have evidence for the presence of long range correlations .
the prepeak is even more intense than that for the sodium silicate at the same wave vector transfer .
the origin of this feature in the hydrous liquid seems to be different from that of the sodium silicate , since the partial structure factor for h - h contributes only little .
furthermore , in the hydrous case , the silica matrix itself seems to be modified since the prepeak occurs as well in the partial structure factors for the matrix .
it remains to be investigated whether the prepeak can be observed at ambient temperature .
the data show a shift to higher @xmath29-vectors and a decrease of the intensity at the lowest temperature . + the other experimentally accessible quantity , the q@xmath30 distribution , shows clearly that the o - h groups try to avoid each other .
silicon atoms with two and more si - o dangling bonds or o@xmath18 ( q@xmath30 sites with @xmath31 2 ) are rarely found . despite the strong covalent character of the o - h bond , the o - h units have a high radial mobility and oxygen triclusters have a high tendency to form .
these oxygen triclusters violate the oxygen stoichiometry and the system compensates this violation by the formation of si - o dangling bonds .
+ the dangling bonds and the triclusters constitute mainly the intermediate states for the diffusion mechanism of the hydrogen atoms in the melt . indeed
three possible diffusion mechanisms have been discussed , two of them involving the rupture of an o - h bond .
free water molecules , hydrogen molecules and free hydroxyl groups are not or only rarely found and can not therefore be made responsible for the hydrogen diffusion in the liquid .
+ coming back to one of the motivating questions , the bubble formation , only little can be concluded from this study . on the one hand , it turned out that the o - h groups are very stable species and it requires a certain stoichiometry violation to weaken the o - h bonds ( note that in o - h - o transition units the stoichiometry of the h atom is also violated ) . on the other hand
these bond weakening transition states occur in sufficiently high quantity to assure the hydrogen diffusion .
nevertheless , a clustering of hydrogen and the formation of stable water molecules do not take place , especially at the lower temperature .
if a the formation of bubbles were realized , we would expect a higher signal in the h - h radial distribution function at typical h - h distances found in liquid water ( 2 - 3 ) and perhaps a signal in the q@xmath30 distribution for @xmath31 2 .
hence we do not see evidence for water clustering at the considered temperatures .
we recognize that the influence of temperature , pressure and silicate composition on the bubble formation is not yet investigated and further work of the present kind will be mandatory .
m. p. thanks the bund der freunde der technischen universitt mnchen and the french science ministry for a grant .
the authors thank also andreas meyer and winfried petry from the technische universitt mnchen and helmut schober and mark johnson from the insitut laue - langevin for fruitful discussions and simona ispas for the ns4 liquid simulation data .
the simulations have been performed on the hitachi sr8000 at leibniz rechenzentrum , munich , germany and on the ibm / sp3 at cines , montpellier , france .
300 s. c. kohn , mineral .
magazine * 64 * , 389 ( 2000 ) .
p. f. mcmillan , rev .
mineral . * 30 * , 131 ( 1994 ) .
d. b. dingwell , science * 273 * , 1054 ( 1996 ) .
d. sahagian .
nature * 402 * , 590 ( 1999 ) .
m. nogami and y. abe , phys .
b * 55 * , 12108 ( 1997 ) .
t. norby , sol .
state ionics * 125 * , 1 ( 1999 ) . c. r. helms and e. h. poindexter .
prog . phys . *
57 * , 791 ( 1994 ) .
e. j. friebele and d. l. griscom , mat .
. proc . * 61 * , 319 ( 1996 ) .
j. sarnthein , a. pasquarello , and r. car , phys .
* 74 * , 4682 ( 1995 ) .
j. sarnthein , a. pasquarello , and r. car , phys .
b * 52 * , 12690 ( 1995 ) .
a. pasquarello and r. car , phys
. lett . * 79 * , 1766 ( 1997 ) .
m. benoit , s. ispas , p. jund , and r. jullien , eur .
j. b * 13 * , 631 ( 2000 ) .
s. susman , k.j .
volin , d.l .
price , m. grimsditch , j.p .
rino , r.k .
kalia , p. vashishta , g. gwanmesia , y. wang and r.c .
liebermann , phys .
b * 43 * , 1194 ( 1991 ) .
s. ispas , m. benoit , p. jund , and r. jullien , phys .
b * 64 * , 214206 ( 2001 ) .
s. ispas , m. benoit , p. jund , and r. jullien , j. non - cryst .
solids * 246 * , 307 ( 2002 ) .
j. horbach , w. kob , and k. binder , j. chem . geol . * 174 * , 87 ( 2001 ) . j. horbach , w. kob , and k. binder , phys .
* 88 * , 125502 ( 2002 ) .
n. zotov , h. keppler , a. c. hannon , and a. k. soper , j. non - cryst .
solids * 202 * , 153 ( 1996 ) .
w. smith , g. n. greaves , and m. j. gillan , j. chem .
phys . * 103 * , 3091 ( 1995 ) .
j. oviedo and j. f. sanz , phys .
b * 58 * , 9047 ( 1998 ) . n. p. bansal and r. h. doremus , _ handbook of glass properties _
( acandemic press , inc .
orlando , san diego , new york , austin , london , montreal , sydney , tokyo , toronto 1986 ) .
holleman , and e. wiberg , _ lehrbuch der anorganischen chemie 101 .
( walter de gruyter , berlin , 1995 ) .
h. behrens and m. nowak , phase transitions * 76 * , 45 ( 2003 ) .
b. p. feuston and s. h. garofalini , j. chem .
phys . * 94 * , 5351 ( 1990 ) . e. b. webb and s. h. garofalini , j. non - cryst .
solids * 226 * , 47 ( 1998 ) . m. benoit and w. kob , europhys .
* 60 * , 269 ( 2002 ) .
a. pasquarello , curr .
opinion sol . state mat .
sci . * 5 * , 503 ( 2001 ) .
r. car and m. parrinello , phys .
lett . * 55 * , 2471 ( 1985 ) .
cpmd version 3.3 , j. hutter , a. alavi , t. deutsch , m. bernasconi , s. goedecker , d. marx , m. tuckerman , and m. parrinello .
mpi fr festkrperforschung and ibm zrich research laboratory ( 1995 - 99 ) .
w. kohn and l. sham , phys .
a * 140 * , 1133 ( 1965 ) .
j. p. perdew , k. burke , and m. ernzerhof , phys .
77 * , 3865 ( 1996 ) .
n. troullier and j. l. martins , phys .
b * 43 * , 1993 ( 1991 ) .
van beest , g.j .
kramer , and r.a .
van santen , phys .
* 64 * , 1955 ( 1990 ) .
g. j. martyna , m. e. tuckerman , d. j. tobias , and m. l. klein , mal .
87 * , 1117 ( 1996 ) .
g. j. martyna , m. e. tuckerman , and m. l. klein , j. chem . phys . * 97 * , 2635 ( 1996 ) . m. e. tuckerman and m. parrinello , j. chem .
* 101 * , 1302 ( 1994 ) .
m. benoit , s. ispas , and m.e .
tuckerman , phys .
b * 64 * , 224205 ( 2001 ) .
a. meyer , h. schober , and d. b. dingwell , europhys .
59 * , 708 ( 2002 ) .
i. farnan , s. c. kohn , and r. dupree , geochim .
. acta . * 51 * , 2869 ( 1987 ) .
m. r. zachariah and w. tsang , j. phys . chem . *
99 * , 5308 ( 1995 ) .
m. j. frish , m. head - gordon , g. w. trucks , j. b. foresman , h. b. schlegel , k. rakhavachari , m. a. robb , j. s. binkley , c. gonzales , d. j. defrees , d. j. fox , r. a. whiteside , r. seeger , c. f. melius , j. baker , l. r. martin , l. r. kahn , j. stewart , s. topiol , and j. a. pople .
_ gaussian 90 _ ; gaussian inc .
pittsburgh , pa 1990 .
|
we use _ ab initio _ molecular dynamics simulations to study a sample of liquid silica containing 3.84 wt.@xmath0 h@xmath1o .
we find that , for temperatures of 3000 k and 3500 k , water is almost exclusively dissolved as hydroxyl groups , the silica network is partially broken and static and dynamical properties of the silica network change considerably upon the addition of water .
water molecules or free o - h groups occur only at the highest temperature but are not stable and disintegrate rapidly .
structural properties of this system are compared to those of pure silica and sodium tetrasilicate melts at equivalent temperatures .
these comparisons confirm the picture of a partially broken tetrahedral network in the hydrous liquid and suggest that the structure of the matrix is as much changed by the addition of water than it is by the addition of the same amount ( in mole % ) of sodium oxide . on larger length scales , correlations are qualitatively similar but seem to be more pronounced in the hydrous silica liquid . finally , we study the diffusion mechanisms of the hydrogen atoms in the melt .
it turns out that hosi@xmath1 triclusters and sio dangling bonds play a decisive role as intermediate states for the hydrogen diffusion .
|
electron neutrinos from @xmath0b decay in the sun have been detected at the sudbury neutrino observatory ( sno ) by means of the charged current ( cc ) reaction on deuterium @xcite .
the result , @xmath1 10@xmath2 @xmath3 s@xmath4 ( here and in the following one sigma statistical errors are combined in quadrature with systematical errors ) , is a factor three smaller than the ssm prediction of bahcall et al.@xcite : @xmath5 the cc reaction is sensitive exclusively to @xmath6 , while electron scattering ( es ) also has a small sensitivity to @xmath7 and @xmath8 .
comparison of @xmath9 to the super - kamiokande ( sk ) precision result on es yields a 3.3@xmath10 difference , providing evidence that there is a non - electron flavor active neutrino component in the solar flux .
extraction of this flux , @xmath11 , can be done in a model independent way by exploiting the similarities of the response functions of sno and of sk , see @xcite . in this way
, one determines from the two experiments the total active neutrino flux , @xmath12 , produced by @xmath0b decay in the sun @xcite : @xmath13 the close agreement with the theoretical prediction ( [ fibssm ] ) is an important confirmation of the robustness of ssm calculations , see table [ tabtc ] .
actually , one can go somehow in more detail and use this experimental information to provide an independent estimate of the accuracy of several input parameters that are used in solar model calculations . among others
, we shall consider the solar metal abundance , opacity and p - p fusion cross section .
we shall also consider the central solar temperature .
as well known , this is not an independent parameter , rather it is the result of solar model calculations , mainly dependent on the assumed values for metals , opacity and luminosity .
the boron flux measurement essentially provides a measurement of the central solar temperature .
so far we neglected the possibility of oscillations of @xmath6 into sterile neutrinos .
in fact one can use the experimental result on active neutrinos @xmath14 and the theoretical prediction on the total flux , @xmath15 to provide an upper bound on the flux of sterile neutrinos on earth . when connecting the @xmath0b flux with quantities characterizing the solar interior , an important link is provided by nuclear physics , through the nuclear cross sections of the processes which lead to @xmath0b production .
we shall discuss the relevance of nuclear physics uncertainties in this respect .
the neutrino flux from @xmath0b decay in the sun depends on several inputs @xmath16 , which can be grouped as nuclear and astrophysical parameters , see table [ tab1 ] .
the first group contains the zero energy astrophysical s - factors for the reactions which are involved in the production of @xmath0b nuclei .
@xmath17 refers to p+p@xmath18d+ e@xmath19 + @xmath20 , @xmath21 and @xmath22 refer respectively to @xmath23he+@xmath23he @xmath24he+2p and to @xmath23he+@xmath25he @xmath18 @xmath26be+@xmath27 . once @xmath26be nuclei have been formed , the probability of forming a @xmath0b nucleus is essentially given by the ratio of proton to electron capture on @xmath26be , i.e on @xmath28 .
the @xmath0b neutrino flux depends on several astrophysical inputs .
concerning the dependence on the solar luminosity @xmath29 and age @xmath30 , a more luminous sun requires a higher internal temperature , which implies a higher boron flux .
an older sun is also a more luminous sun . the metal content z / x determines opacity .
if z / x increases , opacity increases and the radiative transfer of the solar luminosity requires higher temperature gradients , which in turn implies higher temperature and a larger @xmath0b flux
. opacity calculation are complex , and also for fixed metal content different codes yield ( slightly ) different opacities .
for this reason we introduce , by means of a scaling parameter @xmath31 , the possibility of a uniform scaling of opacity in the solar interior .
another scaling factor , d , is allowed for varying the calculated diffusion coefficients . if these inputs are changed with respect to the values used in the ssm calculations , @xmath32 , the @xmath0b - neutrino flux @xmath33 changes according to a power law : @xmath34 we have calculated these coefficients by using solar models which include helium and heavy elements diffusion , see table [ tab1 ] .
they are in agreement with the values calculated ( neglecting diffusion ) in @xcite ( see table [ tab1 ] ) and in @xcite
. most of these values can be understood by using simple , semi - quantitative arguments , see e.g. @xcite . as an example
, one can easily show that at equilibrium the production of @xmath26be nuclei scales as @xmath35 @xcite and that the production of @xmath0b nuclei , which is a minor perturbation of the pp - ii chain , scales as @xmath28 .
each input is affected by uncertainties .
the values shown in table [ tab1 ] , 4-th row , correspond to one sigma percentage error .
they have been obtained according to the following criteria : \i ) for the nuclear cross sections , we generally consider the values and uncertainties recommended in ref .
@xcite which are quite similar to those quoted in ref .
concerning @xmath36 , we quote an uncertainty of 9% , taking into account recent measurements @xcite .
\ii ) concerning solar luminosity , age and metals , we adopt the same uncertainties quoted in @xcite , which are derived by recent measurements of the solar constant , by a detailed comparison of meteorite radioactive datings and by analysis of meteoric and photospheric chemical compositions @xcite .
\iii ) regarding the opacity and diffusion uncertainties , there is no experimental guidance and one has to resort to the comparison among different theoretical calculations .
a factor 2.5% is the typical difference between the calculated values of the livermore and los alamos opacities , see @xcite . concerning diffusion , bahcall and loeb @xcite quote and uncertainty of about 15% by comparing results of different theoretical calculations .
( we have found that agreement with helioseismic information fixes diffusion to the level of 10% @xcite . )
we also show in table [ tab1 ] , fifth line , the contributed fractional error , @xmath37 , i.e the contribution of each parameter to the error on the neutrino flux . summing up in quadrature ,
the nuclear uncertainties yield an error on the flux @xmath38=13.3% , which matches closely the error resulting from astrophysical uncertainties calculated according to the same prescription , @xmath39=12.2% .
all in all , the total calculated uncertainty , @xmath40 is about 18% , in agreement with ref .
the experimental result , eq .
( [ fibfogli ] ) , is in excellent agreement with the theoretical prediction , eq .
( [ fibssm ] ) .
the two determinations are affected by a similar error and , as we have just seen , the theoretical error gets comparable contributions form nuclear and astrophysical uncertainties .
one can expect that in the future , with increasing statistics and with better understanding of the systematical uncertainties , the experimental error will be reduced , possibly by a factor of two . in view of this
, one should improve the theoretical calculation by a comparable factor , which requires a better understanding of the nuclear physics and astrophysics which are involved , see again table [ tab1 ] .
we remind that experiments are only sensitive to active neutrinos .
the comparison with the ssm prediction can be used to derive an upper bound on the presence of sterile neutrinos , by requiring that @xmath41 .
this gives : @xmath42 improvement of this bound will require that both theoretical and experimental determinations become more accurate . neglecting the possibility of sterile neutrinos , one can use the @xmath0b - flux measurement as an independent way of estimating the accuracy of several nuclear and astrophysical parameters .
in fact , the measurement of @xmath33 can be interpreted as a way of determining each of the parameters listed in table [ tab1 ] , by means of : @xmath43 in this way the measurement of @xmath33 together with the independent information on @xmath44 ( @xmath45 ) can be translated into a mesurement of @xmath16 . for each parameter in table [ tab1 ] , we have estimated the corresponding accuracy by taking into account the @xmath0b result and the uncertainty on all other parameters , in turn .
the resulting values , @xmath46 are listed in table [ tab1 ] , last row .
we remark the following points : * the opacity scaling factor @xmath31 is determined to be unity with an accuracy of 9.3% .
this is somehow worse than the estimated input accuracy of 2.6% .
we remind however that this uncertainty was essentially estimated form the comparison between two different theoretical calculations .
* the metal content z / x is confirmed within 18% .
again this is worse than the estimated input uncertainty ( 6.1% ) from studies of solar photosphere , which however might not be representative of the metal content of the solar interior if it has been enriched in metals after the sun formation . *
the astrophysical @xmath47-factor for the pp fusion reaction , @xmath17 , is determined wih accuracy of 9% .
this quantity is not measured and it is the result of theoretical errors .
the estimated uncertainty of these calculations ( input error " ) is 1.7% helioseismology allows a few per cent accuracy @xcite . *
the errors on @xmath21 , @xmath22 , @xmath48 , @xmath36 , @xmath29 and @xmath30 are not competitive with more direct measurements in the literature .
a final remark concerns the accuracy of the central solar temperature @xmath49 ( see table [ tabtc ] for theoretical predictions of @xmath49 ) .
solar model builders have been claiming for decades that @xmath49 is known with an accuracy of one per cent or better .
helioseismology has provided strong support to this finding . although helioseismology measures sound speed and not temperature , consistency with helioseismic data has been found only for solar models which produce @xmath49 within one percent of the ssm prediction @xcite .
the measurement of @xmath0b provides an important test in this respect .
in fact , the temperature is not an independent variable .
its precise value is determined by some of the parameters which we have been discussing : the cross section for the pp - reaction , the metal content of the sun , the adopted values for the radiative opacity , the solar age and luminosity and also the diffusion coefficients .
on the other hand , the central temperature is not affected by the other nuclear parameters listed in table [ tab1 ] . in full generality , the relationship between the boron flux and @xmath49 is ( see @xcite ) : @xmath50 where @xmath51 includes the dependence on the nuclear cross sections concerning the exponent @xmath53 in eq .
( [ fitc ] ) , it is weakly dependent on which parameter is being varied for obtaining a changement of @xmath49 .
we have calculated the exponent by using solar models which include helium and heavy elements diffusion , see table [ tabeta ] and fig .
[ figbeta ] . on these grounds
we will take @xmath54 .
we remark that if one neglects diffusion , slightly higher values of @xmath53 are obtained , see table [ tabeta ] last column .
in fact in ref .
@xcite by studying solar models without diffusion a somehow higher value was found , @xmath55 .
anyhow the choice of a slightly different value dos not induce a significant difference in the conclusions .
the agreement between theory and experiment on @xmath33 thus implies that the central temperature of the sun agrees with the ssm prediction to within one per cent : @xmath56 where the error gets comparable contributions from the uncertainty on @xmath33 and from nuclear physics .
|
by combining the results of sno and super - kamiokande one can derive - in the absence of sterile neutrinos - the total neutrino flux produced from @xmath0b decay in the sun .
we use this information to check the accuracy of several input parameters of solar model calculations .
opacity and p - p fusion cross section are constrained by the @xmath0b flux measurement to the level of few per cent .
the central solar temperature is determined to the one - percent level .
we also find an upper limit for the flux on earth of sterile neutrinos .
we discuss the role of nuclear physics uncertainties on these determinations .
|
different kinds of the non - optical quantum state transport using specific phenomena in `` quantum wires '' are investigated very actively during recent few years .
some references may be found in sec .
[ sec : transf ] and sec .
[ sec : walk ] . in the present work
is discussed a compact theoretical approach to the programmable quantum state transfer .
these methods have applications for the coherent control of quantum systems , the theory of quantum communications and computations .
they also uncover promising relations between different models of the quantum information science .
the programmable state transfer is introduced as a particular kind of conditional quantum dynamics in sec .
[ sec : cond ] .
application of these methods to simple motion along lattices controlled by state of a qubit is discussed in sec .
[ sec : qubot ] .
a relation with programmable extension of so - called perfect quantum state transfer with lattices and spin chains is considered in the sec .
[ sec : transf ] and some possibilities of the control of the coined quantum walk are described in the sec . [ sec : walk ] .
an essential part of quantum information processing , is the `` _ conditional quantum dynamics _
, in which one subsystem undergoes coherent evolution that depends on the quantum state of another subsystem '' @xcite .
a simple example of the conditional quantum dynamics may be written as @xcite @xmath0 we can denote hilbert spaces of these two subsystems as @xmath1 ( `` program '' or a control system ) , @xmath2 ( `` data '' or a target system ) and to consider a gate @xmath3 on @xmath4 . if we _ do not accept entanglement between two subsystems _ ,
most general form of conditional or programmable evolution may be expressed as @xcite @xmath5 \to { | p'_u \rangle } \otimes ( u{| d \rangle } ) , \label{prog}\ ] ] where @xmath6 is _ an arbitrary _ state of the target system and @xmath7 is a state of the control system ( `` a program register '' ) _ implementing _ operator @xmath8 . in @xcite
was shown that if two states @xmath9 and @xmath10 of the program register implement two different operators @xmath11 and
@xmath12 , then eq . ( [ prog ] ) implies @xmath13 due to eq .
( [ ortprog ] ) all states of the program register are orthogonal and the dimension of @xmath1 is equal to the number of different operators we need to implement . it was used in @xcite as an inspiration to the theory of stochastic programmable quantum devices , but there are also implications to usual unitary evolution , discussed in the present paper . due to eq .
( [ ortprog ] ) we may without lost of generality to use states of the control register implementing different programs as a new computational basis @xmath14 @xcite . in such a case
the operator @xmath15 eq .
( [ cond ] ) satisfies eq .
( [ prog ] ) for the basis states , _
i.e. _ , @xmath16 = { | k \rangle } \otimes ( u_k{| d \rangle } )
. $ ] a possible change of the state of the control system in eq .
( [ prog ] ) may be described using the composition of @xmath15 with @xmath17 , _
i.e. _ , an arbitrary unitary operator on the first subsystem . for an arbitrary state @xmath18 of the control
register the operator eq .
( [ cond ] ) does not satisfy eq .
( [ prog ] ) , because states of control and target systems become _ entangled _
@xmath19 = \sum_k{\psi_k\bigl[{| k \rangle } \otimes ( u_k{| d \rangle})\bigr]}. \label{cent}\ ] ] _ the programmable quantum state transfer _ is defined in present work as a quantum network ensuring eqs .
( [ cond ] , [ prog ] ) with _ spatially distributed system as a target_. it should be mentioned , that any universal set of quantum gates with distributed quantum systems might include possibility of information transfer , _
e.g. _ , quantum interfaces @xcite intensively use controlled gates with different purposes .
so it is justifiable to restrict discussion here to more specialized class of quantum networks described by eq .
( [ cond ] ) _ displaying a correspondence principle with standard transport models like lattice gases and random walks_.
perfect cloning of arbitrary unknown quantum states is forbidden @xcite and we may limit consideration to transmission without proliferation . a simple model of such a transfer is a linear lattice with @xmath20 sites , two - dimensional control space , and evolution described by eq . ( [ cond ] ) with @xmath21 , where @xmath8 is the shift operator @xmath22 it may be considered as a rudimentary version of a quantum robot @xcite , and a term `` quantum bot '' or `` qubot '' was suggested for such a system @xcite . in fact , eq .
( [ qubot ] ) corresponds to quantum mechanical notation for a simplest example of one - dimensional invertible _ lattice gas cellular automaton _ ( lgca ) @xcite with `` control bit '' denoting direction of motion .
a similar model is also known in the theory of quantum cellular automata @xcite and coined quantum walks @xcite .
( [ qubot ] ) corresponds to cyclic boundary condition and similar quantum extensions for boundaries with reflections and multidimensional lattices may be also constructed using classical lgca @xcite .
it is possible to quantize the `` almost classical '' model of the state transfer eq .
( [ qubot ] ) and adapt it for quantum control or computations with higher dimensional systems @xcite . for example , in the quantum case it is possible to consider `` @xmath23-th roots '' of the operator @xmath8 with properties and a related question about _ continuous generalization _ of this discrete time model using the limit @xmath24 and even about the hamiltonian of such evolution .
there is quite straightforward approach to such a question @xcite using diagonalization of @xmath8 by the discrete fourier transform @xmath25 , _
i.e. _ , a discrete analogue of transition between coordinate and momenta spaces @xmath26 the matrices @xmath27 and @xmath8 form so - called weyl pair and in continuous limit @xmath8 corresponds to translation operator @xmath28 @xcite . due to eq .
( [ fuf ] ) it is possible to suggest an expression for family of `` roots '' @xmath29 , @xmath30 where @xmath31 , @xmath32 and @xmath33 for @xmath34 . a family of operators @xmath29 and @xmath35 resembles _ the weyl system _ for continuous case @xcite , but satisfy necessary weyl commutation relations only for an integer @xmath36 , @xmath37 and should be discussed elsewhere .
it is also possible to suggest a hamiltonian @xmath38 for the gate @xmath8 @xmath39 hamiltonians for networks with more general topology are discussed in @xcite .
coefficients of the matrices eq .
( [ uroot ] ) and the hamiltonian eq .
( [ uham ] ) may be simply calculated directly @xcite @xmath40 where @xmath41 @xmath42 using eq .
( [ uham ] ) , a hamiltonian of the conditional evolution eq .
( [ qubot ] ) may be represented as @xmath43 similar models may switch over different outputs in according to a state of control .
let us consider two - dimensional control space @xmath1 , two coupled lattices with hilbert space @xmath44 as a target , and the hamiltonian @xmath45 on @xmath46 .
it corresponds to transmission along one lattice for state @xmath47 of control qubit with additional switch between two lattices for @xmath48 .
( [ hs2 ] ) is a degenerated case of `` chessman '' hamiltonian for two - dimensional @xmath49 lattice @xcite , _
i.e. _ , @xmath50 , where @xmath51 are directions of moves .
( [ uham ] ) describes a hamiltonian with long - range interaction and the attenuation law approximately proportional to @xmath52 for @xmath53 .
such a hamiltonian may produce some problem with precise experimental engineering
. it would be good to find an equivalent hamiltonian with nearest - neighbour interaction and it is discussed in the next section .
( [ uroot ] ) shows , that only for integer @xmath54 state is localized @xmath55 , contrary to a nonlocal distribution @xmath56 for real @xmath36 .
in @xcite was suggested a hamiltonian for the quantum spin chain for the perfect state transfer .
let us show , that up to change of basis it produces the same evolution as @xmath38 in eq .
( [ uham ] ) .
it is known @xcite , that the hamiltonian for a heisenberg chain used in @xcite is immediately related with a hamiltonian for one particle on a lattice with tunnelling between neighbor sites , _
i.e _ , with a basic model of the present paper .
similar lattice and graph analogues of the chain networks are also well known in the quantum information science @xcite . here
is suggested for simplicity , that for the transport of a qubit state is used dual - rail encoding @xcite , because such a case has more direct relation with lattice models used here . the operator @xmath57 in eq .
( [ uham ] ) is equal to @xmath58 , where @xmath59 corresponds to a `` fictitious '' particle with spin @xmath60 , @xmath61 is plank s constant , and @xmath62 was introduced after eq .
( [ uroot ] ) .
operators @xmath63 , @xmath64 and @xmath65 have the same eigenvalues and there is some operator @xmath66 : @xmath67 .
it is enough to use the composition of the fourier transform and the transition between @xmath63 and @xmath59 basis ( see @xcite ) .
the operator @xmath68 with a strength parameter @xmath69 corresponds to the hamiltonian of _ the perfect state transfer _ introduced in @xcite and resolves the problem with a `` nearest - neighbour representation '' of the hamiltonian @xmath38 for the shift operator . in this basis instead of
the shift matrix @xmath8 we have higher dimensional representation of a rotation @xmath70 @xcite .
unlike the operator @xmath71 , it displays localization of the initial state @xmath47 only for the extreme points of a lattice ( chain ) , but it is enough for the perfect state transfer . a conditional hamiltonian like eq . ( [ hs ] ) for such a transfer is @xmath72 .
more useful is analogue of eq .
( [ hs2 ] ) , _ i.e. _ , @xmath73 , controlling of switch between two different output lines .
such hamiltonians describe a controlled scalar excitation , but for transfer of a qubit state it is enough to double number of lattices . in general , we can consider such a kind of models as a tensor product of three hilbert spaces @xmath74 , _ viz _ , transmitted state , program and distributed target respectively .
it is also can be considered as an extension of a control system to @xmath75 , when only subsystem @xmath1 may affect on transfer by @xmath2 .
most methods discussed here may be used almost without change both for lattices and spin chains , because for the state transfer via spin chains with @xmath23 nodes nowadays @xcite often is used only @xmath23-dimensional subspace of whole @xmath76-dimensional hilbert space and a lattice with @xmath23 nodes may be used instead . for a spin chain
a simple relation with a lattice model exists only for spin - half particles and so , using one lattice with an internal space @xmath77 for the control and the transferred state , we lose the analogy with spin chains .
it is possible to realize the control with some quantum system attached to a single lattice or use multiple parallel lattices or spin chains @xcite .
the design with spin chains should utilize some interactions @xcite for conditional dynamics .
realistic models of quantum information devices appropriate for such purposes may be found in many papers , from earliest suggestions @xcite till more recent works @xcite .
it should be emphasized , that there is some difference between a model of a global control of such chains @xcite and the programmable dynamics discussed in the present paper .
it is usual distinction between general and programmable quantum networks @xcite , between the external control and the transfer driven by an internal state encoding a program of motion .
a coined quantum walk on a circle @xcite may be considered _ formally _ as a special example of conditional quantum dynamics @xmath78 eq .
( [ qubot ] ) with a control register , altered on each step by the hadamard transform @xmath79 ( or symmetric analogue @xmath80 @xcite ) with @xmath81 steps of evolution described by the operator @xmath82 .
the theory of coined quantum walk has interesting outcome to analysis of the programmable quantum networks , because produces a wide set of examples with feeding a control register by nonorthogonal states .
it was mentioned , that for usual theory of ( non - stochastic ) programmable quantum networks @xcite different states of a program should be orthogonal to ensure eq .
( [ prog ] ) and prevent entanglement between the program and a data eq .
( [ cent ] ) .
it is convenient also to compare the coined quantum walk with the _ programmable quantum processors _
@xcite containing third system , `` a tape '' and a gate @xmath83 for altering of a state of a control register after each step of evolution .
so , instead of one gate @xmath3 eq . ( [ prog ] ) is used an analogue of classical processor timing @xmath84 @xcite .
a similar design may be used for programmable implementation of a coined quantum walk controlled by altered coin(s ) .
different models with set of ( random ) coins provide possibility of `` tuning '' from the quantum to classical - like behavior @xcite .
the programmable implementation of such a model could be compared with the generation of ( pseudo)random numbers and the monte - carlo simulations by a classical computer .
let us recall a quantum bot @xmath78 eq .
( [ qubot ] ) with the control register used as a coin space .
an application @xmath85 with coins like @xmath86 , @xmath87 $ ] provide a smooth transition between an uniform motion and behavior of quantum walk @xcite .
it is possible to introduce a simplest controlled coined quantum walk with @xmath23 different coins @xmath88 and three quantum systems with a step composed from two operators : @xmath89 .
here @xmath90 is @xmath15 operator eq .
( [ cond ] ) for first and second systems , @xmath91 is the operator @xmath78 eq .
( [ qubot ] ) on second and third systems .
if first system has a state @xmath14 during @xmath81 steps , @xmath92 is a `` qubot driven '' quantum walk on third system with a coin @xmath88 .
a state of a coin may be entangled with a state of a lattice .
in such a case the second and third systems may be considered as a joint target , controlled by a state of first system .
generalization of @xmath15 eq .
( [ cond ] ) for continuous parameters may be produced by the simple change of the sum to an integral @xcite and let us use smooth tuning of coins like @xmath93 above .
in this work was considered unified approach to different models of the programmable quantum state transfer . it was used some methods of construction of programmable quantum networks with a higher - dimensional target system adapted for specific properties of distributed dynamical models .
it was shown , that a simple `` qubot '' model may be extended to a programmable system associated with a short - range hamiltonian , coinciding with @xmath68 operator for some fictitious particle with high spin and widely used nowadays in the theory of the perfect state transfer . in the paper
was considered only the _ coined _ quantum walk because of particular structure .
formally , the coin space resembles a specific version of a control register and so a programmable model of such a system should use a `` cascade '' with two control registers for a single target system .
it may be formally treated also using a joint system with a coin and a lattice as a new target for control .
it is shown also , that the application of the theory of programmable quantum networks illustrates some useful relations between three models mentioned above : the quantum bots and lattice gas automata ( see sec . [ sec : qubot ] ) , the perfect quantum state transfer ( see sec . [ sec : transf ] ) , and the coined quantum walk ( see sec .
[ sec : walk ] ) .
a. barenco , d. deutsch , a. k. ekert , and r. jozsa , `` conditional quantum dynamics and logic gates , '' _ * 74 * _ , 40834086 ( 1995 ) . m. a. nielsen and i. l chuang , `` programmable quantum gate arrays , '' _ _ * 79 * , 321324 ( 1997 ) ; _ preprint _ quant - ph/9703032 .
vlasov , `` classical programmability is enough for quantum circuits universality in approximate sense , '' _ preprint _ quant - ph/0103119 .
vlasov , `` universal quantum processors with arbitrary radix @xmath94 , '' _ int .
quantum inf .
, icqi 2001 _ , ? ? ; _ preprint _ quant - ph/0103127 .
s. lloyd , a. j. landahl , and j .- j .
e. slotine , `` universal quantum interfaces , '' _ _ * 69 * , 012305 ( 2004 ) ; _ preprint _ quant - ph/0303048 .
w. k. wootters and w. h. zurek , `` a single quantum can not be cloned , '' _ nature _ * 299 * , 802803 ( 1982 ) .
p. benioff , `` quantum robots and environments , '' _ * 58 * _ , 893904 ( 1998 ) ; _ preprint _ quant - ph/9802067 .
vlasov , `` quantum ` bots ' on lattices , '' ( workshop `` think - tank on computer science aspects [ of quantum computation ] , '' institute for scientific interchange , torino , italy , 1930 june 2000 ) , unpublished .
vlasov , `` on quantum cellular automata , '' _ preprint _ quant - ph/0406119 ( 2004 ) .
t. toffoli and n. margolus , `` invertible cellular automata : a review , '' _ physica d * 45 * _ , 229253 ( 1990 ) .
d. a. meyer , `` from quantum cellular automata to quantum lattice gases , '' _ j. stat .
* 85 * _ , 551574 ( 1996 ) ; _ preprint _ quant - ph/9604003 .
d. aharonov , a. ambainis , j. kempe , and u. vazirani , `` quantum walks on graphs , '' in _ proc .
33rd stoc _ , ( acm , new york , 2001 ) 5059 ; _ preprint _ quant - ph/0012090 .
d. gottesman , `` fault - tolerant quantum computation with higher - dimensional systems , '' _ lect . not .
* 1509 * _ , 302313 ( 1999 ) ; _ preprint _ quant - ph/9802007 .
vlasov , `` algebra of quantum computations with higher dimensional systems , '' _ proc .
spie * 5128 * _ , 2936 ( 2003 ) ; _ preprint _ quant - ph/0210049 .
h. weyl , _ the theory of groups and quantum mechanics _ , ( dover publications , new york , 1931 ) . n. n. bogoliubov , a. a. logunov , a. i. oksak , and i. t. todorov , _ general principles of quantum field theory _ , ( nauka , moscow , 1987 ; kluwer , dordrecht , 1990 ) . j. c. baez , i. e. segal , and z. zhou , _ introduction to algebraic and constructive quantum field theory _ , ( princeton university press , princeton , 1992 ) . v. kotk , g. m. nikolopoulos , and i. jex , `` perfect state transfer in networks of arbitrary topology and coupling configuration , '' _ * 75 * _ , 042319 ( 2007 ) ; _ preprint _ quant - ph/0702016 . l. b. w. jolley , _ summation of series _ , ( dover publications , new york , 1961 ) .
m. christandl , n. datta , a. ekert , and a. j. landahl , `` perfect state transfer in quantum spin networks , '' _ * 92 * _ , 187902 ( 2004 ) ; _ preprint _ quant - ph/0309131 .
m. christandl , n. datta , t. c. dorlas , a. ekert , a. kay , and a. j. landahl , `` perfect transfer of arbitrary states in quantum spin networks , '' _ * 71 * _ , 032312 ( 2005 ) ; _ preprint _ quant - ph/0411020 .
r. p. feynman , r. b. leighton , and m. sands , _ the feynman lectures on physics .
iii . quantum mechanics _ , ( addison - wesley , reading , ma , 1965 ) . m. hamermesh , _ group theory and its applications to physical problems _ , ( addison - wesley , reading , ma , 1964 ) .
a. kay and m. ericsson , `` geometric effects and computation in spin networks , '' _ new j. phys .
* 7 * _ , 143 ( 2005 ) ; _ preprint _ quant - ph/0504063 .
e. farhi and s. gutmann , `` quantum computation and decision trees , '' _ * 58 * _ , 915928 ( 1998 ) ; _ preprint _ quant - ph/9706062 .
d. burgarth and s. bose , `` conclusive and arbitrarily perfect quantum state transfer using parallel spin chain channels , '' _ * 71 * _ , 052315 ( 2005 ) ; _ preprint _ quant - ph/0406112 .
d. burgarth , v. giovannetti , and s. bose , `` efficient and perfect state transfer in quantum chains , '' _ j. phys .
a : math . gen . * * 38 * * _ , 67936802 ( 2005 ) ; _ preprint _ quant - ph/0410175 .
s. c. benjamin , `` schemes for parallel quantum computation without local control of qubits , '' _ * 61 * _ , 020301(r ) ( 2000 ) ; _ preprint _ quant - ph/9909007 .
s. c. benjamin and s. bose , `` quantum computing with an always - on heisenberg interaction , '' _ * 90 * _ , 247901 ( 2003 ) ; _ preprint _ quant - ph/0401071 .
s. c. benjamin , `` multi - qubit gates in arrays coupled by ` always on ' interactions , '' _ new j. phys .
* 6 * _ , 61 ( 2004 ) ; _ preprint _ quant - ph/0403077 .
s. lloyd , `` a potentially realizable quantum computer , '' _ science _ * 261 * , 15691571 ( 1993 ) .
g. berman , g. doolen , d. holm , and v. tsifrinovich , `` quantum computer on a class of one - dimensional ising systems , '' _ phys .
lett . a _ * 193 * , 444450 ( 1994 ) .
j. fitzsimons and j. twamley , `` globally controlled quantum wires for perfect qubit transport , mirroring , and computing , '' _ * 97 * _ , 090502 ( 2006 ) ; _ preprint _ quant - ph/0601120 . j. fitzsimons , l. xiao , s. c. benjamin , and j. a. jones , `` quantum information processing with delocalized qubits under global control , '' _ * 99 * _ , 030501 ( 2007 ) ; _ preprint _ quant - ph/0606188 .
vlasov , `` universal hybrid quantum processors , '' _ part .
* 1[116 ] * _ , 6065 ( 2003 ) ; _ preprint _ quant - ph/0205074 .
vlasov , `` programmable quantum networks with pure states , '' in j. e. stones ( ed ) , _ computer science and quantum computing _ , ( nova science publishers , new york , 2007 ) , pp .
3361 ; _ preprint _ quant - ph/0503230 . j. kempe , `` quantum random walks an introductory overview , '' _ contemp
. phys . * 44 * _ , 307327 ( 2003 ) ; _ preprint _ quant - ph/0303081 .
t. a. brun , h. a. carteret , and a. ambainis , `` quantum to classical transition for random walks , '' _ * 91 * _ , 130602 ( 2003 ) ; _ preprint _
quant - ph/0208195 .
j. kok , v. buek , and m. hillery , `` quantum walks with random phase shifts , '' _ * 74 * _ , 022310 ( 2006 ) ; _ preprint _ quant - ph/0607092 .
v. kendon , `` decoherence in quantum walks a review , '' _ math .
struct . in comp .
sci . * ? * _ , ? ?
( 200 ? ) ; _ preprint _ quant - ph/0606016 .
|
a programmable quantum networks model is used in this paper for development of methods of control of a quantum state transport .
these methods may be applied for a wide variety of patterns of controlled state transmission and spreading in quantum systems .
the programmable perfect state transfer and quantum walk , mobile quantum ( ro)bots and lattice gas automata may be described by unified way with such approach . ]
|
from the point of view of fundamental physics , magnesium diboride ( mgb@xmath0 , discovered to be superconducting below 39 k in 2001 @xcite ) is particularly interesting because it is the clearest example of two - band superconductor ever studied @xcite .
however , this simple compound is also interesting because of its possible applications , that could take advantage of its rather low production cost and of its rather high @xmath1 . at present ,
one of the most promising fields of application is superconducting electronics , with the perspective of mgb@xmath0-based devices with good performances operating at a temperature ( @xmath2 10 - 15 k ) accessible to one - stage cryocoolers .
after the early observation of the josephson effect in mgb@xmath0 break junctions @xcite , the efforts in this directions have led to the fabrication of various kinds of josephson and tunnel junctions @xcite as well as prototypal squids @xcite .
point - contact spectroscopy ( pcs ) has proved particularly useful in the fundamental research on mgb@xmath0 , since it allows measuring in a direct way both the gaps of this compound @xcite with great accuracy @xcite .
actually , being a surface - sensitive probe , pcs is also a unique experimental technique to study the surface superconducting properties of the material .
thus , we also used it as a diagnostic tool within a process of optimization of mgb@xmath0 thin films , in view of the fabrication of josephson junctions of different kinds and , finally , of simple superconducting electronic devices ( e.g. squids ) .
in this paper we will first present some results of relevant fundamental interest we obtained by directional pcs in mgb@xmath0 single crystals in the presence of magnetic fields up to 9 t , among which the temperature dependence of the critical field for @xmath3 and @xmath4 , and the temperature and magnetic - field dependence of the gap amplitudes .
we will show how these results address some important subjects of present debate in the fundamental physics of mgb@xmath0 .
then , we will report the results of pcs measurements in mgb@xmath0 thin films grown by different techniques on various substrates .
these measurements provide information ( e.g. about film orientation ) that can be compared to structural and morphological results , such as xrd and afm / stm .
we will also show that pcs can give very useful indications about the conditions of the very surface ( e.g. , presence and effectiveness of impurities , oxidation , degradation ) that are of crucial importance for the fabrication of tunnel and josephson junctions for electronic applications , but _ can not _ be provided by bulk - sensitive techniques .
the state - of - the - art mgb@xmath0 single crystals used in our experiments were produced by j. karpinski s group at the solid state physics laboratory of eth , zurich ( switzerland ) .
the crystal growth procedure occurs under a pressure of 30 - 35 kbar and is described in detail elsewhere ( see for example ref .
@xcite ) . for our purposes
, we had to use crystals with regular shape ( i.e. flat upper surface and sharp edges ) even if they were rather small ( @xmath5 mm@xmath6 at most ) .
the thin films were produced either at the national electrotechnical institute ( ien ) `` galileo ferraris '' in turin ( italy ) or at the lamia laboratory of the national institute for the physics of matter ( infm ) , in genoa ( italy ) .
films on @xmath7-cut sapphire and mgo ( 111 ) ( both with hexagonal surface symmetry , similar to that of mgb@xmath0 ) were produced by using a two - step procedure by the genoa group @xcite .
the first step consists in the pld deposition of an amorphous , non - superconducting precursor layer in uhv ( @xmath8 mbar ) at room temperature , starting from a stoichiometric mgb@xmath0 sintered target .
the second step is an ex - situ annealing process in magnesium atmosphere , necessary to crystallize the superconducting phase .
the samples are sealed in tantalum crucibles under ar pressure with mg lumps , closed in an evacuated quartz tube and kept at 850@xmath9c for half an hour with a following rapid quenching to room temperature .
in addition to sapphire , the group from turin also used amorphous silicon nitride ( sin ) as substrate , which is particularly advantageous in view of the fabrication of bolometers @xcite .
the 500 nm - thick , amorphous sin layer is grown on silicon wafer by lpcvd .
the mgb@xmath0 films are then deposited by simultaneous evaporation of mg and b at constant temperature in a pressure of 5@xmath10 mbar from a mo resistive heater and from a mo crucible by e - beam heating , respectively .
the resulting precursor is annealed _ in situ _
( i.e. in the deposition chamber ) in ar atmosphere , at 500@xmath9c .
the resulting films look golden - brownish and mirror - like , are always very homogeneous and cover the whole surface of the substrates ( about 5 @xmath11 ) .
as described elsewhere @xcite , the point contacts ( of about @xmath12 ) were made by using a small piece of indium or a drop of ag conductive paint .
this ensured better mechanical stability on thermal cycling and reproducibility of the conductance curves with respect to the standard technique that employs a metallic tip pressed against the sample .
the contacts were placed either on the top surface or on the side of the crystals , so as to inject the current mainly along the @xmath7 axis or along the @xmath13 planes , respectively . knowing the direction of current injection is important because of the anisotropy of mgb@xmath0 .
in fact , pcs gives the @xmath14 characteristics and the conductance curves ( d@xmath15/d@xmath16 vs. @xmath16 ) of a normal metal / superconductor junction . in the case of mgb@xmath0 , the conductance across such a junction contains ( separate ) contributions of the @xmath17 and @xmath18 bands @xcite .
since the @xmath17 bands are almost 2d ( and originate from the overlapping of in - plane @xmath19 boron orbitals ) , their contribution to the conductance is maximum for current flowing along the @xmath13 planes .
this is thus the most favorable configuration to observe clearly the @xmath17-band gap together with the @xmath18-band one .
the use of mgb@xmath0 single crystals allowed us to perform these measurements and to test the predictions of the two - band model concerning the amplitudes of the gaps , their temperature- and magnetic field - dependency , and the relative weight of the two bands for @xmath13-plane and @xmath7-axis current injection .
as in single crystals , point contacts on films were made either by using a spot of ag paint or a piece of indium placed on the top surface . in this case , the weight of the @xmath17-band contribution to the total normalized conductance is an indicator of the film orientation .
for perfectly @xmath7-axis oriented films an almost pure @xmath7-axis conductance is expected , with little or no traces of the @xmath17-band gap ( as an example , see the lower curve in figure[fig1]a ) . however , even in highly oriented films the @xmath17-band contribution could be greater than expected if the surface roughness is not small enough , since microcontacts could be established also on the edge of grains or mesas . similarly , even in films with very flat surface ( such as those grown by hybrid physical - chemical vapour deposition ( hpcvd ) @xcite ) conventional pcs can give an excess @xmath17-band contribution due to the partial penetration of the tip in the films .
probably , this explains the recent results by bugoslavskij and coworkers @xcite .
in addition to the information about film orientation , pcs also provides information about possible enhancements or reductions of the superconducting properties at the surface the clean- or dirty - limit conditions at the surface ( i.e. , roughly speaking , the effectiveness of interband scattering due to impurities ) and the existence of a degraded insulating layer , for example of mgo ( which can arise from the reaction with atmosphere ) .
single crystals . ( a ) typical normalized conductance curves of a @xmath13-plane and of a @xmath7-axis contact .
symbols : experimental data . lines : best - fitting curves ( two - band btk model ) .
( b ) symbols : experimental normalized conductance curves of a @xmath7-axis in contact , measured at 4.6 k in zero field ( `` b=0 '' ) and in the presence of a field of 1 tesla ( `` b=1 t '' ) , together with the difference between these two , vertically shifted by 1 for best comparison ( `` diff.+1 '' ) . solid lines : best - fitting curves obtained by using the functions @xmath20 , @xmath21 and @xmath22 defined in the text .
( c ) temperature dependence of the gaps @xmath23 and @xmath24 obtained from the fit of the partial @xmath17- and @xmath18-band conductances ( symbols ) .
dotted lines : bcs - like @xmath25 curves .
solid lines : predictions of the two - band model in the eliashberg formulation @xcite . ]
figure 1 reports some results of pcs in single crystals . in panel
( a ) symbols represent typical conductance curves of @xmath13-plane and @xmath7-axis contacts . in the first case ( upper curve )
four conductance peaks are present , at @xmath26 mv and @xmath27 mv .
these features are connected to the two gaps , @xmath24 and @xmath23 . in @xmath7-axis contacts ( lower curve ) at the same voltages we observe clearly defined maxima and smooth shoulders , respectively .
such a difference is due to the aforementioned anisotropy of the @xmath17 bands , whose contribution to the conductance is minimum for @xmath7-axis current . that this contribution is not zero is clearly shown by the fact that the standard @xmath28-wave , single - band blonder - tinkham - klapwijk ( btk ) model @xcite is not able to fit the conductance curves . on the contrary ,
the same model generalised to the two - band case , where the total _
normalized _ conductance is @xmath29 , gives rather good results ( solid lines ) .
the best fit is obtained with values of the gaps and of the weight @xmath30 @xcite that agree very well with the predictions of the two - band model in the eliashberg formulation @xcite .
however , the accuracy of these values is rather poor since the fitting function contains 7 adjustable parameters : @xmath23 and @xmath24 , the broadening parameters @xmath31 and @xmath32 ( that account for pair - breaking effects ) , the barrier transparency coefficients @xmath33 and @xmath34 ( proportional to the potential barrier height ) , and @xmath30 .
figure [ fig1]b illustrates the procedure we used to improve this accuracy @xcite .
symbols represent the experimental normalized conductance curves of a @xmath7-axis in contact , measured in zero field ( `` b=0 '' ) and in the presence of a field of 1 tesla ( `` b=1 t '' ) , together with the difference between these two , vertically shifted by 1 for best comparison ( `` diff.+1 '' ) .
as discussed elsewhere @xcite , the @xmath18-band contribution to the experimental conductance is strongly suppressed by the field and , at @xmath35 t , it is no longer detectable by our fitting procedure , while the @xmath17-band gap remains practically unchanged . thus , the conductance curve measured in a field of 1 t practically contains the @xmath17-band conductance alone . if the conductance in zero field is given by @xmath36 , the curve measured in @xmath37 t is obtained by putting @xmath38 and
thus has the functional form @xmath39 .
consequently , the difference between the two curves is expressed by @xmath40 and only contains the @xmath18-band contribution .
fitting the experimental curves reported in the figure ( symbols ) with the corresponding functions @xmath20 , @xmath21 and @xmath41 gives indeed very good results .
the fitting curves are reported in figure[fig1]b as solid lines .
the best - fitting parameters are : @xmath42 mev , @xmath43 mev , @xmath44 from the fit of the curve in magnetic field ; @xmath45 mev , @xmath46 mev , @xmath47 from the fit of the difference .
the clear advantage of this procedure , that allows separating the partial @xmath17- and @xmath18-band conductances , is that the number of free parameters in each fitting function is reduced to three ( @xmath30 is fixed to the value one gets from the fit of @xmath20 , i.e. , in the case of figure [ fig1]b , @xmath48 ) with a consequent reduction of the uncertainty on the gap amplitudes .
if this procedure is repeated at any temperature , one obtains the complete temperature dependence of the gaps reported in figure [ fig1]c ( symbols ) .
two bcs - like curves ( dots ) and the @xmath49 curves predicted by the two - band model @xcite ( solid lines ) are reported for comparison .
it is clear that our data are so precise that they reproduce the theoretically - predicted downward deviation of @xmath24 from the bcs - like behaviour .
however , the actual shape of the curve does not correspond to the theoretical one , maybe indicating that , in our samples , the values of some parameters are slightly different from those used in the model .
-plane junction , in magnetic fields of increasing intensity applied parallel to the @xmath7 axis .
symbols : experimental data ; lines : best - fitting curves obtained with the two - band btk model .
( b ) symbols : magnetic field dependence of the gaps as determined from the fit of the curves in ( a ) ( the scale is on the bottom and left axes ) .
lines : dependence of the maximum pair potential , calculated in ref.@xcite on the reduced field @xmath50 ( the scale is on the top and right axes ) .
the reported lines refer to the case where @xmath51 . , scaledwidth=50.0% ] it is clear that the whole procedure described so far ( that gave one of the clearest direct evidences of the existence of two distinct gaps in mgb@xmath0 up to @xmath1 @xcite ) is based on the disappearance of the @xmath18-band contribution to the conductance in a field of about 1 tesla . to obtain the whole magnetic - field dependence of the gaps ,
we measured the conductance curves of ag - paint , @xmath13-plane contacts at @xmath52 k , in the presence of a magnetic field of increasing intensity applied parallel to the @xmath7 axis .
an example of the results is shown in figure [ fig2]a ( symbols ) .
the solid lines in the same graph are the best - fitting curves of the form @xmath53 , where @xmath30 is assumed to be field - independent . above 1 tesla the best fit is obtained by keeping @xmath54 ( for details on the fitting procedure see ref .
the values of the gaps are shown in figure [ fig2]b as solid symbols .
for comparison , superimposed to the experimental @xmath55 curves we report also the theoretically - predicted values of the maximum order parameter ( calculated at the boundary of the vortex - lattice unit cell @xcite ) in the case where the @xmath18-band diffusivity @xmath56 is 5 times greater than @xmath57 .
the comparison is only qualitative , since pcs rather measures an average of the pair potential in different regions of the vortex lattice , but the agreement is surprising .
this indicates that , even in best - quality single crystals , the @xmath18 band is in the moderate dirty limit , in agreement with the results of stm @xcite and de haas - van alphen measurements @xcite .
in the case of figure [ fig2 ] the junction featured very good curves , but broke down before the normal state was reached .
in many other cases we were able to follow the evolution of the conductance curves up to the critical field @xcite , here defined as the field that marks the return to the normal - state conductance . because of the anisotropy of mgb@xmath0 , the values of the critical field depends very much on whether @xmath4 or @xmath58 .
the resulting values of the critical field @xmath59 are reported in figure [ fig3]a as a function of the reduced temperature @xmath60 . , both for @xmath4 and @xmath61 .
symbols : experimental data obtained from pcs ; the critical field is identified with the field that marks the return to the normal - state conductance .
the gray regions are lower bounded by the values of @xmath59 given by bulk - sensitive techniques ( specific heat @xcite , torque magnetometry @xcite , thermal conductivity @xcite ) and upper - bounded by the values measured by transport ( onset ) @xcite .
( b ) temperature dependence of the anisotropy @xmath62 measured in similar single crystals by different techniques : torque magnetometry ( @xmath63 ) , specific heat ( @xmath64 ) , and pcs ( @xmath65 ) .
, scaledwidth=45.0% ] in the same figure , the shaded regions are lower - bounded by the values of the critical field given by bulk measurements ( i.e. thermal conductivity @xcite , specific heat @xcite and torque magnetometry @xcite ) and upper - bounded by the field that , at any temperature , marks the onset of the superconducting transition in the @xmath66 curve @xcite .
this field has been recently identified @xcite with the surface critical field @xmath67
it is clearly seen that , in both the @xmath3 and @xmath4 configurations , our values practically coincide with the bulk critical field at high temperatures , then approach the surface critical field at lower temperature and finally turn again toward the bulk values .
the `` crossover '' from @xmath59 to nearly @xmath67 always occurs at @xmath68 and thus seems to be temperature - induced rather than field - induced .
the more pronounced curvature of our data with respect to other results gives rise to a temperature dependence of the anisotropy ratio @xmath62 ( solid circles in figure [ fig3]b ) which , at @xmath69 , differs very much from that given by bulk techniques ( open symbols ) .
however , it is interesting to note that a similar @xmath70 behaviour was theoretically predicted and reported in ref .
@xcite in the case where @xmath71 .
again , this indicates that at least at the surface our mgb@xmath0 single crystals present a @xmath18 band in the moderate dirty limit , in complete agreement with the results of figure [ fig2 ] . notice that this conclusion has a great fundamental importance since the role and the amount of intraband scattering in mgb@xmath0 is still a matter of debate , and alternative models ( for negligible @xcite and non - negligible @xcite intraband scattering )
have been proposed to explain experimental findings in this compound .
films grown on sin amorphous substrate .
the curves were measured in liquid helium . also indicated
are the positions of the conductance maxima and of the shoulders related to the energy gaps.,scaledwidth=50.0% ] figure [ fig4 ] reports the conductance curves of two in / mgb@xmath0 point contacts made on the surface of a thin film ( `` film 0 '' ) grown on silicon nitride by co - deposition of mg and b and subsequent _ in situ _ annealing .
the resistive critical temperature of the film is @xmath72 k , and the transition is very sharp ( @xmath73 k ) .
the normal - state resistance of the contacts is also indicated .
the conductance curves feature clear maxima , related to the small gap , at @xmath74 mv ( upper curve ) and @xmath75 mv ( lower curve ) .
a shoulder at about @xmath76 mv is clearly seen in the upper curve , but is also present ( even if smoother ) in the lower one . comparing the curves to those obtained in single crystals ( see figure [ fig1 ] )
it is clear that their shape is compatible with a _ preferential _ @xmath7-axis orientation of the films .
however , the rather evident contribution of the @xmath17 band ( especially in the upper curve ) indicates that this orientation is not complete .
film grown on amorphous silicon nitride .
the presence of plate - like grains , with @xmath7 axis either parallel or perpendicular to the film surface , is clearly seen .
( b ) x - ray diffraction spectrum of the same film . the clear [ 001 ] and [ 002 ] peaks , together with the absence of the other reflections indicated , shows a preferential @xmath7-axis orientation .
, scaledwidth=50.0% ] the afm images of the film surface reported in figure [ fig5]a clarify this point .
the main image and its enlargement ( inset ) clearly show vertically- and horizontally - packed plate - like grains , that in pcs measurements behave as a set of crystals in which the current is injected along to the @xmath7 axis or along the @xmath13 planes .
this explains not only the shape of the curves , with the excess of @xmath13-plane contribution , but also its variability from point to point . an analysis of the images reported in figure [ fig5]a gives the average grain size ( @xmath77 nm ) and the rms roughness , that in this case is about 30 nm ( even if values down to 10 nm were obtained ) .
the small roughness is particularly important in view of the fabrication of planar josephson junctions , where the control of the interface smoothness is crucial .
the xrd spectrum reported in figure [ fig5]b shows clear ( 001 ) and ( 002 ) diffraction maxima , coming from the @xmath7-axis oriented main phase , whose intensity is however smaller than in single crystals .
the absence of a clear ( 101 ) reflection , that is the most intense in powders , indicates that most of the film is oriented , in spite of the amorphous sin substrate .
a very intense spurious peak due to the crystalline si substrate ( on which sin is deposited ) is also shown .
in addition to the information on the orientation , two other important indications can be extracted from the andreev - reflection spectra reported in figure [ fig4 ] .
first , the possibility to observe andreev reflection by pcs means that the contact is
s - n , i.e. the potential barrier @xmath78 at the interface is very small @xcite . in other words , there can not be any insulating layer ( for example , of mgo due to the reaction with air ) covering the surface .
this finding is consistent with the remarkable durability of these films that , if kept in suitably dry atmosphere , retain their properties for several months .
the second important indication concerns the `` clean '' or ` dirty' limit conditions at the surface of the film . in this context
, the expressions `` clean '' and `` dirty '' are referred to the _ interband _ scattering . according to the two - band model appeared in literature @xcite , if the interband scattering is sufficiently strong , the critical temperature falls drastically and the two gaps merge in one single isotropic gap @xmath79 , with bcs - like temperature dependence and zero - temperature value @xmath80 mev .
it is clear that the evidence of two distinct gaps in the curves of figure [ fig4 ] indicates that the surface of the films is in clean limit , even if this kind of measurements can not provide a measure of the scattering rate .
figure [ fig6]a reports the conductance curves of a ag - paint spot contact on a 100 nm - thick mgb@xmath0 film grown on @xmath7-cut sapphire by pld and then annealed _ ex situ _ in mg vapour ( `` film 1 '' in refs.@xcite ) .
the critical temperature of the film , measured by transport , is @xmath81 k , with @xmath82 k. the strong @xmath7-axis orientation of the film is witnessed both by the xrd spectrum reported in figure [ fig6]b ( note that the ( 101 ) peak , that is the most intense in powders , is absent ) and by the large values of the low - temperature anisotropy ( @xmath83 ) , one of the highest reported in literature for films .
the film also features in - plane orientation , with a 30@xmath84-tilt with respect to the substrate , as shown by @xmath85-scan measurements @xcite ; other characteristics are reported in ref.@xcite .
the conductance curve of figure [ fig6]a shows maxima at @xmath86 mv and no clear shoulders .
the absence of multigap features and the position of the maxima suggests that the measured gap is very likely to be @xmath79 . since in the dirty limit
the material becomes isotropic , it is clear that , in this case , pcs is unable to give any information about the orientation of the film .
however , it provides a direct evidence of the degradation of the material , e.g. due to the exposition to air and moisture .
this degradation is definitely restricted to the very top layer of the film ( in fact , transport measurements gave reproducible values of the critical field @xcite even if repeated after several months ) and is thus difficult or even impossible to detect with bulk - sensitive techniques . in this sense
, pcs has the unique capability to probe the conditions of the surface , that is particularly useful in view of the fabrication of planar josephson junctions , where the control of the interface is crucial .
thin film produced by pld from stoichiometric target on @xmath7-cut sapphire and subsequent _ ex situ _ annealing ( `` film 1 '' in refs .
arrows indicate the position of the conductance maxima .
( b ) x - ray diffraction spectrum of the same film , from ref .
only the @xmath87 reflections are evident for the mgb@xmath0 phase , meaning that this phase is @xmath7-axis oriented .
note the presence of a spurious mgo phase , maybe related to the degradation of the surface observed by pcs.,scaledwidth=50.0% ] thin film produced by pld on mgo ( 111 ) and _ ex situ _ annealing .
arrows indicate the position of the conductance maxima and of the shoulders related to energy gaps .
( b ) x - ray diffraction spectrum of the same film , showing clear ( 001 ) and ( 002 ) reflections .
the absence of the ( 101 ) peak was verified by tilting the sample by about 0.5@xmath84 to avoid the superposition with the substrate peak .
the film is @xmath7-axis oriented but features no in - plane orientation.,scaledwidth=50.0% ] figure [ fig7]a reports the conductance curve of a ag - paint contact on a mgb@xmath0 film grown on mgo ( 111 ) substrate ( `` film 2 '' in ref.@xcite ) .
the critical temperature of the film is a little higher than that of film 1 , i.e. @xmath88 k with @xmath89 k. as shown by the xrd spectrum reported in figure [ fig7]b , the film is strongly @xmath7-oriented , even if it does not feature in - plane orientation .
the shape of the conductance curve clearly denotes that the contact is in the `` tunnelling '' regime ( large @xmath78 ) rather than in the andreev - reflection regime ( small @xmath78 ) . as in the case of figure [ fig6]a , the curve in fig .
[ fig7]a presents conductance maxima at energies ( @xmath90 mev ) that correspond to the gap in dirty limit , @xmath79 . in this case
, however , additional features are present in the form of shoulders at @xmath91 mev , that should correspond to the small gap @xmath24 .
this result can be explained by admitting that parallel microcontacts are established between the ag glue and the film , in regions of the film where the surface is clean ( and thus @xmath24 is observed ) and dirty ( where @xmath79 is measured ) . in principle , the absence of any clear evidence of @xmath23 in the conductance curves could be interpreted as a further indication of good @xmath7-axis orientation of the clean material ; however , in the present case the measure is probably too noisy and broadened for this conclusion to be definitely drawn . as in the previous case
, the observation of @xmath79 in some regions of the contact speaks in favour of aging effects , that can not be excluded due to the rather long time elapsed between the film deposition and the pcs measurements ( several months ) .
further support to this hypothesis comes from the fact that we could never observe andreev reflection in contacts made on this films . in other words
, we always obtained s - i - n contacts , with a non - negligible potential barrier at the interface , and with tunnelling - like characteristics .
this is probably due to the presence of an oxide layer on the top of the film
i.e. mgo due to the exposition to air and moisture acting as an insulating barrier and preventing the direct contact between the superconductor and the normal metal . as a matter of fact
, some peaks due to mgo are often found in the xrd spectra , as in figure [ fig6 ] ( in the case of film 2 , however , they are not conclusive since the substrate is made of mgo too ) .
it is worth noting that , again , no evidence of this surface degradation was found in critical field measurements , that always gave a very high value of the anisotropy ( @xmath92 ) supporting the spectroscopic evidence of @xmath7-axis orientation .
in conclusion , we have shown that `` soft '' pcs , that employs small in spots or ag - paste drops to make the `` point '' contacts , presents unique advantages for the investigation of the fundamental superconducting properties of mgb@xmath0 as well as for the characterization of the morphological and electronic properties of crystal and film surfaces .
it is clear that these advantages hold for any superconductor , provided that its surface is sufficiently clean , stable in air , metallic and homogeneous to obtain clear andreev - reflection curves .
morever , some of these requirements are not really strict .
for example , even in the presence of an insulating surface layer ( either intentionally grown or consequent to degradation ) this technique permits the creation of useful weak - link sin junctions ( as shown in figure [ fig7 ] ) . in particular , with respect to standard pcs where a metallic tip is pressed against the sample surface
our `` soft '' pcs technique ensures : i ) greater stability and reproducibility of the junctions at the change of temperature and magnetic field ; ii ) absence of damage or perforation of the surface , that prevents unwanted contributions from the underlying bulk material ; iii ) greater control of the directionality of the contact , that can be essential in the study of anisotropic superconductors like mgb@xmath0 . on the other hand ,
the smaller area of the contact usually obtained by metallic tips can be useful in the study of superconductors with non - homogeneous or granular surfaces . in this respect
, a further improvement of the technique with a reduction of the minimum apparent contact area is under study .
this work was done within the project pra `` umbra '' of infm , the asi contract n. i / r/109/02 and the intas project n. 01 - 0617 .
v.a.s . acknowledges the support from rfbr ( project n. 02 - 02 - 17133 ) and the ministry of science and technologies of the russian federation ( contract n. 40.012.1.1.1357 ) .
10 nagamatsu j , nakagawa n , muranaka t , zenitani y and akimitsu j 2001 _ nature _ * 410 * 63 .
liu a y , mazin i i and kortus j 2001 _ phys . rev .
lett . _ * 87 * 87005 .
brinkman a , golubov a a , rogalla h , dolgov o v , kortus j , kong y , jepsen o and andersen o k 2001 _ phys .
_ b * 65 * 180517(r ) .
choi h j , roundy d , sun h , cohen m l and louie s g 2001 _ nature _ * 418 * 758 .
gonnelli r s , calzolari a. daghero d , ummarino g a , stepanov v a , giunchi g , ceresara s and ripamonti g 2001 _ phys . rev . lett _ * 87 * 097001 .
burnell g , kang d -j , lee h n , moon s h , oh b and blamire m g 2001 _ appl .
lett . _ * 79 * 3464 ; carapella g , martucciello n , costabile g , ferdeghini c , ferrando v and grassano g 2002 _ appl .
* 80 * 2949 ; mijatovic d , brinkman a , oomen i , rijnders g , hilgenkamp h , rogalla h and blank d h a 2001 _ appl .
lett . _ * 80 * 2141 .
zhang y , kinion d , chen j , clarke j , hinks d g and crabtree g w 2001 _ appl .
* 79 * 3995 ; burnell g , kang d -j , ansell d a , lee h n , moon s h , oh b , tarte e j and blamire m g 2003 _ supercond .
technol . _
* 16 * 254 ; brinkman a , veldhuis d , mijatovic d , rijnders g , blank d h a , hilgenkamp h and rogalla h 2001 _ appl . phys . lett . _ * 79 * 2420 .
szab p , samely p , kamark j , klein t , marcus j , fruchart d , miraglia s , marcenat c and jansen a g m 2001 _ phys .
lett . _ * 87 * 137005 .
gonnelli r s , daghero d , ummarino g a , stepanov v a , jun j , kazakov s m and karpinski j 2002 _ phys .
lett . _ * 89 * 247004 .
karpinski j , angst m , jun j , kazakov s m , puzniak r , wisniewski a , roos j , keller h , perucchi a , degiorgi l , eskildsen m r , bordet p , vinnikov l and mironov a 2003 _ supercond .
technol . _
* 16 * 221 ; karpinski j , kazakov s m , jun j , angst m , puzniak r , wisniewski a and bordet p 2003 _ physica c _ * 385 * 42 .
ferrando v , amoruso s , bellingeri e , bruzzese r , manfrinetti p , marr d , velotta r , wang x and ferdeghini c 2003 _ supercond .
technol . _
* 16 * 241 .
monticone e , gandini c , portesi c , rajteri m , bodoardo s , penazzi n , dellarocca v and gonnelli r 2003 _ appl .
_ , submitted .
kashiwaya s and tanaka y 2000 _ rep .
phys . _ * 63 * 1641 .
zeng x h _ et al _ 2002 _ _ nature materials__**1 * * 35 - 38 .
bugoslavsky y , miyoshi y , perkins g k , caplin a d , cohen l f , pogrebnyakov a v and xi x x 2003 _ preprint _ cond - mat/0307540 .
blonder g e , tinkham m and klapwijk t m 1982 _ phys . rev .
b _ * 25 * 4515 .
gonnelli r s , daghero d , ummarino g a , stepanov v a , jun j , kazakov s m and karpinski j 2003 _ physica c _ * 385 * 255 - 263 .
gonnelli r s , daghero d , calzolari a , ummarino g a , dellarocca v , stepanov v a , jun j , kazakov s m and karpinski j , 2003 _ preprint _ cond - mat/0308152 , submitted to _ phys .
_ koshelev a e and golubov a a 2003 _ phys .
lett . _ * 90 * 177002 .
eskildsen m r , kugler m , tanaka s , jun j , kazakov s m , karpinski j and fischer 2002 _ phys .
* 89 * 197003 .
carrington a , meeson p j , cooper j r , balicas l , hussey n e , yelland e a , lee s , yamamoto a , tajima s , kazakov s m and karpinski j 2003 _ phys .
_ * 91 * 037003 .
sologubenko a v , jun j , kazakov s m , karpinski j and ott h r 2002 _ phys . rev .
_ b * 65 * 180505(r ) .
_ 2003 _ physica c _ * 385 * 154 .
angst m , puzniac r , wisniewski a , jun j , kazakov s m , karpinski j , roos j and keller h 2002 _ phys .
lett_. * 88 * 167004 .
eltsev yu 2003 _ physica c _ * 385 * 162 .
d. saint - james and p. g. de gennes , physics letters * 7 * , 306 ( 1963 ) . gurevich a 2003 _ phys .
b _ * 67 * 184515 .
dahm t and schopohl n 2003 _ _ phys .
rev . lett.__**91 * * 017001 ; dahm t , graser s and schopol n 2003 _ preprint _ cond - mat/0304194 .
ferrando v , manfrinetti p , marr d , putti m , sheikin i , tarantini c and ferdeghini c 2003 _ preprint _ cond - mat/0307013 .
|
in this paper we highlight the advantages of using point - contact spectroscopy ( pcs ) in multigap superconductors like mgb@xmath0 , both as a fundamental research tool and as a non - destructive diagnostic technique for the optimization of thin - film characteristics .
we first present some results of crucial fundamental interest obtained by directional pcs in mgb@xmath0 single crystals , for example the temperature dependence of the gaps and of the critical fields and the effect of a magnetic field on the gap amplitudes .
then , we show how pcs can provide useful information about the surface properties of mgb@xmath0 thin films ( e.g. @xmath1 , gap amplitude(s ) , clean or dirty - limit conditions ) in view of their optimization for the fabrication of tunnel and josephson junctions for applications in superconducting electronics .
|
theoretical models of interstellar clouds and clumps most frequently assume static or stationary configurations , the clouds being either confined by the external pressure ( @xcite ; bertoldi & mckee 1992 , herafter bm92 ) , or in `` virial equilibrium '' between its self - gravity and some form of internal energy , be it thermal ( e.g. , @xcite ) , micro - turbulent ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ) or magnetic ( e.g. , @xcite ; @xcite , @xcite ) , forming out of instabilities or coagulation of smaller clouds ( see @xcite for a review ) .
nevertheless , the interstellar medium ( ism ) is well known to be highly violent ( e.g. , @xcite ) , and recently it has become increasingly accepted that it is turbulent throughout ( e.g. , @xcite ; see also the volume `` interstellar turbulence '' [ @xcite ] ) .
in such a medium , clouds may naturally form as turbulent density fluctuations as well . within this dynamic , turbulent framework
, it is important to reconsider the meaning and feasibility of pressure cloud confinement , either thermal or turbulent .
this is most adequately done in the context of the virial theorem ( vt ) , which is a very useful tool for describing the balance between all the physical agents in molecular clouds .
customarily , the vt is considered in a lagrangian frame , i.e. , a frame moving with the flow . however , there are circumstances when one would prefer to use an eulerian description . for example
, a lagrangian description is clearly not well suited for describing the propagation of a wave , in which the fluid does not move along with the perturbation , while in the case of an isolated object a lagrangian description fits most naturally ( see also , passot & pouquet 1996 , hereafter @xcite ) . in the interstellar case , if significant mass exchange occurs between a cloud and its surroundings , such as an accreting cloud , an eulerian ( fixed in space ) may be preferable since a lagrangian boundary would undergo severe distortions .
parker ( 1969 ) derived an eulerian form of the virial theorem ( evt ) in tensor form , but he explicitly neglected the mass flux through the surface of the cloud .
more recently , mckee & zweibel ( 1992 ) ( hereafter @xcite ) have written all the terms entering the evt : @xmath6 where @xmath7 is the moment of inertia of the cloud , @xmath8 is the thermal energy , with @xmath9 being the thermal pressure ; @xmath10 is the kinetic energy , @xmath11 is the surface thermal term , @xmath12 is the surface kinetic term , @xmath13 is the magnetic term , with @xmath14 the maxwell stress tensor , @xmath15 is the gravitational term and @xmath16 is the flux of moment of inertia through the surface of the cloud .
sums over repeated indices are assumed unless otherwise stated .
frequently , the surface terms are neglected altogether , especially in observational work ( @xcite ; @xcite ; @xcite ; @xcite ) , although it is also a common practice in theoretical studies , as a consequence of assuming isolated clouds ( e.g. , @xcite ; @xcite , @xcite ) .
the most notable exception is the thermal pressure surface term @xmath17 , which is frequently invoked for `` pressure confinement '' ( e.g. @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) . by analogy
, @xcite have considered the possibility of turbulent pressure confinement by means of the term @xmath18 .
however , in the present paper we will argue that both kinds of pressure confinement require idealized conditions that are not likely to be realized in the actual ism .
if the interstellar medium ( ism ) is globally turbulent , the velocity field is in general locally nonzero , and thermal pressure balance may be irrelevant , since a cloud can still undergo deformation due to the inertial motions , which will induce a nonzero rhs via the kinetic terms @xmath19 , @xmath18 , and @xmath20
. concerning turbulent pressure confinement , its feasibility requires the underlying assumption that the turbulence be microscopic , so that it can be considered isotropic compared to the scale of the cloud .
however , turbulence is inherently a multi - scale phenomenon , and is expected to contain excitation at scales comparable or even larger to that of the cloud .
it is in this context that the clouds may be considered as density fluctuations produced by larger - scale compressive turbulent motions ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; , passot & pouquet 1995 , hereafter @xcite ) .
such clouds are then turbulent fluctuations , and can either rebound , fragment or collapse depending on the compressive energy available , the cooling ability of the flow , the topology of the velocity and magnetic fields , the production of internal turbulent motions and a number of instability mechanisms ( e.g. , ; vishniac 1983 , 1994 ; @xcite ; tohline , bodenheimer & christodoulou 1987 ; @xcite ; @xcite ; @xcite ; @xcite ) . note that in this scenario , clouds and clumps need not be in a static or quasi - static equilibrium at any time .
moreover , the fact that clouds and clumps are observed to be internally turbulent suggests that the turbulence has not been completely dissipated inside .
in fact , it is known that shocks running obliquely through a density gradient will produce further internal turbulence ( @xcite ; @xcite ) . in this paper
, we analyze the density , velocity and magnetic fields in two - dimensional numerical simulations of turbulence in the ism in order to estimate the importance of the transport processes through the cloud boundaries .
we start in [ kin_sec ] by discussing the role of the kinetic terms in the virial theorem and the implications for pressure confinement .
then , after briefly describing the numerical simulations ( [ simulations ] ) , we proceed to a description of our results ( [ results ] ) , starting with a discussion of the density , velocity and magnetic features that arise , and their spatial correlations ( [ topology ] ) .
then we present an evaluation of the surface and volume kinetic terms in the virial theorem , suggesting that their similarity may be indicative that both types of terms are representative of the same phenomenon ( [ eval_surf ] ) .
the super- or sub - alfvnic character of the motions is discussed in [ super_sub ] . in
[ comp_obs ] we present several comparisons with known observational data , such as line profiles observed in cloud complexes ( [ spectra ] ) , the topology of the magnetic field ( [ b_comparison ] ) , and the lifetimes of the clouds ( [ lifetimes ] ) . in [ discussion ]
we then discuss several issues and possible caveats , like whether the scenario proposed here can be applicable to smaller , denser scales , the low probability of forming quasistatic clumps in a turbulent medium , and the effect of the dimensionality of our simulations .
finally , we give a summary and some conclusions in
[ conclusions ] .
it is interesting to note that the present paper may be regarded as theoretical justification for the suggestion made over 45 years ago by @xcite that an alternative to the picture that visualizes the interstellar medium as consisting of a distribution of discrete clouds might be necessary .
as is well known ( e.g. , @xcite ; @xcite ) , the evt can be obtained from the conservative form of the momentum equation , @xmath21 ( where @xmath22 generically represents the various force densities involved in the problem ) by dotting it with the position vector and integrating over a volume @xmath23 fixed in space .
we remind the reader that the full virial theorem , eq.([evt ] ) , holds always because it is a direct consequence of the momentum equation , although this does not necessarily imply that the system is in equilibrium or , in particular , that the linewidth of an interstellar cloud is directly related to its mass and radius ( @xcite ) .
a detailed analysis of the vt in our simulations will be presented elsewhere ( @xcite ) . in the present paper we focus on the relative importance of the surface and volumetric kinetic terms , and their role in shaping the clouds .
it can be easily checked that , in the evt ( eq . [ [ evt ] ] ) , the kinetic energy @xmath24 , the kinetic surface term @xmath25 , and the time - derivative terms @xmath26 and @xmath27 , originate from the lhs in the momentum equation ( eq . [
[ mom ] ] ) .
while the meaning of the kinetic energy term @xmath19 is obvious , the kinetic surface term @xmath18 has two possible interpretations .
one is as half the flux of the instantaneous rate of change of the moment of inertia @xmath28 through the cloud s surface .
alternatively , it can be interpreted as the sum of the ram pressure plus the kinetic stresses , both evaluated at the surface of the cloud : @xmath29 where @xmath30 in the last surface integral . the first term on the rhs of the eq .
( [ surface_sum ] ) is analogous to the thermal pressure surface term .
the second term , on the other hand , reflects the fact that the turbulent motions are not isotropic , giving off - diagonal contributions to the total pressure tensor @xmath31 ( @xcite ) .
this exhibits in a clear way the difference between the isotropic nature of the thermal pressure and the anisotropic nature of the `` turbulent '' pressure .
another important difference between the thermal and `` turbulent '' pressures lies in the fact that the latter is in general expected to involve motions of scales comparable to that of the cloud , as mentioned in [ intro ] .
but since the velocity field carries mass with it , such large - scale motions necessarily imply a mass flux across fixed , eulerian cloud boundaries .
conversely , in a lagrangian description , such turbulent motions should lead to severe distortions of the lagrangian cloud boundary , possibly giving it a fractal character ( e.g. , scalo 1990 ; @xcite ) .
the exception to the scenario depicted above is the case of strongly self - gravitating clumps in ( magneto- ) hydrostatic equilibrium , from which the gas can not escape once it has been captured ( `` decouples '' from the intercloud medium ) . in this case , the cloud should be bounded by an accretion shock and , although strictly speaking there is flux across the shock , the accreted material is not mixed far beyond the shock into the cloud s `` body '' .
however , it seems to us that the question as to whether such quasi - hydrostatic configurations can actually be produced in a turbulent ism remains open , as discussed in
[ gamma ] .
in the following sections , we discuss the topology of the density , velocity and magnetic fields in 2d numerical simulations of the ism based on the model of passot , & pouquet ( 1995 ) ( hereafter @xcite ) , which represents 1 kpc@xmath32 of the ism on the galactic plane , centered at the solar galactocentric distance .
we refer the reader to that paper for the equations and parameters of the model , which includes self - gravity , magnetic fields , parameterized cooling and diffuse heating , the coriolis force , large - scale shear , and parameterized localized stellar energy input due to ionization heating .
the parameterized cooling is as in @xcite , who fitted piecewise power laws to the standard cooling calculations of @xcite and @xcite . as discussed in @xcite , the net result of the combined cooling and diffuse heating laws
is that the flow effectively behaves as a polytrope , with an effective polytropic exponent @xmath33 which is piecewise constant over the temperature range spanned by the flow .
alternatively , it was shown in @xcite that , for the heating and cooling functions used there , the thermal equilibrium values of the temperature and pressure can be expressed as functions of the density in the range 100 k @xmath34 k for non star - forming regions .
therefore , @xmath33 can also be expressed as a piecewise constant over the corresponding density range .
for the fiducial values adopted in the model ( @xcite ) , @xmath33 takes the values ( @xcite ) @xmath35 where densities are in units of @xmath36 and temperatures in kelvins . in eq .
( [ table_gamma ] ) , equilibrium temperature ranges equivalent to the specified density ranges are indicated in parentheses . in the simulations ,
stellar ionization heating is modeled by means of a local heating source which is turned on if the density exceeds a critical value @xmath37 , where the brackets denote an average over the whole volume of the simulation .
note that temperatures over a few @xmath38 k are never reached in the simulations , because supernovae are not included , since these require non - trivial modifications to the code ( @xcite ) .
the maximum densities reached in the simulations are @xmath39 @xmath36 , with @xmath40 k. the simulations we use are refinements of the @xcite model at a higher resolution ( 800@xmath41800 grid points ) , presented in vzquez - semadeni , ballesteros - paredes , & rodrguez ( 1997 ) , hereafter @xcite . the stellar energy injection maintains the turbulence in the medium , promoting further cloud formation .
nevertheless , we turn off star formation shortly before the time at which the data are analyzed , in order to allow for the largest possible density gradients , since otherwise the stellar heating prevents the density from reaching values significantly larger than @xmath42 , by causing the neighboring gas to re - expand .
the various physical quantities are in units of @xmath43 @xmath36 , @xmath44 km s@xmath45 , @xmath46 k and @xmath47 g ( see @xcite ) .
we have developed a cloud - finding numerical algorithm in physical space ( as opposed to the position - velocity space of observational data ) for the density data from the simulations ( @xcite ) .
a cloud is defined as a connected set of points ( pixels ) whose densities are larger than a given threshold @xmath48 .
the clouds obtained through this procedure are clearly quite arbitrary , since as @xmath48 is increased the boundary of the cloud simply moves `` inwards '' .
however , eventually an increase in @xmath48 will lead to a more qualitative change in which a given cloud splits into two or more `` children '' , in a similar manner to the `` structure trees '' used by @xcite .
this procedure is in a sense analogous to performing observations using different tracers sensitive to different density ranges .
alternative cloud identification algorithms based on laocating density maxima ( e.g. , @xcite ) were not investigated , but we do not expect that such procedures would affect the qualitative nature of our results , which , as will be seen , appear to follow quite generally from the physics of the problem . given the above algorithm for defining the clouds , the kinetic terms entering the vt are then evaluated numerically for each cloud , having previously substracted its bulk mass - averaged velocity , defined as @xmath49 in order to measure only the contribution associated to the fluctuations .
we refer to this as measuring the velocities in the `` cloud frame '' .
in figs . [ family]a , b and c we show three hierarchical levels of the density field ( gray scale ) , with the corresponding velocity fields ( arrows ) . in fig .
[ family]c we further denote super- and sub - alfvnic velocities with black and white arrows , respectively ( see [ comp_obs ] ) , where the alfvn speed is defined as @xmath50 in the code s units .
figure [ family]a shows a snapshot of the simulation at @xmath51yr .
the density field ranges from @xmath52 to @xmath53cm3 .
[ family]b and c show two subsequent hierarchical levels of clouds , obtained at @xmath54 and 8 , respectively .
the sizes of the boxes shown in the three panels are 800 , 215 and 45 pixels per side ( 1 pixel = 1.25 pc ) .
we refer to the cloud in fig .
[ family]b as the `` parent '' cloud , and to that in fig .
[ family]c as the `` child '' cloud .
note that only fig .
[ family]c shows the velocity field in the cloud frame . inspecting the plotted density and velocity fields
, we observe that the fluid velocity at the cloud boundaries is in general continuous , at any level of the hierarchy .
sharp changes in the velocity field ( shocks ) are in general oblique and tend to occur mostly at the centermost parts of the clouds , where the highly filamentary density field exhibits `` ridges '' .
this indicates that the density features are produced by colliding streams , a fact that can be seen more accurately in figs .
[ cut]a , b and c. these show cuts , along the @xmath55- ( at @xmath56 , fig .
[ cut]a ) and @xmath57- ( at @xmath58 and @xmath59 , figs .
b and c respectively ) axes , of the density @xmath60 ( solid line ) , and the @xmath55- and @xmath57-components of the velocity , @xmath61 ( dotted line ) and @xmath62 ( long dashed line ) for a region containing the clouds of figs .
[ family]b and c. the axes in these plots indicate pixel number . for added clarity ,
we show in fig .
[ div ] the lines along which these cuts are taken .
the central parts ( 3 - 5 pixels ) of the density peaks are seen to correspond in general to large negative longitudinal gradients @xmath63 ( no sum over @xmath64 ) of the velocity , i.e. , shocks . however , the clouds ( i.e. , large density values ) extend beyond the shocks for many more pixels , where the velocity component exhibits plateaus with superimposed small - scale structure . these plateaus are indicative of the approaching streams ( extended inward motions ) . as an example , consider the peaks labeled `` 1 '' and `` 2 '' in fig . [ cut]a .
peak 1 corresponds to a shock at @xmath58 , and is flanked by an expansion wave ( @xmath65 ) on the left , and by a plateau with @xmath66 on the right , although the plateau itself contains substructure .
this peak also corresponds to a shock along the @xmath57-direction ( fig .
[ cut]b ) .
interestingly , peak `` 2 '' in fig .
[ cut]a does not seem to correspond to a shock along the @xmath55-direction , but in fig .
[ cut]c it can be seen that it is caused by a shock in the @xmath57-direction . in order to further illustrate this point , in fig .
[ div ] we show the same cloud as in fig . [ family]c , but showing the density field as a gray - scale image and the divergence @xmath67 of the velocity field with countours .
local density maxima clearly coincide with local ( negative ) minima of the divergence .
it is worth noting that in figs .
[ cut]a , b and c , larger - scale density features are also seen to correspond to large - scale negative gradients of the velocity field , although with much substructure .
see , for example , the large cloud complex ( multiple density peak ) extending from @xmath68 in fig.[cut]a .
the very largest - scale inward flows are probably of gravitational rather than turbulent origin ( e.g. , @xcite ) , because the simulations do not include supernovae which could induce turbulent motions at those scales , but in the real ism both cases are likely to occur .
there also appears to be a certain amount of spatial correlation between density and vorticity features . in figure [ vort ]
we show the vorticity field ( contours ) superposed onto the density field ( gray - scale ) for the cloud of fig .
[ family]c . in this case , the correlation between local maxima of the density field and maxima ( or minima ) of the vorticity field is not as tight , the critical points of the vorticity being shifted by a few pixels with respect to the density peaks .
nevertheless , there is still a clear correspondence between vorticity and density features .
this correlation can be understood because , as mentioned above , the stream collisions are in general oblique .
it is possible then that the detailed structure inside the clouds consists of shocks and tangential discontinuities .
moreover , the shocks are generally curved and are encountering density variations in the pre - shock gas , so vorticity production is expected behind them ( @xcite ; @xcite ; @xcite ) .
the continuity of the velocity field through the cloud boundaries has an interesting subjective implication . the definition of a cloud by a density threshold is actually seen to be rather arbitrary , since the clouds do not have sharp physical boundaries .
thus , as long as the change in the density threshold does not imply a `` splitting '' of a cloud into its offspring , it only amounts to considering the same cloud out to different distances from the maximum , but without there being any clear `` edge '' to the cloud .
note that this need not be in conflict with observations suggesting that molecular clouds have sharp edges ( @xcite ) , since in this case the observed boundaries may just reflect a transition from molecular to atomic gas , without a discontinuity in the gas mass density ( e.g. , @xcite , @xcite ) .
the velocity fields observed in the simulations indicate a strong flux of the relevant dynamical quantities through the cloud boundaries at all scales .
to quantify this , in fig .
[ kinetic_terms ] we show a plot of the kinetic energy @xmath19 vs. the kinetic surface term @xmath18 for the parent complex shown in fig . [ family]a and all its daughter clouds resulting from setting @xmath694 , 8 , 11 and 16 , and with areas larger than 300 pixels .
the latter requirement is imposed in order to avoid including clouds so small that they are strongly influenced by viscous and diffusive numerical effects .
the clouds shown in figs . [ family]b and c are represented with a different symbol in fig .
[ kinetic_terms ] for identification .
note that both kinds of virial terms are computed in the cloud frame .
it is seen that in general the surface kinetic virial term is of the same order of magnitude as the volume term , i.e. , the kinetic energy contained in the cloud s volume .
this means that both terms are of similar importance in shaping and supporting the clouds . at a more detailed level
, this seems to be a reflection of the continuity of the velocity field .
the flow is entering the clouds , and shocking at their innermost regions .
thus , the two terms refer to essentially the same process , only measuring it at different places ( one over the cloud s volume , the other at the cloud s boundary but weighted by the distance to the center of mass ) , explaining their similar values .
this is analogous to the well known property of the surface and volume thermal pressure terms , that if the pressure is constant the two terms cancel .
it is important to remark that the similarity between these two terms does not imply that the cloud is `` confined '' by turbulent pressure , in the sense that this is not a static configuration .
first , the two terms do not cancel each other exactly , leaving a net contribution available for shaping the cloud ( the contribution @xmath70 to @xmath71 ) .
second , it is important to note that the points in fig.[kinetic_terms ] lie sometimes above and sometimes below the line @xmath72 , indicating that for some clouds the contribution to @xmath71 is positive , but negative for others .
this implies that the kinetic terms sometimes provide net `` support '' and sometimes cause net compression , although in general the clouds are expected to suffer distortions that can not be classified as either one of the above .
thus , the clouds are evolving with time , continuously changing their volume and shape , since they are the density fluctuations produced by the turbulence .
the thermal pressure , as discussed in @xcite , shows little spatial variation ( except near star formation sites where it is much larger ) because @xmath73 in general for the temperature range spanned by the simulations .
in particular , @xmath74 implies an isobaric medium .
note that @xmath75 is in agreement with the elementary fact that in the ism denser regions are colder in that temperature range ( @xcite ) .
in other words , the near constancy of the pressure in this temperature range is a consequence of thermal balance ( giving @xmath76 ) in a medium whose density is determined by the turbulent motions . however , this near `` pressure equilibrium '' has no effect in confining the clouds , which are in a state of constant change .
as stated in @xcite , the pressure is _ slaved _ to the density field , which in turn is determined by the turbulence .
another point worth discussing is the sub- or super - alfvnic character of the velocity , since it has been traditionally argued that the motions in molecular clouds are supersonic but sub - alfvnic ( e.g. , @xcite ) , although more recently it has been claimed by padoan & nordlund ( 1998 ) that this may not be so , but rather that motions may generally be super - alfvnic . since in our simulations clouds
form out of the general ism , they represent a good test to see which of these conditions develop , rather than using some pre - determined assumption about the sub- or super - alfvnic character of the problem ( e.g. , @xcite ; @xcite ) . in fig .
[ family]c the black and white arrows respectively denote super- and sub - alfvnic velocities .
it is seen that _ inside the cloud there are both sub- and super - alfvnic velocities_. since the velocity field does not show large magnitude fluctuations at the transition sites , this can only be understood in terms of a change in the alfvn speed @xmath77 . to see this , in fig .
[ family]c contours of the magnitude @xmath78 of the magnetic field are shown as well .
the contours span the range from 4 to 10 @xmath79 g , in intervals of 1.5 @xmath79 g .
thicker lines indicate progressively larger values of @xmath78 .
the velocity is seen to be sub - alfvnic everywhere outside the cloud . inside the cloud ,
it is super - alfvnic in the lower region of the cloud . in the upper region ,
the velocity is sub - alfvnic due to a combination of large @xmath78-values and low densities . looking down towards the center , another super - alfvnic region
is seen , due mainly to the increase in the density , which reaches @xmath80 @xmath36 at the peak .
however , at the peak , the velocity becomes sub - alfvnic again , this time due to a sharp decrease in the velocity magnitude itself sort of a `` stagnation point '' at the density peak .
these variations are thus due again to the collisions between the magnetized gas streams , which push the field and cause magnetic shocks and field reversals where the fluid shocks occur , and in general produce a rather chaotic flow .
one such reversal is seen to occur on the filament extending out of the cloud towards the lower left corner of fig .
[ family]c .
this is seen on the contours as a decrease in the field s intensity towards the central ridge of the filament .
figure [ vort ] , which shows the magnetic field vectors on top of the density ( gray scale ) and the vorticity ( contours ) , shows this phenomenon more clearly . note that this trans - alfvnic character of the motions may explain why the density peaks are `` fat '' compared to the stream - collision sites : the information of the presence of the collision is often able to travel upstream of the flow .
furthermore , since the collisions are generally oblique , it is likely that magnetosonic waves propagating perpendicular to the collision surface may overtake the perpendicular component of the oblique fluid motion , even if the total fluid velocity is super - alfvnic .
in contrast , it is well known that in burgers turbulence , where no pressure is available , the density structures are as thin as allowed by viscosity exclusively ( see , e.g. , @xcite ) .
the advection of the magnetic field described above is contrary to the usual assumption that the field controls the fluid motions , forcing the gas to flow along the field lines . nevertheless , it is noteworthy that the field strengths developed by the simulations in both the intercloud medium and in the clouds are not unrealistic .
the lowest values of the field occur at the intercloud medium , and are @xmath81 g , while the largest values occur in the densest regions , with values @xmath82 g .
nevertheless , the field is quite intermittent , and does not follow a unique scaling with the density ( @xcite ) , exhibiting significant variations within the cloud of fig .
[ family]c .
moreover , even under near - equipartition conditions between kinetic and magnetic energy , the velocity field should be expected to exert as much of an influence on the magnetic field as viceversa . in [ b_comparison ]
we discuss the degree to which this field topology is consistent with observations .
the scenario presented here differs significantly from the standard view in which clouds and clumps have well defined boundaries ( density discontinuities ) which separate them from a much more tenuous , warmer medium , as in simulations of cloud collisions ( e.g. , @xcite ; @xcite ; @xcite ) , of cloud confinement ( e.g. , @xcite ) , quasi - static contraction ( e.g. , @xcite ) and stability ( e.g. , @xcite ; @xcite ; @xcite ) , etc . on the other hand
, @xcite has suggested that such a standard model may be an over - idealization stemming from the low resolutions and column density ranges of early studies , and proposed instead a much more complex scenario in which clouds do not have a well - defined identity .
our results support this view , and in fact such poorly - defined identity is reflected in the arbitrariness of cloud- and clump - finding algorithms ( e.g. , @xcite ; @xcite ) .
it is necessary , however , to test whether the scenario proposed here is consistent with different observations .
the colliding streams we have described as responsible for cloud formation should manifest themselves as multiple peaks in density - weighted velocity histograms of one velocity component , the equivalent in our simulations of the optically thin line profiles in observational spectral line maps . although frequently gaussian or other smooth curves are used to fit such lines , conferring the idea that the lines are uni - modal and correspond to well - defined , isolated entities , in actuality the profiles generally exhibit multiple peaks ( for recent examples , see , e.g. , @xcite etc . ) which are in fact normally interpreted as `` clumps '' ( e.g. , williams , blitz & stark 1995 , hereafter wbs95 ) . for comparison with observational profiles ,
we show in fig.[hists ] the density - weighted histogram of the @xmath55-component of the velocity ( @xmath61 ) for the `` parent '' cloud showed in fig .
[ family]b . as usual
, this histogram has been obtained using the local speeds in the cloud frame .
this histogram may be compared with the co line profile for the rossete molecular cloud ( rmc ) shown in fig . 4 of wbs95co spectra shown in fig . 5 of that paper ] .
this cloud is in many aspects comparable to the cloud of fig . [ family]b : while the rmc s dimensions ( as deduced from fig.17 in wbs95 ) are @xmath83 pc , the cloud in fig.[family]b is @xmath84 pc .
furthermore , the cloud in fig .
[ family]b has a mean density of @xmath85 @xmath36 , while the gas sampled in the spectrum of fig .
4 of wbs95 has a mean density @xmath86 @xmath36 . comparing the histograms for both of these clouds ,
we note several points in common .
first , both sets of data have fwhms of roughly 6 when only the main features are considered .
second , both sets exhibit high - velocity bumps , at several from the centroid .
finally , and probably most importantly , the `` main '' features are seen to contain substructure in both sets of plots . however , while such features have been traditionally interpreted as `` clumps '' , in our simulations they originate from extended regions within the cloud . in order to explain this phenomenon , we show in fig .
[ interm ] a different representation of the parent cloud ( fig.[family]b ) .
the isocontours denote the density field , the arrows show the velocity field in the frame of the cloud , and the various tones of gray represent zones with nearly the same @xmath55-component of the velocity @xmath61 .
the velocities range from less than @xmath87 to more than @xmath88 .
each tone of gray indicates a velocity interval of 2 .
several issues are then observed .
first , we note that the various peaks in the histogram of fig .
[ hists ] contain contributions from extended and separate regions in space .
second , local density peaks often encompass several velocity intervals , in a similar fashion to the situation for the entire cloud , suggesting a kind of self - similarity .
note then that the relative height in the histogram depends mainly in the mass fraction of the gas with the corresponding value of @xmath61 , since every velocity interval appears to sample a wide range of densities .
third , a given feature ( peak ) in the histogram may have contributions from regions with very different total velocity vectors , which only happen to have a similar @xmath55-component .
the above results suggest that the identification of `` clumps '' as features in velocity space is risky at best , and possibly erroneous , as already pointed out by @xcite .
a final word concerning the velocity field is in order . as pointed out in [ topology ] ,
significant vorticity is found within the clouds .
observationally , this may be detected as velocity gradients along the cloud s projected area on the sky , which indeed have been observed ( e.g. , @xcite ; @xcite ) .
furthermore , @xcite have recently reported possible direct evidence for vorticity in clouds .
concerning the magnetic field topology , the field reversals and bendings mentioned in [ super_sub ] might seem contrary to the widely accepted fact that the magnetic field lines projected on the plane of the sky , as obtained in polarization analysis ( see , e.g. , @xcite ) , are relatively smooth over the area of the clouds .
however , on closer inspection of the field topology in our simulations , there is no contradiction . as can be seen in fig .
[ vort ] , the magnetic field bendings are correlated with bendings of the density features , so that , along elongated density features , the field appears rather smooth .
besides , the projection along the line of sight can make the field lines appear smoother than they are ( @xcite ) . on the other hand , evidence that the interaction of a shock with dense molecular gas is producing a reversal in the line - of - sight field component in the orion / eridanus region has been recently reported ( @xcite ) .
note also that it is by now well established that the field lines do not in general show a unique direction of alignment with the density features ( @xcite ) . in our simulations
this is also true .
although in particular for the case shown in fig .
[ vort ] the field lines are mostly parallel to the density features , in other cases they are perpendicular ( see , e.g. , the concave region at coordinates @xmath89 and @xmath90 in fig .
[ vort ] ) .
an important consideration is whether the turbulent - fluctuation nature of the clouds proposed here gives cloud lifetimes consistent with current estimates .
first of all , we should point out that clouds are not necessarily completely disrupted by the turbulence .
frequently they are simply sheared , split , merged with others , or they can just re - expand into the general medium if they are not very stronly self - gravitating , again indicating that the medium is not in precise pressure balance ( see the videos mentioned above ) . in any case , a conventional estimate for a typical `` lifetime '' can be given bt @xmath91 , where @xmath92 is a characteristic size of the cloud and @xmath93 its internal velocity dispersion .
as an example , we calculate this for the `` daughter '' cloud of fig.[family]c .
we find @xmath94 , and @xmath95 pixels @xmath96 pc .
thus , @xmath97 yr .
this value is consistent with various estimates of gmc lifetimes of order a few @xmath98 yr ( @xcite ; @xcite ; @xcite ; @xcite ) .
the analyses from the previous sections have been performed on simulations which have reasonably realistic cooling functions and a diffuse heating which decays with increasing density ( mimicking self - shielding ) ( @xcite ) . as seen from eq .
( [ table_gamma ] ) , such heating and cooling rates imply effective polytropic exponents typically smaller than 1 .
these heating and cooling rates are most appropriate for diffuse hi structures , which may be very large ( up to @xmath99 kpc , although small - scale structure must exist there as well ) .
the question then arises as to whether the dynamical scenario we have proposed here is applicable to molecular clouds ( mcs ) and their substructure ( with sizes @xmath100 to @xmath101 pc ) .
although the ultimate test for this problem will require new numerical simulations , which will be presented elsewhere , we can discuss the issue to a certain extent here , based on recent results on the role of @xmath33 ( @xcite ; @xcite ; @xcite ) , which is a measure of the thermal characteristics of the flow .
the main difference between the parameters of the simulations presented here and another set more appropriate for mcs is the effective polytropic exponent of the flow .
( note that the only scale - dependent quantities in our simulations are the heating and cooling rates .
all other quantities are scale - free . ) while in general in our simulations @xmath102
( c.f . [ simulations
] ) , it is possible that in molecular gas @xmath103 ( @xcite ) . actually , the situation at densities @xmath104 @xmath36 is rather uncertain , with @xmath33 possibly being @xmath99 between @xmath105 and a few @xmath38 @xmath36 because of the importance of collisional deexcitation and radiative trapping in co lines , and either larger or smaller than unity at larger densities depending on the presence of embedded stellar sources and other factors ( @xcite ) . in any case
, we can discuss the expected effect of the gas having @xmath103 on the nature of the density fluctuations arising in the flow .
it has recently been shown by passot & ( 1998 ) that at larger @xmath33 the density peaks are spatially more extended and of smaller amplitude , while the density minima are very deep . at small @xmath33 ( typically @xmath106 ) , the situation is reversed , the peaks being very concentrated and large , while the minima are not as deep .
this is because the sound speed @xmath107 scales with the density as @xmath108 , causing the peaks to re - expand faster at larger @xmath33 . at @xmath109
a critical case occurs , the fluctuations of @xmath110 being symmetrical with respect to the mean .
@xcite have calculated the density jump for shocks in a barotropic fluid parameterized by @xmath33 .
thus , if @xmath33 is larger at mc densities than in our simulations , we can expect the density peaks to be wider and less pronounced .
but we do not expect the general scenario of clumps being formed by turbulent compressions to change significantly .
the colliding streams , if supersonic ( or super - magnetosonic ) , create non - stationary shock - bounded slabs , which simply re - expand ( if not made self - gravitating by the compression see below ) on different timescales depending on @xmath33 after the compression subsides . if the compressions are subsonic no shocks form , but , for sufficiently small @xmath33 and/or cloud masses close to their jeans mass , collapse may still be induced by the compressions ( @xcite )
. one crucial issue , however , is whether the compressed slabs can become self - gravitating upon the compression , and , if so , whether a quasi - hydrostatic configuration can be formed .
the former problem has been addressed by a number of authors ( @xcite ; @xcite ; @xcite ; vishniac 1983 , 1994 ; @xcite ; @xcite ) .
two main issues are at play here .
one is the total gravitational ( @xmath111 ) and internal ( @xmath112 ) energies of the cloud ( for the purpose of this discussion the `` internal '' energy can be generalized to include all forms of energy that provide support for the cloud against gravity ) .
the other is the _ rate _ at which each energy increases upon compression . if @xmath113 , the cloud is gravitationally unstable , and collapses upon a perturbation
this is essentially an integral version of the jeans criterion ( e.g. , @xcite ) .
however , for three - dimensional collapse , it is well - known that , if @xmath114 , then the collapse will ulimately be halted , since @xmath112 increases faster than @xmath115 .
the corresponding values of @xmath116 for 2- and 1-dimensional collapse are respectively 1 and 0 ( @xcite ) .
conversely , if a cloud is initially stable according to jeans ( @xmath117 ) , it may be made unstable upon compression .
tohline et al .
( 1987 ) have calculated the required mach number of a 3-dimensional compression as a function of @xmath33 for this to occur , provided that @xmath118 ( for @xmath119 they found that the required mach number is independent of the cloud s mass ) .
two- and one - dimensional compressions require @xmath120 and @xmath121 , respectively ( @xcite ) .
the issue is then the following . in order for an initially stable cloud to be pushed over the internal energy `` barrier '' ( @xcite ) by a strong enough turbulent compression ,
it is necessary that @xmath122 , where @xmath123 is the critical @xmath124 for @xmath115 to increase faster than @xmath112 upon the turbulent compression .
from there on , gravitational collapse takes over . since in general turbulent compressions
are expected to be less than three - dimensional , then generally we expect @xmath125 , where @xmath126 is the critical @xmath124 for gravitational collapse .
then , @xmath127 .
thus , we conclude that , if collapse was triggered by a turbulent compression , it can not later be halted by the internal energy , unless @xmath33 changes in the process .
the latter may happen in the late stages of collapse , if the dynamical timescales become shorter than the radiative cooling timescales , causing @xmath33 to approach the actual heat capacity ratio of the gas @xmath124 , rather than the value derived from the condition of thermal equilibrium ( see , e.g. , @xcite ) .
we should insist here that the effects of all forms of support , such as the magnetic pressure and turbulent support , can in principle be included in @xmath33 .
on the other hand , this is a very simplistic description , since in general @xmath33 may be a function of the density , and the anisotropic nature of magnetic support may not be well described by a simple polytropic exponent . in any case , this discussion illustrates our main point : _ quasi - hydrostatic configurations can not be produced out of turbulent fluctuations , unless the flow becomes closer to adiabatic than to isothermal during the collapse_. fluctuations either rebound or collapse .
however , at small @xmath33 , the rebound time is long ( @xcite ; @xcite ) .
we can examine more quantitatively the conditions under which adiabatic compression or collapse would occur if thermal pressure provides the main support mechanism for a clump , by comparing the cooling timescale @xmath128 ( @xmath129 = cooling rate in units of erg @xmath130 ) to the compression time @xmath131 and the collapse time @xmath132 .
for the cooling rate we distinguish two density regimes .
for @xmath133 @xmath36 the cooling is dominated by co line cooling .
we adopt a rough approximation for the co cooling rate ( see references in scalo et al .
1998 ) as @xmath134 , where @xmath135@xmath36 and we have assumed that collisional deexcitation yields a linear ( rather than quadratic ) density dependence at these densities . we recognize that this is a crude representation of the molecular cooling rate , but it suffices for our order of magnitude estimates . at larger densities the cooling will be dominated by gas - grain collisions ( if there are no embedded protostellar sources ) , with a rate @xmath136 ) , where @xmath137 is the grain temperature and the coefficient depends somewhat on the adopted grain parameters .
we assume @xmath138 for our order of magnitude estimates . for compressions , we find that the condition for adiabatic evolution ( @xmath139 ) yields a critical length scale @xmath140pc @xmath141 where @xmath142 is the mach number of the compression , @xmath143k , and @xmath144@xmath36 . for any reasonable mach number adiabatic compression
is only possible for tiny regions of size @xmath3 pc . if a ( non - adiabatic ) compression results in collapse , the condition for adiabatic collapse is @xmath145 . at densities @xmath146 , the adopted co cooling rate results in the condition @xmath147 .
but at such large densities the cooling will be controlled by gas - grain collisions , not molecular line cooling . using the gas - grain cooling rate ,
the condition for adiabatic collapse is found to be @xmath148 .
but at such small densities molecular line cooling would dominate and be non - adiabatic .
thus for collapse induced by compression , adiabatic evolution will never occur until the collapsing entity is so dense that it becomes opaque to its own infrared radiation , which only occurs at very large ( protostellar ) densities .
these estimates indicate that turbulence - induced compressions or collapse will generally be controlled by the effective polytropic index @xmath33 set by the thermal equilibrium condition , since @xmath149 is smaller than the relevant dynamical timescales , except for extremely small scales of compression or at protostellar densities for collapse .
we conclude that quasi - static configurations are unlikely over a very wide range of physical parameters . is there any observational evidence for a dynamical rather than quasistatic state of clumps ?
a preliminary answer appears affirmative .
indeed , @xcite have recently reported on a system which may be interpreted as a collision of streams in l723 .
furthermore , @xcite have reported a core with inward motions that do not correspond to classical gravitational infall models .
a detailed comparison of these data with our scenario will be presented elsewhere . in this paper
we have used two - dimensional simulations , with the known risks involved by them , such as a different distance dependence of gravity , the absence of a magnetic dynamo , and the possibility of reverse cascades ( although this is probably less important in the compressible case , where the energy spectrum may be determined by the shocks rather than by cascade processes see , e.g. , @xcite ) .
this is a necessary price to pay for the larger resolution allowed by the two - dimensionality , which is essential for the type of hierarchical structure analyses performed here .
however , we do not expect the two - dimensionality of our simulations to be a matter of concern for the validity of our results .
the main points raised here are those of the mechanism of cloud formation and its consequences regarding the mass flux through cloud boundaries , the deformation of the magnetic field by the flow , and the interpretation of line profile observational data .
these do not seem to depend on the dimensionality of the problem , since they originate from the shocks and collisions of streams , which occur in both two and three dimensions . in any case ,
intermediate - resolution 3d simulations are currently being performed which will allow us to test these expectations .
in this paper we have conducted a close examination of the topology of the density , velocity and magnetic fields that result in our 2d simulations of the ism at the kpc scale , in particular within the clouds that emerge as the turbulent density fluctuations in that medium .
we found that , in general , the cloud boundaries are rather arbitrary , being defined only by a density threshold criterion , but not by a physical boundary , such as a density or velocity discontinuity .
the same arbitrariness is likely to be present in clouds defined by the tracer used to observe them , since a cloud may seem to `` end '' where the tracer density falls below a threshold density necessary to excite the transition , but this need not correspond to any abrupt drop in the mass density .
we furthermore considered a hierarchy of clouds in the simulations , by using a sequence of three progressively larger density thresholds for defining them . for clouds at all levels of this hierarchy ,
the velocity field in the simulations is continuous across the clouds `` boundaries '' , indicating that the clouds are being formed by colliding gas streams ( @xcite ; @xcite ; @xcite ) , with the density increasing towards the collision site .
the density enhancements need not be confined exclusively to the collision site as in burgers flows , because the velocities involved are typically trans - alfvnic , meaning that the information on the collision can often propagate upstream ( see also @xcite ) .
furthermore , since the collisions are generally oblique , it is likely that magnetosonic waves propagating perpendicular to the collision surface may overtake the oblique fluid motion , even if the total fluid velocity is super - alfvnic .
this scenario is in sharp contrast with static models in which the clouds are immersed in an intercloud medium of lower density but higher temperature , and confined by pressure balance between the two `` phases '' ( e.g. , @xcite ; @xcite ; @xcite ) . in our simulations both `` phases '' exist , but the transition between them is smooth , and the situation , rather than being static , is highly dynamic . to further characterize the role of the flux across the clouds boundaries , we considered the surface ( @xmath18 ) and volume ( @xmath19 ) kinetic terms in the virial theorem for the whole ensemble of clouds at one temporal snapshot in the simulation .
it was found that both terms are typically of the same order of magnitude , indicating that the surface terms contribute an amount comparable to the total kinetic energy of the clouds to their overall virial balance .
this was interpreted as a signature of the fact that , when the clouds are formed by stream collisions , the same flow is responsible for both the total kinetic energy contained within the clouds and for the flux across its boundaries .
we discussed the differences between the thermal and turbulent pressures , emphasizing the fact that while the former is microscopic and isotropic , the latter may involve macroscopic scales , comparable or even larger than the clouds sizes , and is in general anisotropic , as indicated by its off - diagonal contribution to the total pressure tensor @xmath150 .
we thus concluded that turbulent pressure confinement is a contradictory concept , because it involves motions at large scales which by definition cause a change in the clouds boundaries . in turn , we argued that thermal pressure equilibrium in a non - static medium is irrelevant , because inertial motions may still distort or disrupt a cloud , even under thermal pressure balance with its surroundings . additionally , we studied the magnetic field topology and strength in the clouds in the simulations , and their implication on whether the motions are super- or sub - alfvnic , finding both super- and sub - alfvnic motions within the clouds , supporting the suggestion that the field is highly intermittent ( @xcite ) .
we noticed that the field is significantly distorted and advected by the flow , in agreement with the result that the flow within the clouds is trans - alfvnic .
the field tends to be correlated with the density features , although without a unique mode of alignment ( @xcite ) .
this causes the field to look relatively smooth along density features , in agreement with most polarization studies ( @xcite ) , but at the same time being bent where the density filaments are bent as well .
a crucial question is whether the scenario depicted in this paper is consistent with observational data .
we addressed this problem at the level of density - weighted velocity histograms , which can be compared to observational spectral line profiles from real clouds of similar characteristics to those in the simulations .
although there are limitations to this comparison , our velocity histograms are encouragingly similar to the observational line profiles , both exhibiting comparable fwhms ( @xmath151 ) and high - velocity isolated features , and both containing not completely resolved substructure , indicating multiple components .
we emphasized that while this substructure is usually interpreted as isolated `` clumps '' , in our simulations it originates from different regions of the cloud , which may correspond to different `` streams '' and in general have one similar component of the velocity but very different total velocity vector , as found also by @xcite .
we also mentioned observational evidence for field reversals in shocks ( @xcite ) , and noted that the topology of the magnetic field seen in the simulations is consistent with observations ( @xcite ) .
finally , we discussed at considerable length the possibility that the scenario of clouds as turbulent density fluctuations may apply to molecular clouds and their substructure ( clumps and cores ) . essentially , only the adopted cooling and heating functions in our simulations distinguish them from a case more appropriate to molecular clouds , which are believed to behave in a nearly isothermal manner .
based on recent results on the effect of the effective thermal behavior ( @xcite ; @xcite ; @xcite ) , parameterized by the effective polytropic exponent @xmath33 , we conjectured that only the shape ( height and width ) of the density fluctuations is expected to change , but not the general character of the inflowing streams characteristic of their turbulent nature .
we also suggested that quasi - hydrostatic configurations can not be produced from turbulent compressions , and that density fluctuations must either collapse or rebound , unless the effective thermal behavior becomes closer to adiabatic during the collapse ( @xmath152 ) .
however , for small @xmath33 , the rebound times can be quite long
. we demonstrated , using appropriate cooling rates , that adiabatic evolution , and hence quasistatic configurations , are unlikely to occur either during compressions or during compression - induced collapse . in summary
, we conclude this paper by stating that the results presented here are not in contradiction , but rather quite consistent , with observational data .
they seem to be , however , discrepant with the standard interpretation of those data in terms of sharply - bounded clouds in pressure equilibrium with their surroundings , and with the identification of velocity features in spectral - line data with isolated `` clumps '' .
it is interesting that our results give theoretical support to the suggestion made long ago by @xcite that the interstellar medium may not best be visualized as a system of _ discrete _ clouds . as they pointed out ,
the `` cloud model '' is difficult to displace because , as they say , `` ... a tendency to argue in circles can be noted in the literature , in that confirmation for the _ picture _ of interstellar matter as occurring in the form of discrete clouds is sought in the data analyzed '' .
we thank alyssa goodman , thierry passot , luis rodrguez and jonathan williams for carefully reading early versions of the manuscript and providing us with detailed comments .
we also thank paola dalessio and susana lizano for useful discussions .
we acknowledge the hospitality of the star formation group at the harvard - smithsonian cfa .
the simulations were performed on the cray y - mp/4 - 64 of dgsca , unam .
this work has received partial support from grants unam / dgapa in105295 and unam / cray sc-008397 to e.v .- s . and a unam / dgapa fellowship to j.b .- p .
klein , r. i. , mckee , c. f. , & woods , d. t. 1995 , in the physics of the interstellar medium and intergalactic medium , eds .
a. ferrara , c. f. mckee , c. heiles , & p. shapiro ( san francisco : astronomical society of the pacific ) , p. 366 .
, e. 1998 , `` turbulence in molecular clouds '' , in chemistry , physics , and observations of molecules in space , ed . w. f. wall , a. carramiana , l. carrasco , & p. f. goldsmith ( dordrecht : kluwer ) , in press .
|
we examine the idea that diffuse and giant molecular clouds and their substructure form as density fluctuations induced by large scale interstellar turbulence .
we do this by closely investigating the topology of the velocity , density and magnetic fields within and at the boundaries of the clouds emerging in high - resolution two - dimensional simulations of the ism including self - gravity , magnetic fields , parameterized heating and cooling and a simple model for star formation .
we find that the velocity field is continuous across cloud boundaries for a hierarchy of clouds of progressively smaller sizes .
cloud boundaries defined by a density - threshold criterion are found to be quite arbitrary , with no correspondence to any actual physical boundary , such as a density discontinuity .
abrupt velocity jumps are coincident with the density maxima , indicating that the clouds are formed by colliding gas streams .
this conclusion is also supported by the fact that the volume and surface kinetic terms in the eulerian virial theorem for a cloud ensemble are comparable in general .
the topology of the magnetic field is also suggestive of the same process , exhibiting bends and reversals where the gas streams collide .
however , no unique trend of alignment between density and magnetic features is observed .
both sub- and super - alfvnic motions are observed within the clouds in the simulations . in the light of these results , we argue that thermal pressure equilibrium is irrelevant for cloud confinement in a turbulent medium , since inertial motions can still distort or disrupt a cloud , unless it is strongly gravitationally bound .
turbulent pressure confinement appears self - defeating , because turbulence contains large - scale motions which necessarily distort lagrangian cloud boundaries , or equivalently cause flux through eulerian boundaries .
we then discuss the compatibility of the present scenario with observational data .
we find that density - weighted velocity histograms are consistent with observational line profiles of comparable spatial and velocity resolution , exhibiting similar fwhms and similar multi - component structure .
an analysis of the regions contributing to each velocity interval indicates that the histogram `` features '' do not come from isolated `` clumps '' , but rather from extended regions throughout a cloud , which often have very different total velocity vectors . finally , we argue that the scenario presented here may be also applicable to small scales with larger densities ( molecular clouds and their substructure ,
up to at least @xmath0@xmath1 @xmath2 ) , and conjecture that quasi - hydrostatic configurations can not be produced from turbulent fluctuations unless the thermodynamic behavior of the flow becomes nearly adiabatic .
we demonstrate , using appropriate cooling rates , that this will not occur except for very small compressions ( @xmath3 pc ) or until protostellar densities are reached for collapse . '
# 1#1i"13i * u * # 1#1 # 110@xmath4 # 110@xmath5 # 1@xmath4 # 1@xmath5
|
gas in the interstellar medium ( ism ) in disk galaxies can exist in a molecular , atomic or ionised phase .
most of it will be in the form of neutral atomic hydrogen ( hi ) whose hyperfine transition can be detected at the 21-cm wavelength . since the early work of @xcite ,
the neutral atomic ism has been known to exist in two phases , known as the cold neutral medium ( cnm ) and the warm neutral medium ( wnm ) .
@xcite first suggested that hi seen in emission and absorption had different line widths because it arose from two distinct components of the ism .
the narrow absorption spectra are caused by a cold component of the ism ( the cnm ) with a temperature of @xmath4 whereas the broad emission spectra arose from a warm component with a temperature as high as @xmath5 ( the wnm ) .
the existence of these two components was later confirmed by @xcite using interferometric observations of hi emission and absorption spectra towards 35 extragalactic sources situated at intermediate and high galactic latitude .
the cnm and the wnm are known to coexist in pressure equilibrium over a narrow range of pressure @xcite .
the cnm dominates at high densities and pressures whereas the wnm prevails at low densities and pressures .
these two components have been identified in the milky way , in nearby dwarfs , and spiral galaxies by comparing hi seen in emission and absorption as well as from observations of hi emission alone .
@xcite and @xcite analyzed hi emission velocity profiles of seven nearby dwarf galaxies . by fitting these profiles with gaussian components , they found evidence for a broad component with a dispersion ranging from about 8 to 13 @xmath1 as well as a much narrower component with a dispersion ranging from 3 to 5 @xmath1 .
these velocity dispersions are larger than the predicted thermal line widths for the cnm and the wnm , showing that other energy sources are also playing role in broadening the line profiles .
moreover , they found that the narrow component shows a clumpy distribution and tends to be located near star formation regions , whereas the broad component shows a more ubiquitous distribution . because of the similarities of the observed properties of these two components with those of the cnm and wnm in m 31 and in our galaxy , they associated the narrow component with the cnm and the broad component with the wnm phases of the ism .
a similar study was made by @xcite for the local group dwarf galaxy ngc 6822 .
they also found narrow and broad hi components with mean velocity dispersions of 4 @xmath1 and 8 @xmath1 , respectively . here also , the narrow component is usually located near star forming regions , whereas the broad component tends to be found along every line of sight .
@xcite analyzed hi emission of 11 nearby spiral galaxies and found what he called a high brightness filamentary network ( hbn ) of hi emission which he associated with the cnm and a diffuse low brightness emission which he identified with the wnm .
the hbn dominates inside the optical radius @xmath6 where it accounts for 60 - 90 % of the hi emission , whereas the diffuse component dominates outside this radius . by averaging spectra with peak brightness temperature higher than 4@xmath7 within annular ellipses in his sample
, he found that the combined spectra were characterized by a narrow core superposed on broad lorentzian wings .
the width of the narrow line cores allowed @xcite to give an upper limit of 300 k for the hbn .
@xcite also found that the kinetic temperature of the hbn increases with increasing radius .
@xcite interpreted this as a result of the decrease in the mid - plane thermal pressure towards the outer disks .
understanding the properties of the gas in the ism is crucial as it is the fuel for star formation .
most of this star formation occurs in giant molecular clouds ( gmcs ) .
these are self gravitating clouds with masses @xmath8 and sizes @xmath9 ( e.g. * ? ? ?
hydrogen molecules ( @xmath10 ) are the main constituents of gmcs , but unfortunately @xmath10 is not easily observable .
carbon monoxide ( co ) is the second most abundant molecule in gmcs and can be detected at millimeter wavelengths .
typically , co line emission is used to infer the amount and distribution of @xmath10 in galaxies using some co - to-@xmath10 conversion factor ( e.g. , * ? ? ?
* ) . however , the strength of the co emission depends on the metallicity of the ism @xcite and it is therefore difficult to detect in low metallicity dwarf galaxies .
an alternative approach is thus necessary to investigate the distribution of @xmath10 there .
one example of this is to use the dust emission at infrared wavelengths as a tracer of the molecular gas ( e.g. , * ? ? ?
* ) . another approach , which will be the topic of this paper , is to try and identify a precursor of the molecular gas .
if the h@xmath11 component forms from the cooling of the cnm , then identifying the cnm in galaxies can yield independent information on the properties of a component of the gas that is closely linked to star formation . in this paper
, we attempt to detect the cnm and wnm components of the ism by analyzing the shapes of stacked hi emission profiles of galaxies observed as part of the hi nearby galaxies survey ( things , * ? ? ?
things is a large survey of 34 nearby spiral and dwarf galaxies performed with the nrao very large array ( vla ) .
the purpose of things was to obtain high spectral and spatial resolution maps of the neutral hydrogen in these galaxies .
the things galaxies have distances between 2 and 15 @xmath12 , star formation rates of @xmath13 to @xmath14 , absolute @xmath15-luminosities of @xmath16 to @xmath17 mag , total hi masses of ( 0.0114 ) @xmath18 , and metallicities of 7.5 to 9.2 in units of [ 12 + log(o / h ) ] . using the vla b , c and d arrays ,
the things targets were observed at a spatial resolution of @xmath19 ( robust weighting ) or @xmath20 ( natural weighting ) , respectively .
two of the things galaxies were observed at a spectral resolution of 1.3 @xmath1 and the rest at a resolution of either 2.6 or 5.2 @xmath1 ( see table 2 of @xcite ) .
we use the natural - weighted ( and non - residual scaled ) cubes in this paper , though our results are independent of the weighting scheme used .
we refer the reader to @xcite for an explanation on how the data cubes were produced .
we use the task pbcorr in the groningen image processing system ( gipsy ) to correct for the effect of the primary beam of the telescope in the image cubes .
the publicly available things moment maps already are corrected for primary beam attenuation .
this paper is organized as follows . in section [
sec : method ] , we describe the principles of our method for obtaining high signal - to - noise ( s / n ) stacked profiles .
section [ sec : fitting ] describes the fitting and decomposition of the profiles . in section [ sec : sys_effects ] , we discuss various effects that may influence the shapes of the profiles . in section
[ sec : cleansample ] , we describe how we define a clean sample based on the studies of these effects . in section [ sub : asymm_effect ] , we check the reliability of the profile shapes of the clean sample . in section [ sec : globaltrend ] , we check possible correlations between the shapes of the profiles and properties of galaxies or morphology . in section [ sec : moleculargas ] , we investigate whether the narrow component is associated with molecular gas .
we summarize our results in section [ sec : summary ] .
individual velocity profiles in typical hi observations generally have a low ( @xmath21 ) to moderate ( @xmath22 ) s / n . in the things observations , fitting individual , separate profiles in a data cube needs to be done carefully to avoid skewed results due to the effects of noise . here
, we use a method analogous to the _ stacking _ method sometimes used in high - redshift hi observations ( e.g. , * ? ? ?
* ) to construct high s / n profiles .
note that we stack individual spectra within galaxies rather than combining global galaxy spectra .
since individual profiles located at different positions in a galaxy have different radial velocities , the profiles can not be summed as they are , but need to be shifted in velocity to the same reference velocity .
we use the gipsy task shuffle to do this .
this task uses a velocity field to define the amount by which a profile in the corresponding data cube needs to be shifted .
we use gauss - hermite @xmath23 polynomial velocity fields as input for shuffle .
gauss hermite polynomial velocity fields give robust estimates for the radial velocity of the peak of the profile , even in the presence of profile asymmetries . for an extensive description see @xcite . to minimize the effects of noise ,
we only include individual profiles whose peak fluxes exceed the 3@xmath7 noise level in the data cubes .
for each galaxy , we sum the shuffled profiles to produce what we call _ super profiles_. these are high quality profiles , which due to their increased s / n allow us to reliably explore the presence of the cold and warm neutral components of the ism .
however , as information on the individual profiles is lost during the stacking , care needs to be taken that systematic effects and possible asymmetries in the input profiles are quantified , to prevent these from mimicking the presence of multiple components in the super profiles .
this is discussed in detail in section [ sub : asymm_effect ] .
we show in figure [ fig : superpro_ex ] examples of the super profiles of three galaxies ( one quiescent galaxy : ddo154 , one starburst galaxy : ngc 1569 , and an interacting galaxy : ngc 5194 ; the rest are shown in appendix a ) .
these super profiles are summed over the entire hi disks of the galaxies using the 3@xmath7 cut - off mentioned previously .
typically , each super profile is the sum of a few hundred to a few thousand independent individual profiles ( see table [ tab : table_superpro ] ) .
we also plot @xmath24 uncertainties ( as defined in section [ sec : fitting ] ) , though in many cases these are smaller than the symbols used to plot the profiles .
the super profiles are characterized by broader wings and narrower peaks than purely gaussian profiles .
most of the super profiles are symmetrical .
these properties are all similar to what was found for the individual profiles of seven nearby dwarf galaxies studied by @xcite and @xcite , but observed at a much higher s / n .
|
we analyze the shapes of the hi velocity profiles of the hi nearby galaxy survey ( things ) to study the phase structure of the neutral interstellar medium ( ism ) and its relation to global galaxy properties .
we use a method analogous to the stacking method sometimes used in high redshift hi observations to construct high signal - to - noise ( s / n ) profiles .
we call these high s / n profiles _ super profiles_. we analyze and discuss possible systematics that may change the observed shapes of the super profiles . after quantifying these effects and selecting a sub - sample of unaffected galaxies , we find that the super profiles are best described by a narrow and a broad gaussian component , which are evidence of the presence of the cold neutral medium ( cnm ) and the warm neutral medium ( wnm ) .
the velocity dispersion of the narrow component range from @xmath03.4 to @xmath08.6 @xmath1 with an average of [email protected] @xmath1 , whereas that of the broad component range from @xmath010.1 to @xmath024.3 @xmath1 with an average of [email protected] @xmath1 .
we find that the super profile parameters correlate with star formation indicators such as metallicity , fuv - nuv colors and @xmath3 luminosities .
the flux ratio between the narrow and broad components tends to be highest for high metallicity , high star formation rate ( sfr ) galaxies .
we show that the narrow component identified in the super profiles is associated with the presence of star formation , and possibly with molecular hydrogen .
|
using methods from stochastic geometry and spatial statistics , a stochastic model is developed which describes the 3d nanomorphology of composite materials , being blends of two different ( organic and inorganic ) solid phases .
such materials are used , for example , in photoactive layers of hybrid polymer zinc oxide ( zno ) solar cells where the two solid phases play the role of a polymeric electron donor , consisting of , for example , poly(3-hexylthiophene ) , and an inorganic zno - electron acceptor , respectively .
there is a great advantage of polymer solar cells due to their potentially low production costs , in comparison with classical silicon solar cells .
however , the efficiency of polymer solar cells critically depends on the intimacy of mixing of the donor and acceptor semiconductors used in these devices to create charges as well as on the presence of unhindered percolation pathways in the individual solid phases of the composite material to transport positive and negative charges toward electrodes ; see , for example , @xcite .
it is therefore very important to have tools at one s disposal which are suitable to analyze and model the 3d morphology of these materials quantitatively .
so far , no such tools are available in literature due to the fact that imaging of the 3d morphology in high resolution is a difficult task . the first 3d images of photoactive layers in polymer solar cells , gained by means of electron tomography ( et ) , have been published only recently ; see @xcite and @xcite . in the present paper ,
such 3d images are used to fit our model to real data .
the model then helps to get a better insight into the impact of morphology on the performance of polymer solar cells and , simultaneously , it can be used for virtual scenario analyses , where model - based morphologies of solar cells are simulated to identify polymer solar cells with improved nanostructures . the model developed in this paper
is based on methods from stochastic geometry and spatial statistics ; see @xcite and @xcite for comprehensive surveys on recent results in these fields . in particular , stationary marked point processes are considered as models for complex point patterns extracted from et images , where the points are associated with additional information , so - called `` marks . ''
note that point processes in 3d have been used for many years to analyze geometrically complex point patterns ; see , for example , @xcite .
more recently , further case studies in 3d point process modeling have been performed , for example , in @xcite , @xcite and @xcite ; see also @xcite . besides , there are many monographs dealing with point processes in multidimensional spaces and their statistical inference and simulation .
we refer , for instance , to @xcite , @xcite , @xcite , @xcite , as well as @xcite .
the paper is organized as follows .
section [ sec.pre.lim ] briefly describes the considered solar cells and the corresponding image data on which the model is based . in particular , in section [ sec.multi-scale ] , the main ideas of a multi - scale approach to the segmentation of 3d images are summarized , which has been developed recently in @xcite .
the crucial step of this approach is to find an efficient representation of the binarized and morphologically smoothed images by unions of overlapping spheres .
then , in section [ sec.simulationmodel ] , the spatial stochastic model for the zno phase is introduced , separately for morphologically smoothed zno domains ( macro - scale ) and for those parts representing the difference between the smoothed and nonsmoothed binary images ( micro - scale ) . based on unions of overlapping spheres representing the zno domains ,
that is , marked point patterns extracted from et images , a stochastic model is built for the smoothed 3d morphology ( macro - scale ) of the photoactive layers considered in this paper . since a strong correlation of midpoints of spheres in @xmath0-direction
is observed , we propose a multi - layer approach considering sequences of correlated 2d point processes to model the 3d point patterns of midpoints .
the members of these sequences belong to a suitably chosen class of planar poisson cluster processes , being parallel to the @xmath1@xmath2-plane .
in particular , a generalized version of matrn cluster processes is considered , where the cluster points are scattered in uniformly oriented ellipses around their cluster centers ( section [ sec.mod.sli ] ) . to model the 3d point patterns of midpoints , a markov chain with stationary initial distribution is constructed , which consists of highly correlated matrn cluster processes ( section [ sec.spa.bir ] ) .
it can be seen as a stationary point process in 3d , where the radii of spheres are considered as marks of this point process ( section [ sec.sim.rad ] ) .
subsequently , a spatial stochastic model for the micro - scale part of the morphological structure is developed ( section [ sec.simulationmodelmicro ] ) .
it is used to invert morphological smoothing and completes our model for the 3d morphology of hybrid polymer - zno solar cells .
furthermore , a method to fit model parameters to real image data is proposed.=-1 section [ sec.mod.val ] deals with model validation . to evaluate the goodness of fit
, we compare model characteristics which have been computed from real and simulated data , respectively , like the volume fractions of voxels contributing to monotonous percolation pathways through the photoactive layer , the distribution of spherical contact distances , and the probabilities of exciton quenching .
these characteristics have already been considered in @xcite , since they are strongly related with the performance of solar cells .
finally , section [ sec.con.out ] concludes and provides a short outlook regarding possible future research .
in this section some necessary background information regarding the functionality of polymer solar cells is provided , together with corresponding image data on which the model is based .
we consider photoactive layers of hybrid polymer zinc oxide ( zno ) solar cells where the two solid phases play the role of a polymeric electron donor , consisting of , for example , poly(3-hexylthiophene ) , and an inorganic zno - electron acceptor , respectively . upon exposure to light ,
photons are absorbed in the polymer phase and so - called `` excitons , '' that is , photoexcited electron
hole pairs , evolve .
excitons are neutral quasi - particles which diffuse inside the polymer phase within a limited lifetime ; see @xcite . if an exciton reaches the interface to the zno phase , it is split up into a free electron ( negative charge ) in the zno and a hole ( positive charge ) in the polymer phase .
this process is commonly referred to as quenching , because it reduces the intrinsic fluorescent decay of the exciton in the polymer .
provided that the electrons in the zno phase and the holes in the polymer phase reach the electrodes at the top and bottom of the photoactive layer , respectively , photocurrent is generated .
a schematic illustration of the morphology of photoactive layers in hybrid polymer - zno solar cells is shown in figure [ fig.schematicview ] , where the electrodes are supposed to be parallel to the @xmath1@xmath2-plane .
for further information about polymer solar cells and the physical processes therein we refer , for example , to @xcite . ) and electrons ( @xmath3 ) . ]
note that the extent of blending of the two materials has a large impact on the efficiency of these solar cells , because not all excitons are quenched due to their limited lifetimes .
thus , a morphology as displayed in figure [ fig.schematicview ] , where both materials are mixed intimately , is desirable since the excitons are likely to reach the interface and charges can be generated .
in other words , for a morphology which would be ideal with respect to this aspect of functionality , each location of the polymer phase should have a distance to the zno phase that is smaller than the diffusion length of excitons .
for each location within the polymer phase , the fraction of excitons reaching the interface is called the quenching probability at this location .
the mean of these quenching probabilities , that is , the quenching probability at a randomly chosen location of the polymer phase , is called the quenching efficiency . furthermore , the existence of unhindered percolation pathways within both phases , zno and polymer , is crucial since the generated charges have to be transported to the electrodes throughout the phases .
because of the electric field between the electrodes , these pathways should be preferably monotonous .
hence , to obtain solar cells with high efficiency , an intimately mixed morphology with monotonous percolation pathways for both charge carriers is desirable and should be taken into account when producing devices .
the stochastic model developed in the present paper will be used to identify morphologies with improved efficiency by generating virtual morphologies and investigating the transport processes of electrons and excitons , respectively .
this will be the subject of a forthcoming paper .
the image data have been gained by electron tomography ( et ) ; see @xcite and @xcite . in particular , we consider images for three devices with different thicknesses of the photoactive layers : 57 nm , 100 nm and 167 nm .
for each of the three thicknesses , the 3d et images are given as stacks of 2d grayscale images ( being parallel to the @xmath1@xmath2-plane , say ) , which are numbered according to their location in @xmath0-direction .
the sizes of these images in the @xmath1@xmath2-plane are @xmath4 voxels for the 57 nm film , and @xmath5 voxels for the other two thicknesses .
each voxel represents a cube with side length of 0.71 nm .
figure [ fig.raw_data_slices ] shows representative 2d slices for the three film thicknesses , where the darker parts of the images represent the zno phase due to a higher electron density compared to polymer .
the images displayed in figure [ fig.raw_data_slices ] indicate clear structural differences for the three films .
with increasing layer thickness , the separated domains of polymer and zno are getting finer .
in particular , the thinnest film , that is , the photoactive layer with thickness of 57 nm , features large domains of both polymer and zno .
the stochastic 3d model to be fitted takes these morphological differences into account .
more precisely , the model type will be the same for all three layer thicknesses .
only the values of some model parameters will be different for the varying morphologies ; see section [ sec.simulationmodel ] below . to develop a stochastic model for the 3d morphology of hybrid polymer - zno solar cells ,
the 3d et grayscale images have to be binarized appropriately .
binarization is necessary since we need to decide which voxels are classified as polymer and which as zno .
an elementary approach to binarize grayscale images is to use a global threshold : voxels are set to white ( polymer ) if their grayscale value exceeds a certain threshold , and are otherwise set to black ( zno ) .
however , it is difficult to find a single global threshold to binarize the et images because of the irregular brightness of these images .
thus , instead of considering global thresholds , a method of adaptive thresholding has been used for binarization ; see @xcite .
this method is based on techniques of @xcite and @xcite , where the main idea is to construct a threshold surface which is location - dependent and takes local conditions like overexposure or underexposure into account .
examples of binarizing the et images by adaptive thresholding are displayed in figure [ fig.raw_data_slices ] . in this section
we briefly summarize the main ideas of a multi - scale approach to the segmentation of 3d images , which has been developed recently in @xcite .
the crucial step is to find an efficient representation of the binarized and morphologically smoothed images by unions of overlapping spheres .
let @xmath6 denote the zno phase of the binarized images . since the morphology of the set @xmath6 is rather complex ( see figure [ fig.raw_data_slices ] ) , it is difficult to describe this morphology directly , just by a single stochastic model .
we therefore developed a multi - scale approach to represent the zno phase by different structural components .
each of them will be described separately by suitably chosen stochastic models .
more precisely , we distinguish between a macro - scale component of the binarized et images , which is obtained by morphological smoothing , and several micro - scale components , which consist of those voxels that have been misspecified by the morphological smoothing ; see figure [ fig : micromacroscale ] .
the intention of morphological smoothing is to reduce the structural complexity of the binarized et images , that is , to omit very fine structural components such as `` thin zno branches , '' that is , thin zno parts connected to larger zno domains , `` isolated zno particles , '' that is , small zno particles in the polymer domains , and `` polymeric holes , '' that is , small polymeric particles inside the zno domains .
the morphological transformations which we use for smoothing the zno phase @xmath6 of the original binarized et images are twofold : `` dilation '' and `` erosion . ''
the morphologically smoothed version of the set @xmath6 will thus be denoted by @xmath7 . in the next step a stochastic algorithm
[ see @xcite ] is used to efficiently represent the set @xmath7 by a union of spheres , which is denoted by @xmath8 .
this leads to an enormous data reduction .
another advantage of this representation of the set @xmath7 by unions of spheres is that it allows the interpretation of the morphologically smoothed zno phase as a realization of a marked point process where the midpoints of the spheres are the points and the corresponding radii the marks .
recall that in the macro - scale component @xmath7 of the binarized et images as well as in its representation @xmath8 by unions of spheres , some structural details of the original zno phase @xmath6 , like isolated particles , thin branches and polymeric holes , are omitted .
furthermore , the boundaries of zno domains are morphologically smoothed and slightly enlarged by dilation .
hence , when comparing the sets @xmath6 and @xmath8 , we observe that some voxels are misspecified , that is , indicated as zno although originally being polymer , and vice versa .
the set @xmath9 of misspecified voxels is subdivided into several subcomponents , where each of these subcomponents will be modeled separately .
first , two main types of misspecifications are distinguished : outer misspecifications and inner misspecifications ; see figure [ fig.misspecification ] . each zno voxel that is not covered by a sphere , that is , belonging to the set @xmath10 and therefore constituted as polymer , is called an outer misspecification
typically , thin branches and isolated zno particles are outer misspecifications . on the other hand ,
each polymer voxel which is covered by a sphere , that is , belonging to @xmath11 and constituted as zno , is called an inner misspecification .
inner misspecifications are further subdivided into boundary misspecifications and interior misspecifications .
on the one hand , polymer voxels ( belonging to @xmath12 ) , located near the boundary @xmath13 of the macro - scale component @xmath7 and covered by a sphere , are called boundary misspecifications . on the other hand , each inner misspecification which is not a boundary misspecification
is called an interior misspecification .
typically , polymeric holes belong to interior misspecifications .
we now present our approach to stochastic modeling of the 3d nanomorphology of the zno phase in photoactive layers with three different thicknesses which are given by the binarized et images described in section [ sub.tom.gra ] .
note that the model type is the same for all three thicknesses , just the fitted values of some model parameters are different .
this means , in particular , that our model can be used for computer - based scenario analyses with the general objective of developing improved materials and technologies for polymer solar cells . in accordance with the multi - scale representation of the zno phase which has been described in section [ sec.multi-scale ]
, we will establish stochastic simulation models separately for the morphologically smoothed zno domains represented by unions of overlapping spheres , that is , the macro - scale representation of the zno phase , and for the three types of misspecifications , that is , the micro - scale components . to begin with ,
we develop a point - process model which describes the macro - scale component of the zno phase represented by unions of spheres .
this model is constructed in several steps .
first we consider 2d point processes for those midpoints of spheres which belong to single slices of voxels , being parallel to the @xmath1@xmath2-plane .
since a strong correlation of midpoints spheres in @xmath0-direction is observed , we propose a multi - layer approach considering sequences of correlated 2d point processes to model the 3d point patterns of midpoints.=-1 the members of these sequences belong to a suitably chosen class of planar poisson cluster processes .
in particular , elliptical matrn cluster processes are considered , where the cluster points are scattered in ellipses of uniformly distributed orientation around their cluster centers . to model the 3d point patterns of midpoints ,
a markov chain with stationary initial distribution is constructed , which consists of highly correlated matrn cluster processes .
it can be seen as a stationary point process in 3d , where the radii of spheres are considered as marks of this point process .
the mark correlation functions , which have been computed for the radii of spheres extracted from et images , show strong positive correlations for small distances between midpoints .
hence , for a given configuration of midpoints , the radii associated with these midpoints are not modeled just by independent marking , but a certain moving - average procedure is proposed .
note that our model of a stationary marked point process in 3d describing the macro - scale component of the zno phase is not isotropic .
this is in accordance with the nanomorphology observed in the et images ; see @xcite .
to get an idea which class of point processes is suitable to model the midpoints belonging to the individual slices of voxels , we consider the pair correlation function @xmath14 of stationary and isotropic point processes in @xmath15 .
note that @xmath16 is proportional to the relative frequency of point pairs with distance @xmath17 from each other ; see , for example , @xcite .
then , for each of the three photoactive layers with thicknesses of 57 nm , 100 nm , and 167 nm , the values @xmath18 of the pair correlation function have been estimated for distances @xmath19 within some interval @xmath20 ; see figure [ fig.cor.fun ] .
nm , @xmath21 nm , @xmath22 nm ) ; bottom : estimated pair correlation functions ( @xmath23 nm , @xmath24 nm , @xmath25 nm ) . ] since @xmath26 for small @xmath17 , these estimates clearly indicate clustering of points , which can also directly be seen from the point patterns shown in figure [ fig.cor.fun ] .
the clusters appearing in these point patterns seem to be located in relatively small ( bounded ) areas , which corresponds to the fact that @xmath27 for sufficiently large @xmath17 .
the shapes of the clusters are not circular , but rather elongated .
hence , we propose to consider matrn cluster processes , where the cluster points are scattered in ellipses of uniformly distributed orientation around their cluster centers .
, @xmath28 , @xmath29 ) . ]
this class of ( elliptical ) matrn cluster processes in @xmath15 can be described by a vector of four parameters : @xmath30 , where @xmath31 is the intensity of the stationary poisson point process @xmath32 of cluster centers , @xmath33 and @xmath34 with @xmath35 are the semi - axes of ( random ) ellipses @xmath36 centered at the points @xmath37 of the poisson process @xmath38 of cluster centers and rotated around @xmath37 by random angles @xmath39 which are independent and uniformly distributed on the interval @xmath40 , and @xmath41 is the intensity of the stationary poisson processes @xmath42 of cluster members which are released by the cluster centers @xmath37 within the ellipses @xmath43 .
the matrn cluster process is then defined as the random point pattern @xmath44 given by @xmath45 , where the sequences @xmath46 are assumed to be independent .
there are strong similarities between consecutive 2d slices in terms of a high correlation of midpoint locations in @xmath0-direction as well as approximately equal numbers of points .
figure [ fig.successivemidpoints ] shows a series of such consecutive 2d slices from the 57 nm data set . as a consequence , it is not suitable to model the stacks of these 2d point patterns by sequences of independent matrn processes .
but the ( vertical ) correlation structure visualized in figure [ fig.successivemidpoints ] can be taken into account by considering a markov chain of matrn processes .
this allows us to model small displacements of clusters when passing from slice to slice .
furthermore , `` births '' and `` deaths '' of clusters in @xmath0-direction can also be modeled in this way . in other words ,
we consider a certain class of spatial birth - and - death processes with random displacement of points ; see , for example , @xcite .
for each integer @xmath47 , let @xmath48 be a stationary poisson point process in @xmath15 with intensity @xmath49 such that @xmath50 , and let @xmath51 be an independent and identically distributed ( i.i.d . ) sequence of bernoulli random variables , which is independent of @xmath52 , where @xmath53 for some @xmath54 .
note that @xmath52 and @xmath55 will be used in order to model `` births '' and `` deaths '' of cluster centers , respectively . for each integer @xmath47 , an i.i.d .
sequence @xmath56 of random displacement vectors @xmath57 with values in @xmath15 is considered , which is independent of @xmath52 and @xmath58 .
we assume that the random vectors @xmath57 are uniformly distributed in the set @xmath59 , where @xmath60 and @xmath61 denote the size of minimum and maximum displacement , respectively ; @xmath62 . then ,
a ( stationary ) markov chain @xmath63 of matrn processes can be constructed as follows . for @xmath64 ,
let @xmath65 be an elliptical matrn cluster process as introduced in section [ sec.mod.sli ] , that is , @xmath66 .
assume that the `` birth rate '' @xmath49 and the `` survival probability '' @xmath67 satisfy @xmath68 , where @xmath31 is the intensity of the poisson process @xmath38 of cluster centers .
for @xmath69 , the poisson process @xmath70 of cluster centers is then given by @xmath71 .
the poisson process @xmath72 is constructed in the same way as @xmath70 , that is , @xmath73 and so on ; see also figure [ fig.3dstack ] .
the matrn processes @xmath74 are built similarly to the construction of @xmath75 for example , @xmath76 is given by @xmath77 where the sequences @xmath78 are defined in the same way as @xmath79 introduced in section [ sec.mod.sli ] .
the markov chain @xmath63 of matrn processes introduced above can be seen as a stationary point process in 3d .
it possesses seven ( free ) parameters : @xmath80 describing its initial distribution , and @xmath81 describing the transitions from step to step , whereas the `` birth intensity '' @xmath49 of ( new ) cluster centers is given by @xmath82 .
it turned out that suitable choices for @xmath83 are the values of @xmath84 and @xmath85 .
this means that the uniform distribution of the displacement vectors @xmath57 is implemented as ( discrete ) uniform distribution on the @xmath86-neighborhood in the considered slice of voxels .
techniques for fitting the remaining five parameters of the markov chain @xmath63 are discussed in section [ sub.mod.mod ] .
to get an idea which class of mark distributions is suitable to model the radii of spheres , we computed histograms of radii which have been extracted from the et images for each of the three photoactive layers with thicknesses of 57 nm , 100 nm and 167 nm . recall that in the sphere - putting algorithm mentioned in section [ sec.multi-scale ] we only consider spheres with a minimum radius of @xmath87 voxel sizes
. hence , instead of computing histograms for the original radii , say , @xmath88 we computed histograms for correspondingly reduced radii @xmath89 where @xmath90 .
it turns out that for all three film thicknesses , gamma distributions can be fitted quite nicely to the histograms of reduced radii ; see figure [ fig.gamma_distribution ] .
the parameters @xmath91 and @xmath92 of these gamma distributions @xmath93 have been estimated using the method of moments ; see table [ tab.gamma_distribution ] .
nm , @xmath94 nm , @xmath95 nm . ] .
parameters for gamma distributions of radii [ cols="<,^,^,^",options="header " , ] in the next step , the connectivity of the zno phase is considered .
this also is an important characteristic , because only if there is a high connectivity , that is , if many percolation pathways exist , the produced charges can be transported to the electrodes , where current can be gripped .
for estimating the connected and monotonously connected volume fractions of zno we applied the same methods as in @xcite .
however , in general , the ( conditional ) connected and monotonously connected volume fractions of the foreground phase in globally thresholded images do not monotonously depend on the values of global thresholds .
but as shown in figure [ fig.connectivity ] , this is the case for the globally thresholded et data of photoactive layers of polymer - zno solar cells .
hence , also with respect to connected volume fractions , the values for the two extreme thresholds can be seen as upper and lower bounds . with the exception of the 57 nm film
, the values for the connected volume fractions computed from simulated data , given in table [ tab.check_morphological ] , are nicely between the corresponding values obtained from globally thresholded et images .
also , the relative errors between the values obtained for the adaptively thresholded et images and simulated images are rather small ; see table [ tab.rel.err.vol ] .
nm , @xmath94 nm , @xmath95 nm . ] @lcd2.3d2.3d2.3@ & & & & + & & & & + & & & & + 57 nm & adaptive threshold & 0.133 & 0.963 & 0.928 + & simulated data & 0.112 & 0.905 & 0.864 + & relative error & -0.158 & -0.060 & -0.069 + 100 nm & adaptive threshold & 0.211 & 0.980 & 0.888 + & simulated data & 0.215 & 0.971 & 0.910 + & relative error & 0.019 & -0.009 & 0.025 + 167 nm & adaptive threshold & 0.209 & 0.970 & 0.851 + & simulated data & 0.210 & 0.943 & 0.806 + & relative error & 0.005 & -0.028 & -0.053 + and
@xmath96 obtained from globally thresholded et images are plotted as dashed lines , the corresponding results from simulated image data as solid lines ; @xmath97 nm , @xmath94 nm , @xmath95 nm . ] finally , the spherical contact distribution function @xmath98 $ ] of the zno phase is considered , where @xmath99 can be interpreted as a conditional probability that the minimum distance from a randomly chosen location to the zno phase is smaller or equal than @xmath100 , provided that the considered location belongs to the polymer phase . similarly to the situation which we observed for the connected volume fractions considered above , it turns out that the spherical contact distribution functions of the foreground phase in globally thresholded et images depend monotonously on the value of the global threshold .
hence , the estimated contact distribution functions @xmath101 and @xmath96 obtained for the two extreme thresholds can be seen as upper and lower bounds , respectively ; see figure [ fig.spherical ] .
in addition to this , the spherical contact distribution @xmath102 obtained from simulated data is shown in figure [ fig.spherical ] , where @xmath103 for all considered @xmath100 and for all three layer thicknesses . in summary
, we can conclude that our model fits very well to real data regarding the considered structural characteristics . as the model has been developed for analyzing the influence of morphology on the performance of solar cells , we also consider a physical characteristic for model validation which is described in the following section .
quenching efficiency @xmath104 is the probability of a random exciton being quenched ; see section [ sub.pho.act ] .
it is an elementary but important physical characteristic for the efficiency of solar cells . in a hybrid polymer - zno solar cell ,
absorption of light by the polymer phase does not directly yield free charge carriers . instead excitons are formed .
it is only at the interface of the polymer and zno phase that free charges are generated by quenching ( splitting ) of excitons .
it is therefore of the utmost importance that excitons are able to reach this interface .
the exciton diffusion length in conjugated polymers is typically a few nanometers , which puts a considerable constraint on the morphology of polymer solar cells . in other words ,
the efficiency of exciton quenching is very sensitive to morphology , making it a suitable way to validate our model .
suppose that the polymer phase @xmath105 is given in a cubic sampling window @xmath106 .
then , the overall efficiency @xmath104 of exciton quenching can be obtained from the field @xmath107 of local exciton densities in the polymer phase .
the exciton density field @xmath107 can be computed by solving the steady - state diffusion equation [ see @xcite ] @xmath108 where @xmath109 is the diffusion constant , @xmath110 is the exciton life time , and @xmath111 is the rate of exciton generation . as a boundary condition
we require that all excitons at the polymer - zno interface be quenched , that is , @xmath112 for all @xmath113 . for @xmath114
, cyclic boundary conditions are applied in all directions .
the exciton lifetime and exciton diffusion rate in p3ht are taken from the literature : @xmath115 ps and @xmath116 m@xmath117 s@xmath118 ; see @xcite .
the rate of exciton generation @xmath111 is just a scaling factor , where we use a value of @xmath119 m@xmath120 s@xmath118 which is typical for 1-sun conditions .
normalized to @xmath121 for adaptively thresholded et ( left ) and simulated ( right ) data .
the scalebar , specifying the density of excitons , applies to both images . ]
the diffusion equation is solved numerically to a relative error of less than 10@xmath120 .
figure [ fig.quenchingfields ] shows local exciton density fields @xmath107 for adaptively thresholded and simulated data , respectively .
once @xmath107 is known , the quenching efficiency @xmath104 follows from @xmath122 , where @xmath123 is the average exciton density in the polymer domain @xmath12 .
figure [ fig.quenching ] compares the quenching efficiencies for original and simulated data .
the quenching efficiency is also monotonously depending on the global threshold .
hence , the values of @xmath104 obtained for the two extreme thresholds can be seen as lower and upper bounds .
the values of the quenching efficiency for the simulated data lie well within these lower and upper bounds ; see figure [ fig.quenching ] .
the small relative errors displayed in table [ tab.quenching ] show that our model fits very well to real data , where quenching efficiencies of adaptively thresholded et images are compared with those of simulated data .
@ld2.3cc@ & & * 100 nm film * & * 167 nm film * + adaptive threshold & 0.418 & 0.794 & 0.819 + simulated data & 0.416 & 0.805 & 0.834 + relative error & -0.010 & 0.014 & 0.018 +
in the present paper we developed a parameterized stochastic simulation model for the nanostructure of photoactive layers of hybrid polymer - zno solar cells . the model is based on tools from stochastic geometry .
additional to the model itself , we developed a method to fit its parameters to real 3d et image data . to establish our model ,
the adaptively thresholded et images are segmented using a stochastic algorithm which consists of two main steps .
first , the images are morphologically smoothed in order to slightly decrease their structural complexity .
then , the morphologically smoothed binary images are represented by a system of overlapping spheres , which can be interpreted as a realization of a 3d marked point process , where the sphere centers are the locations of points and the corresponding radii are their marks . for the stochastic simulation model ,
we use a correlated vector of 2d elliptical matrn cluster processes , where the points are subsequently marked to create a 3d marked point process . to complete the model , that is , to include the structural details which were omitted due to the morphological smoothing , a stochastic simulation model for this `` micro - scale '' component is developed afterward .
as our stochastic simulation model is fully parameterized , we also developed techniques for the estimation of the model parameters of all model components .
thus , we are able to fit the simulation model to the et image data described in section [ sub.tom.gra ] . finally , we validated the simulation model by comparing structural and physical characteristics computed from simulated image data with the corresponding characteristics obtained from globally and adaptively thresholded et images , respectively .
in particular , the quenching efficiencies computed for realizations of the simulation model agree very well with those of the et images .
hence , our model nicely reflects the diffusion of excitons .
since we were able to fit our model to 3d et data , it has already proved its capability to represent realistic nanostructures of photoactive layers of hybrid polymer - zno solar cells .
the fact that the model is parameter - based enables us to predict morphologies for film thicknesses , which have not ( yet ) been imaged by 3d et , by interpolating or extrapolating the fitted model parameters . due to a strong correlation between morphology and efficiency of polymer solar cells
[ see @xcite ] , the developed simulation model is of significant importance for further investigations of polymer solar cells . in a forthcoming paper
we will also investigate the transport processes of charges to the electrodes as described in @xcite , additional to the structural and physical characteristics considered in the present paper . by generating virtual morphologies , which are generated as realizations of the developed model with different parameter configurations , and investigating the transport processes of electrons and excitons therein
, the spatial stochastic model will be used to identify morphologies of improved efficiency with respect to the considered physical characteristics .
we also remark that the modeling approach developed in the present paper can be applied to various other kinds of image data , including geographical data considered , for example , in ecology .
then , for instance , the markov chain of matrn cluster processes introduced in section [ sec.spa.bir ] may be viewed as a sequence of dependent 2d point processes through time , which can model , for example , the temporal movements of species in a given region .
the authors would like to thank image analysis & stereology for permission to reprint figures [ fig.schematicview][fig:micromacroscale ] taken from @xcite .
part of this work was done while volker schmidt was visiting the isaac newton institute for mathematical sciences , cambridge .
|
a spatial stochastic model is developed which describes the 3d nanomorphology of composite materials , being blends of two different ( organic and inorganic ) solid phases .
such materials are used , for example , in photoactive layers of hybrid polymer zinc oxide solar cells .
the model is based on ideas from stochastic geometry and spatial statistics .
its parameters are fitted to image data gained by electron tomography ( et ) , where adaptive thresholding and stochastic segmentation have been used to represent morphological features of the considered et data by unions of overlapping spheres .
their midpoints are modeled by a stack of 2d point processes with a suitably chosen correlation structure , whereas a moving - average procedure is used to add the radii of spheres .
the model is validated by comparing physically relevant characteristics of real and simulated data , like the efficiency of exciton quenching , which is important for the generation of charges and their transport toward the electrodes . , , , , , , , , and .
|
the classification of three and four manifolds is one of the open questions which has been addressed by both mathematicians and physicists . in particular
, physicists have shown that a class of quantum field theories called topological field theories provides an elegant approach to solve these problems .
the main idea in any quantum field theory is to represent the theory by an action @xmath0 which gives information about the particle content and their interactions .
the interaction strengths are given by coupling constants .
for capturing the topological features of knots or links as shown in fig .
[ aba : fig1 ] , we need a theory which does not change if we alter the shape or size of these knots or links .
one such theory is the chern - simons gauge theory where the action @xmath0 is explicitly metric independent .
hence , chern - simons field theory provides a natural framework to study knots , links and three manifolds .
the action @xmath0 defining the chern - simons theory on a three manifold @xmath1 based on a gauge group @xmath2 is @xmath3 where @xmath4 is the coupling constant and @xmath5 s are the gauge fields or connections matrix - valued in group @xmath2 . the knots or links - for example ,
the trefoil knot @xmath6 carrying representation @xmath7 of the gauge group @xmath2 are described by the expectation value of wilson loop operators @xmath8 $ ] : @xmath9=\langle w_r(c ) \rangle~=~ \frac { \int_m [ { \cal d}a ] ~ w_r(c)~ exp(i s ) } { { \cal z}[m]}~,\ ] ] where @xmath10=\int_m [ { \cal d } a]~ exp(is)\ ] ] is the partition function and @xmath11 $ ] are the knot invariants
. witten s pioneering work @xcite established a three - dimensional definition for knots and links . in particular , jones and homflypt polynomials and their recursion relations were obtained from chern - simons gauge theory based on @xmath12 and @xmath13 gauge groups .
we can relate coupling constant @xmath4 and the rank @xmath14 to the polynomial variables of homflypt polynomials .
similarly , the jones polynomial variable @xmath15 will be related as @xmath16 .
the two main ingredients which go into the evaluation of the polynomial invariants @xmath11 $ ] of knots and links are 1 .
connection between chern - simons theory on the three - dimensional ball to the two - dimensional wess - zumino conformal field theory on the boundary of the three - ball .
2 . using alexander s theorem , any knot or link can be obtained as a closure of braid . in fig . [
abb : fig2](a ) , we illustrate alexander s theorem by re - drawing the trefoil knot as a closure of two - strand braid with three crossings . also , we have diagrammatically shown in fig .
[ abb : fig2](b ) that these knots or links in @xmath17 can be viewed as gluing two three - balls with oppositely oriented @xmath18 boundaries . in this particular trefoil knot example , the @xmath18 boundary has four - punctures .
the connection between chern - simons theory and wess - zumino conformal field theory states that the chern - simons functional integral over a three - ball with a four - punctured @xmath18 boundary corresponds to state @xmath19 which represents four - point correlator conformal block in the wess - zumino conformal field theory .
the suffix @xmath20 on the state is to indicate that the middle two - strands are braided thrice .
in fact , the punctures get exchanged whenever the middle two - strands get braided .
we can denote the no - crossing four - punctured boundary state as @xmath21 and apply a braiding operator @xmath22 thrice to get the state @xmath23 : @xmath24 similarly , the state for the oppositely oriented boundary will be in the dual space . for the above example in fig .
2(b ) , the state is @xmath25 .
the knot invariant is @xmath9= \langle \psi_0 |\psi_3\rangle=
\langle \psi_0 |{\cal b}^3|\psi_0\rangle~. \label { knot1}\ ] ] in order to see the polynomial form , we need to expand the state @xmath26 in an eigenbasis of the braiding operator @xmath22 . in general for the four - punctured @xmath18 boundary , the braiding can be either on the side two - strands or on the middle two strands . for clarity
, we will take the gauge group @xmath27 .
for the four - punctured @xmath18 boundary , we can chose eigenbasis @xmath28 if the braiding is in the side two - strands .
that is , @xmath29 where the suffix @xmath30 on the braiding operator denotes the braiding between first and the second strands and the eigenvalue is @xmath31 .
the basis is shown diagrammatically in fig .
3(a ) where @xmath7 denotes the @xmath12 representation placed on the strands . from the picture ,
the representation @xmath32 will be an element in the tensor product @xmath33 .
similarly , for braiding middle two strands , we choose the basis @xmath34 as shown in fig .
3(b ) where @xmath35 .
clearly , these two basis states must be related by a duality matrix : @xmath36 |\phi_s\rangle~.\ ] ] when the four strands carry the same representation , we can write in shorthand notation the duality matrix as @xmath37 \equiv a_{ts}~.\ ] ] these duality matrices turns out to be proportional to quantum racah coefficients .
the explicit form of @xmath38 racah coefficients and their identities satisfied are available in ref . .
see also papers@xcite . in this example , the braiding involves middle two strands .
hence the state @xmath21 can be expanded in the middle - strand basis @xmath39 : @xmath40 interestingly , the coefficients @xmath41 has to satisfy @xmath42}~=~\sqrt{s_{0t}/s_{00 } } \equiv \sqrt{dim_q t}~,\ ] ] so that two equivalent knots share the same polynomial invariant . here ,
@xmath43 $ ] denotes the polynomial invariant for unknot carrying representation @xmath44 whose form can be written as the ratio of elements of the modular transformation matrix @xmath0 in wess - zumino conformal field theory or in terms of quantum dimensions of the representation @xmath44 of the quantum group as indicated in the above equation .
the knot invariant ( [ knot1 ] ) will be @xmath9~=~ \langle \psi_0 \vert { \cal b}^3 \vert \psi_0 \rangle~=~ \sum_t dim_q t ( \lambda_t ( r , r))^3 ~.\ ] ] so far , we have not introduced orientation on the strands . in general , the braiding eigenvalue depends on the * framing * and also on the relative orientation on the two braiding strands .
two conventional framing are standard framing and blackboard framing .
standard framing is one where the self - linking number of the knot with its frame is zero .
this is useful to obtain ambient isotopy invariants .
the self - linking number matches the crossing number in the blackboard framing .
hence the braiding eigenvalue in the blackboard framing is useful to obtain regular isotopy invariants . in the trefoil example
, we could place parallel orientation in the middle two - strands .
as the crossing sign due to braiding is positive , we called such a braiding as right - handed braiding .
similarly , an inverse braiding leading to mirror of trefoil ( @xmath45 ) will be called left - handed braiding . for parallely oriented strands ,
the right - handed braiding eigenvalue in standard framing is @xmath46 where @xmath47 denotes the quadratic casimirs in the respective @xmath48 and adjoint representation .
@xmath49 will be @xmath50 depending on the representation @xmath44 appears symmetrically or antisymmetrically in the tensor product @xmath51 .
similarly , the left - handed braiding eigenvalue for antiparallely oriented strands is @xmath52 now using the appropriate braiding eigenvalues , the knot invariants can be written as polynomials in the variable @xmath15 .
the method for a four - punctured @xmath18 boundary is generalisable for @xmath53 such four - punctured @xmath18 boundaries as shown in fig .
[ abd : fig4 ] .
we will see in the next section that this building block will be useful to redraw knots like knot @xmath54 and knot @xmath55 as gluing of three - balls with one or more four - punctured @xmath18 boundaries .
the basis state for such a @xmath53-@xmath18 boundaries is @xmath56 where @xmath57 for any @xmath58 . sometimes , it is useful to keep @xmath18 boundaries with more than four - puctures
. then the basis state for a @xmath18 boundary with @xmath59 puctures will be a @xmath59-point conformal block . for braiding @xmath60 s and @xmath61 s , we choose the basis shown in fig .
[ abe : fig5](a ) and ( b ) respectively .
the procedure we elaborated for the trefoil obtained from gluing two three - balls each with a four - punctured @xmath18 boundary is generalisable for any knot . that is , using the building blocks in fig .
[ abd : fig4 ] or three - balls with @xmath59-punctured @xmath18 boundaries , it is not difficult to see that any knot can be obtained either from gluing two three - balls each with a @xmath59-punctured @xmath18 boundaries or from gluing many three - balls with one or more @xmath53 four - puctured @xmath18 boundaries . the method presented here enables direct evaluation of any knot / link polynomial directly without going through the recursive procedure . though we have concentrated on the gauge group @xmath12 , it is straightforward to generalise for any compact semi - simple gauge group@xcite .
the representation @xmath7 must be replaced by conjugate representation @xmath62 ( @xmath63 for @xmath12 ) depending on the oriented strand is outgoing from or incoming to a @xmath18 boundary as shown in fig .
[ abd : fig4 ] .
thus , we can place any representation @xmath7 of any compact semi - simple gauge group on the knot and obtain if we place the defining representation on the strands , we recover some of the well - known polynomials as tabulated below : [ cols="<,<",options="header " , ] we know that these well - known polynomials do not solve the classification problem .
apart from these special cases , chern - simons field theory gives a huge pool of generalised polynomials depending on placing representation @xmath7 of any gauge group other than the defining representation on the strands .
we believe that at least one of these generalised chern - simons invariant will be able to distinguish two inequivalent knots which are not distinguished by the well - known polynomials .
knot theory literature gives a list of chiral knots and mutant knots which are not distinguished by jones , homflypt and kauffman .
we tried to check the ability of generalised knot invariants , from chern - simons field theory , to detect * chirality and mutations*. we address the chirality detection in the following section .
upto 10 crossings , there are two knots : knot @xmath54 and knot @xmath55 which are chiral but their chirality is not detected by the well - known polynomials . in fig .
[ abf : fig6 ] , we have drawn * knot @xmath54 * in two equivalent ways .
clearly , the knot @xmath54 can be obtained as gluing of five building blocks as shown in fig .
[ abg : fig7 ] . for @xmath27 , @xmath64 = ( 2,.7 ) ( 0,0)(1,0)1.8
( 0,.3)(1,0)1.8 ( 0,0)(.3,0)7(0,1).3 ( .7,.5)(-1,0).7 ( 1.1,.5)(1,0).7 ( .8,.5)@xmath59 ( spin @xmath65 representation ) placed on knot , the states for the building blocks can be written down following the methods presented in the previous section and also using the properties of the duality matrix@xcite
. the states for these five building blocks are @xcite @xmath66}(-1)^{3(n - l_1)}q^{-3/2[n(n+2)-l_1(l_1 + 1)]}\vert { \phi_{l_1}^{(1)}}\rangle\\ \nu _ 1(p_4)&=&\sum _ { l_5=0}^n(-1)^{n - l_5}q^{-1/2[n(n+2)-l_5(l_5 + 1)]}\vert { \phi _ { l_5}^{(1)}}\rangle\\ \nu _ 2(p_1;p_2)&=&\sum _ { i_1,j_1,l_2,r=0}^n { a_{l_1r}a_{j_1r}a_{l_2r}\sqrt { [ 2l_2 + 1]}\over \sqrt { [ 2r+1 ] } } \times q^{n(n+2)-l_2(l_2 + 1)}\vert { \phi _ { i_1}^{(1)}}\rangle \vert{\phi _ { j_1}^{(2 ) } } \rangle\\ \nu _ 2(p_2;p_3)&=&\sum _ { l_3=0}^nq^{l_3(l_3 + 1)}\vert { \phi _ { l_3}^{(1 ) } } \rangle \vert { \phi _ { l_3}^{(2)}}\rangle\\ \nu _ 2(p_3;p_4)&=&\sum _ { i_2,j_2,l_4=0}^n(-1)^{l_4}q^{-l_4(l_4 + 1)/2}a_{l_4i_2}a_{l_4j_2}\vert { \phi _ { i_2}^{(1)}}\rangle \vert{\phi _ { j_2}^{(2)}}\rangle~.\end{aligned}\ ] ] here @xmath67 s denote the @xmath18 boundaries as indicated in fig .
abg : fig7 . using the above states ,
the knot invariant is @xmath68 & = & ( -1)^n q^{{-3\over 2}[n(n+2)]}\sum _ { r , l_1,l_2,j_1,j_2=0}\sqrt { [ 2l_1 + 1 ] } \times\\ ~&~&\sqrt { [ 2l_2 + 1]}\sqrt { [ 2j_2 + 1 ] } a_{l_1r}a_{l_2r } a_{j_1r}a_{j_1j_2 } \times \\ ~&~&(-1)^{l_1}q^{{3\over 2}[l_1(l_1 + 1 ) ] } q^{{3\over 2 } [ j_1(j_1 + 1)]}q^{-l_2(l_2 + 1)}q^{j_2(j_2 + 1)}\end{aligned}\ ] ] we checked the general result for the special cases .
that is , @xmath69 gives jones polynomial and @xmath70 gives akutsu - wadati / kauffman polynomial@xcite .
interestingly , for @xmath71 , the polynomial is @xmath72&= & q^{45/2}-q^{41/2}-q^{39/2}+q^{35/2}+q^{23/2}+q^{21/2}-q^{19/2}\\ ~&~&- q^{17/2}+q^{13/2}-q^{9/2}+q^{5/2}+q^{3/2}+ q^{-3/2}+ q^{-5/2}\\ ~&~&-q^{-13/2}-q^{-15/2}+q^{-21/2}+2q^{-23/2}-q^{-27/2}+2q^{-31/2}\\ ~&~&-3q^{-35/2}-q^{-37/2}+q^{-39/2}+q^{-41/2}~.\end{aligned}\ ] ] clearly , @xmath73 ( q ) \neq v_3[9_{42 } ] ( q^{-1})$ ] indicating that @xmath12 chern - simons spin @xmath74 ( @xmath71 in representation @xmath64 ) knot polynomial is powerful to detect chirality .
similar exercise was performed for knot @xmath55 by gluing the four building blocks as shown in fig .
[ abh : fig8 ] .
the knot invariant @xcite is @xmath75&= & ( -1)^n q^{{n(n+2)\over 2}}\sum _ { i , r , s , u , m=0 } \sqrt { { [ 2r+1][2s+1][2u+1]\over [ 2m+1]}}a_{im}\\ ~&~&a_{ms}a_{rm}a_{iu}(-1)^s~q^{-i(i+1)}q^{m(m+1)}q^{-r(r+1)}q^{u(u+1 ) } q^{{3\over 2}s(s+1)}~.\end{aligned}\ ] ] for @xmath71 , we have checked that @xmath76(q ) \neq v_3[10_{71}](q^{-1})$ ] confirming the ability of generalised chern - simons invariant in detecting chirality . with such positive results for chiral knots
, we attempted to check whether generalised chern - simons invariant is capable of detecting mutation operation which we shall present in the following section .
remove a two - tangle region from any knot and do a rotation by @xmath77 about any of the three perpendicular axis .
replace the rotated two - tangle back with suitable reversal of orientation of the strands to give another knot .
this operation is called mutation and the two knots are said to be mutants . in fig .
[ abi : fig9 ] , we have indicated the three mutation operation @xmath78 s .
an example for the mutant knots is the well - known eleven - crossing kinoshita - terasaka and conway knots .
we can formally represent mutant knots shown in fig .
[ abj : fig10 ] as gluing two three - balls with the two - tangle rooms as shown in fig .
[ abk : fig11 ] . that is , gluing fig .
[ abk : fig11](d ) with any of the fig .
[ abk : fig11](a),(b ) and ( c ) gives mutant knots .
interestingly , the states ( a ) , ( b ) , ( c ) in fig .
[ abk : fig11 ] can be obtained by gluing fig .
[ abk : fig11](a ) with the two - boundary states ( a ) , ( b ) and ( c ) in fig .
[ abl : fig12 ] respectively .
clearly , @xmath79 and @xmath80 represent mutation @xmath81 and @xmath82 respectively in fig .
[ abi : fig9 ] respectively . as braid words ,
[ abl : fig12 ] ( a ) , ( b ) , ( c ) are different but the states @xmath83 s are same : @xmath84 where we have used @xmath85 and the operator @xmath86 interchanges the representations on the first and second , the third and fourth punctures in that basis .
similarly , we can show @xmath87 so , the * generalised invariants * of the mutants @xmath88 obtained by gluing fig . [ abk : fig11](a),(b ) and ( c ) with ( d ) are * same*. it appears that the identities of the states for * four - punctured boundaries * play a crucial role in making the states representing the mutation operation @xmath89 to be similar to identity braid @xmath90 . in order to go beyond four - punctured boundary state , we studied composite braiding @xcite . using the representation theory of composite braids , we showed that the composite invariant for knots are sum of the generalised knot invariants .
this implies composite invariants can not detect mutations in knots .
however , some mutant links can be distinguished by composite invariants .
in this article , we have briefly presented the direct evaluation of generalised invariants of knots and links from chern - simons field theory . for @xmath12 gauge group in chern - simons theory
, we get the the colored jones polynomials for the knots and links carrying higher dimensional @xmath12 representations .
there is a huge pool of generalised chern - simons invariants for other gauge groups like @xmath13 , @xmath91 etc with knots and links carrying arbitrary representations .
the hope is that at least one of the invariants will be able to distinguish two inequivalent knots .
we showed that the chiral knots upto 10 crossings are distinguished by the chern - simons field theory invariants .
however , we gave a proof that the process of mutation can not be detected within the chern - simons field theoretic framework .
the proof is for knots and links carrying arbitrary representation of any compact semi - simple gauge group .
discussions during the knot theory conference revealed that there are approaches from quantum groups@xcite and floer homology@xcite whose invariants does distinguish 11-crossing kinoshita - terasaka and conway mutant knot .
it will be interesting to to check whether this method distinguishes other mutant knots .
9 e. witten , commun .
* 121 * ( 1989 ) 351 .
kirillov , n.yu .
reshetikhin , _ representation algebra @xmath92 , q - orthogonal polynomials and invariants of links _ ,
lomi preprint e-9 - 88 ; see also in : new developments in the theory of knots .
t. kohno , world scientific ( singapore , 1989 ) .
kaul and t.r .
govindarajan , nucl . phys . *
b380 * ( 1992 ) 293 .
kaul and t.r .
govindarajan , nucl . phys .
* b393 * ( 1993 ) 392 .
kaul , commun .
phys.*162 * ( 1994 ) 289 .
p. ramadevi , t.r .
govindarajan and r.k .
kaul , nucl . phys .
* b402 * ( 1993 ) 548 p. ramadevi , t.r .
govindarajan and r.k .
kaul , mod .
lett * a9 * ( 1994 ) 3205 .
m.wadati , t.deguchi and y.akutsu , phys . rep . * 180 * ( 1989 ) 247 and references therein .
p. ramadevi , t.r .
govindarajan and r.k .
kaul , mod .
lett * a10 * ( 1995 ) 1635 .
h.r.morton and h.j.ryder , geom . topol . monogr.*1 * ( 1998)365 .
john a. baldwin , w.d.gilam , _ computations of heegard - floer knot homology _ , arxiv : math/0610167 .
|
in this brief presentation , we would like to present our attempts of detecting chirality and mutations from chern - simons gauge theory .
the results show that the generalised knot invariants , obtained from chern - simons gauge theory , are more powerful than jones , homflypt and kauffman polynomials .
however the classification problem of knots and links is still an open challenging problem .
|
f. borceux and d. bourn @xcite , call a category @xmath2 _ homological _ if it is pointed , regular , and protomodular ( in the sense of bourn @xcite ) ; in fact they claim that such categories provide the most convenient setting for non - abelian versions of various _ standard homological lemmas _ , such as snake lemma , 3@xmath33-lemma , etc . taking this viewpoint , one could still try , however , to introduce a more general setting involving a distinguished class @xmath1 of regular epimorphisms , where the homological lemmas are expected to hold only for short exact sequences @xmath4 with @xmath5 in @xmath1 . in particular
, there is no reason to exclude the trivial case , where @xmath2 is an arbitrary category ( say , pointed and with finite limits and finite colimits ) and @xmath1 the class of all isomorphisms in @xmath2 .
this idea goes back to n. yoneda @xcite , whose _ quasi - abelian _ categories can in fact be defined as pairs @xmath0 , where @xmath2 is an additive category in which the short exact sequences @xmath6 with @xmath7 in @xmath1 have the same properties as all short exact sequences in an abelian category .
+ + the purpose of this paper is to present a new notion of relative homological and relative weakly homological categories @xmath0 , such that whenever @xmath2 is a pointed category with finite limits and cokernels / coequalizers , we have : * @xmath8 always is a relative homological category ; * @xmath9 is a relative weakly homological category if and only if @xmath2 is a protomodular category ; * @xmath10 is a relative homological category if and only if it is a relative weakly homological category and if and only if @xmath2 is a homological category ; * @xmath11 is a relative homological category if and only if @xmath2 is a trivial category ; * suitable reformulations of various homological lemmas relative to @xmath1 hold in @xmath2 .
throughout the paper we assume that @xmath2 is a pointed category with finite limits and cokernels , and @xmath1 is a class of morphisms in @xmath2 containing all isomorphisms .
[ def1 ] a pair @xmath0 is said to be a relative homological category , if it satisfies the following conditions : * the class @xmath1 is pullback stable ; * every morphism in @xmath1 is a normal epimorphism ; * @xmath1-short - five lemma holds , i.e. in every commutative diagram of the form @xmath12 \ar[r]^k & a \ar[r]^f \ar[d]_w & b \ar@{=}[d ] \\ k \ar[r]_{k ' } & a ' \ar[r]_{f ' } & b } \ ] ] with @xmath13 , @xmath14 , and with @xmath15 and @xmath16 in @xmath1 , the morphism @xmath17 is an isomorphism ; * the class @xmath1 is closed under composition ; * if @xmath18 and @xmath19 , then @xmath20 ; * if a morphism @xmath15 in @xmath2 factors as @xmath21 , in which @xmath22 and @xmath23 is a monomorphism , then there exists a monomorphism @xmath24 in @xmath2 and a morphism @xmath25 in @xmath1 , such that @xmath26 ; * if in a commutative diagram @xmath27^k \ar[d]_u & a \ar[r]^f \ar[d]^w & b \ar@{=}[d ] \\ k ' \ar[r]_{k ' } & a ' \ar[r]_{f ' } & b}\ ] ] @xmath13 , @xmath14 , and morphisms @xmath15 , @xmath16 , and @xmath28 are in @xmath1 , then @xmath17 also is in @xmath1
. we will also say that @xmath0 is a relative weakly homological category whenever it satisfies conditions ( a)-(e ) .
note that condition [ def1](c ) in fact follows from condition [ def1](g ) . indeed
: under the assumptions of [ def1](c ) , the morphism @xmath29 is in @xmath1 since @xmath1 contains all isomorphisms and condition [ def1](g ) holds . also , it is a well known fact that in this situation @xmath30 . since every morphism in @xmath1 is a normal epimorphism ,
we conclude that @xmath17 is an isomorphism . assuming that condition [ def1](b ) holds
, we can say that the conditions / axioms used here are much weaker than those used by g.janelidze , l. mrki , and w. tholen @xcite .
however , various arguments from @xcite , used there in the proof of the equivalence of the so - called old and new axioms , can be extended to our context to obtain various reformulations of our conditions .
some of them are given in this section .
[ con1 ] * every morphism in @xmath1 is a regular epimorphism ; * if @xmath18 then @xmath31 ; * ( `` relative hofmann s axiom '' ) if if in a commutative diagram @xmath32^f \ar[d]_w & b \ar[d]^{v } \\ a ' \ar[r]_{f ' } & b ' } \ ] ] @xmath15 and @xmath16 are in @xmath1 , @xmath17 is a monomorphism , @xmath33 is normal monomorphism , and @xmath34 , then @xmath17 is a normal monomorphism ; * if in a commutative diagram @xmath35^{e_1 } \ar[dd]_f & \\ & & c \\ b \ar[rru]_{e_2 } } \ ] ] the morphisms @xmath36 and @xmath37 are in @xmath1 and @xmath38 , then there exists a factorization , in which @xmath23 is a monomorphism and @xmath39 is in @xmath1 .
[ theorem1 ] * condition [ def1](b ) implies conditions [ con1](a ) and [ con1](b ) ; * conditions [ def1](a ) , [ def1](c ) , [ con1](a ) and [ con1](b ) imply condition [ def1](b ) .
\(i ) is obvious .
( ii ) : let @xmath40 be a regular epimorphism in @xmath1 , and let @xmath13 and ; then condition [ con1](b ) implies that @xmath41 also is in @xmath1 . to prove that @xmath15 is a normal epimorphism , it is sufficient to show that the canonical morphism @xmath42 is an isomorphism . for , consider the commutative diagram @xmath43_{\bar{h } } \ar@<0.5ex>[r]^{s_1 } \ar@<-0.5ex>[r]_{s_2 } & a \ar[r]^>>>>>q \ar[dr]_f & \mathrm{coker}(k ) \ar[d]^{h } \\
r \ar@/_/[ru]^{r_1 } \ar@/_/@<-1ex>[ru]_{r_2 } & & b } \ ] ] in which : * @xmath44 is the kernel pair of @xmath15 and @xmath45 is the kernel pair of @xmath41 ; since the class @xmath1 is pullback stable ( condition [ def1](a ) ) , the morphisms @xmath46,@xmath47,@xmath48 , and @xmath49 are in @xmath1 . * @xmath50 is induced by @xmath51
. since there are canonical isomorphisms @xmath52 we can apply condition [ def1](c ) to the diagram @xmath53_{\approx } \ar[r ] & s \ar[d]_{\bar{h } } \ar[r]^{s_1 } & a \ar@{=}[d]\\ \mathrm{ker}(r_1 ) \ar[r ] & r \ar[r]_{r_1 } & a}\ ] ] this makes @xmath54 an isomorphism ; since @xmath15 and @xmath41 are regular epimorphisms , the latter implies that @xmath51 also is an isomorphism .
[ theorem2 ] * conditions [ def1](a ) and [ def1](c ) imply condition [ con1](c ) ; * condition [ def1](c ) implies condition [ con1](d ) ; * conditions [ def1](b ) , [ con1](c ) , and [ con1](d ) imply condition [ def1](c ) .
( i ) : according to the assumptions of [ con1](c ) , consider the commutative diagram @xmath12 \ar[r]^k & a \ar[r]^f \ar[d]_w & b \ar[d]^v \\ k \ar[r]_{k ' } & a ' \ar[r]_{f ' } & b ' } \ ] ] in which @xmath15 and @xmath16 are in @xmath1 , @xmath14 , @xmath55 is a morphism with @xmath56 , @xmath17 is a monomorphism , and @xmath33 is a normal monomorphism .
it is easy to see that @xmath55 is in fact the kernel of @xmath15 , and therefore condition [ def1](c ) can be applied to the diagram @xmath12 \ar[r]^k & a \ar[r]^f \ar[d]_{\langle w , f \rangle } & b \ar@{=}[d ] \\ k \ar[r]_>>>>>{h ' } & a'{\times}_{b'}b \ar[r]_<<<<<{{\pi}_2 } & b } \ ] ] where the projection @xmath57 , being the pullback of @xmath16 along @xmath33 , is in @xmath1 by condition [ def1](a ) .
it follows that @xmath58 is an isomorphism , and so @xmath17 is the pullback of @xmath33 along @xmath16 . since normal monomorphisms are pullback stable , we conclude that @xmath17 is a normal monomorphism , as desired .
( ii ) : since @xmath1 contains all isomorphisms , condition [ con1](d ) follows directly from condition [ def1](c ) .
( iii ) : we have to show that if in a commutative diagram @xmath12 \ar[r]^k & a \ar[r]^f \ar[d]_w & b \ar@{=}[d ] \\ k \ar[r]_{k ' } & a ' \ar[r]_{f ' } & b } \ ] ] @xmath15 and @xmath16 are in @xmath1 , and @xmath55 and @xmath59 are their kernels respectively , then @xmath17 is an isomorphism .
it is a well known fact , that in the situation above @xmath17 has zero kernel and zero cokernel . by conditions [ con1](d ) and [ def1](b )
, we have a factorization @xmath60 in which @xmath23 is a monomorphism and @xmath39 is a normal epimorphism in @xmath1 .
moreover , since @xmath17 has zero kernel , @xmath39 is an isomorphism .
therefore @xmath17 is a monomorphism , and applying condition [ con1](c ) to the diagram above , we conclude that @xmath17 is a normal monomorphism . since @xmath17 has zero cokernel , this implies that @xmath17 is an isomorphism , as desired .
combining these two theorems , we obtain : [ cor1 ] the following conditions are equivalent : * a pair @xmath0 is a relative weakly homological category ; * a pair @xmath0 satisfies conditions [ def1](a ) , [ def1](c ) , [ def1](d ) , [ def1](e ) , [ con1](a ) , and [ con1](b ) ; * a pair @xmath0 satisfies conditions [ def1](a ) , [ def1](b ) , [ def1](d ) , [ def1](e ) , [ con1](c ) , and [ con1](d ) . the implications ( i)@xmath61(ii ) , ( ii)@xmath61(i ) , ( i)@xmath61(iii ) , and ( iii)@xmath61(i ) follow from [ theorem1](i ) , [ theorem1](ii ) , [ theorem2](i)-(ii ) , and [ theorem2](iii ) respectively .
[ prop1 ] the following conditions are equivalent : * a pair @xmath0 in which @xmath1 is the class of all split epimorphisms in @xmath2 , is a relative weakly homological category ; * @xmath2 is a protomodular category in the sense of d.bourn @xcite .
the implication ( i)@xmath61(ii ) follows directly from the definitions .
( ii)@xmath61(i ) : the only condition that requires a verification here is [ def1](b ) ; however it holds by proposition 3.1.23 of @xcite , which asserts that in a pointed protomodular category with finite limits , a morphism @xmath15 is a regular epimorphism if and only if @xmath62 . [ prop2 ] if @xmath2 has coequalizers of kernel pairs and @xmath1 is the class of all regular epimorphisms in @xmath2 , then the following conditions are equivalent : * @xmath0 is a relative weakly homological category ; * @xmath0 is a relative homological category ; * @xmath2 is a homological category in the sense of f. borceux and d. bourn @xcite .
( i)@xmath61(iii ) : as follows from ( i ) , the class of all regular epimorphisms in @xmath2 is pullback stable .
therefore , since @xmath2 has kernel pairs and their coequalizers , it admits ( regular epi , mono)-factorization system .
furthermore , using the same arguments as in @xcite , one can show that in this situation , protomodularity is equivalent to the @xmath1-short five lemma .
it follows that @xmath2 is a homological category .
( iii)@xmath61(ii ) : let @xmath2 be a homological category and @xmath1 be the class of all regular epimorphisms in @xmath2 . then all properties we need for @xmath0 to be a relative homological category , are proved in @xcite .
since the implication ( ii)@xmath61(i ) is trivial , this completes the proof .
let @xmath0 be a relative weakly homological category and let @xmath63 be a pair , in which @xmath64 is a category with finite limits and @xmath65 is a class of morphisms in @xmath64 satisfying conditions [ def1](a ) , [ def1](d ) , and [ def1](e ) .
if functor @xmath66 preserves finite limits , then the pair @xmath67 , in which @xmath68 is the class of all morphisms @xmath39 from @xmath1 for which @xmath69 is in @xmath65 , is a relative weakly homological category . in particular we could take @xmath64 to be an arbitrary category with finite limits and @xmath70 to be the class of all split epimorphisms in @xmath64 . according to the existing literature ( see e.g. @xcite ) , an important example is provided by the forgetful functor @xmath71 from the homological category @xmath2 of topological groups to the category @xmath64 of topological spaces ; the class @xmath72 and the corresponding concept of exactness play a significant role in the cohomology theory of topological groups .
this also applies to the classical case of profinite groups , where , however , @xmath72 coincides with the class of all normal epimorphisms , as shown in section i.1.2 of @xcite ; another such result is used in @xcite .
the results of @xcite also suggest considering the forgetful functor from the category of topological groups to the category of groups . on the other hand one can replace topological groups with more general , so - called protomodular ( = semi - abelian ) , topological algebras , which form a homological category due to a result of f. borceux and m. m.clementino @xcite .
let us also mention the following `` trivial '' examples : if @xmath2 is an abelian category , and @xmath1 is the proper class of epimorphisms in @xmath2 in the sense of relative homological algebra ( see e.g. chapter ix in @xcite ) then @xmath0 is a relative weakly homological category .
a pair @xmath0 , in which @xmath1 is the class of all isomorphisms in @xmath2 , always is a relative homological category . a pair @xmath0 , in which @xmath1 is the class of all morphisms in @xmath2 , is a relative homological category if and only if @xmath2 is a trivial category .
[ def2 ] let @xmath0 be a relative homological category .
a sequence of morphisms @xmath73 & a_{i-1 } \ar[r]^{f_{i-1 } } & a_{i } \ar[r]^{f_i } & a_{i+1 } \ar[r ] & \ldots}\ ] ] is said to be : * @xmath1-exact at @xmath74 , if the morphism @xmath75 admits a factorization @xmath76 , in which @xmath77 and @xmath78 ; * an @xmath1-exact sequence , if it is @xmath1-exact at @xmath74 for each @xmath79 ( unless the sequence either begins with @xmath74 or ends with @xmath74 ) .
as easily follows from the definition , the sequence @xmath80 & a \ar[r]^f & b \ar[r]^g & c \ar[r ] & 0}\ ] ] is @xmath1-exact , if and only if @xmath81 and @xmath82 . having this notion of @xmath1-exact sequences , we may consider the relative cases of various homological lemmas from @xcite and @xcite .
+ * relative snake lemma . * using the same arguments as in the proof of the theorem 4.4.2 of @xcite
, we can prove the following [ lemma1 ] let @xmath0 be a relative homological category .
consider the commutative diagram @xmath83 & 0 \ar[d ] & 0 \ar[d ] \\ &
\mathrm{ker}(u ) \ar[r ] \ar[d ] & \mathrm{ker}(v ) \ar[r ] \ar[d ] & \mathrm{ker}(w ) \ar[d ] \\ & a \ar[r]^f \ar[d]_u & b \ar[r]^g \ar[d]_v & c \ar[r ] \ar[d]^w & 0 \\ 0 \ar[r ] & { a ' } \ar[r]^{f ' } \ar[d ] & { b ' } \ar[r]^{g ' } \ar[d ] & { c ' } \ar[d ] \\ & \mathrm{coker}(u ) \ar[r ] \ar[d ] & \mathrm{coker}(v ) \ar[r ] \ar[d ] & \mathrm{coker}(w ) \ar[d]\\ & 0 & 0 & 0 } \ ] ] in which all columns , and the second and the third rows are @xmath1-exact sequences .
suppose @xmath84 , or , more generally , the following conditions hold : then there exists a natural morphism @xmath90 , such that the sequence @xmath91 & \mathrm{ker}(v ) \ar[r ] & \mathrm{ker}(w ) \ar[r]^d & \mathrm{coker}(u ) \ar[r ] & \mathrm{coker}(v ) \ar[r ] & \mathrm{coker}(w)}\ ] ] is @xmath1-exact .
note that the relative snake lemma can be proved in a relative weakly homological category @xmath0 under some additional conditions .
these additional conditions , however , easily follow from [ def1](f ) and [ def1](g ) when @xmath0 is a relative homological category .
being more precise , to construct the morphism @xmath92 we only need condition [ lemma1](a ) , but to prove the @xmath1-exactness of the sequence above , we need the following : * condition [ lemma1](b ) ; * the morphisms @xmath93 and @xmath89 admit a ( normal epi in @xmath1 , mono)-factorization ; * if @xmath94 is a pullback of @xmath95 and @xmath96 , + @xmath97 is a pullback of @xmath98 and @xmath92 , and if @xmath99 is induced by @xmath16 and @xmath100 , then the morphism @xmath101 is in @xmath1 .
+ [ lemma2 ] let @xmath0 be a relative homological category
. if in a commutative diagram @xmath102 & 0 \ar[d ] & 0 \ar[d ] \\ 0 \ar[r ] & a \ar[r]^{f } \ar[d]_{u } & b \ar[r]^{g } \ar[d]_{v } & c \ar[d]^{w } \ar[r ] & 0 \\ 0 \ar[r ] & a ' \ar[r]_{f ' } \ar[d]_{u ' } & b ' \ar[r]_{g ' } \ar[d]_{v ' } & c ' \ar[d]^{w ' } \ar[r ] & 0 \\ 0 \ar[r ] & a '' \ar[r]_{f '' } \ar[d ] & b '' \ar[r]_{g '' } \ar[d ] & c '' \ar[r ] \ar[d ] & 0\\ & 0 & 0 & 0}\ ] ] the three columns and the middle row are @xmath1-exact sequences , then the first row is an @xmath1-exact sequence if and only if the last row is an @xmath1-exact sequence . if @xmath1 is the class of all regular epimorphisms in @xmath2 , then , by proposition [ prop2 ] , @xmath2 is a homological category and all the extra conditions needed for the proofs of the relative homological lemmas are satisfied . in this case , the relative snake lemma coincides with its absolute version " proved in @xcite , and the relative 3@xmath33-lemma coincides with its absolute version " proved in @xcite .
|
we introduce relative homological and weakly homological categories @xmath0 , where `` relative '' refers to a distinguished class @xmath1 of normal epimorphisms in @xmath2 . it is a generalization of homological categories , but
also protomodular categories can be regarded as examples .
we indicate that the relative versions of various homological lemmas can be proved in a relative homological category .
|
beginning with the early work of thouless @xcite , quasi one - dimensional conductors have provided a valuable arena in which to explore the influence of quantum interference effects on the transport properties of weakly disordered phase coherent conductors . by the early 80 s ,
a complete scaling theory of localization in multi - mode wires had been formulated by dorokhov @xcite ( and developed later by mello , pereyra and kumar @xcite ) as a brownian motion of eigenvalues of the transmission matrix the ` dmpk equation ' ( for a recent review see , e.g. , ref . ) . for several symmetry classes , analytic results for the low moments of conductance
were obtained in both the metallic and strongly localized regimes @xcite . lately , a third , and potentially more versatile approach has been developed to investigate quantum transport in ( multi - terminal ) disordered conductors . by formally relating the transmission eigenvalue distribution to a ` multi - component green s function ' ( see below ) , nazarov has shown @xcite that known results in the metallic limit can be inferred from the equations of motion of the average quasi - classical green function @xcite .
the transport properties of a conductor are fully specified by the transmission matrix @xmath1 @xcite .
in particular , the landauer - bttiker formula allows the conductance to be expressed through eigenvalues @xmath2 of the ` squared ' transmission matrix , @xmath3 the statistics of the @xmath2 follow the universal dorokhov distribution , valid in the metallic limit ( i.e. where the dimensionless conductance @xmath4 ) : @xmath5 by solving the corresponding quasi - classical equation in the diffusive limit , eq .
( [ dorokhov ] ) has been established under very general conditions . yet , while the method introduced by nazarov is capable of describing accurately the transmission eigenvalue distribution in the metallic phase , the method fails to account for the phenomena of weak and strong localization . in recent years
, the development of a field theoretic technique to explore phase coherence phenomena in disordered conductors has exposed the strengths and limitations of the quasi - classical scheme .
weakly disordered conductors and superconductors can be classified into one of ten symmetry classes @xcite , their spectral and transport properties specified by a field theory of nonlinear @xmath0-model type @xcite .
it was realized that within the formally exact framework provided by the field theory , the quasi - classical equations of motion represent a saddle - point or mean - field approximation , involving only a single field configuration @xcite .
indeed , drawing on the substantial literature on the quasi - classical equations in the superconducting context ( the eilenberger @xcite and usadel @xcite equations ) , this identification simplified substantially the analysis of physical realizations of the ` novel symmetry ' classes of disordered superconductors introduced by altland and zirnbauer @xcite .
the absence of localization in the quasi - classical theory follows then from the neglect of fluctuations about this field configuration .
describing localization within the @xmath0-model is however far from trivial .
while weak localization corrections to transport can be developed as a systematic perturbation theory , the transition to strong localization is signalled by the growth of contributions non - perturbative in the conductance .
indeed , a calculation by rejaei @xcite of the exact transmission eigenvalue distribution relied on a mapping of the @xmath0-model for the quasi one - dimensional system onto a heat kernel ( see also ref . ) .
such an analysis affords little intuition into the origin of localization within the framework of the @xmath0-model .
in particular , can the transition ( or , in quasi one - dimension , the crossover ) to strong localization be understood in terms of certain @xmath0-model field configurations ?
a preliminary answer to this question was provided in an insightful work by muzykantskii and khmelnitskii @xcite . using the field theoretic approach
, it was demonstrated that the long - time response of a weakly disordered quasi one - dimensional wire to a voltage step was controlled by spatially inhomogeneous saddle - point field configurations of the @xmath0-model action .
non - perturbative in the conductance , these field configurations were associated with rare `` anomalously localized '' states embedded deep within the metallic phase . whether these states provided a caricature of states near
the anderson transition remains to date the subject of debate .
later , in a related activity , it was shown the nucleation of localized quasi - particle states inside the gap of a weakly disordered symmetry broken superconductor were ascribed to spatially inhomogeneous saddle - point configurations of the action @xcite . in this paper
we will show that , for a range of symmetry classes , the physics of strong localization is captured _ exactly _ by the inclusion of a set of new saddle points combined with their associated gaussian fluctuations .
this recalls a previous investigation by andreev and altshuler @xcite of energy level correlations in symmetry broken ( zero - dimensional ) chaotic systems random matrix ensembles belonging to the unitary symmetry class . there it was shown that the two - point correlator of the density of states could be fully recovered from the inclusion of two saddle points together with their associated gaussian fluctuations . in both cases , the coincidence can be traced to the property of semiclassical exactness shared by the different @xmath0-model field theories .
the aim of the present paper is to elucidate this principle in the quasi one - dimensional system comparing results for transmission eigenvalue distributions to those obtained from heat kernel methods .
the novel view of localization presented here may inform the treatment of more complicated problems .
following rejaei , our analysis rests on a relation which allows the ensemble average of the generating function of the transmission matrix @xmath6 ( at a particular energy ) of a quasi one - dimensional sample @xmath7 to be presented as a partition function of a nonlinear @xmath0-model of the appropriate symmetry class . the density of transmission eigenvalues may then be extracted using the relation @xmath8 where @xmath9 denote the eigenvalues of the matrix @xmath10 . from this function
one can infer the transmission eigenvalue density through the relation @xmath11 the moments of the transmission matrix can be cast in terms of the retarded and advanced green s functions @xmath12 of the microscopic hamiltonian of the wire as @xmath13 where @xmath14 denotes the current operator through left and right cross - sections at @xmath15 and @xmath16 respectively , and the trace on the right hand side runs over spatial coordinates .
following nazarov , it is helpful to note that the determinants in eq .
( [ tgen ] ) can be recast in terms of a multi - component green s function according to the relation @xmath17 where , for example , in the case of the numerator of ( [ tgen ] ) , @xmath18 .
thus the generating function @xmath19 may be related to the green s function of an enlarged hamiltonian , which includes an off - diagonal ` vector potential ' . with this representation in hand
, the outline of the remainder of the paper is as follows . in the next section we present a complete analysis of the localization properties of class @xmath20i , one of the superconducting symmetry classes introduced by altland and zirnbauer @xcite .
we outline the @xmath0-model description of this symmetry class and reproduce nazarov s calculation in this context , before introducing the full set of saddle - point trajectories - geodesics on the target space - and performing a complete semiclassical calculation . from the latter
we will extract the transmission eigenvalue density and mean conductance .
the exactness of the results obtained are then verified by heat kernel methods . in the following section
, the techniques developed above will be applied to two additional symmetry classes that are amenable to such a treatment . finally , in the last section
, we will draw conclusions .
superconductors which exhibit both time reversal symmetry and spin rotation invariance belong to symmetry class @xmath20i . by introducing the familiar nambu doublet of electron and hole operators in what we will refer to as the ` particle - hole ' ( ph ) space , in the mean - field approximation
, the quasi - particle gorkov or bogoliubov - de gennes hamiltonian @xmath21 can be written as @xmath22 where @xmath23 and @xmath24 represent real symmetric matrix elements corresponding to the non - interacting single - particle hamiltonian of the quasi one - dimensional disordered wire and the superconducting order parameter respectively . with this structure ,
the matrix hamiltonian @xmath25 satisfies the symmetry relations @xmath26 which in turn implies the identity @xmath27 between the retarded and advanced green s functions . with this definition
, the ratio of determinants which specifies the generating function ( [ tgen ] ) can be conveniently expressed via a supersymmetric field integral , with the numerator arising from a fermionic integral and the denominator from its bosonic counterpart .
since these are not identical , the corresponding partition function will inherit boundary conditions which break the supersymmetry . to identify the soft diffusion modes of the superconducting system
, it is convenient to effect a doubling of the field space of the supervectors @xmath28 to include a ` charge - conjugation ' ( cc ) space @xcite @xmath29 focusing on the transmission eigenvalue distribution at @xmath30 , we notice that if @xmath31 is the @xmath32 component of the matrix green s function @xmath33 in the cc space , then the advanced component @xmath34 is just the ( 2,2 ) component .
thus the enlarged structure required is just the usual cc space .
altogether , this leads to the representation @xmath35}^{{{\cal g}}^{-1}}\psi\big]\end{aligned}\ ] ] where @xmath36 where @xmath37 , and @xmath38 denote the pauli matrices in the cc space . once cast in the form of a functional field integral , it is a straightforward ( if somewhat lengthy ) and standard procedure to show that the low - energy properties of the partition function are contained with an effective field theory of nonlinear @xmath0-model type ( for a detailed discussion of the explicit derivation , we refer to one of the many standard references , e.g. ref .
two points should be remarked upon .
the first is that , since class @xmath20i has two symmetries particle - hole and time reversal ( tr ) symmetry ( see eq .
[ phsym ] ) the supervector needs to be doubled _ twice_. the size of the supermatrix field @xmath39 that appears at an intermediate stage in the derivation is thus @xmath40 ( ph@xmath41cc@xmath41tr@xmath41bf ) , but this is quickly reduced to the @xmath42 field @xmath43 by @xmath44 on the saddle - point manifold ( @xmath45 denote the pauli matrices in ph space ) , with the ph space then disappearing from view . secondly ,
following rejaei , the ` vector potential ' associated with the coupling of the wire to the external leads , can be absorbed into a rigid boundary condition on the field integral .
taking the metallic contacts at @xmath46 to be to clean normal conducting leads , the ensemble averaged partition function assumes the form ( with @xmath47 ) @xmath48 ^ 2 \right)\ ; , \label{sigma}\end{aligned}\ ] ] where the field integral is over @xmath42 supermatrix fields @xmath43 subject to the nonlinear constraint @xmath49 .
finally , the supersymmetry breaking source terms enter the boundary condition , through the rotation matrix @xmath50 specifically , if we choose @xmath51 then the boundary condition at @xmath16 is @xmath52 where @xmath53 and we have obtained the required form of rejaei s relation ( [ tgen ] ) .
an alternative approach that matches nazarov s orginal formulation is to use the explicit expression for @xmath54 that follows from eq .
( [ fdef ] ) and relate it directly to the green s function in the cc space , i.e. @xmath55\right\rangle \nonumber \\ * & = & \left\langle\mathrm{tr}_{\bf r}\left [ \hat v_r { { \cal g}}^{12}_{\mathrm{bb } } \right]\right\rangle\big|_{\theta = i\phi}\nonumber\\ * & = & -\frac{i\pi\nu d}{2 } { { \cal a}}\left\langle \mathrm{tr}_{\mathrm{tr } } \left(q\partial_x q\right)^{12}_{bb}\big|_{x=0 } \right\rangle_q\big|_{\theta = i\phi } , \label{nazf}\end{aligned}\ ] ] where @xmath56 denotes an average with respect to the @xmath0-model action ( [ sigma ] ) , @xmath57 is the cross - sectional area , and @xmath58 is the single particle density of states .
@xmath59 denotes the trace over the tr space .
henceforth , we will work only with the one - dimensional field @xmath60 , assuming that the width of the wire is much smaller than the localization length . to develop a theory of localization from the field integral ,
it is first useful to establish contact with the theoretical framework introduced by nazarov . as discussed in the introduction , the quasi - classical theory formulated by nazarov
is contained within the saddle - point structure of the present action : specifically , a variation of the action with respect to @xmath43 subject to the nonlinear constraint , @xmath61 , obtains the saddle - point equation @xmath62 which expresses conservation of the matrix current @xmath63 .
indeed , associating @xmath43 with the average quasi - classical green function , the saddle - point equation can be interpreted as the generalization of the quasi - classical equation derived in nazarov s early work to the superconducting wire .
the simplest solution satisfying the given boundary conditions is block diagonal in the bf space and takes the form @xmath64 if one inserts this solution into eq .
( [ nazf ] ) , and neglects field fluctuations around the saddle point , one finds @xmath65 then , making use of eq .
( [ density ] ) , one can straightforwardly recover the dorokhov distribution ( [ dorokhov ] ) for a wire of dimensionless conductance @xmath66 .
however , when derived from the supersymmetric field theory , one can see that the saddle - point equation presents not just one solution , but a whole family of solutions in the compact fermionic sector @xmath67 thus we see that even at the level of a purely _ semi - classical _ analysis , nazarov s treatment is missing two elements ; corrections associated with fluctuations around the conventional saddle - point solution ( [ nazsol ] ) , and solutions ( [ windings ] ) that loop multiply around the compact fermionic sector .
we will show that these two ` channels ' of corrections are , respectively , responsible for the phenomenon of weak localization as , indeed , one might have guessed from the connection between the role of fluctuations in the @xmath0-model and diagrammatic perturbation theory and strong localization . to uncover this relation , in the following
we will undertake the semi - classical analysis taking into account fluctuations of the matrix fields at quadratic order around both the conventional and the non - trivial saddle - point field configurations . to perform the field integration over the fluctuations ,
it is necessary to review the structure of the target manifold . for the minimal ( @xmath68 ) nonlinear @xmath0-model associated with the symmetry class @xmath20i , the target space
spanned by matrices @xmath69 turns out @xcite to be in one - to - one correspondence with @xmath70 , by which we mean the lie group @xmath71 extended to an orthosymplectic lie supergroup .
this group structure of the target manifold is the reason for the exactness of the semi - classical approximation @xcite .
the correspondence between @xmath43 and @xmath72 is described in the appendix .
the space of matrices ( [ bc ] ) parameterized by @xmath73 and @xmath74 corresponds to the maximal abelian subgroup @xmath75 @xmath76 while , in terms of group elements @xmath77 , the generating function is @xmath78 ^ 2 \ , dt \right)\,.\ ] ] here we have switched to the dimensionless variables @xmath79 , @xmath80 , where @xmath81 .
( note that , with this definition , @xmath82 . ) with this parametrization , the set of saddle - point solutions discussed in the previous section corresponds to the geodesic trajectories of free particle motion @xmath83 where @xmath84 with @xmath85 .
these configurations are associated with the classical action @xmath86 ^ 2 \ , dt= ( \phi^2 + \theta_n^2 ) / 4t\ ] ] to evaluate the contribution from gaussian fluctuations , it is helpful to set @xmath87 , and express the element @xmath88 through the exponential parametrization @xmath89 , with boundary conditions @xmath90 .
matrices @xmath91 satisfying the @xmath92 lie algebra condition @xmath93 with @xmath94 may be written as @xmath95 then , to get the desired target space with the non - compact boson - boson sector @xmath96 , and compact fermion - fermion sector @xmath97 , one must take the variable @xmath98 to be real , the variable @xmath99 to be imaginary , and the complex variables @xmath100 and @xmath101 to be related by the condition @xmath102 .
the functional integration measure is trivial , @xmath103 , up to corrections ( of order @xmath104 ) that will not affect the present calculation . using the parametrization for @xmath91 given in eq .
( [ xparam ] ) one obtains the quadratic action @xmath105 \right ) \\ * & = & \frac{1}{4}\int dt\ ; \left ( \dot d^2 - \dot e^2 - \dot b \dot c - \frac{i\theta_n}{t } \left ( b \dot c - \dot b c \right ) \right .
\\ & & \left .
\qquad\qquad + 2 \dot\alpha\dot\delta + \frac{\phi - i\theta_n}{t } ( \alpha \dot\delta - \dot\alpha \delta ) \right .
\\ & & \left .
\qquad\qquad + 2 \dot\gamma \dot\beta - \frac{\phi + i\theta_n}{t } ( \gamma \dot\beta - \dot\gamma \beta ) \right)\;.\end{aligned}\ ] ] performing the gaussian functional integral over @xmath98 and @xmath99 is the same as computing the kernel for free particle motion on the two - dimensional euclidean space @xmath106 , and the result is @xmath107 .
performing the remaining gaussian functional integrals gives determinants in the denominator ( bosons : @xmath108 ) and numerator ( fermions : @xmath109 and @xmath110 ) .
collecting all the determinants , one finally obtains the total fluctuation contribution to the partition function , @xmath111 the determinants are to be evaluated on the hilbert space of square - integrable functions @xmath112)$ ] with dirichlet boundary conditions .
if @xmath113 is some complex number , the operator @xmath114 on that hilbert space has eigenfunctions @xmath115 ( @xmath116 ) , with eigenvalues @xmath117 , so its ( unregularized ) determinant is given by @xmath118 taking the logarithmic differential ( to kill the @xmath113independent infinity ) one obtains @xmath119 therefore , @xmath120 and the above ratio of four determinants gives @xmath121 finally , when combined with the exponential of the classical action and summed over @xmath122 , one obtains the following expression for the partition function : @xmath123 note that the operator @xmath124 starts to exhibit negative eigenvalues as soon as @xmath125 exceeds @xmath126 .
this means that the gaussian functional integral over @xmath100 and @xmath102 does not exist in those cases .
thus what we have done evaluating the functional integral in the gaussian approximation around all of the saddle points ( without paying attention to question of existence ) , and summing contributions was a purely formal calculation .
nevertheless , the answer obtained in this way turns out to be exact !
indeed , the validity of this expression can be established both directly and indirectly . in the following
, we will motivate this conclusion by investigating the transmission eigenvalue density and the mean conductance from the partition function .
then , by interpreting the field integral as a heat kernel , we will find an alternative exact method of computation . with the partition function in hand
, it is a straightforward matter to compute the eigenvalue density making use of the relations ( [ tgen])-([density ] ) , @xmath127 the first term of this expression is just the dorokhov distribution , and originates from the @xmath128 ( nazarov s ) configuration . in this context , the second term simply denotes the weak localization correction associated with the @xmath128 configuration .
all higher - order terms of the perturbation expansion in powers of @xmath82 turn out to vanish identically , a direct manifestation of semi - classical exactness .
indeed , the remaining terms in the result ( [ xdensity ] ) depend non - analytically on @xmath129 as @xmath130 , and arise from the non - trivial saddle - point configurations . a non - trivial check on the validity of the expression follows from the requirement that @xmath131 for all @xmath132 .
( one can infer this condition from the joint probability density given by brouwer _
et al . _
the plot of the density function @xmath133 in fig .
[ fig : densities ] shows the characteristic ` crystallization ' of transmission eigenvalues associated with the onset of localization .
the largest @xmath2 ( smallest @xmath134 ) eigenvalue crystallizes at @xmath135 , corresponding to @xmath136 if @xmath137 .
( 1,0.25)(0.2,0 ) ( 0.02,0 ) ( 0.38,-0.05)(0,0)[b]@xmath138 ( 0,0.22 ) ( 0.67,0 ) ( 1.03,-0.05)(0,0)[b]@xmath138 ( 0.65,0.22 ) a second , and more direct way of inferring exponential localization is to investigate the mean ( spin ) conductance of the superconducting wire .
the mean conductance @xmath139 of the thick disordered wire can be obtained by integrating @xmath140 against the eigenvalue density @xmath141 with the result @xmath142 on applying the rescaling @xmath143 , the latter agrees with the expression derived in ref . .
the leading term @xmath144 is the ohmic contribution , while the constant @xmath145 represents the weak localization correction . to extract the large-@xmath129 asymptotics and demonstrate exponential localization ,
the above expression for @xmath146 is inconvenient .
instead , it is convenient to implement an integral transform of the expression which brings it to a form in which the large-@xmath129 asymptotics can be developed . making use of the identities @xmath147 and @xmath148
, one may easily show that @xmath149 noting that the function @xmath150 is the fourier transform of @xmath151 , poisson summation yields @xmath152 finally , introducing the complement of the error function , @xmath153 the expression for @xmath154 may be brought to the final form @xmath155 while the original expression provides access to the small-@xmath129 asymptotics , the second one gives easy access to the large-@xmath129 behaviour , viz . @xmath156 recalling @xmath157 , the latter shows the characteristic exponential dependence of localization with a localization length @xmath158 .
consideration of the eigenvalue density and mean conductance provide strong circumstantial evidence that the expression derived for the generating function is exact .
however , to emphasize the validity of the semi - classical expansion , it is useful to provide an explicit calculation of the partition function as a solution of the euclidean - time ` schrdinger ' equation or heat kernel . by canonical quantization , the path integral with the lagrangian @xmath159 becomes the ` schrdinger equation ' for free - particle quantum motion on @xmath160 .
the schrdinger operator is the negative of the laplace - beltrami operator @xmath161 ( or laplacian for short ) on @xmath160 .
thus the heat kernel satisfies a schrdinger - like equation which is the diffusion ( or heat ) equation on @xmath160 : @xmath162 where @xmath161 acts on the left argument of @xmath163 .
the reduction from the full heat kernel to the radial function @xmath164 on the maximal abelian subgroup @xmath165 takes the diffusion equation into @xmath166 where we continue to denote the radial part of the laplacian by @xmath161 for simplicity .
since @xmath160 has superdimension zero ( there being the same number of bosonic and fermionic degrees of freedom ) the short - time asymptotics of the heat kernel is given by @xmath167 ( in dimensions @xmath168 , the gaussian would be preceded by a factor @xmath169 . ) to proceed , we must draw on some basic facts from the geometry and analysis of the supermanifold @xmath160 : * diagonalization of matrices @xmath170 ( i.e. @xmath171 with @xmath172 ) defines a polar decomposition of @xmath160 . by this decomposition , the @xmath160-invariant berezin integration measure for @xmath160 determines a radial integration measure @xmath173 , which is positive on the weyl chamber chamber @xmath174 \times [ 0,\pi ] \subset { \rm lie}a$ ] .
* by a standard formula @xcite from the theory of lie groups and symmetric spaces ( for a recent pedagogical review see , for example , ref . )
, the measure function @xmath175 for the case at hand is given by @xmath176 * finally , the laplacian on radial functions is given by @xmath177 by introducing complex coordinates @xmath178 and @xmath179 it is apparent that the analytic square root of the measure function , @xmath180 is harmonic : @xmath181 . using this property
one can easily verify that the radial part of the laplacian can be cast in the form @xmath182 it therefore follows that the product @xmath183 satisfies the euclidean heat equation @xmath184 the euclidean heat kernel in two dimensions is known to be @xmath185 .
however , this is not the solution we want here : as mentioned earlier , the heat kernel @xmath164 is subject to zero - dimensional small-@xmath186 asymptotics , @xmath187 .
this short - time behavior is achieved with the choice @xmath188 to confirm that @xmath189 satisfies the euclidean heat equation , one uses the identity @xmath190 and @xmath191 the solution @xmath192 thus obtained is not yet @xmath193-periodic in @xmath74 and hence does not lift to a function on the abelian group @xmath194 . enforcing periodicity by summing over images ,
we obtain : @xmath195 where @xmath196 as before .
this is the correct answer , and it is easily seen to coincide exactly with the semi - classical result for the partition function derived above . as emphasized above , the coincidence of the semi - classical expansion with the exact expression for the partition function is a particular feature of the group manifold structure of the @xmath0-model for the symmetry class @xmath20i . as a result , we can deduce the existence of other novel symmetry classes where the transport properties of the quasi one - dimensional system can be inferred from the structure of the semi - classical expansion .
having established the principle , we now turn to consider extensions of the present scheme to other symmetry classes . here
we consider classes @xmath197iii and @xmath165iii : symmetry @xmath197iii is realized @xcite in superconductors which exhibit time - reversal symmetry but where the @xmath198 spin rotation symmetry is broken by a spin scattering mechanism such as a spin - orbit interaction .
the latter presents a critical testing ground for the present theory as it has already been established by brouwer et al .
@xcite that exponential localization is _ absent _ generically in quasi one - dimensional systems of this symmetry class . for the symmetry class @xmath197iii ,
the corresponding target space of the nonlinear @xmath0-model is a riemannian symmetric superspace of type @xmath199 @xcite .
this means that we are to work again with the complex orthosymplectic lie supergroup , @xmath200 , but now one must select a non - compact symmetric space @xmath201 in the boson - boson sector and the compact group @xmath202 in the fermion - fermion sector . for @xmath68 these spaces are isomorphic to @xmath203 ( the three - hyperboloid ) and @xmath204 respectively .
a maximal commuting subgroup @xmath205 is still formed by diagonal matrices @xmath206 with @xmath207 and @xmath208 $ ] , so the saddle - point configurations ( [ spset ] ) are unchanged . only the fluctuation contribution differs .
the parametrization of the target space by the exponential mapping @xmath209 is the same as for class @xmath20i but with the boson - boson and fermion - fermion sectors interchanged : @xmath210 where @xmath98 is now imaginary , @xmath99 real , and @xmath211 . in view of the duality
( by exchange of the compact and non - compact sectors ) connecting the nonlinear @xmath0-models for the classes @xmath20i and @xmath197iii , all calculations for @xmath197iii are very similar to those for @xmath20i and , for brevity , we simply quote the results here .
the partition function , obtained by performing the sum over geodesics together with the integral over gaussian fluctuations , is given by @xmath212 from rejaei s relation , the corresponding density of transmission eigenvalues is then given by @xmath213 a plot of this function in fig .
[ fig : densities ] shows the ` crystallization ' of transmission eigenvalues for @xmath214 . yet
, exponential localization does not take place , as @xmath215 peaks at @xmath216 ( maximal transmission ) , and the peak amplitude decays only algebraically with increasing @xmath129 . turning to the mean conductance , the partition function yields the following expression : @xmath217 the first expression agrees with the result of ref . ; the second is obtained from it by poisson resummation .
the first two terms , @xmath218 , represent the ohmic and weak anti - localization terms that can be computed from a standard perturbation theory for the mean conductance . as with symmetry class @xmath20i , higher - order corrections from the perturbative expansion vanish identically .
all non - zero corrections are non - perturbative and arise from the non - trivial geodesics in the compact sector . the second expression in eq . ( [ diiicond ] )
presents an anomalous diffusive asymptotics for large @xmath219 : @xmath220 a feature already seen in ref . . finally ,
expressed in the form of the heat kernel , the solution for class @xmath197iii is trivially related by the duality discussed earlier and the validity of the expression for the partition function above may be confirmed straightforwardly . as a final application of semi - classical exactness , we turn now to one of the chiral symmetry classes .
this symmetry class is relevant to the low - energy physics of the chiral dirac operator @xcite and to the random flux model @xcite .
lately , it has been discussed @xcite in the context of the quasi - particle properties of a @xmath98-wave superconductor subject to a smooth random potential .
previous studies of quasi one - dimensional systems in this class @xcite have revealed a surprisingly rich behaviour , including localization - delocalization transitions with a number of critical points . before exploring the properties of the nonlinear @xmath0-model action ,
it is first necessary to confirm the form of rejaei s relation for this symmetry class . for this
we refer to the @xmath0-model formulation of the bond disordered quasi one - dimensional chain given by altland and merkt @xcite .
we consider @xmath221 chains of @xmath222 sites with hopping matrix element @xmath223 along the chains between sites @xmath224 and @xmath225 , where @xmath226 is a ` staggering ' parameter . in the following ,
we suppose that a weak disorder potential ( which , further , lifts time - reversal symmetry ) of strength @xmath227 couples the chains .
we view the single particle hilbert space as the product @xmath228 , with the first factor referring to the odd and even numbered sites along the chain . the sublattice symmetry of the hamiltonian translates to the condition @xmath229 , where @xmath38 represent the pauli matrices in this odd / even subspace .
from this sublattice symmetry there follows the relation @xmath230 the corresponding current operator is simply given by @xmath231 ( if we ignore the disorder and staggering , or evaluate it in the clean , unstaggered leads ) .
thus the moments of the transmission matrix at zero energy are given by @xmath232 where @xmath233 projects onto the @xmath234 site of each chain .
it seems that these moments can not in general be presented as the coefficients in some expansion of a functional determinant .
the average conductance ( @xmath235 ) may however be obtained as the @xmath236 coefficient of the determinant of a hamiltonian with hopping @xmath237 between the @xmath238 and @xmath239 sites , and @xmath240 between the @xmath241 and @xmath242 sites ( since hopping is described by a @xmath243 term in the hamiltonian ) .
we may thus study the usual partition function @xmath244 for @xmath245 ( the matrix @xmath170 being @xmath246 ) .
the conductance can be then identified with the coefficient @xmath146 in the expansion @xmath247 the most general nonlinear @xmath0-model lagrangian for quasi one - dimensional systems belonging to class @xmath165iii ( and in suitable units of length )
depends on two parameters @xmath248 : @xmath249 the target space of the nonlinear @xmath0-model for this class @xcite is of type @xmath250 and is obtained from the complex lie supergroup @xmath251 by picking @xmath252 in the boson - boson sector and @xmath253 in the fermion - fermion sector . for @xmath254
these are @xmath255 and @xmath204 respectively .
the exponential parametrization @xmath209 of the target space is achieved by setting @xmath256 with real commuting variables @xmath257 and anti - commuting variables @xmath258 .
again , @xmath259 parameterizes the maximal abelian subgroup .
based on the geodesics @xmath260 with @xmath261 , the semiclassical approximation for @xmath262 is found to be @xmath263 / 4 t } \;.\ ] ] expanding the result above for the partition function , and using eq .
( [ condrel ] ) gives @xmath264 where the second expression is obtained by poisson resummation of the first .
the result ( [ aiiicond ] ) depends periodically on the parameter @xmath265 with period @xmath193 .
in fact , @xmath74 has the meaning of a topological angle . from ref .
we know that tuning @xmath74 from @xmath266 to @xmath193 amounts to changing by two the number of channels ( which is assumed to be large in order for the @xmath0-model approximation be valid ) . in terms of the bond disordered chains discussed earlier @xmath267 where @xmath268 is the staggering parameter @xmath226 scaled by the strength of the disorder .
we see that without staggering , a delocalized state appears only for odd channel number .
the localization length diverges with critical exponent @xmath269 as the staggering approaches the critical values @xmath270 .
if class @xmath165iii is realized by dirac fermions in a random abelian gauge field , the parameter @xmath271 measures the strength of the gauge disorder .
although the lagrangian @xmath272 at first sight would seem to become unstable at @xmath273 ( the coefficient of @xmath274 becomes negative there ) , this is not really so .
what matters is the _ full _ quadratic form @xmath275 which has the jordan normal matrix form @xmath276 , and is invertible for any @xmath277 .
in fact , the final result ( [ aiiicond ] ) reveals a smooth dependence of the mean conductance on @xmath277 for all physical values ( @xmath271 ) including the fake singularity @xmath273 . finally ,
seeking an exact solution through the heat kernel , an exact calculation of @xmath278 is still possible as the hamiltonian @xmath279 obtained by canonical quantization from @xmath272 can once again be transformed to euclidean form : @xmath280 with @xmath281 . by solving the differential equation @xmath282 with the appropriate @xmath283-function initial conditions at @xmath284
, we exactly recover the semiclassical answer ( [ aiiikern ] ) .
we have confirmed the exactness of our semiclassical analysis for symmetry classes @xmath20i , @xmath197iii , and @xmath165iii . by simply summing over the saddle points of the classical action
recall that these differed by the length of the geodesic looping around the compact fermion - fermion submanifold of the theory and treating the fluctuations in the gaussian approximation , we have obtained the exact result for the partition function .
the underlying reason for the exactness of our calculations is that the path integral on the @xmath0-model manifolds considered here satisfies an infinite - dimensional generalisation of the duistermaat - heckman theorem @xcite .
that is , the integration manifold in eq .
[ g_action ] , being related to the space @xmath285 of based loops in @xmath160 , is symplectic with liouville measure , and the action is the momentum mapping on this manifold .
we have discussed all the symmetry classes where the @xmath0-model target space has a group - like structure , and thus exhausted the situations in which the semiclassical approach is exact .
nevertheless , we hope that this approach may inform future investigations of more taxing localization problems .
one very interesting extension that could be tackled in a similar manner is the critical scaling of the localization length near the band center in situations where zero energy is described by class @xmath197iii or @xmath165iii .
this lies outside of the scope of methods like the dmpk equation .
let us review the structure of the @xmath0-model target space for class @xmath20i . the construction used in ref .
is based on ` copying ' the symmetries of the hamiltonian to symmetries of the auxiliary space of the @xmath286-field . for class @xmath20i ( particle - hole and time reversal symmetry ) , this is achieved by demanding that the fields satisfy @xmath287 the subscript @xmath288 is to remind us that in the formulation of ref . the variables @xmath289 and @xmath290 are ( super)_matrices _ mapping the auxiliary space to the physical hilbert space ( where the hamiltonian acts ) , and vice versa . in the present case we work with the more traditional column and row supervectors @xmath291 and @xmath292 , since : * in the localization problem the @xmath39 field used to decouple the term arising from potential disorder retains the hilbert space structure of the hamiltonian ( particle hole space and spatial index ) , whereas in the random matrix problem that structure is lost on averaging and , more importantly , * we use some of the additional structure ( the cc space ) to enlarge our hamiltonian ( eq . [ enlargedh ] ) . transcribing the conditions ( [ matrixcond ] ) for supervectors gives @xmath293 where we used the fact that @xmath294 and @xmath295 can be chosen to be orthogonal matrices
now in the localization problem @xmath39 is a @xmath40 matrix ( ph@xmath41cc@xmath41tr@xmath41bf - by ` tr ' we mean the space to accommodate time reversal symmetry ) , so we need to check that the conditions ( [ vectorcond ] ) give rise to a @xmath0-model with the same @xmath42 space as in the random matrix problem .
this is straightforwardly verified following ref . by observing that @xmath296 ( we drop the @xmath297 subscript from now on ) ,
arising from a decoupling of potential disorder , has the symmetry @xmath298 the saddle - point manifold is spanned by @xmath299 or @xmath300 thus ( [ qsym ] ) implies that the symmetries of the @xmath42 field @xmath43 are @xmath301 which are the same relations found in ref . .
to construct an embedding of the @xmath302 matrix group - valued field @xmath170 in @xmath43 requires an explicit choice of @xmath294 and @xmath295 .
we use @xmath303 @xmath304 where @xmath305 are the pauli matrices in tr space , and @xmath306 and @xmath307 are projectors in the bose - bose and fermi - fermi sectors .
note that the structure of the vector in cc space in the bosonic sector is in fact then _ not _ given by eq .
[ ccdoubling ] , but rather looks like @xmath308 fortunately this does nt change the boundary conditions .
now the correct embedding is obtained by finding a transformation to diagonalize the matrix @xmath309 @xmath310 while simultaneously sending @xmath311 why ?
the point is that the set of @xmath312 s that preserve the symmetries ( [ qsym ] ) takes the form @xmath313 in such a basis .
the stability group of transformations that send @xmath314 is then given by eq .
[ tdef ] with @xmath315 , so that the saddle - point manifold is parameterized by @xmath316 an explicit form for @xmath317 is @xmath318 let us verify that this takes the maximal abelian subgroup @xmath165 to the @xmath43 matrices of the form ( [ bc ] ) : @xmath319 finally , the form of the action ( [ g_action ] ) in terms of @xmath170 follows from @xmath320 ^ 2&=&2\,\mathrm{str}\left [ \nabla g\nabla g^{-1 } \right]\\ & = & -2\,\mathrm{str}\left [ g^{-1}\nabla g \right]^2.\end{aligned}\ ] ]
|
we use a novel method based on the semi - classical analysis of @xmath0-models to describe the phenomenon of strong localization in quasi one - dimensional conductors , obtaining the density of transmission eigenvalues . for several symmetry classes , describing random superconducting and chiral hamiltonians , the target space of the appropriate @xmath0-model is a ( super)group manifold . in these cases
our approach turns out to be exact .
the results offer a novel perspective on localization .
|
one of the prime unresolved issues in the dynamics and evolution of galaxy disks remains the origin and evolution of large - scale bar and spiral structure .
though the persistence of grand - design spirals has been tied observationally to the presence of bars or companions @xcite , virtually nothing is known about the actual lifetimes of spiral patterns . additionally , despite indications that the relation between bar and spiral pattern speeds ( which @xcite first argued may not be equal ) may be important for understanding the role of bars in angular momentum transfer during secular disk evolution ( e.g. @xcite and @xcite ) , there are as yet unanswered questions about the connections between multiple patterns in different radial zones . while mode - coupling between patterns , which allows efficient outward angular momentum transfer in disks ( @xcite ; @xcite ) seems a most promising link , in 2d n - body simulations with a dissipative gas component @xcite find evidence for spiral structure in the absence of a bar , bar - spiral mode coupling , spiral - spiral mode coupling , and multiple pattern speeds without mode coupling . clearly , to address questions about the persistence of spiral patterns and the relation between bars and spirals requires not only determination of the pattern speed but how it varies with radius
; only with accurate measurement of bar and spiral or inner and outer spirals pattern speeds in the same galaxy can we confirm whether spiral structure is steady or winding , whether bars and spiral pattern speeds are equal or are unrelated , whether mode - coupling exists , and the domain and number of patterns that can be sustained in a disk .
because they are not directly accessible through observation , pattern speeds are often determined with indirect means such as the identification of predicted behavior at resonance radii ( e.g. @xcite ; @xcite ) or kinematic and morphological comparisons of simulated and observed structure ( e.g. @xcite ; @xcite ) .
it is also clearly desirable to employ methods for estimating pattern speeds which do not rely on theoretical models or simulation .
many other pattern speed determinations have therefore centered on the use of the model - independent method of tremaine & weinberg ( 1984 ; hereafter tw ) which presents a rigorous derivation for the pattern speed @xmath1 based on the requirement of continuity using observationally accessible quantities .
the determination of @xmath1 involves surface density - weighted position and velocity line integrals parallel to the galaxy major axis under several essential assumptions .
specifically , the method requires that the disk of the galaxy is flat ( unwarped ) and contains a single , well - defined rigidly - rotating pattern ; that the surface density of a kinematic tracer of a disk component , which must obey continuity , becomes negligibly small at some radius and all azimuths within the map boundary ( thereby critically yielding converged integrals ) ; and that the relation between the emission from this component and its surface density is linear , or if not , suspected deviations from linearity can be modeled . in that
information from all sampled radii is associated with a single , constant pattern speed , the tw calculation poses a challenge for extracting multiple distinct or radially varying pattern speeds .
non - axisymmetric structures beyond a dominant pattern such as a bar in the disk of a galaxy will interfere with the measurement of @xmath1 of the bar ; when a non - axisymmetric disk can be decomposed into two components with different pattern speeds , then the tw estimate is a luminosity- and asymmetry - weighted average of the two patterns @xcite .
the tw estimate for the secondary pattern speed in the inner disk of ngc 6946 observed in h@xmath2 , for example , is estimated to be limited to an uncertainty of as much as 50% given the primary pattern s contribution to the tw integrals @xcite .
such issues notwithstanding , there have been several recent adaptations of the tw method to positive effect .
applications of the tw method to sb0 galaxies using stellar light as a tracer extend the limits of integration in the tw calculation to just past the end of the bar in order to minimize contributions to the tw integrals from non - axisymmetric ( and several magnitudes dimmer ) features beyond the bar ; in such cases , integrating past the structure of interest is found to be sufficient for achieving converged integrals .
the bar pattern speeds in ngc 7079 @xcite and ngc 1023 @xcite , for example , have both been successfully measured in this way
. when there exist more than one pattern in distinct radial zones , however , arguments about the convergence of the tw integrals are less straight - forward . to measure the secondary bar pattern speed in ngc 2950 ,
@xcite and @xcite explore decoupling the inner secondary and outer primary bar pattern speeds by associating each component with unique surface brightness contributions in the tw calculation .
the tw integrals are modified by the presence of the inner pattern based on an assumption about how and where the two patterns decouple ( to first order ) .
this analysis has confirmed the existence of , if not measured , a unique secondary bar pattern speed possibly indicating counter - rotation with respect to the primary bar . at their best improving the accuracy of pattern speed estimates to about 20% @xcite these kinds of adaptations of the tw method for measurement of single or multiple patterns still require assumptions about bar extent based on morphological or kinematic signatures . to separate the observed surface brightness in ngc 2950 into secondary and primary bars components
, @xcite must assume that the patterns indeed decouple at the inner - bar end , or that the outer pattern is axially symmetric at least within the inner s extent .
there , the transition between the two is inferred from the location of a plateau in the tw surface brightness - weighted position integral as the limits of integration are extended from zero . perhaps more critically ,
as investigated in @xmath3 [ sec : sim1 ] , the direct association of bar length measured in this manner ( or perhaps others ) with pattern extent may introduce error into tw calculations .
furthermore , adaptations of the tw method based on this type of identification are likely to be inapplicable for spiral pattern estimation . in spiral galaxies , not only can identifying transitions between patterns be less clear , but non - axisymmetric motions are significantly smaller ( e.g. @xcite ) than in bars ( at least those typically analyzed with the tw method ) .
indeed , the tw method has yet to be tested on a simulated spiral .
a recent modification to the tw calculation ( merrifield , rand & meidt 2006 ; hereafter mrm ) in which @xmath1 is allowed radial variation promises to be an invaluable resource for tests of long - lived density wave theories and for understanding the connection , if any , between bar and spiral pattern speeds . like the tw method ,
the so - called radial tw ( twr ) method uses measurements of observables to extract radially varying pattern speeds .
as first applied using the bima survey of nearby galaxies co observations of the grand - design sb galaxy ngc 1068 ( mrm ) , the twr method returned a spiral pattern speed solution that declines with radius , allowing a winding time for the pattern to be estimated ( e.g. mrm ) .
as described in mrm , the nature of the discretized calculation presents numerical solutions that are highly susceptible to fluctuations as a result of compounded noise in the data . in this paper ( @xmath3 [ sec : regularization ] )
, we develop the twr method with regularization as a means of smoothing intrinsically noisy solutions as well as testing model solutions of different radial dependence ( described in @xmath3 [ sec : smooth ] ) .
these and other commendations notwithstanding , with regularization one may risk introducing an unrealistic prejudice to twr solutions . in @xmath3 [ sec : bias ] , we address a means of identifying when this is likely to occur , and in @xmath3 [ sec : fourier ] describe a scheme for minimizing such regularization - induced bias . using evidence that arises from these considerations , as well as other a priori information , theoretically and observationally motivated models for @xmath4 can be developed which constrain the number and extent of patterns present in a disk .
once solutions for these models are calculated , the goodness of each must be assessed in order to identify the best - fit solution . in @xmath3 [ sec : weight ] , we outline the criteria with which the models are judged and describe our concept of error evaluation in the final solutions . beginning in @xmath3 [ sec : nbody ] ,
we analyze three simulated galaxies with known pattern speeds ( a barred spiral , a slowly winding spiral , and a double barred spiral , marking the first application of the tw method to simulated spiral patterns ) in order to develop a general stratagem for application to real galaxies . as applied to these simulations , we find that the twr method is able to extract multiple pattern speeds with accuracies on the order of ( and , as we will see in some cases , better than ) the traditional tw method . with the barred spiral simulation in @xmath3 [ sec : sim1 ]
we show that twr bar pattern speed measurement presents an improvement over traditional tw bar estimates , particularly when there is evidence of a significant contribution from the spiral pattern to the tw integrals .
we find that the regularized twr method can recover information from both patterns effectively by identifying and treating the bar - to - spiral transition radius ( which the tw values themselves may not indicate ) as a free parameter in the calculation . in @xmath3 [ sec
: ierror ] , we analyze the results of the method in detail , particularly with regard to morphological limitations .
we compare our twr results with tw estimates in @xmath3 [ sec : twtwr ] and examine the influence of systematic errors due to the assumed disk position angle and inclination ( shown to be crucial for tw estimates in @xcite ) on both in @xmath3 [ sec : err ] .
we also explore the reliability of a fixed parameterization for the bar - to - spiral transition radius .
( @xmath3 [ sec : trans ] ) . in the last third of the paper ,
we investigate the prospects for extracting spiral pattern speed solutions that are winding in nature ( @xmath3 [ sec : sim2 ] ) , marking the first application of the tw method to a spiral simulation , and in @xmath3 [ sec : sim3 ] we address the use of the twr method for the purposes of parameterizing an independently rotating nuclear bar in a double barred simulation .
there , the techniques we employ for decoupling and extracting measurements of the pattern speeds of both the primary and secondary bar components may present an interesting corollary to recent attempts with the tw method to measure secondary bar pattern speeds in the presence of a strong primary bar pattern .
we note here that the twr method is a generalized version of the procedure used on ngc 2950 ( see @xcite ) ; separating the surface brightness into two components can be thought of as the coarsest version of the discretization that is the back - bone of numerical twr solutions .
based on our experience with these simulations , we conclude with comments on the applicability of the method to observations of real galaxies in @xmath3 [ sec : caveats ] where we also outline a general prescription for using the twr calculation with regularization .
by proceeding under the aforementioned assumptions of tremaine @xmath5 weinberg ( 1984 ) , but allowing that @xmath1 may possess spatial variation in the radial direction whereby the surface density of the chosen tracer can be written @xmath6 , with appropriate mathematical generalizations , the derivation and measurement of the pattern speed @xmath4 can be made from observable intensities and kinematics of a chosen tracer . following the derivation given in mrm , integrating the continuity equation obeyed by the tracer ( with the replacement @xmath7 ) @xmath8 over @xmath9 and @xmath10 ( thereby eliminating the unobservable @xmath11 and the spatial derivative @xmath12 ) , and changing from cartesian to polar coordinates yields @xmath13 a final integration with respect to @xmath14 results in a volterra integral equation of the first kind for @xmath4 : @xmath15r\right\}\omega_p(r)dr=\int_{-\infty}^\infty \sigma v_y dx \label{eq : volt}\ ] ] where @xmath16 .
note that with constant @xmath1 in equation ( [ eq : volt ] ) , we arrive at the regular tw result @xmath17 which , with normalization by @xmath18 ( e.g. @xcite ) , leads to @xmath19 where @xmath20@xmath21@xmath22=@xmath23 and @xmath20@xmath9@xmath22=@xmath24 . for a galaxy projected onto the sky plane with inclination @xmath25 ( so to distinguish from index @xmath26 ) , both the kernel on the left and the integral on the right of equation ( [ eq : volt ] ) are observationally determined quantities with @xmath27 , @xmath28 , and @xmath29 , where @xmath30 and @xmath31 are the coordinates in the plane of the sky along the major and minor axes , respectively , and @xmath32 is the observed line - of - sight velocity .
solutions can be extracted numerically by replacing the integral on the left with a discrete quadrature for different values of @xmath33 and @xmath34 ( see figure [ fig - quad ] ) whereby equation ( [ eq : volt ] ) is converted to @xmath35 or to a matrix equation of the form : @xmath36 with @xmath37 an upper triangular @xmath38 square matrix . note that numerical quadratures on either side of the galaxy ( @xmath10 @xmath200 or or @xmath10 @xmath220 ) occur independently , providing two measures of @xmath4 .
furthermore , as governed by the information available , the slices which delimit the quadrature on a single side need not be uniformly spaced . in this case
, solutions inherit a variable bin width @xmath39 .
also , the calculation allows for no azimuthal dependence for the pattern speed , which we assume throughout .
the size of @xmath37 depends on the desired coarseness or fineness of the quadrature ; the separation between slices at positions @xmath40 ( limited by either the resolution or the sampling of the data ) translates into a radial bin width ( modulo @xmath41 ) via equation ( [ eq : twr ] ) .
the quadrature , perhaps more critically , depends on the limits of integration in equation ( [ eq : volt ] ) .
these limits @xmath42 should be chosen based on where the integrals have converged . while in the case of a single bar pattern integrating past the structure of interest is often suitable , as shown in @xcite ( figures 9 - 11 ) , in the presence of strong , extended asymmetry tw values are highly dependent on the extent of integration along each slice @xmath26 . in cases where multiple patterns exist in a single disk , then
, it is equally favorable ( and hopefully sufficient ) to extend all integrals to the edge of the surface brightness distribution .
meeting the requirement of integral convergence in this manner as applied to the twr calculation determines the location of the last radial bin @xmath43 associated with elements @xmath44 along each slice . for a given radial bin width , with the requirement that @xmath43 equals @xmath45 we are presented with the size of @xmath37 as well as the outermost slice position , since @xmath43 must also equal @xmath46 .
one should check to see that @xmath47 , the last entry in @xmath37 associated with the outermost slice , is associated with a fully converged @xmath48 ( which achieves convergence at least by the map boundary ) . since @xmath37 is an upper triangular matrix ,
the @xmath49 can be solved for via simple back - substitution . in this way
, solutions are generated from the outermost to the innermost radius ( from light gray to black in figure [ fig - quad ] ) according to @xmath50 with @xmath51 ( from eq .
[ eq : twr ] ) .
such solutions , however , are especially susceptible to wildly oscillatory behavior as errors from large radii propagate inward .
these compounded errors arise from uncertainties in the velocities and intensities ( which get translated into the @xmath52 ) and can be particularly severe since the outermost bins often cover the lowest signal - to - noise regions in the data .
the errors thus introduced can be systematic .
figure [ fig - unregularized ] shows an example of ( unregularized ) twr solutions for two different binnings using data from the barred spiral simulation of @xmath3 [ sec : sim1 ] . by increasing the bin width the largest oscillations
are reduced , but ( as will become clearer ) even tripling the bin width will not necessarily limit the propagation of noise to the level required for the extraction of realistic solutions .
( a more thorough discussion of the implications for twr solutions will be deferred until the beginning of @xmath3 [ sec : sim1 ] . )
though the initial application of the twr method on a real galaxy , namely ngc 1068 ( mrm ) , showed little of the oscillatory behavior common to noisy , discretized volterra - type solutions ( outside @[email protected] kpc , anyway ) , the solutions were generated over relatively few bins ( only five at most , over the region @xmath53=1.5 to 2.8 kpc ) .
in general , while large bin widths can often minimize the propagation of noise in the calculation , they can be expected to compromise solutions , as will be discussed in the sections to follow ; naturally , the smaller the quadrature element , the more accurately the true radial variation of the pattern speed can be ascertained .
this is particularly critical as applied to disks sustaining multiple patterns where , as described later , a certain degree of radial precision is required for accurate separation of the pattern speeds . barring a large , limiting resolution
, one should expect to be able to perform a sufficiently smooth quadrature wherein the number of elements in @xmath37 becomes large .
since more elements in @xmath37 ( and more bins over which to generate solutions ) result in intrinsically noisy behavior an effect most pronounced in the inner - most bins gaining a finer , more accurate quadrature often means forfeiting control of the solution . by combining regularization with the twr calculation to force a smooth solution , however
, one can counter this effect while maintaining the required precision .
regularization also serves to alleviate the impact of non - global features that are most likely not included in the overall pattern ( and which can singularly introduce large errors into the integrals ) . since rapid fluctuations in @xmath4
are penalized , discrepant points need not be avoided or ignored ( as demanded in performing the twr calculation on ngc 1068 in mrm ) . while one may also use the alternative , which would be to fit models of @xmath1 directly to equation ( [ eq : twr ] ) and perform a grid - search to find the best model form and coefficients , we pursue regularization here , its speed making it preferred .
our procedure entails the following . as a modification to the @xmath55 estimator minimized by solutions @xmath49 of equation ( [ eq : twr ] ) , namely @xmath56 with implicit sum over @xmath26 ( and @xmath57 ) and errors @xmath58 representing the measurement error of the @xmath59 data point @xmath60
, we introduce a regularizing operator , or smoothing functional @xmath61 , containing a priori information in the manner of tikhonov - miller regularization ( @xcite ; @xcite ) in which ( in matrix form ) solutions @xmath62 minimize @xmath63 here , the elements of @xmath64 and @xmath65 are @xmath66 and @xmath67 , respectively , and the role of @xmath68controlling the relative amount of @xmath55 minimization on the left to entropy maximization on the right is explicit
. reduced to a linear set of normal equations , this minimization returns smoothed solutions according to a modified version of equation ( [ eq : twr ] ) : @xmath69 note that the regularizing functional , not necessarily upper triangular , introduces an anticipatory quality to solutions @xmath49 whereby all bins at the same radius are coupled . furthermore , solving for components @xmath49 no longer only involves a procedure like back - substitution , but requires rather an l - u decomposition ( for instance ) as well .
the real power in applying regularization to twr calculations is in the freedom to choose how the smoothness of solutions is achieved . for the purposes of distinguishing between different possible radial dependences for @xmath70
, we choose @xmath61 to reflect a priori assumptions based on simple expectations from theory and observation .
model solutions , then , each incorporating its own @xmath61 , represent smoothed , testable realizations of the pattern speed .
these we restrict to simple forms in order to minimize the additional amount of information to be extracted from the data relative to the traditional tw method .
for polynomial solutions , we consider only constant , linear and quadratic radial dependence .
the elements of the smoothing @xmath61 are associated with the minimization of the @xmath71 derivative of @xmath70 for each polynomial solution of order @xmath72 .
for instance , for linear solutions this entails minimizing @xmath73 whereupon @xmath74 one may also choose a form for @xmath61 that identifies two or more distinct regions of independent radial behavior by invoking a step - function model . for the case of a barred spiral with a constant bar and quadratic spiral , for instance
, this corresponds to minimizing @xmath75 the elements of @xmath61 with @xmath76 reflect the a priori assumption that the bar pattern speed is constant , while those for @xmath77 associate a quadratically varying pattern speed with the spiral .
the index @xmath78 , a free parameter , locates the radial bin where the transition between the two patterns occurs .
obviously , the number of available bins constrains the order of the polynomial in a given radial zone .
once we have chosen @xmath61 , we initially choose @xmath68 to reflect comparable amounts of @xmath55 minimization and regularization by letting @xmath79 .
since we are in the business of generating solutions based on particular models , @xmath68 is modified to arrive at the regularization required to return solutions of a given type .
this modification generally consists of an increase in @xmath68 over @xmath80 .
consider how the regularizing parameter @xmath68 regulates the degree of smoothness of the solution to the weight placed on the data : with @xmath68=0 , equation ( [ eq : tichonov ] ) corresponds to @xmath55 minimization ( and becomes an unbiased estimator with the smallest variance ) , however yielding highly oscillatory solutions , while @xmath68@xmath81@xmath82 corresponds to a maximally smooth estimator with non - vanishing variance .
fitting data sets with different spatial coverage will change the effect of @xmath68 on the solution ( e.g. larger bins require less regularization ) . the most appropriate choice for @xmath68 ( and @xmath61 ) should be made on a galaxy - by - galaxy basis , according to the quality of information to be extracted from observations .
by imposing assumptions about the smoothness of the pattern speed , regularization inevitably introduces complications for extracting realistic solutions . to understand how these arise ,
consider the solution for a barred spiral galaxy .
the nature of the calculation ( from out to in ) has implications for the accuracy of the bar estimates , in particular .
not only do the @xmath49 for bins covering the bar rely on the greatest number of matrix elements ( and errors therein ) , but the bar estimate depends critically on the solution for the spiral and all outer bins via equation ( [ eq : outin ] ) .
consequently , merely requiring the pattern speed in the outer bins to be constant with regularization out to the edge of the surface density ( for instance)effectively removing fluctuations that might better fit the data will have consequences for the bar solutions .
so while the regularization is particularly fast and effective for tests for the radial behavior of patterns , it can also hinder the realization of accurate solutions . for the simulations studied here
, the risk of regularization - induced bias is inherited from the adopted quadrature .
recall our requirement that all slices cover the full extent of the ` emission'so as to insure all integrals be fully converged and relatedly , that the last matrix element governs the outermost slice position . in the barred spiral simulation of @xmath3 [ sec : sim1 ] , for example , such an extensive quadrature presents us with outermost slices that pass through a region where there is simply no discernible pattern ( as indicated by the surface brightness distribution and its fourier decomposition ; see next section ) . while these slices themselves do not provide direct estimates of the patterns of interest , the corresponding bin values are necessary for calculating the bar and spiral solutions .
moreover , the quality of these solutions will be intimately related to the treatment of the outermost bins .
we therefore find that identifying , and reducing the influence of , the compromised zone by not enforcing regularization on these bins to be essential for accurate pattern speed measurement . with the above concerns in mind , we have tested and used the following scheme .
given slices that pass through an outer region which either contains little information from a strong pattern , is suspected of sustaining multiple patterns , or displays only faint emission , we choose in such a case to let the values in the outermost bins be calculated without regularization with the restriction , only , that they minimize the @xmath55 .
once a particular bin at @xmath83 , the cut radius has been reached , regularization is imposed with all remaining inward bins generated accordingly . in our procedure
this corresponds to using an @xmath61 indexed by the cut bin @xmath84 that is lowest - block zero .
and the ` cut ' radius identifies the location in the disk where the outermost discernible pattern ends .
note that this is in contrast to altogether ignoring the outer portion of the disk .
we prefer this procedure for two reasons : 1 ) as qualified later in @xmath3 [ sec : twtwr ] , when each integral is truncated within the disk , the quadrature is at greater risk of ignoring information critical for characterizing the patterns uniformly throughout the disk and 2 ) like the transition radius @xmath85 , we can easily incorporate @xmath83 as a free , though restricted , parameter in our models . in practice , extracting the bar and spiral pattern speeds for the simulated barred spiral in @xmath3 [ sec : sim1 ] involves generating a group of solutions with various bar - to - spiral pattern transition radii for a given ` cut ' location . throughout the analysis
, we choose the cut radius to reflect a priori knowledge of the outermost measurable pattern s termination radius estimated from the surface brightness and its fourier decomposition .
when referred to , the power in each fourier component , or mode @xmath86 , is given by the norm of the complex fourier amplitude @xmath87 where @xmath88 is the angular coordinate of each of the @xmath45 particles at each measured radius .
for the barred spiral in @xmath3 [ sec : sim1 ] , for example , we combine evidence from the surface density where beyond the bar there is enhanced spiral surface density from only a limited radial zone with the fourier spectrum to identify a region in the disk outside the spiral which is susceptible to regularization - induced bias .
specifically , the fourier power spectrum ( figure [ fig - bspowersem ] ) shows that at @[email protected] the second clean hump in @xmath86=2 power decreases to almost zero , marking the end of the spiral .
past this radius , the ( strong ) @xmath86=2 component ( between @[email protected] - 8.0 ) is not associated with visible spiral structure ( see figure [ fig - simi ] ) . reckoning this outer zone to be incompatible with a simple pattern speed model , then
, we consider only the inward bar and primary spiral pattern speeds to be measurable with regularization .
figure 19 of @xcite for this simulation ( figure [ fig - d06 ] in this paper ) confirms this ; not only does the spiral pattern terminate at @[email protected] , but beyond this radius the pattern speed is multi - valued ( this , of course , would be indiscernible in a real galaxy ) . in this case , imposing form with regularization on bins of suspect quality and behavior outside the spiral will likely impair the solution of interest .
we therefore restrict the cut bin for the barred spiral simulation to 4.5@xmath20@[email protected] , representative of where the primary spiral pattern terminates in the disk .
as mentioned in @xmath3 [ sec : sim1 ] , this step is substantiated by our finding that a cut radius of @xmath83=4.8 is one of several @xmath55 minima given a range of possible cut radii . and
furthermore , solutions generated in this manner are judged to overall provide a considerably better fit to the data than solutions where regularization is imposed out to the edge of the surface brightness ( according to the scheme described in the next section ) .
given the data , we simultaneously generate model solutions with different radial dependences for direct comparison for each side ( y@xmath220 or y@xmath200 ) independently .
we then average the two like - model solutions together to construct a global solution .
( in those instances when model solutions include a cut bin , the averaging occurs over the regularized part of the solutions only , in order to maintain the ` unbiased ' quality of the unregularized part of the solutions for each side of the galaxy . )
note that while the assumption that the patterns are indeed global is not overly inspired for the simulated galaxies studied here , putting this into practice on a real galaxy requires that the assumed galaxy kinematic parameters are accurate and that such symmetry exists . with each global model solution we generate a complete set of @xmath20@xmath21@xmath22 using equation ( [ eq : twr ] ) .
we then judge each model through the @xmath89 ( @xmath55 per degree of freedom ) goodness - of - fit estimator of the reproduced to actual @xmath20@xmath21@xmath90 given measurement errors @xmath91 , in keeping with the standard tw analysis .
note that with this choice the @xmath58 in equation ( [ eq : chisq ] ) are related to the errors @xmath91 by @xmath92 .
that is , the calculation fits to the @xmath60 given errors @xmath58 while our @xmath55 estimator considers the differences from @xmath20@xmath21@xmath93 given errors @xmath94 .
while for a real galaxy inaccuracies in the assumed position angle ( pa ) have the largest potential for introducing errors into @xmath20@xmath21@xmath93 , we prefer that the measurement errors @xmath95 reflect random noise in the data , only .
( systematic errors prove more practically assessed through direct tests of the sensitivity of the results to departures from the nominal values for the pa or inclination , for instance . ) for the simulations studied here , then , we obtain errors @xmath91 under the assumption that the inverse mirror image of each @xmath20@xmath21@xmath93 on one side of the galaxy should be the same as on the other side ( i.e. the patterns are symmetric ) . we then assign a global error @xmath95 to each slice where @xmath95=@xmath96/2n ( and @xmath45 is the number of bins / slices used in the twr calculation on a single side ) . in practice ,
the simplest @xmath97 weighting schemes are either uniform weighting for all slices or weighting by the intensity , which should give more weight to slices where the signal is strongest .
we have chosen the former since we are interested in @xmath1 over a broad radial range , and prefer that our result is not dominated by just the slices with the highest signals ( which can vary dramatically ) .
furthermore , the choice of assigning an identical error @xmath95 to each slice carries with it an implicit weighting scheme for equation ( [ eq : tichonov ] ) .
for an exponential surface brightness profile , for example , slices on either side of the galaxy will have progressively smaller @xmath18 as @xmath98 increases .
this corresponds to errors @xmath58 proportional to @xmath18 , then , that grow larger from out to in . since in most cases the uniform weighting scheme will be in actuality most restrictive of the outer bins the goodness of which will affect the solution inward this choice is particularly well suited for the twr calculation
. given our choice of weighting scheme , there are two important considerations which demand that we calculate the @xmath55 over all slices .
first , the inner - most bins contribute to the @xmath20@xmath21@xmath22 in only the inner most slices and hence contribute relatively negligibly to the @xmath55 , despite possibly larger weights @xmath52 ( reflective in part of a surface brightness that is centrally peaked , say ) than bins at larger radius .
this is especially true when an inner pattern appears only over a small fraction of the total bins , and the reproduced @xmath20@xmath21@xmath22 of even those slices that pass directly through the inner pattern still rely ( perhaps predominantly ) on the solution out to the largest radial bin .
however , since the goodness of the inner bins is directly related to the goodness of the bins at larger radii , by considering all slices we more effectively judge the whole solution .
secondly , in cases when some number of outermost bins are calculated without regularization , it is critical to account for the ( largely positive ) effect that these bins have on the solution inward , especially from model to model . in our current scheme , the values of the ` unregularized ' bins are not quite identical to those in the completely unregularized solution since the quantity that they minimize still includes participation from non - zero elements in the smoothing functional @xmath61 and what is currently a model - dependent @xmath68 .
that is , slight variation in the values of the ` unregularized ' bins from model to model is apparent , and we can not ignore the minor differences this introduces to the regularized part of the solution . though this is a minor effect , by considering the @xmath20@xmath21@xmath22 of all slices in the @xmath55 we prevent solution - preference based on the ` unregularized ' bins ( which act essentially as stand - ins ) from being introduced .
the first consideration above also prevents us from calculating judicious error bars on the solutions according to the variation of individual model parameters over a typical @xmath55 confidence interval ; in practice , the value of an inner pattern with minimal radial extent can change considerably with little effect on the @xmath55 . indeed
, we find that the errors generated according to such a prescription are unrepresentative of the goodness of the solutions as returned by the calculation . in @xmath3 [ sec : ierror ] , we describe the dependence of both inner and outer speeds on the location assigned to the transition between the two .
an obvious progression for future applications of the twr method would be an exploration of the covariance of what we consider here , to first order , ` free parameters ' , especially for the purposes of improved error estimation .
presently , however , we construct error bars for model solutions by considering the range of parameters in the best solutions at different assumed projections . as will pertain to the sections that follow , by considering an overall solution in this manner we can fairly account for the uncertainty introduced for real galaxies by the reality that each can be sampled at only a single pa .
in order to establish guidelines for applying the regularized calculation to observations of real galaxies , we next perform tests of the method on simulations with known pattern speeds .
each case invokes unique models for @xmath4 which we motivate and discuss in detail .
the procedure for engaging the method with maximum accuracy then follows from careful examination of the quality of solutions given the available information .
though quite detailed , these individual studies together constrain general scenarios and practices to extrapolate onto observations of similarly structured , real galaxies .
we use three simulations in this study .
the first , which we refer to here as simulation i , constitutes a barred galaxy with spiral structure .
originally presented in @xcite ( hereafter d06 ) where it is referred to as run l2.t12 , it consists of a live disk immersed in a rigid halo . a complete description of the model parameters are contained in d06 .
the second ( spiral ) simulation , which we refer to here as simulation ii , is unpublished .
it was designed with the main aim of generating strong spiral structure using the groove mode mechanism of @xcite and @xcite in which dynamical instability develops from a groove , or narrow feature , in the phase - space density at a particular angular momentum
. a trailing spiral wave is generated , and at the lindblad resonances of the wave , further grooves develop such that the instability is recurrent .
like simulation i , it consists of a rigid halo and live disk , but also includes a live bulge component .
the bulge constitutes @xmath99 of the baryonic mass and is sufficiently concentrated that a bar is very slow in forming .
the disk has toomre-@xmath100 ; in order that a strong spiral was seeded , @xmath101 of disk particles in a narrow angular momentum range were removed leaving 4@xmath10210@xmath103 - 169480 particles including the bulge .
the result , as can be seen in figure [ fig - sim2 ] , is the formation of a strong but transient spiral .
the last simulation in this paper , simulation iii , is a double - barred galaxy generated using the method of @xcite .
this high - resolution simulation consists of live disk and bulge components in a rigid halo potential .
the model has @xmath104 million equal mass particles , with @xmath105 million in the disk and @xmath106 million in the bulge such that the bulge has mass @xmath107 , where @xmath108 is the disk mass .
the initial toomre-@xmath109 of the disk is @xmath110 .
the formation of the secondary bar is induced by making the bulge rotate ( to mimic a pseudobulge ) .
more details of the simulation can be found in @xcite where it is referred to as run d. as in d06 , all lengths and velocities are here presented in natural units .
we analyze a snapshot of each simulation at a single time step with the disk in the @xmath111 plane . by rotating the system about the @xmath112 axis
we assign a line - of - sight direction to establish the kinematical major axis .
another rotation about the @xmath9 axis gives the system an inclination @xmath25 ( chosen throughout at @xmath25=45@xmath113 , unless otherwise specified ) .
the snapshot is then projected onto the sky - plane where @xmath30=@xmath9 and @xmath31=@xmath114 . for a given slice spacing ,
the slices along which the calculations occur are aligned perpendicular to the line - of - sight direction ( parallel to the kinematical major axis ) .
the orientation of these slices , which is identical to the disk pa in a real observation , is designated uniquely the slice angle ( sa ) in the studies that follow . the bar and spiral structure in this simulation , first presented in d06 ,
is featured out to @[email protected] , clear in the surface density ( figure [ fig - simi ] ) and its fourier decomposition ( figure [ fig - bspowersem ] ) . beyond @[email protected] ,
the fourier decomposition indicates the possible presence of a third pattern . with step - models for @xmath4 ,
then , we might reasonably extract pattern speeds for three distinct structures .
however , the @xmath86=2 mode between 5.0@xmath20@[email protected] is not associated with a strong surface density enhancement . and
as remarked upon in @xmath3 [ sec : fourier ] , figure 19 in d06 reproduced here in figure [ fig - d06 ] shows that the pattern in this radial zone is maintained by multiple distinct pattern speeds .
( note that a real galaxy would not be disposed to the analysis provided with this type of plot .
it is available for this simulation , only , and we include it here for the sake of comparison ) . in a clear account of regularization - induced bias , test solutions based on a three - pattern speed model have considerably larger @xmath55 than those parameterizing a bar and single spiral .
we therefore reject models with a third pattern speed and use our ` cut ' scheme where solutions are generated without regularization up to a cut radius @xmath83 which parameterizes the end of the primary spiral pattern .
the actual pattern speeds of the bar and spiral structure to be reproduced by our solutions are @xmath115=0.29 for the bar , and a constant @xmath116=0.18 for the spiral , as estimated from figure [ fig - d06 ] ( from which we also estimate a bar - to - spiral pattern transition radius @[email protected] ) . in our models of @xmath4
, we express the a priori assumption that the bar pattern speed is constant by using an @xmath61 like that in equation ( [ eq : sdiv ] ) but where , for @xmath77 , @xmath61 is reflective of either a constant , linear , or quadratically varying spiral pattern speed .
solutions with spirals of order 0 , 1 , and 2 , then , have a total of 4 , 5 , and 6 degrees of freedom , respectively . the free parameter @xmath78 we restrict for all models such that 1.8@xmath20@[email protected] , according to bar - length estimates from figure [ fig - simi ] ( @[email protected] ) and figure [ fig - bspowersem ] . for @xmath39=0.3 bins , we require a total of 71 slices ( 35 on each side ) to reach the edge of the surface brightness at @[email protected] .
this places the cut bin between 15@xmath20@xmath84@xmath2020 , according to the previously motivated restriction 4.5@xmath20@[email protected] estimated from figures [ fig - simi ] and [ fig - bspowersem ] . in light of the discussion in @xmath3 [ sec : weight ] , we construct errors for our estimates to reflect the expected accuracy of twr bar and spiral solutions given a particular observational scenario . specifically , we perform the calculation for a range of sas spanning the upper half - plane of the galaxy ( quadrants i and ii ) , namely @xmath11815@xmath113 , @xmath11845@xmath113 , and @xmath11875@xmath113 .
each sa corresponds to a unique disk pa .
the resultant bar and spiral estimates generated using the @xmath39=0.3 bin width ( to be discussed at length in the following sections ) are listed in table [ tab - bsall ] and the average and rms of the best - fit solutions for this sa range are shown in figure [ fig - bssol ] .
there , horizontal error bars represent the dispersion in @xmath85 and @xmath83 in the solutions based on variations from sa to sa for the @xmath39=0.3 radial bin width .
rewardingly , the best - fit solutions are quite accurate ; the comprehensive spiral and bar estimates in figure [ fig - bssol ] are 6.7% and 8.3% from their actual values ( with @xmath548% error in @xmath115 but a slightly larger error of @xmath5418% in @xmath116 ) .
so , too , are the determinations of @xmath85 and @xmath83 , according to figures [ fig - bspowersem ] and [ fig - d06 ] .
furthermore , our solutions correctly reproduce the functional form of the spiral pattern speed . at all
sas , out of all solutions with @xmath85 and @xmath83 within their restricted ranges , the lowest reduced @xmath55 solution corresponds to a constant spiral .
figure [ fig - vslice ] shows a comparison between the actual and best - reproduced @xmath20@xmath21@xmath22 at each slice position @xmath40 used in the calculation for the 15@xmath113 slice angle .
the @xmath20@xmath21@xmath22 reproduced by the best - fit constant bar and constant spiral solutions are shown in the left panel , while those from the optimum ( lowest-@xmath119 ) solution with a quadratic spiral are plotted in the right . the close reproduction of the actual pattern speeds by the solutions in figure [ fig - bssol ] occasions further evaluation of strictly unregularized twr calculations .
consider the values in radial bins inside @[email protected] in typical unregularized solutions for this simulation ( figure [ fig - unregularized ] ) .
that there is little to no indication of @xmath115=0.29 in the left plot is perhaps not surprising : the large number of bins which accompany the choice of the small bin width would seem to guarantee a high level of noise propagated throughout the solution .
however , in the slightly more stable solution with the wider bin width , the inner bins are still unrepresentative of the actual pattern speed in this zone
. we can understand this as a systematic error introduced by the noise which not only propagates but also compounds as the full solution assembles from the outermost to the innermost radius ; @xmath4 in a given bin reflects errors from all exterior radial bins making the value in that bin more likely far removed from the actual value .
rcccc + sa&*@xmath115*&*@xmath116*&@xmath85&@xmath83 + 75@xmath113&0.325&0.149&3.0&5.7 + 45@xmath113&0.327&0.111&2.4&4.8 + 15@xmath113&0.343&0.184&2.4&4.5 + -15@xmath113&0.269&0.172&2.7&4.8 + -45@xmath113&0.313&0.199&3.0&5.4 + -75@xmath113&0.303&0.203&2.4&6.0 + -&0.29&0.18&2.5&5.0 + the use of simulations which can be studied at multiple projections provides us with perhaps the most critical assessment for the accuracy of twr solutions .
figure [ fig - bssol ] suggests that the twr method should perform well for any given viewing angle .
however , though still quite small , the rms in each estimate is largely reflective of the non - trivial effect that the orientation of the pattern with respect to the slice angle used in the calculation can have on solutions .
we can understand the origin of the differences in solutions for the range of sas in table [ tab - bsall ] by considering the impact of the limited azimuthal range of the bright spiral enhancement ( clear from the surface brightness distribution in figure [ fig - simi ] where the spiral extends almost perpendicular to the bar major axis ) .
that is , at all slice angles the quadrature accumulates fragmentary information from the spiral since only some of the slices that cross the full radial zone of the spiral pattern intersect the strong spiral structure . but whereas both the bar and spiral estimates seem to suffer at sas in quadrant ii ( i.e. positive sas ) , our measurements of @xmath116 in quadrant i are quite accurate . according to the morphology , in quadrant ii it appears that the limited sampling of the spiral asymmetry implicit in slices other than those that also pass through the bar entails slightly less accurate spiral estimation . to interpret the distinction between solutions from the two quadrants ,
consider the combined influence of regularization and our chosen weighting scheme . specifically , since the regularizing @xmath61 induces the coupling of all bins within the same radial zone , even when the spiral - zone crossing slices do not intersect the strong spiral enhancement , the bins there inherit information from bins at the same radius from slices which do intersect the arms .
according to our weighting scheme , however , this coupling is not uniform ; the degree of support at each azimuth is influenced by the measurement errors for each slice which grow larger from out to in . as a result
, @xmath116 is best constrained by information from slices passing through the outer radial zone of the spiral alone ( not those passing through the bar ) .
so when in quadrant ii the spiral asymmetry appears in only the inner slices , @xmath116 is less precise then when these outer slices clearly intersect the spiral arms .
according to equation ( [ eq : outin ] ) , since the bar estimate is directly related to that of the spiral , the result for these sas is error in both @xmath116 and @xmath115 .
the corresponding determination for the radial domain of the bar pattern , on the other hand , is not as obviously sensitive to this issue .
not only is the bar - end reasonably well defined in both quadrants , but information from the bar which contributes to the parameterization of @xmath85 is reinforced with regularization in the manner described above .
the spiral termination radius , too , seems fairly consistent from sa to sa .
but since the asymmetry is weaker ( and there is less information ) at that location in the disk , we find that this parameter requires the most restriction ( indeed , our determinations of @xmath83 completely span the allowed range ) .
overall , then , our determinations for @xmath85 and @xmath83 are stable and accurate , each with less than 11% error , even in quadrant ii .
nevertheless , since the pattern speed estimates from sas in quadrant i seem to comparatively benefit from the high quality of information from both patterns , we conclude that position angles which provide the most uniform slice coverage of all patterns in the disk are preferred . the largest disparity between step - model solutions from various slice angles can largely be attributed to limitations in determining the location of the transition between the two patterns .
that is , solutions are affected by the finite bin width inherent to the numerical calculation ; slight incompatibility between the actual transition and that to which the solution is limited ( given the bin width ) can result in errors around 10% .
this is a more pervasive effect than morphology alone and more obvious with the use of a slightly larger bin width than @xmath39=0.3 .
nonetheless , we can make several informed inferences about the result of the finite bin width by considering the nature of the twr calculation .
specifically , since the transition determines the contributions of inner and outer patterns to the integral on the right hand side of equations ( [ eq : twint ] ) and ( [ eq : twr ] ) through the matrix elements @xmath52 , a mismatch between the transition bin and the actual transition radius will corrupt the separation of the contributions of the two patterns .
note that this is precisely the source of the covariance between model parameters intimated in @xmath3 [ sec : weight ] . to understand the effects of ( minor ) radial pattern misassignment ,
consider a step - function solution that parameterizes the extent of a bar pattern speed along with that of a spiral .
when the transition bin between the two patterns underestimates the actual transition by a fraction of the bin width , for instance , we would expect the numerical calculation to effectively subtract off from @xmath60 a contribution from the ( lower ) outer pattern where a higher pattern actually exists ( e.g. eq.[[eq : outin ] ] ) , slightly raising the value of the inner pattern speed .
conversely , we expect an overestimation of the transition to result in a slightly lower inner pattern speed .
more subtle effects can occur , however , depending on the geometry of the patterns , as illustrated for this simulation in @xmath3 [ sec : trans ] . for a range of sas
, we emphasize that the resultant errors in possible combination with an undersampled transition are minimal , as long as the bin size is sufficiently small , and are not the result of vastly different transitions for each sa ; we find that the transition between patterns in the best solutions is generally relatively stable with changes in slice angle .
this may be unexpected from the perspective of the traditional tw method since the projected length presented to the slices by the bar will depend on the relative orientation of the slices with its major axis ( and slice - orientation errors tend to be large , as detailed next for this simulation ) . in light of the above results ,
we next examine the improvements available to pattern speed estimation using the twr calculation relative to the tw method .
we specifically compare the bar pattern speed estimates arrived at using the twr method with those using fully extended tw integrals .
though tw estimates of @xmath115 in the presence of a secondary structure may also be attempted using truncated integrals that extend to just past the end of the bar ( such that information from the bar alone is dominant ) , we examine the former case for two reasons : 1 ) to compare the two methods under identical conditions ( i.e. using the same data points @xmath20@xmath21@xmath93 ) and 2 ) to study the influence of the relatively weak spiral ( and evaluate the assumption of negligible non - axisymmetric motions beyond the bar ) .
rccc + sa&@xmath45 slices&*@xmath1*&@xmath120 + 75@xmath113&7&0.270&@xmath1180.010 + 45@xmath113&11&0.282&@xmath1180.062 + 15@xmath113&17&0.208&@xmath1180.007 + -15@xmath113&15&0.248&@xmath1180.026 + -45@xmath113&11&0.294&@xmath1180.008 + -75@xmath113&9&0.233&@xmath1180.053 + though we use fully extended integrals to perform the tw calculation even when making estimates of the bar pattern speed the innermost slices clearly supply evidence for a bar pattern speed that is distinct from that of the other structure in the disk .
figure [ fig - tw45 ] shows a typical plot of @xmath20@xmath21@xmath22 vs. @xmath20@xmath9@xmath22 for this simulation where the inner eleven slices are indeed best fit by a steeper slope than for all slices . for all of the other sas studied in the previous section
, we measure @xmath1 with a unique number of bar - crossing slices .
this is intended to reproduce an optimal tw observing strategy that makes use of only those slices that intersect the enhanced bar surface density .
that is , since the projected length presented to the slices by the strong bar structure depends on the relative orientation of slices with the major axis of the bar , the number of slices that intersect the bar enhancement varies from sa to sa ( from seven to 17 for this simulation at the six studied sas ) .
table [ tab - bstw2 ] lists this optimal number of slices @xmath45 at each sa along with the corresponding pattern speed estimate .
all entries in the table correspond to slopes of best - fit straight lines ( and the corresponding intrinsic scatter ) in plots of @xmath20@xmath21@xmath22 vs. @xmath20@xmath9@xmath22 .
rcccc + & & + & * @xmath1*&@xmath120&@xmath1&@xmath120 + 75@xmath113&0.234&@xmath1180.004&0.115&@xmath1180.008 + 45@xmath113&0.217&@xmath1180.045&0.094&@xmath1180.018 + 15@xmath113&0.208&@xmath1180.007&0.157&@xmath1180.009 + -15@xmath113&0.216&@xmath1180.016&0.121&@xmath1180.007 + -45@xmath113&0.278&@xmath1180.009&0.166&@xmath1180.009 + -75@xmath113&0.175&@xmath1180.010&0.128&@xmath1180.010 + upon inspection , table [ tab - bstw2 ] seems to suggest that even in the presence of the spiral asymmetry information from the bar is maximal in the bar - crossing tw integrals . despite also reflecting a contribution from the spiral pattern , these bar estimates are fairly accurate ( though presumably not as accurate as would be the case for a strictly sb0 galaxy such as ngc 7079 ) .
however , as in table [ tab - bstw ] , if we extend the slice coverage at each sa out to @[email protected] using the inner 17 slices ( at @xmath122=0.3 spacing)closer to the full extent of the radial zone of the bar pattern , according to figure [ fig - bssol]then the quality of the tw estimates diminishes with the inner , bar pattern speed estimates approaching that from a fit to all slices with @xmath123 @xmath205.1 ; even the true bar - crossing slices contain non - negligible information from the spiral pattern . unlike the tw method , the twr method is not relegated to the use of only those slices where the bar contribution is maximized . in principle , the inner solutions are accessible through the very use of information from beyond the bar ( namely from the zone of the spiral ) , and are improved when this information is radially coupled ( e.g. by regularization ) .
this aspect of the calculation allows for the return of pattern speed solutions without reference to an assumed pattern extent , and provides , moreover , an independent means of determining the radial domain of patterns .
of course , this is not to imply that a deficiency of usable slices in tw estimates prevents accurate pattern speed measurement , or determination of pattern extent , for that matter ; using slices that cross primarily through the enhanced emission from the bar , tw estimates from sb0 galaxies have been successfully used to observationally confirm that bars end at or inside their corotation radii .
however , it does suggest that in the presence of non - negligible asymmetry exterior to the bar , tw bar estimates are susceptible to errors introduced by the use of slices positioned near the bar end which are presumed to reflect the bar pattern speed but in reality include a significant contribution from this structure . consider a typical plot of @xmath20@xmath21@xmath22 vs. @xmath20@xmath9@xmath22 for bar - crossing slices .
the best - fit slope determination for @xmath1 from such a plot is primarily governed by slices with the largest @xmath20@xmath21@xmath22 and @xmath20@xmath9@xmath22 . since in sb0 galaxies @xmath20@xmath21@xmath22 and @xmath20@xmath9@xmath22 approach zero in slices at or near the projected bar end ( since they are presumably too far past the strong bar structure to contain information about the bar and mark , rather , a return to axisymmetry in the disk ) , even when tw estimates consider these ` near zero ' slices , they contribute minimally to estimates for @xmath1 .
( see @xcite where studies with a simulated sb0 include a number of slices sampling the full extent of the bar . )
but when there is considerable asymmetric structure present beyond the end of the bar , similarly positioned slices will reflect this contribution , impairing measurement of the true bar pattern speed . in the barred spiral simulation
, this consequence can be characterized upon inspection of @xmath20@xmath9@xmath22 , @xmath20@xmath21@xmath22 and @xmath124=@xmath20@xmath21@xmath22/@xmath20@xmath9@xmath22 as a function of the limit of integration @xmath125=@xmath126 in a typical slice at @xmath10=1.2 ( see figure [ fig - integrals75b ] for sa=-75@xmath113 ) .
@xmath124 for this slice , which according to our estimate for @xmath117 crosses the outer region of the bar , actually seems more reflective of the spiral pattern speed ( beyond the discontinuity at @[email protected] ) .
indeed , past @[email protected] where @xmath20@xmath9@xmath22 crosses zero we can infer that this bar - crossing slice contains substantial participation from the spiral ; both @xmath20@xmath9@xmath22 and @xmath20@xmath21@xmath22 decrease to a dip between @xmath126=4.0 and 6.0 before reaching a plateau .
this is evidence that , despite the relative weakness and limited extent of the spiral as indicated by the surface brightness distribution , the asymmetry in the disk should not be considered dominated by the bar alone .
moreover , this is a clear indication that the twr calculation , which identifies and effectively removes the spiral contribution from the fully extended @xmath127-weighted velocity integrals , can improve upon traditional tw bar estimates that also use fully extended integrals .
of course , as displayed by figure [ fig - integrals75b ] , integrating only between @xmath128@xmath54 2.5 may alone provide a reasonable bar pattern speed and relieve all other tw slices from the spiral contribution .
however , this would provide only a single pattern speed estimate where two are possible ; using the twr method and fully extended integrals , both @xmath115 _ and _ @xmath116 can be measured .
figure [ fig - integrals75b ] also raises a crucial point related to the required limit of integration along each slice in the twr calculation : if there is a clear plateau in the integrals reached before the edge of the map boundary , why not simply truncate the integrals where they have converged ( common to tw estimates ) rather than use integrals extending to the edge of the surface brightness and which present outermost bins that we demand must be ` cut ' ( that is , calculated without regularization ) , anyway ?
the plateau reached at @[email protected] would suggest that truncating the integral there could suitably account for information from the major sources of asymmetry in the disk .
however , this same distinction is not clearly shared by all slices , especially those at large @xmath121 .
if we associate the plateau in this slice at @xmath10=1.2 with the limit of integration @xmath125=8.2 and hence the total radial extent @xmath129 required of the quadrature where @xmath130 , then this locates the outermost slice position at @xmath131=@xmath132 as well as the extent of the integral along this slice .
it is easy to check that ( at least for this slice orientation ) the integral has not achieved convergence by this point , nor have most other integrals in the disk by @xmath129 . rather than risk ignoring information critical for characterizing the patterns uniformly throughout the disk , then , we choose to include in the quadrature all information out to the edge of the surface density .
indeed , this serves to perform a function similar to truncating tw integrals .
the difference in the two procedures arises from the fact that while the limits of integration for each individual slice can be adjusted for a given structure of interest in the tw calculation , the quadrature in the twr method delineates specific bounds which must encompass complete information from all extended patterns in the disk . in this section
we use the barred spiral simulation to consider the errors introduced to twr pattern speed estimates in real galaxies .
we can expect errors in the assumed @xmath133 to dominate errors in the twr calculation , given that such errors translate significantly to inaccuracies in the traditional tw calculation via the line - of - sight velocity integral which is also , of course , a prominent feature of the radial tw equation .
table [ tab - paerror ] summarizes the results for a standard pa error of @xmath134 on the sas in quadrant i chosen for their advantages , as evidenced by the discussion in @xmath3 [ sec : ierror ] .
the average and rms for bar and spiral estimates from both the twr and the traditional tw methods are listed .
( the tw ` bar ' estimates are obtained by fitting to the inner , nominal number of slices listed in table [ tab - bstw2 ] , while all slices with @[email protected] are considered in the ` spiral ' estimates . )
even this small @xmath135 introduces considerable errors ( relative to the known pattern speeds ) to both types of bar estimates .
these errors in @xmath115 can be many times larger than the formal rms .
but whereas the errors are comparable in the twr and tw bar estimates , the errors in the spiral estimates tend to be smaller with twr than tw .
rewardingly , with this error not only are the twr spiral solutions still definitively constant and accurate to @xmath5415% , but the radial domain of both pattern speeds are still well - determined .
the transition between the bar and spiral @xmath85 and the termination radius of the strong spiral pattern @xmath83 are effectively unchanged from the @xmath136 case ; for both @xmath137 and @xmath138 we find @[email protected] @[email protected] .
rcccc + & & + & * @xmath115*&*@xmath116*&*@xmath115*&*@xmath116 * + tw*&0.179&0.162&0.354&0.115 + & @xmath1180.055&@xmath1180.058&@xmath1180.04&@xmath1180.02 + twr&0.214&0.211&0.380&0.158 + & @xmath1180.031&@xmath1180.045&@xmath1180.042&@xmath1180.023 + besides the effects on @xmath133 measurement as studied by @xcite , galaxy inclination and ellipticity play perhaps more prominent roles as sources of error in twr solutions relative to traditional tw estimates .
presumably , large inclination errors will prevent the association of information into accurate radial bins , given that @xmath139 .
we expect this effect to be minimal at moderate inclinations since @xmath140 , and most significant at small inclinations where one would generally find that the difficulty in inferring in - plane morphology and kinematics makes the the tw method impractical in any case . at a moderate , 45@xmath113 inclination
we find that the barred spiral solutions for @xmath118 3@xmath113 inclination error differ from the actual pattern speeds by only the change in @xmath141 introduced by the line - of - sight velocity .
we have shown that the regularized twr method can be used to parameterize the number and radial domain of multiple pattern speeds in a single disk .
formally , the contribution of each to the line - of - sight velocity integral is established through the designation of a transition between patterns . in our scheme
this transition is a free parameter , but the method , of course , could plausibly assimilate other transition - identification methods to similar effect , much like that in the @xcite adaptation of the tw method . specifically , in @xcite a plateau in the integrals @xmath142 with variation in @xmath126 is associated with the transition from an inner to an outer pattern ( see @xcite for details ) . this transition is then used to separate the disk surface brightness into two unique components ( one for an inner secondary bar , one for an outer primary bar ) , thereby governing the decoupling of the pattern speeds . we here pursue this type of diagnostic for the case of the simulated barred spiral in order to test the reliability of employing the twr method with such independent evidence for pattern extent .
figure [ fig - macint ] plots the values of @xmath142 / @xmath143 as a function of @xmath144 for five bar - crossing slices ( @xmath145=0.0 , @xmath146=-0.6 , @xmath147=-0.3 , @xmath148=0.6 , @xmath149=0.9 ) at sa=45@xmath113 where @xmath144=@xmath150 .
rcccc + sa&*@xmath115*&*@xmath116*&@xmath85&@xmath83 + 75@xmath113&0.376&0.193&0.9&5.7 + 45@xmath113&0.348&0.152&0.9&4.8 + 15@xmath113&0.367&0.191&1.2&4.5 + -15@xmath113&0.336&0.186&1.2&4.5 + -45@xmath113&0.336&0.212&1.2&5.7 + -75@xmath113&0.313&0.200&1.2&5.7 + for this projection , there seems to be a plateau at @xmath151 .
we note that this value is smaller than the bar pattern extent indicated by our best - fit solutions ( and the major axis bar length estimated by inspection of the surface brightness distribution ) and furthermore , the same analysis performed over the range of sas does not always as clearly show the same behavior .
presumably , this particular value is more indicative of the bar minor axis length than the full radial zone of the bar pattern , since slices at a 45@xmath113 sa sample along @xmath152 perpendicular to the bar major axis ; other slice orientations are similarly limited to sampling the bar according to its specific projection . allowing that information from the full radial zone of the bar
is not manifest in this type of indicator , we here proceed to assess the consequences for inner and outer pattern speed estimates when the transition between the two patterns is misidentified .
table [ tab - bsmac ] lists the twr bar and spiral estimates for each of the sas studied in the previous sections where we have limited the transition to 0.9@xmath20@[email protected] . at all
sas , this error of several bins in @xmath85 causes an overestimation of the inner pattern speed .
the outer pattern speed , on the other hand , though slightly raised , is still reassuringly accurate .
we can interpret the inaccuracy in @xmath115 , then , as the result of the mis - association of information from one pattern to the other via equation ( [ eq : outin ] ) , as discussed at the end of @xmath3 [ sec : ierror ] .
additionally , we can attribute the greater inaccuracy in solutions from quadrant ii to the reasons discussed in @xmath3 [ sec : ierror ] .
though the subtleties in table [ tab - bsmac ] are most likely specific to this simulation , we emphasize that systematic pattern speed errors introduced by transition mis - identification are generic to the nature of the calculation .
one of the greatest strengths of the twr calculation is that the transition is in principle a free parameter ( within limits ) and need not be restricted to a single , pre - determined value .
we therefore recommend letting the results of the twr calculation speak for themselves : given sufficient resolution and reasonable measurement errors , step - model solutions with the most realistic transition should be recognizable by how well they reproduce the actual @xmath20@xmath21@xmath22 .
since the transition determines the separation of the patterns by interpreting the contribution made to these integrals by each , the natural result is the most accurate determination of the pattern speeds possible . in the previous section we showed that the twr method is capable of detecting and measuring a constant spiral pattern speed that spans less than a third of the disk . here
, we test the aptitude of the twr method in measuring a radially varying spiral pattern speed that subsists over a large radial zone . since the strong spiral surface density enhancement in this simulation ( figure [ fig - sim2 ] ) has only moderate azimuthal range like the spiral in @xmath3 [ sec : sim1 ] , we further explore the likely limitations intrinsic to detecting spiral nonaxisymmetry with a given slice orientation .
the two - armed spiral featured in this simulation extends over a large portion of the disk and is strong both in the surface brightness distribution ( figure [ fig - sim2 ] ) and as traced by departures from axisymmetric rotation ( streaming motions ) in the velocity field ( figure [ fig - sim2vel ] ) .
we estimate the extent of the spiral structure from that of the dominant @xmath86=2 component in the fourier power spectrum plotted in figure [ fig - pplot30s ] . with the expectation , then , that the spiral structure exists between 0.5@xmath153@[email protected] , we restrict our spiral pattern speed solutions to the radial zone @xmath85@xmath20@xmath53@xmath20@xmath83 bordered at the innermost and outermost radii by two independent sets of unregularized bins .
( for our models , 0@xmath20@[email protected] and 2.8@xmath20@[email protected] . ) between the radii marked by the free (
though restricted ) parameters @xmath85 and @xmath83 , we allow each spiral pattern speed solution to vary with radius as an @xmath71 order polynomial where @xmath72=0 to 2 . the two sections of ` place - holding ' unregularized bins serve to isolate the solution in the radial range of interest ; by minimizing errors due to either incorrectly incorporating or imposing an ill - prescribed form to bins outside the radial zone of the spiral , we reduce the introduction of inaccuracies to the spiral pattern speed solution .
the first , outermost section covers the radial zone identified by visual inspection of the surface brightness ( and substantiated by the fourier power spectrum ) where the dominant spiral ceases to extend and where the weakness of the fourier components suggests a region without a noticeable pattern . as a result
, we consider this zone unsuitable for reliable extraction of a pattern speed .
( this last point can also be evidenced by traditional tw values from slices that pass solely through the outermost disk ; the low counts there lead to a large degree of variation in @xmath1 estimates from slice to slice ) .
the second , innermost section covers the radial zone inside @[email protected] . within this radius
we do not expect to be able to extract a realistic pattern speed solution since neither the morphology , velocity field , nor the power spectrum indicate a departure from axisymmetry .
lacking such evidence in the innermost radii , it would be entirely possible to proceed without the use of unregularized inner bins ( as would be the case when the indicators for such a measure are perhaps less obvious ) .
indeed , allowing an independent regularized solution to exist at @xmath53@xmath20@xmath85 , the twr calculation still performs well ; since these bins cover a rather small region of the disk and thereby contribute minimally to the @xmath55 through only the innermost slices , we find that the spiral solution outside of this region closely resembles that in the case where the inner bins are unregularized
. however , we proceed in the manner described above with the expectation that , if only minorly , our spiral pattern speed solution will be improved .
the actual pattern speed at the time of the snapshot shown in figure [ fig - sim2 ] as derived from the time evolution of the phase of the @xmath86=2 component is plotted as a function of radius in figure [ fig - phases ] .
this plot confirms that within @[email protected] , the pattern speed is ill - quantified , with the values for @xmath1 at the inner - most radii oscillating between positive and negative values outside of the vertical range of the plot . at the largest radii
, the pattern speed is characterized by scatter presumably reflective of the lack of a noticeable pattern in figure [ fig - sim2 ] .
the pattern speed between 1.0@xmath153@[email protected] to be reproduced by our solutions shows high - order variation with radius . inside of @[email protected] where the pattern speed is at a maximum ( @[email protected] ) , the pattern seems to be unwinding , while at larger radii the pattern speed decreases with increasing radius .
+ + this behavior is only modestly indicated by traditional tw estimates . in plots of @xmath20@xmath21@xmath22
vs. @xmath20@xmath9@xmath22 ( see figure [ fig - tw51 ] for a comparison of plots generated at four sas ) , adjacent slices trace out a figure - of - eight shape characteristic of complex radial behavior ( namely winding ) . however , the evidence for a variable best - fit slope ( expected for a pattern speed that unwinds and winds ) is not comprehensive or even readily apparent in all cases .
( in fact , all slices are seemingly well fit with a single slope . )
the radial dependence of @xmath1 becomes more apparent upon inspection of the variation of @xmath1 with slice position ( see figure [ fig - twslice51 ] ) , but again , the radial dependence gets smoothed out since @xmath1 for each slice is the result of averaging over all sampled radii .
furthermore , for the four sas shown in figure [ fig - twslice51 ] , there is no single radial behavior implied by all . the twr solutions , on the other hand , are capable of clearly displaying high - order variation in @xmath4 .
the average and rms of the best - fit solutions with @xmath39=0.2 bins at six different sas ( @xmath11815@xmath113 , @xmath11845@xmath113 , and @xmath11875@xmath113 ) is plotted in figure [ fig - ssol ] between the average values of @xmath85 and @xmath83 . for each sa
, we find that the best - fit solution is quadratically varying , correctly reproducing the winding behavior present in the actual pattern speed .
however , the parameters @xmath85 , @xmath83 and the size and location of @xmath154 differ slightly in each solution .
these differences tend to reflect the influence of morphology and slice orientation , as in @xmath3 [ sec : ierror ] .
the value of the peak in @xmath4 , for example , is higher and more pronounced in solutions from quadrant ii than quadrant i ; at positive ( negative ) sas @[email protected] ( 0.27 ) , on average . together with differences in the location where @xmath155 ( which varies for the six sas within the range 1.6@xmath20@[email protected] ) , upon averaging , the result is a slightly under - measured peak value occurring at @[email protected] ( in accord with the actual location ) .
the vertical error bars on bins at large and small @xmath53 similarly reflect variations in the location of @xmath83 and @xmath85 with sa . principally , the solution suffers contamination from bins that perhaps yet contain information from the inner or outer axisymmetric zones .
but we also find that solutions in quadrant ii tend to decrease from @xmath154 to a value @xmath83 that is further in by about 10% than for solutions using slices at the perpendicular orientation .
nevertheless , we find the average @[email protected] and @[email protected] to be in agreement with the bounds of the spiral pattern indicated by the disk surface density and its fourier decomposition . despite these pa - dependent effects , the values implied by the average solution ( @xmath154 , @xmath156 , and @xmath157 )
are accurate to within 5% , 11% and 10% , respectively . that we have correctly reproduced the high - order variation of @xmath4 ,
regardless of sa , however , is perhaps the most remarkable aspect of the twr solutions , even though the detectable variation is only at the 30% level .
naturally , our solution for @xmath4 lends itself to pattern winding time estimates . with the average values for the maximum and minima implied by our solution we can estimate the winding time of the pattern according to @xmath158 for the outer spiral arm , for example
, we estimate an average time to wind @xmath159= 71.64 which is less than 10% from the actual winding time @xmath160=78.53 observed from the time evolution of the simulation .
( this , of course , assumes that @xmath4 does not vary over this time . ) as this simulation would indicate , even without uniform slice coverage , though it may be slightly more difficult to determine with confidence the radial domain of the pattern ( given large errors in @xmath85 and @xmath83 ) , the overall shape , or functional form , for @xmath4 can be ascertained .
of course , this is largely influenced by the adopted measurement errors @xmath95 for each slice and the quality of a priori information that can be gathered and employed . with larger errors @xmath95 , for instance , the @xmath55 criterion becomes less discriminating , and it may be difficult to distinguish between several different radial dependences for @xmath4 . additionally , without clear evidence that limits where the spiral pattern terminates , we risk misidentifying intrinsic radial variation .
indeed , if we restrict @xmath83 to less than that implied by the best - fit solutions and search instead for solutions at a second @xmath83-@xmath161 minimum the pattern speed solutions for all six sas are constant between @[email protected] and @[email protected] . as may be expected , the average value for this constant pattern speed @[email protected] is similar to that suggested by traditional tw estimates where @[email protected] on average . in this section
we address the use of the twr method for the purposes of nuclear bar detection and measurement using the double barred sb0 simulation pictured in figure [ fig - simiii ] . in performing the regularized twr calculation
we again act under the assumption of multiple patterns in distinct radial zones .
our models for @xmath4 parameterize unique , constant pattern speeds for both the primary and secondary bars , known to have pattern speeds of @xmath165=0.23 and @xmath166=0.41 , respectively . from inspection of the surface density and its fourier decomposition ( for which the power spectrum is plotted in figure [ fig-2bpower ] ) , we associate the drop in power of the @xmath86=2 component at @[email protected] and again at @[email protected] with the end of each bar . in step - models for @xmath4 ,
then , we restrict the secondary - to - primary bar and primary bar - to - disk transitions to within 0.4@xmath20@[email protected] and 2.5@xmath20@[email protected] , respectively . to isolate the bars from the rest of the disk we extend the quadrature to the edge of the surface density and employ our ` cut ' procedure in light of the argument set forth at the end of @xmath3 [ sec : twtwr ] .
this is particularly compelling here since , despite the apparent axisymmetry beyond the primary bar in this sb0 simulation , the fourier decomposition shows power in the @xmath86=2 mode beyond @[email protected] , especially in the last third of the disk .
if the asymmetry in this radial zone ( which appears only very weakly in the surface density ) is sustained by an ill - defined , non - unique or perhaps unrigid pattern for which the tw / twr assumptions break - down then associating it with a measurable pattern speed will likely introduce regularization - induced bias to the solution for the interior patterns of interest .
critically , asymmetry such as this may prevent clear integral convergence beyond the primary bar end .
so , too , can its presence in the integrals be expected to degrade the reliability of pattern speed estimates for the structures of interest .
removing the influence of the information in this outer radial zone by calculating the bins there without regularization is our best chance for accurate pattern speed measurement . given that the departure from axisymmetry manifest by a small nuclear bar will be relatively minor compared with that of other patterns in the disk , this is especially relevant for accurate measurement of @xmath116 .
in this case , non - axisymmetry on a comparable scale may easily upset this structure s contribution to the integrals and , if prescribed an incorrect pattern speed model , may introduce consequences for the inner most bins in solutions . our procedure for this simulation , however , does not quite involve calculating without regularization all bins up to the patterns of interest ( i.e. to the end of the primary bar ) .
as in all cases , but particularly here where we are compelled to ` cut ' approximately two thirds of the disk , leaving a large portion of the total bins unregularized may begin to reintroduce unamendable propagating noise . to reduce this risk , careful attention has been paid to the development and testing of models which prevent the destabilization of solutions .
( indeed , the appropriate balance between noise and stability in models for @xmath4 must be explored on a galaxy by galaxy basis . ) here , we find that the most stable and realistic models for @xmath4 are those which include a third , constant pattern past the end of the primary bar . from this we might infer that , though weak and difficult to discern in figure [ fig - simiii ] , there exists a spiral pattern outside the primary bar , perhaps corresponding to the @xmath86=2 component beyond @[email protected] which remains clear , though modest out to @[email protected] .
indeed , we assume that the third minimum in the power of the @xmath86=2 component at this radius corresponds to the end of the spiral pattern , and moreover , since counts are low in the rest of the disk , that the bins beyond @[email protected] are best calculated without regularization .
we note , however , that we do not necessarily expect to measure a realistic pattern speed in this third radial zone . compiling this evidence for two bars and a possible spiral we search for the best - fit solutions parameterizing two constant pattern speeds ( one for the primary bar , one for the inner secondary bar ) out to the end of the primary bar , in addition to a third constant pattern speed @xmath116 restricted to extend out to 4.3@xmath20@[email protected] .
we find that solutions generated in this manner provide much more accurate estimates for the primary and secondary bars compared with solutions that are either regularized over the full extent of the surface density or unregularized up to the end of the primary bar . for the purposes of further establishing favorable conditions for nuclear bar detection
, we adopt a small bin width @xmath39=0.15 in the quadrature . as described in @xmath3 [ sec : ierror ] and @xmath3 [ sec : trans ] , the mismatch by a fraction of a bin width or more between the actual transition and that to which the solution is confined can have consequences for both inner and outer twr pattern speed estimates .
real nuclear bars will need to be well resolved in order to accurately separate the contributions of the two bars .
the secondary bar in this simulation is known to be non - rigidly rotating ; in @xcite and in @xcite , the amplitudes and pattern speeds of secondary bars formed in purely collisionless @xmath45-body simulations through the introduction of a rotating pseudobulge oscillate as the bars rotate through the companions in which they are nested . in @xcite , tw estimates of the secondary bar pattern speed measured using bulge - only kinematics ( the bulge supports the nuclear bar alone and a primary bar contribution need not be accounted for in the tw integrals ) are subject to marked errors consistent with an origin in non - rigid rotation .
these errors result in estimates of @xmath166 too high on one side of the galaxy and too low on the other , in accord with being a manifestation of the oscillations driving radial pulsations which contribute with different signs on the two sides of the galaxy .
cancellation between measurements from both sides of the galaxy in global regularized solutions reduces the effect of the oscillations . in the discussion to follow ,
our focus is on sources of error in the twr calculation other than this intrinsic effect .
rcccccc + sa&*@xmath166*&*@xmath165*&*@xmath116*&@xmath167&@xmath168&@xmath83 + 75@xmath113&0.420&0.231&0.146&0.75&3.15&4.35 + 45@xmath113&0.548&0.278&0.178&0.75&3.15&4.35 + 15@xmath113&0.407&0.231&0.063&0.6&2.7&4.8 + -15@xmath113&0.404&0.264&0.087&0.75&2.7&5.25 + -45@xmath113&0.477&0.259&0.233&0.9&3.15&4.35 + -75@xmath113&0.375&0.225&0.028&0.9&3.15&4.5 + -&0.41&0.23&-&0.8&3.0&5.0 + the twr estimates from solutions with @xmath39=0.15 bins at six sas ( @xmath11815@xmath113 , @xmath11845@xmath113 , and @xmath11875@xmath113 ) are listed in table [ tab-2btwr ] and the average and rms for this sa range is shown in figure [ fig-2bssol3pas ] . compared with solutions generated using @xmath39=0.3 bins in the calculation ,
these solutions are much more stable and accurate ; at less than 8% from their actual values , the comprehensive primary and secondary bar pattern speeds evidently benefit from the integrity of the calculated transitions between patterns ( we find @[email protected] , @[email protected] , and @[email protected] ) .
though the twr estimates in figure [ fig-2bssol3pas ] are accurate , the rms in each suggests 8 to 13% error on the part of the method .
in addition to the intrinsic non - rigid rotation - error expected for the secondary bar measurement , this may yet indicate the slight misdesignation of transitions between patterns , and we can also assume that ( though we do not necessarily consider its measurement realistic ) the third measured pattern speed has a non - trivial influence on the inner pattern speeds , as must @xmath165 on @xmath166 , for instance . with so many free parameters in our model more than in any other used in this paper identifying unequivocally
the dependence of each on the others is difficult ( indeed , the rms in each estimate likely reflects the combined consequence of many such inter - dependencies ) . as in previous sections , however , we can begin to understand the largest source of variability in the estimates from sa to sa by considering the influence of morphology and slice orientation on the solutions .
if we consider that the major axes of the two bars are oriented at @xmath5445@xmath113 from one another at this time step ( that is , with the secondary bar aligned along the x - axis , the primary bar is aligned along the line @xmath10=@xmath9 , or in the @xcite convention , @xmath169=45@xmath113 ) , then slices at @xmath11845@xmath113 are oriented along the primary bar minor / major axis . in this case , even in the presence of the secondary bar @xmath20@xmath21@xmath22 and @xmath20@xmath9@xmath22 for the inner primary - bar crossing slices fluctuate noisily about zero .
this behavior is more than enough to prevent accurate tw estimates for the primary bar ( which @xcite find are less reliable when bars are aligned within @xmath11820@xmath113 of a principle axis ) , and so it is perhaps not surprising that our twr estimates at @xmath11845@xmath113 sas show rather large inaccuracies .
however , what we find requires slightly more interpretation : it is the secondary bar estimate , and not @xmath165 , that is compromised . unlike the results in @xmath3 [ sec : sim1 ] with which we could argue that the regularized twr method is , in principle , fairly equipped to accommodate any alignment of patterns ( and where even slices at @xmath11815@xmath113 from the bar axis return accurate twr estimates ) , here we are not only dealing with more model parameters ( and their intrinsic covariance ) , but the innermost pattern now has a much smaller extent than the structures in the disk .
consequently , the secondary bar occupies the smallest fraction of bins in the innermost slices where , moreover , the errors @xmath58 are the largest .
presumably , in this exploitable state the secondary bar estimate is sacrificed for the primary bar @xmath165 in the regularized calculation ; at sa=@xmath11845@xmath113 , our measurement of @xmath166 is overestimated by @xmath5425% , while @xmath165 is good to within 11% , on average .
( we can assume that such imprecision also reflects the effect of the secondary bar s non - rigid rotation . ) at other slice orientations where the primary bar is more favorably sampled , the trade - off is less severe ( though modestly pa - dependent ) .
an impromptu calculation using particles in the bulge only ( motivated by the @xcite strategy ) seems to confirm the influence of the primary bar in the twr calculation at sa=45@xmath113 .
the error in our lone estimate @xmath166=0.42 in this case is significantly less than when the primary bar is present and also comparable to that in bulge - only estimates at other sas ; for the three sas in quadrant ii tested , we find @[email protected] . even if the twr method can not accurately measure the pattern speed of a pulsating nuclear bar aside from with global regularized solutions , in particular
it is nevertheless appropriate for use in characterizing and constraining @xmath166 , especially in the presence of a primary bar .
the comparative worth of twr solutions can be deduced in light of the distinction between the traditional tw and the twr methods .
our twr pattern speed measurement for the primary bar is comparable to that from the traditional tw calculation ; using slices covering @xmath1702.25 ( such that the bar contribution is maximal in all slices ) , the tw method returns @[email protected] . the twr secondary bar estimate , on the other hand , greatly improves upon the tw estimate @[email protected] available using slices covering @xmath1700.6 .
since the nuclear bar appears in only a very small number of slices and its contribution is , moreover , easily overwhelmed in the extended tw integrals , the tw method only modestly recovers @xmath166 . in the twr calculation , however , the few bins that cover the nuclear bar are supplied with large weights @xmath52 compared with the rest of the information along each slice .
together with a suitable bin width , this effectively isolates information from throughout the nuclear bar extent from that of all other patterns in the disk .
the quality with which @xmath166 can be constrained then depends on how well the contribution from these other patterns is identified and removed from slices intersecting the nuclear bar , and moreover relies on regularization over both sides of the galaxy to construct a global solution . aside from the unavoidable error introduced by the non - rigid rotation of nuclear bars expected from n - body simulations , from our current study it seems likely that similar calculations may be limited according to the degree of resolution and the relative size of the secondary bar .
that is , the former dictates the precision with which the secondary - to - primary bar transition can be determined , and the latter sets the leverage supplied by the secondary bar to the @xmath55 .
furthermore , given that the measurement of an inner , nuclear bar pattern speed is subsequently largely affected by how well the other patterns present in the disk can be measured , success with the twr method requires appropriate models for @xmath4 that account for such concerns .
the successes of the twr method as applied to simulated data should be obtainable with real galaxies , provided that adequate attention is paid to several considerations .
since the calculation foremost requires integration along slices parallel to the galaxy major axis which reflect information from all patterns of interest in the disk , observations must be able to present a number of these .
as governed by the resolution or sampling of the data , the placement of these slices must be able to define a quadrature wherein the position @xmath10 of the outermost slice corresponds to that of the last matrix element along the slice requiring the largest limit of integration @xmath125 for integral convergence ; when convergence can be reached only clearly at the map boundary ( perhaps far beyond the extent of the patterns ) , this is essential for correctly accounting for all information along each slice .
establishing the slice positions and orientation for an accurate quadrature also clearly requires superior knowledge of kinematic and morphological parameters . in general , the same restrictions to the quality of input data in traditional tw calculations
apply to the twr method .
as described in @xmath3 [ sec : err ] , errors in the assumed pa which impair tw estimates can also introduce considerable inaccuracy to the twr solutions .
the inclination angle of the disk must also be well - determined and should be preferably restricted to only moderate values ( which , beyond the observational requirements of the tw method , will keep errors from corrupting the association of information into accurate radial bins ) . additionally ,
though more easily overcome in traditional tw estimates @xcite , errors in the systemic velocity and galaxy center might prove critical for the twr solutions since each side of the disk provides an independent solution for @xmath4 .
care must be taken not to impair the prevailing symmetry along each slice , given that each integral bears more than one estimate .
the incorrect placement of radial bins according to a mis - assigned galaxy center position , for example , could significantly over- or under - estimate the actual pattern speed , and moreover make assessment of the true radial variation unlikely .
the regularization procedure developed here itself makes further demands on the quality and amount of information necessary to perform the calculation .
but by keeping the amount of information beyond that required of the tw method to a minimum , and using standard diagnostics such as fourier decomposition , the requisite set of a priori assumptions can be invoked quite reliably .
as long the information is accessible , requiring at the least theoretical motivation to develop testable models , and limited in principle only by the quality of information from which it is to be gathered , then the regularization should proceed without impediment . of course ,
unlike the simulated galaxies studied here where there is plenty of signal throughout the disk , observations of real galaxies may present sensitivity issues . while regions of low signal - to - noise in the outer disk can be superseded using the ` cut ' procedure developed here
, high quality information from the rest of the disk is an obvious priority for the method ; the departures from axisymmetry induced by all patterns present in the disk must be clearly detectable .
not only does the calculation depend on the presence ( or lack ) of these signatures in both the surface density and in the velocity field but the mere identification of the number and domain of patterns is critical for developing appropriate models for @xmath4 .
this latter necessity may be hard met since , for instance , it will be rare to observe galaxies with surface densities that can be fourier - decomposed as cleanly as is possible with simulated galaxies .
furthermore , unlike simulations , it is impossible to establish whether or not there exists more than a single pattern speed at each measured radius in real galaxies . since the models developed with regularization here are incompatible with non - unique pattern speeds , for real galaxies , the choice of when and where to consider a transition or to keep solutions unregularized may be based on a more qualitative assessment of where clear structure ends .
inevitably , the combination of the above considerations ( related , overall , to the quality of the data ) will determine the extent to which the model for the true radial behavior of @xmath4 can be differentiated from other models .
that is , the @xmath55 criterion with which we judge the goodness of solutions becomes less discriminating the larger the measurement error @xmath95 . since the adopted measurement error @xmath95 for each slice used in the calculation and in the @xmath55 estimator must necessarily incorporate observational errors based on random noise in the data , with severe enough errors different model solutions from real , imperfect data may be indistinguishable .
in addition , systematic errors ( likely dominated by pa uncertainty ) will undeniably challenge the accuracy of solutions . in all applications of the method
it is critical to assess the influence of these errors through direct tests of the sensitivity of solutions to departures from the nominal values of pa , inclination , and kinematic center , for instance .
clearly , this makes 2-d coverage desirable ; here , 2 - 3@xmath113 uncertainties in the pa alone are shown to introduce around 15% error in measurements of @xmath4 for the barred spiral simulation
. insufficient resolution or sampling may also impair twr solutions from real galaxies .
a large adopted bin width not only limits the detectable radial variation in @xmath4 , but also restricts how well multiple patterns can be separated in the resultant quadrature ; a mismatch between the actual transition and that to which the solution is confined can have consequences for the estimates of both inner and outer patterns .
naturally , depending on the models to be tested and relative size of the disk , a resolution - constrained bin width is not guaranteed to impair solutions for all galaxies .
we nonetheless foresee that the only true way to preserve the integrity of solutions is with high - resolution observations .
data cubes lend themselves well to analysis with the twr method , since unlike long - slit spectroscopic observations , the galaxy pa , inclination , systemic velocity and ( kinematic ) center can be derived with errors from the data using a tilted - ring analysis on the first moment of the cube .
additionally , multiple slices can be defined with a single observation .
so between radio and sub - mm investigations of spiral structure , for instance , and ifu spectroscopy with which double bar systems ( and eventually double barred spirals , given larger , more sensitive ifus ) can be studied , applications of the twr method could be extensive . of course , like all applications of the tw method , the observed tracer
must be found to obey continuity and
the relation between the intensity of the tracer and the surface density must be linear or well - determined everywhere . reviews of several possible tracers argued to suitably obey the tw continuity requirement can be found throughout the literature , but we note here that the work of @xcite studying the effect of dust on tw measurements of bars may find meaningful extension in future twr studies of multiple patterns in late - type galaxies .
there , model dust lane features associated with bars introduce errors on the order of 20 - 40% @xcite .
in addition to these errors , twr solutions could possibly be prone to increased error from spiral dust lanes at larger radii . though it is beyond the scope of this work to make a detailed assessment of the sensitivity of twr solutions to dust
, we argue that such noise could well be mitigated through the use of regularization , and expect no greater an effect in twr measurements than tw , which moreover , will be apparent with the use of only optical tracers . in the immediate future
, we plan to apply the twr method to several high - resolution bima song co observations of molecule dominated galaxies to search for spiral winding , relations between bar and spiral pattern speeds , and spiral - spiral mode coupling .
( these observations include single - dish data and therefore do not suffer from missing flux which would be a violation of the continuity requirement . ) for those galaxies with isms not dominated by molecular gas , we plan to combine the co with hi data to make total column density maps ( assuming the ionized component is negligible ) .
since the co - h@xmath173 conversion factor is critical in combining the co and hi maps ( and , of course , in establishing molecule dominance in the former case ) , it will be necessary to test the sensitivity of the twr method to the adopted conversion factor for such combinations .
additionally , since warped disks ( common in hi ) are a violation of the tw assumptions , we will also perform tests to determine if our cut - off scheme can be used to circumvent the warp and thereby extract solutions from the rest of the disk . for these applications of the twr method
, we plan to construct measurement errors @xmath120 for each slice that reflect uncertainties related to the flux cutoff chosen in creating the moment maps .
the effect of pa and other systematic errors will be assessed by testing the sensitivity of solutions to departures from the nominal values .
although in the interest of testing our strategy for each simulated galaxy is somewhat tailored to its unique properties , with the above caveats in mind our studies have enabled us to develop a general and reasonably objective prescription for applying the regularized twr calculation : 1 . establish the bin width and the corresponding number of slices ( not necessarily uniformly spaced ) that are required to achieve converged integrals using an @xmath38 quadrature . for the purposes of measuring multiple patterns in a single disk , this will likely extend to the map boundary .
2 . compile a priori information by inspecting the surface density , its fourier decomposition , and the velocity field for indications of patterns and to establish the expected number and domain of measurable pattern speeds .
this should include the identification of regions in the disk susceptible to regularization - induced bias .
3 . develop theoretically and/or observationally motivated models which parameterize @xmath4 according to the a priori information .
4 . incorporate measurement errors into a single @xmath95 ( and @xmath120 ) for each slice .
these should represent uncertainties in the adopted intensity noise level , and/or other random noise - related errors ; systematic errors are preferably determined through direct testing ( see item 9 ) .
develop the weighting scheme for a reduced @xmath55 estimator which accounts for the total degrees of freedom for the models to be tested .
this should reflect expectations for which slices , if any , are most critically to be reproduced by the models .
generate preliminary solutions for the models . at this point , the degree of regularization required to return solutions according to type should be explored .
7 . with finalized solutions , use equation ( [ eq : twr ] ) to generate a complete set of @xmath20@xmath21@xmath22 for each and calculate the corresponding @xmath174 .
use the @xmath174 to identify the best - fit solution .
test the sensitivity of the results to other systematic effects peculiar to the observation , e.g. adopted pa and/or the co - h@xmath173 conversion factor , for instance .
in this paper we have shown that regularizing the twr calculation is an effective means of smoothing intrinsically noisy solutions for more precise measurement of @xmath4 .
specifically , ( barring a large , limiting resolution ) regularization admits the use of a much smaller bin width than that required to achieve comparable smoothness in the unregularized calculation .
this affords improved assessment of radial variation as well as more accurate determination of the transitions between multiple pattern speeds ( and thus of the values of the pattern speeds themselves , in principle ) .
moreover , with the regularized twr calculation , different theoretically and observationally motivated models for the radial dependence of @xmath4 can be tested in fairly short time and with only the minor addition of information compared to the unregularized twr and tw methods . with a simple scheme for generating @xmath71 order polynomial solutions which can be incorporated into step models
, we have shown that the twr method can be used to parameterize the radial domains of multiple pattern speeds . together with a priori information identifying zones in the disk which may be incompatible with measurement ( either because they are characterized by low signal - to - noise or show no evidence for a pattern )
, we can further constrain the extent of patterns while optimally reducing regularization - induced bias in pattern speed solutions . as applied to three simulated galaxies ,
we find that the twr method developed in this manner performs with a high degree of accuracy ( with less than 15% error ) both in measurement and in extracting information about the true functional form for the pattern speed . tests on a simulation of a barred spiral galaxy indicate that not only can the constant pattern speed for a relatively weak spiral be reliably reproduced , but information about both the pattern speed and the radial extent of the bar pattern can also be extracted .
( indeed , we find that the bar pattern speed estimate is strengthened by the proper use of information from beyond the bar end . ) and though the bar pattern speed estimates is highly susceptible to systematic errors with pa errors introducing the largest uncertainty to twr pattern speeds , as with tw estimates we find that the identification of the transition between the two is relatively stable .
the twr method can also be effectively employed to measure patterns that are winding in nature . in a simulation of a two - armed spiral
, the best - fit twr solutions from several slice orientations correctly reproduce the high - order radial variation of the pattern speed , despite modest indication that not all orientations supply the same authority ( this , of course , would seem to depend on the morphology of this spiral , in particular ) . indeed , though the twr method can , in principle , handle any ( presumably random ) alignment of patterns , in all of the simulations studied , slice orientations which provide the most uniform coverage of the patterns are preferred .
this is of particular importance for nuclear bar pattern speed measurement , as found in tests of the method on a double barred simulation . since the innermost bins which provide the foremost leverage on the nuclear bar are also the most susceptible to errors from throughout the disk , confident measurement requires all other patterns to be well constrained .
in principle , comparable accuracy should be achievable on real galaxies .
however , these tests do not constrain how well the twr method can perform under severe observational limitations which may commonly arise . not only can determinations of the pa , inclination , and dynamical center be subject to considerable errors given low - quality data , but identifying constraints on the patterns present in the disk to be incorporated into models for @xmath4 could prove challenging . additionally , though regularization can reduce the impact of noise on solutions , large measurement errors for each slice could make discriminating between several possible models for @xmath4 difficult . and
most critically , since the nature of the numerical calculation relies on a relatively small bin width to achieve its greatest accuracy , without high resolution , some observations may not afford practical solutions .
nevertheless , if restricted to high resolution observations with adequate sensitivity , and given radially stable kinematic parameters , twr solutions can be used to study the connection between multiple patterns and the nature of spiral winding .
so , too , can we expect progressively more satisfactory applications of the method ; though the number of galaxies to which the method can be successfully applied is limited by the current generation of instruments , in the future , larger ifus , alma , and eventually , ska should yield much higher quality data with larger areal coverage and higher angular resolution .
this prospect in itself should warrant future studies with the twr method .
+ this research was supported in part by nasa through the american astronomical society s small research grant program .
s. e. m. acknowledges support from a nasa - funded new mexico space grant consortium graduate research fellowship . v.
p. d. is supported by an rcuk fellowship at the university of central lancashire .
j. s. acknowledges support from a harlan j. smith fellowship .
this material is based on work partially supported by the national science foundation under grant ast 03 - 06958 to r. j. r. corsini , e. m. , debattista , v. p. , aguerri , j. a. l. 2003 , apj , 599 , l29 debattista , v. p. & sellwood , j. a. 1998 , apj , 493 , l5 debattista , v. p. & sellwood , j. a. 2000 , apj , 543 , 704 debattista , v. p. & shen , j. 2007 , apj , 654 , 127 debattista , v. p. , corsini , e. m. & aguerri , j. a. l. 2002 , mnras , 332 , 65 debattista , v. p. , gerhard , o. , sevenster , m. n. 2002 , mnras , 334 , 355 debattista , v. p. 2003 , mnras , 342 , 1194 debattista , v. p. & williams , t. b. 2004 , apj , 605 , 714 debattista , v. p. , mayer , l. , carollo , c. m. , moore , b. , wadsley , j. & quinn , t. 2006 , apj , 645 , 209 elmegreen , b. g. , elmegreen , d. m. , seiden , p. e. 1989 , apj , 343 , 602 elmegreen , b. g. , elmegreen , d. m. , chromey , f. r. , hasselbacher , d. a. & bissell , b. a. 1996 , aj , 111 , 2233 fathi , k. , toonen , s. , falcon - barroso , j. , beckman , j. e. , hernandez , o. , daigle , o. , carignan , c. & de zeeuw , t. 2007 , apjl , 667 , 137 garcia - burillo , s. , combes , f. , gerin , m. 1993 , a&a , 274 , 148 gerssen , j. & debattista , v. p. 2007 , mnras , 378 , 189 kormendy , j. & norman , c. a. 1979 , apj , 233 , 539 maciejewski , w. 2006 , mnras , 371 , 451 masset , f. & tagger , m. 1997 , a&a , 318 , 747 merrifield , m. r. & kuijken , k. 1995 , mnras , 274 , 933 merrifield , m. r. , rand , r. j. , & meidt , s. e. 2006 , mnras , 366 , 17 miller , k. 1970 , siam journal on mathematical analysis , vol 1 , 52 rand , r. j. & wallin , j. f. 2004 , apj , 614 , 142 rautiainen , p. & salo , h. 1999 , a&a , 348 , 737 rautiainen , p. & salo , h. 2005 , apjl , 631 , 129 roberts , w. & stewart , g. 1987 , apj , 314 , 10 sellwood , j. a. & kahn , f. d. 1991 , mnras , 250 , 278 sellwood , j. a. & lin , d. n. c. 1989 , mnras , 240 , 991 sellwood , j. a. & sparke , l. s. 1988 , mnras , 231 , 25 shen , j. & debattista , v. p. 2007
, in preparation sygnet , j. f. , tagger , m. , athanassoula , e. & pellat , r. 1988 , mnras , 232 , 733 tikhonov , a. n. & arsenin , v. y. 1997 , solutions of ill - posed problems ( new york : wiley ) tremaine , s. & weinberg , m. d. 1984 , apj 282 , l5 zimmer , p. , rand , r. j. & mcgraw , j. t. 2004 , apj , 607 , 285
|
at the intersection of galactic dynamics , evolution and global structure , issues such as the relation between bars and spirals and the persistence of spiral patterns can be addressed through the characterization of the angular speeds of the patterns and their possible radial variation .
the radial tremaine - weinberg ( twr ) method , a generalized version of the tremaine - weinberg method for observationally determining a single , constant pattern speed , allows the pattern speed to vary arbitrarily with radius . here
, we perform tests of the twr method with regularization on several simulated galaxy data sets .
the regularization is employed as a means of smoothing intrinsically noisy solutions , as well as for testing model solutions of different radial dependence ( e.g. constant , linear or quadratic ) .
we test these facilities in studies of individual simulations , and demonstrate successful measurement of both bar and spiral pattern speeds in a single disk , secondary bar pattern speeds , and spiral winding ( in the first application of a tw calculation to a spiral simulation ) .
we also explore the major sources of error in the calculation and find uncertainty in the major axis position angle most dominant . in all cases ,
the method is able to extract pattern speed solutions where discernible patterns exist to within @xmath0 of the known values , suggesting that the twr method should be a valuable tool in the area of galactic dynamics . for utility
, we also discuss the caveats in , and compile a prescription for , applications to real galaxies .
|
the simulation results presented in these supplementary materials are obtained in the following way .
we first distribute randomly a number @xmath8 of potential activity centers uniformly in the unit circle .
then , at each time step , we add a worker @xmath0 at a random position in the circle , and compute the cost functions @xmath72 for all potential activity centers @xmath1 .
we then connect the worker with the center maximizing the cost function , and add one to the value of the traffic corresponding to this center .
we repeat this procedure until all workers are connected to a center .
+ strictly speaking , traffic congestion do not arise only at the exact location of the activity center , but in an area of typical size @xmath73 around that center .
people who are not commuting to this specific center , but who have to travel close to it when commuting to their working place , might add to the traffic in the surroundings of this center .
therefore , we should consider the traffic @xmath74 in the area around the center @xmath1 , that is to say the number of people whose commute path crosses the area of size @xmath73 around this center . for the sake of simplicity and clarity ,
we choose to ignore this in our model and leave it for further investigation .
1 in the main text suggests the existence of two different types of polycentric regimes that we called attractivity - driven and distance - driven polycentric regimes . in the attractivity - driven regime
, individuals decide to work at the most attractive center , provided that the traffic is not too large .
therefore , as the traffic increases , new centers are going to appear in the decreasing order of attractivity . in the distance - driven regime , however , individuals tend to connect to the closest center , and thus the centers should appear in a random order .
+ more specifically , the attractivity - driven regime appears when the term @xmath75 $ ] is negligible compared to the attractivity @xmath24 at small values of traffic .
then , maximising the value of @xmath72 amounts to maximising the attractivity of the center @xmath1
. this will typically be the case when : on the other hand , when @xmath77 , the attractivity of the centers becomes irrelevant and maximising @xmath72 amounts to connecting the closest center @xmath1 to @xmath0 .
an important consequence is that the assumptions used to derive eq . 6 and eq .
10 in the main text are not valid in the distance - driven regime .
in fact , there can not be any stable monocentric state in the distance - driven regime and we start from the beginning ( after a few iterations ) with a polycentric state .
therefore , the difference between attractivity - driven and distance - driven polycentric regime draws the distinction between a system which grows from a single center , and a system where several centers appear from the beginning .
this may be interpreted as the emergence of a single city when people can afford to travel a longer distance than the extension of the system ( attractivity - driven ) , and of a system of cities when they can not ( distance - driven regime ) .
+ in between these two regime lies an intermediate regime , which is neither completely driven by space , not completely driven by the attractivity .
this interplay between space and attractivity makes this intermediate regime difficult to explore analytically , we thus limit ourselves in this letter to the two extreme regimes .
we now show that if the parameters are such that there exists a monocentric state , then the system necessary evolves to an attractivity - driven polycentric structure .
let us assume that the system is in a monocentric state : @xmath26 individuals commute to the center @xmath78 characterised by the attractivity @xmath25 .
let us assume that the monocentric state is unstable and that the @xmath79th individual @xmath0 has two different possibilities .
first , she could choose to go to the next most attractive subcenter characterised by the attractivity @xmath80 , the largest attractivity among all the remaining potential centers .
this translates into @xmath81\ ] ] in general , the most attractive center is at a random fraction of the system size @xmath82 , that is to say @xmath83 .
we are then led to eq .
6 of the main text ( see further in the supplementary material for a derivation ) @xmath84 the second possibility is for the individual to choose the closest center , characterized by @xmath24 .
we then have @xmath85\ ] ] since @xmath1 is the closest subcenter to @xmath0 we can write @xmath86 and after a simple calculation we obtain a new value for @xmath36 ( in the limit of large @xmath8 ) @xmath87 we immediately observe here that for all values of the parameters , we always have @xmath88 .
this result indicates that if a monocentric state exists , the transition is always towards an attractivity - driven polycentric structure . or conversely , that there can not be any stable monocentric state in the distance - driven regime . in order to verify the ordering in the apparition of secondary centers
and therefore the existence of two distinct regimes , we increase the population in a certain configuration until all the the centers are populated . each time a new center is populated , we note its attractivity . at the end , we compare the list obtained with the list of all the attractivities of the centers in decreasing order and compute kendall s @xmath89 function defined by : where @xmath91 is the number of centers , @xmath8 the number of concordant pairs and @xmath92 the number of discordant pairs .
a pair @xmath93 is said to be concordant if @xmath94 and @xmath95 or if @xmath96 and @xmath97 .
it is said to be discordant otherwise .
one can see that if the orderings are identical we have @xmath98 , if they are completely opposite we have @xmath99 and @xmath100 if there is no correlation between the two lists .
[ fig : kendall ] shows the evolution of @xmath101 with the value of @xmath21 .
we see that we have @xmath100 for small values of @xmath21 , indicating that the centers appear in a random order , and that we are in a distance - driven regime . on the other hand , for large values of @xmath21 we have @xmath98 , indicating that the centers appear in increasing order of their attractivity , thus that we are in an attractivity - driven regime .
the study of the nature of the transition between the two regimes goes beyond the scope of this paper , and we leave it for further investigation . + with @xmath102 for @xmath103 , @xmath104 , and @xmath105 .
the curve shows the average for 1000 configurations , and the standard deviation is shown by the shaded area .
the dotted black lines highlight the extreme values @xmath100 and @xmath106 [ fig : kendall],scaledwidth=45.0% ] we confirm numerically the absence of a stable monocentric state in the distance - driven regime , and its existence in the attractivity - driven regime .
we plot on fig .
[ fig : pstar ] the value of @xmath36 , the population at which a second center appears for different values of @xmath107 .
we see that when @xmath108 , @xmath109 , which means that there is no stable monocentric state . on the other hand ,
when @xmath110 , @xmath36 is different from @xmath111 , there exists a stable monocentric regime .
we show later in these supplementary materials that we can compute an analytical expression for @xmath36 when @xmath112 . , the value of population at which a second center appears , with @xmath107 for several values of @xmath82 ( @xmath104,@xmath103,@xmath105 , average of 10000 configurations ) .
all the plots collapse on the same curve , and indicate that there is no stable monocentric state for @xmath113 .
[ fig : pstar],scaledwidth=45.0% ] in the derivation of eq . 13 in the main text
, we assume that cities are systems such that @xmath110 ( and eq .
11 holds ) while still exhibiting some spatial coherence . by spatial coherence , we mean that even if we are in a regime both controlled by the attractivity and space . in particular
, it implies that the overlap between the basins of attraction of each subcenter is relatively small . in order to justify this assumption , we study numerically the behaviour of the average commute distance @xmath114 .
+ if we are in a regime where the individuals connect to the closest center , we have @xmath115 where @xmath41 is the number of active centers . on the other hand , if we are in a regime where individuals tend to connect to the most attractive centers , then the average commute distance scales as @xmath116 .
we thus plot @xmath117 as a function of @xmath107 on fig .
[ fig : coherence_length ] .
we can see that @xmath117 increases as @xmath107 increases : people travel smaller distances in the distance - driven regime that in the attractivity - driven regime . yet , for values of @xmath107 which are not too large ( typically 10 ) the value of the average commute distance lies between the two extremes , closer to the value corresponding to the distance - driven regime , which supports the assumption that has been made in the main text to derive eq . 13 : up to large values of @xmath107 for which we know from figure 1 that we are in the attractivity driven regime , we can still use the assumption of the existence of well - defined basins of attraction for each subcenter with small overlap with each other . with the ratio @xmath118 computed over 10000 generated systems ( @xmath104 , @xmath119 , @xmath103 ) .
we generate the systems such that all the centers are occupied , @xmath120 . for small values of @xmath102
we are in the distance - driven regime , people commute to the closest center and the ratio is close to one . on the other hand , when @xmath118 is large , people commute to the most attractive center disregarding of the distance and the ratio is close to @xmath121 . in the intermediate regime ,
the transition is relatively slow , meaning the system keeps a spatial coherence for a reasonable range of @xmath102.[fig : coherence_length],scaledwidth=45.0% ] in this section , we explicit the details of the calculation of the population @xmath36 for which a first secondary center appears in a city within our model .
we are here interested in the regime @xmath112 in which the monocentric state is stable for small values of the population ( cf fig.2 ) .
we assume that the system is in a situation where all the existing workers are connected to a single center @xmath111 , the most attractive of all the possible centers .
we would like to know what happens when we add the next workers , that is to say whether a secondary center is going to appear , and if so , for what value of the population . + a secondary center @xmath41 will appear when the @xmath122 worker @xmath0 is added to the system if : following our assumptions , the traffic to @xmath111 is given by @xmath124 and the traffic to @xmath41 by @xmath125 .
therefore , the previous equation , in combination with the definition of @xmath72 reads : which is so far an exact result . in order to derive the average value of @xmath36
, we make some further assumption .
first , because households and activity centers are distributed at random , the distance to the different centers is taken to be approximately @xmath128 , the typical size of the system . also , we know from order statistics @xcite that in the case of a random variable @xmath22 uniformly distributed in @xmath129 $ ] , such that @xmath130 we have on average .
therefore , the first secondary center to appear is necessarily the center @xmath41 which maximises @xmath132 , that is to say the second most attractive center with attractivity @xmath80 .
when @xmath133 we find that a secondary center appears when the population @xmath26 of the city is such that : in order to check the validity of our assumptions and thus of the previous formula , we numerically generate a lot of systems for different values of @xmath21 , @xmath82 , @xmath8 and @xmath17 and determine the population @xmath36 at which a secondary center appears .
we then plot the average of @xmath36 as a function of @xmath135 ( fig .
[ fig : collapse_nstar ] ) and see that all the plots collapse on the same curve , confirming the above expression of @xmath36 . .
shows the average of the measured @xmath36 on many simulated configurations of the system as a function of @xmath135 , for different values of the parameters .
the points all collapse on the same curve , confirming our expression for @xmath36.[fig : collapse_nstar],scaledwidth=45.0% ] we saw previously that there always exists a value of the population size over which a secondary center appears .
the same expression for @xmath36 also implies that even in the regime @xmath112 there is a range of parameters for which the monocentric regime simply does not exist , that is when @xmath136 .
assuming @xmath8 and @xmath17 are fixed , for every value of the capacity @xmath16 there exists a value @xmath137 of @xmath21 given by : under which there is no monocentric regime and the system exhibits a polycentric structure from the beginning of its evolution .
therefore , we witness a transition from a monocentric to a polycentric regime only if the parameters are such that @xmath139 and @xmath140 in the derivation of @xmath141 in the main text , we assume that the traffic @xmath44 of the @xmath142 existing centers is rougly the same : @xmath143 .
a simple argument supports this assertion : in order for the @xmath37 center to appear , eq . 7 in the main text
must be satisfied , i.e. @xmath144 must be larger than all the @xmath145 $ ] . seen differently , it means that as long as a center @xmath1 is such that @xmath72 is larger than @xmath144 , people will connect to that center .
therefore , all the centers @xmath1 are going to ` fill up ' until @xmath72 is too large .
we can thus expect that , on average , all the centers will have the same traffic .
of course , the influence of space complicates this reasoning , and to convince the reader , we plot the distribution of @xmath44 in the regime @xmath112 on fig . [
fig : traffic_distribution ] .
the expression given in the main text for @xmath141 obviously relies on the fact that the @xmath24 are uniformly distributed , and it is legitimate to wonder how the scaling would be modified it their distribution was different .
it is usefull to note that the previously derived exponent is the result of two contributions .
first , @xmath151 from the assumption that the commuters are approximately equally distributed among the existing centers .
we do not expect this to change with @xmath24 s distribution and thus we still have @xmath152 .
second , @xmath153 from the maximum of the difference @xmath154 , which obviously depends on the distribution of @xmath22 .
nevertheless , we can reasonably expect @xmath155 where @xmath156 is an increasing function of @xmath41 and decreasing function of @xmath8 .
if we only consider the evolution of @xmath156 with @xmath41 , the function can be approximated around the current value of @xmath41 as : which is still sublinear .
the value of @xmath160 might change with @xmath41 depending on the distribution , but it will always be positive .
thus , the dependence of @xmath41 with @xmath26 might not be a power - law , but it will definitely be sublinear .
we assume that the random variables possess a probability density function @xmath163 defined on @xmath164 , and a cumulative distribution function @xmath165 defined on the same interval .
the probability density function of the @xmath166 _ largest _ sampled value @xmath167 is given by : @xmath173 $ ] is thus a decreasing function of @xmath172 .
therefore , @xmath174 $ ] is an increasing function of @xmath172 .
thus , assuming that the values of @xmath22 are drawn from a pareto distribution , we still have a sublinear dependance of the number of subcenters in the population . in order to verify this formula
, we numerically generate many configurations of the system and we measure the mean @xmath141 for several values of k and several values of @xmath17 .
we then plot the measured average @xmath176 as a function of @xmath177 ( fig .
[ fig : collapse_nk ] ) and see that all the plots collapse on the same curve , confirming the accuracy of our expression for @xmath141 . .
shows the average of the measured @xmath141 on many simulated configurations of the system as a function of @xmath178 , for different values of the parameters .
the points all collapse on the same curve , confirming our expression for @xmath141 .
the plots have been obtained for @xmath179 , @xmath148 , @xmath105 and @xmath180 , in the attractivity - driven polycentric regime .
[ fig : collapse_nk],scaledwidth=45.0% ] extracting the number of activity centers from the zip code business patterns ( zbp ) data is a priori a non - trivial task .
zbp data provide , among other things , the total number of employees per zip code area in the united states every year between 1994 and 2010 . because zip code areas can have different areas , we first decide to normalise the number of employees to the area and obtain the employment density per zip code area .
+ if we further sort the employment densities in decreasing order and plot the employment densities of the different zip code areas as a function of their rank in the ordering ( supplementary figure 3 ) , we obtain a curves which decreases faster than exponentially for the first ranks , and then exponentially .
we interpret this result as there existing a natural scale in the number of subcenters . indeed ,
if the decrease is exponential , one can write for the employment density : where @xmath182 is the typical value of the rank in this situation , and that we would interpret here as the number of subcenters .
however , the rank - plot is not strictly exponential and we therefore define the number of subcenters using a threshold value @xmath183 . if we denote by @xmath184 the highest employee density in the city , we define @xmath185 as : and the threshold value of employment @xmath189 for @xmath190 with two dashed lines .
the number of subcenters is taken to be the number of values of the employment density between those two values.[fig : definition],scaledwidth=45.0% ] in order to plot the population as a function of the number of subcenters , we first extract the number of subcenters for every city in the dataset and for every year between 1994 and 2010 , as indicated in the previous section .
+ we then apply the following treatment to the data : * if a given city corresponds to a single zip code , we only keep one point for that city : its population and @xmath194 center ; * we perform a kolmogorov - smirnov @xcite test between the rank - plots of the employment density in 1994 and 2010 for the remaining cities . if the probability that the two curves are different is greater than a certain threshold @xmath195 , we keep all the points for the city .
otherwise , we just keep a single point . at the end of this process
, we obtain points that can be understood as coming from different realisations of a city .
the following section gives the robustness of the results with regards to the value of the threshold @xmath195 we choose .
the empirical results presented in the main text will a priori depend on the value of the @xmath183 we choose .
we summarize the results obtained when fitting @xmath196 assuming a power - law dependance and a multiplicative noise in table [ table : robustness ] .
we see that for a reasonable range of values for @xmath183 we always have a sublinear behaviour , and that given the estimated variance of the measured exponents , they are all compatible with each other . the exponent derived from the plot on fig .
3 in the main text is very sensitive to the first ( k=1 ) and last point ( k=26 ) .
indeed , a fit excluding these points gives @xmath197 .
this exponent still agrees with the prediction of our model of @xmath198 , and would correspond to a very large value for @xmath199 implying that the traffic dependence in eq .
3 is a sharp threshold function .
this indicates , if anything , that our empirical results should be confirmed through other studies based on different measures and datasets .
griffith , _ the professional geographer _ * 33 * , 189 - 196 ( 1981 )
. v. dokmeci & l. berkoz , _
european planning studies _ * 2 * , 193 - 205 ( 1994 ) .
mcmillen & s.c .
smith , _ journal of urban economics _ * 53 * , 321 - 338 ( 2003 ) .
m. marsili , y .- c .
zhang , _ physical review letters _ * 80*:2741 ( 1998 ) .
m. batty , _ cities and complexity : understanding cities through cellular automata , agent - based models , and fractals _ ( mit press , cambridge , ma , 2005 ) .
|
empirical evidence suggest that most urban systems experience a transition from a monocentric to a polycentric organisation as they grow and expand .
we propose here a stochastic , out - of - equilibrium model of the city which explains the appearance of subcenters as an effect of traffic congestion .
we show that congestion triggers the instability of the monocentric regime , and that the number of subcenters and the total commuting distance within a city scale sublinearly with its population , predictions which are in agreement with data gathered for around 9000 us cities between 1994 and 2010 .
as cities grow , they evolve from monocentric organisations where all the activities are concentrated in the same geographical area usually the central business district to more distributed , polycentric organisations @xcite .
traditional approaches in spatial economics have attempted to describe the phenomenon within the framework of equilibrium models of the city @xcite .
these models are traditionally based on the concept of agglomeration economies to explain why economical activities tend to group and the spatial distribution of wages and rents across the urban space .
however , these approaches fail at giving a satisfactory quantitative account @xcite of the polycentric transition of cities .
first , they describe a city as being in an equilibrium characterised by static spatial distributions of households and business firms . however , the equilibrium assumption is unsupported as cities are out - of - equilibrium systems and their dynamics is of particular interest for practical applications @xcite .
second , these models integrate so many interactions and variables that it is difficult to understand the hierarchy of processes governing the evolution of cities , which ones are fundamental and which ones are irrelevant . yet
, traffic congestion is not explicitly taken into account in the existing models , despite being mentioned in the economics literature as a possible reason for the polycentric transition @xcite .
lastly , the models do not make any quantitative prediction and are therefore unsupported by data .
+ we present in this letter a stochastic , out - of - equilibrium model of the city which relies on the assumption that the polycentric structure of large cities might find its origin in congestion , irrespective of the particular local economic details .
we are able to reproduce many stylized facts and , most importantly , to derive a general relation between the number of activity centers of a city and its population .
finally , we verify this relation against the employment data from around 9000 cities in the u.s . between 1994 and 2010 .
following recent interdisciplinary efforts to construct a quantitative description of cities and their evolution @xcite , we deliberately omit certain details and focus instead on basic processes .
we thereby aim at building a minimal model which captures the complexity of the system and is able to account for qualitative as well as quantitive stylized facts .
the model we propose is by essence dynamical and describes the evolution of cities organisation as their populations increase .
we focus on car congestion mainly due to journey - to - work commutes and its effect on the job location choice for individuals . according to fujita and ogawa s classical model @xcite in spatial economics , an individual living at location
@xmath0 will choose to work at the location @xmath1 that maximises the net income after the deduction of rent and commuting costs @xcite @xmath2 where @xmath3 is the average wage paid by business firms located at @xmath1 ( and thus varies from one location to another ) , @xmath4 is the land rent at @xmath0 , and @xmath5 is the commuting cost between @xmath0 and @xmath1 .
the wage and the land rent result from the interplay between the households and companies locations , agglomeration effects being taken into account .
the commuting cost , on the other hand , does not usually take congestion into account and is taken proportionally to the euclidean distance @xmath6 ( where @xmath7 is the transportation cost per unit of distance ) in most studies .
the time scales involved in the evolution of cities are usually such that the employment turnover rate is larger than the relocation rate of households . on a short time scale
, we can thus focus on the process of job seeking alone , leaving aside the problem of the choice of residence . in other words
, we assume the coupling between both processes to be negligible : we assume that each inhabitant newly added to the city has a random residence location and we concentrate on understanding how such an inhabitant chooses its job among a pool of @xmath8 potential activity centers ( which we suppose are also randomly distributed among the city ) .
the active subcenters are then defined as the subset of potential centers which have a nonzero incoming number of individuals . as a result of these assumptions , a worker living at @xmath0 will choose to work at the center @xmath1 such that the quantity @xmath9 is maximum .
+ we now discuss the form of the two terms @xmath3 and @xmath10 .
the problem of determining the ( spatial ) variations of the average wage @xmath3 at location @xmath1 is very reminiscent of some problems encountered in fundamental physics .
indeed , the wage depends on many different factors , ranging from the type of company , the education level of the inhabitant , the level of agglomeration , etc . ,
and in this respect is not too different from quantities that can be measured in a large atom made of a large number of interacting particles .
in this situation , physicists found out that although it is possible to write down the corresponding equations , not only is it impossible to solve them , but it is also not really useful .
in fact , they found out @xcite that a statistical description of these systems , relying on random matrices could lead to predictions that agree with experimental results .
we wish to import in spatial economics this idea of replacing a complex quantity such as wages , which depends on so many factors and interactions , by a random one .
we therefore decide to account for the interaction between activity centers and people by taking the wage as proportional to a random variable @xmath11 $ ] such that @xmath12 where @xmath13 defines the maximum attainable average wage in the considered city . as for the transportation cost @xmath5
, we choose it to be proportional to the commuting time between @xmath0 and @xmath1 . in a typical situation where passenger transportation is dominated by personal vehicles , this commuting time not only depends on the distance between the two places but also on the traffic between @xmath0 and @xmath1 , the vehicle capacity of the underlying network , and its resilience to congestion .
the bureau of public road formula @xcite proposes a simple form taking all these factors into account . in our framework , it leads to the following expression for the commuting costs @xmath14\ ] ] where @xmath15 is the traffic per unit of time between @xmath0 and @xmath1 and @xmath16 is the typical capacity of a road ( taken constant here ) .
the quantity @xmath17 is a parameter quantifying the resilience of the transportation network to congestion .
we further simplify the problem by assuming that the traffic @xmath15 is only a function of the subcenter @xmath1 and therefore write @xmath18 the total traffic incoming in the subcenter @xmath1 ( see supplementary material @xcite for a short discussion ) . in summary ,
our model is defined as follows . at each time step , we add a new individual @xmath0 located at random in the city , who will choose to work in the activity area @xmath1 ( among @xmath8 possibilities located at random ) such that the following quantity @xmath19 \label{eq : cost_function}\ ] ] is maximum ( we omitted irrelevant multiplicative factors ) .
the quantity @xmath20 is interpreted as the maximum effective commuting distance that people can financially withstand . depending on the relative importance of wages , distance and congestion , the model predicts the existence of three different regimes : the monocentric regime ( top fig .
[ fig : model_results ] ) , the distance - driven polycentric ( middle fig .
[ fig : model_results ] ) regime and the attractivity - driven polycentric ( bottom fig .
[ fig : model_results ] ) regime . ,
title="fig:",scaledwidth=40.0% ] , title="fig:",scaledwidth=40.0% ] , title="fig:",scaledwidth=40.0% ] from now on , we will assume that @xmath21 is large enough so that a monocentric state exists for small values of the population .
in this regime , the value of @xmath22 prevails and the monocentric state evolves to an attractivity - driven polycentric structure as the population increases ( if @xmath21 is too small , the monocentric regime does not exist see the supplementary material @xcite for more details on these points ) . starting from a small city with a monocentric organisation ,
the traffic is negligible and @xmath23 which implies that all individuals are going to choose the most attractive center ( with the largest value of @xmath24 , say @xmath25 ) .
when the number @xmath26 of households increases , the traffic will also increase and some initially less attractive centers ( with smaller values of @xmath22 ) might become more attractive , leading to the appearance of new subcenters characterized by a nonzero number of commuters . more precisely , a new subcenter @xmath1 will appear when for an individual @xmath0 , we have @xmath27 .
the traffic so far is @xmath28 and @xmath29 which leads to the equation @xmath30\ ] ] we assume that there are no spatial correlations in the subcenter distribution , so that we can make the approximation @xmath31 .
the new subcenter will thus be such that @xmath32 is minimum implying that it will have the second largest value denoted by @xmath33 . for a uniform distribution ( details of the calculation
can be found in the supplementary material @xcite , section 2 ) , on average @xmath34 leading to a critical value for the population @xmath35 whatever the system considered , there will therefore _ always _ be a critical value of the population above which the city becomes polycentric ( which can be smaller than one , in which case there is no monocentric regime at all , see the supplementary material @xcite ) .
the monocentric regime is therefore fundamentally unstable with regards to population increase , which is in agreement with the fact that no major city in the world exhibits a monocentric structure .
we note that the smaller the value of @xmath17 ( or the larger the value of the capacity @xmath16 ) , the larger the critical population value @xmath36 which means that cities with good road systems are capable of absorbing large traffic show a monocentric structure for a longer period of time .
having established that cities will eventually adopt a polycentric structure , we can wonder how the number of subcenters varies with the population .
we compute the value of the population at which the @xmath37 center appears .
we still assume that we are in the attractivity - driven regime and that , so far , @xmath38 centers have emerged with @xmath39 @xcite , with a number of commuters @xmath40 , respectively .
the next worker @xmath0 will choose the center @xmath41 if @xmath42 } z_{ij}\ ] ] which reads @xmath43 } \left\ { \eta_j - \frac{d_{ij}}{\ell } \left [ 1 + \left ( \frac{t(j)}{c}\right)^\mu\right ] \right\}\ ] ] the distribution of traffic @xmath44 is narrow @xcite , which means that all the centers have roughly the same number of commuters @xmath45 . as above
we also assume that the distance between the workers households and the activity centers is typically @xmath46 .
the previous expression now reads @xmath47 } \left ( \eta_j \right ) - \eta_k\ ] ] following our definitions , @xmath48 } \left ( \eta_j \right ) = \eta_1 $ ] . according to order statistics ,
if the @xmath24 are uniformly distributed , we have on average @xmath49 .
it follows from these assumptions that ( 1 ) the @xmath37 center to appear is the @xmath37 most attractive one ( 2 ) the average value of the population @xmath50 at which the @xmath37 center appears is given by : @xmath51 conversely , the number @xmath41 of subcenters scales sublinearly with population as @xmath52 it is interesting to note that this result is robust : the dependence is sublinear , _ whatever the distribution _ of the random variable @xmath22 ( see the supplementary material @xcite for a discussion on this point ) .
we can therefore conclude that , probably very generally and under mild assumptions , the number of activity subcenters in urban areas scales sublinearly with their population where the prefactor and the exponent depend on the properties of the transportation network of the city under consideration .
a previous study @xcite showed that the total miles driven daily in a city the ` total commuting distance'scales with the population as @xmath53 where @xmath54 0.5 , 1\right[$ ] , which the authors interpreted as cities having neither totally centralized nor totally decentralized structures .
we can discuss this result within the framework of our model in the following way .
if the system was in the pure attractivity - driven regime , we would have @xmath55 .
but , if we assume that we are in an intermediate regime where eq .
[ eq : prediction ] holds , and where the system exhibits spatial coherence @xcite , we can write the total length of the commutes as @xmath56 inverting the result from eq .
[ eq : prediction ] we therefore get @xmath57 where @xmath58 $ ] .
our model is thus consistent with the fact that the total traveled miles scales with population with a non - trivial exponent comprised in @xmath59 $ ] .
+ we now test the prediction given by eq .
( [ eq : prediction ] ) . for that purpose , we collected data for the number of employees per zip code in the united states that were collected over a span of 16 years , between 1994 and 2010 @xcite , as well as the population of all cities in the us between 1994 and 2010 @xcite .
we estimate the number of subcenters by constructing the rank plot of the employment density @xmath60 ( number of employees per @xmath61 ) for each zip code of a given urban area @xcite .
these plots display a decay as fast as an exponential ( fig . [ fig : rank_plots ] ) which implies that there exists a natural scale for the rank , that we interpret here as the typical number of activity centers .
it also implies that any reasonable method should give an estimate of the number of subcenters of the same order of magnitude ( which would not be the case for slowly decaying functions such as power laws for example ) .
we first note that for some cities typically large ones with stable populations ( fig .
[ fig : rank_plots ] , left) the employment spatial statistics remained stable over the period of study .
for other cities , we observe large variation of the number of subcenters ( fig .
[ fig : rank_plots ] , right ) .
we then plot ( fig .
[ fig : data ] ) the population @xmath26 of cities ( with population @xmath62 ) versus the estimated number of subcenters @xmath41 ( the dispersion in the scatter plot probably results from the fact that different cities have different resilience levels to congestion ) . on average ,
we observe a power law dependence with exponent @xmath63 ( the result is robust with regards to the estimate of @xmath41 , see the supplementary material @xcite for more details ) .
inverting this relation gives us the number of subcenters as a function of the population @xmath64 with @xmath65 .
this result is strikingly in agreement with the prediction given by our model : the number of subcenters in a city scales sublinearly with its population . + ( @xmath66 ) .
see the supplementary material @xcite for more details on the computation of @xmath41 and the robustness of the results .
[ fig : data],scaledwidth=50.0% ] using the measured value of @xmath67 and eq .
( [ eq : beta ] ) we can estimate the exponent of the scaling of @xmath68 with the population and find @xmath69 which agrees very well with the value @xmath70 measured in @xcite directly on the data of the daily total miles driven in more than @xmath71 cities in the us .
+ while agglomeration economies seem to be the basic process explaining the existence of cities and their spectacular resilience , this study brings evidence that congestion is the driving force that tears them apart .
the nontrivial spatial patterns observed in large cities can thus be understood as a result of the interplay between these competing processes .
we believe that the present model represents an important step towards a quantitative , predictive science of cities .
more generally , this microscopic approach is an interesting example of an out - of - equilibrium model : it is governed by local optimization with saturation effects , leads to different regimes , and is characterized by nontrivial dynamical exponents . in this respect ,
we believe that this discrete approach might be of use in the study of pattern formation in biology which has been so far explored from a global optimisation perspective @xcite or used as a coarse - grained approach to reaction - diffusion equations with a density - dependent diffusion coefficient @xcite to compute quantities that are out of reach within the current methods .
acknowledgements mb thanks h. berestycki , a. flammini , and k. mallick for stimulating discussions .
mb acknowledges funding from the eu commission through project eunoia ( project no .
fp7-dg.connect-318367 ) .
|
the maximum resolution achievable with conventional optical techniques is determined by the classical diffraction limit .
the minimum optical spot diameter can be expressed as @xmath0 , where @xmath1 is the wavelength in air , and @xmath2 is the numerical aperture(n is the refractive index and @xmath3 is the convergence semiangle ) .
fortunately , the diffraction limit can be circumvented by use of scanning near - field optical systems , where resolutions less than 50 nm can be achieved . unfortunately , near - field techniques have been troubled by low transmission efficiencies , and therefore poor signal to noise ratios .
although recent research have improved the transmission efficiency considerably , it is still only @xmath4 at 100 nm spot size@xcite .
another way to increase the resolution is by application of a solid immersion lens(sil)@xcite .
this method is currently not capable of the same resolution as the near - field techniques , but the light transmission efficiency is considerably better .
the aim of the present paper is to describe how the polarization , phase and amplitude at the exit pupil influences the intensity distribution near the focal plane of a sil .
although this is not a new issue in optics(see ref.@xcite and references therein ) , we believe it is of interest to gain further understanding of sil - systems as they may become an integral part of the future data storage systems . since our focusing system has high na and the focal point
is placed many wavelengths away from the aperture , the diffracted field near the focal plane can be calculated using the debye approximation@xcite .
let us consider focusing through a dielectric interface(see fig .
the electric field inside medium @xmath5 can be written as@xcite : @xmath6 ds_{ix}ds_{iy } \label{a}\ ] ] where @xmath7 is the wavenumber , @xmath8 is the unit vector along a typical ray , @xmath9 is the solid angle formed by all the geometrical rays , @xmath10 is the vector pupil distribution , which accounts for the polarization , phase and amplitude distribution at the exit pupil .
we can find the electric field near the focal plane by matching the fields in the first and second medium at the interface , @xmath11 .
the resulting electric field in the second medium becomes@xcite : @xmath12exp[ik_{2}s_{2z}z ] exp[ik_{1}(s_{1x}x+s_{1y}y ) ] ds_{1x}ds_{1y } \label{c}\ ] ] @xmath13 is a complex constant(which will be ignored in the rest of this paper ) .
the unit wave - vector is defined in spherical coordinates : @xmath14 \label{d}\ ] ] the position vector can be written as(see also fig .
[ f1 ] ) : @xmath15 \label{e}\ ] ] this gives the following diffraction integral : @xmath16 sin(\theta_{1})d\theta_{1 } d\phi \label{f}\ ] ] @xmath17 where @xmath3 is the convergence semiangle , @xmath18 is the polarization and @xmath19 represents the amplitude and phase distribution at the exit pupil .
we will for the rest of this paper assume that our optical system obeys the sine condition , @xmath20 , see e.g. ref.@xcite .
@xmath21 and @xmath22 \label{i}\ ] ] @xmath23 represents the aberration function introduced due to the mismatch in refractive index . a detailed derivation of this integral was first presented in reference@xcite .
we have presented it in a slightly different form to enlight the further discussion in this paper .
we will consider two particular cases : 1)focusing of electromagnetic waves in a homogeneous media(air ) . in this case @xmath24 .
2)focusing of electromagnetic waves with a sil , which means that d=0(note that @xmath25 ) . in both cases
the aberration function @xmath23 is identically zero@xcite .
the state of the polarization incident on the focusing system will influence the resolution near the focal plane(see e.g. @xcite ) . to discuss this question quantitatively ,
we must find a general expression for the polarization vector .
we assume a incident polarization which may in general depend on the polar and azimuthal angle : @xmath26\ ] ] the polarization vector can be written as@xcite : @xmath27^{-1}il^{(1)}crp_{0}\ ] ] @xmath28\ ] ] which describes the rotation of the co - ordinate system around the optical axis ; @xmath29\ ] ] which describes the change of polarization on propagation through the lens ; @xmath30\ ] ] which describes a rotation of the co - ordinate system into s and p - polarized vectors ; @xmath31\ ] ] which represents the transmission(fresnel coefficients ) through the plane dielectric interface .
this results in : @xmath32 + b[t_{p}cos(\theta_{2})sin(\phi)cos(\phi)-t_{s}sin(\phi)cos(\phi ) ] \\
a[t_{p}cos(\theta_{2})cos(\phi)sin(\phi)-t_{s}sin(\phi)cos(\phi ) ] + b[t_{p}cos(\theta_{2})sin^{2}(\phi)+t_{s}cos^{2}(\phi ) ] \\ - at_{p}sin(\theta_{2})cos(\phi)-bt_{p}sin(\theta_{2})sin(\phi ) \\ \end{array } \right]\ ] ] in the special case of x - polarized incident light(@xmath33 and b=@xmath34 ) this matrix reduces to@xcite ; @xmath35\ ] ] it is seen that different angles have different amplitudes , and that the polarization is dependent on @xmath36 .
another possibillity was explored by quabis et .
@xcite , who found that radial polarized light may increase the resolution .
we extend their analysis to a sil system .
we may write @xmath37 and @xmath38 : @xmath39\ ] ] when inserted in eq .
( [ f ] ) , this gives : @xmath40 @xmath41 @xmath42 @xmath43 @xmath44 where @xmath45 , and @xmath46 is the bessel function of the first kind , of order @xmath47 .
the z - component is completely independent of the azimuthal angle , and its importance increases with increasing na .
let us assume that we are able to find a medium which is only sensitive to the z - component of the polarization .
that is , we may use the z - component , and disregard the x and y - components .
[ f2 ] shows the energy density at the bottom surface(the focal plane ) of a sil with @xmath48(and @xmath49 ) , for both radial(solid line ) and x - polarized incident light(the dashed line represents @xmath50 , whereas the dotted line represents @xmath51 ) . in this example
, we assume that @xmath52 and the wavelength @xmath53
[ f2 ] indicates that it is possible to increase the resolution , and perhaps reproduce the result predicted by scalar theory .
however , it is worth noting that in practice one must combine the radial polarizer in front of the focusing system with an annular aperture , both to increase resolution and to avoid a singularity on the axis .
this was discussed by quabis et .
therefore , if one could find a material which is sensitive only to the z - component , a more detailed calculation using an annular aperture is required to determine the real resolution .
the spatial amplitude and phase distribution(at the exit pupil ) will also influence the energy distribution near the focal plane(see e.g. @xcite ) . as with the incident polarization , there are countless ways to alter the phase and transmittance .
we will limit our discussion to the special class of amplitude / phase filters called annular apertures(see ref.@xcite and references therein ) .
annular apertures that modifies the transmittance and/or phase distribution at the exit pupil may improve the resolution at the expense of higher sidelobes . in this section
we will discuss the effect of such an aperture in front of a focusing system containing a sil .
the transmittance through an annular aperture can be expressed as : @xmath54 where @xmath55 are complex constants(phase and amplitude ) for the various zones in the aperture , and @xmath45 .
if the incident light is polarized in the x - direction(a=1 , b=0 ) , the resulting field can be found by inserting the transmittance into eq .
( [ f ] ) : @xmath56 \label{o}\ ] ] @xmath57 @xmath58 where @xmath59 @xmath60 @xmath61 @xmath62 @xmath63 @xmath64 where @xmath65 is defined to be zero , and we have used that @xmath66 to derive these equations .
similar expressions can be applied to the case of radial polarized light . from eqs .
( [ r ] ) , ( [ s ] ) and ( [ t ] ) we can see that the total field is a sum of electric fields from zones with increasing angular extent , and we expect that the terms with the smallest @xmath67 must be carefully balanced against each other in order to reduce the energy in the sidelobes . for high na systems , the z - component becomes particularly important , and may increase the sidelobe intensity substantially . thus it is necessary to keep the center peak ratio high in order to avoid a dominating z - component , which may increase the the sidelobes as well as the spot size . in total
one must not only balance the terms from the various zones , buth also keep the sidelobes due to the z-(and y)component at an acceptable level .
this is a difficult task , and is best treated by numerical analysis , e.g. by binary search methods@xcite . to keep the physics simple ,
we will limit ourselves to apertures with two and three zones .
the profiles(the time - averaged energy density distributions ) will be normalized , for simple comparison between systems . in order to compare with the situations occuring in a sil - system ,
we first observe what happens when we place a three - zone aperture in front of a system focusing in a homogeneous media .
as pointed out by ando@xcite , the center - peak intensity ratio can be maximized by using a phase aperture , and the sidelobes can be made small by maintaining the same phase for the light passing through the center and outer portion of the aperture .
on the basis of these results , let us assume that @xmath68 , @xmath69 and @xmath70 .
such an aperture can be produced by e.g. lithographical methods , and was experimentally tested by ando et . al.@xcite .
let @xmath71 , @xmath72 , @xmath73 and @xmath53 .
[ f3 ] shows the focused beam profile for the three - zone phase aperture when @xmath50(solid line ) and @xmath51(dash - dotted line ) . for comparison
we have also plotted the profile of the focused beam with no aperture when @xmath50(dashed line ) and @xmath51(dotted line ) .
the resolution due to the three - zone aperture is improved as compared to no aperture , but the peak sidelobe intensity is almost @xmath74 .
we could probably bring this number down by increasing the number of zones in the annular aperture .
[ f4 ] shows the axial energy density distribution for the three - zone annular aperture(solid line ) and no annular aperture(dashed line ) .
note that the focal depth is larger with the three - zone aperture .
the last ten years much research has been done on sil systems@xcite . as mentioned in the introduction , a major goal has been to improve the resolution and focal depth . to that end ,
an interesting possibility is the application of an annular aperture in front of a sil .
as have been pointed out by milster @xmath75 @xmath76@xcite , plane waves incident on the bottom surface of the sil will experience total reflection above the critical angle , @xmath77 .
thus we may divide the plane waves at the exit pupil into two parts ; a homogeneous part where the angles are smaller than the critical angle at the interface , and an inhomogeneous part where the plane waves experience total reflection at the interface , corresponding to evanescent waves below the bottom surface of the sil .
let us first see what happens when we put a simple two - zone annular aperture with @xmath78 and @xmath79 in front of the sil .
the dotted lines in figs .
[ f5 ] and [ f6 ] show the transverse profiles behind the bottom surface of the sil(@xmath80 , @xmath81 and @xmath53 ) when @xmath50 and @xmath51 , respectively .
we assume that @xmath82 and @xmath83 . for comparison
, the profiles for a sil with no annular aperture are also shown(the dashed lines in figs .
[ f5 ] and [ f6 ] ) .
note that as the center(dark ) disk increases , the profile becomes more compressed , but the sidelobes increases .
the increase in resolution is more pronounced when @xmath51 than @xmath50 , since the z - component of the electric field depends on @xmath84(which is zero when @xmath51 , see eq .
( [ q ] ) ) .
[ f7 ] shows the axial distribution behind the sil with(dotted line ) or without(dashed line ) a two - zone annular aperture .
when the center disk increases , the focal depth decreases , contrary to what happens during focusing in a homogeneous media@xcite .
this behaviour occur since the evanescent waves becomes more dominating when we block out the homogeneous waves , and since the amplitude of evanescent waves are decreasing exponentially behind the sil . to increase the light transmission efficiency , one may replace the two - zone amplitude aperture with an axicon , thus creating a bright ring .
next we place a three - zone phase aperture(with the same properties as in the previous section ) in front of the sil . in particular , we are interested in observing what happens when 1)@xmath85 and 2)@xmath86 .
the first case(@xmath85 ) is in some ways similar to focusing in a homogeneous media .
that is , we expect the spot size to decrease , if the convergence semiangles have proper values .
however , we also expect the focal depth to decrease , since we balance terms with homogeneous waves against each other , thus increasing the importance of the evanescent waves . as an example , let us consider the same sil - system as above , but now replace the two - zone aperture with a three - zone annular aperture with @xmath87 , @xmath72 and @xmath73 . the dash - dotted lines in figs .
[ f5 ] and [ f6 ] shows the resulting profiles when @xmath50 and @xmath51 , respectively .
note that the profile is slightly smaller than the dashed line(no annular aperture ) when @xmath51 , whereas for @xmath50 it is slightly larger .
the dash - dotted line in fig .
[ f7 ] shows the axial distribution .
as expected , the focal depth is smaller with the annular aperture . in the second case(@xmath88 )
the opposite behaviour may take place .
that is , we are able to balance terms with inhomogeneous waves against each other , thus increasing the importance of the homogeneous waves . in this way we can produce some kind of enhancement of the focal depth , and
if the the angles @xmath89 and @xmath90 are properly chosen , the beam profile may become narrower as well . as an example we consider an annular aperture with @xmath91 and @xmath92 .
the solid lines in figs .
[ f5 ] and [ f6 ] are representing the transverse profiles when @xmath50 and @xmath51 , respectively .
now the situation is almost opposite compared with the previous annular aperture .
that is , the profile is smaller than the dashed line(no aperture ) when @xmath50 , but larger when @xmath51 . the axial distribution(solid line in fig .
[ f7 ] ) is decaying more slowly than with no annular aperture , which confirms the discussion above .
the choice of incident polarization , phase and amplitude may change the intensity distribution near the focal plane of a focusing system .
it was seen that by impinging radial polarization on a sil system , we may obtain a rotational symmetric z - component of the focused wavefront with spot size similar to that predicted by scalar theory .
we also discussed how annular apertures may change the resolution and focal depth .
it was shown that when we place a three - zone phase aperture in front of a focusing system , both the resolution and focal depth may increase . when applied to sil - systems
, we observed that it was possible to manipulate the contributions from the homogeneous and inhomogeneous waves behind the sil by changing the phase and transmittance distribution at the aperture . in this way it may be possible to improve both the resolution and focal depth .
future investigations may be directed towards the use of more than three zones to improve the resolution and focal depth .
we would like to point out that only a few special cases were discussed in this paper , and that our examples were not optimized for any particular applications .
i would like to thank all the people working in the superconductivity and magnetooptics group at the university of oslo for creating a very active and inspiring enviroment , which resulted in the ideas presented in this work .
the research was financially supported by the norwegian research council .
|
by altering the polarization , phase and amplitude at the exit pupil , the intensity distribution near the focal plane of a solid immersion lens(sil ) system can be changed .
we have studied how the resolution and focal depth changes for a few particular cases .
it was seen that by impinging radial polarization on a sil system , we may obtain a rotational symmetric z - component of the focused wavefront with spot size similar to that predicted by scalar theory .
we also observed that it was possible to manipulate the contributions from the homogeneous and inhomogeneous waves behind the sil by changing the amplitude and phase distribution at the aperture . in this way it may be possible to improve both the resolution and focal depth of the system .
|
( hd 161797 = hip 86974 = wds 17465 + 2743 ) is the third closest quadruple star system to the sun @xcite at a distance of 8.3 pc @xcite .
the system consists of the g5iv primary @xcite surrounded by several m dwarfs .
the closest companion is the ab component , which has long been known from astrometric variations of the primary @xcite and from radial velocity ( rv ) variations @xcite .
the companion was first imaged by @xcite with the adaptive optics ( ao ) system on the 2.5 m telescope at mt .
wilson observatory .
there is also the bc pair , which is separated from the primary by 35 and is a pair of m dwarfs in a [email protected] yr period @xcite .
both @xcite and @xcite concluded that the pair was physically bound to @xmath0 her due to common proper motion .
there is also a wider d component with a separation of several hundred arcseconds , which is also an m dwarf .
@xcite concluded that this is not a physical companion because the proper motion differs dramatically from that of the primary .
while binary stars are common , 50%@xmath24% of f6-g2 stars have companions @xcite , quadruples are much less common .
quadruples come in two major catagories , 3@xmath41 systems , where a fourth star orbits a triple system in a wide orbit and , 2 + 2 systems , which consist of two binaries orbiting a common center of mass .
her is the latter type , which is the more common type @xcite .
@xcite found that 10% of nearby solar - type stellar binaries were actually 2 + 2 quadruples and this is likely linked to stellar system formation processes .
the @xmath0 her system is one of the best - studied 2 + 2 quadrupole systems ; therefore it serves as an archetype for understanding stellar system formation .
improving the characterization of this system is the motivation for this analysis .
the primary star in the system exhibits solar - like oscillations @xcite and these have been used to measure the mass of the primary @xcite using asteroseimology ( see section [ orbit_analysis ] ) .
a mass determination from an orbital analysis will serve as a useful crosscheck between these two techniques and can be used to improve the theoretical modeling of observed oscillation frequencies @xcite . the nature of the @xmath0 her ab companion has been discussed for over 20 years .
most indications are that it is an m - dwarf , but there are persistent possibilities that it is substellar .
@xcite used a monte carlo technique to compute probability density functions of the mass of the close companion to @xmath0 her a based on limited rv data and hypothetical , future , single measurements of the angular separation .
the study concluded that if the separation was 14 then the companion would have a probable mass on the boundary between substellar and stellar .
the angular separation measurements of @xcite fell at the boundary between the two regimes . from the computed @xmath5-band magnitude of the companion
, they determined that the companion had to be redder than an m5v and could be substellar as @xcite suggested .
@xcite also observed the companion in the near - infrared with the mt .
wilson ao system .
they were able to measure the photometry of the companion in @xmath6 and @xmath7 filters . using the _ ri _ photometry from @xcite and their _
hk _ photometry and model atmospheres they conclude that the mass should be @xmath80.13 @xmath9 ( equivalent to an m5v ; @xcite ) .
they note that the _ ri _ photometry is anomalously bright compared to the model , though the near - infrared photometry agreed with the models to within the error bars .
they concluded that the companion is stellar .
@xcite observed the system with the mmt in the mid - infrared and estimated the spectral class of the companion to be m4@xmath21 . combining their results with that of @xcite
, they computed the ( _ k_-_m _ ) color and noted that it was too red for an m4v star , but was similar to an early t - dwarf . with the possibility that the companion was a brown dwarf , we set out to resolve the prior conflicting color information and determine its spectral type with near - infrared spectroscopy .
we also collected multi - epoch astrometry to compute an resolved orbit for the system , allowing for the estimation of the individual masses of the objects .
we observed @xmath0 her at the advanced electro - optical system ( aeos ) 3.6 m telescope with the site ao system and the visible imager ( visim ) camera @xcite on two separate dates : 2003 july 6 and 2005 april 24 . both data sets consist of 1000 frames using a bessel _ i_-band filter .
after collection , any saturated frames were discarded and the remaining frames were debiased , dark subtracted and flat fielded .
the frames were weighted by their peak pixel , which is proportional to their strehl ratio and then co - added using a shift - and - add routine .
the resulting image was analyzed with the program _
fitstars _ @xcite .
the 2005 image is shown in figure [ images]b .
error bars on the astrometry and photometry were assigned using the method in @xcite .
the extracted astrometry and differential photometry are given in table [ astrometry ] .
the table lists the our new results as well as prior measures of the system .
the table gives the ut besselian date of the observation , the measured position angle ( @xmath10 ) , the measured separation ( @xmath11 ) , the differential magnitude and the filter used for the observation .
the final column is either the instrument that was used for the observation or a reference if it was a literature result .
the table does not include null results from the literature .
we used the project 1640 ( p1640 ) coronagraph @xcite mounted on the palm-3000 ao system @xcite at the palomar 5 m telescope to observe @xmath0 her three times in 2012 and 2015 . during the observations
the primary was placed behind the occulting disk .
the first 2012 data set used the astrometric grid spots @xcite , allowing the measurement of the astrometry of the companion even though the primary is occulted by the coronagraph .
two nights later on 2012 june 14 ut , we collected an additional data set , but this time the astrometric grid spot was not used and we were unable to measure the astrometry of this data set .
our final observation was on 2015 march 31 ut and it used the astrometric grid spots .
all the data were reduced using the project 1640 data reduction pipeline described in @xcite . since then
, a few upgrades have been made to the pipeline that reduce lenslet - lenslet cross talk .
the pipeline processing produces an image of the object in each of 32 wavebands , resulting in a data cube .
the astrometric grid spots produce four spots on each frame of the data cube .
the location of the spots is dependent on the wavelength of the frame . to measure the astrometry
we fit a gaussian to each of the four astrometric grid spots .
then we identified the location of the primary by fitting lines to the vertical and the horizontal spots .
the primary is located where these two lines intersect .
the final astrometry is the average of all the astrometry measured from 29 of the data slices from both of the image cubes with grid spots .
we only use 29 image slices because several images in the atmospheric water bands did not produce meaningful results due to poor signal .
the plate scale and position angle offset of the detector were computed from observations of calibration binaries taken during each observing run .
the astrometry is presented in table [ astrometry ] .
we also collected near - ir images of the system on multiple dates using the pharo near - infrared camera @xcite .
like p1640 , pharo is also mounted on the palm-3000 ao system on the 5 m hale telescope .
we collected data on 29 august 2013 ut , 14 may 2014 ut and 31 march 2015 ut . all observations used the 25 field of view with a pixel scale of 25 mas pixel@xmath12 .
the observations used a variety of near - ir filters .
after collection , each frame was debiased , sky subtracted , and flat fielded .
we again used the @xmath13 algorithm to measure the astrometry and photometry of the binary and assigned error bars using the technique described in @xcite we also examined archival pharo observations .
@xcite observed @xmath0 her in 2002 june with pharo in coronagraphic mode in _ ks _ filter .
it was not originally detected with the automated detection algorithm due to nearby residual light from the central star s characteristic `` waffle pattern '' , which the algorithm interpreted as a high noise level in that localized region of the image .
a visual inspection of the archival image clearly shows the companion .
the companion s relative astrometry was extracted by visual inspection .
uncertainties were dominated by the ability to determine the center star s position behind the coronagraphic mask .
we conservatively assigned a 2 pixel ( 008 ) uncertainty in the center star s position ( see discussion in @xcite ) .
we located images of @xmath0 her aa , ab in the archives of the keck ii telescope .
the keck ii data were collected with nirspec s slit camera ( scam ) @xcite in ao mode on 2001 may 15 ut with the nirspec-7 filter with a central wavelength of 2.222 and a fwhm of 0.805 .
we were unable to identify any calibration data for scam for the 2001 data .
in the images the star was placed in different positions across the image . to calibrate the images
, we fit a plane to each image and subtracted it off .
we then subtracted off a median of all the deplaned images .
this resulted in 12 images .
we then measured the astrometry of each image using _
fitstars_. one of the images is shown in figure [ images]a .
we set the error bars equal to the standard deviation of the measurements .
the resulting astrometry and photometry are given in table [ astrometry ] .
her aa , ab showcasing an image from each of the instruments used in this paper .
the instruments and dates of the images are : a ) nirspec / scam from 2001 , b ) visim from 2005 , c ) p1640 from 2012 , and d ) pharo from 2013 .
all images have been rotated so that north is up and east is to the right .
orbital motion of @xmath0 her ab is clearly seen as it rotates clockwise .
each image has a horizontal size of approximately 34 .
the p1640 image is an occulted image and is a single slice of the image cube at a wavelength of 1.720 . in the pharo image ,
the close object to the lower left of the primary is a ghost from a neutral density filter in the camera .
, title="fig:",width=151,height=151 ] her aa , ab showcasing an image from each of the instruments used in this paper .
the instruments and dates of the images are : a ) nirspec / scam from 2001 , b ) visim from 2005 , c ) p1640 from 2012 , and d ) pharo from 2013 .
all images have been rotated so that north is up and east is to the right .
orbital motion of @xmath0 her ab is clearly seen as it rotates clockwise .
each image has a horizontal size of approximately 34 .
the p1640 image is an occulted image and is a single slice of the image cube at a wavelength of 1.720 . in the pharo image ,
the close object to the lower left of the primary is a ghost from a neutral density filter in the camera .
, title="fig:",width=151,height=151 ] her aa , ab showcasing an image from each of the instruments used in this paper .
the instruments and dates of the images are : a ) nirspec / scam from 2001 , b ) visim from 2005 , c ) p1640 from 2012 , and d ) pharo from 2013 .
all images have been rotated so that north is up and east is to the right .
orbital motion of @xmath0 her ab is clearly seen as it rotates clockwise .
each image has a horizontal size of approximately 34 .
the p1640 image is an occulted image and is a single slice of the image cube at a wavelength of 1.720 . in the pharo image ,
the close object to the lower left of the primary is a ghost from a neutral density filter in the camera .
, title="fig:",width=151,height=151 ] her aa , ab showcasing an image from each of the instruments used in this paper .
the instruments and dates of the images are : a ) nirspec / scam from 2001 , b ) visim from 2005 , c ) p1640 from 2012 , and d ) pharo from 2013 .
all images have been rotated so that north is up and east is to the right .
orbital motion of @xmath0 her ab is clearly seen as it rotates clockwise .
each image has a horizontal size of approximately 34 .
the p1640 image is an occulted image and is a single slice of the image cube at a wavelength of 1.720 . in the pharo image ,
the close object to the lower left of the primary is a ghost from a neutral density filter in the camera . , title="fig:",width=151,height=151 ] lccccl 1998.5359 & 167@xmath23 & [email protected]&[email protected] & _ i _ & @xcite + 1998.5360 &
164@xmath23 & [email protected]&[email protected] & _ r _ & @xcite + 2000.38 & [email protected] & [email protected] & [email protected] & _ br@xmath14 _ & @xcite + 2000.77 & [email protected] & [email protected] & [email protected] & _ br@xmath14 _ & @xcite + 2001.3678 & [email protected] & [email protected] & [email protected] & n-7 & nirspec / scam + 2001.80 & 184@xmath28 & [email protected] & [email protected] & _ h _ & @xcite + 2001.80 & 184@xmath28 & [email protected] & [email protected] & _ k _ & @xcite + 2002.4622 & 190@xmath22 & [email protected] & ... & _ ks _ & pharo + 2003.5112 & [email protected] & [email protected] & [email protected] & _ i _ & visim + 2005.3126 &
[email protected] & [email protected] & [email protected] & _ i _ & visim + 2006.28070 & [email protected] & [email protected] & ... & _ m _ & @xcite + 2006.28072 & [email protected] & [email protected] & [email protected] & _ m _ & @xcite + 2012.4473 & [email protected] & [email protected] & ... & _ yjh _ & p1640 + 2013.7426 & [email protected]&[email protected] & [email protected] & _ br@xmath14 _ & pharo + 2014.3669 & [email protected]&[email protected] & [email protected] & _ j _ & pharo + 2014.3669 & [email protected]&[email protected] & [email protected] & _ h _ & pharo + 2014.3669 & [email protected]&[email protected] & [email protected] & _ ch@xmath15 short _ & pharo + 2014.3669 & [email protected]&[email protected] & [email protected] & _ kcont _ & pharo + 2014.3669 & [email protected]&[email protected] & [email protected] & _ ks _ & pharo + 2014.3669 & [email protected]&[email protected] & [email protected] & _ br@xmath14 _ & pharo + 2015.2432 & [email protected]&[email protected] & [email protected]&_br@xmath14 _ & pharo + 2015.2433 & [email protected]&[email protected] & [email protected]&_j _ & pharo + 2015.2433 & [email protected]&[email protected] & [email protected]&_ch@xmath15 short _ & pharo + 2015.2459 & [email protected]&[email protected] & ... & _ yjh _ & p1640
to create a spectrum of the stellar companion , we performed aperture photometry on each of the 32 images in the p1640 data cubes .
this was done with the _ aper.pro _ routine which is part of the idl astrolib and is an adaptation of _ daophot _ @xcite .
a photometry aperture and a sky annulus were centered on the companion .
the radii of the apertures were set equal to 3.5@xmath16 rounded to the nearest pixel size , where @xmath17 is the central wavelength of each image and @xmath18 is the aperture of the telescope .
the annulus size was set to avoid the primary star and the occultation spot .
the spectrum is the measured power in each slice as function of wavelength .
the extracted spectrum is a convolution of the object spectrum , the instrumental spectral response function ( srf ) and the atmospheric srf @xcite . in some of our previous results ,
we have computed the combined srf of the atmosphere and instrument from unocculted images of the primary @xcite .
@xmath0 her a is so bright that it saturates our detector on even the shortest exposures .
we computed the srf from another star . on 2012
june 12 ut , there were no other suitable stars to extract an srf , so we used the srf extracted from hd 129814 on 2012 june 14 ut .
the airmass of the @xmath0 her observations on 12 june was 1.02 , while the airmass of hd 129814 was 1.04 . on 2012
june 14 ut , @xmath0 her was observed at an airmass of 1.01 , and again we used the srf of hd 129814 . on 31 march 2015 ut ,
@xmath0 her was observed at an air mass of 1.01 .
we used the srf computed from hd 74360 from the same night with an airmass of 1.02 . after the spectrum of the companion was extracted from the p1640 data , we compared it against the spectra in the infrared telescope facility ( irtf ) spectral library @xcite .
these include fgkm main sequence stars and lt brown dwarfs .
the template spectra were binned and smoothed in order to produce the equivalent spectra to having the star observed by p1640 .
then each template spectrum was compared against the measured spectrum using the sum of the squares of the residual ( ssr ) as a metric @xcite .
the best fit reference spectrum was the one with the minimum value to the ssr .
in figure [ spectra ] we plot the normalized spectra for @xmath0 her ab for the three observations overplotted with the spectra for m2v , m4v , and m6v stars taken from the irtf spectral library .
there are three m4v stars in the spectral library and we have plotted all three of them on this chart . this shows the variation between spectra of the same type due to age and metallicity .
figure [ metric ] shows the sum of the squares of the residual ( ssr ) metric as a function of spectral type for each of the three observations . for spectral types having multiple entries in the irtf spectral library , multiple values of the ssr metric
are calculated and shown .
the 2012 june 12 ssr metric indicates the star has a spectral type of m2v - m5v .
the 2012 june 14 data have the highest value of the ssr , indicating the worst fit overall , but the ssr metric strongly prefers the m4v result .
the 2015 march 31 result is a m4v - m5v . from these three results ,
we conclude that @xmath0 her ab has a spectral type of m4@xmath21v .
there have been a number of prior mass estimates of the primary star .
@xcite derived a mass of 1.14 @xmath9 with uncertainties of 5% by comparing the measured effective temperature against the luminosity on theoretical evolutionary tracks . using the same technique , @xcite derived a mass of 1.13 @xmath9 .
@xcite compared measured spectra to models to compute a mass of [email protected] @xmath9 .
@xcite detected solar - like oscillations in the primary with asteroseismology .
@xcite modeled these oscillations and computed a mass of @xmath0 her a of @xmath19 @xmath9 and an age of 6.433 @xmath2 0.04 gyr .
the authors note that models with mass 11.1 @xmath9 and age 6.26.7 gyr can also reproduce the non - asteroseismic and asteroseismic constraints .
there have been two measurements of the primary s mass with long baseline interferometery .
@xcite calculated a mass of 1.118 @xmath9 .
@xcite produced a estimate of [email protected] @xmath9 . with the exception of the @xcite analysis ,
the results are all approximately 1.1 @xmath9 and we have used this value below .
in addition to the astrometry shown in table [ astrometry ] , we also have radial velocity ( rv ) data from the lick planet search @xcite .
@xmath0 her a was monitored from 1987 till 2013 . over this time period , 332 rv data points were collected .
the data have a median error of 2.26 m / s .
while not tracing out a complete orbit , the data clearly show acceleration .
we generated mean points for the astrometric data taken the same day .
we computed a combined orbital solution using the orbitx code @xcite .
orbitx uses the levenberg - marquardt method to solve for all the orbital elements .
figure [ visual_orbit ] illustrates the new visual orbital solution , plotted together with all published data as well as the new data in table [ astrometry ] .
previously published observations are indicated by open circles , and measures from table [ astrometry ] are represented by filled circles .
@xmath20 " lines connect each measure to its predicted position along the new orbit ( shown as a thick solid line ) .
a dot - dash line indicates the line of nodes , and a curved arrow in the lower right corner of each figure indicates the direction of orbital motion in addition to providing the figure orientation .
the scale of the orbit is indicated .
finally , the previously published orbit of @xcite is shown as a dashed ellipse .
figure [ rv ] shows the computed spectroscopic orbit and the archival rv data are overplotted .
the rv data include error bars , but they are hard to see due to the tiny errors on each data point .
table [ orbital_elements ] lists the solution for the nine orbital elements : @xmath21 ( period , in years ) , @xmath22 ( semi - major axis , in arcseconds ) , @xmath23 ( inclination , in degrees ) , @xmath24 ( longitude of nodes , equinox 2000.0 , in degrees ) , @xmath25 ( epoch of periastron passage , in fractional julian year ) , @xmath26 ( eccentricity ) , @xmath27 ( longitude of periastron , in degrees ) k1 ( semiamplitude of the primary , in kms@xmath12 ) and v0 ( the systemic velocity , in kms@xmath12 ) . for comparison purposes
, the table also lists the orbital elements from the two previous astrometric orbits for the system from @xcite and @xcite , with the exception of semi - major axis : since the previous orbits were based on the motion of the photocenter , the derived semi - major axis from these previous studies is not expected to be the same .
there were no error bars on the elements from either of the heintz orbits .
@xcite estimated that three quarters of the orbit had been observed , which would be about 45 years .
the three orbits have consistent solutions for the inclination , longitude of node and eccentricity .
we have found a longer period than the previous orbits , but that is not unusual as astrometric orbits often have large errors .
the orbital elements are most appropriately characterized as provisional .
the solution should be good enough to provide reasonable ephemerides over the next decade , but the elements will require correction over the course of a complete orbit to be considered approximately correct . despite the uncertainty , the quantity 3 @xmath28 , should not be grossly erroneous and provides useful information on the mass of the system .
using the combined orbit and the measured parallax @xcite , we are able to compute a mass sum of of @xmath0 her aa , ab as 1.42 @xmath9 with a minimum mass of the secondary of 0.21@xmath9 .
using the previously derived mass of the primary from the literature of 1.1@xmath9 produces an estimate of the secondary mass of 0.32 @xmath9 .
the secondary spectral type determined in [ spectral_type ] corresponds to a mass of 0.14 - 0.36@xmath9 @xcite , which is in general agreement with our estimate of the secondary mass from the combined orbit solution .
lccc p ( yr ) & 63 & 65 & [email protected] + a ( ) & - & - & [email protected] + i ( ) & 68 & 68 & [email protected] + w ( ) & 80.0 & 81.8 & [email protected] + t & 1951.0 & 1951.0 & [email protected] + e & 0.34 & 0.32 & [email protected] + w ( ) & 94.2 & 92 & 214@xmath216 + k1 ( kms@xmath12 ) & - & - & [email protected] + v0 ( kms@xmath12 ) & - & - & [email protected] her aa , ab . the black line is the computed orbit , while the individual data points and their associated error bars are overplotted . , width=264 ] her aa , ab .
the solid thin line orbit is the astrometric orbit from measuring the shift in the photo - center from @xcite .
the broken line through the origin is the line of nodes .
open circles are previously published astrometry values and the filled circles are the new values from this paper .
the axes are labeled in units of arcseconds.,width=264 ]
we have shown that @xmath0 her ab is a stellar companion with a spectral class of m4@xmath21v .
this was estimated from multiple epochs of low - resolution near - infrared spectra of the companion .
in addition we have computed a combined visual and spectroscopic orbit for the system , producing a mass estimate for the secondary which is in agreement with the expected mass of a m4v star .
much of the previous confusion about the stellar versus sub - stellar nature of @xmath0 her ab comes from combining photometric results from different observations .
while that is normally a fine practice , in this case it proved to be a problem .
the near - infrared photometry of @xcite appears to be incorrect as it does not agree with the numerous pharo observations in h and ks nor the nirspec / scam observation at 2.2 .
yet their estimate of the spectral type of the companion from h - k color agrees with ours from near - infrared spectroscopy . it may be that there was an offset in the photometry from both filters
this would also explain the anomalous @xmath29 color computed by @xcite from their mid - ir data combined with the @xcite near - infrared photometry , which they reported as too red for a m4v star .
we have computed a preliminary orbit for the @xmath0 her aa , ab binary .
a continuing effort is needed to collect additional astrometric and radial velocity measurements of the system to improve the orbit determination .
additional rv data in the near future would be especially useful as the system has most likely just undergone an inflection point in the rv curve .
additional data will produce a more accurate estimate of the masses of the two stars .
the mass of the primary can then be compared against the mass of the primary measured from asteroseimology .
combining independent observations with asteroseismology is crucial to advance progress in the theoretical modeling of observed oscillation frequencies , and the validation of asteroseismic relations to derive fundamental stellar properties @xcite .
our thanks to nils turner for assistance with the mt . wilson log books .
we also thank andrei tokovinin for assistance with the latest version of his orbitx code .
a portion of the research in this paper was carried out at the jet propulsion laboratory , california institute of technology , under a contract with the national aeronautics and space administration ( nasa ) .
this work was partially funded through the nasa roses origins of solar systems grant nmo710830/102190 .
project 1640 is funded by national science foundation grants ast-0520822 , ast-0804417 , and ast-0908484 .
in addition , part of this work was performed under a contract with the california institute of technology funded by nasa through the sagan fellowship program .
ja is supported by the laboratory for physical sciences , college park , md , through the national physical science consortium graduate fellowship program .
j.c . was supported by the u.s .
national science foundation under award no .
r.n . was funded by the swedish research council s international postdoctoral grant no .
637 - 2013 - 474 .
the members of the project 1640 team are also grateful for support from the cordelia corporation , hilary and ethel lipsitz , the vincent astor fund , judy vale , andrew goodwin , and an anonymous donor .
this paper is based on observations obtained at the maui space surveillance system operated by the us air force research laboratory@xmath30s directed energy directorate and at the hale telescope , palomar observatory .
this research made use of the keck observatory archive ( koa ) , which is operated by the w. m. keck observatory and the nasa exoplanet science institute ( nexsci ) , under contract with the national aeronautics and space administration .
nexsci is sponsored by nasa s origins theme and exoplanet exploration program , and operated by the california institute of technology in coordination with the jet propulsion laboratory .
this research made use of the washington double star catalogue maintained at the u.s .
naval observatory , the simbad database , operated by the cds in strasbourg , france and nasa s astrophysics data system .
|
@xmath0 her is a nearby quadruple system with a g - subgiant primary and several low mass companions arranged in a 2@xmath1 architecture . while the bc components have been well characterized , the ab component has been detected astrometrically and with direct imaging but there has been some confusion over its nature , in particular whether the companion is stellar or substellar . using near - infrared spectroscopy we are able to estimate the spectral type of the companion as a m4@xmath21v star .
in addition , we have measured the astrometry of the system for over a decade .
we combined the astrometry with archival radial velocity measurements to compute an orbit of the system . from the combined orbit ,
we are able to compute the mass sum of the system . using the estimated mass of the primary
, we estimate the mass of the secondary as 0.32 m@xmath3 , which agrees with the estimated spectral type .
our computed orbit is preliminary due to the incomplete orbital phase coverage , but it should be sufficient to predict ephemerides over the next decade .
|
let us begin by considering a single two - level system interacting with an arbitrary environment .
the hamiltonian of the system is @xmath0 , the environment hamiltonian is @xmath1 , and there is some interaction between the system and the environment that is described by the hamiltonian @xmath2 .
the total system - environment hamiltonian is thus @xmath3 . at time @xmath4
, we prepare the system state @xmath5 . in the usual treatment of the quantum zeno and anti - zeno effects ,
repeated projective measurements described by the projector @xmath6 are then performed on the system with time interval @xmath7 .
the survival probability of the system state after one measurement is then @xmath8 $ ] , where @xmath9 denotes taking the trace over the system and the environment , @xmath10 is the initial combined state of the system and the environment , and the evolution of the system state due to the system hamiltonian itself has been eliminated via a suitable unitary operation just before performing the measurement @xcite . assuming that the system - environment correlations can be neglected , the survival probability after @xmath11 measurements can be written as @xmath12^n = e^{-\gamma(\tau)n\tau}$ ] , thereby defining the effective decay rate @xmath13 .
it should be noted that the behaviour of the effective decay rate @xmath13 as a function of the measurement interval allows us to identify the zeno and anti - zeno regimes .
namely , if @xmath13 increases as @xmath7 increases , we are in the zeno regime , while if @xmath13 decreases if @xmath7 increases , we are in the anti - zeno regime @xcite .
we now consider an alternative way of repeatedly preparing the initial state with time interval @xmath7 .
once again , we start from the initial system state @xmath5 . after time @xmath7 , we know that the state of the system is given by the density matrix @xmath14 $ ] , where once again the evolution due to the free system hamiltonian has been removed .
now , instead of performing the projective measurement @xmath6 , we perform an arbitrary projective measurement given by the projector @xmath15 .
the survival probability is then @xmath16 $ ] , and the post - measurement state is @xmath17 . by performing a unitary operation @xmath18 on the system state on a short timescale , where @xmath19 , we can again end up with the initial state @xmath5 after the measurement .
this process can then , as before , repeated again and again to repeatedly prepare the system state @xmath5 . once again , if the correlations between the system and the environment can be neglected , we can write the effective decay rate as @xmath20 .
but now , we can , in principle , via a suitable choice of the projector @xmath15 , obtain a larger survival probability ( and a correspondingly smaller decay rate ) than what was obtained with repeatedly using projective measurements given by the projector @xmath6 .
the question , then , is what is this projector @xmath15 that should be chosen to maximize the survival probability ? for an arbitrary quantum system , it is difficult to give a general condition or formalism that will predict this optimal projective measurement .
however , most studies of the effect of repeated quantum measurements on quantum systems have been performed by considering the quantum system to be a single two - level system @xcite .
let us now show that if the quantum system is a two - level system , then it is straightforward to derive a general method for calculating the optimal projective measurements that need to be performed as well as an expression for the optimized decay rate .
we start from the observation that the system density matrix at time @xmath7 , just before the measurement , can be written as @xmath21 where @xmath22 is the bloch vector of the system state .
we are interested in maximizing the survival probability @xmath16 $ ] .
it is clear that we can also write @xmath23 where @xmath24 is a unit vector corresponding to the bloch vector for the projector @xmath15 . using eqs . and
, we find that the survival probability is @xmath25 it should then be obvious how to find the optimal projective measurement @xmath15 that needs to be performed .
the maximum survival probability is obtained if @xmath24 is parallel to @xmath22 .
if we know @xmath26 , we can find out @xmath22 .
consequently , @xmath24 is simply the unit vector parallel to @xmath22 .
once we know @xmath24 , we know the projective measurement @xmath15 that needs to be performed .
the corresponding optimal survival probability is given by @xmath27 now , if we ignore the correlations between the system and environment , which is valid for weak system - environment coupling , we can again derive the effective decay rate of the quantum state to be @xmath28 .
we now investigate the optimal effective decay rate for a variety of system - environment models . to begin , we consider the paradigmatic population decay model .
the system - environment hamiltonian is ( we use @xmath29 throughout ) @xmath30 where @xmath31 is the energy difference between the two levels , @xmath32 is the standard pauli matrix , @xmath33 and @xmath34 are the raising and lowering operators , and @xmath35 and @xmath36 are the annihilation and creation operators for mode @xmath37 of the environment . it should be noted that here we have made the rotating - wave approximation .
this system - environment hamiltonian is widely used to study , for instance , spontaneous emission @xcite .
we consider the very low temperature regime .
we initially prepare the system - environment state @xmath38 that describes the two - level system to be in the excited state and the environment oscillators to be in their ground state .
ordinarily , in the studies of the qze and the qaze , the system is repeatedly projected onto the excited state with time interval @xmath7 . as discussed before
, we , on the other hand , allow the system to be projected onto some other state such that the effective decay rate is minimized . to find this optimal projective measurement , we need to understand how the bloch vector of the system evolves in time .
due to the structure of the system - environment hamiltonian , the system - environment state at a later time @xmath7 can be written as @xmath39 , where @xmath40 means that the two - level system is in the ground state and that mode @xmath37 of the environment has been excited .
it then follows that the density matrix of the system at time @xmath7 is @xmath41 we consequently find that the components of the bloch vector of the system are @xmath42 , while @xmath43 .
thus , we have a nice interpretation for the dynamics of the system .
initially , the system is in the excited state .
as time goes on , the coherences remain zero , and the probability that the system makes a transition to the ground state increases . in other words , initially the bloch vector of the system is a unit vector with @xmath44 .
the bloch vector then decreases in magnitude ( while keeping the @xmath45 and @xmath46 components zero ) until the size of the bloch vector becomes zero .
the bloch vector thereafter flips direction and increases in length until effectively @xmath47 .
since the direction of the bloch vector corresponding to the optimal measurement is parallel to the bloch vector of the system , we find that if the measurement interval is short enough such that @xmath48 , then we should keep on applying the projector @xmath49 . on the other hand , if the measurement interval is large enough so that @xmath50 , then we should rather apply the projector @xmath51 , and then , just after the measurement , apply a @xmath52 pulse so as to end up with the system state @xmath53 again .
in other words , the time @xmath54 at which the bloch vector flips direction is of critical importance to us and needs to be found out in order to optimize the effective decay rate . to find this time , we assume that the system and the environment are weakly coupled and thus we can use a master equation to analyze the dynamics of the system . since we numerically solve this master equation
, we might as well put back in the non - rotating wave approximation terms so that the system - environment interaction that we consider in solving the master equation is @xmath55 .
the master equation that we use can be written as @xmath56 + \int_{0}^{t } ds \big \ { [ \bar{f}(t , s)\rho_s(t ) , f ] c_{t s } + \ ; \text{h.c . }
\big \},\ ] ] where the system hamiltonian is @xmath57 , the system - environment interaction hamiltonian has been written as @xmath58 with @xmath59 and @xmath60 , @xmath61 , and h.c .
denotes the hermitian conjugate . here
the environment correlation function @xmath62 is defined as @xmath63 $ ] where @xmath64 and @xmath65 with @xmath66 the partition function . to find the environment correlation function ,
we introduce the spectral density function @xmath67 , where
@xmath68 parametrizes the system - environment coupling strength , @xmath69 characterizes the ohmicity of the environment , and @xmath70 is the cutoff frequency .
we can then numerically solve this differential equation to find the system density matrix at any time @xmath71 and consequently the bloch vector of the system .
we consequently know the optimal projective measurement that needs to performed .
we now present a computational example ( see the previous page ) .
we plot the single measurement survival probability [ see fig .
[ popdecaymerged2](a ) ] and the effective decay rate [ see fig .
[ popdecaymerged2](b ) ] as a function of the measurement interval @xmath7 .
the dotted lines illustrate what happens if we keep on projecting the system state onto @xmath53 . for a small measurement interval ,
the optimal measurement is @xmath49 since the system bloch vector has only a positive @xmath72-component . on the other hand ,
if we realize that when the measurement interval is long enough such that , between each measurement , the bloch vector flips direction , then to maximize the survival probability , we should project instead onto the state @xmath73 and then apply @xmath52 pulse . doing so , we can obtain a higher survival probability or , equivalently , a lower effective decay rate .
this is precisely what we observe from the figure . for this population decay case
, we find that if the measurement interval is larger than @xmath74 , then we are better off by performing the measurement @xmath51 .
there is also a small change in the anti - zeno behaviour . for our modified strategy of repeatedly preparing the quantum state
, we find that beyond measurement interval @xmath75 , there is a sharper signature of anti - zeno behaviour as compared to the usual strategy of repeatedly measuring the excited state of the system .
we now analyze our strategy for the case of the pure dephasing model @xcite .
the system - environment hamiltonian is now given by @xmath76 the difference now compared to the previous population decay model is that the system - environment coupling term now contains @xmath77 instead of @xmath78 .
this difference implies that the diagonal entries of the system density matrix @xmath79 ( in the @xmath80 basis ) can not change - only dephasing can possibly take place , which is the reason that this model is known as the pure dephasing model@xcite .
furthermore , this pure dephasing model is exactly solvable .
the off - diagonal elements of the density matrix undergo both unitary time evolution due to the system hamiltonian and non - unitary time evolution due to the coupling with the environment . assuming the initial state of the total system is the standard product state @xmath81 , the off - diagonals of the density matrix @xmath82_{mn}$ ] , once the evolution due to the system hamiltonian itself is removed , are given by @xmath82_{mn } = [ \rho_{s}(0)]_{mn } e^{- \gamma(\tau)}$ ] where @xmath83@xcite .
writing an arbitrary initial state of the system as @xmath84 , it is straightforward to find that @xmath85 the optimal survival probability obtained using optimized measurements is then @xmath86 where eq . has been used .
on the other hand , if we keep on preparing the initial state @xmath5 by using the projective measurements @xmath6 , we find that @xmath87 we now analyze eqs . and to find conditions under which we can lower the effective decay rate by using optimized projective measurements .
it is clear that if the initial state , in the bloch sphere picture , lies in the equatorial plane , then @xmath88 while @xmath89 . in this case ,
eqs . and
give the same survival probability .
thus , in this case , there is no advantage of using our strategy of optimized measurements as compared with the usual strategy .
the reasoning is clear . in the bloch sphere picture ,
the magnitude of the time evolved bloch vector of the density matrix reduces such that the time evolved bloch vector is always parallel to the bloch vector of the initial pure system state .
as argued before , the optimal projector to measure at time @xmath7 , @xmath90 , must be parallel to the bloch vector of the density matrix at time @xmath7 .
hence in this case , the optimal projector to measure is @xmath91 , corresponding to the initial state . the computational example shown in fig .
[ dephasing1](a ) illustrates our predictions . on the other hand ,
if we make some other choice of the initial state that we repeatedly prepare , we can easily see that our optimized strategy can give us an advantage .
we simply look for the case where the evolved bloch vector ( after removal of the evolution due to the system hamiltonian ) no longer remains parallel with the initial bloch vector . upon inspecting eq .
, we find that our optimized strategy can be advantageous if @xmath92 ( excluding , of course , the cases @xmath93 ) .
in other words , if the bloch vector of the initial state does not lie in the equatorial plane , then the bloch vector of this state at some later time will not remain parallel to the initial bloch vector . in this case then , our optimal measurement scheme will give a higher survival probability as compared to repeatedly measuring the same initial state .
this is illustrated in fig .
[ dephasing1](b ) where we show the effective decay rate and the survival probability after a single measurement for the initial state specified by the bloch vector @xmath94 . after the time at which the transition between the zeno and the anti - zeno regimes occurs
, we clearly observe that the decay rate is lower when one makes the optimal projective measurements .
although this difference on first sight may not appear as very significant , if we perform a relatively large number of repeated measurements , the difference is very significant .
for example , even for three measurements with measurement interval @xmath95 , we find that the quantum state has @xmath96 greater survival probability with the optimized measurements as compared with the usual unoptimized strategy of repeatedly preparing the quantum state .
another computational example has been provided in fig .
[ dephasing2 ] where the initial state is now given by the bloch vector @xmath97 . in fig .
[ dephasing2](a ) we have again illustrated that our optimized strategy of repeatedly preparing the quantum state is better at protecting the quantum compared to the usual strategy . in fig .
[ dephasing2](b ) we have shown how the optimal projective measurement that needs to be performed changes with the measurement interval @xmath7 . in order to do so , we have parametrized the bloch vector corresponding to @xmath15 using the usual spherical polar angles @xmath98 and @xmath99 .
note that the value of the azimuthal angle @xmath99 is expected to remain constant since we have @xmath100 .
on the other hand , the optimal value of the polar angle @xmath98 changes with the measurement interval .
this is also expected since as the system dephases , @xmath101 , ensuring that @xmath102 .
thus , for long measurement intervals , the system bloch vector becomes effectively parallel to the @xmath103 .
it follows that @xmath104 for long measurement intervals .
these predictions are borne out by the behaviour of @xmath98 and @xmath99 in fig .
[ dephasing2](b ) .
we now consider the more general system - environment model given by the hamiltonian @xmath105 where @xmath106 can be understood as the tunneling amplitude for the system , and the rest of the parameters are defined as before .
this is the well - known spin - boson model@xcite , which can be considered as an extension of the previous two cases in that we can now generally have both population decay and dephasing taking place .
experimentally , such a model can be realized , for instance , using superconducting qubits @xcite and the properties of the environment can be appropriately tuned as well @xcite .
once again , assuming that the system and the environment are interacting weakly with each other , we can use the master equation that we have used before [ see eq . ] to find the system density matrix as a function of time .
we now have @xmath107 and @xmath108 .
it should be remembered that once we find the density matrix just before the measurement @xmath26 , we remove the evolution due to system hamiltonian via @xmath109 .
let us first choose as the initial state @xmath110 ( or , in the words , the state that is paramterized by @xmath111 and @xmath112 on the bloch sphere ) . in fig .
[ sb1 ] , we plot the behaviour of the effective decay rate as a function of the measurement interval using both our optimized strategy ( the solid red lines ) and the unoptimized usual strategy ( the dotted , blue curves ) .
it is clear from fig .
[ sb1](a ) that for relatively short measurement intervals , there is little to be gained by using the optimal strategy .
as we have seen before in the pure dephasing case , for the state in the equatorial plane of the bloch sphere , there is no advantage to be gained by following the optimized strategy . on the other hand , for longer time intervals @xmath7
, population decay can be considered to become more significant .
this then means we that see significant difference at long measurement intervals if we use the optimized strategy .
this is precisely what we observe in fig .
[ sb1](b ) . for completeness ,
let us also investigate how the the effective decay rate depends on the functional form of the spectral density . in fig .
[ sb2](a ) and fig .
[ sb2](b ) , we investigate what happens in the case of a sub - ohmic and a super - ohmic environment respectively .
the case of a sub - ohmic environment with @xmath113 is similar to the case with @xmath114 ( ohmic environment ) - once again , the optimal projective measurements decrease the decay rate substantially only at long periods of time . for the case of a super - ohmic environment with @xmath115
[ see fig .
[ sb2](b ) ] , we find that the optimal projective measurements do not substantially lower the decay rate , even for long times .
thus , it is clear that the ohmicity of the environment plays an important role in determining the usefulness of using the optimal projective measurements .
let us now revert to the ohmic environment case to present more computational examples .
first , if @xmath116 , then the effect of dephasing is much more dominant than the effect of population decay . results for this case are illustrated in fig .
[ sb2](c ) .
we see that there is negligible difference upon using the optimal measurements .
this agrees with what we found when we analyzed the pure dephasing model .
we also analyze the opposite case where the effect of population decay is more dominant than the effect of dephasing .
this is done by setting @xmath117 .
we consider higher temperature in this case .
we now observe differences between the unoptimized and optimized decay rates for relatively short times [ see fig . [ sb2](d ) ] , and the difference becomes even bigger at longer times .
in fact , while we observe predominantly only the zeno effect with the unoptimized measurements , we observe very distinctly both the zeno and the anti - zeno regimes with the optimized measurements .
we extend our study to a collection of two - level systems interacting with a common environment .
this hamiltonian can be considered to be a generalization of the usual spin - boson model to a large spin @xmath118 @xcite , where @xmath119 is the number of two - level systems coupled to the environment .
physical realizations include a two - component bose - einstein condensate @xcite that interacts with a thermal reservoir via collisions @xcite . in this case , the system - environment hamiltonian is given by @xmath120 where @xmath121 and @xmath122 are the usual angular momentum operators and the environment is again modeled as a collection of harmonic oscillators .
we first look at the pure dephasing case by settting @xmath123 . in this case
, the system dynamics can be found exactly .
the system density matrix , in the eigenbasis of @xmath122 , after removal of the evolution due to the system hamiltonian can be written as @xmath124_{mn } = [ \rho(0)]_{mn } e^{- i\triangle(t ) ( m^2 - n^2 ) } e^{- \gamma(t ) ( m - n)^2}$ ] . here
@xmath125 has been defined before , and @xmath126}{\omega_k^2}$ ] describes the indirect interaction between the two - level systems due to their interaction with a common environment . for vanishingly
small time @xmath71 , @xmath127 . on the other hand , as @xmath71 increases
, the effect of @xmath128 becomes more pronounced .
thus , we expect significant differences as compared to the single two - level system case for long measurement intervals .
however , it is important to note that we can no longer find the optimal measurements using the formalism presented before since our system is no longer a single two - level system . in principle
, we need then need to carry out a numerical optimization procedure in order to find the projector @xmath15 such that the survival probability is maximized . rather than looking at all possible states
@xmath17 , we instead restrict ourselves to the su(2 ) coherent states since these projective measurements are more readily experimentally accessible .
in other words , we look at @xmath15 where @xmath129 and @xmath130 with the states @xmath131 being the angular momentum eigenstates of @xmath122 .
suppose that we prepare the coherent state @xmath132 with a fixed , pre - determined value of @xmath133 repeatedly . in order to do so , we project , with time interval @xmath7 , the system state onto the coherent state @xmath134 .
after each measurement , we apply a suitable unitary operator to arrive back at the state @xmath135 . again assuming the system - environment correlations are negligible
, we find that @xmath136^{2j } \bigg [ \frac{|\eta|}{1 + |\eta|^2 } \bigg ] ^{2j } \sum_{m , n = -j}^{j } ( \zeta^ { * } \eta)^m ( \eta^ { * } \omega)^n \ ; \binom{2j}{j + m}\binom{2j}{j + n } e^{- i\triangle(\tau ) ( m^2 - n^2 ) } e^{- \gamma(\tau ) ( m - n)^2 } \bigg \}.\end{aligned}\ ] ] for equally spaced measurement time intervals , we numerically optimize eq . over the variables @xmath137 and @xmath138 .
we present a computational example in fig .
[ lsj1](a ) .
we take as the initial state the su(2 ) coherent state with @xmath111 and @xmath139 and we let @xmath140 .
this is simply the generalization of the pure dephasing model that we have looked at before to @xmath140 .
previously , there was no difference in the optimized and unoptimized probabilities .
now , we see that because of the indirect interaction , there is a very noticeable difference . where we observe the zeno regime with the unoptimized measurements ,
we instead see the anti - zeno regime with the optimized measurements .
furthermore , the survival probability can be significantly enhanced using the optimized measurements . for completeness
, we have also considered the more general case with @xmath141 . in this case , the system dynamics can not be solved exactly , so we resort again to the master equation to find the system dynamics . with the system dynamics known
, we again find the projector , parametrized by @xmath138 and @xmath137 , such that the decay rate is minimized .
results are illustrated in [ lsj1](b ) .
once again , using optimizing projective measurements change the zeno and anti - zeno behaviour quatitatively as well as qualitatively .
our central idea is that instead of repeatedly preparing the quantum state of a system using only projective measurements , we can repeatedly prepare the quantum state by using a combination of both projective measurements and unitary operations .
this then allows us to consider the projective measurements that yield the largest survival probability .
if the central quantum system is a simple two - level system , we have derived an expression that optimizes the survival probability , or equivalently the effective decay rate .
this expression implies that the optimal projective measurement at time @xmath7 corresponds to the projector that is parallel to the bloch vector of the system s density matrix at that time .
we consequently applied our expression for the optimized survival probability to various models . for the population decay model
, we found that beyond a critical time @xmath142 , we should flip the measurement and start measuring the ground state rather than the excited state .
for the pure dephasing model , we found that for states prepared in the equatorial plane of the bloch sphere , it is optimal to measure the initial state - determining and making the optimal projective measurement has no effect on the effective decay rate .
in contrast , for states prepared outside of the equatorial plane , the effect of making the optimal projective measurement substantially lowers the effective decay rate in the anti - zeno regime . in the general spin - boson model
, we have found that there can a considerable difference between the effective decay rate if we use the optimal measurements .
we then extended our analysis by analyzing the case of large spin systems . in this case , we find that the indirect interaction between the two - level systems causes the optimal measurements to be even more advantageous . the results of this paper show that by exploiting the choice of the measurement performed , we could substantially decrease the effective decay rate for various cases .
this allows us to ` effectively freeze ' the state of the quantum system with a higher probability .
experimental implementations of the ideas presented in this paper are expected to be important for measurement - based quantum control .
|
measurements in quantum mechanics can not only effectively freeze the state of the quantum system ( the quantum zeno effect ) but also accelerate the time evolution of the system ( the quantum anti - zeno effect ) . in studies of the quantum zeno and anti - zeno effects , a quantum state , usually an excited state of the system , is prepared repeatedly by projecting the quantum state onto the initial state again and again . in this paper
, we repeatedly prepare the initial quantum state in a different manner . instead of only performing projective measurements , we allow unitary operations to be performed , on a very short time - scale , after each measurement .
we can then repeatedly prepare the initial quantum state by performing some projective measurement and thereafter , after each measurement , performing a suitable unitary operation to repeatedly end up with the initial state .
our objective thereafter is to find the projective measurements that need to repeatedly performed such that we obtain the minimum possible effective decay rate of the quantum state .
consequently , the quantum state is maximally protected from its environment .
we find that these optimized projective measurements keep on changing depending on the measurement time interval .
we explicitly find the projective measurements that need to be performed , as well as the optimized effective decay rate , for a variety of system - environment models such as the population decay model , the pure dephasing model , the spin - boson model , and the large spin - boson model .
we find that there can be considerable differences between this optimized effective decay rate and the usual decay rate obtained by repeatedly projecting onto the initial state .
in particular , the zeno and anti - zeno regimes can be considerably modified .
rapid repeated measurements can slow down the time evolution of a quantum system , an effect known as the quantum zeno effect ( qze ) @xcite .
it has also been found that if the measurements are not rapid enough , an opposite effect , known as the quantum anti - zeno effect ( qaze ) , can occur - the measurements can actually accelerate quantum transitions @xcite .
both the qze and the qaze have attracted widespread interest both theoretically and experimentally due to their relevance to the foundations of quantum mechanics as well as possible applications in quantum technologies @xcite .
generally , the analysis of qze and qaze for open quantum systems assumes that one continues to repeatedly prepare the initial state of the quantum system by repeatedly projecting the system onto the initial state .
however , one can prepare the initial state of the system repeatedly in a different way . instead of restricting ourselves to performing the same projective measurement again and again
, we can instead think of allowing ourselves to perform a unitary operation after each projective measurement .
this means that we are then no longer restricted with performing the same projective measurement to repeatedly prepare the initial state .
rather , we can now perform any projective measurement , followed by a suitable unitary operation .
we can then use this freedom in the choice of the projective measurement to think about maximizing the survival probability of the system quantum state .
our goal is to find the projective measurements that maximize this survival probability , or , equivalently , minimize the effective decay rate .
thus we arrive at a strategy that allows us to maximally protect the quantum state .
we can then study the quantum zeno and anti - zeno effects with such optimal projective measurements .
we start off by considering a single two - level system coupled to an environment . starting from an arbitrary initial state ,
our goal is to find the projective measurements followed by the unitary operators that can be used to repeatedly prepare the initial state such that the survival probability of the quantum state is maximized .
we find a general condition that allows us to find such projective measurements and unitary operators independent of the particular form of the environment .
namely , we find that the dynamically evolving bloch vector of the quantum system determines the optimal projective measurement and , in turn , the optimal survival probability .
we find that , in general , the optimal projective measurements depend on the time interval between the measurements .
this contrasts sharply with the usual scenario where the same projective measurement is performed repeatedly , regardless of the measurement interval .
applications of the condition derived are illustrated by studying the usual decay rate and the optimized decay rate for the population decay model , the pure dephasing model , and the spin - boson model .
since the behaviour of the effective decay rate allows us to identify the quantum zeno and anti - zeno regimes , we can instead use this optimized decay rate to look at the quantum zeno and anti - zeno effects .
we then consider a collection of two - level systems interacting with a common environment . in this case , finding the optimal projective measurement is a very difficult problem .
however , we can restrict ourselves to measurements that project onto spin coherent states since these are the measurements that can be performed relatively easily experimentally .
for this scenario , we again derive the optimal measurements and the optimized decay rate for the pure dephasing case as well as the more general scenario with both dephasing and relaxation present . once again , we find that there can be signficant differences between the unoptimized decay rate and the optimized decay rate .
these differences translate to differences in the qze and the qaze .
|
in our ongoing araucaria project ( e.g. gieren et al .
2005a ) , we are applying a number of different stellar standard candles to independently determine the distances to a sample of nearby galaxies .
the systematic differences between the distance results obtained for the individual galaxies from the various stellar candles will be analyzed in forthcoming papers .
this analysis is expected to lead to a detailed understanding of how the various stellar techniques , which are fundamental to calibrate the first rungs of the distance ladder , depend on metallicity and age . while the objects we use for the distance determinations are usually detected from optical wide - field imaging surveys of the target galaxies ( e.g. pietrzyski et al .
2002b ) , the most accurate distance work is then done from follow - up near - infrared images which virtually eliminate reddening as a significant source of error on the results .
examples of this very successful approach are the cepheid work on the sculptor galaxies ngc 300 ( gieren et al .
2005b ) , ngc 55 ( gieren et al . 2008 ) and ngc 247 ( gieren et al .
2009 ) , and the red clump star work on the lmc ( pietrzyski & gieren 2002 ) . the araucaria project has also been developing a new spectroscopic distance indicator , viz . the flux - weighted gravity- luminosity relationship ( fglr ) for blue supergiants ( kudritzki et al . 2003 ; 2008 ) which is able to yield distances accurate to 5% to galaxies containing massive blue stars out to about 10 mpc from low - resolution spectra ( kudritzki et al .
2008 , urbaneja et al .
2008 ) .
thanks to several recent theoretical and empirical studies , evidence has been mounting that rr lyrae ( rrl ) stars are excellent standard candles in the near - infrared spectral range , providing distance results which are superior to the traditional optical method ( e.g. bono 2003a ) .
longmore et al .
( 1986 ) were the first to show that rrl variable stars follow a period - luminosity ( pl ) relation in the near - infrared k - band .
their pioneering work was followed by liu & janes ( 1990 ) , jones et al .
( 1996 ) , and skillen et al .
( 1993 ) , who applied infrared versions of the baade - wesselink method to calibrate the luminosities and distances of rrl stars .
a very comprehensive analysis of the ir properties of rrl stars was given by nemec et al .
the first theoretical constraints on the k - band pl relation of rrl stars are based on non - linear convective pulsation models which were presented by bono et al .
dallora et al .
( 2004 ) later demonstrated that the k - band pl relation for rrl stars appears to have a very small scatter for globular clusters , with small intrinsic spread in metallicity for this type of stars .
further theoretical explorations of the rrl period - mean magnitude - metallicity relations in near - infrared passbands were carried out by bono et al .
( 2003b ) , catelan et al .
( 2004 ) , and cassisi et al .
most recently , sollima et al . (
2008 ) analyzed near - infrared k - band data of rrl stars in some 15 galactic globular clusters and provided the first empirical calibration of the period - luminosity - metallicity ( plz ) relation in the @xmath1 band .
all these existing theoretical and empirical studies have suggested that the rrl star k - band plz relation appears to be a superb means to determine accurate distances to galaxies hosting an abundant old stellar population .
we have therefore started to include this tool in the araucaria project distance work . in previous papers , we determined the distance to the sculptor dwarf galaxy ( pietrzyski et al .
2008 ) , and to the lmc ( szewczyk et al .
2008 ) from this method . in this paper
, we apply the technique to a sample of rr lyrae stars distributed over the smc , and determine the distance to the smc for the first time with the infrared rr lyrae technique .
during several observing runs of the araucaria project , we have obtained all the data analyzed and presented in this paper with the sofi infrared camera attached to the eso new technology telescope .
the large field setup was used to yielding a @xmath2 arcmin field of view and a scale of @xmath5 arcsec per pixel .
it was possible to collect deep @xmath6 and @xmath7 images of six distinct smc fields during 3 nights in 2007 , and of three more fields over 2 nights in 2008 .
all but one nights were found to have photometric conditions .
detailed information on each observed field can be found in table [ tabfields ] .
figure [ figfields ] illustrates the location of the observed fields on an image of the smc . in order to take into account the rapid sky level variations in the ir passbands ,
we have been using a dithering technique during the observations .
total integration times per field were 60 min in the @xmath7 band , and 15 min in the @xmath6 band .
for all the reductions and calibrations , the pipeline developed in the course of the araucaria project was used . as a first step , the subtraction of sky level subprocess ( including the masking of stars with the iraf xdimsum package )
was applied ( pietrzyski & gieren 2002 ) .
later , each single image was flatfielded and stacked into the final deep field .
psf ( point spread - function ) photometry with aperture corrections was then performed in the same way as described in pietrzyski et al .
( 2002b ) .
the photometry was calibrated onto the standard system using observations of 15 standard stars from the ukirt list ( hawarden et al .
all of them were observed under photometric conditions at different airmasses spread in between the regular target fields acquisition .
the data obtained during the non - photometric night was compared with nearby photometric data and the zero - point difference was estimated . thanks to the large number of standard stars observed , the accuracy of our photometric zero points was estimated to be as good as @xmath8 mag .
the results were compared with the two micron all sky survey ( 2mass ) catalog for common stars . this way we could determine the zero point differences which are given in table [ tab2mass ] .
the calibrated nir magnitudes for all rrl stars identified in our science target fields are presented in table [ taballstars ] .
all uncertainties given in the tables are standard deviations .
the final sample of 34 rrl stars analyzed in our sofi / ntt fields was chosen by rejecting all stars which were blended with or contaminated by other close - by objects in the frames .
then it was cross - identified with the optical gravitational lensing experiment ( ogle ) catalog of rrl stars in the smc ( soszyski et al .
the positions of the rrl stars in the @xmath1 , @xmath9 color - magnitude diagram are shown in figure [ figcmd ] .
all variables have at least one measurement collected during photometric conditions . in the case of rrl stars which were observed more than once
, we took a straight average of the random - phase magnitudes , which is expected to lead to a better approximation of their mean magnitudes . following the identification from the ogle catalog ( soszyski et al .
2002 ) , we fundamentalize the rrc periods by adding @xmath10 in order to combine the rrab and rrc stars . we have corrected the observed magnitudes for extinction , adopting the reddening values published by udalski et al . ( 1999 ) which are shown in table [ tabfields ] . adopting the reddening law from schlegel
et al ( 1998 ) , we calculate the following selective extinctions in the different bands : @xmath11 and @xmath12 . the final photometric data for the sample of rrl stars are presented in table [ tabrrlyr ] .
the pl relations for the @xmath0 and @xmath1 bands derived from our magnitudes are shown in figure [ figpmd ] . the relatively large scatter seen in figures [ figcmd ] and [ figpmd ]
are mainly caused by four factors : ( 1 ) the random single - phase nature of our ir measurements , which represents the mean magnitude of an rrl variable only to @xmath13 mag ( e.g. del principe et al .
2006 ) , ( 2 ) the metallicity spread among the rrl stars in smc , ( 3 ) the depth extension of the smc in the line of sight , and ( 4 ) the accuracy of our single measurements which is @xmath14 mag for stars of magnitudes of @xmath15 mag in the @xmath1-band .
we compare the pl relations derived from the observed rrl stars against the existing theoretical ( bono et al .
2003 ; catelan et al . 2004 ) and empirical ( sollima et al . 2008 ) relations .
table [ tabpl ] lists our results for the slope and zero - point values of the pl relations and as well as those from the theoretical and empirical ones .
in order to derive the apparent distance moduli to the smc from our data , we used the following calibratons of the nir pl relations of mixed population rrl stars : @xmath16 \qquad\mbox{-- sollima et al .
( 2008)}\ ] ] @xmath17 \qquad\mbox{-- bono et al . ( 2003b)}\ ] ] @xmath18 @xmath19 we recall that the calibration of sollima et al .
( 2008 ) was constructed for the 2mass photometric system , while the calibrations of catelan et al .
( 2004 ) and bono et al .
( 2003 ) are usable for the glass or bessel & brett systems , respectively .
therefore we transformed our own data , calibrated onto the ukirt system ( hawarden et al .
2001 ) , to the glass and bessel & brett systems using the transformations presented by carpenter ( 2001 ) before calculating distances using the calibrations of catelan et al .
( 2004 ) and bono et al .
( 2003 ) . since there is no signifficant difference between the @xmath1-band of 2mass and ukirt systems ( carpenter 2001 ) , we did not apply any transformations to our data while using the sollima et al .
( 2008 ) calibration . assuming the mean metallicity of rrl stars in the smc to be @xmath20 = -1.7 $ ] dex ( e.g. udalski 2000 ) we have calculated the @xmath1- and @xmath0-band distance moduli for our sample of rrl stars using the final photometric magnitudes as given in table [ tabrrlyr ] .
the fits for the relations @xmath21 to both sets of data are displayed in figure [ figpmdfit ] .
the true distance moduli of the smc obtained from the different plz calibrations are summarized in table [ tabdist ] . as our best distance determination
, we adopt the average value from the different calibrations , yielding @xmath22 mag .
the uncertainty on this value will be discussed in the next section .
the distance moduli obtained from several independent theoretical and empirical calibrations of the infrared rr lyrae plz relation are consistent .
the maximum difference of 0.06 mag between the results from the calibrations of sollima et al .
( 2008 ) and bono et al .
( 2003b ) is certainly not significant taking into account all the uncertainties , which affect the whole process of constructing these calibrations .
it is interesting to note that a very similar difference between the distance moduli derived using these two calibrations was recently obtained by pietrzyski et al .
( 2008 ) for the sculptor galaxy ( @xmath20 = -1.83 $ ] dex ) .
therefore , perhaps there is just a zero point offset in the sense that the distances from the calibration of sollima et al .
( 2008 ) are slightly shorter compared to those from the calibration of bono et al .
( 2003b ) .
taking into account the errors associated with the adopted calibrations , mean metallicity , photometric zero point and absorption correction , we estimate the systematic error of our distance determination to be of 0.12 mag .
therefore our best distance modulus determination to the smc is : @xmath3 ( statistical ) @xmath23 ( systematic ) mag .
this value agrees very well with the value of @xmath24 mag derived from k - band photometry of red clump stars ( pietrzyski , gieren & udalski 2003 ) .
it also agrees , within the combined @xmath25 uncertainties , with the true smc distance modulus derived from cepheid variables ( e.g. udalski 2000 ; barnes et al . 1993 ) .
the distance modulus of the smc found in this paper is 0.39 mag larger than the value for the lmc we found from the same technique ( szewczyk et al .
2008 ) . again
, this difference is very much in line with modern values found from other techniques , like cepheid variables ( 0.5 mag ; udalski 2000 ) , and red clump stars ( 0.47 mag ; pietrzyski , gieren & udalski 2003 ) .
we do not observe any significant difference in the dispersions of the nir p - l relations defined by the rr lyrae variables in the smc ( this paper ) , and the lmc ( szewczyk et al . 2008 ) , arguing against a more significant distance spread in the line of sight in the smc than in the smc , as usually found for younger tracers like cepheids .
the results of our deep infrared imaging and photometry of 34 rr lyrae stars spread over nine @xmath2 arcmin fields in the smc are presented .
our data show two clear sequences in the period luminosity plane , which correspond to the rrc and rrab stars .
after fundamentalizing the rrc periods to the period of the rrab stars by adding @xmath26 , both groups were merged , and the distance moduli to the smc were determined using different theoretical and empirical calibrations .
our final adopted distance agrees very well with modern results obtained from a number of different techniques .
our results confirm that the rr lyrae period - luminosity - metallicity relations in the near - infrared passbands are potentially a very good tool for accurate distance measurements . in future programs , particularly for the lmc , we want to reduce the systematic uncertainty of our distance determinations with this method by providing multi - phase jk observations of the rr lyrae variables to determine their mean magnitudes with better precision .
wg and gp gratefully acknowledge financial support for this work from the chilean center for astrophysics fondap 15010003 , and from the basal center for astrophysics and related tecnologies ( cata 2007 pfb 06 ) .
support from the polish grant n203 002 31/046 and the focus subsidy of the fundation for polish science ( fnp ) is also acknowledged .
it is a special pleasure to thank the support astronomers at eso - la silla for their expert help in the observations , and the eso opc for the generous amounts of observing time at the ntt allocated to our programme .
ccccccccc 1 & sc2_fi & 00:41:57.60 & -73:04:30.00 & 2007 - 08 - 27 & @xmath27 & @xmath28 & clr & @xmath29 + & & & & 2007 - 08 - 28 & @xmath30 & @xmath31 & std & + 2 & sc4_fii & 00:47:55.20 & -73:18:36.00 & 2007 - 08 - 27 & @xmath32 & @xmath33 & clr & @xmath34 + & & & & 2007 - 08 - 28 & @xmath35 & @xmath36 & std & + 3 & sc4_fiii & 00:46:55.20 & -73:15:36.00 & 2007 - 08 - 27 & @xmath37 & @xmath38 & clr & @xmath34 + & & & & 2007 - 08 - 28 & @xmath39 & @xmath40 & std & + 4 & sc4_fiv & 00:48:04.80 & -72:53:24.00 & 2007 - 08 - 27 & @xmath41 & @xmath41 & clr & @xmath34 + & & & & 2007 - 11 - 24 & @xmath42 & @xmath43 & std & + 5 & sc7_fv & 00:56:43.20 & -73:10:30.00 & 2007 - 08 - 28 & @xmath44 & @xmath45 & std & @xmath46 + 6 & sc8_fvi & 00:57:50.40 & -72:36:36.00 & 2007 - 08 - 28 & @xmath47 & @xmath47 & std & @xmath48 + 7 & sc7_fvii & 00:55:09.15 & -72:40:12.80 & 2008 - 12 - 14 & @xmath49 & @xmath50 & std & @xmath46 + 8 & sc7_fviii & 00:56:11.44 & -73:02:08.70 & 2008 - 12 - 14 & @xmath51 & @xmath52 & std & @xmath46 + 9 & sc10_fix & 01:04:38.40 & -72:48:00.00 & 2008 - 12 - 15 & @xmath53 & @xmath54 & std & @xmath55 [ tabfields ] ccccccc + ogle004211.68 - 730329.2 & ab & @xmath60 & sc2_fi & @xmath61 & @xmath62 & + ogle004225.29 - 730349.8 & ab & @xmath63 & sc2_fi & @xmath64 & @xmath65 & + ogle004126.84 - 730355.2 & ab & @xmath66 & sc2_fi & @xmath67 & @xmath68 & + ogle004154.36 - 730539.3 & ab & @xmath69 & sc2_fi & @xmath70 & @xmath71 & + ogle004156.46 - 730641.7 & ab & @xmath72 & sc2_fi & @xmath73 & @xmath74 & blend + ogle004817.88 - 731815.2 & ab & @xmath75 & sc4_fii & @xmath76 & @xmath77 & + ogle004744.69 - 731919.1 & ab & @xmath78 & sc4_fii & @xmath79 & @xmath80 & + ogle004805.65 - 732033.9 & ab & @xmath81 & sc4_fii & @xmath82 & @xmath83 & + ogle004724.96 - 731427.2 & ab & @xmath84 & sc4_fiii & @xmath85 & @xmath86 & + ogle004649.00 - 731544.5 & ab & @xmath87 & sc4_fiii & @xmath88 & @xmath89 & + ogle004650.36 - 731652.2 & ab & @xmath90 & sc4_fiii & @xmath91 & @xmath92 & + ogle004755.07 - 725141.5 & ab & @xmath93 & sc4_fiv & @xmath94 & @xmath95 & + ogle004805.17 - 725144.4 & ab & @xmath96 & sc4_fiv & @xmath97 & @xmath98 & + ogle004754.45 - 725212.1 & ab & @xmath99 & sc4_fiv & @xmath100 & @xmath101 & + ogle004756.26 - 725217.6 & ab & @xmath102 & sc4_fiv & @xmath103 & @xmath104 & + ogle004800.43 - 725222.7 & ab & @xmath105 & sc4_fiv & @xmath106 & @xmath107 & + ogle004758.51 - 725430.6 & ab & @xmath108 & sc4_fiv & @xmath109 & @xmath110 & + ogle004744.89 - 725530.1 & c & @xmath111 & sc4_fiv & @xmath112 & @xmath113 & + ogle004817.37 - 725536.5 & ab & @xmath114 & sc4_fiv & @xmath115 & @xmath116 & + + ogle004211.68 - 730329.2 & ab & @xmath60 & sc2_fi & @xmath117 & @xmath118 & + ogle004225.29 - 730349.8 & ab & @xmath63 & sc2_fi & @xmath119 & @xmath120 & + ogle004126.84 - 730355.2 & ab & @xmath66 & sc2_fi & @xmath121 & @xmath122 & + ogle004154.36 - 730539.3 & ab & @xmath69 & sc2_fi & @xmath123 & @xmath124 & + ogle004156.46 - 730641.7 & ab & @xmath72 & sc2_fi & @xmath125 & @xmath126 & blend + ogle004817.88 - 731815.2 & ab & @xmath75 & sc4_fii & @xmath127 & @xmath128 & + ogle004744.69 - 731919.1 & ab & @xmath78 & sc4_fii & @xmath129 & @xmath130 & + ogle004805.65 - 732033.9 & ab & @xmath81 & sc4_fii & @xmath131 & @xmath132 & + ogle004724.96 - 731427.2 & ab & @xmath84 & sc4_fiii & @xmath127 & @xmath127 & + ogle004649.00 - 731544.5 & ab & @xmath87 & sc4_fiii & @xmath133 & @xmath134 & + ogle004650.36 - 731652.2 & ab & @xmath90 & sc4_fiii & @xmath135 & @xmath136 & + ogle005656.87 - 730850.2 & ab & @xmath137 & sc7_fv & @xmath138 & @xmath139 & + ogle005658.96 - 730850.6 & ab & @xmath140 & sc7_fv & @xmath141 & @xmath142 & contaminated + ogle005616.96 - 730901.6 & ab & @xmath143 & sc7_fv & @xmath144 & @xmath145 & + ogle005614.52 - 731023.8 & ab & @xmath146 & sc7_fv & @xmath147 & @xmath148 & blend + ogle005612.31 - 731222.5 & ab & @xmath149 & sc7_fv & @xmath150 & @xmath151 & + ogle005728.85 - 723454.6 & ab & @xmath152 & sc8_fvi & @xmath153 & @xmath154 & misidentified + ogle005803.73 - 723602.0 & ab & @xmath155 & sc8_fvi & @xmath156 & @xmath157 & blend + ogle005806.86 - 723811.7 & ab & @xmath158 & sc8_fvi & @xmath159 & @xmath160 & + ogle005737.26 - 723819.1 & ab & @xmath161 & sc8_fvi & @xmath162 & @xmath163 & + ogle005752.74 - 723901.8 & ab & @xmath164 & sc8_fvi & @xmath165 & @xmath166 & + + ogle004755.07 - 725141.5 & ab & @xmath93 & sc4_fiv & @xmath167 & @xmath168 & + ogle004805.17 - 725144.4 & ab & @xmath96 & sc4_fiv & @xmath169 & @xmath170 & + ogle004754.45 - 725212.1 & ab & @xmath99 & sc4_fiv & @xmath171 & @xmath172 & + ogle004756.26 - 725217.6 & ab & @xmath102 & sc4_fiv & @xmath173 & @xmath174 & + ogle004800.43 - 725222.7 & ab & @xmath105 & sc4_fiv & @xmath175 & @xmath176 & + ogle004758.51 - 725430.6 & ab & @xmath108 & sc4_fiv & @xmath177 & @xmath178 & + ogle004744.89 - 725530.1 & c & @xmath111 & sc4_fiv & @xmath179 & @xmath180 & + ogle004817.37 - 725536.5 & ab & @xmath114 & sc4_fiv & @xmath181 & @xmath182 & + + ogle005451.72 - 723850.4 & c & @xmath183 & sc7_fvii & @xmath184 & @xmath185 & contaminated + ogle005527.97 - 724136.8 & c & @xmath186 & sc7_fvii & @xmath187 & @xmath188 & + ogle005621.27 - 730046.9 & c & @xmath189 & sc7_fviii & @xmath190 & @xmath191 & + ogle005606.56 - 730242.1 & ab & @xmath192 & sc7_fviii & @xmath193 & @xmath194 & + ogle005609.42 - 730323.2 & ab & @xmath195 & sc7_fviii & @xmath196 & @xmath197 & + + ogle010409.82 - 724611.9 & ab & @xmath198 & sc10_fix & @xmath199 & @xmath200 & + ogle010442.62 - 724627.5 & ab & @xmath201 & sc10_fix & @xmath202 & @xmath203 & + ogle010450.99 - 724700.3 & ab & @xmath204 & sc10_fix & @xmath205 & @xmath206 & + ogle010501.92 - 724932.8 & ab & @xmath207 & sc10_fix & @xmath208 & @xmath209 & + ogle010448.82 - 724945.3 & ab & @xmath210 & sc10_fix & @xmath211 & @xmath212 & + ogle010432.13 - 724958.4 & ab & @xmath213 & sc10_fix & @xmath214 & @xmath215 & [ taballstars ] ccccccccc ogle004211.68 - 730329.2 & ab & @xmath60 & sc2_fi & @xmath216 & @xmath217 + ogle004225.29 - 730349.8 & ab & @xmath63 & sc2_fi & @xmath218 & @xmath219 + ogle004126.84 - 730355.2 & ab & @xmath66 & sc2_fi & @xmath220 & @xmath221 + ogle004154.36 - 730539.3 & ab & @xmath69 & sc2_fi & @xmath222 & @xmath223 + ogle004817.88 - 731815.2 & ab & @xmath75 & sc4_fii & @xmath224 & @xmath225 + ogle004744.69 - 731919.1 & ab & @xmath78 & sc4_fii & @xmath226 & @xmath227 + ogle004805.65 - 732033.9 & ab & @xmath81 & sc4_fii & @xmath228 & @xmath229 + ogle004724.96 - 731427.2 & ab & @xmath84 & sc4_fiii & @xmath230 & @xmath231 + ogle004649.00 - 731544.5 & ab & @xmath87 & sc4_fiii & @xmath232 & @xmath233 + ogle004650.36 - 731652.2 & ab & @xmath90 & sc4_fiii & @xmath234 & @xmath235 + ogle004755.07 - 725141.5 & ab & @xmath93 & sc4_fiv & @xmath236 & @xmath237 + ogle004805.17 - 725144.4 & ab & @xmath96 & sc4_fiv & @xmath238 & @xmath239 + ogle004754.45 - 725212.1 & ab & @xmath99 & sc4_fiv & @xmath240 & @xmath241 + ogle004756.26 - 725217.6 & ab & @xmath102 & sc4_fiv & @xmath242 & @xmath243 + ogle004800.43 - 725222.7 & ab & @xmath105 & sc4_fiv & @xmath244 & @xmath245 + ogle004758.51 - 725430.6 & ab & @xmath108 & sc4_fiv & @xmath246 & @xmath247 + ogle004744.89 - 725530.1 & c & @xmath111 & sc4_fiv & @xmath248 & @xmath249 + ogle004817.37 - 725536.5 & ab & @xmath114 & sc4_fiv & @xmath250 & @xmath251 + ogle005656.87 - 730850.2 & ab & @xmath137 & sc7_fv & @xmath252 & @xmath253 + ogle005616.96 - 730901.6 & ab & @xmath143 & sc7_fv & @xmath254 & @xmath255 + ogle005612.31 - 731222.5 & ab & @xmath149 & sc7_fv & @xmath256 & @xmath257 + ogle005806.86 - 723811.7 & ab & @xmath158 & sc8_fvi & @xmath258 & @xmath259 + ogle005737.26 - 723819.1 & ab & @xmath161 & sc8_fvi & @xmath260 & @xmath261 + ogle005752.74 - 723901.8 & ab & @xmath164 & sc8_fvi & @xmath262 & @xmath263 + ogle005527.97 - 724136.8 & c & @xmath186 & sc7_fvii & @xmath264 & @xmath265 + ogle005621.27 - 730046.9 & c & @xmath189 & sc7_fviii & @xmath266 & @xmath267 + ogle005606.56 - 730242.1 & ab & @xmath192 & sc7_fviii & @xmath268 & @xmath269 + ogle005609.42 - 730323.2 & ab & @xmath195 & sc7_fviii & @xmath270 & @xmath271 + ogle010409.82 - 724611.9 & ab & @xmath198 & sc10_fix & @xmath272 & @xmath273 + ogle010442.62 - 724627.5 & ab & @xmath201 & sc10_fix & @xmath274 & @xmath275 + ogle010450.99 - 724700.3 & ab & @xmath204 & sc10_fix & @xmath276 & @xmath277 + ogle010501.92 - 724932.8 & ab & @xmath207 & sc10_fix & @xmath278 & @xmath279 + ogle010448.82 - 724945.3 & ab & @xmath210 & sc10_fix & @xmath280 & @xmath281 + ogle010432.13 - 724958.4 & ab & @xmath213 & sc10_fix & @xmath282 & @xmath283 [ tabrrlyr ] ccccc free fit & @xmath284 & @xmath285 & @xmath286 & @xmath287 + sollima et al .
( 2008 ) & @xmath288 & @xmath127 & @xmath289 & @xmath290 + bono et al .
( 2003 ) & @xmath288 & @xmath127 & @xmath291 & @xmath292 + catelan et al .
( 2004 ) & @xmath293 & @xmath294 & @xmath295 & @xmath296 [ tabpl ]
|
we have obtained deep infrared @xmath0 and @xmath1 band observations of nine @xmath2 arcmin fields in the small magellanic cloud ( smc ) with the eso new technology telescope equipped with the sofi infrared camera . in these fields ,
34 rr lyrae stars catalogued by the ogle collaboration were identified . using different theoretical and empirical calibrations of the infrared period - luminosity - metallicity relation , we find consistent smc distance moduli , and find a best true distance modulus to the smc of @xmath3 ( statistical ) @xmath4 0.12 ( systematic ) mag which agrees well with most independent distance determinations to this galaxy , and puts the smc 0.39 mag more distant than the lmc for which our group has recently derived , from the same technique , a distance of 18.58 mag .
|
no - cloning theorem is one of the most fundamental theorem in quantum computation and quantum information[1 ] .
the theorem states that there does not exist any process , which turns two distinct nonorthogonal quantum states @xmath0,@xmath1 into states @xmath2,@xmath3 respectively.this restrictions can be successfully utilized in quantum cryptography[2 ] .
although we can not copy an unknown quantum state perfectly but one can always do it approximately . beyond the no - cloning theorem , one can clone an arbitrary quantum state perfectly with some non - zero probability[3].in the past years , much progress has been made in designing quantum cloning machine .
a first step towards the construction of approximate quantum cloning machine was taken by buzek and hillery in 1996 [ 4].they showed that the quality of the copies produced by their machine remain same for all input state.this machine is popularly known as universal quantum cloning machine ( uqcm ) .
later this uqcm was proved to be optimal [ 5 ] .
after that the different sets of quantum cloning machines like the set of universal quantum cloning machines , the set of state dependent quantum cloning machines ( i.e. the quality of the copies depend on the input state ) and the probabilistic quantum cloning machines were proposed .
entanglement[6 ] , the heart of quantum information theory , play a crucial role in computational and communicational purposes . therefore , as a valuable resource in quantum information processing , quantum entanglement has been widely used in quantum cryptography [ 7,19],quantum superdense coding [ 8 ] and quantum teleportation [ 9 ] .
an astonishing feature of quantum information processing is that information can be `` encoded '' in non - local correlations between two separated particles .
the more `` pure '' is the quantum entanglement , the more `` valuable '' is the given two - particle state .
therefore , to extract pure quantum entanglement from a partially entangled state , researchers had done lot of works in the past years on purification procedures[10].in other words , it is possible to compress locally an amount of quantum information.now generally a question arises : whether the opposite is true or not i.e. can quantum correlations be `` decompressed '' ?
this question was tackled by several researchers [ 11,12 ] using the concept of `` broadcasting of quantum inseparability '' .
broadcasting is nothing but a local copying of non - local quantum correlations . that is the entanglement originally shared by a single pair
is transferred into two less entangled pairs using only local operations .
suppose two distant parties a and b share two qubit entangled state + @xmath4 is real and @xmath5 complex and the parameters satisfying the relation @xmath6 .
+ the first qubit belongs to a and the second belongs to b. each of the two parties now perform local copier on their own qubit and then it turns out that for some values of @xmath7 , + ( 1 ) non - local output states are inseparable , and + ( 2 ) local output states are separable .
+ in classical theory one can always broadcast information but in quantum theory , broadcasting is not always possible .
h.barnum et.al . showed that non - commuting mixed states can not be broadcasted [ 16 ] .
however for pure states broadcasting is equivalent to cloning .
+ v.buzek et.al .
were the first who showed that the decompression of initial quantum entanglement is possible , i.e. that from a pair of entangled particles , two less entangled pairs can be obtained by local operation .
that means inseparability of quantum states can be partially broadcasted ( cloned ) with the help of local operation .
they used optimal universal quantum cloners for local copying of the subsystems and showed that the non - local outputs are inseparable if @xmath8 further s.bandyopadhyay et.al .
[ 12 ] studied the broadcasting of entanglement and showed that only those universal quantum cloners whose fidelity is greater than @xmath9 are suitable because only then the non - local output states becomes inseparable for some values of the input parameter @xmath7 .
they proved that an entanglement is optimally broadcast only when optimal quantum cloners are used for local copying and also showed that broadcasting of entanglement into more than two entangled pairs is not possible using only local operations .
i.ghiu investigated the broadcasting of entanglement by using local @xmath10 optimal universal asymmetric pauli machines and showed that the inseparability is optimally broadcast when symmetric cloners are applied [ 21 ] .
+ motivated from the previous works on broadcasting of entanglement , we investigate the problem of secretly broadcasting of three - qubit entangled state between two distant partners with universal quantum cloning machine and then the result is generalized to generate secret entanglement among three parties .
three - qubit entanglement between two distant partners can be generated as follows : let us suppose that the two distant partners share an entangled state @xmath11 .
the two parties then apply optimal universal quantum cloning machine on their respective qubits to produce four qubit state @xmath12 .
one party ( say , alice ) then performs measurement on her quantum cloning machine state vectors .
after that she inform bob about her measurement result using goldenberg and vaidman s quantum cryptographic scheme based on orthogonal states . getting measurement result from alice , other partner ( say , bob ) also performs measurement on his quantum cloning machine state vectors and using the same cryptographic scheme , he sends his measurement outcome to alice .
since the measurement results are interchanged secretly so alice and bob share secretly four qubit state .
they again apply the cloning machine on one of their respective qubits and generate six qubit state @xmath13 .
therefore , each parties have three qubit each . among six qubit state
, we interestingly find that there exists two three qubit state shared by alice and bob which are entangled for some values of the input parameter @xmath14 .
+ in the second part , we investigate the problem of secret entanglement broadcasting among three distant parties . to solve this problem ,
we start with the result of the first part i.e. we assume that the two distant partners ( say , alice and bob ) shared a three qubit entangled state . without any loss of generality , we assume that among three qubits , two are with alice and one with bob
. then alice teleport one of the qubit to the third distant partner ( say , carol ) .
after the completion of the teleportation procedure , we find that the three distant partners shared a three qubit entangled state for the same values of the input parameters @xmath14 as in the first part of the protocol . + in broadcasting of inseparability
, we generally use peres - horodecki criteria to show the inseparability of non - local outputs and separability of local outputs .
+ * * peres - horodecki theorem [ 13,14]:**the necessary and sufficient condition for the state @xmath15 of two spins @xmath16 to be inseparable is that at least one of the eigen values of the partially transposed operator defined as @xmath17 is negative .
this is equivalent to the condition that at least one of the two determinants + + @xmath18= [ cols="^,^,^ " , ] + + is negative .
+ + for the security of the broadcasting of entanglement , we use l.goldenberg et.al . quantum cryptographic scheme which was based on orthogonal states
the cryptographic scheme is described by figure - i .
+ all the previous works on the broadcasting of entanglement deals with the generation of two 2-qubit entangled state starting from a 2-qubit entangled state using either optimal universal symmetric cloner [ 4,5 ] or asymmetric cloner [ 24,25 ] .
the generated two qubit entangled state can be used as a quantum channel in quantum cryptography , quantum teleportation etc .
the advantage of our protocol over other protocols of broadcasting is that we are able to provide a protocol which generates secret quantum channel between distant partners .
the introduced protocol generate two 3-qubit entangled state between two distant partners starting from a 2-qubit entangled state and also provide the security of the generated quantum channel . not only that we also generalize our protocol from two parties to three parties and show that the generated 3-qubit entangled states can serve as a secured quantum channel between three parties . now to hack the quantum information , hackers have to do two things : first , they have to gather knowledge about the initially shared entangled state and secondly , they have to collect information about the measurement result performed by two distant partners . therefore , the quantum channel generated by our protocol is more secured and hence can be used in various protocols viz .
quantum key distribution protocols [ 22,23 ] .
+ we then distribute our work in the remaining three sections .
in section 2 , we present our idea with a specific example for broadcasting of three - qubit entangled state shared between two distant partners . in section 3
, we generalize this idea to generate three - qubit entangled state shared between three distant parties .
to implement the idea , we use the concept of entanglement swapping .
the last section is devoted to the conclusion .
in this section , firstly we define broadcasting of three qubit entanglement , open entanglement and close entanglement .
+ let the previously shared entangled state ( 1 ) described by the two qubit density operator be @xmath19 . using b - h quantum cloning machine twice by the distant partners ( alice and bob ) on their respective qubits
, they generate total six - qubit state @xmath20 between them . therefore , alice has three qubits 1,2 and 5 and bob possesses three qubits 3 ,
4 and 6. + * definition-1 : * the three - qubit entanglement is said to be broadcast if ( i ) any of the two local outputs ( say ( @xmath21,@xmath22 ) in alice s side and ( @xmath23,@xmath24 in bob s side ) are separable ( ii ) one local output ( say @xmath25 in alice s side and @xmath26 in bob s side ) is inseparable and associated with these local inseparable output , two non - local outputs ( say ( @xmath27,@xmath28 ) and ( @xmath29,@xmath30 ) are inseparable .
+ + * definition-2 : * an entanglement is said to be closed if each party has non - local correlation with other parties .
for instance , any three particle entangled state described by the density operator @xmath31 is closed if @xmath32,@xmath25 and @xmath28 are entangled state .
otherwise , it is said to be an open entanglement .
+ closed entanglement and open entanglement is shown in figure - vi and figure - vii respectively .
+ + now we are in a position to discuss our protocol for secretly broadcasting of three qubit entangled state .
we start the protocol with two qubit entangled state @xmath33 shared between two distant partners popularly known as alice and bob .
particles 1 and 3 possessed by alice and bob respectively .
alice and bob then operates quantum cloning machine on their respective qubits . after cloning procedure
, alice perform measurement on the quantum cloning machine state vector and send the measurement result to bob . after getting measurement result from alice ; bob perform measurement on his quantum cloning machine state vector and send the measurement result to alice . consequently
, the two distant partners share a four qubit state @xmath34 .
now alice has two qubits 1 and 2 and bob 3 and 4 respectively .
both of them again operates quantum cloning machine on one of the qubits , they possess . as a result
, the distant parties now share six qubit state @xmath13 in which three qubits 1,2 and 5possessed by alice and remaining three qubits 3,4 and 6 possessed by bob . now
if there exists two 3-qubit entangled state between two distant partners for some values of the input parameter @xmath14 , then only we can secretly broadcast 3-qubit entangled state using only universal quantum cloning machine .
the word secretly is justified by observing an important fact that the transmission of measurement result from alice to bob and bob to alice has been done by using goldenberg and vaidman s quantum cryptographic scheme . therefore , message regarding measurement results can be transmitted secretly between two distant partners .
hence , the broadcasted three - qubit entangled state is only known to alice and bob and not to the third party eve. as a result , these newly generated three - qubit entangled states can be used as a secret quantum channel in various quantum cryptographic scheme .
+ now to understand our protocol more clearly , we again discuss the whole protocol below by considering a specific example .
+ + + let the two particle entangled state shared by two distant partners alice and bob is given by @xmath4 is real and @xmath5 is complex with @xmath35 .
this initial entangled state is shown in figure - ii .
+ + + the b - h quantum copier is given by the transformation + @xmath36 @xmath37 where @xmath38 and @xmath39 are orthogonal quantum cloning machine state vectors .
+ alice and bob then operates b - h quantum cloning machine locally to copy the state of their respective particles . therefore , after operating quantum cloning machine , alice and bob both of them are able to approximate clone the state of the particle and consequently the combined system of four qubits is given by @xmath40|q_{0}\rangle^a+[(\frac{\sqrt{2}\alpha}{3}|\psi^+\rangle|00\rangle+ \frac{\sqrt{2}\beta}{3}|11\rangle|\psi^+\rangle)|q_{0}\rangle^b+(\frac{\alpha}{3}|\psi^+\rangle|\psi^+\rangle + { } \nonumber\\&&\frac{2\beta}{3}|1111\rangle)|q_{1}\rangle^b]|q_{1}\rangle^a\end{aligned}\ ] ] the subscript 1,2 and 3,4 refers to two approximate copy qubits in the alice s and bob s side respectively . also @xmath41 and @xmath42 denotes quantum cloning machine state vectors in alice s and bob s side respectively .
this fact is explained by figure - iii . +
alice then performs measurement on the quantum cloning machine state vectors in the basis @xmath43 .
thereafter , alice inform bob about her measurement result using goldenberg and vaidman s quantum cryptographic scheme based on orthogonal states which is discussed in the previous section . after getting measurement result from alice ,
bob also performs measurement on the quantum cloning machine state vectors in the basis @xmath44 and then using the same cryptographic scheme , he sends his measurement outcome to alice . in this way alice and bob interchange their measurement results secretly .
+ + + + after measurement , let the state shared by alice and bob is given by @xmath45\end{aligned}\ ] ] where @xmath46 represents the normalization factor .
+ afterward , alice and bob again operates their respective cloners on the qubits 2 and 4 respectively and therefore , the total state of six qubits is given by @xmath47\end{aligned}\ ] ] now our task is to see whether we can generate two 3-qubit entangled state from above six qubit state or not . to examine the above fact
, we have to consider two 3-qubit state described by the density operators @xmath48 and @xmath31 . to understand more clearly , see figure - iv . + the density operator @xmath48 is given by @xmath49\end{aligned}\ ] ] the density operator @xmath31 describes the other three qubit state looks exactly the same as @xmath48 .
+ now to show the state described by the density operator @xmath48 is entangled , we have to show that the two qubit states described by the density operators @xmath29,@xmath50 and @xmath26 are entangled i.e. we have to show that there exist some values of the input state parameter @xmath14 for which the three - qubit state is a closed entangled state .
+ the reduced density operators @xmath29,@xmath50 and @xmath26 are given by @xmath51\end{aligned}\ ] ] @xmath52\end{aligned}\ ] ] now using peres - horodecki theorem , we find that the state described by the density operators @xmath50 and @xmath29 are entangled if @xmath53 and the state described by the density operator @xmath26 is entangled if @xmath54 .
therefore , we can say that the state described by the density operator @xmath48 is a closed three qubit entangled state if @xmath54 .
similarly , the other reduced density operator @xmath31 describe a closed entangled state if @xmath54 .
+ also the other two - qubit state described by the density operators @xmath21,@xmath22,@xmath23 and @xmath55 are given by @xmath56\end{aligned}\ ] ] these density operators are separable only when @xmath57 .
hence , broadcasting of three - qubit entangled state is possible when @xmath54 .
+ now , our task is to find out how is the entanglement distributed over the state i.e. how much are the two qubit density operators @xmath50 , @xmath29 and @xmath26 are entangled . to evaluate the amount of entanglement , we have to calculate the concurrence defined by wootters [ 20 ] and hence entanglement of formation .
+ wootters gave out , for the mixed state @xmath15 of two qubits , the concurrence is + @xmath58 where the @xmath59 , in decreasing order , are the square roots of the eigen values of the matrix @xmath60 and @xmath61 denotes the complex conjugation of @xmath62 in the computational basis @xmath63 and @xmath64 is the pauli operator . + the entanglement of formation @xmath65
can then be expressed as a function of c , namely @xmath66 after a little bit calculation , we find that the concurrence and hence the entanglement of formation depends on the probability @xmath14 .
therefore , we have to calculate the amount of entanglement in the 2-qubit states described by the reduced density operators @xmath67 and @xmath26 in the range @xmath54 because the two qubit reduced density operators are entangled in this range of the input state parameter @xmath68 .
since concurrence depends on @xmath14 so it varies as @xmath14 varies .
therefore , when @xmath54 , the concurrences for the mixed states described by density operators @xmath69 varies from 0.17 to 0.29 while the concurrence for the mixed states described by density operators @xmath26 varies from 0.08 to 0.15 respectively .
using the relation ( 16 ) and the values of concurrence , we find that the entanglement of formation for the density oprators @xmath70 varies from 0.06 to 0.15 while the entanglement of formation for the density operator @xmath26 varies from 0.01 to 0.03 respectively .
therefore , the generated three - qubit entangled state is a weak closed entangled state in the sense that the amount of entanglement in the two - qubit density operators are very low .
further , the above results shows that the entanglement between the qubits 1 and 6 ( 1 and 4 ) is higher than between the the qubits 4 and 6 .
+ furthermore , if the measurement results are either @xmath71 + or @xmath72 , then the two 3-qubit state described by the density operators @xmath48 and @xmath31 are different and the broadcasting is possible for @xmath73 or @xmath74 according to the outcomes .
also if the outcome of the measurement is @xmath75 , then the state described by the density operators @xmath48 and @xmath31 are identical and the broadcasting is possible for @xmath76 . + +
in this section , we attempt to answer a question : can we secretly generate two 3-qubit entangled state shared between three distant partners using locc ?
the answer is in affirmative .
now we show below that the 3-qubit entangled state shared between three distant partners can be generated by two different processes .
+ to generate three - qubit entangled state between three distant partners , we require only two well - known concept : ( i ) quantum cloning and ( ii ) entanglement swapping + + entanglement swapping [ 17,18 ] is a method that enables one to entangle two quantum systems that do not have direct interaction with one another .
s.bose et.al .
[ 17 ] generalized the procedure of entanglement swapping and obtained a scheme for manipulating entanglement in multiparticle systems .
they showed that this scheme can be regarded as a method of generating entangled states of many particles .
an explicit scheme that generalizes entanglement swapping to the case of generating a 3-particle ghz state from three bell pairs has been presented by zukowski et.al . the standard entanglement swapping helps to save a significant amount of time when one wants to supply two distant users with a pair of atoms or electrons ( or any particle possessing mass ) in a bell state from some central source .
the entanglement swapping can be used , with some probability which we quantify , to correct amplitude errors that might develop in maximally entangled states during propagation . in this work ,
we use the concept of entanglement swapping in the generation of three - qubit entanglement between three distant partners .
+ now we are in a position to discuss the protocol for secretly generation of two 3-qubit entangled state between three distant partners via quantum cloning and entanglement swapping .
+ + let us suppose for the implementation of any particular cryptographic scheme , three distant partners alice , bob and carol wants to generate two three qubit entangled state between them . to do the same task ,
let us assume that initially alice - bob and carol - alice share two qubit entangled states described by the density operators @xmath19 , @xmath77 , where alice has qubits 1 and 8 , bob and carol possess qubit 3 and 7 respectively .
then alice and bob adopting the broadcasting process described in the previous section to generate two three - qubit entangled state in between them .
therefore , alice and bob now have two 3-qubit entangled state described by the density operators @xmath48 and @xmath31 where alice has qubits 1,2and 5 and bob possesses 3,4 and 6. now we are in a position for the illustration of the generation of 3-qubit entangled between three parties at distant places by using the concept of entanglement swapping .
+ without any loss of generality , we take a three - qubit entangled state between two distant parties described by the density operator @xmath31 . + the density operator @xmath31 can be rewritten as @xmath78\end{aligned}\ ] ] where qubits 2 and 5 possessed by alice and qubit 3 possessed by bob respectively .
+ to achieve the goal of the generation of three qubit entangled state between three distant partners , we proceed in the following way : + let alice and carol shared a singlet state @xmath79 alice then perform bell state measurement on the particles 2 and 8 in the basis + @xmath80 , where @xmath81 , @xmath82 + if the measurement result is @xmath83 , then the 3-qubit density operator is given by @xmath84 + { } \nonumber\\&&\frac{\alpha\beta^*}{27}(|001\rangle\langle111|-|001\rangle\langle100|+|000\rangle\langle110|-|011\rangle\langle110| ) + \frac{\alpha\beta}{27}(-|110\rangle\langle011|+|110\rangle\langle000|{}\nonumber\\&&+|111\rangle\langle001|-|100\rangle\langle001| ) + \frac{|\beta|^2}{36}[\frac{2}{3}(|010\rangle\langle010|+|001\rangle\langle001|+|110\rangle\langle110|+|101\rangle\langle101| ) { } \nonumber\\&&+\frac{1}{3}(|011\rangle\langle011|-|011\rangle\langle000|-|000\rangle\langle011|+|000\rangle\langle000| + |111\rangle\langle111|-|111\rangle\langle100|-{}\nonumber\\&&|100\rangle\langle111|+|100\rangle\langle100|)]\end{aligned}\ ] ] after bell - state measurement , alice announces publicly the measurement result .
thereafter , alice , bob and carol operate an unitary operator @xmath85 on their respective particles to retrieve the state described by the density operator @xmath31 .
+ if the measurement result is @xmath86 or @xmath87 or @xmath88 then accordingly they operate an unitary operator @xmath89 or @xmath90 or @xmath91 on their respective particles to retrieve the state described by the density operator @xmath31 .
+ hence , we find that after getting the measurement result , each party ( alice , bob and carol ) apply the suitable unitary operator on their respective particles to share the 3-qubit entangled state in between them , which is previously shared between only two distant partners alice and bob .
+ also we note an important fact that the generated 3-qubit entangled state is totally secret between three distant partners because the outcome of the measurement on the machine state vector is totally unknown to the eavesdropper .
furthermore , the reduced density operator describing 3-qubit state between two distant partners and the reduced density operator describing 3-qubit state between three distant partners are entangled for the same range of @xmath14 .
+ + we can understand the above protocol pictorially ( figure - v and figure - vi ) : + figure - v : alice and bob share a 3-qubit entangled state described by the density operator @xmath31 .
alice and carol share a singlet state described by the density operator @xmath92
. then alice perform bell - state measurement ( bsm ) on particles 2 and 8 of the joint state described by the density operator @xmath93 .
+ figure - vi : finally , after applying suitable unitary operators , 3-qubit entangled state described by the density operator @xmath31 is generated between three distant partners alice , carol and bob .
+ therefore , in this section we describe the secretly generation of 3-qubit entangled state between three distant partners starting from 3-qubit entangled state shared between two distant partners using quantum cloning and entanglement swapping .
this quantum channel generated by the above procedures can be regarded as a secret quantum channel because the result of the measurement on the machine state vectors transmitted secretly by quantum cryptographic scheme .
in this work , we present a protocol for the secret broadcasting of three - qubit entangled state between two distant partners . here we should note an important fact that the two copies of three - qubit entangled state is not generated from previously shared three - qubit entangled state but from previously shared two - qubit entangled state using quantum cloning machine .
they send their measurement result secretly using cryptographic scheme so that the produced copies of the three - qubit entangled state shared between two distant parties can serve as a secret quantum channel .
we also extend this idea to create three - particle entangled state secretly between three distant partners using quantum cloning and entanglement swapping procedure .
+ + * acknowledgement * + s.a wishes to acknowledge the support given by csir project f.no.8/3(38)/2003-emr-1 , new delhi .
s.a would like to thank samir kunkri for useful discussion with him .
@xmath94 $ ] w.k wootters and w.h.zurek , nature 299,802(1982 ) + @xmath95 $ ] a.ekert , phys.rev.lett . 67 ( 1991)661 + @xmath96 $ ]
l.m.duan and g.c.guo,phys.rev.lett .
80 ( 1998)4999 + @xmath97 $ ] v.buzek and m.hillery , phys.rev.a 54 ( 1996)1844 + @xmath98 $ ] d.bruss , p.d.divincenzo , a.ekert , c.a.fuchs , c.macchiavello and j.a.smolin , phys.rev.a 57 ( 1998)2368 + @xmath99 $ ] einstein , podolsky and rosen , phys.rev .
47 ( 1935)777 + @xmath100 $ ] c.h.bennett and g.brassard , proceedings of ieee international conference on computers , system and signal processing , bangalore , india , 1984 , pp.175 - 179 + @xmath101 $ ] c.h.bennett and s.j.weisner , phys.rev.lett.69 ( 1992)2881 + @xmath102 $ ] c.h.bennett , g.brassard , c.crepeau,r.jozsa , a.peres and w.k.wootters,phys.rev.lett
. 70 ( 1993)1895 + @xmath103 $ ] c.h.bennett , h.j.bernstein , s.popescu and b.schumacher , phys.rev.a 53 ( 1996)2046 ; d.deutsch , a.ekert , r.jozsa , c.macchiavello,s.popescu and a.sanpera , phys.rev.lett .
77 ( 1996)2818 + @xmath104 $ ] v.buzek , v.vedral , m.b.plenio , p.l.knight and m.hillery , phys.rev.a 55 ( 1997)3327 + @xmath105 $ ] s.bandyopadhyay , g.kar , phys.rev.a 60 ( 1999)3296 + @xmath106 $ ] a.peres , phys.rev.lett .
77 ( 1996)1413 + @xmath107 $ ] m.horodecki , p.horodecki , r.horodecki , phys.lett.a 223 ( 1996)1 + @xmath108 $ ] l.goldenberg and l.vaidman , phys.rev.lett .
75 ( 1995)1239 + @xmath109 $ ] h.barnum , c.m.caves , c.a.fuchs , r.jozsa and b.schumacher , phys.rev.lett .
76 ( 1996)2818 + @xmath110 $ ] s.bose , v.vedral and p.l.knight , phys.rev.a 57 ( 1998)822 + @xmath111 $ ] m.zukowski , a.zeilinger , m.a.horne and a.k.ekert , phys.rev.lett .
71 ( 1993)4287 + @xmath112 $ ] p.w.shor and j.preskill , phys.rev.lett .
85 ( 2000)441 + @xmath113 $ ] w.k.wootters , phys.rev.lett .
80 ( 1998)2245 + @xmath114 $ ] i.ghiu , phys.rev.a 67 ( 2003)012323 + @xmath115 $ ] a.cabello , phys.rev.a 61 ( 2000)052312 + @xmath116 $ ] c.li , h - s song and l.zhou , journal of optics b : quantum semiclass
. opt . 5 ( 2003)155 + @xmath117 $ ] n.j .
cerf , phys .
84 , 4497 ( 2000 ) .
+ @xmath118 $ ] n.j .
cerf , j. mod .
47 , 187 ( 2000 ) .
|
in this work , we investigate the problem of secretly broadcasting of three - qubit entangled state between two distant partners .
the interesting feature of this problem is that starting from two particle entangled state shared between two distant partners we find that the action of local cloner on the qubits and the measurement on the machine state vector generates three - qubit entanglement between them .
the broadcasting of entanglement is made secret by sending the measurement result secretly using cryptographic scheme based on orthogonal states .
further we show that this idea can be extended to generate three particle entangled state between three distant partners .
|
the fluctuations at the interface between two fluid phases at thermodynamic equilibrium have been studied very extensively starting from the beginning of this century @xcite , and particular investigation has concerned those at the interfaces of critical fluids @xcite . although the features of equilibrium interfacial fluctuations are now relatively well known , the behavior of an interface under nonequilibrium conditions is still not well understood .
many experiments have been performed to detect an effective nonequilibrium surface tension in miscible fluids @xcite but the results obtained are mainly qualitative . in this paper
we will present experimental results about the behavior of interfacial fluctuations during the diffusive remixing of partially miscible phases . in this system
the interface between two fluid phases is being crossed by a macroscopic mass flow .
it is well known that the fluctuations in the equilibrium states before and after the diffusive partial remixing are controlled by surface tension and gravity .
it is not clear , however , what happens to the interface during the remixing : does the interface temporarily dissolve ? is there a surface tension during the transient , and how is it related to the evolving macroscopic state ?
we try to address these problems by means of low - angle light scattering measurements of the correlation function of fluctuations .
+ the sample considered is a near critical binary mixture kept below its critical temperature t@xmath0 , so that it is macroscopically separated into two bulk phases by a sharp horizontal interface .
the diffusive remixing is started by raising the temperature to a value closer to t@xmath0 , but still below it .
+ we report data at equilibrium showing the expected interface capillary waves , and data obtained out of equilibrium during partial diffusive remixing , showing for the first time that capillary waves at the interface are still present and coexist with `` giant '' nonequilibrium fluctuations in the bulk . by combining time - resolved measurements with predictions for the light scattered by the fluctuations , we are in the position to derive data for the time evolution of the nonequilibrium surface tension .
these data show for the first time that during the diffusive remixing the surface tension attains almost instantaneously its final equilibrium value , and this is consistent with a fast rearrangement of the concentration profile in the neighborhood of the interface .
traditionally , interfacial fluctuations have been studied by means of dynamic surface light scattering techniques @xcite .
these techniques allow to determine the power spectrum of the light scattered from the excitations , and have been used very extensively to characterize the equilibrium properties of interfacial fluctuations in simple fluids and binary mixtures @xcite .
surface light scattering is usually performed by sending a probe beam on the interface at the angle of total reflection , scattered light being collected around the specular reflection angle and partially recombined with the main beam , so that a heterodyne signal is obtained .
however , dynamic scattering techniques are not very well suited to study time - dependent nonequilibrium processes , as the time needed to accumulate an adequate statistics of the fluctuations is often much longer than the time related to changes in the macroscopic state of the fluid . to bypass this problem
we have used a unique low - angle static light scattering setup . although the fluctuations timescales are not accessible with this instrument , it allows to determine in a fraction of a second the static light scattered at 31 wavevectors distributed within a two decades range , making it a very useful tool to study fluctuations around a time - dependent macroscopic state .
this instrument , described in detail elsewhere @xcite , typically investigates a wavevector range corresponding to @xmath1 . + our setup
is configured to detect the transmitted static scattered intensity , the probe beam being sent vertically at normal incidence . in this way light is scattered both at the interface and in the bulk layers above and below it .
it can be easily shown that the structure factor in the transmission geometry is proportional to the usual one in reflection . +
a critical binary mixture is an ideal sample to study the nonequilibrium fluctuations during diffusion at an interface .
although other partially miscible fluids could be used to perform this experiment , the use of a critical mixture allows to tune the timescale of the macroscopic diffusion process by adjusting the temperature difference from the critical point .
moreover experimental runs can be iterated simply by cycling the temperature .
+ the sample is a 4.5 mm thick horizontal layer of the binary mixture aniline
cyclohexane prepared at its critical consolution concentration ( c=0.47 w / w aniline ) .
its critical temperature t@xmath0 is about 30@xmath2c , and was determined to within @xmath30.01@xmath2c before each experimental run . while slowly decreasing the temperature of the single phase above t@xmath0 , t@xmath0 was taken as the temperature where a sudden increase of turbidity was observed .
+ the light scattering cell is a modification of the rayleigh - bnard one already used to investigate fluctuations in a thermal diffusion process @xcite , configured to keep the sample at a uniform temperature . the mixture is sandwiched between two massive sapphire windows whose temperature control , achieved with peltier plates , is as good as 3mk over a period of one week , temperature differences between the two plates being kept to about 2mk . +
a typical measurement sequence involves the following procedure .
an optical background is recorded with the mixture in its one phase region at 5.5 k above t@xmath0 , where the bulk fluctuations amplitude is many orders of magnitude smaller than the one of equilibrium and nonequilibrium fluctuations in our experiment .
this optical background mostly contains contributions due to dust and imperfections of the optical elements , and it is subtracted to all subsequent measurements .
the system is then let phase separate at 3.5k below t@xmath0 , where the concentration difference between the two bulk phases is @xmath4@xcite .
great care is dedicated to eliminate wetting drops at the optical windows .
after a few hours the intensity distribution scattered by the system at thermodynamic equilibrium is recorded , scattered light being mostly due to the capillary waves at the interface . the nonequilibrium process
is then started by suddenly increasing the temperature at 0.1@xmath2c below t@xmath0 .
the concentration difference between the phases has to readjust by means of a diffusive process .
the two macroscopic phases at this temperature are not completely miscible . at equilibrium
they will have concentration difference of @xmath5@xcite and they will be separated by a new interface . the scattered intensity distribution is recorded during this transient , until the system reaches thermodynamic equilibrium after about 24 hours .
it is well known that the interfacial fluctuations at the equilibrium states preceding and following the diffusion process are overdamped capillary waves @xcite , characterized by a @xmath6 power spectrum which exhibits a gravitational stabilization at small wavevectors .
the main problem we want to address is how the fluctuations behave during the transient between the equilibrium states .
the evolution of the scattered intensity distributions can be roughly divided into two stages . in the first stage , represented in fig . 1 , the intensity distribution , initially due to the equilibrium interface excitations , increases and develops a bump , which represents the appearance of a typical lengthscale
. the initial data set , being the least intense , is particularly affected by the subtraction of the optical background .
the time elapsed during this stage roughly corresponds to the thermal time needed to increase the temperature of the sample ( about 200 s ) . in the second stage , represented in fig . 2 , the intensity distribution decreases , until eventually the bump disappears , and the @xmath6 equilibrium power spectrum of capillary waves is recovered .
notice that light scattered at large wavevectors does not change with time during this stage .
we will see that this is related to the readjustment of the concentration profile across the interface having taken place . + to analyze our data we assume that the sample may be depicted as the superposition of two thick bulk layers separated by a thin interface layer , and that these layers scatter light independently from each other .
therefore we are considering two sources of scattering , the interface fluctuations scattering @xmath7 and the fluctuations in the bulk phases scattering @xmath8 , so that the total intensity distribution is : @xmath9 where , as outlined in the appendix , @xmath10 and @xmath11 in eqs .
( [ dis2 ] ) and ( [ dis3 ] ) , @xmath12 is the wavevector of light in vacuum , @xmath13 , @xmath14 , @xmath15 are the mixture s refraction index , weight - fraction concentration and density , respectively .
@xmath16 is @xmath17 and @xmath18 is the gravity acceleration .
+ @xmath19 is the concentration difference across the interface and @xmath20 is the total sample concentration difference minus @xmath19 , that is the concentration difference that falls in the bulk phases .
the rolloff wavevectors @xmath21 and @xmath22 , given by @xmath23^{\frac{1}{2 } } \label{qcap}\end{aligned}\ ] ] and @xmath24^{\frac{1}{4 } } , \label{qro}\end{aligned}\ ] ] characterize the onset of gravitational stabilization at large lengthscales of capillary and bulk fluctuations , respectively @xcite . in eq .
( [ qro ] ) @xmath25 represents the largest concentration gradient in the bulk phases at a certain time @xcite .
we use eqs .
( [ dis1])-([dis3 ] ) to fit the experimental data shown in figs .
in order to limit the number of fitting parameters we take advantage of the fact that the concentration difference @xmath26 across the whole sample does not change during the so called free - diffusive regime @xcite , as the diffusive remixing initially involves only layers of fluid close to the interface . therefore @xmath27 and @xmath28
are related by @xmath29 where @xmath30 is the time required for diffusion to occur over the sample height @xmath31 .
@xmath32 is of the order of 7000s for our sample , by assuming d=@xmath33 ( this is the equilibrium value at 3k below t@xmath0 , see @xcite and references therein ) . by imposing the reference value @xmath34 @xcite and by fitting the experimental data using eqs .
( [ dis1])-([dis3 ] ) , we are able to determine three parameters : the concentration difference across the interface and the rolloff wavevectors @xmath21 and @xmath22 . + results for @xmath19 are presented in fig .
3(a ) as a function of time . initially @xmath19 has its equilibrium value @xmath35 .
after about 40 seconds from the temperature increase it begins to drop , and decreases for about 300s , finally stabilizing to the constant value @xmath36 .
this is roughly a factor of two larger than the reference equilibrium value @xcite .
this discrepancy is due to the difficulty ( both theoretical and in the fitting procedure ) of clearly ascribing the measured scattered intensity to either interface or bulk phases , and an estimate of our rather large fitting and systematic error is shown with the error bars in fig .
this error is such that our data can not be considered fully quantitative , but it does not affect the features we discuss . + by combining the results for @xmath19 and @xmath21 , since from eq .
( [ qcap ] ) @xmath37 we are in the unique position to obtain the time evolution of the interfacial surface tension during the nonequilibrium process .
experimental results for the surface tension are shown in fig .
3(b ) , which represents the main accomplishment of this work .
the two crosses mark the value of the equilibrium surface tension at the initial and final temperature , extrapolated from the reference data from atack and rice @xcite .
the first data point represents the equilibrium surface tension measured with our light scattering setup .
the agreement with the reference value is good .
after the diffusion process is started the surface tension drops about two orders of magnitude , until after about 300s it stabilizes to a constant value .
the asymptotic value of the surface tension is about a factor of two larger than the reference value , a good result considering the wide range of values spanned .
although these results are only partially quantitative , fig .
3 unambiguously shows for the first time that the properties of the nonequilibrium interface rapidly attain their equilibrium values .
this equilibration time is very small compared with the one associated to readjustments of the bulk phases ( which corresponds to about one day ) , and it is comparable to the time needed to increase the temperature of the sample . notice that the surface tension evolution does not show the initial delay seen on the @xmath19 evolution .
this indicates that the surface tension is probably following the local temperature almost instantly , whereas @xmath19 does not change until diffusion has occurred over the fluctuations characteristic lenghtscales . with the diffusion coefficient given above this time
is about 30s for the smallest wavevectors observed .
+ we are now in the position to comment the fast growth , shown in fig . 1 , of the scattered intensity distributions at intermediate and large wavevectors .
soon after the diffusion process is started the concentration difference across the interface decreases at its equilibrium value . according to the concentration conservation eq .
( [ dis4 ] ) , a strong concentration difference @xmath38 is rapidly created in the bulk phases , and a large concentration gradient quickly grows near the interface .
this gives rise to velocity - induced concentration fluctuations described by eq .
( [ dis3 ] ) @xcite .
the growth of @xmath39 during the free diffusive regime is shown by the circles in fig . 4 , and it mirrors the results for @xmath40 in fig .
when enough time has passed for diffusing particles to reach the macroscopic boundaries ( about 5000s as from fig .
4 ) , the sample enters the restricted diffusion regime , where the total concentration difference across the sample begins to change , and eq .
( [ dis4 ] ) does not hold any more .
however , according to fig . 3 , we can now assume that the interfacial parameters have attained their asymptotic values .
in this way we can fit the scattered intensity distributions to determine the concentration difference across the bulk phases .
results for @xmath38 obtained in this way are shown by the squares in fig .
the evolution of this process is very similar to that already observed during the free diffusion of completely miscible phases @xcite . from the fitting of the nonequilibrium bulk data we also determine the rolloff wavevector @xmath22 , which corresponds to the bump observed in the light scattering data in figures 1 - 2 . as outlined above and
thoroughly described in ref.@xcite , fluctuations at wavevectors smaller than @xmath22 are stabilized by gravity , which frustrates the @xmath41 divergence at small wavevectors .
the rolloff wavevector is plotted in fig . 5 as a function of time .
the results displayed in fig . 5 show that the variation of @xmath22 roughly corresponds to a factor of 1.4 .
@xmath22 is mostly determined by the layers of fluid where the concentration gradient is largest @xcite , that is the bulk layers close to the interface . from eq .
( [ qro ] ) , we can estimate that the variation of the concentration gradient close to the interface is rather small , roughly corresponding to a factor four .
this behavior suggests that the bulk concentration gradient is pinned to the interface concentration profile .
we have reported light - scattering measurements of the correlation function of fluctuations at the interface and in the bulk phases of a critical binary mixture undergoing diffusive partial remixing . from the intensity distributions
it is clear that even during remixing a sharp interface separates the two bulk phases , and that this interface is roughened by capillary waves similar to those at equilibrium .
our data have been analyzed , yielding qualitative information on the concentration profile of the system .
in particular our main results are the time evolution of the interface concentration difference and the surface tension , showing how these parameters relax to the final equilibrium values quickly compared to the system s macroscopic equilibration time .
+ we thank dr .
giuseppe gonnella for useful comments .
work partially supported by the italian space agency ( asi ) . in this appendix
we shall outline how one may derive the structure factor of fluctuations in an out of equilibrium binary mixture where a concentration gradient and a diffusive concentration flux are present .
we consider a fluid described by a @xmath42 dependent concentration gradient @xmath43 .
layer by layer , surfaces of uniform concentration can be defined . from a macroscopic point of view
these surfaces are horizontal planes . in the presence of fluctuations
the surfaces are corrugated , due to the motion of parcels of fluid in the vertical direction . as customary
, we will indicate by @xmath44 the vertical displacement of the surfaces from their mean position at @xmath42 .
+ in such a system light scattered with scattering vector * q * is proportional to the mean square amplitude of the roughness mode * q * , @xmath45
. we shall show how this quantity may be calculated for the two limiting cases of a sharp interface and a linear concentration gradient , extending the fluctuating hydrodynamics treatment applied in @xcite . + throughout this paper we are dealing with small fluid velocities and overdamped motion , and equations will be approximated accordingly , see for example the discussion in @xcite .
we will consider the scattering of light by long - wavelength fluctuations , which scatter light mostly at small scattering angles in the forward direction .
+ we shall suppose fluctuations to be generated independently at different heights , and shall first consider a single fluctuation generated in a layer @xmath46 at @xmath47 .
then , for every wavevector * q * , one has that @xmath48 will be maximum for @xmath49 where the fluctuation was generated , and exponentially decreasing with distance from @xmath47 like @xmath50 .
this is a consequence of the hydrodynamic equations for a viscous fluid in the case of overdamped motion @xcite .
+ the hydrodynamic equations describing motion in an incompressible viscous mixture can be linearized for small velocities @xcite as : @xmath51 @xmath52 @xmath53 where @xmath54 is the fluctuating velocity , @xmath14 the fluctuating concentration , @xmath55 is the kinematic viscosity , @xmath56 is the pressure , @xmath57 is the concentration gradient averaged over typical fluctuations timescales and @xmath58 is the mass diffusion flux .
equation ( [ divvzero ] ) is the usual mass conservation equation for an incompressible fluid , eq . ( [ motointerfaccia ] )
is the linearized navier stokes equation , and eq . ( [ convdiff ] ) is the convection - diffusion equation .
the macroscopic diffusive flux @xmath58 is proportional to @xmath59 , where @xmath60 is the chemical potential of the mixture ; it is non
zero where @xmath60 is not constant .
+ equiconcentration surfaces are corrugated by thermal velocity fluctuations in the vertical direction , and these are described by the @xmath42 component of eq .
( [ motointerfaccia ] ) .
+ we change variable from @xmath14 to @xmath61 in eq .
( [ convdiff ] ) , and simplify the equation by assuming that concentration diffusion obeys fick s law @xmath62 , where @xmath63 is the `` mutual diffusion '' coefficient . + from eq .
( [ convdiff ] ) we take the @xmath42 component of the velocity , @xmath64 , and substitute it into equation ( [ motointerfaccia ] ) , obtaining , after fourier transform in @xmath65 and @xmath66 , an equation of motion for non - equilibrium fluctuations : @xmath67\,&=&\nonumber\\=\ , 2\,\nu\,q^2\,\left[\,-i\,\omega\,h_{{\bf q},\omega}(z)\,+\,2d\,q^2\,h_{{\bf q},\omega}(z ) \,\right]\,-\,\frac{1}{\rho}\,q\,p_{{\bf q},\omega}(z)\,&+&\,\frac{1}{\rho}\,q\,s_{{\bf q},\omega},\nonumber\\
\label{motointerfacciah2diff}\end{aligned}\ ] ] where the continuity of tangential stresses on the fluctuating surface has been imposed and a stocastic force term @xmath68 has been added to describe the onset of thermal spontaneous velocity fluctuations .
the correlation function of this stocastic force is assumed to be ( see @xcite ) @xmath69 as in equilibrium .
+ the pressure _
p _ in eq .
( [ motointerfacciah2diff ] ) may be exerted by the external gravity force and by internal capillary forces and is measured against the average fluid pressure . because of eqs .
( [ divvzero])-([motointerfaccia ] ) , in deriving eq .
( [ motointerfacciah2diff ] ) we have imposed that the pressure @xmath70 induced by the fluctuation in layer @xmath47 decays exponentially with distance from the fluctuation layer to the average pressure , like @xmath71 .
this pressure term depends on the local concentration profile .
it can be written explicitly for fluctuations involving a `` sharp interface '' profile and for those in the bulk phases , by considering the gravity and capillary forces on the fluctuations .
we shall consider a fluctuation to be a sharp - interface fluctuation if @xmath47 is close to the interface position @xmath72 . instead ,
if @xmath73 may be considered linear within a range of @xmath74 around @xmath47 , the fluctuation will be considered a bulk fluctuation .
our treatment is approximate in that we are supposing that every fluctuation falls within one of these cases .
+ interface fluctuations and bulk fluctuations give rise to different dynamics .
+ in the case of a sharp interface , the pressure acting on a fluid element at height @xmath75 , when a fluctuation occurs at @xmath72 bending the interface , is : @xmath76 .
\label{frozagravcapb}\end{aligned}\ ] ] substituting this pressure in eq .
( [ motointerfacciah2diff ] ) one may , with algebra similar to that in @xcite , calculate the spectrum of interface fluctuations : + @xmath77 where the static term is : @xmath78 at equilibrium , when the mutual diffusion coefficient @xmath79 , the spectrum linewidth is the same as that of the well - known equilibrium interface @xcite : gravity dominates at small wavevectors and capillary forces at large ones .
however , during the nonequilibrium process , diffusion is effective in relaxing large wavevector fluctuations . at these wavevectors
the linewidth becomes the usual diffusive @xmath80 one , and this diffusive contribution is also apparent in the static structure factor where a new term proportional to @xmath81 is present in the denominator . with our setup
it was not possible to study the structure factor at large enough wavevectors to check the result and , as far as we know , this feature has never been observed .
+ the other case is that of a bulk phase where the concentration gradient is approximately linear . here the pressure opposing a fluid element at height @xmath47 when a fluctuation occurs bending the layer at height @xmath47 is : @xmath82\,\frac{2}{q } , \label{frozagravcapa}\end{aligned}\ ] ] where @xmath83 stands for @xmath84 . + from mean field theories of binary systems @xcite
it follows that @xmath85 .
if the concentration gradient is small , the capillary term becomes negligible compared to the gravitational one , and we shall drop it from now on in this case . as before , from eq .
( [ motointerfacciah2diff ] ) one may calculate the correlation function of bulk fluctuations , recovering the spectrum calculated and commented in ref .
@xcite : + @xmath86 where the static term is @xmath87 although these bulk fluctuations have the same origin as capillary waves , namely velocity fluctuations parallel to the concentration gradient , the transition from a sharp to a diffuse interface radically modifies both the static and the dynamic structure factor of the fluctuations .
+ we shall now outline how one may evaluate the intensity of light scattered by these fluctuations .
suppose a plane wave having wavevector @xmath12 in vacuum and intensity @xmath88 is propagating in the vertical direction in a sample having index of refraction @xmath89 .
the roughness of the equiconcentration surfaces due to a fluctuation at @xmath47 introduces a phase dependancy on @xmath90 .
one easily sees that the resulting scattered intensity per solid angle is @xmath91 , where @xmath92 is the optical path variation induced by the fluctuation given by @xmath93 this integral is similar to that required to calculate the pressures of eqs .
( [ frozagravcapb ] ) and ( [ frozagravcapa ] ) and , depending on the system s local concentration profile at @xmath47 , it can be easily approximated in the same two limiting cases considered above , sharp interface or concentration gradient in bulk phase .
scattering from the whole sample is given by integration over the sample thickness of the scattering due to fluctuations arising in a single layer , weighed with the probability of being in that layer and with the density of fluctuations , so that one integrates the single fluctuation intensity in @xmath94 .
+ if the sample comprises both bulk regions and a sharp interface having a concentration difference @xmath27 , the total scattered light is then : @xmath95 ^ 2\left|h_{\bar{z}}({\bf q } , t)\right|^2 \right]\nonumber\\ \label{scattintb3}\end{aligned}\ ] ] we have used eq .
( [ scattintb3 ] ) to fit our data after approximating the bulk phase integral by considering that most of the scattering is due to the layers with the greatest @xmath96 .
l.i.mandelstam , _ ann .
der physik _ * 41 * , 609 ( 1913 ) .
r.h.katyl,and u.ingard , _ phys .
_ * 20 * , 248 ( 1968 ) .
j.s.huang , and w.w.webb , _ phys .
* 23 * , 160 ( 1969 ) .
j.zollweg , g.hawkins , and g.b.benedek , _ phys .
* 27 * , 1182 ( 1971 ) .
m.a.bouchiat , and j.meunier , _ j. de physique colloque _ * 33 * , c1 - 141 ( 1972 ) .
j. mech . , b / fluids _ * 9 * , 565 ( 1990 ) .
s.e.may , and j.v.maher , _ phys .
lett . _ * 67 * , 2013 ( 1991 ) .
p. petitjeans , _ c. r. acad .
paris _ * 322 * , 673 ( 1996 ) .
d.h.vlad , and j.v.maher , _ phys .
* 59 * , 476 ( 1999 ) .
d.langevin , _ light scattering by liquid surfaces and complementary techniques .
_ ( dekker , new york , 1992 ) .
m.a.bouchiat , and j.meunier , _ j. de physique _ * 32 * , 5611 ( 1971 ) .
m.carpineti et al . , _ phys . rev .
a _ * 42 * , 7347 ( 1990 ) . a.vailati , and m.giglio , _ phys .
lett . _ * 77 * , 1484 ( 1996 ) .
d.atack , and o.k.rice , _ discuss .
faraday soc . _
* 15 * , 210 ( 1953 ) .
a.vailati , and m.giglio , _ nature _ * 390 * , 262 ( 1997 ) .
a.vailati , and m.giglio , _ phys .
e _ * 58 * , 4361 ( 1998 ) .
d.brogioli , a.vailati , and m.giglio , _ phys .
e _ * 61 * , r1 ( 2000 ) .
c.cohen , j.w.h.sutherland , and j.m.deutch , _ phys .
* 2 * , 213 ( 1971 ) .
j.w.cahn , and j.e.hilliard , _ j. chem .
_ * 28 * , 258 ( 1958 ) .
s.a.safran , _ statistical thermodynamics of surfaces , interfaces and membranes . _
( addison - wesley , reading , 1994 ) .
|
we have performed small - angle light - scattering measurements of the static structure factor of a critical binary mixture undergoing diffusive partial remixing .
an uncommon scattering geometry integrates the structure factor over the sample thickness , allowing different regions of the concentration profile to be probed simultaneously .
our experiment shows the existence of interface capillary waves throughout the macroscopic evolution to an equilibrium interface , and allows to derive the time evolution of surface tension .
interfacial properties are shown to attain their equilibrium values quickly compared to the system s macroscopic equilibration time .
pacs numbers : 68.10.-m , 68.35.fx , 05.40.-a , 68.35.rh .
|
the drell - yan process @xcite , in which a charged lepton pair is produced in a hadron - hadron interaction via the electromagnetic @xmath0 process , has provided unique information on parton distributions . in particular , the drell - yan process has been used to determine the antiquark contents of nucleons and nuclei @xcite , as well as the quark distributions of pions , kaons , and antiprotons @xcite .
such information is difficult , if not impossible , to obtain from dis experiments .
as the drell - yan process can be well described by next - to - leading order qcd calculations @xcite , a firm theoretical framework exists for utilizing the drell - yan process to extract the parton distributions . during the last two decades , a series of fixed - target dimuon production experiments ( e772 , e789 , e866 ) have been carried out using 800 gev / c proton beam at fermilab . at 800 gev / c ,
the dimuon data contain drell - yan continuum up to dimuon mass of @xmath1 gev as well as quarkonium productions ( j/@xmath2 , @xmath3 , and @xmath4 resonances ) . the drell - yan process and quarkonium productions often provide complementary information , since drell - yan is an electromagnetic process via quark - antiquark annihilation while the quarkonium production is a strong interaction process dominated by gluon - gluon fusion at this beam energy .
the fermilab dimuon experiments covers a broad range of physics topics .
the drell - yan data have provided informations on the antiquark distributions in the nucleons @xcite and nuclei @xcite .
these results showed the surprising results that the antiquark distributions in the nuclei are not enhanced @xcite , contrary to the predictions of models which explain the emc effect in term of nuclear enhancement of exchanged mesons .
moreover , the drell - yan cross section ratios @xmath5 clearly establish the flavor asymmetry of the @xmath6 and @xmath7 distributions in the proton , and they map out the @xmath8-dependence of this asymmetry @xcite
. pronounced nuclear dependences of quarkonium productions have been observed for j/@xmath2 , @xmath3 , and @xmath4 resonances @xcite .
several review articles covering some of these results are available @xcite . in this article
, we will focus on the recent results from experiment e866 and future prospect of dimuon experiments at fermilab and j - parc .
in the cern na51 @xcite and fermilab e866 @xcite experiments on proton - induced dimuon production , a striking difference was observed for the drell - yan cross sections between @xmath9 and @xmath10 .
as the underlying mechanism for the drell - yan process involves quark - antiquark annihilation , this difference has been attributed to the asymmetry between the up and down sea quark distributions in the proton . from the @xmath11 ratios
the bjorken-@xmath8 dependence of the sea - quark @xmath12 flavor asymmetry has been extracted @xcite . at @xmath13 of 500 gev using various pdfs . ]
future fixed - target dimuon experiments have been proposed at the 120 gev fermilab main injector ( fmi ) and the 50 gev j - parc facilities .
the fermilab proposal @xcite , e906 , has been approved and is expected to start data - taking around 2011 . two dimuon proposals ( p04 @xcite and p24 @xcite )
have also been submitted to the j - parc for approval .
the lower beam energies at fmi and j - parc present opportunities for extending the @xmath14 and the nuclear antiquark distribution measurements to larger @xmath8 ( @xmath15 ) . for given values of @xmath16 and @xmath17 ,
the drell - yan cross section is proportional to @xmath18 , hence a gain of @xmath19 times in the drell - yan cross sections can be obtained at the j - parc energy of 50 gev . since the perturbative process gives a symmetric @xmath20 while non - perturbative processes are necessary to generate an asymmetric @xmath20 sea , it would be very important to extend the drell - yan measurements to kinematic regimes beyond the current limits .
another advantage of lower beam energies is that a much more sensitive study of the partonic energy loss in nuclei could be carried out using the drell - yan nuclear dependence @xcite . to disentangle the @xmath12 asymmetry from the possible charge - symmetry violation effect @xcite
, one could consider @xmath21 boson production in @xmath22 collision at rhic . an interesting quantity to be measured is the ratio of the @xmath23 and @xmath24 cross sections @xcite .
it can be shown that this ratio is very sensitive to @xmath12 .
an important feature of the @xmath21 production asymmetry in @xmath22 collision is that it is completely free from the assumption of charge symmetry .
another advantage of @xmath21 production in @xmath9 collision is that no nuclear effects need to be considered .
finally , the @xmath21 production is sensitive to @xmath20 flavor asymmetry at a @xmath25 scale of @xmath26 6500 gev@xmath27c@xmath28 , significantly larger than all existing measurements .
this offers the opportunity to examine the qcd evolution of the sea - quark flavor asymmetry .
figure 1 shows the predictions @xcite for @xmath22 collision at @xmath29gev .
the mrs s0@xmath30 corresponds to the @xmath31 symmetric parton distributions , while the other three parton distribution functions are from recent global fits with asymmetric @xmath14 sea - quark distributions .
figure 1 clearly shows that @xmath21 asymmetry measurements at rhic could provide an independent determination of @xmath12 .
first results from the lhc are also expected soon , and will provide additional information @xcite .
unlike the electromagnetic drell - yan process , quarkonium production is a strong interaction dominated by the subprocess of gluon - gluon fusion at 800 gev beam energy .
therefore , the quarkonium production cross sections are primarily sensitive to the gluon distributions in the colliding hadrons .
the @xmath4 production ratio , @xmath32 , is expected to probe the gluon content in the neutron relative to that in the proton @xcite .
cross section ratios for @xmath4 resonances as a function of @xmath17 .
the corresponding ratios for drell - yan cross sections are also shown .
the error bars are statistical only . ]
the @xmath33 ratios for @xmath34 production are shown in fig . 2 as a function of @xmath17 @xcite .
these ratios are consistent with unity , in striking contrast to the corresponding values for the drell - yan process , also shown in fig . 2 .
the difference between the drell - yan and the @xmath4 cross section ratios clearly reflect the different underlying mechanisms in these two processes . for @xmath4 production
, the dominance of the gluon - gluon fusion subprocess at this beam energy implies that @xmath35 .
figure 2 shows that the gluon distributions in the proton ( @xmath36 ) and neutron ( @xmath37 ) are very similar over the @xmath17 range @xmath38 .
the overall @xmath39 ratio , integrated over the measured kinematic range , is @xmath40 .
the @xmath4 data indicate that the gluon distributions in the proton and neutron are very similar .
the upcoming fermilab e906 experiment is expected to provide a precise measurement of the @xmath41 ratio .
these data should further test the equality of the gluon distributions in the proton and neutron .
moreover , these data could identify emc effects for deuteron .
it is interesting to note that analogous ratios measured in dis and the drell - yan can not be used to determine the emc effects for deuteron , since different parton distributions are being probed for the hydrogen and the deuterium targets .
if one assumes that the gluon distributions are identical in proton and neutron , then any deviation of the @xmath41 ratio from unity would be attributed to the emc effects of gluon distributions in deuteron .
despite a quarter century of significant experimental and theoretical effort , the specific origins of the observed @xmath42 dependence of the nuclear quark distributions have yet to be unambiguously identified .
attempts to explain the emc effect have led to a large collection of theoretical models @xcite , many of which are capable of describing the essential features of the data , however the underlying physics mechanisms in each model are often very different .
a new calculation of the modifications of nucleon quark distributions in the nuclear medium has recently been reported @xcite . in this approach by clot , bentz and thomas ( cbt )
, the nambu jona - lasinio model is used to describe the coupling of the quarks in the bound nucleons to the scalar and vector mean fields inside a nucleus .
these nucleon quark distributions are then convoluted with a nucleon momentum distribution in the nucleus to generate the nuclear quark distributions @xcite .
an important feature of this model is that for @xmath43 nuclei ( where @xmath44 and @xmath45 refer to the number of neutrons and protons ) the isovector
vector mean field ( usually denoted by @xmath46 ) will affect the up quarks differently from the down quarks in the bound nucleons .
therefore , this model has a novel prediction that the @xmath47 and @xmath48 quarks have distinct nuclear modifications for @xmath43 nuclei .
semi - inclusive dis ( sidis ) on heavy nuclear targets , in which the flavor of the struck quark is tagged by the detected hadron , is a promising experimental tool to search for the flavor - dependent emc effect .
recently , lu and ma @xcite pointed out that charged lepton sidis off nuclear targets and the deuteron can be used to probe the flavor content of the nuclear quark sea , which can help distinguish between the various models of the emc effect .
indeed , a sidis experiment @xcite aiming at a precise determination of flavor dependence of the emc effect has also been proposed at the upgraded @xmath49gev jlab facility .
( upper left ) , @xmath50 ( upper right ) and @xmath51 ( lower ) versus the predictions using nuclear pdfs of the cbt model for tungsten ( red solid ) and @xmath52 nuclear matter ( blue dashed ) . ]
the flavor dependence of the emc effect is a promising experimental observable to distinguish among the plethora of models that can describe the emc effect . as discussed earlier
, the nuclear dependence of proton - induced drell - yan process measured at fermilab has shown the surprising results that the antiquark distributions in the nuclei are not enhanced @xcite , contrary to the predictions of many emc models .
pion - induced drell - yan process provides another experimental tool with which search for flavor - dependent effects in the nuclear modification of the nucleon structure functions @xcite . to explore the sensitivity of pion - induced drell - yan processes to a flavor - dependent emc effect
, we consider the three ratios @xmath53 , @xmath54 and @xmath55 , where @xmath42 represents a nuclear , d a deuteron and h a hydrogen target . assuming isospin symmetry , which implies @xmath56 , @xmath57 , @xmath58 , @xmath59 and keeping only the dominant terms in each cross - section , one readily obtains @xmath60
@xmath61 @xmath62 the up and down nuclear quark distributions are labeled by @xmath63 and @xmath64 respectively , @xmath65 is the up quark distribution in the deuteron and @xmath66 the up quark distribution in the proton .
eqs . 1 - 3
demonstrate that these drell - yan cross - section ratios are very sensitive to the flavor dependence of the emc effect .
3 shows the comparison @xcite of the calculations of the pion - induced drell - yan cross - section ratios with the existing data .
the top left panel shows the ratio of @xmath67 from the na10 experiment @xcite .
these plots contain both the @xmath68 and @xmath69gev data sets , which are very similar .
the calculations are performed at @xmath70gev , since most of the data was obtained at this energy .
the pdfs of the cbt model @xcite at a fixed @xmath25 of @xmath71gev @xmath28 , which is approximately the mean @xmath25 of the na10 experiment , were used .
the top right panel shows the ratio @xmath72 from the na3 experiment @xcite .
the data were collected using a @xmath73gev @xmath74 beam and the @xmath25 range covered was @xmath75gev@xmath28 .
the calculations are performed for @xmath76gev and @xmath77gev@xmath28 . the solid curves in fig . 3 are calculations using the flavor - dependent @xmath78 and @xmath79 from the cbt model with @xmath80 , corresponding approximately to the @xmath81 values for the au , w and pt nuclei .
the dashed curves correspond to the calculated ratios using the nuclear pdfs from the cbt model for @xmath52 .
since @xmath82 in this case , the dashed curves are representative of the predictions for flavor - independent emc models .
figure 3 shows that the na10 data do not exhibit a clear preference for the flavor - dependent versus flavor - independent nuclear pdfs .
in contrast , the na3 data strongly favor the calculations using flavor - dependent nuclear pdfs .
the existing drell - yan data are not sufficiently accurate yet , although the na3 data clearly favor the flavor - dependent over the flavor - independent nuclear pdfs .
precise future pion - induced drell - yan experiments can provide unique constraints that will help distinguish the various theoretical models and most importantly shed new light on the origins of the emc effect .
the study of the transverse momentum dependent ( tmd ) parton distributions of the nucleon has received much attention in recent years as it provides new perspectives on the hadron structure and qcd @xcite .
these novel tmds can be extracted from semi - inclusive deep - inelastic scattering ( sidis ) experiments .
recent measurements of the sidis by the hermes @xcite and compass @xcite collaborations have shown clear evidence for the existence of the t - odd sivers functions .
these data also allow the first determination @xcite of the magnitude and flavor structure of the sivers functions and the nucleon transversity distributions .
the tmd and transversity parton distributions can also be probed in drell - yan experiments . as pointed out @xcite long time ago , the double transverse spin asymmetry in polarized drell - yan , @xmath83 , is proportional to the product of transversity distributions , @xmath84 . the single transverse spin asymmetry , @xmath85 , is sensitive to the sivers function @xcite , @xmath86 of the polarized proton ( beam or target ) .
even unpolarized drell - yan experiments can be used to probe the tmd distribution function , since the cos@xmath87 azimuthal angular dependence is proportional to the product of two boer - mulders functions @xcite , @xmath88 .
a unique feature of the drell - yan process is that , unlike the sidis , no fragmentation functions are involved .
therefore , the drell - yan process provides an entirely independent technique for measuring the tmd functions .
furthermore , the proton - induced drell - yan process is sensitive to the sea - quark tmds and can lead to flavor separation of tmds when combined with the sidis data .
finally , the intriguing prediction @xcite that the t - odd tmds extracted from dis will have a sign - change for the drell - yan process remains to be tested experimentally .
no polarized drell - yan experiments have yet been performed .
however , some information on the boer - mulders functions have been extracted recently from the azimuthal angular distributions in the unpolarized drell - yan process .
the general expression for the drell - yan angular distribution is @xcite @xmath89 where @xmath90 and @xmath91 are the polar and azimuthal decay angle of the @xmath92 in the dilepton rest frame .
boer showed that the @xmath93 term is proportional to the convolution of the quark and antiquark boer - mulders functions in the projectile and target @xcite .
this can be understood by noting that the drell - yan cross section depends on the transverse spins of the annihilating quark and antiquark .
therefore , a correlation between the transverse spin and the transverse momentum of the quark , as represented by the boer - mulders function , would lead to a preferred transverse momentum direction .
pronounced @xmath94 dependences were indeed observed in the na10 @xcite and e615 @xcite pion - induced drell - yan experiments , and attributed to the boer - mulders function .
the first measurement of the @xmath94 dependence of the proton - induced drell - yan process was recently reported for @xmath9 and @xmath10 interactions at 800 gev / c @xcite .
in contrast to pion - induced drell - yan , significantly smaller ( but non - zero ) cos@xmath87 azimuthal angular dependence was observed in the @xmath9 and @xmath10 reactions . while the pion - induced drell - yan process is dominated by annihilation between a valence antiquark in the pion and a valence quark in the nucleon , the proton - induced drell - yan process involves a valence quark in the proton annihilating with a sea antiquark in the nucleon .
therefore , the @xmath9 and @xmath10 results suggest @xcite that the boer - mulders functions for sea antiquarks are significantly smaller than those for valence quarks .
future fixed - target dimuon experiments have been proposed at the 120 gev fermilab main injector and the 50 gev j - parc facilities . as discussed earlier
, the fermilab e906 experiment will extend the @xmath20 asymmetry measurement to larger @xmath8 region .
another goal of this experiment is to determine the antiquark distributions in nuclei at large @xmath8 using nuclear targets .
new information on the quark energy loss in nuclei is also expected .
as discussed earlier , an advantage of lower beam energies is that a much more sensitive study of the partonic energy loss in nuclei could be carried out @xcite . with the possibility to accelerate polarized proton beams at j - parc @xcite , the spin structure of the proton
can also be investigated with the proposed dimuon experiments .
in particular , polarized drell - yan process with polarized beam and/or polarized target at j - parc would allow a unique program on spin physics complementary to polarized dis experiments and the rhic - spin programs .
specific physics topics include the measurements of t - odd boer - mulders distribution function in unpolarized drell - yan , the extraction of t - odd sivers distribution functions in singly transversely polarized drell - yan , the helicity distribution of antiqaurks in doubly longitudinally polarized drell - yan , and the transversity distribution in doubly transversely polarized drell - yan .
it is worth noting that polarized drell - yan is one of the major physics program at the gsi polarized antiproton experiment ( pax ) .
the compass experiment at cern will also measure @xmath74-induced drell - yan on transversely polarized targets to extract sivers functions from the single - spin - asymmetry .
two recent letter - of - intent have also been submitted to rhic for measuring polarized drell - yan at rhic - spin , one in collider mode and the other utilizing internal targets .
it is clear that the drell - yan process will continue to provide unique information on parton distributions in the forseeable future .
99 s.d . drell and t.m .
yan , phys .
lett . * 25 * , 316 ( 1970 ) ; ann .
( ny ) * 66 * , 578 ( 1971 ) .
mcgaughey , j.m .
moss , and j.c .
peng , annu .
nucl . part .
sci . * 49 * , 217 ( 1999 ) .
kenyon , rep .
* 45 * , 1261 ( 1982 ) ; k. freudenreich , int . j. mod
a5 * , 3643 ( 1990 ) .
w.j . stirling and m.r .
whalley , j. phys . *
g19 * , d1 ( 1993 ) .
et al . _ ,
lett . * 69 * , 1726 ( 1992 ) .
et al . _ ,
lett . * 80 * , 3715 ( 1998 ) .
j.c . peng _
et al . _ ,
d * 58 * , 092004 ( 1998 ) .
et al . _ ,
d * 64 * , 052002 ( 2001 ) .
* 64 * , 2479 ( 1990 ) .
et al . _ ,
lett . * 83 * , 2304 ( 1999 ) .
d.m . alde _
et al . _ ,
lett . * 66 * , 133 ( 1991 ) .
d.m . alde _
et al . _ ,
lett . * 66 * , 2285 ( 1991 ) .
et al . _ ,
* 72 * , 1318 ( 1994 ) .
et al . _ ,
d * 52 * , 4251 ( 1995 ) .
g.t . garvey and j.c .
peng , prog .
. phys . * 47 * , 203 ( 2001 ) .
reimer , j. phys .
g * 34 * , s107 ( 2007 ) .
a. baldit _
b * 332 * , 244 ( 1994 ) . http://www.phy.anl.gov/mep/drell-yan ; d. geesaman , p. reimer , _ et al .
_ , fermilab proposal p906 ( 1999 ) .
http://j-parc.jp/nuclpart/pac_0606/pdf/p04-peng.pdf ; j.c .
sawada , _ et al .
_ , j - parc proposal p04 ( 2006 ) .
http://j-parc.jp/nuclpart/pac_0801/pdf/goto.pdf ; y. goto , h. sato , _ et al .
_ , j - parc proposal p24 ( 2007 ) . g.t . garvey and j.c .
peng , phys .
lett . * 90 * , 092302 ( 2003 ) .
ma , phys .
* b274 * ( 1992 ) 111 .
c. boros , j.t .
londergan and a.w .
thomas , phys .
* 81 * ( 1998 ) 4075 .
peng and d.m .
jansen , phys .
b354 * ( 1995 ) 460 .
r. yang , j.c .
peng and m. grosse - perdekamp , phys .
lett . * b6 80 * ( 2009 ) 231 .
g. piller and a.w .
thomas , z. phys .
c * 70 * , 661 ( 1996 ) .
* 100 * ( 2008 ) 062301 .
d. f. geesaman , k. saito and a. w. thomas , ann . rev . nucl . part .
sci . * 45 * , 337 ( 1995 ) .
p. r. norton , rept .
* 66 * , 1253 ( 2003 ) .
i. c. clot , w. bentz and a. w. thomas , phys .
b * 642 * , 210 ( 2006 ) .
i. c. clot , w. bentz and a. w. thomas , phys .
lett . * 102 * , 252301 ( 2009 ) .
b. lu and b .- q .
ma , phys .
c * 74 * , 055202 ( 2006 ) .
d. dutta , d. gaskell , k. hafidi _
_ , pr12 - 09 - 004 , jlab proposal to pac-34 , unpublished ( 2009 ) .
d. dutta , j.c .
peng , i.c .
clot , and d. gaskell , [ arxiv : nucl - ex/1007.3916 ] .
p. bordalo _ et al .
b * 193 * , 368 ( 1987 ) .
a. michelini , cern - ep/81 - 128 ( 1981 ) ; j. badier _
et al . _ ,
b * 104 * , 335 ( 1981 ) . v. barone , a. drago , and p. g. ratcliffe , phys . rep . * 359 * , 1 ( 2002 ) .
a. airapetian _
* 94 * , 012002 ( 2005 ) ; * 101 * , 152002 ( 2009 ) . v. yu
et al . _ ,
lett . * 94 * , 202002 ( 2005 ) ; m. alekseev _
et al . _ ,
b * 673 * , 127 ( 2009 ) .
w. vogelsang and f. yuan , phys .
d * 72 * , 054028 ( 2005 ) ; m. anselmino _
et al . _ ,
d * 72 * , 094007 ( 2005 ) ; m. anselmino _
et al . _ ,
j. * a39 * , 89 ( 2009 ) .
ralston and d.e .
soper , nucl .
b * 152 * , 109 ( 1979 ) .
d. sivers , phys .
d * 41 * , 83 ( 1990 ) .
d. boer and p.j .
mulders , phys .
d * 57 * , 5780 ( 1998 ) .
collins , phys .
b * 536 * , 43 ( 2002 ) .
lam and w.k .
tung , phys .
d * 18 * , 2447 ( 1978 ) .
d. boer , phys .
d * 60 * , 014012 ( 1999 ) .
s. falciano _
et al . _ , z. phys .
c * 31 * , 513 ( 1986 ) .
et al . _ ,
d * 39 * , 92 ( 1989 ) ; j.g .
et al . _ ,
phys . rev .
d * 44 * , 1909 ( 1991 ) .
lett . * 99 * , 082301 ( 2007 ) ; * 102 * , 182001 ( 2009 ) .
b. zhang , z. lu , b .- q .
ma , and i. schmidt , phys .
d * 77 * , 054011 ( 2008 ) .
|
dimuon production has been studied in a series of fixed - target experiments at fermilab during the last two decades .
highlights from these experiments , together with recent results from the fermilab e866 experiment , are presented .
future prospects for studying the parton distributions in the nucleons and nuclei using dimuon production are also discussed .
address = university of illinois at urbana - champaign , urbana , il 61801
|
only a fraction of the volume occupied by the intra - cluster medium is currently mapped with x - ray observations .
the typical measures of the surface brightness , temperature and metal abundance extend out to a fraction of the virial radius , @xmath5 , and only in a few cases meaningful estimates at @xmath6 ( @xmath1 ) is the radius enclosing a mean matter over - density of 500 ( 200 ) times the critical density of the universe .
compared to the cluster virial radius these radii correspond to @xmath7 and to @xmath8 . ] or beyond are obtained .
because of this , more than two - thirds of the typical cluster volume , where the primordial gas is accreted and the dark matter ( dm ) halo is forming , are still unknown for what concerns the gas / dm mass distribution and the related thermodynamical properties .
this poses a limitation in our ability to characterize the physical processes presiding over the formation and evolution of clusters , and thus to use them as high - precision cosmological tools ( see ettori & molendi 2010 and references therein ) . in general the surface brightness distribution , requiring lower counts statistic , is easier to be characterized and has been successfully studied with _
rosat _ pspc ( e.g. the outskirts of the perseus cluster in ettori et al .
1998 ; more systematic studies have been carried out from vikhlinin et al .
1999 and neumann 2005 ) and with _ chandra _
( e.g. ettori & balestra 2009 ) . on the other hand
the measure of the gas temperature is more strongly limited from the background emission .
current exposures with _ chandra _ ( e.g. vikhlinin et al . 2006 ) and _ xmm _
( e.g. leccardi & molendi 2008 ) enabled the reconstruction of the azimuthally - averaged temperature profile for few tens of nearby galaxy clusters out to @xmath9 .
a step forward has been recently obtained from the observations of a handful sample of nearby x - ray luminous objects with the japanese satellite _ suzaku _ that , despite the relatively poor psf and small fov of its x - ray imaging spectrometer ( xis ) , benefits from the modest background associated to its low earth orbit ( see results on pks0745 - 191 , george et al . 2009 ; a2204 , reiprich et al .
2009 ; a1795 , bautz et al . 2009 ; a1413 , hoshino et al .
2010 , a1689 , kawaharada et al.2010 ) .
however as consequence of the small fov ( and of the large solid angle covered from the bright nearby clusters ) these observations provide often discrepant constraints over a few arbitrarily chosen directions . a number of numerical and semi - analitical works in the last few years aimed at providing reliable models of the thermal properties of galaxy clusters at their outskirts , from the theoretical point of view ( e.g.loken et al.2002 ; roncarelli et al.2006 ; burns , skillman & oshea 2010 ; lapi , fusco - femiano & cavaliere 2010 ; cavaliere , lapi & fusco - femiano 2010 ; vazza et al.2010 ) . quite recently ,
a detailed analysis of the cluster profiles up to @xmath10 was reported by burns et al.(2010 ) , using a sample of clusters simulated with the enzo code ( see sec.[sec : methods ] ) . in this article
, we aim at characterizing the intrinsic azimuthal scatter ( due to the 3d structure of clusters ) that should affect at some level the measurements of the thermodynamical properties in the cluster outskirts , and at providing a guidance in interpreting the existing and incoming observations . to this end
, we extended a method proposed by roncarelli et al.(2006 ) to compute the cluster profiles by avoiding the bias caused by small collapsed gas clumps , and we measured the level of azimuthal scatter around the average clusters profiles in a total sample of 29 massive galaxy clusters . the paper is organized as follows : in sec.[sec : methods ] we describe the set of simulations ; in sec.[sec : results ] we present the average properties of the azimuthal scatter of the gas density , temperature , entropy and x - ray luminosity , with a comparison with the observational constraints available .
the discussion of our results is reported in sec.[sec : conclusions ] . in the appendix ,
additional test on the physical and numerical convergence of the main results of the paper are discussed .
$ ] kev for cluster e14 of the enzo sample at @xmath11 .
the left panel is for the original data , the right panel is for the data after the 99 per cent filtering .
the side of the slices is @xmath12mpc / h , the pixel size is @xmath13.,scaledwidth=49.0%,scaledwidth=40.0% ] .
the dotted / dashed lines refer to each cluster in the sample , the thick solid lines show the average profiles for each of the two samples.,title="fig:",scaledwidth=45.0%,scaledwidth=40.0% ] .
the dotted / dashed lines refer to each cluster in the sample , the thick solid lines show the average profiles for each of the two samples.,title="fig:",scaledwidth=45.0%,scaledwidth=40.0% ] .
the dotted / dashed lines refer to each cluster in the sample , the thick solid lines show the average profiles for each of the two samples.,title="fig:",scaledwidth=45.0%,scaledwidth=40.0% ]
we analyze two independent sets of high - resolution ( @xmath14kpc / h ) cosmological simulations of galaxy clusters , extracted from large cosmological boxes ( @xmath15 mpc ) and re - simulated with two of the most tested numerical codes for high performance computing : the adapative mesh refinement code enzo and the tree smoothed particles hydrodynamics code gadget2 .
the main cosmological and numerical parameters adopted are listed in table 1 .
a data - set of 20 non - radiative re - simulations of galaxy clusters with total masses @xmath16 was produced with the 1.5 public version of enzo 1.5(norman et al.2007 and references therein ) , with a few physical implementation ( vazza et al.2010 , hereafter va10 ) .
for the specific details about the simulational setup and the post - processing procedure adopted we address the reader to va10 ; a public archive of part of this data - sample is currently available at _
massive halos in enzo runs were identified following a spherical over - density method , applied to the simulated outputs at @xmath11 ( e.g. gheller , pantano & moscardini 1998 ) .
all clusters with total masses larger than @xmath17 were extracted and sampled at the highest available resolution in the simulation @xmath13 out to the distance of @xmath18 from the cluster centers . a second sample of 9 non - radiative re - simulations of galaxy clusters with masses @xmath19 was obtained with gadget2 ( springel 2005 ) . the details for the simulational setup of these runs can be found in dolag et al.(2005 ) and roncarelli et al.(2006 ) , hereafter ro06 .
also in this case , the halos at @xmath11 were identified basing on a spherical over - density method ; the extraction procedure was tailored in order to have a sample of obviously non - merging systems across the mass range , discarding double systems or systems with ongoing strong interactions .
[ tab : tab1 ] .numerical and cosmological parameters for the simulations : code , mass resolution for dm , gravitational softening , maximum gas spatial resolution @xmath20 , cosmological dm density , cosmological baryon density , normalization of the primordial density power spectrum .
gadget2 runs adopted @xmath21 while enzo runs adopted @xmath22 . [ cols="^,^,^,^,^,^,^,^ " , ] , for n=2 ( top panels ) and n=4 ( bottom panels ) .
the meaning of colors and linestyles is an in fig.[fig : prof_scatter ] .
the points derived from a1795 ( bautz et al.2009 ) are reported in the top panels ( as triangles ) , while the points derived from pks0745 - 191 ( george et al.2009 ) are reported in the bottom panels ( as squares).,title="fig:",scaledwidth=23.5% ] , for n=2 ( top panels ) and n=4 ( bottom panels ) .
the meaning of colors and linestyles is an in fig.[fig : prof_scatter ] .
the points derived from a1795 ( bautz et al.2009 ) are reported in the top panels ( as triangles ) , while the points derived from pks0745 - 191 ( george et al.2009 ) are reported in the bottom panels ( as squares).,title="fig:",scaledwidth=23.5% ] , for n=2 ( top panels ) and n=4 ( bottom panels ) .
the meaning of colors and linestyles is an in fig.[fig : prof_scatter ] .
the points derived from a1795 ( bautz et al.2009 ) are reported in the top panels ( as triangles ) , while the points derived from pks0745 - 191 ( george et al.2009 ) are reported in the bottom panels ( as squares).,title="fig:",scaledwidth=23.5% ] , for n=2 ( top panels ) and n=4 ( bottom panels ) .
the meaning of colors and linestyles is an in fig.[fig : prof_scatter ] .
the points derived from a1795 ( bautz et al.2009 ) are reported in the top panels ( as triangles ) , while the points derived from pks0745 - 191 ( george et al.2009 ) are reported in the bottom panels ( as squares).,title="fig:",scaledwidth=23.5% ]
we thank c.gheller and r.brunino for their priceless support at hpc - cineca ( italy ) , and g.brunetti of fruitful discussions .
we acknowledge the help of d.fabjan in the treatment of gadget simulations data .
f.v . acknowledges the usage of computational resources under the cineca - inaf 2008 - 2010 agreement and the 2009 key project `` turbulence , shocks and cosmic rays electrons in massive galaxy clusters at high resolution '' , and partial financial support from the grant for1254 from deutschen forschungsgemeinschaft .
s.e . and m.r .
acknowledge the financial contribution from contracts asi - inaf i/023/05/0 and i/088/06/0 .
m. r. acknowledges the support of grant anr-06-jcjc-0141 . k.d . acknowledges the support by the dfg priority programme 1177 and additional support by the dfg cluster of excellence `` origin and structure of the universe '' . 200 bautz , m. w. , et al .
2009 , pasj , 61 , 1117 burns , j. o. , skillman , s. w. , & oshea , b. w. 2010 , apj , 721 , 1105 cassano , r. , ettori , s. , giacintucci , s. , brunetti , g. , markevitch , m. , venturi , t. , & gitti , m. 2010 , apjl , 721 , l82 cavaliere , a. , lapi , a. , & fusco - femiano , r. 2010 , arxiv:1010.4415 dolag k. , vazza f. , brunetti g. , tormen g. , 2005 , mnras , 364 , 753 ettori , s. , fabian , a. c. , & white , d. a. 1998 , mnras , 300 , 837 ettori , s. , & balestra , i. 2009 , a&a , 496 , 343 ettori , s. , & molendi , s. 2010 , arxiv:1005.0382 frenk , c. s. , et al .
1999 , apj , 525 , 554 george , m. r. , fabian , a. c. , sanders , j. s. , young , a. j. , & russell , h. r. 2009 , mnras , 395 , 657 hoshino , a. , et al .
2010 , arxiv:1001.5133 kawahara , h. , reese , e. d. , kitayama , t. , sasaki , s. , & suto , y. 2008 , apj , 687 , 936 kawaharada , m. , et al . 2010 , apj , 714 , 423 lapi , a. , fusco - femiano , r. , & cavaliere , a. 2010 , a&a , 516 , a34 leccardi , a. , & molendi , s. 2008 , a&a , 486 , 359 liedahl , d. a. , osterheld , a. l. , & goldstein , w. h. 1995 , apjl , 438 , l115 loken , c. , norman , m. l. , nelson , e. , burns , j. , bryan , g. l. , & motl , p. 2002 , apj , 579 , 571 mitchell , n. l. , mccarthy , i. g. , bower , r. g. , theuns , t. , & crain , r. a. 2009 , mnras , 395 , 180 neumann , d. m. 2005,a&a , 439 , 465 norman m. l. , bryan g. l. , harkness r. , bordner j. , reynolds d. , oshea b. , wagner r. , 2007 , arxiv , 705 , arxiv:0705.1556 rasia , e. , mazzotta , p. , borgani , s. , moscardini , l. , dolag , k. , tormen , g. , diaferio , a. , & murante , g. 2005 , apjl , 618 , l1 romeo , a. d. , sommer - larsen , j. , portinari , l. , & antonuccio - delogu , v. 2006 , mnras , 371 , 548 roncarelli , m. , ettori , s. , dolag , k. , moscardini , l. , borgani , s. , & murante , g. 2006 , mnras , 373 , 1339 reiprich , t. h. , et al .
2009 , a&a , 501 , 899 springel , v. , & hernquist , l. 2003 , mnras , 339 , 289 springel v. , 2005 , mnras , 364 , 1105 springel v. , 2010 , mnras , 401 , 791 valdarnini , r. 2010 , arxiv:1010.3378 vazza , f. , brunetti , g. , gheller , c. , & brunino , r. 2010 , newa , 15 , 695 vazza , f. 2010 , mnras , 1724 voit , g. m. , kay , s. t. , & bryan , g. l. 2005 , mnras , 364 , 909 vikhlinin , a. , forman , w. , & jones , c. 1999 , apj , 525 , 47 vikhlinin , a. , kravtsov , a. , forman , w. , jones , c. , markevitch , m. , murray , s. s. , & van speybroeck , l. 2006 , apj , 640 , 691 wadsley , j. w. , veeravalli , g. , & couchman , h. m. p. 2008 , mnras , 387 , 427
we briefly present here additional tests to assess the reliability of the results against a few numerical effects that might play a role in the computation of the average azimuthal scatter .
first , we show the additional results for the profiles and the azimuthal scatter for the the two most massive clusters of the gadget2 runs ( g1a and g8a ) , in re - simulations where cooling , star formation and feedback from galactic winds where self - consistently followed during the run .
the star formation is followed by adopting a sub - resolution multiphase model for the interstellar medium , including also the feedback from supernovae and galactic outflows ( springel & hernquist 2003 ) . in these runs ,
the efficiency of supernovae to power galactic winds has been set to 50 per cent , which turns into a wind speed of @xmath23 .
these re - simulations have been extracted from the same `` _ cfs _ '' dataset of roncarelli et al.(2006 ) . in the top panel of fig.[fig : appendix ] we show the average gas density ( top lines ) and gas entropy ( bottom lines ) profiles for g1a and g8a at z=0 , after the `` 99 percent '' filtering procedure ( sec.[subsec : method ] ) , for the reference physical run adopted in the main article ( solid lines ) and for the _ csf _ model ( dashed ) . except for the sizable increase of the core gas density and the decrease of the core gas entropy , as an obvious effect of radiative cooling
, the agreement for all the radii outside @xmath24 is remarkable .
this is in line with the early results already reported by roncarelli et al.(2006 ) and romeo et al.(2006 ) . in the bottom panel of fig.[fig : appendix ] we show the corresponding radial trend of the azimuthal scatter as defined in eq.[eq : scatter ] and keeping @xmath25 sectors ( the bold lines refer to gas density and the thin lines refer to gas entropy ) .
for both clusters in this case the differences in the azimuthal scatter are sizable when the two physical runs are compared ( @xmath0 per cent at all radii ) ; however all profiles scatter around the value of @xmath26 at all radii from @xmath27 , with no evident trend with radius . this may be well explained by the small scale effect of cooling sub - clumps at all radii in the _ csf _ run : the `` 99 per cent filtering '' procedure outlined in the main paper efficiently removes the pointlike contribution from these clumps , however cooling make them more compact and enhances their survival within the cluster atmosphere and their efficiency in injecting shocks and small scale disturbances in the icm ( e.g. valdarnini 2010 ) , which can slightly change the local scatter behaviour , even if not the overall trend of azimuthal scatter in the clusters .
these results make us confident that the adoption of more complex physical modeling compared to the non - radiative runs investigated in the main paper can not change our conclusion in any sizable way , for all the cluster volume outside of cores . in a second test
, we checked the stability of our results against the adoption of a larger line of sight across the projected cylinder , increasing its size from @xmath28 ( as used in the main paper ) to @xmath29 .
this was motivated by the fact that the overlap of not virialized regions and large scale environment along the line of sight towards the cluster may vary the reported values of the azimuthal scatter .
we thus re - computed the azimuthal scatter of gas density , gas temperature and gas entropy for cluster g1a of the gadget2 sample , by considering a line of sight of @xmath30 ( fig.[fig : appendix2 ] ) .
the test clearly shows that the difference in the azimuthal scatter of all quantities remains essentially the same within a few percent , with all trends un - altered .
we thus conclude that our results are well constrained for a total line of sight of @xmath31 . sectors .
the thick lines refer to gas density , the thin lines refer to gas entropy ; the meaning of colors and line - styles is as in the left panel.,title="fig:",scaledwidth=23.5%,scaledwidth=23.0% ] sectors .
the thick lines refer to gas density , the thin lines refer to gas entropy ; the meaning of colors and line - styles is as in the left panel.,title="fig:",scaledwidth=23.5%,scaledwidth=23.0% ]
|
we study the azimuthal scatter in the radial profiles of x - ray luminous galaxy clusters , with two sets of high - resolution cosmological re - simulations obtained with the codes enzo and gadget2 .
the average gas profiles are computed for different angular sectors of the cluster projected volume , and compared with the mean cluster profiles at each radius from the center .
we report that in general the level of azimuthal scatter is found to be @xmath0 per cent for gas density , temperature and entropy inside @xmath1 , and @xmath2 per cent for x - ray luminosity for the same volume .
these values generally doubles going to @xmath3 from the cluster center , and are generally found to be higher ( by @xmath4 per cent ) in the case of perturbed systems .
a comparison with results from recent _ suzaku _ observations is discussed , showing the possibility to simply interpret the large azimuthal scatter of observables in light of our simulated results .
[ firstpage ] galaxy : clusters , general methods : numerical intergalactic medium large - scale structure of universe 0.4 cm
|
defined on network - like structures are encountered in a large number of scenarios including molecular interaction in biological systems , computer networks , sensor networks , social and citation networks , the internet and the world wide web , power grids , transportation networks @xcite , and many more .
such data can be visualized as a set of scalar values , known as a graph signal , lying on a particular structure , i.e. , a graph . in computer graphics ,
data defined on any geometrical shape described by polygon meshes can be formulated as a graph signal @xcite .
+ the irregular structure of the underlying graph , as opposed to the regular structure in case of time - series and image signals dealt in classical signal processing @xcite , imposes a great challenge in analysis and processing of graph signals .
fortunately , recent work toward the development of important concepts and tools , extending classical signal processing theory , including sampling and interpolation on graphs @xcite , graph - based transforms @xcite , and graph filters @xcite have enriched the field of graph signal processing .
these tools have been utilized in solving a variety of problems such as signal recovery on graphs @xcite , clustering and community detection @xcite , graph signal denoising @xcite , and semi - supervised classification on graphs@xcite .
+ transforms aimed at frequency analysis of graph signals facilitate efficient handling of the data and remain at the heart of graph signal processing . in literature , there exist two frameworks for frequency analysis and processing of graph signals : ( i ) laplacian matrix based approach , and ( ii ) weight matrix based approach .
the existing laplacian based approach @xcite is limited to the analysis of graph signals lying on undirected graphs with real non - negative weights . in the laplacian based approach , eigendecomposition of the graph laplacian @xmath1
is used for frequency analysis of graph signals .
mathematically , graph fourier transform ( gft ) of a graph signal @xmath2 is defined as @xmath3 , where @xmath4 $ ] is the matrix in which columns are the eigenvectors of @xmath1 . here
, the frequency ordering is based on the quadratic form and turns out to be `` natural '' .
natural frequency ordering means that the small eigenvalues correspond to low frequencies and vice - versa .
on the other hand , the weight matrix based framework @xcite , also referred as the discrete signal processing on graphs ( dsp@xmath5 ) framework , has been built on the graph shift operator .
the weight matrix @xmath6 of the graph acts as the shift operator and the shifted version of a graph signal @xmath2 can be found as @xmath7 .
shift operator is treated as the elementary linear shift invariant ( lsi ) graph filter , which essentially is a polynomial in @xmath6 .
based on the shift operator , authors define total variation ( tv ) on graphs that they utilize for ordering of graph frequencies . in dsp@xmath5 ,
the graph fourier transform of a graph signal @xmath2 is defined as @xmath8 , where @xmath9 is the matrix with the jordan eigenvectors of @xmath6 as its columns ( @xmath6 is decomposed in jordan canonical form as @xmath10 ) .
the jordan eigenvectors of @xmath6 are used as graph fourier basis and the eigenvalues of @xmath6 act as graph frequencies .
+ although dsp@xmath5 is applicable to directed graphs with negative or complex edge weights as well , it does not provide `` natural '' intuition of frequency .
the eigenvalue of @xmath6 with _ maximum _ absolute value acts as the _ lowest _ frequency and as one moves away from this eigenvalue in the complex frequency plane , the frequency increases . thus , frequency ordering is not natural and there is an overhead of frequency ordering as well . also , the interpretation of frequency is not intuitive for example , in general , a constant graph signal results in low as well as high frequency components in the spectral domain .
+ in this paper , we redefine graph fourier transform ( gft ) under dsp@xmath0 . in the new definition of gft ,
the jordan eigenvectors of the directed laplacian matrix are treated as the graph fourier basis and the corresponding eigenvalues constitute the graph spectrum .
the directed laplacian matrix of a graph is a simple extension of the symmetric laplacian discussed in @xcite to directed graphs . to redefine gft under dsp@xmath0
, we first propose a shift operator derived from the directed laplacian .
then , we utilize this shift operator to define total variation ( tv ) of a graph signal which is used for frequency ordering .
we observe `` natural '' frequency ordering as well as better intuition as compared to the existing gft definition under dsp@xmath5 approach .
moreover , the new definition of gft links the dsp@xmath0 framework to the existing laplacian based approach . we also show that considering our proposed shift operator , the lsi filters under dsp@xmath0 become polynomial in the directed laplacian .
+ rest of this paper is organized as follows . in section [ sec : freqanalysis ] , we discuss the directed laplacian of a graph followed by the proposed shift operator and total variation on graph .
then , we redefine the graph fourier transform based on directed laplacian .
we describe lsi graph filters in section [ sec : filtering ] and then conclude the paper in section [ sec : conclusion ] .
first , we present directed laplacian matrix of a graph and then , derive the shift operator from it .
next , we define total variation of a graph signal which is utilized in frequency ordering . finally , we redefine graph fourier transform under the dsp@xmath0 framework . a graph signal is a collection of values defined on a complex and irregular structure modeled as a graph .
a graph is represented as @xmath11 , where @xmath12 is the set of vertices ( or nodes ) and @xmath6 is the weight matrix of the graph in which an element @xmath13 represents the weight of the directed edge from node @xmath14 to node @xmath15 .
moreover , a graph signal is represented as an @xmath16-dimensional vector @xmath17 ^t \in \mathbb{c}^{n}$ ] , where @xmath18 is the value of the graph signal at node @xmath15 and @xmath19 is the total number of nodes in the graph . as discussed in @xcite ,
the graph laplacian for undirected graphs is a symmetric difference operator @xmath20 , where @xmath21 is the degree matrix of the graph and @xmath6 is the weight matrix of the graph .
however , in case of directed graphs ( or digraphs ) , the weight matrix @xmath6 of a graph is not symmetric .
in addition , the degree of a vertex can be defined in two ways @xcite in - degree and out - degree . in - degree of a node
@xmath15 is estimated as @xmath22 , whereas , out - degree of the node @xmath15 can be calculated as @xmath23 .
we consider in - degree matrix and define the directed laplacian @xmath1 of a graph as @xmath24 where @xmath25 is the in - degree matrix .
[ fig : dirgraphexample](a ) shows an example weighted directed graph and the corresponding matrices are shown in fig .
[ fig : dirgraphexample](b)-(d ) .
clearly , the laplacian for directed graph is not symmetric , nevertheless , it follows some important properties : ( i ) sum of each row is zero and hence , @xmath26 is surely an eigenvalue , and ( ii ) real parts of the eigenvalues are non - negative for a graph with positive edge - weights .
+ 0.24 0.24 @xmath6 = @xmath27 + 0.24 @xmath28 = @xmath29 0.24 @xmath30 = @xmath31 there also exist a few other definitions of graph laplacian ( normalized as well as combinatorial ) for directed graphs @xcite .
however , we choose eq .
( [ eq : outlap ] ) as the definition of graph laplacian for our analysis .
we identify the shift operator from the graph structure corresponding to a discrete time periodic signal and then extend it to arbitrary graphs . +
a finite duration periodic discrete signal can be thought of as a graph signal lying on a directed cyclic graph shown in fig .
[ fig : dircyclic ] . indeed , fig .
[ fig : dircyclic ] is the support of a periodic time - series having a period of five samples .
the directed laplacian of the graph is @xmath32 consider a discrete time finite duration periodic signal @xmath33^t$ ] defined on the graph shown in fig .
[ fig : dircyclic ] . shifting the signal @xmath34 by one unit to the right results in the signal @xmath35^t$ ] .
this shifted version of the signal can also be found as @xmath36 where @xmath37 is treated as the shift operator ( matrix ) .
+ we extend the notion of shift to arbitrary graphs and consequently use @xmath37 as the shift operator for graph signals .
hence , the shifted version of a graph signal @xmath2 can be calculated as @xmath38 in contrast to the use of the weight matrix as the shift operator , we opt for the pre - mentioned shift operator which involves the directed laplacian .
our selection of this shift matrix results in a better and simpler frequency analysis which will become evident in the subsequent sections .
total variation ( tv ) of a graph signal is a measure of total amplitude oscillations in the signal values with respect to the graph .
as discussed in @xcite , tv of a graph signal @xmath2 with respect to the graph @xmath39 can be given as @xmath40 where @xmath41 is the derivative of the graph signal @xmath2 at vertex @xmath15 and is defined as the difference between the values of the original graph signal @xmath2 and its shifted version at vertex @xmath15 : @xmath42 from eq .
( [ eq : shift ] ) , ( [ eq : tv1 ] ) , and ( [ eq : graphgrad ] ) , we have @xmath43 observe that the quantity @xmath44 at node @xmath15 is the sum of the weighted differences ( weighted by corresponding edge weights ) between value of @xmath2 at node @xmath15 and values at the neighboring nodes . in other words , @xmath45 provides a measure of variations in the signal values as we move from node @xmath15 to its adjacent nodes . thus , @xmath46-norm of the quantity @xmath44 can be interpreted as the absolute sum of the _ local variations _ in @xmath2
. we will utilize tv given by eq .
( [ eq : graphtv ] ) to estimate variations of the graph fourier basis and subsequently to identify low and high frequencies of a graph .
dsp@xmath0 derives analogy from the perspective of lsi filtering . in dsp@xmath0 ,
lsi graph filters are polynomials in the graph weight matrix @xmath6 and as a result the eigenvectors of @xmath6 become the eigenfunctions of lsi graph filters .
analogous to the fact that the complex exponentials are invariant to lsi filtering in classical signal processing , the eigenvectors of @xmath6 are utilized as the graph fourier basis .
+ we redefine the graph fourier transform under dsp@xmath0 .
we consider the jordan eigenvectors of the graph laplacian matrix as the graph fourier basis that are invariant to an lsi graph filter ( discussed in section [ sec : filtering ] ) , which is a polynomial in @xmath1 under proposed shift operator .
using jordan decomposition , the graph laplacian is decomposed as @xmath47 where @xmath48 , known as the jordan matrix , is a block diagonal matrix similar to @xmath1 and the jordan eigenvectors of @xmath1 constitute the columns of @xmath9 .
we define gft of a graph signal @xmath2 as @xmath49 here , @xmath9 is treated as the graph fourier matrix whose columns constitute the graph fourier basis .
inverse graph fourier transform can be calculated as @xmath50 in this definition of gft , the eigenvalues of the graph laplacian act as the graph frequencies and the corresponding jordan eigenvectors act as the graph harmonics .
the eigenvalues with small absolute value correspond to low frequencies and vice - versa ; we will prove that shortly .
thus , the frequency order turns out to be _
natural_. + before discussing ordering of frequency , we consider a special case when the laplacian matrix is diagonalizable . when the graph laplacian is diagonalizable , eq .
( [ eq : lap_jordan ] ) is reduced to : @xmath51 here , @xmath52 is a diagonal matrix containing the eigenvalues @xmath53 of @xmath1 and @xmath54 \in \mathbb{c}^{n\times n}$ ] is the matrix with columns as the corresponding eigenvectors of @xmath1 .
note that for a graph with real non - negative edge weights , the graph spectrum will lie in the right half of the complex frequency plane ( including the imaginary axis ) .
for an undirected graph with real weights , the graph laplacian matrix @xmath1 is real and symmetric . as a result
, the eigenvalues of @xmath1 turn out to be real and @xmath1 constitutes orthonormal set of eigenvectors . hence
, the jordan form of the laplacian matrix for undirected graphs can be written as @xmath55 where @xmath56 , because the eigenvectors of @xmath1 are orthogonal in undirected case .
consequently , gft of a signal @xmath2 can be given as @xmath57 and the inverse can be calculated as @xmath58 . note that the graph spectrum will lie on the real axis of the complex frequency plane , given the weight matrix is real .
moreover , if the weights are non - negative as well , the graph spectrum will lie on the non - negative half of the real axis .
this coincides with the gft presented in @xcite , where only undirected graphs with real and non - negative weights were considered .
thus , the new definition of gft unifies dsp@xmath0 and laplacian based approach .
we utilize the definition of tv given by eq .
( [ eq : graphtv ] ) to quantify oscillations in the graph harmonics , and subsequently to order the frequencies .
the eigenvalues for which the corresponding proper eigenvectors have small variations are labeled as low frequencies and vice - versa .
the frequency ordering is established by theorem [ th : frequency_ordering ] .
[ th : frequency_ordering ] let @xmath59 and @xmath60 be the eigenvectors with unit @xmath46-norm corresponding to two distinct complex eigenvalues @xmath61 of the laplacian matrix @xmath1 of the graph @xmath39 . if @xmath62 , then the tvs of these eigenvectors with respect to the graph @xmath39 satisfy @xmath63 let @xmath64 be the proper eigenvector corresponding to eigenvalue @xmath65 of the laplacian matrix @xmath1 of the graph @xmath39 , then @xmath66 .
now , using eq .
( [ eq : graphtv ] ) , tv of @xmath64 with respect to the graph @xmath39 is calculated as @xmath67 if we scale all eigenvectors to have the same @xmath46-norm , then from the above analysis we note that @xmath68 0.26 0.18 + 0.26 0.18 note that tv of a proper eigenvector is directly proportional to the absolute value of the corresponding eigenvalue .
therefore , all the proper eigenvectors corresponding to the eigenvalues having equal absolute value will have the same tv . as a result
, distinct eigenvalues may sometimes yield exactly same tv . because of this reason ,
sometimes frequency ordering is not unique .
however , if all the eigenvalues are real , the frequency ordering is guaranteed to be unique .
[ fig : freqorder ] shows the visualization of frequency ordering in the complex frequency plane ; @xmath69 corresponds to zero frequency and as we move away from the origin , the eigenvalues correspond to higher frequencies .
+ an important point worth mentioning here is that we achieve same order of frequency as given by theorem [ th : frequency_ordering ] even if we use quadratic form ( 2-dirichlet form ) in place of tv .
2-dirichlet form @xcite of a signal @xmath2 is defined as @xmath70 substituting @xmath41 from eq .
( [ eq : graphgrad ] ) , we have @xmath71 now , quadratic form of a proper eigenvector @xmath64 corresponding to frequency @xmath65 can be calculated as @xmath72 . therefore , if all the eigenvectors are scaled to have same @xmath73-norm , then @xmath74 therefore , the frequency ordering based on the quadratic form will be same as given by theorem [ th : frequency_ordering ] .
let us consider a directed graph shown in fig .
[ fig : dirgraphexample](a ) . performing jordan decomposition of the laplacian matrix
, we find the fourier matrix @xmath9 and jordan matrix @xmath48 as follows .
@xmath75 @xmath76 eigenvalue @xmath77 corresponds to the highest frequency of the graph .
also note that the tvs of the eigenvectors corresponding to the frequencies @xmath78 and @xmath79 are equal because both of the frequencies have same absolute value .
the magnitudes of the gft coefficients of the graph signal @xmath80^t$ ] are plotted in fig .
[ fig : dirgraphspectrum ] .
^t$ ] defined on the graph shown in fig . [ fig : dirgraphexample](a ) . ] the jordan eigenvector of @xmath1 corresponding to the zero eigenvalue is given as @xmath81^t$ ] .
therefore , for a constant graph signal , gft will have only a single non - zero coefficient at zero frequency ( eigenvalue ) .
for example , consider a constant graph signal @xmath82^t$ ] residing on the graph shown in fig .
[ fig : dirgraphexample](a ) .
gft of the signal is given by @xmath83^t$ ] , magnitudes of which are plotted in fig .
[ fig : dirgraphspectrumconst ] . ^t$ ] defined on the graph shown in fig . [ fig : dirgraphexample](a ) . ]
we can observe the presence of only zero frequency component in the spectrum of a constant graph signal .
this is `` an evidence '' to the intuition that the variation in a constant graph signal is zero as we travel from a node to the other node connected by a directed edge .
in contrast , the graph fourier transform defined in dsp@xmath0 fails to give this basic intuition .
an lsi graph filter is a linear filter for which shifted version of the filter output is same as the filter output to the shifted input .
that is , if @xmath84 , where @xmath85 is the output of the filter @xmath86 corresponding to input @xmath87 , then the filter @xmath86 is shift - invariant ( si ) . the condition for a graph filter to be lsi is given by theorem [ th : filterlsi ] .
[ th : filterlsi ] a graph filter @xmath86 is lsi if the following conditions are satisfied : 1 .
geometric multiplicity of each distinct eigenvalue of the graph laplacian is one .
the graph filter @xmath86 is a polynomial in @xmath1 , i.e. , @xmath86 can be written as @xmath88 where @xmath89 are called filter taps . from eq .
( [ eq : shift ] ) , the shift operator is @xmath90 .
then , for a filter @xmath86 to be lsi , the following must be satisfied : @xmath91 in other words , the filter is lsi if the matrices @xmath1 and @xmath86 commute .
this is true when @xmath86 is a polynomial in @xmath1 , given that the characteristic polynomial and the minimal polynomial of @xmath1 are equal or the geometric multiplicity of every distinct eigenvalue is one @xcite .
considering a @xmath92-tap graph filter and substituting @xmath93 and @xmath94 in eq .
( [ eq : lsi_filter ] ) , we get @xmath95 , which is the shift operator discussed in section [ subsec : shiftop ] .
thus , the shift operator is a first order lsi filter
. also , the graph fourier basis , which are the jordan eigenvectors of @xmath1 , are the eigenfunctions of an lsi filter described by eq .
( [ eq : lsi_filter ] ) .
in this paper , under the dsp@xmath0 framework , we considered the jordan eigenvectors of the directed laplacian as graph harmonics and the corresponding eigenvalues as the graph frequencies , and utilized that to redefine gft . for this purpose
, we proposed a shift operator derived from the directed laplacian of a graph .
our proposed definition of gft under dsp@xmath0 framework , provides natural frequency interpretation and links dsp@xmath0 to the laplacian based approach .
we also showed that by considering our proposed shift operator , lsi filters become polynomial in the directed laplacian .
s. k. narang , a. gadde , and a. ortega , `` signal processing techniques for interpolation in graph structured data , '' _ in proc .
ieee international conference on acoustics , speech signal processing _ ,
vancouver , may 2013 , pp .
5445 - 5449 .
a. anis , a. gadde , and a. ortega , `` towards a sampling theorem for signals on arbitrary graphs , '' _ in proc .
ieee international conference on acoustics , speech signal processing _
, may 2014 , pp .
3864 - 3868 .
d. i. shuman , s. k. narang , p. frossard , a. ortega , and p. vandergheynst , `` the emerging field of signal processing on graphs : extending high - dimensional data analysis to networks and other irregular domains , '' _ ieee signal processing magazine _ , vol .
83 - 98 , may 2013 .
a. sandryhaila and j. m. f. moura , `` big data analysis with signal processing on graphs : representation and processing of massive data sets with irregular structure , '' _ ieee signal processing magazine _ , vol .
80 - 90 , september 2014 .
s. segarra , a. g. marques , g. leus , and a. ribeiro , `` reconstruction of graph signals : percolation from a single seeding node , '' _ in global conf . on signal and info .
process.(globalsip ) _ , december 2015 , orlando , florida , u.s.a . x. dong , p. frossard , p. vandergheynst , and n. nefedov , `` clustering on multi - layer graphs via subspace analysis on grassmann manifolds , '' _ ieee transactions on signal processing _ , vol .
905 - 918 , february 2014 .
v. n. ekambaram , b. ayazifar g. fanti , and k. ramchandran , `` wavelet - regularized graph semi - supervised learning , '' _ in proc .
ieee glob .
signal information process .
_ , austin , tx , dec . 2013 ,
423 - 426 .
|
in this paper , we redefine the graph fourier transform ( gft ) under the dsp@xmath0 framework .
we consider the jordan eigenvectors of the directed laplacian as graph harmonics and the corresponding eigenvalues as the graph frequencies .
for this purpose , we propose a shift operator based on the directed laplacian of a graph . based on our shift operator ,
we then define total variation of graph signals , which is used in frequency ordering .
we achieve natural frequency ordering and interpretation via the proposed definition of gft .
moreover , we show that our proposed shift operator makes the lsi filters under dsp@xmath0 to become polynomial in the directed laplacian .
graph signal processing , graph fourier transform , directed laplacian .
|
with the rapid progress made in upgrading and testing a series of advanced interferometric gravitational - wave detectors such as ligo @xcite , virgo @xcite , and kagra @xcite , there are now great expectations that in the next five years we will witness the first direct detection of gravitational waves .
prime sources for such a detection are binary systems of compact objects , namely , binary systems comprising either two black holes , a black hole and a neutron star , or two neutron stars . the latter configuration , in particular , is potentially a very interesting one , as it will represent the most common source , with a realistic expected detection rate of @xmath0 @xcite .
a detection of gravitational waves from binary neutron stars would yield a wealth of information about the chirp mass , the orientation , and the localization of the binary but also possibly the mass , spin and radius of the individual stars @xcite . in turn , this information could set constraints on the equation of state ( eos ) of the matter in their interior . indeed , a number of recent investigations have revealed that it is possible to set serious constraints on the properties of the neutron - star structure and eos , either when using the inspiral signal only @xcite , or when exploiting the rich spectral features of the postmerger signal @xcite . at the same time
, the merger of a binary system containing at least one neutron star represents arguably the most attractive scenario to explain the complex phenomenology associated with short gamma - ray bursts ( sgrbs ) , although many alternatives exist [ see ref .
@xcite for a recent review ] . while such a scenario was suggested already 30 years ago @xcite , numerical simulations ( see , e.g. , refs .
@xcite ) and new observations @xcite have put this scenario on firmer grounds . in particular
, what these simulations have shown is that the merger of a binary system of neutron stars inevitably leads to the formation of a metastable object , which we dub the binary - merger product or bmp .
depending on the total mass and mass ratio of the binary , the bmp can either be a supramassive neutron star ( smns ) , that is , a star with mass above the maximum mass for nonrotating configurations @xmath1 but below the maximum mass for uniformly rotating configurations @xmath2 , with @xmath3 @xcite ; a hypermassive neutron star ( hmns ) , that is , a star more massive than a smns ; or a black hole .
since the mass distribution in neutron - star binaries is peaked around @xmath4 @xcite , this implies that @xmath5 .
although this can not be excluded , there is also no observational evidence that such massive neutron stars exist . ] .
the general - relativistic hydrodynamical modelling of binary neutron stars has seen very considerable progress over the last decade ( see , e.g. , refs .
@xcite ) , and it has now reached a rather mature state .
in fact , it is presently possible to calculate inspiral waveforms having a phase accuracy comparable to that of binary black - hole simulations thanks to the use of high - order methods with high - order convergence rates @xcite or of very high - resolution and long inspirals @xcite . at the same time
, the space of parameters is also being carefully investigated , both in terms of the variety of the eoss considered @xcite , and of the treatment of radiative losses via neutrino cooling @xcite .
when magnetic fields are present , on the other hand , the bulk of work carried out so far is considerably more limited .
investigations of the impact that magnetic fields have on the dynamics of the binary have in fact started with the first short - inspiral works in refs .
@xcite , which were complemented by the longer simulations in ref .
@xcite . the latter work , together with ref .
@xcite , also investigated the possibility that magnetic fields could have an impact on the gravitational - wave signals emitted by these systems during the inspiral .
the conclusions reached were that , for realistic strengths @xmath6 , the presence of magnetic fields could not be revealed by detectors such as advanced ligo or advanced virgo .
the astrophysical implications of the merger of a magnetized neutron - star binary were explored in ref .
@xcite , where it was shown that instabilities in the torus orbiting the black hole amplify the magnetic field by 3 orders of magnitude and generate a magnetic - jet structure characterized by an ordered poloidal magnetic field along the black - hole spin axis . the broad consistency with the observations in terms of black - hole spin , torus mass and accretion rate , and magnetic - field topology offered the first evidence that the merger of magnetized neutron stars can provide the basic conditions for the central engine of sgrbs .
considerable effort has also been dedicated to investigating the properties of the hmns under more controlled conditions .
for instance , using ultrahigh spatial resolutions but axisymmetric initial data , ref .
@xcite has provided the first evidence from three - dimensional global simulations that a magnetorotational instability ( mri ) @xcite is likely to develop during the lifetime of the hmns ( see also refs .
@xcite for earlier work in two dimensions ) .
in addition , again using axisymmetric initial data and different magnetic field configurations , it has been shown that a magnetically driven wind can be launched from the outer layers of the hmns as a result of its differential rotation @xcite .
these works have also highlighted that for realistic magnetic field topologies the wind is baryon loaded and quasi - isotropic , with bulk velocities of @xmath7 @xcite .
more recently , instead , the use of subgrid modeling as an effective way to describe the turbulent dynamics that develops in the shear layer between the two neutron stars at merger , has suggested that amplifications of up to 5 orders of magnitude are possible @xcite , although these amplifications are not produced in direct simulations @xcite , even at very high resolution @xcite .
finally , progress has taken place also on the derivation of improved numerical techniques , such as those in ref .
@xcite , where the significant advantages of a vector - potential approach and of a lorentz gauge were discussed .
all of the works mentioned above have been carried out within the ideal - magnetohydrodynamic ( imhd ) approximation , in which the electrical conductivity is assumed to be infinite . under these conditions ,
the magnetic flux is conserved and the magnetic field is frozen into the fluid , being simply advected with it .
this is a very good approximation for the stellar interior before the merger because it neglects any effect of resistivity on the dynamics of the plasma .
after the merger , however , there will be spatial regions with very hot plasma where the electrical conductivity is finite and the resistive effects , most notably , magnetic reconnection , will take place . an obvious improvement over the imhd description is the use of the general - relativistic resistive - magnetohydrodynamic ( rmhd ) equations , which provide a complete magnetohydrodynamic ( mhd ) description of regions with a high conductivity , such as the stellar interiors , and of regions with small conductivity , such as the electrovacuum exterior .
furthermore , when the conductivity is set to zero , it yields the maxwell equations in vacuum , thus allowing for the study of the magnetic - field evolution also well outside the stellar magnetosphere @xcite . partly because of the increased complexity of the equations and partly because of the additional difficulties posed by their numerical solution ( the equations easily become stiff in regions of high conductivity ) , rmhd simulations have started only rather recently .
most of the work so far has focussed on problems in flat spacetimes @xcite , but general - relativistic investigations have also been carried out on fixed spacetimes @xcite .
indeed , together with the work carried out in refs .
@xcite , those reported here are , to the best of our knowledge , the only rmhd simulations of the dynamics of binary neutron stars in general relativity .
more specifically , we have followed the inspiral , merger , and collapse to a black hole of a neutron - star binary in which the stars have the same gravitational mass of @xmath8 and are modelled with a simple ideal - fluid eos . complementing the work reported in refs .
@xcite , which concentrated on the electromagnetic emission during the inspiral and at the merger , the focus of the simulations reported here is that of assessing the impact that resistive effects have on the dynamics of the binary after the merger and through to collapse to a black hole . to this scope
we have carried out a close comparison with an equivalent simulation performed for the same binary within the traditional imhd approximation . in this way it has been possible to determine both the similarities between the two regimes and the novel features .
the most important of such features is the evidence the survival time of the hmns before collapse to a black hole increases in a rmhd simulation .
this difference is associated to a less efficient magnetic braking in the resistive regime , in which matter is no longer perfectly advected with the flow , but can move across magnetic - field lines . in turn
, this reduces the transport of angular momentum away from the central regions of the hmns , increasing its lifetime .
interestingly , a longer - lived magnetized hmns is of help in those models of sgrbs which invoke the existence of a magnetarlike object produced after the merger @xcite .
another important result of the simulations reported here , which have been performed at higher resolutions and with a different grid structure than those in the previous work of ref .
@xcite , is the confirmation that a magnetic - jet structure is formed in the low - density funnel produced by the black hole
torus system .
both in rmhd and in imhd , the magnetic field is predominantly toroidal in the highly conducting torus and predominantly poloidal in the nearly evacuated funnel .
furthermore , because of the effective decoupling between the matter and the electromagnetic fields achieved in the rmhd simulations , the magnetic - jet structure is coherent on the largest scale of our system .
however , as in the imhd case , also in these rmhd simulations , the magnetic - jet structure does not lead to an ultrarelativistic outflow .
reconnection processes or neutrino annihilation occurring in the funnel could potentially increase the internal energy in the funnel and launch a relativistic outflow ; none of these effects is modeled in our simulations .
the plan of the paper is as follows . in sec .
[ sec : binaries : approach ] we briefly review the mathematical setup of our simulations , concentrating mostly on the discussion of the general - relativistic rmhd equations used and on the expression of the generalized ohm s law we have employed . section [ sec : numsetup ] , on the other hand , is dedicated to illustrate the numerical strategy employed in the solution of the combined set of the einstein and rmhd equations , including the properties of our computational grid , of our matching to the low - density exterior , and of our initial data .
the core of the paper is represented by sec .
[ sec : binaries : results ] , where we present our results .
after a brief overview , we discuss the magnetic - field topology and magnetic - jet structure produced in the simulations , as well as the comparison with the imhd case .
such a comparison goes over a number of aspects , from the angular - momentum transfer and lifetime of the hmns , over to the black hole torus system , and the electromagnetic luminosities .
finally , sec . [
sec : binaries : conclusions ] contains a conclusive summary of our results and the prospects for future research .
although this is not the focus of this work , an illustration of the magnetic - jet structure obtained in the imhd simulations is presented in appendix [ sec : appendix : imhd ] for completeness .
we use a spacelike signature @xmath9 and a system of units in which @xmath10 unless stated differently .
much of the numerical setup used in these simulations has been presented in greater detail in other papers @xcite , and for compactness we will review here only the basic aspects , referring the interested reader to the papers above for additional information .
however , given its importance here , we will dedicate some space to a review of our fully general - relativistic rmhd framework , which was first presented in ref .
@xcite and represents the extension of the special - relativistic rmhd formalism discussed in ref .
a similar but independent extension has been presented recently in ref .
@xcite , which describes the first 3 + 1 general - relativistic rmhd implementation in fixed spacetimes .
we start by presenting the augmented maxwell equations @xmath11 where @xmath12 is the 4-metric , @xmath13 is the faraday tensor , @xmath14 is the maxwell tensor , @xmath15 is the electric four - current density , and @xmath16 are two auxiliary scalar variables added to the maxwell equations to control the constraints for the magnetic and electric parts , respectively ( see below ) . after a standard 3 + 1 splitting of spacetime
, the maxwell and faraday tensors can be decomposed in terms of the electric ( @xmath17 ) and magnetic ( @xmath18 ) fields measured by an observer moving along the normal direction @xmath19 ( i.e. , normal or eulerian observer ) as @xmath20 with @xmath21 , @xmath22 the determinant of the 4-metric and @xmath23 the levi - civit symbol .
the same can be done for the electric four - current @xmath24 where @xmath25 and @xmath26 are the charge density and the electric current density for an eulerian observer , respectively . using these definitions and performing a 3 + 1 decomposition of eqs . and with respect to the normal vector @xmath27
, we arrive at the following evolution equations , @xmath28 where @xmath29 is the spatial 3-metric , @xmath30 is the trace of the extrinsic curvature @xmath31 , @xmath32 is the lapse and @xmath33 is the shift 4-vector .
we recall that @xmath34 denotes the lie derivative along the shift vector and that @xmath35 is related to the four - dimensional levi - civit tensor via @xmath36 or alternatively @xmath37 , where @xmath38 now is the determinant of the 3-metric .
the scalar fields @xmath39 measure the deviation from the constrained solution , with @xmath40 driving the solution of eq .
toward the zero - divergence condition @xmath41 , and @xmath42 driving the solution of eq . toward the condition
@xmath43 . this driving is exponentially fast and over a time scale @xmath44 .
this approach , named hyperbolic divergence cleaning in the context of imhd , was proposed in ref .
@xcite as a simple way of solving the maxwell equations and enforcing the conservation of the divergence - free condition for the magnetic field .
this method has been extended to the resistive relativistic case in refs .
. an obvious consequence of the maxwell equations in rmhd is the conservation law associated with the electric charge @xmath45 which provides an evolution equation for the charge density @xmath46 combining the mhd and maxwell equations we obtain the following set of evolution equations , which we write in a flux - conservative form as @xmath47 = 0\ , , \end{split } \label{charge_evol } \\ \begin{split } \partial_t ( \sqrt{\gamma } d ) + \partial_k [ \sqrt{\gamma } ( -\beta^k d + \alpha v^k d ) ] = 0\ , , \end{split } \label{rest_mass_evol } \\ \begin{split } \partial_t ( \sqrt{\gamma } \tau ) + \partial_k \{\sqrt{\gamma } [ -\beta^k \tau + \alpha ( s^k - v^k d)]\ } = \sqrt{\gamma } ( \alpha s^{lm } k_{lm } - s^k \partial_k \alpha)\ , , \end{split } \label{energy_density_evol } \\ \begin{split } \partial_t ( \sqrt{\gamma } s_i ) + \partial_k [ \sqrt{\gamma } ( -\beta^k s_i + \alpha s^k_{\ i } ) ] = \sqrt{\gamma } \left[\frac{\alpha}{2 } s^{lm } \partial_i \gamma_{lm}+ s_k \partial_i \beta^k - ( \tau + d ) \partial_i \alpha \right]\ , . \label{momentum_evol } \end{split}\end{aligned}\ ] ] [ eq : rmhdconservative ] the fluid variables @xmath48 , @xmath49 , and @xmath50 are the conserved rest - mass density , the conserved energy density , the conserved momentum , and the fully spatial projection of the energy - momentum tensor , respectively .
their explicit definitions are therefore @xmath51 here , @xmath52 is the lorentz factor , where @xmath53 are the components of the fluid 4-velocity and @xmath54 are the components of the 3-velocity as measured by the eulerian observer .
furthermore , @xmath55 is the enthalpy , with @xmath56 the total energy density , @xmath57 the pressure , @xmath58 the specific internal energy , and @xmath59 the rest - mass density @xcite .
the important difference between the rmhd and imhd equations is that they involve stiff relaxation terms that pose serious numerical limitations on the time evolution of the equations .
for this reason , a distinct class of implicit - explicit evolution methods has been developed , the rkimex schemes @xcite , which we have presented in detail in ref .
@xcite . to close the system of equations presented above ,
a relation for the electric current density in terms of the other fields is necessary , just like ohm s law provides a prescription for the spatial conduction current to be proportional to the electric field .
a generalized ohm s law provides the necessary coupling of the electric current density to the electromagnetic and matter fields .
previous work toward relativistic versions of the generalized ohm s law includes the investigations in refs .
@xcite . in one of the simpler cases the spatial conduction current can be considered as proportional to the electric field measured by the comoving observer .
therefore the electric four - current density can be written as the superposition of an advective and a conductive current @xcite , which takes the form of the generalized ohm s law @xmath60 here , @xmath61/w$ ] is the electric charge density measured in the rest frame comoving with the fluid , and should be contrasted with @xmath62 , which is instead the charge density measured by the eulerian observer .
similarly , @xmath63 and @xmath64 are , respectively , the electric field and the electrical conductivity of the medium ( which is a rank-2 symmetric tensor ) as measured in the same frame . in collisional plasmas the current in the comoving frame can be considered to be carried by the mobile electrons , with charge @xmath65 , and the conductivity tensor becomes @xmath66 with @xmath67 being the magnetic field in the comoving frame , @xmath68 , @xmath69 , and @xmath70 the electron collision time scale . expressing the four - current density of eq . in terms of the fields measured by the eulerian observer we arrive at @xmath71 + \nonumber \\ \phantom{= } & w\sigma\xi^2(e^\alpha b_\alpha)[b^\mu-\epsilon^{\mu\nu\alpha}v_\nu e_\alpha-(v_\nu b^\nu)v^\mu)]\ , \label{eq : ohm}\end{aligned}\ ] ] in deriving eq .
we have made the implicit assumption that the collision frequency between particles is much larger than the typical oscillation frequency of the plasma , which , we recall , is defined as @xmath72 .
this implies that electrons and ions can reach equilibrium on very short time scales and any correction due to the mass difference between electrons and protons can be neglected . as a result , there is no global charge separation and the plasma is neutral .
note also that the first term in eq . accounts for an isotropic scalar law for the current , while the rest represent anisotropies due to the presence of a magnetic field in the comoving frame . ideally , it would be desirable to have a well - defined prescription of the conductivity tensor as a function of the fluid properties , @xmath73 , which stems from the microphysical properties of the plasma ( see ref .
@xcite for a recent discussion ) .
in practice , however , we are far from having such a prescription in the extreme physical conditions characterizing merging neutron stars . however ,
if the collision time scale is much smaller than the electron cyclotron period , where @xmath74 is the cyclotron frequency and represents the frequency at which electrons gyrate perpendicular to the magnetic - field lines . ] or , equivalently , if @xmath75 , electrons do not have sufficient time to gyrate perpendicular to the magnetic field lines . under these conditions ,
the isotropic part of the conductivity is the dominant one ( electrons essentially slide along the magnetic - field lines ) , and expression can be approximated as @xmath76 . as a result , the spatial three - current density in the eulerian frame coming from the generalized ohm s law can be simplified so that @xmath77\,,\end{aligned}\ ] ] where we used the fact that the four - current density can be also written as @xmath78 . in our simulations ,
the conductivity @xmath79 is chosen to be either a constant or a function of the rest - mass density and a discussion will be presented in sec . [
sec : rmhdatmo ] .
the last term in eq . represents the hall effects , which , however , we set to zero for simplicity .
the evolution of the spacetime ( i.e. , of the 3-metric , extrinsic curvature , and conformal factor ) is obtained using the ` mclachlan ` code , which implements the bssnok formulation of the einstein equations @xcite employing three - dimensional finite - differencing operators for calculating the fluxes on the right - hand side of the einstein equations @xcite .
the time integration is carried out using the third - order accurate strong stability preserving implicit - explicit runge kutta scheme outlined in ref .
the time step on each grid is limited by the courant friedrichs lewy ( cfl ) condition @xcite and hence the courant coefficient is set to be @xmath80 on all refinement levels .
a kreiss oliger type dissipation is added to the spacetime evolution equations to ensure that any high - frequency noise produced during the evolution ( mostly at the refinement levels boundaries ) is damped .
the damping parameter in the evolution equation for the shift was carefully chosen to be @xmath81 so that the additional term does not become stiff on the coarser grids .
the general - relativistic rmhd ( gr - rmhd ) equations are solved with the ` whiskyrmhd ` code @xcite , which uses high - resolution shock - capturing schemes @xcite . in particular , the reconstruction of the conserved variables is achieved via the piecewise parabolic method @xcite , while the fluxes are calculated through the approximate riemann solver introduced by harten lax van leer
einfeldt @xcite and which requires only knowledge about the maximum characteristic speeds of the system , i.e. , the speed of light in this case .
the system of rmhd equations is closed by describing the fluid as ideal and with a @xmath82-law eos @xmath83 where @xmath84 and @xmath85 .
this eos is clearly an idealization .
much more sophisticated eoss have been implemented in our numerical infrastructure @xcite and have been used by other groups in general - relativistic magnetohydrodynamic ( grmhd ) simulations @xcite . however , this idealization is probably adequate at this stage as here we are mostly interested in assessing the differences introduced in the dynamics of the bmp by resistive effects .
these effects suffer from even larger uncertainties than those associated to the different eoss .
our code makes use of the ` cactus ` @xcite computational framework , which allows us to employ a box - in - box vertex - centered adaptive mesh refinement grid hierarchy that tracks the `` center of mass '' of the stars as they orbit each other .
this was achieved via the ` bnstracker ` thorn implemented by w. kastaun and the ` carpet ` driver @xcite .
the numerical domain consists of six levels of refinement with the resolution doubling between adjacent refinements .
in addition , we employ moving refinement boxes in order to track the high - density regions . the outer boundary of the computational domain is located at @xmath86 .
the finest resolution during the inspiral and merger is @xmath87 , but an extra refinement level is activated right after the merger with a resolution of @xmath88 .
it is important to remark that , although in a small region of spatial extent @xmath89 , this resolution is considerably higher than the one used in ref .
@xcite ( i.e. , @xmath90 with a spatial extent of @xmath91 ) , where the first evidence was given that the merger of a binary system of magnetized neutron stars can lead to the formation of a magnetic - jet structure . hence , although in a resistive framework , the simulations reported here can be considered as a `` higher - resolution '' counterpart of those presented in ref .
we also note that when considering initial data consisting of a binary in quasicircular orbits ( see sec .
[ sec : bns : lifetime ] ) we do not activate the extra refinement level after the merger , hence keeping a resolution of @xmath92 to describe the hmns .
we do this so as to have two different resolutions to describe the hmns and hence to study the numerical consistency of the solution . to reduce the computational cost associated with the numerical evolution of an equal - mass binary , we use a reflection - symmetry condition across the @xmath93 plane and a @xmath94-symmetry condition across the @xmath95 plane .
these conditions respect the symmetries of the scenario we are investigating .
the outer boundary conditions are set by using simple zeroth - order extrapolation of the hydrodynamic variables .
for the electromagnetic and spacetime variables , we employ simple sommerfeld radiative boundary conditions because of the nature of the fields .
given the long time scale over which our simulations are carried out , reflections due to imperfections of the boundary conditions experience several domain crossings before the end of the simulation .
we plan to improve on this in the future by the application of maximally dissipative boundary conditions that minimize the effects of reflections .
our goal with the use of a rmhd framework is that of modelling the exterior of the neutron stars as an electrovacuum , where both the conductivity and the charge density are negligibly small , so that electromagnetic fields should obey in these regions the maxwell equations in vacuum . on the other hand , we want to model the interior of the stars as highly conducting , so that our equations recover the imhd limit in such regions
there are several different ways to achieve this ; see , for example , refs . @xcite
each of them has , in our view , its advantages and disadvantages . however , because they all try to model the difficult transition region between two regimes that are intrinsically different , they all represent an approximation .
this difficulty is not specific to this problem but is a typical feature of physical problems as , for example , in the transition from an optically thick to an optically thin regime in radiative - transfer calculations .
our approach is also an approximation and is similar to the one in refs .
@xcite in the sense that the matching from the stellar interior to the stellar exterior is achieved through a carefully chosen conductivity profile .
more specifically , the conductivity profile adopted is directly related to the conserved rest - mass density ( and hence to the rest - mass density ) and given by @xmath96},\,0\right]\ , , \label{eq : bns : conductivityprofile}\end{aligned}\ ] ] where @xmath97 corresponds to a uniform scalar conductivity .
the parameters @xmath98 , and @xmath99 determine how sharp the transition to the exterior is .
for the simulations reported here , we have chosen represents the ohmic diffusion time scale , with @xmath100 the typical length scale of the field . for @xmath101 , @xmath102 , which is obviously much larger than the timescale over which our simulations are carried out , i.e. , @xmath103 . ] , @xmath104 , @xmath105 and @xmath106 , where @xmath107 and @xmath108 are the values of the conserved and primitive rest - mass density in the `` atmosphere . '' we recall that , as in other eulerian hydrodynamics and mhd codes , also our ` whiskyrmhd ` code makes use of a very low rest - mass density fluid to handle the evolution of the mhd equations in regions which are associated to the exterior of the stars .
in such a region , we follow the same prescription initially implemented in ref .
@xcite and then adopted in essentially all of the simulations performed with our code , even in its imhd incarnation @xcite .
in essence , we treat as atmosphere any region in the computational domain which is below @xmath108 , which we take here to be @xmath109 , that is approximately 8 orders of magnitude smaller than the maximum rest - mass density . , where @xmath110 @xcite .
] in such a region , we set the fluid 3-velocity to zero , the rest - mass density to a floor value , and the specific internal energy to the value it assumes for a fluid following a polytropic eos @xcite .
in addition , and differently from the imhd implementation of ref .
@xcite , the use of the prescription automatically sets the conductivity in the atmosphere to zero , so that the maxwell equations reduce there to the maxwell equations in vacuum .
hence , in our prescription the atmosphere is de facto a cold , static , uniform fluid in which electromagnetic waves propagate as if in vacuum .
a few remarks should be made about modelling the neutron stars exterior , which in reality is expected to be a highly conducting , low - density , possibly magnetically dominated plasma in very strong magnetic fields .
first , with our prescription for the atmosphere and with the conductivity , the latter is zero also in regions that are not at the atmosphere level but close to it , i.e. , at @xmath111 .
we do this because the strong winds that are produced at the merger and later on rapidly fill the computational domain .
a direct consequence of this is that the atmosphere is present only close to the boundaries and therefore an imhd prescription would be met essentially everywhere .
yet , we are interested in deviations from perfect - flux freezing and in particular in regions where plasma is tenuous and the magnetic fields essentially decouple from the matter .
we can effectively achieve this by setting @xmath112 in any region that is below @xmath113 .
this approach therefore accomplishes our goal of decoupling in the low - density regions the evolution of the electromagnetic fields from the dynamics of the plasma .
of course , an electrovacuum prescription where the rest - mass density is nonzero is , strictly speaking , inconsistent , but we believe that this is a tolerable inconsistency , given that the rest - mass density in these regions takes essentially the smallest values in the whole domain .
second , albeit somewhat arbitrary , our approach suffers from the same uncertainties of other approaches suggested in the literature to match the two different regimes ( see , e.g. , refs .
ideally , its robustness can be validated by varying the free parameters @xmath99 and @xmath108 , although this is something that admittedly we are not able to do here because of the computational costs involved .
more importantly , we find that this approach allows us to take an important step beyond the previous imhd treatment presented in ref . @xcite .
third , while matching to a force - free regime is appealing , it can be rather dangerous in the physical conditions encountered _ after _ the merger ( by contrast , the force - free approximation is probably very good before the merger or in black hole
neutron - star binaries , when only magnetospheric effects are expected to take place ) .
our calculations reveal in fact that the exterior of the conducting matter ( either the hmns or the torus ) is always matter dominated and the plasma beta parameter , i.e. , the ratio of the gas - to - magnetic pressure , is the one normally used in plasma physics , but the inverse of the one employed in other numerical - relativity calculations , e.g. , ref . @xcite . ]
@xmath114 , is at least @xmath115 and of the order of @xmath116 in the polar regions .
this is far from the condition of @xmath117 , where the matter inertia can be neglected and the force - free approximation is a good one . finally , despite the fact that our implementation offers a control over the amount of resistivity in different parts of the flow , the choice of realistic values for the resistivity is far from trivial ( see , e.g. , ref .
@xcite ) and not addressed at all in these simulations .
as already mentioned in sec .
[ sec : ms : rmhd ] , in order to ensure that the magnetic field is essentially divergence free , we have employed the divergence cleaning method of ref .
@xcite , which , however , requires choosing a suitable value for the constant @xmath118 [ cf .
, eqs . and ] . in the simulations presented here , we have set @xmath119 .
this value does not lead to stiffness problems in the coarser grids and at the same time provides a rapid damping of the constraint violation on a time scale @xmath120 .
when after the merger the hmns collapses to a black hole , steep gradients appear in the rest - mass density and cover just few grid points of the finest grid .
if the resolution is sufficiently low , these gradients are simply dissipated numerically and the evolution can proceed without problems . however , for the rather high resolutions used here for the finest grid , i.e. , @xmath121 , the numerical dissipation is smaller and the gradients are not removed , leading to failures in the conversion of the conserved variables to the primitive ones . to counter this problem
we reset to atmosphere the hydrodynamical variables in a mask inside which the lapse function goes below a threshold , e.g. , @xmath122 .
this reset is done only for the hydrodynamical variables , while the spacetime and electromagnetic ones are evolved as usual .
furthermore , the reset is applied in practice to a handful of cells , well inside the apparent horizon and thus not influencing the matter dynamics .
the initial data consist of a magnetized binary neutron - star system of total arnowitt deser misner ( adm ) mass @xmath123 and an initial orbital separation of @xmath124 .
each star has a baryon mass equal to @xmath125 and an equatorial radius of @xmath126 , so that the initial separation corresponds to approximately @xmath127 .
the initial orbital velocity is @xmath128 , and the maximum rest - mass density is @xmath129 .
lacking self - consistent initial data for magnetized binaries , our initial data are generated by the ` lorene ` library as an unmagnetized irrotational binary in equilibrium on a quasicircular orbit @xcite .
the magnetic field is then superimposed on the unmagnetized constraint - satisfying solution .
following refs .
@xcite , the initial magnetic field is fully contained inside the stars and purely poloidal .
this is achieved after prescribing the toroidal vector potential @xmath130 to have the form @xmath131
^ 2\ , , \label{eq : rmhda_b}\ ] ] where @xmath132 determines the point at which the magnetic field goes to zero ( typically before it reaches the surface ) .
the resulting ( poloidal ) magnetic field is just the curl of the vector potential and leads to a maximum magnetic - field strength of @xmath133 at the pressure maximum .
because the initial data constructed in this way are not a solution of the full einstein euler maxwell system , they will introduce an increased violation of the constraint equations , which amount to @xmath134 and @xmath135 in the @xmath136-norm , of the hamiltonian and momentum constraints , respectively .
these values should be compared with those before the introduction of the magnetic field and which are about 1 order of magnitude smaller , i.e. , @xmath137 and @xmath138 .
the perturbations introduced by the addition of the magnetic fields are small enough so as not to have a significant effect on the dynamics of the binary . indeed , our experience is that the @xmath136-norm of the constraints relaxes to values comparable to those of simulations of unmagnetized binaries after about one crossing time or , equivalently , one orbit .
obvious ways to improve this approach , and which could become important for an accurate modelling of the gravitational - wave emission during the inspiral , exist . among them are the use of consistent initial data for magnetized binaries or the simulation of binaries with much larger separations , so that the system has several orbits to reach a more consistent mhd equilibrium .
in addition , because the main focus of this work is the assessment of resistive effects on the postmerger dynamics and hence a comparison between imhd and rmhd simulations , our interest in an accurate treatment of the initial data is rather limited here . as a result , and once again to reduce the computational costs , we accelerate the inspiral as first suggested in ref . @xcite .
more specifically , for most of our simulations we modify the initial linear momenta by adding an initial inward radial velocity which is @xmath139 of the orbital velocity .
this reduces the number of orbits at this separation from @xmath140 to only @xmath141 . of course
, the constraint violations introduced in this way are even larger than those discussed above with the introduction of magnetic fields - norm of the hamiltonian and momentum constraints for binaries with reduced initial momenta with the corresponding violations for binaries on quasicircular orbits .
the latter amount to @xmath142 for the hamiltonian constraint and to @xmath143 for the average of the momentum constraints . ] , but we have also verified that this does not introduce qualitative differences by comparing the results of these grazing collisions with those obtained from the corresponding binaries in quasicircular orbits .
the dynamics of the same neutron - star binary considered here has been previously investigated in hydrodynamic simulations @xcite , in imhd simulations @xcite , and , more recently , also in rmhd simulations @xcite . in all cases , it was shown that after the merger this specific model forms a rapidly rotating hmns with a high degree of differential rotation .
the transient object collapses later to form a rotating black hole of mass @xmath144 .
the magnetic - field strengths chosen here and in previous works @xcite are not sufficiently high to affect the bulk dynamics of the binary during the inspiral @xcite , which can be considered to be equivalent to the purely hydrodynamical case for all practical purposes . in the following subsections ,
we focus on the results obtained from the application of our rmhd implementation .
we start by providing a general description of the basic features of the rmhd dynamics ( sec .
[ sec : bns : features ] ) and then move to make a detailed comparison with an imhd simulation of the same neutron - star binary ( sec .
[ sec : bns : comparison ] ) .
we note that our focus here is different from the one in ref .
@xcite , which matched the resistive description to a force - free one to study the interaction of the two stellar magnetospheres before the merger . here ,
on the other hand , we are mostly interested in the postmerger object and on the effects that resistivity has on the dynamics of the hmns and subsequent black hole torus system .
because of this , and because the scenario we are investigating is polluted by the large baryonic winds produced after merger , our resistive matching is made to an electrovacuum exterior . in this sense , the work carried out here and in ref .
@xcite provide a complementary description of the dynamics of magnetized binary neutron stars in full general relativity .
0.2 cm 0.2 cm the dynamics of the binary when evolved within the rmhd framework is summarized in figs .
[ fig : bns : rhob ] and [ fig : bns : rhob6 ] . the first one , in particular , reports two - dimensional cuts in the @xmath145 and @xmath146 planes of the rest - mass density @xmath59 ( left panel ) , of the specific internal energy @xmath58 ( middle panel ) , and of the modulus of the magnetic field @xmath147 ( right panel ) . marked with white lines
are the projection of the magnetic - field lines on the different planes , while marked with green solid lines are the isocontours of the rest - mass density .
the snapshots refer to times @xmath148 ( top row ) , @xmath149 ( middle row ) , and @xmath150 ( bottom row ) .
the positions of the two stars are marked with @xmath151 and @xmath152 symbols , and the stars trajectories are marked with red dashed lines .
given the small initial orbital separation of @xmath124 and the reduced linear moment , the two stars merge very rapidly .
more specifically , at approximately @xmath153 , the two stellar surfaces start entering in contact , although the actual merger takes place at @xmath154 . as the merger takes place ,
the two stellar cores become significantly distorted by the large tidal fields and produce spiral arms . at the leading edges of these spiral arms , the specific internal energy increases through shock heating ( cf . , the middle column of fig [ fig : bns : rhob ] ) .
the time @xmath155 in the top row of fig .
[ fig : bns : rhob ] corresponds to one orbit of the binary , which is sufficient for the magnetic field to diffuse over the thin transition layer close to the surface of the stars in an attempt to settle to a new equilibrium configuration ( cf .
, the discussion in sec .
iv.c.1 of ref .
once the magnetic field has diffused out of the star it continues to propagate also in regions that are dynamically treated as atmosphere and where the electrical conductivity is set to zero as if the medium was an electrovacuum . in this way , we achieve a rather smooth transition between the highly conducting stellar interior and the electrovacuum exterior .
this is shown in fig .
[ fig : bns : sigma ] , where we show the electrical conductivity at three reference times on the @xmath146 plane . note that , the region in black corresponds to our electrovacuum but does not coincide with the atmosphere .
indeed , the region in black is filled with tenuous plasma as can be seen in the left columns of fig .
[ fig : bns : rhob ] . note also that at this time the magnetic - field topology is still predominantly poloidal in the exterior of the star .
however , a toroidal component is also being generated as the highly conducting material in the stellar interior shears the poloidal magnetic field in the lower - density , high - conductivity spiral arms .
0.2 cm 0.2 cm as mentioned above , at @xmath149 the merger takes place , at least as measured from the position of the first peak in the gravitational - wave amplitude .
when this happens , a vortex sheet is created between the two stars , which could lead to the onset of a kelvin helmholtz instability and the generation of a large - scale and ultrastrong magnetic field @xcite ( see the middle row of fig . [
fig : bns : rhob ] ) .
it is presently a matter of debate whether such large - scale magnetic fields can be produced with amplifications of several orders of magnitude .
present direct simulations are not able to reach the resolutions necessary to resolve the turbulent motion produced by the instability @xcite .
the results obtained so far with direct , very high - resolution simulations , either local @xcite or global @xcite , indicate that the amplification of the magnetic field is of a factor 20 at most , most likely because resistive instabilities disrupt the kelvin
helmholtz unstable vortex @xcite ( but see also ref .
@xcite for recent simulations with subgrid modeling which could lead to much larger amplifications ) . in the simulations reported here
, we find that the magnetic - field magnitude increases slightly less than an order of magnitude during this stage ( see fig . [ fig : bns : mag ] and the discussion in sec .
[ sec : bns : comparison ] ) .
this moderate growth of the magnetic field is possibly due to insufficient resolution and our inability to capture the dynamics of the relevant scales . on the other hand
, it could also indicate that a kelvin helmholtz instability simply does not develop .
a possible reason is that the time scale of the instability might be longer than the dynamical time scale and the shearing motion could very rapidly get destroyed as the two stellar cores collide on a time scale of a fraction of a millisecond .
we should also note that resistive effects might become important at this stage as the reconnection of magnetic - field lines might lead to the acceleration of matter due to ohmic heating .
we believe that higher - order schemes , as those presented in refs .
@xcite , could help resolve this issue .
the bar - deformed hmns produced after the merger is differentially rotating , and magnetic braking transfers angular momentum from the inner core to the outer parts of the star .
the spiral arms widen and merge together generating more shock heating and dissipation .
a magnetically driven wind as a result of differential rotation is launched from the outer layers of the hmns @xcite .
the wind could play an important role in the modelling of short gamma - ray bursts ( sgrbs ) , which show an extended x - ray emission @xcite .
the wind is not constant in time , but rather characterized by a bursty activity in which high internal energy plasma blobs ( i.e. , local concentrations of specific internal energy a few kilometers in size ) are launched from near the black - hole horizon and propagate along the @xmath156-direction ( bottom row of fig .
[ fig : bns : rhob ] ) .
interestingly , the bursts observed in the specific internal energy of the hot rotating halo that forms around the central object are anticorrelated with the bursts observed in the modulus of the magnetic field ; a more detailed discussion on these bursts follows in sec .
[ sec : bns : bursts ] .
unfortunately , the resolution used here is insufficient to be able to track the development of an mri , which is , however , expected to develop @xcite and could significantly amplify the magnetic field .
blind to this effect , our simulation shows that at @xmath157 ( see the bottom row of fig .
[ fig : bns : rhob6 ] ) magnetic braking has managed to store enough rotational energy in the winding of the magnetic - field lines so that the inner core of the star is now less differentially rotating .
the direct consequence of this is that the hmns collapses to a black hole of mass @xmath158 and dimensionless spin @xmath159 , as measured from the apparent horizon @xcite .
we postpone the discussion on how angular momentum is transported outward and how this affects the lifetime of the hmns to sec .
[ sec : bns : lifetime ] . at time @xmath160 ,
the black hole is surrounded by a thick accretion torus that is responsible for confining and collimating the magnetic field along the @xmath156-axis ( cf . , the top row of fig .
[ fig : bns : rhob6 ] ) .
the properties of the black hole - torus system are shown in table [ tab : bns : bhtorus ] at approximately @xmath161 after the collapse and show that the mass of the torus is @xmath162 . the magnetic - field topology and matter
dynamics soon after the collapse are highly turbulent , but the high degree of symmetry introduced by the black hole , which is gravitationally dominant over the torus , rapidly establishes some order in this system . after about one orbital period ,
in fact , the torus becomes essentially axisymmetric and the matter in the polar region is rapidly accreted onto the black hole , giving rise to a funnel where the rest - mass density reaches values close to that of the atmosphere ( cf . , the bottom row of fig .
[ fig : bns : rhob6 ] ) .
0.5 cm 0.5 cm 0.5 cm after black - hole formation , the plasma dynamics in the funnel is far from being stationary and continues with repeated bursts having a period of about @xmath163 . while a more detailed discussion of these bursts is postponed to sec .
[ sec : bns : bursts ] , it is useful to remark here that the ejected material does not have sufficient energy to reach large distances away from the black hole .
this is probably due to the fact that , although the magnetic field is comparatively strong , the material in the funnel is still matter dominated , with @xmath164 .
such large values are not particularly surprising since the initial magnetic field in the stars is rather small , i.e. , @xmath165 , and is not amplified significantly . at the same time , the torus angular velocity profiles have become nearly keplerian , and the mri could develop ( cf . , fig .
[ fig : bns : omega ] ) .
however , as for the hmns , also the spatial resolution of the grid covering the torus is too small to capture the fastest growing modes of the instability in the torus ( see discussion in sec . [
sec : bns : torus ] ) . for completeness , we report in fig .
[ fig : bns : epsvel ] three different two - dimensional cuts on the @xmath146 plane of the specific internal energy and of the velocity field in this plane .
the times selected are the same as those presented in fig .
[ fig : bns : sigma ] , and are useful to gain information on the properties of the flow during the hmns stage and after the collapse to a black hole . note that during the hmns stage the flow consists mostly of the intense magnetically driven outgoing wind discussed above ( left panel ) .
however , once a funnel is produced , the flow is ingoing at heights @xmath166 but becomes outgoing for @xmath167 ( middle and right panels ) .
the location of this stagnation point varies with time and follows the expansion of hot fluid which is produced by the accretion process and that can be followed via the increases in the specific internal energy . in all cases considered ,
the vertical flow very close to the black hole is _
ingoing_. as mentioned above , the high degree of symmetry near the rapidly rotating black hole induces a quick rearrangement of the matter and of the magnetic fields . as a result , the magnetic - field topology at time @xmath168 changes in the funnel and develops a dominant poloidal component , giving rise to a well - defined magnetic - jet structure , which is almost axisymmetric .
this result is similar to what was already found in the simulations of ref .
@xcite , with the important difference that this configuration has been reached with a higher spatial resolution and a consistent treatment of the resistivity . -0.55 cm -0.40 cm -0.40 cm -0.55 cm -0.40 cm -0.40 cm the rest - mass density of the plasma in the funnel is close to that of the atmosphere , but also slightly larger . hence , given our choice of the conductivity profile in eq . , the conductivity is essentially zero everywhere in the funnel ( see the middle and right panels of fig .
[ fig : bns : sigma ] ) .
this has two important consequences .
first , the dynamics of the electromagnetic fields in this region is not that prescribed by the imhd equations but rather that of electromagnetic waves in vacuum . at the same time , because of its ( comparatively ) small rest - mass density and pressure , the matter in the funnel tends to move along the field lines .
we should clarify that this behavior is not achieved because the test - particle limit of the rmhd equations is reached ( the rest - mass density and pressure are in fact nonzero ) but rather because the matter in the funnel can only move in the vertical direction , either accreting onto the black hole or moving outward ( the matter in the funnel has low or zero specific angular momentum ) .
modelling the dynamics of the matter in the funnel is among the most challenging aspects of these calculations .
although the matter there has the largest magnetic - pressure support ( i.e. , @xmath169 ) , this is still about 4 orders of magnitude away from being magnetically dominated .
the reason for this behavior is most likely due to the comparatively weak magnetic fields that we are able to build in the funnel at these resolutions and the short evolution times .
second , the zero - conductivity plasma in the funnel is just adjacent to the high - conductivity plasma of the torus ; this large jump in the conductivity helps in preserving the magnetic - jet structure and in providing a natural agent for the collimation of the flow at low latitudes .
it is useful to remark that in close analogy with what found in ref .
@xcite the magnetic - jet structure produced here is _ not _ a relativistic outflow . instead , it can just be viewed as an almost quasistationary magnetic structure confining the tenuous plasma in the funnel and confining it away from the dynamics of the ultradense plasma in the torus .
in fact , despite the resistive losses , the plasma in the funnel does not have yet sufficient internal energy to be able to launch a relativistic outflow .
it is possible that the strong magnetic fields in the vicinity of the black hole could provide the conditions for electromagnetic extraction of the black hole s rotational energy through the blandford
znajek mechanism @xcite or through a generalized penrose process @xcite .
alternatively , the energy required for launching a relativistic outflow could also be efficiently deposited along the baryon - poor funnel by reconnection processes not fully modelled here or by neutrino pair annihilation @xcite .
clearly , additional work is needed to assess the robustness of our modelling of the funnel region and to assess whether and how energy can be deposited in the magnetic - jet structure
. a closer look at the magnetic - jet structure is offered in fig .
[ fig : rho_and_b_xz_large ] , which shows two - dimensional snapshots on the @xmath146 planes of the rest - mass density ( top row ) and of the magnetic field .
the two columns refer to @xmath150 ( left column ) , when the hmns is just about to collapse , and to @xmath168 ( right column ) , when the black hole
torus system is toward reaching a quasistationary equilibrium .
the different panels in fig .
[ fig : rho_and_b_xz_large ] should be compared with the corresponding ones in figs .
[ fig : bns : rhob6 ] and [ fig : bns : magtorpol ] but are represented here on much larger spatial scales ( see also appendix [ sec : appendix : imhd ] for a closer comparison of the magnetic - field structure in the imhd simulations ) .
interestingly , the magnetic - jet structure extends well beyond the scale of direct influence of the black hole and shows a coherent structure on scales of @xmath170 .
the scale of the magnetic structure is much larger than that of the accreting torus ( i.e. , @xmath171 ) and of the black hole ( i.e. , @xmath172 ) .
additional information on the magnetic - field topology can be appreciated by considering three - dimensional views of the magnetic - field strength and field lines .
this can be seen in figs .
[ fig : bns : magtorpol ] , where we show a three - dimensional snapshot of the toroidal ( left panel ) , poloidal ( middle panel ) , and total magnetic field ( right panel ) .
it is important to remark that a well - defined magnetic - jet structure has been recently reported also in ref .
@xcite from imhd simulations of the merger of a black hole neutron - star binary .
in addition to a jet structure not very different from the one reported here , the authors in ref . @xcite are also able to produce a sustained outflow from the accretion torus . at the same time , the imhd simulations reported in ref .
@xcite , where a very high spatial resolution was used , do not reveal the formation of such a magnetic - jet structure .
it is difficult to assess at the moment the origin of these differences , partly because of the limited amount of information provided on the simulations in ref .
. a more extended discussion of the properties of the magnetic - field topology and dynamics , made along the lines suggested in this paper ( see also sec . [
sec : spd ] ) would be useful to clarify if there are really differences and their origin .
when studying the magnetic - field topology , i.e. , whether it is mostly poloidal or toroidal , it is unavoidable to discuss where certain measurements are made , as the magnetic field can be at the same time mostly poloidal and mostly toroidal but in two different regions .
a quick inspection of the three panels suggests that the magnetic field in the low - density funnel is predominantly poloidal ( clearly shown in the middle panel of fig .
[ fig : bns : magtorpol ] ) .
it is quite natural to expect that the magnetic field will be essentially poloidal near the rotation axis , where matter has a specific angular momentum that is intrinsically small .
equally natural is to expect that a toroidal component will start to develop away from the axis and as one approaches the regions filled by the torus .
this behavior can be explained by the fact that some of the magnetic - field lines in the funnel are anchored to the highly conducting material at the edges of the torus , which is rotating at nearly keplerian velocities .
indeed , the left panel of fig .
[ fig : bns : magtorpol ] shows that the magnetic field acquires a toroidal component near the edges of the funnel , and then becomes essentially toroidal in the torus .
this is also shown in fig .
[ fig : bns : mag : fieldlines ] , which offers three - dimensional snapshots of the rest - mass density @xmath59 ( left panel ) , of the specific internal energy @xmath58 ( middle panel ) , and of the modulus of the magnetic field @xmath173 ( right panel ) at @xmath174 .
also reported are the magnetic - field lines , of which the three - dimensional representation confirms that the magnetic field is mostly poloidal in the magnetic - jet structure , acquiring a twist and a kink when it reaches the edges of the funnel , and becoming essentially toroidal inside the torus .
finally , we note that a vigorous outflow develops at the interface between the magnetic - jet structure and the torus .
the resulting shearing boundary layer could be the site for the development of a kelvin helmholtz instability , which unfortunately we can not investigate at the present resolutions and without a more sophisticated treatment of the conductivity in the transition between large and small values ( the instability does potentially develop in a particularly difficult region ) .
it is clear , however , that the dynamics along the torus walls has all the potential of yielding interesting observational features and should be investigated in the future , possibly using advanced numerical techniques such as those presented in ref .
@xcite .
a word of caution should be spent before concluding this section .
while the behavior described above appears reasonable and is possibly the expected one on the basis of rather simple considerations , this behaviour is ultimately the result of our choice for the conductivity profile in eq . and of our choice of rest - mass density in the atmosphere .
the large computational costs associated with these simulations prevent us from presenting at this point a systematic investigation of how sensitive the results are on the choice for @xmath79 and @xmath108 .
we are aware that this represents a limitation of our investigation , which we plan to resolve with future simulations .
as described in the previous section , the dynamics of the rmhd simulations is rather similar , at least qualitatively , to the one observed in imhd simulations of the same binary presented in ref .
yet , there are some important differences , and these can be best appreciated if we perform a careful comparison of the two evolutions . to this scope , we have performed additional simulations of the same binary discussed in the previous section when , however , the set of equations solved are those of general - relativistic imhd .
we note that this was necessary because the simulations in ref .
@xcite used a different grid structure and resolution but also investigated quasicircular initial data in contrast to the reduced linear momenta we have considered here . in the following
we present the results of this side - by - side comparison , first using two - dimensional spacetime diagrams and then moving to standard one - dimensional snapshots .
we have found that a very efficient way of carrying out a comparison between two mhd evolutions which are qualitatively similar is to use two - dimensional spacetime diagrams .
this technique , which was first introduced in ref .
@xcite , provides a color - coded evolution of various scalar quantities as measured along principal axes ( e.g. , the @xmath175- and @xmath156-directions ) and has the advantage of summarizing simply even rather complex dynamics . as representative examples
, we show in figs .
[ fig : bns : spt_far1 ] and [ fig : bns : spt_far2 ] the differences between the rmhd and imhd implementations in the evolution of the rest - mass density @xmath59 ( top left panel ) , of the specific internal energy @xmath58 ( top right panel ) and of the magnetic field norm @xmath173 ( bottom panel ) .
note that each panel is split into two diagrams , with the left one referring to the rmhd solution and the right one to the imhd solution .
furthermore , while fig .
[ fig : bns : spt_far1 ] refers to quantities measured along the @xmath175-axis and hence is representative of motions on the equatorial plane , fig .
[ fig : bns : spt_far2 ] refers to the @xmath156-axis and is therefore representative of motions in the polar region . in all panels we indicate with a solid green line the evolution of the apparent horizon ( see the discussion in ref .
@xcite on how to interpret such a line ) .
note that the values in the color bars are saturated and do not correspond to the minimum and maximum values of the corresponding fields .
it is clear from both figures that the evolution of all quantities is very similar during the inspiral , when indeed the imhd and rmhd evolutions should be mathematically identical given that our fields are contained in the stars .
as the binary reaches the merger , matter is expelled from the stars mainly in the equatorial plane , as is evident in fig .
[ fig : bns : spt_far1 ] , which also shows that the ejected matter moving through the low - density medium generates shocks that heat up the plasma and appear as thick lines .
the subsequent bursts of matter are mainly associated with the fundamental mode ( @xmath176-mode ) of oscillation of the hmns @xcite .
the equatorial ejections are also accompanied by four or five subsequent `` bursts '' along the @xmath156-axis , starting in both simulations at approximately @xmath177 , and can be seen in fig .
[ fig : bns : spt_far2 ] . clearly , the violent oscillations experienced by the hmns launch matter essentially isotropically . we have already mentioned that the differentially rotating hmns does collapse promptly to a black hole by first rearranging its angular velocity profile and mass distribution through magnetic braking . during this phase , angular momentum is transported outward , with the outer fluid elements moving further away from the star and the inner ones moving toward the center as a result of the angular - momentum losses .
clearly , this redistribution of angular momentum will be different in the rmhd and imhd evolution , with the latter being more efficient in transporting angular momentum outward ( in imhd the fluid can only move along magnetic - field lines , while it can also partially cross them in rmhd ) . as a result
, the collapse can take place slightly earlier , occurring at @xmath178 in the imhd evolution and at @xmath179 in the rmhd simulation ( cf .
, figs .
[ fig : bns : spt_far1 ] and [ fig : bns : spt_far2 ] ) .
we have already commented that one of the major differences between the imhd and rmhd simulations is that our implementation of the latter allows for a description of propagating electromagnetic waves in vacuum . by contrast , the atmosphere treatment of the imhd implementation is such that it does not evolve the magnetic field as the fluid velocities are reset to zero there . as a result , already after the first @xmath180 the magnetic field manages to diffuse out of the stars , heating up the plasma in the outer layers and forcing it to expand . in this way , magnetic energy is converted into internal energy , and therefore the magnetic field at the center of the stars is ( slightly ) lower than in the corresponding imhd simulation .
in contrast , the magnetic field along the @xmath156-axis in the rmhd simulation is higher than in the imhd simulation in the first few milliseconds .
indeed , the differences between the imhd and rmhd simulations become particularly evident after black - hole formation , when the funnel is evacuated and the magnetic - jet structure is built ( see the bottom panels of figs .
[ fig : bns : spt_far1 ] and [ fig : bns : spt_far2 ] ) .
the magnetic diffusivity in the rmhd simulation acts so rapidly that in a bit more than one crossing time the whole computational domain in the rmhd run is filled with electromagnetic fields despite the fact that the magnetic field was initially constrained to the stellar interior .
note that the modulus of the magnetic field along the @xmath156-axis is about 2 orders of magnitude larger in the rmhd simulation ( cf .
, @xmath181 in imhd vs @xmath182 in rmhd ) .
this is due ( mainly ) to the intense currents produced by the rapidly rotating torus and ( partly ) to the magnetic - field diffusion of the strong magnetic field in the torus , which diffuses across the walls of the funnel .
this behavior of the magnetic field is in fact similar to the one observed in a stable magnetized star with extended magnetic fields which was studied in ref .
@xcite , where after an initial transient , the system relaxed to a solution consisting of a large - scale , nearly electrovacuum , dipolar magnetic field configuration in the exterior , which was anchored to the highly conducting neutron star .
the larger values of the magnetic field in the funnel are particularly encouraging as strong magnetic fields are necessary to produce a large acceleration along the @xmath156-direction .
the maximum lorentz factors achieved after the collapse in both simulations is @xmath183 , with the highest values occurring at the end of the simulations .
the magnetic field in the torus is also stronger in the rmhd simulation than the corresponding field in the imhd run , although the differences in this case are only of 1 order of magnitude .
the reason for this has to be found in the fact that in the rmhd simulation the torus is more massive and therefore able to sustain larger amounts of magnetic fields ( as for isolated stars , also a self - gravitating torus in mhd equilibrium will be able to sustain stronger magnetic fields for increasing masses ; cf . ,
table [ tab : bns : bhtorus ] ) .
the top panel of fig .
[ fig : bns : rho ] reports the evolution of the maximum of the rest - mass density @xmath184 for both the imhd ( blue solid line ) and the rmhd simulation ( green solid line ) .
after the merger takes place at @xmath185 , the maximum rest - mass density oscillates at the @xmath176-mode frequency and experiences a sudden drop when an apparent horizon is found since matter inside the apparent horizon is excluded from the calculation of @xmath184 .
it is quite obvious that the behavior of @xmath184 is different in the two simulations , with the hmns in the rmhd surviving for a longer time .
note , however , that the first increase in the central rest - mass density in the top panel of fig . [ fig : bns : rho ] is slightly larger in the rmhd case , most likely because of the resistive increase of the internal energy at the merger . also shown in the top panel of fig .
[ fig : bns : rho ] is the evolution of a simulation in which the set of rmhd equations is used together with a large and uniform conductivity @xmath186 ( red solid line ) . in this case , despite the very different set of equations solved and the different approach for enforcing the divergence - free condition of the magnetic field , the rmhd evolution should mimic the imhd one .
this is indeed the case , and it provides considerable confidence on the robustness of our rmhd approach .
it suggests that the delayed collapse is an effect associated with the choice of physical resistivity and not due to numerical artifacts . despite the complex dynamics , the differences between the imhd and rmhd runs are not difficult to explain . as mentioned in the previous section , in fact ,
an important difference between the imhd and rmhd simulations is that in the latter the magnetic field can not be perfectly locked with the plasma . as a result ,
the imhd evolution is more efficient in redistributing the angular momentum in the system and , in particular , in transporting it outward .
this magnetic - braking process deprives the hmns core of the angular - momentum support , and this leads to an earlier collapse . clearly , since the conductivity in the hmns interior is very high also in the rmhd simulation ( although not infinite ) , magnetic flux freezing is very efficient here as well and the differences in the dynamics of the imhd and rmhd simulations can only be small .
this explains why overall the time of collapse varies by only @xmath187 .
in addition , the important differences in this dynamics are also expected to take place in the outer layers of the hmns , where the conductivity decreases as a response to the conductivity profile .
we should note that the difference in the lifetime of the hmns is not related to the use of initial data with modified momenta , but is present also for a binary of which the initial data is on a quasicircular orbit .
this is shown in the bottom panel of fig .
[ fig : bns : rho ] , where the corresponding quantities are shown , and where it is clear that also in this case the hmns collapses earlier to a black hole in the imhd simulation .
we also recall that for the binary in quasicircular orbit no additional refinement level is added after the merger .
hence , when comparing in fig .
[ fig : bns : rho ] the rmhd evolution of the binaries with reduced momenta and on quasicircular orbits , one is also effectively comparing the evolution of the hmns at two different resolutions , i.e. , @xmath188 and @xmath189 , respectively . the fact that they both yield a longer lifetime of the hmns provides an indirect validation of the numerical consistency of the rmhd solution . because the differences in the magnetic braking between the rmhd and imhd implementations are intrinsically small , it is not easy to show that it is exactly these differences that are responsible for the earlier collapse of the hmns in the imhd simulations . however , such evidence is offered in fig .
[ fig : bns : angularmomentum ] , which reports the spacetime diagrams along the @xmath175-direction of the specific angular momentum @xmath190 @xcite .
( see @xcite for a discussion in the differences in the two definitions ) . ]
the left panel refers to @xmath191 , the specific angular momentum of the rmhd run , while the middle panel refers to the corresponding quantity for the imhd simulation , @xmath192 , and the right panel refers to the relative difference : @xmath193 .
a rapid inspection of the left and middle panel reveals that in both the rmhd and imhd simulations the specific angular momentum increases outward ( as it should for a rotating fluid satisfying the rayleigh stability criterion ) , but also that the profiles are not constant in time and show instead periodic variations that are in phase in the two simulations .
these variations reflect the large oscillations of the hmns , and , indeed , the oscillations in @xmath194 take place at the same frequency as those in the rest - mass density and shown in the other spacetime diagrams in fig .
[ fig : bns : spt_far1 ] ( note the different scale in the @xmath175-axis ) . however , there are important small differences in the dynamics of @xmath191 and @xmath192 , which are apparent in the right panel of fig .
[ fig : bns : angularmomentum ] , where regions in red indicate that the specific angular momentum in these regions is higher ( of @xmath195 ) in the imhd simulation than in the rmhd simulation , while blue regions are exactly the opposite .
it is apparent that the differences are larger in the central parts of the hmns , while the specific angular momenta are very similar in the outer layers , i.e. , for @xmath196 .
we recall that an excess of specific angular momentum at a given position on the @xmath175-axis reflects fluid elements that are rotating at a larger frequency ( @xmath197 in the newtonian limit ) and this is indeed what one would expect if the magnetic fields and the fluid are tightly coupled .
hence , the red regions in the right panel fig . [ fig : bns : angularmomentum ] can be taken to signal a more efficient transfer of angular momentum from the inner regions of the hmns .
a direct consequence of this transfer of angular momentum is the appearance of blue regions adjacent to the red ones and signalling therefore fluid elements that have slowed down .
although the differences in @xmath191 and @xmath192 are small , the transfer continues steadily and with an increased rate up to @xmath198 , when a much larger transfer of angular momentum takes place .
this signals the onset of the instability to gravitational collapse in the imhd simulation , which effectively takes place soon after , i.e. , at @xmath199 .
three remarks should be made before concluding this section .
first , the fact that an rmhd simulation with a uniform conductivity yields the same collapse time as an imhd simulation gives us great confidence about the correctness of the rmhd evolution with nonuniform conductivity . second , although the difference in the survival time between the rmhd and imhd evolution is here rather small , it can be much larger if smaller values of the resistivity are chosen for the stellar interior and for less massive hmnss ; unfortunately present astronomical observations do not set any stringent constraint on the values of the conductivity at these temperatures , rest - mass densities , and magnetic fields .
more importantly , however , a longer survival time is a useful new result in the modelling of binary neutron stars , as it points out that the hmns can survive on comparatively longer time scales than those computed so far in pure hydrodynamics or in imhd simulations .
this is not a minor detail as the most recent modelling of sgrbs with an extended x - ray emission invokes the existence of a magnetized hmns that is able to survive on time scales of the order of @xmath200 before collapsing to a black hole @xcite .
finally , computational constraints have prevented us from extending much past black - hole formation the evolution of the rmhd / imhd simulations with initial data on quasicircular orbits .
nevertheless , the fact that already the dynamics of the hmns is unaffected by the initial reduction in linear momenta and that the hmns collapses to a black hole earlier in both cases , provides us with confidence that the magnetic field dynamics discussed in sec .
[ sec : bns : magjet ] will be very similar also for binaries on quasicircular orbits .
this will also be the focus of our future work . in secs .
[ sec : bns : magjet ] and [ sec : spd ] , we have already discussed the properties of the evolution of the magnetic fields , but have not quantified in detail how the magnitude of the magnetic field changes with time and how this evolution varies in the rmhd and imhd simulations .
this is done now in fig .
[ fig : bns : mag ] , where we report the evolution of the maximum of the modulus of the magnetic field ( black lines ) but also of its toroidal ( red lines ) and poloidal ( blue lines ) components , either in the rmhd simulation ( left panel ) or in the imhd simulations ( middle panel ) .
note that the evolution of the magnetic field in the two implementations is very similar during the inspiral , but also that this changes considerably after the merger .
while in both cases the toroidal magnetic field grows exponentially , the growth is of about 1 order of magnitude in the imhd simulation but of a factor 2 smaller for the rmhd simulation .
furthermore , the magnetic field reaches values of @xmath201 just before the collapse in the imhd simulation and @xmath202 only just after the merger in the rmhd simulation .
this different behavior is not difficult to explain and is simply due to the fact that the shearing of magnetic - field lines is less efficient in rmhd because of the finite conductivity of the matter . because the growth at the merger is mostly due to the shear layer between the two impacting stars
, it is quite natural that a resistive calculation will lead to a smaller magnetic field , quite independently of how well the instability is resolved . as for the rest - mass density ( cf . , fig .
[ fig : bns : rho ] ) , the collapse of the hmns to a black hole leads to a rapid decrease of the maximum value of the magnetic field , as shown in fig .
[ fig : bns : mag ] where the vertical lines signal the first appearance of the apparent horizon .
also in this case , the strongest magnetic fields are hidden inside the horizon , and the maximum values reported are those relative to the magnetic field in the torus . as remarked already when commenting on the spacetime diagrams in sec .
[ sec : spd ] , the larger values of the magnetic field in the rmhd simulation are the result of a more massive torus produced in this case ( cf . , table [ tab : bns : bhtorus ] ) . in the right panel of fig .
[ fig : bns : mag ] we complement the evolution of the magnetic fields with the evolution of the @xmath136-norm of the divergence of the magnetic field for the rmhd simulation ( green dashed line ) and for the imhd one ( blue solid line ) after it is multiplied by @xmath203 .
we recall that the imhd simulation makes use of a constrained transport scheme @xcite , and hence it is able to maintain the violations of this constraint down to machine precision . the rmhd simulation , on the other hand , makes use of a divergence - cleaning scheme @xcite , which is widely known to be less efficient in suppressing the violations .
yet , the purpose of making this comparison is mostly that of highlighting that the divergence - cleaning approach used here may not be as efficient as the constrained transport , but it yields nevertheless very small violations .
indeed , the evolution of the @xmath136-norm of the ratio between the divergence and the magnetic field strength , i.e. , @xmath204 and not shown in fig .
[ fig : bns : mag ] , is of the order of @xmath205 , thus similar to the values reported by similar works in the literature @xcite .
note that the late - time moderate growth of the divergence in the rmhd simulation is due to the amplification of the magnetic fields in the torus , and it would be interesting to investigate if a dynamically adapted dissipation parameter for the divergence cleaning method could help reduce such a growth . are shown for the rmhd simulation with a green dashed line and for the imhd simulation with a solid blue line .
the time of collapse for the rmhd case is depicted with a black dashed line , while it is depicted with a black dotted line for the imhd one . ] in secs .
[ sec : bns : features ] and [ sec : spd ] , we have already anticipated that a wind is produced after the merger , either as a result of shock heating at the merger or because of magnetic winding and consequent pressure imbalance in the outer layers of the hmns or because of neutrino losses @xcite .
in addition to an almost quasistationary and quasi - isotropic wind , both the rmhd and the imhd simulation show the existence of mildly anisotropic and quasiperiodic launching of low rest - mass density , high internal energy blobs of matter that we will refer to as bursts .
overall , seven bursts are launched during the total time of the simulation , with five bursts relative to the hmns stage and two being produced after black - hole formation . in particular , the first two bursts eject material that is moving through the low - density atmosphere with an average speed of @xmath206 but that decelerate as they move away from the hmns , reaching a final outward velocity of @xmath207 .
this behavior is in part due to the natural conversion of kinetic energy to binding energy but also to the interaction of the bursting material with the slow isotropic wind .
this interaction , which is obviously accompanied by shocks , provides a damping mechanism on the propagation of the ejected material .
however , the slow down of the baryon rich material could be revived later on if the slow wind is impacted by a faster , baryon poor wind , as suggested in the `` two - winds '' model of ref .
@xcite .
we have investigated the properties of the bursts by studying the evolution of the specific internal energy as measured by an observer on the @xmath156-axis .
this is reported in fig .
[ fig : bns : bursts ] for an observer at @xmath208 , for both the rmhd simulation ( green dashed line ) and the imhd one ( blue solid line ) . clearly , after the merger the specific internal energy exhibits a quasiperiodic behavior in both simulations , with a very rapid increase .
the increase in the specific internal energy is of slightly less than 2 orders of magnitude and is followed by a slower decay . as already mentioned in sec .
[ sec : spd ] , this behavior can be associated with the oscillations of the hmns and is observed also in the evolution of the rest - mass density ( cf . , the five peaks in fig [ fig : bns : rho ] and in fig . [ fig : bns : spt_far2 ] , top left panel ) .
the characteristic frequency of these peaks is related to the @xmath176-mode frequencies of the bar - deformed hmns @xcite and thus these bursts occur every @xmath209 .
similar peaks ( although less marked ) can be observed also in the magnetic field along the @xmath156-axis ( cf . , fig .
[ fig : bns : spt_far2 ] , bottom panel ) .
the difference is that they occur slightly later than the specific internal energy ones , possibly indicating a conversion of kinetic energy into magnetic energy and vice versa .
the rest - mass density in the blobs of low - density material ejected in the bursts depends on height , but , at a distance of @xmath210 along the @xmath156-axis , it is @xmath211 , while the magnetic field , at the same location , has a strength of @xmath212 .
the bursting activity continues also after the collapse , but with somewhat different properties .
first , the frequency is now set by the radial epicyclic frequencies of the oscillating torus as deduced , for instance , when analyzing the time series of the specific internal energy at a @xmath213 on the @xmath175-axis ( see refs .
@xcite for an introduction to these frequencies in rotating tori ) .
second , the rest - mass density of the blobs ejected is smaller and of the order of @xmath214 , while the magnetic field oscillates between @xmath215 and is stronger probably as a result of the magnetic field increase in the funnel . with only two bursts observed after black - hole formation
, the time span is too short to reach a firmer conclusion on the origin of the bursts after the collapse .
however , all present evidence seems to suggest that the postmerger bursts are triggered by an increased mass - accretion rate as the torus approaches the black hole during the inward phase of its epicyclic oscillation .
of course also resistive reconnection processes could lead to the conversion of magnetic energy into internal energy and thus may be invoked to explain this phenomenology .
we find this not a likely explanation mostly because of the ordered magnetic field structure that builds up in the funnel and which seems rather stationary . clearly , a more detailed study of long - term evolutions with different prescriptions for the conductivity profiles is necessary before reaching more robust conclusions .
we finally note that the outflows produced either by shock - heating , magnetically driven winds or by the periodic bursts , eject a substantial amount of matter .
more specifically , the total rest - mass flux across a spherical surface located at @xmath216 is measured to be @xmath217 , which amounts to a total of @xmath218 ejected from the beginning of the simulation and over the survival time of the hmns ( i.e. , @xmath179 ) .
we also note that the mass - ejection rates reported here are larger than those obtained in ref .
@xcite , where mass fluxes of @xmath219 were reported . however , this is not particularly surprising and for a number of reasons .
first , the hmns considered here is the self - consistent result of a binary merger , while the one studied in ref .
@xcite was built using an axisymmetric differentially rotating equilibrium with a standard ( but somewhat arbitrary ) law of differential rotation .
second , as mentioned in sec .
[ sec : id ] , the linear momenta in our initial data are artificially reduced to accelerate the inspiral .
this also leads to a more violent merger and to larger mass losses at least till black - hole formation .
after the hmns collapses , in fact , the mass flux saturates at @xmath220 . finally , the initial magnetic field in ref .
@xcite is about 2 orders of magnitude larger , and this facilitates substantially the loss of mhd equilibrium at the surface of the hmns and thus the mass loss . because the magnetic field is not strong enough to alter the torus dynamics significantly @xcite , it is natural to expect the two solutions ( imhd and rmhd ) to be very similar in terms of dynamical properties once a nearly quasistationary state is established after the collapse to a black hole .
this expectation is confirmed in the upper panel of fig .
[ fig : bns : omega ] , which shows the rest - mass density distribution along the @xmath175-direction at about @xmath161 after the formation of the apparent horizon . clearly , the two distributions are very similar but not identical . the rmhd simulation , in particular , yields larger rest - mass densities in the outer portions of the torus , which , in turn , are responsible for larger rest masses ( cf . , table [ tab : bns : bhtorus ] ) and stronger magnetic fields ( see the discussion in sec . [
sec : spd ] ) .
furthermore , the rmhd simulation also yields a larger torus as measured from the average position on the @xmath175-axis where the rest - mass density falls below @xmath221 .
-axis for the rmhd ( green dashed line ) and imhd ( blue solid line ) simulations at time @xmath222 after the apparent - horizon formation .
_ bottom panel _ : the same as above but for the angular velocity ; also shown as a reference is a keplerian profile ( black dashed line ) . ]
.properties of the black hole
torus system at @xmath223 after the appearance of the apparent horizon for both the resistive and imhd simulations . shown
are the mass and dimensionless spin of the black hole , as well as the rest mass and size of the torus as estimated with a cutoff on the rest - mass density at @xmath224 . [ cols="<,^,^,^,^",options="header " , ] additionally , the bottom panel of fig .
[ fig : bns : omega ] illustrates the angular velocity profile along the @xmath175 direction @xmath225 for the rmhd implementations ( green dashed line ) and for the imhd one ( blue solid line ) , at the same time as the upper panel .
also shown as a black dashed line is a reference keplerian profile , i.e. , scaling like @xmath226 and which is well matched by both distributions . as remarked in ref .
@xcite , the fact that the outer parts of the torus have a quasi - keplerian behavior has two important implications .
first , it suggests that the tori will be stable and not subject to dynamical instabilities that would lead to their rapid destruction ( see , e.g. , refs .
second , a quasi - keplerian profile also provides optimal conditions for the development of an mri in the torus , thus opening the possibility of further amplification of the magnetic fields that are present after the collapse of the hmns .
1.5 cm despite the exploratory nature of the simulations carried out here , we have computed for both the rmhd and imhd implementations the total electromagnetic luminosity @xmath227 emitted .
this has been estimated as a surface integral of the poynting flux over spherical surfaces placed at representative coordinate radii @xmath228 , where @xmath228 has been varied to guarantee that the measurement is an asymptotic one and is not affected by the local plasma dynamics .
we recall , in fact , that , because in the imhd approximation the magnetic fields are locked with the plasma , the electromagnetic luminosity estimates can be heavily influenced by the presence of matter in the outer regions and thus not correspond to a genuine amount of electromagnetic energy flux leaving the system .
unfortunately , there is no simple way within the imhd approximation of determining whether the integral of the poynting flux computed on the numerical grid is genuinely asymptotic .
however , it is certainly reassuring if the values of @xmath227 are independent from the extraction radius . the evolution of the electromagnetic luminosity is illustrated in fig . [ fig : bns : poynting ] , where we report it as computed at three different extraction radii , @xmath229 .
note that the luminosity during the inspiral phase of the rmhd simulation is much larger than the corresponding one in the imhd simulation because of the diffusion of the magnetic field across the stars surfaces , that fills the entire domain with vacuum electromagnetic fields . as remarked already in sec .
[ sec : rmhdatmo ] and [ sec : spd ] , these magnetic fields are in areas which are treated as atmosphere from a hydrodynamical point of view , but where the maxwell equations are solved with zero conductivity , so that the electromagnetic fields can propagate freely .
we note that a nominal value of @xmath112 does not imply that the electromagnetic luminosity will be zero , since the motion of the compact stars will introduce perturbations in the external electric and magnetic fields , and thus a net poynting flux ( see ref .
@xcite for the electromagnetic emission of inspiralling binary black holes in electrovacuum ) .
indeed , we have found that the electrovacuum luminosity before the merger is @xmath230 , which is smaller than the one reported in ref.@xcite ( i.e. , @xmath231 ) , where the stellar exteriors were modelled in the force - free approximation .
although it has already been found that the electrovacuum luminosity is slightly smaller than the force - free one for the same system ( see refs .
@xcite ) , the differences found here are larger than expected and this may be due to the rather different way in which the exterior regions of stars are treated .
by contrast , the electromagnetic luminosity before the merger in the imhd simulation , where the magnetic fields are always contained inside the stars , is essentially zero . after the merger , the electromagnetic luminosity grows rapidly of about 2 orders of magnitude , essentially as a result of the growth of the magnetic field already discussed in the left panel of fig .
[ fig : bns : mag ] ( we recall that the electromagnetic luminosity should scale quadratically with the magnetic field ) . during the postmerger phases , the luminosity ranges from @xmath232 to @xmath233 , to reach values up to @xmath234 after the collapse of the hmns to a black hole . in the left panel of fig .
[ fig : bns : poynting ] , the luminosity computed on a surface of radius @xmath235 ( black dot - dashed line ) does not overlap with those computed on larger radii ( dark - blue solid line and light - blue dashed solid lines ) , signalling that this radius is too close to the central object and contaminated by the presence of matter .
fortunately , however , the luminosities at @xmath236 and @xmath237 are very close to each other , confirming the robustness of these measurements . by contrast , the three luminosities in the imhd simulation reported in the right panel of fig .
[ fig : bns : poynting ] provide three different values for the luminosity , indicating that at least two of them ( i.e. , those at the smaller extraction radii ) are probably contaminated by the presence of bound matter and hence not reasonable .
we have presented general - relativistic simulations of the inspiral and merger of binary neutron stars when evolved solving the coupled set of the einstein equations and those of rmhd .
our main interest here has been to assess the impact that resistive effects have on the dynamics of these binaries , and which are usually investigated in the more idealized framework of imhd .
because the differences with an imhd description could be rather small in certain stages of the process ( e.g. , during the inspiral ) , we have carried out a close comparison between two simulations evolving the same binary , either in the context of rmhd or in that of imhd .
more specifically , we have studied the dynamics of an equal - mass binary system of neutron stars with a total adm mass @xmath123 and an initial orbital separation of @xmath124 .
the stars are initially irrotational and with zero magnetic field .
a dipolar magnetic field is therefore added before the evolution is started , which is entirely contained inside the stars , at least initially .
furthermore , to reduce computational costs and `` accelerate '' the inspiral , we have slightly reduced the linear momenta of the initial data as done in ref .
@xcite , so that the merger occurs after approximately one orbit . a crucial goal of our rmhd approach has been that of attaining a smooth resistive description from the highly conducting , high - density stellar interior , out to regions of very low - density plasma , where the electromagnetic fields decouple from the fluid .
falling between these two regimes is the large amount of high - density , small - velocity material that is ejected during and after the merger by the hmns and , once the latter collapses to a black hole , by the accreting torus .
this material , occupies a large portion of the computational domain and is produced either by the spiral arms launched at the merger , or by the magnetic winding and consequent pressure imbalance in the outer layers of the hmns .
neutrino losses can also be a source of a wind , but we do not model this here . while there are several ways of potentially reaching a smooth transition between the imhd limit in the stellar interior and an electrovacuum behavior , we have here adopted the same approach we have extensively investigated with isolated neutron stars in ref .
in essence , this matching is achieved through a carefully chosen conductivity profile , where the conductivity is directly related to the conserved rest - mass density and is set to zero once the latter reaches a value close to the atmospheric floor .
this prescription has at least two free parameters .
first , they ensure that the transition region covers only a thin layer close to the surface of the star .
second , they guarantee that this layer remains `` thin '' even in the first steps of the evolution , when the outer layers of the star expand due to a nonzero pressure in the atmosphere . while we have set these two parameters to sensible values , their influence on the results still needs to be fully explored .
overall , we have found that there are many similarities between the rmhd and imhd evolutions , but also one important difference , namely , that the survival time of the hypermassive neutron star , which increases in a rmhd simulation .
the increased lifetime of the hmns appears to be due to a less efficient magnetic - braking mechanism in the resistive regime , in which matter can move across magnetic - field lines , so that the outward transport of angular momentum is reduced . this interpretation is supported by the analysis of the evolution of the specific angular momentum , and it shows that the transport is more efficient in the imhd simulation .
an extended lifetime of the hmns could have intriguing astrophysical consequences , since a longer - lived magnetized hypermassive neutron star brings support to the recent modelling of sgrbs in terms of long - lived magnetarlike objects produced by the merger @xcite .
another important result of these simulations is the confirmation that a magnetic - jet structure is formed in the low - density funnel produced by the black hole torus system .
we note that these simulations have been carried out at higher resolutions and with a different grid structure than those in ref .
@xcite . in the rmhd simulations
the magnetic - jet structure is far more regular , essentially axisymmetric , and extending out to the largest scale in our system .
this is most likely the result of the effective decoupling established between the dynamics of the plasma and that of the electromagnetic fields . in the imhd simulations , a magnetic - jet structure is still present , but on the scale of the torus
this difference is due to the fact that a decoupling of the electromagnetic fields from the plasma is not possible in the imhd approximation , and the magnetic field follows tightly the turbulent dynamics of the matter . in this case , the magnetic - field lines are almost parallel to the @xmath156-axis ( in analogy with what was shown in ref .
@xcite ) and the topology becomes more turbulent on large scales . in both regimes , the magnetic field is predominantly toroidal in the highly conducting torus and predominantly poloidal in the nearly evacuated funnel , although in the imhd simulation , this happens near the rotation axis .
the matter in the funnel does not have an internal energy sufficiently large to launch a relativistic outflow .
however , it is reasonable to expect that reconnection processes or neutrino annihilation occurring in the funnel , none of which we model here , could potentially increase the internal energy in the funnel .
the final comment of this work is in fact a caveat .
while the dynamics of the magnetic - field results presented here appears reasonable , matching the expectations for this type of system as well as previous simulations , we should remark that our results are ultimately dependent on the choice made for ohm s law and for the conductivity profile .
again , while our choice is a very conservative and a plausible one , it represents a choice nevertheless .
the large computational costs associated with these simulations have prevented us from presenting a systematic investigation of how sensitive the results are on the choices for ohm s law , for the conductivity profile , or on the treatment of the atmosphere .
all of these issues deserve further investigation and will be the focus of our future work .
we thank w. kastaun for the visualization library used for the two - dimensional plots and are grateful to b. mundim and k. takami for their valuable comments on the manuscript .
we appreciate useful discussions with n. andersson , i. hawke and c. palenzuela regarding ideas and further applications of resistive - mhd in astrophysical scenarios .
finally , we thank b. giacomazzo for help with the initial data and the ideal - mhd code .
partial support comes from the science and technology facilities council stfc , grant no .
st / h002359/1 , the dfg grant no .
sfb / transregio 7 , the `` newcompstar '' cost action mp1304 , the loewe - program in hic for fair .
the simulations were performed on supermuc at lrz - munich , on datura at aei - potsdam , on the dirac bluegene / q cluster at university of edinburgh - acsp33 , and on the loewe at csc - frankfurt .
although the imhd simulations are not the focus of this paper , for completeness we provide in this appendix a rapid overview of the properties of the magnetic - jet structure as obtained within this approximation .
the essence of the results is shown in fig .
[ fig : rho_and_b_xz_large.imhd ] , which represents the equivalent of fig .
[ fig : rho_and_b_xz_large ] but within imhd .
the different panels show large - scale , two - dimensional snapshots on the @xmath146 planes of the rest - mass density ( top row ) , and of the magnetic field ( bottom row ) . the left column of fig .
[ fig : rho_and_b_xz_large.imhd ] refers to @xmath238 ( left column ) , when the hmns has not yet collapsed to a black hole , while the right column refers to @xmath239 , when a black hole has already been formed . because in the imhd approximation the magnetic fields are tightly locked with the matter , it does not come as a surprise that no ordered magnetic field structure seems to develop before the hmns collapses and forms a black hole .
this is because the dynamics of the plasma is quite turbulent at the merger and during the hmns stage .
however , after a black hole is formed , a well - ordered magnetic - field structure appears as the system reaches a quasistationary state .
we note again that the formation of a magnetic - jet structure occurs around the black - hole rotation axis . differently from the corresponding rmhd simulation ,
the magnetic - jet structure here is not very regular on large scales , and it is necessary to go down to the length scale of the torus , as shown in fig . [
fig : rho_and_b_xz.imhd ] , for the magnetic - jet structure to become evident .
note that the magnetic - field lines are almost parallel to the @xmath156-axis , in analogy with what was shown in ref .
@xcite . finally , although we are here using only the projection of the magnetic - field lines on the @xmath146 plane , the magnified view in fig .
[ fig : rho_and_b_xz.imhd ] reveals that the magnetic field in the low - density funnel is still predominantly poloidal , although not as ordered as in the rmhd simulation ( cf . , fig . [
fig : bns : rhob6 ] , left and right panels ) . at the same time , and not shown here for compactness , the magnetic field is essentially toroidal in the torus .
|
we have studied the dynamics of an equal - mass magnetized neutron - star binary within a resistive magnetohydrodynamic ( rmhd ) approach in which the highly conducting stellar interior is matched to an electrovacuum exterior . because our analysis is aimed at assessing the modifications introduced by resistive effects on the dynamics of the binary after the merger and through to collapse , we have carried out a close comparison with an equivalent simulation performed within the traditional ideal magnetohydrodynamic approximation .
we have found that there are many similarities between the two evolutions but also one important difference : the survival time of the hypermassive neutron star increases in a rmhd simulation .
this difference is due to a less efficient magnetic - braking mechanism in the resistive regime , in which matter can move across magnetic - field lines , thus reducing the outward transport of angular momentum .
both the rmhd and the ideal magnetohydrodynamic simulations carried here have been performed at higher resolutions and with a different grid structure than those in previous work of ours [ l. rezzolla , b. giacomazzo , l. baiotti , j. granot , c. kouveliotou , and m. a. aloy , astrophys .
j. letters 732 , l6 ( 2011 ) ] , but confirm the formation of a low - density funnel with an ordered magnetic field produced by the black hole torus system . in both regimes
the magnetic field is predominantly toroidal in the highly conducting torus and predominantly poloidal in the nearly evacuated funnel .
reconnection processes or neutrino annihilation occurring in the funnel , none of which we model , could potentially increase the internal energy in the funnel and launch a relativistic outflow , which , however , is not produced in these simulations .
|
the analysis of energy spectra of different physical bounded systems has always been an interesting subject in quantum mechanics .
the energy level spacings as well as the existence of degeneracies ( crossings ) in such systems have been widely studied during recent years .
the number of degrees of freedom , the separability of the problem and the number of free parameters involved , among others , are very important features which have to be taken into account when a given quantum spectrum is analyzed .
starting from percival s ideas @xcite for systems with more than one degree of freedom , two kinds of spectra have been distinguished .
the regular spectrum , whose level spacing is characterized by a poisson distribution @xcite , is associated with integrable problems ( the harmonic oscillator is an exception ) .
the other one , the irregular spectrum closely characterized by spectral statistic of the gaussian ensembles , corresponds to non - integrable systems @xcite .
this statistical behavior can be related to crossings and repulsions between levels with the same symmetry when the spectra are analyzed as a function of the parameters of the problem .
in fact , integrable systems depending on one parameter exhibit many crossings while non - integrable systems show repulsions and double - hyperbola curves ( avoided crossings ) , rather than degeneracies .
so , for one parameter dependent systems , the presence of crossings or avoided crossings are quantum fingerprints of the properties of the classical dynamics .
when the system depends on two parameters , the surfaces of energy ( that define in such cases the eigenenergies ) will cross in continuum curves leading to diedric intersections when the system remains integrable or they will intersect at isolated points ( `` diabolical points '' ) if the systems is not integrable ( assuming that it has time reversal symmetry)@xcite .
distributions of avoided crossings ( for one parameter dependent systems ) according to properties such that the closest approach or the mean and difference between the slopes of the involved levels were already established for a generic quantum system employing parametric random - matrix models @xcite . the same models , extended to two and more parameters , were used to study distributions of diabolical points @xcite and other relevant distributions of singular points in the spectra @xcite . on the other hand
, it is well known that classical periodic orbits are essential elements to develop semiclassical quantization methods . not only for integrable systems ( through the berry - tabor formula @xcite ) but also for the non- integrable ones ( using the gutzwiller methods @xcite )
the spectral density can be formally described in terms of closed orbits of the classical system .
therefore , it would not be surprising that other quantum densities such as densities of degeneracies or densities of avoided crossings can be related to the periodic orbits . in the present paper ,
we find the density of degeneracies as an expansion in terms of the periodic orbits for classical integrable systems depending on one parameter . in section [ densi ]
we introduce the density of degeneracies .
the semiclassical version can be written as a sum of a smooth part and oscillating contributions depending on the periodic orbits of the classical system .
section [ suave ] is devoted to compute the smooth part of the density of degeneracies and related distributions for two specific integrable systems whose hamiltonians depend on a single parameter in a different functional form .
we study the rectangular billiard of sides @xmath0 and @xmath1 where the parameter is the ratio @xmath2 ( shape parameter ) and the aharonov - bohm cylindrical billiard where the parameter is the magnetic flux , that is @xmath3 .
the oscillating contributions are computed in section [ oscpart ] .
finally section [ cr ] is devoted to concluding remarks .
we have included an appendix that contains the appropriate derivations of the semiclassical density of degeneracies when the quantum system has not kramer s degeneracies ( like the aharonov - bohm cylindrical billiard ) .
we consider an integrable system whose hamiltonian depends on a single parameter @xmath4 .
that is : @xmath5 where @xmath6 and @xmath7 are the action variables . to obtain the eigenvalues
, we can employ the e.b.k .
semiclassical quantization rule @xmath8 where @xmath9 and @xmath10 are the maslov index @xcite .
so , we establish the quantum eigenenergies @xmath11 by @xmath12 when we consider the eigenenergies as a function of the parameter @xmath4 , the spectrum shows degeneracies ( crossings ) .
they occur whenever @xmath13 given @xmath14 and @xmath15 , this equation determines the values of the parameter @xmath4 for which the eigenenergies labeled by @xmath14 and @xmath15 are degenerated , @xmath16 we define the density of degeneracies @xmath17 as the number of crossings that occurs in the energy interval @xmath18 $ ] and in the parameter interval @xmath19 $ ] . therefore , using the e.b.k .
rule , we can write @xmath17 as follows : @xmath20 where we disregard as in ref.@xcite possible degeneracy factors . employing the poisson sumation formula in eq .
( [ rho ] ) we write , @xmath21 } \nonumber \\ & & \times \int_{\vec{i } \geq 0 } \int_{\vec{i}^{\prime } \geq 0 } d^{n}i \ ; d^{n}i^{\prime } \delta\left ( e - h(\vec{i},\mu ) \right ) \delta\left(\mu - l(\vec{i},\vec{i}^{\prime})\right ) \exp{\left [ i \frac{2 \pi}{\hbar } ( \vec{m } \cdot \vec{i } + \vec{m}^{\prime } \cdot \vec{i}^{\prime } ) \right ] } \ ; . \end{aligned}\ ] ] in the following , to make the expressions more handled , we assume two degrees of freedom . to eliminate the @xmath22-functions , we change the integration variables as follows , @xmath23 where we have defined @xmath24 therefore @xmath25 the partial derivative @xmath26 is defined by the implicit equation @xmath27 and we obtain @xmath28 where @xmath29 is the partial derivative evaluated in @xmath30 .
after integration over @xmath31 and @xmath32 follows @xmath33 } \nonumber \\ & & \times \int_{i_{1 } \geq 0 } \int_{i_{1}^{\prime } \geq 0 } di_{1 } di_{1}^{\prime } \frac { \left| \partial_{\mu}{h}-\partial_{\mu}{h^{\prime } } \right| } { \omega_{2 } \omega_{2}^{\prime } } \exp{\left [ i \frac{2 \pi}{\hbar } ( \vec{m } \cdot \vec{i } + \vec{m}^{\prime } \cdot \vec{i}^{\prime } ) \right ] } \ ; , \end{aligned}\ ] ] where @xmath34 the first term in the eq.([2d ] ) ( with @xmath35 ) corresponds to replace in expression eq .
( [ rho ] ) the quadruple sum by the quadruple integral .
it is a smooth function ( i.e. non oscillating ) of @xmath11 and @xmath4 , @xmath36 starting from eq .
( [ smooth ] ) we can determine other relevant distributions .
as an example , we can obtain the smooth part of the distribution of crossings @xmath37 according to the difference between the slopes of the levels , @xmath38 to obtain such distribution ( in the following dcds ) , we perform the change of the integration variables in eq .
( [ smooth ] ) @xmath39 therefore @xmath40 where the @xmath41 s must be considered as a function of @xmath42 and @xmath4 . leaving out the integration over @xmath43 , follows @xmath44 we stress that the dependence on @xmath43 in eq.([emuv ] ) is not only given by the prefactor @xmath43 but also by the integrand and the limits of integration in the integral over @xmath45 .
the other terms in eq.([2d ] ) contain oscillating functions and we utilize the stationary phase technique to evaluate them .
the conditions of stationary phase lead to the periodicity conditions , namely @xmath46 because @xmath47 s ( @xmath48 s ) are positive , the sums over @xmath49 s ( @xmath50 s ) in eq.([2d ] ) are restricted to the first and the third quadrants .
the numbers @xmath51 define the topology of the periodic orbits .
however , there could be pairs of periodic orbits that having the same topology , they are related to each other through time reversal transformation ( non - self retracing orbits ) .
for the terms @xmath52 ( @xmath53 ) we perform first the integration over the variable @xmath54 and next we evaluate the second integral using the stationary phase approximation .
finally , we obtain @xmath55 \cos{\left(s(m_{1},m_{2})/\hbar + \theta(m_{1},m_{2})\right)}\ ] ] where we have defined @xmath56 @xmath57 are the values of the actions given by eq .
( [ perio ] ) . on the other hand , for @xmath58 or @xmath59 and @xmath60 or @xmath61 we obtain @xmath62
in the present section we compute the smooth part of the density of crossing starting from expression eq.([smooth ] ) for two specific systems .
we will consider two kinds of billiards whose hamiltonian depend on the parameter @xmath4 in a different way . at first we study the well known rectangular billiard .
that is a spinless particle in a two dimensional rectangular box of sides @xmath0 and @xmath1 . the hamiltonian in terms of the action variables is @xmath63 in this case we will consider the crossings as a function of the shape parameter @xmath2 .
we fix the area of the box @xmath64 as a constant to conserve invariant the smooth part of the density of states .
so the hamiltonian can be written as @xmath65 where @xmath66 . in the following
we will use the hamiltonian given by eq.([renor ] ) and we call the corresponding energy as @xmath67 . taking into account that @xmath68 and setting @xmath69 , we obtain , after replacing in eq .
( [ smooth ] ) @xmath70 \ ; , \ ] ] that after elementary integrations leads to @xmath71 to compare this result with the exact one obtained by quantum calculation , we integrate the above density of crossings over the parameter @xmath4 . in this way
we establish the mean number of crossing in the finite interval @xmath72 $ ] with energies between @xmath73 and @xmath74 .
@xmath75 which is a linear function of @xmath73 .
figure [ figure1 ] shows the predictions of the preceding result for @xmath76 and @xmath77 ( solid line ) and @xmath78 ( dashed line ) together with the exact calculations .
now we apply eq .
( [ emuv ] ) to obtain the dcds .
we define the relative difference of slopes as @xmath79 where @xmath80 and @xmath81 is the highest value of @xmath43 for a given value of the energy @xmath73 and the parameter @xmath4 ( in the present example @xmath82 ) . by changing variables in eq .
( [ suacua ] ) @xmath83 and leaving out the integration over @xmath84 , we obtain @xmath85 where @xmath86 therefore , the joint distribution eq .
( [ jointca ] ) results factorizable as @xmath87 .
this means that the distribution of crossings according to the relative difference between the slopes of the levels is a global property of the system .
it holds for any region in the plane @xmath88 .
equation ( [ g1 ] ) is normalized to one and it gives the fraction of crossings whose relative difference of slopes is in the interval @xmath89 $ ] and it is independent on the values of @xmath73 and @xmath4 .
for example , the distribution of the number of crossings in the intervals @xmath90 $ ] and @xmath91 $ ] according to the difference of the slopes is @xmath92 that is , if we perform an histogram according to the relative difference between the slopes for all the crossings in the intervals @xmath90 $ ] and @xmath91 $ ] , the smooth part of such histogram will be given by eq .
( [ g1 ] ) .
figure [ figure2 ] shows the distribution eq .
( [ g1 ] ) together with the histogram resulting from the exact quantum calculation computing the relative difference between the slopes of the crossings that occur in the region @xmath93 and @xmath94 ( @xmath95 crossings ) . as a second example
, we consider the aharonov - bohm cylindrical billiard .
that is a spinless particle confined in a two dimensional cylindrical shell of height @xmath0 and radius @xmath96 axially threaded by a confined magnetic flux @xmath97 .
the hamiltonian is ( taking @xmath45 non negative ) @xmath98 that can be rewritten as @xmath99 where @xmath100 , @xmath101 and @xmath102 .
as in the first example we define the energy @xmath103 .
now , we will consider the crossings as a function of the normalized magnetic flux @xmath97 .
that is we set @xmath3 .
taking into account that @xmath104 and setting @xmath69 , we obtain : @xmath105 \;.\end{aligned}\ ] ] to derive eq.([flux ] ) it is necessary to take into account in eq.([rho ] ) , the effect of the breaking of the time reversal invariance ( see appendix ) . after performing the integrations , eq.([flux ] )
gives @xmath106 this smooth density of crossings is independent on the parameter @xmath4 ( the flux ) .
therefore the number of crossing between @xmath73 and @xmath107 per unit of flux is given by eq .
( [ cilcru ] ) .
figure [ figure3 ] shows @xmath108 given by eq .
( [ cilcru ] ) and the exact quantum calculation for a system with @xmath109 . starting from eq .
( [ flux ] ) , it is not difficult to establish the dcds
. let distinguish the crossings according to the relative sign between the slopes .
we label by a plus sign @xmath110 ( minus sign , @xmath111 ) the crossings between levels with equal ( different ) sign of their slopes
. we can discriminate in expression eq .
( [ flux ] ) the contributions from both kind of crossings .
the first and the second integral in the square bracket correspond to crossings with the same sign of the slopes while the third integral corresponds to crossings with different sign .
thus we find @xmath112 although the smooth part of the density of crossings depends on the energy @xmath73 and on the shape parameter @xmath113 , the fractions of each kind of crossings are the same for all cylinders and they are independent on the energies . to establish the dcds we proceed as follows .
we define the relative jump of the slopes as : @xmath114 for crossings with same sign of the slopes and @xmath115 for crossings with different sign of the slopes . in this way @xmath84 results @xmath116 for crossings labeled @xmath110 while it is @xmath117 for crossings @xmath111 . thus by changing variables in eq .
( [ flux ] ) and leaving out the integration over @xmath84 we obtain @xmath118 where @xmath119 equation ( [ dislope ] ) gives the mean number of crossings per unit of flux that occur in the interval of energy @xmath120 $ ] and such that the relative difference between the slopes lies in the interval @xmath89 $ ] per unit of flux . for @xmath121
there are two contibutions .
the first integral corresponds to crossings with the same sign of the slopes and the second one corresponds to crossings between levels with different sign of the slopes . for @xmath122 , only crossings with different sign of the slopes can occur .
let us remark that the joint distribution @xmath123 results factorizable ( as in the rectangular billiard ) @xmath124 with @xmath125 as in the rectangular billiard , distribution eq .
( [ gv ] ) holds irrespective of the values of the flux and the energy .
therefore , the distribution of crossing below an energy @xmath73 according to the value of @xmath84 will be given by @xmath126 figure [ figure4 ] shows @xmath127 predicted by eq .
( [ gvv ] ) together with the exact quantum results obtained computing the crossings for @xmath128 between @xmath129 ( @xmath130 crossings ) .
we have discriminated the crossings according to the same sign ( @xmath131 , dashed line ) or different sign ( @xmath132 , solid line ) between the slopes of the levels .
we remark that there are @xmath133 crossings @xmath110 and @xmath134 crossings @xmath111.these results are consistent with eq .
( [ fracci ] ) that implies @xmath135
in this section we calculate the oscillating contributions to the density of crossings eq .
( [ osmain ] ) . they are determined by the periodic orbits that satisfy condition eq .
( [ perio ] ) . given the hamiltonian eq .
( [ hamilr ] ) the periodic orbits are classified according to the topology @xmath136 being @xmath137 the number of bounces on a side of length @xmath138 before the periodic orbits are closed .
taking into account eq .
( [ hamilr ] ) and the requirement eq .
( [ perio ] ) , we have : @xmath139 and @xmath140 therefore from eqs .
( [ action ] ) and ( [ theta ] ) follows : @xmath141 on the other hand , using expression eq .
( [ accior ] ) , we have @xmath142 we replace eqs .
( [ acpere ] ) , ( [ srec ] ) and ( [ derirec ] ) in eq . ( [ osmain ] ) and performing elemental integrations we obtain the oscillating contributions : @xmath143 \nonumber \\ & & \times \cos{\left ( 2\pi \sqrt{(\frac{m_{1}^{2}}{\mu}+m_{2}^{2 } \mu ) \epsilon } -\frac{\pi}{4 } \right ) }
\ ; .\end{aligned}\ ] ] on the other hand , replacing eqs .
( [ acpere ] ) , ( [ srec ] ) and ( [ derirec ] ) in eq .
( [ gral ] ) we obtain @xmath144 unlike the smooth part eq .
( [ smoothr ] ) , both contributions involve information about the individual crossings .
the calculation of @xmath145 implies a quadruple sum over @xmath146 while @xmath147 is a double sum .
however if we consider only @xmath147 we have a satisfactory resolution of the individual crossings in the plane @xmath88 as can be seen in fig .
[ figure5 ] .
this figure shows a region in the plane @xmath88 where there are three crossings . in fig .
[ figure5](a ) the exact quantum levels are plotted .
figure [ figure5](b ) shows the oscillating part eq .
( [ osmain ] ) considering @xmath148 and @xmath149 up to 150 . in this example
, the periodic orbits are labeled by the number of rotations around the axis of the cylinder ( @xmath148 ) and the number of bounces on the base of the cylindrical surface ( @xmath149 ) before the periodic orbits are closed .
taking into account that @xmath150 and @xmath151 we obtain ( for details see the appendix ) : @xmath152^{1/4 } \left [ \frac{m_{1}}{\sqrt{m_{1}^2+\gamma m_{2}^2 } } \arcsin{\left(\frac{m_{1}}{\sqrt{m_{1}^2+\gamma m_{2}^2 } } \right ) } \right .
\nonumber \\ & & \left.+\frac{\sqrt{\gamma } m_{2}}{\sqrt{m_{1}^2+\gamma m_{2}^2 } } \right ] \nonumber\\ & & \times \cos{\left ( 2\pi \sqrt{\frac{\epsilon}{\gamma}(m_{1}^{2}+ \gamma m_{2}^{2})}-\frac{\pi}{4 } \right ) } \cos{(2\pi m_{1 } \mu ) } \ ; , \end{aligned}\ ] ] where @xmath153 if @xmath154 and @xmath155 , and @xmath156 \nonumber \\ & & + \left [ \frac{m_{1}}{(m_{1}^2+\gamma m_{2}^2 ) } + \frac{m_{1}^{\prime}}{(m_{1}^{\prime 2}+ \gamma m_{2}^{\prime 2})}\right ] \nonumber \\ & & \times \left [ \cos{\left(2\pi \left(\sqrt{\frac{\epsilon}{\gamma } ( m_{1}^{2}+\gamma m_{2}^{2})}+ \sqrt{\frac{\epsilon}{\gamma } ( m_{1}^{\prime 2}+ \gamma m_{2}^{\prime 2})}-1/4\right)\right)}\cos{\left(2\pi(m_{1}-m_{1}^{\prime})\mu \right ) } \right.\nonumber \\ & & \left.+ \cos{\left(2\pi\left(\sqrt{\frac{\epsilon}{\gamma}(m_{1}^{2}+ \gamma m_{2}^{2})}- \sqrt{\frac{\epsilon}{\gamma}(m_{1}^{\prime 2}+ \gamma m_{2}^{\prime 2})}\right)\right)}\cos{\left(2\pi(m_{1}+m_{1}^{\prime})\mu \right ) } \right ] \ ; .\end{aligned}\ ] ] as in the example of the previous subsection , @xmath147 eq .
( [ osc1ci ] ) , gives enough information to determine individual crossings , as can be seen in fig .
[ figure6 ] . in fig .
[ figure6](a ) is depicted a region of the plane @xmath88 where the exact quantum levels show three crossings .
figure [ figure6](b ) shows the oscillating contribution given by eq .
( [ osc1ci ] ) taking into account @xmath148 and @xmath149 up to 150 . unlike the rectangular billiard , in the present case , the dependence on the parameter @xmath4 of the oscillating contributions is quite simple . therefore eqs .
( [ osc1ci ] ) and ( [ osc2ci ] ) can be easily integrated to obtain the oscillating part of the number of crossings in the energy interval @xmath157 $ ] per unity of flux , @xmath158 as we have pointed out in section [ suave ] , the smooth part of the density of crossings depends only on the energy @xmath73 , so the integrated smooth part ( the first integral in the right hand side of eq . ( [ titi ] ) ) is given by eq .
( [ cilcru ] ) . for the oscillating contributions , the integration over @xmath4 of the terms in eq .
( [ osc1ci ] ) vanishes unless @xmath159 ( because they are proportional to @xmath160 ) .
therefore we obtain : @xmath161 in terms of the classical motion , the contribution eq .
( [ integ1 ] ) is originated by the orbits that having @xmath159 , correspond to an irrotational motion with bounces between the bases of the cylinder .
figure [ figure7](a ) shows the exact quantum calculation of the integrated density of crossings obtained through an histogram .
the sum of the contribution eq .
( [ integ1 ] ) ( up to @xmath162 ) and the smooth part eq .
( [ cilcru ] ) is shown in fig .
[ figure7](b ) .
we can see that the contribution of the oscillating term eq .
( [ integ1 ] ) originates the sharp peaks that are present in the exact calculation . by inspection of eq .
( [ integ1 ] ) , we expect a sharp peak whenever @xmath73 tunes a value such that @xmath163 for all @xmath149 being @xmath164 an integer . for @xmath165
this condition leads to @xmath166 .
therefore we conclude that the peaks correspond to the regions of energy where a `` head - rotational band '' state appears .
such states ( the first state for a given @xmath167 ) are the less affected by the change of the flux ( @xmath4 ) ( they have the smallest slope because they have angular momentum equal to zero ) and they contribute to the density of crossings in a relevant way because they will cross with all the other states that go through ( upward or downward ) this region of energy
. the integration of the terms corresponding to the other oscillating contribution eq .
( [ osc2ci ] ) also vanishes unless @xmath168 .
taking into account these contributions , the integrated density of crossings is shown in fig .
[ figure7](c ) .
in the present paper we have derived a semiclassical expression for the density of degeneracies , @xmath17 that occur in quantum integrable systems depending on a single parameter @xmath4 .
we have applied our results to two specific systems that depend on the parameter in a different functional form , obtaining a quite satisfactory agreement with the exact quantum calculation .
our results show that the density of crossings is strongly dependent on the derivative of the hamiltonian with respect to the parameter . therefore , unlike the density of states ,
even the smooth part of the density of crossings can depend on the energy @xmath11 and on the parameter @xmath4 in a diverse way according to the functional form of @xmath169 .
in particular , the smooth part of @xmath170 for the rectangular billiard , where @xmath4 is the ratio between its sides , results @xmath171 . taking into account that the origin of the crossings between eigenenergies of integrable billiards as a function of a shape parameter has the same and very simple interpretation in terms of geometrical arguments @xcite , we expect that the dependence @xmath171 of the smooth part of the density of crossings , holds for any integrable billiard ( like the elliptic billiard where @xmath4 is the ratio between the axes of the elliptical box or the annular billiard where @xmath4 is the ratio between the outer and inner radii ) . on the other hand , for the bunimovich stadium @xcite , there are numerical evidences that the smooth part of the integrated density of narrowly avoided crossings ( that is the number of narrowly avoided crossings per unit of @xmath11 for a given interval of @xmath4 , being @xmath4 the shape parameter ) , shows a linear dependence on the energy @xmath11 .
this suggests that the integrated distribution of narrowly avoided crossings when @xmath4 is a shape parameter would have the same functional dependence on @xmath11 ( linear ) for other irregular billiards . for the aharonov - bohm cylindrical billiard ,
the smooth part of the density of crossings is @xmath172 without dependence on @xmath3 ( the magnetic flux ) .
such functional form should hold for other aharonov - bohm integrable billiards and this is the case for the aharonov - bohm annular billiard @xcite .
moreover in ref.@xcite it is shown , by numerical computation , that the distribution of spacing in flux between crossings that a given level has , is poissonian .
the present work provides the framework to study properties of the crossing spacing distribution , employing the two point correlation function for the density of degeneracies .
[ [ section ] ] the present appendix is devoted to show how to handle the sum eq .
( [ rho ] ) for the aharonov - bohm cylindrical billiard .
this system having one degree of freedom of rotation , it has not kramer s degeneracies .
that is , conjugate states by time reversal transformation are non degenerated . at first , let us consider the billiard without flux .
the eigenenergies are @xmath173 where @xmath174 is the quantum number related to the rotation around the axis of the cylinder @xmath175 and @xmath176 is the quantum number related to libration motion parallel to axis @xmath177 .
the system has kramers s degeneracy .
a quantum eigenstate @xmath178 and its conjugate @xmath179 have the same eigenenergy if @xmath180 . to take into account these degeneracies in the sum eq .
( [ rho ] ) over @xmath181 we must include a factor @xmath182 in each term such that @xmath183 when the cylinder is threaded by a magnetic flux @xmath97 , the eigenenergies of a conjugate pairs of eigenstates are @xmath184 therefore , conjugated pairs of eigenstates are no longer degenerated .
in such a situation , to preserve the sum eq .
( [ rho ] ) over @xmath181 , we must distinguish two kinds of states according to the dependence on the flux that their eigenenergies have . moreover , eq . ( [ para ] ) which determines the values of the parameter for which the crossings occur , changes according to the same dependence .
thus , eq .
( [ rho ] ) for the cylindrical billiard means @xmath185 where @xmath186 and @xmath187 are the values of the flux determined by the roots of @xmath188 respectively .
now we employ the poisson s formula . the first sum in eq .
( [ rhocil ] ) that takes into account the crossings between energy levels that increase as a function of @xmath97 ( this term includes the crossings between levels @xmath189 and @xmath190 ) can be written as , @xmath191 where we have taken into account that @xmath192 and @xmath193 .
to elimate the @xmath22 functions we employ eqs.(1.7 - 1.18 ) together with the first equation of eq .
( [ vinc ] ) .
we obtain @xmath194 now , we redefine @xmath195 . after replacing it
follow @xmath196 where we have omitted the bar in the integration variables . in the same way , the second sum in eq .
( [ rhocil ] ) ( using the second equation of eq .
( [ vinc ] ) ) and the third and fourth ( using the third equation of eq .
( [ vinc ] ) ) are @xmath197 @xmath198 @xmath199 equation ( [ con ] ) takes into account the crossings between energy levels that decrease as a function of @xmath97 while eqs .
( [ con+- ] ) and ( [ con-+ ] ) correspond to crossings between levels that increase and decrease as a function of @xmath97 .
for the smooth part , we set @xmath35 in the sum of eqs .
( [ con++]- [ con-+ ] ) and taking into account that the lower limit in the integration over @xmath200 can be taken as 0 ( because @xmath201 implies @xmath202 and @xmath203 ) we obtain eq .
( [ flux ] ) . to obtain the first contribution to the oscillating part @xmath204 eq .
( [ osmain ] ) , we set @xmath205 and we integrate over @xmath45 .
then , we approximate the integration over @xmath206 using the stationary phase thechnique obtaining eq .
( [ osc1ci ] ) . for the second contribution
@xmath207 we evalute both integrals ( over @xmath45 and @xmath208 ) in the phase stationary condition .
the resulting eq .
( [ osc2ci ] ) is long but straightforward .
however we will remark some steps . after replacing the phase stationary condition for the actions and reducing the sums over @xmath209 and @xmath210 to the positive quadrants we will have in eqs .
( [ con++ ] ) and ( [ con ] ) terms like ( we write the prefactors as @xmath211 ) , @xmath212 respectively .
these terms can be appropriately combined using trigonometric identities in the form @xmath213 \;.\ ] ] in the same way the terms of eqs .
( [ con+- ] ) and ( [ con-+ ] ) result @xmath214 therefore the sum of these terms is @xmath215 \;,\ ] ] and eq .
( [ osc2ci ] ) follows .
let us remark that eqs .
( [ desdo1 ] ) and ( [ desdo2 ] ) are a consequence of the breaking of the time reversal symmetry when the magnetic flux is present .
in such a case , twin classical orbits ( related to each othe by the time reversal transformation ) that have the same action when the systems has the time reversal symmetry split their actions when the flux is applied @xcite .
this work was partially supported by ubacyt ( tw35 ) , pict97 03 - 00050 - 01015 and conicet .
+ 99 i. c. percival , _ j. phys .
b : at . mol .
phys . _ * 6 * , l229 ( 1973 ) . m. v. berry and m. tabor , _ proc
. r. soc . a _ * 356 * , 375 ( 1977 ) .
t. pechucas , _ phys .
lett . _ * 51 * , 943 ( 1983 ) .
a. bohigas , m.j.giannoni and c.schmit , _ j.physique.lett .
_ * 45 * , l1015 ( 1984 ) .
t. yukawa , _ phys .
lett . _ * 54 * , 1883 ( 1985 ) .
m. v. berry and m. wilkinson , _ proc .
a _ * 392 * , 45 ( 1984 ) . m. wilkinson , _
j. phys . a : math .
* 21 * , 1173 ( 1988 ) .
m. wilkinson , _
a : math . gen . _ * 22 * , 2795 ( 1989 ) .
m.wilkinson and e.j.austin , _ phys .
a _ * 47 * , 2601 ( 1993 ) .
p. walker , m.j.snchez and m.wilkinson , _ j.math.phys .
_ * 37 * , 5019 ( 1996 ) .
m. v. berry and m. tabor , _ proc .
a _ * 349 * , 101 ( 1976 ) .
m. c. gutzwiller _ chaos in classical and quantum mechanics _
( springer , new york,1990 ) .
a.j.s.traiber , a.j.fendrik and m.bernath , _
j.phys.a:math.gen._*22 * , l365 ( 1989 ) .
m.j.snchez , e.vergini and d.wisniacki , _ phys.rev . e _ * 54 * , 4812 ( 1996 ) . a.j.fendrik and m.j.snchez , submitted to epjb , cond - mat/9811272 .
o.bohigas , m.j.giannoni , a.m.ozorio de almeida and c.schmit , _ nonlinearity _ * 8 * , 203 ( 1995 ) .
|
the spectrum of eigenenergies of a quantum integrable system whose hamiltonian depends on a single parameter shows degeneracies ( crossings ) when the parameter varies .
we derive a semiclassical expression for the density of crossings in the plane energy - parameter , that is the number of crossings per unit of energy and unit of parameter , in terms of classical periodic orbits .
we compare the results of the semiclassical formula with exact quantum calculations for two specific quantum integrable billiards .
-0.25 in 6.75 in +
|
trapping and manipulating ultracold neutral atoms in an array of optical potentials opened a vast new research area in the field of quantum fluids and solid state modeling @xcite . demonstrating the superfluid to mott insulator quantum phase transition of interacting bosons in an optical lattice laid the foundation to study and simulate strongly correlated quantum many - body systems with a big impact on condensed matter physics @xcite .
ultracold atoms in optical lattices are now one of the most flourishing directions of experimental and theoretical quantum physics @xcite . besides fundamental physics issues on entanglement , measurement and decoherence @xcite , more and more applications appear for the quantum simulations for a wide range of puzzling effects in solid state physics .
a bec of a dilute bosons , in which a very large number of atoms occupy the same quantum state of motion , was experimentally realized by several groups @xcite .
optical control of the quantum state of such bec opened the door to a new generation of experiments @xcite .
in particular the loading of the bec into a standing wave formed by pairs of counter propagating far detuned laser beams creates a synthetic controllable periodic lattice @xcite .
trapped atoms still can tunnel from site to site , while interacting and repelling each other when occupying the same site .
the stronger the laser field , the deeper the lattice and the slower the hopping rate of the atoms , while at the same time the on - site interaction between the atoms becomes stronger .
thus the laser light allows to switch between a weakly interacting gas , where atoms form a superfluid , to a strongly correlated regime , where atoms avoid each other and localize into individual sites forming a perfectly regular array called mott insulator phase .
as this crossover happens at @xmath0 as a function of an external parameter , it is called a quantum phase transition @xcite .
it is theoretically well described by a bose - hubbard model @xcite and soon was experimentally demonstrated @xcite .
an increasing degree of control of experimental parameters in optical lattices has been demonstrated up to the level of addressing and controlling single atoms at a specific lattice site @xcite . in a parallel development
the dynamics of the electromagnetic field modes in a high - q optical cavity resonantly coupled to quantized excitations of atoms has been extensively studied .
many phenomena discussed as gedanken - experiments to exhibit the quantum physics of coupled systems and measurements were actually performed experimentally and lead to the recent 2012 physics nobel prize @xcite . the first experimental realizations of cavity qed used single rydberg atoms in superconducting microwave resonators . later ground state
alkali atoms in high - q optical resonators allowed to reach the strong coupling limit as well , where light matter coupling dominates decoherence @xcite .
other implementations using solid state technology followed more recently @xcite .
it is of course a natural development to combine these two threads and investigate ultracold atomic lattices within optical resonators @xcite . as a major experimental step in this direction ,
several groups @xcite achieved operations of a bec of trapped ultracold atoms within an optical high - q cavity deeply in the strong resonant coupling regime .
it seems thus that no major technical obstacles prevent the implementation of an optical lattice within an optical cavity .
an alternative route for realizing an optical lattice within a cavity is based on a co - planar waveguide microwave resonator resonantly coupled to the collective hyperfine state of a bec of ultracold atoms in a wire surface trap fabricated on a superconducting atom chip @xcite . this system is predicted to allow the experimental study of magnetic dipole interactions in the strong coupling regime @xcite .
placing ultracold atoms within optical cavities leads to a wealth of interesting dynamical phenomena already in the dispersive regime , where the light frequency is far detuned from the atomic resonance .
this involves a dynamic interplay of light forces , non - local couplings and field dynamics @xcite . in this review
instead we concentrate on resonant interaction of quantized light modes and trapped atoms in the mott phase , which we treat as a new type of matter closely related to an atomic crystal with widely controllable lattice geometry @xcite . while in some general cases we consider coupling of the light to motional degrees of freedom as well , our main emphasis is on collective internal excitations dynamics .
this reduced description already allows to study many solid state effects in a very generic form .
the wide and precise controllability of different lattices and coupling parameters allows to change the number of atoms per site , the geometry , and the symmetry of the system .
these facts give the present system a decisive advantage over conventional solid state crystals where parameters are hardly tunable and particular effects can not be easily singled out .
such toy implementations provide to us a more detailed and deeper understanding of solid state effects .
mathematically one extends the common bose - hubbard atom model to include atoms in excited states formally described as a second kind of bosons within the optical lattice . the system is then represented by a two - component bose - hubbard model exhibiting a much richer phase diagram @xcite .
optical lattices allow us to go beyond the weakly interacting regime of bose gas into strongly correlated systems of solid state physics @xcite . including the internal atomic level structure bears a strong analogy to the excitonic dynamics of molecular crystals , described by `` frenkel excitons '' @xcite .
the electronic excitations delocalize in the lattice due to electrostatic interactions , e.g. resonant dipole - dipole interactions . by exploiting the lattice symmetry one is lead to introduce wave - like collective electronic excitations
called `` excitons '' as fundamental system excitations @xcite . for this expansion to make sense
, the transfer time has to be faster than the exciton lifetime , so that collective states of electronic excitations dominate the electrical and optical system properties .
excitons thus provide an ideal basis to describe resonant effects in optical lattices .
in fact the atomic recoil heating due to emission and absorption of a photon is significantly suppressed as the momentum is shared among all atoms in the lattice and strongly suppress diffusion and heating in the mott insulator .
we present studies of the properties and dynamics of excitons for several generic lattice types of different geometries and dimensionality , starting from finite chains , infinite chains to stacks of two dimensional planar lattices @xcite .
boundary conditions of finite systems give rise to standing wave excitons , while extended systems produce propagating wave solutions . as a special case a mott insulator with two atoms per site can also possess dark on - site localized excitons decoupled from the remaining lattice sites @xcite .
the energy delocalization of excitons among many particles amounts to the occurrence of intrinsic entanglement . to generate and preserve significant entanglement
requires sufficient isolation from the environment in order to bring decoherence and dissipation to a minimum , which is a big challenge for optically excited systems
. here spontaneous emission due to the coupling to free space radiation is a major source of dissipation and decoherence @xcite .
excitons in multi - atom systems exhibit significantly modified spontaneous emission compared to a single atom @xcite , which we have extensively studied for different geometries and dimensions @xcite . while often excitons exhibit enhanced coupling to excitation lasers and vacuum modes and fast or superradiant decay , in many cases one can get long lived or even dark and metastable excitations , which can only decay optically into a narrow set of wave - vectors analogous to bragg angles . if the lattice spacing is smaller than the transition wavelength , the exciton wave vector can be larger than the free space photon wave - vector , so that no optical decay is possible .
nevertheless one can still engineer coherent transfer of such excitons among different lattice regions or different optical lattice planes @xcite .
this effect should have important applications in optical information storage and transfer based on large ensembles .
interestingly , while the individual dipole - dipole energy transfer has an inverse cubic dependence on the distance , collective states within one lattice plane only couple with an exponentially decreasing strength to neighboring planes , which generally allows simplification to a nearest neighbor coupling model . tailored light - matter interaction is of great importance to fundamental physics tests as well as high end technical applications @xcite and in particular for the development of quantum information technologies @xcite .
new frontiers of cavity qed involving ultracold atoms trapped in intra - cavity optical lattices form a natural , but only very recently opened , research direction @xcite .
as above , we concentrate here on atoms with an internal electronic transition close to resonance with a cavity mode , while the optical lattice is generated by external classical fields which are off - resonance to any of the internal atomic transitions forming the dynamical part of the system .
the case of cavity generated optical potentials also exhibits a wealth of interesting phenomena @xcite .
here we generalize the two - component bose - hubbard model to include cavity photons and their coupling to electronic excitations of the cold atoms @xcite .
the phase diagram including superfluid and mott regions is affected by the coupling to cavity photons which mediate long range interactions and enhance tunneling via field fluctuations .
even though using a rather simple model we see that the quantum phase transition to a perfectly ordered mott state occurs at much deeper optical lattices than in free space . in cavities , excitons and photons
are coherently coupled to form new quasi - particles called polaritons @xcite .
as the cavity mirrors exhibit a finite transition probability , the detection of the emitted light provides us with detailed information about the system properties @xcite .
the corresponding optical spectra give a largely nondestructive observation tool for different quantum phases with minimal back - action @xcite .
as an additional example the formation of molecules in the optical lattice introduces optical anisotropy in our artificial crystal .
this manifests itself in the spectra through the cavity photon polarization mixing @xcite .
again our studies range from finite lattices inside optical cavities of spherical mirrors up to two dimensional lattices between planar cavity mirrors . as an interesting property for the search of lower temperatures , the elastic scattering of excitons and polaritons by the defects or vacancies in the mott insulator phase
can be observed via cavity transmission spectra without destroying the lattice state as it occurs in high aperture microscope detection via fluorescence @xcite .
similarly atoms excited to higher bloch bands change the dynamics of excitons and polaritons @xcite .
optical properties of a gas of ultracold atoms with coherently dressed two - level and three - level systems in a mott insulator phase of an optical lattice have been studied by @xcite .
cooperative effects of ultracold molecules in optical lattices are also treated @xcite .
recently rydberg atoms become of large interest for optical lattices .
rydberg excitations in bose - einstein condensates of rubidium atoms loaded into quasi - one - dimensional traps and in optical lattices experimentally have been realized @xcite .
moreover , the formation of excitons in a regular and flexible chain of rydberg atoms as introduced through the interplay between excitonic and atomic motion was proposed theoretically by @xcite .
free space optical lattices have been formed in many geometries from one dimensional to three dimensional . in a parallel development
also evanescent fields have been used to trap atoms and create new lattice geometries . in a particularly interesting and potentially important example , tapered optical fibers were used to trap neutral atoms @xcite .
these are coupled to the propagating fiber modes via evanescent fields surrounding the fiber @xcite .
such a hybrid combination of atoms and a solid state devices allows trapping and , simultaneously , optically interfacing @xcite .
fibers with diameters smaller than the wavelength exhibit strong transverse confinement and a pronounced evanescent field , which by using two colors can be designed to create an array of deep optical micro - traps .
it was recently realized with cesium atoms , where it was shown that the atoms can be efficiently trapped and interrogated with a resonant light field sent through the nanofiber @xcite .
this setup provides an ideal test system to study one dimensional excitons , for which we developed a microscopic theory .
the strong collective coupling between atoms and field modes naturally leads to the formations polaritons in this setup @xcite .
as one of the most interesting consequences we could show that one can create polaritons with a strong fiber mode contributions but with virtually no coupling to the free space radiation modes .
this should constitute an ideal photon - atom ensemble interface for coherent quantum information storage and processing .
the review is organized as follows . in section 2
we extend the bose - hubbard model to exploit electronically excited atoms . in section 3
we introduce excitons into a system of ultracold atoms in optical lattices .
we discuss their formation , dispersion and life times for different geometries and dimensionality .
section 4 is opened by discussing the quantum phase transition within a cavity followed by introducing cavity polaritons in the strong coupling regime , where various possible physical realizations are studied .
defects of atoms excited into higher bloch bands or vacancies in the mott insulator phase are treated in section 5 .
the conclusions are contained in section 6 .
optical lattices created by off resonance laser fields constitute a regular array of micro - traps into which a dilute gas of ultracold atoms can be loaded @xcite .
the atoms experience an optical lattice potential with lattice constant of half the wave length of the lattice laser with different depth for each internal atomic state . at ultra low temperatures , @xmath1
, atoms are localized at the lowest bloch band of the lattice . in such structures
, atoms can hop among different sites through quantum tunneling and interact locally via scattering .
the tunneling process is mainly among nearest neighbor sites with the hopping parameter @xmath2 , and the atom - atom interactions are efficient only for atoms at the same lattice site with an effective strength @xmath3 , which is taken to be a repulsive determined by the @xmath4-wave scattering length .
this system is well described by a bose - hubbard model @xcite represented by the hamiltonian @xmath5 where @xmath6 and @xmath7 are the creation and annihilation operators of an atom at site @xmath8 , and @xmath9 is the atom on - site energy , which includes the external trap potential , @xmath10 is the chemical potential accounting for a possible exchange of atoms with an external atomic reservoir .
the @xmath8 summation is over the lattice sites , and the bracket indicates a possible hopping only between nearest neighbor sites .
the bose - hubbard model predicts a quantum phase transition between the superfluid and the mott insulator phase by changing the external laser field intensity , that is changing the kinetic versus the repulsion energy of the quantum gas . in the limit of @xmath11
the atom hopping among the lattice sites dominates the dynamics and the atomic states are spread out over the whole lattice and one gets a superfluid . in the opposite limit @xmath12 , the on - site atom - atom interaction dominates and the atoms localize deterministically at individual sites , so that we have a mott insulator .
the corresponding phase diagram can be calculated in the mean - field approach @xcite . in fig . [ 1 ] the phase diagram is plotted in the plane @xmath13 which is scaled by @xmath14 , where @xmath15 is the number of nearest neighbors .
the superfluid ( sf ) phase appears outside the shaded three mott insulator ( mi ) phase regions , which are for one , two , and three atoms per site , that is @xmath16 , where we assumed @xmath17 . in the mean field approximation
the systems dimensionality enters only through @xmath15 , where in a hyper - cubic lattice for the one dimension @xmath18 we have @xmath19 , for @xmath20 we have @xmath21 , and for @xmath22 we have @xmath23 .
vs. @xmath24 .
the superfluid sf and the three mott insulator mi regions , for @xmath16 , are shown .
figure reprinted with permission from @xcite copyright 2009 by american physical society.,width=321 ] in the present review we study collective electronic excitations of an atomic mott insulator .
hence we have to introduce an excited atom state , which modifies the above phase diagram as we now have two - level atoms treated here as two kinds of atoms , namely ground state atoms and atoms excited to a fixed internal state .
when loaded on an optical lattice due to different polarizability and stark shift of the ground and excited states each group of atoms experiences a different optical lattice potential , but with ground and excited potentials showing extrema at the same positions , as schematized in fig .
[ 2 ] . for certain , magic , laser wave lengths ,
light shift yield nevertheless identical potentials .
this is important e.g. to create an optical lattice clock @xcite .
close to @xmath0 the ground and the excited atoms will again occupy only their lowest bloch band .
we consider the two groups of atoms with different internal states , hyperfine or electronic , as two kinds of bosons .
therefore , we generalize the single - component bose - hubbard model into two - component bose hubbard model @xcite , with the hamiltonian @xmath25 where @xmath6 and @xmath7 are the creation and annihilation operators of a ground state atom at site @xmath8 , respectively , with the on - site energy @xmath26 , and chemical potential @xmath27 ; @xmath28 and @xmath29 are the creation and annihilation operators of an excited state atom at site @xmath8 , respectively , with on - site energy @xmath30 , and chemical potential @xmath31 ; where @xmath32 , with @xmath33 being the effective atomic transition frequency , which can include ac stark - shifts relative to the free atomic transition frequency . here
@xmath34 and @xmath35 are the hopping parameters of ground and excited state atoms , respectively .
@xmath36 and @xmath37 are the on - site ground and excited state atom - atom interactions , respectively , and @xmath38 is the on - site ground - excited atom interaction related to scattering between ground and excited state atoms .
this extended model leads to a much richer phase diagram .
now we present one typical example of the phase diagram , which is calculated in using the mean - field approach . in figures [ 3 ] we plot the phase diagram for the plane @xmath39 vs. @xmath40 , and @xmath41 vs. @xmath42 , which are scaled by @xmath43 and @xmath44 .
we used for the energies @xmath45 and @xmath46 , and for the scaled atom - atom coupling the number @xmath47 . here the excited and ground phase transition lines split , the transition line ( le ) is for excited state atoms , and ( lg ) for ground state atoms . beside the superfluid phase ,
we get three other regions .
the mott insulator region with one ground state and one excited state atom per site , that is @xmath48 .
the sm region , where the excited state atoms are in the mott insulator phase with one atom per site , and the ground state atoms are in the superfluid phase .
the ms region , where the ground state atoms are in the mott insulator phase with one atom per site , and the excited state atoms are in the superfluid phase .
we like to mention here that more complex quantum phases are suppressed under the present mean field theory , e.g. , that of the anti - ferromagnetic type in which the mott insular includes an excited atom at one site and a ground one in the nearest neighbor sites @xcite . vs. @xmath40 , and @xmath41 vs. @xmath42 .
we used @xmath47 , with @xmath45 and @xmath46 , for @xmath48 . beside the sf and the mi regions ,
the sm region is associated to the mott insulator of excited atoms and superfluid phase of ground atoms , and the ms region to the mott insulator of ground atoms and superfluid phase of excited atoms .
the full - line lg determines the sf - mi transition of the ground atoms , and the dashed - line le that transition for the excited atoms .
figure reprinted with permission from @xcite copyright 2009 by american physical society.,width=321 ]
as shown above even for two state atoms one can prepare a perfectly filled mott phase for the atoms , which we will use as the basis of our further considerations involving resonant excitations .
we focus on a single atom per site , but eventually allow for two atoms per site as in subsection 3.5 .
the mott insulator constitutes an artificial crystal with a lattice constant much larger than the average dimension of the atom .
therefore , each atom retains its identity except from the internal shifts and there are no overlaps among electronic wave functions of atoms setting at different sites as in conventional solids .
we conclude that our artificial lattice is similar to molecular or noble atom crystals but allow for the wide controlablity of the system parameters , as the artificial lattice is formed by variable external laser field .
solid molecular crystals , however , are fixed by the chemical bonds , e.g. van der waals forces , and can be only slightly controlled in applying external fields .
frenkel excitons play a central role in optical and electrical properties of molecular crystals @xcite . here
we ask the question related to the possibility of the formation of frenkel like - excitons for optical lattices .
first we discuss the possibility of energy transfer among atoms at different sites induced by electrostatic interactions or resonant dipole - dipole interactions , which is the main mechanism behind the formation of excitons .
lets take the simplest case of two atoms setting on a lattice at nearest neighbor sites such that they are separated by the lattice constant @xmath49 .
the atoms are considered to be two - level systems with transition energy @xmath50 .
if one atom is excited and the second is in the ground state , can this excitation be transferred among the two atoms , in which the excited atom decays to the ground state and the ground state one becomes excited .
here we consider only energy transfer without charge transfer ?
this process is possible through the resonance dipole - dipole interaction @xcite with the energy transfer parameter @xmath51\left(1 - 3\cos^2\theta\right)\right .
\nonumber \\ & -&\left.\frac{\cos\left(q_aa\right)}{q_aa}\left(1-\cos^2\theta\right)\right\},\end{aligned}\ ] ] where @xmath52 is the atomic transition wave number , as @xmath53 , and @xmath54 is the single excited atom damping rate @xcite , which is given by @xmath55 , with @xmath10 the transition dipole . here
@xmath56 is the angle between the transition dipole and vector distance between the two atoms .
the energy transfer is efficient in the limit of @xmath57 , where the electronic excitation can transfer among the atoms before it spontaneously decays , and this limit is achievable up to @xmath58 . for a typical optical lattice
we have , e.g. , @xmath59 , with @xmath60 .
for @xmath61 we get @xmath62 . for @xmath63
we obtain @xmath64 , and for @xmath65 we obtain @xmath66 , in which the energy transfer is possible .
note that in the limit @xmath67 , with efficient energy transfer , the electrostatic interaction limit is applicable @xcite . in this limit and in neglecting the radiative corrections , we get @xmath68 using the previous numbers , @xmath63 yields @xmath69 , and @xmath65 yields @xmath70 , which are slightly different from the above exact results . for smaller @xmath71
we get much better agreement .
figure [ 4](a ) reports @xmath72 as a function of @xmath71 for eqs .
( [ exact ] ) and ( [ appr ] ) at the polarization direction @xmath63 , and in ( b ) for @xmath65 .
the results justify the use of electrostatic dipole - dipole interactions for optical lattices when @xmath67 . vs. @xmath71 , for ( a ) @xmath63 and ( b ) @xmath65 .
the full line is for equation ( [ exact ] ) , and the dashed line for equation ( [ appr ] ) .
figure reprinted with permission from @xcite copyright 2012 by european physical society.,title="fig:",width=245 ] vs. @xmath71 , for ( a ) @xmath63 and ( b ) @xmath65 .
the full line is for equation ( [ exact ] ) , and the dashed line for equation ( [ appr ] ) .
figure reprinted with permission from @xcite copyright 2012 by european physical society.,title="fig:",width=245 ] energy transfer of electronic excitations among close lattice sites induced by dipole - dipole interaction tends to delocalize the excitation coherently in the lattice . exploiting the lattice symmetry we can construct collective excitations in such systems , which are called excitons . in momentum space such de - localized electronic excitations can be represented by propagating excitation waves . in the following we explicitly calculate these excitons for various optical lattices with different geometries and dimensionality .
we concentrate now in the mott insulator phase with one atom per site @xcite .
the atoms as before are assumed to be two - level systems , and the electronic excitation can transfer among atoms at different sites , e.g. among sites @xmath73 and @xmath74 , with the transfer parameter @xmath75 . as discussed above ,
such a process can be efficient in the limit of @xmath76 , where an electronic excitation could transfer before the decay of the excited atom .
the excitation hamiltonian in general can be written in the second quantized form as @xmath77 where @xmath78 and @xmath79 are the creation and annihilation operators of an electronic excitation at site @xmath73 .
the first term is for the on - site excitation and the second for the energy transfer .
we neglect terms that do not conserve the number of excitations .
the operators are of the spin - half type , and they behave on - site as fermions , and as bosons among different sites . but for a single excitation , or for very small number of excitations relative to the number of lattice sites , saturation effects can be neglected , and then the possibility of having two or more excitations to appear at the same lattice atom is negligible @xcite .
therefore , from this point on we assume the excitations to behave as bosons with the commutation relation @xmath80=\delta_{nm}$ ] .
we treat first a long one dimensional optical lattice @xcite , as in fig .
the lattice length is enough long to neglect edge effects , which allows the use of the periodic boundary condition .
the previous hamiltonian can be easily diagonalized in applying the fourier transform @xmath81 where @xmath82 is the number of lattice sites , and @xmath83 is the position of an atom at site @xmath73 . here
@xmath84 is a wave number that in using periodic boundary conditions takes the values @xmath85 , with @xmath86 , and @xmath87 is the lattice length .
this solution holds also for a ring of an atomic chain . in applying the transformation
the diagonal hamiltonian reads @xmath88 where the exciton dispersion is given by @xmath89 .
as the energy transfer is a function of the interatomic distance , @xmath90 , the exciton dynamical matrix is given by @xmath91 .
note that in place of a discrete energy level of the excited atom we get an energy band for the excitons .
, lattice constant @xmath49 , and energy transfer parameter @xmath2.,width=321 ] the single exciton state of wave number @xmath84 is obtained from the vacuum state , @xmath92 in which all atoms are in the ground state , in applying the exciton creation operator , namely @xmath93 , where @xmath94 . here
@xmath95 stands for a state where only atom @xmath73 is excited and all the others are in the ground state .
the single exciton is a coherent state built of a superposition of electronic excitations at all sites , where the probabilities to find an excitation at each site are equal .
this state is a collective electronic excitation and is considered as a wave that propagates to the left or the right in the lattice with a fixed wave number @xmath84 .
a superposition of such states can be used to build a more physical state of a wave packet of excitons .
we assume now energy transfer only among nearest neighbor sites . in the @xmath67 limit , the transfer parameter @xmath2 is given by eq .
( [ appr ] ) and shows a strong dependence on the @xmath56 angle between the directions of the transition dipole and of the lattice .
the transfer parameter changes from being negative ( attractive ) at @xmath96 to positive ( repulsive ) at @xmath97 .
the dispersion relation now reads @xmath98 with a bandwidth of @xmath99 . for small wave numbers , that is in the limit of @xmath100 , or for long wavelength excitons , we get the dispersion @xmath101 , where we defined the exciton effective mass @xmath102
the exciton in this limit is a quasi - particle that propagates in the lattice with a given wave number and an effective mass . at @xmath103
we get @xmath104 , with a shift produced by the dipole - dipole interaction . in this limit of long wavelength excitons , beyond the nearest neighbor interactions
the long range dipole - dipole interaction yields the correction @xmath105 , where @xmath106 @xcite .
such a shift has a large influence on optical lattice clock states @xcite , but can be controlled through the direction of the transition dipole @xcite , and even can be brought to zero at the angle of @xmath107 . , and with one atom per site .
the transition dipole @xmath10 is plotted which makes an angle @xmath56 with the lattice axis . for a finite optical lattice of @xmath82 sites with one atom per site ,
the lattice length is @xmath108,,width=321 ] next we treat a finite optical lattice chain of @xmath82 two - level atoms @xcite , as schematized in fig .
the lattice length is taken to be smaller than the atomic transition wavelength . in considering energy transfer only between nearest neighbor atoms
, the hamiltonian can be diagonalized by introducing standing wave excitons @xmath109 with @xmath110 . as a mathematical trick to find appropriate solutions ,
we add two additional empty sites at the two lattice edges , at @xmath111 and @xmath112 , where the collective electronic excitation wave function vanishes .
the exciton dispersion then reads @xmath113 the single electronic transition energy splits into @xmath82 discrete collective electronic excitation energies .
the modes are divided into two groups , symmetric and antisymmetric modes . in the next section
we will show that the antisymmetric modes are dark , and the symmetric ones are bright .
furthermore the first bright mode is node less and found to be super radiant and dominates the optical properties . .
the directions of the in - plane wave vector @xmath114 and the transition dipole @xmath10 are indicated.,width=264 ] let us consider the case of a two dimensional planar optical lattice @xcite . for simplicity
we treat the case of a square symmetry with lattice constant @xmath49 , and one two - level atom per site , as in fig . .
the position of a lattice site is defined by the in - plane vector @xmath115 , where @xmath116 , with @xmath117 the number of lattice sites .
the electronic excitation hamiltonian is given by @xmath118 which can be diagonalized by applying the transformation @xmath119 to get the diagonal exciton hamiltonian @xmath120 with the energy dispersion @xmath121 where we defined the distance between two sites by @xmath122 and introduced as previously the @xmath2 transfer parameter .
the exciton is now represented by a wave that propagates in the lattice with in - plane wave vector @xmath114 , where we have @xmath123 .
the nearest neighbor interaction yields the dispersion with @xmath124 $ ] , which for small wave numbers that is @xmath100 with @xmath125 , reads @xmath126 , with the exciton effective mass @xmath102 . at @xmath103
we get @xmath127 with the dipole - dipole interaction shift . in the limit of long wavelength excitons , beyond the nearest neighbor interactions
the long range dipole - dipole interaction yields the correction @xmath128 , where in using the ewald s summation method @xcite we get @xmath129 .
the exciton damping rate through the spontaneous emission into free space can be calculated in using fermi s golden rule @xcite and shows significant deviation from a single excited atom @xcite . for a finite atomic chain
the standing wave excitons are divided into two groups of dark and bright modes , which exhibit metastable or superradiant propagating excitons .
let us mention here an important point concerning the atomic recoil effect via the emission of a photon .
as the exciton is a collective electronic excitation delocalized in the lattice , where the exciton momentum is shared among all lattice atoms , the emitted photon inherits the momentum and energy of the exciton , which requires the photon wave vector component parallel to the lattice to be equal to the exciton wave vector . in this process
the whole lattice reacts and absorbs the momentum collectively and not only a single atom .
hence , the recoil heating is significantly suppressed , and we can conclude that excitonic interactions will not heat the lattice significantly and destroy the mott insulator .
we start with a long optical lattice with one atom per site @xcite as in fig [ 7 ] .
we assume that a single exciton is excited in the lattice with a fixed wave number @xmath84 , and a given polarization @xmath56 for the transition dipole direction relative to the lattice axis .
the damping rate of the exciton into the radiation field of the free space is given by @xmath130 where @xmath131 is the previous one dimensional exciton dispersion , e.g. as for the nearest neighbor interactions in eq .
( [ onedispersion ] ) . for some wave numbers @xmath84 and certain angles @xmath56 the damping rate can be much larger than the single excited atom one , as we get a superradiant enhancement analogous to the dicke model .
however , for other @xmath84-s and @xmath56-s the damping rate can be much smaller than the single excited atom one , or even completely vanish . in general the damping expression shows the existence of a critical wave vector @xmath132 that is given by @xmath133 , beyond which the damping rate is zero and the excitations can no longer decay radiatively . for @xmath134 the damping rate at
the critical wave vector @xmath132 is zero , and for @xmath135 the damping rate at @xmath132 is finite and equal to @xmath136 . we conclude that eq .
( [ damping ] ) holds up to the critical wave number , @xmath132 , where beyond @xmath132 no free space photon exists which conserves both energy and momentum component parallel to the lattice . , for ( a ) @xmath96 and ( b ) @xmath137 .
the damping rate is zero beyond the critical wave number @xmath138 .
the dashed - line is for a single excited atom damping rate .
figure reprinted with permission from @xcite copyright 2010 by european physical society.,title="fig:",width=245 ] , for ( a ) @xmath96 and ( b ) @xmath137 .
the damping rate is zero beyond the critical wave number @xmath138 .
the dashed - line is for a single excited atom damping rate .
figure reprinted with permission from @xcite copyright 2010 by european physical society.,title="fig:",width=245 ] we exhibit this behavior in two plots of the damping rate using the following parameters : transition energy @xmath59 , lattice constant @xmath139 , and transition dipole @xmath140 . in fig .
[ 8](a ) we plot as a function of @xmath141 for @xmath134 the damping rate compared to the damping rate of a single atom .
note that for small wave numbers , @xmath142 , the exciton damping rate is much larger than the free atom damping rate , and these states are superradiant states .
with increasing wave numbers the damping rate drops and reaches the single atom rate , and beyond a critical wave number @xmath132 the damping rate becomes zero and no damping is obtained beyond @xmath132 .
the critical wave number is obtained here at @xmath138 . in fig . [ 8](b ) the same plot is for @xmath135 . here
the damping rates are larger than the single atom one for small wave numbers , @xmath142 , but now the damping rate increases with increasing the wave number .
namely , the large wave number states became more superradiant states .
the maximum damping rate is obtained at the critical wave number @xmath132 , and the damping rate is plotted up to that value , as beyond this point the excitons are metastable .
note that excitons can appear up to the boundary of the brillouin zone at @xmath143 , and hence excitons in the region @xmath144 are metastable .
these excitons are not dark , only they are decoupled from the free space photons , but they can be coupled to confined photons , as presented in subsection 4.5 through the coupling to tapered nanofiber photons @xcite .
collective electronic excitation damping rate rapported to that of a single excited atom @xmath84 , for @xmath145 . figure reprinted with permission from @xcite copyright 2012 by european physical society.,width=321 ] the results are drastically changed for a finite one dimensional atomic chain of only a few atoms @xcite which is short relative to the atomic transition wave length . in this case
we obtained that the exciton states are divided into two groups of symmetric and antisymmetric standing waves .
for each collective mode we can define a collective transition dipole , which is assumed to be localized at the center of the lattice , for the limit of short chains .
the antisymmetric excitons with even wave numbers @xmath146 are dark with zero damping rates , that is @xmath147 .
the symmetric excitons with odd wave numbers @xmath148 are bright with the damping rates @xmath149 where the lattice length is @xmath108 , and the exciton dispersion is given in eq .
( [ findis ] ) .
the first mode @xmath150 , the one without nodes , is superradiant , with a damping rate nine times larger than the second bright mode @xmath151 . in fig .
[ 9 ] the collective electronic excitation damping rate rapported to that of a single excited atom , @xmath152 is plotted versus @xmath84 , using the previous parameters , @xmath96 and @xmath145 .
interesting results can be also obtained for planar optical lattices @xcite .
we deal here , as before , with two dimensional large optical lattice with one atom per site and of square symmetry , as plotted schematically in fig .
we calculate the damping rate into free space of an exciton with in - plane wave vector @xmath114 .
now the dipole makes an angle @xmath56 with the normal to the lattice plane , and an additional angle @xmath153 is needed that is between the in - plane component of the dipole and @xmath114 .
the result is @xmath154 where the exciton dispersion is defined in eq .
( [ twodis ] ) , and @xmath155 .
the most important result here is that for @xmath156 we have @xmath157 , where @xmath158 , as beyond the singularity point the damping rate becomes imaginary .
the results should be compared with the radiative damping rate of a single excited atom @xmath54 .
we present numerical results using the typical previous numbers for optical lattices , except for simplicity we neglect the @xmath114 dependence of the exciton energy , negligible compared to the @xmath159 one .
[ 11](a ) shows @xmath160 vs. @xmath159 for @xmath96 , where the transition dipole has only component normal to the lattice plane . here
@xmath161 is @xmath153 independent .
the damping rate starts from zero , where the long wave length excitons are metastable , and increases with @xmath84 to become superradiant with damping rate larger than @xmath54 .
close to @xmath162 the damping rate diverges , and beyond that value it jumps back to zero , all excitons between @xmath132 and the brillouin zone boundary at @xmath163 being metastable . fig . [ 11](b ) shows @xmath160 vs. @xmath159 for @xmath164 , where the transition dipole has equal in - plane and normal components .
we take the in - plane wave vector @xmath114 to be parallel to the in - plane dipole , that is @xmath165 .
very interesting behavior appears here for the damping rate . for small in - plane wave numbers the excitons are superradiant with damping rate larger than @xmath54 , and decreases in increasing @xmath84 till it becomes zero at @xmath166 where excitons around this point are metastable . for larger @xmath84
the damping rate start to increase and diverges at @xmath167 , and then jump back to zero beyond this point . vs. @xmath155 , for ( a ) @xmath96 , which is @xmath153 independent , and for ( b ) @xmath164 with @xmath165 .
here @xmath168 .
figure reprinted with permission from @xcite copyright 2011 by american physical society.,title="fig:",width=245 ] vs. @xmath155 , for ( a ) @xmath96 , which is @xmath153 independent , and for ( b ) @xmath164 with @xmath165 . here
figure reprinted with permission from @xcite copyright 2011 by american physical society.,title="fig:",width=245 ] the dark and metastable states have an important potential use for the quantum information processing and the memory devices @xcite but their excitation is an experimental challenge .
the dark state can be addressed by applying external fields or by adding a structural asymmetry .
the metastable ones can be coupled to confined photons in using different optical resonators as shown in subsection 4.5 .
the bright states are coupled directly to photons through the dipole coupling , and due to their strong dipole moment they can easily be tuned to achieve the strong coupling regime , making the atomic chain an optimal candidate as an active material for cavity qed . up to this point
we have treated only the case of an optical lattice in the mott insulator phase with one atom per site , but the case of two atoms per site is of large interest and can be easily realized in the present optical lattice experiments @xcite .
furthermore , this case shows novel phenomena beyond that of one atom per site , e.g. the appearance of dark and localized excitons @xcite . as before we consider two - level atoms , in which the ground and excited optical lattice potentials have minima at the same positions , and the atoms are located in the lowest bloch band .
moreover , we neglect overlaps between electronic wave functions for on - site and off - site atoms .
namely , we neglect the on - site formation of molecules from the two atoms , and we consider only transfer of electronic excitations . for planar optical lattices ,
the electronic excitation hamiltonian is as in eq .
( [ planar ] ) , but with an additional index for the two on - site atoms @xmath169 where @xmath170 and @xmath171 run over the two on - site atoms , which are denoted by @xmath172 and @xmath173 . for simplicity , we assume energy transfer only among on - site atoms with average coupling parameter @xmath174 , and among nearest - site atoms with coupling parameter @xmath175 , and we consider a lattice of square symmetry .
the above hamiltonian can be diagonalized in two steps , first the diagonalization related to the two on - site atoms , and second related to the site positions .
the first transformation is taken in introducing the on - site symmetric and antisymmetric states @xmath176 to get @xmath177 with the energies @xmath178 and @xmath179 .
the on - site energy states are represented in fig .
the second transformation is given by @xmath180 , which yields the diagonal hamiltonian @xmath181 the on - site symmetric states are delocalized in the lattice and form excitons with energy dispersion @xmath182 $ ] .
the antisymmetric states are on - site localized and form no excitons in the lattice ; they are dispersion - less with energy @xmath179 .
the on - site symmetric state is found to be superradiant with damping rate of two time a single excited atom , that is @xmath183 , and the antisymmetric state is dark with @xmath184 .
these results can be obtained from eq .
( [ dampingfinite ] ) of subsection 3.4 for a finite chain with @xmath185 , as presented also in @xcite . , and a transition energy @xmath186 .
( b ) the two diagonal states with the splitting energy of @xmath187.,width=321 ] dark localized states can in principle also be excited through nonlinear optical processes or the decay of multiply excited atomic states @xcite . moreover
, the addition of asymmetry to the system through the application of external fields or the deformation of the lattices sites , e.g. in using different laser frequencies and intensities for different directions , can make the antisymmetric states to be slightly bright and to delocalize in the lattice .
this asymmetry can be applied in a controllable way for the writing and reading of the antisymmetric states .
another possibility is the use of an optical super - lattice with two on - site potential minima .
a single atom is taken to be localized at each minimum , and the two on - site atoms are now separated by a given distance .
this leads to the exciton hamiltonian @xmath188 , where @xmath189 stands for the two exciton branches , with the symmetric state dispersion @xmath190 $ ] , where @xmath191 , here @xmath192 and @xmath193 .
the antisymmetric dispersion @xmath194 $ ] , where @xmath195 .
the antisymmetric state can transfer among nearest neighbor sites with the parameter @xmath196 , and to be delocalized and to form a second exciton branch , with a narrow energy band , and is slightly bright with a finite damping rate .
the formation of an on - site molecule out of the two on - site atoms significantly changes the optical properties , as the diatomic molecules can be oriented on the lattice due to the laser polarization or the external trapping fields .
this makes the system anisotropic similar to uni - axial crystals @xcite . as presented with more details in subsection 4.3 , the localization of the anisotropic lattice between planar cavity mirrors results in cavity photon polarization mixing , which can be observed through linear spectra . in this subsection
we discuss the exciton coherent transfer between two different optical lattices @xcite .
we start with two atoms that are separated by a vector distance @xmath197 and have transition dipoles @xmath198 and @xmath199 .
the transition dipoles are equal for identical atoms , but they can have in general different orientations in space .
the dipole - dipole interaction is given by @xmath200 which behaves as inverse cube of the inter - atomic distance .
this electrostatic interaction holds for inter - atomic distances smaller than the atomic transition wave length @xmath201 , where @xmath202 .
this interaction can be used also for two finite atomic chains , with lengths and inter - chain distances smaller than @xmath201 , where the transition dipoles are taken to be of the collective electronic excitations .
the energy transfer can be blocked among chains of different lengths as the collective state energies can be off resonance @xcite .
, with one atom per site and lattice constant @xmath49 .
the transition dipole makes an angle @xmath56 with the lattice direction.,width=321 ] because things can be completely different for extended systems , first we present the coherent exciton transfer among two long parallel one dimensional optical lattices @xcite , as schematized in fig .
the two lattices are separated by a distance @xmath203 , with one atom per site and lattice constant @xmath49 .
we consider an exciton with a given wave number @xmath84 that propagates in one optical lattice with a fixed polarization @xmath56 , and we derive the coupling parameter for the exciton transfer into the other lattice .
we calculate this parameter in considering the long range dipole - dipole interaction among each pair of atoms in applying the ewald s summation method of @xcite . in the limit of small wave numbers , that is @xmath204 , and for large inter - lattices distance , that is @xmath205 so that @xmath206 , we obtain @xmath207 the formation of excitons affect strongly the transfer parameter , as in place of the inverse cube dependence on the inter - atomic distance for the transfer parameter we get exponentially decay parameter behavior .
now we consider the case of two planar parallel optical lattices @xcite , which are separated by a distance @xmath208 , as schematized in fig .
[ 15 ] . for an exciton with in - plane wave vector @xmath114 that propagates in one lattice , we calculate the transfer parameter into the other lattice in applying once again the ewald s summation method .
the coupling parameter is @xmath209 where we defined @xmath210 with the in - plane dipole @xmath211 , and the unit vector @xmath212 .
this result shows a strong anisotropic effect . , with one atom per site and lattice constant @xmath49 .
the transition dipole includes in - plane and normal components.,width=321 ] the most important feature here is the exponential decay behavior as a function of the inter - lattices distance of the exciton coupling among the two optical lattices . even though the resonance dipole - dipole interaction has inverse cubic dependence of the distance between two localized dipoles , the appearance of excitons induce exponentially decrease dependence for the exciton collective dipole among the two lattices . for several parallel lattices
the result allows the use of the assumption of energy transfer among only nearest neighbor planes . such coherent coupling of the exciton transfer among the two lattices can be responsible for the formation of hybrid excitons in the whole systems , in which an exciton can coherently oscillate between the two lattices .
the excitation hamiltonian for two parallel planar optical lattices is exactly as in eq .
( [ twoatoms ] ) , but here @xmath170 and @xmath171 stand for the two lattices .
the same diagonalization steps lead to the exciton hamiltonian @xmath188 , with two hybrid exciton branches .
the symmetric excitons have the dispersion @xmath213 , and the antisymmetric dispersion is @xmath214 . here
@xmath215 is the intra - lattice coupling , and @xmath216 is for the inter - lattices , which is given in eq .
( [ interlattices ] ) .
note that the two hybrid excitons have an energy split of @xmath217 . also here
the antisymmetric modes are dark with zero damping rate , and the superradiant ones are bright with two times the single exciton damping rate .
in the recent years few experiments implemented a dilute boson gas of ultracold atoms in an optical cavity @xcite , and they reached the strong coupling regime because the collective coupling enhancement is enhanced by the square root of the atom number . in the experiment of @xcite
the optical lattice was formed by off resonance cavity photons , with the resulting force effects investigated in @xcite .
optical lattices within an optical cavity operating in the resonant regime are not yet implemented .
@xcite have introduced optical lattices as a new type of active materials for cavity qed . beside the importance of such a system for fundamental physics and its wide range of applications ,
the appearance of superradiant collective states in optical lattices implies the confinement of the electromagnetic field within a cavity .
moreover , as large one and two dimensional lattices give rise for metastable states , which decouple to the free space photons , they appeal to confined photons for efficient coupling .
we begin by answering the crucial question concerning the influence of the cavity photons on the previously calculated phase diagrams @xcite .
namely , how the different mott insulator phases are affected due to the coupling to cavity photons .
the optical lattice is taken to be localized between cavity mirrors , as in fig . [
two electromagnetic fields are required , the classical laser field which is off - resonance and used to form the optical lattice , and the quantized cavity field .
a single cavity mode is assumed to be close to resonance with a single atomic transition .
the previous two - component bose - hubbard model is generalize to include the coupling to cavity photons
. this coupling changes the behavior of the bosons through the emission and absorption of cavity photons , and that is expected to modify the phase diagram .
is smaller than the mode waist @xmath218.,width=188 ] the two - component bose - hubbard model within a cavity in the resonant regime is represented by the hamiltonian @xmath219 where here @xmath220 and @xmath221 , with @xmath222 and @xmath223 are the ground and excited state atoms chemical potentials , respectively . @xmath224 and @xmath225 are taken to be site independent , hence we dropped their index @xmath8 .
we use also @xmath226 , where @xmath227 is the cavity mode energy and @xmath228 is the photon chemical potential .
furthermore , the photon - excitation coupling @xmath229 is assumed to be site independent , which is the case for a homogeneous optical lattice parallel to the cavity mirrors .
we also defined the mean atom and photon number operators , by @xmath230 , @xmath231 , and @xmath232 . here
@xmath233 and @xmath49 are the creation and annihilation operators of a cavity mode , and close to resonance one has @xmath234 .
the modified phase diagram is calculated in applying the consistent mean - field theory suggested in @xcite , as the conventional bogoliubov approximation does not predict the correct quantum phase transition @xcite . in this mean - field theory
the atoms are subject to the mean - field of the neighboring sites and of the other kind of bosons .
hence the hamiltonian can be separated into on - site terms and the rest , which can be written as a sum of a free part and an interacting part , and where we drop the site index in the following .
then the interaction part can be treated in perturbation theory . in this method
atom - atom interactions are treated exactly while the hopping and the coupling to photons are treated approximately .
more complex quantum phases are discussed in @xcite beyond the mean - field theory .
we present an example of phase diagram by plotting the scaled chemical potentials as a function of the atom - atom interactions .
the system in total has a large number of open parameters , so that we can not represent the whole phase space . for our arguments here
we choose the parameters in such a way to allow the existence the mott insulator phase and to emphasize the modifications brought about by the coupling to the cavity photons .
we concentrate on the case with two atoms per site , that is in the case @xmath235 .
we assume one photon in average per site , that is @xmath236 , where @xmath237 , although this assumption is not so crucial . for @xmath236
we have a large number of photons in the cavity , so that photon number fluctuations are suppressed .
as we concentrate in the following on the influence of the coupling to the cavity photons , we assume @xmath238 , and all the parameters will be scaled by @xmath14 .
furthermore , we assume zero ground state energy , @xmath239 , and the cavity mode in resonance with the electronic excitation , that is @xmath240 . moreover , by assuming a resonant pump - cavity mode the chemical potential for cavity photons becomes zero , that is @xmath241 . in the following phase diagram plots we use @xmath242 and for the excitation - photon scaled coupling we used @xmath243 , where @xmath244 .
fig.[17](a ) reports the phase diagram for the planes @xmath27 vs. @xmath36 , and @xmath31 vs. @xmath37 .
we used @xmath245 for the scaled atom - atom coupling .
similar to the case without a cavity , the plot shows the superfluid and the mott insulator phases for one ground or excited state atom per site , that is @xmath48 , with the le and lg transition lines . beside the sf and mi mott regions
, we have as before the sm and ms regions .
the results of fig . [ 17](a ) should be compared to those of fig . [ 3 ] for the phase diagram of a two - component bose - hubbard model without coupling to cavity photons .
we deduce that the effect of the coupling to cavity photons is to shift the mott insulator phase to larger atom - atom interactions , by one order of magnitude in the present case .
as we used @xmath241 , the ground state transition line is shifted relative to the excited state one .
the coupling to cavity photons tend to delocalize the excited atoms through the process in which an excited atom jumps to the ground state by emitting a cavity photon , in turn absorbed by another atom in the lattice located farther apart .
in this situation the atom - atom interaction competes against the atom hopping and the coupling to cavity photons , and to recover the mott insulator phase within a cavity one need to apply a deeper optical lattice .
[ 17](b ) reports the scaled chemical potentials @xmath27 and @xmath31 as a function of @xmath246 , for @xmath245 , and @xmath247 .
it is clear that the mott insulator phase appears only for a limited range of coupling parameters , and for a limited range of mean cavity photon number per site , which can range from few cavity photons up to about one photon per lattice site . beyond some critical values of @xmath246
the atom - atom interaction is not large enough to form the mott insulator phase and only the superfluid phase persists .
more mott insulator phases with other number of atoms per site are presented in @xcite . vs. @xmath36 , and @xmath31 vs. @xmath37 , for @xmath48 .
we have @xmath242 , @xmath243 , @xmath245 , and @xmath236 .
the dashed line is for the excited state atoms , and the full line is for the ground state atoms .
( b ) scaled phase diagram : @xmath248 vs. @xmath246 , for @xmath247 .
figure reprinted with permission from @xcite copyright 2009 by american physical society.,title="fig:",width=245 ] vs. @xmath36 , and @xmath31 vs. @xmath37 , for @xmath48 .
we have @xmath242 , @xmath243 , @xmath245 , and @xmath236 .
the dashed line is for the excited state atoms , and the full line is for the ground state atoms .
( b ) scaled phase diagram : @xmath248 vs. @xmath246 , for @xmath247 . figure reprinted with permission from @xcite copyright 2009 by american physical society.,title="fig:",width=245 ] in all the above discussion we assumed long life times for both the excited atoms and the cavity photons .
this assumption can be a good approximation , e.g. , for rydberg atoms in a high - q cavity .
the overlaps between the mott insulator phases of different number of atoms per site make it possible to choose the quantum phase such that following the decay of an excited atom the systems may remain in one of the mott insulator phases .
the lifetime of the cavity photons has less influence on the mott insulator phase , as the leak of a photon in the mott insulator keeps the system inside the same phase zone .
the calculated phase diagrams provide us with a new tool for controlling the system quantum phases and their transitions . in place of controlling the external laser fields , which form the optical lattice , it is possible to control the cavity photons .
for example we can chose the different quantum phases by fixing the average number of photons within the cavity , which can be easily achieved via controlling the external pump . here
we introduce optical lattice ultracold atoms into an optical resonator as an active material in a cavity qed @xcite . to represent the main results of this
set - up we consider a planar optical lattice that enclosed in the middle between planar cavity mirrors , as represented in fig .
[ 21](a ) .
we start by assuming an ideal cavity and later we include the damping of cavity photons into the outside radiation field .
the cavity photons are free to propagate in the cavity plane with a given in - plane wave vector @xmath114 , and they have discrete modes in the perpendicular direction with wave numbers @xmath249 , where @xmath250 is the distance between the mirrors and @xmath74 is an integer @xmath251 . in the perpendicular direction we consider only a single mode , the one that close to resonance with the atomic transition , e.g. the one with @xmath252 with a maximum field at the lattice plane , the other modes having higher energies . in this part
we consider isotropic optical lattice , hence cavity photons with a fixed polarization , i.e. te and tm modes or any coherence superposition of them . in subsection 4.3
we treat an anisotropic optical lattice of uni - axial symmetry that give rise to photon polarization mixing . , and
its orthogonal direction , @xmath253 , are reported.,title="fig:",width=226 ] , and its orthogonal direction , @xmath253 , are reported.,title="fig:",width=226 ] for the coupled exciton - photon system , in the dipole interaction and for the rotating wave approximation , the hamiltonian is written as @xmath254 here @xmath255 are the creation and annihilation operators of a cavity photon with in - plane wave vector @xmath114 .
the cavity photon dispersion reads @xmath256 where @xmath250 is the distance between the cavity mirrors , and @xmath257 is the cavity medium dielectric constant .
the coupling parameter , e.g. for a lattice of square symmetry , is given by @xmath258 in the case of excitons and photons close to resonance having small wave numbers , that is @xmath259 for @xmath260 , the excitons can taken to be dispersion - less with @xmath261 .
translational symmetry leads to coupling between excitons and photons with the same in - plane wave vectors , namely to a momentum conservation . in the strong coupling regime where the exciton and photon damping rates are smaller than the coupling parameter
, the hamiltonian can be diagonalize by introducing polaritons , through the transformation @xmath262 where the exciton and photon amplitudes are @xmath263 with @xmath264 , and we defined the exciton - photon detuning by @xmath265 . in this regime the real system excitations are polaritons , quasi - particles formed by a coherent superposition of excitons and photons , with the system coherently oscillating between those two .
the polariton hamiltonian is given by @xmath266 with the polariton dispersions @xmath267 we get two polariton branches with a splitting of @xmath268 , the intersection point appearing at @xmath269 and the vacuum rabi splitting equal to @xmath270 .
the polaritons are denoted as upper and lower branches .
the plots of fig . [ 19](a ) report the results for the upper and lower polariton branches in addition to the cavity photon and exciton dispersions , using the typical parameters listed in the following .
the exciton energy around small in - plane wave vectors is taken to be @xmath271 , and a zero detuning between the exciton and the cavity photon dispersions at zero in - plane wave vector .
the transition dipole is @xmath272 , and the optical lattice constant is @xmath273 .
the exciton - photon coupling parameter is @xmath274 , by neglecting the @xmath84 dependence for small in - plane wave vectors . at zero in - plane wave vector
the splitting is clearly given by the rabi frequency .
[ 19](b ) reports the @xmath275 and @xmath276 exciton and photon weights for the lower polariton branch . at the exciton - photon intersection point
the polaritons are half exciton and half photon , that is @xmath277 . at large wave vectors
the lower branch becomes excitonic , that is @xmath278 , and the upper branch becomes photonic , that is @xmath279 . .
the dashed - dotted line denotes for the exciton dispersion , and the dashed - dotted parabola denotes for the cavity photon dispersion .
( b ) the exciton and photon weights vs. the in - plane wave vector @xmath84 for the lower polariton branch .
figure reprinted with permission from @xcite copyright 2007 by american physical society.,title="fig:",width=245 ] .
the dashed - dotted line denotes for the exciton dispersion , and the dashed - dotted parabola denotes for the cavity photon dispersion .
( b ) the exciton and photon weights vs. the in - plane wave vector @xmath84 for the lower polariton branch . figure reprinted with permission from @xcite copyright 2007 by american physical society.,title="fig:",width=245 ] the above discussion considered an ideal cavity with perfect mirrors . but in order to observe the system eigenmodes , we need to couple the internal cavity modes to the external world .
this is done using the standard input - output formalism , which is well proved for high - q cavities @xcite . in this approach
the internal cavity field modes and the external free radiation field are quantized separately and weakly coupled via the cavity mirrors .
this leads to the cavity field damping but also allows to calculate consistently the relation between the cavity input and output fields . to include the atomic spontaneous emission as an energy loss term
, we add a phenomenological effective exciton damping .
this approach finally allows to calculate the transmission @xmath280 , reflection @xmath281 , and absorption @xmath282 spectra for a given incident field in the cavity configuration of fig .
[ 21](a ) . for the previously investigated system we plot in fig .
[ 20 ] the transmission spectra for different wave vectors . here
we choose the following parameters for the damping rates : exciton damping rate @xmath283 , upper and lower mirror damping rates @xmath284 .
the peaks of the transmission correspond to the two polariton branches , which are the real eigenmodes of the system .
the plot indicates that high transmission peaks are obtained for polaritons which are much more photonic than excitonic . at zero wave vector , as the polariton is half exciton and half photon , the upper and the lower spectra are identical . as the wave vector increase the upper branch becomes more photonic than excitonic , and the opposite for the lower one .
hence , the line - width of the upper branch becomes wider , which equals the cavity line - width , and the lower one becomes narrower , which equals the atom line - width . the reflection and absorption spectra can be easily derived in the same manner .
the present results for the mott insulator with one atom per site are expected to be significantly different from the other cases with other number of atoms per site and also from the superfluid phase . for different in - plane wave vectors .
the two peaks correspond to the upper and lower polariton branches at each in - plane wave vector .
figure reprinted with permission from @xcite copyright 2007 by american physical society.,width=264 ] the widely used time - of - flight method for observing the different quantum phases in optical lattices with their quantum phase transitions is a destructive observation tool @xcite . as the laser creating the lattice is switched off and after waiting some time a picture is taken , the system is destroyed and in each observation one need to prepare a new system .
we suggest to use the optical linear spectra as a non - destructive observation tool for the quantum phases .
the appearance of exciton and polariton provides a direct observation tool that can be repeated without modifying the quantum phase .
later we present other optical spectra , and the comparison between them allows us to extract information concerning the quantum phases .
the case of a two dimensional optical lattice in the mott insulator phase with two atoms per site was investigated in subsection 3.5 . here
this system is introduced into a cavity where the optical lattice is taken to be localized in the middle and parallel to two planar cavity mirrors @xcite , as in fig . [ 21](a ) .
the main objective is to examine how the previously derived dark and bright excitons behave within a cavity in the strong coupling regime .
we start by considering the case of asymmetric optical lattice in which the two on - site atoms are separated by an average distance .
the electronic transitions are coupled to the cavity photons with the interaction hamiltonian @xmath285 , with @xmath286 for the two atoms .
the coupling parameter @xmath287 between atom @xmath170 at site @xmath8 and a photon with wave vector @xmath114 is given by @xmath288 , where @xmath289 is the atom position . in terms of on - site
symmetric and antisymmetric operators we get @xmath290 , with @xmath291 for the two modes , where we defined the symmetric and antisymmetric coupling parameters @xmath292 , and @xmath293 . as the on - site
interatomic distance is much smaller than the atomic transition wavelength , the symmetric state coupling is much larger than the antisymmetric one .
the coupling can be controlled through the external laser fields creating the optical lattice in using different lasers at different directions .
it is possible to switch the antisymmetric coupling on and off by converting the lattice site from asymmetric into symmetric one . for symmetric optical lattices we have approximately @xmath294 , where @xmath295 and @xmath296 .
then the antisymmetric excitation states are almost decoupled from the cavity - photons , and they can be considered as dark states . only the symmetric excitation states are strongly coupled to the cavity - photons . in this case by using the symmetric and antisymmetric excitonic picture in the momentum space , we get the total hamiltonian @xmath297 where @xmath298 , and @xmath299 . in the strong coupling regime
the symmetric excitons and the cavity photons are coherently mixed to form the new system excitations which are the polaritons .
the above hamiltonian first part can be easily diagonalize to give the following polariton hamiltonian : @xmath300 here we can use exactly the definition of polaritons by replacing the exciton energy with the symmetric one and the coupling with the symmetric one .
the antisymmetric states are on - site localized and dark , but they can be written and read in a controllable way , this fact making them useful as memory devices . in the present subsection
we assume the transition dipole to have a fixed direction @xcite . this case can occur for optically pumped atoms in a fixed zeeman sub - level of a given angular momentum . also molecules can have oriented transition dipole , e.g. diatomic molecules , where the electronically excited states have transition dipoles oriented along the molecule axis . in combination with external traps and fixed optical lattice laser polarizations the atomic and molecular transition dipole are aligned along a specific direction.this orientation that makes the lattice to be anisotropic with uniaxial symmetry , as represented in fig [ 21](b ) for a planar optical lattice .
the anisotropic properties can be implemented as a signature for the formation of on - site molecules in the optical lattice .
for example the formation of homonuclear diatomic molecules out of the two on - site atoms in the mott insulator phase can be manifested through the anisotropic optical spectra .
we will show that the electronic excitations and the cavity photons of both te and tm polarizations are coherently mixed to form polariton branches , which can be directly observed in the linear optical spectra .
for example , for an incident field of te polarization we obtain both te and tm transmitted and reflected fields . the planar optical lattice is taken to be localized in the middle and parallel to cavity mirrors , as presented in fig . [ 21](a ) . here
we generalize the previous presentation of the cavity photons in order to include the polarizations .
the electromagnetic field is free in the cavity plane with in - plane wave - vector @xmath114 , and is quantized in the perpendicular , @xmath15 , direction with wave numbers @xmath301 , where @xmath250 is the distance between the cavity mirrors , and @xmath74 takes integer numbers , ( @xmath302 ) .
thus for each cavity mode @xmath303 we have two possible polarizations te and tm , which will be denoted by the indexes @xmath304 and @xmath305 , respectively .
the two cavity photon polarizations , @xmath306 , are assumed to be degenerate . because we will consider only cavity modes with @xmath252 to be close to resonance to the atomic transition , we have for the two polarizations @xmath307 , with the cavity photon dispersion @xmath308 given by eq . .
the cavity photon hamiltonian is @xmath309 , where @xmath310 and @xmath311 are the creation and annihilation operators of a cavity photon in the mode @xmath312 , respectively .
the electric field operators and the field spatial functions are reported by @xcite .
we introduce the following unit vectors @xmath313 along the @xmath15 axis , @xmath314 along the @xmath114 in - plane wave vector , that is @xmath315 , and @xmath316 , as illustrated in fig .
[ 21](b ) .
the optical lattice is located at @xmath317 , the cavity middle , and parallel to the cavity mirrors .
the coupling between the atomic transition and the cavity modes is given by the hamiltonian @xmath318 , where the coupling parameter is @xmath319 .
for example , we take @xmath320 to be real , e.g. for @xmath321 , see fig .
21(a ) , and also we have @xmath322 and @xmath323 , to get @xmath324 and @xmath325 , where @xmath326 , and @xmath327 .
the total hamiltonian of the coupled excitations and photons is given by @xmath328 in the strong coupling regime the real system eigenmodes are cavity polaritons , which are obtained by diagonalizing the above hamiltonian .
we obtain the polariton hamiltonian @xmath329 , characterized by three polariton branches , with the eigenfrequencies @xmath330 and @xmath331 , where @xmath332 and @xmath333 , with the excitation - photon detuning @xmath334 . for the case of exciton frequency of @xmath335 , with a distance between the cavity mirrors of @xmath336 , zero detuning between the atomic transition and the cavity photon at @xmath337 , at the angle @xmath164 , transition dipole of @xmath272 , and lattice constant @xmath338 , fig .
[ 22 ] reports the three polariton frequency dispersions @xmath339 as a function of @xmath84 . vs. @xmath84 , at @xmath164 and other parameters reported in the text .
figure reprinted with permission from @xcite copyright 2009 by institute of physics.,width=321 ] in the limit of large detuning , where @xmath340 , we get @xmath341 , @xmath342 , and @xmath331 . the upper @xmath343 branch is a photon with a shift due to the coupling to the excitation , and the lower @xmath344 one is an excitation with a light shift due to the coupling to the cavity photon .
the middle polariton branch is pure photonic . in that limit
we get birefringence @xcite , determined by two refracted cavity fields , the ordinary field of the @xmath343 branch , and the extraordinary field of the @xmath344 branch .
the polariton operators are in general a coherent superposition of excitations and photons of both polarizations , namely we have @xmath345 , with the relation @xmath346 , where the excitation and photon amplitudes of the upper @xmath344 and the lower @xmath347 polariton branches are @xmath348 and @xmath349 , and the excitation and photon amplitudes of the middle @xmath343 polariton branch are @xmath350 and @xmath351 . for each photon polarization , in the strong coupling regime the longitudinal component along the transition dipole coupled to the excitation contributes to the upper and lower polariton branches . instead the transverse component orthogonal to the dipole and decoupled from the excitation , creates the photonic middle branch . for each @xmath114 ,
a direction exists where the polarization direction of a photon is orthogonal to the transition dipole , and hence no interaction between this photon and the material is obtained . , at @xmath352 .
in ( a ) the @xmath304 polarized field transmission spectra , @xmath353 vs , of @xmath304 polarized incident field ; in ( b ) the @xmath304 polarized field reflection spectra , @xmath354 ; in ( c ) the absorption spectra , @xmath355 ; in ( d ) the @xmath305 polarized field transmission and reflection spectra , @xmath356 and @xmath357 . figure adapted with permission from @xcite copyright 2009 by institute of physics.,title="fig:",width=245 ] , at @xmath352 . in ( a )
the @xmath304 polarized field transmission spectra , @xmath353 vs , of @xmath304 polarized incident field ; in ( b ) the @xmath304 polarized field reflection spectra , @xmath354 ; in ( c ) the absorption spectra , @xmath355 ; in ( d ) the @xmath305 polarized field transmission and reflection spectra , @xmath356 and @xmath357 .
figure adapted with permission from @xcite copyright 2009 by institute of physics.,title="fig:",width=245 ] , at @xmath352 . in ( a ) the @xmath304 polarized field transmission spectra , @xmath353 vs , of @xmath304 polarized incident field ; in ( b ) the @xmath304 polarized field reflection spectra , @xmath354 ; in ( c ) the absorption spectra , @xmath355 ; in ( d ) the @xmath305 polarized field transmission and reflection spectra , @xmath356 and @xmath357 . figure adapted with permission from @xcite copyright 2009 by institute of physics.,title="fig:",width=245 ] , at @xmath352 . in ( a ) the @xmath304 polarized field transmission spectra , @xmath353 vs , of @xmath304 polarized incident field ; in ( b ) the @xmath304 polarized field reflection spectra , @xmath354 ; in ( c ) the absorption spectra , @xmath355 ; in ( d ) the @xmath305 polarized field transmission and reflection spectra , @xmath356 and @xmath357 .
figure adapted with permission from @xcite copyright 2009 by institute of physics.,title="fig:",width=245 ] to observe the system eigenmodes we need to couple the internal system to the external world .
our observation tool is linear optical spectra . in assuming non - perfect mirrors , the internal system get coupled to their environment . for an incident external field with a fixed in - plane wave vector and polarization we calculate the transmission , reflection , and absorption spectra , as shown in fig . [
the lifetime of the excited state is included phenomenologically . for a te or @xmath304 polarized incident field , fig . [ 23 ] report the transmission , reflection and absorption spectra of the @xmath304 polarized fields , @xmath353 , @xmath354 , and @xmath355 , respectively , as a function of frequency , @xmath358 , at different angles @xmath56 and @xmath352 . for the excitation and photon
damping rates we used @xmath359 hz and @xmath360 hz . for @xmath353 and @xmath354 the three peaks and three dips correspond to the three polariton branches . at @xmath96
we get zero transmission and complete reflection .
the largest transmission peaks and the deepest reflection dips are obtained at @xmath97 . for the absorption spectra
@xmath355 only two peaks are obtained correspond to the upper and lower branches , as the middle branch is pure photonic and no absorption take place , where the absorption is only for the polariton excitation part . also here zero absorption at @xmath96 , and maximum absorption at @xmath97 , are obtained . fig . [ 23](d ) reports the transmission and reflection spectra of the @xmath305 polarized fields , @xmath356 and @xmath357 , which are equal for identical cavity mirrors .
even though the incident field is @xmath304 polarized we get transmission and reflection of @xmath305 polarized fields , with maximum at @xmath164 , and zeros at @xmath96 and @xmath97 .
in subsection 3.2 we studied collective electronic excitations in a finite linear atomic chain , and in subsection 3.4 we derived their damping rate into free space radiation .
now we localize the finite chain in the middle between two identical spherical cavity mirrors such that the chain axis is perpendicular to the cavity axis @xcite , as shown in fig . [
only the lowest gaussian mode in the cavity is assumed to be close to resonance to the atomic transition .
the electric field operator across the cavity waist , chosen along the optical lattice axis at @xmath317 and as a function of the distance from the cavity axis @xmath361 , is given by @xcite @xmath362 where @xmath363 is the photon polarization unit vector , @xmath364 the beam waist , and @xmath365 the mode volume given by @xmath366 . here
@xmath367 is the cavity photon energy , and @xmath368 is the distance between the cavity mirrors .
the collective states are divided into two groups of symmetric and antisymmetric modes .
the antisymmetric modes are found to be dark and decoupled from the cavity photons , while the symmetric ones are bright . for odd @xmath84-s , in the dipole interaction approximation , the coupling parameter is given by @xmath369 the lowest mode , @xmath370 , the one without nodes , has the highest coupling parameter , that is nine times larger than the next bright mode of @xmath371 .
this mode dominates the optical properties of the system and can easily achieve the strong coupling regime within a cavity .
it includes @xmath372 of the oscillator strength , and is termed superradiant mode .
hence , in the following we consider only the coupling of such superradiant mode to the cavity photons .
the hamiltoian of the coupled superradiant mode and the cavity mode is given by the jaynes - cumming model @xmath373 , where now the creation and annihilation operators @xmath374 and @xmath375 stand for the superradiant mode respectively , and @xmath233 and @xmath49 for the cavity photon .
the coupling parameter is @xmath376 , and the superradiant mode energy is @xmath377 .
the hamiltonian can be easily diagonalized using the transformation @xmath378 , to get @xmath379 , with the two dressed eigenenergies @xmath380 , where @xmath381 and the detuning is @xmath382 .
the eigenstates are coherent superpositions of the superradiant mode and the cavity photon .
the superradiant mode amplitudes are @xmath383 , and the cavity photon amplitudes are @xmath384 . at the intersection point , with @xmath385 , the dressed state is half a superradiant state and half s photon .
but for large positive detuning the @xmath347 branch tends to the superradiant state and the @xmath344 branch becomes photon , and vice versa for the negative detuning . to emphasize the energy transfer influence within a cavity , we compare the results to those without dipole - dipole interactions . the electronic excitation is described now by the hamiltonian @xmath386 .
the excitation - photon coupling hamiltonian reads @xmath387 , where the coupling parameter is @xmath388 .
we define the collective excitation operator @xmath389 with @xmath390 , to get the hamiltonian @xmath391 , where @xmath392 . in fig .
[ 25 ] we compare the vacuum rabi splittings , @xmath393 , as a function of the atom number for the two cases , with and without the resonance dipole - dipole interaction .
it appears that for a large atom number the vacuum rabi splitting for independent atoms is larger than that of interacting atoms .
we conclude that the vacuum rabi splitting is reduced by the dipole - dipole interactions and deviates from the square - root behavior .
this result can be used as a signature for the formation of collective electronic excitations in atomic chains . vs. the atom number @xmath82 , for interacting ( full line ) , and non - interacting ( dashed line ) , optical lattice ultracold atoms .
the difference increases with the atom number .
figure reprinted with permission from @xcite copyright 2009 by european physical society.,width=321 ] recently new directions have been opened for the light - matter coupling in using optical fibers @xcite . in tapered optical fibers the atoms
are trapped outside the fiber and couple to the evanescent field surrounding the fiber .
the guided modes of ultrathin optical fibers exhibit pronounced evanescent field that give rise to an array of optical microtraps .
the system was realized recently for cesium atoms localized in one dimensional optical lattice parallel to the nanofiber @xcite .
we study here a simplified model related to this setup , based on a tapered optical nanofiber and two optical lattices @xcite .
we consider two identical parallel one - dimensional optical lattices with one atom per site , separated by a distance @xmath203 and located at the two opposite sides of the fiber , at a distance @xmath208 from the fiber surface , as schematized in fig .
[ 26 ] . and distance @xmath203 between them , are located parallel to a fiber , at the two opposite sides .
an incident field is injected from the left side of the fiber and the transmitted field is observed on the right side .
the transition dipole @xmath10 makes an angle @xmath56 with the lattice direction.,width=283 ] the electronic excitation hamiltonian in the two optical lattices is given by the one - dimensional version of the planar hamiltonian of eq .
( [ twoatoms ] ) @xmath394 here @xmath395 run over all the lattice sites , and @xmath396 stand for the two lattices @xmath397 , with @xmath398 are the creation and annihilation operators of an electronic excitation at site @xmath73 in lattice @xmath170 , respectively .
the coupling parameter @xmath399 is for resonant dipole - dipole interactions , and gives rise to the excitation transfer among two lattice sites .
the transfer term includes transfers among atoms in the same lattice , @xmath400 and @xmath401 , and transfers among atoms at different lattices , @xmath402 and @xmath403 .
as before , the hamiltonian can be diagonalized in two steps .
first , in the lattice sites in using the transformation @xmath404 . here
@xmath405 is the position of site @xmath73 in lattice @xmath170 .
second , we diagonalize the hamiltonian relative to the two lattice indexes , @xmath396 , by applying the transformation @xmath406 .
the new states represent entangled states between excitons from the two lattices , which are symmetric and antisymmetric states .
the symmetric state is denoted by @xmath407 and takes the plus sign @xmath344 , the antisymmetric state is denoted by @xmath408 and takes the minus sign @xmath347 .
the hamiltonian casts into the diagonal form @xmath409 , where the eigenenergies are @xmath410 and @xmath411 .
we define the exciton dynamical matrix by @xmath412 , using @xmath413 , with @xmath414 , where the dipole - dipole interaction is a function of the distance between the two atoms . for identical lattices , inside the same lattice we have @xmath415 , and among different lattices we have @xmath416 .
excitons in one dimensional optical lattices are presented in subsection 3.2 , and the coherent transfer parameter of excitons among two parallel lattices is discussed in subsection 3.6 . for the nanofiber photons we use a simplified picture .
the one - dimensional optical fiber modes are given by the hamiltonian @xmath417 .
the photon dispersion is taken to be of the form @xmath418 , where the wave number @xmath419 results of the fiber transverse confinement , and @xmath420 is the fiber average dielectric constant .
the electric field operator outside the fiber is @xmath421 where @xmath363 is the photon linear polarization unit vector , @xmath365 the normalization volume , and @xmath422 the mode function which includes the fiber field complexity .
the exciton - photon coupling hamiltonian is given by @xmath423 , with the coupling parameter @xmath424 .
only the symmetric excitons are coupled to the fiber photons , where the antisymmetric ones are dark and decoupled from the photons .
the total hamiltonian reads @xmath425 where due to translational symmetry along the lattice and fiber axis , the hamiltonian is separated for each @xmath84 . in the strong coupling regime
we define the fiber polaritons by diagonalizing the hamiltonian to get @xmath426 we can use the polaritons definition of subsection 4.2 in replacing the exciton and photon dispersions with the present one dimensional ones . here
we present the results for a system with the following parameters : lattice constant @xmath61 ; transition energy @xmath59 ; transition dipole @xmath427 at an angle @xmath135 with the lattice axis ; distance between the two optical lattices @xmath428 ; fiber dielectric constant is @xmath429 ; mode cross area @xmath430 , and mode function at the lattice position @xmath431 .
the cavity mode energy at @xmath337 is taken to be in resonance with the free atom transition energy , that is @xmath432 , where @xmath433 .
the polariton dispersions are similar to those plotted in figs .
the exciton - photon intersection point is here to at @xmath337 , where the polaritons are split by the rabi splitting . to get the linear optical spectra we consider an incident field from the far left side of the fiber , and we calculate the transmission spectra .
the coupling of the fiber field to the external field at the two far edges of the fiber is included in the parameter @xmath434 , and the fiber photon damping rate is taken to be @xmath435 which is assumed to be @xmath84 independent . the symmetric exciton damping rate is included phenomenologically , and is taken to be @xmath436 as discussed in subsection 3.4 .
the polariton damping rate is assumed to be @xmath437 ; a small damping rate of @xmath438 is introduced for the photon , and the fiber edge coupling parameter is assumed to produce a bandwidth @xmath439 .
[ 27](a ) reports the transmission spectra for a wavenumber @xmath440 , where at resonance we get a dip in the transmission spectrum .
[ 27](b ) representing a close look at the minimum shows a small shift relative to the free atom transition , produced by the dipole - dipole interaction . for this case
the symmetric exciton line width is @xmath441 , larger than the free atom line width of @xmath442 . , and @xmath439 .
( b ) a close look around the minimum . figure reprinted with permission from @xcite copyright 2010 by institute of physics.,title="fig:",width=245 ] , and @xmath439 .
( b ) a close look around the minimum . figure reprinted with permission from @xcite copyright 2010 by institute of physics.,title="fig:",width=245 ]
up to this point we treated ideal optical lattices in the mott insulator phase with a fixed number of atoms per site and in which all atoms are localized in the lowest bloch band . in this section
we partly relax this idealization and include different kinds of disorders into the system . in the following
we introduce mainly two kinds of disorders .
first , we exploit the excitation of on - site atoms into higher bloch bands , where we concentrate in the lowest bands @xcite .
second , in the mott insulator phase we consider localized defects that appear at several separated sites in the lattice @xcite .
we treat defects of vacancies distributed randomly in the lattice in the mott insulator phase with one atom per site . for deep optical lattices in the mott insulator the lowest bloch bands
are well separated and can be represented by vibrational modes of a quantum harmonic oscillator .
the transition of atoms between different vibrational states can be induced , e.g. , by inelastic raman scattering of light , or by on - site atom - atom interactions and atom hopping among nearest neighbor sites .
we present an alternative mechanism for the excitation and de - excitation of atoms between different bloch bands through the coupling of the internal electronic excitation to the external atomic motion via electrostatic interactions @xcite .
we consider here the mott insulator with one atom per site where the atoms are treated as two - level systems . in general ground and excited atoms in optical lattices experience different optical lattice potentials , and can have minima at different positions , the fact that induces the coupling of electronic excitations to the vibrational modes .
we show that the on - site electronic excitation is dressed by a cloud of virtual vibrational modes that renormalize the excitation energy . in the appropriate regime ,
the excitation transfers among nearest neighbor sites due to resonance dipole - dipole interactions and can be accompanied by the emission or absorption of vibrational modes .
we start from the electronic excitation hamiltonian @xmath443 .
the internal atomic transition frequency at site @xmath8 of position @xmath444 is @xmath445 , where we use the approximation @xmath446 for the ground and excited state energies @xmath447 . here
@xmath448 is the change in the local transition energy due to an on - site atomic vibration excitation with @xmath449 vibrational quanta , and @xmath450 is vibration - free transition frequency with zero quanta @xmath451 .
we take @xmath452 to be the average spatial growth of the atomic size at site @xmath8 in the @xmath453 internal state due to the excitation of the atom to a higher vibrational state with @xmath74 quanta .
@xmath448 is calculated relative to the local lattice ground state @xmath451 , where @xmath454 is included in @xmath450 , and @xmath452 is measured relative to @xmath455 of the vibration - less state .
the excitation can exchange among atoms at different sites @xmath8 and @xmath456 due to dipole - dipole interactions , and it parametrized by the coupling parameter @xmath457 .
this coupling depends on the average atomic vibration shift @xmath452 and thus on the local vibrational excitation . in a deep lattice
, @xmath452 is a small deviation relative to the lattice constant @xmath49 for the lowest vibrational modes .
to lowest orders in this small perturbation we then can split the hamiltonian into the form @xmath458 . here
@xmath459 is the internal excitation hamiltonian , which is obtained for atoms in the ground vibrational states , @xmath460 is the atom vibration hamiltonian , for excited and ground state atoms , and the coupling hamiltonian @xmath461 is derived perturbatively for atoms excited to higher vibrational states .
to the zero order , for atoms in the ground vibrational states , we thus get @xmath462 , where the transition frequency @xmath463 is the same at each site , and @xmath464 is the transfer parameter among atoms in the lowest vibrational states .
this is exactly the hamiltonian we treat before in subsection 3.1 , and that results in excitons for optical lattices .
vibrational excitations of atoms in the ground and excited state potentials are approximately described by the harmonic hamiltonian @xmath465 where @xmath466 and @xmath467 are the vibration frequency for ground and excited state atoms , respectively . @xmath468 and @xmath469 are the creation and annihilation operators of a vibrational mode at site @xmath8 for ground and excited state atoms , respectively .
the corresponding atomic displacement operators are @xmath470 where @xmath471 is the atomic mass . here ,
ground and excited state atoms are considered as two different kinds of bosons , where each has its own optical lattice potential , as in subsection 2.2 . in the transition between the ground and the excited state one kind of bosons
is destroyed and another created .
to first order in the perturbation series with respect to @xmath472 , we get excitation - vibration coupling by the hamiltonian @xmath473 , where @xmath474 is for the on - site part , and @xmath475 for the transfer part . processes in the first order terms include only a single vibrational quanta . to treat processes of more than a single vibrational quanta one needs to consider higher order terms of the perturbation series .
as we treat here only the first order terms we drop the index @xmath476 off @xmath448 . the on - site part is given by @xmath477-m^g_i\ \left[\left(b_i^{\dagger}b_i\right)b_i+b_i^{\dagger}\left(b_ib_i^{\dagger}\right)\right]\right\},\ ] ] where the coupling parameter is @xmath478 related to the slope of @xmath479 . as @xmath479 is a function of the atomic wave function at site @xmath8 and in a fixed bloch band
, the parameter @xmath480 is related to the correlations between atomic wave functions at the initial and the final bloch bands .
the four possible transitions are presented in fig .
the transfer part is given by @xmath481\ b_i^{\dagger}b_j,\ ] ] where the coupling parameter is @xmath482 which is related to the derivative of @xmath457 .
in the limit of on - site excitation - vibration coupling which is stronger than the transfer coupling , namely in the limit of @xmath483 , we can apply the following canonical transformation @xcite of @xmath484 and @xmath485 , with @xmath486 $ ] , to get @xmath487 and @xmath488 , and with @xmath489 .
the new effective local excitation represents an electronic excitation dressed by a cloud of on - site virtual vibrations and can be considered as an excitation - polaron .
namely , any electronic transition is accompanied by excitation and annihilation of ground and excited state vibrational modes . here
we assume the excitation - vibration coupling parameters to be site independent , by defining @xmath490 and @xmath491 . in terms of the new operators
, the excitation hamiltonian reads @xmath492 , where @xmath493 , with @xmath494 . in this limit
the effect of the on - site excitation - vibration coupling is a simple renormalization of the electronic excitation energy by a shift of @xmath495 , where @xmath496 .
this shift might be quite small but it needs to be considered for precision measurements as in a lattice clock transition energy @xcite .
the vibration hamiltonian is now given by @xmath497 .
the excitation - vibration coupling hamiltonian of the transfer part is written as @xmath498\ \tilde{b}_i^{\dagger}\tilde{b}_j$ ] .
this term describes vibration changes induced by excitation transfer among different sites .
four possible processes are plotted in fig . [ 29 ] .
they show the emission and absorption of a vibration due to the excitation transfer among nearest neighbor sites . with an energy transfer accompanied by the emission of ground state vibration at site @xmath456 .
process ii ( dashed - line ) for @xmath499 with an energy transfer accompanied by the emission of excited state vibration at site @xmath8 .
( b ) process iii ( full - line ) for @xmath500 with an energy transfer accompanied by the absorption of ground state vibration at site @xmath8 .
( dashed - line ) for @xmath501 with an energy transfer accompanied by the absorption of excited state vibration at site @xmath456.,title="fig:",width=245 ] with an energy transfer accompanied by the emission of ground state vibration at site @xmath456 .
process ii ( dashed - line ) for @xmath499 with an energy transfer accompanied by the emission of excited state vibration at site @xmath8 .
( b ) process iii ( full - line ) for @xmath500 with an energy transfer accompanied by the absorption of ground state vibration at site @xmath8 .
( dashed - line ) for @xmath501 with an energy transfer accompanied by the absorption of excited state vibration at site @xmath456.,title="fig:",width=245 ] in terms of transformed operators the interaction hamiltonian can then be transformed into a momentum space representation in terms of exciton operators , with the exciton hamiltonian @xmath502 , for two dimensional optical lattices .
the vibration operators , even though representing on - site localized vibrations , can be rewritten equally in the momentum space by applying @xmath503 and @xmath504 , to get @xmath505 , which have flat dispersions , namely @xmath506-independent .
the exciton - vibration coupling now reads @xmath507 where @xmath508 .
the hamiltonian @xmath461 describes scattering of excitons between different wave vectors involving the emission and absorption of a ground or excited state vibrational quantum .
such a process is possible in the limit of @xmath509 , where the exciton band width is larger than the vibration energy .
the scattering conserves energy and momentum , and as the vibration dispersion is a flat one , the vibration can absorb any amount of momentum from the exciton .
the exciton transition rate within a spontaneous emission is @xmath510 , which results of the fermi golden rule .
the exciton - vibration interaction will serve as an important source for relaxation of excitons toward lower energies of the exciton band and therefore a thermal equilibrium in optical lattices .
the exciton - vibration coupling yields polariton - vibration coupling @xcite . in using the inverse previous transformation @xmath511
, we get the polariton - vibration interaction by @xmath512 where the exciton amplitudes indicate that the interaction is due to the polariton excitonic parts .
this interaction describes scattering of polaritons between states with different wavevectors by the emission or absorption of vibrational quanta . as
the vibrations are dispersionless , one need to take care only of the energy conservation .
a scattering takes place between different momentum states in the lower and the upper polariton branches .
but if the vibration energy equals the splitting energy between the two branches , the polaritons can jump between the two branches by the emission or absorption of vibrations .
the polariton - vibration interaction can represent a significant source for polaritons relaxation toward the minimum energy at @xmath337 . using the fermi golden rule ,
the damping rate for the spontaneous emission of vibrations at zero temperature @xmath0 off polaritons at the lower branch is given by @xmath513 .
the appearance of the exciton amplitudes indicate that the process is much more efficient in regions where the polaritons are more excitonic than photonic .
the appearance of some defects is unavoidable due to imperfections in the dynamical formation of the mott insulator phase , in particular there can be missing or extra atoms at some sites .
we investigate here the effect of such defects on the dynamical properties of excitons and cavity polaritons @xcite .
we concentrate on the case of a low defect number , where the exciton and cavity polariton picture still holds and these quasi - particles are only scattered by single defects .
namely , the distance between each two defects is large enough for the formation of coherent excitons and cavity polaritons , which propagate as free quasi - particles between each two scattering processes . in principle
the defects can move by hopping among the lattice sites , but the corresponding time scale is so long that they can be considered frozen . here
we concentrate on one type of defect which is induced by a missing atom , that in the case of one atom per site corresponds to the presence of vacant sites . in calculating the exciton and cavity polariton elastic scattering amplitude of such defects ,
we show that vacancies behave like hard sphere scatterers .
we suggest cavity polaritons as a tool to detect such defects and show how the scattering can be controlled by changing the exciton - photon detuning .
we treat a two dimensional optical lattice filled with one atom per site in the mott - insulator phase .
a single atom missing at the origin ( @xmath514 ) then creates an impurity in the artificial lattice of ultracold atoms , which for sufficient lattice depth stays localized and will not hop among the lattice sites as shown in fig .
we study the scattering of excitons off such a missing atom .
the addition of the impurity to the ideal atom crystal needs only a small change to the system hamiltonian @xmath515 presented in subsection 3.3 , whose components now read @xmath516 and @xmath517 , where @xmath518 represents free excitons , and @xmath519 is the exciton dispersion with in - plane wave vector @xmath114 .
@xmath365 is the impurity hamiltonian at the origin . to get the above form
, we add and subtract the impurity hamiltonian @xmath365 to the whole hamiltonian @xmath520 , that is @xmath521 , and then we define the ideal case hamiltonian by @xmath522 . for an exciton the impurity thus appears as a potential well located at the origin , with depth @xmath523 and radius @xmath49 .
we will not try to find the self - consistent eigenstates of @xmath520 and consider only the scattering problem . because we can neglect trapping of an exciton in the impurity potential as a vacant site can not absorb the trapping energy , we consider only the scattering process of an exciton off the impurity and calculate its scattering amplitude .
the initial exciton initially with wavevector @xmath114 very far from the impurity , is scattered elastically into the wavevector @xmath524 also very far , with @xmath525 . as the impurity is very deep with depth @xmath50 , in order to calculate the scattering amplitude we need to proceed beyond the born approximation even though the perturbation is confined to a radius @xmath49 .
we apply the schwinger - lippmann equation @xcite in the case of small wave vector scattering excitons , that is @xmath100 . for this case we can use the parabolic approximation for the dispersion of the excitons , @xmath526 , with
@xmath527 the exciton effective mass , and for isotropic atoms in a square lattice of cubic symmetry as in subsection 3.3 @xmath528 .
we obtain the scattering amplitude @xmath529},\ ] ] where the effective exciton band width is defined by @xmath530 .
the dipole - dipole interaction energy between different sites in the optical lattice is small , and hence the excitons have a small band width . in the present case
we have the limit of @xmath531 so that the scattering amplitude is @xmath532 , which exactly reproduces the result for the scattering off a hard disk of radius @xmath49 with the scattering cross section defined by @xmath533 .
we thus conclude that an impurity generated by a missing atom in an optical lattice acts effectively like a hard disk of radius @xmath49 .
hence if a large number of incident excitons with identical wave vector are scattered elastically off the impurity , we get a ring of radius @xmath84 of scattered excitons .
the scattering amplitude is plotted in fig . [ 31 ] as a function of wave vectors , @xmath84 , for lattice constant @xmath273 .
the singularity at @xmath337 is the two dimensional signature .
, for zero detuning and a lattice constant @xmath273 .
figure reprinted with permission from @xcite copyright 2008 by institute of physics.,width=245 ] we now add cavity mirrors to our lattice as described in subsection 4.2 , where in the strong coupling regime , the system eigenstates are the cavity polaritons . as
a polariton is a coherent superposition of an exciton and a photon it can be scattered off an impurity due its excitonic part .
the corresponding hamiltonian then reads @xmath534 and @xmath535 , where the upper and lower polariton branches are presented explicitly in subsection 4.2 .
we calculate the polariton scattering off such impurity .
again the scattering is elastic so that an incident polariton in branch @xmath361 with wave vector @xmath114 will scatter into a polariton with wave vector @xmath524 in the same branch .
we consider the scattering of small wave vector polaritons , that is in the limit @xmath204 in the lower branch . in this limit
the polariton dispersion can considered approximately parabolic with a polariton effective mass of @xmath536 .
this is of the order of the cavity photon effective mass @xmath537 .
hence , the lower polariton branch dispersion is taken to be @xmath538 , and we have in general @xmath539 .
the scattered states will also have small wavevectors due to the energy conservation . as the upper branch has higher energies
its contributions are negligibly small .
in addition the upper branch for larger wavevectors is mainly photonic with small excitonic part , and thus it contributes only weakly to the impurity scattering .
we consider the case around zero detuning between the excitons and photons .
therefore the excitonic weight @xmath540 will change from half around zero wave vector up to one for large wave vectors , where the lower branch became excitonic .
because we neglect the contribution of the upper polariton branch , we drop the branch index , and all the parameters will be only for the lower branch . for simplicity
, we use a model for the lower branch dispersion in place of the real one .
the lower branch is divided into two parts : the first part between @xmath541 is taken to be of a parabolic dispersion with a polariton effective mass , where @xmath542 ; and the second part between @xmath543 , where @xmath163 is the brillouin boundary , is taken to be dispersionless with energy equal to the exciton energy at zero wave vector , with @xmath544 .
the intersection point between the two parts is fixed by @xmath545 .
the calculation yields the scattering amplitude @xmath546 where we defined the polariton effective band width @xmath547 , and @xmath548 . as @xmath549 , we have @xmath550 , and the scattering amplitude is negative , that is @xmath551 .
thus the above impurity effective potential is repulsive for the polaritons . in the case of zero detuning ,
that is @xmath552 , for small wavevectors @xmath553 , @xmath554 of the order of the exciton - photon coupling , @xmath555 , where @xmath556 , we get @xmath557 .
this scattering amplitude equals that of the scattering from an effective potential of a square barrier potential of height @xmath554 and width @xmath49 . for the case of negative detuning , in the limit of @xmath558 , where @xmath554 is of the order of the exciton - photon detuning , the scattering amplitude is @xmath559 .
this scattering amplitude equals that of the scattering from an effective potential of a square barrier potential of height @xmath560 and width @xmath49 .
hence we find that the scattering amplitude can be controlled by changing the exciton - photon detuning
. polaritons . figure reprinted with permission from @xcite copyright 2008 by institute of physics.,width=245 ] using the following parameters : atomic energy @xmath561 , exciton - photon coupling @xmath562 , lattice constant @xmath273 , approximately normal incident waves with @xmath563 , the scattering amplitude is calculated as a function of the exciton - photon detuning , as reported in fig .
it is clear that maximum scattering is obtained for zero detuning . as the detuning increases ,
either positive or negative , the @xmath103 polaritons become more photonic and they are much less scattered off the impurity . as the atomic energy @xmath186 is much larger than the exciton - photon coupling energy @xmath564 , the scattering amplitude shows only a negligibly small asymmetry for negative and positive detunings , but we expect a clear resonance of the scattering amplitude around a resonant value @xmath565 . considering a stream of incident polaritons with the same wave vector @xmath566 we then should see a ring of radius @xmath84 of scattered polaritons .
the effect of defects on the dynamics of excitons in optical lattices can be generalized to the case of the mott insulator phase with several atoms per site that contain defects of extra or less atoms at specific sites , as we presented for the case of two atoms per site with a defect of a singly occupied site .
excitons and polaritons in an optical lattice filled with ultracold atoms in the mott insulator phase represent an idealized form of an optically excitable solid state crystal .
modeling by a two component bose - hubbard model allows us to derive the quantum phase diagram identifying the regions of mott insulator phases .
excitons occur as natural elementary excitations in such a system including energy transfer induced by resonant dipole - dipole interaction .
we extract the condition for the appearance of excitons in such artificial crystals by comparing between the transfer parameter and the excited atom decay .
excitons in the mott insulator phase are collective delocalized electronic excitations in the lattice with their momentum distributed among many lattice sites . propagating excitons are associated to large systems , while systems with boundaries require standing wave excitons . in general their damping rates are large and deviate from single atom decay , where excitons can be metastable or superradiant depending on their wave vectors and polarizations .
dark or metastable excitons open the door for storage and manipulation of photonic qubits with a wide range of applications for quantum information processing
. coherent transfer of excitons among different optical lattices differs from hopping of electronic excitations among two separated atoms . here
the transfer amplitudes decay exponentially with the inter - lattices distance .
lattice excitons can be used as active medium for a cavity qed in the resonant regime .
cavity photons affect the boundaries between quantum phases and in these systems the mott insulator phases appear only for deeper optical lattices .
this gives new tools to observe and control the quantum phase transitions via the average number of cavity photons . in the strong coupling regime
the excitons and photons are coherently coupled to form new quasi - particles termed polaritons .
optical linear spectra are presented as a nondestructive observation tool for the system different quantum phases .
systems of various dimensionality and geometries are investigated , ranging from a finite chain between spherical cavity mirrors up to a two - dimensional optical lattice between planar cavity mirrors .
moreover , the formation of molecules in optical lattices results in an anisotropic crystal that can be manifested through cavity a photon polarization mixing in the linear spectra .
we presented also a set - up of two parallel one - dimensional optical lattices formed by an optical nanofiber and that provides a tool for a direct excitation of one - dimensional excitons through the coupling to fiber photons .
moreover the system gives rise to one - dimensional polaritons that dominate the optical properties in the strong coupling regime .
we suggested a new mechanism for the excitation and de - excitation of atoms into higher bloch bands through the energy transfer process , and we presented it as a significant source for the thermalization of excitons and polaritons . the elastic scattering of excitons and polaritons off vacancies are used for verifying the presence of various defects that can appear naturally in optical lattices .
the discussion in the present review though presented for optical lattice ultracold atoms can be adapted directly for various systems containing lattices of active materials .
for example , one can imagine a lattice of semiconductor nano - structures , e.g. , chains of quantum dots , arrays of color centers in solids , organic molecules ordered on a matrix , and rydberg atoms in a lattice .
the results can be extended into various directions , just to mention the implementation of excitons and polaritons for the physical realization of quantum information processing .
the parameter controllability and the wide range of lifetimes of excitons make them a strong candidate as qubits .
but efficient quantum processing implies entangled qubits , which are achievable via interaction among such qubits .
therefore our results should be extended to exploit exciton - exciton interactions through saturation and coulomb forces , which also can find application for photonics using optical lattices .
we expect this research to guide future experiments of electronically excited optical lattice ultracold atoms both in free apace and for cavity qed .
the work was supported by the austrian science fund ( fwf ) through the lise - meitner program ( m977 ) and the stand - alone project ( p21101 ) , and by the darpa quasar program .
bajcsy , m. , hofferberth , s. , balic , v. , peyronel , t. , hafezi , m. , zibrov , a. s. , vuletic , v. , lukin , m. d. , 2009 .
efficient all - optical switching using slow light within a hollow fiber .
102 , 203902 .
bakr , w. s. , peng , a. , tai , m. e. , ma , r. , simon , j. , gillen , j. i. , folling , s. , pollet , l. , greiner , m. , 2010 . probing the superfluid - to - mott insulator transition at the single - atom level .
science 329 , 547 .
goban , a. , choi , k. s. , alton , d. j. , ding , d. , lacroute , c. , pototschnig , m. , thiele , t. , stern , n. p. , kimble , h. j. , 2012 .
demonstration of a state - insensitive , compensated nanofiber trap .
109 , 033603 .
nayak , k. p. , melentiev , p. n. , morinaga , m. , kien , f. l. , balykin , v. i. , hakuta , k. , 2007 .
optical nanofiber as an efficient tool for manipulating and probing atomic fluorescence .
express 15 , 5431 .
vetsch , e. , reitz , d. , sague , g. , schmidt , r. , dawkins , s. t. , rauschenbeutel , a. , 2010 . optical interface created by laser - cooled atoms trapped in the evanescent field surrounding an optical nanofiber .
104 , 03603 .
viteau , m. , bason , m. g. , radogostowicz , j. , malossi , n. , ciampini , d. , morsch , o. , arimondo , e. , 2011 .
rydberg excitations in bose - einstein condensates in quasi - one - dimensional potentials and optical lattices .
107 , 060402 .
|
ultracold atoms uniformly filling an optical lattice can be treated like an artificial crystal .
an implementation including the atomic occupation of a single excited atomic state can be represented by a two - component bose - hubbard model .
its phase diagram at zero temperature exhibits a quantum phase transition from a superfluid to a mott insulator phase .
the dynamics of electronic excitations governed by electrostatic dipole - dipole interactions in the ordered regime can be well described by wave - like collective excitations called excitons . here
we present an extensive study of such excitons for a wide range of optical lattice geometries and of different dimensionality including boundary effects in finite lattices . as they are coupled to the free space radiation field
, their decay depends on the lattice geometry , polarization and lattice constant .
their lifetimes can vary over many orders of magnitude from metastable propagation to superradiant decay .
particularly strong effects occur in one dimensional atomic chains coupled to tapered optical fibers where free space emission can be completely suppressed and only a superradiant interaction with the fiber modes takes place .
we show that coherent transfer of excitons among spatially separated optical lattices can be controlled and represents a promising candidate for quantum information transfer . for an optical lattice within a cavity the excitons
are coupled to cavity photons and the resulting collective cavity qed model can be efficiently formulated in terms of polaritons as elementary excitations .
their properties are explicitly calculated for different lattices and they constitute a non - destructive monitoring tool for important system properties as e.g. the atomic quantum statistics .
even the formation of bound states and molecules in optical lattices manifests itself in modified polariton properties as e.g. an anisotropic optical spectrum .
partial dissipation of the exciton energy in the lattice leads to heating , which can be microscopically understood through a mechanism transferring atoms into higher bloch bands via a resonant excitation transfer among neighboring lattice sites .
the presence of lattice defects like vacancies in the mott insulator induces a characteristic scattering of polaritons , which can be optically observed to monitor the lattice integrity .
our models can be applied to simulate and understand corresponding collective phenomena in solid crystals , where many effects are often masked by noise and disorder .
ultracold atoms , optical lattices , quantum phase transitions , bose - hubbard model , mott insulator phase , excitons , dark states , superradiance , cavity qed , cavity polaritons , tapered nanofibers 37.10.jk , 42.50.-p , 71.35.-y , 71.36.+c h zoubi moved recently to : _ max - planck institute for the physics of complex systems , noethnitzer str .
38 , 01187 dresden , germany .
|
the electron component in the primary cosmic radiation ( cr ) consists of both electrons and positrons . a number of measurements of their energy spectrum by balloon - borne experiments @xcite and by a single space mission @xcite have shown that the electron intensity , about 1% of the proton intensity at 10 gev , decreases more rapidly with energy than that of cosmic - ray nuclei .
although existing data exhibit sizeable statistical and systematic uncertainties , it is clear that the power - law index of the electron spectrum above 10 gev is larger than 3.0 , in contrast to the proton index of 2.7 .
high - energy electrons in the galactic magnetic field lose energy through synchrotron radiation and by inverse compton collisions with low - energy photons .
the measurement of the energy spectrum of primary electrons can give information on the galactic magnetic field as well as on the confinement time of electrons in the galaxy and on the distribution of particle sources in our galaxy .
positrons are believed to be mainly created in secondary production processes resulting from the interaction of cr nuclei with the interstellar medium .
early measurements @xcite capable of separating electrons and positrons have shown that the positron fraction @xmath1 is around 10% in the 1 - 100 gev region , even though this data deviate significantly from predictions of secondary production models . in this paper
we report details of the analysis procedure and energy spectrum reconstruction using of data collected in the first six months of operation by the large area telescope ( lat ) instrument on board the fermi satellite @xcite .
the lat is a gamma - ray telescope based on the conversion of gamma - rays into electron - positron pairs and is arranged in a @xmath2 array of 16 identical towers .
each tower consists of a silicon micro - strip detectors tracker ( tkr ) followed by a segmented csi calorimeter ( cal ) , to reconstruct the gamma - ray direction and energy . a custom - designed data acquisition module ( tem , tower electronics module )
is located below the calorimeter .
tracker towers are covered by a segmented scintillator anti - coincidence system ( acd ) , consisting of panels of plastic scintillator read out by wave - shifting fibers and photo - multiplier tubes , used to reject cosmic charged particle background ( electrons , protons , heavier nuclei ) .
the acd is also covered by a thermal blanket and micrometeorite shield .
an aluminum grid supports the detector modules and the data acquisition system and computers , which are located below the cal modules .
the lat detector is described in more detail in @xcite .
each of the tracker tower modules is composed of a stack of 19 stiff , lightweight carbon composite panels called trays : one bottom tray , 17 mid trays , and one top tray .
a tray panel is a composite structure : an aluminum honeycomb core is closed by two carbon - fiber face sheets , bonded to four carbon - carbon close - outs to form a sandwich structure for mounting payload . on the top and bottom surfaces of the tray two kapton flexible circuits are glued to supply the bias potential to the silicon microstrip detectors ( ssds ) .
the silicon detectors are bonded on both sides of a panel , with the strips on the top layer parallel to those on the bottom .
each layer hosts 16 ssds in an array of four ladders , each one consisting of four wafers glued head to head and bonded together to form a single detector assembly .
an array of tungsten ( w ) converter foils is glued to the bottom surface of all but the three lowest trays , between the panel and the flexible circuits and detectors , to match the active area of each wafer .
the first 12 layers of tungsten are each 2.7@xmath3 radiation length ( r.l . ) in thickness , while the following four layers are 18% r.l .
thick to increase the photon detection efficiency in the gev and above energy range .
only the top and bottom trays support silicon detectors on one side only .
each tracker module has 18 x , y tracking planes , consisting of 2 layers ( x and y ) of single - sided silicon strip detectors .
the total thickness of the tkr on - axis is 1.5 r.l .
each cal module is composed by 96 csi(tl ) crystals ( @xmath4 ) supported by a carbon fiber composite structure and read out from both ends with silicon pin photodiodes .
the crystals are optically isolated from each other and are arranged horizontally in 8 layers of 12 crystals each .
the total on - axis depth of the cal is 8.6 r. l. each calorimeter module layer is aligned @xmath5 with respect to its neighbors , forming an x , y ( hodoscopic ) array .
the level of segmentation is sufficient to allow spatial imaging of the shower and accurate reconstruction of its direction because each csi crystal provides three spatial coordinates for the energy deposited within : two discrete coordinates from the physical location of the crystal in the array and the third , more precise , coordinate determined by measuring the light yield asymmetry at the ends of the crystal along its longitudinal dimension .
each crystal end is equipped with two photodiodes : a large one with cross section @xmath6 , and a smaller one with cross section @xmath7 .
each photodiodes is readout with two front end electronic channels with different gain , to cover the large dynamic range of energy deposition in the crystal .
the large photodiodes cover the range 2 mev - 1.6 gev , while the small photodiodes cover the range 100 mev - 100 gev .
a calibration charge injection signal can be fed directly to the front end of the pre - amplifiers @xcite .
the longitudinal segmentation enables energy measurements . from the longitudinal shower profile
, we derive an unbiased estimate of the initial electron energy by fitting the measurements to an analytical description of the energy - dependent mean longitudinal profile @xcite .
the lat trigger system takes primarily inputs from the acd , tkr , and cal detectors .
in addition , a periodic trigger is also configurated to perform the charge injection calibration . after an event passes the hardware trigger it is inspected by on - board software filters ,
each configured to identify events likely to be useful for one or more scientific or calibration purposes .
there are 4 types of filters : * gamma : the purpose is to select gamma - ray candidates and events that deposit at least 20 gev in the cal ; * heavy ion : the purpose is to perform calibration on high - energy scales ; * mip : the purpose is to select non iteracting charged particles ( protons ) ; * diagnostic : the purpose is to select an unbiased event sample for filter and background performance studies the lat on - board filters can be configured .
if any filter accepts an event , it is included in the lat data stream and forwarded to the solid state recording for transmission to the ground .
the lat event data contains tkr , cal and acd sub - system information , for istance tkr hit strips and time over threshold ( tot ) , cal and acd amplitude signals , and trigger primitive .
the nominal science operation on orbit includes the gamma , diagnostic and heavy ion filters . in particular , since the gamma filter is configured to accept all events that deposit at least 20 gev in the calorimeter , high - energy events , including electrons , are available for analysis on the ground . to identify the cr electron component
, we first used the acd signal to select charged particle events that penetrated the detector .
as for the analysis developed for extracting lat photon data @xcite , the electrons selection essentially relies on the lat capability to discriminate electromagnetic ( em ) and hadronic showers based on their longitudinal and lateral development , as measured by both the tkr and cal detectors .
em showers start developing in the tkr , while most of the energy is absorbed in the cal .
the measurement of the lateral shower development is a powerful discriminator between more compact em showers and wider hadronic showers .
the cuts have been tuned using a mc data sample of pure electrons and a realistic cr nuclei model @xcite .
we select on variables ( measured or derived quantities ) that map the distribution of tkr clusters around the main track , and in the cal second - order moments of the energy distribution around the shower axis .
a further selection derives from the different distributions of energy and hits in the acd between em and hadron - initiated showers .
similarly to the lat photon background rejection analysis , the remaining necessary boost in the rejection power is obtained by combining two probability variables that result from training classification trees ( ct ) to distinguish between em and hadron events @xcite .
two cts are used , one built with tkr variables , and a second one based on cal variables , which describe the complete event topology .
the variables given most weight by the cts are the same or equivalent to those described above .
the classifiers allow selection of the electrons through a multitude of parallel paths , each with different selections , that map the many different topologies of the signal events into a single , continuous probability variable that is used to simultaneously handle all valid selections .
the tkr and cal electron probabilities are finally combined to create an energy - dependent selection that identifies electrons with greater efficiency and optimized background rejection with respect to a single sequence of cuts .
the cuts used in the analysis are : 1 .
@xmath8 : gammafilter + at least 20 gev energy deposit in the cal + trigger coincidence from acd , tkr and cal + at least one good track reconstructed in the tkr with angle with to respect the lat z - axis less than 72 deg + at least 7 r. l. in the cal + reconstucted angle with respect to local zenith less than 105 deg to reject eventual earth albedo photons + at least 1 acd tile fired 2 .
@xmath9 : @xmath8 + energy deposit on acd tiles equivalent to a single charged particle 3 .
@xmath10 : @xmath9 + large number of tkr clusters in a region around the good track and the average tkr time over threshold ( tot ) larger than the single charged particle value 4 .
@xmath11 : @xmath10 + narrow energy distribution in the cal csi crystals around the shower 5 .
@xmath12 : @xmath11 + ct analysis ( see text).,width=240 ] the data sample presented here has been collected from aug .
4th , 2008 to jan .
31st , 2009 in sky survey scanning operational mode . fig .
[ fig1 ] shows the fraction of events as function of the reconstructed energy and cuts with respect to the @xmath8 .
the @xmath9 , @xmath10 and @xmath11 remove a large number of events in all energy regions , while the @xmath12 improves the electron / hadron separation above 100 gev @xcite .
it is worth to point out that event selection so far is not optimized versus the incident angle of incoming particles , even though it is explicitly energy - dependent to suppress the larger high - energy background .
the residual hadronic background in the final data sample has been estimated by means of a mc full simulation folding cosmic ray ( background model ) that the lat encounters in space .
the background model includes primary charged cosmic rays and earth albedo photons @xcite .
the residual hadronic background has been estimated from the average rate of hadrons that survive electron selection in the simulations .
fig [ fig2 ] shows the energy distribution for final candidate electrons ( open circles ) and the background ( hadron ) contamination ( filled squares ) .
the hadron contamination ranges from about 4% at 20 gev to 21% at 1 tev .
more than 4 m candidate electron events above 20 gev have been selected , and 544 events before background subtraction have been collected in the last energy bin ( 770 gev-1tev ) . the final electron data sample has been evaluated subtracting bin by bin the background events from the candidate electrons ( filled circles in fig .
[ fig2 ] ) .
events considered for the electron analysis require to fail the acd vetoes developed to select photon events @xcite .
this removes the vast majority of the potential gamma - ray contamination .
in fact , the geometry factor for photons , determined with electron selection cuts , is less than 8% of that for electrons at 1 tev .
the ratio between photon and electron fluxes is negligible at low energies and rises to around 20% at the highest energy .
this estimate is obtained from a simple , conservative extrapolation to high energies of the egret all - sky spectrum at gev energies using a @xmath13 power law @xcite .
the overall gamma contamination in the final electron sample is therefore always less than 2% .
this contamination is not subtracted from the final electrons data sample .
the electron / positron flux is evaluated by correcting the observed number of events with the instrument response function ( irf ) .
the electron irf ( also called smearing matrix ) is evaluated from monte carlo simulation , in which a trial spectrum of input pure electrons is simulated .
the smearing matrix describes the full migration from the `` true '' energy bins ( mc energy ) to the reconstructed ones .
the elements of the smearing matrix represent the probabilities to observe a given effect that falls in a reconstructed energy bin from a cause in a given true energy bin .
the smearing matrix also describes the losses of events due to cuts applied to extract the signal , which occur due to finite acceptance ( geometric factor ) of the detector .
the electron efficiency factor is computed as the ratio between the reconstructed number of events after applying the cuts and the number of generated events as function of the true energy .
[ fig3 ] shows the electron acceptance factor as function of true energy evaluated multiplying the efficiency for the area and solid angle used for the generation of the monte carlo data sample . in absence of energy dispersion , i.e. the true and the reconstructed energy are the same ,
the flux could be evaluated simply by dividing the number of observed counts by the detector acceptance . on the other hand , to include the finite energy dispersion effect the flux is evaluated using a deconvolution ( unfolding ) method .
the deconvolution method adopted in the present analysis is based on bayes theorem @xcite .
the unfolding analysis allows to reconstruct the number of events belonging to each bin of true energy . the integral flux in the @xmath14 energy bin ( in units of events per unit area , solid angle and unit time )
is then evaluated dividing the reconstructed events in the bin for the area of the sphere used for the generation of the monte carlo data sample , solid angle and for the live time . the differential flux ( in units of events per unit area , solid angle , unit time and unit energy )
is then evaluated dividing the integral flux for the corresponding energy bin width .
[ fig4 ] shows the differential electron / positron flux as function of energy in the range from 24 gev to 1 tev .
the overall lat - fermi electron spectrum can be fitted with a simple power law model , @xmath15 , with spectral index @xmath16 .
as function of energy .
open circles : average energy in each is taken for @xmath17 ; filled circles : median energy in each bin assumig a power law distribution with spectral index @xmath18 is taken for @xmath17 .
only statistic errors are shown.,width=240 ] fig .
[ fig5 ] shows the differential energy spectrum multiplied by @xmath17 in order to display eventual features that are otherwise difficult to discern in the fig .
two kind of energy values have been considered for calculating @xmath17 : the mean energy in each bin ( open circles ) ; and median energy in each bin assumig a power law distribution with spectral index @xmath18 ( filled circles ) , since the overall energy spectrum follows a simple power law model with spectral index close to 3 .
the relative differences of the mean bin center with respect to the median to evaluate @xmath19 is less than 8% at 1 tev .
the systematic uncertanties assessment provides a maximum error less than 20% at 1 tev @xcite . the absolute lat energy scale , at this early stage of the mission ,
is determined with an uncertainty of @xmath20 .
the systematic uncertanties are not shown in the energy spectrum .
the primary cosmic - ray electron / positron energy spectrum has been measured from 20 gev to 1 tev using six months of lat - fermi instrument .
a detailed comparison of this result with the ones reported of many groups over the years is presented in ref .
an interpretation of the results is presented in ref .
@xcite .
99 j. nishimura et al . , astrophys .
j. , 238 , 394 ( 1980 ) ; j. nishimura et al . , adv .
space res . , 19 , 767 ( 1997 ) ; t. kobayashi et al . ,
j. 601 ( 2004 ) 340 m. aguillar et al .
366 ( 2002 ) 331 o. adriani et al .
, arxiv:0810.4995 a. a. abdo et al . ,
_ measurement of the cosmic ray @xmath21 spectrum from 20 gev to 1 tev with the fermi large area telescope _ ,
* 102 * ( 2009 ) 181101 w.b .
atwood et al . , _ the large area telescope on the fermi gamma - ray space telescope mission _ , apj 697 ( 2009 ) 1071 , arxiv:0902.1089v1 [ astro-ph.im][arxiv:0902.1089v1 [ astro-ph.im ] ] a.a .
abdo et al . _
the on - orbit calibration of the fermi large area telescope _ submitted to astrophys .
j. ; arxiv:0904.2226 c. sgr , j. bregeon and l. baldini _ instrument simulation for the analysis of cosmic ray electrons with fermi lat _ , these proceedings p. sreekumar et al .
, astrophys .
j. * 494 * ( 1998 ) 523 g. dagostini , nucl .
methods in phys .
a 362 * ( 1995 ) 487 l. latronico _ measurement of the cosmic ray electron spectrum from 20 gev to 1 tev with the fermi large area telescope _ , these proceedings d. grasso _ astrophysical interpretations of the high energy cosmic ray electron spectrum measured with the fermi space telescope _ , these proceedings
|
the large area telescope ( lat ) instrument on board the fermi satellite consists of a multi - layer silicon - strip tracker interleaved with tungsten converters ( tkr ) , followed by a csi crystal hodoscopic calorimeter ( cal ) .
sixteen tkr and cal modules are assembled in a 4@xmath04 array .
a segmented anticoincidence plastic scintillator ( acd ) surrounds the tkrs .
the primary cosmic - ray electron / positron energy spectrum has been measured from 20 gev to 1 tev using a dedicated event analysis that ensures efficient electron detection and reduced hadron contamination .
results from detailed monte carlo simulations have been used to reconstruct the observed energy spectrum to the primary cosmic ray spectrum .
we present here details of the analysis procedure and the energy spectrum reconstruction .
cosmic rays ; spectral analysis ; electrons .
|
the type ia supernova remnant g327.6 + 14.6 , or sn 1006 , is the remnant of the explosion which took place in ad 1006 .
the distance to this source is between 1.4 and 2.1 kpc ( green 1988 ) .
its mean expansion rate has been measured as @xmath1 arcsec / yr , which implies a present expansion speed of @xmath2 km / s for an average distance of 1.8 kpc . the time dependence of the expansion ( @xmath3 ) is consistent with sedov expansion , or with a forward / reverse shock pair moving into constant - density material ( moffett , goss , & reynolds 1993 ) .
the asca detection of power law x - ray emission from the bright northeastern ( ne ) and southwestern ( sw ) rims of sn 1006 led koyama et al .
( 1995 ) to infer the existence of electrons accelerated by the first order fermi mechanism up to energies of @xmath4 tev .
the detection of tev @xmath5-rays from this system will prove the existence of such ultrarelativistic electrons ( de jager et al .
1995 , mastichiadis 1996 , pohl 1996 ) . in section 2 we will compile the non - thermal radio to x - ray spectrum , which may be due to synchrotron emission from these electrons , and in section 3 we will compare the acceleration and loss timescale for electrons accelerated by the first order fermi process , and show that the resulting maximum electron energy predicts a synchrotron cutoff frequency which is consistent with observations .
the best - fit spectrum also leads to the derivation of the magnetic field strength @xmath6 associated with the shell . in section 4
we will calculate the inverse compton ( ic ) @xmath5-ray spectrum associated with this best - fit model spectrum from section 3 .
the detection of tev @xmath5-rays from sn 1006 will enable the observer to determine the magnetic field strength , and an important parameter associated with fermi acceleration from the observed @xmath5-ray flux .
the radio morphology shows two bright arcs towards the ne and sw ( see e.g. reynolds & gilmore 1993 ) , with a composite spectrum which is given by @xmath7 and shown in fig .
1 . the individual radio flux measurements for the total emission from sn 1006 ( reproduced from fig . 1 of reynolds 1996 )
are also shown for comparison .
rosat imaging observations of sn 1006 shows that the soft x - ray remnant is dominated by thermal ( @xmath8 kev ) emission towards the se , whereas the x - ray counterparts of the ne and sw radio rims ( or hard aegis " ) are characterized by a power law spectrum which dominates the total emission from the remnant above @xmath9 kev ( willingale et al . 1996 ) .
spatially resolved spectral results by asca have shown that these bright rims are responsible for @xmath10 of the total x - ray flux from sn 1006 above 1 kev , and the interior of the remnant also contains this non - thermal emission , which dominates the thermal component above 2 kev ( koyama et al .
1995 ) . the results from earlier non - imaging instruments are therefore useful to obtain the non - thermal spectrum above 1 kev from the entire remnant .
for example : tenma ( koyama et al .
1987 ) , exosat ( jones & pye 1989 ) , and ginga ( ozaki et al .
1994 ) obtained well - constrained spectral energy indices ranging between @xmath112.0 and @xmath112.3 for energies between 1.5 kev and @xmath12 kev .
this is confirmed by the energy index of @xmath13 of the ne rim found by asca ( koyama et al .
the exosat me field of view was just large enough to cover the entire remnant , without significant background contamination , and the energy spectrum above 1.5 kev ( derived from the given energy flux ) is given by @xmath14 which is shown together with the radio spectrum in fig . 1 .
the change from a @xmath110.56 radio energy index to a @xmath15 x - ray energy index above @xmath9 kev is indicative of a spectral turnover or break around @xmath16 kev , where the spectral energy index should be around @xmath111 ( depending on the sharpness of the cutoff ) .
in fact , soft x - ray imaging observations of the hard aegis by the einstein sss ( becker et al .
1980 ) and exosat le ( jones & pye 1989 ) resulted in a consistent spectral energy index of @xmath111.2 .
willingale et al .
( 1996 ) used rosat pspc observations to separate the thermal and non - thermal energy fluxes of the hard aegis and the rest of the remnant , with the same finding as koyama et al .
( 1995 ) that the non - thermal soft x - ray flux from the hard aegis contributes to @xmath10 of the total non - thermal soft x - ray flux of @xmath17 ergs/@xmath18/s ( for the 0.1 kev to @xmath19 kev range ) .
assuming the average spectral index of @xmath111.2 found by einstein and exosat ( willingale et al .
1996 obtained a spectral index for the sw limb only ) , but using the rosat normalization for the total non - thermal flux from sn 1006 , the resulting energy spectrum for this energy range is therefore @xmath20 and is indicated by its spectral index in fig .
the non - thermal synchrotron emission of sn1006 from shock accelerated electrons has been discussed by reynolds & chevalier ( 1981 ) , ammosov et al .
( 1994 ) and , more recently , by reynolds ( 1996 ) . here
we apply the method given in mastichiadis ( 1996 - henceforth m96 ) .
this allows us to treat the problem in a self - consistent manner by calculating the electron distribution function at each instant in time from the solution of a time - dependent kinetic equation for electrons .
the physical picture is according to the ` onion - shell - model ' ( bogdan & vlk 1983 ) as applied to time - dependent acceleration ( ball & kirk 1992 ) .
therefore we assume that electrons are accelerated in the ( parallel ) shock wave of the expanding supernova remnant and at each instance there is a flux of relativistic electrons which escape downstream and subsequently radiate . for these
we assume to have a power law spectrum in energy , i.e. @xmath21 with @xmath22 and an exponential cut - off for @xmath23 ( webb et al .
for the determination of the normalization factor we followed m96 ( after an original suggestion by drury 1992 ) and set @xmath24 where @xmath25 is the radius of the supernova and @xmath26 is the velocity of the shock .
furthermore , the maximum energy electrons can achieve at a time t depends on whether synchrotron losses become dominant or not and it is given by the minimum of ( lagage & cessarsky 1983 , webb et al .
1984 , m96 and reynolds 1996 ) @xmath27 and @xmath28 where @xmath29 is the shock velocity in units of km / sec , b is the magnetic field strength in gauss , @xmath30 is the supernova remnant age in years , and @xmath31 with @xmath32 being the compression ratio of the shock connected with the spectral index @xmath33 through the relation @xmath34 .
also @xmath35 is the gyrofactor , i.e. the ratio of the particle s mean free path to its gyroradius .
the case @xmath36 corresponds to a simple bohm diffusion .
the relevant energy losses which were included in the kinetic equation were adiabatic due to the expansion of the remnant , synchrotron and inverse compton .
the only free parameter here is the magnetic field as the photon fields on which the electrons lose energy by inverse compton scattering are rather well determined ( see section 4 below ) . adopting standard supernova parameters , i.e. a total explosion energy of @xmath37 erg , an initial shock velocity of @xmath38 km / sec and an external matter density of @xmath39 h - atom/@xmath40 ( willingale et al .
1996 ) we solved numerically the electron kinetic equation which includes a source term and loss terms as these were described above ( for more details see m96 ) .
the obtained electron distribution function was folded with the synchrotron emissivity to obtain the radiated spectrum as this should be observed at a time @xmath41 years .
this way we have made numerical fits to the radio / x - ray data by treating as free parameters the strength of the magnetic field downstream , the gyrofactor @xmath35 and the shock compression ratio @xmath32 .
fig.1 shows the best fit to the radio and x - ray spectra ( shown here as full lines ) which was obtained for @xmath42 g and @xmath43 ( corresponding to @xmath44 ) .
this fit holds for @xmath45 .
as @xmath35 increases above this value the spectrum starts to break due to synchrotron losses and no satisfactory fit could be found .
therefore x - ray data alone might not be able to constrain @xmath35 significantly .
more constraints on @xmath35 can be placed from future tev observations as we will show in the next section .
we proceed now by giving a simple qualitative picture which will help in the understanding of our results : according to the standard picture of cosmic ray acceleration in supernova remnants ( see , for example , dorfi 1991 ) both @xmath46 and @xmath47 peak close to the transition of the supernova remnant from its free expansion to its sedov phase which occurs at @xmath48 . for example , @xmath49 for @xmath50 and @xmath51 for @xmath52 . similarly ,
@xmath53 for @xmath54 and @xmath55 for @xmath56 .
therefore the radiation observed now should be dominated by the particles accelerated at @xmath57 .
using eqn .
( 4 ) we get that the maximum particle energy at @xmath58 is given by @xmath59 @xmath60 is a factor ( of order unity ) that reflects the fact that at @xmath61 the velocity is not any more @xmath62 but somewhat smaller since deceleration of the shock should have started before that time .
this in turn implies that the energy @xmath47 as given by eqn .
( [ esedov ] ) is , strictly speaking , an upper limit of the maximum energy the electrons can attain during acceleration .
equation ( [ esedov ] ) is the maximum energy electrons can attain as long as this is smaller than @xmath63 ( equation [ [ elos ] ] ) . a comparison of the two expressions for @xmath61 yields that this is indeed the case as long as @xmath64 g. if this relation is satisfied then synchrotron losses do not become important at _ any _ epoch , since for @xmath52 , @xmath51 while @xmath65 .
> from the above discussion it becomes evident that if radiation losses do not become important then the synchrotron spectrum that the current model predicts can be fairly well represented by the synchrotron spectrum which is derived from a simple electron distribution function ( power law times exponential cutoff ) of the form @xmath66 } 4\pi d^2,\ ] ] with units in total number of electrons per erg while @xmath67 is in ergs . the best - fit value for the magnetic field is given by @xmath68 @xmath69 g while @xmath70 .
the area factor @xmath71 allows us to calculate photon spectra in terms of the flux at earth if @xmath72 is used .
using the best fit value for @xmath6 the total energy content in relativistic electrons is @xmath73 erg .
this in turn gives us an efficiency of cosmic ray electron production of @xmath74 in rough agreement with the hypothesis that about a few percent of the supernova energy available for acceleration goes to electrons .
several soft photon fields may contribute to the inverse compton ( ic ) scattering of relativistic electrons into the @xmath5-ray region .
de jager ( 1996 ) discussed the production of @xmath5-ray spectral features introduced by the scattering of several thermal photon spectra by the electron cutoff ( with maximum electron energy @xmath75 ) in snr w44 , which may explain the egret detection ( 2eg j1857 + 0118 , thompson et al .
1996 ) of this remnant .
the cutoff in the spectrum derived in section 3 will similarly produce @xmath5-ray bumps on a @xmath76 plot ( as in fig .
1 ) at @xmath5-ray energies @xmath77 ( in the thomson limit ) or @xmath78 ( in the klein - nishina limit ) , with @xmath79k representing the cmbr , @xmath80k the galactic dust contribution , @xmath81k representing the population ii stars , and @xmath82k representing the contribution from population i stars , with densities given by skibo ( 1993 ) .
arendt ( 1989 ) also did not detect any significant fir emission associated with swept up dust in sn 1006 , so that there would be no contribution to soft photons from sources inside or near sn 1006 , in contrast with w44 where the emission from swept - up dust dominates all soft photon fields .
we include all soft photon fields , using the full cross section for isotropic ic as reviewed by blumenthal & gould ( 1970 ) in the calculation of the @xmath5-ray spectrum .
the ic spectra calculated from the best - fit electron spectrum ( eqn . [ [ best ] ] ) for different @xmath35-values are also shown in fig . 1 , and it is clear that the @xmath5-ray energy flux peaks at a few tev ( nearly independent of the value of @xmath35 ) , the dominant source of this bump being the cmbr photons scattered in the thomson limit .
the integral flux above 1 tev is given by @xmath83 with @xmath35 constrained to values below @xmath84 .
the crab spectrum in the range 0.5 to 10 tev ( djannati - ata 1995 ) is also shown for comparison in fig . 1 , and it is clear that we may expect sn 1006 to be a stronger source than the crab at 1 tev if @xmath85 .
it is also clear that the egret upper limit of sn 1006 shown in fig .
1 does not constrain @xmath35 significantly , but tev observations will provide valuable limits .
finally we would like to note that the tev flux as given above is about an order of magnitude higher than the expected flux from the nuclear component of cosmic rays assumed to be accelerated also at the supernova shock ( drury et al .
the recent x - ray observations of sn 1006 ( koyama et al . 1995 ) make a strong case for particle acceleration in supernova remnants .
a necessary consequence of this is that sn 1006 can be a source of tev radiation as the accelerated electrons will scatter off the ambient soft photons to very high energies . in the present paper we calculated the expected tev flux by following the basic principles of first order fermi shock acceleration to fit the radio / x - ray observations and consequently we used the derived electron distribution function to calculate the tev emissivity .
in this case we found that the tev flux depends sensitively on the gyrofactor @xmath35 of the accelerated electrons and consequently one can use any future tev observations of sn 1006 to put limits on this .
we find that the egret upper limit ( derived from thompson et al .
1996 ) does not constrain @xmath35 significantly , but imaging tev observations have no difficulties observing sources nearly 10 times weaker than the crab .
we may therefore be able to detect sn 1006 at tev energies if @xmath35 is not too large .
the ne and sw limbs may be detectable by the atmospheric imaging cherenkov technique ( see e.g. weekes et al .
1989 ) if a search for the ne and sw rims are made .
however , the flux for each rim would be @xmath86 of the flux given in eqn .
( [ ftev ] ) for the total tev emission from sn 1006 , but by superimposing the pixels from the tev image corresponding to the ne and sw rims , the total flux should be about 75% of the flux given in eqn .
( [ ftev ] ) .
am would like to thank the deutsche forschungsgemeinschaft for support under sonderforschungsbereich 328 and the staff and members of the space research unit of the potchefstroom university for their hospitality .
|
supernova 1006 is the first shell type supernova remnant to show evidence of particle acceleration to tev energies . in the present paper
we examine this possibility by modeling the observed x - ray non - thermal emission in terms of synchrotron radiation from fermi accelerated electrons .
the predicted synchrotron spectrum fits the radio and non - thermal component of the observed soft x - ray to hard x - ray emission quite well .
these particles can produce tev gamma rays by inverse compton scattering on the microwave radiation and other ambient fields , and the derived electron distribution is also used to calculate the expected inverse compton flux .
we find that if the remnant is characterised by a magnetic field strength lower than @xmath0 g , then the tev flux can be higher than that of the crab nebula .
about 75% of the tev emission from sn 1006 is expected to be concentrated in the synchrotron bright ne and sw rims ( the hard aegis " ) of the remnant , which would allow a sensitive search if the atmospheric imaging cherenkov technique is used .
|
although first - principle calculations in general , and dft in particular , have provided a practical ( i.e. numerical tractable ) path for solving the electronic structure problem , they have introduced new numerical challenges on their own . within the single electron picture ,
the hamiltonian operator depends on the occupied eigenfunctions and the resulting eigenvalue problem becomes fully non - linear ( i.e. @xmath0 ) . in practice , this _ non - linear eigenvector problem _ is commonly addressed using a self - consistent field method ( scf ) wherein a series of linear eigenvalue problems ( i.e. @xmath1 ) , needs to be solved iteratively until convergence @xcite . successfully reaching convergence by performing scf iterations is of paramount importance to first - principle quantum chemistry and solid - state physics simulations software .
a typical self - consistent iteration procedure for the discretized dft / kohn - sham problem is represented in figure [ fig1 ] . basic self - consistent procedure for solving the ( discretized ) non - linear eigenvector problem with @xmath2 real symmetric and @xmath3 symmetric positive definite ( using non - orthogonal basis functions ) , and obtaining the ground state electron density @xmath4 ( with a factor @xmath5 for spin ) . for the dft / kohn - sham problem , the hamiltonian @xmath6 = -\delta + v_h[n]+v_{xc}[n]+v_{ext}$ ] ,
is composed of the hartree potential @xmath7 ( solution of poisson equation ) , the exchange - correlation potential @xmath8 , and other external potential @xmath9 including the ionic potential . ]
traditional scf mixing methods employ successive approximation iterates of a fixed point mapping to generate the new input electron density at each cycle .
examples of such methods include newton - broyden , or other anderson technique , and pulay mixing techniques using direct inversion of the iterative subspace ( diis ) @xcite .
using diis , the new input electron density at iteration @xmath10 , is generated from a linear combination of a series of previous trial densities ( i.e. @xmath11 ) by minimizing the successive residual errors between input and output densities .
three main numerical difficulties arise from standard scf procedures : ( i ) the linear eigenvalue problem needs to be solved repeatedly a large number of times ; ( ii ) the robustness of the self - consistent iterations is very sensitive to the choice of the initial guess ; and ( iii ) there is , as of yet , no robust and efficient general - purpose iterative numerical scheme for addressing the non - linear coupling with guaranteed convergence .
although much progress has been made to improve the converge rate of scf techniques @xcite , the iterations can still be found to converge very slowly or unreliably @xcite . in this paper
, we present an efficient alternative to traditional scf iteration techniques that is ideally suited to address the aforementioned numerical difficulties when solving the discretized non - linear eigenvector problem .
the scf problem is fully revisited using a general - purpose numerical strategy derived from a modification of the feast eigenvalue algorithm @xcite .
we show that the new approach , named nlfeast ( for non - linear feast ) , takes naturally advantage of the intrisic subspace iterations procedure of the feast algorithm to achieve global convergence , while the non - linearity is only addressed at a level of a reduced system ( which can be solved even approximately using any scf procedures ) .
the outline of this paper is as follows : in section [ sec : feast ] we briefly summarize the numerical steps and the main properties of the feast algorithm presented in ref . for solving the linear eigenvalue problem . in section [ sec : nlfeast ]
we present the generalization of feast for solving the non - linear eigenvector problem . numerical results and capabilities of the new nlfeast approach
are then presented and discussed in section [ sec : result ] .
within the scf procedure in figure [ fig1 ] , solving the linear and symmetric eigenvalue problem at a given iteration step becomes the most time - consuming part of the electronic structure calculations .
in such computations , one can identify two main challenges : ( i ) discretization techniques that accommodate atomistic systems with a high level of accuracy can lead to large size system matrices * \{h , s } * , and ( ii ) the number of eigenpairs needed to compute the electron density is proportional to the number of atoms in the system . in order to characterize large - scale nanostructures and complex systems of current technological interest , many thousands of eigenpairs are indeed needed . in this regard , progress in large - scale electronic structure calculations can be tied together with advances in numerical algorithms for addressing the eigenvalue problem . the recent feast algorithm @xcite , which uniquely combines accuracy , robustness , high - performance and ( linear ) parallel scalability , is ideally suited for addressing the electronic structure problem .
feast is a general algorithm that can be used for solving the linear eigenvalue problem to obtain all the eigenvalues and eigenvectors within a given search interval ( e.g. @xmath12 $ ] ) .
feast s main computational tasks consist of solving a small number of independent linear systems with multiple right - hand sides along a complex contour and one reduced dense eigenvalue problem that is orders of magnitude smaller than the original one ( the size of this reduced problem is of the order of the number of eigenpairs inside the search interval ) .
a full and detailed accounting of the implementation of the feast algorithm is given in ref .
, and the basic procedure is briefly summarized in figure [ fig2 ] . basic feast procedure for solving the generalized eigenvalue problem of size @xmath13 with @xmath2 real symmetric and @xmath3 symmetric positive definite ( spd ) , and obtaining all the @xmath14 eigenpairs within a given interval @xmath12 $ ] .
the density matrix appears implicitly in step-2 , using the complex contour integration of the green s function . in practice ,
the vectors @xmath15 in step-2 can be computed using a high - order numerical integration such as gauss - legendre quadrature , where only a small number of linear systems , @xmath16 , need to be solved , one for each of a number of specific gauss nodes @xmath17 ( associated with the weights @xmath18 ) along a complex contour @xmath19 , i.e. @xmath20 worthy of particular note are steps 2 and 4(c ) which , combined , generate a subspace * q * that spans the eigenvectors solutions in the user - defined interval .
this is done by multiplying a given approximate eigenvector solution at iteration @xmath10 by the density matrix @xmath21 i.e. @xmath22 where the new subspace @xmath23 is then obtained at the feast iteration @xmath24 . denoting by the @xmath14 eigenvectors within the search interval , we note that the density matrix is formally given by @xmath25 . in practice ,
the spectral projection ( [ eq : rho ] ) is conveniently obtained by numerical integration of a set of solutions of independent linear systems defined along a complex contour .
it can be shown that feast consists of a reformulation of the subspace iteration technique for spectral projectors @xcite , which leads to an extremely robust and accurate numerical procedure . by this
means the linear feast algorithm achieves very high convergence rate of the eigenvectors subspace , and allows the non - linear extension of feast ( i.e. nlfeast ) to achieve the kind of performance that will be demonstrated in this paper .
the feast algorithm holds all the following important intrinsic properties : * using a high - order gauss - legendre quadrature , @xmath26 to @xmath27 contour points suffice for feast to consistently converge in @xmath283 iterations to obtain up to thousands of eigenpairs with machine accuracy ; * all multiplicities are naturally captured ; * no ( explicit ) orthogonalization procedure is required ; * pre - computed subspace can be reused as suitable initial guess ( i.e. step-1 in figure [ fig2 ] ) for solving a series of eigenvalue problems that are close one another .
one can ideally take advantage of this feature for addressing , in particular : ( i ) step-3 of the scf procedure in figure [ fig1 ] along the iterations , ( ii ) bandstructure calculations @xcite along the @xmath10 space , or ( iii ) time - dependent propagation using spectral decomposition along the time - domain @xcite ; * the inner linear system at each contour point can be solved using either direct or iterative methods ; * the algorithm can be extended efficiently for solving non - hermitian problems ( e.g. for performing complex bandstructure calculations @xcite ) ; * efficient parallel implementations can be naturally addressed at three different levels : ( i ) many search intervals can be run independently ( no overlap ) , ( ii ) each linear system can be solved independently along the complex contour ( e.g. simultaneously on different compute nodes ) , and ( iii ) the linear systems can be solved in parallel ( the multiple right sides can be parallelized as well ) . since the search intervals can be arbitrarily narrowed within a parallel environment ( i.e. increasing the ratio @xmath29 ) , the algorithm complexity is directly dependent on solving a single linear system of size @xmath13 with @xmath30 right - hand - sides ( i.e. @xmath16 in figure [ fig2 ] ) .
here , we propose to further extend the capabilities of the feast algorithm in order to solve nonlinear eigenvector problems of the form @xmath31}{\bf x}_i = e_i { \bf sx}_i,\quad { \bf n}=2\sum_i |{\bf x}_i|^2 , \label{eq : nl}\ ] ] so that it may be used as a direct and efficient alternative to scf methods conventionally based on a `` schrdinger - poisson '' iterative procedure with density mixing schemes .
recently yang et al .
@xcite have proposed an alternative approach to traditional scf iterations using a direct constrained minimization ( dcm ) algorithm for solving the nonlinear problem ( [ eq : nl ] ) .
dcm consists of constructing and updating new search directions for the eigenvector subspace by solving a much smaller optimization problem which takes the form of a reduced nonlinear eigenvector problem .
these results have motivated this current work , since a reduced eigenvalue problem can also be obtained from the feast algorithm , which also provides an optimal framework for performing subspace iterations .
indeed , once the search subspace @xmath32 is obtained by the contour integration in step-2 of figure [ fig2 ] ( from a given hamiltonian @xmath6 $ ] ) , the resulting reduced problem in step-3 can be expressed in a non - linear form i.e. @xmath33\phi}=\epsilon { \bf s_q\phi } , \quad \mbox { with } \quad { \bf n}=2\sum_i |{\bf q\phi}_i|^2 , \label{eq_nlr}\ ] ] with @xmath34=q^t h[n]q$ ] . using a dft / kohn - sham model , @xmath35 accounts for the projection of the non - linear hartree and exchange - correlation terms onto the working search subspace @xmath32 .
one of the primary difference between linear feast and the proposed non - linear feast algorithm , named nlfeast , then lies in step-3 with the formation and solution of the reduced non - linear eigenvector problem .
this latter , in turn , can be solved using the traditional scf procedure presented in figure [ fig1 ] .
since the non - linearity appears at the level of the reduced system alone , more robust non - linear schemes that would have been too expensive to use on the large size original system ( [ eq : nl ] ) , can potentially be considered for addressing the reduced system ( [ eq_nlr ] ) ( e.g. specific newton - raphson method @xcite ) .
as in the linear case , once the solution of the reduced system is obtained , the subspace is updated and the feast algorithm can come back on track and perform subspace iterations up until convergence . as it will be demonstrated in our experimental results in section [ sec : result ] , however , the algorithm essentially converges faster when the size of the subspace is increased with each feast subspace iteration .
these results suggest that the size of the search subspace should not be made static for the non - linear problem .
this then constitutes a major and innovative difference with the original feast algorithm that aims at improving the robustness of the subspace iteration technique .
the raleigh - ritz procedure ( i.e. step-3 ) then uses a new search subspace @xmath36 which consists not only of the subspace most recently produced by solution of the contour integration at step @xmath10 , i.e. @xmath37 , but also of all ( or some ) of the subspaces generated in previous iterations such that @xmath38 .
the subspace provided by the contour integration can be appended to @xmath39 indefinitely until convergence , or this can be done such that only a finite number ( typically @xmath40 or @xmath41 ) of the most recently generated subspaces are retained .
interestingly , this procedure is reminiscent of the one used in diis where new electron densities are constructed from a subspace consisting of previous generated electron densities and density residuals .
however , traditional mixing schemes act on the electron density alone , and it is not always guaranteed that the density can be expressed as a linear combination of previously generated densities .
in contrast , the proposed approach is expected to be much more robust by essence since a larger and refined search subspace containing all the eigenvectors of interests , is very likely to span the solution of the non - linear problem .
the resulting nlfeast algorithm is given in figure [ fig3 ] .
feast procedure for solving the non - linear eigenvector problem of size @xmath13 with @xmath2 real symmetric and @xmath3 symmetric positive definite ( spd ) , and obtaining all the @xmath14 lowest occupied states within a given interval @xmath12 $ ] and hence the ground state electron density . here ,
step 3-c of the algorithm is making use of a traditional scf procedure for solving the non - linear reduced system .
we note that in step-3a , the column vectors of the newly generated subspace are appended to the matrix of the old subspace , increasing the total subspace size by @xmath30 .
this can be done such that only a finite number of the most recently generated subspaces are retained , or the subspace size can be increased until convergence .
additionally , a singular value decomposition ( svd ) is performed on @xmath36 ( i.e. step 3-b ) , such that only the left singular vectors @xmath42 are used in the rayleigh - ritz procedure . ]
we note that a singular value decomposition ( svd ) also needs to be performed on the search subspace @xmath36 ( i.e. step 3-b ) , where only the left - singular vectors are computed and retained .
in contrast to the linear feast algorithm , one does not want to truncate the size of the search subspace for constructing a positive definite reduced matrix @xmath43 .
this latter is by definition rank deficient and the orthogonalization of the subspace can not then be avoided ( the svd step can also be replaced by a qr decomposition , although the singular values could potentially provide additional information about the convergence rate @xcite ) . in our experiments thus far this step has proven to be computationally inexpensive relative to other parts of the algorithm , and even so , it can be shown that the svd of @xmath44 can be updated at each successive feast iteration with complexity similar to solving a single reduced eigenvalue problem in step-3c @xcite .
finally , like other schemes for solving nonlinear eigenvector equations , nlfeast requires an initial guess for the density @xmath45 ( step-1a of figure [ fig3 ] ) .
however , the quality of the initial guess is not important as far as achieving convergence is concerned , as it will be shown in section [ sec : result ] . in summary , nlfeast can be seen as an inversion of the usual process of solving the nonlinear eigenvector problem ( [ eq : nl ] ) .
using scf - diis , one iteration corresponds to a traditional scf iteration as presented in figure [ fig1 ] which requires , in particular , solving a large linear eigenvalue problem as well as the calculation of the hartree potential by solving poisson s equation .
this scheme presents then two iterative procedures : one outer associated with the scf , and one inner due to the eigenvalue solver ( using a `` black - box '' eigenvalue solver , those inner iterations -intrinsic to eigenvalue algorithms- are most often hidden to the user ) . as discussed earlier
, this linear eigenvalue problem can ideally be solved using the linear feast algorithm ( in figure [ fig2 ] ) which converges in very few iterations of contour integration ( typically @xmath46 ) .
using nlfeast , however , those two iterative procedures happen now in reverse order .
the outer iterations are intrinsic to the feast algorithm ( i.e. subspace iterations ) , while the inner iterations are used for solving the non - linear reduced system ( [ eq_nlr ] ) .
in addition to offering a more robust mathematical approach , nlfeast is expected to considerably reduce the eigenvalue solve time since : ( i ) the reduced non - linear problem ( [ eq_nlr ] ) is orders of magnitude smaller in size as compared to the original one ( [ eq : nl ] ) , and ( ii ) the reduced problem does not need to be solved very accurately to guarantee the convergence of the outer - iterations .
in this section , we propose to demonstrate the numerical efficiency of the proposed nlfeast algorithm for the non - linear eigenvector problem ( [ eq : nl ] ) .
dft / kohn - sham / lda calculations have been performed on various molecules using our in - house all - electron simulation framework that uses a real - space cubic finite element discretization .
some details of our modeling and real - space discretization setup have been provided in ref . .
the pulay - diis technique has been used for performing the traditional scf approach defined in figure [ fig1 ] , as well as for solving the feast non - linear reduced system in step-3c of figure [ fig3 ] .
both diis procedures make use of a subspace composed of five successive generations of the electron densities , but the maximum number of diis iterations for solving the reduced system in nlfeast has been fixed to three ( i.e. the non - linear reduced system is then solved only approximately ) . for most of the examples , indeed , increasing the number of the inner iterations further has had no effect on the overall nlfeast convergence rate . in all our experiments conducted thus far , including various molecules from @xmath47 to @xmath48 ,
the nlfeast algorithm has outperformed scf - diis both in terms of convergence rate and execution time .
three representative examples are shown in figure [ fig4 ] for the silane , benzene and caffeine molecules .
the relative error on the total energy at each iteration is used as the measure of convergence for both approaches , as this is the most directly comparable measure between diis - scf and nlfeast . as discussed in section [ sec : nlfeast ] , the meaning of the outer - iterations is different for both algorithm since , for nlfeast , it directly represents the number of contour integrations and non - linear reduced systems . since diis - scf also takes advantage of the linear feast algorithm , it is possible to have a direct comparison between the main numerical operation counts of diis - scf and nlfeast .
table [ tab_compcost ] summarizes the number of feast contour integrations and poisson system solves needed by both approaches for obtaining the same convergence accuracy ( i.e. @xmath49 ) for the three molecules .
.[tab_compcost ] comparison of the number of contour integrations and poisson system solves required to reach convergence ( set at @xmath49 ) using nlfeast and scf - diis and the dft / kohn - sham / lda model for the three molecules @xmath50 , @xmath51 and @xmath52 .
the number of additional iterations needed to construct the initial subspace ( not reported in figure [ fig4 ] ) has also been taken into consideration ( i.e. one additional contour for nlfeast and five diis iterations for scf - diis ) . with enough parallelism for feast , the cost of each contour integral can be straightforwardly reduced to that of solving a single complex linear system , whereas the solution of poisson s equation consists of solving a single real - valued linear system . [
cols="<,^,^,^,^,^,^ " , ] the size of the search subspace @xmath36 ( in step-3a of figure [ fig3 ] ) may be limited by only retaining a finite number of the most recent subspaces , or it may be extended indefinitely until convergence . for the numerical experiments presented in figure [ fig4 ] and table [ tab_summary ] ,
the rate of convergence is expected to reach a maximum because all generated subspaces have been retained .
figure [ fig5 ] shows the results of several numerical experiments that were performed for our selected three molecules using different sizes for the search subspace @xmath39 . from these results ,
one observe that the rate of convergence of nlfeast is directly related to the number of subspaces that are retained , but a high rate of convergence can still be obtained by retaining only three to four of the most recent subspaces . by retaining a limited number of subspaces , it becomes also possible to increase the size of @xmath30 to consider a larger number of electrons for a given search interval without resorting yet to explicit parallelism . in the case of benzene and caffeine
, one also notes that the algorithm does not appear to converge at all when only a single @xmath53 subspace is used , this subspace being the one that was most recently generated .
although this is not necessarily typical and the solution may eventually converge using a more efficient approach for solving the non - linear reduced problem ( e.g. using a larger number of inner iterations for the feast - diis problem which has been fixed to three in our simulations ) , it highlights the importance of extending the search subspace size for the success of this algorithm . at each iteration of nlfeast , the approximate subspace solution is improved through multiplication by the density matrix of the most current hamiltonian ( see equation [ eq : rho ] ) .
this step , step 2 in figure [ fig3 ] , is accomplished by performing a numerical contour integration of the green s function multiplied by the approximate solution . in practice
, a @xmath54-point gauss quadrature can efficiently be used here , which involves summing the solutions of @xmath54 separate linear systems .
for all the results presented so far , @xmath27 gauss contour points were used to perform the quadrature .
figure [ fig6 ] presents a new set of convergence results for our three selected molecules , while considering the variation of the number of gauss points from as low as @xmath41 to as high as @xmath55 .
although increasing the number of gauss points can help to improve the convergence rate , the effect here is not quite as dramatic as the effect of increasing the subspace size as discussed previously .
we note here very similar convergence behaviors between nlfeast and the feast algorithm for the linear problem .
this latter admits a mathematical convergence proof that shows that the actual convergence rate depends on both the accuracy of the contour integration and the size of the search subspace @xcite . like other schemes for solving nonlinear eigenvector equations , nlfeast requires an initial guess for the density @xmath45 .
clearly , a good initial guess can provide faster convergence , but for nlfeast the quality of the initial guess is not important as far as achieving convergence is concerned .
unlike other means of performing scf iterations , our algorithm is capable of achieving convergence even when given an extremely poor initial guess including the extreme case of no initial guess at all ( i.e. @xmath56 ) .
figure [ fig7 ] shows the results of a number of numerical experiments wherein the initial guess for the electron density was set to zero .
each experiment eventually resulted in convergence .
the algorithm s performance when given such a poor initial guess is partly related to the size of the system ; more electrons typically means that a larger number of iterations is required before convergence is reached .
performance also appears to depend on the particular molecule under consideration . in figure
[ fig7 ] , @xmath57 requires a larger number of iterations to reach a high level of convergence , despite being the second smallest system ( in terms of number of electrons ) .
it is likely that , for some molecules , the algorithm converges towards a local energy minimum before ultimately finding its way to the global minimum , which results in its progress being more delayed than it would be with a molecule where such a detour does not occur .
the purpose of figure [ fig7 ] is just to illustrate the robustness of the algorithm by considering an extreme ( academic ) case .
as shown in table [ tab_summary ] , a conventional initial guess has , so far , guaranteed low residual convergence in less than @xmath58 feast iterations for all the molecules that we have experimented with .
a new eigensolver - based strategy , named nlfeast , is derived from a generalization of the feast eigenvalue algorithm for solving the non - linear eigenvector problem i.e. @xmath0 such as the one arising from the dft / kohn - sham electronic structure model . by providing a fundamental and practical numerical solution for addressing the non - linearity with the occupied eigenvectors ,
nlfeast offers a very efficient and robust alternative to the traditional self - consistent procedure using density - mixing schemes such as pulay - diis .
several numerical experiments using the dft / kohn - sham / lda model , have demonstrated the significant potential of nlfeast to outperform the traditional scf mixing techniques in terms of convergence rate , robustness , and numerical operation counts .
strictly speaking , nlfeast should not be considered as a direct competitor to current methods and it can rather be seen as an alternative formulation on how to handle the non - linearity . indeed , the resulting non - linear reduced system can , in turn , be addressed using any scf procedures . in our simulations ,
the reduced system did not need also to be solved accurately since the overall convergence of our scheme benefits from the robustness of the feast subspace iterations ( which also limits the number of inner scf iterations ) .
this feature could potentially be further exploited while considering hartree - fock or dft hybrid models where the construction of the hamiltonian system is more involved than solving a single poisson equation .
we note that a practical implementation of the technique can be achieved effectively using the feast solver package @xcite . as such
, the migration of electronic structure codes making use of scf mixing schemes , could proceed within two steps : ( i ) integration of the feast package as the main linear eigenvalue solver , ( ii ) reverse - order the scf - process as it was stated in section [ sec : nlfeast ] .
since the feast solver is reverse communication ready , it provides the flexibility to build a customized nlfeast with specific computational modules for the problem at hand ( independently of the physical and discretization models ) . in this paper ,
the applicability of nlfeast has also been demonstrated for a given search interval which can include several hundred states . in order to go beyond the current performance capabilities of nlfeast for addressing much larger molecular systems , including nanostructures of current technological interest , explicit parallelism
will become necessary .
the parallel treatment of nlfeast should directly benefit from the intrinsic parallel capability of the feast algorithm ( including the three level of parallelism mentioned in section [ sec : feast ] ) .
for the all - electron model , in particular , the algorithm can act on different energy ranges ( with no overlap for the contour integrations ) , in order to capture core or valence electrons independently .
for example , the results in figure [ fig8 ] clearly illustrate the various core and valence regions within the energy spectrum distribution for a given molecular configuration . with many search intervals present in the simulation ,
however , one would also need to address the parallel scalability for solving the non - linear reduced system . using the linear feast within a traditional scf process
, there is no issue , since the number of reduced systems is equal to the number of intervals . using nlfeast
, however , the situation becomes less straightforward since the construction of the non - linear reduced system depends ( in principle ) on the solutions of the contour integrations for all search intervals .
the development of new direct or iterative schemes for solving the resulting reduced eigenvalue problem that scale linearly with the number of search intervals will need to be addressed in future research .
the authors wish to acknowledge helpful discussions with prof .
joseph jerome ( the george washington university ) , prof . ahmed sameh ( purdue university ) , and dr .
ping tak peter tang ( intel corporation ) .
this material is supported by nsf under grant # eccs-0846457 and intel corporation .
25ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop _ _ ( , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop `` , '' ( ) , @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) , @noop * * , ( ) @noop * * , ( ) @noop `` , '' ( ) , @noop `` , '' ( ) , @noop `` , '' ( ) ,
|
the self - consistent procedure in electronic structure calculations is revisited using a highly efficient and robust algorithm for solving the non - linear eigenvector problem i.e. @xmath0 .
this new scheme is derived from a generalization of the feast eigenvalue algorithm to account for the non - linearity of the hamiltonian with the occupied eigenvectors . using a series of numerical examples and the dft - kohn / sham model
, it will be shown that our approach can outperform the traditional scf mixing - scheme techniques by providing a higher converge rate , convergence to the correct solution regardless of the choice of the initial guess , and a significant reduction of the eigenvalue solve time in simulations .
|
reciprocity is the symmetry of a physical system with respect to the exchange of a source and detector .
for example , in systems which demonstrate nonreciprocal behavior the transmission loss and/or phase delay depends on the direction of propagation @xcite . in general ,
reciprocal symmetry can be violated in multi - mode devices and the most well - known examples are ferrite - based circulators and isolators @xcite .
such devices are useful in several practical contexts : circulators for instance are commonly employed in both optical @xcite and cryogenic microwave systems @xcite to reduce reflections and , more importantly , to protect from amplifier backaction on the device - under - test .
likewise , in superconducting quantum information , circulators are often used to direct the amplified reflected signal from lumped - element superconducting parametric amplifiers .
since the most common schemes for producing nonreciprocity involve the use of high magnetic fields , they are often incompatible with the goal of integrating several measurement circuits into smaller and smaller physical volumes while also limiting the magnetic flux near sensitive superconducting circuits @xcite . in the superconducting quantum information and microwave engineering fields ,
this has motivated a push to understand how nonreciprocity can be generated using alternative methods @xcite and this general pursuit has been paralleled by similar efforts in optics @xcite .
although these different ideas are embodied in different physical implementations , they all share a common mathematical description based in coupled - mode theory . in this paper , we develop a scheme for determining the scattering matrix of an arbitrary coupled - mode system in terms of _ directed graphs_. in doing so , we show that nonreciprocity is a consequence of asymmetric interference between different connecting paths in a graph , as well as judiciously - placed dissipation in ancillary modes .
we emphasize that while a direct solution of the coupled - mode equations can always be found by means of standard mathematical techniques , the reverse problem of _ synthesizing _ a multi - mode system with desired properties ( gain , impedance match , directionality ) can be much more difficult . with the approach we describe in this work ,
the synthesis of a multi - mode device is translated into the process of building a directed graph whose edges are subject to specific conditions . since this approach is agnostic to physical implementation , it may benefit similar efforts based on optical , mechanical , and hybrid systems .
as many of these systems are parametrically coupled systems , we frame this work in the context of frequency conversion and amplification processes . this paper is organized as follows in section [ sections : reciprocity ] we introduce normalization and matrix conventions for extending the coupled - mode equations of motion to several modes .
this facilitates the definition of reciprocity in parametrically coupled systems with three or more modes . in section [ section : graphrepresentation ] we visualize the matrix representation of the equations of motion using directed graphs and show how the scattering matrix can be computed by use of subgraphs which connect the input / output modes .
we then use this picture to discuss the mechanism for nonreciprocity in general parametrically - coupled systems . in sections [ section :
threemodeamps ] and [ section : directionality ] we show how directed graphs can be used as both a synthesis and analysis tool . as examples ,
we synthesize a novel 3-mode parametric amplifier in which gain , directionality , and impedance match are present ( section [ section : threemodeamps ] ) .
finally , we analyze the isolation properties of a complicated system , the dc - squid , by casting it as a 10-mode scattering problem ( section [ section : directionality ] ) based on earlier work by kamal _ et al .
_ @xcite .
we conclude the paper with a discussion of how this graphical approach might be effectively applied in future work .
resonant and parametrically coupled modes are typically described by the langevin equations of motion @xcite . since in this work
we are primarily focused on the _ topology _ of the coupling in a given system , we utilize a normalized matrix form of these equations and build a graph representation of multimode coupling dynamics .
this approach has an additional advantage in that it allows one to deal with systems with several modes in different physical oscillator incarnations as one may encounter in hybrid electro- and opto - mechanical systems . while the transition from the time - domain equations of motion to their matrix representation is given in _
appendix a _ , in this section we present only the matrix variables that are necessary for building the graphs .
we begin by considering a set of @xmath0 coupled resonators , described by a set of @xmath1 internal mode amplitudes . in general
, we allow the number of modes to exceed the number of resonators @xmath2 such that several modes may reside within a single resonator .
this system can be a set of parametrically or resonantly coupled oscillators , but we are not going to assume a particular physical implementation for now , in order to derive results that are as general as possible . we will call @xmath3 and @xmath4 the natural oscillation frequency and total dissipation rate for the resonator in which mode @xmath5 resides
. the complex coupling rate between modes @xmath5 and @xmath6 will be denoted by @xmath7 . in order to describe both mode frequency conversion and amplification , we divide the modes into two subsets of @xmath8 internal mode amplitudes @xmath9 and @xmath10 internal mode amplitudes @xmath11 in the frequency basis , with each set of modes corresponding to the driven response at frequencies @xmath12 and @xmath13 .
we assume that @xmath14 in the case of frequency conversion between modes @xmath5 and @xmath6 ( for @xmath15 or @xmath16 ) and @xmath17 in the case of parametric amplification ( for @xmath18 and @xmath19 , or @xmath20 and @xmath21 ) .
for example , the coupling matrix for a 2-mode frequency converter or resonantly coupled oscillator system corresponds to the case @xmath22 , while a conventional parametric amplifier couples two modes , one at a positive frequency ( @xmath23 ) and the other at a negative frequency ( @xmath24 ) @xcite . with this general prescription for defining our mode basis
, we can perform an input / output analysis of any system of coupled resonators and calculate the corresponding scattering matrix connecting the vector of input fields @xmath25 to the vector of output fields @xmath26 ( see _ appendix a _ ) .
the scattering matrix can in general be expressed as @xmath27 where @xmath28 is a @xmath29 matrix given by @xmath30\ ] ] and @xmath31 where we have defined an overall normalization prefactor @xmath32{\prod_{\ell=1}^{n_m}{\gamma_\ell}}},\ ] ] and environmental coupling matrix , @xmath33 .
\label{eq : kmatrix}\ ] ] in general @xmath34 , with the equal sign if the internal dissipation rate of mode @xmath5 is zero .
the diagonal elements of @xmath28 are complex normalized detunings between the driven response frequencies and the natural resonator frequencies and include the mode dissipation rates , while the off - diagonal elements are normalized coupling coefficients . in principal ,
the normalizations above are unnecessary , but they allow one to cast the langevin equations of motion ( see _ appendix a _ ) in a simple matrix form ( as @xmath28 , the `` langevin matrix '' ) , emphasizing the underlying structure of the connections between modes .
we show below that this structure , as revealed in the graph representation of @xmath28 , allows one to draw immediate conclusions about the ( non)reciprocity in multimode coupled systems .
[ [ a - formal - definition - of - reciprocity . ] ] a formal definition of reciprocity .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in order to proceed , we must formally define reciprocity within the context of several modes , possibly oscillating at different frequencies ( as found in parametrically coupled systems ) .
the scattering matrix ( equation [ eq : scattering ] ) describes a fully reciprocal system ( transmission between _ any _ two modes is reciprocal ) if it obeys the general constraint @xcite , @xmath35 here the matrix @xmath36 corresponds to a set of phase shifts in the mode basis @xmath37,\ ] ] where we can assume without loss of generality that @xmath38 .
the matrix @xmath36 therefore corresponds to an arbitrary redefinition of the mode phases .
this is similar to the concept of a common mode phase shift , but generalized to several modes . a system that violates condition ( [ eq : reciprocity ] )
is nonreciprocal and is characterized by an asymmetric phase and/or amplitude transmission coefficient between at least one pair of modes .
we note that equation [ eq : reciprocity ] differs from the definition commonly found in textbooks@xcite , @xmath39 .
in fact , equation [ eq : reciprocity ] is valid for more general multimode parametric systems in which the mode frequencies can differ . as such
, it is a gauge - invariant definition of nonreciprocity @xcite . since the environmental coupling matrix @xmath40 in equation
[ eq : kmatrix ] is diagonal , it follows from equation [ eq : scattering ] that reciprocity is guaranteed if and only if the langevin matrix , @xmath28 , obeys the same similarity transformation as in equation [ eq : reciprocity ] , @xmath41 for some set of phase shifts @xmath42 . above
, we noted that all resonant and parametric systems considered here are constrained to couplings which are related by pairwise ( anti)-conjugation , _ i.e. _ , @xmath43 . placing this constraint in equation [ eq : reciprocityh ] we obtain @xmath44 where @xmath45 is the phase of the coupling coefficient @xmath46 between modes @xmath5 and @xmath6 and @xmath47 is an integer .
the system of equations [ eq : reciprocityh2 ] has at most a set of @xmath48 independent relations ( if all off - diagonal elements of @xmath28 are nonzero ) in @xmath49 unknown phases @xmath42 .
the system is therefore overdetermined and the reciprocity condition ( eqn .
[ eq : reciprocity ] ) is , for parametrically coupled systems , satisfied only for specific choices of coupling phases . in ( [ eq : m_matrix ] ) .
@xmath28 is the adjacency matrix of this graph , _
i.e. _ , every element @xmath50 in @xmath28 is the weight of an oriented edge from vertex @xmath6 to vertex @xmath5 in the graph .
( b ) calculation of the determinant @xmath51 and ( c ) the cofactor @xmath52 by use of permutation subdigraphs .
the cofactor @xmath52 is proportional to the scattering coefficient @xmath53 ( _ see text_).,title="fig:",width=16 ]
the conditions for equations [ eq : reciprocityh2 ] to have a unique solution , and therefore for the system to be reciprocal , can be visualized by drawing a _ directed graph _ representing the matrix @xmath28 ( see figure [ fig : determinant_calculation ] ) .
the graph has @xmath1 vertices and we associate a weighted directed branch ( `` edge '' ) between vertices @xmath5 and @xmath6 to every element @xmath46 in the matrix @xmath28 .
the diagonal elements @xmath54 and @xmath55 of @xmath28 are the weights of self - loops in the graph . in other words ,
all of the coupling terms in @xmath28 are shown as directed branches , and all mode detuning and dissipation is represented by self - loops in the graph . within this representation ,
the conditions for a unique solution for the entire system can be obtained by summing equations [ eq : reciprocityh2 ] along every closed loop @xmath56 , yielding a simple result , @xmath57 where @xmath58 is an integer .
equation [ eq : reciprocityh3 ] , reminiscent of kirchoff s first law in electrical circuits , is analogous to the reciprocity condition in jaynes - cummings lattices described in @xcite .
we stress , however that equation [ eq : reciprocityh3 ] is applicable to any linear coupled - mode system bearing arbitrarily complex connections and is not confined to the nearest - neighbor , planar graph couplings shown here .
moreover , in the case of parametrically coupled devices in particular , the vertices of our graphs represent modes that may physically correspond to energy that is differentiated by both space and frequency . as such
, the graphs are a very general representation of a coupled mode system and can describe coupling of energy at several different frequencies as well as physical ports . in order for a system to be _ nonreciprocal
_ , the condition [ eq : reciprocityh3 ] must be violated .
this , however , only expresses a minimal requirement for phase nonreciprocity . we must further define a condition for which a system will also demonstrate amplitude nonreciprocity .
specifically , we may ask which conditions allow the amplitude isolation ratio @xmath59 to be different than 1 . from ( [ eq : scattering ] ) it follows that the scattering coefficient @xmath60 scales with the corresponding element of the inverse langevin matrix , @xmath61 .
the problem of calculating the scattering coefficients is therefore reduced to the problem of inverting the langevin matrix which , as we show below , can also be conveniently cast in terms of directed graphs .
the inverse langevin matrix can be expressed in terms of the transpose of its cofactor matrix @xcite @xmath62 the scattering coefficient is then given by @xmath63 where @xmath64 is the kronecker delta function .
the elements of the cofactor matrix @xmath65 are equal to the determinant of the matrix obtained from @xmath28 after removing the @xmath6-th row and the @xmath5-th column ( _ i.e. _ , the @xmath66 minor ) , multiplied by @xmath67 @xcite .
the determinant of @xmath28 and the cofactor @xmath65 can be easily calculated from the graph in figure [ fig : determinant_calculation](a ) by the following procedure ( _ see _ refs @xcite ) . given a graph , we call a subgraph @xmath68 a `` permutation subdigraph '' if every vertex in @xmath68 has one and only one outgoing edge and one and only one incoming edge @xcite . the total weight @xmath69 of a subgraph @xmath68 is the product of the weights of all of its edges times @xmath70 , where @xmath71 is the number of cycles in @xmath68 ( by cycle we mean a sequence of distinct connected vertices , where the first and last vertex coincide )
. the determinant of the matrix @xmath28 is the sum of weights of every possible permutation subdigraph of the graph associated to @xmath28 @xcite , as shown in figure [ fig : determinant_calculation](b ) .
the cofactor matrix @xmath65 is the coefficient of the @xmath72 element of the matrix @xmath28 in the laplace expansion of the determinant @xmath51 .
therefore , in order to calculate @xmath65 , we need to compute only the permutation subdigraphs in the expansion of @xmath51 that contain the @xmath66 edge , with the @xmath66 edge removed . in other words , for every path @xmath8 from vertex @xmath6 to vertex @xmath5 , we consider the collection of all the permutation subdigraphs @xmath73 of the graph of @xmath28 containing @xmath8 , with the @xmath66 edge removed as in figure [ fig : determinant_calculation](c ) .
the cofactor @xmath65 is then given by @xmath74 where @xmath8 varies over all the paths connecting vertex @xmath5 to vertex @xmath6 and @xmath75 varies over all the subdigraphs containing the path @xmath8 .
the minus sign is due to the fact that , by removing the @xmath66 edge , we reduced the number of cycles in the subdigraph by 1 .
an example of this procedure is outlined in figure [ fig : determinant_calculation](b ) .
the graphs @xmath76 in the expression for the cofactor elements have a simple physical interpretation : they represent the possible scattering mechanisms that connect modes @xmath5 and @xmath6 .
since , from ( [ eq : scattering2 ] ) , the ratio between the scattering coefficients @xmath77 is equal to the ratio between the corresponding cofactors @xmath78 , the problem of analyzing isolation in a given multi - mode system can be reduced to evaluating the differences between the specific graphs that connect @xmath5 to @xmath6 and _ vice versa_. ( b ) and @xmath79 ( c ) .
the element @xmath80 of the cofactor matrix is calculated by summing the weights of the permutation subdigraphs containing a path from vertex @xmath81 to vertex @xmath82 , with the edge from vertex @xmath82 to vertex @xmath81 removed.,title="fig:",width=16 ] as a more concrete example , we apply this procedure to a simple three - mode system in which all modes are coupled via frequency conversion ( @xmath83 ) and are therefore described by the following matrix ( see figure [ fig : determinant_3mode ] ) , @xmath84,\ ] ] where @xmath54 and @xmath46 are normalized detunings and couplings as before .
the isolation between ports @xmath81 and @xmath82 can be calculated by use of the graphs representing the cofactors ( see figure [ fig : determinant_3mode](b ) and ( c ) ) yielding the expression , @xmath85 where , in the last line , we have removed all normalizations . from ( [ eq : iso3mode ] ) we observe that if the total dissipation of mode `` 3 '' , @xmath86 , then @xmath87 and there is no isolation between modes 1 and 2 . in this case mode 3
serves as an internal lossless mode and the system behaves as a simple reciprocal two - mode system .
here , dissipation plays a crucial role in violating reciprocity , as it breaks the symmetry between scattering elements which are otherwise related by complex conjugation .
assuming finite dissipation in mode 3 , we find the condition for maximum isolation @xmath88 , @xmath89 in other words , to achieve maximum isolation , the sum of the phases of the couplings along the loop connecting the three modes , @xmath90 , must be equal to the detuning angle , @xmath91 $ ] . at zero detuning in mode 3 , @xmath92 , this loop phase must be @xmath93 for perfect isolation .
equivalently , one can interpret this isolation condition , eqn .
[ eq : maxiso3 ] , as interference between the two possible coupling paths connecting modes 1 to 2 in the graph ( figure [ fig : determinant_3mode ] ) .
although we focus on the isolation here , condition [ eq : maxiso3 ] also yields unit transmission in the forward direction , @xmath94 , for @xmath95 .
in fact , one can continue to build a 3-mode circulator by finding similar constraints for the 2@xmath963 and 1@xmath963 coupling .
and its graph representation . the squid is equivalent to a nonlinear inductor modulated by the magnetic flux through the squid loop .
one pump at frequency @xmath97 is used to obtain parametric amplification between mode 1 at frequency @xmath98 and mode 2 at frequency @xmath99 .
the second pump at frequency @xmath100 generates frequency conversion between mode 1 and mode 3 at frequency @xmath101 .
a third pump at frequency @xmath102 closes the loop and generates parametric amplification between modes 2 and 3 . in this configuration non - reciprocal trans - gain
can be obtained between modes 1 and 2 .
( c ) and ( d ) subgraphs used to calculate the forward and reverse gain between modes 1 and 2.,title="fig:",width=16 ] traditional single - stage lumped - element parametric amplifiers , such as the one shown in figure [ fig:3mode_amp](a ) , operate in reflection - mode such that a circulator is needed in order to achieve unidirectionality . in
@xcite a single - stage amplifier with two parametric pumps at the same frequency was introduced , but in addition to forward gain it also had unity reverse gain .
the graph methodology introduced here allows us to revisit the problem of gain and isolation from another perspective .
in particular , our discussion above identified the importance of introducing a multi - edge loop to obtain nonreciprocity . here
we analyze a closed - loop amplifier which we term the `` delta '' amplifier , shown schematically in figure [ fig:3mode_amp](b ) both as a device concept and as its corresponding graph .
this amplifier can provide forward gain and zero reverse gain as discussed below .
as drawn , the delta amplifier requires three independent pumps to generate three pairs of coupling edges @xmath103 .
two of these pairs create amplification ( _ i.e. , _
@xmath104 , @xmath105 ) , while the third pair couples through frequency conversion ( @xmath106 ) .
one concept for how this might be implemented is shown schematically in figure [ fig:3mode_amp](b ) as three resonators that share a common current anti - node in which a flux - driven squid is operated as a parametrically modulated inductor . in principle
, one might drive two of these couplings using the same pump frequency@xcite , but for clarity we discuss only the case in which each of the three couplings is driven independently , which has the advantage of increased flexibility due to the increased number of degrees of freedom .
regardless of particular implementation , we can recognize the presence of a closed loop connecting all three modes , as well as the integration of amplification as a coupling process . from this bare description
we recognize that with three independent pumps we are constrained to optimizing only three of the nine possible scattering elements . in laboratory use
, we can choose to optimize for amplification in the `` forward '' direction ( @xmath107 ) , isolation in the `` reverse '' direction ( @xmath108 ) , and low input reflection coefficient ( input match , _
i.e. _ , @xmath109 ) to prevent unwanted reflections of signals propagating away from a device - under - test .
we begin by writing down the expressions for the forward and reverse transmission from equation [ eq : scattering2 ] , @xmath110{\gamma_1\gamma_2\gamma_3}}\frac{\beta^*_{12}\delta_3+\beta_{31}\beta_{23}}{|m| } \\
s_{12 } & = i\frac{\sqrt{\gamma^{ext}_1\gamma^{ext}_2}}{\sqrt[3]{\gamma_1\gamma_2\gamma_3}}\frac{-\beta_{12}\delta_3 - \beta^*_{23}\beta_{31}^*}{|m|},\end{aligned}\ ] ] where @xmath51 is the determinant of the coupling matrix represented in the graph shown in figure [ fig:3mode_amp](b ) .
assuming ideal isolation , @xmath111 , we find an expression similar to equation [ eq : maxiso3 ] , @xmath112 which provides an overall constraint on the coupling amplitudes and phases as a function of the complex detuning of mode 3 . under this primary constraint
, we then set the input reflection coefficient to zero , @xmath113{\gamma_1\gamma_2\gamma_3}}\frac{|\beta_{23}|^2-\delta^*_2 \delta_3}{|m|}-1=0,\ ] ] giving a constraint on the determinant , @xmath114 if external coupling dominates , @xmath115 , in the limit of small detunings @xmath116 the gain reduces to @xmath117 note that this form gives the gain at a fixed frequency where the input match is perfect ( equation [ deltas11 ] ) . by substituting equations
( [ delta_gain ] ) and ( [ isodelta ] ) in equation ( [ deltas11 ] ) at resonance we can determine the three coupling rates .
in the limit of high gain , @xmath118 , with the exact value depending on the gain @xmath68 . we have calculated the resulting @xmath119-parameters for the delta amplifier in figure [ fig : delta_sym ] under various constraints on the pump detunings for the case of uniform dissipation rates @xmath120 . in practice
, one might fix the pump frequencies , locking the three detunings together , and measure the @xmath119-parameters shown in figure [ fig : delta_sym](d ) . for this example calculation
, we selected values for the pump amplitudes to fix the nominal power gain to @xmath12120 db while obtaining significant isolation . the device bandwidth is determined according to equation ( [ delta_gain ] ) that fixes the gain - bandwidth product for specific dissipation rates .
we note here that the device bandwidth is comparable to traditional two - mode superconducting lumped parametric amplifiers , while the center frequency could be tuned by changing the frequencies of the pumps . finally , we find that the calculated output @xmath122 shows reflection gain at the output port , and is consistent with the limitations of having only three parameters ( the independent pumps ) to optimize three of the four scattering parameters between ports 1 and 2
. however , this might not be a limitation in practical use where one is more concerned with the backaction of the amplifier on a quantum circuit on the input side ( port 1 ) .
[ [ added - noise - of - the - three - mode - amplifier . ] ] added noise of the three - mode amplifier .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we computed the equivalent input added noise @xmath123 of the delta amplifier to check that the device can approach the standard quantum limit .
we assumed that half a photon of noise is injected into every port and calculated the total output noise spectral density from the scattering parameters .
the contribution from ports @xmath82 and @xmath124 , divided by the device power gain @xmath68 , constitutes the equivalent added input noise , as outlined in @xcite . for a symmetric device , in the limit of large gain and at zero detuning ,
the result is : @xmath125 where we can neglect @xmath126 as a consequence of the isolation condition .
the results of the numerical calculations for nonzero detuning are shown in figure [ fig : delta_sym ] .
the fact that the delta amplifier is quantum - limited may be surprising at first , as one might expect the third mode to contribute extra noise beyond the quantum limit .
however the third mode acts only as an effective dissipative coupling between modes @xmath81 and @xmath82 @xcite .
we note that the parametric amplification scheme that is the subject of ref .
@xcite constitutes a three mode device in which the added noise was determined to be at the standard quantum limit .
in fact , the delta amplifier is similar to that device except for the addition of an amplification branch which serves to create a closed loop in its graph description , introducing the necessary interference for directionality .
are coupled through internally generated pump signals at multiples of the josephson frequency @xmath127 . at each frequency
there are two doubly degenerate modes , corresponding to the common and differential excitation of the junction phases . ,
title="fig:",width=12 ] while the previous discussion focused on using a graphical approach to synthesize a novel amplifier , it can also be used to identify aspects of other systems which enable both gain and directionality . in this section
we use graphs to analyze a complicated multi - mode system , the dc - squid amplifier .
dc - squids are a type of superconducting amplifier with noise temperatures approaching the standard quantum limit @xcite .
since dc - squids also have inherently low power dissipation ( in addition to high gain and directionality @xcite ) , they have been the subject of several applied superconductivity efforts to produce reliable amplifiers for microwave quantum information measurements . in ref .
@xcite , kamal _ et al .
_ showed that the directionality of the squid is a consequence of multiple parametric amplification and frequency conversion processes due to frequency mixing between the input microwave signal and internally - generated parametric pumps at multiples of the josephson frequency .
this approach casts the dc - squid as a multi - mode coupling problem and , as such , is also amenable to a graph description . here
we describe the parametric coupling in the squid following the description in ref .
@xcite by use of graphs to map the connections between modes .
in particular , we will direct our attention to the interference and dissipation that must be present to generate nonreciprocity in the presence of gain .
a dc - squid consists of a pair of josephson junctions inside a superconducting loop as in figure [ fig : squid ] .
each junction is characterized by a phase difference @xmath128 across its terminals , but it is more convenient to recast the dynamics of the dc - squid in terms of the common and differential phase difference @xmath129 @xcite .
@xmath130 can be expressed as the sum of fast oscillating terms at multiples of the josephson frequency @xmath131 , where @xmath132 is the magnetic flux quantum , and slow oscillating terms at the signal frequency @xmath133 .
the fast oscillating terms act as parametric pumps causing mode conversion between @xmath133 and the mixing product frequencies @xmath134 .
the lowest - dimensional model that can capture nonreciprocity involves two parametric pumps and a total of 10 modes at 5 frequencies @xmath135 .
each mode is doubly - degenerate , because for each frequency there are a common and a differential mode excitation . some scattering subgraphs to describe the squid
are shown in figure [ fig : squid_graph_approximate ] , where , for example , @xmath136 indicates the coupling rate between the common modes at frequency @xmath133 and @xmath137 .
the input and output modes are shown at the very bottom and top respectively .
calculation of the mode coupling rates as a function of the internally generated pump amplitudes is detailed in _
appendix b_. the number of subgraphs to describe the squid is too large for analytical calculations
. however if the squid bias current is high enough , we can look for an approximate expression for the squid isolation for small values of the bias parameter @xmath138 , where @xmath139 is the critical current of the junctions and @xmath140 is the bias current .
the pump amplitudes and the coupling coefficients can then be calculated for small @xmath141 and we can simplify our calculations by considering only the subgraphs of leading order in @xmath141 . these subgraphs are shown in figure [ fig : squid_graph_approximate ] .
the order of the graph is given by the product of its edges and therefore depends on the length of the path connecting the input to the output modes as well as the order of the edge weights ( coupling rates ) .
the order of the coupling rates increases with the difference between the mode drive frequencies ( for example @xmath142 ) .
moreover a small ( @xmath143 ) direct coupling exists between modes at the same frequency , as explained in _
appendix b_. the dissipation rates for the common and differential modes when the squid is terminated into ideal infinite transmission lines , are given by @xmath144 with @xmath145 , @xmath146 , @xmath147 , @xmath56 is the inductance of the squid loop , and @xmath148 and @xmath149 are the junction shunt resistance and capacitance .
since the squid is a nonresonant system , the mode frequencies @xmath150 in equation ( [ m_elements1 ] ) are zero .
we can approximately compute the squid isolation by use of the lowest - order subgraphs shown in figure [ fig : squid_graph_approximate ] .
the first conclusion we can draw , based on our previous discussion , is that non - zero dissipation rates in ( [ gammas ] ) are essential to obtain nonreciprocity in the squid . a change in the impedance terminations at the various mode frequencies @xmath134 will change the amount of nonreciprocity , _ viz .
_ , modes that are terminated in a short or an open circuit ( corresponding to zero dissipation rates ) will not contribute to the squid isolation .
moreover , it is important that @xmath151 .
in fact , by taking into account the symmetries between the mode couplings , the weight of the @xmath152 order subgraphs in figure [ fig : squid_graph_approximate ] is given by @xmath153 + 4(\gamma_c-\gamma_d)\operatorname{im}\left[\beta^{cc}_{\omega+\omega_j,\omega}\beta^{cd}_{\omega,\omega+\omega_j}\right ] , \ ] ] apart for a common factor equal to the product of the internal modes normalized detunings .
if @xmath154 , the second term in equation [ lowest_order ] is zero and the squid becomes reciprocal at first order .
the same cancellation happens in the @xmath155- order subgraphs in figure [ fig : squid_graph_approximate ] and in the @xmath156-order subgraphs , with the exception the subgraphs in the bottom - left corner of figure [ fig : squid_graph_approximate ] . in a practical device isolation
can however be restored for @xmath157 if the resonant coupling rate @xmath158 is increased . in a real device
we expect such a rate to be higher , due to the stray capacitive coupling between the input and output as well as other parasitics @xcite . finally , in figure [ fig : squid_graph_approximate2 ] we show the computed isolation and power gain of the squid as a function of the bias points for the exact numerical solution and for the lowest - order approximations .
we find that graphs up to @xmath156 order in @xmath141 are enough to accurately describe the isolation properties of the squid .
the exact power gain was also computed numerically by extracting the impedance matrix from the scattering matrix of the squid and then calculating the gain as discussed in @xcite . in principle
the power gain may also be computed from the subgraphs , but the number of necessary graphs is large , even at lowest order . in this system , graphs help identify the elements of a complex multimode parametric scattering problem that contribute to the isolation properties . ) based on the linearized coupled - mode equations and the approximate solutions including subgraphs up to the @xmath152 , @xmath155 , and @xmath156 order for @xmath159 , @xmath160 and @xmath161 .
( b ) exact power gain.,title="fig:",width=16 ]
in this work we introduced a graph theoretical representation that can be used as a combinatorial accounting tool to analyze nonreciprocity in coupled - mode systems and used it to derive a minimal unilateral parametric amplifier and analyze the condictions for nonreciprocity in squid amplifiers . an abstract graph associated to the coupled - mode system
was used to compute the input / output scattering coefficients between modes and derive general constraints that need to be satisfied in order for nonreciprocity to occur . in order for a multi - mode system to be reciprocal
the sum of the phases along any loop has to be @xmath162 or @xmath163 . if this condition is violated , phase and/or amplitude nonreciprocity is present and this condition can be interpreted as interference between different permutations of connecting paths .
we also find that dissipation in the remaining disjoint vertices / modes is _ crucial _ for amplitude nonreciprocity / isolation .
specifically , it breaks the symmetry between forward and backward multi - mode scattering processes such that one may obtain constructive loop interference in one direction , and destructive interference in the reverse direction . as a result
we were able to design a novel three - mode parametric amplifier , characterized by forward gain , input match and reverse isolation that can be integrated on a superconducting chip .
although the scattering matrix of a multi - mode system can always be computed numerically using more conventional approaches , graphs offer a new perspective that can be useful in the problem of synthesizing new multi - mode systems and devices .
for instance , this can be leveraged as a design tool , allowing one to reduce a set of desired device characteristics to a minimal description that can be utilized to design new amplifier / converter concepts such as the delta amplifier . as an approach ,
it is conceptually similar to the analysis of electric networks by means of circuit representations of the network components , where graphical manipulation rules are used to simplify the circuit analysis .
we also showed , with the complex example of the dc - squid , that graphs aid in identifying the critical elements that create isolation in the presence of forward gain .
finally , graphs may provide a useful approach to engineering nonreciprocity in recent multi - mode parametrically coupled hybrid systems such as the electrical - mechanical - optical ( three - mode ) transduction bridge @xcite and electro - mechanical amplifiers @xcite .
in general , problems like these are particularly well - suited to a graph - based approach since many of the interesting properties one is interested in ( gain , isolation , gain - bandwidth product , _ etc .
_ ) actually arise from the structure of the coupling network ( the topology of the graph ) and not the physical particulars of a given implementation .
the authors would like to thank prof.m.devoret ( yale ) and dr.archana kamal ( mit ) for motivating discussions .
in this section we formally derive equation [ eq : scattering ] .
we begin by considering a set of @xmath0 coupled resonators described by : @xmath164 where @xmath165 is a vector of @xmath0 internal complex normal mode amplitudes and @xmath166 describes the set of input signal drives for each of these modes .
we assume that @xmath165 and @xmath166 can be expressed as a sum of periodic functions . in this case harmonic balance
can be used to rewrite equation [ system_eq ] as a system of @xmath1 coupled - mode equations : @xmath167 where @xmath168 $ ] is a vector of @xmath169 input stimuli and @xmath170 $ ] is a vector of @xmath1 internal modes that can be expressed as : @xmath171 and the corresponding conjugate expressions for @xmath21 . here , we allow for the number of modes to exceed the number of resonators , @xmath172 and label the drive / response frequencies with an @xmath173 , while the mode itself is labeled with a subscript @xmath5 so that `` @xmath174 '' denotes the drive / response frequency of mode @xmath5 which can be based within any of the @xmath0 resonators depending on how the coupling is directed within the function @xmath175 .
the diagonal matrix @xmath176 describes coupling to the environment through the external dissipation rates @xmath177 .
the diagonal elements @xmath178 of the @xmath179 coupling matrix @xmath180 can then be expressed in this mode basis as @xcite : @xmath181 and the off - diagonal elements @xmath182 are given by @xmath183 in equations [ helements ] and [ helements2 ] @xmath3 and @xmath4 are the natural frequency and total dissipation rate for the resonator in which mode @xmath5 resides , while the `` * '' indicates complex conjugation , and @xmath7 is the complex coupling coefficient between modes @xmath5 and @xmath6 .
we further assume that @xmath14 if @xmath15 or @xmath16 and @xmath17 for @xmath18 and @xmath19 , or for @xmath20 and @xmath21 .
for example , the coupling matrix for a 2-mode frequency converter or resonantly coupled oscillator system corresponds to the case @xmath22 , while a conventional parametric amplifier couples two modes , one at a positive frequency ( @xmath23 ) and the other at a negative frequency ( @xmath24 ) @xcite . with this general prescription for defining our mode basis
, we can perform an input / output analysis of the system ( equation [ eq : coupled_modes ] ) and calculate the vector of output ( scattered ) fields @xmath26 by means of the relation @xcite : @xmath184 in general , the phase factor @xmath185 depends on the nature of the coupling to mode @xmath5 . for simplicity , we set @xmath186 , as one would obtain with a small series capacitance or weakly reflecting mirror
. other choices of coupling will have the effect of modifying the identity matrix on the r.h.s . of equation [ eq : scattering ] .
by substituting equation [ eq : modes ] and [ eq : modes2 ] into equation [ eq : coupled_modes ] and by use of the input / output boundary conditions ( equation [ inputoutput ] ) , we can obtain the expression for the scattering matrix in equation [ eq : scattering ] .
a dc - squid consists of a pair of josephson junctions inside a superconducting loop as in figure [ fig : squid ] .
each junction is characterized by a phase difference @xmath128 across its terminals . defining the common and differential phase difference @xmath129 , the dynamical equations describing the system are @xcite : @xmath187 where @xmath188 , and @xmath189 .
@xmath139 is the junction critical current and @xmath140 is the bias current of the squid , @xmath132 is the magnetic flux quantum , @xmath56 is the loop geometric inductance and @xmath148 is the equivalent shunt resistance across the squid .
the system ( [ squid_equations ] ) can be linearized around the following bias point : @xmath190 where @xmath191 are small perturbations . in
@xcite it was shown that by applying a harmonic balance procedure , the phases @xmath192 can be expressed as @xmath193 where the components @xmath194 are internally generated pumps at frequency @xmath195 , where @xmath196 is the static dc voltage across the squid .
the pumps cause parametric frequency mixing between the various signal modes @xmath197 . in
@xcite the harmonic balance equations were solved for the case @xmath198 to show how nonreciprocity arises from the interference between the different parametric conversion paths . as shown in the main text
, we can employ the graph representation of a coupled - mode system to write explicitly the interference conditions needed to maximize the amount of amplitude nonreciprocity between two modes . in the @xmath198 case we can describe the squid as a 10-mode system .
the 10 modes have frequencies @xmath199 and are doubly degenerate ( for each frequency there are a common and a differential excitation ) . in order to calculate the scattering parameters , we need to find the coupling coefficients between the modes as a function of the internal pump amplitudes .
this can be obtained by linearising equations ( [ squid_equations ] ) for small signal mode - amplitudes and equating terms at the same frequency in order to obtain a set of coupled mode equations [ eq : coupled_modes ] . after taking symmetries into account ,
the system can be described by the following five coupling coefficients @xmath200 where for example @xmath201 corresponds to the coupling between the common modes at frequency @xmath133 and @xmath137 .
10 url # 1#1urlprefix[2][]#2 fan s , baets r , petrov a , yu z , joannopoulos j d , freude w , melloni a , popovi m , vanwolleghem m , jalas d _ et al . _ 2012 _ science _ * 335 * 3838
|
in this work we derive the general conditions for obtaining nonreciprocity in multi - mode parametrically - coupled systems .
the results can be applied to a broad variety of optical , microwave , and hybrid systems including recent electro- and opto - mechanical devices . in deriving these results
, we use a graph - based methodology to derive the scattering matrix .
this approach naturally expresses the terms in the scattering coefficients as separate graphs corresponding to distinct coupling paths between modes such that it is evident that nonreciprocity arises as a consequence of multi - path interference and dissipation in key ancillary modes .
these concepts facilitate the construction of new devices in which several other characteristics might also be simultaneously optimized . as an example
, we synthesize a novel three - mode unilateral amplifier design by use of graphs .
finally , we analyze the isolation generated in a common parametric multi - mode system , the dc - squid .
|
semiconductor double - quantum - well systems in a quantizing magnetic field develop spontaneous inter - layer coherence when the wells are brought into close proximity.@xcite spontaneous coherence leads to a variety of fascinating transport effects including counterflow superfluidity and anomalous interlayer tunneling , and to unusual charged excitations such as merons . a convenient way to describe these ground states is to use a pseudospin language in which the _ which layer _ degree - of - freedom is mapped to a @xmath8 pseudospin . in this language ,
the ground state of a bilayer at total filling factor @xmath9 is an easy - plane pseudospin ferromagnet . at higher filling factors ,
still more exotic states occur , for example states in which the pseudospin orientation varies in space and a charge - density - wave is formed.@xcite interest has recently been growing in the strong - magnetic - field ordered states of graphene bilayers .
single layer graphene@xcite is a two - dimensional honeycomb lattice network of carbon atoms .
bilayer graphene@xcite consists of two graphene layers separated by a fraction of a nanometer . in the normal bernal stacking structure ,
one of the two honeycomb sublattice sites in each layer has a near - neighbor in the other layer , and one does not .
this arrangement producesbilayerexpt , mccann a set of landau levels with energies @xmath10 where @xmath11 is the effective cyclotron frequency and @xmath12 .
all landau levels except @xmath2 are four - fold degenerate ; electronic states are specified by @xmath13 , valley - index ( @xmath3 or @xmath4 ) and spin - index , in addition to the usual label used to specify guiding center states within a landau level .
the @xmath2 landau level has an additional two - valued quantum degree - of - freedom because states with both @xmath0 and @xmath1 landau - level character have zero kinetic energy .
most of the new physics discussed in this paper is related to the property@xcite that electric dipoles can be constructed by forming wavefunctions with coherence between @xmath0 and @xmath1 components .
a second peculiarity of the @xmath2 state is that wavefunctions associated with the @xmath3 valley are localized in one layer , while wavefunctions associated with the @xmath4 valley are localized in the opposite layer .
layer and valley indices are thus equivalent .
it is convenient to use pseudospins to represent both layer ( or equivalently valley ) and the landau - level orbital character degrees of freedom .
the wavefunction for an electron in the @xmath2 landau level is therefore the direct product of a standard guiding center factor and three spinors that capture its dependence on spin , layer , and orbital - landau - index ( @xmath14 ) character .
we refer to the final spinor as the orbital spinor , and to the set of eight landau levels with zero kinetic energy as the bilayer graphene octet.@xcite in neutral graphene the octet is half - filled at all magnetic field strengths .
the presence of the octet in bilayers is revealed experimentally by a jump in the quantized hall conductivity @xcite from @xmath15 to @xmath16 when the charge density is tuned across neutrality in moderately disordered samples . in a recent paper@xcite
some of us predicted that quantum hall effects would appear at all integer filling factors between @xmath17 and @xmath18 in samples of quality sufficientnomuraprl to make interactions dominant relative to unintended disorder .
electron - electron interactions acting alone are expected to lift the degeneracy of the bilayer octet and induce gaps at the fermi level by producing a set of spontaneously broken symmetry states with spin , valley and orbital pseudospin polarizations .
the octet degeneracy lifting is expected@xcite to follow a set of hund s rules in which spin polarization is maximized first , then layer polarization to the greatest extent possible , and finally orbital polarization to the extent allowed by the first two rules .
hall plateaus at all integer filling factors intermediate between @xmath17 and @xmath18 have indeed now been discovered in experimental studies of suspended bilayer graphene samples and bilayer graphene on sio@xmath19/si substrates,@xcite opening up the opportunity to study a rich and still relatively unexplored@xcite family of novel broken symmetry states .
the odd filling factor cases are expected to be most interesting because all three pseudospins are expected to be polarized .
the present paper focuses on the physics associated with the competition between layer and orbital pseudospins at landau levels @xmath7 and @xmath20 at field strengths sufficient to produce maximal spin polarization and reduce the importance of landau - level mixing . in this limit
a negative filling factor @xmath6 is equivalent to a positive filling factor @xmath21 since the two states differ only through the presence in the latter case of inert filled majority spin landau levels . in a previous paper@xcite we studied the quantum hall states which occur at @xmath22 and @xmath23 in the same strong field regime , emphasizing the key role played by the potential energy difference between graphene layers which we refer to here as the bias potential @xmath24 .
the @xmath25 ground state at zero bias is an inter - layer coherent state with orbital index @xmath0 that supports counterflow superfluidity .
one particularly interesting property of this state is that the superfluid density , the coefficient that relates the counterflow supercurrent to the spatial gradient of interlayer phase , vanishes . correspondingly , the state s goldstone mode dispersion is quadratic in wavector @xmath26 , in contrast to the linear dispersion found in coherent semiconductor bilayers and in standard superfluids .
we also found that the uniform ground state has a long - wavelength instability at any non - zero potential - difference bias @xmath27 where @xmath28 is the critical bias at which all @xmath2 charge is tranferred to a single layer . in ref .
, we argued that the instability is probably towards a state in which the direction of the inter - layer pseudospin varies in space . for larger bias @xmath29
the ground state is uninteresting ; the charge is completely in one layer ( or valley ) and in the orbital state @xmath30 the orbital pseudospinwave mode corresponding to transitions between the @xmath0 and @xmath1 orbital states is gapped at a frequency @xmath31 where @xmath32 is the inter - layer tunneling energy in the bernal stacking .
this mode , _ which is an intra - landau level excitation _
, has a finite oscillator strength and will absorbbilayercr electromagnetic radiation .
this behavior contrasts with the standard kohn s theorem@xcite behavior in normal 2deg s which implies that only inter - landau level excitations produce absorption .
surprisingly the phase diagrams for @xmath7 and @xmath33 states differ qualitatively from the corresponding @xmath34 and @xmath23 phase diagrams .
the source of the difference is a competition in the @xmath35 case between interaction and single - particle effects which are reinforcing in the @xmath34 case .
the end result is that the large @xmath24 ground state at @xmath35 places electrons in a coherent combination of @xmath0 and @xmath1 orbital states , and that electric dipoles are consequently spontaneously present in the ground state .
this paper analyzes the dependence of bilayer properties on @xmath24 and explores some of the consequences of the unusual orbitally ordered dipole state . at small bias
we find that the bilayer s @xmath35 ground state is an inter - layer coherent state , much like the corresponding @xmath34 state except that the coherence is between orbitals with @xmath1 character .
this state has a gapless pseudospin wave mode with linear dispersion , like coherent semiconductor bilayers .
the state also has a gapped orbital pseudospin collective mode .
because the orbital spinor carries an electric dipole , this mode has a finite oscillator strength and absorbs electromagnetic radiation , again much like the @xmath34 case .
this mode should be visible in a microwave spectoscopy experiment .
inter - layer coherence decreases with bias until a new ground state is reached that has both inter - layer and orbital coherences . in this mixed state
, the low - energy orbital and inter - layer pseudospin modes are both gapped .
because the modes are coupled , both show up in the microwave absorption spectra .
the collective excitations are highly anisotropic in this phase and we find that the intensity of the absorption depends strongly on the orientation of the electric field of the incident microwaves . the new physics of the @xmath7 case emerges in its simplest form at still stronger bias potentials . both orbital levels in the bottom layer
are then completely filled while only one of the two top layer landau levels is filled .
spontaneous orbital coherence then develops in the top layer .
this spontaneous orbital coherence leads to a _
orbital pseudospin mode .
some of the properties of this state have been studied independently in a recent paper by shizuya@xcite , who also pointed out that orbital coherence is responsible for the existence of a finite density of electrical dipoles with a net polarization .
these dipoles collectively and spontaneously point in some arbitrary direction in the @xmath36 plane . as discussed by shizuya@xcite their orientation can however be controlled by an external electric field parallel to the plane of the bilayer . in this paper , we use an effective pseudospin model to highlight other interesting features of the orbitally - coherent state .
in particular we demonstrate the presence of a dzyaloshinskii - moriya ( dm ) interaction@xcite between orbital pseudospins and show that it leads to an anisotropic softening of the orbital pseudospin mode at a finite wavevector . for strong enough inter - layer bias
, the dm induces an instability toward a pseudospin spiral state .
the orbital pseudospin mode in the high bias regime is gapless and will lead , in the presence of disorder , to strong absorption of electromagnetic waves at very small frequencies .
our paper is organized in the following way . in section
ii , we discuss the non - interacting states of the graphene bilayer within a two - band low - energy model . here
we introduce the aspect of the electronic structure that is responsible for interaction and band effects which are competing at @xmath7 and are reinforcing at @xmath22 . in section iii
, we derive the hamiltonian of the graphene two - dimensional electron gas ( 2deg ) truncated to @xmath2 levels in the hartree - fock approximation .
we use this hamiltonian to derive the equation of motion for the single - particle green s function in section iv and to obtain the order parameters for the various phases which occur at @xmath37 .
section v describes the generalized random - phase approximation ( grpa ) ( or equivalently the time - dependent hartree - fock approximation ( tdhfa ) ) that we use to derive the collective excitations .
the phase diagram for @xmath7 as a function of bias is obtained in section vi .
we then study the collective excitations of the inter - layer coherent phase in section viiand those of the orbital coherent phase in section viii .
finally microwave absorption in the different phases is studied in section ix and we conclude with a brief summary and some suggestions for future work in section x.
in a graphene bilayer with bernal stacking , the two basis atoms of the top layer are denoted by @xmath38 and @xmath39 and those of the bottom layer by @xmath40 and @xmath41 with atoms @xmath38 sitting directly above atoms @xmath41 .
the band structure of the bilayer is calculated using a tight - binding model with in - plane nearest - neighbor tunneling ( with strength @xmath42 ev ) and @xmath43 tunneling ( with strength @xmath44 ev ) .
( for a review of bilayer graphene , see ref . .
) the low - energy ( @xmath45 ) excitations of this model for electrons in the valleys @xmath46 with @xmath47 and @xmath48 can be studied by using the effective two - band model developped in ref . .
using the basis @xmath49 for @xmath50 and @xmath51 for @xmath52 , the effective two - band hamiltonian derived in ref . is @xmath53 where @xmath54 and @xmath55 in this equation , @xmath56 is the bias potential between the two layers , the effective mass @xmath57 with @xmath58 the bare electronic mass , @xmath59 is the triangular lattice constant and @xmath60 is the distance between neighboring carbon atoms in the same plane .
the kets @xmath61 correspond to the atomic sites in different layers that are not directly above one another .
the hamiltonian of the 2deg in a perpendicular magnetic field is obtained by making the substitution @xmath62 ( with @xmath63 ) in eq .
( [ one ] ) .
the vector potential @xmath64 is defined such that @xmath65 in a magnetic field,@xmath66where we have defined the orbital ladder operators @xmath67 with the magnetic length @xmath68 and the parameter @xmath69(tesla ) .
the effective cyclotron frequency @xmath70 at zero bias , the landau levels have energies @xmath71 with @xmath12 in this paper , we study the phase diagram of the 2deg in the @xmath2 landau level . while levels with @xmath72 are four - fold degenerate ( counting spin and valley quantum numbers ) , level @xmath2 has an extra _
orbital _ degeneracy due to the fact that landau level orbitals @xmath0 and @xmath1 have zero kinetic energy .
the states in @xmath2 are thus member of an octet of landau levels that are degenerate if we neglect the zeeman and bias potential energies .
we assume that the zeeman coupling is strong enough to assure maximal spin - polarization , which allows this degree - of - freedom to be neglected . the eigenfunctions and corresponding energies for @xmath2
are then given by@xmath73for the @xmath74 valley and by @xmath75for the @xmath76 valley , using this time the basis @xmath77 for all states .
it is quite clear from these equations that the valley @xmath78 eigenstates are localized in the top(bottom ) layer . for @xmath2 , the layer index is thus equivalent to the valley index .
( for @xmath79 , the spinors have different orbital indices @xmath14 in different layers . ) the functions @xmath80 are the landau gauge ( @xmath81 ) eigenstates of an electron with guiding center @xmath82 , and @xmath83 is the wave function of a one - dimensionnal harmonic oscillator . note that with our choice of gauge , the action of the ladder operators on the states @xmath84 is given by @xmath85 and @xmath86 at finite bias , the parameter @xmath87 lifts the degeneracy between the two orbital states as we show in fig .
[ niveaux ] .
the splitting is however very small . for positive bias , the @xmath0 orbital state in the bottom(top )
layer is lower(higher ) in energy than the @xmath1 orbital state . the orbital states @xmath88 form a two - level system in each valley and we associate them with an _
orbital _ pseudospin .
similarly , the two states @xmath89 are associated with a _ valley _ pseudospin .
we remark that the effective two - band model slightly overestimates the gap @xmath90 between the @xmath0 and @xmath1 orbital states .
the inset in fig .
[ niveaux ] shows the difference between the gap calculated in the two - band model and in the original four - band system . in the region
where the dm interaction driven instability occurs , the difference between the two gaps is however very small .
the inset shows the behavior of the gap evaluated in the original ( solid line ) and two - band ( dashed line ) models with bias . ]
we now add the coulomb interaction to the non - interacting hamiltonian @xmath91 .
we assume that the magnetic field is strong enough so that we can restrict the hilbert space to the @xmath2 landau level and neglect landau level mixing .
we also assume the 2deg to be fully spin polarized ( we comment on this later ) .
we write the electron field operator as @xmath92and@xmath93so that the hartree - fock hamiltonian is given by ( here and in the rest of this paper , we use the convention that repeated indices are summed over ) @xmath94 where @xmath95 is the landau level degeneracy and all energies are measured in units of @xmath96 where @xmath97 is the effective dielectric constant at the position of the graphene layers .
the single - particle energies @xmath98 include capacitive contributions and are defined by @xmath99 , \label{ener}\]]with @xmath100 the number of filled levels in valley @xmath101 , @xmath102 the total number of filled levels , @xmath103 the valley ( or equivalently layer ) index and @xmath88 for the two orbital state indices . in deriving eq .
( [ hamil ] ) , we have taken into account a neutralizing positive background so that the @xmath104 contribution is absent in the hartree term . this convention is indicated by the bar over the summation .
note that for positive bias , the bottom layer ( @xmath105 valley ) is at a lower potential than the top layer ( @xmath74 valley ) .
the density operators in eq .
( [ hamil ] ) , are defined by @xmath106 where @xmath107 creates an electron in state @xmath108 in the landau gauge .
the intralayer @xmath109 and inter - layer @xmath110 hartree and fock interactions are given by @xmath111and@xmath112where @xmath113 is the inter - layer separation in the bernal stacking .
the form factors which appear here,@xmath114capture the character of the two different orbital states .
detailed expressions for the hartree and fock interactions parameters are given in appendix a.
the states with no pseudospin texture at integer filling factors have uniform electronic density and density - matrices that vanish for @xmath115 .
letting @xmath116 @xmath117 the hamiltonian of eq .
( [ hamil ] ) reduces to @xmath118these order parameters are conveniently calculated by defining the time - ordered matsubara green s function @xmath119 since , at time zero , we have @xmath120 in the hartree - fock approximation , the equation of motion for the single - particle green s function is@xmath121 g_{n_{1},n_{2}}^{a , b}\left ( \omega _ { n}\right ) \label{iterationp } \\ + u_{n_{1},n_{3}}^{a , c}g_{n_{3},n_{2}}^{c , b}\left ( \omega _ { n}\right ) = \hslash \delta _ { a , b}\delta _ { n_{1},n_{2 } } ,
\notag\end{gathered}\]]where @xmath122 is a fermionic matsubara frequency and @xmath123 the system of eqs .
( [ iterationp ] ) can be solved in an iterative way by using some initial values for the parameters @xmath124 in ref .
, we solved this equation keeping valley , orbital , and spin indices .
we showed that the solutions of the hartree - fock equations for the balanced bilayer ( @xmath125 ) follow a hund s rules behavior .
the spin polarization is maximized first , then the layer polarization is maximized to the greatest extent possible , and finally the orbital polarization is maximized to the extent allowed by the first two rules . in the absence of bias , the ordering of the first four states ( with spin up ) is given by @xmath126 in this order .
the next four states follow the same order but with spin down .
the occupation of these eight states are given by the filling factor @xmath127 ranging from @xmath22 ( state @xmath128 with spin up fully filled ) to @xmath129 ( all eight states filled ) .
to simplify the notation , we define@xmath130so that @xmath131 with @xmath132 it is easy , using eq .
( [ iterationp ] ) , to prove the sum rules@xmath133where @xmath134
in order to compute the collective excitations , we define the two - particle matsubara green s function@xmath135where again @xmath136 are orbital indices and @xmath137 are valley indices . to derive the equation of motion for these response functions in the generalized random - phase approximation ( grpa ) , we proceed in the following way .
we first derive the equation of motion for @xmath138 in the hartree - fock approximation ( hfa ) using the heisenberg equation of motion @xmath139 , \]]where @xmath140 is the hamiltonian of eq .
( [ hamil ] ) with the averages removed , @xmath141 is the chemical potential and @xmath13 the number operator ( not to be confused with the landau level index ) . after evaluating the commutators ,
we linearize the resulting equation by writing @xmath142 .
we get the grpa equations of motion by keeping the terms up to linear order in @xmath143 in the homogeneous states at integer fillings , @xmath144 so that we get the set of equations @xmath145 \chi _ { n_{1},n_{2},n_{3},n_{4}}^{\left ( 0\right ) a , b , c , d}\left ( \mathbf{q}% , \omega _ { n}\right ) \notag \\ & = & \hslash \left\langle \rho _ { n_{1},n_{4}}^{a , d}\right\rangle \delta _ { n_{2},n_{3}}\delta _ { b , c}-\hslash \left\langle \rho _ { n_{3},n_{2}}^{c , b}\right\rangle \delta _ { a , d}\delta _ { n_{1},n_{4 } } \label{chihfaa } \\ & & + x_{n_{1}^{\prime } , n_{1},n_{3}^{\prime } , n_{2}^{\prime } } ^{a , b^{\prime } } \left ( 0\right ) \left\langle \rho _ { n_{1}^{\prime } , n_{2}^{\prime } } ^{a , b^{\prime } } \right\rangle \chi _ { n_{3}^{\prime } , n_{2},n_{3},n_{4}}^{\left ( 0\right ) b^{\prime } , b , c , d}\left ( \mathbf{q}% , \omega _ { n}\right )
\notag \\ & & -x_{n_{1}^{\prime } , n_{4}^{\prime } , n_{2},n_{2}^{\prime } } ^{a^{\prime } , b}\left ( 0\right ) \left\langle \rho _ { n_{1}^{\prime } , n_{2}^{\prime } } ^{a^{\prime } , b}\right\rangle \chi _ { n_{1},n_{4}^{\prime } , n_{3},n_{4}}^{\left ( 0\right ) a , a^{\prime } , c , d}\left ( \mathbf{q},\omega _ { n}\right ) , \notag\end{aligned}\]]and @xmath146where @xmath147 is a bosonic matsubara frequency .
the retarded response functions are obtained , as usual , by taking the analytic continuation @xmath148 by defining super indices @xmath149 representing the combinations @xmath150 etc .
, we can represent the response functions and interactions matrices as @xmath151 matrices and then write the grpa equation in the matrix form : @xmath152 \chi \left ( \mathbf{q},\omega \right ) = b\left ( \mathbf{q}\right ) , \label{grpamodes}\ ] ] where @xmath153 are @xmath151 matrices ( with @xmath154 the unit matrix ) .
the matrices @xmath155 and @xmath156 depend on the @xmath157 evaluated in the hfa .
we will give later the precise form of these matrices for the phases studied in this paper .
the frequencies of the collective excitations are given by the eigenvalues of the matrix @xmath158 there are in total 4 zero modes , corresponding to unphysical intra - level transitions , and 12 non - zero modes corresponding to inter - level transitions .
the latter occur in six positive - negative energy pairs , corresponding to excitation and deexcitation partners .
of the six collective excitations identified in this way , three are pauli - blocked at @xmath7 and appear in our calculations as dispersionless modes that have zero weight in all physical response properties .
the three remaining excitation modes are physical and for @xmath159 correspond to the interaction - coupled transitions indicated in fig
. modes . note that , in the limit @xmath160 , @xmath161 so that eq .
( [ chigrpa ] ) gives @xmath162 . in this limit
, the collective mode frequencies correspond to transitions between eigenstates of the hartree - fock hamiltonian @xmath163 as illustrated in fig .
[ modes ] for the case @xmath164 and zero bias .
the energy of these eigenstates include the non - interacting energies and the self - energy corrections .
the @xmath165 limit of these sames modes , however , also includes the polarization and excitonic corrections that , in a feynman diagram description of the grpa , are captured by bubble and ladder diagram summations .
these effects make the modes dispersive .
, the first three energy states of the hartree - fock hamiltonian are filled and the only possible transitions are those indicated by the three arrows .
the frequency of these transitions corresponds to the frequency of the collective excitations calculated in the grpa in the limit in which the wavevector @xmath166 we illustrate here the case of zero bias . ]
the properties of the ground state at @xmath22 have been studied in detail in ref . .
at zero bias , the ground state has all electrons in the @xmath128 state and can be described as an @xmath168 layer - pseudospin ferromagnet with orbital character @xmath0 .
the bias @xmath56 acts as an effective external magnetic field that forces the layer pseudospins out of the @xmath36 plane . above a critical bias @xmath169 ,
all electrons are in the bottom layer and the layer - pseudospin is correspondingly fully polarized .
the ground state is then given by @xmath170 and is unchanged if the bias is further increased . the @xmath7 state on which we focus here differs from the @xmath22 state because of a competition between single - particle and interaction energy effects which emerges only in the former case . as illustrated in fig .
( [ niveaux ] ) , single - particle effects captured by the bilayer effective hamiltonian favor occupation of the @xmath1 orbital when three of the octet s eight levels are occupied ( @xmath7 ) .
note that this tendency is independent of the sign of @xmath56 .
exchange interactions , on the other hand , always favor a state in which as many @xmath0 orbitals as possible are occupied .
as we explain below , a compromise is reached by forming a state with coherence between @xmath0 and @xmath171 orbitals .
this physics is enriched by the same tendency toward interlayer coherence which occurs at @xmath22 and@xcite in semiconductor bilayers . indeed ,
our calculations show that the phase diagram at @xmath7 is much more complex than at @xmath22 .
_ at zero bias _
, numerical solution of the hfa equations leads to occupied @xmath172 and @xmath173 states .
the order parameters are then given by @xmath174the ground state is an @xmath168 layer - pseudospin ferromagnet with orbital character @xmath175 the hamiltonian is invariant with respect to the orientation of the pseudospins in the @xmath36 plane so that this phase supports a goldstone mode . the choice of phase in eq .
( [ phase ] ) has the pseudospins pointing along the @xmath176 axis . at finite bias @xmath177 ,
the pseudospins are pushed out the @xmath36 plane i.e. the two layers have unequal population .
the occupation of the four states are in this case@xmath178and@xmath179with inter - layer coherence reflected by @xmath180 we define the critical bias @xmath181 as the bias at which the inter - layer coherence @xmath182 it is given by@xmath183where the fock interactions @xmath184 and @xmath185 are defined in appendix a. as an example , for @xmath186 t , @xmath187 the energy of this inter - layer - coherent state ( ics ) for @xmath177 is @xmath188where the components of the layer pseudospin are given by@xmath189with the convention that pseudospin up is state @xmath190 while pseudospin down is state @xmath191 the hfa equations have a separate set of solutions , favored at larger bias voltages , in which the lower layer is maximally occupied , and the ground state has upper - layer orbital coherence instead of inter - layer coherence .
this solution has @xmath192@xmath193and inter - orbital coherence is signalled by the density - matrix components @xmath194here @xmath195is the critical bias above which all charges in the upper layer are transferred to state @xmath190 and the orbital coherence is lost . in a pseudospin model with the convention : pseudospin up for state @xmath196 and pseudospin down for state @xmath197 , the orbital pseudospin components are given by @xmath198this orbital - coherent phase has all pseudospins tilted slightly away from the @xmath199 axis by an angle@xmath200at the critical bias @xmath201 , @xmath202 this critical bias @xmath201 is very large ; at @xmath186 t , @xmath203 which is near the limit of validity of the effective two - band model i.e. @xmath204 .
for @xmath205 , all electrons are in state @xmath197 and there is no further change with bias of the ground state .
the orbital - coherent state ( ocs ) has an energy given by@xmath206 + % \frac{1}{12}\frac{d}{\ell } \notag \\ & & -\frac{5}{24}\left [ x_{1}\left ( 0\right ) + x_{16}\left ( 0\right ) + 2x_{7}\left ( 0\right ) \right ] \label{ener22 } \\ & & -\frac{1}{6}\left [ x_{1}\left ( 0\right ) -x_{16}\left ( 0\right ) \right ] \left\langle s_{z , k}\right\rangle \notag \\ & & -\frac{1}{6}\left [ x_{1}\left ( 0\right ) + x_{16}\left ( 0\right ) -4x_{7}\left ( 0\right ) \right ] \left\langle s_{z , k}\right\rangle ^{2 } \notag \\ & & -\frac{1}{3}x_{4}\left ( 0\right ) \left\langle \mathbf{s}_{\bot , k}\right\rangle ^{2}. \notag\end{aligned}\]]this energy is independent of the azimuthal angle @xmath207 of the pseudospin vector .
we can thus , without loss of generality , take @xmath208 as real . ]
( [ ener1 ] ) and ( [ ener22 ] ) give @xmath209 above a critical bias @xmath210 .
for biases @xmath211 , $ ] a mixed state with both inter - orbital _ and _ inter - layer coherence would be lower in energy than a state with only interlayer coherence . solving the full hartree - fock equations ,
we find that the crossover from the inter - layer coherent state to the inter - orbital coherent state occurs continuously via an intermediate state with both orders . the boundaries of the intermediate phase must be determined numerically .
we find that the boundaries of this mixed state ( ms ) are given on the left by a new critical bias @xmath212 and on the right by @xmath213 the intermediate phase region broadens with magnetic field as can be seen in fig .
[ biasnum1 ] .
figure [ pordre ] shows the evolution of the inter - layer coherence @xmath214 and the orbital coherence @xmath208 with bias at magnetic field @xmath186 t. the orbital coherence sets in before the inter - layer coherence decreases to zero thus creating the mixed state region identified in this figure by a non zero value of the density - matrix component @xmath215 .
( note that @xmath216 is not given by eq .
( [ m3 ] ) in the mixed state . ) the coherences @xmath217 and @xmath218 which involve a mixing of valley and as well as orbital indices , are non - zero only in the intermediate mixed - state region of the phase diagram .
all density - matrix components vary continuously with inter - layer bias . as can be seen from figs .
[ biasnum1 ] and [ pordre ] , the orbital - coherent phase starts at @xmath219 with a finite orbital coherence @xmath216 .
were it not for the presence of the inter - layer - coherent and mixed states , @xmath220 would start at zero bias and be given by eq .
( [ m3 ] ) for all biaises .
the mixed and inter - layer coherent states are confined to relatively small inter - layer bias voltages ; at larger values of @xmath56 the ground state is a relatively simple state with only orbital coherence .
the exploration of collective excitation properties of this state is one key objective of this paper . and inter - orbital @xmath221 coherences ( left axis ) with bias at filling factor @xmath222 and magnetic field @xmath186 t. on the right axis , one of the non zero coherences @xmath223 in the mixed state . ]
the density - matrix equations of motions which describe the three @xmath7 spin - diagonal dispersive collective modes ( see fig . [ modes ] ) are usually simplest when written in the basis of the bonding and antibonding single - particle hartree - fock eigenstates . for the interlayer coherent state ( ics ) ( the ground state for @xmath224 @xmath225 ) we find that , @xmath226where @xmath227 is the angle between the two - dimensional wavevector @xmath228 and the @xmath176 axis , and @xmath229 and @xmath230 refer to the states @xmath231where @xmath232@xmath233and@xmath234the matrix @xmath235 depends only on the modulus of the wavevector @xmath228 so that the dispersions are isotropic .
this matrix is given by @xmath236 where@xmath237the variables in this matrix are defined in appendix b. the grpa dispersions for the three collective modes are shown in fig .
modesnum1 for zero bias and a magnetic field of @xmath186 t. in the limit @xmath238 , the frequencies of the dispersive modes correspond to transitions between the hfa energy levels indicated in fig .
[ modes ] as expected .
mode @xmath239 in fig .
[ modesnum1 ] is a goldstone mode consisting of a precession of the inter - layer pseudospin @xmath240 around the @xmath176 axis .
we refer to it as the inter - layer pseudospin mode ( ipm ) .
mode @xmath241 is an orbital pseudospin mode ( opm ) consisting of a precession of the orbital pseudospins around their local equilibrium position .
mode @xmath242 involves both a layer and orbital pseudospin flip and has a large gap . at finite bias , the @xmath242 dispersive modes in fig .
[ modes ] are coupled together while at zero bias , modes @xmath239 and @xmath242 completely decouple from mode @xmath241 as is clear from eq .
( [ f1 ] ) . in the inter - layer coherent phase at zero bias and for @xmath243 t.
the numbers refer to the transitions indicated in fig . modes . ] ) dispersion relation of the collective modes in the inter - layer coherent phase at @xmath167 for zero bias and for @xmath186 t. mode @xmath241 is the orbital pseudospin mode while mode @xmath239 is the inter - layer pseudospin mode . ] from eq .
( [ f1 ] ) , it is easy to see that the dispersion of the opm at zero bias ( given by the @xmath244 block in the lower right of the matrix @xmath245 ) is given by@xmath246and has a gap given by ( see appendix b)@xmath247 .
\label{gaporbi}\]]this gap is small but visible in fig .
[ modesnum1petitq ] .
we find numerically , that as the bias is increased , the gap @xmath248 decreases until it reaches zero at the phase boundary of the mixed state .
the dispersion of the ipm is linear in @xmath26 for @xmath249 ( with @xmath250 for @xmath186 t ) as in a semiconductor bilayerspielmanprl .
this findings differ qualitatively from the @xmath251 case for which we found@xcite an ipm with @xmath252 dispersion and a gapless opm . at @xmath22 and zero bias ,
the @xmath128 level is filled .
the @xmath252 ipm dispersion in that case occurs because the possibility of mixing @xmath1 wavefunctions with @xmath253 wavefunctions in excited states allows the inter - layer phase stiffness to vanish . at @xmath7 ,
it is not possible to make the corresponding admixture . in pseudospin language
, a finite bias pushes the layer pseudospins @xmath254 out of the @xmath36 plane but the hamiltonian of the system remains independent of the orientation of the perpendicular component of these pseudospins in the @xmath36 plane .
the ipm therefore remains gapless for @xmath255 it acquires a gap in the mixed and ocs . at filling factor @xmath22 ,
the inter - layer - pseudospin mode becomes unstable at finite bias .
this indicates that the uniform inter - layer - coherent state can not in fact be the ground state at finite bias .
we see no such instability at @xmath256 in the absence of a bias , eq .
( [ f1 ] ) shows that modes @xmath239 and @xmath242 are coupled through @xmath13 and @xmath257 these interactions involves the coulomb interaction matrix elements @xmath258 and @xmath259 ( see appendix a for their definitions ) .
these interactions do not conserve total @xmath0 or @xmath1 quantum numbers .
such interactions do not occur in usual semiconductor 2deg where spin and layer pseudospin indices are conserved .
in this section , we consider collective excitations of the orbital coherent state ( ocs ) which is the ground state in the region @xmath260 , which covers the large region of bias voltages from small values to the largest values for which the two - band effective model applies .
the occurrence of the interesting ocs state at high bias voltages is a consequence of competition between single - particle and interaction effects as explained earlier . by studying its collective interactions
we reveal a dzyaloshinskii - moriya ( dm ) interaction between orbital pseudospins and demonstrate that for large bias voltages it drives an instability to an orbital pseudospin spiral state .
the fact that @xmath261 in the ocs implies that there is a finite density of electric dipoles in this phase as first pointed out in ref . .
to show this , we write the total electronic density ( including the two valleys ) as@xmath262where@xmath263with @xmath264 .
the functions @xmath265 are defined in eqs .
( [ k1]-[k4 ] ) . in our pseudospin langage for the orbital states
, we have the relations ( for the @xmath3 valley ) @xmath266 , \\ \rho _ { + } ^{k}\left ( \mathbf{q}\right ) & = & \rho _ { 0,1}^{k , k}\left ( \mathbf{q}% \right ) = \rho _ { x}^{k}\left ( \mathbf{q}\right ) + i\rho _ { y}^{k}\left ( \mathbf{% q}\right)\end{aligned}\]]and so the density operator in the @xmath3 valley can be written as@xmath267where we have defined @xmath268 and similarly for @xmath269 with @xmath270 now , if the 2deg is in an external electric field @xmath271 we have for the coupling hamiltonian@xmath272 \phi \left ( \mathbf{q}\right ) .
\notag\end{aligned}\]]with the electric field in the plane of the 2deg , this coupling can be written , in real space , as@xmath273 \notag \\ & & + \sqrt{2}\ell \int d\mathbf{r}\left [ \overline{\rho } _ { j , x}\left ( \mathbf{r% } \right ) e_{x}\left ( \mathbf{r}\right ) -\overline{\rho } _ { j , y}\left ( \mathbf{% r}\right ) e_{y}\left ( \mathbf{r}\right ) \right ] .
\notag\end{aligned}\]]the inter - orbital coherence is zero in the @xmath4 valley and so @xmath274 eq . ( [ dipole ] ) implies that the dipole density in the 2d plane is@xmath275or@xmath276 the orientation of the dipole density is set by the phase of the density - matrix component @xmath277 @xmath278 which specifies the phase of the spontaneously established coherence between @xmath0 and @xmath1 orbitals in the ground state . for our choice of the spontaneously established phase of @xmath279 in eq .
( [ m3 ] ) , the dipoles are oriented along the @xmath176 axis . because the @xmath4 valley ( bottom layer ) landau levels are maximally filled , there is no inter - orbital coherence and therefore no contribution to the electric - dipole density from the @xmath76 valley .
collective modes dispersions for the ocs are plotted in fig .
3modesgrandq . for relatively small values of @xmath56
they are very similar to those represented in fig .
[ modesnum1 ] for the ics ; the main changes occur at small wavevector as can be seen in the inset of fig .
the inter - layer pseudospin mode is gapped in the ocs while the orbital pseudospin mode is gapless , with a very anisotropic dispersion as shown in fig .
[ petitsq ] .
we now discuss the physics of the orbital pseudospin mode . in the orbital - coherent state at filling factor @xmath280 bias @xmath281 @xmath282 and magnetic field @xmath283 t. mode
@xmath241 is the goldstone mode due to orbital coherence .
modes @xmath239 and @xmath242 involve an inter - layer transition and are gapped .
the inset shows the dispersion of mode @xmath241 ( gapless ) and mode @xmath239 ( gapped ) at small @xmath284 ] at finite wavevector @xmath228 , the orbital pseudospin mode ( opm ) corresponds to a precession of the orbital pseudospins around their equilibrium orientation in the ground state as illustrated in fig .
the ground state is described by spinors @xmath285where @xmath286 $ ] has been defined in eq .
( angle ) . , bias @xmath287 @xmath282 and magnetic field @xmath186 t. curve ( 1 ) @xmath288 curve ( 2 ) @xmath289 and curve ( 3 ) @xmath290 ] around its equilibrium position with polar angle @xmath291 and azimuthal angle @xmath292 .
this orientation is choosen as the orbital pseudospin quantization axis in the pseudospin wave model .
the angle @xmath293 is the polar angle measured from the equilibrium orientation of the pseudospins . ]
the collective mode corresponds to spatially coherent rotations of this spinor around its ground state value . for this reason ,
it is convenient to choose the orbital pseudospin quantization axis along the direction @xmath294 . in order to do this , we define ` bonding ' and ` antibonding ' electron creation operators by @xmath295 below we use the convention that pseudospin up corresponds to state @xmath296 and pseudospin down to state @xmath297 and we denote the pseudospin by @xmath298 in our grpa system of equations for the collective modes , the orbital - pseudospin wave mode is decoupled from all other modes . below
we follow one possible strategy for explaining the physics of this mode by comparing our microscopic grpa equation of motion to the equations of motion of an effective orbital pseudospin model , and using the comparison to identify the effective pseudospin interactions .
since collective modes correspond to small oscillations of the pseudospin around its quantization direction , we can use an effective model which has interactions only between transverse spins .
( quantization direction interactions can be represented as transverse interactions because of the spin - magnitude constraint . )
we write the pseudospin effective hamiltonian in momentum ( @xmath228 ) space : @xmath299where @xmath300 since @xmath301 in the real space version of eq .
( [ pseudospinh ] ) depends on @xmath302 only , it follows that we can always write @xmath303 and hence @xmath304 .
because the real - space interactions must be real we also have the usual property that @xmath305 . combining these two identities
we can conclude that @xmath306 and @xmath307 are real and even in @xmath228 , while @xmath308 has even real and odd imaginary contributions .
the real parts of @xmath309 and @xmath310 are identical while their imaginary parts differ in sign .
as we emphasize further below , the dm interaction is captured by the imaginary part of @xmath309 .
using these properties and the commutation relation,@xmath311 = \frac{1}{2}i\delta _ { \mathbf{q},-\mathbf{q}% ^{\prime } } , \]]the equations of motion of the pseudospin model are : @xmath312with the dispersion relations@xmath313note that the first term on the right hand side of eq .
( [ disper ] ) is real and odd in @xmath228 and that it represents the contribution of the dm interaction to the collective mode frequency . comparing with our microscopic grpa results for the equations of motion and collective mode frequencies , we obtain the following expressions for the pseudospin effective interactions : @xmath314 -\sin ^{2}(\theta _ { b})\left ( \frac{1}{2}a_{1}+\re \left [ a_{2}\right ] \right ) \label{integ } \\
j_{yy } & = & \frac{1}{2}a_{3}-\re \left [ a_{2}\right ] , \notag \\
j_{xy } & = & -\frac{i}{2}\allowbreak \sin ( \theta _ { b})a_{4}-\cos ( \theta _ { b})\im \left [ a_{2}\right ] , \notag \\ j_{yx } & = & \frac{i}{2}\allowbreak \sin ( \theta _ { b})a_{4}-\cos ( \theta _ { b})\im \left [ a_{2}\right ] , \notag\end{aligned}\]]with @xmath315 , \notag\end{aligned}\]]where @xmath227 is the angle between the wavevector @xmath228 and the @xmath176 axis .
all interactions are defined in appendix a. all hartree and fock interaction terms in eqs .
( [ inter ] ) , @xmath316 and @xmath317 , are real and depend only on the modulus of @xmath318 we see from the structure of eqs .
( [ integ ] ) that the dispersion relation has the symmetry @xmath319 because @xmath320 ( from eq . ( [ angle ] ) ) , it follows that the dispersion at small bias is the same as the dispersion near the critical bias as first pointed out in ref . .
the physical content of the various terms in eqs .
( [ integ ] ) is most easily identified from their long - wavelength forms . from eqs .
( [ integ ] ) we find that at small @xmath26 and small @xmath321 : @xmath322 the pseudospin rotations which change @xmath323 correspond to changes in the angle @xmath324 on the orbital pseudospin bloch sphere relative to the ground state value @xmath325 as illustrated on fig .
[ blochs ] . for this reason @xmath326 remains finite ( is massive ) as @xmath327 when the potential bias is finite unlike the other couplings .
the terms proportional to @xmath328 in @xmath329 , @xmath330 in @xmath331 and @xmath332 in @xmath333 and @xmath334 are simply electrostatic interactions between changes generated when the dipole orientation varies in space .
( recall that the charge density is equal to the divergence of the dipole density . )
these terms are the long - wavelength limits of the hartree interactions captured by the grpa theory .
the imaginary contribution to @xmath335 is the dm interaction whose physics we discuss below .
the eigenvector for the pseudospin motion , at small @xmath26 and small @xmath56 , has @xmath336 if @xmath337 and @xmath338 if @xmath339 so that the long wavelength collective modes are elliptical precessions with minor axis along the massive @xmath340 direction and major axis along the @xmath341 direction which contributes dipolar electrostatic energy .
the long wavelength goldstone collective mode energy therefore has unusual square root dispersion : @xmath342for @xmath343 , we have the linear dispersion : @xmath344we see later that the dm interaction assumes a larger importance at larger bias potentials and shorter wavelengths .
the orbital coherent state occurs at finite bias and is preempted at small biases by the interlayer coherent state .
it is nevertheless interesting to examine the artificial limit in which @xmath345 , but layer degrees of freedom are still not in play . in that limit
all electrons would be in the @xmath0 orbital , there would be no electric dipoles in the ground state , and the exchange parameters would be given by@xmath314 \\
j_{yy } & = & \frac{1}{2}a_{3}-\re \left [ a_{2}\right ] , \notag \\ j_{xy } & = & -\im \left [ a_{2}\right ] , \notag \\ j_{yx } & = & -\im \left [ a_{2}\right ] .
\notag\end{aligned}\]]the dispersion relation would then be given by @xmath346which is isotropic .
the long - wavelength limit dispersion would become @xmath347similar to @xmath22 behavior@xcite .
as explained previously the dm interaction is captured by the imaginary part of @xmath348 .
when this contribution to the orbital pseudospin hamiltonian is isolated it yields an interaction of the standard@xcite dm form : @xmath349 \left ( \mathbf{s}\left ( \mathbf{q}\right ) \times \mathbf{s}% \left ( -\mathbf{q}\right ) \right ) \cdot \widehat{\mathbf{z}}.\]]an examination of eqs .
( [ inter ] ) shows that this interaction is not due to electrostatic dipole interactions , and instead to the exchange vertex corrections i.e. to the interactions @xmath350 and @xmath351 . in appendix c
, we analyze the exchange energy of quantum hall ferromagnets quite generally and show that dm interactions are the rule rather than the exception when the two - states from which the pseudospin is constructed have the same spin .
the physics of exchange interaction contributions to collective mode energies is most simply described by making a particle - hole transformation for occupied states , as discussed in appendix c. the exchange interaction at momentum @xmath352 can then be related@xcite to the attractive interaction between an electron and a hole separated by @xmath353 .
dm interactions occur when the pseudospin state of the electron or hole give rise to cyclotron orbit charge distributions which do not have inversion symmetry , a property that holds here because of dipole formation .
these distortions of the cyclotron orbit are irrelevant when @xmath352 is very large but become important for @xmath354 . in the case of a simple parabolic band @xmath9 quantum hall ferromagnet , for example
, this picture@xcite provides a simple understanding of the full spin - wave dispersion .
the dm interaction is strongest at @xmath355 , _
i.e. _ when @xmath356 .
we have found that over a broad range of @xmath56 values the dm interaction is strong enough to induce an instability of the uniform coherent state . when viewed as a classical complex - variable quadratic form for gaussian energy fluctuations , the orbital pseudospin hamiltonian , eq .
( [ pseudospinh ] ) , is positive definite provided that @xmath357 is positive , @xmath358 is positive , and @xmath359explicit numerical calculations show that the first two stability requirements are always satisfied , but that because of the dm interaction , the third is not satisfied when @xmath360 is large . in the grpa
we find that the opm first becomes soft at @xmath361 when @xmath362 @xmath363 .
( we have @xmath364 at this value of @xmath56 . ) fig .
figdm shows the instability of the orbital - pseudospin mode at @xmath365 we remark that the instability occurs at a positive ( negative ) value of @xmath366 in @xmath367 the higher - energy collective modes ( not shown in the figure ) show no sign of instability . and bias @xmath368 @xmath369 ] the eigenvector with positive frequency of eq .
( [ jdisper ] ) has@xmath370it follows , using eq .
( [ condi ] ) , that , at the dm instability , the energy is lowered by forming coupled density - waves in @xmath340 and @xmath371 pseudospin components with : @xmath372 the real part of the @xmath333 coupling , due mainly to the dipole electrostatic energy , is very small at the instability wavevector because the @xmath2 landau level can not support rapid spatial variation , as we have verified by explicit calculation . because @xmath358 is real , it follows that @xmath323 and @xmath373 spatial variations are out of phase by nearly exactly @xmath374 . if the magnitudes of the @xmath323 and @xmath373 components were identical , this would imply a spiral ground state . because @xmath375 and @xmath358 are not identical at the instability , the spiral is somewhat distorted .
it must be kept in mind , however , that the dm instability may be preempted by a first order transition to a state with lower energy and a more complex pseudospin pattern .
a fuller exploration of the properties of these states , including their properties in the presence of an external electric field , is beyond the scope of the present work .
the mixed state occurs between the inter - layer coherent and inter - orbital coherent phases as shown in fig .
[ biasnum1 ] .
the width of this region in the phase diagram increases with magnetic field . in this phase , all order parameters @xmath376 are finite so that this phase has both inter - layer and inter - orbital coherence .
we show the behavior of some of these order parameters with bias in fig .
[ pordre ] .
the order parameters @xmath377 and @xmath378 which flip both valley and orbital indices are non zero in this phase .
the collective modes in the ms are obtained numerically by solving eq .
( grpamodes ) .
the ms has three dispersive modes , as in the other two phases we studied .
the dispersions of these modes differ from the dispersions in the other two phases at small wavevector only .
we show in fig .
[ mixed](a ) the dispersion of the inter - orbital mode and in fig .
[ mixed](b ) the dispersion of the inter - layer coherent mode along the @xmath176 or @xmath379 axis and at @xmath380 ( @xmath381 ) from the @xmath176 axis .
in contrast with the other two phases , both modes are now gapped for @xmath56 not at the boundaries of this phase . as in the inter - orbital phase ,
the dispersions are highly anisotropic .
we show in the next section that this phase has a distinct signature in the microwave absorption spectrum .
@xmath382 and @xmath186 t along directions @xmath383 and @xmath381 ( @xmath380 from the @xmath176 axis ) . ]
the collective modes discussed in the previous sections can be detected in microwave absorption experiments , as we now show .
we write the current operator , projected onto @xmath2 and valley @xmath3 , as @xmath384where @xmath385 is the vector potential of the external electromagnetic field , @xmath386 and @xmath387 is given in eq .
( [ deux ] ) .
in second quantization , the total current is given by @xmath388with the field operators defined in eqs .
( [ field1],[field2 ] ) and @xmath389 we find that @xmath390where @xmath391and similarly for @xmath392 the same result for @xmath393 can be obtained by calculating the polarization current@xmath394with the dipole density @xmath395 defined in eq .
( [ dipoleq ] ) and using the heisenberg equation of motion @xmath396 .$ ] we note that the orbitally coherent state has spontaneous currents in its ground state , a very exceptional property .
it appears likely that , in finite systems , these currents should flow perpendicular to system boundaries , forcing domain structures in the pseudospin magnetization texture , consistent with expectations based on the electrostatic energy of the dipole density .
we define the total current - current correlation function matsubara green s functions as @xmath397where @xmath398 and @xmath399with @xmath400 and @xmath401 note that @xmath402 are defined so that @xmath403 the microwave absorption for an electric field oriented along the direction @xmath404 is given by@xmath405 e_{0}^{2 } \label{mwabsorption } \\ &
= & -\frac{1}{2}\left ( \frac{e^{2}}{h}\right ) \left ( \frac{\delta _ { b}}{% \gamma _ { 1}}\right ) ^{2}\omega _ { c}^{\ast 2 } \notag \\ & & \times \sum_{i , j = k , k^{\prime } } \im \left [ \xi _ { i}\xi _ { j}\frac{\chi _ { \rho _ { i,\overline{\alpha } , j,\overline{\alpha } } } ^{r}\left ( \omega \right ) } { \omega + i\delta } \right ] e_{0}^{2 } , \notag\end{aligned}\]]where we have assumed a uniform electric field @xmath406 and taken the analytic continuation @xmath407 of @xmath408 to get the retarded response function .
the response functions @xmath409 are calculated in units of @xmath410 so that @xmath411 is the power absorbed per unit area . in eq .
( [ mwabsorption ] ) we have neglected a diamagnetic contribution to the current response which becomes important at low frequencies .
our grpa correlation functions are given by eq .
( [ grpamodes ] ) and numerical results for the absorption in the inter - layer coherent phase are shown in fig .
exactly the same result is obtained , in this phase , if the electric field is set in the @xmath379 direction , _
i.e. _ the absorption is isotropic .
we see that the signal in the absorption is at a frequency corresponding to the orbital - pseudospin mode ( see fig .
modesnum1petitq ) .
the frequency @xmath6 of this mode at @xmath104 decreases with bias while the absorption intensity increases with bias . since @xmath412 ghz ( using @xmath413 for graphene on sio@xmath19 substrate )
, the frequency of the orbital pseudospin mode at zero bias is @xmath414 ghz @xmath415 in the microwave regime .
note that mode @xmath242 in fig .
modesnum1 is also present in the absorption at finite bias and that its frequency has a higher value outside of the microwave regime . the absorption in the mixed state is shown in fig .
[ absomixed ] on a logarithmic scale .
the three lower peaks show the absorption from the inter - layer coherent mode which is gapped in the mixed state .
the other three peaks are from the inter - orbital mode .
the gaps in these two modes increase with bias until they reach a maximum around @xmath416 for @xmath186 t. the gaps then decreases with bias . the intensity of the absorption increases with bias for both modes . in fig .
[ absomixed ] , the electric field is set along the @xmath379 axis .
the absorption is at least @xmath417 times lower if the electric field is oriented along the @xmath176 axis i.e. it is highly anisotropic in the mixed state . with our choice of phase for the ground state of the mixed state , the pseudospins ( and so the electric dipoles according to eq .
( [ dipoled ] ) ) are oriented in the @xmath418 direction .
their motion for @xmath419 is an oscillation in the @xmath36 plane about the @xmath379 axis .
for such a configuration , the electromagnetic absorption is strongest for fields along the @xmath379 axis , which is what we observe in our calculation . in the orbital - coherent phase ,
the orbital - pseudospin mode is gapless and decoupled from the two other gapped modes . since the orbital - coherent mode couples strongly to external electric fields
, we can expect anomalous low - frequency absorption in this state , similar to the drude absorption of a metal .
this interesting and unusual absorption feature is likely to be highly sensitive to disorder .
its detailed analysis lies beyond the scope of the present paper . above @xmath420 ,
the ground state has @xmath421 and there is no orbital coherence anymore .
the opm then has a gap that is proportional to @xmath422 the opm becomes visible in the absorption in this phase while the other modes do not . in summary , we see that each of the three phases in the phase diagram at @xmath37 has a different signature in the microwave absorption spectrum .
we remark that the frequencies in fig .
[ absomixed ] are quite small .
but , they can be increased by increasing the magnetic field . for example , fig .
[ abso40 ] shows the gap in the inter - layer and inter - orbital pseudospin modes in the three phases at a magnetic field of @xmath423 t. t for an electric field in the @xmath379 direction and for biases @xmath424 in units of @xmath282 ] and @xmath425 t. ]
we have studied the phase diagram and collective excitations of a spin - polarized bilayer graphene 2deg at filling factors @xmath7 and @xmath426 , as a function of a bias electric potential which shifts electrons between layers .
our study is based on the hartree - fock approximation for the mean - field ground state , and the grpa for the collective modes and response functions .
we predict phase transitions between the following sequence of states with increasing potential bias : ( 1 ) an inter - layer - coherent state ( ics ) with a zero gap inter - layer pseudospin mode ( ipm ) and an orbital pseudospin mode ( opm ) with a small gap .
this gap can be detected in microwave absorption experiment .
its frequency decreases with bias while its intensity increases with bias ; ( 2 ) a mixed state ( ms ) with both inter - layer and inter - orbital coherence .
both the ipm and opm are gapped and visible in microwaves in this phase
. moreover , the intensity of the absorption is highly sensitive to the direction of the external electric field ; ( 3 ) an orbital coherent state ( ocs ) with orbital coherence concentrated in one layer only .
( the second layer is completely filled . )
the ocs state is a very simple one in which the low potential layer is filled and the high potential layer has a gap induced between it s two landau levels by spontaneously establishing coherence between states with @xmath1 and @xmath0 orbital character .
this state has a number of quite unusual properties , including electric dipoles and associated uniform currents in its ground state .
the ocs has a gapless ( goldstone ) opm and a gapped ipm .
both modes are absent from the absorption spectrum .
the dispersions of the collective modes are also highly anisotropic in this phase .
the phase is unstable at a finite wavevector due to the presence of a dzyaloshinksii - moriya exchange interaction .
we believe that the instability will lead to the formation of a ground state with a non - uniform pseudosopin pattern .
these properties are associated with competition between an electron - electron interaction term which favors @xmath0 orbital occupation , and single - particle terms in the effective two - band model of ref . which favors @xmath1 orbital occupation more strongly at larger inter - layer potential difference .
we note that this band model neglect the @xmath427 hopping term , which connects sites @xmath428 and @xmath429@xcite . strictly speaking , the results obtained in our paper are valid if @xmath430 the effect of the @xmath431 term is to add a negative correction to the energy of the @xmath1 orbital which is independent of the valley index , magnetic field and bias . more precisely : @xmath432 with @xmath433 it is not clear from the literature what is the precise value of the @xmath434 parameter . in ref . , the value @xmath435 is obtained by comparing the tight - binding model for bilayer graphene with the slonczewski - weiss - mcclure tight - binding model for bulk graphitemcclure . in bulk graphite , @xmath436 ev . with @xmath437 , the correction @xmath438 for @xmath439
@xmath440 it follows that this term may have an effect on the existence of the interlayer coherent and mixed phases , and certainly on the location of their phase boundaries , @xmath225 and @xmath441 , since level @xmath1 may be below level @xmath0 in _ both _ layers .
the dzyaloshinksii - moriya transition , however , occurs at a large bias of @xmath442 @xmath96 in a region where the single - particle energy of level @xmath1 is already well below that of level @xmath443 for this reason we believe that the physics in this region should not be affected by @xmath444 more numerical work is needed , however , to assess the precise influence of this term .
macdonald was supported by the nsf under grant dmr-0606489 and by the welch foundation .
r. ct was supported by a grant from the natural sciences and engineering research council of canada ( nserc ) .
y. barlas was supported by a grant from the state of florida .
computer time was provided by the rseau qubcois de calcul haute performance ( rqchp ) .
99 h. a. fertig , phys .
b * 40 * , 1087 ( 1989 ) ; a. h. macdonald , p. m. platzman and g. s. boebinger , phys .
lett . * 65 * , 775 ( 1990 ) ; r. ct , l. brey , and a.h .
macdonald , phys .
b * 46 * , 10239 ( 1992 ) ; xiao - gang wen and a. zee , phys .
lett . * 69 * , 1811 ( 1992 ) ; k. moon _
b * 51 * , 5138 ( 1995 ) ; j. p. eisenstein and a. h. macdonald , nature * 432 * , 691 ( 2004 ) .
l. brey and h. a. fertig , phys .
b * 62 * , 10268 ( 2000 ) ; r. ct and h. a. fertig , phys . rev .
b * 65 * , 085321 ( 2002 ) . a. k. geim and k. s. novoselov , nature materials * 6 * , 183 ( 2007 ) ; a. h. c. neto , f. guinea , n. m. r. peres , k. s. novoselov , and a. k. geim , rev . mod
. phys . * 81 * , 109 ( 2009 ) ; a. k. geim , science * 324 * , 1530 ( 2009 ) .
k. s. novoselov _ et .
nature physics * 2 * 177 ( 2006 ) .
e. mccann and v. i. falko , phys .
lett . * 96 * , 086805 ( 2006 ) .
m. koshino and t. ando , phys . rev .
b * 73 * , 245403 ( 2006 ) .
yafis barlas , r. ct , k. nomura , and a. h. macdonald , phys .
* 101 * , 097601 ( 2008 ) .
k. nomura and a.h .
macdonald , phys .
* 96 * , 256602 ( 2006 ) .
benjamin e. feldman , jens martin , and amir yacoby , nature physics , * 5 * , 889 ( 2009 ) ; y. zhao , p. cadden - zimansky , z. jiang , and p. kim , phys .
* 104 * , 066801 ( 2010 ) . for a commentary and references to related work see kostya novoselov , nature physics * 5 * , 862 ( 2009 ) .
yafis barlas , r. ct , j. lambert , and a. h. macdonald , phys .
lett . * 104 * , 096802 ( 2010 ) .
k. shizuya , phys . rev . * b * 79 , 165402 ( 2009 ) .
cyclotron resonance in bilayer graphene is discussed in d. s. l. abergel and v. i. falko , phys .
b * 75 * , 155430 ( 2007 ) and d. s. l. abergel and t. chakraborty , phys .
lett . * 102 * , 056807 ( 2009 ) .
w. kohn , phys .
rev . * 123 * , 1242 ( 1961 ) .
i. dzyaloshinksii , j. phys .
solids * 4 * , 241 ( 1958 ) ; t. moriya , phys . rev .
* 120 * , 91 ( 1960 ) .
a. h. castro neto , f. guinea , n. m. r. peres , k. s. novoselov , a. k. geim , rev .
phys . * 81 * , 109 ( 2009 ) .
i. b. spielman , j. p. eisenstein , l. n. pfeiffer , and k. w. west , phys .
* 87 * , 036803 ( 2001 ) . c. kallin and b. i. halperin , phys .
b * 30 * , 5655 ( 1984 ) .
b. partoens and f. m. peeters , phys .
b * 74 * , 075404 ( 2006 ) . j. c. slonczewski and p. r. weiss , phys
* 109 * , 272 ( 1958 ) ; j. w. mcclure , phys
. rev . * 108 * , 612 ( 1957 ) .
see for example : a. h. macdonald , in proceedings of the 1994 les houches summer school on mesoscopic quantum physics , edited by e. akkermans et al .
( elsevier science , amsterdam , 1995 ) , pp .
659 - 720 .
we first give the definitions of the hartree , @xmath445 and fock , @xmath446 interactions in eqs .
( [ hart1],[fock1],[hart2],[fock2 ] ) : @xmath447and @xmath448@xmath449where @xmath227 is the angle between the wavevector @xmath450 and the @xmath176 axis and @xmath451 the interactions @xmath452 are obtained by multiplying @xmath453 by @xmath454 where @xmath455 is the inter - layer separation .
the interactions @xmath456 and @xmath457 are obtained by removing the term @xmath458 and the phase factor @xmath459 or @xmath460 in @xmath453 and @xmath461 .
for example @xmath462 while @xmath463 the fock interactions are defined by@xmath464@xmath465@xmath466@xmath467@xmath468@xmath469@xmath470and @xmath471@xmath472the interactions @xmath473 are obtained by multiplying the integrand by @xmath474 . the interactions @xmath475 and @xmath476
are obtained by removing the imaginary term @xmath458 and the phase factor .
the combinations:@xmath477to define @xmath478 we follow the same procedure as for @xmath479 some useful constants : @xmath480@xmath481@xmath482@xmath483
the collective modes at @xmath7 and @xmath484 involve the matrices @xmath485 in eq .
( [ f1q ] ) .
the elements of this matrix are defined by @xmath486@xmath487@xmath488@xmath489@xmath490@xmath491@xmath492@xmath493@xmath494@xmath495@xmath496@xmath497 to get the gap given by eq .
( [ gaporbi ] ) , we use @xmath498
it is possible to derive a rather general and instructive expression for the exchange energy of a pseudospin-@xmath499 quantum hall ferromagnet , for the case in which the pseudospin texture varies in one direction only . in the following
we take this direction to be the @xmath340 direction and choose a landau gauge in which the guiding centers orbits are localized as a function of this coordinate .
if we are interested only in the dependence of energy on pseusospin texture we can assume that every guiding center orbital is occupied by one electron , but leave the pseudospin of that orbital arbitrary .
it is not necessary to immediately specify the orbital character of the states that form the pseudospin and we refer to them for the moment as state a and state b. in the case of immediate interest in this paper , state a has @xmath0 orbital character and state b has @xmath1 orbital character .
we discuss some other examples below .
the pseudospin texture can be specified by the two - component pseudospinors at each guiding center , @xmath500or by the guiding center dependent direction cosines of the pseudospin orientation : @xmath501 where @xmath502 and @xmath503 are the pseudospin orientation polar and azimuthal angles .
when a specific gauge choice is convenient we use @xmath504 and @xmath505 .
the exchange energy of a pseudospin quantum hall ferromagnet is @xmath506the dependence of exchange energy on pseudospin texture can be exhibited explicitly by using the property that each @xmath507 term in the fourier expansion of the two - particle interaction matrix elements can be separated into factors that depend on @xmath508 and @xmath509 independently .
it follows that @xmath510where the greek indices @xmath511 , @xmath512 represents the filling factor of the guiding center states , @xmath513@xmath514 is the magnetic length , @xmath515 is the fourier - transform of the electron - electron interaction and @xmath516 is an interaction form factor .
( [ eq : jab ] ) follows from the following expression for the plane - wave matrix elements : @xmath517where the overbar accent denotes complex conjugation .
the character of the cyclotron orbitals in the two nearly degenerate landau levels is captured by the single - particle form factors @xmath518 .
for example if pseudospin @xmath154 has orbital landau level index @xmath519 , @xmath520 where @xmath521 is the familiar two - dimensional electron gas landau - level form factor@xcite , commonly used in the analysis of many - different properties in the quantum hall regime . note that @xmath522 for the example of quantum hall ferromagnetism discussed in this paper , @xmath523 and @xmath524 when the bias potential is strong enough to yield complete layer polarization .
( the analysis in this section does not apply when both orbital and layer degrees of freedom are in play . )
when the @xmath525 and @xmath229 orbitals have opposite spins , a common occurrence in quantum hall ferromagnetism , @xmath526 .
examples in which the @xmath525 and @xmath229 orbitals are centered in different two - dimensional layers require a slight generalization of the present discussion which we do not explicitly address . using eq .
( [ fipi ] ) we find ( leaving the wavevector dependence of the single - particle and interaction form factors implicit ) that @xmath527 \notag \\ 4f^{0,y } & = & -2\mathrm{im}[\bar{f}^{a , a}f^{a , b}+\bar{f}^{b , a}f^{b , b } ] \notag \\
4f^{0,z } & = & |f^{a , a}|^{2}-|f^{b , b}|^{2}-|f^{a , b}|^{2}+|f^{b , a}|^{2 } \notag \\ 4f^{x , x } & = & 2\mathrm{re}[f^{a , a}\bar{f}^{b , b}+f^{b , a}\bar{f}^{a , b } ] \notag \\ 4f^{y , y } & = & 2\mathrm{re}[f^{a , a}\bar{f}^{b , b}-f^{b , a}\bar{f}^{a , b } ] \notag \\ 4f^{z , z } & = & |f^{a , a}|^{2}+|f^{b , b}|^{2}-|f^{a , b}|^{2}-|f^{b , a}|^{2 } \notag \\ 4f^{x , y } & = & 2\mathrm{im}[f^{a , a}\bar{f}^{b , b}+f^{b , a}\bar{f}^{a , b } ] \notag \\ 4f^{x , z } & = & 2\mathrm{re}[f^{a , a}\bar{f}^{b , a}-\bar{f}^{b , b}f^{a , b } ] \notag \\ 4f^{y , z } & = & -2\mathrm{im}[f^{a , a}\bar{f}^{b , a}-\bar{f}^{b , b}f^{a , b}].\end{aligned}\]]these results are obtained after replacing the pseudospinors by pseudospin direction cosines using @xmath528note that because the exchange energy is real , all the interaction
form factors are real .
the diagonal interaction form factors are even functions of @xmath228 , while the off - diagonal form factors have both even and odd contributions .
the interaction form factors capture the influence of the shape of the pseudospin state dependent cyclotron orbits on exchange energies . in the limit @xmath529
, orthogonality implies that @xmath530 . if follows that @xmath531 .
the exchange energy expression can be written in an alternate form by fourier transforming along the direction perpendicular to the guiding center orbitals , defining @xmath532where @xmath533 is the number of guiding center orbitals in a landau level .
the exchange energy then takes a form from which spin - wave dispersions can be simply read off : @xmath534with the momentum space exchange integral given by @xmath535because @xmath536 is real , @xmath537 .
similarly , for the diagonal elements of @xmath538 , the property that the diagonal elements of @xmath539 are even functions of @xmath228 implies that @xmath540 . combining both properties , we conclude that the diagonal momentum space interactions are real . the off - diagonal elements , however , can have both even real and odd imaginary contributions . for @xmath541 , the integration over @xmath542 in eq .
( [ jalphabeta ] ) has contributions from small @xmath543 only , allowing us to set @xmath544 .
the integral can then be recognized as an inverse fourier transform of the electron - electron interaction from momentum space back to real space .
it follows that , @xmath545this is the familiar peculiarity of quantum hall systems in which exchange interactions at large momenta are most simply understood@xcite after a particle hole transformation which converts them into hartree interactions between electrons and holes . in strong fields the particle and hole in a magnetoexciton with momentum @xmath352
are separated in real space by @xmath546 .
for very large @xmath352 , the separation between particle and hole is larger than the cyclotron orbit sizes and the interaction is approximately given by the interaction between point charges . for smaller values of @xmath352
the size and shape of the cyclotron orbit , represented in momentum space by @xmath547 becomes important .
imaginary off - diagonal contributions to @xmath548 are responsible for dm - like interactions in quantum hall ferromagnets . in the example discussed in this paper for instance ,
the @xmath525 orbital has orbital landau level index @xmath0 , while the @xmath229 orbital has the same spin and @xmath1 .
it follows that @xmath549 and @xmath550 , where @xmath551 is a laguerre polynomial , and that @xmath552 where @xmath553 is the orientation angle of @xmath228 .
it follows that @xmath554 vanishes .
the generalized random - phase - approximation that we use in the main text for collective mode calculations is equivalent to linearized pseudospin - wave theory . for fluctuations around a ground state with pseudospin orientation polar angle @xmath325 ( as discussed in the main text ) , the @xmath555 and @xmath556 pseudospin interactions give rise to dm - like interactions whose strength is proportional to @xmath557 .
the dm interactions can be viewed in particle - hole language as a consequence of the property that the interaction between an electron with pseudospin in the @xmath558 plane at @xmath82 and a hole with pseudospin in the @xmath559 direction at @xmath560 is not invariant under the interchange of @xmath82 and @xmath560 .
this property reflects the pseudospin dependence of dipole and other higher order multipoles in the cyclotron orbit cloud of an electron with a particular pseudospin .
the analysis presented in this section , can be applied to any quantum hall ferromagnet .
one elementary example is the case where the pseudospin orbitals share landau level index @xmath14 and have opposite spins . in this case
@xmath561 the pseudospin model has isotropic heisenberg interactions and the pseudospin - wave excitation energy expression @xmath562 , follows immediately from an expansion of the exchange energy function to quadratic order .
|
in a graphene bilayer with bernal stacking both @xmath0 and @xmath1 orbital landau levels have zero kinetic energy .
an electronic state in the @xmath2 landau level consequently has three quantum numbers in addition to its guiding center label : its spin , its valley index @xmath3 or @xmath4 , and an orbital quantum number @xmath5 the two - dimensional electron gas ( 2deg ) in the bilayer supports a wide variety of broken - symmetry states in which the pseudospins associated these three quantum numbers order in a manner that is dependent on both filling factor @xmath6 and the electric potential difference between the layers . in this paper , we study the case of @xmath7 in an external field strong enough to freeze electronic spins . we show that an electric potential difference between layers drives a series of transitions , starting from interlayer - coherent states ( ics ) at small potentials and leading to orbitally coherent states ( ocs ) that are polarized in a single layer .
orbital pseudospins carry electric dipoles with orientations that are ordered in the ocs and have dzyaloshinskii - moriya interactions that can lead to spiral instabilities .
we show that the microwave absorption spectra of icss , ocss , and the mixed states that occur at intermediate potentials are sharply distinct .
|
the observable spectra of the nuclei of galaxies depend on both the underlying emission sources and the subsequent obscuration and reprocessing of their light by material along the line of sight .
dust is common in these nuclear regions , so its spectral signature is often observed at mid - infrared ( mid - ir ) wavelengths . in particular , the silicate cross section peak near 10 produces a spectral feature that is observed frequently in absorption @xcite and sometimes in emission @xcite .
@xcite examine local active galactic nuclei ( agns ) and ultraluminous infrared galaxies ( ulirgs ) with the striking result that the average spectra of various classes of galaxies show marked differences in the 10 silicate feature ( their fig .
1 ) . we seek a physical explanation for the differences among these families .
moreover , the underlying origin of the distinct classes must also account for the most extreme example , the deep absorption of the ulirg iras 08572 + 3915 .
@xcite present a local sample of agns and ulirgs observed with the low - resolution modules of the infrared spectrograph @xcite aboard the _ spitzer space telescope _
they divide the sample into four classes : ulirgs based on ir luminosity , and quasars , seyfert 1 galaxies , and seyfert 2 galaxies based on optical criteria .
although some galaxies belong to more than one group , the average spectra of the groups are very different .
the silicate feature emerges clearly in emission in the quasar spectrum , while it appears in absorption in all other cases .
the absorption is deepest in the ulirg spectrum and shallowest in the seyfert 1 spectrum .
we illustrate the significant differences with example spectra from the four classes ( figure [ fig : iras08572 ] ) .
we quantify the silicate feature , defining the `` feature strength '' @xmath0 in terms of the the observed ( @xmath1 ) and underlying continuum ( @xmath2 ) flux evaluated at wavelength @xmath3 . to determine the continuum
well , we fit each spectrum over the rest wavelength intervals 5.07.0 , 14.014.5 , and 25.040.0 .
we perform a linear fit to each of these three segments , measuring @xmath4 vs. @xmath5 , then fit a spline to connect them .
we apply this approach directly to the continuum - dominated spectra , which include most of the quasars . in other cases , however , we modify the procedure to accomodate spectral features , such as high - ionization emission lines , polycyclic aromatic hydrocarbon ( pah ) emission , and ice absorption , excluding these regions from the continuum fits . for example , we shift the lower bound of the long - wavelength bandpass to 26.5 to avoid [ ] @xmath6 in many of the seyfert galaxies . when pah emission is strong , the short - wavelength bandpass is reduced to a continuum measurement at 5.5 , and in ice - dominated spectra , we use continuum measurements at 5.2 and 5.6 .
( see spoon et al .
2006b for detailed examples of the continuum fits . )
we measure the strength @xmath7 at the extremum of the continuum - subtracted spectrum .
positive values of @xmath7 indicate emission , while negative values of @xmath7 indicate absorption . in absorption
, this definition of feature strength corresponds to the apparent optical depth , _
i.e. _ , attenuation by a factor of @xmath8 , with @xmath9 conventionally evaluated at @xmath10 ( e.g. * ? ? ?
* ; * ? ? ?
the four classes show characteristic differences in strength , with @xmath11 , -0.18 , -0.54 , and -1.5 in the average spectra of the quasars , seyfert 1 galaxies , seyfert 2 galaxies , and ulirgs , respectively .
the most extreme example , iras 08572 + 3915 , has @xmath12 @xcite . while the ulirgs exhibit a range of @xmath7 , including a few cases of silicate emission where agns are identified optically , the ulirgs are generally deeply absorbed , with @xmath13 in 67% and @xmath14 in 30% of this archival sample @xcite
. moreover , all galaxies having @xmath15 are ulirgs .
a purely absorbing foreground screen can not produce a deep silicate feature because the dust emission of an optically thick screen is much greater than the transmitted external radiation .
denote by @xmath16 the planck function .
dust emission at wavelength @xmath3 from a region with optical depth @xmath17 is @xmath18 if the temperature @xmath19 is constant .
because the dust intensity is @xmath16 when the source is optically thick and @xmath20 when it is optically thin , uniform temperature dust will never produce an absorption feature under any circumstances .
a temperature gradient is essential for an absorption feature .
deep absorption requires dust geometry conducive to large gradients .
consider a cloud illuminated from outside by a radiative source . if the cloud s dimensions are much smaller than the distance to the source , then the heating flux is constant across the cloud s volume . without radiative transfer effects , the dust temperature is uniform throughout the cloud .
in contrast , arranging the same dust in a geometrically thick shell around the same heating source produces a large temperature gradient because of the spatial dilution of the flux with radial distance .
radiative transfer in optically thick dust affects both geometries similarly : the external radiation is absorbed within a short distance from the illuminated face .
this creates a large temperature differential close to the surface but only a modest gradient in the deeper layers .
we can therefore expect the absorption feature to have only limited depth in the case of externally illuminated clouds . _
a deep feature requires the radiation source to be deeply embedded in dust that is thick both optically and geometrically_. to verify this conclusion , we performed exact radiative transfer calculations with the code dusty @xcite .
the slab geometry serves as a proxy for an externally illuminated cloud , and a spherical shell represents deeply embedding dust .
we consider a range of total optical depths , described in terms of the optical depth at 5500 , @xmath21 .
the sed of the underlying source spectrum is a broken power law , characteristic of an agn .
in fact , as @xcite and @xcite demonstrate for similar problems , the resulting mid - ir spectral shape is insensitive to the heating spectrum .
the mid - ir spectrum is ultimately a product of the dust , and ir colors indicate its geometry @xcite .
the model dust is that of the galaxy , composed of 53% silicate and 47% graphite .
we use the absorption and scattering cross sections of @xcite and the grain size distribution of @xcite . changing the silicate fraction does not alter any of the general trends we report .
crystalline silicates , however , produce sharper features in the resulting sed , and we do not consider them further . in both the geometrically thin slab and spherical shell cases , the inner edge of the obscuring region is located where dust sublimates .
assuming @xmath22 k , this inner radius @xmath23 pc , where @xmath24 is the luminosity of the central source in units of @xmath25 . while our concern here is the nuclei of galaxies , the results are equally applicable to other situations , such as dust - embedded stars .
the total optical depth completely describes the slab . for the spherical shell
we consider variations of the geometric thickness and the density distribution .
figure [ fig : strength ] summarizes the model results , showing the 10 feature strength , @xmath7 , vs. @xmath21 for the slab and several spherical shell models . in the slab ,
the deepest absorption approaches @xmath26 at @xmath27 . with the adopted extinction curve , @xmath28 ,
so the actual optical depth @xmath29 while the apparent optical depth is only 1.1 . the depth of the feature does not provide an accurate measurement of the total optical depth along the line of sight , and the slab geometry can not produce the large observed apparent optical depths .
the spherical shell results depend on the total shell thickness and the density profile , as well as the total optical depth .
we considered shell thickness , @xmath30 , ranging from 1.5 to 1000 .
very thin shells are similar to the slab solutions , so we plot only representative thick shells , with outer radius @xmath31@xmath32 .
the radial density distribution is a power law , proportional to @xmath33 .
thick spherical shells _ can _ produce extremely deep silicate features , with the magnitude of the feature strength increasing with decreasing @xmath34 and increasing @xmath35 .
even with the spherical geometry , the apparent optical depth systematically underpredicts the actual line - of - sight optical depth .
several combinations of optical depth and shell geometry can produce the large strength we observe in iras 08572 + 3915 .
a flux ratio helps to distinguish among them .
we compare @xmath36 at 14 and 30 , where the continuum dominates the observed flux . as figures [ fig : strength ] and [ fig : specind ] together illustrate , models with @xmath37 , moderate thickness ( @xmath38 a few hundred @xmath39 ) , and moderately large optical depth ( @xmath40 a few hundred ) agree with both the observed mid - ir color and the large magnitude @xmath7 of iras 08572 + 3915 .
none of the @xmath41 or 2 solutions fits both the feature depth and the 14/30 flux ratio .
however , the @xmath41 solutions at lower optical depth ( @xmath42 ) are consistent with the average values of both @xmath7 and flux ratio measured in the ulirgs .
the observed flux densities separately indicate the intrinsic luminosity of the buried source . comparing the observations of iras 08572 + 3915 with the simulated sed of the preferred solution , in which @xmath43 , @xmath44 , and @xmath45
, the 14 and 30 measurements yield @xmath46 and 1.1 , respectively .
both values are comparable to the luminosity derived from the iras fluxes , @xmath47 , at a distance of 250 mpc .
we also calculate the mass of the obscuring shell , adopting a standard gas - to - dust ratio @xmath48 .
we find @xmath49 , where @xmath50 depends on the radial profile . in terms of shell
thickness , @xmath51 , @xmath52 for @xmath53 , @xmath54 for @xmath44 , and @xmath55 for @xmath56 .
thus , in iras 08572 + 3915 , @xmath57 .
the strength of the absorption feature is a function of the temperature contrast in the dusty material .
the extremely deep features we observe in many ulirgs require a steep temperature gradient . we can broadly account for the differences among the spherical models in terms of the temperature distribution of the dust . in the limit of a thin shell ( @xmath58 )
, the result approaches the slab solution .
the geometric dilution of the heating is also reduced when the material is very concentrated ( i.e. , @xmath34 is large ) . in general ,
the apparent optical depth increases with true optical depth . for a given radial density profile
, the larger spheres typically exhibit deeper features because the temperature range over them is large . the exceptions to these general trends are a consequence of the details of radiative transfer through the material , and the results are sensitive to both temperature variation and optical depth .
for example , any shell will become heated well throughout at sufficiently large optical depth , reducing the temperature gradient and feature strength .
the onset of this effect depends on shell thickness and @xmath34 ; it is apparent in the @xmath30 = 100 solutions for @xmath41 , and to a lesser extent for @xmath37 .
we find the deepest absorption in the @xmath41 or 1 models with moderate size .
we checked the effect of the chemical composition of the dust , varying the silicate fraction from 20 to 80% .
the general trends of @xmath7 with geometry and optical depth remain the same in these additional simulations . _
increasing the silicate abundance does not produce a deeper feature _ ; only the scaling of strength relative to total optical depth changes with chemical composition because the ratio of @xmath59 to total optical depth changes .
the slab temperature profile is similar to that of any externally heated cloud whose dimensions are much smaller than its distance to the source .
the dark and bright faces of the slab are distinct : the silicate feature appears in absorption only from the dark side , while viewing the bright side always results in an emission feature , independent of the cloud s total optical depth .
the spectrum of a clumpy medium mixes the contributions of many individual clouds , each of which presents to the observer either its dark or bright side @xcite .
only a contrived cloud configuration can produce a very deep absorption feature . in a single slab ,
the feature is deepest when the temperature on the illuminated face is around 200400 k and the slab is viewed from its dark side along the line of sight to the source . unlike dust sublimation
, there is no reason to single out the distance corresponding to this particular temperature for the location of all clouds .
similarly , there is no reason for the observer to be in a preferred direction , one that views the dark sides exclusively .
in contrast , only the dark side of a closed shell can be observed .
however , once the shell is fragmented into clouds , it offers some direct views of portions of its illuminated inner surface .
the emission from the bright faces fills the absorption trough , reducing the depth of the feature .
the true distribution of obscuring material is most likely inhomogeneous , consisting of both dense clouds and a smooth intercloud medium , but the relative importance of these two components differs between the ulirgs and other galaxies .
the lack of deep absorption implies that the cloud - dominated medium is most relevant to the agns .
indeed , at high spatial resolution , where the agn is isolated , the silicate absorption is never deep @xcite , and complete models of obscuration by clouds can reproduce the feature in detail @xcite .
individual seyfert galaxies observed with _
spitzer _ may show deep absorption when the larger ( up to 20-kpc scale ) galactic contribution becomes comparable to that of the agn .
nonetheless , the absence of extreme absorption in the composite spectra of both seyfert 1 and seyfert 2 galaxies @xcite indicates that clouds dominate the obscuration of their agns . the deep silicate absorption that is characteristic of many ulirgs requires a central source embedded in obscuring material that is thick both optically and geometrically .
@xcite noted similarly that only centrally - concentrated buried sources can produce deep 3 dust absorption features .
while the medium may be inhomogeneous , the smooth component of the distribution must dominate the 10 spectral region .
otherwise , emission from bright cloud faces would fill in the absorption trough , reducing the magnitude of @xmath7 .
given the scaling of the shell size with luminosity , the observed deep absorption features therefore indicate obscuration on scales of a few 100 pc .
agns may be present in these galaxies , but they must be deeply buried .
the distinction of these ulirgs is the continuous geometry of the covering medium , not the optical depth along the line of sight or the average covering fraction .
indeed , many seyfert 2 galaxies have large column densities along the lines of sight to their nuclei , but they offer clear views of their narrow line regions and some illuminated cloud faces . the geometry of the obscuration determines the emergent sed across the entire ir bandpass .
our simulations demonstrate that the far - ir spectral shapes change significantly with the geometric variations , despite having the same heating source . in general
, the geometries that produce the deepest silicate absorption also have the reddest far - ir spectra , yet even modest optical depths ( @xmath60 ) can have the peak of @xmath61 at wavelengths greater than 60 .
clumpy and smooth distributions produce different far - ir signatures @xcite . in a normal agn , a clumpy medium ( the torus )
obscures the central engine , and the resulting far - ir flux ratios successfully separate agns from star - forming galaxies @xcite .
however , the nature of the heating source is erased by dust reprocessing .
the only qualification is that the source provide uv photons to heat the dust , which both starburst and agn spectra do . when an agn is embedded in a thick , smooth , dusty blanket , its peak flux density shifts to longer wavelengths .
far - ir flux ratios can not be used to identify the powerful sources of radiation in the deeply - embedded ulirgs . because of the extreme obscuration , the emergent near - ir emission of the deeply embedded nuclear sources is weak .
the observed near - ir emission of ulirgs is often stronger than the deeply embedded models predict . even in iras 08572 + 3915 , where the sed declines steeply below 8 ( fig .
1 ) , the near - ir emission exceeds the model results . here
the observed 3.7/14 flux ratio is 0.05 , but the model embedded source does not account for more than 5% of the observed flux at @xmath62 . in these cases ,
the component responsible for the observed near - ir emission must be different from the one dominating at mid - ir wavelengths .
the additional near - ir sources do not fill in the deep silicate absorption troughs , so their spectra must decline steeply above 5
. the photospheric emission of stars may therefore be the origin of the observed near - ir light . because different sources dominate the observed flux in different spectral regions , a near- to mid - ir flux ratio is not generally a measurement of a single physical entity . in summary ,
we have observed characteristic differences in the mid - ir silicate feature among four classes of galaxies .
the underlying physical distinctions of optical depth and obscuring geometry account for these differences .
optically thin regions produce silicate emission , which we observe in the composite quasar spectrum .
modest silicate absorption shows that clumpy obscuration dominates in the seyfert 1 and seyfert 2 galaxies . the extremely deep absorption in many ulirgs , especially that of iras 08572 + 3915 , requires their nuclear sources to be deeply embedded in optically and geometrically thick material .
we thank d. weedman for contributions to this work and an anonymous referee for useful suggestions .
nal acknowledges work supported by the nsf under grant no . 0237291 .
me acknowledges support from nsf ast-0507421 and nasa nng05gc38 g .
carico , d. p. , sanders , d. b. , soifer , b. t. , elias , j. h. , matthews , k. , & neugebauer , g. 1988 , , 95 , 356 de grijp , m. h. k. , lub , j. , & miley , g. k. 1987 , , 70 , 95 draine , b. t. 2003 , , 598 , 1017 glass , i. s. , & moorwood , a. f. m. 1985 , , 214 , 429 granato , g. l. , & danese , l. 1994 , , 268 , 235 hao , l. , et al .
2005 , , 625 , l75 hao , l. , spoon , h. w. w. , weedman , , d. w. , houck , j. r. , & armus , l. 2006 , submitted to houck , j. r. , et al .
2004 , , 154 , 18 imanishi , m. , & maloney , p. r. 2003 , , 588 , 165 ivezi , .
, & elitzur , m. 1997 , , 287 , 799 ivezi , .
, & elitzur , m. 2000 , , 534 , l93 ivezi , . , nenkova , m. & elitzur , m. , 1999 , user manual for dusty , university of kentucky internal report ( accessible at http://www.pa.uky.edu/@xmath63moshe/dusty ) mason , r. e. , geballe , t. r. , packham , c. , levenson , n. a. , elitzur , m. , fisher , r. s. , & perlman , e. 2006 , , 640 , 612 mathis , j. s. , rumpl , w. , & nordsieck , k. h. 1977 , , 217 , 425 nenkova , m. , ivezi ' c , .
, & elitzur , m. 2002 , , 570 , l9 neugebauer , g. , green , r. f. , matthews , k. , schmidt , m. , soifer , b. t. , & bennett , j. 1987 , , 63 , 615 pier , e. a. , & krolik , j. h. 1993 , , 418 , 673 rieke , g. h. 1978 , , 226 , 550 roche , p. f. , aitken , d. k. , & smith , c. h. 1991 , , 252 , 282 roche , p. f. , packham , c. , telesco , c. m. , radomski , j. t. , alonso - hererro , a. , aitken
< , d. k. , colina , l. , & perlman , e. 2006 , , 367 , 1689 rowan - robinson , m. 1980 , , 44 , 403 shi , y. , et al .
2006 , , in press ( astro - ph/0608645 ) siebenmorgen , r. , haas , m. , krgel , e. , & schulz , b. 2005 , , 436 , l5 spoon , h. w. w. , et al . 2006a , , 638 , 759 spoon , h. w. w. , keane , j. v. , tielens , a. g. g. m. , lutz , d. , moorwood , a. f. m. , & laurent , o. 2002 , , 385 , 1022 spoon , h. w. w. , marshall , j. a. , houck ,
j. r. , elitzur , m. , hao , l. , armus , l. , brandl , b. r. , & charmandaris , v. 2006b , submitted to spoon , h. w. w. , moorwood , a. f. m. , lutz , d. , tielens , a. g. g. m. , siebenmorgen , r. , & keane , j. v. 2004 , , 414 , 873 sturm , e. , et al .
2005 , , 629 , l21 weedman , d. w. , et al .
2005 , , 633 , 706 werner , m. w. , et al .
2004 , , 154 , 1 _ spitzer _ spectra of the four galaxy classes ( _ solid lines _ ) and their fitted continua ( _ dotted lines _ ) .
the quasar , seyfert 1 , and seyfert 2 examples are typical of their classes .
these plotted spectra are scaled by factors of @xmath64 , 300 , and 30 , respectively .
the spectrum of iras 08572 + 3915 is not scaled , and this galaxy exhibits the most extreme absorption , with @xmath12 .
these mid - ir spectra have been published by shi et al .
( 2006 ; pg 1211 + 143 and ngc 5643 ) and spoon et al .
( 2006a ; iras 08572 + 3915 ) .
near - ir data ( _ diamonds _ ) are from @xcite , @xcite , @xcite , and @xcite . , height=672 ] strength of the 10 silicate feature ( eq .
1 ) vs. optical depth ( measured at 5500 ) for a slab and several spherical shell models .
the corresponding optical depth at 9.5 is indicated along the upper axis .
emission results in positive values of @xmath7 ; absorption results in negative values .
the apparent silicate optical depth , @xmath65 , when the feature appears in absorption is indicated along the right axis .
the apparent optical depth is never large for the slab ( _ thick green line _ ) .
thin spherical shells are similar to the slab result .
several thick spherical shell models are plotted ( _ thin lines _ ) , with the radial density profile ( @xmath66 ) and the shell thickness ( @xmath30 ) indicated .
the thick gray line shows @xmath7 for iras 08572 + 3915 .
very large absolute values of @xmath7 , which correspond to large apparent optical depth , are possible only with the shell geometry . ]
flux ratio between 14 and 30 vs. optical depth .
colors and line styles as in fig .
[ fig : strength ] .
the combination of @xmath7 and this mid - ir color limits the acceptable models .
we observe @xmath67 in iras 08572 + 3915 , which rules out all @xmath68 ( _ blue _ ) spectra .
given the large strength , the preferred solutions have @xmath69 ( _ red _ ) , with @xmath70 a few hundred , and @xmath71 a few hundred . ]
|
the silicate cross section peak near 10 produces emission and absorption features in the spectra of dusty galactic nuclei observed with the _ spitzer space telescope_. especially in ultraluminous infrared galaxies , the observed absorption feature can be extremely deep , as iras 08572 + 3915 illustrates .
a foreground screen of obscuration can not reproduce this observed feature , even at large optical depth .
instead , the deep absorption requires a nuclear source to be deeply embedded in a smooth distribution of material that is both geometrically and optically thick .
in contrast , a clumpy medium can produce only shallow absorption or emission , which are characteristic of optically - identified active galactic nuclei . in general ,
the geometry of the dusty region and the total optical depth , rather than the grain composition or heating spectrum , determine the silicate feature s observable properties .
the apparent optical depth calculated from the ratio of line to continuum emission generally fails to accurately measure the true optical depth .
the obscuring geometry , not the nature of the embedded source , also determines the far - ir spectral shape .
|
there has been recent progress in understanding physics of strongly correlated electronic systems and their electronic structure near a localization delocalization transition through the development of dynamical mean
field theory ( dmft ) @xcite . merging this computationally tractable many body technique with realistic local density
approximation ( lda ) @xcite based electronic structure calculations of strongly correlated solids is promising due to its simplicity and correctness in both band and atomic limits . at present , much effort is being made in this direction including the developments of a lda+dmft method anisimovkotliar , lda++ approach @xcite , combined gw and dmft theory @xcite as well as applications to various systems such as la@xmath0sr@xmath1tio@xmath2 @xcite , v@xmath3o@xmath2 @xcite , fe and ni licht , ce @xcite , pu @xcite , and many others . for a review , see ref . .
the development of _ ab initio _ dmft scheme requires fast methods and algorithms to solve the anderson impurity model @xcite in general multiorbital case .
present techniques based on either non crossing approximation ( nca ) or iterative perturbation theory ( ipt ) are unable to provide the solution to that problem due to a limited number of regimes where these methods can be applied @xcite .
the quantum monte carlo ( qmc ) technique @xcite is very accurate and can cope with multiorbital situation but not with multiplet interactions .
also its applicability so far has been limited either to a small number of orbitals or to unphysically large temperatures due to its computational cost .
recently some progress has been achieved using impurity solvers that improve upon the nca approximation @xcite , but it has not been possible to retrieve fermi liquid behavior at very low temperatures with these methods in the orbitally degenerate case . in this paper
we explore the possibility to use interpolation for the self energy of the quantum impurity model in general multiorbital situation .
we do not attempt to develop an alternative method for solving the impurity problem , but follow the ideology of lda theory where approximations were designed by analytical fits @xcite to the quantum monte carlo simulations for homogeneous electron gas @xcite .
numerically very expensive qmc calculations for the impurity model display smooth self
energies at imaginary frequencies for a wide range of interactions and dopings , and it is therefore tempting to design such an interpolation . we also keep in mind that for many applications a high precision in reproducing the self energies may not be required .
one of such applications is , for example , the calculation of the total energy ce , nature , science which , as well known from lda based experience , may not be so sensitive to the details of the one electron spectra . as a result , we expect that even crude evaluations of the self
energy shapes on imaginary frequency axis may be sufficient for solving many realistic total energy problems .
another point is a computational efficiency . bringing full self consistent loops with respect to charge densities @xcite and
other spectral functions require many iterations towards the convergency which may not need too accurate frequency resolutions at every step .
this view should , of course , be contrasted with calculation of properties such as the low energy spectroscopy and especially transport where delicate distribution of spectral weight at low energy , namely the imaginary part of the analytically continued self energy , needs to be computed with much greater precision . here , extensions of the interpolative algorithms should be implemented and its beyond the scope of the present work .
we can achieve a fast interpolative algortihm for the self energy utilizing a rational representation .
the coefficients in this interpolation can be found by forcing the self energy to obey several limits and constrains .
for example , if infinite frequency ( hartree fock ) limit , high frequency moments of the self energy , low frequency mass renormalization , number of particles as well as the value of the self energy at zero frequency are known from independent calculation , the set of interpolating coefficients is well defined . in this work ,
we explore the slave boson gutzwiller approach @xcite and the hubbard i approximation hubbard1 to determine these coefficients . comparing interpolating results with more accurate calculations using the quantum monte carlo method
we find that the slave boson approach predicts the parameters of interpolation with a good accuracy while the hubbard i method fails in a number of regimes .
thus , interpolation for the self energy at all frequencies can be obtained from the method designed to work exclusively at low
frequency limit .
our paper is organized as follows .
in section ii we discuss rational interpolation for the self energy and list the constraints . in section iii
we discuss methods for solving anderson impurity model based on the slave boson gutzwiller and the hubbard i approximations which can be used to find these constraints .
we also present our numerical comparisons with the qmc data between various quantities such as the number of particles and the quasiparticle residue .
in section iv we discuss the interpolational approach and compare the self energies and the green functions with the qmc results .
section v is the conclusion .
to be specific , we concentrate on the anderson impurity hamiltonian @xmath4 , \label{ham}\end{aligned}\]]describing the interaction of the impurity levels @xmath5 with bands of conduction electrons @xmath6 via hybridization @xmath7 .
@xmath8 is the coulomb repulsion between different orbitals in the @xmath9band .
inspired by the success of the iterative perturbation theory @xcite , in order to solve the anderson impurity model in general multiorbital case , we use a rational interpolative formula for the self energy .
this can be either encoded into a simple form of pole expansion@xmath10or , alternatively , in a form of continuous fraction expansion@xmath11the coefficients @xmath12 , @xmath13 , in eq .
( [ rat ] ) or the coefficients @xmath14 @xmath15 , ... in eq .
( cfr ) are to be determined together with the hartree fock value of the self energy @xmath16 . in principle , both representations are equivalent and there is a well defined non
linear relationship between the parameters entering ( [ rat ] ) and ( [ cfr ] ) .
our basic assumption is that only a few poles in the rational representation ( [ rat ] ) or a few coefficients in continuous fraction expansion ( [ cfr ] ) is necessary to reproduce an overall frequency dependence of the self
energy necessary for the accuracies of the total energy calculations .
extensive experience gained from solving hubbard and periodic anderson model within dmft at various ratios of the on site coulomb interaction @xmath17 to the bandwidth @xmath18 shows the appearance of lower and upper hubbard bands as well as renormalized quasiparticle peak in the spectrum of one electron excitations @xcite . to describe such three peak structure
we expect that either two or at most three pole formulae for the self energy will suffice . thus , four ( six ) unknown parameters need to be determined when using two ( three ) pole interpolating formulae . in order to fix the coefficients we can explore several constrains for the self energy : _ a ) _ _ hartree fock value .
_ in the limit @xmath19 the self energy takes its hartree fock form @xmath20 _ b ) _ _ moments . _ the subleading high frequency behavior of the self energy can be used to determine some additional coefficients . in the case of the three pole approximation , we determine coefficients @xmath21,@xmath22 and @xmath23 from the first three moments of the self
energy @xmath24 @xmath25 as follows @xmath26 @xmath27 @xmath28 while in the two pole approximation only coefficient @xmath21 should be determined from the high frequency moment , as stated in eq .
( [ aa1 ] ) above .
energy moments can be expressed by the occupancies and correlation functions that can be calculated within an approximation like slave boson gutzwiller or hubard i. the expressions for the self
energy moments are worked out explicitely in the appendix a. _ c ) mass renormalization . _
if independent determination of the quasiparticle residue , @xmath29 , exists , the following relationship holds @xmath30 _ d ) number of particles .
_ sum over all matsubara frequencies of the green s function gives number of particles , i.e.@xmath31where @xmath32defines the impurity green function and @xmath33 is the hybridization function . _ e ) _ _ friedel sum rule .
_ this is a relation between the total density and the real part of the self energy at zero frequency .
it determines zero frequency value of the self
energy @xmath34 formally , some of the constrains [ ( _ c _ ) and ( _ e _ ) ] hold for zero temperature only but we expect no significant deviations as long as we stay at low temperatures .
we thus see that the interpolational scheme is defined completely once a prescription for obtaining parameters such as @xmath35 @xmath36 and @xmath37 is given . for this purpose
we will test two popular methods : slave boson gutzwiller method @xcite as described by kotliar and ruckenstein @xcite and the well known hubbard i approximation @xcite .
we compare these results against more accurate but computationally demanding quantum monte carlo @xcite calculations .
notice that once the parameters such as @xmath29 are computed from a given approximate method , some of the quantities such as the total number of particles , @xmath36 , and the value of the self energy at zero frequency , @xmath38 can be computed fully self consistently .
they can be compared with their non self consistent values .
if the approximate scheme already provides a good approximation for @xmath36 and satisfies the friedel sum rule , the self consistency can be avoided hence accelerating the calculation .
indeed we found that inclusion of the self
consistency improves the gutzwiller results in the vicinity of the mott transition only where the hubbard bands get increasingly important .
we now give the description of our approximate algorithms and then present the comparison with the qmc calculations .
the quantum monte carlo method is a powerful and manifestly not perturbative approach in either interaction @xmath17 or the bandwidth @xmath18 . in the qmc method one
introduces a hubbard stratonovich field and averages over it using the monte carlo sampling .
this is a controlled approximation using different expansion parameter , the size of the mesh for the imaginary time discretization .
unfortunately it is computationally very expensive as the number of time slices and the number of hubbard stratonovich fields increases .
extensive description of this method can be found in ref . .
we will use this method to benchmark our calculations with approximate but much faster algorithms described below . a fast approach
to solve a general impurity problem is the slave boson method @xcite . at the mean field level
, it gives the results similar to the famous gutzwiller approximation @xcite .
however , it is improvable by performing fluctuations around the saddle point .
this approach is accurate as it has been shown recently to give the exact critical value of @xmath17 in the large degeneracy limit at half filling florens .
the main idea is to rewrite atomic states consisting of @xmath39 electrons @xmath40 , @xmath41 with help of a set of slave
bosons @xmath42 . in the following ,
we assume @xmath43 symmetric case , i.e. , equivalence between different states @xmath44 for fixed @xmath39 .
formulae corresponding to a more general crystal field case are given in appendix b. the creation operator of a physical electron is expressed via slave particles in the standard manner @xcite . in order to recover the correct non interacting limit at the mean field level , the bose fields @xmath45 can be considered as classical values found from minimizing a lagrangian @xmath46 corresponding to the hamiltonian ( [ ham ] ) .
two lagrange multipliers @xmath47 and @xmath48 should be introduced in this way , which correspond to the following two constrains : @xmath49 the physical meaning of the first constrain is that the sum of probabilities to find atom in any state is equal to one , and the second constrain gives the mean number of electrons coinciding with that found from @xmath50 . a combinatorial factor @xmath51 arrives due to assumed equivalence of all states with @xmath39 electrons .
minimization of @xmath46 with respect to @xmath52 leads us to the following set of equations to determine the quantities @xmath52 : @xmath53\psi _ { n}+nbt\sum_{i\omega } \delta ( i\omega ) g_{g}(i\omega ) [ lr\psi _ { n-1}+\psi _ { n}bl^{2}]+\]]@xmath54=0 , \label{guz}\]]where @xmath55 , determines the mass renormalization , and the coefficients @xmath56 , @xmath57 are normalization constants as in refs . .
@xmath58 is the total energy of the atom with @xmath39 electrons within @xmath43 approximation .
( [ guz ] ) , along with the constrains ( [ one ] ) , ( [ two ] ) constitute a set of non linear equations which have to be solved iteratively . in practice , we consider eq .
( [ guz ] ) as an eigenvalue problem with @xmath48 being the eigenvalue and @xmath52 being the eigenvectors of the matrix .
the physical root corresponds to the lowest eigenvalue of @xmath48 which gives a set of @xmath52 determining the mass renormalization @xmath59 since the matrix to be diagonalized depends non
linearly on @xmath52 via the parameters @xmath60 and @xmath61 and also on @xmath47 , the solution of the whole problem assumes the self consistency : ( i ) we build an initial approximation to @xmath52 ( for example the hartree
fock solution ) and fix some @xmath47 , ( ii ) we solve eigenvalue problem and find new normalized @xmath52 , ( iii ) we mix new @xmath52 with the old ones using the broyden method @xcite and build new @xmath62 and @xmath61 . steps ( ii ) and ( iii ) are repeated until the self consistency with respect to @xmath52 is reached . during the iterations we also vary @xmath47 to obey the constrains .
the described procedure provides a stable computational algorithm for solving aim and gives us an access to the low frequency green s function and the self energy of the problem via knowledge of the slope of @xmath63 and the value @xmath64 at zero frequency .
the described slave boson method gives the following expression for the self energy : @xmath65the impurity green function @xmath66 in this limit is given by the expression @xmath67 as an illustration , we now give the solution of eq .
( [ guz ] ) for non degenerate case ( @xmath68 and at the particle hole symmetry point with @xmath69 .
consider a dynamical mean
field theory for the hubbard model which reduces the problem to solving the impurity model subject to the self consistency condition with respect to @xmath70 .
starting with the semicircular density of states ( dos ) , the self consistency condition is given by @xmath71 with parameter @xmath72 being the quarter of bandwidth .
we obtain the following simplifications : @xmath73 , @xmath74 , @xmath75 and @xmath76^{-1}.$ ] the sum @xmath77 appeared in eq .
( [ guz ] ) scales as @xmath78 with the constant @xmath79 being the characteristic of a particular density of states and approximately equal to 0.2 in the semicircular dos case .
the self consistent solution of eq .
( [ guz ] ) is therefore possible and simply gives @xmath80 .
the mott transition occurs when no sites with double occupancies can be found , i.e. when @xmath81 a critical value of @xmath82 with @xmath18 being the bandwidth . for @xmath83 ,
this gives @xmath84 and reproduces the exact result for @xmath85 within a few percent accuracy .
as degeneracy increases , critical @xmath17 is shifted towards higher values . from numerical calculations we obtained the following values of the critical interactions in the half
filled case @xmath86 for @xmath87 ( @xmath88level)@xmath89 @xmath90 for @xmath91 ( @xmath92level)@xmath89 and @xmath93 for @xmath94 ( @xmath9level ) . density density correlation function @xmath95 ( where @xmath96 ) for local states with @xmath39 electrons is proportional to the number of pairs formed by @xmath39 particles @xmath97 . since the probability for @xmath39 electrons to be occupied is given by : @xmath98 , the physical density
density correlator can be deduced from : @xmath99 .
similarly , the triple occupancy can be calculated from : @xmath100 .
let us now check the accuracy of this method by comparing it s results with the qmc data .
we consider the two band hubbard model within @xmath101 orbitally degenerate case .
hybridization @xmath102 satisfies the dmftself consistency condition of the hubbard model on a bethe lattice@xmath103where @xmath72 is a quarter of the bandwidth @xmath18 .
the coulomb interaction is chosen to be @xmath104 ( @xmath92 is the half bandwidth ) which is sufficiently large to open the mott gap at integer fillings .
all calculations are done for the temperature @xmath105 .
we first compare the average number of electrons vs. chemical potential determined from the slave bosons which is plotted in fig .
[ fig1](a ) .
this quantity is sensitive to the low
frequency part of the green function which should be described well by the present method .
we see that it reproduces the qmc data with a very high accuracy . on concentration , and ( c ) density density correlation function , @xmath106 versus filling , @xmath107 , in the two band hubbard model.,title="fig:",scaledwidth=50.0% ] + the quasiparticle residue @xmath108 versus filling @xmath107
is plotted in fig .
[ fig1](b ) .
the slave boson method gives the fermi liquid and provides a good estimate for the quasiparticle residue .
indeed , the agreement with the qmc results is satisfactory .
also , recent findings @xcite suggest that at half filling @xmath109 deduced from slave bosons becomes exact when @xmath110 .
still the discrepancy of the order of 30% can be found .
in fact , we have performed several additional calculations for other degeneracies ( @xmath111 and @xmath112 ) and for various parameters regimes .
the trend to underestimate mass renormalization by about 30% can be seen in almost all cases ( it disappears only when @xmath17 approaches zero ) .
this empirical finding is useful as we can use the reduced values of @xmath109 needed for our interpolating formula .
[ fig1](c ) shows the density
density correlation function @xmath113 as a function of average occupation @xmath107 .
the discrepancy is most pronounced for fillings @xmath114 [ see the inset of fig .
[ fig1](c ) ] where the absolute values of @xmath115 are rather small .
although our slave boson technique captures only the quasiparticle peak , it gives the correlation function in reasonable agreement with the qmc for dopings not too close to the mott transition .
now we turn to the hubbard i approximation @xcite which is closely related to the moments expansion method @xcite .
one introduces hubbard operators and the corresponding green function @xmath116 defined for these hubbard operators . within @xmath43
we are summing over certain set of hubbard operators which induce transitions between @xmath39times and @xmath117times occupied local level .
the impurity green function is @xmath118where the matrix @xmath119^{-1}=[g_{nm}^{at}(i\omega ) ] ^{-1}-\delta ( i\omega ) , n , m=0,n-1 $ ] .
@xmath120 is the atomic green function @xmath121the coefficients @xmath122 are the probabilities to find atom with @xmath39 electrons similar to the coefficients @xmath123 introduced above .
they can be also used to find the averages @xmath124 in the same way .
these numbers are normalized to unity , @xmath125 and are expressed via diagonal elements of @xmath116 as follows :
the mean number of electrons can be measured as follows : @xmath127 or as : @xmath128 note that when @xmath129 , @xmath66 is reduced to @xmath130 i.e. hubbard i reproduces the atomic limit .
setting @xmath131 gives @xmath132^{-1}$ ] , which is the correct band limit .
unfortunately , at half filling this limit has a pathology connected to the instability towards mott transition at any interaction strength @xmath17 .
to see this , we consider a dynamical mean
field theory for the hubbard model .
using semicircular density of states , we obtain @xmath133^{-1}g_{at}(i\omega ) $ ] and conclude that for any small @xmath17 @xmath134the system opens a pathological gap in the spectrum .
clearly , using hubbard i only , the behavior of the green function at @xmath135 can not be reproduced .
this emphasizes the importance of using the slave boson treatment at small frequencies .
let us now check the accuracy of this method against qmc .
we again consider the two band hubbard model within @xmath101 symmetry as in the case of the slave boson method described earlier .
the @xmath136 is plotted in fig .
[ fig2](a ) .
the hubbard i approximation does not give satisfactory agreement with the qmc data because it misses the correct behavior at low frequencies . on concentration , and ( c ) density density correlation function , @xmath137 versus filling , @xmath107 , in the two band hubbard model.,title="fig:",scaledwidth=50.0% ] + the comparisons for @xmath138 is plotted in fig . [ fig2](b ) .
these numbers are extracted from the low
frequency slope of the imaginary part of the self
energy on the imaginary axis .
surprizingly , the hubbard i @xmath139s show a relatively good behavior when comparing to the qmc method .
however , the patology of this approximation at half filling would predict @xmath140 for any @xmath17 , which is a serious warning not to use it for extracting the quasiparticle weight .
[ fig2](c ) shows @xmath141 as a function of average occupation @xmath107 . as this quantity
is directly related to the high frequency expansion one may expect a better accuracy here . however , comparing fig .
[ fig2](c ) and fig .
[ fig1](c ) , it is clear that the slave boson method gives more accurate double occupancy .
this is due to the fact that the density matrix obtained by the slave boson method is of higher quality than the one obtained from the hubbard i approximation .
the results of these calculations suggest that the slave boson gutzwiller method shows a good accuracy in comparison with the qmc .
therefore it can be used to determine the unknown coefficients in the interpolational form of the self energy , eqs .
( [ rat ] ) or ( [ cfr ] ) . at the same time
the use of the hubbard approximation should be avoided .
interestingly , while more sophisticated qmc approach captures both the quasiparticle peak and the hubbard bands this is not the case for the slave boson mean field method . to obtain
the hubbard bands in this method fluctuations need to be computed , which would be very tedious in a general multiorbital situation .
however the slave boson method delivers many parameters in a good agreement with the qmc results , and hence it can be used to obtain the parameters that enter the rational approximation . in this way slave
boson approach generates a better green function with quasiparticles and hubbard bands _ at no extra computational cost_.
we now turn to the comparison of the self
energie obtained using the formula ( [ cfr ] ) and the corresponding green functions after the formula ( [ imp ] ) against the predictions of the quantum monte carlo method .
we have studied the results of the two pole and the three pole interpolations which require the determination of either 4 or 6 coefficients .
the interpolational formula with two poles needs finding the values of @xmath142 , @xmath109 and the first high frequency moment of the self
energy @xmath143 the condition that the total number of electrons found from the green function ( [ imp ] ) is the same as predicted by the slave boson method can be used to fix the fourth coefficient .
to explore the three
pole interpolation , two more conditions are required .
we have used additional high frequency moments @xmath144 and @xmath145 which can be determined if one computes correlational functions @xmath124 , @xmath146 , @xmath147 .
the precise functional form of these moments is given in appendix a. .,scaledwidth=32.0% ] we perform the comparison of the interpolational scheme against qmc for two characteristic values of the concentrations : @xmath148 representing the case @xmath114 and @xmath149 representing the case @xmath150 .
both cases are chosen to be close to the mott transition .
this is a hard test for our approach as we have found that the predictions of the interpolative scheme and the qmc data are practically indistinguishable in the metallic region far from the mott transition ( @xmath151 ) . the two band hubbard model within @xmath101 symmetry , @xmath104 , and @xmath152 self consistency condition is used .
.,title="fig:",scaledwidth=32.0% ] + fig .
[ fig3 ] shows the comparison of the real and imaginary parts of the self energy and the green function obtained by the two and three pole interpolations with the qmc results .
both interpolational schemes give pretty good agreement at all frequencies .
these results are obtained with the values of mass renormalizations @xmath109 reduced by 30% which gives a better agreement with @xmath153 as we discussed in section iii . the results with bare
@xmath154 deduced from the slave boson method are also found to be quite accurate but not as good as we advertise in fig .
we have also performed the comparison for the one electron density of states which are shown in fig .
one can see the appearance of the lower and upper hubbard bands as well as renormalized quasiparticle peak . note a complete match in the positions of the hubbard band as well as correct bandwidth for the quasiparticles as compared to the quantum monte carlo calculations . to obtain such remarkable agreement , the use of renormalized values of @xmath109 is essential .
.,title="fig:",scaledwidth=32.0% ] + in fig .
[ fig5 ] , we show our results for the case @xmath149 using the two and three pole interpolational formula ( [ cfr ] ) . as far as the two pole scheme is concerned , the overall agreement is satisfactory but a discrepancy exists in the region of intermediate imaginary frequencies .
it is directly related to the misplaced upper hubard band as can be seen in fig .
this discrepancy disappears in the three
pole interpolation as is evident from the fig .
comparison of the one electron density of states is shown in fig .
[ fig6 ] . clearly , the two pole interpolation is inadequate in this regime , while the three pole interpolating scheme gives correct position of hubbard bands and quasiparticle peak , all basic features of the anderson impurity problem .
finally , we would like to comment on the renormalization of the quasiparticle residue @xmath109 beyond the slave boson mean field result .
while we found empirically that the 30% renormalization significantly improves our fits to the qmc data , over a broad region of parameters , this can not hold everywhere since @xmath109 should go to @xmath155 when @xmath17 approaches zero .
the reduction of @xmath109 beyond the mean field results , in the regions where the correlations are strong , can be thought to arise from fluctuation corrections of the slave boson around the saddle point value . a computationally efficient way to evaluate these corrections , which will also give rise to the quasiparticle lifetime not included in this paper , are important open problems left for future work .
.,title="fig:",scaledwidth=32.0% ] +
to summarize , this paper shows the possibility to extract parameters for the interpolating self energy of the anderson impurity model such as the self energy at zero frequency , the mass renormalization , the density and the high frequency moments using computationally very efficient slave boson mean field approach .
these data are then used to reconstruct the self energy at all frequencies using a rational interpolation formula .
we find that three pole interpolation predicts the self energy quite accurately as compared to the results of more sophisticated approaches such the quantum monte carlo method . as a general conclusion
, we have thus found that interpolative formulae with several poles are sufficient to fit the self energies in a broad range of the parameters of the quantum impurity problem and the slave boson mean field method is capable to deliver all necessary ingredients for this algorithm .
the approach can be used for the evaluation of lda+dmft total energies and is a very direct way to obtain spectra with a three peak structure , characteristic of the strongly correlated regime , with very little computational effort .
the work was supported by nsf dmr grants 0096462 , 02382188 , 0312478 , 0342290 and us doe grant no de - fg0299er45761 .
the authors also acknowledge the financial support from the computational material science network operated by us doe and from the ministry of education , science and sport of slovenia .
high frequency moment expansions for the green function and the self energy can be worked out : @xmath156where the moments of the green function are given by the following formulae@xmath157the expansion for the self energy is obtained as follows@xmath158where the self
energy moments are given by @xmath159^{2}-\sum_{k}|v_{k}|^{2}=\sum_{\beta \beta ^{\prime } } u_{\alpha \beta } u_{\alpha \beta ^{\prime } } ( \langle n_{\beta } n_{\beta ^{\prime } } \rangle -\langle n_{\beta } \rangle \langle n_{\beta ^{\prime } } \rangle ) , \label{sa1 } \\ \sigma _ { \alpha } ^{(2 ) } & = & g_{\alpha } ^{(4)}-2g_{\alpha } ^{(2)}g_{\alpha } ^{(3)}+[g_{\alpha } ^{(2)}]^{3}-\sum_{\mathbf{k}}|v_{\alpha } ( \mathbf{k})|^{2}e_{\mathbf{k}\alpha } = \notag \\ & = & \sum_{\beta \beta ^{\prime } \beta ^{\prime \prime } } u_{\alpha \beta } u_{\alpha \beta ^{\prime } } u_{\alpha \beta ^{\prime \prime } } ( \langle n_{\beta } n_{\beta ^{\prime } } n_{\beta ^{\prime \prime } } \rangle -2\langle n_{\beta } n_{\beta ^{\prime } } \rangle \langle n_{\beta ^{\prime \prime } } \rangle + \langle n_{\beta }
\rangle \langle n_{\beta ^{\prime } } \rangle \langle n_{\beta ^{\prime \prime } } \rangle ) \notag \\ & + & \epsilon _ { f\alpha } \sum_{\beta \beta ^{\prime } } u_{\alpha \beta } u_{\alpha \beta ^{\prime } } ( \langle n_{\beta } n_{\beta ^{\prime } } \rangle -\langle n_{\beta } \rangle \langle n_{\beta ^{\prime } } \rangle ) + \sum_{\mathbf{k},\beta } u_{\alpha \beta } ^{2}v_{\beta } ( \mathbf{k})\langle \left ( 2n_{\alpha } -1\right ) d_{\beta } ^{\dagger } c_{\mathbf{k}\beta } \rangle , \label{sa2}\\ \sigma_{\alpha}^{(3 ) } & = & g_{\alpha}^{(5)}-2 g_{\alpha}^{(4 ) } g_{\alpha}^{(2)}-\left[g_{\alpha}^{(3)}\right]^2 + 3 g_{\alpha}^{(3 ) } \left[g_{\alpha}^{(2)}\right]^2-\left[g_{\alpha}^{(2)}\right]^4-\sum_{{\mathbf k}}|v_{\alpha}({{\mathbf k}})|^2 e_{{{\mathbf k}}{\alpha}}^2= \nonumber\\ & = & \sum_{{\beta}_1{\beta}_2{\beta}_3{\beta}_4}u_{{\alpha}{\beta}_1}u_{{\alpha}{\beta}_2}u_{{\alpha}{\beta}_3}u_{{\alpha}{\beta}_4 } \left(\langle n_{{\beta}_1}n_{{\beta}_2}n_{{\beta}_3}n_{{\beta}_4}\rangle-2\langle n_{{\beta}_1}n_{{\beta}_2}n_{{\beta}_3}\rangle\langle n_{{\beta}_4}\rangle-\langle n_{{\beta}_1}n_{{\beta}_2}\rangle\langle n_{{\beta}_3}n_{{\beta}_4}\rangle \right . \nonumber\\&&\qquad\qquad\qquad\qquad\qquad\qquad + \left . 3\langle n_{{\beta}_1}n_{{\beta}_2}\rangle\langle n_{{\beta}_3}\rangle\langle n_{{\beta}_4}\rangle- \langle n_{{\beta}_1}\rangle\langle n_{{\beta}_2}\rangle\langle n_{{\beta}_3}\rangle\langle n_{{\beta}_4}\rangle \right)\nonumber\\ & + & 2\epsilon_{\alpha}\sum_{{\beta}_1{\beta}_2{\beta}_3}u_{{\alpha}{\beta}_1}u_{{\alpha}{\beta}_2}u_{{\alpha}{\beta}_3 } \left(\langle n_{{\beta}_1}n_{{\beta}_2}n_{{\beta}_3}\rangle-2\langle n_{{\beta}_1}n_{{\beta}_2}\rangle\langle n_{{\beta}_3}\rangle+\langle n_{{\beta}_1}\rangle\langle n_{{\beta}_2}\rangle\langle
n_{{\beta}_3}\rangle \right)\nonumber\\ & + & \epsilon_{\alpha}^2\sum_{{\beta}_1{\beta}_2}u_{{\alpha}{\beta}_1}u_{{\alpha}{\beta}_2 } \left(\langle n_{{\beta}_1}n_{{\beta}_2}\rangle-\langle n_{{\beta}_1}\rangle\langle n_{{\beta}_2}\rangle\right)+{\cal n}_{\alpha}^{(3)}. \label{sa3}\end{aligned}\ ] ] the nonlocal part of the self energy moment , which explicitly depends on hybridization matrix @xmath160 and vanishes in the atomic limit , is explicitly quoted only for the second moment ( last term in eq .
( [ sa2 ] ) ) . in the third order ,
the nonlocal term is denoted by @xmath161 and has not yet been evaluated . in practical calculations ,
we neglected it s contribution . within @xmath43
symetric case the formulae are simplified .
first , introduce the definitions : @xmath162where @xmath123 is probability for @xmath39 times occupancy ( with @xmath39 electrons on the local level ) .
then @xmath163 , \label{fu2 } \\ & & \sum_{\beta \beta ^{\prime } \beta ^{\prime \prime } } u_{\alpha \beta } u_{\alpha \beta ^{\prime } } u_{\alpha \beta ^{\prime \prime } } \langle n_{\beta } n_{\beta ^{\prime } } n_{\beta ^{\prime \prime } } \rangle = u^{3}\left [ ( n-1)(n-2)(n-3)\langle nnn\rangle + 3(n-1)(n-2)\langle nn\rangle + ( n-1)\langle n\rangle \right ] , \label{fu3 } \\ & & \sum_{\mathbf{k}\beta } u_{\alpha \beta } ^{2}v_{\beta } ( \mathbf{k})\langle \left ( 2n_{\alpha } -1\right ) f_{\beta } ^{\dagger } c_{\mathbf{k}\beta } \rangle = ( n-1)(zu)^{2}(1 - 2\kappa ) \int { \frac{d\xi } { \pi } } f(\xi ) \mathrm{im}\{\delta ( \xi ) g_{g}(\xi ) \}. \label{fuv}\end{aligned}\ ] ] introducing the notations@xmath164we , for example , obtain the following formuale@xmath165 u^{2 } , \label{sf1 } \\
\sigma ^{(2 ) } & = & ( n-1)(n-2)(n-3)\langle nnn\rangle u^{3}+(n-1)(n-2)\langle nn\rangle ( 3u^{3}+\epsilon _ { f}u^{2}-2(n-1)\langle n\rangle u^{3 } ) , \notag \\ & + & ( n-1)^{3}\langle n\rangle ^{3}u^{3}-(n-1)^{2}\langle n\rangle ^{2}(2u^{3}+\epsilon _ { f}u^{2})+(n-1)\langle n\rangle ( u^{3}+\epsilon _ { f}u^{2})+(n-1)(zu)^{2}(1 - 2\kappa ) { \delta g}_{g}. \label{sf2}\end{aligned}\ ] ]
in the crystal field case we assume that @xmath166fold degenerate impurity level @xmath167 is split by a crystal field onto @xmath168 sublevels @xmath169 .
we assume that for each sublevel there is still some partial degeneracy @xmath170 so that @xmath171 in limiting case of @xmath43 degeneracy , @xmath172 , and in non degenerate case , @xmath173 .
we need to discuss how a number of electrons @xmath39 can be accommodated over different sublevels @xmath174 . introducing numbers of electrons on each sublevel , @xmath36
, we obtain @xmath175 note the restrictions : @xmath176 @xmath177 and @xmath178 . in @xmath43 case , @xmath179 , and in non degenerate case , @xmath180 , @xmath36 is either 0 or 1 .
total energy for the shell with @xmath39 electrons depends on particular configuration @xmath181 @xmath182 .
\label{slacrflev}\ ] ] many body wave function is also characterized by a set of numbers @xmath181 , i.e. @xmath183 energy @xmath184 remains degenerate , which can be calculated as product of how many combinations exists to accommodate electrons in each sublevel , i.e. @xmath185 let us further introduce probabilities @xmath186 to find a shell in a given state with energy @xmath187 sum of all probabilities should be equal to @xmath155 , i.e.@xmath188 there are two green functions in gutzwiller method : impurity green function @xmath189 and quasiparticle green function @xmath190 where matrix coefficients @xmath191 represent generalized mass renormalizations parameters .
all matrices assumed to be diagonal and having diagonal elements numerated as follows : @xmath192 each element in the green function is represented as follows@xmath193@xmath194and determines a mean number of electrons in each sublevel@xmath195the total mean number of electrons is thus : @xmath196 hybridization function @xmath197 is the matrix assumed to be diagonal and having diagonal elements numerated as follows : @xmath198 mass renormalizations @xmath199 are determined in each sublevel .
the generalization of the non linear equations ( [ guz ] ) has the form@xmath202 \psi _ { n_{1} ... n_{g}}+ \notag \\ & & \sum_{\alpha = 1}^{g}n_{\alpha } [ t\sigma _ { i\omega } \delta _ { \alpha } ( i\omega ) g_{g\alpha } ( i\omega ) ] b_{\alpha } \left [ r_{\alpha } l_{\alpha } \psi _ { n_{1} ... n_{\alpha } -1 ... n_{g}}+b_{\alpha } l_{\alpha } ^{2}\psi _ { n_{1} ... n_{\alpha } ... n_{g}}\right ] + \notag \\ & & \sum_{\alpha = 1}^{g}(d_{\alpha } -n_{\alpha } ) [ t\sigma _ { i\omega } \delta _ { \alpha } ( i\omega ) g_{g\alpha } ( i\omega ) ] b_{\alpha } \left [ r_{\alpha } l_{\alpha } \psi _ { n_{1} ... n_{\alpha } + 1 ... n_{g}}+b_{\alpha } r_{\alpha } ^{2}\psi _ { n_{1} ... n_{\alpha } ... n_{g}}\right ]
. \label{slacrfscf}\end{aligned}\ ] ] k. held _ et al .
_ , psi k newsletter # * 56 * ( april 2003 ) , p. 65 ; a. i. lichtenstein , m. i. katsnelson , and g. kotliar , in _ electron correlations and materials properties _ , ed . by a. gonis ,
n. kioussis and m. ciftan ( kluwer academic , plenum publishers , 2002 ) p. 428 .
|
a rational representation for the self - energy is explored to interpolate the solution of the anderson impurity model in general orbitally degenerate case .
several constrains such as the friedel s sum rule , high frequency moments and the value of quasiparticle residue are used to establish the equations for the coefficients of the interpolation .
we test two fast techniques , the slave boson mean field and the hubbard i approximation to determine the coefficients .
the obtained self energies are compared with the results of numerically exact quantum monte carlo method .
we find that using the slave boson mean field approach we can construct an accurate self
energy for all frequencies via the proposed interpolation procedure .
|
constraint satisfaction problems are fundamental computational problems that arise in many areas of theoretical computer science . in recent years
, a considerable amount of research has been concentrated on the classification of those constraint satisfaction problems that can be solved in polynomial time , and those that are computationally hard . in this paper
, we contribute to this line of research and generalize an important tractability condition from finite domain constraint satisfaction to a broad class of infinite domain constraint satisfaction problems .
we work with the following definition of constraint satisfaction problems ( csps ) , which is well - adapted to treat csps over infinite as well as finite domains .
the definition is based on the concept of a _ homomorphism _ between relational structures , and equivalent to the standard definition for finite domain csps .
a _ ( relational ) structure _ @xmath1 consists of a ( not necessarily finite ) _ domain _ @xmath2 ( or simply @xmath1 when no confusion can arise ) , and a set of relations on @xmath2 , each of a finite positive arity .
each relation is named by a _ relation symbol _ @xmath3 ; the corresponding relation in @xmath1 is denoted by @xmath4 .
the set of all relation symbols is called the _ signature _ of the structure .
a _ homomorphism _ from a relational structure @xmath1 to a relational structure @xmath5 over the same signature is a mapping @xmath6 such that for each relation symbol @xmath3 , and tuple @xmath7 , it holds that @xmath8 , where @xmath9 is applied component - wise to @xmath10 .
the existence of a homomorphism from @xmath1 to @xmath5 is denoted by @xmath11 . for a fixed structure , traditionally denoted by @xmath12 , with finite relational signature @xmath13 the _ constraint satisfaction problem for @xmath12 _ ( denoted by @xmath14 )
is the following problem .
* csp(@xmath12 ) * + instance : a finite structure @xmath1 over the signature @xmath13 .
+ question : is there a homomorphism from @xmath1 to @xmath12 ?
+ to give an example , the three - colorability problem can be formulated as @xmath15 , where @xmath16 is the complete graph with three elements .
a basic example of an infinite - domain csp is @xmath17 , where @xmath18 is the strict linear ordering of the rationals .
jeavons and co - authors @xcite made the ground - breaking observation that for finite structures @xmath12 , the complexity of @xmath14 is captured by the _
polymorphisms _ of @xmath12 , defined as follows . when @xmath19 is a @xmath20-ary function , and @xmath3 is an an @xmath0-ary relation over @xmath21 , then we say that _ @xmath9 preserves @xmath3 _ if for all @xmath0-tuples @xmath22 , we have @xmath23,\dots , t_k[1 ] ) , \dots , f(t_1[n],\dots , t_k[n ] ) ) \in r.\ ] ] a _ polymorphism _ of a relational structure @xmath12 with domain @xmath24 is a function from @xmath25 to @xmath21 that preserves all relations of @xmath12 . in other words ,
a polymorphism is a homomorphism from @xmath26 to @xmath12 , for some @xmath20 .
the exploitation of polymorphisms for classifying the complexity of csps is sometimes referred to as the _ universal - algebraic approach_. indeed , very often tractability of @xmath14 is linked to polymorphisms of @xmath12 with certain ` good properties ' . for a finite domain @xmath27 , it is known that @xmath14 is np - hard unless @xmath12 has a polymorphism @xmath19 satisfying @xmath28 for all @xmath29 , and it is conjectured that @xmath14 can be solved in polynomial - time whenever @xmath12 has such a polymorphism @xcite a _ cyclic term _ when it satisfies the additional requirement of being idempotent ; @xmath30 for all @xmath31 .
they state the conditions for np - hardness and the conjecture for polynomial - time tractability in terms of the absence or presence of such cyclic terms among the polymorphisms of the _ core of @xmath12 _ , cf . section [ sect : mc - cores ] .
it is not hard to verify that their condition is equivalent to the one given here . ] .
this conjecture is known to hold in several special cases when @xmath9 satisfies stronger identities .
we will now look at one such case .
an operation @xmath32 is called * _ idempotent _ if @xmath33 ; * _ commutative _ if @xmath34 for all @xmath35 ; * _ associative _ if @xmath36 for all @xmath37 ; * a _ semi - lattice operation _ if it is commutative , associative , and idempotent . jeavons , cohen , and gyssens @xcite proved that for every finite structure @xmath12 with a polymorphism that is a semi - lattice operation , the problem @xmath14 can be solved in polynomial time . in this paper , we present a generalization of this result to a large class of infinite - domain csps .
all infinite structures considered are assumed to be countably infinite . to state the result
, we need the following definitions . a bijective homomorphism with an inverse that is also a homomorphism is called an _ isomorphism_. an _ automorphism _ of a relational structure @xmath12 is an isomorphism between @xmath12 and itself , and the set of all automorphisms of @xmath12 is denoted by @xmath38 . for a subset @xmath39 of the domain of @xmath12 , the _ orbit of @xmath39 _ in @xmath12 is the set @xmath40 .
when @xmath39 is of cardinality @xmath0 , then we call the orbit of @xmath39 in @xmath12 an _ orbit of @xmath0-subsets_. if the number of orbits of @xmath0-subsets of @xmath12 is at least @xmath41 for some @xmath42 and all sufficiently large @xmath0 , then we say that @xmath12 has _ exponential growth_. otherwise , we say that @xmath12 has sub - exponential growth , or that @xmath12 is a _ sub - exponential structure_. note that every finite structure is a sub - exponential structure since for @xmath0 greater than the domain size , there are no @xmath0-subsets at all , hence zero orbits .
but also the structure @xmath18 ( and all structures with domain @xmath43 whose relations are first - order definable in @xmath18 ) is sub - exponential : it has only one orbit of @xmath0-subsets , for all @xmath0 ( see e.g. @xcite ) .
our main result is the following .
[ thm : main ] let @xmath12 be a sub - exponential structure with finite relational signature . if @xmath12 has a semi - lattice polymorphism , then @xmath14 can be be solved in polynomial time .
finite domain structures with a semi - lattice polymorphism can be solved by a standard technique , known as _ establishing arc - consistency _
the situation is different for sub - exponential structures with a semi - lattice polymorphism .
consider for instance the structure @xmath44 .
it has the same automorphism group as @xmath18 and hence is sub - exponential as well .
this structure has the function @xmath45 on @xmath43 as a polymorphism , but it has been shown that there is no @xmath20 such that this problem can be solved by establishing @xmath20-consistency @xcite .
tractability of @xmath46 instead follows from the following more general result of @xcite : every structure @xmath12 with domain @xmath47 that has the same automorphism group as @xmath18 and that is preserved by the minimum function ( or the maximum function ) has a polynomial - time tractable csp .
our result will be a proper generalization of this result and of the mentioned result of jeavons , cohen , and gyssens .
let us remark that for a general infinite structure , a semi - lattice polymorphism does not suffice to ensure tractability . for an arbitrary subset @xmath48 of @xmath49 , let @xmath50 be the structure @xmath51 .
every such structure has min ( and max ) as a semi - lattice polymorphism .
we claim that csp(@xmath50 ) and csp(@xmath52 ) are different problems for distinct subsets @xmath48 and @xmath53 of @xmath49 .
let @xmath54 be any element in @xmath55 , where @xmath56 denotes the symmetric difference of the sets , and let @xmath57 be the instance on variables @xmath58 containing the constraints @xmath59 , @xmath60 for @xmath61 , and @xmath62 .
then @xmath57 is a satisfiable instance for precisely one of the problems csp(@xmath50 ) and csp(@xmath52 ) .
it follows that there are as many pairwise distinct csps of this type as there are subsets of @xmath49 , i.e. , uncountably many .
however , there are no more than countably many algorithms , hence csp@xmath63 is undecidable for some @xmath48 .
the example of the previous paragraph relies on the fact that for all structures @xmath50 , there is an _
infinite _ number of orbits of @xmath64-subsets .
when the number of orbits of @xmath0-subsets of @xmath12 is finite for all @xmath0 , then the structure is called _
@xmath65-categorical _ in model theory @xcite . for @xmath65-categorical structures ,
polymorphisms still capture the computational complexity of @xmath14 @xcite .
our main result can thus be seen as a contribution to the further extension of the universal - algebraic approach from finite to @xmath65-categorical structures .
our paper is structured as follows . in section [
sect : alg ] we introduce a new algorithmic technique to solve infinite - domain constraint satisfaction problems , and present a reduction of @xmath14 for structures @xmath12 with a semi - lattice polymorphism to an _ efficient sampling algorithm _ for @xmath12 .
the basic idea is that when there is such an efficient sampling algorithm for @xmath12 , then we can use the arc - consistency procedure for finite domain csps to solve @xmath14 ( we actually use the _ uniform _ version of the arc - consistency procedure where both @xmath1 and a finite template @xmath5 are part of the input ) .
in fact , this technique works under a slightly more general assumption on @xmath12 : instead of requiring the existence of a semi - lattice polymorphism , we only require that @xmath12 has _ totally symmetric _
polymorphisms of all arities .
the next part of our paper , section [ sect : classification ] , is devoted to the proof that all sub - exponential structures @xmath12 with totally symmetric polymorphisms of all arities admit such an efficient sampling algorithm . here , our proof is based on a _ classification _ of those structures @xmath12 .
we would like to remark that the general algorithmic technique is applicable also for many structures with totally symmetric polymorphisms of all arities that are _ not _ sub - exponential , and this will be illustrated by some examples in section [ sect : beyond ] .
in fact , we make the conjecture that when @xmath12 is an @xmath65-categorical structure with a semi - lattice polymorphism , then @xmath14 is in p. however , unlike the case of sub - exponential structures , we can not provide a classification result like the one in section [ sect : classification ] for this more general case , and so this remains an interesting open question .
one of the basic building blocks of our algorithm will be the arc - consistency procedure .
we start this section by recalling , in the case of finite relational structures , the connection between the applicability of this procedure , homomorphisms from the _ set structure _ , and the existence of _ totally symmetric polymorphisms_. let @xmath5 be a finite structure with a finite relational signature , and let @xmath1 be an instance of @xmath66 . the _ arc - consistency procedure _ ( ac ) applied to the problem @xmath67 works by reducing a set of possible images for each variable in @xmath1 .
if such a set becomes empty during the procedure , it follows that there can be no homomorphism , so ac rejects .
otherwise , ac accepts .
an algorithm for ac is shown in figure [ fig : alg - ac ] .
we say that arc - consistency _ solves _ the problem @xmath66 if , for every instance @xmath1 , the procedure accepts _ if and only if _ @xmath11 .
arc - consistency is sometimes called _ hyperarc - consistency _ when applied to structures with relations of arity greater than two .
it can be implemented to run in time that is polynomial in @xmath68 , take e.g. ac-3 @xcite .
the _ set structure _ of @xmath5 , denoted by @xmath69 , has as vertices all non - empty subsets of the domain of @xmath5 .
for every @xmath20-ary relation @xmath70 , we have @xmath71 iff for every @xmath72 and @xmath73 , there exists @xmath74 such that @xmath75 .
a @xmath20-ary function @xmath76 is called _ totally symmetric _
if for all @xmath77 we have @xmath78 we say that a structure @xmath5 has _ totally symmetric polymorphisms of all arities _ if , for each @xmath79 , there is a @xmath20-ary polymorphism of @xmath12 that is totally symmetric .
the following is well - known , cf .
@xcite .
[ thm : equiv ] let @xmath5 be a finite structure with a finite relational signature .
the following are equivalent . 1 .
the arc - consistency procedure solves @xmath66 .
2 . there is a homomorphism @xmath80 .
the structure @xmath5 has totally symmetric polymorphisms of all arities .
it is clear that when @xmath12 has a semi - lattice operation @xmath9 , then it also has a totally symmetric polymorphism @xmath81 of arity @xmath0 , for each @xmath82 : @xmath83 from theorem [ thm : equiv ] , it thus follows that arc - consistency solves @xmath66 whenever @xmath5 is a finite relational structure with a semi - lattice polymorphism . in our arguments
, we only need the weaker condition on totally symmetric polymorphisms of all arities ; this gives us a stronger result .
the other component of our algorithm will be a procedure to efficiently `` sample '' appropriate finite substructures of an infinite structure @xmath12 .
formally , we make the following definition .
let @xmath12 be a structure over a finite relational signature .
we say that an algorithm is a _ sampling algorithm for @xmath12 _ if , given a positive integer @xmath0 , it computes a finite structure @xmath5 that is isomorphic to a substructure of @xmath12 such that @xmath11 if and only if @xmath84 , for every instance @xmath1 with @xmath85
. a sampling algorithm is called _ efficient _ if its running time is bounded by a polynomial in @xmath0 .
we are now ready to give an outline of our algorithm .
let @xmath12 be a sub - exponential structure over a finite relational signature , and assume that @xmath12 has totally symmetric polymorphisms of all arities .
the main idea of our algorithm is to reduce @xmath14 to an appropriate _ uniform _ finite domain csp .
that is , when given an instance @xmath1 of @xmath14 , we reduce to the following problem : decide whether @xmath1 maps homomorphically to @xmath5 , where @xmath5 is a finite substructure of @xmath12 returned by a sampling algorithm for @xmath12 on input @xmath86 , and @xmath5 is considered as part of the input .
the algorithm is given in figure [ fig : alg - csp ] .
note that we rely on arc - consistency for deciding whether @xmath1 maps homomorphically to @xmath5 .
hence , for this approach to work , we need to establish the following . 1
. there should be an efficient sampling algorithm which samples some @xmath5 from @xmath12 .
2 . the arc - consistency procedure applied to @xmath87 should accept if and only if @xmath11 .
the first condition implies that the size of @xmath5 is polynomial in the size of @xmath1 , and since ac can be implemented to run in time that is polynomial in @xmath68 , it follows that our algorithm will be polynomial in @xmath86 .
the second condition ensures that the algorithm gives the correct answer for every instance @xmath1 of @xmath14
. it will be the purpose of section [ sect : classification ] to prove that an efficient sampling algorithm for @xmath12 exists .
we state this result as follows .
[ thm : alg - cond ] let @xmath12 be a sub - exponential structure with a finite relational signature and totally symmetric polymorphisms of all arities . then there is an efficient sampling algorithm for @xmath12 .
we next prove one part of a generalization of the equivalence between the second and third item of theorem [ thm : equiv ] to infinite domains .
this result has a converse for all @xmath65-categorical structures , cf .
section [ sect : disc ] .
[ lem : totsymtohomo ] let @xmath12 be a structure over a finite relational signature .
if @xmath12 has totally symmetric polymorphisms of all arities , then @xmath88 for all finite substructures @xmath89 .
let @xmath39 be a finite substructure of @xmath12 , and let @xmath9 be an @xmath54-ary totally symmetric polymorphism of @xmath12 , where @xmath90 and @xmath20 is the maximum arity of any relation in @xmath12 .
let @xmath91 be the function defined on a non - empty set @xmath92 by @xmath93 , where the list of arguments of @xmath9 is padded to length @xmath54 by elements already occurring in @xmath94 .
we claim that @xmath95 is a homomorphism @xmath88 .
we must show that @xmath96 for an arbitrary relation @xmath3 , and tuples @xmath97 . for each @xmath98 ,
let @xmath99 be a set of @xmath20-tuples such that @xmath100 \mid t \in t_i \
} = u_i$ ] .
note that @xmath100 \mid t \in t_j , 1 \leq j \leq k \ } = u_i$ ] , and that we can choose @xmath101 to have size @xmath102 , for each @xmath72 .
let @xmath103 be an enumeration of @xmath104 and @xmath105 be an enumeration of @xmath101 .
we then have @xmath106,\dots , t^{1}_{l(1)}[i],\dots , t^{k}_1[i],\dots , t^{k}_{l(k)}[i],t^{k}_{l(k)}[i ] , \dots , t^{k}_{l(k)}[i]),\ ] ] for all @xmath98 .
since @xmath9 is a polymorphism of @xmath12 , and @xmath107 for all @xmath72 and @xmath108 , it follows that @xmath96 .
the correctness and efficiency of our algorithm now follows from the previous lemma in conjunction with the existence of an efficient sampling algorithm for @xmath12 .
[ thm : correctness ] let @xmath12 be a structure over a finite relational signature with totally symmetric polymorphisms of all arities .
if there exists an efficient sampling algorithm for @xmath12 , then the algorithm in figure [ fig : alg - csp ] correctly solves @xmath14 in polynomial time .
let @xmath1 be the input structure and let @xmath5 be the structure returned by the sampling algorithm for @xmath12 on input @xmath86 .
the sampling algorithm runs in polynomial time in @xmath86 , so the size of @xmath5 will be polynomial in @xmath86 . since ac(@xmath109 ) can be implemented to run in time polynomial in @xmath68 , it follows that the entire algorithm runs in polynomial time . to show correctness ,
note that if ac(@xmath109 ) rejects , then @xmath110 which is equivalent to @xmath111 since @xmath5 was produced by the sampling algorithm for @xmath12 on input @xmath86 .
we may therefore safely reject .
otherwise , ac@xmath87 terminates with a non - empty list @xmath112 for each @xmath113 .
furthermore , for each @xmath20-ary relation @xmath4 , tuple @xmath114 , index @xmath72 , and element @xmath115 , we know that @xmath116 contains a tuple @xmath117 with @xmath118 . in other words
@xmath119 , so the function @xmath120 is a homomorphism . by assumption , @xmath12 has totally symmetric polymorphisms of all arities , so lemma [ lem : totsymtohomo ] implies the existence of a homomorphism @xmath121 . in conclusion
, we have a homomorphism @xmath122 . as a direct corollary of theorem [ thm : alg - cond ] and theorem [ thm : correctness ]
, we get the following result , which also implies our main result , theorem [ thm : main ] .
[ thm : yyy ] let @xmath12 be a sub - exponential structure with a finite relational signature . if @xmath12 has totally symmetric polymorphisms of all arities , then @xmath14 is solvable in polynomial time .
the purpose of this section is to prove theorem [ thm : alg - cond ] , showing that every sub - exponential structure has an efficient sampling algorithm .
our approach is based on a _ classification _ of sub - exponential structures with totally symmetric polymorphisms of all arities .
the general outline of this classification is as follows .
we first present an argument that reduces the classification task to those sub - exponential structures @xmath12 that are _ model - complete cores _ and have totally symmetric polymorphisms of all arities ( section [ sect : mc - cores ] ) .
from there on , the classification follows a decomposition of the automorphism group of @xmath12 .
the next step is the reduction to those structures @xmath12 having a _ transitive _ automorphism group ( section [ ssect : general ] ) .
we then use the fact that the automorphism group of @xmath12 has only finitely many congruence relations to further reduce the classification task to the case that the automorphism group of @xmath12 is _ primitive _ ( section [ ssect : finest ] ) . combining the central theorem from @xcite on primitive permutation groups with a sub - exponential number of orbits of @xmath0-subsets with cameron s theorem on highly set - transitive permutation groups
@xcite we finish the classification in section [ ssect : primitive ] .
we find it instructive to give a ` top - down ' presentation of the classification proof , rather than starting from special cases , and to assemble more general sub - exponential structures from specific ones .
we thus take a decomposition approach , and show first how to describe the most general case in terms of its components .
this sometimes leads to forward - references of results , but we believe that the reader will be compensated by a more accessible presentation . before we start , we recall a basic fact which will frequently be used in the following arguments , and which explains the interaction between permutation group theory and logic for @xmath65-categorical structures .
let @xmath12 be a relational structure . in this paper
, we say that a relation @xmath3 over the domain of @xmath12 is _ first - order definable _ in @xmath12 if there is a first - order formula @xmath123 such that @xmath124 is true in @xmath125 if and only if @xmath126 .
we say that a relational structure @xmath127 is first - order definable in @xmath12 if @xmath127 and @xmath12 have the same domain and every relation @xmath3 in @xmath127 is first - order definable in @xmath12 .
we say that two relational structures are _ first - order interdefinable _ if one of them is first - order definable in the other , and vice versa .
[ thm : ryll ] let @xmath12 be @xmath65-categorical .
then a relation @xmath3 is preserved by all automorphisms of @xmath12 if and only if @xmath3 has a first - order definition in @xmath12 .
in particular , two structures are first - order interdefinable if and only if they have the same automorphism group .
this theorem makes possible a translation of terminology between logic and permutation groups .
we illustrate its use with the following , which will be needed later on .
congruence _ of a permutation group is an equivalence relation that is preserved by all permutations in the group .
a permutation group is called _ primitive _ if the only congruences are the equivalence relation with just one equivalence class , and the equivalence relation where all equivalence classes are of size one ; it is called _ imprimitive _ otherwise . by theorem [ thm : ryll ] , the congruences of the automorphism group of an @xmath65-categorical structure @xmath12 are precisely the first - order definable equivalence relations of @xmath12 .
we will say that @xmath12 is _ primitive _ if its automorphism group is primitive .
a permutation group @xmath128 on a countable set @xmath94 is called _ closed _ if and only if it is the automorphism group of a relational structure with domain @xmath94 .
the topological explanations for this terminology can be found in @xcite , theorem 2.4.4 .
the corresponding topology is called the _ topology of pointwise convergence _ on @xmath129 , where @xmath129 denotes the set of all permutations of @xmath94 . in this topology ,
the open sets are unions of sets of the form @xmath130 , for @xmath0-tuples @xmath131 , @xmath132 of elements of @xmath94 .
a subset @xmath133 of @xmath134 is _ dense ( in @xmath128 ) _ if the closure of @xmath133 with respect to this topology equals @xmath128 .
an endomorphism of a relational structure @xmath12 is a homomorphism from @xmath12 to @xmath12 ; we denote the set of all endomorphisms of @xmath12 by @xmath135 .
a relational structure is called a _ core _ if every endomorphism of @xmath12 is an embedding into a relational structure @xmath56 is an isomorphism between @xmath12 and an induced substructure of @xmath56 . ] . for a relational structure @xmath56 , a _ core of @xmath56 _ is a core structure @xmath12 that is homomorphically equivalent to @xmath56 , that is , there is a homomorphism from @xmath12 to @xmath56 and vice versa .
a first - order formula is called _ primitive positive _ if it is of the form @xmath136 where @xmath137 are atomic formulas .
the importance of primitive positive definitions in this paper comes from the fact that relations with a primitive positive definition in a relational structure @xmath12 are preserved by the polymorphisms of @xmath12 .
the motivation of these definitions for constraint satisfaction with finite templates @xmath56 comes from the following facts . *
every finite relational structure @xmath56 has a core @xmath12 , and @xmath12 is unique up to isomorphism .
* when @xmath12 is a core of @xmath56 , then @xmath12 and @xmath56 have the same csp . * in a finite core structure @xmath12 , every orbit of @xmath0-tuples is an @xmath0-tuple of elements of @xmath12 , then the _ orbit of @xmath10 ( in @xmath12 ) _ is the set @xmath138 . ]
is primitive positive definable .
these properties have been generalized to @xmath65-categorical structures .
to precisely state the result , we need the following concepts .
a structure @xmath12 is _ model - complete _ if every embedding of @xmath12 into itself preserves all first - order formulas .
we later need the following .
[ lem : mc - cores ] an @xmath65-categorical relational structure @xmath56 is a model - complete core if and only if the group of automorphisms of @xmath56 is dense in the endomorphism monoid of @xmath56 , that is , for every endomorphism @xmath9 and finite subset @xmath48 of @xmath139 , there is an automorphism @xmath140 of @xmath56 agreeing with @xmath9 on @xmath48 .
[ thm : mc - cores ] let @xmath56 be an @xmath65-categorical relational structure
. then 1 .
[ enum : hascore ] @xmath56 is homomorphically equivalent to a model - complete core @xmath12 ; 2 .
[ enum : unique ] the structure @xmath12 is unique up to isomorphism , and @xmath65-categorical or finite ; 3 .
[ enum : ppdef ] in @xmath12 , every orbit of @xmath0-tuples is primitive positive definable . for our classification project ( and our algorithmic result )
, it therefore suffices to study the csps for model - complete cores of sub - exponential structures .
let us first show that the model - complete core of a sub - exponential structure is again sub - exponential .
this follows from the following more general result .
[ prop : orbit - non - growth ] let @xmath56 be an @xmath65-categorical relational structure , and let @xmath12 be its model - complete core .
then for every @xmath0 , the number of orbits of @xmath0-subsets in @xmath12 is at most the number of orbits of @xmath0-subsets in @xmath56 .
let @xmath9 be a homomorphism from @xmath56 to @xmath12 , and @xmath140 be a homomorphism from @xmath12 to @xmath56 .
since @xmath12 is a core , it follows that @xmath141 is an embedding , so @xmath140 is injective .
it now suffices to show that when two @xmath0-subsets @xmath142 of @xmath12 are mapped by @xmath140 to two @xmath0-subsets @xmath143 in the same orbit of @xmath0-subsets in @xmath56 , then @xmath144 and @xmath145 lie in the same orbit of @xmath0-subsets in @xmath12 .
let @xmath146 be an @xmath0-tuple listing all the elements in @xmath144 , let @xmath147 be an automorphism of @xmath56 that maps @xmath148 to @xmath149 , and let @xmath150 .
since @xmath151 lists all the elements of @xmath149 , we can arrange the elements of @xmath145 into an @xmath0-tuple @xmath152 such that @xmath153 . by theorem [ thm : mc - cores]([enum : ppdef ] ) , there are primitive positive definitions @xmath154 and @xmath155 of the orbits of @xmath146 and @xmath152
. since @xmath140 , @xmath147 , and @xmath9 preserve primitive positive formulas , the tuple @xmath156 satisfies @xmath154 .
but @xmath157 , and hence @xmath158 also satisfies @xmath155 .
therefore , @xmath154 and @xmath155 define the same orbit of @xmath0-tuples , and so @xmath146 and @xmath152 are in the same orbit .
this implies that @xmath144 and @xmath145 are in the same orbit of @xmath0-subsets .
in general it might not be true that the model - complete core of a sub - exponential structure with a semi - lattice polymorphism has again a semi - lattice polymorphism .
a finite example of this situation can be derived from proposition 5.2 in @xcite .
this example shows a finite poset with a semi - lattice polymorphism which retracts to a poset without a semi - lattice polymorphism . by introducing constants ,
the latter structure can be turned into a core of the former .
however , we always have the following .
let @xmath56 be an @xmath65-categorical relational structure with an @xmath0-ary totally symmetric polymorphism .
then the model - complete core of @xmath56 also has an @xmath0-ary totally symmetric polymorphism .
let @xmath12 be the model - complete core of @xmath56 , and let @xmath159 and @xmath160 be homomorphisms .
when @xmath9 is an @xmath0-ary totally symmetric polymorphism of @xmath56 , then @xmath161 defined by @xmath162 is totally symmetric , and a polymorphism of @xmath12 .
once we have a sampling algorithm for the core @xmath12 of a structure @xmath56 , we obtain a sampling algorithm for @xmath56 as follows .
let @xmath160 be a homomorphism and let @xmath163 be a sample returned by the sampling algorithm for @xmath12 on input @xmath0 .
we then let the sample @xmath5 of @xmath56 be the substructure induced by the image of @xmath163 under @xmath164 . for any structure @xmath1 of size @xmath0
it follows that @xmath165 implies @xmath84 which implies @xmath166 , so @xmath5 is indeed a sample of @xmath56 .
the structure @xmath5 can be computed in polynomial time in the size of @xmath163 which in turn is polynomial in @xmath0 , so @xmath5 can be computed in polynomial time in @xmath0 .
we conclude that it suffices to show theorem [ thm : alg - cond ] for the special case of sub - exponential structures that are model - complete cores and have totally symmetric polymorphisms of all arities .
let @xmath12 be a sub - exponential model - complete core .
since @xmath12 is sub - exponential , it has in particular a finite number of orbits of 1-subsets , called _ orbits _ for short .
a structure is called _ transitive _ if it has only one orbit .
[ prop : trans - mc - cores ] let @xmath12 be an @xmath65-categorical model - complete core , and let @xmath127 be the expansion of @xmath12 by all primitive positive definable relations .
let @xmath48 be an orbit of @xmath12 , and @xmath167 be the restriction of @xmath127 to @xmath48 .
then @xmath167 is a transitive model - complete core .
first observe that every automorphism @xmath147 of @xmath12 is also an automorphism of @xmath127 and preserves @xmath48 , and hence @xmath168 is an endomorphism of @xmath167 . since the same also applies to the inverse of @xmath147 , we have that also @xmath168 has an inverse in @xmath169 , and therefore is an automorphism of @xmath167 .
so , the restriction of an automorphism of @xmath12 to @xmath48 is an automorphism of @xmath167 , and therefore @xmath167 is transitive . to show that @xmath167 is a model - complete core , let @xmath170 be an endomorphism of @xmath167 , and let @xmath10 be a @xmath20-tuple of elements from @xmath48 .
then any primitive positive formula that holds on @xmath10 in @xmath12 also holds on @xmath171 , since @xmath167 is the restriction of an expansion of @xmath12 by all primitive positive definable relations .
since @xmath12 is a model - complete core , the orbits of @xmath20-tuples are primitive positive definable in @xmath12 by theorem [ thm : mc - cores](3 ) , and hence there is an automorphism @xmath147 of @xmath12 that maps @xmath170 to @xmath171 .
then @xmath168 is an automorphism of @xmath167 .
this shows that @xmath172 is dense in @xmath169 , and the statement follows from lemma [ lem : mc - cores ] .
the following result is proved in section [ ssect : finest ] and section [ ssect : primitive ] .
[ thm : transitive ] let @xmath56 be a transitive sub - exponential model - complete core with totally symmetric polymorphisms of all arities . then @xmath56 is either a structure of size 1 , or it is isomorphic to a structure which is first - order interdefinable with @xmath18 .
any such structure with a finite relational signature has an efficient sampling procedure .
so we are left with the task to design an efficient sampling algorithm for @xmath12 using the efficient sampling algorithms for each substructure of @xmath12 induced by an orbit of @xmath12 . for this , we need to analyze how the automorphism group @xmath128 of @xmath12 is built from its _ transitive constituents _ , that is , from the permutation groups of the form @xmath173 on @xmath48 where @xmath48 is an orbit of @xmath12 . in general , we only know that @xmath128 is a subdirect product of its transitive constituents ( see , e.g. , @xcite ) . in our case , we can make this decomposition more precise , since we have a good knowledge of the group @xmath174 .
[ lem : convex ] [ lem : convexorbits ] let @xmath12 be a sub - exponential structure with an @xmath175-invariant linear order @xmath176 defined on the union of two orbits @xmath48 and @xmath53
. then @xmath48 and @xmath53 are convex with respect to @xmath176 .
assume to the contrary that there are elements @xmath177 , @xmath178 such that @xmath179 .
since @xmath180 and @xmath181 lie in the same orbit , there is an automorphism @xmath147 of @xmath12 such that @xmath182 .
let @xmath183 for @xmath184 , and let @xmath185 for @xmath186 .
since @xmath147 preserves the order @xmath176 , we have @xmath187 , and @xmath188 . by repeated application of @xmath147
, we obtain @xmath189 for all @xmath184 , and finally @xmath190 .
hence , we can encode sequences @xmath191 in subsets @xmath89 of size @xmath54 by letting @xmath192 iff @xmath193 and @xmath194 iff @xmath195 .
different sequences of length @xmath54 then correspond to different orbits of @xmath54-subsets in @xmath12 .
this contradicts the assumption that @xmath12 is sub - exponential .
we now describe the automorphism groups of sub - exponential structures where the transitive constituents @xmath196 for all orbits @xmath197 of @xmath128 are isomorphic to @xmath174 : the following theorem shows that in this case @xmath128 is precisely what cameron @xcite calls the _ intransitive action _ of the direct product @xmath198 on @xmath199 .
we will say that a permutation group @xmath134 is _ transitive on a subset @xmath200 _ if for all @xmath201 , there exists a permutation @xmath202 such that @xmath203 .
otherwise , we say that @xmath128 is _ intransitive on @xmath204_. [ thm : no - interaction ] let @xmath12 be a sub - exponential structure with automorphism group @xmath128 and orbits @xmath205 , and for @xmath206 let @xmath207 be the transitive constituent of @xmath128 on @xmath104 .
assume that @xmath207 is isomorphic to @xmath174 for all @xmath98 .
then , @xmath208 is an automorphism of @xmath12 if and only if @xmath209 for each @xmath98 .
since @xmath207 is isomorphic to @xmath174 for each @xmath72 , there is by theorem [ thm : ryll ] a first - order definable dense linear order @xmath210 on each orbit @xmath104 .
we will use @xmath176 to collectively denote these orders , and rely on the context to determine which orbit it applies to .
the result is trivial for @xmath211 , so assume that @xmath212 .
let @xmath213 be the intransitive action of @xmath214 on @xmath199 .
it suffices to show that @xmath128 is dense in @xmath213 ; since @xmath128 is closed , they must then be equal .
assume for the sake of contradiction that @xmath128 is not dense in @xmath213 .
then , there are finite sequences @xmath215 and @xmath216 in @xmath12 , and an automorphism @xmath217 such that @xmath218 , but @xmath219 for all @xmath220
. furthermore we can choose @xmath215 and @xmath216 so that @xmath221 for some @xmath222 . by applying @xmath223 to @xmath224 and @xmath147
, we may then assume that @xmath225 and @xmath226 . by a _
@xmath227-interval _ , we will mean an inclusion - maximal convex subset of @xmath228 , for some @xmath72 , where convexity is evaluated with respect to @xmath104 .
now @xmath229 lies in some @xmath227-interval @xmath230 , and since @xmath231 , we have @xmath232 as well . by assumption
we have @xmath233 for all @xmath234 , so @xmath235 is intransitive on @xmath230 . on the other hand ,
if we let @xmath236 , where @xmath48 is the orbit of @xmath12 containing @xmath230 , then @xmath237 is clearly transitive on each @xmath238-interval contained in @xmath48 : each @xmath238-interval contained in @xmath48 is an orbit , and its corresponding transitive constituent of @xmath237 is isomorphic to @xmath239 . in particular , @xmath240 is transitive on @xmath230 .
it therefore follows that we can find a subsequence @xmath241 of @xmath227 containing @xmath238 , and an element @xmath242 such that @xmath243 is transitive on @xmath230 but @xmath244 is not . by lemma [ lem : convexorbits ]
, we have that every orbit of @xmath245 contained in @xmath230 is convex .
since @xmath244 is intransitive on @xmath230 , it follows that there is an initial segment @xmath246 definable in @xmath245 .
let @xmath247 , @xmath248 , and pick an automorphism @xmath249 such that @xmath203 .
now @xmath250 and @xmath251 is definable in @xmath252 .
let @xmath253 and for @xmath254 , let @xmath255 . by repeating this procedure , we get an increasing sequence of sets @xmath256 which are all definable in @xmath257 , where @xmath258 . for @xmath254 ,
pick @xmath259 , so that each of the elements @xmath260 lies in a different orbit of @xmath257 .
we now encode a set @xmath261 as the subset @xmath262 of @xmath12 consisting of the elements in @xmath241 , @xmath263 , and @xmath264 .
let @xmath265 be another set with encoding @xmath266 , and let @xmath202 be an automorphism such that @xmath267 .
then @xmath147 has to fix @xmath241 and @xmath263 , i.e. , @xmath147 must be an automorphism of @xmath257 .
therefore @xmath147 can not map @xmath260 to @xmath268 for @xmath269 , so @xmath39 must be equal to @xmath270 .
the set @xmath262 has size at most @xmath271 , and since we can fix the size of @xmath241 , we conclude that the number of distinct orbits of @xmath54-subsets of @xmath12 is at least @xmath272 .
this contradicts the sub - exponentiality of @xmath12 , so @xmath128 must be dense in @xmath213 , and the result follows .
together with the remarks of section [ sect : mc - cores ] , the following implies theorem [ thm : alg - cond ] . every sub - exponential model - complete core @xmath12 with totally symmetric polymorphisms of all arities and a finite relational signature @xmath13 has an efficient sampling algorithm . for a given @xmath0
, we have to compute a @xmath13-structure @xmath5 in polynomial time in @xmath0 such that for all structures @xmath1 of size @xmath0 we have @xmath84 if and only if @xmath11 .
let @xmath197 be the orbits of @xmath12 .
let @xmath127 be the expansion of @xmath12 by all primitive positive definable relations , and let @xmath273 be the structures induced in @xmath127 by the orbits @xmath197 of @xmath12 . by proposition [ prop : trans - mc - cores ] , @xmath274 is a transitive model - complete core , for all @xmath275
. since @xmath274 also has totally symmetric polymorphisms of all arities ( obtained as the restrictions of the totally symmetric polymorphisms of @xmath12 to @xmath104 ) , we can apply theorem [ thm : transitive ] , and conclude that each of the structures @xmath274 either has size 1 , or is isomorphic to a structure which is first - order interdefinable with @xmath18 .
we will now prove that an efficient sampling procedure exists in the case when each of the structures @xmath274 is isomorphic to a structure which is first - order interdefinable with @xmath18 .
the proof can easily be modified to handle the case when some of the structures @xmath274 have size 1 .
let @xmath54 be the maximal arity of @xmath13 , and let @xmath276 be the signature that contains a relation symbol for each at most @xmath54-ary primitive positive definable relation in @xmath13 .
for all @xmath275 , let @xmath277 be the structure obtained from @xmath278 by removing all relations except the relations for the symbols from @xmath276 .
then @xmath277 has a finite signature , and we conclude by theorem [ thm : transitive ] that @xmath277 has an efficient sampling procedure .
let @xmath279 be the @xmath276-structure produced by this sampler for @xmath277 on input @xmath0 ; we can assume that it has exactly @xmath0 vertices @xmath280
. the output of our algorithm will be the @xmath13-reduct of a @xmath276-structure @xmath5 with vertex set @xmath281 .
since the orbits are primitive positive definable in @xmath12 , there is a unary relation symbol @xmath282 in @xmath276 for each orbit @xmath48 of @xmath12 .
the structure @xmath5 will be such that @xmath283 , for all @xmath275 and @xmath284 .
note that by theorem [ thm : no - interaction ] , all such structures @xmath5 will be isomorphic . for @xmath285 , we now add tuples to the relation @xmath70 as follows .
we first fix a partition @xmath286 of the arguments of @xmath3 into at most @xmath20 parts .
let @xmath287 be the arguments of @xmath3 of the @xmath72-th part . for each @xmath72 the relation @xmath288,\dots , t[p^i_{l(i ) } ] )
\mid t \in r \text { and } t[p^j_l ] \in u_j \text { for all } j \leq k , l \leq l(j ) \big \}\end{aligned}\ ] ] is primitive positive definable in @xmath12 , and there is a relation symbol @xmath289 for this relation in @xmath276 .
it is clear that given @xmath286 and the relations @xmath289 of @xmath279 we can efficiently compute the relation @xmath290,\dots , t[p^i_{l(i ) } ] ) \in r(p , i ) \text { for all } i \leq k \big \ } \ ; .\ ] ] it follows from theorem [ thm : no - interaction ] that @xmath3 equals the union of @xmath291 over all partitions @xmath286 of the arguments of @xmath3 and since there is a constant number of such partitions , @xmath3 can be computed in polynomial time .
we are left with the task to prove theorem [ thm : transitive ] .
we now proceed to show that when @xmath12 is a transitive sub - exponential model - complete core with totally symmetric polymorphisms of all arities , then it is also primitive . combining this result with the following two theorems
shows that @xmath12 is isomorphic to a structure with a first - order definition in @xmath292 .
this allows us to finish the proof of theorem [ thm : transitive ] in the next section .
[ thm : macpherson ] let @xmath128 be a primitive but not highly set - transitive permutation group on an infinite set @xmath94 . if @xmath293 is a real number with @xmath294 , then @xmath128 has more than @xmath41 orbits of @xmath0-subsets of @xmath94 , for all sufficiently large @xmath0 .
[ thm : cameron ] a permutation group @xmath128 on an infinite set is highly set - transitive iff it is isomorphic to the automorphism group of a structure with a first - order definition in @xmath292 .
we will need two more lemmas .
let @xmath3 and @xmath39 be two binary relations .
an _ alternating closed walk on @xmath3 and @xmath39 of length @xmath295 _ is a sequence of elements @xmath296 , with @xmath297 , and such that @xmath298 and @xmath299 , for @xmath300 .
[ lem : aclwalk ] let @xmath3 and @xmath39 be two binary relations that are preserved by a totally symmetric function @xmath81 of arity @xmath301 . if there is an alternating closed walk on @xmath3 and @xmath39 of length @xmath295 , then @xmath302 . since @xmath298 for @xmath300 , we have @xmath303 , for @xmath304 and @xmath305 .
similarly , since @xmath299 for @xmath300 , we have @xmath306 , for @xmath307 and @xmath308 .
note that @xmath309 and that @xmath310
. therefore , @xmath303 and @xmath311 , hence @xmath312 .
an _ orbital _ of @xmath12 is an orbit of @xmath175 acting component - wise on ordered pairs of elements of @xmath12 .
every structure always has the _ trivial orbital _
@xmath313 .
[ lem : k1 ] let @xmath12 be a model - complete core with finitely many orbitals and a totally symmetric polymorphism @xmath81 of arity @xmath0 for all @xmath301 .
let @xmath94 be an equivalence class of a first - order definable equivalence relation on @xmath12 .
if @xmath314 for some @xmath315 and @xmath316 , then @xmath317 . by assumption , there exists a smallest integer @xmath318 such that @xmath319 .
let @xmath320 , and for @xmath321 , let @xmath322 be the orbital of @xmath12 containing the tuple @xmath323 .
then , we have the inclusion @xmath324 . since @xmath12 has finitely many orbitals , we can find integers @xmath325 such that @xmath326 .
in fact , we can do this while ensuring that @xmath327 .
note that @xmath328 , and that @xmath329 , for all @xmath72 .
in particular , the following sequence is an alternating closed walk on @xmath330 and @xmath331 of length @xmath332 .
@xmath333 as @xmath12 is a model - complete core , each orbital is primitive positive definable in @xmath12 , and hence preserved by @xmath81 for each @xmath301 . from lemma [ lem : aclwalk ]
, it now follows that the orbitals @xmath330 and @xmath334 intersect , and therefore they must be equal .
this implies that the tuples @xmath335 and @xmath336 are in the same orbital , so there exists an automorphism @xmath337 of @xmath12 which fixes @xmath338 and maps @xmath339 to @xmath340 . since @xmath337 fixes @xmath338 , we have @xmath341 , and from the choice of @xmath342 , we have @xmath343 .
it follows that @xmath344 , so @xmath345 , and hence @xmath346 as well . due to the choice of @xmath347 ,
this is only possible if @xmath348 , so we conclude that @xmath147 preserves @xmath94 .
we are now ready to prove the main result of this section .
[ prop : transisprim ] let @xmath12 be a transitive sub - exponential model - complete core with totally symmetric polymorphisms of all arities . then @xmath12 has size 1 , or it is infinite and primitive .
assume that @xmath12 has size at least 2 .
let @xmath349 and @xmath350 denote the congruence of @xmath175 with equivalence classes of size 1 , and the congruence with a single equivalence class , respectively .
let @xmath351 be an inclusion - maximal congruence from the set of all congruences different from @xmath350 .
existence of @xmath351 follows from @xmath12 having finitely many first - order definable equivalence relations , and the existence of @xmath349 .
we want to show that @xmath351 must in fact be @xmath349 from which it follows that @xmath175 is primitive .
let @xmath27 .
by @xmath352 we will denote the set of equivalence classes of @xmath351 . for @xmath31 ,
let @xmath353 $ ] denote the equivalence class of @xmath351 containing @xmath338 , and for @xmath315 , let @xmath354 denote the function on @xmath352 which maps @xmath353 $ ] to @xmath355 $ ] for each equivalence class @xmath353 $ ] of @xmath351 .
( it follows from @xmath351 being a congruence that @xmath354 is well - defined . )
we then have that @xmath356 is a permutation group on @xmath352 .
let @xmath357 .
for each equivalence class @xmath358 , choose @xmath359 such that @xmath360 = x_i$ ] .
let @xmath361 and @xmath362 be two @xmath0-subsets of @xmath363 such that @xmath364 for some @xmath315 . then , @xmath365 , so the number of orbits of @xmath0-subsets of @xmath133 is sub - exponential whenever @xmath12 is sub - exponential .
assume that @xmath133 is finite .
pick two distinct equivalence classes @xmath94 and @xmath204 of @xmath351 .
since @xmath12 is transitive , we can find @xmath315 such that @xmath366 .
but if @xmath133 is finite , then @xmath314 for some @xmath316 , so @xmath317 by lemma [ lem : k1 ] , a contradiction .
therefore @xmath133 must be infinite .
congruences of @xmath133 are in one - to - one correspondence with the congruences of @xmath175 containing @xmath351 .
as the latter are precisely @xmath351 and @xmath350 , it follows that @xmath133 is primitive .
it now follows from theorem [ thm : macpherson ] that @xmath133 is highly set - transitive , and so from theorem [ thm : cameron ] that @xmath133 is isomorphic to the automorphism group @xmath367 of a structure with a first - order definition in @xmath292 . the group @xmath367 either has one or two non - trivial orbitals .
if it only has one non - trivial orbital , then so does @xmath133 , hence for any two distinct equivalence classes @xmath94 and @xmath204 of @xmath351 , there is an automorphism @xmath147 of @xmath12 such that @xmath368 and @xmath369 .
again by lemma [ lem : k1 ] , it follows that @xmath317 , a contradiction .
so @xmath367 has two non - trivial orbitals , one of which is the order @xmath176 on @xmath43 . via the isomorphism
, @xmath133 thus has a non - trivial orbital @xmath176 which is a linear order on the equivalence classes of @xmath351 .
let @xmath3 be a binary relation on @xmath12 defined by @xmath370 iff @xmath353 < y[e]$ ] . for @xmath315
, we have that @xmath354 preserves @xmath176 , so @xmath147 preserves @xmath3 .
assume now that the equivalence classes of @xmath351 have size greater than 1 .
we can then encode a sequence in @xmath371 as a set @xmath372 : choose @xmath373 so that @xmath374 for @xmath375 , encode a value 0 in position @xmath72 by enforcing @xmath376 , @xmath377 , and encode a value 1 by enforcing @xmath378 .
the relations @xmath351 , @xmath379 , and @xmath3 are all preserved by @xmath175 , so if two @xmath295-subsets encodes distinct sequences , then they must be contained in distinct orbits of @xmath295-subsets .
hence , the number of orbits of @xmath295-subsets is greater than or equal to @xmath380 , which contradicts @xmath12 being sub - exponential .
so the equivalence classes of @xmath351 are of size 1 , i.e. , @xmath381 , and @xmath12 is primitive .
assume that @xmath12 is a transitive sub - exponential model - complete core with totally symmetric polymorphisms of all arities .
if @xmath12 is of size greater than 1 , then it is infinite and primitive ( proposition [ prop : transisprim ] ) , so it is highly set - transitive ( theorem [ thm : macpherson ] ) and hence isomorphic to a structure @xmath127 with a first - order definition in @xmath18 ( theorem [ thm : cameron ] ) . to prove the first part of theorem [ thm : transitive ]
, we have to show that @xmath127 is in fact first - order interdefinable with @xmath18 .
if the binary relation @xmath176 is an orbital of @xmath127 , then @xmath176 is first - order definable in @xmath127 by theorem [ thm : ryll ] , and interdefinability follows .
otherwise , the smallest orbital of @xmath127 that contains @xmath176 also contains a pair @xmath382 such that @xmath383 .
it follows that @xmath127 has an automorphism @xmath147 such that @xmath203 and @xmath384 , i.e. , @xmath385 .
since @xmath127 has totally symmetric polymorphisms of all arities , we can apply lemma [ lem : k1 ] to deduce @xmath386 , a contradiction . to prove the second part of theorem [ thm : transitive ]
, we show that any structure with a finite relational signature and a first - order definition in @xmath18 has an efficient sampling algorithm .
a relational structure @xmath12 is _ homogeneous _ ( or _ ultrahomogeneous _ ) if every isomorphism between finite induced substructures of @xmath12 can be extended to an automorphism of @xmath12 .
a structure @xmath12 with signature @xmath13 _ admits quantifier - elimination _ if every first - order @xmath13-formula is over @xmath12 equivalent to a quantifier - free @xmath13-formula .
an @xmath65-categorical structure admits quantifier - elimination iff it is homogeneous @xcite .
let @xmath387 : = \{1,\dots , n\}$ ] .
[ thm : primitive ] every structure @xmath12 with a finite relational signature @xmath13 and a first - order definition in @xmath18 has an efficient sampling algorithm . on input @xmath0 , the output of the algorithm is a structure of size @xmath0 .
the structure @xmath18 is homogeneous , and hence admits quantifier - elimination .
let @xmath388 be first - order definitions of the relations @xmath389 of @xmath12 , written in quantifier - free conjunctive normal form .
for a given @xmath0 , we compute a finite @xmath13-structure @xmath5 with domain @xmath387 $ ] as follows , where @xmath387 $ ] is viewed as a subset of @xmath390 . for each @xmath72 , let @xmath391 be the @xmath392-ary relation over the domain @xmath387 $ ] that contains all @xmath392-tuples that satisfy @xmath393 ( where @xmath392 is the number of free variables of @xmath393 ) .
since @xmath393 is of constant size in @xmath0 , the relation @xmath391 can be computed in time @xmath394 .
the resulting structure @xmath395 ; r^b_1,\dots , r^b_k)$ ] is clearly a substructure of @xmath12 .
moreover , if @xmath1 is a finite @xmath13-structure with @xmath0 elements , and @xmath396 is a homomorphism from @xmath1 to @xmath12 , then by high set - transitivity of @xmath12 there is an automorphism @xmath147 of @xmath12 such that @xmath397 $ ] .
hence , @xmath1 homomorphically maps to @xmath12 if and only if @xmath1 homomorphically maps to @xmath5 .
in this section we illustrate with a couple of examples the possibility of extending our tractability result beyond sub - exponential structures . all examples of @xmath65-categorical structures with totally symmetric polymorphisms of all arities that we are aware of have a first - order interpretation over the structure @xmath18 .
interpretations are a central concept from model theory , which we briefly recall in the following .
let @xmath276 and @xmath13 be relational signatures , @xmath56 a @xmath276-structure , and @xmath12 a @xmath13-structure .
@xmath398-dimensional ( first - order ) interpretation @xmath230 of @xmath12 in @xmath56 _ consists of ( cf .
@xcite ) 1 . a @xmath276-formula @xmath399 ; 2 . for each atomic @xmath13-formula @xmath400 , a @xmath276-formula @xmath401 , where @xmath402 are disjoint @xmath398-tuples of distinct variables ; and 3 .
a surjective map @xmath403 ; such that for all atomic @xmath13-formulas @xmath404 and all @xmath405 , @xmath406 holds in @xmath12 if and only if @xmath407 holds in @xmath56 .
our interest in interpretations also stems from the following result .
[ lem : interpretq ] every structure @xmath12 with a finite relational signature @xmath13 and a @xmath398-dimensional interpretation in @xmath18 has an efficient sampling algorithm . on input @xmath0 ,
the output of the algorithm is a structure of size at most @xmath408 .
let @xmath230 be a @xmath398-dimensional interpretation of @xmath12 in @xmath292 . on input @xmath0
, we compute a finite @xmath13-structure @xmath5 which is the induced substructure of @xmath12 on the domain @xmath409^d))$ ] . for an @xmath54-ary @xmath285
, we do the following .
let @xmath410 be the interpretation in @xmath292 of the atomic formula @xmath411 , given in quantifier - free conjunctive normal form .
we now evaluate @xmath410 on each sequence of @xmath398-tuples , @xmath412^d)$ ] .
the formula @xmath413 is true iff the tuple @xmath414 is in @xmath70 .
since @xmath404 is of constant size in @xmath0 , it follows that we can compute the relation @xmath70 in @xmath415 time .
the signature @xmath13 is finite , so there is a relation of highest arity , independent of @xmath0 , which provides the upper bound on the time complexity of the algorithm . next , let @xmath1 be a finite @xmath13-structure with @xmath0 elements , and assume that @xmath416 is a homomorphism .
this implies @xmath11 as well : the image of @xmath1 under @xmath417 has at most @xmath0 elements , @xmath418 .
let @xmath419 be tuples in @xmath420 such that @xmath421 , and let @xmath422 ) be the function such that @xmath423 for all @xmath72 .
note that @xmath419 contains at most @xmath424 distinct values of @xmath43 . by high set - transitivity of @xmath292
, it follows that there is an automorphism @xmath147 of @xmath292 such that @xmath425 $ ] for each @xmath72 and @xmath108 , where @xmath426 .
now for each relation symbol @xmath285 , we have @xmath427 iff @xmath428 holds in @xmath292 iff @xmath429 holds in @xmath292 , and this in turn is true if and only if @xmath430 .
it follows that @xmath431 is a homomorphism from @xmath1 to @xmath5 .
below , we give a number of examples of structures with exponential growth that are interpretable in @xmath18 , and have semi - lattice operations .
let ` @xmath176 ' , ` @xmath432 ' , and ` @xmath433 ' denote the usual inequality and equality relations on @xmath43 .
let @xmath434 be the relational structure over @xmath435 with binary relations @xmath436 for @xmath437 , where @xmath438 is defined by @xmath439 .
this structure is exponential . to obtain a lower bound on the growth rate of @xmath434 ,
pick an @xmath0-subset @xmath440 , and assume that the projection of @xmath1 on the second component contains exactly @xmath20 distinct values . for @xmath98 ,
let @xmath441 denote the number of elements that have the @xmath72th largest second component value .
then , @xmath442 determines a _ composition of @xmath0 _ ,
i.e. , an expression for @xmath0 as an ordered sum of positive integers . two sets @xmath443 which determine different compositions must be in different orbits . thus , the number of orbits of @xmath0-subsets of @xmath444 is at least @xmath445 , the number of compositions of @xmath0 . hence , theorem [ thm : yyy ] does not apply to @xmath434 . instead
, tractability can be inferred as follows .
the structure @xmath434 has a two - dimensional interpretation @xmath230 in @xmath18 .
the formula @xmath446 is always true , @xmath447 is the identity on @xmath435 , and the interpretations of @xmath3 and @xmath39 are as given above . furthermore , it is easy to verify that @xmath434 is invariant under the semi - lattice operations given by ( component - wise ) @xmath448 and @xmath449 .
tractability of @xmath450 now follows from theorem [ thm : correctness ] and lemma [ lem : interpretq ] . for any @xmath451
, this example can be generalized to a structure with @xmath452 @xmath398-ary relations , and with a @xmath398-dimensional interpretation in @xmath292 .
the csp of each such structure is polynomial - time solvable .
[ fig:1 ] ( -1,-1 ) rectangle ( 1,1.5 ) ; ( ) at ( 0,-0.5 ) @xmath453 ; ( c ) at ( 0.5cm,0 ) [ ] ; ( b ) at ( 0,0.75 cm ) [ ] edge[edge](c ) ; ( a ) at ( -0.5cm,0 ) [ ] edge[edge](c ) edge[edge ] ( b ) ; ( -1,-1 ) rectangle ( 1,1.5 ) ; ( ) at ( 0,-0.5 ) @xmath454 ; ( c ) at ( 0.5cm,0 ) [ ] ; ( b ) at ( 0,0.75 cm ) [ ] edge[edged](c ) ; ( a ) at ( -0.5cm,0 ) [ ] edge[edged](c ) edge[edged ] ( b ) ; ( -1,-1 ) rectangle ( 1,1.5 ) ; ( ) at ( 0,-0.5 ) @xmath455 ; ( c ) at ( 0.5cm,0 ) [ ] ; ( b ) at ( 0,0.75 cm ) [ ] edge[edge](c ) ; ( a ) at ( -0.5cm,0 ) [ ] edge[edged](c ) edge[edged ] ( b ) ; ( -1,-1 ) rectangle ( 1,1.5 ) ; ( ) at ( 0,-0.5 ) @xmath456 ; ( c ) at ( 0.5cm,0 ) [ ] ; ( b ) at ( 0,0.75 cm ) [ ] edge[edged](c ) ; ( a ) at ( -0.5cm,0 ) [ ] edge[edged](c ) edge[edge ] ( b ) ; the _ age _ of a structure @xmath12 is defined as the class of all finite structures isomorphic to a substructure of @xmath12 .
let @xmath457 be a signature with two binary relation symbols , and let @xmath458 be the set of structures in fig .
[ fig:1 ] , where the tuples of @xmath3 ( @xmath39 ) are given by the solid ( dashed ) arrows .
let @xmath459 be the class of all finite structures with signature @xmath457 for which every three - element substructure is isomorphic to a structure in @xmath262 .
it can be shown that @xmath459 is an _ amalgamation class _ so that its _ frass limit _ exists ( cf .
theorem 6.1.2 in @xcite ) .
this is the up to isomorphism unique countable homogeneous structure with age @xmath459 .
the following describes a relational structure @xmath460 with age @xmath459 which can be verified to be homogeneous .
it follows that @xmath460 is isomorphic to the frass limit of @xmath459 .
let @xmath461 , where @xmath462 is the relation @xmath463 , and @xmath464 is the relation @xmath465 . the growth rate of @xmath460 can be bounded as in the previous example , and here it turns out that the number of orbits of @xmath0-subsets is precisely @xmath445 .
the structure @xmath460 also has a two - dimensional interpretation in @xmath18 and semi - lattice polymorphisms given by ( component - wise ) @xmath448 and @xmath449 , so tractability follows once again from theorem [ thm : correctness ] and lemma [ lem : interpretq ] .
let @xmath466 be the following relational structure .
the domain @xmath467 is the disjoint union of two copies of @xmath43 .
the binary relation @xmath468 defines a perfect matching between the elements of @xmath48 and the elements of @xmath53 , and the binary relation @xmath176 defines a dense linear order on @xmath467 such that @xmath469 for all @xmath470 and @xmath471 , and for @xmath472 , we have @xmath473 iff @xmath474 for the elements @xmath177 with @xmath475 .
the structure @xmath476 is invariant under the semi - lattice operations given by @xmath448 and @xmath449 defined with respect to the order @xmath176 on @xmath467 .
it has two orbits and @xmath477 is isomorphic ( as an abstract group ) to @xmath239 . by theorem [ thm : no - interaction
] , this implies that @xmath476 does not have sub - exponential growth .
but @xmath476 has a 2-dimensional interpretation @xmath230 in @xmath292 , so @xmath478 is polynomial - time solvable : let @xmath479 be the formula @xmath480 , and let @xmath481 be the copy of @xmath338 in @xmath48 if @xmath482 and the copy of @xmath338 in @xmath53 if @xmath383 . the matching @xmath468 on @xmath483 and @xmath484
is interpreted by the formula @xmath485 and the order @xmath176 on @xmath483 and @xmath484 is interpreted by the formula @xmath486 .
every finite structure @xmath5 has an interpretation @xmath230 in @xmath292 , in fact , even in @xmath487 .
let @xmath488 and the dimension of the interpretation be @xmath489 .
the formula @xmath490 is true if and only if for exactly one @xmath72 it holds that @xmath491 .
equality is interpreted by the formula @xmath492 it is now straightforward to write down first - order formulas @xmath493 that interpret the relations of @xmath5 .
when @xmath3 is @xmath20-ary , then the formula @xmath493 is a disjunction of conjunctions with @xmath494 variables @xmath495 .
for each tuple @xmath496 , the disjunction contains the conjunct @xmath497 .
the examples in this section suggest the following question . [
quest : interpretations ] is it true that all @xmath65-categorical relational structures with totally symmetric polymorphisms of all arities have a first - order interpretation over @xmath292 ?
in this article we prove that constraint satisfaction problems for templates @xmath12 where the number of orbits of @xmath0-subsets of @xmath12 grows sub - exponentially in @xmath0 can be solved in polynomial time when @xmath12 has a semi - lattice polymorphism .
in fact , we showed the stronger result which only requires the existence of totally symmetric polymorphisms of all arities , instead of requiring the existence of a semi - lattice polymorphism .
this algorithmic result can be showed in two stages : 1 . in the first stage , we reduce csp@xmath498 , for structures @xmath12 with totally symmetric polymorphisms of all arities , to solving certain uniform finite domain csps , and to the task to find an efficient sampling algorithm for @xmath12 .
2 . in the second stage ,
we _ classify _ sub - exponential structures @xmath12 with totally symmetric polymorphisms of all arities , and use the classification to verify that there always exists an efficient sampling algorithm for @xmath12 .
the reduction presented in the first stage crucially relies on the fact that when @xmath12 has totally symmetric polymorphisms of all arities , then for all finite induced substructures @xmath39 of @xmath12 the set structure of @xmath39 homomorphically maps to @xmath12 ( lemma [ lem : totsymtohomo ] ) .
we want to remark that this connection has a converse when @xmath12 is @xmath65-categorical .
[ lem : omegaequiv ] let @xmath12 be an @xmath65-categorical structure over a finite relational signature .
then @xmath12 has totally symmetric polymorphisms of all arities if and only if @xmath88 for all finite substructures @xmath89 .
the forward direction was proved in lemma [ lem : totsymtohomo ] .
the remaining direction can be proved using a common technique for constructing homomorphisms to @xmath65-categorical structures .
given an arbitrary positive integer @xmath0 , we want to produce an @xmath0-ary totally symmetric polymorphism @xmath9 of @xmath12 .
the idea of the proof is as follows : let @xmath499 be an enumeration of the elements of @xmath27 , and let @xmath500 . for each @xmath79 ,
let @xmath501 be the set of homomorphisms from @xmath502)$ ] to @xmath12 .
introduce an equivalence relation @xmath503 on @xmath501 by defining @xmath504 iff @xmath505 for some automorphism @xmath147 of @xmath12 .
let @xmath506 denote the set of equivalence classes of @xmath501 under @xmath503 .
arrange the elements of @xmath507 into a forest containing at least one infinite tree : each @xmath508 is defined to be the root of a separate tree , and for each @xmath212 , and @xmath509 , define the parent of @xmath510 to be the equivalence class containing the restriction of @xmath511 to the non - empty subsets of @xmath512 .
this definition is independent of the choice of representative in @xmath513 , so each equivalence class in @xmath506 , @xmath212 , has precisely one parent .
since @xmath12 is @xmath65-categorical , it follows that there are finitely many equivalence classes for a fixed @xmath20 .
hence , there are finitely many trees and each tree is finitely branching in each node . by assumption , @xmath506 is non - empty for each @xmath79 , so some tree has unbounded height .
now , knig s tree lemma implies the existence of an infinite path @xmath514 in some tree .
assume that there are representatives @xmath515 , such that @xmath516 is the restriction of @xmath511 to the non - empty subsets of @xmath512 .
we show that this path can be extended indefinitely : choose @xmath517 arbitrarily and let @xmath518 be its restriction to the non - empty subsets of @xmath519 .
then , there exists an automorphism @xmath147 such that @xmath520 .
it follows that @xmath521 is the restriction of @xmath522 to the non - empty subsets of @xmath519 , hence we can define @xmath523 . now , for any @xmath0-tuple @xmath524 over @xmath21 , define @xmath525 , where @xmath54 is an any integer such that @xmath526 . by the construction of the sequence @xmath527 ,
the function @xmath9 is a well - defined totally symmetric @xmath0-ary function on @xmath21 . to verify that @xmath9 is a polymorphism of @xmath12 , let @xmath528 be an @xmath347-ary relation , and @xmath529 .
let @xmath530,\dots , t_n[i]\}$ ] , for @xmath531 .
assume without loss of generality that @xmath54 has been chosen large enough so that @xmath532 . then @xmath533,\dots , t_n[i ] ) = f_m(u_i)$ ] for all @xmath72 .
by definition of the set structure , we have @xmath534)}$ ] . since @xmath535 is a homomorphism ,
we conclude that @xmath536 , so @xmath9 is indeed a polymorphism . because of the general applicability of the algorithmic approach in section [ sect : alg ] , and because of the fact that the known structures of exponential growth that have totally symmetric polymorphisms seem to be well - behaved ( see section [ sect : beyond ] )
, we make the following conjecture .
|
a famous result by jeavons , cohen , and gyssens shows that every constraint satisfaction problem ( csp ) where the constraints are preserved by a _
semi - lattice operation _ can be solved in polynomial time .
this is one of the basic facts for the so - called _ universal - algebraic approach _ to a systematic theory of tractability and hardness in finite domain constraint satisfaction .
not surprisingly , the theorem of jeavons et al . fails for arbitrary infinite domain csps .
many csps of practical interest , though , and in particular those csps that are motivated by _ qualitative reasoning calculi _ from artificial intelligence , can be formulated with constraint languages that are rather well - behaved from a model - theoretic point of view . in particular , the automorphism group of these constraint languages tends to be _ large _ in the sense that the number of orbits of @xmath0-subsets of the automorphism group is bounded by some function in @xmath0 . in this paper
we present a generalization of the theorem by jeavons et al . to infinite domain csps where the number of orbits of @xmath0-subsets grows sub - exponentially in @xmath0 , and prove that preservation under a semi - lattice operation for such csps implies polynomial - time tractability . unlike the result of jeavons et al .
, this includes many csps that can not be solved by datalog .
= [ draw , circle , fill = black , text = white , minimum size=2pt , inner sep=0pt ] = [ -latex ] = [ -latex , dashed ]
|
living entities , on scales from birds to individual cells , organize in complex patterns with collective behaviors that serve important biological functions .
examples range from the flocking of birds @xcite to the sorting and organization of cells in morphogenesis @xcite .
work over the last ten years has shown that many aspects of this complex organization can be captured by physical models based on a minimal set of rules or interactions , leading to the emergence of the new field of active matter @xcite .
this is defined as a distinct category of non - equilibrium matter in which energy uptake , dissipation and movement take place at the level of discrete microscopic constituents .
the active matter paradigm has additionally inspired the development of ingenious synthetic chemical and mechanical analogues , such as `` active '' colloids : micron - size spheres partly coated with a catalyst that promotes the decomposition of one of the components of the ambient fluid , resulting in self - propulsion of the colloidal particles @xcite .
collections of such active synthetic particles have been shown to spontaneously assemble in coherent mesoscale structures with remarkable life - like properties @xcite .
active systems exhibit rich emergent behaviors , where a collection of many interacting entities shows large - scale spatial or temporal organization in states with novel macroscopic properties .
for instance , a dense swarm of bacteria can behave collectively as a living fluid with novel rheology @xcite , self - organize in complex regular patterns @xcite , exhibit turbulent motion @xcite , or ` freeze ' into a solid - like biofilm @xcite .
this type of behavior is of course well known in inert inanimate matter that exhibits transitions between different phases upon the tuning of an external parameter , such as temperature , or the application of external forces that perturb the system at its boundaries ( e.g. , shear stresses ) or globally ( e.g. , an electric or magnetic field ) .
it acquires , however , a new unexplored richness in active systems that are tuned out of equilibrium by energy generated internally by each unit .
the active matter paradigm aims at describing and classifying the behavior of this new class of non - equilibrium systems .
it does so by drawing on our understanding of familiar states of matter and of the transitions between them as controlled by interactions between atoms and molecules .
new states of matter arise when we put together many units that are individually driven or motile . how are these new states formed ? are they controlled solely by local interactions among the active particles or
is chemical signaling required to understand the emergence of these new states ?
can we classify and describe them and control the transitions between such states as we know how to do with familiar inert matter ?
[eq : thetai ] ) for @xmath0 and @xmath1 , reproduced from ref .
@xcite with permission from the royal society of chemistry .
the red region corresponds to a phase separated system .
the blue region corresponds to a glass as characterized by the behavior of the msd .
the glass would be replaced by a crystalline state in a monodisperse suspension .
the dotted line is the mean - field spinodal line given by @xmath2 from eq .
( [ eq : phis ] ) for @xmath1 . ]
recently , a number of ingenious synthetic systems have been engineered that show the emergent behavior of living active systems .
these include autophoretic colloids @xcite , rollers @xcite , and droplets @xcite .
the simplest realization of such `` colloidal microswimmers '' is obtained by immersing spherical janus colloids created by coating a hemisphere of a gold bead with platinum in a solution rich in hydrogen peroxide @xmath3 ) @xcite .
the difference in the consumption rate of @xmath4 at the gold and platinum sides maintains an asymmetric concentration of solute on the two hemispheres , resulting in propulsion of the particles along their symmetry axis . in this and
other catalytic colloidal swimmers interactions and propulsion can be tuned in a controlled way , allowing systematic studies up to moderate densities ( see fig .
one remarkable phenomenon shown by these systems is spontaneous assembly in macroscopic clusters @xcite .
this phenomenon is distinct from equilibrium assembly that arises from attractive interactions between the particles and has been shown to be driven by the nonequilibrium interplay of motility and crowding .
the nonequilibrium pressure equation of state of active colloids has been probed experimentally via sedimentation measurements @xcite , revealing a motility - induced effective adhesion that can strongly suppress the pressure at moderate density .
a review of recent experimental findings can be found in ref .
@xcite and is beyond the scope of the present article that focuses on minimal models of active colloids that have in many cases predicted and then qualitatively reproduced many of the experimental observations .
colloids have played a key role in condensed matter physics as model systems for atomic materials where pair interaction can be customized and equilibrium phase transitions and glassy behavior can be investigated with optical microscopy @xcite . although many active entities , from bacteria to birds , are elongated in shape and order in states with local or global liquid crystalline order leading to collective flocking , the growing body of work on synthetic active systems has shown that even spherical active particles can exhibit novel behaviors arising from the irreversible dynamics of each constituent , providing an excellent system for the quantitative testing of active matter theories .
this has led to extensive theoretical and numerical studies of minimal models consisting of self - propelled spheres with purely repulsive interactions known as active brownian particles ( abp ) @xcite .
these systems do not exhibit aligned or flocking states , but form active gases , liquids and solids , as summarized in the phase diagram shown in fig .
[ fig : pd ] .
the interplay between motility and steric effects is responsible for intriguing new phenomena , including motility - induced phase separation in the absence of any attractive interactions @xcite , casimir - type forces @xcite , and ratchet effects @xcite .
the complexity of behavior that arises in these minimal models is truly remarkable .
importantly , the abp model has demonstrated that many aspects of the emergent behavior of active systems do not require biochemical signaling , but are captured by physical contact interactions .
it has additionally provided an excellent playground for addressing fundamental questions about the non - equilibrium statistical mechanics of active systems and whether equilibrium - like notions , such as effective temperature or equations of state , may be useful to describe them . in this article
we review recent advances on the theoretical description of collections of active brownian particles ( abp ) defined as spherical self - propelled particles with purely repulsive interactions by organizing their behaviors in terms of the new active gases , liquids and solid phases formed by these systems .
the work described demonstrates that some key aspects of the emergent behavior of active systems , such as the tendency to spontaneously cluster in large compact structures and to accumulate at surfaces exerting organized forces on the environment , can be described in terms of the nonequilibrium interplay of motility and crowding , without invoking attractive interactions nor biochemical signaling .
we believe that minimal models of the type described here will continue to provide important tools for advancing our understanding of the nonequilibrium statistical mechanics of active matter .
the rest of this paper is organized as follows . in section [ model ] the minimal model of abp
is presented , emphasizing its parameters , limiting cases , and critical values .
next , in section [ idealgas ] , we consider the properties of active gases in the context of recent work characterizing their mechanical properties and defining a pressure equation of state . from
the dilute ideal active gas limit , tuning the rotational pclet number and increasing density beyond a critical value results in phase separation , where significant groups of active particles find their self - propulsion velocity caged due to increased interactions , yielding an active liquid coexisting with the active gas .
this has been characterized by equations of mean velocity and critical density , which are presented and supported by simulation .
this motility - induced phase separation ( mips ) is analyzed in section [ mips ] using continuum equations .
the high - density limit , discussed in section [ crystals ] , is one in which solids and glasses of active colloids are formed , and this is distinguished from the glasses or crystalline states that results from high density collections of passive colloids .
finally , a discussion and outlook is provided in section [ discussion ] , raising current questions regarding the interplay of noise and damping in the current models of active systems .
the rich behavior of active colloids has been studied using a minimal model of self - propelled particles ( spps ) that allows for both analytical and numerical progress . in this model
hydrodynamic interactions are neglected and the ambient fluid is assumed to only provide friction , rendering the dynamics overdamped .
each colloid is modeled as a spherical particle of radius @xmath5 , with an orientation defined by the axis of self - propulsion . in the following
we will discuss both monodisperse systems , where all disks have the same radius @xmath6 , and polydisperse systems , where @xmath6 will denote the mean radius and the radii are uniformly distributed with @xmath7 polydispersity . while most of the work on this model known in the literature as abp has been carried out in two dimensions ( @xmath8 ) , and we will restrict ourselves to this case here , many of the results described also hold in three dimensions @xcite . each particle is characterized by the position @xmath9 of its center and its orientation @xmath10 which in @xmath8 corresponds to a single angle @xmath11 .
the dynamics is then described by coupled langevin equations ( see figure [ fig : abp_sketches ] ) @xmath12 where @xmath13 is the active ( self - propulsion ) speed and @xmath14 the mobility .
the particles interact via short - range radial repulsive forces @xmath15 , where @xmath16 is a pair interaction that depends on the separation @xmath17 of the centers of the two particles .
both soft and hard repulsive forces have been used in the literature , and the qualitative behavior of the system is not affected by the details of the pair repulsion , provided it is pairwise , radial , and decoupled from the angular dynamics .
here we assume soft repulsive forces @xmath18 , with @xmath19 , @xmath20 , and @xmath21 if @xmath22 and @xmath23 otherwise .
we consider @xmath24 disks in an area @xmath25 , with @xmath26 the number density and @xmath27 the packing fraction
. along the direction @xmath28 defined by the angle @xmath11 it makes with the @xmath29 axis .
the angle @xmath11 is subject to angular noise determined by the diffusion rate @xmath30 , corresponding to a persistence time @xmath31 .
the jerky line that ends at the particle is a typical trajectory .
right : the pairwise repulsive interaction is proportional to the overlap @xmath32 . ]
the random force in eq .
( [ eq : ri ] ) describes thermal noise , with zero mean and correlations @xmath33 , where @xmath34 is the translational diffusion coefficient and @xmath35 the thermal energy .
the direction of self propulsion fluctuates due to non - thermal processes .
this is described by the random torque @xmath36 which is also chosen with zero mean and correlations @xmath37 , with @xmath38 the rotational diffusion rate , which is treated as an independent parameter because in many realizations , including bacterial suspensions @xcite and active colloids @xcite , the rotational noise is athermal .
it is instructive to note that eqs .
( [ eq : ri ] ) and ( [ eq : thetai ] ) can be combined into a single equation of the form @xcite @xmath39 where @xmath40 is a non - markovian noise with zero mean and correlations @xmath41 with @xmath42 the diffusion constant of the persistent motion . in other words
the active forcing on the right - hand side of eq .
( [ eq : ri ] ) is equivalent to non - markovian noise with a persistence time @xmath43 . this key property of abp arises because in this minimal model of active colloids the single - particle angular dynamics is decoupled from interactions and from the angular dynamics of other particles . as a result , abps do not exert torques on each other nor on the walls of a container .
this simplification has allowed substantial theoretical progress , but has important limitations discussed in section 6 .
note that eqs.[eq : ri ] and [ eq : xi ] are distinct from gaussian colored noise due to the bounded speed distribution here . in the limit @xmath44 , for fixed @xmath45
, the exponential decay can be approximated by a delta function , @xmath46 . in this limit ,
a single active brownian particle behaves like a thermal brownian particle with an effective temperature @xmath47 @xcite . in interacting abp the thermal limit is only realized , however , if the orientational correlation time @xmath48 is much smaller than all time scales present in the system . in particular , @xmath48 must be small compared to the mean free time between collisions , @xmath49 , and to the interaction time @xmath50 .
the dimensionless ratio @xmath51 is especially relevant for phase separation , whose onset has been interpreted in terms of a critical value of @xmath52 @xcite .
a recent study has also pointed out the crucial role played by @xmath52 in active suspensions , where hydrodynamic interactions are important @xcite .
we will see below that abp can be tuned through active gas , liquid and solid phases by tuning activity , as measured by a dimensionless rotational pclet number @xmath53 and the packing fraction @xmath54 , hence proportional to @xmath55 . here
we prefer to use the rotational pclet number , which is commonly used in the active matter literature and is a direct measure of persistence . ] .
it is useful to first examine the behavior of an ideal gas of spps . in the absence of interactions ,
each particle performs a persistent random walk .
the dynamics is characterized by the persistence time @xmath48 of the path and the associated persistence length @xmath56 .
the rotational pclet number also represents the ratio of the persistence length to the size of the particles , @xmath57 and provides a measure of persistence or activity .
the mean square displacement ( msd ) of a single spp can be calculated exactly and it is given by @xmath58 ^ 2\rangle=4d_tt+2{\mathrm{v}}_0 ^ 2\tau_\text{r}\left[t-\tau_\text{r}\left(1-\text{e}^{-t/\tau_\text{r}}\right)\right]\;,\ ] ] with @xmath59 .
the dynamics is ballistic for @xmath60 , with @xmath61 ^ 2\rangle\sim { \mathrm{v}}_0 ^ 2t^2 $ ] , and diffusive for @xmath62 , with @xmath61 ^ 2\rangle\sim 4[d_t+d_a]t$ ] , and @xmath63 . the same expression is obtained for run - and - tumble bacteria , with the replacement @xmath64 and @xmath65 the tumble rate , indicating that the coarse - grained dynamics is insensitive to whether changes in direction of self - propulsion are continuous or discrete .
a detailed comparison of rotational and run - and - tumble dynamics can be found in ref .
for both bacteria and active colloids the thermal diffusion coefficient @xmath66 is typically two orders of magnitudes smaller than the active diffusion @xmath67 .
this justifies neglecting thermal noise in eq .
( [ eq : ri ] ) .
recently there has been substantial theoretical effort towards characterizing the mechanical properties of an active gas and understanding whether this non - equilibrium system can be described by a pressure equation of state @xcite . in equilibrium
, the pressure @xmath68 of a fluid can be defined in three equivalent ways : ( i ) as the mechanical force per unit area on the walls of the container , ( ii ) from thermodynamics as the derivative of a free energy , and ( iii ) as the trace of the hydrodynamic stress tensor of the fluid , which in turn represents the momentum flux in the system . in equilibrium
all three definitions give the same expression and the pressure is a state function .
recent work by us and others has shown that the pressure is also a state function for an active fluid of spherical spps with purely repulsive interactions @xcite , but becomes dependent on the nature of the walls as soon as the active particles exert torques on each other or on the boundaries , as would for instance be the case for non - spherical spps @xcite . starting with the familiar virial expression @xcite , the contribution from interparticle interactions to the pressure @xmath69 of a passive fluid with pairwise forces @xmath70
can be written as the trace of the stress tensor @xcite ( see , e.g. , section 8.4 of @xcite ) , @xmath71 where @xmath72 is the area of the system and the brackets denote an average over the noise .
this expression , often referred to as irving - kirkwood formula @xcite , holds generally under the assumption of pairwise interactions and highlights the meaning of pressure as describing moment flux across a unit line ( or a unit plane in three dimensions ) . for passive gases in thermal equilibrium to this direct contribution from interaction one must add the momentum carried by free streaming particles , which gives the ideal gas pressure @xmath73 , with @xmath26 the number density . in a fluid of spps
there is an additional contribution to the pressure that describes the flux of self - propulsive force across a unit line .
this contribution , unique to active systems , was first identified in refs .
@xcite and was dubbed ` active ' or ` swim ' pressure .
we will use the latter term here that has now become accepted in the literature . in ref .
@xcite it was shown that the swim pressure can be written in a virial - type form as @xmath74 where @xmath75 is the propulsive force on each particle .
the total pressure of an active gas is then @xmath76 . in the remainder of this section
we will consider active systems at @xmath77 and ignore the thermal ideal gas contribution .
it was shown in ref .
@xcite that for repulsive self - propelled disks the pressure calculated as the sum of eqs .
( [ eq : pd ] ) and ( [ eq : ps ] ) is indeed identical to the force per unit area on the walls of the container , demonstrating that in this system the pressure is a state function . when the active units exchange torques among themselves and/or with the walls of the container , as is the case for non - spherical particles or in the presence of orientation - dependent interactions , the force on a wall depends on the properties of the wall and the definition of a mechanical state function that plays the role of pressure does not seem possible @xcite .
in the absence of interactions @xmath78 and one can readily calculate the pressure @xmath79 of an active ideal gas using eq .
( [ eq : ps ] ) and assuming that the initial position of the particle is not correlated with the noise , with the result @xmath80 . in the limit of large systems
, one can let @xmath81 in this expression to obtain the pressure of an ideal active gas in the thermodynamic limit , @xmath82 in other words the swim pressure of an ideal active gas is simply the pressure of an ideal gas at the temperature @xmath83 .
although eq .
( [ eq : pideal ] ) replaces the kinetic contribution to the pressure that one would have in a thermal passive gas , we will see below that , unlike the ideal gas pressure of thermal systems , this contribution is strongly renormalized by interactions .
additionally , active systems are much more sensitive to finite size effects and care is required when taking the thermodynamic limit .
if the persistence length @xmath84 is comparable to or exceeds the linear size @xmath85 of the container , then particles travel ballistically and one can estimate the pressure as @xmath86 where the last approximate equality holds when @xmath87 . for high activity , the active gas behaves like a highly diluted knudsen gas and the pressure depends on the size of the container .
it has also been shown recently that boundary curvature can strongly affect the density and force distribution in abp @xcite .
care must therefore be taken in numerical simulations as the mapping of a dilute active gas of purely repulsive spherical colloids onto a thermal gas at an effective temperature @xmath88 only holds when @xmath89 .
strong finite - size effects are observed when @xmath84 is a fraction of @xmath85 . in this case
active gases behave quite differently from thermal ones : they accumulate at the walls of the container @xcite or around fixed impurities , exert casimir - type forces @xcite , and spontaneously sort when different in size or activity @xcite .
pclet numbers in excess of @xmath90 , corresponding to persistence lengths of order @xmath90 particle sizes , are easily reached in experiments in active colloids @xcite and finite size effects should be observable in microfluidic devices . at higher density ,
interactions become important .
it is instructive to note that the kinetic contribution to the pressure of thermal brownian particles can also be calculated through a virial - type expression akin to the one defining the swim pressure of abp , given by @xmath91 , with @xmath92 the thermal noise .
assuming , as customary , that the initial position of the particles is not correlated with the noise , and making use of the fact that the noise is uncorrelated in time , this expression gives exactly @xmath93 at all densities @xmath94 .
in contrast , the stochastic propulsive force entering the definition of the swim pressure , eq .
( [ eq : ps ] ) , is correlated in time ( see eq .
( [ eq : xi ] ) ) , leading to the dependence of the swim pressure on interparticle forces . the total pressure of a monodisperse system of repulsive spps is shown in fig .
[ fig : pressure](a ) as a function of packing fraction for several values of the persistence length . at low density
the system behaves like a gas of ` hot ' repulsive colloids with effective temperature @xmath83 , as shown in the inset . at intermediate density
the swim pressure is strongly suppressed due to collisional slowing down ( fig .
[ fig : pressure](c ) ) and the total pressure remains almost constant over a range of packing fractions .
this is the region where the system spontaneously phase separates , as discussed below , and a fluid of repulsive spps behaves qualitatively like a passive fluid of attractive particles . at high density
the swim pressure is negligible and repulsive interactions dominate , yielding the sharp increase of the pressure associated with the transition to a glassy or solid state , as in passive systems .
the pressure of active janus colloids was recently measured in sedimentation experiments @xcite , confirming the prediction of the simulations ( see fig .
[ fig : pressure](b ) ) .
these experiments also confirmed that dilute active gases behave like ` hot ' passive colloids with an effective temperature enhanced by activity and that this mapping holds even in the presence of interactions up to moderate densities , but below the region of spontaneous phase separation .
it was shown in ref .
@xcite that in the thermodynamic limit the swim pressure given in eq .
( [ eq : ps ] ) can be recast in an intuitive form , given by @xmath95 where @xmath96 is the mean velocity of the particles along the direction of self - propulsion , given by @xmath97 .
we note that this expression only holds if @xmath89 and in fact reduces to the bulk pressure of an active ideal gas , @xmath98 , in the absence of interactions . the suppression of the swim pressure due to collisional slowing down
can then be modeled phenomenologically at moderate densities as a linear decay of the propulsion speed , with @xmath99 for @xmath100 and @xmath101 for @xmath102 . in general @xmath103
will depend on @xmath104 .
this linear decay of @xmath96 with @xmath94 has been seen in simulations @xcite and justified by relating the linear decay rate to the pair correlation function @xcite and by estimating it via a kinetic argument @xcite as arising from the fact that at finite density particles can be stalled for the duration @xmath105 of collisions with other particles . at low density , where the mean free time between collisions @xmath106 exceeds @xmath105 , the mean speed
can then be written as @xmath107 .
the collision time @xmath105 was treated as a constant fitting parameter in ref .
@xcite , but will in general depend on pclet number .
it can be estimated as controlled by two delay mechanisms : the time @xmath108 it takes two interacting particles to move around each other and the reorientation time @xmath109 . since the collision time @xmath105 will be controlled by the faster of these two processes , we add the two rates to obtain @xmath110 .
this gives @xmath111 $ ] , with @xmath112 where @xmath113 is a number of order unity .
in our simulations ( see fig . [
fig : pressure ] ) , we find that this expression works well for @xmath114 , with @xmath115 .
the swim pressure for moderately dense gases can then be written as @xmath116 $ ] . at low density ,
the direct part of the pressure can be estimated as @xmath117 , where @xmath118 is the force scale and @xmath119 is the typical displacement of a particle between collisions @xcite .
this expression holds as long as the particles do not overlap .
the most striking phenomenon exhibited by our minimal model of active colloids is motility - induced phase separation ( mips ) , where an active fluid of particles with purely repulsive interactions spontaneously phase separates into a dense liquid surrounded by an active gas @xcite .
this remarkable phenomenon arises from the collisional slowing down of the active particles speed @xmath96 that yields the suppression of the swim pressure shown in fig .
[ fig : pressure](c ) , combined with the fact that the lack of detailed balance allows self - propelled particles to accumulate in the regions where @xmath96 ) is small @xcite . while experiments often see strong clustering , but not complete phase separation @xcite , mips has now been seen in numerical simulations for a variety of repulsive interaction , ranging from soft harmonic potential to wca and hard spheres , in both two and three dimensions @xcite .
the phase behavior of the system , as obtained from simulations of the polydisperse limit of the model described in section [ model ] , is summarized in fig .
[ fig : pd ] .
three regimes can be identified : a homogeneous fluid phase , a solid or frozen state at high packing fraction and relatively low activity , and a regime where the system shows macroscopic phase separation into two bulk phases . the phase separated regime exists in an intermediate range of packing fraction and activity , corresponding to the region where the pressure remains approximately constant with increasing packing fraction . for soft repulsion
, it changes continuously from a high density liquid cluster surrounded by a gas at lower packing fractions , to a `` hole '' of gas phase inside a densely packed liquid at higher packing fractions @xcite .
the existence of a homogeneous liquid phase intermediate between the solid and the phase separated state seems to be a generic feature , but is not fully understood .
in contrast , the suppression of phase separation at high self - propulsion speed is specific to the soft nature of the repulsive potential used in our simulations , where for @xmath120 particles can just pass through each other .
the threshold @xmath121 for phase separation has been estimated in various ways in the literature .
a naive estimate is obtained by assuming that phase separation occurs when @xmath122 , corresponding to the condition that a particle experiences many collisions before changing direction , i.e. , it gets trapped by other particles , with the result @xmath123 .
the same estimate was obtained by redner _
@xcite via a kinetic argument that equates the flux @xmath124 of particles that enter the cluster to the flux out of the cluster , @xmath125 , with @xmath126 a fitting parameter .
the dependence on @xmath38 arises because particles must turn around to point outward before they can leave the cluster . by equating these fluxes
one can estimate the density of the gas as @xmath127 and calculate the fraction @xmath128 of particles in the cluster as a function of @xmath94 and @xmath104 . the condition @xmath129 gives a criterion for the phase separation , corresponding to @xmath130 .
these naive estimates predict that for arbitrarily large @xmath104 phase separation will occur no matter how low the density .
this behavior is not , however , born out by simulations that find a minimum threshold of density below which there is no phase separation . and @xmath131 .
the parameters are @xmath132 , so @xmath50 is the time unit , @xmath133 and @xmath134 . the rotational diffusion has been set to give a pclet number of @xmath135 . ]
another approach is to use continuum equations for an active fluid where motility suppression is incorporated through the density - dependent propulsion speed @xmath96 that controls the swim pressure @xcite , and then employ linear stability analysis to locate the spinodal .
the dynamics is governed by equations for the density @xmath136 and the polarization density @xmath137 that describes the local orientation of the particles axis of self - propulsion , given by @xcite @xmath138\;,\\ \label{p } & & \partial_t{\bm p}=-d_\text{r}{\bm p}-\bm\nabla\pi(\rho)+k\nabla^2{\bm p}\ ; , \end{aligned}\ ] ] where @xmath139 , with @xmath140 the pressure . for times
@xmath141 , one can neglect the time derivative of the polarization in eq .
( [ p ] ) relative to the damping term and eliminate @xmath142 to obtain a nonlinear diffusion equation for the density , given by @xmath143\;,\ ] ] with effective diffusivity @xmath144 the linear stability of a homogeneous state of constant density @xmath145 can be analyzed by examining the dynamics of density fluctuations , @xmath146 .
the decay of fluctuations is controlled by a diffusive mode with diffusion constant @xmath147 that can become negative , signaling the instability of the uniform state .
the onset of the instability corresponds to @xmath148 , which determines the spinodal line .
when @xmath1 the threshold for phase separation in @xmath149 .
this expression does yield a minimum density below which there is no phase separation , as seen in simulations @xcite , but it fails to account for the observed lower bound for the value of @xmath104 required for phase separation .
a lower bound is obtained when @xmath150 . in this case one
finds @xmath151 where @xmath152 represents the minimum value of @xmath104 required for phase separation .
thermal diffusion suppresses phase separation and drives this critical @xmath104 to larger values .
the condition @xmath153 for phase separation can also be written as @xmath154 , indicating that the observed spinodal decomposition requires a sufficiently small @xmath38 or long persistence time .
the expression given in eq .
( [ eq : phis ] ) with the choice of minus sign corresponds to the dotted line in fig .
[ fig : pd ] .
it is , however , unclear whether thermal diffusion is really the mechanism responsible for the lower bound on @xmath104 , given the simulations leading to fig .
[ fig : pd ] were carried out with @xmath1 and do show a finite threshold .
although a finite value of @xmath66 can be generated by interactions even if @xmath1 at the single particle level in the numerical model , this discrepancy suggests that other mechanisms not yet understood may be at play here . for finite @xmath38
one must in general retain the dynamics of the polarization .
the analysis of the modes in this case can be found in ref .
@xcite and yields a growth rate with the characteristic behavior expected for a spinodal , with a maximum growth rate at wavevector @xmath155 . by including noise in the continuum model , one can examine the spectrum of fluctuations and evaluate the structure factor @xmath156 . at small wavevector
one finds @xmath157 , with @xmath158^{-1/2}$ ] a correlation length that diverges on the spinodal line @xcite .
both the divergence of @xmath159 at @xmath160 and the scaling @xmath161 are consistent with the behavior at an equilibrium gas - liquid spinodal line .
although the numerics can be fitted to this behavior @xcite , the divergence of @xmath162 is generally masked by nucleation processes that further support the liquid - gas analogy @xcite .
a true divergence is expected at the critical point at @xmath163 , but it has not yet been possible to extract numerical values for the critical exponents . it has also been shown numerically that the coarsening dynamics closely resembles that of an equilibrium spinodal decomposition and that activity weakly affects on the growth in time of the domains size @xmath164 that is found to be very close to the cahn - hilliard form , @xmath165 , describing the coarsening of a passive system with no momentum conservation have been reported in simulations @xcite , but it is not clear whether these simulations had indeed reached the asymptotic regime . ] although differences exist when gradient terms accounting for interfacial phenomena are included in the theory @xcite .
in contrast , experiments in bacteria and active colloids generally do not see complete separation in two bulk phases , but rather observe the assembly of finite - size clusters that persist for very long times . in bacterial suspensions birth / death events that eliminate density conservation have been shown to arrest the phase separation yielding the formation of finite - size clusters @xcite .
attractive interactions have also ben suggested as a possible mechanism for arresting the phase separation @xcite , but the mechanism that control the size of finite clusters in active colloids remain to be understood . the @xmath166 divergence of the structure factor as @xmath167 is also the signature of the giant number fluctuations ( gnf ) predicted and observed ubiquitously in the ordered state of flocking models @xcite .
mips and gnf are , however , two distinct phenomena . in repulsive spps
the large density fluctuations occur in a disordered state and are associated with a critical point and phase separation .
the large density fluctuations that occur in the ordered state of active fluids with both polar and nematic order are the consequence of the coupling of a conserved field - the density - to a non - conserved noisy polarization field .
the @xmath166 behavior of the correlation is in fact a generic feature of systems where the deterministic dynamics of a conserved field couples to non - conserved noise , resulting in power - law decay of correlations @xcite , and is also ubiquitous in driven granular materials with inelastic interactions @xcite .
the onset of mips suggests that at moderate density a system of repulsive active colloids behaves like a thermal system with attractive interactions .
this correspondence was recently demonstrated by correlating the pressure measured in sedimentation experiments of active colloids with numerical simulations @xcite .
it has additionally been explored in models of brownian particles with gaussian colored noise correlated over a time @xmath48 , where the colored nature of the noise can be incorporated as an effective interaction in a many - body fokker - planck equation @xcite . using this approach
it has been shown recently that for weakly persistent motion ( small @xmath48 ) activity can be recast in the form of an effective pair potential with an attractive part of strength that increases with increasing @xmath104 @xcite .
on the other hand , when the pair interparticle potential actually has an attractive component , activity can yield a weak repulsive contribution to the effective potential , hence competing with attraction by promoting particle escape from the potential well .
in other words , self - propulsion can actually drive a system that has undergone bulk phase separation due to attractive interactions back into a uniform state .
it should be noted , however , that the abp model and the model with gaussian colored noise are characterized by different speed distributions , which is gaussian in the latter , but bounded by @xmath13 in abp and in run - and - tumble models .
simulations have additionally demonstrated that the competition of attraction and activity yields an intermediate regime where complete phase separation is arrested and particles assemble in a gel - like structure that has been seen in simulations in both two and three dimensions @xcite finally , it has been shown that hydrodynamic interactions can suppress phase separation by aligning the particles and therefore suppressing the self - trapping responsible for cluster formation @xcite .
on the other hand , strong clustering is often observed in active systems even in the presence of aligning rules @xcite , indicating that more work needs to be done to understand the effect of aligning torques on the mips of repulsive active particles .
at large packing fraction abp have been shown to freeze in solid states .
while this may not seem surprising since purely repulsive systems are known to form crystalline and glassy states in equilibrium , it is not a priori obvious that such states can survive in the presence of continuous energy input at the microscopic scale .
key to the formation of active solid states is the fact that steric repulsion strongly suppresses motility at high density , as discussed in section [ mips ] .
this results in particle caging and in what was first described by henkes and collaborators as ` active jamming ' @xcite , although that work was for a model of repulsive spps with alignment interactions , where confinement was required to impede collective flocking . in equilibrium freezing in two dimensions
is a continuous , two - step process , where the system goes from a liquid to a hexatic liquid state with quasi - long range bond orientational order , and then to a crystalline state with long - range orientational order and quasi - long range translational order .
crystallization can be quantified using both structural and dynamical criteria .
the onset of translational order identified via structural probes generally coincides with criteria for dynamical arrest , such as the vanishing of the long - time diffusion coefficient @xmath168 ^ 2\rangle$ ] .
crystalline order has been seen in phase separating active systems both in experiments and simulations @xcite . in simulations of monodisperse systems
the dense phase in the phase separated regime exhibits substantial structural order resembling a colloidal crystal near the hexatic - crystal transition , but with unusual super - diffusive behavior @xcite .
crystalline clusters have also been seen in the experiments , such as the ` living crystals ' formed by light activated colloids engineered by palacci _
et al . _
numerical simulations have also examined the onset of bond - orientational order in abp @xcite and its correlation with structural arrest , concluding that the onset of order remains continuous , but is shifted to higher density that in equilibrium systems @xcite . additionally , bialke _ et al . _
@xcite reported the existence of an intermediate region between the liquid and the solid states characterized by large structural fluctuations .
it is tempting to identify this phase an an active hexatic phase similar to but distinct from the equilibrium hexatic that has been observed for instance in superparamagnetic colloidal particles confined to an air - water interface @xcite .
finally , in models with aligning interactions the interplay of activity , alignment and structural order can lead to partially ordered liquid crystalline resting and traveling states that are beyond the scope of the present article @xcite .
our own work has focused on polydisperse systems , where even in equilibrium crystalline order is bypassed in favor of glassy or jammed states .
[ fig : msd](a ) shows a simulation snapshot near the active glass transition for a non - aligning , polydisperse system . in this case
the characterization of glassy behavior relies on criteria for dynamical arrest , such as the strong suppression of diffusion at intermediate times , when particles are persistently caged .
[ fig : msd](b ) shows the evolution of the mean square displacement of a polydisperse assembly of repulsive spps over a range of packing fractions . at small and intermediate packing fraction
the long - time dynamics remains diffusive , but at high packing fraction the msd is bounded over the time of the simulation , indicating a jammed or frozen state . the fluctuations in the actively jammed state ,
see fig . [
fig : msd](a ) , are intimately connected to the low frequency or ` soft ' modes of a solid that are associated with the jamming transition of passive thermal systems @xcite . at small @xmath13
active systems form solid phases that can be analyzed in terms of a complete set of normal modes with normal frequencies @xmath169 determined by the inter - particle interactions @xcite . as in the familiar description of lattice vibrations in terms of phonons ,
the lowest frequencies correspond to large scale collective structural rearrangements of the system .
work on passive granular matter has shown that such low frequency modes have much lower energy barriers than high frequency modes @xcite .
the jammed state is then a steady state with structural rearrangements involving a handful of low frequency modes that also provide easy paths for unjamming as @xmath13 is increased .
this hints at a generic connection between the dynamics of jammed active systems at low @xmath38 and the kinetic arrest of passive glassy systems , as explored in refs .
note that the prominence of low frequency modes in the active jamming dynamics is strongly linked to the polarization - velocity alignment mechanism , and more work is necessary to disentangle the interplay of activity , alignment and caging . from our simulations of polydisperse
abp with soft repulsive interactions , eqs .
( [ eq : ri][eq : thetai ] ) , for @xmath170 showing ballistic growth at small times and the crossover from diffusive to bounded behavior with increasing packing fraction at long time .
the dashed lines have slopes @xmath171 and @xmath172 . ] one interesting feature of the onset of glassy behavior in repulsive spps is the shift of the glass transition to packing fraction higher than the value @xmath173 obtained for thermal systems in the limit of vanishing motility , @xmath174 @xcite .
this shift increases with increasing persistence time @xmath48 and has been reported in both simulations of spps with hard repulsions where activity pushes the glass transition towards the random close packed limit of @xmath175 and in our simulations of soft disks ( see phase diagram of fig .
[ fig : pd ] ) , where we find @xmath176 for @xmath177 .
finally , ni _ et al . _ have shown numerically that doping a hard sphere glass with a small number of active particles promotes crystallization bypassing the glass transition @xcite .
this suggests that a small concentration of active particles may be used as a means for facilitating the formation of large crystalline states that would be otherwise hindered by defects and grain boundaries .
collections of repulsive self - propelled particles , also known as active brownian particles , provide a minimal realization of an active system with rich non - equilibrium behavior .
they exhibit active gas , liquid and solid phases with novel mechanical properties . at low density
they seemingly behave as ` hot ' colloids with an effective temperature controlled by activity ( but see ref .
@xcite ) and a novel swim pressure unique to active systems . at intermediate density
they resemble attractive colloids and exhibit a motility - induced phase separation . at high density
they form solid and glassy phases that have provided new insights in the physics of glassy and jammed solids .
mapping abp s onto an `` equivalent '' equilibrium system has proven a powerful method , as evidenced by the concepts of effective temperature and effective attraction .
such mappings , however , are necessarily limited in scope .
activity - induced attraction is antagonistic to actual attraction @xcite .
the hot colloid model fails in the presence of a rapidly varying external potential , e.g. , a stiff wall @xcite . as a result ,
the application of these concepts is sometimes subject to controversy and one should keep in mind that active systems are fundamentally nonequilibrium .
the key property that has allowed much theoretical progress in the description of this minimal active system is that the noisy single - particle angular dynamics is decoupled from interactions .
in other words active brownian colloids do not exert torques on each other nor on the walls of a container . as a consequence
the effect of activity can be mapped onto a non - markovian noise of strength controlled by the self - propulsion speed and with time correlation given by the persistence time @xmath48 , as shown in eqs .
( [ eq : ri - one][eq : xi ] ) .
this has led to renewed interest in stochastic models of brownian particles with gaussian colored noise , for which exact results exist in the literature @xcite .
although these models are not precisely equivalent to abp due to the different speed distributions , useful insight has been obtained from these studies @xcite . on the other hand , torques are likely to be present in most experimental realizations , as they can arise from the non - spherical shape of the particles , from lateral friction , from hydrodynamic interactions or explicit aligning rules , and will generally modify the collective behavior of the systems .
it has already been shown , for instance , that the pressure is a state function only for the case of repulsive spherical active brownian particles , independent of the nature of the interaction of the system with the walls .
the presence of torques may also arrest the bulk phase separation , leading instead to clustering or micro - phase separation , as ubiquitously seen in experiments .
work remains to be done to fully understand how alignment , induced either by anisotropic steric effects or by aligning interactions , modifies the phase behavior of active colloids and to make _ quantitative _ contact with experiments . in the limit of small persistence time ( @xmath178 ) , active brownian colloids are equivalent to brownian colloids .
the opposite limit of @xmath179 is intriguing and has not been fully explored . in this limit
the active drive becomes correlated in time and resembles quenched disorder . at high density
the fluctuations patterns tend to exhibit both spatially and temporally long - ranged correlations @xcite and resemble those seen in sheared passive granular matter @xcite , suggesting an intriguing connection that needs to be further investigated . finally ,
quite unexplored is the effect of noise in the continuum dynamics as described for instance by eqs .
( [ rho ] ) and ( [ p ] ) .
most work using phenomenological continuum models to describe active matter has focused on deterministic equations or has included white gaussian noise in the linear theory to evaluate correlation functions @xcite . on the other hand
, the noise in active systems is in general multiplicative in the density @xcite and may , for instance , provide a mechanisms for selecting nonlinear dynamical states .
the minimal model of active brownian colloids may provide an excellent playground for beginning the exploration of the role of multiplicative noise in mesoscopic theories of active matter .
mcm thanks xingbo yang and lisa manning for their contribution to some aspects of the work reviewed here and for fruitful discussions .
mcm was supported by nsf - dmr-305184 .
mcm and ap acknowledge support by the nsf igert program through award nsf - dge-1068780 .
mcm , ap and dy were additionally supported by the soft matter program at syracuse university .
ap acknowledges use of the syracuse university htc campus grid which is supported by nsf award aci-1341006 .
yf was supported by nsf grant dmr-1149266 and the brandeis center for bioinspired soft materials , an nsf mrsec , dmr-1420382 .
10 url # 1`#1`urlprefixhref # 1#2#2 # 1#1 m. ballerini , n. cabibbo , r. candelier , a. cavagna , e. cisbani , i. giardina , v. lecomte , a. orlandi , g. parisi , a. procaccini , m. viale , v. zdravkovic , interaction ruling animal collective behavior depends on topological rather than metric distance : evidence from a field study , proceedings of the national academy of sciences 105 ( 4 ) ( 2008 ) 12321237 .
http://dx.doi.org/10.1073/pnas.0711437105 [ ] .
( * ) m. c. marchetti , j. f. joanny , s. ramaswamy , t. b. liverpool , j. prost , m. rao , r. a. simha , hydrodynamics of soft active matter , rev .
( 2013 ) 11431189 .
http://dx.doi.org/10.1103/revmodphys.85.1143 [ ] .
j. r. howse , r. a. l. jones , a. j. ryan , t. gough , r. vafabakhsh , r. golestanian , self - motile colloidal particles : from directed propulsion to random walk , phys .
99 ( 2007 ) 048102 . http://dx.doi.org/10.1103/physrevlett.99.048102 [ ] .
j. palacci , b. abcassis , c. cottin - bizonne , c. ybert , l. bocquet , colloidal motility and pattern formation under rectified diffusiophoresis , phys .
104 ( 2010 ) 138302 .
http://dx.doi.org/10.1103/physrevlett.104.138302 [ ] . w. c. k. poon , from clarkia to escherichia and janus : the physics of natural and synthetic active colloids , in : c. bechinger , f. sciortino , p. ziherl ( eds . ) , proceedings of the international school of physics enrico
fermi course clxxxiv physics of complex colloids , ios press , 2013 , pp .
317. http://dx.doi.org/10.3254/978-1-61499-278-3-317 [ ] .
( * ) j. palacci , s. sacanna , a. p. steinberg , d. j. pine , p. m. chaikin , living crystals of light - activated colloidal surfers , science 339 ( 6122 ) ( 2013 ) 936940 . http://dx.doi.org/10.1126/science.1230020 [ ] .
r. soto , r. golestanian , http://link.aps.org/doi/10.1103/physreve.91.052304[self-assembly of active colloidal molecules with dynamic function ] , phys .
e 91 ( 2015 ) 052304 . http://dx.doi.org/10.1103/physreve.91.052304 [ ] .
http://link.aps.org/doi/10.1103/physreve.91.052304 c. dombrowski , l. cisneros , s. chatkaew , r. e. goldstein , j. o. kessler , self - concentration and large - scale coherence in bacterial dynamics , phys .
93 ( 2004 ) 098103 . http://dx.doi.org/10.1103/physrevlett.93.098103 [ ] .
l. hall - stoodley , j. w. costerton , p. stoodley , http://www.nature.com/nrmicro/journal/v2/n2/abs/nrmicro821.html[bacterial biofilms : from the natural environment to infectious diseases ] , nature reviews microbiology 2 ( 2 ) ( 2004 ) 95108 .
http://www.nature.com/nrmicro/journal/v2/n2/abs/nrmicro821.html p. j. lu , d. a. weitz , http://dx.doi.org/10.1146/annurev-conmatphys-030212-184213[colloidal particles : crystals , glasses , and gels ] , annual review of condensed matter physics 4 ( 1 ) ( 2013 ) 217233 .
http://dx.doi.org/10.1146/annurev-conmatphys-030212-184213 [ ] .
i. theurkauff , c. cottin - bizonne , j. palacci , c. ybert , l. bocquet , dynamic clustering in active colloidal suspensions with chemical signaling , phys .
( 2012 ) 268303 .
http://dx.doi.org/10.1103/physrevlett.108.268303 [ ] .
a. bricard , j .- b .
caussin , n. desreumaux , o. dauchot , d. bartolo , emergence of macroscopic directed motion in populations of motile colloids , nature 503 ( 2013 ) 9598 .
http://dx.doi.org/10.1038/nature12673 [ ] .
j. bialk , t. speck , h. lwen , active colloidal suspensions : clustering and phase behavior , journal of non - crystalline solids 407 ( 2015 ) 367 375 , 7th idmrcs : relaxation in complex systems .
[ ] .
m. b. wan , c. j. olson reichhardt , z. nussinov , c. reichhardt , rectification of swimming bacteria and self - driven particle systems by arrays of asymmetric barriers , phys .
( 2008 ) 018102 .
http://dx.doi.org/10.1103/physrevlett.101.018102 [ ] .
j. stenhammar , d. marenduzzo , r. j. allen , m. e. cates , http://xlink.rsc.org/?doi=c3sm52813h[phase behaviour of active brownian particles : the role of dimensionality ] , soft matter 10 ( 10 ) ( 2014 ) 14891499 . http://dx.doi.org/10.1039/c3sm52813h [ ] .
http://xlink.rsc.org/?doi=c3sm52813h ( * ) f. ginot , i. theurkauff , d. levis , c. ybert , l. bocquet , l. berthier , c. cottin - bizonne , nonequilibrium equation of state in suspensions of active colloids , phys .
x 5 ( 2015 ) 011004 . http://dx.doi.org/10.1103/physrevx.5.011004 [ ] .
a. solon , m. cates , j. tailleur , active brownian particles and run - and - tumble particles : a comparative study , the european physical journal special topics 224 ( 7 ) ( 2015 ) 12311262 .
http://dx.doi.org/10.1140/epjst/e2015-02457-0 [ ] .
s. a. mallory , a. ari , c. valeriani , a. cacciuto , anomalous thermomechanical properties of a self - propelled colloidal fluid , phys .
e 89 ( 2014 ) 052303 . http://dx.doi.org/10.1103/physreve.89.052303 [ ] .
a. p. solon , j. stenhammar , r. wittkowski , m. kardar , y. kafri , m. e. cates , j. tailleur , pressure and phase equilibria in interacting active brownian spheres , phys .
114 ( 2015 ) 198301 .
http://dx.doi.org/10.1103/physrevlett.114.198301 [ ] .
( * ) a. p. solon , y. fily , a. baskaran , m. e. cates , y. kafri , m. kardar , j. tailleur , pressure is not a state function for generic active fluids , nat
. phys . 11
( 2015 ) 673678 .
j. h. irving , j. g. kirkwood , the statistical mechanical theory of transport processes .
iv . the equations of hydrodynamics , the journal of chemical physics 18 ( 6 ) ( 1950 ) 817829 . http://dx.doi.org/http://dx.doi.org/10.1063/1.1747782 [ ] .
y. fily , a. baskaran , m. f. hagan , http://xlink.rsc.org/?doi=c4sm00975d[dynamics of self - propelled particles under strong confinement ] , soft matter 10 ( 30 ) ( 2014 ) 5609 .
http://dx.doi.org/10.1039/c4sm00975d [ ] .
http://xlink.rsc.org/?doi=c4sm00975d ( * ) i. buttinoni , j. bialk , f. kmmel , h. lwen , c. bechinger , t. speck , dynamical clustering and phase separation in suspensions of self - propelled colloidal particles , phys .
lett . 110 ( 2013 ) 238301 .
http://dx.doi.org/10.1103/physrevlett.110.238301 [ ] .
j. stenhammar , a. tiribocchi , r. j. allen , d. marenduzzo , m. e. cates , continuum theory of phase separation kinetics for active brownian particles , phys .
( 2013 ) 145702 .
[ ] .
m. e. cates , j. tailleur , when are active brownian particles and run - and - tumble particles equivalent ?
consequences for motility - induced phase separation , epl ( europhysics letters ) 101 ( 2 ) ( 2013 ) 20010 .
m. e. cates , d. marenduzzo , i. pagonabarraga , j. tailleur , arrested phase separation in reproducing bacteria creates a generic route to pattern formation , proceedings of the national academy of sciences 107 ( 26 ) ( 2010 ) 1171511720 .
http://dx.doi.org/10.1073/pnas.1001994107 [ ] .
r. aditi simha , s. ramaswamy , hydrodynamic fluctuations and instabilities in ordered suspensions of self - propelled particles , phys .
( 2002 ) 058101 .
http://dx.doi.org/10.1103/physrevlett.89.058101 [ ] .
g. grinstein , d .- h .
lee , s. sachdev , conservation laws , anisotropy , and `` self - organized criticality '' in noisy nonequilibrium systems , phys .
64 ( 1990 ) 19271930 . http://dx.doi.org/10.1103/physrevlett.64.1927 [ ] .
r. matas - navarro , r. golestanian , t. b. liverpool , s. m. fielding , hydrodynamic suppression of phase separation in active suspensions , phys .
e 90 ( 2014 ) 032304 . http://dx.doi.org/10.1103/physreve.90.032304 [ ] .
a. zttl , h. stark , hydrodynamics determines collective motion and phase behavior of active colloids in quasi - two - dimensional confinement , phys .
( 2014 ) 118101 .
http://dx.doi.org/10.1103/physrevlett.112.118101 [ ] .
a. m. menzel , t. ohta , h. lwen , http://link.aps.org/doi/10.1103/physreve.89.022301[active crystals and their stability ] , phys .
e 89 ( 2014 ) 022301 . http://dx.doi.org/10.1103/physreve.89.022301 [ ] .
l. chen , j. toner , http://link.aps.org/doi/10.1103/physrevlett.111.088701[universality for moving stripes : a hydrodynamic theory of polar active smectics ] , phys .
( 2013 ) 088701 .
http://dx.doi.org/10.1103/physrevlett.111.088701 [ ] .
http://link.aps.org/doi/10.1103/physrevlett.111.088701 t. c. adhyapak , s. ramaswamy , j. toner , http://link.aps.org/doi/10.1103/physrevlett.110.118102[live soap : stability , order , and fluctuations in apolar active smectics ] , phys .
( 2013 ) 118102 . http://dx.doi.org/10.1103/physrevlett.110.118102 [ ] . http://link.aps.org/doi/10.1103/physrevlett.110.118102 g. szamel , e. flenner , l. berthier , http://link.aps.org/doi/10.1103/physreve.91.062304[glassy dynamics of athermal self - propelled particles : computer simulations and a nonequilibrium microscopic theory ] , physical review e 91 ( 6 ) ( 2015 ) 062304 . http://dx.doi.org/10.1103/physreve.91.062304 [ ] . http://link.aps.org/doi/10.1103/physreve.91.062304 c. maggi , u. m. b. marconi , n. gnan , r. di leonardo , multidimensional stationary probability distribution for interacting active particles , sci . rep . 5 ( 2015 ) 10742 .
|
minimal models of active brownian colloids consisting of self - propelled spherical particles with purely repulsive interactions have recently been identified as excellent quantitative testing grounds for theories of active matter and have been the subject of extensive numerical and analytical investigation .
these systems do not exhibit aligned or flocking states , but do have a rich phase diagram , forming active gases , liquids and solids with novel mechanical properties .
this article reviews recent advances in the understanding of such models , including the description of the active gas and its swim pressure , the motility - induced phase separation and the high - density crystalline and glassy behavior .
active matter , phase separation , active glasses , swim pressure
|
extension of the unprecedentedly accurate ams-02 data@xcite on positron and electron cross - sections to energies above the electroweak scale has major implications . before the publication of the data shown in fig .
1 , it was possible to argue@xcite that the positron content of cosmic rays is the consequence of galactic collisions of high energy primaries via established physics .
now , it is unarguably clear that a new phenomenon of some kind is necessarily involved .
= 3.8 in = 2.7 in @xmath1 = 2.7 in figure 1 ams-02 data various proposals have been put forward but the simplest and most obvious explanation is that there is a new electroweak scale strong interaction , within the primary collisions , that plentifully produces positrons and electrons equally .
this explanation is a radical proposition within the current theory paradigm but the consequences are historic if what is seen is ( as i already suggested@xcite for the lower energy results ) the qcd interaction of the color sextet quark sector of the theory@xcite that i call qud . very significantly ,
the neutral nucleon states of this quark sector provide a form of dark matter that seems to have just the physical properties now thought to be essential .
qud is a very special massless field theory that was first discovered @xcite because of it s unique high - energy unitarity properties .
it was then realized that , amazingly , this theory might underly , unify , and provide an origin for , the full standard model .
it is an infra - red fixed - point su(5 ) theory within which novel massless fermion anomaly dynamics lead to a massive bound - state s - matrix in which the states and interactions are@xcite-@xcite , apparently , those of the standard model .
there is also a higgs - like boson resonance and so qud provides an immediate explanation@xcite for the emergence of the `` nightmare scenario '' at the lhc
. it would be hard to exaggerate the significance of establishing that qud underlies the standard model .
it is self - contained , can not be added to , involves no explicit parameters , and contains both neutrino masses and dark matter .
it has to be entirely right , including the dark matter sector , or else it is simply wrong .
so far , there is much apparent success and no obvious failure . that the multi - regge construction of the s - matrix is very far outside the mainstream , both conceptually and technologically ,
may explain , perhaps , why it s virtues remain largely unappreciated .
the most radical paradigm change involved is that consistency with the underlying su(5 ) unification is achieved , in the multi - regge ( infinite momentum ) framework , via the infra - red enhancement of standard model interactions by a `` wee parton vacuum '' of anomalous wee gauge bosons coupled to massless fermion anomalies .
( the wee gauge bosons are multi - reggeon configurations that are non - local generalizations of the familiar anomaly current . ) as i will outline , the derivation of this crucial result utilises elaborate multi - regge theory .
it is in sharp contrast to the conventional wisdom that additional short - distance interactions are needed to achieve unification of the forces of the standard model .
the color sextet quark sector is the only new qud physics beyond the standard model .
it produces both electroweak symmetry breaking and , as already referred to , a very special form of strongly interacting dark matter .
the existence of the sextet quark sector is crucial for the unitarity properties of the s - matrix and , perhaps as a consequence , the dynamical properties of the qcd sector are modified in a manner which improves considerably the comparison with experiment . particularly important is the absence of glueballs which , for several years has been a deep puzzle for qcd , but which is vital for qud dark matter .
the regge pole nature of the pomeron ( in first approximation ) is also apparent .
this has recently been clearly demonstrated@xcite experimentally at the lhc - even though it is commonly believed that it is not a property of qcd . because of the large wee parton anomaly color factors involved , sextet qcd interactions are stronger than those of the triplet sector , even though the masses are all electroweak scale .
also , although the anomaly dynamics is subtle , it is the same for the triplet and sextet sectors .
consequently , the sextet matter sector will have many properties analogous to the familiar triplet sector , with the difference being that sextet matter will be more massive , more strongly interacting , and predominantly neutral .
there are two sextet flavors which , therefore , produce sextet `` pions '' and `` nucleons '' as the physical states .
the sextet pions become longitudinal vector bosons , while the sextet quark neutrons ( neusons ) are stable , as i will discuss , and provide dark matter . from the ams-02 data shown in fig . 1 we see that , as the electroweak scale is approached ( and beyond ) there is a dramatic increase in positron production and a corresponding enhancement of electron production .
the phenomenon involved is most directly reflected , obviously , in the increase in the positron fraction .
if qud is responsible , then we are looking at `` low - energy '' products of the sextet quark interaction .
this interaction will ( in parallel with the triplet quark hadronic sector ) be dominated by the production of sextet pions that will be manifest , physically , as the production of multiple vector boson states .
the leptonic decays of the vector bosons will produce electrons and positrons plentifully , in equal numbers , just as seen by ams .
unfortunately , as i have lamented often in previous papers , the kinematic focus of the lhc is completely wrong for the discovery of the qud strong interaction .
nevertheless , supporting evidence for the qud explanation of ams data has now been seen , albeit at the extreme short - distances studied at the lhc .
an excess cross - section for vector boson pair production of the kind predicted by qud ( i.e. @xmath2 and @xmath3 , but not @xmath4 , @xmath5 , or @xmath6 ) , that increases with energy , has now been measured by both atlas@xcite and cms@xcite .
nevertheless , it will surely be ironic if it happens that , once again , the discovery of a new interaction is viewed historically as made by cosmic ray physics when it could / should , so easily , have been discovered in an accelerator - certainly the lhc , perhaps even the tevatron .
sextet protons ( prosons ) have an electromagnetic mass and decay into neusons .
therefore , the neutral neusons can be produced as charged prosons that decay , and so they should appear as ultra - high - energy cosmic rays .
it is crucially important , as i have already implied , that there are no glueballs in the qud s - matrix .
this is because the physical states all have chiral properties associated with anomaly pole components coupled to anomalous wee gluons .
would - be glueballs do not have an anomaly pole component and so can not appear as physical states .
also , because of the absence of hybrids , there are no triplet / sextet mesons that could provide an interaction between the triplet and sextet sectors .
consequently , the high - energy pomeron provides the only qcd interaction between the two sectors and so dark matter neusons have the fundamental property that they have a strong interaction with normal matter ( i.e. hadrons ) , but only in very high energy collisions .
as discussed in @xcite , a pomeron interaction can involve either a rapidity gap or , most often , a high multiplicity hadron state spread uniformly across part of the rapidity axis .
the high - energy dominance of sextet states via pomeron interactions , potentially , provides an explanation for the dominance of dark matter production in the ultra - high - energy processes responsible for early universe formation .
it also implies that prosons will be preferentially produced in the present universe , relative to protons , by the highest energy strong interaction processes .
hence , after cosmic acceleration , neusons produced by prosons ( and antineusons produced by antiprosons ) should be the dominant cosmic rays , at the highest energies .
when first discovered , cold dark matter was conjectured to be very different from qud dark matter .
now it seems that more and more of the properties of qud appear in the latest models .
composite models are routinely considered@xcite and it is recognized that strong self - interactions are essential with cross - sections even bigger than those of standard nuclear cross - sections .
both properties are intrinsic for qud sextet dark matter .
this is fortunate since , as we stated above , if qud is correct then dark matter must be neusons or antineusons ( and , perhaps , neuson or antineuson composites ) .
ironically , the most recent dark matter models invoke@xcite a complete ( hidden sector ) non - abelian gauge theory , in addition to the standard model , to produce very strong interactions . a priori
, it might seem to be much simpler to have a higher mass sector of qcd produce the interaction .
however , the conventional expectation is that , because of the evolution of @xmath7 , a higher mass sector would necessarily be weaker interacting .
this is why a `` hidden sector '' , entirely new , gauge theory is invoked .
fortunately , the s - matrix anomaly dynamics of qud implies that the higher mass qcd sextet sector will , indeed , be more strongly interacting than the lower mass triplet sector .
the sextet sector slows down the evolution of @xmath8 beyond the electroweak scale , but this is not the most important phenomenon .
more importantly , the anomalous wee gluon enhancement of the basic interactions is much stronger for the higher color sextet sector . as a result ,
a stronger interacting , more massive sector and a less strongly interacting , lighter sector actually coexist within the same gauge theory .
in @xcite i discussed a variety of cosmic ray phenomena that can be interpreted as evidence for the qud sextet quark interaction , including the presence of neusons as a major uhe component that is responsible for all the confusion surrounding the energy dependence of the mass composition .
i also discussed why a , large cross - section , electroweak scale interaction is the simple and obvious explanation for a cosmic ray mystery that has been well - known for many years , without any agreement on the phenomena involved . the spectrum knee shown in fig . 2 ,
is widely believed to be astrophysical in origin . however , it is such a pronounced ( relatively ) local effect in the spectrum that it is surely far more plausible that the atmospheric interaction is responsible , with the apparent increase of slope due to the production of undetected particles .
that these particles could not be dark matter in any of the forms that in the past were commonly proposed was seen , not surprisingly , as a major argument against such an interaction .
= 3.5 in figure 2 the cosmic ray spectrum knee the qud sextet quark interaction in the atmosphere can be the origin of the knee , as the threshold for double pomeron production ( via either a rapidity gap or a high multiplicity hadron state ) of dark matter neusons and antineusons . as we will discuss , we expect neusons to have a mass @xmath9 gev .
in addition , flavor quantum numbers ( `` sextet isospin '' ) require that they be produced multiply - with two being the minimum .
consequently , the energy scale for the knee is , essentially , set by double pomeron production of massive states with a mass @xmath10 1 tev , which is very reasonable phenomenologically .
there is , however , a minor subtlety .
neusons will be produced both by the atmospheric collisions of normal hadronic primaries and also by the atmospheric collisions of incoming primary neusons . because of the large neuson mass the corresponding thresholds will be close in energy@xcite . in both cases , produced neusons will lose energy via a small ( including zero ) number of subsequent collisions and then `` disappear '' , thus producing the knee .
qud was initially uncovered@xcite via a search for a connection between the critical pomeron and gauge theory reggeon interactions .
the reggeon field theory critical pomeron is the only known description of rising total cross - sections ( an essential requirement to match with an asymptotically - free gauge theory at short distances ) that satisfies full multiparticle t - channel unitarity and all s - channel unitarity constraints .
the rft supercritical phase occurs in superconducting qcd , and the critical behavior requires that asymptotic freedom be saturated .
the only realistic possibility is qcd@xmath11 , with six color triplet quarks and two color sextet quarks .
this is a real possibility if `` sextet pions '' are responsible for electroweak symmetry breaking !
consequently , the critical pomeron requires a specific form of electroweak symmetry breaking that introduces , moreover , a higher color qcd interaction ( of the kind seen by ams ! ) in addition , a detailed analysis shows that an infra - red fixed - point is also needed and so qcd@xmath11 must contain only massless quarks - which is certainly not physically realistic ! at first sight this is fatal for the physical realization of the critical pomeron .
fortunately , there is a resolution .
the way out is provided by the inclusion of the electroweak interaction . insisting on both asymptotic freedom and no anomaly , the required sextet quark sector and the electroweak interaction embed uniquely@xcite in *
qud @xmath12 su(5 ) gauge theory with left - handed massless fermions in the @xmath13 representation . under @xmath14 * @xmath15
@xmath16 @xmath17 strikingly , both the triplet quark and lepton sectors are very close to the standard model . also , qud is real _ \{vector - like } _ with respect to @xmath18 .
the @xmath19 quantum numbers are not right , but the short - distance lepton anomaly is correct .
having first discovered how the reggeon anomaly dynamics of qcd@xmath11 leads to the critical pomeron , i realized that , with the same dynamics , qud could actually be physically realistic .
all elementary leptons and quarks would be confined and remain massless , with infra - red anomalies dominating the dynamics
! the standard model could be an effective theory obtained ( in principle ) by integrating out the elementary leptons . at first sight ,
finding the bound - states and interactions of a massless field theory , such as qud , looks hopeless .
fortunately , multi - regge theory comes to the rescue .
the multi - regge region involves multiple infinite - momenta and it is well - known that infinite momentum wee partons can play a vacuum role . in the next section
we outline how qud infra - red divergences produce wee partons that accompany multi - regge s - matrix interactions and states that are , relatively simple and , essentially , determined by the parity properties of the fermion couplings .
the elementary leptons and triplet quarks are sufficient to produce the known states of the standard model .
moreover , with an infra - red fixed - point and no exact chiral symmetries , all bound - states will acquire masses via reggeon diagram interactions .
the sextet sector , which we will be discussing further , is responsible for electroweak symmetry breaking as anticipated .
the color octet sector was certainly not asked for .
however , it plays a crucial role , via large @xmath20 anomalies , in producing the generation structure of the physical states .
consequently , the potentially profound possibility is raised that the full array of standard model physics could actually be uniquely determined , via qud , by the necessity to accompany the critical pomeron and provide a complete , massive , unitary s - matrix .
the following is a very brief description of the procedure@xcite for constructing the bound - state s - matrix using multi - regge reggeon diagrams .
initially , all reggeons are given masses using scalar condensates .
fermion masses are removed first , leaving ( fundamentally ) chirality transitions that break su(5 ) to su(3)@xmath21u(1)@xmath22 , but only in reggeon anomaly vertices .
gauge boson masses are then smoothly removed ( via complementarity ) , producing a sequence of increasing reggeon global symmetries @xmath23 @xmath24 the last scalar removed is asymptotically free and so the @xmath20 cut - off limit @xmath25 can be taken between the su(4 ) and su(5 ) limits . with @xmath26 many fermion loops violate ward identities and all left - handed gauge boson interactions ( except those mixing with sextet pions ) are eliminated by exponentiating reggeization divergences .
a complete infra - red analysis ( using properties of reggeon interaction kernels ) shows that all divergences exponentiate via reggeization , apart from an overall divergence which produces `` universal wee partons '' that select the physical states and interactions , as illustrated in fig .
the wee partons are gauge bosons that form an adjoint su(5 ) representation with anomalous color parity ( @xmath27 ) that keeps them outside the exponentiations of reggeization . instead , the divergence they produce couples to anomaly vertices that produce anomaly poles via chirality transitions .
u(1 ) anomaly poles couple the wee partons in distinct reggeon channels .
chiral goldstone anomaly poles are a central element of all the bound - states .
( goldstone pions are produced within qcd as shown in fig .
a goldstone anomaly pole can be characterized either as a bound - state of the intermediate fermion pair ( one of which is initially unphysical and carries zero momentum ) or , after the chirality transition , as a reggeon state containing wee partons .
the reggeon state will have an su(5 ) singlet projection , while the bound - state , because of the chirality transition , will violate this symmetry .
= 4.7 in figure 3 .
reggeon diagrams contributing to the overall divergence .
figure 4 . the anomaly pole goldstone pion .
left - handed gauge bosons contribute as wee gauge bosons only when accompanying the sextet pion / gauge boson interaction .
after @xmath25 and the su(5 ) limit , the critical pomeron and the massless photon appear together . also , color octet chiral anomalies at @xmath28 produce@xcite bound - states in standard model generations .
the resulting `` standard model '' vector interactions between bound - states are -0.25 the sextet pions ( @xmath29 ) provide a mass for the electroweak vector bosons via anomaly interactions of the kind illustrated in fig .
5 . figure 5 .
vector boson mass generation this gives a mass @xmath30 ( @xmath31 ) as a wee gluon integral multiplied by a sextet anomaly factor , while the chirality transition provides a sextet flavor quantum number that prevents the exponentiation that would result if the massless vector boson interacted with the wee partons .
the perturbative coupling of the @xmath32 and @xmath33 is retained and the sextet flavor symmetry becomes the su(2)@xmath34 of the standard model .
the elementary qud coupling is kept very small by the fixed - point , but physical standard model couplings will emerge if the infinite sums of wee gauge boson anomaly color factors enhance couplings sufficiently .
the anomaly factors clearly give @xmath35 three standard model generations of physical hadrons and leptons appear via the octet anomalies that remain at infinite light - cone momentum after the full su(5 ) symmetry is restored .
when described in terms of fermions only , without anomalous wee gauge bosons but before the chirality transitions , all bound - states are su(5 ) singlets composed of five elementary fermions , two of which are non - dynamical color octets forming an `` octet pion '' large @xmath20 anomaly contribution . in general ,
anomaly vertex mixing , combined with fermion and wee parton color factors , will produce a wide range of bound - state mass scales .
in addition to the color octets , lepton bound states contain three elementary leptons with two producing an anomaly pole .
the electron is very close to elementary because the anomaly pole involved disturbs the dirac sea minimally .
the muon has the same constituents , but in a different anomaly pole dynamical configuration that will obviously generate a significant mass .
the very small qud coupling should be the origin of desirably small neutrino masses - that have no color or electric charge anomaly enhancement via wee partons .
anomaly color factors imply @xmath36 non - leptonic bound - states are triplet or sextet quark mesons and baryons .
there are no hybrids and , because the anomaly pole bound - states are necessarily chiral goldstones , there are no glueballs .
two qud triplet quark generations give standard model hadrons - that mix appropriately .
the physical b quark is a mixture of all three qud generations .
there are two `` exotic '' triplet quarks with charges -4/3 & 5/3 that have no chiral symmetry and so do not produce light ( anomaly pole ) bound - states .
the left - handed `` top quark '' forms an electroweak doublet with one of the exotic quarks and does not appear in low mass states .
however , a bound - state involving the right - handed top quark remains that mixes with the sextet @xmath37 to give two mixed - parity scalars@xcite - the @xmath38 and the @xmath39 .
the @xmath39 has an electroweak scale mass , while the @xmath38 has a mass between the triplet and sextet scales . ``
tree - unitarity '' implies that the combined @xmath38 and @xmath39 couplings should reproduce@xcite the standard model electroweak couplings of the higgs boson .
therefore , the @xmath38 could be the `` higgs boson '' discovered at the lhc .
if so , the `` qud higgs '' is , predominantly a `` top / anti - top '' resonance . at first sight ,
top quark physics is very different from the standard model .
however , because the sextet @xmath40 reproduces the standard model final states ( automatically at an electroweak scale mass ! ! ) it is hard , experimentally , to distinguish the difference@xcite .
a fundamental outcome is that the `` top quark mass '' of the standard model is electroweak scale simply because a sextet quark resonance is , essentially , responsible , for the events .
an important , experimentally distinguishing , feature of qud `` top quark physics '' is that the @xmath39 should be seen ( and may have already been seen@xcite ) decaying into @xmath3 pairs at the lhc - at the experimental @xmath41 threshold ( where @xmath42 is the standard model top quark ) .
it is well - known that , based on feynman graph color factors , triplet and sextet quark momentum scales are expected to satisfy ( approximately ) the `` casimir scaling '' rule @xmath43 where @xmath44 and @xmath45 are dynamical momenta and @xmath46 and @xmath47 are casimirs for triplet and sextet quarks respectively .
equivalently , in going from triplet to sextet graphical contributions , @xmath7 is effectively replaced by @xmath48 . ) for su(3 ) there are two casimir operators which , in terms of the generators @xmath49 can be written as @xmath50 @xmath51 appears predominantly in anomaly color factors and it is at the origin of the large magnitude of sextet anomaly color factors . since @xmath52 and @xmath53 , we can write @xmath54 the presence of the sextet sector affects both the evolution of @xmath7 and the extraction of parton distribution functions . in the early studies of jet cross - sections at the tevatron an infamous `` jet excess ''
was discovered that could have ( should have ? ) been viewed as direct evidence for the sextet sector .
the results for both @xmath7 and the jet cross - section , shown in fig . 7 , are just as would be expected if the excess were due to double pomeron vector boson pair production .
( again , pomeron exchange implies@xcite either a rapidity gap or , much more often , production of a high multiplicity state spread fairly uniformly across that part of the rapidity axis . )
the vector boson decays would be via the ( dominant ) hadronic modes in this case , as opposed to the leptonic mode involved in the ama-02 electroweak scale positron / electron excess shown in fig . 1 .
unfortunately , as is well known , more sophisticated jet algorithms and appropriately modified gluon distributions were searched for until the conclusion could be reached that there is no significant jet excess .
moreover , it has now become accepted that it is necessary / appropriate for large non - perturbative factors to be included in all applications of perturbative qcd at the lhc . as a result , `` non - perturbative '' hadronization corrections , that will surely obscure any large @xmath58 behavior due to sextet quark physics , are routinely included in jet cross - section analyses . as we have elaborated on in @xcite , because there is no derivation of the parton model , the current phenomenological determination of parton distributions from existing strong interaction data is consistent only if new short - distance physics is all that remains to be discovered
. if new electroweak scale qcd physics is present , then the conventional parton model assumptions and procedure will clearly be inadequate .
it is very fortunate , and may be of enormous historic importance , that the positrons recorded by ams-02 , unlike the jets seen by cdf , can not be made to disappear by appealing to parton model ambiguities in the physics analysis and interpretation of the experimental results . if the conclusion from the ams-02 results is that the physics `` discovered '' in fig .
7 is `` real '' then the implications for lhc physics are immeasurable .
that the ( triplet quark ) proton is lighter than the neutron is entirely due to the fact that the current mass of the @xmath59 quark is less than that of the @xmath60 quark .
electromagnetic effects , alone , would make the proton heavier .
remarkably , the current quark mass difference produces an effect that is of the same order of magnitude , but is just a little larger , and of the opposite sign , than the electromagnetic mass difference . because of the absence of hybrid triplet / sextet states , the lightest of the sextet nucleons must be stable . also , sextet quark current masses must be zero if sextet pions are to mix , via anomaly poles , with the massless @xmath61 and @xmath62 states to give them masses .
in fact , in qud , the su(5 ) gauge bosons can directly transform triplet quarks to leptons , but can not do the same for sextet quarks .
consequently , massive triplet quarks , but not sextet quarks , would be generated ( by self - energy corrections ) in an effective lagrangian containing the massive physical leptons - obtained by `` integrating out '' the elementary massless leptons .
the sextet nucleon mass difference has to be entirely electromagnetic in origin , and so it is the neuson ( @xmath63 ) and the antineuson ( @xmath64 ) that are stable . if the @xmath39 produces `` top events '' , as discussed in the last section , then the sextet quark dynamical mass is approximately given by the standard model top quark mass
this implies the neuson mass should be @xmath65 and the proson mass should be just a little higher .
since triplet and sextet quarks do not combine to form bound states it can be assumed that sextet nucleons also do not form bound states with triplet nucleons .
more particularly , perhaps , if pion exchange provides the binding force for nucleons to form nuclei , the distinct quark content of sextet and triplet nucleons implies that there is no common `` pion '' that can bind them .
the neuson is , therefore , neutral , stable , and ( because of the dominance of sextet states ) will be the dominant , heavy , stable state produced in high energy cross - sections .
consequently , it will be dominantly produced in the high energy interactions that are believed to have been responsible for the formation of the early universe . if it does not form bound states with normal quark matter it will abundantly form cold dark matter , in the form of ( sextet ) nuclei , etc .
( perhaps sextet pions can exist inside sextet nuclei and provide the binding force . ) as a result , the existence of the sextet nucleon sector provides a natural explanation for the dominance of dark matter in the universe .
conversely , if it can be established that neusons are responsible for dark matter , the dominance of dark matter in the universe can be regarded as evidence confirming that sextet quark states dominate high energy cross - sections .
_ this is all the more remarkable because , as we already emphasized , qud is self - contained , can not be added to , involves no parameters , and is entirely right , including the dark matter sector , or is simply wrong . _
a. r. white , `` high - energy unitarity and the standard model '' .
proceedings of gribov-75 : hep - ph/0508125 .
the path from high - energy unitarity to the critical pomeron , then to qcd@xmath11 via the supercritical pomeron , and finally to qud , is outlined ( with references ) .
|
the latest ams-02 electron / positron precision data add to the spectrum knee as direct cosmic ray evidence for an electroweak scale strong interaction .
in addition , there is significant evidence for a strong self - interaction of dark matter .
qud is a unique massless su(5 ) field theory with an anomaly - generated bound - state s - matrix that could be an unconventional origin for the standard model .
the electroweak scale color sextet quark enhanced qcd interaction is the only new physics .
production of multiple vector bosons , that acquire masses via sextet quark pions , will give the ams positron and electron cross - sections - related vector boson pair production having been seen at the lhc .
stable sextet quark neutrons ( neusons ) provide a novel form of very strongly interacting dark matter that has the desired experimental properties .
large rapidity hadronic production of neusons will produce the knee .
-0.4 in 8.5 in 1=0.5 in -0.5 in plus 1000pt minus 1000pt # 1 # 1= to#1 to # 1 # 1 # 1 to 2 # 1 # 1 to 1 in # 1 # 1#2#1#2 # 1#2#3 2 # 1#2 # 3 # 1#1| # 1| # 1 # 1#1 # 1#2#1| .
# 2 # 1/ # 1| # 1| # 1#1 # 1 #
1#20=1=1>0.51-.500-.50-.5110>1 .50-.51 # 1#2 # 1#2 # 1 # 1\{. # 1 # 1=@xmath0= = # 1 # 1,by1 = 1 [ # 1]==,[#1]= # 1-#1- # 1(#1 ) # 1/#2 # 1@xmath0 # 1 # 1 # 1#2#3*#1 * , # 2 ( 19#3 ) # 1[#1 ] # 1 # 1 # 1 # 1(#1 ) ams-02 , strongly self - interacting dark matter , and qud alan r. white 0.6 cm argonne national laboratory 9700 south cass , il 60439 , usa .
|
the quantification of entanglement in a multipartite quantum system is one of the most important and challenging topics of quantum information theory .
the first attempt of measuring entanglement was based on the violation of the so called bell s inequalities @xcite .
it was thought that the amount of non - classical correlations in a bipartite quantum system was intrinsically related to the level of violation of these family of inequalities .
nonetheless , the generalization of such inequalities to multipartite states is not possible in general .
furthermore , it was already shown that , even for the bipartite case , the chsh ( clauser - horne - shimony - holt ) form of bell s inequalities does not constitute a good measure of quantum correlations in the sense that there are states which do not violate it , but can be purified by local operations and classical communication ( locc ) to a state which does @xcite .
therefore , different measures of entanglement were introduced based on the concept of asymptotic distillability @xcite . for pure states ,
the von neumann entropy @xcite of the reduced density matrix @xmath0 is exactly the number of bell states which can be asymptotically distilled by locc from @xmath1 .
it satisfies all desired properties ( normalization , convexity , local unitary operations invariance , etc ... ) of an entanglement measure [ 6 ] .
several attempts were made in order to generalize this measure ( the entropy entanglement - @xmath2 ) for mixed sates .
the most successful are the entanglement cost ( @xmath3 ) , the entanglement of formation ( @xmath4 ) , the relative entropy of entanglement ( @xmath5 ) and the distillable entanglement @xmath6 @xcite .
they all coincide with the entropy of entanglement in the pure states set , showing their relationship with asymptotic distillability properties , and satisfies most of the desirable properties .
however , none of these have an operational expression , i.e. , they can not be calculated for arbitrary states .
furthermore , they are defined only for bipartite states , not expressing the interesting properties of multipartite systems @xcite .
other measures of entanglement for mixed states , not related with @xmath2 , have also been proposed .
the robustness of entanglement ( @xmath7 ) quantifies the minimal amount of mixing with separable states needed to destroy the original entanglement presented in @xmath8 @xcite .
it follows easily from its definition that it can be used to quantify multipartite entanglement .
however , it is calculable only for a few specialized bipartite density operators .
the negativity @xmath9 is the only example of a computable quantifier @xcite .
nonetheless , since it is based on the negativity of the eigenvalues of the partial transpose matrix @xmath10 of the system , it does not quantify the entanglement of positive - partial - transpose ( ppt ) states .
moreover , it is defined only for bipartite correlations . in this paper
we introduce a new computable measure for entanglement , which quantifies all different kinds of multipartite entanglement .
its definition is based on the concept of entanglement witness . according to @xcite , an @xmath11-partite density operator @xmath12(the space of bounded operators acting on the hilbert space @xmath13 ) is non - separable iff there
exists a self - adjoint operator @xmath14 which detects its entanglement , i.e. , such that @xmath15 and @xmath16 for all @xmath17 separable .
this condition follows from the fact that the set of separable states is convex and closed in @xmath18 .
therefore , as a conclusion of the hahn - banach theorem , for all entangled states there is a linear functional which separates it from this set .
it will be necessary to consider only normalized entanglement witnesses such that @xmath19 .
a hermitian operator @xmath20 is an optimal ew ( oew ) for the density operator @xmath21 if @xmath22 for every ew @xmath23 .
although the above definition is different from the one introduced in @xcite , the optimal ews of both criteria are equal .
it is important to stress the close relation between bell s inequalities and entanglement witnesses .
since an ew is a hermitian operator , it is associated with a possible experiment whose expectation value in all separable states is positive .
therefore , one may consider witness operators as generalized bell s inequalities , where the violation of the latter is replaced by a negative expectation value .
of course bell s inequalities have an important role in the refutation of hidden loval variables theories for quantum mechanics , not shared by ews .
nonetheless , in what concerns the detection of entanglement , entanglement witnesses can be considered just generalized bell s inequalities , with the locality condition of the later relaxed .
thus , the optimal entanglement witness of a state @xmath21 can be identified with the experiment which most detects its entanglement , i.e. , which has the lowest possible expectation value .
we will show that the result of this _ optimal _ experiment might be used to quantify entanglement .
the witnessed entanglement @xmath24 of a multipartite state @xmath25 is given by @xmath26 the lack of an operational procedure to calculate entanglement measures in general is ultimately related to the complexity of distinguishing entangled from separable mixed states , which was shown to be np - hard @xcite . since an operational measure would also be a necessary and sufficient test for separability , we should not expect to find one .
nevertheless , a procedure to calculate oew for all states with arbitrary probability and precision was recently introduced @xcite .
although the optimization of entanglement witnesses is also np - hard @xcite , the optimization of approximate ew , operators which are positive for almost all separable states , can be realized very efficiently with a chosen probability . as the amount of violation ( the percentage of negative expectation values over the separable states set )
can also be chosen , it is possible to reach the exact oew with any desired precision .
thus , the witnessed entanglement ( we ) introduced earlier can be approximately calculated for all mixed states .
besides having an interesting physical interpretation and an efficient numerical method of computation , the witnessed entanglement fulfills most of the usual requirements of a good entanglement measure @xcite .
these properties can be summarized as : \(i ) e(@xmath27 ) = 0 if , and only if , @xmath27 is separable .
\(ii ) local unitary operations leave @xmath28 invariant , i.e , @xmath29 \(iii ) ( monotonicity under locc ) entanglement can not increase under local operations and classical communication ( locc ) protocols .
@xmath30 where @xmath31 is a superoperator implementable by locc .
\(iv ) ( continuity ) for every @xmath32 and density operators @xmath8 and @xmath27 , there is a real number @xmath33 such as @xmath34 where @xmath35 @xmath36 is any norm defined for finite dimensional systems .
note that @xmath37 depends of the chosen norm .
\(v ) ( convexity ) mixing of states does not increase entanglement . @xmath38 for all @xmath8 , @xmath27 and @xmath39 . the witnessed entanglement satisfies properties ( i ) , ( ii ) , ( iv ) and ( v ) .
( i ) : from the definition of entanglement witnesses we have that @xmath40 for all separable states @xmath27 . therefore
, @xmath41 conversely , as there exists an ew for every entangled state , @xmath42 is positive for any non - separable density operator .
( ii ) : we will prove by absurd that @xmath42 is invariant under local unitary operations .
suppose that @xmath43 , where @xmath44 then , @xmath45 where @xmath46 is an entanglement witness . but equation ( 9 ) contradicts the fact that @xmath47 is the oew for @xmath8 . if we suppose that @xmath48 , the same argument can be applied to @xmath49 .
therefore , @xmath50 must hold .
( iv ) : let @xmath47 and @xmath51 be optimal entanglement witnesses for @xmath8 and @xmath27 , respectively . if either @xmath8 or @xmath27 is separable , the result follows trivially .
we then assume that @xmath52 as the intersection of the ews set with the set of hermitian matrices with unity trace is compact , we have that for some norm @xmath35 @xmath53 , @xmath54 where @xmath55 is a real number .
every norm of finite demensional operators are equivalent , i.e. , for every finite dimensional operator @xmath56 and any two chosen norms @xmath35 @xmath53 and @xmath35 @xmath36 , there always exists rela numbers @xmath11 and @xmath57 such that @xmath58 thus , one sees that equation ( 14 ) is valid for every norm .
we can assume without loss of generality that @xmath59 ( d ) .
hence , @xmath60 + tr(w_{\rho}\sigma ) - tr(w_{\sigma}\sigma)|.\ ] ] since @xmath51 is a oew for @xmath27 , @xmath61 . from ( d )
we also have @xmath62| \geq |tr[(w_{\rho } - w_{\sigma})\sigma]|$ ] . therefore , @xmath63 + tr(w_{\rho}\sigma ) - tr(w_{\sigma}\sigma)| \leq |tr[w_{\rho}(\rho - \sigma)]| \leq ||w_{\rho}||_{hs}||\rho - \sigma||_{hs},\ ] ] where we have used the cauchy - schwartz inequality for the hilbert - schmidt inner product . from eq .
( 17 ) and conditions ( b ) and ( c ) we then have that for every norm @xmath35 @xmath36 , @xmath64 for some real numbers @xmath65 .
( v ) : the convexity follows straightforwardly from definition 2 and the concavity and convexity of the @xmath66 and @xmath67 functions , respectively .
although @xmath68 is not a monotone under general locc maps , we can state the following : for every unital locc map @xmath69 , @xmath70 it suffices to show that @xmath42 is not increasing under separable operations .
let @xmath31 be a uninal trace - preserving separable quantum superoperator .
then , @xmath71 where @xmath72 is the dual map of @xmath31 , which is unital , since @xmath31 is trace preserving , and separable , and @xmath73 is the optimal ew for @xmath74 . , @xmath75 is also an ew .
indeed , @xmath76 for all @xmath77 , since @xmath23 is an ew and @xmath78 .
now we have to prove that @xmath79 . using that the channel is unital @xmath80 , i.e. @xmath81 , @xmath82
a @xmath11-partite density operator @xmath21 is a @xmath57-separable state if it is possible to find a decomposition for it such that in each pure state term at most m parties are entangled among each other , but not with any member of the other group of @xmath83 parties .
since the subspace of m - separable density operators is convex and closed , it is also possible to apply the hahn - banach theorem to it and establish the concept of entanglement witness to @xmath84-partite entanglement . in order to do that ,
consider the index set @xmath85 .
let @xmath86 be a subset of p which has at most m elements .
then @xmath23 is an @xmath84-partite entanglement witness if : @xmath87 equation @xmath88 assures that the operator @xmath23 is positive for all m - separable states .
thus , it is also possible to determine optimal @xmath84-partite entanglement witnesses for every @xmath84-partite entangled state and establish an order in this set with @xmath42 .
we have considered so far oew of the most general form .
we will show that it is possible to construct another measure of entanglement for non - positive - partial - transpose - states ( npptes ) , reducing the search space of ew .
witnesses operators can be classified as decomposable ( d - ew ) or not decomposable ( nd -ew)@xcite . in the case of bipartite entanglement , a hermitian operator @xmath23 is a @xmath89 if it can be written in the form of @xmath90 where p and q are positive operators .
it is clear from this definition that these ews can not detect pptes .
a @xmath91 is a hermitian operator which can not be written as @xmath92 .
it was already shown that @xmath23 is a @xmath91 if , and only if , it detects pptes [ 12 ] .
if we restrict our search of optimal ew to @xmath89 , we will still have a measure of entanglement , the decomposable witnessed entanglement ( @xmath93 ) , which satisfies the same properties of @xmath42 , despite the fact that it will be null for both separable and pptes .
the advantage of this new measure is that it can be calculated in a straightforward manner .
@xmath94 @xmath95 , where @xmath96 stands for the minimum eigenvalue of a. if @xmath97 is separable or ppt , then @xmath98 will be non - negative and @xmath93 will be null . if @xmath97 is nppt , since @xmath99 , it is easy to see that the optimal d - ew will be @xmath100 , where @xmath101 is the projector in the subspace of the minimum eigenvalue of @xmath102 .
the negativity ( @xmath103 ) is another example of an entanglement measure which is both computable and null for pptes @xcite .
it is well known that , for systems of two qubits , @xmath103 is equal to twice the absolute value of the negative eigenvalue of the partial transpose of the state .
thus , in this dimension , it coincides with @xmath93 .
the decomposable witnessed entropy of entanglement is particularly interesting for 2 x 2 and 2 x 3 systems .
since there are no ppt entangled states in these cases , the optimal @xmath89 is the exact oew and @xmath42 and @xmath93 are equal .
@xmath104 and @xmath105 for all 2 x 3 and 2 x 2 states , respectively .
we have seen that , up to a dimension normalization factor , @xmath106 can be identified with the result of an experiment which has the lowest possible expectation value in @xmath8 , while it has non - negative expectation value in all separable states .
@xmath106 is also related to the amount of mixing with the maximally random state @xmath107 needed to destroy the original entanglement presented in @xmath8 .
this quantity , introduced in [ 9 ] , is called random robustness ( @xmath108 ) and is given by the minimum @xmath109 for which @xmath110 is separable .
@xmath111 , for every multipartite density operator @xmath8 . since @xmath112 is the minimum value for which @xmath113 is separable , @xmath114 .
therefore , @xmath115 .
it was proved in @xcite that , for bipartite pure states , @xmath108 is given by the product of the two biggest schmidt coefficients with the dimension of the composite hilbert space .
given a pure bipartite state @xmath116 with decreasing ordered schmidt coefficients @xmath117 , @xmath118 .
since @xmath108 is an upper bound for the robustness @xmath7 @xcite , it is possible to use the results presented in @xcite and establish the following order between different measures : @xmath119
as an example , consider the following state of three qubits : @xmath120 , where @xmath121 and @xmath122 .
we have calculated @xmath42 for genuine tripartite entanglement , bipartite entanglement of the possible partitions ( ab - c , a - bc and ac - b ) , and bipartite entanglement of the reduced density matrices ( a - b , a - c and b - c ) .
the results are plotted in fig .
1 . since the state is symmetric with respect to exchange of the particles , the amount of bipartite entanglement in the three partitions and in the three reduced density matrices are the same .
the value of @xmath93 , which was also calculated for the three partitions , and @xmath42 are the same , showing that it is not possible to reach a bound entangled state mixing the identity with this family of states .
note in addition that the the witnessed entanglement gives the same order than the relative entropy of entanglement @xcite and the geometric measure of entanglement @xcite for the w and ghz states . as a function of @xmath123 for the tripartite entanglement ( 1 ) , for the bipartite entanglement in any of the three partitions ( 2 ) , for the bipartite entanglement in any of the three reduced density matrices ( 4 ) , and @xmath93 as a function of @xmath123 for the bipartite entanglement in any of the three partitions ( 3 ) .
in summary , we have introduced a new measure of entanglement which is operational and can be used to quantify all different types of multipartite entanglement .
it has two interesting physical interpretations and is connected to several other different measures .
the authors wish to thank marcelo o. terra cunha for his precious comments and revisions .
partial financial support from the brazilian agencies cnpq , institutos do milnio - informao quntica and fapemig .
99 j. s. bell , _ on the problem of hidden variables in quantum mechanics _ , rev .
phys . * 38 * , 447 - 452 ( 1966 ) ; j. s. bell , _ on the einstein - podolsky - rosen paradox _ ,
physics * i * , 195 - 200 ( 1964 ) .
n. gisin , phys .
a * 210 * , 151 ( 1996 ) .
m. keyl , _ fudamentals of quantum information theory _ ( 2002 ) quant - ph/0202122 ; m. horodecki , p. horodecki and r. horodecki , _ limits for entanglement measures _ , phys . rev . lett . * 84 * , 2014 - 2017 ( 2000 ) . c. h. bennett , h. j. bernstein , s. popescu and b. schumacher , _ concetrating partial entanglement by local operations _ , phys .
a * 53 * , 2046 ( 1996 ) .
m. donald , m. horodecki and o. rudolph , _ the uniqueness theorem for entanglement measures _ , quant - ph/0105017 . c. h. bennett , d. p. divincenzo , j.a .
smolin and w. k. wooters , _ mixed - state entanglement and quantum error correction _ , phys . rev .
a * 54 * , 3824 ( 1996 ) ; v. vedral and m. b. plenio , _ entanglement measures amd purification procedures _ , phys .
a * 57 * , 1619 ( 1998 ) ; g. vidal , _ entanglement monotones _ , journ .
mod . opt . * 47 * , 355 ( 2000 ) .
b. m. terhal , phys .
a * 271 * , 319 ( 2000 ) ; m. lewenstein _
et al _ , phys .
a * 62 * , 052310 ( 2000 ) .
g. vidal and b. tarrach , phys .
a * 59 * , 141 - 155 ( 1999 ) .
g. vidal and r.f .
werner , phys .
a * 65 * , 032314 ( 2002 ) . m. horodecki , p. horodecki and r. horodecki , phys .
lett . a * 223 * , 8 ( 1996 ) .
m. lewenstein , b. kraus , j. i. cirac and p. horodecki , quant - ph/0005014 . l. gurvits , _ proceedings of the thirty - fifth acm symposium on the theory of compunting _
( acm press , new york , 2003 ) , p.p .
f. g. s. l. brando and r. o. vianna , quant - ph/0405063 f. g. s. l. brando and r. o. vianna , quant - ph/0405008 d. bruss , _ characterizing entanglement _ , j. math .
* 43 * , 4237 ( 2002 ) m. a. nielsen , j. kempe , phys .
lett . * 86 * , 5184 - 5187 ( 2001 ) .
p. horodecki , phys .
lett . a * 232 * , 333 ( 1997 ) .
v. vedral and m.b .
plenio , phys .
a * 57 * , 1619 ( 1998 ) t.c .
wei and paul m. goldbart , phys .
a * 68 * , 042307 ( 2003 ) .
m. keyl , phys . rep . * 369 * , 431 ( 2002 ) .
|
we present a new measure of entanglement for mixed states
. it can be approximately computable for every state and can be used to quantify all different types of multipartite entanglement .
we show that it satisfies the some of the usual properties of a good entanglement quantifier and derive relations between it and other entanglement measures .
|
in a 2deg , the equation of motion for the electrons is given by euler s equation @xcite @xmath112 where @xmath113 is the electrons mechanical momentum , @xmath11 their charge , @xmath38 the mean time between scattering events , @xmath114 the external potential energy , and @xmath115 the local pressure .
this hydrodynamic description involves the momentum distribution @xmath116 of the fluid s velocity field , rather than the momenta of individual particles , requiring the use of the convective derivative @xmath117 $ ] . for a noninteracting single - component 2deg
, the internal energy in an area @xmath118 is @xmath119 , where @xmath120 is the number of electrons and @xmath121 the fermi energy . from the energy ,
the local pressure @xmath122 can be determined . because @xmath123 , the force due to the internal fermi pressure is proportional to the gradient of the density , and is ( coincidentally ) directly analogous to the effect of contact interactions in a bec .
the sfhd equations described in the text were derived from the time - dependent gross - pitaevskii equation ( gpe ) @xcite @xmath124 with the order parameter ( or wavefunction ) @xmath125 and inverse mass tensor @xmath126 ( which may in general be anisotropic ) . by writing the order parameter @xmath127 in terms of the density @xmath128 , the phase @xmath129 , and the vector potential @xmath130 ( where the integral is evaluated over the position of the particle @xmath131 ) , and neglecting the quantum pressure term @xcite , the gpe can be reformulated as the continuity and euler equations : @xmath132\label{eq : continuitymethods}\\ 0 = & \
\partial_t \mathbf{p}_{\rm m}(\mathbf{r},t ) + \boldsymbol{\nabla}\left[\tfrac{1}{2}\mathbf{p}_{\rm m}(\mathbf{r},t ) \cdot \check{\mathbf{m}}^{-1}\cdot \mathbf{p}_{\rm m}(\mathbf{r},t ) + u(\mathbf{r},t)\right.\nonumber\\ & \left .
\qquad\qquad\qquad\qquad+gn(\mathbf{r})\right ] \label{eq : eulermethods}+ \tau^{-1 } \mathbf{p}_{\rm m}(\mathbf{r},t),\end{aligned}\ ] ] where @xmath133 is the mechanical momentum and @xmath134 is the canonical momentum . a phenomenological damping constant @xmath38
is also included , which is attributed to loss of atoms from the condensate due to photon scattering from the raman beams .
it can be shown @xcite that such loss , through exchange of atoms with the thermal cloud , results in effective momentum damping within the linear approximation ( below ) in this context , the important property of superfluidity is its irrotationality , and this condition , @xmath135 , places a further constraint upon the dynamics .
when @xmath136 is sufficiently small @xcite and the landau gauge vector potential @xmath137 is assumed , the equilibrium condition ( determined by setting the time derivatives to zero in equations ( [ eq : continuitymethods ] ) and ( [ eq : eulermethods ] ) ) is found when @xmath138 .
the solutions of equations and in equilibrium yield the time - independent _ in situ _ density distribution of the bec , which is of the usual thomas - fermi form @xcite , but with effective trapping frequencies @xmath139 , @xmath140 .
the anisotropy parameter @xmath141 is determined implicitly @xcite for each @xmath56 .
the @xmath56-dependence of these effective trapping frequencies @xmath142 and @xmath143 shifts the resonance frequencies of the effective modes as a function of synthetic magnetic field strength @xmath136 .
as written , equations and apply in any dimension . assuming 2d , a diagonal uniform mass tensor ( @xmath144 ) and the spatially dependent 2d interaction term , equations and are transformed using standard vector identities to equations ( 1 ) and ( 2 ) . in order to compare calculations using the sfhd equations to our data
, we used the three - dimensional ( 3d ) form , with the mean - field interaction parameter @xmath145 ( where @xmath146 is the s - wave scattering length ) , and the anisotropic effective mass tensor @xmath147 where @xmath148 is the effective mass calculated from the modified dispersion relation along @xmath14 , which arises from raman dressing @xcite . to calculate the dynamics under a time - periodic modulation of the trapping potential
, we assumed a perturbative change @xmath149 of the trapping potential energy @xmath150 as described in the main text , and for convenience used a complex time dependence : @xmath151 . under this approximation ,
the linearization of equations and gives @xmath152\label{eq : lincontinuity}\\ 0 = & \partial_t\delta \mathbf{p}(\mathbf{r},t ) + \boldsymbol{\nabla } \left[(\mathbf{p}_0(\mathbf{r } ) - q^*\mathbf{a}^*(\mathbf{r}))\cdot \check{\mathbf{m}}^{-1}\cdot\delta \mathbf{p}(\mathbf{r},t ) \right .
\nonumber\\ & \left .
\qquad\qquad+ \delta u ( \mathbf{r})+g\delta n(\mathbf{r},t)\right]+\tau^{-1 } \delta \mathbf{p}(\mathbf{r},t ) \label{eq : lineuler},\end{aligned}\ ] ] where we modelled the time - dependent density and momentum as the equilibrium distributions plus a small contribution oscillating at the drive frequency : @xmath153 and @xmath154 .
( note that for our experiments @xmath155 . ) in the 3d calculations , the density modulation was assumed to be of the form @xmath156 and the momentum modulation @xmath157 where @xmath158 [ which ensures irrotationality : @xmath159 . when the synthetic field is present ,
non - zero coefficients of the @xmath160 terms are required to obtain a solution and indicate a hall response . to extract the temporal dependence of @xmath161 and @xmath162 , we calculated the responses at equally spaced @xmath163 $ ] , taking the imaginary part of @xmath164 $ ] and @xmath165 $ ] at each @xmath166 to obtain the response to the @xmath167 drive .
to compare these results with our tof measurements , we propagated the calculated _ in situ _ density and momentum distributions through 36.2 ms tof @xcite using the sfhd equations and with @xmath51 , @xmath150 and @xmath1 set to zero .
we projected these numerical solutions for the density profile into a two - dimensional plane to mimic absorption imaging and analyzed them in the same manner as the data .
this theoretical model is fit to the experimentally determined second - order moments ( fig .
3 ) . as an additional check , we compared the experimental results determined by direct integration with results obtained by fitting thomas - fermi profiles to the density distributions @xmath81 and comparing the fit parameters to those expected from the sfhd model . both methods of analysis were in excellent agreement with the model , and each other .
after standard laser cooling and trapping procedures @xcite , our experiments began with nearly pure @xmath58 becs of approximately @xmath59 atoms in the @xmath168 state .
the condensate was trapped at the intersection of two 1064 nm laser beams , propagating along @xmath169 ( with @xmath170 focussed to @xmath171 radii of @xmath172 and @xmath173 ) and @xmath174 ( @xmath175 with @xmath176 ) , crossing near their focii .
we implemented a synthetic vector potential using raman dressing @xcite .
a ( real ) uniform magnetic field @xmath177 zeeman split adjacent @xmath178 levels by @xmath179 .
two raman beams at @xmath180 nm counterpropagated along @xmath181 ( with frequencies @xmath182 ) and coupled the three @xmath178 levels with a strength of @xmath183 for the tof measurements and @xmath184 for the _ in situ _ measurements , where @xmath185 is the recoil energy .
the detuning @xmath186 from raman resonance was set to zero at the centre of the trap .
to realize a synthetic magnetic field , we introduced a spatial dependence to the detuning with a ( real ) magnetic field gradient along @xmath3 such that @xmath187 ; values of @xmath188 range from 0 to 640 hz / m in this experiment .
the resulting vector potential @xmath189 gives a _
synthetic _ magnetic field @xmath190 , with a cyclotron frequency @xmath191 .
the raman dressing in this geometry altered the dispersion relationship only along @xmath14 , giving an effective mass @xmath192 .
we modulated the confining potential energy @xmath193 , with amplitude @xmath194 and frequency @xmath70 to drive an ac current along @xmath14 .
to allow the system to reach steady - state , we smoothly turned on the modulation amplitude in 500 ms from zero to @xmath195 for the tof images , and to @xmath196 for the _ in situ _ images ( a greater modulation amplitude was required to obtain an acceptable signal _ in situ _ ) .
this larger modulation corresponded to a change of the _ in situ _ size of the condensate of @xmath197% peak - to - peak .
we verified that the drive was in the linear response regime with measurements at @xmath69 both less and greater than the values used here , and found that the response was proportional to @xmath69 .
after modulating at full - strength for a time @xmath73 , we prepared the system for imaging : in the final 0.5 ms during which the bec remained trapped , the magnetic field gradient was removed and the magnetic bias field was aligned along @xmath2 for imaging .
this process , rapid compared to the periods associated with trap frequencies , did not noticeably affect @xmath81 . upon removing the optical trapping potential , a weak 100 s optical pumping pulse ,
detuned @xmath198 from the @xmath199 to @xmath200 transition , transferred about 2 % of the atoms ( randomly sampled from the distribution ) to the @xmath201 hyperfine manifold .
these atoms were resonantly imaged in 40 s on the @xmath202 cycling transition .
the peak optical density was generally about @xmath203 .
we obtained the second - order moments described in the text ( equation ( 3 ) ) from the 2d density profiles @xmath87 , shown in fig . 2 .
we checked that the thermal ( non - bec ) component of the cloud did not significantly contribute to the optical density signal in these images by using fits to the appropriate 2d atomic density distribution of a mixed ( mean - field bec @xmath204 thermal ) trapped cloud . since our results were not affected ( within experimental uncertainty ) by including the thermal component
, we opted to ignore it in favour of a simpler analysis .
we determined the region over which we calculated the second - order moments as follows : we fit @xmath205^{3/2}$ ] , with a 2d gaussian , and analyzed those points within the area for which the gaussian s value was @xmath206 of its peak . repeating
this analysis with a @xmath207 limit gave similar results , confirming that the @xmath208 limit effectively captures the entire bec .
for the _ in situ _ data presented in fig . 4 , the absorption of the optical pumping pulse as it propagated through the cloud caused a small asymmetry to the observed @xmath201 density profiles .
to `` smooth over '' this asymmetry , we assume a 3d thomas - fermi density profile of the form : @xmath209 ^ 2\!- \left[\frac{y}{r_y(t)}\right]^2\!-\nonumber\\ & \qquad\qquad\qquad\left[\frac{z}{r_z(t)}\right]^2\!- s_{xy}(t)xy\bigg\},\end{aligned}\ ] ] where @xmath210 is the chemical potential ; @xmath211 is the time - dependent thomas - fermi radius of the cloud along @xmath212 with equilibrium radius @xmath213 ; and @xmath214 is a skew parameter characterizing the hall response , whose equilibrium value is zero .
fits of equation , integrated over @xmath215 , to the observed density give radii @xmath103 , @xmath104 and skew parameter @xmath216 .
exact integration over the density profile ( equation ) gives the second - order moments in terms of the thomas - fermi parameters : @xmath217 $ ] and @xmath218 $ ] .
lock - in analysis is performed on the normalized second - order moments to find the amplitude response , as plotted .
we calculated the `` in - phase '' response by multiplying the time - dependent signals @xmath75 , @xmath76 and @xmath78 by @xmath167 and extracting the average of the product to give @xmath219 , @xmath220 and @xmath221 .
similarly , we calculated the `` out - of - phase '' response by multiplying by @xmath222 .
the displayed uncertainties in these values represent the standard deviation of these means .
we calculated the amplitude response of each moment as the root - sum - square of the in- and out - of - phase responses .
after modulating at full - strength for a time @xmath73 , we suddenly removed the optical trap ( in @xmath223 ) . during the following 2
ms of tof , we adiabatically swept the raman detuning @xmath224 from 0 to @xmath225 , adiabatically transferring ( deloading ) the raman - dressed spin - momentum superposition state to @xmath226 . during the sweep , the subsequent turn - off of the raman beams , and the remainder of tof , the condensate underwent free expansion , initially driven by the mean field .
we absorption - imaged the expanded atomic distribution along @xmath2 after 36.2 ms tof . from the resulting two - dimensional density profile @xmath87 , we determined the second - order moments described by equation ( 3 ) .
the discrete sums were analyzed over a region defined as described above .
to account only for the dynamics in the condensate , we removed the ( @xmath227 to @xmath228 ) thermal background from the images by means of a gaussian fit to the region outside the condensate .
we extracted the lock - in response as above .
|
measurement techniques based upon the hall effect @xcite are invaluable tools in condensed matter physics . when an electric current flows perpendicular to a magnetic field , a hall voltage develops in the direction transverse to both the current and the field . in semiconductors ,
this behaviour is routinely used to measure the density and charge of the current carriers ( electrons in conduction bands or holes in valence bands ) internal properties of the system that are not accessible from measurements of the conventional resistance . for strongly interacting electron systems
, whose behaviour can be very different from the free electron gas , the hall effect s sensitivity to internal properties makes it a powerful tool ; indeed , the quantum hall effects @xcite are named after the tool by which they are most distinctly measured instead of the physics from which the phenomena originate . here
we report the first observation of a hall effect in an ultracold gas of neutral atoms , revealed by measuring a bose - einstein condensate s transport properties perpendicular to a synthetic magnetic field @xcite .
our observations in this vortex - free superfluid are in good agreement with hydrodynamic predictions @xcite , demonstrating that the system s global irrotationality influences this superfluid hall signal . and @xmath0 .
at the microscopic level ( right ) , the electrons move in curved paths under the influence of the magnetic field , and scatter from random impurities ( red crosses ) with a rate @xmath1 ( drude model ) .
* b , * the experiment used a 3d bec , flattened along @xmath2 and elongated along @xmath3 , trapped at the intersection of two 1064 nm laser beams ( pink ) .
a nearly uniform synthetic magnetic field @xmath4 ( green arrow ) was created via the combination of two raman laser beams ( red arrows ) counterpropagating along @xmath5 , and a magnetic field gradient along @xmath3 , making a vector potential @xmath6 .
* c , * we generated a current in the bec by modulating the amplitude of the external trapping potential along @xmath5 .
the bec was alternately relaxed ( top ) and compressed ( bottom ) .
the thick black lines schematically indicate the extent of the bec in the absence of any current and the blue ellipses indicate the regions occupied by the bec as the system was modulated .
when @xmath7 ( left ) , the transport was primarily along @xmath5 , while when @xmath8 ( right ) , the current was deflected along @xmath9 : a superfluid analogue of the hall effect.,width=89 ] microscopically , the hall effect results from the lorentz force @xmath10 experienced by particles with charge @xmath11 and velocity @xmath12 moving in a uniform magnetic field @xmath13 . in the @xmath14-@xmath3 plane perpendicular to @xmath15 , this force acts on a current with density @xmath16 to generate an electrochemical potential gradient @xmath17 normal to @xmath18 , where the hall part of the 2d resistivity tensor @xmath19 is antisymmetric . in conventional metals and semiconductors ,
the hall resistivity @xmath20 is related to the carriers charge @xmath11 and density @xmath21 , but not to the dissipative resistivity tensor @xmath22 , where @xmath23 is the @xmath24 identity matrix .
typically , experiments measure a sample s longitudinal and transverse voltages @xmath0 and @xmath25 ( fig .
[ fig : setup]a ) from which the resistivity tensor can be inferred @xcite . here , we report an analogous transport measurement of the full resistivity tensor , including the antisymmetric contributions from the hall effect , in a flattened elongated bec subjected to a synthetic magnetic field @xmath26 ( in which only the charge - field product @xcite @xmath27 is defined ) . the transport characteristics of systems with many particles and sufficiently strong interactions ( coulomb repulsion in electron gases and plasmas , or @xmath28-wave contact interactions in becs ) resemble those of classical fluids and are described by hydrodynamics .
these hydrodynamics can describe ultracold bose @xcite and fermi @xcite gases , or characterize the collective modes
plasmons @xcite and magnetoplasmons @xcite of 2d electron gases ( 2degs ) .
we show that a bec in a synthetic magnetic field obeys hydrodynamic equations similar to those describing a 2deg in a uniform magnetic field . in a simple drude model @xcite description of a 2deg in a magnetic field ( fig .
[ fig : setup]a ) , scattering from impurities gives rise to resistance while a position - dependent potential @xmath29 controls the electron density .
the collective electron dynamics can be expressed in terms of hydrodynamic continuity and euler equations : @xmath30\label{eq : drudecontinuity}\\ 0 & = \frac{d \mathbf{p}_{\rm m}(\mathbf{r})}{dt}+ \left(\begin{array}{cc } \tau^{-1 } & \text{--}\omega_\mathrm{c } \\
\omega_\mathrm{c } & \tau^{-1 } \end{array}\right ) \mathbf{p}_{\rm m}(\mathbf{r } ) + \boldsymbol{\nabla}\left[u(\mathbf{r } ) + gn(\mathbf{r})\right ] , \label{eq : drudeeuler}\end{aligned}\ ] ] where @xmath31 ; @xmath32\mathbf{j}$ ] is the mechanical momentum of particles with mass @xmath33 , charge @xmath11 , and density @xmath21 ; @xmath34 is the cyclotron frequency for magnetic field strength @xmath35 ; @xmath36 is the potential energy ; @xmath37 accounts for the effects of fermi pressure in this noninteracting 2deg @xcite ; @xmath38 is the momentum relaxation time due to scattering from impurities ; and @xmath39 $ ] is the convective derivative .
the matrix in equation is proportional to the resistivity tensor [ through the factor @xmath40 and has equal diagonal components from @xmath41 ( conventional resistance ) and antisymmetric off - diagonal components from @xmath42 ( hall resistance ) . to compare trapped interacting becs
with the 2degs described above , the superfluid hydrodynamic ( sfhd ) equations may be extended @xcite to include a time - independent artificial magnetic field @xmath43 . the resulting 2d continuity and euler equations
are identical to equations and , complete with the off - diagonal antisymmetric components of the resistivity tensor the hall term proportional to the cyclotron frequency @xmath44 ( see supplementary information ) .
the bec s trap provides the external potential energy @xmath45 with trap frequencies @xmath46 along @xmath47 . the interaction term @xmath48 accounts for the mean - field energy due to contact interactions in the bec . in principle , the resistivity tensor s diagonal components are zero due to the bec s superfluidity ; however , atom loss can imitate damping and we retain a phenomenological dissipative term proportional to @xmath1 .
though the systems with which we work are fully three - dimensional , the dynamics of interest are in the plane perpendicular to @xmath27 and are described by these 2d equations . for quantitative comparisons to data , calculations were performed using the 3d version of equations and ( see supplementary information ) . by expressing the sfhd equations in terms of the gauge - invariant mechanical momentum @xmath49 , which describes the actual motion of particles , we depart from the conventional expressions in terms of the canonical momentum @xmath50 ( where @xmath51 is the artificial vector potential such that @xmath52 ) , which is directly related to the gradient of the superfluid order parameter s phase .
the distinction between @xmath53 and @xmath49 has tangible consequences when considering the condition of irrotationality @xmath54 .
when a synthetic magnetic field is present , irrotationality requires equilibrium fluid flow , with @xmath55 .
this need not be the case at places like vortex cores , where superfluid density is zero ; however , in our trapped bec , vortices are energetically allowed only when @xmath56 is greater than a critical value @xcite .
unless noted , we remained in the irrotational regime where @xmath56 is below this critical value . in this experiment , we studied mass transport in a bec subject to a synthetic magnetic field @xmath57 created using raman dressing @xcite .
while previous experiments used rotating traps to exploit the equivalence between the coriolis and lorentz forces @xcite , methods that use atom - light coupling @xcite instead of rotation allow for greater geometric freedom and potentially increased field strengths @xcite .
the raman technique used here enabled our experiment s highly anisotropic becs , whose 4:1 aspect ratio ( fig .
[ fig : setup]b ) mimicked a typical bar - like hall geometry and facilitated identification of the hall response .
our experiments began with a nearly - pure @xmath58 bec of about @xmath59 atoms in the @xmath60 internal state @xcite confined in a crossed optical dipole trap with frequencies @xmath61 hz ( fig .
[ fig : setup]b ) . laser beams counterpropagating along @xmath14 raman - coupled spin states @xmath62 of differing momentum to create new dressed eigenstates
spin - momentum superpositions @xcite .
the dressed atoms experienced a synthetic vector potential @xcite , which in the presence of an appropriate real magnetic field gradient gave @xmath63 and the associated synthetic charge - field product @xmath64 . to measure transport
, we generated a mass current ( a velocity field ) in the bec by perturbatively modulating the trap along @xmath14 ( fig .
[ fig : setup]c ) .
we modified the potential energy @xmath65 , where @xmath66 is the spring constant along @xmath67 , by adding a time - dependent drive @xmath68 with amplitude @xmath69 and drive frequency @xmath70 .
to ensure linear response , we chose @xmath71 ; to avoid transients , we smoothly increased @xmath69 from zero in @xmath72 . after driving in steady - state for a time @xmath73 , we measured the atomic density , either _ in situ _ or after time - of - flight ( tof ) .
we characterized the system s temporal response to the drive by repeating this procedure for 64 equally spaced values of @xmath73 , the longest of which encompassed eight cycles of modulation . when the geometry of the trap deformation is well - overlapped with one of the system s low - energy collective excitations @xcite eigenmodes of the sfhd equations
the response is enhanced as @xmath70 approaches that mode s resonance frequency .
such collective modes have been measured in both bose and fermi gases @xcite . in this work ,
we predominantly drove the quadrupole - like `` @xmath74 '' eigenmode ( whose frequency is interaction - dependent ) corresponding to compression and expansion primarily along @xmath14 ( an oscillatory @xmath75 response ) , with smaller out - of - phase contributions , due to interactions , along @xmath3 and @xmath2 ( @xmath76 and @xmath77 responses ) .
when the synthetic magnetic field is non - zero , correlated transport along @xmath14 and @xmath3 appears ( an @xmath78 response ) resulting from mixing with the scissors mode @xcite .
this @xmath79-dependent correlated transport is the hall effect .
had we excited the `` @xmath80 '' dipole ( sloshing ) mode , we would have observed the hall effect , but would not have probed internal properties related to interactions , as described by kohn s theorem @xcite .
we absorption - imaged the time - dependent column density distributions @xmath81 either _ in situ _ , to directly determine the density response , or after tof , to determine the momentum response . for _ in situ
_ measurements , we imaged the atoms immediately following release from the trap ( @xmath82 ) . for tof measurements
, we allowed the atoms to expand freely for 36.2 ms before imaging . during the first 2 ms of tof , we deloaded @xcite the raman - dressed atoms into the @xmath83 bare state before removing the raman dressing fields .
these density distributions offer snapshots of the dynamically evolving bec at @xmath73 . for each @xmath73
, we extract three independent second - order moments of the density distribution , @xmath75 , @xmath76 , and @xmath78 : @xmath84 where @xmath85 was measured from the centre of the density distribution @xmath86 .
figure [ fig : measurehall ] shows @xmath87 at three points during the modulation cycle without ( figs .
[ fig : measurehall]a - c ) and with ( figs .
[ fig : measurehall]d - f ) the synthetic magnetic field .
while the compression of the cloud along the drive direction @xmath14 was always the most evident signal , a nonzero synthetic magnetic field ( @xmath88 ) tilted the clouds as the velocity field responded to the drive current , macroscopically manifesting the bec s superfluid hall resistivity . in figs .
[ fig : measurehall]d - f , the hall signal @xmath78 oscillates about zero and reaches its minimum , zero , and maximum , respectively .
the temporal responses of the second - order moments are plotted in figs .
[ fig : measurehall]g ( @xmath89 ) and [ fig : measurehall]h ( @xmath88 ) for @xmath90 .
this modulation frequency is above resonance ( fig .
[ fig : quadandamp ] ) : close enough for a strong signal , but far enough to spectrally resolve the natural mode from the driven mode . while @xmath75 and @xmath76 were not significantly affected with the introduction of the synthetic field , a weak @xmath56-dependence of the parallel @xmath75 and perpendicular @xmath76 moments arises from variations in the effective trapping frequencies as field strength increases @xcite .
figure [ fig : quadandamp ] shows the spectrum of the response across the low - frequency @xmath74 resonance .
we measured the time - dependent second - order moments @xmath75 , @xmath76 , and @xmath78 and used a lock - in technique to determine the response amplitudes @xmath91 , @xmath92 , and @xmath93 for each moment both in - phase and out - of - phase with the drive .
we used the tof method to ensure a strong signal even far from resonance . at a moderate synthetic magnetic field strength [ @xmath94 , no vortices entered and equations , apply .
we simultaneously fit all six @xmath70-dependent responses to the solutions of the linearized sfhd equations ( propagated through tof without linearization , see supplementary information ) and extracted three free parameters : the phenomenological damping term @xmath95 , which is the diagonal part of the resistivity matrix in equation ; the effective trap frequency , @xmath96 hz ; and the drive strength @xmath97 .
the latter two parameters are in reasonable agreement with the directly measured quantities @xmath98 hz and @xmath99 .
while the above data are well - characterized by and are in good agreement with a sfhd description and its tof propagation , _ in situ _ measurements provide a clearer demonstration of hall physics . for different values of @xmath56 , we measured the becs time - dependent density profiles and calculated normalized second - order moments @xmath100 , @xmath101 , and @xmath102 from fits to skewed thomas - fermi distributions with radii @xmath103 and @xmath104 ( see supplementary information ) . as in fig . [ fig : measurehall ] , the modulation drive frequency was @xmath105 hz .
figure [ fig : omegac ] shows the amplitude response of the normalized second order moments along with simultaneous fits to the linear - response predictions of sfhd using a single fit parameter , @xmath106 ( with the previously measured @xmath107 and @xmath1 as fixed parameters ) . for small @xmath56 , we find good agreement between our measurements and sfhd .
the nonlinearity of @xmath108 at high fields is caused by the shift of the @xmath74 resonance away from the fixed drive frequency @xmath70 as @xmath56 increases @xcite .
for small @xmath56 , where the resonance is nearly constant , the slope of @xmath108 vs @xmath56 is a measure of the hall resistivity @xmath109 averaged over the cloud s changing density . the good agreement between our measurements and
the irrotational sfhd model demonstrates the utility of this new hall technique as a simple , quantitative way to probe microscopic properties , like the superfluid hall resistivity , of a quantum degenerate gas . beyond @xmath110 ,
indications of vortices were evident in tof ; therefore , we expect that the predictions of vortex - free sfhd no longer apply . in the large-@xmath79 limit of `` diffused vorticity , '' @xcite
in which many vortices have entered the bec and `` ordinary '' fluid behaviour with @xmath111 is valid , calculations show that the superfluid hall signal approaches zero , but does not become negative as in some superconducting materials @xcite . inspired by condensed - matter techniques , we have demonstrated hall physics in a bec and showed that it is sensitive to the properties of an interacting , irrotational bec .
this macroscopic hall measurement technique can be extended to probe the microscopic properties of more complicated ultracold configurations , such as 2d systems with vortices ( arising from thermal fluctuations @xcite or synthetic magnetic fields @xcite ) , systems with spin - orbit coupling @xcite , and quantum hall systems @xcite .
10 url # 1`#1`urlprefix[2]#2 [ 2][]#2 .
_ _ * * , ( ) . , & .
_ _ * * , ( ) . , & .
_ _ * * , ( ) . , , , &
_ _ * * , ( ) . , & .
_ _ * * , ( ) . .
_ _ * * , ( ) . .
_ _ * * , ( ) . , , & .
_ _ * * , ( ) . , & .
_ _ * * , ( ) . .
_ _ * * , ( ) . & .
_ _ * * , ( ) . & _ _ ( , ) . , , , & .
_ _ * * , ( ) . , & .
_ _ * * , ( ) . .
_ _ * * , ( ) . , , &
_ _ * * , ( ) . .
_ _ * * , ( ) . , , , & .
_ _ * * , ( ) . .
_ _ * * , ( ) . , , & .
_ _ * * , ( ) .
_ et al . _ . _
_ * * , ( ) . .
_ _ * * , ( ) .
_ et al . _ . _
_ * * , ( ) . & .
_ _ * * , ( ) .
_ et al . _ . _
_ * * , ( ) . & .
_ _ ( ) . , , , & . _ _ * * , ( ) . & .
_ _ * * , ( ) . , & .
_ _ * * , ( ) . & _ _ ( , , ) . , & .
_ _ * * , ( ) . & .
_ _ * * , ( ) .
_ et al . _ . _
_ * * , ( ) .
* acknowledgements * we appreciate conversations with j. v. porto .
this work was partially supported by onr ; aro with funds from both darpa s ole program and the atomtronics muri ; and the nsf through the physics frontier center at jqi . l. j. l. acknowledges support from nserc ; k. j .- g . acknowledges conacyt ; and m. c. b. acknowledges nist - arra . + + * author contributions * all authors contributed to writing of the manuscript
. l. j. l. executed the final calculations , and led the data taking effort in which the remaining authors participated .
w. d. p. and i. b. s. supervised this work and i. b. s. proposed the initial measurements in accordance to preliminary numerical and analytic calculations .
+ + * author information * correspondence and requests for materials should be addressed to i. b. s. ( [email protected] ) .
|
the experimental realization of graphene in 2004@xcite opened the gates to a whole scientific field of quasi - twodimensional materials with the promise of novel applications and peculiar physical phenomena by virtue of their reduced dimensionality . among these ,
the transition metal dichalcogenides ( tmd ) of molybdenum and tungsten are an interesting addition due to their intrinsic semiconducting nature and high chemical stability both in bulk and few - layer phases .
analogously to graphite , these materials assume a hexagonal crystal structure with stacked layers of quasi - twodimensional atomic layers that are bound together through van - der - waals forces .
the strong spin - orbit interaction and missing inversion symmetry in the monolayer materials cause a significant split of the valence band edge with an associated valley pseudospin and magnetic moment .
this gives rise to coupled spin- and valley physics in these materials@xcite that manifest , for instance , in optically induced spin and valley polarizations due to valley - dependent circular dichroism and a combination of the spin and valley hall effects .
at the same time , being indirect semiconductors in their bulk forms , the attenuation of interlayer interaction causes a transition from indirect to direct band gap materials between bi- and monolayer tmd structures and a strong enhancement of photoluminescence quantum yield for decreasing material thickness@xcite .
these findings suggest a strong light - matter interaction in these low - dimensional materials , which , apart from the insights from the scientific point - of - view , suggest possible novel applications of two - dimensional tmds in thin and flexible optoelectronic devices , such as photodiodes@xcite and photodetectors@xcite , single - photon emitters or in spin- and valleytronic devices@xcite .
however , their quasi-2d nature is a double - edged sword in the sense that the low thickness limits the absorption efficiency and optical quantum yields of the tmd materials .
a number of strategies have thus recently been reported to increase the light - matter interaction and thus the absorption and emission efficiency by placing tmd materials in optical microcavities and enhancing the coupling of light to excitonic@xcite or trionic@xcite dipoles through formation of cavity polaritons .
another approach is tuning the light - matter interaction through resonant exciton - plasmon coupling in hybrid systems of tmds in plasmonic lattices , _
e.g. _ gold nanoantennas@xcite or silver nanodisk arrays@xcite .
at the heart of these interesting physics are the excited states in the form of excitons and trions .
the low - dimensionality of the tmd systems allows for long - range coulomb interaction channels through the adjacent environment of reduced electric field screening and gives rise to tightly bound exciton and trion states that dominate experimental optical spectra of doped@xcite and undoped@xcite few - layer tmds and have binding energies that are an order of magnitude higher than in conventional bulk systems . on the other hand ,
the corresponding excitonic wavefunctions are fairly extended , with radii on the order of several nm , and can be well described in the framework of mott - wannier theory@xcite .
this mixed wannier - mott and frenkel character of excitons is well - known for carbon nanotubes@xcite and other two - dimensional materials@xcite .
however , the low thickness renders the materials highly sensitive to the environment .
another problem arises from peak broadening from the strong electron - phonon coupling@xcite and makes determination of the exciton binding energy as difference between exciton peak energy and the band edge difficult .
correspondingly , the derived binding energies for the prominent sub - gap excitons in molybdenum and tungsten tmds depend highly on the details of the experimental setup and , to the best of our knowledge , experimental studies of the exciton binding energy have been so far limited to monolayer tmds . a variety of strategies for the measurement of the excitonic binding energies have been reported , including _
( i ) _ application of a modified , nonhydrogenic , rydberg model to one - photon photoluminescence excitation ( ple ) spectra@xcite , and determination of the electronic band edge through _ ( ii ) _ scanning tunneling spectroscopy ( sts)@xcite , _ ( iii ) _
photocurrent spectroscopy@xcite and _ ( iv ) _ two - photon ple@xcite spectroscopy . for monolayer mos@xmath1 ,
this yielded experimental binding energies of the lowest - energy exciton between 220 and 570mev@xcite , while the reported binding energies for other monolayer mo and w dichalcogenides are typically in the range of 0.3 - 1.0ev@xcite . beyond the exciton binding energy ,
several groups have reported experimental studies on the peculiar exciton and trion dynamics and the photocarrier relaxation pathways@xcite , that provide insights into the interplay of excited quasiparticles with the underlying electronic band structure .
this is beneficial for understanding , _
e.g. _ , the mechanism of charge separation of interlayer excitons in photodevices based on heterostructures of stacked tmd materials@xcite . on the theoretical side , a variety of studies of the optical properties of quasi-2d tmds
have been reported based on _ ab initio _ calculations employing the excitonic bethe - salpeter equation and various analytical approaches . here , a main focus was the interpretation of experimental optical spectra and the theoretical derivation of relevant exciton binding energies , bohr radii , and quasi - particle band gaps . in a similar fashion to experiments ,
the reported values show a significant dependence on the computational details : the obtained binding energies of the lowest - energy exciton in mos@xmath1 and similar systems are in the range of 0.2 - 1.0ev@xcite , with the absolute peak positions moving accordingly .
one cause of this significant scattering is the peculiar anisotropic screening of the coulomb interaction in two - dimensional systems that obtains a distinct dependence on the in - plane momentum transfer _ q _ in the vicinity of the _ @xmath3 _ point and requires sufficiently dense _
q_-point grids or analytical modeling to yield reliable results@xcite . based on this
, recent _ ab initio _ studies investigated the exciton radiative lifetimes in mos@xmath1/ws@xmath1 and mose@xmath1/wse@xmath1 heterobilayer structures@xcite and the temperature - dependent effect of electron - phonon coupling on the simulated optical spectra and obtained good quantitative agreement with experiments . on the other hand , analytic methods based on tight - binding or parametrized hamiltonians
lack the flexibility of first principles approaches , but possess the necessary computational simplicity to go beyond static descriptions of isolated excitons and to study higher charge carrier complexes , such as trions@xcite and biexcitons@xcite , exciton dynamics@xcite , carrier density dependent optical spectra@xcite or the excitonic bandstructure@xcite .
however , while the nature of the strongest excitonic transitions has been established for monolayer mos@xmath1 , possible effects of the changed electronic dispersions compared to monolayer mos@xmath1 from different chalcogen species ( as in mose@xmath1 and mote@xmath1 ) or interlayer interactions have so far not been addressed , to the best of our knowledge .
insights into the spatial distribution of the exciton wavefunctions in monolayer and few - layer materials and the differences compared to the well - studied monolayer mos@xmath1 could contribute to the understanding of interlayer excitons in tmd heterostructures and few - layer systems and related experimental results , such as recently discovered interlayer resonant raman modes in few - layer mos@xmath1@xcite .
based on theoretical simulations , we thus analyse in this paper the contribution of excitonic states to the absorption spectra in mono- and bilayer mos@xmath1 , mose@xmath1 , and mote@xmath1 and their distributions in real and reciprocal space . we show that the changes in electronic dispersion due to interlayer interactions and different chalcogen atoms move the dominant region of band nesting between valence and conduction band and affect the composition of the dominant transitions close to the electronic band gap .
further , we confirm the existence of interlayer excitons in all three bilayer materials due to the out - of - plane character of the conduction band orbitals at the band nesting points .
( color online ) photoluminescence spectra of mono- and bilayer mos@xmath1 on si / sio@xmath1 substrates at an excitation energy of 2.33ev .
the sharp peaks just below 2.3ev are the raman modes .
the shown experimental data has been published previously in ref . .
we calculated the groundstate properties of all considered materials from density functional theory ( dft ) within the common perdew - burke - ernzerhof ( pbe ) approximation .
a 13x13x1 k - point sampling and normconserving pseudopotentials , including semicore states for molydenum , together with a cutoff energy of 1200ev were employed on this stage .
the dft electronic bandstructure and wavefunctions served as input for the solution of the excitonic bethe - salpeter equation ( bse)@xcite on discrete 30x30 k - point samplings of the brillouin zone to obtain the dielectric functions including electron - hole effects for the studied dichalcogenides . a sufficient amount of valence and conduction bands to converge the derived optical spectra was included in the calculations .
the electronic eigenvalues from dft were shifted by oneshot g@xmath0w@xmath0 quasiparticle corrections , using 24x24 k - point grids , a cutoff energy of 300ev , and 500/800 bands for the monolayer / bilayer systems .
the dielectric functions and bandstructures of the monolayer systems were corrected _ a posteriori _ for spin - orbit effects following the approach in ref . .
supercells of 20x20x1 unit cells were used for the plots of the excitonic wavefunctions in real space and the position of the hole was fixed at the mo atoms in the center of this unit cell .
further details and parameters for trilayer mos@xmath1 can be found in the supplementary material .
while a number of groups have studied mos@xmath1 in both its mono- and bilayer forms , simulations that include electron - hole effects are to date scarce for mose@xmath1@xcite and mote@xmath1@xcite .
figure [ fig : absorption ] shows the obtained imaginary parts of the dielectric functions from our calculations .
the spectrum of 1l - mos@xmath1 , shown in fig .
[ fig : absorption ] ( a ) agrees well with recent reports@xcite in both shape and energies supports our approach for the other materials as well .
inclusion of electron - hole effects induces a number of prominent excitonic peaks that dominate the low - energy imaginary part of the dielectric function : the direct band gap at the _ k _ point in the brillouin zone gives rise to two peaks , _ a _ and _ b _ , that are split from the local spin - orbit coupling and form the fundamental transitions in experimental optical spectra , see fig .
[ fig : expspectrum ] . we calculate a binding energy of the _ a _ exciton of 0.51ev relative to the fundamental electronic band gap of 2.69ev from our g@xmath0w@xmath0 calculations , which fits well into the range of reported values for the _ a _ exciton of 0.2 - 0.57ev from experiments@xcite and to recent theoretical calculations@xcite .
apart from the transition forming the absorption onset , the dielectric function features a prominent peak at an energy of 2.9ev , which has been previously linked to the prominent _c_ feature in photoluminescence spectra at 2.8ev@xcite and consists of a number of strong optical excitations that we will analyse in sec .
[ sec : decomposition ] . while the obtained binding energies of the _ a _ ( and _ b _ )
excitons show a good agreement with experiment , our calculations systematically overestimate the absolute peak positions , a fact that our calculations share with previous reports .
we attribute parts of this overestimation to the employed g@xmath0w@xmath0 approximation ( we refer to the supplementary material for details ) and the neglect of temperature effects in our calculations .
the latter have been shown recently@xcite to red - shift the predicted a and b peak positions in mos@xmath1 by about 0.1ev .
the dielectric functions of 1l - mose@xmath1 and 1l - mote@xmath1 appear qualitatively similar to that of 1l - mos@xmath1 , but show more structure , with the main _c_ peak being split into a number of subpeaks of similar height , _ ca _ and _
cb_. for mose@xmath1 , these two dominant subpeaks appear at energies of 2.45ev and 2.55ev , above the electronic band gap , while in case of mote@xmath1 the _ ca _ transition moves below the electronic band gap .
this is accompanied by a decrease of the fundamental ( direct ) band gap along the series s@xmath2se@xmath2te due to the increase of the in - plane lattice constants .
this increase weakens the hybridization of mo @xmath4 orbitals with chalcogen @xmath5 orbitals that is responsible for opening of the fundamental band gap@xcite .
correspondingly , the optical band gap decreases along the series as well .
our g@xmath0w@xmath0 calculations predict a direct band gap of 2.33ev for 1l - mose@xmath1 , resulting in a binding energy of the _ a _ exciton of 0.48ev , which is slightly decreased compared to mos@xmath1 and in good agreement with recent studies based on keldysh electrostatic potentials@xcite .
the _ a _ peak appears at an energy of 1.86ev with a spin - orbit - splitting of the _ b _ peak by 166mev .
these values can be compared to the recent report by ugeda _ _ et .
al__@xcite , who studied layers of mose@xmath1 on bilayer graphene by sts and photoluminescence spectroscopy and bse@gw calculations .
experimentally , they obtained an electronic band gap of e@xmath6=2.18ev , and an exciton binding energy of e@xmath7=0.55ev , which places the _ a_-peak position at 1.63ev .
as for mos@xmath1 , our calculated exciton binding energy is thus close to experimental values , while the error in the predicted peak position appears to be contained in an overestimated g@xmath0w@xmath0 band gap compared to experiment . on the other hand , the theoretical calculations in@xcite yielded a larger exciton binding energy of 0.63ev , which compensated for the overestimated band gap ( e@xmath6=2.26ev ) and restored the experimental peak position .
substrate screening was identified to be the main cause of error between theoretical and experimental values in that case and we expect this to add to previously mentioned inaccuracies due to coarse _ k_-point samplings . for mote@xmath1 , we obtain a g@xmath0w@xmath0 band gap of 1.78ev and an _ a _ peak at 1.36ev , which , again , is about 0.2ev higher than the _ a_-peak energy of 1.1ev from photoluminescence experiments reported in a recent study@xcite . while the theoretical calculations in@xcite yielded a g@xmath0w@xmath0 band gap very similar to ours , they predicted a significantly higher exciton binding energy of about 0.6ev that led to a good agreement of the simulated and predicted a peak energies and resembles the earlier results by komsa _ _ et .
the available theoretical and experimental data thus suggest a binding energy of the a exciton between 0.4 and 0.65ev . for the bilayer materials , it is well established that stacking two layers of mos@xmath1 leads to a transition from direct to indirect semiconductor due to the additional interlayer interactions . at the same time , the direct electronic band gap decreases relative to the monolayer forms due to weaker quantum confinement of the electronic wavefunctions . for bilayer mos@xmath1 ,
our g@xmath0w@xmath0 calculations predict a reduction of the direct band gap to a value of 2.37ev .
we note that we did not include corrections for spin - orbit interactions in the bilayer systems .
this causes an underestimation of the splitting of the _ a _ and _ b _ peaks in our calculations , which arises solely from band splitting by virtue of interlayer interactions of 100mev and introduces an additional error on the band gap of order 50mev . on the other hand ,
the binding energy of the _ a _ peak significantly decreases compared to monolayer mos@xmath1 to a value of 0.13ev .
this places the _ a _ peak at an energy of 2.24ev , _
i.e. _ slightly blue - shifted compared to the _ a _ peak in monolayer mos@xmath1 , compare to the simulated absorption spectrum of monolayer mos@xmath1 in absence of spin - orbit corrections in fig .
[ fig : mos2 - 1l - a ] .
this is in good agreement with a recent theoretical study@xcite and matches the slight blueshift in the experimental spectra in fig .
[ fig : expspectrum ] . on the other hand ,
the relative positions of the _ a _ peaks of mono- and bilayer systems appears to depend on the sample and both spectra with blue - shifted@xcite and red - shifted@xcite bilayer _ a _ peak have been reported .
we attribute the reduced binding energies of the _ a _ exciton to the proximity of the adjacent layer , which increases the screening of the coulomb interaction between the electron - hole pair . at the same time , changes in the electronic structure due to the second layer split the high _ c _ peak from monolayer mos@xmath1 into two dominant peaks , _ c _ and _ e _ , at energies of 2.7 and 2.9ev in 2l - mos@xmath1 .
we note that our calculations can only account for direct transitions and thus do not include the low - energy peak _i_ that is observed in photoluminescence spectra due to the fundamental indirect band gap in few - layer mos@xmath1 , refer to fig .
[ fig : expspectrum ] .
inclusion of such indirect transitions would lead to a shift of oscillation strength from the _ a _ exciton to the _i_ transition compared to our simulations .
we find similar results for 2l - mose@xmath1 and 2l - mote@xmath1 .
the binding energy of the _ a _ peak of 2l - mose@xmath1 greatly decreases to a value of e@xmath7=0.14ev compared to the electronic band gap of e@xmath6=2.11ev and the a peak position of 1.97ev is slightly above that of the monolayer form .
the dominant _ c _ peaks from 1l - mose@xmath1 contain a significantly lower oscillation strength in 2l - mose@xmath1 in our calculations , while most of the spectral weight is shifted to a tall peak at higher energies . in 2l - mote@xmath1
, the binding energy of the _ a _ exciton of e@xmath7=0.17ev continues the trend and is slightly larger than for the other two bilayer materials .
a possible reason might be the increasing layer separation along the series 2l - mox@xmath1 with x = s , se , te , that is mirrored in the increase of out - of - plane lattice constants in the bulk materials .
the larger layer separation might weaken the effect of an adjacent layer on the screening that counteracts and cancels the binding energy trend observed for the monolayers . in the following sections we will have a closer look on the detailed compositions of the peaks in the calculated dielectric functions .
for the sake of clarity , we will start with the dielectric function in the absence of spin - orbit effects , as shown in fig .
[ fig : mos2 - 1l - a ] ( a ) . in agreement with expectations , we find that the _ a _ peak is caused by excitations from the valence band maxima at the _ k _ and _ k _ points in the brillouin zone to the lowest conduction band , see fig . [
fig : mos2 - 1l - a ] ( b ) . in real space , this strong confinement to the _ k _ point manifests in a relatively large bohr radius of the exciton wavefunction of about 16 , if we define the bohr radius as full - width - half - maximum ( fwhm ) of the upper envolope and place the hole on the central mo atom in the supercell .
the excited electrons are mainly in @xmath8 and @xmath9 orbitals , which dominate the conduction band edge around _ k _ and confine the exciton wavefunction to the mos@xmath1 plane .
a number of additional transitions with lower dipole - strength appear in the energy range of 2.45 - 2.55 ev . without inclusion of spin - orbit coupling ,
these transitions give rise to two additional features between the _ a _ and _ c _ peaks , see fig .
[ fig : mos2 - 1l - a ] ( a ) .
the first of these transitions is strongly localized in reciprocal space and corresponds to a transition between the valence band maximum and conduction band minimum at the _ k _ point , suggesting that it is indeed the first excited state of the _ a _ transition , _
a@xmath10_. this is confirmed by a node in the real - space representation of the excitonic wavefunction [ plotted in fig .
[ fig : mos2 - 1l - a ] ( c ) ] at a distance of 23.5from the hole .
the second peak corresponds to transitions between the band edges very close to the _ k _ point and is more delocalized in reciprocal space compared to _ a _ and _ a@xmath10_. our analysis reveals two nodes in the excitonic wavefunction at distances of roughly 11.5and 34from the hole , suggesting that this transition is the second excited state , _
a@xmath11 _ , of the _ a _ peak .
( color online ) reciprocal and real space wavefunctions of ( a ) the _ c3@xmath12 _ and ( b ) the _ c4@xmath12 _ transitions contributing to the _ c _ peak in the dielectric function of mos@xmath1 . both dominant contributions and the a transition
are indicated in the electronic bandstructure in ( c ) . as in fig .
[ fig : mos2 - 1l - a ] , we did not include spin - orbit effects for the sake of simplicity . ]
the picture slightly changes if spin - orbit effects are included , see fig .
[ fig : absorption ] ( a ) .
the strong localization at the _ k _ point leads to splitting of _ a _ , _
a@xmath10 _ and _ a@xmath11 _ by 148mev into three pairs of transitions , _ _
a__@xmath2__a__+_b _ , _ _
a@xmath10__@xmath2__a@xmath10__+_b@xmath10 _ and _ _ a@xmath11__@xmath2__a@xmath11__+_b@xmath11_. these transitions occur between pairs of valence and conduction bands ( @xmath13 and @xmath14 ) that satisfy the laporte rule of @xmath15 , @xmath16 for the angular momentum quantum number @xmath17 and the spin quantum number @xmath18 .
this has two consequences : _
( i ) _ transitions are only allowed between spin - matched mo @xmath4 and s @xmath5 orbitals , which satisfy the required conditions .
_ ( ii ) _ the absence of inversion symmetry imposes a reverse order of the spin - orbit split bands at _ k _ compared to the _ k _ point . hence , the excited electrons possess @xmath19 and @xmath20 at the _ k _ and _ k _ points , respectively , in the _ a _ exciton , while @xmath21 and @xmath22 for the _ b _ exciton , giving rise to optical dichroism .
we further find that the dark _
exciton is resonant with the _ b _ transition in our calculations , while _
b@xmath10 _ is almost degenerate to the considerably brighter _
a@xmath11 _ transition .
our calculations thus suggest that the two peaks between _ b _ and _ c _ in fig . [
fig : absorption ] ( a ) do not correspond to _ a@xmath10 _ and _ b@xmath10 _ but to the spin - orbit split excitations of the _ a _ and _ b _ transitions , _ a@xmath11 _ and _
b@xmath11_. the _ c _ peak is made up of a number of transitions in the range of 2.7 - 3.0 ev .
two dark contributions , _
c1@xmath23 _ and _ c2@xmath23 _ ( not shown ) , appear at the lower end of the peak at energies of 2.75 and 2.79ev .
they correspond to @xmath13 transitions in the vicinity of _ k_/_k _ , from points on the _ @xmath3_-_k _ ( _ c1@xmath23 _ ) and _ k_/_k_-_m _ lines ( _ c2@xmath23 _ ) .
the corresponding transitions from @xmath24 to @xmath25 , _
c1@xmath26 _ and _ c2@xmath26 _ , are shifted to higher energies due to the spin - orbit splitting of the valence band top around the _ k _ and _ k _ points . approximating the binding energy as the difference between the transition energy and the electronic band gap on the point of the main contribution on the _ @xmath3_-_k _ or _
k_-_m _ lines yields binding energies on the order of 0.005ev , which suggests that both _ c1@xmath27 _ and _ c2@xmath27 _ are un- or weakly bound electron - hole pairs .
plots of the representation in reciprocal space and the exciton wavefunction of _
c1@xmath12 _ can be found in the supplementary material .
the bright transitions _ c3 _ and _ c4 _ appear at slightly higher energies of 2.86 - 2.92ev and carry almost the entire oscillator strength of the _ c _ peak , see fig .
[ fig : absorption ] ( a ) . despite their energetic similarity
, we found significant differences between the two transitions : the reciprocal space representation in fig .
[ fig : mos2 - 1l - c ] ( a ) shows that _ c3 _ is dominated by contributions from points on the _ k_-_m _
lines , with weaker , six - fold degenerated contributions around a point _
x _ at about @xmath28 of the _
@xmath3_-_k _ lines . as for the other transitions , the close vicinity to _
k _ and _ k _ in reciprocal space leads to a split of _
c3 _ into two transitions _
c3@xmath23 _ and _ c3@xmath26 _ that are separated by about 0.07ev .
the corresponding exciton wavefunction , shown in fig .
[ fig : mos2 - 1l - c ] ( a ) , inherits a distinct out - of - plane character ( not shown ) around the hole position from @xmath29 orbitals that strongly contribute to the conduction band at the _ x _ point , which contrasts the intra - plane nature of the _ a _ and _ b _ excitons .
interestingly , the probablity of presence of the excited electron exhibits what appears like additional rings surrounding a core with radius on the order of 20 , which makes the exciton wavefunction fairly extended .
we find similar features for the c1 transitions .
it is thus possible that higher order excitations , _
e.g. _ of _ a _ , are mixed into the _ c1 _ and _ c3 _ transitions .
in contrast , the second dominant peak , _ c4 _ , is almost entirely composed of six - fold degenerated contributions at _ x _ , with relatively minor additional contributions near _ k _ ( and _ k _ ) , see fig . [
fig : mos2 - 1l - c ] ( b ) . the split transitions _ c4@xmath23 _ and _ c4@xmath26 _ between @xmath30/@xmath24 and @xmath31/@xmath25 are almost degenerate as neither the highest valence nor the lowest conduction bands show a noticeable spin - orbit splitting at _ x _ , leaving only a relatively minor energy separation of 0.013ev from the contributions near _ k _ and _ k_. the relative expansion of the excitonic transition in reciprocal space leads to a considerable confinement of the exciton wavefunction in real space with a small bohr radius of 11compared to that of the _ a _ and _ b _ excitons . correspondingly , we find a increased binding energy of about 0.7ev compared to the _ x _ point . these results for
_ c4 _ are in good agreement with previous reports in@xcite . on the other hand , _
c3 _ has not been described before , to the best of our knowledge .
the similarity of reciprocal space representations and transition energy might suggest that the separation of _
c3 _ and _ c4 _ might be an artifact due to symmetry breaking during our calculations .
we thus tested the effect of changes in the electronic structure or the sampling in the bse kernel , but found our results to be robust .
a number of additional features can be found in an energy window < 3ev in our calculations .
two additional strong transitions , _
c5@xmath26 _ and _ c6@xmath26 _ , form a shoulder to _ c _ at energies around 3ev and each have about half the oscillator strength of _ c3@xmath12_. these peaks
are composed of unbound electron - hole pairs with wavevectors from a triangular - shaped region around the _ k _ and _ k _ points and vanishing contributions from the entire _
@xmath3_-_k _ line .
the spin - orbit splitting pushes _
c5@xmath23 _ to an energy of about 2.92ev , where it is almost degenerate with the _ c3@xmath26 _ transition and thus is expected to be a by - product of the excitation of _ c3@xmath26_. the plateau following the _ c _ peak [ fig .
[ fig : absorption ] ( a ) ] is formed by a number of smaller transitions .
particularly notable are the _
e2@xmath12 _ transitions that appear at an energy of 3.145ev and have the strongest oscillation strength .
they consist mainly of contributions within a circle spanned by the six degenerate _ x _ points and an additional feature at a point _x_ [ fig .
[ fig : mos2 - 1l - c ] ( c ) ] along the _ @xmath3_-_k _ line . correspondingly , the spin - orbit splitting of _
e2@xmath27 _ is relatively minor , and on the order of 0.02ev .
the excitonic wavefunction in real space is well localized within a radius of about 20and indicates a bound excitonic state .
considering the energy difference to _
c4@xmath12 _ and the dipole strength , part of _ e2@xmath12 _ could conceivably be composed of an excited state of _
c4@xmath12 _ with additional contributions around _
@xmath3 _ and on the _
@xmath3_-_k _ lines ; however , our calculations do not indicate a node in the excitonic wavefunction .
we refer to the supplementary material for the reciprocal and real space representations of _
e2@xmath12_. for mose@xmath1 , the _ a _ and _ b _ transitions and their excitations correspond to those of mos@xmath1 by virtue of the similar electronic dispersion around the _ k _ and _ k _ points , with an energy separation of 168mev .
the excitonic wavefunction , depicted in fig .
[ fig : mose2 - 1l - dce ] , matches the one of the _ a _ exciton of mos@xmath1 , with a similar bohr radius of 17 . the energy separation between the _ a_/_b _ pair and the _ c _ peaks decreases compared to mos@xmath1 . as a result , the peak following _ a@xmath11 _ [ see fig . [
fig : absorption ] ( b ) ] does not solely originate from _
b@xmath11 _ as in mos@xmath1 , but has another contribution from the _ d2 _ peak .
this peak corresponds to the _
c1 _ peak in mos@xmath1 and was pushed to lower energies by the decrease of the band around _
k _ and _ k_. fig .
[ fig : mose2 - 1l - dce ] ( a ) shows our calculated reciprocal space and real space representation of the _ d2 _ transition .
similar to _ c1 _ in mos@xmath1 , the excitonic wavefunction suggests a weakly or unbound electron - hole pair . for excitation energies larger than 2.4ev ,
the differences in electronic band structure away from the _ k _ and _ k _ points , induced by exchanging the chalcogen atoms , significantly affect the dielectric function .
the dominant contributions come from the _ c2 _ transitions [ fig .
[ fig : absorption ] ( b ) ] , which are of similar strength to the _ a _ and _ b _ peaks .
transitions give another strong contribution and appear similar to _ c2 _ , but are more delocalized in reciprocal space and have about half the oscillation strength of _ c2 _ , on the same level as the _ a@xmath11 _ and _ b@xmath11 _ transitions .
the plots of reciprocal space representations in fig .
[ fig : mose2 - 1l - dce ] ( b ) , ( c ) reveal two main qualitative differences between the dominant _ c _ transitions in mos@xmath1 and mose@xmath1 : for one , _
c2 _ and ( particularly ) _
c3 _ in mose@xmath1 show a significantly stronger weight of contributions in the vicinity of the _ k _ and _ k _ points compared to _
c3 _ and _ c4 _ in mos@xmath1 , see figs . [
fig : mos2 - 1l - c ] ( a),(b ) and [ fig : mose2 - 1l - dce ] ( c),(d ) .
this causes a noticeable spin - orbit splitting into subtransitions _
c2@xmath12 _ and _ c3@xmath12 _ that give rise to the two peaks _ ca _ and _ cb _ in the dielectric function , with an additional contribution of _ d2@xmath32
_ to _ ca_. the second difference is found for the contributions away from the _ k _ and _ k _ valleys . while these originate mainly from a point closer to _ @xmath3 _ in mos@xmath1 , this six - fold degenerated _ x _ point in mose@xmath1 appears further along the _ @xmath3_-_k _ line and conincides with the _ x _ point in mos@xmath1 [ fig .
[ fig : mos2 - 1l - c ] ( b ) ] . the excitonic wavefunctions are well localized in real space , with somewhat larger extension than the _
c4 _ transition in mos@xmath1 , we derive bohr radii of 10and 19for the _
c2 _ and _ c3 _ transitions , respectively .
a band of transitions of roughly equal strength make up the high _ e _ peak between 2.7 and 2.9ev in the dielectric function , see fig .
[ fig : absorption ] ( b ) .
the strongest contribution comes from the _ e1@xmath26 _
, with an oscillation strength similar to that of _ d2@xmath12 _ or _
c3@xmath12 _ , and with a second transition , _ e2@xmath26 _ , is close in energy . the split - off transitions _ e1@xmath23 _ and _ e2@xmath23 _ contribute to the _ cb _ peak .
both transitions share some similarity to _ c3 _ , with main contributions at points at around @xmath33__k__-_m _ and at the _ x _ point .
the corresponding wavefunctions are relatively extended but well - defined and could thus suggest a bound excitonic state .
( color online ) reciprocal space representations of the ( a ) _ c3@xmath12 _ , ( b ) _
c4@xmath12 _ and ( c ) _
c8@xmath12 _ transitions of mote@xmath1 .
( d ) electronic bandstructure of mote@xmath1 without spin - orbit effects and sketches of the _ a _ , _
c3 _ and _ c4 _ transitions . ] in monolayer mote@xmath1 , the spin - orbit splitting at _ k _ and _ k _ of 220mev is large enough to push the _ b _ transition between the first two excitations of _ a _ , see fig .
[ fig : absorption ] ( c ) .
peak thus consists of a superposition of _ b _ with _
a@xmath10 _ and
_ a@xmath11_. the excitonic wavefunction ( not shown ) assumes the spherical shape seen for mos@xmath1 and mose@xmath1 and we obtain a slightly increased bohr radius of about 17from the fwhm . as for mose@xmath1 ,
the _ c _ peak is split into two subpeaks _ ca _ and _ cb_. _ ca _ mainly consists of the _ c4@xmath23 _ transition with smaller contributions from _
c2@xmath23 _ and _ c3@xmath23_. analysis of the reciprocal space representations show that _
c2@xmath12 _ correspond to the _ d2@xmath12 _ transitions in mose@xmath1 , while _
c3 _ and _ c4 _ show strong similarities to the _ c2@xmath12 _ and _ c3@xmath12 _ excitons in mose@xmath1 , see fig .
[ fig : mote2 - 1l - c ] ( a ) .
transitions at the _ x _ point on the _ @xmath3_-_k _ line contribute less to the overall oscillation strength of the exciton compared to the other materials .
we show in sec .
[ sec : nesting ] that the evolution of the _ x _ point position and its relative contributions along the chalcogen series can be understood from band nesting effects in the electronic band structure . as for the other materials , the excitonic wavefunctions of the _ c3@xmath12 _ and _ c4@xmath12
_ transitions are well localized in real space , but have a larger extension while _ ca _ consists of well - separated excitonic transitions , the composition of _ cb _ appears to be more complex .
we find that the _ cb _ peak is almost completely made up of two bright sub - peaks of large oscillation strength : the higher subpeak at 1.92ev is formed by a superposition of the _ c3@xmath26 _ and _ c4@xmath23 _ transitions with an energy separation of only 0.001ev .
this originates in the smaller energy separation between _
c4@xmath23 _ and _ c4@xmath26 _ of 150mev due to the significantly larger contribution at the _ x _ point , where the spin - orbit splitting of @xmath30/@xmath24 and @xmath31/ @xmath25 is smaller than at _ k_. the second , weaker , sub - peak at 1.89ev consists of the _ c2@xmath26 _ and _ c8@xmath23 _ transitions , which are degenerate ( @xmath34ev ) in our calculations . as fig .
[ fig : mote2 - 1l - c ] ( b ) shows , _
c8@xmath23 _ introduces contributions outside of the vicinity of the _ k _ and _ k _ points , particularly from transitions from @xmath30 and @xmath24 to @xmath31 along the _ k_-_m _ line .
the boundary condition e@xmath35(k)=e@xmath35(-k ) at _ m _ acts similarly to inversion symmetry and , together with time - reversal symmetry e@xmath35(k)=e@xmath36(-k ) , induces a kramer s degeneracy on the one - electron states of different spins , thus lifting the spin - orbit splitting of the valence band maximum . in a similar fashion as for the higher peak , this leads to a slightly decreased energy separation of _
c8@xmath23 _ and _ c8@xmath26 _ ( @xmath37e=150mev ) .
the latter forms the _ d _ shoulder in the dielectric function in fig .
[ fig : absorption ] ( c ) .
additional prominent poles in the dielectric function appear at higher energies .
we refer to the supplementary material for details . due to the weak interlayer coupling compared to intralayer bonding in layered tmds , one would expect a relatively small effect of the material thickness on the qualitative nature of the direct excitations .
a significant effect here comes from the previously mentioned band splitting due to interlayer interactions , particularly at the _ @xmath3 _ and _ q _ points that causes a direct - to - indirect bandgap transitions for 1l@xmath22l , see fig .
[ fig : mos2 - 2l ] ( a ) . in bilayer structures ( or any even layer number ) , the added inversion symmetry due to the ab stacking reverses the order of spin - orbit split bands between layers ( not included in our presented calculations ) and lifts the spin - orbit splitting of the peaks at the absorption onset .
in contrast to the monolayer structure , the _ a _ peak seen in experiments is thus caused by transitions from two degenerate bands of different total angular momentum at the valence band top at _
k_/_k _ to two degenerate conduction bands , while the _ b _ peak originates from the corresponding ( degenerate ) spin - orbit- and interlayer - split bands .
figure [ fig : mos2 - 2l ] ( b ) shows the calculated reciprocal space representation of the _ a _ peak of bilayer mos@xmath1 and the real space excitonic wavefunctions projected onto the x - y and x - z planes , respectively . while the excitonic wavefunction is @xmath18-like as in monolayer , the bohr radius of the _ a _ transition increases to a value of 20as a result of the stronger dielectric screening in the bilayer material .
on the other hand , we find a relatively small spill - over of the excitonic wavefunction into the second layer if an electron - hole pair is excited in the upper layer .
this agrees well with recent resonant raman studies@xcite , where few - layer mos@xmath1 acts as a superposition of @xmath38 quasi - independent layers if in resonance with the _ a _ and _ b _ excitons .
the confinement to one plane can be readily explained by the composition of the band edges at the @xmath39 point , which consist of mo @xmath40 , @xmath41 ( valence band ) or mo @xmath42 ( conduction band ) orbitals that are hybridized with sulfur @xmath8 and @xmath9 states and thus extend mainly in - plane .
peak in the imaginary dielectric function in fig .
[ fig : absorption ] ( d ) consists of a superposition of _
a@xmath11 _ and a second peak of low oscillation strength , _
d1_. interestingly , this darker second peak is a direct transition at the _ @xmath3 _
point , which is energetically closer to the absorption onset due to the strong splitting from to interlayer hybridization of the @xmath29 orbitals that make up the valence band maximum at _
@xmath3_. it appears from its delocalized excitonic wavefunction that this transition does not form a bound electron - hole pair . the strongest single transition in the spectrum is _
c1 _ that makes up the _ c _ peak in the imaginary dielectric function .
it appears to correspond to the _ c3 _ transition in monolayer mos@xmath1 , with a considerable contribution away from the vicinity of the _ k _ and _ k _ points .
however , the off-_k_ contributions appear significantly closer to the _ @xmath3 _ point , see fig . [
fig : mos2 - 2l ] ( c ) , which should render the nature of the transitions more _ @xmath3_-like compared to monolayer mos@xmath1 .
this @xmath29 character of the conduction band results in a strong interlayer nature of the excitonic wavefunction , which significantly spills over the neighbouring layers if the upper layer is excited .
this couples the few - layer structure for optical measurements that are resonant with the _ c _ exciton and , _
e.g. _ activates interlayer resonant raman modes in mos@xmath1@xcite . on the other hand , we find that the modified orbital makeup of the _ c _ transition changes the shape and the extent of the excitonic wavefunction within the plane of the excited mos@xmath1 layer as well . the interlayer character of the _ c _ exciton also carries over to the trilayer system , where the exciton wavefunction is significantly delocalized over all three layers .
in contrast , the exciton wavefunction of the _ a _ exciton is confined to the excited layer , see fig . [
fig : mos2 - 3l ] .
two additional bright transitions of similar oscillation strength give rise to the dominant _ e _ peak in the dielectric function of 2l - mos@xmath1 at an energy around 2.9ev , see fig .
[ fig : absorption ] ( d ) . while bearing some resemblance in reciprocal space to the _ c1 _
transition , the contributions are mainly localized within the hexagonal region around the _
@xmath3 _ point and on the _
@xmath3_-_k _ line and might correspond to the _ c4 _ in monolayer mos@xmath1 .
however , the calculated excitonic wavefunctions appear delocalized within and between the layers , see fig .
[ fig : mos2 - 2l ] ( d ) . our results for bilayer mose@xmath1 draw a similar picture as those for mos@xmath1 .
the strongest transition in our calculations has a similar reciprocal space representation as the _ c3 _ excitons in the monolayer material and forms the _ c _ peak in the imaginary dielectric function [ fig .
[ fig : absorption ] ( e ) ] . as for mos@xmath1
, we find a noticeable out - of - plane component of the excitonic wavefunction .
the main factor is the orbital composition of the involved first valence and conduction bands at the _ x _ points at around @xmath43 of the _ @xmath3_-_k _
line , which consist of selene p@xmath44 and p@xmath45 and mo @xmath4 states , while the first valence band has a larger se p@xmath46 character .
( color online ) reciprocal space representations and corresponding excitonic wavefunction in real space ( projected onto the the @xmath47-@xmath48 plane ) of the _ a _ and _ c _ transitions in trilayer mos@xmath1 corresponding to the _ a _ and _ c1 _ excitons in 2l - mos@xmath1 . as for the other materials ,
the hole was placed on the central mo atom in the upper - most layer . ]
compared to mos@xmath1 and mose@xmath1 , the dielectric function of 2l - mote@xmath1 features a number of bright transitions that dominate the dielectric function below an energy of 2.0ev , see fig .
[ fig : absorption ] ( f ) .
the _ c _ and _ d _ peaks , see fig . [
fig : absorption ] , are dominated by transitions between the two highest valence bands @xmath30 and @xmath24 to the two lowest conduction bands that in reciprocal space show a striking resemblance to _ c3 _ and _ c4 _ in the monolayer . here ,
transition is mainly localized at the _ k _ point , with a relatively low contribution at the _ x _ point away from _
k_. the excitonic wavefunction shows a relatively small bohr radius of 13within the excited layer , but , as for mos@xmath1 and mose@xmath1 , is delocalized over the bilayer structure .
in contrast , the _ d3 _ transition , which dominates the _ d _ peak , has no contribution from _ k _ or its immediate neighbourhood , but mainly consists of transitions at the _ x _ point .
the corresponding excitonic wavefunction appears to be very extended with a radius of more than 25 , indicating a weakly bound electron - hole pair .
a number of higher features appear in the dielectric functions of both 2l - mose@xmath1 and 2l - mote@xmath1 above the bright c transitions and should be accessible through suitable laser energies .
peak in mote@xmath1 is predominantly a transition from @xmath24 to the third conduction band @xmath49 at the _ k _ ( and _ k _ ) point .
the broad _ g _ peak in mose@xmath1 is composed of band transitions that mainly involve the _ @xmath3 _ point .
plots of the reciprocal and real space representations of selected transitions of mose@xmath1 and mote@xmath1 and of the _ e _ transitions in mos@xmath1 can be found in the supplementary information . following fermi s golden rule ,
two factors are important for optical absorption of a solid material : one is the dipole matrix element of a specific transition between valence band @xmath50 and conduction band @xmath51 at a given _ k_-point @xmath52 .
the second factor is the joint density of states @xmath53 , i.e. the number of different absorption channels for a given energy @xmath54 .
the joint density of states between @xmath50 and @xmath51 is commonly defined as @xmath55 where integration is performed over a constant energy surface @xmath56 in reciprocal space@xcite .
it is clear from this definition that major contributions to the joint - density of states come from quasi - stationary regions in reciprocal space where @xmath57 is small or vanishing . on one hand , this is the case at points with a conduction band minimum and valence band maximum , where the individual gradients vanish , i.e. @xmath58 and thus causes strong absorption at the band edges . on the other hand ,
the condition is also fulfilled at points in the bandstructure , where @xmath59 and @xmath60 , but @xmath61 .
carvalho _ _ et al.__@xcite recently suggested by dft calculations that such points of `` band nesting '' give rise to the dominant peaks in the optical conductivity of various transition metal dichalcogenides .
figure [ fig : nesting ] ( a)-(c ) show plots of the gradient of the energy difference between the gw corrected first valence and conduction bands of the three monolayer materials over the first brillouin zone in the absence of spin - orbit interaction . for monolayer mos@xmath1
, we find a range of points with small or vanishing gradient in the brillouin zone , specifically at the high symmetry points and along the _ @xmath3_-_k _ direction .
most notably , the plot reveals a hexagonally shaped area around the _
@xmath3 _ point with a low gradient that , when compared to fig .
[ fig : mos2 - 1l - c ] , fits well to the off-_k _ contributions to the _
c3 _ and _ c4 _ transitions .
another point of vanishing gradient difference appears at @xmath62 of the _ @xmath3_-_k _
line , which originates from a region close to the _ q _ valley in the electronic dispersion , see the bandstructure plot in fig .
[ fig : nesting ] ( a ) , where both conduction and valence band exhibit similar curvature . due to the band gap of about 3.8ev , this region does not noticeably influence the dielectric function up to energies of 3ev , where the _ e2 _ and _ e3 _ transitions contain first contributions from the band nesting point . replacing s with se
has a significant effect on the electronic structure that change the picture , see fig .
[ fig : nesting ] ( b ) .
the relative splitting of valence and conduction band at the _ @xmath3 _ point increases compared to mos@xmath1 , increasing the steepness of the conduction band between _
@xmath3 _ and _ q _ , and simultaneously pushing the valence band maximum at _
@xmath3 _ towards lower energies , thus decreasing the gradient in _
@xmath3_-_q _ direction .
this lifts the band nesting that mos@xmath1 exhibits between _
@xmath3 _ and _ q _ and increases the transition energy relatively to the fundamental electronic band gap .
correspondingly , the contribution of this point to the low - energy part of the absorption spectra is negligible .
this leaves the band nesting point between _ q _ and _ k _ , which is barely affected , as the changes in valence and conduction band dispersion cancel each other . at the same time
, the band gap at _ q _ decreases relatively to _ k _ and brings transitions at this point closer to the low - energy part of the absorption spectrum compared to mos@xmath1 .
the band nesting at this point thus strongly contributes to the dominant _ c2 _ and _ c3 _ transitions in mose@xmath1 , compare fig . [
fig : nesting ] ( b ) and fig .
[ fig : mose2 - 1l - dce ] ( c),(d ) .
this trend continues for mote@xmath1 ; the splitting at _ @xmath3 _ further increases , which counteracts the general lowering of the conduction band compared to the valence band due to the increased interatomic distances and results in a high - gradient difference region around the _
@xmath3 _ point that does not contribute to the low - energy absorption .
the band nesting point
_ x _ between _ q _ and _ k _ also appears for mote@xmath1 and strongly contributes to the _ c3 _ and _ c4 _ transitions in fig .
[ fig : absorption ] ( c ) .
this explains the noticeable similarity in nature of the dominant _ c _ transitions in mote@xmath1 and mose@xmath1 compared to the _ c3 _ and _ c4 _ transitions in mos@xmath1 .
further , the band nesting around _ x _ forms a band of low gradient difference between neighbouring
_ m _ points that surrounds _
k _ and _ k _ and contributes to a relative delocalization of the contributions to excitonic transitions that can , for instance , be seen for the _ c8 _ transition depicted in fig .
[ fig : mote2 - 1l - c ] ( c ) .
on the other hand , we have seen in the previous section that the interlayer interaction in few - layer materials is another source for modifications in the electronic dispersion compared to their monolayer forms that affects the quality of optical excitations . in bilayer mos@xmath1 , this particularly concerns the region around the _
point , where hybdridization of the out - of - plane sulfur @xmath29 orbitals between the two layers induces a splitting of the valence band by 0.72 ev .
this increases the band energy gradient around _
@xmath3_. at the same time , we find that the energy difference between the conduction band edges at _ @xmath3 _ and _ q _ only negligibly changes compared to 1l - mos@xmath1 . as fig .
[ fig : nesting ] ( d ) shows , this shifts the point of band nesting on the _
@xmath3_-_k _ lines and forms a compressed region of low gradient difference between valence and conduction band around _
@xmath3_. the obtained band nesting region fits well to the position of the off-_k _ contributions to the _ c _ exciton in our calculations and thus suggests that these contributions are caused by a superposition of transitions with possibly weak oscillation strength that are amplified by the high joint density of states .
based on _ ab initio _ calculations , we showed that the composition and nature of the excitonic peaks in molydenum dichalcogenides is affected by the chalcogen species and interlayer interactions in the system .
the bandstructures of the studied mono - and bilayer systems exhibit singularities in the joint - density of states between the first valence and conduction bands that are shifted along the _ @xmath3_-_k _ line as a consequence of changes in the electronic dispersion .
this particularly concerns the prominent _ c _ peak that has been observed in mos@xmath1 and should also appear in the other materials as well .
we predict that the moved band nesting point induces a sizeable spin - orbit splitting in the _ c _ peak for mose@xmath1 and mote@xmath1 due to its greater vicinity to the _ k _ point , and a corresponding increasing peak broadening along the series mos@xmath63mose@xmath63 mote@xmath1 .
further , we confirm the noticeable interlayer character of the excitonic wavefunction of the _ c _ peak transition in few - layer mos@xmath1 , as opposed to the intralayer character of the fundamental _ a _ transition , and show that a similar behaviour can be expected for mose@xmath1 and mote@xmath1 as well .
this has interesting implications for experimental measurements on n - layer dichalcogenide materials , which thus could be forced to act more decoupled ( excitation resonant with _ a _ exciton ) or coupled ( excitation resonant with _ c _ exciton . )
the authors gratefully acknowledge the north - german supercomputing alliance ( hlrn ) for providing the computational ressources used for the simulations in this work .
this work was supported by the european research council ( erc ) under grant number 259286 and the german research foundation ( dfg ) within spp1459 `` graphene '' .
the authors thank nils scheuschner for discussions and useful input .
62ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) * supplementary materials : * + * light - matter interactions in twodimensional transition metal dichalcogenides : dominant excitonic transitions in mono- and few - layer mox@xmath1 and band nesting * + +
simulated electronic bandstructures of monolayer mos@xmath1 from g@xmath0w@xmath0 calculations and k - point samplings of varying density .
the dispersion appears converged for a grid 12x12 points in the quasi - two - dimensional brillouin zone , while the effect of denser samplings is a rigid relative shift of conduction and valence bands .
this can be traced back to the preculiar structure of the static dielectric function @xmath64(g , g ) for small * q * , refer to@xcite . ]
the necessary input for the calculations of the optical properties was calculated through density functional theory on the level of the perdew - becke - ernzerhof ( pbe ) approximation as implemented in the quantum espresso package@xcite .
using the atomic code , we generated two sets of scalar- and fully relativistic normconserving pseudopotentials for mo , s , se and te and used a cutoff energy of 1200 ev in our calculations .
here , we included the mo 4s and 4p semi - core electrons in the set of valence electrons , while we found that semi - core states for se and te can be neglected without affecting the results .
integrations in reciprocal space were performed on a discrete grid of 13x13x1 points in the brillouin zone . using these parameters , we optimized the atomic positions and inplane lattice constants until the interatomic forces and the stress was below thresholds of 0.01ev / and 0.01gpa , respectively , while keeping a vacuum layer of at least 25in @xmath51 direction in order to minimize spurious interactions between periodic images .
here , the bi- and trilayer structures were modeled as ab - stacked mox@xmath1 sheets , with inversion symmetry for the bilayer materials . a proper description of the interlayer interaction is crucial@xcite in the few - layer structures to obtain the large valence band splitting at the @xmath3 point that is observed in arpes measurements@xcite . for this purpose , we included semi - empirical van - der - waals corrections from the pbe+d3@xcite scheme , which yields excellent predictions of the in- and out - of - plane lattice constants of layered systems .
our obtained lattice parameters for the studied mono- , bi- and trilayer systems and the bulk materials are reported in table [ tab : latconsts ] .
we then calculated the optical properties using the berkeleygw code@xcite .
the optical spectra including electron - hole interactions were calculated by solving the bethe - salpeter equation ( bse ) , using 4/6/9 valence bands and 6/12/18 conduction bands for monolayer / bilayer / trilayer systems .
the input static dielectric function was calculated using 200/300/350 unoccupied bands for monolayer / bilayer / trilayer systems and an energy cutoff of 300ev .
following the typical approach , the electronic bandstructures from dft were corrected by g@xmath0w@xmath0 quasiparticle energies before entering the diagonalization of the bse kernel matrix .
we employed the static remainder technique@xcite that allowed us to limit the number of empty bands to 500/800/900 in the monolayer / bilayer / trilayer systems while obtaining sufficiently converged transition energies for our purposes .
the divergence of the coulomb potential and spurious interactions with neighbouring cells was treated through a wigner - seitz truncation scheme@xcite . using the hybertsen - louie plasmon - pole model@xcite , we calculated the g@xmath0w@xmath0 corrections for the monolayer and bilayer systems on a grid of 24x24x1 k - points and a cutoff energy of 300 ev for the screened coulomb interaction .
based on a series of test calculations in mos@xmath1 , see fig .
[ fig : gw - mos2 ] and refs.@xcite , we estimate the accuracy of our obtained electronic band gaps to be of order 0.1ev .
we then used wannier interpolation to obtain the quasiparticle correction for each band on the k - point grid used for the calculations of the optical spectra . for better comparability
, we chose to employ 30x30x1 k - point grids for the bse calculations of all studied mono- and bilayer systems , while a 21x21x1 k - point grid and energy cutoffs of 250ev were used for both bse and g@xmath0w@xmath0 calculations of trilayer mos@xmath1 .
spin - orbit effects were included a posteriori into the spectra and the quasi - particle band structures of the monolayer systems following the scheme reported previously by louie _ _ et al.__@xcite .
the bi- and trilayer systems were not corrected as the _ a posteriori _ correction does not appear to work very well in these cases due to the degeneracy of the valence band maximum at the @xmath39 point .
.[tab : latconsts ] obtained lattice constants and layer separations for all considered systems and the bulk materials . [ cols="^,^,^,^",options="header " , ]
for comparison with the results in the main text , we here show additional calculated spectra for the six studied mono- and bilayer molybdenum dichalcogenides . fig . [
fig : dielect ] shows the calculated imaginary parts of the dielectric functions without spin - orbit coupling effects for the monolayer systems . on the other hand , we calculated the frequency - dependent absorption coefficient from the obtained dielectric functions ( including spin - orbit interaction ) and show the resulting simulated absorption spectra in fig .
[ fig : absorption ] .
simulated absorption spectra as derived from the calculated dielectric functions .
spin - orbit interactions were included _ a posteriori_. the peaks were broadened by a lorentzian of width 0.05ev . ]
this subsection contains plots of the reciprocal- and real - space representations of additional transitions of monolayer mos@xmath1 contributing to the _ c _ and _ e _ peaks in the dielectric function , compare to fig . 1 ( a ) in the main manuscript , that might be of interest to the reader .
reciprocal space representations of selected optical transitions in monolayer mos@xmath1 contributing to the _ c _ and e peaks in the dielectric fucntion .
the real - space wavefunctions of _
c1 _ and _ e2 _ are shown on a 20x20 supercell , for the hole being fixed in the center .
the positions of the layers in the @xmath47-@xmath48 plots are indicated by gray dashed lines .
all transitions are sketched in the electronic bandstructures . ]
[ fig : mose2 - 1l ] contains reciprocal space representations for the _ c1 _ and _ d1 _ transitions , which contribute to the _ d _ and _ ca _ peaks in the dielectric function , and of a number of transitions that make up the broad _ e _ peak .
reciprocal space representations of the _ c1 _ , _ d1 _ and several _ e _ transitions in 1l - mose@xmath1 . as in [ fig : mos2 - 1l ] , the transitions are sketched in the bandstructures and the real - space wavefunctions are plotted using a 20x20 supercell . ]
reciprocal space representations of additional transitions and sketches in the electronic band structures for monolayer mote@xmath1 .
the real - space wavefunction for _ a / b _ is shown for the hole being fixed at the central mo atom . ] as fig .
[ fig : mote2 - 1l-1 ] for the main transitions of the _ f _ , _ g _ , _ h _ , and _
j _ peaks in the dielectric function in fig . 1 ( c ) of the main text .
] fig .
[ fig : mote2 - 1l-1 ] shows the reciprocal space representations of additional transitions ( c2 , c6 and c10 ) , which contribute to the two split _ ca _ and _ cb _ peaks in the calculated dielectric function . as mentioned in the main text
, the _ a / b _ and _ a**/b * * _ transitions correspond to the transitions of the same name in mos@xmath1 .
in addition to the transitions mentioned in the main text , a number of strong transitions of negligible binding energy show up at higher energies and form the distinct _
f _ , _ g _ , _ h _ and _ j _ peaks in the imaginary part of the dielectric function ( refer to fig . 1 ( c ) of the main text ) , which might be excited through suitable lasers . _
f@xmath23 _ and _ f@xmath26 _ are electronic transitions between @xmath13 and @xmath13 , respectively , at the _ q _ points and the connection lines , thus forming a hexagonal region in reciprocal space .
g _ and _ h _ transitions , on the other hand , correspond to direct transitions at the _ m _ point and on the _
@xmath3_-_m _ line .
all three features are electronic transitions between the valence and conduction band edges , _
i.e. _ @xmath13 and @xmath13 . due to the low valence band curvature at the _ q _ point ,
see the bandstructures in fig .
[ fig : mote2 - 1l-2 ] , we expect the holes to have a significantly larger effective mass compared to that at _ k _ and _
k_. the _ j@xmath23 _ peak at 3.06ev shows some similarity in shape to the _ c4@xmath12 _ transition in mos@xmath1 and is localized in reciprocal space at six degenerate points at about @xmath28 along the _ @xmath3_-_k _ line .
the high oscillation strength suggests a van - hove singularity in the joint density of states at these reciprocal space points between @xmath65 and @xmath49 . the corresponding _
peak for @xmath66 @xmath67 is split by about 200mev .
[ fig : mos2 - 2l ] shows plots for the _ d1 _ and _ e11 _ transitions mentioned in the main text .
_ d1 _ corresponds to a direct transition at the @xmath3-point , which becomes energetically accessible due to the reduced quantum confinement in the bilayer material compared to monolayer mos@xmath1 .
as mentioned in the main text , the corresponding wavefunction is very delocalized over real - space , both within the excited mos@xmath1 layer and into the neighbouring layer .
the plot shows one of two degenerate transitions : one originating from the upper layer , while the other originates from the lower layer .
the reciprocal space representation of _ e11 _ resembles _
c4 _ in the monolayer structure , but shows a more complicated real - space representation .
reciprocal space representations and real space excitonic wavefunctions ( sliced through the upper molybdenum layer , and projected on the @xmath47-@xmath48 plane ) of selected optical transitions in bilayer mos@xmath1 . as for the monolayer materials , we used a 20x20 supercell to expand the excitonic wavefunction in real space and the hole was fixed at the central mo atom in the upper layer . ] reciprocal space representations and real space excitonic wavefunctions ( sliced through the upper molybdenum layer , and projected on the @xmath47-@xmath48 plane ) of selected optical transitions in bilayer mose@xmath1 . ] here , we show selected transitions from our calculations of bilayer mose@xmath1 ( fig . [ fig : mose2 - 2l ] ) and mote@xmath1 ( fig . [
fig : mote2 - 2l ] ) that did not find space in the main text .
the transitions are labeled after fig . 1 ( e ) and ( f ) .
in both materials , we find _ a _ and _ b _ transitions that correspond to the transitions of same name in 2l - mos@xmath1 and are localized in - plane .
the _ c4 _ ( mose@xmath1 ) and _
( mote@xmath1 ) show a great resemblance to both _
c1 _ in 2l - mos@xmath1 and the dominant _ c _ transitions in the monolayer structures : both transitions have significant off-_k _ components that appear at points of high band nesting ( see fig .
[ fig : nesting ] ( e ) and ( f ) ) .
the excitonic wavefunctions show a distinct localization in real - space and spill over into the non - excited layer , albeit seemingly to a somewhat lesser extent than in 2l - mos@xmath1 , compare _
c4 _ in fig .
[ fig : mose2 - 2l ] and _ c1 _ in fig .
[ fig : mote2 - 2l ] to _ c1 _ in fig . 7 ( c ) of the main text .
similar to bilayer mos@xmath1 , we find transitions at the @xmath3-point for 2l - mose@xmath1 , which form the dominant contributions to the _ g _ peak in the dielectric function .
reciprocal space representations and real space excitonic wavefunctions ( sliced through the upper molybdenum layer , and projected on the @xmath47-@xmath48 plane ) of selected optical transitions in bilayer mote@xmath1 . ]
here , we show additional plots of the energy gradient differences between first valence and conduction bands for bilayer mose@xmath1 and mote@xmath1 and trilayer mos@xmath1 that were derived from the quasiparticle bandstructures . for comparison , the plots for the monolayer systems and bilayer mos@xmath1 were included .
plots of the difference of the energy gradients between the first valence band , v@xmath68 , and the first conduction band , c@xmath68 for mono- and bilayer mos@xmath1 , mose@xmath1 and mote@xmath1 and trilayer mos@xmath1 .
band nesting occurs in regions with low gradient difference ( bright areas ) . ]
14ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( )
|
we report _ ab initio _ calculations of the dielectric function of six mono- and bilayer molybdenum dichalcogenides based in a bethe salpeter equation+g@xmath0w@xmath0 ( bse@g@xmath0w@xmath0)ansatz , focussing on the excitonic transitions dominating the absorption spectrum up to an excitation energy of 3.2ev .
our calculations suggest that switching chalcogen atoms and the strength of interlayer interactions should affect the detailed composition of the high c peaks in experimental optical spectra of molybdenum dichalcogenides and cause a significant spin - orbit - splitting of the contributing excitonic transitions in monolayer mose@xmath1 and mote@xmath1 .
this can be explained through changes in the electronic dispersion around the fermi energy along the chalcogen series s@xmath2se@xmath2te that move the van - hove singularities in the density of states of the two - dimensional materials along the _ @xmath3_-_k _ line in the brillouin zone .
further , we confirm the distinct interlayer character of the _c_ peak transition in few - layer mos@xmath1 that was predicted before from experimental data and show that a similar behaviour can be expected for mose@xmath1 and mote@xmath1 as well .
|
the relation between galactic star formation history and the interstellar medium ( ism ) is recorded in the distribution of elements we observe today .
the principal signatures are : ( a ) time - integrated metallicity distribution functions ( mdfs ) , as exhibited by long - lived stars ; ( b ) instantaneous mdfs at a given time , for example , the present - day dispersion in the inhomogeneous ism ; ( c ) abundance ratios of elements with varying origin and histories ; and ( d ) age - metallicity relations .
a unified understanding of these tracers is fundamental to solving the puzzle of galaxy assembly and evolution . over the 30 years of work on chemical evolution modeling ,
the constraints offered by ( b ) , the instantaneous mdf , have been relatively neglected .
however , tinsley ( 1975 ) showed early on that chemical inhomogeneity significantly affects the other chemical signatures .
edmunds ( 1975 ) also argued that simplistic inhomogeneous evolution of overlapping supernova remnants , with no mixing , yields a present - day ism that is orders of magnitude _ smoother _ than what is presently observed .
this issue is only rarely discussed ( e.g. , roy & kunth 1995 ) . but recently , the significance of dispersions in the chemical signatures has been reemphasized ( e.g. , wyse 1995 ; van den hoek & de jong 1997 ) .
we are therefore overdue to reexamine inhomogeneous chemical evolution .
picking up threads from the early work , i present here an analytic model that predicts the metallicity dispersion from first principles in the limiting case of no mixing . as a reference
, we return to the standard , simple model ( e.g. , tinsley 1980 ; pagel & patchett 1975 ; schmidt 1963 ) , which assumes : ( 1 ) a closed system , ( 2 ) initially metal - free , ( 3 ) constant stellar initial mass function ( imf ) , and ( 4 ) chemical homogeneity at all times . most previous analytic investigations of inhomogeneous evolution were tied to the simple model , and assumed a given dispersion in metal production : tinsley ( 1975 ) adopted a fixed metallicity dispersion and propagated this through the simple model ; while searle ( 1977 ) and malinie et al .
( 1993 ) considered an ensemble of regions that individually follow the simple model such that they yield a given dispersion .
a different , perhaps simpler , approach , returns to the concept of overlapping regions of contamination , as discussed qualitatively by , e.g , edmunds ( 1975 ) .
we adopt the tenets , except for ( 4 ) , of the simple model .
consider an initially metal - free ism , in which a first generation of star forming regions is randomly distributed , occupying a volume filling factor @xmath1 .
we assume that these individual regions have a distribution of metal production @xmath3 , which is the probability density function for obtaining metallicity @xmath4 at any point affected by star formation .
note that it is not necessary to assume anything specific about the sizes of the individual regions , but only the relative total volumes of contaminated to primordial gas , given by @xmath1 .
we now consider @xmath2 subsequent generations of star formation .
the individual regions also fall randomly in the ism , but for each generation , occupy the same filling factor @xmath1 as the first . the role of this last assumption will become apparent in 4 below . the probability that any given point in the ism is occupied by @xmath5 overlapping regions
is given by the binomial distribution in @xmath2 and @xmath1 : @xmath6 and the probability of a point retaining primordial metallicity is , @xmath7 for our purposes , we define the metallicity @xmath4 to be the mass of metals per unit volume ; the coexisting mass density of h is assumed to be spatially uniform . in accordance with our assumption of no mixing , we assume that the metallicity at any given point is the sum of those contributed by the individual overlapping regions @xmath8 : @xmath9 to obtain the instantaneous mdf after @xmath2 generations , we sum the distributions @xmath10 for individual sets of regions occupied by @xmath5 objects , weighted by their @xmath11 : @xmath12 to derive the total mdf of all objects ever created over @xmath2 generations , we must reduce the total gas mass after each generation @xmath13 by a factor @xmath14 , to account for the conversion of gas into stars .
we assume the reduction increment @xmath15 to be constant , so that the present - day gas fraction @xmath16 .
the total , time - integrated mdf is then : @xmath17 where @xmath18 is @xmath11 for a given generation @xmath13 . note that this treatment for gas consumption assumes that @xmath18 remain unaffected by the reduction of gas .
we now assume that @xmath3 remains the same for all generations . for large @xmath5
, the central limit theorem applies , which predicts that the summed mdf approximates a normal distribution with mean and variance given by @xmath5 times the mean @xmath19 and variance @xmath20 of @xmath3 .
thus , @xmath21 is given simply by : @xmath22 where @xmath11 is given by equation [ binomial ] .
likewise , to estimate @xmath23 for large @xmath13 , the binomial distribution approximates a normal distribution with mean and variance given by @xmath24 thus equation [ ntot_gen ] may be written : @xmath25 for small @xmath2 , equations [ nz ] and [ ntot ] break down unless @xmath3 describes a normal distribution .
several considerations suggest that @xmath3 should be given by a power - law . as a rough estimate
, we consider the volume affected by a star - forming region to be that of the superbubble generated by its @xmath26 core - collapse supernovae ( sne ) .
following oey & clarke ( 1997 ; hereafter oc97 ) , the mechanical power @xmath27 , where @xmath28 myr is the life expectancy of @xmath29 for ob associations , and @xmath30 erg is the individual sn energy .
the mechanical luminosity function of @xmath29 is assumed to be a power - law with index @xmath31 ( oc97 , eq . 1 ) .
if the superbubbles grow until they are confined by the ambient pressure , the standard , adiabatic shell evolution implies that their final radii will be related to @xmath29 as @xmath32 ( oc97 , eq .
the total affected volume is therefore , @xmath33 in contrast , the total mass of metals produced by @xmath26 sne is given by , @xmath34 where @xmath35 is the mean yield of metals per sn .
taking @xmath36 , we find that @xmath37 for the superbubble population , showing that the largest superbubbles generate the lowest metallicities , owing to dilution into larger volume .
the mdf for superbubbles is , @xmath38 where @xmath39 is the size distribution for objects at their final radii ( oc97 , eq .
the final relation is obtained for @xmath40 , a value typical , perhaps , for all galaxies ( oey & clarke 1998 ) . here
we see that there are larger numbers of metal - rich objects owing to the larger numbers of smaller superbubbles .
finally , the probability density for @xmath4 at any given point in the ism is : @xmath41 for @xmath40 , we obtain , @xmath42 since @xmath3 is a probability density function , its integral must be unity , and @xmath43 gives the appropriate normalization .
thus , for a given generation of star formation , @xmath3 is weighted toward low - metallicity regions because of their large sizes .
the limiting values @xmath44 and @xmath45 are determined by the corresponding @xmath46 and @xmath47 . for a salpeter ( 1955 )
imf of slope 2.35 and sn progenitor masses between 8 and 120 @xmath48 , we use @xmath49 , based roughly on models by woosley & weaver ( 1995 ) . we take @xmath50 pc , a rather arbitrary value corresponding to the characteristic parameter @xmath51 of oc97 ; and @xmath52 pc , corresponding to individual sn remnants . these yield @xmath53 and
@xmath54 , or , taking o as a tracer of primary elements , minimum and maximum [ o / h ] of 3.0 and 1.3 relative to solar .
the adopted yield of @xmath55 may be somewhat high ( e.g. , woosley & weaver 1995 ) , but the mean @xmath3 may be somewhat modified depending on @xmath44 and @xmath45 .
we can now construct monte carlo models of the mdf for small @xmath2 .
we generate the components @xmath10 by drawing from @xmath3 ( equation [ fz_pwr ] ) @xmath5 times and summing the drawn @xmath8 ( equation [ zsum ] ) .
this is repeated 5000 times to obtain distributions for @xmath10 .
equations [ nz_gen ] and [ ntot_gen ] then provide the instantaneous and total mdfs .
the histograms in figures [ montecarlo]@xmath19 and @xmath56 show these models of @xmath21 and @xmath23 for @xmath57 and @xmath58 , and panels @xmath59 and @xmath60 show the same for @xmath61 generations .
the hatched bar in panels @xmath19 and @xmath59 shows the fraction of primordial gas remaining , as determined by @xmath1 .
the analytic curves show the corresponding results from equations [ nz ] and [ ntot ] . after 2 generations at @xmath58
, we can still see the component power - law distributions in the monte carlo models , and there is gross deviation from the analytic approximation .
however , at only @xmath61 , we can already see how the stochastic model is approaching the analytic version . in panel @xmath62 at @xmath63 ,
the models are in close agreement .
in contrast to this rapid evolution , figure [ montecarlo]@xmath64 demonstrates the effect of a small @xmath1 .
the model for @xmath65 is similar to that in panel @xmath60 for @xmath66 .
although these are not identical , we can see that it is the product @xmath0 that characterizes the evolutionary state , as a result of equation [ binomstats ] .
thus , _ the relative filling factor of contamination has the same importance as the number of contaminating generations .
_ while this statement may seem intuitively obvious , it is worth emphasizing , since the implications are profound .
figure [ halo ] shows data for the outer galactic halo mdf ( laird etal .
1988 ; dash - dot line ) , with [ fe / h ] converted to [ o / h ] following pagel ( 1989 ) .
the solid histogram shows an inhomogeneous monte carlo model for @xmath67 .
there is remarkable similarity between the observations and our rudimentary model , especially in the qualitative shape , having a metal - rich tail and large ( @xmath68 dex ) dispersion .
the data also agree well , coincidentally , with the simple model ( dotted line ; e.g. , tinsley 1980 , equation 4.3 ) for present - day metallicity @xmath69 .
both models take @xmath70 . however , it is essential to note that _ the shapes of the two models result from entirely different processes_. the high - metallicity turnover in the simple model results purely from consumption of gas into stars .
figure [ bulge ] shows the galactic bulge , a relatively evolved system , with data from ibata & gilmore ( 1995 ) .
a simple model having @xmath71 requires @xmath72 to resemble the data ( dotted line ) .
the dashed line shows a similar homogeneous model with no gas consumption , i.e. , a linear relation with equal numbers of stars created at all metallicities ( @xmath73 ) .
the difference between these two homogeneous models is thus entirely due to the depletion of gas into stars , which also applies to the simple model for the halo ( figure [ halo ] ) .
in contrast , the high - metallicity tail of the inhomogeneous halo model represents the vestige of the @xmath3 power - law ; while the low - metallicity turnover results from the central limit theorem , as the component distributions progress toward gaussians ( cf .
figure [ montecarlo ] ) .
this offers an alternative interpretation of the mdf vs. the simple model .
in addition , it reveals the importance of the high - metallicity tail in relatively unevolved systems , for probing the parent distribution @xmath3 .
an evolved inhomogeneous model like the galactic bulge should approach the same limit as the homogeneous simple model , since the metallicities contributed by @xmath3 constitute a progressively smaller fraction of the current @xmath74 of the ism ( edmunds 1975 ) .
figure [ bulge ] indeed shows that all the models essentially coincide in the low - metallicity tail . however , at the high - metallicity end , the form of the inhomogeneous model mainly reflects @xmath75 for the most recent contaminating generations .
these become progressively narrower ( equation [ binomstats ] ) , but do not truncate like the simple model .
the inhomogeneous model matches the observed bulge mdf for @xmath76 ; while the simple model would require @xmath77 to be 10 times larger for @xmath70 , with almost half the distribution at @xmath78 [ o / h ] @xmath79 .
thus , it is especially noteworthy that the inhomogeneous model _ agrees with both the galactic halo and bulge mdfs by varying only the single parameter , nq . _ the predicted shape , dispersion , and mean @xmath74 are linked , thus the simultaneous agreement for these features is highly encouraging .
it is also clear that the supersolar mdf is a vital discriminant between the models .
malinie etal .
( 1993 ) emphasize the importance of reproducing not only the low - metallicity tail , but also the high - metallicity drop - off of the mdf .
figure [ gdwarf ] shows histograms of the empirical galactic disk g - dwarf mdf for stars in solar neighborhood , with [ fe / h ] converted to [ o / h ] as before .
the solid curve shows the analytic inhomogeneous model for @xmath80 ; the dashed line shows a model with a lower yield @xmath81 , which is thereby 10 times more evolved for the given mean metallicity .
finally , the dotted line shows the simple model for @xmath71 .
all models take @xmath82 .
the simple model deviates most strongly from the data , illustrating the `` g - dwarf problem . ''
although the inhomogeneous models do not solve the problem , their mdfs are slightly more peaked , thus more similar to the data .
note also that rocha - pinto & maciel ( 1996 ) truncated their derived distribution below an effective [ o / h ] @xmath83 .
the predicted [ o / h ] fwhm dispersions in the present - day , instantaneous mdf are 0.27 and 0.08 dex for the models with @xmath84 and 320 , respectively .
these values can be compared to the present - day observed dispersion in the solar neighborhood evidenced , for example , by the well - known 0.3 dex deviation of the orion nebula from solar metallicity .
in summary , i present a rudimentary , one - zone model which considers overlapping areas of contamination with no mixing .
this model thus represents a limiting case opposite to the homogeneous simple model .
an under - appreciated point is that the filling factor @xmath1 of contamination is as important as the number of star - forming generations @xmath2 .
the product @xmath0 therefore constitutes a single parameter that describes a system s evolutionary state .
@xmath1 and @xmath2 may be independent , and are roughly associated with the global star formation efficiency and age , respectively .
thus a system with low @xmath1 can result in a present - day metal - poor ism , but having old stars ( e.g. , i zw 18 ) ; and a high @xmath1 can yield an old , and simultaneously , metal - rich population ( e.g. , the galactic bulge ) . by varying _
only _ the parameter @xmath0 , the model can remarkably match both the galactic halo and bulge metallicity distributions , and slightly improve the disk g - dwarf problem .
as expected , relatively unevolved systems are sensitive to the parent @xmath3 .
for example , the high - metallicity tail of the galactic halo mdf is consistent with a power - law @xmath3 .
however , evolved systems are independent of @xmath3 and show progressively decreasing dispersions .
further development of this inhomogeneous model , and investigation of additional empirical constraints , are currently underway .
i am grateful to cathie clarke , mike fall , and rosie wyse for detailed discussions and suggestions , and also to annette ferguson for early conversations .
i also enjoyed discussions with many people during recent conferences .
edmunds , m. g. , 1975 , apss , 32 , 483 ibata , r. a. & gilmore , g. , 1995 , mnras , 275 , 605 laird , j. b. , rupen , m. p. , carney , b. w. , & latham , d. , 1988 , aj , 96 , 1908 malinie , g. , hartmann , d. h. , clayton , d. d. , & mathews , g. j. , 1993 , apj , 413 , 633 oey , m. s. & clarke , c. j. 1997 , mnras , 289 , 570 oey , m. s. & clarke , c. j. 1998 , aj , 115 , 1543 pagel , b. e. j. , 1989 , in _ evolutionary phenomena in galaxies , _
j. e. beckman & b. e. j. pagel ( eds . ) , ( cambridge : cambridge u. press ) , 201 pagel , b. e. j. & patchett , b. e. , 1975 , mnras , 172 , 13 rocha - pinto , h. j. & maciel , w. j. , mnras , 279 , 882 roy , j .-
& kunth , d. , 1995 , aa , 294 , 432 salpeter , e. e. , 1955 , apj , 121 , 161 schmidt , m. , 1963 , apj , 137 , 758 searle , l. , 1977 , in _ the evolution of galaxies and stellar populations _ , b. m. tinsley & r. b. larson , ( eds . ) , ( new haven : yale univ .
obs . ) , 219 tinsley , b. m. , 1975 , 197 , 159 tinsley , b. m. , 1980 , fund . cosmic phys . , 5 , 287 van den hoek , l. b. & de jong , t. , 1997 , aa , 318 , 231 woosley , s. e. & weaver , t. a. , 1995 , apjs , 101 , 181 wyse , r. f. g. , 1995 , in _ stellar populations _ , p. c. van der kruit & g. gilmore , ( eds . ) , iau symp .
164 , ( dordrecht : kluwer ) , 133 wyse , r. f. g. & gilmore , g. , aj , 1995 , 110 , 2771
|
a rudimentary , one - zone , closed - box model for inhomogeneous chemical evolution is offered as an alternative reference than the simple model in the limit of no mixing .
the metallicity distribution functions ( mdfs ) of galactic halo and bulge stars can be matched by varying a _ single _ evolutionary parameter , @xmath0 .
@xmath1 is the filling factor of contaminating regions and @xmath2 is the number of star - forming generations .
therefore , _ q and n have equivalent roles _ , and combinations of @xmath2 and @xmath1 yield systems with different metallicities at any given age .
the model also revises interpretation of observed mdfs .
unevolved systems probe the parent distribution of metal production @xmath3 , for example , the high - metallicity tail of the halo distribution agrees with a power - law @xmath3 . #
1to 0pt#1
|
the issue of universality classes in nonequilibrium statistical systems is often linked to the existence of conservation laws [ ] . normally , the number of conservation laws is finite , leading to the occurrence of dynamically disjoint sectors whose number grows as a power of the system size . however , certain dynamical processes involving composite objects exhibit strongly broken ergodicity , with the number of disjoint sectors growing exponentially with system size , as a result of having an extensive number of conservation laws .
examples studied earlier include deposition and evaporation ( @xmath0 ) [ ] , and diffusion ( @xmath1 ) [ ] of bunches of particles .
if these moves are interpreted as involving trimers or dimers ( in general @xmath2-mers ) , then the @xmath2-mers in question do not keep their identity , and can reconstitute in time .
the dynamical moves in these models do not connect configurations in different sectors , implying that the steady state is not unique and depends strongly on the initial condition .
moreover , the form of the long - time decay of time - dependent correlations in the steady state varies strongly from one sector to another . in one dimension , both the partitioning of phase space into many sectors and the accompanying dynamical diversity
could be understood in terms of a nonlocal construct known as the irreducible string ( is ) which is an invariant of the motion .
the is provided a convenient label for each sector [ ] . moreover ,
the position of the elements of the is were the relevant ` slow ' variables in the problem , and thus governed the long time dynamics in different sectors .
the dynamical diversity found in different sectors ( ranging from different power law decays to stretched exponentials ) could be accounted for in terms of the differences in the is from sector to sector [ ] . in turn , the is can be used to construct an extensive number of conservation laws , although it turns out that these involve nonlocal combinations of the site occupancies [ ] .
all such studies known to us involve a symmetric movement of the is .
here we study the dynamical consequences of an asymmetric ( directed ) motion of the is .
this brings in new phenomena associated with driven diffusive systems , such as kinematic waves and effects of the kardar parisi zhang ( kpz)/burgers nonlinearity , present in the 1-@xmath3 asymmetric simple exclusion process ( asep ) [ ] .
we investigate these effects by studying the directed diffusion of reconstituting dimers ( ddrd ) .
the model resembles the diffusing reconstituting dimer model studied in ref .
[ ] , the only difference being that we allow for only forward motion of the dimers ( @xmath4 ) , in contrast to the two - way motion studied in that paper .
the construction of the is is the same for symmetric / asymmetric motion , but the dynamics of dimers ( and of the is ) is quite different in the directed case .
it turns out that there is a correspondence between the ddrd and the well - studied asep , though of a generalised sort , leading to extra features including additional wheeling motion of sites , oscillations of correlation functions , and other new kinetic effects .
these typically result from the combination of collective driving and the invariant , but moving , is .
interestingly , there is a very useful correspondence between the ddrd and a three - species exclusion process .
further , the problem of forward - moving non - reconstituting ( hard ) dimers which keep their identity is recovered in the ddrd , in a particular sector . as in the previous studies of dynamic diversity ,
monte carlo ( mc ) simulation is a key ingredient in trying to understand the dynamics .
in particular we study time - dependent correlation functions in different sectors and use analytic reductions related to the specific form of the is and the motion of its elements to correlate the diverse behaviour seen in the simulations with properties of the asep .
remarkably close correspondences in nonuniversal as well as scaling properties are seen .
the development begins ( section 2 ) with an introduction to the model and then moves on in the same section to the following sequence of emerging topics : equivalence to the three - species process and correspondences with the asep , whose properties are summarized . in section 3
we study the kinetics within a particularly simple sector in which the problem is tantamount to that of non - reconstituting dimers . new effects such as wheeling of sites in the equivalent asep , resulting from the combination of driving and the composition of dimers , are shown to have important consequences for autocorrelation functions .
these show early oscillations and later a decay , whose form ( exponential or power law ) , is decided by a critical condition related to wheeling and kinematic wave velocities .
the long time behaviour shows scaling , and universality in the sense of data collapse to asep scaling functions , but with sector - specific parameters .
the theoretical and mc results are extended to more general sectors in section 4 , via investigations of sublattice currents , sublattice current - density relations , and consequent kinematic wave velocities .
section 5 is mainly concerned with spatial correlations in a particular sector , involving mc results and their exact analysis via one of the equivalent models .
section 6 is a concluding discussion .
the ddrd model consists of a ring of @xmath5 sites , each of which may be singly occupied ( occupation variable @xmath6 ) or empty ( @xmath7 ) .
any of the @xmath8 possible configurations is then an @xmath5-bit binary string .
the system evolves stochastically through the move @xmath9 , which represents the directed diffusion ( rightward hopping ) , of reconstituting dimers ( ddrd ) .
this is equivalent to stochastic hopping of holes two steps to the left ( i.e. staying on the same sublattice ) provided the intervening site is occupied .
the process is the totally asymmetric alternative to the fully symmetric reconstituting dimer diffusion process ( drd ) [ ] which is known to be strongly nonergodic due to the existence of a conserved string of variables , the is .
all states linked by the dynamic process have the same is , so the phase space divides into many sectors , @xmath10 in number , where @xmath11 is the golden number ( @xmath12 . all these properties are shared by the asymmetric generalisations .
the is for a given sector can be obtained from any state in the sector by deleting from its @xmath5-bit string any pair of adjacent @xmath13 s , and repeating the procedure until no more deletions are possible ( the result is independent of the order of deletion ) .
for instance , the configuration @xmath14 leads to the is @xmath15 . an equivalent representation of configurations , of the is , and of the process , uses characters @xmath16 related to the binary variables by @xmath17 [ ] . with the periodic boundary conditions used here any ddrd configuration
can be uniquely decomposed into a configuration of @xmath18 s , @xmath19 s , and @xmath20 s .
for example , the configuration @xmath21 of the last paragraph can be written as @xmath22 . moreover
, the totally asymmetric dimer hopping move @xmath9 corresponds to either @xmath23 , or @xmath24 ; the former move involves dimer reconstitution .
the is construction now corresponds to deletion of all @xmath18 s , so the is is a string of only @xmath19 s and @xmath20 s , e.g. for the configuration @xmath21 defined above the is is @xmath25 . the absence of any exchange of @xmath19 s and @xmath20 s in the ddrd process verifies the conservation of the is . in terms of the characters @xmath16 ,
the ddrd process has an obvious and important analogy with the ( totally ) asymmetric exclusion process ( asep ) , in which mutually excluding particles hop to nearest neighbour vacancies on the right . in this correspondence
the dimers @xmath18 play the role of the asep particle , and the @xmath19 s , and @xmath20 s of the is correspond to asep vacancies . in order to exploit the correspondence here and later
, we use @xmath26 to denote the numbers of dimers , 10 pairs , and single 0 s , respectively .
then the lengths of the ddrd lattice , the is , and the asep lattice are @xmath27 , and @xmath28 , respectively .
the particle density in the asep is @xmath29 , and the fraction of zeros of c type is @xmath30 .
then , in the case of a periodic is , if the periodic unit [ ... ] of the is contains @xmath31 @xmath19 s and @xmath32 @xmath20 s , we have @xmath33 .
it is straightforward to show from these definitions that @xmath34 .
this ratio provides a metric factor particularly important for converting known currents , tagged hole velocities , and particle densities for the asep into corresponding ddrd quantities .
an important point is that the asep image of a fixed ddrd site is not fixed , but moves forward and wheels around the periodic ring .
the causes and consequences of wheeling are discussed in detail in section 3 .
we conclude the section by recalling some facts about the asep that we will need later . on a ring
, the asep has a product - measure uniform steady state ( ss ) , with current @xmath35 at asep particle density @xmath36 .
density fluctuations move as a kinematic wave through the system [ ] , with velocity @xmath37 .
also , in the long time scaling regime , the asep density - density correlation function @xmath38 is of the form @xmath39 it is known [ eq.(4.8 ) of [ ] ] that in the limit of large @xmath40 , @xmath41 with @xmath42 , so the correlation function is exponentially decaying in time at large time @xmath43 .
the exception is when the co - moving condition @xmath44 applies , a special case being the autocorrelation function ( @xmath45 ) when @xmath46 ( making @xmath47 ) . in these cases
the correlation function shows @xmath48 power law decay in time . in view of the ddrd - asep correspondence
, one might expect similar behaviour for asymmetrically diffusing dimers , unlike symmetrically diffusing dimers of ref .
[ ] , where power law decay is always seen .
detailed considerations of the correspondence show that the results are broadly in agreement with these expectations , but modified by a variety of interesting extra features discussed below .
we denote a sector whose is contains no @xmath19 elements as a null sector . in terms of the notation introduced earlier
, this refers to @xmath49^{{\cal l}}$ ] with @xmath50 and @xmath51 . in any such sector , the ddrd with an even number of particles is isomorphic to directed diffusion of nonreconstituting or ` hard ' dimers .
the reason is that every cluster of 1 s must contain an even number of particles ( otherwise the is would not be all @xmath52 s ) , and hence a cluster can be labeled as @xmath53 , with a @xmath54 at the start and @xmath55 at the end .
each successive @xmath56 pair can be thought of as a dimer , which retains its identity forever : the elementary move @xmath9 is @xmath57 .
the problem of nonreconstituting hard dimers ( and more generally @xmath2-mers ) is of interest in several contexts , and some aspects have been studied earlier [ ] .
this isomorphism allows us to obtain the correlation function for the reconstituting dimer problem , in the special case of the null sector , from the derivation given in section 5 for the non - reconstituting case .
we now summarize the results for dynamic properties of interest in null sectors
. a more detailed account will appear in [ ] . by following the motion of a hole in the ddrd and its image in the asep , we find that the ratio of the currents @xmath58 ( for ddrd ) , and @xmath59 is equal to the metric factor @xmath60 .
evidently , we then have @xmath61 . 0.3 cm _ . _
an important effect shows up in a more detailed consideration of the mapping between sites in the asep and in the ddrd .
a fixed site in the ddrd problem corresponds to a moving site in the asep
it wheels around the ring at a finite mean velocity .
this happens because if we follow the asep image of a single jump @xmath9 in the ddrd , we find that the image of the central site of the triplet advances by one unit , while the images of the other two sites remain unchanged . in time
@xmath43 , the displacement of a mapped site is thus @xmath62 where @xmath63 is the wheeling velocity and @xmath64 is a zero - mean variable which represents the effect of stochasticity in the motion . it can be shown [ ] that in the null sector , the wheeling velocity is given by @xmath65 the wheeling motion of asep sites has been verified by direct observation of mapped site motion in mc simulations , which shows a small jitter corresponding to @xmath64 around the predicted average wheeling velocity
. 0.3 cm _ . _
the considerations of the previous paragraph have important consequences for the ddrd : correlation functions are like those of the asep only if wheeling of sites ( @xmath66 ) is allowed for .
this suggests that in the long time scaling regime , if the effects of jitter arising from the stochastic part @xmath64 can be ignored , the density - density correlation function @xmath67 of the ddrd in the null sector will be of the form @xmath68 \ , t^{-2/3}\,\big)\,.\ ] ] for the autocorrelation function ( @xmath45 ) the argument of @xmath69 is proportional to the difference of the wheeling velocity and the kinematic wave speed @xmath70 of the asep , @xmath71 .
then , if @xmath69 is similar to the scaling function @xmath72 of the asep , the autocorrelation function @xmath73 of the null sector will decay as @xmath74 at large times , where @xmath75 is a constant proportional to @xmath76 . on the other hand , in the special case where @xmath76 vanishes , late - time slower - than - exponential decay
should be expected , that is @xmath77 where @xmath78 falls slowly , e.g. approaching a constant .
notice that @xmath76 vanishes at the ` compensating ' value @xmath79 of the asep particle concentration given by @xmath80 which would correspond to an is @xmath49^{\cal l}\,$ ] of length @xmath81
. 0.3 cm _ . _ to support these expectations , we have conducted extensive simulations of forward - diffusing , reconstituting dimers in a variety of situations . since autocorrelation functions of interest
refer to spontaneous fluctuations in steady state ( ss ) , we avoided the relaxation of the system by generating directly ss configurations , thus saving considerable cpu time . in the equivalent asep , all configurations in ss are equally weighted under periodic boundary conditions and can be translated to ddrd configurations .
this enables us to average our measurements over a large number of samples to reduce scatter in the data . in null sectors with
is of length @xmath82 , such initial conditions are obtained by random deposition of @xmath83 monomers on a ring of @xmath84 sites .
subsequently , each monomer is duplicated by adding another one over an extra adjacent location specially added for that purpose , i.e. @xmath85 , which leave us with a ring of @xmath5 sites and @xmath86 randomly distributed dimers .
notice that this generating procedure is not equivalent to a random sequential adsorption of dimers in the original system . had the latter method been applied it would necessarily introduce correlations between dimers because of shielding effects among them [ ] .
we then evolved our ddrd system using the stochastic microscopic rules referred to above .
after a sequence of @xmath5 update attempts at random locations , the time scale is increased by one unit , i.e. @xmath87 , irrespective of these attempts being successful . also , measurements at shorter ( longer ) time intervals can also be allowed by updating the system in @xmath88 @xmath89 microsteps , and increasing @xmath43 as @xmath90 . typically , we considered rings of @xmath91 sites and averaged our measurements over @xmath92 histories starting from the independent configurations constructed above . in fig.[s0.ps ]
we show the resulting autocorrelation functions for various particle densities @xmath93 in null sectors of length @xmath94 , or equivalently , at different concentrations @xmath95 of the associated asep system .
the full line corresponds to the compensating condition @xmath96 , thus making @xmath97 , and is the case where slower than exponential decay is seen .
specifically , it is consistent with a large time behavior @xmath98 , in turn suggesting that the slow decaying function @xmath99 referred to after eq .
( 3 ) approaches a constant at large @xmath43 .
this is in marked contrast to all other is fractions studied which always give rise to exponential decays , either below or above @xmath79 .
this constitutes a strong confirmation of the theory . in preparation for the analysis of ss currents in more general situations
( see section 4 ) , we also estimated numerically these quantities by measuring the three - point correlations involved in the assisted hopping , namely @xmath100 on each of sublattices @xmath13 and @xmath101 , where @xmath102 denotes a vacancy .
as will be discussed below , the compensating condition also corresponds to the vanishing of both current derivatives with respect to particle densities .
the inset of fig.[s0.ps ] displays the currents of sector @xmath49 $ ] ( both equivalent in this case ) , as a function of the particle density ( @xmath103 ) , closely following the @xmath58 current discussed above in this section .
as expected , its extremum occurs very near to @xmath104 namely , close to @xmath105 .
_ the above results would indicate that the long time autocorrelations of the ddrd and asep systems are similar provided that the kinematic wave in each system is stationary . despite the existence of jitter between site locations in the two systems , fig.[asep.ps ]
provides a strong indication that this might be the case provided large times are considered ( @xmath106 ) . on the other hand ,
the short - time differences between both systems ( see inset of fig.[asep.ps ] ) can be qualitatively understood by combining wheeling with the equal - time spatial structure , which , as will be discussed in section 5 , has a finite ss correlation length .
moreover , analogues with the asep would suggest attempting a collapse of all the null sector ddrd autocorrelation data at @xmath107 .
fig.[colla.ps ] shows the data collapse for the ddrd autocorrelation @xmath108 at representative @xmath36 s in the null sector , as a plot of @xmath109\,$ ] versus @xmath110 .
there is a convincing collapse to a line corresponding to @xmath111 versus @xmath112 , where @xmath72 is the asep scaling function [ eq.([scaling ] ) ] in which @xmath113 . as for @xmath114 and @xmath115 , they are fitting parameters which depend on @xmath36 , and their numerical values [ ] closely follow those arising from the analysis of [ ] .
the argument @xmath40 of @xmath116 and the overall prefactor required to give the autocorrelation function for the asep are given on dimensional grounds as functions of @xmath36 in [ ] and [ ] ( in particular , see eqs.(1.3 ) to ( 1.10 ) in [ ] ) , and involve @xmath117 and @xmath118 together with @xmath43 in forms consistent with the scaling variables given in section 2 .
there are also numerical factors , which differ between the two papers . our data collapse for the ddrd yields values for @xmath114 which for most @xmath36 are within 1 per cent of the corresponding theoretical asep values [ ] ; likewise for @xmath115 , except for a persistent difference of a factor of 2 ( possibly owing to the definition of @xmath119 in [ ] ) .
apart from this , the ddrd autocorrelations correspond in every detail to the asep ones except at early times .
the data in fig.[colla.ps ] do not lie in the asymptotic regime where @xmath111 is linear in @xmath112 .
this exponential decay is achieved only for @xmath120 , as displayed in upper inset ; the corresponding times are very large and the autocorrelations too small as to render them numerically accessible by direct simulation .
size dependences are then noticeable unless @xmath121 is large compared to such times , so very large system sizes are required in the mc .
further , the implied large equilibration times mean that for large @xmath5 it would not be feasible to prepare initial states by relaxation . in our simulations , initial states
are obtained via the asep by exploiting its product measure property , as described above .
in non - null sectors , the value of @xmath122 is a measure of reconstitution that occurs in that sector .
we have examined some sectors with periodic is s and find pronounced oscillations in the autocorrelation functions e.g. figs .
4 and 5 correspond to sectors @xmath123^{{\cal l}/3}$ ] ( @xmath124 ) and @xmath125^{{\cal l}/3}$ ] ( @xmath126 ) , with is s composed of repeating units @xmath127 and @xmath128 respectively .
the time periods for such periodic sectors are readily calculated , since the oscillations are due to the alternations of 1 s and 0 s arriving at a site .
_ in this new scenario the initial steady configurations were prepared by means of a slight variation of the generating procedure described for null strings ( section 3 ) .
consider for instance the sector @xmath123^{{\cal l}/3}$ ] .
then , we can identify two types of ` single site holes ' @xmath129 and @xmath130 , within an asep ring of @xmath131 sites over which @xmath132 ` monomers ' ( later on recast as dimers @xmath133 ) , are randomly adsorbed . in going backwards from asep to ddrd
, we first christen the asep vacancies as @xmath19 or @xmath20 , in the order in which they occur in the is .
while reconstructing the ddrd configuration , we expand out the @xmath19 s to be 10 s and let the @xmath20 s be 0 s .
the resulting configuration is updated by our usual stochastic rules .
this was carried out on ddrd rings of @xmath134 sites and averaged typically over @xmath135 independent histories .
particle densities , @xmath136 and @xmath137 , on even and odd sublattices are preserved throughout [ @xmath138 , whereas both autocorrelations and currents were separately computed in each sublattice .
a similar ss construction and numerical considerations apply to the @xmath125^{{\cal l}/3}$ ] sector , with an evolution which now takes place in two equivalent sublattices of density @xmath139 [ @xmath140
. 0.3 cm _ . _ for sectors such as the above , with periodic is s , a full discussion is possible for currents , compensating conditions , and kinematic waves with velocities given by a density derivative of the appropriate current in the ddrd model .
the compensating condition @xmath141 then corresponds to the vanishing of the kinematic wave velocity given by the current derivative in the ddrd process , at which point we expect slower - than - exponential decays of autocorrelations . in these periodic sectors one can work out analytically the current on each sublattice using the ddrd - asep correspondence given in section 2 , and
hence obtain the critical @xmath79 and the velocities ( details can be found in [ ] ) .
if @xmath142 and @xmath143 are the proportions of zeros on sublattices 1,2 , then the current associated with the movement of zeros on each sublattice is @xmath144 \;,\ ; i = 1,2\,.\ ] ] the corresponding ddrd sublattice kinematic wave velocities @xmath145 are obtained from @xmath146 where @xmath147 are sublattice particle densities ( of 1 s ) in the ddrd .
so @xmath148 vanish at a common value , @xmath79 , of the asep density @xmath36 given by @xmath149 = 0\,$ ] .
this is a quadratic equation for @xmath36 , whose root lying between 0 and 1 is @xmath150/y\,.\ ] ] thus , @xmath79 is a monotonic decreasing function of @xmath151 in the range @xmath152 , varying between 0.5 and @xmath153 . as a check on the result for @xmath79 , for the null sector we have @xmath51 , which implies @xmath154 .
it is easy to obtain the @xmath145 from the @xmath155 by relating the @xmath147 to the particle density @xmath156 in the full ddrd ( from both sublattices ) , and thence to @xmath36 ( in terms of which we have @xmath155 ) .
the result is @xmath157 independent of sublattice label @xmath158 , implying that the two sublattice velocities are the same , as well as the @xmath79 at which they vanish . for comparison with mc results [ eq.([currents ] ) ] , we can obtain similarly , in any sector with a periodic is , the current on sublattice @xmath158 of the ddrd in terms of the particle density @xmath147 on sublattice @xmath158 : @xmath159 /[2 z_i + y \,(1 -\rho_i)]\,.\ ] ] the analytic results for @xmath123^{{\cal l}/3}$ ] and @xmath125^{{\cal l}/3}$ ] , obtained by inserting their respective @xmath151 and @xmath160 , are the full lines in the insets in figs.[bcc.ps ] and [ bbc.ps ] respectively ( skewed parabolae as functions of sublattice densities ) . like the null string currents discussed in section 3 ( inset of fig.[s0.ps ] ) , they agree closely with the data points resulting from the simulations .
the sublattice autocorrelation functions shown in the corresponding main panels are at the densities of the maxima of the respective sublattice currents , and each show a @xmath48 power law long - time decay , verifying in turn that they are at the critical density . moreover , in analogy to null string cases , it was found that on departing from these critical conditions the observed power law decay changes abruptly to exponential .
in the null sector @xmath49^{{\cal l } = n_c}$ ] with an even number of particles , the correlation function in the dimer problem can be evaluated exactly ( below ) using the isomorphism to non - reconstituting dimers together with the @xmath56 representation introduced in section 3 .
since the presence of a @xmath54 particle at site @xmath158 implies and is implied by the presence of an @xmath55 particle at site @xmath161 , the ( unsubtracted ) particle - particle correlation function can be reduced to @xmath162 where @xmath163 is the ( unsubtracted ) correlation function obtained from all configurations having an @xmath55 particle at each of sites @xmath158 and @xmath164 .
the weight of each such configuration can be found by mapping it to the corresponding configuration in the equivalent asep .
allowing for the weights of the @xmath55 s at @xmath158 and @xmath164 and those of @xmath165 dimers and @xmath166 holes between , and for multiplicities , and summing over @xmath165 ( from 0 to the integer part of @xmath167 ) provides the required @xmath163 .
after reductions it becomes @xmath168 \ , [ 1 - ( -x)^{r-1}]\,,\ ] ] ( for details , see [ ] ) .
inserting into eq.([spatial ] ) above , and subtracting @xmath169 ^ 2 $ ] provides the subtracted pair correlation function @xmath170 .
this is oscillatory , and has a decaying envelope with correlation length @xmath171^{-1}\,$ ] .
fig.[ss.ps ] shows mc results for the ( subtracted ) spatial correlations @xmath170 , in the null sector for various @xmath36 .
the actual @xmath170 s show the predicted oscillation @xmath172 , and the correlation lengths agree closely with the theory .
this investigation of the simplest model combining strongly broken ergodicity with driving has uncovered a wealth of properties , some expected and others not foreseen . in particular the model shows an interesting interplay between the collective ( and scaling ) kinetics of the basic driven model ( the asep ) and effects of the invariant is which characterises the sector in the ddrd .
monte carlo simulations shows man of these features in striking fashion . despite the richness and complexity of the combined model ,
much has been understood analytically , for sectors with null or periodic is s .
the scaling behaviour seen so far appears to be in the universality class of the asep .
the relationships of nonuniversal variables , like currents , densities ( including critical densities ) , and velocities etc . , have been largely understood .
there are however certain important generalisations which have not been discussed here , which we now briefly mention : additional special sectors , alternative boundary conditions , and @xmath2-mer generalisations .
we can expect , following the discussion of ref .
[ ] for the drd , new effects in sectors having is s with various types of structural correlation .
provided critical conditions ( @xmath79 s ) exist in which autocorrelations show power law time decay , the power law for such sectors need no longer be related simply to the asep exponents .
one would then have different decay patterns , according to sector ( ` dynamic diversity ' ) . the combination of the methods developed here with those of ref .
[ ] might well allow analytic treatment of such effects . changing from periodic boundary conditions to open ones with boundary injection
could strongly alter the behaviour .
for the asep , that change produces non - trivial spatial correlations and dynamics and , most important of all , a steady state nonequilibrium phase transition .
density profiles , correlations , and dynamics differ in the different phases .
we would expect all these features to occur in the corresponding boundary - driven version of the ddrd , in cases ( like injection / ejection of dimers ) so long as the injection and ejection processes keep the is in tact .
again , the detailed properties would vary from sector to sector , and the methods used here should prove useful in this generalised situation .
finally , generalisations from reconstituting dimers to @xmath2-mers , deserve consideration . as with dimers , the problem of nonreconstituting @xmath2-mers which is of current interest [ ]
is contained in null sectors of the general problem .
the sector - wise mapping to the asep still holds , but the number of sectors is much larger for larger @xmath2 , as multiplicities in the types of asep holes increase with @xmath2 . the sector - wise study of the dynamics remains to be explored systematically .
we are grateful to d. dhar , g. schtz , r. zia , c. godrche , p. ferrari , h. spohn , s. majumdar , s. chatterjee , m. evans , and m. moore for helpful observations and correspondence .
mb s stay at the newton institute was supported through epsrc grant 531174 .
mdg acknowledges support of conicet , argentina , under grants pip 5037 and pict ancypt 20350 .
the work of rbs was supported by epsrc under the oxford condensed matter theory grants gr / r83712/01 and gr / m04426 .
m. prhofer and h. spohn , `` in and out of equilibrium '' , ed .
v. sidoravicius , progress in probability vol .
* 51 * , 185 - 204 , birkhauser , boston ( 2002 ) ; cond - mat/0101200 .
tables of scaling functions are also given by m. prhofer and h. spohn in http://www-m5.ma.tum.de/kpz/ ( 2002 ) .
^{\cal l}\,$ ] with different lengths @xmath82 . from bottom
to top they refer to @xmath173 and @xmath174 ( compensating condition @xmath175 ) .
the latter case is consistent with a @xmath48 large time decay , whereas the other situations give rise to exponential decays .
the inset displays steady currents of null sectors with particle densities @xmath176 , closely following the analytical current @xmath177 derived from section 3 , and reaching a maximum near vanishing wave velocities.,title="fig : " ] -3 cm and @xmath178 ( compensating conditions in both systems ) . above @xmath179 mc steps they both decay as @xmath180 and with very similar amplitudes .
the short time behavior is shown by the inset displaying the ratio @xmath181 which involves the wheeling effect referred to in section 3.,title="fig : " ] -3 cm ^{\cal l}$ ] for non - critical regimes .
they refer respectively to @xmath182 ( squares ) , 1/5 ( circles ) and , 1/3 ( triangles ) .
the curvature of the asep scaling function @xmath183 ( solid line , ref .
[ ] ) , evidences that the collapse achieved through the fitting parameters @xmath184 is yet very far from asymptotia .
the latter is only reached above @xmath185 , as hinted by the actual behavior of @xmath183 exhibited in the inset .
nevertheless , the resulting values of @xmath114 and @xmath115 are understandable in terms of the analysis of @xmath183 given in refs . [ ] along with the wheeling velocity [ eq.([wheel ] ) ] . for displaying purposes , all early time data ( @xmath186 ) have been pruned.,title="fig : " ] -3 cm string with length density @xmath187 .
the large time behavior follows closely a @xmath48 power law decay ( denoted by the rightmost lower dashed line ) . as in fig.[bcc.ps ] , the initial oscillations can be accounted for by the periodicity of the is elements [ 10100 ] .
the inset contains the steady currents ( equivalent in both sublattices ) , which follow entirely their analytical counterparts @xmath188 , in eq.([tcurrents ] ) .
the wave velocity vanishes at the current maximum , on approaching the main panel regime at @xmath189 .
, title="fig : " ] -3 cm ^{\cal l}$ ] for @xmath190 1/2 ( triangles ) , 1/3 ( circles ) , 1/5 ( squares ) , and 1/10 ( rhomboids ) .
the solid lines are fitted with slopes ( inverse correlation length ) @xmath191 , in agreement with eqs.([spatial ] ) and ( [ correlation ] ) .
the actual oscillations of @xmath192 also follow the behavior predicted in section 5 . ]
|
0.2 cm we discuss dynamical aspects of an asymmetric version of assisted diffusion of hard core particles on a ring studied by g. i. menon _ et al .
_ in j. stat phys .
* 86 * , 1237 ( 1997 ) .
the asymmetry brings in phenomena like kinematic waves and effects of the kardar - parisi - zhang nonlinearity , which combine with the feature of strongly broken ergodicity , a characteristic of the model .
a central role is played by a single nonlocal invariant , the irreducible string , whose interplay with the driven motion of reconstituting dimers , arising from the assisted hopping , determines the asymptotic dynamics and scaling regimes .
these are investigated both analytically and numerically through sector - dependent mappings to the asymmetric simple exclusion process .
|
in this section we describe the main results of the paper .
( in case any of the terminology appears to be new to the reader , the corresponding definitions are given in the next section . )
if a contact structure on a @xmath2-manifold is cooriented , then every legendrian knot ( i.e. a knot that is everywhere tangent to the contact distribution ) has a natural framing ( a continuous normal vector field ) . hence when studying legendrian knots in such contact manifolds the main question is to distinguish those of them that realize isotopic framed knots . on the other hand a cooriented contact structure @xmath3 on a manifold @xmath4 gives rise to a nondegenerate vector field @xmath5 in @xmath6 that coorients the contact structure . clearly if two legendrian knots @xmath7 and @xmath8 are isotopic as legendrian knots , then they are also isotopic in the category of knots that are everywhere nontangent to @xmath5 .
this observation leads to the following definition .
let @xmath9 be a contact manifold with a cooriented contact structure , and let @xmath5 be the nondegenerate vector field on @xmath6 that coorients the contact structure .
a knot @xmath7 in @xmath4 is said to be _ pseudo - legendrian _ if it is everywhere nontangent to @xmath5 .
two pseudo - legendrian knots @xmath7 and @xmath8 ( in @xmath10 ) are _ pseudo - legendrian isotopic _ if there exists an isotopy @xmath11\times s^1\rightarrow m$ ] such that @xmath12 , @xmath13 , and @xmath14 $ ] the knot @xmath15 is pseudo - legendrian ( with respect to @xmath5 ) .
clearly if @xmath7 and @xmath8 are legendrian isotopic legendrian knots , then they are also pseudo - legendrian isotopic ( with respect to any nondegenerate vector field @xmath5 that coorients @xmath3 ) .
vassiliev invariants proved to be an extremely useful tool in the study of framed knots , and the conjecture is that they are sufficient to distinguish all the isotopy classes of framed knots .
vassiliev invariants can also be easily defined in the categories of legendrian and of pseudo - legendrian knots .
the study of the groups of vassiliev invariants of legendrian knots was initiated by the work @xcite of d. fuchs and s. tabachnikov where it was proved that the groups of @xmath0-valued vassiliev invariants of legendrian and of framed knots in the standard contact @xmath1 are canonically isomorphic .
later the similar result was proved by j. hill @xcite for the groups of @xmath0-valued vassiliev invariants of legendrian and of framed knots in the spherical cotangent bundle @xmath16 of @xmath17 with the standard contact structure .
the proofs of these isomorphisms were based on the existence of the universal @xmath0-valued vassiliev invariant for these spaces , also known as the kontsevich integral @xcite .
( for @xmath16 such universal invariant was first constructed by v. goryunov @xcite . )
unfortunately the kontsevich integral exists only for a rather limited collection of @xmath2-manifolds .
( recently andersen , mattes , and reshetikhin @xcite constructed such an invariant for manifolds that are @xmath18-fibered over an oriented surface @xmath19 with @xmath20 . ) for this reason the question whether the groups of vassiliev invariants of legendrian and of framed knots are always isomorphic was open for some time .
recently the author used different technique to prove @xcite , @xcite that for any abelian group @xmath21 the groups of @xmath22-valued vassiliev invariants of legendrian and of framed knots are canonically isomorphic for a large class of contact @xmath2-manifolds with a cooriented contact structure .
this class of contact @xmath2-manifolds @xmath9 includes all contact manifolds with a tight contact structure , all contact manifolds that are closed and admit a metric of negative sectional curvature , and all contact manifolds such that the euler class of the contact bundle is in the torsion of @xmath23 . on the other hand @xcite ,
@xcite the author constructed the first known examples of contact manifolds where the groups of vassiliev invariants of legendrian and of framed knots are not canonically isomorphic . in these examples vassiliev invariants of legendrian knots can be successfully used to distinguish legendrian knots that realize isotopic framed knots and that are homotopic as legendrian immersions of @xmath24 .
namely , such examples were constructed for @xmath25 and for any @xmath4 that is an orientable total space of an @xmath24-bundle over a nonorientable surface of genus bigger than one .
this brought up a question what information about legendrian knots can be captured with the help of vassiliev invariants of legendrian knots . here
we answer this questionby proving the following theorem , that says that the groups of vassiliev invariants of legendrian and of pseudo - legendrian knots are always canonically isomorphic .
let @xmath22 be an abelian group , let @xmath9 be a contact @xmath2-manifold with a cooriented contact structure , and let @xmath5 be a nondegenerate vector field that coorients @xmath3 .
let @xmath26 be a connected component of the space of legendrian immersions of @xmath24 and let @xmath27 be a connected component of the space of pseudo - legendrian immersions of @xmath24 ( with respect to @xmath5 ) that contains @xmath26 .
( such a component always exists since a path in @xmath26 corresponds to a path in @xmath27 .
moreover one can show , see proposition [ components ] , that every component of the space of pseudo - legendrian immersions of @xmath24 contains a unique component of the space of legendrian immersions of @xmath24 . )
[ first ] the groups of @xmath22-valued vassiliev invariants of legendrian knots from @xmath26 and of pseudo - legendrian knots from @xmath27 are canonically isomorphic
. see theorem [ isomorphism ] . in particular , if @xmath8 and @xmath28 are two legendrian knots that are homotopic as legendrian immersions and that realize isotopic framed knots , and @xmath29 is a vassiliev invariant of legendrian knots such that @xmath30 , then @xmath8 and @xmath28 are not isotopic as pseudo - legendrian knots .
this means that the only information about a legendrian knot that can be captured using vassiliev invariants of legendrian knots is the pseudo - legendrian isotopy class of the legendrian knot .
we work in the smooth category . in this paper @xmath22
is an abelian group ( not necessarily torsion free ) , and @xmath4 is a connected oriented @xmath2-dimensional riemannian manifold ( not necessarily compact ) . a _ contact structure _ on a @xmath2-dimensional manifold @xmath4 is a smooth field @xmath31 of tangent @xmath32-dimensional planes , locally defined as a kernel of a differential @xmath33-form @xmath34 with non - vanishing @xmath35 .
a manifold with a contact structure possesses the canonical orientation determined by the volume form @xmath36 .
the standard contact structure in @xmath1 is the kernel of the @xmath33-form @xmath37 .
a _ contact element _ on a manifold is a hyperplane in the tangent space to the manifold at a point . for a surface @xmath19
we denote by @xmath38 the space of all cooriented ( transversally oriented ) contact elements of @xmath19 .
this space is the spherical cotangent bundle of @xmath19 and it is one of the classical examples of contact manifolds .
its natural contact structure is the distribution of tangent hyperplanes given by the condition that the velocity vector of the incidence point of a contact element belongs to the element .
a contact structure is _ cooriented _ if the @xmath32-dimensional planes defining the contact structure are continuously cooriented ( transversally oriented ) .
a contact structure is _ oriented _ if the @xmath32-dimensional planes defining the contact structure are continuously oriented . since every contact manifold has a natural orientation we see that every cooriented contact structure is naturally oriented and every oriented contact structure is naturally cooriented .
a contact structure is _ parallelizable _ ( _ parallelized _ ) if the @xmath32-dimensional vector bundle @xmath39 over @xmath4 is trivializable ( trivialized ) .
since every contact manifold has a canonical orientation , one can see that every parallelized contact structure is naturally cooriented .
a contact structure @xmath3 on a manifold @xmath4 is said to be _ overtwisted _ if there exists a @xmath32-disk @xmath40 embedded into @xmath4 such that the boundary @xmath41 is tangent to @xmath3 while the disk @xmath40 is transverse to @xmath3 along @xmath41 .
not overtwisted contact structures are called _
tight_. a _ curve _ in @xmath4 is an immersion of @xmath24 into @xmath4 .
( all curves have the natural orientation induced by the orientation of @xmath24 . ) a _ framed curve _ in @xmath4 is a curve equipped with a continuous unit normal vector field .
a _ legendrian curve _ in a contact manifold @xmath9 is a curve in @xmath4 that is everywhere tangent to @xmath3 .
if the contact structure on @xmath4 is cooriented , then every legendrian curve has a natural framing given by the unit normals to the planes of the contact structure that point in the direction specified by the coorientation . to a legendrian curve @xmath42 in a contact manifold @xmath9 with a parallelized contact structure one can associate an integer that is the number of revolutions of the direction of the velocity vector of @xmath42 ( with respect to the chosen frames in @xmath3 ) under traversing @xmath42 according to the orientation .
this integer is called the _ maslov number _ of @xmath42 .
the set of maslov numbers enumerates the set of the connected components of the space of legendrian curves in @xmath1 ( cf .
[ h - principlelegendrian ] ) . for a contact manifold
@xmath9 with a cooriented contact structure fix a nondegenerate vector field @xmath5 that coorients the contact structure .
a _ pseudo - legendrian curve _ in @xmath43 is a curve that is nowhere tangent to @xmath5 .
clearly every legendrian curve in @xmath9 _ realizes a pseudo - legendrian curve . _
( this means that if @xmath44 is a legendrian curve in @xmath9 , then it is also a pseudo - legendrian curve in @xmath10 . )
a _ knot ( framed knot ) _ in @xmath4 is an embedding ( framed embedding ) of @xmath24 into @xmath4 . in a similar way we define legendrian knots , and pseudo - legendrian knots in a contact manifold
@xmath9 with a cooriented contact structure . a _ singular ( framed ) _ knot with @xmath45 double points is a curve ( framed curve ) in @xmath4 whose only singularities are @xmath45 transverse double points .
an _ isotopy _ of a singular ( framed ) knot with @xmath45 double points is a path in the space of singular ( framed ) knots with @xmath45 double points under which the preimages of the double points on @xmath24 change continuously . in a similar way
we define singular legendrian and pseudo - legendrian knots and the notion of isotopy of singular legendrian knots and of singular pseudo - legendrian knots .
an @xmath22-valued framed ( resp .
legendrian , resp .
pseudo - legendrian ) knot invariant is an @xmath22-valued function on the set of the isotopy classes of framed ( resp .
legendrian , resp .
pseudo - legendrian ) knots .
a transverse double point @xmath46 of a singular knot can be resolved in two essentially different ways .
we say that a resolution of a double point is positive ( resp .
negative ) if the tangent vector to the first strand , the tangent vector to the second strand , and the vector from the second strand to the first form the positive @xmath2-frame .
( this does not depend on the order of the strands ) .
if the singular knot is legendrian ( resp .
pseudo - legendrian ) , then these resolution can be made in the category of legendrian ( resp .
pseudo - legendrian ) knots . a singular framed ( resp .
legendrian , resp .
pseudo - legendrian ) knot @xmath47 with @xmath48 transverse double points admits @xmath49 possible resolutions of the double points .
the sign of the resolution is put to be @xmath50 if the number of negatively resolved double points is even , and it is put to be @xmath51 otherwise .
let @xmath29 be an @xmath22-valued invariant of framed ( resp .
legendrian , resp .
pseudo - legendrian ) knots .
the invariant @xmath29 is said to be of _ finite order _ ( or _ vassiliev invariant _ ) if there exists a nonnegative integer @xmath45 such that for any singular knot @xmath52 with @xmath48 transverse double points the sum ( with appropriate signs ) of the values of @xmath29 on the nonsingular knots obtained by the @xmath49 resolutions of the double points is zero .
an invariant is said to be of order not greater than @xmath45 ( of order
@xmath53 ) if @xmath45 can be chosen as the integer in the definition above .
the group of @xmath21-valued finite order invariants has an increasing filtration by the subgroups of the invariants of order @xmath53 .
[ h - principlelegendrian ] _
@xmath54-principle for legendrian curves .
_ for @xmath9 a contact manifold with a cooriented contact structure , we put @xmath55 to be the total space of the fiberwise spherization of the contact bundle , and we put @xmath56 to be the corresponding locally trivial @xmath24-fibration .
the @xmath54-principle proved for the legendrian curves by m. gromov ( @xcite , pp.338 - 339 ) says that the space of legendrian curves in @xmath9 is weak homotopy equivalent to the space of free loops @xmath57 in @xmath55 .
the equivalence is given by mapping a point of a legendrian curve to the point of @xmath55 corresponding to the direction of the velocity vector of the curve at this point .
in particular the @xmath54-principle implies that the set of the connected components of the space of legendrian curves can be naturally identified with the set of the conjugacy classes of elements of @xmath58 .
[ description]_description of legendrian and of pseudo - legendrian knots in @xmath1 . _
the contact darboux theorem says that every contact @xmath2-manifold @xmath9 is locally contactomorphic to @xmath1 with the standard contact structure that is the kernel of the @xmath33-form @xmath37 .
a chart in which @xmath9 is contactomorphic to the standard contact @xmath1 is called _ a darboux chart . _
legendrian knots in the standard contact @xmath1 are conveniently presented by the projections into the plane @xmath59 .
identify a point @xmath60 with the point @xmath61 furnished with the fixed direction of an unoriented straight line through @xmath59 with the slope @xmath62 .
then the curve in @xmath1 is a one parameter family of points with non - vertical directions in @xmath17 .
while a generic regular curve has a regular projection into the @xmath59-plane , the projection of a generic legendrian curve into the @xmath59-plane has isolated critical points ( since all the planes of the contact structure are parallel to the @xmath62-axis ) .
hence the projection of a generic legendrian curve may have cusps .
a curve in @xmath1 is legendrian if and only if the corresponding planar curve with cusps is everywhere tangent to the field of directions . in particular
this field is determined by the curve with cusps .
a pseudo - legendrian knot in @xmath63 can be depicted as follows .
let @xmath5 be a unit vector field on @xmath1 that coorients the contact structure .
choose a system of coordinates @xmath64 in @xmath1 so that @xmath5 is parallel to the @xmath65-axis and points in the same direction
. then a pseudo - legendrian knot @xmath47 can be depicted by the standard knot diagram in @xmath64 .
since @xmath47 is pseudo - legendrian it means that the velocity vector of @xmath47 at every point is not parallel to the @xmath65-axis .
a pseudo - legendrian isotopy of a pseudo - legendrian knot can be depicted by a sequence of second and third reidemeister moves .
( the first reidemeister move does not occur , since during this move the velocity vector of one of the points on the kink becomes parallel to the @xmath65-axis . )
let @xmath9 be a contact manifold with a cooriented contact structure , let @xmath5 be a nondegenerate vector field that coorients the contact structure , let @xmath26 be a connected component of the space of legendrian curves in @xmath9 , and let @xmath27 be the connected component of the space of pseudo - legendrian curves in @xmath10 that contains @xmath26 .
( such a component exists because a legendrian curve @xmath44 in @xmath9 is pseudo - legendrian in @xmath10 .
moreover as it is shown in proposition [ components ] every component of the space of pseudo - legendrian curves in @xmath10 contains a unique component of the space of legendrian curves in @xmath9 . )
let @xmath66 ( resp .
@xmath67 ) be the group of @xmath22-valued order @xmath53 invariants of legendrian ( resp .
pseudo - legendrian ) knots from @xmath26 ( resp . from @xmath27 ) .
clearly every invariant @xmath68 restricted to the category of legendrian knots in @xmath26 is an element @xmath69 .
this gives a homomorphism @xmath70 .
[ donotdistinguish ] @xmath71 , for every @xmath72 and for every legendrian knots @xmath73 such that @xmath8 and @xmath28 are pseudo - legendrian isotopic knots in @xmath10
. for the proof of theorem [ donotdistinguish ] see section [ proofdonotdistinguish ] .
[ equivalent ] the following two statements i and are equivalent .
: : @xmath74 is an isomorphism .
: : @xmath71 for every @xmath72 and for every legendrian knots @xmath73 such that @xmath8 and @xmath28 are pseudo - legendrian isotopic knots in @xmath10 .
the proof of theorem [ equivalent ] becomes obvious when the mapping from the legendrian isotopy classes of legendrian knot from @xmath26 to the pseudo - legendrian isotopy classes of pseudo - legendrian knots from @xmath27 is surjective .
however the famous bennequin inequality shows that this map is not surjective even when @xmath9 is the standard contact @xmath1 .
the proof of theorem [ equivalent ] is given in section [ proofequivalent ] .
combining theorems [ donotdistinguish ] and [ equivalent ] we get the following .
[ isomorphism ] the groups @xmath66 and @xmath67 of @xmath22-valued vassiliev invariants of legendrian knots from @xmath26 and of pseudo - legendrian knots from @xmath27 are canonically isomorphic .
_ some useful facts proved by d. fuchs and s. tabachnikov @xcite .
_ there are two types of cusps arising under the projection of the part of a legendrian knot that is contained in a darboux chart to the @xmath59-plane ( see [ description ] ) .
they are formed by cusps for which the branch of the projection of the knot going away from the cusp is locally located respectively above or below the tangent line at the cusp point .
( see figure [ twocusp.fig ] a and b respectively . ) for a legendrian knot @xmath47 and @xmath75 we denote by @xmath76 the legendrian knot obtained from @xmath47 by the modification corresponding to an addition of @xmath77 cusp pairs of the first type and @xmath78 cusp pairs of the second type to the projection of the part of @xmath47 located in a darboux chart . 1 : : let @xmath8 and @xmath28 be legendrian knots in the standard contact @xmath1 that realize isotopic unframed knots . then for any @xmath79 large enough there exist @xmath80 such that the legendrian knot @xmath81 is legendrian isotopic to @xmath82 .
2 : : if there exists @xmath83 such that legendrian knots @xmath84 and @xmath85 are legendrian isotopic , then every vassiliev invariant of legendrian knots takes equal values on @xmath8 and on @xmath28 . 3 : : the number @xmath45 from the previous observation exists if the ambient contact manifold is @xmath1 and the legendrian knots @xmath8 and @xmath28 belong to the same component of the space of legendrian curves and realize isotopic framed knots .
the first two observation are true for any contact @xmath2-manifold ( since the proof of the corresponding facts is local ) .
but the number @xmath45 from the statement of the third observation does not exist in general .
in the case of the ambient manifold being @xmath1 fuchs and tabachnikov showed the existence of such @xmath45 using the explicit calculation involving the maslov classes and bennequin invariants of legendrian knots . however in order for the bennequin invariant to be well - defined the knots have to be zero - homologous , and in order for the maslov class to be well - defined the knots have to be zero - homologous or the contact structure has to be parallelizable . clearly to prove theorem [ donotdistinguish ]
it suffices to show that the number @xmath45 from the third observation exists for any legendrian knots @xmath8 and @xmath28 that realize isotopic pseudo - legendrian knots in @xmath10 .
( we assume that the contact structure @xmath3 on @xmath4 is cooriented . ) analyzing the proof of fuchs and tabachnikov one verifies that for @xmath92 large enough the legendrian isotopy @xmath93\rightarrow m$ ] changing @xmath87 to @xmath94 can be chosen so that for every @xmath95 $ ] the legendrian knot @xmath96 is contained in a thin tubular neighborhood @xmath97 of @xmath98 and is isotopic ( as an unframed knot ) to @xmath99 inside @xmath97 .
every pseudo - legendrian knot is naturally framed .
put @xmath100 and @xmath101 to be the framed knots corresponding respectively to @xmath102 and to @xmath103 for two framed knots @xmath100 and @xmath104 realizing unframed knots that are isotopic inside @xmath97 there is a well - defined @xmath105-valued obstruction to be isotopic inside @xmath97 in the category of framed knots .
this obstruction is the difference of the self - linking numbers of the inclusions of @xmath106 and @xmath104 into @xmath1 induced by an identification of @xmath97 with the standard solid torus in @xmath1 .
( one verifies that for @xmath106 and @xmath104 that are isotopic as unframed knots inside @xmath97 this difference does not depend on the choice of the identification of @xmath97 with the standard solid torus in @xmath1 . ) from the formula for the bennequin invariant stated in @xcite one gets that the value of the obstruction for @xmath87 to be isotopic as a framed knot to @xmath8 inside @xmath107 is equal to @xmath108 . similarly the value of the obstruction for @xmath94 to be isotopic as a framed knot to @xmath28 inside @xmath109 is equal to @xmath110 .
clearly the value of the obstruction for @xmath100 to be isotopic to @xmath103 inside @xmath97 does not depend on @xmath46 , and we get that @xmath89
. identify @xmath24 with the circle in @xmath0 that consists of the numbers of the absolute value one .
put @xmath111 $ ] , @xmath112 $ ] , @xmath113 $ ] , and @xmath114 $ ] to be the four nonoverlapping arcs of @xmath24 with the end points at @xmath115 .
1 : : @xmath117}$ ] is a homotopy of loops that changes @xmath8 to @xmath87 and happens in a thin tubular neighborhood of @xmath8 . 2 : : @xmath118}$ ] is ( up to the reparametrization ) the legendrian isotopy @xmath119 that changes @xmath87 to @xmath94 .
3 : : @xmath120}$ ] is a homotopy of loops that changes @xmath94 to @xmath28 and happens in a thin tubular neighborhood of @xmath28 . 4
: : @xmath121}$ ] is ( up to the reparametrization ) the isotopy @xmath122 that changes @xmath28 to @xmath8 .
consider the oriented @xmath17 bundle @xmath123 that is induced by @xmath116 from the contact bundle @xmath3 on @xmath4 .
since the isotopies @xmath124 and @xmath119 were chosen so that they are @xmath125 close , we get that the homology class realized by @xmath116 in @xmath126 is zero .
thus the euler class @xmath127 of @xmath128 is @xmath129 .
put @xmath138 to be the @xmath105-valued obstruction to extend the nonzero section of @xmath130 $ ] over @xmath134)=s^1\times \{1\}\cup s^1\times \{i\}$ ] given by the velocity vectors of @xmath8 and of @xmath87 to the nonzero section of @xmath130 $ ] over @xmath139 $ ] .
( it is the relative euler class that takes values in @xmath140 , \p ( s^1\times [ 1,i]))$ ] . )
similarly put @xmath141 to be the obstructions to extend the nonzero sections of @xmath142 over @xmath143 ) , \p ( s^1\times [ -1 , -i]),$ ] and @xmath144)$ ] that were described above to the nonzero sections respectively over @xmath145 , s^1\times [ -1 , -i],$ ] and @xmath146 $ ] . for a legendrian knot @xmath47 in the standard contact @xmath1 put @xmath148 to be the maslov class of @xmath47 . in
@xcite it is shown that @xmath149 for @xmath150 and for the legendrian knots @xmath47 and @xmath151 in the standard contact @xmath1 .
a straightforward verification based on this equality shows that @xmath152 the velocity vectors of the legendrian knots @xmath153 define the nonzero section of @xmath154 $ ] that extends the nonzero section over @xmath155)$ ] defined by the velocity vectors of @xmath87 and of @xmath94 .
thus @xmath156 since @xmath157 $ ] the knot @xmath158 is nowhere tangent to @xmath5 we get that for every point of @xmath103 the projection to @xmath3 along @xmath5 of the velocity vector of @xmath103 at this point is nonzero . thus the isotopy @xmath159 defines the nonzero section of @xmath160 over @xmath161 $ ] that is the extension of the nonzero section of @xmath160 over @xmath162)$ ] defined by the velocity vectors of @xmath8 and of @xmath28 .
hence we get that @xmath163 from the identities @xmath89 and @xmath90 one gets that @xmath165 and @xmath166 .
assume that @xmath167 .
( the case where @xmath168 is treated similarly . ) put @xmath169 .
it is easy to show that since @xmath87 and @xmath82 are legendrian isotopic , then @xmath170 and @xmath171 are also legendrian isotopic .
( basically one can keep the @xmath172 extra cusp pairs close together on a small piece of the projection of the part of the knot contained in a darboux chart during the whole isotopy process . ) but @xmath170 and @xmath171 are obtained from @xmath8 and @xmath28 by the modification corresponding to the addition of @xmath173 pairs of cusps of each of the two types , and we can take @xmath45 from the observation * 2 * to be @xmath173 .
there are four types of small kinks that can be added to a pseudo - legendrian knot .
they are kinks of types @xmath174 , shown in figure [ stabilization1.fig ] .
( here the first number in the pair corresponds to the increment to the rotation number of a knot and the second to the increment to the selflinking number of the knot that occurs if one adds a kink to a pseudo - legendrian knot in @xmath1 . )
let @xmath175 be a pseudo - legendrian knot .
for @xmath176 the @xmath177-_stabilization _
@xmath178 of @xmath175 is a pseudo - legendrian knot obtained from @xmath175 by the addition of @xmath179 and of @xmath180 it is easy to verify that the pseudo - legendrian isotopy type of the knot @xmath178 does not depend on the places on @xmath47 where the kinks are added .
( to get the same fact for singular pseudo - legendrian knots one observes that it is possible to pull a kink through a transverse double point of a singular pseudo - legendrian knot by a pseudo - legendrian isotopy of the singular knot . )
we say that a pseudo - legendrian knot @xmath175 _ admits finitely many symmetric stabilizations _
if there exists nonzero @xmath181 such that @xmath182 and @xmath175 are pseudo - legendrian isotopic .
similarly we introduce the notion of singular pseudo - legendrian knot @xmath183 that _ admits finitely many symmetric stabilizations . _
( proposition [ finitestab ] implies that , if @xmath175 admits finitely many symmetric stabilizations then for every @xmath176 the knot @xmath178 also admits finitely many symmetric stabilizations . )
as it is shown in figure [ kink2.fig ] the pair of kinks of types @xmath184 and @xmath185 can be cancelled by a pseudo - legendrian isotopy .
similar considerations show that a pair of kinks of types @xmath186 and @xmath187 also can be cancelled by a pseudo - legendrian isotopy .
[ stab ] let @xmath8 and @xmath28 be pseudo - legendrian ( with respect to @xmath5 ) knots that are @xmath125-close isotopic to each other as unframed knots .
then there exist @xmath176 such that @xmath8 is pseudo - legendrian isotopic to @xmath192 .
identify @xmath194 with the solid torus in @xmath1 such that both the axis of the torus and the vector field are parallel to the @xmath195-axis .
similar to [ description ] we can depict the pseudo - legendrian knots @xmath8 and @xmath28 by their knot diagrams obtained by projection to an annulus @xmath196 along the @xmath195-axis .
since @xmath8 and @xmath28 are isotopic inside @xmath193 as unframed knots , we get that the knot diagram of @xmath8 in @xmath196 can be changed to a knot diagram of @xmath28 in @xmath196 by a sequence of reidemeister moves .
the second and the third reidemeister moves can be done in the pseudo - legendrian category .
the first reidemeister move can not be done in the pseudo - legendrian category since under it the velocity vector of a knot at one of the points becomes parallel to the @xmath195-axis , and hence the knot becomes tangent to @xmath5 at one point .
as it is shown in figure [ kink2.fig ] one can create ( or annihilate ) a @xmath197-pair of kinks by a pseudo - legendrian isotopy .
similar considerations show that one can create ( or annihilate ) a @xmath198-pair of kinks by a pseudo - legendrian isotopy . using this observation
we can imitate the isotopy of unframed knots that changes @xmath8 to @xmath28 by a pseudo - legendrian isotopy .
namely , if for example the isotopy that changes @xmath8 to @xmath28 as an unframed knot involves the first reidemeister that creates the @xmath184-kink , instead of this move we perform the pseudo - legendrian isotopy that creates a @xmath197-pair of kinks , make the @xmath185 kink very small and keep it very small during the later isotopy process .
[ finitestab ] let @xmath8 and @xmath28 be pseudo - legendrian knots , and @xmath176 be such that @xmath8 and @xmath192 are pseudo - legendrian isotopic .
then for any @xmath203 the knots @xmath204 and @xmath205 are also pseudo - legendrian isotopic .
_ proof of proposition [ finitestab ] .
_ make the @xmath206 and @xmath207 extra kinks used to obtain @xmath204 very small and concentrate them on a small piece of @xmath8 .
keep them small and close together during the isotopy process that was connecting @xmath8 and @xmath192 . as a result
we get a pseudo - legendrian isotopy between @xmath204 and @xmath208 .
finally proposition [ composestabilization ] says that @xmath208 is pseudo - legendrian isotopic to @xmath205 . .
[ legendrianstabilization ] let @xmath47 be a pseudo - legendrian knot , that is pseudo - legendrian isotopic to a legendrian knot ( i.e. its pseudo - legendrian isotopy type is realizable by a legendrian knot ) .
then for any @xmath75 the pseudo - legendrian isotopy class of @xmath76 is also realizable by a legendrian knot
. _ proof of proposition [ legendrianstabilization ] .
_ one observes that if @xmath42 is a legendrian knot that is pseudo - legendrian isotopic to @xmath47 , then the legendrian knot obtained by the modification shown in figure [ twocusp.fig ] b is pseudo - legendrian isotopic to @xmath209 . similarly the legendrian knot obtained by the modification shown in figure [ twocusp.fig ] c is pseudo - legendrian isotopic to @xmath210 .
performing the two modifications @xmath77 and @xmath78 times respectively we get the legendrian knot that is pseudo - legendrian isotopic to @xmath76 .
_ in fact every component of the space of pseudo - legendrian curves contains exactly one component of the space of legendrian curves , but to prove this statement , see proposition [ components ] we have to use proposition [ decrease ] that is in turn based on proposition [ atmostonecomponent ] . _ _ proof of proposition [ atmostonecomponent ] .
_ the @xmath54-principle says that the connected components of the space of legendrian curves in @xmath9 are in one to one correspondence with the conjugacy classes of elements of @xmath58 . under this identification
the connected component of the space of legendrian curves that contains a legendrian curve @xmath42 corresponds to the conjugacy class of @xmath211 , where the loop @xmath212 is obtained by mapping @xmath213 to the point of @xmath55 that corresponds to the velocity vector of @xmath42 at @xmath214 .
let @xmath215 and @xmath216 be legendrian curves that are in @xmath27 . consider a pseudo - legendrian homotopy @xmath217\times s^1\rightarrow m$ ] that connects @xmath216 and @xmath218 .
observe that a pseudo - legendrian curve @xmath47 also defines a loop @xmath219 in @xmath55 by mapping @xmath46 in @xmath24 to the point of @xmath55 that corresponds to the projection of the velocity vector of @xmath47 at @xmath220 to a contact plane along @xmath5 .
( since @xmath47 is pseudo - legendrian the projection is nonzero . )
thus the family of pseudo - legendrian curves @xmath221 defines the free homotopy of loops @xmath222 and @xmath223 .
hence by the @xmath54-principle @xmath216 and @xmath218 belong to the same component of the space of legendrian curves .
[ everythingmadesingular ] the statements of propositions [ composestabilization ] , [ stab ] , [ finitestab ] , [ legendrianstabilization ] are true if one substitutes everywhere in the statements of these propositions pseudo - legendrian knots by singular pseudo - legendrian knots , and legendrian knots by singular legendrian knots .
the proof of this proposition is straightforward .
one has to observe that it is possible to pull a kink through a transverse double point of a singular pseudo - legendrian knot by a pseudo - legendrian isotopy of the singular knot .
[ decrease ] let @xmath9 be a contact manifold with a cooriented contact structure and let @xmath5 be a vector field that coorients the contact structure .
let @xmath26 be a connected component of the space of legendrian curves in @xmath9 and let @xmath27 be the corresponding component of the space of pseudo - legendrian ( with respect to @xmath5 ) curves .
a : : let @xmath224 be a pseudo - legendrian knot , then there exists @xmath225 such that the knot @xmath226 is pseudo - legendrian isotopic to a legendrian knot from @xmath26 .
b : : let @xmath227 be a singular pseudo - legendrian knot ( whose only singularities are @xmath77 transverse double points ) , then there exists @xmath225 such that the knot @xmath228 is pseudo - legendrian isotopic to a singular legendrian knot in @xmath26 .
_ proof of lemma [ decrease ] .
_ we prove statement * a * of lemma [ decrease ] .
( the proof of statement * b * of lemma [ decrease ] is obtained in the same way as the proof of statement * a * of lemma [ decrease ] , see proposition [ everythingmadesingular ] . ) the theorem of chow and rashevskii @xcite , @xcite says that that there exists a legendrian knot @xmath42 that is @xmath125-small isotopic to @xmath47 as an unframed knot . by proposition
[ stab ] we get that there exist @xmath176 such that @xmath229 is pseudo - legendrian isotopic to @xmath42 .
proposition [ legendrianstabilization ] says that the pseudo - legendrian isotopy classes of @xmath230 and of @xmath231 are also realizable by legendrian knots . by propositions [ composestabilization ] and [ finitestab
] @xmath232 is pseudo - legendrian isotopic to @xmath230 and @xmath233 is pseudo - legendrian isotopic to @xmath231 .
using this fact we get that there exist @xmath234 such that @xmath235 is pseudo - legendrian isotopic to a legendrian knot .
the pseudo - legendrian homotopies shown in figure [ kink2.fig ] and [ twokinksingular.fig ] imply that @xmath235 is in @xmath27 .
since by propositions [ atmostonecomponent ] there is only one component of the space of legendrian curves contained in @xmath27 , we get that @xmath235 is realizable by a legendrian knot from @xmath26 .
this finishes the proof of statement * a * of lemma [ decrease ] .
_ proof of proposition [ components ] .
_ let @xmath27 be a connected component of the space of pseudo - legendrian curves , and let @xmath224 be a pseudo - legendrian knot . in the proof of statement * a * of lemma [ decrease ] we showed that there exists @xmath83 such that @xmath226 is realizable by a legendrian knot . on the other hand the pseudo - legendrian isotopy shown in figure [ kink2.fig ] and the pseudo - legendrian homotopy shown in figure [ twokinksingular.fig ] imply that @xmath226 is in @xmath27 .
thus @xmath27 contains at least one component of the space of legendrian curves .
proposition [ atmostonecomponent ] says that @xmath27 contains at most one component of the space of legendrian curves .
thus it contains exactly one component of the space of legendrian curves . the fact that statement i of theorem [ equivalent ] implies statement ii is clear .
thus we have to show that statement ii implies statement .
this is done by showing that there exists a homomorphism @xmath236 such that @xmath237 and @xmath238 .
[ definitionofpsi]_definition of @xmath242 .
_ if the isotopy class of the knot @xmath243 is realizable by a legendrian knot @xmath244 , then put @xmath245 .
the value @xmath246 is well - defined because if @xmath247 is another knot realizing @xmath47 , then @xmath248 by statement i of theorem [ isomorphism ] .
if a pseudo - legendrian knot @xmath250 admits finitely many symmetric stabilizations , then by lemma [ decrease ] and proposition [ legendrianstabilization ] the knot @xmath47 and all the @xmath249 admit a legendrian realization by a knot from @xmath26 .
thus we have defined the value of @xmath242 on all the pseudo - legendrian knots from @xmath27 that admit finitely many symmetric stabilizations .
if @xmath250 admits infinitely many symmetric stabilizations , then either * 1 ) * all the isotopy classes of @xmath251 , @xmath252 , are realizable by legendrian knots from @xmath26 or * 2 ) * there exists @xmath252 such that @xmath251 is realizable by a legendrian knot from @xmath26 ( see [ decrease ] ) and @xmath253 is not realizable by a legendrian knot from @xmath26 .
( in this case @xmath254 etc . also are not realizable by legendrian knots from @xmath26 , see [ legendrianstabilization ] . ) in the case * 1 ) * the value of @xmath242 is already defined on all the pseudo - legendrian knots from @xmath27 that are pseudo - legendrian isotopic to @xmath249 , for some @xmath181 .
if @xmath253 is realizable by a legendrian knot @xmath42 , then consider a singular legendrian knot @xmath260 with @xmath48 double points that are vertices of @xmath48 small kinks such that we get @xmath42 if we resolve all the double points positively staying in the class of the legendrian knots .
( to create @xmath260 we perform the first half of the homotopy shown in figure [ kink.fig ] in @xmath261 places on @xmath42 . )
let @xmath262 be the set of the @xmath49 possible resolutions of the double points of @xmath260
. for @xmath263 put @xmath264 to be the sign of the resolution , and put @xmath265 to be the nonsingular legendrian knot obtained via the resolution @xmath266 .
since @xmath29 is an order @xmath53 invariant of legendrian knots we get that @xmath267 ( a straightforward verification shows that if we resolve @xmath77 double points of @xmath260 negatively , then we get the isotopy class of @xmath268 . )
this finishes the proof of the proposition .
let @xmath227 be a singular pseudo - legendrian knot with @xmath48 double points .
let @xmath262 be the set of the @xmath49 possible resolutions of the double points of @xmath52
. for @xmath269 put @xmath264 to be the sign of the resolution , and put @xmath270 to be the isotopy class of the knot obtained via the resolution @xmath266 .
if the isotopy class of @xmath52 is realizable by a singular legendrian knot from @xmath26 , then identity holds for @xmath52 , since @xmath29 is an order @xmath273 invariant of legendrian knots ( and the value of @xmath242 on a pseudo - legendrian knot @xmath224 realizable by a legendrian knot @xmath244 was put to be @xmath274 ) .
if @xmath275 admits finitely many symmetric stabilizations , then all the isotopy classes of singular pseudo - legendrian knots from @xmath27 that are isotopic to @xmath276 , for some @xmath181 are realizable by singular legendrian knots from @xmath26 , and we get that identity holds for @xmath52 .
if @xmath52 admits infinitely many symmetric stabilizations and all the isotopy classes of singular pseudo - legendrian knots @xmath277 , @xmath181 , are realizable by singular legendrian knots from @xmath26 , then automatically holds for @xmath52 .
if @xmath52 admits infinitely many symmetric stabilizations but not all the isotopy classes of @xmath276 @xmath181 , are realizable by singular legendrian knots from @xmath26 . then let @xmath252 be such that @xmath278 is realizable by a singular legendrian knot from @xmath26 and such that @xmath279 is not realizable by a singular legendrian knot from @xmath26 .
( the existence of such @xmath280 follows from lemma [ decrease ] .
note that as it follows from the version of proposition [ legendrianstabilization ] for singular knots , see [ everythingmadesingular ] , the pseudo - legendrian isotopy classes of @xmath281 etc .
are also not realizable by singular legendrian knot from @xmath26 . )
proposition [ legendrianstabilization ] says that @xmath282 , @xmath283 , are realizable by singular legendrian knots from @xmath26 and hence identity holds for @xmath282 , @xmath284 . using proposition [ mainidentity ] and the fact that identity holds for @xmath282 , @xmath284 , we show that holds for @xmath279 .
namely , @xmath285 considering the values of @xmath68 on the @xmath49 possible resolutions of a singular pseudo - legendrian knot with @xmath261 singular fragments shown in the middle part of figure [ twokinksingular.fig ] we get that @xmath62 should satisfy identity .
hence @xmath238 and this finishes the proof of theorem [ isomorphism ] .
* acknowledgments .
* i am very grateful to riccardo benedetti , yasha eliashberg , alexnadre kabanov , stefan nemirovskii , christian okonek , sergei tabachnikov , and carlo petronio for the their motivating questions and suggestions as to what information about the legendrian knots could be captured using vassiliev invariants .
and i am especially thankful to riccardo benedetti and carlo petronio who were , to my knowledge , the first to conjecture that a fact similar to the one we prove here is true .
these results were obtained during my stay at the max - planck - institute fr mathematik ( mpim ) , bonn , and at the institute for mathematics , zrich university , and i would like to thank the directors and the staff of the mpim , and the staff of the zrich university for hospitality and for providing the excellent working conditions .
99999 j. e. andersen , j. mattes , n. reshetikhin : _ quantization of the algebra of chord diagrams , _ math .
cambridge philos .
, vol . * 124 * no . 3 ( 1998 ) ,
451467 r. benedetti , c. petronio : _ combed @xmath2-manifolds with concave boundary , framed links , and pseudo - legendrian links , _
preprint math.gt/0001162 at the http://xxx.lanl.gov w. l. chow : _ ber systemevon linearen partiellen differentialgleichungen erster ordnung , _ math
* 117 * ( 1939 ) , pp .
98105 d. fuchs and s. tabachnikov : _ invariants of legendrian and transverse knots in the standard contact space , _ topology vol . * 36 * , no . 5 ( 1997 ) ,
10251053 d. fuchs and s. tabachnikov : joint results , private communication with s. tabachnikov ( 1999 ) v. goryunov : _ vassiliev type invariants in arnold s @xmath288-theory of plane curves without direct self - tangencies , _ topology vol .
* 37 * no . 3 ( 1998 ) , pp .
m. gromov : _ partial differential relations , _ springer - verlag , berlin heidelberg ( 1986 ) j. w. hill : _ vassiliev - type invariants in @xmath289-theory of planar fronts without dangerous self - tangencies , _ c. r. acad .
paris sr .
* 324 * no . 5 ( 1997 ) , pp .
537542 m. kontsevich : _
vassiliev s knot invariants , _ i.m .
gelfand seminar , pp .
137150 , adv .
soviet math .
, 16 , part 2 , amer .
soc . , providence , ri ( 1993 ) .
t. q. t. le and j. murakami : _ the universal vassiliev - kontsevich invariant for framed oriented links , _ compositio math .
, vol . * 102 * no . 1 ( 1996 ) pp .
k. rashevskii : _ about the possibility to connect any two points of a completely nonholonomic space by an admissible curve , _ uchen .
libknecht ped .
ser . mat .
* 2 * ( 1938 ) , pp .
8394 v. tchernov : _ arnold - type invariants of wave fronts on surfaces _ , to appear in topology ; preprint http://xxx.lanl.gov/ math.gt/9901133 ( 1999 ) v. tchernov : _ finite order invariants of legendrian , transverse and framed knots in contact @xmath2-manifolds ,
_ preprint http://xxx.lanl.gov math.sg/9907118 ( 1999 ) v. tchernov : _ vassiliev invariants of legendrian , of transverse and framed knots in contact 3-manifolds _ , submitted to topology ; preprint math.sg/0005002 at the http://xxx.lanl.gov
|
the study of the vassiliev invariants of legendrian knots was started by d. fuchs and s. tabachnikov who showed that the groups of @xmath0-valued vassiliev invariants of legendrian and of framed knots in the standard contact @xmath1 are canonically isomorphic .
recently we constructed the first examples where vassiliev invariants of legendrian and of framed knots are different , and vassiliev invariants of legendrian knots distinguish legendrian knots that are isotopic as framed knots and homotopic as legendrian immersions .
this raised the question what information about legendrian knots can be captured using vassiliev invariants . here
we answer this question by showing that for any contact @xmath2-manifold with a cooriented contact structure the groups of vassiliev invariants of legendrian knots and of knots that are nowhere tangent to a vector field that coorients the contact structure are canonically isomorphic .
|
the inflationary phase of the early universe @xcite is a very important ingredient in modern cosmology .
the reason for this is twofold .
first , inflation solves the longstanding problems of standard hot big - bang cosmology , namely the horizon , flatness and monopole problems .
furthermore , it produces the primordial perturbations for the structure that we observe today .
the recent spectacular cmb data from the wmap satellite @xcite have strengthen the inflationary idea , since the observations indicate an _ almost _ scale - free spectrum of gaussian adiabatic density fluctuations , just as predicted by simple models of inflation . in the prospect of future cmb experiments , like @xcite , constructing inflationary models which predict detectable primordial gravitational waves
( tensor - to - scalar ratio ) and non - gaussianities has attracted a lot of interest lately . as far as tensor - to - scalar ratio is concerned , at present
the limit is @xmath0 and it easy to build inflationary models that predict negligible tensor perturbations .
however , it would be interesting to have at our disposal models which predict a tensor - to - scalar ratio at the future detection limit , @xmath1 .
in fact , as it is known that current superstring inflationary models ( see e.g. @xcite ) predict negligible tensor perturbations @xcite ( see however @xcite ) , this would be challenging for testing string theory . on the other hand
, it is known that single field inflationary models with a minimally coupled scalar field predict primordial non - gaussianities at a non - detectable level @xcite , since the non - linear parameter @xmath2 .
however , the planck satellite is expected to detect the non - linear parameter if @xmath3 @xcite .
this is the reason why great effort is devoted to the construction of inflationary models which predict large non - gaussianities @xcite . in the present work
we show that a simple model with a single scalar field and with a non - trivial kinetic term can lead to detectable tensor - to - scalar ratio @xmath4 and non - linear parameter @xmath5 .
scalar field models with a non - canonical kinetic term have been discussed in k - inflation @xcite , in which inflation is not due to the potential but rather to the kinetic term in the lagrangian , and in k - essence models @xcite , which are designed to address the issue why the cosmic acceleration has recently begun .
a previous work @xcite is closely related to ours . in this
, the authors have shown that within the same class of models it is possible to have enhanced production of gravitational waves without conflict with observations .
however , non - gaussianities are not - discussed in this paper and in fact , as we will show below , the particular model @xcite predicts negligible @xmath6 .
the reason for that is that in the model of @xcite the sound speed @xmath7 . in the present work
however we study a different model in which the sound speed can be very low , @xmath8 , and therefore it is possible to have large @xmath4 and @xmath5 at the same time .
our work is organized as follows .
the article consists of four sections , of which this introduction is the first .
we present the basic formula in the second section and then we discuss slow - roll inflation in a simple model in section 3 . finally the fourth section is devoted to some conclusions .
our framework is four - dimensional general relativity coupled to a single scalar field with a general lagrangian @xmath9 , where @xmath10 is the scalar field and @xmath11 is the standard kinetic term .
therefore , our model is described by the action @xmath12 and we have set newton s constant @xmath13 .
the energy - momentum tensor for the scalar field is given by @xmath14 where @xmath15 denotes differentiation with respect to @xmath16 .
we can recast this energy - momentum tensor into the form of the energy - momentum tensor for a perfect fluid @xmath17 where @xmath18 are the energy density and the pressure of the fluid respectively .
the hydrodynamical quantities @xmath19 are given in terms of @xmath20 as follows @xmath21 in the case in which the lagrangian of the scalar field is of the form @xmath22 , the energy density is given by @xmath23 therefore , the equations of motion for our system are just the first friedmann equation and the equation for energy conservation @xmath24 where @xmath25 is the hubble parameter , @xmath26 is the scale factor and the overdot denotes differentiation with respect to cosmic time . defining the sound speed of the scalar field @xmath27 the equation for energy conservation takes the form @xmath28 which generalizes the usual klein - gordon equation of a canonical scalar field in a friedmann - robertson - walker background .
during the slow - roll phase the inflationary dynamics can be quantified in terms of the three parameters @xmath29 where @xmath30 , @xmath31 , @xmath32 is the reduced planck mass ( in the units we are using here @xmath33 ) and a prime denotes differentiation with respect to the scalar field .
we assume that @xmath34 during the slow - roll inflationary phase . furthermore , assuming that @xmath22 the equations of motion in the slow - roll approximation take the form @xmath35 since we ignore @xmath36 in the equation of motion for the scalar field , and @xmath37 is the dominant term in the expression for the energy density . upon considering a concrete model in which @xmath38
we find that the speed of sound is a constant and given by @xmath39 and it can be very low for large @xmath40 , @xmath41 . in addition , the equations of motion now take the form @xmath42 notice that in this model @xmath43 and the end of inflation is determined by the condition @xmath44 .
finally we define the number of e - folding @xmath45 as well as the quantities that connect the theoretical model to observations .
these are the spectral index and the amplitude of perturbations both for scalar and tensor perturbations .
the perturbations in k - inflation in linear order have been computed in @xcite . in linear perturbation theory
the scalar and tensor perturbations evolve independently from each other and therefore scalar perturbations can not produce gravitational waves
. however this is possible if second order effects are taken into account @xcite .
since k - inflation generalizes the usual inflation with a canonical scalar field , we expect that we could study the generation of gravitational waves by second order effects in k - inflation generalizing the results of @xcite .
this however lies beyond the scope of the present work .
the indices and amplitudes of scalar and tensor perturbations are given by @xmath46 the tensor - to - scalar ratio is defined by @xmath47 and is given by @xmath48 in standard inflation ( @xmath49 ) the tensor spectral index and the tensor - to - scalar ratio are not independent , and one has the so - called `` consistency relation '' @xcite , @xmath50 , which in principle can be tested against observations .
however , in k - inflation this relation does not hold any more . instead of that we have @xmath51 .
thus , at least in principle , kinetic inflation is phenomenologically distinguishable from standard inflation @xcite .
if in fact @xmath52 the production of gravitational waves is substantially larger than expected in standard inflationary models .
this was pointed out in @xcite . as in @xcite ,
the equation of motion for the scalar field in the slow - roll approximation can be easily solved and the solution corresponds to a constant kinetic term @xmath16 .
however , this is not important for what we would like to point out
. furthermore , we are not going to concern ourselves with the amplitude of scalar perturbations or the scalar spectral index , since in the present work our main interest is being on the tensor - to - scalar ratio and the non - gaussianity .
we just remark in passing that the observable quantities , @xmath53 etc , depend on the three model parameters @xmath54 through the slow - roll parameters . for a given value of @xmath40 or @xmath55 we can determine the values of @xmath56 by requiring that we must reproduce the observable values of @xmath57 and @xmath58
. then we can compute the rest of the observable quantities , namely the tensor - to - scalar ratio @xmath4 and the non - linear parameter @xmath6 , to which we now turn .
let us now focus on the primordial non - gaussianity . a rough measure of the non - gaussianities is provided by the non - linear parameter @xmath6 @xmath59 where @xmath60 is the gravitational potential which sources the temperature anisotropies in the cmb @xcite . in our model the parameter @xmath6 in leading order is given by @xcite @xmath61 according to @xcite
, @xmath6 is given by the expression @xmath62 where @xmath63 for usual inflation with a canonical scalar field , @xmath64 and therefore @xmath65 as we mentioned before .
in dbi inflation @xcite @xmath66 are different than zero and the sound speed @xmath67 . in a generic k - essence model @xmath66 are also different than zero , but the sound speed can be either larger than unity , @xmath7 or lower than unity , @xmath67 . in our model in the present work @xmath68 constant , @xmath69 , @xmath8 and @xmath70 different than zero .
it easy to see that for a large sound speed , @xmath71 , the non - linear parameter @xmath72 , which is just one order of magnitude compared to the canonical scalar field case , @xmath73 , while for a tiny sound speed , @xmath8 , the non - linear parameter @xmath74 . in this case @xmath6 can be much greater and therefore within the detectable range .
for example , let us consider a case in which @xmath75 , @xmath76 and @xmath77 .
then we obtain for @xmath4 and @xmath6 the values @xmath78 either of which is within the detectable range . in the figure below
we show @xmath6 as a function of @xmath4 and with the two horizontal lines we show the region that corresponds to a detectable range both for @xmath4 and @xmath6 . [ cols="^,^ " , ] in fact our discussion shows that even the form of the potential @xmath79 is not important .
what is important here is the form of the kinetic term , @xmath80 .
the key quantity in the present work is the sound speed and the two facts that it is constant and that is may become low enough so that we can have a large non - linear parameter @xmath6 , but not a negligible tensor - to - scalar ratio @xmath4 . therefore we expect that in our simple inflationary model it is possible to obtain both @xmath4 and @xmath6 in the detectable range independently of the form of the inflaton potential @xmath81 .
a final remark regarding the sound speed is in order here .
it is known that for a canonical scalar field the speed of sound equals unity .
however in models with a general scalar field the speed of sound can be different from unity and in fact it can even be @xmath82 .
for a recent discussion on superluminal fields and causality one can see @xcite . in our work
we have considered a case in which the speed of sound is very small , @xmath8 . in the past
k - essence models received some criticism because of possible superluminal propagation of the field fluctuations @xcite .
however it was shown recently that causal paradoxes do not arise in this kind of models @xcite .
in the present work we have studied slow - roll inflation with a single scalar field and non - trivial kinetic term for the inflaton . in particular , we have considered a simple concrete model in which the lagrangian for the generalized scalar field is of the form @xmath22 , where @xmath83 is the non - trivial kinetic term and @xmath79 is the inflaton potential . in this model
the speed of sound is constant and it can be very low for @xmath41 .
we have shown that in this simple model it is possible to have large tensor - to - scalar ratio and large primordial non - gaussianities .
this is a potentially interesting result in the prospect of new forthcoming observations .
in fact our main result is probably true for any form of the inflaton potential @xmath81 , as it turns out that the crucial point in our discussion is the form of the non - trivial kinetic term .
we would like to thank the anonymous reviewer for his / her valuable comments and suggestions .
this work was supported by project `` particle cosmology '' .
a. h. guth , phys .
d * 23 * ( 1981 ) 347 ; + d. h. lyth and a. riotto , phys .
* 314 * ( 1999 ) 1 [ arxiv : hep - ph/9807278 ] ; + a. r. liddle and d. h. lyth , `` cosmological inflation and large - scale structure , '' cambridge university press , cambridge , 2000 .
wmap , d. n. spergel _
_ , astrophys
. j. suppl . *
148 * , 175 ( 2003 ) , astro - ph/0302209 ; + d. n. spergel _ et al . _
[ wmap collaboration ] , arxiv : astro - ph/0603449 .
[ planck collaboration ] , arxiv : astro - ph/0604069 .
s. h. henry tye , arxiv : hep - th/0610221 ; + j. m. cline , arxiv : hep - th/0612129 ; + r. kallosh , arxiv : hep - th/0702059 .
r. kallosh and a. linde , jcap * 0704 * ( 2007 ) 017 [ arxiv:0704.0647 [ hep - th ] ] .
a. krause , arxiv:0708.4414 [ hep - th ] .
j. m. maldacena , jhep * 0305 * ( 2003 ) 013 [ arxiv : astro - ph/0210603 ] .
e. komatsu _ et al .
_ [ wmap collaboration ] , astrophys . j. suppl .
* 148 * ( 2003 ) 119 [ arxiv : astro - ph/0302223 ] .
d. h. lyth and d. wands , phys .
b * 524 * ( 2002 ) 5 [ arxiv : hep - ph/0110002 ] ; + f. bernardeau and j. p. uzan , phys .
d * 66 * ( 2002 ) 103506 [ arxiv : hep - ph/0207295 ] ; + f. bernardeau and j. p. uzan , phys . rev .
d * 67 * ( 2003 ) 121301 [ arxiv : astro - ph/0209330 ] ; + e. silverstein and d. tong , phys .
d * 70 * ( 2004 ) 103505 [ arxiv : hep - th/0310221 ] ; + j. e. lidsey and i. huston , arxiv:0705.0240 [ hep - th ] ; + f. bernardeau and t. brunier , arxiv:0705.2501 [ hep - ph ] .
m. alishahiha , e. silverstein and d. tong , phys .
d * 70 * ( 2004 ) 123505 [ arxiv : hep - th/0404084 ] . c. armendariz - picon , t. damour and v. f. mukhanov , phys .
b * 458 * ( 1999 ) 209 [ arxiv : hep - th/9904075 ] . c. armendariz - picon , v. f. mukhanov and p. j. steinhardt , phys .
* 85 * ( 2000 ) 4438 [ arxiv : astro - ph/0004134 ] ; + c. armendariz - picon , v. f. mukhanov and p. j. steinhardt , phys .
d * 63 * ( 2001 ) 103510 [ arxiv : astro - ph/0006373 ] .
v. f. mukhanov and a. vikman , jcap * 0602 * ( 2006 ) 004 [ arxiv : astro - ph/0512066 ] .
j. garriga and v. f. mukhanov , phys .
b * 458 * ( 1999 ) 219 [ arxiv : hep - th/9904176 ] .
b. osano , c. pitrou , p. dunsby , j. p. uzan and c. clarkson , jcap * 0704 * ( 2007 ) 003
[ arxiv : gr - qc/0612108 ] .
a. r. liddle and d. h. lyth , phys .
* 231 * ( 1993 ) 1 [ arxiv : astro - ph/9303019 ] . e. komatsu and d. n. spergel , phys .
d * 63 * ( 2001 ) 063002 [ arxiv : astro - ph/0005036 ] .
d. seery and j. e. lidsey , jcap * 0506 * ( 2005 ) 003 [ arxiv : astro - ph/0503692 ] .
j. p. bruneton , arxiv : hep - th/0612113 .
c. bonvin , c. caprini and r. durrer , phys .
lett . *
97 * ( 2006 ) 081303 [ arxiv : astro - ph/0606584 ] .
e. babichev , v. mukhanov and a. vikman , arxiv:0708.0561 [ hep - th ] .
|
an inflationary single field model with a non - trivial kinetic term for the inflaton is discussed .
it is shown that it is possible to have large primordial non - gaussianities and large tensor - to - scalar ratio in a simple concrete model with just a scalar field and a generalized kinetic term for the inflaton field .
this is potentially interesting in the prospect of new forthcoming observations .
|
anomalous x - ray pulsars ( axps ) are a sub - class of x - ray pulsars distinguished by pulse periods lying in a narrow range between @xmath11 and @xmath12 , low x - ray luminosities , soft x - ray spectra , no detected optical counterparts , no detected orbital doppler shifts of pulse arrival times , and associations with shell - type supernova remnants ( mereghetti & stella 1995 ; van paradijs , taam & van den heuvel 1995 ) .
the nature of axps is uncertain ; possibilities include an accreting neutron star in a binary , with a low - mass white dwarf or he - burning star as a companion ( mereghetti & stella 1995 ; baykal & swank 1996 ; mereghetti , israel & stella 1998 ) , an isolated neutron star accreting from a residual disk following a phase of common - envelope evolution ( van paradijs et al .
1995 ; ghosh , angelini & white 1997 ) , a non - accreting massive white dwarf ( morini et al . 1988 ; paczyski 1990 ; usov 1994 ) , and a non - accreting magnetar ( thompson & duncan 1996 ; heyl & hernquist 1998 ; kouveliotou et al .
1998 ) . the two axps with well - sampled timing histories , 1e 1048.1@xmath05937 and 1e 2259@xmath1586 ,
are observed to spin down irregularly : the rotation frequency @xmath13 of each object decreases linearly with time @xmath8 on average , but there are ` bumps ' superimposed on this average trend during which @xmath14 fluctuates by a factor of between two and five every five to ten years ( mereghetti 1995 ; baykal & swank 1996 ; baykal et al . 1998 ; oosterbroek et al . 1998 ; and references therein ) . in accreting - star models of axps
, the average spin - down trend is attributed to the accretion torque acting on a near - equilibrium rotator , and the bumps are the result of white torque noise ( e.g. due to disk inhomogeneities ) as in ordinary binary x - ray pulsars ( baykal & swank 1996 ) . in isolated - star models ,
the spin - down trend is attributed to magnetic - dipole braking , and the bumps are analogous to glitches observed in rotation - powered pulsars like vela ( usov 1994 ; heyl & hernquist 1998 ) . in this letter , we present unequivocal new evidence that axps are non - accreting magnetars . in [ sec : axp2 ] , it is shown that as a magnetar spins down it wobbles anharmonically , with a period of five to ten years , due to an effect called ` radiative precession ' .
the spin - down signature of the wobble , calculated theoretically in [ sec : axp3 ] , matches closely the bumpy timing histories of 1e 1048.1@xmath05937 and 1e 2259@xmath1586 .
the theory is used to predict @xmath7 over the next 20 years for both objects and yields a statistical relation between bump recurrence time and average spin - down rate for the axp population as a whole .
implications for the internal and magnetospheric structures of magnetars , including axps and soft gamma - ray repeaters ( sgrs ) , are explored in [ sec : axp4 ] .
a magnetar is a triaxial body in general .
hydromagnetic stresses arising from non - radial gradients of the superstrong internal magnetic field , e.g. between the magnetic poles and equator if the field is dipolar , deform the star , inducing matter - density perturbations @xmath15 and hence a fractional difference @xmath16 between the principal moments of inertia ( goldreich 1970 ; de campli 1980 ; melatos 1998 ) . here
, @xmath17 is the characteristic magnitude of the internal magnetic field , @xmath18 is the isothermal sound speed ( @xmath19 ) , and @xmath20 is the stellar radius ; one has @xmath21 if the internal field is confined to the stellar crust and @xmath22 if it is generated deep inside the star , e.g. in a convective dynamo ( thompson & duncan 1993 ) . in a rotation - powered pulsar with @xmath23 ,
the hydromagnetic deformation is comparable to the elastic deformation arising from shear stresses in the crystalline stellar crust , and the principal axes of inertia are oriented arbitrarily with respect to the axis @xmath24 of the external magnetic dipole ( goldreich 1970 ; de campli 1980 ; melatos 1998 ) . in a magnetar , however , the hydromagnetic deformation is much larger , and @xmath24 is approximately parallel to one of the principal axes ( @xmath25 , say ) ; the alignment is not exact due to the complicated structure of the internal field near its generation site , cf .
the non - axisymmetric , magnetically modified taylor columns seen in simulations of the geodynamo ( glatzmaier & roberts 1996 ) . provided that the rotation axis @xmath26 is not parallel to @xmath27 , as is usually the case for rotation - powered pulsars , the star precesses about @xmath25 with period @xmath28 .
the eulerian precession is not free .
it couples to a component of the vacuum magnetic - dipole torque associated with the asymmetric inertia of the near - zone radiation fields outside the magnetar ; the electromagnetic energy density ( and hence inertia ) of the near - zone radiation fields is greater at the magnetic poles than at the equator by an amount @xmath29 , resulting in an oscillatory , precession - inducing torque ( goldreich 1970 ; good & ng 1985 ; melatos 1998 ) .
the near - field torque exceeds the familiar braking torque ( @xmath30 ) by a factor @xmath31 and acts on a commensurately shorter time - scale , @xmath32 , where @xmath33 is the characteristic electromagnetic braking time . since the near - field torque is directed along @xmath34 , it does not change @xmath13 for a spherical star . nor does it change @xmath13 for an aspherical star , provided @xmath35 is large enough to give @xmath36 , so that the near - field torque averages to zero over one precession period . when the dominant deformation is hydromagnetic , however , one finds @xmath37 , close to unity ( i.e. strong coupling ) provided @xmath17 is moderately larger than @xmath38 as expected ( thompson & duncan 1993 ) . under these circumstances , the star executes an anharmonic wobble , called radiative precession , with period @xmath10 ( @xmath39 ) , and the near - field torque changes @xmath13 on the precession time - scale ( melatos 1998 ) .
we now show that the timing signature of radiative precession matches the observed bumpy spin - down of axps by solving numerically euler s equations of motion for a biaxial , dipole magnet rotating _ in vacuo _ , @xmath41 , \label{eq : axp1 } \\ \dot{\omega}_2 & = & \epsilon\omega_1\omega_3 + \omega_0^{-2}\tau_0^{-1 } [ -a\omega^2\omega_2 + b(-\omega_1\cos\chi+\omega_3\sin\chi ) ( \omega_1\sin\chi+\omega_3\cos\chi ) ] , \label{eq : axp2 } \\ \dot{\omega}_3 & = & - \omega_0^{-2}\tau_0^{-1}\sin\chi [ a\omega^2(-\omega_1\cos\chi+\omega_3\sin\chi ) + b\omega_2(\omega_1\sin\chi+\omega_3\cos\chi ) ] .
\label{eq : axp3}\end{aligned}\ ] ] in ( [ eq : axp1])([eq : axp3 ] ) , subscripts denote components along the principal axes of inertia , @xmath42 is the angle between @xmath24 and @xmath25 , and we have @xmath43 , @xmath44 , and @xmath45 , where @xmath46 is an arbitrary origin ; for derivations of the equations , see goldreich ( 1970 ) and melatos ( 1998 ) .
terms in ( [ eq : axp1])([eq : axp3 ] ) proportional to @xmath35 give rise to eulerian precession , terms proportional to @xmath47 are associated with the near - field torque , and terms proportional to @xmath48 cause secular braking .
a biaxial , dipole magnet rotating _ in vacuo _ is an idealized model of a hydromagnetically deformed magnetar . in reality , such a body is triaxial ( if it is indeed rigid ) , has high - order and/or off - centered multipoles contributing to the near - zone magnetic field , and is surrounded by a plasma magnetosphere .
the values of @xmath49 and @xmath50 reflect , in a coarse way , the magnetization state of the stellar interior and the distribution of magnetospheric currents ( melatos 1998 ) . in figures [ fig : axp1 ] and [ fig : axp2 ] , we plot the computed @xmath7 from ( [ eq : axp1])([eq : axp3 ] ) on top of the x - ray timing histories of 1e 1048.1@xmath05937 and 1e 2259@xmath1586 respectively , extending the theoretical curves 20 years beyond the present as a testable prediction .
the timing history of 1e 2259@xmath1586 is sufficiently well - sampled that only one good fit is possible . in the case of 1e 1048.1@xmath05937 , the solid curve is the favored fit , with @xmath51 , but an alternative fit , with similar average slope and @xmath52 , is also acceptable .
multiple good fits are hard to find .
euler s equations of motion contain just three unknown parameters : @xmath35 , @xmath53 , and @xmath42 .
it is significant that the theory agrees with the observations as well as it does despite its idealized nature particularly as only two of the three parameters are truly free , since one needs strong torque coupling ( @xmath54 , or equivalently @xmath55 ) in order to get any bumps at all .
moreover , the best agreement with observations is achieved for parameter values that are consistent with general physical considerations . if axps are hydromagnetically deformed magnetars with @xmath56 , one expects @xmath57 , @xmath58 , and @xmath42 relatively small ( cf .
@xmath59 for the earth ) , as discussed above .
the theoretical curves do not pass exactly through every available data point , and more departures are expected in the future , e.g. the slope between the last two data points in figure [ fig : axp1 ] is @xmath60 times the solid - curve theoretical slope at that epoch . indeed , a formal estimate of the chi - square for the solid curve in figure [ fig : axp1 ] , taking published timing uncertainties at face value , implies that the fit is poor : one finds a chi - square of @xmath61 with 10 degrees of freedom , notably inferior to multiple - glitch models , for example ( heyl & hernquist 1998 ) .
this is because a biaxial , dipole magnet is an over - idealized model of an axp , as noted above ; the chi - square likelihood improves dramatically for a more realistic model with just two extra parameters , e.g. a non - dipolar near - zone magnetic field and a triaxial ellipsoid of inertia .
however , our aim in this paper is not to model @xmath7 in detail , but to account for key gross features of the data the average spin - down rate , bump recurrence time , and bump amplitude with a simple physical theory . in this regard ,
the agreement with observations is good .
one data point in the timing history of 1e 2259@xmath1586 , at @xmath62 , represents a spin - up event lasting at most @xmath63 at the @xmath64 level of uncertainty ( baykal & swank 1996 ) .
spin - up can not be explained by radiative precession because ( i ) @xmath65 is always negative , even while @xmath66 decreases by up to a factor of five during a bump , and ( ii ) there is no natural way to explain a @xmath63 time - scale .
if taken at face value , the spin - up event must be a different phenomenon , e.g. a discontinuous change of internal structure analogous to the glitches of rotation - powered pulsars like vela ( usov 1994 ; heyl & hernquist 1998 ) .
numerical studies show that the bump recurrence time and average spin - down rate satisfy @xmath67 and @xmath68 , with @xmath69 .
in other words , the greater the magnetic field of an axp , the shorter is its precession period and the greater is its spin - down rate , with @xmath70 in addition , one finds that @xmath65 increases from @xmath71 to @xmath72 over a time @xmath73 during the course of a bump , yielding a bump amplitude @xmath74 this number is similar for all axps .
both the relations ( [ eq : axp4 ] ) and ( [ eq : axp5 ] ) will be subject to observational testing as the population of axps with measured timing histories swells over time .
note that they are statistical relations , with scatter expected about an overall trend .
the detailed structure of the magnetic field inside an axp which does not enter into the idealized theory presented here , except through @xmath42 is likely to differ from object to object , affecting @xmath10 , @xmath75 , and @xmath76 significantly , as the broad ranges of @xmath77 and @xmath78 attest .
why is radiative precession not observed in rotation - powered pulsars with @xmath23 ?
these objects are deformed hydromagnetically like magnetars , with an added elastic deformation , and they spin down electromagnetically .
young pulsars like the crab , with @xmath79 and @xmath80 , ought to precess with period @xmath81 , yet there is no clean evidence of bumpy spin - down in radio timing data , nor of concomitant changes in pulse profile ( e.g. relative height and separation of conal components ) and polarization characteristics ( e.g. position - angle swing ) .
lyne , pritchard & smith ( 1988 ) reported a quasi - sinusoidal variation in crab timing residuals with a period of @xmath82 months but judged it likely to be an artifact of an overlooked glitch .
( see also melatos 1998 . )
the only reliable detection of pulsar precession to date has been the general - relativistic geodetic precession of the binary pulsar psr b1913@xmath116 ( weisberg , romani & taylor 1989 ; kramer 1998 ) .
one possible explanation is that radiative precession is damped in pulsars with @xmath23 .
frictional dissipation inside the star , due to time - dependent elastic strains in the crust and/or imperfect crust - core coupling , is thought to occur on a time - scale @xmath83 ( link , epstein & baym 1993 ) , rapidly aligning @xmath26 with @xmath25 .
( in the case of the earth , dissipation restricts @xmath26 to fluctuating within just @xmath84 of @xmath25 under the action of solar and lunar tides ; see bura & p 1993 ) . in a magnetar , where the magnetic energy exceeds the mechanical energy of rotation
, the stiffening effect of the superstrong magnetic field may hinder the elastic strains and/or sheared fluid flows responsible for internal damping .
a second possible explanation is that conduction currents in the magnetosphere of a pulsar with @xmath23 , where electron - positron pair production is plentiful , modify the electromagnetic torque in such a way as to suppress the precession - inducing near - field component . in a magnetar , where it is thought that pair production is quenched , e.g. by positronium formation ( usov & melrose 1996 ; j. heyl 1999 , private communication ) or photon splitting ( baring & harding 1998 ) , the vacuum magnetic - dipole torque , with its unmodified near - field component , is a closer approximation to reality .
both explanations imply that radiative precession will not be suppressed in sgrs , where one has @xmath85 ( kouveliotou et al .
1998 ) .
what does radiative precession teach us about strongly magnetized neutron stars themselves ?
firstly , the close agreement between theory and observation implies that axps are indeed non - accreting magnetars , and that accretion is not needed to explain fluctuations in @xmath65 , contrary to claims in the literature ( e.g. mereghetti 1995 ) . secondly , if sgrs are magnetars too , as indicated by recent x - ray timing data ( kouveliotou et al .
1998 ) , they ought to exhibit bumpy spin - down just like axps , perhaps punctuated by brief , glitch - like spin - up events during the gamma - ray bursts themselves .
this constitutes a direct test of the magnetar model for sgrs .
thirdly , the fact that bumpy spin - down is seen in axps implies that @xmath17 is at most a few times @xmath38 to ensure strong coupling , i.e. @xmath86 , as discussed above .
this is indirect evidence that the magnetic field of these objects is generated relatively near the stellar surface .
fourthly , radiative precession places an upper limit on the fluid viscosity @xmath87 of a newly born magnetar ; if @xmath87 is too high , @xmath26 aligns with @xmath88 before the stellar crust crystallizes , and there is no precession ( melatos 1998 ) . the upper limit obtained in this way
that the viscous damping time is greater than the crust crystallization time of @xmath89 validates certain semi - quantitative calculations of @xmath87 for pulsars with @xmath23 by cutler & lindblom ( 1987 ) .
we remark in closing that the predicted hydromagnetic deformation ( @xmath57 ) constitutes a misaligned mass quadrupole which generates gravitational radiation
. however , the radiation is too weak to be detected by planned gravitational - wave interferometers like ligo and virgo , because axps are slow rotators and the gravitational - wave amplitude is a strongly rising function of @xmath13 .
i am indebted to lars bildsten for bringing the existence of spin - down irregularities in axps to my attention , and for subsequent discussions .
i also thank jon arons , peter goldreich , and the referee , jeremy heyl , for comments .
this work was supported by nasa grant nag53073 , nsf grant ast9528271 , and by the miller institute for basic research in science through a miller fellowship .
baring , m. g. , & harding , a. k. 1998 , , 507 , l55 baykal , a. , & swank , j. 1996 , , 460 , 470 baykal , a. , swank , j. h. , strohmayer , t. , & stark , m. j. 1998 , , 336 , 173 bura , m. , & p , k. 1993 , gravity field and dynamics of the earth ( berlin : springer - verlag ) cutler , c. , & lindblom , l. 1987 , , 314 , 234 de campli , w. m. 1980 , , 242 , 306 ghosh , p. , angelini , l. , & white , n. e. 1997 , , 478 , 713 glatzmaier , g. a. , & roberts , p. h. 1996 , science , 274 , 1887 goldreich , p. 1970
, , 160 , l11 good , m. l. , & ng , k. k. 1985 , , 299 , 706 heyl , j. s. , & hernquist , l. 1998 , , submitted ( preprint : astro - ph/9808125 ) kouveliotou , c. , et al .
1998 , , 393 , 235 kramer , m. 1998 , , 509 , 856 link , b. , epstein , r. i. , & baym , g. 1993 , , 403 , 285 lyne , a. g. , pritchard , r. s. , & smith , f. g. 1988 , , 233 , 667 melatos , a. 1998 , , submitted mereghetti , s. 1995 , , 455 , 598 mereghetti , s. , israel , g. l. , & stella , l. 1998 , , 296 , 689 mereghetti , s. , & stella , l. 1995 , , 442 , l17 morini , m. , robba , n. r. , smith , a. , & van der klis , m. 1988 , , 333 , 777 oosterbroek , t. , parmar , a. n. , mereghetti , s. , & israel , g. l. 1998 , , 334 , 925 paczyski , b. 1990 , , 365 , l9 thompson , c. , & duncan , r. c. 1993 , , 408 , 194 thompson , c. , & duncan , r. c. 1996 , , 473 , 322 usov , v. v. 1994 , , 427 , 984 usov , v. v. , & melrose , d. b. 1996 , , 464 , 306 van paradijs , j. , taam , r. e. , & van den heuvel , e. p. j. 1995 , , 299 , l41 weisberg , j. m. , romani , r. w. , & taylor , j. h. 1989 , , 347 , 1030
|
the two anomalous x - ray pulsars ( axps ) with well - sampled timing histories , 1e 1048.1@xmath05937 and 1e 2259@xmath1586 , are known to spin down irregularly , with ` bumps ' superimposed on an overall linear trend . here
we show that if axps are non - accreting magnetars , i.e. isolated neutron stars with surface magnetic fields @xmath2 , then they spin down electromagnetically in exactly the manner observed , due to an effect called ` radiative precession ' .
internal hydromagnetic stresses deform the star , creating a fractional difference @xmath3 between the principal moments of inertia @xmath4 and @xmath5 ; the resulting eulerian precession couples to an oscillating component of the electromagnetic torque associated with the near - zone radiation fields , and the star executes an anharmonic wobble with period @xmath6 , where @xmath7 is the rotation frequency as a function of time @xmath8 .
we solve euler s equations for a biaxial magnet rotating _ in vacuo _ ; show that the computed @xmath7 matches the measured timing histories of 1e 1048.1@xmath05937 and 1e 2259@xmath1586 ; predict @xmath7 for the next 20 years for both objects ; predict a statistical relation between @xmath9 and @xmath10 , to be tested as the population of known axps grows ; and hypothesize that radiative precession will be observed in future x - ray timing of soft gamma - ray repeaters ( sgrs ) .
|
fe - based superconductors ( fescs ) provide another avenue to understand unconventional superconductivity.@xcite due to its ease of synthesis , bafe@xmath0as@xmath0 is a prototype for these systems , where its low - temperature ( t @xmath6 k ) ground state is a striped , antiferromagnetic ( afm ) orthorhombic ( @xmath7 ) structure,@xcite often called a spin - density wave ( sdw ) , and which is reproduced in density functional theory ( dft ) calculations.@xcite at nel t@xmath8 ( @xmath9 k ) , both a magnetic and structural transition occurs to a tetragonal ( @xmath10 ) paramagnet.@xcite by doping with a transition - metal on the fe - site or others on ba- and as - sites , superconductivity ( sc ) can be achieved , and similarly with pressure.@xcite there are strong connections between the magnetism and sc . dopants weaken the magnetic state and cooper pairing is , perhaps , driven by increased magnetic fluctuations out of the ground state.@xcite dft has proven successful in modeling the geometry , magnetic ordering , and electronic structure of fescs .
the magnetic ground states of lafeaso , bafe@xmath0as@xmath0 , nafeas , and fete are all correctly predicted.@xcite fermi - surface ( fs ) nesting is apparent from dft calculations and agrees with angle - resolved photoemission ( arpes ) , suggesting an itinerant nature@xcite and which is supported from the spin - wave dispersion.@xcite furthermore , dft explains quantitatively effects of doping on fs nesting , and why cu doping behaves differently than co and ni.@xcite k@xmath11fe@xmath0se@xmath0 ( isostructural to bafe@xmath0as@xmath0 ) does not have the hole pockets needed for fs nesting,@xcite as dft finds.@xcite dft results for bafe@xmath0as@xmath0 show a strong coupling between the structure and magnetism.@xcite planar defects , thus , have been proposed to explain key features in magnetic and transport properties of fescs near / below the structural transition .
mazin and johannes@xcite suggested a model in which low - energy magnetic anti - phase ( apbs ) and 90@xmath12 domain ( dbs ) boundaries proliferate ( fig .
[ fig1 ] ) , which have yet to be tested .
so , are structural and magnetic planar defects energetically favorable and what are their properties ? to answer , we use dft to model potentially operative magnetic ( structural - induced ) defects , both isolated and extended , and explore their stability and properties by varying the structural parameters .
defects can be very important in realistic materials , like bafe@xmath0as@xmath0 . above t@xmath13 , the paramagnetic state
may be realized by mobile apbs and dbs ; below t@xmath13 , with interlayer coherence , apbs become pinned and dbs thermodynamically inaccessible , possibly explaining sensitivity to interlayer elements , large magneto - resistance , features in the differential resistivity ( @xmath14 ) , and invariance of resistivity anisotropy . with orthorhombic distortions ( @xmath15 ) , both structural and concomitant magnetic twins ( fig .
[ fig2 ] ) are observed in bafe@xmath0as@xmath0 along @xmath16 with @xmath17-@xmath18@xcite up to @xmath19-@xmath20@xcite between boundaries . with stress , samples detwin , but twins return upon its removal;@xcite as in yba@xmath0cu@xmath21o@xmath22,@xcite @xmath23 twins terminate on @xmath24 twins .
twins cause anisotropic scattering near afm wavevectors , giving 2-dimensional spin fluctuations .
twins also create stripes of increased diamagnetic response,@xcite and nucleate sc at their boundaries.@xcite recently , niedziela et al.@xcite found by rietveld analysis a bigger orthorhombic ratio ( o = @xmath25 ) for local structural fits ( o = 1.38% ) than global fits ( o = 0.78% ) ; they proposed a high density of _ nano_-twins ( fig .
[ fig2 ] ) account for this discrepancy by its better match to measured pair distribution functions ( pdf ) .
we show that displacements at the _ nano_-twin boundary affect spin alignment , reducing the average `` ordered '' moment . for completeness , we note that , while dft supports the observed sdw for the parent compound , the fe moment ( @xmath26)@xcite is twice that assessed for the average ordered moment from neutron diffraction ( @xmath27).@xcite in fact , various experiments assess very different fe moments .
core - electron spectroscopy@xcite finds @xmath28 , like dft , while @xmath29fe mssbauer@xcite and nuclear magnetic resonance@xcite find @xmath30 , as in diffraction assessments .
for fe - based magnets such a large discrepancy between ordered moments from theory and experiment is unusual .
spin - orbit and hybridization ( controlled by fe / as planar spacing ) in a @xmath31 model explained the small in - plane moments in fe - pnictides.@xcite yet , our dft moments are reduced @xmath32@xmath33 from spin - orbit , but 50 - 100% by slightly reduced fe - as spacing .
dft predicts correct moments at short times ( @xmath3210@xmath34 s ) necessary to yield lattice constants that agree with experiment.@xcite dynamical mean field theory ( dmft ) explains the discrepancy from dft as a result of dynamical fluctuations at the fe sites that reduce the observed moment over longer time scales ( @xmath3210@xmath35 s),@xcite and reproduces the trends in reduced fe moments and renormalized mass across various fescs.@xcite dmft finds fescs are correlated due to intra - atomic exchange from hund s coupling @xmath36 ( 0.3 - 0.6 ev)@xcite ( which reduces the coherence temperature for fermi liquid behavior@xcite ) , not from especially large @xmath37 ( 2.8 - 5.2 ev , as derived from a five band constrained random phase approximation)@xcite or proximity to a mott insulating state . below the coherence temperature , high electron mobility results in moment screening ( over 10@xmath38 .
notably , this scenario does not consider spatial fluctuations , defects , nor their effect on magnetism near / below the phase transition , as explored in the present work .
- , ( b ) @xmath39- planes , and ( c ) @xmath40 db with no strain ( @xmath41=@xmath42 ) .
red ( blue ) circles are `` up '' ( `` down '' ) in - plane moments , as indicated in ( a ) .
hs , ms and ls indicate fe - sites with high- , medium- and low - spin states.,width=264 ] we use dft to simulate various magnetic planar defects , @xmath43 , two types of apbs , a @xmath40 db , twin boundaries , and our modified _
nano_-twin , which are all excitations of the sdw .
figure [ fig1]a and [ fig1]b shows two apb boundaries in the fe - plane parallel to the @xmath44- or @xmath39-planes and fig . [ fig1]c shows a locally unstrained @xmath40 db . figure [ fig2]a shows a typical example of an ideal twin . a modified @xmath1twin with @xmath45-dimensional structural distortion ( consistent with that suggested by niedziela _ _ et al.__@xcite ) is shown in fig .
[ fig2]b with a series of static displacements along @xmath41- and @xmath42-axis in the supercell .
the undisplaced nanotwin with @xmath46-layer of fe separating defect planes is really a magnetic stacking fault ( sf ) ; a nanotwin supercell has very different boundary conditions than a twin , with different far neighbors and distances between defect pairs ; indeed , `` ideal twin '' supercells formed with 1-layer separation between defect planes ( a high density of sfs ) has local environments like the nanotwin , except that twin has symmetric relaxations governed by the supercell periodicity , whereas the nanotwin has asymmetric , localized distortions to match the pdf .
while we show the defect energies are similar , a nanotwin , due to its boundary condition and supercell , may be considered a fluctuating twin nuclei , which can have low - spin fe - sites unavailable in the ideal twin supercell . for nondefected ( parent ) and defected cells we calculate energy per atom and the associated magnetic moments ( bulk is @xmath47 ) .
from this we derive the planar defect energy , @xmath48 , defined as @xmath49 , where @xmath50 and @xmath51 are the total _ energy per atom _ of the defected and nondefected cell , respectively .
@xmath52 is the distance between defect planes and @xmath53 is the volume per atom . while the energy per atom is helpful
, @xmath48 is the appropriate comparison for cost of creating the defect interface and its dependence on defect density and defect volume .
note that 2 defect boundaries are created for twins , hence , @xmath54 is appropriate defect energy . to ( 110 ) , where atom positions in a distorted cell are barycentric weights of the cell corners .
hs , ms and ls are indicated .
, width=328 ] to do this , we use vasp@xcite with plane - wave pseudopotential projected augmented wave ( paw ) basis,@xcite with an energy cut - off of @xmath55@xmath56v .
a monkhorst - pack brillouin zone integration with a @xmath57 * k*-mesh is used for the sdw ( @xmath7 ) structure .
smaller * k*-meshes are used for supercells depending on the length coverage along each axis . for apbs , we constructed doubled ( @xmath45@xmath58@xmath58 ) , quadrupled , and octupled supercells to examine excitations , denoted by 2-apb , 4-apb and 8-apb , respectively ( ) . for apbs ( figs .
[ fig1]a , b ) , we use measured lattice parameters@xcite ( @xmath41=@xmath59 , @xmath42=@xmath60 and @xmath61=@xmath62 ) . for a @xmath40 db , we set @xmath63 = @xmath64 = @xmath65 = @xmath66 to reduce local strain effects , and construct supercells similar to the apbs , denoted as 2-db , 4-db , and 8-db .
twin ( @xmath67$]@xmath68@xmath58 ) supercells ( @xmath69=@xmath70 ) are denoted by ( @xmath71)-n fe - layers between defect planes , and have @xmath72 unit cells along @xmath41 and @xmath73@xmath74 atoms / cell .
nanotwin supercells are denoted 3-n , 5-n , 9-n , and 13-n for fe - layers between isolated nanotwin pairs ; the supercells with the static displacements suggested by niedziela _
et al . _ are more complex because the local distortions must be compensated within the cell ( fig .
[ fig2]b ) .
-plane ( b ) @xmath39-plane , and ( c ) @xmath40 db .
sdw indicates the afm ground state .
insets depict the local fe environment .
, width=321 ]
the energies and moments for apb and db defects relative to the non - magnetic ( nm ) state are shown in fig .
[ fig3 ] ( top ) , and compared to the afm ground state ( sdw ) . in all cases , fe moments have two behaviors : a high - spin state ( hs in fig .
[ fig1 ] ) at sites away from boundaries and a low - spin state ( ls in fig .
[ fig1 ] ) at / near boundaries . for apb(@xmath44 ) , the ls moment falls substantially to @xmath75 from @xmath76 , similar to that found by @xcite while for apb(@xmath39 ) , the ls moment decreases only to @xmath77 .
the two spin states depend on local magnetic environments ( inset fig .
[ fig3 ] ) .
moments do not vary much with the size of the supercells , but these two structures energetically compete with the ground state sdw ( @xmath78v / atom ) . for @xmath40 db ( fig .
[ fig1]c ) , the hs state has a higher moment of @xmath79 due to global strain from changed lattice parameters .
the ls moment decreases slightly to @xmath80 near the boundary .
this defect requires within @xmath81v / atom excess energy to form compared to the sdw .
it is energetically competing with the apb(@xmath39 ) .
both defects are then expected to be present at the same temperature .
the local environment does not play a significant role , suggesting simple models such as counting the number of aligned neighbors is not sufficient to characterize the moments .
but for ( a ) twins ( labels defined in text ) , and nanotwins ( @xmath82@xmath83@xmath68 supercell ) with displacements along ( b ) d@xmath41 and ( c ) d@xmath42 axes .
fe has hs , ms , and ls states due to local environment ( insets).,width=321 ] the energy and moments for twins are shown in interestingly , an fe - atom in a twin has three spin states depending on the local environment .
fe - atoms at the boundary remain in a medium - spin state ( ms in fig . [
a ls state occurs on fe - sites adjacent to the boundary ( fig . [
these fe - sites have the same nearest - neighbor environment as the bulk hs states but differ in the farther neighbors .
these defects can form at a few @xmath84v / atom , albeit @xmath48 is more critical , see below .
the nanotwin energies and moments versus distortion along @xmath85 and @xmath86axis ( in ) are shown in fig .
[ fig4](b , c ) .
similar to twins , there are three fe spin states : a hs bulk ( @xmath47 ) , a ms ( @xmath87 ) at the boundary , and a ls ( @xmath88 ) in the vicinity of the distorted side of the boundary .
the structural perturbations show a stronger effect on the ls moments near defect boundaries , decreasing to as low as @xmath89 .
isolated ( fluctuating ) nanotwins are equally competitive to form as dense twins but with much reduced moments .
energies are affected mostly by the changed magnetic configurations and very little by spatial distortions .
so , magnetic defects drive the short - range structural distortion ( not the other way around ) and can help quench magnetization . .@xmath48 ( @xmath54 for twins ) for various planar defects ( in @xmath4j / m@xmath90 ) .
energies ( @xmath84v / atom ) are relative to sdw .
@xmath91 is dominated by @xmath52 increasing faster than the decrease in ( @xmath92 ) , unlike for apbs or dbs . [
cols="<,^,^,>",options="header " , ] ( top ) and @xmath93e = ( e@xmath94-e@xmath95 ) and `` d '' ( bottom ) versus number of fe - layers.,width=321 ] planar defect energies ( @xmath48 or @xmath96 ) are compared in table [ table1 ] ; they give the relative _ order in which magnetic defects can form and remain after processing_. structural defects can act as pinning sites for magnetic domain walls .
energetically , apb(@xmath39 ) , @xmath40 db and nanotwins ( low - energy spin excitations ) are the most favorable and most likely to persist after annealing .
interestingly , densely - pack twins of a single tetragonal variant are also remarkable very low energy .
these nanoscale defects compete with widely separated twins ( spin kinks ) , which are observed .
such small fluctuating defects will affect the observed average moments , whereas separated twins will affect the magnetic correlation length , see below .
separated twins do form and are stabilized by lattice strain arising from disclinations formed when twins oriented @xmath40 apart ( from the two tetragonal variants ) intersect.@xcite it is the twin - twin interactions that stabilize the mesoscale twins . typically in metals ,
the calculated @xmath97 is monotonically decreasing versus @xmath52 ( the separation of the twin boundaries ) until it plateaus at the measured twin boundary energy ; essentially , the defects interact ( costing energy ) until separated enough that they are screened from one another .
strikingly in bafe@xmath0as@xmath0 , separated twins are higher in energy than dense twins , until a @xmath52 of @xmath98 unit cells ( @xmath99 fe - layers ) , where @xmath97 reaches a maximum ( table [ table1 ] and fig .
[ fig5 ] ) , after which there is a slow convergence of @xmath97 versus @xmath52 ( fig . [ fig5 ] ) . at @xmath100 unit cells ( @xmath32@xmath101 ) ,
@xmath97 has not yet converged , emphasizing the long - range interactions among twins . observed structural twins@xcite are extended well beyond the ones computationally feasible .
thus , higher - density twins should become prominent near the phase transition , where they compete with the ground state .
twin separation @xmath52 is also affected by stress .
equilibrium is typically reached when the applied stress is @xmath102 , which is , however , exceedingly small for isolated twins in bafe@xmath0as@xmath0 . in real samples
twins appear in @xmath40 oriented pairs , where @xmath23 twins terminate on @xmath24 twins ; this configuration is stabilized by lattice strain arising from disclinations,@xcite where the strain is reduced at the cost of increased @xmath52 . with stress ( estimated roughly from a set of disclinations,@xcite and orders of magnitude larger than @xmath103 ) , samples detwin , but twins would ( and do ) return upon its removal.@xcite twins cause anisotropic scattering near afm wavevectors , giving 2-dimensional spin fluctuations , and create stripes of increased diamagnetic response.@xcite while twin separation depends on local defects and stress , it is expected to get a peak in the magnetic susceptibility @xmath104 at @xmath105 , where @xmath106 is the average twin - twin separation where @xmath54 saturates .
the direction of @xmath107 is perpendicular to twin boundaries ( i.e. , @xmath108 to reciprocal - space @xmath109- and @xmath110-axes , where @xmath111 ( @xmath112 ) is along @xmath41- ( @xmath42- ) axis ) .
while the twins dictate the magnetic correlation length , we suggest that small , low - energy excitation can further depress average moments by spatial and temporal averaging , beyond those due to dynamic fluctuations .
nanotwins ( fig .
[ fig2]b ) with no local distortion are like an isolated , ideal defect pair , not a dense set of twins . to understand the effect of short - ranged structural distortion , we have studied a @xmath46-@xmath113 ideal twin with(out ) relaxation in @xmath114-plane for only those atoms near the boundary , more localized than in the nanotwin supercell .
the planar defect energy with(out ) relaxation is @xmath115 ( @xmath116 ) @xmath4j / m@xmath90 .
relaxations along @xmath41- and @xmath42-axis lie within @xmath117% of ideal , close to the best fit to measured pdf,@xcite so the twin and nanotwin are very similar in energy and local structure . unlike for ideal twins , the nanotwin surface energy decreases to its limiting value as the nanotwin - nanotwin distance grows ( table [ table1 ] ) , and it is much lower in energy than extended twins .
thus , a nanotwin may be considered a fluctuating twin nuclei , which has many more ls sites ( fig .
[ fig4 ] ) not available in a twin supercell , with moments as low as @xmath89 near the defect , similarly to the assessed values in bafe@xmath0as@xmath0 .
our calculations support niedziela et al.s suggestion@xcite that nanotwins constitute an important fluctuating excitation in bafe@xmath0as@xmath0 .
-orbital and spin - projected dos of an undistorted nanotwin on hs ( black , heavy line ) and ls ( red , light line ) fe - sites , with majority ( minority ) dos plotted on positive ( negative ) vertical axis .
fermi energy is indicated by vertical dashed line .
@xmath111 , @xmath112 , and @xmath118 directions correspond to @xmath41 , @xmath42 , and @xmath61 respectively.,width=328 ] because the local magnetic configurations play the key role in determining the spin states of fe , we calculated the site- and density of states ( dos ) to understand the electronic - structure origin . figure [ fig6 ] shows the fe @xmath52-projected dos for hs and ls states .
for the bulk ( hs ) states , the major contribution at fermi energy e@xmath119 arises from fe @xmath120 and @xmath121 , also evidenced from arpes.@xcite all the other orbital components exhibit a pseudo - gap near e@xmath122 . for ls - fe compared to hs - fe , all the projected dos are shifted towards e@xmath122 .
the most pronounced effect occurs for @xmath120 and @xmath121 character , where majority states for ls fall into a pseudogap for @xmath120 but are peaked for @xmath121 .
although the change of these orbital states is dominated by in - plane fe - spin configurations , small contributions also arise from the hybridization with as @xmath123 and @xmath124 orbitals ( out of the fe - plane ) , eventually altering the fs .
the large difference in the near e@xmath122 ( majority ) dos between the hs and ls state points to the orbital dependent electronic origin for quenched moments .
in summary , we studied competing low - energy , magnetic planar defects in bafe@xmath0as@xmath0 .
the favorable defects are apb(@xmath39 ) , @xmath40db , and nanotwins , but twins ( which are observed ) are favorable through the mesoscale .
the most pronounced reductions in fe - moment are near the boundaries of apbs(@xmath44 ) and nanotwins .
we find that isolated closely - spaced twins ( twin nuclei ) are energetically favorable and correspond to a recently proposed nanotwin suggested to match the pair distribution function from scattering experiment.@xcite nanotwins are energetically insensitive to microscopic displacements near the boundary , in contrast to sensitivity to the as @xmath118 coordinate .
apbs along and are not equally favorable , an anisotropy not anticipated in the mazin and johannes model.@xcite these defects can reduce the fe moment from spatial averaging , an environmental dependence which is not included in dmft.@xcite assessing these defects and their dynamics can affect magnetism , which can be evaluated via monte carlo simulations , and which are planned .
work was supported by the u.s .
department of energy , office of basic energy science , division of materials science and engineering ( seed funding ) , and , for s.n.k . by the center for defect physics , an energy frontier research center at ornl .
ames laboratory is operated for the u.s .
doe by iowa state university under contract de - ac02 - 07ch11358 .
52ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty link:\doibase 10.1103/revmodphys.83.1589 [ * * , ( ) ] link:\doibase 10.1146/annurev - conmatphys-062910 - 140518 [ * * , ( ) ] link:\doibase 10.1146/annurev - conmatphys-070909 - 104041 [ * * , ( ) ] link:\doibase 10.1038/nphys1759 [ * * , ( ) ] link:\doibase 10.1063/1.4752092 [ * * , ( ) ] link:\doibase 10.1103/physrevb.78.020503 [ * * , ( ) ] link:\doibase 10.1103/physrevb.78.094511 [ * * , ( ) ] link:\doibase 10.1103/physrevb.78.214515 [ * * , ( ) ] link:\doibase 10.1103/physrevb.82.024519 [ * * , ( ) ] http://stacks.iop.org/0953-8984/21/i=1/a=012208 [ * * , ( ) ] link:\doibase 10.1103/physrevb.78.104527 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.101.087004 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.102.047005 [ * * , ( ) ] link:\doibase 10.1103/physrevb.78.085104 [ * * , ( ) ] link:\doibase 10.1103/physrevb.80.020504 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.102.177003 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.100.237003 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.110.107007 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.104.137001 [ * * , ( ) ] link:\doibase 10.1103/physrevb.84.054544 [ * * , ( ) ] link:\doibase 10.1103/physrevb.78.220501 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.109.167003 [ * * , ( ) ] link:\doibase 10.1103/physrevb.82.180520 [ * * , ( ) ] http://stacks.iop.org/0256-307x/28/i=5/a=057402 [ * * , ( ) ] link:\doibase 10.1103/physrevb.79.184523 [ * * , ( ) ] link:\doibase 10.1038/nphys1160 [ * * , ( ) ] link:\doibase 10.1103/physrevb.79.060506 [ * * , ( ) ] link:\doibase 10.1103/physrevb.79.180508 [ * * , ( ) ] link:\doibase 10.1103/physrevb.81.184508 [ * * , ( ) ] link:\doibase 10.1080/01418619308213974 [ * * , ( ) ] link:\doibase 10.1103/physrevb.81.184513 [ * * , ( ) ] link:\doibase 10.1103/physrevb.85.024530 [ * * , ( ) ] link:\doibase 10.1103/physrevb.86.174113 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.102.107003 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.101.257003 [ * * , ( ) ] link:\doibase 10.1103/physrevb.79.184519 [ * * , ( ) ] link:\doibase 10.1103/physrevb.84.100509 [ * * , ( ) ] link:\doibase 10.1103/physrevb.85.220503 [ * * , ( ) ] link:\doibase 10.1103/physrevb.79.052504 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.101.126401 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.104.197002 [ * * , ( ) ] link:\doibase 10.1038/nmat3120 [ * * , ( ) ] link:\doibase 10.1146/annurev - conmatphys-020911 - 125045 [ * * , ( ) ] link:\doibase 10.1143/jpsj.79.044705 [ * * , ( ) ] http://stacks.iop.org/1367-2630/11/i=2/a=025021 [ * * , ( ) ] link:\doibase 10.1143/jpsj.77.093711 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/j.physc.2009.03.040 [ * * , ( ) ] link:\doibase 10.1038/nphys2250 [ * * , ( ) ] link:\doibase 10.1103/physrevb.54.11169 [ * * , ( ) ] link:\doibase 10.1103/physrevb.59.1758 [ * * , ( ) ] link:\doibase 10.1103/physrevb.80.144522 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.104.057002 [ * * , ( ) ]
|
in bafe@xmath0as@xmath0 , structural and magnetic planar defects begin to proliferate below the structural phase transition , affecting descriptions of magnetism and superconductivity .
we study using density - functional theory the stability and magnetic properties of competing antiphase and domain boundaries , twins and isolated @xmath1twins ( twin nuclei ) - spin excitations proposed and/or observed .
these nanoscale defects have very low surface energy ( @xmath2-@xmath3 @xmath4jm@xmath5 ) , with twins favorable to the mesoscale .
defects exhibit smaller moments confined near their boundaries making a uniform - moment picture inappropriate for long - range magnetic order in real samples .
_ nano_twins explain features in measured pair distribution functions , so should be considered when analyzing scattering data .
all these defects can be weakly mobile and/or have fluctuations that lower assessed `` ordered '' moments from longer spatial and/or time averaging , and should be considered directly .
|
broad absorption line ( bal ) quasars show absorption from gas with blueshifted outflow velocities of typically 0.1@xmath3 . about 10% of quasars exhibit bal troughs ,
usually attributed either to an orientation effect such that all quasars have bal outflows covering @xmath410% of the sky , or a phase of @xmath4100% covering lasting @xmath410% of the typical quasar lifetime .
understanding bal outflows is necessary since the bal outflow mass loss rates seem comparable to the accretion rates required to power quasars .
unusual bal quasars delineate the full range of physical conditions and parameter space spanned by bal outflows .
the sloan digital sky survey ( york et al . 2000 ; www.sdss.org ) has discovered populations of unusual low - ionization broad absorption line quasars ( ` lobals ' ) , many of them ` felobals ' with absorption from excited feii or feiii ( hall et al . 2002 ) . one population of unusual bal quasars has unprecedented line ratios which are the reverse of those seen in previously known objects ; namely , lobals with stronger absorption in aliii than mgii and felobals with weak or no feii absorption but with feiii even stronger than mgii ( fig .
. the cloudy modeling of de kool et al .
2002 ( his figs . 6 - 7 ) shows that these ratios can be produced in a high density outflow ( log @[email protected] @xmath1 for ionization parameter @xmath7 ) with column density @xmath8 in a narrow range ( @xmath40.2 dex ) such that a partially ionized zone of feii and mgii typically present in agn broad emission line regions does not exist in the bal outflow along our line of sight , but a feiii+aliii zone does .
the high densities and narrow range in @xmath8 required to reproduce the observed line ratios explains why these objects are so rare .
the sdss has recently found a quasar with a narrow feiii - dominant absorption system for which high - resolution spectra could test this explanation .
unfortunately , the most luminous feiii - dominant bal quasar known , sdss 2215 - 0045 , can not test this model since a vlt+uves spectrum shows no narrow features .
however , the uves spectrum does reveal that civ and siiv emission are as weak as aliii .
this might be due in part to a weak absorption trough at @xmath44000kms@xmath2 in addition to the main trough at @xmath415000kms@xmath2 ( fig .
1b , blue arrows ) .
the feiii uv48 @xmath92080 absorption in sdss 2215 - 0045 appears stronger than the uv34 @xmath91910 absorption , which is unphysical . in hall
2002 this was attributed to a previously unknown resonance populating the lower term of uv48 , but we now believe it is due to a complicated underlying continuum . as seen in fig . 1a ,
the unabsorbed continuum of 2215 - 0045 ( thin dotted red line ) can be estimated as a power law plus feii+feiii emission ( using the extreme fe emitter sdss 0923 + 5745 as a template ) , all reddened by @xmath10@xmath11@xmath12-@xmath13@xmath14=0.06 .
due to the strong emission at 1800 and the lack of emission at 2000 , the observed spectrum ( heavy line ) appears consistent with saturated feiii uv48 and uv34 absorption with 35% partial covering .
the lower - ionization mgii absorption has smaller partial covering , which is typical .
the fe - emission template is not a perfect fit ( e.g. , 2215 - 0045 has weaker mgii and ciii ] emission than the template ) , but it shows the spectrum can be fit without invoking a new feiii resonance .
traditionally , objects with troughs @xmath152000 km / s wide , or with absorption solely within 3000 km / s of the quasar redshift , have not been classified as bal quasars .
however , numerous studies have shown that both ` mini - bals ' and many narrow associated systems also arise in the central regions of agn . thus in hall
( 2002 ) we suggested an alternative to the traditional balnicity index which does not discriminate against narrow intrinsic absorbers .
the physics of agn outflows are more easily studied with such objects , since in bals most doublet transitions are blended and can not be used to infer the physical conditions or other parameters of the outflow . for example , studies of numerous intrinsic agn outflows have shown that the troughs are usually saturated . in such cases , the trough depth in high - resolution spectra yields that ion s covering factor as a function of velocity . then , knowing the relative strengths of the various transitions observed , interesting conclusions can be drawn by examining the spectra .
an excellent example is the reddened , extreme - feii - emitting , mini - lobal quasar sdss 1453 + 0029 , which has @[email protected] from an isaac near - ir spectrum ( fig .
2 ) . broadly speaking
, this object has two hei absorption systems within a more extensive ( in velocity space ) set of mgii absorption systems .
this hei absorption arises in an excited state populated by recombination ( heii + e@xmath18 @xmath19 hei + @xmath20 ) .
thus this hei absorption traces the column density of heii , which yields a lower limit on @xmath8 since hydrogen is completely ionized in heii regions .
component a in sdss 1453 + 0029 appears to have equally strong hei@xmath93188 and hei@xmath93889 absorption ( fig .
since the latter transition is intrinsically much stronger , this means that hei@xmath93889 is saturated .
the covering factor of the hei ( @xmath450% ) is less than that of mgii ( @xmath485% ) by inspection .
component b has hei line ratios consistent with hei@xmath93889 being at most only slightly saturated , but its velocity - dependent covering factor reaches a maximum @xmath490% while that of mgii is only @xmath460% .
thus , hei and mgii absorption can behave very differently , even over small velocity ranges , and even when the two components are very likely in physical proximity as indicated by their nearly identical velocities . based on data from eso programs 267.a-5698 and 67.a-580 . funding for the sdss archive has been provided by the alfred p. sloan foundation , the participating institutions , nasa , the nsf , the u.s .
department of energy , the japanese monbukagakusho , and the max planck society .
the sdss is managed by the astrophysical research consortium , for the participating institutions : the university of chicago , fermilab , the institute for advanced study , the japan participation group , the johns hopkins university , los alamos national laboratory , the max - planck - institute for astronomy , the max - planck - institute for astrophysics , new mexico state university , university of pittsburg , princeton university , the usno and the university of washington .
|
among the unusual broad absorption line quasars being found by the sloan digital sky survey ( sdss ) are objects with much stronger absorption in feiii than feii .
these unusual line ratios require a high density in the outflow ( @xmath0 @xmath1 ) .
they should also appear for only a limited range of outflow column densities , which explains their rarity .
previously we suggested that the feiii line ratios were also affected by a resonance ; we now believe this is an artifact of structure in the underlying feii+feiii pseudocontinuum . the sdss is also discovering objects with absorption in rarely seen transitions such as hei .
vlt+uves high - resolution spectra of one such object , the mini - bal quasar sdss 1453 + 0029 , show that it has two hei absorption systems with considerably different properties separated by only 350 km s@xmath2 .
|
high - dimensional data have generated a great challenge in different fields of statistics , computer science , and machine learning . in order to consider cases where the number of covariates is larger than the sample size ,
new methodologies , applicable for the model under some structural constraints , have been developed .
for instance , there have been substantial works under the sparsity assumption including sparse linear regression , sparse covariance matrices estimation or sparse inverse covariance matrices estimation ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in this paper , we focus on the problem of _ low rank matrix recovery and uncertainty quantification_. there have been quite a few work on estimating low rank matrices in the matrix regression setting ( also named the trace regression setting , the matrix compressed sensing setting , or the quantum tomography setting when the parameter is a density matrix ) . many authors ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) considered the exact recovery of a low - rank matrix based on a subset of uniformly sampled entries .
also @xcite considered matrix recovery based on a small number of noisy linear measurements in the framework of restricted isometry property ( rip ) .
@xcite proved non - asymptotic bounds on the frobenius risk , and investigated matrix completion under a row / column weighted random sampling .
@xcite proposed a nuclear norm minimisation method and derived a general sharp oracle inequality under the condition of restricted isometry property .
very recently , @xcite considered a rank - one projection model and used constrained nuclear norm minimization method to estimate the matrix .
@xcite considered a specific quantum tomography problem where the parameter is a density matrix ( for more details , plasase see subsection [ ss : q ] ) , and @xcite proved that the quantum tomography design setting satisfies the rip .
in addition , @xcite proposed an estimator based on an entropy minimisation for solving a quantum tomography problem .
@xcite adapt the iterative hard thresholding method ( first introduced in the sparse linear regression setting , see e.g. * ? ? ?
* ; * ? ? ?
* ) to the problem of low rank matrix recovery in the case where the noise is non - stochastic and of small @xmath0 norm .
this procedure has the advantage of being very computationally efficient . in the same vein but applied to the more challenging stochastically noisy setting , @xcite introduced a soft thresholding technique that provides efficient result in this setting in frobenius norm , see also @xcite for other thresholding methods in related settings that provide results also in frobenius norm .
another important problem is on understanding the uncertainty associated to these statistical methodologies , by e.g. characterizing the limiting distribution of the efficient estimators . yet
results in this area for high dimensional models are still scarce , available mainly for the sparse ( generalised ) linear regression models @xcite . in the papers
@xcite , the authors focus first on constructing an estimator for the sparse parameter that has good properties in @xmath1 risk , and they use then this result to exhibit the limiting distribution of their estimator . knowing this limiting distribution immediately enables the construction of tests and confidence sets for low dimensional subsets of parameters . a similar achievement ,
i.e. the construction of an estimator that has an explicit limiting distribution , does not exist in the low rank matrix recovery setting . to the best of our knowledge ,
moreover , all the theoretical results from the above papers on the estimation of the parameter in the noisy setting are derived in frobenius risk neither in the entrywise matrix @xmath2 risk , nor in the operator norm ( i.e. the largest singular value ) . in our paper
, we consider the problem of constructing an estimator for low - rank matrix in a stochastically noisy high - dimensional setting , under the assumption that a rip - type isometry condition is satisfied ( see assumption [ ass : designbis ] ) .
we provide ( in theorem [ th : mainthm2 ] ) error bounds for our estimator in all @xmath3 schatten norms for @xmath4 .
we prove in particular that this estimator has optimal frobenius and operator norm risk by proving that this estimator has optimal @xmath2 risk performance uniformly over any change of orthonormal basis .
in addition , a slight modification of our estimator has an explicit gaussian limiting distribution with bounded bias in operator norm ( see theorem [ thm : asymnorm2 ] ) ; and in the particular case when the design consists in uncorrelated gaussian entries , we prove that the bias in @xmath1 entry wise norm is bounded as well , which is immediately useful for testing hypotheses and constructing confidence intervals for each parameter of interest , similar to the ideas in @xcite . moreover
our estimator is computationally efficient with an explicit algorithm .
the proposed algorithm is inspired by the iterative hard thresholding , that refines its estimation of the matrix by iteratively estimating the low rank sub - space where the matrix s image is defined .
it requires only @xmath5 iteration steps to converge approximately , and the computational complexity of the method is of order @xmath6 where @xmath7 is the dimension of the matrix , and @xmath8 is the sample size . in the experiment section
we first provide some simulations where we illustrate the efficiency of our method and explain how it can be used to create a confidence interval for the entries of the low rank matrix .
we then apply our method to a specific _ quantum tomography application _ , namely multiple ion tomography ( see , e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , where the assumptions required by our method are naturally satisfied ( see e.g. * ? ? ?
* ; * ? ? ?
finally we compare our method with other existing estimation methods for the trace regression setting @xcite using the gradient descent implementation of @xcite and also regularized maximum likelihood based procedures @xcite . as a complement in the supplementary material ,
we adapt our method to the setting of sparse linear regressionm and provide an estimator that has an explicit limiting distribution ( recovering the results of @xcite ) .
for @xmath9 , @xmath10 and @xmath11 , we write @xmath12 for the hard thresholded version of @xmath13 at level @xmath14 , i.e. for the vector @xmath15 such that @xmath16 for @xmath17 . for @xmath18 and @xmath19 , we write @xmath20 for the standard @xmath0 norm of @xmath13 , and @xmath21 for the standard @xmath1 norm of @xmath13 . for a @xmath22 complex matrix @xmath23 , we write @xmath24 as the conjugate transpose of @xmath23 .
we write @xmath25 for the trace of @xmath23 , and @xmath26 for the matrix whose diagonal entries are the same as @xmath23 while its non - diagonal entries are all zeros .
we write the entry - wise matrix norm of @xmath23 as @xmath27 , and its squared frobenius norm as @xmath28 .
we write also the operator norm of @xmath23 as @xmath29 , where the @xmath30 are the singular values of @xmath23 , and the schatten @xmath3 norm of @xmath23 for @xmath31 as @xmath32 - and note that @xmath33 . for @xmath9 , we write @xmath34 for the hard thresholded version of @xmath23 at level @xmath14 for each entry , i.e. for the matrix @xmath35 such that @xmath36 for @xmath37 .
let @xmath38 .
let @xmath39 be the set of @xmath40 matrices , and @xmath41 be the set of @xmath40 complex matrices of rank less than or equal to @xmath42 .
let us also write @xmath43 for the set of orthonormal matrices in @xmath44 . for @xmath45
, we consider the matrix regression problem where for any @xmath46 , @xmath47 where @xmath48 is an i.i.d .
vector of gaussian white noise , i.e. @xmath49 ( but our results hold in the same way for any sub - gaussian independent noise @xmath50 : see remark [ rem : noise ] ) , and @xmath51 but @xmath52 .
let us write @xmath53 for the linear operator going from @xmath44 to @xmath54 , and such that for any @xmath55 , @xmath56 the model can be rewritten as @xmath57 where @xmath58 .
this matrix regression model is directly related to the quantum tomography model ( in which case the design @xmath53 is often chosen to be the random pauli design @xcite , but it is also related to e.g. matrix completion @xcite .
we state the following assumption on the design operator @xmath53 . [
ass : designbis ] let @xmath59 . for any @xmath60
, it holds that @xmath61 where @xmath62 .
[ rem : gaussian ] the above assumption is very related to the restricted isometry property . typically ,
for uncorrelated gaussian design with mean @xmath63 and variance @xmath64 entries , it will hold with probability larger than @xmath65 for @xmath66 where @xmath67 is a universal constant . for the pauli design used in quantum tomography , it will hold with probability larger than @xmath65 for @xmath68 where @xmath67 is a universal constant @xcite - see subsection [ ss : q ] for a description of description of a quantum tomography setting in which the pauli matrices represent measurements .
as a generalization of sparsity constraints in linear regression models , we impose a rank @xmath69 constraint on a matrix @xmath70 .
that is , we require the rows ( or columns ) of @xmath71 lie in some @xmath42-dimensional subspace of @xmath72 .
this type of rank constraint arises in numerous applications such as quantum tomography , matrix completion , and matrix compressed sensing ( see e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
our method considers the parameters @xmath73 .
the parameter @xmath74 is a small probability that will calibrate the precision of the estimate : the theoretical results that we will prove later for this estimate will hold with probability @xmath65 , and the smaller @xmath74 , the larger the constant in the bound ( see theorem [ th : mainthm2 ] ) .
the parameter @xmath75 is an upper bound on two times the actual low rank of the parameter @xmath71 .
it does not need to be tight , and the final results will not depend on it as long @xmath76 ( see assumption [ ass : designbis ] and theorem [ th : mainthm2 ] ) .
the parameter @xmath77 is an upper bound on the frobenius norm of the parameter @xmath71 .
it again does not need to be tight , but constants in the proof will scale with it .
we set the initial values for the estimator @xmath78 and the threshold @xmath79 such that@xmath80 we update the thresholds @xmath81 where @xmath82 , @xmath83 is an universal constant ( see lemma [ helo ] ) and @xmath84 . set now recursively , for @xmath85 , @xmath86 , @xmath87 and let @xmath88 be two orthonormal matrices that diagonalise @xmath89 .
then we set @xmath90 this procedure provides a sequence of estimates , and as we will prove in the next subsection , this sequence is with high probability close to the true @xmath71 as soon as @xmath91 is of order @xmath92 ( see theorems [ th : mainthm2 ] and [ thm : asymnorm2 ] ) .
[ rem : para ] note that although we describe this method using many quantities , in fact while implementing our method we only need to set up four quantities : @xmath93 and the stopping time @xmath91 .
we describe in equation how to implement a good stopping rule , and in subsection [ ss : disc ] how to choose the three first parameters .
in particular , @xmath79 can be chosen in a data driven way .
this method is related to iterative hard thresholding ( iht ) , a method that has been developed for the sparse regression setting ( see e.g. * ? ? ?
* ; * ? ? ?
it is less straightforward to see this in this setting , as in the sparse regression setting where we adapt also our method in subsection [ ss : sr ] , and for a more comprehensive discussion of the relation between our method and iht , see the remark [ rem : iht ] .
note that iht algorithms have been proved to work in settings where the noise is small and non - stochastic ( see e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , but to the best of our knowledge , there are no results on iht in a stochastically noisy setting .
[ [ main - result - for - our - thresholded - estimator ] ] main result for our thresholded estimator + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we now provide a theorem that guarantees that the estimate @xmath94 after @xmath95 iterations has at most rank @xmath42 , and its entry - wise @xmath2 risk and frobenius risk are bounded with the optimal rates ( see the first point in subsection 3.3 ) .
[ th : mainthm2 ] assume that assumption [ ass : designbis ] is satisfied and that @xmath96 . let @xmath97 .
we have that for a constant @xmath98 it holds that with probability larger than @xmath65 and for any @xmath99 @xmath100 and also that @xmath101 and also that for any @xmath102 @xmath103 the above theorem proves among other things that our estimate attains the minimax optimal schatten @xmath3 risk , which other estimates in the literature also attain for e.g. @xmath104 .
the first interesting property is that our proposed estimator has an explicit algorithmic form and is very computationally efficient .
another interesting additional property is that it is also minimax - optimal in operator norm ( or entry - wise matrix @xmath1 risk ) , and that the entry - wise error is not more than @xmath105 with high probability _ for any orthonormal change of basis of the matrix @xmath71_. this is a strong result since the entry - wise norm is not invariant by orthonormal change of basis while the frobenius norm is .
this result is already useful for measuring the uncertainty of an estimate ( in particular since it does not require the a priori knowledge of the rank of the matrix @xmath71 ) .
[ [ asymptotic - normality - results ] ] asymptotic normality results + + + + + + + + + + + + + + + + + + + + + + + + + + + + to prove asymptotic normality , we slightly modify the estimator defined in theorem [ th : mainthm2 ] . consider the estimator @xmath94 of theorem [ th : mainthm2 ] ( with @xmath97 ) and define @xmath106.\ ] ] [ thm : asymnorm2 ] set @xmath107 then we have @xmath108 where @xmath109 .
let @xmath97 .
the two following bounds hold for the bias term @xmath110 under two different assumptions . *
assume that assumption [ ass : designbis ] is satisfied for some @xmath111 and that @xmath112 .
if the rank of @xmath71 is smaller than @xmath113 and if its frobenius norm is bounded by @xmath77 , there is a constant @xmath98 such that with probability larger than @xmath65 @xmath114 * assume that the elements in the design matrices @xmath115 are i.i.d .
gaussian with mean @xmath63 and variance @xmath64 , and that @xmath116 , we have that @xmath117 note that this implies the previous result .
this theorem , which is in the spirit of the works in the context of sparse linear regression of @xcite , implies that there exists an estimator of @xmath71 that has a gaussian limiting distribution , and whose rescaled bias @xmath110 with respect to @xmath71 can be bounded @xmath118 in operator norm under our assumption [ ass : designbis ] and @xmath119 in @xmath1 norm as well in the specific case where the design is gaussian .
[ rem : noise ] theorems [ th : mainthm2 ] and [ thm : asymnorm2 ] are proved for a gaussian noise @xmath50 , but these results are easily generalisable to any independent , sub - gaussian noise , with a similar but more technical proof ( based on talagrand s inequality ) .
the results of theorem [ thm : asymnorm2 ] would however be modified in that the random variable @xmath120 , conditioned on the design @xmath53 , would then not be exactly gaussian , but have a limiting gaussian distribution using the central limit theorem . [ [ stopping - rule - r ] ] stopping rule @xmath91 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + theorem [ th : mainthm2 ] is satisfied after @xmath121 iterations of our thresholding strategy , and so we know what is a theoretical value for @xmath91 so that our strategy works .
however , it is possible to propose a data driven stopping rule that will perform well . for a desired precision @xmath122
, we propose to stop the algorithm as soon ( after having thresholded a last time ) as @xmath123 let us write @xmath124 for the time where the stopping rule stops .
the following result holds for the estimator stopped at this stopping rule .
[ prop : srule ] assume that assumption [ ass : designbis ] is satisfied and that @xmath125 , i.e. @xmath126 . let @xmath127 in ( [ eq : sr ] .
the estimator @xmath128 satisfies with probability larger than @xmath65 and for any @xmath99 @xmath129 and also that @xmath130 which implies for any @xmath4 @xmath131 moreover @xmath124 is such that @xmath132 this empirical stopping rule , that does not require the tuning of any additional parameters we introduced the desired precision @xmath133 .
the precision @xmath133 is a very natural quantity to choose for the experimenter
one can set e.g. @xmath134 for an @xmath135 optimal solution with respect to the solution outputted by an algorithm that runs for an infinitely long time ] , is guaranteeing minimax optimal results in less than @xmath92 iterations .
note that theorem [ thm : asymnorm2 ] would also hold using this stopping rule - this can be proved in the same way as theorem [ prop : srule ] is proved .
[ [ comparison - of - our - results - with - the - literature ] ] comparison of our results with the literature + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + our theorem [ th : mainthm2 ] gives bounds for our estimators in all schatten @xmath4 norms ( including the operator norm , and therefore uniform entry wise bounds in all rotation basis ) .
a first point is that our results are minimax optimal in both frobenius and operator norm .
the corresponding lower bound in frobenius norm can be found in e.g. theorem 5 of @xcite ( under an assumption related to our assumption [ ass : designbis ] ) or @xcite under an assumption that is the same as ours .
the corresponding lower bound in operator norm can be found in e.g. @xcite .
in addition to this , the paper @xcite even contains further lower bounds results proving that the operator norm rate @xmath105 ( and associated schatten @xmath136 norm @xmath137 ) is optimal also in the case of quantum tomography that we use in our experiments later on , i.e. under the additional assumptions that the parameter is a density matrix and that the design is random pauli . to the best of our knowledge ,
our method is the first iterative method that has such an optimality property in operator norm - for instance , the paper @xcite provides results for schatten norms with @xmath138 $ ] , but not for other schatten norms .
besides , we proved in theorem [ thm : asymnorm2 ] a slight modification of our estimator has an explicit gaussian limiting distribution , and this is , again to the best of our knowledge , the first iterative method for low rank matrix recovery that has such a property . on top of that ,
the computational complexity of our algorithm is low as for any procedure based on iterative hard thresholding : see the papers @xcite .
our assumption [ ass : designbis ] is a strong rip condition .
but it is in particular satisfied in the interesting application of multiple ion tomography for the natural pauli design as soon as the number of settings is large enough , see subsection [ ss : q ] .
operator norm bounds are particularly interesting since they provide an entrywise bound up to any change of orthonormal basis . in particular
, they provide a bound on the eigen values - and since these bounds do not depend on the true rank @xmath42 , they can be used to implement conservative confidence sets .
moreover as highlighted in the papers @xcite , having a bound on the entrywise risk , and then an estimator with explicit limiting distribution , is interesting in that it can be used to construct tests and confidence intervals for subsets of coordinates of the parameter @xmath71 .
we illustrate this point in the simulation section ( see section [ s : sim ] ) , where a confidence set is constructed using the limiting distribution .
note however that the bound on the bias term @xmath110 in @xmath1 norm in theorem [ thm : asymnorm2 ] requires the fact that the design is gaussian . on the other hand ,
the bound on the bias term @xmath110 in operator norm in theorem [ thm : asymnorm2 ] requires only the fact that our assumption [ ass : designbis ] is satisfied . [
[ stopping - rule - r-1 ] ] stopping rule @xmath91 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + our theorems are satisfied after @xmath121 iterations of our thresholding strategy , and so we know what is a theoretical value for @xmath91 so that our strategy works .
we also defined an empirical stopping rule , see and theorem [ prop : srule ] .
we use this stopping rule in practice for all our experiments in section . [
[ calibration - of - the - parameters - of - the - proposed - method ] ] calibration of the parameters of the proposed method + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + our method is not parameter free - there are three quantities that need to be calibrated .
the two first ones , that we write @xmath139 and @xmath140 , enter in the definition of the thresholds sequence @xmath141 . @xmath139 controls the rate at which we make our threshold decay , and @xmath142 is the quantity toward which it converges when @xmath91 goes to infinity . the last quantity , namely @xmath79 is the initialisation of the threshold sequences .
here are some comments on how to choose these quantities : * * rate of decay @xmath139 :* in theory the parameter @xmath139 can be taken between @xmath64 and @xmath143 where @xmath75 is an upper bound on the rank of the parameter and @xmath144 is the constant associated to the design such that assumption [ ass : design ] is satisfied
. it might not be really possible to compute exactly @xmath75 or @xmath144 without more assumptions on the design .
but there is at least one design that is interesting in practice , namely the random pauli design for quantum tomography , that is such that we have an upper bound on @xmath144 for all @xmath75 that is of order @xmath145 with high probability , i.e. it decays with @xmath8 ( the same holds in gaussian design up to the @xmath146 term ) . in this design
if @xmath8 is large enough , we know that taking @xmath147 will work - we do not want to take @xmath139 too close to @xmath64 since the bound on the performance of the estimator scales with @xmath148 .
* * smallest threshold calibration @xmath140 * the interpretation and theoretical value of @xmath140 is clear : it should be taken to be larger than the @xmath74 quantile of the lhs quantity defined in equation divided by @xmath149 .
now since we do not have access to this quantile , we calibrate it in the experimental section of this paper as an empirical estimator of the asymptotic quantile described above ( using theorem [ thm : asymnorm2 ] ) . *
* initialisation threshold @xmath79 :* the constant @xmath79 needs to be taken as an upper bound on the frobenius norm of @xmath71 .
note first that estimating from the data an upper bound on @xmath150 is easy under assumption [ ass : design ] , i.e. the quantity for @xmath151 @xmath152 overestimates @xmath150 . in our simulations
, we propose a slightly more refined heuristic upper bound and use the same @xmath79 and @xmath153 in subsection [ s : sim ] and [ ss : q ] . to conclude on the practical tuning of the constants
, we would like to emphasize that at least in practical situation , namely quantum tomography , we have enough information about both the design and the noise level to know that the above calibration will be working , provided that the target matrix is indeed low rank .
so although the tuning of parameters is always a tricky issue for any algorithm , in at least this specific application , our algorithm can be used as it is .
in this section we present some experiments , first some simulation for the construction of confidence intervals , and then a formal comparison of our thresholded estimator with other methods on specific quantum tomography problem , namely multiple ion tomography .
we performed experiments for low - rank matrix recovery , with matrix dimension @xmath7 .
we consider a gaussian design where each @xmath154 and are independent .
we also consider a gaussian uncorrelated noise @xmath49 .
we consider a parameter @xmath71 of rank @xmath42 that is stochastically generated in an isotropic way as @xmath155 we implemented our method choosing a data - driven heuristic for the choice of our parameters .
we first set @xmath156 we set for any @xmath86 @xmath157 i.e. the empirical risk , and @xmath158 where @xmath159 is the @xmath160 quantile of a @xmath161 random variable .
@xmath162 replaces here @xmath140 , and is a heuristic high probability bound on the error for each coordinate .
we set @xmath163 which is by construction higher than the frobenius norm of @xmath71 with high probability , and @xmath164 where we select @xmath147 ( we take @xmath165 so that the decay is not too fast , but also so that @xmath148 is not too large ) .
we also use the heuristic stopping rule described in equation , i.e. we iterate until @xmath166 for @xmath167 .
we also construct , using the limiting distribution results provided in theorem [ thm : asymnorm2 ] , a confidence set for the all the entries of @xmath71 that is such that for any entry @xmath168 , we set the confidence interval @xmath169,\ ] ] where @xmath170 where @xmath171 .
we provide several results , depending on the values of @xmath172 , averaged over @xmath173 iterations of simulations .
for these simulations , we present three kinds of results : * a first set of graphs ( figure [ fig:1 ] ) presents , for different values of @xmath174 , and in function of the sample size @xmath8 , the logarithm of the rescaled frobenius risk of the estimate @xmath175 , i.e. @xmath176 * a second set of graphs ( figure [ fig:2 ] ) presents , for different values of @xmath174 , and in function of the sample size @xmath8 , the logarithm of the averaged diameter of the confidence intervals @xmath177 , i.e. @xmath178 * a last set of graphs ( figure [ fig:3 ] ) presents , for different values of @xmath174 , and in function of the sample size @xmath8 , the averaged coverage probability of the confidence intervals @xmath177 , i.e. @xmath179 all these graphs also exhibit @xmath180 confidence intervals ( upper and lower @xmath181 quantile values from 100 iterations ) around their means ( dotted lines in the graphs , the solid line being the mean ) . , for different values of @xmath174 .
the solid line is the average over @xmath173 iterations , the dotted lines form @xmath180 confidence intervals.,scaledwidth=100.0% ] , for different values of @xmath174 .
the solid line is the average over @xmath173 iterations , the dotted lines form @xmath180 confidence intervals.,scaledwidth=100.0% ] , for different values of @xmath174 .
the solid line is the average over @xmath173 iterations , the dotted lines form @xmath180 confidence intervals.,scaledwidth=100.0% ] , for different values of @xmath174 .
the solid line is the average over @xmath173 iterations , the dotted lines form @xmath180 confidence intervals.,scaledwidth=100.0% ] , for different values of @xmath174 .
the solid line is the average over @xmath173 iterations , the dotted lines form @xmath180 confidence intervals.,scaledwidth=100.0% ] , for different values of @xmath174 .
the solid line is the average over @xmath173 iterations , the dotted lines form @xmath180 confidence intervals.,scaledwidth=100.0% ] , for different values of @xmath174 .
the solid line is the average over @xmath173 iterations , the dotted lines form @xmath180 confidence intervals.,scaledwidth=100.0% ] , for different values of @xmath174 .
the solid line is the average over @xmath173 iterations , the dotted lines form @xmath180 confidence intervals.,scaledwidth=100.0% ] , for different values of @xmath174 .
the solid line is the average over @xmath173 iterations the dotted lines form @xmath180 confidence intervals.,scaledwidth=100.0% ] , for different values of @xmath174 .
the solid line is the average over @xmath173 iterations the dotted lines form @xmath180 confidence intervals.,scaledwidth=100.0% ] , for different values of @xmath174 .
the solid line is the average over @xmath173 iterations the dotted lines form @xmath180 confidence intervals.,scaledwidth=100.0% ] , for different values of @xmath174 .
the solid line is the average over @xmath173 iterations the dotted lines form @xmath180 confidence intervals.,scaledwidth=100.0% ] these figures exhibit different behaviours depending on the difficulty of the problems ( increasing with @xmath3 and more importantly with @xmath42 ) .
the graphs in figure [ fig:1 ] for @xmath182 ( and @xmath183 ) exhibit first a very fast decay of the risk , until some critical threshold @xmath184 where @xmath185 seems to be between @xmath186 and @xmath187 . at this point
, one can actually observe that the method recovers in most case the true rank @xmath42 of the matrix , whereas it before recovered only a smaller rank approximate of @xmath71with a too small @xmath8 , it could not distinguish all the signal from the noise , and the fact that it gradually does for larger @xmath8 explains the fast decay of the logarithm of the rescaled risk .
after that , the curve has a kink and the decay becomes slower ( the theory predicts that the logarithm of the rescaled risk should decrease with @xmath8 as @xmath188 ) .
after this kink , all the @xmath42 rank directions " have been identified , and the logarithm of the rescaled risk starts decreasing slower , according to the theoretical rate of @xmath188 .
the graphs in figure [ fig:1 ] for @xmath189 ( and @xmath183 ) exhibit mainly the first regime , since @xmath42 is larger and the second regimes comes for larger values of @xmath8empirically , we can observe that the method recovers most of the time all @xmath42 directions " as soon as @xmath190 for @xmath191 , as soon as @xmath192 for @xmath193 .
a parallel evolution can be observed in figure [ fig:2 ] , for the logarithm of the average length of the confidence intervals .
it is not at all surprising since this length is supposed to reflect the risk .
the averaged coverage of these intervals in figure [ fig:3 ] is in average larger than @xmath194 in all cases , and in more than @xmath180 of the cases , it is higher than @xmath195 in all cases , which makes our method reliable .
an important application that satisfies our assumptions and to which our method can be applied is _ quantum tomography _
, i.e. the estimation of quantum states .
we consider the popular problem of multiple ion tomography , i.e. the problem of estimating the joint quantum state of @xmath196 two - dimensional systems ( qubits ) , as encountered in ion trap quantum tomography , see @xcite or @xcite for textbooks on this problem .
such a system s _ quantum state _ can be represented by a positive semi - definite unit trace complex matrix @xmath71 ( a _ density matrix _ ) of dimension @xmath197 . for each individual qubit , the experimenter can measure one of the three pauli observables described by the @xmath198 by @xmath198 _ pauli matrices _
@xmath199 , where @xmath200,\ , \sigma^2=\left [ \begin{array}{cc } 0 & -i\\ i & 0 \end{array } \right],\ , \sigma^3 = \left [ \begin{array}{cc } 1 & 0\\ 0 & -1 \end{array } \right],\ ] ] and each of these measurements may yield one of two outcomes , denoted by @xmath201 and @xmath202 respectively .
therefore , a full experiment ( i.e. an experiment that describes the measurement for each of the @xmath196 qubits ) is then defined by a setting @xmath203 where each @xmath204 for @xmath205 , which specifies which of the @xmath206 pauli observables is measured for each qubit . for each fixed setting @xmath207 , the measurement produces random outcomes @xmath208 , with expected probability @xmath209 where @xmath210 where @xmath211 is the eigen projector of the the @xmath198 by @xmath198 pauli matrix @xmath212 associated to the eigen value @xmath213 ( we remind that the @xmath198 by @xmath198 pauli matrices have eigen values of either @xmath201 or @xmath202 ) .
now set @xmath214,\ ] ] for the last @xmath215 pauli matrix such that @xmath216 form an orthogonal basis of @xmath217 .
let @xmath218 be the outcome of an experiment given a setting @xmath203 ( where each @xmath219 ) .
write @xmath220 for a setting where a subset @xmath221 of the @xmath196 matrices @xmath212 of this setting have been replaced by @xmath222 , and @xmath223 for the outcome where the same subset @xmath224 of the @xmath196 elements @xmath213 have been replaced by @xmath64 .
since the only eigen value of @xmath222 is @xmath64 , the outcome of the measurement of a qubit by @xmath222 is always @xmath64 .
this implies in particular that the distribution of @xmath223 as described above , is the same as the distribution of the outcome of an experiment when the measurement setting is @xmath220 .
for this reason , measuring according to setting @xmath207 actually gives information about all settings @xmath220 for any subset @xmath224 of @xmath225 .
thus instead of measuring all settings which are tensor products of @xmath226 pauli matrices @xmath227 , it is enough to measure all settings which are tensor products of @xmath226 pauli matrices @xmath228 , as they provide information about corresponding settings that involve @xmath222 . therefore if one measures all @xmath229 settings that correspond to the settings @xmath203 where each @xmath204 , we have observations about all measurements direction , and our measurement setting is complete
now we are interested in also dealing with situations where one does not want to observe all @xmath229 settings , and where one has only a number of settings @xmath230 .
we consider a random measurement setting as in @xcite , i.e. each measurement setting @xmath207 is chosen uniformly at random ( each @xmath212 is chosen uniformly at random among @xmath231 ) .
let @xmath232 be the total number of measurement setting chosen in this random way . for each chosen measurement setting @xmath207 , we perform @xmath14 repetition of the experiment and observe @xmath14 independent outcomes .
so for each chosen measurement setting @xmath233 with @xmath234 , we observe @xmath14 independent outcomes @xmath235 which are observations according to setting @xmath233 .
it is often convenient to express the information contained by a measurement @xmath236 in a way that involves tensor products of @xmath226 pauli matrices , rather than their spectral projections , see @xcite .
the set of matrices that are created by @xmath196 tensor products of @xmath226 pauli matrices @xmath227 is exactly the set of @xmath237 pauli matrices rescaled by @xmath238 ( introduced briefly in remark [ rem : gaussian ] ) i.e. the @xmath239 rescaled pauli basis .
indeed let @xmath240 , then one easily verifies that @xmath241 where @xmath242 is the expectation according to the outcome @xmath218 when measurement @xmath207 is chosen .
in this sense , the measurement described by the @xmath239 rescaled pauli matrix @xmath243 can be measured by the parity of the spins @xmath244 that one gets when performing measurement @xmath207 : in fact @xmath245 is a random variable such that its value is @xmath64 with probability @xmath246 and @xmath202 with probability @xmath247 - and its expectation is @xmath248 as noted .
we write @xmath249 for this distribution .
let us go back to our experimental setup .
as remarked before , each of the @xmath232 quantum measurement settings according to setting @xmath233 ( with @xmath234 ) , where @xmath250 and where each @xmath251 for @xmath252 provides us with @xmath7 information in the sense of our trace regression model , i.e. we observe at each measurement @xmath233 , for each replication @xmath253 and for all @xmath254 @xmath255 in our trace regression model , we can average the observations @xmath256 and we have the the following averaged observations for any @xmath257 and for all @xmath254 @xmath258 where @xmath259 is the @xmath260 repetition ( among @xmath14 iterations ) of the observation and where @xmath261 is the averaged noise and is such that @xmath262 , and such that @xmath261 is sub - gaussian has a sum of bounded random variables and satisfies @xmath263 for any @xmath264 . now in order for our assumption [ ass : designbis ] to be satisfied for rank @xmath42 matrices for a large enough number of settings @xmath232 , we need to rescale our data .
we set @xmath265 where @xmath266 is the cardinality of @xmath224 and where @xmath267 is the rescaled noise and is such that @xmath268 , and such that @xmath269 is sub - gaussian and satisfies @xmath270 for any @xmath264 .
it is a direct consequence from the results of @xcite and from our remark [ rem : gaussian ] that if @xmath271 , then with high probability on the random draws of our settings we have that assumption [ ass : designbis ] is satisfied for rank @xmath42 matrices .
we can therefore apply our method and other low rank recovery methods such as trace regression methods to our rescaled data @xmath272 we let @xmath273 so that @xmath274 , and let @xmath275 with @xmath276 and consider @xmath277 measurement settings . these settings with small @xmath42 and @xmath278 were chosen since we are more interested in the truncated measurement setting ( such that @xmath230 ) . for the replication , we consider @xmath279 .
using the data ( [ rescaleddata ] ) , we estimate @xmath71 by three methods our proposed method ( iht ) , the truncated maximum likelihood estimator ( mle ) for the high dimensional multiple ion tomography model as described in @xcite , and nuclear norm penalization ( nnp ) method computed using a gradient descent method ( e.g. * ? ? ?
we use the same tuning parameters as in ( [ emprisk ] ) , ( [ upsilonr ] ) , ( [ t1 ] ) , and ( [ tr ] ) and also we select @xmath280 and the same stopping rule as we describe in subsection [ s : sim ] . for the stepsize of the gradient descent method used to compute the nnp estimator , we follow the recommendation in subsection 3.1 in @xcite .
norm ( linf ) and shatten @xmath64 norm ( shatten ) of @xmath281 in function of @xmath278 ( and therefore in function og the number of settings @xmath232 ) , for different values of @xmath7 for replication @xmath282 and @xmath283 using the three methods described.,scaledwidth=100.0% ] figure [ fig:4 ] and [ fig:5 ] present the result when the number @xmath14 of replications is @xmath7 and when the true rank is 1 or 2 respectively , and for four different values of @xmath7 .
we provide average values of squared frobenius norm , operator norm , entrywise @xmath2 norm , and shatten @xmath64 norm of @xmath284 averaged over 100 iterations .
red dots , blue blank triangles , and black asterisks are average value of iht , mle , and nnp , respectively .
intuitively when @xmath278 increases these risks will decrease .
our estimator shows almost comparable result to the mle except in the case where @xmath285 and @xmath286 . in this case ,
iht estimates @xmath71 by @xmath63 a few times and pretty well for most cases so that on average the frobenius norm is still large .
figures [ fig:6 ] and [ fig:7 ] present the result when the replication is @xmath287 and when the true rank is 1 or 2 respectively , and for four different values of @xmath7 .
they show similar patterns as for the cases @xmath282 , but we can see that iht performs well especially for @xmath288 .
an interesting feature that all these pictures illustrate is that iht performs the best relatively to other methods for a large number of replication @xmath14 , and for difficult problems with high @xmath7 and @xmath42 ( see in particular the plots in figure [ fig:7 ] for large @xmath7 ) .
note that our method is computationally more efficienct than the other two methods in the sense that when @xmath289 , iht takes about 40 seconds while mle ( and even nnp ) takes about 2.5 minutes for one iteration , on a regular laptop , i.e. it is about four times slower .
norm ( linf ) and shatten @xmath64 norm ( shatten ) of @xmath281 in function of @xmath278 , for different values of @xmath7 for replication @xmath282 and @xmath290 using three methods.,scaledwidth=100.0% ] norm ( linf ) and shatten @xmath64 norm ( shatten ) of @xmath281 in function of @xmath278 , for different values of @xmath7 for replication @xmath291 and
@xmath283 using three methods.,scaledwidth=100.0% ] norm ( linf ) and shatten @xmath64 norm ( shatten ) of @xmath281 in function of @xmath278 , for different values of @xmath7 for replication @xmath291 when @xmath290 using three methods.,scaledwidth=100.0% ]
for convenience in writing the proofs , we introduce the following quantities .
we write , for integers @xmath292 , the vectorisation of a @xmath293 matrix ( where @xmath294 ) @xmath23 by stacking the rows of @xmath295 as @xmath296 we write the kronecker product between two matrices @xmath23 and @xmath77 as @xmath297 .
consider the @xmath298 matrix @xmath299 such that @xmath300 for @xmath46 and for @xmath301 where @xmath302 : @xmath303 = \left [ \begin{array}{cccccccc } x^1_{1,1 } & x^1_{1,2 } & \ldots & x^1_{1,d } & \ldots & x^1_{d,1 } & \ldots & x^1_{d , d } \\
x^2_{1,1 } & x^2_{1,2 } & \ldots & x^2_{1,d } & \ldots & x^2_{d,1 } & \ldots & x^2_{d , d } \\ \vdots & \vdots & \vdots & \vdots & \vdots & \vdots & \vdots & \vdots \\ x^n_{1,1 } & x^n_{1,2 } & \ldots & x^n_{1,d } & \ldots & x^n_{d,1 } & \ldots & x^n_{d , d } \end{array } \right ] .
$ ] let @xmath304 be the set of vectorization of matrices in @xmath305 , that is , @xmath306 . if @xmath307 , then @xmath308 contains a vector @xmath309 of dimension @xmath310 such that @xmath311 for @xmath312 . assumption [ ass : designbis ] can be rewritten as follows in this vectorized new notation .
[ ass : design ] let @xmath59 . for any @xmath60
, it holds that @xmath313 where @xmath62 .
assumption [ ass : design ] actually implies the following lemma that bounds the scalar products rather than the norms .
[ lem : tra ] if assumption [ ass : design ] holds , then for any @xmath314 , we have that @xmath315 the proof of this lemma is in subsection [ proof : tra ] . let @xmath316 be the set of vectors of @xmath317 and of norm @xmath64 .
[ [ explicit - writing - of - the - quantities ] ] 1 .
explicit writing of the quantities + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for any matrices @xmath318 of dimension @xmath319 with @xmath320 , we set @xmath321 \big)v \\ & = u^t \big(\frac{1}{n}\sum_{i\leq n } ( x^i)^t y_i^r \big)v = : u^t \hat \psi^r v\end{aligned}\ ] ] where @xmath322 and @xmath323 .
also we set @xmath324 and @xmath325 note that @xmath326 by linearity of the trace .
let @xmath327 , then @xmath328 is a scalar : @xmath329 let @xmath330 be the column vector of dimension @xmath310 such that @xmath331 , that is , @xmath332 for @xmath333 .
note that @xmath334 , and that @xmath335 .
consider the @xmath298 matrix @xmath299 such that @xmath300 for @xmath46 .
consider the column vector @xmath336 where @xmath337 .
then we have @xmath338 where here @xmath339 is the classic vectorial scalar product on @xmath54 . also by definition of @xmath340 and @xmath341
@xmath342 the last equation implies that @xmath343 [ [ bound - on - the - stochastic - term ] ] 2 . bound on the stochastic term + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we first bound the second term in ( [ eq : dec ] ) with the following lemma . [ helo ]
assume that @xmath344 ( note that @xmath345 for @xmath346 ) .
it holds with probability larger than @xmath65 that @xmath347 where @xmath83 is an universal constant .
its proof is in subsection [ proof : helo ] .
lemma [ helo ] implies that on an event of probability larger than @xmath65 , we can bound the stochastic term in ( [ eq : dec ] ) @xmath348 let @xmath349 be an event of probability larger than @xmath65 where the above holds . [
[ bound - on - the - first - term - in - eqdec - provided - that - the - rank - kr - of - psir - is - smaller - than-2k ] ] 3 . bound on the first term in ( [ eq : dec ] )
provided that the rank @xmath350 of @xmath351 is smaller than @xmath352 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + let us assume , only for this paragraph 3 .
of the proof , that the rank @xmath350 of @xmath351 is smaller than @xmath353 . by lemma
[ lem : tra ] , we can apply equation ( since @xmath354 ) , and combining this with the fact that @xmath355 , we have @xmath356 by combining equation , and , and using @xmath357 , we then have in the case that @xmath358 that on @xmath349 @xmath359 since the previous result holds in the worst case of @xmath360 , we directly have on @xmath349 the corresponding entrywise result whenever @xmath361 @xmath362 by definition , we know @xmath363 and @xmath364 , which gives on @xmath349 whenever @xmath361,@xmath365 note also by definition of the thresholding process , the matrix @xmath366 is such that it is diagonal with all diagonal elements smaller than @xmath153 .
let @xmath367 and @xmath368 for @xmath369 . by elementary calculations , we have @xmath370 and so we have that @xmath371 by definition of @xmath94
, we have @xmath372 so this implies that @xmath373 combining this with equation , we obtain that on @xmath349 and whenever @xmath361 @xmath374 [ [ induction ] ] 4 .
induction + + + + + + + + + + + + we now stop assuming that the rank @xmath350 of @xmath351 is smaller than @xmath352 , and we consider the general case . we are going to prove by induction that on @xmath349 , for any integer @xmath86 , we have that ( i ) the rank of @xmath375 is smaller than @xmath42 , and ( ii ) @xmath376 . for @xmath377 , since @xmath378 , then its rank is @xmath63 and is therefore bounded by @xmath42 and ( i ) is satisfied . moreover , since @xmath379 , then ( ii ) is satisfied as well .
now assume that ( i ) and ( ii ) hold on @xmath349 for a given @xmath91 ( as it holds for @xmath380 not only on @xmath349 but on the entire probability space ) .
by induction assumption ( i ) , we have that on @xmath349 the rank of @xmath375 is smaller than @xmath42 , which implies that the rank of @xmath381 is smaller than @xmath382 . because we have that @xmath361 , equation applies and on @xmath349 @xmath383 by definition of @xmath153 and since @xmath384 ( since @xmath353 ) .
moreover , in the same way , we have that on @xmath349 ( see equation since @xmath361 ) @xmath385 let us now state the following lemma .
[ lem : mati ] let @xmath386 be a matrix ( with @xmath387 ) , with singular values @xmath388 ordered in decreasing order ( all positive ) . for any @xmath389 and for any collection of orthogonal vectors
@xmath390 , we have @xmath391 write @xmath392 for the singular values of @xmath393 ordered in decreasing order and all positive ( and @xmath394 for the diagonalising matrices ) .
let @xmath395 be the matrices that diagonalise @xmath71 and order its singular values in decreasing order on the diagonal and write @xmath396 for its singular values ( all positive ) .
by lemma [ lem : mati ] , we know that , for any @xmath397 , @xmath398 therefore , on @xmath349 , by equation , we know that for any @xmath397 @xmath399 so since all @xmath400 that are smaller than @xmath153 are thresholded for constructing @xmath94 ( we remind that the @xmath401 are the diagonal elements of the diagonal matrix @xmath402 , that is thresholded at level @xmath153 in the construction of @xmath94 ) , it means that on @xmath349 , the rank of @xmath94 is smaller than the rank of @xmath71 , i.e. it is smaller than @xmath42 .
this proves the first part of the induction ( i ) .
now let @xmath403 be the matrices that diagonalise @xmath404 , and let @xmath405 . by , we have that on @xmath349 @xmath406 now since the rank of both @xmath94 and @xmath71 are smaller than @xmath42 on @xmath349 , we know that the rank of @xmath404 , and thus of @xmath407 , is smaller than @xmath352 . therefore , we have since @xmath407 is diagonal and has therefore only @xmath352 non - zeros elements that on @xmath349 @xmath408 which implies that on @xmath349 , since the frobenius norm is invariant by rotation @xmath409 this concludes part ( ii ) of the induction and therefore , it concludes the induction . [ [ conclusion ] ] 5 . conclusion + + + + + + + + + + + + + by the previous induction , we know that on @xmath349 , we have @xmath410 and also that @xmath411 and also that for any @xmath4 @xmath412 this concludes the proof since for @xmath91 larger than @xmath413 with @xmath414 a large enough constant , we have by definition of the sequence @xmath153 that @xmath415 first , note that for @xmath416 , we have @xmath417 thus , without loss of generality , we consider @xmath418 . we know that @xmath419 and @xmath420 this gives @xmath421 by assumption [ ass : design ] ,
using @xmath422 , we have for @xmath314 , @xmath423 this concludes the proof . since @xmath49
, we have that @xmath424 this implies that ( using a gaussian tail probability @xmath425 for @xmath426 when @xmath427 ) with probability larger than @xmath65 @xmath428 since @xmath299 satisfies the assumption [ ass : design ] with @xmath346 , we have that @xmath429 which implies that for any @xmath430 , we have @xmath431 equation implies together with equation that for a matrix @xmath430 , with probability larger than @xmath65 , @xmath432 where @xmath433 . to obtain the bound for the supremum of the quantity in ( [ eq : ber2 ] ) over all @xmath434 , we consider the approximating set @xmath435 whose property is described as follows
. let @xmath436 .
let , for any @xmath437 , @xmath438 be a @xmath439 covering set of @xmath440 .
here we use a classic result ( * ? ? ? * lemma 3.1 ) , saying that the @xmath441-covering numbers of @xmath440 is bounded by @xmath442 .
thus the cardinality of @xmath438 is smaller than @xmath443 .
let @xmath444 be the event such that for all @xmath445 and for each vector in @xmath446 , it holds that @xmath447 where @xmath448 , where @xmath449 is a large constant . by equation , and since @xmath450
we know that @xmath451 holds with probability @xmath452 for each @xmath453 and for each vector @xmath446 . by a union bound , we have that @xmath454 since @xmath449 .
let now @xmath455 such that @xmath456 .
it is possible to write @xmath23 as @xmath457 where each @xmath458 belongs to @xmath438 , and where the @xmath459 are such that @xmath460 .
we have on @xmath444 that @xmath461 this concludes the proof .
let @xmath462 be the singular vectors of @xmath463 , i.e. @xmath464 .
let @xmath465 .
the dimension of @xmath224 is @xmath466 .
let now @xmath467 be the vectorial sub - space that is orthogonal to @xmath468 .
its dimension is @xmath469 .
since @xmath470 , there is at lest one unitary vector in @xmath471 .
let @xmath472 be this vector , since it is in @xmath224 , it can be written as @xmath473 where for @xmath474 , we have @xmath475 and @xmath476 .
therefore , we have that @xmath477 consider @xmath478 .
so we have that @xmath479 since the @xmath480 are orthonormal , and since the singular values are positive and ordered in decreasing order .
this concludes the proof . from the proof of theorem
[ th : mainthm2 ] , equation , we know that the estimator @xmath128 satisfies @xmath481 i.e. the desired result in operator norm and from which the results on the rank and in the other schatten norm follow because of the thresholding . also since the sequence @xmath153 is an arithmetico - geometrical sequence converging to @xmath482 , of arithmetic term @xmath483 and of geometric term @xmath139
, it is clear that @xmath124 is such that @xmath484 i.e. @xmath132 this concludes the proof . by definition , we have that @xmath485
let @xmath486 and let @xmath487 be the vector with all element equal to @xmath63 except the @xmath488 entry which is equal to @xmath64 , and we consider that @xmath489 .
we have by definition and using representations ( [ simple1 ] ) and ( [ simple2 ] ) that @xmath490 and @xmath491 note that given @xmath492 , @xmath493 and @xmath494 the following lemma is a concentration inequality that holds in gaussian design . [ lem : concer ] assume that the design is gaussian .
let @xmath55 .
we have that with probability larger than @xmath65 ( on the design ) @xmath495 let @xmath496 .
we have @xmath497 where @xmath498 is the @xmath499 identity matrix .
this implies that @xmath500 where @xmath501 is the chi square distribution with @xmath466 degrees of freedom . by bernstein s inequality ,
we thus have ( since the @xmath502 distribution is sub - gaussian ) that , with probability larger than @xmath65 , @xmath503 where @xmath83 is an universal constant .
this implies that , with probability larger than @xmath65 , @xmath504 this concludes the proof . combining lemma [ lem : concer ] with pythagoras s theorem as in the proof of lemma [ lem : tra ]
, we have that for any @xmath505 , with probability larger than @xmath65 , @xmath506 by a union bound , this implies that with probability larger than @xmath65 , @xmath507 where @xmath83 is a universal constant .
this concludes the proof ( in remarking that the above quantity is arbitrarily small when @xmath508 ) .
[ [ acknowledgements ] ] acknowledgements + + + + + + + + + + + + + + + + we would like to thank richard nickl , richard samworth and rajen shah for insightful comments and discussions .
part of this work was produced when ac was in the statslab in the university of cambridge .
ac s work is supported since 2015 by the dfg s emmy noether grant musyad ( ca 1488/1 - 1 ) .
# 1 acharya , a. , t. kypraios , and m. guta ( 2015 ) .
efficient quantum tomography with incomplete measurement settings . .
agarwal , a. , s. negahban , and m. j. wainwright ( 2012 ) .
fast global convergence of gradient methods for high - dimensional statistical recovery . *
40*(5 ) , 24522482 .
bickel , p. , y. ritov , and a. tsybakov ( 2009 ) .
simultaneous analysis of lasso and dantzig selector . * 37 * , 17051732 .
blumensath , t. and m. e. davies ( 2009 ) .
iterative hard thresholding for compressed sensing . *
27*(3 ) , 265274 .
bunea , florentina , y. she , and m. wegkamp ( 2011 ) .
optimal selection of reduced rank estimators of high - dimensional matrices . *
39*(2 ) , 12821309 .
butucea , c. , m. gu , and t. kypraios ( 2015 ) .
spectral thresholding quantum tomography for low rank states . _
17_(11 ) , 113050 .
cai , t. and h.h zhou ( 2012 ) .
optimal rates of convergence for sparse covariance matrix estimation . *
40 * , 23892420 .
cai , t. and a. zhang ( 2015 ) .
rop : matrix recovery via rank - one projections . * 43 * , 102138 .
cands , e. and t. tao ( 2010 ) .
the power of convex relazation : near - optimal matrix completion . *
56 * , 20532080 .
cands , e. j. and y. plan ( 2011 ) .
tight oracle bounds for low - rank matrix recovery from a minimal number of random measurements . *
57*(4 ) , 23422359 .
cands , e. and b. recht ( 2009 ) .
exact matrix completion via convex optimization .
* 9 * , 717772 .
carpentier , a. , j. eisert , d. gross , and r. nickl ( 2015 ) .
uncertainty quan- tification for matrix compressed sensing and quantum tomography problems . .
chen , y. and m. wainwright ( 2015 ) .
fast low - rank estimation by projected gradient descent : general statistical and algorithmic guarantees .
flammia , s. t , d. gross , y .- k .
liu , and j. eisert .
quantum tomography via compressed sensing : error bounds , sample complexity and efficient estimators .
, 14(9):095022 , 2012 .
friedman j. and t. hastie and r. tibshirani ( 2008 ) .
sparse inverse covariance estimation with the graphical lasso . *
9*(3 ) , 432441 . van de geer , s. , p. bhlmann , y. ritov , and r. dezeure ( 2014 ) .
on asymptotically optimal confidence regions and tests for high - dimensional models .
* 42*(3 ) , 11661202 .
goldfarb , d. and s. ma ( 2011 ) .
convergence of fixed - point continuation algorithms for matrix rank minimization . * 11 * , 183210 .
gross , d. ( 2011 ) . recovering low - rank matrices from few coefficients in any basis . *
57*(3 ) , 15481566 .
gross , d. , y .- k .
liu , s. t flammia , s. becker , and j. eisert .
( 2010 ) quantum state tomography via compressed sensing .
, 105(15):150401 .
guta , m. , t. kypraios , and i. dryden ( 2012 ) .
rank - based model selection for multiple ions quantum tomography .
_ 14 _ , 105002 .
haeffner , h. , w. haensel , c. f. roos , j. benhelm , d. c. al kar , m. chwalla , t. koerber , u. d. rapol , m. riebe , p. o. schmidt , c. becher , o. ghne , w. dur , and r. blatt ( 2005 ) .
scalable multi - particle entanglement of trapped ions .
_ 438 _ , 643 .
holevo , a. s. ( 2001 ) . .
springer .
huang , h. and s. ma and c - h .
zhang ( 2008 ) .
adaptive lasso for sparse high - dimensional regression models . * 18 * , 16031618 .
javanmard , a. and a. montanari .
( 2014 ) confidence intervals and hypothesis testing for high - dimensional regression . *
15 * ( 1 ) , 28692909 .
klopp , o. matrix completion by singular value thresholding : sharp bounds .
* 9 * , ( 2 ) .
23482369 .
knight , k. and w. fu ( 2000 ) .
asymptotics for lasso - type estimators .
* 28 * , 13561378 .
koltchinskii , v. ( 2011 ) . on neumann entropy penalization and low - rank matrix estimation . *
39*(6 ) , 2936 - 2973 .
koltchinskii , v. , k. lounici , and a. b. tsybakov ( 2011 ) .
nuclear - norm penalization and optimal rates for noisy low - rank matrix completion . *
39*(5 ) , 23022329 .
koltchinskii , v. , d. xia ( 2015 ) . optimal estimation of low rank density matrices .
arxiv preprint arxiv:1507.05131 .
liu , y.k .
universal low - rank matrix recovery from pauli measurements .
, 16381646 .
meinshausen , n. and p. bhlmann ( 2006 ) .
high dimensional graphs and variable selection with the lasso . *
34 * , 14361462 .
needell , d. and j.a .
tropp ( 2009 ) .
cosamp : iterative signal recovery from incomplete and inaccurate samples .
* 26 * , 301321 .
negahban , s. and wainwright , m. ( 2011 ) .
estimation of ( near ) low - rank matrices with noise and high - dimensional scaling . * 39 * , 10691097 .
nickl , r. and van de geer , s. ( 2014 ) .
confidence sets in sparse regression .
* 41*(6 ) , 2852 - 2876 .
nielsen , m. a. and i. l. chuang ( 2000 ) . .
cambridge : cambridge university press .
recht , b. ( 2011 ) . a simpler approach to matrix completion .
* 12 * , 34133430 .
tanner , j. and k. wei ( 2012 ) .
normalized iterative hard thresolding for matrix completion .
* 35 * , s104s125 .
zhang , c - h . and
zhang , s - s . (
confidence intervals for low dimensional parameters in high dimensional linear models . *
76 * , 217242 .
the method that we proposed and studied in the low rank matrix recovery setting can be adapted and simplified to accommodate another setting : the sparse linear regression setting .
we explain how to construct an estimator based on iht , and prove that the estimator is efficient in @xmath0 and @xmath2 norm , and provide the limiting distribution of a simple modification of our estimate .
consider the linear model @xmath513 where @xmath514 is a @xmath515 matrix , the signal vector @xmath516 is @xmath42-sparse ( @xmath517 ) , and @xmath48 is an i.i.d .
vector of gaussian white noise , i.e. @xmath49 ( as in the matrix regression , we do not need the gaussian assumption and our results hold with sub - gaussian independent noise ) , and @xmath518 .
we denote the sample covariance matrix by @xmath519 .
[ rem : ass ] suppose @xmath514 is from a distribution whose covariance matrix is @xmath524 .
let the minimum eigenvalue @xmath525 and the maximum eigenvalue @xmath526 and @xmath527 } \sigma_{ii } \leq 1 $ ] .
assume that @xmath528 has independent sub - gaussian rows with zero mean and sub - gaussian norm @xmath529 .
then from the paper @xcite , for @xmath530 , with probability larger than @xmath531 with @xmath532 , there exists a computationally feasible @xmath35 such that @xmath533 holds . in this case
, we can take @xmath534 .
this algorithm takes again three parameters : @xmath535 and @xmath77 .
we have the same interpretation for @xmath74 as in the matrix regression setting , @xmath75 is an upper bound on two times the sparsity of @xmath536 ( again , it does not need to be tight as long as @xmath537 is small enough ) , and @xmath77 is a loose bound on the @xmath1 norm of @xmath536 .
then we update thresholds in each iteration @xmath540 , by @xmath541 where @xmath542 where @xmath543 . recall that the pseudo inverse @xmath35 of @xmath544 and @xmath537
are taken from assumption [ ass : matrix ] .
set now recursively , @xmath545 and @xmath546 this procedure provides a sequence of estimates , and as we will prove in the next subsection , this sequence is with high probability close to the true @xmath536 as soon as @xmath91 is of order @xmath92 ( see theorems [ th : mainthm ] and [ thm : asymnorm ] ) .
[ rem : iht ] the proposed method modifies iterative algorithms ( see e.g. * ? ? ?
* ; * ? ? ?
the usual ( normalised ) iht algorithm updates the estimate using @xmath547 where @xmath548 is a hard thresholding operator that keeps the largest @xmath42 elements of a vector and @xmath549 is a stepsize that can have the interpretation of a gradient step when it is much smaller than @xmath64 .
the difference is in the thresholding ; we update thresholds while they pick the largest @xmath42 values after adjusting the added parts .
most importantly , previous works on this estimator only considered the case of a deterministic ( small ) noise , so their analysis is not applicable in our model where the noise is stochastic .
[ th : mainthm ] assume that assumption [ ass : matrix ] is satisfied and that @xmath551 .
let @xmath552 .
we have that with probability larger than @xmath65 , for any @xmath99 , @xmath553 where @xmath554 and @xmath555 .
to prove asymptotic normality , we slightly modify the estimator defined in theorem [ th : mainthm ] .
this is similar to the de - sparsified lasso by @xcite in the sense that we also use a de - sparsified version of our estimator .
consider the estimator @xmath550 of theorem [ th : mainthm ] ( with the same @xmath560 ) and @xmath35 in assumption [ ass : matrix ] , and define @xmath561 [ thm : asymnorm ] suppose that the same assumptions and notation used in theorem [ th : mainthm ] hold .
then , writing @xmath562 and @xmath563 , we have @xmath564 where @xmath565 .
if @xmath566 ( e.g. for designs as in remark [ rem : ass ] , we have @xmath567 so it suffices that @xmath568 ) then we also have @xmath117 by an union bound ( with hoeffding s inequality ) we know that with probability larger than @xmath65 @xmath571 let @xmath349 be the event of probability @xmath65 where the previous equation is satisfied . by equation ( [ eq : superpoule ] ) we know that on @xmath349 , since @xmath581 is @xmath42 sparse @xmath582 since @xmath583 . since @xmath584 , we have that on @xmath349 , by equation ( [ equ ] ) , all the coordinates @xmath466 of @xmath585 such that @xmath586 are set to @xmath63 .
this implies that the support of @xmath585 ( and thus the support of @xmath587 ) is included in the support of @xmath536 on @xmath349 .
therefore , @xmath585 is @xmath588sparse on @xmath349 .
also , still since @xmath584 , we have that @xmath589 and this implies together with equation ( [ equ ] ) that on @xmath349 , we have @xmath590 this concludes the proof for @xmath591 .
the induction is complete , and we have that the previous equation holds for all @xmath86 .
it is equivalent to the fact that on @xmath349 ( and thus with probability larger than @xmath65 ) , for all @xmath86 @xmath592 and the support of @xmath550 is included in the support of @xmath536 . by definition , @xmath597 given @xmath514 , we know that @xmath120 is a linear function of the gaussian vector @xmath50 , thus @xmath598 now we prove the bound for @xmath110 .
note that using ( [ eq : pseudoinv ] ) and @xmath599 , for a sufficiently large @xmath8 , we have a constant @xmath67 such that @xmath600 then using the result from theorem [ th : mainthm ] , with probability at least @xmath65 , we have as long as @xmath601 @xmath602 as @xmath603 .
|
we consider the problem of low rank matrix recovery in a stochastically noisy high dimensional setting .
we propose a new estimator for the low rank matrix , based on the iterative hard thresholding method , and that is computationally efficient and simple .
we prove that our estimator is optimal both in terms of the frobenius risk , and in terms of the operator norm risk , i.e. in terms of the entry - wise risk uniformly over any change of orthonormal basis .
this result allows us to provide the limiting distribution of the estimator . in the case where the design is gaussian
, we prove that the entry - wise bias of the limiting distribution of the estimator is small , which is of great interest for constructing tests and confidence sets for low dimensional subsets of entries of the low rank matrix .
|
extreme scattering events ( ese ) are a class of dramatic decreases ( @xmath3 50% ) in the flux density of radio sources near 1 ghz for several weeks bracketed by substantial increases (; fig .
[ fig : lightcurve ] ) .
because of their simultaneity at different wavelengths and light travel time arguments , eses are likely a propagation effect @xcite .
first identified toward extragalactic sources , eses have since been observed toward pulsars @xcite . to date , the only other observational constraints on the structures responsible for eses besides the light curves are the lack of pulse broadening and the variation in the pulse times of arrival during the pulsar eses .
this paper summarizes constraints obtained during the ese toward the quasar 1741@xmath0038 .
we discuss faraday rotation measurements by in
[ lazio : sec : rm ] , vlbi imaging by in
[ lazio : sec : vlbi ] , and absorption measurements by in
[ lazio : sec : hi ] .
we present our conclusions in [ lazio : sec : conclude ] .
figure [ fig : lightcurve ] shows the ese of 1741@xmath0038 with the epochs of the various observations indicated .
at each epoch , the polarization position angle @xmath4 was measured at 6 to 10 frequencies and then fit , by minimizing @xmath5 and accounting for @xmath6 ambiguities , as a function of the observing wavelength , @xmath7 .
the same procedure was used for 1741@xmath0038 and for the parallactic angle calibrator 1725@xmath8044 .
based on the relative invariability of rm toward 1725@xmath8044 and 1741@xmath0038 , conclude @xmath9 rad m@xmath10 during the .
this upper limit implies an upper limit on the mean magnetic field parallel to the line of sight of @xmath1 for a typical value of the free electron column density through the structure , @xmath11 pc @xmath12 @xcite .
this upper limit is much larger than the typical interstellar field strength but enhancement of the ambient field to @xmath13 mg is possible within a shock front @xcite .
alternately , @xmath14 may be small if the field is disordered or if the ionized region is not magnetized .
comparison of the visibility data during the ese to those obtained after the ese ( 1994 july 8) show the source to be more resolved during the .
the excess angular broadening is @xmath15 mas , implying that the ese structure contributed a scattering measure @xmath16 . in turn
, the pulse broadening of a background pulsar should be @xmath17 @xmath18s at 1 ghz , consistent with the observed lack of broadening during pulsar eses @xcite .
the refractive models commonly used to explain eses predict that the source s flux density and angular diameter should be highly correlated .
we observe an _ anti - correlation_. simple stochastic broadening models require much more scattering ( 2 mas ) than is observed .
we consider it likely that both refractive defocussing and stochastic broadening are occurring .
we were unable to test a key prediction of refractive models
angular position wander of the background source because these observations had no absolute position information .
a second prediction is multiple imaging . during this ese , any secondary image(s )
must have been extremely faint ; multiple imaging , with the secondary image slightly offset from the primary , is unlikely to explain the increase in angular diameter because no other effects of strong refraction are seen ( cf . also ) .
we also observe little , if any , ese - induced anisotropy in the vlbi images .
if ese lenses are filamentary structures @xcite , they must be extended _ along _ the line of sight , a possibility also suggested by .
at all epochs the opacity spectra show the presence of a strong absorption feature near 5 and a typical rms determined outside the line of @xmath19 .
there is no gross change in the absorption line during the ese nor do any additional absorption components appear . between any two epochs @xmath20 ( @xmath21 ) .
this upper limit implies a neutral column density change of @xmath22 for a structure with a spin temperature @xmath23
. proposes interstellar tiny ( au)-scale atomic structures ( tsas ) in order to explain small ( angular ) scale changes in opacity .
tsas would have @xmath24 and @xmath25 k. our @xmath26 limit marginally excludes a connection between tsas and eses .
propose au - scale molecular clouds in the galactic halo precisely to explain eses .
the clouds would be cold , @xmath27 k , and dense enough to be opaque in the line .
eses would result from the photoionized skins of the clouds .
we see no @xmath28 features ( walker 2000 , private communication , has since suggested @xmath29 )
. however , the clouds could have velocities approaching 500 , while the observed velocity range is no more than 250 significant absorption could have been present outside of our velocity range .
salient aspects of this observational program are * @xmath30 rad m@xmath10 implying a magnetic field within the scatterer of @xmath1 . * no change in the vlbi structure , except a 0.7 mas _ increase _ in the angular diameter .
this increase is _ not _ consistent with that expected from a purely refractive model : eses must result from both broadening and defocussing within the ionized structures . *
@xmath31 implying that the column density associated with the ese structure is @xmath32 .
tiny - scale atomic structures are marginally ruled out ; -opaque , halo molecular clouds would be excluded , but the observed velocity range covers only 25% of the allowed range .
many people helped make these observations possible , most notably m. claussen , a. clegg , b. dennison , r. fielder , k. johnston , and e. waltman .
the nrao is a facility of the national science foundation operated by .
radio astronomy research at the nrl is supported by the office of naval research .
|
we review observational constraints on the structures responsible for extreme scattering events , focussing on a series of observations of the quasar pks 1741@xmath0038 .
vla observations were conducted to search for changes in the rotation measure and absorption during the ese , while vlbi observations sought ese - induced changes in the source s image .
no rm changes were found implying @xmath1 mg , and no opacity changes were found implying @xmath2 .
no multiple imaging was observed , but the diameter of the source _ increased _ by 0.7 mas , contrary to what is predicted by simple refractive lens modeling of eses .
we summarize what these limits imply about the structure responsible for this .
|
the study of multiparticle production in hard processes provides one with an important information about both the qcd partonic branching processes and the transition from the coloured partons to the colourless hadrons .
the vast amount of experimental data collected in various hard collisions processes convincingly confirms that the inclusive characteristics of qcd jet systems can be successfully described within the analytical perturbative approach to multiparticle production , see e.g. @xcite .
this approach is based on the so - called modified leading logarithmic approximation ( mlla ) @xcite and on the concept of local parton hadron duality ( lphd ) @xcite .
in particular , the experiment clearly demonstrates that the bright colour coherence effects survive the hadronization stage and are distinctively visible in the data .
this has been known for quite a long time for annihilation processes ( see e.g. @xcite ) . recently the very impressive data from hera @xcite and from the tevatron @xcite have become available .
the lphd allows one to relate the ( sufficiently ) inclusive hadronic observables to the corresponding quantities computed for the cascading partonic system .
one of the well known ( but still striking ) predictions of the perturbative scenario is the depletion of the soft particle production and the resulting approximately gaussian shape of the inclusive distribution in the variable @xmath10for particles with energy @xmath11 in a jet of energy @xmath12 ( the so - called `` hump - backed plateau '' ) @xcite .
this behaviour is observed in all inclusive energy distributions ; the charged particle spectrum for not too small momenta prove to be in surprisingly good agreement with the mlla - lphd predictions reflecting the qcd cascading picture of multiple hadroproduction .
moreover , the data collected in various hard scattering processes ( , dis , @xmath13 ) clearly demonstrate a remarkable universality of particle spectra assuming the proper ( mlla - based ) choice of the cascading evolution variables , equivalent to the @xmath14 energy @xmath15 .
the most challenging is the soft momentum end of the particle spectra ( p @xmath16gev ) where the non - perturbative dynamics could invalidate the perturbatively based expectations .
an attempt to stretch the perturbative predictions to the limit of their applicability ( or better to say , beyond it ) has been performed in previous papers of the present authors ( see e.g. @xcite ) . in these papers
it was shown that ( after the proper modifications ) the perturbatively based formulae allow a sufficiently smooth transition into the soft momentum domain .
the non - perturbative hadronization effects are encoded in the transverse momentum cutoff @xmath17 which can be motivated by the space - time picture of hadroproduction in qcd jets , see e.g. @xcite .
explicit predictions for distributions of hadrons in the soft region need some additional assumptions on mass effects .
scaling phenomena are expected to be of more general nature .
let us recall that the gluons of long wave length are emitted by the total colour current which is independent of the internal structure of the jet and is conserved when the partons split .
applying the lphd hypothesis one then expects that the hadron spectrum at very low momenta @xmath18 should be nearly independent of the jet energy @xcite .
as discussed in @xcite , the low momentum data support the basic ideas of qcd coherence and lphd .
quantitatively , the analysis was performed in terms of the invariant particle density @xmath19 for annihilation into hadrons at low momenta in quite a wide @xmath14 energy region ( from adone to lep-2 ) .
the spectra were found to be in a good agreement with the scaling behaviour and with analytical perturbative expectations which became sensitive to the running of the coupling @xmath20 at small scales @xcite .
the new h1 data@xcite follow these predictions as well , thus confirming the universality of soft particle production .
in this paper we study further the phenomenology of the soft particle production for both the charged and identified hadrons .
the analytical calculations are compared with the quantity @xmath21 which has some practical advantages over the invariant density considered before .
we derive a scaling prediction for particles at low transverse but arbitrary longitudinal momenta , which generalizes our previous results and which for very small @xmath22 explains the so - called fan - invariance experimentally observed long ago @xcite .
an important test of this line of approach towards soft particle production is the sensitivity to the colour charge of the primary emitters : the density of low momentum particles roughly doubles when going from a primary @xmath3 to a @xmath4 antenna .
it is interesting to note that this prediction applies to the soft particles already at present energies , whereas for the total multiplicity this doubling needs much higher energies .
the sensitivity of the low momentum particle density to the primary colour charges can be studied in experimentally accessible processes , such as in @xmath23 jets @xcite . here
this process is discussed further and we derive predictions for the soft particle production in quark and gluon exchange reactions which are observed in photoproduction or dis at hera or in @xmath1 collisions .
in refs . @xcite an approximate solution of the mlla evolution equation for the inclusive energy distribution @xmath24 of soft gluons , originating from a primary parton @xmath25 , has been derived .
this solution can be written as @xmath26 with @xmath27 here we have used the logarithmic variables @xmath28 , @xmath29 and @xmath30 with @xmath11 the particle energy and @xmath31 the jet virtuality @xmath32 for a jet of primary momentum @xmath12 and half opening angle @xmath33 ) ; @xmath34 is the respective colour factor , i.e. , @xmath35 and @xmath36 ; @xmath37 denotes the anomalous dimension of multiplicity and is related to the @xmath38 running coupling by @xmath39 or @xmath40 with @xmath41 , @xmath42 ; @xmath43 is the qcd - scale and @xmath44 and @xmath45 are the numbers of colours and of flavours respectively , @xmath46 . the shower evolution is terminated by the transverse momentum cut - off @xmath47 .
the first term in ( [ duetermini ] ) corresponds to the emission of a single gluon and yields the leading contribution for @xmath48 .
it is proportional to the colour charge of the primary parton .
furthermore , this term does not depend on the @xmath14 energy , contrary to the higher order terms which provide the rise of the spectrum for large @xmath11 with increasing @xmath15 .
the spectrum vanishes at @xmath48 as @xmath49 in order to relate parton and hadron distributions at low momentum @xmath50 , one has to make some additional assumptions on how to include mass effects . in @xcite
it was required that the invariant density @xmath51 for hadrons approaches a constant limit when @xmath52 as observed experimentally .
here it is worth to recall that at low @xmath18 @xmath53 where @xmath54 are the standard analogues of the dis structure functions @xmath55 , see e.g. @xcite .
as well known , @xmath56 are related to the matrix elements of the current commutators and should be regular when @xmath52 . for spectra which vanish as in ( [ dslimit ] ) a plausible relation between partons and hadrons can be achieved by relating their distributions as @xcite @xmath57 at the same energies @xmath58 where @xmath59 and @xmath60 is a normalization parameter .
if hadrons from both hemispheres are added , @xmath60 should be replaced by 2@xmath60 . for large energies @xmath61 ( [ phrel ] )
coincides with the usual lphd relation @xmath62 . in @xcite
we have considered the invariant density @xmath63 in the limit of vanishing rapidity @xmath64 and transverse momentum @xmath22 or , equivalently , for vanishing momentum @xmath65 , i.e. , @xmath66 where the factor 1/2 takes into account that both hemispheres are added in the limit @xmath67 .
then , since @xmath68 for small @xmath18 , the invariant hadronic density @xmath69 approaches the finite , energy independent limit as in ( [ izero ] ) @xmath70 in the fixed @xmath71 limit @xmath72 is replaced by @xmath37 and @xmath73 . with prescription ( [ phrel ] ) one obtains a good description of the moments of the energy spectrum @xmath74 by the mlla formulae with @xmath17 = 270 mev in a wide energy range @xcite , as well as of the particle spectrum in the soft region ( @xmath75 gev ) @xcite .
certainly at the moment there is no unique recipe allowing one to perform a smooth transition between the perturbative and non - perturbative regimes .
also there are certain approximations involved in the analytical results for low energy particles . in refs .
@xcite an alternative prescription has been discussed : @xmath76 where @xmath77 is the so - called limiting spectrum @xcite which solves the mlla evolution equations for @xmath78 = 0 in the full kinematic region , except very close to the boundary @xmath79 = y. at this phase space boundary the limiting spectrum approaches the finite limit @xmath80 , which equals numerically 1.069 ( 1.055 ) for @xmath45=3(5 ) . basing on the prescription ( [ phrel2 ] ) one arrives at the alternative formula for the invariant hadronic density which in the case of annihilation reads @xmath81 with @xmath82 .
as it is easy to see this prescription also guarantees that the hadronic density reaches a constant limit at @xmath52 .
we shall compare here the expectation based on eq .
( [ phrells ] ) with the experimental data for charged pions @xmath83 .
the mlla limiting spectrum taken at @xmath84 = 150 mev has been found to provide a good description of pion spectra at relativistic energies , see e.g. @xcite . recall that the low momentum region in charged particle spectra is dominated by the pions .
the energy dependence of the soft particle production rate was investigated previously for annihilation in terms of the invariant particle density @xmath85 in the wide energy interval ( from adone to lep-2 ) @xcite .
recently , new data on soft particle production became available from dis at hera @xcite . in the following we compare various results for the quantity @xmath86 as a function of the momentum @xmath87 of the registered particle . when presented in terms of this observable
the data points are independent of the choice of the theoretical parameter @xmath17 ; in this case , however , we are dealing with a quantity which is not directly related to the lorentz - invariant phase space . using this quantity may facilitate the comparison between different experimental results .
1 shows the momentum dependence of @xmath86 for various @xmath14 energies @xmath15 in annihilation in comparison with the theoretical predictions . in fig .
1a the distributions of charged particles are presented in comparison with the approximate mlla result ( [ dmlla ] ) with prescription ( [ phrel ] ) for the relation between partons and hadrons .
the theoretical prediction describes well the approach to the scaling limit for low momenta and also the shape and energy dependence of the spectra with the choice @xmath88 mev , as already discussed in @xcite .
the detailed behaviour at the very small momenta @xmath89 depends on the approximation at the parton level involving large production angles ( @xmath90 ) and also on the assumed relation between partons and hadrons in this region .
this point is illustrated in fig .
1b , where we show the theoretical predictions with the alternative choice of the low momentum behaviour discussed in sect .
2 , based on the approximate analytical mlla solution for @xmath91 ( limiting spectrum ) and relation ( [ phrel2 ] ) ; this form describes the pion spectra for @xmath92 gev with the mass parameter @xmath93 mev which is in agreement with the earlier fits to the pion spectra @xcite . with this prescription
the particle production rate in the soft limit @xmath94 is lower than in fig .
1a , but it also approaches an energy independent value in this limit and that is the essential prediction of the qcd approach based on coherent soft gluon emission .
looking at the data in fig . 1 more closely , there seems to be a certain increase of the soft particle rate at lep energies .
this is illustrated in fig . 2 a and b where the data below and above @xmath95 gev respectively in annihilation show separately a very good approach to an energy independent limit at small @xmath87 , whereas there is a difference in absolute rates of 20 - 30% in the two energy regions .
this is also visible in fig .
3a where we show the @xmath14 energy dependence of three values of the momentum @xmath87 .
we also include here the recent data from hera @xcite which closely match the results . at the lowest momentum @xmath96 gev
we see an almost energy independent production rate , but a small jump at the lep-1 energy is visible . a rise at the higher energies could partially be caused by the weak decay products of heavy quark particles ( perhaps also particles with strange quarks ) which would add incoherently to the spectrum of particles produced directly from the soft gluons .
this could happen in particular at the @xmath97 because of the larger branching ratio into @xmath98-quarks .
however , the experiment @xcite clearly shows that at low momenta ( @xmath99 gev ) the effect induced by the @xmath100 events is below 5% .
it would be interesting to clarify quantitatively in more details which part of the observed rise could be explained by the overall effect of the weak decays .
a certain small increase of the rate is actually also expected from the theory at the nonzero momenta ( see the curves in fig .
3a ) . for larger
momenta the rise of the production rate with energy is more pronounced and this trend is reproduced by the theoretical calculation for the full range of energy where the data are available ( @xmath101 gev ) . an approximate energy independence of the rate for the low momentum kaons and protons is well tested for @xmath92 gev where data at some common low momentum value are available . a comparison in figs . 3a and
3b suggests that the spectra become practically energy independent when the particle momenta are of the order of their mass or smaller .
it should be noted that the energy independence of the soft particle rate is by no means trivial .
we recall , in particular , that the particle density @xmath102 at central rapidity @xmath103 rises with @xmath14 energy .
this is illustrated in fig .
4 for tasso and opal data where rather precise results are available at @xmath104 gev .
whereas in this energy range ( 14 to 91 gev ) the plateau height rises by about 70% the corresponding rise for the soft particles at @xmath104 gev is only about 10% .
this is consistent with the qcd picture of particle production : the central plateau rises asymptotically with the same exponent @xmath105 as the full particle multiplicity and this rise comes mainly from the high transverse momentum particles .
the soft particle production is suppressed because of the coherence in the emission from all colour sources . in fig .
5 we consider for different identified particles with mass @xmath106 the rescaled spectra @xmath107 which are expected to approach energy independent dimensionless numbers in the soft limit @xmath94 .
first we see again , that in the pion spectra the violation of energy independence occurs at lower momenta than in the spectra of the heavier particles .
furthermore , these rescaled spectra of pions and kaons , separated by an order of magnitude around @xmath108 gev seem to converge towards a common value in the soft limit .
a similar convergence is also suggested for the protons although the lack of precise data at small momenta exclude a firm conclusion .
we find an approximate universality of the quantity @xmath109 for the particle species considered here .
this corresponds to our previous observation @xcite that the invariant density behaves like @xmath110 in the soft limit . in our theoretical picture
this quantity is given by @xmath111 ( and not @xmath112 , for example ) .
to the extent that @xmath17 is related to the particle mass such a universality could be anticipated .
it would be interesting to have more experimental information on low momentum particles to test better the validity of such a universal behaviour ; new hera data with identified particles at such small momenta could be helpful as well .
as discussed before the gluon spectrum in the soft limit becomes independent of energy of the primary parton above about 1 gev .
this energy independence can be observed , for example , in the @xmath14 of annihilation or in the breit frame of dis .
this result can be easily generalized .
we make a lorentz transformation along the initial parton direction , then in the new frame the initial partons have different energies but the soft particle production rate should remain unaltered as long as the primary energies are still sufficiently large ( and the colour of the primary parton has not changed ) . consequently , the quantity @xmath114 should be independent of rapidity @xmath64 for such transformations @xcite and this behaviour should still hold approximately for @xmath22 not much larger than the hadron masses . therefore ,
if the coherence argument is applied to this class of frames one obtains the prediction of scaling for the distribution @xmath115 equivalently , for @xmath116 with production angle @xmath117 one expects @xmath118 an approximate scaling behaviour of this type but for the distributions normalized to unity was found by the tasso collaboration @xcite in the energy range from 14 to 34 gev long ago and was called fan invariance . in the following we will show explicitly how this scaling behaviour arises in the perturbative qcd calculations ; furthermore we predict how the abovementioned scaling behaviour is violated .
this will lead to the suggestion of other variables in which scaling is restored in a wider kinematical region .
the double differential distribution @xmath119 in @xmath120 and angle @xmath117 is derived from the @xmath121-distribution of particles @xmath122 in a jet of half opening angle @xmath123 by ( see also @xcite ) @xmath124 with @xmath125 for small angles . for larger angles
an appropriate variable is @xmath126 instead of @xmath127 @xcite . for the application at small @xmath22
we take the approximate expression ( [ dmlla ] ) for @xmath128 and obtain ( with @xmath129 ) @xmath130 within the dla this distribution ( the first factor in the first term ) is given by @xmath131 \label{dndla}\ ] ] @xmath132 for small @xmath22 the first term in the brackets from the single gluon bremsstrahlung dominates and is energy independent .
this is the term which corresponds to the flat plateau of radiation in the low @xmath22 limit ; it is also responsible for the fan invariance as follows from the exact relation @xmath133 with @xmath134 defined in terms of the momentum @xmath87 .
the second term gives the energy dependent correction which vanishes in the soft limit @xmath135 ; the @xmath14 energy @xmath15 enters only through the variable @xmath136 with @xmath137 .
the mlla correction involves the exponent @xmath138 and its derivative @xmath139 so it again depends only on @xmath140 and @xmath121 ( for soft particles one can assume @xmath141 ) .
therefore our perturbative results fulfil the scaling law @xmath142 i.e. the dependence on the @xmath14 energy enters only through the variable @xmath121 or @xmath143 . in this way the violation of the fan invariance can be partly absorbed into the scaling variables @xmath121 or @xmath143 .
the analytical predictions have been derived for the soft gluon radiation ( with energies smaller than , say , @xmath144(1 gev ) ) , while the angles are measured with respect to the primary parton axis .
there are two effects which are beyond the perturbative calculations .
first , the direction of the primary parton is not directly known , but may be related to the jet axis or to the initial hadron or photon direction . here
the situation can be improved by measuring the energy - multiplicity correlation @xcite taking the momentum of each particle in the event in turn as jet axis with the particle energy fraction as a weight .
the angular correlation of this type has been studied using monte carlo events with favourable results @xcite .
secondly , we can only perform calculations at the parton level , so in comparison with experiment an assumption on the effect of hadronization has to be made .
this will be described in the next subsection . according to the lphd concept
, we assume that parton and hadron spectra are proportional to each other if momenta are large compared to @xmath17 .
this picture has so far only been applied to observables which are integrated over @xmath113 , while we are now interested in distributions differential in @xmath113 .
since perturbative formulae explicitly contain the cut - off @xmath145 , they can not indeed be directly compared with the data at small @xmath113 , but they have to be modified in order to provide a smooth behaviour in this region . there is no unique prescription to be followed in order to build this bridge between parton and hadron level , and what we are proposing here is a simple phenomenological ansatz which is consistent with kinematical requirements and the scaling behaviour in the soft limit . in analogy to the relation ( [ phrel ] ) applied to the energy spectra in section 2 , we take the transverse momentum cut - off @xmath17 as an effective mass of the hadron with @xmath146 and we then replace @xmath147 in eqs .
( [ dnmlla ] ) and ( [ dndla ] ) on the @xmath148 as well as on the @xmath149 this choice guarantees the proper kinematical limits @xmath150 and @xmath151 for parton and hadron spectra , while they coincide for large @xmath113 .
furthermore we use as before @xmath152 .
so for hadrons one can write ( dropping the label @xmath153 everywhere ) @xmath154 with the function @xmath155 defined by ( [ ffun ] ) at the parton level , whereby the replacements ( [ htop ] ) are applied .
once again , relations such as ( [ htop ] ) are not an essential part of the perturbative lphd picture , rather they should be considered as a plausible extension into the region @xmath156 , consistent with kinematics and scaling properties . combining now ( [ edndpkt ] ) and ( [ fandef ] ) , one obtains quantitative theoretical predictions for the angular distribution of particles .
results for different values of @xmath157 at @xmath15 = 14 gev are shown in fig .
[ figfan ] for @xmath17 = 0.5 gev and @xmath158 .
choosing such parameters , the results describe reasonably well the experimental data obtained by the tasso collaboration@xcite in the region of low @xmath157 , while some deviation is visible for larger values of @xmath157 is largely determined by the position of the maximum in the angular distribution and is larger than the value taken above for the energy spectrum .
a unification is possible in principle by calculating the energy spectrum from the integration over the modified double differential spectrum @xmath159 , but this is not attempted here . ] .
the large values of @xmath160 and of angles correspond , however , to large values of the transverse momentum @xmath113 , where our approximations are not valid any more .
this point is particularly important for the study of `` fan invariance '' , i.e. , the energy independence of the renormalized angular distribution : the distributions presented by the tasso collaboration@xcite are normalized to unity in the full angular region and therefore not directly accessible by our calculations . a better test of the scaling law ( [ fan ] )
can be obtained from the direct study of the double differential distribution @xmath161 in absolute normalization ( i.e. , normalized to the particle multiplicity ) as a function of the transverse momentum @xmath113 at fixed @xmath160 .
[ energy ] shows the theoretical predictions in mlla with the same parameters as in fig .
[ figfan ] at @xmath160 = 0.8 gev at four different values of @xmath14 energy @xmath15 .
a good scaling behaviour is visible in the region of very low transverse momentum , whereas the position of the peak and the tail are found to smoothly depend on the @xmath14 energy .
if the distributions were rescaled to the same area as in @xcite , the overall energy dependence would be rather small . in order to better understand the origin of this energy dependence , it is useful to separately look at the contribution of the different terms in the theoretical formula .
this is shown in fig .
[ approx ] , where the @xmath162 distributions at @xmath160 = 0.8 gev and @xmath15 = 34 gev are shown for different approximations : the dla ( born term only ) , dla ( born plus next order term ) and mlla .
whereas the born term of dla , which is independent of @xmath14 energy , is dominant at small @xmath113 , the other terms become important with increasing @xmath113 and give rise to the energy dependence visible in fig .
[ energy ] .
as the next order term becomes important for larger @xmath113 , we also compare in fig .
[ approx ] with the prediction from the limiting spectrum which corresponds to an all order summation within a high energy approximation ( taking the same @xmath17 , @xmath78 = 0 and normalization ( @xmath60 = 3.5 ) adjusted to match the mlla result ) .
it is remarkable how close are the predictions from the mlla and from the limiting spectrum .
this result suggests that retaining only the first two terms in the dla solution gives indeed a reasonable approximation in the region of low transverse momentum .
the deviation from the scaling behaviour ( [ fan ] ) is due to the weak logarithmic dependence of the correction terms on the primary energy . as a consequence of ( [ edndpkt ] ) with @xmath163 the invariant density ,
after our modification ( [ htop ] ) , still follows the scaling behaviour @xmath164 i.e. , for fixed @xmath113 it depends on particle and @xmath14 energies only through @xmath165 .
this scaling behaviour ( [ neweq ] ) is expected to hold in a wider kinematic region than eq .
( [ fan ] ) .
the nearly energy independent soft end of the particle spectrum has been interpreted above as a consequence of the coherent soft gluon emission together with the lphd .
alternatively , one may argue that at these low momenta of few hundred mev we observe the purely hadronic phenomena not related to the perturbative qcd at all .
it is therefore important to test further the perturbative nature of the soft particle production .
the production rate in the limit ( [ izero ] ) can not be predicted in absolute terms , but we note the dependence of this rate on the colour charge factors @xmath166 and @xmath44 for the jets originating from a primary quark or gluon respectively .
therefore , we have argued in @xcite that the sensitivity of the soft end of the spectrum to the colour of the primarily produced hard partons provides a crucial test of the perturbatively based picture and can teach us about the region of its validity .
the prediction for the ratio of multiplicities in gluon and quark jets @xmath167 has been made long ago @xcite but holds only at asymptotic energies . at existing energies
this ratio is measured to be smaller but not in disagreement with the perturbative qcd analysis ( see , for example @xcite ) .
our present discussion concerns only the soft particles : in this kinematic regime the asymptotic predictions on the spectrum , in particular the scaling properties , are well satisfied and therefore the asymptotic ratio @xmath168 in this limit could be asymptotic too without contradicting the global results . therefore , in order to test this scenario , one would like to compare the soft particle production in @xmath3 and @xmath4 events . whereas there are well measured spectra down to small momenta of about 200 mev from quark jets in and dis processes ( see figs
. 1 and 2 ) the direct production of a @xmath4 final state is more difficult . may become a prospective source of clear information @xcite ] to perform such a comparison nevertheless we have discussed in @xcite two different approaches .
first , there is the comparatative study of two jet events arising from reactions with primary hadrons and photons mediated by quark or gluon exchange processes .
if the exchange is sufficiently hard the two types of processes yield primary colour sources of the desired @xmath3 and @xmath4 types .
we will dicuss below how these limits can be derived explicitely from the hard scatterings in the limit of small momentum transfer .
the phenomenological studies along this line are described in papers @xcite .
secondly , one may consider final states with more than two partons .
in certain collinear limits such final states approach a @xmath3 or @xmath4 type antennae .
a simple example is provided by the process @xmath169 which for the configurations with angle @xmath170 or @xmath171 behaves like a @xmath3 and for @xmath172 behaves like a @xmath4 final state @xcite .
such extreme configurations are in general not realistic .
however , we can study the radiation perpendicular to the event plane ( say , in the cone with not too large opening angle @xmath173 ) , then the particle density at low momenta should vary by the factor @xmath174 when going from the @xmath3 to the @xmath4 type configuration and this variation can be predicted also for the intermediate kinematical range .
similarly , one can study such an associated perpendicular radiation in parton - parton scattering processes with quark or gluon exchange at any angle ( see fig .
9 ) . in the following we will present the expectations for various processes ; the main results for @xmath169 have been given already in @xcite .
it was realized long ago that the overall structure of soft particle angular distribution in multijet events in hard processes is governed by the underlying dynamics of colour at small distances .
an instructive known example is the perturbative explanation @xcite of the particle angular flow in @xmath175 events ( string effect
the multiplicity flows build up a colour portrait of an event which can be used as a natural partonometer mapping the primary interaction short - distance process @xcite . in these studies the drag effect on the particles within the production plane is caused by the primary colour dipoles .
more recently , the azimuthal angular distribution around the jet directions @xcite have been studied .
the partonometry ideas are strongly supported by a wealth of experimental data from @xmath175 annihilation ( see e.g. @xcite ) and the tevatron @xmath176 collider @xcite . in the present application we are interested in the particle production at low momenta @xmath87 of order @xmath17 .
a straightforward calculation is possible for the radiation perpendicular to the production plane of the jets .
otherwise , in multijet events , the cut - off condition @xmath145 is more difficult to implement .
also , in this case @xmath113 appears as argument of @xmath71 and does not change for boosts of partons within the plane . in the collinear limits for the 3-jet events one obtains the radiation transverse to the jet for which the coherence arguments apply in the first place . in the simplest case of a boosted @xmath3 pair , or equivalently , in @xmath177 the radiation pattern of gluons with momentum @xmath87 in direction @xmath178
is given by @xcite @xmath179 where @xmath180 and @xmath181 for @xmath182 and @xmath183 in direction @xmath184 and @xmath185 .
for the radiation perpendicular to the plane we simply obtain @xmath186 with the relative angle @xmath187 between the primary partons .
this yields the two limits @xmath188 for parallel and antiparallel momenta respectively .
as we are interested only in the soft particles and the relative variations between different processes or configurations we restrict ourselves to the single gluon bremsstrahlung of order @xmath71 as in ( [ nqqbg ] ) .
the mlla corrections will become important for larger momenta ( see e.g. the discussion in @xcite ) .
the soft gluon radiation in the @xmath189 process is given by the antenna pattern ( defined as in ( [ nqqbg ] ) ) @xmath190 \label{wqqbg}\ ] ] where @xmath191 refer to @xmath192 and this bremsstrahlung pattern holds , irrespective of whether the partons are incoming or outgoing .
with ( [ whperp ] ) one can obtain the particle density in the direction perpendicular to the production plane .
it is convenient to normalize the production rate of the soft particles to the corresponding rate in the standard process @xmath193 in its rest frame with @xmath194 according to ( [ ijdef ] ) and define the ratio @xmath195 for a general process @xmath87 . for the perpendicular radiation in the 3 jet events one finds @xmath196 .
\label{rperpqqg}\ ] ] in our approximation the momentum spectrum does not depend on the angles @xmath187 between the jets
. it will be interesting to study the spectrum @xmath197 in such a cone experimentally and to find out whether the angular dependence of the particle density in ( [ rperpqqg ] ) holds down to small momenta .
it is also interesting to note the difference of this prediction to the large @xmath44 approximation in which the @xmath189 event is treated as a superposition of two @xmath3 dipoles ( see , e.g. , @xcite ) . in this case
the last term in eq .
( [ rperpqqg ] ) drops out and @xmath198 .
this yields @xmath199 \qquad \hbox{(large $ n_c$ ) } \label{rperpln}\ ] ] predictions from these formulae for @xmath200 are presented in table [ tableperp ] for various relative angles @xmath187 .
note , in particular , the limiting case @xmath200 = 1 for soft or collinear primary gluon emission and the proper @xmath4 limit for the parallel @xmath201 ( @xmath202 = 0 ) configuration , as expected .
the role of the large-@xmath44 limit can be investigated also by studying the production rate in 3-jet events normalized to the sum of rates from the corresponding 2-jet events ( dipoles ) with opening angle @xmath203 and @xmath204 respectively : @xmath205 this ratio measures directly the deviation from the large-@xmath44 limit @xmath206= 1 and thereby from the @xmath3-dipole approximation .
this approximation is not necessarily limited towards soft particle production . for the simple case of mercedes - like events ( @xmath207 ) one obtains @xmath208 .
the 2-jet rates for relative angle @xmath187 which appear in the denominator of eq .
( [ rtilde ] ) could be found experimentally from the corresponding @xmath209 final states .
.prediction for the ratio @xmath210 from ( [ rperpqqg ] ) and its large-@xmath44-approximation ( [ rperpln ] ) for different configurations of the @xmath189 events ( @xmath211 , @xmath212 , @xmath213 ) . [ cols="^,^,^",options="header " , ] let us list a few further results : _ a ) _ a particularly simple situation is met for mercedes - type events where no jet identification is necessary for the above measurements .
_ b ) _ the large angle radiation is independent of the mass of the quark ( for @xmath214 ) . _
c ) _ the above predictions for the ratios @xmath215 are derived for the soft particles according to the bremsstrahlung formula ( [ nqqbg ] ) .
one may also consider the particle flow integrated over momentum , as in the discussion of the string / drag effect . in this case one
has to include all higher order contributions which take into account the fact that the soft gluon is part of a jet generated from a primary parton .
then the angular flow @xmath216 is given by the product of the radiation factor @xmath217 and a cascading factor@xcite .
for the ratio of multiplicity flows one obtains the same predictions ( [ rperpqqg ] ) and ( [ rtilde ] ) as the cascading factors cancel
. it will be interesting to see to what extent the predicted angular dependence for both quantities the multiplicity flows and the soft particle yields are satisfied experimentally .
_ d ) _ the similarity of particle flows in 3-jet and 2-jet events with corresponding angles @xmath203 , @xmath204 , as expected in the large-@xmath44 approximation , should also apply to further details of the final state such as the particle ratios .
since the lorentz transformation along the boost direction produces a larger drag for heavier particles , one may expect that after the boost the @xmath218 and @xmath219 ratios for soft particles decrease and then the same is true for soft particles in 3-jet events in comparison to 2-jet events in their rest frame . next we consider the photoproduction of di - jets .
the soft bremsstrahlung depends only on the momentum vectors of the partons but not on the virtuality of the photon .
our results can be applied to @xmath0 as well as to @xmath1 collisions .
the di - jet photoproduction has been studied recently in detail at hera @xcite . in di - jet production one
can distinguish in the leading order qcd approach the direct and the resolved processes @xcite .
in the first case the photon participates directly in the hard scattering subprocess and transfers a large fraction ( @xmath220 ) of its primary energy to the secondary jets ; in the second case , the hard scattering subprocess involves the partons in the protons and also in the photons and the energy fraction is smaller ( @xmath221 ) .
the data indeed show two peaks in the distribution of @xmath222 corresponding to the two types of processes .
the direct processes are mediated by @xmath223 the so - called photon - gluon - fusion and qcd - compton processes respectively .
their differential cross sections are given by @xmath224 where @xmath225 is the quark charge and @xmath226 with the scattering angle @xmath227 in the di - jet @xmath14 .
the cross section is maximal for small @xmath228 ( or @xmath229 ) where it is dominated by quark exchange
. in the resolved processes the partons ( @xmath230 and @xmath231 ) in the photon interact with the partons of the proton . we restrict ourselves in this discussion to the small angle scattering where the gluon exchange contribution dominates and consider the subprocesses @xmath232 ( in processes @xmath233 and @xmath234 the @xmath182 s
could be replaced by the @xmath183 s ) .
the different exchanges in the direct and resolved processes cause different distributions @xmath235 and @xmath236 respectively and this difference is clearly seen in the data @xcite . therefore in this case , by choosing suitable intervals of @xmath222 , one can select processes of different colour structure and one can check whether or in which kinematical interval the low momentum particles are sensitive to it . to this end
we calculate the respective production rates for the soft particles . for the given total energy the hard process depends on 3 variables which we choose here for simplicity as the incoming parton energy fractions @xmath237 and @xmath238 and the momentum transfer @xmath228 .
the cross section for producing the di - jets with and without accompagnying soft gluons is then given by @xmath239 where @xmath240 denotes the parton structure functions at scale @xmath241 .
we are interested in the density of the soft radiation , i.e. in the ratio @xmath242 where often in practical applications an integration over ranges of the kinematical variables @xmath243 and @xmath228 has to be performed .
if only one subprocess @xmath244 $ ] contributes we come back to relations analogous to ( [ nqqbg ] ) .
+ the angular factors required in ( [ dsigsoft ] ) for the partons @xmath182 , @xmath183 and @xmath231 involved in processes a ) and b ) in ( [ gamgl ] ) and ( [ gamq ] ) are given by ( [ wqqbg ] ) with the proper relabeling of the partons ( see also @xcite ) @xmath245 , \label{na}\\ w^b(\vec n)&= & n_c \left[{(\widehat{34})}+{(\widehat{24})}-\frac{1}{n_c^2}{(\widehat{23})}\right ] . \label{nb}\end{aligned}\ ] ] for the case of perpendicular radiation in the di - jet rest frame we have simply @xmath246 if either of the two direct processes a ) or b ) dominates we obtain for the respective rates , normalized as in ( [ rperpdef ] ) @xmath247.\label{rpb}\end{aligned}\ ] ] this means that the soft radiation in the photon gluon fusion process is the same as in the standard process @xmath193 irrespectively of the scattering angle @xmath248 . in case of the qcd - compton process we find in the limit of small angles @xmath249 which is dominated by quark exchange @xmath250 . in the case of backward scattering with @xmath251 ( i.e. the transition @xmath252 at small angle )
we have effectively a colour octet channel and find @xmath253 as in case of a @xmath4 final state . in the large @xmath44 approximation we can consider the bremsstrahlung in the quark exchange processes as coming from a single colour dipole whereby the second is inactive ( for @xmath254 ) ; in case of octet exchange the radiation comes from two dipoles and adds incoherently .
if the two jets are not identified one has to add the cross sections for scatterings at angles @xmath248 and @xmath255 .
one finds @xmath256+g(y)\left[1+y - x / n_c^2\right]}{g(x)+g(y ) } \label{rbsym}\end{aligned}\ ] ] with @xmath257 , @xmath258 and @xmath259 .
so the latter ratio rises from @xmath260 at @xmath254 to @xmath261 at @xmath262 .
therefore , in the kinematic region where one of the processes dominates one can check the variation of the soft particle density with scattering angle @xmath248 , analogously to the case of @xmath263 . in the more general situation one has to use the superposition ( [ dsigsoft ] ) with the structure functions as weights . in the further discussion
we restrict ourselves to the scattering at small angles where quark or gluon exchanges dominate to simplify the discussion .
then in both processes a ) and b ) the quark exchange with @xmath250 dominates and this result also holds in a general superposition ( [ dsigsoft ] ) of symmetrized direct processes . for small angles @xmath264 the expressions for the radiation patterns
simplify significantly @xcite .
the antenna patterns accompagnying the @xmath265 parton scattering through the t - channel exchange are then given by @xmath266 , \label{nc}\\ w^d(\vec n ) & = & n_c \left[{(\widehat{13})}-\frac{1}{n_c^2}{(\widehat{24})}+\frac{1}{2}\left\ { { ( \widehat{12})}+{(\widehat{34})}+ { ( \widehat{14})}+{(\widehat{23})}\right\}\right ] , \label{nd}\\ w^e(\vec n)&=&w^f(\vec n)\;=\ ; 2c_f [ { ( \widehat{14})}+{(\widehat{23 } ) } \nonumber\\ & \;&+\frac{1}{2n_cc_f}\left\ { 2({(\widehat{12})}+{(\widehat{34})})- { ( \widehat{14})}-{(\widehat{23 } ) } - { ( \widehat{13})}-{(\widehat{24})}\right\ } ] , \label{nef}\\ w^g(\vec n)&= & 2c_f [ { ( \widehat{12})}+{(\widehat{34 } ) } \nonumber\\ & \;&+\frac{1}{2n_cc_f}\left\ { 2({(\widehat{14})}+{(\widehat{23})})- { ( \widehat{12})}-{(\widehat{43 } ) } - { ( \widehat{13})}-{(\widehat{24})}\right\ } ] , \label{neg}\end{aligned}\ ] ] considering the subprocesses @xmath267 separately we obtain the ratios @xmath268,\label{rpd}\\ r_\perp^{e , f}&=&(1+\cos \theta_s)+\frac{1}{n_cc_f},\label{rpe}\\ r_\perp^{g}&=&2+\frac{1}{2n_cc_f}(3\cos \theta_s -1).\label{rpg}\end{aligned}\ ] ] at low scattering angles @xmath269 where the gluon exchange dominates all these ratios @xmath270 are @xmath271 and for the important processes
c ) and d ) they vary only slowly with angle
. therefore also for the superposition ( [ dsigsoft ] ) of all processes the ratio will be close to ( [ rpi ] ) for not too large angles . on the other hand ,
these ratios differ markedly ( by about a factor of 2 ) from @xmath272 for the direct photoproduction . in the large @xmath44 limit and at @xmath273 @xmath274 correspond to the incoherent sum of two @xmath3 antennae - dipoles , while @xmath272 are given by only one @xmath3 antenna . therefore , when comparing the perpendicular radiation pattern in direct and resolved photoproduction of di - jets one expects an increased particle density in the latter case by about a factor of 2
. it will be interesting to find out experimentally whether such a difference occurs already at such low momenta as a few hundred mev .
such a finding would provide an important support for the relevance of perturbative qcd , in particular of the colour coherence , to the soft particle production .
the tests which are interesting from our point of view are very similar to those discussed for photoproduction in the last subsection .
first there are the subprocesses ( [ ggto])-([qqto ] ) of parton parton collisions ( at small angles ) which we discussed in connection with resolved photoproduction . secondly , there are the processes with a final state direct photon ( or , equivalently a weak vector boson w or z ) produced together with a hadronic jet with opposite transverse momenta ( see , e.g. @xcite ) .
these processes correspond to the direct processes in photoproduction .
the relevant subprocesses are @xmath275 and the soft bremsstrahlung is again found from ( [ wqqbg ] ) with appropriate relabeling @xmath276 , \label{nhb}\\ w^{b'}(\vec n)&= & n_c \left[{(\widehat{12})}+{(\widehat{24})}-\frac{1}{n_c^2}{(\widehat{14})}\right ] . \label{nha}\end{aligned}\ ] ] then we obtain for the perpendicular radiation @xmath277\label{rpbp}\end{aligned}\ ] ] just as in case of direct photoproduction .
there is an interesting difference with the case of photoproduction . at high primary energies the parton collisions with initial gluons are more likely .
therefore in photoproduction the photon - gluon fusion process ( a ) without angular dependence in the soft particle production rate is favoured whereas at the hadron collider the other process ( @xmath278 ) with angular dependence is more likely .
this offers the possibility to study the angular dependence of the soft particle production in the same process .
the analytical perturbative approach to multiparticle production together with the assumption of lphd has proven to be very successful in the description of various inclusive characteristics of jets .
it is of importance to investigate further the potential and limitations of this picture , in particular in the soft region where hadronization effects could in principle wash out the perturbative predictions .
the soft particles follow a striking prediction of the perturbative analysis .
the particle density at low momentum is nearly energy independent over two orders of magnitude in the @xmath14 energy of @xmath279 annihilation and this behaviour is supported now also by the data from @xmath280 collisions . more generally ,
the analytic calculations suggest a scaling law for particles at low transverse momentum ; such a property was observed already long ago in the petra energy range .
the scaling law is derived analytically for soft gluons which are emitted coherently from all other more energetic colour sources . in the perturbative description
this process of single gluon emission is dominated by the born term of @xmath281 which is energy independent . at larger @xmath113 gluons
are more frequently produced through cascading processes of higher order in @xmath71 .
remarkably , the hadrons follow this prediction demonstrating the relevance of lphd also for soft particles .
it is very important to further clarify whether the observed scaling behaviour is just a general property of hadronization or should be considered as consequence of the perturbative qcd as in our picture . in the latter case
the density of the low momentum particles should vary in a well defined way with the type of reactions and orientations of the primary partons in the hard process . in this paper
we derive predictions for processes for which data are already available .
experimental studies of the proposed type would clarify at which scale of the low transverse momentum or energy the behaviour expected from the perturbative analysis sets in .
we would like to thank a. de angelis , yu.l .
dokshitzer , d. kant , p. mttig , r. orava , g. thompson , t. sjstrand and e.a .
de wolf for useful discussions . 99 a. bassetto , m. ciafaloni and g. marchesini , phys . rep . , * c100 * ( 1983 ) 201 . yu .
l. dokshitzer , v. a. khoze , a. h.mueller and s. i. troyan , rev .
phys . , * 60 * ( 1988 ) 373 ; basics of perturbative qcd " , ed .
j. tran thanh van , editions frontires , gif - sur - yvette , 1991 . for a recent phenomenological status of the analytical perturbative approach ,
see v.a .
khoze and w. ochs , int .
j. mod .
* a12 * ( 1997 ) 2949 .
l. dokshitzer and s. i. troyan , proc.19th winter school of the lnpi , vol . 1 , p.144 ; leningrad preprint lnpi-922 ( 1984 ) .
a. h. mueller , nucl .
* b213 * ( 1983 ) 85 ; erratum quoted ibid . ,
* b241 * ( 1984 ) 141 .
i. azimov , yu . l. dokshitzer , v. a.khoze and s. i. troyan , z. phys .
, * c27 * ( 1985 ) 65 and * c31 * ( 1986 ) 213 .
, s. aid et al . , nucl .
* b445 * ( 1995 ) 3 .
, c. adloff et al . ,
nucl . phys . *
b504 * ( 1997 ) 3 .
zeus coll .
, m. derrick et al .
, z. phys .
* 67 * ( 1995 ) 93 .
, f. abe et al . , phys .
* d50 * ( 1994 ) 5562 ; + d0 coll .
, b. abbott et al . , hep - ex/9706012 , to be published in phys .
lett . ; + d0 coll .
cullen - vidal , fermilab preprint , fermilab conf-96 - 304-ei ( 1996 ) .
, a. korytov , in proc . of qcd 96 , ( montpellier , france , 1996 ) , eds .
s. narison , nucl .
suppl . ) * 54a * ( 1997 ) 67 .
l. dokshitzer , v. s. fadin and v. a.khoze , phys .
, * 115b * ( 1982 ) 242 ; z. phys . , * c15 * ( 1982 ) 325 .
a. bassetto , m. ciafaloni , g. marchesini and a. h. mueller , nucl .
phys . , * b207 * ( 1982 ) 189 .
s. lupia and w. ochs , phys .
* b365 * ( 1996 ) 339 ; + mpi - pht/97 - 26 , hep - ph/9704319 , april 1997 , z. phys . * c * in press .
v. a. khoze , s. lupia and w. ochs , phys . lett .
* b394 * ( 1997 ) 179 ; + proc .
workshop on multiparticle production , eds .
hwa , w. kittel , w.j .
metzger and d.j .
schotanus , june 1966 , nijmegen , the netherlands , world scientific ( 1997 ) , p. 358
( hep - ph/9610348 ) .
s. lupia , in proc . of qcd 96 , ( montpellier , france , 1996 ) , eds .
s. narison , nucl .
. suppl . ) * 54a * ( 1997 ) 55 ; + w. ochs , in proc .
xxxvi cracow school of theor .
phys . , ( zakopane , poland , 1996 ) , acta phys . pol . * b27 * ( 1996 ) 3505 .
s. lupia and w. ochs , mpi - pht/97 - 46 , hep - ph/9707393 , july 1997 , phys . lett .
* b * in press .
v. a. khoze , in in @xmath97 physics " , cargse 1990 , eds .
m. lvy , j - l .
basdevant , m. jacob , d. speiser , j. weyers and r. gastmans , nato asi , series b : physics vol .
261 , p. 419 .
tasso coll .
, m. althoff et al . , z. phys .
* c22 * ( 1984 ) 307 .
ioffe , v.a .
khoze and l.n .
lipatov , hard processes , vol . 1 , north holland publ . ,
amsterdam ( 1984 ) .
l. dokshitzer , v. a. khoze and s. i.troyan , int .
j. mod .
phys . * a7 * ( 1992 ) 1875 . yu . l. dokshitzer , v. a. khoze and s. i.troyan , z. phys . ,
* c55 * ( 1992 ) 107 .
delphi coll .
, p. abreu et al . , phys . lett .
* b347 * ( 1995 ) 447 .
tasso coll .
, w. braunschweig et al . , z. phys .
* c47 * ( 1990 ) 187 .
opal coll .
, m. z. akrawy et al . , phys .
lett . , * b247 * ( 1990 ) 617 .
mark i coll .
, j. l. siegrist et al . , phys . rev .
* d26 * ( 1982 ) 969 .
topaz coll .
, r. itoh et al . , phys .
b345 * ( 1995 ) 335 .
aleph coll .
, d. buskulic et al . , z. phys .
* c73 * ( 1997 ) 409 argus coll .
, s. behrends et al .
d31 * ( 1985 ) 2161 ; + tpc/2@xmath282 coll .
, h. aihara et al .
* 61 * ( 1988 ) 1263 ; + opal coll .
, r. akers et al .
, z. phys . *
c63 * ( 1994 ) 181 .
aleph coll .
, r. barate et a;. , preprint cern / ppe 96 - 186 , to appear in phys .
mark ii coll .
, j. f. patrick et al .
49 ( 1982 ) 1232 ; + l3 coll . , b. adeva et al . ,
b259 * ( 1991 ) 199 ; + delphi coll . ,
p. abreu et al . , phys
b372 * ( 1996 ) 172 ; + opal coll .
, m. z. akrawy et al.,z
c72 * ( 1996 ) 191 ; + opal coll .
, k. ackerstaff et al .
, preprint cern - ppe/97 - 015 , to appear in z. phys .
c tasso coll .
, r. brandelik et al .
, phys . lett .
* b94 * ( 1980 ) 444 ; + tasso coll . , w. braunschweig et al .
, z. phys . *
c42 * ( 1989 ) 180 .
l. dokshitzer and m. olsson , nucl .
* b396 * ( 1993 ) 137 .
w. ochs and j. wosiek , z. phys . * c68 * ( 1995 ) 269 .
delphi coll .
, p. abreu et al .
, nucl . phys . *
b444 * ( 1995 ) 3 .
l. dokshitzer , v. a. khoze and s. i.troyan , sov .
j. nucl . phys . , * 46 * ( 1987 ) 712 ; * 50 * ( 1989 ) 505 .
s. j. brodsky and j. f. gunion , phys .
rev.lett . ,
* 37 * ( 1976 ) 402 .
opal coll .
, g. alexander et al . , phys . lett .
* b388 * ( 1996 ) 659 .
v. a. khoze , in proceedings of the workshop physics and experiments with @xmath283 linear colliders " , saariselk , finland , 1991 , eds .
r. orava , p.eerola and m. nordberg ( world scientific , singapore ) 1992 , p.547 .
l. dokshitzer , v. a. khoze and s. i.troyan , sov .
j. nucl .
, * 47 * ( 1988 ) 881 ; + j. w. gary , phys . rev .
, * d49 * ( 1994 ) 4503 .
i. azimov , yu . l. dokshitzer , v. a.khoze and s. i. troyan , phys .
, * b165 * ( 1985 ) 147 .
b. andersson , g. gustafson , g. ingelman and t.sjstrand , phys .
, * 97 * ( 1983 ) 31 ; + b. andersson , g. gustafson and t.sjstrand , phys .
lett . , * b94 * ( 1980 ) 211 .
t. sjstrand , computer physics commun . * 82 * ( 1994 ) 74 .
zeus coll .
, m. derrick et al .
, phys . lett .
* b348 * ( 1995 ) 665 ; * b384 * ( 1996 ) 401 .
, s. aid et al .
, z. phys . *
c70 * ( 1996 ) 17 .
butterworth , ucl / hep 97 - 04 , contribution to ringberg workshop new trends in hera physics
, tegernsee , germany , may 1997 ( hep - ex/9707001 ) . r.k .
ellis , g. marchesini and b.r .
webber , nucl .
* 286 * ( 1987 ) 643 ; erratum nucl .
* 294 * ( 1987 ) 1180 .
j. ellis , v.a .
khoze and w.j .
stirling , z. phys .
* c75 * ( 1997 ) 287 .
khoze and w.j .
stirling , z. phys .
* c76 * ( 1997 ) 59 .
andreev , v.a .
khoze and a.i .
lebedev , sov .
lebedev inst . report ( 1989 ) ; n8 , 82 ; v.a .
khoze , a.i .
lebedev and j.a .
vazdik , mod .
lett . , * a9 * ( 1994 ) 1665 .
t. sjstrand and m. van zijl , phys . rev .
* d36 * ( 1987 ) 2019 .
j. m. butterworth , j. r. forshaw , t. sjstrand and j. k. storrow , j. phys .
* g22 * ( 1996 ) 883 .
* fig.1 . a * : charged particle distribution @xmath86 as a function of particle momentum @xmath87 . experimental data at various @xmath14 energies@xcite are compared with predictions using @xmath284 $ ] with @xmath285 computed in mlla from eqs .
( [ dmlla],[duetermini ] ) ( @xmath17 = 0.27 gev , @xmath60 = 0.45 ) , which approach a common limit for @xmath286 0 .
the detailed shape in the dashed region is particularly sensitive to parton level approximations . *
b * : same distribution for charged pions@xcite in comparison with predictions using @xmath287 $ ] with @xmath285 computed from the limiting spectrum ( @xmath17 = 0.155 gev , @xmath60 = 1.125 ) . * fig .
2 . a * : the same observable as in fig .
* 1a * , but only the data collected at @xmath14 energies from 3 up to 58 gev@xcite are shown . *
b * : the same observable as in * a * , but only the data collected at lep at @xmath14 energies @xmath15 = 91 , 133 , 161 gev@xcite are shown . * fig .
3 . a * : charged particle distribution @xmath86 at three values of momentum @xmath87 = 0.2 , 0.7 and 1.2 gev as a function of @xmath14 energy @xmath15 . experimental data from @xmath279 annihilation experiments@xcite and from @xmath280 experiment @xcite in the breit frame . in the latter case
two sets of data are shown , corresponding to different ways of sampling the data ( see @xcite ) ; in both sets data have been multiplied by a factor 2 to obtain a common normalization with @xmath279 data and a selection on the energy flow has been applied @xcite .
the solid line corresponds to formula ( [ phrel1 ] ) and the dashed one to formula ( [ phrells ] ) . * b * : distribution @xmath86 of charged kaons and of protons at fixed values of momentum as a function of @xmath14 energy .
experimental data from @xmath279 experiments@xcite . * fig . 4 * : rapidity density at @xmath288 , @xmath289 , and distribution @xmath290 at momentum @xmath87 = 0.25 gev for charged particles in @xmath279 annihilation as a function of @xmath14 energy @xmath15 .
data from @xcite . * fig . 6 * : theoretical prediction at @xmath15 = 14 gev for the single inclusive particle density @xmath291 as a function of the angle @xmath117 at four different values of @xmath157 = 0.15 , 0.3 , 0.5 , 0.8 gev ( from right to left ) according to mlla and eq .
( [ htop ] ) ( with parameters @xmath17 = 0.5 gev and @xmath78 = 0.05 ) . * fig . 7
* : theoretical predictions as in fig .
* 6 * for @xmath162 as a function of transverse momentum @xmath113 at fixed value of @xmath160 = 0.8 gev at four different values of @xmath14 energy @xmath15 = 14 ( solid ) , 34 ( long - dashed ) , 91 ( short - dashed ) and 172 gev ( dotted line ) . *
fig . 8 * : comparison of different theoretical predictions for the @xmath162 distribution as a function of transverse momentum @xmath113 at fixed value of @xmath160 = 0.8 gev and @xmath15 = 34 gev : dla born term ( long - dashed ) , dla ( short - dashed ) , mlla ( solid ) with @xmath17 = 0.5 gev and @xmath78 = 0.05 and limiting spectrum ( dotted line ) with the same @xmath17 and @xmath78 = 0 ( normalization adjusted by factor @xmath60 = 3.5 ) .
|
the spectrum of partons in a qcd jet becomes independent of the primary energy in the low momentum limit .
this follows within the perturbative qcd from the colour coherence in soft gluon branching .
remarkably , the hadrons follow such behaviour closely , suggesting the parton hadron duality picture to be appropriate also for the low momentum particles .
more generally , this scaling property holds for particles of low transverse and arbitrary longitudinal momentum , which explains an old experimental observation ( `` fan invariance '' ) . further tests of the perturbatively based picture for soft particle production are proposed for three - jet events in annihilation and di - jet production events in @xmath0 , @xmath1 and @xmath2 collisions .
they are based upon the difference in the intensity of the soft radiation from primary @xmath3 and @xmath4 antennae . * perturbative universality in soft particle production * valery a. khoze @xmath5 , sergio lupia @xmath6 , wolfgang ochs @xmath6 @xmath7 _ infn - laboratori nazionali di frascati , p.o .
box 13 , + i-00044 frascati , italy _
+ + @xmath8 _ th division cern + ch-1211 , geneva 23 _
+ + @xmath9 _ max - planck - institut fr physik + ( werner - heisenberg - institut ) + fhringer ring 6 , d-80805 munich , germany _
|
the formation of structures due to the deposition of particles is a topic of growing interest , and has challenged theoretical and experimental researchers in the field of material physics @xcite . from the experimental point of view there is a real possibility of developing thin film devices with important technological applications and , at the same time , the theoretical physicists can apply the tools well known in the realm of statistical equilibrium physics to describe these new nonequilibrium phenomena .
it is well established that some characteristics of the growing surfaces present scale - invariant properties , so that quite different growing processes exhibit very similar scaling behavior , which we believe is universal .
the temporal evolution of a surface formed by the deposition of particles is usually described in terms of some scale exponents .
these scaling exponents define the most fundamental characteristics of the growth processes , so that we can put into the same universality class many different processes that have the same values of the scaling exponents .
in this work we are interested in the description of the morphology of a surface formed by adding particles to an initially flat two - dimensional substrate .
particles having one unit height and linear size @xmath7 land horizontally onto the surface and are not allowed to diffuse . in this way , we calculate the interface width , @xmath8 , a function that determines the roughness of the interface , where @xmath9 is the side of a two - dimensional square substrate and @xmath10 is time variable . in order to calculate the interface width , we determine the vertical height of a given point of the surface relative to the substrate , @xmath11 , where * r * gives the position of the cells on the substrate . the roughness @xmath8 is defined as the mean square fluctuation of the height , @xmath12^{2}\rangle^{1/2}$ ] , where @xmath13 is the average value of the surface height at a given instant of time @xmath10 . for a large number of growth models , family and vicsek
proposed a scaling relation , relating the surface roughness with the linear size of the lattice and time @xcite .
this scaling relation is written as @xmath14 where the scaling function @xmath15 is a constant when @xmath0 is very large and @xmath16 when @xmath17 .
the exponent @xmath6 is related to the saturation of the interface width at long times , @xmath18 is called the dynamic exponent and @xmath19 is the exponent that measures the evolution of the interface width at the initial times of deposition .
they are not independent , and are related by the equation @xmath20 the kinetics of growing interfaces can also be studied by means of stochastic differential equations , which describe the evolution , in space and time , of the points at the growing surface .
these equations capture the essence of the discrete models and serve to put them in the proper universality class .
the most important universality classes are defined by the edwards - wilkinson ( ew ) equation @xcite : @xmath21 and by the kardar - parisi - zhang ( kpz ) equation @xcite : @xmath22 where @xmath23 represents the random fluctuations in the deposition processes , with zero configurational average , and uncorrelated in space and time .
generally , the linear equation ( ew ) represents random deposition models , where correlations between nearest neighbor sites are present , as for instance , models including surface relaxation .
an important generalization of the ew equation is the kpz equation that includes a non - linear term @xmath24 in the ew equation .
it takes into account lateral correlations and it is useful to describe growth processes occuring along the local normal to the surface . it represents a wide variety of processes of surface growth and non - equilibrium interfaces , such as those related to the formation of porous surfaces @xcite , corrosion processes of metallic surfaces @xcite , dissolution of a crystalline solid in a liquid medium @xcite , etc .
there are many studies in the literature concerning the properties of the growth models where calculations have been done analytically , by solving the stochastic differential equations , employing mean - field calculations , or through the extensive use of monte carlo simulations @xcite .
the simplest non - trivial known discrete models are the random deposition model with surface relaxation and the ballistic deposition model , described by the equations ew and kpz , respectively .
discrete atomistic models that are commonly presented in the literature , are in general related to the deposition of identical particles , with the same size as the lattice parameter of the substrate . in some studies ,
different models of particle deposition are combined @xcite , or even different species of particles are deposited at the same time , a feature observed in some systems @xcite .
interfaces generated by the deposition of particles larger than one cell of the underlying substrate have been considered , especially for models describing the growth surface due to polymer chains deposition @xcite .
some other important questions related to the thin - film growth have received some attention recently , such as the lattice geometry and temperature effects @xcite .
recently , we have investigated the surface growth generated by the random deposition of particles of different sizes on a one - dimensional substrate @xcite .
we have shown that the roughness , growth and dynamic exponents of the model are independent of the size of the particles .
the results of our simulations have shown that the roughness evolves in time following three different behaviors .
the roughness at the initial times behaves as in the random deposition model . at intermediate times
, the surface roughness grows slowly , and finally , at long times , it enters into the saturation regime .
the scaling exponents of the model are the same as those predicted by the villain - lai - das sarma equation . in the present study
we investigate the properties of the model in two dimensions .
particles can be deposited independently in the @xmath0 and @xmath1 directions of the substrate .
we also observe three different regimes for the temporal evolution of the interface width .
however , contrary to the findings in one dimension , the growing exponent determined at intermediate times depends strongly on the size of the particle that is considered for deposition . for particles whose size is larger than two units of the lattice parameter of the substrate , this exponent is always larger than the one observed at the initial times of deposition .
while in one dimension a new particle can be accomodated in one of the two ends of an already deposited particle , independently of its size , in two dimensions , the new deposition can be done in the @xmath25 positions around a deposited particle , where @xmath7 is the linear size of the particle .
this dependence on @xmath7 makes the growth in two dimensions to be non - universal .
this paper is organized as follows . in sec .
ii we give a brief description of the model , along with the aggregation rules for the deposition of linear n - mers . in sec .
iii we describe the details involved in the monte carlo simulations , and in sec .
iv we present the results of our simulations for the scaling exponents , showing the anomalous behavior of the growth exponent as a function of the size of the particles at intermediate times . finally , in sec .
v , we present our conclusions .
we consider in this work a simple model that can mimic the growth of thin films due to the deposition of linear n - mers on a flat substrate . in this model
the particles are dropped randomly over a square lattice containing @xmath26 unit cells .
all particles to be deposited are one unit height and the columns where they land are increased by at least one unit , because voids are created during the deposition process . at the beginning of the deposition
, we assume that we have a flat surface , and that particles land always horizontally on the surface .
as we will see next , after some initial steps in the growth , the interface width behaves in an unusual way depending strongly on the size of the particle that is being deposited .
the particles deposited on the substrate are linear n - mers of size ( 1 x 1 x n ) , with @xmath27 .
they are added to the substrate in accordance with the rules of the random deposition model .
a particle is aggregated only if the site for deposition coincides with the midpoint of the particle and there is enough space to accomodate it in the @xmath0 or @xmath1 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 .
as we will see next , during a unit of time , we try to deposit @xmath28 particles , which roughly means the deposition of one layer .
therefore , depending on the size of the particle , many trials of deposition are not successful in the unit of time
. other possible mechanisms of deposition could be used , as for instance , a particle landing from above at a random site and stop when the adsorbate is touched . in this case particles
would not be rejected , and during one unit of time more than one layer is incorporated to the substrate .
we performed monte carlo simulations on square lattices of side @xmath9 , with @xmath9 ranging from @xmath29 to @xmath30 , measured in units of the lattice parameter of the substrate . for @xmath31 , simulations were done only for particles of size @xmath32 and @xmath33 .
the simulations were carried out in @xmath34 dimensions , and the resulting deposit is a porous three - dimensional structure .
we also assumed periodic boundary conditions in both directions of the substrate .
as in our simulations we want that one unit of time corresponds to the formation of a new plane of deposited particles parallel to the substrate , it depends on the size of the particle that is being considered for deposition .
if we are depositing particles of linear size @xmath7 , the unit of time , which is measured in monte carlo steps ( mcs ) , corresponds to @xmath28 trials of deposition over the surface .
according to the growth rules of the model , as established in the last section , depending on the size of the particle to be deposited , during one mcs some particles are not incorporated to the surface , because we do not allow any two particles cross themselves in the same plane .
therefore , a particle that is not added to the surface is reflected , and the trial of deposition is lost .
the algorithm for the random deposition of particles is the following : first a cell on a plane parallel to the substrate is randomly selected , and this cell position is defined as the midpoint of the next particle to be deposited .
then , we also choose randomly , the direction , @xmath0 or @xmath1 , for the deposition .
the particle falls horizontally , and is aggregated only if there is enough space around the cell in the selected direction . otherwise ,
as established in the model description section , the trial fails and the particle is not incorporated to the substrate .
we start the deposition process at time @xmath35 , which corresponds to an initially flat substrate .
we record the interface width @xmath8 , as a function of time for different values of the side @xmath9 of the square lattice and different particle sizes @xmath7 .
we also considered deposition of particles of different sizes with a statistical weight selected from a poisson distribution as we have done in our previous work @xcite . to get reliable results we averaged over a large number of samples .
for the smallest lattice considered in this study @xmath36 samples are sufficient to get a good statistics .
we exhibit in fig . 1 the log - log plot of the interface width as a function of time for the deposition of n - mers , with @xmath37 , on a square substrate of side @xmath38 .
we observe that the time dependence of the roughness can be divided in three different regimes . at the initial times of deposition ,
less than @xmath39 mcs , the growth exponent @xmath2 is close to that of the random deposition of particles where spatial correlation between cells can be neglected . at intermediate times , in the range from @xmath36 to @xmath40 mcs , the growth exponent exhibits a completely different behavior from the one observed in one dimension @xcite .
the correlation between sites becomes important due to the effect of excluded area that a given deposited particle imposes to the others .
however , at long times , more than @xmath41 mcs in fig .
1 , the number of particles that are reflected from the surface stabilizes , and we reach a constant value for the interface width . this peculiar behavior was found in the electrophoretic deposition of polymer chains on a two - dimensional substrate @xcite . on a square lattice of side @xmath38 .
the values of the growth exponents @xmath2 and @xmath3 are indicated in the figure.,scaledwidth=70.0% ] we show in fig .
2 the behavior of the interface width for the deposition on a square substrate of side @xmath42 and different particle sizes . in fig .
2d we are plotting the interface width when we consider the deposition of a mixture of particles , with sizes in the range from @xmath43 to @xmath44 , which are randomly selected from a poisson distribution like one used in our previous work . except for @xmath45 , the interface width saturates as shown in fig . 2 , and we observe the same trends as seen in fig .
1 : at the initial times we have a growth exponent @xmath2 , very close to @xmath46 , typical of uncorrelated growths , a new exponent @xmath3 at intermediate times , which depends on the size of the particle being deposited , and finally , the interface width saturates at long times of deposition . 0.2 cm .
the values of the growth exponents @xmath2 and @xmath3 are indicated in the plots . ( a )
@xmath45 , ( b ) @xmath47 , ( c ) @xmath48 and ( d ) mixture of particles with sizes @xmath27.,title="fig : " ] + .
the values of the growth exponents @xmath2 and @xmath3 are indicated in the plots .
( a ) @xmath45 , ( b ) @xmath47 , ( c ) @xmath48 and ( d ) mixture of particles with sizes @xmath27.,title="fig : " ] + .
the values of the growth exponents @xmath2 and @xmath3 are indicated in the plots .
( a ) @xmath45 , ( b ) @xmath47 , ( c ) @xmath48 and ( d ) mixture of particles with sizes @xmath27.,title="fig : " ] + .
the values of the growth exponents @xmath2 and @xmath3 are indicated in the plots .
( a ) @xmath45 , ( b ) @xmath47 , ( c ) @xmath48 and ( d ) mixture of particles with sizes @xmath27.,title="fig : " ] + now we turn our attention to the dependence of the growth exponent @xmath3 and the roughness exponent @xmath6 on the length of particles . for each particle size
we observe a saturation of the interface width , which depends on the side @xmath9 of the square lattice . at very long times
, we expect that roughness scales with the substrate size as @xmath49 .
the exponent @xmath6 can be estimated from @xmath50 after we extrapolate the effective exponents defined by the equation @xcite @xmath51}{\ln2}.\ ] ] in fig .
3 we plot the exponent @xmath52 as a function of @xmath53 for three different particle sizes .
when we fit the data points , we find , for large values of @xmath9 , a value close to @xmath54 . within the error bars , we assume that this is the value of the exponent @xmath6 for all the particle sizes considered in this study , except for @xmath45 , where it is not defined . on the other hand , we plot in fig .
4 , the exponent @xmath3 , which describes the time evolution of the interface width at intermediate times , as a function of the particle size , and for three different linear sizes of the substrate .
the dependence of @xmath3 on @xmath9 is very weak , while it depends strongly on the size of the particle .
this is evident when the size of particle changes from @xmath32 to @xmath33 .
for @xmath32 the growth exponent @xmath3 is smaller than @xmath2 , while it is larger than @xmath2 for @xmath55 . as to be expected , the coverage by dimers is more effective than the corresponding coverage by linear n - mers larger than three units of the underlying lattice parameter .
for instance , while the jamming coverage for dimers @xcite is @xmath56 , the corresponding value for trimers @xcite is @xmath57 . as a function of @xmath53 for three different particle sizes as indicated in the figure . from the fit we find that @xmath58 for large values of @xmath9.,scaledwidth=70.0% ] as a function of the particle length @xmath7 for three different sides @xmath9 of the square lattice as indicated in the figure.,scaledwidth=70.0% ] we display in table i the data for the deposition of particles with sizes in the range @xmath59 to @xmath44 , as well as for the deposition of a mixture of particles , onto a square substrate of side @xmath42 . at the initial growth stage denoted by the growth exponent @xmath2 , the deposition is clearly driven by an uncorrelated mechanism
however , at intermediate times , we observe a sudden jump in the value of the exponent @xmath3 when the particle size changes from @xmath32 to @xmath33 .
this behavior appears to be due to the increase in the spatial correlations with the size of the particles .
the excluded area formed around each deposited particle increases with its size .
the increase in the spatial correlations is more striking when we consider the sizes @xmath45 and @xmath32 .
when we go from @xmath45 , for which we always have @xmath60 close to @xmath46 at any time , with no spatial correlations , to @xmath32 , the exponent @xmath61 , measured at intermediate times , decreases to a value close to @xmath62 .
we also collected the figures of the saturation value of the interface width for the same set of particle sizes .
as it is almost constant , it corroborates the data of fig .
3 , where the same fact was observed for all lattice sizes considered in this work .
therefore , by assuming that @xmath63 , and taking the values of @xmath3 from table i , we can estimate the values of the dynamical critical exponent @xmath18 through the relation @xmath64 , which are also displayed in table i. as to be expected , the values we find for the deposition of a mixture of particles are intermediary between those determined for the smallest and largest particles .
although we have considered deposition of particles as large as @xmath65 , fig .
4 shows that the rate of increasing the value of the exponent @xmath3 is lower for higher values of @xmath7 .
the same trend was observed in simulations performed in one dimension @xcite .
contrary to one - dimensional results , where the scaling exponents were related to the villain - lai - das sarma equation , in the case of deposition of particles of different sizes in two dimensions , the growth exponents are non - universal .
as far as we know , there is no continuum growth equation that can describe this non - trivial model for deposition of n - mers in two dimensions .
0.2 cm .growth exponents @xmath2 and @xmath3 , saturation value of the interface width , and the dynamic exponent @xmath18 for the deposition of @xmath66 and for a mixture ( last line ) of @xmath66 with sizes in the range from @xmath45 to @xmath65 onto a square lattice of side @xmath42 . [ cols="^,^,^,^,^",options="header " , ]
we have considered in this work a growth model for the deposition of linear n - mers , with @xmath67 , over a square lattice .
the particles are one unit height , and are randomly deposited in two perpendicular directions parallel to the substrate . by employing monte carlo simulations , we calculated the interface width as a function of time .
we have shown that three different regimes emerge from calculations . at the initial times ,
the behavior is typical of an uncorrelated growth , given by the exponent @xmath68 , for whatever particle size @xmath7 . at intermediate times
, the growth exponent is no more universal . while for @xmath32 its value is @xmath69 , it jumps to @xmath70 for @xmath33 , showing that spatial correlations increase with particle size .
this happens because the excluded region around each deposited particle increases with its linear size @xmath7 .
for @xmath71 the rate of increasing the value of @xmath61 is lower for higher values of @xmath7 .
finally , at very long times and for very large lattices , the interface width attains a constant value @xmath72 , which is independent of the length @xmath7 of the deposited particle . despite the same deposition model in one dimension
is described by the villain - lai - das sarma continuum equation , in two dimensions it is clearly non - universal , with growth and dynamical exponents depending on the length of the deposited particles .
|
we present some results of monte carlo simulations for the deposition of particles of different sizes on a two - dimensional substrate .
the particles are linear , height one , and can be deposited randomly only in the two , @xmath0 and @xmath1 directions of the substrate , and occupy an integer number of cells of the lattice .
we show there are three different regimes for the temporal evolution of the interface width . at the initial times
we observe an uncorrelated growth , with an exponent @xmath2 characteristic of the random deposition model . at intermediate times
, the interface width presents an unusual behavior , described by a growing exponent @xmath3 , which depends on the size of the particles added to the substrate . if the linear size of the particle is two we have @xmath4 , otherwise we have @xmath5 , for all other particle sizes . after a long time the growth reaches the saturation regime where the interface width becomes constant and is described by the roughness exponent @xmath6 , which is nearly independent of the size of the particle .
similar results are found in the surface growth due to the electrophoretic deposition of polymer chains .
contrary to one - dimensional results the growth exponents are non - universal .
|
the dynamics of nonlinear oscillators and circuits are very often studied with external force being of the form @xmath0 or @xmath1 .
other forms of forces such as square wave , rectified sine wave , rectangular wave , etc . can be generated and applied to dynamical systems .
the study of the effect of such periodic forces will be helpful to choose a suitable external drive in creating and controlling nonlinear behaviours . in recent years
there are reports on the effect of different forces on certain nonlinear phenomena @xcite .
analysis of features of a particular dynamics with various periodic forces and a detailed comparative study of effects induced by them will be of great use .
it is also important to explore the utility and applicability of analytical methods such as multiple - scale perturbation theory and melnikov method to the systems driven by periodic forces other than @xmath0 and @xmath1 . motivated by the above , in the present paper we wish to study the occurrence of horseshoe chaos in duffing oscillator driven by different periodic forces applying melnikov analytical method .
recently , this method has been applied to study the duffing and duffing - van der pol oscillators with different external perturbations @xcite .
the melnikov method has also been applied to other nonlinear systems . in the present paper , we consider the perturbed duffing double - well oscillator @xmath2\;,\end{aligned}\ ] ] [ eq1 ] where@xmath3 is the damping constant,@xmath4 is the natural frequency,@xmath5 is the stiffness constant which plays the role of nonlinear parameter , @xmath6 is an external periodic force , @xmath7 is a small parameter , @xmath8 0 , @xmath9 0 .
we study the occurrence of horseshoe chaos with different periodic forces such as sine wave , square wave , symmetric saw - tooth wave , asymmetric saw - tooth wave , rectified sine wave and modulus of sine wave . in section 2
, we obtain the melnikov threshold condition for the transverse intersection of homoclinic orbits for the system ( [ eq1 ] ) separately for each of the above periodic forces . in section 3
, we plot the melnikov threshold curve in the ( @xmath10 ) parameter space for all the forces where @xmath11 and @xmath12 are the amplitude and frequency of the external periodic force .
we verify the analytical prediction with the numerically calculated critical values of @xmath11 at which transverse intersections of stable and unstable manifolds of the saddle occur .
the melnikov threshold value is also compared with the onset of asymptotic chaos wherever possible . only for the modulus of sine wave long time transient motion instead of asymptotic chaos near the melnikov threshold curve
we characterize the transient dynamics using the average transient time . in section 4
, we consider the system ( [ eq1 ] ) driven by two sine and modulus of sine forces and analyze the effect of the amplitude and phase of the second force .
starting from horseshoe chaos and asymptotic chaos we show the possibility of suppressing them .
the control regions can be identified by the melnikov method .
finally , we end up with conclusion in section 5 .
the unperturbed part of the system ( [ eq1 ] ) with @xmath7=0 has one saddle point ( @xmath13 ) = ( 0,0 ) , and two center type fixed points , ( @xmath13 ) = ( @xmath14 . the two homoclinic orbits connecting the saddle to itself are given by @xmath15 the melnikov function @xmath16 measures the distance between the stable manifold @xmath17 and unstable manifold @xmath18 of a saddle .
when the two orbits are always separated then the sign of @xmath16 always remains same .
@xmath16 oscillates when the orbits @xmath19 and @xmath20 intersect transversely ( horseshoe dynamics ) .
a zero of @xmath16 corresponds to a tangential intersection .
the occurrence of transverse intersections implies that the poincar map of the system has the so - called horseshoe chaos @xcite . for the duffing eq.([eq1 ] ) , the melnikov function is @xmath21 d\tau . \label{eq3}\ ] ] in the following we calculate the melnikov function for the system ( [ eq1 ] ) with different periodic forces . for the system ( [ eq1 ] ) driven by the force @xmath22 ,
the melnikov integral ( [ eq3 ] ) is worked out as @xmath23 where @xmath24 for the square wave @xmath25 where @xmath26@xmath27 is sign of @xmath28 ,
using its fourier series we obtain @xmath29 where @xmath30 for the symmetric saw - tooth wave of the form @xmath31 the melnikov function is @xmath32 where @xmath33 . \label{eq7b}\end{aligned}\ ] ] for the asymmetric saw - tooth wave @xmath34 we obtain @xmath35 for the rectified sine wave @xmath36 we find @xmath37 where @xmath38 for the modulus of sine wave @xmath39 , we find [ eq11 ] @xmath40
in this section , we compute the melnikov threshold values for horseshoe chaos and compare it with numerical prediction . for the sine wave force
the threshold condition for transverse intersections of stable manifolds @xmath41 and unstable manifolds @xmath42 is @xmath43 for the other forces @xmath16 is a convergent series . in these cases
the threshold values of @xmath11 can be determined numerically . for the rectified sine wave force
@xmath44 s in eq.([eq10b ] ) have to be calculated numerically . throughout our study
we fix the parameters in eq.([eq1 ] ) as @xmath45 , @xmath46 , @xmath47 and @xmath48 .
@xmath49 can be calculated numerically as follows . for the forces other than sine and modulus of sine wave , we can not write the sufficient condition for the existence of simple zeros of @xmath16 .
therefore , we identify the occurrence of homoclinic bifurcations by numerically measuring the time @xmath50 elapsed between two successive zeros of @xmath16 .
@xmath50 is calculated for a range of amplitude of the external force .
the value of @xmath11 at which first time @xmath16 changes sign and thereby giving finite @xmath50 is the melnikov threshold value for homoclinic bifurcation .
figure ( [ fig1 ] ) shows the plot of @xmath49 versus @xmath51 , the number of terms in the summation , eq.([eq5a ] ) , for the square wave force .
@xmath49 converges to a constant value with increase in @xmath51 . for @xmath52 10 ,
the variation in @xmath49 is negligible .
similar result is found for the other forces also . in our numerical calculation of @xmath49
we fix @xmath51 = 50 .
figure ( [ fig2 ] ) shows the plot of the threshold curves for horseshoe chaos in the @xmath53 parameter plane . below the threshold curve
no transverse intersection of stable and unstable manifolds of the saddle occurs and above the threshold curve the transverse intersection of orbits of the saddle occurs .
the threshold curves are nonintersecting .
smooth variation of @xmath49 is found for all the forces considered in our study .
the variation of @xmath49 with @xmath12 is similar for all the periodic forces . among the six forces
, @xmath49 is maximum for the asymmetric saw - tooth wave and is minimum for the square wave .
thus , onset of horseshoe chaos can be either delayed or advanced by an appropriate choice of periodic force .
we verify the analytical prediction by numerically computing the stable and unstable manifolds of the saddle . in fig.([fig3 ] ) we plotted the orbits of the saddle for two values of @xmath11 @xmath54 one for @xmath55 and another for @xmath56 for each of the forces . for clarity only part of the manifolds
are shown . in the left side subplots , for @xmath57 ,
the stable and unstable orbits are well separated . in the right side subplots , for @xmath56 ,
we can clearly notice transverse intersections of orbits at certain places .
the critical values of @xmath49 and @xmath58 ( at which onset of cross - well chaos occurs ) for various forces are given in table 1 .
the numerical result agrees well with the theoretical prediction .
& force & @xmath59 & & @xmath58 + 1 . & sine wave & 0.3565 & & 0.3833 + 2 .
& square wave & 0.2745 & & 0.2910 + 3 . & symmetric saw - tooth wave & 0.4629 & & 0.4792 + 4 . & asymmetric saw - tooth wave & 0.7155 & & 0.7326 + 5 . &
rectified sine wave & 0.5026 & & 0.5192 + 6 .
& modulus of sine wave & 0.3985 & & @xmath60 + for the system ( [ eq1 ] ) driven by modulus of sine wave onset of cross - well chaos ( asymptotic ) is not observed near the melnikov threshold value . however , transient motion followed by a long time periodic motion and hysteresis phenomenon are observed .
different paths are followed when @xmath11 is increased from @xmath61 to @xmath62 and decreased from @xmath62 to @xmath61 .
onset of jumping motion from left - well to right - well is found to occur at @xmath63 .
long time transient motion is found for @xmath11 values just below @xmath64 . for a value of @xmath11
, we calculated the transient time @xmath65 , the time taken to settle on the right - well attractor for a set of 900 initial conditions in the region @xmath66 $ ] , @xmath67 $ ] . then , average transient time @xmath68 is calculated for a range of values of @xmath11 .
figure ( [ fig4 ] ) shows the variation of @xmath68 as a function of @xmath69 .
@xmath68 is found to scale as @xmath70 .
in this section , we consider the problem of controlling of chaos by the addition of second weak periodic forces .
we consider only the two forces , namely , sine wave and modulus of sine wave .
the effect of other forces can also be studied . in eq.([eq1 ] ) we choose @xmath71 where @xmath72 is the phase shift .
recently , zambrano et al @xcite studied suppression of chaos by the phase difference between the two external periodic forces .
wang et al @xcite found the occurrence of strange non - chaotic attractor in the duffing oscillator for a range of values of phase @xmath72 with the force @xmath73 . for the system ( [ eq1 ] ) , the melnikov function is [ eq14 ] @xmath74 where @xmath75 and @xmath76 are given by eq.([eq4b ] ) and @xmath77 [ eq14b ] the choice @xmath78 and @xmath79 is trivial . for @xmath80 a melnikov condition similar to eq.([eq12 ] ) can not be written for arbitrary values of @xmath12 and @xmath81 .
so , we numerically compute the time @xmath50 and identify the parametric regime where @xmath82 . in the absence of the second periodic force
the horseshoe chaos is found to occur for @xmath11=0.359 and @xmath12 = 1 .
we study the possibility of suppression of chaos by the addition of the second periodic force . figure ( [ fig5 ] ) shows the plot of @xmath83 against @xmath81 for @xmath84 and @xmath79 . in this figure @xmath83
is zero ( that is , @xmath50 is infinity ) only at @xmath81 = 0.5 , 1.5 and @xmath85 .
this implies that horseshoe chaos does not occur for these three values of @xmath81 . for other values of @xmath81 ,
both @xmath86 and @xmath87 oscillate and hence @xmath83 are nonzero .
these results are verified numerically .
next , we study the effect of @xmath72 . for @xmath88 and @xmath89 ( which gives @xmath90 ) , the condition for @xmath16 to change sign
is given by @xmath91 the equality sign corresponds to tangential intersections of the orbits of saddle .
we denote @xmath92 and @xmath93 are the two roots of the eq.([eq15 ] ) with equality sign .
figure ( [ fig6 ] ) shows the plot of @xmath94 versus @xmath72 for @xmath95 , and @xmath96 . in the regions @xmath97 , @xmath98 and @xmath99 enclosed by the continuous and dashed curves in fig.([fig6 ] ) , the sign of @xmath16 remains same and transverse intersection of orbits of the saddle does not occur . in the remaining region @xmath16 oscillates and
horseshoe chaos occurs .
the above result is also confirmed by direct numerical simulation .
we note that suppression of horseshoe chaos can be achieved for a range of values of @xmath72 , @xmath81 and @xmath94 .
a striking result is that suppression of asymptotic chaotic motion is also found in the regions @xmath97 , @xmath98 and @xmath99 . in fig.([fig7 ] ) bifurcation diagram and maximal lyapunov exponent ( @xmath100 ) as a function of @xmath94 are reported . for @xmath101 the maximal lyapunov exponent @xmath100
is found to be negative in the interval @xmath102 $ ] while the interval of @xmath103 in which suppression of horseshoe chaos predicted by the melnikov method is @xmath104 $ ] . for @xmath105 , suppression of asymptotic chaos
is found in the interval @xmath106 $ ] .
the melnikov method predicted interval of @xmath103 for suppression of horseshoe chaos is @xmath107 $ ] .
similar results are observed for various values of @xmath72 in the regions @xmath97 , @xmath98 and @xmath99 . for the duffing oscillator driven by two modulus of sine wave @xmath108 where @xmath75 and @xmath109 are given by the eqs.([eq4b ] ) and ( [ eq11 ] ) respectively and @xmath110 is same as @xmath109 except that @xmath12 in @xmath109 is now replaced by @xmath81 .
we numerically compute the time @xmath50 and identify the parametric regions where @xmath111 . figure ( [ fig8 ] )
shows the plot of threshold values of @xmath112 versus @xmath72 for @xmath95 , and @xmath113 . in the regions @xmath97 , @xmath98 and @xmath99 enclosed by the continuous and dashed curves in fig.([fig8 ] ) , the sign of @xmath16 remains same and transverse intersection of orbits of the saddle does not occur . in the remaining region @xmath16 oscillates and
horseshoe chaos occurs .
the above result is also verified by numerical simulation .
for example , fig.([fig9 ] ) shows the part of the stable and unstable orbits in the poincar map for two values of @xmath112 chosen outside and inside the region @xmath98 for @xmath12 = @xmath114 , @xmath115 and @xmath116 .
the threshold values of @xmath112 are @xmath117 = 0.615 and @xmath93 = 0.393 .
when @xmath112 is varied from say @xmath118 horseshoe chaos does not occur in the interval @xmath119 . in fig.([fig9]a )
transverse intersections are seen for @xmath120 . for @xmath121 ( lying in the interval [ @xmath122 , @xmath117 ] ) the stable and unstable orbits
are well separated ( fig.[fig9]b ) .
suppression of horseshoe chaos can be achieved for a range of values of @xmath72 , @xmath81 and @xmath112 .
the bifurcation diagram shows absence of cross - well chaos ( asymptotic ) near the melnikov threshold value .
however , long transient motion followed by a periodic motion and hysteresis phenomenon are observed .
in the present paper we have performed a numerical and analytical studies of homoclinic bifurcation in the duffing oscillator driven by different periodic forces and addition of second periodic forces . applying the melnikov analytical method we obtained the threshold condition for onset of horseshoe chaos , that is , transverse intersections of stable and unstable branches of homoclinic orbits . for the modulus of sine wave instead of cross - well chaos a long time transient motion followed by a period-@xmath124 solution is observed . analytical prediction of horseshoe chaos is found to be in good agreement with numerical simulation for all the forces .
suppression of horseshoe chaos and asymptotic chaos are found for a range of amplitude of the second periodic force and phase shift between the two external forces . with the good agreement obtained between theoretical and numerical predictions
we emphasize that the melnikov analysis can be successfully used to predict the onset of chaos in the presence of weak periodic perturbations and also the effect of such perturbations on regular and chaotic dynamics . in the duffing oscillator driven by each of the different periodic forces considered in our study transverse intersection of left - well homoclinic orbits and transverse intersection of right - well homoclinic orbits
occur at same critical values of , say , the amplitude @xmath11 of the force .
this is because in the absence of driving force and damping the potential is symmetric and the two homoclinic orbits connecting saddle to itself satisfy the relation @xmath125 .
it is of interest to consider the situation where @xmath126 . in this case
, it is possible to have onset horseshoe chaos ( as well as asymptotic chaos ) in the two wells at different threshold values .
this will be analyzed in future .
chacon r , control of homoclinic chaos by weak periodic perturbations ( world scientific , singapore , 2005 ) .
konishi k , generating chaotic behaviours in an oscillator driven by periodic forces , phys .
lett . a 2003 ; 320 ; 200 - 206 .
leung a.y.t , zengrong l , suppressing chaos for some nonlinear oscillators , int .
chaos 2004 ; 14 ; 1455 - 1465 .
ge z.m , leu w.y , anticontrol of chaos of two degrees of freedom loud speaker system and synchronization of different order systems , chaos , solitons and fractals 2004 ; 20 ; 503 - 521 .
lai y.c , liu z , nachman a , zhu l , suppression of jamming in excitable systems by aperiodic stochastic resonance , int . j. bifur .
chaos 2004 ; 14 ; 3519 - 3539 .
gandhimathi v.m , rajasekar s , kurths j , effects of the shape of periodic forces on stochastic resonance , int .
j. bifur . chaos ( 2008 , in press ) .
nana nbendjo b.r , salissou y , woafo p , active control with delay of catastrophic motion and horseshoe chaos in a single - well duffing oscillator , chaos , solitons and fractals 2005 ; 23 ; 809 - 816 .
tang y , yang f , chen g , zhou t , classification of homoclinic tangencies for periodically perturbed systems , chaos , solitons and fractals 2006 ; 28 ; 76 - 89 .
wang r , deng j , jing z , chaos control in duffing system , chaos , solitons and fractals 2006 ; 27 ; 249 - 257 .
balibrea f , chacon r , lopez m.a , reshaping induced order - chaos routes in a damped driven helmholtz oscillator , chaos , solitons and fractals 2005 ; 24 ; 459 - 470 .
wang x , lai y.c , lai c.h , effect of resonant - frequency mismatch in attractors , chaos 2006 ; 16 ; 023127 .
yang j , feng w , jing z , complex dynamics in josephson system with two external forcing terms , chaos , solitons and fractals 2006 ; 30 ; 235 - 256 . jing z , yang z , jiang t , complex dynamics in duffing - van der pol oscillators , chaos , solitons and fractals 2006 ; 27 ; 722 - 747 .
gao h , chen g , global and local control of homoclinic and heteroclinic bifurcations , int .
chaos 2005 ; 15 ; 2411 - 2432 .
guckenheimer j , holmes p , dynamical systems and bifurcations of vector fields ( springer , new york , 1990 ) .
wiggins s , introduction to applied nonlinear dynamical systems and chaos ( springer , new york , 1990 ) .
zambrano s , allaria e , brugioni s , leyva i , meucci r , sanjuan m.a.f , arecchi f.t , numerical and experimental exploration of phase control of chaos , chaos 2006 ; 16 ; 013111 . versus @xmath51 , the number of terms in the summation in eq.([eq5a ] )
, for @xmath127 = 0.5 , @xmath128 = 1 , @xmath129 = 1 and @xmath12 = 1 when the system ( [ eq1 ] ) is driven by the square wave force .
variation in @xmath49 converges to a constant value with increase in @xmath51 . ]
plane for the system ( [ eq1 ] ) driven by the forces ( a ) asymmetric saw - tooth wave , ( b ) rectified sine wave , ( c ) symmetric saw - tooth wave , ( d ) modulus of sine wave , ( e ) sine wave and ( f ) square wave . the values of the parameters in eq.([eq1 ] ) are @xmath127 = 0.5 , @xmath128 = 1 and @xmath129 = 1 . ] ) .
the system is driven by ( a - b ) sine wave , ( c - d ) square wave , ( e - f ) symmetric saw - tooth wave , ( g - h ) asymmetric saw - tooth wave , ( i - j ) rectified sine wave and ( k - l ) modulus of sine wave . the other parameters are @xmath127 = 0.5 , @xmath128 = 1 , @xmath129 = 1 and @xmath12 = 1 . left side subplots are for @xmath55 while the right side subplots are for @xmath130 versus @xmath81 for the system ( [ eq1 ] ) with @xmath71 , @xmath133 , @xmath134 , @xmath45 , @xmath46 , @xmath47 , @xmath48 and @xmath135 .
continuous curve corresponds to positive sign while dashed curve corresponds to negative sign in eq.([eq14 ] ) . ] of eq.([eq15 ] ) with equality sign .
horseshoe dynamics does not occur in the regions @xmath97 , @xmath98 , and @xmath99 enclosed by the curves .
continuous and dashed curves represent @xmath92 and @xmath93 respectively .
the other parameters are @xmath134 , @xmath136 , @xmath137 , @xmath138 and @xmath47 . ]
|
the effect of the shape of six different periodic forces and second periodic forces on the onset of horseshoe chaos are studied both analytically and numerically in a duffing oscillator .
the external periodic forces considered are sine wave , square wave , symmetric saw - tooth wave , asymmetric saw - tooth wave , rectified sine wave , and modulus of sine wave .
an analytical threshold condition for the onset of horseshoe chaos is obtained in the duffing oscillator driven by various periodic forces using the melnikov method .
melnikov threshold curve is drawn in a parameter space .
for all the forces except modulus of sine wave , the onset of cross - well asymptotic chaos is observed just above the melnikov threshold curve for onset of horseshoe chaos . for the modulus of sine wave long time transient motion followed by a periodic attractor
is realized .
the possibility of controlling of horseshoe and asymptotic chaos in the duffing oscillator by an addition of second periodic force is then analyzed .
parametric regimes where suppression of horseshoe chaos occurs are predicted .
analytical prediction is demonstrated through direct numerical simulations .
starting from asymptotic chaos we show the recovery of periodic motion for a range of values of amplitude and phase of the second periodic force .
interestingly , suppression of chaos is found in the parametric regimes where the melnikov function does not change sign . , , , duffing oscillator , melnikov method , horseshoe chaos , asymptotic chaos . 05.45.+b
|
gnu octave is a numerical computing platform based on programming with matrices as a fundamental construct .
several modeling and simulation research work benefit through the reduced time and complexity offered by matrix programming environments .
modeling and simulation applications when developed through exploratory analysis , or initial speculative analysis of parameters , optimization in not the key goal .
however , once a stable model is created , running parametric search over a large search space requires the creation of optimized models . optimizing computational models written in matrix languages , for time and space usage , help to directly speed up the search process by reducing time and resources ( memory , diskspace ) .
there are many ways of optimizing program performance , like tracing , logging , and profiling .
profiling is the only method where the runtime is proportional to program size , unlike the other methods which depend on the execution time of the program . clearly ,
for large scale parametric searches , we can not use tracing or logging for optimization , due to enormous log data that needs post - analysis .
this makes profiling as an efficient tool for program optimization . profiling for matrix programming languages like gnu octave , help identify resource usage , inefficient function usage , program flow , piecemeal function run - times and a complete idea of the running times spent on subroutines as a percentage of the program lifetime .
the landmark work in profiling is the gnu debugger ( gdb ) @xcite , which initiated the idea of a call - graph profiler . since gdb
, many types of profilers have been proposed and built , including dynamic instrumented profilers , static sampling profilers , flat profilers , call graph profilers .
the java virtual machine ( jvm ) provides a complete infrastructure for dynamic program analysis , and encourages custom built profilers for querying and collecting statistics from the jvm @xcite .
this is called the java virtual machine profiling interface ( jvmpi ) , and represents the state - of - the art in profiling . using jvmpi many successful profilers can be built for generic profiling or integrated into the existing applications .
we present a call - graph profiler for matrix based languages , implemented on the freely available ( open source ) gnu octave platform .
with knowledge of octave internals @xcite , we chose gnu octave platform for creating profiler .
our work addresses concerns of optimization of numerical computational models , discrete simulations , and exploratory analysis , through the use of profiling .
important metrics for the profiler are minimal sampling - time , minimal memory and resource usage from profiler , dynamic data collection , meaningful output presentation .
we build a dynamic instrumentation profiler , with dispatch built into the octave interpreter , which allows us to create profilers of increasing complexity from flat profiler till the call - graph profiler .
the terminology in use while describing the profiler statistics are described as follows 1 .
_ total time _ : time taken for the subroutine to run , excluding the runtime of subroutines it calls .
this includes only the computational times , and not the times for functions called during computation .
self time _ : the run - time of a subroutine including its calls to the sub - functions .
this is greater than or equal to the total - time of the subroutine .
3 . _ average time _ : the average of the total times , over all calls to the subroutine . the self time , and total time are reported as averages over the total calls .
4 . _ percentage time _ : the fraction of the program runtime for which the given subroutine has been executing . 5 . _
number of calls _ : total number of times a subroutine was called .
6 . _ cumulative time _ :
this gives the sum of the times of all subroutines , that have lesser self time than the current function .
results are sorted in descending order of self time , and increasing cumulative time . in this paper
, we use the terms function and subroutine interchangeably
. the rest of the paper is organized as follows . in sec
ii , we report the profiling api required to create profilers on gnu octave .
the algorithms used for the flat profiler and the call - graph profiler are discussed in sec iii , iv .
benchmarking results of the profilers are presented in sec v , to illustrate accuracy and confidence on the profiled results
. finally , in sec vi , we review the features and limitations of our profiler framework for octave , and indicate the future work required to integrate the profiler into the main gnu octave project .
most profiling systems work by collecting data of a program at runtime .
profilers collect statistics , and the time of occurrence for each piece of information . to enable collection of profiling events like function invocation , object creation , deletion , function call completion , and such interpreter - system specific data ,
profilers need an event delivery mechanism from the interpreter itself .
this makes building a profiler for gnu octave language as two tasks ; building interpreter api event delivery mechanism to the profiler , and then building profilers to make use of the events .
this separation of concerns , was inspired from the profiler design of languages like python @xcite , and ruby @xcite .
the rest of this section describes the events of interest to our profiler , and api for the event delivery system .
the simplest profiler ( flat profilers , sec iii ) , need information about the occurrence of two events ; the function call , and function return .
since octave is not an object oriented system we do not have object creation , deletion , nor member function invocation and such events can not be reported.in a more complicated profiler ( call - graph profiler , sec iv ) , the same events ( function call , return ) have attributes like the function - caller and function - callee passed to the profiler program .
we do not trap events like variable state changed , entering a program section corresponding to line number in source code , or leaving such a program section , etc .
the implications of these events are discussed in sec vi . the api design of the event - reporting system is based on a global singleton class _ octave_profiler _ , which is instantiated by the octave interpreter .
the class _ octave_profiler _ , has a member variable _ profiler_fcn _ which serves as a event reporter function .
this is set by the profiler program , using a call to _
static bool set_profiler(profiler_function profile);_. when set by a profiler , the _ profiler_fcn _ is invoked from within the octave interpreter whenever execution invokes a function or returns from a function .
the profiler stops profiling , by invoking the _
static bool clear_profiler ( ) ; _ which returns the handle _
profiler_fcn _ to a null , and prevents event delivery .
the event delivery is managed by function + _ static void send_event ( const octave_function fcn , profiler_function_type ftype , profiler_call_state cstate ) ; _ + from the internals of the octave interpreter . in this invocation _
fcn _ is the function in question , and _ cstate _ indicates execution has entered or returned from a function .
send_event _ function abstracts the call sequence _ profile_fcn(fcn , ftype , cstate ) ; _ by keeping it private . this api in short provides a hook for the profilers to start / stop receiving events , and a base class that implements this mechanism .
function profilers are written by deriving from the primitive base class _ profile_base_. the base class provides timing information , and the setup tear - down for the events as discussed above .
the user level profiler interface is through the function _ profile _ , which has usage : * profile [ onoffinfo ] [ graphflat ] * , where the options are indicated in square brackets
. a typical use case would be : 1 .
[ profile_usage ] profile on graph _ % use a call - graph profiler _ 2 .
bch ( ) _ % invoke the script _ 3 .
profile info _
% stop profiling & printout the results_. the _ profile _ is responsible for creating instances of call - graph or flat profilers , and pass on the start , stop , print requests to these profilers using api of the profiler , as 1 .
_ static void start_profiling ( ) _ 2 . _ static void stop_profiling ( ) _ 3 .
_ void print_profile ( ) _ the profilers to extend the base class _ profile_base_. this completes the description of the octave side profiling api , and the user interface for profilers .
flat - file profilers are very simple , and only count the average statistics of the program . the only attribute information saved by flat - profilers are saved in the structure , _ call_elem _ , which has fields + 1 . long int ncalls ; 2 .
double total_time ; 3 .
double self_time ; 4 . std::string key ; for each function that is invoked in the program .
the meaning of each field , is self explanatory and the terms are defined in the introduction .
nowhere are absolute times measured , and programs rely on relative time separation of events .
the timing information is stored in the class objects of _ time_elem_. + 1 .
double delta ; //incremental time from the previous routine that a function is called 2 .
double tick ; //time for running our kids / child functions the relative times of events noted down at each event , are finally added up to obtain the statistics for flat - profiler output .
the class _ profiler_flat _ implements the flat - profiler , by extending _
profile_base_. the algorithm is implemented using a stack data structure .
the essential algorithm is the same for the complex call - graph profiler too .
the profiler dispatch function delivers the function invoke and return events by interfacing with the profiler api , describe in sec ii .
the profiler function for the flat - profilers is given as , + _ static void profile_func(const octave_function * fcn , profiler_function_type ftype , profiler_call_state cstate ) ; _ this function further delivers the events to the particular call - processing routines that handle event call , return separately . the function _
profile_func ( ) _ is modeled as a template pattern , that delegates the events to particular handlers .
the algorithm for the flat - profiler is summarized as , 1 .
when profiler is started , note starting time .
register the profile event handler 3 . on call event
: + push a _
time_elem _ instance set to zero , into time-stack.this time - stack is a false call - stack , as it mirrors the interpreter call - stack , functions are invoked and returned .
4 . on return event
check if hashtable has an instance of record for the given function .
otherwise create a new _ call_elem _ instance for this function and set the name to the function .
2 . increase the number of calls on this record by 1 .
compute the relative time difference between the call and return events ; use the time_elem object on the top of time - stack .
4 . add the total time to the call record s corresponding field .
add the self time to the call record s corresponding field.compute self time by subtracting from total time , the value of tick .
update the record in the hashtable , indexed by the function name as key .
if the call - stack of time , is not empty add the cost of this call , to the parent in the parents , time element tick field .
repeat the steps 3 - 4 , till stopped .
clear the profiling handler , and receive no more events .
once profiling is stopped , prepare to print output .
compute % times .
sort the hashtable entries according to the total - time field of record .
print out according to descending order of total times .
it is important to note the source of this algorithm is obtained from profilers for popular programming languages @xcite , @xcite .
we attribute the idea to the python , and ruby implementations .
call - graph profiler builds the profiling output with the program execution , as a directed graph with arcs .
the arcs point from the caller to the callee , and conveys the time of execution of the callee function .
second order statistics and more than averages can be obtained by sifting through the profiling data , and it becomes much valuable than flat profiling .
there are particular cases where flat profiling information is not helpful ; in general numerical routine execution times depend on the size of the input argument , and the average total time used for routines that are not o(1 ) , skew the profiled data .
call - graph profilers side step such problems by assigning second order statistics , which include the self , average , and total times for each arc of a function call , and profiled function s complexity can be clearly observed without skewing the data . from definition of a call - graph , the parent - child relationships ( caller - callee relations ) from the profiled information are also immediately available .
it is to be noted , that in our implementation not every parent - child relationship is saved , and the data is averaged for each unique caller - callee information , in order to reduce the profiler output to a meaningful subset .
data structures derived from _
call_elem _ , and _
time_elem _ with extra variables , to contain the caller - callee relationship records are used
. the call - graph profiler is implemented in the _ class profile_callgraph _ which as in the flat - profiler derives from the _ profile_base _ class .
the call - graph profiler is itself , so to speak , an incremental improvement over the flat - profiler .
it s profiler event reported , dispatch and logging mechanisms are similar to flat - profiler , and not reiterated here .
much of the algorithm of the call - graph profiler is very - similar to the flat - profiler .
the differences remain ; 1 . function call - event
: when call - stack is empty , all call events are added to the toplevel callee hashtable .
this saves the caller - callee information .
function return - event : the returning function is added as a callee to top - of - stack ( tos).then the caller for this returning function set in the hash - table , and its timing record updated . similarly the callee for the tos function is set as the returning function , and the caller records updated .
3 . printing : the data is printed out as a tree , after sorting according to descending order .
the printing of results follows a tree like pattern , illustrating the call - graph nature of the program execution .
to evaluate the flat and call - graph profilers a test case comprising of a communication system simulation program was evaluated .
the program , and associated files were about 1672 lines of octave code , excluding comments .
this code set is chosen for its availability as much as its similar performance on the flat - profiler and the call - graph profilers , due to the constant input modulation sizes used all over the simulation program .
the profiling is carried on at the toplevel program using the sequence of calls to _ profile _ function mentioned in sec [ profile_usage ] .
the flat profiler , gives an average performance of the functions across the runtime of the program .
the run time of the programs are reported in seconds , while the ms / call indicates milliseconds / call .
it should be noted that measured results are more finely - granular than the ones reported .
results are rounded - off due to formatting constraints . from the results in table 1 ,
we see that cpu hogging function is gf_add which takes about 29% of the program runtime .
this information , along with the self - times and number of calls can be used to arrive at possible optimization candidates .
.flat - profiler results [ cols= " < , < , < , < , < , < , < " , ] the call - graph profiler output is more involved as shown in fig 1 .
the important difference is , the call times , count information are collected as the caller - callee basis , and reported so .
the call - graph profiler in this case helps to identify functions that perform on a o(@xmath0 ) complexity basis .
the voluminous output of the call - graph profiler is reduced to the first few lines for brevity sake , is presented in the figure 1 .
the same benchmark program was run on the call - graph profiler as well .
[ fig_ncalls)a ] there is a significant performance hit due to the profiling . in our design ,
we explicitly compensate for the profiler runtime , and this is not a problem .
the reported overhead times are found after compensation , and for the flat - profiler .
the reported overhead time can only be accounted for , using a free parameter computed before profiling on each profiling session .
this is called the _ bias _ value , as reported in the python profiler @xcite . in our profiler design
we do not include such free parameters .
such an overhead observed can only be attributed to the times that are not computable within the profiler .
our hypothesis attributes the time due to the interpreter s delay in invoking the profiler for each function call and return events .
this agrees well with the observed o(n ) overhead time dependence on the number of function calls . in figure 2 ,
a tight - loop function was profiled with a number of calls , to obtain the overhead information presented in the graph .
the linear trend observed in the overhead time seems to justify the apparent constant overhead time for the interpreter which can not be compensated without computing a constant bias factors . from figure 2 , the bias factor would seem to be the slope of the overhead time , which can be estimated to be around an overhead time of 8.1706x@xmath1 seconds / call .
this is however a free parameter , and dependent on implementation details .
it should be noted that , most profilers suffer from the performance hit due to the profiling overhead . on a more general note , from the benchmark tests we observe the total overhead times to be less than 0.5% of the total program runtime .
our design has optimized the overhead time compared consistently from our initial prototype by compensating for each measurable profiling time .
the call - graph profiler performs with a larger overhead compared to the flat - profiler ; we estimate a rough factor of @xmath2 increase in the overhead time , for the call - graph profiler .
the explanation for this variation we think , to be the memory creation and cleanup associated with the data structures used to build the call - graph .
also a non - trivial i / o times are associated with the call - graph procedures .
[ fig_ncalls_b ] certain features which are not implemented at present are not limitations to the profiler .
these include 1 .
resource profiling for opened - files , network - connections , database handles ; 2 .
arguments passed from the caller - callee function are not traced ; 3 .
event filtering , and selective profiling .
the limitations of the profiler reported below include features that can not be added in the current design . 1 .
memory profiling needs deeper access to the interpreter than the present framework can allow .
2 . line stepping and watch on variables are not possible , and more appropriate for a debugger .
non - local exits are not traced ; this means uncaught exceptions are not profiled , and would end in a aborting of execution .
this is however classified as a bug in the user s octave script program .
we also note that , complete integration of the profiler into the codebase of the gnu octave project requires a different approach to creating the profiler - api
. such an profiler mechanism would work by walking the abstract syntax tree ( ast ) , and passing function call and return events to the profiler . from our experimental work , we see this as feasibility to bring the advantages of the call - graph profiler to octave , in the future .
in this paper we have demonstrated a flat profiler and a call - graph profiler for matrix based programming language like gnu octave . we have reported the profiling overhead , benchmark the performance for both the profiler . further the limitations and possible extensions on this design are enumerated . 16 susan l. graham ,
peter b. kessler , and marshall k. mckusick .
gprof : a call graph execution profiler , proceedings of the sigplan 82 , pp 120126 .
sheng liang , deepa viswanathan , comprehensive profiling support in the java virtual machine , 5th usenix conference coots 99 .
guido van rossum et - al , python ( www.python.org ) profiler code : python - svn / lib / profile.py .
yukihiro matsumoto , ruby ( www.ruby-lang.org ) profiler code : ruby / lib / profiler.rb .
john w. eaton , octave project ( www.gnu.org/software/octave ) , accessed feb,2007 .
muthiah annamalai , hemant kumar , leela velusamy , octave - gtk : a gtk binding for gnu octave , proceedings of octave meeting 2006 , arxiv : cs/0604073v2 [ cs : se ] 28 apr , 2006 .
|
we report the design and implementation of a call - graph profiler for gnu octave , a numerical computing platform .
gnu octave simplifies matrix computation for use in modeling or simulation .
our work provides a call - graph profiler , which is an improvement on the flat profiler .
we elaborate design constraints of building a profiler for numerical computation , and benchmark the profiler by comparing it to the rudimentary timer start - stop ( tic - toc ) measurements , for a similar set of programs .
the profiler code provides clean interfaces to internals of gnu octave , for other ( newer ) profiling tools on gnu octave .
|
this conference takes place in the whitsunday group of islands discovered on whitsunday ( june 3 ) , 1770 , by the english explorer , capt .
james cook .
surprisingly , the stated goal of cook s voyage was astronomical in nature .
as described in richard hough s excellent biography ( 1997 ) of capt .
cook , the story has echoes in today s searches for other worlds .
johannes kepler and edmund halley ( 1716 ) predicted that venus would traverse across the face of the sun in 1761 and 1769 .
these astronomers realized that comparing the timing of the transit from multiple locations would yield a measurement of cosmological importance , the earth - sun distance or astronomical unit ( au ) , then known to no better than a factor of two ( sellers , 2001 ; and _ www.dsellers.demon.co.uk/ venus / ven_ch4.htm)_. just as today s scientists argued that the hubble space telescope was necessary to determine the hubble constant , and thus the scale of the universe , so too did the astronomers of the 18@xmath1 century argue for an expedition to measure the transit of venus and thus to determine the scale of the solar system .
since the 1761 transit was poorly observed due to bad weather , in 1767 the royal society established a committee to `` report on places where it would be advisable to make the observations , the methods to be pursued , and the persons best suited to carry out the work '' of observing the 1769 transit . after a period of deliberation , the royal society , acting much like today s national academy of science , forwarded a proposal to king george iii calling for an expedition to tahiti to observe the transit .
the proposal offered a remarkably modern set of arguments ( hough , 1994 , p. 44 ) : * it highlighted the practical applications of the results , noting that transit measurement would `` contribute greatly to the improvement of astronomy on which navigation so much depends . ''
* it noted with concern potential international competition : `` the french , spaniards , danes and swedes are making the proper dispositions for the observation thereof ... the empress of russia has given directions for having the same observed ... it would cast dishonor [ on the british nation ] should they neglect to have the correct observations made of this important phenomenon . '' * and finally , the proposal included an estimate of the cost of the expedition , exclusive of launch vehicle , that would prove to be an underestimate by a thoroughly modern factor of @xmath2 . ``
the expense would amount to _
44_4,000 , exclusive of the expense of the ship .
the royal society is in no condition to defray the expense . ''
king george approved the project .
the royal navy provided the launch vehicle , the collier _ endeavour _ , and lieutenant james cook .
the royal society provided an observer , charles green , and the gentleman - naturalist joseph banks , who also contributed more than _ 44_10,000 of his funds . on august 26 , 1768 , cook set off with 94 passengers and crew and supplies for 18 months including 604 gallons of rum and 4 tons of beer .
they arrived in tahiti one year later , six weeks before the transit , and set up an observatory at `` point venus . ''
the observations were successfully made on june 3 , 1769 .
the rest of the voyage was not without event as cook undertook to fulfill the second ( and secret ) part of his admiralty orders , namely to search for the mysterious southern continent .
he circumnavigated new zealand s north and south islands and explored the east coast of australia from cape hicks , past sydney and botany bay up to the great barrier reef . on sailing up the whitsunday passage , just a few miles from hamilton island where we now sit ,
cook noted `` this land is diversified by hill and valley , wood and lawn , with a green pleasant experience . ''
a few days later , however , he ran aground on the reef , which he termed an `` insane labyrinth '' , nearly sinking before limping to shore for a month of repairs .
cook returned to england in 1771 with transit data which , when combined with measurements from other sites , led to the determination of the astronomical unit to within 3% of the modern value , a major advance in observational cosmology ( sellers 2001 ) .
two hundred years after cook s voyage , scientists have started to consider the challenge of finding life on planets beyond our own . as summarized in woolf and angel ( 1998 ) and beichman et al .
( 1999 , 2000 , and 2002 ) , modern technology offers a realistic opportunity to address this ancient question . in march , 2000 ,
the terrestrial planet finder ( tpf ) project at jpl selected four university - industry teams to examine a broad range of instrument architectures capable of directly detecting radiation from terrestrial planets orbiting nearby stars , characterizing their surfaces and atmospheres , and searching for signs of life .
over the course of two years the four teams , incorporating more than 115 scientists from 50 institutions worked with more than 20 aerospace and engineering firms . in the first year of study , the contractors and the tpf science working group ( tpf - swg ) examined over 60 different ideas for planet detection .
four main concepts , including a number of variants , were selected for more detailed study . of these concepts ,
two broad architectural classes appear sufficiently realistic to the tpf - swg , to an independent technology review board , and to the tpf project that further technological development is warranted in support of a new start around 2010 . _
the primary conclusion from the effort of the past two years is that with suitable technology investment , starting now , a mission to detect terrestrial planets around nearby stars could be launched within a decade .
_ the detection of earth - like planets will not be easy .
the targets are faint and located close to parent stars that are @xmath31 million ( in the infrared ) to @xmath31 billion times ( in the visible ) brighter than the planets
. however , the detection problem is well defined and can be solved using technologies that can be developed within the next decade .
we have identified two paths to the tpf goal of finding and characterizing planets around 150 stars out to distances of about 15 pc : @xmath4 at visible wavelengths , a large telescope ( a 4x10 m elliptical aperture in one design and an 8x8 m square aperture in another ) equipped with a selection of advanced optics to reject scattered and diffracted starlight ( apodizing pupil masks , coronagraphic stops , and deformable mirrors ) offers the prospect of directly detecting reflected light from earths .
@xmath4 at mid - ir wavelengths , nulling interferometer designs utilizing from three to five 3@xmath54 m telescopes located on either separated spacecraft or a large , 40 m boom can directly detect the thermal radiation emitted by earths .
the tpf - swg established that observations in either the optical / near - infrared or thermal infrared wavelength region would provide important information on the physical characteristics of any detected planets , including credible signposts of life .
in fact , the two wavelengths provide complementary information so that in the long run , both would be desirable .
the choice of wavelength regime for tpf will , in the estimation of the tpf - swg , be driven by the technological readiness of a particular technique .
the tpf science working group ( tpf - swg ) established a design reference program to give broad guidelines for defining architectures for tpf .
the goals for tpf were set out at the december , 2000 , meeting of the tpf - swg : _ primary goal for the terrestrial planet finder ( tpf ) _ : tpf must detect radiation from any earth - like planets located in the habitable zones surrounding 150 solar type ( spectral types f , g , and k ) stars .
tpf must : 1 ) characterize the orbital and physical properties of all detected planets to assess their habitability ; and 2 ) characterize the atmospheres and search for potential biomarkers among the brightest earth - like candidates . _ the broader scientific context _ : our understanding of the properties of terrestrial planets will be scientifically most valuable within a broader framework that includes the properties of all planetary system constituents , e.g. both gas giant and terrestrial planets , and debris disks .
some of this information , such as the properties of debris disks and the masses and orbital properties of gas giant planets , will become available with currently planned space or ground - based facilities .
however , the spectral characterization of most giant planets will require observations with tpf .
tpf s ability to carry out a program of comparative planetology across a range of planetary masses and orbital locations in a large number of new solar systems is by itself an important scientific motivation for the mission
. _ astrophysics with tpf _ :
an observatory with the power to detect an earth orbiting a nearby star will be able to collect important new data on many targets of general astrophysical interest .
architectural studies should address both the range of problems and the fundamental new insights that would be enabled with a particular design .
early tpf - swg discussions made it apparent that observations in either the visible or mid - infrared portions of the spectrum were technically feasible and scientifically important .
a sub - committee of the tpf - swg was established under the leadership of dave des marais to address the wavelength regimes for tpf .
the conclusions of their report ( des marais _ et al .
_ 2002 ) can be summarized briefly as follows : @xmath4 photometry and spectroscopy in either the visible or mid - ir region would give compelling information on the physical properties of planets as well as on the presence and composition of an atmosphere .
@xmath4 the presence of molecular oxygen ( o@xmath6 ) or its photolytic by - product ozone ( o@xmath7 ) are the most robust indicators of photosynthetic life on a planet .
even though h@xmath6o is not a bio - indicator , its presence in liquid form on a planet s surface is considered essential to life and is thus a good signpost of habitability .
@xmath4 species such as h@xmath6o , co , ch@xmath8 , and o@xmath6 may be present in visible - light spectra ( 0.7 to 1.0 @xmath9 m minimum and 0.5 to 1.1 @xmath9 m preferred ) of earth - like planets . an ozone band at 0.3 @xmath9 m and a general rise in albedo due to rayleigh scattering are among the few features in the blue - uv part of the spectrum . the lines of these species can be resolved with spectral resolving powers of @xmath10 .
@xmath4 species such as h@xmath6o , co@xmath6 , ch@xmath8 , and o@xmath7 may be present in mid - infrared spectra of earth - like planets ( 8.5 to 20 @xmath9 m minimum and 7 to 25 @xmath9 m preferred ) .
these lines ( except ch@xmath8 ) can be resolved with spectral resolving powers of @xmath11 .
@xmath4 the influence of clouds , surface properties ( including the presence of photosynthetic pigments such as chlorophyll ) , rotation , etc
. can have profound effects on the photometric and spectroscopic appearance of planets and must be carefully addressed with theoretical studies in the coming years ( e.g. ford , seager , and turner 2001 ) . in conclusion
, the tpf - swg agreed that either wavelength region would provide important information on the nature of detected planets and that the choice between wavelengths should be driven by technical considerations .
after an initial year during which the four study teams investigated more than 60 designs , the teams plus jpl identified four architectural classes ( with a number of variants ) worthy of more intensive study .
high level descriptions of these architectures are given below ; more detailed information is available in the summary of the recent architecture studies ( beichman _ et al .
_ 2002 ) , the final reports from the teams , and , for the separated spacecraft interferometer , the tpf book ( 1999 ) .
two groups ( ball and boeing - svs ) investigated the potential for a visible light coronagraph to satisfy tpf s goals .
while there are differences between the designs , there are major similarities : 1 ) a large optical surface ( 4 @xmath12 10 m for ball , 8@xmath128 m for boeing - svs ) ; 2 ) a highly precise , lightweight primary mirror equipped with actuators for figure control with surface quality of order 1 - 5 nm depending on spatial frequency ; and 3 ) a variety of pupil masks ( square , gaussian , or other ( spergel and kasdin 2001 ) and/or lyot stops ) to suppress diffracted starlight . in the case of the ball designs , a key component was a small deformable mirror with @xmath13 elements capable of correcting residual mid - spatial frequency errors to @xmath14 and stable to @xmath15 . in the ball design , the combination of pupil masks and the deformable mirror reduces the ratio of starlight ( scattered or diffracted)to planet light to approximately unity over an angular extent between @xmath16 and @xmath17 . with these features ,
the ball systems are able to conduct a survey of 150 stars with images taken at 3 epochs for confirmation and orbital determination in less than half a year .
the boeing - svs system , as proposed , takes more time to complete such a survey because without a deformable mirror the ratio of starlight to planet light is about 100 times worse than in the ball design ; addition of a deformable mirror would result in comparable performance for the two telescopes . in under a day per star ,
the ball system could detect ( snr=5 at spectral resolution @xmath18 ) various atmospheric tracers , including o@xmath6 , a critical signpost for the presence of photosynthetic life .
the study teams pointed out that the potential for ancillary science was particularly impressive for the visible systems , since it would be straight - forward to add a complement of traditional astronomical instruments , e.g. uv - optical imagers and spectrographs .
operated on an 8 - 10 m telescope , such instruments would represent a giant advance over the present uv - optical performance of the hubble space telescope .
of particular interest would be the ability to make diffraction limited images at uv - wavelengths with @xmath19 milli - arcsec resolution .
future studies will have to assess whether the specialized requirements of a planet finding , e.g. an off - axis secondary , might compromise the general astrophysics potential of a visible / ultra - violet system .
conversely , nasa will have to weigh whether specialized needs such as high uv throughput requiring special coatings and careful attention to contamination issues might significantly increase the cost of the observatory or compromise its planet - finding performance .
the greatest technical risk for the visible coronagraph is in the development , manufacturing and implementation of a large primary mirror with ultra - low wavefront errors as well as components associated with starlight suppression .
the coronagraphs themselves are functionally simple and although the demands for system performance are challenging , none are thought to be insurmountable .
however , the problem of fabricating and launching a large ( 8 - 10 m ) mirror can not be overemphasized .
the tpf project s independent technology review board noted that there exists no capability to fabricate such a high precision ( 3 - 5 times better than hubble s mirror , 5 - 10 times better than ngst s mirror ) , lightweight optical element for ground or space .
but even if a 8 - 10 m system proves to be too difficult to implement on the tpf timetable , a 2 - 4 m class telescope could demonstrate high dynamic range coronagraphic imaging and carry out an exciting scientific program . such a system could find earths only around the closest dozen stars because of its degraded angular resolution , but it could find and characterize jupiters around many more distant stars . a telescope of this scale might fit into budget of a discovery mission .
lockheed martin and jpl examined two versions of the infrared nulling interferometer : structurally connected and separated spacecraft .
the lockheed martin study concluded that a structurally connected infrared interferometer with four 3.5 m diameter telescopes on a fixed 40 m baseline comes close to achieving tpf s goals .
the system uses four collinear telescopes arranged as two interleaved bracewell nulling interferometers to reject star light adequately so that stellar leakage does not compromise the overall system noise .
the array would be rotated around the line of sight to the star over a 6 - 8 hour period .
the telescopes can be combined in different pairs to achieve the short and long baselines needed to observe distant or nearby stars .
the nulled outputs of the combined pairs are combined again to yield a @xmath20 null or an effective @xmath21 null with phase chopping . , where @xmath22 , with @xmath23 the interferometer baseline and @xmath24 the off - axis angle away from the star .
higher powers of @xmath25 require more complex optical configurations , but yield broader , deeper nulls . ]
the separated spacecraft version of the nulling interferometer was described in the 1999 tpf report ( beichman _ et al .
_ 1999 ; also , see woolf and angel 1998 ) .
it uses a different arrangement of telescopes to produce a deeper , @xmath26 , null that can be tuned to resolve most effectively the habitable zone around each target . because the stellar leakage is reduced in this design , the stability requirements are relaxed relative to the structurally connected interferometer .
the @xmath26 null is , however , less efficient in its use of baseline , requiring roughly 1.5 - 2 times longer baselines than the structurally connected system . providing a 80 - 100 m baseline leads , in turn , to the likely requirement for a more complex separated spacecraft system .
thus , a near - term study must investigate whether a 40 m system can satisfy tpf s goals . if not , then nasa should pursue aggressively the development of a separated spacecraft nulling interferometer .
it should be mentioned that the european space agency ( esa ) has studied a two dimensional , separated spacecraft array of infrared telescopes for its darwin mission .
an industrial study by alcatel found that this version of a planet - finding mission was technically feasible .
the ancillary science possible with an interferometer is likely to be more specialized than for a 8 - 10 m visible telescope equipped with general purpose instruments .
however , the prospect of a telescope with ngst - like sensitivity , but with 10 @xmath12 better angular resolution , imaging the cores of protostars , active galaxies , and high redshift quasars is an exciting one .
the largest area of technical risk for the infrared interferometers is not in the performance of the individual components but in the operation of the various elements as a complete system .
most of the required elements are either under development and making good progress , or are reasonable extensions of technology being developed for missions and ground observatories that will be in place well before tpf needs them .
however , the system complexity of the separated spacecraft system can not be overemphasized .
such a system would demand at least one precursor space mission : a formation flying interferometer , such as the starlight project , to validate the complex control algorithms and beam transport needed for this version of tpf .
this conference offers a happy opportunity for researchers pursuing two quite different techniques for finding extraterrestrial life to come together .
origins program has focused on a search for planets and primitive life , in part because of political considerations that ended nasa s involvement in seti over a decade ago .
we are fortunate that dedicated scientists like jill tarter and frank drake , as well as far - seeing donors such as paul allen , have continued to pursue seti in the context of the vibrant research activities we heard described at this conference . despite the political barriers between nasa s programs and seti , the unity of these efforts can be seen through an examination of the drake equation @xmath27 @xmath28 twentieth century astronomers determined the first two terms , the star formation rate in the galaxy and the fraction of stars that are of solar type .
present day radial velocity studies combined with future transit experiments ( the kepler mission ; borucki et al .
2001 ) and astrometric observations ( the space interferometric mission , sim ) will determine @xmath29 , the fraction of stars with planets , and @xmath30 , the number of earth - like planets in the habitable zone , both statistically and around our nearest neighboring stars .
tpf will characterize the earth - like planets , habitable or not , and provide the first measurements ( or upper limits ) of @xmath31 , the fraction of suitable planets that develop life .
thus , within a generation , we will have well established values for the next three terms in the drake equation .
however , astronomical observations can not determine the final three terms in the equation : the fraction of planets that develop intelligent life , the fraction of intelligent civilizations that can ( or want ) to communicate , and the lifetime of a communicative civilization .
these terms are the realm of the seti .
however , it is the basic frustration of seti that we can not separate these remaining terms without at least one successful contact .
this inability to interpret a negative result makes seti an inherently non - scientific experiment despite its highly technological nature . on the other hand
, seti is an important program of _ exploration _ that must be carried out with our best technology and with great scientific rigor in light of the importance of a positive outcome . despite
the interesting papers presented at this conference , useful information on the last terms in the drake equation will not come from terrestrial analogy .
such attempts remind one of capt .
cook s critics who tried to deduce the existence of a grand southern continent by analogy with the northern hemisphere ( hough 1994 , p. 220,306 ) .
it took cook s explorations to rebut these claims and discover the truth .
similarly , it will take rigorous searches ( or serendipitous discovery ) to determine whether there is another planet with _
intelligent _ life in the universe .
nasa , together with its potential international partners , has begun to address the challenge of looking for habitable planets and seeking signs of life beyond the solar system .
captain cook s voyages remind us that we have asked these questions before and , after much hard work , have been rewarded with new continents to explore . as halley said in his paper predicting the 1761 and 1769 transits of venus ( halley 1716 ; quoted in sellers 2001 ) : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ `` we therefore recommend again and again , to the curious investigators of the stars to whom , when our lives are over , these observations are entrusted , that they , mindful of our advice , apply themselves to the undertaking of these observations vigorously . and for them we desire and pray for all good luck , especially that they be not deprived of this coveted spectacle by the unfortunate obscuration of cloudy heavens , and that the immensities of the celestial spheres , compelled to more precise boundaries , may at last yield to their glory and eternal fame . ''
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ cook s voyage , in the service of science and exploration , resonates with us today as we use transits of extra - solar planets and other techniques to search for new worlds and for life beyond earth .
that cook , banks , and the english society that dispatched them had multiple motivations for making this voyage some noble , some base only highlights the similarities with modern exploration where personal ambition , local politics , and practical applications mix with the search for scientific truth . today
s scientists and policy makers would do well to consider the linkages between science and exploration that future voyages of discovery may enable .
the technologies needed to look for other habitable worlds are within our grasp . we need only the will ( and the funding ) to undertake the search .
the research described in this paper was performed at the jet propulsion laboratory , california institute of technology , under a contract with the national aeronautics and space administration .
this work was supported by the tpf project .
the author acknowledges valuable contributions from tpf science working group and the contractor teams , as well as the dedication and hard work of dan coulter and chris lindensmith at jpl .
the hospitality and dedication of the conference organizers and beauty of the conference site were appreciated by all the participants .
|
over two hundred years ago capt .
james cook sailed up whitsunday passage , just a few miles where we now sit , on a voyage of astronomical observation and discovery that remains an inspiration to us all .
since the prospects of our visiting planets beyond our solar system are slim , we will have to content ourselves with searching for life using remote sensing , not sailing ships .
fortunately , a recently completed nasa study has concluded that a terrestrial planet finder could be launched within a decade to detect terrestrial planets around nearby stars . a visible light coronagraph using an 8 - 10 m telescope , or an infrared nulling interferometer , operated on either a @xmath0 m structure or separated spacecraft , could survey over 150 stars , looking for habitable planets and signs of primitive life .
such a mission , complemented by projects ( kepler and eddington ) that will provide statistical information on the frequency of earth - sized planets in the habitable zone , will determine key terms in the `` drake equation '' that describes the number of intelligent civilizations in the universe .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in epsf
|
the knowledge from first principles of the electronic structure of atoms , molecules and solids is contained in the @xmath0-electron schrdinger equation that , within the born - oppenheimer approximation ( i.e. , at fixed nuclei positions @xmath1 ) , reads ( _ 1, ... ,_n)(_1_1, ... ,_n_n)= e(_1_1, ...
,_n_n ) , [ eq_schro ] with @xmath2 where @xmath3 is the external potential due to the @xmath4 nuclei of charges @xmath5 at positions @xmath6 , and hartree atomic units , @xmath7 ( reduced planck s constant ) @xmath8 ( electron mass ) , @xmath9 ( bohr radius ) , @xmath10 ( electron charge ) , have been used .
since electrons obey the fermi - dirac statistics , @xmath11 must be antisymmetric under particle exchange , @xmath12 , where @xmath13 denotes the three - dimensional electronic position and @xmath14 the spin degree of freedom ( @xmath15 or @xmath16 ) . in what follows
we will be only concerned with the search for the ground - state energy , i.e. , the lowest eigenvalue @xmath17 of eq .
( [ eq_schro ] ) .
the methods that both chemists and physicists have developed to find approximate solutions of eq .
( [ eq_schro ] ) can be roughly divided in two large groups : wave - function methods ( traditional quantum chemistry methods @xcite , quantum monte carlo @xcite ) and density methods ( density functional theory @xcite , and density matrix functional theory @xcite , that is somehow in between the two groups ) .
simplistically , wave - function methods start from an approximation for the antisymmetric , normalized , @xmath0-electron wave - function , @xmath18 , and take advantage of the variational principle , e_0 e_0^approx=__approx_approx| | _ approx . since @xmath18 is a @xmath19 dimensional object , wave - function methods are in general computationally expensive when the number of particles increases .
density functional theory ( dft ) uses the electron density @xmath20 as a basic variable , n()=n__1 ...
_n|(_1,_2_2 , ... , _
n_n)|^2d_2 ... d_n , [ eq_dens ] a much simpler quantity to handle , resulting in a low computational cost that allows to reach system sizes much larger than those accessible to wave - function methods . at given number of electrons @xmath0 ,
the hohenberg and kohn theorem @xcite tells us that the ground - state density @xmath20 of eq .
( [ eq_dens ] ) completely determines ( except for an additive constant ) the external potential @xmath21 of eq .
( [ eq_vne ] ) ( for the sake of simplicity we only consider physical hamiltonians with a nondegenerate ground state ) .
since the kinetic energy operator of eq ( [ eq_t ] ) and the electron - electron interaction of eq . ( [ eq_vee ] ) are the same for all systems , a _ universal _ functional @xmath22 of the density @xmath20 can be defined as @xcite f[n;_ee,]=_n| + _ ee| , [ eq_fmel ] where , to keep the connection with what we will do in the next sections , we have explictly shown the dependence of @xmath22 on the electronic interaction @xmath23 and on the kinetic energy operator @xmath24 . the minimum search in eq .
( [ eq_fmel ] ) is performed over all antisymmetric wave - functions @xmath11 that yield the density @xmath20 ( by definition @xmath20 also gives , by integration , the number of electrons @xmath0 ) .
the universal functional @xmath22 can be also defined as a legendre transform @xcite f[n;_ee , ] = _ v \ { _ @xmath25 denotes a local one - body operator of the form of eq .
( [ eq_vne ] ) with @xmath21 replaced by @xmath26 .
if the exact form of the functional @xmath22 was known , the variational principle would tell us that e_0 = _
n \{f[n;_ee,]+n()v_ne ( ) d}. [ eq_en ] in practice , because we rely on approximations for @xmath22 , the energy estimated by carrying out the minimization in eq .
( [ eq_en ] ) can be lower than the exact @xmath17 .
the kohn - sham @xcite approach to dft introduces another density functional @xmath27 $ ] , t_s[n]f[n;0,]=_v \ { _ |+|-n ( ) v ( ) d } , [ eq_ts ] where , as in the rest of this work , @xmath28 always denotes the wave - function of a spin-@xmath29 fermionic system with zero electron - electron interaction , i.e. , in the majority of cases , a single slater determinant ( the fundamental idea of kohn - sham has been to introduce the fermionic statistic in the construction of @xmath27 $ ] ) . since dealing with noninteracting particles
is computationally simple , kohn and sham @xcite proposed to search for that particular noninteracting system which has the same ground - state density of the physical one .
this defines a model system which is usually called kohn - sham ( ks ) system .
the difference between @xmath30 $ ] and @xmath27 $ ] defines the hartree - exchange - correlation functional @xmath31 $ ] , e_hxc[n]=f[n;_ee,]-t_s[n ] that needs to be approximated .
the functional derivative , @xmath32/\delta n(\rv)=v_{\rm hxc}(\rv)$ ] , determines the one - body ks potential @xmath33 that forces the @xmath0 noninteracting electrons to have the same density of the physical system .
the ks - dft relies thus on the assumption that , given a physical ground - state density @xmath20 , it is possible to find a noninteracting system which has the _ same ground - state density_. from the functional @xmath31 $ ] the classical electrostatic hartree term is usually extracted , e_h[n]=dd , and the remaining , unknown , part is called exchange - correlation energy , e_xc[n]=f[n;_ee,]-t_s[n]-e_h[n ] .
the hartree functional @xmath34 $ ] describes the electrons as if they were classical charge distributions .
it is a simple functional of the density , and yields an important part of the energy of a many - electron system , but is nonzero also for a one - electron density .
this `` self - interaction part '' of @xmath34 $ ] is cancelled by the exact @xmath35 $ ] , but this does not occur for most of the current approximate functionals , which suffer of the so called `` self - interaction error '' .
the success of ks - dft is mostly due to the fact that simple approximations ( local - density approximation and generalized gradient corrections ) for @xmath35 $ ] and its functional derivative provide practical estimates of thermodynamical , structural and spectroscopic properties of atoms , molecules and solids .
however , with the current approximations , ks - dft is still lacking in several aspects , in particular it fails to handle near - degeneracy correlation effects ( rearrangement of electrons within partially filled shells ) and to recover long - range van der waals interaction energies .
the inaccuracy of ks - dft stems from our lack of knowledge of @xmath35 $ ] , and much effort is put nowadays in finding new approximations to this term @xcite .
a trend in the current research is to construct _ implicit _ functionals of the density : in particular , the exchange - correlation functional is divided into exact exchange e_x[n]= |_ee|- e_h[n ] , [ eq_exks ] and the remaining correlation energy @xmath36=e_{\rm xc}[n]-e_x[n]$ ] .
the exact exchange cancels the self - interaction error of @xmath34 $ ] , and is an implicit functional of @xmath20 through the slater determinant @xmath28 .
the corresponding kohn - sham potential must be determined via the optimized effective potential method ( oep ) @xcite , e_0=_v\{_v| + _
ee+_ne |_v+e_c[n__v ] } , [ eq_oepks ] where @xmath37 is the ground state of the noninteracting hamiltonian , @xmath38 .
the construction of a correlation energy functional @xmath36 $ ] to be used with exact exchange in eq .
( [ eq_oepks ] ) is still an open problem . in this work ,
we review some basic ideas , results , and open questions of a different approach : instead of trying to approximate the ks @xmath35 $ ] , we change the model system defined by eq .
( [ eq_ts ] ) . before discussing the choice of different model hamiltonians , we report some equations that will be used in the next sections . an exact formula for the functional @xmath31 $ ] can be obtained via the adiabatic connection formalism @xcite : by varying a real parameter @xmath39 , the interaction @xmath40 between the electrons ( we have defined @xmath41 to denote pairwise interactions that only depend on the electron - electron distance ) is switched on continuously from zero to @xmath42 , while the density is kept fixed by an external one - body potential @xmath43 .
each hamiltoninan @xmath44 along this adiabatic connection has a ground - state wavefunction @xmath45 that yields , by construction , the same density @xmath20 for each @xmath39 . if @xmath46 and @xmath47 , the ks hartree - exchange - correlation energy is given by @xcite e_hxc[n]=_0^a d_0^dr_124r_12 ^ 2 f^(r_12 ) , [ eq_adiaks ] where the spherically and system - averaged pair density ( apd ) @xmath48 is obtained by integrating @xmath49 over all variables but @xmath50 , f^(r_12 ) = _ _ 1 ... _n @xmath51 .
equation ( [ eq_ts ] ) shows that the ks approach to dft introduces a model hamiltonian in which @xmath23 is set equal to zero and the one - body potential , @xmath52 , is different from @xmath21 ( and is obtained by imposing the condition that the ground - state electron density of the model hamiltonian be the same of the physical system ) . solving the noninteracting ks hamiltonian instead of the fully interacting hamiltonian of eq .
( [ eq_schro ] ) is obviously very practical .
the exact ground - state energy of the physical system can , in principle , be obtained from the ks slater determinant @xmath28 via the functional @xmath35 $ ] . however , there are cases in which the restriction that the model system be noninteracting makes the search for an approximate @xmath35 $ ] seem like a daunting task .
consider the simple example of the h@xmath53 molecule ( two electrons in the field of two nuclei of unitary charge spaced by a distance @xmath54 ) . in fig .
[ fig_h2 ] we report the quantity @xmath55 that enters in eq .
( [ eq_adiaks ] ) for the ks system and for the physical system ( @xmath56 and @xmath57 , respectively ) at two internuclear distances @xmath54 . as shown by eq .
( [ eq_adiaks ] ) , the change in the apd @xmath58 when we switch from the ks system to the physical ( fully interacting ) system determines the functional @xmath31 $ ] .
if this change is not drastic , we can expect that universal approximations for @xmath31 $ ] can work relatively well .
this is , e.g. , the case of the h@xmath53 molecule at the equilibrium distance @xmath59 a.u . , reported in fig .
[ fig_h2 ] .
but when we stretch the molecule ( e.g. , in the extreme case @xmath60 considered in the same figure ) , we see that the two apd are completely different , which means that the term @xmath31 $ ] becomes very important : it has to correct the very different nature of the ks wavefunction with respect to the physical one .
this effect is completely system - dependent and it is thus very difficult to include in a _
universal _ functional of the density . of eq .
( [ eq_intra ] ) for the h@xmath53 molecule in the ks system ( @xmath56 ) and in the physical system ( coulombic interaction ) .
the internuclear equilibrium distance @xmath59 and the extreme stretched molecule at @xmath60 are considered.,title="fig:",width=253 ] of eq .
( [ eq_intra ] ) for the h@xmath53 molecule in the ks system ( @xmath56 ) and in the physical system ( coulombic interaction ) .
the internuclear equilibrium distance @xmath59 and the extreme stretched molecule at @xmath60 are considered.,title="fig:",width=253 ] cases like this appear when we have near - degenerate levels in the kohn - sham system ( in the case of the h@xmath53 molecule described above , the energies of the two ks states @xmath61 and @xmath62 become closer and closer as @xmath54 increases ) .
the basic idea reviewed in this paper is to remove the constraint that the model system be noninteracting in order to keep a reasonable resemblance between the model wavefuntion and the real one .
obviously , the model system must still be less expensive to solve than the physical one .
a possible approach is to define a density functional for a `` partial '' electron - electron interaction @xmath63 , smaller than the coulomb interaction @xmath64 , @xmath65 , but different from zero ( we restrict our choice to pairwise interactions that only depend on the electron - electron distance ) . following the idea of the adiabatic connection of sec .
[ sec_adiabatic ] , we can make our partial interaction depend on a real parameter @xmath66 in such a way that @xmath67 when @xmath68 and @xmath69 when @xmath66 tends to some positive value @xmath70 , and define f^f[n;^ , ] = _ v \ { _ functional @xmath71 $ ] switches from the kohn - sham one of eq .
( [ eq_ts ] ) to the physical one of eq .
( [ eq_flieb ] ) , @xmath72 & = & f[n;0,\hat{t}]=t_s[n ] , \\ f^{\mu\to a}[n ] & = & f[n;\hat{w}_{ee},\hat{t } ] . \end{aligned}\ ] ] in analogy with the ks approach , we can ask that the model system with interaction @xmath73 have the same density of the physical one .
this fixes the external one - body potential ( that we call @xmath74 ) in our model hamiltonian @xmath75 .
the ground - state wavefunction of our model system is denoted @xmath76 : it is a multideterminantal wavefunction that must be computed with one of the standard methods of quantum chemistry ( configuration interaction , coupled cluster , multi - configuration self - consitent field , ... ) . in general ,
if @xmath66 is not too large ( i.e. , if @xmath77 is still much smaller than the full coulomb interaction ) , few determinants describe @xmath76 quite accurately , so that the computational cost can be kept low .
the difference @xmath78 $ ] between the physical functional @xmath30 $ ] and the partially - interacting @xmath71 $ ] , ^=f[n;_ee,]-f[n;^ , ] , [ eq_fcomp ] is what we need to approximate , together with its functional derivative that determines @xmath79 . how to choose @xmath73 ?
two points are important to determine the partial interaction : ( i ) that we can design reasonable approximations for the corresponding @xmath78 $ ] , and ( ii ) that we can solve adequately the hamiltonian @xmath75 of the model system .
a convenient choice seems to be a long - ranged interaction , i.e. , a @xmath73 that behaves as @xmath42 for large @xmath80 , but that is softer than @xmath42 for small @xmath80 .
the reasons for this choice are ( i ) short - range correlation effects seem to be more transferable from one system to another @xcite , and they should thus be more easily described in terms of an approximate _ universal _ density functional ; ( ii ) the traditional wavefunction methods of quantum chemistry can reasonably describe @xmath76 at a lower cost than the fully interacting system , because of the smaller interaction and because of the absence of the electron - electron cusp . for practical reasons ( analytic matrix elements for both gaussians and plane - waves , i.e. the most commonly used basis sets in quantum chemistry and solid - state physics , respectively ) , a common choice is w^(r_12)= , [ eq_erf ] where @xmath81 is the error function ( and we thus have @xmath82 ) .
other possibilities for @xmath73 have been also explored : they are all , like eq .
( [ eq_erf ] ) , arbitrary , and an open question is wether there is a way to determine @xmath73 according to some optimal criteria .
the approach resulting from the choice of eq .
( [ eq_erf ] ) is the one that we describe more in details in the next sec .
[ sec_lrsr ] . before doing that
, for completness we also briefly introduce a different choice of the model hamiltonian . another way to modify
the model system is by adding a nonlocal one - body operator , @xmath83 , multiplied by a real , positive constant @xmath84 @xcite , f^g[n ] f[n;+g_nl,_ee ] .
[ eq_nonlo ] by setting @xmath83 equal to the projector ( here reported for closed - shell systems ) onto the virtual kohn - sham orbitals @xmath85 , _ nl(,)=(-)-_i=1^n/2_i ( ) _ i^*( ) , the functional @xmath86 $ ] of eq . ( [ eq_nonlo ] )
switches from the physical one at @xmath87 to the kohn - sham one of eq .
( [ eq_ts ] ) in the limit @xmath88 ( since @xmath89 is always @xmath90 , as @xmath88 the model system minimizes its energy by making @xmath91 , that is , by occupying only the ks orbitals ) .
this choice of the model system is not further detailed here ; the interested reader may find additional information in the literature @xcite .
following the same steps of sec .
[ sec_adiabatic ] , we can write an exact formula for the functional @xmath78 $ ] of eq . ( [ eq_fcomp ] ) in terms of the apd @xmath92 .
it is convenient to use for the adiabatic connection the same partial interaction @xmath73 of eq .
( [ eq_erf ] ) chosen to determine the model system , ^=_^ d _ 0^dr_124r_12 ^ 2 f^(r_12 ) .
[ eq_ffromac ] this formula is identical to eq .
( [ eq_fcomp ] ) , except from the fact that the integration over the coupling constant starts from the positive value @xmath66 instead than zero . as in ks - dft ,
the functional @xmath78 $ ] can be divided into an hartree term , _
h^= d d n()n( ) , [ eq_ehcomp ] and an exchange - correlation term _
xc^=^-_h^that needs to be approximated .
the functional @xmath93 $ ] can , in turn , be divided into exchange and correlation in two different ways .
we can , in fact , define an exchange functional by using the kohn - sham determinant @xmath28 , _
x^= |_ee-^|- _ h^ , [ eq_excomp ] and then define the usual correlation energy functional @xmath94 $ ] , ^_c[n]=^_xc[n]-_x^ , [ eq_eccomp ] but we can also define a multideterminantal ( md ) exchange functional @xcite by using the wavefunction @xmath76 , _ x , md^= ^|_ee-^|^- _ h^ , [ eq_exmd ] and then a corresponding correlation energy , ^_c , md[n]=^_xc[n]-_x , md^. [ eq_ecmd ] these two ways of splitting exchange and correlation play a role only if we implement an optimized effective potential - like scheme , in analogy with what is usually done for the exact - exchange ks - dft @xcite [ see eq .
( [ eq_oepks ] ) ] . in the case of the multideterminantal exchange of eq .
( [ eq_exmd ] ) the problem can be reformulated as @xcite e_0=_v^\{^_v^| + _
ee+_ne |^_v^+^_c , md[n_^_v^ ] } , where @xmath95 is obtained by solving the schrdinger equation corresponding to the hamiltonian @xmath96 .
this equation is the generalization of eq .
( [ eq_oepks ] ) to the case of the multideterminantal model system @xmath76 . from eq .
( [ eq_ffromac ] ) it is possible to derive exact properties of the functionals of eqs .
( [ eq_excomp])([eq_ecmd ] ) in the @xmath68 limit @xcite and in the @xmath97 limit @xcite . in the first case ,
all the functionals @xmath98 , @xmath99 [ eqs .
( [ eq_excomp ] ) and ( [ eq_eccomp ] ) ] and @xmath100 , @xmath101 [ eqs .
( [ eq_exmd ] ) and ( [ eq_ecmd ] ) ] tend to the ks functionals of sec .
[ sec_ksdft ] .
this limit can be studied with perturbation theory : one finds that the way in which the functionals approach the ks ones depends on wether the system is confined @xcite ( atoms , molecules ) or extended @xcite .
the large-@xmath66 limit is the most interesting , since , as shown by eq .
( [ eq_ffromac ] ) , it always lies in the range of @xmath66-values for which we want to construct approximations . in this limit ,
all the functionals of eqs .
( [ eq_excomp])([eq_ecmd ] ) vanish and , since @xmath102 , eq .
( [ eq_ffromac ] ) shows that their large-@xmath66 behavior is determined by the short - range part ( small @xmath80 ) of the spherically and system - averaged pair density @xmath92 .
it is possible to show @xcite that , when @xmath97 , the small-@xmath80 part of @xmath92 is dominated by f^(r_12)=f(0 ) , [ eq_expaf ] where @xmath103 is the `` on - top value '' ( zero electron - electron distance ) of the physical ( i.e. , corresponding to the coulomb interaction ) @xmath104 , and the function @xmath105 is equal to @xcite p_1(y ) = + ( + ) ( y ) .
[ eq_p1 ] inserting eq .
( [ eq_expaf ] ) ( and , for the case of @xmath106 $ ] , the ks @xmath107 , corresponding to the ks determinant @xmath28 ) into eq .
( [ eq_ffromac ] ) we find , for unpolarized systems @xcite @xmath108 = -\frac{\pi}{4\mu^2}\int d\rv n(\rv)^2 + o\left(\frac{1}{\mu^4}\right ) , \label{eq_largemuex } \\ & & \overline{e}_c^{\mu\to\infty}[n ] = \frac{\pi}{\mu^2}\left[f(0)-\frac{1}{4}\int d\rv n(\rv)^2\right ] + f(0)\frac{4\sqrt{2\pi}}{3\ , \mu^3}+o\left(\frac{1}{\mu^4}\right ) , \label{eq_largemuec } \\ & & \overline{e}_{\rm x , md}^{\mu\to\infty}[n ] = \frac{\pi}{\mu^2}\left[f(0)-\frac{1}{2}\int d\rv n(\rv)^2\right ] + f(0)\frac{4\sqrt{\pi}(2\sqrt{2}-1)}{3\ , \mu^3}+o\left(\frac{1}{\mu^4}\right ) , \label{eq_largemuexmd } \\ & & \overline{e}_{\rm c , md}^{\mu\to\infty}[n ] = -f(0)\frac{4\sqrt{\pi}(\sqrt{2}-1)}{3\ , \mu^3}+o\left(\frac{1}{\mu^4}\right ) .
\label{eq_largemuecmd}\end{aligned}\ ] ] these equations tell us that ( i ) if we use the definition of exchange with the ks determinant , the functional @xmath109 $ ] is , for large @xmath66 , exactly described by a local functional of the density ; ( ii ) all the other three functionals involve the physical `` on - top '' @xmath103 , which is usually not available ( its knowledge would require a very accurate calculation for the physical system ! ) . however , it is possible to construct approximations for @xmath103 .
for instance , for many systems the estimate of @xmath103 from the local density approximation ( i.e. , by transfer from the uniform electron gas model ) is rather good @xcite .
another possibility is to use eq .
( [ eq_expaf ] ) to estimate @xmath103 from @xmath110 .
this estimate has the advantage of being without self - interaction error @xcite .
following the same historical path of ks - dft , the simplest approximation one can think of is the local density ( lda ) , _
xc^=dn ( ) [ _ xc(n())- _ xc^(n ( ) ) ] , where @xmath111 and @xmath112 are the exchange - correlation energy per particle of an electron gas of uniform density @xmath113 with interaction @xmath42 @xcite and @xmath73 @xcite , respectively .
generalized - gradient approximations ( gga ) for @xmath114 have been also constructed @xcite along similar lines of ks - dft @xcite . in this context
, we only mention a conceptually simple approximation based on the exact properties of the previous sec .
[ sec_prop ] .
it consists @xcite in a rational interpolation between a given gga ks functional at @xmath115 , and zero at @xmath97 .
the rational interpolation is constrained to recover the exact large-@xmath66 behavior of eqs .
( [ eq_largemuex])-([eq_largemuec ] ) , using the lda approximation for the physical on - top @xmath103 .
all these ways of constructing approximate functionals are based on the assumption of transferability of short - range exchange and correlation effects from one system ( the uniform electron gas ) to the others .
another strategy that came out recently , and that seems very well suited to describe short - range correlation effects , consists in generating realistic apd @xmath116 along the adiabatic connection [ with @xmath117 , to be inserted in eq .
( [ eq_ffromac ] ) ] by solving simple `` radial '' ( unidimensional ) equations @xcite for a set of `` effective geminals '' @xmath118 , @xmath119 \psi^{\mu'}_i(r_{12 } ) = \epsilon_i^{\mu'}\ , \psi_i^{\mu'}(r_{12 } ) \nonumber \\ & & \sum_i \vartheta_i|\psi_i^{\mu'}(r_{12})|^2 = f^{\mu'}(r_{12 } ) .
\label{eq_eff}\end{aligned}\ ] ] the effective interaction @xmath120 is ( in principle ) the lagrange multiplier for the exact @xmath116 ( i.e. , corresponding to the wavefunction @xmath121 along the adiabatic connection ; in practice @xmath120 is approximated ) . in these equations
the rule for the occupancy @xmath122 of the effective geminals is chosen , for spin compensated systems , to be slater - determinant - like : occupancy 1 for even @xmath123 ( singlet symmetry ) , occupancy 3 for odd @xmath123 ( triplet symmetry ) , up to @xmath124 occupied geminals .
this rule has been applied to solve the effective equations ( [ eq_eff ] ) in the uniform electron gas at @xmath125 ( coulombic interaction ) , with rather accurate results @xcite when combined with simple approximations for the effective interaction potential @xmath126 . as said in the previous paragraphs , the model wavefunction @xmath76 is , in most cases , computed with one of the traditional wavefunction methods of quantum chemistry .
we report here an illustrative example of such calculation , and we then give in the next sec .
[ sec_results ] an overview of recent results obtained using different methods for computing @xmath76 , combined with different approximations for the functionals .
since the only purpose of this paragraph is to investigate the wavefunction part of the multideterminantal dft , we only consider two very small systems , the he and the be atoms , for which it has been possible to construct very accurate potentials @xmath127 @xcite . in this way ( i )
we have essentially no approximation on the functional part and we can focus our attention on the effect of approximations on the calculation of @xmath76 , and ( ii ) we can produce accurate benchmark results to test our calculations .
of course , this is not the general procedure in which we are finally interested .
the general procedure rather consists in using a given approximation ( see sec . [ sec_approximations ] ) for the functional @xmath128 $ ] and the corresponding potential , and combining it with a wavefunction method to calculate @xmath76 , to finally obtain the total ground state energy of the physical hamiltonian .
this procedure has been followed with very promising results in the works reviewed in the next sec .
[ sec_results ] . here , in order to illustrate the efficiency of approximate wavefunction methods to treat @xmath76 , we thus proceed as follows @xcite .
we first construct , for each @xmath66 , the model hamiltonian @xmath129 using an accurate potential @xmath130 @xcite and compute accurately its ground - state energy , @xmath131 , at the multi - reference configuration interaction with singles and doubles ( mrcisd ) level .
we then compute various approximate ground - state energies , @xmath132 , by using approximate configuration inteaction ( ci ) type wave functions @xmath133 expanded into linear combinations of all the few slater determinants generated from small orbital spaces @xmath134 .
the orbitals used are the natural orbitals of the coulombic system calculated at the mrcisd level .
the accuracy of the approximation for @xmath133 can be assessed by looking at the difference between @xmath135 and @xmath136 @xmath137 the differences @xmath138 are plotted as a function of @xmath66 in fig .
[ fig_ci_he ] for the he atom with the orbital spaces @xmath139 , @xmath140 and @xmath141 .
one sees that , in the coulombic limit , @xmath142 , the reduction of the orbital space leads to important errors in the energy . when @xmath66 is decreased , i.e. when the interaction is reduced , the errors due to limited orbital spaces get smaller and smaller . for instance , at @xmath143 , using only the single - determinant wave function @xmath144 , leads to an error @xmath145 of less than @xmath146 hartree .
the case of the be atom with the orbital spaces @xmath140 and @xmath141 is reported in fig .
[ fig_ci_be ] . because of the near - degeneracy of the @xmath147 and @xmath148 levels , the inclusion of @xmath148 configurations in the wave function is important , quite independently of the electron - electron interaction .
indeed , the difference @xmath149 remains large for almost all @xmath66 s . on the contrary , the error of the calculation where the @xmath148 orbitals are included , @xmath150 , quickly falls off when @xmath66 is decreased .
again , for @xmath143 for instance , the error @xmath150 given by the few - determinant ci - type wave function @xmath151 is less than @xmath146 hartree .
therefore , the modification of the interaction enables to increase the accuracy of ci - type wave function calculations , or equivalently for a fixed target accuracy , decrease the effort of the calculation by reducing the orbital space . the crucial point for this effect to appear seems to be the reduction of the electron - electron interaction compared to the coulomb interaction rather than the long - range character of the modified interaction . where @xmath76 is an accurate wave function and @xmath133 are approximate wave functions generated from small orbital spaces @xmath139 , @xmath140 and @xmath141 , as a function of @xmath66 for the he atom .
] where @xmath76 is an accurate wave function and @xmath133 are approximate wave functions generated from small orbital spaces @xmath140 and @xmath141 , as a function of @xmath66 for the be atom . ]
the idea of using the partially interacting model system of eq .
( [ eq_fmu ] ) has been explored with different techniques to compute the wavefunction @xmath76 , and using different approximations for the functionals of sec .
[ sec_functionals ] .
we review very briefly some of the corresponding results .
the lda functional for @xmath93 $ ] has been combined with configuration interaction ( ci ) to handle the wavefunction @xmath76 in refs .
@xcite ; the method has been applied to atoms and small molecules . again using lda for the short - range functional , in ref .
@xcite it has been shown that long - range van der waals forces in rare gas dimers can be well described by using second - order perturbation theory for @xmath76 .
a short - range gga functional has been combined with the coupled - cluster ( cc ) method to describe @xmath76 , with very good results for small molecules , both for the closed and the open shell cases @xcite : in particular , the results from multideterminantal dft for small and medium basis - set sizes are better of both the pure dft result and the pure cc result . in ref .
@xcite , a different gga functional @xcite has been used , and @xmath76 has been determined by multi - configuration self consitent field ( mcscf ) . the corresponding application to systems in which near - degeneracy effects play a major role is promising , although self - interaction errors in the functionals are still a problem in some cases @xcite .
realistic @xmath116 from eqs .
( [ eq_eff ] ) have been generated for two - electron atoms in ref .
@xcite : with a very simple approximation for @xmath120 ( inspired to the one used for the electron gas at @xmath125 ) , extremely accurate @xmath93 $ ] have been obtained .
this new strategy of producing functionals is still at a very early stage , and several steps are needed for its development . in particular , we need better approximations for @xmath120 , and the implementation of an efficient algorithm to couple the radial equations ( [ eq_eff ] ) with the method used to solve the model hamiltonian @xmath75 . in conclusions , changing
the model hamiltonian in dft calculations seems to be a promising alternative to the construction of better exchange - correlation functionals for standard kohn - sham dft .
of course , the method has been under development only in the last decade , and it still needs further improvement and investigation in many aspects .
see , e.g. , j.a .
pople , _ nobel lecture : quantum chemical models _ , rev .
( 1999 ) * 71 * , 12671274 .
see , e.g. , w.m.c .
foulkes , l. mitas , r.j .
needs , and g. rajagopal , _ quantum monte carlo simulations of solids _ , rev .
( 2001 ) * 73 * , 3383 .
see , e.g. , w. kohn , _ nobel lecture : electronic structure of matter wave functions and density functionals _ , rev .
( 1999 ) * 71 * , 12531266 .
see , e.g. , a.j .
coleman and v.i .
yukalov , _ reduced density matrices : coulson s challenge _ ,
springer , berlin , 2000 .
p. hohenberg and w. kohn , _ inhomogeneous electron gas _ phys .
( 1964 ) * 136 * , b864-b871 .
m. levy , _ universal variational functionals of electron densities , 1st - order reduced density - matrices , and natural spin - orbitals and solution of the v - representability problem _ , proc .
( 1979 ) * 76 * , 6062 - 6065 .
lieb , int .
j. quantum chem . , _ density functionals for coulomb - systems _ ( 1983 ) * 24 * , 243 - 277 . w. kohn and l.j .
sham , _ self - consistent equations including exchange and correlation effects _ , phys . rev .
( 1965 ) * 140 * , a1133-a1138 .
see , e.g. , a. e. mattsson , _ in pursuit of the `` divine '' functional _ , science ( 2002 ) , * 298 * , 759 - 760 .
see , e.g. , s. kmmel and j.p .
perdew , _ optimized effective potential made simple : orbital functionals , orbital shifts , and the exact kohn - sham exchange potential _ ,
b ( 2003 ) * 68 * , 035103 1 - 15 ; w. yang and q. wu , _ direct method for optimized effective potentials in density - functional theory _ , phys
89 * , 143002 1 - 4 ( 2002 ) , and references therein .
see , e.g. , a. savin , f. colonna , r. pollet , _ adiabatic connection approach to density functional theory of electronic systems _ , int .
j. quantum chem . ( 2003 ) * 94 * , 166 - 190 , and references therein .
w. yang , _ generalized adiabatic connection in density functional theory _ , j. chem .
( 1998 ) , * 109 * , 10107 - 10110 .
see , e.g. , k. burke , j. p. perdew , and m. ernzerhof , _ why semilocal functionals work : accuracy of the on - top pair density and importance of system averaging _ , j. chem . phys . ( 1998 ) * 109 * , 3760 - 3771 .
j. rey and a. savin , _ virtual space level shifting and correlation energies _ , int .
j. quantum chem .
( 1998 ) , * 69 * , 581 - 590 . c. gutl ,
_ espace orbitalaires et thorie de la fonctionelle de la densit _ , phd thesis , universit pierre et marie curie ( paris , france ) . full text ( in french ) available at http://www.lct.jussieu.fr/dft/ g.e .
engel , and w.e .
pickett , _ investigation of density functionals to predict both ground - state properties and band structures _
b ( 1996 ) * 54 * , 8420 - 8429 .
j. toulouse , p. gori - giorgi and a. savin , _ a short - range correlation energy density functional with multi - determinantal reference _ , theor .
( 2005 ) , * 114 * , 305 - 308 .
j. toulouse , f. colonna , a. savin , _ long - range / short - range separation of the electron - electron interaction in density functional theory _
rev . a ( 2004 ) * 70 * , 062505 1 - 16 . s. paziani , s. moroni , p. gori - giorgi , and g.b .
bachelet , _ local - spin - density functional for multideterminant density functional theory _ ,
b ( 2006 ) * 73 * , 155111 1 - 9 .
gori - giorgi and a. savin , _ properties of short - range and long - range correlation energy density functionals from electron - electron coalescence _
rev . a ( 2006 ) * 73 * , 032506 1 - 9 .
gill and r.d .
adamson , _ a family of attenuated coulomb operators _ , chem .
phys . lett .
( 1996 ) * 261 * , 105 - 110 .
d. m. ceperley and b.j .
alder , _ ground state of the electron gas by a stochastic method _ ,
( 1980 ) * 45 * , 566 - 569 .
j. toulouse , a. savin , and h .- j .
flad , _ short - range exchange - correlation energy of a uniform electron gas with modified electron - electron interaction _ ,
j. quantum .
( 2004 ) * 100 * , 1047 . j. toulouse , f. colonna , and a. savin , _ short - range exchange and correlation energy density functionals : beyond the local density approximation _ , j. chem
( 2005 ) * 122 * , 014110 1 - 10 . j. heyd , g.e .
scuseria , and m. ernzerhof _ hybrid functionals based on a screened coulomb potential _ , j. chem .
( 2003 ) , * 118 * 8207 - 8215 .
e. goll , h .- j .
werner , and h. stoll , _ a short - range gradient - corrected density functional in long - range coupled - cluster calculations for rare gas dimers _ , phys .
( 2005 ) * 7 * , 3917 - 3923 .
e. goll , h .- j .
werner , h. stoll , t. leininger , p. gori - giorgi , and a. savin , _ a short - range gradient - corrected spin density functional in combination with long - range coupled - cluster methods : application to alkali - metal rare - gas dimer _ , chem .
, in press .
j. p. perdew , k. burke , and m. ernzerhof , _ generalized gradient approximation made simple _ , phys .
( 1996 ) * 77 * , 3865 - 3868 .
gori - giorgi and a. savin , _ simple model for the spherically- and system - averaged pair density : results for two - electron atoms _ , phys .
rev . a ( 2005 ) * 71 * , 032513 1 - 9 .
gori - giorgi and a. savin , _ system - adapted correlation energy density functionals from effective pair interactions _
, phil . mag .
( 2006 ) * 86 * , 2643 - 2659 .
p. gori - giorgi and j.p .
perdew , _ short - range correlation in the uniform electron gas : extended overhauser model _ ,
b ( 2001 ) * 64 * , 155102 1 - 8 . b. davoudi , m. polini , r. asgari , and m. p. tosi , _ self - consistent overhauser model for the pair distribution function of an electron gas in dimensionalities d = 3 and d = 2 _ , phys .
b ( 2002 ) * 66 * , 075110 1 - 8 .
t. leininger , h. stoll , h .- j .
werner , and a. savin , _ combining long - range configuration interaction with short - range density functionals _ , chem .
( 1997 ) * 275 * , 151 - 160 .
r. pollet , a. savin , t. leininger , and h. stoll , _ combining multideterminantal wave functions with density functionals to handle near - degeneracy in atoms and molecules _ , j. chem .
( 2002 ) * 116 * , 1250 - 1258 . j. g. angyan , i. c. gerber , a. savin , and j. toulouse , _ van der waals forces in density functional theory : perturbational long - range electron interaction corrections _ ,
rev . a ( 2005 ) * 72 * , 012510 1 - 9 .
j. toulouse , _ extension multidterminantale de la mthode de kohn - sham en thorie de la fonctionnelle de la densit par dcomposition de linteraction lectronique en contributions de longue porte et de courte porte _ , phd thesis , universit pierre et marie curie ( paris , france ) . full text ( in french ) available at http://www.lct.jussieu.fr / toulouse/.
|
the formalism of kohn and sham uses a specific ( model ) hamiltonian which highly simplifies the many - electron problem to that of noninteracting fermions .
the theorem of hohenberg and kohn tells us that , for a given ground state density , this hamiltonian is unique . in principle , this density can be chosen as that of the real , interacting system . to obtain the energy , or other properties of the real system ,
approximations are needed . working with non interacting fermions
is an important simplification , but it may be easier to produce approximations with different choices of the model hamiltonian . the feature that the exact density is ( ideally ) reproduced can be kept in the newly defined fictitious systems . using model hamiltonians having the same form as the physical one ,
that is , being built of one- and two - body operators , allows to approach the physical hamiltonian arbitrarily close , and thus a systematic reduction of the approximations .
|
over 50 years , fermi liquid theory serves as the basis for our understanding of the behavior of electrons in metals @xcite .
developed by landau to describe the behavior of @xmath0 at low @xmath1 @xcite and based on a minimal number of postulates , it allows to describe the behavior of interacting fermions not only qualitatively , but also quantitatively ( see e.g. , @xcite ) .
the key postulate of fermi liquid theory is that the interaction between fermions with energies near the fermi energy is weak in any fermi liquid , and does not qualitatively change the structure of low - energy fermionic states compared to a fermi gas . in the green s function language
, the condition of weak scattering near the fermi surface implies that the fermionic self - energy @xmath2 is linear in @xmath3 and in @xmath4 at smallest frequencies and smallest @xmath5 . in particular , at @xmath6 ,
@xmath7 the imaginary part of @xmath8 , responsible for the scattering from one electronic state into the other , and hence , for a finite lifetime of a given fermionic state , must be smaller than @xmath3 , i.e. , should behave at low frequencies as @xmath9 the conditions specified by ( [ 1 ] ) and ( [ 2 ] ) imply that at low energies , the dominant effect of electron - electron interaction is the shift of the energy levels , proportional to the deviation of the energy level from the fermi surface . at the same time , the levels themselves remain intact , i.e. , they are not destroyed by the interaction .
( [ 1 ] ) and ( [ 2 ] ) imply that in the limit of small @xmath3 and small @xmath10 , the green s function for interacting fermions has the same form as for non - interacting fermions , modulo the renormalization of the fermi velocity and the overall @xmath11 factor : @xmath12 where @xmath13 , and @xmath14 is the effective mass .
since the corrections due to electron - electron interactions do not change the functional form of the fermionic propagator , the thermodynamic characteristics of a fermi system , which describe the system s reaction to small external perturbations such as temperature or magnetic field , must retain the same functional form as for free fermions .
this , in particular , implies that the specific heat is linear in @xmath1 at low temperatures , @xmath15 , and the static spin susceptibility reduces to a constant at @xmath16 .
the linear in @xmath1 specific heat and the constant spin susceptibility are the two fundamental properties of a fermi liquid in any @xmath17 . in this communication
, we consider the situation when the system at @xmath16 is close to a density - wave instability at @xmath18 .
the most straightforward example is a system near a ferromagnetic quantum critical point ( fqcp ) .
spin fluctuations can be either isotropic or anisotropic , i.e. , of ising type , due to spin - orbit coupling .
our conclusions are applicable for both cases . for definiteness , we consider isotropic case .
we assume that the system is weakly coupled far away from the transition .
this implies that the corrections to a fermi liquid are generally small .
near the transition , however , the bosonic mode that mediates interaction between fermions becomes soft , and self - energy corrections increase @xcite . in @xmath19 , the dimensionless coupling constant @xmath20 scales as a positive power of the ferromagnetic correlation length @xmath21 : @xmath22 , where @xmath23 is the coupling , @xmath24 is the interatomic spacing , and @xmath25 is of order of the fermionic bandwidth . at @xmath26 , @xmath27 is small if , as we assume , the interaction is small compared to @xmath28 . however , as @xmath21 increases , @xmath20 also increases and diverges at a fqcp , where @xmath29 .
the issue brought about in several recent publications @xcite is whether in this situation a fermi liquid survives right up to a critical point , or it is destroyed already at some distance away from criticality .
the possibility that a fermi liquid is destroyed before a fqcp is based on the observation @xcite that if the fully renormalized interaction is static , then the effective mass @xmath30 diverges already at @xmath31 , i.e. , at some distance from a fqcp .
it was suggested @xcite that a new intermediate phase termed as `` fermionic condensate '' may emerge near a fqcp .
we will show that the @xmath32 dependence of the effective mass holds only as long as the retardation of the interaction is a small perturbation , and the self - energy predominantly depends on @xmath33 : @xmath34 .
we show that , as @xmath21 increases , the self - energy crosses over from @xmath35 to @xmath8 already at @xmath36 . above the crossover , the expression for the mass changes from @xmath37 to @xmath38 .
then , as the system enters into the strong coupling regime @xmath39 , the effective mass increases , but it does not diverge until the system reaches a fqcp .
as a result , no new intermediate phase phase appears near fqcp .
a similar situation has recently been found in @xcite for a transition in a spatially isotropic system into a density - wave state at at finite @xmath40 .
there is a qualitative similarity between the cases @xmath40 considered in @xcite and @xmath18 considered here in both cases the effective mass does not diverge before fqcp .
however , the physics of the crossover is somewhat different in the two cases ( see below ) .
in our analysis we neglect two other peculiarities of the system behavior near a fqcp .
the first is the non - analytic momentum dependence of the static spin susceptibility @xcite .
it either makes the ferromagnetic transition first order , or leads to spiral distortion of a second - order ferromagnetic ordering .
the second is the pairing instability @xcite .
it leads to the development of a dome on top of the fqcp where the system becomes a @xmath41-wave superconductor .
these two peculiarities compete with each other , but both affect the system behavior in the immediate vicinity of the transition , when the locality is already self - generated . in this paper
we will be primarily interested in the origin for the locality , and how the self - energy crosses over from @xmath35 to @xmath8 .
following earlier studies @xcite we assume that the system behavior near a fqcp is described by the spin - fermion model , in which fermion - fermion interaction is mediated by collective spin fluctuations peaked at @xmath18 : @xmath42 where @xmath43 .
this model describes the physics at low energies and can , in principle , be obtained from the original hubbard - type model of short - range electron - electron interaction on a lattice , after fermions with energies comparable to @xmath28 are integrated out @xcite . within rpa approximation ,
this procedure is illustrated in fig.[fig_rpa ] .
this separation of scales into a low - energy part and a high - energy part makes sense only when @xmath44 as we will see below that quantum - critical behavior extends to energies of order @xmath45 .
the collective spin fluctuations are low - energy degrees of freedom near fqcp , and they remain in the theory after the high - energy degrees of freedom are eliminated . for fermions ,
the restriction to low - energies implies that their dispersion can be linearized near the fermi surface : @xmath46 . the schematic derivation of the effective interaction within the rpa approximation .
the dashed line is the effective bosonic propagator . near sdw instability ,
the vertices between solid and dashed lines contain spin @xmath47 matrices . ] to keep @xmath23 smaller than @xmath28 and , at the same time , bring the system to a ferromagnetic instability requires extra assumptions . within the rpa approximation @xcite , the full spin susceptibility @xmath48 , where @xmath49 is the static spin susceptibility of free fermions .
ferromagnetic instability then occurs only when @xmath50 .
there are two ways out of this .
first , if the range of the interaction @xmath51 is much larger than the interatomic spacing ( i.e. , @xmath52 ) , the condition for fqcp becomes @xmath53 , i.e. , the smallness of @xmath54 is consistent with the closeness to fqcp @xcite .
second , one can use the fact that the renormalizations outside rpa make effective coupling frequency dependent @xmath55 .
the ferromagnetic instability is then determined by @xmath56 , while the coupling constant @xmath20 involves @xmath57 at small @xmath58 .
the separation of scales then becomes possible if @xmath59 .
we assume that either the range of the interaction , or its frequency dependence validate the condition @xmath60 near a fqcp .
consider first the self - energy within the perturbation theory for eq .
( [ hint ] ) .
the self - energy diagram to first order in @xmath61 is presented in fig .
[ fig_1]a . in analytic form , at @xmath16 @xmath62 where @xmath63 and @xmath46 .
the full green s function is related to @xmath64 as @xmath65 . since we are interested in finite @xmath66 and finite deviations from the fermi surface , it is convenient to subtract the constant @xmath67 from the self - energy in ( [ 1.7 ] ) before evaluating the integral .
this constant @xmath67 accounts for the renormalization of the chemical potential and does not affect the physics near fqcp .
we will just neglect it below .
expanding the fermionic dispersion @xmath68 at small @xmath69 as @xmath70 , where @xmath71 is the angle between @xmath72 and @xmath73 , we obtain @xmath74 where @xmath75 , and @xmath76[\omega + \omega + i v_f ( \delta k + q \cos \theta)]}. \label{1.11}\end{aligned}\ ] ] we see from ( [ 1.10 ] ) that if @xmath77 is non - singular at small @xmath66 and @xmath78 , the self - energy scales as @xmath66 and @xmath78 .
this implies that the self - energy modifies the parameters in the green s function but does not affect its functional form compared to free fermions , in agreement with the key postulate of the fermi liquid theory .
( a , b ) the self - energy diagram with the bare interaction @xmath61 , and the effective interaction @xmath79 .
( c ) the diagram for the full self - energy .
it contains full fermionic propagator , full vertices , and the fully renormalized @xmath80 . ]
although @xmath81 is non - singular , it has to be evaluated with extra care by two reasons .
first , for a constant @xmath61 , the integrand in ( [ 1.11 ] ) formally diverges at large @xmath82 and @xmath69 , hence the ordering of the integrations over momentum and frequency is relevant .
second , the integrand in ( [ 1.11 ] ) contains two poles separated only by @xmath3 and @xmath83 .
in this situation , one can not just set @xmath84 in evaluating @xmath81 , but rather should consider the limit @xmath85 .
a similar situation occurs near a qcp in which the instability occurs at a finite @xmath69 @xcite .
to evaluate the integral at finite @xmath78 and @xmath66 , we note that @xmath61 can be approximated by a constant only at small @xmath69 , while at large momenta , @xmath61 eventually vanishes for any realistic potential .
the vanishing of @xmath61 at large @xmath69 provides the ultraviolet regularization of the integral in ( [ 1.11 ] ) .
since the regularization is provided by the momentum dependence , one should first integrate over frequency in infinite limits ( the corresponding integral in ( [ 1.11 ] ) is convergent ) , and then integrate over @xmath73 . for practical purposes , it is convenient to introduce @xmath86
. then @xmath87 . performing the integration over @xmath88 , we find that the frequency integral is nonzero when @xmath89 and @xmath90 have different signs , and the two poles in ( [ 1.11 ] ) are located in different half - planes of frequency . for @xmath89 within this interval , the integration over frequency yields @xmath91[\omega + \omega + i v_f ( \delta k + q_1 ) ] } \nonumber \\ & & = - \frac{1}{i\omega_m - v_f \delta k } \label{1.13}\end{aligned}\ ] ] for other values of @xmath89 , the poles are in the same half - plane of frequency , and the frequency integral vanishes . since the result of the frequency integration in ( [ 1.13 ] ) does not depend on @xmath89 , the subsequent integration over @xmath89 just gives the length of the interval where the poles are in different half - planes of frequency , i.e. , @xmath78 .
as @xmath78 is vanishingly small , @xmath92 can be safely replaced by @xmath93 .
we then obtain @xmath94 where @xmath95 accordingly , @xmath96 we see that the self - energy depends only on momentum . this result
could indeed be anticipated as the interaction does not depend on frequency .
one could eliminate the frequency dependence in @xmath64 at the very first step of the calculations by shifting the frequency variable in eq .
( [ 1.7 ] ) from @xmath88 to @xmath97 .
we will see , however , that the integration procedure presented above is more appropriate as it allows for straightforward extensions to the case when the interaction does depend on frequency
. substituting the self - energy from ( [ 1.17 ] ) into the green s function , we obtain @xmath98 , i.e. , the effective mass @xmath99 is @xmath100 by analyzing the computation of @xmath14 , we see that the renormalization of the fermionic mass indeed comes from intermediate fermionic states very near the fermi surface ( @xmath101 ) .
the regions away from the vicinity of the fermi surface do not contribute to the mass renormalization .
this is again in agreement with the fermi liquid theory .
note that from the mathematical point of view , the renormalization of the effective mass is a typical example of an `` anomaly '' @xcite .
indeed , @xmath102 because of the double pole in ( [ 1.11 ] ) , but the limit @xmath103 of @xmath81 is finite because the double pole is splitted into two single poles , and at finite @xmath78 there exists a range where these two poles are in different half - planes of frequency .
the smallness of the region where this happens is compensated by the smallness of the distance between the two poles . as a result , the momentum and frequency integration in eq .
( [ 1.11 ] ) yields @xmath104 , which is finite at a nonzero @xmath78 and depends on the ratio @xmath105 .
the same term @xmath106 is present in the particle - hole bubble at small momentum and frequency transfer and accounts for the difference between @xmath107 and @xmath108 in the fermi liquid theory @xcite .
the same type of analysis can be performed for the spin susceptibility , and the result is that the renormalization of @xmath109 due to fermion - fermion interaction also comes from `` anomalous '' terms associated with the splitting of a double pole in the particle - hole bubble at small momentum and frequency transfer . from this perspective
, landau fermi liquid theory can be viewed as the `` theory of anomalies '' .
further , we see from ( [ 1.17 ] ) that the quasiparticle @xmath11 factor is not renormalized by the self - energy and remains @xmath110 .
where the renormalization of @xmath11 comes from ? to answer this question , we return to eq .
( [ 1.11 ] ) .
the renormalization of @xmath11 is related to the renormalization of the frequency dependence of the green s function .
the latter comes from a regular part in @xmath77 , for which @xmath111 . as we said , for a purely static @xmath61 , @xmath102 .
suppose now that the effective interaction acquires some dependence on frequency after high - energy fermions are integrated out .
this frequency dependence must be analytic , as relevant fermionic energies are @xmath112 , i.e. , are much larger than @xmath88 . quite generally then , @xmath113 where @xmath114 is the spin velocity . as the frequency dependence comes from fermions , @xmath115 . for the interaction as in ( [ 1.21 ] ) , the frequency integral in eq .
( [ 1.11 ] ) for @xmath116 has a double pole at @xmath117 , but also has two extra poles at @xmath118 .
these two extra poles are present in both half - planes of frequency .
hence , even if the frequency integration is extended to the half - plane where there is no double pole , the result of frequency integration remains non - zero , and @xmath116 becomes finite . as the two new poles are not associated with the splitting of a double pole , @xmath116 is generally determined by fermionic states far away from the fermi surface . in particular , at @xmath119 ,
the poles in @xmath120 are located at @xmath121 . by this reason , @xmath11 is the input parameter for the fermi - liquid theory , but it can not be evaluated within the fermi liquid theory , which deals only with fermions near the fermi surface .
we see therefore that , in general , the second - order self - energy @xmath122 has two contributions .
the first , regular contribution , comes from fermionic states far from the fermi surface and yields @xmath123 the second , anomalous contribution , comes from fermionic states near the fermi surface and yields @xmath124 substituting both terms into the expression for @xmath125 , we see that the anomalous self - energy gives rise to the mass renormalization , while the regular part of the self - energy accounts for the renormalization of @xmath11 : @xmath126 the above consideration can be extended to an arbitrary strong interaction .
the full self - energy diagram is presented in fig .
[ fig_1]b .
it contains full green s function , the fully renormalized interaction , and the full vertices .
as long as the full green s function at low energies has the fermi liquid form of eq .
( [ 3 ] ) , the self - energy has an anomalous term which comes from the integration over electronic states very near the fermi surface . for these contribution ,
the fully renormalized vertices can be approximated by their values at @xmath127 and @xmath128 .
the integration over @xmath88 and over @xmath89 then proceeds the same way as before and yields @xmath129 where now @xmath130 here @xmath131 is the fully renormalized interaction , and @xmath132 accounts for the vertex renormalization .
observe that @xmath133 in ( [ 1.25 ] ) is a bare fermi velocity .
the regular contribution can be rather complex because both @xmath134 and @xmath135 depend on frequency .
still , however , the regular self - energy depends only on @xmath116 and is given by @xmath136 substituting the full self - energy into the equation for the full green s function , we obtain @xmath137 eq .
( [ 1.27 ] ) determines @xmath11 , while eq .
( [ 1.28 ] ) determines the mass renormalization in terms of @xmath11 and the coupling constant @xmath20 . as we discussed , the low - energy theory based on the spin - fermion model is valid only when the spin - fermion interaction @xmath23 is much smaller than the fermionic bandwidth @xmath138 .
the dimensionless @xmath20 scales with @xmath54 , and is a small number within the spin - fermion model when @xmath119 . alternatively speaking , far away from quantum criticality , the spin - fermion model is only valid in the weak coupling limit .
consider , however , what happens with @xmath20 , given by ( [ 1.24 ] ) when the system approaches fqcp , and @xmath21 diverges .
substituting @xmath139 into ( [ 1.24 ] ) , and assuming that the vertex renormalization is non - singular at small @xmath69 , we obtain @xmath140 for @xmath141 , the integral over @xmath69 is infrared convergent , even when @xmath142 .
the integral is then determined by large @xmath143 and yields @xmath27 , that is , @xmath20 remains small even at fqcp . for @xmath144 ,
the situation is , however , different . for @xmath145 , the integral in ( [ 2.2 ] )
depends logarithmically on the lower limit @xcite , and @xmath146 we see that , despite the smallness of the @xmath54 ratio , the coupling constant diverges at fqcp .
this implies that in the immediate vicinity of fqcp , the weak coupling approximation becomes invalid . for @xmath19 ,
the divergence of @xmath20 becomes a power law : @xmath147 .
for @xmath148 we have @xcite @xmath149 before we proceed with the analysis of the crossover in the self - energy , it is instructive to compare the diagrammatic calculations of @xmath14 with the fermi liquid theory .
for definiteness , we consider the @xmath150 case . in the fermi liquid theory ,
@xmath14 is related to the quasiparticle interaction function as @xmath151 where @xmath152 is the charge component of the quasiparticle interaction function @xmath153 , which also contains the spin component @xmath154 .
@xmath71 is the angle between @xmath155 and @xmath156 , which both are at the fermi surface , @xmath157 is the solid angle . the diagrams for the quasiparticle interaction function @xmath158 to first order in the interaction .
] to first order in the interaction , the quasiparticle interaction function coincides , up to the overall minus sign , with the antisymmetrized interaction at zero momentum transfer . for spin - mediated interaction ,
we have ( see fig .
[ fig_2 ] ) @xmath159 using @xmath160 we obtain @xmath161 substituting this into ( [ 1.33 ] ) , we obtain @xmath162 this expression coincides with the lowest - order diagrammatic result . to see this , one can evaluate the mass renormalization explicitly , using @xmath61 from eq .
( [ hint ] ) and @xmath163 , and compare with ( [ 1.15 ] ) , ( [ 1.18 ] ) .
alternatively , one can re - evaluate the diagrammatic expression to reproduce ( [ 1.33_a ] ) . a way to do
this is to integrate over frequency in ( [ 1.7 ] ) without expanding @xmath164 in @xmath69 .
the frequency integration yields @xmath165 subtracting a constant @xmath166 from @xmath35 and expanding the difference to first order in @xmath167 , we obtain @xmath168 using @xmath169 , introducing @xmath170 as both @xmath171 and @xmath172 are near the fermi surface , and shifting the integration variable from @xmath73 to @xmath172 , we obtain @xmath173 substituting this into the expression for the green s function , we reproduce eq .
( [ 1.33_a ] ) .
the extension of the analogy between the fermi liquid theory and diagrammatic technique to an arbitrary strong interaction requires some care .
the diagrammatic theory yields , at arbitrary interaction strength , @xmath174 the fermi liquid formula , eq .
( [ 1.33 ] ) contains @xmath175 instead of @xmath176 . in arbitrary fermi liquid
, @xmath152 is related to the charge component of the vertex function @xmath107 as @xmath177 @xcite .
the full vertex @xmath107 includes all regular vertex renormalizations , but does not include anomalous vertex corrections from the particle - hole bubble at small momentum and frequency transfer ( the latter disappear in the limit @xmath178 ) . the diagrammatic series for @xmath107 . the diagrams labeled as `` b '' must be included , but the diagrams labeled as `` a '' should be excluded as their inclusion would amount to double counting . ]
we see that fermi liquid and diagrammatic expressions for @xmath14 are equivalent if the relation between @xmath179 and @xmath175 is @xmath180 , i.e. , is the same as in ( [ 1.35_c ] ) , but with one power of @xmath11 . to understand how this @xmath11 appears ,
consider the diagrammatic series for @xmath107 in terms of @xmath181 , fig.[fig_3 ] the diagrams of fig .
[ fig_3]a should be neglected as their inclusion would amount to double counting - these renormalizations are already incorporated into the effective interaction @xmath182 , when high - energy fermions are integrated out in the rg - type procedure .
the diagrams of fig .
[ fig_3]b , however , should be included into @xmath107 as in the rg treatment they vanish because of double poles . since @xmath107 does not include anomalous contributions from particle - hole bubbles , the diagrams in fig .
[ fig_3]b are non - zero only if @xmath183 depends on frequency , otherwise the frequency integral of the convolution of two fermionic propagators vanishes .
we remind that when the effective @xmath182 depends on frequency , @xmath184 is non - zero , and @xmath11 renormalizes down from @xmath110 .
the same factor @xmath116 appears in the diagrammatic series for @xmath185 . for weakly momentum dependent @xmath183 , the diagrams for @xmath185 factorize and form geometric series . in this situation
, @xmath185 reduces to @xmath186 using the relation ( [ 1.27 ] ) between @xmath11 and @xmath116 , we find @xmath187 , hence @xmath188 using ( [ 1.38 ] ) , we immediately find that the diagrammatic expression for the effective mass , eq .
( [ 1.33_b ] ) , is equivalent to the expression for the effective mass in the fermi liquid theory , eq .
( [ 1.33 ] ) . for strong momentum
dependent @xmath183 , the diagrammatic series for @xmath189 can not be factorized , and we did nt find how to prove eq .
( [ 1.38 ] ) by summing up the diagrams .
we now return to the results for the effective mass , and the quasiparticle @xmath11-factor , eqs ( [ 1.27 ] ) and ( [ 1.28 ] ) .
assume momentarily that the interaction is purely static .
then @xmath110 and @xmath190 as we said , when the system approaches the fqcp and @xmath191 , @xmath20 increases and eventually becomes larger than @xmath192 . according to ( [ 2.1 ] ) , at @xmath193 , the effective mass diverges
, i.e. , the fermi - liquid state becomes unstable .
this effect was noticed in @xcite .
the authors of @xcite argued that for @xmath194 , the fermi liquid state is replaced by the new state of matter which they termed as `` fermionic condensate '' .
the issue we now address is whether eq .
( [ 2.1 ] ) is valid at @xmath195 .
this equation was obtained assuming that @xmath110 .
we will argue that the renormalization of @xmath11 can only be neglected at parametrically small @xmath20 , while at @xmath195 , the renormalization of @xmath11 is essential .
when this renormalization is included , the effective mass remains finite at @xmath195 , and diverges only at fqcp , where @xmath20 becomes infinite . to understand this ,
consider the renormalization of the @xmath11 factor in more detail .
we remind that it originates from the frequency dependence of the the full vertices and the fully renormalized effective interaction @xmath196 .
the fully renormalized @xmath196 differs from @xmath79 by the bosonic self - energy @xmath197 ( see fig . [ fig_4 ] ) : @xmath198 the contribution to @xmath199 from high - energy fermions is already incorporated into @xmath120 and must therefore be neglected to avoid double counting .
however , the polarization bubble contains also the landau damping term which comes from internal fermions in the bubble with energies smaller than @xmath88 .
this term is not included into @xmath120 as in the the rg procedure that gives @xmath120 one assumes that internal frequencies are much larger than external frequencies .
the diagram for the bosonic self - energy @xmath199 . ] we verified aposteriori that the frequency dependence of the effective interaction is more relevant than the frequency dependence of the full vertices ( the latter leads only to minor corrections at @xmath200 ) .
accordingly , we assume that the full vertices remain static , absorb the vertex renormalization factor @xmath201 into @xmath23 , and focus on @xmath196 .
the frequency dependence of @xmath202 comes from two sources .
first , @xmath203 already possess some regular frequency dependence , see eq .
( [ 1.21 ] ) .
second , @xmath199 also introduces frequency dependence into @xmath196 . for small @xmath69 and even smaller @xmath204 ,
@xmath205 substituting it into ( [ dec_29_1 ] ) and using ( [ 1.21 ] ) for @xmath120 , we obtain @xmath206 where @xmath207 , and we used the facts that @xmath115 and @xmath208 . at small frequencies and small momenta , the landau damping term obviously dominates over the regular , @xmath209 term in ( [ 2.8 ] ) .
we therefore will only keep the landau damping term .
we now use eq .
( [ 2.8 ] ) to compute @xmath116 and @xmath11 ( see eqs .
( [ 1.22 ] ) and ( [ 1.27 ] ) ) .
according to eq .
( [ 1.11 ] ) , @xmath210 integrating in ( [ 2.9 ] ) first over frequency and then over momentum and introducing @xmath211 we obtain @xmath212 where @xmath213 , and at large @xmath214 , @xmath215 vanishes as some power of @xmath214 . in @xmath148 , @xmath216 and @xmath217 .
we now analyze the result . that at large @xmath214 , @xmath116 is small
could be anticipated as @xmath218 corresponds to @xmath219 , in which case the frequency integral in ( [ 2.9 ] ) vanishes , i.e. , @xmath102 .
less anticipated is , however , the fact that in the opposite limit of small @xmath214 , @xmath220 , and @xmath221 becomes independent on @xmath222 .
this implies that at small @xmath214 , the regular term in the fermionic self - energy is of the same order as the singular term .
furthermore , one can easily verify that at @xmath223 , @xmath116 comes from fermions in the vicinity of the fermi surface , just as the anomalous part of @xmath224 .
combining regular and anomalous terms in the self - energy , we then obtain @xmath225 the @xmath226dependence of @xmath64 is canceled out between @xmath227 and @xmath228 .
we see that the self - energy is totally different at small @xmath214 and at large @xmath214 . at large @xmath214 ,
i.e. , at some distance away from fqcp , the anomalous part of the self - energy well exceeds the regular part , and @xmath34 . at small @xmath214 , both regular and anomalous parts of the self - energy are of the same order , and the sum of the two yields @xmath229 .
in this last case , @xmath230 , hence @xmath231 eq . ( [ 1.28_1 ] ) implies that @xmath14 does not diverge before fqcp .
the crucial issue is at what @xmath232 the system experiences the crossover from @xmath35 to @xmath8 .
if the crossover does not occur up to @xmath233 , the effective mass still diverges at @xmath234 , and the eventual transformation to @xmath8 becomes meaningless .
however , the crossover occurs already at small @xmath20 , when the calculations are under control . to see this
, we note that @xmath122 is the scaling function of the single parameter @xmath214 .
the crossover in the self - energy then obviously occurs at @xmath235 . using @xmath236 and the definition of @xmath232
, we obtain that @xmath237 hence the crossover from @xmath35 to @xmath8 occurs at @xmath238 this is the main result of the paper .
we see that the crossover in the self - energy occurs already at small @xmath232 , well before the effective mass apparently diverges .
this implies that eq .
( [ 2.1 ] ) is only applicable at small @xmath20 , and it can not be extrapolated to @xmath31 . at a first glance , the fact that @xmath239 and @xmath240 are comparable above the crossover to @xmath8 , invalidates the idea that the self - energy in a fermi liquid theory can be viewed as an anomaly .
it turns out , however , that the self - energy @xmath241 can be also identified with the anomaly , however the regularization procedure must be different from the one we used to obtain @xmath242 . to understand this , consider again the full self - energy , eqs .
( [ 1.10 ] ) , ( [ 1.11 ] ) , but now substitute into ( [ 1.11 ] ) the full @xmath196 instead of the bare @xmath120 .
the anomalous part of the self - energy is associated with the splitting of the double pole and is @xmath243 , where @xmath244 is determined by the integral @xmath245 in sec.[se ] , we evaluated this integral by integrating over frequency first .
the frequency integration restricted the integral over @xmath89 to @xmath246 ( for positive @xmath78 ) .
the result then was @xmath247 and it yielded @xmath248 , i.e. , the self - energy below the crossover .
however , as we already have said , the @xmath249 integral in ( [ 2.14 ] ) is formally ultraviolet divergent , and the result of the integration over momentum and frequency depends on the order of the integration . to check this ,
let s integrate over @xmath250 first .
the integration proceeds in the same way as before , but now @xmath251 where @xmath252 .
the integral over @xmath253 is nonzero when the two poles are located in different half - planes of complex @xmath254 , i.e. , when @xmath255 ( for @xmath256 ) . performing the integration ,
we obtain @xmath257 using this result , we obtain @xmath258 , which coincides with the self - energy above the crossover . comparing @xmath259 and @xmath260
, we see that they differ by a constant @xmath261 . which of the two results is the correct one ?
the answer depends on the functional form of @xmath262 which regularizes the integrals ( [ 2.14 ] ) and ( [ 2.16 ] ) . indeed , eqs .
( [ 2.14 ] ) and ( [ 2.16 ] ) are valid as long as @xmath263 can be approximated by a constant and absorbed into @xmath232 . at small @xmath89 and @xmath88 ,
this is true if the system is not at fqcp .
however , at large @xmath89 and @xmath88 , the effective interaction obviously vanishes .
this implies that once we leave @xmath264 in the integrand for @xmath244 , the @xmath265 integral over @xmath89 and @xmath88 will become ultraviolet convergent
. then @xmath244 must be the same independent on what integration is performed first .
now , @xmath264 vanishes at large @xmath88 and at large @xmath89 , i.e. , both can serve as cutoff .
the momentum cutoff at @xmath266 is imposed by the lattice .
the frequency cutoff is due to the fact @xmath199 is linear in @xmath88 , hence @xmath267 vanishes at large frequencies . for @xmath268 ,
the frequency dependence of @xmath267 is negligible , and the regularization of the integral for @xmath244 is provided by the momentum cutoff at @xmath269 . in this situation ,
it is natural to integrate over @xmath88 first in infinite limits , and then integrate over @xmath89 up to the cutoff .
after the integral over @xmath88 is evaluated , the subsequent integration over @xmath89 is restricted to a much narrower range than the cutoff , and we obtain @xmath270 independent on the cutoff . as an exercise , we verified that the same result @xmath270 is obtained at @xmath268 by integrating first over @xmath89 , up to the cutoff , and then integrating over frequency .
after the frequency integration , the momentum cutoff disappears from @xmath271 . for @xmath223 ,
the situation is different as for typical @xmath272 , the landau damping term in @xmath267 exceeds the static part of the effective interaction .
this implies that the regularization is now provided by the dynamical part of @xmath267 . in this situation ,
it is natural to first integrate over @xmath89 , in infinite limits , and then integrate over @xmath88 .
the result is @xmath273 .
again , as an exercise , we verified that @xmath273 at @xmath223 is reproduced if we integrate over @xmath88 first , and then integrate over @xmath89 .
the landau damping term in @xmath267 is relevant at the intermediate stages of the calculations , but disappears from the final answer .
this consideration implies that at @xmath223 , the self - energy @xmath122 can again be divided into the anomalous and regular parts @xmath274 , but now @xmath275 and @xmath276 where @xmath277 is small at @xmath223 . in particular , for @xmath148 , @xmath278 .
we now consider what happens when the dimensionless coupling @xmath20 becomes large , i.e. , the system falls into the strong coupling regime . for definiteness , we focus on @xmath148 . in general , the strong coupling case can be hardly treated diagrammatically , as one needs to include infinite series of both self - energy and vertex correction diagrams . in our case
, the situation , however , simplifies as we still have @xmath54 as the small parameter .
as we just said , in this situation , only the anomalous self - energy , which depends on frequency , becomes large when @xmath279 , whereas the regular part of the self - energy remains small in @xmath280 even when @xmath281 .
we verified that in this situation , + ( i ) the quasiparticle density of states @xmath282 = n_0 ~{\text sign}~ \omega \label{dec_30_1}\ ] ] remains the same as for free fermions in the presence of @xmath283 , + ( ii ) vertex corrections at @xmath284 remain small in @xmath280 even when @xmath20 becomes large + ( iii ) the corrections to the landau damping term from various insertions into the particle - hole bubble are also small in @xmath280 , i.e. , @xmath199 is still determined by eq .
( [ 2.7 ] ) , even at strong coupling .
the smallness of vertex corrections and corrections to @xmath199 is the key element of eliashberg - type theories @xcite . if the self - energy depended on @xmath33 , eq .
( [ dec_30_1 ] ) would not hold and vertex corrections and corrections to @xmath199 would not be small . without vertex corrections and corrections to ( [ 2.7 ] ) , the full self - energy at _ arbitrary _ @xmath20 above the crossover is still given by eq .
( [ 1.10 ] ) , but with the overall @xmath11 , with @xmath285 instead of @xmath133 , and with @xmath196 given by ( [ 2.8 ] ) .
evaluating the self - energy and neglecting the regular piece in @xmath122 , we obtain @xmath286 substituting this result into @xmath287 , we obtain a set of two coupled equations for @xmath11 and @xmath288 : @xmath289 these equations are generally different from the weak coupling version , eqn .
( [ 1.28_1 ] ) due to the presence of the extra @xmath290 in the equation for @xmath291 .
however , solving these two equations , we obtain the same result as before : @xmath292 this equivalence with the weak - coupling result is the consequence of the fact that @xmath293 is the invariant of the set ( [ 2.19_a ] ) .
( [ 2.20_a ] ) is another key result of the paper .
we see that the effective mass remains finite all the way up to a fqcp , where it diverges together with @xmath20 , and fermi liquid description breaks down .
simultaneously , the quasiparticle @xmath11 factor gradually decreases with increasing @xmath20 and only vanishes at a fqcp .
this implies that the intermediate phase near a fqcp does not develop both at weak and strong coupling .
upper panel .
the scaling function @xmath294 from eq .
middle and lower panels : imaginary and real parts of the fermionic self - energy near the ferromagnetic quantum - critical point .
observe that @xmath295 is almost linear in @xmath3 in a wide frequency range . ] for completeness , we also present the result for the fermionic self - energy @xmath296 at arbitrary frequencies @xcite . substituting @xmath297 into the integral for the self - energy and integrating over @xmath89 ,
we obtain , for positive @xmath3 @xmath298 where @xmath299 .
the result can be cast into @xmath300 the scaling function @xmath294 is plotted in fig.[fig_5 ] at small @xmath254 @xmath301 . at large @xmath254 , @xmath302 then , at small @xmath303 , we recover the fermi - liquid behavior , eqs .
( [ 2.18_a ] ) , ( [ 2.20_a ] ) . at @xmath304 ,
the system falls off into the quantum - critical regime , and @xmath305 , where @xmath306 does not depend on @xmath21 .
note that still , @xmath307 , i.e. , quantum - critical behavior is confined to low frequencies . at the fqcp , @xmath308 vanishes , and @xmath309 behavior extends down to @xmath310 @xcite . the schematic behavior of the self - energy at various frequencies . ] the system behavior at various frequencies is schematically presented in fig .
[ fig_6 ] . at the fqcp , @xmath308 vanishes , and @xmath309 behavior extends down to @xmath310 . for arbitrary @xmath311 , in the quantum - critical regime , @xmath312 . in three dimensions , in the quantum - critical regime , @xmath313 , i.e. , in real frequencies @xmath314 ( a marginal fermi liquid behavior ) . the quantum - critical behavior in @xmath315 and
@xmath316 has been extensively studied recently both theoretically @xcite and experimentally @xcite , and we refer the reader to the existing literature on this subject .
to conclude , in this paper we considered a metal near an itinerant ferromagnetic quantum - critical point .
we assumed that the residual interaction @xmath23 between fermions and low - energy ferromagnetic collective spin excitations is smaller than the fermionic bandwidth @xmath317 .
the dimensionless coupling constant @xmath20 is then small away from a ferromagnetic quantum - critical point , at @xmath119 .
we argued that at @xmath191 , the dimensionless coupling increases as the system approaches a fqcp , and @xmath20 diverges at a fqcp .
we computed fermionic self - energy and found that at weak coupling , @xmath318 , and @xmath30 .
if this behavior extended to @xmath195 , the effective mass would diverge at some distance away from a fqcp .
however , we found that already at small @xmath319 , the self - energy crosses over from @xmath35 to @xmath8 , and the result for the effective mass changes to @xmath320 .
this implies that the effective mass does not diverge , and the fermi liquid description remains valid everywhere in the paramagnetic phase .
we argued that the result @xmath320 remains valid all the way up the fqcp , where @xmath14 diverges together with @xmath20 .
our consideration is complimentary to the recent analysis of the crossover from @xmath35 to @xmath8 in @xmath265 isotropic systems near a density - wave instability at a finite @xmath321 @xcite . in both cases ,
the crossover occurs at small @xmath322 , i.e. , before the system enters into the strong coupling regime .
still , the finite @xmath69 case considered in @xcite and the @xmath18 case considered here are physically different .
in particular , the landau damping term in @xmath199 scales as @xmath88 at @xmath323 and as @xmath324 at vanishingly small @xmath69 .
less obvious but physically relevant difference is that near the transition at @xmath321 , the crossover to @xmath8 occurs at vanishingly small coupling if we set the fermionic bandwidth to infinity @xcite . in other words the crossover near a qcp at a finite @xmath321 occurs at @xmath322 , where @xmath28 literally has the meaning of the bandwidth . at the same time , near the ferromagnetic transition
, the crossover occurs at a _ finite _
@xmath20 even if one sets the bandwidth to infinity and linearizes the fermionic spectrum , as we actually did . in this case , @xmath28 has the meaning of the fermi energy @xmath325 , rather than of the bandwidth .
the present analysis does not invalidate completely the idea that the effective mass can diverge before the system reaches a quantum - critical point .
we proved that this does not happen as long as @xmath54 is small , and the calculations in the crossover regime are under control .
what happens at @xmath326 is an open issue , and we refrain from speculating what the system behavior may be . in any event , for @xmath23 comparable to the bandwidth , one can not depart from the spin - fermion model as the very idea that one can integrate out high energy fermions and obtain an effective model for low - energy degrees of freedom becomes meaningless .
it is my pleasure to thank m. i. kaganov for numerous conversations during the completion of this work .
i acknowledge useful conversations with v. galitski , v. khodel , d. maslov , c. pepin , and v. yakovenko .
i am thankful to c. pepin and v. yakovenko for critical reading of the manuscript and useful comments .
the work was supported by nsf dmr-0240238 and the center for material theory at umd .
a. a. abrikosov , l. p. gorkov , and i. e.dzyaloshinski , _ methods of quantum field theory in statistical physics _ , ( dover publications , new york , 1963 ) ; e. m. lifshitz and l. p. pitaevski , _ statistical physics _ , ( pergamon press , 1980 ) .
z. wang , w. mao and k. bedell , phys .
* 87 * , 257001 ( 2001 ) ; k.b .
blagoev , j.r .
engelbrecht and k. bedell , phys .
lett * 82 * , 133 ( 1999 ) ; r. roussev and a. j. millis , phys .
b * 63 * , 140504 ( 2001 ) . s.a .
brazovskii , sov .
jetp * 41 * , 85 ( 1975 ) ; d. belitz , t. r. kirkpatrick , and t. vojta , phys . re v. b * 55
* , 9452 ( 1997 ) ; s. das sarma and e. h. hwang , phys . rev .
lett . * 83 * , 164 ( 1999 ) .
g. y. chitov and a. j. millis , , 5337 ( 2001 ) ; , 0544414 ( 2001 ) ; a. v. chubukov and d. l. maslov , phys .
b * 68 , * 155113 ( 2003 ) ; _ ibid . _ * 69 * , 121102 ( 2004 ) ; a. v. chubukov . c. pepin and j. rech , , 147003 ( 2004 ) ; d. belitz , t.r .
kirkpatrick , and joerg rollbuehler , , 155701 ( 2004 ) .
monthoux and g.g .
lonzarich , phys .
b * 59 * , 14598 ( 1999 ) .
s. b. treiman , r. jackiw , and d. j. gross , _ lectures on current algebra and its applications _ ( princeton university press , princeton , 1972 ) , see also b.l .
altshuler et al , europhys .
* 44 * , 401 ( 1998 ) ; r. haslinger and a. v. chubukov , , 214508 ( 2003 ) .
eliashberg , sov .
jetp * 11 * , 696 ( 1960 ) ; d.j .
scalapino in _ superconductivity _ , vol . 1 ,
r. d. parks , dekker inc . n.y .
1969 ; f. marsiglio and j.p .
carbotte , in ` the physics of conventional and unconventional superconductors ' , eds .
k.h . bennemann and j.b .
ketterson ( springer - verlag ) .
the @xmath309 behavior has also been found in other systems , by similar reasons : j. polchinski , nucl .
b * 422 * , 617 ( 1994 ) ; b. l. altshuler , l.b .
ioffe , and a. j. millis , phys .
b * 52 * , 5563 ( 1995 ) ; v. oganesyan , s. a. kivelson , and e. fradkin , phys .
b * 64 * , 195109 ( 2001 ) ; h.y .
kee and y.b .
kim , j. phys : cond .
matter * 16 * , 3139 ( 2004 ) ; i. vekhter and a. chubukov , , 016405 ( 2004 ) .
p. gegenwart , j. custers , t. tayama , k. tenya , c. geibel , g. sparn , n. harrison , p. kerschl , d. eckert , k .- h .
mller , and f. steglich , j. low temp .
phys . * 133 * , 3 ( 2003 ) .
j. custers , p. gegenwart , h. wilhelm , k. neumaier , y. tokiwa , o. trovarelli , c. geibel , f. steglich , c. pepin , and p. coleman , nature , * 424 * , 524 - 527 ( 2003 )
|
we analyze the behavior of interacting fermions near a ferromagnetic stoner instability .
we show that the landau damping of the spin susceptibility is a relevant perturbation near a ferromagnetic quantum - critical point ( fqcp ) .
we argue that , as the system approaches a fqcp , the fermionic self - energy crosses over from predominantly momentum dependent away from the transition to predominantly frequency dependent in the immediate vicinity of the transition .
we argue that due to this self - generated locality , the quasiparticle effective mass does not diverge before a fqcp is reached .
|
a quantitative theory of propagation of elastic waves in a fluid - saturated porous solid was formulated in the classic papers by biot [ 1 ] .
one of the major findings of biot s work was that there was a breakdown in poisseuille flow above a certain characteristic frequency specific to the fluid - saturated porous material .
biot theoretically studied this phenomenon by considering the flow of a viscous fluid in a tube with longitudinally oscillating walls under an oscillatory pressure gradient .
apart from the fundamental interest , the investigation of the dynamics of fluid in porous media under oscillatory pressure gradient and oscillating pore walls is of prime importance for the recently emerged technology of acoustic stimulation of oil reservoirs [ 2 ] .
for example , it is known that the natural pressure in an oil reservoir generally yields no more than approximately 10 percent oil recovery .
the residual oil is difficult to produce due to its naturally low mobility , and the enhanced oil recovery operations are used to increase production .
it has been experimentally proven that there is a substantial increase in the net fluid flow through porous space if the latter is treated with elastic waves .
however , there is a fundamental lack of understanding of the physical mechanisms of fluid mobilization in saturated rock through the effect of elastic waves ; the theory of such mobilization virtually does not exist .
biot s theory can be used to describe the interaction of fluid - saturated solid with the sound for a classic newtonian fluid ; however , oil and other hydrocarbons exhibit significant non - newtonian behavior [ 3 ] .
the aim of this paper is therefore to incorporate non - newtonian effects into the classical study of biot [ 1 ] .
recently , del rio , lopez de haro , whitaker [ 4 ] presented a study of enhancement in the dynamic response of a viscoelastic ( maxwell ) fluid flowing in a stationary ( non - oscillating ) tube under the effect of an oscillatory pressure gradient .
we combine this theory with the effect of the acoustic oscillations of the walls of a tube introduced by biot [ 1 ] , providing a complete description of the interaction of maxwell fluid , filling the pores , with acoustic waves . finally , in order to emphasize that the consept of dynamic permeability is an adequate way of desciprtion of the phenomenon , we note that this concept has been widely used before [ 5 ] . in the next section
we formulate our model , whereas the last section will conclude with the discussion of the results .
in this section we present our model of maxwell fluid flowing in a cylindrical tube whose walls are oscillating longitudinally and the fluid is subject to an oscillatory pressure gradient .
we give analitical solutions of the problem in the frequency domain .
the governing equations of the problem consist of the continuity equation for the incompressible fluid , @xmath2 and the linearized momentum equation , @xmath3 here , @xmath4 , @xmath5 , @xmath6 denote velocity , pressure and mass density of the fluid , whereas @xmath7 represents the viscous stress tensor .
we describe the viscoelatic effects of the fluid using maxwell s model , which assumes that @xmath8 where @xmath9 is the viscosity coefficient and @xmath10 is the relaxation time .
now , let @xmath11 be a velocity of the wall of the tube which oscillates in time as @xmath12 , where @xmath13 is the angular frequency .
the flow of fluid in a cylindrical tube with longitudinally oscillating walls can be descibed by a single component of the velocity , namely , its @xmath14-component @xmath15 ( @xmath14 axis is along the ceterline of the tube ) .
we use the cylindrical coordinate system @xmath16 in the treatment of the problem .
we introduce the relative velocty @xmath17 as @xmath18 .
thus , assuming that all physical quantities vary in time as @xmath12 , we arrive at the following master equation for @xmath17 @xmath19 here , we have introduced the following notations : @xmath20 which is a sum of the applied pressure gradient and force exerted on the fluid from the oscillating wall of the tube and @xmath21 , which is @xmath22 . note that a no - slip boundary condition at the wall is assumed .
the solution of eq.(4 ) can be found to be [ 1 ] @xmath23 where @xmath24 is the bessel function and @xmath25 . applying a no - slip boundary condition @xmath26 at the wall of the tube , where @xmath27 is its radius ,
we finally get @xmath28= -{{xa^2(1 - i \omega t_m)}\over { \nu } } { { 1}\over{(\beta a)^2 } } \left[1 - { { j_0(\beta r)}\over{j_0(\beta a ) } } \right ] . \eqno(5)\ ] ] defining the cross - section averaged velocity as @xmath29 we obtain @xmath30 \equiv k(\omega ) x. \eqno(6)\ ] ] here @xmath31 is the dynamic permeability [ 4 ] that describes the frequency dependent response of the tube to the applied total force on the fluid .
simple comparison reveals that eq.(6 ) resembles closely eq.(6 ) from ref . [ 4 ] , with the only difference being that we have @xmath32 in place of their @xmath33 . note that in the case of stationary tube walls ( @xmath34 ) eq.(6 ) exactly coincides with eq.(6 ) from ref.[4 ] .
simple calculation ( applying lhospital rule for the @xmath35 uncertanty ) shows that @xmath36 .
thus , following ref.[4 ] , we will introduce the dimensionless dynamic permeability as @xmath37 , which will be used later ( see fig.7 ) .
note , that we were able to easily reproduce fig . 1 from ref .
[ 4 ] , confirming the existence of sharp resonances of @xmath38 in the elastic regime ( see below for the definition of this regime ) . following the work of biot [ 1 ] we calculate the stress at the wall @xmath39 , @xmath40 note , that when @xmath41 this expression obviously coincides with the corresponding newtonian form .
the total friction force is @xmath42 .
following biot we calculate the ratio of total friction force to the average velocity , i.e. @xmath43 } \over{(1 -i \omega t_m ) } } \left[1 - { { 2 j_1(\beta a)}\over { ( \beta a ) j_0(\beta a ) } } \right]^{-1}. \eqno(8)\ ] ] simple analysis reveals that @xmath44 which corresponds to the limiting case of poiseuille flow .
following biot [ 1 ] , we also introduce a function @xmath0 with @xmath1 being @xmath45 in the following manner @xmath46 thus , @xmath47 } \over{(1 -i \kappa^2 / \alpha ) } } \left[1 - { { 2 j_1(\kappa \sqrt { i + \kappa^2 / \alpha})}\over { \kappa \sqrt { i + \kappa^2 / \alpha } j_0(\kappa \sqrt { i + \kappa^2 / \alpha } ) } } \right]^{-1}. \eqno(9)\ ] ] note , that @xmath0 measures the deviation from poiseuille flow friction as a function of frequency parameter @xmath1 , as introduced by biot [ 1 ] . in eq.(9 )
, @xmath48 denotes the deborah number [ 4 ] , which is defined as the ratio of characteristic time of viscous effects @xmath49 to the relaxation time @xmath10 , i.e. @xmath50 . as noted in ref .
[ 4 ] , the value of the parameter @xmath48 determines in which regime the system resides . beyond a certain critical value ( @xmath51 ) ,
the system is dissipative , and viscous effects dominate . on the other hand , for small @xmath48 ( @xmath52 ) the system exhibits viscoelastic behavoir which we call the elastic regime .
note , that the newtonian flow regime can be easily recovered from eq.(9 ) by putting @xmath53 .
we plot this limiting case in fig .
as it can be seen from the figure , both @xmath54 $ ] and @xmath55 $ ] exactly coincide with the newtonian limiting case studied in biot s work ( see fig . 4 in ref.[1 ] ) .
this graph demonstrates a breakdown in poiseuille flow as frequency increases ( recall that @xmath56 ) .
in all our calculations we have used polynomial expansions of @xmath24 and @xmath57 with absolute error not exceeding @xmath58 % .
thus , our calculation results are accurate to this order .
a finite - but - large @xmath48 regime is shown in the next two figures .
2 corresponds to the case when @xmath59 .
we see in fig .
2 that the real and imaginary parts of @xmath0 start to deviate from the newtonian fluid behavior at _ large _ frequencies . in fig .
3 solutions correspond to the case when @xmath60 ; thus we see how viscoelastic effects become pronounced already at _ low _ frequencies .
4 presents the behavior of @xmath54 $ ] and @xmath55 $ ] when @xmath61 .
as it can be seen from the graph , sharp resonances appear on the curves .
this feature can be explained by the fact that in this case @xmath48 is less than @xmath62 , which means that the system switched from the dissipative ( viscous ) regime to the elastic one .
5 shows the solutions when @xmath63 . in this case
we see more irregular behavior of @xmath54 $ ] and @xmath55 $ ] with a number of irregular spikes . the extreme non - newtonian ( elastic ) regime is studied in fig .
6 where we plot the solutions for the case when @xmath64 . in this case
notable change is that there are fewer but more pronounced spikes .
$ ] is close to zero for most of the frequencies , and only at certain frequencies we see sharp resonances . in this regime
the system acts a _ window _ for these frequencies .
another noteworthy observation is that , on one hand , as long as @xmath65 ( figs . 2 and 3 ) , @xmath54 $ ] is always greater than its initial value , i.e. @xmath54 > 1 $ ] , and for large frequencies it reaches a ceratin asymptotic value . on the other hand ,
when @xmath52 ( figs . 4 - 6 ) , we observe an overall decrease in @xmath54 $ ] with the increase in frequency , i.e. , @xmath54 < 1 $ ] for all @xmath1 s . in fig .
7 we have also studied the behavior of the dimensionless dynamic permeability @xmath66 as a function of @xmath67 , the dimensionless frequency defined as @xmath68 , for the case when @xmath69 . since @xmath52 ,
we observe sharp resonances in the dynamic permability at certain frequences , which can be explained by the non - newtonian behavior of the fluid .
in this paper we have studied non - newtonian effects on the dynamics of fluids in porous media by calculating a flow of maxwell fluid in a tube , which oscillates longitudinally and is subject to an oscillatory pressure gradient .
the present study investigates novelties brought about into the classic biot s theory [ 1 ] of propagation of elastic waves in a fluid - saturated porous solid by inclusion of non - newtonian effects .
we have used time fourier transform and converted the problem to the frequency domain .
we have calculated the dynamic permeability , thus modifying the work of del rio , lopez de haro , whitaker [ 4 ] by inclusion of the effect of longitunally oscillating tube walls .
we investigated how the function @xmath0 , which measures the deviation from poiseuille flow friction as a function of frequency parameter @xmath1 , is modified by non - newtonian effects .
our work thus provides a combined theory of non - newtonian flow in a longitudinally oscillating tube , which constitutes the basis for a realistic model of the effects of elastic waves in a fluid - saturated porous space .
the present analisys clearly demonstrates the existence of a transition from a dissipative to elastic regime ( as @xmath48 decreases ) , in which sharp enhancements of flow ( resonances ) occur .
\(a ) we studied modifications brought about by non - newtonian effects into biot s theory .
the investigation of the function @xmath0 is important for a number of applications , since @xmath0 uniquely determines the response of a realistic fluid - saturated porous medium to the elastic waves .
thus , determination of @xmath0 for non - newtonian ( maxwell ) fluid is necessary to guide e.g. , oil - field exploration applications .
\(b ) as we have seen from fig . 7 , non - newtonian effects cause substantial enhancements in dynamic permeability .
we were not able to determine in the literature what value of @xmath48 a natural crude oil would have .
however , as it can be seen from fig.7 for @xmath70 , we can get up to 60 times increase in permeability at certain resonant frequences .
lower @xmath48 s yield even more drastic emhancements . at any rate , we obtained an analytical expression for @xmath0 ( eq.(9 ) ) , which can provide the behavior of this function for any given @xmath48 . in the introduction section ,
the practical impact of the possibility of acoustic stimulation of oil reservoirs has been outlined .
this result clearly demonstrates that in the crude oil , that can be modelled as a maxwell fluid , there are certain resonant frequencies at which oil production can be increased significantly if the well is irradiated with elastic waves at these frequencies .
biot , j. acoust .
, * 28 * , 179 ( 1956 ) ; m.a .
biot , j. acoust .
, * 28 * , 168 ( 1956 ) i.a .
beresnev and p.a .
johnson , geophys .
, * 59 * , 1000 ( 1994 ) ; t. drake and i. beresnev , the american oil & gas reporter , september 1999 , p.101 c. chang , q.d .
nguyen , h.p .
ronningsen , j. non - newtonian fluid mech . , * 87 * , 127 ( 1999 )
williamson , k. walters , t.w .
bates , r.c .
coy and a.l .
milton , j. non - newtonian fluid mech .
, * 73 * , 115 ( 1997 ) ; g.a .
nunez , g.s .
ribeiro , m.s .
arney , j. feng and d.d .
joseph , j. rheol .
, * 38(5 ) * , 1251 ( 1994 ) ; l.t . wardhaugh and d.v .
boger , j. rheol .
, * 35(6 ) * , 1121 ( 1991 ) j.a .
del rio , m. lopez de haro and s. whitaker , phys .
, * e58 * , 6323 ( 1998 ) d.l .
johnson , j. koplik and r. dashen , j. fluid mech .
, * 176 * , 379 ( 1987 ) ; m .- y .
zhou and p. sheng , phys . rev . *
b39 * , 12027 ( 1989 ) ; m. avellaneda and s.t .
torquato , phys .
fluids a * 3 * , 2529 ( 1991 ) ; m. sahimi , rev . mod .
phys , * 65 * , 1393 ( 1993 ) ; p.sheng and m .- y .
zhou , phys .
lett . , * 61 * , 1591 ( 1998 )
|
we analyzed effects of elasticity on the dynamics of fluids in porous media by studying a flow of a maxwell fluid in a tube , which oscillates longitudinally and is subject to oscillatory pressure gradient .
the present study investigates novelties brought about into the classic biot s theory of propagation of elastic waves in a fluid - saturated porous solid by inclusion of non - newtonian effects that are important , for example , for hydrocarbons . using the time fourier transform and transforming the problem into the frequency domain
, we calculated : ( a ) the dynamic permeability and ( b ) the function @xmath0 that measures the deviation from poiseuille flow friction as a function of frequency parameter @xmath1 .
this provides a more complete theory of flow of maxwell fluid through the longitudinally oscillating cylindrical tube with the oscillating pressure gradient , which has important practical applications .
this study has clearly shown transition from dissipative to elastic regime in which sharp enhancements ( resonances ) of the flow are found .
|
it is well known that molecular clouds are magnetized with compressible magnetic turbulence determining most of their properties ( see elmegreen & falgarone 1996 , stutzki 2001 ) .
star formation ( see mckee & tan 2002 , elmegreen 2002 , pudritz 2001 ) , cloud chemistry ( see falgarone 1999 and references therein ) , shattering and coagulation of dust ( see lazarian & yan 2003 and references therein ) are examples of processes for which knowledge of turbulence is absolutely essential .
there are many excellent reviews that deal the theory of turbulent molecular clouds and numerical simulations ( see falgarone 1999 , vazquez - semadeni et al .
2000 , mac low & klessen 2003 ) .
this _ short _ review is focused on the recently uncovered basic properties of mhd turbulence and their implications for understanding of molecular clouds .
we also briefly deal with recovering of the 3d statistics of turbulent velocity from observations , which is a theoretical problem in itself .
a more detailed discussion of mhd turbulence can be found in the companion review by chandran , while observational aspects of turbulence in molecular clouds are discuss in the review by falgarone ( this volume ) . why do we expect molecular clouds to be turbulent ?
a fluid of viscosity @xmath0 becomes turbulent when the rate of viscous dissipation , which is @xmath1 at the energy injection scale @xmath2 , is much smaller than the energy transfer rate @xmath3 , where @xmath4 is the velocity dispersion at the scale @xmath2 .
the ratio of the two rates is the reynolds number @xmath5 . in general ,
when @xmath6 is larger than @xmath7 the system becomes turbulent .
chaotic structures develop gradually as @xmath6 increases , and those with @xmath8 are appreciably less chaotic than those with @xmath9 . observed features such as star forming clouds are very chaotic for @xmath10 .
this makes it difficult to simulate realistic turbulence .
the currently available 3d simulations containing 512 grid cells along each side can have @xmath6 up to @xmath11 and are limited by their grid sizes .
therefore , it is essential to find _ scaling laws _ " in order to extrapolate numerical calculations ( @xmath12 ) to real astrophysical fluids ( @xmath10 ) .
we show below that even with its limited resolution , numerics is a great tool for _ testing _ scaling laws .
kolmogorov theory provides a scaling law for _ incompressible _ _
non_-magnetized hydrodynamic turbulence ( kolmogorov 1941 ) .
this law provides a statistical relation between the relative velocity @xmath13 of fluid elements and their separation @xmath14 , namely , @xmath15 .
an equivalent description is to express spectrum @xmath16 as a function of wave number @xmath17 ( @xmath18 ) .
the two descriptions are related by @xmath19 .
the famous kolmogorov spectrum is @xmath20 .
the applications of kolmogorov theory range from engineering research to meteorology ( see monin & yaglom 1975 ) but its astrophysical applications are poorly justified and the application of the kolmogorov theory can lead to erroneous conclusions ( see reviews by lazarian et al . 2003 and lazarian & yan 2003 ) let us consider _ incompressible _ mhd turbulence first .
there have long been understanding that the mhd turbulence is anisotropic ( e.g. shebalin et al .
substantial progress has been achieved recently by goldreich & sridhar ( 1995 ; hereafter gs95 ) , who made an ingenious prediction regarding relative motions parallel and perpendicular to magnetic field * b * for incompressible mhd turbulence .
an important observation that leads to understanding of the gs95 scaling is that magnetic field can not prevent mixing motions of magnetic field lines if the motions are perpendicular to the magnetic field .
those motions will cause , however , waves that will propagate along magnetic field lines .
if that is the case , the time scale of the wave - like motions along the field , i.e. @xmath21 , ( @xmath22 is the characteristic size of the perturbation along the magnetic field and @xmath23 is the local alfven speed ) will be equal to the hydrodynamic time - scale , @xmath24 , where @xmath25 is the characteristic size of the perturbation perpendicular to the magnetic field .
the mixing motions are hydrodynamic - like .
they obey kolmogorov scaling , @xmath26 , because incompressible turbulence is assumed . combining the two relations above we can get the gs95 anisotropy , @xmath27 ( or @xmath28 in terms of wave - numbers ) . if we interpret @xmath22 as the eddy size in the direction of the local magnetic field ( larson 1981 ) for velocity and cloud size . ] and @xmath25 as that in the perpendicular directions , the relation implies that smaller eddies are more elongated .
gs95 predictions have been confirmed numerically ( cho & vishniac 2000 ; maron & goldreich 2001 ; cho , lazarian & vishniac 2002a , hereafter clv02a ; see also clv03a ) ; they are in good agreement with observed and inferred astrophysical spectra ( see clv03a ) .
however , the gs95 model considered incompressible mhd , but the media in molecular clouds is _
highly compressible_. does any part of gs95 model survives ?
literature on the properties of compressible mhd is very rich ( see reviews by pouquet 1999 , cho & lazarian 2003 and references therein ) .
higdon ( 1984 ) theoretically studied density fluctuations in the interstellar mhd turbulence .
matthaeus & brown ( 1988 ) studied nearly incompressible mhd at low mach number and zank & matthaeus ( 1993 ) extended it . in an important paper matthaeus et al .
( 1996 ) numerically explored anisotropy of compressible mhd turbulence .
however , those papers do not provide universal scalings of the gs95 type .
the complexity of the compressible magnetized turbulence with magnetic field made some researchers believe that the phenomenon is too complex to expect any universal scalings for molecular cloud research .
high coupling of compressible and incompressible motions is often quoted to justify this point of view .
below we shall provide arguments that are suggestive that the fundamentals of compressible mhd can be understood and successfully applied to molecular clouds . in molecular clouds
the regular magnetic field is comparable with the fluctuating one .
therefore for most part of our discussion , we shall discuss results obtained for @xmath29 , where @xmath30 is the r.m.s .
strength of the random magnetic field .
turbulent support of molecular clouds ( see review by mckee 1999 ) critically depends on the rate of turbulence decay . for a long time magnetic fields were thought to be the means of decreasing dissipation of turbulence .
numerical calculations by mac low et al .
( 1998 ) and stone et al . (
1998 ) indicated that compressible mhd turbulence decays as fast as the hydrodynamic turbulence .
this gives rise to a belief that it is the compressibility that is responsible for the rapid decay of mhd turbulence .
this point of view has been recently subjected to scrutiny in cho & lazarian ( 2002 , 2003a , henceforth cl02 and cl03 , respectively ) . in these papers
a technique of separating different mhd modes was developed and used ( see fig . 1 ) .
this allowed us to follow how the energy was redistributed between these modes .
how is this idealized incompressible model related to the actual turbulence in molecular clouds ?
compressible mhd turbulence is a highly non - linear phenomenon and it has been thought that different types of perturbations or modes ( alfven , slow and fast ) in compressible media are strongly coupled .
nevertheless , one may question whether this is true .
a remarkable feature of the gs95 model is that alfven perturbations cascade to small scales over just one wave period , while the other non - linear interactions require more time .
therefore one might expect that the non - linear interactions with other types of waves should affect alfvenic cascade only marginally .
moreover , since the alfven waves are incompressible , the properties of the corresponding cascade may not depend on the sonic mach number .
the generation of compressible motions ( i.e. _ radial _ components in fourier space ) from alfvenic turbulence is a measure of mode coupling .
how much energy in compressible motions is drained from alfvenic cascade ? according to closure calculations ( bertoglio , bataille , & marion 2001 ; see also zank & matthaeus 1993 ) , the energy in compressible modes in _ hydrodynamic _ turbulence scales as @xmath31 if @xmath32 .
cl03 conjectured that this relation can be extended to mhd turbulence if , instead of @xmath33 , we use @xmath34 .
( hereinafter , we define @xmath35 , where @xmath36 is the mean magnetic field strength . )
however , since the alfven modes are anisotropic , this formula may require an additional factor .
the compressible modes are generated inside the so - called goldreich - sridhar cone , which takes up @xmath37 of the wave vector space .
the ratio of compressible to alfvenic energy inside this cone is the ratio given above .
if the generated fast modes become isotropic ( see below ) , the diffusion or , `` isotropization '' of the fast wave energy in the wave vector space increase their energy by a factor of @xmath38 .
this results in @xmath39^{-1 } , \label{eq_high2}\ ] ] where @xmath40 and @xmath41 are energy of compressible and alfven modes , respectively .
( [ eq_high2 ] ) suggests that the drain of energy from alfvenic cascade is marginal when the amplitudes of perturbations are weak , i.e. @xmath42
. results of calculations shown in fig .
2 support the theoretical predictions .
we may summarize this issue in the following way . for the incompressible motions to decay fast
, there is no requirement of coupling with compressible motions follows @xmath43 which can be understood if we account for the fact that the energy is being injected at the scale smaller than the scale of the system .
therefore some energy originally diffuses to larger scales through the inverse cascade .
our calculations ( cho & lazarian , unpublished ) , stimulated by illuminating discussions with chris mckee , show that if this energy transfer is artificially prevented by injecting the energy on the scale of the computational box , the scaling of @xmath44 becomes closer to @xmath45 . ] .
the marginal coupling of the compressible and incompressible modes allows to study modes separately .
is it feasible to obtain scaling relations for the compressible mhd turbulence ?
some hints about effects of compressibility can be inferred from the gs95 seminal paper .
more discussion was presented in lithwick & goldreich ( 2001 ) , which deals with electron density fluctuations in the gas pressure dominated plasma , i.e. in high @xmath46 regime ( @xmath47 ) .
the incompressible regime corresponds to @xmath48 , so it is natural to expect that for @xmath49 the gs95 picture would persist .
lithwick & goldreich ( 2001 ) also speculated that for low @xmath46 plasmas the gs95 scaling of slow modes may be applicable . a detailed study of compressible mode scalings is given in cl02 and cl03 .
our considerations above about the mode coupling can guide us to predict the mode scaling . indeed ,
if alfven cascade evolves on its own , it is natural to assume that slow modes exhibit the gs95 scaling . indeed , slow modes in gas pressure dominated environment ( high @xmath46 plasmas ) are similar to the pseudo - alfven modes in incompressible regime ( see gs95 ; lithwick & goldreich 2001 ) . the latter modes do follow the gs95 scaling . in magnetic pressure dominated environments ( low @xmath46 plasmas ) ,
slow modes are density perturbations propagating with the sound speed @xmath50 parallel to the mean magnetic field .
those perturbations are essentially static for @xmath51 .
therefore alfvenic turbulence is expected to mix density perturbations as if they were passive scalar .
this also induces the gs95 spectrum .
the fast waves in low @xmath46 regime propagate at @xmath52 irrespectively of the magnetic field direction . in high @xmath46 regime ,
the properties of fast modes are similar , but the propagation speed is the sound speed @xmath50 . thus the mixing motions induced by alfven waves should marginally affect the fast wave cascade .
it is expected to be analogous to the acoustic wave cascade and hence be isotropic .
results of numerical calculations from cho & lazarian ( cl03 ) for magnetically dominated media similar to that in molecular clouds are shown in fig .
3 . they support theoretical considerations above .
why would we care about those scalings ?
how wrong is it to use kolmogorov scalings instead ?
dynamics , chemistry and physics of molecular clouds ( see falgarone 1999 ) presents a complex of problems for which the exact scalings may be required to a different degree .
if we talk about dynamics of interstellar dust or propagation of cosmic rays , one _ must _ account for the actual scalings and couplings of different modes ( see reviews by lazarian & yan 2003 , lazarian et al .
there are other problems , e.g. turbulent heat transport where the exact scaling of modes seems to be less important ( cho et al .
our considerations above assume that the gas is fully ionized .
molecular clouds are partially ionized .
what is the consequence of this ?
an obvious effect of neutrals is that they do not follow magnetic field lines and thus produce viscosity
. in hydrodynamic turbulence viscosity sets a minimal scale for motion , with an exponential suppression of motion on smaller scales .
below the viscous cutoff the kinetic energy contained in a wavenumber band is dissipated at that scale , instead of being transferred to smaller scales .
this means the end of the hydrodynamic cascade , but in mhd turbulence this is not the end of magnetic structure evolution . for viscosity
much larger than resistivity , there will be a broad range of scales where viscosity is important but resistivity is not . on these scales
magnetic field structures will be created by the shear from non - damped turbulent motions , which amounts essentially to the shear from the smallest undamped scales . indeed , this new regime of turbulence has been discovered ( see fig .
4 ) ! a theoretical model for this new regime can be found in lazarian , vishniac , & cho ( 2003 ; hereafter lvc03 ) .
it explains the spectrum @xmath53 as a cascade of magnetic energy to small scales under the influence of shear at the marginally damped scales .
the spectrum is similar to that of the viscous - convective range of passive scalar in hydrodynamic turbulence ( see , for example , batchelor 1959 ; lesieur 1990 ) , although the study in lazarian , vishniac , & cho ( lvc03 ) suggests that the physical origin of it is different . a study confirming that the @xmath54 spectrum is not a bottleneck effect is presented in cho , lazarian , & vishniac ( 2003b ) .
the mechanism is based on the solenoidal motions and therefore the compressibility should not alter the physics of this regime of turbulence . for compressible simulations ( see fig .
4 ) the inertial range is much smaller due to numerical reasons , but it is clear that the viscosity - damped regime of mhd turbulence persists . the magnetic fluctuations , however , compress the gas and thus cause fluctuations in density .
these density fluctuations may have important and yet unexplored consequences for the small scale structure of the molecular clouds .
there are indications that small - scale structure is indeed present in molecular gas ( marscher , moore , & bania 1993 ) as well as in other partially ionized phases of the ism ( see heiles 1997 ) .
those structures can be produced by turbulence with a spectrum substantially more shallow than the kolmogorov one , e.g. with the spectrum @xmath53 ( see deshpande 2000 ) .
below the viscous scale the fluctuations of magnetic field obey the damped regime shown in fig .
[ fig_viscous]b and produce density fluctuations . for typical cold neutral medium gas ( see draine & lazarian 1998 ) , these fluctuations can be as large as @xmath55 au and less for denser gas .
an important prediction in lvc03 is that the intermittency of magnetic structures increases with the decrease of the scale .
this prediction was confirmed by numerical simulations in cho , lazarian & vishniac ( 2003 ) , which showed that the filling factor of magnetic field was decreasing with the increase of the wavenumber .
does the effect of neutrals amounts only to emergence of the viscosity - damped regime of mhd turbulence ?
the answer to this question is negative .
partial ionization provides the whole range of new interesting effects .
some of them are studied in lvc03 . here
we mention those .
first of all , it is clear that whether ions and neutrals act as one fluid in molecular clouds depends on whether the eddy turnover rate @xmath56 is longer or shorter than the rate @xmath57 of neutral - ion collisions . if @xmath58 , neutrals decouple from ions and develop _ hydrodynamic _ kolmogorov - type cascade .
indeed , the damping rate for those hydrodynamic motions @xmath57 and below the decoupling scale the hydrodynamic motions evolve without much hindrance from magnetic field . magnetic fields with entrained ions develop the viscosity - damped mhd cascade until ion - neutral collisional rate gets longer than the dynamical rate of the intermittent magnetic structures . after that the turbulence reverts to its normal mhd cascade which involves only ions . if @xmath59 up to the scale at which neutral viscosity damps turbulent motions , the viscosity - damped regime emerges at the scale where kinetic energy associated with turbulent eddies is dissipated .
similarly to the earlier case when the when ion - neutral collisions get insufficient to preserve pressure confinement of the small scale magnetic filaments , outbursts of ordinary ionic mhd turbulence will take place .
the turbulence will be intermittent both in time and space because of the disparity of time scales at which turbulence evolves in the viscosity - damped and free ionic mhd regimes .
we plan to test those predictions with a two fluid mhd code .
substantial advances in understanding of mhd turbulence in ordinary and viscosity - damped regime make it most important to test the correspondence of theoretical expectations and observational reality ( see review by ostriker 2003 ) .
important advances in obtaining high resolution spectral data ( see falgarone et al .
2000 ) make the testing feasible .
the turbulence spectrum characterizes the distribution of energy over spatial scales and reflects the processes of dissipation and energy injection . the gs95 model would also gives the kolmogorov spectrum of alfvenic motions if averaging over @xmath60 directions is performed .
this stems from the fact that for the inertial range @xmath61 and therefore @xmath62 .
it is known that studies of stochastic density provide only indirect insight into turbulence . a stationary ( not evolving ) density distribution is indistinguishable from the result of active turbulence . however , doppler shifted spectral lines carry information about turbulent velocity .
the problem is that both velocity and density fluctuations contribute to the observed fluctuations .
therefore numerous attempts to study turbulence in diffuse interstellar medium and molecular clouds using emission lines ( see munch 1958 , odell 1986 , odell & castaneda 1987 , miesch & bally 1994 , reviews by scalo 1987 , falgarone 1999 ) were facing uncertainty of what quantity was actually measured .
studies of the velocity field have been attempted at different times with velocity centroids ( e.g. munch 1958 , miesch , scalo , & bally 1999 , ossenkopf & mac low 2002 , miville - deschenes et al . 2003 ) .
however , it has long been realized that the centroids are affected , in general , by both velocity and density fluctuations ( see stenholm 1989 ) .
a criterion of when centroids indeed reflect the velocity statistics was obtained in lazarian & esquivel ( 2003 , henceforth le03 ) . without this testing , it is not clear _ a priori _ what is actually being measured .
an important recent development is a quantitative description of spectral data from turbulent media obtained in lazarian & pogosyan ( 2000 , 2003 henceforth lp00 and lp03 , respectively ) , where the fluctuations of intensity in channel maps were related to the statistics of velocity and density .
lp00 introduced a new technique that was termed velocity channel analysis ( henceforth vca ) . within vca the separation of velocity and density contributions
is obtained by changing the thickness of the analyzed slice of the position - position- velocity ( ppv ) data cube .
the vca was successfully tested numerically in lazarian et al .
( 2001 ) and esquivel et al .
( 2003 ) and applied to the small magellanic cloud ( stanimirovic & lazarian 2002 ) .
lp03 accounts for the absorption in the turbulent media .
thus it substantially extends the range of spectral lines that can be used for turbulence studies .
note that the importance of lp00 and lp03 studies is not limited to the development of particular toolkit of how to interpret channel maps .
these works provide theoretical description that can be used within different techniques .
for instance , lp03 establishes limitations on turbulence spectra that can be recovered by another promising technique , namely , spectral correlation function ( see rosolowsky et al .
1999 ) .
other important tools have also been studied recently . principal component analysis ( pca ) of the emission data ( heyer & schloerb 1997 , brunt & heyer 2002ab ) provide a new promissing way to characterise observations .
recent testing showed that it provides statistics different from power spectra ( brunt et al .
wavelet analysis ( see gill & henriksen 1990 ) , spectral correlation functions ( see rosolowsky et al . 1999 ,
padoan , rosolowsky & goodman 2001 ) , genus analysis ( see lazarian , pogosyan & esquivel 2002 ) are other statistical tools . they can provide statistics and insight complementary to power spectra .
synergy of different techniques should provide the necessary insight into turbulence and enable the comparison of observational statistics with theoretical expectations .
there are whole classes of processes for which we are not sure even about the sign of the effect , for instance , whether turbulence supports or compresses molecular clouds . or consider imbalanced turbulence , i.e. the turbulence where the flow of energy in one direction is larger than the flow of energy in the opposite direction .
this situation is typical for interstellar medium with its localized sources of energy .
clv02a speculated that imbalanced turbulence can propagate over larger distances and feed energy to clouds without star formation .
to what degree does this process get modified in the presence of compressibility ?
alfvenic modes in imbalanced turbulence live longer and the interaction between density fluctuations and the alfven mode becomes more important . in the field of observational studies
, the situation looks remarkably promising . with high resolution surveys available ( see falgarone et al .
2000 ) , studies of turbulence statistics should at last become a mainstream research .
the prospects of studies of turbulent velocity are most encouraging . indeed , for the first time , one has understanding of how to get statistics of velocity and density from channel maps ( lp00 , lp03 ) , and when the velocity centroids reflect the statistics of velocity ( le03 ) .
moreover , a quantitative description of the statistics of the spectral line data cubes ( see lp00 , lp03 ) allows us to devise new techniques of turbulence studies .
apart from testing of the particular scaling laws , this research should identify sources and injection scales of the turbulence ( see the companion review by falgarone et al .
1998 ) . is turbulence in molecular clouds a part of a large scale ism cascade ( see armstrong et al .
how does the share of the energy within compressible versus incompressible motions vary within the galactic disk ?
there are examples of questions that can be answered in future .
while this review deals mostly with power spectra , higher order statistics will be widely used in the future .
recent numerical research that employed higher order statistics ( muller & biskamp 2000 , clv02a , boldyrev et al .
2002 , cho , lazarian & vishniac 2003b ) showed it to be a promising tool .
for instance , the distinction between the old iroshnikov - kraichnan and the gs95 model is difficult to catch using power spectra with a limited inertial range , but is quite apparent for fourth order statistics .
the difference in physical consequences of whether the turbulence dissipates in shocks or in intermittent vortices may be very substantial .
our discussion in 2 suggests that incompressible motions do dissipate via vortices .
the scaling of their intermittency with the reynolds number is still to be established . in the meantime , obtaining higher order statistics with spectral line observations is a challenging problem .
higher order statistics obtained from observational data were reported for observed velocity in falgarone et al .
( 1994 ) and for density in padoan et al .
according to falgarone & puget ( 1995 ) and falgarone et al .
( 1995 ) ( see also review by falgarone ) , the intermittency in vorticity distribution can result in the outbursts of localized dissipation that make tiny regions within cold diffuse clouds chemically active .
the testing of those ideas is done in pety & falgarone ( 2000 ) , where synthetic maps obtained using _ hydrodynamic _ simulations were analyzed . since then
, more results indicating the hydrodynamic simulations may to some degree reflect the physics of mhd turbulence have emerged .
first of all , clv02a found out that the fluid motions perpendicular to @xmath63 are _ identical _ to hydrodynamic motions .
moreover , for low ionization the turbulence in neutrals and ions decouple with turbulence in neutrals forming a hydrodynamic cascade ( see 5 ) .
studies of the statistics of magnetic field in molecular clouds is the next challenging problem ( see crutcher , heiles & troland 2003 , ostriker 2003 ) .
better understanding of grain alignment ( see review by lazarian 2003 and references therein ) allows to better identify variations of polarization with the variations of magnetic field .
however , this important research has not gained sufficient momentum yet .
1 . understanding of molecular clouds requires understanding of the basics of mhd turbulence .
mhd turbulence is not a mess . scaling relations for its modes
have been established recently .
. fast decay of mhd turbulence is not due to strong coupling of compressible and incompressible motions .
the transfer of energy from alfven to compressible modes is small .
the alfven mode develops on its own and decays fast .
doppler shifts imprinted in spectra lines provide an excellent way of testing theoretical expectations .
advances in understanding how this information can be extracted from spectroscopic data allow this . + * acknowledgments * _ a.l .
acknowledges the support of nsf grant ast-0125544_.
|
recent advances in understanding of the basic properties of compressible magnetohydrodynamic ( mhd ) turbulence call for revisions of some of the generally accepted concepts . first , mhd turbulence is not so messy as it is usually believed .
in fact , the notion of strong non - linear coupling of compressible and incompressible motions is not tenable .
alfven , slow and fast modes of mhd turbulence follow their own cascades and exhibit degrees of anisotropy consistent with theoretical expectations .
second , the fast decay of turbulence is not related to the compressibility of fluid . rates of decay of compressible and incompressible motions are very similar .
third , viscosity by neutrals does not suppress mhd turbulence in a partially ionized gas .
instead , mhd turbulence develops magnetic cascade at scales below the scale at which neutrals damp ordinary hydrodynamic motions .
the implications of those changes of mhd turbulence paradigm for molecular clouds require further studies .
those studies can benefit from testing of theoretical predictions using new statistical techniques that utilize spectroscopic data .
we briefly discuss advances in development of tools using which the statistics of turbulent velocity can be recovered from observations .
|
0.5 cm it is remarkable that different data sets of complementary nature such as supernovae , baryon oscillations , galaxy clustering , microwave back ground and weak lensing all taken together strongly support the late time acceleration of universe . in the standard lore , one assumes that the history of universe is described by the general relativity ( gr ) .
the late time acceleration can easily be captured in this frame work by introducing a scalar field with large negative pressure known as _ dark energy _ @xcite . in view of the fine tuning problem ,
the scalar field models , specially those with tracker like solutions , are more attractive compared to the models based on cosmological constant . at present , observations are not in a position to reject or to establish the dark energy metamorphosis . a host of scalar field models have been investigated in the literature .
the scalar field models can fit the data but lack the predictive power .
it then becomes important to seek the support of these models from a fundamental theory of high energy physics .
0.5 cm one can question the standard lore on fundamental grounds .
we know that gravity is modified at small distance scales ; it is quite possible that it is modified at large scales too where it has never been confronted with observations directly .
it is therefore perfectly legitimate to investigate the possibility of late time acceleration due to modification of einstein - hilbert action .
it is tempting to study the string curvature corrections to einstein gravity amongst which the gauss - bonnet correction enjoys special status .
a large number of papers are devoted to the cosmological implications of string curvature corrected gravity@xcite .
these models suffer from several problems .
most of these models do not include tracker like solution and those which do are heavily constrained by the thermal history of universe .
for instance , the gauss - bonnet theory with dynamical dilaton might cause transition from matter scaling regime to late time acceleration allowing to alleviate the fine tuning and coincidence problems . however , it is difficult to reconcile this model with nucleosynthesis@xcite .
another possibility of large scale modification is provided by non - locally corrected gravity which typically involves inverse of dalembertian of ricci scalar .
the non - local construct might mimic dark energy ; the model poses technical difficulties and there has been a little progress in this direction@xcite .
the large scale modification may also arise in extra dimensional theories like dgp model which contains self accelerating brane .
apart from the theoretical problems , this model is heavily constrained by observation
. 0.5 cm on purely phenomenological grounds , one could seek a modification of einstein gravity by replacing the ricci scalar by @xmath0 .
the @xmath0 gravity models have been extensively investigated in past five years@xcite .
the @xmath0 gravity theories giving rise to cosmological constant in low curvature regime are plagued with instabilities and on observational grounds they are not distinguished from cosmological constant .
the recently introduced models of @xmath0 gravity by hu - sawicki and starobinsky ( referred as hss models hereafter ) with disappearing cosmological constant@xcite have given rise to new hopes for a viable cosmological model within the framework of modified gravity ( @xmath0 gravity model with similar properties is proposed in ref.@xcite ) .
these models contain minkowski space time as a solution in the low curvature regime which is an unstable solution . in high curvature regime
these models reduce to cosmological constant .
both the first and the second derivatives of @xmath0 with respect to r are positive . the positivity of the first derivative ensures that the scalar degree of freedom , a characteristic of any f(r ) theory , is not tachyonic where as the positivity of second derivative tells us that graviton is not ghost thereby guaranteing the stability . in starobinsky parametrization@xcite ,
@xmath0 is given by , @xmath4 $ ] .
the hss models can evade solar physics constraints provided that the model parameters are chosen properly . an important observation has recently been made by by appleby @xmath1 battye and forolov@xcite ( see also @xcite )
. the minimum of scalaron potential which corresponds to dark energy can be very near to @xmath5 or equivalently @xmath6 . as pointed out in ref.@xcite
, the minimum should be near the origin for solar constraints to be evaded .
hence , it becomes most likely that we hit the singularity if the parameters are not fine tuned . 0.5 cm in order to check whether the @xmath0 gravity theory is cosmological viable or
not , it is necessary that this theory must be compatible with the observations . in this work
, we study the starobinsky model using the data from the recent observations which include @xmath2 , hubble parameter at various red - shifts and the baryon acoustic oscillation ( bao ) peak from sloan digital sky survey ( sdss ) .
0.5 cm this paper is organised as follows . in section
ii , we describe the general properties of the model highlighting the fine tuning problem . the friedmann equation and the special cases of trace equation
is studied in section iii .
the cosmological constraints from the recent observations are described in section iv .
finally , section v contains the results and discussions .
0.5 cm in this section we shall revisit f(r ) models with disappearing cosmological constant .
these models have potential capability of being distinguished from @xmath3cdm and could lead to a viable cosmological model .
however , even at the background level this class of models are fine tuned . in
what follows we shall explicitly bring out these features specializing to starobinsky parametrization .
0.5 cm the action of @xmath0 gravity is given by@xcite , @xmath7 \sqrt{-g } \quad d^4 x,\ ] ] which leads to the following equation of motion @xmath8 here prime @xmath9 denotes the derivatives with respect to @xmath10 .
the f(r ) gravity theories apart from a spin-2 object necessarily contain a scalar degree of freedom .
taking trace of eq.([eq : freqn ] ) gives the evolution equation for the scalar degree of freedom , @xmath11 it would be convenient to define scalar function @xmath12 as @xmath13 which is expressed through ricci scalar once f(r ) is specified . 0.5
cm we can write the trace equation ( equation @xmath14 ) in the term of @xmath15 and @xmath16 as @xmath17 0.5 cm the potential can be evaluated using the following relation @xmath18 0.5 cm recently hu , sawicki and starobinsky proposed a functional form of @xmath0 with the desirable properties of a viable model @xcite . in this paper
we shall consider the model in starobinsky parametrization : @xmath19.\ ] ] here n and @xmath20 are greater than zero .
and @xmath21 is of the order of presently observed cosmological constant , @xmath22 .
the properties of this model can be summarized as follows : 1 . in the absence of matter
@xmath23 is always a solution , @xmath24 . however , as @xmath25 , flat space time is unstable .
2 . for @xmath26 , @xmath27 .
the high - curvature value of the effective cosmological constant is @xmath28 .
the stability conditions for the adopted model are @xmath29 0.5 cm in the starobinsky model the scalar field @xmath12 is given by @xmath30 0.5 cm we can compute @xmath31 for a given value of @xmath32 . in case of @xmath33
, we have @xmath34 where @xmath35 . in case and of @xmath36
we have , @xmath37 0.5 cm in the frw background , the trace equation ( equation @xmath14 ) can be rewritten in the convenient form @xmath38 versus @xmath12 for @xmath39 and @xmath40 .
the red spot marks the initial condition for evolution of @xmath12.,width=377 ] 0.5 cm the time - time component of the equation of motion @xmath41 gives the hubble equation @xmath42 we recover einstein gravity in the limit @xmath43 .
the simple picture of dynamics which appears here is the following : above infrared modification scale @xmath44 , the expansion rate is set by the matter density and once the local curvature falls below @xmath45 the expansion rate gets effect of gravity modification . 0.5 cm for pressure less dust
, the effective potential has an extremum at @xmath46 for a viable late time cosmology , the field should be evolving near the minimum of the effective potential .
the finite time singularity inherent in the class of models under consideration severely constraints dynamics of the field . in the case @xmath36 and @xmath47 .
the initial condition is as marked by the red spot in figure [ fig : n2l1p2v].,width=377 ] the effective potential has minimum which depends upon @xmath32 and @xmath20 . for generic values of the parameters ,
the minimum of the potential is close to @xmath48 , corresponding to infinitely large curvature .
thus , while the field is evolving towards minimum , it can easily oscillate to a singular point .
however , depending upon the values of parameters , we can choose a finite range of initial conditions for which scalar field @xmath12 can evolve to the minimum of the potential without hitting the singularity .
for instance , let us begin with the initial conditions as shown in figure @xmath49 given by @xmath50 ( @xmath51 ) .
we choose @xmath52 then @xmath53 becomes @xmath54 for a given value of @xmath10 ( see eq .
@xmath55 ) .
the minimum of the potential for @xmath56 and @xmath57 corresponds to @xmath58 or @xmath59 which is indeed close to singularity .
the numerical results are shown in figures [ fig : n2l1p2v ] and [ fig : n2l1p2r ] . versus @xmath20 for different values of @xmath32 .
with increase in @xmath32 , @xmath60 moves towards zero ( corresponding to infinitely large value of @xmath10 ) for smaller values of @xmath20 . the curves from bottom to top correspond to @xmath61 respectively.,width=377 ] 0.5 cm we find that the range of initial conditions allowed for the evolution of @xmath12 to the minimum without hitting singularity shrinks as the numerical values of parameters @xmath32 and @xmath20 increase .
this is related to the fact that for larger values of @xmath32 and @xmath20 , the minimum fast moves towards @xmath5 , see figure [ fig : phmin ] .
0.5 cm for example , in the case of @xmath62 , we find that @xmath63 which is close to zero .
the initial value of @xmath12 which is quite near to minimum also gives the divergence of @xmath64 and @xmath65 at the finite time ( see fig .
[ fig : n1l4r ] ) . 0.5 cm on the other hand , one can see from the figure [ fig : phmin ] that @xmath60 also shifts to zero when @xmath32 is increased . for @xmath66 and @xmath67
we have @xmath68 .
the range of initial conditions for which the scalar field can evolve to the minimum is very small . 0.5
cm we know that in case of large curvature , the quantum effects become important leading to higher curvature corrections . keeping this in mind ,
let us consider the modification of starobinsky s model , @xmath70 then @xmath12 becomes @xmath71.\ ] ] 0.5 cm when @xmath72 is large the first term which comes from @xmath69 dominates . in this case
, the curvature singularity @xmath73 corresponds to @xmath74 .
hence , in this modification , the minimum of the effective potential is separated from the curvature singularity by the infinite distance in the @xmath75 plane .
0.5 cm for @xmath36 , @xmath12 and @xmath76 are given by @xmath77 0.5 cm for @xmath78 and @xmath79 , we have a large range of the initial condition for which the scalar field evolves to the minimum of the potential . though the introduction of @xmath80 term formally allows to avoid the singularity but can not alleviate the fine tuning problem as the minimum should be brought near to the origin to respect the solar constraints .
last but not the least one could go beyond the approximation ( see equation ( [ extm ] ) ) by iterating the trace equation and computing the corrections to @xmath10 given by equation ( [ extm ] ) . as pointed by starobinsky@xcite , such a correction might become large in the past .
this may spoil the thermal history and thus needs to be fine tuned .
the aforesaid discussion makes it clear that hss models are indeed fine tuned and hence very delicate . for @xmath33 and @xmath81.,width=377 ] , @xmath82 and @xmath83 in presence of @xmath80 correction .
the minimum of the effective potential in this case is located at @xmath84.,width=377 ]
the friedmann equation can be casted in the form , @xmath85 where @xmath86 . to parameterise the system in a more convenient form
we define : @xmath87 since @xmath88 we have @xmath89 .
the @xmath90 and @xmath91 for the starobinsky model are of the form @xmath92 , \nonumber \\
f ' & = & 1-\frac{2n\lambda}{x } \left(1+\frac{1}{x^2}\right)^{-(1+n ) } , \\ h_0 ^ 2 f '' & = & \frac{2n\lambda}{3\omega_{\lambda } } \left[-1+\frac{(2n+1)}{x^2}\right ] \left(1+\frac{1}{x^2}\right)^{-(2+n)}. \nonumber\end{aligned}\ ] ] 0.5 cm the trace equation can be re - written as @xmath93 = \frac{\omega_{m}}{\omega_{\lambda}}.\ ] ] here @xmath94 is the matter density of the universe at a given redshift .
@xmath95 , is present fractional matter density and @xmath96 , is the present fractional density of the vacuum . 0.5 cm the friedmann equation ( equation ( [ nfreid ] ) )
can now be written as @xmath97 where @xmath0 and its derivatives are as given above .
note that we can recover the usual friedmann equation for @xmath98 , the case for which our action reduces to einstein hilbert action . 0.5 cm in this paper we test the viability of @xmath0 cosmology .
the friedmann equation evaluated at @xmath99 imposes a constraint on the free parameters @xmath32 and @xmath20 . 0.5 cm to check the compatibility of starobinsky model with observations we need to first obtain the hubble parameter @xmath100 , as a function of redshift @xmath101 .
the hubble equation in its present form depends on the curvature @xmath10 , through @xmath102 and its derivatives .
therefore , we need to solve the trace equation ( equation ( [ treqn ] ) ) , relating curvature with redshift . in order to study the behavior of trace equation , which does not admit a simple solution for general @xmath32 , we first analyse its quantitative behavior for large values of @xmath101 , corresponding to @xmath103 .
therefore , the trace equation can be simplified considerably and in the leading order we recover @xmath104 similarly , the friedmann equation in the leading order gives @xmath105 thus at large redshift we recover the standard @xmath3cdm cosmology for @xmath82 , independently of @xmath32 ( see figure [ fig : hzn1 ] ) .
0.5 cm there is no general solution for the trace equation and therefore we analyse this model for fixed values of @xmath32 .
it was shown by capozziello and tsujikawa @xcite that we need @xmath106 to satisfy solar system tests . in this work
we consider model with @xmath33 and @xmath107 .
the trace equation in this case simplifies to a quintic equation @xmath108 where @xmath109 .
this equation can not be solved analytically for general @xmath110 in terms of other parameters . since @xmath103 , we approximate this equation with a quartic equation . this approximation is found to be sufficiently accurate for @xmath103 ( see fig . [
fig : fitn12 ] ) . we need to solve for x ( @xmath111 ) using the trace equation .
we have two quartic equations @xmath112 and @xmath113 for n = 1 and 2 case , respectively .
we shall solve the quatric equation with @xmath114 and @xmath115 .
let us first consider the case of @xmath116 and look for the roots of @xmath117 at @xmath99 .
we numerically analyse the eq .
26 and find that for generic values of @xmath20 ( @xmath118 ) , two of the four roots of quatric equation are always imaginary .
the friedman equation puts a constraints on the values of @xmath20 such that @xmath119 for @xmath120 , see eq . 23 .
this condition fixes the value of @xmath20 to @xmath107 for one of the roots , the other root corresponds to @xmath121 .
we should now check the viability of the roots by invoking the non - zero values f the redshift . as z increases ,
i.e. , we move to past , @xmath122 should decrease as r should increase during a viable cosmic evolution which happens in case of @xmath82 .
however , for the root corresponding to @xmath123 , x increases in the past and @xmath110 quickly goes beyond its normal range @xmath124 and we need not consider this root at all .
therefore we have only one root obtained at @xmath125 which produces the correct final asymptotic state as a de sitter model when z goes to -1 ( see figure [ fig : eos ] ) . in this figure
we plot the effective equation of state with redshift : @xmath126 once @xmath20 is fixed there is no other free parameter in the theory .
the hubble parameter can now be plotted for the best fit value of @xmath20 .
the trace equation in this case is a seventh degree polynomial in @xmath110 @xmath127 again we can approximate this equation by a quartic equation in @xmath110 using the fact that @xmath103 ( see fig.[fig : fitn22 ] ) . in order to calculate the roots of this equation
, we followed the same procedure as described for @xmath128 case . here
also @xmath82 reproduces the correct late time behavior ( see figure [ fig : eos ] ) .
in this paper , we check the compatibility of the model with the @xmath2 data and the measurements of baryon acoustic oscillation peak .
we find the values of @xmath129 to see how well the @xmath0 model with @xmath130 and @xmath131 accommodate the observations . here , for a given model , we calculate @xmath129 using h(z ) data , bao measurement and for the joint data ( h(z ) + bao ) .
simon , verde and jimenez ( 2005 ) employed differential ages of passively evolving galaxies to get the hubble parameter as a function of redshift , @xmath2 @xcite .
the nine data points of h(z ) with @xmath132 have been obtained by using absolute ages of 32 galaxies taken from the gemini deep deep survey ( gdds ) and the archival data . 0.5 cm for the adopted @xmath0 model , we calculate the values of @xmath129 for @xmath133 and @xmath36 .
for this we define @xmath134 0.5 cm here @xmath135 is the present day value of the hubble constant , @xmath136 is the expected value of the hubble constant in the @xmath0 cosmology at redshift @xmath137 for a particular @xmath32 and @xmath135 .
@xmath138 is the observed value and @xmath139 is the corresponding @xmath140 uncertainty in the measurement .
the sum is over all observed data points ( nine in number )
. 0.5 cm as the value of @xmath129 is highly sensitive to the value of @xmath135 , we marginalize over @xmath135 to obtain the modified @xmath129 . for this
we define the likelihood function as : @xmath141 here @xmath142 is the prior probability function for @xmath135 which is gaussian : @xmath143,\ ] ] with @xmath144 as the value of @xmath135 ( and @xmath145 is the error in it ) as suggested by independent observations . in this paper , we also study the effect of different priors on the result .
we use two set of priors : 0.3 cm @xmath146 km / s / mpc , as obtained from the median statistics analysis of 461 measurements of @xmath135 @xcite .
0.3 cm @xmath147 km / s / mpc , as suggested by the chandra x - ray observatory results @xcite . 0.4 cm 0.3 cm we observe that the value of the modified @xmath148 is sensitive to the choice of the prior . before recombination ,
the universe was in the completely ionized state .
the cosmological perturbations in the relativistic primordial baryon - photon plasma produced acoustic oscillations .
these oscillations were imprinted in the form of peaks in the late time power spectrum of non - relativistic matter .
this acoustic peak which is predicted at the measured scale of @xmath149 is detected in the large scale correlation function of 46478 sample of luminous red galaxies in the sloan digital sky survey@xcite .
it is described by a dimensionless parameter @xmath150 , @xmath151^{1/3 } % = 0.469 \pm 0.017\ ] ] where @xmath152 , @xmath153 is the dimensionless comoving distance to @xmath154 , @xmath155 is given by @xmath156 . 0.5 cm we define @xmath157 , with @xmath158 .
the values of @xmath159 for @xmath128 and @xmath36 are listed in table 1 .
.here @xmath160 . [ cols="<,<,<,<,<,>",options="header " , ]
the observed late time acceleration of the universe is one of the major unsolved problem in the cosmology .
it may hint at the breakdown of einstein gr .
this has lead to modification of the einstein theory of gravity .
one of the attractive possibility to modify this theory is to replace the ricci scalar @xmath10 with the generic function @xmath0 in the hilbert action .
0.5 cm @xmath0 theories are usually studied by two methods : _ metric _ and _ palatini _ approach .
the palatini approach is explored extensively in the literature both theoretically and observationally @xcite ( see the other ref s given in these papers ) .
this formulation gives second order differential field equation which can explain the late time behavior of the universe .
0.5 cm the metric approach of @xmath0 theory leads to a fourth order non linear differential equation in terms of scale factor .
this equation is difficult to solve both analytically and numerically even for the special cases .
therefore , not much observational tests have been performed on @xmath0 theories based on metric formulation .
our work is an attempt to check the concordance of @xmath0 theory of gravity using the metric approach with some of the cosmological observations . in particular , we have explored the starobinsky model in which function @xmath0 is analytic , satisfying the condition @xmath161 .
this model also passes the solar system and laboratory tests successfully for large values of @xmath32 . in this work
we study the @xmath0 model for @xmath162 and @xmath56 .
0.5 cm in order to study the cosmological viability of this theory , we investigate this class of model with two observational tests .
the first method is based on the hubble parameter versus redshift data , @xmath2 .
the hubble parameter is related to the differential age of the universe through this form @xmath163 by estimating the @xmath164 , one can obtain directly the hubble parameter , @xmath2 at different redshifts .
the h(z ) data has one major advantage , unlike in the standard candle approach ( sne ia ) : the hubble function is not integrated over .
the other important feature of this test is that differential ages are less sensitive to systematic errors as compared to the absolute ages @xcite .
this observational @xmath2 have been used earlier also to constrain various other dark energy models @xcite . 0.5 cm in this work , we check the compatibility of @xmath0 model with @xmath116 and @xmath56 with the @xmath2 data .
since @xmath135 is a nuisance parameter , we marginalise over @xmath135 .
we further combine the results obtained from @xmath2 data set with the bao data . to perform the joint test , we define the quantity : @xmath165 the results are given in the table 1 .
it appears that @xmath36 is favored by the observations .
this fact is also in agreement with solar and laboratory test .
the other important conclusions are as follow : * the variation of @xmath2 with redshift becomes independent of @xmath32 after the redshift around @xmath166 , see fig .
7 . before this redshift ( @xmath167 )
, there is a small difference between the behavior of @xmath2 with @xmath101 for @xmath133 and @xmath131 models .
this variation is still within the error bars of the data points . at higher
redshift ( i.e when @xmath168 ) we recover the standard @xmath3cdm universe for @xmath169 , for all the values of @xmath32 .
hence the thermal history of the universe is correctly reproduced by this model . *
the expansion history for this model with @xmath170 matches exactly with the standard cosmological model ( @xmath3cdm ) with equation of state parameter @xmath171 and @xmath172 ( see fig .
the variation in parameters @xmath173 and @xmath96 ( keeping @xmath174 ) leads to very small changes in values of @xmath20 , keeping it close to 2 . *
as shown in table 1 , it seems that the present observational data used in this work prefer @xmath36 over @xmath33 , as indicated by @xmath175 per degree of freedom , @xmath176 .
however , @xmath177 can not be ruled out . *
as stated earlier , the algebraic expressions become cumbersome for larger values of @xmath32 .
it is quite possible that observational tests discussed in this paper may be compatible with values of @xmath32 larger than two which are also consistent with solar and laboratory tests .
however , it should be emphasized that for a given value of @xmath20 , the minimum of the scalaron potential gets closer to zero for larger values of @xmath32 making the model vulnerable to singularity , see fig .
3 . 0.5 cm due to large error bars in the data sets used
, the variation of the expansion history of the universe studied in this paper can easily be accommodated by the observations , thereby making the models close to @xmath3cdm at the background level . at present
the sample of @xmath2 data is too small and the error bars are large . in the future , large amount of precise @xmath2 data is expected to become available .
this will not only reveal the fine features of the expansion history of the universe but also tightly constrain the cosmological parameters .
0.5 cm there is a further need to explore this model with time based observational tests , like age of the universe and high redshift objects .
it is known that the evolution of the age of the universe with the redshift vary from model to model .
so it is possible that the model of the universe which are able to reproduce the total age of the universe at @xmath178 , may not accommodate the objects at high redshift@xcite .
therefore the time based observational test may play the key role in short - listing the viable dark energy model in the near future .
it is also expected that the study of matter power perturbations would allow to distinguish this model from the standard cosmology .
we thank v. sahni and s. tsujikawa for useful comments .
ad @xmath179 dj thank a. mukherjee and s. mahajan for providing the facilities to carry out the research work .
this work is supported by japanese - indian collaboration project ( grant no dst / int / jsps / project-35/2007 ) .
the research by sn has been supported in part by the monbu - kagaku - sho of japan under grant no.18549001 and global coe program of nagoya university provided by the japan society for the promotion of science ( g07 ) .
99 v. sahni and a. a. starobinsky , int .
j. mod .
d * 9 * , 373 ( 2000 ) ; t. padmanabhan , phys .
rept . * 380 * , 235 ( 2003 ) ; p. j. e. peebles and b. ratra , rev .
phys . * 75 * , 559 ( 2003 ) ; e. j. copeland , m. sami and s. tsujikawa , int . j. mod
. phys . , * d15 * , 1753(2006)[hep - th/0603057 ] ; l. perivolaropoulos , arxiv : astro - ph/0601014 ; j. frieman , m. turner and d. huterer , arxiv:0803.0982 .
d. g. boulware and s. deser , phys .
lett . *
55 * , 2656 ( 1985 ) ; s. nojiri and s. odintsov , gen .
* 36 * ( 2004 ) 1765 [ arxiv : hep - th/0308176 ] ; phys . lett .
b * 576 * ( 2003 ) 5 [ arxiv : hep - th/0307071 ] ; phys .
b * 599 * ( 2004 ) 137 [ arxiv : astro - ph/0403622 ] ; [ arxiv : hep - th/0310045 ] ; a. a. starobinsky , jetp . lett . * 86 * , 157 ( 2007 ) .
s. a. appleby and r. a. battye , phys .
* b * 654 ( 2007 ) . s. nojiri and s. d. odintsov , phys .
b * 657 * ( 2007 ) 238 [ arxiv:0707.1941 [ hep - th ] ] ; s. nojiri and s. d. odintsov , phys . rev .
d * 77 * ( 2008 ) 026007 [ arxiv:0710.1738 [ hep - th ] ] ; g. cognola , e. elizalde , s. nojiri , s. d. odintsov , l. sebastiani and s. zerbini , phys . rev .
d * 77 * ( 2008 ) 046009 [ arxiv:0712.4017 [ hep - th ] ] .
s. a. appleby and r. a. battye , arxiv:0803.1081 .
|
we study the @xmath0 theory of gravity using metric approach .
in particular we investigate the recently proposed model by hu - sawicki , appleby @xmath1 battye and starobinsky . in this model ,
the cosmological constant is zero in flat space time .
the model passes both the solar system and the laboratory tests .
but the model parameters need to be fine tuned to avoid the finite time singularity recently pointed in the literature .
we check the concordance of this model with the @xmath2 and baryon acoustic oscillation data .
we find that the model resembles the @xmath3cdm at high redshift .
however , for some parameter values there are variations in the expansion history of the universe at low redshift .
|
all graphs considered in this paper are finite , undirected , and have no loops or multiple edges .
let @xmath10 and @xmath11 denote the sets of vertices and edges of @xmath0 , respectively .
an @xmath12-biregular bipartite graph @xmath0 is a bipartite graph @xmath0 with the vertices in one part all having degree @xmath13 and the vertices in the other part all having degree @xmath14 .
a partial edge - coloring of @xmath0 is a coloring of some of the edges of @xmath0 such that no two adjacent edges receive the same color .
if @xmath15 is a partial edge - coloring of @xmath0 and @xmath16 , then @xmath17 denotes the set of colors of colored edges incident to @xmath18 .
an edge - coloring of a graph @xmath0 with colors @xmath1 is called an interval @xmath2-coloring if for each @xmath3 there is at least one edge of @xmath0 colored by @xmath4 , and the colors of edges incident to any vertex of @xmath0 are distinct and form an interval of integers .
a graph @xmath0 is interval colorable , if there is @xmath19 for which @xmath0 has an interval @xmath2-coloring .
the set of all interval colorable graphs is denoted by @xmath20 . for a graph @xmath21 , the greatest value of @xmath2 for which @xmath0 has an interval @xmath2-coloring
is denoted by @xmath22 .
the concept of interval edge - coloring was introduced by asratian and kamalian @xcite . in @xcite
they proved the following theorem .
[ mytheorem1 ] if @xmath0 is a connected triangle - free graph and @xmath21 , then @xmath23 . in particular , from this result
it follows that if @xmath0 is a connected bipartite graph and @xmath21 , then @xmath24 .
it is worth noting that for some families of bipartite graphs this upper bound can be improved .
for example , in @xcite asratian and casselgren proved the following [ mytheorem2 ] if @xmath0 is a connected @xmath12-biregular bipartite graph with @xmath25 and @xmath21 , then @xmath26 . for general graphs , kamalian proved the following [ mytheorem3]@xcite . if @xmath0 is a connected graph and @xmath27 , then @xmath28 .
the upper bound of theorem [ mytheorem3 ] was improved in @xcite .
[ mytheorem4]@xcite . if @xmath0 is a connected graph with @xmath29 and @xmath21 , then @xmath30 .
on the other hand , in @xcite petrosyan proved the following theorem .
[ mytheorem4 ] for any @xmath31 , there is a graph @xmath0 such that @xmath21 and @xmath32 . for planar graphs ,
the upper bound of theorem [ mytheorem3 ] was improved in @xcite .
[ mytheorem6]@xcite .
if @xmath0 is a connected planar graph and @xmath21 , then @xmath33 . in this note we give a short proof of theorem [ mytheorem3 ] based on theorem [ mytheorem1 ] .
we also derive a new upper bound for the greatest possible number of colors in interval edge - colorings of regular graphs .
let @xmath34 and @xmath35 be an interval @xmath36-coloring of the graph @xmath0 . define an auxiliary graph @xmath37 as follows : it is easy to see that @xmath43 is an edge - coloring of the graph @xmath37 with colors @xmath48 and @xmath49 for @xmath47 .
now we present an interval @xmath50-coloring of the graph @xmath37 . for
that we take one edge @xmath51 with @xmath52 , and recolor it with color @xmath53 .
clearly , such a coloring is an interval @xmath50-coloring of the graph @xmath37 .
since @xmath37 is a connected bipartite graph and @xmath54 , by theorem [ mytheorem1 ] , we have in a similar way as in the prove of theorem [ mytheorem3 ] , we can construct an auxiliary graph @xmath37 and to show that this graph has an interval @xmath50-coloring .
next , since @xmath37 is a connected @xmath60-regular bipartite graph with @xmath61 and @xmath54 , by theorem [ mytheorem2 ] , we have
|
an edge - coloring of a graph @xmath0 with colors @xmath1 is called an interval @xmath2-coloring if for each @xmath3 there is at least one edge of @xmath0 colored by @xmath4 , and the colors of edges incident to any vertex of @xmath0 are distinct and form an interval of integers . in this paper
we prove that if a connected graph @xmath0 with @xmath5 vertices admits an interval @xmath2-coloring , then @xmath6 . we also show that if @xmath0 is a connected @xmath7-regular graph with @xmath5 vertices has an interval @xmath2-coloring and @xmath8 , then this upper bound can be improved to @xmath9
. + keywords : edge - coloring , interval coloring , bipartite graph , regular graph
|
since the mid 1980s narrow - line seyfert 1 galaxies ( nls1s ; * ? ? ?
* ) have become a field of extensive study in agn science .
nls1s are crucial for our understanding of the agn phenomenon , because they are most likely agn at an early stage ( e.g. * ? ? ?
they possess relatively low - mass black holes and high eddington ratios @xmath4 .
nls1s are characterized by extreme properties , such as steep soft and hard x - ray spectra , strong x - ray variability , and strong optical fe ii emission ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the nls1 mkn 335 ( @xmath5 = @xmath6 , @xmath7 = @xmath8 , z=0.026 ) is a well - known bright soft x - ray agn and has been the target of most x - ray observatories .
it was seen as a bright x - ray agn by uhuru @xcite and einstein @xcite .
@xcite reported a strong soft x - ray excess found in the exosat spectrum , which was confirmed by bbxrt observations @xcite .
ginga observations of mkn 335 suggested the presence of a warm absorber in the source @xcite . during rosat observations it also appeared bright and with a strong soft x - ray excess @xcite .
the x - ray spectrum during the 1993 asca observation @xcite was either interpreted by the presence of a warm absorber @xcite or by x - ray reflection on the disk @xcite .
beppo - sax observations of mkn 335 also confirm the presence of a strong soft x - ray excess @xcite and a small or moderate compton reflection component .
_ xmm - newton _ observed mkn 335 in 2000 and again in 2006 @xcite .
mkn 335 is exceptional in showing evidence for an unusually broad wing in the iron line @xcite .
the wing is required , if the xmm spectrum is explained in terms of reflection ; it is not , if a partial covering interpretation is adopted .
high - amplitude variability provides important new constraints to distinguish between these different spectral models .
mkn 335 was observed by _ swift _
@xcite in 2007 may and appeared to be dramatically fainter in x - rays than seen in all previous observations . in this letter
we report on this historical low x - ray flux state of mkn 335 and compared the continuum properties of the _ swift _ with previous _ xmm - newton _ observations . throughout the paper spectral indexes
are denoted as energy spectral indexes with @xmath9 .
luminosities are calculated assuming a @xmath10cdm cosmology with @xmath11=0.27 , @xmath12=0.73 and a hubble constant of @xmath13=75 km s@xmath14 mpc@xmath14 corresponding to a luminosity distance d=105 mpc .
all errors are 90% confidence unless stated otherwise .
_ swift _ observed mkn 335 on 2007 may 17 and 25 and june 30 to july 02 for 4.8 , 8.2 , and 8.7 ks ( table[obs_log ] ) , respectively , with its x - ray telescope ( xrt ) in photon counting mode ( pc mode ) and in all 6 filters of the uv - optical telescope ( uvot ) .
x - ray data were reduced with the task _ xrtpipeline _ version 0.11.4 .
source and background photons were extracted with _
xselect _ version 2.4 , from circles with radii of 47@xmath15 and 189@xmath15 , respectively .
the spectral data were re - binned with at least 20 photons per bin _ grppha _ version 3.0.0 .
the 0.3 - 10.0 kev spectra were analyzed with _
xspec _ version 12.3.1x @xcite .
the auxiliary response files were created with _ xrtmkarf _ and corrected using the exposure maps , and the standard response matrix _
swxpc0to12_20010101v008.rmf_. the uvot data were coadded for each segment in each filter with the uvot task _ uvotimsum _ version 1.3 .
source photons in all filters were selected in a circle with a radius of 5@xmath15 .
uvot magnitudes and fluxes were measured with the task _ uvotsource _ version 3 .
the uvot data were corrected for galactic reddening ( @xmath16 ; * ? ? ? * ) . _ xmm - newton _ observed mkn 335 in 2000 and 2006 for 37ks and 133 ks , respectively ( see table[obs_log ] ) . during the 2000 observation the optical monitor ( om ) did photometry in the v , b , u , and m2 filters .
during the 2006 observation the uv grism was used exclusively . the _ xmm - newton _
epic pn data were analyzed using the xmmsas version _ xmmsas_20060628_1801 - 7.0.0_. the 2000 amd 2006 observations were performed in full - frame and small window , respectively . because the 2000 observation was severely affected by pileup , photons from a 20@xmath15 source - centered circle were excluded .
the source photons in the 2006 pn data were selected in a radius of 1@xmath17 and background photons of both observations from a source - free region close by with the same radius .
the spectra were rebinned with 100 photons per bin . in order to compare the photometry in the om with the uvot we selected 5 field stars with similar brightness in v as mkn 335 . only in b and m2 the om magnitudes
had to be adjusted by @xmath180.10 mag and @xmath190.30 mag , respectively .
none of the _ xmm - newton _ and _ swift _ spectra can be fitted by a single absorbed power law model . in the literature a variety of spectral models have been applied to the x - ray data of mkn 335 , including warm absorption @xcite , partial covering @xcite and reflection @xcite .
fits with a warm absorber model ( _ absori _ ) and blackbody plus power law model yield unacceptable results . while fits to the _ xmm - newton _ 2000 data yield acceptable fits by using an absorbed broken power law model with the absorption column density fixed to the galactic value ( 3.96@xmath20 @xmath1 ; * ? ? ?
* ) , the _ xmm - newton _ 2006 and _ swift_spectra require additional components . we used a partial covering absorber model with an underlying power law and broken power law spectral models . table[xray_res ] summarizes the results from the x - ray spectral analysis .
figure[mkn335_xspec ] displays the _
spectra fitted with a power law and partial covering absorber
. fits to each spectrum were first performed separately .
subsequently all the _ swift _ spectra were fitted simultaneously in xspec with the power law spectral slopes tied and the absorber parameters and the normalizations left to vary .
the results are listed in table[xray_res ] and suggest a development of the partial covering absorber over time .
the most dramatic change is from the 2006 _ xmm - newton _ to the first _
swift_observation when the absorber became nearly opaque and only 2% of the x - ray emission can be seen directly . in this case
the absorption column density changes from @xmath21 @xmath1 with a covering fraction of 0.45 during the 2006 _ xmm - newton _ observation to about @xmath22 @xmath1 and a covering fraction of 0.98 during the first swift observation .
we fitted all three _ swift _ spectra simultaneously in xspec by tying the covering fraction @xmath2 and spectral indices together .
this fit suggests a change in the absorption column density @xmath23 of the partial covering absorber by a factor of 2 within a week between the 2007 may 17 and 25 observations .
alternatively , we also fitted the spectra with @xmath23 tied and @xmath2 left as a free parameter .
an f - test gives an f - value of 7.8 that these two fits are different and a probability p=0.006 of a random result .
leaving @xmath23 free gives a significantly better result than leaving @xmath2 free to vary . in the rest - frame 0.2 - 2.0 kev band the observed fluxes ( only corrected for galactic absorption ) seem to be highly variable and between the _ xmm - newton _
2000 and the first _ swift _ observation we found variability by a factor of 30 .
however , when correcting also for intrinsic absorption the unabsorbed restframe 0.2 - 2.0 kev fluxes from rosat to swift are comparable . during the rosat all - sky survey observation a flux of 4@xmath24 w m@xmath25 was found @xcite . correcting for a partial covering absorber in the _ xmm - newton _ and _ swift _ spectra we found that the flux varied only by factors of 4 - 6 as listed in table[xray_res ] .
the x - ray spectra of mkn 335 in the higher and more typical flux state can be well described as arising from an incident power law and reflection component ( e.g. * ? ? ?
* ; * ? ? ?
however , the low - flux spectra are difficult to reproduce by simply rescaling the high - state models or by varying the relative contribution of each component .
a modified reflection model , which self - consistently describes the high- and low - flux states is being investigated and is presented in gallo et al .
( in prep ) .
as shown by the spectral energy distribution ( sed ) in figure[mkn335_sed ] there was no dramatic variability in the uv data between the 2000 _ xmm - newton _ om and 2007 _ swift _ uvot observations , although during the 2007 may 17th observation mkn 335 was about 0.2 mag fainter .
the uv / optical spectral slopes are on the order of @xmath26 , except for the 2007 may 17 observation when it was @xmath27 .
the uv to x - ray spectral slope @xmath28=0.384 log(@xmath29 ) .
] was significantly steeper during the _ swift _
observations with @xmath28=1.91 and 1.65 during the _ swift _ segments 001 and 002 , respectively . during the 2000 _ xmm - newton _ observation ,
however , an @xmath28=1.32 was measured , consistent with the value given by @xcite .
we reported the _ swift _ observations of the nls1 mkn 335 when it was in its lowest x - ray flux state ever observed .
historically , mkn 335 has exhibited x - ray variability by about a factor of a few ( e.g. * ? ? ?
* ; * ? ? ?
* ) , although the source has always remained rather bright at least until the last x - ray observation with _ suzaku _ in 2006 june ( j. larsson , 2007 priv .
however , sometime between 2006 june and 2007 may the observed flux dropped by a factor of more than 30 including a dramatic change in its sed .
the x - ray spectrum has become progressively more complex as the x - ray flux has diminished , indicative of either absorption or reflection ( e.g. * ? ? ?
the 2 - 10 kev high - flux spectrum in 2000 did not appear overly complex and the high - energy continuum could be simply fitted with a power law .
the 2006 _ xmm - newton _ data , however , can be fitted with a partial covering absorber model ( see also * ? ? ?
* ) , suggesting that the absorber started moving in the line of sight before 2006 january .
partial covering of the central light source has been invoked since the early days of agn x - ray spectroscopy ( e.g. * ? ? ?
* ) , and quite often to describe the x - ray spectrum of nls1 ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
its presence is also indicated by narrow absorption lines ( which appear to be saturated but do not reach zero intensity ) in uv spectra of broad absorption line ( bal ) quasars ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
however , the geometry and physics of partial coverers are still not well understood .
one possible geometry consists of thick blobs of gas , partially covering parts of the accretion disk ( e.g. * ? ? ?
* ) . in the case of mkn 335
, the clouds must cover only the inner parts of the disk , since we find that the uv emission is not highly variable between the _ xmm - newton _ observation in 2000 and the _ swift _ observations in 2007 , while the x - rays vary dramatically .
note that the fits to the may 17 and may 25 swift spectra suggest a change in the partial covering absorber column densities by a factor of about 2 .
this timescale is consistent with e.g. the absorber toy model suggested by @xcite , where thick clouds at 10 - 100 schwarzschild radii partially obcure the central region and causing the x - ray variability . alternatively ,
a partial covering situation may arise if our line - of - sight passes through an accretion - disk driven wind which is launched at intermediate disk radii ( e.g. * ? ? ?
* ; * ? ? ?
if such a wind varies with time and/or is inhomogeneous , different parts of the central source would be covered at different times . in both partial covering geometries ,
the physics is still uncertain . in the case of dense blobs : how are they confined and what is their origin ( e.g. * ? ? ?
* ) . in case of disk - driven winds
: what is the driver of these massive outflows ( e.g. * ? ? ?
the high column density we need in our swift spectral fits is similar to those frequently observed in bal quasars ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . in this context
, it is interesting to note that similarities between nls1 galaxies and bal quasars have been pointed out repeatedly ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
in one specific case , that of the x - ray transient nls1 galaxy wpvs 007 , the onset of heavy x - ray absorption @xcite is indeed accompanied by the onset of uv bals @xcite . when correcting for the effects of intrinsic absorption we found that the x - ray flux of mkn 335 originating from the central engine has been very similar in the rest - frame 0.2 - 2.0 kev band between the rosat observations and the most recent _ swift_observations . using these fluxes , the intrinsic variability is only a factor of about 4 - 6 , which is quite normal for an agn , in particular for a nls1 .
the change in intrinsic flux between the first and second swift observations is about a factor of 3 within a week .
if a partial covering absorber is the correct model this flux change implies that the soft x - ray scattering region can only be a few light days in diameter which is consistent with the @xcite toy model .
the deep low - state of mkn 335 discovered with _ swift_provides us with a rare chance to scrutinize the properties of x - ray low - state agn in general .
mkn 335 is unique with respect to being relatively bright during its low - state .
therefore , follow - up observations of mkn 335 in its current low - state are highly encouraged
. we will continue our monitoring with _ swift _ in order to find the timescales on which the agn switches from a low to high state , but also deep _ xmm - newton _ observations , optical spectrocsopy and spectropolarimetry are needed to clarify the nature of the current low - state .
we want to thank paul obrien and our referee for useful comments on the manuscript .
we also want to thank the _ swift _ pi neil gehrels for approving our too request and the _ swift _ science planners mike stroh and sally hunsberger for fitting the observations into the _ swift_schedule .
we acknowledge the use of public data from the _ xmm - newton _ and swift data archives . 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 .
swift is supported at psu by nasa contract nas5 - 00136 .
this research was supported by nasa contract nnx07ah67 g ( d.g . ) .
abrassart , a. , & czerny , b. , 2000 , , 356 arnaud , k. a. , 1996 , asp conf .
ser . 101 : astronomical data analysis software and systems v , 101 , 17 ballantyne , d.r . ,
iwasawa , k. , & fabian , a.c . ,
2001 , , 323 , 506 barlow , t.a . , hamann , f. , sargent , w.l.w .
, 1997 , aspc , 128 , 13 bianchi , s. , matt , g. , haardt , f. , maraschi , l. , nicastro , f. , perola , g.c . , petrucci , p.o . , & piro , l. , 2001 , , 376 , 77 boller , t. , brandt , w.n . ,
& fink , h.h . , 1996 , , 305 , 53 boroson , t.a . ,
& green , r.f . , 1992 , , 80 , 109 boroson , t.a . , 2002 , , 565 , 78 brandt , w.n .
, laor , a. , & wills , b.j . , 2000 , ,
528 , 637 brandt , w.n . , & gallagher , s.c . ,
2000 , newar , 44 , 461 crummy , j. , fabian , a.c . ,
gallo , l.c . , & ross , r.r . , 2006 , , 365 , 1067 czerny , b. , & janiuk , a. , 2007 , , 464 , 167 dickey , j.m . , & lockman , f.j .
, 1990 , , 28 , 215 doroshenko , v.t . , sergeev , s.g . ,
merkulova , n.i . ,
sergeeva , e.a . , & golubinsky , y.v . , 2005 , , 437 , 87 dunn , j.p . ,
jackson , b. , deo , r.p . , farrington , c. , das , v. , & crenshaw , d.m .
, 2006 , , 118 , 572 edelson , r.a . ,
krolik , j.h . , & pike , g.f . , 1990 , , 359 , 86 elvis , m. , 2000 , , 545 , 63 gallagher , s.c . , brandt , w.n .
, chartas , g. , & garmire , g.p . , 2002 , , 567 , 37 gallo , l.c . ,
tanaka , y. , boller , th . ,
fabian , a.c . ,
vaughan , s. , & brandt , w.n .
, 2004 , , 353 , 1064 gallo , l.c . , 2006 , , 368 , 479 gallo l.c . , grupe , d. , & komossa , s. , 2007 , , in prep .
gehrels , n. , et al .
, 2004 , apj , 611 , 1005 george , i.m . ,
turner , t.j . ,
yaqoob , t. , netzer , h. , laor , a. , mushotzky , r.f . ,
nandra , k. , & takahashi , t. , 2000 , , 531 , 52 gondoin , p. , orr , a. , lumb , d. , & santos - lleo , m. , 2002 , , 388,74 green , p.j . , & mathur , s. , 1996 , , 462 , 637 grupe , d. , , 127 , 1799 grupe , d. , thomas , h .- c . , & beuermann , k. , 2001 , , 367 , 470 grupe , d. , mathur , s. , & elvis , m. , 2003 , , 126 , 1159 grupe , d. , wills , b.j .
, leighly , k.m . , & meusinger , h. , 2004a , , 127 , 156 grupe , d. , mathur , s. , & komossa , s. , 2004b , , 127 , 3161 grupe , d. , schady , p. , leighly , k.m . ,
komossa , s. , obrien , p.t .
, & nousek , j.a . , 2007 , , 133 , 1988 guilbert , p.w . , & rees , m.j . , 1988 , , 233 , 475 halpern , j.p . , 1982 , phd thesis , harvard university hamann , f. , 1998 , , 500 , 798 holt , s. , et al .
, 1980 , , 241 , l13 kuncic , z. , celotti , a. , rees , m.j . , 1997 , , 284 , 717 leighly , k.m . , 1999 , , 125 , 297 leighly , k.m . ,
1999b , , 125 , 317 leighly , k.m .
, et al . , 2007 , in prep longinotti , a.l . , sim , s.a . , nandra , k. , & cappi , m. , 2007a , , 374 , 237 longinotti , a.l . ,
sim , s.a . , nandra , k. , cappi , m. , & oneill , p. , 2007b
, asp conf .
series , vol . , eds markowitz , a. , & edelson , r. , 2004 , , 617 , 939 mathur , s. , 2000 , , 314 , l17 nandra , k. , & pounds , k.a .
, 1994 , , 268 , 405 oneill , p.m. , nandra , k. , cappi , m. , longinotti , a.l . , &
sim , s.a .
, 2007 , , accepted , astro - ph/0708.0751 osterbrock , d.e . , & pogge , r.w . , 1985 , , 297 , 166 pounds , k.a . , stanger , v.j . ,
turner , t.j . , king , a.r . ,
& czerny , b. , 1987 , , 224 , 443 proga , d. , 2007 , , 661 , 693 schlegel , d. j. , finkbeiner , d. p. , & davis , m. 1998 , , 500 , 525 tanaka , y. , boller , t. , & gallo , l. , in merloni , a. , nayakshin , s. , sunyaev , r.a . , eds , growing black holes : accretion in a cosmological context , springer , berlin , p. 290
tananbaum , h. , peters , g. , forman , w. , giacconi , r. , jones , c. , & avni , y. , 1978 , , 223 , 74 tananbaum , h. , et al . , 1979 , , 234 , l9 turner , t.j .
, et al . , 1993 , , 407 , 556 wills , b.j . ,
brandt , w.n .
, & laor , a. , 1999 , , 520 , l91 lcccr xmm 2000 & 12 - 25 17:18 & 12 - 26 02:06 & 36910 + xmm 2006 & 01 - 03 19:10 & 01 - 05 08:03 & 133251 + _ swift _
001/2007 & 05 - 17 00:32 & 05 - 17 05:37 & 4860 + _ swift _ 002/2007 & 05 - 25 00:01 & 05 - 25 19:23 & 8084 + _ swift _ 003/2007 & 06 - 28 00:01 & 06 - 28 14:37 & 2932 + _ swift _
004/2007 & 06 - 30 00:13 & 06 - 30 14.52 & 2837 + _ swift _ 005/2007 & 07 - 02 14:47 & 07 - 02 21:18 & 2979 ccccccccccccc xmm 2000 & ( a ) & [email protected] & 1.76@xmath31 & [email protected] & & & 13.10 & & 513/419 + & ( b ) & [email protected] & & & 7.7@xmath32 & [email protected] & 13.10 & 12.72 & 561/419 + xmm 2006 & ( a ) & [email protected] & [email protected] & [email protected] & & & 13.17 & & 2420/1327 + & ( c ) & [email protected] & [email protected] & [email protected] & 55.1@xmath33 & 0.45@xmath34 & 13.17 & 12.91 & 1973/1325 + & ( d ) & [email protected] & [email protected] & [email protected] & 51.0@xmath35 & 0.43@xmath34 & 13.17 & 12.93 & 2611/1910 + _ swift _ 001 & ( c ) & 2.05@xmath36 & & & 31.1@xmath37 & 0.98@xmath38 & 14.63 & & 9/9 + & ( e ) & [email protected] & [email protected] & [email protected] & 38.2@xmath39 & 0.95@xmath34 & 14.75 & 13.48 & 2611/1910 + & ( f ) & [email protected] & & & 20.0@xmath40 & 0.93 & 14.71 & 13.55 & 87/85 + & ( g ) & [email protected] & & & 20.5@xmath41 & 0.94@xmath42 & 14.70 & 13.51 & 87/86 + & ( h ) & [email protected] & & & 12.8@xmath43 & 0.91@xmath44 & 14.72 & 13.69 & 95/86 + _ swift _ 002 & ( c ) & [email protected] & & & 10.4@xmath43 & 0.93@xmath38 & 14.25 & 13.10 & 52/48 + & ( e ) & [email protected] & [email protected] & [email protected] & 10.4@xmath45 & 0.88@xmath38 & 14.27 & 13.37 & 2611/1910 + & ( f ) & [email protected] & & & 10.4@xmath46 & [email protected] & 14.25 & 13.10 & 87/85 + & ( g ) & [email protected] & & & 10.9@xmath47 & 0.94@xmath42 & 14.25 & 13.06 & 87/86 + & ( h ) & [email protected] & & & 12.8@xmath43 & 0.94@xmath42 & 14.25 & 13.02 & 95/86 + _ swift _ 003 - 005 & ( c ) & [email protected] & & & 16.6@xmath48 & 0.93@xmath49 & 14.49 & 13.32 & 23/26 + & ( e ) & [email protected] & [email protected] & [email protected] & 17.5@xmath50 & 0.89@xmath51 & 14.50 & 13.56 & 2611/1910 + & ( f ) & [email protected] & & & 16.4@xmath52 & 0.94@xmath38 & 14.48 & 13.28 & 87/85 + & ( g ) & [email protected] & & & 15.8@xmath53 & 0.94@xmath42 & 14.47 & 13.28 & 87/86 + & ( h ) & [email protected] & & & 12.8@xmath43 & 0.92@xmath38 & 14.48 & 13.36 & 95/86
|
we report the discovery of the narrow - line seyfert 1 galaxy mkn 335 in an extremely low x - ray state . a comparison of _ swift _ observations obtained in may and june / july 2007 with all previous x - ray observations between 1971 to 2006 show the agn to have diminished in flux by a factor of more than 30 , the lowest x - ray flux mkn 335 has ever been observed in .
the _
swift_observations show an extremely hard x - ray spectrum at energies above 2 kev .
possible interpretations include partial covering absorption or x - ray reflection from the disk . in this letter
we consider the partial covering interpretation .
the _ swift _
observations can be well fit by a strong partial covering absorber with varying absorption column density ( @xmath0 @xmath1 ) and a covering fraction @xmath2=0.9 - 1 .
when corrected for intrinsic absorption , the x - ray flux of mkn 335 varies by only factors of 4 - 6 . in the uv mkn 335
shows variability in the order of 0.2 mag .
we discuss the similarity of mkn 335 with the highly variable nls1 wpvs007 , and speculate about a possible link between nls1 galaxies and broad - absorption line quasars .
|
let @xmath1 be a smooth @xmath3-dimensional complex projective variety .
boucksom , demailly , paun , and peternell have recently shown that the dual of the cone of numerical classes of effective divisors on @xmath1 is the closure of the cone of numerical classes of movable curves ( * ? ? ?
* theorem 0.2 ) ; see also @xcite .
their theorem is analogous to a theorem of kleiman that says that the dual of the cone of numerical classes of ample divisors is the closure of the cone of numerical classes of effective curves @xcite .
debarre and lazarsfeld asked whether these results might generalize in the following way : for @xmath4 , is the dual of the cone of numerical classes of divisors on @xmath1 whose stable base locus has dimension less than @xmath2 generated by the numerical class of a natural collection of curves , such as curves moving in a family that sweeps out a subvariety of dimension @xmath2 ? this paper gives an affirmative answer to their question in the toric case , with a slight twist the curves that one must consider include not only curves on @xmath1 , but also curves on small modifications of @xmath1 .
recall that a small modification @xmath5 is a birational map that is an isomorphism in codimension 1 .
if @xmath6 is a subvariety of @xmath1 , we say that @xmath7 maps @xmath6 birationally to @xmath8 if @xmath7 is defined on an open set meeting @xmath6 and maps a dense open subset of @xmath6 isomorphically onto a dense open subset of @xmath8 . if @xmath9 is a divisor on @xmath1 , then we write @xmath10 for the divisor @xmath11 on @xmath12 .
[ main ] let @xmath1 be a complete @xmath0-factorial toric variety with dense torus @xmath13 , and let @xmath14 be a divisor on @xmath1 .
the following are equivalent : 1 .
the stable base locus @xmath15 has dimension less than @xmath2 .
2 . for every @xmath13-invariant subvariety @xmath16 of dimension @xmath2 , and for every small modification @xmath5 that maps @xmath6
birationally to @xmath8 , with @xmath12 projective and @xmath0-factorial , and for every irreducible curve @xmath17 on @xmath12 moving in a family sweeping out @xmath8 , @xmath18 recall that the stable base locus @xmath15 of a @xmath0-divisor @xmath14 is the set - theoretic intersection of the base loci of the complete linear systems @xmath19 for all positive integers @xmath20 such that @xmath21 is integral .
let @xmath22 be the space of @xmath0-divisors on @xmath1 , modulo numerical equivalence .
since the stable base locus of the sum of two divisors is contained in the union of their stable base loci , the set of classes in @xmath22 that are represented by divisors whose stable base locus has codimension greater than @xmath2 is a convex cone ( for a general numerical class , the stable base locus is independent of the choice of representative ( * ? ? ? * and 10.3 ) ) .
we write @xmath23 for the closure of this cone in @xmath24 .
this notation is suggestive of the fact that @xmath23 is also the closure of the cone of numerical classes of divisors that are ample in codimension @xmath2 , that is , the cone of numerical classes of divisors @xmath14 such that there is an open set @xmath25 , with @xmath26 ample , whose complement has codimension greater than @xmath2 .
the cone @xmath27 of pseudo - effective divisors , the closure of the cone of numerical clases of effective divisors , is filtered by the cones @xmath28 : @xmath29 we have an analogous filtration of the cone @xmath30 of pseudo - effective curves , as follows .
let @xmath31 be the space of 1-cycles in @xmath1 with coefficients in @xmath32 , modulo numerical equivalence .
the intersection pairing on @xmath1 induces a natural isomorphism @xmath33 .
let @xmath34 be the the convex cone in @xmath31 generated by classes of irreducible curves moving in a family that sweeps out a @xmath2-dimensional subvariety of @xmath1 .
so @xmath35 if a curve @xmath17 moves in a family that sweeps out a @xmath2-dimensional subvariety , and if @xmath15 has dimension smaller than @xmath2 , then @xmath17 can be moved off of the support of some effective divisor linearly equivalent to a multiple of @xmath14 , and hence @xmath36 is nonnegative .
so there are natural inclusions @xmath37 kleiman s theorem says that equality holds for @xmath38 , and the theorem of boucksom , demailly , paun , and peternell says that equality holds for @xmath39 . as lazarsfeld points out , one does not expect equality for @xmath4 ; in light of the equality for @xmath39 , equality for general @xmath2 would say , roughly speaking , that the restriction of a big divisor to its stable base locus is not pseudo - effective , which seems too strong ( a classical theorem of fujita says that the restriction of a divisor to its stable base locus is not ample ( * ? ? ?
* theorem 1.19 ) ) .
another heuristic reason to believe that equality should not hold for general @xmath2 comes from the consideration of small modifications @xmath40 , with @xmath12 complete and @xmath0-factorial .
numerical equivalence on @xmath1 and on @xmath12 coincide ( this follows , for instance , from kleiman s characterization of numerically trivial line bundles ( * ? ? ?
* xiii , theorem 4.6 ) ) .
so the map @xmath41 induces an isomorphism @xmath42 .
the global sections of @xmath14 and of @xmath10 are canonically identified , so if @xmath6 is contained in @xmath15 , and if @xmath7 maps @xmath6 birationally to @xmath8 , then @xmath8 is in the stable base locus of @xmath10 . in particular , @xmath43 is identified with @xmath44 .
but there is no natural way to identify @xmath45 and @xmath46 compatibly with the intersection pairing .
we now introduce additional cones of numerical classes of 1-cycles which take into account @xmath2-movable curves on small modifications of @xmath1 .
let @xmath5 be a small modification , with @xmath12 complete and @xmath0-factorial .
we define @xmath47 to be the image of the convex cone generated by numerical classes of irreducible curves @xmath17 on @xmath12 moving in a family that sweeps out the birational image of a @xmath2-dimensional subvariety of @xmath1 under the identifications @xmath48 there is an obvious inclusion @xmath49 with this notation , theorem [ main ] can be restated as follows . [ main ]
let @xmath1 be a complete @xmath3-dimensional @xmath0-factorial toric variety .
then @xmath50 where the sum is over all small modifications @xmath40 such that @xmath12 is projective and @xmath0-factorial .
it seems unclear whether a statement like theorem [ main ] should be true in general , because it is not clear whether a general variety @xmath1 will have enough @xmath0-factorial projective small modifications . by (
* lemma 1.6 ) , this depends essentially on whether @xmath1 has enough divisors @xmath14 whose section ring @xmath51 is finitely generated .
* problem .
* let @xmath1 be a projective @xmath3-dimensional @xmath0-factorial complex variety .
is the inclusion @xmath52 an equality ?
the following example illustrates the necessity of considering curves on small modifications ; we give an example of a threefold @xmath1 such that @xmath53 is properly contained in @xmath54 .
[ projective bundle ] _ assume these claims are true .
choose in @xmath55 an open cone @xmath25 containing @xmath56 sufficiently small so that any class in @xmath25 pairs negatively with @xmath57 and positively with the @xmath58 .
any effective representative of a class in @xmath25 must have a component that lies in @xmath57 , and @xmath57 is disjoint from the @xmath58 , so @xmath25 does not contain the class of any irreducible curve on @xmath1 . by perturbing a supporting hyperplane cutting out the extremal ray spanned by @xmath56 , we can find a closed half space @xmath59 containing @xmath60 but not containing @xmath56 . since @xmath53 is generated by classes in @xmath60 , @xmath53
is contained in @xmath59 and does not contain @xmath56 .
_ _ however , consider the small modification @xmath61 given by flopping the three curves @xmath62 .
the variety @xmath12 can be realized as a projectivized vector bundle over the blow up of @xmath63 at @xmath64 , @xmath65 , and @xmath66 , @xmath67 where @xmath68 is the line bundle given by the strict transform of a smooth cubic passing through the @xmath69 .
let @xmath70 be the section of @xmath71 whose image is @xmath72 .
then @xmath56 is represented on @xmath12 by @xmath73 , where @xmath17 is the strict transform of a smooth conic in @xmath63 passing through the @xmath69 , and @xmath73 moves in a family that sweeps out the surface @xmath72 in @xmath12 . _
for the remainder of this paper , we fix a complete simplicial fan @xmath74 in an @xmath3-dimensional latticed vector space @xmath75 .
let @xmath76 be the corresponding toric variety over some fixed algebraically closed field , and let @xmath13 be the dense torus in @xmath1 .
let @xmath77 be the rays ( or 1-dimensional cones ) of @xmath74 .
let @xmath78 be the primitive generator of @xmath79 , and let @xmath80 be the corresponding prime @xmath13-invariant divisor on @xmath1 .
see @xcite for details on the correspondence between fans and toric varieties . since @xmath74 is complete and simplicial , @xmath81 is @xmath0-factorial , and numerical and rational equivalence coincide on @xmath1 .
we have a short exact sequence @xmath82 where the map from @xmath83 takes @xmath84 to @xmath85 .
the dual short exact sequence is @xmath86 where @xmath87 is the character lattice of @xmath13 , and the map from @xmath83 takes @xmath88 to the numerical class of the divisor @xmath89 .
if @xmath90 , then the polytope @xmath91 is defined by @xmath92 if @xmath14 is integral , the characters @xmath93 , for all @xmath94 , form a basis for @xmath95 .
the vanishing locus of @xmath96 is supported exactly on the union of the divisors @xmath80 such that @xmath97 .
it follows that the base locus , and hence the stable base locus , of any divisor on @xmath1 is @xmath13-invariant .
see @xcite for more details .
if @xmath98 is the @xmath13-invariant subvariety corresponding to a cone @xmath99 , then @xmath98 is not contained in the stable base locus of @xmath14 if and only if the class of @xmath14 in @xmath55 is in the cone @xmath100 spanned by the classes of @xmath13-invariant divisors not containing @xmath98 .
in other words , @xmath101 since base loci are @xmath13-invariant , and the codimension of @xmath98 is the dimension of @xmath102 , @xmath103 since each @xmath100 is a rational polyhedral cone , it follows that @xmath28 is also a rational polyhedral cone .
furthermore , the dual of @xmath28 is given by @xmath104 from the description of @xmath100 and the exact sequences above , we have @xmath105 the cones @xmath106 , for maximal cones @xmath102 , appear prominently in the bunches of cones " in @xcite .
we now prove the claims made in example [ projective bundle ] .
_ suppose @xmath107 , @xmath108 , and @xmath109 with @xmath74 being the fan whose maximal cones are @xmath110 then @xmath111 is the variety considered in example [ projective bundle ] , @xmath112 for @xmath113 and @xmath114 , @xmath115 is the strict transform of @xmath116 , and @xmath117 is the strict transform of @xmath118 .
see @xcite for the construction of the fan corresponding to a projectivized split vector bundle on a toric variety .
the following diagram illustrates the fan @xmath119 corresponding to @xmath12 , as well as the fan @xmath74 ; the intersection of each fan with the hyperplane @xmath120 is shown .
see @xcite for the changes in the fan corresponding to a flop .
_ @xmath121 @xmath122 _ we claimed that the classes of @xmath123 , and @xmath124 give a basis for @xmath55 .
this now follows immediately from the short exact sequence @xmath125 and the fact that @xmath126 is a basis for @xmath127 .
it remains to show that the class @xmath128 given by @xmath129 , @xmath130 , and @xmath131 spans an extremal in @xmath54 .
recall that @xmath132 for @xmath133 , @xmath134 is in @xmath135 , and hence @xmath56 is not in @xmath136 .
therefore , it suffices to show that @xmath56 spans an extremal ray in @xmath137 , which is clear : @xmath56 spans the inward normal to the face of @xmath138 spanned by the classes of @xmath139 , and @xmath124 . _
by the exact sequence @xmath82 a numerical class @xmath128 is represented uniquely by an @xmath140-tuple of real numbers @xmath84 , such that @xmath141 , and these numbers satisfy the relation @xmath142 conversely , any such @xmath140-tuple @xmath143 corresponds to some numerical class in @xmath31 . given a cone @xmath99
, we say that a ray @xmath79 is adjacent to @xmath102 if there is a cone @xmath144 containing both @xmath102 and @xmath79 .
[ curves move ] let @xmath17 be an irreducible curve in @xmath1 , with @xmath145 .
then there is a cone @xmath99 such that 1 .
@xmath17 moves in a family sweeping out @xmath98 , 2 .
@xmath146 unless @xmath79 is adjacent to @xmath102 , 3 .
@xmath147 unless @xmath79 is in @xmath102 .
let @xmath102 be the unique cone such that the orbit @xmath148 contains an open dense subset of @xmath17 .
the action of @xmath13 moves @xmath17 in a family sweeping out @xmath98 .
furthermore , if @xmath79 is not adjacent to @xmath102 , then @xmath80 is disjoint from @xmath98 , so @xmath149 .
similarly , if @xmath79 is not in @xmath102 , then @xmath80 does not contain @xmath148 , and hence does not contain @xmath17 , so @xmath150 .
proposition [ curves move ] restricts the numerical classes of 1-cycles that can be represented by a positive scalar multiple of an irreducible curve .
we now show , by explicitly constructing curves with given numerical properties , that these are the only nontrivial restrictions . [ constructing curves ] let @xmath102 be a cone in @xmath74 , and let @xmath151 be integers such that 1 .
@xmath152 , 2 .
@xmath146 unless @xmath79 is adjacent to @xmath102 , 3 .
@xmath147 unless @xmath79 is in @xmath102 .
then there is an irreducible curve @xmath17 that moves in a family sweeping out @xmath98 such that @xmath145 for all @xmath153 .
we first consider the case @xmath154 .
so we are given nonnegative integers @xmath151 such that @xmath152 , and must construct a curve @xmath17 sweeping out @xmath1 such that @xmath145 for all @xmath153 .
choose distinct elements @xmath155 in the base field , let @xmath156 be the one - parameter subgroup corresponding to @xmath157 , and consider the rational map @xmath158 given by @xmath159 where the product is given by the group law on @xmath13 .
since @xmath1 is complete , @xmath160 extends to a regular morphism @xmath161 .
it is straightforward to check , using local coordinates , that if @xmath146 then @xmath162 is disjoint from @xmath80 , and if @xmath163 is positive then @xmath164 furthermore , in the latter case , @xmath165 is a point in the dense orbit @xmath166 , along which @xmath80 is locally principal , and the local intersection multiplicity is @xmath167 the action of @xmath13 moves @xmath162 in a family sweeping out @xmath1 .
we now consider the case of a nonzero cone @xmath102 in @xmath74 .
let @xmath168 be the quotient lattice @xmath169 and let @xmath170 be the fan in @xmath171 whose cones are the projections of cones of @xmath74 containing @xmath102 , so @xmath98 is the toric variety corresponding to @xmath170 @xcite .
the rays of @xmath170 are the projections of the rays of @xmath74 adjacent to @xmath102 , but not contained in @xmath102 ; after renumbering , we may assume that these rays are @xmath172 .
let @xmath173 be the image of @xmath79 under the projection @xmath174 , let @xmath175 be the primitive generator of @xmath173 in @xmath168 , let @xmath176 be the positive integer such that @xmath177 , and let @xmath178 be the divisor in @xmath98 corresponding to @xmath173 , for @xmath179 . by hypothesis , @xmath180 in @xmath168 .
hence there is a curve @xmath17 sweeping out @xmath98 such that @xmath181 for @xmath179 .
we claim that @xmath17 is the required curve .
let @xmath182 be the natural inclusion , so @xmath183 . by the basic properties of toric intersection theory , as developed in @xcite , @xmath184 , so @xmath185 for @xmath179 .
for @xmath186 , if @xmath79 is not adjacent to @xmath102 , then @xmath80 is disjoint from @xmath98 and hence @xmath149 . and
if @xmath79 is in @xmath102 , then @xmath187 is uniquely determined by the condition @xmath188 .
therefore @xmath145 , for all @xmath189 , as required .
in order to prove the conjecture in the toric case , we need to know that we have enough small modifications to work with . although it would suffice to construct enough small modifications torically , it is perhaps helpful to observe that every small modification @xmath40 , with @xmath12 projective , is toric , in the sense that @xmath12 has the structure of a toric variety with dense torus @xmath13 such that @xmath7 is @xmath13-equivariant
. this fact follows from ( * ? ? ?
* proposition 1.11 ) ; it can also be seen directly by choosing a very ample divisor @xmath190 on @xmath12 .
then @xmath190 is the birational transform of some divisor @xmath191 on @xmath1 , which is linearly equivalent to a @xmath13-invariant divisor @xmath192 .
the natural map @xmath193 is @xmath13-equivariant , and agrees with the map to @xmath12 , embedded by @xmath190 , up to a projective linear change of coordinates . roughly speaking
, the following proposition says that , given a cone @xmath102 and a collection of rays close to @xmath102 " and surrounding @xmath102 " in @xmath74 , there is a complete simplicial fan @xmath119 containing @xmath102 , whose rays are exactly the rays of @xmath74 , such that each of the rays in the given collection is adjacent to @xmath102 in @xmath119 .
such a fan corresponds to a complete @xmath0-factorial small modification of @xmath1 .
[ small modifications ] let @xmath194 span a cone @xmath102 in @xmath74
. assume there exists @xmath195 such that 1 .
@xmath196 is linearly independent , 2 .
there is a linear relation @xmath197 with @xmath198 for all @xmath189 , 3 .
@xmath199 is not contained in the convex cone @xmath200 , for @xmath201 .
then there is a complete simplicial fan @xmath119 containing @xmath102 , whose rays are exactly the rays of @xmath74 , such that @xmath202 is adjacent to @xmath102 in @xmath119 for @xmath203 .
furthermore , @xmath119 can be chosen such that @xmath204 is projective
. choose large positive numbers @xmath205 , and small positive numbers @xmath206 .
let @xmath207 be the polytope in @xmath208 @xmath209 and let @xmath210 be the fan whose nonzero cones are the cones over the faces of @xmath207 .
it is straightforward to check that @xmath211 is a face of @xmath207 , and so @xmath212 is a cone in @xmath210 , for @xmath203 . by construction ,
the rays of @xmath210 are a subset of the rays of @xmath74 , and , provided that the @xmath213 are sufficiently general , @xmath210 is simplicial .
let @xmath119 be constructed by successive star subdivisions of @xmath210 with respect to the rays of @xmath74 that are not in @xmath210 .
so @xmath119 is also simplicial , and since , by hypothesis , none of the rays that are added lie in the cone spanned by @xmath214 , the cones @xmath212 in @xmath210 remain unchanged in @xmath119 . in particular , @xmath102 is a cone in @xmath119 , and @xmath79 is adjacent to @xmath102 in @xmath119 for @xmath203 . it remains to check that @xmath204 is projective .
since @xmath207 is a convex polytope , @xmath215 is projective , and since @xmath119 is constructed from @xmath210 by a sequence of star subdivisions , @xmath204 is constructed from @xmath215 , which is projective , by a sequence of blow ups .
so @xmath204 is also projective .
let @xmath128 be a numerical class spanning an extremal ray of @xmath216 . since @xmath217 @xmath56 spans an extremal ray of @xmath218 for some @xmath219-dimensional cone @xmath144 .
we will show that there is a face @xmath220 , a small modification @xmath40 that is birational on @xmath221 , and an irreducible curve @xmath17 sweeping out the birational transform @xmath8 on @xmath12 , such that @xmath17 represents the numerical class @xmath56 .
since @xmath56 is in @xmath218 , @xmath56 is given by @xmath141 , with @xmath147 for @xmath222 .
after renumbering , we may assume that @xmath223 let @xmath102 be the face of @xmath224 spanned by @xmath225 .
now we have @xmath226 if @xmath227 then , by proposition [ constructing curves ] , there is a curve @xmath17 moving in a family that sweeps out @xmath1 such that @xmath228 for all @xmath189 .
so we may assume @xmath229 .
since @xmath199 is not in @xmath102 for @xmath230 , we must also have @xmath231 . by propositions [ constructing curves ] and [ small modifications ]
, it will therefore suffice to show that @xmath232 is linearly independent , and that @xmath199 is not contained in the convex cone @xmath200 for @xmath201 .
suppose there is a linear relation @xmath233 .
then we have a class @xmath234 given by @xmath235 for small @xmath236 , the classes @xmath237 and @xmath238 lie in @xmath218 , but not in the ray spanned by @xmath56 , and @xmath239 can be written as @xmath240 contradicting the assumption that @xmath56 spans an extremal ray of @xmath218 . similarly , if @xmath199 is in the cone spanned by @xmath214 for some @xmath241 , then we have a linear relation @xmath242 , with all of the @xmath243 . so there is a class @xmath234 given by @xmath244 now @xmath245 is contained in @xmath246 , which is contained in @xmath216 , since @xmath247 and we have assumed @xmath229 . for small positive @xmath236 , @xmath238 is also contained in @xmath248 .
then @xmath56 can be written as @xmath249 contradicting the assumption that @xmath56 spans an extremal ray .
* acknowledgments .
* i thank m. hering , a. kronya , and m. musta for helpful conversations related to this work , and o. debarre for useful comments on an earlier draft of this paper .
i am especially grateful to r. lazarsfeld for suggesting this question , and for valuable advice .
|
we show that the dual of the cone of divisors on a complete @xmath0-factorial toric variety @xmath1 whose stable base loci have dimension less than @xmath2 is generated by curves on small modifications of @xmath1 that move in families sweeping out the birational transforms of @xmath2-dimensional subvarieties of @xmath1 .
we give an example showing that it does not suffice to consider curves on @xmath1 itself .
|
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.