article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
image compression is a topic of interest since the beginning of digital imaging , remaining relevant continuously due to the ongoing improvement in image resolution .
while standard approaches are based on orthogonal bases and frames like cosine transforms or wavelets , an alternative route based on ideas from partial differential equations has emerged recently ( cf .
@xcite ) . in the latter case particular attention
is paid to compressions from which cartoons can be reconstructed accurately avoiding the artefacts of the above mentioned standard approaches by a direct treatment of edges .
roughly speaking , their idea is to detect edges and store the image value in pixels on both sides of an edge .
the remaining parts of the image are completed by harmonic inpainting .
an alternative interpretation of the two - sided pixel values , worked out more clearly in the osmosis setting of @xcite , is that a vector field @xmath0 corresponding to the normal derivatives of the image at the edge location is stored .
the inpainting of the image @xmath1 then corresponds to a solution of @xmath2 where @xmath3 is the image domain into which @xmath0 is extended by zero off the edges . in this paper
we start from reinterpretation of the pde - based compression in terms of sparsity , which directly translates into a variational framework .
the key observation is that with the edge detection and zero extension of @xmath0 one essentially looks for a sparse vector field that leads to a certain precision in the reconstruction via ( [ eq : inpainting ] ) . in the spirit of the predominant sparsity regularization of imaging problems we study a direct variational approach
: we minimize an @xmath4-type norm of the vector field subject to the constraint that @xmath1 reconstructed via ( [ eq : inpainting ] ) approximates a given image @xmath5 up to a certain tolerance , i.e. @xmath6 as we shall discuss below a more rigorous statement of the problem takes into account that @xmath0 needs to be interpreted as a vectorial radon measure on @xmath3 ( similar to gradients of @xmath7-functions ) in a continuum setting .
the properties of a limiting continuum model appear to be of particular advantage for the compression issues , since one expects to concentrate the measure @xmath0 on a set of lebesgue measure zero .
this means that for a suitable discrete approximation of the model the number of pixels we need to store the vector field in divided by the total number of pixels tends to zero .
a direct consequence is an increase in the compression rates as the image resolution increases , a highly desirable property .
apart from compression we shall investigate models like the above one as a regularization for more general imaging problems .
the relation to denoising models can readily be seen when the above constraint of approximating @xmath5 is incorporated via a lagrange functional .
indeed , there exists a lagrange parameter @xmath8 such that ( [ eq : model0 ] ) is equivalent to @xmath9 hence , we may also interpret the norm of @xmath0 as an implicit regularization of @xmath0 and simply replace the first term by an arbitrary data term to treat other imaging problems .
this is more apparent when we replace the constraint by a natural special solution @xmath10 . in this case ( [ eq :
model1 ] ) is just the rof - model for denoising ( cf .
@xcite ) and indeed an equivalence relation holds in spatial dimension one . in higher spatial dimensions
it becomes apparent that a key role is played by the curl @xmath11 .
if the curl of @xmath0 vanishes , it simply becomes a gradient vector field and hence we again recover the rof - model .
this motivates to study further generalizations of the model also penalizing @xmath11 . with the interpretation that @xmath0 is related to the gradient of @xmath1 ,
the additional term becomes a higher order regularization effectively .
functionals of this type are currently studied in particular to reduce staircasing artefacts in total variation regularization ( cf .
@xcite ) , and indeed we shall be able to draw very close analogies to the recently very popular tgv - approach ( cf .
the reduction of staircasing is confirmed by computational experiments .
however , in the denoising case we shall see that the divergence part alone does not suffice for appropriate smoothing .
the remainder of the paper is organized as follows : in section 2 we discuss the model and its variants .
then , we proceed to a discussion of some theoretical properties in section 3 , which provide some insights into the sparse vector field model .
section 4 introduces a numerical solution based on primal - dual optimization methods , which is used for some experimental studies in section 5 .
finally , we provide a conclusion and directions for future research .
in order to obtain an appropriate formulation of ( [ eq : model1 ] ) we proceed as in @xcite and interpret @xmath0 as a @xmath12-dimensional radon measure on @xmath13 . the regularization functional is then @xmath14 where ( [ eq : inpainting ] ) is to be understood in a weak form as well .
hence , ( [ eq : model1 ] ) is rewritten as @xmath15 the model can be formulated as in recent approaches for denoising by defining @xmath16 ( in the sense of distributions ) . with @xmath17 being the characteristic function of the set @xmath18 we obtain @xmath19 one observes that the regularization functional is now an infimal convolution of total variation and a functional of @xmath20 .
the same structure is apparent in the recently popularized tgv - model ( cf .
@xcite ) , which in the analogous setting reads @xmath21 a major difference of the tgv approach to our new model is the fact that in our case only the divergence of @xmath0 respectively @xmath22 is penalized , which might be too weak to achieve suitable regularization properties .
this can obviously be realized if we add additional regularization terms depending on @xmath11 , which is natural since a divergence - free vector field is constant if and only if its curl vanishes .
note that @xmath23 , hence @xmath24 , i.e. we can formulate regularization either on @xmath0 or on @xmath22 .
the most general formulation is given by @xmath25 with convex functionals @xmath26 and @xmath27 , which however exceeds the scope of this paper and is left as subject of future research .
we finally recast the above results in terms of the regularization they induce on @xmath1 . in this respect
we also discuss the boundary conditions in ( [ eq : inpainting ] ) , respectively its weak formulation .
natural boundary conditions are no - flux conditions @xmath28 on @xmath29 , which means the used weak formulation of ( [ eq : inpainting ] ) is @xmath30 hence , we can define the regularization functional @xmath31 $ ] @xmath32 once we have defined the regularization terms it is straight - forward to extend the variational model to other imaging tasks , e.g. by just changing the data fidelity .
moreover , we can consider bregman iterations ( cf .
@xcite ) @xmath33 as well as other scale space methods such as the gradient flow ( cf .
@xcite ) @xmath34 and the inverse scale space method ( cf .
in the following we further discuss some properties of the regularization functional @xmath35 defined via ( [ eq : rdefinition ] ) . to avoid obvious technicalities with constants ,
we restrict ourselves to the space @xmath36 if @xmath3 is a bounded domain .
we start with some topological properties induced by @xmath35 : let @xmath3 be a sufficiently regular domain . then there exists a constant @xmath37 such that @xmath38 for all @xmath39 .
moreover , dom@xmath40 is a subspace of @xmath41 and @xmath35 is a norm on dom@xmath42 . finally , @xmath43 we have @xmath44 for @xmath45 we define @xmath22 as the weak solution of the poisson equation @xmath46 with homogeneous neumann boundary conditions and mean value zero .
thus , using the weak formulations we have @xmath47 regularity of solutions of the poisson equation yields continuity of @xmath22 and the existence of a constant @xmath48 such that @xmath49 for all @xmath45 .
hence , @xmath50 which yields the estimate of the @xmath4-norm .
the one - homogeneity and triangle inequality follow in a straight - forward way from the definition , consequently @xmath35 is a norm on a subspace of @xmath51 .
estimate ( [ rtvestimate ] ) is obtained since @xmath52 satisfies ( [ eq : inpaintingweak1 ] ) , hence the infimum over all admissible @xmath0 is less or equal to the total variation . a next step towards the understanding of properties of @xmath35
is an investigation of its subdifferential , with subsequent consequences for optimality conditions of ( [ eq : model1a ] ) .
for brevity we use a formal approach based on lagrange multipliers .
we have @xmath53 if and only if @xmath54 for solutions @xmath55 of the saddle - point problem @xmath56 for given @xmath1 and the lagrangian is defined as @xmath57 thus , we find @xmath58 and the optimality conditions for the saddle point problem yield ( [ eq : inpaintingweak1 ] ) and @xmath59 it is instructive to compare the subgradients of @xmath35 with those of tv . indeed with similar reasoning
one can show that @xmath60 if @xmath61 for a vector field @xmath62 and @xmath10 .
this means that if we can write @xmath63 we also obtain @xmath53 . in particular
this opens the door towards a simple verification whether solutions of the rof - model are also solutions of the sparse vector field model ( [ eq : model1a ] ) .
one simply has to inspect the subgradient in the optimality condition and check whether the associated vector field @xmath64 can be written as a gradient .
we will exemplify this in the case of the most well - known example for the rof model , the reconstruction of the indicator function of a ball on @xmath65 ( cf .
this function is an eigenfunction of tv , i.e. there exists @xmath66 ( depending on the radius @xmath35 of the ball ) such that @xmath67 it is easy to see that @xmath68 , where @xmath69 is the signed distance function of the ball ( cf .
@xcite ) and @xmath26 satisfies @xmath70 hence , we have @xmath71 for @xmath72 , which implies that @xmath73 , with the same value .
the results in @xcite imply that the variational model ( [ eq : model1a ] ) reconstructs data @xmath5 being the indicator function of a ball in the form @xmath74 , with @xmath75 depending on @xmath66 and @xmath76 .
moreover , the bregman iteration and inverse scale space methods reconstruct @xmath5 exactly after a finite number of iterations respectively finite time .
finally we return to the original idea of compressing an image by encoding a sparse vector field . for this sake
it is desireable that @xmath0 has support on a set of small ( or even zero ) lebesgue measure .
thinking about the continuum case as a limit of discrete pixel images , the asymptotic property of zero lebesgue measure means that the image ( in 2d ) can be encoded by a number of values proportional to the square root of the number of pixels .
consequently the compression rate of such a pde - based approach should improve with higher image resolution , which is highly relevant given the current trend of screen and camera resolution .
we already see from the example of the indicator function of a ball above that we can expect the method to encode a piecewise constant image by vector fields concentrated on the edge sets . for more complicated images the vector field potentially needs to have a larger support to obtain a suitable reconstruction , since away from the support of @xmath0 the function @xmath1 is just harmonic . a better understanding of the compression properties would need a characterization of the structure of minimizers , similar to @xcite . while the one - dimensional case is equivalent to total variation regularization and hence always yields @xmath0 concentrated on a set of zero lebesgue measure ( cf .
@xcite ) , the multi - dimensional case is less clear and left as an interesting topic of future research .
further studies on the compression properties will be carried out below by computational experiments .
in order to solve our minimization problem ( [ eq : model1a ] ) numerically , we at first need to discretize it . thereto we will adopt the notation of the continuous functions @xmath1 , @xmath0 , @xmath5 and the operators @xmath77 , @xmath78 and @xmath79 , but from now on , we are thereby referring to their discretized versions , which we shall comment on in the following . for simplification , we assume the normalized images to be quadratic , i.e. @xmath80^{n \times n}$ ] .
the pixel grid can be written as @xmath81 , where @xmath82 denotes the spacing size .
we use forward finite differences with neumann boundary conditions for the discretization of the gradient of @xmath1 and in order to preserve the adjoint structure the divergence is discretized with backward finite differences .
moreover , in this discrete setting the @xmath12-dimensional radon measure on @xmath13 becomes the discrete l1-norm .
considering @xmath0 as being related to the gradient of @xmath1 , the regularization term in problem can be interpreted as the discrete total variation norm .
we decided to use its isotropic version which is given by @xmath83 consequently , the minimization problem reads : @xmath84 or equivalently : @xmath85 where @xmath17 is again the characteristic function of the set @xmath18 . defining @xmath86
one can easily see that we can calculate a solution of the above problem by applying a version of the recently very popular primal - dual algorithms ( cf . )
designed for efficiently solving general minimization problems of the form @xmath87 where f and g are proper convex lower - semicontinuous functionals .
+ @xmath88 , @xmath89 $ ] @xmath90 , @xmath91 , @xmath92 @xmath93 @xmath94 @xmath95 we decided to use the first - order primal - dual algorithm as proposed by chambolle and pock ( cf .
@xcite ) given by algorithm [ alg : cp ] . adopting their notation
we can now derive the updates concerning our minimization problem .
thereto we at first have to calculate the dual functional @xmath96 . since in our case @xmath26 is the characteristic function of the set @xmath97 , it is straightforward to see that @xmath98 equals zero .
hence the update for the dual variable @xmath99 simplifies to @xmath100 .
next we consider the update for the primal variable @xmath101 . as the subdifferentials of @xmath27 with respect to @xmath1 and @xmath0
are independent of @xmath0 and @xmath1 , respectively , we can update each component of @xmath101 separately ( see algorithm 2 ) . using the norm of the sum of the primal and dual residual given by ( cf .
@xcite ) @xmath102 where @xmath103 as a stopping criterion the implementation of our minimization problem can be summarized by algorithm [ alg : model1 ] , where the two - dimensional isotropic shrinkage operator is defined by : @xmath104 image @xmath5 , @xmath8 , @xmath88 , @xmath89 $ ] , max no of iterations , @xmath105 @xmath106 , @xmath107 , @xmath91 , @xmath108 , @xmath109 @xmath110 @xmath111 @xmath112 shrinkage(@xmath113 )
@xmath114 @xmath115 as already mentioned in section [ chap : variational_model ] the model discussed so far can be further extended by considering for example bregman iterations as proposed by osher and coworkers @xcite . to incorporate this iterative regularization method in our algorithm
, we use their `` adding - back - the - noise '' formulation such that the update for @xmath1 in the previously introduced routine is replaced by @xmath116 besides , the existing implementation is extended by an outer loop over a given number of bregman iterations in which @xmath82 is updated by @xmath117 after each complete cycle of the inner loop .
in the following we present some results for the cases of image compression and denoising discussed above . as an example we chose the frequently used image `` trui '' ( @xmath118 pixels ) , making the approach comparable to previous results such as @xcite .
however , since the size of this image does not correspond to modern hd resolutions , we also created two similar images with sizes of @xmath119 and @xmath120 pixels , respectively .
we start by discussing the compression of the cartoon part of an image , which is illustrated in figure [ fig : comparisonofparametervalues ] for the trui test image .
we plot the relative error vs. the non - zero gradient ratio , which means the number of pixels with non - zero @xmath0 divided by the total number of pixels . on the left
we show a comparison of the sparse vector field ( svf ) model with the classical rof model , which demonstrates the improved compression properties . on the right
we plot the results for the variational svf model compared to the bregman iteration ( numbers correspond to bregman iterations ) , which illustrates that no significant improvement can be obtained with respect to compression by the latter . in figure
[ fig : principleofimagecompressionalgorithm ] we display the results of the compression and the corresponding vector fields for @xmath121 .
one observes that the support of @xmath0 corresponds well to an edge indicator , confirming the relation to the approach in @xcite .
the reconstructed image seems to preserve the main edges well , but does not have the strict piecewise constant behaviour as total variation regularization , which seems attractive for further reconstruction tasks .
we also investigate the behaviour for higher resolution . in order to mimic increasing resolution we simply downscale the test images to @xmath122 times the number of original pixels ,
@xmath123 $ ] .
we then perform compression at fixed error tolerance ( corresponding to constant @xmath66 when appropriately scaled ) for the images of different size and finally plot the non - zero gradient ratio vs. @xmath122 in figure [ comparisonofnon - zerogradientratios ] .
our expectation that due to continuum limit and the potential convergence towards a concentrated measure the ratio decreases with increasing resolution is well - confirmed for the trui image as well as for a similar image at higher resolution .
finally we display the result of the svf model for image compression performed on a high definition image and compare it to a jpg image with the same compression rate in figure [ fig : comparisonwithjpg_highresolutiontestjoana ] .
indeed we achieve an improved psnr with the svf model .
we also mention that several further compression steps on @xmath0 can be carried out in an analogous way to @xcite , which will lead to highly improved rates , but is beyond the scope of this paper .
+ + we finally give an outlook towards other tasks such as denoising with sparse vector fields .
for this sake we compare the model with results of the classical rof model and choose in both cases @xmath66 such that the psnr to the original image is maximized .
we illustrate the result in figure [ fig : denoisingtrui ] , which appears to be representative for all our tests .
one observes that the reduced staircasing in the svf model compared to the rof model is less visible , which is due to point like artefacts that were not present without noise .
this results in a lower psnr than for the rof model , which is consistent for all our tests .
the reason for the artefacts is that @xmath0 is too sparse in this case and does not encode the contours anymore .
this can be seen in the last image of figure [ fig : denoisingtrui ] . ,
middle left ) , svf denoising result ( @xmath124 , middle right ) , norm of vector field @xmath0 in svf model ( right).,title="fig:",height=132 ] , middle left ) , svf denoising result ( @xmath124 , middle right ) , norm of vector field @xmath0 in svf model ( right).,title="fig:",height=132 ] , middle left ) , svf denoising result ( @xmath124 , middle right ) , norm of vector field @xmath0 in svf model ( right).,title="fig:",height=132 ] , middle left ) , svf denoising result ( @xmath124 , middle right ) , norm of vector field @xmath0 in svf model ( right).,title="fig:",height=132 ]
we have introduced the svf model for image compression motivated by diffusion inpainting and found several interesting connections to tv - type regularization methods .
the svf approach leads to significantly sparser vector fields than the gradients of total variation , which appears quite useful for compression tasks , and seems not to suffer from staircasing artefacts , which appears attractive for other reconstruction tasks .
however , the denoising performance of the svf model is not convincing , since it creates point artefacts at reasonable choice of the regularization parameter .
this is probably due to the fact that the norm induced by the corresponding regularization is too weak ( there is an upper but no lower bound in terms of tv ) . for future improvement it seems natural to consider regularization on the curl of the vector field as well , such that @xmath0 becomes again concentrated on contours rather than scattered points .
in particular we suggest to study the more general problem @xmath125 for positive @xmath126 and @xmath127 where again @xmath128 ( see ) . due to the exact penalization properties of one - norms we expect that the rof model corresponds to the case of @xmath126 and @xmath127 sufficiently large , while the svf model in this paper is @xmath129 and @xmath127 sufficiently large .
this work has been supported by erc via grant eu fp 7 - erc consolidator grant 615216 lifeinverse .
mb acknowledges support by the german science foundation dfg via exc 1003 cells in motion cluster of excellence , mnster , germany .
jg acknowledges support by the cambridge biomedical research centre .
the authors thank tim lpmeier ( mnster ) and johannes hjorth ( cambridge ) for the acquisition of hd photographs used as test data .
chambolle , a. , caselles , v. , cremers , d. , novaga , m. , pock , t. .
an introduction to total variation for image analysis . in : fornasier , m. ( ed . ) , theoretical foundations and numerical methods for sparse recovery , degruyter , berlin , 263 - 340 ( 2010 ) .
mainberger , m. , weickert , j. : edge - based image compression with homogeneous diffusion . in : computer analysis of images and patterns . ed .
by jiang , x. , petkov , n. , pp .
springer , berlin , heidelberg ( 2009 ) weickert , j. , hagenburg , k. , breuss , m. , vogel , o. : linear osmosis models for visual computing .
energy minimization methods in computer vision and pattern recognition .
springer berlin heidelberg ( 2013 ) | this paper introduces a novel variational approach for image compression motivated by recent pde - based approaches combining edge detection and laplacian inpainting .
the essential feature is to encode the image via a sparse vector field , ideally concentrating on a set of measure zero .
an equivalent reformulation of the compression approach leads to a variational model resembling the rof - model for image denoising , hence we further study the properties of the effective regularization functional introduced by the novel approach and discuss similarities to tv and tgv functionals .
moreover we computationally investigate the behaviour of the model with sparse vector fields for compression in particular for high resolution images and give an outlook towards denoising .
+ + * keywords : * image compression , denoising , reconstruction , diffusion inpainting , sparsity , total variation |
granular media are large conglomerations of discrete , solid particles , such as sand , gravel , or powder , with unusual , interesting dynamics @xcite .
a one - dimensional system of spherical beads in a lattice is one of the simplest representation of granular media substrates , wherein each bead represents a grain of material . in this approximation ,
the position of a particular bead is based on forces resulting from its interaction with its two nearest neighbors @xcite .
this context has proved especially fruitful for investigating numerous aspects of the nonlinear dynamic response of such bead chain systems @xcite .
a particular focal point of emphasis has been on the study of one - dimensional granular crystals .
the availability of a wide variety of materials and bead sizes , as well as the tunability of the response within the weakly or strongly nonlinear regime renders such crystals an ideal playground for the investigation of a variety of fundamental concepts ranging from nonlinear waves and discrete breathers to shock waves , defect modes and bifurcation phenomena among many others .
however , this tunability also makes these crystals promising candidates for a wide variety of engineering applications such as shock and energy absorbing materials @xcite , actuating and focusing devices @xcite , and sound scramblers or filters @xcite .
one aspect that has not been studied , to the best of our knowledge , in such prototypical granular lattices is that of directed transport via the so - called ratcheting effect . directed ratchet transport ( drt ) , is defined as the directed transmission of an entity despite the lack of a net external force acting upon it @xcite .
this phenomenon has been associated with applications in dc current in semiconductors @xcite , the motion of fluxons in josephson junctions @xcite , bose - einstein condensates @xcite , cold atoms in optical lattices @xcite , among many others .
furthermore , drt occurring in granular systems has been associated with the study of molecular motors @xcite . as detailed in ref .
@xcite , the emergence of drt behavior is associated with the breaking of symmetries , which can be achieved by either a reshaping of the system s potential or by introducing an external forcing @xcite . for instance , drt is present when a granular material is placed in a vertically vibrating sawtooth surface profile @xcite .
typically , drt is studied ( for a single particle or a collection of particles ) when the external input acts on the system as a whole @xcite .
our aim in this work , on the other hand , is to force a _
single _ particle to achieve global drt in the context of granular crystals . in what follows in section
ii , we present the basic ( fermi - pasta - ulam type ) model that is widely accepted as representing the one - dimensional dynamics of a granular crystal @xcite with parameters that are adapted from recent experiments on the field such as refs . @xcite .
we then proceed to use the tunability of the system through actuating one bead within the chain by means of suitable biharmonic forcing that will be the source of our drt through its induced breaking of time - reversal and half - period time shift symmetries .
in particular , in section iii , we will propose a biharmonic forcing of the system involving the simplest pair of two frequencies @xmath0 relevant for such drt ( i.e. , with @xmath1 coprimes and @xmath2 odd ) , namely @xmath3 and @xmath4 @xcite . in section iv
, we will develop diagnostic quantities evaluating the relative magnitude of the clearly discernible in our numerical computations drt .
we will analyze the dependence of the induced asymmetry in the crystal response on both the frequency of the drive @xmath5 , as well as on the relative strength of the two terms in the biharmonic forcing , as controlled by the corresponding parameter @xmath6 .
the former analysis will separate different regimes in our observation of drt , namely the non - permanent deformation of the crystal that we will refer to as temporal ratcheting and the permanent deformation thereof that we will refer to as spatial ratcheting .
the clear distinction between these two regimes is an especially intriguing feature of our current setup .
the latter analysis ( over @xmath6 ) will provide a means for optimizing the ensuing transport which can both be theoretically understood and , in principle , experimentally exploited .
finally , we consider in section v , the modification of the above features in the more experimentally realistic setup incorporating dissipation .
we find there that the relevant phenomenology is modified dramatically , including even a potential reversal of the direction of the current ( for sufficiently low driving frequencies ) .
finally , in section vi we summarize our findings and present a number of directions for future study .
to ensure that the beads remain in contact , we consider a horizontal lattice that is precompressed on both ends with a force @xmath7 resulting in a static bead displacement @xmath8 ( see fig .
[ fig : beadlattice ] ) .
the existence of the precompression also serves to ensure that a linear spectrum of excitations exists in the lattice ( see details below ) . with these considerations , based on the hertzian law of spherical
point contacts , a system comprised of @xmath9 identical beads can be described by the following newtonian equation @xcite : @xmath10_+^\frac{3}{2 } - a[\delta_0+u_{i}-u_{i+1}]_+^\frac{3}{2}\ ] ] where @xmath11_+ = { \rm max}\left\{0,y\right\}$ ] , @xmath12 is the bead mass , @xmath13 is the displacement of the center of the @xmath14th bead from its equilibrium position , and @xmath15 is the hertzian constant calculated as @xmath16 where @xmath17 , @xmath18 , @xmath19 are , respectively , the bead s radius , young s elastic modulus , and poisson s ratio .
the static displacement is @xmath20 . in line with the experiments of refs .
@xcite , the parameter values listed in table [ tab : defparam ] were used .
.default parameters for bead - lattice system .
[ cols="^,^,^",options="header " , ] as we will show later , a critical factor affecting the presence / type of drt is the acoustic phonon band cutoff frequency .
plane wave solutions to the system follow the dispersion relation @xmath21 @xcite , where @xmath22 is the wave number and @xmath23 is the temporal frequency . we see that this relationship is periodic ( with period @xmath24 ) and that there is a cutoff frequency @xmath25 , above which plane wave solutions can not propagate .
the maximal frequency value occurs at the boundaries of the @xmath26 interval , which correspond to the smallest allowable wavelength .
substituting @xmath27 into the dispersion relation yields the cutoff frequency , @xmath28 therefore , @xmath29 defines the range of propagating frequencies , called the acoustic band .
frequencies @xmath30 lie within the band gap and can not propagate through the lattice as plane waves . with the parameter values given in table [ tab : defparam ]
, we have @xmath31 khz . in terms of angular frequency , @xmath32 40.31 @xmath33 , which is the critical frequency used from this point forward . notice that all the frequencies that will be mentioned hereafter will be measured in @xmath33 .
typically , drt behavior is observed in the velocity of a single particle or in that of a coherent structure such as a solitary wave @xcite . in the case of the granular crystal
though , drt will be observed ( and examined ) throughout the system as a whole .
consider a lattice where each bead begins at its equilibrium position with no initial velocity . to introduce energy into the system , the @xmath34th bead , located at the center of the lattice , is controlled by the following biharmonic , periodic function @xmath35,\ ] ] where @xmath36 is time
, @xmath37 is the amplitude , @xmath5 is the frequency , @xmath6 is the biharmonic weight , and @xmath38 is a phase . to maintain uniformity on each side of @xmath34th bead , we assume @xmath9 is odd .
the motivation for choosing to control the displacement of the central bead is that we envisage the possibility of doing experimental drt studies in the future where the position of the central bead will be controlled by an actuator .
an essential characteristic of this functional form is that it has a zero - integral over one period , indicating that the function is not biased in any direction . in other words , the @xmath34th bead
s temporal center of mass , relative its equilibrium position , is zero .
consequently , any directed behavior observed must be attributed to drt rather than a preferential direction for the input .
it is relevant to note that the prescription of the motion of the @xmath34th bead is tantamount to introducing a force , with the same characteristics , into its nearest neighbors through the equations of motion ( [ eomcomp ] ) . for @xmath39 each bead orbit on one side of
the lattice corresponds directly to an orbit on the other side of the lattice traveling in the opposite direction , translated by a half - period delay .
these orbits exactly cancel each other out and thus there is no drt . however , when @xmath40 , the symmetry of @xmath41 is broken and drt can occur @xcite .
the system is numerically solved using a fourth - order runge - kutta scheme .
the conservation of total energy is used to determine an appropriate time step .
the final integration time @xmath42 varies based on the frequency @xmath5 , but is always selected so that it is an integer multiple of @xmath43 , the period of @xmath41 .
@xmath9 also varies with @xmath5 , but is always sufficiently large so that energy from @xmath34th bead s oscillation never reaches the 1st or @xmath9th bead .
we consider values of @xmath5 ranging from 10 to 40 and set @xmath6 equal to @xmath44 . in section
[ sec : ip&r ] , we demonstrate that these parameters result in drt towards the right - hand side of the lattice . it is possible to change this direction by adding an additional phase mismatch between the two harmonics of the driver @xmath41 ( results not shown here ) . figure [ fig : varyamp ] illustrates drt magnitude ( quantification is discussed in section [ sec : ip&r ] ) for values of @xmath37 ranging from @xmath45 to @xmath46 .
these relatively small values of the forcing amplitude @xmath37 ensure that a ( relatively ) small amount of energy is introduced into the system so the beads always remain in contact with each other .
the relationship between @xmath37 and the drt magnitude is clearly nonlinear , as the average slope of the lines in the log - log plot in fig .
[ fig : varyamp ] is about 2.9725 , which indicates an essentially cubic ( gain ) relationship . based on these findings
, the remaining simulations have @xmath47 in order to exploit most of the nonlinear gain but also avoid losing contact between all beads for all times .
in order to quantify the drt displayed by the bead chain , let us define the quantity @xmath48 corresponding to the average of the hertzian forces on either side of the @xmath14th bead integrated over time .
this choice of drt measure is inspired by the fact that a piezo embedded inside a bead precisely measures the average of the hertzian forces felt by the adjacent beads .
it is important to mention at this stage that drt could be captured using many possible measures .
in fact , we also used , instead of @xmath48 , the actual forces acting on each bead and other combinations thereof and the results are qualitatively similar ( results not shown here ) .
we should point out that it is essential to consider the entire space of possible phases @xmath38 in @xmath41 .
this allows the full spectrum of the function to be sampled without biasing any direction based on the initial phase of the driver . to do this in our numerical experiments , we consider sixteen values of @xmath38 , equally spaced throughout one period of @xmath41 and define @xmath49 as the average of @xmath48 over these phases
. to create profiles which will allow drt behavior to be observed , we compare the normalized difference of @xmath49 for pairs of beads equidistant from the center bead , that is @xmath50 where @xmath51 ) . by monitoring @xmath52 over time ,
a drt profile is constructed ( see fig . [
fig : impprof ] ) . + + + in fig .
[ fig : impprof ] , for each value of @xmath53 , a contour plot illustrating the drt profile as a function of @xmath54 and @xmath36 is provided .
additionally , the right panels show the asymmetry indicator profile of eq .
( [ impdif ] ) ) at a set of particular times , written in terms of oscillations of the center bead .
a non - zero value of @xmath52 indicates the preferential transport of force in one direction , that is , drt .
we see that , after transient behavior , all significant values of @xmath52 are negative , indicating the presence of drt towards the right hand side of the lattice ( see eq . ) . figures [ fig : impprof](a),(b ) , illustrate the drt profiles for @xmath55 , where both @xmath5 and @xmath56 are below @xmath57 . after a transient time interval
, we observe a drt `` wave '' , or cone , advancing as time progresses ( and leaving no drt behind it ) .
for this reason we call this behavior
_ temporal drt_. for a given time , let @xmath58 denote the value of @xmath54 at the outer edge of the cone ( that effectively travels at the speed of sound within the medium ) , @xmath59 denote the value of @xmath54 at the inner edge of the cone , and @xmath60 denote the value of @xmath54 for which @xmath61 , and therefore drt , is maximal . for @xmath62 ,
we have @xmath63 since the energy from the forcing function has not yet reached beads offset this far from the center bead .
for the region defined by @xmath64 , there exists a positive , approximate - linear relationship ( with respect to @xmath54 ) describing the magnitude of the drt .
similarly , for @xmath59@xmath65 , there exists a negative approximate linear relationship describing drt magnitude .
the drt wave has already moved through the region defined by @xmath66 .
the characteristic property of this class of behavior , which we call class i , is that for @xmath66 , @xmath67 , indicating drt is _ no longer present _ shortly after the wave has left a region .
class ii behavior is observed by considering a forcing frequency of 20 , as shown in figs .
[ fig : impprof](c),(d ) . here
, @xmath57 is greater than @xmath5 and slightly larger than @xmath56 . after allowing for transient time
, the regions defined by @xmath62 , @xmath64 , and @xmath59@xmath65 , display the same qualitative behavior as the previous case .
however , what clearly distinguishes this region is that for @xmath68 we see a qualitatively different result , namely , for each @xmath54 , @xmath52 is approximately equal to a nonzero constant .
this is indicative of an equilibrium drt state defined by the spatial extent of the region through which the ratcheting wave has already passed .
this effect and the `` kink''-like pattern that it leads to ( rather than the pulse like structure of class i ) in the context of the asymmetry indicator @xmath52 is hereafter referred to as _
spatial drt_. this fundamental distinction of regimes of temporal and spatial drt is , arguably , one of the most interesting traits observed herein , and to our knowledge , has not been reported before , although we believe that it should be more general than the particular realization considered herein . in figs .
[ fig : impprof](e),(f ) , @xmath5 = 30 and thus @xmath69 and @xmath70 . as a result , a different behavior that will be characterized hereafter as belonging to class iiia is observed .
the features are similar to class ii , but now the magnitude of the spatial drt is approximately equal to @xmath71 , the maximal temporal drt magnitude
. otherwise said , the tail of the kink associated with the asymmetric deformation of the lattice , rather than having the linear profile of class ii , it is essentially flat .
as shown in figs .
[ fig : impprof](g),(h ) , where @xmath72 , as @xmath5 approaches @xmath57 , the drt profile remains qualitatively the same , but the slope of the ( approximately ) linear relationship for @xmath64 decreases . in our kink - based visualization of the corresponding @xmath52 s , this regime
is associated not with the translation of the structure over the lattice which roughly preserves its shape , as in class iiia .
instead , it appears associated predominantly with the widening of the relevant spatial structure in this regime that we will refer to as class iiib . for @xmath73 , neither of the plane waves comprising @xmath41 can propagate . as a result ,
the spatial and temporal drt behavior breaks down .
this can be discerned in figs .
[ fig : freq - drt ] , and [ fig : freq - vel ] where the drt magnitude and velocity ( that we will now proceed to define more precisely ) approach zero as @xmath5 approaches @xmath57 . to explore the effects of parameter variation on the magnitude of the drt
, we now establish an additional ratcheting metric .
for each of the drt profiles under consideration , @xmath74 represents the maximal amount of temporal ratcheting at any given time .
we use this value as the temporal drt metric . on the other hand , when spatial ratcheting is present , it is first observed by comparing beads adjacent to the @xmath34th bead , that is at @xmath75 . as time progresses , the behavior spreads out from the center bead and is also observed for larger values of @xmath54 .
we observe that @xmath52 is approximately the same for each @xmath54 exhibiting spatial drt behavior ; therefore the magnitude of spatial ratcheting can be quantified by means of @xmath76 .
both @xmath74 and @xmath76 vary slightly over time .
nevertheless , we have verified that this variation is small and , therefore , choose to measure @xmath76 and @xmath74 at the final integration time @xmath77 .
figure [ fig : freq - drt ] illustrates the magnitude of spatial and temporal drt , based on the above diagnostic , for a wide range of representative frequencies . for class
i ( @xmath78 ) , the magnitude of temporal drt slowly increases with the frequency .
this regime corresponds to both input frequencies of the forcing ( @xmath5 and @xmath56 ) being below the cutoff frequency @xmath79 . as it can be noticed in fig .
[ fig : freq - drt ] , the spatial drt starts appearing once the second harmonic ( @xmath56 ) of the driver gets closer to the cutoff frequency ( see left vertical dashed line ) .
this seems to be an effect of the nonlinear response of the system that `` widens '' the region of the cutoff frequency .
in fact , class ii corresponds to the region of frequencies where the second harmonic of the driver transitions from being transmitted to completely being stopped due to the cutoff frequency .
it is interesting that for class ii and iiia , defined by @xmath80 corresponding to @xmath69 but @xmath56 close to ( under or over ) @xmath57 , the temporal drt magnitude is approximately constant .
however , as the first harmonic ( @xmath5 ) starts getting close to the cutoff frequency ( see right vertical dashed line ) , naturally , both spatial and temporal drt start to disappear and eventually vanish , as expected , once both , first and second , driver harmonics are inside the forbidden gap .
the effect of the first harmonic starting to approach the cutoff frequency begins at , approximately , @xmath81 , corresponding to the onset of class iiib behavior , the magnitude of temporal drt begins to sharply decrease as @xmath5 approaches @xmath57 .
this lack of drt for higher frequencies is consistent with the drt breaking down for @xmath73 .
on the other hand , spatial drt significantly increases as we move from class i to class ii and subsequently iiia , and it also , in turn , sharply decreases in the case of class iiib .
we define the _ outer - cone horizon _ as the location of the onset of temporal drt and the _ inner - cone horizon _ as the location of the onset of spatial drt .
the velocities of the outer and inner horizon are calculated numerically for values of @xmath5 ranging from 10 to 40 . for each frequency
, the velocity of the outer - cone horizon approximately corresponds to the sound velocity of the system , @xmath82 ( see filled dots in fig . [
fig : freq - vel ] ) .
this is a consequence of the nonlinearity of the system that `` mixes '' the frequencies introduced by the forcing and thus excites all modes .
in contrast to the behavior observed by the outer - cone velocity , as shown in fig . [
fig : freq - vel ] ( see open dots ) , the inner - cone velocity strongly depends on the forcing frequency . for small values of @xmath5 , where both @xmath5 and @xmath56 are smaller than @xmath57 , there is essentially no spatial drt , as the two cones propagate with essentially the same speed forming the `` pulse '' observed in the asymmetry indicator @xmath52 .
as the frequency increases through class i and toward class ii , the velocity of the inner - cone horizon decreases significantly until the threshold @xmath70 is crossed . at that point
the inner - cone velocity approaches zero . within class ii , the continuously decreasing velocity of the inner - cone forms the tail of the kink discussed in connection to fig .
[ fig : impprof ] . past the point of @xmath83 ,
the inner - cone velocity abruptly increases to a value similar to that for the lower frequencies .
subsequently , in class iiia , the velocity decreases as the case where both @xmath5 and @xmath56 are larger than @xmath57 is approached .
class iiib corresponds to a larger rate of decreasing velocity .
interestingly , the spatio - temporal wave velocities are independent of the choice of @xmath6 .
it is interesting to point out at this stage that the presence of drt in our system is a direct consequence of the symmetry breaking provided by the external forcing when @xmath84 [ see eq .
( [ biharmonic2 ] ) ] .
however , it is also important to note that nonlinearity is also a key ingredient for the presence of drt .
in fact , if the hertzian forces in eq .
( [ eomcomp ] ) are replaced by linear ( hooke ) springs , drt is no longer present .
the absence of drt for the linear force case is a consequence of the fact that the drt corresponding to a phase @xmath38 is the negative of the one for phase @xmath85 , where @xmath43 is the period of the driver , and thus providing a cancellation when drt is averaged through all possible phases of the driver . finally , it is also important to stress that , as discussed earlier and depicted in fig .
[ fig : varyamp ] , the magnitude of drt is , approximately , proportional to the cube of the forcing amplitude .
therefore , nonlinearity in the system is not only necessary for observing drt , but it also provides a nonlinear enhancement of the drt magnitude with respect to the input amplitude .
@xmath86{fig6a.eps } \\
\includegraphics[width = 5.5cm]{fig6b.eps } \end{array}$ ] in ref .
@xcite , it was shown that when considering the forcing function @xmath41 , the magnitude of drt behavior is due to two competing effects : the increase in the degree of symmetry breaking and the decrease in the transmitted impulse over a half - period , which is denoted as effective symmetry breaking .
it was shown that the drt behavior is optimally enhanced for @xmath87 .
we now demonstrate that this result holds for the granular crystal case under consideration .
we consider forcing frequencies of @xmath5 = 17.5 and 30 .
figure [ fig : eta - drt ] illustrates the spatio - temporal drt magnitude as a function of @xmath6 . as is expected , for @xmath88 , the single harmonic does not induce drt behavior . as @xmath6 increases , the magnitude of drt increases until a maximum value for ratcheting is reached at @xmath89 ( in our case since drt is towards the left , the maximum ratcheting effect corresponds to a minimum for drt ) , with the exception of spatial drt for @xmath5 = 17.5 .
the magnitude then decreases until drt is again not present at @xmath90 . to further explore and quantitatively appreciate this result ,
consider a generic biharmonic forcing function @xmath91 , where @xmath92 are amplitudes , @xmath93 are phases , @xmath5 is the frequency and @xmath94 and @xmath95 are coprimes .
it can be shown that the ratchet velocity @xmath96 where @xmath97 is a system - dependent constant @xcite . with the parameters in @xmath41
, we have @xmath98 .
observe that this function has a minimum ( for @xmath99 ) at @xmath87 , which is consistent with the numerical simulations . by setting @xmath100 equal to the numerically - calculated drt magnitude at @xmath87
, we can solve for the free parameter @xmath97 .
these curves are shown in fig .
[ fig : eta - drt ] .
there is a striking agreement between the calculated numerical drt magnitudes and theoretical curves .
we see less of a correspondence for spatial drt for @xmath101 . in general , for other frequencies considered
, the temporal drt matched the theoretical curves more consistently than spatial drt , particularly for @xmath5 near @xmath102 and @xmath57 .
as a side note , it is possible to draw physical intuition for the optimal biharmonic weight being at 2/3 if one considers the ideal ratcheting forcing : a sawtooth function . expanding a sawtooth function in fourier series and keeping only the first two harmonics it is straightforward to show that their ratio is 2 . in our case , when @xmath103 , we precisely get a ratio between the two harmonics of @xmath104 . in other words , the optimal ratcheting forcing , i.e. @xmath103 , is the best possible approximation to a sawtooth function when using two harmonics .
to more closely represent physical reality , dissipation is introduced into the uniform granular crystal by augmenting eq . :
@xmath105_+^\frac{3}{2 } - a[\delta_0+u_{i}-u_{i+1}]_+^\frac{3}{2}-\frac{m}{\tau}\dot{u}_i,\ ] ] where @xmath42 is a dissipation constant set equal to @xmath106 .
this value was chosen to match the dissipation constant used in the experiments of ref .
@xcite . to investigate the effects of friction
, we numerically integrate the four representative frequency cases presented earlier and illustrate the corresponding results in fig .
[ fig : impproffric ] . in order to allow sufficient time for transient behavior , @xmath77 , the final integration time , and
@xmath9 were considerably larger than for the non - friction simulations . + + + when friction is present , we observe that all cases exhibit qualitatively similar behavior .
namely , unlike the frictionless case , the velocity of the temporal drt is not set by the sound velocity of the system , but instead decreases as time progresses .
furthermore , in all dissipative cases , we have observed that spatial drt is weakened by the presence of dissipation and that all cases present a similar spatial structure which seems to involve a progressively widening ( i.e. , dispersing ) kink state .
it is worth noticing that for @xmath107 , the drt profile is slightly different than for the other cases .
the maximal temporal drt value , @xmath74 now occurs at @xmath108 .
furthermore , as time increases , @xmath109 for @xmath110 .
this is indicative of a small amount of spatial drt .
in fact , the behavior is similar to class iiib for the frictionless system . for @xmath55 ,
the presence of friction radically changes the characteristics of the drt profiles .
temporal drt is present , but for @xmath111 periods , all non - zero values of @xmath52 are now positive , indicating that the drt direction has _ reversed _ and now favors left propagation . for @xmath112 , the minimum of @xmath52 is less than zero illustrating the remnants of rightward drt . in each of the drt profiles provided in fig .
[ fig : impproffric ] , we observe two regimes of drt propagation . initially , the temporal drt `` wave '' travels at the sound velocity of the system .
however , as time progresses , the dissipation tend to slow down the propagation of this wave .
similarly , dissipation is also responsible for progressively damping out the drt magnitude .
eventually , a steady state solution is reached where the energy being pumped into the system by the forcing is balanced by dissipation .
finally , it is worth mentioning that , despite taking the optimal value for the forcing amplitude to exploit maximum drt gain , the drt magnitude is on the order of @xmath113@xmath114 which amounts to a @xmath115@xmath116 of biased transport between left and right propagation .
although these values are relatively small , drt should be possible to measure in current experimental setups .
in this work , a one - dimensional granular chain ( crystal ) was considered where the position of the center bead was prescribed by a biharmonic forcing function .
this functional form is known to induce ratcheting . yet , in our case , a distinguishing characteristic was the system - wide emergence ( i.e. in space - time ) of directed ratchet transport ( drt ) in the force profiles .
the regimes where temporal ( transient ) ratcheting and spatial ( i.e. , with a permanent spatial `` imprint '' over the lattice ) ratcheting were identified as a function of the system s frequency .
the relationship between the frequencies @xmath5 and @xmath56 of the forcing function and the cutoff frequency @xmath57 of the system determined the characteristics of the observed drt and its separation into different classes . in the class i pertaining to temporal ratcheting , a drt
wave " traveled away from the center bead at the sound velocity of the system . once the temporal drt wave moved through a region , a steady - state was induced in this region , wherein all bead pairs exhibited similar drt magnitude . if this value was non - zero , it corresponded to class ii and the so - called spatial drt .
the modification of the form of spatial ratcheting past the regime where @xmath83 gave rise to yet another regime that was referred to as class iii .
the frequency , @xmath5 , and biharmonic weight , @xmath6 of the forcing function were varied so that the response of the magnitude of drt and the velocity of the drt waves " could be determined .
while the wave velocity was independent of the biharmonic weight , @xmath87 maximized the magnitude of spatio - temporal drt , in accordance with the expectations of refs .
friction was subsequently introduced into the system , leading to weakening of the ratcheting effect and a rather uniform spatial form of its profile in classes ii and iii .
yet , it was class i that was most significantly affected by the inclusion of friction within the system , which resulted in drt switching directionality from right to left . having paved the way for the consideration of ratchet effects in granular crystals , there are numerous directions along which the present study can be extended .
it is certainly of interest to attempt to expand the range of considered materials and parameters ( as well as that of heterogeneous systems such as dimers , trimers @xcite ) and of a wider range of forcing frequencies and displacement parameters .
a key aspect of such a broader parametric effort is to try to maximize the relevant drt , so as to render it more accessible to potential experiments .
another important direction of particular interest is to attempt to expand the present considerations to the realm of higher dimensional granular crystals .
recent efforts have made these gradually more accessible to experimental investigations @xcite and hence such ratcheting efforts would be extremely timely and relevant to consider .
carlin and y.k .
proceedings of the ieee , * 53 * , 11 ( 1965 ) 1788 .
f. falo , p.j .
martnez , j.j .
mazo , and s. cilla , europhys .
lett . , * 45 * ( 1999 ) 700 . f. falo , p.j
. martnez , j.j .
mazo , t.p .
orlando , k. segal , and e. tras , appl .
a , * 75 * ( 2002 ) 263 .
ustinov , c. coqui , a. kemp , y. zolotaryuk , and m. salerno .
* 93 * ( 2004 ) 087001 .
f. jlicher , a. ajdari , and j. prost .
* 69 * ( 1997 ) 1269 .
s. flach , o. yevtushenko , and y. zolotaryuk .
* 84 * ( 2000 ) 2358 .
p. reimann .
phys . rep .
, * 361 * ( 2002 ) 57 .
rapaport comp .
147 * ( 2002 ) 141 .
n. boechler , g. theocharis , s. job , p.g .
kevrekidis , m.a .
porter , and c. daraio .
* 104 * ( 2010 ) 244302 . | directed - ratchet transport ( drt ) in a one - dimensional lattice of spherical beads , which serves as a prototype for granular crystals , is investigated .
we consider a system where the trajectory of the central bead is prescribed by a biharmonic forcing function with broken time - reversal symmetry . by comparing the mean integrated force of beads equidistant from the forcing bead , two distinct types of directed transport can be observed _spatial _ and _ temporal _ drt . based on the value of the frequency of the forcing function relative to the cutoff frequency ,
the system can be categorized by the presence and magnitude of each type of drt .
furthermore , we investigate and quantify how varying additional parameters such as the biharmonic weight affects drt velocity and magnitude . finally , friction is introduced into the system and is found to significantly inhibit spatial drt .
in fact , for sufficiently low forcing frequencies , the friction may even induce a switching of the drt direction . |
the question on the formation of the present - day massive spheroidals is one of the most debated issues of galaxy evolution and it is strongly linked to the general problem of structure formation in the universe ( see @xcite for a recent review ) . in one scenario , massive spheroidals are formed at early cosmological epochs ( e.g. @xmath0 ) through the `` monolithic '' collapse of the whole gas mass@xcite .
such a formation would be characterized by an episode of intense star formation , followed by a passive evolution ( or pure luminosity evolution , ple ) of the stellar population to nowadays . in marked contrast
, the hierarchical scenarios predict that massive spheroidals are the product of rather recent merging of pre - existing disk galaxies taking place mostly at @xmath1@xcite . in hierarchical scenarios ,
fully assembled massive field spheroidals at @xmath2 are rare objects@xcite , and the spheroids of cluster ellipticals were assembled before those of field ellipticals@xcite . from an observational point of view ,
a direct way to test the above scenarios is to search for massive field ellipticals at @xmath2 and to compare their number with the model predictions ( see the introduction of @xcite for a recent review on observational tests ) .
since the near - ir light is a good tracer of the galaxy mass@xcite , @xmath3-band imaging provides an important possibility to perform surveys aimed at selecting massive ellipticals at high-@xmath4 . a galaxy with a stellar mass of about @xmath5 m@xmath6 is expected to have @xmath7 for @xmath8@xcite , thus implying that moderately deep @xmath3-band surveys can efficiently select massive galaxies .
a first selection criterion to find ellipticals at @xmath2 is to apply a color threshold to @xmath3-band selected galaxies . in the framework of passive evolution , such a threshold
is set by the colors expected for a galaxy at @xmath2 formed at a given @xmath9 .
for instance , according to the bruzual & charlot ( 1999 ) spectral synthesis models ( @xmath10 , salpeter imf ) , a very red color of @xmath11 would allow to select @xmath12 passively evolving galaxies formed at @xmath13 ( @xmath14=50 km s@xmath15 mpc@xmath15 , @xmath16 ) , thus allowing to search for elliptical candidates formed at early epochs .
however , searches based on color selection criteria gave discrepant results : some works showed that the number of such red galaxies is lower compared to the predictions of ple@xcite , whereas others did not confirm such a deficit up to redshifts of about two@xcite . in order to avoid the possible biases ( e.g. star formation ) present in the color selection technique , another approach is to derive the fraction of ellipticals by taking spectra of all the galaxies in @xmath3-selected samples irrespectively of colors@xcite ( see also http://www.arcetri.astro.it/@xmath17k20/ ) . this method allows to overcome the putative problem of ellipticals missed because bluer than the adopted color threshold due to a low level of residual star formation@xcite . finally , the third possibility to find @xmath2 ellipticals is to select galaxies according to their morphology and surface brightness profiles ( with or without an associated color selection criterion ) .
this approach was adopted for example by @xcite for @xmath18 , and by @xcite for @xmath2 .
while the results seem to agree with no or little number density evolution for early type galaxies at @xmath1 , the analysis of the @xmath2 samples led again to discrepant results , thus making the question on @xmath2 ellipticals even more controversial .
after several small field surveys ( @xmath171 - 60 arcmin@xmath19 , see references in previous section ) leading to discrepant results , the most recent success was provided by a wide field survey for extremely red objects ( eros)@xcite . such a survey ( the widest so far : 700 arcmin@xmath19 to @xmath20 , with a sub - area of 447 arcmin@xmath19 to @xmath21 ) provided a complete sample of about 400 objects with @xmath22 suitable for reliably constraining the number density of high-@xmath4 elliptical candidates .
the main results of such a survey are the detection of strong angular clustering of eros ( an order of magnitude larger than that of field galaxies ; see fig . 1 - 2 ) , and the accurate estimate of the surface density of elliptical candidates at @xmath2 .
the observed clustering can explain the previous discrepant results on the surface density of @xmath2 ellipticals as due to strong field - to - field variations ( the `` cosmic variance '' ) , and it suggests that most eros are ellipticals rather than dust reddened starbursts ( see @xcite for more details ) .
finally , even in the conservative case where up to 70% of eros are _ not _ ellipticals , the observed surface density ( complemented by the results of @xcite ) is in good agreement with the predictions of ple ( fig .
3 ) , suggesting that most field ellipticals were fully assembled at least by @xmath23@xcite .
this result does not imply that the formation of massive spheroidals occurred necessarily through a `` monolithic collapse '' scenario , but it simply constrains the epoch when the formation took place , and it implies that , if ellipticals formed through merging , this occurred mostly at @xmath24 .
c c c in addition to wide field imaging , the existence of galaxies with the colors expected in the case of passive evolution and with de vaucouleurs @xmath25 surface brightness profiles consistent with being dynamically relaxed spheroidals at @xmath2 has significantly grown thanks to hst deep imaging ( e.g @xcite ) .
the spectroscopic confirmation of high-@xmath4 ellipticals is extremely challenging because of their faintness both in the optical and in the near - ir , and because of the few characteristic spectral features present in their spectra : mainly the strong 4000 continuum break , the weaker breaks at 2600 , 2900 and 3260 , and a handful of absorptions detectable with the present - day largest telescopes and very long integration times ( fig .
moreover , for @xmath26 , the 4000 break falls in a critical spectral region where the optical and near - ir spectrographs are less efficient and the atmosphere severely hampers the observations . despite such difficulties , the keck telescopes and
the eso vlt are confirming the existence of @xmath2 passively evolving ellipticals with old ages ( @xmath27 gyr ) consistent with being formed at remote cosmological epochs@xcite ( fig .
4(a ) ) , as well as ellipticals displaying a low level of star formation indicated by the possible detection of weak [ oii]@xmath283727 emission ( see fig .
4(b ) ; see also@xcite )
the major problem affecting the statistical studies of @xmath2 ellipticals is the `` pollution '' of color - selected samples by a fraction of star forming galaxies reddened by strong dust extinction@xcite .
the fraction of dusty eros is currently unknown , but its accurate estimate is crucial to infer a reliable surface density of high-@xmath4 ellipticals `` cleaned '' by the contamination of dusty systems .
this goal can be reached observing complete samples of eros with optical+near - ir spectroscopy ( when feasible ) , deep hst imaging and submm photometry . although based on small and/or incomplete samples , recent submm and hst observations suggested that the dusty galaxies are probably segregated among the reddest eros with @xmath29 or @xmath30@xcite .
the other major problem in the identification of @xmath2 ellipticals is that a large fraction of eros are beyond the spectroscopic limits of the present largest telescopes ( e.g. @xmath31 , @xmath32 ) .
this strongly limits our ability to spectroscopically confirm the nature of a high-@xmath4 elliptical candidate .
the large binocular telescope will play an important role in the study of high-@xmath4 ellipticals . as a single - dish telescope , deep spectroscopy in the ranges 0.7 - 1.0@xmath33 m and 0.9 - 1.8@xmath33 m will be possible with the mods and the lucifer spectrographs respectively .
this will allow us to enlarge the samples of spectroscopically identified high-@xmath4 ellipticals and , for instance , to infer their luminosity function and 3d clustering at @xmath2 . as a diffraction limited telescope
, the lbt will provide high angular resolution near - ir images that will be crucial in deriving the surface brightness profiles and , for instance , to extend the kormendy relation and to attempt the tolman test at @xmath2 .
finally , in case of objects unfeasible with spectroscopy , the combination of high resolution imaging and optical + near - ir photometry ( possibly done with _ special medium band _ filters ) will allow us first to confirm the spheroidal nature of the elliptical candidates , and then to reliably estimate their photometric redshifts .
99 renzini a. , cimatti a. 2000 , in the hy - redshift universe : galaxy formation and evolution at high redshift " , ed .
bunker & w.j.m .
van breugel , a.s.p .
series vol .
193 , in press ( astro - ph/9910162 ) eggen o.j . , lynden - bell d. , sandage a. 1962 , apj , 136 , 748 larson r.b .
1974 , mnras , 173 , 671 kauffmann g. 1996 , mnras , 281 , 487 baugh c.m . , cole s. , frenk c.s .
1996 , mnras , 283 , 1361 gavazzi g. , pierini d. , boselli a. 1996 , a&a , 312 , 397 kauffmann g. , charlot s. 1998 , mnras , 297 , l23 schade d. et al . 1999 ,
apj , 525 , 31 zepf s.e .
1997 , nature , 390 , 377 franceschini a. et al .
1998 , apj , 506 , 600 barger a.j .
et al . 1999 ,
aj , 117 , 102 totani t. , yoshii j. 1997 , apj , 501 , l177 benitez n. et al . 1999 ,
apj , 515 , l65 broadhurst t.j .
, bouwens r.j . 1999 ,
apj , 530 , l53 scodeggio m. , silva d.r .
2000 , a&a , 359 , 953 jimenez r. et al .
1999 , mnras , 305 , l16 cowie l.l .
1996 , aj , 112 , 839 cohen j.g .
et al . 1999 ,
apj , 512 , 30 eisenhardt p. et al .
2000 , in the birth of galaxies " , xth rencontres de blois , ed .
b. guiderdoni et al .
( astro - ph/0002468 ) cimatti et al .
2001 , in preparation ( see http://www.arcetri.astro.it/@xmath17k20/ ) menanteau f. et al .
1999 , mnras , 309 , 208 treu t. , stiavelli m. 2000 , apj , 524 , l27 moriondo g. , cimatti a. , daddi e. 2000 , a&a , in press daddi e. et al .
2000 , a&a , in press thompson d. et al . 1999 ,
apj , 523 , 100 daddi e. , cimatti a. , renzini a. 2000 , submitted stiavelli m. et al . 1999 ,
a&a , 343 , l25 .
spinrad h. et al .
1997 , apj , 484 , 581 liu m.c .
2000 , aj , 119 , 2556 soifer b.t . 1999 ,
aj , 118 , 2065 cimatti a. et al . 1999 ,
a&a , 352 , l45 cimatti a. , andreani p. , rttgering h. , tilanus r. 1998 , nature , 392 , 895 dey a. et al . 1999 ,
apj , 519 , 610 smail i. et al .
1999 , mnras 308 , 1061 andreani p. , cimatti a. , loinard l. , rttgering h.j.a .
2000 , a&a , 354 , l1 gear w.k .
2000 , mnras , in press ( astroph/0007054 ) | the most recent observational results on the search for high redshift field ellipticals are reviewed in the context of galaxy formation scenarios .
the perspectives for large binocular telescope ( lbt ) observations are also discussed . |
cold dark matter ( cdm ) theory successfully describes many aspects of the formation of large - scale structure in the universe @xcite .
however , mismatches do exist between its predictions and observations , such as the cusp - core controversy , missing satellites @xcite and the angular momentum problem @xcite . in particular
, the cusp - core issue has provoked much debate .
cdm simulations consistently yield density profiles with steeper inner slopes ( power - law exponent between @xmath71 and @xmath71.5 ) than observational studies which have found a range of slopes , including constant density cores in dark matter dominated low surface brightness galaxies @xcite and shallow slopes in clusters with gravitationally lensed arcs @xcite . these results , among others , have initiated discussion about the role baryons play in softening simulated cores @xcite and the observational effects that may mask cusps in low surface brightness galaxies @xcite .
recent @xmath0-body results demonstrate that , at the current resolution of simulations , the central power - law exponent does not converge to a universal value @xcite . in numerical simulations ,
a softened gravitational potential is used to prevent the macro - particles ( 10@xmath810@xmath9 ) from experiencing artificially strong two - body interactions ( * ? ? ?
* for example ) . the softening length , @xmath2 , is chosen to maximise the resolution while suppressing two - body effects over the simulation running time
in view of the ongoing disagreement regarding the form of @xmath3 , it is important to clarify _ analytically _ , by a code - independent argument , whether the choice of @xmath2 affects the physics of the system and hence @xmath3 . in this paper , we employ the framework of statistical mechanics @xcite , drawing upon recent results on phase transitions in @xmath0-body systems with attractive power - law potentials @xcite .
self - gravitating particles behave qualitatively differently to many other statistical systems because gravity is an unscreened , long - range force .
they are best examined within the microcanonical ensemble , where the energy and number of particles are fixed and phases with negative specific heat are allowed .
we apply the results of the classical theory of self - gravitating , @xmath0-body systems to demonstrate the effects of introducing a short distance cutoff in numerical simulations , in particular the effect on stability .
the study of the thermal stability of self - gravitating systems has a long history .
@xcite showed that spherical systems of point particles in a box with reflecting walls are gravitationally unstable below a critical temperature , collapsing catastrophically to a central point .
@xcite generalised this work to a spherical system of @xmath0 classical hard spheres in contact with a heat bath , showing the gravothermal instability to be a general feature of self - gravitating systems held at a constant temperature .
@xcite , investigating point fermions obeying the pauli exclusion principle , showed that a _ stable _ low temperature phase can exist if the gravitational potential is softened , transforming the gravothermal instability to a phase transition to the low temperature phase . in this paper , we extend these results and use them to reinterpret some of the ambiguous results of numerical simulations of cdm haloes discussed above .
a detailed comparison with preceding analytic work is presented in section [ comparison ] .
section 2 briefly reviews the formalism for treating @xmath0 self - gravitating , collisionless particles statistically . in section 3 , we apply the formalism to compute @xmath3 analytically as a function of @xmath1 , the total energy , and @xmath2 .
the result is a thermodynamic phase diagram that contains both collapsed and extended haloes . in section 4 , we locate published @xmath0-body simulations on the phase diagram and show they are biased towards the collapsed phase .
this phase is unstable for @xmath10 , suggesting that collapsed haloes are an artificial by - product of the softened potential ; there is no immediate reason to expect agreement between simulated and observed profiles unless the gravitational potential is appreciably softened in nature .
we emphasise at the outset that it is not our intention to reproduce realistic cdm haloes with non - zero angular momentum and hierarchical clustering ; rather , we demonstrate in a code - independent manner how the softening used in @xmath0-body simulations may artificially alter the density profiles found .
the properties ( e.g. energy , entropy ) and collective behaviour ( e.g. gravothermal catastrophe ) of a self - gravitating gas of cdm particles in thermodynamic equilibrium take different values when computed in different statistical ensembles because the long - range nature of the gravitational potential renders the system inseparable from its environment @xcite . in this paper , we follow previous studies by considering the self - gravitating gas in the microcanonical ensemble ( mce ) , whose features are constant energy , volume and particle number .
particles do not evaporate from the system over time and the walls of the container are perfectly reflecting .
the mce is more appropriate than the canonical ensemble ( ce ) for three reasons : ( i ) it is unclear how to construct an external heat bath ( required by the ce ) for a long - range potential , because the system interferes with the environment @xcite ; ( ii ) states with negative specific heat are inaccessible in the ce @xcite ; and ( iii ) the equilibrium density profile in the violently relaxed ( smoluchowski ) limit is the singular isothermal sphere in the ce , contrary to observations @xcite .
the density of states , @xmath11 , is the volume of the ( 6@xmath121)-dimensional surface of constant energy @xmath1 in phase space ( @xmath13 ) , where ( @xmath14 ) are the co - ordinates and momenta of the @xmath15-th particle . at any one moment ,
the system occupies one point in the 6@xmath0-dimensional phase space . for particles of equal mass @xmath16
, one has @xmath17d^{3n}pd^{3n}x,\ ] ] where the first and second sums give the kinetic and potential energy , and the integral is over phase space volume .
the gravitational potential , @xmath18 , is given by @xmath19 or , if the potential is artificially softened over a characteristic length @xmath20 , by @xmath21^{-1/2}$ ] .
the thermodynamic entropy @xmath22 ( up to a constant ) and the temperature @xmath23 of the system are defined in terms of @xmath11 : @xmath24 @xmath25 these quantities are hard to interpret when assigned to a system far from equilibrium .
note that @xmath11 diverges for @xmath26 and @xmath27 ; any two particles can be brought arbitrarily close together , liberating an infinite amount of potential energy , so that the co - ordinate space integral diverges @xcite .
this is a serious problem because it is impossible to achieve thermodynamic equilibrium if @xmath11 diverges ; the system does not have time to sample the infinite number of possible microstates with equal probability @xcite . if the dark matter particles are fermions , the pauli exclusion principle does prevent this problem .
however , the fraction of the phase space volume sampled by @xmath0 mildly relativistic cdm particles in a time @xmath28 , given by @xmath29 , is exceedingly small for most proposed cdm particles , e.g. @xmath30m@xmath31/gev for self - interacting dark matter @xcite .
the density of states is evaluated in the continuum ( mean - field ) limit by integrating over momentum and then expressing the remaining configurations as a functional integral over possible density profiles @xmath32 @xcite , @xmath33,\ ] ] where the effective dimensionless action @xmath34 and dimensionless energy @xmath35 are quantities defined by @xcite . in ( [ functional ] ) and ( [ action ] ) , and throughout the remainder of this paper , the density profile and position co - ordinates @xmath36 are written as dimensionless quantities , relative to the total mass @xmath37 and outer radius @xmath38 of the system , with @xmath39 and @xmath40 . upon evaluating the functional integral by a saddle point method ( which involves extremising the action ) , ( [ functional ] )
reduces to three coupled integral equations describing the density profile @xmath32 , the central density @xmath41 and the inverse temperature @xmath42 . for a newtonian potential ,
one has @xmath43 , \\\label{coupled2 } & & \frac{1}{\rho_0}=\int^{1}_{0 } 4{\pi}{x^2_2 } \exp\left[\frac{2\pi\beta}{x_2}\int^{1}_{0}\rho(x_1)x_1(|x_2+x_1|-|x_2-x_1|){dx_1}\right ] dx_2 , \\\label{coupled3 } & & \frac{3}{2\beta}=\xi + 4\pi^2\int^{1}_{0}\int^{1}_{0}\rho(x_1)\rho(x_2)x_1x_2(|x_1+x_2|-|x_1-x_2|)dx_1dx_2.\end{aligned}\ ] ] note that the factor @xmath44 in ( [ coupled2 ] ) was omitted due to a typographical error by @xcite .
the solutions to these equations describe the density profile , entropy and temperature of an equilibrium system for a given energy . to obtain analogous equations for the softened potential , we replace @xmath45 everywhere with @xmath46^{1/2}$ ] .
this extension is valid in the mean - field formalism for @xmath2 small : correction terms are @xmath47 @xcite .
the mean - field limit is only meaningful physically for @xmath48 , otherwise @xmath11 diverges .
@xcite verified the mean - field results against monte - carlo simulations and an alternative analytic method known as the mayer cluster expansion , where the density of states is expanded as a combinatorial series in the dilute limit ( @xmath49 ) .
the different approaches are in accord in the dilute limit . in order to verify the mean - field approach in the high - density limit ( @xmath50 ) , relevant to the extended and collapsed phases studied here , we need to calculate the correction terms in this limit , following @xcite
a project outside the scope of this paper . nevertheless , to give a rough idea of these corrections , we note ( by analogy ) that they are of order @xmath51 in the ce , where @xmath52 in the ce is a proxy for @xmath53 in the mce .
note that the classical thermodynamic limit ( @xmath54 constant as @xmath55 ) does not apply for gravitating systems ; thermodynamic quantities are finite if proportional to @xmath56 as @xmath55 .
we solve ( [ coupled])([coupled3 ] ) for the radial density profile @xmath32 by the following iterative relaxation scheme @xcite : given the current iterate of the profile , @xmath57 , apply ( [ coupled3 ] ) , ( [ coupled2 ] ) and ( [ coupled ] ) to compute @xmath58 , @xmath59 and @xmath60 in that order , then apply @xmath61 until the convergence criterion @xmath62 is satisfied .
we typically adopt @xmath63 and @xmath64 ( @xmath65 ) in this work .
the softening can be introduced into this scheme in two ways : ( i ) as a nonzero lower limit of integration in the integrals in ( [ coupled])([coupled3 ] ) ; and ( ii ) in the potential , @xmath66^{-1/2}$ ] .
both approaches were tested and found to produce qualitatively similar behaviour ; we concentrate on the latter in this work as it is more closely allied to @xmath0-body simulations . with no softening present in the gravitational potential , a stable solution of ( [ coupled])([coupled3 ] ) formally exists above a cutoff energy @xmath67 .
the density profile of the halo exhibits a flat central core , with @xmath68 as @xmath69 , and near - isothermal wings , with @xmath32 @xmath70 @xmath71 ( @xmath722.2 ) as @xmath73 , as illustrated in figure [ nice_density](a ) .
we refer to it as the _ extended _ phase .
it agrees with the solution for secondary infall onto a spherical perturbation @xcite and behaves asymptotically like the spherical , thermally conducting polytrope @xcite and infinite - dimensional brownian gas @xcite . for @xmath74 and @xmath10 , a formal solution of ( [ coupled])([coupled3 ] ) does not exist
. the entropy and temperature are discontinuous at this cutoff energy as shown in figure [ nice_density](b ) .
this is the well - known gravothermal catastrophe @xcite .
note that , for @xmath75 , the singular isothermal sphere @xmath76 , @xmath77 and @xmath78 is always a solution of ( [ coupled])([coupled3 ] ) , as can be verified analytically , but it is not stable and so the iterative procedure never converges to it , but rather to figure [ nice_density](a ) .
if the gravitational potential is softened , a stable phase exists for all values of @xmath80 . for @xmath81 ,
the halo is extended as in figure [ nice_density](a ) , with a flat core and near - isothermal envelope , @xmath82 . however , for @xmath74 , the halo is collapsed .
figure [ collapse_density](a ) displays the density profile of such a collapsed halo for @xmath83 ( note that both axes are logarithmic ) .
the halo has a steep dirac peak ( ` cusp ' ) at @xmath84 : @xmath32 is flat for @xmath85 , decreases as a large inverse power of @xmath86 for @xmath87 and flattens for @xmath88 . for intermediate energies in the range @xmath89 ,
the system is bistable : the halo can be either extended or collapsed depending on the initial conditions and the route to equilibrium .
figure [ collapse_density](b ) , a plot of entropy and temperature as a function of energy , illustrates this bistability and the hysteresis to which it can lead .
@xmath90 and @xmath91 jump discontinuously at both @xmath92 and @xmath93 .
if @xmath80 enters the intermediate range from below , the halo remains collapsed until @xmath80 exceeds @xmath92 .
alternatively , if @xmath80 enters from above , the halo remains extended until @xmath80 is reduced below @xmath93 @xcite .
the critical energy @xmath93 , and the collapsed and extended profiles we obtain , are consistent with previous analyses @xcite .
for example , @xcite find a phase transition at @xmath94 , consistent with our value @xmath95 , and a critical reciprocal temperature at the transition in the range @xmath96 for @xmath97 , consistent with @xmath98 in this paper ( @xmath83 ) .
a more precise comparison is prohibited by the adoption of the ce rather than the mce in most previous work .
the van der waals model proposed by @xcite is an exception ; it is examined in detail in the following section .
the phenomenon of bistability was overlooked until the work of @xcite .
the behaviour of the system depends somewhat on the choice of the relaxation parameter @xmath99 , defined at the start of this section .
phase_table ] displays the minimum softening length for which the system makes the transition to a stable collapsed phase , for a given value of @xmath99 .
the phase transition from the extended to the collapsed phase is increasingly delicate as @xmath2 decreases : for larger values of @xmath99 , the system is less likely to reach the critical point where the phase transition occurs and thus remains in the extended phase . although the effect is numerical , it potentially reflects the relative likelihoods of the possible routes that the real system can take to equilibrium .
.minimum softening , @xmath100 , for which a stable collapsed solution is found for a given relaxation parameter @xmath99 .
[ phase_table ] [ cols="<,<",options="header " , ]
we have investigated the equilibrium configurations of @xmath0 self - gravitating collisionless particles , interacting via a softened gravitational potential , in the mce and mean - field limit . below a critical energy , @xmath101 , a system with
@xmath102 exists in a stable , collapsed phase .
this phase is unstable for pure gravity ( @xmath10 ) . above another critical energy @xmath103 , both softened and unsoftened systems exist in an stable , extended phase . in the intermediate region @xmath104 ,
both the collapsed and extended phases are accessible ; the detailed route to equilibrium determines which one is picked out .
the density profiles for the extended and collapsed phases are qualitatively different .
the extended profile has a flat core and near - isothermal outer envelope .
the collapsed profile is a centrally condensed dirac peak , whose logarithmic slope depends on @xmath2 .
we compare our results with published @xmath0-body simulations by using the softening parameter , @xmath2 , size , @xmath105 , and concentration parameter , @xmath106 , to place simulated haloes on the @xmath80-@xmath2 phase diagram .
we find that many published simulations inadvertently sample the collapsed phase only , even though this phase is unstable for pure gravity and arguably irrelevant astrophysically .
we remind the reader that we neglect several effects that are important in real cdm haloes , such as hierarchical clustering , nonzero angular momentum , and cosmological expansion .
our results elucidate some of the artificial behaviour that a softened potential can introduce ; they are not a substitute for a full @xmath0-body calculation .
we are grateful to bruce mckellar for extensive discussions on the theoretical basis of the mean - field equations , and the anonymous referee for useful comments that improved the manuscript .
this work was supported by the australian research council discovery project grant 0208618 .
cmt acknowledges the funding provided by an australian postgraduate award .
antonov , v. a. 1962 , vest .
leningrad univ .
, 7 , 135 aronson , e. b. & hansen , c. j. 1972 , , 177 , 145 athanassoula , e. 2004 , in iau symposium 220 , dark matter in galaxies , ed .
s. d. ryder , d. j. pisano , m. a. walker , and k. c. freeman , ( san francisco : asp ) , 255 athanassoula , e. , fady , e. , lambert , j. c. & bosma , a. 2000 , , 314 , 475 bertschinger , e. 1985 , , 58 , 39 binney , j. & tremaine , s. 1987 , galactic dynamics , ( princeton u. p. ) chabanol , m .-
corson , f. & pomeau , y. 2000 , europhys .
lett . , 50 , 148 davis , m. , efstathiou , g. , frenk , c. s. & white , s. d. m. 1985 , , 292 , 371 de blok , w. j. g. , mcgaugh , s. s. & rubin , v. c. 2001 , , 122 , 2396 de vega , h. j. & sanchez , n. 2002 , nuclear phys .
b , 625 , 409 ghigna , s. , moore , b. , governato , f. , lake , g. , quinn , t. & stadel , j. 2000 , , 544 , 616 hamilton , a. j. s. , kumar , p. , lu , e. & matthews , a. , 1991 , , 374 , 1 hayashi , e. , navarro , j.f .
, taylor , j.e . , stadel , j. & quinn , t. 2003 , , 584 , 541 huang , k. 1987 , statistical mechanics , ( 2nd ed . ; john wiley & sons inc . )
hertel , p. & thirring , w. 1971 , commun .
phys . , 24 , 22 huss , a. , jain , b. & steinmetz , m. 1999 , , 308 , 1011 huss , a. , jain , b. & steinmetz , m. 1999 , , 517 , 64 ispolatov , i. & cohen , e. g. d. 2001 , , 64 , 056103 klypin , a. , kravtsov , a. v. , valenzuela , o. & prada , f. 1999 , , 522 , 82 lynden - bell , d. & eggleton , p. p. 1980
, , 191 , 483 lynden - bell , d. & wood , r. 1968 , , 138 , 495 mo , h. j. , mao , s. & white , s. d. m. 1998 , , 295 , 319 moore , b. , governato , f. , quinn , t. , stadel , j. & lake , g. 1998 , , 499 , l5 moore . b. , ghigna , s. , governato , f. , lake , g. , quinn , t. , stadel , j. & tozzi , p. 1999a
, , 524 , l19 moore , b. , quinn , t. , governato , f. , stadel , j. & lake , g. 1999b , , 310 , 1147 navarro , j. & benz , w. 1991 , , 380 , 320 navarro , j. , frenk , c. & white , s. d. m. 1996 , , 462 , 563 navarro , j. , frenk , c. & white , s. d. m. , 1997 , , 490 , 493 navarro , j. f. , hayashi , e. , power , c. , jenkins , a. r. , frenk , c. s. , white , s. d. m. , springel , v. , stadel , j. & quinn , t. r. 2004 , , 349 , 1039 padmanabhan , t. 1990 , , 188 , 285 padmanabhan , t. 2002 , in lecture notes in physics 602 , dynamics and thermodynamics of systems with long range interactions , ed .
t. dauxois , s. ruffo , e. arimondo , and m. wilkens , ( berlin : springer verlag ) , 165 padmanabhan , t. & engineer , s. 1998 , , 493 , 509 peebles , p. j. e. 1982 , , 263 , 1 reed , d. , governato , f. , verde , l. , gardner , j. , quinn , t. , stadel , j. , merritt , d. & lake , g. 2003 , preprint ( astro - ph/0312544 ) sand , d. j. , treu , t. , smith , g. p. & ellis , r. s. , 2004 , , 604 , 88 shen , j. , sellwood , j.a .
2004 , , 604 , 614 sire , c. & chavanis , p .- h .
2002 , , 66 , 046133 spergel , d. n. & steinhardt , p. j. 2000 , , 84 , 3760 splinter , r. j. , mellott , a. l. , shandarin , s. f. , & suto , y. 1998 , , 497 , 38 swaters , r. a. , madore , b. f. , van den bosch , f. c. & balcells , m. 2003 , , 583 , 732 thacker , r. j. , couchman , h. m. p. 2001
, , 555 , l17 van albada , t. s. , , 201 , 939 van kampen , e. 2000 , preprint ( astro - ph/0002027 ) | the statistical mechanics of @xmath0 cold dark matter ( cdm ) particles interacting via a softened gravitational potential is reviewed in the microcanonical ensemble and mean - field limit .
a phase diagram for the system is computed as a function of the total energy @xmath1 and gravitational softening length @xmath2 . for softened systems ,
two stable phases exist : a collapsed phase , whose radial density profile @xmath3 is a central dirac cusp , and an extended phase , for which @xmath3 has a central core and @xmath3 @xmath4 @xmath5 at large @xmath6 . it is shown that many @xmath0-body simulations of cdm haloes in the literature inadvertently sample the collapsed phase only , even though this phase is unstable when there is zero softening . consequently , there is no immediate reason to expect agreement between simulated and observed profiles unless the gravitational potential is appreciably softened in nature . |
the positron - production target for the ilc positron source is driven by a photon beam generated in an helical undulator placed at the end of main electron linac @xcite .
the undulator length is chosen to provide the required positron yield .
the source is designed to deliver 50% overhead of positrons .
therefore , the positron yield has to be 1.5 positrons per electron passing the undulator .
the required active length of the undulator is about 75 meters for the nominal electron energy of 250 gev , the undulator @xmath0-value has been chosen to be 0.92 , the undulator period is 11.5 mm and a quarter - wave transformer is used as optical matching device ( omd ) .
the photon first harmonic energy cutoff is 28 mev , the average energy of photons is about 29 mev and the average photon beam power is about 180 kw in a train of 2625 bunches with a frequency of 5 hz . although only relatively small fraction of total photon beam energy deposited in the target ( about 5% ) , the peak energy density deposited in target is high due to the small opening angle of the synchrotron radiation in the helical undulator resulting in a small photon spot size on the target .
for example , for 500 meters space between the undulator and target , the average radius of the photon beam is approximately 2 mm and the peak energy density could achieve 120 j / g in the 0.4 radiation length thick titanium - alloy target rotated with 100 m / s tangential velocity .
there is no experimental data indicating the upper limit of the peak energy density deposited by photons in the titanium alloy material with 90% of titanium , 6% of aluminium and 4% of vanadium .
the analysis of the electron beam induced damage to the slc positron target @xcite and the simulations of thermal shock @xcite show that the energy deposition limit is about 30 j / g for tungsten with 25% of rhenium target irradiated by 33 gev electrons and a general criteria of failure due to an equivalent ( von - mises ) stress of 50% of tensile strength may apply to this target material @xcite .
the thermal structural modeling of a rotated titanium target irradiated by helical undulator photons has been performed for the nlc by w. stein and j. sheppard @xcite .
they recommend to consider as `` safe '' thermal stresses below one third to one half of the yield stress . in this paper ,
the thermal stress in the ilc positron source target has been estimated for the sb2009 set of parameters @xcite .
the energy transfer from the photon beam into temperature of target material and the structural deformation and mechanical stress coupled with this temperature rise due to complexity of these time - dependent , cross - coupled and nonlinear processes can not be treated with the highest level of details @xcite .
therefore , the choice of simulation tools and reasonable approximations and simplifications plays an important role .
the energy deposition in the target has been calculated in fluka @xcite .
an amount of energy is counted as deposited if after collisions the primary or secondary particles have energies lower than the energy cut - offs .
the fluka default cut - offs were used : 1.511 mev for electrons and positrons and 333 kev for photons .
figure [ au : fig - edep - fluka ] shows the `` original '' fluka data distribution ( i.e. without any scaling factors ) of the energy deposited close to the back side of the target .
the energy is given in units of gev per cubic centimeter and per impinging on the target photon . and
@xmath1 ) of the energy [ gev/(ph@xmath2 ) ] deposited close to the backside of ti6al4v target calculated in fluka.,width=230 ] the temperature rise @xmath3 in the target for given a energy deposition @xmath4 has been calculated according to the following equation @xmath5 where @xmath6 is the number of electrons per bunch ( @xmath7 ) , @xmath8 is the photon yield ( 1.94 photons per electron and per 1 meter of undulator ) , @xmath9 is the length of undulator ( 70 meters ) , @xmath10 is the number of bunches crossing the same volume / bin , @xmath11 is the target density ( 4.49 g/@xmath2 ) and @xmath12 is the specific heat capacity ( 0.523 j/(g k ) ) . the temperature data in a 1.48 cm thick cylindrical titanium target after the first 100 bunches has been imported into ansys @xcite .
the temperature distribution on the back side of the target is shown in figure [ au : fig - temperature - backside ] .
the maximal increase of temperature per bunch is about 2.2 k. as a first step , a statical ansys model of the target material response to the heat load ( see , figures [ au : fig - edep - fluka ] and [ au : fig - temperature - backside ] ) has been applied .
the total deformation and equivalent von - mises stress are shown in figs . [
au : fig - deform ] and [ au : fig - stress ] .
the maximum of equivalent stress is about 100 mpa on the back side of the target in the circular area around the photon beam axis with a radius of approx
this stress is about 12% of the tensile yield strength for titanium alloy ( the properties of ti6al4v alloy , grade 5 can be found , for example , in ref .
to simulate the time evolution of thermal stress in the positron source target , the target movement has been analyzed more accurately and ansys transient ( explicit ) model of deformation and stress has been used .
we consider the tangential velocity of the target rim ( 1 meter in diameter ) of 100 meters per second as velocity in @xmath1 direction in a cartesian system .
the energy deposited after one pulse ( 1312 bunches with 554 ns bunch separation ) as function of @xmath1 coordinate is shown in fig .
[ au : fig - edep - rot ] .
this figure shows also the energy deposited by a single bunch and the corresponding temperature rise . both profiles on fig .
[ au : fig - edep - rot ] are plotted for highest energy deposition : in the @xmath13-direction close to the target back side and in the middle of bunch(es ) in @xmath14-direction .
the bunch overlapping factor is defined as the ratio of the maximal deposited energy after a complete bunch train with respect to the maximum after just one bunch .
this factor for the nominal sb2009 source parameters is about 59 .
@xmath15 figure [ au : fig - tempstress - rot ] shows the temperature and equivalent stress in the `` rotated '' target .
the target has been cut in the middle plane in order to show the distributions inside the target . the static ansys model for the equivalent stress after one bunch train does not take into account thermal diffusion and
thus overestimates the stress induced in the target .
@xmath16 to reduce the effect of thermal diffusion on the stress and to study the time - dependent dynamic effects , another model has been used . in this model the temperature distribution after one single bunch has been scaled with the above - mentioned bunch overlapping factor .
the cylindrical geometry of the target has been chosen to keep the symmetry of the model and to reduce the computing time .
figure [ au : fig - temperature59 ] shows the temperature distribution after `` 59 bunches '' .
the total deformation after 59 bunches is plotted in fig .
[ au : fig - deformation59 ] and the evolution in time of maximal deformation is shown in fig .
[ au : fig - deformation59-vs - time ] . the starting time ( 0 sec . ) corresponds to the end of the pulse .
the reflections from the target surfaces and interference of the waves result in the series of maxima at the level about 25% of the initial deformation .
the deformation transverse to the beam axis ( radial deformation ) contributes only minor ( about one third ) to the total deformation .
the time dependence of the dominating longitudinal ( @xmath13-component ) velocity is presented in fig .
[ au : fig - velocityz59-vs - time ] showing the positive velocity directed out of the target and negative velocity .
figure [ au : fig - velocityz59 ] shows a snapshot of the time evolution for the @xmath17-distribution after one pulse and with 0.1 @xmath18s delay at the moment when the negative velocity has reached the maximum .
the @xmath1-component of deformation and velocity are also shown in figs .
[ au : fig - deformationy59 ] and [ au : fig - velocityy59-vs - time ] .
because of geometry and beam symmetry , the deformation , maximal and minimal @xmath19-dependencies on time are symmetrical ( mirrored ) too .
it has to be noted that the transient effects during the pulse were not considered and thus the all velocities are starting from zero level at the end of the beam pulse .
-component of velocity after 59 bunches.,height=192 ] -axis after 59 bunches and 0.1 @xmath18s delay.,height=220 ] -axis ) deformation after 59 bunches.,height=192 ] -component of velocity after 59 bunches.,height=192 ] the time evolution of the maximal equivalent ( von - mises ) stress in the target is plotted in fig .
[ au : fig - stress59-vs - time ] .
the stress distributions after one beam pulse and additional 0.1 @xmath18s delay are shown in fig .
[ au : fig - stress59 ] .
the peak stress value is about 160 mpa which corresponds to 18% of tensile yield strength .
this level of stress can be considered as acceptable . s delay ( right).,title="fig:",height=160 ] s delay ( right).,title="fig:",height=160 ]
the energy deposition in the ilc positron source target has been simulated in fluka for the sb2009 set of parameters .
the peak energy density in the rotated titanium - alloy target is about 120 j / g for the conservative choice of a magnetic focusing device ( quarter - wave transformer ) and 250 gev electron beam energy .
the different simplified ( static and transient ) ansys models have been used to estimate the thermal stress induced by fast temperature rise and thermal expansion of the target .
the peak stress is about 160 mpa .
it is less then 20% of tensile yield strength .
such stress will not damage the target .
in the future , also the cooling of the target has to be added in model .
the procedure used so far , in which the deposited energy is converted into temperature , has to be eliminated and the direct import of the heat source into ansys can additionally improve the accuracy of stress estimations .
the thermal and structural effects in the target have to be also simulated taking into account the time structure of the bunch train .
we would like to thank the organizers and the host of posipol 2011 for this fruitful and encouraging workshop and for hospitality .
v. k. bharadwaj , y. k. batygin , j. c. sheppard , d. c. schultz , s. bodenstein , j. gallegos , r. gonzles , j. ledbetter , m. lpez , r. romero , t. romero , r. rutherford and s. maloy , _ proc . of pac 2001 _ , 2123 ( 2001 ) . | to achieve the extremely high luminosity for colliding electron - positron beams at the future international linear collider @xcite ( ilc ) an undulator - based source with about 230 meters helical undulator and a thin titanium - alloy target rim rotated with tangential velocity of about 100 meters per second are foreseen .
the very high density of heat deposited in the target has to be analyzed carefully .
the energy deposited by the photon beam in the target has been calculated in fluka . the resulting stress in the target material after one bunch train has been simulated in ansys .
desy-12 - 018 |
cosmic strings are one - dimensional topological defects formed after phase transitions .
they are considered to make up a weblike structure in the universe , so - called _ the cosmic - string network_. cosmic strings could be a probe for the early phases of the universe long before the cosmic microwave background ( cmb ) epoch .
they have a potential to reveal the physics during the phase transition of fields in the early universe , and also be a potential source of gravitational waves @xcite and an extra source of cmb anisotropy @xcite .
the simplest classical field - theoretic model to describe the string formation is the abelian - higgs ( ah ) model , where there are a complex scalar field with the self - coupling constant @xmath3 and a @xmath4 gauge field with the gauge coupling constant @xmath5 ( see e.g. the textbook @xcite ) .
the basic properties of cosmic strings in the ah model can be classified by a single parameter , @xmath1 , where @xmath6 and @xmath7 are the masses of the scalar field and the gauge field , respectively , acquired after the spontaneous breaking of @xmath4 .
the case with @xmath8 is called `` critical coupling '' or `` bogomolnyi coupling '' , and the cases with @xmath2 and @xmath9 are called type - i and type - ii ( cosmic ) strings , respectively .
these names stemmed from the classification of superconductors and are not to be confused with those of superstring theories .
historically , numerical simulations on the dynamical formation of the string network in the expanding universe have been performed in the formulation based on the nambu - goto action ( e.g. see @xcite ) . in the nambu - goto simulations ,
strings are treated as zero - width strings , and the detailed interactions between strings playing an important role at the reconnection can not be incorporated .
hence the interactions of two strings are usually introduced by hand so that the strings reconnect stochastically . on the bounty of the rapid development of computers , it has been becoming possible to directly simulate the formation , evolution and extinction of strings in the basis of the field - theoretic models on the lattice .
a pioneer work of the field - theoretic simulations was done in ref .
@xcite . after this work
, some groups have tried to perform simulation of the ah strings .
most studies assumed @xmath8 . with this assumption ,
consistency with one of the semianalytic models , velocity - dependent one - scale model , was studied in @xcite , and the impacts of strings on the cmb were studied in @xcite .
focusing on string interactions , ref .
@xcite clarified that there are no interactions between parallel straight strings for @xmath8 in the minkowski spacetime , while there is the repulsive force between type - ii strings and the attractive force between type - i strings . due to the attractive feature of the parallel type - i strings
, they can form a bound state which could affect the characteristics of the resultant network . in refs .
@xcite , the authors found the nonintercommutation process accompanying a temporal bound state in the collisions of strings with low velocity and small collision angle .
a similar feature can be seen in the cosmic superstring network @xcite where there are two kinds of strings , f - strings and d - strings , and they can form a bound state called fd - strings @xcite . to investigate such a strong interaction between strings , urrestilla and vilenkin have performed simulations of scalar fields with a @xmath10 gauge field and they observed a small fraction of bound states formed in the string network @xcite . as another interesting feature of type - i strings , it was reported that extremely high velocity collisions also result in nonintercommutation @xcite .
these interesting characteristics could affect the type - i string network .
as for the type - ii strings , there are a large number of network simulations @xcite , which have been mainly used for studies on the cosmologically generated axions .
note that the targets of these simulations are global strings corresponding to the extreme type - ii case , @xmath11 . in most field theories including the ah model , coupling constants
are expected to be of the same order .
therefore many of the previous works on cosmic strings have dealt with the critical coupling or weakly type - i / ii strings . however , as a special case , it is also reported that a kind of minimally supersymmetric standard model prefers @xmath12 ( extreme type - i strings ) @xcite .
hence we stress that it is still an open question what the preferred value of @xmath0 in the universe is , and field - theoretic simulations of the type - i string network including their interesting characteristics are needed . in this paper , we perform simulations of the abelian - higgs model with various choices of @xmath2 in the radiation - dominated universe . in an expanding background , the strings feel additional effective repulsive force , dragging effect by the cosmic expansion , in between them .
hence , it is expected that the properties of the resultant string network depend on not only the strength of the intrinsic attractive force between strings , but also the hubble parameter at the string formation epoch . to investigate the characteristics of the network , we solve the field equations of the scalar field and the gauge field in a numerical way .
this paper is organized by the following sections . in sec .
[ sec : basic ] , we give the field equations to be solved and set up the model used throughout this paper . in sec .
[ sec : setup ] , we define some numerical parameters for the following numerical simulations . in sec . [
sec : identification ] , we discuss how to find the string cores and define the estimator of the correlation length . in sec . [
sec : result ] , we show numerical results of the network simulations of type - i strings varying the parameter @xmath0 .
then , in sec . [ sec : conclusion ]
, we conclude this paper .
in addition , we check the numerical convergence of our results in appendix [ appsec : check ] , and related with this , the @xmath0 dependence of the width of strings for their static configurations is shown in appendix [ appsec : vortex ] .
finally , in appendix [ appsec : smoothing ] , we explain our procedure to estimate the effective number of strings in the horizon - sized box .
we use the units such that @xmath13 .
in the spatially homogeneous and isotropic space - time , the metric @xmath14 is parameterized by the scale factor @xmath15 as @xmath16 the action of the ah model is @xmath17 where the symbol @xmath18 denotes the complex conjugate and @xmath19 here we introduced the complex scalar field @xmath20 , and the @xmath4 gauge field , @xmath21 . varying this action
, we obtain the field equations for @xmath20 and @xmath21 in arbitrary gauge as @xmath22 throughout this paper , we take a gauge condition @xmath23 .
then the evolution equations to be solved in the friedmann universe become @xmath24 where the prime ( @xmath25 ) denotes the derivative with respect to the conformal time @xmath26 , @xmath27 is the comoving hubble parameter defined as @xmath28 , and @xmath29 the constraint equation given by the @xmath30 component of eq .
( [ eq : eqfulla ] ) becomes @xmath31 in our simulations , we impose the periodic boundary condition on the boundaries of the computational domain .
therefore the volume integral of eq .
( [ eq : const ] ) is trivially satisfied , and hence we do not consider this equation hereafter .
we consider the following temperature - dependent potential : @xmath32 the transition temperature @xmath33 is found as the temperature at which the effective mass of @xmath20 deriving from the second derivative of the potential vanishes . after the phase transition ,
the scalar field starts to oscillate around the true vacuum given by @xmath34 the masses of the scalar field @xmath20 and gauge field @xmath35 after the phase transition are given as @xmath36 and @xmath37 , respectively , in the zero temperature limit .
the ratio of these masses , @xmath38 plays an important role for the characteristics of strings and the string network at the late time in this paper is different from that in the literature , where the mass of the scalar field is calculated as @xmath39 and hence the parameter @xmath0 is written by @xmath40 . ] . in this paper , to investigate the string network constituted by the type - i strings , we set @xmath41 .
throughout this paper , we assume the radiation - dominant universe and no backreaction from the scalar and gauge fields onto the background geometry
. then the friedmann equation is simply given by @xmath42 where @xmath43 is the effective massless degrees of freedom .
the vacuum expectation value of the scalar field at the zero temperature @xmath44 determines the normalization of the typical energy scales of @xmath45 , @xmath46 and also the time scale and the spatial scale . for convenience ,
we introduce another parameter @xmath47 defined as @xmath48 .
we fix @xmath49 throughout this paper except in the last part of sec .
[ subsec : correlation ] .
note that the constant @xmath50 does not appear anywhere except in eq .
( [ eq : friedmann2 ] ) .
hence it is not needed to set a specific value to @xmath50 or @xmath43 . in what follows the symbols with the subscript @xmath51 ( @xmath52 )
refer to the quantities at the initial ( final ) time of each simulation .
normalizing the initial value as @xmath53 , the scale factor can be written as @xmath54 .
the hubble parameter is then recast as @xmath55 .
the comoving hubble is denoted as @xmath27 and is given by @xmath56 .
the initial , final and transition times are , respectively , given by @xmath57 our numerical simulations are performed before the phase transition and end sufficiently after it , namely , @xmath58 and @xmath59 .
the computational domain is a comoving box with the side length @xmath60 , and we define a new quantity @xmath61 to measure the relative size of the box at a given time compared to the horizon scale .
we use @xmath62 and @xmath63 as numerical parameters instead of the box size @xmath60 and the final time @xmath64 . in particular , we choose @xmath65 to avoid the contamination from the finiteness of the computational domain throughout the simulations . in the end , we are left with the physical parameters @xmath66 , two of them being independent . as for the initial conditions
, we set the values of @xmath20 on the assumption that the scalar field stays in the thermal bath with the temperature @xmath46 @xcite .
reference @xcite provides the equal - time correlation function of @xmath67 . subtracting the contribution from the infinite vacuum energy , we obtain the equal - time correlation function of @xmath67 , @xmath68 where @xmath69 , and @xmath70 represents the ensemble average at the finite temperature without the contribution from the vacuum energy .
the integrand in the right - hand side gives the power spectra of @xmath71 .
if we take a limit , @xmath72 and @xmath73 , we obtain the variances , @xmath74 . in our simulations , we use eq .
( [ eq : initpower ] ) with @xmath75 .
firstly we give the initial condition for @xmath76 at each grid point , @xmath77 , in the fourier space by generating the gaussian random numbers for @xmath78 according to the above power spectrum , and the homogeneous random numbers between @xmath79 and @xmath80 for the phase of @xmath76 .
then , using the inverse fourier transformation , we obtain the initial condition for @xmath81 in the real space . as for the time derivative , @xmath82
, we simply set @xmath83 everywhere .
next , @xmath84 is determined from the constraint equation with the fourier transformation given in eq .
( [ eq : const ] ) , while @xmath35 is set to be zero .
we expect that the details of the initial condition do not crucially affect the final behavior of strings after the phase transition .
we use the staggered grid where @xmath67 lies at the grid points ; @xmath85 connects at each link two grid points , @xmath86 and @xmath87 , where @xmath88 is the grid spacing ; and @xmath89 represents the unit vector parallel to the axis of @xmath90 .
the fiducial values of the numerical parameters are listed in table [ tab : param ] .
we use these values in most of our simulations , except when we check the reliability of our numerical results , dependence on the resolution and the box size . in solving eqs .
( [ eq : eqphi])([eq : eqa ] ) for @xmath20 and @xmath35 , we use the second - order finite difference scheme for spatial derivatives and the second - order leapfrog scheme for time evolution .
to implement this time - evolution scheme , we define @xmath91 and eliminate the first - derivative term , @xmath92 or @xmath93 . then we solve the equation of @xmath94 , instead of @xmath20 .
in addition , in order to compare with the previous studies @xcite , we implement the press - ryden - spergel ( prs ) algorithm @xcite in the last part of sec .
[ sec : result ] . in the prs prescription
, the coupling constants , @xmath3 and @xmath5 , become time dependent , @xmath95 where @xmath96 and @xmath97 are the initial values , and note that @xmath0 keeps its constancy during the simulations .
this algorithm is effective for the lattice simulations in the expanding universe since the expanding lattice spacing can forever , in principle , follow the width of the strings .
.the fiducial values of numerical parameters .
the parameters enclosed in parentheses are determined from other ones .
[ cols="^,^,^ " , ]
in order to measure the total length of the strings in the network , we first have to identify the location of the cores of strings .
however there is great ambiguity in how we identify them . in this paper , we implement the method developed in ref .
@xcite . here
we briefly explain this algorithm .
let us consider a cube constituted by eight neighbor grid points on which the scalar field stays .
focusing on one of six surfaces of the cube , if the phase of the scalar field on the surface becomes monotonically larger or smaller along its four sides and eventually the sum of the differences of phases between two neighbor grid points on the surface comes to @xmath80 , we can judge that a string passes the surface in principle . instead of this direct method
, we consider a complex plane representing the complex scalar field , and divide it into three domains . according to ref .
@xcite , there is an advantage to dividing the plane inhomogeneously as @xmath98 , @xmath99 , and @xmath100 . on this plane
, we plot four values of @xmath101 on a surface of the unit grid cube .
we judge that a string passes on the surface if each domain possesses at least one @xmath101 .
we repeat this process for six surfaces of the cube , and determine the surfaces across which the string passes .
then we connect the central points of the surfaces by straight line segments .
( therefore , the length of a line segment is @xmath88 or @xmath102 where @xmath88 is the grid spacing . ) after repeating this procedure for all cubes , we can identify the location of strings in the simulation box .
finally , the total length of the strings can be estimated by summing the lengths of the line segments . in ref .
@xcite , the authors used an estimator of the comoving correlation length of the string network by measuring the total length of the strings in the simulation box , @xmath103 where @xmath104 is the comoving volume of the simulation box , and @xmath105 is the total comoving length of strings in the box , estimated with the method discussed in the previous section .
the quantity @xmath106 represents the averaged comoving separation of two neighboring strings , and thus is identical to the correlation length from the viewpoint of the so - called one - scale model .
if the scaling regime is reached , @xmath106 should grow proportionally to the conformal time , @xmath107 . in this paper
, we estimate the correlation length from simulations using this estimator .
if the scaling regime starts , the string energy should behave as @xmath108 where @xmath109 is the physical correlation length . in order to directly see whether the network really gets to the scaling regime , we define and calculate the string energy and check whether it evolves according to this scaling .
given an action @xmath110 , the energy density can be derived as @xmath111 where @xmath112 denotes the comoving observer s four - velocity . using eq .
( [ eq : action ] ) , this can be expressed as @xmath113 + \frac{1}{2a^4}\left ( |a'_i|^2 + f_{12}^2+f_{23}^2+f_{31}^2 \right ) .
\label{eq : energy}\ ] ] next we consider the width of a string to estimate the energies possessed by each string . in this paper
, we do not directly calculate the width , but recognize the region for @xmath114 with a given constant @xmath115 , e.g. @xmath116 , as a portion of strings , namely , @xmath117 where @xmath118 is the heaviside step function .
in fig . [ fig:3d ] , we plotted three different time slices of the scalar field with @xmath119 and @xmath120 as an example , which shows the isosurface with @xmath121 . from left ,
they correspond to @xmath122 , and @xmath123 , respectively .
note that the phase transition takes place at @xmath124 in this simulation .
after the phase transition , the scalar field starts to oscillate around the true vacuum in the so - called mexican hat potential . just after the transition , the oscillation is still so strong that it is not obvious whether the strings are actually forming , as shown in the left panel of fig .
[ fig:3d ] . after a while , due to the hubble friction
, the scalar field in most parts of the computational box is settling down to the true vacuum , and satisfies @xmath125 .
then the strings begin to appear , as shown in the center panel of fig .
[ fig:3d ] . at this phase , we observed the pulsating behavior of strings : that is , the width of strings oscillates in time .
this feature can also be seen in the time evolution of string energy shown in the next subsection . at the later time
, strings clearly appear in the box , as shown in the right panel of fig .
[ fig:3d ] . even at this time
, the string pulsation can be still observed , and for some strings , we found a phenomenon such that a wave packet is moving on a string , for instance , on a small loop with a clump shown in the left center of the box .
next we identify all strings at each time step , and calculate their energies to check whether the string network is in the scaling regime .
to calculate their energies , we do not use the string core identification technique discussed in sec .
[ subsec : string_cores ] since it is difficult to set the string width which changes in time , and it does not matter where the string core is . instead , we fix @xmath115 , and we identify the region satisfying @xmath114 as strings
. then the string energy is given by eq .
( [ eq : rhostr ] ) . in fig .
[ fig : energy ] , we show the time evolution of the string energy for @xmath120 and @xmath119 with @xmath126 and @xmath127 . for @xmath128 , we find that the relation @xmath129 is approximately satisfied at the late time , if we average the oscillatory behavior , and that the transition time seems insensitive to the choice of @xmath115 .
this observation indicates that the network would lie in the scaling regime . in fig .
[ fig : corr ] , we show the time evolution of the correlation length estimated by eq .
( [ eq : def_xi ] ) for various values of @xmath0 including weakly type - ii regime with @xmath119 .
each of them is the average of 10 realizations , and the error bars indicate one @xmath130 .
we found that the correlation length is strongly dependent upon @xmath0 , particularly for @xmath131 .
our findings are as follows .
first , the time of the start of the string formation is common to all the cases and is about @xmath132 .
moreover , there is a general tendency that the correlation length @xmath106 becomes smaller . for @xmath133 ,
@xmath134 becomes smaller , which indicates the initial density of the string network would be larger . for @xmath135 , @xmath134 does not change so much , but the increasing rate of @xmath106 , or @xmath136 , becomes smaller as @xmath0 approaches @xmath137
finally , as @xmath0 goes below @xmath137 , @xmath138 rises again while the low increasing rate is maintained .
the simulation results imply that the network with the smaller value of @xmath0 tends to take a smaller @xmath106 , or be denser in the comoving box .
to see this fact from another aspect , we calculate the effective number of strings in a virtual box whose volume is @xmath139 , which hereafter we refer to as `` horizon - sized box . ''
focusing on a string , it occupies an area @xmath140 on the surface perpendicular to the string by definition . hence , considering a cross - sectional surface of the simulation box , one can imagine that @xmath141 strings pass across the surface . according to this naive expectation , we estimate the effective number of strings in the horizon - sized box by @xmath142 if @xmath106 behaves approximately as @xmath143 where @xmath144 and @xmath145 are constants and we neglect the oscillatory behavior , @xmath146 at the late time becomes @xmath147 where we used @xmath148 in the radiation - dominant universe .
this equation indicates that the network density in the horizon - sized box is determined by the increasing rate of @xmath106 , and also the correspondence to the fact that the number of strings in the horizon - sized box is constant in the scaling regime .
we estimate the ensemble average , @xmath149 , using the data obtained by all realizations , and then we obtain the late - time number of strings in the simulation box , @xmath150 . the details of the procedure to obtain @xmath150 are explained in appendix [ appsec : smoothing ] . briefly speaking ,
we first remove the oscillatory behavior of @xmath106 in each realization , and then using the least - square method , we fit the data in the range of @xmath151 to the ansatz @xmath152 where @xmath153 and @xmath154 is dependent on the value of @xmath0 . how to choose @xmath154 is discussed in appendix [ appsec : check ] .
figure [ fig : nlate ] shows @xmath150 against @xmath0 .
the error bars are due to the variation of @xmath144 in each realization .
we found that @xmath150 becomes obviously larger for @xmath155 , while it is almost constant and only a few strings exist for @xmath156 .
although @xmath157 realized the maximum number of strings , it might be premature to conclude that the network with @xmath157 becomes densest in the horizon - sized box .
as discussed in appendix [ appsec : check ] , the reliable range of simulation data for @xmath158 is relatively short . in order to extend the reliable range ,
many more spatial resolutions are required to resolve the steep change of the gauge field around the string core , indicated by @xmath159 for @xmath160 , as discussed in appendix [ appsec : vortex ] . back to the network density in the comoving box
, we performed additional simulations for @xmath161 and @xmath162 with fixed @xmath50 where @xmath47 is defined in sec .
[ sec : setup ] , and check whether the critical value of @xmath0 realizing the densest network in the comoving box depends on @xmath47 . for the large value of @xmath47 , the phase transition takes place at the earlier universe since the transition temperature is given by @xmath163 . in fig .
[ fig : diffh ] , the phase transition at higher energies with @xmath161 provides a smaller critical @xmath0 around @xmath164 , whereas the critical @xmath0 is not so clear for @xmath165 , but would be around @xmath166 .
these facts imply that the string network density in the comoving box is strongly related to the dragging effect by the cosmic expansion .
finally , we investigate the effect of the prs algorithm , given in eq .
( [ eq : prs ] ) , on the correlation lengths of the simulated type - i string network .
we choose @xmath120 and @xmath137 , and also @xmath167 as a reference .
their correlation lengths are shown in fig .
[ fig : prs ] . in this figure
, we also plotted the corresponding results shown in fig .
[ fig : corr ] . as we mentioned before
, there is a strong dependence of @xmath106 on the value of @xmath0 .
in contrast , if we use the prs algorithm , the @xmath0 dependence completely disappears , and thus it seems that we can obtain the universal value of the gradient of @xmath106 and @xmath150 .
focusing on the case with the critical coupling ( red line ) , the gradient of @xmath106 is almost insensitive to whether we use the prs or not .
this feature has been expected since there would be no interactions , or sufficiently weak , between strings in this case , except at the impact points of the reconnection process .
therefore , even if we vary the value of the coupling constants @xmath3 and @xmath5 in time , the characteristics of the whole network are not so affected .
in contrast , this is not the case with the type - i strings , which have the intrinsic string - string interactions . to use the prs algorithm for them
corresponds to weakening the interactions on purpose as time proceeds .
hence , the network would lack its characteristics depending on the value of @xmath0 . consequently
, we would like to claim the need to carefully apply the prs algorithm to the cases , except those with the critical coupling .
note that , for those with the prs , the scaling regime starts later than those without the prs .
this is because @xmath3 is decreased at and after the phase transition .
smaller @xmath3 means that the potential becomes more flat and thus it takes more time for the field to sufficiently relax and then to satisfy @xmath125 .
we numerically studied the formation and time evolution of type - i cosmic - string networks in the abelian - higgs model by three - dimensional lattice simulations in a box with @xmath168 grid .
figure [ fig : energy ] was useful to check that the network actually enters the scaling regime .
then we particularly focused on the dependence of the correlation length @xmath106 on the parameter @xmath0 . in the type - i regime ( @xmath2 )
, the gauge field plays an important role for the interaction between strings . as seen in fig .
[ fig : corr ] , we found that the time dependence of the correlation length is strongly dependent on the value of @xmath0 .
more concretely , we found that there seems to be a critical value of @xmath0 with which the string network becomes densest in the expanding universe , defined here does not mean the critical coupling , namely , @xmath8 . ] and that this critical value becomes smaller , if the energy scale of the phase transition becomes higher ; see fig .
[ fig : diffh ] .
furthermore , we found that the effective number of strings in a box with the volume @xmath139 , as defined in eq . , is almost constant for @xmath169 , and the number tends to suddenly increase for @xmath170
; see fig .
[ fig : nlate ] .
the figure also indicates that the number of strings has a peak at @xmath157 .
however , it would be premature to conclude that @xmath157 actually realizes the densest network since the reliable range of simulation data is not sufficiently wide due to the shortage of the spatial resolution of our simulations for @xmath158 .
the critical value of @xmath0 seems to depend on the energy scale of the phase transition .
we found that the phase transition at higher energies provides a smaller critical @xmath0 , whereas the value becomes larger if the phase transition takes place at lower energies .
this fact implies that the critical @xmath0 realizing the densest network is determined not only from the strength of the gauge interaction , but also from the environmental effect , namely , the cosmic expansion . in order to clarify the origin of the critical @xmath0
, it would be needed to deeply investigate the string - string interaction in the friedmann background .
so far , field - theoretic simulations of string network formation have been performed with the press - ryden - spergel ( prs ) algorithm where @xmath3 and @xmath5 are varied in time to maintain the constancy of the comoving width of a string .
this algorithm is effective for the lattice simulations in the expanding universe since the expanding lattice spacing can forever , in principle , follow the width of the strings . in order to investigate the validity of this algorithm for the type - i strings
, we have also performed the simulations with the prs algorithm . as a result , the interesting properties mentioned above completely disappeared , and hence we can not find any differences among the results with different values of @xmath0 .
this result indicates that the prs algorithm should not be applied to type - i strings , if one focuses on the epoch soon after the phase transition where the string - string interaction would be strong .
however , there is a subtlety in the connection between the two results with and without the prs algorithm in fig .
[ fig : prs ] . naively thinking , we can speculate that , at the sufficiently late time , the mean separation of the strings would become large enough for them to terminate the interactions with each other .
this fact would mean that the correlation length evolves along with the results with the prs algorithm shown in fig .
[ fig : prs ] at the late time , since the change of string width must be negligible at the sufficiently late time .
in other words , it is expected that the gradient of @xmath106 without the prs would become larger at some time when the string - string interaction can be neglected , and then the gradient becomes similar to that with the prs .
unfortunately , with our present computer resources , we could not follow the simulations up to such a transition point , and thus this is still an open question .
we thank professor a. vilenkin for giving useful comments on this work . t.
h. is supported by jsps grant - in - aid for young scientists ( b ) no . 23740186 , and also by mext hpci strategic program .
y. s. is supported in part by mext through grant - in - aid for scientific research on innovative areas no .
. k. t. is supported by jsps grant - in - aid for young scientists ( b ) no .
23740179 , by mext through grant - in - aid for scientific research on innovative areas no . 24111710 , and partially by jsps grant - in - aid for scientific research ( b ) no .
d. y. was supported by grant - in - aid for jsps fellows
we check the robustness of our numerical results to the resolution .
figure [ fig : resolution ] shows that the correlation length with @xmath120 , 0.4 and 0.8 when we vary the number of grids with a fixed box size , @xmath60 . in the comoving
coordinate , a string seems to become thinner in time , and thus the simulation is broken down when the grid can no longer resolve the string .
this fact reflects that the end time of each simulation becomes later as @xmath171 is increased .
moreover , just before the breakdown , @xmath106 tends to be flat , while it grew almost linearly .
therefore , the reliable results would be obtained only in the region where two results with different resolutions overlap . due to this shortage of resolutions at the late time , we use only the relatively reliable part of simulation data in the finite time range , @xmath172 , when we estimate the gradient of @xmath106 in sec .
[ subsec : correlation ] or appendix [ appsec : smoothing ] . for all cases , we fix @xmath173 corresponding to the starting time of the scaling regime , and basically @xmath174 which is the end time of simulations . from the results in fig . [
fig : resolution ] , we choose @xmath175 for @xmath157 , and @xmath176 for @xmath120 .
consider the axial - symmetric string configuration in the minkowski spacetime , so - called abrikosov - nielsen - olesen vortex @xcite . according to ref .
@xcite , using the cylindrical coordinates @xmath177 that originated from the center of the string , the scalar field and the gauge field can be represented by the following forms : @xmath178 where @xmath179 is the winding number of the string . substituting them into eqs .
( [ eq : eqfullphi ] ) and ( [ eq : eqfulla ] ) with @xmath180 , and neglecting the time dependence , the governing equations for @xmath181 and @xmath182 are given by @xmath183 where @xmath184 and we neglect the temperature - dependent terms in the potential @xmath185 .
the boundary conditions are given as @xmath186 for @xmath187 and @xmath188 for @xmath189 .
with these conditions , we solved the above equations numerically .
then we calculated the half - value widths of @xmath181 and @xmath182 , the value of @xmath190 satisfying @xmath191 or @xmath192 , for the various @xmath0 as the estimator of the string width .
figure [ fig : vortexwidth ] shows the @xmath0 dependence of the half - value widths for @xmath193 strings with @xmath119 .
clearly the string cores consisting of the scalar field and the gauge field get thin for type - i strings ( @xmath2 ) . in other words , the large @xmath5 with a fixed value of @xmath3 or the small @xmath3 with a fixed value of @xmath5 produces thin strings .
in particular , the half - value width of the gauge field is more strongly dependent on @xmath0 than that of the scalar field .
this property requires the finer resolution of the computational domain , particularly for @xmath2 .
the following is the flow chart to estimate @xmath150 defined in eq .
( [ eq : nlate ] ) from the raw simulation data of @xmath106 shown in fig .
[ fig : corr ] . * determining the reliable range of @xmath106 data , @xmath154 , discussed in appendix [ appsec : check ] . * smoothing @xmath106 to remove the oscillatory behavior .
* fitting each smoothed @xmath106 to a linear function of @xmath26 to obtain the gradient of @xmath106 , and averaging all realizations to obtain the expectation value of the gradient and its variance .
first of all , according to the convergence check against the spatial resolution discussed in appendix [ appsec : check ] , we determined the reliable range of @xmath194 data , @xmath154 .
next , let us consider the smoothing process for the raw data of @xmath194 containing the oscillations .
our final goal is to fit @xmath106 to a linear function such as @xmath195 .
hence the smoothing process should not affect the gradient @xmath144 .
the simplest treatment would be the averaging with neighbor points in the time domain .
defining @xmath196 for @xmath197 and @xmath198 , this averaging can be described as @xmath199 where @xmath200 represents the @xmath201 th smoothed data , and @xmath202 is the original raw data .
we repeat this process until @xmath203 .
this formula can be derived by approximating the equation @xmath204 with the second - order central difference formula . in fig .
[ fig : smoothing ] , the blue line represents the resultant smoothed @xmath106 with @xmath205 times iterations , while the red line represents the original data .
it is clear that only the oscillatory behaviors are successfully removed , and the global gradient does not change during this process .
finally , we fit the linear regime of the smoothed @xmath106 in @xmath206 to a linear function , @xmath207 . according to this procedure
, we obtain ten independent values of @xmath144 from the ten sets of simulation data for a given @xmath0 .
then we can calculate the expectation value of @xmath144 , @xmath149 , and its variance , @xmath208 . finally , using the definition of @xmath150 given in eq .
( [ eq : nlate ] ) , we plot fig .
[ fig : nlate ] , where the error bar indicates @xmath209 .
t. damour and a. vilenkin , phys . rev
. lett .
* 85 * ( 2000 ) 3761 [ gr - qc/0004075 ] .
t. damour and a. vilenkin , phys .
d * 64 * ( 2001 ) 064008 [ gr - qc/0104026 ] .
t. damour and a. vilenkin , phys . rev .
d * 71 * ( 2005 ) 063510 [ hep - th/0410222 ] .
x. siemens , v. mandic and j. creighton , phys .
* 98 * ( 2007 ) 111101 [ astro - ph/0610920 ] .
x. siemens , j. creighton , i. maor , s. ray majumder , k. cannon and j. read , phys .
d * 73 * ( 2006 ) 105001 [ gr - qc/0603115 ] .
b. p. abbott _ et al .
_ [ ligo scientific collaboration ] , phys . rev .
d * 80 * ( 2009 ) 062002 [ arxiv:0904.4718 [ astro-ph.co ] ]
. b. p. abbott _ et al . _ [ ligo scientific and virgo collaborations ] ,
nature * 460 * ( 2009 ) 990 [ arxiv:0910.5772 [ astro-ph.co ] ] .
s. olmez , v. mandic and x. siemens , phys .
d * 81 * ( 2010 ) 104028 [ arxiv:1004.0890 [ astro-ph.co ] ] .
s. kuroyanagi , k. miyamoto , t. sekiguchi , k. takahashi and j. silk , phys .
d * 86 * ( 2012 ) 023503 [ arxiv:1202.3032 [ astro-ph.co ] ] .
n. bevis , m. hindmarsh , m. kunz and j. urrestilla , phys .
* 100 * ( 2008 ) 021301 [ astro - ph/0702223 ] .
k. takahashi , a. naruko , y. sendouda , d. yamauchi , c. -m .
yoo and m. sasaki , jcap * 0910 * ( 2009 ) 003 [ arxiv:0811.4698 [ astro - ph ] ] .
d. yamauchi , k. takahashi , y. sendouda , c. -m .
yoo and m. sasaki , phys .
d * 82 * ( 2010 ) 063518 [ arxiv:1006.0687 [ astro-ph.co ] ] .
d. yamauchi , y. sendouda , c. -m .
yoo , k. takahashi , a. naruko and m. sasaki , jcap * 1005 * ( 2010 ) 033 [ arxiv:1004.0600 [ astro-ph.co ] ] .
r. battye and a. moss , phys .
d * 82 * ( 2010 ) 023521 [ arxiv:1005.0479 [ astro-ph.co ] ] .
d. yamauchi , k. takahashi , y. sendouda and c. -m .
yoo , phys .
d * 85 * ( 2012 ) 103515 [ arxiv:1110.0556 [ astro-ph.co ] ] . c. dvorkin , m. wyman and w. hu , phys . rev .
d * 84 * ( 2011 ) 123519 [ arxiv:1109.4947 [ astro-ph.co ] ] .
j. urrestilla , n. bevis , m. hindmarsh and m. kunz , jcap * 1112 * ( 2011 ) 021 [ arxiv:1108.2730 [ astro-ph.co ] ] .
s. kuroyanagi , k. miyamoto , t. sekiguchi , k. takahashi and j. silk , phys .
d * 87 * ( 2013 ) 023522 [ arxiv:1210.2829 [ astro-ph.co ] ] .
p. a. r. ade _ et al .
_ [ planck collaboration ] , arxiv:1303.5085 [ astro-ph.co ] . a. vilenkin and e. p. s. shellard , cambridge university press , 1994 .
a. a. fraisse , c. ringeval , d. n. spergel and f. r. bouchet , phys .
d * 78 * ( 2008 ) 043535 [ arxiv:0708.1162 [ astro - ph ] ] .
g. r. vincent , m. hindmarsh and m. sakellariadou , phys .
d * 56 * ( 1997 ) 637 [ astro - ph/9612135 ] .
a. albrecht and n. turok , phys
. rev .
d * 40 * ( 1989 ) 973 .
a. albrecht and n. turok , phys .
* 54 * ( 1985 ) 1868 .
d. p. bennett and f. r. bouchet , phys .
d * 41 * ( 1990 ) 2408 .
b. allen and e. p. s. shellard , phys .
* 64 * ( 1990 ) 119 .
c. j. a. p. martins and e. p. s. shellard , phys .
d * 73 * ( 2006 ) 043515 [ astro - ph/0511792 ] .
k. d. olum and v. vanchurin , phys .
d * 75 * ( 2007 ) 063521 [ astro - ph/0610419 ] . c. ringeval , m. sakellariadou and f. bouchet , jcap * 0702 * ( 2007 ) 023 [ astro - ph/0511646 ]
. j. j. blanco - pillado , k. d. olum and b. shlaer , phys .
d * 83 * ( 2011 ) 083514 [ arxiv:1101.5173 [ astro-ph.co ] ] .
g. vincent , n. d. antunes and m. hindmarsh , phys .
* 80 * ( 1998 ) 2277 [ hep - ph/9708427 ] .
j. n. moore , e. p. s. shellard and c. j. a. p. martins , phys .
d * 65 * ( 2001 ) 023503 [ hep - ph/0107171 ] .
n. bevis , m. hindmarsh , m. kunz and j. urrestilla , phys .
d * 75 * ( 2007 ) 065015 [ astro - ph/0605018 ] .
p. mukherjee , j. urrestilla , m. kunz , a. r. liddle , n. bevis and m. hindmarsh , phys .
d * 83 * ( 2011 ) 043003 [ arxiv:1010.5662 [ astro-ph.co ] ] .
n. bevis , m. hindmarsh , m. kunz and j. urrestilla , phys .
d * 82 * ( 2010 ) 065004 [ arxiv:1005.2663 [ astro-ph.co ] ] . l. m. a. bettencourt and r. j. rivers , phys .
d * 51 * ( 1995 ) 1842 [ hep - ph/9405222 ] .
l. m. a. bettencourt , p. laguna and r. a. matzner , phys .
* 78 * ( 1997 ) 2066 [ hep - ph/9612350 ] .
p. salmi , a. achucarro , e. j. copeland , t. w. b. kibble , r. de putter and d. a. steer , phys .
d * 77 * ( 2008 ) 041701 [ arxiv:0712.1204 [ hep - th ] ] .
s. sarangi and s .- h .
h. tye , phys .
b * 536 * , 185 ( 2002 ) [ arxiv : hep - th/0204074 ] .
n. t. jones , h. stoica and s .- h .
h. tye , phys .
b * 563 * , 6 ( 2003 ) [ arxiv : hep - th/0303269 ] .
e. j. copeland , r. c. myers and j. polchinski , jhep * 0406 * , 013 ( 2004 ) [ arxiv : hep - th/0312067 ] . g. dvali and a. vilenkin , jcap * 0403 * , 010 ( 2004 ) [ arxiv : hep - th/0312007 ] . j. urrestilla and a. vilenkin , jhep * 0802 * ( 2008 ) 037 [ arxiv:0712.1146 [ hep - th ] ] .
a. achucarro and r. de putter , phys .
d * 74 * ( 2006 ) 121701 [ hep - th/0605084 ] .
m. yamaguchi , phys .
d * 60 * ( 1999 ) 103511 [ hep - ph/9907506 ] .
m. yamaguchi and j. yokoyama , phys .
d * 66 * ( 2002 ) 121303 [ hep - ph/0205308 ] .
m. yamaguchi and j. yokoyama , phys .
d * 67 * ( 2003 ) 103514 [ hep - ph/0210343 ] .
pen , u. seljak and n. turok , phys .
* 79 * ( 1997 ) 1611 [ astro - ph/9704165 ] .
r. durrer , m. kunz and a. melchiorri , phys .
d * 59 * ( 1999 ) 123005 [ astro - ph/9811174 ] .
m. yamaguchi , j. yokoyama and m. kawasaki , phys .
d * 61 * ( 2000 ) 061301 [ hep - ph/9910352 ] .
m. yamaguchi , m. kawasaki and j. yokoyama , phys .
* 82 * ( 1999 ) 4578 [ hep - ph/9811311 ] .
y. cui , s. p. martin , d. e. morrissey and j. d. wells , phys .
d * 77 * ( 2008 ) 043528 [ arxiv:0709.0950 [ hep - ph ] ] .
m. hindmarsh , s. stuckey and n. bevis , phys .
d * 79 * ( 2009 ) 123504 [ arxiv:0812.1929 [ hep - th ] ] .
w. h. press , b. s. ryden and d. n. spergel , astrophys .
j. * 347 * ( 1989 ) 590 .
a. a. abrikosov , sov .
jetp * 5 * ( 1957 ) 1174 [ zh .
teor . fiz .
* 32 * ( 1957 ) 1442 ] .
h. b. nielsen and p. olesen , nucl .
b * 61 * ( 1973 ) 45 . | we study the network of type - i cosmic strings using the field - theoretic numerical simulations in the abelian - higgs model . for type - i strings
, the gauge field plays an important role , and thus we find that the correlation length of the strings is strongly dependent upon the parameter @xmath0 , the ratio between the masses of the scalar field and the gauge field , namely , @xmath1 .
in particular , if we take the cosmic expansion into account , the network becomes densest in the comoving box for a specific value of @xmath0 for @xmath2 . |
it is well - known that the fate of a star after it has evolved away from the main sequence ( ms ) depends on its mass .
the accepted scenario for the post - ms evolution of low to intermediate mass stars is that , following the helium flash , burnout of he occurs in the core on the horizontal branch .
after evolution up the asymptotic giant branch , the star sheds its outer envelope , which is illuminated as a planetary nebula ( pn ) by the still - hot stellar core .
however , in as many as 20% of cases ( blcker @xcite ) the star , as it evolves towards the white dwarf ( wd ) region of the hr diagram , re - ignites a residual helium shell in a vltp and retraces its evolutionary track to the right to become a born again red giant ( bag ; see herwig @xcite and references therein ) .
the final evolution to a wd is predicted to take roughly a few centuries , thus representing a very rapid ( and hence seldom seen ) phase of stellar evolution .
a very small number of stars is known to have undergone vltps , e.g. sakurai s object ( v4334 sgr ) , v605 aql and fg sge .
all are c - rich , h - poor , have extensive dust shells , and lie at the centre of a pn .
observations of bags in the past @xmath0 years have shown that the rate at which the final evolution occurs has been seriously _ underestimated . _
we have a long - standing programme of observations of bags at infrared ( ir ) wavelengths , some extending at least over a couple of decades . these included ground - based observations ( e.g. woodward , @xcite ; tyne , @xcite ; gehrz , @xcite ) and observations using the infrared spectograph ( e.g. evans @xcite ) and the forcast grism spectrometer on the stratospheric observatory for infrared astronomy ( ) .
fg sge underwent its vltp in 1880 .
the 536 spectrum as observed by and is shown in fig .
there is clearly a dust continuum , at a temperature @xmath1 k. this is considerably lower than that determined by either woodward ( @xcite ) or gehrz ( @xcite ) , who found that there was a major dust - formation event sometime between 1983 and 1992 .
they also found that , over the period 19922001 , the dust temperature was consistently within @xmath2 k of 1000 k , indicating that mass - loss and dust production was continuing .
more recent data show that , sometime between 1992 and 2006 ( the date of the first observation ) mass loss had ceased and that the dust shell had started cooling . of interest
is the difference between the underlying 650 k blackbody and the irs spectrum , shown in fig . [ bags ] .
there is clear evidence for hydrocarbon `` unidentified infrared '' ( uir ) emission , and absorption by acetylene ( c@xmath3h@xmath3 ; 13.688 ) ; this demonstrates that the mass lost in the 19922001 episode ( gehrz , @xcite ) was carbon - rich .
also highlighted are two weak emission features at 33.45 and 34.75 ( see below ) .
the forcast observation , obtained in 2014 , show that the dust has further cooled ( see fig .
[ bags ] ) .
the vltp of sakurai s object occurred in 1996 .
thorough analyses of the ir development and the implications for the evolution of the ( carbon ) dust produced are given by tyne ( @xcite ) and hinkle & joyce ( @xcite ) .
this object also showed c@xmath3h@xmath3 in absorption , as well as hcn and a range of polyynes ( hc@xmath4n ; evans @xcite ) .
the ejection of an optically thick carbon dust shell commenced in late 1997 ( evans @xcite ; hinkle & joyce @xcite ) ; following the dust ejection event the central star has been completely ( @xmath5 ) obscured , and has only recently ( 2015 february ) become visible at @xmath6 .
exquisite observations by chesneau ( @xcite ) with the mid - ir interferometer midi / vlti showed that the dust was confined to a disc .
the cooling of the dust shell from 620 k to 180 k , as observed from the ground ( ukirt+jcmt ) and from space- and air - borne ( , ) observations , is shown in fig .
the hcn isotopologues in the irs spectra show that the @xmath7c/@xmath8c ratio is @xmath9 , consistent with that deduced from the fundamental and first overtone co absorption ( eyres , @xcite ; pavlenko , @xcite ) , and with its vltp status .
the nature of ck vul is unclear . once thought to be the oldest `` old nova '' ( shara , @xcite ) it erupted in 1670 it has also been suggested that it is a bag ( evans , @xcite and references therein ) , a diffusion - induced nova ( miller bertolami , @xcite ) , and a stellar merger ( kaminski , @xcite ) .
the irs spectrum , of both the central object and the inner nebula , shows a number of uir features on a 170 k blackbody continuum ; the difference between the irs spectrum of the central object , and a 170 k blackbody is shown in fig .
there are a number of ionic features ( e.g. [ o ] at 25.89 ) and we again see uir features that are completely different from those commonly seen in novae ( helton , @xcite ) .
we also note the two features at 33.45 and 34.75 , also present in fg sge ( see above ) .
they are labelled as being due to [ s ] and [ si ] in fig .
[ bags ] but they are extraordinarily strong in ck vul compared to their strength in fg sge .
there seems to be no reasonable way in the case of ck vul to account for these features in terms of ionic emission only
. it may be that they arise either in entirely different environments in the two objects , or in different carriers .
observations of the bag phenomenon at mid - to - far ir ( 538 ) wavelengths provide a wealth of information ( e.g. about the mineralogy of the dust , far ir fine structure lines ) and hence about the circumstellar environment of bags that can be obtained by no other means .
observations of bags using are continuing and will be reported elsewhere
. 99 blcker , t. , 2001 , ap&spsci , 275 , 1 chesneau , o. , , 2009 , a&a , 493 , l17 evans , a. , , 2002 , mnras , 332 , l35 evans , a. , , 2006 , mnras , 373 , l75 eyres , s. p. s , , 2004 , mnras , 350 , l9 gehrz , r. d. , , 2005 , apj , 623 , 1105 helton , l. a. , , 2011 , in _ pahs and the universe _ , eas publ .
series , vol .
46 , 407 herwig , f. , 2005 , ara&a , 43 , 435 hinkle , k. h. , & joyce , r. r. , 2014 , apj , 785 , 146 miller bertolami , m. m. , , 2011 , mnras , 415 , 1396 pavlenko , ya . v. , , 2004 , a&a , 417 , l39 ; kaminski , t. , , 2015 , nature , 520 , 322 shara , m. m. , , 1985 , apj , 294 , 271 tyne , v. h. , , 2002 , mnras , 334 , 875 woodward , c. e. , , 1993 , apj , 408 , l37 | we describe the evolution of the carbon dust shells around very late thermal pulse ( vltp ) objects as seen at infrared wavelengths .
this includes a 20-year overview of the evolution of the dust around sakurai s object ( to which olivier made a seminal contribution ) and fg sge .
vltps may occur during the endpoint of as many as 25% of solar mass stars , and may therefore provide a glimpse of the possible fate of the sun . |
over the last several decades , with new evidence , the objects we call galaxies " have become much larger . extended dark - matter halos " were proposed to produce flat rotation curves at large radii in disk galaxies ( rubin et al . 1980 ) , and
a corona " of hot interstellar gas at the galaxy s virial temperature was predicted by spitzer ( 1956 ) to provide pressure confinement of high - latitude clouds .
more recently , astronomers have observed galactic kinematic tracers ( blue horizontal branch stars , globular clusters , satellite galaxies ) to distances of 50 - 250 kpc , and x - ray absorption - line ( ) spectroscopy and stacked soft x - ray emission have provided evidence of large reservoirs of hot ionized gas in milky way halo ( miller & bregman 2013 ) and the outskirts of external galaxies ( soltan 2006 ; anderson et al .
2013 ) . in ultraviolet spectroscopy , the cosmic origins spectrograph ( cos ) on the _ hubble space telescope _
( hst ) has recently detected extended ( 100 - 150 kpc ) reservoirs of highly ionized oxygen ( ) around star - forming galaxies ( tumlinson et al .
2011 , 2013 ; stocke et al . 2013 ) likely created by outflows of metal - enriched gas from star formation .
thus , our picture of galaxies has evolved to a system of stars and gas embedded in an extended dark - matter halo , often associated with an even larger gaseous circumgalactic medium ( cgm ) .
galaxies must also be viewed in a cosmological context , in which most of the baryonic matter in the universe is unseen ( persic & salucci 1992 ) and likely distributed through the intergalactic medium ( igm ) in a cosmic web " shaped by dark - matter structure and inefficient galaxy formation ( cen & ostriker 1999 ; dav et al .
1999 ; smith et al . 2011 ) . observations and
modeling ( shull et al . 2012 ) suggest that 60 - 80% of the cosmological baryons reside in the low - redshift , multi - phase igm , with perhaps 20% in collapsed form ( galaxies , groups , clusters ) .
although these baryon fractions have some uncertainty , current research is devoted to understanding the physical conditions , spatial extent , and evolution of the gas at distances of 100 kpc to a few mpc from galaxies . at low redshift ( @xmath13 )
the influence of galactic winds and metal injection out to @xmath14 mpc has been inferred from the association of qso absorption systems of and with nearby galaxies ( penton et al .
2002 ; prochaska et al .
2011 ; stocke et al .
2006 , 2013 ) .
galactic outflows have been detected in absorption toward intermediate - redshift galaxies ( steidel et al .
2004 ; martin et al .
2012 ; tripp et al .
2011 ) , while at higher redshifts , @xmath15 , rudie et al .
( 2012 ) found an enhancement of circumgalactic ( @xmath16 kpc ) absorbers in a sample of 886 star - forming galaxies probed by 15 background qsos .
thus , the connection between galaxies and extended absorption systems seems secure .
perhaps because of the recent nature of these discoveries , a semantic problem has arisen regarding the proposed structures : halo , cgm , igm .
when does gas cease to be circumgalactic and become intergalactic ?
are the edges of galaxies defined by gravity or gas outflows ?
are quasar absorption lines the extended halos of intervening galaxies ( bahcall & spitzer 1969 ) or filaments of intergalactic gas ( sargent et al .
similarly , the phrase circumgalactic medium " appears to have replaced the concept of a gaseous halo " or
galactic corona " of hot interstellar gas at the galaxy s virial temperature .
ionized gas with high covering factor has been detected above the galactic disk in uv absorption - line surveys in metal ions such as ( sembach et al .
2003 ) and ( shull et al . 2009 ; lehner & howk 2011 ) and in soft x - ray absorption lines of or at @xmath17 ( nicastro et al . 2002 ; mckernan et al . 2005 ;
wang et al . 2005 ) .
the uv absorbers appear to come primarily from gas within 2 - 10 kpc of the disk plane , elevated by supernovae and star formation in the disk .
the galactic x - ray absorption suggests hot gas at @xmath18 k , but its radial extent is controversial .
it may come from a 50-kpc halo ( anderson & bregman 2010 ; gupta et al .
2012 ) although absorption toward background agn ( fang et al . 2006 ; hagihara et al .
2010 ) and x - ray binaries ( yao & wang 2005 ; hagihara et al .
2011 ) suggests that much of the absorption comes within several kpc of the disk .
the disk model is consistent with both x - ray observations and total mass considerations ( collins et al .
2005 ; fang et al .
2006 ) .
the intent of this paper is to improve the definition of galaxy halos as regions of strong gravitational influence , using dynamical principles and observational constraints .
section 2 discusses physical measures of the spatial extent of large ( @xmath19 ) galaxies including the milky way and andromeda .
we discuss the somewhat arbitrary and occasionally misused definition of virial radius " .
somewhat better defined are the gravitational radius " , @xmath20 , derived from galactic mass and potential energy , the gravitational sphere of influence , accretion radius , and tidal radius .
section 2.3 develops a physically realistic definition of @xmath21 for halos of mass @xmath3 , assembled primarily at redshifts @xmath4 with further mass accretion down to the present epoch .
these new virial radii are typically 50 - 60% the sizes used to analyze hot halo gas with hst / cos .
section 3 discusses four estimates of galaxy extent : ( 1 ) recent kinematical studies of the milky way and andromeda ( m31 ) ; ( 2 ) the galactopause " where outflow ram pressure balances thermal pressure of the cgm ; ( 3 ) qso absorber cross sections derived from metal absorption - line frequency in redshift ; and ( 4 ) virial radii and halo masses obtained from galaxy abundance - matching .
we also discuss recent estimates of the mass and size of the milky way and andromeda halos from kinematics of stars , galactic satellites , and the local group .
most of our estimates suggest a smaller spatial extent ( @xmath6 kpc ) for galaxies of mass @xmath22 , comparable to the milky way and m31 .
section 4 concludes with applications of the new definition of virial radius to observations of extended gas around galaxies made with hst / cos , and to recent mass and size measurements for the milky way and m31 .
this section begins with an overview of the cosmological context , in which galaxies form from density perturbations that turn around from the hubble expansion background and collapse into dynamical ( virial ) equilibrium , but with continued infall of dark matter and gas .
the environment of these structures will be influenced by feedback " from star formation within galaxies , through outflows that escape to the igm or become stalled and return to the galaxy . both infall and
outflow processes depend on the gravitational attraction of the galaxy and its halo , as measured by parameters such as escape velocity , accretion radius , and tidal radius .
below , these measures of gravitational influence are explored , leading to a practical definition of virial radius , consistent with collapse in the past ( @xmath23 ) with continued mass infall to the present .
the classic overdensity @xmath24 ( often rounded to 200 ) was originally derived analytically for a spherical top - hat density perturbation in an einstein - desitter universe . the mean background density , @xmath25 , following the behavior of the expansion parameter @xmath26 .
the perturbation begins its collapse at a turn - around " density , @xmath27 , at a redshift @xmath28 .
for example , if all the matter in the local group were associated with a spherically distributed mass @xmath3 , the proper size of the mass perturbation at turnaround would be @xmath29 ^{1/3 } \approx ( 1.2~{\rm mpc } ) \left ( \frac { m_{\rm lg } } { 5\times10^{12}~m_{\odot } } \right ) ^{1/3 } \left [ \frac { 1.5 } { 1+z_{\rm turn } } \right ] \ ; .\ ] ] here , we have scaled to the local group mass , @xmath30 at a turnaround redshift @xmath31 , corresponding to the recently measured dynamical history of the milky way and andromeda system ( van der marel et al .
these galaxies turned around from the hubble flow even earlier , perhaps at @xmath32 , and had correspondingly smaller radii , probably @xmath33 mpc .
the fact that a given galaxy halo is assembled over a range of redshifts suggests an improved definition of virial radius , @xmath34 , for a galaxy observed at redshift @xmath35 but assembled primarily at redshift @xmath36 , when half its halo mass had collapsed .
this change in definition increases the background matter density , @xmath37 , and makes the virial radii smaller .
we return to this new formalism in sections 2.3 and 3.4 . for extended self - gravitating systems without a sharp boundary ,
it is useful to define a characteristic gravitational radius " ( binney & tremaine 2008 ) as @xmath38 , where @xmath39 is the gravitational potential energy , @xmath40 the gravitational radius also bears a close relation to the half - light radius " , @xmath41 , defined by spitzer ( 1969 ) for spherical stellar systems .
for many mass distributions , @xmath42 , so that @xmath43 and @xmath44 .
for example , a homogeneous sphere of mass @xmath45 and radius @xmath46 has @xmath47 and @xmath48 . for the density distribution of plummer s model , @xmath49^{-5/2}$ ] , @xmath50 and @xmath51 , where @xmath52 is the plummer scale length .
the structure of collapsed halos has been well studied through cosmological n - body simulations ( navarro , frenk , & white 1997 ; springel et al .
2005 ; klypin et al .
these numerical experiments show that dark - matter halos collapse into structures with cuspy cores and extended halos .
the collapsed structures in these simulations have been fitted to various radial profiles , such as nfw ( navarro , frenk , & white 1997 ) , for which the density , potential , and enclosed mass are , @xmath53 ^2 } \\
\phi(r ) & = & - \left ( 4 \pi g \rho_0 r_s^2 \right ) \frac { \ln ( 1 + r / r_s ) } { r / r_s } \\
m(r ) & = & \left ( 4 \pi \rho_0 r_s^3 \right ) \left [ \ln \left ( 1 + \frac{r}{r_s } \right ) - \frac { r / r_s}{1 + r / r_s } \right ] \ ; .\end{aligned}\ ] ] here @xmath54 is a characteristic radius defined by the break in slope and related to the virial radius through @xmath55 , with the concentration parameter " , @xmath56 , typically between 520 .
the mass enclosed with the virial radius is written as @xmath57 \ ; , \ ] ] where @xmath58 and @xmath56 are found to be correlated in cosmological simulations ( navarro et al .
1997 ; prada et al .
2012 ) . increasing the concentration , @xmath56 ,
adds mass and extent to the halo .
for example , changing @xmath56 from 5 to 20 doubles the enclosed mass , @xmath59 .
the circular velocity , @xmath60 , reaches a maximum @xmath61 at radius @xmath62 , and equal to @xmath63 } \ ; .\ ] ] for the nfw model , the mass diverges logarithmically with radius , and the gravitational radius is formally undefined without a cut - off . by integrating equation ( 2 ) out to the virial radius , @xmath64 , we can define a finite nfw potential energy and gravitational radius , @xmath65 } { \left [ \ln ( 1+c ) - \frac { c}{1+c } \right]^2 } \\ r_g ^{\rm ( nfw ) } & = & gm_{\rm vir}^2 / \vert w_{\rm vir } \vert = ( 2 r_s ) \ ; \frac { \left [ \ln ( 1+c ) - \frac { c}{1+c } \right]^2 } { \left [ 1 - \frac { \ln ( 1+c ) } { ( 1+c ) } - \frac { 1}{(1+c ) } \right ] } \ ; .\end{aligned}\ ] ] this nfw gravitational radius ranges from @xmath66 ( for @xmath67 ) to @xmath68 ( for @xmath69 ) .
table 1 summarizes the ratio of gravitational radius to characteristic scale lengths for several commonly used density distributions .
another observationally useful characteristic radius is the gravitational radius of influence " ( ferrarese & ford 2005 ; binney & tremaine 2008 ) defined as the radius at which the kepler velocity around a central mass @xmath45 equals the transverse velocity dispersion , @xmath70 , of surrounding stars , @xmath71 in the galactic context of this paper , we assume that @xmath45 is the total mass of the galaxy within radius @xmath72 , scaled as @xmath73 .
the gravitational influence of this mass can traced by luminous halo stars or galaxy satellites with dispersion @xmath74 .
the dynamical dependence of @xmath75 ( observed in the stellar component ) on @xmath45 will soften the linear dependence of @xmath76 . in the milky way ,
radial velocity dispersions , @xmath77 km s@xmath78 , of different halo populations have been measured out to 80 kpc ( xue et al .
2008 ; gnedin et al .
2010 ) . from radial velocities of blue horizontal branch stars between 100 - 150 kpc ,
deason et al .
( 2012 ) find even lower dispersions , @xmath79 km s@xmath78 , although values for satellites are somewhat larger .
thus , depending on assumptions about the ratio , @xmath80 , the value of @xmath72 for the milky way halo may be somewhat larger than 200 kpc .
this formula is analogous to that used in bondi accretion of gas with isothermal sound speed @xmath81 , for which one defines an accretion radius " @xmath82 here again , we have scaled to a @xmath83 galaxy surrounded by a gaseous halo at its virial temperature @xmath84 with @xmath85 for mean atomic weight @xmath86 appropriate for a fully ionized plasma with @xmath87 by number .
the fact that these two raidii are similar is a consequence of the chosen gas temperature and stellar velocity dispersion , which are expected to be in near equilibrium . in standard cosmological terminology , the virial radius , @xmath0 ,
is defined such that the mean density of a halo of mass @xmath3 equals the virial overdensity " ( variously denoted as @xmath88 or @xmath89 ) times the mean density of the universe .
unfortunately , some confusion has arisen in the proper value of @xmath88 , much of it over convention : whether the overdensity is expressed relative to the critical ( closure ) density , @xmath90 , or to the matter density , @xmath91 .
both conventions are used in the literature , but if one is not careful in using a consistent definition of @xmath88 , the estimated virial radius will be too small by a factor @xmath92 and the mean matter density of the halo too large by @xmath93 .
there are other misapplications of the virial radius because the initial collapse to dynamical equilibrium occurred in the past , when @xmath94 was higher , followed by additional infall during galaxy assembly . for these reasons ,
we develop a physically motivated definition of @xmath95 for a halo of mass @xmath3 , observed at redshift @xmath35 but assembled at an earlier epoch , @xmath36 .
the redshift of halo assembly @xmath96 is taken as the epoch of major virialization ( @xmath97 ) when half the halo mass had collapsed ,
a useful convention introduced by tacchella , trenti , & carillo ( 2013 ) hereafter denoted ttc13 .
the basic concept of virialization follows from the assumption of conservation of kinetic energy and potential energy during collapse ( @xmath98 ) together with the virial theorem ( @xmath99 ) .
the gravitating system is assumed to collapse to half its turnaround radius at twice the turnaround time for cycloidal collapse .
the collapsed system has then increased in mean density by a factor of 8 since turnaround . in an einstein - desitter universe ( @xmath100 ) , the expanding background density , @xmath25 , has dropped by a factor of 4 for an expansion parameter @xmath26 .
thus , one obtains the classic virial overdensity , @xmath101 , where @xmath102 and @xmath103 are evaluated at the times of turnaround and virialization , respectively .
cole & lacey ( 1996 ) explored the structure of halos in n - body simulations of clustering in an @xmath100 universe .
they found that the radius , @xmath104 enclosing a mean overdensity of 178 , accurately separated the virialized halo interior , in approximate dynamic equilibrium , from the exterior where matter was still falling in .
this property was noted by binney & tremaine ( 2008 ) , who recommended using @xmath105 .
the situation changes in a flat @xmath106cdm cosmology ( @xmath107 ) where both the expanding background and collapse dynamics differ from the @xmath100 case .
eke et al .
( 1996 ) studied virialization dynamics for flat models with @xmath108 , but they defined overdensity relative to _ closure _ density . for values of @xmath109 at @xmath110 , they found @xmath111 . in figure 1 of their paper , they plotted the dependence of @xmath89 on @xmath112 , which serves as a proxy for higher redshift , since @xmath113 approaches 1 at the higher redshifts when most systems collapsed .
bryan & norman ( 1998 ) provided an approximate fit , @xmath114 $ ] where @xmath115 .
this expression must be divided by @xmath113 to be consistent with the convention that overdensity is relative to _ matter _ density .
we define @xmath0 through the relation @xmath116 , where the mean matter density at the virialization ( or assembly ) redshift is @xmath117 , and the parameters @xmath118 and @xmath119 are defined at the current epoch ( @xmath120 ) .
further discussion is given by klypin et al .
( 2011 ) and van der marel et al .
( 2012 ) , both of whom follow the convention that @xmath88 is relative to the mean _ matter _ density .
their expressions are scaled to a product , @xmath121 , evaluated at @xmath120 for which @xmath122 and @xmath123 . however , many papers in the literature do not account for the fact that virial collapse occurred at earlier times , at redshift @xmath97 when @xmath124 . for a flat @xmath106cdm universe with @xmath125 at @xmath110
, one finds @xmath126 ( relative to @xmath94 ) and a product @xmath127 . for the more typical collapse at higher redshifts , the product @xmath128 , appropriate for the decrease in @xmath129 as @xmath130 .
this overdensity of 200 is slightly above the classical value @xmath131 for dynamical reasons discussed by binney & tremaine ( 2008 ) .
consider a galaxy observed at redshift @xmath35 , associated with halo mass @xmath132 and assembled ( at @xmath96 ) with half its current mass .
the virialization redshift @xmath97 is taken to be @xmath96 , determined from standard cosmological collapse criteria ( lacey & cole 1993 ; sheth & tormen 1999 ) and computed as in trenti , perna , & tacchella ( 2013 ) .
table 2 lists the assembly redshifts , which range from @xmath133 ( at @xmath134 ) to @xmath135 ( at @xmath136 ) for various galaxies observed at @xmath137 with stellar masses @xmath2 . based on these new criteria for galaxy collapse and assembly ,
the virial radius is : @xmath138 ^{-1/3 } \left ( 1+z_a \right)^{-1 } \ ; .\ ] ] although the relation between @xmath139 exhibits a small anti - correlation ( table 2 ) , it does contain some numerical scatter which should be kept in mind .
this expression differs from often - used formulae by the over - density scaling and by the factor @xmath140 , reflecting the fact that most galaxies underwent virialization in the past , not at @xmath120 . after this initial ( half - mass ) assembly , their proper size changes gradually because of continued mass infall into the halo .
to lowest order , this slight growth in @xmath0 ( @xmath141 ) is captured in eq .
( 12 ) by using the _ current _ halo mass ( @xmath3 ) .
second - order effects of late infall or adiabatic compression may occur in the radial mass distribution , which in the nfw formalism changes the scale parameter @xmath54 and concentration @xmath56 .
some of the infalling matter may not attain virial equilibrium , as the orbital timescale is quite long , @xmath142 .
this formalism assumes spherical infall and a constant density profile ; both approximations change in more realistic collapse scenarios .
for example , after accounting for central condensation of the collapsed structure , rubin & loeb ( 2013 ) find a reduced virial overdensity .
in addition to infall , galaxy halos can be probed by fast stars injected radially outward .
we therefore discuss the radius at which the fastest galactic stars reach maximum apogalactic distance in a dark matter halo potential . as a simple model that illustrates the basic effect , we assume an isothermal sphere distribution in density , @xmath143 and enclosed mass , @xmath144 leading to a constant circular velocity @xmath145 . for this density and mass distribution ,
the equation of motion is @xmath146 , for a star launched with radial velocity @xmath147 from radius @xmath148 , leading to the first integral , @xmath149 = - ( 2 \sigma^2 ) \ln ( r / r_0 ) \ ; .\ ] ] if the star coasts to rest ( @xmath150 at @xmath151 ) one has @xmath152 and thus @xmath153 \ ; .\ ] ] equation ( 14 ) illustrates the sensitivity of @xmath154 to launching conditions ( @xmath155 and @xmath156 ) and halo mass ( velocity dispersion @xmath75 ) .
for example , a star with @xmath157 km s@xmath78 injected radially from @xmath158 kpc into a halo with @xmath159 km s@xmath78 would have a maximum extent of just 40 kpc .
the stellar ejection distance doubles ( @xmath160 kpc ) if the launch velocity @xmath161 km s@xmath78 .
these distances are not surprising for velocities comparable to the escape speed from the solar circle within the milky way ( @xmath162 kpc ) , recently estimated from rave stellar survey data to be @xmath163 km s@xmath78 ( piffl et al . 2014 ) . repeating the above calculation for the maximum radius in an nfw mass distribution and potential ( equations 3 , 4 , 5 ) we find , @xmath164 } \left [
\frac { \ln \left ( 1 + r_0 / r_s \right ) } { r_0 / r_s } - \frac { \ln \left ( 1 + r_{\rm max } / r_s \right ) } { r_{\rm max } / r_s } \right ] \ ; .\ ] ] this equation can be solved for the velocity @xmath155 required for a star to reach radius @xmath165 from initial radius @xmath156 .
combined with equation ( 7 ) for the maximum circular velocity , @xmath61 , in an nfw halo , we can write the velocity @xmath155 needed to escape to @xmath166 as @xmath167 ^{1/2 } \ ; , \ ] ] independent of @xmath56 . if we take @xmath168 as the radius at @xmath61 , we find @xmath169 .
this agrees reasonably well with the milky way values , if we assume @xmath170 km s@xmath78 and obtain an escape velocity @xmath171 km s@xmath78 .
a related size estimate comes from the hypothetical radius , @xmath172 , to which one must extend an isothermal density distribution to maintain a constant ( flat ) circular velocity , @xmath173 \ ; .\ ] ] of course , rotation curves do not remain flat at large radii , but @xmath172 provides a lower limit on the extent of the isothermal model , as noted by smith et al .
( 2007 ) . for @xmath163
km s@xmath78 and @xmath174 km s@xmath78 , one finds a truncation radius @xmath175 kpc for @xmath158 kpc , with a large range ( 40 to 110 kpc ) arising from uncertainties in determining @xmath176 and @xmath177 . at larger radii , the circular velocity , @xmath178 likely declines , particularly if the halo density distribution falls off as @xmath179 as in the nfw model .
the total ( virial ) masses of galaxies include substantial mass beyond 50 - 80 kpc , particularly for halos with @xmath180 .
extra mass accretion may account for the discrepancy between low and high values for the milky way mass estimates .
it is also consistent with the physically motivated revision to the method in which the virial radius is calculated at the galaxy assembly time " at redshift @xmath36 ( section 2.3 ) .
we return to these observational issues in section 3.4 . with the current state of observations ,
these kinematic constraints suggest halo sizes of 150 - 200 kpc for @xmath5 galaxies .
future observations of rotation curves , @xmath178 , and more distant kinematic tracers ( blue horizontal branch stars , hypervelocity stars , globular clusters , and satellites ) will be of great help .
the rave survey data will be greatly improved by the _ gaia _
mission , leading to better determinations of escape velocity from the solar circle , and later with lsst , through studies of galaxy satellites , bhb stars , and rr - lyrae stars .
despite their proximity , the milky way ( mw ) and andromeda ( m31 ) galaxies remain surprisingly uncertain in their inferred mass and extent .
recent mw mass estimates range from just under @xmath83 ( smith et al . 2007 ;
xue et al . 2008 ; gnedin et al .
2010 ; deason et al .
2012 ) to values between @xmath181 ( boylan - kolchin et al . 2013 ; nesti & salucci 2013 ) .
some of the lower masses are inferred for stellar tracer distances well within @xmath0 .
for example , gnedin et al . (
2010 ) use hypervelocity stars at distances beyond 25 kpc to find @xmath182 .
xue et al . ( 2008 ) use 2400 blue horizontal branch stars to estimate @xmath183 .
however , they extrapolate to find a virial mass @xmath184 for two models of the circular velocity curve .
busha et al .
( 2011 ) use lmc and smc kinematics to infer @xmath185 .
boylan - kolchin et al .
( 2013 ) provide a thorough discussion of the previous mw mass measurements , both large and small , but they base their mass on the high probability that the leo i satellite is gravitationally bound . for a leo i distance ( @xmath186 kpc ) and galactocentric space velocity ( 200 km s@xmath78 ) , they find a median milky way virial mass @xmath187 with an asymmetric 90% confidence interval of @xmath188\times10^{12}~m_{\odot}$ ] and @xmath189 kpc for the median mass .
nesti & salucci ( 2013 ) use dark - matter halo models to estimate @xmath190 but with substantial uncertainties for both burkert and nfw density profiles .
piffl et al .
( 2014 ) use the inferred escape speed at the solar circle , @xmath191 km s@xmath78 from the rave survey , to estimate @xmath192 and a virial radius @xmath193 kpc .
watkins et al .
( 2010 ) find @xmath194 within 300 kpc from 26 satellite galaxies , and mcmillan ( 2011 ) proposes a mass model with @xmath195 .
based on the recent dynamical estimates out to larger distances , we conclude that the milky way virial mass is likely close to @xmath196 .
mass estimates for the andromeda ( m31 ) galaxy include @xmath197 within 300 kpc from 23 satellite galaxies ( watkins et al .
2010 ) and @xmath198 within 200 kpc from kinematics of 53 globular clusters ( veljanoski et al .
tamm et al .
( 2012 ) used photometry and the m31 rotation curve , together with several mass distributions , to find a dark - matter halo mass @xmath199 , stellar mass @xmath200 , and virial radius 189 - 213 kpc .
fardal et al .
( 2013 ) used bayesian simulation models of the giant southern stream around andromeda to derive @xmath201 .
they note that this mass , and those recently inferred for the milky way , alleviate the tension between virial mass estimates and the local group timing mass , recently estimated at @xmath202 ( van der marel et al .
2012 ) . because both the milky way and andromeda halos have likely accreted substantial mass beyond their observable radii ( 60 - 80 kpc ) , their virial masses are taken here to be @xmath203 and @xmath204 , with a sum , @xmath205 , close to the inferred timing mass of the local group .
simulations suggest that the remaining mass may consist of diffuse dark matter " produced by the destruction of sub - halos ( gao et al .
cosmic scatter could reduce the local group mass from the value above , as has been suggested by van der marel et al .
( 2012 ) and gonzalez , kravtsov , & gnedin ( 2013 ) .
another measure of gravitational influence is from tidal effects between nearby galaxies .
we define the characteristic tidal distance in terms of masses and densities of the two galaxies , treated as deformable fluid spheres ( chandrasekhar 1963 ) of masses ( @xmath206 , @xmath207 ) and radii ( @xmath208 , @xmath209 ) , @xmath210 in the case of the milky way ( mw ) and andromeda ( m31 ) system , internal tidal influences are seen in the form of large - scale streams of stars ( fardal et al .
2013 ; ibata et al .
2014 ) , but no obvious tails or bridges produced by external tides are detected .
ibata et al . (
2014 ) find a smooth stellar - light halo in m31 out to 150 kpc , with internal streams from tidally stripped satellites .
usually the roche formula is applied to situations with @xmath211 .
however , in the case of the mw - m31 system , we assume @xmath212 to estimate that @xmath213 for each galaxy . here , @xmath208 can be regarded as the effective radius of either galaxy , each of mass @xmath214 . if the distance between the two galaxy centers is @xmath215 kpc ( van der marel et al .
2012 ) , the absence of strong mutual tidal effects limits their tidal radius to @xmath216 .
equation ( 18 ) then implies that most of the stellar and gaseous matter in the two galaxies is confined within radius @xmath217 kpc .
the fluid approximations in making this estimate are large , and more accurate , time - dependent models and searches for external tidal effects would be helpful to set a better limit .
many galaxies undergoing active star formation have been spectroscopically observed at @xmath218 ( steidel et al .
2004 ) to have large - scale outflows with bulk velocities of 200 - 300 km s@xmath78 . for a sample of over 200 star - forming galaxies at redshifts @xmath219 , blueshifted absorption at greater than 100 km s@xmath78
is seen toward 20% of the population ( martin et al .
because the outflows are collimated , the outflow fraction " must be signiifcantly higher than the blueshifted fraction " .
some 2.5% have velocities greater than 200 km s@xmath78 , with the largest doppler components suggesting velocities up to 500 km s@xmath78 .
the outflow fraction depends on the observed star - formation rate ( sfr ) and is three times higher for galaxies with sfr @xmath220 compared to those at @xmath221 . if these outflows persist at the observed rates for @xmath14 gyr , as martin et al .
( 2012 ) suggest , they could remove a large mass of baryons , up to @xmath222 . such metal - enriched outflows could be responsible for the large baryon reservoirs of gas and metals around star - forming galaxies observed in absorption as part of the cos - halos project .
tumlinson et al .
( 2011 , 2013 ) found substantial column densities , @xmath223 @xmath224 out to distances of 100 - 150 kpc , toward a sub - sample of 28 galaxies with active star formation .
these -bearing halos surround galaxies with sfrs of @xmath225 yr@xmath78 and specific sfrs greater than @xmath226 yr@xmath78 .
how far will these galactic winds expand before stalling ?
several simulations ( furlanetto & loeb 2003 ; oppenheimer & dav 2008 ) estimate that these outflows expand to @xmath227 kpc , distances inferred observationally in metal - line auto - correlation of absorbers ( martin et al .
a simple estimate illustrates the main effects , as winds encounter surrounding gas in the cgm and produce a galactopause " where the wind ram pressure , @xmath228 , equals the confining gas pressure of the cgm , @xmath229 .
here , @xmath230 is the mass - loss rate of a conical outflow into total solid angle @xmath231 , @xmath232 is the mass density , and @xmath233 is the wind velocity .
observers and modelers typically relate @xmath234 to the sfr through a mass - loading factor " , defined here as @xmath235 .
this mass loading has been inferred from optical and x - ray data to lie in the range @xmath236 ( strickland & heckman 2009 ) . in a careful study of the dwarf
starburst galaxy ngc 1569 , martin , kobulnicky , & heckman ( 2002 ) use observations of the sfr , h@xmath237 emission , metallicities , and x - ray emission from the outflow to estimate @xmath238 for the mass ratio of x - ray emitting wind to stellar ejecta . with standard assumptions about the ratio of ejecta mass to star formation
, this range corresponds to mass loading factors @xmath239 . for the cgm gas pressure in low - redshift , actively star - forming galaxies
, we use the observed column densities ( @xmath240 @xmath224 ) and spatial impact parameters ( 100 - 150 kpc ) from cos - halos data , which suggest total hydrogen column densities @xmath241 @xmath224 for solar oxygen abundances , ( o / h ) @xmath242 , and ionization fractions @xmath243 .
stocke et al .
( 2013 ) estimated a mean pressure @xmath244 @xmath245 k ( with a range from 3 to 60 @xmath245 k ) for warm clouds in the cgm of low - redshift galaxies . using x - ray absorption through the galactic halo , miller & bregman ( 2013 ) modeled thermal pressures @xmath246 $ ] @xmath245 k for halos with characteristic scales of [ 50 , 100 ] kpc and gas temperature @xmath247 .
thus , in the inner portions of the cgm that would initially confine the outflows , the gas pressure is scaled to @xmath248 , appropriate for hydrogen number density @xmath249 , halo virial temperature @xmath250 , and fully ionized gas with he / h @xmath251 by number . for wind speeds @xmath252 ,
the approximate radius of the galactopause is @xmath253 the combination of parameters ( @xmath254 ) is expected to be of order unity , while bipolar wind solid angles @xmath231 typically cover 10 - 40% of @xmath255 sterradians for starburst galaxies ( veilleux , cecil , & bland - hawthorn 2005 ) . if the galaxy wind breaks through the higher - density cgm , the igm pressure will be considerably lower , and the metal - enriched winds will expand well beyond 200 kpc , depending on the duration of the starburst outflow .
simulations of the low - redshift igm ( e.g. , smith et al . 2011 ) find a wide range of intergalactic gas pressures in the ly@xmath237 forest ( @xmath256 ) and hotter -bearing gas ( @xmath257 ) .
in the above estimate , the star - formation rates were scaled to @xmath258 . in the sample of active star - forming galaxies at @xmath259 ( martin et al .
2012 ) , the observed sfrs range from 20 - 100 @xmath260 , while in the sample of 28 active cos - halos galaxies at @xmath261 ( tumlinson et al .
2013 ) the sfrs are 110 @xmath260 .
these rates likely reflect the general decline in sfr history among all galaxies , from @xmath262 to the present epoch .
indirect estimates of the sizes of extended gaseous halos and metal - enriched outflows around galaxies can be made from observations of qso absorption lines . in a pioneering paper ,
bahcall & spitzer ( 1969 ) proposed that these absorbers are produced in extended halos of normal galaxies " and used the line frequency , @xmath263 , the number of absorbers per unit redshift , to constrain the halo cross section .
their estimate , @xmath264 kpc , for the radius of a spherical halo is comparable to recent estimates from uv spectra of low - redshift qso absorbers obtained from hst ( shull et al .
1996 ; stocke et al . 2006
; prochaska et al .
2011 ) . those uv studies related line frequency of absorbers in ( ly@xmath237 ) or the doublet ( 1032 , 1038 ) to the absorber cross section and space density , @xmath265 , of galaxies , using a luminosity function to extrapolate to dwarf galaxy scales . in this formalism , the igm absorbers are associated with the nearest - neighbor galaxy , a procedure expected to be correct statistically , even though some absorbers have alternate possible galaxies within 200600 kpc ( stocke et al.2013 ) . at redshifts @xmath13 and ignoring cosmological effects , the number of of absorbers per unit redshift is @xmath266 , from which one infers an absorber size @xmath267 ^{1/2 } \approx \left ( 157~{\rm kpc } \right ) \left [ \frac { d{\cal n } / dz } { 20 } \right]^{1/2 } \left [ \frac { \phi_{\rm gal } } { 0.06~{\rm mpc}^{-3 } } \right ] ^{-1/2 } \\ \ ; .\ ] ] this formula uses the observed @xmath268 down to 10 m equivalent width for the 1032 absorption line in low - redshift absorber surveys ( danforth & shull 2008 ; tilton et al .
we adopt a hubble constant @xmath269 , a hubble length @xmath270 , and a galaxy luminosity function , @xmath271 with normalization @xmath272 at b - band absolute magnitude @xmath273 and faint - end slope @xmath274 given by the millennium galaxy catalog ( driver et al .
2005 ) and converted to @xmath275 scale .
the galaxy space density is scaled to 0.06 mpc@xmath276 , a value appropriate for extrapolation of @xmath277 down to dwarf galaxies of luminosity @xmath278 , which have @xmath279 times the space density of galaxies with @xmath280 for @xmath281 .
these sizes are consistent with surveys that attempt to match the nearest - neighbor galaxy , in 3d space , to low - redshift absorbers ( stocke et al . 2006 , 2013 ) .
the galaxy spatial distributions in the surveys suggest that the nearest galaxies to absorbers lie at distances @xmath14 mpc from an @xmath5 galaxy and at @xmath6 kpc from a galaxy at @xmath12 .
observers often lack direct measurements of galaxy masses .
instead , they estimate halo mass , @xmath282 , from the inferred stellar mass @xmath2 or galaxy luminosity @xmath283 ( shankar et al .
2006 ) using the technique of abundance matching " ( behroozi et al . 2010
; moster et al . 2010 , 2013 ) .
by matching the abundance of dark - matter halos in simulations to observed galaxy distributions in stellar mass and luminosity , they attempt to characterize the relationship between the stellar masses of galaxies and the masses of the dark matter halos in which they live .
moster et al .
( 2010 ) obtain a parameterized stellar - to - halo mass relation by populating halos and subhalos in an n - body simulation with galaxies and requiring that the observed stellar mass function be reproduced . in a different approach , tacchella , trenti , & carollo ( ttc13 )
assume that the rest - frame uv luminosity and stellar mass of a galaxy are related to its dark - matter halo assembly and gas infall rate .
galaxies are assumed to experience a burst of star formation at the halo assembly time ( redshift @xmath284 ) followed by a constant star formation rate sustained by steady gas accretion that provides the dominant contribution to the uv luminosity at all redshifts .
the model is calibrated by constructing a galaxy luminosity versus halo mass relation at @xmath285 via abundance matching .
after this luminosity calibration , the model naturally fits the @xmath285 stellar mass function and correctly predicts the evolution of both luminosity and stellar mass functions from @xmath286 to @xmath110 .
generically , halo - matching methods have many similarities , including the flattening at high @xmath2 of the relation that maps stellar mass @xmath2 into halo mass @xmath3 . the similar relation between luminosity @xmath283 of the central galaxy and
the dark - matter halo has been explained ( cooray & milosavljevic 2005 ) as the result of a decline in merging efficiency for the accretion of satellites .
flattening of the @xmath2 curve at high @xmath3 predicts that ( @xmath110 ) galaxies with @xmath287 reside in very massive halos , @xmath288 , although errors in @xmath2 ( or @xmath283 ) can create large uncertainties in the inferred @xmath3 .
a more subtle effect is that scatter in @xmath2 is amplified asymmetrically in @xmath3 ; simply adopting the mean relation for @xmath289 leads to errors ( behroozi et al .
various halo - matching methods differ in specific details , including the redshift evolution of the ratio @xmath290 . here , we adopt the ttc13 formalism , which has a clear physical basis for mass assembly and rest - frame uv luminosity evolution . the cos - halos study ( tumlinson et al . 2013 ; werk et al .
2013 ) used the moster et al .
( 2010 ) technique , whereas this paper adopts the ttc13 formalism .
both methods find that galaxies with @xmath291 correspond to large halo masses .
however , as seen in table 3 , there are differences between the two methods in the estimated halo masses and virial radii of the 44 galaxies in the cos - halos sample .
the inferred values of @xmath3 and @xmath0 are particularly uncertain for galaxies with @xmath291 . comparing values in columns ( 5)(8 )
, one sees that the new method gives virial radii smaller by factors 0.4 - 0.8 .
statistically , the ratio of mean virial radii computed with the two methods is @xmath292 , with the largest change occurring for lower - mass halos , @xmath293 ( all masses are expressed in @xmath294 units ) .
table 4 summarizes the statistical differences for the 44 galaxies in this sample , including subsets of galaxies at low , moderate , and high stellar masses .
this decrease in @xmath0 occurs primarily because we evaluate the galaxy collapse and virialization at overdensity , @xmath88 , at the redshifts of galaxy assembly , @xmath295 , rather than at the observed redshifts @xmath137 .
in addition , the conversion from @xmath2 to @xmath3 using modeling or abundance matching differs between methods , particularly in its dependence on redshift .
another difference is that we define @xmath88 , relative to the _ matter _
density rather than the _ critical _ density . in a few instances with @xmath291 and large stellar masses , @xmath296
, we find large virial radii , @xmath297 kpc , and anomalously large massive halos , @xmath298 .
this result occurs because halo - matching algorithms and simulated luminosity functions find large stellar masses to be rare and associate them with very massive halos .
applied to the large local group galaxies , halo - matching gives high values for the milky way ( @xmath299 and @xmath9 kpc for @xmath300 ) and m31 ( @xmath301 and @xmath302 kpc for @xmath303 ) .
unfortunately for cgm studies , halo - matching for @xmath291 ( high @xmath2 ) galaxies remains an uncertain technique , as demonstrated in figure 1 of stocke et al .
this paper reviews current observations and inferences about the spatial extent of galaxies , and the extent of their gravitational influence .
these included several dynamical measures : the gravitational radius " @xmath38 , the radius of influence " @xmath304 , and the accretion radius " @xmath305 .
clearly , these definitions are related for systems near dynamical equilibrium in the gravitational potential of the galactic halo .
for example , in section 2.1 and table 1 , we found a ratio @xmath306 for nfw halos with concentrations ranging from @xmath69 to @xmath67 .
various estimates of radial extent from rotation curves and injection of fast stars into extended halos suggest 150 - 200 kpc extents for halo masses of @xmath83 .
kinematic observations of the milky way and m31 give masses @xmath307 within 60 - 80 kpc .
the inferred galactic escape velocity , @xmath163 km s@xmath78 , together with the local circular velocity @xmath308 km s@xmath78 , is consistent with halos extending beyond 80 - 100 kpc .
somewhat larger halo masses , @xmath309 and @xmath310 ( all masses in @xmath294 units ) have been inferred for the milky way and m31 out to larger radii , using globular clusters , satellite galaxies , and stellar streams .
their mass sum , @xmath311 , is comparable to recent estimates of the timing mass of the local group , @xmath312 ( van der marel 2012 ) , although a recent study ( gonzalez et al .
2013 ) used cosmological simulations of halo pairs and a likelihood correction to find a lower value , @xmath313 .
for either of these masses , the milky way and andromeda are likely to have radial extents @xmath6 kpc .
the answer to the question posed in the title to this paper depends on how one wishes to define or observe the edge of a galaxy " .
does a galaxy end because of gravity or outflows ?
this paper suggests that the best definition of @xmath0 is in terms of gravitational binding .
dark matter halos have very large extents , but galactic outflows can drive gas out of the stellar disk , into the ( bound ) halo , and often into the ( unbound ) cgm and igm .
we have formulated a new definition of virial radius in this spirit , based on galaxy assembly in the past , when the background density was higher . as a result
, we find virial radii smaller by factors of 0.5 - 0.6 , with important consequences for assessing whether the extended absorption seen with cos is bound halo gas , or unbound gas on its way to the igm . *
a revised formulation of virial radius , @xmath314 , expresses the critical overdensity @xmath88 relative to the background matter density , @xmath315 , evaluated at @xmath96 , the redshift of galaxy assembly when half the halo mass had collapsed . here , @xmath96 depends on @xmath3 and is inferred from stellar mass @xmath2 . for dynamical consistency ,
the circumgalactic medium of large galaxies is defined by their gravitationally bound regions , estimated from the virial radius or gravitational radius , @xmath316 .
* for galaxies with halo mass @xmath317 observed by hst / cos at redshifts @xmath318 , the assembly redshifts are @xmath319 .
because virialization was largely determined by early collapse , when the background density was higher , we adopt @xmath320 and @xmath0 is reduced by a factor @xmath321 compared to @xmath110 .
* halo masses can be estimated from stellar masses by halo - matching based on a physically motivated model ( ttc13 ) for galaxy assembly consistent with rest - frame uv luminosity functions from @xmath322 .
applied to 44 cos - halos galaxies , this formalism yields virial radii smaller by factors of 0.5 - 0.6 .
galaxies with @xmath323 ( @xmath324 ) have @xmath325 kpc , consistent with their observed radial extent . *
if confined by cgm gas pressure , the wind galactopause is predicted to occur at distances @xmath326 kpc for sfr = 10 - 30 @xmath294 yr@xmath78 , mass - loading factors @xmath327 , and the observed range of cgm gas pressures and wind velocities .
stronger galaxy winds could break out into the lower density igm , with radial extents beyond 200 kpc determined by the strength , collimation , and duration of the outflow . * for dynamical masses of the large local group galaxies , we adopt : milky way ( @xmath328 ) and m31 ( @xmath329 ) with ( gravitational ) radial extents of 150 - 200 kpc . in our new formalism , the milky way stellar mass , @xmath8 , would correspond to somewhat higher halo mass @xmath299 and virial radiius @xmath9 kpc for assembly at @xmath10 .
the inferred @xmath3 for m31 is anomalously high ( @xmath301 ) , which suggests caution in using halo - matching to obtain @xmath3 and @xmath0 for massive , luminous galaxies ( @xmath291 ) . as noted earlier , the new definition of @xmath0 is important in deciding whether gas in the vicinity of galaxies is gravitationally bound or unbound .
whether galaxies are open or closed boxes " helps to determine the metal evolution of galaxies and the extent of metal - pollution of the igm .
it also suggests that astronomers should be careful in making interchangeable use of the terms cgm and halo .
matter in the galactic halo is _ gravitationally bound _
, while the circumgalactic ( gaseous ) medium may in fact be _ unbound outflows _ that will merge into the igm .
what many observers are calling the cgm is probably gas at the cgm igm interface . in observations of large metal - enriched reservoirs around star - forming galaxies ( tumlinson et al .
2013 ; stocke et al .
2013 ) , the radial offset distances @xmath46 of the qso sight lines were normalized to @xmath0 , with much of the metal - enriched gas within @xmath330 . with the revised ( smaller ) virial radii ( table 3 and figure
1 ) some of this metal - enriched gas extends beyond the region of strong gravitational influence , probably on its way out to the igm .
this trend is seen in the 28 actively star - forming galaxies in the cos - halos sample , whereas the undetected around 16 passive galaxies all had sight lines with @xmath330 . in future work
, both data sets will be examined carefully with the new formalism .
i thank crystal martin , michele trenti , massimo ricotti , john stocke , brian keeney , blair savage , and vasily belokurov for helpful discussions , and evan tilton and joshua moloney for comments on the manuscript .
this research was supported by the astrophysical theory program at the university of colorado boulder ( grant nnx07-ag77 g from nasa ) .
i am grateful to the institute of astronomy at cambridge university for their stimulating scientific environment and support through the sackler visitor program . lccc uniform sphere & @xmath331 & @xmath46 & @xmath332 + plummer model & @xmath333^{-5/2}$ ] & @xmath52 & @xmath334 + jaffe model & @xmath335 & @xmath336 & @xmath337 + hernquist model & @xmath338 & @xmath336 & @xmath339 + nfw model ( @xmath340 ) & @xmath341 & @xmath54 & @xmath342 + nfw model ( @xmath343 ) & @xmath341 & @xmath54 & @xmath344 + nfw model ( @xmath345 ) & @xmath341 & @xmath54 & @xmath346 + nfw model ( @xmath347 ) & @xmath341 & @xmath54 & @xmath348 + 9.00 & 1.709 & & & 12.25 & 1.112 + 9.25 & 1.666 & & & 12.50 & 1.070 + 9.50 & 1.620 & & & 12.75 & 1.026 + 9.75 & 1.572 & & & 13.00 & 0.979 + 10.00 & 1.531 & & & 13.25 & 0.936 + 10.25 & 1.486 & & & 13.50 & 0.891 + 10.50 & 1.439 & & & 13.75 & 0.844 + 10.75 & 1.394 & & & 14.00 & 0.806 + 11.00 & 1.346 & & & 14.25 & 0.765 + 11.25 & 1.302 & & & 14.50 & 0.722 + 11.50 & 1.255 & & & 14.75 & 0.684 + 11.75 & 1.216 & & & 15.00 & 0.645 + 12.00 & 1.155 & & lccccccrrr j0226 + 0015(268,22 ) & 0.22744 & 0.51 & 10.8 & 12.62 & 12.73 & 303 & 178 & 1.03 & 80 + j0401 - 0540 ( 67,24 ) & 0.21969 & 0.37 & 10.2 & 12.08 & 11.86 & 200 & 85 & 1.19 & 86 + j0803 + 4332(306,20 ) & 0.25347 & 1.63 & 11.3 & 13.48 & 13.84 & 581 & 462 & 0.83 & 79 + j0910 + 1014 ( 34,46 ) & 0.14274 & 0.60 & 10.6 & 12.48 & 12.39 & 279 & 133 & 1.09 & 116 + j0910 + 1014(242,34 ) & 0.26412 & 2.16 & 11.5 & 13.76 & 14.40 & 716 & 747 & 0.74 & 139 + j0914 + 2823 ( 41,27 ) & 0.24431 & 0.37 & 9.8 & 11.87 & 11.48 & 169 & 61 & 1.26 & 104 + j0925 + 4004(196,22 ) & 0.24745 & 1.48 & 11.3 & 13.45 & 13.84 & 569 & 462 & 0.83 & 86 + j0928 + 6025(110,35 ) & 0.15400 & 0.52 & 10.8 & 12.65 & 12.73 & 317 & 178 & 1.03 & 94
+ j0935 + 0204 ( 15,28 ) & 0.26228 & 0.79 & 11.0 & 12.88 & 13.13 & 365 & 251 & 0.95 & 114 + j0943 + 0531(106,34 ) & 0.22839 & 0.74 & 10.8 & 12.61 & 12.73 & 300 & 178 & 1.03 & 125 + j0943 + 0531(216,61 ) & 0.14311 & 0.72 & 11.0 & 12.80 & 13.13 & 382 & 251 & 0.95 & 154 + j0943 + 0531(227,19 ) & 0.35295 & 0.24 & 9.6 & 11.66 & 11.31 & 141 & 69 & 1.29 & 95 +
j0950 + 4831(177,27 ) & 0.21194 & 1.44 & 11.2 & 13.30 & 13.59 & 511 & 371 & 0.88 & 94 + j1009 + 0713(204,17 ) & 0.22784 & 0.28 & 9.9 & 11.90 & 11.56 & 174 & 66 & 1.24 & 62 + j1009 + 0713(170,9 ) & 0.35569 & 0.29 & 10.3 & 12.06 & 11.98 & 189 & 94 & 1.16 & 45 + j1016 + 4706(274,6 ) & 0.25195 & 0.23 & 10.2 & 12.10 & 11.86 & 202 & 85 & 1.19 & 24 + j1016 + 4706(359,16 ) & 0.16614 & 0.44 & 10.5 & 12.35 & 12.24 & 251 & 117 & 1.12 & 46 + j1112 + 3539(236,14 ) & 0.24670 & 0.52 & 10.3 & 12.17 & 11.98 & 214 & 94 & 1.16 & 54 + j1133 + 0327(110,5 ) & 0.23670 & 1.94 & 11.2 & 13.19 & 13.59 & 515 & 371 & 0.88 & 14 + j1133 + 0327(164,21 ) & 0.15449 & 0.19 & 10.1 & 12.09 & 11.76 & 206 & 78 & 1.21 & 56 + j1157 - 0022(230,7 ) & 0.16378 & 0.55 & 10.9 & 12.72 & 12.91 & 334 & 207 & 1.00 & 20 + j1220 + 3853(225,38 ) & 0.27371 & 0.66 & 10.8 & 12.53 & 12.73 & 279 & 178 & 1.03 & 159 + j1233 + 4758(94,38 ) & 0.22210 & 0.70 & 10.8 & 12.58 & 12.73 & 295 & 178 & 1.03 & 137 + j1233 - 0031(168,7 ) & 0.31850 & 0.37 & 10.6 & 12.30 & 12.39 & 230 & 133 & 1.09 & 33 + j1241 + 5721(199,6 ) & 0.20526 & 0.20 & 10.2 & 12.10 & 11.86 & 205 & 85 & 1.19 & 21
+ j1241 + 5721(208,27 ) & 0.21780 & 0.20 & 10.1 & 12.02 & 11.76 & 192 & 78 & 1.21 & 92 + j1245 + 3356(236,36 ) & 0.19248 & 0.20 & 9.9 & 11.81 & 11.56 & 178 & 66 & 1.24 & 117 + j1322 + 4645(349,11 ) & 0.21418 & 0.58 & 10.8 & 12.50 & 12.73 & 303 & 178 & 1.03 & 39 + j1330 + 2813(289,28 ) & 0.19236 & 0.24 & 10.3 & 12.22 & 11.98 & 225 & 94 & 1.16 & 91 + j1342 - 0053(157,10 ) & 0.22702 & 1.08 & 11.0 & 12.67 & 13.13 & 345 & 251 & 0.95 & 37 + j1342 - 0053(77,10 ) & 0.20127 & 0.25 & 10.5 & 12.34 & 12.24 & 247 & 117 & 1.12 & 34 + j1419 + 4207(132,30 ) & 0.17925 & 0.55 & 10.6 & 12.46 & 12.39 & 272 & 133 & 1.09 & 92 + j1435 + 3604(126,21 ) & 0.26226 & 0.33 & 10.4 & 12.20 & 12.01 & 218 & 97 & 1.15 & 86 + j1435 + 3604(68,12 ) & 0.20237 & 1.37 & 11.1 & 13.08 & 13.35 & 433 & 302 & 0.92 & 40 + j1437 + 5045(317,38 ) & 0.24600 & 0.46 & 10.2 & 12.06 & 11.86 & 196 & 85 & 1.19 & 149 + j1445 + 3428(232,33 ) & 0.21764 & 0.29 & 10.4 & 12.26 & 12.01 & 230 & 97 & 1.15 & 117 + j1514 + 3619(287,14 ) & 0.21223 & 0.18 & 9.7 & 11.81 & 11.39 & 164 & 57 & 1.28 & 49 + j1550 + 4001(197,23 ) & 0.31247 & 1.78 & 11.4 & 13.50 & 14.10 & 578 & 577 & 0.79 & 106 + j1550 + 4001(97,33 ) & 0.32179 & 0.82 & 10.9 & 12.54 & 12.91 & 311 & 207 & 1.00 & 155 + j1555 + 3628(88,11 ) & 0.18930 & 0.54 & 10.5 & 12.38 & 12.24 & 254 & 117 & 1.12 & 35 + j1617 + 0638(253,39 ) & 0.15258 & 2.65 & 11.5 & 14.03 & 14.40 & 912 & 747 & 0.74 & 104
+ j1619 + 3342(113,40 ) & 0.14137 & 0.19 & 10.1 & 12.12 & 11.76 & 211 & 78 & 1.21 & 100 + j2257 + 1340(270,40 ) & 0.17675 & 0.67 & 10.9 & 12.78 & 12.91 & 348 & 207 & 1.00 & 120 + j2345 - 0059(356,12 ) & 0.25389 & 0.80 & 10.9 & 12.63 & 12.91 & 304 & 207 & 1.00 & 48 + all galaxies & 44 & 10.61 & 12.53 & 12.55 & 315 & 200 & 0.563 + @xmath349 & 12 & 10.23 & 12.12 & 11.89 & 207 & 81 & 0.422 + @xmath350 & 10 & 10.84 & 12.62 & 12.80 & 309 & 190 & 0.613 + @xmath351 & 7 & 11.34 & 13.53 & 13.97 & 626 & 534 & 0.845 + | our current view of galaxies considers them as systems of stars and gas embedded in extended halos of dark matter , much of it formed by the infall of smaller systems at earlier times .
the true extent of a galaxy remains poorly determined , with the virial radius " ( @xmath0 ) providing a characteristic separation between collapsed structures in dynamical equilibrium and external infalling matter .
other physical estimates of the extent of gravitational influence include the gravitational radius , gas accretion radius , and galactopause " arising from outflows that stall at 100 - 200 kpc over a range of outflow parameters and confining gas pressures .
physical criteria are proposed to define bound structures , including a more realistic definition of @xmath1 for stellar mass @xmath2 and halo mass @xmath3 , half of which formed at assembly redshifts " ranging from @xmath4 .
we estimate the extent of bound gas and dark matter around @xmath5 galaxies to be @xmath6 kpc .
the new virial radii , with mean @xmath7 kpc , are 40 - 50% smaller than values estimated in recent hst / cos detections of and absorbers around galaxies . in the new formalism
, the milky way stellar mass , @xmath8 , would correspond to @xmath9 kpc for half - mass halo assembly at @xmath10 .
the frequency per unit redshift of low - redshift absorption lines in qso spectra suggests absorber sizes @xmath11 kpc when related to intervening @xmath12 galaxies .
this formalism is intended to clarify semantic differences arising from observations of extended gas in galactic halos , circumgalactic medium ( cgm ) , and filaments of the intergalactic medium ( igm ) .
astronomers should refer to _ bound gas _ in the galactic halo or cgm , and _ unbound _ gas at the cgm - igm interface , on its way into the igm . |
quantum entanglement is one of the most fascinating characteristic of quantum physics , a fantastic application of entanglement @xcite is quantum teleportation , which plays a key role in the field of quantum communication . since the seminal work of bennett et al .
@xcite , teleportation has been the research interest of researchers and a number of work both in theory and experiments has been devoted to it @xcite . up to now the teleportation has been studied in different branches , such as directly and network controlled teleportation @xcite ; discrete - variables and continuous - variables teleportation @xcite ; prefect and probabilistic teleportation @xcite and so on .
in fact , one of the key problem of teleportation is how to construct an usefulness quantum channel , different channels will yield different results , some channels can be used to realize perfect teleportation , while some others can only enable probabilistic teleportation .
because of the inevitable interaction with its surroundings , correlations in quantum states are difficult to maintain @xcite , therefore the probabilistic teleportation @xcite has been widely discussed in recent years . a necessary and sufficient condition for realizing perfect teleportation and
successful teleportation has been given in @xcite .
based on the bell basis measurement , we found that if the channel parameter matrix ( cpm ) is unitary , then one can always realize a perfect teleportation ( i.e. , the successful probability @xmath1 ) , if the cpm is invertible but not unitary , however , one can only realize a probabilistic teleportation ( i.e. , the successful probability @xmath2 ) . motivated by the idea of ref .
@xcite , in which the authors introduced an auxiliary qubit to realize probabilistic teleportation , we propose some unitary transformation methods in probabilistic teleportation in this work .
these methods are based on different entanglement matching coefficients @xmath0 which corresponding to different unitary transformations . through detailed analysis with the bell basis and generalized bell basis measurement in two probabilistic teleportation processes
, we suggest a general probability of successful teleportation , which is not only determined by the entanglement degree of transmission channel and measurement methods , but also related to the unitary transformation in the teleportation process . for example
, if we teleport the unknown one - qubit state @xmath3 via the channel state @xmath4 , then the whole probability of successful teleportation is @xmath5 , where @xmath6 .
our conclusion covers and complements the results of ref . @xcite . as different @xmath0
will give different kinds of evolution methods , one can have more flexible and selectable evolution method experimentally .
suppose alice wants to send an unknown one - qubit state @xmath3 to bob @xmath7 where @xmath8 , with @xmath9 being taken to be 0 or 1 and a repeated index denotes summation .
the general two - qubit state @xmath10 used as the quantum channel can be expressed as follows @xmath11 where @xmath12 . if alice adopts the standard bell basis measurement ( bm ) @xmath13 @xmath14 on her particles , i.e. , @xmath15 and @xmath16 .
then with the denotation of bell basis @xcite , the total state of the system can be rewritten as @xmath17 where @xmath18 is the element of @xmath19 after alice s measurement , the total state will collapse to @xmath20 obviously , based on the bm method , all the @xmath21 are unitary .
so if the cpm @xmath22 is unitary , one can always realize the perfect teleportation ( i.e. , the whole probability @xmath1 ) , if @xmath22 is invertible but not unitary , one can only realize a probabilistic teleportation ( i.e. , the whole probability @xmath2 ) . in ref .
li et al . presented a protocol of probabilistic teleportation by introducing an auxiliary qubit state @xmath23 and performing an unitary transformation on bob s state .
they employ a partially entangled state as the quantum channel , that is @xmath24 with @xmath25 being real numbers and @xmath26 .
the cpm @xmath27 is obviously invertible but not unitary , so bob can not directly retrieve the state by acting @xmath28 on the collapsed state @xmath29 . with the standard bell basis @xmath30 , the total state of the system can be rewritten as @xmath31.\end{aligned}\ ] ] after alice s bm @xmath32 , bob will get the unnormalized state as follows @xmath33 in order to obtain the original state , one can introduce an auxiliary qubit state @xmath23 @xcite in bob s state , now bob s state can be express as @xmath34 with the standard two - qubit basis @xmath35 with parameter @xmath0 for the particles ( a,3 ) @xmath36 where we called @xmath0 the entanglement matching coefficient of bob s evolution . to ensure the transformation @xmath37 to be unitary , we demand @xmath38
. there are different unitary transformation methods with different @xmath0 . after bob s evolution
, the state @xmath39 turns out to be @xmath40 certainly , @xmath41 is not normalized .
now bob performs measurement on the auxiliary qubit @xmath42 , if the measurement outcome is @xmath43 , the teleportaton fails , if the measurement outcome is @xmath44 , the teleportation is successfully accessed and bob s state becomes @xmath45 and bob s @xmath46 probability . from eq .
( 7 ) one can obtain the bell state @xmath47 occurring probability as @xmath48,\end{aligned}\ ] ] similarly , the bell state @xmath49 , @xmath50 and @xmath51 occurring probability are @xmath52 , p^3_a = p^4_a=\frac{1}{{2}}[(a\beta)^2+(b\alpha)^2],\end{aligned}\ ] ] if @xmath53 then @xmath54 , which is just the prefect teleportation .
now we compute the probability of bob for obtaining the original state from the state @xmath39 .
the normalized state corresponding to the state in eq .
( 11 ) is latexmath:[\[\begin{aligned } after bob s successful measurement on @xmath44 , the probability @xmath56 of obtaining the original state from the state @xmath39 is @xmath57 we consider alice s measurement and bob s different operations in the teleportation process .
for alice s each measurement and bob s operation , the probability of obtaining the initial state is @xmath58 where @xmath59 is just the square of coefficient of the state @xmath60 in eq .
( 12 ) . summing all the contributions of @xmath61 , we obtain the whole probability of successful teleportation as @xmath62 where @xmath38 . if @xmath63 , we take @xmath64 , then the optimal probability is @xmath65 .
when @xmath53 then @xmath66 , which is just the case for prefect teleportation , and for this special case @xmath37 is the same as that in ref .
@xcite , i.e. @xmath67 for different matching coefficients @xmath0 , one can adopt different kinds of unitary transformation .
for example , when @xmath68 the whole probability of successful teleportation is @xmath69 and our unitary transformation matrix turns out to be @xmath70 next we discuss some difference between the probability @xmath71 and @xmath72 for two kinds of unitary transformation . for arbitrary @xmath26 and @xmath73 , there are always @xmath74 .
20 ) is an unitary transformation for obtaining the optimal probability .
however , the condition @xmath26 and @xmath75 yields @xmath76 , and therefore one can only obtain probability @xmath77 . when @xmath68 we have @xmath69 , and the normalization condition @xmath26 gives rise to @xmath78 with @xmath79 , thus one can only attain the probability @xmath80 for @xmath73 . because @xmath6 , when @xmath81 ( here @xmath75 and @xmath82 ) , then @xmath83 .
for this case there are a little difference between the probability @xmath71 and @xmath84 for the two kinds of unitary transformation ( see fig .
1 ) . fig .
comparison of the probability with different channel parameters .
different matching coefficients @xmath0 correspond to different unitary transformations .
although @xmath85 , it provides one with more flexible selection for probabilistic teleportation .
considering now alice makes a generalized bell basis measurement ( gbm ) , these are @xmath86 where @xmath87 and @xmath88 the transformation matrix @xmath89 between the generalized bell basis and computation basis @xmath90 is @xmath91 let us reconsider the aforementioned one - qubit teleportation , under the generalized bell basis , the total state of the system is @xmath92 after alice s gbm @xmath93 , bob will get the corresponding unnormalized state as follows @xmath94 when bob introduced an auxiliary qubit state @xmath23 for @xmath95 , and make an unitary transformation @xmath96 to the state @xmath97 , with @xmath98 where @xmath99 , the state @xmath39 becomes @xmath100 if bob s measurement outcome is @xmath23 , the teleportation is successfully implemented . considering alice s bell basis measurement @xmath101 and bob s evolution on @xmath39 , the probability of successful teleportation is @xmath102 similarly , after bob s @xmath103 , @xmath104 and @xmath105 transformation to the corresponding states and measurement on the auxiliary qubit , the probability of successful teleportation respectively are @xmath106 thus the whole probability of successful teleportation is @xmath107 next we discuss the optimal probability of successful teleportation by entanglement matching for the following two cases : \(1 ) @xmath108 . for this case , @xmath109 , @xmath110 , so we take @xmath111 in @xmath96 and @xmath104 , @xmath112 in @xmath103 and @xmath105 , for which one can obtain @xmath113 , @xmath114 .
the optimal whole probability is @xmath115 \(2 ) @xmath116 . in this case ,
@xmath109 , @xmath117 , so we take @xmath118 , @xmath119 , for which we have @xmath113 , @xmath120 .
the optimal whole probability is @xmath121 from the above analysis , one can see that the optimal probability of successful teleportation is determined by the smaller value of @xmath122 and @xmath123 , i.e. , the optimal probability is determined by the entanglement degree of alice s measurement or the quantum channel .
however , for @xmath124 , the whole probability of successful teleportation is @xmath125 .
therefor , the general probability of successful teleportation is not only determined by the factors of the channel and measurement , but also related to the unitary transformation during teleportation process .
the unavoidable influence of environment always induces degradation of quantum correlations , therefor the study of probabilistic teleportation is significant for quantum information processing . in this paper , we generalized the protocol of probabilistic teleportation by introducing an auxiliary qubit and the unitary transformation methods . moreover , through the analysis based on the bell basis and generalized bell basis measurement in two probabilistic teleportation , we suggested a general probability of successful teleportation , which is not only determined by both the entanglement degree of transmission channels and the measurement methods , but also related to unitary transformation in teleportation process , i.e. , @xmath72 , although @xmath126 ( the optimal @xmath37 transformation ) .
however in experiment , it is more important to realize successful teleportation . as different entanglement matching coefficients @xmath0 will give different @xmath37 evolution methods , so one can have more flexible selectable evolution method experimentally .
+ * acknowledgments * + this work was supported in part by the national natural science foundation of china under grant no .
10902083 , the natural science foundation of shaanxi province under grant nos . 2009jq8006 , 2009jm6001 and 2010jm1011 .
r. horodecki , p. horodecki , m. horodecki , k. horodecki , rev .
81 ( 2009 ) 865 .
a. abliz , h. j. gao , x. c. xie , y. s. wu , w. m. liu , phys .
a 74 ( 2006 ) 052105 .
z. g. li , s. m. fei , z. d. wang , w. m. liu , phys . rev . a 79 ( 2009 ) 024303 . c. h. bennett , g. brassard , c. crpeau _ et al .
70 ( 1993 ) 1895 .
d. bouwmeester , j. w. pan , k. mattle _ et al .
_ , nature 390 ( 1997 ) 575 .
m. a. nielsen , e. knill , r. laflamme , nature 396 ( 1998 ) 52 .
h. y. dai , p. x. chen , c. z. li , j. opt .
b : quantum semiclass . opt .
6 ( 2004 ) 106 .
y. yeo , w. k. chua , phys . rev .
96 ( 2006 ) 060502 .
z. s. yuan , _ et al .
_ , nature phys 454 ( 2008 ) 1098 . m. l. hu , phys .
lett . a 375 ( 2011 ) 922 . m. l. hu , phys .
lett . a 375 ( 2011 ) 2140 .
a. karlsson , m. bourennane , phys .
a 58 ( 1998 ) 4394 .
f. g. deng , _ et al .
a 72 ( 2005 ) 022338 .
z. x. man , y. j. xia , n. b. an , phys .
a 75 ( 2007 ) 052306 .
g. f. shi , x. q. xi , x. l. tian , r. h. yue , opt .
commun . 282
( 2009 ) 2460 .
s. l. braunstein , h. j. kimble , phys .
80 ( 1998 ) 869 .
p. marian , t. a. marian , phys .
a 74 ( 2006 ) 042306 .
p. marian , t. a. marian , int .
j. quantum inf . 6
( 2008 ) 721 .
w. l. li , c. f. li , g. c. guo , phys .
a 61 ( 2000 ) 34301 .
b. s. shi , y. k. jiang , g. c. guo , phys .
lett . a 268 ( 2000 ) 161 .
m. l. hu , phys .
a 374 ( 2010 ) 3520 .
m. l. hu , h. fan , ann .
( 2012 ) 851 .
h. lu , chin .
lett . 18 ( 2001 ) 1004 .
j. fang , y. lin , s. zhu , x. chen , phys .
a 67 ( 2003 ) 014305 .
t. gao , z. x. wang , f. l. yan , chin . phys .
( 2003 ) 2094 .
f. l. yan , h. w. ding , chin .
23 , 17 ( 2006 ) .
w. x. jiang , _ et al . _
lett 24 ( 2007 ) 1144 .
x. l. tian , x. q. xi , int .
j. quantum inform . 7 ( 2009 ) 927 .
x. l. tian , _
et al . _ , opt .
commun . 282
( 2009 ) 4815 .
x. l. tian , x. q. xi , mod .
b 23 ( 2009 ) 2261 . | we proposed a general transformation in probabilistic teleportation , which is based on different entanglement matching coefficients @xmath0 corresponding to different unitary evolution which provides one with more flexible evolution method experimentally . through analysis based on the bell basis and generalized bell basis measurement for two probabilistic teleportation
, we suggested a general probability of successful teleportation , which is not only determined by the entanglement degree of transmission channels and measurement methods , but also related to the unitary transformation in the teleportation process .
epsf = 18truecm = 24truecm = -10 mm -2.0truecm pacs : 03.67.hk , 03.65.ta |
a massive star builds up onion - like layers of different chemical elements synthesized by hydrostatic nuclear burning processes during its lifetime . at the end of its evolution ,
the innermost fe core collapses into a neutron star , which triggers a core - collapse supernova ( sn ) explosion .
the detailed process of the explosion is complicated and poorly understood , but a consensus from theoretical studies suggests that the explosion should be asymmetric and turbulent , especially near the core ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* and references therein ) .
multi - dimensional numerical simulations have shown that the explosion also leads to extensive mixing and inversion among the stratified layers by hydrodynamic instabilities ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
a distinct method to explore the explosion dynamics of sne , therefore , would be to investigate the detailed chemical and kinematic properties of sn ejecta material in nearby young galactic supernova remnants ( snrs ) where the imprints of explosion remain .
cassiopeia a ( hereafter cas a ) , at the age of @xmath3 years @xcite , is one of the best studied young galactic snrs .
its sn explosion was classified as type iib based on the optical spectra of light echoes @xcite , which implies that the progenitor was probably a star of @xmath4@xmath5 that had lost a significant portion , but not all , of its h - rich envelope before the explosion .
over the past several decades , cas a has been extensively studied in almost all wavebands from radio to gamma - rays .
the complex spatial distribution of _ shocked _ sn ejecta , such as the x - ray - emitting fe - rich ejecta plumes beyond the si - rich material and o / s - rich optical knots outlying the bright ejecta shell ( e.g. , * ? ? ? * ; * ? ? ?
* ) , indicates an explosion resulting in an inversion of the chemical layers .
furthermore , the inhomogeneous distribution of _ unshocked _ sn ejecta radiating si and ti emission lines @xcite implies that the explosion was turbulent near the progenitor core . in the visible waveband ,
many bright optical knots are presented in and around cas a. they have been classified into two major groups based on their proper motions and line - of - sight velocities : ( 1 ) fast - moving knots ( fmks ) and ( 2 ) quasi - stationary flocculi ( qsfs )
. the fmks show large proper motions and high radial velocities , corresponding to expansion velocities of up to @xmath6 ( e.g. , * ? ? ?
* ; * ? ? ?
they have spectral features strongly enhanced in o and other heavy elements ( e.g. , s and ar ) that are mostly synthesized from the nuclear burning process in a deep stellar layer , while showing no detectable h or he emission lines ( e.g. , * ? ? ?
* ; * ? ? ?
based on their high expansion velocities and chemical composition , they have been regarded as dense material ejected from the disrupted layer of the progenitor after the sn explosion .
the dynamical and chemical properties of the qsfs are very different from those of the fmks .
the qsfs have considerably slower velocities ( @xmath7 ) , and so their typical expansion age is @xmath8 years @xcite , which is much larger than the age of the remnant .
they are bright in [ ] and h@xmath9 emission , with a handful of other h and he emission lines , in optical spectra ( e.g. , * ? ? ?
* ; * ? ? ?
considering these properties , qsfs are believed to be dense circumstellar medium ( csm ) blown out from the progenitor prior to the sn explosion .
in addition , some intermediate optical knots , that is , the so called fast - moving flocculi ( fmfs ) or nitrogen knots ( nks ) , were reported by @xcite . while their spectra , which show strong [ ] 6548 , 6583 accompanied by weak h@xmath9 without any lines of o and s , are analogous to those of qsfs @xcite , their proper motions are larger than @xmath10 , corresponding to @xmath11 @xcite .
most of them have been found outside of the bright main ejecta shell .
therefore , these outlying n - rich knots have been interpreted as being fragments of the progenitor s photosphere expelled by the sn blast wave at the time of explosion .
the dynamical and chemical properties of these different types of optical knots have provided important clues to unveiling the snr s origin and evolution .
for example , the expansion center and age were determined from the proper motions of the fmks ( e.g. , * ? ? ?
* ; * ? ? ?
* and references therein ) , and the three - dimensional ( 3d ) structure of the ejecta knots reconstructed from spectral mapping observations has shown that the sn ejecta is expanding spherically but is systematically receding at a speed of @xmath12 at a distance of 3.4 kpc @xcite .
the dense , slow - moving qsfs indicate that the progenitor had undergone significant and inhomogeneous mass loss during the red supergiant phase @xcite .
although this velocity - based classification is an efficient way to classify the knots as sn ejecta and csm , which have distinctive expansion velocities , i.e. , a few @xmath13 vs. a few @xmath14 , it encounters limitations when characterizing the sn ejecta material from the different nucleosynthetic layers .
according to previous numerical simulations for core - collapse sn explosions , the radial velocity profiles of heavy elements in sn ejecta are almost identical @xcite .
several models in @xcite predict the velocity separation of up to @xmath15 between different heavy elements .
their velocity profiles , however , are very broad ( a few @xmath13 ) , so that their distributions largely overlap . these numerical simulations may imply that sn material from different nucleosynthetic layers is barely distinguishable in velocity space . in order to comprehend the explosion dynamics , therefore , a more systematic classification of sn ejecta based on their chemical composition is needed . in this paper , we report the results of broad near - infrared ( nir ) spectroscopic observations toward the main ejecta shell of cas a , focusing on classification of the emission knots based on their _ spectrochemical _ properties . the nir study of cas a has been relatively limited in the literature , although there are many bright forbidden lines of various elements in the nir waveband , some of which may arise from the deep nucleosynthetic layers . as far as we are aware , the only nir spectroscopic study covering the entire _ jhk _ bandpass was conducted by @xcite , who obtained low - resolution ( @xmath16 ) spectra of five fmks and three qsfs that were previously known .
they showed that the spectra of the fmks are dominated by [ ] 1.029 , 1.032 , 1.034 , 1.037 ( hereafter [ ] 1.03 multiplets ) and [ ] 1.188 , as well as two high - ionization si lines , [ ] 1.963 and [ ] 1.430 , while those of the qsfs show strong he i 1.083 accompanied by h emission lines .
a dozen bright [ ] lines are also detected in both fmks and qsfs .
our spectra confirm these features . here
we present the spectrochemical classification of knots and discuss their characteristics .
the organization of the paper is as follows . in section [ sec - obs ]
, we outline our spectroscopic observations and data reduction procedures .
an explanation of how we identified individual knots from two - dimensional ( 2d ) dispersed images and derived their spectral properties is given in section [ sec - ide ] . in sections [ sec - pca ] and [ sec - dis ] ,
we carry out a classification of the knots using principal component analysis ( pca ) and discuss the origin of knots in different classes .
finally , the paper is summarized in section [ sec - sum ] .
we carried out nir spectroscopic observations of cas a using triplespec mounted on the palomar 5 m hale telescope .
triplespec is a cross - dispersed nir spectrograph that provides simultaneous wavelength coverage from 0.94 to 2.46 at a spectral resolving power of @xmath17 @xcite .
the spectrograph uses two adjacent quadrants , i.e. , @xmath18 pixels , of a rockwell scientific hawaii - ii array .
the slit width and length are 1 and 30 , respectively . on 2008
june 29 and august 8 , we obtained spectra at eight slit positions along the main ejecta shell ( figure [ fig - slit ] ) .
the orientations of slits 1 and 4 were set perpendicular to the shell , while those of the others were largely parallel to the shell .
the nearby a0v star hd223386 was observed as a spectroscopic standard right before and/or after the target observations at similar airmasses .
in addition to the target spectra , we also obtained spectra of sky background by dithering along the slit or taking spectra of nearby sky , depending on the complexity of the target fields .
the total on - source exposure time at each slit position ranges from 300 to 1800 s. the detailed parameters of the spectroscopic observations are provided in table [ tbl - log ] .
we developed an idl - based data reduction pipeline to reduce the obtained triplespec data .
the pipeline first performed subtraction of a dark frame and flat fielding , followed by subtraction of sky background emission .
for the latter , we used sky background emission obtained in a dithered frame or in a frame from nearby empty sky , depending on the complexity of the source emission in a given slit . because of the non - uniform dispersion by the three cross - dispersing prisms in triplespec @xcite , some orders of the triplespec spectra are severely curved . by carefully tracing the spectra of standard stars and airglow emission lines
, we obtained a 2d wavelength solution for each order that can correct this effect in a satisfactory manner .
we conducted fifth - order order polynomial fits to the wavelengths of the oh airglow emission lines @xcite in the triplespec spectra to obtain the wavelength solutions at 0.5 @xmath19 uncertainty for each order .
heliocentric velocity correction was also performed when calculating the velocity of the emission line detected in the triplespec spectra .
we used an a0v - type standard star ( hd223386 ) in the photometric calibration and confirmed that the fluxes of the [ ] 1.644 emission line are consistent with those from the narrow - band imaging observations of the same areas ( see section [ sec - obs - img ] ) .
in 2005 august 28 and 2008 august 11 , we performed nir imaging observations for the remnant using the wide - field infrared camera ( wirc ; * ? ? ?
* ) attached to the palomar 5 m telescope .
the camera consists of a single @xmath20 rockwell hawaii - ii hgcdte nir detector with a pixel scale of @xmath21 pixel@xmath22 , which provides a field of view of @xmath23 .
we used an [ ] narrow - band filter that has a mean wavelength of 1.644 and a bandwidth of 252 .
while the single exposure time per frame in 2005 and 2008 is 60 s and 200 s , the multiple dithering observations yield total integration times of 1200 s and 5400 s , respectively .
we also obtained h - continuum narrow - band images ( mean wavelength of 1.570 and bandwidth of 236 ) in order to subtract the bright stars in the [ ] narrow - band images .
the average seeing throughout our observations was @xmath24 fwhm .
we followed standard procedures for the reduction of nir imaging data .
first , the dark and sky background were subtracted from each individual dithered frame .
then , all of the frames were divided by the normalized flat image .
the astrometric solution was derived on the basis of unsaturated stars around the remnant in the two micron all sky survey ( 2mass ) point - source catalog ( psc ; * ? ? ?
we coadded all dithered frames which were astrometrically aligned . in terms of photometric calibration ,
h_-band magnitude in the 2mass system was used by assuming that the [ ] narrow - band magnitude is the same as the _ h_-band magnitude .
the uncertainty in the zero - point magnitude we derived is less than 0.1 mag , corresponding to the 10% of the flux .
figure [ fig-2dspec ] shows 2d dispersed images of strong emission lines detected in our triplespec spectra : red for [ ] 1.644 , green for he i 1.083 , and blue for [ ] 1.188 + [ ] 0.953 .
the emission features are complex , often with multiple peaks , and so the identification of individual ` knots ' by visual inspection is not straightforward .
we used the idl routine clumpfind @xcite , which was developed for the identification of clumps in molecular clouds .
the routine identifies ` clumps ' by searching for local maxima above some intensity threshold and following them down them to low - intensity levels . for a given
2d dispersed image , a `` mask '' locating individual knots was generated from [ ] 1.644 emission features or , if they are weak , [ ] 0.953 or [ ] 1.188 emission features , and shifted along the wavelength to find other emission lines associated with the knots .
the details of this knot identification process are given in the supplement material of @xcite . in total
, we identified 63 knots of distinctive kinematical and spectral properties in the 2d dispersed images .
figure [ fig - clumps ] shows their locations and table [ tbl - knot ] lists their positions , sizes , and radial velocities .
we extracted one - dimensional ( 1d ) spectra of individual knots using their mask files ( figure [ fig - clumps ] ) .
we identified 46 emission lines in total and derived their parameters by performing single gaussian fits to the detected lines .
table [ tbl - flx ] lists the derived line widths and fluxes .
[ ] 1.644 line is detected in all 63 knots and their fluxes are also listed in table [ tbl - knot ] . among the 46 emission lines ,
43 lines are previously detected lines in snrs @xcite , whereas three lines of [ ] at 2.145 , 2.218 , and 2.242 are detected for the first time in snrs .
the [ ] lines originate from transitions between levels in @xmath25 g and @xmath25h terms with high excitation energies ( @xmath26 k ) and have been detected in a few objects , such as sgra@xmath27/irs16 complex ( * ? ? ?
* and references therein ) , classical novae @xcite , h ii regions @xcite , and planetary nebulae @xcite . among the previously reported lines , on the other hand , the o i 1.1286 , 1.1287 , [ ] 1.1881 , and [ ]
1.430 lines are not detected in our spectra .
@xcite reported detection of the o i lines in the fmks in cas a , but we were unable to confirm the detection with our spectra .
they also reported detection of the highly ionized [ ] 1.430 line , but again we were unable to confirm the detection .
the [ ] 1.1881 line was included in the list of identified lines in three qsfs in cas a and the kepler snr by @xcite , but we consider that this was a misidentification of the [ ] 1.1883 line . the expected flux of the [ ] 1.1881 line in typical conditions ( e.g. , @xmath28 k , @xmath29
@xmath30 ) in snrs is almost negligible , i.e. , its flux relative to the [ ] 1.257 line is @xmath31 , whereas the [ ] 1.1883 line could be as strong as @xmath32 of the [ ] 1.257 line for cosmic abundance , or even higher if fe atoms are locked into dust grains @xcite .
in order to systematically characterize the spectral properties of the 63 knots , which have 46 emission lines in total , we applied the pca method .
pca measures the variances among the original input variables ( i.e. , brightness of the emission lines in this study ) and then sets new orthogonal axes called principal components ( pcs ) along the largest variances . therefore , the largest variance ( or information ) is contained in the first pc ( pc1 ) , the second most in pc2 , and so forth . if there are significant correlations among the original input variables , then the majority of the information is confined within the first few pcs , which makes it possible to categorize the objects into a few groups based on the first few pcs . before performing the pca
, we apply an extinction correction using the line ratio of the [ ] 1.257 and 1.644 emission .
these two lines originate from the same upper level ( @xmath33 ) and therefore their intrinsic flux ratio ( @xmath34_{\rm int}$ ] ) depends on the einstein a - coefficients ( @xmath35 ) and their wavelengths , i.e. , @xmath34_{\rm int}=(a_{ki,1.257}/1.257)/(a_{ki,1.644}/1.644)$ ] .
the a - coefficients , however , are considerably uncertain so the line ratio ranges from 0.98 to 1.49 ( * ? ? ?
* ; * ? ? ?
* and references therein ) in literature .
we adopt a line ratio of 1.36 , which is the value suggested by @xcite and @xcite .
( we found that the uncertainty of the theoretical line ratio does not affect our classification of the knots because they are well grouped in pc spaces , as we will show in section [ sec - pca - result - pc ] .
the criteria of the groups , however , may change depending on the intrinsic line ratio we adopt . in section [ sec - pca - result - pc ]
, we will describe this in more detail . ) then , by comparing the observed line flux ratio to the intrinsic ratio , we obtain the extinction of the knots and deredden the observed fluxes of all of the lines .
we use the general interstellar extinction curve derived from a carbonaceous - silicate grain model with a milky way size distribution for @xmath36 @xcite .
the derived extinctions are listed in table [ tbl - knot ] . in our previous work @xcite
, we showed that the extinction toward the west is systematically larger than that toward the east , which is consistent with the previous optical / x - ray extinction estimates ( see figure 1 in * ? ? ?
we further showed that the extinctions of red - shifted knots are systematically higher than those of the blue - shifted knots , implying the presence of a large amount of sn dust inside and around the main ejecta shell ( see * ? ? ? * for details ) .
the lines from the same upper level in the dereddened spectral data do not provide independent information any more , so that the number of attributes in the pca are now reduced from 46 to 23 . in order to prevent a few bright lines dominating the pca ,
the line intensities are standardized by subtracting the mean and dividing by the standard deviation .
since we use the mean - subtracted data , the zero pcs represent the location of mean brightness , not the location of the zero fluxes of the lines ( hereafter convergent point ) , which is important for our classification ( see section [ sec - pca - result - pc ] ) . in order to get the pc coefficients of a knot indicating the convergent point
when all the emission lines of a knot get close to zero flux , we add one artificial knot into the data set that has emission lines with zero flux .
table [ tbl - pca ] contains the relative and cumulative fraction of variances contributed by the 10 most significant pcs . the first three pcs account for the majority , i.e. , @xmath3785% , of the spectral information ; thus , we use them in our classification of the knots .
figure [ fig - atr ] shows the projection of the 23 attributes on the combination plane of the three most significant pcs .
( this type of plots is known as an _ h - plot _ ; see @xcite and @xcite . ) while figure [ fig - atr](a ) and [ fig - atr](b ) show 2d projections on the planes of ( pc1 vs. pc2 ) and ( pc1 vs. pc3 ) , figure [ fig - atr](c ) shows 3d projections for the three pcs together .
the lengths of the vectors in the plots are proportional to the fractional contributions by the spectral line to the given pc , and their quadratic sum is equal to unity . in figure [ fig - atr](d ) , we visualize the 3d projections using the coordinate of ( longitude vs. latitude ) representing the two projection angles on the surface of a sphere .
we see in figure [ fig - atr ] that the attributes can be largely divided into three groups , each of which is composed of several strongly correlated spectral lines .
( note that the cosine of the angle between the vectors on the plots measures the linear correlation between the emission lines ; see @xcite for example . ) the first group ( hereafter ` he group ' ) is composed of h i , he i , and [ ] lines .
these lines are almost perfectly correlated with each other .
the lengths of their vectors are close to unity in the pc1-pc3 plane , which means that these spectral lines are properly accounted for by pc1 and pc3 . the second group ( ` s group ' ) is composed of [ ] , [ ] , and ionized s emission lines .
these lines are also strongly correlated with each other and mostly contributed by pc2 and pc3 in the direction orthogonal to the he group lines . the last group ( ` fe group ' ) is composed of ionized fe emission lines , i.e. , [ ] and [ ] lines .
these lines are rather loosely correlated but are still generally well separated from the lines in the other two groups .
the [ ] 2.046 and 2.224 lines in particular appear somewhat distinct from the other [ ] lines
. this might be caused by the higher excitation energies of the two lines than those of the other lines , i.e. , @xmath38 k vs. @xmath39 k. we plot the pc coefficients of the 63 knots on the pc planes in figure [ fig - obj ] .
the central positions of the planes , where ( pc1 , pc2 , pc3 ) = ( 0 , 0 , 0 ) , represent the spectrum made by averaging all of the spectra of all 63 knots . in the lower panels of the figure , which are the enlarged plots of the central areas of the pc planes ,
we draw dashed lines in order to group the knots ( see below ) originating from ( pc1 , pc2 , pc3 ) = ( -0.10 , 0.23 , 0.10 ) .
this convergent point is the location of the virtual knot with zero flux ( section [ sec - pca - method ] ) , and the radial distance from the convergent point is proportional to the brightness of the emission line .
the distributions of the pc coefficients in figure [ fig - obj ] are very similar to those in figure [ fig - atr ] .
there appear to be three groups of knots in figure [ fig - obj ] that have pc coefficients similar to those of the three groups in figure [ fig - atr ] , i.e. , the he , s , and fe groups .
we therefore group the knots in figure [ fig - obj ] as he - rich , s - rich , and fe - rich knots using the dashed lines . as a result
, we identify 7 he - rich knots , 45 s - rich knots , and 11 fe - rich knots .
figure [ fig-1dspec ] shows sample 1d spectra of the three groups .
as expected , the he - rich knots have strong lines of he i 1.083 compared to [ ] and some of them also show several emission lines of h i , [ ] , [ ] , and [ ] as well . the h i and [ ] lines are detected only in the he - rich knots .
the s - rich knots also have bright [ ] lines , but they have much stronger lines of [ ] 0.953 , [ ] 1.03 multiplets , [ ] 1.188 , and [ ] 1.963 .
the low - ionized emission lines of refractory elements , i.e. , [ ] and [ ] , are also detected in several s - rich knots .
although some s - rich knots show the he i 1.083 line , their intensities are significantly less than those of the he - rich knots .
the fe - rich knots have strong lines of [ ] and some of them have a weak he i 1.083 line as well . the brightest fe - rich knot ,
knot 10 in slit 5 , also emits [ ] lines in the _ k _ band . as in the he - rich knots ,
no lines of c , si , and s are detected in the fe - rich knots .
the three knot groups are easily distinguishable when the flux ratios of [ ] 1.03 multiplets , he i 1.083 , and [ ] 1.188 are compared ( figure [ fig - fluxratio ] ) .
the he - rich knots are well separated from the other groups in f(he i-1.083)/f([]-1.644 ) ,
i.e. , the ratio is greater than two for the he - rich knots but lower than two for the s - rich and fe - rich knots .
the f([]-1.03)/f([]-1.644 ) and f([]-1.188)/f([]-1.644 ) of the he - rich knots are mostly smaller than a few times 1.0 and 0.1 , respectively .
although the s - rich and fe - rich knots are not clearly distinguished in the f([]-1.03)/f([]-1.644 ) and f([]-1.188)/f([]-1.644 ) comparisons , we see that these ratios are higher in the s - rich knots than the fe - rich knots , i.e. , f([]-1.03)/f([]-1.644 ) @xmath40 and f([]-1.188)/f([]-1.644 ) @xmath41 for the s - rich knots and vice versa .
it is worth noting that the flux ratios of the knots in figure [ fig - fluxratio ] and the criteria mentioned above are based on the assumption that the intrinsic flux ratio of [ ] 1.257 and 1.644 is 1.36 ( section [ sec - pca - method ] ) .
if we adopt different line ratios , e.g. , 0.98 to 1.49 ( * ? ? ?
* ; * ? ? ?
* and references therein ) , then the criteria of those three flux ratios will be f(he i-1.083)/f([]-1.644 ) = 1.02.4 , f([]-1.03)/f([]-1.644 ) = 2.36.0 , and f([]-1.188)/f([]-1.644 ) = 0.190.34 .
figure [ fig - hist ] compares the distributions of the knot sizes , radial velocities , and line widths of the three knot groups .
the angular sizes of most of the knots are in the range 27(or 0.030.1 pc at a distance of 3.4 kpc ) and there is no significant difference among the three groups in their sizes , although some of the s - rich and fe - rich knots are as large as 10 . on the other hand , there are significant differences in their radial velocities and line widths .
the radial speeds of he - rich knots are @xmath42 , while those of s - rich knots range from @xmath43 to @xmath44 with a median of @xmath45 .
the radial velocities of fe - rich knots range from @xmath46 to @xmath47 with a median of @xmath48 . in line width , the he - rich knots have widths of 510 , while the s- and fe - rich knots have widths of 1035 .
( note that our spectral resolution at 1.64 is @xmath49 . )
figure [ fig - hist ] also compares the distribution of the [ ] 1.644 line fluxes among the three knot groups .
while the s - rich and fe - rich knots have a similar distribution with an increased number of knots that have faint [ ] emission , the pattern is absent in the he - rich knots .
there is no apparent correlation among the four physical parameters of the knots .
one of the physical parameters of the knots that can be straightforwardly obtained is electron density using [ ] lines originating from levels with similar excitation energies , because their ratios are mainly determined by electron densities ( e.g. , * ? ? ?
[ ] 1.644 and 1.677 are such lines , and figure [ fig - ne ] compares their expected ratios as a function of density for the assumed temperatures of 5000 , 10,000 , and 20,000 k ( left panel ) with observed values as a function of the [ ] 1.644 flux ( right panel ) .
we see that the ratio is quite insensitive to temperature and can be used to estimate electron density in the range @xmath50@xmath51 .
the electron density of the he - rich knots is a few @xmath52 , while the s - rich knots show electron densities over a broad range of @xmath53 to @xmath51 .
the fe - rich knots have somewhat lower ( @xmath54@xmath52 ) electron densities compared to the other two groups .
there appears to be no correlation between the electron densities and [ ] 1.644 line fluxes .
in this section , we discuss the origin of the knots using their spectral characteristics described in the previous section .
the he - rich and s - rich knots have quite distinct spectral properties ; the he - rich knots have high f(he i-1.083)/f([]-1.644 ) and low f([]-1.03)/f([]-1.644 ) , f([]-1.188)/f([]-1.644 ) , while the s - rich knots have low f(he i-1.083)/f([]-1.644 ) and high f([]-1.03)/f([]-1.644 ) , f([]-1.188)/f([]-1.644 ) .
their kinematic properties are also quite different ; the he - rich knots have low ( @xmath42 ) line - of - sight speeds , while the s - rich knots have high ( up to @xmath55 ) line - of - sight speeds .
these spectral and kinematical properties suggest that the he - rich knots are dense , slow - moving csm swept up by the sn blast wave , while the s - rich knots are fast - moving sn ejecta that have been shocked .
the same conclusion was reached by @xcite , who performed an abundance analysis using [ ] 1.188 and [ ] 1.257 lines .
they showed that the relative abundance of p ( a major product of the stellar ne - burning layer ) to fe ( in number ) for the he - rich knots is similar to the solar abundance , whereas that of the s - rich knots is 10100 times higher than the solar abundance .
the characteristics of the two types of knots match well with those of qsfs and fmks known from previous optical studies ( see section [ sec - int ] for a summary of their properties ) . similar to the optical qsfs ,
the he - rich knots have bright he i lines together with h i and [ ] lines ; all the he - rich knots have a he i 1.083 line , 6 out of the 7 show h i pa@xmath56 , br@xmath57 lines , and the three brightest ones also have [ ] 1.040 , 1.041 lines .
like the optical fmks dominated by lines of ionized heavy elements o , s , ar , the s - rich knots show strong s lines plus [ ] and [ ] lines ; all the s - rich knots have [ ] 1.03 multiplets , and 39 out of 45 s - rich knots also have [ ] 1.188 and/or [ ] 1.963 lines .
indeed , the overall spectra of he - rich and s - rich knots are similar to the nir spectra of qsfs and fmks obtained by @xcite , respectively .
furthermore , the radial velocities of the two nir knot groups are well consistent with those of the optical groups . for example , the radial velocity of the he - rich knots is @xmath58 , while the generally accepted radial velocity of qsfs is @xmath59 @xcite .
in addition , the median line - of - sight velocity of s - rich knots is @xmath60 , while the systematic velocity of the fmks is @xmath61 @xcite .
rich emission lines of si , p , and s in the s - rich knots and fmks imply that they are the sn ejecta originated from the ne- and o - burning layers .
the two bright emission lines , [ ] 1.188 and [ ] 1.644 , have comparable excitation energies and critical densities , and so their line ratios are strongly dependent on their abundance ratio , @xmath62(p / fe ) @xcite . as seen in figure [ fig - fluxratio ]
, there is a large scatter in this line ratio for s - rich knots , which implies that the abundance ratio @xmath62(p / fe ) varies almost two orders of magnitude for these knots .
we also found that 13 out of 45 s - rich knots have clear but relatively weak emission lines of he i and/or [ ] .
the detection of the he , c , and fe lines in the s - rich ejecta , which are either lighter or heavier elements than the o - burning materials , might infer microscopic mixing during the sn explosion . in many s - rich knots , a highly ionized si line , [ ] 1.963 , is detected , while in a few s - rich knots , a [ ] 1.645 line is also detected .
the detection of si in very different ionization stages indicates the broad range of temperatures in the s - rich knots . in section [ sec - pca - result ] , we found that the fe - rich knots exhibit intermediate characteristics between he - rich and s - rich knots ; they emit strong [ ] lines without any si , p , and s lines , but have high line - of - sight speeds of up to @xmath63
a few knots also emit an he i 1.083 line .
the high velocities , however , suggest that they are not dense qsfs represented by the he - rich knots .
their line widths are also considerably broader than those of he - rich knots , i.e. , 1035 vs. 510 ( figure [ fig - hist ] ) . on the other hand , the missing si , p , and s lines
indicate that the abundances of these ne- and o - burning elements are very low in these fe - rich knots .
we can consider two possible explanations regarding the origin of the fe - rich knots : ( 1 ) swept - up csm around contact discontinuity ( cd ) or ( 2 ) shocked sn ejecta enriched with fe elements that had been synthesized in explosive si burning .
the ambient medium that the cas a sn blast wave is propagating into is believed to be csm with an @xmath64 density distribution ( e.g. , * ? ? ? * ) .
in such a case , 1d similarity solutions show that the shocked csm accumulates at the cd with infinite density asymptotically @xcite .
we thus expect `` dense '' csm expanding at a speed comparable to the shocked sn material . in the real situation ,
however , this interacting region between the shocked sn ejecta and the shocked csm is hydrodynamically unstable and distorted , with the density of the shocked csm limited to @xmath65 times the density at the ambient shock @xcite . the temperature of the shocked csm near the cd , therefore , may be lower than the typical temperature ( @xmath66 kev ) of the shocked csm @xcite , but probably no more than a factor of 10 , and all of the fe in the shocked csm will be in high ionization stages . furthermore , h i lines are not detected in all fe - rich knots with an upper limit of f(h i - pa@xmath56)/f([]-1.257 ) @xmath67 . note that the observed ratio of these line intensities ranges between 0.05 and 10 for snrs while it is @xmath68 for orion , which should be the typical ratios for shocked and photoionized gases of cosmic abundance , respectively @xcite .
therefore , h must be depleted in fe - rich knots .
the non - detection of he and n lines might also indicate that the abundances of these ` circumstellar ' elements are very low in fe - rich knots , although this needs to be confirmed from other waveband observations .
( see the next paragraph for an explanation of the faint he lines detected in some fe - rich knots . )
we may therefore conclude that the fe - rich knots are not likely the shocked csm .
this leaves the second possibility , i.e. , the fe - rich knots are fe - enriched sn ejecta .
the high velocities and large velocity widths are consistent with sn ejecta being swept up by the reverse shock .
the low [ ] and [ ] fluxes compared to the [ ] flux , however , implies that the abundances of p and s , which are ne- and o - burning materials , are very low , which is in sharp contrast to the s - rich knots .
these characteristics strongly suggest that _ the fe - rich knots are most likely `` pure '' fe ejecta synthesized in the deepest stellar interior . _
the weak he i 1.083 line detected in some fe - rich knots could be due to an @xmath9-rich freeze - out process during the explosive si burning ; just after the explosion , complete si burning with an @xmath9-rich freeze - out occurs under high temperatures and low density conditions in the stellar deep layer , and many alpha particles are `` frozen out '' without participating in further nucleosynthetic processes @xcite .
similar dense , fe - predominant ejecta , likely from the @xmath9-rich freeze - out process , have been detected in another young core - collapse snr g11.2 - 0.3 @xcite .
pure fe ejecta have been detected in x - rays ( see below ) but not in optical or nir wavebands .
this is surprising considering the extensive optical / nir studies carried out for cas a since its discovery .
figure [ fig - knotpos ] partly gives an answer . in the right panel of figure [ fig - knotpos ]
, red is an [ ] 1.644 narrow - band image while green and blue are _ hubble space telescope _
( _ hst _ ) acs / wfc f850lp and f775w images which are dominated by ionized s and o lines , respectively @xcite .
previous optical observations had been mostly toward the northern ejecta shell bright in ionized o , s , and ar lines ( e.g. , * ? ? ? * ; * ? ? ?
* ; * ? ? ?
* ) or toward fmks outside the main shell ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
figure [ fig - knotpos ] , however , shows that the southwestern ( sw ) main ejecta shell , which is bright in the [ ] line but faint in the ionized o and s lines , is the region where fe - rich ejecta can be found .
indeed , our result shown in the left panel of figure [ fig - knotpos ] confirms this ; 9 out of 11 fe - rich knots are located in slits 46 .
( note that the compact red knots in the interior and beyond the sw shell are mostly qsfs , and they are bright in [ ] 6548 , 6583 and h@xmath9 images too @xcite . )
the slit positions are determined from an [ ] 1.644 image , so that some of them were placed toward the red portions of the main ejecta shell and , by decomposing the emission into individual velocity components , we were able to identify fe - rich ejecta .
it is worth noting that @xcite also noted the bright [ ] emission in the sw shell in their [ ] 1.644 image of cas a. meanwhile , we can see that the [ ] 17.9 emission is much brighter than the emission from o - burning elements such as ar and s in the sw shell in the _ spitzer _ mid - infrared maps of ionic lines @xcite .
our result suggests that this [ ] emission - predominant area , i.e. , the red area of the sw ejecta shell in figure [ fig - knotpos ] , is probably mainly composed of fe ejecta .
it is not easy to identify the counterpart of fe - rich knots in optical images because several velocity components are usually superposed along the line of sight toward the main ejecta shell .
the brightest fe - rich knot ( knot 10 in slit 5 ; hereafter k10 ) , however , is somewhat isolated and we can identify its counterpart . in figure [ fig - k10 ] , the upper two images are [ ] 1.644 images at different epochs and they show that k10 is a clump of @xmath69 elongated along the slit .
the two [ ] images clearly show that the clump is moving fast tangentially . the proper motion is measured @xmath70 yr@xmath22 , implying a tangential velocity of @xmath71 . in the lower f775w and f850lp images
, we see diffuse faint emission at the position of k10 ( see the red contour ) .
its brightness distribution is different , with two small ( @xmath72 ) bright spots ( s1 and s2 in the figure ) in the lower part of the clump .
one of these bright spots , s1 , is coincident with an s - rich knot ( knot 9 in slit 5 ) , which is spatially coincident with k10 but has a line - of - sight velocity ( @xmath47 ) that is very different from that of k10 ( @xmath73 ) .
the other bright spot , s2 , must also be due to an s - rich knot not detected in our spectroscopy .
so excluding these two bright knots , k10 appears faint in f775w and f850lp images .
we suspect that most of the diffuse emission in the f775w and f850lp images is due to optical [ ] lines .
this can be confirmed by optical spectroscopy .
recently , there have been optical spectral mapping observations of cas a @xcite and , in principle , a similar analysis can be done to detect fe - rich knots , although the optical [ ] lines , e.g. , [ ] 7155 and 8617 lines , will be much fainter than the [ ] 1.644 line because of large interstellar extinction ( @xmath7410 mag ) toward cas a ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . the distribution and amount of fe ejecta in cas a have been a subject of controversy .
previous x - ray studies detected hot and diffuse `` pure '' fe ejecta with mass @xmath75 that might have formed by @xmath9-rich freeze - out during the complete si burning @xcite .
these x - ray - emitting shocked fe ejecta are distributed mainly in the southeastern and northern regions of the remnant ( figure [ fig - fedist ] ) . on the other hand ,
the hard x - ray emission from the radioactive decay of @xmath2ti has been detected in the interior of the main ejecta shell ( figure [ fig - fedist ] ; * ? ? ?
* ) . since @xmath2ti is essentially synthesized in complete si burning with @xmath9-rich freeze - out in the innermost region ( e.g. , * ? ? ?
* ) , @xmath2ti traces `` pure '' @xmath76ni or its stable nuclei @xmath76fe . the majority of the observed @xmath2ti is inside the reverse shock and therefore from unshocked fe ejecta . the inferred mass of the unshocked fe ejecta is @xmath75 @xcite .
such unshocked fe ejecta , however , have not yet been detected , presumably because they are cool ( @xmath77 k ; e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . instead
, the x - ray - emitting fe ejecta located just outside the @xmath2ti emission was attributed to the corresponding shocked ejecta ( figure [ fig - fedist ] ; * ? ? ?
however , the missing x - ray - emitting fe ejecta toward the south and northeast directions from the explosion center have been puzzling .
we note that the nir [ ] -bright , red portion of the sw shell appears to be in contact with the interior @xmath2ti - emitting region ( figure [ fig - fedist ] )
. this might be the case for the small red portion near slit 2 in the northeastern shell , too .
therefore , if these [ ] -bright regions are composed of shocked , dense fe ejecta , as implied from our spectroscopic result , it explains why we do not see x - ray - emitting diffuse fe ejecta toward these directions ; the unshocked `` pure '' fe ejecta traced by the radioactive @xmath2ti emission in the interior of the main shell is composed of both dense and diffuse ejecta , and when these are swept up by a reverse shock , we observe either nir [ ] emission or x - ray emission depending on their densities .
we do not see nir emission associated with the central bright @xmath2ti emission but this could be because the shock is face - on , as suggested by the large red - shifted central velocity ( 11003000 @xmath78 ) of the @xmath2ti line @xcite , or because the reverse shock has not yet reached the dense , unshocked fe ejecta .
recent multi - dimensional simulations show that such global asymmetry in fe ( or @xmath76ni ) density can arise from the low - mode convection of the innermost region just after the core bounce ( e.g. , * ? ? ?
future nir spectral mapping observations revealing the 3d distribution of the dense fe ejecta will be helpful for understanding the sn explosion dynamics of the innermost region .
we have carried out nir spectroscopic observations toward the main ejecta shell of the young snr cas a. in total , 63 individual knots were identified from eight slit positions by using a clump - finding algorithm .
each of these knots has distinct kinematical and spectral properties . within the _ jhk _ spectral range ( 0.942.46 )
, we found 46 emission line features including a dozen bright [ ] lines , forbidden lines of other metallic species , and h and he lines .
we employed the pca method to classify the knots based on their relative line fluxes into three distinctive groups : he - rich knots of pre - supernova circumstellar wind material , plus s - rich and fe - rich knots of sn ejecta material .
the he - rich and s - rich knots correspond to qsfs and fmks studied in the visible waveband , while fe - rich knots , showing in general only [ ] emission lines , are likely ` pure ' dense fe ejecta from the innermost layer of the progenitor .
we summarize our main results as follows .
the pca showed that the nir spectral lines can be grouped into three groups : ( 1 ) group 1 , composed of h i and he i lines together with [ ] lines , ( 2 ) group 2 , composed of forbidden lines of si , p , and s , and ( 3 ) group 3 , composed of forbidden fe lines .
the lines in the first two pcs are strongly correlated with each other , while the correlation is rather weak among the forbidden fe lines in group 3 .
these three spectral groups of the emission lines are almost independent in 3d pc space ( figure [ fig - atr ] ) .
the distribution of the knots in the pc planes matches well with the above spectral groups , and we classified the knots into three groups : ( 1 ) he - rich , ( 2 ) s - rich , and ( 3 ) fe - rich knots .
the knots belonging to these three groups are well separated from each other in f([]-1.03)/f([]-1.644 ) vs. f(he i-1.083)/f([]-1.644 ) plane ( figure [ fig - fluxratio ] ) , so that one may use these line ratios to classify the knots in cas a. it would be interesting to determine whether this classification methodology applies for other core - collapse snrs . \3 .
the he - rich knots show bright emission lines of he i 1.083 and [ ] together with [ ] and h i lines .
their line - of - sight speeds are small ( @xmath42 ) . from these chemical and kinematical characteristics ,
we conclude that the he - rich knots are dense csm swept up by the sn blast wave .
these knots correspond to the previously known qsfs .
the s - rich knots show strong forbidden lines of s together with [ ] and [ ] , and their line - of - sight speeds reach a few 1000 @xmath78 .
these chemical and kinematical properties indicate that the s - rich knots are dense sn ejecta material mostly originating from the o - burning layers and swept up by a reverse shock .
these knots correspond to the fmks detected in previous optical studies .
the fe - rich knots only show strong [ ] and [ ] lines , and no or weak he i 1.083 lines . like the s - rich knots
, they have large line - of - sight speeds ( up to @xmath63 ) and broad line widths ( 1035 ) , but they do not show the lines from si , p , and s. some fe - rich knots show he i 1.083 but their fluxes compared to the [ ] lines are much weaker than those of the he - rich knots .
these spectroscopic properties suggest that the fe - rich knots are most likely `` pure '' dense fe ejecta from the innermost layer of the sn . the comparison of [ ] 1.644 images with the _ hst _ acs / wfc f850lp and f775w and _ nustar _
@xmath2ti images reveals that these fe ejecta are mainly distributed in the sw main ejecta shell , just outside the unshocked @xmath2ti in the interior .
this supports that there could be a large amount of unshocked `` pure '' fe ejecta associated with @xmath2ti .
together with the diffuse , x - ray - emitting `` pure '' fe ejecta detected by _ chandra _ , our result implies that the fe ejecta synthesized in the innermost region develop large - scale non - uniformity during the sn explosion and are expelled asymmetrically .
this seems to be consistent with the low - mode , convection - driven sn explosion model .
we wish to thank the anonymous referee for the very useful comments and suggestions which helped us improve the quality of the paper .
we thank brian grefenstette and fiona harrison for providing the _ nustar _ data .
we also want to thank john raymond and sung - chul yoon for helpful discussions .
the interactive 3d figures were made by using asymptote which is a descriptive vector graphics language .
this research was supported by basic science research program through the national research foundation of korea(nrf ) funded by the ministry of science , ict and future planning ( 2014r1a2a2a01002811 ) .
cccrccr 2008 jun 29 & 1 & 23:23:28.54 + 58:50:32.2 & 0.0 & both & 50.50 & 300 @xmath79 4 + & 8 & 23:23:18.13 + 58:50:14.4 & 13.4 & ab & 51.00 & 300 @xmath79 5 + & 2 & 23:23:37.60 + 58:49:59.7 & 288.7 & ab & 50.25 & 300 @xmath79 4 + & st & 23:48:53.97 + 59:58:44.3 & 0.0 & ab & 30.00 & 30 @xmath79 6 + 2008 aug 08 & 3 & 23:23:41.11 + 58:48:45.2 & 36.1 & ab & 45.25 & 300 @xmath79 6 + & 4 & 23:23:24.77 + 58:47:12.2 & 12.0 & ab & 46.25 & 300 @xmath79 6 + & 5 & 23:23:18.50 + 58:47:25.1 & 307.3 & os & 30.00 & 300 @xmath79 1 + & 6 & 23:23:18.33 + 58:47:23.5 & 307.3 & os & 30.00 & 300 @xmath79 1 + & 7 & 23:23:13.57 + 58:47:49.2 & 5.2 & ab & 45.25 & 300 @xmath79 6 + & st & 23:48:53.97 + 59:58:44.3 & 0.0 & ab & 30.00 & 30 @xmath79 8 + cc|c|rcrrr 1 & 1 & 23:23:28.54 + 58:50:47.6 & 8.75 & s & 8.0 ( 0.5 ) & @xmath80 ( 2 ) & 589 ( 8) + 1 & 2 & 23:23:28.54 + 58:50:34.1 & 3.25 & s & 10.6 ( 0.7 ) & @xmath81 ( 2 ) & 165 ( 4 ) + 1 & 3 & 23:23:28.54 + 58:50:33.8 & 5.00 & he & 5.8 ( 0.1 ) & @xmath82 ( 2 ) & 793 ( 5 ) + 1 & 4 & 23:23:28.54 + 58:50:32.3 & 5.50 & s & 3.3 ( 2.0 ) & @xmath83 ( 17 ) & 96 ( 13 ) + 1 & 5 & 23:23:28.54 + 58:50:31.1 & 5.75 & s & 4.9 ( 0.2 ) & @xmath84 ( 2 ) & 383 ( 3 ) + 1 & 6 & 23:23:28.54 + 58:50:27.6 & 3.00 & s & 3.3 ( 0.5 ) & @xmath85 ( 3 ) & 132 ( 4 ) + 1 & 7 & 23:23:28.54 + 58:50:26.6 & 5.75 & fe & 9.8 ( 0.4 ) & @xmath86 ( 2 ) & 561 ( 10 ) + 1 & 8 & 23:23:28.54 + 58:50:25.8 & 3.00 & s & 3.4 ( 0.4 ) & @xmath87 ( 2 ) & 198 ( 5 ) + 2 & 1 & 23:23:35.19 + 58:50:06.0 & 4.50 & fe & 5.5 ( 0.9 ) & @xmath88 ( 6 ) & 204 ( 10 ) + 2 & 2 & 23:23:35.10 + 58:50:06.3 & 5.75 & s & 9.3 ( 3.6 ) & @xmath89 ( 14 ) & 75 ( 10 ) + 2 & 3 & 23:23:35.83 + 58:50:04.3 & 8.00 & s & 4.2 ( 1.7 ) & @xmath90 ( 20 ) & 140 ( 12 ) + 2 & 4 & 23:23:38.21 + 58:49:58.1 & 9.25 & s & 7.0 ( 0.5 ) & @xmath91 ( 3 ) & 480 ( 10 ) + 2 & 5 & 23:23:38.21 + 58:49:58.1 & 6.50 & s & 11.0 ( 1.0 ) & @xmath92 ( 3 ) & 339 ( 12 ) + 2 & 6 & 23:23:38.70 + 58:49:56.8 & 3.75 & s & 9.2 ( 1.7 ) & @xmath93 ( 2 ) & 164 ( 7 ) + 3 & 1 & 23:23:41.64 + 58:48:50.9 & 6.75 & s & 8.2 ( 0.3 ) & @xmath94 ( 3 ) & 460 ( 6 ) + 3 & 2 & 23:23:41.28 + 58:48:47.0 & 5.00 & s & 7.6 ( 0.3 ) & @xmath95 ( 3 ) & 310 ( 4 ) + 3 & 3 & 23:23:40.92 + 58:48:43.2 & 7.25 & s & 7.3 ( 0.1 ) & @xmath96 ( 2 ) & 972 ( 4 ) + 3 & 4 & 23:23:40.52 + 58:48:38.9 & 6.50 & s & 7.2 ( 0.3 ) & @xmath97 ( 4 ) & 434 ( 8) + 3 & 5 & 23:23:40.16 + 58:48:35.1 & 5.75 & s & 6.2 ( 1.2 ) & @xmath98 ( 7 ) & 121 ( 7 ) + 4 & 1 & 23:23:24.90 + 58:47:16.8 & 5.25 & fe & 15.4 ( 1.4 ) & @xmath99 ( 5 ) & 240 ( 6 ) + 4 & 2 & 23:23:24.82 + 58:47:13.9 & 3.75 & s & 10.6 ( 0.6 ) & @xmath100 ( 4 ) & 289 ( 4 ) + 4 & 3 & 23:23:24.77 + 58:47:12.2 & 5.75 & fe & 7.5 ( 0.4 ) & @xmath101 ( 3 ) & 298 ( 5 ) + 4 & 4 & 23:23:24.73 + 58:47:10.7 & 6.25 & s & 9.9 ( 0.3 ) & @xmath102 ( 3 ) & 606 ( 5 ) + 4 & 5 & 23:23:24.73 + 58:47:10.7 & 4.00 & fe & 6.6 ( 0.4 ) & @xmath103 ( 4 ) & 236 ( 4 ) + 4 & 6 & 23:23:24.72 + 58:47:10.2 & 5.75 & s & 10.3 ( 0.3 ) & @xmath104 ( 3 ) & 577 ( 6 ) + 5 & 1 & 23:23:17.21 + 58:47:32.7 & 5.25 & s & 11.5 ( 0.8 ) & @xmath105 ( 3 ) & 313 ( 7 ) + 5 & 2 & 23:23:17.28 + 58:47:32.3 & 5.25 & s & 8.8 ( 0.5 ) & @xmath106 ( 3 ) & 561 ( 10 ) + 5 & 3 & 23:23:17.34 + 58:47:32.0 & 4.50 & fe & 8.7 ( 0.9 ) & @xmath107 ( 4 ) & 286 ( 8) + 5 & 4a & 23:23:17.26 + 58:47:32.4 & 6.25 & s & 8.3 ( 0.3 ) & @xmath108 ( 3 ) & 537 ( 7 ) + 5 & 4b & 23:23:17.34 + 58:47:32.0 & 6.25 & he & 8.3 ( 0.1 ) & @xmath108 ( 3 ) & 2149 ( 7 ) + 5 & 5 & 23:23:17.62 + 58:47:30.3 & 5.25 & s & 9.3 ( 0.5 ) & @xmath109 ( 3 ) & 456 ( 6 ) + 5 & 6 & 23:23:17.88 + 58:47:28.8 & 4.75 & s & 8.8 ( 0.8 ) & @xmath110 ( 4 ) & 273 ( 8) + 5 & 7 & 23:23:18.00 + 58:47:28.0 & 3.75 & s & 7.2 ( 0.6 ) & @xmath111 ( 3 ) & 215 ( 5 ) + 5 & 8 & 23:23:18.72 + 58:47:23.8 & 5.00 & s & 10.8 ( 0.5 ) & @xmath112 ( 3 ) & 492 ( 9 ) + 5 & 9 & 23:23:19.41 + 58:47:19.7 & 1.75 & s & 11.9 ( 1.8 ) & @xmath113 ( 8) & 99 ( 5 ) + 5 & 10 & 23:23:19.49 + 58:47:19.3 & 9.75 & fe & 7.9 ( 0.0 ) & @xmath114 ( 3 ) & 4959 ( 8) + 6 & 1 & 23:23:17.02 + 58:47:31.3 & 3.50 & s & 6.6 ( 0.9 ) & @xmath115 ( 6 ) & 245 ( 8) + 6 & 2 & 23:23:17.07 + 58:47:31.0 & 3.25 & s & 9.0 ( 0.7 ) & @xmath116 ( 4 ) & 340 ( 7 ) + 6 & 3 & 23:23:16.94 + 58:47:31.7 & 4.25 & s & 10.2 ( 0.3 ) & @xmath117 ( 3 ) & 667 ( 7 ) + 6 & 4a & 23:23:17.22 + 58:47:30.0 & 2.75 & he & 10.0 ( 2.1 ) & @xmath118 ( 6 ) & 58 ( 6 ) + 6 & 4b & 23:23:17.30 + 58:47:29.6 & 2.75 & s & 10.4 ( 3.6 ) & @xmath118 ( 6 ) & 34 ( 6 ) + 6 & 5 & 23:23:17.22 + 58:47:30.0 & 3.00 & s & 7.4 ( 0.8 ) & @xmath119 ( 5 ) & 272 ( 6 ) + 6 & 6 & 23:23:17.53 + 58:47:28.2 & 6.25 & s & 10.4 ( 0.4 ) & @xmath120 ( 3 ) & 714 ( 8) + 6 & 7 & 23:23:17.99 + 58:47:25.5 & 4.50 & s & 7.8 ( 0.3 ) & @xmath121 ( 3 ) & 623 ( 8) + 6 & 8 & 23:23:18.47 + 58:47:22.6 & 2.75 & s & 9.3 ( 0.7 ) & @xmath122 ( 4 ) & 171 ( 5 ) + 6 & 9 & 23:23:18.75 + 58:47:21.0 & 2.75 & fe & 8.3 ( 0.5 ) & @xmath123 ( 4 ) & 258 ( 5 ) + 6 & 10 & 23:23:19.04 + 58:47:19.3 & 3.50 & fe & 7.3 ( 0.4 ) & @xmath124 ( 3 ) & 283 ( 5 ) + 6 & 11 & 23:23:19.29 + 58:47:17.8 & 6.50 & fe & 6.9 ( 0.2 ) & @xmath125 ( 3 ) & 676 ( 7 ) + 6 & 12 & 23:23:19.52 + 58:47:16.4 & 4.25 & fe & 9.9 ( 1.7 ) & @xmath126 ( 6 ) & 92 ( 6 ) + 7 & 1 & 23:23:13.66 + 58:47:56.4 & 3.25 & s & 10.2 ( 0.7 ) & @xmath127 ( 3 ) & 161 ( 3 ) + 7 & 2 & 23:23:13.62 + 58:47:53.2 & 5.50 & he & 8.4 ( 0.2 ) & @xmath128 ( 3 ) & 424 ( 6 ) + 7 & 3 & 23:23:13.51 + 58:47:44.2 & 4.00 & he & 8.5 ( 0.2 ) & @xmath129 ( 2 ) & 647 ( 9 ) + 7 & 4 & 23:23:13.49 + 58:47:42.0 & 2.75 & he & 8.5 ( 0.3 ) & @xmath130 ( 3 ) & 480 ( 10 ) + 7 & 5 & 23:23:13.46 + 58:47:39.7 & 3.25 & he & 9.1 ( 0.2 ) & @xmath131 ( 3 ) & 516 ( 7 ) + 8 & 1 & 23:23:18.56 + 58:50:28.1 & 3.25 & s & 8.1 ( 0.9 ) & @xmath132 ( 3 ) & 180 ( 7 ) + 8 & 2 & 23:23:18.43 + 58:50:24.0 & 6.00 & s & 7.6 ( 0.5 ) & @xmath133 ( 3 ) & 369 ( 9 ) + 8 & 3 & 23:23:18.22 + 58:50:17.1 & 6.25 & s & 7.3 ( 0.4 ) & @xmath134 ( 2 ) & 443 ( 9 ) + 8 & 4 & 23:23:18.07 + 58:50:12.3 & 5.25 & s & 9.2 ( 0.4 ) & @xmath135 ( 3 ) & 364 ( 6 ) + 8 & 5 & 23:23:17.93 + 58:50:07.7 & 4.50 & s & 10.8 ( 1.6 ) & @xmath136 ( 6 ) & 170 ( 8) + 8 & 6 & 23:23:17.88 + 58:50:06.2 & 4.00 & s & 13.6 ( 1.7 ) & @xmath137 ( 3 ) & 139 ( 5 ) + 8 & 7 & 23:23:17.76 + 58:50:02.3 & 4.75 & s & 9.4 ( 0.7 ) & @xmath138 ( 4 ) & 378 ( 10 ) + 8 & 8 & 23:23:17.76 + 58:50:02.1 & 4.00 & s & 6.8 ( 0.8 ) & @xmath139 ( 5 ) & 239 ( 12 ) + 8 & 9 & 23:23:17.67 + 58:49:59.1 & 4.00 & s & 8.1 ( 0.6 ) & @xmath140 ( 3 ) & 294 ( 6 ) + cc|lr|rr|r 1 & 1 & [ s iii ] @xmath141 - @xmath142 & 0.95311 & 7.2 ( 0.2 ) & 9233 ( 261 ) & + 1 & 1 & [ c i ] @xmath143 - @xmath142 & 0.98241 & @xmath144 ( @xmath144 ) & @xmath144 ( 15 ) & + 1 & 1 & [ c i ] @xmath141 - @xmath142 & 0.98503 & @xmath144 ( @xmath144 ) & @xmath144 ( 15 ) & + 1 & 1 & [ s ii ] @xmath145 - @xmath146 & 1.02867 & 8.4 ( 0.1 ) & 2317 ( 18 ) & line - fix + 1 & 1 & [ s ii ] @xmath147 - @xmath146 & 1.03205 & 8.4 ( - ) & 3182 ( - ) & line - fix + 1 & 1 & [ s ii ] @xmath145 - @xmath148 & 1.03364 & 8.4 ( - ) & 2185 ( 13 ) & line - fix + 1 & 1 & [ s ii ] @xmath147 - @xmath148 & 1.03705 & 8.4 ( - ) & 1058 ( - ) & line - fix + 1 & 1 & [
n i ] @xmath147 - @xmath149 & 1.03979 & @xmath144 ( @xmath144 ) & @xmath144 ( 12 ) & + 1 & 1 & [ n i ] @xmath145 - @xmath149 & 1.04074 & @xmath144 ( @xmath144 ) & @xmath144 ( 14 ) & + 1 & 1 & [ s i ] @xmath141 - @xmath142 & 1.08212 & 9.9 ( 0.5 ) & 207 ( 15 ) & + 1 & 1 & he i @xmath150 - @xmath151 & 1.08302 & @xmath144 ( @xmath144 ) & @xmath144 ( 11 ) & + 1 & 1 & h i pa@xmath57 & 1.09381 & @xmath144 ( @xmath144 ) & @xmath144 ( 9 ) & + 1 & 1 & [ s i ] @xmath143 - @xmath142 & 1.13059 & @xmath144 ( @xmath144 ) & @xmath144 ( 13 ) & + 1 & 1 & [ p ii ] @xmath143 - @xmath142 & 1.14682 & 7.8 ( 0.7 ) & 293 ( 48 ) & oh - cont + 1 & 1 & [ p ii ] @xmath141 - @xmath142 & 1.18828 & 9.8 ( 0.1 ) & 736 ( 15 ) & + 1 & 1 & [ fe ii ] @xmath152 - @xmath153 & 1.24854 & @xmath144 ( @xmath144 ) & @xmath144 ( 7 ) & + 1 & 1 & [ fe ii ] @xmath154 - @xmath155 & 1.25214 & @xmath144 ( @xmath144 ) & @xmath144 ( 8 ) & + 1 & 1 & [ fe ii ] @xmath156 - @xmath33 & 1.25668 & 10.2 ( 0.3 ) & 388 ( 15 ) & + 1 & 1 & [ fe ii ] @xmath157 - @xmath155 & 1.27035 & @xmath144 ( @xmath144 ) & @xmath144 ( 18 ) & + 1 & 1 & [ fe ii ] @xmath154 - @xmath158 & 1.27878 & 4.0 ( 0.6 ) & 30 ( 6 ) & + 1 & 1 & h i pa@xmath56 & 1.28181 & @xmath144 ( @xmath144 ) & @xmath144 ( 12 ) & + 1 & 1 & [ fe ii ] @xmath159 - @xmath153 & 1.29427 & 9.3 ( 1.0 ) & 133 ( 22 ) & oh - cont + 1 & 1 & [ fe ii ] @xmath157 - @xmath158 & 1.29777 & @xmath144 ( @xmath144 ) & @xmath144 ( 12 ) & + 1 & 1 & [ fe ii ] @xmath152 - @xmath33 & 1.32055 & 17.5 ( 2.6 ) & 92 ( 18 ) & + 1 & 1 & [ fe ii ] @xmath154 - @xmath153 & 1.32778 & 12.3 ( 2.1 ) & 67 ( 15 ) & + 1 & 1 & [ fe ii ] @xmath160 - @xmath153 & 1.53347 & 8.7 ( 0.9 ) & 95 ( 12 ) & oh - cont + 1 & 1 & [ fe ii ] @xmath161 - @xmath158 & 1.59947 & 9.8 ( 1.1 ) & 76 ( 11 ) & oh - cont + 1 & 1 & [ si i ] @xmath143 - @xmath142 & 1.60683 & @xmath144 ( @xmath144 ) & @xmath144 ( 5 ) & + 1 & 1 & [ fe ii ] @xmath160 - @xmath33 & 1.64355 & 13.7 ( 0.1 ) & 589 ( 8 ) & + 1 & 1 & [ si i ] @xmath141 - @xmath142 & 1.64545 & @xmath144 ( @xmath144 ) & @xmath144 ( 7 ) & + 1 & 1 & [ fe ii ] @xmath162 - @xmath155 & 1.66377 & 14.0 ( 1.5 ) & 73 ( 11 ) & oh - cont + 1 & 1 & [ fe ii ] @xmath161 - @xmath153 & 1.67688 & 15.4 ( 0.5 ) & 146 ( 8 ) & + 1 & 1 & [ fe ii ] @xmath162 - @xmath158 & 1.71113 & @xmath144 ( @xmath144 ) & @xmath144 ( 9 ) & + 1 & 1 & [ fe ii ] @xmath163 - @xmath155 & 1.74494 & @xmath144 ( @xmath144 ) & @xmath144 ( 5 ) & + 1 & 1 & [ fe ii ] @xmath163 - @xmath158 & 1.79710 & @xmath144 ( @xmath144 ) & @xmath144 ( 62 ) & + 1 & 1 & [ fe ii ] @xmath162 - @xmath153 & 1.80002 & @xmath144 ( @xmath144 ) & @xmath144 ( 240 ) & + 1 & 1 & [ fe ii ] @xmath161 - @xmath33 & 1.80939 & @xmath144 ( @xmath144 ) & @xmath144 ( 689 ) & + 1 & 1 & [ si vi ] @xmath146 - @xmath148 & 1.96287 & 16.4 ( 0.6 ) & 501 ( 27 ) & + 1 & 1 & [ fe ii ] @xmath164 - @xmath165 & 2.04601 & @xmath144 ( @xmath144 ) & @xmath144 ( 20 ) & + 1 & 1 & he i @xmath166 - @xmath167 & 2.05813 & @xmath144 ( @xmath144 ) & @xmath144 ( 12 ) & + 1 & 1 & [ fe ii ] @xmath168 - @xmath165 & 2.13277 & @xmath144 ( @xmath144 ) & @xmath144 ( 9 ) & + 1 & 1 & [ fe iii ] @xmath169 - @xmath170 & 2.14511 & @xmath144 ( @xmath144 ) & @xmath144 ( 10 ) & + 1 & 1 & h i br@xmath57 & 2.16553 & @xmath144 ( @xmath144 ) & @xmath144 ( 11 ) & + 1 & 1 & [ fe iii ] @xmath171 - @xmath172 & 2.21779 & @xmath144 ( @xmath144 ) & @xmath144 ( 13 ) & + 1 & 1 & [ fe ii ] @xmath173 - @xmath174 & 2.22379 & @xmath144 ( @xmath144 ) & @xmath144 ( 10 ) & + 1 & 1 & [ fe iii ] @xmath169 - @xmath175 & 2.24209 & @xmath144 ( @xmath144 ) & @xmath144 ( 13 ) & + ccrr 1 & 9.8 & 42.7 & 42.7 + 2 & 5.1 & 22.2 & 64.9 + 3 & 4.7 & 20.4 & 85.3 + 4 & 1.1 & 4.7 & 90.0 + 5 & 1.0 & 4.2 & 94.2 + 6 & 0.4 & 1.8 & 96.0 + 7 & 0.3 & 1.3 & 97.4 + 8 & 0.2 & 0.7 & 98.1 + 9 & 0.1 & 0.5 & 98.6 + 10 & 0.1 & 0.4 & 98.9 + | we report the results of broadband ( 0.952.46 ) near - infrared spectroscopic observations of the cassiopeia a supernova remnant . using a clump - finding algorithm in two - dimensional dispersed images ,
we identify 63 ` knots ' from eight slit positions and derive their spectroscopic properties .
all of the knots emit [ ] lines together with other ionic forbidden lines of heavy elements , and some of them also emit h and he lines .
we identify 46 emission line features in total from the 63 knots and measure their fluxes and radial velocities .
the results of our analyses of the emission line features based on principal component analysis show that the knots can be classified into three groups : ( 1 ) he - rich , ( 2 ) s - rich , and ( 3 ) fe - rich knots .
the he - rich knots have relatively small , @xmath0 , line - of - sight speeds and radiate strong he i and [ ] lines resembling closely optical quasi - stationary flocculi of circumstellar medium , while the s - rich knots show strong lines from o - burning material with large radial velocities up to @xmath1 indicating that they are supernova ejecta material known as fast - moving knots .
the fe - rich knots also have large radial velocities but show no lines from o - burning material .
we discuss the origin of the fe - rich knots and conclude that _ they are most likely `` pure '' fe ejecta synthesized in the innermost region during the supernova explosion . _ the comparison of [ ] images with other waveband images shows that these dense fe ejecta are mainly distributed along the southwestern shell just outside the unshocked @xmath2ti in the interior , supporting the presence of unshocked fe associated with @xmath2ti . |
[ n1n2n3table ] @xmath1 [ brandnewsquaretable ] @xmath2 the well - known squares in figure [ franklinsquares ] were constructed by benjamin franklin .
the square f2 was introduced separately and hence is generally known as _
the other 8-square_. the entries of the squares are from the set @xmath3 , where @xmath4 or @xmath5 . every integer in this set occurs in the square exactly once . for these squares ,
the entries of every row and column add to a common sum called the _ magic sum_. the @xmath6 squares have magic sum 260 and the @xmath7 square has magic sum 2056 . in every half row and half column the entries add to half the magic sum .
the entries of the main bend diagonals and all the bend diagonals parallel to it add to the magic sum .
in addition , observe that every @xmath8 sub - square in f1 and f2 adds to half the magic sum , and in f3 adds to one - quarter the magic sum .
the property of the @xmath9 sub - squares adding to a common sum and the property of bend diagonals adding to the magic sum are _ continuous properties_. by continuous property we mean that if we imagine the square is the surface of a torus ; i.e. opposite sides of the square are glued together , then the bend diagonals or the @xmath10 sub - squares can be translated and still the corresponding sums hold . in fact , these squares have innumerable fascinating properties .
see @xcite , @xcite , and @xcite for a detailed study of these squares . from now on ,
row sum , column sum , or bend diagonal sum , etc .
mean that we are adding the entries of those elements .
based on the descriptions of benjamin franklin , we define _ franklin squares _ as follows ( see @xcite ) .
consider an integer , @xmath11 such that @xmath12 .
let the magic sum be denoted by @xmath13 and @xmath14 .
we define an _
@xmath15 franklin square _ to be a @xmath16 matrix with the following properties : 1 .
every integer from the set @xmath3 occurs exactly once in the square .
consequently , @xmath17 2 .
all the the half rows , half columns add to one - half the magic sum .
consequently , all the rows and columns add to the magic sum .
all the bend diagonals add to the magic sum , continuously .
all the @xmath9 sub - squares add to @xmath18 , continuously .
consequently , all the @xmath19 sub - squares add to @xmath20 , and the four corner numbers with the four middle numbers add to @xmath21 .
we call permutations of the entries of a franklin square that preserve the defining properties of the franklin squares _
symmetry operations _ , and two squares are called _ isomorphic _ if we can get one from the other by applying symmetry operations . rotation and reflection
are clearly symmetry operations .
see @xcite for more symmetry operations like exchanging specific rows or columns . in @xcite
, we showed that f1 and f2 are not isomorphic to each other .
constructing franklin squares are demanding and only a handful such squares are known to date .
we showed how to construct f1 , f2 , and f3 using methods from algebra and combinatorics in @xcite . in the same article ,
for the first time since benjamin franklin , we constructed new franklin squares n1 , n2 , and n3 , given in table [ n1n2n3table ] . we proved that these squares were not isomorphic to each other nor to f1 or f2 .
in other words , these squares were really new . those methods being computationally challenging are not suitable for higher order franklin squares
moreover , the constructions used computers and hence lacked the intrigue of franklin s constructions . in @xcite
, we followed benjamin franklin closely , and used elementary techniques to construct a franklin square of every order . with these techniques we are able to construct f1 and f3 , but not f2 . in section [ constructsection ] , we modify the methods in @xcite to construct f2 . in section [ newconstructsection ] , we create new franklin squares b1 and b2 , given in table [ brandnewsquaretable ] , but this time using elementary techniques , in keeping with the true spirit of benjamin franklin .
in this section , we show how to construct the franklin square f2 .
we present a quick recap of our construction of f1 in @xcite . note that @xmath23 throughout this article .
the @xmath24 square was divided in to two halves : the left half consisting of first four columns and the right half consisting of the last four columns .
the left half is constructed by first placing the numbers @xmath25 in a specific pattern as shown in figure [ f1leftconstruct ] .
then , the number @xmath26 is placed in the same row that contains @xmath27 , where @xmath28 , such that @xmath27 and @xmath26 are always equidistant from the middle of the left half . see figure [ f1leftconstruct ] .
similarly , the second half is constructed by first placing numbers @xmath29 in a specific pattern .
finally , the numbers @xmath26 is placed in the same row that contains @xmath27 , where @xmath30 , such that @xmath27 and @xmath26 are always equidistant from the middle of the right half .
this procedure is explained in figure [ f1rightconstruct ] .
the algorithm for generating the pattern of @xmath31 in f1 is given in @xcite .
we also generalized this procedure to construct a franklin square for any given order in @xcite .
from now on , throughout the article , when we say _ pattern of a square _ , we mean the pattern of the numbers @xmath31 in the square . in our search for new franklin squares ,
it is natural to swap rows for columns in the above method .
thus , we divide the @xmath6 square in to two halves : the top half consisting of the first four rows and the bottom half consisting of the last four rows .
we place the numbers @xmath32 in the bottom half , and the numbers @xmath33 in the top half in the pattern shown in figure [ columnsconstruct ] . this time we place the numbers @xmath26 in the same columns as @xmath27 , equidistant from the middle of each part .
the franklin square s we get from this procedure is isomorphic to f1 : check that we can derive this square from f1 by rotating it and then reflecting it .
the square s is not interesting because it is essentially the same as f1 .
but the pattern of entering the numbers @xmath31 in s is algorithmic since it is just a row version of the pattern of f1 . that is , instead of filling two columns at a time , we fill two rows at a time .
we modify the pattern of s to get a pattern for f2 .
after that , we place @xmath26 in the same column as @xmath27 , for @xmath34 , equidistant from the center of each half , as usual . see figure [ f2constructfig ] for the construction of f2 .
the construction of f2 involved moving columns of in the @xmath35 pattern of the square s and then switching elements along the diagonals .
this pattern is not easily generalized to higher orders .
but in the next section , we modify the patterns of known squares to create new squares .
in this section we construct the new franklin squares b1 and b2 ( see table [ brandnewsquaretable ] ) .
to begin with , we look at the patterns of the numbers @xmath36 appearing in the franklin squares n1 , n2 , and n3 .
these are shown in figure [ nipatternsfig ] .
we found that even in these squares that were constructed using hilbert bases ( see @xcite ) , the strategy of finding the pattern of the franklin square , and then placing @xmath27 and @xmath26 in the same row or column , as the case may be , equidistant from the center of relevant half of the square , worked .
observe that the patterns of n1 and n2 are derived by modifying the pattern of f2 .
so in their constructions , we will place @xmath27 and @xmath26 in the same columns . on the other hand , since
, n3 is a permutation of the pattern of f1 , we place @xmath27 and @xmath26 in the same rows while constructing n3 .
benjamin franklin s patterns always restrict the entries @xmath37 to one half of the square .
observe that n2 is the only square with this property . guided by these observations
, we modified the pattern of f2 to build the new franklin squares b1 and b2 .
see figure [ b1b2pattern ] for the patterns of b1 and b2 .
since the patterns were derived from f2 , to construct them , we will place @xmath27 and @xmath26 in the same columns .
b1 and b2 are not isomorphic to each other or any other known squares because the columns were made different by construction . | in this article , we reveal how benjamin franklin constructed his second @xmath0 magic square .
we also construct two new @xmath0 franklin squares . |
piro et al . ( 1999 ) and yoshida et al . ( 1999 ) report the detection of an iron emission line feature in the x ray afterglow spectra of grb 970508 and grb 970828 , respectively .
both lines are characterized by a large flux and equivalent width ( ew ) compared with the theoretical previsions made in the framework of the hypernova and compact merger grb progenitor models ( ghisellini et al .
1999 , bttcher et al .
the line detected in grb 970508 is consistent with an iron @xmath0 line redshifted to the rest frame of the candidate host galaxy ( @xmath1 , metzger et al .
1997 ) , while grb 970828 has no measured redshift and the identification of the feature with the same line would imply a redshift @xmath2 . the line fluxes ( equivalent widths ) are @xmath3 erg @xmath4 s@xmath5 ( @xmath6 kev ) and @xmath7 erg @xmath4 s@xmath5 ( @xmath8 kev ) for grb 970508 and grb 970828 , respectively .
as indicated by the flux uncertainties , these line features are revealed at a statistical level of @xmath9 . a strong iron emission line unambiguously points towards the presence , in the vicinity of the burster , of a few per cent of iron solar masses concentrated in a compact region .
thus the presence of such a line in the x ray afterglow spectrum would represent the `` rosetta stone '' for unveiling the burst progenitor . to date
, three main classes of models have been proposed for the origin of gamma ray bursts ( grb ) : neutron star neutron star ( ns ns ) mergers ( paczyski 1986 ; eichler et al .
1989 ) , hypernovae or failed type ib supernovae ( woosley 1993 ; paczyski 1998 ) and supranovae ( vietri & stella 1998 ) . in the ns
ns model the burst is produced during the collapse of a binary system composed of two neutron stars or of a neutron star and a black hole . interacting neutrinos or the spin down of the remnant black hole should power a relativistic outflow and , eventually , the production of @xmath10ray photons .
hypernovae are the final evolutionary stage of very massive stars ( @xmath11 ) , whose core implodes in a black hole without the explosion of a classical supernova .
again , the rotational energy of the remnant black hole is tapped by a huge magnetic field , powering the burst explosion . while ns
ns mergers should take place outside the original star formation site , due to the large peculiar velocities and long lifetime of the coalescing binary systems , hypernovae take place in the dense and possibly iron rich molecular cloud where the massive star was born .
the supranova scenario ( vietri & stella 1998 ) assumes that , following a supernova explosion , a fast spinning neutron star is formed , with a mass that would be supercritical in the absence of rotation . as radiative energy losses spin it down in a time scale of months to years , it inevitably collapses to a kerr black hole , whose rotational energy can then power the grb . a supernova remnant ( snr )
is naturally left over around the burst location .
since the detected lines are both characterized by a flux of several @xmath12 erg s@xmath5 , we adopt as reference a line with a flux as @xmath13 and adopt cgs units . ]
@xmath14 erg @xmath4 s@xmath5 .
this in itself constrains both the amount of line emitting matter and the size of the emitting region .
we assume that the emitting region is a homogeneous spherical shell centered in the grb progenitor , with radius @xmath15 and width @xmath16 .
a limit to the size of the emitting region can be set , _
independently from the flux variability _ , imposing the flux level times the minimum duration of the emission ( @xmath17 ) to be less than the total available energy supplied by the burst fluence . allowing for an efficiency @xmath18 of the burst photon reprocessing into the line ( @xmath19 , see ghisellini et al .
1999 ) we have : @xmath20 where @xmath21 is the total grb fluence . note that this result , obtained with independent arguments , agrees with the @xmath22 day variability time
scales of the detected features . to obtain a lower limit to the total amount of mass ( see lazzati et al .
1999 ) we consider a parameter @xmath23 given by the ratio between the total number of iron line photons divided by the number of iron nuclei : @xmath23 is the number of photons produced by a single iron atom .
this number can be constrained to be less than the total number of ionizations an ion can undergo when illuminated by the burst or x ray afterglow photons . for a grb located at @xmath24 , we obtain km s@xmath5 mpc@xmath5 , @xmath25 and @xmath26 . ] : @xmath27 where @xmath28 it the total energy of the burst photons , @xmath29 the energy of an ionizing photon and @xmath30 the ionization cross section of the iron k shell .
the total mass is given by the ratio of the total number of produced photons divided by @xmath23 times the mass of the iron atom .
we have : @xmath31 where @xmath32 is the time during which the line is observed in units of @xmath33 seconds and @xmath34 is the iron abundance in solar units .
if such a large amount of mass were uniformly spread around the burst location , it would completely stop the fireball .
in fact ( see e.g. wijers , rees & mszros 1997 ) the fireball is slowed down to sub relativistic speeds when the swept up mass equals its initial rest mass . with a typical baryon load of @xmath35 ,
the mass predicted in eq .
[ minmass ] would stop the fireball after an observer time @xmath36 s , i.e. almost one day .
any surviving long wavelength emission should then decay exponentially in the absence of energy supply .
the only way to reconcile a monthly lasting power
law optical afterglow with iron line emission is through a particular geometry , in which the line of sight is devoid of the remnant matter .
this implies ( see also fig .
[ fig1 ] ) that the burst emission is only moderately ( if at all ) beamed , since the burst photons must illuminate the surrounding matter and the line of sight simultaneously .
the limits on the mass discussed in the above section are very general and independent of the mechanism producing the line photons .
lazzati et al .
1999 give a complete analysis of three mechanisms that are able to generate such an intense line feature .
in this scenario , burst ( or x ray afterglow ) photons keep all the iron in the vicinity of the burst fully ionized .
if the density of iron and free electrons is large and the plasma is cool , recombination of free electrons with iron nuclei is efficient and iron line photons are produced .
however , we require that both @xmath37 @xmath38 and that @xmath39 k. during the burst inverse compton interactions of free electrons with @xmath10ray photons heat the plasma to @xmath40 k , and recombination is inefficient .
this scenario may , however , be marginally consistent with grb 970508 data if the ionization flux is provided by an afterglow with an high energy tail but with typical compton temperature of @xmath41 k. a supernova remnant of several solar masses , some months after the explosion of the supernova , has an emission integral of the same order of magnitude of the intergalactic medium of a galaxy cluster .
these systems emit @xmath42 kev iron lines due to collisional excitation of hydrogenoid iron ions .
in addition , the iron richness of a supernova remnant can be up to ten times larger than in a cluster of galaxies , allowing for the emission of iron line with several kev equivalent widths .
in addition to the line , the heated snr produces bremsstrahlung radiation at a level @xmath43 erg @xmath4 s@xmath5 , compatible with a standard x ray afterglow at @xmath44 .
if some dense matter ( e.g. a snr with less than a month of life ) is present in close vicinity with the burster , part of the incident radiation is reflected toward the observer .
this same physical process is efficient in the disc of seyfert galaxies , and a strong iron line is produced since the cross section of the iron atom is much larger than that of free electrons . for a @xmath44 burst
, lines with up to @xmath45 erg @xmath4 s@xmath5 can be produced . in steady state ,
@xmath46 ev ew lines are produced ( if the material covers half of the sky ) , but in gamma
ray bursts the reflected component is seen when the burst emission has already faded and the ew can be much higher ( even a few kev ) . given the fast variability properties of the detected lines , this model is favored with respect to recombination and thermal emission .
the presence of dense matter at rest in the reference frame of the burster ( and not of the fireball ) is a discriminating parameter for the main models of burst progenitor .
here we analyze the two main models , binary neutron star mergers and hypernovae , and a particular model , the supranova , that naturally accounts for the surrounding iron rich matter .
the production of lines in the fireball has been discussed in mszros & rees 1998a .
they find that lines ( particularly iron lines ) can be produced inside the fireball , but we should observe them at a frequency @xmath47 , where @xmath48 is the frequency of the line in the laboratory . in the case of grb 970508
the observed frequency of the line is consistent with an iron line at the same redshift of the burster .
an incredible fine tuning is necessary to explain the feature as an optical uv line blueshifted in the x ray band due to the fireball expansion .
the line must hence be produced by the interaction of the burst photons with cold matter that surrounds the burster .
the lifetime of a binary system composed of a couple of compact objects ( ns ns or bh ns ) ranges from 100 million to 10 billion years , mainly depending on the initial separation and on the mass of the binary system components ( see e.g. lipunov et al 1995 ) . due to the kick velocity that a compact object inherits from the supernova explosion that generated it , @xmath49 km s@xmath5 on average ( kalogera et al .
1998 ) , these binary systems merge well outside their formation site and , possibly , even outside the host galaxy . assuming an hydrogen density @xmath50 @xmath38 and a solar iron abundance , a sphere of radius : @xmath51 is necessary to meet the requirements of eq .
[ minmass ] . however , the radius of the above equation is five orders of magnitude larger than the maximum allowed value ( cf .
[ maxrad ] ) .
if we insert in eq .
[ radio ] the density appropriate for a molecular cloud , in which a very massive star is thought to be located , we obtain @xmath52 cm for @xmath53 @xmath38 that , for an iron rich cloud , is marginally consistent with the limit of eq .
[ maxrad ] . to firmly rule out the possibility of an intense iron line produced in the hypernova scenario , we analyze each emission mechanism discussed in sect .
reflection can be immediately excluded since a radius @xmath54 cm cloud is necessary to have a thomson thick mirror .
the efficiency of thermal emission is proportional to the square of the density .
lazzati et al .
1999 estimate that both a density @xmath55 @xmath38 and a mass of several solar masses ( see also sect . [ emec ] ) are necessary to produce such an intense iron line . in the hypernova scenario ,
the density is 5 orders of magnitude smaller , giving a flux 10 orders of magnitude fainter .
the recombination time in a medium with the considered density is too long ( see bttcher et al . 1999 ) , however a fluorescence line may be emitted due to ionization of the inner shell of almost neutral iron atoms .
the more optimistic estimate of the line flux produced by fluorescence in a molecular cloud is given in ghisellini et al .
1999 , with an upper limit @xmath56 erg @xmath4 s@xmath5 .
even if we consider the presence of a pre hypernova wind with @xmath57/yr , the total mass contained in a radius of @xmath58 cm would be a few percent of a solar mass , leaving the above discussion unaffected .
we hence conclude that in a standard hypernova scenario it is not possible to produce an intense iron line .
mszros & rees 1998b discussed the possible production of an intense iron line
if a compact companion of the hypernova has left a dense torus of material ( @xmath59 ; @xmath60 @xmath38 ) at a distance @xmath61 cm from the burster .
another progenitor model can account for this matter in a more natural way .
in the model presented by vietri & stella 1998 the burst explosion is preceded by a supernova . the neutron star left over by the supernova explosion is supermassive , i.e. its mass exceeds the limiting mass for a neutron star , but the implosion in a black hole is prevented by the fast rotation . as the system spins down , due to radiative losses , the neutron star shrinks until a limiting state is reached and the star collapses in a black hole .
the duration of this transient phase depends from the strength of the magnetic field embedded in the star and from the initial spin velocity , but should typically last between several months and a dozen of years . if we consider a supernova exploded a year before the burst , the density of its shell remnant at the burst onset is : @xmath62 where @xmath63 is the mass of the ejecta in solar units , @xmath64 is their speed in units of @xmath65 cm s@xmath5 and the width of the shell has been assumed @xmath66 .
the shell radius is @xmath67 cm , in good agreement with the constraints derived above .
if we consider that a snr can have an iron richness ten times the solar value ( woosley 1988 ) , this scenario is the most natural one to explain the intense iron line observed in the afterglow of grb 970508 and grb 970828 .
the recent explosion of grb 990123 has gathered the attention of the gamma
ray astronomers due to the huge fluence and large distance of this burst .
the early x ray afterglow flux was measured by the _ beppo_-sax satellite at a level of @xmath68 erg s@xmath5 ( heise et al .
1999 ) about 6 hours after the burst trigger .
the first results on the spectral analysis gives a `` featureless spectrum '' ( heise et al . 1999 ) . if an iron line feature with the same ew of those of grb 970508 and grb 970828 were present in the afterglow of grb 990123 , a firm detection would have been established : since the underlying continuum flux of the latter burst is @xmath69 times larger than in the other two cases , the signal to noise ratio of the line would have been 10 times larger . moreover , due to the larger redshift of grb 990123 ( z=1.6004 , kulkarni et al .
1999 ) the redshifted line energy lies in a region where the detector efficiency is the largest .
this could have led to a @xmath70 line feature .
there are , however , two reasons why an iron line is not expected in the afterglow of grb 990123 .
first , ( lazzati et al . 1999 ) the line intensity is limited by the recombination time and not by the ionization time .
this means that , even in the case of grb 970508 - a weak burst - the ionizing flux was more than enough to fully exploit the matter surrounding the progenitor .
if an iron line with flux @xmath71 erg @xmath4 s@xmath5 was present in the x ray afterglow of the 23 january burst , its ew would be decreased to the very low value of @xmath72 ev , hence well below the detection threshold of the _ beppo_-sax . finally , the break in the power law decrease of the optical light curve of grb 990123 afterglow has been interpreted as a sign of beaming in the burst itself ( kulkarni et al . 1999 ; fruchter et al 1999 ) .
if this interpretation is correct , even if a cloud were present in close vicinity with the burster , its illumination would have been minimal , preventing the formation of an intense line .
we have analyzed how the detection of an iron line redshifted to the rest frame of the burster can be used to constrain the progenitor models .
indeed , due to the relativistic outflows involved in the fireball model , very poor information of the progenitor is contained in the gamma ray burst photons , and
the only mean to investigate the progenitor nature is through the interaction of the photons with the ambient medium .
in particular , we have analyzed three different models predicting very different conditions of the external medium : the merging of two compact objects in a binary system , the hypernova and the supranova .
we derived a ( model independent ) lower limit to the mass required to produce the line , ruling out the ns
ns merger model and , with a careful analysis , the hypernova .
the supranova , which naturally accounts for a shell remnant of iron rich material surrounding the burster seems the most promising model capable to produce an intense iron line . due to the paucity of the statistical significance of the line detections ,
more observations are needed to draw firm conclusions .
we show that , despite its luminosity , grb 990123 is not the ideal test for the line detection and more observations with more efficient satellites are needed .
bttcher , m. , dermer , c. d. , crider , a. w. , & liang , e. p. 1999
, , 343 , 111 eichler , d. , livio , m. , piran , t. , & shramm , d. n. 1989 , nature , 340 , 126 fruchter a. s. , et al . 1999 ,
apj submitted ( astro - ph/9902236 ) heise , j. , et al .
1999 , iauc n. 7099 kalogera , v. , kolb , u. , & king , a. r. 1998 , , 504 , 967 kulkarni , s. , et al .
1999 , nature , 398 , 389 lazzati , d. , campana , s. , & ghisellini , g. 1999 , , 304 , l31 lipunov , v. m. , postnov , k. a. , prokhorov , m. e. , & panchenlo , i. e. 1995 , , 454 , 593 mszros , p. , & rees , m. j. 1998a , , 502 , l105 mszros , p. , & rees , m. j. 1998b , , 299 , l10 metzger , m. r. , et al .
1997 , nature , 387 , 878 paczyski , b. 1986 , , 308 , l43 paczyski , b. 1998 , , 494 , l45 piro , l. , et al .
1999 , , 514 , l73 vietri , m. , & stella .
l. 1998 , , 507 , l45 wijers , r. a. m. j. , rees , m. j. , & mszros , p. 1997
, , 288 , l51 woosley , s. e. 1988 , , 330 , 218 woosley , s. e. 1993 , , 405 , 273 yoshida , a. , et al . 1999 , in press | the discovery of a powerful and transient iron line feature in the x ray afterglow spectrum of grb 970508 and grb 970828 , if confirmed , would be a major breakthrough for understanding the nature the progenitor of gamma
ray bursts .
we show that a large mass of iron very close to the burster is necessary to produce the emission line .
this in itself strongly limits the possible progenitor of the gamma ray event , suggesting the former explosion of a supernova , as predicted in the supranova model ( vietri & stella 1998 ) . the line emission process and the line intensity depend strongly on the age , density and temperature of the remnant .
the simultaneous observation of the iron line and of a power law optical afterglow lasted for one year strongly suggest that the burst emission is isotropic .
recent observations of grb 990123 are also discussed . |
given graphs @xmath0 and @xmath1 , say that @xmath3 if every 2-edge - coloring of @xmath1 results in a monochromatic copy of @xmath0 in @xmath1 .
using this notation , the ramsey number @xmath4 of @xmath0 is the minimum @xmath5 such that @xmath6 . instead of minimizing the number of vertices , one can minimize the number of edges .
define the _ size - ramsey number _ @xmath7 of @xmath0 to be the minimum number of edges in a graph @xmath1 such that @xmath3 .
more formally , @xmath8 the study of size - ramsey numbers was proposed by erds , faudree , rousseau and schelp @xcite in 1978 . by definition of @xmath4
, we have @xmath9 .
since the complete graph on @xmath4 vertices has @xmath10 edges , we obtain the trivial bound @xmath11 chvtal ( see , _
e.g. _ , @xcite ) showed that equality holds in ( [ clique ] ) for complete graphs .
in other words , @xmath12one of the first problems in this area was to determine the size - ramsey number of the @xmath5 vertex path @xmath13 . answering
a question of erds @xcite , beck @xcite showed that @xmath14 since @xmath15 for any graph , beck s result is sharp in order of magnitude .
the linearity of the size - ramsey number of paths was generalized to bounded degree trees by friedman and pippenger @xcite and to cycles by haxell , kohayakawa and uczak @xcite .
beck @xcite asked whether @xmath7 is always linear in the size of @xmath0 for graphs @xmath0 of bounded degree .
this was settled in the negative by rdl and szemerdi @xcite , who proved that there are graphs of order @xmath5 , maximum degree 3 , and size - ramsey number @xmath16 .
they also conjectured that for a fixed integer @xmath17 there is an @xmath18 such that @xmath19 where the maximum is taken over all graphs @xmath0 of order @xmath5 with maximum degree at most @xmath17 .
the upper bound was recently proved by kohayakawa , rdl , schacht , and szemerdi @xcite . for further results about the size - ramsey number
e.g _ , the survey paper of faudree and schelp @xcite .
somewhat surprisingly the size - ramsey numbers have not been studied for hypergraphs , even though classical ramsey numbers for hypergraphs have been studied extensively since the 1950 s ( see , e.g. , @xcite ) , and more recently @xcite . in this paper
we initiate this study for @xmath2-uniform hypergraphs .
a _ @xmath2-uniform hypergraph _
@xmath20 ( _ @xmath2-graph _ for short ) on a vertex set @xmath21 is a family of @xmath2-element subsets ( called edges ) of @xmath21 .
we write @xmath22 for its edge set . given @xmath2-graphs @xmath20 and @xmath23 ,
say that @xmath24 if every 2-edge - coloring of @xmath25 results in a monochromatic copy of @xmath26 in @xmath25 .
define the _ size - ramsey number _ @xmath27 of a @xmath2-graph @xmath20 as @xmath28
motivated by extending the basic theory from graphs to hypergraphs , we prove results for cliques , trees , paths , and bounded degree hypergraphs . for every @xmath2-graph @xmath26 , we trivially have @xmath29 where @xmath30 is the ordinary ramsey number of @xmath20 . our first objective was to generalize ( [ eq : chvatal ] ) to 3-graphs , which shows that equality holds for graphs .
it is fairly easy to obtain a lower bound for @xmath31 that is quadratic in @xmath32 , but we were only able to do slightly better .
[ cliquethm ] @xmath33 .
the following basic questions remain open .
[ ques:1 ] is @xmath34 ?
[ ques:2 ] for @xmath35 let @xmath36 .
define @xmath37 to be the hypergraph obtained from @xmath38 by removing one edge .
is it true that @xmath39 ? clearly , the affirmative answer to the latter gives a negative answer to question [ ques:1 ] .
given integers @xmath40 and @xmath5 , a @xmath2-graph @xmath41 of order @xmath5 with edge set @xmath42 is an _
@xmath43-tree _ , if for each @xmath44 we have @xmath45 and @xmath46 for some @xmath47 .
we are able to give the following general upper bound for trees .
[ thm : ub_tree ] let @xmath40 be fixed integers .
then @xmath48 one can easily show that this bound is tight in order of magnitude when @xmath49 ( see section [ sec : trees ] for details ) .
the situation for @xmath50 is much less clear .
let @xmath51 be fixed integers .
is it true that for every @xmath5 there exists a @xmath2-uniform @xmath43-tree @xmath52 of order at most @xmath5 such that @xmath53 here is another related question pointed out by fox @xcite .
let us weaken the restriction on the edge intersection in the definition of @xmath41 .
let @xmath54 be a @xmath2-graph of order @xmath5 with edge set @xmath42 such that for each @xmath44 we have @xmath45 .
let @xmath51 be fixed integers .
is @xmath55 polynomial in @xmath5 ? given integers @xmath56 and @xmath57 , we define an _
@xmath43-path @xmath58 _ to be the @xmath2-uniform hypergraph with vertex set @xmath59 $ ] and edge set @xmath60 , where @xmath61 and @xmath62 .
in other words , the edges are intervals of length @xmath2 in @xmath59 $ ] and consecutive edges intersect in precisely @xmath43 vertices .
the two extreme cases of @xmath49 and @xmath63 are referred to as , respectively , _ loose _ and _ tight _ paths .
clearly every @xmath43-path is also an @xmath43-tree .
thus , by theorem [ thm : ub_tree ] we obtain the following result .
@xmath64 our first result shows that determining the size - ramsey number of a path @xmath58 for @xmath65 can easily be reduced to the graph case .
[ pathprop ] let @xmath66 . then ,
@xmath67 clearly , this result is optimal .
determining the size - ramsey number of a path @xmath58 for @xmath68 seems to be a much harder problem . here
we will only consider tight paths ( @xmath69 ) . by
we get @xmath70 the most complicated result of this paper is the following improvement of ( [ pathbound ] ) .
[ thm : tightpath ] [ thm : tight_path_k ] fix @xmath71 and let @xmath72 .
then @xmath73 the gap in the exponent of @xmath5 between the upper and lower bounds for this problem remains quite large ( between 1 and @xmath74 ) .
we believe that the lower bound is much closer to the truth .
indeed , the following question still remains open .
is @xmath75 ?
if true , then since @xmath76 , this would imply the linearity of the size - ramsey number of all @xmath43-paths .
our main result about bounded degree hypergraphs is that their size - ramsey numbers can be superlinear .
this is proved by extending the methods of rdl and szmerdi @xcite to the hypergraph case .
[ thm : bounddegree ] let @xmath35 be an integer .
then there is a positive constant @xmath77 such that for every @xmath5 there is a @xmath2-graph @xmath20 of order at most @xmath5 with maximum degree @xmath78 such that @xmath79 there are several other problems to consider such as finding the asymptotic of the size - ramsey number of cycles and many other classes of hypergraphs . in general , they seem to be very difficult .
therefore , this paper is the first step towards a better understanding of this concept . in the next sections we prove these result for cliques ( section [ sec : cliques ] ) , trees ( section [ sec : trees ] ) , paths ( section [ sec : paths ] ) , and hypergraphs with bounded degree ( section [ sec : bounded ] ) .
* proof of theorem [ cliquethm ] .
* we show that if @xmath23 is a 3-graph with @xmath80 for @xmath81 , then @xmath82 .
induction on @xmath83 .
if @xmath84 , then there is a 2-coloring of @xmath85 with no monochromatic @xmath86 .
since @xmath87 , this coloring yields a 2-coloring of @xmath23 with no monochromatic @xmath86 .
suppose that @xmath88 .
since @xmath80 , there are @xmath89 and @xmath90 in @xmath91 with @xmath92 .
otherwise , @xmath93 a contradiction .
let @xmath89 and @xmath90 be such that @xmath94 .
define @xmath95 as follows : @xmath96 and @xmath97 clearly , @xmath98 and @xmath99 . by the inductive hypothesis
there is a 2-coloring @xmath100 of the edges of @xmath95 with no monochromatic @xmath101 .
let @xmath102 .
thus , @xmath103 and @xmath104 .
if there exists @xmath105 such that @xmath106 and @xmath107 $ ] is monochromatic , then set @xmath108 .
if there exists @xmath109 such that @xmath110 and @xmath111 $ ] is monochromatic , then set @xmath112 .
we continue this process obtaining @xmath113 where @xmath114 $ ] is monochromatic , @xmath115 , and @xmath116 $ ] contains only monochromatic cliques of order at most @xmath117 .
now we define a 2-coloring @xmath118 of @xmath23 . a. if @xmath119 , then @xmath120 . b. if @xmath121 and @xmath122 , then @xmath123 .
c. if @xmath124 and @xmath125 , then @xmath126 takes the opposite color to the color of @xmath114 $ ] .
d. if @xmath124 and @xmath127 , then color @xmath126 arbitrarily .
now suppose that there is a monochromatic clique @xmath128 in @xmath23 .
such a clique must contain @xmath90 .
now there are two cases to consider . if @xmath129 , then the subgraph of @xmath95 induced by @xmath130 is also a monochromatic copy of @xmath101 , a contradiction . otherwise , @xmath131 .
thus , @xmath132 and @xmath133 .
observe that @xmath134 and @xmath135 .
but this yields a contradiction @xmath136 for @xmath81 .
first for convenience we recall the definition of a hypertree . given integers @xmath40 and @xmath5 , recall that a @xmath2-graph @xmath41 of order @xmath5 with edge set @xmath42 is an _
@xmath43-tree _ , if for each @xmath44 we have @xmath45 and @xmath46 for some @xmath47 .
* proof of theorem [ thm : ub_tree ] .
* fix @xmath137 .
we are to show that @xmath138 .
recall that a _ partial steiner system @xmath139 _ is a @xmath2-graph of order @xmath140 such that each @xmath141-tuple is contained in at most one edge . due to a result of rdl
@xcite it is known that there is a constant @xmath142 such that for every @xmath143 there is an @xmath144 with the number of edges satisfying @xmath145 ( see also @xcite for similar results ) .
it is easy to observe that for @xmath146 every @xmath147-tuple is contained in at most @xmath148 edges .
fix @xmath40 .
let @xmath149 , where the constant @xmath150 is defined as @xmath151 let @xmath23 be a @xmath152 satisfying .
observe that if @xmath153 , then @xmath23 can be viewed as a complete @xmath2-graph of order @xmath140 .
clearly , @xmath154 . it remains to show that for any @xmath155 tree , @xmath156 .
degree _ of a set @xmath157 ( @xmath158 ) by @xmath159 and for @xmath160 a _ minimum ( non - zero ) @xmath43-degree _ by @xmath161 first observe that for any 2-coloring of the edges of @xmath23 , there is a monochromatic sub - hypergraph @xmath162 with @xmath163 . indeed , suppose that @xmath164 is colored with blue and red colors .
assume by symmetry that the red hypergraph @xmath165 has at least @xmath166 edges .
set @xmath167 .
if there exists @xmath168 with @xmath169 , then let @xmath170 ( we remove @xmath171 and all incident to @xmath171 edges ) .
now we repeat the process .
if there exists @xmath172 with @xmath173 , then let @xmath174 .
continue this way to obtain hypergraphs @xmath175 where either @xmath176 or @xmath177 is empty hypergraph .
but the latter can not happen , since the number of removed edges from @xmath165 is less than @xmath178 now we greedily embed @xmath52 into @xmath179 . at every step we have a connected sub - tree @xmath180 .
assume that we already embedded @xmath181 edges of @xmath52 obtaining @xmath182 .
let @xmath183 be such that @xmath184 for some @xmath185 .
observe that there is always an edge @xmath186 such that @xmath187 . indeed ,
if @xmath188 , then this is true since @xmath189 and @xmath190 and every @xmath191-tuple of vertices of @xmath162 is contained in at most one edge in @xmath162 .
otherwise , if @xmath192 , first we find a set @xmath193 such that @xmath194 and @xmath195 is contained in an edge of @xmath162 , and next apply the previous argument to @xmath195 .
thus , we can extend @xmath182 to @xmath196 , as required . as mentioned in the introduction
, it would be interesting to decide whether theorem [ thm : ub_tree ] is tight up to the hidden constant .
this is definitely the case for @xmath49 .
indeed , let @xmath52 be a @xmath2-uniform star - like tree of order @xmath5 defined as follows .
assume that @xmath197 divides @xmath198 .
@xmath52 consists of @xmath199 arms @xmath200 ( each with two edges ) : @xmath201 , where @xmath202 and all @xmath203 vertices are pairwise different ( see figure [ fig : star ] ) . with @xmath204 arms each of length 2 .
] assume that @xmath156 and color @xmath205 by red if degree ( in @xmath23 ) of every vertex in @xmath126 is less than @xmath199 ; otherwise @xmath126 is blue .
since @xmath206 and there is no red copy of @xmath52 , there must be a blue copy of @xmath52 .
every edge in such a copy has at least one vertex of degree at least @xmath199 ( in @xmath23 ) .
since @xmath52 has @xmath199 vertex disjoint edges and every edge ( in @xmath23 ) can intersect at most 3 of those disjoint edges , @xmath207
in this section we prove proposition [ pathprop ] and theorem [ thm : tightpath ] .
* proof of proposition [ pathprop ] . *
let @xmath1 be a graph satisfying @xmath208 and @xmath209 ( _ cf .
we construct a @xmath2-graph @xmath23 as follows .
replace every vertex @xmath210 by an @xmath43-tuple @xmath211 ( different for every @xmath90 ) and each @xmath212 by @xmath213 where @xmath214 are different for every edge @xmath126 , too .
thus , @xmath23 is a @xmath2-graph with @xmath215 and @xmath216 .
now color @xmath217 .
this coloring ( uniquely ) defines a coloring of @xmath218 .
since @xmath1 contains a monochromatic copy of @xmath13 , @xmath23 also contains a monochromatic copy of @xmath58 .
consequently , @xmath219 and the proof is complete .
we now turn to the main result of this section which we restate for convenience . *
[ thm : tightpath ] * [ thm : tight_path_k ] fix @xmath71 and let @xmath72 . then @xmath73 first we prove an auxiliary result . in order to do it we state some necessary notation . set @xmath220 for a graph @xmath221
let @xmath222 be the set of all cliques of order @xmath43 and let @xmath223 .
let @xmath224 and @xmath225 be a family of pairwise vertex - disjoint cliques .
define @xmath226 as the number of @xmath2-cliques of @xmath0 which @xmath227 vertices form a vertex set of some @xmath228 and the remaining vertex is from @xmath229 .
similarly , let @xmath230 be the number of @xmath2-cliques in @xmath0 which @xmath227 vertices form a vertex set of some @xmath228 and the remaining vertex is from @xmath231
. finally , let @xmath232 ( for @xmath233 ) be the number of @xmath2-cliques containing at least one vertex from @xmath234 .
[ prop : auxg_k ] let @xmath71 be an integer and let @xmath235 .
then there exists a graph @xmath221 of order @xmath5 ( for sufficiently large @xmath5 ) satisfying the following : a. for every @xmath236 , @xmath237 , and every @xmath238 , @xmath239 , vertex disjoint @xmath240-cliques such that @xmath241 we have @xmath242 [ property : ik ] b. for every @xmath233 , @xmath243 , @xmath244 [ property : iiik ] c. the total number of @xmath2-cliques satisfies @xmath245 where @xmath246 [ property : iik ] it suffices to show that the random graph @xmath247 with @xmath248 and @xmath249 satisfies _ a.a.s .
_ occurs _ asymptotically almost surely _ , or _
_ for brevity , if @xmath250 . ]
below we will use the following bounds on the tails of the binomial distribution @xmath251 ( for details , see , _
e.g. _ , @xcite ) : @xmath252 first we show that @xmath0 _ a.a.s .
_ satisfies .
fix an @xmath253 and @xmath254 with @xmath255 . observe that without loss of generality we may assume that @xmath256 .
note that @xmath257 .
thus , @xmath258 and ( applied with @xmath259 ) implies @xmath260 now we bound from above the number of all possible choices for @xmath261 and @xmath262 .
clearly we have at most @xmath263 choices for @xmath261 .
observe that the number of choices for @xmath262 can be bounded from above by the number of ways of choosing an ordered subset of vertices of size @xmath264 . indeed
, suppose that @xmath265 is such a choice .
then @xmath262 can be defined as @xmath266 .
thus we conclude that there are at most @xmath267 ways to choose @xmath261 and @xmath262 .
hence , by @xmath268 similarly , since @xmath269 , @xmath270 and since @xmath271 , @xmath272 inequality ( applied with @xmath259 ) yields @xmath273 therefore , we deduce that @xmath274 consequently , by and we get that _ a.a.s._@xmath275 for any choice of @xmath261 and @xmath262 .
this finishes the proof of . for each vertex @xmath276 ,
let @xmath277 denote the number of @xmath2-cliques of @xmath0 which contain @xmath90 .
in order to show that _ a.a.s .
_ @xmath0 also satisfies , we will first estimate @xmath277 for each @xmath276 .
the standard application of ( applied with @xmath278 and @xmath259 ) with the union bound imply that _ a.a.s .
_ the degree of every vertex @xmath279 satisfies @xmath280 the number of @xmath2-cliques which contain @xmath90 is equal to the number of @xmath240-cliques in the neighborhood of @xmath90 .
therefore , in order to show it suffices to bound the number of @xmath240-cliques in any set of size at most @xmath281 .
let @xmath282 with @xmath283 .
first we will decompose all @xmath240-tuples of @xmath284 into linear @xmath240-uniform hypergraphs @xmath285 with @xmath286 and @xmath287 for every @xmath288 .
that means that each @xmath240-tuple of @xmath284 belongs to exactly one @xmath289 and each pair of elements of @xmath284 appears in at most one @xmath240-tuple in @xmath290 .
the existence of such a decomposition follows from a more general result of pippenger and spencer @xcite ( see also @xcite ) .
let @xmath291 be the random variable that counts the number of @xmath240-tuples of @xmath289 which appear as @xmath240-cliques of @xmath0 .
observe that @xmath292 .
therefore for each @xmath181 , @xmath293 and by ( with @xmath294 ) @xmath295 consequently , the union bound over all subsets @xmath282 of size @xmath147 and over all @xmath181 for each @xmath288 implies @xmath296 since @xmath297 grows like a polynomial in @xmath5 .
therefore it follows that _ a.a.s._@xmath298 where @xmath299 in a similar way one can show that @xmath300 where @xmath301 note that equation gives the bound @xmath302 which proves part .
now we finish the proof of .
since each @xmath2-clique is counted exactly @xmath2 times , the number of @xmath2-cliques is _ a.a.s . _
at least @xmath303 it follows now from and that given a set @xmath304 , @xmath243 , the number of @xmath2-cliques of @xmath0 which intersect @xmath234 is _ a.a.s . _ at most @xmath305 finally observe that , together with the choice of @xmath150 yield that @xmath306 implying condition , as required .
now we are ready to prove main result of this section .
* proof of theorem [ thm : tightpath]*. we show that there exists a @xmath2-graph @xmath23 with@xmath307 such that any two - coloring of the edges of @xmath23 yields a monochromatic copy of @xmath308 .
let @xmath0 be a graph from proposition [ prop : auxg_k ] . set @xmath309 and let @xmath217 be the set of @xmath2-cliques in @xmath0 .
we prove that such @xmath23 is a ramsey @xmath2-graph for @xmath310 with @xmath311 , where @xmath235 .
take an arbitrary red - blue coloring of the edges of @xmath312 and assume that there is no monochromatic @xmath310 .
we will consider the following greedy _ procedure _ which at each step finds a blue tight path of length @xmath181 labeled as @xmath313 . 1 .
let @xmath314 be the _ trash _ set of @xmath240-tuples and @xmath315 be the set of _ unused _ vertices and set @xmath316 .
at any point in the process , if @xmath317 , then stop .
[ step : start_k ] ( in this step @xmath318 . ) if possible , then choose a blue edge from @xmath319 and label its vertices by @xmath320 and then set @xmath321 .
otherwise , if not possible , stop .
[ step : adding_k ] ( in this step @xmath322 . )
let @xmath323 be the labels of the last @xmath227 vertices of the constructed blue path .
if possible , select a vertex @xmath324 for which @xmath325 form a blue edge .
label @xmath89 as @xmath326 , set @xmath327 and @xmath328 .
repeat this step until no such @xmath89 can be found .
( in this step also @xmath322 . )
let @xmath323 be the labels of the last @xmath227 vertices of the constructed blue path which can not be extended in a sense described in step .
remove these @xmath227 vertices from the path and set @xmath329 and @xmath330 . after this removal
there are two possibilities : a. if @xmath331 , then put back @xmath332 to @xmath319 ( i.e. @xmath333 ) , set @xmath316 , and return to step ; b. otherwise , return to step .
+ this procedure will terminate under two circumstances : either @xmath317 or no blue edge can be found in step .
first let us consider the case when @xmath317 , that means , there are @xmath334 vertex disjoint @xmath240-tuples in @xmath262 .
denote by @xmath261 the vertex set of the blue path which was obtained when @xmath317 .
clearly , @xmath335 , otherwise there would be a blue @xmath310 .
we are going to apply proposition [ prop : auxg_k ] with sets @xmath261 and @xmath262 .
notice that every edge of @xmath23 which contains a @xmath240-tuple from @xmath262 and the remaining vertex from @xmath336 must be colored red .
( this is because for a @xmath240-tuple to end up in @xmath262 , there must have been no vertex @xmath89 in step that could extend the blue path . )
it also follows from step that each @xmath240-tuple in @xmath262 is contained in at least one blue edge .
thus , proposition [ prop : auxg_k ] implies that @xmath337 .
that means that the number of red edges which contain a @xmath240-tuple from @xmath262 and the remaining vertex from @xmath319 is at least @xmath78 times the number of blue edges with a @xmath240-tuple from @xmath262 .
now remove all the blue edges from @xmath23 which contain a @xmath240-tuple from @xmath262 and denote such @xmath2-graph by @xmath338 . perform the above procedure on @xmath338 .
this will generate a new trash set @xmath339 .
observe that @xmath340 , since every edge of @xmath338 which contains a @xmath240-tuple from @xmath262 must be red .
again , if @xmath341 , then we use the same argument as above to find that the number of red edges in @xmath338 which contain a @xmath240-tuple from @xmath339 and the remaining vertex from @xmath319 is at least @xmath78 times the number of blue edges in @xmath338 with a @xmath240-tuple from @xmath339 .
indeed , we can again apply the inequality from proposition .
this is because @xmath342 is smaller than the number of all blue edges in @xmath23 containing a @xmath240-tuple from @xmath339 , while ( since we do not remove red edges ) @xmath343 remains same in both @xmath338 and @xmath23 . now remove the blue edges from @xmath338 which contain a @xmath240-tuple from @xmath339 obtaining a @xmath2-graph @xmath344 .
keep repeating the procedure until it is no longer possible . at some point
, we will run out of blue edges in @xmath345 for some @xmath346 , and the procedure will terminate prematurely in step . in this case
@xmath347 , @xmath348 and @xmath319 has no blue edges .
however , there still may be some blue edges which contain a vertex from @xmath349 .
proposition [ prop : auxg_k ] ( applied for @xmath350 ) implies that the number of such edges is at most @xmath244 let @xmath351 and @xmath352 be the numbers corresponding to @xmath226 and @xmath230 obtained at the end of the procedure applied to @xmath353 .
thus , @xmath354 for each @xmath355 .
let @xmath356 and @xmath357 denote the number of red and blue edges in @xmath23 .
observe that @xmath358 furthermore , since all sets @xmath359 are mutually disjoint , each red edge in @xmath23 containing a @xmath240-tuple from some @xmath359 can be only counted at most @xmath2 times .
thus , @xmath360 consequently , by and , we get @xmath361 and so @xmath362 the conclusion is that there are more red edges than there are blue edges in @xmath23 .
if we reverse the procedure and look for a red path instead of a blue one , we will conclude that there are more blue edges than red edges .
since these two statements contradict each other , the only way to avoid both statements is if a monochromatic path exists .
in this section we prove theorem [ thm : bounddegree ] , which states that hypergraphs with bounded degree can have nonlinear size - ramsey numbers .
* proof of theorem [ thm : bounddegree ] .
* we modify an idea from rdl and szemerdi @xcite . for simplicity
we only present a proof for @xmath363 , which can easily be generalized to @xmath35 .
the hypergraph @xmath20 will be constructed as the vertex disjoint union of graphs @xmath364 each of which is a tree with a path added on its leaves .
next we will describe the details of such construction .
let @xmath366 consider a binary 3-tree @xmath367 on @xmath368 vertices rooted at vertex @xmath369 ( see figure [ fig : b ] ) .
denote by @xmath370 the set of all its leafs .
call the edge containing @xmath369 the _ root edge_. observe that latexmath:[\[\label{eq : v(b ) } let @xmath373 by an automorphism of @xmath262 . since the root edge @xmath126 is the unique edge with exactly one vertex of degree 1 , @xmath374 .
the other two vertices of @xmath126 are permuted by @xmath373 .
consequently , @xmath373 permutes two vertices of every other edge .
hence , it is easy to observe that the order of the automorphism group of @xmath262 satisfies @xmath375 now consider a tight path @xmath377 of length @xmath378 placed on the leaves @xmath370 in an arbitrary order . considering labeled vertices of @xmath370
there are clearly @xmath379 such paths .
label them by @xmath200 for @xmath380 .
let @xmath359 be vertex disjoint copies of @xmath262 and @xmath381 , where @xmath382 .
let @xmath373 be an isomorphism between @xmath364 and @xmath383 .
since the only vertices of degree 4 are on paths @xmath200 and @xmath384 , @xmath385 .
thus , @xmath386 and so @xmath359 and @xmath387 are isomorphic .
thus , the number of pairwise non - isomorphic @xmath364 s is at least clearly , @xmath392 .
furthermore , by , we get @xmath393 moreover , @xmath394 and the independence number of @xmath20 satisfies @xmath395 indeed , let @xmath396 be an independent set of size @xmath397 .
we estimate the number of edges @xmath398 between sets @xmath399 and @xmath400 .
first observe that @xmath401 next , since each triple between @xmath399 and @xmath400 intersects one of the partition classes on 2 vertices and @xmath402 , @xmath403 this implies that @xmath404 and so .
recall that @xmath20 consists of @xmath413 pairwise non - isomorphic copies of @xmath364 .
we estimate the number of copies of @xmath364 s contained in a sub - hypergraph induced by @xmath414 .
first fix an edge @xmath126 in @xmath415 $ ] and count the number of copies of @xmath364 s for which @xmath126 is a root edge .
since @xmath416 for each @xmath417 , we get that this number is at most @xmath418 where the factor 3 counts the number of choices for the root vertex , the next factors count the number of possible @xmath359 s with @xmath126 as a root , and the last factor counts the number of paths on the set of leafs . thus , there is an @xmath419 such that @xmath420 appears in @xmath415 $ ] at most @xmath421 times .
denote by @xmath162 the sub - hypergraph consisting of root edges from all copies of @xmath420 in @xmath415 $ ] .
thus , @xmath422 color edges in @xmath162 together with edges incident to @xmath423 blue ; otherwise red .
clearly , there is no red copy of @xmath20 , since there is no red copy of @xmath420 .
moreover , there is no blue copy of @xmath20 , since every blue sub - hypergraph of order @xmath424 has an independent set of size at least @xmath425 which is strictly bigger than @xmath426 ( _ cf . _ ) .
r. faudree and r. schelp , _ a survey of results on the size ramsey number _ , paul erds and his mathematics , ii ( budapest , 1999 ) , bolyai soc .
11 , jnos bolyai math .
budapest , 2002 , pp . | the size - ramsey number of a graph @xmath0 is the minimum number of edges in a graph @xmath1 such that every 2-edge - coloring of @xmath1 yields a monochromatic copy of @xmath0 .
size - ramsey numbers of graphs have been studied for almost 40 years with particular focus on the case of trees and bounded degree graphs .
we initiate the study of size - ramsey numbers for @xmath2-uniform hypergraphs .
analogous to the graph case , we consider the size - ramsey number of cliques , paths , trees , and bounded degree hypergraphs .
our results suggest that size - ramsey numbers for hypergraphs are extremely difficult to determine , and many open problems remain . |
in previous work @xcite we have investigated quantum gravity in @xmath0 dimensions with negative cosmological constant on the torus , using an approach involving quantum holonomy matrices .
this followed on from earlier work by one of us with regge and zertuche @xcite based on the traces of the holonomies . in @xcite we focused on the quantum geometry that arises from introducing a constant quantum connection , from which the holonomy matrices and their traces are obtained ( in the sector where these matrices are diagonal ) .
some interesting features emerged , in particular a quantum version of the well - known goldman bracket for loops on a surface . in the present article
we describe some new developments in the understanding of this quantum geometrical picture .
the classical action of @xmath0 gravity with negative cosmological constant @xmath2 , in the dreibein formulation , was related by witten @xcite to chern - simons theory for the gauge group @xmath3 .
thus the classical phase space of this model of @xmath0 gravity corresponds to the moduli space of flat @xmath3 connections on the spatial manifold , which we take to be the torus @xmath4 . treating each @xmath5 factor
separately , we are then led to consider pairs of commuting @xmath5 matrices ( up to simultaneous conjugation by an element of @xmath5 ) , representing the holonomies of the flat connection along two generating cycles of the torus ( up to gauge transformation ) .
note that the fundamental group of the torus is generated by two cycles , with a single relation saying that the generators commute . for a detailed description of the moduli space of flat @xmath5 connections on the torus ,
see @xcite . the diagonal sector of this moduli space
can be parametrised by constant connections @xcite on the torus of the form a= ( r_1 dx + r_2 dy ) ( clcr 1 & 0 + 0 & -1 ) [ conn ] where @xmath6 are coordinates on the torus ( with periodicity @xmath7 ) , or on its covering space @xmath8 .
the connections are constant in the sense that the functions @xmath9 are constant functions ( independent of @xmath10 and @xmath11 ) . after integration ,
this yields the holonomies : u_i = _ _
i a = ( clcre^r_i & 0 + 0 & e^- r_i ) i=1,2 [ hol1 ] where @xmath12 are generating cycles of the fundamental group of the torus
. in chern - simons theory components of the connection @xmath13 have non - trivial poisson brackets amongst themselves . at the level of the connection components @xmath14
this translates into the poisson bracket \ { r_1 , r_2 } = - , which is quantised by @xmath15 = \frac{i\hbar \sqrt{-\lambda}}{4}. \label{rcommrel}\ ] ] therefore we will consider a quantised phase space parametrised by constant quantum connections of the form : = ( _ 1 dx + _ 2 dy ) ( ll 1 & 0 + 0 & -1 ) .
[ quconn ] after integration along the generating cycles the wilson loops , i.e. traces of the holonomies , of this connection are precisely the variables used by nelson and regge @xcite .
this connection also yields the description in terms of quantum holonomy matrices of nelson and picken @xcite .
we remark also that the relation between the parameters @xmath9 ( a pair for each @xmath5 factor ) and the adm variables for @xmath0 gravity has been elucidated in @xcite .
our purpose here is to further explore the quantum geometry arising from the assignment of a quantum matrix ( or its trace ) to a general class of loops on the torus .
it is convenient to identify loops on the torus @xmath16 with paths on its covering space @xmath1 .
we therefore consider all loops on the torus as being represented by piecewise - linear ( pl ) paths between integer points on the @xmath17 plane , and work with this description , bearing in mind that paths upstairs i.e. in @xmath1 represent loops downstairs i.e. on the torus @xmath16 .
in particular , a natural class of paths @xmath18 to consider are those straight paths that start at the origin and end at an integer point @xmath19 . they will be denoted @xmath20 . using the quantum connection we assign a quantum matrix to any such straight path by _
( m , n)= _ ( m , n ) = ( ll e^m_1 + n_2 & 0 + 0 & e^-m_1 -n_2 ) . [ umn ] this assignment is straightforwardly extended to any pl path between integer points by assigning a quantum matrix to each linear segment of the path , in an analogous fashion to , and multiplying the matrices in the same order as the segments appear along the path .
this prescription obviously coincides with the general relation : p_p = p _ p . [ hol ] where @xmath21 denotes path - ordering . in the upstairs picture , two homotopic loops on the torus
are represented by two pl paths on the plane , @xmath22 , with the same integer starting point and the same integer endpoint .
it was shown in @xcite that the following relationship holds for the respective quantum matrices : @xmath23 where @xmath24 denotes the signed area enclosed between the paths @xmath25 and @xmath26 , and @xmath27 the signed area between two such pl paths is defined as follows : for any finite region @xmath28 enclosed by @xmath25 and @xmath26 , if the boundary of @xmath28 consists of oriented segments of @xmath25 and @xmath29 ( the orientation reversal of @xmath26 ) , and as such is globally oriented in the positive ( anticlockwise ) , or negative ( clockwise ) sense , this gives a contribution of @xmath30 , or @xmath31 respectively , to the signed sum @xmath24 ( otherwise the contribution is zero ) .
see figure [ signedarea ] .
the signed area between @xmath25 and @xmath26 is the area of the the horizontally shaded region minus the area of the vertically shaded region ] we note for the discussion in the next sections that the wilson loops , or traces of the holonomies , are related by the same `` area phases '' , i.e. @xmath32 where @xmath33 . in @xcite this relation between the holonomy matrices for homotopic loops / paths
was interpreted as a @xmath34-deformed surface group representation , i.e. a deformation of a representation of the fundamental group in terms of matrices .
there we used the fact that signed area between paths has the following two properties : @xmath35 where @xmath36 denotes the concatenation of paths @xmath25 and @xmath26 .
another perspective on this comes from recent work by one of us with martins @xcite , where the notion of @xmath37-dimensional holonomy was explored . in the present context , @xmath37-dimensional holonomy
may be described as the assignment of an element of a group @xmath38 to each of a pair of homotopic loops , and simultaneously the assignment of an element of a group @xmath39 to the homotopy between the two loops , subject to certain consistency relations . amongst these relations
, there is the requirement that the elements assigned to the homotopies behave well under vertical and horizontal composition , which translates precisely to the properties of signed area as in .
we note also that a crucial ingredient in the construction of @xcite is a connection @xmath37-form with values in the lie algebra of @xmath39 , and indeed in the present context there is a natural @xmath37-form , namely the non - vanishing curvature of the connection ( non - vanishing because of the non - commutativity of the components @xmath40 and @xmath41 , as pointed out in @xcite ) .
thus there are strong hints that the quantum holonomies of @xmath0 gravity may be interpreted in terms of @xmath37-dimensional holonomy .
when analysing the behaviour of the wilson loops @xmath33 for the quantum connection in @xcite a link with the goldman bracket @xcite emerged .
this bracket is a poisson bracket for functions @xmath42 , defined on homotopy classes of loops on a surface , which for @xmath43 takes the following form ( see @xcite thm . 3.14 , 3.15 and remark ( 2 ) , p. 284 ) : \{t(_1 ) , t(_2 ) } = _ s _ 1 _ 2 ( _ 1,_2,s)(t(_1s_2 ) - t(_1s_2 ^ -1 ) ) .
[ gold ] here @xmath44 denotes the set of ( transversal ) intersection points of @xmath45 and @xmath46 and @xmath47 is the intersection number of @xmath45 and @xmath46 at the intersection point @xmath48 , i.e. ( for simple intersections ) @xmath49 if the angle between the tangent vector of @xmath45 at @xmath48 and the tangent vector of @xmath46 at @xmath48 is between @xmath50 and @xmath51 degrees , and @xmath52 if it is between @xmath51 and @xmath53 degrees .
finally @xmath54 and @xmath55 denote loops which are _ rerouted _ at the intersection point @xmath48 , meaning that we follow @xmath45 as far as the intersection point @xmath48 , then follow @xmath46 ( or the inverse loop @xmath56 ) from @xmath48 back to @xmath48 , and finally proceed along @xmath45 from @xmath48 back to the starting point of @xmath45 . for a fuller discussion and
examples see section 5 of @xcite .
we found that the wilson loops @xmath57 , for @xmath20 a straight path corresponding to a loop on the torus , satisfied a quantum version of the goldman relation .
in fact two different quantisations emerged , a straightforward one , and a refined one , which we will now describe .
first , for straight paths @xmath58 and @xmath59 , the goldman bracket takes the form : \{t(m , n ) , t(s , t ) } = ( mt - ns ) ( t(m+s , n+t)- t(m - s , n - t ) ) .
[ straightforward ] here @xmath60 is the total intersection number between @xmath25 and @xmath26 , and this factor appears because there are effectively @xmath60 simple intersection points and the corresponding reroutings @xmath61 are all homotopic to the straight path @xmath62 , with an analogous statement for the negative reroutings @xmath63 . by a simple direct calculation ,
it was shown in @xcite that the wilson loops satisfy : @xmath64= ( q^{\frac{mt - ns}{2}}-q^{-\frac{mt - ns}{2 } } ) \left(\hat{t}(m+s , n+t ) - \hat{t}(m - s , n - t)\right ) \label{qgb1}\ ] ] i.e. a quantisation of , with the total intersection number @xmath60 replaced by a quantum total intersection number ( the first factor on the r.h.s . of ) .
we will call this the straightfward quantisation of the goldman bracket .
a more refined quantisation was also obtained in @xcite , where each rerouting appears as a separate term , and the area phases of these different but homotopic reroutings are taken into acccount .
this takes the following form : = _ s p_1
p_2 ( q^(p_1,p_2,s ) - 1)(p_1sp_2 ) + ( q^-(p_1,p_2,s ) - 1)(p_1sp_2 ^ -1 ) [ qgold ] which quantises ( with loops @xmath65 substituted by paths @xmath18 ) by replacing the intersection numbers @xmath66 by quantum intersection numbers @xmath67 .
an example is given by the choice @xmath68 , @xmath69 , where the refined bracket is given by = _
s = p , r , q ( q^-1 -1 ) ( ( 1,2)s(2,1 ) ) + ( q-1 ) ( ( 1,2)s(-2,-1 ) ) , [ qgoldexp ] where the terms on the r.h.s .
come from the positive and negative reroutings at three intersection points denoted @xmath70 , see figure [ p20a ] .
we then calculate the area phases of each rerouting relative to the corresponding straight path @xmath71 or @xmath72 , and using we obtain : @xmath73 & = & ( q^{-1}-1)(q^{\frac{3}{2 } } + q^{\frac{1}{2 } } + q^{-\frac{1}{2 } } ) \hat{t}(3,3 ) + ( q-1 ) ( q^{-\frac{3}{2 } } + q^{-\frac{1}{2 } } + q^{\frac{1}{2 } } ) \hat{t}(-1,1 ) \nonumber \\ & = & ( q^{-\frac{3}{2 } } - q^{\frac{3}{2 } } ) ( \hat{t}(3,3 ) - \hat{t}(-1,1)).\end{aligned}\ ] ] i.e. the straightforward form of this commutator . and
@xmath74 for @xmath75 , title="fig:",height=151 ] and @xmath74 for @xmath75 , title="fig:",height=151 ] in the next section we describe some new insights which allow a better understanding of intersections and reroutings , and prepare the way to address some outstanding issues concerning the two quantisations and ( see the final section ) .
in @xcite we observed that the paths that appear after rerouting two straight paths at an intersection point all pass through an integer point in the parallelogram spanned by the two paths . here
we will show from a new viewpoint that this is necessarily the case , and we shall see that it is also a very useful way to examine the nature of the reroutings .
we look at reroutings of the type @xmath76 , where @xmath25 and @xmath26 are both straight paths in @xmath8 starting at the origin , corresponding to loops on the torus @xmath77 .
this means that we follow @xmath25 up to the intersection point @xmath28 , reroute along @xmath26 starting at the point @xmath28 until @xmath26 returns to @xmath28 , and complete the remainder of the first path @xmath25 from @xmath28 back to the origin .
note that the intersection point @xmath28 may occur at the origin of @xmath8 ; in that case we follow @xmath26 straight away . instead of studying the intersections between @xmath25 and @xmath26 by reducing both paths to a fundamental domain , as we did in @xcite , we work directly in @xmath8 .
fix the path @xmath25 and parallel translate @xmath26 to start at a new integer point , denoted @xmath78 , in such a way that it intersects @xmath25 at @xmath28 .
of course , all integer points are the same when projecting down to the torus , but this shift of @xmath26 gives a very clear picture of where @xmath28 is located along both paths simultaneously .
we introduce the following notation for paths on the plane : @xmath79 denotes the subpath of the path @xmath18 going from the point @xmath13 on @xmath18 to the point @xmath80 on @xmath18 . also for any path @xmath18 , let @xmath81 denote the integer endpoint of @xmath18 .
note that for straight paths @xmath20 , the notation for the endpoint and for the path itself coincide , i.e. @xmath82 , but when the path is non - straight , there is a distinction between @xmath18 and @xmath81 .
this latter situation will occur when we try and extend the refined bracket to non - straight paths - see the outstanding issues listed in the final section .
now write @xmath83 , to indicate how @xmath25 is divided into two segments by the intersection point @xmath28 . here
@xmath25 starts at the origin @xmath50 , but the parallel translated @xmath26 starts at the integer point @xmath78 , so we require a notation to describe the shift or parallel translation of paths to start at different points : @xmath84 will denote the path @xmath18 parallel translated to start at @xmath78 instead of its original starting point .
( of course , projected down to the torus , the original @xmath18 going from the origin to an integer point , and the shifted path @xmath84 give rise to identical loops ) . in this way ,
we can write down an explicit algebraic expression for the rerouting @xmath76 represented as a path in the plane ( see figure [ rerouting ] ) : @xmath85 where the start and endpoints @xmath78 and @xmath86 of @xmath87 are related by : @xmath88 . the rerouting @xmath76 ] this detailed notation for rerouted paths should prove useful in the future for manipulating paths algebraically and making precise general statements
. however the approach in this section also makes it very clear why intersections occur and how they give rise to reroutings . in concrete terms , for an intersection to occur , we need either a starting point @xmath78 for @xmath87 such that @xmath87 intersects @xmath25 in a point @xmath28 ( which may be the origin , but not the endpoint of @xmath25 ) , or equivalently we need an endpoint @xmath86 such that the appropriate @xmath87 ending in @xmath86 intersects @xmath25 in a point @xmath28 as before . looking first at the possible starting points @xmath78 , we see that they are the integer points lying in a `` pre - parallelogram '' with vertices @xmath89 , @xmath90 , @xmath91 and the origin @xmath50
. see figure [ preparallelogram ] .
here we should exclude any integer points lying on the edge between @xmath90 and @xmath91 , since the corresponding paths @xmath87 intersect @xmath25 at its endpoint , and we should also exclude any integer points lying on the edge between @xmath89 and @xmath90 , to avoid double counting , as we are including the integer points lying along the edge between @xmath50 and @xmath91 . the parallelogram and pre - parallelogram ( dotted line ) for @xmath25 and @xmath26 , showing some of the integer points ( black dots ) and the corresponding parallel - translated copies of @xmath26 ] equivalently , we can look at the endpoints @xmath86 lying inside the parallelogram generated by @xmath25 and @xmath26 , i.e. with vertices @xmath50 , @xmath91 , @xmath93 and @xmath94 . here
we will exclude the integer points lying on the edge between @xmath91 and @xmath95 and those lying along the edge between @xmath50 and @xmath91 ( these endpoints correspond to the starting points we excluded from the pre - parallelogram ) .
from both perspectives we can readily verify that the number of intersections is correct , since the total intersection number is : @xmath96 the total number of intersection points ( counting multiplicities ) , all of which have the same sign of the intersection number , is therefore the modulus of @xmath97 , i.e. in geometric terms , the area of the parallelogram or the pre - parallelogram . in turn , this area is given by a classical theorem of pick @xcite , which states that the area @xmath98 of a planar polygon @xmath99 with vertices at integer points of the plane is given in terms of the number of interior integer points @xmath100 and the number of boundary integer points @xmath101 as follows : @xmath102 this value is evidently the same as the number of integer starting points @xmath78 in the pre - parallelogram , or equivalently the number of integer end points @xmath86 in the parallelogram generated by @xmath25 and @xmath26 , since we have excluded two edges , i.e. half the integer points lying along the interior of the edges and three out of four of the vertices ( this is accounted for by the @xmath52 in the formula , since @xmath103 , the single remaining vertex ) .
a convenient feature of this approach is that it handles reducible paths in a natural manner .
a reducible path is one such that @xmath104 , where @xmath105 are all integers , with @xmath106 .
if @xmath26 is reducible , this means that there are integer points along @xmath26 other than the endpoints . in terms of the pre - parallelogram analysis
, there will be @xmath107 integer starting points @xmath78 for each intersection point @xmath28 along @xmath25 .
the intersection point @xmath28 therefore has intersection number @xmath108 and the rerouting at @xmath28 will appear with multiplicity @xmath107 .
these issues were briefly touched upon at the end of the penultimate section of @xcite .
a final point that emerges from this discussion is that , as we observed in @xcite , obviously any rerouting must pass through one of the integer points inside the parallelogram generated by @xmath25 and @xmath26 ( namely the appropriate endpoint @xmath86 ) .
we have presented a new perspective on intersections and reroutings which holds promise of further progress on the following outstanding and interrelated issues : * the quantum intersection numbers which naturally appear in the two quantisations and have a different appearance , and in particular they are symmetric , for the straightforward bracket , under the interchange @xmath109 .
this is not the case for the refined bracket . *
although there are strong arguments suggesting that , for straight paths , the refined bracket and the straightforward bracket are equivalent , a full proof is still lacking . *
part of the previous problem is dealing with reducible paths , i.e. paths such that @xmath104 , where @xmath105 are all integers , with @xmath106 .
* we would like to extend the refined bracket beyond straight paths on the left to a more general class of paths , and ultimately get a definition of the refined bracket which closes on the class of paths . *
the antisymmetry of the refined bracket for two straight paths is not manifest .
* we must prove the jacobi identity for the refined bracket extended to a wider class of paths .
in particular , the tools are now ready to start analysing the refined bracket when the paths on the l.h.s .
are no longer straight . in the first instance
we hope to address a phenomenon that occurred when we performed direct calculations to see how the refined bracket might extend to non - straight paths : we encountered some unexpected extra phases when the second path @xmath26 in the rerouting @xmath76 was non - straight , and the first path @xmath25 was straight .
these difficulties did not occur when the first path was non - straight and the second straight , i.e. the discrepancy was clearly related to the fact that the middle part of the rerouting was non - straight . in future work
we also hope to explore in more detail the link with @xmath37-dimensional holonomy @xcite , as discussed at the end of section 2 , and gain further understanding of the elegant quantum geometry that emerges through the quantised goldman bracket , relating it e.g. to noncommutative geometry @xcite or to other quantisations of the goldman bracket @xcite .
this work was supported by the istituto nazionale di fisica nucleare ( infn ) of italy , iniziativa specifica fi41 , the italian ministero delluniversit e della ricerca scientifica e tecnologica ( miur ) , and the _ programa operacional cincia e inovao 2010 _ , project number poci / mat/60352/2004 , financed by the _ fundao para a cincia e a tecnologia _ ( fct ) and cofinanced by the european community fund feder . | in the context of quantum gravity for spacetimes of dimension @xmath0 , we describe progress in the construction of a quantum goldman bracket for intersecting loops on surfaces . using piecewise linear paths in @xmath1
( representing loops on the spatial manifold , i.e. the torus ) and a quantum connection with noncommuting components , we review how holonomies and wilson loops for two homotopic paths are related by phases in terms of the signed area between them . paths rerouted at intersection points with other paths occur on the r.h.s . of the goldman bracket . to better understand their nature we introduce the concept of integer points inside the parallelogram spanned by two intersecting paths , and
show that the rerouted paths must necessarily pass through these integer points .
+ + pacs numbers : 04.60.kz , 02.20.uw + mathematics subject classification : 83c45 |
morphology is one of the main characteristics of galaxies , and the morphological classification has been central to many advances in the picture of galaxy formation and evolution .
different correlations between morphology and other galaxy properties have been studied , including the relation with stellar mass ( e.g. , * ? ? ?
* ) , colour ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , luminosity ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , environment ( e.g. , * ? ? ?
* ) , black hole mass ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , nuclear activity ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , and x - ray properties ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
+ the methods of morphological classification of galaxies can be separated into three groups : i ) visual ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , for classifying the nearby and well resolved galaxies , ii ) parametric , based on the galaxy physical ( e.g. * ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
? * ; * ? ? ?
* ; * ? ? ?
* ) and mathematical parameters , where assuming an analytic model for fitting the galaxy ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , they perform the decomposition of well resolved galaxies and provide the properties of different structures , and finally , iii ) non - parametric , which does not assume any particular analytic model .
+ the non - parametric methods are based on measuring the different galaxy quantities that correlate with the morphological types , i.e colours ( e.g. * ? ? ?
* ) , spectral properties ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , or light distribution ( e.g. * ? ?
? * ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the non - parametric methods are less time - consuming in comparison with other methods , and can provide an easy and fast separation between regular and irregular ( or disturbed ) sources and/or early- and late - type galaxies , down to intermediate redshifts ( @xmath01.5 ) , or higher if dealing with space based , or good seeing ground - based data .
over the past years different morphological diagrams have been applied , relating galaxy light concentration and asymmetry or smoothness parameters either to classify galaxies @xcite , or to select merger candidates @xcite .
however , some of the previous works showed that these parameters can depend significantly on e.g. the aperture definition for measuring the galaxy flux / radius @xcite , and/or on the resolution and signal - to - noise ratio ( hereafter s / n ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
different trends of these parameters can be observed in relation to the data quality , spatial resolution and depth . since a reliable morphological classification is essential for galaxy formation and evolution studies , it is crucial to disentangle how strong is this impact on each morphological parameter when dealing with different data sets . in most of previous works
this observational bias was analysed either for a particular parameter , either for a particular observational condition ( survey ) , or for a particular sample of galaxies . in this work we present , for the first time ,
a systematic study , where in a consistent way the analysis was carried out for all parameters , on large sample of galaxies , and in relation with different observational conditions .
+ we analysed the observational bias that might affect each of the six commonly used morphological parameters : abraham concentration index , gini coefficient , conselice - bershady concentration index , m20 moment of light , asymmetry index , and smoothness .
we studied how this bias depends on the spatial resolution and magnitude / redshift distributions of galaxies , and how it affects the diagnostic diagrams used to classify galaxies .
we used a visually classified sample of local galaxies and simulated them to map the observational conditions of three different ground- and space - based deep surveys : the advanced large homogeneous area medium band redshift astronomical survey ( * ? ? ? * alhambra ) , the subaru / xmm-_newton _ deep survey ( * ? ? ?
* sxds ) , and the cosmos evolution survey ( * ? ? ?
* cosmos ) . with this analysis
our main goal is to measure a set of morphological parameters in the real ( local ) conditions and in the simulated conditions of non - local surveys in order to study how the spatial resolution and data depth affect each parameter and the commonly used morphological diagrams .
we emphasize that is out of our aim to classify the galaxies in the three selected non - local surveys , neither reclassify local galaxies once they were scaled to map the conditions of alhambra , sxds , and cosmos .
comparing the parameters of local galaxies measured before and after shifting them in redshift and magnitude , we were able to put constraints on the main diagrams , observing how the position and the shape of the regions typical of early- and late - type galaxies change in the local and non - local conditions .
moreover , analysing the morphological diagrams in ground - based and space - based surveys , and at different magnitude cuts , we quantified how strong is the impact from spatial resolution and survey depth , respectively .
the results obtained in this work can be applied to surveys similar to alhambra , sxds , and cosmos .
+ the paper is organized as follows : in sec .
[ sec_data ] we described the used local and high - redshift samples and the corresponding surveys . the applied methodology is described in sec .
[ sec_method ] .
the results obtained for local and simulated galaxies are showed in sec .
[ sec_results ] . in sec .
[ sec_discussion ] we discussed how the impact from data depth and spatial resolution affects the morphological parameters and commonly used diagnostic diagrams for galaxy classification . finally , we summarised our results in sec . [ sec_summary_conclusions ]
+ we assume the following cosmological parameters throughout the paper : @xmath1=0.7 , @xmath2=0.3 , and h@xmath3=70 km s@xmath4 mpc@xmath4 . unless otherwise specified , all magnitudes are given in the ab system @xcite .
to test how the observational bias affects the analysed morphological parameters , we used a sample of 3000 local galaxies at 0.01@[email protected] ( with a mean redshift of 0.04 ) , observed in the sloan digital sky survey ( sdss ) data release 4 ( dr4 ) down to an apparent extinction - corrected magnitude of @xmath6@xmath716 , and visually classified by nair & abraham ( 2010 ; hereafter n&a ) , using the @xmath6 and @xmath8 bands .
the number of local galaxies is selected as a compromise between the computing time and the classification accuracy , since the computing time to train the support vector machine ( svm ) through the galsvm code used in this work ( see sec .
[ sec_method ] ) is totally dependent and very sensitive to the size of the training data set @xcite .
the galaxies were selected randomly out of @xmath014,000 sources contained in the n&a catalogue , making sure that the selected sub - sample is representative in terms of the general properties of the whole data set ( see * ? ? ? * and their fig .
3 ) : @xmath6 band magnitude , redshift , @xmath6-@xmath8 colour , morphological classification , and inclination in the case of the late - type galaxies .
on the other hand , figure [ fig_local_distgeneral ] shows the redshift , g - band magnitude , and n&a morphological classification ( t - type in their work ) distributions of the selected local sample .
as can be seen from the t - type histogram , the selected sample occupies all range of morphologies , from elliptical to irregular galaxies . in this work
, all analyses were performed dividing galaxies into three morphological groups : + * 1 ) early - type galaxies ( hereafter et ) * - with t - type@xmath50 , including elliptical ( c0 , e0 , e+ ) , lenticular ( s0- , s0 , s0 + ) and s0/a galaxies from the n&a classification , + * 2 ) early spirals ( hereafter lt_et ) * - with 0@xmath7t - type@xmath54 , including sa , sab , sb , and sbc , and + * 3 ) late spirals and irregular galaxies ( hereafter lt_lt ) * - with t - type@xmath94 , including sc , scd , sd , sdm , sm , and i m galaxies . + we stress that throughout this work , we do not re - classify local galaxies and every time we specify the morphological type , it refers to that from n&a classification . here
we want to measure a set of the most used morphological parameters in real conditions of the local sample and in the simulated conditions of non - local surveys in order to evaluate how the spatial resolution and data depth affect each parameter .
et , lt_et , and lt_lt galaxies represent 45% , 38% , and 17% of the selected local sample , respectively , completely consistent with the whole n&a sample .
+ irregular galaxies constitute only @xmath71% of all sources in the n&a catalogue .
with such a small population , we were not able to provide reliable statistical analyses in comparison to the other two broader groups , ell / s0 and spirals .
this is the reason why we do not study irregulars separately , and divide instead all late - type galaxies into two subgroups : earlier and later . in our lt_lt defined group , galaxies classified as irr by n&a
form only 2% of the whole population , being therefore insignificant for affecting the results represented in this paper .
+ in this section we describe the properties of non - local surveys used for studying the impact from spatial resolution and data depth on the morphological classification of galaxies . since our work is statistical , to select the surveys we considered the following criteria : + - a large number of detected galaxies , either a deep survey covering a smaller area , and/or a shallower survey covering a large area , + - surveys with accurate photometry as well as accurate photometric redshift measurements available , and + - surveys with different properties in terms of spatial resolution and depth .
+ finally , as a result of these criteria , we selected three non - local surveys for our analysis : deep ground - based surveys alhambra and sxds ( alhambra being shallower than sxds , but with a large covered area and accurate photometric and photometric redshift measurements , and sxds as an example of the deepest available ground - based data ) , and cosmos as an example of deep space - based surveys .
we had to take into account the number of selected surveys , since for each of them the applied computational procedure is time consuming . in the following sections we describe each of these surveys .
their basic properties are summarised in table [ tab_surveys ] .
we can not compare directly the results obtained between the three analysed non - local surveys : first , because they have different depth , so the analysed magnitude cuts are different ( see below ) , and secondly they have different photometric information available , so the bands that we used are different as well .
we tested each survey separately , and tried to represent the obtained results depending on the properties of each one . [
cols="^,^,^,^,^,^,^",options="header " , ] * * column description : * * diagram * - commonly used morphological diagrams ( for the description of each parameter see sec .
[ sec_method_param ] ) ; * mag_cut * - for each survey , and in each diagram , the contamination levels are provided at three corresponding magnitude cuts ( see sec . [ sec_method_highz ] ) ; * cont .
et * , * cont . lt_et * , and * cont .
lt_lt * - contamination level of et , lt_et , and lt_lt simulated local galaxies , respectively , with the other two morphological types at the corresponding magnitude cuts . for intermediate- and high - redshift galaxy samples , and large data sets , typical for current ( and future )
deep photometric surveys , the application of automated non - parametric methods , that use the parameters analysed within this work , is one of the ways to obtain the information about the morphology . in some of the latest works ,
the classification based on non - parametric methods was done using a set of parameters simultaneously , therefore decreasing the impact of spatial resolution and depth ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
however , for classifications based on only one , two , or three parameters , special attention should be taken in relation to the magnitude and redshift distribution of the analysed sample , and the data quality .
+ the asym index was used in some of previous works to select mergers and disturbed galaxies @xcite .
asym is especially sensitive to the survey noise , as showed in sec .
[ sec_method_highz ] , and to both , the spatial resolution and depth , as showed in sec .
[ sec_results_morph_simul ] . in surveys similar to cosmos , for sources fainter than f814=23.0
, we again do not recommend the use of a single criterion based on asym when classifying them morphologically , since the contamination levels of 50% highest density regions of et and lt rise to above 40% . in the case of ground - based surveys ,
we do not recommend the use of this single criterion in any case ( except for maybe the brightest galaxies ) , since the contamination levels are much higher , taking the noise effect especially into account ; we suggest instead the simultaneous use of multiple diagnostic diagrams . + over the past ten years , the m20-gini diagram ( fig .
[ fig_morphdiag_m20_gini ] ) found wide applications .
it was mainly used in identifying galaxy mergers and peculiar ( disturbed ) galaxies ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) ; often it was used in the analysis where different types of galaxies , e.g. ulirgs , dust obscured galaxies , star - forming galaxies , or quiescent galaxies , were compared with control samples @xcite ; or for selecting et and lt galaxies @xcite .
+ from our results , in alhambra and sxds surveys even for the brightest analysed sources ( [email protected] and @[email protected] , respectively ) the contamination levels between the et and lt galaxies in m20-gini diagram are about 30% . in cosmos
however , at the lowest magnitude cut [email protected] ( z@xmath111 ) the contamination is insignificant ( @xmath1110% , except for the youngest spiral / irregular galaxies ) , but already at the second magnitude cut [email protected] ( [email protected] ) the contamination is @xmath920% .
we represented in fig .
[ fig_morphdiag_m20_gini ] the limits established by @xcite to separate between normal et and lt galaxies and major mergers , in the local universe ( dotted lines ) , and at 0.2@[email protected] ( solid lines ) , using the all - wavelength extended groth strip international survey ( aegis ) hst / acs data .
the solid - lines criterion ( non - local sample ) was later applied by @xcite to [email protected] sources to select major mergers , reporting a population of @xmath015% of misclassified sources , and was used in other works , as mentioned above . from the established limits we can see again
how sensitive the diagram is to spatial resolution and data depth , changing significantly between the ground- and space - based surveys .
while in cosmos the contamination by normal galaxies in the region assigned for major mergers , does not exceed the values measured by the authors ( even at higher redshifts ) , in the analysed ground - based surveys the contamination can be @xmath970% .
+ when applying the m20-gini diagram to classify galaxies and select mergers , we highly recommend to use it always with additional criteria when dealing with ground - based data , among which : a combination with other morphological diagrams , a careful selection of the control sample and evaluation of the diagram effectiveness , and/or application of probabilistic approaches . in space - based surveys like cosmos to distinguish between the et and lt sources additional criteria should be taken into account at least for galaxies fainter than f814=23.0 .
+ finally , different combinations of 2 - 3 parameters analysed in the previous sections , were used in many studies to distinguish between the et and lt galaxies , and also to study the structure of galaxies with other properties , like sfr , mass , environment , nuclear activity , etc ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the combination between the concentration parameters ( cabr and/or ccon ) and asym was also used for selecting et and lt sources @xcite , while in some cases the combination between the gini and ccon was used to study the host properties of agn and quasars ( e.g. * ? ? ?
the ccon - asym - smooth diagram was used in classifying both galaxies and mergers @xcite .
we found smooth to be especially sensitive to both spatial resolution and depth , and the most unstable parameter of all analysed .
moreover , together with asym , we also found it to be more sensitive to noise , as showed in sec .
[ sec_method_highz ] .
as already mentioned , we do not recommend the use of these two parameters in the ground - based surveys if the noise effects are not taken into account carefully . in surveys similar to cosmos
, they can be used in combination with other parameters but for sources brighter than f814=24.0 .
the combination of only asym and smooth gives high contamination levels even in the case of cosmos ( as showed in fig .
[ fig_morphdiag_cosmos_3plots_twomagbins ] and table [ tab_cont_alh_sxds_cosmos ] ) . + we suggest the consultation of table [ tab_cont_alh_sxds_cosmos ] and fig .
[ fig_morphdiag_ccon_cabr ] to [ fig_morphdiag_cosmos_3plots_twomagbins ] when using diagnostic diagrams in morphological studies , especially if they are based on 2 - 3 parameters .
we discussed in sec .
[ sec_results_morphdiagram ] the effectiveness of a particular diagnostic diagram in the ground - based surveys and noted that in cases similar to both alhambra and sxds , special care has to be taken into account when classifying galaxies at all magnitude cuts ( redshifts ) .
our suggestion when using ground - based data is to avoid the diagnostic diagrams based on 2 - 3 parameters , and to apply instead all morphological parameters simultaneously , and to use the statistical approaches providing a probability for a galaxy to belong to a given morphological type ( e.g. , approaches similar to that used in the galsvm code and tested in huertas - company et al .
diagrams that include m20 moment of light showed to be more affected by spatial resolution , and less effective in separating galaxies .
moreover , the noise should be taken into account if using asym and smooth parameters .
+ in space - based surveys similar to cosmos , even at the highest magnitude cuts ( [email protected] ) , in all diagnostic diagrams the contamination levels are significantly lower than at the brightest magnitudes in the ground - based surveys .
the observational bias is insignificant for all parameters at magnitudes brighter than f814=21.0 ( except for the youngest spiral and irregular galaxies ) , as shown in sec [ sec_results_morphdiagram ] ; however when going to fainter magnitudes the contamination increases .
therefore , even when dealing with data sets similar to cosmos , we remind that eventual contamination levels are present at fainter magnitudes ; the diagnostic diagrams that show to be more effective in this case are ccon - cabr , ccon - gini , ccon - smooth , and/or cabr - asym . to apply again , when possible , different diagnostics will minimise the effect of the observational bias . + in order to merge all the analysed morphological parameters , and to provide an additional and statistical test of our previous results , a linear discriminant analysis ( lda ; as implemented by pedregosa et al .
2011 ) was performed .
the lda method allows us to find the linear combination of measured parameters that best separates two or more classes of objects or events .
we tested the combination of the four morphological parameters ( cabr , gini , ccon , and m20 ) to distinguish between et , lt_et , and lt_lt galaxies in the three analysed non - local surveys ( alhambra , sxds , and cosmos ) . knowing that asym and smooth are found to be affected by noise
, we performed the lda without these two parameters .
however we also performed a lda including all the six parameters for comparison and to check the effect of noise since both asym and smooth parameters resulted to be the ones more significantly affected by noise , mainly in the two ground - based surveys , as we described in sec .
[ sec_method_highz ] .
( left panel ) and @xmath12 ( right panel ) eigenvectors resulting from the lda of the local sample , without including the smooth and asym parameters .
the red cross - hatched , blue hatched , and green solid histograms correspond to the et , lt_et , and lt_lt galaxies , respectively . as it can be seen , @xmath13 is the main morphology discriminator , whereas @xmath12 only slightly improves the classification for the lt_et galaxies .
[ his_lda],scaledwidth=49.0% ] the training phase of the lda was obtained by fitting the coefficients over the measured parameters of the galaxies of the local sample .
we found that all the possible discrimination is based on two eigenvectors ( @xmath13 and @xmath12 ) both being linear combinations of the four parameters entering in the classification .
the first one , @xmath13 , is dominated by gini , ccon , and cabr , whereas the second one , @xmath12 , is based on cabr , gini , and m20 .
almost all the power of the discrimination arise from the @xmath13 eigenvector whereas @xmath12 only slightly helps to the classification of lt_et galaxies , as it can be seen in fig .
[ his_lda ] where the distributions of both @xmath13 and @xmath12 for the three morphological types are shown .
the three concentration indexes show to be more stable , discriminating better between different morphological types in comparison with m20 .
an attempt to improve the lda was also carried out by doing a second lda iteration including only those galaxies that show the same input and predicted output morphology .
we did not find any significant change , and therefore did not proceed further with this additional step . as a result of the lda of the local sample we found that about 30% of the objects show a different input morphology than the one predicted by the resulting eigenvectors of the lda .
this fact could be caused by errors in the measured parameters or even in the original classification .
to illustrate the results obtained by the lda we present in figure [ his_lda ] , the histogram of the predicted morphologies after applying the eigenvector discrimination over the local sample , where for each output morphological type ( et , lt_et , lt_lt ) we show the contribution of galaxies with different input morphology . as it clearly can be seen ,
whereas et galaxies tend to be well classified , lt_et spread between adjacent input types , and lt_lt are split between both input late - types . after obtaining the eigenvectors that better discriminate the morphology in the local sample , lda transformations to the ( @xmath13 and @xmath12 ) space were performed for each non local sample ( alhambra , sxdx , cosmos ) .
the results so obtained at the three magnitude cuts and for each non - local sample are presented in table [ tab_lda_4param ] . in this table , for each survey , the first column represents the three corresponding magnitude cuts , the second column corresponds to the total number of galaxies ( ngal ) at each magnitude cut and sample ( i.e. galaxies with valid measured parameters after applying the filtering criteria explained in sec .
[ sec_results_morphdiagram ] ) while the score column represents the ratio between the galaxies having the same lda output morphology as the input one normalized to the total number of galaxies ngal ; a score of 1 would indicate a perfect lda classification .
we also tested the results when including asym and smooth . in table
[ tab_lda_all ] we present the results obtained from the lda taking into account all morphological parameters . in general , the scores obtained after removing these two parameters ( see table [ tab_lda_4param ] ) are found to be similar to those that would be obtained including also asym and smooth , however the total number of valid galaxies increase significantly and this is more notable in the case of the ground base surveys ( alhambra and sxds ) . if asym and smooth are excluded , the total number of galaxies with good morphology is found to be larger , thus the final classification get worst when including these two parameters as they are indeed affected by noise . only in surveys similar to cosmos , the inclusion of asym and smooth would have a slightly positive net effect . .
et galaxies are found to be well classified , however lt_et spread between adjacent input types , and lt_lt galaxies are split between both late - types .
[ his_pred],scaledwidth=45.0% ] table [ tab_lda_types ] corresponds to the obtained scores ( ratio of galaxies with same output / input morphology over total galaxies ) for each morphological type and magnitude cut in each survey , as resulting from the lda with neither smooth nor asym included . as can be seen , in all surveys , et galaxies are better recognised at the brighter magnitude cuts , whereas lts tend to be better recognised at fainter magnitude cuts .
in this paper we present for the first time a systematic study of the impact from spatial resolution and depth on the six morphological parameters commonly used for galaxy morphological classification : abraham concentration index , gini coefficient , conselice - bershady concentration index , m20 moment of light , asymmetry , and smoothness .
we studied how strong the impact is for three data sets that have different spatial resolutions and depth : alhambra and sxds , as an example of ground - based data , and cosmos , as an example of deep space - based data .
our results correspond to maximum analysed redshift values of : 1.0 , 2.2 , and 1.6 in the three surveys , respectively , where the provided values describe the distribution of 95% of each sample .
we used a sample of 3000 early- and late - type local galaxies , with the available visual morphological classification , and we measured their parameters in 2 cases : + - first , the reference values , that correspond to the real galaxy redshifts ( magnitudes ) , and + - second , the simulated values , obtained after moving the local galaxies to the observational conditions and magnitude / redshift distribution of three selected surveys .
+ comparing the results obtained in these two cases we showed how each parameter changes in each survey , at three particular magnitude cuts . finally , we analysed and quantified how the impact from spatial resolution and depth affects some of the most used morphological diagrams , gave some suggestions for the galaxy / merger classification studies , and used the lda statistical approach to analyse the most effective combination of parameters to distinguish between different types .
+ in the following , we summarise some of our main findings : + - all six analysed morphological parameters suffer from significant biases related to the spatial resolution and data depth .
+ - the impact of the spatial resolution on the morphology is much stronger in comparison with the data depth , being therefore the most responsible for changing the parameters in the ground - based surveys , making in general the galaxies to appear less concentrated and more symmetric .
+ - we stress that asym and smooth are more sensitive to noise effects .
survey noise should be taken into account carefully when using these two parameters in morphological classification of galaxies , especially when dealing with ground - based data .
+ - m20 results to be also significantly affected in all surveys , changing both the shape and range of its distribution at all brightness levels .
however , m20-gini is the most used diagram for selecting interacting ( merging ) systems .
we highly recommend the use of this diagram simultaneously with other morphological parameters , when dealing with ground - based data sets , and in space - based surveys like cosmos at least when studying sources fainter than f814=23.0 .
+ - ccon shows to be more sensitive to the spatial resolution in comparison to gini and cabr .
it works therefore better for space - based data , whereas the other two concentration indexes behave better for ground - based surveys .
+ - in surveys similar to alhambra , when analysing the highest density regions of early- and late - type galaxies in the main morphological diagnostic diagrams , the impact from spatial resolution and data depth introduces contamination levels of 5 - 25% for et and 3 - 35% for lt galaxies ( depending on the diagram ) , for the brightest galaxies with [email protected] . the diagrams that seem to work the best in this case are those that do not include m20 moment of light .
at the faintest analysed magnitudes ( [email protected] ) the contamination levels increase significantly , being as high as 60 - 100% , making each of the diagnostics useless if used separately .
similar results are obtained in the case of sxds , but at higher magnitude cuts ( of order 1 - 2 ) . moreover , in both surveys , at all magnitude cuts , classification ot lt_lt galaxies suffers very high levels of contamination . taking all this into account ,
when dealing with ground - based data sets , we suggest to avoid the use of 2 - 3 parameter diagnostic diagrams in morphological classification , and to apply instead the use of all morphological parameters simultaneously , and to use statistical approaches based on probability distributions that the galaxy is et or lt .
+ - in space - based surveys similar to cosmos , even at the highest magnitude cuts ( [email protected] ) , the contamination levels are significantly lower than those at the brightest magnitudes in the ground - based surveys .
the observational bias is insignificant for all parameters at magnitudes brighter than f814=21.0 ( except for the lt_lt galaxies ) .
however , when going to fainter magnitudes the contamination increases , and depending on the diagram it changes from 2 - 25% at [email protected] , to 6 - 35% for et , and 20 - 70% for lt galaxies at [email protected] . in surveys similar to cosmos
, we again suggest to use several diagnostics to classify galaxies when going to magnitudes fainter than f814=23.0 .
+ - through lda analysis we obtained that the combination of cabr , gini , and ccon parameters , is the most effective to distinguish between different morphological types .
+ the results presented in this paper can be directly applied to any survey similar to alhambra , sxds and cosmos , and also can serve as an upper / lower limit to take into account when classifying galaxies using shallower / deeper data sets .
we thank chris lintott for accepting to review this paper , giving us constructive comments that in our believe improved the paper significantly .
we also appreciate that he signed his report .
we thank kaz sekiguchi and the sxds team for informing us about the stage of the photometric redshift measurements .
mp acknowledge financial support from jae - doc program of the spanish national research council ( csic ) , co - funded by the european social fund .
this research was supported by the junta de andaluca through project tic114 , and the spanish ministry of economy and competitiveness ( mineco ) through projects aya2010 - 15169 , aya2013 - 42227-p , and aya2013 - 43188-p .
the work uses the observations collected at the centro astronmico hispano alemn ( caha ) at calar alto , operated jointly by the max - planck institut fur astronomie and the instituto de astrofsica de andaluca ( csic ) .
the cefca is funded by the fondo de inversiones de teruel , supported by both the government of spain ( 50% ) and the regional government of aragn ( 50% ) . in this work
, we made use of virtual observatory tool for operations on catalogues and tables ( topcat ) .
iraf is distributed by the national optical astronomy observatories , which are operated by the association of universities for research in astronomy , inc .
, under cooperative agreement with the national science foundation 99 abraham , r. g. , et al .
, 1994 , apj , 432 , 75 abraham , r. g. , et al . , 1996 , apjs , 107 , 1 abraham , r. g. , van den bergh , s. , nair , p. , 2003
, apj , 588 , 218 adams , t. f. , 1977 , apjs , 33 , 19 andrae , r. , jahnke , k. , & melchior , p. , 2011
, mnras , 411 , 385 andrae , r. , melchior , p. , & jahnke , k. , 2011 , mnras , 417 , 2465 aparicio - villegas , t. , et al .
2010 , aj , 139 , 1242 baillard , a. , et al . , 2011 , a&a , 532 , 74 baldry , i. k. , et al .
, 2004 , apj , 600 , 681 baldwin , j. a. , phillips , m. m. , & terlevich , r. , 1981 , pasp , 93 , 5 barden , m. , et al . , 2012 , mnras , 422 , 449 bell , e. f. , et al . , 2003 , apjs , 149 , 289 bentez , n. , et al .
2000 , apj , 536 , 571 bentez , n. , et al .
2009 , apj , 692 , l5 bershady , m. a. , jangren , a. , & conselice , c. j. , 2000 , aj , 119 , 2645 bertin , e. & arnouts , s. , 1996 , a&as , 117 , 393 blanton , m. r. , 2003 , apj , 594 , 186 b@xmath14hm , a. , et al .
2013 , a&a , 549 , 46 brammer g. b. , van dokkum p. g. , coppi p. , 2008 ,
apj , 686 , 1503 bussmann r. s. , et al . , 2009 , apj , 693 , 750 carollo , c. m. , et al . , 2013 ,
apj , 773 , 112 casata , p. , et al .
, 2007 , apjs , 172 , 270 chen , y. , et al .
2010 , apj , 712 , 1385 choi , y .- y . ,
woo , j .- h . , &
park , c. , 2009 , apj , 699 , 1679 chung , j. , et al .
, 2013 , apj , 767 , 15 cibinel , a. , et al . , 2013 ,
apj , 776 , 72 cirasuolo , m. , magliocchetti , m. & celotti , a. , 2005 , mnras , 357 , 1267 conselice , c. j. , et al . , 2000 , apj , 529 , 886 conselice , c. j. , et al . , 2003 , apj , 147 , 1 conselice , c. j. , 2006 , mnras , 373 , 1389 conselice , c. j. , et al . , 2008 , mnras , 386 , 909 cotini , s. , et al . , 2013 , mnras , 431 , 2661 deng , x .- f . , 2013 , raa , 13 , 651 de souza , r. e. , gadotti , d. a. , & dos anjos , s. , 2004 , apjs , 153 , 411 de vaucouleurs , g. 1948 , annales dastrophysique , 11 , 247 doi , m. , fukugita , m. , & okamura , s. , 1993 , mnras , 264 , 832 folkes , s. r. , lahav , o. , & maddox , s. j. , 1996 , mnras , 283 , 651 furusawa , h. , 2008 , apjs , 176 , 1 gabor , j. m. , et al .
, 2009 , apj , 691 , 705 graham , a. w. , et al . , 2001a , apj , 563 , l11 graham , a. w. , trujillo , i. & caon , n. , 2001b , aj , 122 , 1707 gunn , j. e. , et al . , 1998 ,
aj , 116 , 3040 hartley , w. g. , et al . , 2013 , mnras , 431 , 3045 heckman , t. m. , 1978 , pasp , 106 , 113 hickox , r. c. , et al . , 2009 , apj , 696 , 891 ho , l. c. , filippenko , a. v. , & sargent , w. l. , 1995 , apjs , 98 , 477 hogg , d. w. , et al . , 2003 , apj , 585 , 5 holwerda , b. w. , et al . , 2014 , apj , 781 , 12 huertas - company , m. , et al . , 2007 ,
a&a , 468 , 937 huertas - company , m. , et al . , 2008 ,
a&a , 478 , 971 huertas - company , m. , et al . , 2009 ,
a&a , 505 , 83 huertas - company , m. , et al . , 2011 , a&a , 525 , 157 huertas - company , m. , et al . , 2014 , arxiv:1406.1175 humason , m. l. , 1931 , apj , 74 , 35 hung , c .-
, et al . , 2014 , apj , 791 , 63 ilbert , o. , et al . , 2009 , apj , 690 , 1236 jimnez - teja , y. & bentez , n. , 2012 , apj , 745 , 150 kauffmann , g. , et al .
, 2003 , mnras , 346 , 1055 kelly , b. c. & mckay , t. a. , 2004 , apj , 127 , 625 kelly , b. c. & mckay , t. a. , 2005 , aj , 129 , 1287 kelm , b. , focardi , p. , & sorrentino , g. , 2005 , a&a , 442 , 117 koekemoer , a. , et al . , 2007 , apjs , 172 , 196 kormendy , j. & richstone , d. , 1995 , ara&a , 33 , 581 leauthaud , a. , et al . , 2007 , apjs , 172 , 219 lanyon - foster , m. m. , et al . , 2012 ,
mnras , 424 , 1852 lawrence , a. , et al . , 2007 , mnras , 379 , 1599 lilly , s. , et al . , 2007 , apjs , 172 , 70 lintott , c. j. , et al
. , 2008 , mnras , 389 , 1179 lintott , c. j. , et al . , 2011 , mnras , 410 , 166 lisker , t. , 2008 , apjs , 179 , 319 lotz , j. m. , primack , j. , madau , p. , 2004 , aj , 128 , 163 lotz , j. m. , et al .
, 2008 , apj , 672 , 177 lotz , j. m. , et al .
2010a , mnras , 404 , 575 lotz , j. m. , et al . 2010b , mnras , 404 , 590 lotz , j. m. , et al .
2011 , apj , 742 , 103 marconi , a. & hunt , l. k. , 2003 , apj , 589 , l21 mclure , r. j. , dunlop , j. s. , & kukula , m. j. , 2000 , mnras , 318 , 693 melbourne , j. et al . , 2007 , apj , 660 , 81 mendez , a. j. , et al . , 2011 , apj , 736 , 110 moles , m. , et al . 2008 ,
aj , 136 , 1325 molino , a. , et al .
2014 , mnras , 441 , 2891 morgan , w. w. , & mayall , n. u. , 1957 , pasp , 69 , 291 nair , p. b. & abraham , r. , 2010 , apjs , 186 , 427 ngan , w. et al .
, 2009 , mnras , 396,1211 oke , j. b. & gunn , j. e. , 1983 , apj , 266 , 713 pearson , k. , 1895 , proceedings of the royal society of london , 58 , 240 peng , c. j. , et al . , 2002 , aj , 124 , 266 peng , c. j. , et al . , 2010 , aj , 139 , 2097 pentericci , l. , et al . , 2010 ,
a&a , 514 , 64 petty , s. m. , et al . , 2014 , [ arxiv:1408.2511p ] pierce , c. m. , et al . , 2007 ,
apj , 660 , l19 povi , m. , et al .
, 2009a , apj , 706 , 810 povi , m. , et al .
, 2009b , apj , 702 , l51 povi , m. , et al . , 2012 ,
a&a , 541 , 118 povi , m. , et al . , 2013 , mnras , snchez - almeida , j. , et al . , 2010 ,
apj , 714 , 487 scoville , n. , et al .
2007 , apjs , 172 , 1 sekiguchi , k. , et al .
2004 , annual report of the national astronomical observatory of japan , 5 , 40 srsic , j. l. , 1963 , baaa , 6 , 41 simard , l. , et al . , 2002 , apjs , 142 , 1 simard , l. , et al .
, 2011 , apjs , 196 , 11 simpson , c. , et al .
, 2012 , mnras , 421 , 3060 simpson , c. , et al . , 2013 , mnras , 433 , 2647 stott , j. p. , et al . , 2013 , mnras , 430 , 1158 strateva , i. , et al . , 2001 , aj , 122 , 1861 tasca , l. a. m. & white , s. d. m. , 2011 , a&a , 530 , 106 tukey , j. w. , 1977 , _
exploratory data analysis _ , addison - wesley .
isbn 0 - 201 - 07616 - 0 .
oclc 3058187 urrutia , t. , lacy , m. , & becker , r. h. , 2008 , apj , 674 , 80 villforth , c. , et al .
2014 , mnras , 439 , 3342 wang , t. , et al . , 2012 ,
apj , 752 , 134 willett , k. w. , et al . , 2013 , mnras , 435 , 2835 weiner , b. j. , et al .
, 2005 , apj , 620 , 595 yamauchi , c. , et al . , 2005 ,
aj , 130 , 1545 | we consistently analyse for the first time the impact of survey depth and spatial resolution on the most used morphological parameters for classifying galaxies through non - parametric methods : abraham and conselice - bershady concentration indices , gini , m20 moment of light , asymmetry , and smoothness .
three different non - local datasets are used , alhambra and sxds ( examples of deep ground - based surveys ) , and cosmos ( deep space - based survey ) .
we used a sample of 3000 local , visually classified galaxies , measuring their morphological parameters at their real redshifts ( z@xmath00 ) .
then we simulated them to match the redshift and magnitude distributions of galaxies in the non - local surveys .
the comparisons of the two sets allow to put constraints on the use of each parameter for morphological classification and evaluate the effectiveness of the commonly used morphological diagnostic diagrams .
all analysed parameters suffer from biases related to spatial resolution and depth , the impact of the former being much stronger . when including asymmetry and smoothness in classification diagrams , the noise effects must be taken into account carefully , especially for ground - based surveys .
m20 is significantly affected , changing both the shape and range of its distribution at all brightness levels .
we suggest that diagnostic diagrams based on 2 - 3 parameters should be avoided when classifying galaxies in ground - based surveys , independently of their brightness ; for cosmos they should be avoided for galaxies fainter than f814=23.0 .
these results can be applied directly to surveys similar to alhambra , sxds and cosmos , and also can serve as an upper/ lower limit for shallower / deeper ones .
[ firstpage ] surveys ; galaxies : morphology ; galaxies : fundamental parameters ; |
in the recent years , communications over multiple - antenna fading channels has attracted the attention of many researchers .
initially , the main interest has been on the point - to - point multiple - input multiple - output ( mimo ) communications @xcite . in @xcite and @xcite ,
the authors have shown that the capacity of a mimo point - to - point channel increases linearly with the minimum number of the transmit and the receive antennas .
more recently , new information theoretic results @xcite , @xcite , @xcite , @xcite have shown that in multiuser mimo systems , one can exploit most of the advantages of multiple - antenna systems .
it has been shown that in a mimo broadcast system , the sum - capacity grows linearly with the minimum number of the transmit and receive antennas @xcite , @xcite , @xcite . to achieve the sum capacity , some information theoretic schemes , based on dirty - paper coding , are introduced .
dirty - paper coding was originally proposed for the gaussian interference channel when the interfering signal is known at the transmitter @xcite .
some methods , such as using nested lattices , are introduced as practical techniques to achieve the sum - capacity promised by the dirty - paper coding @xcite
. however , these methods are not easy to implement . as a simple precoding scheme for mimo broadcast systems ,
the channel inversion technique ( or zero - forcing beamforming @xcite ) can be used at the transmitter to separate the data for different users . to improve the performance of the channel inversion technique , a zero - forcing approximation of the dirty paper coding ( based on qr decomposition ) is introduced in @xcite ( which can be seen as a scalar approximation of @xcite ) . however , both of these methods are vulnerable to the poor channel conditions , due to the occasional near - singularity of the channel matrix ( when the channel matrix has at least one small eigenvalue )
. this drawback results in a poor performance in terms of the symbol - error - rate for the mentioned methods @xcite . in @xcite ,
the authors have introduced a _
vector perturbation technique _ which has a good performance in terms of symbol error rate .
nonetheless , this technique requires a lattice decoder which is an np - hard problem . to reduce the complexity of the lattice decoder , in @xcite , the authors have used lattice - basis reduction to approximate the closest lattice point ( using babai approximation ) . in this paper
, we present a transmission technique for the mimo broadcast channel based on the lattice - basis reduction .
instead of approximating the closest lattice point in the perturbation problem , we use the lattice - basis reduction to reduce the average transmitted energy by reducing the second moment of the fundamental region generated by the lattice basis .
this viewpoint helps us to : ( i ) achieve a better performance as compared to @xcite , ( ii ) expand the idea for the case of unequal - rate transmission , and ( iii ) obtain some analytic results for the asymptotic behavior ( snr@xmath1 ) of the symbol - error - rate for both the proposed technique and the perturbation technique of @xcite .
the rest of the paper is organized as the following : sections ii and iii briefly describe the system model and introduce the concept of lattice basis reduction . in section iv , the proposed method is described and in section v , the proposed approach is extended for the case of unequal - rate transmission . in section
vi , we consider the asymptotic performance of the proposed method for high snr values , in terms of the probability of error . we define the precoding diversity and the outage probability for the case of fixed - rate users . it is shown that by using lattice basis reduction , we can achieve the maximum precoding diversity .
for the proof , we use a bound on the orthogonal deficiency of an lll - reduced basis .
also , an upper bound is given for the probability that the length of the shortest vector of a lattice ( generated by complex gaussian vectors ) is smaller than a given value .
using this result , we also show that the perturbation technique achieves the maximum precoding diversity . in section vii ,
some simulation results are presented .
these results show that the proposed method offers almost the same performance as @xcite with a much smaller complexity .
as compared to @xcite , the proposed method offers almost the same performance .
however , by sending a very small amount of side information ( a few bits for one fading block ) , the modified proposed method offers a better performance with a similar complexity .
finally , in section viii , some concluding remarks are presented .
we consider a multiple - antenna broadcast system with @xmath2 transmit antennas and @xmath3 single - antenna users ( @xmath4 ) .
consider @xmath5^{t } $ ] , @xmath6^{t } $ ] , @xmath7^{t } $ ] , and the @xmath8 matrix @xmath9 , respectively , as the received signal , the transmitted signal , the noise vector , and the channel matrix .
the transmission over the channel can be formulated as , @xmath10 the channel is assumed to be raleigh , i.e. the elements of @xmath9 are i.i.d . with the zero - mean unit - variance complex gaussian distribution and
the noise is i.i.d .
additive gaussian .
moreover , we have the energy constraint on the transmitted signal , @xmath11 .
the energy of the additive noise is @xmath12 per antenna , i.e. @xmath13 . the signal - to - noise ratio ( snr ) is defined as @xmath14 . in a broadcast system
, the receivers do not cooperate with each other ( they should decode their respective data , independently ) .
the main strategy in dealing with this restriction is to apply an appropriate precoding scheme at the transmitter .
the simplest method in this category is using the channel inversion technique at the transmitter to separate the data for different users : @xmath15 where @xmath16 , and @xmath17 is the hermitian of @xmath18 .
moreover , @xmath19 is the transmitted signal before the normalization ( @xmath20 is the normalized transmitted signal ) , and @xmath21 is the data vector , i.e. @xmath22 is the data for the _ i_th user . for @xmath23 ( the number of transmit antennas and the number of users are equal ) ,
the transmitted signal is @xmath24 the problem arises when @xmath9 is poorly conditioned and @xmath25 becomes very large , resulting in a high power consumption .
this situation occurs when at least one of the singular values of @xmath9 is very small which results in vectors with large norms as the columns of @xmath26 .
fortunately , most of the time ( especially for high snrs ) , we can combat the effect of a small singular value by changing the supporting region of the constellation which is the main motivation behind the current article .
when the data of different users are selected from @xmath27 $ ] , the overall constellation can be seen as a set of lattice points . in this case ,
lattice algorithms can be used to modify the constellation .
especially , lattice - basis reduction is a natural solution for modifying the supporting region of the constellation .
lattice structures have been frequently used in different communication applications such as quantization or decoding of mimo systems .
a real ( or complex ) lattice @xmath28 is a discrete set of @xmath29-d vectors in the real euclidean space @xmath30 ( or the complex euclidean space @xmath31 ) that forms a group under ordinary vector addition .
every lattice @xmath28 is generated by the integer linear combinations of some set of linearly independent vectors @xmath32 , where the integer @xmath33 , @xmath34 , is called the dimension of the lattice @xmath28 .
the set of vectors @xmath35 is called a basis of @xmath28 , and the matrix @xmath36 $ ] , which has the basis vectors as its columns , is called the basis matrix ( or generator matrix ) of @xmath37 .
the basis for representing a lattice is not unique .
usually a basis consisting of relatively short and nearly orthogonal vectors is desirable .
the procedure of finding such a basis for a lattice is called _ lattice basis reduction_. a popular criterion for lattice - basis reduction is to find a basis such that @xmath38 is minimized . because the volume of the lattice is @xmath39 . ]
does not change with the change of basis , this problem is equivalent to minimizing the orthogonality defect which is defined as @xmath40 the problem of finding such a basis is np - hard @xcite .
several distinct sub - optimal reductions have been studied in the literature , including those associated to the names minkowski , korkin - zolotarev , and more recently lenstra - lenstra and lovasz ( lll ) @xcite .
an ordered basis @xmath41 is a _ minkowski - reduced basis _ @xcite if * @xmath42 is the shortest nonzero vector in the lattice @xmath28 , and * for each @xmath43 , @xmath44 is the shortest nonzero vector in @xmath28 such that @xmath45 may be extended to a basis of @xmath28 .
minkowski reduction can be seen as a greedy solution for the lattice - basis reduction problem . however , finding minkowski reduced basis is equivalent to finding the shortest vector in the lattice and this problem by itself is np - hard .
thus , there is no polynomial time algorithm for this reduction method . in @xcite , a reduction algorithm ( called _ lll algorithm _ )
is introduced which uses the gram - schmidt orthogonalization and has a polynomial complexity and guarantees a bounded orthogonality defect . for any ordered basis of @xmath28 ,
say @xmath46 , one can compute an ordered set of gram - schmidt vectors , @xmath47 , which are mutually orthogonal , using the following recursion : @xmath48 where @xmath49 is the inner product .
an ordered basis @xmath50 is an _ lll reduced basis _
@xcite if , * @xmath51 for @xmath52 , and * @xmath53 where @xmath54 , and @xmath55 is the gram - schmidt orthogonalization of the ordered basis and @xmath56 for @xmath57 .
it is shown that lll basis - reduction algorithm produces relatively short basis vectors with a polynomial - time computational complexity @xcite .
the lll basis reduction has found extended applications in several contexts due to its polynomial - time complexity . in @xcite ,
lll algorithm is generalized for euclidean rings ( including the ring of complex integers ) . in this paper , we will use the following important property of the complex lll reduction ( for @xmath58 ) : [ see @xcite ] let @xmath59 be an @xmath60-dimensional complex lattice and @xmath61 $ ] be the lll reduced basis of @xmath37 . if @xmath62 is the orthogonality defect of @xmath63 , then , @xmath64
assume that the data for different users , @xmath65 , is selected from the points of the integer lattice ( or from the half - integer grid @xcite ) .
the data vector @xmath21 is a point in the cartesian product of these sub - constellations . as a result ,
the overall receive constellation consists of the points from @xmath66 , bounded within a @xmath67-dimensional hypercube . at the transmitter side , when we use the channel inversion technique , the transmitted signal is a point inside a parallelotope whose edges are parallel to vectors , defined by the columns of @xmath68 .
if the data is a point from the integer lattice @xmath69 , the transmitted signal is a point in the lattice generated by @xmath68 .
when the squared norm of at least one of the columns of @xmath68 is too large , some of the constellation points require high energy for the transmission .
we try to reduce the average transmitted energy , by replacing these points with some other points with smaller square norms .
however , the lack of cooperation among the users imposes the restriction that the received signals should belong to the integer lattice @xmath69 ( to avoid the interference among the users ) .
the core of the idea in this paper is based on using an appropriate supporting region for the transmitted signal set to minimize the average energy , without changing the underlying lattice .
this is achieved through the lattice - basis reduction .
when we use the continuous approximation ( which is appropriate for large constellations ) , the average energy of the transmitted signal is approximated by the second moment of the transmitted region @xcite .
when we assume equal rates for the users , e.g. @xmath70 bits per user ( @xmath71 bits per dimension ) , the signal points ( at the receiver ) are inside a hypercube with an edge of length @xmath72 where @xmath73 therefore , the supporting region of the transmitted signal is the scaled version of the fundamental region of the lattice generated by @xmath74 ( corresponding to its basis ) with the scaling factor @xmath75 .
note that by changing the basis for this lattice , we can change the corresponding fundamental region ( a parallelotope generated by the basis of the lattice and centered at the origin ) .
the second moment of the resulting region is proportional to the sum of the squared norms of the basis vectors ( see appendix a ) .
therefore , we should try to find a basis reduction method which minimizes the sum of the squared norms of the basis vectors .
figure 1 shows the application of the lattice basis reduction in reducing the average energy by replacing the old basis with a new basis which has shorter vectors . in this figure , by changing the basis @xmath76 ( columns of @xmath74 ) to @xmath77 ( the reduced basis ) , the fundamental region @xmath78 , generated by the original basis , is replaced by @xmath79 , generated by the reduced basis . among the known reduction algorithms , the minkowski reduction can be considered as an appropriate greedy algorithm for our problem .
indeed , the minkowski algorithm is the successively optimum solution because in each step , it finds the shortest vector .
however , the complexity of the minkowski reduction is equal to the complexity of the shortest - lattice - vector problem which is known to be np - hard @xcite
. therefore , we use the lll reduction algorithm which is a suboptimum solution with a polynomial complexity .
assume that @xmath80 is the lll - reduced basis for the lattice obtained by @xmath68 , where @xmath81 is an @xmath82 unimodular matrix ( both @xmath83 and @xmath84 have integer entries ) .
we use @xmath85 as the transmitted signal where @xmath86 is the precoded data vector , @xmath87 is the original data vector , and @xmath88 is the length of the edges of the hypercube , defined by ( [ eq : ar ] ) . at the receiver side ,
we use modulo operation to find the original data : @xmath89 @xmath90 in obtaining ( [ eq:9 ] ) from ( [ eq:8 ] ) , we use the fact that @xmath91 and @xmath92 have integer entries . in this method , at the beginning of each fading block , we reduce the lattice obtained by @xmath68 and during this block the transmitted signal is computed using ( [ eq : transmod ] ) . neglecting the preprocessing at the beginning of the block ( for lattice reduction ) , the complexity of the precoding is in the order of a matrix multiplication and a modulo operation . therefore , the complexity of the proposed precoding method is comparable to the complexity of the channel inversion method .
however , as we will show by the simulation results , the performance of this method is significantly better , and indeed , is near the performance of the perturbation method , presented in @xcite . in the perturbation technique @xcite ,
the idea of changing the support region of the constellation has been implemented using a different approach . in @xcite , @xmath93 is used as the precoded data , where the integer vector @xmath94 is chosen to minimize @xmath95 .
this problem is equivalent to the closest - lattice - point problem for the lattice generated by @xmath96 ( i.e. finding the lattice point which is closer to @xmath97 ) .
therefore , in the perturbation technique , the support region of the constellation is a scaled version of the voronoi region @xcite of the lattice . in the proposed method
, we use a parallelotope ( generated by the reduced basis of the lattice ) , instead of the voronoi region .
although this approximation results in a larger second moment ( i.e. higher energy consumption ) , it enables us to use a simple precoding technique , instead of solving the closest - lattice - point problem . for the lattice constellations , using a parallelotope instead of the voronoi region ( presented in this paper )
is equivalent with using the babai approximation instead of the exact lattice decoding ( previously presented in @xcite ) .
the only practical difference between our lattice - reduction - aided scheme and the scheme presented in @xcite is that we reduce @xmath26 , while in @xcite , @xmath98 is reduced .
as can be observed in the simulation results ( figure 2 ) , this difference has no significant effect on the performance .
however , the new viewpoint helps us in extending the proposed method for the case of variable - rate transmission and obtaining some analytical results for the asymptotic performance .
the performance of the proposed lattice - reduction aided scheme can be improved by combining it with other schemes , such as regularization @xcite or the v - blast precoding @xcite , or by sending a very small amount of side information . in the rest of this section , we present two of these modifications . in @xcite , the authors have proposed a regularization scheme to reduce the transmitted power , by avoiding the near - singularity of @xmath18 . in this method , instead of using @xmath99 , the transmitted vector is constructed as @xmath100 where @xmath101 is a positive number . to combine the regularized scheme with our lattice - reduction - aided scheme , we consider @xmath102 as the matrix corresponding to the reduced basis of the lattice generated by @xmath103 .
when we use the regularization , the received signals of different users are not orthogonal anymore and the interference acts like extra noise .
parameter @xmath104 should be optimized such that the ratio between the power of received signal and the power of the effective noise is minimized @xcite . in practical systems ,
we are interested in using a subset of points with odd coordinates from the integer lattice . in these cases
, we can improve the performance of the proposed method by sending a very small amount of side information .
when the data vector @xmath21 consists of odd integers , using the lattice - basis reduction may result in points with some even coordinates ( i.e. @xmath105 has some even elements ) , instead of points with all - odd coordinates in the new basis . for this case , in ( [ eq : transmod ] ) , the set of precoded data @xmath106 is not centered at the origin , hence the transmitted constellation ( which includes all the valid points @xmath107 ) is not centered at the origin . therefore , we can reduce the transmitted energy and improve the performance by shifting the center of the constellation to the origin .
it can be shown that the translation vector is equal to @xmath108^{\mathtt{t } } + [ 1+i , 1+i,\cdots , 1+i]^{\mathtt{t } } ) \mod 2 $ ] where @xmath109 .
when we use this shifted version of the constellation , we must send the translation vector to the users ( by sending 2 bits per user ) at the beginning of the block .
however , compared to the size of the block of data , the overhead of these two bits is negligible .
the above idea of using a shift vector can be also used to improve the perturbation technique ( if we only use the odd points of the lattice ) . after reducing the inverse of the channel matrix and obtaining the bits ( corresponding to the shift vector ) at the beginning of each fading block , the closest point to the signal computed in equation ( [ eq : transmod ] )
can be found by using the sphere decoder .
then , the transmitted signal is obtained by @xmath110 where @xmath111 is the zero - one shift vector , which is computed for users at the beginning of the fading block , and the perturbation vector @xmath112 is an even integer vector such that the vector @xmath113 has the minimum energy . this method which can be considered as _ modified perturbation method _ outperforms the perturbation method in @xcite .
when we are not restricted to the odd lattice points , using ( [ eq : modpert ] ) instead of @xmath114 does not change the performance of the perturbation method .
it only reduces the complexity of the lattice decoder @xcite .
in the previous section , we had considered the case that the transmission rates for different users are equal . in some applications , we are interested in assigning different rates to different users
. consider @xmath115 as the transmission rates for the users ( we consider them as even integer numbers ) .
equation ( [ eq : transmod ] ) should be modified as @xmath116 where the entries of @xmath117^{t } $ ] are equal to @xmath118 also , at the receiver side , instead of ( [ eq:8 ] ) and ( [ eq:9 ] ) , we have @xmath119 @xmath120 if we are interested in sum - rate , instead of individual rates , we can improve the performance of the proposed method by assigning variable rates to different users .
we assume that the sum - rate ( rather than the individual rates ) is fixed and we want to reduce the average transmitted energy . to simplify the analysis
, we use the continuous approximation which has a good accuracy for high rates . considering continuous approximation ,
the sum - rate is proportional to the logarithm of the volume of the lattice with basis @xmath121 and the average energy is proportional to the second moment of the corresponding prallelotope , which is proportional to @xmath122 ( see appendix a ) .
the goal is to minimize the average energy while the sum - rate is fixed .
we can use another lattice generated by @xmath123 with the same volume , where its basis vectors are scaled versions of the vectors of the basis @xmath124 , according to different rates for different users .
therefore , we can use @xmath125 instead of @xmath126 ( where @xmath127 is a unit determinant @xmath128 diagonal matrix which does not change the volume of the lattice ) . for a given reduced basis @xmath124 ,
the product of the squared norms of the new basis vectors is constant : @xmath129 the average energy corresponding to the new lattice basis should be minimized .
when we use the modified basis @xmath130 instead of @xmath131 , the average energy is proportional to @xmath132 ( see appendix a ) . according to the arithmetic - geometric mean inequality , @xmath132 is minimized iff @xmath133 therefore , @xmath134 having the matrix @xmath124 , the columns of matrix @xmath135 can be found using the equation ( [ eq : unequal ] ) and @xmath136 can be obtained by ( [ eq : unequal2 ] ) .
now , for the selection of the reduced basis @xmath63 , we should find @xmath124 such that @xmath137 is minimized .
because @xmath138 is given , the best basis reduction is the reduction which maximizes @xmath139 , or in other words , minimizes the orthogonality defect . in practice , we use discrete values for the rate , and sometimes , we should assign the rate zero to some users ( when their channel is very bad ) . in this case , for the rate assignment for other users , we use the lattice reduction on the corresponding sublattice . it should be noted that the average transmit power is fixed per channel realization and no long - run averaging is considered , and no long - run power allocation is used .
in this section , we consider the asymptotic behavior ( @xmath140 ) of the symbol error rate ( ser ) for the proposed method and the perturbation technique .
we show that for both of these methods , the asymptotic slope of the ser curve is equal to the number of transmit antennas . by considering the outage probability of a fixed - rate mimo broadcast system
, we will show that for the ser curve in high snr , the slope obtained by the proposed method has the largest achievable value .
also , we analyze the asymptotic behavior of the outage probability for the case of fixed sum - rate . we show that in this case , the slope of the corresponding curve is equal to the product of the number of transmit antennas and the number of single - antenna users .
when we have the channel - state information ( csi ) at the transmitter , without any assumption on the transmission rates , the outage probability is not meaningful .
however , when we consider given rates @xmath141 for different users , we can define the outage probability @xmath142 as the probability that the point @xmath143 is outside the capacity region . for a mimo broadcast system with
@xmath144 transmit antennas , @xmath145 single - antenna receivers ( @xmath146 ) , and given rates @xmath141 , @xmath147 define @xmath148 as the probability that the capacity of the point - to - point system corresponding to the first user ( consisting of @xmath144 transmit antennas and one receive antenna with independent channel coefficients and csi at the transmitter ) is less than @xmath149 : @xmath150 where @xmath151 is the vector defined by the first row of @xmath152 .
note that the entries of @xmath151 have iid complex gaussian distribution with unit variance .
thus , its square norm has a chi square distribution .
we have , @xmath153 @xmath154 @xmath155 @xmath156 we are interested in the large values of @xmath157 . for @xmath158 , @xmath159 @xmath160 @xmath161 where @xmath162 is a constant number .
now , @xmath163 @xmath164 @xmath165 @xmath166 @xmath167 according to the definition of @xmath168 , @xmath169 .
therefore , @xmath170 we can define the diversity gain of a mimo broadcast constellation or its _ precoding diversity _ as @xmath171 where @xmath172 is the probability of error .
similar to ( * ? ? ?
* lemma 5 ) , we can bound the precoding diversity by @xmath173 . thus , based on theorem 2 , the maximum achievable diversity is @xmath144 .
we show that the proposed method ( based on lattice - basis reduction ) achieves the maximum precoding diversity . to prove this , in lemma 1 and lemma 2
, we relate the length of the largest vector of the reduced basis @xmath63 to @xmath174 ( the minimum distance of the lattice generated by @xmath17 ) . in lemma 3
, we bound the probability that @xmath174 is too small .
finally , in theorem 3 , we prove the main result by relating the minimum distance of the receive constellation to the length of the largest vector of the reduced basis @xmath63 , and combining the bounds on the probability that @xmath174 is too small , and the probability that the noise vector is too large .
consider @xmath175 $ ] as an @xmath176 matrix , with the orthogonality defect @xmath177 , and @xmath178 $ ] as the inverse of its hermitian ( or its pseudo - inverse if @xmath179 ) .
then , @xmath180 and @xmath181 consider @xmath182 as an arbitrary column of @xmath121 .
the vector @xmath182 can be written as @xmath183 , where @xmath184 is orthogonal to @xmath185 for @xmath186 .
now , @xmath187 $ ] can be written as @xmath188 where @xmath189 is a unit - determinant @xmath190 matrix ( a column operation matrix ) : @xmath191 @xmath192 @xmath193^{\mathtt{h } } [ \mathbf{b}_{1} ...
\mathbf{b}_{i-1 } \mathbf{b}_{i}^{\prime } \mathbf{b}_{i+1} ... \mathbf{b}_{m}]\right ) .\ ] ] according to the hadamard theorem : @xmath194^{\mathtt{h } } [ \mathbf{b}_{1} ... \mathbf{b}_{i-1 } \mathbf{b}_{i}^{\prime } \mathbf{b}_{i+1} ... \mathbf{b}_{m}]\right ) \leq\ ] ] @xmath195 therefore , @xmath196 @xmath197 also , @xmath198 results in @xmath199 and @xmath200 for @xmath201 .
therefore , @xmath202 now , @xmath203 and @xmath204 , both are orthogonal to the @xmath205-dimensional subspace generated by the vectors @xmath206 ( @xmath207 ) . thus , @xmath208 @xmath209 @xmath210 the above relation is valid for every @xmath211 , @xmath212 . without loss of generality , we can assume that @xmath213 : @xmath214 @xmath215 similarly , by using ( [ eq:15 ] ) , we can also obtain the following inequality : @xmath216 consider @xmath175 $ ] as an lll - reduced basis for the lattice generated by @xmath68 and @xmath217 as the minimum distance of the lattice generated by @xmath218 .
then , there is a constant @xmath219 ( independent of @xmath220 ) such that @xmath221 according to the theorem 1 , @xmath222 consider @xmath223 $ ] . by using lemma 1 and ( [ eq : delta ] ) , @xmath224 the basis @xmath63 can be written as @xmath225 for some unimodular matrix @xmath83 : @xmath226 noting that @xmath227 is unimodular , @xmath223 $ ] is another basis for the lattice generated by @xmath98 .
therefore , the vectors @xmath228 are vectors from the lattice generated by @xmath98 , and therefore , the length of each of them is at least @xmath229 : @xmath230 @xmath231 @xmath232 assume that the entries of the @xmath233 matrix @xmath9 has independent complex gaussian distribution with zero mean and unit variance and consider @xmath234 as the minimum distance of the lattice generated by @xmath220 .
then , there is a constant @xmath235 such that @xmath236 proof : see appendix b. for a mimo broadcast system with @xmath144 transmit antennas and @xmath145 single - antenna receivers ( @xmath146 ) and fixed rates @xmath141 , using the lattice - basis - reduction method , @xmath237 consider @xmath238 $ ] as the lll - reduced basis for the lattice generated by @xmath239 .
each transmitted vector @xmath240 is inside the parallelotope , generated by @xmath241 ( where @xmath242 are constant values determined by the rates of the users ) .
thus , every transmitted vector @xmath240 can be written as @xmath243 for each of the transmitted vectors , the energy is @xmath244 @xmath245 @xmath246 thus , the average transmitted energy is @xmath247 where @xmath248 . the received signals ( without the effect of noise ) are points from the @xmath249 lattice .
if we consider the normalized system ( by scaling the signals such that the average transmitted energy becomes equal to one ) , @xmath250 is the squared distance between the received signal points . for the normalized system
, @xmath251 is the energy of the noise at each receiver and @xmath252 is the energy of the noise per each real dimension . using ( [ eq:60 ] ) , for any positive number @xmath253 , @xmath254 @xmath255 + using lemma 2 , @xmath256 @xmath257 + the @xmath258 matrix @xmath259 has independent complex gaussian distribution with zero mean and unit variance .
therefore , by using lemma 3 ( considering @xmath260 and @xmath261 ) , we can bound the probability that @xmath174 is too small .
* case 1 * , @xmath262 : @xmath263 @xmath264 @xmath265 @xmath266 where @xmath267 is a constant number and @xmath268 is the euler number .
if the magnitude of the noise component in each real dimension is less than @xmath269 , the transmitted data will be decoded correctly .
thus , we can bound the probability of error by the probability that @xmath270 is greater than @xmath271 for at least one @xmath272 , @xmath273 . therefore , using the union bound , @xmath274 @xmath275 @xmath276 @xmath277 @xmath278 @xmath279 @xmath280 @xmath281 for the product terms in ( [ eq:56 ] ) , we can bound the first part by ( [ eq:55 ] ) . to bound the second part
, we note that @xmath282 has real gaussian distribution with variance @xmath283
. therefore , @xmath284 now , for @xmath285 , @xmath286 @xmath287 @xmath288 @xmath289 where @xmath290 is a constant number which only depends on @xmath2 .
thus , @xmath291 + according to theorem 2 , this limit can not be greater than @xmath144 . therefore , @xmath292 * case 2 * , @xmath293 : for the @xmath258 matrix @xmath259 , we use the first inequality in lemma 3 ( by considering @xmath260 and @xmath294 ) to bound the probability that @xmath174 is too small : @xmath295 @xmath296 @xmath297 @xmath298 . the rest of proof is similar to the case 1 .
perturbation technique achieves the maximum precoding diversity in fixed - rate mimo broadcast systems . in the perturbation technique , for the transmission of each data vector @xmath299 , among the set @xmath300 , the nearest point to the origin
is chosen .
the transmitted vector in the lattice - reduction - based method belongs to that set .
therefore , the energy of the transmitted signal in the lattice - reduction - based method can not be less than the transmitted energy in the perturbation technique .
thus , the average transmitted energy for the perturbation method is at most equal to the average transmitted energy of the lattice - reduction - based method .
the rest of the proof is the same as the proof of theorem 3 .
when the sum - rate @xmath301 is given , similar to the previous part , we can define the outage probability as the probability that the sum - capacity of the broadcast system is less than @xmath301 .
for a mimo broadcast system with @xmath144 transmit antennas , @xmath145 single - antenna receivers , and a given sum - rate @xmath302 , @xmath303 for any channel matrix @xmath9 , we have @xcite @xmath304 where @xmath305 is a diagonal matrix with non - negative elements and unit trace . also , @xcite @xmath306
the entries of @xmath152 have iid complex gaussian distribution with unit variance .
thus @xmath307 is equal to the square norm of an @xmath308-dimensional complex gaussian vector and has a chi square distribution with @xmath309 degrees of freedom .
thus , we have ( similar to the equations [ eq:21]-[eq:27 ] , in the proof of theorem 2 ) , @xmath310 @xmath311 @xmath312 where @xmath313 is a constant number . now , @xmath314 @xmath315 @xmath316 by using ( [ eq:74 ] ) , ( [ eq:76 ] ) , and ( [ eq:79 ] ) : @xmath317 the slope @xmath318 for the ser curve can be easily achieved by sending to only the best user .
similar to the proof of theorem 3 , the slope of the symbol - error rate curve is asymptoticly determined by the slope of the probability that @xmath319 is smaller than a constant number , where @xmath320 is the entry of @xmath9 with maximum norm . due to the iid complex gaussian distribution of the entries of @xmath9 , this probability decays with the same rate as @xmath321 , for large @xmath322 .
however , although sending to only the best user achieves the optimum slope for the ser curve , it is not an efficient transmission technique because it reduces the capacity to the order of @xmath323 ( instead of @xmath324 ) .
figure 2 presents the simulation results for the performance of the proposed schemes , the perturbation scheme @xcite , and the naive channel inversion approach .
the number of the transmit antennas is @xmath325 and there are @xmath326 single - antenna users in the system .
the overall transmission rate is 8 bits per channel use , where 2 bits are assigned to each user , i.e. a qpsk constellation is assigned to each user . by considering the slope of the curves in figure 2
, we see that by using the proposed reduction - based schemes , we can achieve the maximum precoding diversity , with a low complexity .
also , as compared to the perturbation scheme , we have a negligible loss in the performance ( about 0.2 db ) .
moreover , compared to the approximated perturbation method @xcite , we have about 1.5 db improvement by sending the bits , corresponding to the shift vector , at the beginning of the transmission . without sending the shift vector ,
the performance of the proposed method is the same as that of the approximated perturbation method @xcite .
the modified perturbation method ( with sending two shift bits for each user ) has around 0.3 db improvement compared to the perturbation method .
transmit antennas and @xmath326 single - antenna receivers with the rate @xmath327 bits per channel use per user . ]
figure 3 compares the regularized proposed scheme with v - blast modifications of zero - forcing and babai approximation for the same setting . as shown in the simulation results ( and also in the simulation results in @xcite and @xcite ) , the modulo - mmse - vblast scheme does not achieve a precoding diversity better than zero forcing ( though it has a good performance in the low snr region ) .
however , combining the lattice - reduction - aided ( lra ) scheme with mmse - vblast precoding or other schemes such as regularization improves its performance by a finite coding gain ( without changing the slope of the curve of symbol - error - rate ) .
combining both the regularization and the shift vector can result in better performance compared to other alternatives .
transmit antennas and @xmath326 single - antenna receivers with the rate @xmath327 bits per channel use per user . ) . ]
figure 4 compares the performances of the fixed - rate and the variable - rate transmission using lattice - basis reduction for @xmath328 transmit antennas and @xmath329 users . in both cases ,
the sum - rate is 8 bits per channel use ( in the case of fixed individual rates , a 16qam constellation is assigned to each user ) .
we see that by eliminating the equal - rate constraint , we can considerably improve the performance ( especially , for high rates ) .
in fact , the diversity gains for the equal - rate and the unequal - rate methods are , respectively , @xmath145 and @xmath330 .
transmit antennas and @xmath329 single - antenna receivers with sum - rate 8 bits per channel use . ]
a simple scheme for communications in mimo broadcast channels is introduced which is based on the lattice reduction technique and improves the performance of the channel inversion method .
lattice basis reduction helps us to reduce the average transmitted energy by modifying the region which includes the constellation points .
simulation results show that the performance of the proposed scheme is very close to the performance of the perturbation method .
also , it is shown that by using lattice - basis reduction , we achieve the maximum precoding diversity with a polynomial - time precoding complexity .
in this appendix , we compute the second moment of a parallelotope whose centroid is the origin and its edges are equal to the basis vectors of the lattice .
assume that @xmath331 is an @xmath332-dimensional parallelotope and @xmath333 is its second moment .
the second moment of @xmath334 is @xmath335 .
the parallelotope @xmath331 can be considered as the union of @xmath336 smaller parallelotopes which are constructed by @xmath337 , where @xmath338 , @xmath339 , is a basis vector .
these parallelotopes are translated versions of @xmath334 with the translation vectors @xmath340 , @xmath341 .
the second moments of these parallelotopes are equal to @xmath342 , @xmath343 . by the summation over all these second moments
, we can find the second moment of @xmath331 .
@xmath344\ ] ] @xmath345 @xmath346 @xmath347
lemma 3 states that the probability that a lattice , generated by @xmath60 independent @xmath348-dimensional complex gaussian vectors , @xmath349 , with a unit variance per each dimension , has a nonzero point inside a sphere ( centered at origin and with the radius @xmath350 ) is bounded by @xmath351 for @xmath352 , and @xmath353 for @xmath354 .
we can assume that @xmath355 ( for @xmath356 , lemma 3 is trivial because the probability is bounded ) . when @xmath357 , the lattice consists of the integer multiples of the basis vector @xmath358 .
if the norm of one of these vectors is less than @xmath359 , then the norm of @xmath358 is less than @xmath359 . consider the variance of the components of @xmath358 as @xmath360 .
the vector @xmath358 has an @xmath361-dimensional complex gaussian distribution , @xmath362 .
therefore , the probability of this event is , @xmath363 when the variance of the components of @xmath358 is equal to one , we have , @xmath364 consider @xmath366 as the lattice generated by @xmath367,@xmath368, ... ,@xmath369 .
each point of @xmath366 can be represented by @xmath370 , where @xmath371 are complex integer numbers .
the vectors @xmath367,@xmath368, ... ,@xmath369 are independent and jointly gaussian . therefore , for every integer vector @xmath372 , the entries of the vector @xmath373 have complex gaussian distributions with the variance the @xmath332-dimensional complex integer points @xmath380 , such that @xmath381 , can be considered as the centers of disjoint unit - volume cubes .
all these cubes are inside the region between the @xmath382-dimensional spheres , with radii @xmath383 and @xmath384 .
therefore , the number of @xmath332-dimensional complex integer points @xmath380 , such that @xmath381 , can be bounded by the volume of the region between these two @xmath382-dimensional spheres .
thus , this number is bounded by @xmath385 for some constant are some constant numbers . ]
therefore , @xmath387
@xmath388 each point of @xmath395 can be represented by @xmath396 , where @xmath397 belongs to the lattice @xmath398 and @xmath399 is a complex integer .
consider @xmath400 as the sphere with radius @xmath350 and centered at @xmath401 .
now , @xmath396 belongs to @xmath402 iff the @xmath403 belongs to @xmath400 .
also , the sphere @xmath400 includes a point @xmath404 iff @xmath405 includes @xmath406 , where @xmath407 is the sphere centered at @xmath408 with radius @xmath409 ( see figure 4 ) .
therefore , the probability that a lattice point exists in @xmath410 is equal to the probability that @xmath411 is in at least one of the spheres @xmath412 , @xmath413 .
assume that the minimum distance of @xmath398 is @xmath436 .
the spheres with the radius @xmath437 and centered by the points of @xmath398 are disjoint .
therefore , the number of points from the ( @xmath438)-dimensional complex lattice @xmath398 , such that @xmath439 , is bounded by @xmath440 ( it is bounded by the ratio between the volumes of ( @xmath441)-dimensional spheres with radii @xmath442 and @xmath443 ) . also , the number of points from @xmath398 , such that @xmath444 , is bounded by @xmath445 ( it is bounded by the ratio between the volumes of the region defined by @xmath446 and the sphere with radius @xmath443 ) : to bound the second term of ( [ eq:104 ] ) , we note that for @xmath456 , the radii of the spheres @xmath457 are less or equal to @xmath458 , and the centers of these spheres lie on the @xmath459-dimensional complex subspace containing @xmath460 .
also , the norm of these centers are less than @xmath418 .
therefore , all of these spheres are inside the region @xmath461 which is an orthotope centered at the origin , with @xmath462 real dimensions ( see figure 5 ) : also , according to the gaussian distribution of the entries of @xmath466 ( which have variance @xmath467 on each real dimension ) , we can bound the third term of ( [ eq:104 ] ) as , @xmath468 by using ( [ eq:123 ] ) , ( [ eq:125 ] ) , and ( [ eq:126 ] ) , c. b. peel , b. m. hochwald , and a. l. swindlehurst , `` a vector - perturbation technique for near - capacity multiple - antenna multi - user communications - part ii : perturbation , '' _ ieee trans .
_ , pp . 537 544 , march 2005 .
v. tarokh , n. seshadri , and a. r. calderbank , `` space - time codes for high data rate wireless communication : performance criterion and code construction , '' _ ieee trans .
info . theory _ ,
44 , pp . 744765 , mar .
1998 . c. windpassinger , r. f. h. fischer , and j. b. huber , `` lattice - reduction - aided broadcast precoding , '' in _
5th international itg conference on source and channel coding ( scc ) _ , ( erlangen , germany ) , pp . 403408 , january 2004 .
r. fischer , `` lattice - reduction - aided equalization and generalized partial response signaling for point - to - point transmission over flat - fading mimo channels , '' in _
6th international itg - conference on source and channel coding _ ,
( munich , germany ) , april 2006 . g. d. forney and l. f. wei , `` multidimensional constellations - part i : introduction , figures of merit , and generalized cross constellations , '' _
ieee j. select areas commun .
_ , vol . 7 , pp .
877892 , august 1989 .
c. b. peel , b. m. hochwald , and a. l. swindlehurst , `` a vector - perturbation technique for near - capacity multiple - antenna multi - user communications - part i : channel inversion and regularization , '' _ ieee trans .
_ , pp . 195 202 , jan . 2005 . | a simple scheme for communication over mimo broadcast channels is introduced which adopts the lattice reduction technique to improve the naive channel inversion method . lattice basis reduction helps us to reduce the average transmitted energy by modifying the region which includes the constellation points .
simulation results show that the proposed scheme performs well , and as compared to the more complex methods ( such as the perturbation method @xcite ) has a negligible loss .
moreover , the proposed method is extended to the case of different rates for different users . the asymptotic behavior ( snr@xmath0 ) of the symbol error rate of the proposed method and the perturbation technique , and also the outage probability for the case of fixed - rate users is analyzed .
it is shown that the proposed method , based on lll lattice reduction , achieves the optimum asymptotic slope of symbol - error - rate ( called the precoding diversity ) .
also , the outage probability for the case of fixed sum - rate is analyzed .
0.6 cm * communication over mimo broadcast channels using lattice - basis reduction * + 0.8 cm mahmoud taherzadeh , amin mobasher , and amir k. khandani 0.5 cm coding & signal transmission laboratory + department of electrical & computer engineering + university of waterloo + waterloo , ontario , canada , n2l 3g1 + |
the nature of the lowest - lying @xmath0 excited states ( denoted as @xmath3 state in the following ) in deformed nuclei has been a long standing problem in nuclear physics and studied by various approaches @xcite .
traditionally they have been considered to be collective excitations such as the @xmath2-vibration@xcite . in recent years
there has been calculations along this line based based on the algebraic collective model of rowe and co - workers @xcite as well as analytical solutions of the bohr hamiltonian with a certain kind of potential and a deformation - dependent mass term @xcite .
there are also calculations using the interacting boson model ( ibm ) with different truncated hamiltonians @xcite . in these calculations
the @xmath3 state is supposed to be a pure collective excitation , and its excitation energy can be fitted together with the yrast and the @xmath4-bands .
other studies do not assume that the @xmath5 states are purely collective .
for example , in the quasiparticle - phonon model ( qpm ) the @xmath5 excitations have been investigated in several rare - earth nuclei @xcite . in these
works the first @xmath5 excitations are described as one - phonon states , with the result that the phonons are built by several ( not many ) two - quasiparticle configurations .
a similar conclusion is also obtained in the studies of the @xmath5 excitations in @xmath6gd and @xmath7er with the projected shell model ( psm ) @xcite .
these studies reveal the relevance of the two - quasiparticle states in the @xmath3 states . however , the shortcoming of the latter models is that they do not work well in transitional nuclei .
recently , there has been also calculations of the low - lying excited states with the quasi - random phase approximation with skyrme forces @xcite and the five dimensional collective bohr hamiltonian with the gogny force @xcite . in the treatment of transitional nuclei ,
the generator coordinate method ( gcm ) is widely used @xcite . the gcm is a very flexible microscopic model which handles shape vibrations as the bohr hamiltonian but that at the same time allows to incorporate single particle degrees of freedom .
it has been applied with great success with effective forces , like skyrme @xcite , gogny @xcite or relativistic @xcite , to consider situations where the collective degrees of freedom play a relevant role , like shape coexistence in transitional nuclei . in recent years
the gcm has been used also with separable forces to calculate the first @xmath0 excitations in several gd , dy and er isotopes ( including the transitional @xmath8 isotones ) @xcite . in this work
the quadrupole deformation @xmath2 is used as the generator coordinate and after projecting on angular momentum and particle number the hill - wheeler equation @xcite is solved . in these calculations
the @xmath5 states are interpreted as shape vibrations .
the calculated excitation energies and transition probabilities of the @xmath3 states are in good agreement with the experimental data .
however , this study is restricted to isotopes with neutron number @xmath9 .
an attempt to extend this study to the heavier er isotopes with @xmath10 was not successful .
considering the above mentioned studies with the qpm and psm , one may guess that the failure of the gcm calculation in heavier er isotopes is due to the lack of two - quasiparticle states in the model space . with the inclusion of the two - particle states
, one could at least expect results as good as the psm calculations for the heavier er isotopes with @xmath10 .
in fact , there are several hints indicating that the first @xmath0 excitations in these er isotopes are not as collective as a pure shape vibration .
a measurement of the lifetime of the @xmath3 excitation in @xmath11er @xcite showed that the transition probability of the @xmath3 state in this nucleus is not as large as expected for a @xmath2-vibration .
the best candidate for the @xmath2-vibration in this nucleus is the third @xmath0 excited state . in @xmath7er ,
the measured e2 transition probability from the @xmath3 to the @xmath12 state is also very small @xcite . the @xmath13 value of @xmath14 w.u .
is much smaller than the value of tens of w.u .
expected for a @xmath2-vibration @xcite .
there are also some hints in this direction from the above mentioned calculations with the bohr model and the ibm . in the calculations with the bohr model , the parameters are determined by fitting the experimental spectra , and the transition rates are then calculated with these parameters .
it turns out that the @xmath13 rates from the @xmath5 states are always overestimated , sometimes by one order of magnitude .
this indicates that the calculated @xmath3 states are `` more collective '' than they should be . in the ibm calculations
the parameters are also determined by fitting the experimental data .
it was found in ref .
@xcite that the parameters determined in this way for @xmath7er do not follow the smooth trend as a function of the neutron number . however , if one assumes that the lowest collective @xmath0 excitation is the second @xmath0 excitation in @xmath7er , one would get smoothly varying parameters .
this also indicates that the first @xmath0 excitation in @xmath7er may not be a collective @xmath2-vibration .
the purpose of this paper is to generalize the above mentioned gcm calculation @xcite by simultaneously considering different nuclear shapes and their two - quasiparticle excitations in the gcm ansatz .
we apply the new theory to perform a systematic study of the first @xmath0 excitations in the er isotopes from @xmath15 to @xmath16 .
since the light isotopes are very soft in @xmath2 and the heavy ones strongly deformed , it is expected that the role played by the collective and the single particle degrees of freedom as well as their coupling will be elucidated .
it is expected that the inclusion of the two - quasiparticle states will improve the description of these @xmath0 states in the er isotopes with larger neutron numbers .
it should be mentioned that the gcm calculation with two - quasiparticle states has been used in ref .
@xcite in a very restricted configuration space for the study of ge and zn isotopes . in sect.[sect2
] we give a presentation of the theoretical framework and some numerical details .
the results are shown and discussed in sect.[sect3 ] .
we give a summary of our conclusion in sect.[sect4 ] .
in our calculations we use the separable pairing plus quadrupole hamiltonian @xcite @xmath17 where the operators @xmath18 , @xmath19 , @xmath20 and @xmath21 ( @xmath22 for protons and @xmath23 for neutrons ) are given by : @xmath24 with the quadrupole operator defined by @xmath25 and @xmath26 runs from @xmath27 to @xmath28 . as mentioned in the introduction as an ansatz for our wave functions we use the shape parameter @xmath2 to generate wave functions @xmath29 with this shape . for this purpose
we solve the hartree - fock - bogoliubov equation with constraints on the total quadrupole moment @xmath30 and the average particle number .
the wave function @xmath29 of the energy minimum for a given @xmath2 value is provided by @xmath31 with the lagrange multipliers @xmath32 , @xmath33 and @xmath34 determined by the constraining conditions : @xmath35 the relation between the quadrupole operator and the deformation parameter @xmath2 is given by @xmath36 with @xmath37 and @xmath38 fm . for each hfb vacuum @xmath29
there is a set of corresponding quasiparticle operators @xmath39 satisfying @xmath40 the second components of our gcm ansatz are the two - quasiparticle states , defined by @xmath41 finally , the hfb vacua and the two - quasiparticle states are projected onto good angular momentum and particle number .
thus the complete ansatz for our wave function has the form @xmath42 where the index @xmath43 runs over the set @xmath44 and @xmath45 labels the different states with angular momentum @xmath46 . in this calculation
the axial symmetry is preserved , and each of the hfb vacua and the two - quasiparticle states has a good quantum number @xmath47 .
therefore the summation over @xmath47 is omitted in eq .
the projection operators in the above expression are given by @xcite : @xmath48 for the angular momentum projection , and @xmath49,\ ] ] for the particle number projection .
it has been shown in refs .
@xcite that the particle number projection may cause troubles in the case that the exchange terms of the interaction are neglected .
for this reason in our calculations we will not neglect any term .
minimisation of the energy with respect to the coefficients @xmath50 leads to the hill - wheeler ( hw ) equation @xcite @xmath51 which has to be solved for each value of the angular momentum .
the gcm norm- and energy - overlaps have been defined as : @xmath52 to cope with the problem of the linear dependence one first introduces an orthonormal basis defined by the eigenvalues @xmath53 and eigenvectors @xmath54 of the norm overlap : @xmath55 this orthonormal basis is known as the natural basis and , for @xmath53 values such that @xmath56 , the natural states are defined by : @xmath57 obviously , a cutoff @xmath58 has to be introduced in the value of the norm eigenvalues to avoid linear dependences @xcite .
then , the hw equation is transformed into a normal eigenvalue problem : @xmath59 from the coefficients @xmath60 we can define the so - called collective wave functions @xmath61 that satisfy @xmath62 and are equivalent to a density of probability . concerning the number of two - quasiparticle states considered in the gcm ansatz of eq
. [ eqwf ] , in this calculation we set the following energy cutoff condition : @xmath63 where @xmath64 and @xmath65 represent the quasiparticle energies of the states @xmath66 and @xmath67 .
@xmath68 is the energy of the hfb state with deformation @xmath2 given by @xmath69 with @xmath70 the solution of eq .
[ hfb_eq ] and @xmath71 represents the deformation of the hfb minimum .
@xmath72 is the energy minimum of the potential energy as a function of @xmath2 .
the convergence with this cutoff has been checked and it is very good .
the hamiltonian of eq . [ hamiltonian ] has been used extensively in many psm calculations with great success @xcite .
we use the following strengths : the monopole pairing strength is @xmath73 , with the sign @xmath74 for neutrons and @xmath75 for protons and the quadrupole pairing strength is @xmath76 .
the quadrupole - quadrupole strength can not be taken from the psm because in that approach it is fixed by the experimentally measured @xmath2 deformation and it must be adjusted separately for each nucleus , see @xcite .
for the strength of the quadrupole force , we take @xmath77 . here
@xmath78 and @xmath79 can be either protons or neutrons . for protons @xmath80 and for neutrons @xmath81 .
furthermore , @xmath82 with @xmath83 mev and @xmath84 is the oscillator length ( @xmath85 ) .
the choice of the quadrupole strength is inspired by the work of baranger and kumar , ref .
these authors performed a very good analysis of the pairing plus quadrupole interaction @xcite .
they carried out very extended calculations along the nuclide chart within different approaches .
in particular they calculated ground states properties and potential energy surfaces in the hfb approach as well as in the beyond mean field approach in the framework of the bohr collective hamiltonian @xcite .
all these facts have contributed to the wide acceptance of the baranger kumar approach .
unfortunately this model is not appropriate for some beyond mean field approaches since the single particle model space used by baranger and kumar consist of only two mayor shells and it requires the use of a core to obtain good experimental agreement .
the core provides , among others , a contribution to the moment of inertia which is impossible to combine with an angular momentum projection . as discussed in ref .
@xcite with a single particle space of three shells there is no need to introduce a core . the effective charge for calculating the e2 transitions are @xmath86 for protons and @xmath87 for neutrons as in ref .
@xcite . in our calculations
three major shells are used as the single particle space . for rare earth nuclei
they are @xmath88 for neutrons and @xmath89 for protons @xcite . for the single - particle energies we use the @xmath90 and @xmath26 nilsson parameters of ref .
@xcite for neutron and protons .
furthermore , the i@xmath91 orbit in the @xmath92 shell for neutrons is shifted by @xmath93 . the h@xmath94 orbit in the @xmath95 shell for protons
is shifted by @xmath96 , and the rest orbits in the same shell are shifted by @xmath97 .
the purpose of these modifications of the single - particle energies is to obtain reasonable energy curves for either soft or well deformed er isotopes . without these modifications ,
the energy curves for @xmath98er will be extremely soft and obviously unreasonable .
it should be noticed that we do not determine our parameters by fitting the observed excitation energies of the @xmath3 states , as in other calculations @xcite .
the er isotopes studied in this work range from the soft ones ( with neutron number @xmath15 or @xmath99 ) to the well deformed ones ( up to @xmath16 ) . a very useful information is provided by the potential energy as a function of the deformation parameter @xmath2 . the hfb ground state energy as a function of the deformation parameter @xmath2 is provided by eq .
[ e_ground ] .
the corresponding energy curves for the different er isotopes are shown in fig .
[ erpes ] as dashed lines .
the hfb energies have been shifted in such a way that the energy of the @xmath100 value of each nucleus coincide with successive multiples of 6 mev .
the horizontal dotted lines are plotted as an energy reference . in fig .
[ erpes ] the transition from soft to rigid deformation is clearly visible . for @xmath98er
the energy curve is very soft with two shallow coexisting minima at the prolate and oblate side . in this case
one could expect large shape fluctuations and the excitation energy of the shape vibrational state should be relatively low . with increasing neutron number prolate and oblate minima develop around @xmath101 , with growing @xmath2 as n increases and a clear predominance of the prolate over the oblate shape .
large changes are found up to @xmath102 ( @xmath11er ) , for larger @xmath103 the potential energy curves look rather similar .
for the very well deformed isotopes one would expect small shape fluctuations around the prolate minimum , and the genuine shape vibrational excitations should appear at higher excitation energies .
these curves qualitatively agree with the energy curves obtained with the baranger and kumar calculations of ref .
@xcite . in the same figures
we also plot the particle number and angular momentum projected ( pnamp ) energies defined by @xmath104 where we have introduced @xmath105 .
the pnamp energies are relative to the corresponding hfb values .
the about 2 mev energy gain obtained in the pnamp approach with respect to the hfb one at @xmath100 is due only to the particle number projection . at larger deformation
we obtain about 4 mev energy gain from both projections .
the small energy gain obtained from the angular momentum projection at very small deformations as compared with the larger ones has as a consequence that the prolate and the oblate minima are deeper than in the hfb approach .
we have also plotted the values of the experimental deformations , where available @xcite , on the projected curves . as one
can see these values are located very close to the absolute minimum of the corresponding energy curves .
er calculated in the hfb approach , dashed lines , and in the pnamp one , continuous lines .
the bullet on each curve represents the experimental value of the deformation @xcite.,scaledwidth=50.0% ] the next step is the solution of the hill - wheeler equation , eq .
[ hw_nat ] , which provides the energies and wave functions of the ground and excited states . in fig .
[ gcmvs2qp ] the experimental values for the excitation energies of the @xmath3 states for the erbium isotopes are plotted ( black squares ) .
we observe an increasing behavior of the excitation energies with the neutron number @xmath103 up to @xmath102 , then a marked decrease up to @xmath106 and again an increase towards @xmath16 .
the rough structure of this behavior is understood in general terms : looking at a nilsson diagram we find a deformed shell closure at @xmath102 and a less pronounced one at @xmath16 . since for shell closures
one expects higher excitation energies , the mentioned closures roughly explain the observed behavior . in our model , see eq .
[ eqwf ] , we have two main degrees of freedom , namely , the shape vibrations and the two - quasiparticle excitations . to disentangle the different contributions we present calculations in three different model spaces .
the first model space includes hfb vacua of all deformations and does not include any two - quasiparticle state , i.e. the ansatz of eq .
[ eqwf ] is replaced by @xmath107 the third model space correspond to the full ansatz of eq .
[ eqwf ] and includes the hfb vacua of all deformations , as well as two - quasiparticle states built on each of them .
the results of the three calculations for the @xmath3 states are plotted in fig .
[ gcmvs2qp ] : red bullets for model space 1 , blue triangles for model 2 and magenta diamonds for model 3 .
we observe in models 1 and 2 a complementary behavior .
while the @xmath2 vibrations provide better results for @xmath108 , the two - quasiparticle degree of freedom provides better results for @xmath109 .
these quantitative results reinforce what one qualitatively would conclude from fig .
[ erpes ] .
interestingly , the results of model 3 demonstrate how both degrees of freedom combine to reproduce the experimental results quite well . in a more careful analysis one
observes that while for @xmath110 the shape fluctuations already provide rather good results and for @xmath111 the two - quasiparticle states are sufficient , for @xmath112 both degrees of freedom are clearly necessary .
energies calculated by three different model spaces .
the model spaces 1 , 2 and 3 correspond to the wave functions of eq .
[ eqwf1 ] , eq .
[ eqwf2 ] and eq .
[ eqwf ] , respectively .
the experimental data @xcite is also shown for comparison.,scaledwidth=50.0% ] bands of @xmath1er , compared with the experimental data .
the data are taken from @xcite.,scaledwidth=50.0% ] the eigenvalues of the hill - wheeler equation , eq . [ hw_eq ] , with the full model space for different angular momenta are shown in fig .
[ erspectrum ] for the yrast band and the band built on the @xmath3 states ( denoted as @xmath3 band in the following ) for the er isotopes . in the plots we have subtracted from the energies the rotational energy of a rigid rotor .
the rotational constants have been fixed in such a way that the energy of the @xmath12 state is zero for each isotope .
the values of a in the plots are @xmath113 @xmath114 for @xmath115 , respectively , and 0.013 for @xmath116 .
the levels have been classified into bands according to their transition probabilities .
notice that the members of the @xmath3 bands do not coincide in general with the lowest excited state with angular momentum @xmath46 , i.e. with @xmath117 state . for the @xmath3 bands and for very high angular momentum , the band structure may be dominated by two - quasiparticle and four - quasiparticle states . at present
the four - quasiparticle configurations are not included in our model space , therefore we only show these bands up to @xmath118 .
we observe that while the light er isotopes deviate from a rigid rotor the heavier ones get close to it with increasing mass number .
a similar behavior is observed for the excitation energy of the members of the @xmath3 band .
their excitation energies are very low for the light erbium isotopes and higher for the heavier ones .
this behavior is clearly related to the collectivity of the states .
the members of the @xmath3 band in the light erbium isotopes are collective whereas those in the heavier ones are less collective .
these facts will show up in the transition probabilities , see figs .
[ erbe2crossing]-[e0 ] . in general
we find a good agreement with the experimental data .
in particular the good agreement with the experiment for the low spin members of the yrast bands indicates that the ratio @xmath119 is correctly reproduced .
this ratio is plotted in fig .
[ fig_4/2 ] .
it clearly reflects the transition from soft nuclei to well deformed ones .
the description of this transition is only possible because of shape mixing within the framework of the gcm .
the higher spin members of the yrast bands are also well reproduced by the inclusion of the two - quasiparticle states in the gcm ansatz . with these two - quasiparticle states we are able to describe band - crossing phenomenon at high spins .
although our main topic are the @xmath3 states , we can see the effect of the gcm as well as two - quasiparticle states in the systematic reproduction of the yrast bands . for the erbium isotopes.the data
are taken from @xcite.,scaledwidth=50.0% ] not only the yrast states are well described , the excitation energies of the @xmath3 bands are also reasonably well reproduced as shown in fig .
[ erspectrum ] .
this is an indication that our model space contains the important degrees of freedom for the bands based on the @xmath3 states .
many possible excitation modes have been proposed over the years for these @xmath0 excitations . besides the @xmath2-vibrations and two - quasiparticle excitations mentioned in sect.[sect1 ] ,
there are other suggestions such as phonon excitation based on the @xmath4-vibration @xcite or pairing isomers @xcite .
it is very difficult to include all of these possible degrees of freedom in a model space .
based on the good agreement with the experimental results it seems that in these er isotopes , the important degrees of freedom are the shape vibration and the two - quasiparticle states .
it should also be noticed that the moments of inertia of the @xmath3 bands are also well reproduced , especially in @xmath120er , in which the @xmath3 bands have larger moments of inertia than the yrast bands .
values for the yrast bands and the @xmath3 bands of @xmath1er , compared with available data .
the data are taken from @xcite.,scaledwidth=48.0% ] another relevant information is provided by the intraband e2 transition probabilities . in fig .
[ erbe2 ] we plot the calculated @xmath121 values along the yrast and the @xmath3 bands together with the available experimental information .
we first discuss the yrast bands .
the calculated results are in reasonable agreement with the data except for @xmath98er .
in particular , the transition probability from the @xmath12 to the @xmath122 state increases roughly with the neutron number , corresponding to the transition towards the well deformed region .
the decrease of the experimental @xmath13 values at high spin , which is a result of the band crossing , is qualitatively reproduced by the theoretical results .
the calculated intraband @xmath13 for @xmath98er disagree with the observed data .
however , they are in qualitative agreement with the measured intraband @xmath13 in @xmath123dy , which is an isotone of @xmath98er
. this may indicate that the disagreement of the intraband @xmath13 in @xmath98er does not mean a general failure of our model for soft nuclei .
it may be related to some unknown reason in @xmath98er , which is not taken into account in our model .
it might be suggested that @xmath98er is @xmath4-unstable and we should include the @xmath4 degree of freedom to describe the intraband @xmath13 . the theoretical values for the intraband @xmath13 along the @xmath3 bands are also shown in the fig .
[ erbe2 ] .
they are of the same order as those of the yrast bands .
the decrease at around @xmath118 is due to the band crossing with a two - quasiparticle band .
values for @xmath1er , compared with experimental data in neighbouring nuclei with the same neutron numbers .
the data are from @xcite except for @xmath7er and @xmath124er , which are taken from ref .
@xcite and ref .
@xcite , respectively.,scaledwidth=50.0% ] besides the intraband e2 transitions , the interband ones taking place between the @xmath3 and the @xmath12 states are also important .
these transition rates are related to the quadrupole collectivity of the initial @xmath0 state .
therefore to see the performance of our model it is important to compare the calculated interband @xmath13 values with experimental data .
however , for the er isotopes these transitions probabilities has been only measured for @xmath11er @xcite , @xmath7er @xcite and @xmath124er @xcite . in order to examine the neutron number dependence of this interband @xmath13 , we also include the data measured in some sm , gd and yb nuclei with the same neutron number as the er isotopes .
the calculated interband @xmath13 values are shown in fig .
[ erbe2crossing ] together with the data measured in sm , gd , er and yb isotopes .
it is found that the interband @xmath13 values , if the @xmath125 dependence is ignored , decrease with growing neutron number .
this trend is reproduced by our calculations . for @xmath98er with @xmath15 ,
the calculated interband @xmath13 is around 100 w.u .
, which is of the order estimated for a @xmath2-vibration @xcite ( although the estimation is made with the assumption of well deformed nuclei ) .
this suggests that the @xmath3 state in this nuclei may be dominated by shape vibration .
on the other hand , the calculated interband @xmath13 for @xmath7er is less than 0.4 w.u . , which suggests that in this case the @xmath3 state may be dominated by two - quasiparticle excitations .
these suggestions are in accordance with what was inferred from fig .
[ gcmvs2qp ] .
note that the variation of the interband @xmath13 within this nuclei is as large as three orders of magnitude , which indicates that these @xmath0 excitations are of very different structure .
this large variation range has been well reproduced by our calculation , which is another justification of our gcm+2qp model space .
er , compared with the experimental data where available .
the data are taken from @xcite for @xmath126er and from @xcite for @xmath11er.,scaledwidth=50.0% ] another important observable are e0 transitions .
they provide important information about the mixing content of wave functions @xcite . in the absence of mixing one
expects pure configurations and thereby small matrix elements and larger ones in the presence of mixing .
the matrix element of the monopole transition @xmath127 from the state @xmath128 to the @xmath129 is given by @xmath130 with @xmath131 for protons and @xmath132 for neutrons .
the states @xmath128 and @xmath129 are given by eq .
the e0 strength is usually measured by the dimensionless quantity @xmath133 related to @xmath127 by @xmath134 where @xmath135 is the nuclear radius and @xmath38 fm . in fig .
[ e0 ] we display the @xmath133 values for the transitions from the first two members ( i.e. , @xmath136 and @xmath137 of the @xmath5 band to the corresponding states of the ground band for the erbium isotopes together with the experimental values .
the theoretical values for @xmath138 , as expected , are very similar to the @xmath139 ones .
we observe an exponential decrease of the strength with increasing mass number as in the previous case of @xmath140 displayed in fig . [ erbe2crossing ] .
this behaviour clearly reflects the different stages of collectivity of the members of the @xmath3 band .
the agreement with the experimental results for @xmath141er and @xmath7er is good whereas for @xmath11er we obtain a peak which is not observed experimentally .
this is probably related to the change of the most relevant two - quasiparticle configuration from neutrons in @xmath141er to protons in @xmath11er , see table [ tab ] .
er and @xmath6er for transitions from the @xmath3 band to the ground band compared with the experimental data where available .
the data are taken from @xcite.,scaledwidth=50.0% ] now we want to analyse the mixing of the unperturbed @xmath3 band and the ground band in the rotor model .
the simplest model for mixing of rotational bands was provided by mikhailov for well deformed nuclei @xcite .
this model assumes that the origin of the mixing of the two bands is the coriolis force . assuming that the intrinsic quadrupole moment of the bands is the same one obtains @xcite the simple formula @xmath142 ,
\label{eq : mik}\ ] ] the superscript @xmath143 denotes the excited band , in our case the @xmath3 band and @xmath144 the ground band . if the assumptions of the model are fulfilled @xmath145 and @xmath146 are constants @xcite and the plot , the so - called mikhailov plot , of the left hand side of eq .
[ eq : mik ] as a function of @xmath147 should be a straight line .
notice that for each angular momentum @xmath148 the final states can have @xmath149 .
an isotope that satisfies the requirements of the model and for which there are some experimental results is @xmath124er . in fig .
[ mikhailov ] we represent the calculated transition probabilities together with the experimental data .
we obtain a qualitative agreement with the experimental data and the fit of a straight line to the calculated points is a good approximation indicating the validity of the model for this nucleus . in contrast
we also display in the same figure the calculated values for the nucleus @xmath6er . in this case
we find that the mikhailov predictions are not satisfied indicating that the mixing of the bands has a different origin and that the assumptions of the model are not satisfied , see below .
another analysis often used in the analysis of vibrational bands @xcite is performed by mixing unperturbed bands but without assuming any specific form for the perturbation , the mixing parameters being adjusted as to obtain a good fit to the experiment .
we have performed such an analysis to our data in the terms of ref .
@xcite for the nuclei @xmath124er and @xmath6er and the results , not quoted here , are similar to the one based on the mikhailov plots . for the description of @xmath124er
it is enough to mix the unperturbed ground and the @xmath150 bands whereas for @xmath6er further bands will be required .
states in @xmath1er .
the thick continuous black lines represent @xmath151 and the thick dashed red lines @xmath152 as a function of the deformation parameter @xmath2 for the ground states .
the thin blue and magenta lines represent the corresponding amplitudes for the @xmath3 states.,scaledwidth=50.0% ] in the discussions above we compared the results of the calculations with the measured observables and demonstrated the reliability of our method . in the following
we will focus on the collective wave functions . according to eq .
[ norm_coll_wf ] , the normalization of the collective wave function eq .
[ coll_wf ] for a given state @xmath45 and angular momentum @xmath46 can be written as @xmath153 the quantity @xmath154 is related to the contribution of the @xmath2-constrained ground state ( hfb vacuum ) , @xmath155 , to the amplitude of the collective wave function , while @xmath156 is related to the amplitude of the two - quasiparticle state , @xmath157 .
the separation of eq .
[ coll_sep ] makes sense because it allows to differentiate the collective degrees from the single particle ones .
we concentrate on the wave functions of the ground and first excited state , @xmath158 and @xmath28 , respectively , and for @xmath159 . in fig .
[ ercollective-2 ] we plot the quantities @xmath151 and @xmath152 as a function of the deformation parameter @xmath2 for the different erbium isotopes .
the hfb vacua contributions to the @xmath122 state are represented by thick black solid lines . in the top panels
one observes that for the @xmath160er isotopes there is a contribution from the oblate part . for the heavier isotopes
only the prolate part is different from zero .
furthermore it can be seen that the width of the amplitude distribution gets smaller as the neutron number increases corresponding to the depth of the prolate well .
this is in accordance with the conclusion drawn from the energy curves shown in fig .
[ erpes ] .
the contribution of the two - quasiparticle states to the wave function of the @xmath122 states , thick dashed red lines , is negligible for the oblate part and rather small for the prolate part of all isotopes though it increases with growing neutron number .
that means that for the ground states the relevance of the two - quasiparticle states is always small .
the wave functions of the ground states are dominated by the hfb vacua , as one would expect .
the amplitudes for the @xmath3 states are also plotted in fig .
[ ercollective-2 ] .
the hfb vacua part is represented by thin solid blue lines and the two - quasiparticle part by short dashed thin magenta lines .
notice that the two - quasiparticle part of the @xmath3 states peaks close to the @xmath2-values where the vacuum part of the @xmath122 states does as it should be .
a first glance reveals that now the two - quasiparticle parts play a much more important role than for the @xmath122 states . as a matter of fact
the hfb vacua contributions are larger than the two - quasiparticle part only for the @xmath161er isotopes while for @xmath162er the latter one predominates .
interestingly the hfb vacua part contribution to the wave functions displays a node structureer the thin solid blue lines have two nodes , but it is just because some of the values around @xmath100 are very small and seems to vanish .
the real nodes corresponds to positive @xmath2 values . ] for @xmath163er characteristic for shape vibrations @xcite . for @xmath164er
the hfb vacua contributions are very small and do not show a nodal structure .
the picture that emerges from this analysis is that the @xmath3 states in the light erbium isotopes are dominated by shape vibrations while in the heavier ones the two - quasiparticles excitations predominate .
though at first sight it seems a smooth change from one nucleus to the other , small details of the wave function may cause large changes in some observables .
thus the increase in the @xmath165 transition probability from @xmath7er to @xmath124er , see fig .
[ erbe2crossing ] , can be explained noticing that the hfb vacua contribution to the wave function of the @xmath122 state state is the same as the @xmath122 . ]
peaks at higher deformation than the corresponding to @xmath7er . the span of several orders of magnitude observed in fig .
[ erbe2crossing ] for the interband @xmath13 transition probabilities from the @xmath3 states is not surprising if one considers the large variety of wave functions .
but for angular momentum @xmath166 , top panels , and @xmath167 , lower panels , and two nuclei @xmath6er , left panels , and @xmath11er , right panels.,scaledwidth=50.0% ] to know the evolution of the composition of the wave functions with the angular momentum we have plotted in fig .
[ ercollective - j ] the same quantities as in fig .
[ ercollective-2 ] but for angular momenta @xmath166 and @xmath167 and only for the nuclei @xmath6er and @xmath11er . for the soft @xmath6er and for the @xmath168 member of the yrast band
we observe a decrease of the hfb vacua contribution , as compared with @xmath169 , specially for small prolate deformations and in particular for the oblate part which vanishes ( not shown here ) .
the two - quasiparticle contribution now increases vigorously and its maximum is about half of the hfb vacua .
concerning the @xmath170 state of the @xmath3 band we observe the same tendency but now the increase of the two - quasiparticle part is even stronger than before . for @xmath171 and for both states
the hfb vacua contribution diminishes even more and the two - quasiparticle part increases .
furthermore for the yrast state we observe a shift of the hfb vacua contribution to larger deformations and another of the two - quasiparticle part to smaller deformations . for @xmath11er , right panels ,
the role of the quasiparticles is more important than for @xmath6er , in particular for the member of the @xmath3 band . in fig .
[ ercollective-2 ] we have presented the contribution of the two - quasiparticle states for each deformation @xmath2 for the erbium isotopes . to evaluate the relevance of the two - quasiparticle excitations for a given nucleus we need the total contribution of the two - quasiparticle to all @xmath2 s , i.e. , the last term of eq .
[ coll_sep ] .
this term is represented in fig .
[ er2qpweight ] for the er isotopes .
the full squares are for the ground state and the bullets for the @xmath3 state .
it is shown that the relevance of the two - quasiparticle excitation for the @xmath3 states increases with the neutron number up to @xmath102 and then remains more or less constant .
in particular , the @xmath3 state is to a large extent a two - quasiparticle excitation in the@xmath172er isotopes .
states in @xmath1er , measured by the summation of the absolute square of the collective amplitudes over the two - quasiparticle states and all shapes .
the contributions from the most relevant two - quasiparticle configurations are also shown.,scaledwidth=50.0% ] we are also interested to know the indices @xmath173 of the pair of quasiparticles providing the largest contribution to the sum just discussed .
this contribution is given by @xmath174 and plotted in fig .
[ er2qpweight ] as triangles and diamonds for the ground state and the @xmath3 state , respectively .
it is seen that the most relevant two - quasiparticle configuration accounts for more or less half of the total contribution of all two - quasiparticle configurations .
such a structure of the two - quasiparticle combinations is in qualitative accordance with the results of the qpm @xcite and psm @xcite calculations .
the nilsson quantum numbers of the most relevant two - quasiparticle configurations are listed in table.[tab ] .
they are built by time - reversal conjugate states and are the lowest two - quasiparticle states .
this is also in accordance with the qpm and psm results .
.[tab ] the most relevant two - quasiparticle configurations in the @xmath3 states in @xmath1er . [ cols="^,^",options="header " , ]
in this work the generator coordinate method ( gcm ) is generalised by including two - quasiparticle excitations built on generating functions of different deformations as well as the ground states . in this way the collective vibrational shape fluctuations and the non - collective quasiparticle excitation are taken into account within a common framework .
with such a model space and the separable pairing plus quadrupole hamiltonian , calculations for er isotopes with neutron number @xmath175 have been performed .
the strengths of the pairing and quadrupole force used in the calculation are determined in a systematic way , and the single - particle energies are adjusted in order to get reasonable potential energy curves .
the calculated spectra of the yrast bands and the bands built on the @xmath3 states are in good agreement with experimental data .
the intraband ( along the yrast bands ) and interband ( from the @xmath3 states to the @xmath12 states ) e2 transitions are also reasonably well reproduced ( except for the intraband @xmath13 in @xmath98er ) .
the transition from soft to rigid deformation with the increasing neutron number is also well reproduced by our calculations .
the properties of the @xmath3 states are studied in detail .
the good agreement between the calculated excitation energies of these @xmath0 states and the experimental data suggests that the two degrees of freedom , i.e. shape vibration and two - quasiparticle excitation , are sufficient for the description of these @xmath0 states .
the large variation of the @xmath176 values , shown in our calculated results as well as in the measured experimental data for several neighbouring nuclei , indicates that the role played by these two degrees of freedom in the @xmath3 states are very different for different isotopes .
studies of the collective wave functions of these @xmath0 states show that in @xmath98er the @xmath3 state is dominated by shape vibration , while the two - particle excitation only plays a minor role . on the other hand , in @xmath7er the @xmath3 state
is dominated by two - quasiparticle excitations , while the contribution from the shape vibration is almost negligible .
the situation in other er isotopes lie between these two extremes , in which both degrees of freedom are necessary for a reasonable description . generally speaking ,
the relevance of two - quasiparticle excitations increase with the neutron number in the er isotopes studied .
the contribution of shape vibration decrease with the neutron number .
the exception is @xmath124er in which the shape vibrations still make a larger contribution to the @xmath3 state as compared with @xmath7er .
one limitation of this calculation is that the axial symmetry is preserved and the @xmath4-degree of freedom is not taken into account . as a consequence we do not have @xmath4-vibrational states in our calculations .
therefore we could not study the e2 transition between the @xmath3 states to the @xmath4-vibrational state , which is also an important quantity discussed in the studies of the nature of the @xmath3 states . because of these we shall include the @xmath4 degree of freedom in our future work .
we would lastly mention that a potential application of the theory would be the calculation of neutron pair transfer to investigate the @xmath150 bands @xcite .
two neutron transfer in deformed nuclei is usually calculated in the hfb or qrpa theories @xcite . for deformed superfluid nuclei the particle number and the angular momentum are not conserved quantities and the use of a symmetry conserving theory like the one presented in this work may reveal important clues on the pair transfer phenomenon .
fang - qi chen gratefully acknowledges discussions with prof .
this work was supported by the spanish ministerio de economa y competitividad under contracts fpa2011 - 29854-c04 - 04 , fpa2014 - 57196-c5 - 2-p .
99 p. e. garrett , j. phys .
g : nucl . part .
* 27 * , r1 ( 2001 ) a. bohr and b. r. mottelson , _ nuclear structure _ , world scientific , singapore , 1988 d.j .
rowe , t.a .
welsh , and m.a .
caprio , phys .
c 79 , 054304 ( 2009 ) d. bonatsos , p. e. georgoudis , d. lenis , n. minkov , c. quesne , phys .
c * 83 * , 044321 ( 2011 ) d. bonatsos , p. e. georgoudis , n. minkov , d. petrellis , c. quesne , phys . rev .
c * 88 * , 034316 ( 2013 ) .
chou , n. v. zamfir , r. f. casten , phys .
c * 56 * , 829 ( 1997 ) e. a. mccutchan , n. v. zamfir , r. f. casten , phys .
c * 69*,064306 ( 2004 ) i. bentley , s. frauendorf , phys .
c * 83 * , 064322 ( 2011 ) s. zerguine , p. van isacker , a. bouldjedri , phys . rev .
c * 85 * , 034331 ( 2012 ) v. g. soloviev , a. v. sushkov , n. yu , shirikova , phys .
c * 51 * , 551 ( 1995 ) n. lo iudice , a. v. sushkov , n. yu .
shirikova , phys .
c * 72 * , 034303 ( 2005 ) n. lo iudice , a. v. sushkov , n. yu .
shirikova , phys . rev .
c * 70 * , 064316 ( 2004 ) yang sun , ani aprahamian , jing - ye zhang , ching - tsai lee , phys .
c * 68 * , 061301(r ) ( 2003 ) d. bucurescu et al , phys .
c 73 , 064309 ( 2006 ) j. terasaki and j. engel , phys .
c * 84 * , 014332 ( 2011 ) n. hinohara , m. kortelainen , and w. nazarewicz , phys .
c * 87 * , 064309 ( 2013 ) j .- p .
delaroche , m. girod , j. libert , h. goutte , s. hilaire , s. pru , n. pillet , and g. f. bertsch , phys .
c * 81 * , 014303 ( 2010 ) .
p. ring , p. schuck , _ the nuclear many - body problem _ ( springer , new york , 1980 ) m. bender , p .- h .
heenen and p .-
reinhard , rev .
phys . * 75 * , 121 ( 2003 ) .
rodrguez and j.l .
egido , phys .
* 99 * , 062501 ( 2007 ) .
t. niki , d. vretenar , and p. ring , phys .
c * 74 * , 064309 ( 2006 ) .
fang - qi chen , yang sun , peter ring , phys .
c * 88 * , 014315 ( 2013 ) d. l. hill and j.a .
wheeler , phys . rev . * 88 * , 1102 ( 1953 ) p. e. garrett , m. kadi , c. a. mcgrath , v. sorokin , min li , minfang yeh , phys . lett .
b 400 , 250 ( 1997 ) t. hrtlein , m. heinebrodt , d. schwalm , c. fahlander , eur . phys .
j. a 2 , 253 ( 1998 ) m. didong , h. mther , k. goeke , a. faessler , phys .
c * 14 * , 1189 ( 1976 ) h. mther , k. goeke , k. allaart , a. faessler , phys .
c * 15 * , 1467 ( 1977 ) k. hara , y. sun , int .
phys e * 4 * , 637 ( 1995 ) f. doenau , phys .
c 58 ( 1998 ) 872 .
m. anguiano , j. l. egido and l.m .
. phys . * a 696 * , 467 - 493 ( 2001 ) j. m. yao , j. meng , p. ring , and d. pea arteaga , phys .
c * 79 * , 044312 ( 2009 ) m. baranger and k. kumar , nucl .
a * 110 * , 529 ( 1968 ) k. kumar and m. baranger , nucl .
a * 110 * , 529 ( 1968 ) k. kumar and m. baranger , nucl .
phys . a * 92 * , 608 ( 1967 ) t. bengtsson , i. ragnarsson , nucl .
a 436 , 14 ( 1985 ) s. raman , c. w. nestor jr , and p. tikkanen atom .
data nucl .
78,1 ( 2001 ) www.nndc.bnl.gov/ensdf/ r. f. casten , p. von brentano , phys .
c 50 , r1280(r ) ( 1994 ) i. ragnarsson , r. a. broglia , nucl .
a 263 , 315 ( 1976 ) m. baglin , nucl .
data sheets 109 , 1103 ( 2008 ) d. d. dijulio _ et al _ , eur .
j. a * 47 * , 25 ( 2011 ) j.l .
wood , k. heyde , w. nazarewicz , m. huyse , p. van duppen , phys .
215 , 101 ( 1992 ) j. l. wood , e. f. zganjar , c. de coster , k. heyde , nucl .
a 651 323 ( 1999 ) t. kibedi , r. h. spear , atom .
data nucl .
89 77 ( 2005 ) ( for er166 ) v. m. mikhailov , izv .
nauk sssr , ser .
28 , 308 ( 1964 ) [ transl .
ussr , phys .
28 , 225 ( 1964 ) ] ; 30 , 1334 ( 1966 ) [ transl . 30 , 1392 ( 1966 ) ] .
r. f. casten , _ nuclear structure from a simple perspective _ ( oxford university press , new york , 2001 ) l. l. riedinger , n. r. johnson and j. h. hamilton , phys .
179 * , 1214 ( 1969 ) r. a. broglia , o. hansen and c. riedel , _ advances in nuclear physics _ vol . 6 , ed .
m. baranger and e. vogt ( new york : plenunm ) p. 287
j. l. egido and j. o. rasmussen , phys .
c36 , 316 ( 1987 ) | the ground and first excited @xmath0 states of the @xmath1er isotopes are analyzed in the framework of the generator coordinate method .
the shape parameter @xmath2 is used to generate wave functions with different deformations which together with the two - quasiparticle states built on them provide a set of states . an angular momentum and particle number projection of the latter
spawn the basis states of the generator coordinate method . with this ansatz and using the separable pairing plus quadrupole interaction we obtain a good agreement with the experimental spectra and e2 transition rates up to moderate spin values .
the structure of the wave functions suggests that the first excited @xmath0 states in the soft er isotopes are dominated by shape fluctuations , while in the well deformed er isotopes the two - quasiparticle states are more relevant . in between both degrees of freedom are necessary . |
the authors would like to thank the w.m .
keck foundation for the financial support , akos vertes who provided the laser for our experiments , infrared fiber systems , silver spring , md , who provided the optical fibers for this study , william rutkowsky for helping with the instrumentation , andrew gomella and craig s pelissier for helping with the detector calibration process , jyoti jaiswal , mary ann stepp and gauri tadvalkar for helping with the cell culture , and alexander jeremic for providing us the facility to culture cell samples . | we report the near - field ablation of material from cellulose acetate coverslips in water and myoblast cell samples in growth media , with a spot size as small as 1.5 @xmath0 m under 3 @xmath0 m wavelength radiation .
the power dependence of the ablation process has been studied and comparisons have been made to models of photomechanical and plasma - induced ablation .
the ablation mechanism is mainly dependent on the acoustic relaxation time and optical properties of the materials .
we find that for all near - field experiments , the ablation thresholds are very high , pointing to plasma - induced ablation as the dominant mechanism .
near - field scanning optical microscopy ( nsom)@xcite is a promising technique that overcomes the diffraction limit of conventional optical microscopy @xcite and by doing so has created a number of potential applications in biological imaging .
the combination of nsom for ablation with mass spectrometry is of particular interest to obtain detailed molecular information with spatial resolution better than that of the conventional optical spectrometry .
a step in this direction , ultraviolet - nsom - based mass spectrometry with a lateral resolution of 170 nm in ambient conditions , has demonstrated soft ablation capabilities @xcite .
an improvement would be ablating in the infrared rather than the uv regime so that the native water in a cell plays the role of the ablation matrix due to its strong absorption at 2940 nm @xcite . by this approach
, cells can be probed _ in - vivo _ or _ in - vitro _ , but in the far - field , the spatial resolution is limited by diffraction effects and by the quality of available optics to a spot size of about 50@xmath1 .
there are in the literature a number of reports of ablation of conventional solids@xcite and organic molecules @xcite .
there are fewer in which the energy delivered to the sample is characterized well enough to measure ablation thresholds @xcite . in these ,
the ablation thresholds are many orders of magnitude larger than would be expected for far - field approaches .
schematic diagram of nsom setup , modified for ir ablation ( a ) ir fiber tip , attached to the tuning fork , scale bar:1 mm .
( b ) sem image of an ir fiber tip etched using modified tube etching technique @xcite , scale bar:200 @xmath0m.,width=268 ] we report the first integration of ir with near - field techniques applied to the ablation of live cells .
we obtain ablation features as small as 1.5 @xmath1 under 3 @xmath1 wavelength radiation , in hard and soft materials , and in air and water environments .
the ablation threshold and fluence dependence of these processes are discussed here .
of particular interest is the successful ablation of cellular samples in growth media , which we describe in this paper .
the experiments described here are performed with an nsom apparatus , modified for operation in the mid - ir region @xcite . in all experiments ,
the mid - ir output of a nd : yag laser , coupled to a pumped - optical - parametric oscillator ( opo ) ( set to 2940 nm , 100 hz , 5 ns pulse width ) is used @xcite . a schematic diagram of the experimental setup is shown in fig.[scan1 ] .
a schwarzschild objective is used to focus the laser beam and couple it to an nsom tip , which is fabricated by tube etching @xcite from a short stub of germanium - based glass fiber @xcite .
afm images of the sample , using normal force feedback , are obtained using the same tip . to calibrate the fluence through the tip
, laser power measurements are made using a pbse photoconductive detector @xcite positioned beneath the fiber tip in place of the sample . to set a baseline for understanding the ablation process , material from a cellulose acetate plastic coverslip is ablated in air in the far - field configuration .
the ablation threshold for the sample is determined by conducting a series of single - spot ablations . in these
, the laser beam is focused at normal incidence on the surface of the sample with a repetition rate of 100hz .
the sample is simultaneously observed under the microscope , and the ablation threshold is defined to be the fluence below which we do not observe visible permanent modification of the surface of the sample . in later experiments , a cellulose acetate plastic coverslip is ablated in air and in water by near - field techniques .
laser ablation in the water medium is achieved by immersing the cellulose acetate cover slip in water , 2 mm - deep , and then coupling the laser beam into the tip , which is partially immersed in the water .
the near - field ablation threshold is determined by comparing the afm images for different laser powers .
afm images of the cellulose acetate coverslip before and after ablation in water are shown in fig .
[ scan ] .
3d topographic image of a plastic cover slip in water ( a ) before and ( b ) after near - field ablation .
( c ) profile of the crater , whose size is 1.25 @xmath0 m , as measured by the full - width at half maximum ( fwhm ) of the profile .
scalebar : 2 @xmath0m , width=336 ] myoblast cells are studied in specially prepared sample holders formed from a pdms ring molded directly on a glass cover slip to hold the sample and surrounding media during the process of scanning and ablating .
cells from the c2c12 cell line are cultured directly in these molded wells .
plated cells attach to the glass surface of the sample holders during the incubation period .
topographic images of myoblast cells in media before and after ablation are shown in fig .
[ myo2 ] . under the near - field laser illumination conditions of our experiment
, it is clear from the afm images that the damage is localized to a well defined region , approximately 2.5 @xmath0 m in size . from the profile images of the craters
, we can see that there is some redeposition of the ablated material in the vicinity of the ablation crater , as is the case with the cellulose acetate material .
topographic image of myoblast cells in media ( a ) before ablation .
( b ) after ablation .
( c ) profile of the crater , whose size is 2.5 @xmath0 m .
scalebar : 10 @xmath0 m .
, width=336 ] in our study , the ablation thresholds for a single material , cellulose acetate , varied greatly , from 0.22 j/@xmath2 to 301 j/@xmath2 as seen in table i. with such a big disparity between the near - field and the far - field ablation thresholds , the question naturally arises about whether the mechanism has differs . in the far - field regime ,
the low thresholds are characteristic of a photomechanical ablation mechanism that starts with thermoelastic stresses arising from local expansion of the sample upon high - intensity laser heating@xcite .
the thermo - accoustic process is favored when the pulse width and spot size are suitable to locally contain the thermal and accoustic energy deposited by the beam @xcite .
for example , the region of the sample ablated by each pulse is set by the thermal diffusion length @xmath3 @xcite , @xmath4 where @xmath5 is the laser pulse width , and @xmath6 is the thermal diffusivity of the material . for the cellulose acetate and myoblast samples , the thermal diffusion length , @xmath3 , is approximately 25 nm , small enough to minimize lateral thermal damage .
heating leads to local thermal expansion , and ablation occurs when conditions for stress or inertial confinement are achieved .
that is , for laser pulse durations shorter than a characteristic acoustic relaxation time@xcite , @xmath7 here @xmath8 is the speed of sound , and @xmath9 is the optical absorption coefficient of the material
. a theoretical model for the photomechanical mechanism has been developed by a number of research groups @xcite , which indicates the possibility for thermo - accoustic ablation in our samples . for cellulose acetate , the speed of sound in the sample is @xmath10 m / s and the absorption coefficient is @xmath11 m@xmath12 , to yield an acoustic relaxation time @xmath13 of about @xmath14 ns .
this is much longer than the laser pulse width and satisfies the inertial confinement condition @xmath15 . for cellular samples in growth media
, we expect the properties of water to determine the length and time scales for the ablation process .
the absorption coefficient of water is of the order of @xmath16 m@xmath17 and the velocity of the sound is 1497 m / s , for which the acoustic relaxation time is equal to 0.5 ns .
comparing to our laser pulse width of 5 ns , the condition for the stress confinement , @xmath18 , is not satisfied , and photomechanical stress is not likely to be a factor , or at least will be reduced significantly .
.results of ablation threshold studies on cellulose acetate and myoblast cells in far - field ( ff ) and near - field ( nf ) regime .
each of these has been reproduced at least five times .
the laser wavelength is 2940 nm , and the pulse width is 5 ns [ cols="^,^,^,^ " , ] the measured thresholds for the near - field ablation for all cases reported here and for a number reported in the literature @xcite are much higher than expected for photomechanical mechanism .
this is not surprising for the ablation of cells but is so for the cellulose acetate . in both cases
the measured ablation threshold points to a second , higher energy mechanism , plasma induced ablation @xcite . here , the intense laser beam ionizes molecules in the sample , and subsequent collisions within the ablated plume and with ambient gas molecules lead to the formation of a hot dense plasma above the sample surface .
the vapor plume then expands perpendicularly from the surface and is further ionized by incoming radiation .
consequently , the plasma absorbs more energy from the trailing part of the laser pulse by photoionization or inverse bremsstrahlung processes @xcite .
near - field ablation alters the conditions for plasma - induced ablation .
the laser beam s path through the plasma is a narrow gap of only about 10 nm before reaching the sample , and hence the interaction with the plasma is weaker .
furthermore , the presence of a sharp probe in the vicinity of the sample leads to localization of the ionizing field and the probe itself physically blocks the expansion of the plume in the direction normal to the sample , leading to the movement of the plume laterally away from the ablated spot @xcite . in water ,
the conditions for plasma - induced ablation are enhanced and we observe a decrease in the threshold fluence .
this change is primarily due to the different optical properties of the water , namely the refractive index , which leads to an overall reflectivity decrease from @xmath19 to @xmath20 upon immersion of the sample in water .
( the reflectivity is calculated in the usual way , r = @xmath21 , here @xmath22 = 1.48 , the refractive index of cellulose acetate . )
overall , the absorptivity of water - sample system is larger than that of the air - sample system @xcite .
also , in water , the plasma is confined and there is a delay in its expansion .
hence , the induced pressure created by the laser sample interaction is much greater in water than in air and the plasma is compressed @xcite , an effect that is further enhanced by near - field confinement of the light . as has been observed by a number of groups @xcite high fluences
are required to achieve near - field ablation in organic @xcite and metallic samples @xcite .
likewise , higher fluences lead to plasma induced ablation conditions for the myoblast cells . due to the presence of water above and below the cell membrane , the induced plasma pressure in the myoblast cell samples
is enhanced compared to that of the cellulose acetate sample .
this effect is further amplified by the near - field confinement of the light .
the reflectivity of myoblast cell samples , in growth media , is found to be @xmath23 , assuming @xmath24,@xcite which helps to explain the smaller threshold fluence of myoblast cells compared to cellulose acetate . in conclusion , as a first step towards _ in - vitro _ mass spectral analysis of biological samples , we have successfully demonstrated the ablation of hard and soft materials in liquids , and , in particular , of biological samples in nutrient media . to understand the ablation mechanism , we have measured the ablation threshold of different types of samples .
the mechanisms that are involved in the ablation process are mainly photomechanical and plasma induced ablation for samples in air . since the acoustic relaxation time of water
rich samples is of the order of the laser pulse width , the ablation mechanism in cellular samples is easily seen to be plasma induced rather than photomechanical . for cellulose acetate ,
the necessity for this large increase in the near - field ablation threshold in air is not as clear .
however , for cells in media , their thermal and acoustic properties lead to conditions that favor the plasma - induced mechanism over the photomechanical one . |
observations of protoplanetary disks around young stars in star - forming regions support the widely accepted idea that planet formation is a common by - product of star formation ( e.g. beckwith 2000 ) .
our solar system may be a typical planetary system in which earth - like planets accrete near the host star from rocky debris depleted of volatile elements , while giant gaseous planets accrete in the ice zones ( @xmath1 au ) around rocky cores ( boss 1995 , lissauer 1996 ) .
when the rocky cores in the ice zones reach a critical mass ( @xmath2 ) runaway gaseous accretion ( formation of jupiters ) begins and continues until a gap in the protoplanetary disk forms or the disk dissipates ( papaloizou and terquem 1999 , habing 1999 ) .
the presence of metals is then a requirement for the formation of both earths and jupiters .
we can not yet verify if our solar system is a typical planetary system or how generic the pattern described above is .
the doppler technique responsible for almost all extrasolar planet detections ( mayor 1995 , butler 2000 and references therein ) is most sensitive to massive close - orbiting planets and is only now becoming able to detect planetary systems like ours , i.e. , jupiters at @xmath1 au from nearby host stars .
the doppler technique has found more than 40 massive ( @xmath3 ) extrasolar planets in close ( @xmath4 au ) , often eccentric orbits around high metallicity host stars ( schneider 2000 ) .
i refer to _ all _ of these giants as ` hot jupiters ' because of their high mass and proximity to their central stars .
approximately 5% of the sun - like stars surveyed possess such giant planets ( marcy and butler 2000 ) .
thus there is room in the remaining 95% for stars to harbour planetary systems like our solar system .
it is not likely that giant planets have formed in situ so close to their host stars ( bodenheimer 2000 ) .
it is more likely that after formation in the ice zone , these giants moved through the habitable zone , destroyed nascent earths ( or precluded their formation ) and are now found close to their host stars ( lin 1996 ) .
how this migration occurred is an active field of research .
however , independent of the details of this migration , recent detections of extrasolar planets are telling us more about where earths are not , than about where earths are .
the aims of this paper are to use the most recent observational data to quantify the metallicity range compatible with the presence of earths and estimate the age distribution of earth - like planets in the universe .
the outline of the analysis is as follows : 1 .
compare the metallicity distribution of stars hosting hot jupiters with the metallicity distribution of stars in the solar neighborhood to obtain the probability of hosting hot jupiters ( and therefore the probability of destroying earths ) 2 .
assume that starting in extremely low metallicity stars , the probability to produce earths increases linearly with metallicity ( this assumption is discussed in section 2.2 ) 3 .
combine items 1 and 2 above to estimate the probability of harbouring earths as a function of metallicity ( fig . 1 ) 4 .
use current estimates of the star formation rate in the universe ( fig .
2a ) and observations of high redshift metallicities to estimate the metallicity distribution of star - forming regions as a function of time ( fig .
combine 3 and 4 above to estimate the age distribution of earth - like planets in the universe ( fig .
figure 1 shows the metallicity distribution of 32 stars hosting hot jupiters whose metallicities have been published ( gonzalez 2000 , table 1 , butler 2000 , table 4 and references therein ) .
the fact that these hosts are significantly more metal - rich than sun - like stars in the solar neighborhood has been reported and discussed in several papers , including gonzalez ( 1997 , 1998 , 2000 ) , ford , rasio and sills ( 1999 ) , queloz ( 2000 ) , butler ( 2000 ) .
the metallicity distribution of sun - like stars in fig .
1 is a linear combination of similar histograms in sommer - larsen ( 1991 ) and rocha - pinto and maciel ( 1996 ) .
these two references were chosen because their g dwarf samples are taken from the solar neighborhood and , although they are not identical to the metallicity distribution of the target stars that have been searched for planets using the doppler technique , they are good representatives of these stars .
let the observed metallicity distribution of the sun - like stars hosting giant planets be @xmath5 , where @xmath6 \equiv log(fe / h ) - log(fe / h)_{\odot } \approx log(z / z_{\odot})$ ] , and @xmath7 is the mass fraction of metals in the sun . in fig .
1 , the distribution of sun - like stars , @xmath8 , has been normalized so that the 32 host stars represent @xmath9 ( the average planet - finding efficiency given in marcy and butler 2000 ) of the total .
that is , each bin of @xmath8 has been rescaled such that @xmath10 .
target stars have not been selected for metallicity .
although doppler shifts can be measured with slightly more precision in metal - rich stars , and metal - rich stars are slightly brighter for a given spectral type ( leading to a malmquist bias ) , these two selection effects are estimated to be minor compared to the difference between the distributions in fig . 1 ( butler 2000 ) . for a given metallicity , an estimate of the relative probability that a star will host
a hot jupiter is the ratio of the number of stars hosting hot jupiters to the number of stars targeted , = .
this is plotted in fig . 1 and labelled `` probability of destroying earths '' .
at low metallicity , @xmath11 is low and remains low until solar metallicity , where it rises steeply .
this probability predicts that more than @xmath12 of sun - like stars with @xmath13 will have a doppler - detectable hot jupiter , @xmath14 of @xmath15 stars will have one and @xmath16 of solar metallicity stars ( @xmath17 ) will have one .
these predictions are also supported by independent observations : * a star with extremely high metallicity was included in the target list because of its high metallicity ( @xmath18 ) .
a planet , bd-10 3166 , was found around it ( butler 2000 ) .
this star was not included in fig .
1 because of selection bias , but this result does support the probability calculated here : @xmath19 .
* thirty - four thousand stars in the globular cluster tucanae 47 ( @xmath20 ) were monitored with hst for planets transiting the disks of the hosts .
fifteen or twenty such transits were predicted based on a @xmath16 planet - finding efficiency ( assumed to be independent of metallicity and stellar environment ) . none has been found ( gilliland 2000 ) .
this result is consistent with the probability calculated here : @xmath21 , but gilliland ( 2000 ) also suggest that the lack of planets could be due to high stellar densities disrupting planetary stability . the width of the ` probability of destroying earths ' region has been set by the errors on the terms in eq .
1 . the region is broad enough to contain @xmath11 s calculated when alternative estimates of @xmath8 are used singly or in combination ( e.g. sommer - larsen 1991 , rocha - pinto and maciel 1996 , favata 1997 ) and when a range of planet - finding efficiencies are assumed ( @xmath22 to @xmath23 ) .
thus , the curve is fairly robust to variations in both the estimates of the metallicity distribution of the target stars and to varying estimates of the efficiency of finding hot jupiters .
when larger numbers of hot jupiters are found and the metallicity distribution of the target stars is better known , the new @xmath11 should remain in ( or very close to ) the region labelled `` probability of destroying earths '' in fig
the probability of producing earths is zero at zero metallicity and increases as metals build up in the universe .
the qualitative validity of this idea is broadly agreed upon ( trimble 1997 , whittet 1997 ) but it is difficult to quantify . during star formation ,
varying degrees of fractionation transform a stellar metallicity disk into rocky and gaseous planets .
simulations of terrestrial planet formation by wetherill ( 1996 ) suggest that the mass of rocky planets within 3 au is approximately proportional to the surface density of solid bodies in the protoplanetary disk . in the low surface density regime , ( @xmath24 )
, where finding enough material to make an earth is a problem , the number of planets in the mass range @xmath25 increases roughly in proportion to the surface density , i.e. , to the metallicity .
this increase is not because the overall number of planets increases but because the masses ( of a constant number of planets ) increase , bringing them into the earth - like mass range .
these simulations may be the best evidence we currently have to support the idea that in the low metallicity regime , the probability of forming earths is linearly proportional to metallicity .
this also suggests that the earliest forming earths orbit minimal metallicity stars and are at the low mass end of whatever definition of ` earth - like ' is being used .
similar considerations apply to jupiter formation but at a slightly higher surface density threshold . weidenschilling ( 1998 )
finds that a @xmath26 disk surface density is not quite enough to initiate runaway jupiter formation but that a modest increase in surface density will .
these simulations support the standard core accretion models of planet formation and suggest that planet formation ( both rocky and gaseous ) is enhanced when more metals are available . in this analysis
i make the simple assumption that the ability to produce earths is zero at low metallicity and increases linearly with metallicity of the host star .
specifically , let @xmath27 be the relative ability to produce earths as a function of metallicity .
i assume that : * @xmath28 ( earth production is proportional to the abundance of metals ) * @xmath29 .
that is , at very low metal abundance , @xmath30 , the probability of producing earths is @xmath31 . to represent the uncertainty in this zero probability boundary condition , the range @xmath32 is shown in fig .
this assumption is discussed later . * the most metallic bin , @xmath33 ,
is assigned the probability of 1 : @xmath34 .
the probability of a stellar system harbouring earths , @xmath35 , is the probability of producing earths times the probability of not destroying them , = [ 1- ] .
this probability of harbouring earths is plotted in fig .
1 . starting at low metallicity , it rises linearly and then gets cut off sharply at @xmath36 .
it peaks at @xmath37 , has a mean of @xmath38 and a median of @xmath39 .
the @xmath40 confidence range is @xmath41 $ ] .
@xmath35 can be used to focus terrestrial planet search strategies .
for example , to maximize the chances of finding earths , nasa s terrestrial planet finder ( tpf ) should look at stars with metallicities within the 68% confidence range and in particular near the peak of @xmath35 . also , since there is a radial metallicity gradient in our galaxy , @xmath35 can be used to define a galactic metallicity - dependent habitable zone analogous to the water - dependent habitable zones around stars .
this can be done by replacing ` @xmath42 ' in eqs . 3 , 4 and 6 with galactic radius
` @xmath43 ' and replacing the @xmath44 with the density of sun - like stars @xmath45 .
the sun ( @xmath46 \equiv 0 $ ] ) is more metal - rich than @xmath47 of local sun - like stars and less metal - rich than @xmath47 of the stars hosting close - orbiting extrasolar planets .
the high value of @xmath48 ( compared to neighboring stars ) and the low value compared to hot jupiter hosts may be a natural consequence of the goldilocks metallicity selection effect discussed here ( see also gonzalez 1999 ) .
models need to simultaneously explain the presence of hot jupiters close to the host star , the high metallicity of the host stars ( specifically the steepness of the rise in @xmath49 seen in fig .
1 ) as well as the small eccentricities of the closest orbiting planets and the large eccentricities of the planets further out . planet / planet gravitational scattering may provide a natural way to explain these features ( weidenschillling and marzari 1996 , rasio and ford 1996 and lin and ida 1997 ) .
higher metallicity of the protoplanetary disk enhances the mass and/or number of giant planets , thereby enhancing the frequency of gravitational encounters between them .
simulations with up to nine planets have been done ( lin and ida 1997 ) and apparently , `` the more the merrier '' .
that is , the more planets there are , the more likely one is to get scattered into a sub - au orbit .
the least massive planets suffer the largest orbital changes .
thus the least massive are more likely to be ejected , but also are more likely to be gravitationally scattered into orbits with small periastrons which can become partially circularised either by tidal circularization or by the influence of a disk ( provided the disk has not dissipated before the scattering ) .
the observational determination of the star formation rate ( @xmath50 ) of the universe has been the focus of much current work which we summarize in fig .
2 a. various sources indicate a @xmath50 at high redshift an order of magnitude larger than the current @xmath50
. initial estimates in which a peak of star formation was found at redshift @xmath51 are being revised as new evidence indicates that there may be no peak in the star formation rate out to the maximum redshifts available ( @xmath52 ) .
let us restrict our attention to the set of sun - like stars ( spectral types f7 - k1 , in the mass range @xmath53 ) that have ever been born in the universe .
since @xmath16 of the mass that forms stars forms sun - like stars , the star formation rate as a function of time , @xmath44 can be multiplied by @xmath54 to yield the age distribution of sun - like stars in the universe . here
, the standard simplification is made that the stellar initial mass function is constant .
if low mass star formation is suppressed in low metallicity molecular clouds ( nishi and tashiro 2000 ) then the 1.5 gyr delay between @xmath50 and @xmath55 ( section 5 ) is even longer . if all sun - like stars formed planets irrespective of their metallicity , then the planet formation rate in the universe would equal @xmath56 , shown in fig .
2a . however , fig .
1 indicates that metallicity is a factor which should be taken into account .
thus , we estimate the earth - like planet formation rate , @xmath57 , orbiting sun - like stars as pfr(t ) = a sfr(t ) f(t ) , where @xmath58 is the fraction of stars being formed at time @xmath42 which are able to harbour earths .
if all sun - like stars formed planets and metallicity had no effect on terrestrial planet formation , we would have @xmath59 . if we knew
that on average one out of every thousand sun - like stars had an earth - like planet , ( and this number did not depend on the metallicity of the star ) , then the planet formation rate would be @xmath60 , which is just a rescaling of the @xmath50 in fig .
a plausible first approximation could have @xmath61 .
that is , the higher the average metallicity of the universe , the higher the efficiency with which star formation produces earths . in this analysis , however , this guess is improved on by taking into account the dispersion of metallicity of star forming regions around the mean at any given time , as well as by including the metallicity dependent selection effect for harbouring earths . when these are taken into consideration , @xmath58 becomes an integral over metallicity , f(t ) p ( , ( t ) ) d , where @xmath62 was derived above and @xmath63 is a gaussian parametrisation of the metallicity distribution of star - forming regions in the universe ( eq .
the universe started off with zero metallicity and a complete inability to form earths .
the metallicity of the universe gradually increased as a result of star formation and its by - products : various types of stellar novae and stellar winds .
various observations form a consistent picture of the gradual increasing metallicity of the universe ( fig .
the star formation rate plays a dual role in this analysis since stars make planets ( @xmath64 , eq .
3 ) and stars make metals , @xmath65 .
integration of this last proportionality yields the increasing mean metallicity , @xmath66 , of star forming regions in the universe _
0^t sfr(t)dt ~(t ) . the resulting @xmath66 ,
is plotted in fig .
2b ( thick line ) .
the grey area around @xmath66 reflects the spread in the estimates of the @xmath50 ( grey area in @xmath67 ) .
metallicity observations in our galaxy and at large redshifts are available to check the plausibility of this integral and are shown in fig .
2b . at any given time @xmath42 , some star forming regions have low metallicity while some have high metallicity .
we parametrize this spatial dispersion around the mean by a time - dependent gaussian centered on @xmath66 : p(,(t ) ) = exp [ ] .
the current metallicity distribution of ob stars in the thin disk ( gummersbach 1998 ) , which may be our best estimate of the current mean metallicity of star - forming regions in the universe , is used to normalize this function , @xmath68 , and provide the dispersion , @xmath69 .
performing the integral in eq .
4 and inserting the result into eq . 3 yields an estimate of the terrestrial planet formation rate in the universe which is also the age distribution of earths orbiting sun - like stars in the universe .
this distribution is plotted in fig . 2c and indicates that the average age of earths around sun - like stars is @xmath70 billion years .
the error bar represents the uncertainty in the @xmath50 ( shown in fig .
2a ) as well as the range of assumptions about the low metallicity tail of @xmath71 , discussed below .
thus , the average earth in the universe is @xmath0 billion years older than our earth . and ,
if life exists on some of these earths , it will have evolved , on average , @xmath72 billion years longer than we have on earth . among these earths ,
@xmath73 are older than our earth while @xmath74 are younger .
@xmath40 of earths in the universe are between @xmath75 and @xmath76 gyr old while @xmath12 are between @xmath77 and @xmath78 gyr old .
the time delay between the onset of star formation and the onset of terrestrial planet formation is the difference between the x - intercept of the thin and thick solid lines in fig .
2c . this delay is @xmath79 gyr and has an important dependence on the low metallicity tail of @xmath62 , specifically , on the metallicity for which @xmath80 has been assumed . to estimate the dependence of the main result on this assumption , both a high and a low metallicity case ( @xmath81 and @xmath82 )
have been considered .
that is , i have used the two boundary conditions @xmath83 and @xmath84 and the variation they produce in the result , to compute representative error bars .
the resulting variation is about one half of the variation due to the uncertainty in the @xmath50 .
if rocky planets can easily form around stars with extremely low metallicity due to high levels of fractionation during planet formation , then the lower limit used here , @xmath85 , may not be low enough .
these linear metallicity variations yield error estimates but other possibilities exist .
the masses of earth - like planets and the ability of a stellar system to produce them may not be linear functions of metallicity .
for example , there may be a strongly non - linear dependence on metallicity such as a metallicity threshold below which earths do not form and above which they always do .
if that were the case then the @xmath55 plotted in fig . 2c
would shift to the right or left depending on where the threshold is . in this analysis
i have assumed that the moons of hot jupiters do not accrete into earth - like planets .
this speculation has not been explored in any detail . if true , hot jupiters would destroy earths , but would also help create alternative sites for life . however , the delayed onset of planet formation compared to star formation derived here would be largely unchanged .
the cratering history of the moon tells us that the earth underwent an early intense bombardment by planetesimals and comets from its formation 4.56 gyr ago until @xmath86 gyr ago . for the first 0.5 gyr ,
the bombardment was so intense ( temperatures so high ) that the formation of early life may have been frustrated ( maher and stevensen 1988 ) . the earliest isotopic evidence for life dates from the end of this heavy bombardment @xmath87 billion years ago ( mojzsis 1996 ) .
thus , life on earth seems to have arisen as soon as temperatures permitted ( lazcano and miller 1994 ) . to interpret fig . 2c as the age distribution for life in the universe several assumptions need to be made . among them
are : 1 .
life is based on molecular chemistry and can not be based on just hydrogen and helium .
2 . the dominant harbours for life in the universe are on the surfaces of earths in classical habitable zones .
3 . other time - dependent selection effects which promote or hamper the formation of life ( supernovae rate ?
, gamma ray bursts ? , cluster environments ? ) are not as important as the metallicity selection effect discussed here ( norris 2000 ) .
life is long - lived .
if life goes extinct on planets then the @xmath55 needs to have its oldest tail chopped off to represent only existing life .
this paper is an attempt to piece together a consistent scenario from the most recent observations of extrasolar planets , the star formation rate of the universe and the metallicity evolution of the star - forming regions of the universe .
the precision of all of these data sets is improving rapidly . with more than 2000 stars
now being surveyed , we expect more than @xmath88 giant planets to be detected in the next few years .
the metallicities of target lists are also under investigation .
thus , the uncertainties in the metallicities of target stars and stars hosting planets will be reduced ( reducing the error bars in both the numerator and denominator of eq .
planet / planet interactions may explain the hot jupiter / high metallicity correlation but at least two other ( non - mutually exclusive ) explanations exist : 1 ) metallicity enhanced migration of giants in protoplanetary disks ( e.g. murray 1998 ) and 2 ) infall of metal - rich accretion disks onto the host stars , precipitated by the in - spiraling of large planets ( e.g. gonzalez 1998 , quellin and holman 2000 ) .
the infall of metallicity - enhanced material probably occurs in all migration or interaction scenarios .
however , if the outer convective zones of g dwarfs are thick enough to mix and dilute this material ( laughlin and adams 1999 ) then the analysis done here requires no significant modification for metallicity enhancement .
if the dilution is not effective then the observed metallicity of a star will not be a faithful indicator of the star s true metallicity and therefore will not be a good indicator of the probability to produce earths as assumed here .
the results obtained here for the metallicity and age distributions of earth - like planets in the universe are easily testable . over the next decade or two
, intensive efforts will be focused on finding earths in the solar neighborhood .
microsecond interferometry ( sim ) and even higher angular resolution infrared interferometry ( tpf & irsi - darwin ) as well as micro - lensing planet searches ( planet ) and high sensitivity transit photometry ( corot ) all have the potential to detect earth - like planets .
these efforts will eventually yield metalllicity and age distributions for the host stars of earth - like planets that can be compared to figs . 1 and 2c .
in addition , figs . 1 and 2c can be used to focus these efforts .
* acknowledgemnets * + i acknowledge g. gonzalez , f. rasio , j. lissauer , a. fernandez - soto , d. whittet , g. wetherill , p. butler , k. ragan and r. depropris for helpful comments and discussion . this work has been supported by an australian research council fellowship . | planets like the earth can not form unless elements heavier than helium are available .
these heavy elements , or ` metals ' , were not produced in the big bang .
they result from fusion inside stars and have been gradually building up over the lifetime of the universe .
recent observations indicate that the presence of giant extrasolar planets at small distances from their host stars , is strongly correlated with high metallicity of the host stars .
the presence of these close - orbiting giants is incompatible with the existence of earth - like planets .
thus , there may be a goldilocks selection effect : with too little metallicity , earths are unable to form for lack of material , with too much metallicity giant planets destroy earths . here
i quantify these effects and obtain the probability , as a function of metallicity , for a stellar system to harbour an earth - like planet .
i combine this probability with current estimates of the star formation rate and of the gradual build up of metals in the universe to obtain an estimate of the age distribution of earth - like planets in the universe .
the analysis done here indicates that three quarters of the earth - like planets in the universe are older than the earth and that their average age is @xmath0 billion years older than the earth .
if life forms readily on earth - like planets as suggested by the rapid appearance of life on earth this analysis gives us an age distribution for life on such planets and a rare clue about how we compare to other life which may inhabit the universe .
psfig = |
the formation of gas giant planets requires significant amounts of gas and dust to be present in the circumstellar environment of a young ttauri star .
the lifetime of protoplanetary disks is accordingly an important observable to constrain planet formation . to infer disk lifetimes ,
a number of previous studies have targeted young star - forming regions to measure the fraction of stars that exhibit either ongoing accretion or hot circumstellar dust or both .
these fractions appear to be a strong function of the age of a star - forming region , monotonically decreasing from @xmath480% to 0% within @xmath010myr ( e.g. , * ? ? ?
* ; * ? ? ?
the functional shape appears to be fit well by an exponential decay with a time constant of @xmath53myr @xcite .
the fact that a large fraction of all young stars is bound in multiple systems @xcite has a strong effect on the conclusions that can be drawn from this observation because a ) the presence of stellar binary companions has a strong effect on disk evolution , and b ) a large number of binary stars typically remains undetected and will `` contaminate '' single star measurements .
[ [ the - effect - of - stellar - binarity - on - disk - evolution . ] ] * the effect of stellar binarity on disk evolution . * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + sub - mm observations show that multiple stars exhibit a much reduced total dust mass compared to undisturbed systems @xcite .
this can be explained by the dynamical interaction of a disk with the stellar companion . for wide binaries
this leads to a truncation of the outer disk to @xmath01/51/2 times the binary separation @xcite .
this implies shorter disk lifetimes , given that mass accretion rates in binaries are comparable to those of single stars @xcite . and indeed , the data of ( * ? ? ?
* ; * ? ? ?
* see also @xcite ) imply an exponential decay of binary circumstellar disk fractions with a time constant of @xmath61.3myr , much shorter than that found by previous studies that do not focus on binary stars .
the fact that disks in binary stars follow a different evolutionary path may help to constrain planet formation scenarios : @xcite found that stars with close ( @xmath7au ) binary companions are less likely to be orbited by a low - mass ( @xmath21m@xmath8 ) gas planet than wider binaries and single stars .
one possible interpretation calls for a slow gas planet - forming process for planets @xmath31m@xmath8 too slow to complete in close binaries before the disk disperses .
more massive planets , in contrast , may form through a rapid process that completes equally often around single stars and around even closely separated binary components where disk lifetimes are shorter .
[ [ the - difficulty - of - identifying - single - stars . ] ] * the difficulty of identifying single stars . * +
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + while above considerations and the fact that sample sizes are typically small paint a noisy but consistent picture of disk evolution in ttauri _ binaries _ , the evolution of a disk in an undisturbed system like that of a single star can not be easily observed .
this is partly due to our lack of knowledge about the true multiplicity status of young stars : efficient high - angular resolution imaging searches for companions do not reach much below @xmath00.1@xmath9 ( @xmath010au at the distance of the nearest star - forming regions of @xmath0100pc ) , and few young stars have been subject to radial velocity ( rv ) monitoring over baselines @xmath110yr ( equivalent to @xmath45au ) .
this leaves the 510 au separation range mostly unstudied for stellar companions in the best - sampled nearby regions , and typically the range of unassessed separations is even larger . owing to this observational effect
, the fraction of undetected binary companions in any previous disk frequency survey can be @xmath030% or larger , depending on the distance of the targeted region and the availability of suitable survey data ( * ? ? ?
* ; * ? ? ?
* _ submitted _ ) .
the previously derived disk fractions in these samples accordingly represent a mixture of binary and single star disk data compromising their potential to provide physical information about disk decay . [
[ disk - evolution - around - single - stars . ] ] * disk evolution around single stars . *
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + applying a statistical correction for undetected binarity , @xcite study the early evolution of single stars and find observational evidence for a disk fraction evolution that may be different from an exponential decay . in qualitative agreement with disk evolution models by @xcite ,
disk frequencies stay close to 100% until @xmath023myr of stellar evolution and then rapidly decay before @xmath06myr of age .
such a change would have strong implications for planet formation .
for example , gas giant planets may form on much longer timescales or start to form at a later time
than typically assumed .
while providing interesting qualitative evidence , the results reported by @xcite are based on unresolved binaries with no information about whether circumstellar material is distributed around one or both stars .
furthermore , their study relies on the sum of a variety of disk indicators which probe different physical aspects of a disk .
the values presented by @xcite accordingly represent upper limits to the true accretion and inner dust disk frequency of single stars . to gain a better understanding of the evolution of undisturbed disk material in the first few myr of stellar evolution , a consistent and direct measurement of disk frequency around single stars is needed . as current instruments do not enable the identification of binary companions at all separations and mass ratios , the frequency of undetected binary companions and the disk fraction in multiples must be taken into careful consideration to infer single star values . in the following
we describe our previous and ongoing efforts .
we measured the frequency of disks around 52 spatially resolved multiple stars with separations between @xmath0251000au in the orion nebula cluster and chamaeleoni star - forming regions @xcite . for each stellar component ,
our measurements trace both the presence of ongoing accretion based on brackett-@xmath10 ( br@xmath10 ) emission as well as hot inner dust inferred from near - infrared color excess .
we compare this binary data set to single stars in chamaeleoni ( * ? ? ?
* _ submitted _ ) . in order to compile a sample with as little binary contamination as possible
, we queried the high - angular resolution imaging survey by @xcite for stars without stellar companions between @xmath050500au and brighter than @xmath11mag .
stars with known radial velocity companions were removed from the sample .
a total of 54 single star candidates were observed with sofi / ntt @xmath12-band spectroscopy to assess the presence of emission at the wavelength of br@xmath10 . using a monte carlo simulation
, we simulated the overall companion distribution of our sample and find that , even after excluding all adaptive optics ( ao ) and rv companions , there remains a @xmath030% chance of multiplicity for any star in the sample . using the results from our chamaeleoni binary study @xcite
, we correct the measured single star accretion frequency for the bias introduced by undetected binarity . as these measurements are based on a well - defined set of suitable disk indicators ( br@xmath10 , nir color excess ) and good coverage by ao and rv companion searches , the resulting disk evolution measurements represent the most robust assessment of the relative abundance of disks around the individual components of binaries and single stars to date .
figure [ fig1 ] shows our assessment of single and binary star disk fraction as a function of age .
close binaries ( @xmath13au ) appear to disperse their disks on much smaller timescales than wider binaries , both for our accretion and hot inner disk measurements .
_ right : _ hot circumstellar dust as inferred from @xmath14@xmath12 and @xmath12@xmath15 excess @xcite .
the blue continuous and dotted lines show exponential decay functions measured by ( * ? ? ?
* @xmath16myr , @xmath17myr ) , not corrected for undetected multiplicity .
the dashed cyan line shows the disk evolution model by @xcite.,scaledwidth=78.0% ] we furthermore robustly confirm that the inferred single star accretor fraction ( @xmath18% ; * ? ? ?
* _ submitted _ ) is about 6 times larger than that of close binaries .
the single star accretor fraction appears to be slightly larger but consistent with both wide binaries and the decay curve found by previous surveys without binary correction ( e.g. , * ? ? ? * ) . in particular , our new measurement of the single star accretor fraction is less than 1@xmath19 larger than previous estimates of the accretion frequency in chamaeleoni ( @xmath041% , based on * ? ? ?
the data disfavor a scenario where all stars retain their disks until 23myr @xcite .
it has recently been suggested that ages of star forming regions are much older than previously assumed @xcite .
if the ages of the targeted orion nebula cluster and chamaeleoni regions turn out to be a factor of @xmath02 larger than assumed here , then agreement between the data and the model is preserved .
single stars are hard or even impossible to identify in most star - forming regions .
this is because contamination with binary companions that are inaccessible to today s observational methods is on the order of @xmath430% for a typical star - forming region like chamaeleoni . as close binaries ( @xmath3100au )
have been shown to exhibit disk evolution timescales that are shorter than those of wide binaries and single stars , typical disk frequency studies are biased by undetected binary companions . rather than following an exponential decay , as suggested by studies without binary correction , the fraction of single stars with disks may stay constant and close to 100% until @xmath023myr , followed by a rapid ( @xmath23myr ) drop , according to disk evolution models that include viscosity , photoevaporation , and planet formation .
we conducted the first dedicated single star disk frequency measurement in chamaeleoni .
we find a significantly higher accretor fraction among single stars than around components of close binary stars ( @xmath3100au ) in the same region .
the single star disk frequency appears to be inconsistent with the very high disk frequency at 2myr suggested by previous models and statistical considerations .
part of the discrepancy between the observations and models might be due to systematic age uncertainties .
more measurements of single and binary star disk frequencies are needed to infer the evolution of undisturbed disks as a function of time and to constrain the qualitative and quantitative evolution of disk material in both truncated and undisturbed disks . | based on a large number of observations carried out in the last decade it appears that the fraction of stars with protoplanetary disks declines steadily between @xmath01myr and @xmath010myr .
we do , however , know that the multiplicity fraction of star - forming regions can be as high as @xmath150% and that multiples have reduced disk lifetimes on average . as a consequence
, the observed roughly exponential disk decay can be fully attributed neither to single nor binary stars and its functional form may need revision .
observational evidence for a non - exponential decay has been provided by @xcite , who statistically correct previous disk frequency measurements for the presence of binaries and find agreement with models that feature a constantly high disk fraction up to @xmath03myr , followed by a rapid ( @xmath22myr ) decline .
we present results from our high angular resolution observational program to study the fraction of protoplanetary disks of single and binary stars separately .
we find that disk evolution timescales of stars bound in close binaries ( @xmath3100au ) are significantly reduced compared to wider binaries .
the frequencies of accretors among single stars and wide binaries appear indistinguishable , and are found to be lower than predicted from planet forming disk models governed by viscous evolution and photoevaporation . |
the ap star @xmath1 equ ( hd 201601 , bs 8097 ) is one of the brightest objects of this class , with the apparent luminosity @xmath5 mag .
the exact spectral type of this object is a9p ( srcreu subclass ) .
the magnetic field of @xmath1 equ has been studied for more than 50 years , starting from october 1946 ( see babcock 1958 ) .
the longitudinal magnetic field @xmath0 of this star does not exhibit periodic variations in time scales typical of stellar rotation , @xmath6 days .
such a variability of the @xmath0 field was observed in most ap stars .
the above effect is commonly interpreted as the result of stellar rotation ( oblique dipole model ) .
the first measurements by babcock ( 1958 ) showed that the value of the longitudinal magnetic field @xmath0 of @xmath1 equ was positive in 194652 , and approached nine hundred g. from that time on the value of @xmath0 slowly decreased and even changed sign in 1970/71 .
one could interpret the magnetic behavior of @xmath1 equ either as secular variations , or variations caused by extremely slow rotation .
if the latter picture is correct , then the corresponding magnetic and rotational periods are in the range from 72 to 110 years ( bonsack & pilachowski 1974 ; leroy et al .
1994 ; bychkov & shtol 1997 ; scholz et al .
1997 ) .
the behavior of the @xmath0 field in @xmath1 equ was investigated by many authors in the second half of the twenty_th _ century . for this research
we compiled @xmath0 observations published by bonsack & pilachowski ( 1974 ) , scholz ( 1975 ; 1979 ) , borra & landstreet ( 1980 ) , zverko et al .
( 1989 ) , mathys ( 1991 ) , bychkov et al .
( 1991 ) , bychkov & shtol ( 1997 ) , scholz et al .
( 1997 ) , mathys & hubrig ( 1997 ) , hildebrandt et al .
( 2000 ) , leone & kurtz ( 2003 ) and hubrig et al .
( 2004 ) .
we included in this paper our unpublished magnetic @xmath0 measurements which were obtained during the past seven years .
all the new magnetic observations showed , that the slow decrease of the @xmath0 field in @xmath1 equ apparently reached the minimum in 19962002 and has actually started to increase .
in this paper we determined the accurate parameters of secular variability of @xmath1 equ : the period @xmath7 , the amplitude and the time of zero phase for @xmath0 variations , which were approximated by a sine wave .
we support the hypothesis that the long - term @xmath0 variation in @xmath1 equ is a periodic feature .
possible origin of this variation can not be uniquely determined , see discussion in section [ sec : discussion ] of this paper .
we have performed spectropolarimetric observations of zeeman line splitting for @xmath1 equ at the coude focus of the 1-m optical telescope ( special astrophysical observatory , russian academy of sciences ) .
zeeman spectra were obtained with the echelle spectrograph gecs ( musaev 1996 ) . we have put the achromatic analyser of circularly polarised light in front of the spectrometer slit .
images of the zeeman echelle spectra were recorded from ccd detectors in standard fits format .
final reduction of the archived spectra was performed with the standard midas software ( monin 1999 ) .
effects of instrumental polarisation on @xmath0 measurements obtained with this instrument were investigated by bychkov et al .
( 1998 , 2000 ) .
table [ tab : saores ] presents the full set of our @xmath0 measurements of @xmath1 equ ( total 33 @xmath0 points ) .
the meaning of the first 3 columns is obvious .
the fourth column gives the number @xmath8 of spectral lines which were used for the measurement of @xmath0 for a given exposure .
time length @xmath9 of the exposure ( in min ) is given in the last column of table [ tab : saores ] .
on average , the value of a single @xmath0 number listed in table [ tab : saores ] was obtained after averaging of @xmath0 measurements obtained in 500 - 1300 spectral lines .
standard deviation @xmath10 for the resulting value of @xmath0 was computed in the standard manner as the error of an arithmetic mean value .
errors @xmath10 determined in the above way reached rather low values in several observations listed in table [ tab : saores ] . in 2005/2006 we plan to verify the reality of such @xmath10 by a special program of @xmath0 observations .
actually we accept these errors _ bona fide _ and note the following properties of our @xmath0 measurements .
the referee pointed out that a few pairs of @xmath0 measurements of one night in table [ tab : saores ] differ by only a few g , which is substantially less than the corresponding standard deviation @xmath10 .
we can explain this only as a purely random effect , and do not see any reason for it either in the acquisition of observational data or their reduction .
secondly , series of measurements taken within a few nights generally show a scatter of the order of 100 g , which is much higher than the standard errors @xmath10 in table [ tab : saores ] .
the latter are of the order of @xmath11 g , and such a discrepancy suggests that our standard deviations are systematically underestimated , and are in fact of the order of 100 g. on the other hand , such a scatter of @xmath12 100 g is not inconsistent with the short - term variability of light and the longitudinal magnetic field @xmath0 in @xmath1 equ in time scales of minutes or above it .
leone & kurtz ( 2003 ) recently discovered periodic variations of the longitudinal magnetic field @xmath0 in @xmath1 equ over the pulsation period of this star , @xmath13 min . the estimated amplitude @xmath14 g for this period ,
therefore , these variations at least can contribute to the scatter of our @xmath0 points collected in table [ tab : saores ] .
study of the rapid periodic @xmath0 variations on a time scale of minutes was also presented in bychkov et al .
( 2005b ) for @xmath1 equ .
they did not found conclusive evidence of such variations above the noise level at @xmath15 g. we also performed spectral analysis of the full set of 298 @xmath0 time series from years 19462004 .
we concluded that there are no short - period field variations with periods above ca . 1 day , but were not able to extend our analysis for shorter periods , see section 4 of this paper .
.measurements of @xmath0 in @xmath1 equ ( hd 201601 ) . [ cols="<,>,^,>,^",options="header " , ]
magnetic observations presented in table [ tab : saores ] represent completely new data .
they cover time span of ca .
7 years and include the phase when the effective magnetic field @xmath0 in @xmath1 equ apparently reached its minimum value , and then the slow decrease of @xmath0 observed in the recent @xmath12 50 years has been reversed .
this fact is of extraordinary importance , because it allows one for a fairly accurate determination of the magnetic period and the amplitude of @xmath0 variations in @xmath1 equ .
we have compiled the set of 298 observations of the @xmath0 field in @xmath1 equ , scattered in the literature , and appended our measurements .
these data cover the time period 19462004 ( 58 years ) .
they are displayed in fig .
[ fig : long ] .
note , that the @xmath0 measurements obtained by babcock ( 1958 ) apparently cover the phase of the maximum longitudinal magnetic field in @xmath1 equ .
the set of @xmath0 measurements analysed in this paper is rather heterogeneous .
the data have been obtained by several different observers over a long time period using various instruments and techniques , and it is impossible to estimate or test credibly their systematic and random errors , particularly for the earliest observations of the longitudinal magnetic field in @xmath1 equ .
therefore , we arbitrarily assumed that systematic errors of the @xmath0 observations are equal to zero .
in other words , all the @xmath0 points for @xmath1 which were found in the literature are fully compatible .
random errors of individual @xmath0 points frequently were given in the source papers , and are denoted by vertical bars in fig .
[ fig : long ] .
these errors were not directly available for the earliest photographic measurements by h.w .
babcock ( 1958 ) and bonsack & pilachowski ( 1974 ) .
we adopted here the estimated error for babcock s data equal 238 g , and 151 g for bonsack & pilachowski .
these numbers were obtained in our thorough reanalysis of the earliest papers dealing with measurements of stellar magnetic fields , cf .
section 3.1 in bychkov et al .
( 2003 ) .
determination of the period and other parameters of the apparent magnetic variability for @xmath1 equ was performed in the following manner .
assuming that the run of the observed longitudinal field @xmath0 with time @xmath16 can be approximated by a sine wave @xmath17 \ , , \label{equ : sigma1}\ ] ] we determined all four parameters : the period @xmath18 , the average field @xmath19 , the amplitude @xmath20 and the time of zero phase @xmath21 using the iterative technique of nonlinear fitting .
starting values of @xmath18 , @xmath19 , @xmath20 , @xmath21 and their standard deviations were found by our computer code for the nonlinear least squares method ( bychkov et al .
the final values and their errors were then computed with the public domain code `` nlfit.f '' , which is designed for curve and surface fitting with the levenberg - marquardt procedure ( odrpack v. 2.01 subroutines ) .
the code is available at the site www.netlib.org .
fitting of a sine wave to all the 298 @xmath0 points with errors as in fig .
[ fig : long ] gave very poor results with the @xmath22 for a single degree of freedom @xmath23 .
such fits are unacceptable , and in case of @xmath1 equ the poor fit is the result of underestimated errors of many @xmath0 points .
many @xmath0 observations presented in fig .
[ fig : long ] have very low errors , which sometimes are less than 20 g. our new @xmath0 points , which are collected in table [ tab : saores ] , also are of such a high formal accuracy .
we can not judge , whether an apparent scatter of @xmath0 points in fig .
[ tab : saores ] is due to unrealistic error estimates or the intrinsic short - term variability of the longitudinal magnetic field in @xmath1 equ .
the estimated random error of @xmath0 points about the starting sine wave equals to 213 g. for the final fitting of a sine we assumed that all the 298 points have identical errors of 213 g. final values of the fitted parameters and their standard deviations @xmath24 for the sine phase curve are given below .
in other words , a parameter range from @xmath25 to @xmath26 is just the true 68% confidence interval for this parameter .
the above fit of a sine wave with uniform errors of 213 g is very good , with @xmath27 .
the effect of inhomogeneity in the @xmath0 time series plus the possible existence of rapid magnetic variability in @xmath1 equ were compensated by the increase of the random error , and neither should influence the above parameters of secular magnetic variability in @xmath1 equ .
the standard parameter @xmath28 was defined for the oblique rotator model of an ap star .
it is related to the angle @xmath29 between the magnetic dipole axis and the rotational axis , and the angle @xmath30 between the rotational axis and the line of sight ( preston 1967 ) : @xmath31 parameters @xmath32 and @xmath33 of the @xmath0 sine wave for @xmath1 equ are given by note , that the meaning of @xmath34 and @xmath35 for use in eq .
[ eqn : rrr ] is different : @xmath34 denotes there the value of magnetic intensity which has the higher absolute value , and @xmath36 has the lower absolute value . in this way we obtained the value of @xmath28 for @xmath1 equ equal to @xmath37 .
bychkov et al .
( 2005a ) presented an extensive catalog of the magnetic phase curves and their parameters for 136 stars on the main sequence and above it .
we quoted there the previously estimated period for @xmath1 equ , @xmath38 , which was obtained on the basis of a shorter series of @xmath0 data .
this paper and the new , more accurate @xmath39 represents a major revision of the previously known magnetic period of @xmath1 equ .
significant scatter of the observed points in the long - term run of @xmath40 in fig .
[ fig : long ] suggests the search for short - term periodicities .
we applied the strategy of prewhitening to the set of available @xmath0 measurements , and removed the principal sine - wave variations from the data .
prewhitened data were then analysed with the method developed by kurtz ( 1985 ) , and with his fortran code ( kurtz 2004 ) .
such a search for peaks in the @xmath0 amplitude spectrum of @xmath1 equ in this paper was restricted to trial periods higher than 1 day .
this is because many of the earlier magnetic observations for this star either have poorly determined the time of measurement , or have long times of exposure ( see e.g. babcock 1958 ) .
the star @xmath1 equ exhibits rapid nonradial pulsations and the corresponding @xmath0 with the period @xmath41 min ( leone & kurtz 2003 ) and , possibly , with simultaneous shorter periods ( bychkov et al .
none of them were analysed in this paper .
we have identified two additional periods of statistically low significance in the range @xmath42 , see fig .
[ fig : short ] : 3 mm @xmath43 days , amplitude @xmath44 g @xmath45 days , amplitude @xmath46 g 3 mm both peaks in the amplitude spectrum in fig . [ fig : short ] exhibit low signal to noise ratio , with noise level at ca .
80 g. the period @xmath47 is close to 1 year .
since most of the existing @xmath0 observations for @xmath1 equ were performed in months july - november , then the peak @xmath47 in the amplitude spectrum represents a false period which most likely reflects the average 1-year repetition time in the acquisition of the existing magnetic measurements .
we believe that the peak @xmath48 in the amplitude spectrum of the @xmath0 field of @xmath1 equ is the random effect of a pure noise .
the peak is very narrow , in fact , it only appears in a single bin of a very dense discrete frequency mesh .
kurtz ( 1983 ) discussed the possible existence of the period of @xmath49 days in his photometric observations of @xmath1 equ in 1981 .
that period was of low probability , but possibly could be identified with the real rotational period in this star .
we do not confirm the existence of the 38 day period in long - term @xmath0 observations of @xmath1 equ , see fig .
[ fig : short ] .
there exist three possible explanations for the observed long - term behavior of the longitudinal magnetic field in @xmath1 equ : 3 mm * 1
. * precession of the rotational axis ( lehmann 1987 ) .
* 2 . * solar - like magnetic cycle ( krause & scholz 1981 ) , * 3 . *
rotation with the period of 91.2 years .
3 mm the ap star @xmath50 hd 201601 in fact is a binary system .
one can assume , that the gravitational force from the secondary companion can cause precession of the ap star .
as the result , the angle between the rotational axis and the direction towards the earth varies periodically .
therefore , changes of the aspect can in principle cause apparent variations of the longitudinal magnetic field @xmath0 or the amplitude of its variations
. effects of precession in long - period ap stars were studied by lehmann ( 1987 ) , who showed that the oblateness of stars caused by the rotational or magnetic flattening is not adequate to produce observable precession effects .
the only exception was 52 her , where the observed behavior of the star could be interpreted as a precessional motion .
the above considerations indicate that the precession theory does not convincingly explain @xmath0 variations in this star .
the idea by krause & scholz ( 1981 ) that we actually observe the solar - like magnetic cycle in @xmath1 equ in which the global magnetic field reverses its polarity , can not be easily verified by the existing observations of the global longitudinal magnetic field @xmath0 .
moreover , one can note that such an idea requires the existence of a mechanism in the interior of @xmath1 equ which ensures the transfer of huge magnetic energy into electric currents and vice versa .
note that the required efficiency of such a mechanism and the amplitude of magnetic field variations in @xmath1 equ is ca .
four orders of magnitude larger than that in the sun in a similar timescale . following the widely accepted picture of an ap star ,
we believe that the magnetic field of @xmath1 equ can be approximated by a dipole located in the center of the star .
the dipole is inclined to the rotational axis of @xmath1 equ .
we assume that the magnetic field is stable and remains frozen in the interior of a rotating star at the time of observations , i.e. during at least of 58 years .
therefore , slow variations of the @xmath0 field in @xmath1 equ are caused by an extremely slow rotation , in which case our @xmath51 .
such an explanation is supported to some extent by polarimetric measurements by leroy et al .
( 1994 ) .
we plan to perform high accuracy polarimetric measurements of @xmath1 equ with the new version of minipol .
the device was constructed to measure the angles and the degree of linear polarisation of stellar radiation , and will be operational at the special astrophysical observatory in 2006 .
we also expect that shall be able to verify the extremely slow rotation of @xmath1 equ measuring the rate of change for the polarisation angle of stellar radiation .
the ap star @xmath1 equ ( hd 201601 ) exhibited slow and systematic decrease of the longitudinal magnetic field @xmath0 starting from 1946 , when the global magnetic field of this star was discovered ( babcock 1958 ) .
we have compiled the full set of 298 existing @xmath0 measurements , which consists of the @xmath0 data published in the literature and our observations obtained during recent 7 years .
the latter magnetic data ( 33 @xmath0 points ) were measured with the echelle spectrograph in the coude focus of the 1-m telescope at the special astrophysical observatory .
our newest observations showed that the longitudinal magnetic field @xmath0 of @xmath1 equ reached its local minimum and started to rise in 1998 - 2004 .
all the available data cover the time period of 58 years ( 1946 - 2004 ) and include both phases of the maximum and minimum @xmath0 . assuming that the secular variability of the @xmath0 field is a periodic feature , we determined parameters of the magnetic field curve in @xmath1 equ and give the value of its period , @xmath52 years , with the zero phase ( maximum of @xmath0 ) at @xmath53 jd @xmath54 .
sine - wave fit to the @xmath0 phase curve yields @xmath55 g and @xmath56 g. spectral analysis of the 58-year long @xmath0 time series essentially do not show the existence of shorter periods , down to trial periods of @xmath12 1 day .
more specifically , there are no real shorter periods in the run of the longitudinal magnetic field @xmath0 with amplitudes exceeding the noise level of 80 g.
we are grateful to john d. landstreet , the referee , for his criticism and suggestions regarding our computations and the manuscript .
we thank don kurtz for providing his fortran software used here to compute the amplitude spectrum of @xmath1 equ .
this research was supported by the polish committee for scientific research grant no
. 1 p03d 001 26 . | we present an analysis of the secular variability of the longitudinal magnetic field @xmath0 in the roap star @xmath1 equ ( hd 201601 ) .
measurements of the stellar magnetic field @xmath0 were mostly compiled from the literature , and append also our 33 new @xmath0 measurements which were obtained with the 1-m optical telescope of special astrophysical observatory ( russia ) .
all the available data cover the time period of 58 years , and include both phases of the maximum and minimum @xmath0 .
we determined that the period of the long - term magnetic @xmath0 variations equals @xmath2 years , with @xmath3 g and @xmath4 g. [ firstpage ] stars : magnetic fields stars : chemically peculiar
stars : individual : hd 201601 |
the process @xmath1 is already one of the key processes at lep2 , and will be of similar importance at future linear @xmath4 colliders . hence it is not surprising that considerable theoretical effort has gone into the precise prediction of the cross - section in the standard model ( sm ) , both for on- and off - shell w bosons ( @xcite , see @xcite for a review ) . for a process well accessible both experimentally and theoretically in the sm ,
one of the obvious questions to ask is whether it can tell us anything about physics beyond the sm .
supersymmetric extensions play a special role because they , like the sm , allow to make precise predictions in terms of a set of input parameters .
previous calculations in supersymmetric theories include the complete one - loop corrections in spontaneously broken supersymmetry @xcite , sfermion - loop effects in the mssm @xcite , and also the complete mssm corrections to the closely related triple - gauge - boson vertex @xcite . in this paper
the complete one - loop corrections for @xmath1 in the mssm including real bremsstrahlung in the soft - photon approximation are presented .
the results are rather small : the maximum deviation from the standard model within the scanned region of parameter space is @xmath5 for unpolarized and transversally polarized w bosons , and @xmath6 for longitudinal w bosons .
the outline of this paper is as follows . in sect .
[ sect : kin¬ ] the kinematics and notation are fixed .
[ sect:1loop ] describes the details of the one - loop calculation .
the results of the calculation and the scan over the mssm parameter space are presented in sect.[sect : results ] .
[ sect : conclus ] finally gives the conclusions .
the reaction studied here is @xmath7 where @xmath8 and @xmath9 represent the momenta and helicities of the external particles , respectively .
the incoming particles travel along the @xmath10 axis and are scattered into the @xmath11@xmath10 plane . neglecting the electron mass , the explicit representations of the momenta and the polarization vectors in the centre - of - mass system are @xmath12 where @xmath13 is the velocity of the electrons , @xmath14 is the momentum of the w bosons , @xmath15 is the beam energy , and @xmath16 is the scattering angle . the polarized differential cross - section is obtained from the helicity amplitudes @xmath17 as @xmath18 in the following , only unpolarized electrons are considered , since that is the situation at lep2 . in this case
the cross - section has to be averaged over the initial helicities according to @xmath19 polarization combinations are given by a sequence of four letters , uult , where l , t , and u denote longitudinal ( @xmath20 ) , transverse ( @xmath21 ) , and unpolarized particles .
since the coupling of the electron to any of the higgs particles in the sm or the mssm is suppressed by a factor @xmath22 , the higgs - exchange diagrams can safely be neglected .
the tree - level diagrams are then the same for the sm and the mssm , @xmath23 and @xmath24 exchange in the @xmath25-channel and neutrino exchange in the @xmath26-channel ( see fig .
[ fig : born ] ) .
the tree - level results shall not be discussed here as this has been done in detail elsewhere ( explicit formulas can be found in @xcite ) . the tree - level diagrams . ]
the calculation of the @xmath0 radiative corrections has been performed in t hooft feynman gauge .
ultraviolet ( uv ) divergences were treated within dimensional regularization . for the renormalization
the on - shell scheme @xcite was used , following the formulation worked out in @xcite . in the absence of susy particles at tree level ,
only those counter - terms appear that are already present in the sm .
the only change is that now the self - energies from which the renormalization constants are derived have to be calculated in the mssm . to @xmath0 ,
the squared matrix element is given by @xmath27 where @xmath28 and @xmath29 denote the sum of the contributing tree - level and one - loop feynman diagrams , respectively , and @xmath30 is the qed correction factor from real bremsstrahlung in the soft - photon approximation .
the feynman diagrams were generated with @xcite which in its current version uses algorithms that can deal with supersymmetric theories @xcite .
the resulting amplitudes were algebraically simplified using @xcite and then converted to a fortran program .
the package @xcite was used to evaluate the one - loop scalar and tensor integrals . for a single point in phase and parameter space
the fortran program runs about 3.6 ms in the sm and 102 ms in the mssm .
scans over parameter space in the mssm typically take several hours . apart from obvious checks such as uv- and ir - finiteness ,
the sm results of @xcite were fully confirmed and the effects of sfermion - loops in @xcite could be reproduced for various scenarios of sfermion masses to within a very small constant shift ( @xmath31 ) which is an effect of the different renormalization schemes used and hence of higher order . with the presence of external gauge bosons
one might expect sizable gauge cancellations and hence instabilities in the numerical code .
to safeguard against such problems , the sm values were computed once in the conventional way and once in the background - field method @xcite , where gauge cancellations should be greatly reduced , and an agreement to 9 digits was found . even though the same comparison can not easily be done for the mssm
( a corresponding model file is currently not available ) , a similar stability is expected since the gauge structure of the mssm is essentially the same as in the sm . as an alternative measure of stability , one can compare the individual contributions from self - energies , vertices , and boxes with their sum . to show that the cancellations are harmless numerically , the following table lists the individual contributions ( each of them renormalized ) that make up the amplitude at a particular point in parameter space for longitudinal w bosons , for which the largest cancellations can be expected .
@xmath32 in addition to the virtual diagrams , real photon emission from the external legs has to be taken into account to cancel the infrared ( ir ) divergences which arise from the exchange of massless photons .
the ir divergences are regularized by an infinitesimal photon mass @xmath33 . in the soft - photon limit
the cross - section for real photon emission is proportional to the born cross - section , @xmath34 with the soft - photon factor @xmath30 given by @xmath35 here @xmath36 is the maximum energy of the emitted photons , @xmath37 is the charge of the @xmath38th external particle , and the sign in front of the product @xmath39 is @xmath40 if particles @xmath38 and @xmath41 are both either incoming or outgoing , and @xmath42 otherwise .
the basic integrals needed for the evaluation of have been worked out in @xcite .
the absolute magnitude of the @xmath0 corrections is largely determined by the qed contributions and depends strongly on the photon - energy cutoff @xmath36 through logarithms of the form @xmath43 . without an experimentally motivated choice of @xmath44 , the overall size of the one - loop corrections can be shifted more or less at will using different values of @xmath36 . since the intention here is only to show the differences between the sm and the mssm calculations , the concrete value of @xmath36 is rather unimportant . in this calculation
a soft - photon cutoff energy of @xmath45 has been used as in @xcite .
the cancellation of the ir divergences has been checked numerically by varying the photon - mass @xmath33 on which the result must not depend .
a variation of @xmath33 from 1 to @xmath46 gev leaves the first 10 digits of the cross - section invariant . while soft - photon radiation is sufficient to cancel ir divergences , it is an approximation , valid only if @xmath36 is small compared to all relevant energy scales .
if this is not the case , hard - photon radiation must be taken into account , too .
however , the hard - photon corrections at @xmath0 are exactly the same as in the sm , therefore they have been omitted in the present calculation .
this section lists the contributing one - loop diagrams .
the diagrams can be separated into self - energy contributions , vertex corrections , and box corrections .
note that diagrams which are already present in the sm are omitted in the figures .
the self - energy contributions fall into two categories , corrections to the @xmath23 and @xmath24 propagator in the @xmath25-channel , and corrections to the @xmath47 propagator in the @xmath26-channel .
these diagrams are shown in fig.[fig : selfdiags ] . the mssm self - energy corrections .
braces indicate that there is one diagram for the first members of all braced lists , one for the second members , etc .
a sfermion with index @xmath38 accounts for six particles , @xmath48 . ]
the vertex diagrams can be grouped into corrections to the initial- and final - state vertex in the @xmath25-channel ( figs .
[ fig : vertsidiags ] and [ fig : vertsfdiags ] ) , and corrections to the @xmath26-channel vertices ( fig.[fig : verttdiags ] ) .
the mssm contributions to the initial - state vertex in the @xmath25-channel . ] the mssm contributions to the @xmath25-channel final - state vertex .
braces indicate that there is one diagram for the first members of all braced lists , one for the second members , etc . a sfermion with index @xmath38 accounts for six particles , @xmath49 .
] finally , the box diagrams are displayed in fig .
[ fig : boxdiags ] . the mssm contributions to the @xmath26-channel vertices . ] the mssm box diagrams . ]
for the sm parameters the following numerical values are used : @xmath50 m_e & = 0.51099907\mev,\quad & m_u & = 53.8\mev,\quad & m_d & = 53.8\mev,\quad \\ m_\mu & = 105.658389\mev , & m_c & = 1.50\gev , & m_s & = 150\mev , \\
m_\tau & = 1777\mev , & m_t & = 174\gev , & m_b & = 4.7\gev .
\end{aligned}\ ] ] the masses of the up and down quarks are effective parameters which are adjusted such that the five - flavour hadronic contribution to @xmath51 is 0.02778 @xcite , @xmath52 [ [ higgs - sector ] ] higgs sector + + + + + + + + + + + + the neutral higgs sector is fixed by choosing a value for @xmath53 and for the mass @xmath54 of the cp - odd neutral higgs boson @xmath55 .
for the other neutral - higgs masses , which receive significant radiative corrections @xcite , the two - loop approximation formula of @xcite is used .
there are only small radiative corrections for the charged higgs masses and the following equation holds for @xmath56 @xcite .
@xmath57 \ln\frac{\msusy}{m_{t , r}}\,,\end{gathered}\ ] ] where @xmath58 is a universal soft - susy - breaking mass introduced in the next paragraph and @xmath59 is the running top mass .
[ [ sfermions ] ] sfermions + + + + + + + + + for simplicity , all soft - susy - breaking parameters are assumed equal and mixing between sfermion generations is neglected , so that @xmath60 then , the sfermion mass matrix is given by @xcite @xmath61 where the elements in braces apply to @xmath62 and @xmath63 , respectively .
[ [ charginos - and - neutralinos ] ] charginos and neutralinos + + + + + + + + + + + + + + + + + + + + + + + + + the chargino mass matrix @xcite @xmath64 and the neutralino mass matrix @xcite @xmath65 are diagonalized with unitary matrices @xmath66 , @xmath67 , and @xmath68 such that @xmath69 this diagonalization is done numerically using the subroutines of the lapack library @xcite . the @xmath70 gaugino - mass parameter @xmath71 which appears as a further input parameter in @xmath72
is fixed , as usual , by the susy - gut relation @xmath73 the remaining input parameters are then scanned over the following regions : @xmath74 when the particle masses are calculated from these input parameters , the fortran program checks whether they are consistent with the current exclusion limits and automatically omits already - excluded points in parameter space from the scan .
the following bounds are used : @xmath75 in fig .
[ fig : absscan ] the absolute values of the total and differential cross - section in the sm and mssm is shown for three different polarizations : uuuu ( all particles unpolarized ) , uutt ( transverse w bosons ) , and uull ( longitudinal w bosons ) .
the thickness of the curves reflects the range of the allowed values within the parameter scan . the total cross - section and the differential cross - section at @xmath76 gev for the sm and the mssm .
the thickness of the curves give the range of allowed values within the parameter scan .
the polarizations are denoted by : uuuu ( all particles unpolarized ) , uutt ( transverse w bosons ) , and uull ( longitudinal w bosons ) . ]
[ fig : relscan ] shows the relative corrections in the sm and mssm with respect to the born cross - section .
the overall magnitude of the corrections is largely determined by the well - understood qed logarithms , as discussed in sect .
[ sect : qed ] .
the bands show the variation of the cross - section within the parameter scan .
the bands for the sm and mssm overlap slightly , the mssm band being the lower one .
the two bands are of comparable width of the order of a few percent for all polarizations .
the dominating polarizations over the entire energy range are uu@xmath77@xmath78 , and this explains why the plots for the uuuu and uutt polarizations in fig . [ fig : relscan ] are quite similar . on the other hand it is not possible to trace the origins of the radiative corrections by disentangling the self - energy , vertex , and box contributions since in the presence of external gauge bosons there are delicate gauge cancellations . the relative difference between the one - loop corrected and the born cross - section .
the cross - hatched bands indicate the minimum and maximum reached within the scan over the parameter space .
the polarizations are denoted by : uuuu ( all particles unpolarized ) , uutt ( transverse w bosons ) , and uull ( longitudinal w bosons ) . ] the relative deviation between the sm central value and the mssm band is plotted in fig .
[ fig : smmssm ] .
while the largest corrections ( @xmath79 at 1 tev ) are seen for purely longitudinal w bosons , one has to keep in mind that the cross - section for longitudinally polarized w bosons is much smaller than for the transverse polarizations . in the transverse polarizations and in the unpolarized case
the maximum deviation between the sm and the mssm is roughly 1.5% .
the maximum deviation is reached in all polarizations for light susy particles , for example in the unpolarized case ( uuuu ) the lightest higgs , stop , chargino , and neutralino masses at the point of maximum deviation are @xmath80 , @xmath81 , @xmath82 , and @xmath83 , respectively . the relative difference between the sm central value @xmath84 and the mssm bands of fig .
[ fig : relscan ] .
the polarizations are denoted by : uuuu ( all particles unpolarized ) , uutt ( transverse w bosons ) , and uull ( longitudinal w bosons ) . ]
the variation of the cross - section with the scanned mssm parameters is shown in fig .
[ fig : variation ] .
the largest variation is connected with the soft - susy - breaking mass @xmath58 and confirms the idea of the previous calculations @xcite that important contributions come from the sfermion sector .
nevertheless , the contributions from the other sectors are similar in size and can not be neglected .
the variation of the mssm cross - section for unpolarized particles with the scanned mssm parameters .
the variation of a parameter @xmath85 is computed as the difference between maximum and minimum of the cross - section for fixed @xmath85 , averaged over all values of @xmath85 . ]
the calculation was checked for a possible scheme dependence by repeating it using the input values @xmath86 instead of @xmath87 .
this is not entirely straightforward since @xmath88 enters not only in loop corrections , but more directly in the kinematics of the process .
for instance , the threshold of @xmath89 is different in the sm and mssm so that it is not admissible to plot the ratio as in fig . [
fig : smmssm ] . to compute @xmath90 , the fit formula of @xcite , approximating the two - loop result , together with the mssm contributions to @xmath91 @xcite have been used .
the results are shown for longitudinal w bosons , for which the mssm corrections are largest , in fig .
[ fig : relscangf ] and differ only slightly from the former result .
the same as the lower diagram in fig .
[ fig : relscan ] ( the relative one - loop corrections for longitudinal w bosons ) , but calculated from the input parameter set @xmath92 instead of @xmath93 .
the cross - hatched bands indicate the minimum and maximum reached within the scan over the parameter space . ]
the mssm corrections to @xmath1 are probably too small to be detected at lep2 , taking into account that the convolution with decay amplitudes in the final and photon radiation in the initial state tends to smear the corrections . with the projected accuracy of a linear collider , however
, corrections of this magnitude will play a role . yet
even a linear collider may have a hard time disentangling the mssm contributions because the one - loop electroweak corrections include sudakov logarithms of the form @xmath94 , whereas for susy corrections these logarithms are power - suppressed @xcite . extrapolating from 20% one - loop electroweak corrections at tev energies
, the higher - order electroweak contributions may be comparable in magnitude to the one - loop mssm corrections .
although the calculation presented here is limited to on - shell w bosons and hard - bremsstrahlung effects have not been considered , these effects are the same for the sm and the mssm , hence it should be possible to use the mssm cross - section presented here in a straightforward way in existing monte carlo generators , for example @xcite .
the fortran code for this process is available at http://www.hep-processes.de .
i thank w. hollik for discussions , c. schappacher for help with the implementation of the mssm parameter scan in fortran , and g. weiglein for proofreading the manuscript .
parts of this calculation have been performed on the qcm computer cluster at the university of karlsruhe , supported by the deutsche forschungsgemeinschaft ( forschergruppe `` quantenfeldtheorie , computeralgebra und monte - carlo simulation '' ) .
m. lemoine and m. veltman , _ nucl .
_ * b164 * ( 1980 ) 445 ; + m. bhm , a. denner , t. sack , w. beenakker , f. berends , and h. kuijf , _ nucl .
_ * b304 * ( 1988 ) 463 ; + j. fleischer , f. jegerlehner , and m. zralek , _ z. phys .
_ * c42 * ( 1989 ) 409 .
ross and j.c .
taylor , _ nucl . phys . _
* b51 * ( 1973 ) 125 , e : _ nucl .
* b58 * ( 1973 ) 643 ; + a. sirlin , _ phys .
_ * d22 * ( 1980 ) 971 ; + j. fleischer and f. jegerlehner , _ phys .
rev . _ * d23 * ( 1981 ) 2001 ; + k. i. aoki et al .
. prog .
* 73 * ( 1982 ) 1 ; + m. bhm , w. hollik , and h. spiesberger , _ fortschr . phys . _
* 34 * ( 1986 ) 687 . | the complete @xmath0 corrections including soft - photon bremsstrahlung to the process @xmath1 in the mssm are calculated for on - shell w bosons .
the relative difference between the mssm and standard model corrections is generally quite small .
the maximum deviation from the standard model within the scanned region of parameter space is @xmath2 for unpolarized and transversally polarized w bosons , and @xmath3 for longitudinal w bosons .
+ pacs numbers : 12.60.jv , 13.10.+q , 12.15.lk . |
the original method of maximum entropy , maxent @xcite , was designed to assign probabilities on the basis of information in the form of constraints .
it gradually evolved into a more general method , the method of maximum relative entropy ( abbreviated me ) @xcite - caticha07 , which allows one to update probabilities from arbitrary priors unlike the original maxent which is restricted to updates from a uniform background measure .
the realization @xcite that me includes not just maxent but also bayes rule as special cases is highly significant .
first , it implies that me is _ capable of reproducing every aspect of orthodox bayesian inference _ and proves the complete compatibility of bayesian and entropy methods .
second , it opens the door to tackling problems that could not be addressed by either the maxent or orthodox bayesian methods individually .
the main goal of this paper is to explore this latter possibility : the problem of processing data plus additional information in the form of expected values . when using bayes rule it is quite common to impose constraints on the prior distribution . in some cases these constraints
are also satisfied by the posterior distribution , but these are special cases . in general , constraints imposed on priors do not propagate to the posteriors .
although bayes rule can handle _ some _ constraints , we seek a procedure capable of enforcing _ any _ constraint on the posterior distributions .
after a brief review of how me processes data and reproduces bayes rule , we derive our main result , the general canonical form of the posterior distribution for the problem of simultaneous updating with data and moment constraints .
the final result is deceivingly simple : bayes rule is modified by a canonical exponential factor .
although this result is very simple , it should be handled with caution : once we consider several sources of information such as multiple constraints we must confront the problem of non - commuting constraints .
we discuss the question of whether they should be processed simultaneously , or sequentially , and in what order .
our general conclusion is that these different alternatives correspond to different states of information and accordingly we expect that they will lead to different inferences . as an illustration , the multinomial example of die tosses
is solved in some detail for two problems .
they appear superficially similar but are in fact very different .
the first die problem requires that the constraints be processed sequentially .
this corresponds to the familiar situation of using maxent to derive a prior and then using bayes to process data .
the second die problem , which requires that the constraints be processed simultaneously , provides a clear example that lies beyond the reach of bayes rule .
our first concern when using the me method to update from a prior to a posterior distribution is to define the space in which the search for the posterior will be conducted .
we wish to infer something about the value of a quantity @xmath0 on the basis of three pieces of information : prior information about @xmath1 ( the prior ) , the known relationship between @xmath2 _ and _ @xmath1 ( the model ) , and the observed values of the data @xmath3 . and @xmath2 to represent one or many unknown variables , @xmath4 , and one or multiple experiments , @xmath5 .
] since we are concerned with both @xmath2 _ and _ @xmath1 , the relevant space is neither @xmath6 nor @xmath7 but the product @xmath8 and our attention must be focused on the joint distribution @xmath9 .
the selected joint posterior @xmath10 is that which maximizes the entropy,@xmath11=-\tint dxd\theta ~p(x,\theta ) \log \frac{p(x,\theta ) } { % p_{\text{old}}(x,\theta ) } ~,~ \label{entropy}\]]subject to the appropriate constraints .
all prior information is codified into the _ joint prior _ @xmath12 .
both @xmath13 ( the familiar bayesian prior distribution ) and @xmath14 ( the likelihood ) contain prior information .
the new information is the observed data @xmath15 , which in the me framework must be expressed in the form of a constraint on the allowed posteriors .
the family of posteriors @xmath9 that reflects the fact that @xmath2 is now known to be @xmath16 is such that @xmath17this amounts to an _ infinite _ number of constraints on @xmath9 : for each value of @xmath2 there is one constraint and one lagrange multiplier @xmath18 . maximizing @xmath19 , ( [ entropy ] ) , subject to the constraints ( [ data constraint ] ) plus normalization , @xmath20 + \tint dx\,\lambda ( x)\left [ \tint d\theta ~p(x,\theta ) -\delta ( x - x^{\prime } ) % \right ] \right\ } = 0~,\]]yields the joint posterior , @xmath21where @xmath22 is a normalization constant , and @xmath23 is determined from ( [ data constraint ] ) , @xmath24the final expression for the joint posterior is@xmath25and the marginal posterior distribution for @xmath1 is@xmath26which is the familiar bayes conditionalization rule . to summarize : @xmath27 is updated to @xmath28 with @xmath29 fixed by the observed data while @xmath30 remains unchanged .
we see that in accordance with the minimal updating philosophy that drives the me method _ one only updates those aspects of one s beliefs for which corrective new evidence ( in this case , the data ) has been supplied_.
here we generalize the previous section to include additional information about @xmath1 in the form of a constraint on the expected value of some function @xmath31 , @xmath32we emphasize that constraints imposed at the level of the prior need not be satisfied by the posterior . what we do here differs from the standard bayesian practice in that we _ require _ the constraint to be satisfied by the posterior distribution . maximizing the entropy ( [ entropy ] ) subject to normalization , the data constraint ( [ data constraint ] ) , and the moment constraint ( [ < f > ] ) yields the joint posterior,@xmath33where @xmath22 is a normalization constant,@xmath34the lagrange multipliers @xmath23 are determined from the data constraint , ( [ data constraint]),@xmath35so that the joint posterior becomes@xmath36 is determined by imposing that the posterior @xmath10 satisfy ( [ < f > ] ) .
this yields an implicit equation for @xmath37 , @xmath38note that since @xmath39 the resultant @xmath37 will depend on the observed data @xmath16 . finally , the new marginal distribution for @xmath1 is@xmath40for @xmath41 ( no moment constraint ) we recover bayes rule . for @xmath42 bayes rule
is modified by a canonical exponential factor .
the me method allows one to process information in the form of constraints .
when we are confronted with several constraints we must be particularly cautious . in what order
should they be processed ? or should they be processed at the same time ?
the answer depends on the nature of the constraints and the question being asked .
we refer to constraints as _ commuting _ when it makes no difference whether they are handled simultaneously or sequentially .
the most common example is that of bayesian updating on the basis of data collected in multiple experiments : for the purpose of inferring @xmath1 it is well - known that the order in which the observed data @xmath43 is processed does not matter .
the proof that me is completely compatible with bayes rule implies that data constraints implemented through @xmath44 functions , as in ( [ data constraint ] ) , commute .
it is useful to see how this comes about .
when an experiment is repeated it is common to refer to the value of @xmath2 in the first experiment and the value of @xmath2 in the second experiment .
this is a dangerous practice because it obscures the fact that we are actually talking about _ two _ separate variables .
we do not deal with a single @xmath2 but with a composite @xmath45 and the relevant space is @xmath46 .
after the first experiment yields the value @xmath47 , represented by the constraint @xmath48 , we can perform a second experiment that yields @xmath49 and is represented by a second constraint @xmath50 .
these constraints @xmath51 and @xmath52 commute because they refer to _ different _ variables @xmath53 and @xmath54 .
an experiment , once performed and its outcome observed , can not be _ un - performed _ and its result can not be _ un - observed _ by a second experiment .
thus , imposing one constraint does not imply a revision of the other . in general constraints need not commute and when this is the case the order in which they are processed is critical .
for example , suppose the prior is @xmath55 and we receive information in the form of a constraint , @xmath56 .
to update we maximize the entropy @xmath57 $ ] subject to @xmath56 leading to the posterior @xmath58 as shown in figure 1 .
next we receive a second piece of information described by the constraint @xmath59 . at this point
we can proceed in essentially two different ways : * ( a ) sequential updating . *
having processed @xmath60 , we use @xmath58 as the current prior and maximize @xmath61 $ ] subject to the new constraint @xmath59 .
this leads us to the posterior @xmath62 . *
( b ) simultaneous updating .
* use the original prior @xmath63 and maximize @xmath57 $ ] subject to both constraints @xmath56 and @xmath59 simultaneously .
this leads to the posterior @xmath64 .
as the current prior and maximize @xmath65 $ ] subject to both constraints @xmath60 and @xmath59 simultaneously .
fortunately , and this is a valuable check for the consistency of the me method , it is easy to show that case ( c ) is equivalent to case ( b ) .
whether we update from @xmath66 or from @xmath58 the selected posterior is @xmath67 . ] to decide which path ( a ) or ( b ) is appropriate , we must be clear about how the me method treats constraints . the me machinery interprets a constraint such as @xmath60 in a very mechanical way : all distributions satisfying @xmath68 are in principle allowed and all distributions violating @xmath60 are ruled out .
updating to a posterior @xmath58 consists precisely in revising those aspects of the prior @xmath66 that disagree with the new constraint @xmath60 .
however , there is nothing final about the distribution @xmath58 .
it is just the best we can do in our current state of knowledge and we fully expect that future information may require us to revise it further . indeed , when new information @xmath59 is received we must reconsider whether the original @xmath56 remains valid or not .
are _ all _ distributions satisfying the new @xmath59 really allowed , even those that violate @xmath60 ? if this is the case then the new @xmath59 takes over and we update from @xmath58 to @xmath69 . the constraint
@xmath60 may still retain some lingering effect on the posterior @xmath62 through @xmath70 but in general @xmath60 has now become obsolete .
alternatively , we may decide that the old constraint @xmath60 retains its validity .
the new @xmath59 is not meant to revise @xmath60 but to provide an additional refinement of the family of allowed posteriors . in this case
the constraint that correctly reflects the new information is not @xmath59 but the more restrictive @xmath71 .
the two constraints should be processed simultaneously to arrive at the correct posterior @xmath64 . to summarize : sequential updating is appropriate when old constraints become obsolete and are superseded by new information ; simultaneous updating is appropriate when old constraints remain valid .
the two cases refer to different states of information and therefore _ we expect _ that they will result in different inferences .
these comments are meant to underscore the importance of understanding what information is being processed ; failure to do so will lead to errors that do not reflect a shortcoming of the me method but rather a misapplication of it .
this is a loaded die example illustrating the appropriateness of sequential updating .
the background information is the following : a certain factory makes loaded dice . unfortunately because of poor quality control , the dice are not identical and it is not known how each die is loaded .
it is known , however , that the dice produced by this factory are such that face @xmath72 is on the average twice as likely to come up as face number @xmath73 .
the mathematical representation of this situation is as follows .
the fact that we deal with dice is modelled in terms of multinomial distributions .
the probability that casting a @xmath74-sided die @xmath75 times yields @xmath76 instances for the @xmath77 face is@xmath78where @xmath79 with @xmath80 , and @xmath81 with @xmath82 .
the generic problem is to infer the parameters @xmath1 on the basis of information about moments of @xmath1 and data @xmath83 .
the additional information about how the dice are loaded is represented by the constraint @xmath84 .
note that this piece of information refers to the factory as a whole and not to any individual die .
the constraint is of the general form of ( [ < f>])@xmath85for this particular factory @xmath86 , and all @xmath87 except for @xmath88 and @xmath89 .
now that the background information has been given , here is our first example .
we purchase a die . on the basis of our general knowledge of dice
we are led to write down a joint prior @xmath90(the particular form of @xmath13 is not important for our current purpose so for the sake of definiteness we can choose it flat . ) at this point the only information we have is that we have a die and it came from a factory described by @xmath60 .
accordingly , we use me to update to a new joint distribution .
this is shown as @xmath58 in figure 1 .
the relevant entropy is@xmath11=-\tsum\limits_{m}\tint d\theta ~p(x,\theta ) \log \frac{% p(x,\theta ) } { p_{\text{old}}(x,\theta ) } ~,\]]where @xmath91maximizing @xmath19 subject to normalization and @xmath60 gives the @xmath58 posterior@xmath92where the normalization constant @xmath93 and the lagrange multiplier @xmath94 are determined from@xmath95the joint distribution @xmath96 can be rewritten as@xmath97 to find out more about this particular die we toss it @xmath75 times and obtain data @xmath98 which we represent as a new constraint@xmath99our goal is to infer the @xmath1 that apply to our particular die .
the original constraint @xmath60 applies to the whole factory while the new constraint @xmath59 refers to the actual die of interest and thus takes precedence over @xmath100 as @xmath101 we expect @xmath60 to become less and less relevant .
therefore the two constraints should be processed sequentially . using me ,
that is ( [ solution b ] ) , we impose @xmath59 and update from @xmath102 to a new joint distribution ( shown as @xmath103 in figure 1)@xmath104marginalizing over @xmath105 and using ( [ p1 ] ) the final posterior for @xmath1 is@xmath106where@xmath107 the readers will undoubtedly recognize that ( [ posterior a ] ) is precisely the result obtained by using maxent to obtain a prior , in this case @xmath108 given in ( [ p1 ] ) , and then using bayes theorem to take the data into account .
this familiar result has been derived in some detail for two reasons : first , to reassure the readers that me does reproduce the standard solutions to standard problems and second , to establish a contrast with the example discussed next .
here is a different problem illustrating the appropriateness of simultaneous updating .
the background information is the same as in the previous example .
the difference is that the factory now hires a quality control engineer who wants to learn as much as he can about the factory .
his initial knowledge is described by the same prior @xmath109 , ( [ p0 ] ) . after some inquiries he is told that the only available information is @xmath110 .
not satisfied with this limited information he decides to collect data that reflect the production of the whole factory .
randomly chosen dice are tossed @xmath75 times yielding data @xmath111 which is represented as a constraint,@xmath112the apparent resemblance with ( [ c2a ] ) may be misleading : ( [ c2a ] ) refers to a single die , while ( [ c2b ] ) now refers to the whole factory .
the goal here is to infer the distribution of @xmath1 that describes the overall population of dice produced by the factory .
the new constraint @xmath113 is information in addition to , rather than instead of , the old @xmath60 : the two constraints should be processed simultaneously . from ( [
joint posterior ] ) the joint posterior is or from @xmath58 we obtain the same posterior @xmath114 . ]
latexmath:[\[p_{\text{new}}^{(b)}(m,\theta ) = \delta ( m - m^{\prime } ) p_{\text{old}}(\theta over @xmath105 the posterior for @xmath1 is @xmath116where the new normalization constant is @xmath117this looks like the sequential case , ( [ posterior a ] ) , but there is a crucial difference : @xmath118 and @xmath119 . in the sequential updating case ,
the multiplier @xmath120 is chosen so that the intermediate @xmath58 satisfies @xmath60 while the posterior @xmath69 only satisfies @xmath59 . in the simultaneous updating case
the multiplier @xmath37 is chosen so that the posterior @xmath114 satisfies both @xmath60 and @xmath59 or @xmath71 .
ultimately , the two distributions @xmath121 are different because they refer to different problems : @xmath122 refers to a single die , while @xmath123 applies to all the dice produced by the factory . functions , had the constraints been processed sequentially but in the opposite order , first the data @xmath59 , and then the moment @xmath60 , the resulting posterior would be the same as for simultaneous update to @xmath124 . ]
the realization that the me method incorporates bayes rule as a special case has allowed us to go beyond bayes rule to process both data and expected value constraints simultaneously . to put it bluntly , anything one can do with bayes can also be done with me with the additional ability to include information that was inaccessible to bayes alone .
this raises several questions and we have offered a few answers .
first , it is not uncommon to claim that the non - commutability of constraints represents a _ problem _ for the me method .
processing constraints in different orders might lead to different inferences and this is said to be unacceptable .
we have argued that , on the contrary , the information conveyed by a particular sequence of constraints is not the same information conveyed by the same constraints in different order . since different informational states should in general lead to different inferences , the way me handles non - commuting constraints should not be regarded as a _ shortcoming _ but rather as a _ feature _ of the method .
second , we are capable of processing both data and moments .
is this kind of information of purely academic interest or is it something we might encounter in real life ? at this early stage
our answer must be tentative : we have given just one example the die factory which we think is fairly realistic .
however , we feel that other applications ( e.g. in econometrics and ecology ) can be handled in this way as well.giffinecon07,giffineco07 finally , is it really true that this type of problem lies beyond the reach of bayesian methods ?
after all , we can always interpret an expected value as a sample average in a sufficiently large number of trials .
we can always construct a large imaginary ensemble of experiments .
entropy methods then become in principle _ superfluous _ ; all we need is probability . the problem with inventing _ imaginary _
ensembles to do away with entropy in favor of mere probabilities , or to do away with probabilities in favor of more intuitive frequencies , is that the ensembles are just what they are claimed to be , imaginary .
they are purely artificial constructions invented for the purpose of handling incomplete information .
it seems to us that a safer way to proceed is to handle the available information directly as given ( i.e. , as expected values ) without making additional assumptions about an imagined reality .
* acknowledgements : * we would like to acknowledge valuable discussions with c. cafaro , k. knuth , and c. rodrguez
. 9 e. t. jaynes , phys .
rev . * 106 * , 620 and * 108 * , 171 ( 1957 ) ; r. d. rosenkrantz ( ed . ) , _
e. t. jaynes : papers on probability , statistics and statistical physics _
( reidel , dordrecht , 1983 ) ; e. t. jaynes , _ probability theory : the logic of science _
( cambridge university press , cambridge , 2003 ) .
j. e. shore and r. w. johnson , ieee trans .
theory * it-26 * , 26 ( 1980 ) ; ieee trans .
theory * it-27 * , 26 ( 1981 ) .
j. skilling , the axioms of maximum entropy , _ maximum - entropy and bayesian methods in science and engineering _ , g. j. erickson and c. r. smith ( eds . ) ( kluwer , dordrecht , 1988 ) .
a. caticha , relative entropy and inductive inference , _ bayesian inference and maximum entropy methods in science and engineering _ , g. j. erickson and y. zhai ( eds . ) , aip conf
* 707 * , 75 ( 2004 ) ( arxiv.org/abs/physics/0311093 ) .
a. caticha and a. giffin , updating probabilities , _ bayesian inference and maximum entropy methods in science and engineering _ , ed . by ali mohammad - djafari ( ed .
) , aip conf . proc . * 872 * , 31 ( 2006 ) ( http://arxiv.org/abs/physics/0608185 ) .
a. caticha , information and entropy , presented at the _
27th international workshop on bayesian inference and maximum entropy methods in science and engineering _ , saratoga springs , ny , 2007 .
a. giffin , updating probabilities with data and moments : an econometric example , to be presented at the _ 3rd econophysics colloquium _ , ancona , italy , 2007 .
a. giffin , updating probabilities with data and moments : an ecological example , to be presented at the _ 7th international conference on complex systems _ , boston , 2007 .
here we pursue the calculation of the posterior ( [ posterior b ] ) in more detail . to be specific we choose a flat prior , @xmath125 . then , dropping the superscript ( b ) , where @xmath127 differs from @xmath128 in ( [ zeta ] ) only by a combinatorial coefficient , @xmath129and @xmath37 is determined from ( [ f ] ) which in terms of @xmath127 now reads @xmath130 .
a brute force calculation gives @xmath127 as a nested hypergeometric series , @xmath131where each @xmath132 is written as a sum of @xmath133 functions , the index @xmath135 takes all values from @xmath136 to @xmath137 and the other symbols are defined as follows : @xmath138 , @xmath139 , and @xmath140with @xmath141 .
the terms that have indices @xmath142 are equal to zero ( i.e. @xmath143 etc . ) .
a few technical details are worth mentioning : first , one can have singular points when @xmath144 . in these cases
the sum must be evaluated as the limit as @xmath145 second , since @xmath146 and @xmath147 are positive integers the gamma functions involve no singularities .
lastly , the sums converge because @xmath148 .
the normalization for the first die example , ( [ z2 ] ) , can be calculated in a similar way .
currently , for small values of @xmath74 ( less than 10 ) it is feasible to evaluate the nested sums numerically ; for larger values of @xmath74 it is best to evaluate the integral for @xmath127 using sampling methods . a more detailed version of the multinomial example
is worked out in giffinecon07 . | we use the method of maximum ( relative ) entropy to process information in the form of observed data and moment constraints .
the generic canonical form of the posterior distribution for the problem of simultaneous updating with data and moments is obtained .
we discuss the general problem of non - commuting constraints , when they should be processed sequentially and when simultaneously . as an illustration , the multinomial example of die tosses
is solved in detail for two superficially similar but actually very different problems .
address = department of physics , university at albany
suny , albany , ny 12222,usa |
this contribution considers the problem of constructing detectors appropriate for doing short - baseline neutrino physics at the first muon collider complex .
the physics motivations for these detectors are discussed elsewhere in these proceedings@xcite .
since the proposed experiments are short - baseline , the physics being considered is primarily the high - energy physics of neutrino - nucleon deep - inelastic scattering ; however , the final section of the paper considers an oscillation experiment possible with the lowest energy neutrino beam .
the muon collider is expected to use a series of recirculating linacs to accelerate the muons before injection into a collider ring .
any segment along the muon s trajectory that is straight will necessarily create a collimated neutrino beam with an angular divergence of approximately @xmath0 .
the recirculating linacs ( rlas ) will have @xmath1 in which acceleration takes place , and the interaction points in the collider rings will also have @xmath2-@xmath3 m straight sections .
we wrote a simple monte carlo of the muon collider straight sections to predict fluxes , based on the workshop muon accelerator parameters .
we used the @xmath4 beam , and assumed the beam polarization and divergence were zero .
fluxes were predicted far downstream of each straight section in order to allow for shielding .
we considered only a `` dumb '' shield option in which sufficient concrete shielding to range out the muons from the primary muon beam was used .
table [ tab : flux ] gives a table of the distances of a proposed experiment downstream of each straight section in the rlas and collider .
.for each straight section in the rlas and for several different collider scenarios , the maximum muon energy , the required shielding distance , the @xmath5 beam size and the @xmath5 dis events in millions per g/@xmath6/yr are shown .
[ cols="<,^,^,^,^,^,^",options="header " , ]
the fluxes detailed above represent improvements by @xmath7 to @xmath2 orders of magnitude over contemporary high energy neutrino beams events / kg / yr with a mean energy of @xmath8 gev@xcite ; beams at the cern ps used by nomad and chorus had a mean energy of @xmath9 gev and observed @xmath10 events / kg / yr@xcite . ]
. it would therefore be possible to run an experiment with conventional high - mass sampling target - calorimeters and observe perhaps @xmath11 events per year in previously explored energy regimes . = however , sampling calorimeters would have some difficulties in the fmc environment .
high mass sampling calorimeters are only designed for the detection of or since and events are difficult to separate from neutral current interactions of other neutrino species , especially when the final state lepton has a low fraction of the initial neutrino energy . therefore , unless the muon beam polarity of the fmc were reversible , only neutrinos or anti - neutrinos could be observed in a single detector .
furthermore , high mass , sparsely instrumented detectors are sensitive to pile - up . as per pulse event rates in table [ tab : backg - pp ] show , this is a particular concern downstream of the high - energy recirculating linacs where the beam is concentrated in time in a relatively small number of turns .
figure [ fig : nutev - pileup ] shows an extreme illustration of the difficulties of dealing with pile - up in a sample - calorimeter .
assuming these difficulties could somehow be overcome , it is also unclear what the novel physics available from @xmath11 neutrino interactions in dense material would be . in general ,
precision measurements from neutrinos are systematics limited in such detectors@xcite and @xmath12@xcite in the ccfr experiment ] and searches for rare phenomena are background limited .
= 0.5 probably more interesting than the above detector technology would be one where the target is small enough to vary its composition for studies of nuclear effects on nucleon structure or even spin physics in neutrino deep inelastic scattering .
this is possible in a geometry more closely resembling traditional fixed - target experiments : target , followed by spectrometry and calorimetry and muon identification as shown in figure [ fig : ft - schematic ] .
the problems of tracking , particle - id and calorimetry in such detectors are well - known and will not be discussed here .
such a detector would presumably be able to identify the outgoing lepton in charged - current events , possibly with trds in the downstream tracking providing enhanced identification
. this detector would be able to tag the production of charmed particles by observing high momentum final state leptons , and with a sufficiently fine - grained upstream tracker , charmed final state particles could be tagged _ via _ detached vertices@xcite .
physics goals of an experiment using this detector are described elsewhere in these proceedings@xcite .
the most serious technical difficulties of such detectors is pile - up in the dense calorimetry and muon systems . a typical iron - scintillator sampling calorimeter / muon shield of @xmath13 ( @xmath14 m of steel or @xmath15 kg/@xmath6 )
would observe the per pulse background rates shown in table [ tab : backg - pp ] .
background rates at the collider are probably manageable , but the background rates at the rlas , particularly rla3 , could be difficult .
= = another avenue that becomes possible with the fmc neutrino beams is searching for neutrino oscillations in a modest - sized experiment . figure [ fig : rla1-qe ] shows the expected energy spectrum of quasi - elastic events downstream of rla1 . in a water - density 0.05m@xmath16 target
, one would expect to observe approximately @xmath17 million events per year .
this sets the scale for a truly `` table - top '' scale neutrino oscillation experiment . with a flavor selected beam of and , the simplest search experimentally appearance is background - laden at these energies and @xmath18 appearance , aside from being heavily suppressed by @xmath19 , requires a more sophisticated detector . ]
is for wrong - sign muon ( @xmath4 ) appearance .
the primary backgrounds would come from mis - identifying the muon charge or from production of charged pions which could decay or be misidentified as muons . by
conclusively identifying @xmath20 charge in a spectrometer , tracking the particle from the point of appearance and observing its decay with a characteristic muon lifetime , these backgrounds could presumably be kept very low . a naive version of a detector designed for this measurement appears in figure [ fig : table - top - det ] . with proper granuality of tracking and careful optimization of the magnetic field for a broad acceptance in momentum
, even such a small scale detector could reach @xmath21 sensitivity if naive background estimations are correct .
we have presented outlines of detector designs for short - baseline neutrino experiments at the first muon collider complex .
this facility would make available neutrino beams of unprecedented intensity , and we argue that the best way to take advantage of this is to design novel detectors , such as small targets or `` table - top '' neutrino experiments to explore different aspects of neutrino scattering or neutrino oscillations than those studied in the past .
harris and k.s .
mcfarland , `` a small target neutrino deep inelastic scattering experiment at the first muon collider '' , these proceedings .
r. bernstein _ et al .
_ , `` the nutev ssqt '' , fermilab - tm-1884 , apr 1994 . j. altegoer _
( nomad collaboration ) , cern - ppe-97 - 059 , may 1997 .
e. eskut _ et al .
_ ( chorus collaboration ) , cern - ppe-97 - 033 , mar 1997 .
seligman _ et al .
_ , `` a new determination of @xmath22 from neutrino - nucleon scattering '' , nevis - report-292 , jan 1997 , to appear in phys . rev .
et al . _ ,
`` a precision measurement of electroweak parameters in neutrino - nucleon scattering '' , fermilab - pub-97 - 001-e , jan 1997 , to appear in z. phys .
* c*. bruce king , these proceedings . | we consider possible detector designs for short - baseline neutrino experiments using neutrino beams produced at the first muon collider complex .
the high fluxes available at the muon collider make possible high statistics deep - inelastic scattering neutrino experiments with a low - mass target . a design of a low - energy neutrino oscillation experiment on the `` tabletop '' scale is also discussed . |
our main theorem gives a relation between _ fibrations _ ( or _ surface bundles _ ) and a related notion of _ formal fibrations_. by a fibration we shall mean a smooth map @xmath0 , where @xmath1 and @xmath2 are smooth , oriented , compact manifolds and @xmath3 is a submersion ( i.e. @xmath4 is surjective ) .
a cobordism between two fibrations @xmath5 and @xmath6 is a triple @xmath7 where @xmath8 is a cobordism from @xmath9 to @xmath10 , @xmath11 is a cobordism from @xmath12 to @xmath13 , and @xmath14 is a submersion which extends @xmath15 .
[ defn : formal - fib ] a. an _ unstable formal fibration _ is a pair @xmath16 , where @xmath17 is a smooth proper map , and @xmath18 is a bundle epimorphism .
b. a _ stable formal fibration _ ( henceforth just a formal fibration ) is a pair @xmath16 , where @xmath3 is as before , but @xmath19 is defined only as a _
stable _ bundle map .
thus for large enough @xmath20 there is given an epimorphism @xmath21 , and we identify @xmath19 with its stabilization @xmath22 . a cobordism between formal fibrations @xmath23 and @xmath24 is a quadruple @xmath25 which restricts to @xmath26 . c. the formal fibration induced by a fibration @xmath17 is the pair @xmath27 , and
a formal fibration is _ integrable _ if it is of this form .
our main theorem relates the set of cobordism classes of fibrations with the set of cobordism classes of formal fibrations .
let us first discuss the _ stabilization _ process ( or more precisely `` stabilization with respect to genus '' .
this should not be confused with the use of `` stable '' in `` stable formal fibration '' . in the former
, `` stabilization '' refers to increasing genus ; in the latter it refers to increasing the dimension of vector bundles . )
suppose @xmath28 is a formal fibration ( we will often suppress the bundle epimorphism @xmath19 from the notation ) and @xmath29 is an embedding over @xmath2 ( i.e. @xmath30 ) , such that @xmath3 is integrable on the image of @xmath20
. then we can _ stabilize _
@xmath3 by taking the fiberwise connected sum of @xmath1 with @xmath31 ( along @xmath20 ) , where @xmath32 is the torus .
if @xmath3 happens to be integrable , this process increases the genus of each fiber by 1 .
our main theorem is the following .
[ thm : main0 ] let @xmath33 be a formal fibration which is integrable over the image of a fiberwise embedding @xmath34 .
then , after possibly stabilizing a finite number of times , @xmath3 is cobordant to an integrable fibration with connected fibers
. we will also prove a relative version of the theorem .
namely , if @xmath2 has boundary and @xmath3 is already integrable , with connected fibers , over a neighborhood of @xmath35 , then the cobordism in the theorem can be also be assumed integrable over a neighborhood of the boundary .
theorem [ thm : main0 ] is equivalent to madsen - weiss `` generalized mumford conjecture '' @xcite which states that a certain map @xmath36 induces an isomorphism in integral homology . in the rest of this introduction we will explain the equivalence and introduce the methods that go into our proof of theorem [ thm : main0 ] .
our proof is somewhat similar in ideas to the original proof , but quite different in details and language .
the general scheme of reduction of some algebro - topological problem to a problem of existence of bordisms between formal and genuine ( integrable ) fibrations was first proposed by d. b. fuchs in @xcite .
that one might prove madsen - weiss theorem in the form of theorem [ thm : main0 ] was also suggested by i. madsen and u. tillmann in @xcite .
let @xmath37 be a compact oriented surface , possibly with boundary @xmath38 .
let @xmath39 denote the topological group of diffeomorphisms of @xmath37 which restrict to the identity on the boundary .
the classifying space @xmath40 can be defined as the orbit space @xmath41 and it is a classifying space for fibrations : for a manifold @xmath2 , there is a natural bijection between isomorphism classes of smooth surface bundles @xmath42 with fiber @xmath37 and trivialized boundary @xmath43 , and homotopy classes of maps @xmath44 . the _ mapping class group _ is defined as @xmath45 , i.e. the group of isotopy classes of diffeomorphisms of the surface .
it is known that the identity component of @xmath39 is contractible ( as long as @xmath46 or @xmath47 ) , so @xmath40 is also a classifying space for @xmath48 ( i.e. an eilenberg - mac lane space @xmath49 ) . when @xmath50 , this is also related to the _ moduli space _ of riemann surfaces ( i.e. the space of isomorphism classes of riemann surfaces of genus @xmath51 ) via a map @xmath52 which induces an isomorphism in rational homology and cohomology .
mumford defined characteristic classes @xmath53 for @xmath54 and conjectured that the resulting map @xmath55 \to h^*(\mathcal{m}_g;{\mathbb { q}\,})\end{aligned}\ ] ] is an isomorphism in degrees less than @xmath56 .
this is the original _
mumford conjecture_. it is convenient to take the limit @xmath57 . geometrically , that can be interpreted as follows .
pick a surface @xmath58 of genus @xmath51 and with one boundary component . also pick an inclusion @xmath59 .
let @xmath60 be the union of the @xmath58 over all @xmath51 , i.e. a countably infinite connected sum of tori .
we will consider fibrations @xmath42 with trivialized `` @xmath60 ends '' .
-end , height=170 ] we are considering pairs @xmath61 where @xmath62 is a smooth fiber bundle with fiber @xmath60 , and @xmath63 is a _ germ at infinity _ of an embedding over @xmath2 .
this means , for @xmath2 compact , that representatives of @xmath20 are defined on the complement of some compact set in @xmath64 , and their images contain the complement of some compact set in @xmath65 .
let us describe a classifying space for fibrations with @xmath60 ends .
let @xmath66 be the the mapping telescope ( alias _ homotopy colimit _ ) of the direct system @xmath67 @xmath68 is a classifying space for fibrations with @xmath60 ends .
any compact @xmath69 is contained in @xmath70 for some finite @xmath51 .
let @xmath71 be the complement of @xmath58 .
let us consider for a moment fibrations with fiber @xmath60 and an embedding as in ( [ eq:6 ] ) , but which is actually defined on @xmath72 .
call such bundles _
@xmath73-trivialized_. specifying a @xmath73-trivialized bundle is the same thing as specifying a fibration @xmath74 with connected , compact fibers , and trivialized boundary @xmath75 ( namely @xmath76 is the complement of the image of @xmath20 ) .
thus , the disjoint union @xmath77 is a classifying space for @xmath73-trivialized bundles ( notice @xmath78 is independent of @xmath73 ) .
a @xmath73-trivialized bundle is also a @xmath79-trivialized bundle , and increasing @xmath73 is represented by a `` stabilization '' self - map @xmath80 . in the representation , @xmath81 maps @xmath82 to @xmath83 and this is induced by the same map as in .
now the statement follows by taking the direct limit as @xmath84 , and noticing that @xmath68 is the homotopy colimit of the direct system @xmath85 in this section we define a space @xmath86 and interpret it as a classifying space for formal fibrations .
the forgetful functor from fibrations to formal fibrations is represented by a map @xmath87 we then consider the same situation , but where fibrations and formal fibrations have @xmath60 ends .
this changes the source of the map to @xmath68 , but turns out to not change the homotopy type of the target .
we get a map @xmath88 representing the forgetful functor from fibrations with @xmath60 ends to formal fibrations with @xmath60 ends .
the space @xmath86 is defined as the thom spectrum of the negative of the canonical complex line bundle over @xmath89 . in more detail , let @xmath90 be the grassmannian manifold of oriented 2-planes in @xmath91 .
it supports a canonical 2-dimensional vector bundle @xmath92 with an @xmath93-dimensional complement @xmath94 such that @xmath95 .
there is a canonical identification @xmath96 the thom space @xmath97 is defined as the one - point compactification of the total space of @xmath98 , and the identification ( [ eq:2 ] ) induces a map @xmath99 .
the space @xmath100 is defined as the direct limit @xmath101 like we did for @xmath40 , we shall think of @xmath100 as a classifying space , i.e. interpret homotopy classes of maps @xmath102 from a smooth manifold @xmath2 in terms of certain geometric objects over @xmath2 .
recall the notion of _ formal fibration _ from definition [ defn : formal - fib ] above .
a cobordism @xmath25 of formal fibrations is a _ concordance _ if the target cobordism is a cylinder : @xmath103 $ ] .
[ lemma : pt1 ] there is a natural bijection between set @xmath104 \end{aligned}\ ] ] of homotopy classes of maps , and the set of concordance classes of formal fibrations over @xmath2 .
this is the standard argument of pontryagin - thom theory . in one direction ,
given a map @xmath105 , one makes the adjoint map @xmath106 transverse to the zero section of @xmath98 and sets @xmath107 .
then @xmath1 comes with a map @xmath108 and the normal bundle of @xmath109 is @xmath110 .
this gives a stable isomorphism @xmath111 and hence a stable epimorphism @xmath112 . in the other direction ,
given a formal fibration @xmath16 with @xmath113 , we pick an embedding @xmath109 . letting @xmath114
be the normal bundle of this embedding , we get a `` collapse '' map @xmath115 we also get an isomorphism of vector bundles over @xmath1 @xmath116 let @xmath117 be a classifying map for the kernel of the stable epimorphism @xmath118 ( this is a two - dimensional vector bundle with orientation induced by the orientations of @xmath2 and @xmath1 ) , so we have a stable isomorphism @xmath119 . combining with ( [ eq:8 ] )
we get a stable isomorphism @xmath120 . by adding @xmath121 we get a stable isomorphism @xmath122 which we can assume is induced by an unstable isomorphism ( since we can assume @xmath123 )
@xmath124 this gives a proper map @xmath125 and hence a map of thom spaces @xmath126 .
compose with ( [ eq:10 ] ) and take the adjoint to get a map @xmath127 .
finally let @xmath128 to get a map @xmath102 . the homotopy class of the resulting map @xmath102 is well defined and depends only on the concordance class of the formal fibration @xmath17 .
a fibration naturally gives rise to a formal fibration , and this association gives rise to a map of classifying spaces which is the map ( [ eq:5 ] ) .
we would like to make this process compatible with the stabilization procedure explained in section [ sec : diff - groups - mapp ] above .
to this end we consider formal fibrations with @xmath73-trivialized @xmath60 ends .
this means that @xmath17 is equipped with an embedding over @xmath2 @xmath129 and that @xmath16 is integrable on the image of @xmath20 .
of course , we also replace the requirement that @xmath1 be compact by the requirement that the complement of the image of @xmath20 be compact .
[ lemma : pt2 ] formal fibrations with @xmath73-trivialized ends are represented by the space @xmath100 .
this is similar to the proof of lemma [ lemma : pt1 ] above . applying the pontryagin - thom construction from the proof of that lemma to the projection @xmath130
gives a path @xmath131 . applying the pontryagin - thom construction to an arbitrary
@xmath73-trivialized formal fibration gives a path @xmath132 whose restriction to @xmath133 is @xmath134 .
the space of all such paths is homotopy equivalent to the loop space @xmath135 .
increasing @xmath73 gives a diagram of stabilization maps @xmath136^s \ar[r ] & { \omega^\infty { \mathbb { c}\,}p^\infty_{-1 } } \ar[d]^s\\ { \coprod_g b { \mathrm{diff}}(f_{g,1 } ) } \ar[r ] & { \omega^\infty { \mathbb { c}\,}p^\infty_{-1}. } } \end{aligned}\ ] ] on the right hand `` formal '' side , the stabilization map is up to homotopy described as multiplication by a fixed element ( multiplication in the loop space structure . see the proof of lemma [ lemma : pt2 ] . ) in particular it is a homotopy equivalence , so the direct limit has the same homotopy type @xmath137 .
taking the direct limit we get the desired map @xmath138 this is the map . the target of this map
should be thought of as the homotopy direct limit of the system @xmath139 and as a classifying space for formal fibrations with @xmath60 ends .
( in some sense it is a `` coincidence '' that the classifying space for formal fibrations and the classifying space for formal fibrations with @xmath60 ends have the same homotopy type ) . for a pair @xmath140 of spaces ,
oriented bordism @xmath141 is defined as the set of bordism classes of continuous maps of pairs @xmath142 for smooth oriented compact @xmath143-manifolds @xmath1 with boundary @xmath144 . to be precise ,
a bordism between two maps @xmath145 is a map @xmath146 , where @xmath8 is a compact , oriented manifold with boundary with corners , so that @xmath147 , where @xmath148 and @xmath149 is a cobordism between closed manifolds @xmath150 and @xmath151 , and the map @xmath152 such that @xmath153 . for a single space @xmath2 set @xmath154 .
oriented bordism is a _ generalized homology theory_. this means that it satisfies the usual eilenberg - steenrod axioms for homology ( long exact sequence etc ) except for the dimension axiom . in particular a map @xmath155 induces an isomorphism @xmath156 if and only if the relative groups @xmath157 all vanish .
the following result is well known .
it follows easily from the atiyah - hirzebruch spectral sequence ( for completeness we give a geometric proof in appendix b ) .
[ lem : bordism - to - homology ] let @xmath158 be a continuous map of topological spaces .
then the following statements are equivalent .
a. @xmath159 is an isomorphism for @xmath160 and an epimorphism for @xmath161 .
b. @xmath162 is an isomorphism for @xmath163 and an epimorphism for @xmath161 .
in particular , @xmath3 induces an isomorphism in homology in all degrees if and only if it does so in oriented bordism .
we apply this to the pair @xmath164 . interpreting @xmath68 and @xmath86 as classifying spaces for fibrations , resp .
formal fibrations , with @xmath60 ends , we get the following interpretaion . [ lemma : interpret - relative - bord ] there is a natural bijection between the relative oriented bordism groups @xmath165 and cobordism classes of formal fibrations @xmath166 with @xmath60 ends .
the formal fibration is required to be integrable over a neighborhood of @xmath35 , and cobordisms @xmath167 are required to be integrable over a neighborhood of @xmath168 .
that induces an isomorphism in integral homology ( madsen - weiss theorem ) is now , by lemma [ lem : bordism - to - homology ] , equivalent to the statement that the relative groups @xmath169 all vanish . by lemma [ lemma :
interpret - relative - bord ] , this is equivalent to [ thm : main1 ] any formal fibration @xmath17 with @xmath60 ends is cobordant to an integrable one .
if @xmath16 is already integrable over @xmath35 , then the cobordism can be assumed integrable over @xmath170 .
theorem [ thm : main1 ] is our main result .
it is a geometric version of madsen - weiss theorem .
it is obviously equivalent to theorem [ thm : main0 ] above ( with its relative form ) .
j. harer proved a homological stability theorem in @xcite which implies precise bounds on the number of stabilizations needed in theorem [ thm : main0 ] . at the same time
, it will be an important part of the proof of the same theorem ( as it does in @xcite ) .
roughly it says that the homology of the mapping class group of a surface @xmath37 is independent of the topological type @xmath37 , as long as the genus is high enough .
the result was later improved by ivanov ( @xcite ) and then by boldsen @xcite .
we state the precise result .
consider an inclusion @xmath171 of compact , connected , oriented surfaces .
let @xmath172 , and let @xmath173 denote the complement of @xmath37 . thus @xmath174 .
there is an induced map of classifying spaces @xmath175 a map @xmath176 classifies a fibration @xmath42 with fiber @xmath177 and boundary @xmath43 , where @xmath178 .
lifting it to a map into @xmath40 amounts to extending the embedding @xmath179 to an embedding @xmath180 over @xmath2 .
the most general form of harer stability states that the map ( [ eq:3 ] ) induces an isomorphism in @xmath181 for @xmath182 , where @xmath51 is the genus of @xmath37 .
consequently , by lemma [ lem : bordism - to - homology ] , it induces an isomorphism in oriented bordism @xmath183 for @xmath184 or , equivalently , the relative bordism group @xmath185 vanishes for @xmath184 .
thus , harer stability has the following very geometric interpretation : for any fibration @xmath186 with fiber @xmath177 and boundary @xmath43 , @xmath3 is cobordant to a fibration @xmath187 via a cobordism @xmath188 ( which is a fibration with trivialized boundary @xmath189 , which restricts to @xmath190 ) where the embedding @xmath191 extends to an embedding @xmath192 moreover , this can be assumed compatible with any given extension @xmath193 over the boundary of @xmath2 .
here we assume @xmath194 as above , that @xmath37 and @xmath177 are connected , and that @xmath37 has large genus . if the fibration has @xmath60 ends , the genus assumption is automatically satisfied , and we get the following corollary .
[ thm : harer - geometric ] let @xmath195 be compact surfaces with boundary ( not necessarily connected ) .
let @xmath17 be a fibration with @xmath60 ends , and let @xmath196 be a fiberwise embedding over @xmath2 , such that in each fiber the complement of its image is connected .
, height=226 ] then , after possibly changing @xmath17 by a bordism which is the trivial bordism over @xmath35 , the embedding @xmath20 extends to an embedding of @xmath197 . explicitly , the bordism in the theorem is a fibration @xmath14 with @xmath60 ends , where the boundary @xmath198 is partitioned as @xmath199 with @xmath200 and @xmath201 .
the extension of @xmath20 is a fiberwise embedding @xmath202 over @xmath11 .
given a formal fibration @xmath203 with @xmath204 ends , the overall aim is to get rid of all singularities of @xmath3 after changing it via _ bordisms_. our first task will be to simplify the singularities of @xmath3 as much as possible using only _
homotopies_. the simplest generic singularities of a map @xmath205 are _
folds_. the fold locus @xmath206 consists of points where the rank of @xmath3 is equal to @xmath207 , while the restriction @xmath208 is an immersion .
in the case when @xmath209 , which is the case we consider in this paper , we have @xmath210 .
a certain additional structure on folded maps , called an _ enrichment _ , allows one to define a homotopically canonical _ suspension _ , i.e. a bundle epimorphism @xmath211 , such that @xmath16 is a formal fibration .
the enrichment of a folded map @xmath3 consists of * an @xmath143-dimensional submanifold @xmath212 such that @xmath213 , and the restriction of @xmath3 to each connected component of @xmath214 is an embedding @xmath215 ; * a trivialization of the bundle @xmath216 with a certain additional condition on the behavior of this trivialization on @xmath213 .
of course , existence of an enrichment is a strong additional condition on the fold map . in section
[ sec : enriched ] , we explain how to associate to an enriched folded map @xmath217 a formal fibration @xmath218 , where @xmath219 is a bundle epimorphism associated to the enrichment @xmath220
. the main result of section [ sec : prelim ] is theorem [ thm : epi - folds ] , which proves that any formal fibration can be represented in this way ( plus a corresponding relative statement ) .
this is proved using the @xmath221-principle type result proven in @xcite .
note that theorem [ thm : epi - folds ] is a variation of the main result from @xcite and can also be proven by the methods of that paper .
also in section [ sec : prelim ] we recall some basic facts about folds and other simple singularities of smooth maps , and discuss certain surgery constructions needed for the rest of the proof of theorem [ thm : main1 ] .
this part works independently of the codimension @xmath222 , and hence the exposition in this section is done for arbitrary @xmath223 .
each fold component has an index which is well defined provided that the projection of the fold is co - oriented . assuming this is done , folds in the case
@xmath224 can be of index @xmath225 and @xmath226 .
we call folds of index @xmath227 _ hyperbolic _ and folds of index @xmath228 _ elliptic_. it is generally impossible to get rid of elliptic folds by a homotopy of the map @xmath3
. however , it is easy to do so if one allows to change @xmath3 to a _ bordant _
. this bordism trades each elliptic fold component by a parallel copy of a hyperbolic fold , see figure [ mw11 ] and section [ sec : to - hyperbolic ] below .
a similar argument allows one to make all fibers @xmath230 , connected ( comp .
@xcite ) .
a generalization of harer s stability theorem to enriched folded maps , see theorem [ thm : harer2 ] below , plays an important role in our proof . in section [ sec : proof - harer ] below we deduce theorem [ thm : harer2 ] from harer s theorem [ thm : harer - geometric ] by induction over strata in the stratification of the image @xmath231 of the fold , according to multiplicity of self - intersections .
let @xmath3 be an enriched folded map with hyperbolic folds and with connected fibers .
let @xmath232 be one of the fold components and @xmath233 its image .
for the purpose of this introduction we will consider only the following special case .
first , we will assume that @xmath232 is homologically trivial . as we will see , when @xmath234 this will always be possible to arrange . in particular , @xmath235 bounds a domain @xmath236 .
next , we will assume that the fold @xmath232 has index 1 with respect to the outward coorientation of the boundary of the domain @xmath237 .
in other words , when the point @xmath238 travels across @xmath235 inside @xmath237 then one of the circle in the fiber @xmath239 collapses to a point , so locally the fiber gets disconnected to two discs , see figure [ mw13 ] . the inverse index 1 surgery makes a connected sum of two discs at their centers .
note that on an open collar @xmath240 along @xmath232 in @xmath237 there exists two sections @xmath241 such that the @xmath242-sphere @xmath243 is the `` vanishing cycle '' , for the index @xmath244 surgery when @xmath245 travels across @xmath235 .
moreover , the enrichment structure ensures that the vertical bundle along these sections is trivial .
if one could extend the sections @xmath246 to all of @xmath237 preserving all these properties , then the fiberwise index @xmath244 surgery , attaching @xmath244-handle along small discs surrounding @xmath247 and @xmath247 , @xmath248 , would eliminate the fold @xmath235 .
this is one of the fold surgeries described in detail in section [ sec : surgery - general ] .
though such extensions @xmath249 need not exist for our original folded map @xmath3 , harer stability theorem in the form [ sec : surgery - general ] states that there is an enriched folded map @xmath250 , bordant to @xmath3 , for which such sections do exist , and hence the fold @xmath235 could be eliminated .
as already mentioned , section [ sec : prelim ] recalls basic definition and necessary results and constructions involving fold singularities . in section [ sec : folds - cusps ] we define folded maps .
the goal of section [ sec : double - folds ] is theorem [ thm : special - folded ] which is an @xmath221-principle for constructing so - called _ special folded _
maps , whose folds are organized in pairs of spheres . this theorem is a reformulation of the wrinkling theorem from @xcite . we deduce theorem [ thm : special - folded ] from the wrinkling theorem in appendix a. in section [ sec : enriched ] we define the notion of enrichment for folded maps and prove that an enriched folded map admits a homotopically canonical suspension and hence gives rise to a formal fibration . the rest of section [ sec : prelim ] will prove that any formal fibration is cobordant to one induced by an enriched folded map .
section [ sec : surgery - general ] is devoted to fold surgery constructions which we use later in the proof of the main theorem .
these are just fiberwise morse surgeries , in the spirit of surgery of singularities techniques developed in @xcite . for further applications we need a version of theorem [ thm :
special - folded ] applicable to a slightly stronger version of formal fibrations @xmath16 when @xmath251 is a surjective map between the _ non - stabilized _ tangent bundle . in section [ sec : destab ] we explain the modifications which are necessary in the stable case , and in section [ sec : epi - folds ] we formulate and prove theorem [ thm : epi - folds ] , reducing formal fibrations to enriched folded maps . in section [ sec : cob ] we introduce several special bordism categories and formulate the two remaining steps of the proof : proposition [ prop : enriched - hyperbolic ] which allows us to get rid of elliptic folds , and proposition [ prop : hyperbolic - fibrations ] which eliminates the remaining hyperbolic folds . section [ sec : harer - generalized ] is devoted to the proof of the harer stability theorem for folded maps ( theorem [ thm : harer2 ] ) .
we conclude the proof of theorem [ thm : main1 ] in section [ sec : main - proof ] by proving proposition [ prop : enriched - hyperbolic ] in [ sec : to - hyperbolic ] and proposition [ prop : hyperbolic - fibrations ] in [ sec : to - fibrations ] . in section [ sec : misc ] we collect two appendices . in appendix
a we deduce theorem [ thm : special - folded ] from the wrinkling theorem from @xcite .
appendix b is devoted to a geometric proof of lemma [ lem : bordism - to - homology ] .
let @xmath1 and @xmath2 be smooth manifolds of dimension @xmath252 and @xmath143 , respectively .. ] for a smooth map @xmath253 we will denote by @xmath206 the set of its singular points , i.e. @xmath254 a point @xmath255 is called a _
fold _ type singularity or a _ fold _ of index @xmath73
if near the point @xmath256 the map @xmath3 is equivalent to the map @xmath257 given by the formula @xmath258 where @xmath259 and @xmath260
. we will also denote @xmath261 and write @xmath262 . for @xmath263
this is just a non - degenerate index @xmath73 critical point of the function @xmath264 . by a _ folded map
_ we will mean a map with only fold type singularities .
given @xmath265 and an @xmath266 we will call a @xmath267-dimensional sphere @xmath268 the _ vanishing cycle _ of the fold over the point @xmath269 .
the normal bundle of the image @xmath270 of the fold is a real line bundle over @xmath232 .
a _ coorientation _ of @xmath235 is a trivialization of this line bundle .
a choice of coorientation allows one to provide each fold component @xmath232 with a well - defined index @xmath81 , which changes from @xmath81 to @xmath271 with a switch of the coorientation .
the normal bundle of @xmath232 is @xmath272 , and the second derivatives of @xmath3 gives an invariantly defined non - degenerate quadratic form @xmath273 .
denote @xmath274 .
there is a splitting @xmath275 which is defined uniquely up to homotopy by the condition @xmath276 .
given an orientable @xmath277-dimensional manifold @xmath232 , let us consider the map @xmath278 given by the formula @xmath279 where @xmath280 and @xmath281 .
the singularity @xmath282 consists of two copies of the manifold @xmath232 : @xmath283 the manifold @xmath284 is a fold of index @xmath73 , while @xmath285 is a fold of index @xmath286 , with respect to the coorientation of the folds in the image given by the vector field @xmath287 where @xmath288 is the coordinate @xmath289 .
it is important to notice that the restriction of the map @xmath290 to the annulus @xmath291 is an _
embedding_. , height=151 ] although the differential @xmath292 degenerates at points of singularity @xmath293 , it can be canonically _ regularized _ over @xmath294 , an open neighborhood of the annulus @xmath295 .
namely , we can change the element @xmath296 in the jacobi matrix of @xmath290 to a positive function @xmath297 , which coincides with @xmath296 on @xmath298 $ ] for sufficiently small @xmath299 .
the new bundle map @xmath300 provides a homotopically canonical extension of the map @xmath301 to an epimorphism ( fiberwise surjective bundle map ) @xmath302 we call @xmath303 the _ regularized differential _ of the map @xmath290 .
a map @xmath304 defined on an open @xmath305 is called a _
double @xmath232-fold _ of index @xmath306 if it is equivalent to the restriction of the map @xmath290 to @xmath307 .
for instance , when @xmath308 and @xmath232 is a point , a double @xmath232-fold is a morse function given in a neighborhood of a gradient trajectory connecting two critical points of neighboring indices . in the case of general @xmath143 , a double @xmath232-fold
is called a _
spherical double fold _ if @xmath309 .
it is always easy to create a double @xmath232-fold as the following lemma shows .
this lemma is a parametric version of the creation of two new critical points of a morse function .
[ lm:2fold - creation ] given a submersion @xmath310 of a manifold @xmath311 , a closed submanifold @xmath312 of dimension @xmath313 such that @xmath314 is an embedding with trivialized normal bundle , and a splitting @xmath315 of the vertical bundle @xmath316 over @xmath317 , one can construct a map @xmath318 such that * @xmath319 coincides with @xmath3 near @xmath320 ; * in a neighborhood of @xmath232 the map @xmath319 has a double @xmath232-fold , i.e. it is equivalent to the map restricted to @xmath321 , where the frames @xmath322 along @xmath323 provide the given trivializations of the bundles @xmath324 and @xmath325 ; * @xmath326 and @xmath327 are homotopic via a homotopy of epimorphisms fixed near @xmath320 .
there exists splittings @xmath328\times d^d$ ] , where @xmath329 is the unit @xmath330-disc , and @xmath331 $ ] of neighborhoods @xmath332 in @xmath311 and @xmath333 in @xmath2 , such that the map @xmath3 has the form @xmath334\times [ -1,1]^d\ni ( v , z , x=(x_1,\dots , x_d))\mapsto ( { \overline{v}}=f(v),z)\in { \overline c}\times[-2,2].\ ] ] consider a @xmath335 function @xmath336\to[-2,2]$ ] which coincides with @xmath337 on @xmath338 $ ] , with @xmath339 near @xmath340 , and such that @xmath341 are its only critical points .
take two cut - off @xmath342-functions @xmath343 $ ] such that @xmath344 on @xmath345 $ ] , @xmath346 on @xmath347 $ ] , @xmath348 on @xmath349 $ ] and @xmath350 on @xmath351 $ ] . set @xmath352 and define first a map @xmath353 by the formula @xmath354 and then extend it to @xmath1 , being equal to @xmath3 outside @xmath355 .
the regularized differential of a linear deformation between @xmath3 an @xmath356 provides the required homotopy between @xmath326 and @xmath327 .
[ rem : killing ] it can be difficult to eliminate a double @xmath232-fold . even in the case @xmath357 this is one of the central problems of morse theory .
a map @xmath253 is called _ special folded _
, if there exist disjoint open subsets @xmath358 such that the restriction @xmath359 is a submersion ( i.e. has rank equal @xmath143 ) and for each @xmath360 the restriction @xmath361 is a spherical double fold .
_ in addition , _ we require that the images of all fold components bound balls in @xmath2 .
the singular locus @xmath206 of a special folded map @xmath3 is a union of @xmath277-dimensional double spheres @xmath362 .
it is convenient to fix for each double sphere @xmath363 the corresponding annulus @xmath364 which spans them .
notice that although the restriction of @xmath3 to each annulus @xmath365 is an _ embedding _ , the restriction of @xmath3 to the union of all the annuli @xmath366 is , in general , only an _ immersion _
, because the images of the annuli may intersect each other . using an appropriate version of the transversality theorem we can arrange by a @xmath367-small perturbation of @xmath3 that all combinations of images of its fold components intersect transversally .
the differential @xmath368 can be regularized to obtain an epimorphism @xmath369 . to get @xmath370 we regularize @xmath371 for each double fold @xmath361 . in our proof of theorem [ thm : main1 ]
we will use the following result about special folded maps .
[ thm : special - folded ] * [ special folded mappings]*[thm : special ] let @xmath372 be an epimorphism which covers a map @xmath373 .
suppose that @xmath3 is a submersion on a neighborhood of a closed subset @xmath374 , and that @xmath37 coincides with @xmath326 over that neighborhood .
then if @xmath223 then there exists a special folded map @xmath375 which coincides with @xmath3 near @xmath376 and such that @xmath377 and @xmath37 are homotopic rel .
moreover , the map @xmath51 can be chosen arbitrarily @xmath379-close to @xmath3 and with arbitrarily small double folds .
let us stress again the point that a special folded map @xmath17 by definition has only spherical double folds , each fold component @xmath380 is a sphere whose image @xmath381 is embedded and bounds a ball in @xmath2 . in the equidimensional case ( @xmath382 )
it is not possible , in general , to make images of fold components embedded .
see appendix a below .
theorem [ thm : special ] is a modification of the wrinkling theorem from @xcite .
we formulate the wrinkling theorem ( see theorem [ thm : wrinkled ] ) and explain how to derive [ thm : special ] from [ thm : wrinkled ] in appendix a below .
special folded mappings give a nice representation of ( unstable ) formal fibrations . as a class of maps
, it turns out to be too small for our purposes .
namely , we wish to perform certain constructions ( e.g. surgery ) which does not preserve the class of special folded maps .
hence we consider a larger class of maps which allow for these constructions to be performed , and still small enough to admit a homotopically canonical extension to a formal fibration .
this is the class of enriched folded maps .
recall that a folded map is , by definition , a map @xmath383 which locally is of the form ( [ eq : fold ] ) . in this section
we study a certain extra structure on folded maps which we dub _ framed membranes_. [ defn : membrane ] let @xmath384 be closed manifolds and @xmath17 be a folded map .
a _ framed membrane _ of index @xmath73 , @xmath385 , for @xmath3 is a compact , connected @xmath143-dimensional submanifold @xmath386 with boundary @xmath387 , together with a _ framing _
@xmath388 where @xmath389 are _ trivialized _ subbundles of @xmath390 of dimension @xmath73 and @xmath391 , respectively , such that a. the restriction @xmath392 is an embedding ; b. @xmath393 are transversal to each other and to @xmath394 ; c. there exists a co - orientation of the image @xmath235 of each fold component @xmath395 such that @xmath396 ; thus , over @xmath397 we have @xmath398 , while over @xmath399 @xmath400 splits as @xmath401 , where @xmath402 is a line bundle contained in @xmath403 . a boundary component of a membrane @xmath404 is called _ positive _ if @xmath405 , and _ negative _ otherwise .
we will denote by @xmath406 and @xmath407 , respectively , the union of positive and negative boundary components of @xmath404 .
note that the coorientation of a component @xmath408 implied by the definition of a framed membrane is given by inward normals to @xmath409 if @xmath410 , and by outward normals to @xmath409 if @xmath411 .
the index of the fold component @xmath232 is equal to @xmath73 in the former case , and to @xmath286 in the latter one .
we will call a framed membrane @xmath412 _ pure _ if either @xmath413 , or @xmath414 .
otherwise we call it _
mixed_. switching the roles of the subbundles @xmath325 and @xmath324 gives a _ dual _ framing @xmath415 .
the index of the framed membrane @xmath416 equals @xmath391 , and we also have @xmath417 . [ defn : enrichment ] an _ enriched _ folded map is a pair @xmath418 where @xmath17 is a folded map and @xmath419 is an _ enrichment _ of @xmath3 , consisting of finitely many disjoint framed membranes @xmath420 in @xmath1 such that @xmath421 , where @xmath404 is the union of the @xmath422 .
we point out that while the definition implies @xmath3 is injective on each @xmath422 , the images @xmath423 need not be disjoint .
[ ex : double - fold ] the double @xmath232-fold @xmath290 defined by has the annulus @xmath424 as its membrane . together with the frame @xmath425 , where the subbundles @xmath324 and @xmath325 are generated , respectively by @xmath426 along @xmath323 the membrane defines a _
canonical enrichment _ of the double @xmath232-fold @xmath290 . in particular , _ any special folded map has a canonical enrichment . _ note that we have @xmath427 and @xmath428 . from the tubular neighborhood theorem and a parametrized version of morse s lemma , we get [ lemma - replacing - eq:14 ] let @xmath412 be a connected framed membrane for an enriched folded map @xmath429 .
let @xmath430 be a connected component of @xmath404 , and @xmath431 its image .
then there is a tubular neighborhood @xmath432 of @xmath232 with coordinate functions @xmath433 and a tubular neighborhood @xmath434 with coordinate functions @xmath435 such that we have - @xmath436 along @xmath232 ; - the vector field @xmath437 defines the coorientation of @xmath235 implied by the framing of @xmath404 ; - the vector fields @xmath438 belong to @xmath324 and provide its given trivialization , while @xmath439 provide the given trivialization of @xmath325 ; - in these local coordinates , @xmath3 is given by @xmath440 where @xmath441 and @xmath442 coincides with @xmath443 , where the signs in the above formulas coincide with the sign of the boundary component @xmath232 of the framed membrane @xmath412 . while the use of normal form [ lemma - replacing - eq:14 ] is convenient but it is not necessary .
indeed , it is obvious that all the stated properties can be achieved by a @xmath444-small perturbation of our data near @xmath232 , and this will suffice for our purposes .
a _ suspension _ of a folded map @xmath205 is a surjective homomorphism @xmath445 such that @xmath446 , where @xmath447 is the projection @xmath448 and @xmath449 is the inclusion .
the main reason for considering enrichments of folds is that an enriched folded map admits a suspension whose homotopy class depends only on the enrichment .
[ prop : canonical - enrichment ] to an enriched folded map @xmath418 we can associate a homotopically well defined suspension @xmath450 .
the suspension @xmath451 will be of the form @xmath452 where @xmath453 is a 1-form , @xmath2 is a section of @xmath454 , and @xmath455 is a function . the 1-form @xmath456 is defined as @xmath457 near @xmath458 , using the local coordinate @xmath288 on @xmath237 . to extend it to a 1-form on all of @xmath1 , we will extend the function @xmath288 .
we first construct convenient local coordinates near @xmath404 .
the map @xmath17 restricts to a local diffeomorphism on @xmath459 . the local coordinate functions @xmath460 near @xmath399 from the normal form [ lemma - replacing - eq:14 ] extend to @xmath461 in such a way that the vector fields @xmath462 along @xmath404 generate the bundle @xmath324 , while @xmath463 along @xmath404 generate the bundle @xmath325 .
the tubular neighborhood theorem then gives a neighborhood @xmath464 with coordinate functions @xmath465 so that the fibers of @xmath466 are the fibers of @xmath3 , or , more precisely , @xmath467 .
( the function @xmath466 and the function @xmath468 of the lemma [ lemma - replacing - eq:14 ] are not directly related ; in fact they have codomains of different dimension . ) on this overlap , the function @xmath288 of lemma [ lemma - replacing - eq:14 ] can be expressed as a function of the local coordinates .
indeed , we have @xmath469 , so @xmath470 so @xmath471 if we taylor expand the square root , we get @xmath472 for positive functions @xmath473 .
the function @xmath474 extends over @xmath475 ( use the same formula in the local coordinates of the tubular neighborhood of @xmath397 ) , and hence we can also extend @xmath288 to a neighborhood of @xmath404 inside @xmath476 , such that on @xmath475 is satisfies ( [ eq:15 ] ) .
extend @xmath288 to all of @xmath1 in any way , and let @xmath457 .
we have defined a bundle map @xmath477 which is surjective whenever @xmath478 . near @xmath404 , @xmath479 precisely when @xmath480 .
it remains to define the section @xmath481 of @xmath482 .
pick a function @xmath483 $ ] such that @xmath484 near @xmath458 , is negative on @xmath397 and equal to @xmath485 on @xmath486 , and which is equal to @xmath487 outside a small neighborhood of @xmath404 . then set @xmath488 [ rem : neg - stab ]
changing the sign of @xmath489 in the formula provides another suspension of the enriched folded map @xmath418 which we will denote by @xmath490 .
if @xmath143 is even then the two suspensions @xmath491 and @xmath490 are homotopic .
most ( but not all ) of the data of an enrichment @xmath492 of a folded map @xmath17 can be reconstructed from the suspension @xmath493 .
if we write @xmath494 in the matrix form , the manifold @xmath404 is the set of points with @xmath495 and @xmath496 not surjective .
the partition of @xmath497 into @xmath498 is determined by the coorientation of images of the fold components . on the other hand ,
the splitting @xmath499 can not be reconstructed from the suspension .
[ lm : special - enrichment ] let @xmath205 be a special folded map .
then the suspension @xmath450 ( as well as the suspension @xmath500 ) of the canonically enriched folded map @xmath501 is homotopic to its stabilized regularized differential @xmath502 .
this can be seen in the local models . as in the previous section ,
we study a folded map @xmath17 with cooriented fold images .
let @xmath503 be a connected component of index @xmath73 , and let @xmath504 be its image . for @xmath505 , the fibers @xmath506 has a singularity .
there are two directions in which we can move @xmath256 away from @xmath235 to resolve the singularity and get a manifold .
the manifold we get by moving @xmath256 to the positive side ( with respect to the coorientation ) differs from the manifold we get by moving @xmath256 to the negative side by a surgery of index @xmath73 , i.e. it has an embedded @xmath507 instead of a @xmath508 .
if @xmath235 bounds an embedded domain @xmath509 , then one can try to prevent the surgery from happening , or , which is the same , to perform an inverse morse surgery fiberwise in each fiber @xmath510 . in this section
we describe this process , which we call _ fold eliminating surgery _ in more detail .
we begin with a local model for the surgery .
let @xmath511 be an @xmath143-dimensional oriented manifold with collared boundary @xmath512 .
the collar consists of an embedding @xmath513 \to p$ ] , mapping @xmath514 .
extend @xmath511 by gluing a bicollar @xmath515 $ ] , @xmath516 } u.\end{aligned}\ ] ] let @xmath517 be a quadratic form of index @xmath73 on @xmath518 : @xmath519 where @xmath520 .
let @xmath521 be the domain @xmath522 we are going to use the map @xmath523 $ ] as a prototype of a fold .
the boundary of the ( possibly singular ) fiber @xmath524 can be identified with @xmath525 via the diffeomorphism @xmath526 the map @xmath527 $ ] is a folded map which has @xmath528 as its fold of index @xmath529 with respect to the coorientation of the fold defined by the second coordinate of the product @xmath530 $ ] .
, height=151 ] given a smooth function @xmath531 $ ] we define @xmath532 we then have @xmath533 . given a one - parameter family of functions @xmath534 $ ] , @xmath535 $ ] , we denote @xmath536\,|\ , q ( x ) = \varphi_t(p)\}\,,\cr z^{\varphi_t } & = \{(p , x , t)\in { { \partial}}{\widetilde}p\times h \times[0,1]\,|\ , q ( x ) = \varphi_t(p)\}\,,\cr r^{\varphi_t } & = p^{\varphi_t}\cap \{||x_+||=2\}.\cr \end{split}\ ] ] we have @xmath537 .
we will consider the projection @xmath538
\to \tilde p \times [ 0,1]\end{aligned}\ ] ] and especially its restriction to the subsets . using ,
the set @xmath539 can be identified with @xmath540 ) \times s^{k-1 } \times s^{d - k-1}$ ] via a diffeomorphism over @xmath541 $ ] . in particular
we get a diffeomorphism @xmath542 \to r^{\varphi_t},\end{aligned}\ ] ] which scales the @xmath543 coordinates .
in fact it can be seen to be given by the formula @xmath544 although we shall not need this explicit formula .
the restriction @xmath545 $ ] is our `` template cobordism '' , and we record its properties in a lemma .
[ lm : model1 ] \a ) suppose that @xmath242 is not a critical value of @xmath546 .
then @xmath547 is a smooth manifold of dimension @xmath548 , and the projection @xmath549 is a folded map with the fold @xmath550 , which projects to @xmath551 . in particular , the map @xmath552 is non - singular if @xmath546 does not take the value @xmath242 . the fold @xmath232 has index @xmath73 with respect to the coorientation of @xmath235 by an outward normal vector field to the domain @xmath553 .
\b ) let @xmath534 $ ] , @xmath535 $ ] , be a one - parameter family of functions such that @xmath242 is not a critical value of @xmath554 and @xmath555 and of the function @xmath556\to[-1,1 ] $ ] defined by @xmath557 , @xmath558 $ ] .
we also assume @xmath559 is independent of @xmath560 for @xmath256 near @xmath512
. then @xmath561 $ ] is a folded cobordism between the folded maps @xmath562 and @xmath563 .
we have @xmath564 $ ] , so together with we get a diffeomorphism @xmath565 \to z^{\phi_t } \cup r^{\phi_t}. \end{aligned}\ ] ] we will need to apply the above lemma to two particular functions on @xmath566 . recall that @xmath567 \subset \tilde p$ ] denotes the bicollar . take @xmath568 and pick a function @xmath555 with the following properties : * @xmath569 near @xmath570 , * @xmath571 on @xmath572 , * for @xmath573 , @xmath574 is a non - decreasing function of @xmath575 , * @xmath576 for @xmath577
. we will write @xmath578 and @xmath579 for @xmath580 and @xmath581 , and denote by @xmath582 and @xmath583 the respective projections @xmath584 and @xmath585 .
similarly , we will use the notation @xmath586 and @xmath587 instead of @xmath588 and @xmath589 .
the map @xmath584 is non - singular , while the map @xmath585 has a fold singularity with image @xmath590 .
the index of this fold with respect of the outward coorientation to the boundary of @xmath511 is equal to @xmath73 .
taking linear interpolations between @xmath554 and @xmath555 in one order or the other , we get folded cobordisms in two directions between the map @xmath584 and @xmath591 .
we will denote the corresponding cobordisms by @xmath592 and @xmath593 , respectively .
the projections @xmath594 $ ] and @xmath595 $ ] are folded bordisms in two directions between @xmath596 and @xmath597 .
we think of @xmath598 as a one - parameter family of ( possibly singular ) manifolds , interpolating between @xmath599 and @xmath600 . using the trivialization
, these manifolds all have the same boundary , so @xmath601 and @xmath602 may be used as local models for cobordisms .
they allow us to create , or annihilate a fold component , respectively .
we describe the fold eliminating surgery more formally in the next section and leave the formal description of the inverse process of fold creating surgery to the reader .
in fact fold creating surgery will not be needed for the proof of the main theorem
. in the context of enriched folded maps there are two versions of fold eliminating surgery .
one will be referred to as _
membrane eliminating_. in this case the membrane will be eliminated together with the fold .
the second one will be referred to as _ membrane expanding .
_ in that case the membrane after the surgery will be spread over @xmath603 , the image of @xmath511 in the target . for the membrane eliminating case
we choose the submanifold @xmath604\ ] ] as a template membrane for the folded bordism @xmath605 .
next we choose the subbundles @xmath324 and @xmath325 spanned by vector fields @xmath606 and @xmath607 , respectively , as a template framing .
note that with this choice we have @xmath608 and the index of the membrane @xmath609 is equal to @xmath610 . for the membrane expanding surgery
we choose as a template membrane the submanifold @xmath611\ ] ] with boundary @xmath612 as the membrane for the folded bordism @xmath595 $ ] .
we choose the subbundles @xmath324 and @xmath325 spanned by vector fields @xmath613 and @xmath614 , respectively , as a template framing .
note that with this choice we have @xmath615 and the index of the membrane @xmath616 is equal to @xmath73 .
note that in the membrane eliminating case the restriction of the membrane @xmath609 to @xmath579 projects diffeomorphically onto @xmath617 , while in the membrane expanding case the restriction of the membrane @xmath616 to @xmath579 projects diffeomorphically onto @xmath618 .
_ membrane eliminating surgery_. let @xmath619 be an enriched folded map and @xmath412 one of its membranes .
suppose that the framed membrane @xmath412 is pure and assume first that @xmath413 , and that the index of the membrane is equal to @xmath620 .
note that in this case the boundary fold @xmath399 has index @xmath73 with respect to the outward coorientation of @xmath409 .
consider the model enriched folded map @xmath621 where @xmath511 is diffeomorphic to @xmath404 .
fix a diffeomorphism @xmath622 .
let @xmath623 denote a neighborhood of @xmath624 . according to lemma [ lemma - replacing - eq:14 ]
there exist an extension @xmath625 of the embedding @xmath626 and an embedding @xmath627 such that * the diagram @xmath628^{\psi}\ar[d]^{\pi^1}&m\ar[d]^f\\ u \ar[r]^{{\widetilde}\psi } & x}\ ] ] commutes ; * @xmath629 ; * the canonical framing of the membrane @xmath630 is sent by @xmath631 to the given framing of the membrane @xmath404 . the data needed for eliminating the fold @xmath458 by surgery consists of an extension of @xmath631 to all of @xmath600 such that * the diagram @xmath632^{\psi}\ar[d]^{\pi^1}&m\ar[d]^f\\ { \widetilde}p \ar[r]^{{\widetilde}\psi } & x}\ ] ] commutes ; * @xmath633 ; * the canonical framing of the membrane @xmath634 is sent by @xmath631 to the given framing of the membrane @xmath404 . [
constr : membrane - elim ] fold eliminating surgery consists of replacing @xmath579 by @xmath635 with the projection @xmath582 .
more precisely , cut out @xmath636 $ ] from @xmath637 $ ] and glue in @xmath638 along the identification .
this gives an enriched folded map @xmath639 $ ] which is a cobordism starting at @xmath3 , and ending in an enriched folded map where the fold @xmath458 , _ together with its membrane _
@xmath404 , has been removed .
, height=151 ] , height=264 ] the case @xmath414 can be reduced to the previous one by the following procedure .
let @xmath640 be the dual framing of @xmath404 , see section [ sec : enriched ] above .
then @xmath641 , @xmath642 , @xmath643 , and the membrane @xmath416 has index @xmath644 .
hence , we can use for the membrane eliminating surgery the above template of index @xmath391 , and then switch the framing of the constructed membrane in the cobordism to the dual one .
let @xmath645 be a domain in @xmath2 with boundary bounded by image @xmath235 of a fold @xmath232 of index @xmath73 with respect to the outward orientation of @xmath646 .
suppose that the membrane @xmath404 adjacent to @xmath232 projects to the complement of @xmath511 in @xmath2 , i.e. @xmath647 , and @xmath410 .
the case @xmath648 can be reduced to the positive by passing to the dual framing as it was explained above in the membrane eliminating case .
consider the template enriched folded map @xmath621 as in section [ sec : surgery - template ] .
let @xmath626 denote the inclusion @xmath649 . according to lemma [ lemma - replacing - eq:14 ]
, there exist an extension @xmath625 of the embedding @xmath626 and an embedding @xmath627 such that * the diagram @xmath628^{\psi}\ar[d]^{\pi^1}&m\ar[d]^f\\ u \ar[r]^{{\widetilde}\psi } & x}\ ] ] commutes ; * @xmath650 ; * the canonical framing of the membrane @xmath651 is sent by @xmath631 to the given framing of the membrane @xmath404 . in this case
the data needed for eliminating the fold @xmath458 by surgery consists of an extension of @xmath631 to all of @xmath600 such that * the diagram @xmath652^{\psi}\ar[d]^{\pi^1}&m\ar[d]^f\\ { \widetilde}p\ar[r]^{{\widetilde}\psi } & x}\ ] ] commutes ; * the canonical framing of the membrane @xmath653 is sent by @xmath631 to the given framing of the membrane @xmath404 .
[ constr : membrane - exp ] fold eliminating surgery consists of replacing @xmath579 by @xmath578 with the projection @xmath654 .
exactly as in construction [ constr : membrane - elim ] we get an enriched folded map @xmath639 $ ] which is a cobordism starting at @xmath3 , and ending in an enriched folded map where the fold @xmath458 has been removed .
both constructions eliminate the fold @xmath399 .
the difference between them is that in the membrane expanding case , the above surgery spreads the membrane @xmath404 over the domain @xmath511 .
the embedding @xmath655 required for the surgeries in constructions [ constr : membrane - elim ] and [ constr : membrane - exp ] is determined up to isotopy by slightly simpler data . to any smooth manifold @xmath511 with collared boundary ,
let @xmath656 $ ] be the function defined in section [ sec : surgery - template ] and let @xmath657 this is a closed manifold , which up to diffeomorphism depends only on @xmath511 .
in fact , it is diffeomorphic to the boundary of @xmath658 ( after smoothing the corners of @xmath658 ) .
the projection @xmath659 restricts to a folded map @xmath660 with fold @xmath512 of index @xmath73 with respect to the outward orientation of the boundary of @xmath511 .
we have an embedding @xmath661 given by @xmath662 . the normal bundle of this embedding has a canonical frame given by projections of the frame @xmath663 to @xmath664 .
we also have an embedding @xmath665 as @xmath666 , which identifies @xmath512 with the folds of the projection @xmath667 , and the normal bundle of @xmath668 is framed by @xmath669 let us denote @xmath670 .
thus we have @xmath671 let @xmath672 be an enriched folded map .
\a ) * membrane eliminating case . *
let @xmath673 be a framed membrane with @xmath413 .
basis _ for membrane eliminating surgery consists of a pair @xmath674 , where @xmath511 is a compact @xmath143-manifold with boundary , @xmath675 is an embedding , and @xmath676 is a framing of the normal bundle of @xmath221 , such that the following conditions are satisfied . * @xmath677 ; * the map @xmath678 factors through an embedding @xmath679 , i.e. @xmath680 , and hence @xmath681 ; * the vectors
@xmath682 belong to @xmath683 and along @xmath684 coincide with the given framing of the bundle @xmath685 . *
the vectors @xmath686 defines the prescribed framing of the bundle @xmath687 ; * @xmath688 is disjoint from membranes of @xmath492 , other than @xmath404 . , where @xmath689 , @xmath690,height=113 ] \b ) * membrane expanding case .
* let @xmath645 be a domain in @xmath2 bounded by folds of index @xmath73 with respect to the outward orientation of @xmath646 .
let @xmath232 be the union of the corresponding fold components , and @xmath412 the union of framed membranes adjacent to @xmath232 .
suppose that @xmath410 and @xmath691 .
basis _ for membrane expanding surgery consists of a pair @xmath674 , where @xmath692 is an embedding , and @xmath676 is a framing of the normal bundle of @xmath221 , such that the following conditions are satisfied . *
the map @xmath678 factors through an embedding @xmath693 , i.e. @xmath680 ; * @xmath694 and coincides with the given framing of the bundle @xmath695 over the fold @xmath696 ; * @xmath697 and @xmath688 is disjoint from any other membranes different from @xmath404 . * the vector field @xmath698 is tangent to @xmath404 and inward transversal to @xmath399 .
note that in both cases it follows from the above definitions that @xmath699 is a folded map with the definite fold @xmath700 . given a basis @xmath701 , one can extend , uniquely up to isotopy the embedding @xmath221 to an embedding @xmath702 such that the diagram or commutes .
this , in turn , enables us to perform a membrane eliminating or membrane expanding surgery .
[ rem : fold - creating]_fold creating surgeries .
_ fold creating surgeries are inverse to fold eliminating surgeries . for our purposes
we will need only one such surgery which creates a fold of index 1 with respect to the _ inward _ co - orientation of its membrane .
a basis of such a surgery is given by a pair @xmath701 , where @xmath221 is an embedding @xmath703 over a domain @xmath704 disjoint from the folds of the map @xmath3 , and @xmath705 is a framing of the vertical bundle @xmath706
. see figure [ mw17 ] .
let us review fold eliminating surgeries in the case @xmath224 .
these surgeries can be of index @xmath242 , @xmath244 , @xmath707 or @xmath226 .
let @xmath232 be a union of fold components whose projections bound a domain @xmath704 . in the membrane eliminating case , @xmath511 is the projection @xmath708 of the membrane which spans @xmath232 . in the membrane expanding case
the membranes adjacent to @xmath232 projects to the complement of the domain @xmath511 .
all fold indices below are with respect to the outward coorientation of the boundary of the domain @xmath645 .
_ index 0 .
_ we have @xmath709 , i.e. the basis of the surgery in this case is a framed embedding @xmath710 which sends @xmath711 to the fold @xmath232 .
only the membrane expanding surgery is possible in this case .
when a point @xmath712 approaches @xmath711 from outside , a spherical components of the fiber @xmath506 dies .
the surgery prevents it from dying end prolongs its existence over all points of @xmath511 . _ index 1 . _
the surgery basis in this case consists of 2 sections @xmath713 , together with framings of the bundle @xmath714 over them . as @xmath715 the sections @xmath716 converge to the same point @xmath717 , @xmath718 . in the membrane eliminating case , one of these sections is the membrane @xmath404 .
the manifold @xmath719 is obtained by a fiberwise index 1 surgery ( i.e. the connected sum ) along the framed points @xmath720 and @xmath721 , @xmath722 .
this eliminates the fold @xmath232 together with the membrane @xmath404 in the membrane eliminating case , and spreads the membrane over @xmath511 in the membrane expanding case . in the latter case
the newly created membrane is a section over @xmath511 which takes values in the circle bundle over @xmath511 formed by central circles of added cylinders @xmath723 $ ] .
_ index 2 .
_ the surgery basis in this case is a circle - subbundle over the domain @xmath724 , i.e. a family of circles in fibers @xmath725 @xmath726 , which collapse to points in @xmath232 when @xmath256 converges to a boundary point of @xmath511 . in the fold eliminating case , the membrane @xmath404 is a section over @xmath511 of this circle bundle .
the surgery consists of fiberwise index 2 surgery of fibers along these circles , which eliminates the fold @xmath232 together with its membrane in the eliminating case , and spreads it over @xmath511 in the expanding one .
_ index 3 .
_ the basis of the surgery in this case is a connected component of @xmath1 which forms an @xmath727-bundle over @xmath511 .
the 2-spheres collapse to points of @xmath232 when approaching the boundary of @xmath511 .
the surgery eliminates this whole connected component , in particular removing the fold and its membrane .
the membrane expanding case is not possible for @xmath728 . so far
( in theorem [ thm : special - folded ] and lemma [ lm : special - enrichment ] ) we have related _ unstable _ formal fibrations to ( enriched ) folded maps . in theorem
[ thm : main1 ] we need to work with _
stable _ formal fibrations ( because that is what @xmath729 classifies ) . in this section
we study the question of whether an epimorphism @xmath730 can be `` destabilized '' , i.e.homotoped to be of the form @xmath731 for some unstable epimorphism @xmath732 .
this is not possible in general of course ( the obstruction is an euler class ) .
instead we prove the following .
[ prop : destabilization ] let @xmath733 be a bundle epimorphism with underlying map @xmath17 .
assume @xmath1 and @xmath2 are connected .
then there is a compact codimension 0 submanifold @xmath734 which is homotopy equivalent to a simplicial complex of dimension at most 1 , such that the following hold , after changing @xmath3 and @xmath494 by a homotopy ( in the class of bundle epimorphisms ) .
a. @xmath735 is folded and has an enrichment @xmath419 such that * @xmath736 if @xmath737 ; * @xmath738 or @xmath739 in the case @xmath357 ; b. @xmath494 is integrable near @xmath740 , i.e. it equals @xmath741 there . c. @xmath494 destabilizes outside @xmath9 , i.e. it equals @xmath742 there , for some bundle epimorphism @xmath743 .
the strategy of the proof of the proposition is as follows .
first we forget about ( iii ) in the proposition , and only worry about how @xmath19 and @xmath494 looks like on @xmath9 .
we prove that this can be done for a large class of possible @xmath9 s .
after that , we consider the obstruction to destabilizing @xmath494 outside @xmath9 without changing it on @xmath9 .
this obstruction is essentially an integer , and we prove that @xmath9 can be chosen so that the obstruction vanishes .
we first give a local model for the enriched folded map @xmath744 .
let @xmath745 $ ] be the interval , and let @xmath746 be a simplicial complex .
we will consider @xmath747 as the subset @xmath748 .
let @xmath749 be a regular neighborhood of @xmath750 , and let @xmath751 be a regular neighborhood of @xmath752 .
let @xmath753 be the projection .
in order to avoid confusion we will write @xmath754 , and hence @xmath755 .
we can assume that @xmath756 is a folded map , with fold @xmath757 which has index 0 with respect to the inward coorientation of @xmath758 .
let @xmath759 $ ] be a bicollar of @xmath320 , i.e. an embedding @xmath760 , which maps @xmath761 and @xmath762 $ ] to @xmath311 .
let @xmath763 .
we construct a folded map @xmath764 in the following way . first pick a function @xmath765 \to [ -1,1 ] \times { \mathbb { r}}$ ] with the following properties .
a. @xmath766 for @xmath767 ; b. @xmath768 for @xmath769 , @xmath770 for @xmath771 , and @xmath772 ; c. @xmath773 .
in particular @xmath19 is an immersion of codimension 1 , and @xmath774 $ ] is a folded map with fold @xmath775 .
extend to an immersion @xmath776 \times { \mathbb { r}}\to [ -1,1]\times { \mathbb { r}}$ ] with the property that @xmath777 for @xmath767 . recall
that @xmath778 $ ] and construct a codimension 0 immersion @xmath779 by @xmath780 for @xmath781 .
extend to a codimension 0 immersion @xmath782 by @xmath783 for @xmath784 .
for @xmath785 , let @xmath786 be the first coordinate of @xmath787 .
differentiating @xmath788 then gives a bundle map @xmath789 with underlying map @xmath790 .
we record some of its properties in a lemma .
[ lem : local - model ] a. the map @xmath788 is homotopic in the class of submersions ( = immersions ) to the map @xmath791 .
b. let @xmath792 and @xmath297 and @xmath793 be as above .
then @xmath297 is a folded map .
the fold is @xmath794 , and the image of the fold is also @xmath795 . near @xmath740 ,
the bundle map @xmath793 is _ integrable _ , i.e. @xmath796 .
c. let @xmath797 be the projection and define a bundle map @xmath798 by @xmath799 .
it covers the folded map @xmath800 , which has fold @xmath801 .
the image of the fold is @xmath802 , and it has index 0 with respect to the inward coorientation of @xmath803 .
d. [ item:1 ] a membrane for the underlying map @xmath221 can be defined as @xmath804 with the framing @xmath805 , and @xmath806 .
this defines an enrichment @xmath419 for @xmath221 .
finally , @xmath807 is integrable over @xmath740 and @xmath808 .
we leave this as an easy exercise for the reader .
cf also section [ sec : basis - surgery ] and the proof of proposition [ prop : canonical - enrichment ] .
let us also point that we could equally well have based the construction on a map @xmath809
\to [ -1,1]\times { \mathbb { r}}$ ] defined as @xmath19 above , except that we replace the condition @xmath773 by @xmath810 . using this map as a basis for the construction gives a bundle epimorphism @xmath811 which also satisfies the conclusion of lemma [ lem : local - model ] , except that ( [ item:1 ] ) gets replaced by @xmath812 .
let @xmath733 be as in the proposition .
by phillips theorem we can assume that @xmath494 is induced by a submersion @xmath813 pick cubes @xmath814 and @xmath815 .
we regard @xmath816 and let @xmath817 denote the projection to the first @xmath143 coordinates .
we can assume that @xmath818 .
the space of submersions @xmath819 is , by phillips theorem , homotopy equivalent to @xmath820 which is connected , and hence we may assume that @xmath821 .
( here we used @xmath822 . in the case @xmath823 , we get @xmath824 which has two path components , but after possibly permuting coordinates on @xmath825 , we may assume that @xmath826 is in the same path component as @xmath827 . )
hence , after a homotopy of @xmath631 in the class of submersions , we may assume that @xmath828 . by lemma [ lem : local - model](i )
, we can assume after a further deformation of @xmath631 in a neighborhood of @xmath829 , that it agrees with the map @xmath830 .
this proves ( i ) and ( ii ) in the proposition .
remember that the domain @xmath9 is a regular neighborhood of a simplicial complex @xmath750 .
the local model in lemma [ lem : local - model ] worked for any such @xmath376 , but in the proof of ( i ) and ( ii ) in the proposition we used that @xmath376 had dimension at most 1 . on the other hand , @xmath376 could still be arbitrary within that restriction .
it remains to prove that @xmath494 can be destabilized outside @xmath9 , for a suitable choice of @xmath376 .
there is an obstruction to doing this , which we now describe .
let @xmath81 denote a section of @xmath831 such that @xmath832 is the constant section @xmath833 .
this defines @xmath81 uniquely up to homotopy ( in fact @xmath81 is unique up to translation by vectors in the kernel of @xmath494 ) . over @xmath9 the epimorphism @xmath494
is induced by a composition @xmath834 and on @xmath9 we may choose @xmath81 so that @xmath835 takes @xmath81 to a unit vector in the @xmath836-direction .
another relevant section is the constant section @xmath837 .
we have @xmath838 for @xmath839 .
our aim is to change @xmath494 by a homotopy and achieve @xmath838 for all @xmath245 outside @xmath9 . in each fiber , @xmath840 , so by induction of cells of @xmath841 , we can assume that @xmath842 outside @xmath843 , since @xmath844 can be built using cells of dimension at most @xmath845 .
it remains to consider @xmath846 .
let @xmath847 be the section which agrees with @xmath81 on @xmath9 and with @xmath848 outside @xmath9 .
thus @xmath81 and @xmath847 are both sections of @xmath849 which equal @xmath848 outside @xmath843 .
we study their homotopy classes in the space of such sections . using stereographic projection ,
the fiber of @xmath850 at a point @xmath851 can be identified with the one - point compactification of @xmath852 .
hence we can think of sections as continuous vector fields on @xmath1 , which are allowed to be infinite .
the section at infinity is @xmath853 . in this picture
we have the following way of thinking of @xmath847 : for @xmath854 , @xmath855 is a unit vector orthogonal to @xmath795 pointing outwards . moving @xmath245 away from @xmath795 to the _ inside
_ makes @xmath855 smaller and it gets zero as we get far away from @xmath795 . moving @xmath245 away from @xmath795 to the _ outside _ makes @xmath855 larger and it gets infinite as we get far away from @xmath795 .
the section @xmath847 depends up to homotopy only on the simplicial complex @xmath750 , hence the notation .
there is a bijection between @xmath856 and sections of @xmath857 which agree with @xmath848 outside @xmath843 .
the bijection takes @xmath847 to @xmath858 .
the tangent bundle @xmath859 is trivial over @xmath843 , so the space of such sections is just the space of pointed maps @xmath860 and homotopy classes of such are classified by their degree , which is an integer .
we have assumed @xmath861 so using the standard embedding @xmath862 we can work entirely inside @xmath863 .
the geometric interpretation of @xmath847 given above can then be rephrased even more conveniently .
let @xmath864 be the retraction in the tubular neighborhood , and let @xmath865 for @xmath866 .
pick any continuous extension of @xmath867 to @xmath843 with the property that when @xmath868 , we have @xmath869 up to homotopy there is a unique such extension because we are picking a point @xmath870 in a contractible space . then @xmath871 . to calculate the degree of the corresponding map we perturb even further .
remember that any simplicial complex @xmath376 has a standard vector field with the following property : the stationary points are the barycenters of simplices and the flowline starting at a point @xmath245 converges to the barycenter of the open cell containing @xmath245 .
let @xmath872 be the time @xmath220 flow of this vector field , and define a vector field @xmath873 just like @xmath874 , except that we replace the right hand side of by @xmath875 for some small @xmath876 .
the resulting vector field vanishes precisely at the barycenters of @xmath376 , and the index of the vector field at the barycenter of a simplex @xmath877 is @xmath878 .
the claim now follows from the poincar - hopf theorem .
let us first consider the case @xmath879 .
we have proved that all possible sections of @xmath849 which agree with @xmath848 outside @xmath843 , are homotopic to @xmath847 for some @xmath376 .
then we can choose @xmath376 such that @xmath880 . since @xmath847 and @xmath81 agree on @xmath9
there is a homotopy of @xmath81 , fixed over @xmath9 , so that @xmath838 for all @xmath881 .
this homotopy lifts to a homotopy of bundle epimorphisms @xmath733 , and then ( iii ) is satisfied . for @xmath357 we may not be able to choose a @xmath882 with the required euler characteristic , since subcomplexes of 1-manifolds always have non - negative euler characteristic .
however , vector fields of negative index can be achieved as @xmath883 , and that is the vector field that arises if we use the negative suspension @xmath884 .
the following theorem summarizes the results of section [ sec : prelim ] .
[ thm : epi - folds ] let @xmath733 be a bundle epimorphism .
suppose that @xmath223 and @xmath494 is integrable in a neighborhood of a closed set @xmath885 .
then there is a homotopy of epimorphisms @xmath886 , @xmath887 $ ] , fixed near @xmath323 , which covers a homotopy @xmath888 , such that @xmath889 is folded , and @xmath890 for some enrichment @xmath419 of @xmath891 . if @xmath892 then the image @xmath504 of each fold component @xmath380 of @xmath891 bounds a domain in @xmath2 . first use proposition [ prop : destabilization ] to make @xmath19 enriched folded over a domain @xmath9 , such that @xmath494 destabilizes outside @xmath9 . then use lemma [ lm : special - enrichment ] to make @xmath893 special enriched outside @xmath9 .
let us rephrase the results of the previous section more systematically , and put them in the context of the overall goal of the paper .
so far we have mainly studied the relation between formal fibrations and enriched folded maps .
let us formalize the result .
we consider various bordism categories of maps @xmath17 such that @xmath1 and @xmath2 are both oriented and @xmath2 is a compact manifold , possibly with boundary , and which satisfy the following two conditions : c1 .
@xmath3 has @xmath204 ends , i.e. there is ( as part of the structure ) a germ at infinity of a diffeomorphism @xmath894 such that @xmath895 , where @xmath487 is a germ at infinity of the projection @xmath896 .
this trivialized end will be called the _ standard end _ of @xmath1 .
there is a neighborhood @xmath311 of @xmath35 such that @xmath897 is a fibration ( i.e. smooth fiber bundle ) with fiber @xmath204 .
when @xmath898 we will always assume that all fold components are homologically trivial , and in particular , the image @xmath381 of any fold component @xmath899 bounds a domain in @xmath2 .
note that this condition is preserved by all fold surgeries which we discussed above in section [ sec : surgery - general ] .
we have been studying the following bordism categories of maps .
a. @xmath900 is the category of _ fibrations _ ( smooth fiber bundles ) with fiber @xmath204 , which satisfy c1 and c2 .
b. @xmath901 is the category of enriched folded maps , satisfying c1 and c2 . c. @xmath902 is the category of _ formal fibrations _
, i.e. bundle epimorphisms @xmath733 with underlying map @xmath17 , such that @xmath3 satisfies c1 and c2 , and such that @xmath903 near @xmath904 .
are the objects in bordism categories , bordisms in the categories @xmath900 , @xmath901 , and @xmath902 are required to be trivial ( as bordisms , and not as fibrations ! ) over a neighborhood of @xmath35 .
we have functors @xmath905 the functor @xmath906 is the obvious inclusion .
everything we said in chapter [ sec : prelim ] works just as well with the conditions c1 and c2 imposed , and hence proposition [ prop : canonical - enrichment ] gives the functor @xmath907 . in this setup
, the main goal of the paper is to prove that any object in @xmath902 is cobordant to one in @xmath900 .
theorem [ thm : epi - folds ] , which also works for manifolds with boundary , says that @xmath908 is essentially surjective : any object in @xmath902 is cobordant to an element in the image of @xmath908 .
it remains to see that any object of @xmath901 is cobordant to one in @xmath900 .
in fact theorem [ thm : epi - folds ] is a little stronger : any object of @xmath902 is cobordant to one in the image of @xmath908 using only homotopies of the underlying maps , i.e. no cobordisms of @xmath1 and @xmath2 . in contrast , comparing @xmath900 to @xmath901 involves changing @xmath1 and @xmath2 by _ surgery_. the surgery uses the membranes in the enrichment , and also makes crucial use of a form of harer s stability theorem .
in fact , it is convenient to work with a slight modification of the category @xmath901 .
let @xmath909 be the category with the same objects as @xmath901 , but where we formally add morphisms which create double folds along a submanifold @xmath232 , i.e. singularities of the form ( [ eq : double - fold2 ] ) .
we also formally add inverses of these morphisms .
when @xmath892 we additionally require the manifold @xmath232 to be homologically trivial .
by example [ ex : double - fold ] , this has a canonical enrichment , and we formally add morphism in @xmath909 in both directions between the map @xmath17 and the same map @xmath910 with a double @xmath232-fold singularity created .
according to [ lm : special - enrichment ] , adding a double @xmath232-fold together with its canonical enrichment changes @xmath450 only by a homotopy .
hence the functor @xmath911 extends to a functor @xmath912 . in the proof of our main theorem ,
@xmath901 is just a middle step , and it turns out to be more convenient to work with the modified category @xmath909 .
we leave it as an exercise to the reader to show that when @xmath892 there is , in fact , no real difference : if two objects in @xmath909 are cobordant , then they are already cobordant in @xmath901 .
we shall not need this fact .
we prove that any object of @xmath909 is cobordant to one in @xmath900 in two steps .
let @xmath913 be the subcategory where objects and morphisms are required to satisfy the following conditions .
a. folds are hyperbolic , i.e. have no folds of index 0 and 3 .
b. for all @xmath914 , the manifold @xmath915 , i.e. the fiber minus its singularities , is connected .
let @xmath916 be the category with the same objects , but where we formally add morphisms ( in both directions ) which create double folds .
our main result , theorem [ thm : main1 ] , follows from theorem [ thm : epi - folds ] and the following two propositions .
[ prop : enriched - hyperbolic ] let @xmath223 .
any enriched folded map @xmath917 is bordant in the category @xmath918 to an element in @xmath919 [ prop : hyperbolic - fibrations ] let @xmath920 .
any enriched folded map @xmath921 is bordant in the category @xmath922 to a fibration from @xmath923 .
the proof of the latter uses harer stability .
this is the only part of the whole story in which @xmath224 is used in an essential way .
the reason for the condition @xmath223 in the first proposition is explained in appendix a.
the main ingredient in the proof of [ prop : hyperbolic - fibrations ] is harer s stability theorem [ thm : harer - geometric ] . in this section
we will deduce a version of [ thm : harer - geometric ] for enriched folded maps .
the proof of the following main theorem of this section will be given in section [ sec : proof - harer ] .
section [ sec : nodal ] contains necessary preliminary constructions .
[ thm : harer2 ] let @xmath924 be an enriched folded map .
let @xmath925 be a closed domain with smooth boundary transversal to the images of the folds and @xmath195 be compact surfaces with boundary .
let @xmath926 be a fiberwise embedding over @xmath320 whose image does not intersect any fold or membrane and such that the complement of its image in each fiber is connected , even after removing the folds of @xmath3 .
then , after possibly changing @xmath418 by a bordism which is constant outside @xmath927 , the embedding @xmath20 extends to an embedding of @xmath928 .
the following corollary of theorem [ thm : harer2 ] will be the key ingredient in the proof of proposition [ prop : hyperbolic - fibrations ] .
[ cor : basis - existence ] let @xmath929 be an enriched folded map .
let @xmath930 be fold components of @xmath3 and @xmath931 their image .
assume that the @xmath932 , @xmath933 , are disjoint and their union bounds together a domain @xmath934 , all folds @xmath935 have the same index with respect to the outward co - orientation of @xmath512 and that one of the following conditions holds : m1 .
there exists a pure framed membrane @xmath404 which spans @xmath936 and projects to @xmath511 ; m2 .
all framed membranes adjacent to @xmath430 project to the complement of @xmath511 and the folds @xmath937 are either all positive or all negative ( recall that the boundary of a membrane is split into a positive and negative part ) .
then @xmath418 is bordant in the category @xmath922 to an element @xmath938 such that * the bordism is constant over the complement of @xmath939 ; * @xmath938 has no more membranes than @xmath418 ; * @xmath319 admits a basis for a surgery eliminating the fold @xmath232 .
the surgery eliminates the membrane of @xmath232 in case m1 and spreads it over @xmath511 in case m2 .
consider a slightly smaller domain @xmath940 , so that @xmath941 is an interior collar of @xmath711 in @xmath511 .
if the index of @xmath232 with respect to the outward coorientation of @xmath235 is @xmath244 then the @xmath242-dimensional vanishing cycles over points of @xmath320 form two sections @xmath942 of the map @xmath3 . in case m1 we can assume that one of these sections , say @xmath943 , consists of the points of the membrane @xmath404 .
the local structure near the membrane allows us to construct fiberwise embeddings @xmath944 and @xmath945 such that @xmath946 extends the section @xmath943 , @xmath947 and @xmath948 . applying theorem [ thm : harer2 ] with @xmath949 , @xmath950 , and @xmath951
, we construct a basis for a membrane eliminating surgery which removes the fold @xmath232 . in the case
m2 the enrichment structure for the membranes adjacent to @xmath232 provides us with an extension of sections @xmath952 and @xmath953 to disjoint fiberwise embeddings @xmath954 such that @xmath955 . to conclude the proof in this case we apply [ thm : harer2 ] with @xmath956 , @xmath957 , and @xmath958
suppose now that the index of @xmath232 is @xmath707 .
consider first the case m2 .
then the vanishing cycles over @xmath320 define a fiberwise embedding @xmath959 over @xmath320 which extends to a fiberwise embedding @xmath960 disjoint from all folds and their membranes , where @xmath323 is the annulus @xmath961 $ ] .
it follows from the definition of the category @xmath962 that the complement of the image of @xmath20 is fiberwise connected even after all singularities being removed
. hence we can apply theorem [ thm : harer2 ] with @xmath963 and @xmath964 to construct a basis for a membrane expanding surgery eliminating the fold @xmath232 .
finally , in the case @xmath965 each vanishing cycle @xmath966 , has a unique point @xmath967 which is also in the membrane @xmath404 of the fold @xmath232 .
this point is the center of an embedded disk @xmath968 , and the framed membrane gives a fiberwise embedding @xmath969 over @xmath311 , which over the boundary extends to a fiberwise embedding @xmath970 over @xmath795 .
hence , we are in a position to apply theorem [ thm : harer2 ] with @xmath949 and @xmath964 . consider a quadratic form @xmath971 take the handle @xmath972 and denote by @xmath973 the level set @xmath974 , @xmath975 $ ] .
when passing through the critical value @xmath242 , the level set @xmath973 experiences a surgery of index @xmath244 , i.e. changes from a @xmath707-sheeted to a @xmath244-sheeted hyperboloid .
the critical level set @xmath976 is the cone @xmath977 .
let us fix diffeomorphisms @xmath978 $ ] which send boundary circles of @xmath973 to @xmath979 , @xmath980 $ ] .
we will write @xmath981 $ ] for @xmath982 .
we will call the singularity of @xmath976 a _ node _ and call surfaces with such singularities
_ nodal_. a singular surface @xmath983 is called _ @xmath73-nodal _ if it is a smooth surface in the complement of @xmath73 points @xmath984 , while each of these points has a neighborhood diffeomorphic to @xmath976 .
the function @xmath474 is a folded map @xmath985 with the origin @xmath986 as its fold @xmath987 .
if we want to add to it the structure of an _ enriched _ folded map , there should be considered four local possibilities for the choice of an enriched membrane depending on the membrane index and the sign of @xmath988 as the membrane boundary component : a. _ index @xmath242 membrane with negative boundary _ , @xmath989 , @xmath990 ; b. _ index @xmath244 membrane with positive boundary _ , @xmath991 , @xmath992 ; c. _ index @xmath244 membrane with negative boundary_. @xmath993 , @xmath994 d. _ index @xmath707 membrane with positive boundary _ , @xmath995 , @xmath996 . in the cases ( i)-(ii ) the fold has index 1 and in the cases ( iii)-(iv ) index 2 with respect to the outward orientation of the boundary of the projection @xmath997 of @xmath404 .
a _ @xmath73-nodal fibration _
@xmath998 is a fiber bundle whose fibers are @xmath73-nodal surfaces , equipped with @xmath73 disjoint fiberwise embeddings @xmath999 over @xmath11 , such that the complement of the images of the @xmath1000 forms a smooth fiber bundle over @xmath1001 .
let @xmath998 be a @xmath73-nodal fibration .
denote @xmath1002 and construct a manifold @xmath1003 together with a map @xmath1004 as follows .
set @xmath1005 where @xmath1006 , @xmath1007 , are gluing diffeomorphisms defined by the formula @xmath1008 for @xmath982 , @xmath1009 and @xmath1010 for @xmath1011 .
the map @xmath1012 extends to a map @xmath1013 as equal to the projection @xmath1014 for @xmath1015 and equal to the map @xmath1016 on the @xmath1017-th copy of @xmath1018 glued with the attaching map @xmath1019 .
note that the map @xmath356 has @xmath73 fold components which are mapped to the hypersurfaces @xmath1020 .
thus @xmath1021 is the locus of @xmath73-multiple intersection of images of fold components of @xmath356 .
we will call the folded map @xmath1013 the _ universal unfolding _ of the @xmath73-nodal fibration @xmath998 .
the following lemma , which follows from the local description of an enriched folded map in a neighborhood of its fold , see lemma [ lemma - replacing - eq:14 ] , shows that the universal unfolding describes the structure of a folded map over a neighborhood of the locus of maximal multiplicity of fold intersection .
[ lm : k - multiple ]
let @xmath1022 be a folded map with cooriented hyperbolic folds .
suppose that all combinations of ( projections of ) fold components intersect transversally among themselves and with @xmath1023 .
let @xmath73 be the maximal multiplicity of the fold intersection and @xmath1024 be one of the components of the @xmath73-multiple intersection .
then @xmath1025 is a submanifold with boundary @xmath1026 , and the restriction @xmath1027 is a @xmath73-nodal fibration .
let @xmath1028 be the universal unfolding of the @xmath73-nodal fibration @xmath1029 .
then there exist embeddings @xmath1030 and @xmath1031 which extend the inclusions @xmath1032 and @xmath1033 such that the diagram @xmath1034^{\phi}\ar[d]^{{\widehat}f } & m\ar[d]^f\\ { \widehat}z \ar[r]^{\phi } & x}\ ] ] commutes .
if the folded map is enriched then , depending on indices of the membranes adjacent to the intersecting folds , and signs of the folds as boundary components of the membranes , one can arrange that the pre - images of the membranes and their framings under the embedding @xmath1035 coincide with the submanifolds @xmath1036 , @xmath1037 , and their defined above model framings in the corresponding copies of @xmath1018 in @xmath1003 .
notice that theorem [ thm : harer - geometric ] is easily generalized to @xmath73-nodal fibrations as follows .
[ thm : harer - geometric - nodal ] let @xmath195 be compact surfaces with boundary ( not necessarily connected ) .
let @xmath17 be a @xmath73-nodal fibration with @xmath60 ends , and let @xmath196 be a fiberwise embedding over @xmath2 , such that its image in each fiber is disjoint from the nodes , and that in each fiber the complement of its image is connected , even after removing all nodes .
then , after possibly changing @xmath17 by a bordism which is the trivial bordism over @xmath35 , the embedding @xmath20 extends to an embedding of @xmath197 , still disjoint from nodes and with connected complement .
let @xmath619 be an enriched folded map from @xmath919 .
let @xmath925 be a compact domain with smooth boundary .
we assume that all combinations of ( projections of ) fold components intersect transversally among themselves and with @xmath320 .
let @xmath73 be the maximal multiplicity of the fold intersection .
we denote by @xmath1038 , @xmath1039 , the set of intersection points of multiplicity @xmath1040 in @xmath311 , and set @xmath1041 .
thus we get a stratification @xmath1042 , and we have @xmath1043 . set @xmath1044 .
note that @xmath1045 is a closed submanifold of @xmath311 with boundary @xmath1046 .
the map @xmath1047 is a @xmath73-nodal fibration .
the membranes which are not adjacent to the fold components intersecting along @xmath1045 , together with their framings define fiberwise embeddings @xmath1048 over @xmath1045 , disjoint from the image of @xmath20 and from each other .
let us apply theorem [ thm : harer - geometric - nodal ] , the nodal version of harer s stability theorem , to the nodal fibration @xmath1049 and the fiberwise embedding @xmath1050 where @xmath1051 and @xmath1052 are disjoint unions of @xmath1053 and @xmath1054 , respectively , with @xmath1055 copies of the disc @xmath1056 . as a result
, we find a bordism @xmath1057 ( in the class of @xmath73-nodal fibrations ) between the @xmath73-nodal fibrations @xmath1058 and @xmath1059 which is constant over @xmath1060 and such that the fiberwise embedding @xmath1061 , extends to a fiberwise embedding @xmath1062)\cup({{\partial}}_+y_k= u_k')\right)\times { \widetilde}\sigma_2 \to w_k.\ ] ] let @xmath1063 be the universal unfolding of the @xmath73-nodal fibration @xmath1064 .
we view @xmath1065 as a bordism between the universal unfoldings @xmath1066 and @xmath1067 of the @xmath73-nodal fibrations @xmath1068 and @xmath1069 . the embeddings @xmath1070 extends to a fiberwise embedding @xmath1071 over @xmath1072 . according to lemma [ lm : k - multiple ] the restriction of @xmath3 to a tubular neighborhood of @xmath1045 is isomorphic to the universal unfolding @xmath1073 of the @xmath73-nodal fibration @xmath1068 . in other words , there exist embeddings @xmath1074 and @xmath1075 which extend the inclusions @xmath1076 and @xmath1077 such that the diagram @xmath1078^{\phi_k}\ar[d]^{{\widehat}f_k}&m\ar[d]^{f}\\ { \widehat}x_k \ar[r]^{\phi_k } & x}\ ] ] commutes .
moreover , @xmath1079 can be chosen in such a way that the framed membranes adjacent to intersecting folds correspond to model framed membranes of lemma [ lm : k - multiple ] .
let us glue the bordism @xmath1080 to the trivial bordism @xmath1081 using the attaching maps @xmath1082 : @xmath1083 and then smooth the corners .
the folded map @xmath1084 resulting from this construction is a bordism between @xmath205 and @xmath1085 , which is trivial over the complement of @xmath1086 .
model framed membranes of lemma [ lm : k - multiple ] provide us with a canonical extension to @xmath1087 of framed membranes adjacent to @xmath1088 . on the other hand ,
the restriction of @xmath1089 to @xmath1090 allows us to extend all the other framed membranes .
thus the constructed map @xmath1084 together with the extended enrichment is a bordism in the category @xmath919 .
the fiberwise embedding @xmath1091 extends to a closed neighborhood @xmath1092 in @xmath719 .
consider the domain @xmath1093 .
the maximal multiplicity of fold intersection over @xmath1094 is equal to @xmath610 .
hence , we can repeat the previous argument to extend @xmath1089 over the stratum @xmath1095 , possibly after changing it by another bordism in the category @xmath919 .
continuing inductively we find the required extension to the whole domain bounded by @xmath320 .
as it was already mentioned above , theorem [ thm : main1 ] follows from theorem [ thm : epi - folds ] and propositions [ prop : enriched - hyperbolic ] and [ prop : hyperbolic - fibrations ] .
we prove these propositions in the next two sections .
let @xmath418 be an enriched folded map with @xmath17 .
first , we get rid of non - hyperbolic folds .
let @xmath1001 be a non - hyperbolic fold component .
the following procedure , which is illustrated in fig .
[ mw11 ] , replaces @xmath1001 by a parallel hyperbolic fold .
let @xmath404 be the membrane of @xmath1001 , and @xmath1096 \subset x$ ] be an interior collar of @xmath1097 in @xmath1098 .
let us recall that the map @xmath3 has a standard end , where it is equivalent to the trivial fibration @xmath1099 .
let @xmath1100\subset n$ ] and @xmath1101\subset n$ ] , so that @xmath1102
. let us lift @xmath1103 to an annulus @xmath1104 , @xmath1105 . write @xmath1106 @xmath1107 , @xmath1108 , @xmath1109 , so that @xmath1110 and @xmath1111 .
using lemma [ lm:2fold - creation ] we can create a double fold with the membrane @xmath323 and folds @xmath1112 of index @xmath244 and @xmath1113 of index @xmath242 with respect to the coorientation of @xmath1114 by the second coordinate of the splitting @xmath1115 $ ] .
the folds @xmath1113 and @xmath1001 have index @xmath242 with respect to the outward coorientation of @xmath1116 , and hence we can use a membrane expanding surgery to kill both folds , @xmath1001 and @xmath1113 , and spread their membranes over @xmath1117 . as a result of this procedure
we have replaced @xmath1001 by a hyperbolic fold @xmath1112 .
it remains to make the fibers of @xmath1118 connected .
we begin with the following lemma from @xcite .
[ lm : discs ] let @xmath1119 be an enriched folded map without folds of index @xmath242 .
then there exist disjoint @xmath143-discs @xmath1120 , @xmath933 , embedded into @xmath1 , such that * @xmath1121 is embedding @xmath1122 for each @xmath1123 ; * @xmath1124 , where @xmath212 is the union of all membranes ; * for each @xmath1125 each irreducible component of @xmath1126 intersects at least one of the discs @xmath1120 at an interior point .
note first , that the statement is evident for any _ fixed _
hence , without controlling the disjointness of the disks @xmath1120 the statement just follows from the compactness of @xmath2 .
one can choose the required _ disjoint _
disks @xmath1120 using the following trick : fix a function @xmath1127 and take the disks @xmath1120 such that each disk belongs to its own level hypersurface of @xmath221 .
when we choose such disks for @xmath1125 one needs to avoid the points @xmath1128 where the level hypersurface is tangent to the fiber @xmath1129 .
it can be done by a small perturbation of disks , if the complement of all `` bad '' points ( for all @xmath1125 ) is _ open _ and _ dense _ in @xmath1129 for all @xmath1125 . but thom s jet transversality theorem asserts that this is a generic situation for functions @xmath1127 .
let the disks @xmath1130 be as in lemma [ lm : discs ] .
let us also consider discs @xmath1131 , @xmath933 , where @xmath1132 are disjoint points at the end @xmath1133 of the fiber @xmath37 .
next , using each pair @xmath1134 , @xmath933 , as a basis for an index 1 fold creating surgery we create new folds @xmath1135 of index @xmath244 with the discs @xmath1136 serving as membranes , while making all the fibers connected , see remark [ rem : fold - creating ] . as a result of this step
we arrange all fibers @xmath1137 , to be connected .
this completes the proof of proposition [ prop : enriched - hyperbolic ] .
let @xmath1138 be the collection of ( connected ) framed membranes of @xmath1139 .
we are going to inductively remove all of them . if the membrane @xmath1140 is pure then , in view of corollary [ cor : basis - existence ] , we can assume , after a possible change of @xmath3 by a bordism in the category @xmath919 , that there is a basis for the fold surgery which removes @xmath1141 together with the membrane .
suppose now that the membrane @xmath1140 is mixed .
consider first the case @xmath892 . by our assumption
each boundary component @xmath1142 of @xmath1143 bounds in this case a domain @xmath236 . if @xmath1144 then the domain @xmath237 is uniquely defined .
if @xmath2 is closed then we fix a point @xmath1145 and denote by @xmath237 the domain which does not contain @xmath256 .
there exists exactly one boundary component @xmath232 of @xmath1140 such that @xmath1146 . for any other boundary component @xmath1147 of @xmath1140
we have @xmath1148 .
we will refer to @xmath232 as the exterior fold of @xmath1140 , and to all other boundary folds of @xmath1140 as its interior folds .
first , we apply corollary [ cor : basis - existence ] in order to create a basis for a surgery which eliminates each interior fold @xmath1147 and spreads the membrane @xmath1140 over @xmath1149 . after applying this procedure to all interior folds of @xmath1140
we will come to the situation when @xmath1150 is connected , and hence the membrane @xmath1140 is pure , which is the case we already considered above .
applying the same procedures subsequently to the membranes @xmath1151 we complete the proof of proposition [ prop : hyperbolic - fibrations ] when @xmath892 .
finally consider the case @xmath357 , i.e. @xmath1152 or @xmath1153 .
we assume for determinacy that @xmath1152 , i.e. @xmath3 is a morse function .
a mixed framed membrane of @xmath3 connects two critical points @xmath1154 of @xmath3 of index @xmath244 and @xmath707 , and with critical values @xmath1155 , respectively . for
a small @xmath1156 let us consider vanishing circles @xmath1157 of critical points @xmath1158 and @xmath1159 , where we denote @xmath1160 .
choose an oriented embedded circle @xmath1161 which intersects @xmath1162 transversally in one point .
let @xmath323 denote the annulus @xmath961 $ ] , and @xmath1163 be a small 2-disc centered at a point @xmath1164 .
choose a fiberwise embedding @xmath1165 over @xmath1166 $ ] , such that @xmath1167 and the linearization of @xmath73 along @xmath1168 provides the given framing of the membrane @xmath404 .
let us also choose embeddings @xmath1169 and @xmath1170 such that @xmath1171 , @xmath1172 , and for @xmath1173 we have @xmath1174 .
let us apply theorem [ thm : harer2 ] to the morse function ( @xmath1175 the folded map ) @xmath1176 with @xmath1177 , @xmath1178 and the embedding @xmath1179 , which is equal to @xmath1180 on @xmath1181 and to @xmath73 on @xmath1182 .
theorem [ thm : harer2 ] then allows us , after possibly changing @xmath3 in its bordism class in @xmath1183 , to extend @xmath1184 to a fiberwise over @xmath1185 map @xmath1186 which coincides with @xmath73 on @xmath1187 .
choosing a metric for which the cylinder @xmath1188 is foliated by gradient trajectories , one of which is @xmath404 , we come to the situation when we can apply the standard morse theory cancellation lemma , see for instance @xcite , to kill both critical points @xmath1189 and @xmath1158 together with their membrane @xmath404 .
the cancellation deformation is inverse to the double fold creation , and hence it can be realized by a bordism in the category @xmath1190 .
this completes the proof of theorem [ thm : main1 ] .
let @xmath892 . given a map @xmath205 , a point @xmath255
is called a _ cusp _ type singularity or a _ cusp _ of index @xmath1191 if near the point @xmath256 the map @xmath3 is equivalent to the map @xmath1192 given by the formula @xmath1193 where @xmath1194 .
the set of cusp points is denoted by @xmath1195 .
it is a codimension @xmath244 submanifold of @xmath206 which is in the above canonical coordinates is given by @xmath1196 the vector field @xmath1197 along @xmath1195 is called the _
characteristic vector _ field of the cusp locus .
it can be invariantly defined as follows .
note that for any point @xmath1198 there exists a neighborhood @xmath1199 in @xmath2 such that @xmath1200 can be presented as a union of two manifolds @xmath1201 with the common boundary @xmath1202 , the common tangent space @xmath1203 at the point @xmath1204 , and the common outward coorientation @xmath114 of @xmath1205 . on the other hand
, the differential @xmath326 defines an isomorphism @xmath1206 hence , there exists a vector field @xmath11 transversal to @xmath1207 in @xmath859 along @xmath1195 , whose projection defines the coorientation @xmath114 of @xmath1208 in @xmath1209 for all points @xmath1198 .
one can show that any vector field @xmath11 defined that way coincides with the vector field @xmath1197 for some local coordinate system in which the map @xmath3 has the canonical form .
note that the line bundle @xmath1210 over @xmath1195 is always trivial .
indeed , @xmath1210 can be equivalently defined as the kernel of the quadratic form @xmath1211 , and thus one has an invariantly defined cubic form @xmath1212 which does not vanish .
the bundle @xmath1213 can be split as @xmath1214 , so that the quadratic form @xmath1215 is positive definite on @xmath1216 and negative definite on @xmath695 .
the singularity @xmath1221 is the @xmath277-dimensional sphere @xmath1222 whose equator @xmath1223 consists of cusp points of index @xmath1191 .
the upper hemisphere @xmath1224 consists of folds of index @xmath81 , while the lower one @xmath1225 consists of folds of index @xmath1226 .
the radial vector field @xmath1227 serves as a characteristic vector field of the cusp locus .
although the differential @xmath1228 degenerates at points of @xmath1229 , it can be canonically _ regularized _ over @xmath1230 , an open neighborhood of the disk @xmath1231 .
namely , we can substitute the element @xmath1232 in the jacobi matrix of @xmath1233 by a function @xmath297 which coincides with @xmath1234 on @xmath1235 and does not vanish along the @xmath143-dimensional subspace @xmath1236 the new bundle map @xmath1237 provides a homotopically canonical extension of the map @xmath1238 to an epimorphism ( fiberwise surjective bundle map ) @xmath1239 .
we call @xmath1240 the _ regularized differential _ of the map @xmath1233 .
a map @xmath304 defined on an open ball @xmath305 is called a _
wrinkle _ of index @xmath1241 if it is equivalent to the restriction @xmath1242 .
we will use the term `` wrinkle '' also for the singularity @xmath1243 of a wrinkle @xmath3 .
notice that for @xmath357 the wrinkle is a function with two nondegenerate critical points of indices @xmath81 and @xmath1226 given in a neighborhood of a gradient trajectory which connects the two points .
a map @xmath253 is called _ wrinkled _ if there exist disjoint open subsets @xmath358 such that the restriction @xmath1244 is a submersion ( i.e. has rank equal @xmath143 ) and for each @xmath360 the restriction @xmath361 is a wrinkle .
the singular locus @xmath206 of a wrinkled map @xmath3 is a union of @xmath277-dimensional spheres ( wrinkles ) @xmath1245 .
each @xmath1246 has a @xmath1247-dimensional equator @xmath1248 of cusps which divides @xmath1246 into two hemispheres of folds of two neighboring indices .
the differential @xmath1249 can be regularized to obtain an epimorphism @xmath1250 . to get @xmath370 we regularize @xmath371 for each wrinkle
@xmath361 .
[ wrinkled mappings][thm : wrinkled ] let @xmath1251 be an epimorphism which covers a map @xmath1252 .
suppose that @xmath3 is a submersion on a neighborhood of a closed subset @xmath374 , and @xmath37 coincides with @xmath326 over that neighborhood
. then there exists a wrinkled map @xmath1253 which coincides with @xmath3 near @xmath376 and such that @xmath1254 and @xmath37 are homotopic rel .
moreover , the map @xmath51 can be chosen arbitrarily @xmath379-close to @xmath3 and with arbitrarily small wrinkles .
we are going to modify each wrinkle to a spherical double fold using _ cusp elimination surgery _ , which is one of the surgery operations studied in @xcite .
unlike fold elimination surgeries described above in section [ sec : surgery - general ] cusp elimination surgery does not affect the underlying manifold and changes a map by a homotopic one . for maps @xmath1256
the operation is shown on fig.[mw1 ] .
let @xmath1257 be a connected component of the cusp locus .
let @xmath11 be the characteristic vector field of @xmath232 .
suppose that the bundles @xmath1258 and @xmath1210 over @xmath232 are trivialized , respectively , by the frames @xmath1259 a _ basis _ for a cusp eliminating surgery consists of an @xmath277-dimensional submanifold @xmath1260 bounded by @xmath232 , together with an extension of the above framing as a trivialization of the normal bundle @xmath114 of @xmath323 in @xmath1 , such that let us extend @xmath323 to a slightly bigger manifold @xmath1265 such that @xmath1266 , and extend the framing over @xmath1265
. one can show ( see @xcite and @xcite ) that there exists a splitting @xmath1267 of a tubular neighborhood of @xmath1265 in @xmath1 , such that in the corresponding local coordinates @xmath1268 and @xmath1269 the map @xmath3 can be presented as a composition @xmath1270 where @xmath221 is an immersion and @xmath37 has the form @xmath1271 where the function @xmath1272 satisfies @xmath1273 on @xmath1274 and @xmath1275 on @xmath1276 , @xmath1277\to[0,1]$ ] is a cut - off function equal to @xmath244 near @xmath242 and to @xmath242 near @xmath244 , and @xmath1156 is small enough .
consider another function @xmath1278 which coincides with @xmath1279 outside @xmath1280 and such that @xmath1281 denote @xmath1282 , @xmath535 $ ] , and consider homotopies @xmath1283 @xmath1284 , and @xmath1285 .
the homotopy @xmath1286 is supported in @xmath311 and hence can be extended to the whole manifold @xmath1 as equal to @xmath3 on @xmath1287 .
the next proposition is straightforward . 1 .
the homotopy @xmath1286 removes the cusp component @xmath232 .
the map @xmath1288 coincides with @xmath1289 outside @xmath311 , has only fold type singularities in @xmath311 , and @xmath1290 2 .
suppose that @xmath1291 consists of only fold points and that the restriction of the map @xmath3 to @xmath1292 is an embedding .
then the restriction @xmath1293 is an embedding provided that the neighborhood @xmath1294 in the surgery construction is chosen small enough .
\a ) there exists an embedding @xmath1296 and a framing @xmath705 of the normal bundle to @xmath1297 such that the pair @xmath1298 forms a basis for a surgery eliminating the cusp @xmath1299 of the wrinkle ; the construction is clear from fig.[mw2 ] .
the manifold @xmath323 in this case is obtained from the boundary of the upper semi - ball @xmath1304 by removing the open disc @xmath1305 , and then smoothing the corner . here
@xmath1306 should be chosen small enough so that @xmath323 lie in the prescribed neighborhood of the wrinkle .
unfortunately the embedding @xmath1301 does not satisfies the property @xmath1311 . however , if @xmath223 this can be corrected as follows .
we suppose that the index @xmath1312 ( if @xmath1313 then one should start with an embedding @xmath221 obtained by smoothing the boundary of the _ lower _ semi - ball ) .
let us denote by @xmath1314 the composition @xmath1315 , and by by @xmath1316 and @xmath1317 the projections of @xmath1314 to the first and second factors , respectively .
* @xmath456 vanishes along @xmath1319 together with all its derivatives ; * @xmath1320 ; * the function @xmath1321 has a unique interior critical point , the minimum , at @xmath1322 ; * the map @xmath1323 is an embedding , and the image @xmath1324 does not intersect the image of the wrinkle .
[ prop : wrinkle - to - double ] [ thm : surgery ] there exists a @xmath379-small perturbation of the map @xmath1242 in an arbitrarily small neighborhood of the embedded disk @xmath1330 constructed in [ prop : preparation ] such that the resulting map @xmath1331 is a special folded map with only one double fold ( of index @xmath1332 ) .
moreover , the regularized differentials of @xmath1233 and @xmath1331 are homotopic . for a pair @xmath140 of spaces ,
let @xmath141 denote the set of bordism classes of continuous maps of pairs @xmath1334 for smooth oriented compact manifolds @xmath1335 with boundary @xmath144 .
there is a `` cycle map '' @xmath1336 that maps the class of @xmath3 to @xmath1337 ) \in h_n(x , a)$ ] .
recall that @xmath157 is a `` generalized homology theory '' , i.e. it satifies the same formal properties as singular homology ( the eilenberg - steenrod axioms except the dimension axiom ) .
in particular we have a long exact sequence for pairs of spaces .
if @xmath158 is an arbitrary map , then we can replace @xmath11 by the mapping cylinder of @xmath3 .
the axioms implies there is a natural long exact sequence @xmath1338 where @xmath1339 is the mapping cone and @xmath1340 is the cone point .
it seems well known that a map @xmath1341 induces an isomorphism in oriented bordism if and only if it induces an isomorphism in singular homology ( with @xmath856-coefficients ) .
this can be seen e.g. from the atiyah - hirzebruch spectral sequence .
we offer a more geometrical argument , based on hurewicz theorem .
recall that this says that a map @xmath158 of simply connected spaces is a weak equivalence if and only if it is a homology equivalence .
we will use the _ unreduced suspension _
@xmath1342 of a space @xmath2 .
this is the union @xmath1343 of two cones .
we will regard @xmath1344 as a pointed space , using one of the cone points ( which we denote @xmath1345 ) as basepoints .
it is easily seen that @xmath1342 is simply connected if and only if @xmath2 is arcwise connected .
also it follows from the mayer - vietoris exact sequence that @xmath1346 for any homology theory @xmath1347 ( e.g. singular homology or oriented bordism ) . for the converse
assume @xmath1351 is an isomorphism .
then @xmath1352 is an isomorphism so we can assume that @xmath2 and @xmath11 are arc connected
. then @xmath1342 and @xmath1353 are simply - connected .
then hurewicz theorem implies that @xmath1354 is a weak equivalence .
we proved above that @xmath3 is a homology equivalence if and only if @xmath1348 is a weak equivalence .
using the bordism hurewicz theorem we can prove in the same way that @xmath3 is a bordism isomorphism if and only if @xmath1348 is a weak equivalence .
we sketch a proof of a bordism hurewicz theorem .
let @xmath1355 be the map that maps the homotopy class of a map @xmath1356 to the bordism class of the same map .
the composition of @xmath1357 with the cycle map is the classical hurewicz homomorphism .
we construct an inverse .
let @xmath1362 represent an element of @xmath1363 .
we can assume @xmath1 is connected .
choose a cw - structure on @xmath1364 with only one top cell @xmath1365 .
let @xmath1366}$ ] and @xmath1367}$ ] denote the skeleta in the chosen structure .
these fit into a cofibration sequence @xmath1368}/m^{[n-2 ] } \to m / m^{[n-2 ] } \to
m / m^{[n-1 ] } \to \sigma ( m^{[n-1]}/m^{[n-2]}).\ ] ] the last of these maps is a pointed map @xmath1369 , where @xmath73 is the number of @xmath313 cells in @xmath1364 .
it must induce the zero map in homology , because otherwise we would have @xmath1370 .
hence it is null - homotopic because @xmath1371 . applying the functor @xmath1372 $ ] ,
pointed homotopy classes of maps to @xmath1359 , then gives a isomorphisms @xmath1373},(x , x_0 ) ] \to [ ( m,\partial m ) , ( x , x_0)]\ ] ] notice how surjectivity of the hurewicz homomorphism is easier than injectivity .
surjectivity uses only connectivity estimates , but injectivity uses a property of the attaching map of the top dimensional cell in an oriented manifold .
assume @xmath158 is a map of simply connected spaces that is a bordism isomorphism .
let @xmath1339 be the mapping cone and @xmath1379 the cone point .
it follows from the long exact sequence in oriented bordism that @xmath1380 . hence by the previous theorem ( using that @xmath1339 is simply connected ) we get that @xmath1339 is weakly contractible .
therefore @xmath1381 , so @xmath3 is a homology isomorphism and hence a weak equivalence by the classical hurewicz theorem .
ab99 v.i .
arnold , _ wave front evolution and equivariant morse lemma _ , comm . in
pure and appl .
29(1976 ) , 557 - 582 .
s. boldsen , _ improved homological stability for the mapping class group with integral or twisted coefficients _ , arxiv:0904.3269v1 .
y. eliashberg , _ surgery of singularities of smooth maps _ ,
nauk sssr ser . mat .
, * 36*(1972 ) , 13211347 . d.b .
fuchs , _ quillenization and bordism .
_ funkcional .
i priloen . 8 ( 1974 ) ,
1 , 3642 .
ivanov , _ stabilization of the homology of teichmller modular groups _
, algebra i analiz 1 ( 1989 ) , no .
3 , 110126 . n.v .
ivanov , _ on the homology stability for teichmller modular groups : closed surfaces and twisted coefficients _
, mapping class groups and moduli spaces of riemann surfaces ( gttingen , 1991/seattle , wa , 1991 ) , 149194 , contemp .
, 150 , amer .
soc . , providence , ri , 1993 .
j. harer , _ stability of the homology of the mapping class groups of orientable surfaces _ , ann . of math .
( 2 ) 121 ( 1985 ) , no .
2 , 215249 . i. madsen and u. tillmann , _ the stable mapping class group and @xmath1382 _ , invent .
( 2001 ) , no .
3 , 509544 . i. madsen , and m. weiss , _ the stable moduli space of riemann surfaces : mumford s conjecture _ , ann . of math .
( 2 ) 165 ( 2007 ) , no .
3 , 843941 .
j. milnor , _ lectures on the @xmath221-cobordism theorem _ , princeton university press , princeton , nj , 1965 | we give an alternative proof of madsen - weiss _ generalized mumford conjecture_. our proof is based on ideas similar to madsen - weiss original proof , but it is more geometrical and less homotopy theoretical in nature . at the heart of the argument is a geometric version of _ harer stability _ , which we formulate as a theorem about folded maps . |
this supplementary contains details about our analytical approach to solve for the spectrum and the wavefunctions of both bulk and edge states for a generalized kane - mele ( km ) model @xcite .
we shall consider two kinds of boundary conditions , corresponding to the zigzag and beard edges .
the model hamiltonian reads : @xmath160 we have included a staggered potential where @xmath161 for @xmath162 and @xmath163 for @xmath164 sublattice .
this staggered potential can be used to drive a transition between a trivial and a topological phase . in the above expression , @xmath165 denotes the spin and @xmath166 describes the sublattice pseudo spin components corresponding to the @xmath167 sublattices . as discussed in the main text , there is a gauge degree of freedom for fourier transformation of the fermion creation and destruction operators due to the bi - particle structure of lattice .
the two types of edges considered below correspond to two different gauge choices : 1 ) zigzag edge : @xmath27 and 2 ) beard edge : @xmath168 .
first we shall derive the spectrum of the edge states for the zigzag edge .
after fourier transformation and taking into account that @xmath165 is a good quantum number , we obtain the schrdinger equation @xcite , @xmath169 where @xmath37 , and @xmath170 .
note that @xmath33 , with @xmath34 , @xmath35 and @xmath171 , respectively . in the following
, we will work in units where @xmath172 , for the sake of notational simplicity .
in addition , we treat @xmath173 as an operator and @xmath174 as its eigenvalues . looking for solutions of the form @xmath175 , where @xmath176 leads to the following secular equation : @xmath177^{2}-t^{2}[1 + 4\cos \frac{k_{x}}{2}\cosh \frac{\sqrt{3}\kappa } { 2}]-4t^{2}(\cos \frac{k_{x}}{2})^{2}=0 . \label{secular-1}\ ] ] this equation relates @xmath178 and the eigenvalues , @xmath85 .
it is a quadratic equation of the variable @xmath179 . by denoting @xmath180 , @xmath181 , and @xmath182
, we obtain the two roots : @xmath183 which implies there are four roots for @xmath184 , corresponding to @xmath185 with @xmath186 .
the latter are all functions of the eigenvalue @xmath85 .
note that for the edge states , we have that @xmath187 .
thus , we use the convention that @xmath188 for the function @xmath189 .
next we note that two linearly independent wavefunctions satisfy the open boundary conditions corresponding the zigzag edge , namely @xmath190 for each value of @xmath85 are @xmath191 and @xmath192 , where @xcite @xmath193 hence , the eigenfunctions can be expressed as the linear combination thereof . by introducing a @xmath194 matrix of coefficients @xmath195 _
{ 2\times 2}$ ] , the eigenfunction can be written as follows : @xmath196 .
\label{wave-1}\ ] ] let us substitute thisfunction back into the schrdinger equation , eq . .
because @xmath197 are linearly independent , we obtain the following conditions relating the column vectors of the matrix @xmath198 : @xmath199 where @xmath200 , @xmath201 , @xmath202 , and @xmath203 , respectively .
note that , in the above derivation , we have used the fact @xmath204 , @xmath205 and @xmath206 .
the combinations of equations in the same column in eq.([relationsforl ] ) give the secular equation ( [ secular-1 ] ) , which has been used to relate @xmath207 and spectrum @xmath208 .
the other two independent equations are obtained by combining diagonal terms in eq.([relationsforl ] ) , which yields : @xmath209where @xmath210 .
expressing @xmath207 as functions of @xmath85 , this equation is exactly the constraint for spectrum @xmath85 . in the following
, we will solve this equation .
eq.([relation-1 ] ) implies @xmath211 , where @xmath212 . to have a nontrivial solution for @xmath213
, the condition @xmath214 is required .
@xmath215 + \frac{1}{t}\left [ d_{0}-d_{x}\sigma _ { x}-d_{y}\sigma _ { y}-d_{z}\sigma _ { z}\right ] } { t^{2}\sin \beta _ { 1}\sin \beta _ { 2}},\ ] ] where @xmath216 thus the condition @xmath214 becomes @xmath217 . especially for @xmath218 , we have @xmath219 because @xmath188 . thus it further becomes @xmath220 , which gives the spectra @xmath221 note that eq.([relation-1 ] ) gives an additional constraint for the spectra . from @xmath211 and @xmath219
, we obtain @xmath222 . combining it with @xmath223 ( eq.([relation-1 ] ) ) , we get the constraint @xmath224 the second constraint is @xmath188 .
these constraints restrict a region @xmath225 $ ] , where edge states exist .
we show the resulted spectra in fig.[spectrum - z ] . and
@xmath226 ; central panel : @xmath227 and @xmath228 ; right panel : @xmath227 and @xmath229.,title="fig:",width=5 ] and @xmath226 ; central panel : @xmath227 and @xmath228 ; right panel : @xmath227 and @xmath229.,title="fig:",width=5 ] and @xmath226 ; central panel : @xmath227 and @xmath228 ; right panel : @xmath227 and @xmath229.,title="fig:",width=5 ] +
to investigate the wavefunctions at one of edge , it is helpful to shift the position of system so that the system is on the upper half plane ( @xmath230 with @xmath218 ) . for a semi - infinite system ,
the wavefunction satisfying the boundary condition @xmath231 has a much simpler form @xmath232 the left problems are to solve the @xmath233 matrix @xmath234 and the normalization factor @xmath235 . for each @xmath50
, we have obtain the spectra @xmath236 and the ` wave - vector ' @xmath237 with @xmath238 in the last section .
substituting eq . into the schrdinger eq.([schr-12 ] ) , we obtain @xmath239 _ { 12}/\left\ { \left [ \mathcal{h}_{0}^{s}\right ] _ { 11}-\epsilon_{s}\right\ } \\
1\end{array } \bigg ] \equiv \bigg [ \begin{array}{c } \chi_{s,1 } \\ 1\end{array}\bigg].\ ] ] explicitly , @xmath240-\epsilon_s}\ ] ] note that the system is a lattice structure , and therefore the wavefunction is defined on the discrete lattice points only : @xmath241 where @xmath242 .
the normalization factor is @xmath243 where @xmath244^{-1/2}\ ] ] where the function @xmath245 $ ] . upon denoting @xmath246 as the @xmath247 components of @xmath248
, we find @xmath249 for the case @xmath250 and @xmath251 , which suggesting the bottom edge states ` prefer ' b - sublattice .
for the bulk states with periodic boundary conditions , crystal momentum @xmath252 is treated as good quantum number in both the @xmath253-direction and @xmath20-direction .
thus , upon setting @xmath254 in eq .
( with @xmath255 ) , we obtain the ( bulk ) dispersion : @xmath256^{2}},\ ] ] where @xmath257 . however , for open boundary conditions and in the limit @xmath258 , the spectrum of bulk state is not modified from the above form because the boundary effects become negligible in the thermodynamic limit . on other hand , wavefunctions are modified and become different from bloch waves because of the scattering with the boundary .
thus , from the secular equation , for each @xmath259(@xmath260 is real ) and thus @xmath261 , we can find another root , @xmath262 .
in total four different roots for @xmath178 exist , i.e. , @xmath185 with @xmath186 , corresponding to a same energy @xmath85 .
note that @xmath263 and thus @xmath264 are real .
thus there are two different cases : 1 ) @xmath265 , the plane wave decays at the edge ; and 2 ) @xmath266 , different modes interference with each other : and @xmath226 ; central panel : @xmath227 and @xmath228 ; right panel : @xmath227 and @xmath229.,title="fig:",width=5 ] and @xmath226 ; central panel : @xmath227 and @xmath228 ; right panel : @xmath227 and @xmath229.,title="fig:",width=5 ] and @xmath226 ; central panel : @xmath227 and @xmath228 ; right panel : @xmath227 and @xmath229.,title="fig:",width=5 ] so far , we have obtained the eigenvalues and the complete set of eigenfunctions for the model of eq .
[ hami - rxsm ] .
hence , the green s function can be expressed in terms of them : @xmath317 where @xmath318 . in the real space , @xmath319 , where @xmath320 represents the different components of @xmath321 .
thus , using @xmath322 and @xmath323 , where @xmath324 and @xmath246 are the @xmath325 components of @xmath326 and @xmath248 respectively , we have @xmath327
for beard edge , using the gauge @xmath168 , and following the same steps as for the zigzag case , we obtain the spectrum for edge state , @xmath328}{\sqrt{(t\cos \alpha ) ^{2}+(2\lambda _ { \mathrm{so}}\sin \alpha ) ^{2}}}.\ ] ] in this case , the constraint becomes @xmath329 , where @xmath330 , @xmath331 and @xmath332 .
the resulting band structure is shown in fig .
[ spectrum - b ] .
note that the edge states intersect at @xmath333 @xcite .
the transmission coefficient @xmath334 as a function of the impurity strength @xmath11 and the electron energy , for an impurity on the b sublattice site on the rows @xmath335 and @xmath336 is shown in fig .
[ trans ] . for an impurity on a b sublattice site on the first atomic row ( i.e. @xmath335 ) ( left ) , @xmath337 ( right ) ( the spin - orbit coupling is @xmath338 ) .
, title="fig:",width=3 ] for an impurity on a b sublattice site on the first atomic row ( i.e. @xmath335 ) ( left ) , @xmath337 ( right ) ( the spin - orbit coupling is @xmath338 ) . ,
next , in order to deal with the effects of interactions in a nonperturbative way , we shall rely upon the bosonization technique .
the resulting model is analyzed using the renormalization - group ( rg ) following ref . .
using bosonization , the electron field operator for the right ( @xmath339 ) and left moving ( @xmath17 ) edge electron can be expressed in terms of a set of bosonic fields @xmath340 and @xmath341 as follows : @xmath342 } , \ ] ] where @xmath343 is a short - distance cutoff , @xmath344 is the plasmon velocity , @xmath345 and @xmath346 are the so - called klein factors satisfying @xmath347 , which allows to satisfy the anti - commutation relations between the two fermion chiralities @xmath339 and @xmath17 .
the bosonic fields obey @xmath348 = i\frac{\pi}{2 } \mathrm{sgn}(x^{\prime}-x)$ ] .
the chiral densities are given by @xmath349 after bosonizing the low energy effective model and upon applying a unitary transformation generated by @xmath350 $ ] with @xmath351 , @xmath352 , and using the factor @xmath353 = \partial _ { x}\phi \left ( x\right ) + \zeta \pi \delta \left ( x\right ) $ ] , the forward scattering term @xmath147 can be eliminated from @xmath354 ( cf .
eq . 13 in the main text ) , and the resulting hamitonian , @xmath355 reads@xcite : @xmath356 + 2\frac{y_{b}}{\xi } ( d^{\dagger } d-\frac{1}{2})[u_{r}u_{l}e^{2i\phi _ { 0}}+h.c.]\notag \\ & + \frac{y_{t}}{\xi } \left [ d^{\dagger } ( u_{r}e^{-i\left ( \phi _ { 0}-\lambda \theta _ { 0}\right ) } -u_{l}e^{-i\left ( -\phi _ { 0}-\lambda \theta _ { 0}\right ) } ) + h.c.\right ] , \\ h_{\ast } & = \frac{v}{2\pi } \int dx\left [ k\left ( \partial _ { x}\theta \right ) ^{2}+k^{-1}\left ( \partial _ { x}\phi \right ) ^{2}\right ] -\varepsilon _ { 0}(d^{\dagger } d-\frac{1}{2}).\end{aligned}\ ] ] where @xmath357 , @xmath358 , @xmath359 , @xmath360 , @xmath361 , @xmath362 are dimensionless couplings , and @xmath363 is the so - called luttinger parameter and @xmath364 is the edge plasmon velocity .
here we also use the factor @xmath365 $ ] . in the above equation @xmath366
denotes the distance of the ( low - energy ) bound state from the fermi energy @xmath131 . in what follows we focus on the resonant case for which @xmath367 .
taking into account the scaling behavior @xmath368 and @xmath369 } e^{-i [ \phi_0 ( \tau ) -\lambda \theta_0 ( \tau ) ] } \right\rangle \sim |\tau^{\prime } -\tau| ^{-k/2-\lambda ^{2}k^{-1}/2},\ ] ] and their operator product expansions ( ope ) , using cardy s approach @xcite , we arrive at the following set of rg equations valid to second order in the couplings describing backscattering ( @xmath370 ) and tunneling in and out of the resonant level : @xmath371 y_{t}+y_{t}(y_{b } + v_b ) , \\
\frac{d\delta_f}{d\ln \xi } & = 4 ( 1-\delta_f ) y^2_t,\\ \frac{d v_b}{d\ln \xi } & = ( 1-k ) v_b.\end{aligned}\ ] ] these rg equations are similar to the ones derived in ref . .
they show that for weak to moderate attractive interactions ( i.e. @xmath372 ) , the tunneling operator @xmath373 is flows to strong coupling . on the other hand , both the backscattering interaction ( @xmath374 ) and potential ( @xmath375 ) will be initially suppressed .
eventually , the runaway of @xmath376 drags along @xmath377 and @xmath370 , quickly driving the forward interaction with the level to its fixed point @xmath378 . as a result
, the transmission through the impurity will be suppressed @xcite , as discussed in the main text . | we study the problem of a backscattering impurity coupled to the edge states of a two - dimensional topological insulator . in the regime
where the backscattering potential is larger than the band gap and accounting for electron - electron interactions , it is shown that the system can be described as a resonant level coupled to the one - dimensional ( 1d ) channel of interacting edge electrons .
we discuss the relationship of this system to the model of a ( structureless ) impurity in a 1d interacting electron liquid .
different from the latter model , in the resonant regime transmission is suppressed also for weak to moderately attractive interactions . at zero temperature ,
charge transport in two - dimensional topological insulators ( 2dti ) takes place through 1d edge states , whose metallic character is protected by symmetry @xcite .
for instance , time reversal symmetry ( trs ) prevents the ubiquitous anderson localization from happening in the 1d edge channels of quantum spin hall insulators ( qshi ) in the presence of scalar and spin - orbit ( so ) disorder , and for weak to moderate interactions @xcite .
the qsh effect exhibited by qshi has been observed in hgte / cdte and inas / gasb / alsb quantum wells @xcite .
however , for longer ( @xmath0 m ) edge channels , deviations from the expected conductance quantization of @xmath1 and relatively short mean - free paths have been measured @xcite . despite intense theoretical investigation @xcite on the backscattering ( bs ) mechanisms , a complete understanding of the origin of the finite edge channel resistance has not been achieved .
the proposed mechanisms involve electron - electron scattering , often in combination with scalar , so , or magnetic disorder @xcite . magnetic impurities break trs ( above the kondo temperature ) and lead to bs @xcite . yet ,
whether electron bs yields or not corrections to the dc conductance depends on the microscopic details of the scatterer @xcite . + for the simplest model of an impurity inducing bs in an interacting 1d channel , kane and fisher found @xcite that repulsive electron interactions enhance bs whereas attractive interactions suppress it . in the latter case ,
the impurity potential is screened and the channel becomes perfectly transparent as the temperature @xmath2 .
for repulsive interactions , the channel is effectively cut in two as @xmath3 by the renormalization to strong coupling of the bs potential .
many of kane and fisher s predictions have been confirmed numerically @xcite , analytically using the bethe ansatz @xcite , and experimentally using carbon nanotubes @xcite . for weak interactions , yue
, mateev and glazman @xcite obtained a complementary description by relying on the hartree - fock approximation : the friedel oscillations introduced by the impurity in the electron density modify the scattering potential , leading to a logarithmic singularity @xmath4 in the reflection coefficient .
the logarithm is just the first term in an infinite series which , upon resummation , yields the powerlaw conductance obtained by kane and fisher @xcite . in this article , we will address the question of under which conditions the 1d description studied by kane and fisher description applies to a backscatterer coupled to the edge states of a 2dti ( see fig.[sketch ] ) .
specifically , we show the latter scenario does not universally apply and provide an alternative low - energy model .
as shown below , the physics of an bs impurity in a 2dti is different due to the new features brought about by the dimensionality and electronic structure of the ti .
unlike earlier work @xcite , our focus is on the effective model describing the low - energy physics in the presence of interactions , which have dramatic effects in 1d systems @xcite . indeed , for a strong impurity potential in the bulk of a ti
, bound states appear within the band gap @xcite .
this also happens for large spin @xmath5 magnetic impurity ( or a magnetic cluster ) , at temperatures @xmath6 well above the kondo temperature @xmath7 ( @xmath7 is suppressed for large @xmath5 @xcite ) .
the impurity will be treated as a local zeeman field , whose direction , in order to maximize bs , we shall take perpendicular to the spin quantization axis : @xmath8 where @xmath9 creates one electron at lattice site @xmath10 . here , we are interested in the limit where @xmath11 is larger than the band gap .
the resulting problem is by no means of academic interest as the size of the band gap of 2dti can be made rather small @xcite , especially close to the topological transition . for an bs impurity in a qshi sufficiently far from the edge , the potential leads to two ( bound ) states within the gap . as the impurity moves from the bulk to the edge ,
the bound states resonate with the edge states .
thus , the bs impurity problem at the edge of a 2dti a problem is akin to a resonant level problem coupled to a 1d interacting channel @xcite . before describing the derivation of the above picture , it is worth mentioning that , besides a strongly interacting magnetic impurity at the edge of a qsh insulator , the model considered here also describes a strongly - coupled point contact in a quantum ( anomalous ) hall insulator ( cf .
fig .
[ sketch ] ) .
this system may be also realized in a point contact in a photonic topological material , e.g. ref . .
in order to describe the qshi , we shall use the kane - mele ( km ) model @xcite , which is defined on a honeycomb lattice by the hamiltonian : @xmath12 where @xmath13 describes the _ intrinsic _ so coupling @xcite as an imaginary next nearest neighbor hopping and @xmath14 depends on the electron hoping path ; @xmath15 is the electron spin projection on the axis perpendicular to the 2d plane . in the absence of interactions
( to be treated further below ) , the impurity problem is described by the hamiltonian @xmath16 ( cf .
eqs .
[ eq : bsi ] and [ hami - rx ] ) . in order to solve it
, we first obtain an analytical solution of the km model , eq . in a strip geometry of width @xmath17 with the zigzag edges ( fig.[shift ] ) . in this geometry , the bloch wavevector parallel to the edge , @xmath18 , is a good quantum number .
however , @xmath19 must be treated as an operator ( see below ) .
the wave functions along the @xmath20-axis obey open boundary conditions @xcite .
we first consider the fourier transform of eq . in the bloch basis , where : @xmath21 here @xmath22 is the position of @xmath23 sublattice sites and @xmath24 is total number of unit cells .
because of the bi - partite structure of the honeycomb lattice , the fourier transform of @xmath25 is not unique and depends on the relative phase @xmath26 . this gauge freedom must be fixed by the boundary conditions ( bcs ) .
the appropriate choice for the zigzag edge is @xmath27 , so that the @xmath28th row of the a sublattice overlaps with the @xmath28th row of the b sublattice ( see fig.[shift ] ) .
this maps the honeycomb lattice onto the so - called `` brick wall '' lattice and thus the bcs are @xmath29 at @xmath30 .
+ after identifying the boundary conditions , we proceed to solve the 1d schrdinger equation : @xmath31 where we have used the following notation : @xmath32 and @xmath33 , with @xmath34 , @xmath35 and @xmath36 , respectively ( @xmath37 ) .
in addition , since @xmath15 is a good quantum number , @xmath38 .
below , we look for solutions that are combinations of plane waves @xmath39 .
we are not interested in finite size effects and therefore we focus on the large @xmath17 limit .
thus , the coupling between the two edges vanishes and we obtain the dispersion for the edge states @xcite : @xmath40^{2}}},\ ] ] where the @xmath41 ( @xmath42 ) sign corresponds to the bottom ( top ) edge at @xmath43 ( @xmath44 ) and @xmath45 .
the bands of edge states cross at @xmath46 @xcite ( for a bearded edge they cross at @xmath47 @xcite ) . for @xmath48 , eq .
agrees with the results of ref . .
for the bottom edge states , below we use the notation @xmath49 .
a plot of the bands @xcite for a wide zigzag ribbon obtained by this method can be found in the supplementary material @xcite .
next , we consider the form of the eigenstates .
up to a phase factor depending on @xmath50 , they are a combination of four @xmath51 , where @xmath52 .
there are three kinds of states : i ) @xmath53 and @xmath54 , which is an eigenstate localized at the edge , ii ) @xmath53 and @xmath55 , a resonance state , which is extended in the bulk , and iii ) @xmath56 and @xmath57 a bulk scattering state @xcite . in order to investigate the effect of the impurity on the electronic transport
, we next solve the lippmann - schwinger equation ( lse ) @xmath58 , where @xmath59 is the green s function for eq . .
we assume the bs impurity to be located on the b sublattice at the bottom edge since the wavefunction of edge states on this edge is mostly localized on the b sublattice @xcite . in order to extract the transmission and reflection coefficient of the edge electrons ,
we assume the incident electron has a bloch wave number @xmath60 on the right - moving edge channel , i.e. , @xmath61 and therefore its energy is @xmath62 and group velocity is @xmath63 .
let us introduce @xmath64 and @xmath65 , where @xmath66 corresponds to @xmath67 .
thus , the asymptotic behavior of the wavefunction for @xmath68 becomes @xmath69 and @xmath70 , for @xmath71 , where @xmath72 , and @xmath73 here @xmath74 is the normalization length of system along the edge and @xmath75 is the impurity position . from the above results , the transmission and the reflection coefficients can be extracted as @xmath76 and @xmath77 , respectively .
the energy dependence of the transmission coefficient is shown in fig .
[ transi-2 ] .
note that , when the impurity is located on the first atomic row ( i.e. @xmath78 ) , the transmission is essentially independent on the energy , which makes it similar to a bs impurity in a purely 1d channel .
this behavior arises from the weak coupling induced by the impurity between the edge and bulk states ( owing to the small weight of the bulk states on the @xmath78 row ) , even for relatively large values of @xmath11 .
thus , scattering is dominated by the 1d edge states . however , for the second atomic row and beyond ( i.e. @xmath79 ) , the weight of the bulk - state wavefunctions becomes larger , thus increasing the coupling between bulk and edge states . for large values of @xmath11 ,
a pair of narrow scattering resonances appears within the gap . in the neighborhood of the anti - resonances
the transmission coefficient changes very rapidly with energy and ( on resonance ) vanishes for large @xmath11 . for an impurity on a b sublattice site on ( a ) the first atomic row
( i.e. , @xmath78 ) and ( b ) @xmath80.(the spin - orbit coupling is @xmath81 .
, title="fig:",width=3 ] + in order to understand the emergence of a pair of scattering anti - resonances , we need to consider the poles of the t - matrix : @xmath82^{-1 } \mathcal{v}_{\mathrm{imp}}$ ] . for a strong impurity potential located within the bulk ,
far from the edge , the impurity only couples to the bulk states . in this case , the poles of the t - matrix are obtained from the condition @xmath83=0 $ ] where @xmath84 is the green s function in the bulk .
the latter is real for @xmath85 within the energy gap since the density of states vanishes and it is odd in @xmath85 ( due to the particle - hole symmetry of @xmath25 ) , therefore vanishing at @xmath86 .
thus , @xmath87 for small @xmath85 .
hence , at large @xmath11 , two bound states , appear at @xmath88 corresponding to the two eigenvalues of @xmath89 .
as the impurity location approaches the edge , the bound states will hybridize with the continuum of edge states , leading to the anti - resonances in the transmission coefficient . in the following
, we shall construct a 1d low - energy effective model that describes the edge channels in the presence of the bs impurity at large @xmath90 , where @xmath91 is bulk band gap .
the effective model is valid at energies and temperatures smaller than @xmath92 and therefore only involves the edge degrees of freedom .
the construction of the effective low energy model is aided by using the known symmetries of the system hamiltonian , @xmath16 , as both share the same symmetries . to this end
, we note that the km model in the strip geometry , @xmath93 , is invariant under trs ( @xmath94 ) , spin rotations about the @xmath95-axis ( i.e. @xmath96 , @xmath97 ) , particle - hole symmetry ( @xmath98 ) , and translations by multiples of @xmath99 along the edge direction .
the impurity potential , @xmath100 , breaks all those symmetries , but the composite system described by @xmath101 is invariant under the subgroup span by the combined @xmath102 and @xmath103 transformations . therefore , according to the above discussion , the effective model takes the form of a fano model @xcite describing two discrete levels coupled to the continuum of edge states , which is invariant under @xmath102 and @xmath103 .
since for @xmath104 the position of the resonances approaches the center of the band gap at @xmath105 , we shall focus in the neighborhood of @xmath106 , where linearization the edge state spectrum , i.e. @xmath107 , is a good approximation .
thus , the effective hamiltonian reads @xmath108 + h_{-}\left[d , t_{-}\psi(0)\right],\label{eq : hameff}\\ h_b & = iv_f \int dx\ : \psi^{\dagger } s^{z } \partial_x\psi + v_{b}a_{0}\psi^{\dagger}(0)s^{x}\psi(0 ) , \label{eq : hamb}\\ h_{\pm}&[f,\chi ] = \pm \epsilon_0\ : \left ( f^{\dag } f -\tfrac{1}{2}\right ) + v_{c}a^{1/2}_{0 } \left [ f^{\dagger } \chi + \mathrm{h.c . } \right ] \ ] ] where @xmath109 is the spinor field operator describing the edge states , @xmath110 and @xmath111 are the creation operators of electrons in the bound states with @xmath89 eigenvalue and energy @xmath112 and @xmath113 , respectively , and @xmath114 ; @xmath115 is a short distance cut - off . in the above model , @xmath116 describes a renormalized bs amplitude for the edge electrons , and @xmath117 the tunneling into and out of the bound states .
the reflection coefficient for the effective model reads : @xmath118 we have fitted the numerical results for the impurity system interacting with the edge of a 2dti to this equation , as shown on the left panel of fig .
[ transmissions-2 ] for @xmath119 .
the behavior of the fitted @xmath120 , @xmath121 , and @xmath122 as a function of @xmath11 , is shown in the right panel .
as expected from the above discussion , @xmath122 decreases as @xmath123 .
note also that @xmath124 . in passing ,
we note that for @xmath125 , the resonances disappear and the the reflection coefficient becomes independent of the energy , which applies to an impurity located on the first atomic row ( @xmath78 ) .
( @xmath126 is the band gap ) .
dots are the transmission coefficient obtained numerically for the kane - mele model with a backscatterer at the edge . the red line is the fit to the effective model ( cf .
. [ eq : hameff ] ) .
( right ) effective model parameters as a function of @xmath127.,title="fig:",width=1 ] ( @xmath126 is the band gap ) .
dots are the transmission coefficient obtained numerically for the kane - mele model with a backscatterer at the edge . the red line is the fit to the effective model ( cf .
. [ eq : hameff ] ) .
( right ) effective model parameters as a function of @xmath127.,title="fig:",width=1 ] finally , we study the effect of electron interactions on the above picture .
interactions will be treated non - perturbatively using the bosonization method @xcite , and their characteristic energy scale @xmath128 ( where @xmath129 is the electron charge ) is assumed to be smaller than @xmath126 .
before considering interaction effects , we further project the effective 1d model in eq .
must onto the subspace of excitations with energy @xmath130 ( @xmath131 is fermi energy ) . in particular , when @xmath131 is away from the resonances at @xmath132 , the bound states can be integrated out . to leading order , this yields a renormalized backscattering amplitude : @xmath133 .
\label{eq : jbeff}\ ] ] hence , the model reduces to the one studied by kane and fisher @xcite ( see @xmath134 in eq .
[ eq : hamkf ] below ) , with the bs amplitude @xmath116 replaced by @xmath135 from eq . . on the other hand , for @xmath136 or @xmath137 , the resonant level can not be immediately integrated out , although the bound state further away can . assuming ( without loss of generality ) that @xmath138 yields the following low - energy effective model : @xmath139 + \left ( d^{\dag}d - \tfrac{1}{2}\right ) \notag\\ & \qquad\qquad \times \left [ u_f
\ : \psi^{\dag}(0)\psi(0 ) + u_b \ : \psi^{\dag}(0)s^x \psi(0 ) \right ] , \label{eq : hameff2 } \\ & h_{kf } = h_b + u \int dx\ : \rho_r \rho_l .
\label{eq : hamkf}\end{aligned}\ ] ] here we have also accounted for interactions between the edge electrons ( with amplitude @xmath140 ) and with the resonant level electron ( with amplitudes @xmath141 and @xmath142 ) .
furthermore , integrating out the non - resonant level at @xmath143 modifies the bs potential @xmath144 in @xmath145 by an amount @xmath146 .
forward scattering is also generated but it has been dropped since it can be eliminated by a unitary transformation @xcite .
the hamiltonian in @xmath145 in eq .
is very similar to a model of side - coupled level to an interacting 1d channel , which has been studied by goldstein and berkovits in ref .
@xcite . following these authors , after bosonizing @xcite eq .
, we apply a unitary transformation and eliminate the forward interaction term @xmath147 , at the expense of renormalizing the scaling dimension ( @xmath148 ) of the tunneling , @xmath149 , where @xmath150 , where @xmath151 is the imaginary time and @xmath152 $ ] , where @xmath153 is the luttinger parameter and @xmath154 the velocity of the edge plasmons @xcite .
tunneling into the resonant state becomes relevant in the renormalization - group ( rg ) sense when @xmath155 .
thus , for weak to moderate repulsion ( i.e @xmath156 ) , both tunneling @xmath117 and the bs ( @xmath157 ) renormalize to strong coupling and suppress the transmission across the impurity , as confirmed by the rg flows to 2nd order @xcite .
interestingly , for moderate attraction ( i.e. @xmath158 ) , for which bs is _ navely _ irrelevant @xcite , we have @xmath159 .
thus , tunneling flows to strong coupling and 2nd order rg flows @xcite show that the runaway flow of @xmath117 drags along @xmath142 and @xmath141 , which leads to the narrowing of the anti - resonance in the transmission and suppression of conductance as @xmath3 , also for weak to moderate attraction in the edge channel . in summary
, we have investigated the problem of an impurity inducing backscattering at the edge of a two - dimensional topological insulator ( ti ) . by obtaining a solution of the scattering problem ,
we have derived the 1d effective low - energy hamiltonian describing the system .
strong backscattering induces two discrete levels within the band gap , which couple to the ( interacting ) metallic edge(s ) . when the fermi energy resonates with one of the levels , both moderately attractive or repulsive interactions
suppress the transmission through the impurity . off resonance , the system can be effectively described as an impurity in a tomonaga - luttinger liquid @xcite with a renormalized backscattering potential .
this picture should be robust in the presence of weak uniform rashba spin - orbit , which will introduce an energy dependence in the couplings of the effective model , eq .
, arising from the spin texture of the bulk and edge states . however , energy dependence is an irrelevant perturbation at low energies . besides this , other model regimes and extensions to this work , such as multiple impurity problems , etc . will be explored elsewhere @xcite .
we thank l. glazman , f. guinea , y .- h .
ho , c .-
l .
huang , and x .- p .
zhang for useful discussions .
m.a.c . gratefully acknowledges support by the ministry of science and technology ( taiwan ) under contract number nsc 102- 2112-m-007 - 024-my5 , and taiwan s national center of theoretical sciences ( ncts ) .
99 m. z. hasan and c. l. kane , rev .
mod .
phys . * 82 * , 3045 ( 2010 ) ; x .-
l .
qi and s .- c .
zhang , rev .
mod .
phys . * 83 * , 1057 ( 2011 ) . c. wu , b. a. bernevig , and s. c. zhang , phys .
rev .
lett .
* 96 * , 106401 ( 2006 ) . c. xu and j. e. moore , phys .
rev .
b * 73 * , 045322 ( 2006 ) .
g. dolcetto , m. sassetti , t .- l .
schmidt , nuovo cimento * 39 * , 113 ( 2016 ) .
a. strm , h. johannesson , and g. i. japaridze , phys .
rev .
lett . * 104 * , 256804 ( 2010 ) ; f. crpin , j. c. budich , f. dolcini , p. recher , and b. trauzettel , phys . rev .
b * 86 * , 121106 ( 2012 ) ; f. geissler , f. crpin , and b. trauzettel , phys . rev .
b * 89 * , 235136 ( 2014 ) .
m. knig _
et al . _ ,
science * 318 * , 766 ( 2007 ) ; a. roth _
et al . _ , science * 325 * , 294 ( 2009 ) ; k. c. nowack _ et al . _ ,
nat .
mater .
* 12 * , 787 ( 2013 ) ; g. grabecki _
et al . _ ,
phys .
rev .
b * 88 * , 165309 ( 2013 ) ; g. m. gusev , _ et al .
_ , phys .
rev .
b * 89 * , 125305 ( 2014 ) ; i. knez , r .-
r .
du , and g. sullivan , phys .
rev .
lett . * 107 * , 136603 ( 2011 ) ; k. suzuki , y. harada , k. onomitsu , and k. muraki , phys .
rev .
b * 87 * , 235311 ( 2013 ) ; i. knez _
et al . _ ,
phys .
rev .
lett . * 112 * , 026602 ( 2014 ) ; e. m. spanton _
et al . _
phys .
rev .
lett . * 113 * , 026804 ( 2014 ) ; a. f. young , j. d. sanchez - yamagishi , b. hunt , s. h. choi , k. watanabe , t. taniguchi , r. c. ashoori , and p. jarillo - herrero , nature * 505 * , 528 ( 2014 ) .
t. l. schmidt , s. rachel , f. von oppen , and l. i. glazman , phys .
rev .
lett . * 108 * , 156402 ( 2012 ) ; j. c. budich , f. dolcini , p. recher , and b. trauzettel , phys .
rev .
lett . * 108 * , 086602 ( 2012 ) ; n. lezmy , y. oreg , and m. berkooz , phys .
rev .
b * 85 * , 235304 ( 2012 ) ; j. i. vyrynen , m. goldstein , and l. i. glazman , phys . rev
. lett . * 110 * , 216402 ( 2013 ) ; j. i. vyrynen , m. goldstein , y. gefen , and l. i. glazman , phys . rev .
b * 90 * , 115309 ( 2014 ) ; n. kainaris , i. v. gornyi , s. t. carr , and a. d. mirlin phys . rev .
b * 90 * , 075118 ( 2014 ) .
y. tanaka , a. furusaki , and k. a. matveev , phys .
rev .
lett . * 106 * , 236402 ( 2011 ) .
l. kimme , b. rosenow , and a. brataas , phys . rev .
b 93 , 081301 ( 2016 ) .
e. eriksson , a. strm , g. sharma , and h. johannesson phys .
rev .
b * 86 * , 161103(r ) ( 2012 ) .
j. maciejko , c. liu , y. oreg , x .-
l .
qi , c. wu , s .- c .
zhang , phys .
rev .
lett . * 102 * , 256803 ( 2009 ) . c. l. kane and m. p. a. fisher , phys .
rev .
lett . * 68 * , 1220 ( 1992 ) ; phys .
rev .
b * 46 * , 15233 ( 1992 ) . m. p. a. fisher and l. glazman in mesoscopic electron transport " , edited by l. kowenhoven , g. schn and l. sohn , nato asi series e , kluwer academic publishers ( dordrecht , the netherlands ) ( 1997 )
. l. i. glazman , i. m. ruzin , and b. i. shklovskii , phys .
rev .
b * 45 * , 8454 ( 1992 ) . s .-
j .
qin , m. fabrizio , and l. yu phys .
rev .
b * 54 * , r9643(r ) ( 1996 ) . c. rylands and n. andrei , phys . rev .
b * 94 * , 115142 ( 2016 ) .
z. yao , h. w. ch .
postma , l. balents , and c. dekker , nature ( london ) nature 402 , 273 - 276 ( 1999 ) .
d. yue , l. i. glazman and k. a. matveev , phys .
rev .
b * 49 * , 1966 ( 1994 ) . j. s. van dyke and d. k. morr phys . rev .
b * 93 * , 081401(r ) ( 2016 ) ; report arxiv:1609.01726 ( 2016 ) . a. o. gogolin , a. a. nersesyan , a. m. tsvelik , _ bosonization and strongly correlated systems _ cambridge univeristy press ( cambridge , uk 1999 ) . t. giamarchi , _ quantum physics in one - dimension _ , clarendon press ( oxford , uk 2004 ) .
r. r. biswas , and a. v. balatsky , phys .
rev .
b * 81 * , 233405 ( 2010 ) ; w .- y.shan , j. lu , h .- z .
lu , and s .- q .
shen , phys .
rev .
b * 84 * , 035307 ( 2011 )
. j. r. schieffer , journal of applied physics 38 , 1143 ( 1967 ) .
a. furusaki , phys .
rev .
b * 57 * , 7141 ( 1998 ) .
d. g. polyakov , and i. v. gornyi , phys .
rev .
b , * 68 * , 035421 ( 2003 ) ; y. v nazarov and l. i. glazman , phys . rev .
lett . * 91 * , 126804 ( 2003 ) .
i. v. lerner , v. i. yudson , and i. v. yurkevich , phys .
rev .
lett .
100 , 256805 ( 2008 ) .
m. goldstein , and r. berkovits , phys .
rev .
lett . *
104 * ,
106403 ( 2010 ) . z. wang , y. d. chong j. d. joannopoulos , and m. soljai , nature * 461 * , * 772 * , ( 2009 ) . c. l. kane and e. j. mele , phys .
rev .
lett . * 95 * , 226801 ( 2005 ) ; phys .
rev .
lett . * 95 * , 146802 ( 2005 ) .
b. zhou _ et al .
_ , phys .
rev .
lett .
* 101 * , 246807 ( 2008 ) .
see supplementary material .
g. zhang _
et al . _ , nanoscale , * 6 * , 3259 ( 2014 ) .
h. doh , g. s. jeon , and h. j. choi , arxiv:1408.4507 ( 2014 ) .
u. fano , phys . rev . * 124 * , pp . 1866 ( 1961 ) .
j. h. zheng and m. a. cazalilla , _ to be published_. * supplementary material for strong backscatterer at the edge of a two - dimensional topological insulator " * |
supersymmetry ( susy ) is one of the promising candidates of the physics beyond the standard model .
no signature of susy has been found yet . however , the discovery of the susy particles is guaranteed for @xmath8 tev at the cern large hadron collider ( lhc ) for the minimal supergravity ( msugra ) model . date taking is expected to start from 2007 .
masses of the sparticles will also be measured at the lhc @xcite with reasonable accuracy , which is important to distinguish various susy breaking models .
the lagrangian of supersymmetric theory is highly constrained .
for example the sfermion - fermion - gaugino interaction is restricted to be of the form @xmath9-@xmath10-@xmath11 because @xmath12 and @xmath13 belong to the same chiral multiplet in supersymmetry . on the other hand , @xmath14 and @xmath15 mixing terms
are controlled by the @xmath16 term of the superpotential of the form @xmath17 .
therefore the mixing term is proportional to the yukawa coupling @xmath18 and @xmath6 parameter .
studying such mixing terms determined by the supersymmetry would be an interesting target for collider physics after the discovery of the supersymmetry .
in this paper we study the effect of the neutralino polarization in the decay chain @xmath19 here @xmath20 denotes the lepton from @xmath21 decay . due to the chirality structure of the squark - quark - neutralino coupling , the @xmath22
is polarized .
the polarization of @xmath22 then affects the angular distribution of the slepton in the @xmath23 decay .
the polarization dependence of the angular distribution eventually shows up in the charge asymmetry in the @xmath24 distribution , because the polarization dependent part of the amplitude flips under the charge conjugation transformation .
the effect can be seen at the @xmath25 collider lhc because the number of squarks @xmath26 produced at the lhc is larger than @xmath27 .
the polarization effect is considered in refs .
@xcite , where the slepton is assumed to be purely right - handed .
we study this process taking account of the left - right mixing of the sleptons , which depends on the lepton flavor .
we point out that the left - right mixing significantly affects the charge asymmetry as well as the decay width of @xmath23 for @xmath28 and @xmath29 .
this paper is organized as follows . in the next section ,
we first summarize the relevant formulae and discuss the effect of the left - right mixing qualitatively , then we evaluate the lepton flavor dependence of the charge asymmetry and the decay widths for explicit examples . in sec .
[ sec : simulation ] , we carry out the monte carlo ( mc ) study and discuss the feasibility to study the chiral effect at the lhc .
[ sec : discussion ] is devoted for discussions .
we parameterize the sfermion - fermion - neutralino interaction lagrangian as @xmath30 where @xmath31 and @xmath32 are the suffices for the mass eigenstates of the neutralinos and the sfermions , respectively .
@xmath33 is the su(2 ) gauge coupling constant . for the charged lepton sector ,
the coupling constants @xmath34 and @xmath35 are given as @xmath36 \cos\theta_l + \frac { m_l } { m_w \cos\beta } ( u_n^*)_{i3 } \sin\theta_l , \label{eq : l1 } \\ l_{i2}^{l } & = & \left [ ( u_n^*)_{i2 } + ( u_n^*)_{i1}\tan\theta_w \right ] \sin\theta_l + \frac { m_l } { m_w \cos\beta } ( u_n^*)_{i3 } \cos\theta_l , \label{eq : l2 } \\ r_{i1}^{l } & = & 2(u_n)_{i1 } \tan\theta_w \sin\theta_l , + \frac { m_l } { m_w \cos\beta } ( u_n)_{i3 } \cos\theta_l \label{eq : r1 } \\ r_{i2}^{l } & = & 2(u_n)_{i1 } \tan\theta_w \cos\theta_l , - \frac { m_l } { m_w \cos\beta } ( u_n)_{i3 } \sin\theta_l \label{eq : r2}\end{aligned}\ ] ] where @xmath37 is the unitary matrix which diagonalize the neutralino mass matrix .
the left - right mixing angle of the sleptons @xmath38 is defined as follows .
the slepton mass term is written as @xmath39 where the mass squared matrix @xmath40 is given as @xmath41 in this paper , we neglect lepton flavor mixing and assume that @xmath42 and @xmath6 are real . then @xmath38 is obtained by diagonalizing the slepton mass matrix : @xmath43 where we assume @xmath44 .
the mass eigenstates @xmath45 are defined as @xmath46 the coupling constants for the squark sector are obtained in the same way .
the decay widths of the subprocesses @xmath47 and @xmath48 are written as @xmath49 here and hereafter , we neglect the quark and lepton masses for the kinematics , while keeping them in the coupling constants eqs .
( [ eq : l1])([eq : r2 ] ) .
the momentum configuration of the whole decay chain @xmath50 is described by three angular variables .
the distribution of the decay chain is given as @xmath51 , \label{eq : dist } \\
a(l ) & = & \frac { |l_{2\beta}^{q}|^2 - |r_{2\beta}^{q}|^2 } { |l_{2\beta}^{q}|^2 + |r_{2\beta}^{q}|^2 } \cdot \frac { |l_{2\alpha}^{l}|^2 - |r_{2\alpha}^{l}|^2 } { |l_{2\alpha}^{l}|^2 + |r_{2\alpha}^{l}|^2 } , \label{eq : asym}\end{aligned}\ ] ] where @xmath52 denotes the branching ratio @xmath53 , _
etc_. @xmath54 is the angle between the momenta of the quark and the lepton @xmath55 in the @xmath22 rest frame , @xmath56 is the angle between the two lepton momenta in the slepton rest frame , and @xmath57 is the angle between the decay planes of @xmath58 and @xmath59 . since squark and slepton decays are spherically symmetric in the rest frames of the decaying particles , the angular distribution is flat over @xmath60 and @xmath57 .
the @xmath54 dependence comes from the polarization of @xmath22 and the chirality structure of the sfermion - fermion - neutralino interaction shows up in this angular distribution . qualitatively
if @xmath61 decays into @xmath22 followed by @xmath62 , the lepton favors going in the same ( opposite ) direction to @xmath22 for @xmath63 . instead of the above angular variables , we consider the invariant masses of the combinations of the quark and the leptons , which are directly connected as the observables .
the relations are given as @xmath64.\end{aligned}\ ] ] the polarization effect of @xmath22 can be seen clearly in the charge asymmetry in the @xmath65 distribution , since an @xmath66 is the coefficient of @xmath54 , and it affects the decay distributions for @xmath67 and @xmath68 with opposite signs
. however , the leptons @xmath55 of the neutralino decay are indistinguishable from @xmath69 of the slepton decay .
we therefore study the charge asymmetry between the @xmath70 distributions taking both @xmath55 and @xmath69 into account .
the @xmath71 distribution is the sum of the @xmath72 distribution from the decay chain @xmath73 and the @xmath74 distribution from @xmath75 .
the charge asymmetry in @xmath70 distribution was studied in ref .
@xcite in a case where the slepton is purely right - handed : @xmath76 . in this paper
we take account of the left - right mixing of the sleptons , whose effect depends on the lepton flavor .
we consider a `` typical '' case of the minimal supergravity model .
the su(2 ) and u(1 ) gaugino masses @xmath77 and @xmath78 , respectively , are related to each other as @xmath79 and the magnitude of the higgsino mass @xmath6 is assumed to be larger than @xmath77 , so that the wino component dominates @xmath22 and the bino component dominates @xmath80 . as for the slepton mass matrix , the right - handed slepton mass parameter becomes smaller than the left - handed one due to the running effect .
therefore the lighter slepton , @xmath81 , is @xmath3-like and the heavier one , @xmath82 , is @xmath2-like in the most of the parameter space .
we can safely neglect the left - right mixing of the squarks because the process we consider is the decay of the first generation squark .
the first decay process @xmath83 in eq .
( [ eq : decay ] ) occurs predominantly through the @xmath61-@xmath84-@xmath85 coupling if @xmath86 .
consequently the first factor of the right - hand side of eq .
( [ eq : asym ] ) is very close to unity .
let us now consider the decay process @xmath87 with @xmath88 , @xmath6 , or @xmath29 .
we first note that , in the minimal supergravity model , the slepton left - right mixing angle @xmath38 is approximately written as @xmath89 where @xmath90 and @xmath91 are the universal scalar mass and the gaugino mass at the gut scale , respectively , and @xmath42 is the trilinear scalar coupling constant at the low energy scale .
the numerical factor 0.37 in the denominator is determined by the gauge coupling constants . in the ( s)electron case , the left - right mixing of the selectrons is negligible , _
i. e. _ , @xmath92 for @xmath88 in eqs .
( [ eq : l1])([eq : r2 ] ) , and the charge asymmetry is maximal ( @xmath93 ) .
the effect of the left - right mixing and the yukawa coupling is quite significant for the ( s)tau mode . for @xmath94 , the typical magnitude of @xmath95
is @xmath96 for large @xmath4 in msugra .
magnitudes of the neutralino mixing matrix elements are typically @xmath97 and @xmath98 . substituting these numbers in eqs .
( [ eq : l1 ] ) and ( [ eq : r1 ] ) , we see that @xmath99 can dominate over @xmath100 and the behavior of the charge asymmetry is opposite to the electron case .
the left - right mixing effect may be observed even in the ( s)muon case . since both terms of the right - hand side of eq .
( [ eq : l1 ] ) are approximately proportional to the lepton mass , @xmath101 is enhanced by @xmath102 compared to @xmath103 .
for relatively large @xmath4 , it is possible that @xmath101 and @xmath104 are of the same order of magnitude , as we will see later .
notice that the branching ratio of @xmath105 also depends on the lepton flavor .
the coupling @xmath106 is suppressed due to the small @xmath107 component , therefore even tiny mixing of the left - hand component will have a significant effect .
especially , the decay width for @xmath108 mode is much larger than those for @xmath109 and @xmath110 , and @xmath111 .
we also consider the case where @xmath112 . in this case , both @xmath81 and @xmath82 contribute to the decay chain so that the decay distribution over @xmath70 is more complicated . since @xmath82 is @xmath2-like , the coupling @xmath113 in eq .
( [ eq : l2 ] ) dominates the @xmath114 decay and the branching ratio of @xmath114 is much larger than that of @xmath115 .
the main contribution of the charge asymmetry therefore comes from @xmath82 .
let us study the lepton flavor dependence more quantitatively .
in order to compare with the monte carlo simulations in the next section , we take benchmark points sps1a and sps3 given in ref .
@xcite , where the mass spectrum is calculated by isajet @xcite .
the model points we study are as follows ; * sps1 @xcite : + this is the msugra point where @xmath116 gev , @xmath117 gev , the trilinear coupling at the gut scale @xmath118 gev , @xmath119 and @xmath120 .
the decay @xmath121 is open .
we also show some results for the point where @xmath90 and @xmath91 are the same as those of sps1a but @xmath122 and 20 . for these points ,
the effect of @xmath123 left - right mixing in @xmath22 decay becomes significant .
* sps3 ( or point c of @xcite ) : + @xmath124 gev , @xmath125 gev , @xmath126 , @xmath119 and @xmath120 .
the decay @xmath127 is open for this point .
the decay must show opposite charge asymmetry to that of @xmath128 .
the kinematics of the decay distribution is important for our study .
we therefore list the calculated end points for the decay ( [ eq : decay ] ) in table [ list ] . .
the sparticle masses and end points of @xmath129 distributions arising from the cascade decay @xmath130 at sps1a and sps3 .
the masses and end points for sps1a(@xmath122 , 20 ) are similar to that for @xmath119 for the first and second generation sfermions .
all masses are given in gev .
[ cols="^,^,^,^,^,^,^,^,^",options="header " , ] when @xmath123 left - right mixing is taken into account , @xmath131 and @xmath132 for @xmath133 ( table [ tab : tanbeta ] ) .
the statistical significance of the deviation of the branching ratio @xmath134/$]@xmath135 $ ] is around 8.5 for 300 fb@xmath136 .
the significance of the deviation of charge asymmetry @xmath137 is only around 1 for 300 fb@xmath136 .
the statistical significance at the lhc may be compared with that at the future linear collider ( lc ) at @xmath138 gev , where @xmath139 cross sections is @xmath140 for a left handed beam @xcite .
there are 400 @xmath22@xmath22 events followed by @xmath141 for @xmath142 at sps1a(@xmath133 ) . assuming no background
, the deviation of branching ratios around 5% may be detectable .
finally we discuss the decay distribution of @xmath143 .
ideally the signal distribution would show charge asymmetry near the end point as can be seen in fig .
[ asymmetry ] c ) .
however , as a @xmath29 lepton decays into a meson and @xmath144 or @xmath145 , one can not measure the @xmath29 lepton momentum directly .
when @xmath29 decays into leptons , the arising charged lepton is soft , and they typically have less than 1/3 of the @xmath29 lepton energy .
such large smearing of the energy is not suitable for the charge asymmetry study . on the other hand ,
@xmath29 decay into a meson is a two body decay , and the meson might be detected as a low multiplicity isolated high @xmath146 jet .
a qcd jet might also be mis - identified as a @xmath29 jet , and the probability is the function of @xmath29 tagging efficiency @xmath147 and @xmath29 jet @xmath146 .
qcd events become large background , however they can be removed by subtracting the sum of the distribution of @xmath148 and @xmath149 from that of @xmath150 .
the identification of the hadronic @xmath29 decays has a strong dependence on the details of the detector performance . in the detector simulator atlfast ,
the detection efficiency of the @xmath29 jet and the rate of fake @xmath29 jet is estimated by using parameterizations of the full simulation data , which are described in ref .
however , the misidentification probability of the charge of the @xmath29 jet and the charge distribution of fake @xmath29 jets have not been fully accounted for in the parameterized simulation .
we therefore do not study the charge asymmetry in @xmath29 jets in this paper .
in this paper , we study if susy study at the lhc can provide an information on the chiral nature of the sfermions .
we especially study the charge asymmetry of @xmath151 and @xmath152 distribution which comes from the the decay cascade @xmath153 @xmath154 , where @xmath155 can be either @xmath3 or @xmath2 . when @xmath22 is wino like , the asymmetry @xmath156 is negative near the @xmath0 end point for pure @xmath2 , while it is positive for pure @xmath3 .
we find that the charge asymmetry is significant for two representative msugra points , sps1a and sps3 , where slepton masses are lighter than @xmath157 . for @xmath158 , the branching ratio and @xmath159 distributions may depend on the relation among the left and right couplings of @xmath160 interactions .
the branching ratio and charge asymmetry are proportional to @xmath161 and @xmath162 , respectively .
when @xmath22 is wino like , @xmath101 is proportional to small left - right mixing of @xmath123 , and @xmath163 , while @xmath104 is proportional to the small bino component @xmath107 of @xmath22 . for large @xmath4
, @xmath164 becomes comparable to @xmath107 in msugra points , leading to the non - universalities in @xmath165 and @xmath66 .
we have shown that the lhc can detect @xmath5-@xmath6 non - universality at sps1a(@xmath119-@xmath166 ) for @xmath167 fb@xmath136 if detection efficiencies of @xmath5 and @xmath6 are understood at the lhc .
the discovery of supersymmetry and determination of the low energy lagrangian are the goal of susy study at future colliders .
so far , many physics studies have focused on the determination of susy parameters at benchmark points . on the other hand , the charge asymmetry in the @xmath159 distribution provides an information on the chiral structure of the mssm , which is constrained by nature of supersymmetry .
especially @xmath123 and @xmath168 non - universality is a direct evidence of @xmath169-@xmath170 mixing due to the @xmath16 term .
i. hinchliffe , f. e. paige , m. d. shapiro , j. soderqvist and w. yao , phys .
d * 55 * , 5520 ( 1997 ) [ arxiv : hep - ph/9610544 ] ; s. abdullin _ et al .
_ [ cms collaboration ] , j. phys .
g * 28 * , 469 ( 2002 ) [ arxiv : hep - ph/9806366 ] ; atlas collaboration , cern - lhcc-99 - 14/15 ; h. bachacou , i. hinchliffe and f. e. paige , phys
. rev .
d * 62 * , 015009 ( 2000 ) [ arxiv : hep - ph/9907518 ] ; i. hinchliffe and f. e. paige , phys .
d * 61 * , 095011 ( 2000 ) [ arxiv : hep - ph/9907519 ] ; b. c. allanach , c. g. lester , m. a. parker and b. r. webber , jhep * 0009 * , 004 ( 2000 ) [ arxiv : hep - ph/0007009 ] .
b. c. allanach _ et al .
_ , in _ proc . of the aps / dpf / dpb summer study on the future of particle physics ( snowmass 2001 ) _ ed
n. graf , eur .
j. c * 25 * , 113 ( 2002 ) [ econf * c010630 * , p125 ( 2001 ) ] [ arxiv : hep - ph/0202233 ] . | in supersymmetric theory , the sfermion - fermion - gaugino interactions conserve the chirality of ( s)fermions .
the effect appears as the charge asymmetry in @xmath0 distributions at the cern large hadron collider where jets and leptons arise from the cascade decay @xmath1 .
furthermore , the decay branching ratios and the charge asymmetries in @xmath0 distributions are flavor non - universal due to the @xmath2 and @xmath3 mixing . when @xmath4 is large , the non - universality between @xmath5 and @xmath6 becomes @xmath7 level .
we perform a monte carlo simulation for some minimal supergravity benchmark points to demonstrate the detectability . |
the recent , first , direct detection @xcite of gravitational waves ( gw ) from a binary black hole merger by advanced ligo @xcite has opened a new window for the investigation of astrophysical compact objects .
the new generation gw detectors advanced ligo and advanced virgo @xcite are also expected to reveal an incoming gravitational transient signal from binary neutron stars ( bns ) coalescence and merger , once their sensitivity at higher frequencies will increase . at design sensitivity , the rate of bns signals detected is predicted to be in the interval ( 0.2 - 200 ) per year @xcite , making them the next target for gw detection .
these expected detections present a unique way to learn about the physics of matter at the extreme conditions present in neutron stars and the eos of nuclear matter above the nuclear density .
fully general relativistic simulations of bns started in 1999 @xcite but it is since the crucial breakthroughs of 2005 @xcite that numerical relativity is the main instrument to study the dynamics of the merger of compact objects .
this has lead to the development of community driven public software like the einstein toolkit @xcite and the lorene library @xcite that allow to openly simulate such systems and in particular bns coalescence and merger @xcite .
simulation of bns mergers , where numerical relativity is almost the only available tool , is one of the main endeavors to study the possible gravitational waveforms that will be discovered by the present detectors .
for this reason there is now a considerable effort ( using both public and private codes ) in analyzing the effect of the eos used to describe matter on the gravitational - wave signal ( see @xcite and reference therein ) , and to explore other directions like incorporating in the codes more microphysical ingredients , such as finite temperature nuclear eoss with neutrino emission @xcite and magnetic fields @xcite , to accurately simulate the physics of the post - merger remnant ( either a ( hyper)massive neutron star ( hmns ) or a black hole ( bh ) surrounded by an accretion disk ) .
however , despite all the progress , numerical modeling of bnss should still be considered an exploratory study due to a number of not completely settled issues .
some of those issues are : the techniques for initial data generation and the determination of their accuracy ; the evolution code convergence properties and the selection of the optimal numerical resolution ; and , more importantly , the choice of an appropriate modeling for the neutron star eos , still unknown from the microphysical and the experimental points of view .
this is reflected by the fact that there is also a lot of effort in trying to develop analytical techniques to compute the gravitational - wave signal ( to be compared and validated by numerical relativity ones ) from the inspiral phase of compact binary mergers @xcite , among which the effective one body ( eob ) @xcite approach was particularly successful .
recently , tidal effects have been included in those analytical models @xcite . from the measurements of tidal effects in the inspiral gravitational signal the eos of the neutron star core
can be distinguished @xcite . to contribute to this task ,
many - orbits ( > 10 ) bns simulations with nuclear eoss have been performed lately , comparing their gravitational signals to various eob formulations with tidal effects and taylor - expanded post - newtonian ( pn ) expressions @xcite .
the latter is challenging , because most of the initial data used are eccentric , and the number of orbits used are not , in any sense , close to the @xmath5 175 orbits that were used in similar studies of binary black hole systems @xcite .
the present work improves the results of @xcite focusing on the techniques to extract the gravitational - wave signal from numerical simulations ; the analysis of the eccentricity of the orbits obtained evolving quasi - circular initial data and its impact on the gravitational waveforms ; the dependence of results on the number of simulated pre - merger orbits , and different neutron star eoss . in detail
, we present new long - term ( up to 16 orbits ) equal - mass bns simulations with four different nuclear eoss , starting with four different values of the interbinary distance @xmath6 ( @xmath7 , @xmath8 , @xmath9 , and @xmath2 km ) .
the comparison of simulations starting from different initial orbital frequencies is a necessary and fundamental test to evaluate the accuracy of current numerical bns simulations and their ability to model tidal effects . in particular ,
long numerical simulations are needed to construct hybrid analytical - numerical waveforms for gw detectors data analysis where it is important to know how many orbits before merger can be effectively simulated using current numerical methods and resolutions .
the outline of this paper is the following : in section [ sec : setup ] we describe the numerical methods employed in our code and the initial models we choose to evolve . in section [ sec : gw ]
we describe the techniques we tested for extracting , integrating and extrapolating to null infinity the gravitational - wave signal . in section [ sec : eccentricity ]
we briefly comment on the eccentricity of the orbits from the evolution of quasi - circular initial data .
in section [ sec : comparison ] we present our results about the comparison of simulations of the same model starting with different initial frequencies .
finally , in section [ sec : conclusion ] we summarize the main results of the present work . throughout this paper
we use a space - like signature @xmath10 , with greek indices running from 0 to 3 , latin indices from 1 to 3 , and the standard convention for summation over repeated indices .
the computations are performed using the standard @xmath11 split into ( usually ) space - like coordinates @xmath12 and a time - like coordinate @xmath13 .
our coordinate system @xmath14 ( far from the origin ) has , as it can be checked , almost isotropic coordinates and far from the origin they would have the usual measure unit of `` time '' and `` space '' .
in particular , @xmath13 is time when measured from an observer at infinity .
all computations have been done in normalized computational units ( hereafter denoted as cu ) in which @xmath15 .
we report the radius of the spheres used for gravitational waves extraction in cus .
we used the exact same code setup as our previous study about binary neutron star mergers @xcite , which is why we refer the interest reader to that work and the references within for details of the setup . here
we only summarize for clarity the most important components and numerical algorithms .
a fundamental aspect for us is that this work has been produced using only free , open source software and as such can be reproduced and extended by anyone in the scientific community .
the dynamical evolution of the analyzed models was computed using the einstein toolkit @xcite .
the einstein toolkit is a free , publicly available , community - driven general relativistic ( gr ) code based on the ` cactus ` @xcite computational framework . in particular , we have chosen the eleventh release ( code name `` hilbert '' , et_2015_05 ) .
some local modifications and additions were necessary , all of which are open - source and freely available from the subversion server of the parma university gravity group @xcite , and all of which are planned to be proposed for the next release of the einstein toolkit .
the data are evolved on a cartesian mesh with 6 levels of mesh refinement , each with a double resolution respect to the parent level .
the outer boundary of the grid is set at @xmath16 km from its center , to be able to extract the gravitational signal far from the source .
the standard grid spacing on the innermost level used for this paper is @xmath17 m , which in @xcite we found sufficient to get qualitatively good results .
we used a mirror symmetry across the @xmath18 plane consistent with the symmetry of the problem , in order to reduce the computational costs by a factor two . the grid structure and the code parallelization
are handled by the ` carpet ` code @xcite .
the spacetime metric evolution is performed by the ` mclachlan ` module @xcite , implementing a @xmath11 dimensional split of the einstein equations using the bssn - nok formalism @xcite and solving them with fourth - order finite differences .
the general relativistic hydrodynamics ( grhd ) equations are solved by the module ` grhydro ` @xcite with high resolution shock capturing techniques using the fifth - order weno reconstruction method @xcite and the hlle riemann solver @xcite .
the time integration is performed with a fourth - order runge - kutta method @xcite with a constant courant factor of @xmath19 .
kreiss - olinger dissipation @xcite is applied to the curvature evolution quantities in order to damp high - frequency noise .
we chose in particular the combination of the bssn - nok formulation of the einstein equations and the weno reconstruction method among the possibilities provided by the einstein toolkit because in @xcite we observed that combination of algorithms to enter the convergent regime already at low resolutions , allowing us to obtain reliable results without the need to perform computationally expensive simulations at higher resolution .
our initial data are generated with the open source lorene library @xcite , which is able to compute quasi - equilibrium configurations for binary irrotational neutron stars in quasi - circular orbits .
this is achieved imposing the presence of the helical killing vector @xmath20 . from the imposition of this symmetry
the initial data do not incorporate a radial velocity component coming from the gravitational radiation reaction .
this will lead to the presence of a tangible eccentricity in the stars orbit during their evolution ( see section [ sec : eccentricity ] and ref .
@xcite for more details ) .
the true eos for nuclear matter in an environment similar to a neutron star is still not known , not even assuming a small effect of the temperature , i.e. , a cold neutron star , as expected here for initial data .
we thus have to simulate the effect of different plausible eoss on observable quantities , in the hope to learn about the eos indirectly through these observations .
the four cold eoss for nuclear matter at beta equilibrium that we employed are , in decreasing order of compactness : * the * apr4 * eos @xcite , obtained using variational chain summation methods with the argonne two - nucleon interaction and including also boost corrections and three - nucleon interactions , * the * sly * eos @xcite , based on the skyrme lyon effective nuclear interaction , * the * h4 * eos @xcite , constructed in a relativistic mean field framework including also hyperons contributions and tuning the parameters to have the stiffest possible eos compatible with astrophysical data , and * the * ms1 * eos @xcite , constructed with relativistic mean field theory considering only standard nuclear matter .
.48 @xcite .
right : initial density profiles for the stars using these four eoss , with a baryon mass of @xmath21.,title="fig : " ] + .48 @xcite .
right : initial density profiles for the stars using these four eoss , with a baryon mass of @xmath21.,title="fig : " ] + each of these eoss satisfies the observational constraint of a maximum mass for a non rotating ns higher than @xmath22@xcite . in fig .
[ fig : eos ] and fig .
[ fig : in_dens ] we show the mass over radius curves and the initial density profiles respectively , for non rotating neutron stars for each of these eos .
they have been generated solving the tov equations with the code ` rns ` @xcite .
we followed @xcite parameterizing the eoss as piecewise polytropes , with the following expressions in each density interval @xmath23 $ ] : @xmath24 for each eos we used 7 polytropic pieces , of which the first four ( at lower densities ) always adopt the prescription of @xcite for the stellar crust and the three remaining , instead , use the coefficient found in appendix b of @xcite for each of the four eos models described above .
characteristics of the employed eoss and of their impact on the initial models are listed in tab .
[ table : eos ] .
+ .characteristics of the employed eoss and of the corresponding initial models . in the first three columns we report the adiabatic @xmath25 index of the three highest density polytropic pieces ( starting with density @xmath26 and separated by densities @xmath27 and @xmath28 ) taken from @xcite . in the fourth column
we report the density that separates the crust eos from the ns core eos .
the last three columns show : the gravitational mass at infinite separation , the radius and the compactness @xmath29 of each star , respectively . [
cols="^,^,^,^,^,^,^,^",options="header " , ] +
we presented results from numerical simulations of bns mergers using four different nuclear eoss for the ns interior and starting from four different values for the interbinary distance .
we tested different procedures for the extrapolation of the gravitational wave signal @xmath3 , extracted at finite radii , to null infinity .
in particular , we used for the first time in bns simulations the second order perturbative extrapolation of @xcite , show - casing advantages in reducing the extraction errors with respect to the first - order extrapolation @xcite .
we also showed that the error of the wave extraction procedure , after the extrapolation , is smaller than the typical numerical errors reported for this kind of bns simulations , even in the first part of the signal .
we proposed a simple procedure to integrate the newman - penrose scalar @xmath3 to obtain the gravitational wave strain @xmath4 , based on applying a high - pass digital filter to @xmath4 after a time - domain integration , instead of filtering @xmath3 before a frequency domain integration , which currently is the most common procedure in numerical relativity @xcite .
our procedure gives the cleanest signal when applying it after the extrapolation formula , whose integral terms introduce additional low - frequency noise .
we computed the small orbital eccentricity due to the quasi - circular initial data for all models , showing a dependence on the ns eos but not being able to show any clear dependence on the starting distance between the stars .
we compared the amplitude of our simulations with a recently developed analytical model for eccentric binaries @xcite , showing good agreement in the first part of the signal .
this is a confirmation of the fact that the gravitational wave amplitude oscillations are really an imprint of the orbital eccentricity and of the quality of the simple , approximate , eccentricity calculation procedure we used .
this result confirms also that initial data eccentricity is , until the very last few orbits , the main source of discrepancy between gravitational waves computed with numerical relativity and point - particle post - newtonian approximations . in order to understand the relevance of this measured initial data eccentricity , we compared two analytical waveforms at the same pn order , with the same mass and initial frequency of our @xmath30 km models , one with the same eccentricity of our numerical simulations .
their difference in phase is almost always below typical numerical errors in current bns simulations , except for the last orbit of models with the same eccentricity of our more compact ones ( with apr4 and sly eoss ) .
the detectability of their difference in advanced ligo is small , requiring a maximum distance of @xmath31 mpc for its detection at @xmath32 .
this is due to the shortness of current numerical simulations with respect to the whole frequency band of interferometric detectors
. the effect of eccentricity will be more important when longer numerical simulations will be performed for gw detector data analysis .
techniques for reducing initial data eccentricity have been developed by some groups in the last couple of years @xcite . since the recent beginning of the advanced detector era
, we see an urgent need for public low - eccentricity bns initial data in the numerical relativity community . we compared the results of simulations with the same ns eos and different starting interbinary distance .
their emitted gravitational signals show large phase differences in the last orbits before merger , when aligning them with the standard formula , allowing only for a relative shift in the time variable . when we aligned them allowing for a time dilatation , to account for the longer effective merger time of simulations with @xmath33 km , the remaining phase differences were close to the expected numerical errors .
this effective merger time difference was found to be higher for more compact stars . investigating the post - merger spectrum
, we found that simulations starting from @xmath30 km have the dominant peak at a somewhat lower frequency ( but still consistent with the error ) than the others , while we found a very good agreement with the results of @xcite for simulations starting from a distance of @xmath34 km .
this project greatly benefited from the availability of public software that enabled us to conduct all simulations , namely `` lorene '' and the `` einstein toolkit '' .
we express our gratitude the many people that contributed to their realization .
this work would have not been possible without the support of the suma infn project that provided the financial support of the work of af and the computer resources of the cineca `` galileo '' hpc machine , where part of the simulations were performed .
other computational resources were provided by he louisiana optical network initiative ( qb2 , allocations loni_hyrel15 , and loni_hyrel16 ) , and by the lsu hpc facilities ( supermike ii , allocations hpc_hyrel15 and hpc_hyrel16 ) .
fl is directly supported by , and this project heavily used infrastructure developed using support from the national science foundation in the usa ( grants no .
1212401 , no . 1212426 , no .
1212433 , no .
1212460 ) .
partial support from infn `` iniziativa specifica teongrav '' and by the `` newcompstar '' , cost action mp1304 , are kindly acknowledged .
goodale t , allen g , lanfermann g , mass j , radke t , seidel e and shalf j 2003 the cactus framework and toolkit : design and applications _ vector and parallel processing
vecpar2002 , 5th international conference , lecture notes in computer science _
( berlin : springer ) http://edoc.mpg.de/3341 | we present results from three - dimensional general relativistic simulations of binary neutron star coalescences and mergers using public codes .
we considered equal mass models where the baryon mass of the two neutron stars ( ns ) is @xmath0 , described by four different equations of state ( eos ) for the cold nuclear matter ( apr4 , sly , h4 , and ms1 ; all parametrized as piecewise polytropes ) .
we started the simulations from four different initial interbinary distances ( @xmath1 , and @xmath2 km ) , including up to the last 16 orbits before merger .
that allows to show the effects on the gravitational wave phase evolution , radiated energy and angular momentum due to : the use of different eoss , the orbital eccentricity present in the initial data and the initial separation ( in the simulation ) between the two stars .
our results show that eccentricity has a major role in the discrepancy between numerical and analytical waveforms until the very last few orbits , where `` tidal '' effects and missing high - order post - newtonian coefficients also play a significant role .
we test different methods for extrapolating the gravitational wave signal extracted at finite radii to null infinity .
we show that an effective procedure for integrating the newman - penrose @xmath3 signal to obtain the gravitational wave strain @xmath4 is to apply a simple high - pass digital filter to @xmath4 after a time domain integration , where only the two physical motivated integration constants are introduced .
that should be preferred to the more common procedures of introducing additional integration constants , integrating in the frequency domain or filtering @xmath3 before integration .
_ keywords _ : numerical relativity , gravitational wave , neutron star binaries , einstein toolkit . |
maybe one day , one of my grand - children , at the age of 15 or 16 , will ask me : grand - father , could you explain what is relativity theory ?
my physics teacher lectured about it , talking of rolling trains and of lightnings hitting the railroad , and i understood almost nothing !
this is the discussion i would like to have with her ( or him ) .
do you know the theorem : the diagonals of a parallelogram meet at their middle point ?
yes , i do !
i even know that the converse is true : if the diagonals of a plane quadrilateral meet at their middle point , that quadrilateral is a parallelogram .
and i believe that i know a proof !
good ! you know all the stuff needed to understand the basic idea of relativity theory !
however , we must first think about time and space .
time and space seem to me very intuitive , and yet difficult to understand in deep !
many people feel the same .
the true nature of time and space is mysterious .
let us say that together , time and space make the frame in which all physical phenomena take place , in which all material objects evolve , including our bodies .
we should keep a modest mind profile on such a subject .
we can not hope to understand all the mysteries of time and space .
we should only try to understand some of their properties and to use them to describe physical phenomena .
we should be ready to change the way we think about time and space , if some experimental evidence shows that we were wrong .
but if we do not know what are time and space , how can we hope to understand some of their properties , and to be able to use them ?
by building mental pictures of time and space .
unfortunately we , poor limited human beings , can not do better : we know the surrounding world only through our senses ( enhanced by the measurement and observation instruments we have built ) and our ability of reasoning .
our reasoning always apply to the mental pictures we have built of reality , not to reality itself .
let me now indicate how the mental pictures of time and space used by scientists have evolved , mainly from newton to einstein .
the great scientist isaac newton @xcite ( 16421727 ) used , as mental picture of time , a straight line @xmath0 , going to infinity on both sides , hence without beginning nor end , with no privileged origin .
each particular time , for example now , or three days ago at the sunset at paris , corresponds to a particular element of that straight line .
observe that newton considered , without any discussion , that for each event happening in the universe , there was a corresponding well defined time ( element of the straight line @xmath0 ) , the time at which that event happens .
where is that straight line @xmath0 ?
is it drawn in some plane or in space ? nowhere
! you should not think about the straight line of time @xmath0 as drawn in something of larger dimension .
newton considered time as an abstract straight line , because successive events are linearly ordered , like points on a straight line .
do nt forget that @xmath0 is a mental picture of time , not time itself ! however , that mental picture is much more than a confuse idea : it has very well defined mathematical properties . in modern language
, we say that @xmath0 is endowed with an _ affine structure _ and with an _
orientation_. what is an affine structure ? and what is its use ?
an affine structure allows us to compare two time intervals and to take their ratio , for example to say that one of these intervals is two times the other one .
newton considered the comparison of two time intervals as possible , even when they were many centuries or millenaries apart , and to take their ratio .
in modern mathematical language , that property determines an _ affine structure_. for the mathematician , that property means that we can apply transforms to @xmath0 by sliding it along itself , without contraction nor dilation , and that these transforms ( called _ translations _ ) do not change its properties . for the physicist ,
it means that the physical laws remain the same at all times .
another important property of time : it always flows from past to future . to take it into account
, we endow @xmath0 with an _ orientation _ ; it means that we consider the two directions ( from past to future and from future to past ) as different , not equivalent , for example by choosing the direction from past to future as preferred .
we then say that @xmath0 is _
oriented_. ok , i roughly agree with that mental picture , although it does not account for the main property of time : it flows continuously and we can not stop it ! and what about space ?
newton identified space with the three dimensional space of geometers , denoted by @xmath1 : the space in which there are various figures made of planes , straight lines , spheres , polyhedra , which obey the theorems developed in euclidean geometry : thales and pythagoras theorems , the theorem which says that the diagonals of a parallelogram meet at their middle point , @xmath2 newton used time and space to describe the motion of every object @xmath3 of the physical world as follows .
that object occupies , at each time @xmath4 ( element of @xmath0 ) for which it exists , a position @xmath5 in space @xmath1 .
the motion of of @xmath3 is described by its successive positions @xmath5 when @xmath4 varies in @xmath0 .
let me introduce now a new concept , that of space - time @xcite , due to the german mathematician hermann minkowski ( 18641909 ) .
that concept was not used in mechanics before the discovery of special relativity .
that is very unfortunate , since its use makes much easier the understanding of the foundations of classical mechanics , as well as those of relativistic mechanics .
therefore i use it now , with the absolute time and space of newton , although newton himself did not use that concept .
newton space - time is simply the product set @xmath6 , whose elements are pairs ( called _ events _ ) @xmath7 , made by a point @xmath8 of @xmath1 and a time @xmath4 of @xmath0 .
what is the use of that space - time ?
it is very convenient to describe motions .
for example , the motion of a material particle @xmath9 ( a very small object whose position , at each time @xmath10 , is considered as a point @xmath11 ) , is described by a line in @xmath6 , made by the events @xmath12 , for all @xmath4 in the interval of time during which @xmath9 exists .
that line is called the _ world line _ of @xmath9 .
you will see on figure 1 ( where , for simplicity , space is represented as a straight line , as if it were one - dimensional ) the world lines of three particles , @xmath9 , @xmath13 and @xmath14 .
@xmath15{relpic.1}}\ ] ] * the world line of @xmath13 est parallel to the time axis @xmath0 : that particle is at rest , il occupies a fixed position in the absolute space @xmath1 . *
the world line of @xmath14 is a slanting straight line .
the trajectory of that particle in absolute space @xmath1 is a straight line and its velocity is constant . *
the world line of @xmath9 is a curve , not a straight line .
it means that the velocity of @xmath9 changes with time . for newton , _ rest _ and _ motion _ were absolute concepts : a physical object is at rest if its position in space does not change with time ; otherwise , it is in motion .
it seems very natural .
why should we change this view ?
because nothing is at rest in the universe !
the earth rotates around its axis and around the sun , which rotates around the center of our galaxy . and
there are billions of galaxies in the universe , all moving with respect to the others ! for these reasons , newton s concept of an absolute space was criticized very early , notably by his contemporary , the great mathematician and philosopher gottfried wilhelm leibniz ( 16471716 ) .
but without knowing what is at rest in the universe , how newton managed to study the motions of the planets ?
to study the motion of a body @xmath3 , newton , and after him almost all scientists up to now , used a _
reference frame_. it means that he used another body @xmath16 which remained approximately rigid during the motion he wanted to study , and he made as if that body was at rest
. then he could study the _ relative motion _ of @xmath3 with respect to @xmath16 .
assuming that newton s absolute space @xmath1 exists , we recover the description of absolute motion of @xmath3 by choosing , for @xmath16 , a body at rest in @xmath1 .
the corresponding reference frame is called the _ absolute fixed frame_. the body @xmath16 used to determine a reference frame can be , for example , * the earth ( if we want to study the motion of a falling apple ) , * the trihedron made by the straight lines which join the center of the sun to three distant stars ( if we want to study the motions of the planets in the solar system ) .
all reference frames are not equivalent .
a _ galilean frame _ , also called an _ inertial frame _ , is a reference frame in which the _ principe of inertia _ holds true .
that principle , first formulated for absolute motions in newton s absolute space @xmath1 , says that the ( absolute ) motion of a free particle takes place on a straight line , at a constant speed .
but , as shown by newton himself , that principe remains true for the _ relative motion _ of a free particle with respect to some particular reference frames , the _
galilean frames_. more exactly , let us assume that the principle of inertia holds true for the relative motion of free particles with respect to the reference frame defined by the rigid body @xmath17 .
what happens for the relative motion of these free particles with respect to another reference frame , defined by another rigid body @xmath18 ?
it is easy to see that the principle of inertia still holds true _ if and only if _ the relative motion of @xmath18 with respect to @xmath17 is a motion by translation at a constant speed .
the absolute frame , if it exists , therefore appears as a galilean frame among an infinite number of other gallilean frames , that no measurement founded on mechanical properties can distinguish from the others . for this reason ,
several scientists , following leibniz , doubted about its existence .
leibniz accepted newton s concept of an absolute time , but not that of an absolute space .
his views were not successful during his life , probably because at that time nobody saw how to cast them in a mathematically rigorous setting .
now we can do that ; let me explain how .
we will consider that at each time @xmath10 , there exists a _
space at time @xmath4 _ , denoted by @xmath19 , whose properties are those of the three - dimensional euclidean space of geometers .
we must consider that the spaces @xmath20 and @xmath21 , at two different times @xmath22 and @xmath23 , @xmath24 , have no common element .
leibniz space - time , which will be denoted by @xmath25 ( for universe ) , is the disjoint union of all the spaces @xmath26 for all times @xmath10 .
so , according to leibniz views , we still have a space - time , but no more an absolute space !
the next picture shows , * on the left side , newton space - time @xmath27 , with the two projections @xmath28 and @xmath29 ; * on the right side , leibniz space - time @xmath30 , endowed with only one natural projection onto absolute time @xmath0 , still denoted by @xmath31 ; the horizontal lines represent the spaces @xmath32 , for various values of @xmath10
. @xmath15{relpic.2}}\ ] ] but how do you put together the spaces at various times @xmath19 to make leibniz space - time @xmath30 ? are they stacked in an arbitrary way ?
of course no !
leibniz space - time @xmath30 is a @xmath33-dimensional affine space , fibered ( _ via _ an affine map ) over time @xmath0 , which is itself a @xmath34-dimensional affine space .
its fibres , the spaces @xmath19 at various times @xmath10 , are @xmath35-dimensional euclidean spaces .
the affine structure of @xmath30 is determined by the _ principle of inertia _ of which we have already spoken .
that principle can be formulated in a way which does not use reference frames , by saying : _ the world line of any free particle is a straight line .
_ so formulated , the principle of inertia can be applied to newton space - time @xmath27 and to leibniz space - time @xmath25 as well .
more , it _ determines _ the affine structure of @xmath25 , since one can easily show that the affine structure for which it holds true , if any , is unique . a physical law , the _ principle of inertia _
, is so embedded in the geometry of leibniz space - time @xmath30 . by using a reference frame @xmath16 ,
one can split leibniz space - time into a product of two factors : a space @xmath36 , fixed with respect to that frame , and the absolute time @xmath0 .
but of course , the space @xmath36 depends on the choice of the reference frame @xmath16 .
for that reason , it seems that before 1905 , not many scientists were aware of the fact that by dropping newton s absolute space @xmath1 , they already had completely changed the conceptual setting in which motions are described : * according to newton , absolute space @xmath37 and absolute time @xmath38 were directly related to reality , while space - time @xmath39 was no more than a mathematical object , not very interesting ( he did not use it ) and not directly related to reality ; * but according to leibniz s views , when expressed as done above , it is space - time @xmath30 which is directly related to reality , as well as absolute time @xmath0 ; absolute space @xmath1 no more exists .
einstein @xcite was led to drop leibniz space - time when trying to reconcile the theories used in two different parts of physics : mechanics on one hand , electromagnetism and optics on the other hand .
according to the theory built by the great scotch physicist james clerk maxwell ( 18311879 ) , electromagnetic phenomena propagate in vacuum as waves , with the same velocity in all directions , independently of the motion of the source of these phenomena .
maxwell soon understood that light was an elecromagnetic wave , and lots of experimental results confirmed his views . in leibniz space - time ( as well as in newton space - time ) _ relative velocities behave additively_. in that setting , it is with respect to _ at most one particular reference frame _ that light can propagate with the same velocity in all directions .
physicists introduced a new hypothesis : electromagnetic waves were considered as vibrations of an hypothetic , very subtle , but highly rigid medium called the _
luminiferous ether _
, everywhere present in space , even inside solid bodies .
they thought that it was with respect to the ether s reference frame that light propagates at the same velocity in all directions .
this new hypothesis amounts to come back to newton s absolute space identified with the ether .
there were even physicists who introduced additional complications , by assuming that the ether , partially drawn by the motion of moving bodies , could deform with time !
but if the luminiferous ether really exists , accurate measurements of the velocity of light in all directions should allow the determination of the earth s relative velocity with respect to the ether ! good remark !
these measurements were made several times , notably by albert abraham michelson ( 18521931 ) and edward williams morley ( 18381923 ) , between 1880 et 1887 .
no relative velocity of the earth with respect to the luminiferous ether could be detected .
these results remained not understood until 1905 , despite many attempts .
the most interesting of these attempts was that due to hendrik anton lorentz ( 18531928 ) and george francis fitzgerald ( 18511901 ) .
independently , they proposed the following hypothesis : when a rigid body , for example a rule or the arm of an interferometer , is moving with respect to the luminiferous ether , that body contracts slightly in the direction of its relative displacement .
so that is the famous relativistic contraction my teacher spoke about !
no ! not at all ! lorentz and fitzgerald considered that contraction as a true physical effect of the relative motion of a body with respect to the ether .
this assumption is now completely abandoned , together with the luminiferous ether ! the relativistic contraction of lengths and dilation of times has nothing to do with it : rather than a real phenomenon , it is only an appearance , like the following effect of perspective .
imagine that you look at a 20 centimeters rule , from a distance of , say two meters from its center .
that rule looks shorter when it is not perpendicular to the straight line which joins your eye to its center than when it is .
it may even seem to be reduced to a point when it lies along that straight line .
as we will soon see , the relativistic contraction of lengths and dilation of times has a similar origin .
einstein was the first to understand ( in 1905 ) that the results of michelson and morley experiments could be explained by a deep change of the properties ascribed to space and time . at that time , his idea appeared as truly revolutionary .
but now it may appear as rather natural , if we think along the following lines : _ when we dropped newton space - time in favour of leibniz space - time , we recognized that there is no absolute space , but that space depends on the choice of a reference frame . maybe time too
is no more absolute than space , and depends on the choice of a reference frame ! _
but if we drop absolute time , which properties are left to our space - time ?
in 1905 , einstein implicitly considered that space - time still was a @xmath33-dimensional affine space , which will be called _ minkowski space - time _ and will be denoted by @xmath40 .
he implicitly considered too that _ translations _ of @xmath40 leave its properties unchanged , and he assumed that the _ principe of inertia _ still holds true in @xmath40 when expressed without the use of reference frames : _ the world line of any free particle is a straight line .
_ he also kept the notion of a _
galilean frame_. in @xmath40 , a galilean frame is determined by a direction of straight line ( not any straight line , a _ time - like _ straight line , as we will see below ) .
given a galilean frame @xmath16 , minkowski space - time @xmath40 can be split into a product @xmath41 of a three - dimensional space @xmath36 and a one - dimensional time @xmath42 , which both depend on @xmath16 .
let me recall that in leibniz space - time @xmath30 , a galilean frame @xmath16 allowed us to split @xmath30 into a product @xmath43 of a three - dimensional space @xmath36 , which depended on @xmath16 , and the one - dimensional absolute time @xmath0 , which did not depend on @xmath16 .
that is the main difference between leibniz s and einstein s views about space and time . under these hypotheses ,
the properties of space - time follow from two principles : * the _ principle of relativity : _ all physical laws have the same expression in all galilean frames ; * the _ principle of constancy of the velocity of light : _ the modulus of the velocity of light is an universal constant , which depends neither on the galilean frame with respect to which it is calculated , nor on the motion of the source of that light .
you said that a direction of straight line was enough to determine a galilean frame . but
how is that possible , since we no more have an absolute time ?
that determination will follow from the pinciple of constancy of the velocity of light .
let us call _ light lines _
the straight lines in @xmath40 which are possible world lines of light signals . given an event @xmath44
, the light lines through @xmath3 make a @xmath35-dimensional cone , the _ light cone with apex @xmath3 _ ; the two layers of that cone are called _ the past half - cone _ and _ the future half - cone _ with apex @xmath3 . since it is assumed that translations leave unchanged the properties of space - time , the light cone with another event @xmath45 as apex is deduced from the light cone with apex @xmath3 by the translation which
maps @xmath3 onto @xmath45 . apart from light lines ,
there are two other kinds of straight lines in @xmath40 : * _ time - like straight lines _ , which lie _ inside _ the light cone with any one of their elements as apex ; * and _ space - like straight lines _ , which lie _ outside _ the light cone with any of their element as apex .
i can now explain how the direction of a time - like straight line @xmath46 determines a galilean frame @xmath16 .
that frame is such that the rigid bodies at rest in it are those whose all material points have , as world lines , straight lines parallel to @xmath46 .
the straight lines parallel to @xmath46 will be called the _ isochorous lines _ of the reference frame @xmath16 ; each of these lines is a set of events which all happen at the same place in the space @xmath47 of our frame @xmath16 .
for each event @xmath48 , the set of all other events which occur at the same time as @xmath49 , for the time @xmath42 of our galilean frame @xmath16 , will be called the _ isochronous subspace _ through @xmath49 , for the galilean frame @xmath16 .
it is a @xmath35-dimensional affine subspace @xmath50 of @xmath51 containing the event @xmath49 , and the other isochronous subspaces for @xmath16 are all the @xmath35-dimensional subspaces of @xmath40 parallel to @xmath52 .
they are determined by the property : the length covered by a light signal , calculated in the reference frame @xmath16 , during a given time interval , also evaluated in that reference frame , _ is the same in any two opposite directions . _ in a schematic @xmath53-dimensional space - time ( or in a plane section containing @xmath46 of the true @xmath33-dimensional space - time ) , the direction of isochronous subspaces is easily obtained as shown on the left part of figure 3 : we take the two light lines @xmath54 and @xmath55 through an event @xmath56 ( the red lines on that figure ) ; we take another event @xmath57 , for example in the future of @xmath3 , and we build the parallelogram @xmath58 with two sides supported by @xmath54 and @xmath55 , with @xmath3 as one of its apices and @xmath59 as center .
the isochronous subspaces are all the straight lines parallel to the space - like diagonal @xmath60 of that parallelogram .
three of these lines are drawn ( in blue ) on figure 3 , @xmath61 , @xmath62 and @xmath63 .
a light signal starting from @xmath3 covers , during the time interval between events @xmath3 and @xmath59 , the lengths @xmath64 towards the left and @xmath65 towards the right .
these lengths are equal because @xmath60 is the diagonal of a parallelogram whose center is @xmath59 .
@xmath15{relpic.3}}\ ] ] what for the true @xmath33-dimensional minkowski space - time @xmath40 ? and what are the space @xmath36 and the time @xmath42 of our reference frame @xmath16 ?
it is the same , as shown on the right side of figure 3 .
take the event @xmath66 on the light line @xmath46 such that @xmath59 is the middle point of @xmath67 .
consider the future light half - cone with apex @xmath3 and the past light half - cone with apex @xmath66 .
their intersection is a @xmath53-dimensional sphere @xmath68 .
the unique affine hyperplane @xmath62 which contains @xmath68 is an isochronous subspace for the galilean frame determined by the direction of @xmath69 ( in blue on figure 3 ) .
the other isochronous subspaces for that galilean frame are all the hyperplanes parallel to @xmath70 .
the space @xmath36 is the set of all the isochorous lines , _
i.e _ the set of all straight lines parallel to @xmath46 , and the time @xmath42 the set of all isochronous subspaces .
minkowski space - time @xmath40 splits into the product @xmath41 , or in other words can be identified with that product , because a pair made by an isochorous line and an isochronous subspace determine a unique element of @xmath40 , the event at which they meet .
what happens if you change your galilean frame ?
of course , as for galilean frames in leibniz space - time , the direction of isochorous lines ( the straight world lines of points at rest with respect to the chosen galilean frame ) is changed . moreover ,
contrary to what happened in leibniz space - time , the direction of isochronous subspaces is also changed ! therefore , the chronological order of two events can be different when it is appreciated in two different galilean frames ! up to now
, we have compared the lengths of two straight line segments in @xmath40 only when they were supported by parallel straight lines .
that was allowed by the _ affine structure _ of @xmath40 .
we need more , because the spectral lines of atoms allow us to build clocks and to compare time intervals measured in two different galilean frames .
let @xmath71 and @xmath72 be two straight line segments supported by two different time - like straight lines @xmath46 and @xmath73 , which meet at the event @xmath3 .
let @xmath74 and @xmath75 be the galilean frames determined by the directions of @xmath46 and @xmath76 , respectively .
we assume that the time intervals corresponding to @xmath71 measured in @xmath74 , and to @xmath72 measured in @xmath75 , are the same .
let @xmath77 be the event at which the time - like straight line @xmath73 meets the isochronous subspace @xmath78 containing @xmath59 of the galilean frame @xmath74 ( figure 4 ) .
since the events @xmath59 and @xmath77 are synchronous for @xmath74 , the time interval corresponding to @xmath72 appears longer than the time interval corresponding to @xmath71 when both are observed in the reference frame @xmath74 , by the ratio @xmath79 . that ratio is the _ ratio of dilation of times _ of the galilean frame of @xmath80 , when observed in the galilean frame @xmath74
similarly , @xmath81 is the ratio of dilation of times of the galilean frame @xmath74 when observed in the galilean frame @xmath75 . according to the principle of relativity
, these two galilean frames must play the same role with respect to the other , which implies the equality @xmath82 . by a well known property of hyperbolae , that equality holds _ if and only if @xmath59 and @xmath83 lie on the same arc of hyperbola which has the light lines @xmath55 and @xmath54 ( which meet at @xmath3 and are contained in the two - dimensional plane which contains @xmath46 and @xmath73 ) as asymptotes_. or more generally , on the same hyperboloid with the light cone of @xmath3 as asymptotic cone .
@xmath15{relpic.4}}\ ] ] the comparison of lengths on two non - parallel space - like straight lines is similar to the comparison of time intervals .
let @xmath84 and @xmath85 be two segments supported by two space - like straight lines which meet at the event @xmath3 .
they are of equal length _ if and only if @xmath86 and @xmath87 lie on the same hyperboloid with the light cone of @xmath3 as asymptotic cone_.
the comparison of time intervals and lengths presented above allows a very natural introduction of the pseudo - euclidean metric of minkowski space - time .
the construction of isochronous subspaces in two different galilean frames , as presented above , leads to the formulas for lorentz transformations with a minimum of calculations .
the pictures we have presented allow a very easy explanation of the apparent contraction of lengths and dilation of times associated to a change of galilean frames and a very simple explanation , without complicated calculations , of the ( improperly called ) paradox of langevin s twins
. * acknowledgements . *
the author thanks the team analyse algbrique of the institut de mathmatiques de jussieu and his university for taking in charge his registration fee at this international conference .
einstein , a. , lorentz , h.a . ,
weyl , h. , minkowski , h. , _ the principles of relativity _ , a collection of original papers on the special and general theory of relativity , with notes by a. sommerfeld .
methuen and company , 1923 .
reprinted by dover publications , inc .
, new york .
newton , isaac , _
principes mathmatiques de la philosophie naturelle _ , tomes i et ii , translated by madame la marquise du chastellet , chez desaint et saillant , paris , 1759 . reprinted by the ditions jacques gabay , paris , 1990 .
poincar , henri , , book containing the text of a lecture presented at the congress of the association franaise pour lavancement des sciences ( lille , 1909 ) , the paper dated 23 july 1905 _ sur la dynamique de llectron _ , rendiconti del circolo matematico di palermo * xxi * ( 1906 ) , and a note aux comptes rendus de lacadmie des sciences with the same title dated 15 june 1905 ; gauthier - villars , paris , 1924 ; reprinted by the ditions jacques gabay , paris , 1989 . | the author thinks that the main ideas or relativity theory can be explained to children ( around the age of 15 or 16 ) without complicated calculations , by using very simple arguments of affine geometry .
the proposed approach is presented as a conversation between the author and one of his grand - children . limited here to the special theory ,
it will be extended to the general theory elsewhere , as sketched in conclusion . for agathe , florent , basile , mathis , gabrielle , morgane , quitterie and my future other grand - children |
the latest measurement of the shapiro delay for the millisecond pulsar psrj1614 - 2230 provides reliable lower bound on the maximum mass to be @xmath2 @xcite . this measurement rules out all the equations of state ( eoss ) yielding the maximum mass less than that of the psrj1614 - 2230
of course , the eoss for the nucleonic matter can readily yield the compact stars with masses @xmath3 .
the eoss with hadron - quark phase transition are also compatible with the mass measurement of the psrj1614 - 2230 , provided , the quarks are assumed to be strongly interacting and are in colour superconducting phase @xcite .
however , at large , the maximum mass of the compact stars are found to be well below @xmath0 when the non - nucleonic degrees of freedom like hyperons and kaon condensates are considered @xcite . one might thus infer in the backdrop of previous calculations that the existence of hyperons and kaon condensates are unlikely in the psrj1614 - 2230 .
recently , studies involving role of hyperons on the maximum mass of the compact stars are revisited @xcite .
it is shown that the eos for the hyperonic matter at higher densities can be stiffened in several ways within the relativistic mean - field ( rmf ) models .
the maximum mass for these eoss are above @xmath0 indicating the existence of hyperons in the psrj1614 - 2230 .
the eos of the hyperonic matter are stiffened by including the vector - isoscalar strange @xmath4 mesons within the rmf model , in addition to the usual scalar - isoscalar @xmath5 , vector - isoscalar @xmath6 and vector - isovector @xmath7 mesons @xcite .
the @xmath4 mesons stiffens the eos at densities far beyond the nuclear saturation density .
further , the increase of about @xmath8 in the maximum mass of the compact stars is achieved by varying the hyperon - nucleon potential depths for the @xmath9 and @xmath10 hyperons which are not very well known .
it has been also demonstrated @xcite that maximum mass of the compact stars containing hyperons are well above @xmath11 , provided , the @xmath6 mesons are coupled to hyperons as strongly as to the nucleons .
such strong meson - hyperon couplings stemmed from the breaking of su(6 ) symmetry . another way to stiffen
the eos is to include the non - linear terms for the @xmath4-mesons as suggested in ref .
it may be pointed out that the nuclear eoss employed in ref .
@xcite are generally quite stiff in comparison to those extracted from the heavy - ion data @xcite
. an alternative scenario has also emerged in which compact stars containing hyperons can have masses @xmath3 . in this scenario
, one assumes the presence of hypothetical weakly interacting light bosons ( wilbs ) at densities several times of the nuclear saturation density .
the role of wilbs on the eos or the internal structure of the compact stars crucially depends on the choice of the characteristic scale which is the square of the ratio of the coupling strength to the masses of wilbs .
the values of the characteristic scale of the wilbs are poorly constrained at present by the laboratory data and the compact star observable @xcite .
consequently , the nuclear eoss compatible with the heavy - ion data allow the hyperons to exist in the compact stars with mass @xmath3 @xcite .
the required stiffness of the eos at very high density is achieved by adjusting the characteristic scale of the wilbs . in the present work
we explore the possibility of existence of hyperons in the psrj1614 - 2230 using a diverse set of nuclear eoss calculated within the rmf models .
in particular , we examine whether the nuclear eoss compatible with the heavy - ion data can allow the hyperons to exist in the psrj1614 - 2230 without recourse to the wilbs .
this makes our present investigation quite different than the ones performed earlier .
most of the previous investigations either employed the nuclear eoss which are stiffer in comparison to the heavy - ion data or included the contributions of the wilbs .
the paper is organized as follows . in sec .
ii we describe the theoretical framework based on the rmf model . in sec .
iii , we present our results for the eoss for the matter without and with hyperons . in sec .
iv we present our results for the bulk properties of the static compact stars . finally , we state our conclusions in sec . v.
we use standard and extended versions of the rmf models to compute the bulk properties of the compact stars .
the standard rmf model includes the contributions from the non - linear self - interaction for the @xmath5 meson and extended rmf model includes the contributions from the self and/or cross interaction terms for the @xmath5 , @xmath6 and @xmath7 meson .
we will discuss both versions of the rmf models .
the derivations of the effective lagrangian density and corresponding energy density functionals for the extended rmf model are well documented in refs .
@xcite . in rmf models ,
baryons interact through the exchange of scalar ( @xmath5 ) , vector ( @xmath6 and @xmath7 ) as well as two additional hidden - strangeness ( @xmath5 * and @xmath4 ) mesons .
the baryons considered in this work are nucleon ( @xmath12 ) and hyperons ( @xmath13 ) . the total lagrangian density of rmf model including hyperons plus leptons ( l )
can be written as , @xmath14 where the free baryons lagrangian density is , @xmath15\psi_b,\ ] ] here , @xmath16 is baryons field and the sum is taken over @xmath12 , @xmath17 , @xmath9,and @xmath10 baryons .
the lagrangian density for free mesons involved is , @xmath18 the @xmath19 , @xmath20 and @xmath21 are field tensors corresponding to the @xmath6 , @xmath4 and @xmath7 mesons field , and can be defined as @xmath22 , @xmath23 and @xmath24 .
the lagrangian @xmath25 describing the interactions of the baryons through the mesons is , @xmath26\psi_b , \label{eq : llin}\ ] ] where @xmath27 are the baryons isospin matrices .
the lagrangian describing nucleons self interactions for @xmath5 , @xmath6 , and @xmath7 mesons can be written as , @xmath28 while the free leptons lagrangian density is , @xmath29\psi_l.\ ] ] here @xmath30 is leptons ( electron and muon ) field .
the meson - nucleon coupling constants and nonlinear parameters of rmf models are determined by adjusting them to reproduce ground state properties of finite nuclei and nuclear matter . here
, we employed nl3 , gm1 , tm1 , g2 , bsr2 , bsr6 , bsr9 , bsr13 and bsp parameter sets for meson - nucleon coupling constants and nonlinear parameters @xcite .
these parameterizations of the rmf models are associated with different form for the lagrangian density .
they differ in their non - linear part of the effective lagrangian ( eq . [ eq : lnlin ] ) which are summarized in table [ tab : rmf ] .
for instance , the lagrangian density associated with the parameter sets nl3 and gm1 correspond to the standard rmf model which includes non - linear term only for the self - interactions of the @xmath5 meson . other parameter sets tm1
, g2 and bsr correspond to the extended rmf models which include contributions from the self - interactions of the @xmath6 mesons and/or cross - interactions between @xmath31 and @xmath7 mesons .
the meson - hyperon coupling strengths @xmath32(@xmath33 and @xmath34 ) in eq .
( [ eq : llin ] ) can not be determined very well at present .
the properties of compact stars are quite sensitive to the choices for the values of @xmath35,@xmath36 and @xmath37 .
the variations in the values of @xmath38 and @xmath39 do not appreciably alter the properties of compact stars .
the values of @xmath38 and @xmath39 are kept fixed in our calculations .
we take @xmath40 , this choice is consistent with @xmath41hypernuclear data which yields weak @xmath42 interaction @xcite .
the values of @xmath38 are taken from the su(6 ) quark model .
we make several choices for the values of @xmath43 and @xmath37 .
one of which correspond to those obtained within the su(6 ) quark model .
the su(6 ) quark model values for @xmath36 , @xmath38 and @xmath37 are @xmath44 for given values of @xmath36 the coupling strengths @xmath35 are usually obtained from potential depths for hyperons in the symmetric nuclear matter at the saturation density as , @xmath45 the values of potential depths @xmath46 chosen are as follows @xcite , @xmath47 for the sake of convenience we define , @xmath48 and , @xmath49 it has been shown earlier in ref .
@xcite that the values of @xmath50 , which are larger compared to those for the su(6 ) model ( eq . [ eq : su6 ] ) , can yield heavier compact stars composed of hyperons without affecting the properties of the @xmath51hypernuclei .
recent investigation @xcite suggests that the values of @xmath52 can significantly differ from those given by the su(6 ) quark model .
the breaking of the su(6 ) symmetry yields @xmath53 ( i.e. , @xmath54 ) . in the present work we vary the values of @xmath55 and @xmath56 over a wide range to see in what limits the compact stars with hyperons can satisfy the constraints on the lower bound on the maximum mass imposed by the psrj1614 - 2230 .
we use different versions of the rmf models , as summarized in table [ tab : rmf ] , to compute the eoss for the @xmath57-equilibrated nucleonic and hyperonic matters .
our choice for the models are such that they yield wide variations in the various quantities associated with symmetric and asymmetric nuclear matter at the saturation density . in table
[ tab : nmpro ] we provide the values of some of the quantities associated with nuclear matter at the saturation density @xmath58 , namely , the binding energy per nucleons @xmath59 , incompressibility coefficient for symmetric nuclear matter ( @xmath60 ) , symmetry energy coefficient ( @xmath61 ) , linear density dependence of the symmetry energy coefficient ( @xmath62 ) and various quantities ( @xmath63 ) , ( @xmath64 ) and ( @xmath65 ) .
these quantities are evaluated as follows , @xmath66 where , @xmath67 is the energy per nucleon at a given density @xmath7 and asymmetry @xmath68 . in table
[ tab : nmpro ] , we also list the values of the neutron - skin @xmath69 for the @xmath70pb nucleus .
the values of @xmath60 , @xmath71 and @xmath62 vary over a wide range for the selected rmf model .
this immediately indicates that the behaviour of the eoss for the different rmf models considered should be quite different even at the low densities . in fig .
[ fig : snm ] we plot the eoss for symmetric nuclear matter ( snm ) in terms of pressure versus nucleon density obtained for various rmf models .
the bounds on the eoss as shown by shaded regions are the ones extracted from heavy - ion collision data @xcite .
the eoss for the nl3 , gm1 , bsr2 and bsr6 are stiffer in comparison to those obtained from the heavy - ion data . whereas , the tm1 , g2 , bsr9 , bsr13 and bsp are consistent with the eos from the heavy - ion data .
the eos for the bsr2 and bsr6 parameterizations are very much the same for the reasons as follow .
both of the parameter sets belong to the same type of the rmf model as can be seen from table [ tab : rmf ] .
further , the parameters of the bsr2 and bsr6 parameterizations are obtained by fit to exactly same set of experimental data for bulk properties of finite nuclei , but , for different values of the neutron - skin thickness in @xmath70pb nucleus .
thus , the bsr2 and bsr6 parameterizations are expected to yield different behaviour only for the eoss of the asymmetric nuclear matter .
similar is the case for the eoss obtained with the bsr9 and bsr13 parameterizations .
our main goal is to investigate whether the nuclear eoss which are compatible with the heavy - ion data can allow the hyperons to exist in the @xmath0 pulsar psrj1614 - 2230 . for this purpose , the eoss for the @xmath72equilibrated hyperonic matter are calculated using different values for @xmath52 and @xmath56 .
we present here the eoss only for two different choices for the @xmath52 and @xmath73 .
we refer these choices as xsu6 and x180 . for the case of xsu6 , @xmath74 and @xmath75 which correspond to those given by the su(6 ) quark model as conventionally used .
the x180 refers to @xmath76 and @xmath77 .
the values of @xmath78 coupling strength are determined using the nucleon - hyperon potential depths as described in sec .
it can be easily verified from eqs .
( [ eq : su6 ] , [ eq : xw ] and [ eq : xphi ] ) that the values of @xmath52 and @xmath56 for the case of x180 are augmented by a factor of @xmath79 with respect to their values for the su(6 ) model . in fig .
[ fig : eos_bem ] , the eoss for the nucleonic and hyperonic matters are compared for a few different rmf models .
the eoss for the nucleonic matter are labeled as no hyp. the eoss for the hyperonic matter are labeled as xsu6 and x180 .
it can be seen that the eoss for the hyperonic matter corresponding to x180 are quite close to those obtained for the nucleonic matter .
it should be pointed out that the threshold density is lowest for the @xmath17 hyperons for all the rmf models considered irrespective of the choice of the meson - hyperon coupling strengths .
further , the augmented meson - hyperon coupling strengths results in marginal increase in the values of the threshold densities for the @xmath17 hyperons .
for instance , the threshold densities for @xmath17 hyperons for the nl3(bsp ) parameterizations for the xsu6 and x180 cases are 0.28(0.38 ) and 0.32(0.42 ) @xmath80,respectively .
thus , the augmented meson - hyperon coupling strengths might increases the maximum mass of the compact stars without significantly affecting its hyperonic contents . before employing our eoss for the hyperonic matter to study the properties of the compact stars
, we would like to look into the behaviour of the @xmath51h potentials @xmath81 . the values of @xmath51h for given meson - hyperon coupling strengths are obtained as , @xmath82 where , the values of the fields @xmath31 and @xmath4 are calculated for the pure @xmath17 matter at a given density @xmath7 . in fig .
[ fig : ulh ] , we plot the results for the @xmath81 obtained using the meson - hyperon coupling strengths corresponding to xsu6 and x180 . for the comparison
, we also plot the values of @xmath81 obtained for two different models based on the su(3 ) symmetry @xcite .
the green dashed curve is obtained using ref .
@xcite for @xmath83 which yields highest value for the maximum mass for the compact star .
the values of @xmath84 and @xmath85 for the case of x180 lie within those obtained using other models .
whereas , the potential @xmath86 is somewhat stiffer in comparison to the other models as considered .
the properties of static or non - rotating compact stars for a given eos is obtained by solving the tolman - oppenheimer - volkoff ( tov ) equations @xcite . for the case of rotating compact stars we solve the einstein equations for stationary axi - symmetric spacetime .
the numerical computations are performed using the code written by stergioulas @xcite .
we describe the outer crust region of the compact star using the eos of rster _
et al . _
@xcite which is the recent update of the one given by baym , pethick , and sutherland @xcite . due to the fact that the detailed eos of inner crust indeed is not yet certain ,
the polytropic pressure - energy density relation is used to interpolate the eos for the region between outer crust and the core @xcite .
the core is assumed to be composed of either the nucleonic or the hyperonic matter in @xmath72equilibrium .
the eos of the core is obtained from the different parameter sets of the rmf models as presented in sec .
[ sec : eos ] . in fig .
[ fig : m - r ] we display our results for the mass - radius relationships for the static compact stars composed of @xmath57-equilibrated nucleonic matter ( upper panel ) and hyperonic matter ( lower panel ) .
the shaded region represents the current lower limit on the maximum mass , @xmath87 , of the compact star obtained from the recent mass measurement of the psrj1614 - 2230 .
the mass - radius relationships for the compact stars containing hyperons are obtained using standard values for the meson - hyperon coupling strengths as discussed in the previous section .
it is evident from the upper panel that the maximum masses of the compact star in the absence of hyperons for all the rmf models considered are compatible with the current limit on its lower bound .
the solid circles in the lower panel indicate the minimum mass of the compact star at which the hyperons begin to appear .
this minimum mass ranges from @xmath88 for the various rmf models considered .
the parameter sets nl3 , gm1 , bsr2 and bsr6 yield maximum mass larger than @xmath0 when the hyperons are included .
thus , these parameter sets readily allow the hyperons to appear in the psrj1614 - 2230 .
the nuclear eoss , however , for these cases are not compatible with the heavy - ion data ( see fig . [
fig : snm ] ) . on the other hand , the nuclear eoss for the parameter sets bsr9 , bsr13 , bsp , g2 and tm1 are compatible with the heavy - ion data , but , they yield the maximum mass in the range of @xmath89 with the inclusion of hyperons .
it seems that the nuclear eoss , compatible with heavy - ion data , can possibly allow the hyperons to exist in the psrj1614 - 2230 for substantially larger values of the meson - hyperon coupling strengths in comparison to their standard values . in fig .
[ fig : mr_x180 ] we show the mass - radius relationships for the compact stars obtained by augmenting the meson - hyperon coupling strengths in comparison to their standard values .
as discussed previously , the label x180 corresponds to @xmath76 and @xmath90 which is larger by a factor of @xmath91 with respect to their slandered values taken from the su(6 ) quark model .
we plot the results only for those rmf model for which the nuclear eoss are compatible with the heavy - ion data .
we find that the maximum masses for the tm1 and bsp parameter sets are consistent with the mass of the psrj1614 - 2230 .
it may be pointed out to this end that the radius @xmath92 for the compact star with canonical mass of @xmath93 for the bsp case is 12.7 km which is consistent with @xmath94 km as determined from the recent observations of both transiently accreting and bursting sources @xcite .
the values of @xmath92 determined in ref .
@xcite are independent of assumptions about the composition of the core .
we also look into the hyperon fractions at the maximum mass . in table [
tab : ns ] we present our results for the maximum mass configurations obtained for two different choices of meson - hyperon coupling strengths corresponding to xsu6 and x180 .
these results are compared with their nuclear counterparts having no hyperons .
we also list the values of the partial mass @xmath95 which is composed of the hyperonic matter .
the values of @xmath95 are obtained by integrating the density profile , corresponding to the maximum mass configuration , from the central density to the threshold density at which hyperons begin to appear .
it is interesting to note that the ratio @xmath96 is sizable and almost the same for both the choices of the meson - hyperon coupling strengths .
the maximum mass obtained using augmented meson - hyperon coupling strengths are very close to their nuclear counter parts and yet significant fraction ( @xmath1 ) of the mass is composed of the hyperonic matter .
further , the radius @xmath97 at the maximum mass for the case with no hyperons and the one obtained using augmented meson - hyperon coupling strengths are almost the same .
thus , it seems possible to obtain the maximum mass configuration of the compact star , containing sizable fraction of hyperons , which masquareds its nuclear counterpart .
the maximum mass of the compact stars with hyperons nearly same as their nuclear counterpart has also been found in ref .
@xcite in fig .
[ fig : bsr28 ] we plot the variations of @xmath98 with @xmath52 and @xmath56 for the case of bsp .
the value of @xmath98 increases with increasing @xmath52 and @xmath56 .
the @xmath98 depends strongly on @xmath52 for smaller values of @xmath56 .
as the @xmath56 increases , the value of @xmath98 tends to saturate and becomes more or less independent of @xmath52 . for @xmath99 , the @xmath98 increases by @xmath100 with increase in @xmath52 from 0.5 to 1 . also
, the @xmath98 is less dependent on @xmath56 for larger @xmath52 . to this end
, we would like to remind that the @xmath101 potential depth @xmath102 is taken to be 30 mev .
the value of @xmath103 is not yet certain . in fig .
[ fig : bsr281 ] , we compare our results for the @xmath98 obtained for @xmath104 and 30 mev .
the value of @xmath98 depends somewhat on @xmath103 for smaller values of @xmath105 . for @xmath106 ,
the @xmath98 may increases by @xmath107 with increase in @xmath103 from -30 to 30 mev .
we also calculate the maximum mass of the compact stars rotating with the frequency @xmath108 hz as that of the psrj1614 - 2230 . in fig .
[ fig : bsp_rns ] , we display the variations for the @xmath98 with @xmath52 and @xmath56 .
it can be easily seen that the value of @xmath76 with @xmath109 is sufficient to produce the compact star with mass compatible with the current lower bound on the @xmath98 .
we use various rmf models to explore the possibility of existence of hyperons in the heaviest observed compact star psrj1614 - 2230 having mass @xmath110 .
we have examined the conditions required for the hyperons to exist in the psrj1614 - 2230 when the nuclear eoss are subjected to the constrained imposed by the the heavy - ion data .
the various rmf models are selected in such a way that they result in a diverse set of nuclear eoss .
the values of maximum mass for these models , in the absence of hyperons , are consistent with the mass of psrj1614 - 2230 .
the nuclear eoss which are consistent with the heavy - ion data require larger values of the meson - hyperon coupling strengths in order to allow the hyperons to exist in the psrj1614 - 2230 .
particularly , the coupling of @xmath6 and @xmath4 vector mesons to the hyperons are required to be significantly augmented with respect to their standard values .
these lager values of the coupling strengths increases the maximum mass to the desired limit without significantly affecting the fraction of maximum mass composed of hyperonic matter .
we find that the maximum mass configuration of the compact star with sizable fraction of hyperons masquareds its nuclear counterpart .
.[tab : rmf ] various self - interaction and cross - interaction terms included in the lagrangian density associated with different parameterizations of the rmf models considered in the present work . the index 1 and 0 is used to indicate whether or not the corresponding term is included .
[ cols="^,^,^,^,^,^,^,^ " , ] | the possibility of existence of hyperons in the recently measured @xmath0 pulsar psrj1614 - 2230 is explored using a diverse set of nuclear equations of state calculated within the relativistic mean - field models .
our results indicate that the nuclear equations of state compatible with heavy - ion data allow the hyperons to exist in the psrj1614 - 2230 only for significantly larger values for the meson - hyperon coupling strengths .
the maximum mass configurations for these cases contain sizable hyperon fractions ( @xmath1 ) and yet masquared their counterpart composed of only nucleonic matter . |
a recently proposed model of 3d massive gravity @xcite , dubbed `` minimal massive gravity '' ( mmg ) , has bulk properties that are identical to those of `` topologically massive gravity '' ( tmg ) ( which propagates a single massive spin-@xmath0 mode @xcite ) but its boundary properties ( for ads asymptotics ) are different .
specifically , mmg evades the `` bulk / boundary clash '' of tmg ; this is the impossibility ( for tmg ) of arranging for both central charges of the asymptotic conformal symmetry algebra to be positive while also arranging for the bulk mode to have positive energy . in this paper
we present a greatly simplified , and geometrical , analysis of the unitarity conditions of mmg .
our results confirm those of @xcite but we also consider a slightly larger class of models by leaving free the normalisation of the parameters of the mmg action , and we cut in half the relevant parameter space by establishing equivalence under a `` duality '' transformation in the full parameter space .
the final result is that unitarity restricts the parameters to a connected region of parameter space , up to equivalence .
it has been proposed @xcite that the relevant asymptotic symmetry algebra for 3d gravity with flat - space asymptotics is the 2d `` galilean conformal algebra '' ( gca ) , obtained by contraction of the 2d conformal algebra .
based on this proposal , flat space holography for tmg was initiated in @xcite , and it was argued that unitarity constraints on the central charges of the gca algebra could be satisfied only in a limit for which tmg degenerates to 3d conformal gravity , which has no local degrees of freedom . here we show that the same analysis , applied to mmg , leads to a flat space mmg model that is not conformal and still has local degrees of freedom , although its linearised limit coincides with linearised 3d conformal gravity , implying a linearisation instability .
the tmg action can be written as the integral of a lagrangian 3-form constructed from three lorentz - vector one - forms : the dreibein @xmath1 , the ( dual ) lorentz connection @xmath2 and a lagrange multipler field @xmath3 imposing a zero - torsion constraint @xcite . using a 3d vector algebra notation for lorentz vectors we can write this lagrangian 3-form as @xmath4 = -\sigma e\cdot r + \frac{1}{6}\lambda_0\ , e\cdot e\times e + h\cdot t + \frac{1}{\mu } l_{lcs}(\omega)\ , , \ ] ] where @xmath5 and @xmath6 are the torsion and curvature 2-forms , respectively , and @xmath7 is the lorentz - chern - simons ( lcs ) 3-form for @xmath2 .
this lagrangian 3-form is parametrized by the mass parameter @xmath8 and the dimensionless constants @xmath9 .
we refer the reader to @xcite for details of conventions .
the 1-form fields @xmath10 can be eliminated by their joint equations of motion , leading to an action for the 3-metric alone . for a particular choice of units for the 3d newton constant @xmath11 ( which has dimensions of inverse mass for unit speed of light ) , the resulting action is @xmath12 = \sigma i_{eh}[g;\lambda ] + \frac{1}{\mu } i_{lcs}[g]\ , , \ ] ] where @xmath13 is the ( 3d ) einstein - hilbert action , together with a cosmological constant term ( @xmath14 ) ; the lcs term is now the chern - simons action for the levi - civita affine connection .
variation of this action yields the tmg equation @xmath15 where @xmath16 is the einstein tensor and @xmath17 the cotton tensor .
the asymptotic symmetry algebra for an asymptotic ads vacuum with cosmological constant @xmath18 is the 2d conformal algebra @xcite .
for tmg the ( left / right ) central charges are @xmath19 in the context of an ads / cft interpretation , this is the leading order result in a semi - classical approximation that is valid ( re - instating factors of @xmath20 ) when @xmath21 , and quantum consistency requires @xmath22 for some integer @xmath23 ( see e.g. @xcite ) .
we may assume , without loss of generality , that @xmath24 given this , we see from ( [ cpm ] ) that @xmath25 will be negative and the ( putative ) dual cft non - unitary , unless @xmath26 . however , the spin-2 bulk mode is a `` ghost '' ( i.e. has negative energy ) unless @xmath27 .
this is the bulk / boundary unitarity clash .
the mmg action can not be written in terms of the metric alone but it can be written in terms of the @xmath28-form lorentz - vector fields @xmath29 . the lagrangian 3-form is @xcite @xmath30= l_{tmg}[e,\omega , h ] + \frac{\alpha}{2 } e\cdot h\times h\ , , \ ] ] where @xmath31 is a new dimensionless constant ; for reasons explained in @xcite , the parameters are restricted such that @xmath32 the 1-form fields @xmath10 can not be consistently eliminated from the action but the complete set of equations of motion determine them uniquely in terms of the dreibein and its derivatives . when the resulting equation of motion is expressed in terms of the metric
, one finds that @xmath33 where . ] @xmath34 and @xmath35 the @xmath36-tensor is @xmath37 where @xmath38 ( the 3d schouten tensor ) and the scalar @xmath39 is its trace .
notice that the @xmath36-tensor involves only second derivatives of the metric , so the characteristics of the mmg equation will be those of the tmg equation , found by setting @xmath40 .
we defer to @xcite for an explanation of how the mmg equation manages to be consistent with bianchi identities despite the fact that @xmath41 .
the mmg action is proportional to the integral of the lagrangian 3-form ( [ mmgact ] ) with a proportionality constant that is itself proportional to @xmath42 .
a rescaling of @xmath11 can be compensated by a rescaling of the mmg parameters . specifically , if @xmath43 , for _ positive non - zero _ constant @xmath44 , then the action is unchanged if @xmath45 and @xmath46 given that @xmath47 , this scaling allows us to choose a normalisation for the parameters such that @xmath48 , and this choice was made in @xcite . for tmg
it makes sense to make this choice since setting @xmath49 leads to inconsistent field equations unless one also sets @xmath50 and then one is left with 3d conformal gravity , which has no local degrees of freedom . for mmg
it is perfectly consistent to set @xmath49 , and allowing for this case leads to significant simplifications .
for this reason , we impose no restriction on the parameter @xmath51 .
we could use the scaling symmetry to set @xmath52 .
this is especially attractive since unitarity of mmg ( with ads asymptotics ) requires @xmath53 @xcite so , anticipating this result , we could restrict to @xmath54 .
however , imposing a normalisation condition on @xmath31 would complicate any discussion of the tmg limit .
for this reason , we shall proceed without making any choice of normalisation for the mmg parameters .
let us define a new set of fields @xmath55 for the action ( [ mmgact ] ) by setting @xmath56 this is an invertible field redefinition and therefore has no physical effect .
however , the action in terms of the new fields has exactly the same form as it had for the old fields but in terms of the new parameters @xmath57 in other words , the map @xmath58 takes one mmg model into an equivalent mmg model .
in particular , the mmg field equation ( [ sourcefree ] ) is unchanged because the parameters @xmath59 are @xmath60-inert .
notice that the @xmath60-map is an involution ( @xmath61 ) so it relates _ pairs _ of equivalent models ; it is a `` duality '' in parameter space .
notice also that @xmath62 recall that @xmath63 by definition , so @xmath60 has no fixed points .
it follows that we may choose @xmath64 to have either sign , without loss of generality . every mmg model for which @xmath65 is equivalent to one with @xmath66 .
this result raises a puzzle .
when @xmath66 we can take the @xmath67 limit to recover tmg from mmg . given equivalence under the action of h , there should also be a limit to tmg when @xmath65 , but this can not be a simple @xmath67 limit because this would change the sign of @xmath64 .
the resolution of this puzzle is that there is indeed another , less obvious , tmg limit when @xmath65 , which may be deduced by following the consequences of the @xmath60-map .
one takes @xmath68 in such a way that for , non - zero @xmath31 , @xmath69 where @xmath70 and @xmath71 are finite in the @xmath67 limit .
this limit now yields a tmg model with @xmath72
maximally symmetric solutions of the mmg equation ( [ sourcefree ] ) are characterized by the value of the cosmological constant @xmath73 , which can be defined ( in the context of a vacuum solution ) via the equation @xmath74 .
substitution into the mmg equation yields the relation @xmath75 in the case of an ads vacuum we have a unitarity constraint arising from the requirement of positive central charges for the asymptotic 2d conformal algebra , in addition to the bulk unitarity constraints .
the unitarity conditions found for mmg in @xcite are * no - tachyon condition : @xmath76 * no - ghost condition .
when combined with the no - tachyon condition this becomes the inequality @xmath77 * positivity of both central charges : @xmath78 this inequality is valid on the assumption ( which we make without loss of generality ) that @xmath79 ( recall that @xmath80 ) and saturation of it defines , in analogy to tmg @xcite , the `` chiral '' limit .
properties of chiral mmg have been investigated recently in @xcite . for tmg
it is obvious that the chiral limit implies saturation of the no - tachyon condition .
this is no longer obvious for mmg but it is true nevertheless , as will emerge from the analysis to follow .
we now define new dimensionless variables @xmath81 by @xmath82 observe that the inequality @xmath83 follows immediately from our assumption that @xmath84 .
it also follows that @xmath85 where the final equality follows from ( [ quad ] ) .
the new notation also simplifies the result of @xcite for the central charges of the asymptotic 2d conformal algebra , which can now be written as @xmath86 the unitarity conditions can now be written as linear inequalities in the @xmath81-plane : * no - tachyon : @xmath87 . *
no - ghost : @xmath88 .
* positive central charges : @xmath89 . to summarise ,
the `` allowed '' region in the @xmath81-plane is defined by the linear inequalities @xmath90 within this allowed region , each point on the `` mmg hyperbola '' , @xmath91 corresponds to an mmg model with ads@xmath92 vacuum satisfying all ( bulk and boundary ) unitarity conditions .
observe that the boundary lines @xmath93 and @xmath94 meet _ on the hyperbola _ at the `` chiral '' point with coordinates @xmath95 this confirms our earlier claim that the no - tachyon condition for mmg is saturated at the critical point . the other intersection of the two lines @xmath93 and
@xmath94 is also on the hyperbola , but this corresponds to @xmath96 , and hence to @xmath97 .
so boundary unitarity is violated at this other chiral point , although it is also true ( as we shall see shortly ) that some unitarity condition is violated in any neighbourhood of either chiral point .
we will first use this new geometrical framework to recover standard results for tmg .
we can do this by setting @xmath98 , in which case the mmg hyperbola becomes the tmg parabola @xmath99 the inequalities ( [ ineq4 ] ) simplify to ( i ) linear inequalities in the @xmath81-plane , @xmath100 which exclude the entire plane unless @xmath26 , plus ( ii ) the ( no - ghost ) restriction @xmath27 on the parabola parameters .
these are contradictory requirements , so _
the unitarity conditions can not be satisfied_. the inequalities ( [ ineq4 ] ) imply that @xmath101 and hence that @xmath102 but this contradicts the requirements that @xmath87 and @xmath83 unless @xmath103 equivalently , unitarity requires @xmath104 , in agreement with the result of @xcite . given @xmath53 , the second inequality of ( [ ineq4 ] ) can be written as another upper bound on @xmath105 : @xmath106 and this is a stronger upper bound than @xmath87 when @xmath107 .
otherwise it is weaker , so * @xmath108 if @xmath107 . in this case
@xmath109 but , since @xmath110 , the chiral point is outside the allowed region .
see figure [ ychpve ] .
* @xmath87 if @xmath111 . in this case @xmath112 , so the chiral point is again outside the allowed region .
see figure [ ychnve ] .
we learn from this that _ the chiral point is always outside the allowed region_. ) for @xmath113 and @xmath114 .
the region defined by the inequalities ( [ ineq4 ] ) is shaded in blue .
the chiral point is the intersection of the hyperbola with the solid blue line , which is parallel to the left asymptote.[ychpve ] ] ) for @xmath113 and @xmath115 .
the region defined by the inequalities ( [ ineq4 ] ) is shaded in blue .
note that the right - hand branch is now excluded outright ; the chiral point is at @xmath112 on this excluded branch.[ychnve ] ] the mmg hyperbola crosses the @xmath105-axis at @xmath116 and at the hyperbola degenerates into its asymptotes , but this case is excluded by definition of mmg . ]
@xmath117 for future use we record here that @xmath118 the above facts suggest that we should organize the analysis according to the signs of @xmath64 and @xmath119 .
as observed earlier , we have seen that we may choose either sign for @xmath64 without loss of generality but we will ignore this fact , for now , in order to simplify comparison with the results of @xcite .
this means that we have four generic cases to consider , plus the two special cases for which @xmath120 or @xmath121 .
each of the four generic cases corresponds ( given @xmath122 ) to a definite sign of @xmath123 which we will also need to know . 1 .
@xmath124 . in this case @xmath125 , so the right - hand branch of the hyperbola is the one passing through the origin .
the chiral point is on this branch .
the two conditions imply @xmath126 this corresponds to the third of the possibilities deduced in @xcite .
+ the allowed region in the @xmath81-plane lies to the left of the line @xmath127 , which excludes the entire right - hand branch of the hyperbola .
but @xmath128 , so the entire upper - half of the left - hand branch is inside the allowed region .
2 . @xmath129 .
again @xmath125 , so the right - hand branch passes through the origin .
this case occurs when @xmath130 this corresponds to the first of the possibilities deduced in @xcite .
+ the allowed region now lies to the left of the line @xmath93 , and under the line @xmath131 , which now passes through the chiral point on the _ lower - half _ of the right - hand branch of the hyperbola .
this again means that the entire right - hand branch is excluded while the entire upper - half of the left hand branch is allowed .
3 . @xmath132 . in this case @xmath133 , so the left - hand branch of the hyperbola is now the one passing through the origin .
this case is only realised when @xmath134 this corresponds to part of the second possibility deduced in @xcite .
+ the allowed region lies to the left of the line @xmath93 , but since @xmath135 , the entire upper - half of the left - hand branch of the hyperbola is in the allowed region .
the entire right - hand branch is excluded because the line @xmath131 passes through the chiral point on the _ lower - half _ of the right - hand branch of the hyperbola .
4 . @xmath136 .
again @xmath133 so the left - hand branch of the hyperbola passes through the origin .
this case occurs when @xmath137 this corresponds to another part of the second possibility deduced in @xcite .
+ the allowed region lies to the left of the line @xmath127 .
but @xmath138 , so the entire right - hand branch of the hyperbola is excluded while the left - hand branch is allowed .
notice that @xmath139 is @xmath60-inert , so cases @xmath140 and cases @xmath141 are both `` @xmath60-pairs '' . in all cases only the upper half of the left - hand branch is allowed by unitarity . to summarise , cases 1&2 cover @xmath113 _ except _ @xmath142 and cases 3&4 cover @xmath143 _ except _ @xmath121 .
we now deal with the two special cases @xmath142 and @xmath121 . as these are mapped into each other by @xmath60-duality , we can limit the discussion to @xmath142 without loss of generality . at @xmath49
the previous discussion goes through except that the two figures degenerate into a single figure with the chiral point moving to the origin and the `` chiral boundary line '' becoming @xmath144 .
the upshot is that the four cases have now been reduced to just _ two _ cases , related by @xmath60-duality .
these are @xmath145 and in both cases only the upper half of the left - hand branch of the mmg hyperbola is allowed by unitarity .
the results of @xcite for the regions in the mmg parameter space allowed by unitarity have been recovered , and extended to include the @xmath142 case . however , each of the three allowed parameter regions found in @xcite came with a sign restriction .
what we now aim to show is that this sign restriction is equivalent to the statement that the allowed segment of the mmg hyperbola is on its left - hand branch .
a point on the mmg hyperbola is on the left hand branch if and only if @xmath146 but a sufficient conditon that applies in either case is @xmath147 , or @xmath148 using the definition of @xmath105 given in ( [ xandy ] ) , we may rewrite this inequality as @xmath149 on the other hand , equation ( [ quad ] ) viewed as a quadratic equation for @xmath150 has the solution , as was done there . ]
@xmath151\ , .\ ] ] the sign choice appearing here enters into the unitarity analysis of @xcite . to see the implications of this sign , we substitute for @xmath150 in ( [ signineq ] ) to deduce the inequality @xmath152 as the left hand side must be positive ,
we conclude that the sign choice is correlated with the sign of @xmath153 : * @xmath154 requires the _ bottom sign_. this applies to case 1 .
* @xmath155 requires the _
top sign_. this applies to cases 2 , 3 and 4 .
this result agrees with @xcite .
cases 3&4 , together with the special case @xmath121 , jointly correspond to the second possibility deduced in @xcite .
cases 1&2 correspond to the other two possibilities found in @xcite .
because of the @xmath60-duality , we lose no generality by focusing on the cases @xmath156
if we also allow for @xmath49 . then @xmath66 and every point on the left - hand mmg hyperbola with @xmath83 has @xmath157 . proceeding as before we now find that @xmath158 and by squaring we deduce that @xmath159 each choice of @xmath160 satisfying this inequality corresponds to one point on the upper left - hand branch of the mmg hyperbola .
we also learn that @xmath161 which confirms the reality of the square root in the expression on the left hand side of ( [ proceed ] ) .
recall that @xmath162 .
this shows that the points at which the mmg hyperbola crosses the @xmath105-axis correspond to flat space limits of ads@xmath92 .
there are two such points , one on the right - hand branch and another on the left - hand branch .
the flat space limit on the left - hand branch is on the boundary of the region allowed by bulk and boundary unitarity . strictly speaking , this boundary point in the full parameter space ; this becomes a curve after a choice of normalisation of parameters .
] is not in the allowed region .
although the bulk unitarity conditions have a smooth flat - space limit , the asymptotic symmetry algebra changes . it has been argued in @xcite that the holographic dual to a 3d bulk gravity theory with flat - space asymptotics should be a theory invariant under the 2d `` galilean conformal algebra '' ( gca ) .
this is a wigner - inn contraction of the more usual virasoro@xmath163virasoro symmetry algebra .
the _ non - zero _ commutators of the gca are @xmath164 & = & ( m - n ) l_{m+n } + \frac{c_1}{12 } \left(m^3-m\right ) \delta_{m+n } \nonumber \\ \left [ m_n , l_n \right ] & = & ( m - n ) m_{m+n } + \frac{c_2}{12}\left(m^3-m\right ) \delta_{m+n}\ , , \end{aligned}\ ] ] and the central charges @xmath165 and @xmath166 are limits of linear combinations of the central charges @xmath167 of the 2d conformal algebra : @xmath168 \ , .\ ] ] for mmg we have , using the formula ( [ cpm1 ] ) and the fact that @xmath169 as @xmath170 for fixed @xmath8 , @xmath171 where @xmath172 is one of the two values of @xmath105 on the mmg hyperbola when @xmath173 ; i.e. _ either _ @xmath174 _ or _
as explained in subsection [ subsec : duality ] , we may choose the mmg parameters so that @xmath66 without loss of generality . making this choice (
because the tmg limit is then simple ) we have @xmath125 , so that @xmath174 on the right - hand branch of mmg hyperbola and @xmath175 on the left - hand branch .
thus @xmath176 as only the right - hand branch survives ( as a parabola ) in the tmg limit , we deduce that @xmath177 for flat - space tmg , in agreement with @xcite . on the left - hand branch
we have @xmath178 not only is this a new possibility , not available to tmg , but also it is the value of @xmath166 found by choosing the flat - space limit on the boundary of the segment of the mmg hyperbola allowed by unitarity .
it was argued in @xcite that , under certain conditions , the 2d gca algebra admits a unitary truncation to one copy of the virasoro algebra and the boundary theory becomes the chiral half of a cft .
specifically , the conditions are that @xmath179 where @xmath180 is the eigenvalue of the @xmath181 generator , equal to the graviton mass . in @xcite
these conditions were satisfied by taking a @xmath182 limit in which flat - space tmg degenerates to 3d conformal gravity .
as just observed , mmg has _ two _ flat - space limits , one on the left - hand branch of the mmg hyperbola and one on the right - hand branch , but the condition @xmath183 can not be satisfied on the left - hand branch .
however , setting @xmath142 makes both @xmath166 and @xmath180 vanish at the flat - space limit on the right - hand branch . at this point
, the mmg equation degenerates to @xmath184 taking the trace we deduce that @xmath185 the left hand side of this equation is the curvature - squared scalar of new massive gravity ( nmg ) @xcite .
there are some similarities here with `` new topologically massive gravity '' ( ntmg ) @xcite for which the field equation is the same as ( [ flat ] ) but with the @xmath186-tensor of nmg replacing the @xmath36-tensor of mmg ; as observed in @xcite , both tensors have the same scalar trace .
the @xmath36-tensor does not contribute to linear order in the equation ( [ flat ] ) , so its linearisation ( about flat space ) is just linearised 3d conformal gravity , which does not propagate any modes .
in contrast , the non - linear theory does have local degrees of freedom ; a slight modification of the hamiltonian analysis of @xcite ( to allow for @xmath49 ) shows that the physical phase - space of the non - linear theory has dimension @xmath0 per space point .
there is therefore a linearisation instability of the mmg model with field equation ( [ flat ] ) , analogous to the linearisation instability of ntmg @xcite .
linearisation instabilities of this type occur when the linearised theory has a gauge invariance that is explicitly broken by interactions in the full theory .
this phenomenon does not occur for tmg ( the linearisation instability at the chiral point discussed in @xcite is of another type ) but it does occur for nmg at partially massless ads vacua @xcite , where the accidental gauge invariance of the linearised theory is again linearized weyl invariance .
the ( unique ) ads vacuum at the `` merger point '' of mmg is `` partially massless '' , implying a linearisation instability in that case too @xcite ; in fact , equation ( [ flat ] ) is the flat space limit of the mmg equation at its merger point .
we have presented a geometrical re - analysis of the bulk and boundary unitarity condition found in @xcite for the 3d `` minimal massive gravity '' ( mmg ) model with ads@xmath92 asymptotics .
the four ( un - normalized ) dimensionless parameters of the mmg action are considered in two pairs ; one pair ( essentially , the cosmological constant and cosmological parameter in units of the square of the tmg mass parameter ) parametrises a plane
. the unitarity conditions are linear inequalities in the plane but an ads@xmath92 vacuum exists only for points lying on a hyperbola in this plane .
we have called this the mmg hyperbola ; it degenerates to a parabola in the tmg limit .
both the linear inequalities and the hyperbola depend on the other two parameters , which are the coefficient @xmath51 of the einstein - hilbert term and the coefficient @xmath31 of the new mmg term , which are restricted by the condition @xmath63 .
the linear inequalities are inconsistent unless @xmath53 , in which case they determine an `` allowed '' region in the plane . in all cases ,
only the upper left - hand branch of the mmg hyperbola is in this region , and this corresponds to a sign choice in the quadratic equation that gives the ads radius in terms of a cosmological parameter @xmath160 , which is restricted only by a simple inequality . in our analysis ,
the values of @xmath51 for which @xmath187 are special , and this leads to four cases , depending on the signs of @xmath139 and @xmath64 .
only three cases arose in the analysis of @xcite because @xmath49 was excluded but @xmath188 was not singled out as special . once
both the @xmath49 and @xmath188 cases are included , the number of connected parameter regions allowed by unitarity is reduced to two which differ according to the sign of @xmath64 .
however , we have shown that any mmg model with @xmath66 is equivalent to some mmg model with @xmath65 , and _
vice versa_. we may therefore choose either sign for @xmath64 without loss of generality .
the choice @xmath66 has the advantage that it simplifies the tmg limit , and the net result for this choice is that the only restrictions on the parameter space imposed by unitarity ( given ads@xmath92 asymptotics ) are , up to equivalence , @xmath189 this defines _ a single connected region in the space of parameters of the mmg action _ of the mmg equation ; @xmath53 implies @xmath104 but to find the analogous explicit restrictions on @xmath190 and @xmath191 would require an explicit inversion of the map defined by equations ( [ bars ] ) and ( [ barlamzero ] ) , and this is known only for @xmath50 @xcite . ] .
it excludes tmg .
if we fix the scaling invariance ( induced by a rescaling of the 3d newton constant ) by choosing @xmath54 then the above unitarity restriction reduces ( up to equivalence ) to @xmath192 of course , this is only a necessary condition for the unitarity of any 2d cft dual to mmg , but that is all we can test at the level of semi - classical effective 3d gravity , and it is a test that tmg fails but mmg passes .
we have also used our results to investigate minkowski vacua in the context of the particular proposal for flat - space holography put forward in @xcite , and to compare with the situation for tmg .
one new feature of mmg is the existence of two flat space limits .
one is a minkowski limit of an mmg model with ads asymptotics satisfying all unitarity conditions , but this does not appear to be useful . instead , the flat - space limit that is available to both tmg and mmg is the one for which proposed unitarity conditions on central charges of the gca algebra can be satisfied . in this limit tmg
degenerates to 3d conformal gravity but mmg degenerates to a very different theory with local degrees of freedom , although this are not apparent from the linearised theory .
it thus appears that mmg has implications not only for ads holography but also for flat - space holography .
a.s.a . and p.k.t .
acknowledge support from the uk science and technology facilities council ( grant st / l000385/1 ) .
a.s.a . also acknowledges support from clare hall , cambridge , and from the cambridge trust .
we are grateful to eduardo casali , elias kiritsis and alasdair routh for helpful discussions , and to eric bergshoeff and wout merbis for helpful correspondence .
e. bergshoeff , o. hohm , w. merbis , a. j. routh and p. k. townsend , `` minimal massive 3d gravity , '' class .
quantum grav . 31
( 2014 ) 145008 [ arxiv:1404.2867 [ hep - th ] ] .
s. deser , r. jackiw , and s. templeton , `` topologically massive gauge theories , '' _ annals phys . _
* 140 * ( 1982 ) 372411 .
a. bagchi , r. gopakumar , i. mandal and a. miwa , `` gca in 2d , '' jhep * 1008 * ( 2010 ) 004 [ arxiv:0912.1090 [ hep - th ] ] .
a. bagchi and r. fareghbal , `` bms / gca redux : towards flatspace holography from non - relativistic symmetries , '' jhep * 1210 * , 092 ( 2012 ) [ arxiv:1203.5795 [ hep - th ] ] .
a. bagchi , s. detournay and d. grumiller , `` flat - space chiral gravity , '' phys .
* 109 * ( 2012 ) 151301 [ arxiv:1208.1658 [ hep - th ] ] .
d. grumiller , r. jackiw and n. johansson , `` canonical analysis of cosmological topologically massive gravity at the chiral point , '' in _ fundamental interactions : a memorial volume for wolfgang kummer _ , editors : daniel grumiller , anton rebhan and dimitri vassilevich , world scientific , 2010 , pp.363 - 374 arxiv:0806.4185 [ hep - th ] .
s. carlip , `` the constraint algebra of topologically massive ads gravity , '' jhep * 0810 * ( 2008 ) 078 [ arxiv:0807.4152 [ hep - th ] ] .
w. li , w. song and a. strominger , `` chiral gravity in three dimensions , '' jhep * 0804 * ( 2008 ) 082 [ arxiv:0801.4566 [ hep - th ] ] .
b. tekin , `` minimal massive gravity : conserved charges , excitations and the chiral gravity limit , '' phys . rev .
d * 90 * 081701 [ arxiv:1409.5358 [ hep - th ] ] .
e. a. bergshoeff , o. hohm and p. k. townsend , `` massive gravity in three dimensions , '' phys .
* 102 * ( 2009 ) 201301 [ arxiv:0901.1766 [ hep - th ] ] .
r. andringa , e. a. bergshoeff , m. de roo , o. hohm , e. sezgin and p. k. townsend , `` massive 3d supergravity , '' class .
* 27 * ( 2010 ) 025010 [ arxiv:0907.4658 [ hep - th ] ] .
d. dalmazi and e. l. mendona , `` a new spin-2 self - dual model in d=2 + 1 , '' jhep * 0909 * ( 2009 ) 011 [ arxiv:0907.5009 [ hep - th ] ] .
o. hohm , a. routh , p. k. townsend and b. zhang , `` on the hamiltonian form of 3d massive gravity , '' phys .
d * 86 * ( 2012 ) 084035 [ arxiv:1208.0038 ] .
m. blagojevic and b. cvetkovic , `` extra gauge symmetries in bht gravity , '' jhep * 1103 * ( 2011 ) 139 [ arxiv:1103.2388 [ gr - qc ] ] .
a. s. arvanitakis , `` on solutions of minimal massive 3d gravity , '' arxiv:1501.01808 [ hep - th ] . | a geometrical analysis of the bulk and anti - de sitter boundary unitarity conditions of 3d `` minimal massive gravity '' ( mmg ) ( which evades the `` bulk / boundary clash '' of topologically massive gravity ) is used to extend and simplify previous results , showing that unitarity selects , up to equivalence , a connected region in parameter space .
we also initiate the study of flat - space holography for mmg .
its relevant flat space limit is a deformation of 3d conformal gravity ; the deformation is both non - linear and non - conformal , implying a linearisation instability .
= 4 damtp-2014 - 80 |
a channel @xmath1 with noncausal state information at the sender has capacity @xmath7 as shown by gelfand and pinsker @xcite .
transmitting at capacity , however , obscures the state information @xmath3 as received by the receiver @xmath4 . in some instances we wish to convey the state information @xmath3 itself , which could be time - varying fading parameters or an original image that we wish to enhance .
for example , a stage actor with face @xmath8 uses makeup @xmath9 to communicate to the back row audience @xmath10 . here
@xmath9 is used to enhance and exaggerate @xmath8 rather than to communicate new information .
another motivation comes from cognitive radio systems @xcite with the additional assumption that the secondary user @xmath11 communicates its own message and at the same time facilitates the transmission of the primary user s signal @xmath3 .
how should the transmitter communicate over the channel to `` amplify '' his knowledge of the state information to the receiver ? what is the optimal tradeoff between state amplification and independent information transmission ? to answer these questions , we study the communication problem depicted in figure [ fig : setup ] . here
the sender has access to the channel state sequence @xmath12 , independent and identically distributed ( i.i.d . ) according to @xmath13 , and wishes to transmit a message index @xmath14 : = \{1 , 2 , \ldots , 2^{nr}\}$ ] , independent of @xmath3 , as well as to help the receiver reduce the uncertainty about the channel state in @xmath15 uses of a state dependent channel @xmath16 .
based on the message @xmath17 and the channel state @xmath3 , the sender chooses @xmath18 and transmits it across the channel . upon observing the channel output @xmath4
, the receiver guesses @xmath19 $ ] and forms a list @xmath20 that contains likely candidates of the actual state sequence @xmath3 . without any observation @xmath4 ,
the receiver would know only that the channel state @xmath3 is one of @xmath21 typical sequences ( with almost certainty ) and we can say the uncertainty about @xmath3 is @xmath22 . now upon observing @xmath4 and forming a list @xmath23 of likely candidates for @xmath3 , the receiver s list size is reduced from @xmath24 to @xmath25 .
thus we define the _ channel state uncertainty reduction rate _ to be @xmath26 as a natural measure for the amount of information the receiver learns about the channel state . in other words ,
the uncertainty reduction rate @xmath27 $ ] captures the difference between the original channel state uncertainty and the residual state uncertainty after observing the channel output . later in section [ sec : gaussian ] we will draw a connection between the list size reduction and the conventional information measure @xmath2 that also captures the amount of information @xmath4 learns about @xmath3 .
more formally , we define a @xmath28 code as the encoder map @xmath29
\times \mathcal{s}^n \to \mathcal{x}^n\ ] ] and decoder maps @xmath30\\ l_n & : \mathcal{y}^n \to 2^{\mathcal{s}^n}\end{aligned}\ ] ] with list size @xmath31 the probability of a message decoding error @xmath32 and the probability of a list decoding error @xmath33 are defined respectively as @xmath34 where the message index @xmath17 is chosen uniformly over @xmath35 $ ] and the state sequence @xmath3 is drawn i.i.d .
@xmath36 , independent of @xmath17 .
a pair @xmath37 is said to be achievable if there exists a sequence of @xmath28 codes with @xmath38 and @xmath39 as @xmath40 . finally , we define the optimal @xmath41 tradeoff region , or the _ tradeoff region _ in short , to be the closure of all achievable @xmath41 pairs , and denote it by @xmath42 .
this paper shows that the tradeoff region @xmath42 can be characterized as the union of all @xmath37 pairs satisfying @xmath43 for some joint distribution of the form @xmath44 . as a special case ,
if the encoder s sole goal is to `` amplify '' the state information ( @xmath45 ) , then the maximum uncertainty reduction rate @xmath46 is given by @xmath47 the maximum uncertainty reduction rate @xmath48 is achieved by designing the signal @xmath11 to enhance the receiver s estimation of the state @xmath3 while using the remaining pure information bearing freedom in @xmath11 to provide more information about the state .
more specifically , there are three different components involved in reducing the receiver s uncertainty about the state : 1 .
the transmitter uses the channel capacity to convey the state information . in section [ sec : noncausal ] , we study the classical setup @xcite of coding for memory with defective cells ( example [ ex : memory ] ) and show that this `` source - channel separation '' scheme is optimal when the memory defects are symmetric .
2 . the transmitter gets out of the way of the receiver s view of the state .
for instance , the maximum uncertainty reduction for the binary multiplying channel @xmath49 ( example [ ex : binary ] in section [ sec : noncausal ] ) with binary input @xmath50 and binary state @xmath51 is achieved by sending @xmath52 .
the transmitter actively amplifies the state . in example
[ ex : wdp ] in section [ sec : gaussian ] , we consider the gaussian channel @xmath53 with gaussian state @xmath8 and gaussian noise @xmath54 . here
the optimal transmitter amplifies the state as @xmath55 under the given power constraint @xmath56 .
it is interesting to note that the maximum uncertainty reduction rate @xmath48 is the information rate @xmath57 that could be achieved if both the state @xmath8 and the signal @xmath9 could be freely designed , instead of the state @xmath8 being generated by nature .
this rate also appears in the sum rate of the capacity region expression for the cooperative multiple access channel ( * ? ? ?
* problem 15.1 ) and the multiple access channel with cribbing encoders by willems and van der meulen @xcite .
when the state information is only _ causally _ available at the transmitter , that is , when the channel input @xmath58 depends on only the past and the current channel channel state @xmath59 , we will show that the tradeoff region @xmath42 is given as the union of all @xmath37 pairs satisfying @xmath60 over all joint distributions of the form @xmath61 .
interestingly , the maximum uncertainty reduction rate @xmath48 stays the same as in the noncausal case .
that causality incurs no cost on the ( sum ) rate is again reminiscent of the multiple access channel with cribbing encoders @xcite . the problem of communication over state - dependent channels with state information known at the sender has attracted a great deal of attention .
this research area was first pioneered by shannon @xcite , kuznetsov and tsybakov @xcite , and gelfand and pinsker @xcite .
several advancements in both theory and practice have been made over the years .
for instance , heegard and el gamal @xcite characterized the channel capacity and devised practical coding techniques for computer memory with defective cells .
costa @xcite studied the now famous `` writing on dirty paper '' problem and showed that the capacity of an additive white gaussian noise channel is not affected by additional interference , as long as the entire interference sequence is available at the sender prior to the transmission .
this fascinating result has been further extended with strong motivations from applications in digital watermarking ( see , for example , moulin and osullivan @xcite , chen and wornell @xcite , and cohen and lapidoth @xcite ) and multi - antenna broadcast channels ( see , for example , caire and shamai @xcite , weingarten , steinberg , and shamai @xcite , and mohseni and cioffi @xcite ) .
readers are referred to caire and shamai @xcite , lapidoth and narayan @xcite , and jafar @xcite for more complete reviews on the theoretical development of the field . on the practical side , erez , shamai , and
zamir @xcite proposed efficient coding schemes based on lattice strategies for binning .
more recently , erez and ten brink @xcite report efficient coding techniques that almost achieve the capacity of costa s dirty paper channel . in @xcite , we formulated the problem of simultaneously transmitting pure information and helping the receiver estimate the channel state under a distortion measure .
although the characterization of the optimal rate - distortion tradeoff is still open in general ( cf .
@xcite ) , a complete solution is given for the gaussian case ( the writing on dirty paper channel ) under quadratic distortion @xcite .
in this particular case , optimality was shown for a simple power - sharing scheme between pure information transmission via costa s original coding scheme and state amplification via simple scaling .
recently , merhav and shamai @xcite considered a related problem of transmitting pure information , but this time under the additional requirement of _ minimizing _ the amount of information the receiver can learn about the channel state . in this interesting work , the optimal tradeoff between pure information rate @xmath0 and the amount of state information @xmath62 is characterized for both causal and noncausal setups .
furthermore , for the gaussian noncausal case ( writing on dirty paper ) , the optimal rate - distortion tradeoff is given under quadratic distortion .
( this may well be called `` writing dirty on paper '' . ) the current paper thus complements @xcite in a dual manner
. it is refreshing to note that our notion of uncertainty reduction rate @xmath5 is essentially equivalent to merhav and shamai s notion of @xmath62 ; both notions capture the normalized mutual information @xmath63 .
( see the discussion in section [ sec : gaussian ] . )
the crucial difference is that @xmath5 is to be maximized while @xmath62 is to be minimized .
both problems admit single - letter optimal solutions . the rest of this paper is organized as follows . in the next section ,
we establish the optimal @xmath41 tradeoff region for the case in which the state information @xmath3 is noncausally available at the transmitter before the actual communication .
section [ sec : gaussian ] extends the notion of state uncertainty reduction to continuous alphabets , by identifying the list decoding requirement @xmath64 with the mutual information rate @xmath65 .
in particular , we characterize the optimal @xmath37 tradeoff region for costa s `` writing on dirty paper '' channel .
since the intuition gained from the study of the noncausal setup carries over when the transmitter has causal knowledge of the state sequence , the causal case is treated only briefly in section [ sec : causal ] , followed by concluding remarks in section [ sec : conc ] .
in this section , we characterize the optimal tradeoff region between the pure information rate @xmath0 and the state uncertainty reduction rate @xmath5 with state information noncausally available at the transmitter , as formulated in section [ sec : intro ] .
[ thm : noncausal ] the tradeoff region @xmath42 for a state - dependent channel @xmath66 with state information @xmath3 noncausally known at the transmitter is the union of all @xmath41 pairs satisfying @xmath67 for some joint distribution of the form @xmath44 , where the auxiliary random variable @xmath68 has cardinality bounded by @xmath69 . as will be clear from the proof of the converse , the region given by is convex .
( we can merge the time - sharing random variable into @xmath68 . )
since the auxiliary random variable @xmath68 affects the first inequality only , the cardinality bound on @xmath70 follows directly from the usual technique ; see gelfand and pinsker @xcite or a general treatment by salehi @xcite .
finally , we can take @xmath9 as a deterministic function of @xmath71 without reducing the region , but at the cost of increasing the cardinality bound of @xmath68 ; refer to the proof of lemma [ lemma : equiv ] below .
it is easy to see that we can recover the gelfand
pinsker capacity formula @xmath72 for the other extreme case of pure state amplification , we have the following result .
[ corollary : noncausal ] under the condition of theorem [ thm : noncausal ] , the maximum uncertainty reduction rate @xmath73 is given by @xmath74 thus the receiver can learn about the state @xmath3 essentially at the maximal cut - set rate @xmath75 before we prove theorem [ thm : noncausal ] , we need the following two lemmas .
the first one extends fano s inequality ( * ? ? ?
* lemma 7.9.1 ) to list decoding .
[ lemma : gen - fano ] for a sequence of list decoders @xmath76 @xmath77 with list size @xmath78 fixed for each @xmath15 , let @xmath79 be the sequence of corresponding probabilities of list decoding error . if @xmath80 , then @xmath81 where @xmath82 as @xmath83 .
define an error random variable @xmath62 as @xmath84 we can then expand @xmath85 note that @xmath86 and @xmath87 .
we can also bound @xmath88 as @xmath89 where the inequality follows because when there is no error , the remaining uncertainty is at most @xmath90 , and when there is an error , the uncertainty is at most @xmath91 .
this implies that @xmath92 taking @xmath93 proves the desired result .
the second lemma is crucial to the proof of theorem [ thm : noncausal ] and contains a more interesting technique than lemma [ lemma : gen - fano ] .
this lemma shows that the third inequality can be replaced by a tighter inequality below ( recall that @xmath94 since @xmath95 ) , which becomes crucial for the achievability proof of theorem [ thm : noncausal ] .
[ lemma : equiv ] let @xmath96 be the union of all @xmath37 pairs satisfying . let @xmath97 be the closure of the union of all @xmath37 pairs satisfying @xmath98 for some joint distribution @xmath99 , where the auxiliary random variable @xmath68 has finite cardinality . then @xmath100 since @xmath95 forms a markov chain , it is trivial to check that @xmath101 for the other direction of inclusion , we need some notation .
let @xmath102 be the set of all distributions of the form @xmath103 consistent with the given @xmath13 and @xmath1 , where the auxiliary random variable @xmath68 is defined on an arbitrary finite set .
further let @xmath104 be the restriction of @xmath102 such that @xmath105 for some function @xmath106 , i.e. , @xmath107 takes values @xmath108 or @xmath109 only .
if we define @xmath110 to denote the closure of all @xmath37 pairs satisfying , , and over @xmath104 , or equivalently , if @xmath110 is defined to be the restriction of @xmath111 over a smaller set of distributions @xmath104 , then clearly @xmath112 let @xmath113 be defined as the closure of @xmath41 pairs satisfying . since @xmath114 forms a markov chain on @xmath104 , we have @xmath115 to complete the proof , it now suffices to show that @xmath116 to see this , we restrict @xmath113 to the distributions of the form @xmath117 with @xmath118 independent of @xmath119 , namely , @xmath120 with deterministic @xmath121 , i.e. , @xmath122 is a function of @xmath123 , and call this restriction @xmath124 . since @xmath9 is a deterministic function of @xmath125 and at the same time @xmath126 form a markov chain , @xmath124 can be written as the closure of all @xmath41 pairs satisfying @xmath127 for some distribution of the form @xmath128 satisfying .
but we have @xmath129 and the set of conditional distributions on @xmath130 given @xmath8 satisfying is as rich as any @xmath131 .
( indeed , any conditional distribution @xmath132 can be represented as @xmath133 for appropriately chosen @xmath134 and _ deterministic _ distribution @xmath135 with cardinality of @xmath136 upper bounded by @xmath137 ; see also ( * ? ? ?
* eq . ( 44 ) ) . )
therefore , we have @xmath138 which completes the proof .
now we are ready to prove theorem [ thm : noncausal ] . for the proof of achievability , in the light of lemma [ lemma : equiv ] , it suffices to prove that any pair @xmath139 satisfying , , for some @xmath140 is achievable . since the coding technique is quite standard , we only sketch the proof here .
for fixed @xmath140 , the result of gelfand
pinsker @xcite shows that the transmitter can send @xmath141 bits reliably across the channel .
now we allocate @xmath142 bits for sending the pure information and use the remaining @xmath143 bits for sending the state information by random binning .
more specifically , we assign typical @xmath3 sequences to @xmath144 bins at random and send the bin index of the observed @xmath3 using @xmath145 bits . at the receiving end ,
the receiver is able to decode the codeword @xmath146 from @xmath4 with high probability . using joint typicality of @xmath147
, the state uncertainty can be first reduced from @xmath148 to @xmath149 . indeed , the number of typical @xmath3 sequences jointly typical with @xmath150 is bounded by @xmath151 .
in addition , using @xmath143 bits of independent refinement information from the hash index of @xmath3 , we can further reduce the state uncertainty by @xmath152 .
hence , by taking the list of all @xmath3 sequences jointly typical with @xmath150 satisfying the hash check , we have the total state uncertainty reduction rate @xmath153 by varying @xmath142 , it can be readily seen that all @xmath37 pairs satisfying @xmath154 for any fixed @xmath155 are achievable . for the proof of converse , we have to show that given any sequence of @xmath28 codes with @xmath156 the @xmath41 pairs must satisfy @xmath157 for some joint distribution @xmath158 .
the pure information rate @xmath0 can be readily bounded from the previous work by gelfand and pinsker ( * ? ? ?
* proposition 3 ) .
here we repeat a simpler proof given in heegard ( * ? ? ?
* appendix 2 ) for completeness ; see also ( * ? ?
? * lecture 13 ) . starting with fano s inequality
, we have the following chain of inequalities : @xmath159 where ( a ) follows from the csiszr sum formula @xmath160 and ( b ) follows because @xmath161 is independent of @xmath162 . by recognizing the auxiliary random variable @xmath163 and noting that @xmath164 form a markov chain
, we have @xmath165 on the other hand , since @xmath166 , we can trivially bound @xmath5 by lemma [ lemma : gen - fano ] as @xmath167 similarly , we can bound @xmath168 as @xmath169 where ( a ) follows since @xmath17 is independent of @xmath3 and conditioning reduces entropy , ( b ) follows from the data processing inequality ( both directions ) , and ( c ) follows from the memorylessness of the channel .
we now introduce the usual time - sharing random variable @xmath170 uniform over @xmath171 , independent of everything else . then implies @xmath172 on the other hand , implies @xmath173 where the last equality follows since @xmath174 form a markov chain .
finally , we recognize @xmath175 @xmath176 and note that @xmath177 , @xmath178 , and @xmath179 , which completes the proof of the converse
. roughly speaking , the optimal coding scheme is equivalent to sending the codeword @xmath146 reliably at the gelfand pinsker rate @xmath180 and reducing the receiver s uncertainty by @xmath181 from @xmath4 and the decoded codeword @xmath146 .
it should be noted that @xmath182 has the same form as the achievable region for the dual tradeoff problem between pure information rate @xmath0 and ( minimum ) normalized mutual information rate @xmath183 studied in @xcite .
but we can reduce the uncertainty about @xmath3 further by allocating part @xmath152 of the pure information rate @xmath184 to convey independent refinement information ( hash index of @xmath3 ) . by varying @xmath185 $ ]
we can trace the entire tradeoff region @xmath186 .
it turns out an alternative coding scheme based on wyner
ziv source coding with side information @xcite , instead of random binning , also achieves the tradeoff region @xmath42 . to see this ,
fix any @xmath140 and @xmath187 satisfying @xmath188 and consider the wyner ziv encoding of @xmath3 with covering codeword @xmath189 and side information @xmath190 at the decoder .
more specifically , we can generate @xmath191 @xmath189 codewords and assign them into @xmath144 bins . as before we use the gelfand
pinsker coding to convey a message of rate @xmath141 reliably over the channel .
since the rate @xmath192 is sufficient to reconstruct @xmath189 at the receiver with side information @xmath4 and @xmath146 , we can allocate the rate @xmath152 for conveying @xmath189 and use the remaining rate @xmath193 for extra pure information . forming a list of @xmath3 jointly typical with @xmath194 results in the uncertainty reduction rate @xmath5 given by @xmath195
thus the tradeoff region @xmath196 can be achieved via the combination of two fundamental results in communication with side information : channel coding with side information by gelfand and pinsker @xcite and rate distortion with side information by wyner and ziv @xcite .
it is also interesting to note that the information about @xmath3 can be transmitted in a manner completely independent of geometry ( random binning ) or completely dependent on geometry ( random covering ) ; refer to @xcite for a similar phenomenon in a relay channel problem .
when @xmath10 is a function of @xmath197 , it is optimal to identify @xmath198 and theorem [ thm : noncausal ] simplifies to the following corollary .
[ coro : deterministic ] the tradeoff region @xmath42 for a deterministic state - dependent channel @xmath199 with state information @xmath3 noncausally known at the transmitter is the union of all @xmath41 pairs satisfying @xmath200 for some joint distribution of the form @xmath201 . in particular , the maximum uncertainty reduction rate is given by @xmath202 the next two examples show different flavors of optimal state uncertainty reduction . [
ex : memory ] consider the problem of conveying information using a write - once memory device with stuck - at defective cells @xcite as depicted in figure [ fig : memory ] .
here each memory cell has probability @xmath203 of being stuck at @xmath108 , probability @xmath204 of being stuck at @xmath109 , and probability @xmath205 of being a good cell , with @xmath206 .
it is easy to see that the channel output @xmath10 is a simple deterministic function of the channel input @xmath9 and the state @xmath8 .
now it is easy to verify that the tradeoff region @xmath42 is given by @xmath207 where @xmath208 can be chosen arbitrarily ( @xmath209 ) .
this region is achieved by choosing @xmath210 . without loss of generality , we can choose @xmath211 independent of @xmath8 , because the input @xmath9 affects @xmath10 only when @xmath212 .
there are two cases to consider . 1 .
if @xmath213 , then the choice of @xmath214 maximizes both and , and hence achieves the entire tradeoff region @xmath42 .
the optimal transmitter splits the full channel capacity @xmath215 to send both the pure information and the state information .
( see figure [ fig : graph](a ) for the case @xmath216 . ) 2 .
on the other hand , when @xmath217 , there is a clear tradeoff in our choice of @xmath208 .
for example , consider the case @xmath218 .
if the goal is to communicate pure information over the channel , we should take @xmath214 to maximize the number of distinguishable input preparations .
this gives the channel capacity @xmath219 .
if the goal is , however , to help the receiver reduce the state uncertainty , we take @xmath220 , i.e. , we transmit a fixed signal @xmath221 .
this way , the transmitter can minimize his interference with the receiver s view of the state @xmath8 .
the entire tradeoff region is given in figure [ fig : graph](b ) .
[ ex : binary ] consider the binary multiplying channel @xmath222 , where the output @xmath10 is the product of the input @xmath50 and the state @xmath223 .
we assume that the state sequence @xmath3 is drawn i.i.d . according to @xmath224 .
it can be easily shown that the optimal tradeoff region is given by @xmath225 this is achieved by @xmath210 , independent of @xmath8 .
as in example [ ex : memory](b ) , there is a tension between the pure information transmission and the state amplification . when the goal is to maximize the pure information rate , we should choose @xmath226 to achieve the capacity @xmath227 .
but when the goal is to maximize the state uncertainty reduction rate , we should choose @xmath228 ( @xmath52 ) to achieve @xmath229 . in words , to maximize the state uncertainty reduction rate ,
the transmitter simply clears the receiver s view of the state .
the previous section characterized the tradeoff region @xmath42 between the pure information rate @xmath0 and the state uncertainty reduction rate @xmath230 .
apparently the notion of uncertainty reduction rate @xmath5 is meaningful only when the channel state @xmath8 has finite cardinality ( i.e. , @xmath231 ) , or at least when @xmath232 .
however , from the proof of theorem [ thm : noncausal ] ( the generalized fano s inequality in lemma [ lemma : gen - fano ] ) , along with the fact that the optimal region is single - letterizable , we can take an alternative look at the notion of state uncertainty reduction as reducing the list size from @xmath21 to @xmath233 .
we will show shortly in proposition [ prop : noncausal ] that the difference @xmath234 of the normalized list size is essentially equivalent to the normalized mutual information @xmath235 , which is well - defined for an arbitrary state space @xmath236 and captures the amount of information the receiver @xmath4 can learn about the state @xmath3 ( or lack thereof @xcite ) .
hence , the physically motivated notion @xmath5 of list size reduction is consistent with the mathematical information measure @xmath237 , and both notions of state uncertainty reduction can be used interchangeably , especially when @xmath236 is finite . to be more precise ,
we define a @xmath238 code by an encoding function @xmath29 \times \mathcal{s}^n \to \mathcal{x}^n\ ] ] and a decoding function @xmath239.\ ] ] then the associated state uncertainty reduction rate for the @xmath240 code is defined as @xmath241 where the mutual information is with respect to the joint distribution @xmath242 induced by @xmath243 with message @xmath17 distributed uniformly over @xmath35 $ ] , independent of @xmath3 .
similarly , the probability of error is defined as @xmath244 a pair @xmath37 is said to be achievable if there exists a sequence of @xmath238 codes with @xmath245 and @xmath246 the closure of all achievable @xmath37 pairs is called the tradeoff region @xmath247 .
( here we use the notation @xmath247 instead of @xmath42 to temporarily distinguish this from the original problem formulated in terms of the list size reduction . )
we now show that the optimal tradeoff @xmath248 between the information transmission rate @xmath0 and the mutual information rate @xmath5 has the same solution as the optimal tradeoff @xmath42 between @xmath0 and the list size reduction rate @xmath5 .
[ prop : noncausal ] the tradeoff region @xmath247 for a state - dependent channel @xmath66 with state information @xmath3 noncausally known at the transmitter is the closure of all @xmath41 pairs satisfying @xmath249 for some joint distribution of the form @xmath44 with auxiliary random variable @xmath68 .
hence , @xmath248 has the identical characterization as @xmath42 in theorem [ thm : noncausal ] .
let @xmath250 be the region described by . we provide a sandwich proof @xmath251 , which is given implicitly in the proof of theorem [ thm : noncausal ] .
more specifically , consider a finite partition and @xmath10 is defined as @xmath252_p ; [ y]_q),$ ] where the supremum is over all finite partitions @xmath253 and @xmath170 ; see kolmogorov @xcite and pinsker @xcite . ] to quantize the state random variable @xmath8 into @xmath254 $ ] . under this partition ,
let @xmath255}^{**}$ ] be the set of all @xmath37 pairs satisfying @xmath256 ) \\
\delta & \le h([s ] ) \\
r + \delta & \le i(x , [ s ] ; y)\end{aligned}\ ] ] for some joint distribution of the form @xmath257)p(u , x|[s])p(y|x,[s])$ ] with auxiliary random variable @xmath68 .
consider the original list size reduction problem with state information @xmath254 $ ] and let @xmath255}^*$ ] denote the tradeoff region .
then theorem [ thm : noncausal ] shows that @xmath255}^ { * * } = \mathcal{r}_{[s]}^*$ ] . in particular , for any @xmath258 and @xmath259}^{**}$
] , there exists a sequence of @xmath260 codes @xmath261 such that @xmath262 and @xmath263^n \ne l_n(y^n ) ) \to 0 $ ] .
now from the generalized fano s inequality ( lemma [ lemma : gen - fano ] ) , the achievable list size reduction rate @xmath264 should satisfy @xmath265^n ; y^n ) + n { \epsilon}_n \le i(s^n ; y^n ) + n{\epsilon}_n\ ] ] with @xmath82 as @xmath83 . hence by letting @xmath83 and @xmath266 we have from the definition of @xmath248 that @xmath267}^ { * * } = \mathcal{r}_{[s]}^ * \subseteq \mathcal{r}_i^*.\ ] ] also it follows trivially from repeating the intermediate steps in the converse proof of theorem [ thm : noncausal ] that @xmath268 . finally taking a sequence of partitions with mesh @xmath269 and hence letting @xmath255}^ { * * } \to \mathcal{r}^{**}$ ] , we have the desired result .
since both notions of state uncertainty reduction , the list size reduction @xmath270 and the mutual information @xmath63 , lead to the same answer , we will subsequently use them interchangeably and denote the tradeoff region by the same symbol @xmath42 .
[ ex : wdp ] consider costa s writing on dirty paper model depicted in figure [ fig : wdp ] as the canonical example of a continuous state - dependent channel . here
the channel output is given by @xmath271 , where @xmath18 is the channel input subject to a power constraint @xmath272 , @xmath273 is the additive white gaussian state , and @xmath274 is the white gaussian noise .
we assume that @xmath3 and @xmath275 are independent . for the writing on dirty paper model , we have the following tradeoff between the pure information transmission and the state uncertainty reduction .
[ prop : gaussian ] the tradeoff region @xmath196 for the gaussian channel depicted in figure [ fig : wdp ] is characterized by the boundary points @xmath276 where @xmath277 the achievability follows from proposition [ prop : noncausal ] with trivial extension to the input power constraint .
in particular , we use the simple power sharing scheme proposed in @xcite , where a fraction @xmath278 of the input power is used to transmit the pure information using costa s writing on dirty paper coding technique , while the remaining @xmath279 fraction of the power is used to amplify the state . in other words ,
@xmath280 with @xmath281 independent of @xmath8 , and @xmath282 with @xmath283 evaluating @xmath284 and @xmath285 for each @xmath278 , we recover and .
the proof of converse is essentially the same as that of ( * ? ? ?
* theorem 2 ) , which we do not repeat here . as an extreme point of the @xmath37
, we recover costa s writing on dirty paper result @xmath286 by taking @xmath287 . on the other hand ,
if state uncertainty reduction is the goal , then all of the power should be used for state amplification . the maximum uncertainty reduction rate @xmath288 is achieved with @xmath289 and @xmath290 .
in ( * ? ? ? * theorem 2 ) , the optimal tradeoff was characterized between the pure information rate @xmath0 and the receiver s state estimation error @xmath291 .
although the notion of state estimation error @xmath292 in @xcite and our notion of the uncertainty reduction rate @xmath5 appear to be distinct objectives at first sight , the optimal solutions to both problems are identical , as shown in the proof of proposition [ prop : gaussian ] .
there is no surprise here .
because of the quadratic gaussian nature of both problems , minimizing the mean squared error @xmath293 can be recast into maximizing the mutual information @xmath294 , and vice versa .
also the optimal state uncertainty reduction rate @xmath48 ( or equivalently , the minimum state estimation error @xmath295 is achieved by the symbol - by - symbol amplification @xmath296 .
finally , it interesting to compare the optimal coding scheme to the optimal coding scheme when the goal is to minimize ( instead of maximizing ) the uncertainty reduction @xcite , which is essentially based on coherent subtraction of @xmath9 and @xmath8 with possible randomization .
the previous two sections considered the case in which the transmitter has complete knowledge of the state sequence @xmath3 prior to the actual communication . in this section ,
we consider another model in which the transmitter learns the state sequence on the fly , i.e. , the encoding function @xmath297 \times \mathcal{s}^i \to \mathcal{x } , \qquad i = 1,2,\ldots , n,\ ] ] depends causally on the state sequence .
we state our main theorem .
[ thm : causal ] the tradeoff region @xmath42 for a state - dependent channel @xmath66 with state information @xmath3 causally known at the transmitter is the union of all @xmath41 pairs satisfying @xmath298 for some joint distribution of the form @xmath61 , where the auxiliary random variable @xmath68 has cardinality bounded by @xmath299 . as in the noncausal case , the region is convex . since the auxiliary random variable @xmath68 affects the first inequality only , the cardinality bound @xmath300 ; see shannon @xcite . ) finally , we can take @xmath9 as a deterministic function of @xmath71 without decreasing the region .
compared to the noncausal tradeoff region @xmath301 in theorem [ thm : noncausal ] , the causal tradeoff region @xmath302 in theorem [ thm : causal ] is smaller in general .
more precisely , @xmath302 is characterized by the same set of inequalities as in @xmath301 , but the set of joint distributions is restricted to those with auxiliary variable @xmath68 independent of @xmath8 . indeed , from the independence between @xmath68 and @xmath8 , we can rewrite as @xmath303 which is exactly the same as . thus the inability to use the future state sequence decreases the tradeoff region .
however , only the inequality , or equivalently , the inequality , is affected by the causality , and the sum rate does not change from .
since the proof of theorem [ thm : causal ] is essentially identical to that of theorem [ thm : noncausal ] , we skip most of the steps .
the least straightforward part is the following lemma .
[ lemma : equiv2 ] let @xmath96 be the union of all @xmath41 pairs satisfying . let @xmath97 be the closure of the union of all @xmath37 pairs satisfying , , and @xmath304 for some joint distribution @xmath305 where the auxiliary random variable @xmath68 has finite cardinality . then @xmath100 the proof is a verbatim copy of the proof of lemma [ lemma : equiv ] , except that here @xmath68 is independent of @xmath8 , i.e. , @xmath306 .
the final step follows since the set of conditional distributions on @xmath307 given @xmath8 of the form @xmath308 with deterministic @xmath121 is as rich as any @xmath309 , and @xmath310 with this replacement , the desired proof follows along the same lines as the proof of lemma [ lemma : equiv ] .
as one extreme point of the tradeoff region @xmath42 , we recover the shannon capacity formula @xcite for channels with causal side information at the transmitter as follows : @xmath311 on the other hand , the maximum uncertainty reduction rate @xmath48 for pure state amplification is identical to that for the noncausal case given in corollary [ corollary : noncausal ] . [
corollary : causal ] under the condition of theorem [ thm : causal ] , the maximum uncertainty reduction rate @xmath48 is given by @xmath312 thus the receiver can learn about the state essentially at the maximum cut - set rate , even under the causality constraint .
for example , the symbol - by - symbol amplification strategy @xmath289 is optimal for the gaussian channel ( example [ ex : wdp ] ) for both causal and noncausal cases .
finally , we compare the tradeoff regions @xmath302 and @xmath301 with a communication problem that has a totally different motivation , yet has a similar capacity expression . in (
* situations 3 and 4 ) , willems and van der meulen studied the multiple access channel with cribbing encoders . in this communication problem , the multiple access channel @xmath313 has two inputs and one output .
the primary transmitter @xmath8 and the secondary transmitter @xmath9 wish to send independent messages @xmath314 $ ] and @xmath315 $ ] respectively to the common receiver @xmath10 .
the difference from the classical multiple access channel is that either the secondary transmitter @xmath9 learns the primary transmitter s signal @xmath8 on the fly ( @xmath316 ( * ? ? ?
* situation 3 ) ) or @xmath9 knows the entire signal @xmath3 ahead of time ( @xmath317 ( * ? ? ?
* situation 4 ) ) .
the capacity region @xmath318 for both cases is given by all @xmath37 pairs satisfying @xmath319 for some joint distribution @xmath320 .
this capacity region @xmath318 looks almost identical to the tradeoff regions @xmath301 and @xmath302 in theorems [ thm : noncausal ] and [ thm : causal ] , except for the first inequality .
moreover , has the same form as the capacity expression for channels with state information available at _ both _ the encoder and decoder , either causally or noncausally .
( the causality has no cost when both the transmitter and the receiver share the same side information ; see , for example , caire and shamai ( * ? ?
* proposition 1 ) . )
it should be stressed , however , that the problem of cribbing multiple access channels and our state uncertainty reduction problem have a fundamentally different nature .
the former deals with encoding and decoding of the signal @xmath3 , while the latter deals with uncertainty reduction in an uncoded sequence @xmath3 specified by nature . in a sense ,
the cribbing multiple access channel is a detection problem , while the state uncertainty reduction is an estimation problem .
because the channel is state dependent , the receiver is able to learn something about the channel state from directly observing the channel output .
thus , to help the receiver narrow down the uncertainty about the channel state at the highest rate possible , the sender must jointly optimize between facilitating state estimation and transmitting refinement information , rather than merely using the channel capacity to send the state description .
in particular , the transmitter should summarize the state information in such a way that the summary information results in the maximum uncertainty reduction when coupled with the receiver s initial estimate of the state . more generally , by taking away some resources used to help the receiver reduce the state uncertainty
, the transmitter can send additional pure information to the receiver and trace the entire @xmath37 tradeoff region .
there are three surprises here .
first , the receiver can learn about the channel state and the independent message at a maximum cut - set rate @xmath57 over all joint distributions @xmath321 consistent with the given state distribution @xmath13 .
second , to help the receiver reduce the uncertainty in the initial estimate of the state ( namely , to increase the mutual information from @xmath322 to @xmath323 ) , the transmitter can allocate the achievable information rate @xmath324 in two alternative methods
random binning and its dual , random covering .
thirdly , as far as the sum rate @xmath325 and the maximum uncertainty reduction rate @xmath48 are concerned , there is no cost associated with restricting the encoder to learn the state sequence on the fly .
b. chen and g. w. wornell , `` quantization index modulation : a class of provably good methods for digital watermarking and information embedding , '' _ ieee trans .
inf . theory _ ,
it-47 , no . 4 , pp . 14231443 , 2001 .
a. sutivong , t. m. cover , m. chiang , and y .- h .
kim , `` rate vs. distortion trade - off for channels with state information , '' in _ proc .
inform . theory _ , lausanne , switzerland , june / july 2002 , p. 226 .
h. weingarten , y. steinberg , and s. shamai , `` the capacity region of the gaussian multiple - input multiple - output broadcast channel , '' _ ieee trans .
inf . theory _ , vol .
it-52 , no . 9 , pp .
39363964 , sept . | we consider the problem of transmitting data at rate @xmath0 over a state dependent channel @xmath1 with state information available at the sender and at the same time conveying the information about the channel state itself to the receiver .
the amount of state information that can be learned at the receiver is captured by the mutual information @xmath2 between the state sequence @xmath3 and the channel output @xmath4 .
the optimal tradeoff is characterized between the information transmission rate @xmath0 and the state uncertainty reduction rate @xmath5 , when the state information is either causally or noncausally available at the sender . in particular ,
when state transmission is the only goal , the maximum uncertainty reduction rate is given by @xmath6 .
this result is closely related and in a sense dual to a recent study by merhav and shamai , which solves the problem of _ masking _ the state information from the receiver rather than conveying it . |
the navier - stokes equations ( ns ) , while comprising the complete dynamics of turbulence , have at least two disadvantages for theoretical investigation of the physics of turbulence : ns lacks analytical solution for the case of the fully turbulent state and the nonlinear advection term results in turbulent states of high complexity which tends to obscure the fundamental mechanisms underlying the turbulence .
one approach to overcoming these impediments has been the search for simplifications of ns that retain essential features of the turbulence dynamics .
the linearized navier - stokes equations ( lns ) provide one example of the successful application of this approach in which the power of linear systems theory is made available to the study of turbulence @xcite .
the lns system captures the non - normal mechanism responsible for perturbation growth in ns @xcite .
this linear mechanism retained in lns underlies both the process of subcritical transition to turbulence and the maintenance of the turbulent state @xcite
. however , linear models are unable to capture other essential phenomena in turbulence that are intrinsically nonlinear including establishment of the turbulent mean velocity profile and maintenance of turbulence in a statistically steady state in the absence of extrinsic excitation , the mechanism of which is referred to as the self - sustaining process ( ssp ) . while there remain many further aspects of the turbulent state to be addressed , such as the structures and spectrum of the inertial subrange , a basis of understanding of wall turbulence comprises these fundamental aspects .
one approach to studying these phenomena that has proven successful is to impose simplifications that isolate certain mechanisms of the turbulence dynamics .
an example is restricting the size of the channel @xcite .
the existence of the minimal channel for sustaining turbulence isolated the streak structure component of the ssp and showed that the ssp cycle comprises interaction among the streak , a pair of streamwise vortices of opposite sign ( referred to as a rolls ) and the perturbation field .
in minimal channel turbulence a single streak fits in the channel and the streamwise extent of the channel is of such length that the perturbations can adopt a structure allowing them to extract energy from the mean flow through a time dependent interaction @xcite .
another example is imposing simplification of the dynamics by isolating specific regions of the turbulent flow .
for example , it was found that when the inner and outer regions of the flow are isolated the turbulence was independently maintained in each region arguing that an independent ssp is operating in these regions @xcite .
the minimal channel flow studies referred to above focused attention on the effort to obtain a self - consistent dynamical description of the interaction among the streak , the roll and the perturbation field .
the streaks in minimal simulations are associated with the zero streamwise harmonic ( @xmath0 ) .
this observation motivates constructing a reduced complexity turbulence model based on the simplest closure of the lns equations , which govern interaction between the streamwise mean flow ( with the streak structure included ) and the streamwise varying perturbation field ( characterized by the @xmath1 flow field ) .
this system is obtained by augmenting lns by only the nonlinearity resulting from including the feedback of the @xmath1 perturbations on the @xmath0 streamwise mean flow .
this dynamics , which greatly restricts the nonlinearity of the ns , will be referred to as the restricted nonlinear system ( rnl ) .
remarkably , as will be demonstrated in this work , the rnl self - sustains a realistic turbulent state , not only in minimal channels at low reynolds numbers , but also in large channels and at moderate reynolds numbers .
the rnl system was introduced here as a simple extension of lns that supports self - sustained turbulence .
however , the rnl has deeper roots : it approximates the second order cumulant closure of the ns , which is the basis of the stochastic structural stability theory ( s3 t ) .
s3 t defines a statistical mean state dynamical system and implementations of this system have recently been used to develop theories of turbulence @xcite .
s3 t employs an ensemble closure which produces autonomous statistical mean state dynamics in which turbulent mean states exist as statistical equilibria .
this makes the turbulent state available as an object for stability study , extending classical hydrodynamic stability theory which addresses only the stability of stationary sample state solutions of the ns .
s3 t , in contrast , can be used to determine the structural stability of statistical mean state entities such as attractors ( characterised by a specific probability density function ) , so that for instance , when such an attractor becomes s3 t unstable the fluid state bifurcates to a new attractor characterized by a different probability density function .
an example application of s3 t is to a constant shear flow subjected to homogeneous turbulence excitation in which it is found that a bifurcation occurs when the reynolds number exceeds a critical value resulting in a new stationary state comprising the mean flow with a roll - streak structure and a perturbation field supporting it in a new statistical steady state @xcite . at even higher reynolds number
a saddle - node bifurcation occurs in the s3 t system and the flow transitions to a time dependent state that self - sustains , which is identified with transition to the turbulent state . in the self - sustaining turbulent state
the statistical mean state dynamics of s3 t approaches the deterministic dynamics of the rnl .
recent work has exploited s3 t dynamics to explore the role of streamwise coherent structures in turbulence , including the dynamics of the roll and streak structures @xcite . in this paper , we verify predictions of s3 t for turbulence structure at high reynolds numbers in pressure driven channel flow , by comparing flow statistics , structures , and dynamical diagnostics obtained from the rnl system to results obtained in the s3 t system and to direct numerical simulations ( dns ) of the full ns equations .
consider a pressure driven plane poiseuille flow maintained by application of the time - dependent pressure , @xmath2 , where @xmath3 is the streamwise coordinate .
the wall - normal direction is @xmath4 and the spanwise direction is @xmath5 .
the lengths of the channel in the streamwise , wall - normal and spanwise direction are respectively @xmath6 , @xmath7 and @xmath8 .
the channel walls are at @xmath9 and @xmath10 .
streamwise mean , spanwise mean and time mean quantities are denoted respectively by an overbar , @xmath11 , square brackets , @xmath12=l_z^{-1 } \int_0^{l_z } \bullet ~{\textrm{d}}z$ ] and a wide hat @xmath13 , with @xmath14 sufficiently long .
the velocity , @xmath15 , is decomposed into its streamwise mean value , denoted @xmath16 , and the deviation from the mean ( the perturbation ) , @xmath17 , so that the flow velocity is @xmath18 .
the pressure gradient is similarly written as @xmath19 .
the ns can be then decomposed into an equation for the mean and an equation for the perturbation as follows : @xmath20 [ eq : nse0 ] where @xmath21 is the coefficient of kinematic viscosity .
the @xmath22 components of @xmath23 are @xmath24 and the corresponding components of @xmath25 are @xmath26 .
the streak component of the streamwise mean flow is denoted with @xmath27 and defined as @xmath28 $ ] .
the @xmath29 and @xmath30 are the streamwise mean velocities of the roll vortices .
streamwise mean perturbation reynolds stress components are therefore written as e.g. @xmath31 , @xmath32 . the rnl approximation is obtained by neglecting the perturbation - perturbation interaction terms in eq . .
the rnl system is : @xmath33 [ eq : qle0 ] equation describes the dynamics of the streamwise mean flow , @xmath23 , which is driven by the divergence of the streamwise mean reynolds stresses .
these reynolds stresses are obtained from eq . which incorporates the influence of the time dependent streamwise mean flow @xmath16 on the streamwise varying perturbations @xmath25 but not the nonlinear interaction among the perturbations .
only the interaction of the perturbations directly on the streamwise mean flow , @xmath23 , is retained in the rhs of eq . .
remarkably , rnl self - sustains turbulence by incorporating this one essential nonlinear interaction , in the absence of which a self - sustained turbulent state can not be established @xcite .
@*7c abbreviation & @xmath34/h$]&@xmath35 & @xmath36 & [ @xmath37,@xmath38 $ ] + ns350 & @xmath39$]&@xmath40 & 357.1 & @xmath41 $ ] + rnl350 & @xmath39$]&@xmath42 & 353.5 & @xmath43 $ ] + ns950 & @xmath44$]&@xmath45&939.9 & @xmath46 $ ] + rnl950 & @xmath44$]&@xmath47 & 882.4 & @xmath48 $ ] +
the data were obtained from a dns of eqs . and from the rnl that is directly associated with the dns .
both the dns and its directly associated rnl are integrated with no - slip boundary conditions in the wall - normal direction and periodic boundary conditions in the streamwise and spanwise directions .
the dynamics were expressed in the form of evolution equations for the wall - normal vorticity and the laplacian of the wall - normal velocity , with spatial discretization and fourier dealiasing in the two wall - parallel directions and chebychev polynomials in the wall - normal direction @xcite .
time stepping was implemented using the third - order semi - implicit runge - kutta method .
quantities reported in outer units lengths are scaled by the channel half - width , @xmath49 , and time by @xmath50 and the corresponding reynolds number is @xmath51 where @xmath52 ( @xmath53 is the shear at the wall ) is the friction velocity .
inner units lengths are scaled by @xmath54 and time by @xmath55 .
velocities scaled by the friction velocity @xmath56 will be denoted with the superscript @xmath57 , which indicates inner unit scaling .
parameters for the simulations presented are listed in table [ table : geometry ] .
in this section we compare turbulence diagnostics obtained from self - sustained turbulence in the rnl system , eqs . , to diagnostics obtained from the parallel dns of the ns , eqs .
the geometry and resolution of the ns and rnl cases are given in table [ table : geometry ] .
results are reported for poiseuille turbulence at either @xmath58 or @xmath59 .
the rnl simulations were initialized with an ns state and run until a steady state was established .
the rnl simulations produce self - sustained turbulence with the time mean estimated @xmath60 values reported in table [ table : geometry ] , which are close to the @xmath60 values of the ns turbulent state .
henceforth , the rnl simulations will be identified with the @xmath60 value of the corresponding ns .
the turbulent mean profiles for the ns and rnl simulations with @xmath58 and @xmath59 are shown in outer variables in fig .
[ fig : uprofile350950 ] , while fig .
[ fig : uprofile350950 ] , shows the same data in inner - wall units .
previous simulations in couette turbulence @xcite at lower reynolds numbers ( @xmath61 ) showed very small difference between the mean turbulent profile in ns and rnl .
these simulations at larger reynolds numbers show significant differences in the mean turbulent profiles sustained by ns and rnl .
this is especially pronounced in the outer regions where rnl produces a mean turbulent profile with substantially smaller shear .
both profiles produce a log layer . however , the shear in these logarithmic regions are different : the von krmn constant of ns at @xmath59 is @xmath62 while for rnl it is @xmath63
. formation of a log layer indicates that the underlying dynamics producing the log layer are retained in rnl . because rnl maintains essentially the same stress and variance as ns in the log layer with a smaller shear , rnl is in this sense more efficient than ns in transferring energy from the mean to the perturbations .
a comparison of perturbation statistics in rnl and ns is shown in fig .
[ fig : vel_om_prime ] for @xmath59 .
the streamwise perturbation velocity fluctuations are significantly more pronounced in rnl and the magnitude of the streak in rnl exceeds significantly the streak magnitude in ns in the inner wall region ( cf . fig .
[ fig : vel_om_prime ] ) .
in contrast , the wall - normal and spanwise fluctuations in rnl are less pronounced than in ns ( cf . fig .
[ fig : vel_om_prime ] ) and similarly the streak fluctuations in the outer region are also less pronounced in rnl ( cf .
[ fig : vel_om_prime ] ) .
the structure of the fluctuations of the vorticity components as a function of @xmath4 is shown in fig .
[ fig : vel_om_prime ] . the @xmath64 and @xmath65 fluctuations are similar in ns and rnl .
the large @xmath64 fluctuations are associated with the shear of the streamwise velocity , while the @xmath65 fluctuations are associated with the streamwise streak structure .
only the @xmath66 fluctuations differ appreciably in amplitude between rnl and ns .
this vorticity component is primarily associated with the fluctuations in the streamwise roll circulations that are responsible for maintaining the streak which is central to sustaining the turbulence . despite this difference in the r.m.s .
values of the velocity fluctuations , both rnl and ns produce very similar @xmath67 reynolds stress .
the increased amplitude of the velocity fluctuations in rnl is consistent with the fact that rnl and ns produce nearly the same energy dissipation rate .
the reynolds stress @xmath68}$ ] is the sum of @xmath69}$ ] and @xmath70}$ ] .
comparison of the wall - normal distribution of the time mean of these two components of the reynolds stress is shown in fig . [
fig : uv_f ] . because the turbulence in ns and rnl is sustained with essentially the same pressure gradient , the sum of these reynolds stresses is the same linear function of @xmath4 outside the viscous layer .
the reynolds stress is dominated by the perturbation reynolds stress @xmath70}$ ] in both simulations , with the rnl stress penetrating farther from the wall .
this is consistent with the fact that the perturbation structure in rnl has larger scale .
this can be seen in a comparison of the ns and rnl perturbation structure shown in fig .
[ fig : yzpert ] .
note that the reynolds stress @xmath71}$ ] associated with the streak and roll in the outer region of the ns simulation is larger than that in rnl .
further , the average correlation between the perturbation @xmath72 and @xmath73 fields are almost the same in both simulations while the correlation between the @xmath27 and @xmath29 in the rnl is much smaller than that in ns in the outer layer .
this is seen in a plot of the structure function ( cf .
@xcite ) shown in fig .
[ fig : uv_f ] . turning now to the ns950 and rnl950 simulations , a @xmath74 plane snapshot of the streamwise mean flow component ( corresponding to @xmath0 streamwise wavenumber ) is shown in fig .
[ fig : yzstreak ] .
contours of the streamwise streak flow field , @xmath27 , are shown together with vectors of the streamwise mean @xmath75 field , which indicates the velocity components of the large - scale roll structure .
the presence of organized streaks and associated rolls is evident both in the inner - wall and in the outer - wall region .
figure [ fig:3dstreak ] shows a three - dimensional perspective of the flow of the ns and rnl simulations , in which all of the @xmath76 components of the velocity field are included .
note that in the rnl there is no visual evidence of the @xmath0 roll / streak structure which is required by the restrictions of rnl dynamics to be the primary structure responsible for maintaining the rnl self - sustained turbulent state .
rather , the most energetic streamwise harmonic ( @xmath77 for this channel ) is most prominent and dominates the perturbation structure . a comparison of the spectral energy densities of the velocity fields as a function of streamwise and spanwise wavenumber , @xmath78 , provides an alternative view of the turbulent structure .
the premultiplied spectral energy densities of each of the three contributions to the kinetic energy , @xmath79 , @xmath80 and @xmath81 at channel heights @xmath82 , representative of the inner - wall region , and @xmath83 , representative of the outer - wall region , are shown in fig .
[ fig : spectr_2d ] . near the inner - wall
, rnl produces spanwise streak spacing and rolls similar to those in ns .
the tendency of rnl to favor longer structure is also evident in these figures .
the spectra for the outer region indicate similar large scale structure and good agreement in the spanwise spacing between rnl and ns .
this figure establishes the presence of large - scale structure in the outer region in both rnl and ns .
the corresponding contour plots of normalized premultiplied one - dimensional spectral energy densities as a function of spanwise wavelength and wall - normal distance , shown in fig .
[ fig : lambda_z ] accord with the spectra of toh & itano @xcite and demonstrate again the similarity in structure in ns and rnl and specifically the presence of large structures in the outer flow .
it has been noted that in ns while the dominant large scale structures scale linearly with distance from the wall in the inner - wall region , in the outer regions structures having the largest possible streamwise scale dominate the flow variance at high reynolds number @xcite .
this linear scaling near the wall can be seen in fig .
[ fig : lambda_z ] where the pre - multiplied spectral densities as a function of the distance @xmath4 and of the spanwise wavelength , @xmath84 , are shown , as in @xcite , for both nl and rnl . in both simulations
the spanwise wavelength associated with the spectral density maxima increases linearly with wall distance and this linear dependence is intercepted at @xmath85 ( or @xmath86 ) . beyond @xmath85 structures
assume the widest wavelength allowed in the channel , suggesting that simulations must be performed in longer boxes in future work ( cf .
discussion in jimnez & hoyas @xcite and flores & jimnez @xcite ) .
corresponding contour plots of spectral energy density as a function of streamwise wavelength and wall - normal distance are shown in fig .
[ fig : lambda_x ] .
these plots show that the perturbation variance in the inner wall and outer wall region is concentrated in a limited set of streamwise components which is apparent in fig .
[ fig:3dstreak ] .
the restriction of the streamwise structure is particularly pronounced in the case of rnl in which the outer layer variance peaks at @xmath87 , which scale the wall - normal velocity inherits .
note that the maximum wavelength in these graphs is equal to the streamwise length of the box and not to the infinite wavelength associated with the energy of the streak / roll structure , which as we will see is the most energetic structure in the dns but not in the rnl .
rnl turbulence has the property that when initiated with full ns turbulence it spontaneously transitions to a self - sustaining turbulent state supported by a severely restricted set of streamwise fourier components .
this property is consistent with the structure of the rnl system , which retains only the interaction between the @xmath0 and the @xmath88 components , from which it follows that the only energy source for maintaining a perturbation is its interaction with the mean flow @xcite .
it is remarkable that at these fairly high reynolds numbers only a small set of streamwise harmonics are maintained by this interaction with the mean flow .
moreover , this is the fundamental interaction maintaining rnl turbulence and by implication , given the similarity in the structure and dynamics between them , of ns turbulence as well .
even if the rnl dynamics is initialized with an ns flow state with energy in all @xmath89 components , the rnl turbulence eventually reduces to involve only the @xmath90 fourier components with wavenumber @xmath91 .
we view this transition of ns turbulence to rnl turbulence as a process of distillation by which a small set of structures maintaining the turbulent state is identified , a result that was previously obtained in the case of self - sustained couette turbulence at @xmath92 and @xmath93 @xcite . in this previous work ,
a minimal channel rnl simulation at @xmath92 was shown to self - sustain turbulence by the interaction of only two @xmath76 components : @xmath0 and the first harmonic in the channel . the analogous distillation process for the @xmath59 simulation is shown in fig .
[ fig : enkx_dnsql950 ] .
the time evolution of the energy of the first @xmath94 streamwise varying fourier components in an ns simulation is shown in the left part of the figure ( @xmath95 ) while in the right part is shown the subsequent evolution of these components when at the indicated time the perturbation - perturbation interactions are suppressed so that the turbulence evolves under rnl dynamics .
it is evident from fig .
[ fig : enkx_dnsql950 ] that rnl950 turbulence in a channel with @xmath96 retains only the six fourier components @xmath97 out of the @xmath98 streamwise varying components that are present in the ns simulation .
all components with @xmath99 decay exponentially . as a result
, a transition occurs in which a reduced complexity dynamics maintaining turbulence arises , which self - sustains turbulence despite this greatly restricted support in streamwise wavenumber space .
that rnl maintains a turbulent state similar to that of ns with nearly the same @xmath60 ( @xmath100 vs. @xmath101 ) implies that these systems have approximately the same energy production and dissipation and that the @xmath102 components retained in rnl assume the burden of accounting for this energy production and dissipation .
specifically , the components in ns that are not retained in rnl are responsible for approximately @xmath103 of the total energy dissipation , which implies that the components that are retained in rnl must increase their dissipation by that much .
large scale roll / streak structures are prominent in the inner layer as well as in the outer layer both in ns and in rnl .
the dynamics of this structure can be diagnosed using the time evolution of the energy of each @xmath76 component during the transition from ns to rnl , shown in fig .
[ fig : enkx_dnsql950 ] .
it can be seen that in ns the energy associated with the streamwise mean structure with @xmath0 and @xmath104 is dominant among the structures that deviate from the mean flow , @xmath105 $ ] . in the inner layer
the interaction of roll / streak structures with the @xmath1 perturbation field maintains turbulence through an ssp @xcite .
the rnl system provides an especially simple manifestation of this ssp as its dynamics comprise only interaction between the mean ( @xmath0 ) and perturbation ( @xmath88 ) components . the fact that rnl self - sustains a counterpart turbulent state provides strong evidence that the rnl ssp captures the essential dynamics of turbulence in the inner wall region .
the structure of the rnl system compels the interpretation that the time dependence of the ssp cycle , which might appear at first to consist of a concatenation of random and essentially unrelated events , is instead an intricate interaction dynamics among streaks , rolls and perturbations that produces the @xmath16 which , when introduced in eq .
, results in generation of an evolving perturbation lyapunov structure with exactly zero lyapunov exponent .
s3 t identifies this exquisitely contrived ssp cycle which comprises the generation of the streak through lift - up by the rolls , the maintenance of the rolls by torques induced by the perturbations which themselves are maintained by an essentially time - dependent parametric non - normal interaction with the streak ( rather than e.g. inflectional instability of the streak structure ) @xcite .
vanishing of the lyapunov exponent associated with the ssp is indicative of a feedback control process acting between the streaks and the perturbations by which the parametric instability of the perturbations on the time dependent streak is reduced so that it asymptotically maintains a zero lyapunov exponent .
examination of the transition from the ns to the rnl , shown by the simulation diagnostics in fig .
[ fig : enkx_dnsql950 ] , reveals the action of this controller .
when in eq .
the interaction among the perturbations is switched off , so that the simulation is governed by rnl dynamics , we observe a sudden increase of the energy of the surviving @xmath88 components , cf .
the rapid increase of the energy in the @xmath106 components .
increase of the energy of these components is expected because the dissipative effect of the perturbation - perturbation nonlinearity that acts on these components is removed in rnl .
as these modes grow the ssp cycle quickly adjusts to a new turbulent equilibrium state characterized by reduced streak amplitude and increased energy in the largest streamwise scales .
this ssp cycle is more efficient in the sense that with smaller mean shear a self - sustained turbulence with approximately the same @xmath60 as that in ns is maintained . this turbulent state is dominated by high amplitude fluctuations of the @xmath107 components , as well as in the components associated with the wall - normal and spanwise direction .
this can be seen in a comparison of the ns and rnl perturbation structure ( the velocity field corresponding to @xmath88 ) shown in fig .
[ fig : yzpert ] .
the perturbations in rnl simultaneously reduce the shear of the mean flow and maintain a reduced amplitude streak in the outer layer .
a comparison of the shear , of the r.m.s .
@xmath29 velocity , and of the r.m.s .
streak velocity , @xmath27 , in the outer layer is shown as a function of @xmath4 in fig .
[ fig : uyvrmsusrms ] , from which it can be seen that the reduction of the amplitude of the streak in rnl is equal to the reduction in the mean flow shear .
it is important to note that these dependencies are integral to the ssp cycle and specifically of its feedback control which determines the statistical steady state and must be understood in the context of the cycle . in the discussion above we have assumed that the presence of roll and streak structure in the outer layer in rnl indicates the existence of an ssp cycle there as well , and by implication also in ns .
in order to show this consider the momentum equation for the streamwise streak : @xmath108}\bit{\right)}+{\left(}w \,\partial_z u-{\left[}w \,\partial_z u{\right]}\bit{\right)}}_{\textrm{a } } -\nonumber\\ & \qquad\qquad \underbrace{-{\left(}\overline { v ' \,\partial_y u'-{\left[}v ' \,\partial_y u'{\right]}}{\right)}- { \left(}\overline { w ' \,\partial_z u'-{\left[}w ' \,\partial_z u'{\right]}}{\right)}}_{\textrm{b}}+ \,\nu\ , { \delta}u _ { s}~.\label{eq : us}\end{aligned}\ ] ] term a in eq . is the contribution to the streak acceleration by the ` lift - up ' mechanism and the ` push - over ' mechanism , which represent transfer to streak momentum by the mean wall - normal and spanwise velocities respectively .
term b in eq .
is the contribution to the streak momentum by the perturbation reynolds stress divergence ( structures with @xmath88 ) . in order to identify the mechanism of streak maintenance we examine whether terms a and b accelerate or decelerate the streaks by evaluating from the simulation the contribution of the terms @xmath109 and @xmath110 , where @xmath111 and @xmath112 , to streak maintenance as a function of time and region @xmath113 of the flow ( similar results are obtained by multiplying with @xmath27 to obtain an energy rather than a momentum budget for @xmath27 ) .
the results of this calculation are shown in fig .
[ fig : usdot_time ] for @xmath113 extending over the the inner region , the outer region and the whole channel . in the inner and outer wall regions for both ns and rnl the streak structure
is supported primarily by the lift - up mechanism while the reynolds stress divergences oppose the streak . while the magnitude of the acceleration by the lift - up and the deceleration by the reynolds stress divergence are nearly the same in both ns and rnl in the inner region , in the outer region the acceleration by the lift - up in the rnl is about half of that in the ns due to the smaller mean flow shear in rnl .
the wall - normal structure of the time mean @xmath114 and @xmath115 are shown in fig .
[ fig : usdotyomxdoty ] .
we conclude that in ns and rnl the only positive contributions to the outer layer streaks are induced by the lift - up from the roll circulation , despite the small shear in this region .
we next consider the dynamics maintaining the roll circulation .
we have established that the roll circulation is not only responsible for streak maintenance in the inner layer but also in the outer layer .
we now examine the mechanism of roll maintenance using streamwise averaged vorticity , @xmath116 , in the outer layer as a diagnostic . in order for roll circulation to be maintained against dissipation there must be a continuous generation of @xmath117 .
there are two possibilities for the maintenance of @xmath117 in the outer layer : either @xmath117 is generated locally in the outer layer , or it is advected from the near wall region . from eq .
we obtain that @xmath117 satisfies the equation : @xmath118 term c expresses the streamwise vorticity tendency due to advection of @xmath117 by the streamwise mean flow @xmath75 . because there is no vortex stretching contribution to @xmath117 from the @xmath75 velocity field , this term only advects the @xmath117 field and can not sustain it against dissipation .
however , this term may be responsible for systematic advection of @xmath117 from the inner to the outer layer .
term d is the torque induced by the perturbation field .
this is the only term that can maintain @xmath117 .
the overall budget for square streamwise vorticity in the region @xmath113 , @xmath119 , @xmath120 , is given by : @xmath121}&= \underbrace { - { \left[}\frac1{2}\omega_x^2\,v{\right]}\left.\vphantom{\frac1{2}}\right|^{y = y_2}_{y = y_1 } } _ { h^{-2 } { u_\tau}^3\,i_{c } } + \underbrace{\int_{y_1}^{y_2 } { \textrm{d}}y\,{\left[}\bit\omega_x\times\textrm{term d}{\right]}}_{h^{-2 } { u_\tau}^3\,i_{d}}+ { \,\nu\int_{y_1}^{y_2 } { \textrm{d}}y\,{\left[}\vphantom{{\dot{\raisebox{0pt}[\dimexpr\height+.2ex][\depth]{$\omega$}}}_x^{(\textrm{a})}}\bit\omega_x\,{\delta}\omega_x{\right]}}~. \label{eq : omx^2}\end{aligned}\ ] ] where we have defined @xmath122 and @xmath123 , with @xmath124 and @xmath125 .
term @xmath126 represents the flux of vorticity into the region by the streamwise averaged wall - normal velocity , @xmath29 .
time series of the contributions from @xmath127 and @xmath128 to the @xmath117 production for ns950 and rnl950 , shown in fig .
[ fig : omx^2 ] , demonstrate that @xmath117 is primarily generated in situ by reynolds stress torques .
the corresponding wall - normal structure of the time mean @xmath129 , representing the local contribution to streamwise mean vorticity generation from perturbation reynolds stress induced torques is shown in fig .
[ fig : usdotyomxdoty ] .
note that for ns in the outer layer the streamwise mean vorticity generation by the reynolds stress is strongly positive at each instant .
this is so despite the fact that the r.m.s .
@xmath29 velocity is smaller in the ns than in the rnl , as seen in fig .
[ fig : uyvrmsusrms ] , implying greater streamwise vorticity dissipation in the ns .
it could be argued that the positive value of the generation @xmath130 found in ns and rnl is a consequence of the finite streamwise extent of the channel and as the channel length increases , assuming that there is no systematic correlation between streamwise average torques and streak structure , @xmath130 should decrease as @xmath131 and in the limit of an infinite channel the ns and rnl should sustain no @xmath117 .
however , s3 t theory shows that there is a systematic correlation between streaks and roll generation by perturbation torque resulting from the deformation of the turbulence by the streak and as a result it predicts that in the limit @xmath132 , @xmath130 should asymptote to a finite non - zero value at least in rnl .
having established that the streamwise vorticity in the outer layer is generated in situ from local reynolds stress divergences we conclude that the ssp cycle is operational in the outer layer just as in the inner layer .
we have established that rnl self - sustains turbulence at moderate reynolds numbers in pressure driven channel flow , despite its greatly simplified dynamics when compared to ns .
remarkably , in the rnl system , the turbulent state is maintained by a small set of structures with low streamwise wavenumber fourier components ( at @xmath59 with the chosen channel the ssp involves only the @xmath0 streamwise mean and the next @xmath133 streamwise fourier components )
. not only that , but this minimal turbulent dynamics arises spontaneously when the rnl system is initialized by ns turbulence at the same reynolds number . in this way rnl spontaneously produces a turbulent state of reduced complexity .
rnl identifies an exquisitely contrived ssp cycle which has been previously identified to comprise the generation of the streak through lift - up by the rolls , the maintenance of the rolls by torques induced by the perturbations which themselves are maintained by an essentially time - dependent parametric non - normal interaction with the streak ( rather than e.g. inflectional instability of the streak structure ) @xcite .
the vanishing of the lyapunov exponent associated with the ssp is indicative of a feedback control process acting between the streaks and the perturbations by which the parametric instability that sustains the perturbations on the time dependent streak is reduced to zero lyapunov exponent , so that the turbulence neither diverges nor decays .
we have established that both ns and rnl produce a roll / streak structure in the outer layer and that an ssp is operating there despite the low shear in this region .
it has been shown elsewhere that turbulence self - sustains in the log layer in the absence of boundaries @xcite .
this is consistent with our finding that an ssp cycle exists in both the inner - layer and outer - layer .
the turbulence maintained in rnl is closely related to its associated ns turbulence and both exhibit a log layer , although with substantially different von krmn constants .
existence of a log layer is a fundamental requirement of asymptotic matching between regions with different spatial scaling , as was noted by millikan @xcite .
however , the exact value of the von krmn constant does not have a similar fundamental basis in analysis and rnl turbulence , which is closely related to ns turbulence but more efficient in producing reynolds stress , maintains as a consequence a smaller shear and therefore greater von krmn constant .
specifically , we have determined that the ssp cycle in rnl is characterized by a more energetic and larger scale perturbation structure , despite having a lower amplitude streak and mean shear .
formation of roll / streak structures in the log layer is consistent with the universal mechanism by which turbulence is modified by the presence of a streak in such way as to induce growth of a roll structure configured to lead to continued growth of the original streak .
this growth process underlies the non - normal parametric mechanism of maintaining the perturbation variance in the ssp that maintains turbulence if the reynolds number is large enough @xcite .
this universal mechanism does not predict nor require that the roll / streak structures be of finite streamwise extent and in its simplest form it has been demonstrated that it supports roll / streak structures with zero streamwise wavenumber . from this point of view
the observed length of roll / streak structures is not a consequence of the primary mechanism of the ssp supporting them but rather a secondary effect of disruption by the turbulence . in this work
we have provided evidence that ns turbulence is persuasively related in its dynamics to rnl turbulence . moreover , given that the dynamics of rnl turbulence can be understood fundamentally from its direct relation with s3 t turbulence we conclude that the mechanism of turbulence in wall bounded shear flow is the roll / streak / perturbation ssp that was previously identified to maintain s3 t turbulence .
this work was funded in part by the multiflow program of the european research council .
navid constantinou acknowledges the support of the alexander s. onassis public benefit foundation .
brian farrell was supported by nsf ags-1246929 .
we thank dennice gayme for helpful reviewing comments . | s3 t ( stochastic structural stability theory ) employs a closure at second order to obtain the dynamics of the statistical mean turbulent state . when s3 t is implemented as a coupled set of equations for the streamwise mean and perturbation states , nonlinearity in the dynamics is restricted to interaction between the mean and perturbations .
the s3 t statistical mean state dynamics can be approximately implemented by similarly restricting the dynamics used in a direct numerical simulation ( dns ) of the full navier - stokes equations ( referred to as the ns system ) .
although this restricted nonlinear system ( referred to as the rnl system ) is greatly simplified in its dynamics in comparison to the associated ns , it nevertheless self - sustains a turbulent state in wall - bounded shear flow with structures and dynamics comparable to that in observed turbulence .
moreover , rnl turbulence can be analyzed effectively using theoretical methods developed to study the closely related s3 t system . in order to better understand rnl turbulence and its relation to ns turbulence
, an extensive comparison is made of diagnostics of structure and dynamics in these systems .
although quantitative differences are found , the results show that turbulence in the rnl system closely parallels that in ns and suggest that the s3t / rnl system provides a promising reduced complexity model for studying turbulence in wall - bounded shear flows . |
wave breaking is one of the most complex phenomena that occurs in the near shore region . during propagation of wave from deep to shallow water
, the wave field is transformed due to shoaling .
close to the shoreline , they become unstable and break . in the process of breaking
, energy is redistributed from fairly organized wave motion to small scale turbulence , large scale currents and waves .
+ classical boussinesq theory provides a set of evolution equations for sueface water waves in the combined limit of weak nonlinearity ( characterized by @xmath0 ) and weak dispersion ( @xmath1 ) with the raio @xmath2 .
the parameters represent a wave height to water deth ratio and a water depth to wavelength ratio , respectively .
+ it has been shown by numerous researchers that boussinesq - type equations for varying water depth can describe nonlinear transformation in the shoaling region quite well . in the last couple of decades , a lot of research effort has gone into improving the predictive capability of these equations in the intermediate water - depth and close to the surf zone ( see e.g. nwogu @xcite , madsen @xcite , wei @xcite ) .
it was established that to extend the validity of these equations to the deep water , higher order dispersive terms will have to be retained . to improve the predictive capability close to wave breaking , the higher order nonlinear terms are very important to include in the equation .
however , all these models use additional terms that is artificially added to the momentum equation , which would then reproduce the main characteristic of a breaking wave , i.e. the reduction in wave height to describe wave breaking phenomena .
wave breaking in the software funwave ( funwave is based on the model described by nwogu @xcite ) is modeled by introducing momentum mixing term developed by kennedy et al @xcite .
+ starting with the works of nwogu @xcite and madsen @xcite , most progress have been done to explain wave breaking phenomena .
shen @xcite addresses the problem with partially rotational flow .
vorticity dynamics and formation into the fluid is very important in the wave breaking as well surf zone . to address this problem ,
veeramony & svendsen @xcite derived breaking terms in boussinesq equation assuming flow as a two - dimensional rotational flow . here
, the breaking process is modeled by assuming that vorticity is generated in the roller region of the breaking wave and solving vorticity transport equation to determine the distribution of the vorticity .
this naturally introduces additional terms in the momentum equation which causes wave height reduction as well as changes in the velocity field .
however , since this model is based on stream function formulation , it can not be trivially extended to three - dimensional flow .
the phenomena of wave breaking in boussinesq equations are being modeled using quite few techniques which can preserve the wave shape as well as include energy dissipation mechanism .
shen @xcite developed a generalized form of bousinesq equation in three dimensional by introducing vertical flow field with arbitrary vorticity distribution up to o(@xmath3 .
but he did not describe momentum transport equation with full description of rotational flow .
recently , zou et al @xcite addressed the problem by including the higher order terms in boussinesq equation for two dimensinal flow .
this model solves vorticity distribution based on the parametric form taken form surface roller experimental data . in this paper
, we developed a general analytic form for breaking term for fully nonlinear set of boussinesq equations for three dimensional vertical flow field near surf zone region .
derivation of breaking term from reynold stress based vorticity transport equation was also developed to describe rotational field as a complete model for thorough understanding of wave breaking phenomena .
+ the term nonlinear indicates that no truncation based on powers of @xmath4 is employed .
the paper is organized as follows : section 2 discusses the basic governing equations for continuity and momentum with boundary conditions .
section 3 describes the equation for horizontal and vertical velocity distribution for potential and rotational components . in section 4 ,
the breaking term is derived for velocity transport equation for fully nonlinear case and solved analytic form vorticity transport equation from fourier series expansion .
reynold stresses are analyzed in the breaking region . in last section , results were discussed with conclusion . a parametric analysis of the role eddy viscosity profile is being discussed in detail .
we consider a three - dimensional wave field with free surface @xmath5 propagating over a variable water depth @xmath6 .
as we are primarily concerned with wave breaking , we only consider here wave propagation in shallow water .
wave in this region can be characterized by two non - dimensional parameters @xmath7 and @xmath8 where @xmath9 is the characteristic wave amplitude and @xmath10 the characteristic wave length .
the parameter @xmath11 is a measure of frequency dispersion and @xmath4 that of the non linearity of the wave . in this study , since we are only considering shallow water waves , we only have to consider weakly dispersive waves ( up to @xmath12 ) but have to retain all nonlinear terms . in this paper ,
the variables are non - dimensionalized using following scaling : @xmath13 where the @xmath14 represents the dimensional variables , @xmath15 is the acceleration due to gravity , @xmath16 and @xmath17 are the horizontal components of the velocity in the @xmath18 and @xmath19 directions respectively , @xmath20 is the vertical velocity .
we start with the eulerian equations of continuity and momentum in nondimensionalized form for velocity field @xmath21 as : @xmath22 since the fluid flow is rotational , we also have three dimensional vorticity field @xmath23 in the fluid defined as @xmath24 where @xmath25 the continuity equation then becomes , @xmath26 here @xmath27 the above equations satisfy two boundary conditions for velocity at bottom and at free surface . at the free surface @xmath28 , since particles are free to move with fluid velocity , the kinematic boundary condition is @xmath29and at bottom @xmath30 @xmath31 where @xmath32 is two component horizontal surface velocity .
@xmath34 refer to horizontal derivative with respect to x and y in all subsequent calculations .
the horizontal component for vorticity field @xmath35 can be described as , @xmath36 with @xmath37 as two component horizontal field whereas vertical component of vorticity expressed as @xmath38 this is straightforward calculation from equation ( 6 ) and ( 8) which is the beginning equation in three dimensional vorticity field formulation .
@xmath39 @xmath20 represents the vertical velocity of the flow . in the above equation ,
once @xmath20 solved , horizontal component of velocity @xmath40 can be solved from vorticity relation . in weakly
hydro static case ( @xmath41 ) , solution is typically obtained from iterative perturbation procedure with successive correction term up to @xmath42 . in case of breaking waves where vorticity is very strong , so @xmath43 we assume solution as , @xmath44 and @xmath45 for horizontal and vertical velocity component . under this assumption , poisson equation becomes @xmath46 @xmath47\ ] ] @xmath48 , @xmath49 can be calculated from bottom boundary conditions using equation ( 7 ) separately where the boundary conditions are , @xmath50 at bottom boundary _ z = -h _ .
+ since at any other depth @xmath51 , w is constrained by continuity equation only , so the equation follows @xmath52 where @xmath53 is velocity at any arbitrary depth @xmath54 . in boussinesq equation , one may take depth average or any intermediate velocity for horizontal velocity between bottom and free surface as reference velocity . in the wave breaking zone where the vorticity is developed non uniformly , the equations become simpler with the choice of depth average velocity which includes contribution from surface vorticity gradient .
we assume solution for velocity comes also from rotational contribution due to vorticity at the wave surface .
so the velocity has both potential as well as rotational component , @xmath55 , @xmath56 + we solve @xmath57 and @xmath58 at any depth @xmath59 @xmath60_{z } } _ { r } = - { \bf \bigtriangledown } \cdot ( u_{m } - z_{r } s { \rbrack _ { z } } _ { r } ) + z _ { r } { \bf \bigtriangledown } \cdot s\ ] ] @xmath61 @xmath62 with boundary condition @xmath63 and @xmath64 + equations ( 4 ) - ( 16 ) form basic shallow water boussinesq equations .
in the surf zone , vorticity grows very strongly as a non uniform function over depth . following shen [ 2000 ] , we define reference velocity as @xmath65 in terms of depth average velocity @xmath66 and magnitude of vorticity at free surface @xmath67 with the assumption of @xmath68 .
we set here @xmath69 as linear calibration for @xmath70 does not hold here in presence of nonuniform velocity as wave dispersion properties change both spatially and temporally with vorticity . and boundary condition can be set as @xmath71 integrating equation ( 9 ) from bottom to surface and applying boundary condition to ( 16 ) we get @xmath72 as , @xmath73 where @xmath74 is the vertical velocity distribution generated by horizontal divergence of vorticity added to surface velocity .
now , once @xmath75 is calculated , @xmath76 can be calculated from eqn ( 15 ) with surface boundary condition @xmath77 and @xmath78 finally , we calculate horizontal velocity as @xmath79 + o(\mu^{\rm 4 } ) \end{split}\ ] ] which on averaging over depth yields , @xmath80 - \cfrac{\mu^{\rm 2}}{2 } h_{\rm \eta } [ { \bigtriangledown } \tilde{u } + { \eta } s_{\rm \eta } ] \cdot{\bigtriangledown}h \\ - ( { \bigtriangledown } \cdot \tilde{u } - { \eta}{\bigtriangledown } \cdot s_{\rm \eta}){\bigtriangledown } h + o(\mu^{\rm 4 } ) \end{split}\ ] ] @xmath81 is the average surface velocity contribution due to vorticity and it is significant for suspended sediment particles in the flow .
the term @xmath82 is the change due to depth variation of vorticity @xmath83 .
the total water depth @xmath84 and surface elevation @xmath85 are taken as @xmath86 = z + h and @xmath87 .
the contribution for velocity has and rotational component apart from potential due to vorticity generation .
+ after we redefine @xmath88 and @xmath89 , we express potential and rotational component up to order @xmath90 as , @xmath91 \\ u _ { r}(z ) = \tilde u _ { r } - \bigtriangleup u(z ) + \eta s _ { \eta } + \mu^ { 2 } ( s _ { { wl } } - \bar s _ { { wl } } ) \\ - \cfrac{\mu^ { 2}}{2 } ( \frac{1}{3 } - z^ { 2})d^ { 2 } { \eta } \bigtriangledown ( \bigtriangledown \cdot s _
{ \eta } ) \\ - { \mu^ { 2 } } ( \cfrac{1}{2 } - z)d { \eta } \left [ ( \bigtriangledown \cdot s
_ { \eta})\bigtriangledown h - \bigtriangledown(s _ { \eta } \cdot \bigtriangledown h ) \right ] \end{split}\ ] ] similar expressions for vertical velocity are @xmath92 - \cfrac{\mu^ { 2}}{2 } ( ( \cfrac{1}{3}-z^ { 2 } ) \bigtriangledown \cdot \left [ d^ { 2 } ( h + z ) ( \bigtriangledown ( \bigtriangledown \cdot \tilde u _ { p } ) \right ] \\ - \mu^ { 2}(\frac{1}{2}-z ) \left [ \bigtriangledown \cdot ( h+z ) ( \bigtriangledown ( u _ { p } \cdot \bigtriangledown h ) + ( \bigtriangledown \cdot u _ { p } ) \bigtriangledown h ) \right ] \end{split}\ ] ] @xmath93 \\ - \cfrac{\mu^ { 2}}{2 } \left [ \bigtriangledown \cdot ( h+z ) ( \cfrac{1}{3 } - z^ { 2})d^ { 2 } { \bigtriangledown}(\eta \bigtriangledown \cdot s _
{ \eta } ) \right ]
\\ + { \mu^ { 2 } } \bigtriangledown \cdot \lbrack ( \frac{1}{2 } -z)d \left [ \bigtriangledown(\eta s _ { \eta}.\bigtriangledown h ) - ( \bigtriangledown \cdot \eta s _ { \eta})\bigtriangledown h \right ] \end{split}\ ] ]
the most obvious approach to solve boussinesq equation is to drop notion of pursuing an expansion in powers of @xmath4 and instead use weakly dispersive expression for @xmath94 or horizontal velocity in the form of power series in @xmath95 to evaluate complete surface boundary condition .
we refer to this procedure as @xmath96 in the sense that all of the availble information on velocities is used to evaluate boundary conditions . conventional time dependent boussinesq equations for surface wave height and consequent breaking term calculation are very straight forward and have benn calculated in case of irrotational waves . here
we take up fully nonlinear calculation as vorticity becomes a large fraction of water depth in the surf zone or shoaling waves .
while developing boussinesq equations for horizontal momentum , we retain up to order o(@xmath97 and o(@xmath98 in our fully nonlinear calculation . fully nonlinear boussinesq equations for long wave
have been derived by mei [ 1983 ] for flat bottom and by wei et al @xcite for variable bottom surface in case of irrotational wave .
shen @xcite addressed problems in developing generalized three dimensional irrotational propagating wave field to include rotational motion in general but did not describe the vorticity breaking terms . for horizontal propagation of waves ,
the three dimensional problem can be reduced in terms of two horizontal velocity by integrating over depth and retaining up to order @xmath99 and @xmath100 as horizontal velocity is governed by momentum equation at the surface @xmath101 by , @xmath102 in the surf zone region of sloping beach , waves break due to high vorticity and the breaking of wave later being converted to turbulence .
so horizontal variation of water depth @xmath103 must be considered in this case .
we express surface propagation equation in terms of average velocity description and total time derivative of horizontal momentum can be written as , @xmath104 where surface velocity is given by , @xmath105 we consider @xmath106 for wavy bottom @xmath107 @xmath108 this long wave momentum equation upon simplification over flat bottom case can be compared to the one derived by shen [ 2000 ] the vertical velocity can be obtained similarly , + @xmath109 @xmath110 is defined in previous section .
in contrast to the result by shen @xcite , additional contribution factor here arises from vorticity variation which is significant for surf zone wave .
wei et al @xcite also calculated breaking term for irrotational long wave momentum equation over a variable bottom wave .
the intermediate depth velocity @xmath111 is being used there proportional to h instead of depth average velocity used here which may not be valid inside the fluid .
the use of @xmath112 in our approach avoids this difficulty .
finally we try to generalize equation by solving vorticity from vorticity transport equation in next section .
madsen and svendsen @xcite used a cubic vertical distribution of rotational velocity based on roller jump data which can not considered in three dimension case as it is not guaranteed to bring accuracy in the simulation .
so we try to solve vorticity function from reynold stress based equation .
@xmath113 taking the curl on both sides and use vorticity function @xmath114 we get , @xmath115 @xmath116 is called `` vorticity stretching '' factor and is the gradient in vorticity value . this term leads to change of rotation of material particles present in the flow to the beach .
contribution of this term can not be incorporated from two dimension roller jump data .
we generalize the equation in three dimension as @xmath117 \end{split}\ ] ] after changing the variable from ( x , y , z , t ) to wave following coordinates @xmath118 , we write the vorticity equation as @xmath119 \\ - { \delta}^{\rm 2 } \frac{\sigma u } { ( h+\delta \eta ) } ( \bigtriangledown \cdot \eta ) \cfrac{\partial s}{\partial \sigma } + { \delta}^{\rm 2 } \cfrac{\sigma s}{(h+\delta\eta)}(\bigtriangledown \cdot \eta ) \cfrac{\partial u}{\partial \sigma }
\\ = \nu \left [ { \mu^{\rm 2 } } \bigtriangledown^{\rm 2}s + \cfrac{1}{(h+\delta
\eta ) } \cfrac{\partial^{\rm 2}s}{\partial \sigma^{\rm 2 } } \right ] + o(\mu^{\rm 2 } ) + o(h_{\rm x } ) + o(h_{\rm y } ) \end{split}\ ] ] the boundary conditions in new coordinate system are , @xmath120 after we redefine @xmath121 , which transforms the equation , @xmath122 \\ + { \delta}u(\bigtriangledown \cdot \omega ) - \cfrac{\delta \sigma } { ( h+ \delta \eta ) } ( \bigtriangledown \cdot \eta ) \cfrac{\partial \omega } { \partial \sigma } - \cfrac{\delta \sigma^ { 2}}{(h+\delta \eta ) } ( \bigtriangledown \cdot \eta ) \\
+ \cfrac{\delta u}{(h+\delta \eta ) } \cfrac{\partial \omega } { \partial \sigma } + \cfrac{\delta w}{(h+\delta \eta ) } \omega _ { s } \\ - \cfrac{\delta \omega}{(h+ \delta \eta ) } \cfrac{\partial w } { \partial \sigma } - \cfrac{\delta \sigma \omega _ { s}}{(h+ \delta \eta ) } \cfrac{\partial w}{\partial \sigma } \\ - \delta \omega ( \bigtriangledown \cdot u ) - { \delta}{\sigma } \omega _ { s } ( \bigtriangledown \cdot u ) + \cfrac { \delta^ { 2 } \sigma \omega}{(h + \delta \eta)}(\bigtriangledown \cdot \eta ) \cfrac{\partial u } { \partial \sigma } + \cfrac{\delta^ { 2 } \sigma \omega}{(h+\delta \eta ) } ( \bigtriangledown \cdot \omega)\cfrac{\partial u}{\partial \sigma } \\ + \frac { \delta^ { 2 } \sigma \omega _ { s}}{(h+\delta \eta ) } ( \bigtriangledown \cdot \eta ) \cfrac{\partial u}{\partial \sigma } \end{split}\ ] ] with new boundary , @xmath123 with initial condition @xmath124 .
this additional equation can be solved numerically as done by briganti et al .
@xcite for the two dimensional case or an analytical solution can be formulated as shown by veeramony & svendsen @xcite .
the analytical solution can be calculated by assuming @xmath125 which gives first and second solution as @xmath126 where the solution is @xmath127 assuming @xmath128 and to solve @xmath129 , assume @xmath130 which gives zeroth order solution as @xmath131 to consider @xmath132 where @xmath133 to solve above equation , assume @xmath134 where solution becomes @xmath135 with @xmath136 the solution for vorticity s becomes , @xmath137 to solve breaking term , we need value of @xmath138 for boundary and eddy viscosity value as input data . the parametric form of vertical profile of eddy viscosity is being applied to calculate vorticity function using various experimental data .
the eddy viscosity distribution over the water column @xmath139 is assumed such that its maximum value is located at the water surface except at the roller where the maximum is located at the lower edge of the roller .
the value is estimated by a mixing length hypothesis and @xmath140 is given by , @xmath141 the best fit for @xmath142 based on experimental data for hydraulic jump @xcite , @xmath143 where @xmath144 is the position of the toe of the roller and @xmath145 is the roller length .
figure 1 . shows our analytical calculation time evolution of vorticity profile for @xmath146 .
finally we conclude here by developing a most generalized form fully nonlinear boussinesq equations for wave propagation in surf zone region with variable bathymetry .
the vorticity distribution was calculated using vorticity transport equation(vte ) . in the wave breaking zone , vorticity generated by
the shear stress of current is very strong , we showed in our calculation the contribution to the surface velocity due to vorticity variation which has significant contribution in fluid flow .
these extra terms in generalized equation complicate the numerical technique as these terms are present in the equation in multiple form of equations for vorticity components which has to be solved in coupled solution technique .
veeramony @xcite used simplified formulation by taking constant eddy viscosity value but this oversimplified case may bring inaccuracy in calculation .
briganti et.al @xcite formulated a numerical technique scheme to solve vte using generalized depth variable eddy viscosity @xmath147 in two dimension case . in three dimensional formulation , the nonlinear terms in the vorticity transport equation(vte )
will complicate the calculation and so proper numerical technique have to be developed .
turbulence develops because of the instability of vortical flow , turbulence injection can be explained from vortical function development .
@xmath9 wave amplitude ; @xmath10 characteristic wave length ; @xmath148 vorticity function ; @xmath15 gravitational constant ; @xmath149 bottom friction constant
finnaly we speacially r. dutta thanks office of naval research lab ( nrl ) under grant [ gr001820 ] for financial support of this work .
kennedy , a.b .
, chen q. , kirby j.t . & dalrymple r.a . , boussinesq modeling of wave transformation , breaking and run - up i : one dimension , _ j. waterway , port , coastal & ocean engineering _ , * 126*,(1999 ) , 206 - 214 . | in this report , mathematical model for generalized nonlinear three dimensional wave breaking equations was developed analytically using fully nonlinear extended boussinesq equations to encompass rotational dynamics in wave breaking zone .
the three dimensional equations for vorticity distributions are developed from reynold based stress equations .
vorticity transport equations are also developed for wave breaking zone .
this equations are basic model tools for numerical simulation of surf zone to explain wave breaking phenomena .
the model reproduces most of the dynamics in the surf zone .
non linearity for wave height predictions is also shown close to the breaking both in shoaling as well as surf zone .
* keyword * wave breaking , boussinesq equation , shallow water , surf zone .
pacs : 47.32-y . |
the dynamical evolution of dense star clusters is a problem of fundamental importance in theoretical astrophysics .
star clusters like open and globular clusters are among the simplest stellar systems : they are spherical , they contain no dust to confuse the observations and they appear to have no dark matter .
moreover , they are dynamically old : a typical star in a globular cluster has completed some @xmath0 orbits since the cluster was formed and processes like gravothermal collapse and two - body relaxation occur on timescales comparable with their ages .
thus , they provide the best physical realization of the gravitational n - body problem i.e. to understand the evolution of a system of n point masses interacting only by gravitational forces . in spite of the many advances made in the recent past , many aspects of the problem have remained unresolved like the production of exotic objects ( ferraro et al .
2012 ) , the importance of tidal - shocks in the long term evolution and survival of star clusters in the galaxy ( gnedin , lee & ostriker 1999 ) and the ability to retain dark remnants ( morscher et al . 2013 ; sippel & hurley 2013 ) .
the most direct approach to the simulation of star clusters is through n - body simulations . in these kind of studies
the gravitational forces of stars are directly computed and any additional ingredient like e.g. binaries , tidal field , stellar evolution , etc . can be easily incorporated .
for this reason , in many cases n - body simulations represent the unique tool to face with complex topics within the gravitational n - body problem .
however , several processes involved in the dynamical evolution of a star cluster occur on different timescales , so that a direct scaling of the result of an n - body simulation to larger number of particles is not possible ( baumgardt 2001 ) .
although the grape series of special - purpose computers is steadily increasing in performance and the development of graphic processing units ( gpus ) computing , direct n - body simulation of the evolution of clusters with more than a few percent binaries and a moderate number of stars ( @xmath1 ) is still computationally expensive , with computational timescales of the order of months . until now only open clusters such as m67 and the arches cluster ( hurley et al .
2005 ; harfst , portegies zwart & stolte 2010 ) and loosely bound globular cluster objects such as palomar 4 and palomar 14 ( zonoozi et al .
2011 , 2014 ) have been modelled at the necessary level of sophistication .
alternative numerical methods to simulate the evolution of star clusters have been developed in past years including fluid models ( larson 1970 ; angeletti & giannone 1977a , b ) , orbit - averaged fokker - planck methods ( cohn 1980 ; takahashi 1995 ) and monte carlo simulations ( hnon 1971 , hereafter h71 ; giersz 1998 ; joshi , rasio & portegies - zwart 2000 ) .
monte carlo methods can be regarded as a hybrid between direct n - body integrations and numerical solutions of the fokker - planck equation . in this approach
the system is modelled as a sample of `` superstars '' i.e. a subsample of stars sharing the same mass and integrals of motions . in spherical systems
the motion of each superstar depends only on its energy and angular momentum and on the cluster potential , this last quantity being a unique function of positions and masses of the superstars . through an iterative algorithm
it is therefore possible to follow the evolution of the system once suited perturbations to the integrals of motion of the superstars are applied to account for the effect of two - body interactions . within the family of monte carlo methods
two approaches can be distinguished : _
i ) _ the _ orbit - following _ method ( also known as the `` princeton method '' ; spitzer & hart 1971 ) where the orbits of the superstars are directly computed and _ ii ) _ the _ orbit - averaged _ method ( the `` cornell method '' ; h71 ) where only the energies and angular momenta of superstars are monitored . while orbit - following methods are more suited to follow all those processes occurring on the dynamical timescale ( such as the evaporation of stars from the cluster , the violent relaxation , the tidal shocks and the phase of post - core collapse ) , orbit - averaged methods are computationally less expensive since perturbations to the integrals of motions need to be computed at time - steps that are short compared to the relaxation time and the time consuming integration of the orbits is not required
moreover , it is particularly easy to add more complexity and realism to the simulations one layer at a time and they are particularly easy to be parallelised . for these reasons ,
orbit - averaged monte carlo simulations have been employed by a number of groups to study the dynamical evolution of globular clusters and the dense centers of galaxies . in recent years
a particular effort have been made by these groups to include the effect of a mass spectrum ( giersz 2001 ) , stellar evolution ( joshi , nave & rasio 2001 ) , three- and four - body interactions ( giersz & spurzem 2003 ; fregeau & rasio 2003 ; fregeau et al .
2007 ) and a simplified treatment of a tidal field ( giersz et al . 2013 ; takahashi & baumgardt 2012 ) .
one of the most complex process to be modelled in monte carlo simulations is the escape from a cluster in an external tidal field .
in fact , while stars can escape from an isolated cluster only when they have positive energies , when the cluster moves within a tidal field the effective potential felt by a cluster star is perturbed and has a maximum at a distance ( called `` tidal radius '' ) which depends on the shape of the external field and on the orbital parameters of the cluster .
if the star moves beyond this radius the gravitational attraction of the cluster will not balance the combined effect of the external tidal field and the centrifugal force and the star will escape from the cluster .
for this reason the presence of the external field accelerates the escape process and consequently the whole structural evolution of the system ( spitzer 1987 ) .
since the first pioneering studies by h71 , the effect of a steady external field has been modelled by removing stars able to reach an apocenter larger that the tidal radius .
this last quantity was estimated considering the distance of the lagrangian point in the simple case of an external potential produced by a point mass on a cluster moving on a circular orbit . even in this simple case
, however , this criterion represents only a rough approximation .
indeed , the presence of the external field breaks the spherical symmetry of the effective potential and the size of the tidal radius depends on the direction of the star motion . in practice
, stars can escape from the cluster only close to the direction of the galactic center through the so - called `` 2nd and 3rd lagrangian points '' .
moreover , in real clusters , once a star reach the energy required to escape , it needs several crossing times to reach the right direction ( the so - called `` potential escapers '' ) thus producing a delayed escape .
fukushige & heggie ( 2000 ) derived a simple prescription to estimate the timescale of escape as a function of the excess of energy with respect to the lagrangian point energy level . on the basis of this last result , giersz et al .
( 2013 ) adopted a delayed escape criterion which successfully account for this effect .
the situation is more complex when eccentric orbits are considered : in this case , the hamiltonian is time - dependent and the tidal radius can be only instantaneously determined .
in this situation , the aperture in the phase - space for a star to escape changes with time and stars can have only a limited amount of time to reach such an aperture .
moreover , stars escaping from the cluster can be re - captured when the cluster expand during its motion away from the perigalacticon . to further complicate the picture ,
it has been shown that stars with prograde and retrograde motion escape with different efficencies ( read et al .
2006 ) and there are stars which permanently remain bound to the cluster outside the tidal radius ( `` non - escapers '' ; ross , mennim & heggie 1997 ) . finally , the potential of the milky way can not be realistically approximated as a point mass but consists of many non - spherical components . in a non -
spherical potential orbits are in general non - planar and rapid changes of potential can produce compressive shocks which increase the kinetic energy budget of cluster stars ( gnedin & ostriker 1997 ) .
in this paper we present a new orbit - averaged monte carlo code able to simulate the evolution of a star cluster moving on an eccentric orbit within a realistic external potential . in sect .
2 we describe the code and the modification made to the original algorithm described by h71 . in sect .
3 the recipies to account for the effect of the external field are outlined .
sect . 4 in devoted to the description of the set of simulations and their comparison with n - body simulations .
we summarize our results in sect .
a detailed derivation of the tidal radius and effective potential in complex potentials is provided in the appendix .
the code presented here is an updated version of the orbit - averaged monte carlo method extensively described in h71 ( see also stodolkiewicz 1982 , giersz 1998 and joshi et al .
the basic idea of this approach is to consider the cluster as a sample of superstars characterized by mass ( @xmath2 ) , energy ( @xmath3 ) and angular momentum ( @xmath4 ) per unit mass generating a spherical symmetric potential ( @xmath5 ) .
the evolution of the cluster is divided in time - steps ( @xmath6 ) of variable duration . at each time - step the following steps are performed 1 .
the optimal time - step is determined ( see eq . 10 by joshi et al . 2000 ; see also sect .
[ esc_sec ] ) ; 2 .
a statistical realization of the cluster is performed by placing the superstars at random positions along their orbits .
each star is placed at a given distance from the cluster center according to the inverse of the star velocity at that distance ( see below ; see also sect .
7 of h71 ) ; 3 .
the cluster potential profile is evaluated according to the masses and positions of the superstars ( see below ) ; 4 .
the mechanical work made by the ( internal+external ) potential change on the suparstars is calculated and corrections to the stars energies are applied ( see sect . [ shocks_sec ] ; see also sect .
4 of stodolkiewicz 1982 ) ; 5 .
each superstar is assumed to interact with its nearest neighbor producing a perturbation on its energy and angular momentum ( see sect . 5 of h71 ) .
stars satisfying the escape criterion ( see sect .
[ esc_sec ] ) are removed from the simulation .
the above steps are repeated until the end of the simulation .
a detailed description of the algorithms adopted to perform the above steps is provided in h71 , giersz 1998 and joshi et al .
( 2000 ) and will not be repeated here .
below we describe only the modifications to their approaches regarding the way the cluster potential is determined while in sect .
[ ext_sec ] we extensively describe the adopted escape criterion .
the most time - consuming step of the procedure outlined above is the distribution of the superstars across the cluster .
this is done by extracting a variable @xmath7 ( with @xmath8 ) from the distribution @xmath9 where @xmath10 and @xmath11 are the pericenter and apocenter of the star orbit within the cluster potential and @xmath12 is the radial component of the star velocity at @xmath13 the values of @xmath10 and @xmath11 are first determined from the star energy and angular momentum and the cluster potential at the two zeros of the function @xmath14 then @xmath7 is extracted from @xmath15 using the von neumann rejection technique .
h71 demonstrated that the above algorithm ensures that the probability to extract a given position is proportional to the time spent by the star in that position . according to h71
, the cluster potential at a given distance from the cluster center @xmath16 can be determined in a straightforward way from the positions and masses of the superstars using the poisson equation in its discrete integral form @xmath17 where @xmath18 and @xmath19 are the position and mass of the i - th superstar , @xmath20 is the total number of superstars and @xmath21 is the index such that @xmath22 .
this approach is efficient and adaptive by definition i.e. in the densest regions of the cluster it provides a better sampling of the potential .
however , from a computational point of view , the above procedure is quite expensive since it requires a cycle over the n superstars to find the index @xmath21 which must be repeated at least two times for each superstar to find @xmath10 , @xmath11 and @xmath12 . for this purpose we decided to calculate the cluster potential at the beginning of each time
- step on a grid of @xmath23 evenly spaced radial steps .
the potential at the distance @xmath16 is then determined by linearly interpolating between the two contiguous knots @xmath21 and @xmath24 . in this case
the index @xmath21 is immediately found as @xmath25 .
this simple modification speed up the entire process by a factor of ten .
we developed two independent methods to define the potential in the grid knots which are used in different conditions .
the first method ( hereafter referred as the _ fast method _ ) is to use eq .
[ pot1_eq ] defining at each time - step a step - size of the grid @xmath26 where @xmath27 is the cluster core radius ( casertano & hut 1985 ) and @xmath28 is the superstar number density at @xmath18 calculated using the 50 nearest neighbor superstars .
this is the fastest method and provides a good accuracy for most of the cluster evolution when clusters with moderate concentrations and a large number of superstars are considered .
unfortunately , something is lost in the above modification : in the advanced stages of core collapse a large fraction of stars is contained within the innermost radial bins .
in this situation , the spatial resolution of the grid is not adequate to follow the fast evolution of the cluster core and produces an unrealistic delay of the core collapse .
moreover , another drawback of this method ( which is in common with the canonical method adopted by h71 ) is that the potential depends on the position of the superstars .
as positions are randomly extracted , it is possible that fluctuations in the potential are present when a small number of superstars are considered .
this effect is particularly strong in the cluster core where the potential is determined by few stars and can produce a `` spurious relaxation '' .
h71 have shown that such an effect is negligible when a large number ( @xmath29 ) of particles is used . however
, a pernicious effect is produced by such fluctuations when the correction for the mechanical work made by the potential ( step ( iv ) of the above scheme ; see stodolkiewicz 1982 ) is considered .
indeed , fluctuations are erroneously interpreted as real potential changes introducing spurious corrections in the stars energies . on the long term
, this produces a drift in the total cluster energy accelerating the relaxation process even when a number of superstars as large as @xmath30 is considered . for this reason we developed another method ( hereafter referred as the _ integral method _ ) to determine the cluster potential in the grid knots .
consider a star with energy @xmath31 and angular momentum @xmath32 moving in the potential @xmath5 .
when the star is in the j - th radial bin @xmath33 its radial component of the velocity will be given by eq .
[ vr_eq ] and its radial component of the acceleration will be @xmath34 these quantities have been estimated using the potential and its derivative calculated in the previous time - step . in the approximation that the star moves with a uniformly accelerated motion ,
the time spent to cross the interval ( @xmath35 , @xmath36 ) is @xmath37 the probability to find the star in that interval will be @xmath38 where @xmath39 is the orbital period of the superstar .
the potential at each point of the grid can be then calculated through the relation @xmath40 where @xmath21 is the index such that @xmath41 .
this method has the advantage to depend only on the energies and angular momenta of the stars and not on the randomly extracted positions of the superstars . in practice , it is equivalent to compute the potential from an infinite number of randomly extracted positions thus eliminating the problem of fluctuations .
moreover , it allows to estimate the orbital period of each superstar ( eq .
[ per_eq ] ) which will be used in the escape algorithm ( see sect .
[ esc_sec ] ) .
there are two main drawbacks of this methods : first , to determine @xmath42 and @xmath43 it is necessary to use the potential profile calculated in the previous time - step .
for this reason , at odds with the _ fast method _ , the potential can be computed only on grid of positions which is fixed in time .
therefore , a small step - size is required from the beginning of the simulation to adequately follow the cluster evolution also in the advanced stages of core collapse .
we found that a good sampling of the potential profile during the entire cluster evolution is provided by the choice of @xmath44 where @xmath45 is the cluster core radius ( see eq .
[ rc_eq ] ) at the beginning of the simulation .
second , the above method is computationally expensive and almost all the improvement provided by the adoption of the evenly spaced grid is lost . to optimize the speed of the simulation without losses of accuracy , during the simulation we adopt the _ fast method _ when @xmath46 , and switch to the _ integral method _ when the above condition is not satisfied .
in this section we describe the treatment of the external tidal field in our monte carlo code . here
we considered two kinds of external field potential : the first generated by a point mass galaxy with mass @xmath47 and the second by the analytical bulge+disk+halo potential defined in johnston , spergel & hernquist ( 1995 ; hereafter j95 ) .
the orbit of the cluster within these potentials has been computed starting from its orbital energy and z - component of the angular momentum using a fourth - order runge - kutta algorithm providing an accuracy in terms of energy conservation better than @xmath48 during the entire evolution .
the effect of dynamical friction has been neglected since it is expected to be negligible on the star cluster scale ( gnedin et al .
1999 ) .
the presence of an external field imply an increase of star losses because of two main processes : escape of stars through the cluster boundaries and tidal shocks ( this last process occurring only when the external potential has a bulge / disk component ) .
we discuss the algorithms to include the above processes in the following sections . and @xmath49 moving on a circular orbit at @xmath50 around a point - mass galaxy of @xmath51 .
the dashed , solid and dotted lines ( red , black and blue in the online version of the paper ) indicate the effective potential along the x- , y- and z - axis , respectively.,width=328 ] as already introduced in sect .
[ intro_sec ] , the presence of an external tidal field imply the presence of a tidal cut in the cluster potential . stars with enough energy can cross the cluster tidal radius and evaporate from the cluster on a timescale comparable to the star orbital period .
a star orbiting around a cluster immersed in an external field feels an effective potential given by the combination of the cluster potential ( @xmath52 ) , the external field potential ( @xmath53 ) and a term linked to the angular motion of the cluster . @xmath54 where * @xmath55 * is the angular speed of the cluster and * r * is the position vector of the star in a reference frame centered on the center of mass of the galaxy+cluster system rotating with angular speed @xmath56 .
note that the jacobi integral associated to the above effective potential is conserved only in the particular case of a cluster moving on a circular orbit within a spherical potential .
however , when other orbits / potentials are considered , the timescale on which such an integral changes is longer than the dynamical time of most cluster stars so that we can assume it instantaneously conserved .
consider a cartesian reference system centered on the cluster with the x - axis pointed toward the galaxy center , the y - axis parallel to the galactic plane and directed toward the cluster rotation and the z - axis perpendicular to the previous axes .
the acceleration felt by the a star approaching the tidal radius @xmath57 will be @xmath58 where @xmath59is the position vector of the star , @xmath60 is the cluster mass and @xmath61 is the acceleration due to the cluster angular motion .
the tidal radius is defined as the distance where the projection of the above acceleration on * r * is zero which corresponds to the radius at which the effective potential has a local maximum .
note that , according to the above definition , both the tidal radius and the effective potential depend on the direction of the escape . it can be shown ( see appendix ) that the shortest and less energetic tidal radius occurs in correspondence of the x - direction ( i.e. the 2nd and 3rd lagrangian points ) while in the z - direction
the effective potential is a growing function of the radius and no maxima exist ( see fig .
[ pot ] ) .
for this reason stars escape preferentially from the lagrangian points . to determine wether a star escapes from the cluster we defined three distinct criteria .
the first criterion we adopted is that the star energy and angular momentum allow the motion across the tidal radius . for this purpose we extracted two random numbers ( @xmath62 and @xmath63 )
uniformly distributed between 0 and 1 and defined the coefficients @xmath64 these coefficient univocally define a direction of escape such that @xmath65 .
the tidal radius and effective potential in the defined direction are then calculated ( see appendix ) as well as the mechanical work made by the external potential on the superstar ( @xmath66 ; see below ) . the first criterion is satisfied if @xmath67 the second criterion is based on the fact that the escaping star needs a timescale comparable to its orbital period ( @xmath68 ) to reach the tidal radius .
this produces a delay in the escape process .
this can be crucial when eccentric orbits are considered : in this case , @xmath57 changes with time and the star satisfies eq .
[ crit1_eq ] only in a limited time interval .
if such time interval is short compared to its orbital period , there is only a small probability for the star to escape .
we then extracted a random number @xmath69 uniformly distributed between 0 and 1 and calculated the orbital period of the star to reach the tidal radius using eq .
[ per_eq ] .
we assumed that the the star can escape during the time - step @xmath6 if @xmath70 once the first two cirteria are satisfied the star can escape from the cluster .
after this phase the star moves in the galactic potential as an independent satellite following an epicyclic orbit along either the trailing or the leading arm of the cluster tidal tails . during this motion ,
the distance of the star from the cluster center oscillates according to the orbital phase of the cluster . to definitively escape from the cluster attraction
there is a typical timescale which depends on the star s energy and on the cluster orbital parameters .
moreover , if the cluster follows an eccentric orbit its tidal radius grows when the cluster leaves the perigalacticon and it can possibly exceed the distance of the previously escaped star . in this case
the star is re - captured . to account for this effect , when a star satisfies the two above mentioned criteria its position and velocity
are calculated using eq .
[ pos_eq ] ( @xmath71 ) and @xmath72 where @xmath73 and @xmath74 is a random number uniformly distributed between 0 and 1 .
the above phase - space coordinates are transformed in the galactic reference system , added to the cluster systemic coordinates and the orbit of both the star and the cluster within the galactic potential are followed using a fourth - order hermite integrator with an adaptive timestep for an entire cluster orbital period . during its motion outside the cluster the distance of the star from the cluster and the tidal radius are calculated .
the star is removed from the simulation if the conditions @xmath75 are satisfied during a cluster orbital period . in the above equations @xmath76 is the maximum tidal radius reached by the cluster during an orbital period , @xmath6 is the time - step , @xmath68 is the star orbital period within the cluster and @xmath69 is the random number extracted for the criterion in eq .
[ crit2_eq ] .
the last modification regards the time - step adopted in the simulation when the external field is present . in joshi et a. ( 2001 ) the time - step is defined to ensure small deflection angles when introducing the perturbations to the supertars energies and angular momenta .
however , another requirement is that the jacobi integral should not significantly vary within the time - step .
so , we adoted as time - step the minimum between the time - step defined by joshi et al .
( 2001 ) and 0.01@xmath77 , where @xmath77 is the cluster orbital period .
when a cluster passes through the galactic disk or close to the galactic bulge the gravitational field of these two components exerts a compressive force which is superposed on the cluster s own gravitational field .
this process is known as `` disk / bulge shocking '' ( ostriker , spitzer & chevalier 1972 ; aguilar , hut & ostriker 1988 ) .
the theory of tidal shocks has been studied in the past by many authors and applied to orbit - averaged fokker - planck codes by gnedin & ostriker ( 1997 , 1999 ) and allen , moreno & pichardo ( 2006 ) .
this effect can be viewed as a consequence of the mechanical work made by the external potential during the orbit . indeed ,
any variation of the potential shape during the cluster evolution produces a work on the stars which is equal to the instantaneous potential variation .
the variation of the internal potential can be due to many processes ( secular dynamical evolution , evaporation of stars , stellar evolution , etc . ) and is taken into account using the prescriptions by stodolkiewicz ( 1982 ) . in time - dependent external potentials
( e.g. when an eccentric orbit and/or non - spherical potentials are considered ) also the external potential makes a work on the stars .
this is particularly important when fast changes in the external potential occurs , like in the case of the disk crossing and the perigalactic passages . in this case , the binding energy per unit mass of the superstars located in the outskirts of the cluster changes on a timescale comparable to the dynamical time .
this effect facilitates the escape of stars and can accelerate the cluster dynamical evolution .
in analogy with stodolkiewicz ( 1982 ) , at each time - step we estimated the work made by the external potential on each superstar as the average between the potential changes at the positions of the superstar in two subsequent time - steps @xmath78/2\\\end{aligned}\ ] ] such a correction is updated during the orbit and temporary added to the energy of the superstar only to verify the escape criteria defined above .
this correction can not indeed be permanently added the superstars energies since these are distributed across the cluster adopting the internal potential only ( i.e. assuming the cluster as isolated ) while the external potential is taken into account only in a subsequent step when the escape criteria are verified .
[ cols="<,^,^,^,^,^,^,^,^,^ , > " , ] = 0 , 0.14 , 0.33 and 0.6 are drawn with solid , dashed , dotted and dot - dashed lines , respectively.,width=328 ] but for the set of simulations with 42000 particles.,width=328 ] = 2 , 3 , and 4 kpc are drawn with solid , dashed , and dotted lines , respectively.,width=328 ] we tested the prediction of our code with a set of collisional n - body simulations performed with nbody6 ( aarseth 1999 ) . all simulations ( both monte carlo and n - body ) were run with n=21000 and n=42000 single - mass particles as a compromise to ensure a large number statistics and to limit the computational cost of n - body simulations .
the standard @xmath79 parameter has been used to control the time step and set an energy error tolerance of @xmath80 . with these choices we got a relative error in energy smaller than @xmath81 at the end of every simulation run . for each simulation
the cluster mass contained within the apocentric tidal radius and the lagrangian radii have been calculated and compared .
lagrangian radii have been calculated from the cumulative distribution of distances from the cluster center of the particles .
here we considered the radii containing 1% , 2% , 5% , 10 - 90% of the cluster mass within one apocentric tidal radius .
we considered three different cases : _
i ) _ an isolated cluster with a plummer ( 1911 ) profile ; _ ii ) _ a cluster with a mass @xmath82 , a king ( 1966 ) profile with @xmath83 and @xmath49 orbiting in a galactic potential generated by a point - mass of @xmath84 with an apogalacticon at @xmath85 , and _
iii ) _ the same cluster of ( ii ) orbiting in the bulge+disk+halo galactic potential defined by j95 . in cases _ ( ii ) _ and _ ( iii ) _ a set of simulations with different eccentricities and orbits have been considered .
the whole set of simulations is summarized in table 1 . in fig .
[ plum ] the evolution of the lagrangian radii of the isolated plummer ( 1911 ) model as a function of the initial half - mass relaxation time is compared with that predicted by the n - body simulation with n=8192 particles by baumgardt , hut & heggie ( 2002 ) .
the agreement is excellent with only a small discrepancy near the core collapse for the innermost radii .
the core collapse occurs after @xmath86 when the cluster have lost @xmath873% of its stars .
both quantities are in good agreement with the results of baumgardt et al .
the excellent agreement with the n - body simulation indicates that the monte carlo code well reproduces the relaxation process until the core collapse .
another set of simulations have been performed considering an external tidal field generated by a point - mass galaxy .
in such a potential stars feel the tidal cut but are not subject to disk shocks .
these simulations are therefore suited to test the escape from the cluster boundary both in case of circular and eccentric orbits .
simulations have been run until core collapse .
the subsequent evolution is largely influenced by the presence of binaries which form during the maximum density phase .
as our monte carlo code does still not account for this process it can not reproduce properly such an evolutionary stage .
the evolution of the cluster mass and of the lagrangian radii for this set of simulations are compared with the results of n - body simulations in fig .
[ pm_mass ] and [ pm_lag ] , respectively .
it is apparent that in all the simulations with eccentricity @xmath88 the monte carlo tends to systematically overpredict the mass - loss rate with respect to n - body simulations , although such a discrepancy is always within 5% .
the evolution of the lagrangian radii is also well reproduced during the entire cluster evolution .
a different situation is for the simulation of the most eccentric ( @xmath89 ) orbit . in this case , while the evolution of the bound mass is well reproduced by our monte carlo code , the lagrangian radii are strikingly different . in particular , while the n - body simulation predict an overall contraction of the cluster , the monte carlo code predicts a quick expansion followed by a quick collapse of the core .
a possible reason for such a discrepancy is that in this last orbit the tidal radius penetrates into the cluster at pericenter leaving a significant fraction of cluster stars free to escape
. when this occurs , the criteria defined in sect .
[ esc_sec ] are not adequate anymore and an unrealistically large fraction of stars evaporate from the cluster in a short amount of time .
the loss of potential energy is larger than that in kinetic energy and the cluster expands , furthermore increasing the escape efficiency .
as the number of stars in the cluster becomes smaller the relaxation process speeds up and the core quickly collapse .
summarizing , it appears that the treatment of the external tidal field described in sect .
[ esc_sec ] is effective when the escape rate of stars during a cluster orbital period is smaller than a critical value . in fig .
[ pm_mass42 ] and [ pm_lag42 ] the comparison between the bound mass and lagrangian radii for the same set of simulations with 42000 particles are shown , respectively . also in this case , the agreement is good in all the simulations with moderate eccentricity ( @xmath88 ) both in terms of the mass and of the lagrangian radii evolution .
again , for the most eccentric simulation , while the bound mass evolution is fairly well reproduced , the lagrangian radii of the monte carlo simulation show the same unproper behaviour already noticed in the simulations with 21000 particles .
to properly estimate the range of validity of our code , we performed another set of simulations where a cluster with 21000 particles and the same structural characteristics of the previous simulations has been launched on circular orbits at different distances from the point - mass galaxy ( @xmath902,3 and 4 kpc ) .
the results of such an experiment are shown in fig.s [ pm_testmass ] and [ pm_testlag ] .
as expected , monte carlo simulations related to clusters moving at large galactocentric distances show a good agreement with n - body ones . on the other hand ,
as the tidal field becomes stronger the mass - loss rate predicted by the monte carlo code exceeds that that of the n - body simulations . as a consequence ,
the cluster internal dynamical evolution accelerates and the core collapse is anticipated . according to the escape criteria defined in sect .
[ esc_sec ] , the timescale at which a star escapes from the cluster is of the order of a few dynamical times . while this timescale depends on the energy and angular momentum of each star and on the cluster potential , it is expected to scale with the global quantity @xmath91 .
it is therefore useful to introduce the mass - loss rate per half - mass dynamical time @xmath92 the above parameter has been calculated during the cluster evolution and averaged from the beginning of the simulation to the core collapse ( see table 1 ) . to define a criterion of validity of our simulations we correlated the values of @xmath93 with the discrepancy between bound mass fraction predicted by monte carlo and n - body simulations measured at core - collapse in this set of simulations .
we obtain @xmath93=-0.00007 , -0.00013 and -0.00040 and @xmath94=1.2% , 3.0% and 15.9% for the simulations at 4 , 3 and 2 kpc , respectively . on the basis of the above comparison and defining a resonable agreement at @xmath955%
, we adopt a value of @xmath96 - 0.0002 as a conservative limit of validity of our code .
the last set of simulations considers a cluster immersed in the three - components bulge+disk+halo external potential defined by j95 .
we considered a cluster lying at an initial distance of 10 kpc from the galactic center with two different eccentricities and two different heights above the galactic plane . in each case
the intensities of the tidal shocks are different since the cluster cross the disk and approaches its pericenter at different distances from the galactic center with different velocities .
the three considered orbits in the x - y , x - z and r - z planes are shown in fig .
the mass evolution of the three simulations are compared with those predicted by the associated n - body simulations in fig.s [ compl_mass ] and [ compl_mass42 ] for the set with 21000 and 42000 particles , respectively .
it can be noted that in all the simulations with 21000 particles the agreement is good ( within 5% ) during the entire evolution . on the other hand , in simulations with 42000 particles a tendency of the monte carlo code to underestimate
the cluster mass - loss rate is noticeable .
this is particularly apparent in the simulations with eccentric orbits where the difference with respect to the prediction of the n - body simulation reach @xmath8710% at the core - collapse .
such a discrepancy is in the opposite sense of what observed in simulations run within a point - mass potential , where the mass - loss rates were slightly overpredicted .
note that in the three simulations the average mass - loss rate is @xmath97 i.e. below the critical limit where significant differences have been noticed in the simulations within a point - mass external potential .
it is also interesting to note that the mass evolution of the two simulations with different heights above the galactic plane ( k - j95-e033 and k - j95-e033z ) yield to a quite similar residual mass after 12 gyr .
this is not surprising since at the moment of the disk crossing , although the disk shocks are more intense in simulation k - j95-e033z because of the largest velocity of the cluster , the disk density is in both cases relatively small . also , the pericentric distance of both orbits is 5 kpc , significantly larger with respect to the bulge half - mass radius ( @xmath871.69 kpc ) .
so , in the above cases , both disk and bulge shocks have only a little impact on the cluster structural and dynamical evolution . in fig.s
[ compl_lag ] and [ compl_lag42 ] the evolution of the lagrangian radii of the three simulations are compared with the predictions of the corresponding n - body simulations for the sets with 21000 and 42000 particles , respectively .
again the agreement is good during the entire cluster evolution .
in this paper we presented a new implementation of the monte carlo method to simulate the evolution of star clusters including for the first time the effect of a realistic tidal field and the possibility to consider eccentric orbits .
the effect of the external field has been taken into account considering both the process of evaporation through the cluster boundaries and the effect of tidal shocks .
the adopted algorithm is based on the theory of evaporation taking into account for the random direction of the escape , the time delay due to the star orbital period and the occurrence of the re - capture process .
the comparison with direct n - body simulations indicates an excellent agreement in an isolated cluster and a good agreement ( within 5 - 10% ) in clusters within a tidal field of moderate intensity , in terms of the evolution of both the mass and the lagrangian radii .
this indicates that both the process of two - body relaxation and the evaporation of stars are accurate enough . on the other hand ,
a significant discrepancy is apparent when an extremely intense tidal stress is present ( @xmath98 ) .
the reason of the disagreement relies on the failure of some ( maybe all ) of the criteria adopted to take into account the escape of stars when these apply to a significant fraction of cluster stars .
it is worth noting that such an intense tidal field should be felt by only a small fraction of globular clusters ( only 2 out the 53 globular clusters considered by allen et al .
2006 , 2008 ) .
so this code can provide an efficient tool to study the evolution of present - day globular clusters . on the other hand ,
the code developed here could fail to reproduce the initial stages of cluster evolution : in this case , the fast potential changes due to stellar evolution driven mass loss likely lead to a situation of roche - lobe overfilling where the critical mass loss rate is easily reached .
consider that one of the basic assumptions of the monte carlo method is spherical symmetry which conflicts with the presence of the external field .
in this situation star orbits are not expected to be planar and it is not guaranteed that two superstars with contiguous ranking in distance from the cluster center are neighbors .
so , any treatment of the external field is expected to fail when somehow strong tides are considered .
the performance of the code are also very good : the simulation k - pm - e0 - 21k presented here with 21000 particles takes @xmath8745 minutes ( `` clock on the wall '' time ) on a single five - years old asus machine equipped with a single intel core t5800@2ghz processor , while the corresponding n - body simulation takes @xmath8730 hours with a cluster node equipped with an intel xeon e5645 cpu ( 12 cores ) and a nvidia tesla m2090 ( 512 cuda cores ) . in spite of the relatively small considered number of particles , the above performance is good considering that the code include several cycles which can be in principle easily parallelized thus reducing the computational cost of simulations . assuming a scaling of the computation time with the number of particles as @xmath99 , it is possible to run simulations with @xmath100 with computation time of few days .
the code already include other features to properly simulate the cluster evolution like the inclusion of a mass spectrum , the mass - loss driven by stellar evolution and the direct integration of three- and four - body interactions following the prescriptions by joshi et al .
( 2001 ) and fregeau et al .
however , the predictions of the code with these implemented features have not been tested and are not presented here .
a forthcoming paper will introduce these features in the next future .
as acknowledges the prin inaf 2011 `` multiple populations in globular clusters : their role in the galaxy assembly '' ( pi e. carretta ) .
amb whish to thank the lady davis foundation .
we thank the anonymous referee for his / her helpful comments and suggestions .
we warmly thank holger baumgardt to have provided his n - body simulations and zeinab khorrami , carlo nipoti , luca ciotti and enrico vesperini for useful discussions .
99 aarseth s. j. , 1999 , pasp , 111 , 1333 aguilar l. , hut p. , ostriker j. p. , 1988 ,
apj , 335 , 720 allen c. , moreno e. , pichardo b. , 2006 , apj , 652 , 1150 allen c. , moreno e. , pichardo b. , 2008 , apj , 674 , 237 angeletti l. , giannone p. , 1977
, ap&ss , 50 , 311 angeletti l. , giannone p. , 1977
, a&a , 58 , 363 baumgardt h. , 2001 , mnras , 325 , 1323 baumgardt h. , hut p. , heggie d. c. , 2002 , mnras , 336 , 1069 casertano s. , hut p. , 1985
, apj , 298 , 80 cohn h. , 1980 , apj , 242 , 765 ferraro f. r. , et al . , 2012 , natur , 492 , 393 fregeau j. m. , rasio f. a. , 2007 , apj , 658 , 1047 fregeau j. m. , grkan m. a. , joshi k. j. , rasio f. a. , 2003 , apj , 593 , 772 fukushige t. , heggie d. c. , 2000 , mnras , 318 , 753 giersz m. , 1998 , mnras , 298 , 1239 giersz m. , 2001 , mnras , 324 , 218 giersz m. , spurzem r. , 2003 , mnras , 343 , 781 giersz m. , heggie d. c. , hurley j. r. , hypki a. , 2013 , mnras , 431 , 2184 gnedin o. y. , ostriker j. p. , 1997 ,
apj , 474 , 223 gnedin o. y. , ostriker j. p. , 1999 ,
apj , 513 , 626 gnedin o. y. , lee h. m. , ostriker j. p. , 1999 ,
apj , 522 , 935 harfst s. , portegies zwart s. , stolte a. , 2010 , mnras , 409 , 628 henon m. , 1969 , a&a , 1 , 223 hnon m. h. , 1971 , ap&ss , 14 , 151 hernquist l. , 1990 , apj , 356 , 359 hurley j. r. , pols o. r. , aarseth s. j. , tout c. a. , 2005 , mnras , 363 , 293 johnston k. v. , spergel d. n. , hernquist l. , 1995 , apj , 451 , 598 joshi k. j. , nave c. p. , rasio f. a. , 2001 , apj , 550 , 691 joshi k. j. , rasio f. a. , portegies zwart s. , 2000 , apj , 540 , 969 king i. , 1962 , aj , 67 , 471 king i. r. , 1966 , aj , 71 , 64 larson r. b. , 1970 , mnras , 147 , 323 miyamoto m. , nagai r. , 1975 , pasj , 27 , 533 morscher m. , umbreit s. , farr w. m. , rasio f. a. , 2013 , apj , 763 , l15 ostriker j. p. , spitzer l. , jr . , chevalier r. a. , 1972 , apj , 176 , l51 plummer h. c. , 1911 , mnras , 71 , 460 read j. i. , wilkinson m. i. , evans n. w. , gilmore g. , kleyna j. t. , 2006 , mnras , 366 , 429 ross d. j. , mennim a. , heggie d. c. , 1997 , mnras , 284 , 811 sippel a. c. , hurley j. r. , 2013 , mnras , 430 , l30 spitzer l. , 1987 , princeton , nj , princeton university press spitzer l. , jr . , hart m. h. , 1971 , apj , 164 , 399 takahashi k. , 1995 , pasj , 47 , 561 takahashi k. , baumgardt h. , 2012 , mnras , 420 , 1799 stodolkiewicz j. s. , 1982 , aca , 32 , 63 zonoozi a. h. , haghi h. , kpper a. h. w. , baumgardt h. , frank m. j. , kroupa p. , 2014 ,
mnras , 440 , 3172 zonoozi a. h. , kpper a. h. w. , baumgardt h. , haghi h. , kroupa p. , hilker m. , 2011 , mnras , 411 , 1989
as discussed in sect . [ esc_sec ] , when the cluster is immersed in an external tidal field its stars feel an effective potential due to the combination of the internal cluster potential ( @xmath52 ) , the external field potential ( @xmath53 ) and a term linked to the cluster angular motion .
such an effective potential does not have a spherical symmetry so that the distances of its local maxima ( tidal radii ) are direction dependent . in the next sections the tidal radius and the effective potential as a function of the direction of escape
are derived both for the case of an external field generated by a point - mass and for the complex bulge+disk+halo potential by j95 .
the external potential generated by a point - mass galaxy is @xmath101 where @xmath102 is the mass of the galaxy and @xmath16 the distance of a test particle from the point - mass . consider a star at @xmath16 orbiting around a cluster located at @xmath103 . in the cartesian reference frame centered on the cluster with the x - axis pointed toward the point - mass galaxy , the y - axis in the direction of the rotation and the z - axis in the direction of the angular momentum ,
be @xmath104 the position vector of the star .
the distance of the star from the galaxy will be @xmath105 the tidal radius is defined as the point where the projection of the acceleration felt by the star ( eq .
[ acc_eq ] ) on @xmath106 is zero .
the acceleration is composed by three terms : one associated to the cluster potential , one associated to the external field and another associated to the cluster angular motion .
the term associated to the external field is @xmath108\nonumber\\ & = & \frac{g m_{g}}{|{\bf r'}| |{\bf r}|^{5}}[x^{2}\left(-2 r_{cl}^{2}+|{\bf r'}|^{2}+4 x r_{cl}-3 x^{2}\right)+y^{2}\left(r_{cl}^{2}+|{\bf r'}|^{2}-2 x r_{cl}-3 y^{2}\right)+z^{2}\left(r_{cl}^{2}+|{\bf r'}|^{2}-2 x r_{cl}-3 z^{2}\right)]\nonumber\end{aligned}\ ] ] in the limit @xmath109 the above term can be approximated to @xmath110 .
the acceleration due to the angular motion of the cluster is @xmath111 where @xmath55 is the angular speed of the cluster which , in an elliptic orbit with eccentricity @xmath112 and apogalactic distance @xmath113 is @xmath114 the third term of equation [ pmat_eq ] has null projection on @xmath59 since it is orthogonal to @xmath59 by definition . in the cluster reference system the remaining terms ( corresponding to the coriolis and the centrifugal+tidal acceleration ) have projection on @xmath59 @xmath115 .
the total acceleration in the direction of @xmath59 at @xmath57 will be given by the sum of eq.s [ pmc_eq ] , [ pmf_eq ] and [ pma_eq ] @xmath116 where @xmath117 @xmath4 is the angular momentum of the superstar and @xmath74 is a random number uniformly distributed between 0 and 1 ( see eq .
[ vel_eq ] ) .
it is interesting to note that the coriolis acceleration is directed toward the direction of escape when the star is on a prograde orbit , while it is directed toward the cluster when retrograde orbits are considered .
this is the reason why stars on prograde orbits escape more easily from the cluster ( see also hnon 1969 ; read et al .
2006 ) .
note that over a large number of random extractions of @xmath74 the term @xmath121 has null mean . in this case
the tidal radius will simply be @xmath122 at pericenter and in correspondence to the lagrangian point @xmath123 will be @xmath124 , and the above equation reduces to that reported by king ( 1962 ) .
we consider the three components galactic model by j95 which consists of the superposition of an hernquist ( 1990 ) bulge , a miyamoto & nagai ( 1975 ) disk and a logarithmic halo @xmath127 with @xmath128 , @xmath129 , @xmath130 , @xmath131 , @xmath132 , @xmath133 and @xmath134 . here
we adopted a convenient cylindrical reference system @xmath135 , @xmath136 is the distance from the galactic center and @xmath137 is the angle between the z - axis and the position vector . in this reference system
the cluster will be in @xmath138 .
define an alternative cartesian rotating reference frame centered in the cluster center with the x - axis pointed toward the galactic center , the y - axis parallel to the galactic plane and pointed toward the direction of the cluster rotation and the z - axis perpendicular to the other axes .
consider a star in @xmath139 moving around a cluster immersed in the above external potential .
the coordinate transformation to the first reference system are @xmath140 and @xmath105 in analogy with what done in sect .
[ pm_sec ] , we calculate the projection of the acceleration felt by the star ( eq .
[ acc_eq ] ) on @xmath106 .
the internal acceleration is always given by eq .
[ pmc_eq ] . the term associated to the external field can be calculated separately for the three galactic components @xmath141\nonumber\\ \frac{{\bf r'}}{|{\bf r'}| } \cdot ( { \bf r ' } \cdot { \bf \nabla } ) \nabla \phi_{d}&=&g m_{d } |{\bf r'}|\left\{\frac{1+\frac{a b^{2}}{(b^{2}+z_{cl}^{2})^{\frac{3}{2}}}(\tilde{x}^{2}cos^{2}\theta+\tilde{z}^{2}sin^{2}\theta ) } { \left[r_{cl}^{2}+\left(a+\sqrt{b^{2}+z_{cl}^{2}}\right)^{2}\right]^{\frac{3}{2}}}- 3\frac{\left(r_{cl}^{2}+\frac{2 a z_{cl}^{2}}{\sqrt{b^{2}+z_{cl}^{2}}}\right ) \tilde{x}^{2}+\frac{a^{2 } z_{cl}^{2}}{b^{2}+z_{cl}^{2}}(\tilde{x}^{2}cos^{2}\theta+\tilde{z}^{2}sin^{2}\theta ) } { \left[r_{cl}^{2}+\left(a+\sqrt{b^{2}+z_{cl}^{2}}\right)^{2}\right]^{\frac{5}{2}}}\right\}\nonumber\\ \frac{{\bf r'}}{|{\bf r'}| } \cdot ( { \bf r ' } \cdot { \bf \nabla } ) \nabla \phi_{h}&=&\frac{2 v_{0}^{2 } |{\bf r'}|}{(d^{2}+r_{cl}^{2})^{2}}[(d^{2}+r_{cl}^{2})(\tilde{y}^{2}+\tilde{z}^{2})+(d^{2}-r_{cl}^{2})\tilde{x}^{2 } ] .
\label{complf_eq}\end{aligned}\ ] ] where @xmath142 the acceleration due to the angular motion of the star is given by eq .
[ pmat_eq ] , where @xmath143 and @xmath144 is the z - component of the cluster angular momentum . as in the case of the point - mass potential ,
only the coriolis ( * @xmath145 * ) and the centrifugal+tidal ( * @xmath146 * ) terms of the acceleration have a projection on @xmath59 different from zero .
so @xmath147- \dot{\theta}_{cl}^{2 } |{\bf r'}|(\tilde{x}^{2}+\tilde{z}^{2})+ \frac{2 l_{z } \dot{\theta}_{cl } |{\bf r'}|}{r_{cl}^{2 } } \tilde{y } ( \tilde{x}~cos\theta_{cl}-\tilde{z}~sin\theta_{cl})\nonumber\\ & & -2\dot{\theta}_{cl}(\tilde{x}~v_{z}-\tilde{z}~v_{x})-\frac{2 l_{z}}{r_{cl}^{2}}[\tilde{x}~v_{y}sin\theta_{cl}-\tilde{y } ( v_{x}sin\theta_{cl}+v_{z}cos\theta_{cl})+\tilde{z}~v_{y}cos\theta_{cl } ] .
\label{compla_eq}\end{aligned}\ ] ] it is convenient to introduce the characteristic densities @xmath148\right\}. \label{complb_eq}\end{aligned}\ ] ] where @xmath149 is either the acceleration due to the i - th galactic component ( @xmath150 ) or the centrifugal+tidal acceleration , @xmath4 is the angular momentum of the superstar and @xmath74 is a random number uniformly distributed between 0 and 1 ( see eq .
[ vel_eq ] ) .
it is therefore possible to combine eq.s [ pmc_eq ] , [ complf_eq ] , [ compla_eq ] and [ complb_eq ] to obtain the projection of the total acceleration felt by the star on @xmath59 latexmath:[\[{\bf \nabla \phi_{eff}}=\frac{g m_{cl}}{|{\bf r'}|^{2}}-g \gamma the tidal radius will therefore be @xmath119 with @xmath152 again , the coriolis term has null mean over a large number of random extractions of @xmath74 , so we can calculate @xmath57 assuming @xmath153 .
@xmath154 note that at the lagrangian point @xmath123 the above formula is equivalent to the estimate provided by allen et al . | we present a new implementation of the monte carlo method to simulate the evolution of star clusters .
the major improvement with respect to the previously developed codes is the treatment of the external tidal field taking into account for both the loss of stars from the cluster boundary and the disk / bulge shocks .
we provide recipes to handle with eccentric orbits in complex galactic potentials .
the first calculations for stellar systems containing 21000 and 42000 equal - mass particles show good agreement with direct n - body simulations in terms of the evolution of both the enclosed mass and the lagrangian radii provided that the mass - loss rate does not exceed a critical value .
[ firstpage ] methods : numerical methods : statistical stars : kinematics and dynamics globular clusters : general |
the discovery of millisecond spin periods of neutron stars in low mass x - ray binaries ( lmxbs ) with the _ rossi x - ray timing explorer _ ( _ rxte _ ) has helped elucidate the nature of these sources .
neutron star lmxbs consist of a neutron star accreting from a low mass companion .
as material ( mostly h and he ) is accreted onto the star and gets compressed , it eventually ignites and burns unstably ( see * ? ? ?
this phenomenon is observed as a type i x - ray burst .
type i x - ray bursts have been observed from over @xmath070 lmxbs ( see * ? ? ? * and references therein ) .
the recurrence time of these bursts varies but in some cases it can be as frequent as every few hours ( see * ? ? ?
* for examples ) .
occasionally it is possible to observe the spin of the neutron star modulating the burst emission the so called `` burst oscillations '' @xcite .
burst oscillations have been observed from @xmath018 lmxbs ( see * ? ? ? * for examples ) .
x - ray bursts from lmxbs have been discovered which are @xmath01000 times longer , and thus , that much more energetic , than canonical type i x - ray bursts .
they are aptly named `` superbursts '' .
superbursts are believed to occur by the unstable burning of carbon @xcite .
@xcite discovered highly coherent pulsations during a superburst from 4u 1636@xmath253 .
the pulse train lasted for @xmath0800 s , as opposed to the @xmath010 s long pulse trains observed in type i x - ray bursts .
superbursts are much rarer than type i x - ray bursts .
thus far ten have been observed from eight lmxbs ( see * ? ? ?
between 2005 march 7 and 2005 july 21 the pca galactic bulge scan survey discovered an outburst from the direction of the globular cluster ngc 6440 .
a follow up 1.8-ks long pointed
_ rxte _ observation ( observation identification number 91050 - 03 - 07 - 00 ) was initiated on 2005 june 14 .
the data presented here were acquired from the proportional counter array ( pca ) on board _
rxte_. the pca consists of five identical and independent proportional counter units ( pcus ) .
each pcu is a 90% xenon/10% methane gas filled proportional counter .
each pcu has a collimated 1@xmath3@xmath41@xmath3 field - of - view , 256 spectral channels in the 260 kev range , and a limiting temporal resolution of @xmath01 @xmath5s .
only two pcus ( pcu 0 and pcu 2 ) were operational throughout the span of the observation .
the data were taken in ` e_125us_64m_0_1s ` mode , which returns events to a limiting resolution of 125 @xmath5s and with moderate ( 64 channels as opposed to the full 256 channels ) spectral resolution .
this mode was used because it is not as susceptible to buffer overflows during high countrate data as compared to the less restrictive modes ( e.g. ` goodxenon ` ) . using the pointing of the observation ( 17@xmath648@xmath7528 , -202132 ) and the planetary ephemeris de200 ,
the times of raw events were corrected to the solar system barycenter . binning the events into a 1-s resolution lightcurve
reveals a quickly decaying burst ( fig .
[ fig : lc ] ) .
this burst was reminiscent of canonical type i x - ray bursts from lmxbs , therefore we searched this event for burst oscillations .
we rebinned the raw events using the full spectral band pass into a time series with 0.5/1024 s resolution , which yields an equivalent nyquist frequency of 1024 hz . a leahy normalized ( see * ? ? ?
* ) power density spectrum ( pds ) is displayed in figure [ fig : lc ] ( inset ) .
a prominent peak is clearly seen at 442 hz .
the probability of this peak occurring by chance after accounting for the number of trials ( the total number of frequency bins in our pds ) is @xmath0@xmath8 .
ngc 6440 harbors the bright x - ray transient sax j1748.9@xmath22021 , which exhibited a 409.7 hz burst oscillation @xcite . however , there are many other x - ray sources in the cluster ( see fig [ fig : chandra image ] )
. _ chandra x - ray observatory _ ( _ cxo _ ) imaging by @xcite revealed 24 x - ray sources in ngc 6440 , and they concluded that 45 of these sources are likely quiescent lmxbs .
thus , the phenomenon we discovered is most likely from a different object in the cluster . ] to study the frequency evolution of the pulse we calculated a dynamic @xmath9 statistic .the @xmath9 statistic is analogous to the fourier power spectrum , with the advantage that the data need not be binned , thus allowing us to oversample our data .
the @xmath9 statistic is defined as : @xmath10 where @xmath11 is the total number of photons in each interval , @xmath12 is the total number of harmonics that one deems significant , @xmath13 is the frequency searched over , and @xmath14 is the event time .
the factor in front of the summations normalizes the @xmath9 statistics in an analogous way to leahy normalizing a pds .
we calculated the dynamic @xmath9 power spectrum using a 200-s long window , which was translated across the observation with a step size of 16 s. our dynamic @xmath9 power spectrum is displayed in fig .
[ fig : frequency evolution ] .
notice that the pulsations where highly significant for 576 s. as is common for canonical x - ray bursts , the frequency drifts ; however , unlike the large @xmath01 - 2 hz drifts seen in those , the pulsation here drifts down in frequency only by @xmath02.1@xmath15 hz in 576 s. to further quantify this frequency evolution we determined how the phase of the pulsations varies as a function of time . for each interval with @xmath16 we generated a pulse profile by folding the events in that interval on the frequency determined from the @xmath9 statistic .
we then cross correlated these pulse profiles with a sinusoid of fixed phase to determine the pulse phase as a function of time .
we can model the pulse phase ( @xmath17 ) at a given time ( @xmath18 ) by the following taylor expansion : @xmath19 where @xmath20 is the barycentric frequency , @xmath21 is the frequency derivative , and @xmath22 is some reference epoch . we were able to `` whiten '' our phase residuals with just a frequency derivative ( see fig .
[ fig : frequency evolution ] , bottom ) .
we find @xmath23 hz , @xmath24 hz s@xmath25 at @xmath22=53535.463215 mjd ( utc ) .
the phase residuals after subtracting the model given by eq .
[ eq : phase ] are shown in fig .
[ fig : frequency evolution ] ( bottom panel ) .
notice that the pulse train lasts for 576 s ; this is quite long when compared to those of type i x - ray bursts .
we folded all the events within the 576 s pulse train using our best fit @xmath20 and @xmath21 , and the resulting pulse profile is shown in fig .
[ fig : profile ] .
notice that the pulse profile is highly sinusoidal .
fitting a sinusoid to the pulse profile returned a reduced @xmath26 value of 0.76 for 13 degrees of freedom , affirming the absence of any harmonic content . from the fit
we obtain a fractional pulse amplitude of [email protected]% .
it is possible that the frequency drift we have observed is entirely due to orbital motion .
the observed spin frequency of a neutron star in a circular orbit is given by @xmath28 , \label{eq : orbit}\ ] ] where , @xmath29 is the barycentric frequency at time @xmath30 , @xmath31 is the projected velocity of the neutron star , @xmath32 is the orbital period , and @xmath33 is the orbital phase at time @xmath30 .
unfortunately we do not have a long enough data set to place interesting constraints on @xmath32 or @xmath31 ; however , we can determine whether orbital modulation is consistent with the observed frequency drift .
for example , if we assume , for solely demonstrative purposes , that this system has a projected velocity comparable to the one found by @xcite for 4u 1636@xmath253 ( @xmath34 km @xmath25 ) , then fitting eq .
[ eq : orbit ] to our frequency time series yields a good fit with a reasonable orbital period , i.e. @xmath357@xmath271 hr . hence , binary motion can account for the observed frequency drift .
observing another outburst would be essential in ultimately constraining the orbital parameters of this system .
the timing analysis seems to suggest that this event is in fact very different from a canonical burst oscillation . to study it
further we performed a spectral analysis .
we analyzed the spectral evolution of the burst by breaking the burst up into ten intervals .
the exposure of each interval was selected by demanding that each one contained an equal number of photons . using the last interval as an estimator of the persistent emission
, we fit each interval to a simple photoelectrically absorbed blackbody model while holding the column density fixed to the value found by @xcite for the optical reddening .
we find only subtle evidence for spectral softening , significant only at the 1-@xmath36 level . using the distance estimate to the cluster found by @xcite , @xmath37 kpc , we were able to calculate the luminosity of the burst . at the start of the observation the flux was @xmath02@xmath4@xmath38 erg s@xmath25 , which is @xmath00.1@xmath39 , where @xmath39 is the eddington luminsosity for a neutron star .
the burst lightcurve was well fit by an exponential with a decay timescale of 22 s. unfortunately the observation only caugth the tail end of the burst - like event , we could not therefore characterize the peak luminosity of the burst .
the timing properties of the pulsations are reminiscent of those observed during superbursts .
in particular the highly coherent pulsations and the long pulse trains . the superburst interpretation would be very exciting given the rarity of such events .
however , we could not unambiguously determine this is the case given that we only caught the tail end of the burst . to study this phenomenon
further we extrapolated our luminosity model determined in our spectral analysis before the start time of the observation .
we find that the burst reaches the eddington luminosity @xmath050 s before the start of the observation . now ,
if we assume that we caught the tail of the outburst it is very surprising that it dropped from @xmath0@xmath39 down to @xmath0.01@xmath39 in @xmath077 s. this is much too fast for a superburst - like event .
if , on the other hand , this is a precursor events , such as those seen in superbursts , than it is unclear why the persistent flux did not rise back to @xmath0@xmath39 shortly after as observed in other superbursts .
thus , the spectral analysis of this event suggests it is not a superburst .
the timing properties of the pulsations share all the hallmarks of those observed during superbursts , but the spectral analysis makes this interpretation unlikely
. if this event is not a superburst , then could it be that these highly coherent pulsations have eminated from an accretion - powered millisecond pulsar ?
recently @xcite discovered `` intermittent '' pulsations from the accretion - powered millisecond pulsar ( amp ) hete j1900.1@xmath22455 .
they found that the properties of this pulsar differed from those of the other six known amps .
for example , the pulsations were only present in the first few months of its outburst as opposed to other amps which show pulsations throughout ( see * ? ? ? * for a review ) .
in addition , the pulse amplitude increased at three points in time that were almost coincident with the times of thermonuclear bursts .
the pulsed amplitude subsequently decaded after the bursts .
thus , it is plausible that we have observed a similar phenomenon . following the theoretical work of @xcite and @xcite , @xcite suggest that the reason why the pulsations in hete j1900.1@xmath22455 sometimes `` switch off '' is because the magnetic field is burried by accreted material , thus it can not channel the flow which gives rise to the pulsations in the first place if the event we have discovered is another example of these intermittent pulsations from an unidentified source in ngc 6440 , then a future outburst should reveal more long pulse trains , and if they happen to be correlated with the times of type - i x - ray bursts then this would solidify the connection between this source and hete j1900.1@xmath22455 .
we discovered a 442 hz pulsation during a burst from an x - ray source in ngc 6440 .
we could not establish the energetics of the burst as _
rxte _ only caught the tail end of the event , but based on the timing properties of the pulsation we conclude that it was not a canonical burst oscillation . in particular we find that : the pulse train lasted for 576 s , as opposed to the @xmath010 s long pulse trains observed in type i x - ray bursts .
the signal was highly coherent and drifted down only by @xmath40 hz , as opposed to type - i x - ray burst which exhibit drifts which are three orders of magnitude larger .
the pulsations share all the properties of those observed during superbursts ; however , the complete energetics of this event make this interpretation implausible .
we conclude that the long pulse train is most likely an `` intermittent '' pulsation from an accreting millisecond pulsar such as those seen only in hete j1900.1 - 2455 thus far @xcite .
we note that ngc 6440 is known to harbor a transient x - ray source with a 409.7 hz spin frequency @xcite however , x - ray imaging of the cluster by @xcite has shown that ngc 6440 contains many x - ray sources .
thus , the burst presented here most likely emanated from a different object in the cluster .
fpg is supported by the nasa postdoctoral program administered by oak ridge associated universities at nasa goddard space flight center .
this research has made use of data obtained through the high energy astrophysics science archive research center online service , provided by the nasa / goddard space flight center . | we report on the serendipitous discovery of a 442-hz pulsar during a _ rossi x - ray timing explorer _ ( _ rxte _ )
observation of the globular cluster ngc 6440 .
the oscillation is detected following a burst - like event which was decaying at the beginning of the observation .
the time scale of the decay suggests we may have seen the tail - end of a long - duration burst .
low - mass x - ray binaries ( lmxbs ) are known to emit thermonuclear x - ray bursts that are sometimes modulated by the spin frequency of the star , the so called burst oscillations .
the pulsations reported here are peculiar if interpreted as canonical burst oscillations . in particular
, the pulse train lasted for @xmath0500 s , much longer than in standard burst oscillations .
the signal was highly coherent and drifted down by @xmath02@xmath1hz , much smaller than the @xmath0hz drifts typically observed during normal bursts .
the pulsations are reminiscent of those observed during the much more energetic `` superbursts '' , however , the temporal profile and the energetics of the burst suggest that it was not the tail end nor the precursor feature of a superburst .
it is possible that we caught the tail end of an outburst from a new ` intermittent' accreting x - ray millisecond pulsar , a phenomenon which until now has only been seen in hete j1900.1@xmath22455 @xcite . we note that @xcite reported the discovery of a 409.7 hz burst oscillation from sax j1748.9@xmath22021 , also located in ngc 6440 .
however , _ chandra x - ray observatory _ imaging indicates it contains several point - like x - ray sources , thus the 442 hz object is likely a different source . |
in epidemic studies the analysis of large amounts of personal data is essential . at the same time
the dissemination of the results of those studies , even in a compact and summarized form , can provide some information that can be exploited to identify the row pertaining to a certain individual .
for instance , zip code , gender and date of birth can uniquely identify 87% of individuals in the u.s .
therefore when managing personal data it is of the utmost importance to effectively protect individuals privacy .
one approach to deal with such problem is the @xmath0-anonymity model @xcite .
each row of a given table represents all data regarding a certain individual .
then different rows are clustered together , and some entries of the rows in each cluster are suppressed ( i.e. they are replaced with a @xmath3 ) so that each cluster consists of at least @xmath0 identical rows . therefore each row @xmath4 in the resulting table is clustered with at least other @xmath5 rows identical to @xmath4 , hence the resulting data do not allow to identify any individual . while such formulation is not really sophisticated and has some practical limitations , it is definitely interesting from a theoretical point of view , as witnessed by the rich literature available .
we will focus on separating the cases that can be solved efficiently from those that are intractable , therefore hinting at which strategies are likely or not going to be successfully employed when studying more sophisticated formalizations .
notice that different formulations of the problem have also been proposed @xcite , for example allowing the generalization of entry values , that is an entry value can be replaced with a less specific value @xcite , or considering a notion of proximity among values @xcite .
a parsimonious principle leads to the optimization problem where we want to minimize the number of entries in the table to be suppressed .
the @xmath0-anonymity problem is known to be apx - hard even when the matrix entries are over a _
alphabet and @xmath2 @xcite , as well as when the matrix has @xmath6 columns and @xmath7 ( this time on arbitrary alphabets ) @xcite .
furthermore , a polynomial - time @xmath8-approximation algorithm on arbitrary input alphabet , as well as approximation algorithms for restricted cases are known @xcite . recently
, two polynomial - time approximation algorithms with factor @xmath9 have been independently proposed @xcite . in this paper
we investigate the parameterized complexity @xcite of the problem , unveiling how different parameters are involved in the complexity of the problem .
a first systematic study of the parameterized complexity of the @xmath0-anonymity problem has been proposed in @xcite .
here , we follow the same direction , showing that the problem is w[1]-hard when parameterized by the size of the solution and @xmath0 , and we provide a fixed - parameter algorithm , when the problem is parameterized by the number of columns and the maximum number of different values in any column .
these problems were left open in @xcite . in table [ tab : par - compl - status ] we report the status of the parameterized complexity of the @xmath0-anonymity problem , where in bold we have emphasized the new results presented in this paper .
we recall that a problem @xmath10 parameterized by a set @xmath11 of parameters is in the class fpt @xcite if it admits an exact algorithm with complexity @xmath12 , where @xmath13 is an arbitrary function , and @xmath14 is the size of the input problem , while it is w[i]-hard @xcite , for some @xmath15 if it is unlikely to be fixed - parameter tractable .
we recall that xp @xcite is a superclass of all sets w[@xmath16 .
moreover , proving that a problem @xmath17 with parameter set @xmath18 is np - hard when all parameters in @xmath18 are some constants , implies that @xmath19 xp unless p = np .
.summary of the parameterized complexity status of the @xmath0-anonymity problem ; @xmath20 represents the maximum number of different values in a column , @xmath21 represents the number of columns , @xmath14 represents the number of rows , @xmath0 represents the minimum size of a cluster , @xmath22 represents the size of the solution . [ cols="^,^,^,^,^",options="header " , ] the rest of the paper
is organized as follows . in section [ sec : pre - sec ]
we introduce some preliminary definition and we give the formal definition of the @xmath0-anonymity problem . in section [ sec : par - hard ] we show that the @xmath0-anonymity is w[1]-hard . in section
[ sec : fpt - algo ] we give a fixed parameter algorithm , when the problem is parameterized by the size of the alphabet and the number of columns .
finally , in section [ sec : ap-3 - 3 ] we show that the @xmath1-anonymity problem is apx - hard , even when the rows have length bounded by @xmath1 .
let us introduce some preliminary definitions that will be used in the rest of the paper . given a graph @xmath23 , and @xmath24 , the _ subgraph induced _ by @xmath25 is denoted by @xmath26=(v ' , e')$ ] , where @xmath27 .
a graph @xmath23 is cubic when each vertex in @xmath28 has degree three . given an alphabet @xmath29 , a row @xmath4 is a vector of elements taken from the set @xmath29 , and the @xmath30-th element of @xmath4 is denoted by @xmath31 $ ] .
notice that it is equivalent to consider a row as a vector over @xmath29 or as a string over alphabet @xmath29 .
let @xmath32 be two equal - length rows .
then @xmath33 is the hamming distance of @xmath34 and @xmath35 , i.e. @xmath36\neq r_2[i]\}|$ ] .
let @xmath37 be a set of @xmath38 rows , then a _ clustering _ of @xmath37 is a partition @xmath39 of @xmath37 . given a clustering @xmath40 of @xmath37
, we define the _ cost _ of the row @xmath4 belonging to a set @xmath41 of @xmath17 as @xmath42\neq r_2[j]\}|$ ] , that is the number of entries of @xmath4 that have to be suppressed so that all rows in @xmath41 are identical .
similarly we define the cost of a set @xmath41 , denoted by @xmath43 , as @xmath44\neq r_2[j]\}|$ ] .
the cost of @xmath17 , denoted by @xmath45 , is defined as @xmath46 .
given a set @xmath47 and a clustering @xmath17 of @xmath37 , the cost induced by @xmath17 in set @xmath18 is @xmath48 .
notice that , given a clustering @xmath49 of @xmath37 , the quantity @xmath50 is a lower bound for @xmath51 , since all the positions for which @xmath34 and @xmath35 differ will be deleted in each row of @xmath41 .
we are now able to formally define the @xmath0-anonymity problem ( @xmath0-ap ) .
@xmath0-ap . +
* input * : a set @xmath37 of equal lenght rows over an alphabet @xmath52 . + * output * : a clustering @xmath49 of @xmath37 such that for each set @xmath41 , @xmath53 and @xmath45 is minimum . in
what follows , given a set @xmath18 of parameters , we denote by @xmath54-ap the @xmath0-ap problem parameterized by @xmath18 , thus omitting @xmath0 .
we will consider the following parameters : @xmath21 is the number of columns of the rows in @xmath37 ; @xmath14 is the number of rows in @xmath37 ; @xmath20 is the maximum number of different values in any column of the table ; @xmath0 is the minimum size of a cluster ; @xmath22 is the maximum number of entries that can be suppressed .
let @xmath55 be a solution of the @xmath0-ap problem .
notice that a suppression at position @xmath30 of a row @xmath4 is represented replacing the symbol @xmath31 $ ] with a @xmath3 . given a set @xmath56 of @xmath17 ,
some entries of the rows clustered in @xmath56 are suppressed , so that the resulting rows are all identical to a vector @xmath4 over alphabet @xmath57 ; such a vector is the _ resolution vector _ associated with @xmath56 . given a resolution vector @xmath4
, we define @xmath58 as the number of entries suppressed in @xmath4 , that is @xmath59= * \}|$ ] . given a resolution vector @xmath4 and a row @xmath60 ,
we say that @xmath4 is _ compatible _ with row @xmath61 iff @xmath31 \neq r_i[j]$ ] implies @xmath31=*$ ] . given a row @xmath61 of @xmath37 and a set of resolution vectors @xmath62
, we define the set @xmath63 .
given a set @xmath37 of rows , we define a _ group _ of rows of @xmath37 as a maximal set of identical rows . given a group @xmath64 ,
the _ representative row _ of @xmath64 , denoted by @xmath65 , is any row of @xmath64 , while @xmath66 is the number of rows in @xmath64 and @xmath67 .
a set @xmath37 of rows can be partitioned in groups of identical rows in polynomial time @xcite , therefore we can compute in polynomial time whether a set @xmath37 of rows is @xmath0-anonymous , i.e. @xmath37 can be partioned into groups of size at least @xmath0 .
if this is not possible , then observe that at least @xmath0 entries of @xmath37 must be suppressed to get a solution of the @xmath0-ap problem , that is @xmath68 .
hence @xmath69-ap is in fpt iff @xmath70-ap is in fpt .
consequently our parameterized reduction @xcite will show the fixed - parameter intractability of @xmath69-ap and @xmath70-ap .
we show that @xmath69-ap and @xmath70-ap are w[1]-hard . given an set @xmath37 of equal length rows , @xmath69-ap and @xmath70-ap
ask if there exists a clustering @xmath72 of r such that @xmath73 for each set @xmath41 , and @xmath74 .
we present a parameter preserving reduction from the @xmath75-clique problem , which is known to be w[1]-hard @xcite , to the @xmath69-ap problem . given a graph @xmath23 , an @xmath75-clique is a set @xmath24 where each pair of vertices in @xmath25 are connected by an edge of @xmath76 , and @xmath77 .
the @xmath75-clique problem asks for a subset @xmath25 of the vertices of a given graph @xmath76 inducing an @xmath75-clique in @xmath76 .
clearly the vertices of a @xmath75-clique are connected by @xmath78 edges . given a graph @xmath23 , we use @xmath79 and @xmath80 to denote respectively the number of edges and of vertices of @xmath76 .
we construct the instance @xmath37 of @xmath69-ap associated with @xmath76 .
first , let us define @xmath81 .
the set @xmath37 consists of @xmath82 rows and @xmath83 columns over alphabet @xmath84 .
more precisely , for each edge @xmath85 in @xmath86 , there is a group @xmath87 of @xmath88 identical rows @xmath89 , @xmath90 , where * @xmath91= \sigma_{i , j}$ ] , for @xmath92 ; * @xmath93= 1 $ ] , @xmath94= 1 $ ] ; * @xmath95= 0 $ ] , for @xmath96 and @xmath97 .
moreover , @xmath37 also contains a group @xmath98 made of @xmath99 identical rows equal to @xmath100 .
[ lem : w[1]-hard : prop - base ] let @xmath37 be the instance of @xmath69-ap associated with @xmath76 and consider two rows @xmath101 of @xmath37 , such that @xmath102 and @xmath103 . then , @xmath104 \neq r_x(i , j)[t]$ ] , for each @xmath105 .
[ lem : w[1]-hard-1 ] let @xmath23 be a graph , let @xmath25 be a @xmath75-clique of @xmath76 and let @xmath37 be the instance of @xmath69-ap associated with @xmath76
. then we can compute in polynomial time a solution @xmath17 of @xmath69-ap over instance @xmath37 with cost at most @xmath106 .
[ lem : w[1]-hard-2 ] let @xmath23 be an instance of @xmath75-clique , let @xmath37 be the instance of @xmath69-ap associated with @xmath76 and let @xmath17 be a solution of @xmath69-ap over instance @xmath37 with cost at most @xmath106
. then we can compute in polynomial time a @xmath75-clique @xmath25 of @xmath76 .
first we will prove that @xmath17 must have a set @xmath107 .
assume to the contrary that in @xmath17 there are two sets @xmath108 , @xmath109 containing at least a row of @xmath98 .
notice that @xmath110 while @xmath111 .
moreover , by lemma [ lem : w[1]-hard : prop - base ] , all rows in @xmath108 or @xmath109 must have suppressed the first @xmath112 entries , which results in at least @xmath113 suppressions , contradicting the assumption on the cost of the solution .
hence , @xmath98 is properly contained in a set @xmath114 of @xmath17 , as @xmath115 .
moreover , let @xmath116 be a row of @xmath117 and let @xmath4 be a row of @xmath118 . by lemma [ lem : w[1]-hard : prop - base ] @xmath119 \neq r[t]$ ] for each column @xmath120 , @xmath105 , therefore all entries in the first @xmath112 columns of each row in @xmath114 must be suppressed .
now , let us prove that , for each set @xmath87 of @xmath37 , there exists a set @xmath121 of @xmath17 such that @xmath122 . assume to the contrary that no such set @xmath121 exists , for a given @xmath87 .
then either @xmath123 or there exists a row of @xmath87 clustered together with a row of @xmath124 in @xmath17 , with @xmath125 . in the first case , that is @xmath126 , @xmath127 , by construction all entries of the first @xmath112 columns of the rows in @xmath114
must be suppressed , resulting in at least @xmath128 suppressions and thus contradicting the assumption on the cost of the solution .
consider now the second case , that is there is a set @xmath108 in @xmath17 containing at least a row of two different sets @xmath87 and @xmath124 of @xmath37 .
observe that given @xmath129 and @xmath102 , @xmath4 and @xmath116 differ in the first @xmath112 columns . thus the entries of the first @xmath112 columns of the rows of @xmath114 must be suppressed , resulting in at least @xmath113 suppressed entries and thus contradicting the assumption on the cost of the solution .
hence , for each set @xmath87 of @xmath37 , there exists a set @xmath121 of @xmath17 such that @xmath122 . by our previous arguments
we can assume that @xmath17 consists of the clusters @xmath114 and @xmath121 , for each @xmath130 , and that @xmath131 .
notice that only @xmath114 can contain some suppressed entries .
also @xmath132 , for otherwise we can improve the cost of @xmath17 by moving a row in @xmath133 from @xmath134 to @xmath121 .
now let @xmath135 be the set of edges @xmath136 of @xmath76 such that a row of @xmath87 is in @xmath114 and let @xmath25 be the set of vertices incident on at least an edge in @xmath135 .
then we can show that @xmath26 $ ] is a @xmath75-clique .
notice that the entries in the first @xmath112 columns of @xmath114 must be suppressed , as well as all columns with index @xmath137 such that @xmath138 , since in those columns all rows in @xmath98 have value @xmath139 while some row in @xmath140 have value @xmath141 .
an immediate consequence is that the overall number of suppressed entries is at least @xmath142 .
since , by hypothesis , the number of suppressed entries is at most @xmath143 , then @xmath144 .
notice that , since @xmath145 and @xmath146 , then @xmath140 contains exactly @xmath147 distinct rows corresponding to edges in @xmath135 incident on @xmath25 vertices .
hence @xmath25 induces a @xmath75-clique in @xmath76 . from lemma [ lem : w[1]-hard-1 ] and
[ lem : w[1]-hard-2 ] , our reduction is parameter preserving , therefore @xmath69-ap and @xmath70-ap are w[1]-hard .
in this section we present a fixed - parameter algorithm for the @xmath148-approblem , that is the instance of the ap problem , where the number @xmath21 of columns and the maximum number @xmath20 of different values in any column are two parameters . notice that @xmath0-ap parameterized by exactly one of @xmath20 or @xmath21 is not in fpt , as @xmath0-ap is apx - hard ( hence np - hard ) even when one of @xmath20 or @xmath21 is a constant @xcite . before giving the details of the algorithm ,
let us first introduce some preliminary definitions .
let @xmath37 be an instance of @xmath148-ap , and for each column of @xmath37 with index @xmath30 , @xmath149 , let @xmath150 be the set of different values that the rows of @xmath37 have in column @xmath30 .
notice that @xmath151 , for each @xmath149 .
let @xmath152 and @xmath153 .
assume @xmath154 is a feasible solution of @xmath148-apover instance @xmath37 .
the set @xmath62 consisting of a resolution vector for each set @xmath155 is called _ candidate set _ for solution @xmath148-ap .
let @xmath18 be the set of possible rows of length @xmath21 and having value over alphabet @xmath156 for the position @xmath30 , @xmath149 , then @xmath157 is bounded by @xmath158 . given a candidate set @xmath62 , notice that @xmath159 and that each row @xmath160 must compatible with at least one resolution vector in @xmath62
. given a row @xmath4 and the set @xmath62 of resolution vectors , recall that we denote by @xmath161 the set of resolution vectors of @xmath62 compatible with @xmath4 .
moreover , given a resolution vector @xmath162 , we denote by @xmath163 the number of suppressions in @xmath116 . for each row @xmath164 we define its weight as @xmath165 .
notice that @xmath166 whenever @xmath4 is compatible with a row without suppressions .
informally , the weight of a row is equal to the maximum number of its entries that might be preserved in a solution where @xmath62 is the set of resolution vectors .
finally , we define @xmath167 and @xmath168 for each row @xmath169 .
notice that @xmath170 , for each @xmath171 .
the weights defined above will be used later in section [ subsect_build ] to define the weight function @xmath172 .
let us first describe the general idea of the algorithm . given a candidate set @xmath62 , the algorithm computes an optimal solution @xmath173 associated with a candidate set @xmath159 ( see algorithm [ alg : solving - ap ] ) .
the algorithm consists of two main phases . in the first phase ( section [ subsect_build ] ) , given the set @xmath37 of input rows and the candidate set @xmath62 , the algorithm builds a weighted bipartite graph @xmath174 associated with @xmath37 and @xmath62 . in the second phase ( section [ subsect_solution ] ) a solution of @xmath148-apis computed starting from a maximum weighted matching of the graph @xmath174 .
section [ subsect_opt_solution ] is devoted to prove that the solution computed by the algorithm is optimal .
[ alg : solving - ap ] @xmath175 the set of resolved vectors of length @xmath21 , where each @xmath30-th symbol , @xmath149 , is taken from the alphabet @xmath156 @xmath167 let us consider a candidate set @xmath62 of vectors for an optimal solution of @xmath148-ap . since @xmath159 , there exist at most @xmath177 possible candidate sets of rows @xmath62 , therefore our fpt algorithm computes each candidate set @xmath62 and verifies if there exists a solution @xmath173 with cost at most @xmath22 . in order to verify
if such a solution exists , the algorithm builds a bipartite graph @xmath176 , as described in this section .
the intuitive idea behind the graph is that edges of the graph correspond to possible ways of assigning each row in @xmath37 to a resolution vector @xmath178 .
rows assigned to the same resolution vector @xmath178 are clustered in the solution @xmath173 .
the construction of the vertex set of the graph is based on a a partition of @xmath37 into two disjoint sets called @xmath179 and @xmath180 ( that is @xmath181 ) .
the set @xmath179 consists of those rows @xmath160 belonging to the group @xmath64 such that : @xmath182 , that is @xmath4 belongs to a group of at least @xmath0 identical rows , and there exists a row @xmath183 , such that @xmath184 and @xmath65 are the same vector .
notice that only rows in @xmath185 might have no suppressed entry in a solution @xmath173 .
the vertex set of @xmath176@xmath186 has @xmath187 sets .
two sets ( @xmath188 , @xmath189 ) consist of vertices associated with the rows in @xmath180 , three sets ( @xmath190 , @xmath191 , @xmath192 ) consist of vertices associated with the rows in @xmath179 , and a final set called @xmath193 consists of vertices associated with the rows in @xmath62 . in the latter case notice that for each row @xmath194 in @xmath62 there exist @xmath0 vertices in @xmath193 to ensure that the cluster associated with @xmath194 has size at least @xmath0 .
the vertex set is defined as follows : * for each row @xmath195 , there is a corresponding vertex @xmath196 in @xmath188 and a corresponding vertex @xmath197 in @xmath189 ; * for each group @xmath64 consisting of the set of rows @xmath198 , where each @xmath199 , @xmath200 , there are @xmath0 corresponding vertices in @xmath190 , ( such vertices are denoted by @xmath201 ) , @xmath202 corresponding vertices in @xmath191 ( such vertices are denoted by @xmath203 @xmath204 @xmath205 ) , and @xmath202 corresponding vertices in @xmath192 ( such vertices are denoted by @xmath206 @xmath204 @xmath207 ) ; * for each row @xmath178 , there are @xmath0 corresponding vertices in @xmath193 ( such vertices are denoted by @xmath208 ) .
notice that our graph @xmath176is edge - weighted .
let @xmath172 be the weight function assigning a positive weight to each edge of @xmath176 .
given the set of edges @xmath209 , we denote by @xmath210 .
first , notice that the set @xmath62 consists of two disjoint sets : the set @xmath211consists of those rows in @xmath62 that have no suppressions , while @xmath212 .
each edge connects a vertex of @xmath213 with a vertex of @xmath214 , hence the graph @xmath176is bipartite .
the set @xmath62 consists of two disjoint sets : the set @xmath211consists of those rows in @xmath62 that have no suppressions , while @xmath212 .
intuitevely , each edge represents a possible assignment of a row in @xmath37 to a resolution vector in @xmath62 .
[ alg : matching ] now we are ready to define formally the set of edges @xmath86 of @xmath176and the weight function @xmath172 .
there are five possible kinds of edges . 1 .
let @xmath4 be a row of @xmath180 , and let @xmath194 be a row in @xmath215 .
then there is an edge @xmath216 , for each @xmath217 , with weight @xmath218 .
2 . let @xmath4 be a row in @xmath180 .
then there is an edge @xmath219 with weight @xmath220 .
3 . let @xmath64 be a group consisting of rows @xmath221 , where @xmath61 , for each @xmath222 with @xmath200 , is a row of @xmath179 ; let @xmath116 be the resolution vector of @xmath223 identical to @xmath65 .
then there is an edge @xmath224 , for each @xmath222 with @xmath225 .
all edges @xmath226 have weight @xmath227 .
4 . let @xmath64 be a group consisting of rows @xmath221 , where @xmath61 , for each @xmath222 with @xmath200 , is a row of @xmath179 ; let @xmath194 be a row in @xmath228 .
then there is an edge @xmath229 , for each @xmath222 with @xmath230 and for each @xmath30 with @xmath217 .
all edges @xmath231 have weight @xmath232 .
5 . let @xmath64 be a group consisting of rows @xmath221 , where @xmath61 , @xmath200 , is a row of @xmath179 .
then there is an edge @xmath233 for each @xmath222 with @xmath234 .
all edges @xmath226 have weight @xmath235 . in this section
we prove in lemma [ lem : feas : solution ] that @xmath236 is a clustering of the rows in @xmath37 that is a feasible solution for the @xmath148-approblem .
[ fig : ex - alg ] for an example . since @xmath176bipartite
, we can efficiently compute a maximum weight matching @xmath237 of @xmath176 @xcite .
given a matching @xmath237 of the graph @xmath176 , algorithm [ alg : matching ] computes in polynomial time a clustering @xmath236 of the rows in @xmath37 .
informally , the clustering is computed by assigning the rows in @xmath37 to the resolution vector in @xmath62 , using the edges in the matching @xmath237 .
[ fig : ex - alg ] clll + name&data&@xmath238&group + @xmath34 & aaa&@xmath1 & + @xmath35 & aaa&@xmath1 & + @xmath239 & aaa&@xmath1&[0pt]@xmath240 + @xmath241 & aaa&@xmath1 & + @xmath242 & aba&@xmath243&@xmath244 + @xmath245 & bbb&@xmath243&@xmath246 + @xmath247 & bbc&@xmath243&@xmath248 + cll + name&vectors&@xmath238 + @xmath249 & aaa&@xmath250 + @xmath251 & a*a&@xmath252 + @xmath253 & bb*&@xmath252 + = [ circle , fill = black , minimum size=6pt , inner sep=0pt ] /in 1/7 , 2/6 , 3/5 ( l- ) at ( 7,1.75 * ) [ label = below:@xmath254 ; /in 4/2 , 5/1 ( l- ) at ( 7,1.75 * ) [ label = below:@xmath255 ; /in 6/2 , 7/1 ( l- ) at ( 7,1.75 * ) [ label = below:@xmath256 ; /in 4/2 , 5/1 ( r- ) at ( 10,1.75 * ) [ label = below:@xmath257 ; /in 1/7 , 2/6 , 3/5 ( r- ) at ( 10,1.75 * ) [ label = below:@xmath258 ; ///in 1/1/s_3/1 , 2/2/s_3/2 , 3/3/s_2/1 , 4/4/s_2/2 , 5/5/s_1/1 , 6/6/s_1/2 ( s- ) at ( 0,1 + 1.75 * ) [ label = below:@xmath259 ; //in
1/1/2 , 2/2/2 , 3/3/2 , 4/4/3 ( l- ) node[midway , below ] ( r- ) ; /in 2/1 ( l- )
node[pos=0.63,below ] 20 ( s- ) ; /in 3/4 ( l- ) node[pos=0.83,below ] 20 ( s- ) ; /in 4/3 ( l- ) node[pos=0.7,below ] 20 ( s- ) ; /in 5/3 ( l- ) node[pos=0.3,below ] 20 ( s- ) ; /in 1/2 , 5/4 ( l- ) node[pos=0.63,above ] 20 ( s- ) ; /in 5/5 ( l- )
node[midway , below ] 3 ( r- ) ; //in 1/1/20 , 2/2/20 , 3/3/20 , 4/4/20 , 6/5/21 , 7/6/21 ( l- )
node[pos=0.36,below ] ( s- ) ; notice that , each vertex @xmath260 has only the edge @xmath261 on it , hence we can always add those edges to any matching and the corresponding solution @xmath236 of @xmath148-ap ] .
let @xmath237 be a matching of @xmath176and let @xmath262 be a vertex of @xmath176 , then we say that @xmath262 is _ covered _ by a matching @xmath237 if there exists an edge of @xmath237 for which @xmath262 is one of its endpoints .
moreover , we will say that @xmath237 is _ feasible _ if all vertices in @xmath193 are covered by @xmath237 .
when a matching @xmath237 covers all vertices in @xmath263 and is feasible , it is defined as a _
complete matching_. let @xmath17 be a clustering of an instance @xmath37 of the @xmath148-approblem .
then @xmath17 is _ feasible _ if and only if each set of the partition @xmath17 contains at least @xmath0 rows .
the next part of this section is devoted to show that every maximum weight matching @xmath237 is complete and that clustering @xmath236 is _
feasible_. first , we will show in the next two lemmata that , given @xmath264 , @xmath265 is a threshold that distinguishes between matchings that are feasible and those that are not .
[ lem : eq - matching - weight - t ] let @xmath237 be a matching of @xmath176 , let @xmath266 be the subset of @xmath193 consisting of the vertices of @xmath193 that are covered by @xmath237 , and let @xmath267 be the subset of the edges of @xmath237 that have one endpoint in @xmath266 .
then the total weight of the edges in @xmath267 is exactly @xmath268 .
it is an immediate consequence of the observation that all edges where an endpoint is @xmath269 have the same weight @xmath270 , with @xmath271 .
[ lem : eq - matching ] let @xmath237 be a matching of @xmath176and let @xmath267 be the subset of the edges of @xmath237 that have one endpoint in @xmath193 .
then the total weight of the edges in @xmath267 is at least @xmath272 if and only if @xmath237 is feasible .
let @xmath267 be the subset of the edges of @xmath237 that have one endpoint in @xmath193 , and let @xmath273 be the total weight of edges in @xmath267 .
an immediate consequence of lemma [ lem : eq - matching - weight - t ] is that @xmath274 if and only if @xmath267 is feasible .
assume now that @xmath237 is not feasible , then there exists at least one vertex @xmath275 that is not covered by @xmath237 .
again , a consequence of lemma [ lem : eq - matching - weight - t ] is that @xmath276 .
let @xmath277 be the set @xmath278 . by construction , @xmath279 and @xmath280 is an upper bound on the total weight of @xmath277 , therefore @xmath281 , completing the proof .
using lemmata [ lem : eq - matching - weight - t ] and [ lem : eq - matching ] , we can prove lemma [ lem : feas : solution ] .
[ lem : feas : solution ] let @xmath237 be a maximum weight matching of @xmath176 , then @xmath237 is complete and the solution @xmath236 computed by algorithm [ alg : matching ] is feasible .
this section is devoted to prove that , starting from a maximum weight matching @xmath237 , algorithm [ alg : matching ] computes an optimal solution @xmath236 of @xmath148-ap . in order to prove that any maximum weight matching @xmath237 of the graph @xmath176leads to an optimal solution of @xmath148-apover instance @xmath37
, we are going to prove that @xmath282 if and only if @xmath148-apover instance @xmath37 admits a solution with cost not greater than @xmath22 , and such solution is computed by applying algorithm [ alg : matching ] .
such result will be obtained through a sequence of technical lemmata .
since @xmath237 is a maximum weighted matching , we can assume by lemma [ lem : feas : solution ] that @xmath237 is complete .
given a complete matching @xmath237 , we denote by @xmath283 the set of edges of @xmath237 with one endpoint in @xmath284 and one endpoint in @xmath193 , while we denote by @xmath285 the set of those edges of @xmath237 that have one endpoint in @xmath286 and one endpoint in @xmath287 .
furthermore , let us denote by @xmath288 the set of vertices of @xmath284 that are endpoints of an edge in @xmath283 and by @xmath289 the set of vertices of @xmath286 that are endpoints of an edge in @xmath285 .
notice that by definition of @xmath289 and , by definition of complete matching , @xmath290 .
finally , let us denote by @xmath291 the set of rows in @xmath37 associated with the vertices in @xmath289 .
lemma [ lem : cost : match : part : feas ] shows how the weight of a complete matching @xmath237 is related to the edge weights of @xmath176 .
[ lem : cost : match : part : feas ] let @xmath237 be a complete matching of @xmath176 , and let @xmath292 be the total weight of @xmath237
. then @xmath293 . in the next two lemmata
, we will show that : ( i ) given an instance @xmath37 of @xmath148-ap , if there exists a solution of @xmath148-apover @xmath37 that suppresses at most @xmath22 entries then the graph @xmath176associated with @xmath37 admits a complete matching of @xmath176with total weight @xmath294 ; ( ii ) given a complete matching of the graph @xmath176of total weight @xmath294 , algorithm [ alg : matching ] returns a solution @xmath236 of @xmath148-apthat suppresses at most @xmath22 entries .
these lemmata , coupled with lemma [ lem : feas : solution ] , prove the correctness of algorithm [ alg : matching ] in theorem [ theo : correctness - algorithm ] .
[ lem : from - partially - feasible - solution - to - matching ] let @xmath37 be an instance of @xmath148-ap , let @xmath173 be a feasible solution of @xmath148-ap over instance @xmath37 that suppresses at most @xmath22 entries , let @xmath176be the graph associated with @xmath37 and @xmath62 .
then there exists a complete matching of @xmath176with total weight @xmath294 .
[ lem : feasible_matching - solution ] let @xmath37 be an instance of @xmath148-ap , let @xmath176be the graph associated with @xmath37 , and let @xmath237 be a complete matching of @xmath176of weight @xmath295 . then , starting from the matching @xmath237 of @xmath176 , algorithm [ alg : matching ] computes a feasible solution @xmath236 of @xmath148-apover instance @xmath37 , where there are at most @xmath22 suppressions .
since @xmath237 is complete , for each vertex @xmath296 of @xmath193 , with @xmath217 , there exists an edge @xmath297 for some @xmath298
. then algorithm [ alg : matching ] defines a solution @xmath236 for @xmath148-apassigning , for each edge @xmath299 , the row @xmath4 corresponding to vertex @xmath262 to the set that has resolution vector @xmath194 .
more precisely , row @xmath4 is defined by algorithm [ alg : matching ] as the @xmath30-th element of the set that has resolution vector @xmath194 .
therefore each set associated with a resolution row @xmath178 will consist of at least @xmath0 rows compatible with @xmath194 .
hence @xmath236 is a feasible solution .
recall that @xmath237 has a total weight of at least @xmath300 .
we will prove that @xmath236 induces at most @xmath22 suppressions . by lemma [ lem :
cost : match : part : feas ] , @xmath301 where @xmath302 .
notice that , by definition of @xmath236 , each vertex of @xmath288 corresponds to a row in @xmath37 assigned to a set with a resolution vector in @xmath62 .
such rows associated with @xmath288 induce a cost in @xmath236 of @xmath303 .
furthermore , the vertices of @xmath289 corresponds to rows of @xmath37 inducing a cost of at most @xmath304 .
therefore @xmath236 induces @xmath302 suppressions .
[ theo : correctness - algorithm ] let @xmath37 be an instance of @xmath148-ap . then algorithm [ alg : solving - ap ] returns a solution @xmath236 of cost at most @xmath22 if and only if such a solution exists . by lemma [ lem : feas : solution ] , @xmath236 is feasible . hence if @xmath236 suppresses at most @xmath22 entries
, then @xmath148-apadmits a solution of cost at most @xmath22 . on the other hand , by lemma [ lem :
from - partially - feasible - solution - to - matching ] , if there exists a solution @xmath305 of @xmath37 that suppresses at most @xmath22 entries , then there exists a feasible matching @xmath237 with weight @xmath294 . then , by lemma [ lem : feasible_matching - solution ] , algorithm [ alg : solving - ap ] returns a solution @xmath236 of @xmath148-ap that suppresses at most @xmath22 entries . if @xmath148-apadmits a solution that suppresses at most @xmath22 entries , then there exists a set @xmath306 of resolution vectors such that @xmath307 is a solution for @xmath148-apwith resolution vectors @xmath306 with the property that @xmath307 suppresses at most @xmath22 entries .
now , there exist @xmath308 possible sets of resolution vectors and the construction of graph @xmath176requires @xmath309 .
a maximum matching @xmath237 of a bipartite graph can be computed in polynomial time @xcite and starting from @xmath237 , we can compute a solution of the @xmath148-apin time @xmath310 .
hence the overall time complexity of the algorithm is @xmath311 .
in this section we investigate the computational and approximation complexity of @xmath1-ap(@xmath1 ) , that is @xmath0-ap when each row consists of exactly @xmath1 columns and @xmath2 .
we show that @xmath1-ap(@xmath1 ) is apx - hard via an l - reduction from minimum vertex cover on cubic graphs ( mvcc ) , which is known to be apx - hard @xcite .
due to page limit , we only sketch the proof .
the mvcc problem , given a cubic graph @xmath23 , asks for a smallest @xmath312 such that each edge of @xmath76 has at least one of its endpoints in @xmath313 .
let @xmath314 be instance of mvcc , where @xmath315 and @xmath316 .
the reduction builds an instance @xmath37 of @xmath1-ap(@xmath1 ) associating with each vertex @xmath317 a set @xmath318 consisting of @xmath319 rows , and with each edge @xmath320 a set @xmath321 consisting of @xmath322 rows .
finally , a set @xmath266 of @xmath1 more rows is added to @xmath37 .
now we can describe formally our reduction .
let @xmath318 be the set of rows associated with vertex @xmath317 .
the rows in @xmath318 have values over an alphabet @xmath323 .
the set @xmath318 consists of @xmath319 rows belonging to @xmath187 groups , denoted by @xmath324 , of identical rows .
the representative rows of groups @xmath324 , and the cardinality of the groups , are defined as follows : - @xmath325 , with @xmath326 ; each group @xmath327 , with @xmath326 , consists of exactly two rows ; - @xmath328 , with @xmath326 ; each group @xmath329 , with @xmath326 , consists of exactly one row
. notice that given two rows @xmath330 belonging to different groups of @xmath318 , @xmath331 iff @xmath332 , @xmath333 ( or the converse ) or @xmath334 . given a group @xmath327 , with @xmath326 , each symbol @xmath335
is called the _ private _ symbol of @xmath327 .
the groups of rows @xmath336 , with @xmath337 , are denoted as the _ docking groups _ of @xmath318 , and each of them is associated with a set @xmath338 of rows encoding an edge @xmath339 of @xmath76 .
more precisely , given the set of rows @xmath321 , we denote by @xmath340 the docking group of @xmath318 associated with set @xmath321 . now , let us build the set @xmath321 of rows associated with an edge @xmath136 .
let @xmath340 and @xmath341 be the two docking groups of @xmath318 and @xmath342 respectively , associated with the set @xmath321 .
let @xmath343 and @xmath344 be the private symbols of groups @xmath340 and @xmath341 respectively .
the set @xmath321 consists of @xmath322 rows distributed in @xmath187 groups .
the rows of @xmath321 have values over alphabet @xmath345 .
let us define the representative rows and the cardinality of the groups in @xmath321 : - @xmath346 ; group @xmath347 consists of a single row ; - @xmath348 ; group @xmath349 consists of two rows ; - @xmath350 ; group @xmath351 consists of a single row ; - @xmath352 , with @xmath353 ; each group @xmath354 , with @xmath353 , consists of a single row .
the group of @xmath321 that has two occurrences of symbol @xmath343 shared with @xmath340 is called the @xmath222-group of set @xmath321 , and is denoted as @xmath355 .
notice that , given two rows @xmath330 of @xmath318 , @xmath321 respectively , then @xmath331 iff @xmath356 and @xmath357 . finally , a set @xmath266 of @xmath1 rows @xmath358 are added to @xmath37
. the rows in @xmath266 have values over an alphabet @xmath359 disjoint from any other set @xmath360 , @xmath361 .
each row @xmath362 , and it has hamming distance @xmath1 from any other row of @xmath37 .
therefore for any set @xmath313 containing some rows @xmath363 , all positions of a row in @xmath313 will be suppressed .
now , consider the set @xmath318 .
the following lemma gives a lower bound on the cost of an optimal solution of @xmath1-ap(@xmath1 ) over instance @xmath318 .
[ lem : opt - r_i ] let @xmath318 be a set of rows , then an optimal solution of @xmath1-ap(@xmath1 ) over instance @xmath318 has a cost of at least @xmath319 .
the main idea of the reduction is showing that we can consider a set of solutions , called _
canonical solutions _ , that is solutions where : ( i ) @xmath17 contains exactly one cluster @xmath266 containing only suppressed entries ; ( ii ) each set @xmath318 is associated with either a _ type a _ or a _ type b _ solution ( to be defined later ) , eventually with the contribution of some rows in the sets @xmath321 for a _ type b _ solution ; \(iii ) two sets @xmath318 , @xmath342 are associated with a _ type b _ solution only if there is no edge set @xmath321 in the instance @xmath37 , that is the corresponding vertices @xmath364 , @xmath365 are not adjacent in @xmath76 ; \(iv ) either an edge set is part of a _ type b _ solution of some set @xmath318 and has a total cost of @xmath366 or it has a total cost of @xmath367 .
notice that , by construction , in a canonical solution , rows @xmath368 .
let us define the notions of _ type a _ and _ type b _ solution . given a set @xmath318 and the edge sets @xmath321 , @xmath338 , @xmath369 , a _
type a _ solution for @xmath318 consists of three sets @xmath370 , where @xmath371 , while a _ type b _ solution consists of the following sets : ( i ) three sets @xmath372 , @xmath373 , @xmath374 ; ( ii ) @xmath375 .
lemma [ lem : compute - canonical ] is the main technical contribution of this section .
[ lem : compute - canonical ] let @xmath17 be a solution of @xmath1-ap(@xmath1 ) over instance @xmath37
. then we can compute in polynomial time a canonical solution @xmath305 of @xmath1-ap(@xmath1 ) over instance @xmath37 such that @xmath376 . _ sketch of the proof .
_ by direct inspection , it is immediate to notice that _ type a _ and _ type b _ solutions induce @xmath319 suppression in rows of @xmath318 hence , by lemma [ lem : opt - r_i ] , they are optimal for @xmath318 .
the next step is computing in polynomial time a solution @xmath377 such that each set @xmath318 is associated in @xmath377 only with either a _ type a _ or _ type b _ solution , and such that @xmath378 .
such step is obtained by exploiting the optimality of _ type a _ and _ type b _
solutions for @xmath318 , and some properties of the instance @xmath37 .
then , starting from such solution @xmath377 , we can compute in polynomial time a canonical solution @xmath305 such that @xmath379 .
the main idea to prove this result is that for any two sets @xmath318 , @xmath342 , such that both @xmath318 and @xmath342 are associated with a _ type b _ solution in @xmath377 and @xmath321 is part of the instance @xmath37 , then we can improve the solution by imposing a _ type a _ solution for @xmath318 . a consequence of lemmata [ lem : opt - r_i ] and [ lem : compute - canonical ] and some properties of the instance @xmath37 , is lemma [ lem : apx - hard1 ] .
[ lem : apx - hard1 ] let @xmath17 be a solution of @xmath1-ap(@xmath1 ) over instance @xmath37 of cost @xmath380 , then we can compute in polynomial time a solution of mvcc over instance @xmath76 of size @xmath313 .
let us consider a canonical solution of @xmath1-ap(@xmath1 ) .
first , notice that the three rows @xmath381 , @xmath382 , @xmath383 provide together a cost of @xmath319 . since two sets of rows are associated with a _ type b
_ solution only if there does not exist a set @xmath321 , on the contrary , given an edge set @xmath321 at least one of the set @xmath318 and @xmath342 is associated with a _ type a _ solution .
consequently , the set of rows associated with a _ type a _ solution corresponds to a vertex cover of the graph @xmath76 .
now consider the cost of a canonical solution .
for each set @xmath318 of rows associated with a _ type b _ solution , we can show that each of the three edge sets @xmath321 , @xmath338 , @xmath369 has a cost of @xmath366 . notice that , given an edge set @xmath321 , if both sets @xmath318 , @xmath342 are associated with _ type a _ solutions , then we can show that the edge set @xmath321 has a cost of @xmath367 . accounting this decreasing of the cost of the edge sets to the set @xmath318 of rows with a _ type b _ solution , is equivalent to assign to a _ type b
_ solution a cost equal to @xmath187 , while a _ type a _ solution has a cost equal to @xmath319 .
similarly to lemma [ lem : apx - hard1 ] , we can prove that starting from a solution @xmath313 of mvcc over instance @xmath76 , we can compute in polynomial time a solution @xmath17 of @xmath1-ap(@xmath1 ) over instance @xmath37 of cost @xmath380 .
therefore @xmath1-ap(@xmath1 ) is apx - hard .
10 g. aggarwal , t. feder , k. kenthapadi , s. khuller , r. panigrahy , d. thomas , and a. zhu .
achieving anonymity via clustering . in _ pods _ ,
pages 153162 .
g. aggarwal , t. feder , k. kenthapadi , r. motwani , r. panigrahy , d. thomas , and a. zhu .
anonymizing tables . in _ icdt _ , lncs 3363 , pages 246258 .
g. aggarwal , k. kenthapadi , r. motwani , r. panigrahy , d. thomas , and a. zhu .
approximation algorithms for k - anonymity . , 2005 .
alimonti and v. kann . some apx - completeness results for cubic graphs .
, 237(12):123134 , 2000 .
g. ausiello , p. crescenzi , v. gambosi , g. kann , a. marchetti - spaccamela , and m. protasi . .
springer - verlag , 1999 .
p. bonizzoni , g. della vedova , and r. dondi .
the k - anonymity problem is hard . in _
fct _ , lncs 5699 , pages 2637 .
r. chaytor , p. a. evans , and t. wareham .
fixed - parameter tractability of anonymizing data by suppressing entries . , 18(4 ) : 362 - 375 , 2009 . r. downey and m. fellows . .
springer verlag , 1999 .
r. g. downey and m. r. fellows .
fixed - parameter tractability and completeness ii : on completeness for @xmath384}$ ] .
, 141(1&2):109131 , 1995 .
w. du , d. eppstein , m. t. goodrich , and g. s. lueker . on the approximability of geometric and geographic generalization and the min - max bin covering problem . in _ wads _ , lncs 5664 , pages 242253 . 2009 .
a. gionis and t. tassa .
@xmath0-anonymization with minimal loss of information . , 21(2 ) : 206 - 219 , 2009 . a. meyerson , r. williams .
on the complexity of optimal k - anonymity . in _ pods _ , pages 223228 .
r. niedermeier . .
oxford university press , 2006 .
h. park and k. shim .
approximate algorithms for k - anonymity . in _
sigmod _ , pages 6778
. 2007 .
p. samarati .
protecting respondents identities in microdata release . , 13:10101027 , 2001 .
p. samarati and l. sweeney . generalizing data to provide anonymity when disclosing information ( abstract ) . in _ pods _ ,
page 188 . 1998 .
j. schwartz , a. steger and a. weissl , fast algorithms for weighted bipartite matching . in _ wea _ ,
pages 476487 , 2005 .
l. sweeney .
k - anonymity : a model for protecting privacy .
, 10(5):557570 , 2002 .
[ appendix - lem : w[1]-hard : prop - base ] let @xmath37 be the instance of @xmath69-ap associated with @xmath76 and consider two rows @xmath101 of @xmath37 , such that @xmath102 and @xmath103 .
then , @xmath104 \neq r_x(i , j)[t]$ ] , for each @xmath105 .
[ appendix - lem : w[1]-hard-1 ] let @xmath23 be a graph , let @xmath25 be a @xmath75-clique of @xmath76 and let @xmath37 be the instance of @xmath69-ap associated with @xmath76
. then we can compute in polynomial time a solution @xmath17 of @xmath69-ap over instance @xmath37 with cost at most @xmath106 .
initially let @xmath305 be a solution consisting of clusters @xmath98 , @xmath87 , for each @xmath130 . for each @xmath386 ) , let @xmath387 be the first row of @xmath87 .
compute a new solution @xmath17 consisting of clusters @xmath114 , @xmath121 , for each @xmath130 , where : notice that , since @xmath25 is a @xmath75-clique , @xmath132 .
moreover , by construction , @xmath394 , therefore @xmath17 is a feasible solution for @xmath37 .
notice also that no entries is suppressed in the rows of each set @xmath121 , therefore to determine the cost of @xmath305 it suffices to determine the number of entries deleted in @xmath114 , and we will show that such number is exactly @xmath106 . indeed , by construction , for each column @xmath120 of the first @xmath112 columns , and for each row @xmath102 and @xmath103 , @xmath104 \neq r_x(i , j)[t]$ ] , hence all the entries of the first @xmath112 columns of the rows in @xmath114 must be deleted , resulting in @xmath395 suppressions .
now let us consider the columns with index @xmath396 and @xmath397 .
in such positions , all rows of @xmath98 are equal to @xmath139 , while all rows in the sets @xmath398 , @xmath399 are equal to @xmath141 .
consider the @xmath400 of @xmath401 . as
the corresponding edges are incident on a set of @xmath75 vertex , by construction there exists a set @xmath402 of exactly @xmath75 columns , with @xmath403 , where at least one of the rows in @xmath140 is equal to @xmath141 , while the rows in @xmath98 are all equal to @xmath139 .
since in any other column all rows in @xmath114 have value equal to @xmath139 , hence there are additional @xmath404 suppressions for the columns with index @xmath405 .
overall , the number of suppressions is @xmath406 which , by the choice of @xmath0 is equal to @xmath106 .
first notice that , as @xmath237 is feasible , each vertex of @xmath193 is covered and each vertex @xmath407 , with @xmath217 , is covered by @xmath237 .
assume that @xmath237 is not complete and that a vertex @xmath408 of @xmath409 ( resp .
@xmath410 , with @xmath411 , of @xmath412 ) is not matched .
then , by construction , also the vertex @xmath413 of @xmath414 ( resp .
@xmath415 of @xmath416 ) is not covered by @xmath237 , as @xmath408 ( resp .
@xmath410 ) is the only vertex adjacent to @xmath413 ( resp .
@xmath415 ) in @xmath176 .
hence we can compute the matching @xmath417 by adding all the edges of @xmath237 to @xmath417 and by adding edges @xmath418 , ( resp .
@xmath419 ) for each vertex @xmath408 ( resp .
@xmath410 ) not covered by @xmath237 .
since @xmath236 feasible , all vertices in @xmath193 are covered by @xmath237 .
furthermore , we can assume , by lemma [ lem : prop : match : part : feas ] , that each vertex in @xmath286 is covered by @xmath237 .
hence each row in @xmath37 is assigned by algorithm [ alg : matching ] to a set whose resolution vector is @xmath62 . furthermore algorithm [ alg : matching ] assigns to each set with resolution vector @xmath178 at least @xmath0 rows .
hence the clustering @xmath236 computed by algorithm [ alg : matching ] is feasible .
the total weight @xmath292 of the matching @xmath237 is defined as @xmath422 by lemma [ lem : eq - matching ] and by definition of the weight function @xmath172 , it follows that @xmath423 and by definition of @xmath424 it holds @xmath425 hence @xmath426 by definition of feasible matching and by lemma [ lem : prop : match : part : feas ] , @xmath427 .
furthermore , since @xmath428 , then @xmath429 . by construction @xmath430 and @xmath431 .
hence @xmath432 [ appendix - lem : from - partially - feasible - solution - to - matching ] let @xmath37 be an instance of @xmath148-ap , let @xmath173 be a feasible solution of @xmath148-ap over instance @xmath37 that suppresses at most @xmath22 entries , let @xmath176be the graph associated with @xmath37 and @xmath62 .
then there exists a complete matching of @xmath176with total weight @xmath294 .
since @xmath173 is feasible , we notice that each set of @xmath173 associated with a resolution vector @xmath433 must have cardinality at least @xmath0 .
furthermore , we assume that all the sets of @xmath173 are all associated with different resolution vectors , otherwise we can merge all the sets with the same resolution vector without increasing the cost of @xmath173 .
let @xmath194 be a row of @xmath62 and denote by @xmath434 the set of rows of @xmath37 assigned to the set associated with resolution vector @xmath194 .
starting from @xmath173 we compute incrementally a matching @xmath237 by adding edges .
first , for each set of vertices @xmath435 , @xmath436 , let @xmath437 be the minimum number such that @xmath438 does not have any edge incident on it in @xmath237 .
first , assume that @xmath439 ; add the edge @xmath440 to @xmath237 , for each @xmath436 .
now , assume that @xmath441 .
scan the rows in @xmath442 and for each row @xmath4 in @xmath442 , if @xmath443 add the edge @xmath444 to @xmath237 .
if @xmath445 and belongs to group @xmath64 add the edge @xmath446 to @xmath237 .
if no such @xmath438 exists , then no edge is added to @xmath237 .
notice that by construction , since all sets in @xmath62 have at least @xmath0 rows , then all vertices of @xmath193 are covered by @xmath237 , therefore @xmath237 is feasible . finally add to @xmath237 all edges @xmath447 , @xmath448 , @xmath234 , for each vertex in @xmath449 , @xmath450 respectively that is not already covered in @xmath237 .
hence @xmath237 is complete . given a solution @xmath173 , a resolution vector @xmath194 of @xmath62 and the corresponding matching @xmath237 , consider the order in which the rows of a set @xmath442 are scanned sequentially to construct @xmath237 .
each of the first @xmath0 rows assigned to a cluster with resolution vector equal to @xmath194 , by construction corresponds to an edge of @xmath237 joining a vertex of @xmath288 and a vertex of @xmath193 . since @xmath237 is complete
, those rows have a total cost in @xmath173 of @xmath451 .
the remaining rows of @xmath37 correspond to vertices of @xmath289 .
notice that those rows have a total cost in @xmath173 not larger than @xmath304 . by lemma [ lem :
cost : match : part : feas ] @xmath452 .
since @xmath173 suppresses at most @xmath22 entries of @xmath37 , then @xmath453 , therefore @xmath295 .
[ prop : dist - edge - rows ] let @xmath454 , @xmath455 be two rows of @xmath321 , with @xmath468 and @xmath469 , with @xmath470 .
let @xmath459 , @xmath471 be two rows of @xmath318 and @xmath472 , with @xmath473 , and let @xmath474 be a row of @xmath475 , with @xmath476 or @xmath477 . then : * @xmath478 ; * @xmath463 iff @xmath479 and @xmath480 , with @xmath481 ; * @xmath482 iff @xmath459 is in the docking group @xmath340 of @xmath318 and @xmath454 is in the group @xmath355 ; * @xmath483 only if @xmath459 is in a group adjacent to @xmath340 ; * @xmath484 ; * @xmath485 . in what follows , by an abuse of notation , we may use a group @xmath486 to denote its representative row @xmath487 . fig .
[ fig : graph - dist-1 ] shows the groups of @xmath318 , @xmath342 , @xmath321 .
each group of identical rows si represented with a vertex , while an edge joins two vertices iff the corresponding groups are at hamming distance @xmath141 .
let us consider the set of @xmath319 rows , distributed in @xmath187 groups , of @xmath318 .
as none of the group of @xmath318 consists of at least @xmath1 rows , it follows that any solution of @xmath1-ap(@xmath1 ) suppresses at least one entry in each row of @xmath318 , hence the lemma follows . [ lem : cost - ty_b ] let @xmath18 be a _ type b _ solution of @xmath1-ap(@xmath1 ) over instance @xmath489 , then @xmath18 suppresses @xmath319 entries in the rows of @xmath318 , and @xmath141 entry for each row of @xmath355 , @xmath490 , @xmath491 . by construction each set of a _ type b _ solution containing a docking group of @xmath318 consists of three rows , where exactly one position is suppressed for each row by prop .
[ prop : dist - edge - rows ] .
the cluster @xmath375 consists of @xmath1 rows , where exactly one position is suppressed for each row . by a simple counting argument
, the rows of @xmath318 have a total cost of @xmath319 .
let @xmath17 be a solution of @xmath1-ap(@xmath1 ) over instance @xmath37 and let @xmath321 be an edge set . then we say that @xmath17 induce an _
i - normal _ solution for @xmath321 if it contains the following three sets : ( i ) one set clusters @xmath492 , ( ii ) one set containing @xmath493 , such that exactly two entries ( in columns @xmath141 and @xmath1 ) are suppressed in the rows of @xmath494 , ( iii ) set @xmath495 .
[ lem : cost - edge-10 ] let @xmath17 be a solution of @xmath1-ap(@xmath1 ) over instance @xmath37 and let @xmath321 be an edge set , then @xmath17 suppresses at most @xmath366 entries in the rows of @xmath321 only if it induces an i - normal or j - normal solution for @xmath321 .
first assume that @xmath17 induces an i - normal solution for @xmath321 . by prop .
[ prop : dist - edge - rows ] , it follows that one entry of @xmath355 is suppressed . in the set consisting of the rows in @xmath354 , with @xmath496 , by prop .
[ prop : dist - edge - rows ] two positions for each row are suppressed . finally , by prop . [
prop : dist - edge - rows ] , in the set @xmath497 , exactly one position is suppressed for each row .
now , let us prove that if @xmath17 is a solution that is not i - normal or j - normal for @xmath321 , then @xmath498 .
notice that that each row in @xmath354 , with @xmath496 , has a hamming distance @xmath243 from any other row of @xmath499 , hence at least two entries are suppressed in each solution @xmath17 .
furthermore , notice that each of the four rows in the groups @xmath347 , @xmath349 , @xmath351 must have a cost of at most @xmath141 .
but then , the rows of @xmath349 must be co - clustered with the row of exactly one of @xmath347 , @xmath351 ( w.l.o.g .
@xmath347 ) .
but then @xmath500 , must be co - clustered with @xmath501 .
[ lem : set - x ] let @xmath18 be a solution of @xmath1-ap(@xmath1 ) over instance @xmath37 , then we can compute in polynomial time a solution @xmath62 such that @xmath502 and @xmath62 contains at most one set suppressing three entries for each row .
assume that solution @xmath18 contains sets @xmath503 , with @xmath504 , such that all the positions of the rows in @xmath505 are suppressed .
then we can compute in polynomial time a solution @xmath62 by merging the set @xmath503 in a single cluster @xmath11 . notice that @xmath502 , as in both solution @xmath62 and @xmath18 three positions are suppressed for each row @xmath506 . 1 . for each edge set @xmath321
, @xmath305 has a set @xmath507 containing the rows of groups @xmath508 , such that for each row in @xmath507 exactly two columns ( columns @xmath141 and @xmath1 ) are supprssed ; 2 .
@xmath376 .
first , notice that by prop .
[ prop : dist - edge - rows ] the rows in the groups @xmath509 have distance smaller than @xmath1 only w.r.t
. rows of @xmath321 . furthermore ,
notice that , by construction , each row in @xmath510 may be equal to another row of @xmath37 only in the second position .
assume that there exist clusters @xmath511 ( at most one of these clusters can be empty ) containing the rows of @xmath510 , such that at most two entries are suppressed for each row of the cluster @xmath512 , @xmath513 .
then , for each row in @xmath514 , the positions @xmath141 and @xmath1 are suppressed .
hence , we can merge clusters @xmath511 , without increasing the cost of the solution , obtaining one set that contains the rows @xmath508 .
assume that some rows of @xmath509 are in the cluster @xmath266 and some rows of @xmath509 are in a different cluster @xmath11 , such that at most two entries are suppressed for each row of the cluster @xmath11 .
then we can move the rows of @xmath515 to @xmath11 , decreasing the cost of the solution .
now , assume that these rows are all clustered in set @xmath266 .
it follows that each row of @xmath509 have a cost of @xmath1 .
hence , we can move this set of rows to a new set @xmath509 , decreasing the cost of the solution .
[ lem : cost - vert - adj - opt ] let @xmath318 , @xmath342 be two set of rows and let @xmath321 be an edge set of @xmath37 .
let @xmath17 be a solution of @xmath1-ap(@xmath1 ) over instance @xmath37 that associates a _ type b _ solution with both @xmath318 , @xmath342 .
then we can compute in polynomial time a solution @xmath305 of @xmath1-ap(@xmath1 ) over instance @xmath37 where exactly one of @xmath318 , @xmath342 is associated with a _ type b _ solution and such that @xmath376 .
notice that , by prop .
[ prop : dist - edge - rows ] , the rows of group @xmath349 have hamming distance @xmath141 only from the rows of @xmath347 and @xmath351 .
since in a _ type b _ solution the rows of @xmath347 and @xmath351 are co - clustered with rows of @xmath318 and @xmath342 , it follows by prop .
[ prop : dist - edge - rows ] that the rows of @xmath349 are co - clustered in @xmath17 with rows at hamming distance at least @xmath243 .
hence , @xmath17 suppresses two entries in each row of @xmath349 , and , as @xmath349 consists of @xmath243 rows , at least @xmath516 entries of rows in @xmath349 are suppressed in @xmath17 .
notice that , as @xmath318 and @xmath342 are associated with _ type b _ solutions in @xmath17 , the only rows that can be clustered with the rows of @xmath349 are those of groups @xmath517 , @xmath518 , @xmath519 .
starting from solution @xmath17 , let us compute a solution @xmath305 of @xmath1-ap(@xmath1 ) over instance @xmath37 as follows .
let @xmath321 , @xmath520 , @xmath521 be the edge sets associated with the three edges incident in @xmath364
. modify solution @xmath17 so that @xmath305 induces a _ type a _ solution for @xmath318 , and a j - normal solution for @xmath321 .
moreover , for each row of a group @xmath522 of an edge set @xmath521 , with @xmath477 , co - cluster such group @xmath523 with the cluster containing the rows of @xmath508 . by lemma
[ lem : cost - edge-10 ] the rows of edge set @xmath321 have a total cost of @xmath366 .
notice that by lemma [ lem : cost - edge - mand - set ] , we can assume that @xmath17 has a set @xmath313 containing @xmath508 , such that exactly two entries ( corresponding to the positions @xmath141 and @xmath1 ) are suppressed for each row in @xmath313 .
hence the representative row of @xmath313 has hamming distance @xmath243 from @xmath523 , as they are equal in position @xmath243 .
now , each of these two rows @xmath523 has a cost of @xmath243 in @xmath305 , while it has a cost of least @xmath141 in @xmath17 .
notice that each of the two rows of @xmath349 has a cost of at least @xmath243 in @xmath17 , while it has a cost of @xmath141 in @xmath305 .
hence , @xmath524 .
[ lem : apx - edge - set - cost ] let @xmath17 be a solution of the @xmath1-ap(@xmath1 ) over instance @xmath37 , such that two sets @xmath318 , @xmath342 are not associated with a _ type b _ solution in @xmath17 and @xmath17 induces a total cost of @xmath366 for the rows in @xmath321 .
then at least one of @xmath318 , @xmath342 has cost @xmath367 .
assume that @xmath17 induces a total cost of @xmath366 for the rows in @xmath321 .
notice that the rows in @xmath525 have a total cost of @xmath187 , as by prop .
[ prop : dist - edge - rows ] they are at hamming distance at least @xmath243 from any other row of @xmath37 .
furthermore , the @xmath516 rows of @xmath526 must have a cost of at least @xmath141 in @xmath17 .
notice that , by lemma [ lem : cost - edge-10 ] , @xmath17 induces either an i - normal or j - normal solution for @xmath321 ( w.l.o.g .
we assume that is i - normal ) .
hence @xmath355 is clustered with the rows of @xmath501 ) , while of @xmath527 is clustered with @xmath349 , otherwise some rows of @xmath349 are clustered in @xmath17 with a row at hamming distance at least @xmath243 , hence the total cost of the rows in @xmath321 is greater than @xmath366 .
but then , we claim that @xmath17 induces a cost of at least @xmath367 for the rows of the set @xmath318 .
now , recall that by hypothesis @xmath318 is not associated with a _ type b _ solution in @xmath17 , and let us consider the clusters containing rows of @xmath318 in @xmath17 .
notice that if at least two rows of @xmath318 are clustered with some rows at hamming distance at least @xmath243 , then @xmath17 induces a cost of at least @xmath367 for the set @xmath318 .
recall that group @xmath340 of @xmath318 is clustered only with rows of group @xmath355 of @xmath321 , and consider the cases that either the three groups of rows in @xmath528 are co - clustered , or not . in the former case ,
as @xmath318 is not associated with a _ type b _ solution , it follows that the rows of at least one of the docking group of @xmath318 are clustered with rows at hamming distance @xmath243 ; hence @xmath17 induces a cost of at least @xmath367 for the set @xmath318 . in the latter case ,
let us consider the group of @xmath318 ( w.l.o.g .
@xmath529 ) adjacent to @xmath340 and let @xmath313 be the cluster containing the unique row of @xmath529 .
as the rows in @xmath528 are not co - clustered , it follows that @xmath313 contains a row @xmath4 at hamming distance at least @xmath243 from @xmath529 .
if @xmath530 , then @xmath17 suppresses at least two entries of two rows of @xmath318 , namely @xmath4 and @xmath529 , hence @xmath17 induces a cost of at least @xmath367 in rows of @xmath318 .
if @xmath531 , then @xmath4 must be a row at hamming distance @xmath1 from @xmath529 . indeed by prop .
[ prop : dist - vertex - rows ] and by prop .
[ prop : dist - edge - rows ] , the rows at hamming distance not greater than @xmath243 from @xmath529 belong to @xmath532 .
we have assumed that @xmath533 , and it must be @xmath534 , since by hypothesis @xmath17 induces an i - normal solution for the rows in @xmath321 . hence @xmath529 must have cost equal to @xmath1 and must be part of the cluster @xmath266 in @xmath17 by the lemma [ lem : set - x ] .
it follows that @xmath17 induces a cost of at least @xmath367 for the set @xmath318 .
[ appendix - lem : compute - canonical ] let @xmath17 be a solution of @xmath1-ap(@xmath1 ) over instance @xmath37
. then we can compute in polynomial time a canonical solution @xmath305 of @xmath1-ap(@xmath1 ) over instance @xmath37 such that @xmath376 .
let us consider the solution @xmath17 . before computing a canonical solution @xmath305 , we compute an intermediate solution @xmath377 such that @xmath535 as follows . first , for each set @xmath318
, if @xmath318 is associated with a @xmath536 solution in @xmath17 , then define a @xmath536 solution for @xmath318 in @xmath377 .
otherwise , if each docking vertices @xmath501 of a set @xmath318 is clustered in @xmath17 with the row of group @xmath355 of @xmath321 , then define a _ type b _
solution for @xmath318 in @xmath377 ; else define a _
type a _ solution for @xmath318 in @xmath377 .
furthermore , define a set containing row @xmath358 in @xmath377 .
next , consider the rows of an edge set @xmath321 , and define a clustering of the rows not yet clustered in @xmath377 . if exactly one of @xmath318 , @xmath342 ( w.l.o.g .
@xmath318 ) is associated with a _ type b _ solution in @xmath377 , then define an i - normal solution for @xmath321 in @xmath377 .
else if at least one of @xmath318 , @xmath342 ( w.l.o.g .
@xmath318 ) is associated with a _ type a _ solution in @xmath377 , then define the following solution : one set contains the rows in @xmath537 ; one set contains the rows in @xmath538 .
if both @xmath318 , @xmath342 are associated with a _ type b _ solution in @xmath377 , then define a set @xmath539 in @xmath377 .
now , let us show that @xmath540 .
by lemma [ lem : opt - ty_a ] and by lemma [ lem : cost - ty_b ] it follows that for each row in a set @xmath318 the cost in @xmath377 is optimal .
furthermore , by lemma [ lem : set - x ] , we can assume that @xmath17 contains a set @xmath541 , hence the rows in @xmath542 have all cost @xmath1 in both @xmath17 and @xmath377 .
hence it remains to consider the cost of the edge set @xmath321 .
let @xmath321 be an edge set .
notice that by lemma [ lem : cost - edge - mand - set ] we can assume that @xmath17 contains a set @xmath543 , and by construction @xmath377 contains a set @xmath544 .
hence each row of @xmath354 , with @xmath353 , has a cost equal to @xmath243 in both @xmath17 , @xmath377 .
let us consider the case when both sets @xmath318 and @xmath342 are associated with a _ type b _ solution in both @xmath17 and @xmath377 .
the groups of @xmath321 not co - clustered in a _ type b _ solution of @xmath318 , @xmath342 , are @xmath349 , @xmath517 , @xmath518 , @xmath519 . by construction , as the rows at hamming distance @xmath141 from @xmath349 are clustered in the _ type b _ solution of @xmath318 , @xmath342 in @xmath17 ( hence can not be co - clustered with @xmath349 ) , it follows that the rows @xmath349 must be clustered with a row having hamming distance at least @xmath243 in @xmath17 .
as @xmath377 contains the set @xmath545 and as @xmath17 contains the set @xmath543 , it follows that the cost of the rows in @xmath321 in solution @xmath17 is greater or equal than the cost of the rows in @xmath321 in solution @xmath377 .
let us consider the case when exactly one of the sets @xmath318 and @xmath342 ( w.l.o.g .
@xmath342 ) is associated with a _ type b _ solution in @xmath377 and in @xmath17 . by construction , @xmath318 is associated with a _ type a _ solution in @xmath377 . by lemma [ lem : cost - edge-10 ]
it follows that that @xmath546 , and , as each row in @xmath547 has a cost of @xmath243 in @xmath377 , it follows that each row in @xmath548 has a cost of @xmath141 in @xmath377 . as @xmath17 contains the set @xmath507 , it follows that @xmath17 suppresses two entries in the rows of @xmath547 , hence the cost of the rows in @xmath321 in solution @xmath17 is greater or equal than the cost of the rows in @xmath321 in solution @xmath377 .
let us consider the case when at least one of the sets @xmath318 and @xmath342 ( w.l.o.g .
@xmath318 ) is associated with a _ type b _ solution in @xmath377 and not in @xmath17 .
notice that by construction , the rows in groups @xmath501 , @xmath549 are clustered in both @xmath17 and @xmath377 .
now , if @xmath17 induces a cost of at least @xmath367 for the rows in @xmath321 , since @xmath377 induces a cost of at most @xmath367 for the rows in @xmath321 it follows that @xmath550 .
if @xmath17 induces a cost of @xmath366 for the rows in @xmath321 , then by prop .
[ prop : dist - edge - rows ] @xmath349 must be co - clustered with @xmath527 . then
, it follows that by construction @xmath342 is associated with a _ type a _ solution in @xmath377 and that the rows of @xmath321 have a total cost of @xmath366 in @xmath377 . hence @xmath550 .
now , let us consider the case when both @xmath318 , @xmath342 , are associated with a _ type a _ solution in @xmath377 . in this case , by construction , the rows in the edge set @xmath321 have a total cost of @xmath367 in @xmath377 , while they have a cost of at least @xmath366 in @xmath17 , as the rows in @xmath525 ( contained in the set @xmath507 of @xmath17 ) have a total cost of @xmath187 , while each of the @xmath516 rows of @xmath526 has a cost of at least @xmath141 in @xmath17
assume that the rows of @xmath321 have a total cost of @xmath366 in @xmath17 .
by lemma [ lem : apx - edge - set - cost ] , @xmath17 induces a total cost of @xmath367 for the rows of one of the sets @xmath318 , @xmath342 ( w.l.o.g .
@xmath318 ) . notice that by lemma [ lem : opt - ty_a ] @xmath377 induces a cost of @xmath319 for all the set @xmath318 .
now , let us consider the set @xmath318 and the three edge sets @xmath321 , @xmath338 , @xmath488 . in what follows
, we will consider the cost induced by @xmath17 and by @xmath377 in the set @xmath318 and in some of the edge sets @xmath321 , @xmath338 , @xmath488 .
more precisely , for each edge set @xmath551 in @xmath552 , let us consider its cost together with the cost of @xmath318 only if @xmath553 and @xmath554 are clustered in @xmath17 ( otherwise @xmath551 will be eventually be considered together with @xmath442 ) . by construction the cost of at most two edge sets in @xmath321 , @xmath338 , @xmath488 ( assume w.l.o.g .
@xmath321 , @xmath338 ) are considered together with the cost of @xmath318 , otherwise @xmath553 and @xmath554 would be co - custered in @xmath17 , for each @xmath555 and by construction @xmath318 would be associated with a _ type b _ solution in @xmath377 . since @xmath556 , @xmath557 , @xmath558 , while @xmath559 , @xmath560 and @xmath561 it follows that @xmath562 .
now , we have shown that @xmath540 .
notice that @xmath377 may not be a canonical solution , as there may exist two sets @xmath318 , @xmath342 , with @xmath321 part of the instance , associated with a _ type b _ solution in @xmath377 .
now , applying lemma [ lem : cost - vert - adj - opt ] for each pair of sets @xmath318 , @xmath342 , associated with a _ type b _ solution in @xmath377 , with @xmath321 part of the instance , we can compute a canonical solution @xmath305 such that @xmath563 . hence @xmath564 .
we can define a solution @xmath17 of @xmath1-ap(@xmath1 ) of cost @xmath380 , as follows .
type a _ solution for each @xmath318 associated with a vertex @xmath565 .
each of such sets has a cost of @xmath319 .
define a _ type b _
solution for set @xmath318 associated with a vertex @xmath566 , and define an i - normal solution for the sets @xmath321 , @xmath338 , @xmath369 .
each such set @xmath318 has a cost of @xmath319 , and each edge set in @xmath567 , @xmath338 , @xmath568 has a cost of @xmath366 .
accounting this decreasing of the cost of the edge sets ( from @xmath367 to @xmath366 ) to the set @xmath318 , is equivalent to assign to a _ type b _
solution a cost equal to @xmath187 . for any other edge set @xmath321 add to @xmath17 the following sets : @xmath569 , @xmath570 .
each such edge set has a cost of @xmath367 .
finally , define a set @xmath571 , having a total cost of @xmath319 . | the problem of publishing personal data without giving up privacy is becoming increasingly important . a clean formalization that has been recently proposed
is the @xmath0-anonymity , where the rows of a table are partitioned in clusters of size at least @xmath0 and all rows in a cluster become the same tuple , after the suppression of some entries .
the natural optimization problem , where the goal is to minimize the number of suppressed entries , is hard even when the stored values are over a binary alphabet and as well as on a table consists of a bounded number of columns . in this paper
we study how the complexity of the problem is influenced by different parameters .
first we show that the problem is w[1]-hard when parameterized by the value of the solution ( and @xmath0 ) .
then we exhibit a fixed - parameter algorithm when the problem is parameterized by the number of columns and the maximum number of different values in any column . finally , we prove that @xmath0-anonymity is still apx - hard even when restricting to instances with @xmath1 columns and @xmath2 . |
heterostructures with nanoobjects ( such as quantum dots , nanowires , molecules , etc ) hybridized to one conducting and another superconducting electrode seem to be promising testing fields where the strong electron correlations ( responsible e.g. for coulomb blockade and kondo physics @xcite ) can be confronted with the superconducting order @xcite .
coulomb repulsion between electrons in the solid state physics is known to suppress the local ( @xmath2-wave ) pairing and , through the spin exchange mechanism , eventually promotes the intersite ( @xmath3-wave ) superconductivity @xcite .
mutual relation between such repulsion and the local pairing is however rather difficult for studying , both on theoretical grounds and experimentally . in nanoscopic heterostructures
some of these limitations can be overcome by a suitable adjustment of the hybridization and the gate - voltage positioning of energy levels involved in the charge transfer @xcite .
they enable a controllable changeover between the kondo regime and opposite case dominated by the induced on - dot pairing .
quantum dot ( qd ) coupled with the strength @xmath4 to metallic conductor ( n ) and with @xmath5 to superconducting electrode ( s ) can exhibit the features characteristic both for the on - dot pairing and the kondo effect ( including their coexistence ) @xcite .
their efficiency depends on the ratio @xmath6 . in the limit
@xmath7 the underlying physics is controlled by on - dot pairing and manifests itself e.g. by the particle - hole splitting of the quasiparticle levels . on the other hand for @xmath8
the strong correlations take over .
non - trivial aspects related to such interplay between the coulomb interactions and the proximity induced on - dot pairing has been addressed theoretically using various methods like : the mean field slave boson approach @xcite , the noncrossing approximation @xcite , perturbative scheme @xcite , constrained slave boson technique @xcite , numerical renormalization group @xcite and other @xcite .
also the cotunneling regime of a coulomb blockaded quantum dot sandwiched between a normal and superconducting lead , where charge fluctuations are strongly suppressed , has been discussed emphasizing the role of in - gap resonances @xcite .
as far as the experimental situation is concerned it has been less intensively explored .
the earliest transport measurements for n - qd - s interface have been obtained using the multi - wall carbon nanotube deposited between au and al electrodes @xcite .
those investigations concentrated however on the specific regime @xmath9 , when the coulomb correlations dominated over the proximity effect .
other studies of the same group have been done for similar structures replacing a metallic electrode by a ferromagnet @xcite .
several recent efforts focused on the multiterminal structures involving two normal and one superconducting electrodes as useful schemes for realization of : the crossed andreev reflections tunable via gate voltages @xcite , the cooper pair splitters @xcite , and the qd spin valves @xcite .
= 6.5 cm very useful understanding of a subtle interplay between the correlations and the induced on - dot pairing has been gained from recent measurements by r.s .
et al _ @xcite .
the authors have explored the subgap transport originating from the andreev - type scattering processes for several representative ratios @xmath6 using the self - assembled inas quantum dots deposited between the golden ( n ) and aluminium ( s ) electrodes .
their measurements provided the unambiguous experimental evidence for : a ) particle - hole splitting of the subgap conductance of the andreev states when @xmath10 , and b ) enhancement of the zero - bias andreev conductance due to formation of the kondo resonance at the fermi level of metallic lead , as has been qualitatively suggested by our studies @xcite and also indicated by other groups @xcite .
the present work extends our former studies by taking into account interference effects arising from the additional degrees of freedom . as the simplest prototype for fano - type interference
@xcite we consider the setup ( see figure [ scheme ] ) with a side - attached quantum dot contributing an extra pathway for electrons transmitted between the metallic and superconducting leads .
our analysis is complementary to the previous study by y. tanaka _ et al _ who considered the double quantum dots coupled between n and s leads in a @xmath0-shape setup but assuming @xmath11 , @xmath12 @xcite and in a series configuration @xcite . in section 2
we introduce the microscopic model and briefly outline basic notes on the nonequilibrium subgap transport . in the next section 3
we discuss a unique way in which the fano - type lineshapes might be observed in andreev conductance , focusing on the uncorrelated quantum dots . in the last part ( section 4 )
we discuss the influence of correlations at the interfacial quantum dot which seem to have remarkable influence on the low bias transport .
we end with the summary and suggestions for the future studies .
for description of the heterojunction illustrated in figure [ scheme ] we use the hamiltonian @xmath13 where the double quantum dot ( dqd ) is represented by @xmath14 the energies of each ( @xmath15 ) quantum dot electrons are denoted by @xmath16 and @xmath17 stands for the usual interdot hopping .
we restrict considerations of the correlation effects ( section 4 ) to the coulomb repulsion @xmath18 between opposite spin electrons @xmath19 , @xmath20 at the interfacial quantum dot .
the external reservoirs n and s of charge carriers are described by @xmath21 and correspondingly @xmath22 assuming the isotropic energy gap @xmath23 .
as usually @xmath24 denote the electron energies measured from the individual chemical potentials @xmath25 which become detuned @xmath26 if a bias @xmath27 is applied across the junction inducing the nonequilibrium charge flow @xmath28 .
fano - type quantum interference effects originating from the hopping @xmath17 to side - coupled quantum dot @xmath29 are discussed here assuming that only the interfacial quantum dot @xmath30 is directly coupled to external leads @xmath31 in the wide - band limit approximation it is convenient to introduce the structureless coupling constants @xmath32 which shall be used here as the energy units .
interplay between the proximity induced on - dot pairing , the correlations and the quantum interference effects can be in practice detected by measuring the differential conductance @xmath33 . particularly valuable for this purpose is the low voltage ( subgap ) regime @xmath34 . under such conditions the charge current
is provided by the anomalous andreev scattering in which electrons from the metallic lead are converted into the cooper pairs in superconductor with a simultaneous reflection of the electron holes back to the normal lead . on a formal level
the resulting andreev current can be expressed by the landauer - type formula @xcite @xmath35 , \label{i_a}\end{aligned}\ ] ] where @xmath36 is the fermi distribution function and the transmittance @xmath37 depends on the off - diagonal part ( in the nambu notation ) of the retarded green s function ( [ gf ] ) of the interfacial quantum dot .
fano resonances appear in many physical systems due to the quantum interference of the waves transmitted resonantly via some discrete energy level combined with transmittance contributed from a continuum of other states . in nanoscale physics
such resonances are feasible in a variety of constructions @xcite .
fano lineshapes are present for instance in the electron transport when two external electrodes are in parallel coupled through a quantum dot and directly via a shortcut bridge @xcite .
another simple possibility takes place in the electron tunneling using two quantum dots with considerably different linebroadenings @xcite . in the latter case
the narrower level is responsible for forming the fano resonance on a background of the broader level . in this work
we want to analyze similar interference effects appearing in the anomalous andreev current , which is very specific because of the particle and hole degrees of freedom mixed with one and other . to have a clear picture of the underlying physics let us start by considering the noninteracting case @xmath38 when the green s functions of each quantum dot can be determined exactly .
electron transport of the setup shown in figure [ scheme ] is determined by effective properties of the interfacial quantum dot . for this purpose
we compute the matrix green s function @xmath39 introducing the standard spinor notation @xmath40 and @xmath41 . in the equilibrium conditions
@xmath42 this function depends only on the time difference and its fourier transform obeys the following dyson equation @xmath43 where the term @xmath44 corresponds to the selfenergy of noninteracting case ( @xmath11 ) and @xmath45 accounts for the correlation effects ( discussed in section iv ) .
focusing on the deep subgap regime @xmath46 we obtain that @xmath44 simplifies to ( see the appendix ) @xmath47 when the interference effects caused by the hopping @xmath17 to the side - coupled qd are neglected the expression ( [ sigma0 ] ) becomes static ( @xmath48-independent ) and nontrivial physics of this , so called _ atomic superconducting limit
_ , has been explored in detail by several groups @xcite including ourselves @xcite .
taking into account the quantum interference @xmath49 we show in figure [ dos ] the proximity induced on - dot pairing [ formally arising from the off - diagonal parts of ( [ sigma0 ] ) ] illustrating the energy spectrum @xmath50 obtained for strong coupling to the superconducting lead @xmath51 .
such coupling @xmath52 is responsible for the particle - hole splitting of the effective quasiparticle states formed at @xmath53 whereas the coupling @xmath4 controls their broadening . in the particular case
@xmath54 the quasiparticle peaks appearing at @xmath55 ( where @xmath56 ) are symmetric , but for arbitrary @xmath57 they are weighted by the corresponding bcs coefficients @xmath58 @xcite . on top of such behavior
we clearly notice that hopping to the side - coupled quantum dot induces additional features appearing in the effective spectrum near @xmath59 as the fano resonance and antiresonance . for the case of both metallic leads there would survive just the single fano structure at @xmath60 which in very pedagogical way has been discussed by r. itko @xcite .
fano - type lineshapes ( see the lower panel in figure [ dos ] ) are present only in the weak hopping regime @xmath61 . for increasing @xmath17
the fano structures gradually evolve into separate quasiparticle peaks illustrated in figure [ dos_evol ] .
physically this can be assigned to the induced pairing on the side - attached qd @xmath62 transmitted there indirectly via the interfacial quantum dot .
such effect again qualitatively differs from the structures of the dqd coupled to both metallic leads @xcite .
interrelation between the interference and proximity effect can be practically investigated by measuring the tunneling current . in figure
[ ga ] we show bias voltage @xmath27 dependence of the differential andreev conductance @xmath63 determined at zero temperature from ( [ i_a ] ) over a broad regime covering both the subgap quasiparticle peaks .
figure [ ga_fano ] illustrates the resulting fano - type lineshapes @xmath64 nearby @xmath65 , where @xmath66 and the asymmetry parameter @xmath67 gradually decreases upon increasing the hopping integral @xmath17 .
our results can be thought as extension of the predictions obtained for the normal electron tunneling using the @xmath0-shape dqd coupled to both metallic leads @xcite onto the anomalous andreev current where the particle hole mixing has the essential importance .
coulomb repulsion between electrons of opposite spins can have an important influence on the spectral and transport properties of various nanostructures .
for the case of quantum dots coupled to both conducting leads such interactions are known to be responsible for : a ) the charging effect ( if a given energy level @xmath16 is attempted to be occupied by more than a single electron this costs the system an extra energy @xmath68 ) , b ) the kondo effect when the singlet state is formed between qd and itinerant electrons from the leads @xcite . in spectroscopic properties they are manifested by appearance of the coulomb satellite around @xmath69 and the narrow kondo resonance at the fermi level . for heterostructures with the superconducting electrodes the situation is more complex due to a competition between
the induced on - dot pairing and coulomb repulsion .
the rich interplay between the quantum interference , correlations and proximity effect for the configuration shown in figure [ scheme ] have been so far addressed using the density functional technique @xcite ( which does not capture the kondo physics ) and by the numerical renormalization group calculations @xcite . in latter case
the authors focused on @xmath11 , @xmath12 when the side - attached quantum dot can indirectly form the kondo state with electrons of the metallic lead affecting the andreev transport . to account for the correlation effects predominantly originating from the interfacial quantum dot
we extend here the procedure previously used by us for studying the single quantum dot @xcite .
the main idea is to approximate the correlation selfenergy @xmath45 by the diagonal matrix @xmath70 such assumption ( applied also in the nrg studies @xcite ) can be thought as the simplest ansatz for the many - body selfenergy @xmath45 allowing to combine the proximity effect ( [ sigma0 ] ) with the correlations , brought separately from the particle and hole channels . in more advanced treatments
one should take into account the possible feedback effects between these normal and anomalous channels .
we nevertheless hope that by imposing ( [ u_correction ] ) we can get some insight at least on a qualitative level which might stimulate the future studies . within qualitative studies of the correlation effects
we can describe the coulomb blockade and kondo effects using the following equation of motion expression @xcite @xmath71 [ \tilde{\omega}\!-\!\varepsilon_{1}\ ! -\!u_{1 } \!-\!\sigma_{3}(\omega)]\!+\!u_{1 } \sigma_{1}(\omega ) } { \tilde{\omega}-\varepsilon_{1 } - [ \sigma_{3}(\omega)+u_{1}(1\!-\!n_{1,\sigma } ) ] } \label{ansatz}\end{aligned}\ ] ] where @xmath72 $ ] , @xmath73 and @xmath74 .
we explored the interfacial quantum dot spectrum and the related transport properties at @xmath75 , i.e. well below the kondo temperature .
specific numerical computations have been done for @xmath76 , @xmath77 and symmetric coupling to both external leads giving the optimal conditions for any low - bias features in the andreev current @xcite .
this ratio @xmath78 is a reason why the particle - hole splitting is hardly visible , but otherwise ( for larger @xmath52 ) the kondo peak is either reduced or completely absent @xcite .
upper panel in the figure [ corr_dos ] illustrates the fano resonance / antiresonance around @xmath79 ( where @xmath80 ) obtained for the hopping @xmath81 .
these fano - type interference objects appear on top of the characteristic spectrum consisting of the kondo resonance and the broad quasiparticle peaks seen at @xmath57 and its coulomb satellite at @xmath82 .
such spectrum is the result of interference effects discussed in the preceding section and the correlation features .
the lower panel of figure [ corr_dos ] shows the very specific situation @xmath83 when the kondo and fano structures coincide with each other .
the fano - type resonance seems to play then a dominant role .
differential conductance of the andreev current ( [ i_a ] ) for the @xmath0-shaped double quantum dot system ( [ model ] ) is shown in figure [ kondo_peak ] .
we notice a clear suppression of the zero - bias peak ( present for the single quantum dot case @xmath84 as indicated by the dashed lines ) because of a destructive influence of the fano - type interference ( the solid lines ) .
the subgap andreev current is thus very sensitive to the interplay between the kondo and fano effects . for obvious reasons their most dramatic competition occurs when energy of the side - coupled quantum dot coincides with the kondo resonance , i.e. for @xmath85 .
quantum interference effects destroy then the kondo peak washing out the zero - bias enhancement of the andreev conductance .
we have studied a unique nature in which the fano - type quantum interference manifest itself in the energy spectrum and differential conductance of the heterojunction where a metallic lead is coupled via double quantum dot to superconducting electrode . in the regime of subgap source - drain voltage
@xmath86 nonequilibrium charge transport is contributed only through the anomalous andreev channel when electron from the metallic electrode is converted into the cooper pair ( propagating further in superconductor ) with a simultaneous reflection of hole back to the metallic lead .
transmittance of such andreev scattering is a sensitive probe of the proximity induced on - dot pairing as well as the quantum interference and correlation effects .
since on - dot pairing mixes the particle with hole states the interference effects are doubled in a comparison to similar junctions without the superconducting electrode . in particular , for @xmath0-shape configuration schematically shown in figure [ scheme ] we notice that effective spectrum of the interfacial quantum dot develops the resonance and antiresonance , correspondingly at @xmath87 ( figure [ dos ] ) .
these fano - type structures are present whenever the hopping integral @xmath17 to the side - attached quantum dot ( @xmath29 ) is much smaller than the linebroadening @xmath4 ( whereas @xmath52 merely controls the induced quasiparticle splitting ) . upon increasing @xmath17
the fano - type features disappear , evolving into the new quasiparticle peaks ( figure [ dos_evol ] ) being a consequence of the proximity effect indirectly spread onto the side - attached quantum dot .
correlation effects play an important role with regard to the following aspects : a ) the charging effect which causes appearance of the coulomb satellite near @xmath82 , b ) the kondo singlet state ( when the interfacial quantum dot spin is effectively screened by electrons of the metallic electrode leading to formation of a narrow resonance at @xmath88 ) , and c ) eventual suppression the on - dot pairing .
we have previously shown @xcite that the kondo effect enhances the zero - bias andreev conductance as indeed reported experimentally @xcite . in the present work
we indicate that in the double quantum dots the quantum interference can ( destructively ) affect such feature if the fano - type structures appear nearby the kondo peak
. a more detailed analysis of the fano - kondo interplay could be a challenging task in the future studies .
for this purpose one should resort either to nonperturbative techniques , like the numerical renormalization group , or to some sophisticated perturbative methods capable to interpolate between the limits @xmath89 , @xmath90 and @xmath91 .
we thank b. buka and k.i .
wysokiski for instructive discussions on the fano resonances in nanophysics .
this work is partly supported by the polish ministry of science and education under the grant nn202187833 .
using the standard nambu notation we can express the retarded green s functions of the metallic lead @xmath92 the ( unperturbed ) side - attached quantum dot @xmath93 and the isotropic superconductor @xmath94 in the equation ( [ gs ] ) we applied the bcs coefficients @xmath95 \nonumber \\
u_{\bf k}v_{\bf k } & = & \frac{\delta}{2e_{\bf k } } , \nonumber\end{aligned}\ ] ] where @xmath96 . for the case of uncorrelated quantum dots ( @xmath38 ) we can determine the selfenergy @xmath97 of the interfacial quantum dot from the following equation @xmath98 assuming the wide - band limit we introduce the constant weighed density of states @xmath99 where @xmath100 is the conduction bandwidth
we then easily find that @xmath101 because , according to the kramers - krnig relation , the real part disappears .
in the same way we obtain from a straightforward algebra that @xcite @xmath102 where @xmath103 in the extreme subgap limit @xmath46 the function ( [ gamma ] ) approaches @xmath104 and in consequence @xmath105 which proves the equation ( [ sigma0 ] ) .
11 m. pustilnik and l.i .
glazman , j. phys . : condens .
matter * 16 * , r513 ( 2004 ) .
y. yamada , y. tanaka , and n. kawakami , arxiv/1101.5239 ( 2011 ) .
anderson , p.a .
lee , m. randeria , t.m .
rice , n. trivedi , f.c .
zhang , j. phys . : condens .
matter * 16 * , r755 ( 2004 ) .
nazarov , y.l .
blanter , _ quantum transport : introduction to nanosciene _ , cambridge university press , cambridge ( 2009 ) . r.s .
deacon , y. tanaka , a. oiwa , r. sakano , k. yoshida , k. shibata , k. hirakawa , and s. tarucha , phys .
lett . * 104 * , 076805 ( 2010 ) ; r.s .
deacon , y. tanaka , a. oiwa , r. sakano , k. yoshida , k. shibata , k. hirakawa , and s. tarucha , phys .
b * 81 * , 121308(r ) ( 2010 ) and the supplemented on - line information .
r. fazio and r. raimondi , phys .
lett . * 80 * , 2913 ( 1998 ) ; phys .
lett . * 82 * , 4950 ( 1999 ) ; p. schwab and r. raimondi , phys .
b * 59 * , 1637 ( 1999 ) .
a.a . clerk , v. ambegaokar , and s. hershfield , phys .
b * 61 * , 3555 ( 2000 ) .
cuevas , a. levy yeyati , and a. martin - rodero , phys .
b * 63 * , 094515 ( 2001 ) .
m. krawiec and k.i .
wysokiski , supercond .
technol .
* 17 * , 103 ( 2004 ) .
y. tanaka , n. kawakami , and a. oguri , j. phys .
76 , 074701 ( 2007 ) .
j. bauer , a. oguri , and a.c .
hewson , j. phys . : condens
. matter * 19 * , 486211 ( 2007 ) .
t. hecht , a. weichselbaum , j. von delft , and r. bulla , j. phys .
condens .
matter * 20 * , 275213 ( 2008 ) .
sun , j. wang , and t .- h .
lin , phys .
b * 59 * , 3831 ( 1999 ) ; q .- f . sun , h. guo , and t .- h .
lin , phys .
lett . * 87 * , 176601 ( 2001 ) .
cho , k. kang , and c .- m .
ryu , phys . rev .
b * 60 * , 16874 ( 1999 ) .
y. avishai , a. golub , and a.d .
zaikin , phys .
b * 63 * , 134515 ( 2001 ) ; t. aono , a. golub , and y. avishai , phys .
b * 68 * , 045312 ( 2003 ) .
t. domaski and a. donabidowicz phys .
b * 78 * , 073105 ( 2008 ) ; t. domaski , a. donabidowicz , and k.i .
wysokiski , phys .
b * 78 * , 144515 ( 2008 ) ; phys .
b * 76 * , 104514 ( 2007 ) .
v. koerting , b.m .
andersen , k. flensberg , and j. paaske , phys .
b * 82 * , 245108 ( 2010 ) .
graeber , t. nussbaumer , w. belzig , and c. schnenberger , nanotechnology * 15 * , 479 ( 2004 ) . l. hofstetter , a. geresdi , m. aagesen j. nygard , c. schnenberger , and s. csonka , phys .
* 104 * , 246804 ( 2010 ) .
herrmann , f. portier , p. roche , a. levy yeyati , t. kontos , and c. strunk , phys .
lett . * 104 * , 026801 ( 2010 ) .
j. eldridge , m.g .
pala , m. governale , j. knig , phys .
b * 82 * , 184507 ( 2010 ) .
b. sothmann , d. futterer , m. governale , j. knig , phys .
b * 82 * , 094514 ( 2010 ) .
y. yamada , y. tanaka , and n. kawakami , j. phys .
79 * , 043705 ( 2010 ) .
miroshnichenko , s. flach , and y.s .
kivshar , rev .
* 82 * , 2257 ( 2010 ) .
y. tanaka , n. kawakami , and a. oguri , phys .
b * 78 * , 035444 ( 2008 ) ; j. phys . : conf .
series * 150 * , 022086 ( 2009 ) .
y. tanaka , n. kawakami , and a. oguri , phys .
b * 81 * , 075404 ( 2010 ) .
b. buka , p. stefaski , phys .
lett . * 86 * , 5128 ( 2001 ) .
r. itko , phys .
b * 81 * , 115316 ( 2010 ) .
p. trocha and j. barna , phys .
b * 76 * , 165432 ( 2007 ) .
t. meng , s. florens , and p. simon , phys .
b * 79 * , 224521 ( 2009 ) .
h. haug and a .- p .
jauho , _ quantum kinetics in transport and optics of semiconductors _ , springer verlag , berlin ( 1996 ) .
s. sasaki , h. tamura , t. akazaki , and t. fujisawa , phys . rev .
lett . * 103 * , 266806 ( 2009 ) .
a. karmanyos , i. grace , and c.j .
lambert , phys .
b * 79 * , 075119 ( 2009 ) . | we analyze the quantum interference effects appearing in the charge current through the double quantum dots coupled in @xmath0-shape configuration to an isotropic superconductor and metallic lead . owing to proximity effect
the quantum dots inherit a pairing which has the profound influence on nonequilibrium charge transport , especially in the subgap regime @xmath1 .
we discuss under what conditions the fano - type lineshapes might appear in such andreev conductance and consider a possible interplay with the strong correlation effects . |
large concentrations of matter in the universe are found using optical galaxies as tracers of mass .
systems with a wide range of mass and size were discovered by this technique , from pairs and triplets of galaxies to filaments extending for hunderds of mpc .
do optical galaxy surveys detect all large - scale mass concentrations , or do there exist populations of `` dark '' massive objects ?
one approach for detection of dark systems is through gravitational lensing .
week lensing observations can detect very large scale mass structures ( e.g. , schneider et al .
1998 ) , but this approach remains technically challenging .
compact , massive systems near the line of sight of a distant object can be detected by strong lensing .
at least one such object was found : a massive x - ray cluster at @xmath9 which is very poor optically for its mass ( hattori et al . 1997 , but see also benitez et al .
1998 ) .
x - ray surveys are another avenue for finding compact , massive , optically dark objects , because such objects are likely to contain gravitationally heated x - ray emitting gas . tucker , tananbaum & remillard ( 1995 ) did not find any completely dark x - ray clusters in the _ einstein _ data .
however , x - ray surveys with _
rosat _ did find a new class of objects that could not be discovered optically bright isolated elliptical galaxies surrounded by dark matter and hot gas halos typical of a group or poor cluster .
the first such object was a `` fossil group '' found by ponman et al .
the object appears optically as a giant isolated elliptical galaxy .
its x - ray halo extends for at least 500 kpc ; such an extent implies the group - like total mass .
bhringer et al .
( 1998 ) mention the existence of similar objects in the _ rosat _ all - sky survey .
x - ray observation of an optically selected isolated elliptical revealed the existence of a group - like x - ray halo ( mulchaey & zabludoff 1998 ) .
our large - area _ rosat _ survey of extended x - ray sources ( vikhlinin et al .
1998 , paper i hereafter ) is ideal for the search for optically dark clusters and groups at low redshift . in this _ letter _ , we report a detection of four objects similar to the ponman et al .
fossil group .
we call them x - ray over - luminous elliptical galaxies ( olegs ) .
we show that these objects have gas and dark matter halos extending up to radii of 1 mpc ; they are more numerous and massive than was previously appreciated , and represent an important class of mass concentrations in the universe .
we use @xmath10 , except where the @xmath11-dependence is explicitly given , and @xmath12 .
all uncertainties are reported as 68% confidence intervals .
the @xmath13 and @xmath14 band optical ccd images of serendipitously detected extended x - ray sources from our _ rosat _ pspc catalog ( paper i ) were obtained on the flwo 1.2 m and danish 1.54 m telescopes .
we examined these images and selected those objects that were obviously associated with a bright elliptical galaxy , but had no corresponding concentration of fainter galaxies .
to add confidence to this selection , two further critera were applied .
first , the central galaxy redshift was required to be @xmath15 because galaxy concentration is harder to detect in more distant objects .
second , we required the 0.52 kev x - ray luminosity to be @xmath16erg s@xmath1 .
this luminosity corresponds to poor abell clusters which should be visible as galaxy concentrations in our ccd images .
x - ray fainter objects can be poor groups and therefore we could misidentify them with single , isolated galaxies .
these selections resulted in a sample of four objects ( table [ tab : sample ] ) .
two of them , @xmath17 and @xmath18 , were previously detected in the _ einstein _ surveys ( stocke et al .
1991 and griffiths et al .
1992 , respectively ) and identified with normal elliptical galaxies .
we also rediscovered the ponman et al .
object , @xmath19 . the remaining object , @xmath20 , first appears in our sample .
we have measured the redshift of @xmath20 and adopted the redshifts of others from the literature .
there are several other plausible oleg candidates , however they do not satisfy either the redshift or the x - ray luminosity criteria . in all four objects ,
the x - ray emission is detected with @xmath21 significance out to a large radius @xmath22mpc ( table [ tab : sample ] ) .
the x - ray surface brightness distribution shows no significant deviations from azimuthal symmetry .
our optical images cover an area large enough to observe the entire x - ray emitting region and to determine the local density of background galaxies .
we have measured the optical luminosity by co - adding light from the central galaxy and all fainter galaxies within @xmath23 and subtracting the estimated background contribution .
the resulting @xmath13-band luminosities ( with k- and extinction corrections applied ) are in the range @xmath24 .
these values are uncertain by @xmath25 due to background fluctuations caused by the presence of relatively bright foreground or background galaxies . in all objects
, at least @xmath26 of light ( exluding the background ) comes from the central galaxy ; the contribution from other galaxies does not exceed the level of background fluctuations .
although this object was discovered by _ einstein _ , its large extent in x - rays remained unnoticed prior to the _ rosat _ observation .
the long _ rosat _ exposure and the low background made it possible to trace the diffuse x - ray emission to a large radius and even perform a simple spectral analysis .
the optical and x - ray properties of 1159 + 5531 appear to be representative of the other objects in our sample , and the main conclusions derived from the 1159 + 5531 data can be applied to the entire oleg population .
the x - ray contour map of 1159 + 5531 overlayed on the optical ccd image is shown in fig [ fig:1159:opt ] .
the x - ray surface brightness profile is shown in fig [ fig:1159:x]a .
the surface brightness is detected to @xmath27mpc from the center with @xmath28 significance .
the @xmath29-model fit ( cavaliere & fusco - femiano 1976 ) in the entire radial range yields @xmath30 and a small core - radius value @xmath31kpc .
the central gas density derived from this fit corresponds to the gas cooling time @xmath32yr indicating the presence of the cooling flow . outside the central 200 kpc region , where the surface brightness is unaffected by enhanced cooling flow emission , the best fit @xmath29-model yields @xmath33kpc and @xmath34 .
these large uncertainties arise from the interplay between @xmath35 and @xmath29 .
fortunately , both gas and total mass at large radius depend mostly on the slope of the surface brightness profile and not on the individual values of @xmath35 and @xmath29 .
the surface brightness slope is well - constrained a power law fit , @xmath36 , between @xmath37 and @xmath38kpc yields @xmath39 .
figure [ fig:1159:x]b presents the gas temperatures in four annuli measured by fitting the _ rosat _ spectrum with the raymond & smith ( 1977 ) model .
we fixed the metal abundance at @xmath40 of the solar value and the galactic hydrogen column at @xmath41@xmath42 derived from the radio surveys ( stark et al .
fit to the overall _ rosat _ spectrum with free absorption yielded a consistent value , @xmath43@xmath42 . within 100kpc ,
the temperature is significantly lower than in the 100250 kpc annulus , confirming the existence of the cooling flow .
temperatures are poorly constrained beyond 250kpc because of low statistics and because _ rosat _ is insensitive above 2kev ; nevertheless , the lower limits show that the temperature does not drop below 1kev to at least 800kpc from the center .
the temperature @xmath44kev derived from the @xmath45 correlation for 13kev clusters and groups ( fukazawa 1997 , hwang et al .
1999 ) is consistent with the spectral fit outside the cooling flow region .
we will use this value for the mass calculations below .
the gas mass was calulated by deprojection of the observed x - ray surface brightness profile ( fabian et al . 1981 , white , jones & forman 1997 ) .
the total mass was derived from the hydrostatic equilibrium equation assuming constant temperature at all radii @xmath46 where radii are in units of mpc .
the results of the mass determination at several radii are presented in table [ tab : m:1159 ] .
the formal statistical uncertainty of the total mass is 3040% , including uncertainties both in the slope of the surface brightness profile and in the overall temperature . [
cols="<,^,^,^,^,^ " , ] optically , 1159 + 5531 appears as an isolated giant elliptical residing at the peak of the x - ray emission ( fig [ fig:1159:opt ] ) .
the sensitivity limit of this image corresponds to absolute magnitude @xmath47 at the object redshift . above this sensitivity limit
, we detect 18 galaxies within a projected distance of 100 kpc from the central galaxy , while @xmath48 are expected due to the background estimated beyond 1 mpc .
to measure the total optical luminosity , we integrated the light within 90 kpc of the central galaxy and within 15 kpc of fainter galaxies .
we excluded the objects with peak surface brightness exceeding that of the central galaxy ; the radial profiles of exluded objects showed that all they were stars . the light density due to background galaxies was measured outside a projected distance of 1 mpc from the central galaxy .
the optical luminosities within 0.4 , 0.7 , and 1 mpc with the subtracted background contribution are reported in table [ tab : m:1159 ] .
the optical properties of the central galaxy itself are remarkable .
the diffuse optical light can be traced to @xmath49kpc in @xmath13 , @xmath14 , and @xmath50 bands .
the light profile very accurately follows the de vaucouleurs law with an effective radius @xmath51kpc .
the absolute magnitude is @xmath52 and @xmath53 ( with k- and galactic extinction correction applied ) .
such a high luminosity is rare in field ellipticals but is typical of cd galaxies in clusters ( hoessel 1980 ) .
to summarize , 1159 + 5531 is an object with x - ray properties typical of poor clusters and which contains an isolated galaxy whose optical properties resemble central cluster galaxies .
a detailed analysis of the surface brightness and spectral fitting could be performed only with the 1159 + 5531 data . in the remaining three objects ,
only crude estimates of the total and gas mass are possible . to control the accuracy of these estimates ,
we apply them also to 1159 + 5531 .
the gas temperature is estimated from the @xmath45 correlation ( fukazawa 1997 , hwang et al .
the temperature scatter around the mean @xmath45 relation is only @xmath54 , which is accurate enough for our purposes .
additional confidence is added by the agreement of the @xmath45 temperature estimate and the direct spectral fit in @xmath17 .
the x - ray surface brightness profiles are fit with the @xmath29-model with free normalization and core - radius , but we fix @xmath55 because of low statistics . using the @xmath29-model fit , we derive the central gas cooling time . in all objects except for @xmath20 , it is much shorter than the hubble time , therefore they likely contain cooling flows . the gas mass within @xmath23 , the radius of the detectable x - ray emission , is also derived using the @xmath29-model fit . for a fixed @xmath56 ,
the derived value of gas mass is relatively insensitive to @xmath29 , therefore fixing @xmath55 does not introduces significant errors in the gas mass .
unfortunately , the presence of the cooling flow limits the accuracy of the @xmath29-model gas mass estimates .
for example , a detailed modelling of the @xmath17 surface brightness profile including the cooling flow yields @xmath57 larger gas mass than the global @xmath29-model fit .
the estimated temperature and @xmath55 is substituted into eq ( [ eq : hydro - eq ] ) to derive the total mass .
the estimated gas and total masses are listed in table [ tab : sample ] . for @xmath17 ,
the crude method yields a 30% lower gas mass and 20% lower total mass compared to a more detailed analysis in [ sec:1159 ] .
central galaxies in olegs clearly dominate their surroundings . the brightest galaxy within the projected distance of 500kpc around @xmath17 is 2.9 magnitudes fainter and may be at a different redshift .
olegs do not show any detectable concentration of galaxies in projection down up to 7.5 magnitudes fainter than the central galaxy .
however , we can not completely exclude the existence of a dwarf galaxy population in these systems .
mulchaey & zabludoff ( 1998 ) have studied ngc 1132 , a nearby optically selected galaxy with x - ray properties similar to olegs and @xmath58 , and found a concentration of dwarf galaxies with @xmath59 in the range from @xmath60 to @xmath61 , consistent in number with that in x - ray detected galaxy groups .
the sensitivity of our @xmath17 image is adequate to detect such galaxies .
we do not find any evidence for existence of a dwarf concentration around @xmath17 , but the upper limit on their number is consistent with the composite group profile presented by mulchaey & zabludoff .
the possibility of existence of a dwarf population around @xmath17 and our other objects can be tested only by a detailed redshift survey .
the volume covered by our survey contains 4 objects .
the corresponding bayesian lower and upper 95% confidence limits of the true number of objects are 1.97 and 9.15 , respectively ( see , e.g. , kraft , burrows , & nousek 1992 ) . for given @xmath56
, the survey volume can be calculated using the dependence of solid angle on limiting flux ( paper i ) .
we find a volume of @xmath62mpc@xmath63 for objects with @xmath64erg s@xmath1 ; a slightly larger volume of @xmath65mpc@xmath63 is obtained for objects with @xmath66erg s@xmath1 . conservatively assuming the larger volume
, we obtain the spatial density of olegs of @xmath67mpc@xmath6 at 90% confidence .
this number density is comparable to the number of other objects of similar nature galaxy groups ( compact groups in particular ) and field elliptical galaxies in the corresponding magnitude range . the x - ray luminosity functions from ebeling et al .
( 1997 ) and burns et al .
( 1996 ) show that olegs represent @xmath68 of all clusters and groups with @xmath16erg s@xmath1 . using the x - ray and optical luminosity functions of hickson compact groups ( hcgs ) from ponman et al .
( 1996 ) and sulentic & rabaa ( 1994 ) , we find that olegs outnumber comparably x - ray luminous hcgs by a factor of @xmath69 and are as numerous as hcgs of comparable total optical luminosity . to estimate the number density of field elliptical galaxies , we used the @xmath13-band luminosity function derived from the las campanas redshift survey ( lin et al .
1996 ) . above a limiting absolute magnitude @xmath70 corresponding to the luminosity of our faintest object @xmath18 ( and 2.2 brighter than @xmath71 ) , we find the number density of _ all _ galaxies @xmath72mpc@xmath6 .
most of these bright galaxies can be spectrally classified as early types ( bromley et al .
some of them should lie within rich clusters , so the above number density is an upper limit for the field population .
therefore , most , if not all , of the brightest field ellipticals ( with @xmath73 ) are olegs , i.e. possess x - ray halos extending for hundreds of kpc .
this also means that our oleg sample can not be significantly incomplete .
the contribution of olegs to the total mass in the universe is close to that of rich clusters .
markevitch ( 1998 ) derived the number density @xmath74mpc@xmath6 for @xmath8kev clusters . since the total mass scales with temperature as @xmath75 , olegs ( @xmath76kev ) are 6.5 times less massive than 7kev clusters .
they outnumber such clusters by a factor of @xmath77 , and hence their contribution to the total mass density is similar .
olegs have unusually high mass - to - light ratios . within 5001000 kpc ,
the radius where extended x - ray emission can still be detected , @xmath78 in the @xmath13-band is in the range 250450@xmath79 .
we find similar or slightly higher values of @xmath78 in the @xmath14-band .
our @xmath78 estimates are 23 times higher than the @xmath14-band values for clusters and groups determined from the x - ray mass measurements ( david , jones & forman 1995 ) .
is this difference real or it can be explained by our measurement errors ? the largest uncertainty in @xmath4 is due to the inaccuracy of the total mass measurement .
our total mass estimates are essentially proportional to temperatures estimated from the @xmath45 relation .
a 23-fold overestimate of @xmath80 from this relation is possible , but unlikely , given the good agreement between the estimate and measurement for @xmath81 .
a plausible explanation of the origin of olegs is that they are merged compact galaxy groups ( ponman et al .
1994 , mulchaey & zabludoff 1998 ) .
if this explanation is correct , our results imply that most of the brightest field ellipticals are products of group merging and that these products are more numerous than present - day groups .
the latter is consistent with the short estimated lifetime of compact groups ( mamon 1986 ) .
however , the high values of @xmath4 we find in olegs pose a problem for this scenario because the optical luminosity is unlikely to decrease during the galaxy merging .
even the scarce available data indicate that olegs represent a new interesting class of objects warranting a detailed x - ray and optical study .
these objects are likely to have been undisturbed for a long time , and thus represent the ultimate example of cluster - like systems in hydrostatic equilibrium . forthcoming _
observations of two objects from our sample will result in accurate profiles of the gas and total mass .
if high @xmath4 and low gas fraction is confirmed , this will pose a problem for low @xmath82 estimates from the values of these quantities in clusters .
we thank m. markevitch for useful discussions .
this work was supported by the smithsonian institution and nas8 - 39073 contract .
hq acknowledges support from grant 8970009 and the award of a presidential chair in science . | we detect four isolated , x - ray over - luminous ( @xmath0^{-2}$]erg s@xmath1 ) elliptical galaxies ( olegs ) in our 160 square degree _ rosat _ pspc survey .
the extent of their x - ray emission , total x - ray luminosity , total mass , and mass of the hot gas in these systems correspond to poor clusters , and the optical luminosity of the central galaxies ( @xmath2 ) is comparable to that of cluster cds . however , there are no detectable fainter galaxy concentrations around the central elliptical .
the mass - to - light ratio within the radius of detectable x - ray emission is in the range @xmath3 , which is 23 times higher than typically found in clusters or groups .
these objects can be the result of galaxy merging within a group . however , their high @xmath4 values are difficult to explain in this scenario .
olegs must have been undisturbed for a very long time , which makes them the ultimate examples of systmes in hydrostatic equilibrium .
the number density of olegs is @xmath5 mpc@xmath6 at the 90% confidence .
they comprise 20% of all clusters and groups of comparable x - ray luminosity , and nearly all galaxies brighter than @xmath7 .
the estimated contirubution of olegs to the total mass density in the universe is close to that of @xmath8kev clusters . |
the objective of this article is to design a controller in a _ networked control system _ @xcite that produces sparse control vectors for effective compression before transmissions .
unfortunately , the calculation of optimal sparse vectors will , in general , require significant computational cost and may thereby introduce delays , which are unacceptable for closed - loop operation . to overcome this issue , we subsample the problem to reduce its size and adopt a fast algorithm called _ fista _ ( fast iterative shrinkage - thresholding algorithm ) @xcite .
networked control systems are those in which the controlled plants are located away from the controllers , and the communication should be made through rate - limited communication channels such as wireless networks or the internet @xcite . in networked control systems ,
efficient signal compression or representation is essential to send control data through rate - limited communication channels . for this purpose
, we propose an approach of sparse control signal representation using the _ compressive sampling _ technique @xcite .
our contributions in this paper are ( 1 ) a new strategy for networked feedback control systems based on compressive sampling , ( 2 ) an effective data compression scheme of the control signals with sparse representation , ( 3 ) formulation of the design problem by @xmath0 optimization which can be efficiently solved
. the compressive sampling approach will open up a new vista in control theory . to the best of our knowledge
, so far only a few studies have applied compressive sampling to control : @xcite proposes to use compressive sensing in feedback control systems for perfect state estimation and @xcite proposes sparse representation of transmitted control packets for feedback control with packet dropouts . for remote control systems ,
@xcite also propose to use @xmath0 optimization ( as in this paper ) .
however , @xcite consider only feed - forward control systems .
[ fig : ncs ] shows the networked control system which we consider in this article .
the system consists of a controlled plant @xmath2 , a sensor ( or sampler ) @xmath3 , a decoder or a digital - to - analog ( da ) converter @xmath4 , and a digital controller @xmath5 .
the definitions of these systems are given as follows : plant @xmath2 : : : the controlled plant @xmath2 is modeled by the following state - space representation : @xmath6 where @xmath7 , @xmath8 , and @xmath9 .
sensor @xmath3 : : : the sensor ( or sampler ) @xmath3 converts the continuous - time state @xmath10 into a discrete - time signal @xmath11:={{{\boldsymbol{x}}}_{\text{c}}}(kt)$ ] , @xmath12 , where @xmath13 is the sampling period .
decoder @xmath4 : : : the decoder ( or da converter ) @xmath4 converts a vector valued signal @xmath14 $ ] into a continuous - time signal @xmath15 , @xmath12 via @xmath16 \mapsto u_k : = \sum_{m =- m}^m \theta_m[k]\psi_m \in l^2[0,t ) , \label{eq : psi}\ ] ] where @xmath17 is a positive integer , @xmath18 $ ] is the @xmath19-th element of the vector @xmath14 $ ] , and @xmath20 we call the vector @xmath14 $ ] a _
control vector_. note that the continuous - time signal @xmath21 is band - limited to @xmath22 [ rad / sec ] .
that is , @xmath21 belongs to the following subspace of @xmath23 : @xmath24 .
\label{eq : vm}\ ] ] the input @xmath25 to the plant @xmath2 is defined by @xmath26 , @xmath27 , @xmath12 .
controller @xmath5 : : : the controller @xmath5 uses a continuous - time reference signal @xmath28 , @xmath12 , and the sampled state @xmath11 $ ] to produce the control vector @xmath14 $ ] .
the latter defines the input signal @xmath21 as per ( [ eq : psi ] ) .
we assume that @xmath3 , @xmath4 , and @xmath5 are synchronized at @xmath29 , @xmath12 .
we also assume that we can transmit the control vector @xmath14 $ ] and the sampled state @xmath11 $ ] through communication channels without any delays nor packet dropouts . in this article
, we consider a situation where the size @xmath30 of @xmath14 $ ] is much larger than the size @xmath31 of the state @xmath11 $ ] , and should be compressed because @xmath14 $ ] needs to be transmitted through a rate - limited communication channel . under these assumptions
, we then formulate our control problem .
let @xmath32 be the continuous - time signal @xmath33 on the interval @xmath34 , that is , @xmath35 , @xmath27 , @xmath36 .
we design the controller @xmath5 to achieve the following objectives : 1 .
the first objective is to attenuate the tracking error between the reference @xmath37 and the output @xmath32 on the interval @xmath34 , @xmath12 .
the error is measured by the @xmath1 norm : @xmath38 2 .
the second objective is to reduce the data size of the control vector @xmath14 $ ] which defines the control @xmath21 via ( [ eq : psi ] ) . for this objective
, we adopt the so - called @xmath39-norm of @xmath21 defined by @xmath40\|_0 $ ] , the number of the nonzero elements in @xmath14 $ ] . in general , there is a trade - off between tracking - error attenuation and data - size reduction .
for example , the sparsest solution @xmath41 leads to very large error , and the control @xmath21 which minimizes only the first objective function may not be sparse . to solve this problem , we adopt _
regularization_. the problem is formulated as follows .
[ prob1 ] given reference signal @xmath42 , @xmath12 , find the control @xmath43 ( or the control vector @xmath14 $ ] ) which minimizes @xmath44 where @xmath45 is the regularization parameter to reconcile the trade - off between the tracking error and the sparsity .
the objective function @xmath46 in ( [ eq : prob1 ] ) is defined on an infinite - dimensional signal subspace @xmath47 defined in ( [ eq : vm ] ) .
we here relax the objective function into a finite - dimensional convex @xmath0 optimization problem by using the technique of compressive sampling . since the signals @xmath37 and
@xmath21 are assumed to be band - limited up to the frequency @xmath22 [ rad / sec ] , we can safely discretize the signals by sampling them at a sampling frequency higher than @xmath48 based on shannon s sampling theorem @xcite . however , if @xmath17 is very large , it may take very long time to compute the optimal vector .
it follows that there may exist a large delay in the feedback loop , which may lead to instability and control performance deterioration .
hence it is preferable to use a more efficient method than shannon s sampling . for this purpose
, we adopt the technique of compressive sampling @xcite with _ random sampling _ , which can reduce the computational load for the optimization .
random sampling is modeled as follows : we first split the interval @xmath49 with sampling points @xmath50 , @xmath51 , where @xmath52 is the nyquist rate .
then we randomly choose @xmath5 sampling points ( @xmath53 ) from @xmath54 .
to model this , we define a random matrix @xmath55^\top \in \{0,1\}^{k\times n}$ ] , where @xmath56 are discrete random variables chosen from the uniform distribution on @xmath57 such that @xmath58 , @xmath59 , and @xmath60 is the standard basis in @xmath61 , that is , @xmath62 ( @xmath63 ) denotes a unit vector whose @xmath64-th element is equal to one and the other elements are equal to zero .
the random matrix @xmath65 is re - chosen at every sampling step @xmath66 . by using the random variables
@xmath67 , we define the random sampling points by @xmath68 , @xmath69 , @xmath70
. then we consider random sampling of the output @xmath32 .
the sampled output @xmath71 with the control signal @xmath43 defined in ( [ eq : psi ] ) is computed by @xmath72 { \langle \phi_n , \psi_m \rangle } , \label{eq : ytn}\ ] ] where @xmath73 is the inner product in @xmath23 , and @xmath74{{\boldsymbol{b}}},\quad & t\in[0,t_n),\\ 0,\quad & t\in(t_n , t].\end{cases}\ ] ] define the randomly sampled output vector @xmath75:=[y_k(t_{i(1)}),\dots , y_k(t_{i(k)})]^\top\in { { \mathbb{r}}}^{k}.\ ] ] then by ( [ eq : ytn ] ) , we have @xmath76 = ug{{\boldsymbol{\theta}}}[k ] + uh{{\boldsymbol{x}}}[k]$ ] , where @xmath77 is an @xmath78 matrix defined by @xmath79 , @xmath80 , @xmath81 , and @xmath82 is an @xmath83 matrix defined by @xmath84^\top$ ] .
let @xmath85 : = \left[r_k(t_1),r_k(t_2),\dots , r_k(t_n)\right]^\top\in{{\mathbb{r}}}^n$ ] and @xmath86 , @xmath87=u({{\boldsymbol{r}}}[k]-h{{\boldsymbol{x}}}[k])$ ] .
then the tracking error at the random sampling points @xmath88 is given by @xmath76-{{\boldsymbol{r}}}[k ] = \phi{{\boldsymbol{\theta}}}[k]-{{\boldsymbol{\alpha}}}[k]$ ] .
it follows that the cost function ( [ eq : prob1 ] ) in problem [ prob1 ] is approximately described in a finite - dimensional one : @xmath89 ) : = \|\phi{{\boldsymbol{\theta}}}[k]-{{\boldsymbol{\alpha}}}[k]\|_2 ^ 2 + \mu \|{{\boldsymbol{\theta}}}[k]\|_0 .
\label{eq : j0}\ ] ] the minimization of the cost function ( [ eq : j0 ] ) is still difficult to solve when @xmath17 is large since the optimization is a combinatorial one . to reduce this
, we adopt a convex relaxation by replacing the @xmath90 norm with the @xmath91 norm : @xmath92 ) : = \|\phi{{\boldsymbol{\theta}}}[k]-{{\boldsymbol{\alpha}}}[k]\|_2 ^ 2 + \mu \|{{\boldsymbol{\theta}}}[k]\|_1 .
\label{eq : j1}\ ] ] the cost function @xmath93)$ ] in ( [ eq : j1 ] ) is convex in @xmath14 $ ] and hence the optimal value uniquely exists . to obtain the @xmath0 optimal vector
, we use an iterative algorithm called fista @xcite .
this algorithm is very simple and fast ; it can be effectively implemented in digital devices , which leads to a real - time computation in the feedback loop .
for this algorithm , see @xcite .
in this section , we illustrate simulation results to show the effectiveness of the compressive sampling technique in networked feedback control systems .
the matrices in the state - space representation ( [ eq : plant ] ) of the controlled plant @xmath2 are taken as @xmath94 where @xmath95 and @xmath96 .
we assume the initial state @xmath97 .
the control period @xmath98 is set to be @xmath99 .
the number of the basis functions @xmath100 , or the size of the control vector @xmath14 $ ] is @xmath101 ( @xmath102 ) .
we use the reference @xmath103 the sparsity of the reference @xmath37 is given by @xmath104 .
therefore , @xmath37 is a sparse vector when it is represented by the basis functions @xmath105 defined in ( [ eq : exp ] ) .
that is , the reference @xmath37 is sparse with respect to the basis @xmath100 .
the shortest sampling interval in random sampling is @xmath106 .
we set the number of random sampling @xmath107 .
the iteration steps in fista for minimizing the @xmath0 optimization in ( [ eq : j1 ] ) is 10 .
we run the simulation of the feedback control for @xmath108 , that is , the length of simulation time is @xmath109 .
first , we compute the relation between the regularization parameter @xmath110 in ( [ eq : j1 ] ) and metrics for control performance to be achieved by the optimal control vector @xmath14 $ ] .
we use two metrics : rms ( root mean square ) of the tracking error @xmath111 and the average sparsity of control vector .
@xmath112\}_{k = 0 } ^{100}$ ] .
the rms is defined as : @xmath113 the average sparsity is defined as : @xmath114\|_0}{101}$ ] .
[ fig : mu ] shows the performance as a function of the parameter @xmath110 . to compare the proposed method with a conventional one
, we consider the @xmath1-optimal control which minimizes @xmath115 .
this cost function limits the energy ( i.e. , the @xmath1 norm ) of the control @xmath21 , which has been widely used in control .
the optimal control vector , say @xmath116 $ ] , is given by @xmath117 = ( \mu_2i + g^\top g)^{-1}g^\top ( { { \boldsymbol{r}}}[k]-h{{\boldsymbol{x}}}[k ] ) , \label{eq : l2opt}\ ] ] where we assume the control @xmath21 is in the subspace @xmath47 defined in ( [ eq : vm ] ) .
the rms error performance for the @xmath1-optimal control is also shown in fig .
[ fig : mu ] . in this case , the feedback system becomes unstable for @xmath118 .
the sparsity of the optimal control vector is @xmath119\|_0=101 $ ] for all @xmath120 .
that is , the @xmath1-optimal control does not produce any sparse vectors at all . fig .
[ fig : mu ] suggests that the optimal parameter is @xmath121 for @xmath0 optimization , and @xmath122 for @xmath1 optimization . with these parameters ,
we simulate the feedback control .
[ fig : coef ] ( a ) shows the absolute value of the nonzero elements in the @xmath0 optimal control vector @xmath14 $ ] at @xmath123 .
we can see that the number of the nonzero elements is 8 out of 101 ( the size of the vector ) , and hence the vector is very sparse .
then , the @xmath1 norm of the tracking error @xmath124 on the @xmath66-th period @xmath125 ( @xmath108 ) is shown in the top figure in fig . [ fig : comparison ] . the sparsity history @xmath126\|_0\}_{k=0}^{100}$ ]
is shown in the bottom figure in fig .
[ fig : comparison ] . to compare the proposed method with conventional @xmath1-optimal control , we compute the control vector @xmath116 $ ] by the formula ( [ eq : l2opt ] ) with @xmath127 . fig .
[ fig : coef ] ( b ) shows the absolute value of the nonzero elements in the control vector @xmath116 $ ] at @xmath123 .
we can see that all the elements in this vector are nonzero ( cf fig .
[ fig : coef ] ( a ) ) .
one may think that the vector @xmath116 $ ] is _ compressible _ since almost all the elements are nearly zero . to see the difference , we truncate the full vector @xmath116 $ ] by using the sparsity history in fig . [ fig : comparison ] . the stars in fig .
[ fig : coef ] ( b ) are the 8 elements of the truncated vector .
the tracking error by the truncated vectors is shown in the top figure in fig.[fig : comparison ] .
the proposed method shows the better performance than the truncated @xmath1-optimal control with the same data size .
this shows the effectiveness of our method . in an additional simulation study
, we considered a step function for the reference , that is , @xmath128 , @xmath12 .
this signal is also sparse in the space @xmath47 and produces a sparse control vector ( we omit details due to space limitations ) .
we have studied the use of compressive sampling for feedback control systems with rate - limited communication channels .
simulation studies indicate that the method proposed can effectively compress the signals transmitted .
control vectors are obtained via an @xmath0 optimization , which is solved by the fista algorithm .
future work could include further investigation of bit - rate issues and the study of closed loop stability . | we investigate the use of compressive sampling for networked feedback control systems .
the method proposed serves to compress the control vectors which are transmitted through rate - limited channels without much deterioration of control performance .
the control vectors are obtained by an @xmath0 optimization , which can be solved very efficiently by fista ( fast iterative shrinkage - thresholding algorithm ) .
simulation results show that the proposed sparsity - promoting control scheme gives a better control performance than a conventional energy - limiting @xmath1-optimal control . |
one of the most interesting features of many nonlinear evolution equations is the spontaneous onset of singularities in solutions starting from perfectly smooth initial data .
such a phenomenon , usually called `` blowup '' , has been a subject of intensive studies in many fields ranging from fluid dynamics to general relativity .
whether or not the blowup can occur for a given nonlinear evolution equation is the central mathematical question which , from the physical point of view , has a direct bearing on our understanding of the limits of validity of the corresponding model .
unfortunately , this is often a difficult question .
two famous examples for which the answer is not known are the navier - stokes equation and the einstein equations . once the existence of blowup is established for a particular equation , many further questions come up , such as : when and where does the blowup occur ? what is the character of blowup and is it universal ? can a solution be continued past the singularity ?
in this paper we consider these questions for the yang - mills ( ym ) equations in higher dimensions . in the physical @xmath0 dimensions , where the ym equations are the basic equations of gauge theories describing the weak and strong interactions of elementary particles
, it is known that no singularities can form .
this was shown by eardley and moncrief @xcite who proved that solutions starting from smooth initial data remain smooth for all future times .
the motivation for studying the ym equations in unphysical @xmath1 dimensions for @xmath2 is twofold ( and unrelated to the latest fashion of doing physics in extra dimensions ) . from the mathematical point of view
, it is the obvious thing to ask how the property of global regularity depends on the dimension of the underlying spacetime and whether singularities can form in @xmath1 dimensions for @xmath2 .
however , there is also a less evident physical reason which is motivated by the hope that by understanding the problem of singularity formation for the ym equations one might get insight into the analogous , but much more difficult , problem in general relativity . from this viewpoint in which the ym equations are considered as a toy model for the einstein equations
it is essential that these two equations belong to the same criticality class .
let us recall that the criticality class is defined as the degree @xmath3 in the homogeneous scaling of energy @xmath4 under dilations @xmath5 .
the classification of equations into subcritical ( @xmath6 ) , critical ( @xmath7 ) , and supercritical ( @xmath8 ) is a basis of the heuristic meta - principle according to which subcritical equations are globally regular , while supercritical equations may develop singularities for some ( large ) initial data @xcite . for the ym equations we have @xmath9 , while for the einstein equations @xmath10 .
therefore , the ym equations in @xmath11 have the same criticality , @xmath12 , as the einstein equations in the physical dimension .
another way of seeing this is to note that in @xmath11 the dimension of the ym coupling constant @xmath13=m^{-1 } l^{d-4}$ ] ( in @xmath14 units ) is the same as the dimension of the physical newton s constant @xmath15=m^{-1 } l$ ] .
for the reason just explained , the main body of this paper is focused on the lowest super - critical dimension @xmath11 . in section 3
we show that in this case there exists a countable family of regular ( by regularity we mean analyticity inside the future light cone ) spherically symmetric self - similar solutions labelled by a nonnegative integer @xmath16 ( a nodal number ) .
next , using linear stability analysis we show in section 4 that the number of unstable modes around a given solution is equal to its nodal number .
the role of self - similar solutions in the dynamical evolution is studied in section 5 , where we show that : i ) the @xmath17 solution determines a universal asymptotics of singularity formation for solutions starting from generic `` large '' initial data ; ii ) the @xmath18 solution plays the role of a critical solution sitting at the threshold of singularity formation .
the latter is in many respects similar to the critical behaviour at the threshold of black hole formation in gravitational collapse . in both cases the threshold of singularity ( or black hole )
formation can be identified with the codimension - one stable manifold of a self - similar solution with exactly one unstable mode .
these similarities are discussed in detail in section 6 .
we consider also the cauchy problem for the ym equations in @xmath19 . despite intensive studies of this borderline case ,
the problem of global existence is open . in section 7
we describe numerical simulations which , in combination with analytic results , strongly suggest that large - energy solutions do blow up .
we show that the process of singularity formation is due to concentration of energy and proceeds via adiabatic shrinking of the instanton solution . at the end , a recent attempt of determining the asymptotic rate of shrinking is sketched .
we remark that there are close parallels between ym equations in @xmath1 dimensions and wave maps in @xmath20 dimensions @xcite .
indeed , many of the phenomena described here are mirrored for the equivariant wave maps into spheres in three @xcite and two @xcite spatial dimensions .
sections 5 and 7 of this survey are based on joint work with z. tabor @xcite .
the material of sections 3 and 4 is new .
we consider yang - mills ( ym ) fields in @xmath1 dimensional minkowski spacetime ( in the following latin and greek indices take the values @xmath21 and @xmath22 respectively ) .
the gauge potential @xmath23 is a one - form with values in the lie algebra @xmath24 of a compact lie group @xmath25 . in terms of the curvature
@xmath26 $ ] the action is @xmath27 where @xmath28 is the gauge coupling constant .
hereafter we set @xmath29 .
the ym equations derived from ( [ action ] ) are @xmath30 = 0.\ ] ] as written , this equation is underdetermined because of the gauge invariance @xmath31 where @xmath32 is an arbitrary function with values in @xmath25 . in order to correctly formulate the cauchy problem for equation ( [ ymeq ] ) , one must impose additional conditions which fix this gauge ambiguity .
we shall not discuss this issue here because in the spherically symmetric ansatz , to which this paper is restricted , the gauge is fixed automatically . for simplicity , we take here @xmath33 so the elements of @xmath34 can be considered as skew - symmetric @xmath35 matrices and the lie bracket is the usual commutator . assuming the spherically symmetric ansatz @xcite @xmath36 the ym equations reduce to the scalar semilinear wave equation for the magnetic potential @xmath37 @xmath38 where @xmath39 is the radial laplacian in @xmath40 dimensions .
the central question for equation ( [ wave ] ) is : can solutions starting from smooth initial data @xmath41 become singular in future ? as mentioned above , in the physical @xmath42 dimensions eardley and moncrief answered this question in the negative @xcite .
however , simple heuristic arguments indicate that the property of global regularity enjoyed by the ym equations in @xmath42 might break down in higher dimensions . in order to see why the global behaviour of solutions is expected to depend critically on the dimension @xmath43
, we recall two basic facts .
the first fact is the conservation of ( positive definite ) energy @xmath44 where the coefficient @xmath45 follows from the integration over the angles and taking the trace .
the second fact is scale - invariance of the ym equations : if @xmath46 is a solution of ( [ ymeq ] ) , so is @xmath47 , or equivalently , if @xmath37 is a solution of ( [ wave ] ) , so is @xmath48 . under this scaling the energy scales as @xmath49 ,
hence the ym equations are subcritical for @xmath50 , critical for @xmath19 , and supercritical for @xmath51 . in the subcritical case ,
shrinking of solutions to arbitrarily small scales costs infinite amount of energy , so it is forbidden by energy conservation . in other words ,
transfer of energy to arbitrarily high frequencies is impossible and consequently the cauchy problem should be well posed in the energy norm .
this important fact was proved in @xmath42 by klainerman and machedon @xcite , who thereby strengthened the result of eardley and moncrief . in the supercritical case , shrinking of solutions might be energetically favourable and consequently singularities are anticipated .
in fact , we shall show below that singularities do form already in the lowest supercritical dimension @xmath11 . in the critical dimension
@xmath19 the problem of singularity formation is more subtle because the scaling argument is inconclusive .
in order to set the stage for the discussion of singularity formation we first need to analyze in detail the structure of self - similar solutions of equation ( [ wave ] ) . as we shall see , these solutions play a key role in understanding the nature of blowup . by definition ,
self - similar solutions are invariant under dilations @xmath52 , hence they have the form @xmath53 where a positive constant @xmath54 , clearly allowed by the time translation invariance , is introduced for later convenience . note that for a self - similar solution we have @xmath55 hence the solution becomes singular at the center when @xmath56 ( there is no blowup in the first derivative because regularity demands that @xmath57 ) . thus , each self - similar solution @xmath58 provides an explicit example of a singularity developing in finite time from smooth initial data . substituting the ansatz ( [ ssansatz ] ) into ( [ wave ] ) one obtains the ordinary differential equation @xmath59
as explained in the introduction , because of the expected connections with the einstein equations , we are mainly interested in the lowest super - critical dimension @xmath11 . in this case equation ( [ sseq ] ) reduces to @xmath60 although the similarity coordinate @xmath61 is natural in the discussion of singularity formation , it has a disadvantage of not covering the region @xmath62 , in particular it does not extend to the future light cone of the point @xmath63 .
for this reason we define a new coordinate @xmath64 which covers the whole spacetime : the past and the future light cones are located at @xmath65 and @xmath66 , respectively ; while the center @xmath67 corresponds to @xmath68 ( for @xmath69 ) and @xmath70 ( for @xmath62 ) . in terms of @xmath71 equation ( [ sseq5 ] ) becomes @xmath72 we first consider this equation inside the past light cone , that is for @xmath73 and impose the boundary conditions @xmath74 which follow from the demand of smoothness at the endpoints .
as we shall see below , once a solution to this boundary value problem is constructed , its extension beyond the past light cone can be easily done . to show that equation ( [ main ] ) admits solutions satisfying ( [ bc ] ) we shall employ a shooting technique .
the main idea of this method is to replace the boundary value problem by the initial value problem with initial data imposed at one of the endpoints and then adjusting these data so that the solution hits the desired boundary value at the second endpoint . in the case at hand
we shall shoot from @xmath65 towards infinity . substituting a formal power series expansion about @xmath65 into ( [ main ] )
one finds the asymptotic behaviour @xmath75 where @xmath76 is a free parameter determining uniquely the whole series . in the following a solution of equation ( [ main ] ) starting at @xmath65 with the asymptotic behaviour ( [ aorbit ] )
will be called an @xmath76-orbit . without loss of generality
we may assume that @xmath77 .
we claim that there is a countable set of values @xmath78 for which the @xmath79-orbits exist for all @xmath80 and have the desired asymptotics at infinity ( such orbits will be called connecting ) .
the proof consists of several steps .
0.2 cm _ step 1 ( local existence)_. first , we need to show that @xmath76-orbits do in fact exist , that is , the series ( [ aorbit ] ) has a nonzero radius of convergence . since the point @xmath65 is singular , this fact does not follow from standard theorems . fortunately , in @xcite breitelohner , forgcs , and maison have derived the following result concerning the behaviour of solutions of a system of ordinary differential equations near a singular point : 0.1 cm _ theorem [ bfm ] . consider a system of first order differential equations for @xmath81 functions @xmath82 and @xmath83 @xmath84 where constants @xmath85 and integers @xmath86 and let @xmath87 be an open subset of @xmath88 such that the functions @xmath89 and @xmath24 are analytic in the neighbourhood of @xmath90 for all @xmath91 .
then there exists an @xmath16-parameter family of solutions of the system ( [ bfm ] ) such that @xmath92 where @xmath93 and @xmath94 are defined for all latexmath:[$c\in c , 0.1 cm we shall make use of this theorem to prove the local existence of @xmath76-orbits . in order to put equation ( [ main ] ) into the form ( [ bfm2 ] ) we define the variables @xmath96 and get @xmath97}{2+y}.\ ] ] since the function @xmath98 is analytic near @xmath99 for any @xmath100 and @xmath101 , according to the bfm theorem there exists a one - parameter family of local solutions such that @xmath102 transforming ( [ loc ] ) back to the original variables we obtain the behaviour of @xmath76-orbits . 0.2 cm _
step 2 ( a priori global behaviour)_. it follows immediately from ( [ main ] ) that for @xmath103 a solution can not have a maximum ( resp .
minimum ) for @xmath104 ( resp .
@xmath105 ) . thus ,
once the solution leaves the strip @xmath106 , it can not reenter it ( actually , such a solution becomes singular for a finite @xmath71 ) .
it is also clear that as long as @xmath106 the solution can not go singular . to derive the asymptotics at infinity of @xmath76-orbits that stay in the strip @xmath106 we shall make use of the following functional @xmath107 for solutions of equation ( [ main ] ) we have @xmath108 so @xmath109 is monotone increasing .
now , we shall show that solutions satisfying @xmath106 for all @xmath110 tend to @xmath111 as @xmath112 . to see this , first notice that for such solutions @xmath113 must be negative
because if @xmath114 for some @xmath115 then @xmath116 is strictly positive for @xmath117 so the solution must leave the strip @xmath106 in finite time .
since @xmath118 and @xmath119 , it follows that @xmath113 has a nonpositive limit at infinity which in turn implies by ( [ q ] ) that @xmath120 and by ( [ q ] ) that @xmath121 exists . by lhpital s rule we have @xmath122 and using ( [ main ] ) again
, we get that @xmath123 equals @xmath124 or @xmath125 .
the latter is impossible because then @xmath126 is a global minimum contradicting the fact that @xmath113 increases .
thus , @xmath127 .
0.2 cm _ step 3 _ ( i ) ( _ behaviour of @xmath76-orbits for small @xmath76 _ ) .
rescaling @xmath128 we get @xmath129 as @xmath130 , the solutions of this equation tend uniformly on compact intervals to the solution of the limiting equation @xmath131 with the same initial condition .
this equation can be solved explicitly but for the purpose of the argument it suffices to notice that its solution , call it @xmath132 , is oscillating at infinity .
since @xmath133 up to an arbitrarily large @xmath71 if @xmath76 is sufficiently small , it follows that the number of zeros of the solution @xmath134 tends to infinity as @xmath130 .
0.1 cm ( ii ) ( _ behaviour of @xmath76-orbits for large @xmath76_. ) we rescale the variables , setting @xmath135 , @xmath136 to get @xmath137 as @xmath138 , the solutions of this equation tend uniformly on compact intervals to the solution @xmath139 of the limiting equation @xmath140 .
thus , @xmath141 for large @xmath76 and therefore the @xmath76-orbit crosses @xmath142 for a finite @xmath71 .
0.2 cm _ step 4 ( shooting argument)_. we define the set @xmath143 we know from step 3 that the set @xmath144 is nonempty ( because the @xmath76-orbits with large @xmath76 belong to it ) and bounded below ( because the @xmath76-orbits with small @xmath76 do not belong to it ) .
thus @xmath145 exists . the solution @xmath146 can not cross the line @xmath142 at a finite @xmath71 because the same would be true for nearby solutions , violating the definition of @xmath147 .
thus , @xmath148 for all @xmath71 and hence , by step 2 , @xmath149 .
this completes the proof of existence of the nodeless self - similar solution @xmath150 .
next , let us consider the solution with @xmath151 for small @xmath152 .
by the definition of @xmath147 there must be a point @xmath153 where this solution attains a positive local maximum @xmath154 and since no minima are possible for @xmath155 , it follows that there must be a point @xmath156 where @xmath157 .
we shall show that @xmath158 provided that @xmath159 is sufficiently small . as argued above this
would imply that the solution @xmath134 leaves the strip @xmath106 via @xmath160 . from ( [ q ] )
we have @xmath161 in order to estimate the last integral note that for @xmath117 @xmath162 so @xmath163 .
substituting this into ( [ bla1 ] ) one gets @xmath164 the right hand side of this inequality is equal to @xmath165 for @xmath166 so , by continuity , it remains strictly positive for @xmath167 near @xmath168 . by taking a sufficiently small @xmath159
we can have @xmath167 arbitrarily close to @xmath168 , hence @xmath169 which proves that @xmath76-orbits with @xmath151 have exactly one zero .
this means that the set @xmath170 increases up to some @xmath153 where it attains a positive local maximum @xmath154 and then decreases monotonically up to some @xmath171 where @xmath172 is nonempty .
let @xmath173 . by step 3 ,
@xmath174 exists and is strictly positive .
using the same argument as above we conclude that the @xmath174-orbit must stay in the region @xmath106 for all @xmath71 , hence @xmath175 .
this completes the proof of existence of the self - similar solution @xmath176 with exactly one zero .
the subsequent connecting orbits are obtained by induction .
we conclude that there exists a countable family of self - similar solutions @xmath177 indexed by the integer @xmath178 @xmath16 which counts the number of zeros for @xmath103 .
0.15 cm _ remark .
_ since the sequence @xmath78 is decreasing and bounded below by zero , it has a nonnegative limit @xmath179 .
if @xmath180 , then the @xmath181-orbit can not leave the region @xmath106 for a finite @xmath71 ( because the set of such orbits is clearly open ) hence it must be a connecting orbit with some finite number of zeros . but
this contradicts the fact that the number of zeros of @xmath79-orbits increases with @xmath16 .
hence , @xmath182 .
this implies that for any finite @xmath71 , @xmath177 goes to zero when @xmath183 .
0.15 cm we remark that the existence of the solution @xmath184 was first shown by cazenave , shatah , and tahvildar - zadeh @xcite via a variational method .
0.2 cm the shooting technique is not only a powerful analytical tool ; it is also an efficient numerical method of solving two - point boundary value problems .
the numerical results produced by this method are shown in table 1 and figure 1 .
@xmath185 .the shooting parameters of solutions @xmath186 for @xmath187 . [ cols="^,^,^,^,^,^,^",options="header " , ] @xmath185 we point out that although @xmath7 is not a genuine eigenvalue , it is distinguished from the strictly positive part of the continuous spectrum by the fact that the corresponding non - square - integrable pseudo - eigenfunction , called the zero mode , is subdominant at infinity .
the existence of the zero mode is due to the time translation symmetry , or in other words , the freedom of shifting the blowup time @xmath54 in ( [ ssansatz ] ) . to see this ,
consider the self - similar solution with a shifted blowup time @xmath188 , where @xmath189 . in terms
of the original similarity variables @xmath190 and @xmath191 , we have @xmath192 hence the perturbation generated by shifting the blowup time corresponds to @xmath7 and has the form @xmath193 an alternative way of deriving this result is to take @xmath194 , which is ( [ main ] ) reexpressed in terms of @xmath195 , differentiate it and compare with ( [ schr ] ) . since by construction the solution @xmath196 has @xmath16 extrema
, it follows from ( [ zeromode ] ) that the zero mode @xmath197 has @xmath16 nodes .
this implies , by the standard result from sturm - liouville theory , that the potential @xmath198 has exactly @xmath16 negative eigenvalues , in agreement with the numerical results shown in table 2 .
we conclude that the self - similar solution @xmath186 has exactly @xmath16 unstable modes ( apart from the unphysical zero mode ) .
in particular , the fundamental solution @xmath184 is linearly stable , which makes it a candidate for the attractor .
the rest of this section is a digression concerning a striking regularity which an acute reader might have already noticed in table 2 . namely , the third column of table 2 indicates that for each @xmath199 the first eigenvalue below the continuous spectrum @xmath200 is equal to @xmath201 ( with the numerical accuracy of ten decimal places ) ! this puzzling numerical fact is calling for an explanation .
clearly , it has something to do with the particular form of the nonlinearity since , for instance , the analogous problem for self - similar wave maps from @xmath0 dimensional minkowski spacetime into the 3-sphere does not have this property @xcite .
we suspect that the problem has some hidden symmetry , yet we can not exclude a possibility that the numerics is misleading and the eigenvalues @xmath202 are not precisely equal but their splitting is beyond the numerical resolution .
some insight into this puzzle can be gained by analyzing the limiting case @xmath203 .
recall that @xmath196 tends to zero for any @xmath204 as @xmath205 , hence the sequence of potentials @xmath198 has the following nonuniform limit @xmath206 for the limiting potential @xmath207 the schrdinger equation ( [ schr ] ) can be solved exactly .
the solution that is admissible at infinity ( which as before is the limit - point ) is given by the associated legendre function of the first kind @xmath208 here @xmath209 is one of the roots of @xmath210 ; the second root gives the same solution because @xmath211 with real @xmath212 is real and @xmath213 .
as @xmath214 , the solution ( [ uinf ] ) behaves as @xmath215 so it is always admissible , independently of @xmath3 .
this means that @xmath216 is the limit - circle point and therefore in order to have a well - defined self - adjoint problem we need to impose an additional boundary condition . in the language of spectral theory such a condition is called a self - adjoint extension .
the continuous part of the spectrum is the same for all self - adjoint extensions but the eigenvalues do depend on the choice . in our case
the self - adjoint extension amounts to fixing @xmath217 - the phase of oscillations of the eigenfunctions for @xmath214 .
the natural choice is to require that the eigenfunctions oscillate with the same phase as the zero mode , that is @xmath218 , or equivalently @xmath219 note that under this condition the following diagram commutes @xmath220 substituting ( [ uinf ] ) into ( [ exten ] ) and using the asymptotic expansion ( for real @xmath212 ) @xmath221 we obtain the quantization condition for the eigenvalues @xmath222 this transcendental equation has infinitely many roots which for @xmath223 can be obtained only numerically ( see the last row in table 2 ) .
however , for @xmath224 the exact solution is @xmath225 because ( accidentally ? ) @xmath226 , as can be readily verified using four times the identity @xmath227 .
thus , we showed that the least negative eigenvalue of the limiting potential is equal to @xmath201 .
although this analysis does not resolve the original puzzle why all @xmath228 are equal to @xmath229 , it shows at least that @xmath229 is the accumulation point of this sequence .
the asymptotic distribution of eigenvalues for @xmath230 can be derived from ( [ quant ] ) by using the formula for the asymptotic behaviour of the gamma function for large @xmath231 @xmath232 which yields @xmath233 applying this to ( [ quant ] ) one gets @xmath234 this formula was useful is providing an initial guess in the the numerical root finding procedure for equation ( [ quant ] ) for large @xmath235 .
having learned about self - similar solutions , we are now prepared to understand the results of numerical studies , first reported in @xcite , of the cauchy problem for the ym equation in five space dimensions @xmath236 the main goal of these studies was to determine the asymptotics of blowup .
our numerical simulations were based on finite difference methods combined with adaptive mesh refinement .
the latter were instrumental in resolving the structure of singularities developing on vanishingly small scales .
we stress that a priori analytical insight into the problem , in particular the knowledge of self - similar solutions was very helpful in interpreting the numerical results .
we solved equation ( [ wave5 ] ) for a variety of initial conditions interpolating between small and large data .
a typical example of such initial data is a gaussian ( ingoing or time - symmetric ) of the form @xmath237,\ ] ] with adjustable amplitude @xmath238 and fixed parameters @xmath239 and @xmath240 .
the global behaviour of solutions is qualitatively the same for all families of initial data and depends critically on the amplitude @xmath238 ( or any other parameter which controls the `` strength '' of initial data ) . for small amplitudes
the solutions disperse , that is the energy is radiated away to infinity and in any compact region the solution approaches the vacuum solution @xmath241 .
this is in agreement with general theorems on global existence for small initial data @xcite .
heuristically , this follows from the fact the for a small amplitude the nonlinearity is dominated by the dispersive effect of the linear wave operator . for large amplitudes
we observe the development of two clearly separated regions : an outer region where the evolution is very slow and a rapidly evolving inner region where the solution attains a kink - like shape which shrinks in a self - similar manner to zero size in a finite time @xmath54 .
the kink is , of course , nothing else but the self - similar solution @xmath242 ( rather than @xmath71 ) and abuse the notation by writing @xmath243 to denote @xmath244 . ]
( see figure 3 ) .
we summarize these findings in the following conjecture : _ solutions of equation ( [ wave5 ] ) corresponding to sufficiently large initial data do blow up in finite time in the sense that @xmath245 diverges as @xmath246 for some @xmath247 .
the universal asymptotic profile of blowup is given by the stable self - similar solution : _
@xmath248 we think that the basic mechanism which is responsible for the observed asymptotic self - similarity of blowup can be viewed as the convergence to the lowest `` energy '' configuration . to see this ,
let us rewrite ( [ wave5 ] ) in terms of the similarity variable @xmath61 and the slow time @xmath249 to get @xmath250 in this way the problem of blowup was converted into the problem of asymptotic behaviour of solutions for @xmath251 .
the natural `` energy '' functional associated with this problem is @xmath252 @xmath253 has a minimum at the self - similar solution @xmath184 and saddle points with @xmath16 unstable directions at solutions @xmath186 with @xmath199 . since the wave equation ( [ wave5eta ] ) contains a damping term reflecting an outward flux of energy through the past light cone of the singularity , we suspect ( but can not prove ) that @xmath253 decreases with time .
if so , it is natural to expect that solutions will tend asymptotically to the minimum of @xmath253 .
we already know that solutions with small data disperse and solutions with large data blow up .
the question is what happens in between . using bisection
, we found that along each interpolating family of initial data there is a threshold value of the parameter , say the amplitude @xmath254 , below which the solutions disperse and above which a singularity is formed .
the evolution of initial data near the threshold was found to go through a transient phase which is universal , i.e. the same for all families .
this intermediate attractor was identified as the self - similar solution @xmath255 .
having gone through this transient phase , at the end the solutions leave the intermediate attractor towards dispersal or blowup .
this behaviour is shown in figure 4 for the time - symmetric initial data of the form ( [ gauss ] ) .
0.15 cm the universality of the dynamics at the threshold of singularity formation can be understood heuristically as follows . as we showed above
, the self - similar solution @xmath255 has exactly one unstable mode in other words the stable manifold of this solution has codimension one and therefore generic one - parameter families of initial data do intersect it .
the points of intersection correspond to critical initial data that converge asymptotically to @xmath255 .
the marginally critical data , by continuity , initially remain close to the stable manifold and approach @xmath255 for intermediate times but eventually are repelled from its vicinity along the one - dimensional unstable manifold ( see figure 5 ) . according to this picture the universality of the nearly critical dynamics follows immediately from the fact that the same unstable mode dominates the evolution of all solutions .
more precisely , the evolution of marginally critical solutions in the intermediate asymptotics can be approximated as @xmath256 where @xmath257 is the single unstable mode with the eigenvalue @xmath225 .
the small constant @xmath258 , which is the only vestige of the initial data , quantifies an admixture of the unstable mode for precisely critical data @xmath259 .
the time of departure from the intermediate attractor is determined by the time @xmath260 in which the unstable mode grows to a finite size , i.e. , @xmath261 .
using @xmath262 and substituting @xmath225 , we get @xmath263 .
various scaling laws can be derived from this .
for example , consider solutions with marginally sub - threshold amplitudes @xmath264 .
for such solutions the energy density @xmath265 initially grows at the center , attains a maximum at a certain time @xmath266 and then drops to zero . substituting ( [ inter ] ) into ( [ density ] )
we get that @xmath267 , and hence @xmath268 .
the behaviour of solutions near the threshold of singularity formation described above shares many features with critical phenomena at the threshold of black hole formation in gravitational collapse . to explain these similarities
, we now briefly recall the phenomenology and heuristics of the critical gravitational collapse . consider a spherical shell of matter and let it collapse under its own weight .
the dynamics of this process , modelled by einstein s equations , can be understood intuitively in terms of the competition between gravitational attraction and repulsive internal forces ( due , for instance , to kinetic energy of matter or pressure ) .
if the initial configuration is dilute , then the repulsive forces `` win '' and the collapsing matter will rebound or implode through the center , and eventually will disperse .
on the other hand , if the density of matter is sufficiently large , some fraction of the initial mass will form a black hole .
critical gravitational collapse occurs when the attracting and repulsive forces governing the dynamics of this process are almost in balance , or in other words , the initial configuration is near the threshold of black hole formation .
the systematic studies of critical gravitational collapse were launched in the early nineties by the seminal paper by choptuik @xcite in which he investigated numerically the collapse of a self - gravitating massless scalar field .
evolving initial data fine - tuned to the border between no - black - hole and black - hole spacetimes , choptuik found the following unforseen phenomena near the threshold : + ( i ) universality : all initial data which are near the black hole threshold go through a universal transient period in their evolution during which they approach a certain intermediate attractor , before eventually dispersing or forming a black hole .
this universal intermediate attractor is usually referred to as the critical solution .
+ ( ii ) discrete self - similarity : the critical solution is discretely self - similar , that is it is invariant under dilations by a certain fixed factor @xmath269 called the echoing period .
+ ( iii ) black - hole mass scaling : for initial data that do form black holes , the masses of black holes satisfy the power law @xmath270 where @xmath159 is the distance to the threshold and @xmath271 is a universal ( i.e. , the same for all initial data ) critical exponent . thus , by fine tuning to the threshold one can make an arbitrarily tiny black hole .
put differently , there is no mass gap at the transition between black - hole and no - black - hole spacetimes . what choptuik found for the scalar field , has been later observed in many other models of gravitational collapse , although the symmetry of the critical solution itself was found to depend on the model : in some cases the critical solution is self - similar ( continuously or discretely ) , while in other cases the critical solution is static ( or periodic ) . in the latter case
black hole formation turns on with finite mass .
these two kinds of critical behaviour are referred to as the the type ii or type i criticality , respectively , to emphasize the formal analogy with second and first order phase transitions in statistical physics .
we refer the interested reader to @xcite for an excellent review of the growing literature on critical gravitational collapse .
the present understanding of critical behaviour in gravitational collapse is based on the same phase space picture as in figure 5 , that is , it is associated with the existence of a critical solution with exactly one unstable mode .
this picture leads to some quantitative predictions .
in particular , in the case of type ii critical collapse , an elementary dimensional analysis shows that the critical exponent @xmath271 in the power law @xmath272 is a reciprocal of the unstable eigenvalue of the critical solution . by now
, the similarities between type ii critical gravitational collapse and the dynamics at the threshold of singularity formation in the @xmath273 ym equations should be evident .
this analogy , together with similar results for wave maps in @xmath0 dimensions @xcite , @xcite , shows that the basic properties of critical collapse , such as universality , scaling , and self - similarity , first observed for einstein s equations , actually have nothing to do with gravity and seem to be robust properties of supercritical nonlinear wave equations .
the obvious advantage of toy models , such as the one presented in this paper , is their simplicity which allowed to get a much better analytic grip on critical phenomena than in the case of einstein s equations ; in particular , it was possible to prove existence of the critical solution .
the only characteristic property of type ii critical collapse which so far has not found in simpler models ( besides , of course , the absence of black holes which are replaced by singularities ) is discrete self - similarity of the critical solution
. it would be very interesting to design a toy model which exhibits discrete self - similarity at the threshold for singularity formation because this could give us insight into the origin of this mysterious symmetry .
in this section we consider the cauchy problem for the ym equation in four space dimensions @xmath274 we begin by recalling some facts concerning equation ( [ wave4 ] ) which are be important in understanding the dynamics of singularity formation .
first , we note that , in contrast to @xmath11 , there are no smooth self - similar solution in @xmath19 .
this follows from the fact that in @xmath43 dimensions the local solutions of equation ( [ sseq ] ) near the past light cone behave as @xmath275 , hence they are not smooth if @xmath43 is even ( in particular , they are not differentiable in @xmath19 ) .
although such singular self - similar solutions do exist , they can not develop from smooth initial data and therefore they are not expected to participate in the dynamics .
second , @xmath19 is the critical dimension in the sense that the energy ( [ energy ] ) does not change under scaling .
this means that , even though the model is scale invariant , a nontrivial finite energy static solution may exist the ym coupling constant @xmath276 provides the scale of energy . ] .
in fact , such a static solution is well known @xmath277 this is the instanton in the four - dimensional euclidean ym theory . of course , by reflection symmetry , @xmath278 is also the solution .
since the model is scale invariant , the solution @xmath279 generates an orbit of static solutions @xmath280 , where @xmath281 . to analyze the linear stability of the instanton , we insert @xmath282 into ( [ wave4 ] ) and linearize . in this way
we get the eigenvalue problem ( the radial schrdinger equation ) @xmath283 this problem has a zero eigenvalue @xmath284 which follows from scale invariance .
the corresponding eigenfunction ( so called zero mode ) is determined by the perturbation generated by scaling @xmath285 since the zero mode @xmath286 has no nodes , it follows by the standard result from sturm - liouville theory that there are no negative eigenvalues , and _ eo ipso _ no unstable modes around @xmath279 .
thus , the instanton is marginally stable .
note that the zero eigenvalue lies at the bottom of the continuous spectrum @xmath287 , hence there is no spectral gap in the problem . after these preliminaries , we return to the discussion of the cauchy problem for equation ( [ wave4 ] ) . for small energies the solutions disperse , in agreement with general theorems . for large energies , at first sight
the global behaviour seems similar to the @xmath11 case as before , near the center the solution attains the form of a kink which shrinks to zero size .
however , this similarity is superficial because now the kink is not a self - similar solution ( as no such solution exists ) .
it turns out ( see figure 6 ) that the kink has the form of the _ scale - evolving _
instanton @xmath288 where a scaling factor @xmath289 is a positive function which tends to zero as @xmath290 .
we summarize these findings in the following conjecture : _ solutions of equation ( [ wave4 ] ) with sufficiently large energy do blow up in finite time in the sense that @xmath245 diverges as @xmath246 for some @xmath247 .
the universal asymptotic profile of blowup is given by the instanton .
more precisely , there exists a positive function @xmath291 for @xmath292 such that _
the key question which is left open in this conjecture is : what determines the evolution of the scaling factor @xmath289 ; in particular , what is the asymptotic behaviour of @xmath289 for @xmath290 ? numerical evidence shown in figure 7 suggests that the rate of blowup goes asymptotically to zero , that is ( @xmath294 ) @xmath295 but it seems very hard to determine an exact asymptotics of @xmath289 from pure numerics .
recently , an analytical approach to this problem has been suggested in @xcite .
below we sketch the main idea of this approach .
let @xmath296 be a manifold of rescaled instantons ( a one - dimensional center manifold ) .
assuming that a solution is in a neighbourhood of @xmath297 , we decompose it as @xmath298 here @xmath101 represents a small deviation of the solution from @xmath297 and @xmath299 is the collective coordinate on @xmath297 . to fix the splitting between these two parts we require that @xmath300 , where @xmath301 is the projection on @xmath297 .
plugging ( [ decomp ] ) into ( [ wave4 ] ) we get ( @xmath302 ) @xmath303 where @xmath304 is the linear perturbation operator about the instanton @xmath305 and @xmath306 it is clear from ( [ main2 ] ) that @xmath307 , hence for @xmath101 to decay to zero as @xmath290 , the rate of blowup must go to zero as well .
we stress this point to emphasize that the linear evolution of @xmath289 , predicted for example by the geodesic approximation , is inconsistent with conjecture 2 .
next , by projecting equation ( [ main2 ] ) on @xmath297 and in the orthogonal direction , we get a coupled system consisting of a nonhomogeneous wave equation for @xmath101 and an ordinary differential equation for @xmath299 . solving the first equation for @xmath101 and plugging the result into the second equation , we obtain in the lowest order the following modulation equation @xmath308 from this we get the leading order asymptotics for @xmath309 @xmath310 as shown in figure 7 this result is in rough agreement with numerics .
there are many possible sources of the apparent discrepancy . on the numerical side
there are discretization errors , an error in estimating the blowup time , or errors in computing @xmath299 from the data . on the analytical side
, there might be corrections to ( [ sigal ] ) coming from the bounded region expansion and , more importantly , from the far field behaviour ) is not quite straightforward because of the presence of infrared divergencies which need to be regularized . ] .
finally , and in our opinion most likely reason of discrepancy is that the solution shown in figure 7 has not yet reached the truly asymptotic regime and consequently the higher order corrections to formula ( [ sigal ] ) are still significant .
the issue of blowup rate is closely related to the problem of energy concentration in the singularity . to explain this
, we define the kinetic and the potential energies at time @xmath69 inside the past light cone of the singularity @xmath311 substituting ( [ adiabatic ] ) into ( [ energy_cone ] ) we obtain @xmath312 assuming ( [ rate ] ) , this implies that @xmath313 thus , the energy equal to the energy of the instanton gets concentrated in the singularity .
this means that in the process of blowup the excess energy must be radiated away from the inner region as the solution converges to the instanton .
it is worth pointing out that the concentration of energy is a necessary condition for blowup in the critical dimension . to see this ,
suppose that the solution blows up at time t and assume for contradiction that @xmath314 .
then , by choosing a sufficiently small @xmath152 we can have @xmath315 arbitrarily small .
this implies , by causality and global existence for small energy data , that the solution exists globally in time , contradicting the assumption . for radial equations of the type @xmath316 a stronger result
was proved by struwe @xcite who basically showed that if the solution of ( [ struwe ] ) blows up , then it must do so in the manner described in conjecture 2 . in this sense formation of singularities
is intimately tied with the existence of a static solution .
finally , we address briefly the issue of the threshold of singularity formation . using the technique described in section 5 , along each interpolating family of initial data
we can determine the critical point separating blowup from dispersal .
however , in contrast to the @xmath11 case , we see no evidence for the existence of an intermediate attractor in the evolution of nearly critical data .
this fact , together with a similar result for @xmath317 dimensional wave maps @xcite suggests that in the critical dimension the transition between blowup and dispersal is not governed by any critical solution .
we suspect that the evolution of precisely critical initial data still has the form ( [ adiabatic ] ) but the dynamics of the scaling factor is different than in ( [ sigal ] ) .
this belief is based on the fact that in the evolution of marginally critical initial data we can clearly distinguish a transient phase during which @xmath289 drops very quickly and then , either reaches a minimum and starts growing ( in the case of dispersal ) , or keeps decreasing with the `` normal '' rate ( in the case of blowup ) .
there are two main lessons that we wanted to convey in this survey .
the first lesson is that there are striking analogies between major evolution equations . in particular ,
the mechanism of blowup to a large extent is determined by the criticality class of the model .
these analogies can be used to get insight into hard problems ( such as singularity formation for einstein s equations ) by studying toy models which belong to the _ same _ criticality class .
this approach is in the spirit of general philosophy expressed by david hilbert in his famous lecture delivered before the international congress of mathematicians at paris in 1900 @xcite : _ `` in dealing with mathematical problems , specialization plays , as i believe , a still more important part than generalization .
perhaps in most cases where we seek in vain the answer to a question , the cause of the failure lies in the fact that problems simpler and easier than the one in hand have been either not at all or incompletely solved .
all depends , then , on finding out these easier problems , and on solving them by means of devices as perfect as possible and of concepts capable of generalization . '' _
the second lesson is concerned with the interplay between numerical and analytical techniques .
accurate and reliable numerical simulation of singular behaviour is difficult and hard to assess . in order to keep track of a singularity developing on exceedingly small spatio - temporal scales , one needs sophisticated techniques such as adaptive mesh refinement . for these techniques the convergence and error analysis are lacking so
extreme care is needed to make sure that the computed singularities are not numerical artifacts .
for this reason , in order to feel confident about numerics it is important to have some analytical information , like existence of self - similar solutions . without a theory ,
simulations alone do not provide ample evidence for the existence of a singularity .
we believe that the interaction between numerical and analytical techniques , illustrated here by the studies of blowup , will become more and more important in future as we begin to attack more difficult problems .
0.2 cm _ acknowledgment .
i am grateful to michael sigal and yu .
n. ovchinnikov for permission to announce here the results of yet unpublished joint work .
i thank zbislaw tabor for providing me with data for figure 7 .
i acknowledge the hospitality of the albert einstein institute for gravitational physics in golm and the erwin schrdinger institute for mathematical physics in vienna , where parts of this paper were produced . | this is a survey of recent studies of singularity formation in solutions of spherically symmetric yang - mills equations in higher dimensions .
the main attention is focused on five space dimensions because this case exhibits interesting similarities with einstein s equations in the physical dimension , in particular the dynamics at the threshold of singularity formation shares many features ( such as universality , self - similarity , and scaling ) with critical phenomena in gravitational collapse . the borderline case of four space dimensions
is also analyzed and the formation of singularities is shown to be intimately tied to the existence of the instanton solution . |
according to a nomenclature that has emerged in various physical and mathematical contexts , one refers to the so - called ter - martirosyan skornyakov ( henceforth tms ) operators as a distinguished class of quantum hamiltonians for systems of non - relativistic particles with two - body `` _ _ zero - range _ _ '' ( or `` _ _ contact _ _ '' , or `` _ _ point _ _ '' ) interaction .
this terminology stems from early works in nuclear physics , where it was the nucleon - nucleon coupling to be initially modelled as a `` contact '' interaction .
nowadays the typical experimental realisation is that of ultra - cold atom systems where , by feshbach resonance methods , the two - body scattering length is tuned to a magnitude that exceeds by many orders its nominal value , and the effective range of the interaction shrinks correspondingly to a very small scale , so that to an extremely good approximation the interaction can be considered to be of infinite scattering length and/or zero range . in section [ sec : history_tms ]
we will provide a more diffuse context and references .
informally speaking , tms hamiltonians are qualified by the two characteristics of acting as the @xmath0-body @xmath1-dimensional _ free _ hamiltonian on functions that are supported _ away _ from the `` coincidence hyperplanes '' @xmath2 , and of having a domain that consists of square - integrable functions @xmath3 , possibly with fermionic or bosonic exchange symmetry , which satisfy specific asymptotics when @xmath4 for some or for all particle couples @xmath5 .
this models an interaction supported only on the hyperplanes @xmath2 .
it is customary to refer to this ultra - violet asymptotics as the `` _ _ tms condition _ _ '' .
the explicit form for the tms condition has various versions ( see , e.g. , , , , , , or below ) , all essentially equivalent to each other . noticeably , such asymptotics emerge from different contexts and languages : on the one side the physical heuristics for an `` effective '' low - energy two - body scattering due to an interaction of very short range , on the other side the mathematical theory of self - adjoint extensions of symmetric operators on hilbert space .
this is a fascinating history of reciprocal influence and mutual inspiration between `` early days '' nuclear physics , modern condensed matter physics , and mathematical operator theory and self - adjoint extension theory .
section [ sec : history_tms ] below will partially survey it .
tms hamiltonians represent the modern operator - theoretic approach to multi - particle quantum systems with two - body point interaction , and have an intimate connection to the alternative approach based on energy quadratic forms .
they arise as natural effective models , based on stringent physical heuristics on the behaviour of the many - body wave - function when _ two _ particles come on top of each other . in many circumstances , however , which depend essentially on the mass of the particles and on possible additional symmetries of the system , a formal tms hamiltonian fails to be self - adjoint and each of its self - adjoint extensions accounts for a different behaviour of the system when _ three _ particles get closer and closer to the same point .
the question is then to identify these extensions ( if more than one ) and to study their stability and spectral properties .
this is even more so since quantum systems with zero - range interactions may exhibit two somewhat exotic phenomena , as compared with the case of ordinary finite range potentials : the so - called `` _ _ thomas effect _ _ '' , namely the emergence of an infinite discrete sequence of bound states with negative energy diverging to @xmath6 and eigenfunction collapsing onto the barycentre , and the `` _ _ efimov effect _ _ '' , that consists of an infinite sequence of bound states with negative energy arbitrarily close to zero and eigenfunctions extending on a larger and larger spatial scale , with a non - square - integrable limit .
the tms hamiltonians account for such phenomena as well . in this work
we put the emphasis on the mathematical scheme within which a two - body interaction modelled by means of a tms condition ( a _ physical _ requirement ) does correspond to a self - adjoint hamiltonian ( a _ mathematical _ constraint that ensures the well - posedness and the correct interpretation of the quantum model ) . here
is how the general problem is posed , as we shall elaborate further in the historical review of section [ sec : history_tms ] and then in the concrete settings of sections [ sec:2body - point ] and [ sec:2 + 1 ] * one starts ( mostly in @xmath7 dimensions ) with the operator @xmath8 obtained by _ restricting _ the @xmath0-body free hamiltonian to the regular wave - functions that are supported away from the coincidence hyperplanes @xmath9 .
additional partial or global symmetries for @xmath8 are possible , such as rotational symmetries , or exchange bosonic or fermionic symmetries for some or all of the @xmath0 particles .
@xmath8 is clearly densely defined , symmetric , and positive , and any its self - adjoint extension is naturally interpreted as a model for an interaction supported at @xmath10 . *
the self - adjoint extensions of @xmath8 are restrictions of @xmath11 .
thus , one first characterises the domain and the action of @xmath11 and then one selects a special class of restrictions of @xmath11 , which are obtained by reducing the domain of @xmath11 to only those functions @xmath12 satisfying , for some or all couples of variables @xmath13 , the condition that , if @xmath14 and @xmath15 is a regular change of variables , then @xmath16 in @xmath17 are the conjugate fourier variables to @xmath18 , the function @xmath19 depends on @xmath12 and on the considered couple @xmath5 , but not on @xmath20 , and the constant @xmath21 is prescribed and is @xmath20-independent too .
equation above is one version of the so - called ter - martirosyan skornyakov condition .
it is based on stringent physical heuristics ( see the discussion in section [ sec : history_tms ] and - below ) that allow one to interpret it as an interaction in the @xmath22-channel with _ zero range _ and @xmath23-wave scattering length equal to @xmath24 . *
the tms condition selects a `` physical '' extension @xmath25 of @xmath8 labelled by the conventional parametrisation @xmath26 .
the mathematical problem is then to recognise @xmath25 as a self - adjoint extension of @xmath8 , or to identify and classify its self - adjoint extensions , and then to investigate its spectral and stability properties .
the scheme above , which was made explicit for the first time by minlos @xcite in 1987 ( based on an old seminal but very concise work of minlos and faddeev @xcite in 1961 ) , has been since then the object of numerous investigations that we will quote in section [ sec : history_tms ] the general question of the self - adjoint realisation of @xmath25 and of its stability and spectral properties is still open .
an amount of partial information is available for special cases of @xmath0-body systems . in this work , beyond placing the problem into a historical perspective ( see section [ sec : history_tms ] ) , we discuss how the tms condition can be proved to be a _ self - adjointness condition _ , based on the extension theory specifically tailored for semi - bounded symmetric operators , as developed by kren , viik , and birman .
( we review the main results of this theory in appendix [ app : kvb ] and we refer to @xcite for a comprehensive discussion . ) the use of the kren - viik - birman theory is in fact at the basis of all recent studies on the self - adjoint realisation of the operator @xmath25 selected by the tms condition : in this context , it allows one to reduce the problem of the self - adjointness of @xmath25 on the hilbert space @xmath27 ( with possible symmetries ) to the problem of the self - adjointness of a suitable integral operator on the space of the functions @xmath19 s appearing in the asymptotics .
what we do here is on the first place to elaborate in full detail the precise application of the kren - viik - birman theory to the tms condition , with a discussion and through intermediate results that to our knowledge are not present in the literature .
further , we put for the first time the emphasis on the crucial difference between the tms condition as a _ point - wise _ identity , and the same condition interpreted as a suitable _ functional _ identity , as we now explain . in section
[ sec:2body - point ] , which contains the first group of our results , we discuss the tms condition for the simplest composite system possible , consisting of two particles with point interaction .
this is a well - studied quantum system that is completely understood within the standard self - adjoint extension theory a la von neumann .
first we re - obtain the well - known hamiltonian of the system solely by means of the kren - viik - birman theory , a procedure that we did not manage to find elsewhere in the literature .
this yields an alternative equivalent characterisation of the whole class of self - adjoint extensions of the `` away - from - hyperplanes '' free hamiltonian @xmath8 that represented the starting point of the analysis see step 1 of the general scheme above .
we then show that imposing the tms condition to functions in the domain of @xmath11 reproduces , for all possible values of the scattering length @xmath28 , _ all _ the self - adjoint extensions of @xmath8 .
in particular this shows that the tms condition is a self - adjointness condition . in section [ sec:2 + 1 ] , where the second group of our results is presented , we follow the same approach for a more complicated system consisting of three particles with point interaction we develop our explicit discussion for the so - called `` 2 + 1''-fermionic system , two identical fermions coupled with a third particle of different nature . exploiting again the general results of the kren - viik - birman theory
, we find that the tms condition is a self - adjointness condition , which selects a sub - class of extensions of @xmath11 , only if it is given as a suitable functional identity . as a generic point - wise identity ,
instead , the tms condition is in general not even a condition for a symmetric extension of @xmath8 .
furthermore , we show that the issue of the self - adjoint realisation of the tms condition does indeed boil down to the self - adjointness problem of a simpler integral operator ( that acts on the so - called `` space of charges '' ) as normally given for granted in the literature , _ but _ on different ( more regular ) functional spaces than those considered so far .
this brings us to the natural follow - up of the present analysis , which was in fact our original motivation and that we intend now to develop in a future work .
there are indeed not completely understood discrepancies in the literature between the ranges of the particle masses in which the tms condition for three - body systems is shown to be a self - adjointness condition , or instead to give rise to a symmetric extension of @xmath8 with its own family of self - adjoint extensions .
such discrepancies emerge between the operator - theoretic approach sketched above and an alternative approach through quadratic forms ( i.e. , the construction of a closed and semi - bounded quadratic form such that the function in the domain of self - adjoint operator that realises it display the tms asymptotics ) .
re - visiting the operator - theoretic approach in view of our present findings is likely to account for an explanation .
we briefly elaborate on this point in the final section [ sec : developments ]
in this section we present the historical emergence of tms operators in the physics and mathematics of point interactions .
it therefore should _ not _ be regarded as a complete review on the history of point interactions ! in the 1930 s quantum mechanics began to be applied to the newly observed nuclear phenomena . at first , the decrease by a factor @xmath29 from the atomic to the nuclear scale made it plausible to model the interaction among nucleons as a delta - like interaction . in 1932
wigner @xcite calculated that the nuclear forces interaction must be of very short range and very strong magnitude .
this led three years later first bethe and peierls @xcite and then thomas @xcite to describe the neutron - proton scattering by means of the two - body schrdinger equation in the approximation of a potential of very short range , an approach subsequently developed by fermi @xcite and breit @xcite with the introduction of the so - called `` delta pseudo - potential '' .
formulated in modern terms , the celebrated `` _ _ bethe - peierls contact condition _ _ '' , which is still today ubiquitous in many formal physical treatments , prescribes on the basis of physical heuristics that the wave - function @xmath3 of @xmath0 three - dimensional particles subject to a two - body zero - range interaction of scattering length @xmath30 among particles @xmath31 and @xmath32 behaves asymptotically as @xmath33 where the point - wise limit is meant as @xmath34 for some function @xmath35 of the variable @xmath36 in the hyperplane @xmath2 .
clearly , what makes this approximation appealing , and computationally advantageous , is its dependence on few parameters only ( the @xmath30 s ) , instead of the complete knowledge of the interaction .
while bethe and peierls had studied the problem of _ two _ low - energy nucleons with contact interaction and obtained for the two - body problem , thomas had considered the _ three - body _ problem showing that as the range of the two - body forces tends to zero the ground state of the three - body system can approach @xmath6 , even when the ground state energies of all two - body subsystems remain constant .
this effect , referred to since then as the `` _ _ thomas effect _ _ '' , was the first evidence that the deceptively simple three - body problem with zero - range interaction has a much richer ( and potentially much more complicated ) phenomenology than the analogous two - body problem .
the next extensive study of a system of _ three _ low - energy nucleons appeared some 20 years later , in 1955 , due to ter - martirosyan and skornyakov @xcite , two nuclear physicists who credited landau for the ideas they exploited .
they assumed that the bethe - peierls condition remains valid in _ each _ two - body channel and they used it as boundary condition for solving the eigenvalue problem for the three - body schrdinger equation with formal delta - like two - body potentials .
for brevity let us revisit their conclusion in the simplest case of spinless identical particles , for which they made use of standard centre - of - mass jacobi coordinates @xmath37 , @xmath38 and introduced a formal delta - like potential @xmath39 for wave - functions @xmath40 . by suitably expressing @xmath41 in terms of an auxiliary function @xmath42 ( here @xmath43 and @xmath44 are fourier conjugate variables ) , they found that ( in units @xmath45 and particle mass @xmath46 ) @xmath12 is a bound state of energy @xmath47 whenever @xmath48 is a solution to the integral equation @xmath49 where @xmath50 and @xmath28 is the @xmath23-wave scattering length in each two - body channel .
analogues to equation were later found for other systems with point interactions , among which , to mention those that have received the largest attention , three - body systems with different symmetries ( three distinguishable particles , two identical fermions plus a third particle of different nature , etc . ) , four - body systems with two distinct couples of identical fermions , and more generally @xmath51 systems with @xmath0 identical fermions of one type plus @xmath52 identical fermions of another type .
each equation of this class is today referred to as a `` _ _ ter - martirosyan skornyakov equation _ _ '' .
the subject proved soon to be worth a deeper understanding , despite the effectiveness of the description . on the one hand the manipulations of ter - martirosyan and skornyakov ( as well as the previous ones by bethe , peierls , thomas , fermi , and breit ) were rather formal for mathematical standards and called for a more rigorous justification .
on the other hand , an evidence of some sort of indeterminate physical description emerged when danilov @xcite in 1961 observed that the tms equation has a solution @xmath53 for arbitrary values of @xmath54 , with large momentum asymptotics @xmath55 where @xmath56 is an explicit universal constant and @xmath57 are two further constants that depend on @xmath54 . inspired by ideas ( mainly of gribov ) by which some additional `` experimental '' parameter that can not be computed using only two - body experimental data was needed for the full description of the three - body system , danilov proposed an ad hoc removal of this non - physical continuum of eigenvalues on @xmath58 by constraining the solutions to to have the form with @xmath59 for some additionally prescribed parameter @xmath60 .
@xmath61 was in some vague sense given the meaning of a three - body parameter , as opposite to @xmath62 in which is a parameter of the two - body problem for each couple of bosons . under the restriction ,
equation has a discrete and infinite set of solutions @xmath63 , with energies @xmath64 as @xmath65 according to the asymptotics @xmath66 a quantitative manifestation of the thomas effect . in modern mathematical terms , the phenomenon noted by danilov is understood as follows : the three - body point - interaction hamiltonian implicitly identified by ter - martirosyan and skornyakov by means of the condition for its eigenstates at given two - body scattering length @xmath67 , is _ not _ a self - adjoint operator and it admits a one - parameter family of self - adjoint extensions , labelled by @xmath60 ; for each @xmath61 , the corresponding self - adjoint hamiltonian has a countable discrete spectrum accumulating exponentially to @xmath6 with corresponding eigenfunctions that collapse onto the barycentre ; the union of the negative spectra of all such self - adjoint extensions is the whole negative real line . motivated by the scheme of ter - martirosyan and skornyakov for the three - body problem with point interaction and by danilov s observation , minlos and faddeev @xcite in the same year 1961 provided essentially the whole explanation above , including the asymptotics and , in the form of two beautiful short announcements , albeit with no proofs or further elaborations .
theirs can be considered as the beginning of the mathematics of quantum systems with zero - range interactions .
this is even more so because for the first time the problem was placed within a general mathematical framework , the theory of self - adjoint extensions of semi - bounded symmetric operator , that kren , viik , and birman had developed between the mid 1940 s and the mid 1950 s ( see appendix [ app : kvb ] ) .
a somewhat different approach characterised the start of the mathematical study of the _ two - body _ problem . in 1960 - 1961
, a few months before the works of minlos and faddeev on the three - body problem , berezin and faddeev @xcite published the first rigorous analysis of a three - dimensional model with two particles coupled by a delta - like interaction .
the emphasis was put in realising the formal hamiltonian @xmath68 as a self - adjoint extension of the restriction @xmath69 ( in the relative variable @xmath70 between the two particles ) .
working in fourier transform , they recognised that the latter operator has deficiency indices @xmath71 and they characterised the whole family @xmath72 of its self - adjoint extensions as the operators @xmath73 defined on the domain of @xmath74-functions @xmath75 such that , as @xmath76 , @xmath77 ( for a more direct comparison see in the following we have replaced here the parameter @xmath62 of the notation of @xcite with @xmath78 . ) as - were only announced with no derivation , with a sole reference to the monograph @xcite of akhiezer and glazman on linear operators in hilbert space , we are to understand that berezin and faddeev came to their conclusion by methods of von neumann s self - adjoint extension theory , as presented in ( * ? ? ?
* chapter vii ) , combined with explicit calculations in fourier transform .
this leaves the question open on why they did not approach the extension problem within the same language of kren , viik , and birman , as used by minlos and faddeev for the three - body case . in this language , as we work out in section [ sec:2body - point ] , - would have emerged as a very clean application of the general theory and , most importantly , the asymptotics in would have arisen with a natural and intimate connection with the tms equation .
berezin and faddeev rather focused on re - interpreting the action of the hamiltonian @xmath25 as a renormalised rank - one perturbation of the free laplacian , re - writing in position coordinates as @xmath79 we conjecture that they did not know the old work of bethe and peierls for two nucleons , or they did not consider it relevant in their context , for no word is spent in @xcite to derive the singularity @xmath80 as @xmath81 from their asymptotics . with the subsequent theoretical and experimental advances in nuclear physics the initial playground for models of point interactions
it became clear that the assumption of zero range was only a crude simplification of no fundamental level .
the lack of a physically stringent character for the idealisation of zero range in experimentally observed quantum - mechanical systems , and the somewhat obscure emergence of the unboundedness from below for the self - adjoint realisations of the three - body hamiltonian , decreased the physical interest towards point interactions and left their rigorous study in a relatively marginal position , and the approach of ter - martirosyan and skornyakov quiescent . moreover , after faddeev published in 1963 his fundamental work @xcite on the three - body problem with regular two - body forces , the concern of the physicists switched over to the numerical solutions of the corresponding faddeev equations . in the russian physical literature , mainly under the input of faddeev , methods and models of point interactions , albeit not fully rigorous , moved their applicability to atomic and molecular physics , a mainstream that ideally culminates with the late 1970 s monograph of demkov and ostrovskii @xcite on the `` zero - range potentials '' and their application to atomic physics .
the use of formal delta - like potentials remained for some decades as a tool for a formal first - order perturbation theory ; in addition , the kren - viik - birman self - adjoint extension theory lost ground to von neumann s theory in the literature in english language on the mathematics for quantum mechanics
it rather evolved in more modern forms in application to boundary value problems for partial differential equations , mainly in the modern theory of boundary triplets .
it is the merit of albeverio , gesztesy , and hegh - krohn , and their collaborators ( among whom , streit and wu ) , in the end of the 1970 s and throughout the 1980 s , to have unified an amount of previous investigations by establishing a proper mathematical branch on rigorous models of point interactions , with a systematic study of _ two - body _ hamiltonians and of _ one - body _ hamiltonians with finite or infinitely many _ fixed centers _ of point interaction .
we refer to the monograph @xcite for a comprehensive overview on this production , and especially to the end - of - chapter notes in @xcite for a detailed account of the previous contributions .
the main tools in this new mainstream were : von neumann s extension theory on the first place ( hence with no reference any longer to the methods of kren - viik - birman ) , by which point interaction hamiltonians were constructed as self - adjoint extensions of the restriction of the free laplacian to functions that vanish in a neighbourhood of the point where the interaction is supported ; resolvent identities ( of kren and of konno - kuroda type , see ( * ? ? ? * appendices a and b ) ) by which these self - adjoint extensions were recognised to be finite - rank perturbations of the free laplacian , in the resolvent sense , and were also re - obtained by resolvent limits of schrdinger hamiltonians with shrinking potentials ; plus an amount of additional methods ( dirichlet quadratic forms , non - standard analysis methods , renormalisation methods ) for specific problems .
let us emphasize , in particular , that the original heuristic arguments of bethe and peierls and their two - body contact condition find a rigorous ground based on the fact , which can be proved within von neumann s extension theory ( see , e.g. , ( * ? ? ?
* theorems i.1.1.1 and i.1.1.3 ) ) , that any self - adjoint extension of @xmath82 on @xmath74 has a domain whose elements behave as @xmath83 as @xmath81 , as an @xmath23-wave ( hence a `` low - energy '' ) boundary conditions , for some @xmath84 $ ] . as for the initial three - body problem with two - body point interaction ,
it finally re - gained centrality from the mathematical point of view ( while physically a stringent experimental counterpart was still lacking ) around the end of the 1980 s and throughout the 1990 s .
this was first due to minlos and his school @xcite ( among which melnikov , mogilner , and shermatov ) , by means of the operator - theoretic approach used for three identical bosons by minlos and faddeev , and slightly later due to dellantonio and his school @xcite ( among which figari and teta ) , with an approach based on quadratic forms , where the `` physical '' energy form is first regularised by means of an ultra - violet cut - off and a suitable renormalisation procedure , and then is shown to be realised by a self - adjoint hamiltonian . an alternative direction was started further later by pavlov and a school that included kuperin , makarov , melezhik , merkuriev , and motovilov , @xcite , by indroducing internal degrees of freedom , i.e. , a spin - spin contact interaction , so as to realise semi - bounded below three - body hamiltonians . after a further period of relative quiescence , the subject has been experiencing a new boost , due to the last decade s rapid progress in the manipulation techniques for ultra - cold atoms and , in particular , for tuning the effective @xmath23-wave scattering length by means of a magnetically induced feshbach resonance ( * ? ? ?
* section 5.4.2 ) .
this has made it possible , among others , to prepare and study ultra - cold gases in the so - called `` _ _ unitary regime _
@xcite , i.e. , the case of negligible two - body interaction range and huge , virtually infinite , two - body scattering length ( both lengths being compared to a standard reference length such as the bohr radius ) . in such a regime , unitary gases show properties , including superfluidity , that have the remarkable feature of being universal in several respects @xcite , and are under active experimental and theoretical investigation . as we do not have space here for an outlook on such an active field , we refer to the overview given in the introductory sections of the works @xcite and to the references therein .
let us only underline that from the experimental point of view , zero - range interactions in ultra - cold atom physics are today far from being just an idealisation of real - world two - body potentials with small support and in many realisations the zero - range , delta - like character of the interaction turns out to be an extremely realistic and in fact an unavoidable description . in turn , all this has brought new impulse and motivations to the already developing mathematical research on the subject , with a series of fundamental contributions in the last few years @xcite , many of which provide rigorous ground to experimental or numerical evidence on the physical side .
the hamiltonian of point interaction between two particles in three dimension is well known since the first rigorous attempt @xcite by berezin and faddeev in 1961 , which we have already mentioned in section [ sec : history_tms ] , and the seminal work @xcite by albeverio and hegh - krohn in 1981 . in ( * ? ? ?
* chapter i.1 ) one can find the complete discussion of the self - adjoint realisation of this operator , its explicit domain and action , its resolvent , its spectral properties , its approximation by short - range potentials , and its scattering theory .
in the first part of this section we shall re - obtain this hamiltonian and its main properties within the self - adjoint extension scheme of kren - viik - birman , as opposite to von neumann s scheme used in the above works .
we follow this line both for general reference , because to our knowledge this approach has never been worked out in the literature , and above all because we need to establish the grounds for the second part of this section , where we shall realise the point interaction la ter - martirosyan skornyakov . for the tools from the kren - viik - birman theory we shall make use of
, we refer to the appendix [ app : kvb ] and , more diffusely , to the work @xcite .
the starting point is the operator @xmath85 which is clearly a densely defined , symmetric , closed , and positive operator on the hilbert space @xmath74 .
the variable @xmath86 has the meaning of relative variable between the two particles : after removing the centre of mass of the two - body system , the only relevant problem is in the relative variable .
@xmath8 is the closure of the negative laplacian restricted to the smooth functions compactly supported away from the origin , and @xmath87 the space above is clearly a closed subspace of @xmath88 , and it is also proper , as shown in and below .
the free hamiltonian on @xmath89 is the natural starting point when one aims at constructing a singular interaction supported only at @xmath90 , and any self - adjoint extension of this operator has the natural interpretation of a `` candidate '' hamiltonian for the point interaction . throughout this discussion
it will be convenient to work in fourier transform .
we therefore re - write by means of the following simple lemma : lemma[lemma : dom_hdot_ftransform ] for the operator @xmath8 defined in one has * the domain and the action of @xmath8 are given by @xmath91 * the friedrichs extension of @xmath8 is given by @xmath92 by suitable approximation arguments ( see appendix [ app : approx ] ) , we have @xmath93 moreover , since @xmath94 and @xmath95 are in @xmath74 , then @xmath96 .
hence @xmath97 which , together with @xmath98 , proves part ( i ) . as for part ( ii ) , we first observe that the form domain of @xmath8 , which is the completion of @xmath99 in the @xmath100-norm , is precisely @xmath101\;=\overline{\,h^2_0(\mathbb{r}^3\!\setminus\!\{0\})\;}^{\,\|\,\|_{h^2}}=\;\;h^1_0({\mathbb{r}}^3\!\setminus\!\{0\})\,.\ ] ] since @xmath102 ( see appendix [ app : inclusions ] ) and @xmath103 is the domain of a self - adjoint extension of @xmath8 , namely the self - adjoint @xmath104 on @xmath105 , we conclude that @xmath104 must be the friedrichs extension @xmath106 of @xmath8 , owing to the characterisation of @xmath106 as the unique self - adjoint extension of @xmath8 whose operator domain is contained in @xmath107 $ ] .
as every semi - bounded and densely defined symmetric operator , @xmath8 admits self - adjoint extensions .
we are after the family of such extensions .
the first step is to determine the adjoint of @xmath8 .
proposition[prop : d_hdostar ] let @xmath108 .
* one has @xmath109 * the domain and the action of the adjoint of @xmath8 are given by @xmath110 remark the decomposition of the generic element @xmath111 depends on the chosen @xmath112 , but of course @xmath113 does not , nor does the outcome of @xmath11 applied to @xmath114 , as one sees from . in order to apply the general decomposition formulas of lemma [ lemma : krein_decomp_formula ] we need to deal with an operator with positive bottom .
to this aim we introduce the auxiliary operator @xmath115 which is by construction densely defined , symmetric , and closed , and with bottom @xmath116 . clearly , @xmath117 and @xmath118 . since @xmath119 , then @xmath120 if and only if @xmath121 which by and a standard localisation argument yields .
because of @xmath122 and , we have that @xmath123 this , together with the decomposition formula discussed in appendix [ app : kvb ] ( lemma [ lemma : krein_decomp_formula ] ) , and the characterisation of @xmath124 yield immediately .
the decomposition also implies that the action of @xmath125 on a generic element @xmath126 is the same as the action of @xmath127 on the component @xmath128 , while @xmath129 : this is precisely . as for
, it follows directly from and .
remark the decomposition formula , together with above , gives @xmath130 therefore , expression shows that a generic @xmath111 is in general less regular than @xmath103 , for only the component @xmath131 is in @xmath132 , whereas the component @xmath133 is not .
related to that , shows that whereas @xmath134 , none of the two _ distributions _
@xmath135 and @xmath136 whose difference gives precisely @xmath137 is realised as a square - integrable function ( thus , in the difference the two non - square - integrable singularities cancel out ) . with the above knowledge of @xmath138 and @xmath139 the kren - viik - birman theory provides an explicit prescription to restrict @xmath125 so as to find the whole family of self - adjoint extensions of @xmath140 , and hence of @xmath8 .
theorem[thm:1p1_extensions_birman ] * the self - adjoint extensions of the operator @xmath8 on @xmath74 constitute the one - parameter family @xmath141 where @xmath142 is the friedrichs extension @xmath106 , that is , @xmath143 whereas , for @xmath144 , @xmath145 where @xmath108 is arbitrary .
* each extension @xmath146 is semi - bounded below . in particular , for the bottom @xmath147 of @xmath146 one has @xmath148 * for each @xmath144 the quadratic form of the extension @xmath146 is given by @xmath149\!&=&\!\ !
h^1(\mathbb{r}^3)\dotplus \ker(\mathring{h}^*+\lambda\mathbbm{1 } ) \label{eq : hring - tau_form1}\\ \mathring{h}^{(\tau)}[\phi+u_\xi]\!\!&=&\!\ !
\|\nabla\phi\|_2 ^ 2-\lambda\|\phi+u_\xi\|_2 ^ 2+\lambda\|\phi\|_2 ^ 2+\tau\frac{\;\pi^2}{\sqrt{\lambda}}|\xi|^2 \label{eq : hring - tau_form2 } \end{aligned}\ ] ] for any @xmath150 and any @xmath151 , @xmath152 , where @xmath108 is arbitrary .
fixed @xmath108 , by theorem [ thm : vb - representaton - theorem_tversion ] the self - adjoint extensions of @xmath153 are one - to - one with the self - adjoint operators on hilbert subspaces of @xmath154 , which is the one - dimensional space found in .
the generic case is that of a self - adjoint operator acting on the whole @xmath154 , that is , the map @xmath155 of multiplication by the scalar @xmath144 . in this case
the expression for @xmath156 ( @xmath157 ) reads , by means of and , precisely as .
then follows from and the fact that @xmath158 , and is an immediate consequence of .
the case of the subspace @xmath159 of @xmath154 corresponds by proposition [ prop : parametrisation_sf_sn_tversion ] to the friedrichs extension @xmath160 , where @xmath106 has been determined in of lemma [ lemma : dom_hdot_ftransform ] .
thus , re - writing the generic @xmath161 as @xmath162 for arbitrary @xmath163 and @xmath164 , and comparing it with , one recognises that @xmath165 when @xmath166 .
therefore , @xmath167 and follows from .
this concludes the proof of part ( i ) .
the semi - boundedness of each @xmath146 follows by proposition [ cor : finite_deficiency_index ] , or also proposition [ cor : finite - dimensional ] .
conditions for @xmath146 follow by the general conditions , using the fact that @xmath168 and that @xmath108 is arbitrary , thus also part ( ii ) is proved .
last , we observe that @xmath169=\mathcal{d}[\mathring{h}^{(\tau)}+\lambda\mathbbm{1}]$ ] and @xmath170=(\mathring{h}^{(\tau)}+\lambda\mathbbm{1})[g]-\lambda\|g\|_2 ^ 2 $ ] , thus one deduces - from of theorem [ thm : semibdd_exts_form_formulation_tversion ] applied to @xmath171 : formula is an immediate consequence of ; concerning , formula prescribes the contributions @xmath172 $ ] and @xmath173 to @xmath174 $ ] , which are the first three summands in the r.h.s . of , plus the term @xmath175\;=\;\tau\int_{\mathbb{r}^3}\big|\frac{\xi}{p^2+\lambda}\big|^2\,{\mathrm{d}}p\;=\;\tau\,\frac{\;\pi^2}{\sqrt{\lambda}}\,|\xi|^2\ ] ] which is the fourth summand .
this completes the proof of part ( iii ) .
corollary[cor : delta3d_xcoordinates ] for each self - adjoint extension @xmath146 and for arbitrary @xmath108 one has @xmath176 and @xmath177\!\!\!&=&\!\!\!\big\{g=\phi+\eta\,g_\lambda\,\big|\,\phi\in h^1(\mathbb{r}^3)\,,\;\eta\in\mathbb{c}\,,\;g_\lambda(x)=\frac{e^{-\sqrt{\lambda}|x|}}{4\pi|x|}\big\ } \label{eq : hring - tau_form1_xversion}\\ \mathring{h}^{(\tau)}[\phi+\eta\,g_\lambda]\!\!\!&=&\!\!\ ! -\lambda\|\phi+\eta\,g_\lambda\|_2
^ 2+\|\nabla\phi\|_2 ^ 2+\lambda\|\phi\|_2 ^ 2+\frac{\tau}{8\pi\sqrt{\lambda}}\,|\eta|^2\ , .
\label{eq : hring - tau_form2_xversion}\end{aligned}\ ] ] since @xmath178 the inverse fourier transform of a generic @xmath179 of the form reads @xmath180 where @xmath181 from one computes @xmath182 and plugging this into one obtains .
is the inverse fourier transform of .
follows by taking the inverse fourier transform in , using the characterisation for @xmath183 and formulas and .
is a straightforward re - writing of .
remark[rem : htau_halpha ] in corollary [ cor : delta3d_xcoordinates ] above we have re - derived , apart from an obvious re - definition of the extension parameter @xmath184 , the well - known formulas for @xmath146 previously obtained in the literature by means of von neumann s extension theory . indeed , each @xmath146 is precisely the extension @xmath25 discussed in ( * ? ? ?
* chapter i.1 ) , where @xmath185 by means of , the expression for @xmath186 takes the form of ( * ? ? ?
* eq . ( i.1.1.27 ) ) .
in particular , the analysis of the extension parameter @xmath62 done in ( * ? ? ?
* chapter i.1 ) shows that the two - body point interaction modelled by the hamiltonian @xmath25 has @xmath23-wave scattering length equal to @xmath187 .
remark the function @xmath188 in the decompositions and is customarily referred to as the `` _ _ regular part _ _ '' of the given @xmath114 of the operator domain or the form domain of @xmath146 ,
the difference @xmath189 taking the name of the `` _ _ singular part _ _ '' of @xmath114
. formulas and give , for a generic @xmath190 , @xmath191 which provides once more the interpretation of the point - like character of the interaction modelled by @xmath146 at @xmath90 .
remark owing to the decomposition , a generic @xmath161 displays the characteristic asymptotics of tms - type ( see above ) when @xmath192 . indeed , using the continuity of @xmath188 ,
one derives from @xmath193 and replacing @xmath194 with @xmath35 according to and @xmath184 with @xmath62 given by one obtains @xmath195 in the expression above @xmath108 is arbitrary ( and at fixed @xmath114 the charge @xmath35 is implicitly @xmath112-dependent , see ) and one can therefore read the asymptotics as @xmath196 ; thus , in terms of the scattering length @xmath197 of the interaction , one has @xmath198 which has the form of the tms asymptotics .
remark[rem : renormalis1 ] another customary and equivalent expression for the action of @xmath146 in spatial coordinates is obtained by taking the inverse fourier transform in ( while the inverse fourier transform in yielded ) .
one finds @xmath199 where @xmath200 is the dirac distribution .
the l.h.s .
of is an @xmath201-function for each @xmath161 and the r.h.s
. expresses this @xmath201-function as the difference of two distributions . in general an element @xmath161
does not belong to @xmath103 , in which case @xmath135 is only meant as a distributional derivative : this term has an @xmath201-part plus a distributional ( non - square - integrable ) part @xmath202 which is _ cancelled _ in the difference in the r.h.s . of .
this is consistent with above , where formally ( as @xmath192 ) one obtains a distributional contribution in @xmath135 given by @xmath203 , which is precisely a @xmath200-distribution .
the cancellation occurring in the r.h.s .
of can be regarded as the _ renormalisation _ of @xmath135 needed to give meaning to @xmath204 when @xmath161 . in remark
[ rem : renormalis2 ] below we shall complete this comment by showing that expresses the very same normalisation announced by berezin and faddeev . the functions in @xmath113 have the following asymptotic behaviour .
lemma[lemma : asymptotic_integral ] let @xmath114 be an arbitrary function in @xmath113 . for a fixed @xmath108 let @xmath205 be the decomposition of @xmath114 obtained in proposition [ prop : d_hdostar ] for some @xmath163 and some @xmath206 .
then @xmath207 owing to lemma [ lemma : dom_hdot_ftransform ] , equation , @xmath208 as @xmath76 , so this integral is a @xmath209 contribution in .
as for the two other summands in the decomposition of @xmath179 , we have @xmath210 and @xmath211 which complete the computation of the r.h.s . of .
one _ defines _ the tms extension @xmath212 of @xmath8 , @xmath213 , to be the restriction of @xmath11 to those @xmath114 s of @xmath113 for which in the asymptotics the coefficients of the @xmath214 term and of the @xmath215 term are proportional by a factor @xmath62 , more precisely @xmath216 thus , @xmath217 consists of all @xmath114 s of @xmath113 for which @xmath218 both and express a _
tms condition _ ; it is a constraint on the singular part of the @xmath114 s of @xmath113 that restricts the choice of @xmath219 to those such that @xmath220 condition above is precisely of the form @xmath221 with @xmath222 owing to of theorem [ thm:1p1_extensions_birman ] , this implies at once that the operator @xmath212 selected by the tms condition -- above is precisely the self - adjoint extension @xmath146 of @xmath8 qualified by the parameter @xmath184 given by in terms of @xmath62 .
we also observe that is the inverse of : therefore , by what recalled in remark [ rem : htau_halpha ] , the tms extension @xmath212 models a two - body point interaction with @xmath23-wave scattering length equal to @xmath187 .
thus , * imposing the tms condition on @xmath11 with parameter @xmath62 produces the self - adjoint extension of @xmath8 that gives the point interaction with scattering length @xmath187 ; * the collection of all the tms extensions obtained this way cover the whole family of self - adjoint extensions of @xmath8 : _ all self - adjoint extensions of @xmath8 are of tms
type_. remark[rem : renormalis2 ] with the analysis of this subsection we can supplement remark [ rem : renormalis1 ] above with the following observation . first , one recognises that the condition identified by berezin and faddeev for each self - adjoint extension of @xmath8 is precisely of the tms form .
furthermore , the asymptotics shows that for each @xmath223 the corresponding charge @xmath35 is given by @xmath224 which plugged into yields @xmath225 this is exactly the property announced by berezin and faddeev , and it is the fourier - transformed version of the identity that expresses @xmath226 by means of a suitable renormalisation of @xmath135 .
as discussed in section [ sec : history_tms ] , the problem of a three - particle quantum system with two - body point interaction has been studied since long .
however , one still has a relatively limited knowledge of the corresponding hamiltonians , primarily their self - adjoint realisation ( whereas the information on their stability and spectral properties is only partial ) .
the novel difficulty , as opposed to the two - body case , is due to the fact that the ter - martirosyan skornyakov condition does not select in general a domain of self - adjointness , and one has to further study the self - adjoint extension of the resulting tms operator .
we restrict our discussion to the most studied case , that of two identical fermions in interaction with a third particle of different nature ( the so called `` _ _ 2 + 1 fermionic system _ _ '' ) . after removing the centre of mass ,
the free hamiltonian of a three - dimensional system of two identical fermions of unit mass in relative positions @xmath227 with respect to a third particle of different species and with mass @xmath228 is the operator @xmath229 acting on the hilbert space @xmath230 the subscript ` f ' standing for the fermionic sector of the @xmath201-space , i.e , the square - integrable functions that are anti - symmetric under exchange @xmath231 . following the same path as in subsection [ subsec:2delta ] ,
one therefore starts with the operator @xmath232 where @xmath233 and @xmath234 @xmath8 is a densely defined , closed , positive , and symmetric operator on @xmath235 .
as such , @xmath8 has equal deficiency indices , and in proposition [ prop : d_hdostar_2 + 1 ] below we shall see that they are infinite .
any self - adjoint extension of @xmath8 has a natural interpretation of hamiltonian of point interaction between each fermion and the third particle .
it is convenient first to characterise @xmath99 in fourier transform .
lemma[lemma : dom_hdot_ftransform_2 + 1 ] * the domain of @xmath8 is given by @xmath236 where @xmath237 and the duality product is meant between the spaces @xmath238 and @xmath239 .
the action of @xmath8 is given by @xmath240 where @xmath241 in other words , @xmath99 consists of the @xmath242-functions @xmath243 such that @xmath244 in @xmath245 , where @xmath246 , and @xmath8 acts on such @xmath243 s as the free two - body ( negative ) laplacian . * the friedrichs extension of @xmath8 is given by @xmath247 clearly , for @xmath246 one has @xmath248 , where @xmath249 and @xmath250 is the hyperplane @xmath251 . for a generic @xmath252 a standard trace theorem ( * ? ? ?
* lemma 16.1 ) asserts that @xmath253 , and if in addition @xmath254 , then @xmath255 .
thus , by repeated application of the trace theorem to a function @xmath256 , one finds @xmath257 in @xmath258 .
summarising , @xmath259 one therefore has that @xmath163 is equivalent to @xmath260 let now @xmath163 .
the fourier transforms that follow are all of @xmath201-functions , therefore the corresponding integral expressions are meant as @xmath201-norm limits . from @xmath261 for a.e .
@xmath262 , and from the distributional identity @xmath263 , one finds @xmath264 from this and from one deduces that the function @xmath265 vanishes in @xmath245 . plugging the last identity into yields @xmath266 which proves the first line in .
this result , together with the anti - symmetry of @xmath243 , implies that @xmath267 we now observe that @xmath268 because @xmath269 , and that the map @xmath270 belongs to @xmath239 , because @xmath271 where we used @xmath272 thus , one can regard as the vanishing of a duality product between the spaces @xmath238 and @xmath239 , and since @xmath163 was arbitrary , one concludes the second line in .
equations - are the fourier - transformed version of , and this concludes the proof of part ( i ) . concerning part ( ii ) , we first observe that the form domain of @xmath273 , namely the completion of @xmath274 , is the space @xmath275\;=\;h^1_0((\mathbb{r}^3\times\mathbb{r}^3)\!\setminus\!(\gamma_1\cup\gamma_2))\cap{\mathcal{h}}\,.\ ] ] by a standard approximation with smooth and compactly supported functions ( see appendix [ app : inclusions ] ) , @xmath276 and @xmath277 is the domain of the self - adjoint extension of @xmath273 given by the free negative laplacian on @xmath235 .
therefore , owing to the characterisation of the friedrichs extension @xmath106 as the unique self - adjoint extension of @xmath8 whose operator domain is contained in @xmath107 $ ] , one deduces immediately . in the next proposition
we characterise the adjoint of @xmath8 , which is the preliminary step in order to identify the self - adjoint extensions of @xmath8 within the general scheme of the kren - viik - birman theory .
proposition[prop : d_hdostar_2 + 1 ] let @xmath108 .
* one has @xmath278 * there exist constants @xmath279 such that for a generic @xmath280 one has @xmath281 * the domain and the action of the adjoint of @xmath8 are given by @xmath282 and @xmath283 where @xmath284 and @xmath133 are defined as in above , and @xmath285 in order to prove part ( i ) we use the fact that @xmath286 if and only if for every @xmath287 @xmath288 where we applied in the last step . since @xmath289 ( owing to with @xmath290 ) and @xmath108 , then @xmath291 ( in the sense that each quantity controls the other from above and from below ) , and hence the fact that @xmath292 is equivalent to @xmath293 therefore , the last identity above implies , owing to the second line of , that @xmath294 for some @xmath295 , and hence each @xmath296 is of the form @xmath297 given by . part ( ii ) is taken directly from ( * ? ? ?
* lemma b.2 ) . concerning part ( iii ) , because of @xmath160 and , we have that @xmath298 this , together with the decomposition formula of lemma [ lemma : krein_decomp_formula ] and the characterisation of @xmath139 yield immediately .
the decomposition also implies that the action of @xmath138 on a generic element @xmath299 is the same as the action of @xmath300 on the component @xmath301 , while @xmath302 : this is precisely . as for
, it is an immediate consequence of and .
the space @xmath139 determined in is the kren space of the model ( the `` boundary value space '' , in modern terminology ) .
it is known by the kren - viik - birman theory that the self - adjoint extensions of @xmath8 defined in are parametrised by self - adjoint operators acting on hilbert subspaces of @xmath139 , according to the classification given by theorem [ thm : vb - representaton - theorem_tversion ] in appendix [ app : kvb ] . unlike the two - body model discussed in the previous section , where @xmath139 was one - dimensional , in the three - body model
this space is _ infinite - dimensional _
( compare with ) , which makes the variety of the self - adjoint extensions of @xmath8 much more complicated . in this respect , the extensions of ter - martirosyan skornyakov type form a proper sub - family ( in the two - body case , all self - adjoint extensions were of tms type ) .
eventually one introduces the tms extensions of @xmath8 , analogously to section [ subsec : tms-1 + 1 ] for the two - body model , as restrictions of @xmath11 to domains characterised by special asymptotics of their wave - functions in the vicinity of the coincidence hyperplanes . prior to that , in this subsection we shall develop a general scheme for the identification of a generic self - adjoint extension of @xmath8 , within which we will later select those of tms type .
to this aim , it is convenient first to define the expressions @xmath303 and @xmath304 for fixed @xmath108 , where @xmath305 is given by and @xmath306 since , for arbitrary @xmath307 and @xmath308 , @xmath309 ( owing to a schwartz inequality in the first step and in the second one ) , we see that the integral in is finite for any @xmath308 , @xmath307 , while in general it diverges when @xmath310 , as the example @xmath311 shows .
a similar argument shows that the integral in is finite too at least for @xmath310 .
summarising , @xmath312 is well - defined point - wise for almost every @xmath313 for @xmath308 , @xmath307 , whereas @xmath314 is so ( at least ) for @xmath295 .
the relevance of functions of the form @xmath315 and @xmath316 is due to the fact that they arise in the asymptotic behaviour of the elements of @xmath113 .
lemma[lemma : asymptotic_integral_2 + 1 ] let @xmath114 be an arbitrary function in @xmath113 . for a fixed @xmath108 consider the decomposition of @xmath179 in terms of @xmath317 , @xmath318 , @xmath319 given by . then , as @xmath320 , @xmath321 as a point - wise identity for almost every @xmath322 .
remark for a generic @xmath323 , and correspondingly for a generic charge @xmath310 , the quantity in the l.h.s .
of is _ infinite _ for every finite @xmath20 because , as remarked after the definitions - , the quantity @xmath312 is in general infinite when @xmath310 . instead , when additionally @xmath308 , with @xmath307 , the r.h.s . of is finite ( for almost every @xmath324 ) : this case corresponds to a dense set of @xmath114 s in @xmath113 , and for such @xmath114 s the quantity in the l.h.s .
of is _ finite _ for finite @xmath20 and only diverges , linearly in @xmath20 , as @xmath320 .
splitting @xmath325 according to yields @xmath326 the first summand in the r.h.s .
of is a @xmath209-contribution , owing to lemma [ lemma : dom_hdot_ftransform_2 + 1](i ) . for the second summand ,
which is re - written as @xmath327 one finds @xmath328 whence @xmath329 analogously , for the third summand in the r.h.s .
of one has @xmath330 and @xmath331 whence @xmath332 these findings , re - written with the definitions and , show that the r.h.s .
of is precisely given by formula .
thus , functions in @xmath113 display completely analogous asymptotics to the two - body model .
it is important to observe , however , that whereas was an identity between _ scalars _ , here is a _ point - wise _ almost everywhere identity between _
functions_. this is a crucial difference to keep into account when one imposes the ter - martirosyan skornyakov condition in such asymptotics . to elaborate on this point further at a later stage ,
let us also interpret @xmath333 and @xmath334 as maps between suitable functional spaces .
proposition[prop : t - w ] let @xmath108 .
* for each @xmath335 the expression defines a densely defined and symmetric operator @xmath336 with domain @xmath337 .
moreover , @xmath338 maps continuously @xmath339 into @xmath340 for each @xmath341 .
instead , @xmath342 .
* the expression defines a bounded , positive , and invertible linear operator @xmath343 , and for generic @xmath344 one has @xmath345 remark the choice of @xmath74 as the hilbert space where to study @xmath338 is made here for consistency with the previous literature @xcite , but it has no fundamental reason . as we shall discuss in subsection
[ subsec : tms_2p1 ] below , what is intrinsically fundamental for the self - adjoint extension theory of @xmath8 is the operator @xmath346 on the hilbert space @xmath347 .
\(i ) we re - write as @xmath348 , where @xmath349 the symmetry of @xmath338 on @xmath74 is obvious , since @xmath350 is the multiplication by a real function and @xmath351 is an integral operator with real and symmetric kernel , and so too is the fact that @xmath352 is dense in @xmath74 .
it is also clear that @xmath353 , thus it only remains to prove that @xmath354 , i.e. , @xmath355 . in turn , setting @xmath356 , the last inequality is equivalent to @xmath357 , where @xmath358 it is easily verified ( using to introduce the @xmath359-factors ) that for the _ positive _ function @xmath360 one has @xmath361 and @xmath362 a standard schur test based on ( * ) and ( * * ) implies @xmath357 and hence @xmath354 for @xmath341 and arbitrary @xmath363 .
the function @xmath364 is a counter - example showing that the same bound can not hold for @xmath365 : indeed , clearly @xmath366 for arbitrary @xmath367 , but @xmath368 and hence @xmath369 .
\(ii ) for arbitrary @xmath370 we compute @xmath371 where we used the symmetry under exchange @xmath372 in the second step , in the third step , and in the last step . therefore ,
@xmath373 where we used in the last step , which shows that @xmath374 , that the map @xmath343 is bounded , and that holds true . owing to
, one has @xmath375 , thus @xmath376 is positive .
furthermore , the following chain of implications holds : @xmath377 @xmath378 @xmath379 @xmath380 @xmath378 @xmath381 @xmath378 @xmath382 , where we used in the first implication and in the last one ; this proves that @xmath376 is injective and hence invertible on its range . for a generic @xmath295 one
has this chain of implications : @xmath383 @xmath384 @xmath378 @xmath379 @xmath385 @xmath378 @xmath386 @xmath378 @xmath387 ( again using in the first implication and in the last one ) , hence by duality @xmath388 must be dense in @xmath245 .
since @xmath376 is bounded , then @xmath388 is also closed in @xmath245 , thus @xmath376 is an invertible bijection @xmath389 . as an immediate consequence of proposition [ prop : t - w ] , @xmath390 defines a scalar product in @xmath391 . it is _ equivalent _ to the standard scalar product of @xmath391 , as follows by combining with .
we shall denote by @xmath392 the hilbert space consisting of the @xmath391-functions and equipped with the scalar product @xmath393 .
then the map @xmath394{\;\;\;\cong\;\;\;}\;h^{-1/2}_{w_\lambda}(\mathbb{r}^3)\,,\qquad u_\xi \longmapsto \,\xi \end{split}\ ] ] is an isomorphism between hilbert spaces , with @xmath139 equipped with the standard scalar product inherited from @xmath235 .
one can therefore equivalently parametrise the self - adjoint extensions of @xmath8 in terms of self - adjoint operators acting on hilbert subspaces of @xmath139 or of @xmath392 .
the whole family of such extensions is given by the kren - viik - birman theory through the classification of theorem [ thm : vb - representaton - theorem_tversion ] .
the previous analysis brings us now to the class of operators on @xmath392 ( or , also , operators on @xmath139 ) which identify those self - adjoint extensions of @xmath8 of ter - martirosyan skornyakov type . as it will emerge in the following ,
the crucial point is the possibility of reducing @xmath338 to a @xmath201-closed invariant subspace with values into @xmath245 and to define on the orthogonal complement another @xmath396-valued symmetric operator @xmath397 .
given the resulting @xmath398 one has then to investigate the self - adjointness of @xmath399 on @xmath392 .
we observe that this is related to , but it is _ not _ the same question of the self - adjointness of @xmath338 on @xmath74 .
the study of the self - adjoint extensions of @xmath338 , as a densely defined and symmetric operator on @xmath74 , has been carried on systematically in a series of works by minlos and shermatov @xcite , melnikov and minlos @xcite , shermatov @xcite , and minlos @xcite . in the additional work @xcite by one of us and
co - workers , the friedrichs extension of @xmath338 was studied ( in those regime of masses @xmath228 in which @xmath338 itself is semi - bounded below ) .
it is relevant to remark that in all those works @xmath352 was fixed in spaces of various regularity , at least @xmath400 .
we recall from proposition [ prop : t - w](i ) that @xmath401 because both its multiplicative part @xmath350 and its integral part @xmath351 map separately @xmath400 into @xmath74 : as discussed in @xcite , in the domain of a self - adjoint extension @xmath402 of @xmath338 there are elements @xmath35 for which neither @xmath403 nor @xmath404 is square - integrable , but their difference is , due to a cancellation of singularities in @xmath405 .
more precisely , @xmath338 commutes with the rotations in @xmath105 and , with respect to the canonical decomposition @xmath406 ( where the @xmath407 s are the spherical harmonics on @xmath408 ) , @xmath338 leaves each @xmath409 invariant and is densely defined and symmetric on @xmath409 , thus @xmath410 and @xmath411 therefore , the study of the self - adjointness or of the self - adjoint extensions of @xmath338 boils down to the same study for each @xmath412 .
it is today well - known from the works cited above that for _ even _
@xmath413 s @xmath412 is self - adjoint on @xmath409 , while for _ odd _
@xmath413 s there exist masses @xmath414 such that @xmath412 is self - adjoint on @xmath409 for @xmath415 and it has instead a one - parameter family of self - adjoint extensions for @xmath416 $ ] .
as seen in the proof of proposition [ prop : t - w](i ) , @xmath338 fails to map @xmath417 into @xmath245 and the counter - example considered therein was a function in @xmath418 .
in fact , that failure is exceptional and it does not occur for @xmath419-functions with a sufficient amount of oscillations , as the following proposition shows .
proposition[prop : t - ell ] for each @xmath420 , and in terms of the notation of - , @xmath421 whence , in particular , @xmath422 as discussed already in the proof of proposition [ prop : t - w](i ) , we only need to prove the statement for @xmath351 , the integral part of @xmath338 see .
analogously to - , @xmath423 where @xmath424 acts symmetrically on @xmath425 .
it is standard to derive from and that @xmath426 where @xmath427 is the @xmath413-th legendre polynomial .
thus , proving is equivalent to proving @xmath428 which is in turn equivalent to the boundedness in @xmath429 of the integral operator @xmath430 defined by @xmath431 using and integrating by parts @xmath420 times in @xmath432 yields @xmath433 since @xmath434 , analogously to @xmath435 then @xmath436 where we used in the first and last step , and the formula @xmath437 ( @xmath438 ) in the third step . from @xmath439
we deduce @xmath440 and from @xmath441 we deduce @xmath442 a standard schur test based on ( * ) and ( * * ) implies @xmath443 , thus concluding the proof .
an immediate consequence of propositions [ prop : t - w](i ) and [ prop : t - ell ] is the following .
corollary let @xmath444 be the operator acting as @xmath338 on the hilbert space @xmath445 with domain @xmath446 .
then @xmath444 is densely defined and symmetric , and it maps continuously @xmath447 ( with the @xmath448-norm ) into @xmath449 ( with the @xmath450-norm ) . in turn , the corollary above , together with proposition , show that although in general @xmath451 nevertheless the map @xmath346 can be defined on parts of @xmath347 with values in itself . elaborating further , we obtain the following result .
proposition[lem : selfadj_hierarchy ] the following be given : * two constants @xmath108 and @xmath452 , * the densely defined and symmetric operator @xmath453 on the hilbert space @xmath445 which acts as @xmath338 with domain @xmath446 , * and a densely defined and symmetric operator @xmath397 on the hilbert space @xmath454 with @xmath455 . with respect to the decomposition @xmath456 , let @xmath457 then @xmath458 is a densely defined and symmetric operator on @xmath74 and @xmath459 is a densely defined and symmetric operator on @xmath392 .
moreover , if @xmath460 is a self - adjoint extension of @xmath461 on @xmath392 , then @xmath462 ( where @xmath463 is the isomorphism ) is a self - adjoint operator on @xmath139 .
the statements for @xmath458 and @xmath464 are obvious , and so too is the density of @xmath465 in @xmath392 .
the symmetry of @xmath461 follows from the identity , valid for @xmath466 , @xmath467 where we used the symmetry of @xmath458 in @xmath74 , the fact that @xmath62 is real , and the properties of @xmath376 discussed in proposition [ prop : t - w](ii ) .
remark[rem : selfadj_of_t_not_enough ] if , in addition to the assumptions of proposition [ lem : selfadj_hierarchy ] above , one assumes also that @xmath444 and @xmath397 are self - adjoint on their respective hilbert spaces and hence @xmath458 is _ self - adjoint _ on @xmath74 , then @xmath468 to see this , we use the fact that for generic @xmath469 there exists @xmath470 such that @xmath471 owing to , , and , @xmath472 whereas , for @xmath473 , @xmath474 therefore , ( * ) reads @xmath475 or also ( using @xmath476 ) @xmath477 since @xmath458 is self - adjoint on @xmath74 , the last bound implies @xmath478 , whence the conclusion .
we observe , however , that this argument and the conclusion are not enough to claim that the self - adjointness of @xmath458 implies the self - adjointness of @xmath461 : the latter could still have a larger adjoint and admit self - adjoint extensions .
we thus see that each self - adjoint extension @xmath460 of @xmath461 on @xmath392 identifies one self - adjoint extension of @xmath8 by means of its unitarily equivalent version @xmath479 .
this extension , call it @xmath212 , as prescribed by theorem [ thm : vb - representaton - theorem_tversion ] ( namely the kren - viik - birman extension theory ) is the restriction of @xmath11 to the domain @xmath480 indeed , in comparison with the general formula , @xmath481 has a trivial orthogonal complement in @xmath139 . for each
@xmath223 one deduces from that @xmath482 a comparison between and shows that @xmath217 is obtained as a restriction @xmath113 by imposing the condition @xmath483 as an identity in @xmath139 which , by the unitary equivalence , is equivalent to @xmath484 as an identity in @xmath392 . the self - adjoint extension @xmath212 is given by the restriction of @xmath11 to those elements of @xmath113 whose charges , in terms of the decomposition , instead of belonging generically to @xmath347 are such that @xmath35 belongs to the domain of @xmath460 and @xmath485 is of the form @xmath486 .
we now see that @xmath212 is an operator of tms type . indeed , owing to the definition
, condition implies @xmath487 as an identity in @xmath74 . in turn , owing to the definition , reads @xmath488 plugging - into yields the following asymptotics for elements in @xmath217 as @xmath76 : @xmath489 @xmath490 the asymptotics is completely analogous to for the two - body system , except for the fact that it only holds for non - spherically symmetric charges . if the charge @xmath35 is spherically symmetric , then is modified with an additional @xmath215-term in @xmath20 as done in .
in fact , both and express the same kind of ultra - violet asymptotics at the coincidence hyperplanes , for their formal inverse fourier transformed version reads @xmath491 where @xmath492 is identically zero if the charge @xmath35 associated to @xmath223 belongs to @xmath447 , whereas @xmath493 if @xmath494 . in either case the leading singularity in @xmath495 as @xmath496 is precisely of the form @xmath497 .
we summarise the above analysis by saying that * the operator @xmath212 , defined by - in terms of a self - adjoint extension @xmath460 of @xmath498 on @xmath392 , where @xmath499 , @xmath397 is a densely defined and symmetric operator @xmath397 on @xmath454 with @xmath455 , and @xmath500 is the component of @xmath338 on @xmath501 with domain @xmath502 , is a self - adjoint extension of @xmath8 because it satisfies the condition , which is a special case of the conditions of self - adjoint extension of the kren - viik - birman theory ; * and moreover @xmath212 is a hamiltonian of ter - martirosyan skornyakov type because , through - , _ implies _ the tms asymptotics - .
we observe , however , two fundamental differences with respect to the construction of the point interaction hamiltonian for the two - body system . in the two - body case ,
the tms condition , imposed in the asymptotics of the elements of @xmath113 , turns out to be a _ condition of self - adjoint extension_. in the 2 + 1 fermionic system , instead , * the tms condition is found to hold only for a class of self - adjoint extensions of @xmath8 , those identified a la kren - viik - birman by the _
constraint , * and if one imposes as a point - wise identity valid for a generic class of @xmath35 s , one is _ not _ guaranteed yet to have identified a domain of self - adjointness for an extension of @xmath8 , because in general such an ansatz is not implemented by a self - adjoint operator on @xmath183 .
remark a clarification in retrospective on the emergence and the meaning of the operators @xmath397 and @xmath458 introduced in proposition [ lem : selfadj_hierarchy ] above is surely beneficial at this point .
the key issue that arises evidently in our discussion is that defines a densely defined , symmetric operator @xmath338 on @xmath74 which fails to map a linear space of certain spherically symmetric @xmath448-functions into @xmath503 , whereas it does map into the latter space all the @xmath448-functions that are @xmath201-orthogonal to the spherically symmetric ones . as a consequence ,
unlike what is customarily given for granted at this point in the literature , there arises the issue to make the symmetric operator @xmath346 _ densely defined _ on the space of charges @xmath392 and to realise it self - adjointly .
however , one would like to have a self - adjoint operator on @xmath392 precisely of the form @xmath346 : indeed on the one hand by general facts ( krein - vishik - birman ) a self - adjoint operator on @xmath392 identifies a self - adjoint hamiltonian of point interaction for the three - body system , and on the other hand the structure @xmath346 results in a ter - martirosyan skornyakov condition for the elements in the domain of such an hamiltonian .
the way to retrieve a tms - like hamiltonian was then to cure @xmath346 on the sector of spherical symmetry , by replacing @xmath338 with a modified operator @xmath504 , where @xmath444 is the restriction of @xmath338 to the subspace @xmath505 and @xmath397 is an arbitrary self - adjoint operator in @xmath454 _ with values in _ @xmath503 .
the corresponding @xmath399 is now symmetric and densely defined on @xmath392 and any its self - adjoint extension identifies a tms hamiltonian for the three - body systems .
for such a hamiltonian , the tms condition emerges only when the charge @xmath35 has symmetry @xmath420 ; on the sector @xmath506 what holds instead is the weaker condition .
the latter still prescribes that the generic element @xmath495 of the domain of the tms hamiltonian has a leading singularity @xmath507 as @xmath508 when the charge @xmath35 associated to @xmath114 is spherically symmetric , however no @xmath62-constraint is prescribed between singular and regular part of @xmath114 .
this larger freedom in the unconstrained regular part reflects precisely the arbitrariness of @xmath397 .
for the two - body system with point interaction , imposing the tms asymptotics at scattering length @xmath509 selects the whole one - parameter family of self - adjoint extensions of @xmath8 , the formal free hamiltonian defined away from the coincidence configurations . for larger systems
, @xmath8 has infinite deficiency indices and the tms asymptotics emerge for a proper subclass of extensions of @xmath8 , provided that the charges are taken in a domain of suitable regularity and symmetry .
indeed , except for the two - body case , the tms condition expresses _ point - wise _ asymptotics , which per se is not enough to be a condition of self - adjointness : the latter has to be a suitable _ functional _ condition , such as in the preceding discussion .
recognising a tms condition as a self - adjointness condition , by means of the general classification of self - adjoint extensions given by the kren - viik - birman theory , is an idea that dates back to the original announcements @xcite by minlos and faddeev in 1961 and it has been exploited in a series of works by minlos and collaborators @xcite . to our understanding , however , the issue of making the operator that in our notation reads @xmath346 a well - defined map on the space of charges @xmath347 , more precisely the issue on whether @xmath510 , was never addressed , nor was it noted that @xmath338 fails to map spherically symmetric functions of @xmath417 into @xmath245 ( see proposition [ prop : t - w](i ) and its proof ) .
in fact , in all recent works @xcite the initial domain of @xmath338 is taken to be @xmath400 and hence , owing to proposition [ prop : t - w](i ) , in general @xmath511 consists of @xmath201-functions that can not be pulled back to @xmath512-functions by the inverse of @xmath376 .
also , the ubiquitous statement in the above - mentioned literature , according to which each self - adjoint realisation in @xmath74 of the operator @xmath338 defined formally by identifies ( by general facts of the kren - viik - birman theory ) a self - adjoint extension of @xmath8 that displays the tms asymptotics for the functions of its domain , need be made more precise in two crucial respects .
first , one has to factor out the part of @xmath338 that acts on @xmath454 , the spherically symmetric functions , as we argued in proposition [ lem : selfadj_hierarchy ] and in the discussion that followed from it .
second , even when @xmath338 is self - adjoint on the subspace of higher momentum charges , the corresponding densely defined and symmetric operator @xmath498 on @xmath392 is not necessarily self - adjoint and may in turn admit a multiplicity of self - adjoint extensions , as we observed in remark [ rem : selfadj_of_t_not_enough ] : only a self - adjoint extension @xmath460 of @xmath461 identifies , a la kren - viik - birman , the self - adjoint extension @xmath212 of @xmath8 .
it remains therefore unclear how to relate the range of masses @xmath228 ( identified in @xcite ) in which @xmath338 , initially defined on @xmath513 , is self - adjoint or has a family of self - adjoint extensions on @xmath74 with the actual range of masses in which @xmath8 admits one or more self - adjoint extensions displaying the tms asymptotics .
the fact that @xmath498 may be only symmetric on @xmath392 even when @xmath458 is self - adjoint on @xmath74 should account for a _
larger _ range of masses in which @xmath8 has a multiplicity of tms - like self - adjoint extensions than the range in which @xmath338 has a multiplicity of self - adjoint extensions .
it becomes of great interest now to re - read and understand , in terms of the general classification of self - adjoint extensions of @xmath8 provided by the kren - viik - birman theory , those results that produced tms self - adjoint extensions of @xmath8 through an approach based on quadratic forms on hilbert space ( see @xcite and above all the recent work @xcite ) .
indeed , the quadratic form approach produces a single self - adjoint tms hamiltonian , or alternatively a family of self - adjoint tms hamiltonians , all extensions of @xmath8 , in a regime of masses that differs from what is known from the operator - theoretic approach and that coincides instead to what is found in the physical literature see ( * ? ? ?
* remark 2.4 and proposition 2.2 ) , as well as the discussion around ( * ? ? ?
* eq . ( 1.17 ) ) .
owing to the general picture of the self - adjoint extension theory , each such hamiltonian _ must _ be selected by a condition on the charges realised by a self - adjoint map on @xmath183 as in or , more generally , in .
one should identify such a map and to compare it to its analog in the operator - theoretic approach .
armed with the analysis and the discussion developed here , we plan to address these issues in a follow - up work .
in this appendix we collect the main results of the kren - viik - birman theory of self - adjoint extensions of semi - bounded symmetric operators .
this theory was developed by kren @xcite , viik @xcite , and birman @xcite ) between the mid 1940 s and the mid 1950 s .
for the present formulation and the proof of all the statements that follow we refer to the comprehensive discussion @xcite , as well as to the expository works @xcite . for
any given symmetric operator @xmath514 with domain @xmath515 , let @xmath516 be the `` bottom '' of @xmath514 , i.e. , its greatest lower bound .
hereafter @xmath514 shall be semi - bounded below , meaning therefore @xmath517 .
it is not restrictive to assume henceforth @xmath518 for in the general case one applies the discussion that follows to the strictly positive operator @xmath519 , @xmath520 , and then re - express trivially the final results in terms of the original @xmath514 .
when @xmath514 is densely defined , _
the choice @xmath521 implies that the friedrichs extension @xmath522 of @xmath514 is invertible with bounded inverse defined everywhere on @xmath235 _ : this will allow @xmath523 to enter directly the discussion . in the general case in which
@xmath522 is not necessarily invertible , the role of @xmath523 is naturally replaced by the inverse @xmath524 of any a priori known self - adjoint extension @xmath525 of @xmath514 , which thus takes the role of given `` datum '' of the theory .
moreover , with the choice @xmath521 , the level @xmath526 becomes naturally the reference value with respect to which to express the other distinguished ( canonically given ) extension of @xmath514 , the kren - von neumann extension @xmath527 .
lemma[lemma : krein_decomp_formula ] _
( decomposition formulas ) _ for a densely defined symmetric operator @xmath514 with positive bottom , one has @xmath528
theorem[thm : vb - representaton - theorem_tversion ] _ ( classification of self - adjoint extensions operator version . )
_ let @xmath514 be a densely defined symmetric operator on a hilbert space @xmath235 with positive bottom ( @xmath521 ) .
there is a one - to - one correspondence between the family of all self - adjoint extensions of @xmath514 on @xmath235 and the family of the self - adjoint operators on hilbert subspaces of @xmath529 .
if @xmath530 is any such operator , in the correspondence @xmath531 each self - adjoint extension @xmath532 of @xmath514 is given by @xmath533 theorem[thm : semibdd_exts_operator_formulation_tversion]_(characterisation of semi - bounded extensions . ) _ let @xmath514 be a densely defined symmetric operator on a hilbert space @xmath235 with positive bottom . if , with respect to the notation of , @xmath532 is a self - adjoint extension of @xmath514 , and if @xmath534 , then @xmath535 as an immediate consequence , @xmath536 for any semi - bounded @xmath532 . in particular , positivity or strict positivity of the bottom of @xmath532
is equivalent to the same property for @xmath530 , that is , @xmath537 moreover , if @xmath538 , then @xmath539 theorem[thm : semibdd_exts_form_formulation_tversion]_(characterisation of semi - bounded extensions form version . ) _ let @xmath514 be a densely defined symmetric operator on a hilbert space @xmath235 with positive bottom and , with respect to the notation of , let @xmath532 be a _ semi - bounded _ ( not necessarily positive ) self - adjoint extension of @xmath514
. then @xmath540_tversion } \mathcal{d}[t]\;=\ ; \mathcal{d}[s_t]\,\cap\,\ker s^*\ ] ] and @xmath541\;&=\;\mathcal{d}[s_f]\,\dotplus\,\mathcal{d}[t ] \\
s_t[f+v , f'+v']\;&=\;s_f[f , f']\,+\,t[v , v ' ] \\ & \forall f , f'\in\mathcal{d}[s_f],\;\forall v , v'\in\mathcal{d}[t]\ , .
\end{split}\ ] ] as a consequence , @xmath542 and @xmath543 proposition[prop : parametrisation_sf_sn_tversion]_(parametrisation of @xmath522 and @xmath527 . )
_ let @xmath514 be a densely defined symmetric operator on a hilbert space @xmath235 with positive bottom and let @xmath532 be a _ positive _ self - adjoint extension of @xmath514 , parametrised by @xmath530 according to theorems [ thm : vb - representaton - theorem_tversion ] and [ thm : semibdd_exts_form_formulation_tversion ] .
* @xmath532 is the _ friedrichs extension _ when @xmath544=\{0\}$ ] ( `` @xmath545 '' ) .
* @xmath532 is the _ kren - von neumann extension _ when @xmath546=\ker s^*$ ] and @xmath547 @xmath548 ( @xmath549 ) .
proposition[cor : finite_deficiency_index]_(finite deficiency index . )
_ if @xmath514 is a semi - bounded and densely defined symmetric operator on a hilbert space @xmath235 with finite deficiency index , then * the semi - boundedness of @xmath532 is equivalent to the semi - boundedness of @xmath530 ; * any self - adjoint extension of @xmath514 is bounded below .
proposition[cor : finite - dimensional]_(``finite - dimensional '' extensions are always semi - bounded . ) _ given a semi - bounded and densely defined symmetric operator @xmath514 on a hilbert space @xmath235 , whose bottom is positive , all the self - adjoint extensions of @xmath532 of @xmath514 for which the parameter @xmath530 , in the parametrisation of theorem [ thm : vb - representaton - theorem_tversion ] , is a self - adjoint operator acting on a _ finite - dimensional _ subspace of @xmath529 are semi - bounded . for the occurrence of unbounded below self - adjoint extensions
it is necessary that @xmath550 .
we prove here that the @xmath551-closure of @xmath89 , which is by definition the space @xmath552 , coincides with the space @xmath553 , thus obtaining the identity .
clearly @xmath554 because the @xmath551-convergence implies the point - wise convergence of continuous functions .
thus , given @xmath555 with @xmath556 , we only need to find for arbitrary @xmath307 a function @xmath557 such that @xmath558 given a cut - off function @xmath559 such that @xmath560 \\ \chi(r)\;=\;1\qquad & \textrm { for } \;r\in[2,+\infty)\ , , \end{split}\ ] ] set @xmath561 then , for any @xmath562 , @xmath563 and @xmath564 where @xmath565 depends only on @xmath566 , @xmath567 , and @xmath568 .
correspondingly , each function @xmath569 belongs to @xmath103 and vanishes when @xmath570 .
furthermore , we now show that @xmath571{\;\ ; n\to + \infty\;\;}\;0\,.\ ] ] indeed , @xmath572 follows immediately by dominated convergence and , whereas @xmath573 is obtained with the following argument .
first , one estimates @xmath574 for the first summand in the r.h.s . of the inequality above one
has @xmath575 as @xmath576 again by dominated convergence and . for the second summand , @xmath577{\;\ ; n\to + \infty\;\;}\;0 \end{split}\ ] ] where we used in the first step , a hlder inequality in the second , the sobolev embedding in the third , and dominated convergence in the last , @xmath578 denoting the closed ball of @xmath105 of radius @xmath579 centred at the origin . for the third summand , @xmath580{\;\ ; n\to + \infty\;\;}\;0 \end{split}\ ] ] where in the first step we used , in the second we used the estimate @xmath581 ( where the constant @xmath582 does not depend on @xmath583 ) that follows from morrey s inequality and the sobolev embedding ( see ( * ? ? ?
* section 5.6.2 ) , and in the last step we used dominated convergence .
thus , is proved . as a consequence of above , for the arbitrary @xmath307 fixed at the beginning
there is @xmath584 and @xmath585 such that @xmath586 is a @xmath551-function satisfying @xmath587 we consider now a standard mollification @xmath588 of @xmath589 for some @xmath590 with @xmath591 and @xmath592 , @xmath562 , @xmath86 . then @xmath593 and @xmath594 in @xmath551 as @xmath576
therefore , there is @xmath595 large enough so that @xmath596 satisfies @xmath597 _ and _ @xmath598 ( where @xmath200 is the radius of the ball that contains the support of @xmath589 , see above ) . as a consequence ,
@xmath599 vanishes in @xmath600 : indeed , if @xmath601 and @xmath602 , that is , @xmath603 and @xmath604 , one has @xmath605 and hence @xmath606 because @xmath589 vanishes in @xmath607 .
summarising , we have found @xmath608 such that @xmath609 last , we consider a cut - off function @xmath610 such that @xmath611 \\ \zeta(r)\;=\;0\qquad & \textrm { for } \;r\in[2,+\infty)\ , , \end{split}\ ] ] and set @xmath612 then , in complete analogy to the reasoning above , we see that for @xmath583 large enough each @xmath613 belongs to @xmath614 and @xmath615 in @xmath551 as @xmath576 .
this implies the existence of @xmath616 such that the function @xmath617 satisfies @xmath618 using , , and above in a triangular inequality we finally conclude @xmath619 for a function @xmath620 , which completes the proof of .
we prove here : the proof applies straightforwardly also to obtain .
let @xmath252 .
we want to show that @xmath243 belongs to @xmath621
. since @xmath622 it is enough to find , for arbitrary @xmath307 , a function @xmath623 such that @xmath624 first , since @xmath625 and @xmath626 is dense in @xmath627 , there exists @xmath628 such that @xmath629 given a cut - off function @xmath559 such that @xmath560 \\ \chi(r)\;=\;1\qquad & \textrm { for } \;r\in[2,+\infty)\ , , \end{split}\ ] ] set @xmath630 then , for any @xmath562 , @xmath631 and @xmath632 where here and henceforth @xmath633 denotes the @xmath634-dimensional gradient and @xmath565 depends only on @xmath566 and @xmath567 .
correspondingly , each function @xmath635 belongs to @xmath636 and we now show that @xmath637{\;\ ; n\to + \infty\;\;}\;0\,.\ ] ] indeed , @xmath638 follows immediately by dominated convergence and , whereas @xmath639 follows from the vanishing of both summands in the r.h.s . of the inequality
@xmath640 explicitly , @xmath641 by dominated convergence , owing to , whereas @xmath642{\;\ ; n\to + \infty\;\;}\;0\ , , \end{split}\ ] ] where we used in the first step , a hlder inequality and the estimate @xmath643 in the second step , where @xmath644 depends only on the radius of @xmath645 , the continuous embedding @xmath646 in the third step , and dominated convergence in the last step . as a consequence of above , for the arbitrary @xmath307 fixed at the beginning
there is @xmath584 such that @xmath647 satisfies @xmath648 a triangular inequality based on and then yields , thus completing the proof of . clearly
, also shows that to any function in @xmath649 there is a function in @xmath636 arbitrarily close in the @xmath100-norm , which implies that the spaces @xmath627 and @xmath650 coincide .
this yields the identity in .
the arguments above apply virtually unchanged both for the inclusion @xmath651 and for the identity @xmath652 , with the obvious removal of one variable , thus proving .
we warmly thank g. dellantonio and l. guerini for enlightening discussions on the subject .
, _ theory of linear operators in hilbert space _ , dover publications , inc . ,
new york , 1993 . translated from the russian and with a preface by merlynd nestell , reprint of the 1961 and 1963 translations , two volumes bound as one . , _ the unitary gas and its symmetry properties _ , in the bcs - bec crossover and the unitary fermi gas , w. zwerger , ed .
836 of lecture notes in physics , springer berlin heidelberg , 2012 , pp .
127191 . height 2pt depth -1.6pt width 23pt , _
@xmath0-particle systems with zero - range interactions _ , in proc .
locarno conf .
`` stochastic processes , physics and geometry ii '' , s. albeverio , u. cattaneo , and d. merlini , eds .
, world scientific , singapore , 1995 , pp .
138145 . , _ mathematical aspects of the three - body problem in the quantum scattering theory _ , translated from the russian by ch .
gutfreund . translation edited by l. meroz , israel program for scientific translations jerusalem ; daniel davey & co. , inc .
, new york , 1965 . , _ mathematical theory of non - relativistic two- and three - particle systems with point interactions _ , in cargse lectures in theoretical physics : application of mathematics to problems in theoretical physics ( cargse , 1965 ) , gordon and breach science publ .
, new york , 1967 , pp .
247287 . height 2pt depth -1.6pt width 23pt , _ on pointlike interaction between @xmath0 fermions and another particle _ , in proceedings of the workshop on singular schrdinger operators , trieste 29 september - 1 october 1994 , a. dellantonio , r. figari , and a. teta , eds .
, ilas / fm-16 , 1995 . | for quantum systems of zero - range interaction we discuss the mathematical scheme within which modelling the two - body interaction by means of the physically relevant ultra - violet asymptotics known as the `` ter - martirosyan skornyakov condition '' gives rise to a self - adjoint realisation of the corresponding hamiltonian .
this is done within the self - adjoint extension scheme of kren , viik , and birman .
we show that the ter - martirosyan skornyakov asymptotics is a condition of self - adjointness only when is imposed in suitable functional spaces , and not just as a point - wise asymptotics , and we discuss the consequences of this fact on a model of two identical fermions and a third particle of different nature .
* keywords : * point interactions , self - adjoint extensions , kren - viik - birman theory , ter - martirosyan skornyakov operators . |
the idea of inflation is a very successful paradigm to solve the problems of the standard cosmology and it provides a basis for production and evolution of seeds for large scale structure of the universe [ 1,2 ] . from a thermodynamical viewpoint ,
there are two possible alternatives to inflationary dynamics : standard picture is isentropic inflation referred to as supercooled inflation . in this picture
, universe expands in inflation phase and its temperature decrease rapidly .
when inflation ends , a reheating period introduces radiation into the universe .
the fluctuations in this type of inflation model are zero - point ground state fluctuations and evolution of the inflaton field is governed by ground state evolution equations . in this model
we have not any thermal perturbations and therefore density perturbations are adiabatic ( or curvature ) .
the other picture is a non - isentropic inflation , the so called warm inflation .
warm inflation has no need to introduce reheating phase since interaction between the inflaton and other fields in this scenario produces the radiation energy density . in this picture
, inflation terminates smoothly and radiation regime is dominated without a reheating period .
the fluctuations during warm inflation emerge from some excited states and the evolution of the inflaton has dissipative terms arising from interaction of the inflaton and other fields [ 3,4 ] which affects the inflaton dynamics through a noise term in the equations of motion [ 5,6 ] ( for a comprehensive list of the references on warm inflation , see [ 7 ] ) . the important point in the warm inflation scenario
is that the density fluctuations in this scenario arise from thermal rather than vacuum fluctuations [ 3,6,8 ] and the fluctuations in the radiation produce the entropy ( isocurvature ) perturbations .
the thermal fluctuations during warm inflation lead to production of necessary initial seeds for large scale structure ( lss ) formation .
the warm inflation ends when the radiation is dominated in the universe and the universe enters in a standard big bang phase [ 4,9 ] .
the entropy fluctuations disappear before inflation ends . + the goal of this investigation is to study cosmological perturbations in a braneworld viewpoint of the warm inflation in the presence of interaction between inflaton and modified induced gravity on the brane . among various braneworld scenarios ,
the model proposed by dvali , gabadadze and porrati ( dgp ) [ 10 ] predicts deviations from the standard @xmath2-dimensional gravity even over large distances . in this scenario ,
the transition between four and @xmath3-dimensional gravitational potentials arises due to the presence of an induced gravity term in the brane action .
existence of a higher dimensional embedding space allows for the existence of bulk or brane matter which can certainly influence the cosmological dynamics on the brane . in the dgp setup ,
the bulk is a flat minkowski spacetime , but a reduced gravity term appears on the brane without tension .
this model has a rich phenomenology discussed in [ 11 ] .
maeda , mizuno and torii have constructed a braneworld scenario which combines the randall - sundrum ii ( rs ii ) and the dgp model [ 12 ] . in this combination ,
an induced curvature term appears on the brane in the rs ii scenario which contains an ads bulk .
this model has been called the warped dgp braneworld in the literature [ 13 ] .
the supercooled inflation models in this scenario were studied in minimal and non - minimal cases [ 13 - 17 ] .
warm inflation model on a warped dgp brane in the minimal case has been studied by del campo and herrera [ 18 ] , but here we consider the effects of the non - minimal coupling of the scalar field and modified induced gravity on the brane because one important feature of the inflationary paradigm is the fact that inflaton can interact with other fields such as gravitational sector of the theory .
this interaction is shown by the non - minimal coupling of the inflaton field and modified induced curvature in the spirit of the scalar - tensor theories , which is motivated from several compelling reasons ( for a discussion on the reasons to include an explicit non - minimal coupling between inflaton and gravitational sector in a typical inflation model , see [ 19 ] ) .
in fact , inclusion of the non - minimal coupling in our setup is not just a matter of taste ; it is forced upon us since as has been indicated in [ 19 ] , in most theories used to describe inflationary scenarios , it turns out that a non - vanishing value of the coupling constant can not be avoided .
+ to have a complete treatment of the problem , we consider possible modification of the induced gravity on the brane in the spirit of the @xmath0-gravity .
the main motivation for adopting such a framework is the fact that although inflation is an elegant scenario to resolve some shortcomings of the standard cosmology and provides a causal and predictive theory of structure formation , there are some important and yet unsolved problems in it [ 20 ] .
hierarchy problem , trans - planckian problem and singularity problem are among these problems .
modification of the induced gravity in the spirit of @xmath0 theories may provide a reliable framework to solve at least part of these problems .
in another words , incorporation of the modified induced gravity on the brane in the spirit of higher order gravitational theories may shed more light on these problems .
in fact , @xmath0-gravity is the simplest way to achieve this goal and it is also possible to obtain a nonsingular cosmology in this setup . for a review on @xmath0-gravity and also inflation and cosmic acceleration in modified gravity see [ 21 ] . in this paper
we consider the general form of @xmath0-gravity and we discuss cosmological perturbations in the framework of non - minimal warm inflation on the warped dgp brane .
+ through this paper , a dot on a quantity represents the time derivative and a prim marks derivative with respect to ricci curvature @xmath4 .
consider a 5-dimensional ads bulk spacetime with a single @xmath2-dimensional brane embedded in it .
standard matter , including inflaton field , are localized on the brane but gravity and possibly non - standard matter are free to propagate in the bulk .
the gravity is induced on the brane via interaction of bulk gravitons and matter localized on the brane .
the action of this extension of dgp scenario where brane is foliated in the bulk with warped geometry can be written as follows [ 12 ] @xmath5+\int_{brane}d^{4}x\sqrt{-g}\bigg[\frac{1}{\kappa_{5}^{2 } } k^{\pm}+{\cal{l}}_{brane}(g_{\alpha\beta},\psi)\bigg].\ ] ] in this action the quantities are defined as follows : @xmath6 with @xmath7 are coordinates in the bulk while @xmath8 with @xmath9 are induced coordinates on the brane .
@xmath10 is the 5-dimensional gravitational constant .
@xmath11 and @xmath12 are 5-dimensional ricci scalar and matter lagrangian respectively .
@xmath13 is the trace of the extrinsic curvature on either sides of the brane .
this term is known as the york - gibbons - hawking term [ 22 ] which provides a through framework for imposing suitable boundary conditions on the field equations .
@xmath14 is the effective @xmath2-dimensional lagrangian .
this action is actually a combination of the randall - sundrum ii [ 23 ] and the dgp model [ 10 ] .
consider the brane lagrangian as follows @xmath15 where @xmath16 is a mass parameter , @xmath4 is ricci scalar of the brane , @xmath17 is tension of the brane and @xmath18 is lagrangian of the matters localized on the brane .
assume that bulk contains only a cosmological constant , @xmath19 . with these choices , action ( 1 )
gives either a generalized dgp or a generalized rs ii model : it gives dgp model if @xmath20 and @xmath21 and gives rs ii model if @xmath22 [ 12 ] .
considering a spatially flat frw metric on the brane , the cosmological dynamics on the brane is given by @xmath23,\ ] ] where @xmath24 is corresponding to two possible branches of the solutions in this warped dgp model as a manifestation of the two possible embedding of brane in the bulk .
other quantities are defined as @xmath25^{1/2}$ ] where @xmath26^{1/2}$ ] , @xmath27 with @xmath28 and @xmath29 .
note that by definition , @xmath30 , @xmath31 and @xmath32 is an integration constant where corresponding term in the generalized friedmann equation is called the dark radiation term . since we are interested in the inflation dynamics of our model , we neglect dark radiation term in which follows , on large scales this term slightly suppresses the radiation density perturbations at late times . in a kinetic era
, this suppression is much stronger and drives the density perturbations to zero . ] . in this case ,
generalized friedmann equation ( 3 ) attains the following form @xmath33.\ ] ] this equation is the basis of our forthcoming arguments .
after introducing the warped dgp braneworld scenario , here we consider the case of non - minimal warm inflation in this setup .
we assume that the warm inflation is driven by the non - minimally coupled scalar field @xmath34 with potential @xmath35 on the warped dgp brane , where possible modification of the induced gravity on the brane is taken into account within the general framework of @xmath0-gravity .
the action of this model with a non - minimally coupled scalar field is given as follows @xmath5\ ] ] @xmath36.\ ] ] where @xmath37 is a non - minimal coupling and @xmath0 is a function of the ricci scalar on the brane [ 25 ] .
@xmath38 is lagrangian of the other matters localized on the brane .
variation of the action with respect to @xmath34 gives the equation of motion of the scalar field in this warm inflation scenario @xmath39 @xmath40 is the dissipation coefficient and during inflation period , it is responsible for decay of the scalar field into radiation .
there are several choices for @xmath40 , that is : a constant , a function of the scalar field @xmath34 , a function of temperature @xmath41 and a function of both scalar field and temperature , @xmath42 ( for a recent progress in this direction , see [ 26 ] ) . in the supercooled inflation models @xmath43 and the equation of motion has the standard form @xmath44 .
the energy - momentum tensor of a scalar field non - minimally coupled to induced gravity for a dgp - inspired @xmath0-gravity scenario is @xmath45 @xmath46 so , the energy density and pressure are given by @xmath47 @xmath48,\hspace{1cm}\ ] ] @xmath49 @xmath50 @xmath51.\ ] ] note that @xmath52 and @xmath53 are curvature - dependent parts of the energy density and pressure of the non - minimally coupled scalar field respectively .
the conservation equation for scalar field energy density in this dissipative setup is given by @xmath54 since @xmath40 is responsible for interaction of the scalar field and radiation , it is expected that this coefficient has no dependence on the non - minimal coupling of the scalar field and @xmath0-gravity . the energy density and pressure that contain both scalar field and radiation contributions are @xmath55 and @xmath56 where @xmath57 and @xmath58 are the radiation energy density and pressure respectively .
the conservation equation for the combined system of scalar field and radiation is given by @xmath59 which implies the entropy production .
making use of eq . ( 11 ) and ( 12 ) we get @xmath60 where by equation ( 6 ) , @xmath61 is directly related to the non - minimal coupling .
the basic idea of warm inflation is that radiation production is occurring concurrently with inflationary expansion due to dissipation of the inflaton field system .
the equation of state for radiation field is given by @xmath62 .
therefore , the conservation equation of @xmath63 yields the following result @xmath64 that a part of dynamics is represented by this equation .
+ in the slow - roll approximation where @xmath65 , equation of motion for the scalar field takes the following form @xmath66 where @xmath67 . in warm inflation ,
the radiation production is quasi - stable so that @xmath68 and @xmath69
. therefore we have from ( 15 ) @xmath70 by using equations ( 4 ) , ( 16 ) and ( 17 ) we obtain @xmath71}\bigg],\ ] ] where we have used the definition of the dissipation factor as follows @xmath72 which is a dimensionless parameter .
since during an inflationary era the scalar field energy density dominates over the energy density of the radiation field , that is , @xmath73 , we can assume @xmath74 .
here @xmath75 is the stefan - boltzmann constant and @xmath76 is the number of degrees of freedom for the radiation field , that in the standard cosmology is @xmath77 .
a part of the effects of the non - minimal coupling and dissipation is hidden in the definition of energy density , @xmath78 , which attains the following form by using ( 8) @xmath79,\ ] ] where @xmath80 and @xmath81 . + we define the most important slow - roll parameter as for slow - roll parameter while @xmath24 marks two possible branches of the dgp setup . ] @xmath82 } { \bigg[\rho_{\varphi}+\rho_{0}+\varepsilon\rho_{0}\big({\cal{a}}_{0}^{2}+\frac{2\eta\rho_{\varphi } } { \rho_{0}}\big)^{1/2}\bigg]^{2}},\hspace{.7cm}\ ] ] where by definition @xmath83\ ] ] @xmath84 - 3f'''(r)\frac{\dot{r}^{2}h}{\dot{\varphi}}\bigg).\ ] ] due to complicated form of the equations , here we restrict our analysis to the first order of the non - minimal coupling is constraint to be very close to the conformal coupling , @xmath85 by the recent observations ( see [ 27 ] for instance ) . ] , @xmath86 . since @xmath87 itself is multiplied by @xmath86 in equation ( 21 ) , to have a first order analysis we should consider those terms of @xmath87 that are independent of @xmath86 .
so , we should consider only the terms independent of @xmath86 in the definition of @xmath88 and @xmath89 .
these terms are @xmath90 and @xmath91 , respectively .
+ in comparison with minimal warm inflation on dgp brane as presented in ref .
[ 18 ] , we see that our equation ( 21 ) reduces to equation ( 15 ) of ref . [ 18 ] for @xmath92 . on the other hand ,
for @xmath93 we recover typical expression for non - minimal supercool inflation in the warped dgp brane . the relation between energy densities of radiation and inflaton fields
can be calculated using the slow - roll parameter @xmath94 to find @xmath95}{1+\varepsilon \eta\big({\cal{a}}_{0}^{2}+\frac{2\eta\rho_{\varphi}}{\rho_{0}}\big)^{-1/2}}\bigg].\ ] ] the inflation takes place when the condition @xmath96 ( or equivalently @xmath97 ) is fulfilled .
this condition in our case reduces to the following expression for realization of the warm inflation in our non - minimal setup @xmath98}{1+\varepsilon \eta\big({\cal{a}}_{0}^{2}+\frac{2\eta\rho_{\varphi}}{\rho_{0}}\big)^{-1/2}}\bigg ] > \frac{2(1+r)}{r}\rho_{\gamma}\,\frac{\big(v_{,\varphi}^{2}+\xi\beta\big ) } { \big(v_{,\varphi}-\xi f(r)\varphi\big)^{2}}.\ ] ] the warm inflationary period lasts up to violation of this condition .
the inflationary epoch ends when the @xmath99 is fulfilled and this implies that @xmath98}{1+\varepsilon \eta\big({\cal{a}}_{0}^{2}+\frac{2\eta\rho_{\varphi}}{\rho_{0}}\big)^{-1/2}}\bigg]= \frac{2(1+r)}{r}\rho_{\gamma}\,\frac{\big(v_{,\varphi}^{2}+\xi\beta\big ) } { \big(v_{,\varphi}-\xi f(r)\varphi\big)^{2}}.\ ] ] the second slow - roll parameter in this setup is given by @xmath100\ ] ] @xmath101^{-1}.\ ] ] in the minimal case one has only the first two terms of the right hand side of this expression .
note also that in our setup we consider @xmath102 or equivalently @xmath103 .
the number of e - folds , @xmath104 in the presence of the non - minimal coupling and for a warped dgp - inspired @xmath0-gravity can be written as @xmath105 @xmath106d\varphi.\ ] ] where @xmath107 denotes the value of the scalar field @xmath34 when universe scale observed today crosses the hubble horizon during inflation , and @xmath108 is the value of the scalar field when the universe exits the inflationary phase .
the inhomogeneous perturbations of the frw background are described by the metric in the longitudinal gauge [ 28 , 29 ] @xmath109 where @xmath110 is the scale factor on the brane , @xmath111 and @xmath112 are the metric perturbations .
the radiation and scalar fields interact through the friction term @xmath40 .
the spatial dependence of all perturbed quantities are of the form of plane waves @xmath113 , where @xmath114 is the wave number .
a perturbation of the metric implies , through einstein s equations of motion , a perturbation in the energy - momentum tensor .
the energy - momentum tensor in our setup as defined in equation ( 7 ) is diagonal if we note that @xmath4 is just a function of the cosmic time [ 30 ] .
we note that the two metric perturbations are not equal due to the presence of the anisotropic stress perturbation .
the perturbed weyl contribution to the einstein equations can be parametrized as an effective fluid with anisotropic stress perturbation , and this contribution can not be set to zero .
+ the perturbed field equations can be obtained straightforwardly from einstein field equations . in a warped dgp braneworld model ,
the einstein field equations change to effective equations on the brane given as [ 12 ] @xmath115 where @xmath116 and @xmath117 and @xmath118 is the total stress - tensor on the brane .
also we have @xmath119 where @xmath120 is the five dimensional weyl tensor and @xmath121 is the spacelike unit vector normal to the brane .
the friedmann equation ( 3 ) can be calculated directly from these equations ( see [ 31 ] for instance ) .
so , to obtain perturbed field equations , if we adopt the standard prescription as has been presented in ref . [ 32 ] , we should replace the quantities in the standard picture with corresponding effective quantities .
we note that in the background spacetime , @xmath122 and we can use equation ( 4 ) as friedmann equation in this setup .
but the perturbed frw brane has a nonzero @xmath123 , which encodes the effects of the bulk gravitational field on the brane [ 33 ] and we have use the friedmann equation ( 3 ) where @xmath124 [ 12 ] .
the perturbed @xmath125 field equations are needed to determine the evolution of @xmath126 . in this manner ,
the temporal part of the perturbed field equations are given as @xmath127 @xmath128 @xmath129 @xmath130 where @xmath131 appears from the decomposition of the velocity field as @xmath132 [ 29 ] and we have omitted the subscript @xmath114 .
the last equation is related to the perturbed effective einstein equation for the component @xmath133 that @xmath134 $ ] .
equations ( 32 ) and ( 33 ) above have their standard forms but now in the non - minimal dgp brane world model .
the perturbed total energy density and pressure in longitudinal gauge can be written as @xmath135 and @xmath136 respectively .
the second terms on the right hand side of these two equations are hallmark of the warm inflation , because a perturbation of the metric leads to a perturbation in the stress energy - momentum tensor and in the warm inflationary model the stress - momentum tensor contains the radiation field too . in the dgp brane world model by using the friedmann equation ( 3 ) , one can define an effective gravitational energy density and pressure as @xmath137^{1/2},\ ] ] and @xmath138 ^{-1/2}\ ] ] @xmath139^{1/2}\bigg),\ ] ] respectively where @xmath140 and @xmath141 obey the standard friedmann equation .
in other words , using the standard friedmann equation as @xmath142 and substituting for @xmath143 from equation ( 38 ) , we recover the equation ( 3 ) . the effective pressure is then calculated by using the continuity equation .
now we can rewrite equations .
( 36 ) and ( 37 ) for dgp brane world model as @xmath144^{-1/2 } \big(\delta\rho_{\varphi}-\mu^{2}\delta e_{\,0}^{0}\big)+\delta\rho_{\gamma},\ ] ] @xmath145^{-1/2}\big(\delta p_{\varphi}-\frac{1}{3}\mu^{2}\delta e_{\,0}^{0}\big)\ ] ] @xmath146 ^{-3/2}\big(\delta\rho_{\varphi}-\mu^{2}\delta e_{\,0}^{0}\big)+\frac{1}{3}\delta\rho_{\gamma},\ ] ] where @xmath147 can calculated from the general equation @xmath148 as @xmath149 where @xmath150 can parametrize as an effective fluid , with density perturbation @xmath151 , isotropic pressure perturbation @xmath152 , anisotropic stress perturbation @xmath153 and energy flux perturbation @xmath154 .
indeed , the perturbed weyl contribution to the einstein equations can be parametrized as an effective fluid with anisotropic stress perturbation ( for details , see [ 33 ] ) .
@xmath155 and @xmath156 contain the effects of the non - minimal coupling @xmath157 @xmath158 the last terms in both of these relations are related to the non - minimal coupling of the scalar filed and induced gravity on the brane and can be calculated as follows @xmath159 @xmath160 where @xmath161 so that @xmath162 and @xmath163 are defined as follows @xmath164 and @xmath165 @xmath166\bigg].\ ] ] we need the following relation to calculate equation ( 47 ) explicitly @xmath167\ ] ] to have a complete set of equations for treating perturbations , we perturb equations ( 6 ) and ( 15 ) to find @xmath168 @xmath169 @xmath170\ ] ] @xmath171 we study the effects of the non - minimal coupling of the scalar field and modified induced gravity on the brane in the warm inflation and we compare our results with the minimal case .
note that equation ( 52 ) is the same the corresponding equation for minimal case but other equations mentioned above are changed considerably .
to interpret the evolution of the cosmological perturbations , the scalar perturbations can be decomposed so that : a ) the projection orthogonal to the trajectory which is called entropy or isocurvature perturbation is generated if inflation is driven by more than one scalar field , and b ) the parallel projection corresponds to the adiabatic or curvature perturbations and this type of perturbations are generated if the inflaton field is the only field in inflation period [ 34,35 ] .
note however that these perturbations might even be cross - correlated to the entropy ones [ 36 - 38 ] .
+ since warm inflation paradigm includes two interacting fields , isocurvature ( entropy ) perturbations are expected to be generated due to thermal fluctuations in the radiation field since the scalar and radiation fields interact in a thermal bath [ 32,39,40 ] . + for treating entropy perturbations , we note that @xmath172 and @xmath173 are related together via entropy perturbation @xmath174 [ 32,41 ] @xmath175 where @xmath176 is the sound effective velocity in the fluid composed of the radiation and scalar field non - minimally coupled to modified induced gravity on the warped dgp brane .
@xmath177 is the non - adiabatic pressure perturbation , @xmath178 , which is due to variation of the total equation of state that relates @xmath179 and @xmath143 .
the entropy perturbation @xmath180 represents the displacement between hypersurfaces of uniform pressure and density .
+ using equations ( 40)-(44 ) in equation ( 53 ) , we have @xmath181^{-1/2}\bigg)\ ] ] @xmath182 ^{-1/2}\ ] ] @xmath183 ^{-3/2}\big(\delta\rho_{\varphi}-\mu^{2}\delta e_{\,0}^{0}\big).\ ] ] note that if we set @xmath184 , this expression reduces to the standard model result and all traces of the dgp setup will disappear
. @xmath78 , @xmath185 and @xmath155 have been defined by ( 8) , ( 9 ) and ( 43 ) . using the equations ( 32 ) , ( 33 ) and ( 34 ) we can rewrite this relation as @xmath186 @xmath187 where @xmath188 , @xmath189 and @xmath190 in the non - minimal case are defined as @xmath191\big({\cal{z}}-1\big)}\ ] ] @xmath192 @xmath193^{3/2}{\cal{z}}\bigg)^{-1},\ ] ] and @xmath194^{-1/2}\bigg),\ ] ] respectively .
we use the slow - roll approximation and quasi - stable conditions ( 16 ) and ( 17 ) to write @xmath195{\cal{z}}}{\big(v_{,\varphi}\dot{\varphi}+\dot{\rho_{\xi}}\big){\cal{z } } + 4\mu^{2}e_{\,0}^{0}h\big({\cal{z}}-1\big)}\ ] ] @xmath196 as is obvious from equation ( 55 ) , in addition to dissipation , the non - minimal coupling of the scalar field and modified induced gravity on the brane has a crucial role in the shape of the entropy perturbations ; it is seen in the first two terms ( where a part of the effects of the non - minimal coupling is hidden in the definition of the sound effective velocity , @xmath197 ) and in the last three terms of this equation . in the minimal case ,
equation ( 59 ) leads to @xmath198 which contains dissipation effect in the dgp model .
however , in the presence of the non - minimal coupling between induced gravity and the scalar field , both non - minimal coupling and dissipation affect dynamics of these perturbations in relatively complicated manner . in the minimal case and within the standard model ,
if we consider a small dissipation by setting @xmath199 , the entropy perturbation vanishes for long wavelength and the primordial spectrum of perturbation is due to adiabatic perturbations [ 32 ] .
but in our case , if we set @xmath199 , for long wavelength that @xmath200 , the entropy perturbation is given by @xmath201 @xmath202 where @xmath203{\cal{z}}}{\big(v_{,\varphi}\dot{\varphi}+\dot{\rho_{\xi}}\big){\cal{z } } + 4\mu^{2}e_{\,0}^{0}h\big({\cal{z}}-1\big)}\ ] ] @xmath196 in the absence of the non - minimal coupling , the entropy perturbation reduces to the result of the minimal setup [ 18 ] @xmath204\ ] ] @xmath205 in the minimal standard theory of cosmological perturbations , when there are no dissipation effects , all perturbations are adiabatic and there is no trace of the non - adiabatic perturbations .
but , as we have shown here , in a dgp - inspired non - minimal setup , in the absence of dissipations there is a non - vanishing contribution of the non - adiabatic perturbations .
we note that our inspection shows that this effect is mainly as a result of dgp than non - minimal coupling .
the effect of the non - minimal coupling tends to increase the contribution of the entropy perturbations .
+ the curvature perturbation on a uniform density hypersurface is defined as [ 42,43 ] @xmath206 using the acceleration equation @xmath207 where by definition @xmath208 , the curvature perturbation can be written as @xmath209 from this equation , we deduce [ 44 ] @xmath210 which implies that @xmath211 is a constant if the pressure perturbation is adiabatic on the large scales .
this equation relates the change in the comoving curvature perturbation due to the source @xmath177 ( or equivalently @xmath212 .
using ( 55 ) for long wavelength perturbations , @xmath213 is given by @xmath214 @xmath215 where @xmath216 .
in contrast to the minimal standard case , the entropy perturbations depend not only on the dissipation effects but also they depends on the non - minimal coupling of the scalar field and induced gravity in the dgp setup .
in other words , even with small dissipation , the entropy perturbations are important yet .
it was expected a priori , based on the standard picture , that in the absence of dissipation the perturbation should be adiabatic since just one field is present . however , in our non - minimal dgp - inspired model with modified induced gravity the effects of the non - adiabatic perturbations are present yet and in this case curvature perturbations can not be constant in time and they attain an explicit time - dependence . these are new results for the rest of the theory of cosmology perturbations .
we note that isocurvature perturbations are free to evolve on superhorizon scales , and the amplitude at the present day depends on the details of the entire cosmological evolution from the time that they are formed . on the other hand , because all super - hubble radius perturbations evolve in the same way , the shape of the isocurvature perturbation spectrum is preserved during this evolution [ 45 ] .
in the previous section , we have shown that in the warm inflationary model the entropy perturbations are generated since inflaton and radiation fields interact with each other . here
we are going to obtain scalar and tensorial perturbation for warm inflation and we expect that for @xmath43 , the results of cool inflation will be recovered .
+ we take into account the slow - roll approximation at the large scales , @xmath217 , where we need to describe the non - decreasing adiabatic and entropy modes . in this situation ,
equations ( 51 ) and ( 52 ) become respectively @xmath218 @xmath219 and equation ( 34 ) takes the following form @xmath220^{-1 } \dot{\varphi}\delta\varphi,\ ] ] where we have used the relation of the velocity field as @xmath221 and @xmath222 . now
solve these three equations to find the desired relations .
first , by substituting equation ( 69 ) into equation ( 67 ) , we find @xmath223 @xmath224^{-1}\dot{\varphi}\delta\varphi .\ ] ] following [ 45 ] , we define an auxiliary function as @xmath225 therefore , equation ( 70 ) can be rewritten as @xmath226\ ] ] @xmath227\big[1+\xi\frac { \rho_{\varphi}}{\rho_{0}\eta\mu^{2}}\varphi^{2}f'(r)\big]^{-1}\ ] ] @xmath228^{1/2}\bigg]^{-1}.\ ] ] a solution of this equation is given as @xmath229 , where @xmath230 is an integration constant . from equation ( 71 ) , @xmath231 is given by @xmath232\ ] ] @xmath227\big[1+\xi\frac { \rho_{\varphi}}{\rho_{0}\eta\mu^{2}}\varphi^{2}f'(r)\big]^{-1}\ ] ] @xmath228^{1/2}\bigg]^{-1}\bigg)d \varphi\bigg].\ ] ] for simplicity we define the following quantity @xmath233\ ] ] @xmath227\big[1 + \xi\frac{\rho_{\varphi}}{\rho_{0}\eta\mu^{2}}\varphi^{2}f'(r)\big]^{-1}\ ] ] @xmath228^{1/2}\bigg]^{-1}\bigg)d \varphi.\ ] ] with this definition , equation ( 73 ) can be rewritten as @xmath234,\ ] ] and therefore , the density perturbation is given by ( see [ 45 ] ) @xmath235}{v_{,\varphi}}\delta\varphi.\ ] ] we note that the main result here is the presence of a non - adiabatic pressure contribution due to the entropy perturbation .
this pressure controls the evolution of the curvature perturbation on large scales during inflation .
in fact in the presence of entropy perturbations the primordial curvature perturbation is not constant after horizon crossing , so the relevant value to be compared with observations should be evaluated at most at the end of inflation . in this respect , this quantity should be evaluated at the end of inflation [ 45 ] . for @xmath43 ,
this relation reduces to the density perturbation in a cool non - minimal inflation model in the framework of dgp - inspired modified gravity . in the high dissipation regime , @xmath236 , the fluctuations in the warm inflationary model
generate by thermal interactions rather than quantum fluctuations [ 46 ] @xmath237 where the freeze - out scale at which dissipation damps out the thermally excited fluctuations is defined as @xmath238 although we have used the usual spectrum for the field perturbations in warm inflation , but the presence of the non - minimal coupling is hidden in the hubble parameter @xmath88 .
given that this spectrum have been derived in a complete different set - up , to begin with without coupling of the scalar field to the ricci scalar , it may hold in the scenario studied in this work if we use the form of @xmath88 dependent on the non - minimal coupling .
it is obvious that modified induced gravity on the brane shows itself in the friedmann equation and hence @xmath88
. now equations ( 74 ) and ( 75 ) for @xmath236 can be rewritten as follows @xmath239\ ] ] @xmath240\big[1 + \xi\frac{\rho_{\varphi}}{\rho_{0}\eta\mu^{2}}\varphi^{2}f'(r)\big]^{-1}\ ] ] @xmath228^{1/2}\bigg]^{-1}\bigg)d \varphi,\ ] ] and @xmath241\frac{h\sqrt{3r}t } { v_{,\varphi}^{2}},\ ] ] where a hat on a quantity shows that quantity is computed in the high dissipation regime .
one important quantity in the inflationary cosmology is the scalar spectral index defined as follows @xmath242 in our setup , this quantity in the high - dissipation regime , @xmath243 , becomes @xmath244\ ] ] @xmath228^{1/2}\bigg]^{-1},\ ] ] where @xmath245 is the integrand of equation ( 77 ) .
the running of the spectral index in our setup is given as follows @xmath246 @xmath247 \big(\xi f(r)\varphi - v_{,\varphi}\big)\bigg[\rho_{\varphi}+\rho_{0}+ \varepsilon\rho_{0}\big[{\cal{a}}_{0}^{2}+\frac{2 \eta}{\rho_{0}}\big(\rho_{\varphi}-\mu^{2}e_{\,0}^{0}\big)\big]^{1/2}\bigg]^{-1}\ ] ] @xmath248\big(\xi f(r)\varphi - v_{,\varphi}\big)^{2}\ ] ] @xmath249^{1/2}\bigg]^{-2}.\ ] ] for an inflationary model driven by just one scalar field , the running of the spectral index constraint by the wmap5+sdss+snia combined data is @xmath250 , with @xmath251 cl ( see for instance [ 47 ] and references therein ) . in our case , we see that dissipative effects , modified induced gravity and the non - minimal coupling of the scalar field and induced gravity have the potential to produce a variety of spectra ranging between red and blue ( see [ 3,6,8,17,32,40 ] for realization of these spectral index in different scenarios ) .
now we pay attention to the tensorial perturbations . as it has been mentioned in ref.[48 ] , the generation of tensor perturbations during inflation period produces stimulated emission in the thermal background of gravitational waves .
this process changes the power spectrum of the tensor modes by an extra , temperature - dependent factor given by @xmath252 .
so , the spectrum of tensor perturbations is given by @xmath253^{1/2}\bigg].\ ] ] using equations ( 78 ) and ( 82 ) , in the limit of @xmath254 the tensor to scalar ratio is given by @xmath255\coth\big(\frac{k}{2t}\big)\bigg]_{k = k_{i}},\ ] ] where @xmath256 and @xmath257 denotes the value of @xmath114 when universe scale crosses the hubble horizon during inflation .
the wmap5+sdss+snia combined data gives the values of the scalar curvature spectrum as @xmath258 at @xmath259 and the tensor to scalar ratio at this value of @xmath257 as @xmath260 [ 49 ] .
evidently , these values will set severe constraints on the parameters of our model some of which are studied in the next section .
now we study numerically the case with the following scalar field potential @xmath261 where @xmath262 and @xmath263 are constants .
we consider a modified gravity model with @xmath264 , where @xmath265 and @xmath266 are constant [ 50 ] .
we set also @xmath267^{\frac{1}{2}}$ ] , ( see [ 18 ] ) , and we will restrict ourselves to the high dissipation regime where @xmath254 . in our presentation of the numerical results we take a dissipative coefficient proportional to the hubble parameter , so that @xmath268 is constant . in our calculations we use @xmath269 since we consider the high dissipation regime , @xmath254 . as we will show , the value of the dissipation coefficient has some impact on the results as it is usually the case in the standard warm inflation .
we will come back to the role played by dissipation shortly . from equation ( 78 ) , the scalar power spectrum in our model with exponential potential ( 84 ) is given by @xmath270\big[x+1+\varepsilon \omega_{x}\big]^{\frac{1}{4}}\big[1+\frac{1}{\upsilon}(1+\varepsilon\omega)\big]^ { \frac{1}{4}}\bigg]_{k = k_{i}},\ ] ] where by definition @xmath271^{-1}\big(1+\varepsilon \eta\omega^{-1}\big)+\frac{9}{8}\big[\sqrt{\frac{2}{p}}\frac{\rho_{0}\upsilon } { \mu}+\xi\varphi\big(f_{0}\big(1+\frac{2}{3}n\big)r^{n}\big)\big]\ ] ] @xmath272^{-\frac{3}{2}}\big [ 1+\varepsilon\eta\omega^{-1}\big]\big(\sqrt{\frac{2}{p}}\frac{\rho_{0 } \upsilon}{\mu}+\xi f_{0}r^{n}\varphi\big)\big[x+1+\varepsilon \omega_{x}\big]^{-\frac{1}{2}}\ ] ] @xmath273^{\frac{1}{2}}\big[1+\frac{1}{\upsilon}(1+\varepsilon\omega)\big]^ { -\frac{1}{2}}\bigg]\ ] ] @xmath274^{-1}\big[x+1+\varepsilon \omega_{x}\big]^{-1}\bigg)d\varphi,\ ] ] and other quantities are defined as follows @xmath275 from equation ( 83 ) , the tensor to scalar ratio in this setup is given by @xmath276\big[x+1+\varepsilon \omega_{x}\big]^{\frac{3}{4}}\big[1+\frac{1}{\upsilon } ( 1+\varepsilon\omega)\big]^{-\frac{1}{4}}\coth\big(\frac{k}{2t}\big)\bigg]_{k = k_{i}}.\ ] ] equations ( 85 ) and ( 88 ) with the definitions ( 86 ) and ( 87 ) are very complicated and to have an intuition , we have to study these quantities numerically . using the appropriate values of @xmath277 and @xmath278 as mentioned previously , equations ( 85 ) and ( 88 ) lead us to the following result @xmath279.\ ] ] here the subscript @xmath280 means that the corresponding quantity should be calculated at @xmath281 where we set @xmath259 . from equation ( 89 )
we get @xmath282 where @xmath283 and @xmath284,@xmath285 . in this analysis , we set @xmath286 for both dgp - branches of the model . also we set @xmath287 , @xmath288 , @xmath259 , @xmath289 and @xmath290 .
note that important quantities such as @xmath291 and @xmath292 now depend on the parameters of the model in this warped dgp - inspired framework .
the results of our numerical calculations are shown in figures @xmath293 , @xmath294 , @xmath295 and @xmath2 .
figure @xmath293 shows the spectral index versus @xmath296 for normal ( @xmath297 ) branch of the model .
depending on the values of the conformal coupling , it is possible to have both red and blue spectrum in this model .
we note that positive values of the non - minimal coupling give more reliable results in comparison with observations ( this is supported from other viewpoints too ; see for instance [ 27 ] ) . with positive @xmath86
, our model favors only the red power spectrum .
figure @xmath294 shows the running of the spectral index in normal branch of the model . for @xmath298 , the calculated running in our model
can not be compared with observations and therefore is excluded from our consideration .
figures @xmath295 and @xmath2 show the corresponding results for the self - accelerating branch of the scenario . again
, negative values of the non - minimal coupling are excluded on observational grounds .
these values are essentially related to anti - gravitation .
now we focus on the effect of dissipation on the inflation parameters .
figure 5 shows the variation of the slow - roll parameter @xmath94 versus @xmath299 for different values of the dissipation factor , @xmath268 .
as this figure shows , @xmath94 decreases by increasing the dissipation effect for a fixed value of @xmath299 . in figure 6
we have considered the situation for a continuous variation of the dissipation parameter .
again this figure shows that the slow - roll parameter decreases by increasing @xmath268 .
however , @xmath94 increases by increasing @xmath299 where @xmath300 . now to see the effect of dissipation on the number of e - folds , we define @xmath301\,,\ ] ] which is the integrand of the number of e - folds defined as ( 27 ) .
we rephrase this quantity and also equation ( 27 ) versus @xmath299 .
figure 7 shows the variation of @xmath302 versus @xmath299 for different values of the dissipation factor , @xmath268 .
the number of e - folds is given by the surface enclosed between this curve and two vertical lines located at @xmath303 and @xmath304 corresponding to @xmath107 and @xmath108 respectively .
as this figure shows , the number of e - folds increases by increasing dissipation factor . finally , we note that the self - accelerating solution of the dgp setup is unstable due to existence of ghosts [ 51,52 ] . in our setup ,
incorporation of several new degrees of freedom such as modified induced gravity , non - minimal coupling and the warped geometry of the bulk has provided a relatively wider parameter space .
this wider parameter space may provide a better framework to treat instabilities of the self - accelerating solution .
however this is not an easy task and lies out of our interest here ( for a recent progress in this direction see [ 53 ] ) .
in this paper we have studied cosmological perturbations and their evolution in a braneworld viewpoint of the warm inflation in the presence of interaction between inflaton and induced gravity on the brane .
we have incorporated possible modification of the induced gravity on the brane in the spirit of @xmath0-gravity .
the cosmological perturbations are treated with complete details and the roles played by modification of the induced gravity , dissipation and the non - minimal coupling are discussed .
the main results of our study can be summarized as follows : in the minimal standard theory of cosmological perturbations , when there are no dissipation effects , all perturbations are adiabatic and there is no trace of the non - adiabatic perturbations .
but , as we have shown here , in a dgp - inspired non - minimal setup with modified induced gravity , in the absence of dissipation there is a non - vanishing contribution of the non - adiabatic perturbations . in this setup
, this effect is mainly as a result of dgp ( with modified induced gravity ) than the non - minimal coupling .
the effect of the non - minimal coupling tends to increase the contribution of the entropy perturbations .
the numerical analysis of the parameter space ( which is wide enough due to incorporation of several new degrees of freedom ) shows that it is possible to have both red and blue spectrum and relatively large running of the spectral index depending on the sign of the non - minimal coupling . however , negative values of the non - minimal coupling give results that are not supported by observations and therefore should be excluded from our considerations .
the effect of dissipation is so that the slow - roll parameter @xmath94 decreases by increasing the dissipation factor .
however , the number of e - folds increases by increasing the dissipation factor .
modification of the induced gravity and its coupling to the inflaton field in a warm inflation framework brings a variety of new possibilities which constraint our model in comparison with the recent observations . in other words ,
this wider parameter space provides much more freedom than single , self - interacting scalar field inflation to fit with the observational data .
we are indebted to an anonymous referee for insightful suggestions and comments .
part of this work has been done during b. f. sabbatical leave at the infn institute , frascati , italy .
she would like to express her deepest appreciation to the members of this institute , especially _
professor stefano bellucci _ for kind hospitality and supports .
the work of k. n. is supported financially by the research institute for astronomy and astrophysics of maragha , iran .
10 a. liddle and d. lyth , _
cosmological inflation and large - scale structure _ , cambridge university press , 2000 .
r. h. brandenberger , [ arxiv : hep - th/0509099 ] ; b. a. bassett , s. tsujikawa and d. wands , _ rev .
phys . _ * 78 * , 537 ( 2006 ) , [ arxiv : astro - ph/0507632 ] ; j. e. lidsey _
et al _ , _ rev .
phys . _ * 69 * , 373 ( 1997 ) .
a. berera and l. z. fang , _ phys . rev .
lett . _ * 74 * , 1912 ( 1995 ) .
a. berera , _ phys .
* 75 * , 3218 ( 1995 ) .
a. berera , _ phys .
d _ * 54 * , 2519 ( 1996 ) .
a. berera , _ nucl .
phys b _ * 585 * , 666 ( 2000 ) , [ arxiv : hep - ph/9904409 ] .
a. berera , _ contemporary phys .
_ * 47 * , 33 ( 2006 ) , [ arxiv:0809.4198 ] ; a. berera , i. g. moss and r. o. ramos , _ rept .
phys . _ * 72 * , 026901 ( 2009 ) , [ arxiv:0808.1855 ] ; m. bastero - gil and a. berera , [ arxiv:0902.0521 ] ; a. berera , m. gleiser and r. o. ramos , _ phys . rev
. lett . _ * 83 * , 264 ( 1999 ) , [ arxiv : hep - ph/9809583 ] ; m. bellini , _ class .
grav . _ * 16*,2393 ( 1999 ) , [ arxiv : gr - qc/9904072 ] ; m. bellini , _ nucl .
b _ * 563 * , 245 ( 1999 ) , [ arxiv : gr - qc/9908063 ] ; s. gupta , a. berera , a. f. heavens and s. matarrese , _ phys . rev . d _ * 66 * , 043510 ( 2002 ) , [ arxiv : astro - ph/0205152 ] ; r. h. brandenberger and m. yamaguchi , _ phys . rev .
d _ * 68 * , 023505 ( 2003 ) , [ arxiv : hep - ph/0301270 ] ; l. m. h. hall , ian g. moss and a. berera , _ phys .
d _ * 69 * , 083525 ( 2004 ) , [ arxiv : astro - ph/0305015 ] ; l. m. h. hall , i. g. moss and a. berera , _ phys .
b _ * 589 * , 1 ( 2004 ) , [ arxiv : astro - ph/0402299 ] ; s. gupta , _ phys . rev
. d _ * 73 * , 083514 ( 2006 ) , [ arxiv : astro - ph/0509676 ] ; a. berera , _ grav .
cosmol . _ * 11 * , 51 ( 2005 ) , [ arxiv : hep - ph/0604124 ] ; m. bastero - gil and a. berera , [ arxiv : hep - ph/0610343 ] ; j. c. bueno sanchez , m. bastero - gil , a. berera and k. dimopoulos , [ arxiv:0802.4354 ] ; a. berera , _ contemporary physics _ * 47 * , 33 ( 2006 ) , [ arxiv:0809.4198 ] ; a. berera , i. g. moss and r. o. ramos , [ arxiv:0808.1855 ] ; i. g moss and c. xiong , [ arxiv:0808.0261 ] ; j. m. romero and m. bellini , _
nuovo cim .
b _ * 124 * ( 2009 ) 861 ; r. herrera , [ arxiv:1006.1299 ] .
i. g. moss , _ phys .
b _ * 154 * , 120 ( 1985 ) .
a. berera , _ phys .
d _ * 55 * , 3346 ( 1997 ) .
g. dvali , g. gabadadze and m. porrati , _ phys .
b _ * 485 * , 208 ( 2000 ) , [ arxiv : hep - th/0005016 ] .
a. lue , _ physics reports _ * 423 * , 1 ( 2006 ) [ arxiv : astro - ph/0510068 ] .
kei - ichi maeda , s. mizuno and t. torii , _ phys .
d _ * 68 * , 024033 ( 2003 ) , [ arxiv : gr - qc/0303039 ] .
cai and h. zhang , _ jcap _ * 0408 * , 017 ( 2004 ) , [ arxiv : hep - th/0403234 ] . m. bouhamdi - lopez , r. maartens and d. wands , _ phys .
d _ * 70 * , 123519 ( 2004 ) , [ arxiv : hep - th/0407162 ] .
e. papantonopoulos and v. zamarias , _ jcap _ * 0410 * , 001 ( 2004 ) , [ arxiv : gr - qc/0403090 ] .
h. zhang and z. zhu , _ phys .
b _ * 641 * , 405 ( 2006 ) , [ arxiv : astro - ph/0602579 ] . k. nozari and b. fazlpour , _ jcap _ * 11 * , 006 ( 2007 ) , [ arxiv:0708.1916 ] . s. del campo and r. herrera , _ phys .
b _ * 653 * , 122 ( 2007 ) , [ arxiv : gr - qc/0708.1460 ] .
v. faraoni , _ phys .
d _ * 53 * , 6813 ( 1996 ) ; v. faraoni , phys . rev .
d * 62 * , 023504 ( 2000 ) , [ arxiv : gr - qc/0002091 ] .
r. h. brandenberger , [ arxiv : astro - ph/9711106 ] .
see also r. h. brandenberger , _ lect .
notes phys .
_ * 738 * , 393 ( 2008 ) , [ arxiv : hep - th/0701111 ] ; r. h. brandenberger and x. zhang , [ arxiv:0903.2065 ] .
t. p. sotiriou and v. faraoni , [ arxiv : gr - qc/0805.1726 ] . for inflation
in the spirit of modified gravity , see for instance b. chen , m. li , t. wang and y. wang , _ mod .
* 22 * , 1987 ( 2007 ) , [ arxiv : astro - ph/0610514 ] , g. cognola , e. elizalde , s. nojiri , s.d .
odintsov , l. sebastiani and s. zerbini , _ phys .
d _ * 77 * , 046009 ( 2008 ) , [ arxiv:0712.4017 ] ; see also s. nojiri and s. d. odintsov , _ phys . rev .
d _ * 68 * , 123512 ( 2003 ) , [ arxiv : hep - th/0307288 ] . j. w. york , _ phys .
_ * 28 * , 1082 ( 1972 ) ; g. w. gibbons and s. w. hawking , _ phys
d _ * 15 * , 2752 ( 1977 ) .
see also r. dick , _ class .
grav . _ * 18 * , r1 ( 2001 ) , [ arxiv : hep - th/0105320 ] . l. randall and r. sundrum , _ phys .
lett . _ * 83 * , 4690 ( 1999 ) .
b. gumjudpai , r. maartens and c. gordon , _ class .
grav . _ * 20 * , 3295 ( 2003 ) , [ arxiv : gr - qc/0304067 ] .
see also r. maartens , [ arxiv : astro - ph/0402485 ] .
s. nojiri and s. d. odintsov , _ phys .
b _ * 599 * , 137 ( 2004 ) , [ arxiv : astro - ph/0403622 ] ; s. nojiri , s. d. odintsov , p. v. tretyakov , _ prog . theor
_ * 172 * , 81 ( 2008 ) , [ arxiv:0710.5232 ] ; y. zhang , [ arxiv:0903.0685 ] .
m. szydlowski , o. hrycyna and a. kurek , _ phys .
d _ * 77 * , 027302 ( 2008 ) [ arxiv:0710.0366 ] . v. f. mukhanov , h. a. feldman and r. h. brandenberger , _ phys .
_ * 215 * , 203 ( 1992 ) .
j. bardeen , _ phys .
* 22 * , 1882 ( 1980 ) .
a. riotto , [ arxiv : hep - ph/0210162 ] .
k. nozari , _ jcap _ * 09 * , 003 ( 2007 ) , [ arxiv:0708.1611 ] . h. p. de oliveira and s. e. joras , _ phys .
d _ * 64 * , 063513 ( 2001 ) , [ arxiv : gr - qc/0103089 ] . k. koyama and r. maartens , _ jcap _ * 0601 * , 016 ( 2006 ) , [ arxiv : astro - ph/0511634 ] + d. langlois , r. maartens , m. sasaki and d. wands , _ phys .
* 63 * ( 2001 ) 084009 c. gordon , d. wands , b. a. bassett and r. maartens , _ phys .
d _ * 63 * , 023506 ( 2001 ) , [ arxiv : astro - ph/0009131 ] .
d. langlois and f. vernizzi , _ jcap _ * 02 * , 017 ( 2007 ) , [ arxiv : astro - ph/0610064 ] .
n. bartolo , s. matarrese and a. riotto , _ phys .
d _ * 64 * , 083514 ( 2001 ) .
n. bartolo , s. matarrese and a. riotto , _ phys .
d _ * 64 * , 123504 ( 2001 ) .
n. bartolo , s. matarrese , a. riotto and d. wands , _ phys .
d _ * 66 * , 043520 ( 2002 ) .
de oliveira , _ phys .
b _ * 526*,1 ( 2002 ) .
a. a. starobinsky and j. yokoyama , [ arxiv : gr - qc/9502002 ] ; a. a. starobinsky , s. tsujikawa and j. yokoyama , _ nucl .
b _ * 610*,383 ( 2001 ) .
h. kodama and m. sasaki , prog .
78 , 1 ( 1984 ) .
b. a. bassett , f. tamburini , d. i. kaiser and r. maartens , _ nucl.phys .
b _ * 561 * , 188 - 240(1999 ) , [ arxiv : hep - ph/9901319 ] ; c. gordon , d. wands , b. a. basset and r. maartens , _ phys . rev
. d _ * 63 * , 023506 ( 2001 ) , [ arxiv : astro - ph/0009131 ] .
d. polarski and a. a. starobinsky , _ nucl .
b _ * 385 * , 623 ( 1992 ) .
d. wands , k. a. malik , d. h. lyth and a. r. liddle , _ phys .
d _ * 62 * , 043527 ( 2000 ) .
a. r. liddle and a. mazumdar , phys .
* ( 2000 ) 123507 ; m. a. cid , s. d. campo and r. herrera , _ jcap _ * 10 * , 005 ( 2007 ) , [ arxiv : astro - ph/0710.3148 ] .
a. n. taylor and a. berera , _ phys .
d _ * 62 * , 083517 ( 2000 ) .
ma and x. zhang , _ jcap _ * 03 * , 006 ( 2009 ) , [ arxiv:0812.3421 ] .
k. bhattacharya , s. mohanty and a. nautiyal , _ phys .
lett . _ * 97 * , 251301 ( 2006 ) , [ arxiv : astro - ph/0607049 ] .
e. komatsu _ et al .
_ , _ astrophys
. j. suppl . _ * 180 * , 330 ( 2009 ) , [ arxiv:0803.0547 ] .
g. cognola , e. elizalde , s. nojiri , s. d. odintsov and s. zerbini , _ phys .
d _ * 73 * , 084007 ( 2006 ) , [ arxiv : hep - th/0601008 ] ; see also s. nojiri and s. d. odintsov , _ int
. j. geom .
_ * 4 * , 115 ( 2007 ) , [ arxiv : hep - th/0601213 ] .
k. koyama , _ class .
quantum grav . _
* 24 * , r231 ( 2007 ) [ arxiv : hep - th/0709.2399 ] . c. de rham and a. j. tolley , _ jcap _ * 0607 * , 004 ( 2006 ) [ arxiv : hep - th/0605122 ] .
see for instance : m. sami , [ arxiv:0904.3445 ] ; m. cadoni and p. pani , [ arxiv:0812.3010 ] ; y. shtanov , v. sahni , a. shafieloo and a. toporensky , _ jcap _ * 04 * ( 2009 ) 023 , [ arxiv:0901.3074 ] . | we construct a warm inflation model with inflaton field non - minimally coupled to induced gravity on a warped dgp brane .
we incorporate possible modification of the induced gravity on the brane in the spirit of @xmath0-gravity .
we study cosmological perturbations in this setup . in the case of two field inflation such as warm inflation ,
usually entropy perturbations are generated . while it is expected that in the case of one field inflation these perturbations to be removed , we show that even in the absence of the radiation field , entropy perturbations are generated in our setup due to non - minimal coupling and modification of the induced gravity .
we study the effect of dissipation on the inflation parameters of this extended braneworld scenario . + * pacs * : 04.50.-h , 98.80.-k , 98.80.cq , 98.80.es + * key words * : braneworld gravity , scalar - tensor theories , induced gravity , warm inflation , perturbations + , and + + @xmath1_research institute for astronomy and astrophysics of maragha , + p. o. box 55134 - 441 , maragha , iran _ |
this paper focuses on dimension reduction in @xmath0 , in particular , on the method based on _ cauchy random projections _
@xcite , which is special case of _ linear random projections_. the idea of _ linear random projections _ is to multiply the original data matrix @xmath11 with a random projection matrix @xmath12 , resulting in a projected matrix @xmath13 . if @xmath14 , then it should be much more efficient to compute certain summary statistics ( e.g. , pairwise distances ) from @xmath15 as opposed to @xmath5 .
moreover , @xmath15 may be small enough to reside in physical memory while @xmath5 is often too large to fit in the main memory .
the choice of the random projection matrix @xmath16 depends on which norm we would like to work with .
@xcite proposed constructing @xmath16 from i.i.d .
samples of @xmath17-stable distributions , for dimension reduction in @xmath18 ( @xmath19 ) . in the stable distribution family @xcite ,
normal is 2-stable and cauchy is 1-stable .
thus , we will call random projections for @xmath9 and @xmath0 , _ normal random projections _ and _ cauchy random projections _ , respectively . in _ normal random projections _
@xcite , we can estimate the original pairwise @xmath9 distances of @xmath5 directly using the corresponding @xmath9 distances of @xmath15 ( up to a normalizing constant ) .
furthermore , the johnson - lindenstrauss ( jl ) lemma @xcite provides the performance guarantee .
we will review _ normal random projections _ in more detail in section [ sec_intr_rp ] . for _ cauchy random projections _
, we should not use the @xmath0 distance in @xmath15 to approximate the original @xmath0 distance in @xmath5 , as the cauchy distribution does not even have a finite first moment .
the impossibility results @xcite have proved that one can not hope to recover the @xmath0 distance using linear projections and linear estimators ( e.g. , sample mean ) , without incurring large errors .
fortunately , the impossibility results do not rule out nonlinear estimators , which may be still useful in certain applications in data stream computation , information retrieval , learning , and data mining .
@xcite proposed using the sample median ( instead of the sample mean ) in _ cauchy random projections _ and described its application in data stream computation . in this study , we provide three types of nonlinear estimators : the bias - corrected sample median estimator , the bias - corrected geometric mean estimator , and the bias - corrected maximum likelihood estimator .
the sample median estimator and the geometric mean estimator are asymptotically equivalent ( i.e. , both are about @xmath10 efficient as the maximum likelihood estimator ) , but the latter is more accurate at small sample size @xmath8 . furthermore , we derive explicit tail bounds for the bias - corrected geometric mean estimator and establish an analog of the jl lemma for dimension reduction in @xmath0 .
this analog of the jl lemma for @xmath0 is weaker than the classical jl lemma for @xmath9 , as the geometric mean estimator is a non - convex norm and hence is not a metric .
many efficient algorithms , such as some sub - linear time ( using super - linear memory ) nearest neighbor algorithms @xcite , rely on the metric properties ( e.g. , the triangle inequality ) .
nevertheless , nonlinear estimators may be still useful in important scenarios . * _ estimating @xmath0 distances online _
+ the original data matrix @xmath11 requires @xmath20 storage space ; and hence it is often too large for physical memory .
the storage cost of all pairwise distances is @xmath21 , which may be also too large for the memory .
for example , in information retrieval , @xmath22 could be the total number of word types or documents at web scale .
to avoid page fault , it may be more efficient to estimate the distances on the fly from the projected data matrix @xmath15 in the memory . * _ computing all pairwise @xmath0 distances _ + in distance - based clustering and classification applications , we need to compute all pairwise distances in @xmath5 , at the cost of time @xmath23 . using _
cauchy random projections _
, the cost can be reduced to @xmath24 .
because @xmath14 , the savings could be enormous . * _ linear scan nearest neighbor searching _
+ we can always search for the nearest neighbors by linear scans .
when working with the projected data matrix @xmath15 ( which is in the memory ) , the cost of searching for the nearest neighbor for one data point is time @xmath25 , which may be still significantly faster than the sub - linear algorithms working with the original data matrix @xmath5 ( which is often on the disk ) .
we briefly comment on _ coordinate sampling _
, another strategy for dimension reduction . given a data matrix @xmath26 , one can randomly sample @xmath8 columns from @xmath5 and estimate the summary statistics ( including @xmath0 and @xmath9 distances ) . despite its simplicity , there are two major disadvantages in coordinate sampling .
first , there is no performance guarantee . for heavy - tailed data
, we may have to choose @xmath8 very large in order to achieve sufficient accuracy .
second , large datasets are often highly sparse , for example , text data @xcite and market - basket data @xcite . @xcite and @xcite
provide an alternative coordinate sampling strategy , called _ conditional random sampling ( crs ) _ , suitable for sparse data .
for non - sparse data , however , methods based on _ linear random projections _ are superior .
the rest of the paper is organized as follows .
section [ sec_intr_rp ] reviews _ linear random projections_. section [ sec_results ] summarizes the main results for three types of nonlinear estimators .
section [ sec_median ] presents the sample median estimators .
section [ sec_gm ] concerns the geometric mean estimators .
section [ sec_mle ] is devoted to the maximum likelihood estimators .
section [ sec_conclusion ] concludes the paper .
we give a review on _ linear random projections _ , including _
normal _ and _ cauchy random projections_. denote the original data matrix by @xmath27 , i.e. , @xmath22 data points in @xmath28 dimensions .
let @xmath29 be the @xmath30th row of @xmath5 .
let @xmath31 be a random matrix whose entries are i.i.d
. samples of some random variable . the projected data matrix @xmath13 .
denote the entries of @xmath16 by @xmath32 and let @xmath33 be the @xmath30th row of @xmath15 .
then @xmath34 , with entries @xmath35 , i.i.d .
@xmath36 to @xmath8 , where @xmath37 is the @xmath38th column of @xmath16 . for simplicity , we focus on the leading two rows , @xmath39 and @xmath40 , in @xmath5 , and the leading two rows , @xmath41 and @xmath42 , in @xmath15 .
define @xmath43 to be @xmath44 if we sample @xmath45 i.i.d . from a _ stable distribution _
@xcite , then @xmath46 s are also i.i.d .
samples of the same stable distribution with a different scale parameter . in the family of stable distributions ,
normal and cauchy are two important special cases .
when @xmath45 is sampled from the standard normal , i.e. , @xmath47 , i.i.d . , then @xmath48 because a weighted sum of normals is also normal .
denote the squared @xmath9 distance between @xmath39 and @xmath40 by @xmath49 .
we can estimate @xmath50 from the sample squared @xmath9 distance : @xmath51 it is easy to show that ( e.g. , @xcite ) @xmath52 we would like to bound the error probability @xmath53 by @xmath54 . since there are in total @xmath55 pairs among @xmath22 data points , we need to bound the tail probabilities simultaneously for all pairs . by the bonferroni union bound , it suffices if @xmath56 using ( [ eqn_normal_tail ] ) , it suffices if @xmath57 therefore , we obtain one version of the jl lemma : _ if @xmath58 , then with probability at least @xmath59 , the squared @xmath9 distance between any pair of data points ( among @xmath22 data points ) can be approximated within @xmath60 fraction of the truth , using the squared @xmath9 distance of the projected data after normal random projections .
_ many versions of the jl lemma have been proved @xcite .
note that we do not have to use @xmath61 for dimension reduction in @xmath9 .
for example , we can sample @xmath45 from some _ sub - gaussian distributions _ @xcite , in particular , the following _ sparse projection distribution _ : @xmath62 when @xmath63 , @xcite proved the jl lemma for the above sparse projection , which can also be shown by sub - gaussian analysis @xcite .
recently , @xcite proposed _ very sparse random projections _ using @xmath64 in ( [ eqn_subg_rji ] ) , based on two practical considerations :
* @xmath28 should be very large , otherwise there would be no need for dimension reduction .
* the original @xmath9 distance should make engineering sense , in that the second ( or higher ) moments should be bounded ( otherwise various _ term - weighting _ schemes will be applied ) .
based on these two practical assumptions , the projected data are asymptotically normal at a fast rate of convergence when @xmath64 . of course , _ very sparse random projections _ do not have worst case performance guarantees . in _ cauchy random projections _ , we sample @xmath45 i.i.d . from the standard cauchy distribution ,
i.e. , @xmath65 . by the 1-stability of cauchy @xcite ,
we know that @xmath66 that is , the projected differences @xmath67 are also cauchy random variables with the scale parameter being the @xmath0 distance , @xmath68 , in the original space .
recall that a cauchy random variable @xmath69 has the density @xmath70 the easiest way to see the 1-stability is via the characteristic function , @xmath71 for @xmath72 , @xmath73 , ... , @xmath74 ,
@xmath75 , and any constants @xmath76 , @xmath77 , ... , @xmath78 . therefore , in _ cauchy random projections _
, the problem boils down to estimating the cauchy scale parameter of @xmath79 from @xmath8 i.i.d .
samples @xmath80 .
unfortunately , unlike in _ normal random projections _ , we can no longer estimate @xmath81 from the sample mean ( i.e. , @xmath82 ) because @xmath83 .
although the impossibility results @xcite have ruled out estimators that are metrics , there is enough information to recover @xmath81 from @xmath8 samples @xmath43 , with a high accuracy .
for example , @xcite proposed using the sample median as an estimator .
the problem with the sample median estimator is the inaccuracy at small @xmath8 and the difficulty in deriving explicit tail bounds needed for determining the sample size @xmath8 . +
this study focuses on deriving better estimators and explicit tail bounds for _ cauchy random projections_. our main results are summarized in the next section , before we present the detailed derivations .
casual readers may skip these derivations after section [ sec_results ] .
we propose three types of nonlinear estimators : the bias - corrected sample median estimator ( @xmath84 ) , the bias - corrected geometric mean estimator ( @xmath85 ) , and the bias - corrected maximum likelihood estimator ( @xmath86 ) . @xmath84 and @xmath85 are asymptotically equivalent but the latter is more accurate at small sample size @xmath8 . in addition , we derive explicit tail bounds for @xmath85 , from which an analog of the johnson - lindenstrauss ( jl ) lemma for dimension reduction in @xmath0 follows .
asymptotically , both @xmath84 and @xmath85 are @xmath87 efficient compared to the maximum likelihood estimator @xmath86 .
we propose accurate approximations to the distribution and tail bounds of @xmath86 , while the exact closed - form answers are not attainable .
denoted by @xmath84 , the bias - corrected sample median estimator is @xmath88 where @xmath89 here , for convenience , we only consider @xmath90 , @xmath91 = 1 , 2 , 3 , ... some key properties of @xmath84 : * @xmath92 , i.e , @xmath84 is unbiased . * when @xmath93 , the variance of @xmath84 is @xmath94 @xmath95 if @xmath96 .
* as @xmath97 , @xmath84 converges to a normal in distribution @xmath98 denoted by @xmath85 , the bias - corrected geometric mean estimator is defined as @xmath99 important properties of @xmath85 include : * this estimator is a non - convex norm , i.e. , the @xmath18 norm with @xmath100 . *
it is unbiased , i.e. , @xmath101 . *
its variance is ( for @xmath102 ) @xmath103 * for @xmath104 , its tail bounds can be represented in exponential forms @xmath105 * these exponential tail bounds yield an analog of the johnson - lindenstrauss ( jl ) lemma for dimension reduction in @xmath0 : + _ if @xmath106 , then with probability at least @xmath59 , one can recover the original @xmath0 distance between any pair of data points ( among all @xmath22 data points ) within @xmath107 ( @xmath108 ) fraction of the truth , using @xmath85 , i.e. , @xmath109 . _ denoted by @xmath86 , the bias - corrected maximum likelihood estimator is @xmath110 where @xmath111 solves a nonlinear mle equation @xmath112 some properties of @xmath86 : * it is nearly unbiased , @xmath113 . *
its asymptotic variance is @xmath114 i.e. , @xmath115 , @xmath116 , as @xmath117 .
( @xmath118 ) * its distribution can be accurately approximated by an inverse gaussian , at least in the small deviation range .
based on the inverse gaussian approximation , we suggest the following approximate tail bound @xmath119 which has been verified by simulations for the tail probability @xmath120 range .
recall in cauchy random projections , @xmath121 , we denote the leading two rows in @xmath5 by @xmath39 , @xmath40 @xmath122 , and the leading two rows in @xmath15 by @xmath41 , @xmath42 @xmath123 .
our goal is to estimate the @xmath0 distance @xmath124 from @xmath43 , @xmath125 , i.i.d .
it is easy to show ( e.g. , @xcite ) that the population median of @xmath126 is @xmath81 .
therefore , it is natural to consider estimating @xmath81 from the sample median , @xmath127 as illustrated in the following lemma ( proved in appendix [ app_proof_lem_me ] ) , the sample median estimator , @xmath128 , is asymptotically unbiased and normal . for small samples ( e.g. , @xmath129 ) , however , @xmath128 is severely biased .
[ lem_me ] the sample median estimator , @xmath128 , defined in ( [ eqn_def_me ] ) , is asymptotically unbiased and normal @xmath130 when @xmath90 , @xmath91 = 1 , 2 , 3 , ... , the @xmath131 moment of @xmath128 can be represented as @xmath132 if @xmath133 , then @xmath134 .
+ + for simplicity , we only consider @xmath90 when evaluating @xmath135 .
once we know @xmath136 , we can remove the bias of @xmath128 using @xmath137 where the bias correction factor @xmath138 is @xmath139 @xmath138 can be numerically evaluated and tabulated , at least for small @xmath8 . as an infinite sum .
note that @xmath140 , @xmath141 , is the probability density of a beta distribution @xmath142 .
] obviously , @xmath84 is unbiased , i.e. , @xmath92
. its variance would be @xmath143 of course , @xmath85 and @xmath144 are asymptotically equivalent , i.e. , @xmath145 .
figure [ fig_bme ] plots @xmath138 as a function of @xmath8 , indicating that @xmath128 is severely biased when @xmath129 .
when @xmath146 , the bias becomes negligible .
note that , because @xmath147 , the bias correction not only removes the bias of @xmath128 but also reduces its variance .
the sample median is a special case of sample quantile estimators @xcite .
for example , one version of the quantile estimators given by @xcite would be @xmath148 where @xmath149 and @xmath150 are the .75 and .25 sample quantiles of @xmath151 , respectively .
our simulations indicate that @xmath128 actually slightly outperforms @xmath152 .
this is not surprising .
@xmath152 works for any cauchy distribution whose location parameter does not have to be zero , while @xmath128 takes advantage of the fact that the cauchy location parameter is always zero in our case .
this section derives estimators based on the geometric mean , which are more accurate than the sample median estimators .
the geometric mean estimators allow us to derive tail bounds in explicit forms and ( consequently ) an analog of the johnson - lindenstrauss ( jl ) lemma for dimension reduction in @xmath0 .
recall , our goal is to estimate @xmath81 from @xmath8 i.i.d .
samples @xmath153 . to help derive the geometric mean estimators , we first study two nonlinear estimators based on the fractional moment , i.e. , @xmath154 ( @xmath155 ) and the logarithmic moment , i.e , @xmath156 , respectively , as presented in lemma [ lem_d_log ] .
see the proof in appendix [ app_proof_lem_d_log ] .
[ lem_d_log ] assume @xmath157 .
then @xmath158 from which we can derive two biased estimators of @xmath81 from @xmath8 i.i.d .
samples @xmath80 : @xmath159 whose variances are , respectively , @xmath160 the term @xmath161 decreases with decreasing @xmath162 , reaching a limit @xmath163 in other words , the variance of @xmath164 converges to that of @xmath165 as @xmath162 approaches zero . + note that @xmath165 can in fact be written as the _ geometric mean _ : @xmath166 @xmath164 is a non - convex norm ( @xmath167 ) because @xmath168 .
@xmath144 is also a non - convex norm ( the @xmath167 norm as @xmath169 ) .
both @xmath164 and @xmath144 do not satisfy the triangle inequality .
we propose @xmath85 , the bias - corrected geometric mean estimator .
lemma [ lem_d_gm ] derives the moments of @xmath85 , proved in appendix [ app_proof_lem_d_gm ] .
[ lem_d_gm ] @xmath99 is unbiased , with the variance ( valid when @xmath102 ) @xmath170 the third and fourth central moments are ( for @xmath171 and @xmath172 , respectively ) @xmath173 + the higher ( third or fourth ) moments may be useful for approximating the distribution of @xmath85 . in section [ sec_mle ] , we will show how to approximate the distribution of the maximum likelihood estimator by matching the first four moments ( in the leading terms ) .
we could apply the similar technique to approximate @xmath85 .
fortunately , we do not have to do so because we are able to derive the exact tail bounds of @xmath85 in lemma [ lem_d_gm_tail ] , which is proved in appendix [ app_proof_lem_d_gm_tail ] .
[ lem_d_gm_tail ] @xmath174 where @xmath175 @xmath176 where @xmath177 by restricting @xmath178 , the tail bounds can be written in exponential forms : @xmath179 + an analog of the jl bound for @xmath0 follows from the exponential tail bounds ( [ eqn_exp_right ] ) and ( [ eqn_exp_left ] ) . [ lem_jl_l1 ] using @xmath85 with @xmath180 , then with probability at least @xmath59 , the @xmath0 distance , @xmath81 , between any pair of data points ( among @xmath22 data points ) , can be estimated with errors bounded by @xmath181 , i.e. , @xmath182 .
* remarks on lemma [ lem_jl_l1 ] * : ( 1 ) we can replace the constant `` 8 '' in lemma [ lem_jl_l1 ] with better ( i.e. , smaller ) constants for specific values of @xmath183 .
for example , if @xmath184 , we can replace `` 8 '' by `` 5 '' .
see the proof of lemma [ lem_d_gm_tail ] .
( 2 ) this lemma is weaker than the classical jl lemma for dimension reduction in @xmath9 as reviewed in section 2.1 .
the classical jl lemma for @xmath9 ensures that the @xmath9 inter - point distances of the projected data points are close enough to the original @xmath9 distances , while lemma [ lem_jl_l1 ] merely says that the projected data points contain enough information to reconstruct the original @xmath0 distances . on the other hand ,
the geometric mean estimator is a non - convex norm ; and therefore it does contain some information about the geometry .
we leave it for future work to explore the possibility of developing efficient algorithms using the geometric mean estimator .
+ figure [ fig_hist_d_gm ] presents the simulated histograms of @xmath85 for @xmath185 , with @xmath186 and @xmath187 .
the histograms reveal some characteristics shared by the maximum likelihood estimator we will discuss in the next section : * supported on @xmath188 , @xmath85 is positively skewed . *
the distribution of @xmath85 is still `` heavy - tailed . '' however , in the region not too far from the mean , the distribution of @xmath85 may be well captured by a gamma ( or a generalized gamma ) distribution . for large @xmath8 ,
even a normal approximation may suffice .
figure [ fig_gm_vs_me ] compares @xmath85 with the sample median estimators @xmath128 and @xmath84 , in terms of the mean square errors .
@xmath85 is considerably more accurate than @xmath128 at small @xmath8 .
the bias correction significantly reduces the mean square errors of @xmath128 .
this section is devoted to analyzing the maximum likelihood estimators ( mle ) , which are `` asymptotically optimum . '' in comparisons , the sample median estimators and geometric mean estimators are not optimum .
our contribution in this section includes the higher - order analysis for the bias and moments and accurate closed - from approximations to the distribution of the mle .
the method of maximum likelihood is widely used .
for example , @xcite applied the maximum likelihood method to _ normal random projections _ and provided an improved estimator of the @xmath9 distance by taking advantage of the marginal information .
the cauchy distribution is often considered a `` challenging '' example because of the `` multiple roots '' problem when estimating the location parameter @xcite . in our case , since the location parameter is always zero , much of the difficulty is avoided .
recall our goal is to estimate @xmath81 from @xmath8 i.i.d .
samples @xmath189 .
the @xmath190 joint likelihood of @xmath43 is @xmath191 whose first and second derivatives ( w.r.t .
@xmath81 ) are @xmath192 the maximum likelihood estimator of @xmath81 , denoted by @xmath111 , is the solution to @xmath193 , i.e. , @xmath194 because @xmath195 , @xmath111 indeed maximizes the joint likelihood and is the only solution to the mle equation ( [ eqn_mle ] ) . solving ( [ eqn_mle ] )
numerically is not difficult ( e.g. , a few iterations using the newton s method ) . for a better accuracy ,
we recommend the following bias - corrected estimator : @xmath196 lemma [ lem_mle_asymp ] concerns the asymptotic moments of @xmath111 and @xmath86 , proved in appendix [ app_proof_lem_asymp ] .
[ lem_mle_asymp ] both @xmath111 and @xmath86 are asymptotically unbiased and normal .
the first four moments of @xmath111 are @xmath197 the first four moments of @xmath86 are @xmath198 + the order @xmath199 term of the variance , i.e. , @xmath200 , is known , e.g. , @xcite .
we derive the bias - corrected estimator , @xmath86 , and the higher order moments using stochastic taylor expansions @xcite .
we will propose an inverse gaussian distribution to approximate the distribution of @xmath86 , by matching the first four moments ( at least in the leading terms ) .
the maximum likelihood estimators are tested on msn web crawl data , a term - by - document matrix with @xmath201 web pages .
we conduct cauchy random projections and estimate the @xmath0 distances between words . in this experiment
, we compare the empirical and ( asymptotic ) theoretical moments , using one pair of words .
figure [ fig_bias_var ] illustrates that the bias correction is effective and these ( asymptotic ) formulas for the first four moments of @xmath86 in lemma [ lem_mle_asymp ] are accurate , especially when @xmath202 .
theoretical analysis on the exact distribution of a maximum likelihood estimator is difficult .
, @xmath203 , ... , @xmath204 , the distribution of @xmath111 can be exactly characterized @xcite .
@xcite studied the conditional confidence interval of the mle .
later , @xcite proposed the normal approximation to the exact conditional confidence interval and showed that it was superior to the unconditional normality approximation .
unfortunately , we can not take advantage of the conditional analysis because our goal is to determine the sample size @xmath8 before seeing any samples . ] in statistics , the standard approach is to assume normality , which , however , is quite inaccurate . the so - called _ edgeworth expansion _
improves the normal approximation by matching higher moments @xcite .
for example , if we approximate the distribution of @xmath86 using an edgeworth expansion by matching the first four moments of @xmath86 derived in lemma [ lem_mle_asymp ] , then the errors will be on the order of @xmath205
. however , edgeworth expansions have some well - known drawbacks .
the resultant expressions are quite sophisticated .
they are not accurate at the tails .
it is possible that the approximate probability has values below zero .
also , edgeworth expansions consider the support is @xmath206 , while @xmath86 is non - negative .
we propose approximating the distributions of @xmath86 directly using some well - studied common distributions .
we will first consider a gamma distribution with the same first two ( asymptotic ) moments of @xmath86 .
that is , the gamma distribution will be asymptotically equivalent to the normal approximation .
while a normal has zero third central moment , a gamma has nonzero third central moment .
this , to an extent , speeds up the rate of convergence .
another important reason why a gamma is more accurate is because it has the same support as @xmath86 , i.e. , @xmath188 .
we will furthermore consider a _
generalized gamma _ distribution , which allows us to match the first three ( asymptotic ) moments of @xmath86 .
interestingly , in this case , the generalized gamma approximation turns out to be an inverse gaussian distribution , which has a closed - form probability density .
more interestingly , this inverse gaussian distribution also matches the fourth central moment of @xmath86 in the @xmath207 term and almost in the @xmath208 term . by simulations ,
the inverse gaussian approximation is highly accurate .
note that , since we are interested in the very small ( e.g. , @xmath209 ) tail probability range , @xmath205 is not too meaningful .
for example , @xmath210 if @xmath211
. therefore , we will have to rely on simulations to assess the accuracy of the approximations . on the other hand
, an upper bound may hold exactly ( verified by simulations ) even if it is based on an approximate distribution .
as the related work , @xcite applied gamma and generalized gamma approximations to model the performance measure distribution in some wireless communication channels using random matrix theory and produced accurate results in evaluating the error probabilities .
the gamma approximation is an obvious improvement over the normal approximation .
, the resultant estimator of the squared @xmath9 distance has a chi - squared distribution ( e.g. , ( * ? ? ?
* lemma 1.3 ) ) , which is a special case of gamma . ] a gamma distribution , @xmath212 , has two parameters , @xmath213 and @xmath214 , which can be determined by matching the first two ( asymptotic ) moments of @xmath86 .
that is , we assume that @xmath215 , with @xmath216 assuming a gamma distribution , it is easy to obtain the following chernoff bounds ; and we then choose @xmath217 that minimizes the upper bound . ] : @xmath218 where we use @xmath219 to indicate that these inequalities are based on an approximate distribution .
note that the distribution of @xmath220 ( and hence @xmath221 ) is only a function of @xmath8 as shown in @xcite .
therefore , we can evaluate the accuracy of the gamma approximation by simulations with @xmath222 , as presented in figure [ fig_gamma_tail ] .
figure [ fig_gamma_tail](a ) shows that both the gamma and normal approximations are fairly accurate when the tail probability @xmath223 ; and the gamma approximation is obviously better .
figure [ fig_gamma_tail](b ) compares the empirical tail probabilities with the gamma chernoff upper bound ( [ eqn_gamma_right])+([eqn_gamma_left ] ) , indicating that these bounds are reliable , when the tail probability @xmath224 .
the distribution of @xmath86 can be well approximated by an inverse gaussian distribution , which is a special case of the three - parameter generalized gamma distribution @xcite , denoted by @xmath225 .
note that the usual gamma distribution is a special case with @xmath226 .
if @xmath227 , then the first three moments are @xmath228 we can approximate the distribution of @xmath86 by matching the first three moments , i.e. , @xmath229 from which we obtain @xmath230 taking only the leading term for @xmath231 , the generalized gamma approximation of @xmath86 would be @xmath232 in general , a generalized gamma distribution does not have a closed - form density function although it always has a closed - from moment generating function . in our case , ( [ eqn_ig ] )
is actually an inverse gaussian distribution , which has a closed - form density function . assuming @xmath233 , with parameters @xmath213 and @xmath214 defined in ( [ eqn_ig_parameters ] ) , the moment generating function ( mgf )
, the probability density function ( pdf ) , and cumulative density function ( cdf ) would be ( * ? ? ?
* chapter 2 ) @xcite @xmath234 where @xmath235 is the standard normal cdf , i.e. , @xmath236 .
here we use @xmath237 to indicate that these equalities are based on an approximate distribution .
assuming @xmath238 , then the fourth central moment should be @xmath239 lemma [ lem_mle_asymp ] has shown the true asymptotic fourth central moment : @xmath240 that is , the inverse gaussian approximation matches not only the leading term , @xmath241 , but also almost the higher order term , @xmath242 , of the true asymptotic fourth moment of @xmath86 .
assuming @xmath243 , the tail probability of @xmath86 can be expressed as @xmath244 assuming @xmath243 , it is easy to show the following chernoff bounds : @xmath245 to see ( [ eqn_ig_left ] ) .
assume @xmath246 .
then , using the chernoff inequality : @xmath247 whose minimum is @xmath248 , attained at @xmath249
. we can similarly show ( [ eqn_ig_right ] ) .
+ combining ( [ eqn_ig_left ] ) and ( [ eqn_ig_right ] ) yields a symmetric bound @xmath250 figure [ fig_ig_tail ] compares the inverse gaussian approximation with the same simulations as presented in figure [ fig_gamma_tail ] , indicating that the inverse gaussian approximation is highly accurate .
when the tail probability @xmath251 , we can treat the inverse gaussian as the exact distribution of @xmath86 .
the chernoff upper bounds for the inverse gaussian are always reliable in our simulation range ( the tail probability @xmath120 ) .
it is well - known that the @xmath0 distance is far more robust than the @xmath9 distance against `` outliers . ''
there are numerous success stories of using the @xmath0 distance , e.g. , lasso @xcite , lars @xcite , 1-norm svm @xcite , and laplacian radial basis kernel @xcite .
dimension reduction in the @xmath0 norm , however , has been proved _ impossible _
if we use _ linear random projections _ and _ linear estimators_. in this study , we propose three types of nonlinear estimators for _ cauchy random projections _ : the bias - corrected sample median estimator , the bias - corrected geometric mean estimator , and the bias - corrected maximum likelihood estimator .
our theoretical analysis has shown that these nonlinear estimators can accurately recover the original @xmath0 distance , even though none of them can be a metric .
the bias - corrected sample median estimator and the bias - corrected geometric mean estimator are asymptotically equivalent but the latter is more accurate at small sample size .
we have derived explicit tail bounds for the bias - corrected geometric mean estimator and have expressed the tail bounds in exponential forms . using these tail bounds ,
we have established an analog of the johnson - lindenstrauss ( jl ) lemma for dimension reduction in @xmath0 , which is weaker than the classical jl lemma for dimension reduction in @xmath9 .
we conduct theoretic analysis on the bias - corrected maximum likelihood estimator ( mle ) , which is `` asymptotically optimum . ''
both the sample median estimator and the geometric mean estimator are about @xmath10 efficient as the mle .
we propose approximating its distribution by an inverse gaussian , which has the same support and matches the leading terms of the first four moments of the proposed estimator .
approximate tail bounds have been provide based on the inverse gaussian approximation
. verified by simulations , these approximate tail bounds hold at least in the @xmath252 tail probability range . although these nonlinear estimators are not metrics , they are still useful for certain applications in ( e.g. , ) data stream computation , information retrieval , learning and data mining , whenever the goal is to compute the @xmath0 distances efficiently using a small storage space
. the geometric mean estimator is a non - convex norm ( i.e. , the @xmath18 norm as @xmath100 ) ; and therefore it does contain some information about the geometry .
it may be still possible to develop certain efficient algorithms using the geometric mean estimator by avoiding the non - convexity .
we leave this for future work .
we are grateful to piotr indyk and assaf naor for the very constructive comments on various versions of this manuscript .
we thank dimitris achlioptas , christopher burges , moses charikar , jerome friedman , tze l. lai , art b. owen , john platt , joseph romano , tim roughgarden , yiyuan she , and guenther walther for helpful conversations or suggesting relevant references .
we also thank silvia ferrari and gauss cordeiro for clarifying some parts of their papers .
trevor hastie was partially supported by grant dms-0505676 from the national science foundation , and grant 2r01 ca 72028 - 07 from the national institutes of health .
59 natexlab#1#1url # 1#1 dimitris achlioptas .
database - friendly random projections : with binary coins . , 660 ( 4):0 671687 , 2003 . charu c. aggarwal and joel l. wolf . a new method for similarity indexing of market basket data . in _ proc .
of sigmod _ , pages 407418 , philadelphia , pa , 1999 .
nir ailon and bernard chazelle .
approximate nearest neighbors and the fast transform . in _ proc . of stoc _ ,
pages 557563 , seattle , wa , 2006 .
charles antle and lee bain . a property of maximum likelihood estimators of location and scale parameters . , 110 ( 2):0 251253 , 1969 .
rosa arriaga and santosh vempala .
an algorithmic theory of learning : robust concepts and random projection . in _ proc . of focs _ , pages 616623 , new york , 1999 .
rosa arriaga and santosh vempala .
an algorithmic theory of learning : robust concepts and random projection .
, 630 ( 2):0 161182 , 2006 .
v. d. barnett .
evaluation of the maximum - likelihood estimator where the likelihood equation has multiple roots .
, 530 ( 1/2):0 151165 , 1966
. m. s. bartlett .
approximate confidence intervals , . , 400 ( 3/4):0 306317 , 1953 .
r. n. bhattacharya and j. k. ghosh . on the validity of the formal expansion . , 60 ( 2):0 434451 , 1978 . bo brinkman and mose charikar .
on the impossibility of dimension reduction in @xmath0 . in _ proc . of focs _ , pages 514523 , cambridge , ma , 2003
bo brinkman and mose charikar .
on the impossibility of dimension reduction in @xmath0 . , 520 ( 2):0 766788 , 2005 . olivier chapelle , patrick haffner , and vladimir n. vapnik .
support vector machines for histogram - based image classification . , 100 ( 5):0 10551064 , 1999
. herman chernoff .
a measure of asymptotic efficiency for tests of a hypothesis based on the sum of observations .
, 230 ( 4):0 493507 , 1952 . raj s. chhikara and j. leroy folks . .
marcel dekker , inc , new york , 1989 .
francisco jose de .
a. cysneiros , sylvio jose p.
dos santos , and gass m. cordeiro .
skewness and kurtosis for maximum likelihood estimator in one - parameter exponential family models . , 150 ( 1):0 85105 , 2001 .
sanjoy dasgupta and anupam gupta .
an elementary proof of a theorem of .
, 220 ( 1):0 60 65 , 2003
. inderjit s. dhillon and dharmendra s. modha .
concept decompositions for large sparse text data using clustering . , 420 ( 1 - 2):0 143175 , 2001 . bradley efron , trevor hastie , iain johnstone , and robert tibshirani . least angle regression . , 320 ( 2):0 407499 , 2004 .
eugene f. fama and richard roll .
some properties of symmetric stable distributions . , 630 ( 323):0 817836 , 1968 . eugene f. fama and richard roll .
parameter estimates for symmetric stable distributions . , 660 ( 334):0 331338 , 1971 .
william feller . .
john wiley & sons , new york , ny , second edition , 1971 .
marin ferecatu , michel crucianu , and nozha boujemaa .
retrieval of difficult image classes using svd - based relevance feedback . in _ prof . of multimedia information retrieval _ , pages 2330 , new york , ny , 2004 .
silvia l. p. ferrari , denise a. botter , gauss m. cordeiro , and francisco cribari - neto .
second and third order bias reduction for one - parameter family models .
, 30:0 339345 , 1996 .
r. a. fisher .
two new properties of mathematical likelihood . , 1440 ( 852):0 285307 , 1934 .
p. frankl and h. maehara .
the lemma and the sphericity of some graphs .
, 440 ( 3):0 355362 , 1987 . hans u. gerber . from the generalized gamma to the generalized negative binomial distribution .
, 100 ( 4):0 303309 , 1991 .
i. s. gradshteyn and i. m. ryzhik . .
academic press , new york , fifth edition , 1994 .
gerald haas , lee bain , and charles antle .
inferences for the cauchy distribution based on maximum likelihood estimation . , 570 ( 2):0 403408 , 1970 .
david v. hinkley .
likelihood inference about location and scale parameters .
, 650 ( 2):0 253261 , 1978 .
p. hougaard .
survival models for heterogeneous populations derived from stable distributions . , 730 ( 2):0 387396 , 1986 .
piotr indyk .
stable distributions , pseudorandom generators , embeddings and data stream computation . in _ focs _ , pages 189197 , redondo beach , ca , 2000 .
piotr indyk .
algorithmic applications of low - distortion geometric embeddings . in _ proc . of focs _ , pages 1033 , las vegas , nv , 2001 .
piotr indyk and rajeev motwani .
approximate nearest neighbors : towards removing the curse of dimensionality . in _ proc . of stoc _ , pages 604613 , dallas , tx , 1998 .
piotr indyk and assaf naor .
nearest neighbor preserving embeddings . , 2006
. jens ledet jensen . .
oxford university press , new york , 1995 .
w. b. johnson and j. lindenstrauss .
extensions of mapping into space . , 26:0 189206 , 1984 .
j. f. lawless .
conditional confidence interval procedures for the location and scale parameters of the cauchy and logistic distributions . , 590 ( 2):0 377386 , 1972 .
james r. lee and assaf naor . embedding the diamond graph in @xmath18 and dimension reduction in @xmath0 . , 140 ( 4):0 745747 , 2004 . ping li and kenneth w. church . using sketches to estimate two - way and multi - way associations
. technical report tr-2005 - 115 , microsoft research , ( a shorter version is available at www.stanford.edu/@xmath253pingli98/publications/report_sketch.pdf ) , redmond , wa , september 2005 .
ping li , kenneth w. church , and trevor j. hastie .
conditional random sampling : a sketched - based sampling technique for sparse data .
technical report , department of statistics , stanford university ( www.stanford.edu/~pingli98/publications/crs_tr.pdf ) , 2006 .
ping li , trevor j. hastie , and kenneth w. church .
improving random projections using marginal information . in _ proc .
of colt _ , pittsburgh , pa , 2006 .
ping li , trevor j. hastie , and kenneth w. church .
sub - gaussian random projections .
technical report , department of statistics , stanford university ( www.stanford.edu/~pingli98/report/subg_rp.pdf ) , 2006 .
ping li , trevor j. hastie , and kenneth w. church .
very sparse random projections . in _ proc . of kdd _ , philadelphia , pa , 2006 .
ping li , debashis paul , ravi narasimhan , and john cioffi . on the distribution of for the receiver and performance analysis . , 520 ( 1):0 271286 , 2006 .
gabor lugosi .
concentration - of - measure inequalities . , 2004 .
j. huston mcculloch .
simple consistent estimators of stable distribution parameters . , 150 ( 4):0 11091136 , 1986 .
thomas k. philips and randolph nelson .
the moment bound is tighter than chernoff s bound for positive tail probabilities .
, 490 ( 2):0 175178 , 1995 .
v. seshadri . .
oxford university press inc .
, new york , 1993 .
thomas a. severini . .
oxford university press , new york , 2000 .
gregory shakhnarovich , trevor darrell , and piotr indyk , editors . . the mit press , cambridge , ma , 2005 .
jun shao . .
springer , new york , ny , second edition , 2003 . l. r. shenton and k. bowman .
higher moments of a maximum - likelihood estimate . , 250 ( 2):0 305317 , 1963 .
alexander strehl and joydeep ghosh .
a scalable approach to balanced , high - dimensional clustering of market - baskets . in _ proc . of hipc _ , pages 525536 , bangalore , india , 2000 .
robert tibshirani .
regression shrinkage and selection via the lasso .
, 580 ( 1):0 267288 , 1996 .
m. c. k. tweedie . statistical properties of inverse gaussian distributions . . , 280 ( 2):0 362377 , 1957 .
m. c. k. tweedie . statistical properties of inverse gaussian distributions . . , 280 ( 3):0 696705 , 1957 .
santosh vempala . .
american mathematical society , providence , ri , 2004 .
ji zhu , saharon rosset , trevor hastie , and robert tibshirani . 1-norm support vector machines . in
, 2003 .
v. m. zolotarev . .
american mathematical society , providence , ri , 1986 .
assume @xmath157 . the probability density function ( pdf ) and
the cumulative density function ( cdf ) of @xmath254 would be @xmath255 for @xmath303 , we can prove an exponential bound for @xmath304 .
first of all , note that we do not have to choose the optimum @xmath297 . by the taylor expansion , for small @xmath183
, @xmath305 can be well approximated by @xmath306 it is easy to see that as @xmath312 , @xmath313 .
figure [ fig_gm_constant](a ) illustrates that it suffices to let @xmath314 , which can be numerically verified .
this is why the last step in ( [ eqn_proof_right ] ) holds .
of course , we can get a better constant if ( e.g. , ) @xmath315 . in order to attain @xmath326
, we have to restrict @xmath8 to be larger than a certain value . for no particular reason ,
we like to express the restriction as @xmath327 , for some constant @xmath328 .
we find @xmath329 suffices , although readers can verify that a slightly better ( smaller ) restriction would be @xmath330 .
general formulas for the bias and higher moments of the mle are available in @xcite .
we need to evaluate the expressions in @xcite , involving tedious algebra : @xmath340}{2k\text{i}^2 } + o\left(\frac{1}{k^2}\right ) \\ & \text{var}\left(\hat{d}_{mle}\right ) = \frac{1}{k\text{i } } + \frac{1}{k^2}\left(-\frac{1}{\text{i}}+\frac{[1 ^ 4]-[1 ^ 22]-[13]}{\text{i}^3 } + \frac{3.5[12]^2-[1 ^ 3]^2}{\text{i}^4}\right ) + o\left(\frac{1}{k^3}\right ) \\ &
\text{e}\left(\hat{d}_{mle}-\text{e}\left(\hat{d}_{mle}\right)\right)^3 = \frac{[1 ^ 3]-3[12]}{k^2\text{i}^2}+o\left(\frac{1}{k^3}\right ) \\\notag & \text{e}\left(\hat{d}_{mle}-\text{e}\left(\hat{d}_{mle}\right)\right)^4= \frac{3}{k^2\text{i}^2 } + \frac{1}{k^3}\left(-\frac{9}{\text{i}^2}+ \frac{7[1 ^ 4 ] - 6[1 ^ 22]-10[13]}{\text{i}^4}\right)\\ & \hspace{1.7 in } + \frac{1}{k^3}\left(\frac{-6[1 ^ 3]^2 - 12[1 ^ 3][12]+45[12]^2}{\text{i}^5}\right)+o\left(\frac{1}{k^4}\right),\end{aligned}\ ] ] where , after re - formatting , @xmath341 = \text{e}(l^\prime)^3 + \text{e}(l^\prime l^{\prime\prime } ) , \hspace{0.3 in } [ 1 ^ 4 ] = \text{e}(l^\prime)^4 , \hspace{0.3 in } [ 1 ^ 22 ] = \text{e}(l^{\prime\prime}(l^\prime)^2 ) + \text{e}(l^{\prime})^4 , \\ & [ 13 ] = \text{e}(l^\prime)^4 + 3\text{e}(l^{\prime\prime}(l^\prime)^2 ) + \text{e}(l^\prime l^{\prime\prime\prime } ) , \hspace{0.3 in } [ 1 ^ 3]=\text{e}(l^\prime)^3.\end{aligned}\ ] ] without giving the detail , we report @xmath343 hence @xmath344 = -\frac{1}{2}\frac{1}{d^3 } , \hspace{0.25 in } [ 1 ^ 4 ] = \frac{3}{8}\frac{1}{d^4 } , \hspace{0.25in}[1 ^ 22 ] = \frac{1}{4}\frac{1}{d^4 } , \hspace{0.25in}[13 ] = \frac{3}{4}\frac{1}{d^4 } , \hspace{0.25in}[1 ^ 3 ] = 0.\end{aligned}\ ] ] because @xmath111 has @xmath199 bias , we recommend the bias - corrected estimator @xmath346 whose first four moments are @xmath347 by brute - force algebra . first , it is obvious that @xmath348 then @xmath349 | for _ , is available as a technical report in stanford statistics achive ( report no .
2006 - 04 , june , 2006 ) . ]
dimension reduction in @xmath0 , the method of _ cauchy random projections _ multiplies the original data matrix @xmath1 with a random matrix @xmath2 ( @xmath3 ) whose entries are i.i.d
. samples of the standard cauchy @xmath4 .
because of the impossibility results , one can not hope to recover the pairwise @xmath0 distances in @xmath5 from @xmath6 , using linear estimators without incurring large errors . however , nonlinear estimators are still useful for certain applications in data stream computation , information retrieval , learning , and data mining .
we propose three types of nonlinear estimators : the bias - corrected sample median estimator , the bias - corrected geometric mean estimator , and the bias - corrected maximum likelihood estimator .
the sample median estimator and the geometric mean estimator are asymptotically ( as @xmath7 ) equivalent but the latter is more accurate at small @xmath8 .
we derive explicit tail bounds for the geometric mean estimator and establish an analog of the johnson - lindenstrauss ( jl ) lemma for dimension reduction in @xmath0 , which is weaker than the classical jl lemma for dimension reduction in @xmath9 .
asymptotically , both the sample median estimator and the geometric mean estimators are about @xmath10 efficient compared to the maximum likelihood estimator ( mle ) .
we analyze the moments of the mle and propose approximating the distribution of the mle by an inverse gaussian .
* keywords : * dimension reduction , @xmath0 norm , cauchy random projections , jl bound |
solar flares are most energetic magnetic explosions in the solar activities .
they can increase the emission in a broad range of the electromagnetic spectrum , from radio wavelengths to x- and @xmath0-rays ( fletcher et al . 2011 ) . in the standard solar flare model ,
i.e. the cshkp model ( named after carmichael 1964 ; sturrock 1966 ; hirayama 1974 ; kopp & pneuman 1976 ) , the erupting flux rope stretched magnetic filed lines to induce the magnetic reconnection ; due to the successive reconnections , the flare loops ( fls ; originally named as post - flare loops ) formed and straddled the magnetic polarity inversion line , and their footpoints are heated by the energy transport from the reconnection site to appear as flare ribbons ( frs ) .
however , the standard flare model is basically two - dimensional , and it remains deficient to explain many inherent three - dimensional ( 3d ) observational features , such as the formation of coronal sigmoids ( aulanier et al .
2010 ; green et al . 2011 ; savcheva et al . 2015 ) , the erupting flux rope ( zhang et al . 2012 ) , the moving bright emissions along the frs ( fletcher & hudson 2002 ; del zanna et al .
2006 ; chandra et al .
2009 ) , and the strong - to - weak shear transition in fls ( aulanier et al .
recently , aulanier et al . ( 2012 ) and janvier et al . (
2013 ) proposed a fully 3d flare model that incorporates the standard 2d one in one of its cuts , thus satisfying the principle of correspondence . in the 3d magnetohydrodynamic simulation
, the magnetic field lines passed through the quasi - separatrix layers ( qsls ; priest & dmoulin 1995 ) , and could undergo a successive reconnection , which exchanged their connectivity with neighboring field lines ; the continuous reconnections resulted in the obvious slipping motion along the qsls ( pontin et al . 2005 ; de moortel & galsgaard 2006 ; aulanier et al .
therefore , it is classified as the slipping magnetic reconnection .
as the theoretical models have been developed , some observational cases of the slipping magnetic reconnection have been analysed .
aulanier et al .
( 2007 ) first presented the direct observations of slipping magnetic reconnection in coronal loops by the x - ray telescope onboard hinode .
based on the atmospheric imaging assembly ( aia ; lemen et al . 2012 ) on the solar dynamics observatory ( sdo ; pesnell et al .
2012 ) , dudk et al .
( 2014 ) reported the apparent slipping motion of fls during an eruptive x1.4 flare , and li & zhang ( 2014 , 2015 ) showed two examples of slipping motion of fls and the quasi - periodic pattern of the latter case .
however , the previous observations only focused on the slipping motion in two frs , and the slippage in multiple frs with complex magnetic configuration has never been discussed . in this letter , we present the slipping magnetic reconnections in multiple frs during an x1.2 eruptive flare .
the x1.2 flare occurred in ars 11944 and 11943 on 2014 january 7 , associated with a rapid coronal mass ejection and a fast coronal wave .
we only focus on the flare , and its start , peak , and end times are about 18:04 , 18:32 , and 18:58 ut , respectively .
to study the evolution of frs , we mainly use the observations from the aia on the sdo and the h@xmath1 filtergrams from the global oscillation network group ( gong ) of the national solar observatory .
the aia has 10 euv and uv wavelengths , covering a wide range of temperatures .
the aia image ( 4096 @xmath24096 pixels ) covers the full disk of the sun and up to 0.5 @xmath3 above the limb , with a pixel resolution of 0.6@xmath4 and a cadence of 12 s. the h@xmath1 images are at 6563 with a spatial resolution of 1@xmath4 and a cadence of around 1 minute ( harvey et al .
magnetograms from the helioseismic and magnetic imager ( hmi ; scherrer et al .
2012 ) , another instrument on the sdo , are chosen to check the magnetic field configuration of the eruption region , with a cadence of 45 s and a pixel scale of 0.6@xmath4 . in order to analyze the dynamics of moving structures
, we employ the time - slice approach .
the associated speeds are obtained by the linear fits , assuming that the measurement uncertainty of the selected points is 4 pixels ( @xmath5 mm ) .
the general appearance of the eruption region and the nearby environment before the flare is shown in the hmi magnetogram ( panel a ) and aia 171 image ( panel b ) in figure 1 .
the flare occurred between the ars 11944 and 11943 , but the negative polarity ( n1 ) of ar 11944 barely contributed .
the flare was associated with the leading positive polarity ( p1 ) of ar 11944 and the pair of polarities of ar 11943 ( n2-p2 ) , and it seems that n2 is surrounded by p1 and p2 ( panel a ) .
there was a sigmoidal loop ( sl ) connecting n2 and p1 ( panels b - c ) . to the north of sl
, there were trans - equatorial lines ( tels ) linking ars 11944 and 11946 .
tels were resulted from magnetic reconnection that occurred at the x point between ars 11944 and 11946 ( panel b ) . as tels formed more and more , their south footpoints approached much more close to sl ( boxes in panels d - e ) .
the time - slice plot along the s1 in panel e shows the clear inflow for tels , with a speed of 4.8 km s@xmath6 before the flare onset ( panel i ) , which is close to the upper limit of reconnection inflow speeds ( yokoyama et al .
the close relationship between the inflow and the flare onset demonstrates that the magnetic reconnection between tels and the overlying field of sl s western part invoked the flare .
there formed three normal frs ( panel f ) : the center fr ( fr1 ) , the east fr ( fr2 ) , and the west fr ( fr3 ) . fr1 located at n2 , and fr2 and fr3 lied at p1 and p2 , respectively .
fr2 and fr3 connected fr1 by fls ( black arrows in panel h ) .
frs all experienced a lateral extension , as a result of slipping reconnections .
there appeared hook structures at the eastern end of fr1 and fr2 ( black arrows in panel g ) , and a faint semi - circular secondary fr ( sfr ; named by zhang et al .
2014 ) on p2 ( the white arrow in panel g ) connecting fr3 .
note that the warm coronal loops ( l1 ; the white arrow in panel h ) , connecting sfr and fr1 , survived during the eruptive flare .
the evolution of the multiple frs is shown in 304 images ( first and third columns ) and h@xmath1 filtergrams ( second and fourth columns ) in figure 2 . before the flare ,
fr3 and sfr were at a single continuous bright lane ( white arrows in panel a1 ) .
there were also some filaments ( f1-f3 ; arrows in panel a2 ) around the eruption region . at the beginning of the flare ,
fr1 lay at the edge of n2 as the coast - line of a peninsula ( the black arrow in panel b1 ) ; sequentially , fr1 quickly moved from the north top toward the center of n2 ( the dotted line s2 and the southward black arrow in panel c1 ) .
meanwhile , fr2 extended eastward , and fr3 moved southward ( white arrows in panels b1-d1 ) . as a result of the extension ,
hook - like structures formed at the east ends of fr1-fr2 ( conjugated arrows in panel d1 ) , and there appeared the semi - circular sfr connecting the south end of fr3 . in addition , there appeared mass flow from fr2 to ar 11946 ( black arrows in panels e - f ) , which likely provides another evidence of the magnetic reconnection between tels and the overlying field of sl s western part .
moreover , the f1-f3 seemed to survive during the entire flare ( white arrows in panels e2-h2 ) .
the attractive element of the event is that the bright features moving along sfr ( better seen in the attached animation ) .
the bright features first moved clockwise in west - east direction towards sfr s easternmost point ( northward black arrows in panels c - d of figure 2 ) , and then slipped anticlockwise towards fr3 .
note that the slippage of sfr first occurred along the outer edge , and then moved along the inner edge .
following the slippage , sfr nearly disappeared and was replaced by wider coronal dimmings ( northward white arrows in panels e1-h1 of figure 2 ) , but fr3 were nearly intact ( black arrows in panels g2-h2 of figure 2 ) .
the slipping motion along sfr and the associated dimmings are consistent with the presence of eruption as predicted by the standard 3d model for eruptive flares ( aulanier et al .
2012 ; janvier et al .
next , we focus on the interesting anticlockwise slippage of sfr shown in aia 94 , 193 , 171 , 304 , and 1600 ( figure 3 ) . in the right column , the moving features are indicated by black arrows , and the wider dimmings are pointed out by white arrows in panels a2-d4 .
the hook - like structures of fr1 and fr2 are shown by black arrows in the left column . during the flare , l1 connecting sfr and n2 ( white arrows in panels
b1-c1 ) were intact .
to best analyse the kinetics of fr1 and sfr , we employ the time - slice approach along the dotted lines in figure 2 and figure 3 . according to the time - slice plots ( upper panels of figure 4 ) , the slippage of fr1 started at about 18:04 ut , close to the flare onset , and
its speed is about 384 km s@xmath6 .
the anticlockwise propagation velocity of the moving feature along sfr is about 110 - 130 km s@xmath6 in 304 and 94 , and its onset is about at 18:30 ut . in middle and bottom panels of figure 4 , the eruptive flare is also shown in original ( panels d and i ) and base - difference ( panels e - h ) images of aia 94 ( see the attached animation ) .
the moving features of sfr ( white arrows in panels f - h ) and associated dimmings ( the white arrow in panel i ) are clear . at the beginning of the flare ,
fls were still in a lower altitude , but there were some hot loops ( l2 , panel d ) overlying fls between fr1 and fr2 , which existed before the flare .
as fls grew up , l2 was impulsively heated and expanded . about some minutes later , there formed some new loops ( l3 ; the black arrow in panel f ) connecting fr1 and a remote brightening ( the white arrow in panel e ) . the brightening rooted in the negative polarity ( n3 , see figure 1a ) of f3 ,
consistent with the brightening at the turn of f3 ( westward white arrows in panels e1-f1 of figure 2 ) .
because l3 connected p1 with n3 , there is likely a magnetic reconnection between l2 and the overlying arcades of f3 .
l3 kept extending southward along n3 ( the black arrow in panels h - i ) , and there left a clear channel between l3 and fls . on the other hand , the hook - like structure of fr2 was also fading ( dotted boxes in panels e - g ) , which is possibly related to the formation of l3 .
the apparent slipping motion of the fls and the hook - like structures of frs are consistent with the 3d reconnection in qsls , as predicted by the 3d standard flare model ( aulanier et al .
2012 and janvier et al .
the 3d model is generically with two polarities and two j - shape frs .
this kind of slippage of twin frs have been reported in previous studies ( dudk et al . 2014 ; li & zhang 2014 , 2015 ) . here
, the slipping motions of the flare involved at least two ars and three distinct frs , which may be a consequence of the more complex flux distribution in the photosphere . the semi - circular sfr contributes to complex magnetic configuration , and confirms the presence of a qsl ( aulanier et al .
2007 ) .
the sfr was not connected by fls , and disappeared following the slippage .
does it mean that sfr is a different structure , e.g. footpoints of coronal loops impulsively heated by the flare ( zheng et al .
2015 ) ? note that the eruption was launched between fr1 and fr3 towards southwest ( better seen in 193 of the attached animation ) , and it was closely associated with a fast partial - halo cme in the southwest ] . following the slippage , sfr widened , and coronal dimmings formed within the widened part of sfr ( white arrows in right column of figure 3 ) , which is in accordance with the predictions of the 3d standard model for eruptive flares ( figure 11a of janvier et al .
hence , sfr is the extended hook of fr3 , and represents footpoints of erupting structures as predicted in the 3d standard model for eruptive flares .
the presence of dimmings within the hook is due to the stretching of the field lines opened by the eruption . according to 3d slipping reconnection in qsls predicted by the 3d flare model
, we suggest the scenario of the slippage of sfr in a sketch ( figure 5 ) , which is based on an hmi magnetogram showing the associated magnetic polarities ( p1-p2 , n1-n3 ) .
fr3 and sfr are at a single continuous bright lane ( orange lines ) before the flare ( figure 2a1 ) .
the erupting sl ( the dashed light - blue line ) induced the flare and fls ( yellow lines ) . due to the successive reconnections , fr3
extended clockwise and formed a semi - circular sfr .
the bright features first moved along the outer edge of sfr toward its easternmost point , and continually slipped anticlockwise along the inner edge of sfr , followed by coronal dimmings within the hook ( the shadow in panel b ) .
the slipping directions of sfr is indicated by orange arrows , and the tight narrow hook is similar to that reported by dudk et al .
the presence of dimmings reveals the open of the erupting structure ( the pink lines ) .
note that the erupting structure is just a schematic representation of the eruption , and can itself be fragmented , since all of the fr1-fr3 have hook structures .
in addition , the survived warm coronal loops ( l1 ; red lines ) connected the inner edge of sfr and fr2 . due to the eruption direction
, the dimmings only occurred to the southwest of the closed l1 .
the inner edge of sfr is then simply the topological boundary between closed and open field lines . on the other hand , the hot coronal loops l2 ( green lines ) are impulsively heated , expanded , and reconnected with the magnetic field lines ( black lines ) overlying f3 ( the sparkle ) , which resulted in the formation of l3 ( deep - blue lines ) . as the findings in dudk et al .
( 2014 ) , the tight narrow hook of sfr fade and disappear , due to the moving footpoints of erupting structures . for the fading and disappearance of the hook of fr2 ,
it is possibly because of the successive reconnections between l2 and the overlying loops of f3 .
it suggests that the eruption with complex magnetic configurations involves at least two magnetic reconnections .
all the frs underwent slipping motions in most of aia passbands , especially for fr1 and sfr .
fr1 extended fast with a speed of 384 km s@xmath6 , which is faster than that of previous results ; the anticlockwise propagation velocity of sfr is about 130 km s@xmath6 , which is similar to that of the previous results ( dudk et al .
2014 ; li & zhang 2014 , 2015 ) .
the fast slippage of fr1 reveals the strong - to - weak shear transition of fls , which is likely associated with the untwisting of sl during the eruption . according to the definitions in aulanier et al .
( 2006 ) , the slipping - running and slipping reconnection regimes respectively correspond to super- and sub - alfvnic field line fast slippage . due to
the average alfvn speed for the decaying ar is 500 km s@xmath6 ( aulanier et al .
2012 ) , all the slipping speeds here are in the range of sub - alfvnic speed .
we suggest that the extension and propagation in multiple frs are as a result of slipping magnetic reconnections . benefiting from the high - resolution observations of aia / sdo
, we capture the slipping magnetic reconnections in multiple frs during the x1.2 flare on 2014 january 7 . as the result of slipping motion , multiple frs extended with hook structures . due to the moving footpoints of erupting structures and another magnetic reconnection , hooks disappeared , and/or
was replaced by coronal dimmings .
our results are accordant with the slipping magnetic reconnection regime in 3d standard model for eruptive flares ( aulanier et al . 2006 ; aulanier et al . 2012 ; janvier et al .
2013 ; janvier et al .
more studies of the slipping reconnection will be helpful in understanding the 3d standard flare model ; further observations and theoretical work will be necessary . many thanks to wei liu , ting li , xiaoli yan , and hongqiang song for the constructive discussion .
sdo is a mission of nasa s living with a star program .
the authors thank the sdo team for providing the data .
this work is supported by grants nsbrsf 2012cb825601 , nnsfc 41274175 , and 41331068 , and yunnan province natural science foundation 2013fb085 . | with the observations of the solar dynamics observatory , we present the slipping magnetic reconnections with multiple flare ribbons ( frs ) during an x1.2 eruptive flare on 2014 january 7 .
a center negative polarity was surrounded by several positive ones , and there appeared three frs .
the three frs showed apparent slipping motions , and hook structures formed at their ends . due to the moving footpoints of the erupting structures , one tight semi - circular hook disappeared after the slippage along its inner and outer edge , and coronal dimmings formed within the hook .
the east hook also faded as a result of the magnetic reconnection between the arcades of a remote filament and a hot loop that was impulsively heated by the under flare loops .
our results are accordant with the slipping magnetic reconnection regime in 3d standard model for eruptive flares .
we suggest that complex structures of the flare is likely a consequence of the more complex flux distribution in the photosphere , and the eruption involves at least two magnetic reconnections . |
the low enrichment of @xmath3mo alloy dispersed in an @xmath4 matrix is a prototype for new experimental nuclear fuels @xcite .
when these metals are brought into contact , diffusion in the @xmath5 interface gives rise to interaction phases . also , when subjected to temperature and neutron radiation , phase transformation from @xmath6 to @xmath7 occurs and intermetallic phases develop in the u@xmath8mo@xmath9al interaction zone .
fission gas pores nucleate in these new phases during service producing swelling and deteriorating the alloy properties @xcite .
an important technological goal is to delay or directly avoid undesirable phase formation by inhibiting interdiffusion of @xmath4 and @xmath10 components .
some of these compounds are believed to be responsible for degradation of properties @xcite .
housseau _ et al . _
@xcite , based on the effective diffusion coefficients values calculated from their experimental permeation tests , have demonstrated that these undesirable phases have not influence on the mobility of @xmath10 in @xmath4 . on the other hand ,
bierlin and green @xcite have reported the activation energy values of @xmath10 mobility in @xmath4 , based on the maximum rate of penetration of @xmath10 into @xmath4 .
on the other hand , brossa _ et al . _
@xcite , have produced couples and triplets structures using deposition methods to study the efficient diffusion barriers that should have simultaneously , a good bonding effect and a good thermal conductivity .
the practical interest of a @xmath11 barrier is shown by several publications concerning with the diffusion in the systems @xmath12 , @xmath13 and @xmath14 .
the study of the @xmath0 binary system was , limited to solid samples of the sandwich - type , clamped together by a titanium screw and diffusion treatments have been carried out .
results from this work @xcite , have inspired present calculations .
therefore it is important to study carefully and with special attention the initial microscopic processes that originate these intermetallic phases . in order to deal with this problem we started studying numerically the static and dynamic properties of vacancies and interstitials defects in the @xmath4(@xmath10 ) bulk and in the neighborhood of a @xmath15 interface using molecular dynamics calculations @xcite . here
, we review our previous works @xcite , performing calculation of the tracer diffusion coefficients in binary @xmath0 and @xmath1 alloys , using analytical expressions of the diffusion parameters in terms of microscopical magnitudes . we have summarized the theoretical tools needed to express the diffusion coefficients in terms of microscopic magnitudes such as , the jump frequencies , the free vacancy formation energy and the vacancy - solute binding energy .
then we start with non - equilibrium thermodynamics in order to relate the diffusion coefficients with the phenomenological onsager @xmath16-coefficients .
the microscopic kinetic theory , allows us to write the onsager coefficients in term of the jump frequency rates @xcite , which are evaluated from the migration barriers and the phonon frequencies under the harmonic approximation .
the lattice vibrations are treated within the conventional framework of vineyard @xcite that corresponds to the classical limit .
the jump frequencies are identified by the model developed further by le claire in ref .
@xcite , known as the five - frequency model for f.c.c lattices .
the method includes the jump frequency associated with the migration of the host atom in the presence of an impurity at a first nearest neighbor position .
all this concepts need to be put together in order to correctly describe the diffusion mechanism .
hence , in the context of the shell approximation , we follow the technique developed by allnatt in refs .
@xcite to obtain the corresponding transport coefficients , which are related to the diffusion coefficients through the flux equations . a similar procedure for f.c.c .
structures was performed by mantina et al .
@xcite for @xmath17 , @xmath18 and @xmath19 diluted in @xmath4 but using density functional theory ( dft ) . also , using dft calculations for b.c.c .
structures , choudhury _ et al .
_ @xcite have calculated the tracer self - diffusion and solute diffusion coefficients in diluted @xmath20 and @xmath21 alloys including an extensive analysis of the onsager @xmath16-coefficients . in the present work ,
we do not employ dft , instead we use a classical molecular statics technique coupled to the monomer method @xcite .
this much less computationally expensive method allows us to compute at low cost a bunch of jump frequencies from which we can perform averages in order to obtain more accurate effective frequencies .
although we use classical methods , we have also reproduced the migration barriers for @xmath0 with dft calculations coupled to the monomer method @xcite .
we proceed as follows , first of all we validate the five - frequency model using the @xmath0 system as a reference case for which there is a large amount of experimental data and numerical calculations @xcite . since , the @xmath1 and @xmath0 systems share the same crystallographic f.c.c .
structure , the presented description is analogous for both alloys .
the full set of frequencies are evaluated employing the economic monomer method @xcite .
the monomer is used to compute the saddle points configurations from which we obtain the jumps frequencies defined in the five - frequency model . for the @xmath0 system case ,
our results of the tracer solute and self - diffusion coefficients are in good agreement with the experimental data . in this case
we found that @xmath4 in @xmath11 , at diluted concentrations , migrates as a free specie in the full range of temperatures here considered . in the case of @xmath1 ,
present calculations show that both , the tracer and self - diffusion coefficients agree very well with the available experimental data in ref .
@xcite , although a vacancy drag mechanism could occur at temperatures below 500k , while , for at high temperatures the solute @xmath10 migrates by direct interchange with the vacancy .
the paper is organized as follows : in section [ s1 ] we briefly introduce a summary of the macroscopic equations of atomic transport that are provided by non - equilibrium thermodynamics @xcite . in this way analytical expressions of the intrinsic diffusion coefficients in binary alloys in terms of onsager coefficients are presented .
section [ s2 ] , is devoted to give the way to evaluate the onsager phenomenological coefficients following the procedure of allnat @xcite in terms of the jumps frequencies in the context of the five - frequency model . in section [ s3 ]
we show the methodology used to evaluate the tracer diffusion coefficients for the solvent and solute atoms , as well as , the so called solvent enhancement factor .
finally , in section [ s5 ] we present our numerical results using the theoretical procedure here summarized , which show a perfect accuracy with available experimental data , also we give an expression for the vacancy wind parameter which gives essential information about the flux of solute atoms induced by vacancy flow . the last section briefly presents some conclusions .
isothermal atomic diffusion in binary @xmath22 alloys can be described through a linear expression between the fluxes @xmath23 and the driving forces related by the onsager coefficients @xmath24 as , @xmath25 where @xmath26 is the number of components in the system , @xmath23 describes the flux vector density of component @xmath27 , while @xmath28 is the driving force acting on component @xmath27 .
the second range tensor @xmath24 is symmetric ( @xmath29 ) and depends on pressure and temperature , but is independent of the driving forces @xmath28 . from ( [ eq.2 ] ) the @xmath30 fick s law , which describes the atomic jump process on a macroscopic scale ,
can be recovered . on the other hand , for each @xmath27 component , the driving forces may be expressed , in absence of external force , in terms of the chemical potential @xmath31 , so that @xcite , @xmath32 in ( [ eq.3 ] ) @xmath33 is the absolute temperature , and the chemical potential @xmath31 is the partial derivative of the gibbs free energy with respect to the number of atoms of specie @xmath27 that is , @xmath34 where @xmath35 , is the activity coefficients , which is defined in terms of the activity @xmath36 and @xmath37 , is the molar concentration of specie @xmath27 . for the particular case of a binary diluted alloy @xmath38 with @xmath26 available lattice sites per unit volume , containing molar concentrations @xmath39 for host atoms , @xmath40 of solute atoms ( impurities ) and @xmath41 vacancies , the fluxes in terms of the onsager coefficients are expressed as , @xmath42 @xmath43 and @xmath44 from ( [ jal ] ) and ( [ jbl ] ) , we define @xmath45 @xmath46 in the case of @xmath47 , the diffusion coefficient for the vacancy is given by , @xmath48 in ( [ dateq ] ) and ( [ dbteq ] ) , @xmath49 and @xmath50 are the intrinsic diffusion coefficients for solvent @xmath51 and solute @xmath52 respectively , while @xmath53 is the vacancy diffusion coefficient @xcite . in ( [ dateq ] ) and ( [ dbteq ] ) the quantities @xmath54 are the thermodynamic factors , @xmath55 murch and qin @xcite have shown that the standard intrinsic diffusion coefficients in ( [ dateq ] ) and ( [ dbteq ] ) can be expressed in terms of the tracer diffusion coefficients @xmath56 , @xmath57 which are measurable quantities , and the collective correlation factor @xmath58 ( @xmath59 ) as : @xmath60\phi_a = d^{\star}_{a}\left [ \frac{f_{aa}}{f_a } - \left(\frac{c_a}{c_{s}}\right)\frac{f^{(a)}_{as}}{f_a}\right]\phi_a , \label{dateq1}\ ] ] @xmath61\phi_b = d^{\star}_{s}\left [ \frac{f_{ss}}{f_s } - \left(\frac{c_s}{c_{a}}\right)\frac{f^{(s)}_{as}}{f_s}\right]\phi_b .
\label{dbteq2}\ ] ] the intrinsic diffusion coefficients in ( [ dateq1 ] ) and ( [ dbteq2 ] ) are known as the modified darken equations , where @xmath62 ( @xmath63 ) are the diffusion coefficients of atoms of specie @xmath64 in a complete random walk performing @xmath65 jumps of length @xmath66 per unit time .
the collective correlation factors @xmath58 are related to the @xmath24 coefficients through , @xmath67 and for the mixed terms , @xmath68 the tracer correlation factors @xmath69 , @xmath70 are defined as the ratios @xmath71 and @xmath72 respectively .
the term in square brackets in the second term of equations ( [ dateq1 ] ) and ( [ dbteq2 ] ) , is the vacancy wind factor @xmath73 @xcite . in the next sections , we present the onsager coefficients in terms of the atomic jump frequencies taken from ref .
in order to understand the effect of different vacancy exchange mechanisms on solute diffusion , we adopt an effective five frequency model la le claire @xcite for f.c.c .
lattices , assuming that the perturbation of the solute movement by a vacancy @xmath74 , is limited to its immediate vicinity .
figure [ fig1 ] defines the jump rates @xmath75 ( @xmath76 ) considering only jumps between first neighbors . for them
, @xmath77 implies in the exchange between the vacancy and the solute , @xmath78 when the exchange between the vacancy and the solvent atom lets the vacancy as a first neighbor to the solute ( positions denoted with circled 1 in figure [ fig1 ] ) .
the frequency of jumps such that the vacancy goes to sites that are second neighbor of the solute is denoted by @xmath79 ( sites with circled 2 ) .
the model includes the jump rate @xmath80 for the inverse of @xmath79 .
jumps toward sites that are third and forth neighbor of the solute are all denoted with @xmath81 and @xmath82 respectively while @xmath83 and @xmath84 are used for their respective inverse frequency jumps .
the jump rate @xmath85 is used for vacancy jumps among sites more distant than forth neighbors of the solute atom . in this context , that enables association ( @xmath80 ) and dissociation reactions ( @xmath79 ) , i.e the formation and break - up of pairs , the model include free solute and vacancies to the population of bounded pairs .
it is assumed that a vacancy which jumps from the second to the third shell , with @xmath86 , will never return ( or returns from a random direction ) . as in ref .
@xcite we express @xmath87 and @xmath88 the six symmetry types of vacancy sites that are in the first coordination shell ( first neighbor with the solute ) or in the second coordination shell ( sites accessible from the first shell by one single vacancy jump ) are shown in figure [ fig2 ] .
sites that are equally distant from the solute atom @xmath52 at the origin , and that have the same abscissa ( x - coordinate in fig.[fig2 ] ) share the same vacancy occupation probability @xmath89 , equivalently for @xmath90 .
table [ t2 ] resumes the sites probability with @xmath91 where for @xmath92 there is only one index @xmath64 that is given in crescent order with the distance to the solute atom @xmath52 in a positive abscissa , while @xmath93 denote sites with negative @xmath94 coordinate . for the sites in the @xmath95 plane ( @xmath96 ) ,
the sites are denoted with two indexes as @xmath97 , where the second index @xmath98 is given in crescent order of the distance to the solute atom @xmath52 .
table [ t2 ] denotes the number of different types of sites and the distance of them to the @xmath94 axis .
+ .probability of occurrence of the vacancy at a site of the subset @xmath99 . [ cols="^,^,^,^,^,^,^,^,^",options="header " , ] in figure [ fig26 ] , we establish a comparison of our calculations for @xmath100 with the experimental data in table [ dexp ] , for a molar uranium concentrations @xmath101 .
we see that , experimental values ( filled stars ) in the temperature range of @xmath102^{\circ}c$ ] are in perfect agreement with @xmath100 obtained with the here described procedure . in the temperature range
where there are available experimental data , the @xmath10 mobility is mainly due to direct interchange between the @xmath10 atom and the vacancy . on the other hand ,
the diffusion of @xmath10 in @xmath4 was also calculated in a study of the maximum rate of penetration of @xmath10 into @xmath4 , in the temperature range @xmath103^{\circ}c$ ] @xcite .
the maximum penetration coefficient values in ref .
@xcite were , @xmath104 , @xmath105 and @xmath106 @xmath107 for @xmath108 , @xmath109 and @xmath110 , respectively . from the expression @xmath111 , the activation energy @xmath112 was @xmath113 in cal per mole in the temperature range of @xmath103^{\circ}c$ ] , where @xmath114 is expressed in calories per @xmath115 per mole , and
@xmath116 is a proportionality constant .
the plot @xmath117 vs @xmath118 provides a convenient basis for expressing and comparing penetration coefficients . as a final comment , a recent work by leenaers _
@xcite , presents a great quantity of experimental findings for a real system , where the present model can also be applied .
also performed but not shown here , for the @xmath0 , we have reproduced all the microscopical parameters with @xmath119 atoms using the classical molecular static technique and the siesta code coupled to the monomer method @xcite . in the literature
several researchers have studied the solvent atom - vacancy exchange in terms of the jump frequencies @xmath120 and @xmath121 , in the framework of the random alloy model , as for example in ref .
the authors have performed an extensive monte carlo study of the tracer correlation factors in simple cubic , b.c.c .
. binary random alloys . on the other hand , the kinetic formalism of moleko _ et al . _
@xcite , also describes the behavior of the tracer correlation factors for slow and faster diffusers .
in summary , in this work we present the general mechanism based on non - equilibrium thermodynamics and the kinetic theory , to describe the diffusion behavior in f.c.c diluted alloys .
non equilibrium thermodynamic , through the flux equations , relates the diffusion coefficients with the onsager tensor , while the kinetic theory relates the onsager coefficients in terms of microscopical magnitudes . in this way we are able to write expressions for the diffusion
coefficients only in terms of microscopic magnitudes , i.e. the jump frequencies .
the five frequency model has also been of great utility in order to discriminate the relevant jump frequencies , evaluated from the migration barriers under the harmonic approximation in the context of the conventional treatment by vineyard corresponding to the classical limit .
hence , we have calculated the full set of phenomenological coefficients from which the full set of diffusion coefficients are obtained through the flux equation . in this respect ,
the jump frequencies have been calculated from the migration barriers which are obtained with an economic static molecular techniques ( cmst ) namely the monomer method , that searches saddle configurations efficiently .
although in this work we have performed the treatment for the case of f.c.c .
latices where the diffusion is mediated by vacancy mechanism , a similar procedure can be adopted for other crystalline structures or different diffusion mechanism ( for example , interstitials ) .
we have exemplified our calculations for the particular cases of diluted @xmath0 and @xmath1 f.c.c . binary alloys .
we have found that the tracer diffusion coefficient are in very good agreement with the available experimental data , for both alloys .
present calculations show that qualitatively a vacancy drag mechanism is unlikely to occur for the @xmath0 system . in the case of @xmath1 ,
a vacancy drag mechanism could occur at temperatures below @xmath2k , while above this temperature the solute migrates by a direct interchange mechanism with the vacancy , such as was corroborated in the comparison with the available experimental data .
we have demonstrated that , the cmst is appropriate in order to describe the impurity diffusion behavior mediated by a vacancy mechanism in f.c.c .
this opens the door for future works in the same direction where a similar procedure will be used that includes interstitial defects .
i am particularly grateful to dr .
roberto c. pasianot for help on calculations of the attempt jump frequencies , to dr .
rivas for comments on the manuscript , and to martn urtubey for figure [ fig2 ] .
this work was partially financed by conicet pip-00965/2010 .
references http://www.rertr.anl.gov/ a.m. savchenko , a.v .
vatulin , i.v .
dobrikova , g.v .
kulakov , s.a .
ershov , y.v .
konovalov , proc . of the international meeting of the rertr ,
* s12 * -5 ( 2005 ) .
mirandu , s.f.aric , s.n.balart , l.m .
gribaudo , mat .
charact . ,
* 60 * , 888 ( 2009 ) .
n. housseau , a. van craeynest , d. calais , journal of nuclear materials , * 39 * -2 , 189 - 193 ( 1971 ) .
bierlein and d.r .
green , _ the diffusion of uranium into aluminium _
, physical metallurgy unit metallurgy research ( 1955 ) , url : www.osti.gov/scitech/servlets/purl/4368963 .
f. brossa , h.w .
sheleicher and r. theisen , proceeding of the symposium : new nuclear materials including non - metalic fuells , prague - juli 1 - 5 , 1963 vol ii .
edited by i.a.e.a .
pascuet , v.p .
ramunni and j.r .
fernndez , physica b : condensed matter , * 407 * , 16 , 3295 - 3297 ( 2011 ) .
ramunni , m.i .
pascuet y j.r .
fernndez , proc . of mmm 2010 , microstructure modeling , 719 - 722 ( 2010 ) .
allnat , j. phys .
c : solid state phys .
* 14 * , 5453 - 5466 ( 1981 ) .
allnat , j. phys .
c : solid state phys .
* 14 * , 5467 - 5477 ( 1981 ) .
vineyard , j. phys .
solids * 3 * , 121 ( 1957 ) .
le claire , journal of nuc . mat .
* 69 - 70 * , 70 - 96 ( 1978 ) .
m. mantina , y. wang , l.q .
chen , z.k .
liu and c. wolverton , acta materialia * 57 * 4102 - 4108 ( 2009 ) .
m. mantina , y. wang , r. arroyave , l. q. chen , z. k. liu , and c. wolverton , phys .
* 100 * , 215901 ( 2008 ) .
s. choudhury , l. barnard , j.d .
tucker , t.r .
allen , b.d .
wirth , m. asta and d. morgan , journal of nuclear materials , * 411 * , 1 - 3 , 1 - 14 ( 2011 ) .
ramunni , m.a .
alurralde and r.c .
pasianot , phys .
b * 74 * , 054113 ( 2006 ) .
pasianot , r.a .
prez , v.p .
ramunni and m. weissmann , journal of nuclear materials , * 392 * , 1 , 100 - 104 ( 2009 ) .
. shang , d.e .
kim , c.l .
zacherl and y. wang , journal of applied physics * 112 * , 053515 ( 2012 ) .
a.r . allnat and a.b .
lidiard , _ atomic transport in solids _ , cambridge university press .
howard and a.b .
lidiard , rep .
* 27 * , 161 ( 1964 ) .
murch and z. qin , defect and diffusion forum * 109 * , 1 - 18 ( 1994 ) . th .
heumann , j. phys .
f : metal phys . * 9 * , 10 , 1997 - 2010 ( 1979 ) . j.l .
bocquet , acta metall .
* 22 * ( 1974 ) .
j.r . manning , phys .
rev . * 136 * , a175846 ( 1964 ) .
m. koiwa and s. ishioka , phil . mag . * a47 * , 927 ( 1983 ) .
tucker , r. najafabadi , t.r .
allen and d. morgan , j. of nuc . mat . * 405 * , 216 - 234 ( 2010 ) .
s. divinski and chr .
herzig , intermetallics * 8 * , 1357 - 1368 ( 2000 ) .
i.v . belova and g.e .
murch , philosophical magazine a * 83 * , 3 , 393 - 399 ( 2003 ) .
moleko , a.r .
allnatt and e.l .
allnatt , phil . mag .
a * 59 * , 141 ( 1989 ) .
a. van der ven and g. ceder , phys .
_ 94 _ , 045901 ( 2005 ) .
a. van der ven , j. c. thomas , q. c. xu , b. swoboda and d. morgan , phys .
b * 78 * , 104306 ( 2008 ) .
a. van der ven , hui - chia yu , g. ceder and katsuyo thornton , progress in materials science * 55 * , 61 - 105 ( 2010 ) .
y. mishin , d. farkas , m.j .
mehl and d.a .
papaconstantopoulos , phys .
b , * 59 * 3393(1999 ) .
r. zope and y. mishin , phys .
b * 68 * , 24102 ( 2003 )
pascuet , g. bonni and j. r. fernndez , journal of nuclear materials * 424 * , 158 - 163 ( 2012 ) .
alonso , j.r .
fernndez , p.h .
gargano and g.h .
rubiolo , physica b : condensed matter * 404 * , 18 , 2851 - 2853 ( 2009 ) .
a. f. voter and s. p. chen , _ high temperature ordered intermetallic alloys of materials research society symposium proceedings _ , materials research society , pittsburgh , pennsylvania , 175 ( 1987 ) . c. j. smith , metal reference book , butterworth , london , @xmath122 edition ( 1976 ) .
w. wycisk and m. feller - kniepmeier , j. nucl .
* 69 - 70 * , 616 ( 1978 )
. l. e. murr , _ interfacial phenomena in metals and alloys _ , addison - wesley , reading , ma ( 1975 ) . c. kittel , introduction to solid state physics , wiley - interscience , new york ( 1986 ) .
g. henkelman and h. jnsson , journal of chemical physics * 111 * , 15 : _ surfaces , interfaces , and materials _ ( 2002 ) .
kazuaki toyoura , yukinori koyama , akihide kuwabara , fumiyasu oba and isao tanaka , phys .
b * 78 * , 214303 ( 2008 ) .
n. sandberg , b. magyari - kope and t. r. mattsson , phys .
rev . lett .
* 89 * , 065901 ( 2002 ) .
c. wert and c. zener , physical review , * 76*(8 ) , 1169 - 1175 , ( 1949 ) .
chelsey l. zacherl , www.ccmd.psu.edu/publications/theses/media/zacherl , dissertationfinal.pdf .
bocquet , g. brebec , y. limoge , diffusion in metals and alloys , elsevier science bv , amsterdam , the netherlands , 1996 .
campbell and a.l .
rukhin , acta materialia , * 59 * , 5194 - 5201(2011 ) . w. gust , m.b .
hintz , a. loddwg , h. odelius and b. predel , phys .
( a ) * 64*(1 ) , 187 - 194 ( 1981 ) .
swalin and a. martin , journal of metals * 8 * , 567 , ( 1956 ) .
tsuyoshi yamamoto , toshiyuki takashima and keizo nishida , transactions of the japan institute of metals , * 21 * , 9 , 601 - 608 ( 1980 ) .
a. leenaers , c. detavernier and s. van den berghe , j. of nuc . mat . * 381 * , 242 - 248 ( 2008 ) .
i.v . belova and g.e .
murch , philosophical magazine a * 80 * , 7 , 1469 - 1479 ( 2000 ) . | impurity diffusion coefficients are entirely obtained from a low cost classical molecular statics technique ( cmst ) . in particular , we show how cmst is appropriate in order to describe the impurity diffusion behavior mediated by a vacancy mechanism . in the context of the five - frequency model , cmst allows to calculate all the microscopic parameters , namely : the free energy of vacancy formation , the vacancy - solute binding energy and the involved jump frequencies , from them , we obtain the macroscopic transport magnitudes such as : correlation factor , solvent - enhancement factor , onsager and diffusion coefficients .
specifically , we perform our calculations in f.c.c . diluted @xmath0 and @xmath1 alloys .
results for the tracer diffusion coefficients of solvent and solute species are in agreement with available experimental data for both systems .
we conclude that in @xmath0 and @xmath1 systems solute atoms migrate by direct interchange with vacancies in all the temperature range where there are available experimental data . in the @xmath1 case
, a vacancy drag mechanism could occur at temperatures below @xmath2k .
diffusion , moddeling , numerical calculations , vacancy mechanism , diluted alloys , @xmath0 and @xmath1 systems . |
although transport in low - dimensional quantum systems has intensively been investigated theoretically in the past years , there still is an ongoing interest in understanding the transport phenomena in such systems , including their temperature and length scale dependence @xcite .
those works have often addressed a qualitative classification of the occurring transport types into ballistic or normal diffusive behavior and , in particular cases , the crucial mechanisms which are responsible for the emergence of diffusion have been studied . in this context
the role of non - integrability and quantum chaos is frequently discussed as an at least necessary condition @xcite .
significant theoretical attention has been devoted to spin-@xmath0 chains @xcite , e.g. to the prominent anisotropic heisenberg chain ( xxz model ) @xcite .
most controversial appears the question whether or not the ( finite temperature ) transport in the pure heisenberg chain is ballistic @xcite .
even though there certainly is a large variety of different methods for the investigation of transport in quantum systems , we concentrate here on two of the main approaches in detail which are also most relevant in the context of the present work .
the first approach may be classified as a direct one , since it is rather close to an experimental measurement setup : within the theory of open quantum systems @xcite the model of interest is coupled locally , e.g. at both ends of a spin chain , to reservoirs of the transported quantity , e.g. at different temperature or chemical potential @xcite . due to the coupling to reservoirs , a stationary non - equilibrium state eventually results for which all relevant expectation values such as the current and the spatial density profile of the transport quantity can be evaluated . here
, a vanishing profile corresponds to ballistic behavior , whereas normal diffusive dynamics is associated with a strictly linear profile , according to fourier s law , respectively fick s law , see , e.g. ref . . in the latter case a finite conductivity is simply given by the ratio of the current and the spatial density gradient of the transported quantity .
this conductivity can be also understood in terms of a diffusion coefficient , since transport is driven by a gradient within the considered model and not by an external force @xcite .
one may therefore directly compare with the diffusion constant of a corresponding closed scenario , where transport is not induced by the coupling to baths but by an initially inhomogeneous non - equilibrium density @xcite . in order to simulate the influence of the baths the liouville - von neumann equation for the coherent evolution of the density matrix
is routinely extended by an incoherent damping term , see , e.g. ref . .
the derivation of such a dynamical equation from a microscopic bath model is highly non - trivial and involves a combination of various subtle approximation schemes , e.g. improper approximations may eventually lead to a mathematically correct but physically irrelevant quantum master equation ( * qme*)@xcite .
one often intends to derive a proper * qme * of the lindblad form @xcite , because its special structure allows to apply the numerically efficient method of stochastic unraveling @xcite , e.g. spin-@xmath0 chains with about @xmath1 sites become numerically tractable @xcite .
however , a recently suggested matrix product operator ansatz has been shown to significantly increase this number of available sites up to several dozens @xcite , i.e. finite size effects are drastically reduced and the extracted conductivity may be interpreted as a pure bulk property of the model , if it does not depend crucially on the concrete form and strength of the bath coupling , of course @xcite .
another approach for the investigation of transport in quantum systems is the green - kubo formula ( * kf * ) which was originally derived for electrical conductance by the use of linear response theory @xcite . in that case
the ( frequency - dependent ) electrical field is an external force which perturbs the system and the resulting current of charge through the system is the response to this external perturbation .
the * kf * as such gives the ( frequency - dependent ) conductivity as the linear response coefficient in terms of a current auto - correlation function .
the same approach is also used in the context of gradient - driven transport phenomena , e.g. for transport of energy or heat . in that case the current is driven by a much more complicated mechanism @xcite , especially since it can not be treated as a perturbation to the system .
nevertheless , simply by the replacement of the electrical current , e.g. by the energy current @xcite , the * kf * is used for gradient - driven transport , too . but
a rigorous justification of this replacement remains still a conceptual problem , see ref . or the overview paper ( and the comprehensive literature which is cited therein ) .
however , the * kf * is nowadays a standard method for the investigation of transport in spin chains @xcite , not least due to its direct computability , once a finite piece of the considered system has been exactly diagonalized .
as far as numerics is concerned , exact diagonalization is restricted to spin-1/2 chains with at most 24 sites . but another difficulty arises , if the * kf * is evaluated on the basis of a finite piece : one often distinguishes between ballistic and normal diffusive transport by the notion of the drude weight , essentially the conductivity s singular contribution at zero - frequency , see the review , for example . whenever it is finite , the long - time behavior is expected to be ballistic , and whenever it vanishes , the dc - conductivity as the zero - frequency limit of the conductivity s regular non - singular part determines the long - time behavior .
since for any finite system the conductivity exclusively consists of delta peaks at different frequencies , there apparently are only singular contributions and it therefore is a difficult question how to extract and extrapolate the dc - conductivity from a finite system @xcite .
moreover , the quantitative comparison of the resulting dc - conductivity ( due to a possibly hypothetic external force ) with a diffusion constant ( from non - equilibrium bath scenarios ) is hardly possible without the detailed knowledge about an einstein relation between both quantities . in the present paper
we do not intend to further discuss the above mentioned conceptual and methodological problems which may come along with the * kf*. instead we will present a different but in a sense related approach which particularly is not concerned with the most of those problems . to this end
we will firstly introduce a `` typical '' inhomogeneous non - equilibrium density in sec . [ diffusion ] and then connect the evolution of the variance of this density to a current auto - correlation function at finite times @xcite .
remarkably , in the special limit of infinitely long times this connection will be shown to yield a generalized einstein relation which relates the diffusion constant to the dc - conductivity , i.e. as evaluated by the * kf*. moreover , we will demonstrate that the great advantage of the connection is given by its direct applicability at finite times and for finite systems as well . in this context
we will suggest another concept for the analysis of data which is available from current auto - correlations functions . by the use of the suggested concept we will investigate in secs . [ tilted]-[alternating ] transport in several translationally invariant spin-1/2 chains in the special limit of high temperatures .
we will concretely consider spin transport in the anisotropic heisenberg chain , the pure heisenberg chain within an alternating field , and energy transport in an ising chain which is exposed to a tilted field . by means of
numerically exact diagonalization we indeed observe strong indications for diffusive behavior in the considered spin chains for a range of model parameters and , what is more , we are able to quantitatively confirm the diffusion constants which were found for these systems from non - equilibrium bath scenarios in refs . .
finally , we will close in sec .
[ summary ] with a summary and a conclusion .
in this sec . we are going to introduce our approach to density - driven transport in translationally invariant quantum systems . to this end the next sec .
[ diffusion_a ] firstly presents the pertinent definitions and exclusively describes the general theory which eventually yields the basic eq .
( [ d ] ) for the time - dependent diffusion constant @xmath2 . in the following sec .
[ diffusion_b ] we then motivate to evaluate @xmath2 for finite times @xmath3 and particularly illustrate the concept which will be used for the concrete spin chains in the subsequent secs .
thus , the reader which is not primarily interested in the theoretical details may directly continue with sec .
[ diffusion_b ] . in the present paper
we will investigate translationally invariant , one - dimensional quantum spin systems which are described by a respective hamiltonian @xmath4 . in those quantum systems we will consider an overall conserved transport quantity @xmath5 , i.e. @xmath6 = 0 $ ] .
this transport quantity and the hamiltonian as well are decomposable into @xmath7 formally identical addends @xmath8 , respectively @xmath9 corresponding to different positions , i.e. @xmath10 thus , @xmath8 is a local density of the transported quantity @xmath5 .
note that the @xmath8 may be defined on the positions of the @xmath9 , in between , or both .
the above decomposition is further done in such a way that heisenberg s equation of motion for the local densities @xmath8 reads @xmath11 = \underbrace{\imath \ , [ \hat{h}_{\mu^- } , \hat{x}_\mu ] } _ { \equiv \hat{j}_{(\mu-1 ) } } + \underbrace{\imath \ , [ \hat{h}_{\mu^+ } , \hat{x}_\mu ] } _ { \equiv - \hat{j}_\mu } \ , , \label{continuity}\ ] ] where @xmath12 and @xmath13 represent those local addends of the hamiltonian @xmath4 which are located directly on the l.h.s .
, respectively r.h.s . of @xmath8 .
this apparently implies a kind of locality .
however , such a description can always be at least approximately enforced , if only interactions are reasonably short - ranged . for all quantum systems in the following secs .
[ tilted]-[alternating ] the description will be even exact , because interactions between nearest - neighbors are taken into account solely .
as routinely done , the comparison with a continuity equation suggests the definition of a local current @xmath14 according to the scheme in eq .
( [ continuity ] ) , see ref . , for example .
this definition is consistent , if @xmath15 } _ { -\hat{j}_\mu } + \underbrace{\imath \ , [ \hat{h}_{(\mu+1)^- } , \hat{x}_{(\mu+1 ) } ] } _ { \hat{j}_\mu } = 0 \ , , \ ] ] where the latter holds , if @xmath5 is globally preserved .
the total current @xmath16 is given by @xmath17 once the above decomposition has been established , we can define a certain class of initial states @xmath18 which corresponds to some inhomogeneous , non - equilibrium density . to those ends
let @xmath19 denote the deviation of the local densities @xmath8 from their equilibrium average @xmath20 , where @xmath21 is any stationary equilibrium state , i.e. @xmath22 = 0 $ ] .
then the initial state @xmath18 reads @xmath23 with some realization for the numbers @xmath24 .
the factor @xmath25 is concretely given by @xmath26 and therefore quantifies the equilibrium fluctuations of the transported quantity @xmath5 .
for the special initial state @xmath18 we now consider the actual expectation values @xmath27 it follows that @xmath28 i.e. the sum @xmath29 of the numbers @xmath24 in eq .
( [ initialstate ] ) determines the sum of the actual expectation values @xmath30 .
of particular interest is the spatial variance @xmath31 of the @xmath30 .
it is given by @xmath32 ^ 2 \ , .
\label{definitionw2}\ ] ] if the dynamics of the @xmath30 was indeed generated by a discrete diffusion equation of the form @xmath33 \ , , \label{diffusionequation}\ ] ] then the evolution of this variance would read @xmath34 as long as the @xmath30 vanish at the ends of a chain ( open boundary conditions ) or are reasonably concentrated at a sector of a ring ( closed boundary conditions ) .
even though eq .
( [ diffusionequation ] ) implies eq .
( [ w2 ] ) , the inverse direction generally is not true , of course . however , in ref .
a connection of the form ( [ w2 ] ) has recently been found directly from eq .
( [ definitionw2 ] ) , i.e. by firstly applying heisenberg s equation of motion to eq .
( [ definitionw2 ] ) and subsequently manipulating the resulting equations . the time - dependent diffusion constant @xmath2 reads @xmath35 and is essentially given in terms of a time - integral over the current auto - correlation function @xmath36 i.e. @xmath37 in the limit of high temperatures ( @xmath38 ) .
this limit will be considered throughout this work .
strictly speaking , the above diffusion constant @xmath2 is restricted to the special initial state @xmath18 in eq .
( [ initialstate ] ) , of course . nevertheless , due to its concrete form ,
this initial state represents an ensemble average w.r.t .
to typicality @xcite or , more precisely , the dynamical typicality of quantum expectation values @xcite .
thus , the overwhelming majority of all possible initial states with the same @xmath39 as @xmath18 is also expected to approximately yield the @xmath30 corresponding to @xmath18 , if the dimension of the relevant hilbert space is sufficiently large , see especially ref . .
the latter largeness is certainly fulfilled for all practical purposes . or ,
in other words , the concrete curves for @xmath30 and thus for @xmath31 only slightly `` fluctuate '' around the curve of the ensemble average .
in fact , for the single - particle quantum system in refs . it has been demonstrated that eqs .
( [ w2])-([c ] ) correctly describe the dynamics for all pure initial states @xmath40 which are not created explicitly in order to violate these equations , see ref . .
eventually , let us relate the non - perturbative result for the diffusion constant @xmath2 to the standard result of linear response theory for the dc - conductivity @xmath41 @xcite . in the limit of high temperatures
this conductivity may be written as @xmath42 where @xmath43 is the inverse temperature .
the comparison of eq .
( [ sigmadc ] ) with eq .
( [ d ] ) yields @xmath44 i.e. it leads to an einstein relation .
it is well - known that the integral in eq .
( [ sigmadc ] ) will diverge , whenever @xmath45 , the fourier transform of @xmath46 , has a finite contribution at @xmath47 , see ref . , for example .
as routinely done , we may hence define a drude weight @xmath48 , e.g. @xmath49 if the total current is strictly preserved , i.e. @xmath50 = 0 $ ] , the diffusion constant @xmath2 is completely governed by the drude weight @xmath48 , i.e. @xmath51 for all @xmath3 . consequently , transport is purely ballistic at each time , respectively length scale .
even if the total current does not represent a strictly conserved quantity , a non - zero drude weight directly implies that the diffusion constant @xmath2 is still approximately given by the straight line @xmath52 in the limit @xmath53 .
however , at some finite time or length scale the diffusion constant @xmath2 may nevertheless appear to be almost constant , i.e. @xmath54 , as expected for diffusive behavior .
such a behavior requires that the drude weight @xmath48 is not relevantly large ( on the corresponding scale ) , the finiteness of @xmath48 by itself is not crucial in this context . in fact , it is well - known that the classification of transport types for a given quantum system generally is a concept which crucially depends on the considered time or length scale , see refs . , for example .
the above line of reasoning becomes also relevant in those situations where some model parameter @xmath55 induces a transition from a finite towards a zero drude weight @xmath48 , e.g. where @xmath48 vanishes above some critical value of @xmath55 , say @xmath56 .
then at this critical value @xmath56 a sharp transition from ballistic towards non - ballistic transport is to be expected at the infinite time scale . however , such a transition does not necessarily appear at some finite time , respectively length scale .
even if there was a transition , a sharp one would require at the critical value @xmath56 a sudden jump of the drude weight @xmath48 from a finite and relevantly large number ( on the corresponding scale ) to zero .
we hence do not concentrate merely on drude weights and their finiteness , although those are also discussed , of course . instead of that we focus on the diffusion constant @xmath2 , as defined in eq .
( [ d ] ) , at finite times and for finite systems as well . according to eq .
( [ d ] ) , we can evaluate the underlying current auto - correlation function directly in the time domain , i.e. @xmath46 , instead of the frequency domain , i.e. @xmath45 .
this way we are not concerned with the problems which may arise due to the fact that @xmath45 is a highly non - smooth function for a finite system .
we particularly exploit that the @xmath57-curve does not change at sufficiently short time scales any more , when the size of a system becomes large enough .
thus , at those short time scales , interesting signatures of the infinitely large system may already be extractable for a system with an accessible size . in the following secs . [ tilted]-[alternating ]
this concept will be demonstrated in full detail .
, as given by eq .
( [ d ] ) , for energy transport in the ising chain within a tilted field in the high temperature limit ( @xmath58 ) .
insets zoom in @xmath2 at short @xmath3 .
parameters : @xmath59 , @xmath60 , @xmath61 ( top ) as well as @xmath62 ( bottom ) .
all curves are evaluated numerically ( exact diagonalization ) for chain lengths @xmath63 , @xmath64 , @xmath65 , @xmath1 ( arrows ) .
the circles ( top inset ) represent additional data for @xmath66 , extracted from ref . and computed by the use of approximative numerical integrators .
the cross ( top inset ) indicates the conductivity @xmath67 = 11 from ref .
@xcite , as obtained from a non - equilibrium bath scenario for the same set of parameters . ] of the diffusion constant @xmath2 , the drude weight , for energy transport in the ising chain within a tilted field at high temperatures ( @xmath58 ) .
parameters : @xmath59 and @xmath60 .
all open symbols are evaluated numerically ( by the use of exact diagonalization ) for chain lengths @xmath68 and for @xmath69-components @xmath70 .
the filled square is extracted from ref . and represents an upper bound for @xmath66 . ] in the present section we will study a first example of a translationally invariant spin system .
this system is an ising chain which is exposed to a , say , tilted magnetic field @xmath71 . concretely , its hamiltonian reads ( @xmath72 ) @xcite @xmath73 @xmath74 where @xmath75 and @xmath76 denote the @xmath69-component , respectively @xmath77-component of the total vector @xmath78 .
here , one might think of some magnetic field which originally was in line with the @xmath69-direction and has been rotated about the @xmath79-axis with the angle @xmath80 . in the hamiltonian
( [ h_tilted ] ) the operators @xmath81 ( @xmath82 ) are the standard pauli matrices ( corresponding to site @xmath83 ) ; @xmath84 denotes the coupling strength ; and @xmath7 represents the total number of sites . throughout this work
we use periodic boundary conditions , i.e. @xmath85 . due to the presence of the tilted magnetic field , the only trivial symmetries are translational invariance and mirror symmetry @xcite such that the whole hilbert space is decomposable into @xmath86 decoupled subspaces with similar dimensions . for the present model
the component of the total spin @xmath87 in @xmath88-direction ( and any other direction ) is non - preserved .
therefore magnetization ( or spin ) is not a suitable transport quantity here .
however , this model is appropriate to investigate the transport of energy , i.e. it allows to study the density dynamics of the local quantities @xmath9 , see eq .
( [ h_tilted ] ) .
note that these quantities contain contributions from both the ising interaction and the zeeman energy .
the respective energy current @xmath89 is given by @xcite @xmath90 @xmath91 = - \frac{b_x \
, j}{8 } \ , ( \hat{\sigma}_\mu^z - \hat{\sigma}_{\mu+2}^z ) \ , \hat{\sigma}_{\mu+1}^y \label{j_tilted}\ ] ] and the factor @xmath25 , as defined in eq .
( [ epsilon2 ] ) , reads @xmath92 if the component @xmath75 is identical to zero ( @xmath93 ) , the energy current ( [ j_tilted ] ) is strictly preserved .
furthermore , the hamiltonian ( [ h_tilted ] ) is well - known to be integrable .
but it can become quantum chaotic for @xmath94 , e.g. for the special set of parameters @xmath59 , @xmath61 , @xmath95 ( @xmath96 ) @xcite .
recently , for exactly this parameter set , a strong evidence for diffusive behavior has been found from non - equilibrium bath scenarios @xcite .
[ d_tilted ] ( top ) shows the diffusion constant @xmath97 , as given by eq .
( [ d ] ) , for the above set of parameters .
we observe that @xmath2 increases within the correlation time of the underlying current auto - correlation function but , already on this time scale , essentially the drude weight governs the overall shape of the curve , at least for small lengths @xmath98 .
but when @xmath7 is increased , the drude weight @xmath48 rapidly becomes smaller , i.e @xmath48 decreases faster than a power - law , see fig
. [ drude_tilted ] ( squares ) .
in particular there is no need to assume a finite and relevant value for @xmath48 in the thermodynamic limit @xmath99 .
already for @xmath100 the drude weight @xmath48 is such small that the tendency of the diffusion constant @xmath2 to gradually develop towards a horizontal line becomes visible .
nevertheless , only from those lengths which are available from numerically exact diagonalization a definite conclusion may still be vague , particularly conclusions about the time after which @xmath2 possibly remains constant and about the constant value @xmath101 , too .
fortunately , additional data is available in literature : in ref .
the current auto - correlation function has been evaluated also by the use of an approximative numerical integrator .
based on this data the time evolution of the diffusion constant @xmath2 can be extracted @xcite for the length @xmath66 , see fig .
[ d_tilted ] ( top inset ) . and in fact , for times which are larger than @xmath102 we observe that @xmath2 takes on a constant value @xmath103 .
the latter value further is in excellent agreement with the conductivities @xmath104 from ref . and
@xmath105 from ref .
@xcite , as obtained therein for the same set of parameters from non - equilibrium bath scenarios .
( in these works @xmath67 is denoted by @xmath106 . ) when the @xmath69-component is decreased from @xmath107 down to @xmath108 , diffusive behavior eventually breaks down towards ballistic transport , because the current becomes strictly preserved for @xmath109 .
unfortunately , it is hardly possible to give a critical value for this transition , simply due to the limited system sizes which are accessible by the use of exact diagonalization .
however , such a value may be very unsharp for a continuous transition , i.e. if the drude weight changes smoothly and does not jump suddenly from zero ( or a finite , irrelevantly small number ) to a finite , relevantly large number , cf .
[ diffusion ] .
contrary , when the @xmath69-component is increased from @xmath110 , the indications of diffusive behavior become much more pronounced for the accessible system sizes , cf . fig .
[ d_tilted ] ( bottom ) .
respective diffusion constants may therefore be suggested for this parameter regime .
, as given by eq .
( [ d ] ) , for magnetization transport ( spin transport ) in the anisotropic heisenberg chain ( xxz model ) in the high temperature limit ( @xmath58 ) .
insets zoom in @xmath2 at short @xmath3 .
the curves are evaluated numerically ( by the use of exact diagonalization ) for the anisotropy parameter @xmath111 and for chain lengths @xmath63 and @xmath112 . ] , as given by eq .
( [ d ] ) , for magnetization transport ( spin transport ) in the anisotropic heisenberg chain ( xxz model ) in the high temperature limit ( @xmath58 ) .
insets zoom in @xmath2 at short @xmath3 .
the curves are evaluated numerically ( by the use of exact diagonalization ) for chain lengths @xmath63 , @xmath64 , @xmath65 , @xmath112 ( arrows ) and for anisotropy parameters @xmath113 ( top ) , @xmath114 ( middle ) , as well as @xmath115 ( bottom ) .
one cross ( middle inset ) indicates the conductivity @xmath116 in ref .
@xcite , as found therein from a non - equilibrium bath scenario .
another cross ( bottom inset ) represents the value @xmath117 with the dc - conductivity @xmath118 according to ref . .
] of the diffusion constant @xmath2 , the drude weight , for magnetization transport ( spin transport ) in the anisotropic heisenberg chain ( xxz model ) at high temperatures ( @xmath58 ) .
the symbols are evaluated numerically ( by the use of exact diagonalization ) for chain lengths @xmath119 and for anisotropy parameters @xmath120 .
the line indicates a fit for the special case of @xmath121 . ]
in this section we will investigate transport in the anisotropic heisenberg chain ( or xxz model ) as another and certainly more interesting example of a translationally invariant spin system .
it is described by a hamiltonian of the form ( @xmath122 ) @xcite @xmath73 @xmath123 with the anisotropy parameter @xmath124 .
independent from the concrete choice of @xmath124 , the hamiltonian ( [ h_anisotropic ] ) is integrable in terms of the bethe ansatz , see ref .
, for example . in the presence of an external ( uniform ) magnetic field @xmath125 one may add to the hamiltonian ( [ h_anisotropic ] ) a zeeman term of the form @xmath126 .
however , because the following investigation ( at infinite temperature ) will not depend on the concrete choice of @xmath125 , we set @xmath127 for simplicity . as well - known , the hamiltonian ( [ h_anisotropic ] ) is invariant under rotations about the z - axis , i.e. it commutates with @xmath128 . as a consequence @xmath4
can be diagonalized within decoupled subspaces with dimensions @xmath7 over @xmath129 , where @xmath130 is the quantum number w.r.t .
moreover , due to the translational invariance as well as the mirror symmetry of ( [ h_anisotropic ] ) , the problem can be reduced further by a factor @xmath86 .
since @xmath128 represents a strictly conserved quantity , this model allows to investigate the density dynamics of the local quantities @xmath131 , i.e. magnetization transport ( or spin transport ) . the respective magnetization current @xmath132
is given by @xcite @xmath133 @xmath134 = \frac{j}{4 } ( \hat{\sigma}_\mu^x \hat{\sigma}_{\mu+1}^y -\hat{\sigma}_\mu^y \hat{\sigma}_{\mu+1}^x ) \ , \label{j_anisotropic}\ ] ] and the factor @xmath25 , as defined in ( [ epsilon2 ] ) , takes on the fixed value @xmath135 . similarly , an energy current can be also introduced via the relation @xmath136 $ ] .
but , since its commutator with the hamiltonian ( [ h_anisotropic ] ) vanishes exactly , this current is strictly conserved such that energy transport is purely ballistic at each time- and length - scale ( and for all finite temperatures , see ref . ) . except for the special case of @xmath137 ( xy model ) the magnetization
current is non - preserved . on that account
it could be in principle possible to find diffusive behavior for non - zero anisotropies .
nevertheless , in the literature there is strong evidence that such a behavior is restricted to the regime @xmath138 only @xcite , most promising appear those @xmath124 which are close to @xmath114 @xcite .
those expectations are also due to results on the drude weight @xmath48 @xcite , e.g. bethe ansatz approaches suggest that @xmath48 is finite for the regime @xmath139 and zero for the regime @xmath138 @xcite . still controversial is the special case of @xmath140 , where @xmath48 may be already zero @xcite or not @xcite . for completeness
, we also show numerical results for the drude weight in fig .
[ drude_anisotropic ] , although respective data can be found already in ref . ,
also obtained by the use of exact diagonalization and additionally extrapolated to @xmath99 .
this data can be transferred directly to the present investigation , if the concrete values for the drude weight are multiplied by a factor @xmath141 .
further results on drude weights from exact diagonalization can be found in , e.g. refs . .
[ d_anisotropic_0.5 ] shows the diffusion constant @xmath97 , as defined in eq .
( [ d ] ) , for the anisotropy parameter @xmath111 . for @xmath63
the drude weight @xmath48 determines the overall shape of the @xmath2-curve , i.e. on all time scales @xmath2 is very close to the straight line @xmath52 .
there is only an insignificant deviation from this line at short times scales below the correlation time of the underlying current auto - correlation function , see fig .
[ d_anisotropic_0.5 ] ( inset ) . for @xmath142 ,
i.e. when the length of the chain is more than doubled , the @xmath2-curve remains practically the same . on that account
it appears to be justified to assume a similar curve in the thermodynamic limit @xmath99 .
this assumption is consistent with the extrapolation of the drude weight in ref .
, see fig . [ drude_anisotropic ] ( squares ) , too .
we may therefore suggest purely ballistic behavior for the case @xmath111 . note that sirker et al .
have recently presented results in ref . which point towards a coexistence of diffusive and ballistic dynamics , where the ballistic distribution is small due to a small drude weight
even though we reproduce that the drude weight is large for @xmath143 and does not depend significantly on @xmath7 , we can not exclude the possibility that the drude weight eventually becomes small for @xmath144 , of course . in fig .
[ d_anisotropic ] ( middle ) we display the diffusion constant @xmath2 for the anisotropy @xmath121 .
we observe that @xmath2 increases at short time scales , i.e. within the correlation time of the underlying current auto - correlation function @xmath46 , and then remains approximately constant for an interval at intermediate time scales , i.e. a `` plateau '' is formed at those times .
finally , a renewed increase takes place on long time scales which is completely governed by the drude weight , or by the zero - frequency distribution of @xmath45 .
note that the plateau can not be seen , if zero- and finite - frequency parts are treated separately from each other .
it is a feature which arises from a combination of both contributions .
the above plateau remarkably is already visible for @xmath63 .
moreover , its `` height '' does not change with @xmath7 , while its `` width '' seems to increase gradually , see fig .
[ d_anisotropic ] ( middle inset ) .
the latter increase particularly appears to be plausible , because the drude weight is commonly expected to vanish in the thermodynamic limit @xmath146 , as already outlined above .
we hence make the educated guess that the plateau of @xmath2 will be continued to , say , arbitrary long times , when only @xmath7 becomes sufficiently large , e.g. @xmath99 .
then the height @xmath147 directly determines the concrete value of the diffusion constant , of course . and
indeed , the latter value for @xmath148 also is in excellent agreement with the conductivity @xmath116 in ref .
@xcite , as found therein from a non - equilibrium bath scenario .
although a respective figure is not shown for @xmath149 , such an agreement is additionally obtained with the results in ref .
, namely @xmath150 and @xmath151 .
( in these works @xmath67 is denoted by @xmath106 . )
those quantitative agreements support the correctness of our guess and the emergence of diffusive transport appears to be verified for those @xmath124 which are close to @xmath114 .
independent from the above arguments regarding the long - time extrapolation for @xmath99 , we should mention that for @xmath142 the @xmath2-curves have already converged for those times which are shorter than @xmath152 , see fig .
[ d_anisotropic ] ( middle inset ) . due to
( [ w2 ] ) , this time corresponds to a length scale @xmath153 on the order of about @xmath154 sites .
thus , at least for @xmath155 the dynamics should be known . for larger anisotropy parameters
@xmath124 we find a similar situation : again there is a clearly developing plateau at intermediate time scales , see fig .
[ d_anisotropic ] ( bottom ) for the case @xmath156 . for the accessible system sizes , however
, the final plateau height @xmath148 does not seem to be reached yet , i.e. , @xmath148 can not be read off with the same accuracy , as done before for @xmath157 . for comparison
we hence indicate in fig .
[ d_anisotropic ] ( bottom inset ) the quantity @xmath117 with the dc - conductivity @xmath158 according to ref .
, obtained therein from an analysis on the basis of the standard green - kubo formula .
the good agreement with this result is noticeable . as already mentioned above , it is still controversial , whether or not the drude weight is finite in the limit @xmath159 , when @xmath124 becomes exactly @xmath160 .
however , let us for the moment assume that the drude weight is indeed finite in that limit .
then at the infinite time scale the diffusion constant @xmath2 is completely governed by this non - zero drude weight and increases linearly . at finite time
scales , however , @xmath97 may still appear to be almost constant and feature a plateau like the one for the case @xmath121 . in principle such a plateau can be very wide , if only its height is large in comparison with the drude weight @xmath48 such that the finiteness of @xmath48 is by itself not crucial in this context , cf .
[ diffusion ] .
we hence show in fig .
[ d_anisotropic ] ( top ) the diffusion constant @xmath2 for the special case @xmath140 .
the initial short - time increase directly passes into the final long - time increase , i.e in between a horizontal line is not observable for the accessible system sizes .
nevertheless , we can not exclude the possibility that a plateau with a height on the order of @xmath161 will eventually develop , when the system size is further increased , cf .
[ d_anisotropic ] ( top inset ) .
we may thus compare this value for @xmath101 with the extrapolated drude weight @xmath162 in ref .
@xcite , for example .
( the latter extrapolated @xmath48 is about @xmath135 of the long - time slope for @xmath142 , cf . figs .
[ d_anisotropic ] and [ drude_anisotropic ] . ) by the use of the rough estimation latexmath:[$d \ , @xmath54 can not be wider than some time scale @xmath164 , respectively length scale @xmath165 . thus ,
if the above extrapolation for the drude weight was indeed correct , diffusion would be restricted to very few sites only .
contrary , if the drude weight was zero ( or a finite , irrelevantly small number ) , diffusion could occur for a possible arbitrary number of sites , of course .
, as given by eq .
( [ d ] ) , for magnetization transport ( spin transport ) in the heisenberg chain within an alternating field in the high temperature limit ( @xmath58 ) .
insets zoom in @xmath2 at short @xmath3 .
the curves are evaluated numerically ( by the use of exact diagonalization ) for chain lengths @xmath166 , @xmath167 , @xmath65 , @xmath168 ( arrows ) and for field strengths @xmath169 ( top ) as well as @xmath170 ( bottom ) .
the crosses ( insets ) indicate the conductivities @xmath171 and @xmath172 in ref .
, as obtained from a non - equilibrium bath scenario . ]
of the diffusion constant @xmath2 , the drude weight , for magnetization transport ( spin transport ) in the heisenberg chain within an alternating field at high temperatures ( @xmath58 ) .
the symbols are evaluated numerically ( by the use of exact diagonalization ) for chain lengths @xmath173 and for field strengths @xmath174 . ]
in this section we are going to concentrate exclusively on the pure heisenberg chain , i.e. @xmath140 , and investigate another modification of the hamiltonian ( [ h_anisotropic ] ) which may lead to diffusive transport .
such a modification certainly is the incorporation of disorder , i.e. coupling constants or field strengths which vary randomly from one site to another @xcite .
but the investigation of disorder is numerically rather challenging and conceptually more subtle due to the localization phenomenon , even for a non - interacting single - particle system @xcite .
instead we will consider another scenario without disorder , where the heisenberg chain is exposed to a strictly alternating field .
this scenario is concretely described by the hamiltonian ( [ h_anisotropic ] ) and an additional zeeman term of the form @xcite @xmath175 where @xmath176 is the deviation from the mean @xmath125 . as already done before , we may set @xmath127 for simplicity
. the presence of the zeeman term ( [ h_alternating ] ) does not affect the commutation of @xmath4 and @xmath128 , i.e. the hamiltonian can be diagonalized within decoupled @xmath130-subspaces , too .
as long as @xmath7 is even , there also is translational invariance ( w.r.t two sites ) such that the problem can be reduced further by a factor @xmath177 .
( there is no mirror symmetry for even @xmath7 . )
since @xmath128 still represents a strictly conserved quantity , the present model allows to investigate the transport of magnetization w.r.t . the field parameter @xmath176 .
note that the respective current @xmath132 and factor @xmath25 are identical to those in sec .
[ anisotropic ] .
when @xmath176 is increased from zero and becomes comparative with the coupling strength @xmath84 , the model undergoes a transition to quantum chaos @xcite , i.e. one may assume that the latter transition already is a first pointer towards the onset of diffusion .
even though it is entirely independent from those assumptions , we start our investigation with @xmath169 . for @xmath169
the curve for the diffusion constant @xmath2 has changed from fig .
[ d_anisotropic ] ( top ) into fig .
[ d_alternating ] ( top ) .
in particular the initial short - time increase does not directly pass into the final long - time increase any more .
instead there is an oscillation in between such that , at the first view , the situation appears to be much more complicated than a simple horizontal line .
however , for @xmath178 the @xmath2-curve has already converged until the end of this oscillation , see fig . [ d_alternating ] ( top inset ) .
exactly at this position the curve seems to gradually develop a plateau with a height on the order of @xmath179 .
the plateau becomes visible , since the drude weight rapidly decreases with @xmath7 , i.e. it is about one order of magnitude smaller for @xmath178 than for @xmath63 , cf . fig
. [ drude_alternating ] ( circles ) .
note that the drude weight does not fulfill a simple @xmath180-dependence and decays faster than a power - law .
thus , there is no need to suppose a finite and relevant drude weight in the limit @xmath159 .
remarkably , the above suggested @xmath101 for @xmath181 agrees well with the conductivity @xmath182 from ref .
, as found therein from a non - equilibrium bath scenario .
the deviation from this conductivity is on the order of @xmath183 solely .
however , the small deviation may be explained as follows : in ref .
the conductivity is evaluated by the use of the lindblad equation , i.e. , it is extracted from the steady state of a finite chain which is at both ends weakly coupled to baths at different temperatures .
this weak bath - coupling is essential for the validity of the lindblad approach .
but , for a finite chain , a too weak bath - coupling may also yield a value @xmath67 which is lower than the correct value , say , @xmath101 , see ref . .
since in ref
. the independence of the conductivity from the bath - coupling strength is not analyzed in detail , we suppose that the above deviation results from a slightly too weak coupling of chain and reservoirs . for larger field parameters @xmath176 suggestions are less reliable , since the initial oscillations of the @xmath2-curves become much more pronounced , as already visible in fig .
[ d_alternating ] ( bottom ) .
even though there are oscillations , @xmath2 still remains strictly positive , i.e. those oscillations probably are no pointer towards insulating behavior .
note that definite conclusions can not be made for the parameter regime @xmath184 , since the curves for @xmath2 continuously change from fig .
[ d_anisotropic ] ( top ) into fig .
[ d_alternating ] ( top ) , i.e. one basically is concerned with the problem that the drude weight governs the overall shape of the curve , at least for accessible system sizes .
in the present paper we have investigated transport in several translationally invariant spin-@xmath0 chains in the special limit of high temperatures . we have concretely considered spin transport in the anisotropic heisenberg chain , in the pure heisenberg chain within an alternating field , and energy transport in an ising chain which is exposed to a tilted field . to this end
we have firstly reviewed on a recently derived connection between the evolution of the variance of some `` typical '' inhomogeneous non - equilibrium density and the current auto - correlation function at finite times @xcite . in the limit of infinitely long times this connection was shown to yield a generalized einstein relation which relates the diffusion constant ( in the absence of any external force ) to the dc - conductivity ( as the linear response coefficient in the presence of an external force , i.e. as evaluated from the standard green - kubo formula @xcite ) .
however , we have additionally demonstrated that the great advantage of the above connection is given by its direct applicability at finite times and for finite systems , e.g. at short times interesting signatures of an infinitely large system may be extractable for a system with an accessible size . by means of
numerically exact diagonalization we have indeed observed strong indications for diffusive behavior in the considered spin chains for a range of model parameters .
moreover , the suggested diffusion constants have been found to be in quantitative agreement with recent results on diffusion coefficients which were obtained for the same spin chains from numerically involved investigations of non - equilibrium bath scenarios in refs . .
amongst all those and our findings at high temperatures the emergence of diffusive transport of magnetization appears to be verified in the anisotropic heisenberg chain with the anisotropy parameter @xmath121 @xcite , despite the integrability of the model .
it is known that the onset of quantum chaos is not a sufficient condition for diffusion @xcite but the latter result suggests also that non - integrability is not necessary at all .
unfortunately , it is still an open question whether spin transport in the pure heisenberg chain is diffusive or ballistic . simply by the use of numerically exact diagonalization
we were not able to reach those system sizes which are required for any definite conclusion on that question .
therefore the next step certainly is the application of approximate methods in order to obtain the current auto - correlation function at finite times , either numerical ones , e.g. suzuki - trotter decompositions @xcite , or analytical ones , e.g. projection operator techniques @xcite or moment methods @xcite .
approximative methods will also be in indispensable , when the investigation is extended from one - dimensional spin chains to , e.g. more - dimensional spin lattices .
apart from the above methodic details , a physically interesting question is the dependence of transport on temperature , of course .
since the connection between the evolution of the variance and the current auto - correlation function is not restricted to infinite temperature , further investigations may be done in this direction , too .
however , we expect that for low temperatures the convergence of the diffusion constant at finite times is much slower , when the system size is increased .
thus , for low temperatures conclusions on the basis of a finite system may be less reliable . | we investigate transport in several translationally invariant spin-@xmath0 chains in the limit of high temperatures .
we concretely consider spin transport in the anisotropic heisenberg chain , the pure heisenberg chain within an alternating field , and energy transport in an ising chain which is exposed to a tilted field .
our approach is essentially based on a connection between the evolution of the variance of an inhomogeneous non - equilibrium density and the current auto - correlation function at finite times .
although this relationship is not restricted to the case of diffusive transport , it allows to extract a quantitative value for the diffusion constant in that case . by means of
numerically exact diagonalization we indeed observe diffusive behavior in the considered spin chains for a range of model parameters and confirm the diffusion coefficients which were obtained for these systems from non - equilibrium bath scenarios . |
fluctuations of physical quantities near zero frequency have been investigated by several authors since the papers of @xcite and @xcite .
a general theory on the fluctuation - dissipation theorem , which will be the starting point of this paper , was developed by @xcite . to the best of our knowledge ,
a concrete expression for the low - frequency spectrum of fluctuations of magnetic fields in a thermal plasma was obtained for the first time by @xcite .
they found a peak around @xmath1 magnetic fluctuation which was interpreted as the evanescent energy component of electromagnetic fluctuations `` screened '' in a plasma below the plasma frequency .
the impact of such result into the cosmic microwave background was then investigated by @xcite .
although in these two references the authors claim that the fluctuations were rigorously computed , several approximations were indeed made and they were not able to get an unique formula covering the low- and high - frequency spectrum .
the aim of this paper is to reevaluate the derivation of the spectrum of magnetic fluctuations avoiding approximations in the low - frequency region and also in the transition between the low- and the high - frequency spectrum .
several different behaviors between our results and the previous one , mainly in the low - frequency part of the spectrum , are found and discussed .
the fluctuation - dissipation theorem developed by @xcite is able to deal with the thermal fluctuations inside a plasma in or near thermal equilibrium .
the expression for the magnetic field fluctuation in an homogeneous isotropic non - magnetized equilibrium plasma was obtained by @xcite looking at waves in such a plasma . in an electron - positron plasma , for example
, the magnetic fluctuations in wave number and frequency space is given by @xmath2 \omega^2}\end{aligned}\ ] ] where @xmath3 is the boltzmann constant and @xmath4 and @xmath5 are , respectively , the plasma and the collisional frequencies . here
@xmath6 , and @xmath7 being @xmath8 the particle density inside the plasma , @xmath9 is the atomic number of these constituents , @xmath10 and @xmath11 are respectively the charge and the mass of the electron , and @xmath12 is the lorentz factor given by @xmath13 .
this result can be integrated in @xmath14 to get ( the fourier transform ) @xmath15 and the magnetic energy density is @xmath16 .
the integral of eq .
( [ original - formula ] ) over wave numbers shows a high wave number divergence . according to @xcite , this
is expected since the derivation is based on classical fluid equations of motion and the constant collision frequency @xmath5 is independent of @xmath17 .
however , they prefer to carry on their analyzes in the simpler phenomenological approach . in order to overcome the large @xmath17 dependence
, they first take the limit @xmath18 and then they integrate over @xmath17 to infinity , which corresponds to the vanishing cross section of collisions as @xmath19 .
this is a very delicate point and we will turn back to this point in section [ general ] . for both the high frequency and high wave number
limits the authors emphasized that the expression of eq .
( [ original - formula ] ) has a substantial value only where @xmath20 .
the combined high - frequency and high wave - number limits were get by letting @xmath18 .
the expression for the low - frequency spectrum was obtained by breaking up the @xmath17 integral into two intervals , by introducing a cutoff value @xmath21 , with @xmath22 . in the integration from 0 to @xmath21 ,
@xmath5 was kept finite while in the integral from @xmath21 it was used the approximation @xmath18 .
the expressions obtained for the high and low parts of the spectrum was , respectively : @xmath23 and @xmath24 where @xmath25 is the heaviside step function , @xmath26 , @xmath27 , and @xmath28 .
finally the zero frequency limit of the magnetic fluctuations is give by @xmath29 at this point the frequency spectral intensity was plotted for a temperature @xmath30 k by requiring that the value of @xmath21 ( or @xmath31 ) provide a smooth behavior at the joint between the low - frequency spectrum and the black - body spectrum .
the choice was @xmath32 or ( @xmath33 .
the result for other values of the temperature were presented in another paper by @xcite .
the main claims by these authors was that the intensity of the spectrum does not vary sensitively with @xmath21 and that near @xmath34 the spectrum goes like @xmath35 .
let us now show our results .
we have integrated eq .
( [ original - formula ] ) over @xmath17 analytically , without any approximation , by the partial fractions technique . the exact result for the indefinite integral over
the wave number is : @xmath36\right\}\end{aligned}\ ] ] with the following definitions : @xmath37 is the debye length , @xmath38 , @xmath39 , @xmath40 , @xmath41 , @xmath42 , @xmath43 , @xmath44 , @xmath45 , @xmath46 , @xmath47 , @xmath48 , and @xmath49 where @xmath50 .
note that the general result shows only a linear divergence in @xmath17 restricted to the first term of eq .
( [ indef_integ ] ) .
this term , however , can not be simply discarded by doing the limit @xmath51 before we integrate over @xmath17 , as did by @xcite , since it plays a very important role when the strict limit @xmath52 ( which we prefer to indicate from now on by @xmath53 ) is to be considered , even when @xmath17 is large .
indeed , if we discard it for all large values of @xmath17 it can be shown that the limit @xmath54 of @xmath55 will be negative .
therefore , our result for the definite integral can be put in the form @xmath56 where @xmath57 the term @xmath58 will be taken as @xmath59 with @xmath60 as large as we want .
this will render the confrontation with the previous result easier .
we note that for @xmath61 @xmath62 , @xmath63 .
these three terms can be written as functions of the original parameters , considering that @xmath64 } + \omega(\omega^2 + \eta^2 -\omega_p^2)\right]^{1/2}\ ] ] and @xmath65 } - \omega ( \omega^2 + \eta^2 -\omega_p^2)\right]^{1/2}\ ] ] compared to eqs .
( [ tajima_high])-([zero_lim_tajima ] ) it is immediately evident how our result is different from those of equations , showing a much more complicated dependence of the frequency spectrum on the variable @xmath66 , and on the parameters @xmath4 and @xmath5 , which dependencies on plasma temperature are shown , respectively , in figures [ fig - omega ] . our full result @xmath55 is shown in fig .
[ fig - full ] , where both the normalized ( gray curve ) and non normalized ( green curve ) spectral intensities are given , for @xmath67 @xmath62 e @xmath68
k. and normalized @xmath69 spectral intensities showing the different components of the spectrum , @xmath58 , @xmath70 and @xmath71 , as given by eq .
( [ our - limit ] ) . ] in this figure we have also plotted each one of the terms that contribute to @xmath55 are shown in different colors . the deep we see in this figure near @xmath72 tends to disappear as @xmath0 goes to high values ; as @xmath0 decreases up to @xmath73 @xmath62 , the ordinate of the deep tends to zero .
the detail of the non normalized spectral intensity near @xmath74 is shown in fig .
[ fig - detail ] .
it shows naturally a smooth behavior between the low - frequency spectrum and the blackbody spectrum , which is constructed by hand in @xcite .
.,height=302 ] if we compare this graph to the correspondent one shown in fig .
1(a ) of the paper of @xcite , we see that our result , for @xmath75 , is a classical blackbody radiation spectrum which goes to zero at @xmath76 while their blackbody spectrum has a much greater width ( more than two orders of magnitude ) .
another important qualitative difference between the non approximate and the approximate results is that we found a very peculiar oscillations in @xmath55 for the @xmath77 region of the spectrum , as can be seen from figure [ fig - detail - k10 ] .
these oscillations occur in an @xmath78 region where the classical blackbody spectrum still have a significant value ; there is however , in this case , a strong interference in the total spectrum @xmath55 , eq .
( [ our - result ] ) , due to a change of sign of the function @xmath79 .
such a kind of behavior was found just for cut - off values of the order of @xmath80 @xmath62 . for values of @xmath0 greater than this one
such fluctuations disappear . in any case , this feature confirm our statement that the result can vary sensitively with @xmath0 , contrary to what was sustain by @xcite .
@xmath62 . ] finally , we have studied the behavior of @xmath81 by varying @xmath66 e @xmath5 .
the result is shown in figure [ fig - surf6 ] .
notice that just the peak of the zero - frequency plasma spectrum depends on @xmath5 ( the blackbody part remains unchanged ) .
indeed , the spectral intensity varies two orders of magnitude by varying @xmath5 by two orders too , namely , it goes from @xmath82 , for @xmath83 , to @xmath84 , for @xmath85 .
thus , our result indicates that , when one goes backwards in time , temperature grows , dynamo action is enhanced ( since @xmath5 goes down ) , and the resonance peak of the zero - frequency plasma peak goes down . and
in this paper we have computed the spectrum of magnetic fluctuations of an homogeneous cosmic plasma avoiding any approximations .
several different behaviors between our results and the previous one obtained by @xcite , mainly in the low - frequency part of the spectrum , are found and discussed .
it is important to stress that the exact result indicates that the peak of the zero - frequency spectrum can indeed vary sensitively with the cut - off value @xmath86 . in the light of this new result , and following the papers of @xcite and @xcite , the problem of establishing an upper limit for fractal space dimensionality from cobe data can be revisited .
our results can still be improved towards cosmological applications by computing the fourier transformed volume element @xmath87 in terms of curved riemannian space embedded into general relativistic spacetime .
this will allow us to address the problem of dynamo action in einstein cosmology .
the dynamo effect in plasmas is a competitive effect between convection of the cosmological fluid and the plasma resistivity .
this is the reason why is so interesting to consider the relation between the dynamo action @xmath12 and the plasma resistivity and its frequency . recently ,
some new investigations on this directions were addressed by @xcite , by using the magnetic field correlation tensor in space of negative curvature , and by @xcite in the case of positive curvature .
also recently , @xcite has obtained a constraint on dynamo action from cobe data , using two - dimensional spatial sections of negative curvature of friedmann universe , based on the general relativistic magnetohydrodynamic equation derived by @xcite . in near future , by using the formula of @xcite , @xmath88 where @xmath12 is the magnetic field growth rate ( dynamo action ) , we expect to compute @xmath12 in terms of the magnetic plasma resistivity ( @xmath5 ) and plasma frequency ( @xmath4 ) . in this way we shall estimate the variation of dynamo action in terms of @xmath5 and @xmath4 .
the authors would like to thank garcia de andrade for useful comments . | magnetic fluctuations in a non - magnetized gaseous plasma is revisited and calculated without approximations , based on the fluctuation - dissipation theorem . it is argued that the present results are qualitative and quantitative different form previous one based on the same theorem . in particular , it is shown that it is not correct that the spectral intensity does not vary sensitively with @xmath0 .
also the simultaneous dependence of this intensity on the plasma and on the collisional frequencies are discussed . |
the next generation of wide - area extragalactic submillimetre and millimetre ( hereafter sub - mm ) surveys , for example from the balloon - borne large aperture submillimetre telescope ( blast , devlin et al 2001 ) , laboca on the atacama pathfinder experiment ( apex ) , the scuba2 camera on the james clerk maxwell telescope ( jcmt ) and bolocam - ii on the large millimetre telescope ( lmt ) , will produce large samples ( @xmath2 ) of distant , luminous starburst galaxies .
the dramatic increase in the number of submillimetre detected galaxies requiring follow - up observations makes it unreasonable to expect that a large fraction of their obscured or faint optical and ir counterparts will have unambiguous , spectroscopically - determined redshifts .
an alternative method to efficiently and robustly measure the redshift distribution for large samples of submillimetre galaxies is clearly necessary . given the underlying assumption that we are witnessing high rates of star formation in these submillimetre galaxies , then we expect them to have the characteristic fir peak and steep submillimetre ( rayleigh - jeans ) spectrum which is dominated by thermal emission from dust heated to temperatures in the range @xmath3k by obscured young , massive stars .
the observed radio fir luminosity correlation in local starburst galaxies ( e.g. helou et al .
1985 ) , that links the radio synchrotron emission from supernova remnants with the later stages of massive star formation , is also expected to apply to the submillimetre galaxies .
thus , in recent years , a considerable amount of effort has been invested in assessing the accuracy with which these broad continuum features in the spectral energy distributions ( seds ) of submillimetre galaxies at rest - frame mid - ir to radio wavelengths can be used to provide photometric - redshifts ( hughes et al .
1998 ; carilli & yun 1999 , 2000 ; dunne , clements & eales 2000 ; rengarajan & takeuchi 2001 ; yun & carilli 2002 ; wiklind 2003 ; blain , barnard & chapman 2003 ) . in a contribution to this general investigation
hughes et al .
( 2002 , paperi ) described monte - carlo simulations that used a library of multi - frequency template seds , derived from observations of local starbursts and agn with a wide - range of fir luminosities ( @xmath4 ) and temperatures ( @xmath5 ) , to measure the accuracy of photometric redshifts that could be derived from future 250 , 350 , 500@xmath6 m extragalactic surveys with blast and herschel , and complementary 850@xmath6 m surveys from scuba .
aretxaga et al .
( 2003 , paperii ) then applied the techniques described in paperi to the catalogues of submillimetre galaxies identified in various scuba surveys , and derived photometric redshifts for individual sources using existing radio submillimetre data . in this paper
we use new optical and ir spectroscopic observations of submillimetre galaxies published by chapman et al .
( 2003a ) , and simpson et al .
( 2004 ) to update our previous comparison of spectroscopic and long - wavelength photometric redshifts ( paper ii ) . in section 2
we explain the selection criteria for the sub - mm sources with spectroscopic redshifts included in our analysis , and demonstrate how the accuracy of the photometric redshift prediction varies according to the quality of the radio mm fir photometric data . in section 3
we discuss the significant agreement between the spectroscopic and photometric redshifts , and use these results to challenge the suggestion by blain et al .
( 2003 ) that it is not possible to derive photometric redshifts with an accuracy of @xmath7 or better without adopting an unreasonably tight dispersion in the luminosity - temperature ( l - t ) relation or a limited range of seds in the analysis .
finally , our conclusions are summarized in section4 .
the cosmological parameters adopted throughout this paper are @xmath8 kms@xmath9mpc@xmath9 , @xmath10 , @xmath11 .
over the last year several new optical and ir spectroscopic redshifts have been published for sub - mm sources ( chapman et al .
2003a , simpson et al . 2004 ) , enhancing substantially the number of sources available for checking the reliability of photometric estimates .
many of the sub - mm galaxies selected for spectroscopic study have been extracted from the 8-mjy scuba survey ( scott et al .
2002 ) , aided by deep follow - up observations with the vla ( ivison et al .
2002 ) . of the 10 optical redshifts published by chapman et al .
( 2003a ) , we include in our analysis only those 5 sources with photometric data of sufficient quality .
we briefly discuss the sed properties of the excluded sources : * smm105224.6 + 572119 ( @xmath12 ) , also known as le850.15 from the uk 8-mjy scuba survey of the lockman hole east , is rejected because the sub - mm source has since been shown to be a spurious 850@xmath6 m detection . in scott
2002 ) this source was originally listed as a low s / n ( 3.4@xmath13 ) submm source , having been extracted from one of the more noisy regions of the 8-mjy maps .
subsequently it was highlighted by ivison et al .
( 2002 ) to be very likely erroneous on account of its lack of a radio counterpart ( despite its very bright apparent sub - mm flux ) .
the spurious nature of this source has been supported by the failure to detect the source in a reanalysis of the 8-mjy maps conducted as part of shades ( mortier et al .
2004 ; www.roe.ac.uk/@xmath14ifa/shades ) , and by its non - detection in the iram mambo mm - wavelength maps of the 8-mjy lockman east field ( greve et al .
* smm105207.7 + 571907 ( @xmath15 ) or le850.12 , also from the uk 8-mjy scuba survey , has a bright and statistically secure radio counterpart ( ivison et al .
this object , however , is a known agn , with strong , flat - spectrum , variable radio emission .
therefore , it must be excluded from the analysis presented here , which relies on the fitting of non - variable starburst - dominated spectral templates .
as we show in fig 3 , no sed in our local template library can reproduce the high radio fluxes of this variable radio - source at the spectroscopic redshift , and indeed , in paper ii we already reported that the sed could not be matched at any redshift with the local templates .
* smm123600.2 + 621047 ( @xmath16 ) , whose position ( j2000 ) corresponds to a 1.4ghz radio source ( 123600.150 + 621047.17 ) in richards ( 2000 ) , has a spectrum which presumably arises from an optically - faint ( @xmath17 ) vla radio - source that was followed up with submillimetre photometry , but the robustness of the radio - submm association can not be determined from the published information .
* smm131212.7 + 424423 ( @xmath18 ) still do not have published 1.4ghz data from the ssa13 field ( fomalont et al .
, in prep ) , and we therefore can not reconstruct the radio sub - mm sed of this source . in the analysis described in 2.2
we include the remaining 5 sub - mm sources from the 8-mjy survey studied by chapman et al .
( 2003a ) , namely n2850.1 , n2850.4 , n2850.8 , le850.6 and le850.18 , which have unambiguous radio identifications from ivison et al .
the spectroscopic redshifts for all these sources also appear to be secure , and indeed redshift information for two of them has been published elsewhere : n2850.4 ( smail et al . 2003 ) and n2850.1 ( chapman et al .
2002a ) from deep near - infrared spectroscopy with the subaru telescope , simpson et al .
( 2004 ) have measured spectroscopic redshifts based on two or more emission lines with secure radio and near - infrared identifications for 2 additional 8-mjy scuba sources ( n2850.2 and n2850.12 in the elais n2 field , scott et al .
we therefore include these two sources in the expanded sample analysed here .
we note that simpson et al . also suggested a tentative redshift for le850.3 estimated from a single 2.5@xmath13 emission - line , and also a redshift for cudss14.9 from the putative hk ca absorption lines that need confirmation ( simpson , priv .
communication ) .
these latter two redshifts are not considered robust enough for inclusion in our analysis .
we thus have 7 new redshifts for robust sub - mm sources from the uk 8-mjy scuba survey , which can be added to the heterogenous collection of 8 sub - mm sources with spectroscopic redshifts previously considered by aretxaga et al .
thus , the new spectroscopic data have effectively doubled the size of the comparison sample , as well as extending the redshift baseline out to @xmath19 . [ cols="<,^,^,^,^,^,<",options="header " , ] + since our original analysis in paper ii , new photometric data at 1.2 mm obtained with mambo ( greve et al .
2004 ) have also become available for the sub - mm galaxies first identified in the uk 8-mjy scuba survey areas ( scott et al .
we have therefore incorporated this additional photometric information in a re - derivation of the photometric redshifts of uk 8-mjy scuba sources in the current spectroscopic sample .
we also include n2850.12 which has a 3.4@xmath13 flux determination at 850@xmath6 m , and hence failed our selection criteria in paper ii , but which now has a spectroscopic - redshift ( simpson et al .
2004 ) based on an ir counterpart associated with the 3@xmath13 peak in the radio observations of ivison et al .
( 2002 ) .
we have adopted the fiducial evolutionary model le2 ( the least restrictive one in paper ii , which applied a non - informative prior at @xmath20 , and @xmath21 luminosity evolution to @xmath22 ) to derive the new photometric - redshift estimates which are given in table 1 .
we produce simulated catalogs of sources in mock surveys to the same depth as those in which the real sub - mm galaxies were detected .
the corresponding redshift distributions are then derived by the joint probability of identifying the particular radio mm fir colours and fluxes of the real sources with the mock galaxies in the simulated catalogs . in order to check the stability of the redshift solutions found ,
we have calculated 100 different monte carlos for each source , and for each of these realizations we have derived its redshift probability distribution , its mode and 68% and 90% confidence level intervals . as expected , the sources with photometric redshifts based on 2 or more colours are the most stable and show well defined peaks and little variation in the mode of the redshift distribution of the different simulations .
it is important to note that these photometric redshift estimates include absolute calibration errors of 5 - 20% in the individual radio submillimetre fluxes , and assume no correlation between the luminosity and shape of an sed ( or temperature ) . from our sample of 15 sub - mm galaxies , fig .
1 shows the photometric - redshift spectroscopic - redshift regression plot for those 12 sources that have at least one robust colour based on two or more detections ( @xmath23 ) in the radio - mm - fir regime .
the remaining 3 sources ( n2850.8 , n2850.12 , and w - mm11 ) have a robust detection at only one wavelength and various upper - limits and hence are not included in fig .
1 . figs .
2 and 3 show the comparison of the observed seds of the sources analysed in this paper and the template seds considered in the photometric redshift analysis , illustrating the cases where there is good agreement and catastrophic disagreement , respectively .
all seds in our template library are accepted by more than one of these sources . among the photometric redshifts based on at least two colours ( with well measured fluxes , i.e. @xmath23 , in three or more bands )
the submillimetre galaxy which departs most clearly from the @xmath24 line is n2850.1 ( @xmath25 ) .
our new photometric redshift of n2850.1 , using the most recent upper limit at 1.2 mm ( greve et al .
2004 ) together with the 450 , 850@xmath6 m and 1.4ghz fluxes , is @xmath26 at a 68% confidence level , and @xmath27 at a 90% level .
these values are consistent with the estimates presented in paper ii , but remain strongly inconsistent with the optical spectroscopic redshift ( @xmath28 ) reported by ( chapman et al .
this is not surprising as there has already been considerable debate over whether the optical identification for this sub - mm source is correct . the optical spectrum and redshift for this counterpart , originally published by chapman et al .
( 2002a ) , led these authors to argue that the optically - bright galaxy coincident ( within @xmath29 ) with the radio position of n2850.1 was possibly a foreground galaxy that lenses the sub - mm source .
this argument was based primarily on the fact that the temperature of the dust emission @xmath30 k deduced for the sub - mm source at the optical spectroscopic redshift was 4@xmath13 below that of the local dusty galaxies with the same intrinsic luminosity . in chapman
( 2003a ) , n2850.1 has a revised temperature of @xmath31 k or 6@xmath13 below the temperature distribution of local analogs , and colder than smm22173 + 0014 , which chapman et al .
( 2002a ) also claim is lensed .
it is clear therefore that among the sources considered here , n2850.1 is the most likely example of a bright sub - mm source produced by gravitational lensing by an intermediate redshift galaxy , analogous to the case of hdf850.1 studied in detail by dunlop et al .
proving this beyond doubt remains a challenge , as astrometric information of the quality available to dunlop et al .
( 2004 ) does not yet exist in the case of n2850.1 .
whatever the correct explanation for its properties , it is obvious that n2850.1 is an unusual source , being the only object in the sample of 15 galaxies which can not be fit by any of our sed templates when redshifted to @xmath32 ( see fig . 3 ) . for the sake of transparency and completeness in what follows we therefore quote the overall accuracy of the redshift estimation procedure both with and without the inclusion of n2850.1 in the statistical calculations . in general the agreement between photometric and spectroscopic redshifts is encouraging .
the three sources for which photometric - redshifts are based on a solid detection ( @xmath23 ) at just one wavelength , however , are the least precise due to insufficient photometric constraints , although we note that they are still formally consistent with the spectroscopic redshifts found : n2850.8 at @xmath33 has a photometric redshift @xmath34 at a 68% confidence level ( 1.04.5 at the 90% level ) ; n2850.12 has a photometric redshift of @xmath35 at a 68% confidence level ( 1.5 to 5.0 at the 90% level ) , which is consistent with the @xmath36 ( simpson et al .
2004 ) ; and w - mm11 at @xmath37 , as determined in paper ii , has a @xmath38 at a 68% confidence level ( 2.05.8 at the 90% level ) . despite the strong suspicion that n2850.1 is a lensed scuba galaxy at a redshift @xmath39 , even if we include n2850.1 in our analysis the rms dispersion about @xmath40 is @xmath41 .
this result considers only those 10/15 sub - mm galaxies with at least two measured colours based on 3 or more @xmath23 detections in the radio mm fir regime .
the precision significantly improves to @xmath42 if we exclude n2850.1 .
extending this analysis to the 12/15 sources with at least one colour determined from detections at two or more bands , the measured mean accuracy of the photometric redshifts in the range @xmath43 is @xmath44 and @xmath45 , including and excluding n2850.1 , respectively . finally if we also include sources with only upper - limits in the colours ( for example a single submillimetre detection with a non - detection at radio wavelengths ) , we measure @xmath46 and @xmath47 , including and excluding n2850.1 , respectively .
blain et al . ( 2003 ) have stated that the technique described in papers i and ii assumes a narrow - range of local sed templates , with a tight distribution of dust temperatures , and luminosities , to derive the photometric redshifts .
we emphasise again that the range of dust temperatures in our template library ranges from 25 65k , and that we include 20 local galaxies ( starbursts and agn ) with well - measured seds and fir luminosities spanning the range @xmath48 .
furthermore , it is misleading to suggest that a wide range of dust temperatures ( that may be correlated with luminosity ) should translate directly into a similar redshift uncertainty in our calculations , since the redshift distributions are dominated by those seds that best fit the radio mm fir data .
although all seds contribute to the redshift distributions at some level , but with varying degrees of significance , a well - defined peak can be measured .
the comparison presented here provides reassurance that , by allowing the variety of local template seds to be selected at random , and then scaled to the required fir luminosity to populate the evolving luminosity function , we have offered the photometric - redshift method a library of galaxies with a sufficiently broad range of dust - temperatures , sed shapes and levels of star formation and agn activity from which to find a solution .
we also note that the sensitivities of the current submillimetre experiments in blank - field surveys select only those starburst galaxies with @xmath49 .
this being the case , then perhaps the future choice of seds should be restricted to those galaxies more luminous than this sensitivity limit , in which case our library of seds will only be lacking the highest - luminosity local counterparts ( @xmath50 ) .
if we consider that there exists a luminosity temperature dependence in the seds of starbursts , then we will be missing the seds that peak at the shortest wavelengths in our library , and thus we will be underestimating the redshifts for some fraction of scuba sources .
an encouraging aspect of our study is that the derived errors in our calculations of the individual photometric redshifts ( i.e. the 68% confidence levels , shown as dotted lines in fig . 1 and in column 3 of table 1 ) , are systematically larger than the measured dispersion around the @xmath40 line .
this dispersion is based on the differences between the modes of the photometric redshift distributions and the spectroscopic redshifts .
for instance , the average 68% confidence interval is @xmath51 for those sources with two or more colours derived from three or more detections ( black diamonds in fig . 1 ) , while the measured accuracy of their photometric redshifts is @xmath52 and @xmath53 , including and excluding n2850.1 .
this suggests that our errors are overestimated , possibly due to considering a range of sed templates wider than the true , but still unknown , distribution of sed shapes of sub - mm galaxies .
we note , however , that , with the currently available photometry , all the seds in our template library are accepted by more than one sub - mm galaxy as possible counterparts ( fig . 2 ) . to conclude this paper , and to provide further justification for the results presented in 2.2 , we illustrate in fig.[evol_photz ] how the accuracy of the photometric - redshift method evolves and improves as we include an increasing number of robust photometric measurements
when based on a single detection at 850@xmath6 m ( fig.4a ) the initial photometric - redshift estimates reflect only the prior assumption for the luminosity evolution of the sub - mm population .
the addition of detections or deep limits at @xmath54 m , that sample the rest - frame fir sub - mm peak , place initial constraints on the photometric redshifts ( fig.4b ) .
the 850@xmath6m radio spectral index has been shown to also provide a useful one - colour measure of photometric - redshifts for the sub - mm population ( e.g. carilli & yun 2000 ) , with the ability to robustly identify galaxies at @xmath55 from those at lower - redshifts . hence the use of data in the radio regime ( 1.48 ghz ) in this analysis continues to improve the photometric - redshift accuracy ( fig.4c ) by providing confirmation , and therefore adding more weight to the results determined from the data at @xmath56 m .
finally , with the addition of further mm - wavelength data at 12 mm , we eventually derive a photometric - redshift accuracy of @xmath57 or @xmath58 for all 12 galaxies with at least one well - measured colour if we include or reject n2850.1 .
respectively ( fig.4d ) .
= 8.0 cm + = 8.0 cm + = 8.0 cm + = 8.0 cm +
we have complemented our previous comparison of photometric - redshifts and spectroscopic redshifts for 8 submillimetre galaxies ( aretxaga et al .
2003 ) with 7 new spectroscopic redshifts ( chapman et al .
2003a , simpson et al . 2004 ) and recently published mambo 1.2 mm data ( greve et al .
the increased number of available spectroscopic redshifts for submillimetre galaxies with sufficient accompanying rest - frame radio to fir photometry confirms the reliability of our previous simulations and predictions for the redshifts of sub - mm sources .
this accuracy is all the more impressive given that the photometric - redshifts for 6 of new sources were effectively predicted before the additional 7 spectroscopic redshifts were available ( paper ii ) .
if we consider all available data for those sub - mm galaxies that have at least one colour determination based on two detections , we conclude that the rest - frame radio - fir photometric method can provide redshifts with an accuracy of @xmath59 over the redshift interval @xmath60 .
this redshift uncertainty increases to @xmath57 if the true spectroscopic - redshift of the brightest sub - mm source n2850.1 in the northern elais2 field is @xmath61 ( chapman et al .
there is , however , a more natural explanation for the discrepancy between the photometric - redshift and spectroscopic - redshift of n2850.1 , namely that the sub - mm source is lensed by the foreground optical counterpart ( as already suggested by chapman et al .
2002a ) , analogous to the situation for hdf850.1 , the brightest object in the hubble deep field ( dunlop et al . 2004 ) .
the photometric method currently relies on limited data , restricted to a few low s / n detections at observed radio , millimetre and submillimetre wavelengths .
no optimization of the current method has been made , based on any prior knowledge of those local sed templates that are consistent with the rest - frame data . despite the lack of any assumption in our simulations and analysis about a dependence between the shape of a local sed and luminosity of the redshifted submillimetre source , the method clearly works to a useful accuracy ( e.g. van kampen et al . 2004 ) .
the inclusion of such a dependence and measurements with greater s / n can only reduce the dispersion of possible seds that are consistent with the observational data , and hence this will reduce the width of the redshift probability distributions for individual targets .
one of the difficulties in this analysis has been to select from the limited publically - available information which sub - mm galaxies have secure spectroscopic redshifts derived from unambiguous optical or ir counterparts .
it is therefore encouraging to look forward to the next few years as the essential radio , ( sub)millimetre and fir photometric data become available from facilities such as the vla , gbt , lmt , apex , blast , herschel , spitzer and astro - f for complete and substantial samples of sub - mm sources .
thus we are confident that the combination of these high s / n multi - wavelength rest - frame radio fir data will generate photometric - redshifts with accuracies of @xmath62 for the majority of the individual sub - mm selected galaxies . given this , it will be possible to accurately measure the entire redshift distribution and star formation history of the high-@xmath63 population of heavily - obscured starburst galaxies without having to measure a spectroscopic redshift for every sub - mm source .
ia and dhh gratefully acknowledge support from conacyt grants 39548-f and 39953-f . aretxaga i. et al . , 2003 ,
mnras , 342 , 759 ( paperii ) barger a.j .
et al . , 1999 ,
aj , 117 , 2656 blain a.w . ,
barnard v.e . ,
chapman s.c . , 2003 , mnras , 338 , 733 carilli c.l .
, yun m.s . , 1999 , ap.j . , 513 , l13 carilli c.l . , yun m.s . ,
2000 , ap.j . , 530 , 618 chapman s.c . ,
smail i.r . ,
ivison r.j . ,
blain a.w . , 2002a , mnras , 335 , l17 .
chapman s.c .
et al . , 2002b , apj , 572 , l1 .
chapman s.c .
et al . , 2002c , apj , 573 , 66 .
chapman s.c . ,
blain a.w .
, ivison r.j . ,
smail i.r . , 2003a , nature , 422 , 695 chapman s.c .
et al . , 2003b , apj , 585 , 57 devlin m. et al . , 2001 , in lowenthal j. , hughes d.h .
, eds . , deep millimetre surveys : implications for galaxy formation and evolution , world scientific , p. 59 .
dunlop j.s .
et al . , 2004 ,
mnras , 350 , 769 dunne l. , clements d.l .
, eales s.a
. , 2000 , mnras , 319 , 813 eales s. et al . , 2000 , aj , 120 , 2244 . frayer d.t .
et al . , 1998 ,
apj , 506 , 7 greve t.r .
et al . , 2004 , mnras , submitted ( astro - ph/0405361 ) helou g. , soifer b.t . , rowan - robinson m. , 1985 , apjl , 289 , 7 hu e.m . , ridgway s.e . , 1994 ,
aj , 107 , 1303 hughes d.h .
et al . , 1998 ,
nat , 394 , 241 hughes d.h .
. , 2002 , mnras , 335 , 871 ( paperi ) isobe t. , feigelson e.d . ,
nelson p.i . , 1986 ,
apj , 306 , 490 ivison r.j . , et al . , 2002 , mnras , 337 , 1 mortier , a. et al . ,
2004 , mnras , submitted rengarajan t.n .
, takeuchi t.t . , 2001 , pasj , 53 , 433 richards e.a . ,
2000 , apj , 533 , 611 simpson c. , et al . , 2004 , mnras , 353 , 179 .
smail i. et al . , 2003 ,
mnras , 342 , 1185 van kampen et al .
2004 , mnras , submitted ( astro - ph/0408552 ) wiklind t. , 2003 , apj , 588 , 736 yun m.s .
, carilli c.l . , 2002 ,
apj , 568 , 88 .
scott s. , et al .
, 2002 , mnras , 331 , 817 . | we present a comparison between the published optical , ir and co spectroscopic redshifts of 15 ( sub-)mm galaxies and their photometric redshifts as derived from long - wavelength ( radio mm fir ) photometric data .
the redshift accuracy measured for 12 sub - mm galaxies with at least one robustly - determined colour in the radio mm fir regime is @xmath0 ( r.m.s . ) . despite the wide range of spectral energy distributions in the local galaxies that are used in an un - biased manner as templates , this analysis demonstrates that photometric redshifts can be efficiently derived for sub - mm galaxies with a precision of @xmath1 using only the rest - frame fir to radio wavelength data .
= = = = = = = = # 1 # 1 # 1 # 1 @mathgroup@group @mathgroup@normal@groupeurmn @mathgroup@bold@groupeurbn @mathgroup@group @mathgroup@normal@groupmsamn @mathgroup@bold@groupmsamn = `` 019 = ' ' 016 = `` 040 = ' ' 336 = " 33e = = = = = = = = # 1 # 1 # 1 # 1 = = = = = = = = epsf [ firstpage ] surveys galaxies : evolution cosmology : miscellaneous infrared : galaxies submillimetre |
in the past decades , the vortex state of superconducting networks has been investigated by several groups using different imaging techniques ( scanning hall microscopy @xcite , scanning squid microscopy@xcite or bitter decoration @xcite ) .
the vortex configuration was studied in square or triangular lattices as a function of the magnetic field . in superconducting arrays
the relevant variable is the magnetic frustration , @xmath2 , which represents the vortex filling factor .
@xmath2 is defined as @xmath0 with @xmath3 , the flux quantum , and @xmath4 , the magnetic flux per elementary plaquette .
the vortex pattern reflects the spatial phase configuration of the superconducting order parameter resulting from the competition between the magnetic field and the underlying lattice @xcite . for rational frustration @xmath5 , with @xmath6 and @xmath7 integer numbers ,
a commensurate ground state is generally expected as for example the checkerboard state which was imaged in square lattices at @xmath1 @xcite .
the purpose of this letter is to present the detailed results of a magnetic decoration experiment in a fully frustrated dice network .
the unusual properties of the dice lattice were recently highlighted by the discovery of a peculiar destructive interference phenomenon occurring at @xmath1 for the electronic wave function in the tight - binding model@xcite .
this phenomenon was shown later to manifest itself as a broadening of the superconducting transition and a suppression of the critical current in the corresponding superconducting wire network@xcite .
this observation was indicative of the absence of commensurate vortex state in the dice network , in contrast to the square network which was shown to exhibit a sharp critical current peak at @xmath1@xcite .
we report here new careful decoration experiments on a series of high quality extended niobium dice networks under different magnetic fields .
the analysis of the vortex correlation functions shows unambiguously that the vortex state is fully disordered at @xmath1 and confirms the results suggested by our preliminary experiments on small networks @xcite .
the issue of the vortex configuration is of additional interest in the dice lattice as the vortices are located on the ( dual ) kagom lattice ( fig .
[ kagome ] ) .
since the vortex variable is binary , for @xmath2 smaller than 1 ( in each cell there is either one vortex ( @xmath8 ) or no vortex ( @xmath9 ) ) , we can view this problem as ising spins on the nodes of a kagom lattice .
this general problem has been studied theoretically@xcite with first and second neighbour interaction .
our vortex experiment is the first observation of binary variables on a the kagom lattice .
the imaging was performed by the bitter magnetic decoration technique @xcite .
since the magnetic contrast of vortices in superconducting wire networks@xcite is very weak , we used the so - called flux compression technique@xcite .
the networks ( fig .
[ pattern ] ) were made of niobium wires ( @xmath10 m long , @xmath11 nm wide ) patterned from a @xmath12 nm thick nb layer epitaxially grown on a sapphire substrate at 550@xmath13c .
the patterning was achieved by reactive ion etching in @xmath14 through an aluminum mask prepared by _ e - beam _ lithography@xcite .
only the first @xmath15 nm nb was removed , leaving unpatterned a @xmath16 nm thick niobium layer as a flux compression bottom layer .
the total array size was @xmath17 m long and @xmath18 m wide .
the networks contained about @xmath19 cells , with a wire width homogeneity better than @xmath20 .
the elementary cell area in all networks is @xmath21m@xmath22 and the matching field ( @xmath23 ) was obtained for @xmath24 mt .
the decoration cell was placed inside a double @xmath25-metal shield . before each
run the magnetic field was calibrated using the magnetoresistance curve of a large sns josephson junction array .
the flux accuracy is a few @xmath26 per cell .
we always found the counted vortex density perfectly consistent with our field calibration .
we first apply the magnetic field in the normal state at @xmath27 k. then the sample is slowly cooled down across the niobium transition temperature .
vortices nucleate at the niobium superconducting transition temperature , @xmath28 k for our samples , and , because of the strong network pinning , their configuration freezes out as the temperature decreases a few mk below @xmath29@xcite .
flux compression takes place at the superconducting transition ( @xmath30 k ) of the bottom layer when the network vortex loops convert into abrikosov vortices .
once the temperature is stabilized at @xmath31 k , ni particles are flash - evaporated on the sample under a nominal residual helium pressure of @xmath32 mbar .
the vortex positions are then registrated at room temperature using a scanning electron microscope ( cf fig .
[ pattern ] ) .
this `` one - shot '' technique allows us to image a wide region and collect statistical information on the position of vortices .
in this section we compare the vortex structures at @xmath33 and at @xmath1 .
the datas are extracted from the sem - micrographs and the vortex configuration is then regenerated on a computer . the pictures on fig .
[ configtiers ] and [ configdemi ] are parts of the complete images .
since the bravais cell is composed of three different plaquettes , we represent the vortices with three different colors ( blue , green , and red ) depending on where they are located . to obtain a quantitative information on the degree of disorder
, we have calculated linear correlation functions between vortex variables in the three equivalent directions of the lattice ( see fig .
[ kagome ] ) . by making distinction between thoses three directions , we keep information on the domain shapes , which would be lost on averaging over the three contributions .
the vortex variables @xmath34 are equal to @xmath8 if a vortex is present in the @xmath35 cell and @xmath9 if not .
the three correlation functions , @xmath36 , @xmath37 and @xmath38 are defined as : @xmath39 where the @xmath40 cell is the @xmath41-th neighbour of the @xmath35 cell in the @xmath42 , respectively @xmath43 or @xmath44 , direction ( see fig . [ kagome ] ) and @xmath45 is the mean product @xmath46 for each @xmath35 having at least one @xmath41-th neighbour in the @xmath42 ( respectively @xmath43 or @xmath44 ) direction .
we first consider filling factor @xmath47 where a stable commensurate state is expected .
indeed one can build a simple ordered configuration with 1 vortex every @xmath48 cells , i.e. one vortex per bravais cell .
this ordered phase is three times degenerated due to the lattice symmetry . on fig .
[ configtiers ] , we show two of those phases observed on two different runs performed at magnetic field @xmath49 mt . we have analyzed each time a zone of approximately @xmath50 cells at the network s center .
the actual frustration , as determined from the observed number of vortices , is for the first run : @xmath51 ( fig .
[ configtiers ] left ) .
we observe a huge single domain extending over the whole picture .
for the second run shown in fig .
[ configtiers ] right , the actual frustration is @xmath52 .
the observed effective domain size is approximately 10 cells .
excess vortices seem to gather near the domains walls , as can also be noticed in the former case ( added vortices sit preferentially on the defect lines ) .
this behavior is similar to the one observed in square arrays @xcite .
the correlation functions for @xmath53 are shown on fig .
[ corretiers ] .
as can be seen , the correlation falls from @xmath54 at @xmath55 to approximately @xmath56 at @xmath57 in the three directions and this level of correlation is conserved beyond the 40-th neighbour .
the extended ordered vortex state at @xmath33 , is similar to that observed in the square or triangular lattices with rationnal frustrations @xcite .
this structure remains ordered even within a few percent of excess vortices but the domain size decreases .
this indicates the robustness of the @xmath47-ordered state to extra vortices .
in contrast , at half filling the vortex lattice is incommensurate with the dice lattice .
one must consider two bravais cells .
we performed three different experiments at @xmath1 .
the number of investigated cells was approximately @xmath58 for each experiment .
the measured frustration was @xmath59 around the ideal half filling .
a part of two regenerate images is displayed on fig .
[ configdemi ] .
despite the fact that @xmath1 is a rational filling factor , we did not observe simple ordered phases in any of the three experiments .
rather we observe strongly disordered states reminiscent of the incommensurate states occuring at irrational frustrations @xcite .
the correlation functions ( see fig . [ corrdemi ] ) rapidly drops to approximately @xmath60 beyond @xmath61 .
this residual correlation is a real effect , larger than the experimental uncertainties .
it is due to differences in the probability of populating the three different sites in the bravais cell .
it corresponds for this decoration experiment to a population ratio of @xmath62 , @xmath63 and @xmath64 respectively for the different kind of sites .
this difference could be due to some slight inhomogeneity in the network temperature during cooling or in the wire geometry@xcite .
presently the origin of this small effect is not understood and remains an open problem . fitting the correlation function with an exponential function @xmath65
, we get a correlation length @xmath66 for @xmath1 .
we also imaged a vortex lattice at @xmath67 .
we saw no ordered domains , but we could not make definitive conclusion on the @xmath68-state with this experiment .
this small deviation from the exact @xmath69 indeed corresponds to @xmath70 of the vortex population and might change the configuration more efficiently than for a higher filling factor such as @xmath33 or @xmath1 . in the litterature ,
no observation of ordered vortex state was ever reported in superconducting networks for such small filling factors .
our observation of a very short correlation length at @xmath71 , is indicative of a strongly disordered state .
this observation is consistent with previous transport measurements in superconducting al networks with the same lattice@xcite .
a sharp peak was observed at @xmath33 , consistent with the strongly pinning commensurate state shown in fig .
[ configtiers ] . on the other hand
the suppression of critical current at @xmath1 was indicative of a weakening of the phase rigidity due to the localization of the superconducting wave function in the aharonov bohm cages@xcite .
the absence of ordering is reminiscent of the problem of classical ising spins ( zero or one vortex per cell ) on the nodes of the kagom lattice@xcite .
however , since the condition of nearest neighbour interaction which is central to the prediction of disordered state at all temperatures in the ising model , is presumably not fulfilled for array vortices one can not draw conclusions from this analogy .
very recent theoretical studies of vortices in dice josephson junctions arrays have more relevance as they directly address this issue of vortex ordering .
korshunov @xcite investigated the ground state of superconducting dice arrays with either cosine or quadratic dependence of the coupling energy upon the phase difference .
he found a ground state made of an ordered configuration of vortex triads ( clusters of three adjacent vortices ) . because of the proliferation of zero - energy domain walls , the ground state entropy is large but it is not extensive , in contrast to the case of ising spins on a kagome lattice@xcite where an extensive entropy was found .
therefore , in this model , the true ground state configuration for @xmath1 should be ordered .
in contrast , the observed vortex correlation functions in our decoration experiments do not reveal any signature of order .
also , the histogram of vortex cluster sizes shows no preferred occurrence of triads .
we do not believe that the geometrical irregularities present in experimental sample@xcite are responsible for the observed disordered state .
instead , we believe that it results from the combination of large thermal fluctuations and large entropy of low energy configurational states at the vortex freezing temperature@xcite . indeed , a large number of low energy vortex configurations are likely to exist in the dice lattice @xcite .
however , their difference in energy is much less than the energy barrier for vortex motion which , in the wire network@xcite is mostly determined by the single wire length , independently of the network topology . on cooling down the vortex lattice
may therefore be freezed out in a configuration different from the true ground state . a possible origin of the lack of ordering of the vortex lattice
was recently pointed out by v. cataudella and r. fazio@xcite .
their monte carlo simulations on the fully frustrated josephson junction dice array suggest the existence of a low temperature phase transition below which a glassy dynamics prevents the system to reach the true ground state conjectured in ref@xcite .
this scenario is consistent with our observation but can not be tested in our strong vortex pinning wire network array .
we performed magnetic imaging of the vortex configuration in nb dice networks under several magnetic fields .
we focused our study on magnetic frustrations @xmath33 and @xmath1 .
numerical calculations of the correlations on the observed vortex positions lead to a quantitative characterization of the degree of disorder .
while a robust commensurate state is found at @xmath33 , we observe a strongly disordered vortex configuration at @xmath1 characterized by a vortex correlation length close to one cell size .
it is worth noticing that this feature , particular to the dice lattice , appears here in a system with no geometrical disorder .
the discrepancy with the korshunov prediction that the ground state is ordered suggests that the true ground state is not accessible in our experiment .
further investigation of josephson junction arrays on a dice lattice by transport and high sensitive magnetic microscopy@xcite at much lower temperature are needed to reveal the true nature of the ground state .
we acknowledge fruitful discussions with c.c .
abilio , b. canals , b. douot , s.e .
korshunov , m. feigelman , l. ioffe , p. martinoli , r. mosseri , r. fazio and j. vidal .
discussions within the esf `` vortex matter program '' and the tmr network no .
fmrx - ct97 - 0143 are gratefully acknowledged .
the _ e - beam _ lithography and the sem observations were carried out with cea - leti - plato organization teams and tools . vortex pinning is strong in wire arrays@xcite due to the large energy barrier for vortex crossing a superconduction wire .
this barrier is suppressed in the vicinity of @xmath29 when the superconducting coherence length which defines the vortex core size becomes of order of the wire length .
the condition @xmath72 gives an indication of the onset of pinning by the wire array and determines the freezing temperature of the vortex lattice .
it is reached at @xmath73 mk below @xmath29 .
the _ e - beam _ lithography process may lead to some defects which are repeted in each cell of the network .
for example we can not discard a few percent difference in the width of wires with different angles .
the difference in the probability of populating the three different sites in the bravais cell might be due to such systematic defects . | we have investigated the vortex state in a superconducting dice network using the bitter decoration technique at several magnetic frustrations @xmath0=1/2 and 1/3 .
in contrast to other regular network geometries where the existence of a commensurate state was previouly demonstrated , no ordered state was observed in the dice network at @xmath1 and the observed vortex - vortex correlation length is close to one lattice cell . |
semiconductor nanostructures ( nss ) exhibit increased oscillator strength due to electron hole wave function overlap , and band gap engineering due to the effect of quantum confinement ( qc ) .
thus , materials like si are a viable option for opto - electronics , photonics , and quantum computing.@xcite qc is defined as the modification in the free particle dispersion relation as a function of a system s spatial dimension.@xcite if a free electron is confined within a potential barrier , a shift in the band gap energy is observed , which is inversely proportional to the system size squared , in the effective mass approximation . as a result ,
the emitted photon energy is directly proportional to the gap energy ( @xmath0 ) .
qc often manifests itself in optical experiments when the dimension of the system is systematically reduced and an increase in the absorbed / emitted photon energy is measured corresponding to electron transitional states . for practical applications , utilizing qc effects in nss
requires an understanding of the band structure of a low - dimensional material , how the method of preparation effects the final properties of the ns , and the kinetics/ dynamics of the absorption / emission process .
the confinement potential is determined by the alignment of the respective fermi levels when a material of a @xmath1 is surrounded by a material of a @xmath2 , with @xmath3.@xcite the preparation technique can introduce stress in the system , which changes the band gap energy.@xcite for indirect gap materials phonon processes can effect the recombination mechanism.@xcite the lifetime associated with the recombination event can be altered by the excitation power.@xcite ( for a review of general properties of low - dimensional structures , see refs . .
for a discussion of other higher order effects in nss , see ref . .
) this article is concerned with the electron / hole recombination process in amorphous ( a ) versus crystalline ( c ) nss with different dimensions .
several theoretical models ( e.g. see refs . ) have been applied to ns ; all models are empirical and no one model can model all semiconductor nss . since the parameters of a ns system are dependent upon the preparation method for a particular material , a comprehensive theoretical understanding must test along this dimension as well . in this article
, we consider a relatively simple model of direct e - h recombination using a ` particle in a box ' type model as a perturbation to the effective mass theory .
we use no adjustable parameters and include corrections to the model dependent on the preparation method as known experimentally and/or computationally when needed , thus achieving transparency in the physics involved .
the only parameter tested in this work is the crystallinity , which is shown to effect the strength of confinement ( defined in sec .
[ theory ] ) , because of the different symmetry properties of the electron and hole .
the model is applied to experimental results on crystalline and amorphous si and ge nss , including quantum wells ( qws ) , wires ( q - wires ) and dots ( qds ) . systems of regular shape are chosen to ensure crystallinity is the primary parameter .
for example , data obtained by van buuren et al .
@xcite for high quality ` star - shaped ' samples are difficult to analyse theoretically .
parameters relevant to a particular system are discussed and energy corrections are given when needed .
results are discussed and a mechanism for the differences between the strength of confinement in the amorphous and crystalline system is proposed .
central to the discussion of the confinement strength in nss is how the effective mass ( em ) changes as a function of dimension .
recently , cosentino et al.@xcite demonstrated an increase in the confinement strength due to a reduced em .
we developed a model for the em as a function of dimension and considered its effect in a qw .
modifications to the em parameter include the use of a spatially dependent effective mass ( spdem ) introduced via the von roos hamiltonian @xcite .
this form of the em is appropriate for doped semiconductors , or when there exists a graded potential @xcite . at the interface of a ns where the crystal potential will change , possibly abruptly ,
the em will change .
the influence of the interface on the em is expressed through the bastard type boundary conditions ( b.cs ) @xcite .
proper treatment of the em in a ns is an unresolved problem .
the em is important for theoretical models @xcite , and related to the hopping parameter and carrier mobility in the tight - binding model @xcite .
there is experimental @xcite and theoretical @xcite evidence that the em should depend on ns dimension .
fundamentally , since qc causes an increase in @xmath4-space @xcite , one does expect an increase in the bulk em ( @xmath5 ) .
furthermore , a change in the @xmath6 will modify the bohr radius , therefore , altering the regime in which qc effects can be observed .
the challenge is that experimental measurements of the @xmath6 in a ns are model dependent @xcite , and it is difficult to theoretically scale the em in a ns @xcite .
however , the em provides a natural framework to incorporate the influence of a modified crystal potential due to the interface , which is not adequately accounted for in theoretical models @xcite .
recently , costa filho et al . addressed the problem of a spdem by introducing a characteristic inverse length scale into the translation operator @xcite , given by : @xmath7 @xmath8 defines the inverse length scale and mixes the displacement , @xmath9 , of a carrier particle with the original position , @xmath10 .
note that , this formalism is equivalent to the @xmath11-exponential formalism in tsallis nonextensive thermostatistics @xcite and results in a deformed or contracted space .
tsallis thermostatistics describes a meta - equilibrium system , which provides a natural connection to confined carrier particles in an excited state . from eq .
a modified momentum operator was derived : @xmath12 along with a definition for the spdem @xcite : @xmath13 furthermore , it was shown that a point canonical transformation ( pct ) of a hamiltonian defined with @xmath14 in the kinetic term and with a harmonic potential energy yields an effective hamiltonian for a particle with constant mass in a morse potential @xcite , in perfect analogy with ref . .
therefore , eq . provides a natural framework to describe the anharmonic influence of the potential barrier on the carrier particles in a ns . in this report
, we consider a spdem given by eq . and discuss the physical implications of this term on the confinement energy for a qw .
note that the discussion , presented here generalizes to the case of a quantum dot and wire .
the bohr radius of an electron ( e ) , hole ( h ) or exciton ( x ) is given by , in si units : @xmath15 @xmath16 is the effective mass of the e , h or x , respectively , @xmath17 is the electric charge and @xmath18 is the dielectric constant .
depending on the e or h effective mass , the x - bohr radius is 4.5 nm for si and 24 nm for ge .
the bohr radius defines the spatial dimension of the particles , which determines the range of sizes for which qc can be observed .
we define three regimes of confinement here:@xcite * weak confinement : when the dimension of the system is much larger than @xmath19 and @xmath20 .
in this situation , the appropriate mass in the kinetic term is @xmath21 .
the energy term is dominated by the coulomb energy . * medium confinement : when the dimension of the system is much smaller than @xmath19 , but larger than @xmath20 , then only electrons will experience confinement .
the relevant mass is simply @xmath22 for the kinetic term .
most materials belong to this class . *
strong confinement : when the dimension of the system is much smaller than @xmath19 and @xmath20 . here
both electrons and holes experience confinement and the relevant mass is the reduced mass , @xmath23 , with @xmath24 . in this regime ,
the coulomb term is small and can generally be treated as a perturbation .
below we will use the terms ` weak , ' ` medium ' and ` strong ' to refer to the different regimes of confinement discussed above .
in the ` particle - in - a - box ' model the bulk @xmath0 is taken as the ground state energy .
the effect of reduced dimension is considered as a perturbation to the bulk @xmath0 within the em approximation .
we consider an infinite confinement model and use the bulk em ( the details of the model are in refs . ) .
the expression for the variation of @xmath0 is given by : @xmath25 @xmath26 is the band gap of the bulk material and @xmath27 represents the qd diameter , the qw thickness or the q - wire diameter in what follows .
the calculation was carried out for confinement in 1d , 2d with cylindrical coordinates and 3d with spherical coordinates .
the parameter @xmath28 is given for si and ge in the strong , medium and weak confinement regimes in table [ tbl1 ] .
the change in energy of the conduction band minimum ( @xmath29e@xmath30 ) due to qc is labelled as ` medium confinement ' in table [ tbl1 ] , because a @xmath29e@xmath30 is equivalent to qc of the electron only as defined by our model , where only the electron mass is considered in eq . .
the change in energy of the valence band maximum ( @xmath29e@xmath31 ) due to qc is also listed in table [ tbl1 ] , which is calculated by considering confinement of the hole only , where only the hole mass is considered in eq . .
the other fixed parameter is the appropriate @xmath32 of the bulk system .
.parameter @xmath28 given in eq . for 3d , 2d , 1d confinement and for @xmath29e@xmath30 , @xmath29e@xmath31 .
[ cols="^,^,^,^",options="header " , ]
we have studied the effect of confinement dimensions and crystallinity on the magnitude of the band gap expansion ( as a function of decreasing size ) in si and ge nss ( quantum wells ( qws ) , wires ( q - wires ) and dots ( qds ) . medium and strong confinement models provide the best fit to experimental results ; moreover crystalline materials exhibit medium confinement , while amorphous materials exhibit strong confinement regardless of the confinement dimensions of the system .
this difference in confinement strength was explained by considering the extent of spatial delocalization of the hole .
a possible explanation is hole pinning due to coupling with the vibronic states.@xcite it has previously been reported@xcite that band tail states become strongly delocalized in the amorphous system compared to the crystalline system .
this hole delocalization would partially account for the trends observed in our work
. a lower value of the effective mass is reported for the amorphous system , which accounts for the trends observed in our work , while the hole mass increases and the electron mass decreases as a function of spatial confinement.@xcite we have studied the effect of a spdem with one singularity in a two band em model for qc by introducing a phenomenological model with an inverse characteristic length related to the gaussian width parameter .
this model for the em accounts for electron coupling with the interface and allows for tunnelling effects to be included in the confinement energy . as the thickness of a qw
is reduced , there is increased electron coupling with the interface , which increases the tunnelling probability .
furthermore , we find these results to be a promising first step toward the development of better models of the effective mass parameter in low - dimensional theories .
more work needs to be done to test the results of eq . within more sophisticated theories , such as , @xmath33 , tight - binding , or pseudo - potential methods | we look at the relationship between the preparation method of si and ge nanostructures ( nss ) and the structural , electronic , and optical properties in terms of quantum confinement ( qc ) .
qc in nss causes a blue shift of the gap energy with decreasing ns dimension . directly measuring the effect of qc
is complicated by additional parameters , such as stress , interface and defect states .
in addition , differences in ns preparation lead to differences in the relevant parameter set . a relatively simple model of qc , using a ` particle - in - a - box'-type perturbation to the effective mass theory , was applied to si and ge quantum wells , wires and dots across a variety of preparation methods .
the choice of the model was made in order to distinguish contributions that are solely due to the effects of qc , where the only varied experimental parameter was the crystallinity .
it was found that the hole becomes de - localized in the case of amorphous materials , which leads to stronger confinement effects .
the origin of this result was partly attributed to differences in the effective mass between the amorphous and crystalline ns as well as between the electron and hole .
corrections to our qc model take into account a position dependent effective mass .
this term includes an inverse length scale dependent on the displacement from the origin .
thus , when the debroglie wavelength or the bohr radius of the carriers is on the order of the dimension of the ns the carriers ` feel ' the confinement potential altering their effective mass .
furthermore , it was found that certain interface states ( si - o - si ) act to pin the hole state , thus reducing the oscillator strength . |
a smooth , geometrically finite hyperbolic surface @xmath0 has finite genus , with ends consist of a finite number of hyperbolic funnels and/or cusps .
we ll assume that @xmath1 has infinite area , so there is at least one funnel .
under this assumption , the ( positive ) laplacian @xmath2 has absolutely continuous spectrum @xmath3 and finitely many discrete eigenvalues in @xmath4 , with no embedded eigenvalues .
the resolvent @xmath5 is well - defined for @xmath6 , as long as @xmath7 is not an @xmath8 eigenvalue of @xmath2 . by mazzeo - melrose @xcite and guillop - zworski @xcite
, @xmath9 admits a meromorphic continuation to @xmath10 , with poles of finite rank .
the _ resonances _ of @xmath1 are defined as the poles of this continuation .
we let @xmath11 denote the set of resonances repeated according to the multiplicity given by the rank of the residue of @xmath9 at the pole .
cc resplot101010.pdf ( 25,-3)@xmath12 & resplot101.pdf ( 25,-3)@xmath13 associated to the surface ( or the group ) is the selberg zeta function , @xmath14 , defined as a product over the primitive length spectrum @xmath15 , @xmath16 converging uniformly on compact sets for @xmath17 .
the zeta function admits meromorphic continuation to @xmath18 for any geometrically finite hyperbolic surface .
this follows from the selberg trace formula if the area is finite , and was extended to the general case by guillop @xcite . by patterson - perry @xcite and borthwick - judge - perry @xcite , the divisor of @xmath14
consists of zeros at the resonance set @xmath11 , plus a set of `` topological '' zeros ( depending on the euler characteristic of @xmath1 ) and poles ( depending on the number of cusps . ) in particular , @xmath14 is entire if and only if @xmath1 has no cusps ( equivalent to @xmath19 being convex co - compact ) .
see @xcite for additional background on this link between the resonances and the zeta function .
figure [ resplot101.fig ] shows sample plots of the resonance sets of two hyperbolic surfaces , computed as zeros of the corresponding zeta function by the method described in [ zeta.sec][comp.sec ] .
the curves in the plots are made up of discrete points ; the spacing between resonances is small enough that individual resonances are not necessarily resolved at this scale .
the plots show only @xmath20 because the resonance sets are symmetric with respect to conjugation , by self - adjointness . in our definition of resonance
we use the parameter @xmath21 because of the connection to the selberg zeta function .
a more traditional choice from the point of view of scattering theory would be @xmath22 , which would flip the axes in the plots .
the theory of the zeta function for infinite - area surfaces is of course an extension of the much older selberg theory for compact or finite - area hyperbolic surfaces . in the finite - area case
, selberg @xcite showed that the resonances are confined to a vertical strip @xmath23 .
furthermore , the selberg trace formula leads a weyl law for resonances of the form , @xmath24 proven in this case by mller @xcite and parnovski @xcite .
another result of selberg @xcite shows that the resonances of a finite - area hyperbolic surface accumulate on the critical line @xmath25 , in the sense that for any @xmath26 , there are infinitely resonances in the strip @xmath27
. dyatlov ( * ? ? ?
* eq . ( 1.17 ) ) recently pointed out that under the assumption of finitely many embedded eigenvalues ( a condition which is known to hold generically ) , most resonances lie near the critical line in the sense that @xmath28 for any @xmath26 . by comparison , in the infinite - area case
a great deal is still unknown about the distribution of resonances .
for example , there is no weyl law analogous to .
guillop - zworski @xcite proved that for non - elementary hyperbolic surfaces ( as well as compactly supported perturbations ) , @xmath29 in @xcite , we showed that there exist constants for these upper and lower that depend only on @xmath30 and the sum of lengths of the geodesic funnel boundaries , with the upper constant being sharp in the sense that it reduces to in the finite - area case . the distribution of resonances for infinite - area surfaces has attracted renewed attention lately on the arithmetic side , in the work of bourgain - gamburd - sarnak @xcite .
infinite - area surfaces are not included in the classical selberg theory , since arithmetic fuchsian groups yield quotients of finite volume . however , the main point in @xcite was a generalization of selberg s famous @xmath31 theorem to infinite - area `` congruence '' surfaces .
this amounts to a uniform estimate on the spectral gap between the first ( rightmost ) resonance and the bulk of @xmath11 .
such estimates have applications to upper bounds for the affine linear sieve .
another source of recent interest in the distribution of resonances for infinite - area hyperbolic surfaces comes from the theory of chaotic scattering systems in physics , as exemplified by @xmath32-disk scattering systems . in the last decade
a number of interesting conjectures have arisen for such systems , involving connections between the distribution of resonances near the critical line ( @xmath25 in our case ) and chaotic dynamics of the associated classical system .
the conjectures have been supported by both numerical and experimental evidence , and also by mathematical theorems in some cases . with these recent developments as motivation ,
we present here a numerical investigation in the spirit of hejhal - rackner @xcite .
we will focus on some particular aspects of the distribution of resonances for hyperbolic surfaces : 1 . the asymptotic density of resonances near the critical line : the fractal weyl law .
2 . the spectral gap between the first resonance and the bulk of the resonance set .
3 . concentration of imaginary parts ( quantum decay rates ) at half of the classical escape rate .
we will give further details of the specific conjectures , as well as background and references , in the relevant sections
[ frweyl.sec][decay.sec ] below . for the finite - area case
, the conjectures all correspond to known theorems , by the results mentioned above .
the dimension @xmath33 of the limit set of @xmath19 plays a central role in our investigations .
patterson s famous result @xcite equated the dimension of the limit set with the exponent of convergence of @xmath19 , and showed that @xmath34 was the lowest eigenvalue of the laplacian if @xmath35 .
patterson extended this result to @xmath36 in @xcite , showing that @xmath33 is always the location of the first resonance .
in addition , the value of @xmath33 determines the dimension for the fractal weyl law and the classical escape rate , so it is the crucial parameter in all three conjectures mentioned above .
the basis for our computations is an algorithm for computing the selberg zeta function for a classical schottky group from jenkinson - pollicott @xcite . by a result of button @xcite , any convex co
- compact fuchsian group is representable as a classical schottky group , so in principle this technique would allow the approximation of resonances for any smooth hyperbolic surface without cusps . in practice , however , the number of computations required increases exponentially with the number of generators of the schottky group . to keep the computations manageable
we have confined our attention to groups with 2 generators , which limits us to surfaces with @xmath37 .
another limitation on the method is a severe degradation of the convergence rate for @xmath38 .
we are thus not able to investigate global asymptotics of the form .
fortunately , the resonances of greatest interest are those nearest the critical line , and the convergence rate is stable up to very large values of @xmath39 , provided we choose group parameters so that @xmath33 is relatively small . *
acknowledgment*. i would like to thank maciej zworski for encouragement to develop this project , and for helpful comments and suggestions along the way .
as noted in the introduction , for a geometrically finite hyperbolic surface @xmath40 the resolvent is written @xmath41 and admits a meromorphic continuation to @xmath18 .
the resonance set @xmath11 is then defined to be the set of poles of @xmath9 , with multiplicities given by @xmath42 the corresponding global counting function is @xmath43 the bound from guillop - zworski @xcite gives @xmath44 for more general surfaces with hyperbolic ends , with arbitrary metric inside a compact set , provided the @xmath45-volume ( a regularization of the infinite area ) is nonzero .
the bounds apply in particular to smooth geometrically finite hyperbolic surfaces with only one exception : the parabolic cylinder , @xmath46 , which has only a single resonance . in borthwick @xcite a sharp constant
was identified for the upper bound : @xmath47 where @xmath30 is the euler characteristic and @xmath48 are the boundary lengths of the hyperbolic funnel ends .
this upper bound implies that the lower bound could also be written with a constant that depends only on @xmath30 and @xmath48 .
no weyl law which would make these asymptotics more precise is known for infinite - area hyperbolic surfaces .
the exponent of convergence @xmath33 of a fuchsian group @xmath19 is defined by @xmath49 the limit set @xmath50 of a fuchsian group @xmath19 is the set of accumulation points of an orbit of @xmath19 , with respect to the poincar disk topology .
patterson @xcite and sullivan @xcite famously proved that @xmath33 is the hausdorff dimension of the limit set , @xmath51 the group @xmath19 is cofinite if and only if @xmath52 , so the patterson - sullivan theory is most interesting in the infinite - area case .
for the elementary cases , which consist of @xmath53 and the hyperbolic cylinders obtained when @xmath19 is cyclic ( assuming @xmath1 is smooth ) , we have @xmath54 . for a non - elementary surface @xmath55 . in his original paper , patterson
@xcite proved that if @xmath35 , then @xmath34 is the lowest eigenvalue of @xmath2 .
he extended this result in @xcite , proving that for all @xmath56 , @xmath9 has a simple pole at @xmath57 and no other poles with @xmath58 .
the same proofs give also the corresponding result that @xmath33 is the first zero of the zeta function @xmath14 , defined in .
patterson later expanded on this connection between the resonance set and the selberg zeta function , proving in @xcite a trace formula connecting the logarithmic derivative of @xmath14 to a regularized trace of the resolvent .
( the analogous formula in finite area follows directly from an application of the selberg trace formula , but the regularization problem is more serious in the infinite - area case . ) using this formula as a key ingredient , patterson - perry @xcite proved the following result for convex co - compact surfaces , which was extended to surfaces with cusps in borthwick - judge - perry @xcite : for @xmath40 geometrically finite , the zero set of the zeta function @xmath14 is the union of the resonance @xmath11 ( taken with multiplicities ) and the set of points @xmath59 , @xmath60 , with multiplicities @xmath61 .
if @xmath1 has @xmath62 cusps then @xmath14 has poles of order @xmath62 at @xmath63
. this connection forms the basis for all of the resonance calculations in this paper ; we will describe the methods used to calculating approximate zeros of the selberg zeta function in [ zeta.sec][comp.sec ] .
an argument of button @xcite shows that any convex co - compact fuchsian group is a classical schottky group .
this means that for any smooth , geometrically finite hyperbolic surface @xmath1 without cusps , there is a classical schottky group @xmath19 such that @xmath64 .
suppose the fuchsian group @xmath65 is a classical schottky group with @xmath66 generators .
such a group is defined by a collection of euclidean disks @xmath67 in @xmath68 with centers on the real axis and mutually disjoint closures . if @xmath69 denotes the element of @xmath70 that maps the exterior of @xmath71 onto the interior of @xmath72 , then @xmath19 is freely generated by the transformations @xmath73 .
we adopt the standard convention that @xmath69 is defined cyclically for @xmath74 , such that @xmath75 removing the set of disks @xmath76 from @xmath53 yields a fundemantal domain for the action of @xmath19 , and by triangulating this domain we can easily see that the quotient @xmath40 has euler characteristic @xmath77 .
associated to @xmath19 is a bowen - series map @xmath78 , where @xmath79 , defined simply by @xmath80 the associated ruelle transfer operator , acting on the hilbert space @xmath81 of square - integrable analytic functions , is defined by @xmath82 this operator is trace class on @xmath81 for any @xmath18 , and by pollicott @xcite we can use it to write the selberg zeta function as a determinant : for all @xmath83 , @xmath84 see ( * ? ? ?
5.8 ) for details of the proof in the setting of hyperbolic surfaces .
note that this formula gives a rather direct proof that @xmath14 extends to an entire function in the convex - co - compact case .
( conversely , the existence of poles in the zeta function in the case of a surface with cusps shows that we could not extend these methods to that case . )
our numerical method rests on a algorithm adapted from jenkinson - pollicott @xcite and guillop - lin - zworski @xcite for computing the coefficients in the power series expansion , @xmath85 using a lemma of grothendieck ( * ? ? ?
* lemma 2 ) , pollicott - rocha ( * ? ? ?
* prop 3.6 ) prove that for any compact @xmath86 there exist constants @xmath87 , @xmath88 such that @xmath89 hence we have a series expansion for the zeta function , @xmath90 converging uniformly on compact sets in @xmath68 .
the formula for @xmath91 comes from expanding the exponential in @xmath92 and then collecting powers of @xmath93 .
as noted in @xcite , it is convenient to compute @xmath91 as a sum @xmath94 where the @xmath95 are defined recursively by first setting @xmath96 and then taking @xmath97 and @xmath98 for @xmath99 .
this recursive construction reduces the problem to a computation of @xmath100 . under the decomposition @xmath101 , the component @xmath102 is given by @xmath103^s u(s_i^{-1}z)\bigr|_{z\in d_j}\ ] ] the components of @xmath104 are indexed by @xmath105 so that @xmath106.\ ] ] if @xmath107 , then the holomorphic fixed point formula ( see , e.g. ( * ? ? ? * lemma 15.7 ) ) gives @xmath108 = \frac{(t_\sigma^{-1})'(w_-)^s}{1 - ( t_\sigma^{-1})'(w_-)},\ ] ] where @xmath109 is the repelling fixed point of @xmath110 .
an easy computation shows that @xmath111 , so the formula for @xmath100 becomes @xmath112 from we can see that the information required to compute @xmath14 from consists of the set of lengths @xmath113 for each @xmath114 .
note that the multiplicities in this set do not agree with those of the length spectrum itself , because the cyclic permutations of a given index set are listed separately in @xmath115 . to explain the precise connection , each element @xmath116 of the length spectrum corresponds to a non - trivial conjugacy class within the group .
for each such conjugacy class there is a minimal word length @xmath32 needed to write a representative of the class in terms of the basis @xmath117 .
the length @xmath116 associated to the conjugacy class appears in the set @xmath118 only for @xmath32 given by this minimum word length , but with a multiplicity given by the number of representatives of length @xmath32 within the conjugacy class .
this is the same as the number of distinct cyclic permutations of the word .
jenkinson - pollicott @xcite proposed using the truncated series , @xmath119 as a numerical method for computing @xmath33 , as the convergence turns out to be very rapid near @xmath57 .
guillop - lin - zworski @xcite adapted this technique to study the growth of the zeta function and the resonance count in strips , for schottky reflection groups . in this paper
we apply this same approach to the approximation of @xmath120 for convex co - compact schottky groups , but restricted to groups with two generators . in
, we saw that the information required for the calculation of is the collection of sets @xmath118 for word lengths @xmath121 .
the number of words of each length is @xmath122 so to calculate @xmath123 requires a computation of roughly @xmath124 elements of the length spectrum .
this makes it somewhat impractical to apply the algorithm unless @xmath125 .
the dependence on @xmath126 presents another serious limitation .
the terms in the sum for @xmath100 grow exponentially for @xmath38 .
because of this , the relative errors @xmath127 also grow exponentially as @xmath126 decreases , as illustrated on the right side in figure [ errors.fig ] . while the computations are still feasible for small negative values of @xmath126 in many cases , the range is quite limited in this direction . on the other hand , for @xmath128
the errors grow quite slowly as @xmath129 , so we can handle relatively large values of @xmath39 . cc err1.pdf ( 37,47)@xmath130 ( 98,6)@xmath131 & err2.pdf ( 82,47)@xmath132 ( 99,6)@xmath131 another significant factor in the size of error terms is the presence or absence of short geodesics . if @xmath1 has many short geodesics then the @xmath100 terms become much larger , and the convergence of can be impractically slow in this case as well .
longer geodesics correspond in general to smaller values of @xmath33 , so this limitation restricts our attention to @xmath33 on the order of @xmath133 .
one area where we can gain efficiency is in grouping the elements of @xmath118 according to multiplicities .
the length spectrum @xmath15 of a hyperbolic surface was shown to have unbounded multiplicities by randol @xcite .
more recently , ginzburg - rudnick @xcite investigated the length spectrum multiplicity problem for the case of a free group @xmath19 on two generators @xmath134 , e.g. a classical schottky group with @xmath125 .
as mentioned above , the multiplicity problem in @xmath118 is not the true length multiplicity problem , but is related in a direct way .
define the _ length class _ of an element @xmath135 by @xmath136 : = \bigl\{\sigma'\in{\mathcal{w}}_n:\ > \ell(t_{\sigma ' } ) = \ell(t_{\sigma})\text { for any choice of generators}\bigr\}.\ ] ] the full multiplicity of @xmath137 in @xmath118 will be the size of the class @xmath138 $ ] .
the operations on @xmath139 that clearly preserve the length class are : 1 .
cyclic permutation : @xmath140 2 .
inverse : @xmath141 .
reverse : @xmath142 .
4 . decompose @xmath110 as a word @xmath143 and then reverse the order by @xmath144 a typical word of length @xmath32 generates a length class of multiplicity @xmath145 , consisting @xmath32 permutations each for the word , its reverse , its inverse , and the reverse of its inverse .
words with greater multiplicity can be formed by applying these symmetry operations to composite words . according to (
* conjecture 1 ) , this should be a complete list of operations that preserve the length class .
this is not proven , and in any case the operations do not always lead to distinct elements of @xmath115 , so it remains a difficult combinatorial problem to understand the structure of the length classes for a general value of @xmath32 . as part of the strategy for efficient computation of the functions @xmath100 ,
we first compute a multiplicity table breaking down @xmath115 into length classes .
this is done simply by computing lengths for all elements of @xmath115 for a few different sets of generators and then correlating the results .
since the groups are freely generated , the structure of the length classes depends only on the number of generators .
table [ lnsize.tab ] shows a comparison of the number of length classes to the number of elements of @xmath118 .
.size of @xmath118 versus number of distinct lengths . [ cols="^,^,^",options="header " , ] the argument principle is the obvious technique for locating or counting zeros of an analytic function .
the derivative @xmath146 can be computed as in , simply be replacing @xmath100 with the derivative , @xmath147 and then running the same recursive algorithm used to compute @xmath91 .
however the logarithmic derivative @xmath148 is highly oscillatory , even away from the zeros , as illustrated in figure [ logzp.fig ] . for @xmath126
closer to @xmath45 these oscillations become much more severe , so that numerical integration of @xmath149 over long intervals is extremely slow and prone to errors .
another issue with this approach is that the numerical routines often fail when poles are close to the path . in regions with a high density of poles , accurate estimation of the integrals
requires adjustments to the paths to avoid poles .
logzp.pdf we have found that a much faster and more stable approach to the application of the argument principle is subdivide the sample region into a rectangular grid . along each short horizontal or vertical segment in the grid , we sample @xmath14 at a finely - spaced sequence of points @xmath150 , and then compute @xmath151 we can efficiently compute the resonance counts in a rectangular bin array by first using this sampling approach to obtain @xmath152 on all of the horizontal and vertical edge segments . for the sampling we specify a minimum spacing , typically 0.01 , as well as a minimum number of samples per edge , but otherwise the samples are spaced uniformly .
the stability of the method is easily checked by counting zeros in the same region with varying bin shapes and sizes .
the method yields correct results even when zeros lie extremely close to the edges of the bins , so we are able to avoid the issue of adjusting paths to avoid poles . for the resonance plots shown in figure [ resplot101.fig ] and below , we start by producing a bin count with a very fine mesh , so that each bin is roughly the size of a pixel in the final plot .
then for each bin with a positive count , we locate the corresponding zero using a standard root - finding algorithm , with seed point furnished by the center of the bin .
although bins with multiple resonances are possibly underrepresented by this technique , this has no impact on the plot because the resulting points would be superimposed and indistinguishable .
since we restrict our attention to schottky groups with @xmath125 generators , we can only produce hyperbolic surfaces of euler characteristic @xmath37 .
there are two possible topologies : a surface of genus zero with three funnels or a surface of genus one with one funnel .
trif.pdf ( 26,43)@xmath153 ( 45,65)@xmath154 ( 72,38)@xmath155 a hyperbolic pair of pants is bounded by 3 simple closed geodesics .
there is a unique way to attach funnels to these boundary circles , producing a 3-funnel surface topologically equivalent to a thrice punctured sphere .
we denote such a surface by @xmath156 , where the @xmath157 are boundary lengths of the original pair of pants .
these three real parameters characterize the surface completely up to isometry .
to associate a schottky group to @xmath158 , we set @xmath159 with the parameter @xmath160 chosen by solving @xmath161 . in this scheme , the circles @xmath162 are centered at @xmath163 , and @xmath164 at @xmath165 , as shown in figure [ circles1.fig ] .
circles1.pdf ( 9,15)@xmath166 ( 32,15)@xmath167 ( 63,15)@xmath168 ( 87,15)@xmath169 ( 11,7.5)@xmath170 ( 37,7)@xmath171 ( 49.5,7)@xmath172 ( 59.5,7)@xmath173 ( 85,7.5)@xmath174 ( 52,19)@xmath175 ( 71,27.5)@xmath176 some sample resonance plots for 3-funnel surfaces were shown in figure [ resplot101.fig ] . in the symmetric case , @xmath177 , the resonances are concentrated very strongly a few distinctive resonance curves . along some of these
, for example the curve that starts at the origin , the multiplicity of each resonance is 2 .
the plot on the left in figure [ resplot121212.fig ] shows another example of this structure .
these resonance curves are strongly reminiscent of the resonance families seen in @xmath32-disk scattering systems , as seen , for example , in the plots of numerically calculated resonances in @xcite .
this similarity is perhaps not surprising , as the classical dynamics of a 3-disk scattering system bear a strong resemblance to the dynamics of the symmetric surface @xmath177 .
however , for hyperbolic surfaces the existence of these patterns was not previously predicted .
we believe the origin of the resonance chains can be explained through a symmetry factorization of the zeta function ( work in progress ) .
cc resplot121212.pdf ( 25,-3)@xmath178 & resplot121212p1.pdf ( 25,-3)@xmath179 one very interesting feature , in view of the decay rate conjecture to be discussed in
[ decay.sec ] , is the fact that the hubs where resonance curves meet occur on the line @xmath180 .
this seems to hold true for all of the @xmath177 cases .
the hubs do start to lose some coherence as @xmath39 increases , as seen in the plot on the left in figure [ resplot101.fig ] .
when we perturb slightly from the symmetric case , as shown on the right in figures [ resplot101.fig ] or [ resplot121212.fig ] , we see smaller - scale oscillations of the resonance curves .
the tightly concentrated hubs near @xmath181 remain , but are somewhat broader .
figure [ res10perturbed.fig ] traces the evolution of the patterns as the case shown in figure [ resplot101.fig ] is further perturbed .
ccc resplot10p00.pdf ( 12,-4)@xmath12 & resplot10p02.pdf ( 12,-4)@xmath182 & resplot10p1.pdf ( 12,-4)@xmath13 + & & + resplot10p2.pdf ( 12,-4)@xmath183 & resplot10p4.pdf ( 12,-4)@xmath184 & resplot101011.pdf ( 12,-4)@xmath185 the spacing of the resonances within each resonance curve is extremely consistent .
for example , consider the bottom left arc of the @xmath186 plot in figure [ resplot121212.fig ] . over the first thousand points ( from @xmath187 to @xmath188 ) the distance between resonances decreases monotonically from @xmath189 to @xmath190 .
similarly , for the corresponding arc for @xmath191 , the spacing decreases from @xmath192 at an extremely slow rate .
it is interesting to note that @xmath193 while @xmath194 .
this approximate ( and never exact ) spacing of @xmath195 appears in all of the @xmath177 cases , and is stable under small perturbations . further away from the symmetric case , the pattern of resonance curves changes significantly , as shown in figure [ resplot121415.fig ] .
the curves visible on the right in figure [ resplot121212.fig ] actually flatten out very quickly as we move away from the symmetric case , and the curves visible in figure [ resplot121415.fig ] are actually new patterns of coherence that emerge between points in these flattened curves .
one can see the onset of this coherence in the final two plots of figure [ res10perturbed.fig ] .
the spacing along these curves is also very consistent . for both surfaces
shown shown in figure [ resplot121415.fig ] , the spacing along each curve is approximately @xmath196 .
cc resplot121314.pdf ( 25,-3)@xmath197 & resplot121415.pdf ( 25,-3)@xmath198 torf.pdf ( -1,31)@xmath153 ( 40,30)@xmath154 ( 25,23)@xmath199 a suitable parametrization of the moduli space for a hyperbolic surface of genus one with one funnel was worked out in buser - semmler @xcite .
the parameters consist of two lengths @xmath200 , and an angle @xmath201 , and we will denote the corresponding surface as @xmath202 .
the corresponding generators in @xmath70 are defined by @xmath203 @xmath204 the length parameters are the lengths @xmath205 for @xmath206 of two simple closed geodesics , with @xmath199 is the angle between them , as shown in figure [ torf.fig ] .
if @xmath207 are sufficiently large and @xmath199 is sufficiently close to @xmath208 , then @xmath175 and @xmath176 will be schottky generators , with circles as shown in figure [ circles2.fig ] .
circles2.pdf ( 37,22)@xmath176 ( 53,29)@xmath175 ( 51,14)@xmath199 ( 23,-1)@xmath209 ( 49,-1)@xmath172 ( 51.5,-3)@xmath210 ( 62,-1)@xmath211 ( 92,-2)@xmath212 a portion of the resonance sets for two funneled tori are shown in figure [ resplottor.fig ] . the pattern that appears in the plot for @xmath213 on the left is the result of overlapping resonance curves similar to those seen for the 3-funnel surfaces .
the curves is clearly resolved if we plot on a smaller scale , as illustrated in figure [ rescurve1010.fig ] .
cc resplot1010.pdf ( 25,-3)@xmath214 & resplot1010p01.pdf ( 25,-3)@xmath215 rescurve1010.pdf for funneled tori with parameters further away from the symmetric case , the resonance plots appear incoherent , as shown in figure [ resplot1012.fig ] .
the resonance curves do persist if we perturb only very slightly from the symmetric case , as the plot on the right in figure [ resplottor.fig ] illustrates .
however , the pattern changes quite rapidly as the parameters vary .
presumably patterns of resonance curves are present even in plots such as figure [ resplot1012.fig ] , but overlapping to such an extent that they are difficult to distinguish . resplot1012.pdf
the work of sjstrand @xcite on semiclassical bounds for resonance counting has led to a general expectation for chaotic scattering systems that the number of resonances near the continuous spectrum should satisfy a power law with exponent equal to half of the dimension of the classical trapped set . recently a large number of theoretical @xcite , numerical @xcite and experimental @xcite studies have appeared in support of this conjectural ` fractal weyl law ' . for a geometrically finite hyperbolic surface , @xmath40 ,
a geodesic is trapped if and only if its lift to @xmath53 has endpoints in the limit set @xmath50 introduced in [ prelim.sec ] . since @xmath216 , it follows by a straightfoward calculation that the dimension of the trapped set ( as a subset of @xmath217 ) is @xmath218 .
let us define the resonance counting function for a vertical strip as @xmath219 for the general case of open chaotic systems the fractal weyl conjecture was first made formally in lu - sridhar - zworski @xcite . in our context
the conjecture could be paraphrased as : for a geometrically finite hyperbolic surface @xmath1 , @xmath220 provided that @xmath139 is sufficiently small .
the conjecture is sometimes stated in a stronger form , as an exact asymptotic , @xmath221 for some constant @xmath222 . for finite - area surfaces , since @xmath223 and all resonances are contained in a vertical strip , the fractal weyl law is simply the weyl law .
guillop - lin - zworski @xcite , improving on an earlier result of zworski @xcite , proved a stronger , localized version of the conjectured upper bound : for any @xmath139 , @xmath224 note that this is consistent with a remainder term of @xmath225 for the upper bound in .
this estimate apples to hyperbolic schottky manifolds in any dimension , and has been extended to general asymptotically hyperbolic manifolds by datchev - dyatlov @xcite , with @xmath33 reinterpreted in terms of the upper minkowski dimension of the trapped set . for conformally compact hyperbolic surfaces ,
dyatlov - guillarmou @xcite have established a fractal bound on the remainder term in the weyl asymptotic for the scattering phase , also consistent with the fractal weyl law .
the best known lower bound comes from the trace formula and is considerably weaker than the conjectured growth rate .
guillop - zworski @xcite proved that for any @xmath226 , @xmath227 jakobson - naud @xcite proved a lower bound that shows explicit dependence on @xmath33 , but the growth rate is only logarithmic unless the group is arithmetic . for the 3-funnel surfaces , large scale fluctuations in the growth rate seem to be a persistent feature of the vertical - strip counting function .
figure [ frweyl12.fig ] shows a comparison across several surfaces with nearby parameters , with the bounds as conjectured in the fractal weyl law included for comparison .
( the straight lines are graphs of @xmath228 , for two values of @xmath229 chosen arbitrarily . )
the results are certainly not inconsistent with the fractal weyl law prediction , but within the range of these plots , at least , the curves are too unstable for any meaningful estimate of the growth rate .
ccc frweyl12a.pdf ( 20,40)@xmath230 ( 20,47)@xmath231 & frweyl12b.pdf ( 20,40)@xmath232 ( 20,47)@xmath197 & frweyl12c.pdf ( 20,40)@xmath233 ( 20,47)@xmath234 we can also compare our resonance counts to the localized upper bound . figure [ nbox121314.fig ] shows the plot of this resonance count in boxes of increasing imaginary height , for the surface @xmath235 whose cumulative count was shown in the center graph in figure [ frweyl12.fig ] .
the plateaus occurring on either side of @xmath236 are puzzling . to understand what is causing them , we return to a resonance plot for this surface , in figure [ resplot121314.fig ] . as we noted in [ resplot.sec ] , on each curve of resonances the spacing is remarkably consistent . and
since the curves are nearly vertical , this means the space between imaginary parts is essentially constant on each curve .
so the fluctuations in the resonance count shown in figure [ nbox121314.fig ] are accounted for by resonance curves crossing the imaginary axis , rather than by a change in the density within individual curves .
the plateaus occur in zones where the resonance lines bunch together , creating an interval with no crossings .
nbox121314.pdf ( 100,6)@xmath131 resplot121314b.pdf for longer values of the length parameters , we can calculate the strip counting function for much larger values of the imaginary parts .
figure [ weyl121415.fig ] shows the plot of @xmath237 for the 3-funnel surface @xmath238 , up to @xmath239 .
we see similar large scale oscillations even on this much greater scale .
the plateaus in the resonance count noted above persist for large values of @xmath131 , as shown in figure [ nbox121415.fig ] , and within the range of this plot the heights of the plateau do not vary .
another striking feature of this graph is the quasi - periodic structure ; the pattern of peaks and valleys recurs with a period somewhere around 15000 . from one cycle to the next we see a clear rise in the valleys of the distribution .
if the fractal weyl law does hold , then it would seem to require that those valleys must eventually rise above the ( apparently ) fixed height of the plateaus .
unfortunately , direct evidence of this is beyond the range of our calculations .
weyl121415.pdf nbox121415.pdf ( 99,4)@xmath131 recently , naud @xcite , has proved , for convex co - compact @xmath19 , that for @xmath240 there exists a function @xmath241 such that @xmath242 with @xmath243 but @xmath244 for @xmath245 . in other words ,
the optimal growth rate conjectured in certainly does not hold for @xmath245 .
we can see the evidence of this result by plotting the counting function @xmath246 for several values of @xmath139 , as in figure [ fwsigma.fig ] . for @xmath247
the difference in slope is barely evident , but least - squares fit gives an exponent @xmath248 , versus @xmath249 for @xmath250 . at @xmath251 the difference in slope is much more dramatic ; the exponent from the curve fit is only @xmath252 in this case .
fwsigma.pdf ( 100,36)@xmath251 ( 100,49)@xmath247 ( 100,56.5)@xmath253 for funneled tori the log - log plot of the resonance counting function in strips looks more uniform but still has significant long - range fluctuation in all of the examples checked .
figure [ frweyltor.fig ] shows some examples .
note the apparent concavity in the log - log plots .
this behavior would contradict if it were to continue .
to get some idea what is happening , we examine the corresponding localized resonance count in figure [ nbox1213.fig ] . up to @xmath254 ,
the resonance count indeed seems to have faster - than - linear growth , in disagreement with the proven @xmath255 bound .
the final plateau shows ( fortunately ! ) that this behavior does not continue .
the resulting curve is irregular and difficult to draw any long - range conclusions from , but at least it plausibly satisfies the @xmath255 bound .
cc frweyl1212.pdf ( 20,50)@xmath256 ( 20,44)@xmath257 & frweyl1213.pdf ( 20,50)@xmath258 ( 20,44)@xmath259 nbox1213.pdf ( 98,6)@xmath131 another way to try to understand the nature of this distribution is to plot resonance densities as in figure [ dens1213.fig ] .
this plot reveals lines of higher density that evolve slowly as @xmath39 increases .
the asymptotics of the counting function would seem to be dominated by the evolution of these structures .
dens1213.png ( 7,0)@xmath172 ( 28,0)@xmath260 ( 53,0)@xmath261
in physical terms , the resonances with the greatest real part correspond to the most stable states , and hence play a dominant role in the wave asymptotics .
these wave asymptotics are related to asymptotics of the length spectrum by the trace formula , so the most stable resonances are also the most significant in the related problems of lattice point counting and asymptotics of the length spectrum .
see , for example , naud @xcite . in a similar way
, understanding the rightmost edge of the resonance distribution plays a crucial role in various counting problems in number theory . for examples in the context of this paper , see the work of bourgain - gamburd - sarnak @xcite or bourgain - kontorovich @xcite .
gap999.pdf by the positivity of @xmath2 , any resonances in @xmath6 lie in the interval @xmath262 .
hence there must always be a gap between the first resonance set and the bulk of @xmath11 in this case .
naud @xcite has shown that such a gap exists even for @xmath36 : for some @xmath26 , @xmath263 one can see in various plots above that there can be resonances with @xmath126 extremely close to @xmath33 .
for example , in the resonance plot for @xmath238 in figure [ resplot121415.fig ] , the rightmost line of resonances in nearly vertical near @xmath264 ; the next resonance in the line above that point is @xmath265 . within the class @xmath177 ,
the initial curve of resonances starting from @xmath33 becomes more vertical as @xmath116 increases .
compare , for example , the slopes of these curves in figures [ resplot101.fig ] and [ resplot121212.fig ] .
constant @xmath266 evidently takes on arbitrarily small values , just within the class of 3-funnel surfaces .
another version of the spectral gap problem is to find the `` essential spectral gap '' , defined as @xmath267 jakobson - naud @xcite formulated this notion and proved that @xmath268 for any convex co - compact group @xmath19 .
they also made the following : [ sp.gap ] for a convex co - compact group , @xmath269 note that this is consistent with the situation for finite - area surfaces , for which @xmath223 .
in that case @xmath270 because of the location of the critical line , and selberg s result @xcite on accumulation of resonances shows that in fact @xmath271 .
to investigate the high - frequency behavior of the spectral gap , we define a resonance `` envelope '' function , for a given window size @xmath272 , by @xmath273 the essential gap conjecture can then be rephrased as @xmath274 figure [ gap999.fig ] shows a plot of this function for the surface @xmath275 .
the oscillatory pattern is a clear extension of the structure seen on the left in figure [ resplot101.fig ] .
all of the symmetric @xmath177 cases share this same basic pattern for the short range oscillations .
we also contraction of these oscillations away from the line @xmath276 , but this may a temporary or periodic effect . in the density plot for @xmath186 shown below in figure [ dens121212.fig ]
, we can see that the envelope function plot for that surface would show a very regular oscillation out to @xmath277 , with only a very small decrease in amplitude .
figure [ gap121415.fig ] shows a plot of the envelope function for @xmath238 .
the oscillatory pattern is rather striking ; the shorter oscillations have a period of around 800 , but there is also a clear repeat of the pattern of peaks and valleys on a larger scale , with a period of around 15000 .
this of course corresponds to the periodic structure noted in figure [ nbox121415.fig ] .
the peaks of this longer oscillation occur at @xmath278 , @xmath279 , and @xmath280 .
gap121415.pdf for funneled tori , the resonances are generally more uniformly distributed , and the oscillations in the envelope function are of a much smaller amplitude .
figure [ gap1213.fig ] shows an sample plot , for the surface @xmath281 . a clear opening of the gap visible on the right .
over the course of the plot , which covers @xmath282 , the maximum real part decreases from @xmath283 to @xmath284 .
another interesting feature of the resonance distribution for @xmath32-disk scattering systems noted in lu - sridhar - zworski @xcite was a concentration of the average decay rate of resonant states at half of the classical escape rate .
this observation was made in numerical calculations of the resonances for the 3-disk scattering system .
recent experimental results for microwave scattering @xcite have corroborated the existence of such a concentration .
the physical justification for this phenomenon is actually quite straightforward .
semiclassically ( i.e. for large @xmath285 ) , we expect the probability distributions of resonance states to exhibit , in some average sense , features of the classical flow . in particular , the decay rate of quantum states should be related to the classical process of escape to infinity . since the probability distribution is the modulus square of the wave function , the expected quantum decay rate is actually half of the classical escape rate . to define the classical escape rate , for a small fixed @xmath26 and @xmath286 ,
let @xmath287 be the set of points @xmath288 for which there exists some @xmath289 in the trapped set such that @xmath290 for all @xmath291 $ ] ( distance in the sasaki metric on @xmath292 ) . in other words , @xmath287 contains those momenta for which trajectories remain close to trapped trajectories up to time @xmath293 .
the escape rate is defined as the decay exponent for the volume of this set , @xmath294 using the fact that the escape rate can be computed in terms of the topological pressure of the flow , naud @xcite showed that for a conformally compact hyperbolic surface , @xmath295 on the other hand , the decay rate of a resonant state at @xmath296 is @xmath297 .
hence we should expect a concentration of the real parts of resonances near the line @xmath181 .
naud s estimate gives some theoretical support to this conjecture , showing at least that a reduced density of resonances for @xmath298 .
note also that the predicted concentration is consistent with the picture for finite - area surfaces . in that case
the classical escape rate is zero , and dyatlov ( * ? ? ?
* eq . ( 1.17 ) ) shows that under generic conditions most resonances lie very close to @xmath25 .
we have already noted the hubs of resonance curves near @xmath181 in figures [ resplot101.fig ] and [ resplot121212.fig ] , which shows a concentration of decay rates similar to that seen in the examples cited above .
this concentration continues over a large scale , as illustrated by the density plot in figure [ dens121212.fig ] .
the concentration is most striking in the symmetric cases , but all of the 3-funnel surfaces seem to share the phenomenon of bunching of resonance curves near @xmath181 to some extent .
figure [ rhist.fig ] shows a comparison of two histograms of the distribution of real parts for a portion of the resonance set .
we can track the parameter dependence of this behavior with a density plot of the real parts take over a range of length parameters , as shown in figure [ rdplot.fig ] . for funneled tori ,
the situation is quite different .
the line @xmath181 does not seem to be singled out in the plots of figure [ resplottor.fig ] or figure [ resplot1012.fig ] .
figure [ rhist1213.fig ] shows the histogram of real parts for a range of the resonance set for @xmath281 , which shows no sign of the expected concentration . in the density plot for this surface in figure [ dens1213.fig ] , we see that the curves on which the resonances concentrate are not coalescing near @xmath181 , at least not within the range of the plot .
it is certainly conceivable that these curves do eventually converge on @xmath181 , but within the range of our numerical techniques we are unable to find direct evidence of this .
s. j. patterson , the selberg zeta - function of a kleinian group , _ number theory , trace formulas , and discrete groups : symposium in honor of atle selberg , oslo , norway , july 1421 , 1987 _ , academic press , new york , 1989 .
h. schomerus , k. m. frahm , m. patra , and c. w. j. beenakker , quantum limit of the laser linewidth in chaotic cavities and statistics of residues of scattering matrix poles , _ physica a _ * 278 * ( 2000 ) , 469496 .
a. selberg , remarks on the distribution of poles of eisenstein series , _ festschrift in honor of i. i. piatetski - shapiro on the occasion of his sixtieth birthday , part ii ( ramat aviv , 1989 ) _ , israel math .
, vol . 3 , weizmann , jerusalem , 1990 , pp . | we study the distribution of resonances for geometrically finite hyperbolic surfaces of infinite area by counting resonances numerically .
the resonances are computed as zeros of the selberg zeta function , using an algorithm for computation of the zeta function for schottky groups .
our particular focus is on three aspects of the resonance distribution that have attracted attention recently : the fractal weyl law , the spectral gap , and the concentration of decay rates . |
it is well known that images of the rest - frame uv and optical continua are elongated preferentially along the radio axis in powerful radio galaxies ( prgs ) at redshift ( @xmath9 ) @xmath10 ( e.g. , chambers et al .
1987 ; mccarthy et al .
1987 ) ; the so - called alignment effect .
indeed , many high-@xmath9 ( @xmath11 ) prgs ( hzprgs ) show the alignment effect , and its origin has been in debate in this decade ( e.g. , mccarthy 1993 ) .
various models have been proposed to explain the alignment effect ; e.g. , ( 1 ) scattering of the anisotropic radiation from a central engine of active galactic nuclei ( agn : e.g. , di serego alighieri et al .
1989 ; fabian 1989 ; cimatti et al . 1998 ) , ( 2 ) the jet - induced star formation ( e.g. , chambers et al . 1987 ; mccarthy et al . 1987 ) , and ( 3 ) the jet - induced shock heating ( e.g. , de breuck et al . 2000 ) . in these models , agns play some important roles to create an extended aligned continuum emission at various wavelengths .
the nebula contribution to the extended uv continua is also suggested in some prgs ( e.g. , dickson et al .
1995 ) .
an alternative idea , a pair of elongated bubbles or bi - directional cones associated with the superwind / starburst activity of the host galaxy , is proposed to explain a figure-8 shaped extended nebula in another radio galaxy , mrc 0406 - 244 , at @xmath12 ( taniguchi et al .
2001 ; see also rush et al .
1997 ; pentericci et al .
although the radio - jet activity and the starburst / superwind activity are different physical processes and their timescales are also significantly different , the alignment effect can be understood in the following way ( taniguchi et al .
the superwind could blow as a bipolar wind , which is often observed in many superwind galaxies in the local universe ( see heckman et al .
1990 ) , since it is likely that the gas in the host galaxy is distributed with a disk - like configuration even for a young host galaxy of hzprgs .
the radio jet is also expelled to the two directions perpendicular to the accretion plasma disk . therefore , we can explain the alignment effect if the accretion disk is nearly co - planer to the host gas disk .
since some of these objects are known to experience vigorous star formation activity ( e.g. , dey et al . 1997 ) , this new mechanism can be regarded as one of possible ideas .
study of such superwind nebulae in high-@xmath9 universe seems very useful to understand the formation and evolution of galaxies from a general point of view .
the reason for this is that galaxies are expected to experience the galactic wind ( i.e. , the powerful initial superwind in a late phase of the galaxy formation ) , which is considered to play significant roles to determine global characteristics of the present - day massive elliptical galaxies ( e.g. , arimoto & yoshii 1987 ) . since we know many hzprgs even at redshift @xmath13 3 ( e.g. , rttgering 1997 ; van ojik et al .
1997 ; van breugel et al .
1999 , de breuck et al .
2002 ) thanks to their powerful radio emission , their superwind activities can be used to probe early star formation history in their host galaxies .
powerful 8 - 10 m - class telescopes enable us to investigate them in detail ( e.g. , villar - martin et al .
2000 ) . in this paper , we focus our attention to one of hzprgs , 1243 + 036 (= 4c+03.24 ) , at @xmath14 , which has the aligned ly@xmath0 nebula , and is one of the highest-@xmath9 objects known to show an alignment effect ( see , e.g. , chambers et al . 1990 ; lacy et al . 1994 ) .
this object was identified as a hzprg during the course of spectroscopic survey of ultra steep spectrum sources conducted at eso ( rttgering 1994 , 1995 , 1997 ; van ojik et al .
its detailed follow - up observations were made by van ojik et al .
( 1996 ; hereafter vo96 ) . according to vo96 , the observational properties of the nebula are very complex in morphology and kinematics , and can be summarized as follows km s@xmath4 mpc@xmath4 and a deceleration parameter of @xmath15 for consistency with the previous study of vo96 . with these cosmological parameters , 1
arcsecond corresponds to 6.8 kpc at a distance of 1243 + 036 . ] .
( a ) this object is surrounded by a huge ly@xmath0 halo which extends over 135 kpc .
( b ) its ly@xmath0 luminosity amounts to @xmath16(ly@xmath0 ) @xmath17 ergs s@xmath4 .
( c ) its morphology shows the alignment effect .
( d ) its kinematical properties show the presence of the following two components ; i ) the high - velocity width component with fwhm ( full width at half maximum ) @xmath18 1500 km s@xmath4 at @xmath19 km s@xmath4 ( blueshifted ) , and ii ) the halo component with fwhm @xmath18 250 km s@xmath4 .
the halo component extends over 135 kpc with an average velocity gradient of @xmath18 450 km s@xmath4/135 kpc .
in addition , recent high - resolution images of @xmath20-band , @xmath21-band , and narrow - band near 2.3@xmath22 m ( corresponding to the rest - frame uv - optical continua and redshifted [ oiii]4959 , 5007 emission line , respectively ) have revealed closely - aligned continuum and line emission along the narrow channel of the radio jet ( van breugel et al .
1998 ; pentericci et al .
although the close spatial coincidence between the uv / optical and radio emission strongly suggests a possibility of the radio jet - induced alignment component in this object , the fan- or cone - like appearance of the ly@xmath0 nebula seems to indicate the presence of another nebula component within the same object .
therefore it is interesting to investigate this object in more detail and to explore its origin unambiguously .
observation was made with the faint object camera and spectrograph , focas ( kashikawa et al .
2002 ) , attached at the cassegrain focus of the subaru telescope ( kaifu 1998 ) on april 22 , 2002 ( ut ) to study the ly@xmath0 nebula around 1243 + 036 .
the sky condition was clear and the seeing size ( fwhm ) was 0.6 .
a 300 second @xmath23-band image , which includes the redshifted ly@xmath0 emission at @xmath24 , was obtained without ccd binning ( 0.1 per pixel ) just before the spectroscopic observation for target acquisition purpose .
then , a 0.8 long slit was placed on the peak of the @xmath23-band image at a position angle ( pa ) of 152@xmath25 .
the slit positioning accuracy was estimated to be about 0.5 due to the faintness of the target .
@xmath26 pixel@xmath27 ccd binning was applied at the time of spectroscopy , and a single 1800 sec exposure was taken to give a spectrum covering from 4700 to 9400 with the combination of both the medium - dispersion grism ( 300b ) with 300 grooves mm@xmath4 and the order sorting filter ( y47 ) .
the resultant resolution was 0.3 in space and 2.81 in wavelength per pixel corresponding 150 km s@xmath4 at the redshifted ly@xmath0 emission .
the data reduction was made by using both the idl - based focas data reduction software ( yoshida et al .
2000 ) and the iraf in a standard way ( overscan subtraction , bias subtraction , and flat fielding by using the dome flat images ) .
the wavelength calibration was made by using sky emission lines , and could be done with high accuracy ( @xmath28 ) because the redshifted ly@xmath0 emission is very close to the strong [ o i ] sky emission at 5577 .
we show our @xmath23-band image overlaid on the continuum image taken with a f702w filter of hst ( pentericci et al .
1999 ) in figure 1 .
the @xmath23-band image is smoothed by gaussian convolution to enhance the weak emission at outer parts of the nebula ; the effective seeing ( after the smoothing ) is 1.1 fwhm .
the total magnitude of the nebula is @xmath29 .
it is clearly shown that our @xmath23-band image reveals widely extended nebula component , showing remarkable contrast with the continuum emission , which is tightly aligned to the narrow radio jet ( vo96 ; pentericci et al .
note also that the @xmath23-band image appears very similar to the ly@xmath0 image taken with a narrow - band filter ( vo96 ) and thus our @xmath23-band imaging traces the ly@xmath0 mostly .
however , there is a notable difference between the two at northwest ( nw ) of the nucleus : although the ly@xmath0 image shows an extended ly@xmath0 emission only toward southeast ( se ) of the nucleus , our @xmath23-band image shows an extended nebula toward both nw and se of the nucleus .
since our @xmath23-band filter bandpass covers the whole redshifted ly@xmath0 emission , and the contribution of the continuum emission should be almost negligible for the extended component judging from both the prominent nuclear source and weak emission extended along the jet ( vo96 ; van breugel et al .
1998 ; pentericci et al .
1999 ) , our @xmath23-band image and the ly@xmath0 image of vo96 should reveal the gaseous nebula in a similar way .
therefore this apparent discrepancy in two images needs an explanation .
our ly@xmath0 spectrogram is shown in figure 2 along with the slit position overlaid on our @xmath23-band image .
the spectrogram is also smoothed by 4 by 4 binned - pixel running averaging to enhance the outer weak emission .
after the smoothing , the effective resolution is 1.2 in space , which is almost comparable to that of the smoothed imaging data .
the velocity resolution is 600 km s@xmath4 . in order to check the consistency quantitatively between the spectrogram and the @xmath23-band image
, we made an one - dimensional plot of the spatial ly@xmath0 flux distribution by integrating the @xmath23-band flux over an expected slit position along the slit width directions ( figure 2b ) .
a similar plot was also made from the spectroscopic data by integrating the flux on the spectrogram along the wavelength direction at each spatial position ( figure 2d ) .
these two plots look very similar to each other ; i.e. , both show not only an asymmetric distribution toward se but also a diffuse extension toward nw at a fainter flux level .
this suggests not only that our spectroscopy and imaging data are consistent with each other ( i.e. , the spatial orientation of the spectrogram is right , and the slit was placed at the expected position at the center of the nebula ) , but also that the nw nebula does exist .
it seems noteworthy that the nw component is blueshifted at 5541 or bluer for the most part ( figure 2c ) .
such a blueshifted component could not be detected by vo96 due to rather narrow wavelength coverage of their narrow - band filter ( 5541
therefore , we conclude that there is a pair of elongated nebulae along nw and se directions around 1243 + 036 .
our ly@xmath0 spectrogram ( figure 2c ) reveals complicated structures of the nebula both in space and in velocity .
we group the nebula into the following four components : ( i ) the compact nucleus component , ( ii ) the extended near - systemic component at both sides of the nucleus , ( iii ) the redshifted component aside the nucleus , and ( iv ) the extended blue components at both sides of the nucleus .
we give below a summary of their observational properties and their comparisons with results of vo96 , who conducted similar long - slit spectroscopy .
note that , although the slit position ( centered on the nebula ) and its orientation ( pa=152@xmath25 ) are the same for the two , the slit widths ( 2.5 for vo96 and 0.8 for us ) and the velocity resolutions ( 150 km s@xmath4 for vo96 and @xmath30 km s@xmath4 for us ) are different .
\(i ) the component at @xmath31 around the peak of the ly@xmath0 emission shows an evident blue - asymmetric profile with its peak at 5560 , being consistent with the report by vo96 ( at 5557 ) .
its intensity peak corresponds spatially to that of the rest uv - optical continuum emission with which the radio core showing flatter spectrum is associated ( vo96 ; van breugel et al .
1998 ; pentericci et al .
therefore , we attribute this component to the nucleus component . the blue wing emission can be traced down to @xmath6 km s@xmath4 in the rest frame of the galaxy ( hereafter all velocities are shown with respect to the systemic velocity in the rest frame of the galaxy ) although a part of the wing may be contaminated by the off - nucleus blue components at both sides of the nucleus ( see below ) .
\(ii ) the extended component whose velocity is close to the systemic one is found at both sides of the nucleus , and can be traced out to @xmath32 ( 4 - 5 ) from the nucleus .
it shows no strong velocity shear across the extension ( @xmath33 km s@xmath4 ) although the sky emission at 5577 might affect the measurement of its velocity curve given our lower velocity resolution .
the velocity width is narrower than the nucleus component across the extension since we could not resolved the width by our instrument resolution ( 600 km s@xmath4 fwhm ) .
vo96 detected an `` outer halo '' component extending out to @xmath34 on both sides of the nucleus at almost systemic velocity .
this component has a narrow line width ( @xmath35 km s@xmath4 fwhm ) and shows a large - scale velocity shear of as much as @xmath36 km s@xmath4 .
although our spectroscopic resolution is not high enough to investigate such a velocity curve , our near - systemic component appears to correspond to the outer halo component of vo96 although we detect it only out to @xmath37 .
\(iii ) the redshifted component is detected at @xmath38 nw of the nucleus , se of the nucleus ( i.e. , another side of the nucleus ) in figure 11 of vo96 .
] and is found at redward of the sky emission . due to the strong sky contamination , little information is available for this component .
\(iv ) the blueshifted component at se extends up to @xmath39 from the nucleus and shows strong blue - shifted velocity of @xmath40 km s@xmath4 . at the opposite side of the nucleus ( nw ) ,
another blue component is detected which is much more extended ( @xmath41 @xmath42 from the nucleus ) and shows much wider velocity width ( @xmath7 km s@xmath4 fwhm corrected for the instrumental resolution ) .
the velocity curve shows the bluest velocity of @xmath6 km s@xmath4 at @xmath18 2.5 nw , and the velocity slows down at outer regions ( @xmath40 km s@xmath4 at @xmath43 nw ) .
vo96 has also detected blueshifted and extended component .
note , however , that there might be a difference between their spectrogram and ours .
we could not detect any compact emission - line component at @xmath44 nw from the nucleus ,- band image and spectrogram are consistent with each other in orientation , as shown in section 3.1 , and our @xmath23-band image and their narrow - band ly@xmath0 image ( their figure 9 ) are also consistent with each other in orientation , given the different wavelength coverage of the nebula ( as explained in section 3.1 ) , ensuring that our orientation is right .
therefore , although vo96 claimed that this high - velocity component is found at se , its correct position is probably at the opposite side ( nw ) of the nucleus . in the following , we assume an image flip of the vo96 s spectrogram for the comparison . ]
although vo96 claimed a presence of an enhanced ly@xmath0 emission there which shows blueshifted velocity of @xmath19 km s@xmath4 with a wider line width of @xmath45 km s@xmath4 fwhm .
our spectrogram shows , on the other hand , a component which shows similar kinematical properties ( blueshifted by @xmath46 km s@xmath4 with an even wider line width of @xmath47 km s@xmath4 fwhm ) but is extended more smoothly over @xmath41 . this difference could be explained by the different slit coverage on the nebula between the two spectrograms , i.e. , our narrow slit ( @xmath48 of vo96 s ) could not cover a portion of the nebula where enhanced ly@xmath0 emission is present .
if this is the case , the nebula might have complex sub - structures within a nebula .
we discuss plausible origins of the alignment effect observed in 1243 + 036 .
there are some pieces of evidence indicating that other mechanism than the radio - jet related one works to excite the extended nebula in 1243 + 036 .
first , the previous high - resolution images of @xmath20-band , @xmath21-band , and narrow - band near 2.3@xmath22 m ( corresponding to the rest - frame uv - optical continua and redshifted [ oiii]4959 , 5007 emission line , respectively ) have revealed closely - aligned continuum and line emission along the narrow channel of the radio jet ( van breugel et al . 1998 ; pentericci et al . 1999 ) .
although this indicates an importance of the radio jet activity to the alignment effect , the ly@xmath0 nebula of this galaxy is extended much more widely than the region of the aligned continuum ( figure 1 ) .
second , the extended components are blueshifted at both side of the nucleus .
the jet - induced extranuclear shock , if any , should show a pair of blueshifted and redshifted components at each side of the nucleus because the symmetric propagation of a pair of jet and its counter jet is expected , being inconsistent with the observation . therefore it seems very likely that the effect of the radio jet onto the extended nebula can not explain the excitation of the whole ly@xmath0 nebula .
another idea of the origin of the widely extended ly@xmath0 nebula is the circumgalactic matter ionized by the anisotropic hard radiation from agn .
if this is the case , the matter irradiated by the agn radiation is required to be distributed over a large scale ( @xmath49 kpc ) at a violent kinematical status to explain the velocity shift of @xmath40 km s@xmath4 together with a large velocity dispersion of 1000 - 2000 km s@xmath4 .
one possibility may be that such violent gaseous systems are tidally - induced structures formed through a putative merger event . however , even nearby luminous mergers show an overall velocity differences of @xmath50 km s@xmath4 or less along the tidal tails ( e.g. , hibbard & yun 1999 ; mihos & bothun 1998 ) .
also the wider velocity dispersion seems difficult to be reproduced along the tidal tail ( e.g. , hibbard & yun 1999 ; mihos & bothun 1998 ) .
another candidate for the extended matter is the remaining sub - galactic clumps for the galaxy formation .
however , the range of the clump velocity dispersion would be 300 km s@xmath4 or less because gaseous clumps with a higher velocity dispersion are difficult for them to assemble to form a galaxy .
therefore this idea could not be simply applied for the case of 1243 + 036 .
an alternative idea for the origin of the blue - shifted extended ly@xmath0 nebula is the shock - excited nebula associating with the superwind outflow .
our @xmath23-band image reveals a pair of elongated bubbles or cone - like structures extending into opposite ways from the nucleus .
similar extended nebulae are often seen around the starburst galaxies with the superwind / superbubble activity , such as m82 , ngc 3079 , and arp 220 ( e.g. , heckman et al . 1990 ) .
also similar nebula is found around a high-@xmath9 ( @xmath12 ) radio galaxy mrc 0406 - 244 which shows a pair of shock - excited elongated bubbles associated with the superwind activity ( taniguchi et al .
2001 and references therein ) .
if 1243 + 036 is also the case , the projected velocity field of the nebula would show two components at different velocities each of which corresponds to either front or rear surface of the expanding bubble or outflowing cone ( e.g. , heckman et al .
1990 ) , rather than a single blueshifted component as observed . if the galaxy is surrounded by the dusty halo , which is often found around hzprgs ( van ojik et al .
1997 ) , the redshifted emission from the rear surface of the bubble may be attenuated due to the longer path through the halo .
it seems also possible that the strong sky contamination at just red side of the ly@xmath0 emission might make it very difficult to detect redshifted counterparts of the blueshifted extended components .
therefore the blueshifted and extended components at both sides of the nucleus can be explained by the expanding bubble / outflowing cone model .
the top of the expanding bubble , if not blown out to form conical outflow ( e.g. , heckman et al .
1990 ) , will show a single line - of - sight velocity .
if the superbubble around 1243 + 036 expands almost within the sky plane , then the observed velocity would be blueshifted from the systemic velocity within the bubble , and it goes back closer to the systemic one at the tip .
wider velocity width may be due to the kinematical disturbance at the shock front of the bubble .
since the superwind / superbubble model can reproduce the observed morphological and kinematical properties of the nebula in qualitative ways , this model seems more plausible than the model of irradiated disturbed matter by agn .
although we could not reject a possibility of the contribution of the agn excitation to the nebula , it seems very likely that the superwind activity still plays an important role to bring the disturbed material around the galaxy as observed .
we apply a simple superwind model to the nebula around 1243 + 036 to see whether the observed size and the kinematical conditions can be reproduced quantitatively by the model .
to estimate order - of - magnitude values of the nebula size and the expansion velocity , we apply a simple model of the evolution of a single spherical superbubble . the radius and the velocity of the expanding shell at time @xmath51 ( in units of 10@xmath52 yr ) in the model are @xmath53 and @xmath54 where @xmath55 is the mechanical luminosity released collectively from the supernovae in the central starburst in units of @xmath56 erg s@xmath4 , and @xmath57 is the average hydrogen number density of the inter galactic medium ( igm ) in units of @xmath58 @xmath59 ( mccray & snow 1979 ; koo & mckee 1992a , 1992b ; heckman et al . 1996 ) . since the host galaxy is luminous in blue in the rest - frame of the galaxy ( @xmath60 ; van breugel et al .
1998 ) , it seems to be a massive elliptical galaxy with a mass @xmath61 .
assuming an instantaneous burst of star formation of @xmath62 , the mechanical luminosity is estimated to @xmath63 erg s@xmath4 over a time scale of @xmath64 yrs ( leitherer et al .
1999 ) . and @xmath65 are proportional to just one fifth power of @xmath55 . ] the average hydrogen number density of the igm is estimated following taniguchi & shioya ( 2000 ) , and is @xmath66 at @xmath14 in the adopted cosmology . substituting all these values into the equations , we find @xmath67 kpc and @xmath68 km s@xmath4 .
therefore both the observed size of the blueshifted extended nebula ( @xmath8 kpc in radius , corresponding to 4.5 ) and the maximum blueshift velocity ( @xmath6 km s@xmath4 ) can be roughly reproduced if we assume @xmath69 0.1 0.3 , which is comparable to the age of a nearby superwind nebula of arp 220 ( e.g. , heckman et al . 1996 ) . hence the superwind model can be applicable to the blueshifted extended nebula of 1243 + 036 in a quantitative way .
we found an elongated giant nebula ( @xmath70 kpc ) with blueshifted ( @xmath19 @xmath6 km s@xmath4 ) and broad ( @xmath18 1000 1900 km s@xmath4 fwhm ) ly@xmath0 emission around a hzprg 1243 + 036 at @xmath14 .
the nebula can be recognized as a pair of expanding elongated bubbles or outflowing bi - directional cones at both sides of the nucleus , being much similar to some prototypical superwind galaxies in local universe in both morphological and kinematical points of view .
we show that the superwind model can be applicable to the nebula to explain the size and the kinematical properties of its bubble or cone - like structure . in order to see such an evolved nebula associated with the superwind activity at @xmath14 ,
the onset of the starburst should be at @xmath71 .
the is because it takes several @xmath72 yrs for the superwind to blow out of the host galaxy since the onset of the burst ( arimoto & yoshii 1987 ) , and another a few @xmath73 yrs is required for the nebula to expand to the observed size .
if this would be the case , it seems very likely that 1243 + 036 is in the phase of the galactic wind after the initial burst for the galaxy formation ( arimoto & yoshii 1987 ) . | we report results of our new spatially - resolved , optical spectroscopy of the giant ly@xmath0 nebula around a powerful radio galaxy 1243 + 036 ( 4c+03.24 ) at @xmath1 . the nebula is extended over @xmath2 kpc from the nucleus , and forms a pair of cones or elongated bubbles .
the high - velocity ( @xmath3 km s@xmath4 ; blueshifted with respect to the systemic velocity ) ly@xmath0-emitting components are detected at both sides of the nucleus along its major axis .
the northwestern nebula is more spectacular in its velocity shift ( blueshifted by @xmath5 km s@xmath4 to @xmath6 km s@xmath4 ) and in its width ( @xmath7 km s@xmath4 fwhm ) over @xmath8 kpc scale .
we discuss possible origin of the nebula ; 1 ) the shock - heated expanding bubble or outflowing cone associated with the superwind activity of the host galaxy , 2 ) halo gas photoionized by the anisotropic radiation from the active galactic nuclei ( agn ) , and 3 ) the jet - induced star - formation or shock . the last possibility may not be likely because ly@xmath0 emission is distributed out of the narrow channel of the radio jet .
we show that the superwind model is most plausible since it can explain both the characteristics of the morphology ( size and shape ) and the kinematical structures ( velocity shift and line width ) of the nebula although the photoionization by agn may contribute to the excitation to some extent . |
for the stellar populations of our milky way galaxy ( mwg ) - the halo , the bulge , and the disk ( thick plus thin disk ) - the fundamental questions that have to be addressed are : when , how and on what timescales did the galactic components form , and was there any connection between them ?
if yes , simultaneously or sequentially ? one possible approach to disentangle the evolutionary scenario is to look for evolutionary signatures in age , dynamics , and chemistry of long - lived stars , in the stellar populations within our mwg . at present , two major and basically different strategies for modelling galaxy evolution can be followed : dynamical investigations which include hydrodynamical simulations of isolated galaxy evolution and of protogalactic interactions reaching from cosmological perturbation scales to direct mergers , and , on the other hand , studies which neglect any dynamical effects but consider either the whole galaxy or particular regions and describe the temporal evolution of mass fractions and element abundances in detail . for the case of a closed box a linear relation between the time - dependent metallicity @xmath0 and the initial - to - temporal gas ratio @xmath1
$ ] follows analytically , where the slope is determined by the yield @xmath2 , i.e. the metallicity release per stellar population .
deviations from this simple relation are explained by lower `` effective '' yields @xmath3 due to outflow of metal - rich gas from the ( now open ) volume or infall of low - metallicity ( presumably primordial ) gas .
such dynamical effects can only be properly treated if simulations can account for the energetics , the composition , and the dynamical state of the galactic gas , as well as the relevant interchange processes , in a self - consistent manner .
this includes the pollution of the different gas phases with characteristic elements by means of various stellar mass - loss mechanisms , the gas phase transitions , and the self - consistent large - scale streaming motions of hot intercloud gas ( icm ) or cool gas ( cm ) that falls in from a reservoir of protogalactic gas .
it follows that galactic regions and components experience mutual dynamical interactions and their evolutions are not decoupled .
let us consider a number of chemical peculiarities of different kinds of galaxies which justify the use of sophisticated multi - phase dynamical descriptions of the interstellar medium ( ism ) in studies of the global scenario of galaxy evolution : \1 ) in the solar vicinity at least three severe problems arise by considering the metallicity distribution of f or g dwarfs , stars that live sufficiently long to trace the evolution of a galaxy : the age - metallicity relation , @xmath2 of the metallicity distribution and , finally , the lack of metal - poor g dwarfs ( the well - known g - dwarf problem ) . various influences on the evolution of the solar neighbourhood have therefore been invoked by different authors and are applied by artificial parametrizations ranging from time - dependent accretion of pristine halo gas to temporal variations of the stellar initial mass function ( imf ) .
although they lack self - consistency , the results can often provide very helpful basic insights into influences of distinct effects .
\2 ) it is also important to explain the observed differences of @xmath2 in bulge , disk and halo of the mwg ( pagel 1987 ) .
\3 ) there is now much evidence that metal - absorption line systems ( als ) in qso spectra arise from the gaseous halos of forming galaxies . in spite of the lack of knowledge of their dynamical properties and their origin , as a working hypothesis it is assumed that the early hot halo gas is produced by supernovae typeii ( sneii ) , because the observed abundance ratios agree well with snii yields ( reimers et al .
the metallicity and radial extent of als should therefore appear automatically in galaxy models as the result of chemical evolution .
\4 ) dwarf galaxies ( dgs ) present a variety of morphological types .
their structural and chemical properties differ from those of giant galaxies .
the dwarf irregular galaxies ( dirrs ) appear with lower @xmath4 at the same gas fraction as gss .
this implies that the metal - enriched gas from sneii was lost from the galaxy rather than astrated ( larson 1974 , dekel & silk 1986 ) . yet
many dwarf spheroidals ( dsphs ) which represent the low - mass end of dgs show not only a significant intermediate - age stellar population , but also more recent sf events ( smecker - hane et al .
1994 , han et al .
1997 ) with increasing metallicity , indicating that gas was kept in the system .
\5 ) gss and dgs differ significantly in their n / o ratios .
a fundamental explanation is needed to explain why gss like the mwg reach higher ratios at larger o abundances than dgs , while n / o is almost restricted to around -1.5 for dgs over a wide range in o ( see fig.3 ) .
for systems and sites of low potential energy we know from empirical studies and theoretical investigations that the ism is on average held in balance by counteracting processes like heating and cooling , turbulence and dissipation ( burkert & hensler 1989 , hensler et al .
since these processes are non - linearly coupled , the effect of neglecting one of them will alter the evolution completely .
a number of studies of self - regulated sf exist with particular attention to the influences of stellar radiation , supernova explosions , and the evaporation / condensation balance between the two chemically and dynamically distinct gas phases , the cloudy medium ( cm ) and the hot intercloud medium ( icm ) ( franco & cox 1983 , ikeuchi et al .
1984 , mckee 1989 , bertoldi & mckee 1995 , kppen et al .
1995,1998 ) .
the evolution of dgs is self - regulated and determined by large - scale outflows ( hensler et al .
1993,1998a ) .
external effects like extended dm halos , the igm pressure , etc .
( vilchez 1995 ) could cause the morphological differences of dgs by regulating the otherwise unbound hot gas that can be held in the galactic halo .
the gas could then either be stripped off , or it cools and recollapses , igniting subsequent sf .
self - regulation with snii energy deposition can also characterise the structure and evolution of galactic disks which reach a lower effective gravitational potential in rotational equilibrium ( firmani & tutukov 1992 ; burkert et al .
1992 , rosen & bregman 1995 ) .
to approach global models of galaxy evolution which yield the structural differences and details , adequate treatment of the dynamics of stellar and gaseous components is essential .
at least the following processes have to be taken into account : sn , sf , heating , cooling , stellar mass loss , condensation and evaporation .
this includes the treatment of the multi - phase character of the ism as well as the star - gas interactions and phase transitions . since gas and stars evolve dynamically , and
because several processes both depend on their metallicities and also influence the element abundances in each component , these models are called * chemodynamical * ( _ cd _ ) . it must be emphasized , however , that the number of free parameters in the _ cd _ scheme is not large but actually smaller than in multi - zone models .
the allowed ranges of parameter values are strongly constrained , either because they are theoretically evaluated ( like e.g. evaporation and condensation ) , empirically determined ( like e.g. stellar winds ) , or because they force self - regulation in a way that is independent of the parameterization .
because of limited space we refer the interested reader to more comprehensive descriptions of the _ cd _ treatment and to different applications ( non - rotating galaxies : theis , burkert & hensler 1992 , hensler , burkert & theis 1993 , hensler , gallagher & theis 1998a ; vertical settling of the galactic disk : burkert , truran & hensler 1992 ; disk galaxies : samland & hensler 1996 ; the mwg : samland , hensler & theis 1997 ( sht97 ) , samland 1998 ; dwarf galaxies : hensler & rieschick 1998 , also section 5 ) .
as a striking success of the _ cd _ treatment we will first briefly present some of the results from a published model ( sht97 ) which can be compared to the observational features of the mwg .
the model starts from an isolated spheroidal , rotating but purely gaseous cloud with a mass of @xmath5 , a radius of 50 kpc , and an angular momentum of about @xmath6 pc@xmath7 myrs@xmath8 , corresponding to a spin parameter @xmath9=0.05 .
we assume that the protogalaxy consists initially of cm and icm with a density distribution of plummer - kuzmin - type ( satoh 1980 ) with 10 kpc scalelength .
the initial cm / icm mass division ( 99%/1% ) does not affect the later collapse , because the onset of sf determines the physical state within less than 10@xmath10 years . while the evolutionary phases are described in detail in sht97 ,
here we wish to emphasize the striking agreement of the _ cd _ model after 15 gyrs with first the metallicity distributions of the halo , the bulge , and the solar vicinity ( fig.1 ) , and secondly the radial oxygen gradient within the disk ( fig.2 ) .
convincingly one single _ cd _
model reproduces the different ( otherwise implausible ) @xmath3 in the different regions and demonstrates that they result simply from large - scale streaming effects of the hot gas .
the icm is produced by sneii in overpressure to its surrounding ism and expands until its cooling leads to condensation , i.e. the phase transition to the cm .
in addition , the model shows agreement with the mwg structure , i.e. gas - star content , baryonic mass distributions , velocity distributions , pn and sn rates . from this agreement it may be safe to assume that the temporal evolution of the model is reliable , e.g. the formation epochs of the components and the temporal variation of the radial metallicity gradient ( see sht97 ) .
the enormous energy release by massive stars from their combined wind , radiation and snii explosion leads to violently expanding hot gas bubbles .
they act dynamically on the ambient ism by sweeping it up and squeezing it into dense shells , which break up due to dynamical instabilities .
since massive stars have peeled off their unprocessed shell material during the h - main sequence lifetime , they expel their nucleosynthesized products before and during their wolf - rayet phase and even more intensely by snii explosion . due to its rapid expansion the hot gas engulfs the denser cool clumps and clouds .
the effect is twofold : first , as it passes the clouds the icm significantly perturbs their shape and surface ( elmegreen , this volume ) .
secondly , the contact interface between cm and icm is blown up by heat conduction .
if the cloud is able to get rid of the diffused energy , hot gas can condense onto its surface ; if not , the cloudy material evaporates from the surface and immigrates into the icm . reasonably , this mass exchange by means of evaporation / condensation is a self - regulated process in static media ( kppen et al .
streaming motions , however , can alter this picture and can lead to runaway behaviour in either direction .
certainly this mechanism causes a highly efficient small - scale mixing between the gas phases and by this homogenizes abundances on the local scales of massive star associations . indeed ,
kobulnicki ( this volume ) reports the non - detection of any sizable o , n , and he anomalies from hii regions in the vicinity of young super starclusters in starburst dgs with one exception , ngc 5253 , which reveals a central n overabundance .
while c and n are mainly contributed to the cm by pne from intermediate - mass stars , o and fe are the dominant tracers of snii and snia ejecta , respectively . since the mixture of e.g. n and o can only result from phase transitions between cm and icm , the n / o ratio allows to make qualitative deductions about the mixing direction and its temporal efficiency .
additionally , its radial distribution provides an insight into dynamical effects of the ism . as mentioned in the introduction ( point 5 )
the n / o ratio is smaller in dgs than in gss by almost 0.7 dex , while o is lower by one order of magnitude .
the averaged n / o value for dgs ( see fig.3 ) at -1.46 ( garnett 1990 ) lies only 0.2 dex below the ratio determined from metal - dependent yields ( woosley & weaver 1995 ) for @xmath11and integrated over a salpeter imf .
the almost linear dependence of the nitrogen production on @xmath4 allows for even smaller n / o in dgs because of the generally lower @xmath4 , which explains the observed scatter to even smaller ratios . in order to study and compare abundances and structural signatures in _ cd _ models of dgs
, we have performed 2d simulations of rotating gaseous protogalactic clouds for a large range of initial masses with and without dark matter halos .
the density distribution is again of the plummer - kuzmin - type with the same spin parameter as the above - mentioned mwg model .
here we discuss 10@xmath12 dg models starting with a 2 kpc scalelength .
fig.3 presents a diagram for the n / o vs. o / h ratios of dgs and gss ( also the solar value ) compared with evolutionary tracks by different authors . in contrast to the other models shown ( see garnett 1990 ) which begin and partly remain at too large n / o ranges , both our _ cd _ models commence at very low values due to the delayed nitrogen release by pne and the lower n production in massive stars at low metallicities .
the most evolved track ( 10@xmath12 with dm ) rises rapidly and reaches n / o of -1.8 after 1 gyr and -1.6 after 2 gyrs , respectively .
this small value results numerically as the yield ratio , but as n and o enrich different gas phases , an almost ideal mixing of cm and icm is required . since the hot material can not fully condense onto the clouds , an almost total evaporation of the cm in the vicinity of the sf and snii explosion sites must be invoked .
reasonably , the n mixes perfectly in this case with o in the icm to an almost constant abundance ratio and expands over larger distances within the dg . due to
its cooling the icm forms new condensations of cm where the abundances are observed in hii regions at a constant value ( kobulnicki , this volume ) . in the case of phase transition by means of condensation , only parts of the icm ( and therefore of the o content ) is incorporated into the cm , which leads to higher n / o ratios , but also reveals inhomogeneous n / o distributions .
larger gravitational potentials and resulting mass densities in more massive galaxies lead to a faster cooling of the icm and significantly higher condensation rates which produces a larger n / o . with the same differential mixing processes the observed c / o tendency of dgs ( garnett et al .
1995 ) can also be explained ( rieschick & hensler , in preparation ) .
because of limited space here , we could only briefly demonstrate that various structural and chemical agreement with observations can be achieved self - consistently by global evolutionary _ cd _ models .
if the _ cd _ prescription is applied , important physical processes like large - scale coupling of different galactic regions by dynamical interactions as well as small - scale mixing effects between the gas phases are adequately taken into account , and this substantially fixes the element abundances .
abundances can serve as reliable diagnostic tools of galaxy evolution and provide a chance to deconvolve it in detail , if studies couple the above - mentioned gas processes with the dynamics of gas and stars as well as with their mutual interactions .
is very grateful to the organizers of the conference for their invitation and their kindest hospitality .
the authors achnowledge gratefully cooperations and discussions with j.kppen , m.samland and ch.theis and their contributions to the field of chemodynamics .
we also thank the referee mike edmunds for the careful reading of the text and his comments for clarification .
this work is supported by the _ deutsche forschungsgemeinschaft _ ( dfg ) under grant no .
he 1487/5 - 3 ( a.r . ) , the participation at this meeting by one of us ( g.h . ) under grant no .
he 1487/21 - 1 .
the numerical calculations are partly performed at the computer centers rz kiel , zib berlin , and hlrz jlich .
+ [ -0.8 cm ] bertoldi f. , mckee c.f . , 1995 , _
amazing light _
chiao ( new york : springer ) burkert a. , hensler g. , 1989 , in _ evolutionary phenomena in galaxies _ , eds .
beckmann & b.e.j .
pagel , cambridge university press , p. 230
burkert a. , truran j.s.w . , hensler g. 1992 , _ apj _ , * 391 * , 651 dekel a. , silk j. , 1986 , _ apj _ , * 303 * , 39 firmani c. , & tutukov a. v. , 1992 , _ a&a _ , * 264 * , 37 franco j. , cox d.p . , 1983 , _ apj _ , * 273 * , 243 garnett d.r . , 1990 , _ apj _ , * 360 * , 142 garnett d.r . ,
skillman e.d . , dufour r.j . , et al.,1995 , _ apj _ , * 443 * , 64 han m. , hoessel j.g . , gallagher j.s . , et al
. , 1997 , _ aj
_ , * 113 * , 1001 hensler g. , theis ch . , burkert a. , 1993 , in proc .
3@xmath13 daec meeting _ the feedback of chemical evolution on the stellar content of galaxies _ , eds .
d. alloin & g. stasinska , observatoire de paris , p. 239 hensler g. , gallagher j.s . ,
, 1998a , _ apj _ , * * , submitted hensler g. , kppen j. , samland m. , theis ch . , 1998b , in proc .
of the german - japanese workshop _ galaxy evolution
_ , eds .
n. arimoto & w. duschl , springer , in press hensler g. , rieschick a. , 1998 , iau highlights in astronomy , in press ikeuchi s. , habe a. , tanaka y. d. 1984 , _ mnras _ , * 207 * , 909 kppen j. , theis ch . , hensler g. , 1995 , _ a&a _ , * 296 * , 99 kppen j. , theis ch . , hensler g. , 1998 , _
a&a _ , * 328 * , 121 larson r.b . , 1974 , _ mnras _ , * 169 * , 229 matteucci f. , tosi m. , 1985 , _ mnras _ , * 217 * , 391 mckee , c.f .
1989 , apj , 345 , 782 pagel b.e.j . , 1987 , in : _ the galaxy _ , eds .
g. gilmore & b. carswell , reidel , p. 341
reimers d. , vogel s. , hagen h .- j .
, et al . , 1992 , nature , 360 , 561 rosen a. , bregman j.n . , 1995 , _ apj _ , * 440 * , 634 samland m. , 1994 , phd .
thesis , university of kiel samland m. , 1998 , _ apj _ , * * , in press samland , m. , & hensler , g. 1996 , rev .
, 9 , 277 samland m. , hensler g. , theis ch . , 1997 , _ apj _ , * 476 * , 544 ( sht97 ) satoh c. , 1980 , _ pasj _ , * 32 * , 41 smecker - hane t.a . , stetson p.b . ,
hesser j.e . , et al . , 1994 , _ aj _ ,
* 108 * , 507 theis c. , burkert a. , hensler g. , 1992 , _ a&a _ , *
265 * , 465 vilchez j.m . , 1995 , _
aj _ , * 110 * , 1090 woosley s.e . , weaver t.a . , 1995 ,
_ apjs _ , * 101 * , 181 | since stellar populations enhance particular element abundances according to the yields and lifetimes of the stellar progenitors , the chemical evolution of galaxies serves as one of the key tools that allows the tracing of galaxy evolution . in order to deduce the evolution of separate galactic regions one has to account for the dynamics of the interstellar medium , because distant regions can interact by means of large - scale dynamics . to be able to interpret the distributions and ratios of the characteristic elements and their relation to e.g.the galactic gas content ,
an understanding of the dynamical effects combined with small - scale transitions between the gas phases by evaporation and condensation is essential . in this paper , we address various complex signatures of chemical evolution and present in particular two problems of abundance distributions in different types of galaxies : the discrepancies of metallicity distributions and effective yields in the different regions of our milky way and the n / o abundance ratio in dwarf galaxies .
these can be solved properly , if the chemodynamical prescription is applied to simulations of galaxy evolution . |
in the first days of the discovery of high - temperature superconductivity , anderson@xcite suggested that the two - dimensional hubbard model held the key to the phenomenon . despite its apparent simplicity ,
the two - dimensional hubbard model is a formidable challenge for theorists .
the dimension is not low enough that an exact solution is available , as in one dimension .
the dimension is not high enough that some mean - field theory , like dynamical mean field theory@xcite ( dmft ) , valid in infinite dimension , can come to the rescue . in two dimensions , both quantum and thermal fluctuations are important .
in addition , as we shall see , it turns out that the real materials are in a situation where both potential and kinetic energy are comparable
. we can not begin with the wave picture ( kinetic energy dominated , or so - called weak coupling ) and do perturbation theory , and we can not begin from the particle picture ( potential energy dominated , or so - called strong coupling ) and do perturbation theory . in fact , even if one starts from the wave picture , perturbation theory is not trivial in two dimensions , as we shall see .
variational approaches on the ground state have been proposed,@xcite but even if they capture key aspects of the ground state , they say little about one - particle excitations . even before the discovery of high - temperature superconductivity , it was suggested that antiferromagnetic fluctuations present in the hubbard model could lead to d - wave superconductivity,@xcite a sort of generalization of the kohn - luttinger mechanism@xcite analogous to the superfluidity mediated by ferromagnetic spin fluctuations in @xmath0he.@xcite nevertheless , early quantum monte carlo ( qmc ) simulations@xcite gave rather discouraging results , as illustrated in fig .
[ fig_0 ] . in qmc ,
low temperatures are inaccessible because of the sign problem . at accessible temperatures ,
the d - wave pair susceptibility is smaller than the non - interacting one , instead of diverging .
since the observed phenomenon appears at temperatures that are about ten times smaller than what is accessible with qmc , the problem was left open .
detailed analysis of the irreducible vertex@xcite deduced from qmc did suggest the importance of d - wave pairing , but other numerical work@xcite concluded that long - range d - wave order is absent , despite the fact that slave - boson approaches@xcite and many subsequent work suggested otherwise .
the situation on the numerical side is changing since more recent variational,@xcite dynamical cluster approximation@xcite and exact diagonalization@xcite results now point towards the existence of d - wave superconductivity in the hubbard model . even more recently , new numerical approaches are making an even more convincing case.@xcite and extended @xmath1-wave susceptibilities obtained from qmc simulations for @xmath2 and a @xmath3 lattice .
the solid lines are the non - interacting results . from ref .
the low temperature downturn of d - wave however seems to come from a mistreatment of the sign problem ( d.j .
scalapino private communication).,width=283 ] after twenty years , we should be as quantitative as possible .
how should we proceed to investigate a model without a small parameter ? we will try to follow this path : ( 1 ) identify important physical principles and laws to constrain non - perturbative approximation schemes , starting from both weak ( kinetic energy dominated ) and strong ( potential energy dominated ) coupling .
( 2 ) benchmark the various approaches as much as possible against exact ( or numerically accurate ) results .
( 3 ) check that weak and strong coupling approaches agree at intermediate coupling .
( 4 ) compare with experiment . in brief
, we are trying to answer the question , is the hubbard model rich enough to contain the essential physics of the cuprates , both hole and electron doped ? "
the answer is made possible by new theoretical approaches , increased computing power , and the reassurance that theoretical approaches , numerical and analytical , give consistent results at intermediate coupling even if the starting points are very different .
this paper is a review of the work we have done in sherbrooke on this subject . in the short space provided , this review will not cover all of our work .
needless to say , we will be unfair to the work of many other groups , even though we will try to refer to the work of others that is directly relevant to ours .
we do not wish to make priority claims and we apologize to the authors that may feel unfairly treated .
section [ methodology ] will introduce the methodology : first a method that is valid at weak to intermediate coupling , the two - particle self - consistent approach ( tpsc ) , and then various quantum cluster methods that are better at strong coupling , namely cluster perturbation theory ( cpt ) , the variational cluster approximation ( vca ) also known as variational cluster perturbation theory ( vcpt ) , and cellular dynamical mean field theory ( cdmft ) with a brief mention of the dynamical cluster approximation ( dca ) . in all cases
, we will mention the main comparisons with exact or numerically accurate results that have been used to benchmark the approaches .
in sect . [ results ] we give some of the results , mostly on the pseudogap and the phase diagram of high - temperature superconductors .
more importantly perhaps , we show the consistency of the results obtained by both weak- and strong - coupling approaches when they are used at intermediate coupling .
finally , we compare with experiment in section [ experiment ] .
we consider the hubbard model@xmath4 where @xmath5 ( @xmath6 ) are creation and annihilation operators for electrons of spin @xmath7 , @xmath8 is the density of spin @xmath7 electrons , @xmath9 is the hopping amplitude , and @xmath10 is the on - site coulomb repulsion . in general , we write @xmath11 respectively for the first- , second- and third - nearest neighbor hopping amplitudes . in the following subsections ,
we first discuss how to approach the problem from the weak coupling perspective and then from the strong coupling point of view . the approaches that we will use in the end are non - perturbative , but in general they are more accurate either at weak or strong coupling . even at weak coupling
, the hubbard model presents difficulties specific to two dimensions .
the time - honored random phase approximation ( rpa ) has the advantage of satisfying conservation laws , but it violates the pauli principle and the mermin - wagner - hohenberg - coleman ( or mermin - wagner , for short ) theorem .
this theorem states that a continuous symmetry can not be broken at finite temperature in two dimensions .
rpa gives a finite - temperature phase transition .
the pauli principle means , in particular , that @xmath12 in a model with only one orbital per site .
this is violated by rpa since it can be satisfied only if all possible exchanges of electron lines are allowed ( more on this in the following section ) . since the square of the density at a given site is given by @xmath13 , violating the pauli condition @xmath14 will in general lead to large errors in double occupancy , a key quantity in the hubbard model since it is proportional to the potential energy .
another popular approach is the moriya@xcite self - consistent spin - fluctuation approach@xcite that uses a hubbard - stratonovich transformation and a @xmath15 factorization .
this satisfies the mermin - wagner theorem but , unfortunately , violates the pauli principle and introduces an unknown mode - coupling constant as well as an unknown renormalized @xmath10 in the second - order term . the conserving approximation known as fluctuation exchange ( flex )
approximation@xcite is an eliashberg - type theory that is conserving but violates the pauli principle , assumes a migdal theorem and does not reproduce the pseudogap phenomenon observed in qmc .
more detailed criticism of this and other approaches may be found in refs . . finally , the renormalization group@xcite has the great advantage of being an unbiased method to look for instabilities towards various ordered phases .
however , it is quite difficult to implement in two dimensions because of the proliferation of coupling constants , and , to our knowledge , no one has yet implemented a two - loop calculation without introducing additional approximations.@xcite such a two - loop calculation is necessary to observe the pseudogap phenomenon . the tpsc approach , originally proposed by vilk , tremblay and collaborators , @xcite aims at capturing non - perturbative effects .
it does not use perturbation theory or , if you want , it drops diagrammatic expansions .
instead , it is based on imposing constraints and sum rules : the theory should satisfy ( a ) the spin and charge conservation laws ( b ) the pauli principle in the form @xmath16 ( c ) the local - moment and the local - density sum rules . without any further explicit constraint , we find that the theory satisfies the mermin - wagner theorem , that it satisfies consistency between one- and two - particle quantities in the sense that @xmath17 and finally that the theory contains the physics of kanamori - brckner screening ( in other words , scattering between electrons and holes includes t - matrix quantum fluctuation effects beyond the born approximation ) .
several derivations of our approach have been given,@xcite including a quite formal one@xcite based on the functional derivative baym - kadanoff approach.@xcite here we only give an outline@xcite of the approach with a more phenomenological outlook .
we proceed in two steps . in the first step ( in our earlier work
sometimes called zeroth step ) , the self - energy is obtained by a hartree - fock - type factorization of the four - point function with the _ additional constraint _ that the factorization is exact when all space - time coordinates coincide . functional differentiation , as in the baym - kadanoff approach@xcite , then leads to a momentum- and frequency - independent irreducible particle - hole vertex for the spin channel that satisfies@xcite @xmath18 . the local moment sum rule and the pauli principle in the form @xmath19 then determine double occupancy and @xmath20 .
the irreducible vertex for the charge channel is too complicated to be computed exactly , so it is assumed to be constant and its value is found from the pauli principle and the local charge fluctuation sum rule . to be more specific ,
let us use the notation , @xmath21 and @xmath22 with @xmath23 and @xmath24 respectively bosonic and fermionic matsubara frequencies .
we work in units where @xmath25 @xmath26 and lattice spacing are all unity .
the spin and charge susceptibilities now take the form @xmath27 and @xmath28 with @xmath29 computed with the green function @xmath30 that contains the self - energy whose functional differentiation gave the vertices .
this self - energy is constant , corresponding to the hartree - fock - type factorization . as it would in the trivial hartree - fock factorization ( see previous note ) .
] the susceptibilities thus satisfy conservation laws@xcite .
one enforces the pauli principle @xmath19 implicit in the following two sum rules , @xmath31 where @xmath32 is the density .
the above equations , in addition to@xcite @xmath33 suffice to determine the constant vertices @xmath20 and @xmath34 .
once the two - particle quantities have been found as above , the next step of the approach of ref . ,
consists in improving the approximation for the single - particle self - energy by starting from an exact expression where the high - frequency hartree - fock behavior is explicitly factored out .
one then substitutes in the exact expression the irreducible low - frequency vertices @xmath20 and @xmath34 as well as @xmath35 and @xmath36 computed above .
the exact form for the self - energy expression can however be obtained either in the longitudinal or in the transverse channel . to satisfy crossing symmetry of the fully reducible vertex appearing in the general expression and to preserve consistency between one- and two - particle quantities , one averages the two possibilities to obtain @xcite@xmath37 g_{\sigma}^{(1)}(k+q).\end{aligned}\ ] ]
the resulting self - energy @xmath38 on the left hand - side is at the next level of approximation so it differs from the self - energy entering the right - hand side .
one can verify that the longitudinal spin fluctuations contribute an amount @xmath39 to the consistency condition@xcite
@xmath40 @xmath41 and that each of the two transverse spin components as well as the charge fluctuations also each contribute @xmath42 in addition , one verifies numerically that the exact sum rule@xcite @xmath43/\pi = u^{2}n_{-\sigma}(1-n_{-\sigma})$ ] determining the high - frequency behavior is satisfied to a high degree of accuracy .
the theory also has a consistency check .
indeed , the exact expression for consistency between one- and two - particle quantities should be written with @xmath44 given by @xmath45 instead of with @xmath46 .
in other words @xmath47 @xmath41 should be satisfied instead of @xmath48 @xmath49 which is exactly satisfied here .
we find through qmc benchmarks that when the left- and right - hand side of the last equation differ only by a few percent , then the theory is accurate .
to obtain the thermodynamics , one finds the entropy by integrating @xmath50 times the specific heat @xmath51 so that we know @xmath52 .
there are other ways to obtain the thermodynamics and one looks for consistency between these.@xcite we will not discuss thermodynamic aspects in the present review . at weak coupling in the repulsive model
the particle - hole channel is the one that is influenced directly .
correlations in crossed channels , such as pairing susceptibilities , are induced indirectly and are harder to evaluate .
this simply reflects the fact the simplest hartree - fock factorization of the hubbard model does not lead to a d - wave order parameter ( even though hartree - fock factorization of its strong - coupling version does ) .
the @xmath53-wave susceptibility is defined by @xmath54 with the @xmath55-wave order parameter equal to @xmath56 the sum over @xmath57 being over nearest - neighbors , with @xmath58 depending on whether @xmath57 is a neighbor along the @xmath59 or the @xmath60 axis .
briefly speaking,@xcite to extend tpsc to compute pairing susceptibility , we begin from the schwinger - martin - kadanoff - baym formalism with both diagonal@xcite and off - diagonal@xcite source fields .
the self - energy is expressed in terms of spin and charge fluctuations and the irreducible vertex entering the bethe - salpeter equation for the pairing susceptibility is obtained from functional differentiation .
the final expression for the @xmath55-wave susceptibility is , @xmath61 in the above expression , @xmath62 is the form factor for the gap symmetry , while @xmath63 and @xmath64 stand for both wave - vector and fermionic matsubara frequencies on a square - lattice with @xmath65 sites at temperature @xmath66 the spin and charge susceptibilities take the form @xmath67 and @xmath68 @xmath69 with @xmath29 computed with the green function @xmath30 that contains the self - energy whose functional differentiation gave the spin and charge vertices .
the values of @xmath70 @xmath34 and @xmath71 are obtained as described above . in the pseudogap regime
, one can not use @xmath72 .
instead,@xcite one uses the local - moment sum rule with the zero temperature value of @xmath73 obtained by the method of ref .
that agrees very well with qmc calculations at all values of @xmath74 also , @xmath75 contains self - energy effects coming from spin and charge fluctuations , as described above.@xcite the same principles and methodology can be applied for the attractive hubbard model.@xcite in that case , the dominant channel is the s - wave pairing channel .
correlations in the crossed channel , namely the spin and charge susceptibilities , can also be obtained _ mutatis mutandi _ along the lines of the previous paragraph . to test any non - perturbative approach , we need reliable benchmarks .
quantum monte carlo ( qmc ) simulations provide such benchmarks .
the results of such numerical calculations are unbiased and they can be obtained on much larger system sizes than any other simulation method .
the statistical uncertainty can be made as small as required .
the drawback of qmc is that the sign problem renders calculations impossible at temperatures low enough to reach those that are relevant for d - wave superconductivity .
nevertheless , qmc can be performed in regimes that are non - trivial enough to allow us to eliminate some theories on the grounds that they give qualitatively incorrect results .
comparisons with qmc allow us to estimate the accuracy of the theory .
an approach like tpsc can then be extended to regimes where qmc is unavailable with the confidence provided by agreement between both approaches in regimes where both can be performed
. as a function of filling and for various values of @xmath10 expect for @xmath2 where the dashed line shows the results of our theory at the crossover temperature @xmath76 . from ref .
.,width=188 ] ) dependence of the spin and charge structure factors for different sets of parameters .
solid lines are from tpsc and symbols are our qmc data .
monte carlo data for @xmath77 and @xmath78 are for @xmath79 clusters and @xmath80 ; all other data are for @xmath81 clusters and @xmath82 .
error bars are shown only when significant . from ref .
.,width=264 ] at half - filling @xmath77 .
the solid line is from tpsc and symbols are monte carlo data from ref . . taken from ref .
.,width=207 ] as a function of temperature at zero matsubara frequency .
the filled circles ( bws ) are from ref . .
taken from ref .
.,width=207 ] in order to be concise ,
details are left to figure captions .
let us first focus on quantities related to spin and charge fluctuations .
the symbols on the figures refer to qmc results while the solid lines come from tpsc calculations .
[ fig_25 ] shows double occupancy , a quantity that plays a very important role in the hubbard model in general and in tpsc in particular .
that quantity is shown as a function of filling for various values of @xmath10 at inverse temperature @xmath83 .
[ fig_26 ] displays the spin and charge structure factors in a regime where size effects are not important .
clearly the results are non - perturbative given the large difference between the spin and charge structure factors , which are plotted here in units where they are equal at @xmath84 . in fig .
[ fig_27 ] we exhibit the static structure factor at half - filling as a function of temperature . below the crossover temperature @xmath85
, there is an important size dependence in the qmc results .
the tpsc calculation , represented by a solid line , is done for an infinite system .
we see that the mean - field finite transition temperature @xmath86 is replaced by a crossover temperature @xmath85 at which the correlations enter an exponential growth regime
. one can show analytically@xcite that the correlation length becomes infinite only at zero temperature , thus satisfying the mermin - wagner theorem .
the qmc results approach the tpsc results as the system size grows .
nevertheless , tpsc is in the @xmath87 universality class@xcite contrary to the hubbard model for which @xmath88 so one expects quantitative differences to increase as the correlation length becomes larger .
it is important to note that @xmath85 does not coincide with the mean - field transition temperature @xmath86 .
this is because of kanamori - brueckner screening@xcite that manifests itself in the difference between @xmath20 and the bare @xmath10 .
below @xmath85 , the main contribution to the static spin structure factor in fig .
[ fig_27 ] comes from the zero - matsubara frequency component of the spin susceptibility .
this is the so - called renormalized classical regime where the characteristic spin fluctuation frequency @xmath89 is much less than temperature . even at temperatures higher than that
, tpsc agrees with qmc calculation much better than other methods , as shown in fig .
[ fig_28 ] . for @xmath2 , @xmath90 , @xmath77 , and all independent wave vectors @xmath91 of an @xmath92 lattice .
results obtained from maximum entropy inversion of qmc data on the left panel and many - body tpsc calculations with eq .
( [ self - long ] ) on the middle panel and with flex on the right panel . from ref .
.,width=245 ] , @xmath90 , @xmath77 , @xmath93 , @xmath94 .
upper panels show @xmath95 extracted from maximum entropy on @xmath96 shown on the corresponding lower panels .
( a ) qmc .
( b ) tpsc using eq .
( [ self - long ] ) .
( c ) flex . from ref .
.,width=283 ] below the crossover temperature to the renormalized classical regime , a pseudogap develops in the single - particle spectral weight .
this is illustrated in fig .
[ fig_29].@xcite eliashberg - type approaches such as flex do not show the pseudogap present in qmc .
the size dependence of the results is also quite close in tpsc and in qmc , as shown in fig .
[ fig_30 ] .
susceptibility obtained from qmc simulations ( symbols ) and from the tpsc approach ( lines ) in the two - dimensional hubbard model .
both calculations are for @xmath2 , a @xmath97 lattice .
qmc error bars are smaller than the symbols .
analytical results are joined by solid lines .
the size dependence of the results is small at these temperatures .
the @xmath84 case is also shown at @xmath98 as the upper line .
the inset compares qmc and flex at @xmath99 . from ref
.,width=245 ] the d - wave susceptibility@xcite shown in fig .
[ fig_12 ] again clearly demonstrates the agreement between tpsc and qmc . in particular ,
the dome shape dependence of the qmc results is reproduced to within a few percent .
we will see in sec .
[ results ] how one understands the dome shape and the fact that the d - wave susceptibility of the interacting system is smaller than that of the non - interacting one in this temperature range . from tpsc ( left ) and from a temperature cutoff renormalization group technique from ref .
( right ) .
the corresponding van hove filling is indicated on the upper horizontal axis .
crossover lines for magnetic instabilities near the antiferromagnetic and ferromagnetic wave vectors are represented by filled symbols while open symbols indicate instability towards @xmath53-wave superconducting .
the solid and dashed lines below the empty symbols show , respectively for @xmath100 and @xmath101 , where the antiferromagnetic crossover temperature would have been in the absence of the superconducting instability .
the largest system size used for this calculation is @xmath102 . from ref .
.,width=321 ] ( filled triangles ) and qmc @xmath103 ( open circles ) for @xmath104 and various temperatures ( a ) at @xmath105 and ( b ) at @xmath106 on a @xmath107 lattice .
the dashed lines are to guide the eye . from ref .
.,width=226 ] ( open diamonds ) and @xmath108 ( open squares ) with the results of qmc calculations ( open circles ) for @xmath104 .
right : the momentum dependent occupation number @xmath109 .
circles : qmc calculations from ref . .
the solid curve : tpsc . the dashed curve obtained by replacing @xmath110 by @xmath10 in the self - energy with
all the rest unchanged .
the long - dash line is the result of a self - consistent t - matrix calculation , and the dot - dash line the result of second - order perturbation theory . from ref .
.,width=321 ] lattice .
qmc data for the density of states taken from ref . .
figures from ref .
.,width=321 ] to conclude this section , we quickly mention a few other results obtained with tpsc .
[ fig_32 ] contrasts the crossover phase diagram obtained for the hubbard model at the van hove filling@xcite with the results of a renormalization group calculation.@xcite the difference occurring in the ferromagnetic region is discussed in detail in ref . . finally , we point out various comparisons for the attractive hubbard model .
[ fig_33 ] shows the static s - wave pairing susceptibility , fig .
[ fig_34 ] the chemical potential and the occupation number , and finally fig .
[ fig_35 ] the local density of states and the single - particle spectral weight at a given wave vector .
dmft@xcite has been extremely successful in helping us understand the mott transition , the key physical phenomenon that manifests itself at strong coupling . however , in high dimension where this theory becomes exact , spatial fluctuations associated with incipient order do not manifest themselves in the self - energy . in low dimension ,
this is not the case .
the self - energy has strong momentum dependence , as clearly shown experimentally in the high - temperature superconductors , and theoretically in the tpsc approach , a subject we shall discuss again below .
it is thus necessary to go beyond dmft by studying clusters instead of a single anderson impurity as done in dmft .
the simplest cluster approach that includes strong - coupling effects and momentum dependence is cluster perturbation theory ( cpt).@xcite in this approach , an infinite set of disconnected clusters are solved exactly and then connected to each other using strong - coupling perturbation theory .
although the resulting theory turns out to give the exact result in the @xmath84 case , its derivation clearly shows that one expects reliable results mostly at strong coupling .
this approach does not include the self - consistent effects contained in dmft .
self - consistency or clusters was suggested in ref . and a causal approach was first implemented within dca,@xcite where a momentum - space cluster is connected to a self - consistent momentum - space medium . in our opinion ,
the best framework to understand all other cluster methods is the self - energy functional approach of potthoff.@xcite the form of the lattice green function obtained in this approach is the same as that obtained in cpt , clearly exhibiting that such an approach is better at strong - coupling , even though results often extrapolate correctly to weak coupling . amongst the special cases of this approach , the variational cluster approach ( vca ) , or variational cluster perturbation theory ( vcpt)@xcite is the one closest to the original approach . in a variant , cellular dynamical mean field theory@xcite ( cdmft ) ,
a cluster is embedded in a self - consistent medium instead of a single anderson impurity as in dmft ( even though the latter approach is accurate in many realistic cases , especially in three dimensions ) .
the strong - coupling aspects of cdmft come out clearly in refs . .
a detailed review of quantum cluster methods has appeared in ref . .
even though cpt does not have the self - consistency present in dmft type approaches , at fixed computing resources it allows for the best momentum resolution .
this is particularly important for the arpes pseudogap in electron - doped cuprates that has quite a detailed momentum space structure , and for d - wave superconducting correlations where the zero temperature pair correlation length may extend well beyond near - neighbor sites .
cpt was developed by gros@xcite and snchal@xcite independently .
this approach can be viewed as the first term of a systematic expansion around strong coupling.@xcite let us write the hopping matrix elements in the form @xmath111 where @xmath112 and @xmath32 label the different clusters , and @xmath113 label the sites within a cluster
. then @xmath114 labels all the hopping matrix elements within a cluster and the above equation defines @xmath115 .
we pause to introduce the notation that will be used throughout for quantum cluster methods .
we follow the review article ref . .
in reciprocal space , any wave vector @xmath91 in the brillouin zone may be written as @xmath116 where both @xmath91 and @xmath117 are continuous in the infinite size limit , except that @xmath117 is defined only in the reduced brillouin zone that corresponds to the superlattice .
on the other hand , @xmath118 is discrete and denotes reciprocal lattice vectors of the superlattice . by analogy ,
any position @xmath119 in position space can be written as @xmath120 where @xmath121 is for positions within clusters while @xmath122 labels the origins of the clusters , an infinite number of them .
hence , fourier s theorem allows one to define functions of @xmath123 @xmath117 or @xmath118 that contain the same information as functions of , respectively , @xmath124 @xmath122 or @xmath125 also , we have @xmath126 @xmath127 where @xmath32 is an integer .
sites within a cluster are labelled by greek letters so that the position of site @xmath128 within a cluster is @xmath129 , while clusters are labelled by latin letters so that the origin of cluster @xmath112 is at @xmath130 . returning to cpt
, the green function for the whole system is given by @xmath131 _ { \mu\nu}=\left [ \hat{g}^{(c)-1}(z)-\hat{v}(\tilde{\mathbf{k}})\right ] _ { \mu\nu } \label{gsuperlattice}%\ ] ] where hats denote matrices in cluster site indices and @xmath132 is the complex frequency . at this level of approximation
, the cpt green function has the same structure as in the hubbard i approximation except that it pertains to a cluster instead of a single site . since @xmath133 and @xmath134 @xmath135 the green function ( [ gsuperlattice ] ) may also be written as @xmath136 this form allows a different physical interpretation of the approach . in the above expression ,
the self - energy of the lattice is approximated by the self - energy of the cluster .
the latter in real space spans only the size of the cluster .
we still need an expression to extend the above result to the lattice in a translationally invariant way .
this is done by defining the following residual fourier transform : @xmath137 notice that @xmath138 may be replaced by @xmath139 in the above equation since @xmath140 .
the self - energy functional approach , devised by potthoff@xcite allows one to consider various cluster schemes from a unified point of view .
it begins with @xmath141,$ ] a functional of the green function@xmath142=\phi[g]-\mathrm{tr}((g_{0\mathbf{t}}^{-1}% -g^{-1})g)+\mathrm{tr}\ln(-g ) .
\label{grandpotential}%\ ] ] the luttinger ward functional @xmath143 $ ] entering this equation is the sum of connected vacuum skeleton diagrams . a diagram - free definition of this functional is also given in ref . .
for our purposes , what is important is that ( 1 ) the functional derivative of @xmath143 $ ] is the self - energy@xmath144}{\delta g}=\sigma\label{selfluttinger}%\ ] ] and ( 2 ) it is a universal functional of @xmath145 in the following sense : whatever the form of the one - body hamiltonian , it depends only on the interaction and , functionnally , it has the same dependence on @xmath145 .
the dependence of the functional @xmath146 $ ] on the one - body part of the hamiltonian is denoted by the subscript @xmath147 and it comes only through @xmath148 appearing on the right - hand side of eq .
( [ grandpotential ] ) .
the functional @xmath141 $ ] has the important property that it is stationary when @xmath145 takes the value prescribed by dyson s equation .
indeed , given the last two equations , the euler equation takes the form@xmath149}{\delta g}=\sigma - g_{0\mathbf{t}}% ^{-1}+g^{-1}=0.\ ] ] this is a dynamic variational principle since it involves the frequency appearing in the green function , in other words excited states are involved in the variation . at this stationary point , and only there , @xmath146 $ ] is equal to the grand potential .
contrary to ritz s variational principle , this last equation does not tell us whether @xmath146 $ ] is a minimum or a maximum or a saddle point there .
there are various ways to use the stationarity property that we described above .
the most common one , is to approximate @xmath143 $ ] by a finite set of diagrams .
this is how one obtains the hartree - fock , the flex approximation@xcite or other so - called thermodynamically consistent theories .
this is what potthoff calls a type ii approximation strategy.@xcite a type i approximation simplifies the euler equation itself . in a type iii approximation
, one uses the exact form of @xmath150 $ ] but only on a limited domain of trial green functions .
following potthoff , we adopt the type iii approximation on a functional of the self - energy instead of on a functional of the green function .
suppose we can locally invert eq .
( [ selfluttinger ] ) for the self - energy to write @xmath145 as a functional of @xmath151 we can use this result to write , @xmath152=f[\sigma]-\mathrm{tr}\ln(-g_{0\mathbf{t}}% ^{-1}+\sigma).\ ] ] where we defined @xmath153=\phi[g]-\mathrm{tr}(\sigma g).\ ] ] and where it is implicit that @xmath154 $ ] is now a functional of @xmath155 .
@xmath156,$ ] along with the expression ( [ selfluttinger ] ) for the derivative of the luttinger - ward functional , define the legendre transform of the luttinger - ward functional .
it is easy to verify that@xmath157}{\delta\sigma}=\frac{\delta\phi[g]}{\delta
g}\frac{\delta g[\sigma]}{\delta\sigma}-\sigma\frac{\delta g[\sigma]}% { \delta\sigma}-g =- g\ ] ] hence , @xmath158 $ ] is stationary with respect to @xmath155 when dyson s equation is satisfied@xmath159}{\delta\sigma}=-g+(g_{0\mathbf{t}}% ^{-1}-\sigma)^{-1}=0.\ ] ] to perform a type iii approximation on @xmath156 $ ] , we take advantage that it is universal , i.e. , that it depends only on the interaction part of the hamiltonian and not on the one - body part .
this follows from the universal character of its legendre transform @xmath143 $ ] .
we thus evaluate @xmath156 $ ] exactly for a hamiltonian @xmath160 that shares the same interaction part as the hubbard hamiltonian , but that is exactly solvable .
this hamiltonian @xmath160 is taken as a cluster decomposition of the original problem , i.e. , we tile the infinite lattice into identical , disconnected clusters that can be solved exactly . examples of such tilings are given in fig .
[ fig_42 ] . denoting the corresponding quantities with a prime , we obtain , @xmath161=f[\sigma^{\prime}]-\mathrm{tr}% \ln(-g_{0\mathbf{t}^{\prime}}^{-1}+\sigma^{\prime}).\ ] ] from which we can extract @xmath162 $ ] .
it follows that @xmath163=\omega_{\mathbf{t}^{\prime}}% [ \sigma^{\prime}]+\mathrm{tr}\ln(-g_{0\mathbf{t}^{\prime}}^{-1}+\sigma ^{\prime})-\mathrm{tr}\ln(-g_{0\mathbf{t}}^{-1}+\sigma^{\prime } ) .
\label{sef_eq}%\ ] ] the type iii approximation comes from the fact that the self - energy @xmath164 is restricted to the exact self - energy of the cluster problem @xmath160 , so that variational parameters appear in the definition of the one - body part of @xmath160 . in practice
, we look for values of the cluster one - body parameters @xmath165 such that @xmath166/\delta\mathbf{t}^{\prime}=0 $ ] .
it is useful for what follows to write the latter equation formally , although we do not use it in actual calculations . given that @xmath167 $ ] is the actual grand potential evaluated for the cluster , @xmath168/\partial\mathbf{t}^{\prime}$ ] is canceled by the explicit @xmath165 dependence of @xmath169 and we are left with @xmath170}{\delta\sigma^{\prime } } \frac{\delta\sigma^{\prime}}{\delta\mathbf{t}^{\prime}}\nonumber\\ & = -\mathrm{tr}\left[\left ( \frac{1}{g_{0\mathbf{t}^{\prime}}^{-1}% -\sigma^{\prime}}-\frac{1}{g_{0\mathbf{t}}^{-1}-\sigma^{\prime}}\right ) \frac{\delta\sigma^{\prime}}{\delta\mathbf{t}^{\prime}}\right ] .\end{aligned}\ ] ] given that the clusters corresponding to @xmath165 are disconnected and that translation symmetry holds on the superlattice of clusters , each of which contains @xmath171 sites , the last equation may be written @xmath172\frac{\delta\sigma_{\nu\mu}^{\prime}(i\omega_{n})}{\delta \mathbf{t}^{\prime}}=0 .
\label{eulervca}%\end{aligned}\ ] ] in variational cluster perturbation theory ( vcpt ) , more aptly named the variational cluster approach ( vca ) , solutions to the euler equations ( [ eulervca ] ) are found by looking for numerical minima ( or more generally , saddle - points ) of the functional . typically , the vca cluster hamiltonian @xmath160 will have the same form as @xmath173 except that there is no hopping between clusters and that long - range order is allowed by adding some weiss fields , for instance like in eq .
( [ weiss_eq ] ) below .
the hopping terms and chemical potential within @xmath160 may also be treated like additional variational parameters .
in contrast with mean - field theory , these weiss fields are not mean fields , in the sense that they do not coincide with the corresponding order parameters .
the interaction part of @xmath173 ( or @xmath173 ) is not factorized in any way and short - range correlations are treated exactly .
in fact , the hamiltonian @xmath173 is not altered in any way ; the weiss fields are introduced to let the variational principle act on a space of self - energies that includes the possibility of specific long - range orders , without imposing those orders .
indeed , the more naturally an order arises in the system , the smaller the weiss field needs to be , and one observes that the strength of the weiss field at the stationary point of the self - energy functional generally decreases with increasing cluster size , as it should since in the thermodynamic limit no weiss field should be necessary to establish order .
the cellular dynamical mean - field theory ( cdmft ) is obtained by including in the cluster hamiltonian @xmath160 a bath of uncorrelated electrons that somehow must mimic the effect on the cluster of the rest of the lattice .
explicitly , @xmath160 takes the form @xmath174 where @xmath175 annihilates an electron of spin @xmath7 on a bath orbital labelled @xmath176 .
the bath is characterized by the energy of each orbital ( @xmath177 ) and the bath - cluster hybridization matrix @xmath178 .
this representation of the environment through an anderson impurity model was introduced in ref . in the context of dmft ( i.e. , a single site ) .
the effect of the bath on the electron green function is encapsulated in the so - called hybridization function @xmath179 which enters the green function as @xmath180_{\mu\nu}=\omega+\mu - t_{\mu\nu}^{\prime}-\gamma_{\mu\nu } ( \omega)-\sigma_{\mu\nu}(\omega).\ ] ] moreover , the cdmft does not look for a strict solution of the euler equation ( [ eulervca ] ) , but tries instead to set each of the terms between brackets to zero separately . since the euler equation ( [ eulervca ] ) can be seen as a scalar product , cdmft requires that the modulus of one of the vectors vanish to make the scalar product vanish . from a heuristic point of view
, it is as if each component of the green function in the cluster were equal to the corresponding component deduced from the lattice green function .
this clearly reduces to single site dmft when there is only one lattice site . when the bath is discretized , i.e. ,
is made of a finite number of bath orbitals , the left - hand side of eq .
( [ eulervca ] ) can not vanish separately for each frequency , since the number of degrees of freedom in the bath is insufficient .
instead , one adopts the following self - consistent scheme : ( 1 ) one starts with a guess value of the bath parameters @xmath181 and solves the cluster hamiltonian @xmath160 numerically .
( 2 ) one then calculates the combination @xmath182 ^{-1}+\hat\sigma^{\prime}(i\omega_{n})\ ] ] and ( 3 ) minimizes the following canonically invariant distance function : @xmath183 over the set of bath parameters ( changing the bath parameters at this step does not require a new solution of the hamiltonian @xmath160 , but merely a recalculation of the hybridization function @xmath184 ) .
the bath parameters obtained from this minimization are then put back into step ( 1 ) and the procedure is iterated until convergence . in practice ,
the distance function ( [ dist_func ] ) can take various forms , for instance by adding a frequency - dependent weight in order to emphasize low - frequency properties@xcite or by using a sharp frequency cutoff.@xcite these weighting factors can be considered as rough approximations for the missing factor @xmath185 in the euler equation ( [ eulervca ] ) .
the frequencies are summed over on a discrete , regular grid along the imaginary axis , defined by some fictitious inverse temperature @xmath186 , typically of the order of 20 or 40 ( in units of @xmath187 ) .
even when the total number of cluster plus bath sites in cdmft equals the number of sites in a vca calculation , cdmft is much faster than the vca since the minimization of a grand potential functional requires many exact diagonalizations of the cluster hamiltonian @xmath188 .
the final lattice green function from which one computes observable quantities may be obtained by periodizing the self - energy , as in ref . or in the cpt manner described above in eq .
( [ gcpt ] ) .
we prefer the last approach because it corresponds to the green function needed to obtain the density from @xmath189 and also because periodization of the self - energy gives additional unphysical states in the mott gap@xcite ( see also ref . ) .
the dca@xcite can not be formulated within the self - energy functional approach .
it is based on the idea of discretizing irreducible quantities , such as the energy , in reciprocal space .
it is believed to converge faster for @xmath190 quantities whereas cdmft converges exponentially fast for local quantities.@xcite since dmft becomes exact in infinite dimension , the most difficult challenge for cluster extensions of this approach is in one dimension .
in addition , exact results to compare with exist only in one dimension so it is mostly in @xmath191 that cluster methods have been checked . in @xmath192
there have also been a few comparisons with qmc as we shall discuss .
limit of the one- dimensional hubbard model , as calculated from ( a ) an exact diagonalization of the hubbard model with @xmath193 on a periodic 12- site cluster ; ( b ) the same , but with cpt , on a 12-site cluster with open boundary conditions ; ( c ) the exact solution , taken from ref . ; beware : the axes are oriented differently . in ( a ) and ( b )
a finite width @xmath194 has been given to peaks that would otherwise be dirac @xmath195-functions.,width=264 ] , for @xmath196 and various cluster sizes @xmath197 ( connected symbols ) .
for comparison , the exact diagonalization values of finite clusters with periodic boundary conditions are also shown ( dashed lines ) for @xmath198 and @xmath199 .
bottom : same for the double occupancy .
an extrapolation of the results to infinite cluster size ( @xmath200 ) using a quadratic fit in terms of @xmath201 is also shown , and is accurate to within 0.5% . taken from ref .
.,width=245 ] cpt has been checked@xcite for example by comparing with exact results@xcite for the spectral function at @xmath202 in @xmath191 as shown in fig .
[ fig_53a ] .
[ fig_53b ] shows the chemical potential as a function of density for various values of @xmath10 .
[ fig_54 ] shows the convergence rates for the total energy and for the double occupancy in the @xmath191 half - filled model .
clearly , there is a dramatic improvement compared with exact diagonalizations . ) as a function of @xmath10 , as obtained from various methods : exact diagonalization ( ed ) , cpt and vcpt on a 10-site cluster , quantum monte carlo ( qmc ) and variational monte carlo ( vmc ) . taken from ref .
.,width=207 ] the main weakness of cpt is that it can not take into account tendency towards long - range order .
this is remedied by vcpt , as shown in fig .
[ fig_55b ] where cpt , vcpt are both compared with qmc as a benchmark . despite this agreement
, we should stress that long wave length fluctuations are clearly absent from cluster approaches .
hence , the antiferromagnetic order parameter at half - filling , for example , does not contain the effect of zero - point long wave length transverse spin fluctuations .
this is discussed for example in the context of fig .
9 of ref .
. cluster with @xmath203 bath sites of the density as a function of the chemical potential in the one - dimensional hubbard model for @xmath2 , as compared with the exact solution , dmft and other approximation schemes . taken from ref .
.,width=245 ] cdmft corrects the difficulties of cpt near half - filling by reproducing the infinite compressibility predicted by the bethe ansatz in one dimension as shown in fig .
[ fig_56].@xcite detailed comparisons between the local and near - neighbor green functions@xcite have been performed .
one should note that these results , obtained from exact diagonalization , also need the definition of a distance function ( see eq .
( [ dist_func ] ) above ) that helps find the best bath parametrization to satisfy the self - consistency condition .
this measure forces one to define calculational parameters such as a frequency cutoff and an fictitious temperature defining the matsubara frequencies to sum over .
the final results are not completely insensitive to the choice of fictitious temperature or frequency - weighing scheme but are usually considered reliable and consistent with each other when @xmath186 lies between @xmath204 and @xmath205 .
the cutoff procedures have been discussed in ref . .
the relative merits of dca and cdmft have been discussed for example in refs .
briefly speaking , convergence seems faster in dca for long wave length quantities but cdmft is faster ( exponentially ) for local quantities .
in this section , we outline the main results we obtained concerning the pseudogap and d - wave superconductivity in the two - dimensional hubbard model .
quantum cluster approaches are better at strong coupling while tpsc is best at weak coupling .
nevertheless , all these methods are non - perturbative , the intermediate coupling regime presenting the physically most interesting case .
but it is also the regime where we have the least control over the approximations . as we will show
, it is quite satisfying that , at intermediate coupling , weak - coupling and strong - coupling approaches give results that are nearly in quantitative agreement with each other .
this gives us great confidence into the validity of the results .
as an example of concordance , consider the fact that to obtain spectral weight near @xmath206 at optimal doping in the electron - doped systems , @xmath10 has to be roughly @xmath207 . for larger @xmath208 ( @xmath78 in cpt ) that weight , present in the experiments , disappears .
smaller values of @xmath10 ( @xmath2 in cpt ) do not lead to a pseudogap .
other examples of concordance include the value of the superconducting transition temperature @xmath209 obtained with dca and with tpsc as well as the temperature dependence of souble occupancy obtained with the same two methods . to understand the pseudogap it is most interesting to consider both hole and electron - doped cuprates at once .
this means that we have to include particle - hole symmetry breaking hoppings , @xmath210 and @xmath211 .
we will see in the present section that it is possible to obtain a pseudogap at strong coupling without a large correlation length in the particle - hole or in the particle - particle channels .
by contrast , at weak coupling one does need a long - correlation length and low dimension .
so there appears to be theoretically two different mechanisms for the pseudogap . in units of @xmath212 , for wave vectors along the high - symmetry directions shown in the inset .
( a ) cpt calculations on a @xmath213 cluster with ten electrons ( 17% hole doped ) .
( b ) the same as ( a ) , with 14 electrons ( 17% electron doped ) . in all cases @xmath214 and @xmath215 .
a lorentzian broadening @xmath216 is used to reveal the otherwise delta peaks . from ref .
.,width=245 ] corresponding to fig .
[ fig_62 ] , taken from ref . .
hole - doped case on the left , electron - doped case on the right panel , width=245 ] the top panel in fig .
[ fig_62 ] presents the single - particle spectral weight , @xmath95 or imaginary part of the single - particle green function , for the model with @xmath217 @xmath215 in the hole - doped case , for about @xmath218 doping.@xcite each curve is for a different wave vector ( on a trajectory shown in the inset ) and is plotted as a function of frequency in units of @xmath212 .
this kind of plot is known as energy dispersion curves ( edc ) .
it is important to point out that the theoretical results are obtained by broadening a set of delta function , so that the energy resolution is @xmath216 corresponding roughly to the experimental resolution we will compare with in the next section . at small @xmath196 on the top panel of fig .
[ fig_62 ] , one recovers a fermi liquid . at large @xmath10 , say @xmath78 , the mott gap at positive energy is a prominent feature .
the pseudogap is a different feature located around the fermi energy . to see it better , we present on the left - hand panel of fig . [ fig_63 ] a blow - up in the vicinity of the fermi surface crossing occurring near @xmath219 .
clearly , there is a minimum in @xmath95 at the fermi - surface crossing when @xmath10 is large enough instead of a maximum like in fermi liquid theory .
-@xmath214 , @xmath215 hubbard model , taken from ref . . ,
width=245 ] it is also possible to plot @xmath95 at fixed frequency for various momenta .
they are so - called momentum dispersion curves ( mdc ) . in fig .
[ fig_64 ] we take the fermi energy @xmath220 and we plot the magnitude of the single - particle spectral weight in the first quadrant of the brillouin zone using red for high - intensity and blue for low intensity .
the figure shows that , in the hole - doped case ( top panel ) , weight near @xmath206 survives while it tends to disappear near @xmath219 and @xmath221 .
that pseudogap phenomenon is due not only to large @xmath10 but also to the fact that the line that can be drawn between the points @xmath219 and @xmath221 crosses the fermi surface . when there is no such crossing , one recovers a fermi surface ( not shown here ) .
the @xmath219 to @xmath221 line has a double role .
it is the antiferromagnetic zone boundary , as well as the line that indicates where umklapp processes become possible , i.e. , the line where we can scatter a pair of particles on one side of the fermi surface to the other side with loss or gain of a reciprocal lattice vector .
large scattering rates explain the disappearance of the fermi surface.@xcite we also note that the size of the pseudogap in cpt , defined as the distance between the two peaks , does not scale like @xmath222 at large coupling .
it seems to be very weakly @xmath10 dependent , its size being related to @xmath212 instead .
this result is corroborated by cdmft.@xcite -@xmath210-@xmath211 hubbard model , calculated on a 16-site cluster in cpt , at @xmath223 .
inset : the pseudogap . left : the corresponding fermi energy momentum distribution curve.,width=321 ] the edc for the electron - doped case is shown on the bottom panels of fig .
[ fig_62 ] near optimal doping again .
this time , the mott gap appears below the fermi surface so that the lower hubbard band becomes accessible to experiment .
the edc in fig .
[ fig_71 ] shows very well both the mott gap and the pseudogap .
details of that pseudogap can be seen both in the inset of fig .
[ fig_71 ] or on the right - hand panel of fig .
[ fig_63 ] . while in the hole - doped case the mdc appeared to evolve continuously as we increase @xmath10 ( top panel of fig .
[ fig_64 ] ) , in the electron - doped case ( bottom panel ) the weight initially present near @xmath224 at @xmath2 disappears by the time we reach @xmath78 .
-@xmath210-@xmath211 hubbard model , with @xmath225 and @xmath226 , calculated on a 8-site cluster for @xmath78 in vcpt .
d - wave superconductivity is present in the holde - doped case ( left ) and both antiferromagnetism and d - wave superconductivity in the electron - doped case .
the resolution is not large enough in the latter case to see the superconducting gap .
the lorentzian broadening is @xmath227 . from ref .
, width=245 ] in fig .
[ fig_110 ] we show , with the same resolution as cpt , the mdc for vcpt.@xcite in this case the effect of long - range order is included and visible but , at this resolution , the results are not too different from those obtained from cpt in fig .
[ fig_64 ] .
-@xmath210 , @xmath78 hubbard model , calculated on a 4-site cluster in cdmft . energy resolution , @xmath228 ( left and middle ) .
left : hole - doped dsc ( @xmath229 ) , middle : electron - doped dsc ( @xmath230 ) , right : same as middle with @xmath231 .
note the particle - hole transformation in the electron - doped case . from ref .
, width=283 ] cdmft@xcite gives mdc that , at comparable resolution , @xmath228 , are again compatible with cpt and with vcpt . the middle panel in fig .
[ fig_75 ] is for the electron - doped case but with a particle - hole transformation so that @xmath232 and @xmath233 .
since there is a non - zero d - wave order parameter in this calculation , improving the resolution to @xmath231 reveals the d - wave gap , as seen in the right most figure .
-@xmath234 , @xmath78 hubbard model , calculated on a 4-site cluster in cdmft .
top : normal ( paramagnetic ) state for various densities .
bottom : same for the antiferromagnetic state . from ref . .
, width=321 ] , @xmath234 calculated in cpt , vcpt and qmc . from ref . .
, width=245 ] it has been argued for a while in dca that there is a mechanism whereby short - range correlations at strong coupling can be the source of the pseudogap phenomenon.@xcite to illustrate this mechanism in cdmft , we take the case @xmath235 and compare in fig .
[ fig_81 ] the edc for @xmath78 without long - range order ( top panels ) and with long - range antiferromagnetic order ( bottom panels).@xcite the four bands appearing in figs [ fig_81]a and [ fig_81]d are in agreement with what has been shown@xcite with cpt , vcpt and qmc in fig .
[ fig_55a ] .
evidently there are additional symmetries in the antiferromagnetic case .
the bands that are most affected by the long - range order are those that are closest to the fermi energy , hence they reflect spin correlations , while the bands far from the fermi energy seem less sensitive to the presence of long - range order .
these far away bands are what is left from the atomic limit where we have two dispersionless bands . as we dope , the chemical potential moves into the lower band closest to the fermi energy .
when there is no long - range order ( figs [ fig_81]b and [ fig_81]c ) the lower band closest to the fermi energy moves very close to it , at the same time as the upper band closest to the fermi energy looses weight , part of it moving closer to the fermi energy .
these two bands leave a pseudogap at the fermi energy@xcite , although we can not exclude that increasing the resolution would reveal a fermi liquid at a very small energy scale . in the case
when there is long - range antiferromagnetic order , ( figs [ fig_81]e and [ fig_81]f ) the upper band is less affected while the chemical potential moves in the lower band closest to the fermi energy but without creating a pseudogap , as if we were doping an itinerant antiferromagnet .
it seems that forcing the spin correlations to remain short range leads to the pseudogap phenomenon in this case .
when a pseudogap appears , it is created again by very large scattering rates.@xcite , in the two - dimensional hubbard model for @xmath77 , as calculated from tpsc ( lines with x ) and from dca ( symbols ) from ref . .
taken from ref .
.,width=245 ] in hartree - fock theory , double occupancy is given by @xmath236 and is independent of temperature .
the correct result does depend on temperature .
one can observe in fig .
[ fig_83 ] the concordance between the results for the temperature - dependent double occupancy obtained with dca and with tpsc@xcite for the @xmath235 model .
we have also done extensive comparisons between straight qmc calculations and tpsc.@xcite the downturn at low temperature has been confirmed by the qmc calculations .
it comes from the opening of the pseudogap due to antiferromagnetic fluctuations , as we will describe below .
the concomitant increase in the local moment corresponds to the decrease in double - occupancy .
there seems to be a disagreement at low temperature between tpsc and dca at @xmath237 in fact tpsc is closer to the direct qmc calculation . since we expect quantum cluster methods in general and dca in particular to be less accurate at weak coupling , this is not too worrisome . at @xmath2 the density of states obtained with tpsc and with dca at various temperatures
are very close to each other.@xcite we stress that as we go to temperatures well below the pseudogap , tpsc becomes less and less accurate , generally overemphasizing the downfall in double occupancy .
, @xmath238 at various hole dopings , as obtained from tpsc .
the far left from ref . is the fermi surface plot for @xmath239 hole - doped ca@xmath240na@xmath241cuo@xmath242cl@xmath242.,width=321 ] and two different @xmath10 s , @xmath243 and @xmath244 obtained from tpsc .
the first column is the corresponding experimental plots at @xmath245 and @xmath246 doping in ref . .
from refs .
.,width=321 ] we will come back to more details on the predictions of tpsc for the pseudogap , but to illustrate the concordance with quantum cluster results shown in the previous subsection , we show in fig .
[ fig_84 ] mdc obtained at the fermi energy in the hole doped case for @xmath247 , @xmath248 .
again there is quasi - particle weight near @xmath224 and a pseudogap near @xmath219 and @xmath221 .
however , as we will discuss below , the antiferromagnetic correlation length necessary to obtain that pseudogap is too large compared with experiment .
the electron - doped case is shown in fig .
[ fig_85 ] near optimal doping and for different values of @xmath74 as @xmath10 increases , the weight near @xmath224 disappears .
that is in concordance with the results of cpt shown in fig .
[ fig_64 ] where the weight at that location exists only for small @xmath74 that also agrees with slave - boson calculations@xcite that found such weight for @xmath101 and it agrees also with one - loop calculations@xcite starting from a hartree - fock antiferromagnetic state that did not find weight at that location for @xmath78 .
the simplest hartree - fock approach@xcite yields weight near @xmath224 only for unreasonably small values of @xmath10 . a cartoon explanation of the pseudogap is given in fig .
[ fig_88 ] . at high temperature
we have a fermi liquid , as illustrated in panel i. now , suppose we start from a ground state with long - range order as in panel iii , in other words at a filling between half - filling and @xmath249 . in the hartree - fock approximation we have a gap and the fermion creation - annihilation operators now project on bogoliubov - valentin quasiparticles that have weight at both positive and negative energies . in two dimensions , the mermin - wagner theorem means that as soon as we raise the temperature above zero , long - range order disappears , but the antiferromagnetic correlation length @xmath250 remains large so we obtain the situation illustrated in panel ii , as long as @xmath250 is much larger than the thermal de broglie wave length @xmath251 in our usual units . at the crossover temperature @xmath85 then the relative size of @xmath250 and @xmath252 changes and
we recover the fermi liquid .
we now proceed to sketch analytically where these results come from starting from finite temperature .
details and more complete formulae may be found in refs . .
note also that a study starting from zero temperature has also been performed in ref . .
first we show how tpsc recovers the mermin - wagner theorem .
consider the self - consistency conditions given by the local moment sum rule eq .
( [ suscep ] ) together with the expression for the spin - susceptibility eq .
( [ chi_sp ] ) and @xmath20 in eq .
( [ ansatz ] ) .
first , it is clear that if the left - hand side of the local moment sum rule eq .
( [ suscep ] ) wants to increase because of proximity to a phase transition , the right - hand side can do so only by decreasing @xmath71 which in turns decreases @xmath20 through eq .
( [ ansatz ] ) and moves the system away from the phase transition .
this argument needs to be made more precise to include the effect of dimension .
first , using the spectral representation one can show that every term of @xmath253 is positive .
near a phase transition , the zero matsubara frequency component of the susceptibility begins to diverge .
on can check from the real - time formalism that the zero - matsubara frequency contribution dominates when the characteristic spin fluctuation frequency @xmath254 becomes less than temperature , the so - called renormalized - classical regime .
we isolate this contribution on the left - hand side of the local moment sum rule and we move the contributions from the non - zero matsubara frequencies , that are non - divergent , on the right - hand side . then , converting the wave vector sum to an integral and expanding the denominator of the susceptibility around the wave vector where the instability would occurs to obtain an ornstein - zernicke form , the local moment sum rule eq .
( [ suscep ] ) can be written in the form@xmath255 the constant on the right - hand side contains only non - singular contributions and @xmath256 contains @xmath20 that we want to find . from the above equation
, one finds immediately that in @xmath257 @xmath258 so that the correlation length diverges only at @xmath259 in three dimensions , isotropic or not , exponents correspond to those of the @xmath87 universality class.@xcite to see how the pseudogap opens up in the single - particle spectral weight , consider the expression ( [ self - long ] ) for the self - energy .
normally one has to do the sum over bosonic matsubara frequencies first , but the zero matsubara frequency contribution has the correct asymptotic behavior in fermionic frequencies @xmath24 so that one can once more isolate on the right - hand side the zero matsubara frequency contribution .
this is confirmed by the real - time formalism@xcite ( see also eq .
( [ sigma_reel ] ) below ) . in the renormalized classical regime
then , we have @xmath260 where @xmath261 is the wave vector of the instability . hence , when @xmath262 , in other words at hot spots , we find after analytical continuation and dimensional analysis that @xmath263 clearly , in @xmath264 , @xmath265 vanishes as temperature decreases , @xmath266 is the marginal dimension and in @xmath192 we have that @xmath267 that diverges at zero temperature . in a fermi liquid
that quantity vanishes at zero temperature .
a diverging @xmath265 corresponds to a vanishingly small @xmath268 as we can see from@xmath269 to see graphically this relationship between the location of the pseudogap and large scattering rates at the fermi surface , we draw in fig .
[ fig_91 ] both the fermi surface mdc and , in the lower panels , the corresponding plots for @xmath270 .
note that at stronger @xmath10 the scattering rate is large over a broader region , leading to a depletion of @xmath271 over a broader range of @xmath91 values .
an argument for the splitting in two peaks seen in figs .
[ fig_29 ] and [ fig_88 ] is as follows .
consider the singular renormalized contribution coming from the spin fluctuations in eq .
( [ rc - contribution - sigma ] ) at frequencies @xmath272
taking into account that contributions to the integral come mostly from a region @xmath273 , this expression leads to@xmath274 which , when substituted in the expression for the spectral weight ( [ spectral - weight_1 ] ) leads to large contributions when @xmath275 or , equivalently,@xmath276 which corresponds to the position of the hot spots in fig .
[ fig_85 ] for example .
note that analogous arguments hold for any fluctuation that becomes soft,@xcite including superconducting ones.@xcite the wave vector @xmath261 would be different in each case .
the cpt results of figs .
[ fig_62 ] and [ fig_71 ] clearly show that the pseudogap is different from the mott gap . at finite doping ,
the mott gap remains a local phenomenon , in the sense that there is a region in frequency space that is not tied to @xmath277 where _ for all wave vectors _ there are no states .
the peudogap by contrast is tied to @xmath277 and occurs in regions nearly connected by @xmath278 whether we are talking about hole- or about electron - doped cuprates .
that the phenomenon is caused by short - range correlations can be seen in cpt from the fact that the pseudogap is independent of cluster shape and size ( most of the results were presented for @xmath279 clusters and we did not go below size @xmath280 ) .
the antiferromagnetic correlations and any other two - particle correlations do not extend beyond the size of the lattice in cpt .
hence , the pseudogap phenomenon can not be caused by antiferromagnetic long - range order since no such order exists in cpt .
this is also vividly illustrated by the cdmft results in fig .
[ fig_81 ] that contrast the case with and without antiferromagnetic long - range order .
the cdmft results also suggest that the pseudogap appears in the bands that are most affected by antiferromagnetic correlations hence it seems natural to associate it with short - range spin correlations .
the value of @xmath210 has an effect , but it mostly through the fact that it has a strong influence on the relative location of the antiferromagnetic zone boundary and the fermi surface , a crucial factor determining where the pseudogap is .
all of this as well as many results obtained earlier by dca@xcite suggest that there is a strong coupling mechanism that leads to a pseudogap in the presence of only short - range two - body correlations
. however , the range can not be zero .
only the mott gap appears at zero range , thus the pseudogap is absent in single - site dmft . in the presence of a pseudogap at strong coupling ( @xmath281 ) ,
wave vector is not , so to speak , such a bad quantum number in certain directions . in other words
the wave description is better in those directions . in other directions
that are `` pseudogapped '' , it is as if the localized , or particle description was better .
this competition between wave and particle behavior is inherent to the hubbard model . at the fermi surface in low dimension
, it seems that this competition is resolved by dividing ( it is a crossover , not a real division ) the fermi surface in different sections .
there is also a weak - coupling mechanism for the pseudogap .
this has been discussed at length just in the previous section on tpsc .
another way to rephrase the calculations of the previous section is in the real frequency formalism .
there one finds@xcite that @xmath282\chi_{sp}^{\prime\prime}(\mathbf{q}% ; \omega^{\prime } ) \label{sigma_reel}%\end{aligned}\ ] ] so that if the characteristic spin fluctuation frequency in @xmath283 is much larger than temperature , then @xmath284 $ ] can be considered to act like a window of size @xmath285 or @xmath286 and @xmath287 can be replaced by a function of @xmath288 times @xmath289 which immediately leads to the fermi liquid result @xmath290.$ ] in the opposite limit where the characteristic spin fluctuation frequency in @xmath291 is much less than temperature , then it acts as a window narrower than temperature and @xmath284 $ ] can be approximated by the low frequency limit of the bose factor , namely @xmath292 using the thermodynamic sum rule , that immediately leads to the result discussed before in eq.([imsigma2 ] ) , @xmath293 this mechanism for the pseudogap needs long correlation lengths . in cpt , this manifests itself by the fact that the apparent pseudogap in fig .
[ fig_64 ] at @xmath2 is in fact mostly a depression in spectral weight that depends on system size and shape .
in addition , in contrast to the short - range strong - coupling mechanism , at weak coupling the pseudogap is more closely associated with the intersection of the antiferromagnetic zone boundary with the fermi surface .
which mechanism is important for the cuprates will be discussed below in the section on comparisons with experiments .
the existence of d - wave superconductivity at weak coupling in the hubbard model mediated by the exchange of antiferromagnetic fluctuations @xcite had been proposed even before the discovery of high - temperature superconductivity.@xcite at strong - coupling , early papers@xcite also proposed that the superconductivity would be d - wave .
the issue became extremely controversial , and even recently papers have been published@xcite that suggest that there is no d - wave superconductivity in the hubbard model .
that problem could have been solved very long ago through qmc calculations if it had been possible to do them at low enough temperature .
unfortunately , the sign problem hindered these simulations , and the high temperature results@xcite were not encouraging : the d - wave susceptibility was smaller than for the non - interacting case .
since that time , numerical results from variational qmc,@xcite exact diagonalization@xcite and other numerical approaches@xcite for example , suggest that there is indeed d - wave superconductivity in the hubbard model . in the first subsection ,
we show that vcpt leads to a zero - temperature phase diagram for both hole and electron - doped systems that does show the basic features of the cuprate phase diagram , namely an antiferromagnetic phase and a d - wave superconducting phase in doping ranges that are quite close to experiment@xcite ( the following section will treat in more detail comparisons with experiment ) .
the results are consistent with cdmft.@xcite the fall in the d - wave superconducting order parameter near half - filling is associated with the mott phenomenon .
the next subsection stresses the instability towards d - wave superconductivity as seen from the normal state and mostly at weak coupling .
we show that tpsc can reproduce available qmc results and that its extrapolation to lower temperature shows d - wave superconductivity in the hubbard model .
the transition temperature found at optimal doping@xcite for @xmath2 agrees with that found by dca,@xcite a result that could be fortuitous . but
again the concordance between theoretical results obtained at intermediate coupling with methods that are best at opposite ends of the range of coupling strengths is encouraging .
-wave ( top ) order parameters for @xmath78 , @xmath214 @xmath215 as a function of the electron density ( @xmath32 ) for @xmath294 , @xmath295 and 10-site clusters , calculated in vcpt .
vertical lines indicate the first doping where only @xmath55-wave order is non - vanishing . from ref .
, width=302 ] for various values of @xmath210 , calculated in cdmft on a @xmath296 cluster for @xmath78 .
the positive @xmath210 case corresponds to the electron - doped case when a particle - hole transformation is performed . from ref . .
, width=264 ] in vcpt , one adds to the cluster hamiltonian the terms@xcite @xmath297 with @xmath298 and @xmath299 are respectively antiferromagnetic and d - wave superconducting weiss fields that are determined self - consistently and @xmath300 equal to @xmath301 on near - neighbor sites following the d - wave pattern .
we recall that the cluster hamiltonian should be understood in a variational sense .
[ fig_96 ] summarizes , for various cluster sizes , the results for the d - wave order parameter @xmath302 and for the antiferromagnetic order parameter @xmath303 for a fixed value of @xmath78 and the usual hopping parameters @xmath214 and @xmath215 .
the results for antiferromagnetism are quite robust and extend over ranges of dopings that correspond quite closely to those observed experimentally . despite the fact that the results for d - wave superconductivity still show some size dependence , it is clear that superconductivity alone without coexistence extends over a much broader range of dopings on the hole - doped than on the electron - doped side as observed experimentally .
vcpt calculations on smaller system sizes@xcite but that include , for thermodynamic consistency , the cluster chemical potential as a variational parameter show superconductivity that extends over a much broader range of dopings . also , for small @xmath280 clusters , vcpt has stronger order parameter on the electron than on the hole - doped side , contrary to the results for the largest system sizes in fig .
[ fig_96 ] .
this is also what is found in cdmft as can be seen in fig .
[ fig_138 ] .
it is quite likely that the zero - temperature cooper pair size is larger than two sites , so we consider the results for @xmath280 systems only for their qualitative value .
concerning the question of coexistence with antiferromagnetism , one can see that it is quite robust on the electron - doped side whereas on the hole - doped side , it is size dependent .
that suggests that one should also look at inhomogeneous solutions on the hole - doped side since stripes are generally found experimentally near the regions where antiferromagnetism and superconductivity meet .
-wave ( top ) order parameters as a function of the electron density ( @xmath32 ) for @xmath214 @xmath215 and various values of @xmath10 on a 8-site cluster , calculated in vcpt . from ref . .
, width=302 ] for various values of @xmath10 , and @xmath304 calculated in cdmft on a 4-site cluster . from ref . .
, width=264 ] fig .
[ fig_97 ] shows clearly that at strong coupling the size of the order parameter seems to scale with @xmath305 , in other words it decreases with increasing @xmath74 this is also found in cdmft,@xcite as shown in fig .
[ fig_139 ] for @xmath235 . near
half - filling on @xmath306 lattice .
contrary to the strong coupling case , the d - wave order parameter @xmath307 survives all the way to half - filling at weak coupling , unless we also allow for antiferromagnetism.,width=245 ] if we keep the antiferromagnetic order parameter to zero , one can check with both vcpt and cdmft ( fig .
[ fig_138 ] ) that the d - wave superconducting order parameter goes to zero at half - filling .
this is clearly due to mott localization .
indeed , at smaller @xmath2 for example , the order parameter does not vanish at half - filling if we do not allow for long - range antiferromagnetic order , as illustrated in fig .
[ fig_139 ] for cdmft.@xcite the same result was found in vcpt , as shown in fig.[fig_1].@xcite restoring long - range antiferromagnetic order does however make the d - wave order parameter vanish at half - filling .
there are thus two ways to make d - wave superconductivity disappear at half - filling , either through long antiferromagnetic correlation lengths@xcite or through the mott phenomenon . in the real systems , that are mott insulators and
also antiferromagnets at half - filling , both effects can contribute . in the introduction to this section
, we alluded to qmc calculations for the d - wave susceptibility.@xcite recent results@xcite for that quantity as a function of doping for various temperatures and for @xmath308 @xmath235 are shown by symbols in fig .
[ fig_12 ] . for lower temperatures ,
the sign problem prevents the calculation near half - filling . yet
, the lowest temperature is low enough that a dome shape begins to appear .
nevertheless , comparison with the non - interacting case , shown by the top continuous line , leads one to believe that interactions only suppress d - wave superconductivity .
we can easily understand why this is so . as we already know , the tpsc results obtained from eq .
( [ suscep_d ] ) are very close to the qmc calculations , as shown by the solid lines in fig .
[ fig_12 ] . in the temperature range of interest ,
the main contribution comes from the first term in eq .
( [ suscep_d ] ) .
that term represents the contribution to the susceptibility that comes from dressed quasiparticles that do not interact with each other . since dressed quasiparticles have a lifetime , a pair breaking effect , it is normal that this contribution to the interacting susceptibility leads to a smaller contribution than in the non - interacting case . at the lowest temperature , @xmath98 , the vertex contribution represented by the second term in eq .
( [ suscep_d ] ) accounts for about @xmath309 of the total .
it goes in the direction of increasing the susceptibility .
as we decrease the temperature further in tpsc , that term becomes comparable with the first one . since the vertex in eq .
( [ suscep_d ] ) accounts for the exchange of a single spin wave , equality with the first term signals the divergence of the series , as in @xmath310 .
the divergence of that series represents physically the instability of the normal phase to a d - wave superconducting phase .
this is analogous to the thouless criterion and hence it gives an upper bound to @xmath209 .
in other words , berezinskii - kosterlitz - thouless physics is not included . as a function of doping , @xmath311 , for @xmath235 calculated in tpsc using the thouless criterion . from ref .
.,width=245 ] fig .
[ fig_104 ] shows the tpsc transition temperature for @xmath2 and for @xmath101 .
as we move towards half - filling , located to the left of the diagram , starting from large dopings , @xmath209 goes up because of the increase in antiferromagnetic fluctuations .
eventually , @xmath209 turns around and decreases because of the opening of the pseudogap .
physically , when the pseudogap opens up , weight is removed from the fermi level , the density of states becomes very small , and pairing can not occur any more . in the flex approximation@xcite that does not exhibit a pseudogap,@xcite that downturn , observed already in qmc at high temperature , does not occur . we have observed that in cases where @xmath312 so that the pseudogap opens only in a limited region around hot spots , the downturn can become less pronounced .
the case @xmath313 that corresponds to optimal doping for @xmath2 in fig .
[ fig_104 ] has been studied by dca . in an extensive and systematic study of the size dependence , maier _
@xcite established the existence of a d - wave instability at a temperature that coincides to within a few percent with the result in fig .
[ fig_104 ] .
since very few vortices can fit within even the largest cluster sizes studied in ref . , it is quite likely that the @xmath209 that they find does not include berezinskii - kosterlitz - thouless effects , just like ours .
despite the fact that , again , the concordance between weak and strong coupling methods at intermediate coupling comforts us , the uncertainties in the results found with tpsc and dca force us to also allow for a fortuitous coincidence .
the reduction of the real problem of high - temperature superconducting materials to a one - band hubbard is a non - trivial one .
it has been discussed already in the early days of high @xmath209 superconductivity .
the notion of a zhang - rice singlet@xcite emerged for hole doped systems .
the mapping to a one - band model has been discussed in many references,@xcite and we do not wish to discuss this point further here .
in fact it is far from obvious that this mapping is possible .
it is known that about @xmath314 ev below the fermi surface , that mapping fails in hole - doped systems.@xcite nevertheless , the one - band hubbard model is in itself a hard enough problem for us .
so it is satisfying to see that , in the end , it gives a picture that agrees with experiment in a quite detailed manner for the arpes spectrum near the fermi surface , for the phase diagram as well as for neutron scattering in cases where it can be calculated .
although we will not come back on this point at all , we briefly mention that fitting the spin wave spectrum@xcite for all energies and wave vectors at half filling in la@xmath315cuo@xmath316 gives values of @xmath317 that are close to those used in the rest of this paper.@xcite it is in this context that ring exchange terms are usually discussed .
arpes experiments have played a central role in the field of high - temperature superconductivity .
we can not expect to be able to present the vast experimental literature on the subject .
we refer the reader to a very exhaustive review@xcite and to some less complete but recent ones.@xcite the main facts about arpes have been summarized in ref . .
we comment on their main points one by one , using italics for our paraphrase of the reported experimental observations . _
( i ) the importance of mott physics and the renormalization of the bandwidth from @xmath212 to @xmath305 for the undoped parent compounds_. this renormalization was clear already in early qmc calculations.@xcite we already discussed the presence of four peaks .
the one nearest to the fermi surface at negative energies is the one referred to by experimentalists when they refer to this renormalization .
this band has a dispersion of order @xmath305 ( not shown on fig .
[ fig_55a ] , but see ref . )
. this result also agrees with vcpt as shown in fig .
[ fig_55a ] and cdmft ( fig .
[ fig_81]a ) . as shown in figs .
[ fig_81]a and [ fig_81]d , whether the state is ordered or not the band width is similar .
analytical strong - coupling expansions@xcite and exact diagonalizations also find the same result . to find detailed agreement with experiment , one needs to include @xmath210 and @xmath211.@xcite the evolution of the position of chemical potential for extremely small dopings as discussed in ref
. is not reproduced by the strong - coupling calculations , although the result on chemical potential is somewhat material dependent.@xcite _ ( ii ) in the overdoped case , the fermi surface is well defined . _
although we have not shown any figures concerning this point , vcpt and cdmft show the same result .
hole - doped ca@xmath240na@xmath241cuo@xmath242cl@xmath242 from ref .
.,width=188 ] _ ( iii ) the evolution with doping of the electronic structure has been mapped .
it has shown the importance of antiferromagnetic correlations in the p - type underdoped cuprates and especially in the n - type ones in which the hot - spot physics is still observed at optimal doping .
_ we will come back on the latter point for electron - doped cuprates in the following subsection .
the strong - coupling results obtained with vcpt and cdmft have a resolution of order @xmath318 , which translates into about @xmath319 mev .
this is not enough to accurately measure the fermi velocity , which was found to be doping independent in lsco.@xcite however , this suffices to compare with mdc curves obtained experimentally by integrating over an energy range of about @xmath320 mev , as shown in fig .
[ fig_107 ] obtained in ref . on calcium oxyclorate ca@xmath240na@xmath321cuo@xmath315cl@xmath315 , a @xmath245 hole - doped high temperature superconductor .
the similarities between that figure and the cpt ( fig . [ fig_64 ] ) , vcpt ( fig . [ fig_110 ] ) and cdmft ( fig .
[ fig_75 ] ) results is striking .
the agreement is better when no antiferromagnetic long - range order is assumed , as in the cpt case .
the flattening of the band structure near @xmath219 observed experimentally , can also be seen in cpt by comparing the top and middle edc s taken at small and large @xmath10 respectively on the left panel of fig .
[ fig_63 ] .
this flattening is associated with the pseudogap phenomenon .
recall that the theoretical results were obtained with @xmath322 and @xmath215 .
this in turn implies an electron - hole asymmetry that is observed experimentally .
we come back to this in the following subsection . _
( iv ) the overall d - wave symmetry of the superconducting gap has been observed for both hole and electron doping , supporting the universality of the pairing nature in the cuprates .
_ in the next to next subsection , we discuss the phase diagram for competing antiferromagnetism and d - wave superconductivity and show striking similarities with the observations . _
( v ) a normal - state pseudogap has been observed to open up at a temperature _ @xmath323 _ in the underdoped regime with a d - wave form similar to the one of the superconducting gap .
_ that statement is correct only in the hole - doped compounds . in electron - doped systems
the pseudogap has a form that is not of d - wave shape .
if @xmath209 comes from a universal pairing mechanism , a universal mechanism may also be behind the pseudogap .
as we have already discussed however , there are quantitative differences between strong and weak coupling mechanisms for both @xmath209 and the pseudogap . for electron - doped systems , we made quantitative predictions for the value of @xmath324 that have later been confirmed experimentally .
all this is discussed further below .
to date , in cluster methods the pseudogap temperature has been studied only with dca.@xcite _ ( vi ) a coherent quasiparticle peak below _
@xmath209 _ has been observed near _
whose spectral weight scales with the doping level _ @xmath325 _ in the underdoped regime .
_ we expect that it is a general result that long - range order will restore quasiparticle like excitations in strongly correlated systems because gaps remove scattering channels near the fermi level .
this is clearly illustrated by comparing the upper and lower panels in fig .
[ fig_81 ] that contrast the same spectra with and without antiferromagnetic long - range order .
we have not performed the analysis of our results yet that could tell us whether the spectral weight of the quasiparticle scales with @xmath325 in the hole - underdoped regime .
our resolution may not be good enough to see the quasiparticle peak .
sharpening of the quasiparticle excitations in the superconducting state has however been observed in dca.@xcite _ ( vii ) the presence of an energy scale of about _
@xmath326 mev _ in the quasiparticle dynamics manifests itself through a sharp dispersion renormalization and drop in the scattering rate observed at those energies at different momenta . _ in hole - doped systems there is a kink in the nodal direction that is already seen above @xmath209 while in the antinodal direction it appears only below @xmath209 . the energy scales and doping dependences of these two kinks are also different.@xcite the energy resolution in vcpt and cdmft is not sufficient to distinguish these subtleties . in electron - doped cuprates
experiments@xcite suggest that there is no observable kink feature , in agreement with the results presented in the following subsection .
the arpes spectrum of electron - doped cuprates is strikingly different from that of their hole - doped counterpart .
the fermi energy mdc s for the first quadrant of the brillouin zone@xcite are shown at the top of fig .
[ fig_121a ] for three different dopings .
there is a very clear evolution with doping . at the lowest dopings
, there is no weight near @xmath327 contrary to the hole - doped case shown in fig .
[ fig_107 ] . for all dopings there is weight near @xmath219 instead of the pseudogap that appeared there in the hole - doped case .
the edc s , also shown on the bottom of fig .
[ fig_121a ] , are drawn for a trajectory in the brillouin zone that follows what would be the fermi surface in the non - interacting case .
regions that are more green than red on the corresponding mdc s along that trajectory are referred to as hot spots .
on the edc s we clearly see that hot spots do not correspond to simply a decrease in the quasiparticle weight @xmath328 .
they truly originate from a pseudogap , in other words from the fact that the maximum is pushed away from zero energy .
even though the measurements are done at low temperature @xmath329 the energy resolution of about @xmath320 mev makes the superconducting gap invisible .
what is observed at this resolution is the pseudogap . in units of @xmath212 and wave vector from vcpt for @xmath78 @xmath330 and @xmath331 at the bottom and @xmath332 ( electron - doped ) at the top .
the lorentzian broadening is @xmath333 in the main figure and @xmath334 in the inset that displays the d - wave gap .
top panel is for the electron - doped case in the right - hand panel of fig .
[ fig_110 ] , while bottom panel is for the hole - doped case on the left of fig .
[ fig_110 ] . from ref .
.,width=321 ] electron - doping . from ref . ,
width=245 ] the contrast between the location of the pseudogap in the hole and electron - doped compounds is clearly seen in fig .
[ fig_117 ] obtained from vcpt.@xcite in that figure , the magnitude of the spectral weight is represented by the different colors as a function of frequency ( in units of @xmath212 ) along different cuts of the brillouin zone . in the bottom panel , for the hole - doped case , one observes the pseudogap near @xmath219 . in the top panel , for the electron - doped case
, it is only by zooming ( inset ) on the region for the fermi energy crossing near @xmath219 that one sees the d - wave superconducting gap . at @xmath245 electron - doping
, the pseudogap near @xmath224 is apparent . in this case
there is antiferromagnetic long - range order , but even if we use cpt that does not exhibit long - range order , there appears a pseudogap in that region .
the main difference between cpt and vcpt results is the bending back of the bands ( for example around the symmetry axis at @xmath206 ) caused by halving of the size of the brillouin zone in the antiferromagnetic case .
form factors@xcite are such that the intensity is not symmetric even if the dispersion is . the faint band located at an energy about @xmath212 below the fermi energy near @xmath224
was also found in ref . by a one - loop spin - wave calculation around the hartree - fock antiferromagnetic ground state at @xmath78 .
experimentalists@xcite have suggested the existence of these states .
the vcpt results go well beyond the spin - wave calculation ( dashed lines in fig .
[ fig_117 ] ) since one can also see numerous features in addition to remnants of the localized atomic levels around @xmath335 and @xmath336 .
the optimally doped case is the real challenge for strong - coupling calculations .
the spin - wave approach in ref . never shows the weight near @xmath224 that is seen in experiment ( fig .
[ fig_121 ] ) .
early mean - field calculations by kusko et al.@xcite suggest that this @xmath224 feature appears for @xmath100 .
this is very small compared with @xmath10 of the order of the bandwidth @xmath337 necessary to have a mott insulator at half - filling .
we already discussed in sec .
[ weakandstrongcoupling ] that both cpt and tpsc show that weight near @xmath224 appears for @xmath10 not too large , say of order @xmath207 .
this same result is also obtained in the kotliar - ruckenstein slave boson approach.@xcite along the fermi surface calculated in tpsc ( left ) at optimal doping for @xmath247 , @xmath248 , @xmath338 mev and corresponding arpes data on ncco ( right ) . from ref .
.,width=245 ] along two other directions calculated for @xmath247 , @xmath248 , @xmath338 mev in tpsc ( left column ) and corresponding arpes data on ncco ( right column ) . from ref .
.,width=245 ] since tpsc is valid for a system of infinite size , we present detailed comparisons@xcite with experiment@xcite on nd@xmath339ce@xmath340cuo@xmath316 , an electron - doped cuprate .
we take @xmath247 , @xmath341 results obtained with @xmath342 are very close to those we present . with the values used in cpt , @xmath217 @xmath343 @xmath344 tpsc does not lead to strong enough antiferromagnetic fluctuations to obtain non - trivial effects in the temperature range studied , @xmath345 .
we take @xmath338 mev .
[ fig_121a ] shows the correspondence between edc and mdc .
comparisons with experimental edc at wave vectors along the non - interacting fermi surface appear in fig .
[ fig_122 ] for @xmath244 and @xmath246 doping @xmath346 .
the dashed lines indicate the quite detailed agreement between theory and experiment . at the hot spot ,
( middle dashed line ) , the weight is pushed back about @xmath347 ev and there is a very small peak left at the fermi surface , as in the experiment .
if @xmath10 is not large enough the antiferromagnetic fluctuations are not strong enough to lead to a pseudogap . as in cpt
[ fig_64 ] ) , if @xmath10 is too large the @xmath224 weight disappears , as illustrated earlier in fig .
[ fig_85 ] . in fig .
[ fig_123 ] , cuts along the @xmath348 to @xmath349 and @xmath350 to @xmath351 directions are compared with experiment .
again the peak positions and widths are very close , except for some experimental tails extending in the large binding energy direction .
the theoretical results have similar asymmetry , but not as pronounced .
experimentally , the large binding energy tails ( the background ) are the least reproducible features from sample to sample , especially for wave vectors near the fermi surface " .
the experimental renormalized fermi velocities are @xmath352 m / s and @xmath353 m / s along the zone diagonal and along the @xmath219-@xmath349 direction , respectively .
the corresponding renormalized fermi velocities obtained by tpsc are @xmath354 m / s and @xmath355 m / s , respectively .
the agreement is very good , particularly along the diagonal direction .
the bare fermi velocities are renormalized in tpsc by roughly a factor of two.@xcite along the fermi surface shown in the insets for ( b ) @xmath332 , @xmath243 .
lines are shifted by a constant for clarity . from ref . .
, width=245 ] as we move towards half - filling , we have to increase @xmath10 slightly to find agreement with experiment , as discussed earlier in fig .
[ fig_85 ] .
[ fig_124 ] shows how well the edc s agree for a fermi surface cut at @xmath245 doping @xmath356 .
the increase is expected physically from the fact that with fewer electrons the contribution to screening that comes from thomas fermi physics should not be as good .
this is also consistent with the fact that a larger value of @xmath10 is necessary to explain the mott insulator at half - filling .
it would also be possible to mimic the arpes spectrum by keeping @xmath10 fixed and changing the hopping parameters , but the changes would be of order @xmath357 which does not appear realistic.@xcite mev ) .
filled symbols denote calculated results and empty ones experimental data of ref . and ref .
( @xmath358 ) . from ref . .
, width=245 ] we have already explained that the physics behind the pseudogap in tpsc is scattering by nearly critical antiferromagnetic fluctuations .
this is illustrated in fig .
[ fig_125 ] .
if this explanation is correct , the antiferromagnetic correlation length measured by neutron scattering should be quite large .
the results of the measurement@xcite and of the tpsc calculations are shown in fig .
[ fig_126 ] .
the agreement is again surprisingly good .
as we move to smaller dopings @xmath359 ( not shown ) the agreement becomes less good , but we do expect tpsc to deteriorate as @xmath10 increases with underdoping .
the arrow points to the temperature where edc s shown earlier were calculated .
note however that the neutron measurements were done on samples that were not reduced , by contrast with the arpes measurements mentioned earlier .
we are expecting experiments on this subject .. we should point out that the edc s depend strongly on temperature and on the actual value of the antiferromagnetic correlation length only in the vicinity of the temperature where there is a crossover to the pseudogap regime .
decreasing the temperature makes the @xmath277 peaks near @xmath219 sharper@xcite as observed experimentally.@xcite ( filled circles denote @xmath324 calculated from tpsc , empty ones experimental data extracted from optical conductivity.@xcite ) empty triangles are experimental nel temperatures @xmath360 .
the samples are reduced.@xcite from ref . .
, width=245 ] the arpes pseudogap temperature @xmath324 has been predicted with tpsc.@xcite the predictions are shown by the solid line in fig .
[ fig_127 ] .
the pseudogap temperature observed in optical experiments@xcite is shown by the open circles .
it differs from the arpes result , especially as we move towards optimal doping .
the size of the pseudogap observed in the optical experiments@xcite @xmath361 is comparable to the arpes pseudogap . the solid line in fig .
[ fig_127 ] contains several predictions .
if we look at @xmath362 doping @xmath363 , the line predicts @xmath364 .
experiments that were done without being aware of this prediction@xcite have verified it .
it would be most interesting to do neutron scattering experiments on the same samples to check whether the antiferromagnetic correlation length @xmath250 and the thermal de broglie wave length @xmath252 are comparable at that temperature , as predicted by tpsc . fig .
[ fig_127 ] also predicts that the pseudogap induced by antiferromagnetic fluctuations will disappear at the quantum critical point where long - range antiferromagnetic order disappears , in other words it will coincide with the crossing of the experimentally observed nel temperature ( dashed line with triangles in fig .
[ fig_127 ] ) with the zero temperature axis ( if that crossing is not masked by the superconducting transition ) .
recent optical conductivity experiments@xcite confirm this prediction as well . in tpsc ,
superconducting fluctuations can also lead to a pseudogap by an analogous mechanism.@xcite superconductors , from ref . .
there should also be a pseudogap line on the electron - doped side .
it was not well studied at the time of publication of that paper.,width=245 ] the main features appearing in the phase diagram of high - temperature superconductors are the pseudogap phase , the antiferromagnetic phase and the d - wave superconducting phase .
[ fig_133]@xcite shows the typical diagram with hole doping to the right and electron doping to the left .
zero on the horizontal axis corresponds to half - filling .
there are other features on the phase diagram , in particular checkerboard patterns@xcite or stripe phases@xcite that appear in general close to the region where antiferromagnetism and superconductivity come close to each other . before we try to understand these more detailed features , one should understand the most important phases . in the previous subsection
we have discussed the pseudogap phase , in particular on the electron - doped side ( not indicated on fig .
[ fig_133 ] ) .
a recent review of the pseudogap appears in ref . .
in the following , we discuss in turn the phase diagram and then the nature of the superconducting phase itself and its relation to the mott phenomenon .
( dashed ) and d - wave ( solid ) order parameter obtained from cdmft on a @xmath296 cluster .
the result obtained by forcing @xmath365 is also shown as a thin dashed line.,width=245 ] we have already shown in fig .
[ fig_96 ] the prediction of vcpt for the zero - temperature phase diagram.@xcite here , we just point out how closely the position of the antiferromagnetic phase boundary , appearing in the lower panel , coincides with the experimental phase diagram in fig .
[ fig_133 ] .
( note that electron concentration increases from right to left on this experimental phase diagram ) .
in particular , there is little size dependence to the position of this boundary , ( 6 to 10 sites ) and in addition the dependence on the value of @xmath10 is also weak , as can be seen from fig .
[ fig_97 ] .
hence , the positions of the antiferromagnetic phase boundaries is a robust prediction of vcpt . the cdmft result for a four site cluster in a bath
is shown in fig .
[ fig_136 ] for @xmath366 and @xmath78 .
the agreement with experiment is not as good . despite the useful presence of a bath in cdmft ,
the cluster itself is of size @xmath367 which is probably smaller than the cooper pair size .
we can obtain results closer to those of vcpt by increasing the variational space .
the d - wave superconducting order parameter on the top panel of fig .
[ fig_96 ] shows more size dependence than the antiferromagnetic order parameter . nevertheless , there are some clear tendencies : ( a ) d - wave superconductivity can exist by itself , without antiferromagnetism .
the vertical lines indicate the location of the end of the antiferromagnetic phase for the various system sizes to help this observation .
( b ) the range where d - wave - superconductivity exists without antiferromagnetism , is about three times larger on the hole than on the electron - doped side , as observed experimentally .
( c ) as system size increases , the maximum d - wave order parameter is larger on the hole than on the electron - doped side .
( d ) the tendency to have coexisting antiferromagnetism and d - wave superconductivity is rather strong on the electron - doped side of the phase diagram .
this is observed experimentally@xcite but only over a rather narrow region near optimal doping .
recent experiments@xcite challenge this result , others@xcite indicate that antiferromagnetism can be induced from the d - wave superconducting phases with very small fields .
( e ) on the hole - doped side , d - wave superconductivity and antiferromagnetism coexist for a very narrow range of dopings for system size @xmath368 , for a broad range extending to half - filling for @xmath369 and not at all for @xmath370 . in other words ,
the tendency to coexistence is not even monotonic .
we interpret this result as a reflection of the tendency to form stripes observed experimentally on the hole - doped side.@xcite we can not study systems large enough to allow for striped inhomogeneous states to check this statement .
for the two - band model . from ref . , width=245 ]
the more realistic two - band model has also been studied using dca.@xcite the results are shown on fig .
[ fig_136b ] .
electron concentration increases from right to left .
this phase diagram is very close to that obtained with the same method from the one - band hubbard model@xcite with @xmath217 @xmath371 , @xmath78 .
the qualitative results agree with the other calculations and with experiment : antiferromagnetism extends over a narrower doping range for hole than for electron doping and d - wave superconductivity by itself exists over a broader range for the hole - doped case than for the electron - doped case .
the actual ranges where antiferromagnetism and d - wave superconductivity exist are not in as good an agreement with experiment as in the vcpt case . however , as in cdmft , the system sizes , @xmath280 , are very small .
overall then , quantum cluster methods , vcpt in particular , allow us to obtain from the hubbard model the two main phases , antiferromagnetic and d - wave superconducting , essentially in the observed doping range of the zero - temperature phase diagram . at finite temperature ,
dca and tpsc agree on the value of @xmath209 for the particle - hole symmetric model at @xmath245 doping and @xmath2 .
recent studies of the irreducible vertex using dca@xcite also show that in the weak - coupling limit the particle - particle d - wave channel leads to an instability driven by antiferromagnetic fluctuations as temperature decreases , as found in tpsc . to understand the effect of pressure on the phase diagram , note that @xmath372 should decrease as pressure increases since the increase in the overlap between orbitals should lead mainly to an increase in @xmath212 .
hence , as can be deduced from fig .
[ fig_104 ] , applying pressure should lead to a decrease in the value of @xmath209 at weak coupling , concomitant with the decrease in antiferromagnetic fluctuations that lead to pairing in the weak coupling case .
this is indeed what pressure does experimentally in the case of electron - doped high - temperature superconductors,@xcite reinforcing our argument that near optimal doping they are more weakly coupled .
it is widely known on the other hand that pressure _ increases _ @xmath209 in hole - doped systems . that is consistent with the strong - coupling result that we found in vcpt and cdmft , namely that the maximum d - wave order parameter in that case scales with @xmath373 a quantity that increases with @xmath212 and hence pressure .
whereas in the weak coupling case superconductivity is a secondary phenomenon that occurs after antiferromagnetic fluctuations have built up , in strong coupling they can be two distinct phenomena as can be seen from the phase diagram , even though they arise from the same microscopic exchange interaction represented by @xmath305 .
superconductivity in the underdoped regime is very much non - bcs .
first of all , we notice in fig .
[ fig_139 ] obtained in cdmft@xcite that at strong coupling the d - wave superconducting order parameter vanishes as we move towards half - filling even in the absence of long - range antiferromagnetic order . in other words , the mott phenomenon by itself suffices to destroy d - wave superconductivity .
this conclusion is reinforced by the fact that at weak coupling ( @xmath2 ) where there is no mott localization , d - wave superconductivity survives at half - filling . in the presence of antiferromagnetic long - range order , that last statement would not be true , as confirmed by vcpt calculations in fig .
[ fig_1 ] : at @xmath2 d - wave superconductivity survives at half - filling if we do not allow for antiferromagnetic long - range order but it disappears if we do .
in bcs theory , the presence of an interaction @xmath305 that leads to attraction in the d - wave channel would lead at @xmath374 to d - wave superconductivity at all dopings including half - filling , unless we allow for competing long - range order . at strong coupling ,
no long - range order is necessary to destroy d - wave superconductivity .
superconductivity at strong coupling@xcite also differs from bcs in the origin of the condensation energy .
suppose we do bcs theory on the attractive hubbard model .
then , as in the usual bcs model , kinetic energy is increased in the superconducting state because the fermi surface is no - longer sharp . on the other hand , in the superconducting phase there is a gain in potential energy .
the reverse is true at strong coupling .
this result follows from dca@xcite and is in agreement with the kinetic energy drop in the superconducting state that has been estimated from the _
_ f-__sum rule in optical conductivity experiments.@xcite photoemission data@xcite had also suggested this kinetic energy drop in the superconducting state . a crossover from non - bcs - like to bcs behavior in the condensation mechanism as we go from underdoping to overdoping has also been seen recently experimentally.@xcite we do not seem to have the resolution to find that crossover since the condensation energy becomes very small on the overdoped side .
we expect that crossover from strong to weak coupling will also lead to a change from a kinetic - energy driven to a potential - energy driven pairing mechanism .
this is confirmed by cdmft calculations for the attractive hubbard model.@xcite , @xmath214 as calculated in cdmft on a @xmath296 cluster . from ref .
.,width=245 ] a third way in which superconductivity in the underdoped regime is non - bcs is that the drop in the order parameter as we go towards half - filling is accompanied by an increase in the gap as measured in the single - particle density of states .
fig . 6 of ref . summarizes the experimental evidence for the increase in the size of the gap .
that increase , observed in the cdmft calculation of the gap , is illustrated in fig .
[ fig_141].@xcite that gap has essentially the same size as that observed in the normal pseudogap state@xcite .
high - temperature superconductivity has forced both experimentalists and theorists to refine their tools and to develop new ones to solve the puzzles offered by this remarkable phenomenon . from a theoretical perspective , the original suggestion of anderson@xcite that the physics was in the one - band hubbard model is being confirmed . in the absence of _
ab initio _ methods to tell us what is the correct starting point , such insight is essential .
the non - perturbative nature of the phenomenon has however forced theorists to be extremely critical of each other s theories since none of them can pretend that a small parameter controls the accuracy of the approximations .
if theorists are to convince each other and experimentalists that a solution of the high - temperature superconductivity problem has been found , then the theories have to give quantitative results and to make predictions . unlike most traditional problems in condensed matter physics however , the non - perturbative nature of the problem means that no simple mean - field like theory can be trusted , even if it seems to agree qualitatively with experiment .
in fact several such theories have been proposed@xcite not long after the experimental discovery of the phenomenon but they have not been accepted immediately .
theories have to be internally consistent , they have to agree with exact results whenever they are available , and then they can be compared with experiments . if there is a disagreement with experiment , the starting point ( one - band hubbard model ) needs to be reconsidered . when approaches developed on the basis of weak - coupling ideas agree at intermediate coupling with approaches developed on the basis of strong - coupling ideas , then one gains confidence in the validity of the results .
we have argued that such concordance is now found in a number of cases and that corresponding rather detailed quantitative agreement with experiment can be found . in a non - perturbative context
it becomes essential to also cross check various approaches .
the main theoretical methods that we have discussed are those that we have developed or perfected or simply used in our group : the two - particle self - consistent approach that is based on weak - coupling but non - perturbative ideas ( no diagrams are involved ) , as well as heavily numerical approaches such as qmc and various quantum cluster methods , vcpt and cdmft . based on our own work and that of many others , we think the following experimental facts about high - temperature superconductivity can be reproduced very accurately by calculations for the one - band hubbard model with @xmath10 in the intermediate coupling range ( @xmath375 ) with @xmath376 mev , and hopping parameters @xmath210 and @xmath211 close to the values suggested by band structure calculations,@xcite namely @xmath217 @xmath215 .
\(i ) in the one - band hubbard model the main phases of the zero - temperature phase diagram , namely antiferromagnetic and d - wave superconducting , appear very near the observed ranges for both the hole- and electron - doped cases .
\(ii ) the normal state is unstable to a d - wave superconducting phase in a temperature range that has the correct order of magnitude . as usual
the value of @xmath209 is the most difficult quantity to evaluate since one must take into account kosterlitz - thouless physics as well as the effect of higher dimensions etc , so this level of agreement must be considered satisfying .
\(iii ) the arpes mdc at the fermi energy and the edc near the fermi energy are qualitatively well explained by cluster calculations for both hole- and electron - doped cases .
these comparisons , made at a resolution of about @xmath319 to @xmath320 mev are not very sensitive to long - range order , although order does influence the results .
one is mainly sensitive to the pseudogap , so this is the main phenomenon that comes out from the model .
energy resolution is not good enough to see a kink .
more details about what aspects of arpes are understood may be found in sec .
[ arpes_overview ] .
\(iv ) in the case of electron - doped cuprates , the value of @xmath10 near optimal doping seems to be in the range @xmath377 , which means that it is accessible to studies with tpsc that have better resolution . in that case , the agreement with experiment is very accurate , even if there is room for improvement and a need for further experiments .
in addition , the value of @xmath324 for 13% doping has been predicted theoretically before it was observed experimentally , one of the very rare predictions in the field of high - temperature superconductors .
all of this agreement with arpes data is strong indication that @xmath377 is appropriate to describe electron - doped superconductors near optimal doping .
additional arguments come from the pressure dependence of the superconducting transition temperature @xmath378 which increases with @xmath379 contrary to the strong - coupling result , and from simple ideas on thomas - fermi screening .
the latter would predict that the screened interaction scales like @xmath380 and cpt results do lead to @xmath381 smaller on the electron- than on the hole - doped side.@xcite in addition , the optical gap at half - filling is smaller in electron- than in hole - doped systems .
what is the physics ?
the physics of the antiferromagnetic phase at both weak and strong coupling is well understood and needs no further comment .
for the pseudogap , we have argued that there seems to be two mechanisms , a weak coupling one that involves scattering off critical fluctuations and that is very well understood within tpsc , and a strong - coupling one where there is no need for large correlation lengths .
there is no simple physical picture for the latter mechanism although the fact that it does not scale with @xmath305 but with @xmath212 seems to suggest forbidden hopping .
the pseudogap is clearly different from the mott gap . whether there is a phase transition as a function of @xmath10 that separates the weak and strong coupling regimes or whether there is only a crossover is an open question .
the shape of the mdc s at the fermi energy clearly show in any case that in some directions wave vector is not such a bad quantum number whereas in the pseudogap direction , a localized or almost localized particle - like picture would be appropriate .
in fact the pseudogap occurs near the intersection with the antiferromagnetic zone boundary that turns out to also be the place where umklapp processes are possible . in other words ,
the presence of a lattice is extremely important for the appearance of the pseudogap .
we have seen that with spherical fermi surfaces the fermi liquid survives even for large @xmath10 .
the dichotomy between the wave description inherent to the fermi liquid and the particle ( localized ) description inherent to the mott phenomenon seems to be resolved in the pseudogap phase by having certain directions where electrons are more wave - like and other directions where particle - like ( gapped ) behavior appears .
the latter behavior appears near regions where the presence of the lattice is felt through umklapp processes .
it is clear that when weak - coupling - like ideas of quasiparticles scattering off each other and off collective excitations do not apply , a simple physical description becomes difficult .
in fact , knowing the exact wave functions would give us the solution but we would not know how to understand physically the results .
this lack of simple physical images and the necessity to develop a new discourse is quite apparent for d - wave superconductivity . at weak coupling exchange of slow antiferromagnetic fluctuations
is at the origin of the phenomenon , while at strong - coupling the fact that the maximum value of the d - wave order parameter scales with @xmath305 tells us that this microscopic coupling is important , even though there is no apparent boson exchange .
this is where mean - field like theories@xcite or variational approaches@xcite can help when they turn out to give results that are confirmed by more accurate and less biased methods .
there are many open problems , some of which are material dependent and hence may depend on interactions not included in the simplest hubbard model .
we have already mentioned the problem of the chemical potential shift in arpes for very small dopings@xcite that seems to be somewhat material dependent@xcite .
it would also be important to understand additional inhomogeneous phases that are observed in certain high - temperature superconductors .
that is extremely challenging for quantum cluster methods and unlikely to be possible in the very near future , except for inhomogeneities of very short wave length .
also , we still need to improve concordance between the methods before we can make predictions that are quantitative at the few percent level for all physical quantities .
apart from dca , there are no quantum cluster methods that have been developed yet to study two - particle response functions that are necessary to obtain results on the superfluid density and on transport in general .
transport studies are being completed in tpsc.@xcite such studies are crucial since they are needed to answer questions such as : ( i ) why is it that for transport properties , such as optical conductivity , the number of carriers appears to scale with doping whereas in arpes the surface of the brillouin zone enclosed by the apparent fermi surface appears to scale with the number of electrons ?
is it because the weight of quasiparticles at the fermi surface scales like the doping or because of vertex corrections or because of both ?
( ii ) can we explain a vanishing superfluid density as doping goes to zero@xcite only through mott physics or can competing order do the job.,@xcite after twenty years all the problems are not solved , but we think that we can say with confidence that the essential physics of the problem of high - temperature superconductivity is in the one - band hubbard model .
at least the pseudogap , the antiferromagnetic and the d - wave superconducting phases come out from the model
. refinements of that model may however be necessary as we understand more and more details of the material - specific experimental results .
has a revolution been necessary to understand the basic physics of high - temperature superconductors ? certainly , it has been necessary to change our attitude towards methods of solution .
we have seen that to study intermediate coupling , even starting from weak coupling , it has been necessary to drop diagrams and to rely instead on sum rules and other exact results to devise a non - perturbative approach . at strong coupling we had to accept that numerical methods are essential for progress and that we need to abandon some of the traditional physical explanations of the phenomena in terms of elementary excitations . even though progress has been relatively slow , the pace is accelerating in the last few years and there is hope that in a few years the problem will be considered for the most part solved .
the theoretical methods ( numerical and analytical ) that have been developed and that still need to be developed will likely remain in the tool box of the theoretical physicist and will probably be useful to understand and perhaps even design other yet undiscovered materials with interesting properties .
the success will have been the result of the patient and focused effort of a large community of scientists fascinated by the remarkable phenomenon of high - temperature superconductivity .
the present work was supported by nserc ( canada ) , fqrnt ( qubec ) , cfi ( canada ) , ciar , the tier i canada research chair program ( a .- m.s.t . ) .
we are grateful to our collaborators , g. albinet , s. allen , m. boissonneault , c. brillon , m. capone , l. chen , m. civelli , a .-
dar , b. davoudi , j .- y .
p. delannoy , a. gagn - lebrun , m. j. p. gingras , a. georges , m. guillot , v. hankevych , p. c. w. holdsworth , f. jackson , s. kancharla , g. kotliar , j .- s .
landry , p .-
lavertu , f. lemay , s. lessard , m .- a .
marois , s. pairault , d. perez , m. pioro - ladrire , d. plouffe , d. poulin , l. raymond , s. roy , p. sahebsara , h. touchette , and especially y.m .
we also acknowledge useful discussions with p. fournier , m. greven , i. herbut , k. shen and l. taillefer and we are grateful to v. hankevych and s. kancharla for permission to include some of their unpublished figures in this paper .
arpes : angle resolved photoemission spectroscopy : experiment from which one can extract @xmath382 .
cdmft : cellular dynamical mean field theory : a cluster generalization of dmft that allows one to take into account both wave vector and frequency dependence of the self - energy.@xcite it is best formulated in real space .
dca : dynamical cluster approximation : a cluster generalization of dmft that allows one to take into account both wave vector and frequency dependence of the self - energy based on coarse graining of the self - energy in reciprocal space.@xcite dmft : dynamical mean field theory : this approach is exact in infinite dimension .
it takes the frequency dependence of the self - energy into account and includes both the mott and the fermi liquid limits.@xcite qmc : quantum monte carlo : determinental approach@xcite .
this provides an essentially exact solution to the model for a given system size and within statistical errors that can be made smaller by performing more measurements .
s. allen , a .- m .
tremblay and y. m. vilk , in _ theoretical methods for strongly correlated electrons _ , edited by d. snchal , c. bourbonnais and a .-
tremblay , crm series in mathematical physics , springer ( 2003 ) .
kotliar , s. y. savrasov , k. haule , v. s. oudovenko , o. parcollet , c.a .
marianetti , _ electronic structure calculations with dynamical mean - field theory : a spectral density functional approach _ ( 2005 ) , .
g. biroli and g. kotliar , phys .
b * 71 * , 037102 ( 2005 ) ; b. kyung , g. kotliar , a. -m
. s. tremblay , _ quantum monte carlo study of strongly correlated electrons : cellular dynamical mean - field theory _ , ( 2006 ) . | this is a short review of the theoretical work on the two - dimensional hubbard model performed in sherbrooke in the last few years .
it is written on the occasion of the twentieth anniversary of the discovery of high - temperature superconductivity .
we discuss several approaches , how they were benchmarked and how they agree sufficiently with each other that we can trust that the results are accurate solutions of the hubbard model .
then comparisons are made with experiment .
we show that the hubbard model does exhibit d - wave superconductivity and antiferromagnetism essentially where they are observed for both hole and electron - doped cuprates .
we also show that the pseudogap phenomenon comes out of these calculations . in the case of electron - doped high temperature superconductors , comparisons with angle - resolved photoemission experiments are nearly quantitative .
the value of the pseudogap temperature observed for these compounds in recent photoemission experiments has been predicted by theory before it was observed experimentally .
additional experimental confirmation would be useful .
the theoretical methods that are surveyed include mostly the two - particle self - consistent approach , variational cluster perturbation theory ( or variational cluster approximation ) , and cellular dynamical mean - field theory . |
a viable baryogenesis mechanism aims to explain the observed asymmetry in the baryon density , @xmath0 , and the celebrated sakharov conditions state the necessary ingredients for baryogenesis : ( i ) c and cp violation , ( ii ) non - equilibrium , ( iii ) b number violation .
b number violation is present in the hot universe due to sphaleron processes while c is violated in the electroweak sector of the standard model ( sm ) .
the two important aspects of electroweak baryogenesis ( ewbg)@xcite are transport and cp violation .
ewbg requires a strong first - order electroweak phase transition to drive the plasma out of equilibrium .
the cp violation is induced by the moving phase boundary .
hence it is important to derive transport equations that contain cp - violating quantum effects in a genuine manner . compared to other baryogenesis mechanisms , ewbg has the attractive property that the relevant energy scale will be accessible by the next generation of collider experiments .
the kadanoff - baym equations represent the statistical analog to the schwinger - dyson equations and are of the following form : ( + - p_l m e^- _ k - p_r m^ e^- _ k ) g^ < ( k , x ) = , and all quantities are functions of the center of mass coordinate @xmath1 and the momentum @xmath2 . in the approximation of planar wall profiles ,
the spin is conserved what can be used to partially decouple the equations . in the kadanoff - baym equations ,
the exponential derivative operator is usually expanded which corresponds to a semi - classical expansion in @xmath3 , the so - called gradient expansion . in the following
we will keep the first two orders in @xmath3 .
the simplest example of cp violation in transport equations is given by the one - flavour case with a @xmath4dependent complex phase in the mass term @xcite , @xmath5 .
in this case the constraint equation ( real part of the k .- b .
equation ) leads to the following _ ansatz _ ( the subscript @xmath6 is the spin quantum number ) g^<_s = 2 f_s ( k_0 - _ s ) , _ 0 = , _ s = _ 0 - s , where @xmath7 denotes the distribution function and @xmath8 the energy of the particle in the presence of the wall .
the kinetic equation ( imaginary part of the k .- b .
equation ) is of the vlasov type _ z f_s + f_s _
k_z f_s = , f_s = - + s [ sem_cl_force ] .
note that the second part of the force @xmath9 violates cp and hence sources ewbg .
the multi flavour case can be treated in the linear response approximation , where the green function is split according to @xmath10 and leads to a kinetic equation of the form @xcite ( without using the in general nonalgebraic constraint eqs .
that reproduce the dispersion relations in lowest order ) k_z_z
g + + k_0 g = s ( g^<,eq ) .
the third term is a damping term , taking into account the collision terms .
note that the second term will lead to an oscillation of the off - diagonal densities in the mass eigenbasis , similar to neutrino oscillations .
the right - hand side of this equation contains contributions of higher order in the gradient expansion that give rise to cp violation and ewbg .
in the mssm the dominant contribution to baryogenesis comes from the charginos ( higgsino - wino - mixing ) with the mass matrix _
l = , m(z ) = m_2 & g h_2^ * ( z ) + g h_1^ * ( z ) & _ c , where the susy breaking parameters @xmath11 and @xmath12 contain complex phases .
the cp - violating sources to first order in gradients only contribute to the off diagonal terms in flavour space and hence are suppressed by the oscillation effect .
they read @xcite ^a_&= & 2 g^2 t_c^-4(m_2_c ) ( |m_2|^2 - |_c|^2 ) _
( h_1 h_2)_(0)^3 , + ^b_&= & 2 g^4 t_c^-4(m_2_c ) ( h_1 ^ 2-h_2 ^ 2 ) _ ( h_1 h_2)_(0)^3 , + ^c_&= & -2 g^2 t_c^-2 ( m_2_c ) ( h_2_h_1 - h_1_h_2 ) ( _ ( 0)^0 + 4_(2)^3 ) .
in addition , there is a cp - violating source of second order in the gradient expansion that contributes to the diagonal elements in flavour space and corresponds to the semi - classical force that appears in the one flavour case in eq .
( [ sem_cl_force ] ) ^d_0 = 2 v_w g^2 t_c^-4(m_2_c ) ( h_2 ^ 2_z h_1 + h_1 ^ 2_z h_2 ) _ ( 0)^3 .
the functions @xmath13 and @xmath14 denote certain momentum integrals of the equilibrium distribution functions .
[ mssm_ewbg ] shows the produced baryon asymmetry for a maximal cp - violating phase in the chargino mass matrix using the system of diffusion equations suggested in ref .
@xcite . gev . in the right plot ,
the black area denotes the region in the @xmath15 plane , where ewbg is viable.[mssm_ewbg],title="fig:",width=8 ] gev . in the right plot ,
the black area denotes the region in the @xmath15 plane , where ewbg is viable.[mssm_ewbg],title="fig:",width=5 ] -1.5 cm in the right plot , the black area denotes the region of the parameter space where ewbg is viable .
we conclude that ewbg in the mssm is only possible if : ( i ) the charginos are nearly mass degenerate such that mixing effects are not suppressed .
( ii ) the cp phases in the chargino sector are @xmath16 .
the general nmssm of ref .
@xcite consists of the mssm extended by a gauge singlet and the superpotential w_nmssm = s h_1 h_2 + s^3 + h_1 h_2 + r s + w_mssm .
, scaledwidth=60.0% ] due to the explicit @xmath17 term and the singlet self - couplings , this model provides a rich higgs phenomenology ; however , additional assumptions have to be made to prevent higher - dimensional operators from destabilizing the hierarchy
. it does not suffer from a domain wall problem since there are no discrete symmetries . in the nmssm
, a @xmath18 or @xmath19 symmetry is imposed to solve the domain wall problem without destabilizing the electroweak hierarchy .
the @xmath17 term is forbidden and only induced after electroweak symmetry breaking . thus the @xmath17 problem is solved .
the discrete symmetries also eliminate the singlet self coupling . a rather large value of lambda is needed in the nmssm to fulfill current mass bounds on the higgsinos and charginos , which might lead to a landau pole below the gut scale .
in contrast to the mssm , no light stop is needed in the nmssm or nmssm , since the additional singlet terms in the higgs potential strengthen the phase transition @xcite . in the nmssm case
these terms read : = l_mssm + m_s^2 |s|^2 + ^2 |s|^2 ( h_1^h_1 + h_2^h_2 ) + t_s ( s + h.c . ) + ( a_s h_1 h_2 + h.c . ) . in a simplified scheme without cp violation , a first - order phase transition due to tree - level dynamics
occurs if @xcite m_s^2 < | - m_s a | , a = 2 , ^2 = ^2 2 + ^2 2 .
[ pt_ineq ] fig .
[ pt_nmssm ] displays eq .
( [ pt_ineq ] ) for random nmssm models with a strong pt and shows that this criterion is also decisive if cp violation and the one - loop effective potential are taken into account .
since the trilinear term in the superpotential contributes to the higgs mass , @xmath20 is generically of @xmath16 ; hence two - loop contributions from the charginos to the electron edm are naturally small .
the one - loop contributions to the electron edm can , as in the mssm , be reduced by increasing the sfermion masses .
, scaledwidth=60.0% ] the effective @xmath17 parameter is dynamical in the nmssm and nmssm , and hence its complex phase can change during the phase transition .
this leads to new cp - violating sources in the chargino sector that are of second order in the gradient expansion and do not rely on mixing .
thus , these contributions are not suppressed by the flavour oscillations and mass degenerate charginos are not required for viable ewbg . additionally , the bubble wall tends to be thinner than in the mssm and hence it is rather generic to generate the observed baryon asymmetry @xcite .
[ ewbg_nmssm ] shows the binned bau for a random set of nmssm models with a strong first order pt .
another interesting feature of the nmssm is transitional cp violation .
if universality is violated in the singlet sector of the nmssm , the phase transition can connect a high temperature phase with broken cp and a low temperature phase of conserved cp @xcite . such cp violation can not be detected by zero temperature experiments !
the singlet self coupling is important to stabilize the singlet in this scenario .
singlet extensions of the mssm provide a framework in which electroweak baryogenesis seems to be possible without fine tuning .
a. g. cohen , d. b. kaplan and a. e. nelson , nucl .
b * 349 * ( 1991 ) 727 .
k. kainulainen , t. prokopec , m. g. schmidt and s. weinstock , jhep * 0106 * ( 2001 ) 031 [ hep - ph/0105295 ] .
t. konstandin , t. prokopec and m. g. schmidt , nucl .
b * 716 * ( 2005 ) 373 [ hep - ph/0410135 ] .
t. konstandin , t. prokopec , m. g. schmidt and m. seco , nucl .
b * 738 * ( 2006 ) 1 [ hep - ph/0505103 ] .
p. huet and a. e. nelson , phys .
d * 53 * ( 1996 ) 4578 [ hep - ph/9506477 ] . c. panagiotakopoulos and a. pilaftsis , phys .
d * 63 * ( 2001 ) 055003 [ hep - ph/0008268 ] .
s. j. huber and m. g. schmidt , nucl .
b * 606 * , 183 ( 2001 ) [ hep - ph/0003122 ] ; s. j. huber and m. g. schmidt , proceedings sewm 00 , marseille , 272 - 278 [ hep - ph/0011059 ] .
a. menon , d. e. morrissey and c. e. m. wagner , phys .
d * 70 * ( 2004 ) 035005 [ hep - ph/0404184 ] .
s. j. huber , t. konstandin , t. prokopec and m. g. schmidt , hep - ph/0606298 . | we compare electroweak baryogenesis in the mssm , nmssm and nmssm .
we comment on the different sources of cp violation , the phase transition and constraints from edm measurements . |
pion compton scattering , @xmath10 , allows one to extract the electric and magnetic polarizabilities of the ( charged ) pion . in a classical picture
these polarizabilities characterize the deformation response ( i.e. induced dipole moments ) of a composite system in external electric and magnetic fields . in the proper quantum field
theoretical formulation the electric and magnetic polarizabilities , @xmath11 and @xmath12 , are defined as expansion coefficients of the compton scattering amplitudes at threshold . however , since pion targets are not directly available , real pion compton scattering has been approached using different artifices , such as high - energy pion - nucleus bremsstrahlung @xmath13 , radiative pion photoproduction off the proton @xmath14 , and the crossed channel two - photon reaction @xmath15 . from the theoretical side
there is an extraordinary interest in a precise ( experimental ) determination of the pion polarizabilities . within the framework of current algebra
@xcite it has been shown ( long ago ) that the polarizability difference @xmath16 of the charged pion is directly related to the axial - vector - to - vector form factor ratio @xmath17 measured in the radiative pion decay @xmath18 @xcite . at leading ( nontrivial ) order the result of chiral perturbation theory @xcite , @xmath19 ,
is of course the same after identifying the combination of low - energy constants as @xmath20 .
recently , the systematic corrections to this current algebra result have been worked out in refs.@xcite by performing a full two - loop calculation of pion compton scattering in chiral perturbation theory .
the outcome of that extensive analysis is that altogether the higher order corrections are rather small and the value @xmath21@xmath9
@xcite for the pion polarizability difference stands now as a firm prediction of the ( chiral - invariant ) theory .
the non - vanishing value @xmath22@xmath9 for the pion polarizability sum ( obtained also at two - loop order ) is presumably too small to cause an observable effect in low - energy pion compton scattering .
however , the chiral prediction @xmath23@xmath9 is in conflict with the existing experimental determinations of @xmath24@xmath9 from serpukhov @xcite and @xmath25@xmath9 from mainz @xcite , which amount to values more than twice as large .
these existing experimental determinations of @xmath16 certainly raise doubt as to their correctness since they violate the chiral low - energy theorem notably by a factor 2 . in
that contradictory situation it is promising that the ongoing compass @xcite experiment at cern aims at measuring the pion polarizabilities with high statistics using the primakoff effect .
the scattering of high - energy negative pions in the coulomb field of a heavy nucleus gives ( in the region of sufficiently small photon virtualities ) access to cross sections for @xmath26 reactions .
as an alternative , one could also directly analyze the bremsstrahlung process @xmath27 , omitting the whole kinematical extrapolation from virtual to real photons . in the appendix
we will write down the corresponding fivefold differential cross section @xmath28 including born terms , the pion polarizability difference @xmath16 , and an ( equally ) important pion - loop correction . in any case , the effects of the pion s low - energy structure on ( real or virtual ) compton scattering observables turn out to be relatively small . for center - of - mass energies @xmath29 ( i.e. sufficiently below the prominent @xmath30-resonance ) the differential cross sections @xmath31 in backward directions are reduced at most by about @xmath32 in comparison to the ones of a structureless pion @xcite .
therefore , a precise knowledge of the pure qed radiative corrections to pion compton scattering is indispensable if one wants to extract the pion polarizabilities from the data with good accuracy . in certain kinematical regions
the effects from the pion s low - energy structure and the pure qed radiative corrections may become of comparable size .
although calculations of radiative corrections are abundant in the literature , we are not aware of a detailed and utilizable exposition of the radiative corrections to scalar ( spin-0 ) boson compton scattering .
the case of spin-1/2 electron compton scattering has been treated already in the early days of quantum electrodynamics by brown and feynman @xcite .
the radiative corrections to the thomson limit , valid close to threshold , have been reported in ref.@xcite . in the work of akhundov et al .
@xcite the one - photon loop diagrams to virtual pion compton scattering have been considered , but no accessible sources to the corresponding analytical expressions ( which are necessary for an implementation into data analyses ) are given .
since those previous results were mainly presented in numerical form it is difficult to implement them independently into future data analyses . as a consequence of that deficit the radiative corrections to pion compton scattering are
sometimes merely adapted from the known ones for myons by simply replacing the mass : @xmath33 .
of course , such a substitutional procedure mistreats profound differences in the couplings of photons to scalar spin-0 bosons and spin-1/2 fermions .
the purpose of the present paper is to fill this gap and to present a detailed calculation of the one - photon loop radiative corrections to pion compton scattering , @xmath34 . we give closed - form analytical expressions for the @xmath1 corrections to the two invariant amplitudes @xmath2 and @xmath3 as they emerge from 11 classes of contributing one - loop diagrams .
ultraviolet and infrared divergencies are both treated by the method of dimensional regularization which ensures gauge invariance at every step .
while the ultraviolet divergences can be absorbed in renormalization constants , the cancellation of infrared divergencies requires the inclusion of soft photon radiation below an experimental detection threshold @xmath4 .
the total finite radiative correction depends then logarithmically on the small energy resolution scale @xmath4 .
we find that the radiative corrections become maximal in backward directions , reaching about @xmath5 at a center - of - mass energy of @xmath35 for @xmath36mev .
with such a size and kinematical signature the radiative corrections are not negligible in comparison to the effects from the pion polarizabilities , which also show up preferentially in the backward directions .
furthermore , we include in our calculation of the radiative corrections also the leading pion structure effect through a two - photon contact - vertex proportional to the pion polarizability difference @xmath37@xmath9 .
we find that these structure effects do not modify the relative size and angular dependence of the radiative corrections .
our results can be utilized for analyzing the compass experiment at cern .
we start out with defining the invariant amplitudes for the ( real ) pion compton scattering process : @xmath38 .
it is advantageous to work in the center - of - mass frame and to choose ( in this frame ) the coulomb gauge @xmath39 for the ( transverse ) photon polarization vectors .
the corresponding t - matrix reads then : @xmath40 \bigg\ } \,,\ ] ] with @xmath41 , and @xmath42 and @xmath43 the two independent mandelstam variables .
the inequalities refer to the physical region of pion compton scattering and the third ( dependent ) mandelstam variable @xmath44 is given by @xmath45 . as indicated in eq.(1 ) we view the two ( dimensionless ) invariant amplitudes , @xmath2 and @xmath3 , as functions of @xmath46 and @xmath47 .
the ( three ) tree diagrams of scalar quantum electrodynamics lead to the following contributions : @xmath48 note that the contribution of the @xmath46-channel pole diagram vanishes , since the couplings of the initial and final state photon are both equal to zero in coulomb gauge , @xmath49 , in the center - of - mass frame . performing the sums over transverse photon polarizations and applying flux and appropriate two - body phase space factors
the differential cross section at tree level reads : @xmath50 \over s[s(1+z)+m_\pi^2(1-z)]^2}\,,\ ] ] where @xmath51 has been expressed in terms of the cosine of the cms scattering angle @xmath52 . when including the one - photon loop radiative corrections into the invariant amplitudes , @xmath2 and @xmath3 , the differential cross section gets modified by an additive interference term ( see eq.(3 ) in ref.@xcite ) : @xmath53 } \bigg\ { 2m_\pi^2(1-z)\ , { \rm re}\,a(s , u)^ { ( \gamma-\rm loop)}\\\nonumber & & + ( 1+z)[m_\pi^2(1-z)-s(1+z ) ] \ , { \rm re}\,b(s , u ) ^{(\gamma-\rm loop ) } \bigg\ } \,,\end{aligned}\ ] ] of order @xmath54
. its ratio to the point - like cross section in eq.(3 ) defines the ( virtual part of the ) radiative correction factor @xmath55 as a function of the center - of - mass energy @xmath56 and @xmath57 .
before we turn to the explicit evaluation of the one - photon loop diagrams in the next section we add several remarks about regularization and renormalization .
we use the method of dimensional regularization to treat both ultraviolet and infrared divergencies ( where the latter are caused by the masslessness of the photon ) .
the method consists in calculating loop integrals in @xmath58 spacetime dimensions and expanding the results around @xmath59
. divergent pieces of one - loop integrals generically show up in form of the composite constant : @xmath60 containing a simple pole at @xmath59 .
in addition , @xmath61 is the euler - mascheroni number and @xmath62 an arbitrary mass scale introduced in dimensional regularization in order to keep the mass dimension of the loop integrals independent of @xmath58 .
ultraviolet ( uv ) and infrared ( ir ) divergencies are distinguished by the feature of whether the condition for convergence of the @xmath58-dimensional integral is @xmath63 or @xmath64 .
we discriminate them in the notation by putting appropriate subscripts , i.e. @xmath65 and @xmath66 . in the actual calculation
an infrared divergent term originates typically from a feynman parameter integral of the form : @xmath67 .
since scalar quantum electrodynamics is a multiplicatively renormalizable field theory , all ( ultraviolet ) divergencies can be absorbed in renormalization constants @xmath68 ( see appendix a in ref.@xcite ) .
these establish the ( formal ) relations between ( renormalized ) physical quantities ( @xmath69 , etc . ) and the bare ones .. ] in the present context , we want to remind the reader only about the cancellation between self - energy corrections and vertex corrections for the electric charge @xmath70 .
for that purpose , consider the coupling of a zero - momentum photon with polarization vector @xmath71 to an on - shell pion of momentum @xmath72 .
the tree level coupling @xmath73 supplemented by the one - photon loop corrections to it takes the following form : @xmath74 \bigg\ } \,,\ ] ] where the terms in round brackets correspond to the three diagrams shown in fig.1 in that order ( including for the first two their horizontally reflected partners ) . the same cancellation mechanism ( @xmath75 ) is at work in spinor electrodynamics @xcite , where the central one - photon loop diagram in fig.1 is absent , and the balance in the square bracket of eq.(6 ) reads : @xmath76 .
the effect of the pionic vacuum polarization ( on the photon propagator ) is canceled by the counterterm @xmath77 , corresponding the @xmath78 of the value in spinor electrodynamics @xcite . of the pion wavefunction renormalization .
in total the effects sum to zero . ]
in this section , we present analytical results for the radiative corrections of order @xmath79 to the invariant amplitudes @xmath2 and @xmath3 of pion compton scattering .
our convenient choice of coulomb gauge ( in the center - of - mass frame ) eliminates still all @xmath46-channel pole diagrams since for these either the initial state or final state photon coupling vanishes identically . dropping also the diagrams with a tadpole - type self - energy insertion ( the loop consisting of a single virtual photon line , which is set to zero in dimensional regularization ) we are left with the 20 one - photon loop diagrams shown in fig.2 . for the purpose of an ordering scheme they have been divided into classes i - xi . due to an increasing number of internal pion propagators their evaluation rises in complexity . in order to simplify all calculations , we exploit gauge invariance and employ the feynman gauge where the photon propagator is just proportional to the minkowski metric tensor @xmath80 .
moreover , for a concise presentation of the one - loop amplitudes , @xmath81 and @xmath82 , it is most helpful to work with the dimensionless kinematical variables @xmath83 , @xmath84 and @xmath85 , which obey the relation @xmath86 .
we can now enumerate the analytical expressions for the 11 classes of contributing one - photon loop diagrams . for notational simplicity
we drop the arguments @xmath46 and @xmath47 of the invariant amplitudes @xmath2 and @xmath3 , as well as a superscript indicating the class of diagrams . for this class of diagrams the tree level amplitudes in eq.(2 )
get multiplied by the wavefunction renormalization factor of the pion : . ]
@xmath87 this @xmath47-channel pole diagram involves the once - subtracted ( off - shell ) self - energy of the pion : @xmath88 \,.\ ] ] these @xmath47-channel pole diagrams generate a constant vertex correction factor : @xmath89 \,.\ ] ] these @xmath47-channel pole diagrams generate a @xmath90-dependent vertex correction factor : @xmath91\,.\ ] ] the sum of photon rescatterings in the @xmath46- and @xmath47-channel gives : @xmath92\,.\ ] ] these @xmath47-channel pole diagrams generate again a @xmath90-dependent vertex correction factor : @xmath93\,,\ ] ] these irreducible @xmath46-channel diagrams give : @xmath94 \bigg\}\,,\ ] ] where li@xmath95^{-1 } \ln y$ ] denotes the conventional dilogarithmic function .
these irreducible @xmath47-channel diagrams give : @xmath96 \bigg\}\ , , \\ & & b =- a+ { \alpha \over 2\pi}\ , { \hat t \over \hat u-1}\bigg\{{3\hat u-1 \over 2\hat u^2}\ln(1-\hat u)-{\hat u+1\over 2\hat u } + { 1\over \hat u-1 } \bigg[{\rm li}_2(\hat u)-{\pi^2 \over 6 } \bigg ] \bigg\}\,.\end{aligned}\ ] ] : the @xmath97-dependent vertex correction to the two - photon contact coupling reads : @xmath98 \bigg\}\ ,
. \end{aligned}\ ] ] here , we have introduced the frequently occurring logarithmic loop function : @xmath99 and the auxiliary variable @xmath100 appearing in eq.(16 ) is determined by the relation @xmath101 .
the @xmath46-channel box diagram gives : @xmath102-{1\over 2}\sqrt{4-\hat t}\ , l(t ) \nonumber \\ & & -l^2(t)+\int_1^\infty \!\!{dx \over x-\hat s}\ , { 2-\hat t\over ( x-1)^2+\hat t x } \big[(x+1)\ln x+2(1-x)\sqrt{4-\hat t } \ , l(t)\big]\bigg\ } \
, , \end{aligned}\ ] ] @xmath103 ^ 2 } \bigg[\hat t x ( x+1 ) \ln x \nonumber \\ & & + [ ( x-1)^2+\hat t x](x-1 ) + { 2(x-1 ) \over \sqrt{4-\hat t } } [ \hat t^2x -6\hat t x -2(x-1)^2]l(t ) \bigg]\bigg\}\ , . \end{aligned}\ ] ] the loop integrals involving four propagators are written here in terms of their spectral function representations ( employing the kramers - kronig dispersion relation ) , where the @xmath104-dependent integrands stand for the corresponding imaginary parts . for the numerical evaluation of the relevant real parts of @xmath105 and @xmath106 one has to treat @xmath107 as a principal value integral .
it can be conveniently decomposed into a sum of two nonsingular integrals by the following formula : @xmath108 the @xmath47-channel box diagram gives : @xmath109-{1\over 2}\sqrt{4-\hat t}\ , l(t)\nonumber \\ & & -l^2(t)+\int_1^\infty \!\!{dx
\over x-\hat u}\,{2-\hat t\over ( x-1)^2 + \hat t x } \big[(x+1)\ln x+2(1-x)\sqrt{4-\hat t}\ , l(t)\big ] \bigg\}\ , , \end{aligned}\ ] ] @xmath110\nonumber \\ & & + { 1\over 4 } -l^2(t)+ { \hat t^2 - 2 \hat t \over ( 1-\hat u ) \sqrt{\hat t^2 - 4\hat t}}\bigg[4\big(\xi_{ir}+\ln(1 -\hat u)\big ) \sqrt{-\hat t}\,l(t ) + { \rm li}_2(w ) \nonumber \\ & & -{\rm li}_2(1-w ) + { 1\over 2 } \ln^2 w -{1\over 2 } \ln^2(1-w ) + { \rm li}_2(h_- ) - { \rm li}_2(h_+ ) \bigg]+\int_1^\infty\!\!{dx \over x-\hat u } \ , \nonumber \\ & & \times
{ 2-\hat t\over [ ( x-1)^2+\hat t x]^2 } \bigg\{\hat t ( x+1)(2-\hat t - x ) \ln x+[(x-1)^2+\hat t x]\big(x+\hat
t+{\hat t\over x}-1\big ) \nonumber \\ & & + { 2l(t)\over \sqrt{4-\hat t}}\big[\hat t^3(x+1)+ \hat t^2(3x^2 - 3x-4 ) + 2\hat t(x^3 - 4x^2 + 2x+1)-2(x-1)^3\big ] \bigg\ } \bigg\}\ , , \end{aligned}\ ] ] where the arguments @xmath111 of the dilogarithms li@xmath112 are determined by the relation @xmath113
. several checks of our calculation can be performed now .
first , one verifies that the ultraviolet divergent terms proportional to @xmath65 cancel in the total sums for @xmath105 and @xmath106 separately ( using the relation @xmath114 in the latter case ) . secondly ,
crossing symmetry implies for our chosen set of independent amplitudes that @xmath2 and @xmath115 are symmetric functions under the exchange of the variables , @xmath116 .
this property follows e.g. by relating our set of amplitudes , @xmath2 and @xmath3 defined via eq.(1 ) , to the ( manifestly ) crossing - symmetric functions introduced in refs.@xcite .
while the symmetry condition @xmath117 becomes evident after summing up all contributions , the one for @xmath3 provides a highly nontrivial check between analytical terms and terms represented by dispersion integrals in eqs.(18 - 22 ) . in its consequence
, crossing symmetry interrelates the whole set of one - loop diagrams in fig.2 , which as such is not manifestly crossing - symmetric due to the particular choice of coulomb gauge .
in the next step we have to consider the infrared divergencies . inspection of eqs.(7,16,22 ) reveals that the ( summed ) infrared divergent terms proportional to @xmath66 appear in the same ratio @xmath118 as the ( tree level ) born terms written in eq.(2 ) . as a consequence of that , the infrared divergent virtual ( loop ) corrections multiply the point - like cross section @xmath119 by a factor : @xmath120 \ , \xi_{ir } \,.\ ] ] the unphysical infrared divergence @xmath66 gets canceled at the level of the cross section by the contributions of soft photon bremsstrahlung . in its final effect , the ( single ) soft photon radiation multiplies the tree level cross section @xmath121 by a ( universal ) factor @xcite : @xmath122 which depends on a small energy resolution @xmath4 . working out this momentum space integral by the method of dimensional regularization ( with @xmath64 )
one finds that the infrared divergent correction factor @xmath123 gets removed and the following finite radiative correction factor remains : @xmath124 \ln{m_\pi \over 2\lambda } + { \hat
s+1 \over \hat s-1 } \ln \hat s + \int_0^{1/2}\!\ ! dy \,{(\hat
s+1 ) ( \hat t-2)\over w [ 1- \hat t y(1-y ) ] } \ln { \hat s+1+w \over \hat s+1-w } \bigg\ } \,,\ ] ] with the abbreviation @xmath125 .
one can easily convince oneself that the radiative correction factor @xmath126 vanishes in forward direction @xmath127 .
we note that the terms beyond the ones proportional to @xmath128 are specific for the evaluation of the soft photon correction factor @xmath129 in eq.(24 ) in the center - of - mass frame . as it is written in eq.(25 ) , @xmath130 refers to an ( idealized ) experimental situation where all undetected soft photon radiation would fill in momentum space a small sphere of radius @xmath4 in the center - of - mass frame . in a real experiment
this region will be of different shape with no sharp boundaries due the detector efficiencies etc .
such additional ( experiment specific ) radiative corrections can be accounted for and calculated by integrating the fivefold differential cross section for @xmath131 @xcite over the appropriate region in phase space . by construction
this region excludes the infrared singular domain @xmath132 and thus leads to a finite result .
we are now in the position to present numerical results for the radiative corrections to pion compton scattering .
the complete radiative correction factor is @xmath130 written in eq.(25 ) plus all finite loop amplitudes @xmath105 and @xmath106 inserted into the interference cross section @xmath133 and divided by the tree level cross section @xmath134 .
fig.3 shows in percent the total radiative correction factor for three selected center - of - mass energies @xmath135 as a function of @xmath136 .
the detection threshold for soft photons has been set to the value @xmath137mev .
one observes that the radiative corrections become maximal in backward directions @xmath138 , reaching values up to @xmath5 for @xmath35 .
with such an angular dependence the pure qed radiative corrections have the same kinematical signature as the effects from the pion s low - energy structure ( i.e. pion polarizability difference @xmath139 and pion - loop correction of chiral perturbation theory @xcite ) . in magnitude
they are suppressed by about a factor @xmath140 .
we agree qualitatively with the findings of ref.@xcite that the radiative corrections simulate effects of ( fake ) pion polarizabilities @xmath141@xmath142 corresponding in magnitude to about @xmath143 of the prediction of chiral perturbation theory @xcite .
a proper inclusion of radiative corrections is therefore essential if one wants to extract the pion polarizabilities from compton scattering data with good accuracy .
we also note that when increasing the energy resolution scale @xmath4 the magnitude of the radiative correction goes down .
for example , with @xmath144mev or @xmath145mev the before - mentioned @xmath5 change to @xmath146 or @xmath147 . finally , it is worth to mention that as one approaches ( at fixed scattering angle @xmath148 ) the compton threshold , @xmath149 , all radiative correction effects go away @xcite . in this long - wavelength limit
only the pure thomson amplitude @xmath150 survives .
this non - renormalization theorem is strictly fulfilled by the one - loop amplitudes @xmath2 and @xmath3 written in eqs.(7 - 22 ) . in the same way , a direct computation at zero photon momentum @xmath151 gives for the factor multiplying the thomson amplitude : @xmath152 \ , , \end{aligned}\ ] ] where we have indicated the ( classes of ) contributing diagrams . .
the pion is treated as a structureless spin-0 boson . ]
so far we have treated in our calculation of the radiative corrections the pion as a structureless spin-0 boson .
such an approximation is valid only at very low ( photon ) energies .
the leading pion structure relevant for compton scattering is given by the difference of its electric and magnetic polarizability @xmath153 . in an effective field theory approach
it is easily accounted for by a new two - photon contact - vertex proportional to the squared electromagnetic field strength tensor , @xmath154 .
the s - matrix insertion following from this ( higher - order ) gauge - invariant effective @xmath155-vertex reads : @xmath156 where one photon @xmath157 is ingoing and the other one @xmath158 outgoing . at tree level
the polarizability vertex eq.(27 ) gives rise to the contributions @xmath159 and @xmath160 to the invariant compton scattering amplitudes . in order to prevent any misunderstandings we stress that when writing ( merely ) @xmath161 for the coupling strength in eq.(27 ) , an electric and a magnetic pion polarizability , equal in magnitude and opposite in sign @xmath162 , are always both included .
we can reinterpret the one - photon loop diagrams in fig.2 now in such a way that the two - photon contact - vertex represents the polarizability vertex proportional to @xmath161 . in the case of classv
we have to treat two cases , namely either the upper or lower contact - vertex .
going through the classes i , iii - ix and reevaluating the loop diagrams with the s - matrix insertion from the polarizability vertex , we find the following contributions to the invariant amplitudes @xmath105 and @xmath106 .
@xmath163 @xmath164 \ , , \\ & & b =- a+ { \beta_\pi m_\pi \ , t \over 2\pi}\bigg [ -\xi_{uv}+1-{1\over 2 \hat u } -{(\hat u- 1)^2\over 2\hat u^2 } \ln(1- \hat u ) \bigg ] \ , , \end{aligned}\ ] ] @xmath165 \bigg\}\ , , \\ & & b = { \hat s -1\over \hat u-1 } \,a \ , , \end{aligned}\ ] ] @xmath166 \bigg\}\ , , \\ & & b =- a+ { \beta_\pi m_\pi \,t \over 2\pi } \ , { \hat s -1\over \hat u-1}\bigg\{{\hat u+1\over 2\hat u}+ { 1 - 3\hat u \over 2\hat u^2}\ln(1-\hat u)+{1\over
\hat u-1 } \bigg[{\pi^2 \over 6 } -{\rm li}_2(\hat u)\bigg ] \bigg\}\,,\end{aligned}\ ] ] @xmath167 \bigg\}\ , , \qquad b = 0\ , .
\end{aligned}\ ] ] the contributions from classes iii and iv ( multiplicative vertex correction factors ) vanish due to some special features of the s - matrix insertion eq.(27 ) .
one verifies that the ultraviolet divergent terms proportional to @xmath65 drop out in the total sum for @xmath3 , but not for @xmath2 . in the latter case
the @xmath168-term can be canceled by interpreting the coupling constant in eq.(27 ) as a bare one and splitting it as @xmath169 into the physical coupling constant @xmath12 and a counterterm piece .
of negative mass dimension and therefore on general grounds it will generate a non - renormalizable quantum field theory .
] one also convinces oneself easily that the crossing - symmetry relations are fulfilled for the total sums of @xmath2 and @xmath3 .
the infrared divergent terms proportional to @xmath66 showing up in eqs.(28,35 ) are again canceled by the soft photon bremsstrahlung contributions . as a result
, one has to multiply the partial cross section linear in the polarizability : @xmath170}\,,\ ] ] with the radiative correction factor @xmath130 written down in eq.(25 ) . .
the pion s low - energy structure is included through a polarizability difference @xmath171@xmath142 . ] in fig.4 we show again the radiative correction factor for center - of - mass energies @xmath172 as a function of @xmath136 with a pion polarizability difference of @xmath173@xmath142 included .
the reference cross section is now the sum @xmath174 of point - like and polarizability - improved ( @xmath175 ) cross section .
the radiative correction is of course for each part of relative order @xmath79 .
the detection threshold of soft photons has been kept at the value @xmath176mev . in comparison to fig.3
one observes almost no changes for the three curves in fig.4
. only some slight upward shifts of the maximal values near @xmath177 are visible .
we can therefore conclude that the size and angular dependence of the radiative corrections to pion compton scattering remain practically unchanged when including the leading structure effect in form of the pion polarizability difference @xmath178@xmath142 .
of course , there is yet the pion - loop correction of chiral perturbation theory @xcite . in the cross sections it works partly against the polarizability effects
it is expected that the combined radiative pion - photon two - loop corrections ( emerging from a vast number of two - loop diagrams ) will follow the same trend . in summary , we have presented a detailed ( analytical ) calculation of radiative corrections to pion compton scattering which can be utilized for data analyses .
it is common practice to extract from primakoff events in high - energy pion - nucleus bremsstrah- lung a cross section for pion compton scattering via the equivalent photon method @xcite .
this involves a kinematical extrapolation from ( time - like polarized ) virtual photons to ( space - like transverse ) real photons . while the procedure becomes exact for asymptotically small photon virtualities @xmath179 , it introduced uncertainties into the analysis of real experiments , where only some lower bounds can be reached .
moreover , at the very low photon - virtualities @xmath180 shielding effects of the nuclear charge by inner - shell atomic electrons comes into play as a further complication .
the issue of coulomb - strong interaction interference in pion - nucleus bremsstrahlung has been studied recently in ref.@xcite .
according to this work it is advantageous to restrict oneself to momentum transfers @xmath181mev @xmath182 , since there the ratio between polarizability and born contributions remains essentially the same as for real pion compton scattering .
we want to point out here , that for the purpose of determining the pion polarizabilities ( i.e. the dominant difference @xmath153 ) one can omit the whole extrapolation procedure and directly analyze the pion bremsstrahlung process as it is measured in the laboratory frame .
the relevant ingredients to the cross section ( at sufficiently low momentum transfers @xmath183 where the one - photon exchange dominates ) are summarized in this appendix .
consider the pion bremsstrahlung process , @xmath184 , where a virtual photon transfers the ( small ) momentum @xmath185 from the nucleus ( of charge @xmath186 ) to the pion .
the extremely small recoil energy @xmath187 of the nucleus can be neglected .
the fivefold differential cross section in the laboratory frame reads @xcite : @xmath188 with @xmath189 and @xmath190 the squared amplitude summed over the final state photon polarizations @xmath191 : @xmath192 this expression has been derived from a t - matrix for the virtual compton scattering process , @xmath193 , of the form @xmath194 with @xmath195 and @xmath196 two ( real ) amplitudes . @xmath197 and @xmath198 are the angles between the momentum vectors @xmath199 and @xmath200 and @xmath201 , respectively .
in addition , @xmath202 is the angle between the two planes spanned by @xmath203 and @xmath204 .
the ( three ) tree diagrams of scalar quantum electrodynamics lead to the following contributions to the amplitudes @xmath195 and @xmath196 : @xmath205 with @xmath206 the initial energy of the pion , @xmath207 the final energy of the pion and @xmath208 .
the photon energy is denoted by @xmath209 .
note that the diagram involving the two - photon contact - vertex of scalar qed vanishes here , due to the orthogonality of the time - like virtual photon and the space - like real photon polarizations , @xmath210 .
the contributions from the pion electric and magnetic polarizabilities ( taking @xmath211 as a good approximation ) read : @xmath212 it follows directly from the s - matrix insertion in eq.(27 ) evaluated in the present kinematical situation , @xmath193 . at the same order ( in the small momentum expansion )
there are the pion - loop corrections of chiral perturbation theory @xcite . on the one hand side ,
they introduce the charge form factor of the pion : @xmath213 with @xmath214@xmath215 @xcite the mean square charge radius of the pion , @xmath216mev the pion decay constant , and the logarithmic loop function : @xmath217 as a result , the tree level born terms , @xmath218 and @xmath219 in eq.(39 ) , have then to be multiplied by the pion charge form factor @xmath220 , because a virtual photon ( with @xmath221 ) coupling to the pion can resolve this part of structure of the pion .
moreover and specific for ( virtual ) compton scattering , there is the pion - loop diagram shown in fig.5 .
it describes photon scattering off the pion - cloud around the pion and leads to the following contribution @xcite : @xmath222\nonumber \\ & & + 2 q \sqrt{4m_\pi^2 + q^2}\,\tilde l(q ) -{2q^2 \over \tau } \sqrt{4m_\pi^2 + \tau^2}\ , \tilde l(\tau ) \bigg\ } \ , .
\end{aligned}\ ] ] note that this irreducible loop contribution depends on both , the photon virtuality @xmath180 and the squared invariant momentum transfer @xmath223 between the initial and final state pion , with @xmath224 where @xmath225 . in the cross section
, the pion - loop correction in eq.(43 ) compensates in part the effects from the pion polarizabilities @xcite .
therefore , it is important to include it in the analysis of pion bremsstrahlung data , if one wants to extract reliably the pion polarizability difference @xmath226 .
of course , there are yet the one - photon loop radiative corrections to virtual pion compton scattering @xmath193 ( see fig.2 ) .
these will be presented ( in the same style as done here for real pion compton scattering ) in a forthcoming publication @xcite .
terentev , _ sov . j. nucl
. phys . _ * 16 * , 87 ( 1973 ) .
e. frlez et al . , _ phys .
lett . _ * 93 * , 181804 ( 2004 ) ; hep - ex/0606023 .
donoghue and b.r .
holstein , _ phys .
rev . _ * d40 * , 2378 ( 1989 ) .
u. brgi , _ phys .
lett . _ * b377 * , 147 ( 1996 ) ; _ nucl .
phys . _ * b479 * , 392 ( 1996 ) .
j. gasser , m.a .
ivanov , and m.e .
sainio , _ nucl .
_ * b745 * , 84 ( 2006 ) ; and refs . therein .
antipov et al . , _ phys .
* b121 * , 445 ( 1983 ) ; _ z. phys . _ * c26 * , 495 ( 1985 ) .
j. ahrens et al .
phys . j. _ * a23 * , 113 ( 2005 ) .
compass collaboration : p. abbon et al .
, hep - ex/0703049 . n. kaiser and j.m .
friedrich , _ eur .
phys . j. _ * a36 * , 181 ( 2008 ) .
brown and r.p .
feynman , _ phys .
rev . _ * 85 * , 231 ( 1952 ) .
e. corinaldesi and r. jost , _ helv .
. acta _ * 21 * , 183 ( 1948 ) .
akhundov , d.y .
bardin , and g.v .
mitselmakher , _ sov .
phys . _ * 37 * , 217 ( 1983 ) ; + a.a .
akhundov , d.y .
bardin , g.v .
mitselmakher , and a.g .
olshevskii , _ sov . j. nucl .
phys . _ * 42 * , 426 ( 1985 ) ; + a.a .
akhundov , s. gerzon , s. kananov , and m.a .
moinester , _ z. phys .
_ * c66 * , 279 ( 1995 ) .
m. vanderhaeghen et al . , _ phys .
rev . _ * c62 * , 025501 ( 2000 ) ; and refs . therein .
i.y . pomeranchuk and i.m .
shmushkevich , _ nucl .
* 23 * , 452 ( 1961 ) .
g. fldt and u. tengblad , nucl - th/0807.2700 ; nucl - th/0802.0971 ; _ phys .
rev . _ * c76 * , 064607 ( 2007 ) . c. itzykson and j.b .
zuber , quantum field theory , mcgraw - hill book company , 1980 ; chapter 5.2.4 . c. unkmeir , s. scherer , a.i .
lvov , and d. drechsel , _ phys .
rev . _ * c61 * , 034002 ( 2000 ) .
particle data group ( w.m .
yao et al . ) , _ j. phys . _
* g33 * , 1 ( 2006 ) .
n. kaiser and j.m .
friedrich , in preparation . | we calculate the one - photon loop radiative corrections to charged pion compton scattering , @xmath0 .
ultraviolet and infrared divergencies are both treated in dimensional regularization .
analytical expressions for the @xmath1 corrections to the invariant compton scattering amplitudes , @xmath2 and @xmath3 , are presented for 11 classes of contributing one - loop diagrams .
infrared finiteness of the virtual radiative corrections is achieved ( in the standard way ) by including soft photon radiation below an energy cut - off @xmath4 , and its relation to the experimental detection threshold is discussed .
we find that the radiative corrections are maximal in backward directions , reaching e.g. @xmath5 for a center - of - mass energy of @xmath6 and @xmath7mev .
furthermore , we extend our calculation of the radiative corrections by including the leading pion structure effect ( at low energies ) in form of its electric and magnetic polarizability difference , @xmath8@xmath9 .
we find that this structure effect does not change the relative size and angular dependence of the radiative corrections to pion compton scattering .
our results are particularly relevant for analyzing the compass experiment at cern which aims at measuring the pion electric and magnetic polarizabilities with high statistics using the primakoff effect .
24.5 cm 17.cm -2.2 cm -0.6 cm -0.6 cm # 1#2 # 2 cm n. kaiser and j.m .
friedrich + pacs : 12.20.-m , 12.20.ds , 13.40.ks , 14.70.bh |
low - dimensional molecular conductors provide a fruitful stage to study strong electron correlation effects leading to a wide variety of phase transitions @xcite . in this context ,
among the most - studied families are the quasi - one dimensional ( q1d ) molecular conductors tmttf@xmath1@xmath3 ( tmttf : tetramethyl - tetrathiofulvalene , @xmath3 : monovalent anion ) @xcite .
these salts form a q1d @xmath4-band at quarter - filling in terms of holes with intrinsic dimerization along the conduction axis .
they exhibit various types of phase transitions such as ferroelectric charge ordering ( fco ) , spin - peierls ( sp ) , antiferromagnetic ( af ) , and superconducting ( sc ) transitions by applying pressure or replacement of @xmath3 @xcite . among them , ( tmttf)@xmath1sbf@xmath2 shows a peculiar behavior under pressure ; a cooperative reduction of fco and af phase transition temperatures by the application of pressure has been reported by nmr measurements @xcite .
this result naively does not coincide with the case for typical co transitions , where co suppresses the tendency toward magnetic ordering due to decrease of the effective spin exchange couplings @xcite .
recently we have proposed a mechanism where the fco transition increases the two - dimensionality in the magnetic sector and leads to the stabilization of the af state @xcite . in this study , we have also found that the diagonal inter - chain hopping @xmath0 strongly depends on the choice of anion .
intriguingly , @xmath0 causes geometrical frustration between the chains of spins localized on every dimer , forming q1d exchange couplings as seen from fig .
[ fig.1 ] ( b ) ; we will see this effect more explicitly later . in this paper
, we investigate the role of @xmath0 for the spin ordering in the absence or the presence of fco , and how it affects the cooperative behavior between fco and two - dimensional ( 2d ) af states in tmttf@xmath1sbf@xmath2 .
we investigate a q1d extended hubbard model at 1/4-filling in terms of holes with coulomb interactions and inter - chain hoppings , whose hamiltonian is given by @xmath5 where @xmath6 is the transfer integral between the neighboring sites denoted by @xmath7 , @xmath8 ( @xmath9 ) is the creation ( annihilation ) operator of a hole on the @xmath10th site with spin @xmath11 or @xmath12 , and @xmath13 with @xmath14 .
@xmath15 and @xmath16 are the on - site and the inter - site coulomb interactions , respectively . and
@xmath17 , respectively.,width=226 ] the transfer integrals have been estimated @xcite for two representative members , ( tmttf)@xmath1pf@xmath2 and ( tmttf)@xmath1sbf@xmath2 ( hereafter , we call them as pf@xmath2 salt and sbf@xmath2 salt ) by the first - principles band calculations using the computational code qmas ( quantum materials simulator ) @xcite based on the projector augmented - wave method @xcite with the generalized gradient approximation @xcite .
we determine the values by fitting to the band structures .
the values of transfer integrals are given as @xmath18 @xmath19@xmath20 for pf@xmath2 salt and @xmath21 for sbf@xmath2 salt , in the unit of mev ( notations are shown in fig .
[ fig.1 ] ( a ) ) . from these results
, we can see that the diagonal hoppings @xmath22 are negligible for pf@xmath2 salt but appreciably large for sbf@xmath2 salt , especially @xmath23 . considering these results , we hereafter set the transfer integrals as @xmath24 and choose the diagonal inter - chain transfer integral @xmath0 as a parameter .
we set the on - site coulomb interaction as @xmath25 and impose a constraint on @xmath16 as @xmath26 and @xmath27 to realize the fco pattern shown in fig .
[ fig.3 ] ( b ) .
we perform numerical exact diagonalization on a @xmath28 sites cluster under the periodic boundary condition and calculate the following inter - dimer / intra - dimer charge and spin structure factors given by @xmath29 where @xmath30 is the total number of dimers and @xmath31 denotes the center position of the @xmath10th dimer . here , the inter - dimer / intra - dimer ( @xmath32/@xmath33 ) correlations are detected by the summation / difference in charge and spin densities within each dimer , @xmath34 and @xmath35 with @xmath36 , respectively , where even ( odd ) number is labeled as the site for the left ( right ) side in a dimer , as shown in fig .
[ fig.1 ] ( a ) .
and @xmath10 sites @xmath37 ( the site numbers are shown in fig . [ fig.1 ] ( a ) ) at @xmath38 for @xmath39 , and @xmath40 .
lines are connected between sites in a same chain .
( b ) a schematic figure for 2d af state between dimers . ( c ) inter - dimer spin structure factors @xmath41 at @xmath38 for @xmath39 , and @xmath40 .
lines are guides for eyes.,width=283 ] first , we show the inter - chain real space spin - spin correlations @xmath42 at @xmath38 in fig .
[ fig.2](a ) .
the system is in the dimer - mott insulating regime @xcite , where the intra - chain dimerization and @xmath15 lead to a charge - uniform mott insulator . for @xmath43 , the sign of @xmath42 changes each other between dimers ; there exists af correlation as shown in fig .
[ fig.2](b ) , although the inter - chain correlation is weak due to small @xmath44 . with increasing @xmath45 , this af correlation
is diminished , as seen , for example , in the sign change of @xmath46 , for which the sites are connected by @xmath0 . as a result , the af correlation between inter - chain dimers is weakened . in fig .
[ fig.2](c ) , we show the inter - dimer spin structure factors @xmath41 .
it has peaks at @xmath47 , indicating af correlation between dimers . with increasing @xmath45 , @xmath48
are suppressed , while @xmath49 and @xmath50 are enhanced ; @xmath0 operates as the inter - chain spin frustration parameter . and @xmath10 sites @xmath37 ( the site numbers are shown in fig . [ fig.1 ] ( a ) ) at @xmath51 for @xmath39 , and @xmath40 .
lines are connected between sites in a same chain .
( b ) a schematic figure for 2d af state in fco state .
the size of a circle represents the hole density .
( c ) inter - dimer spin structure factors @xmath41 at @xmath51 for @xmath39 , and @xmath40 .
lines are guides for eyes.,width=283 ] next , in fig . [
fig.3](a ) , we show @xmath42 at @xmath51 for the co region . for @xmath43 , @xmath42 is large ( note the difference in the scale from fig .
[ fig.2](a ) ) only at odd numbers @xmath52 and its sign changes each other between the neighboring charge rich sites , indicating the enhanced development of 2d af correlations in the presence of fco , as shown in fig .
[ fig.3](b ) @xcite .
when @xmath0 is introduced , the absolute value of @xmath53 becomes appreciably larger , in contrast to the case of @xmath38 .
such a behavior is also manifested in @xmath54 , as shown in fig .
[ fig.3 ] ( c ) , where @xmath55 is seen to become large with increasing @xmath45 , while @xmath49 and @xmath50 become small .
these results indicate that @xmath0 enhances the two - dimensionality in magnetic sector and further stabilizes the 2d af state . and ( b ) inter - dimer spin structure factors @xmath56 as functions of @xmath57 for @xmath39 , and @xmath40 .
lines are guides for eyes.,width=207 ] the role of @xmath0 on the 2d af correlation depends on the degree of co correlation . in fig .
[ fig.4 ] , @xmath58 and @xmath59 are plotted as a function of @xmath57 for @xmath39 , and @xmath40 . for each @xmath0
, we see that @xmath60 and @xmath61 cooperatively develops in @xmath62 ; the fco correlation assists in stabilizing the af state .
the noticeable point is that @xmath0 becomes relevant for the af correlation in the fco state , as seen from the enhancement of @xmath63 by @xmath0 only for @xmath62 , whereas the change is small for @xmath64 . on the other hand
, @xmath0 is irrelevant for the fco correlation , i.e. @xmath60 does not depend on @xmath0 .
for insight into the effects of @xmath0 on the magnetic properties , we consider the strong coupling limits and estimate the leading terms of the spin exchange couplings by perturbation calculations with respect to the transfer integrals . in the dimer - mott insulating state for the case of large @xmath15 and small @xmath57 , the spin - exchange couplings between dimers along @xmath65-axis and @xmath66-axis are given by @xmath67 and @xmath68 with the effective on - site coulomb interaction for the dimer units @xmath69 @xcite .
also , in the presence of @xmath0 there is a diagonal inter - chain spin - exchange coupling @xmath70 .
this term causes the spin frustration between inter - chain dimers as seen from fig .
[ fig.1 ] ( b ) and is enlarged toward @xmath71 ; in fact , the @xmath72 dependence of @xmath73 becomes weak with increasing @xmath45 as shown in fig .
[ fig.2 ] ( c ) .
on the other hand , in the basis of the fco state for the case of large @xmath15 and @xmath57 , the coupling between the nearest - neighbor intra - chain charge rich sites is given by @xmath74 from the forth - order perturbation @xcite .
for @xmath43 , the coupling between inter - chain charge rich sites is given by @xmath75 from the second - order perturbation .
since @xmath76 is suppressed with increasing @xmath57 , @xmath77 can become the same order compared to @xmath76 .
this gives the increase of the two - dimensionality in the magnetic sector and induces 2d af state @xcite . in the presence of @xmath0 ,
there is another additional contribution for the coupling between inter - chain charge rich sites @xmath78 from the third - order perturbation process ( for example , @xmath79 and @xmath80 in fig .
[ fig.1 ] ( a ) ) , which is given by @xmath81 .
here , the minus sign comes from the minus term coupling with the transfer integrals in eq .
( [ ehm ] ) .
since @xmath82 is large compared to @xmath83 , this process can become the same order compared to @xmath77 in the intermediate @xmath57 region . as a result ,
this contribution becomes relevant in the presence of fco and the two - dimensionality in the spin sector increases .
this agrees with the behavior of the @xmath57 dependence of @xmath61 shown in fig .
[ fig.4 ] ( b ) .
we can conclude that not only the enhancement of the two - dimensionality in spin sector by fco but also the largeness of @xmath0 in sbf@xmath2 salt compared to other salts promotes stabilization of the coexisting af and fco state observed in the experiments @xcite . finally , we comment on the role of @xmath0 in the presence of electron - phonon couplings , which is relevant for the sp state in both the dimer - mott and the co states @xcite .
previous works show that , in general , the sp state is suppressed and the 2d af state is stabilized with increasing the inter - chain spin exchange coupling @xmath77 @xcite . in pf@xmath2 salt , it is experimentally indicated that , at finite - temperatures above the first order boundary between the two states , a quantum critical behavior is seen , controlled by a hidden quantum critical point @xcite .
the competition between the sp and the 2d af states occurs as well in the presence of fco @xcite .
the role of @xmath0 may enhance the strong quantum fluctuations and bring about difficulties in determining the magnetic phases in the fco state at low temperatures .
one possibility is that the fluctuation leads to the destabilization of ordered phases in the critical region , and results in a spin - liquid state down to the ground state , in between the sp and the 2d af quantum critical points .
this @xmath0 effect on the magnetic properties with electron - phonon couplings remains as a future problem . in summary
, we have investigated the effects of charge order and the diagonal inter - chain transfer integral on the spin ordering in tmttf@xmath84 .
the diagonal @xmath0 works as spin frustration between the inter - chain dimers in the dimer - mott insulating state and reduces the af correlation . on the other hand , in the presence of fco ,
this spin frustration is released and 2d af correlation is developed ; @xmath0 enhances the two - dimensionality in the spin sector .
this assistance of @xmath0 for the cooperative behavior between the fco and the af states agrees with the experiments in tmttf@xmath1sbf@xmath2 .
we thank m. ogata and y. otsuka for discussions and k. furukawa and t. nakamura for providing us the crystal structure data . this work was supported by a grant - in - aid for scientific research in the priority area of molecular conductors ( no .
20110003 , 20110004 ) from mext , japan .
this material is based upon work supported in part by the national science foundation under grant nos .
dmr-0804625 and dmr-1105531 , and a grant - in - aid for scientific research ( no .
20110003 , 20110004 , 21740270 ) from mext , japan .
04 for recent reviews , _ molecular conductors
_ , chem . rev . * 104 * no .
11 ( 2004 ) ; _ special topics _ , _ organic conductors _ , j. phys .
. jpn . * 75 * no.5 ( 2006 ) ; _ the physics of organic superconductors and conductors _ : a. lebed , springer series in materials science ( 2008 ) ; _ special section _ , _ focus on organic conductors _ , sci .
technol . adv .
mater . * 10 * no.2 ( 2009 ) . c. bourbonnais and d. jrome , in _ the physics of organic superconductors and conductors _ , edited by a. lebed , springer series in materials science ( 2008 ) .
w. yu , f. zhang , f. zamborszky , b. alavi , a. baur , c. a. merlic , and s. e. brown , phys .
b * 70 * , 121101(r ) ( 2004 ) .
m. itoi , c. araki , m. hedo , y. uwatoko , and t. nakamura , j. phys .
soc . jpn . *
77 * , 023701 ( 2008 ) .
f. iwase , k. sugiura , k. furukawa , and t. nakamura , j. phys .
78 * , 104717 ( 2009 ) .
h. seo , j. merino , h. yoshioka , and m. ogata , j. phys .
. jpn . * 75 * , 051109 ( 2006 ) .
y. otsuka , h. seo , y. motome , and t. kato , j. phys .
74 * , 3238 ( 2005 ) ; physica b * 404 * , 479 ( 2009 ) . k. yoshimi , h. seo , s. ishibashi , and s. e. brown , arxiv:1110.3573 , to be published in physical review letters .
p. e. bl@xmath85chl , phys .
b * 50 * , 17953 ( 1994 ) .
j. p. predew , k. burke , and m. ernzerhof , phys .
. lett . * 77 * , 3865 ( 1996 ) .
y. ohta , k. tsutsui , w. koshibae , and s. maekawa , phys .
b * 50 * , 13594 ( 1994 ) .
r. t. clay , s. mazumdar , and d. k. campbell , phys .
b * 67 * , 115121 ( 2003 ) .
m. kuwabara , h. seo , and m. ogata , j. phys .
soc . jpn . *
72 * , 225 ( 2003 ) .
s. inagaki and h. fukuyama , j. phys .
. jpn . * 52 * , 3620 ( 1983 ) .
d. s. chow , p. wzietek , d. fogliatti , b. alavi , d. j. tantillo , c. a. merlic , and s. e. brown phys .
81 , 3984 ( 1998 ) . | we theoretically investigate the effects of charge order and spin frustration on the spin ordering in tmttf salts . using first - principles band calculations ,
we find that a diagonal inter - chain transfer integral @xmath0 , which causes spin frustration between the inter - chain dimers in the dimer - mott insulating state , strongly depends on the choice of anion . within the numerical lanczos exact diagonalization method ,
we show that the ferroelectric charge order changes the role of @xmath0 from the spin frustration to the enhancement of the two - dimensionality in spin sector .
the results indicate that @xmath0 assists the cooperative behavior between charge order and antiferromagnetic state observed in tmttf@xmath1sbf@xmath2 . , , , and charge ordering , spin frustration , exact diagonalization method , tmttf salts 71.10.fd , 71.20.rv , 71.30 , 75.30.kz |
ngc 869 and ngc 884 ( @xmath0 and @xmath1 persei , respectively ) are a well known double cluster rich in massive b - type stars , and have been the focus of many studies over the years .
recent studies show that ngc 869 and ngc 884 have nearly identical ages of @xmath8 1314 myr , common distance moduli of dm @xmath8 11.85 , and common reddenings of e(b - v ) @xmath8 0.50.55 ( ( * ? ? ?
* currie et al . 2009 ) , ( * ? ? ?
* slesnick et al . 2002 ) , ( * ? ? ?
* bragg & kenyon 2005 ) ) .
* currie et al .
( 2008 ; hereafter c08 ) ) identified two populations of ngc 869 and ngc 884 stars with detected spitzer mips-24 @xmath9 m excess emission : 20 a and f - type stars with luminous debris disk emission and 57 brighter , earlier stars with weaker excess emission .
they identify most of the latter group as candidate be stars .
however , only 21 were previously listed as be stars ( eg .
* bragg & kenyon 2002 ) , ( * ? ? ?
* slesnick et al . 2002 ) ) . in this study , we analyze blue optical spectra of 92 early - type cluster members , including 16 candidate be stars from c08 , and investigate their near - to - mid infrared ( ir ) excesses . with continued monitoring of these stars in the both the optical and ir regimes , we hope to explore these excesses as a reasonable means for identifying potential be stars within clusters , as well as to investigate the transient natures of the disks surrounding the known be stars in ngc 869 and ngc 884 .
blue optical spectra of 92 members of ngc 869 and ngc 884 were obtained on 2005 november 1415 using the wiyn 3.5 m telescope with the hydra spectrograph .
the observed spectra cover a wavelength range of 42504900 . shown in figure 1
are samples of the model spectral fits used to measure values for @xmath2 sin @xmath3 , @xmath4 , and log @xmath10 for b - type stars .
@xmath2 sin @xmath3 was determined by comparing the he i @xmath114387 , 4471 , 4713 , and mg ii @xmath124481 lines with the kurucz atlas9 models ( ( * ? ? ?
* kurucz 1994 ) ) and taking a weighted average of these four values . for stars having @xmath4
@xmath13 15000 k , the tlusty bstar2006 models ( ( * ? ? ?
* lanz & hubeny 2007 ) ) were used to find @xmath4 and log @xmath10 using the h@xmath14 line . for stars having @xmath4
@xmath15 15000 k , the kurucz atlas9 models were used ( ( * ? ? ? * kurucz 1994 ) ) .
the method of ( * ? ? ?
* huang & gies ( 2006 ; hereafter hg06 ) ) was used to determine log @xmath5 . for
be stars , strmgren photometry available from the webda database was used to derive @xmath4 and log @xmath5 based on the methods of ( * ? ? ?
* mcswain et al . ( 2008 ) ) .
the masses and radii for all stars were determined from the schaller et al .
( 1992 ) evolutionary tracks , which are shown plotted with @xmath4 and log @xmath5 in figure 2 . and on the right is he i @xmath124378 .
the solid line is our observed spectrum while the dashed line displays our model fit to the line , with the computed residual shown above , shifted for clarity .
, title="fig : " ] and on the right is he i @xmath124378 .
the solid line is our observed spectrum while the dashed line displays our model fit to the line , with the computed residual shown above , shifted for clarity .
, title="fig : " ] and log @xmath5 are plotted with the evolutionary tracks of ( * ? ? ?
* schaller et al . ( 1992 ) ) .
the zams mass of each evolutionary track is labeled along the bottom .
normal b - type stars are shown as open diamonds while be stars are filled diamonds .
, title="fig : " ] and log @xmath5 are plotted with the evolutionary tracks of ( * ? ? ?
* schaller et al . ( 1992 ) ) .
the zams mass of each evolutionary track is labeled along the bottom .
normal b - type stars are shown as open diamonds while be stars are filled diamonds .
, title="fig : " ]
sixteen be candidates from c08 are present in our sample or that of hg06 .
three of these 16 stars show no evidence of circumstellar emission in our spectra , though all have been observed to be be stars in the past ( ( * ? ? ?
* keller et al . 2001 ) ) .
ten of the c08 be candidates in our spectra do show emission .
stellar parameters for the remaining three candidates are found in hg06 , thus we can not comment on the presence of emission at the time of observation .
in addition , we find be emission in one star ( no .
1772 ) that was not observed by c08 , and we present results for one additional star ( no .
1268 ) identified as a be star by ( * ? ? ?
* keller et al . ( 2001 ) ) .
these results are summarized in table 1 .
.measured physical parameters for be stars [ cols="<,^,^,^,^,^,^ , < , < " , ] + @xmath16 identification numbers from the webda database .
@xmath17 be candidate in c08 .
@xmath18 stars not showing emission in our observations are likely transient be stars .
@xmath19 identified as be star by ( * ? ? ?
* keller et al . ( 2001 ) ) .
[ sed ] spectral energy distributions ( seds ) for three stars in ngc 869 and ngc 884 are displayed in figure 3 .
ubv magnitudes are from the webda database , jhk@xmath20 are from the 2mass survey , and spitzer [ 8 ] and [ 24 ] @xmath9 m are from c08 .
these magnitudes were then converted to fluxes via the methods detailed in ( * ? ? ?
* bessell et al . ( 1998 ) ) , ( * ? ? ?
* cohen et al . ( 2003 ) ) , ( * ? ? ?
* colina et al . ( 1996 ) ) , ( * ? ? ?
* reach et al .
( 2005 ) ) , and ( * ? ? ?
* rieke et al . ( 2008 ) ) . assuming a constant @xmath21 = 0.52 for ngc 869 and ngc 884 ( ( * ? ? ?
* bragg & kenyon 2005 ) , ( * ? ? ?
* slesnick et al . 2002 ) ) , reddened blackbody curves have been overlaid with these plots to investigate their near - to - mid ir excesses .
all three stars shown in figure 3 are proposed be candidates ( c08 ) , with ngc 884 - 2138 and ngc 869 - 49 having emission present in our optical spectra and observed near - to - mid ir excess .
ngc 884 - 2165 is not in our spectroscopic sample but has previously been identified as a be star and has observed ir excess ( ( * ? ? ? * keller et al .
we have measured the physical parameters of 77 b - type stars and 15 be stars in ngc 896 and ngc 884 .
sixteen be candidates from c08 are present in our sample or that of hg06 .
of these 16 be candidates , 3 stars show no evidence of emission in our optical data and are likely transient be stars .
ten of these be candidates do show emission in our spectra . those be candidates without emission in our spectra should be monitored in the future to further investigate their transient nature . in the future ,
irac 3.6 - 5.8 @xmath9 m data will be combined with the optical and ir fluxes used here to investigate the observed seds .
we will fit the new seds using modern flux models rather than blackbody curves .
modifications accounting for variable reddening throughout the clusters will also be made .
these new sed fits can then be used to model the be disk sizes and temperatures .
we are grateful for travel support provided by the american astronomical society and the international astronomical union .
we would like to thank yale university for providing access to the wiyn telescope at kpno .
institutional support was provided by lehigh university .
this work was also supported by nasa dpr number nnx08av70 g .
rieke , g. h. , blaylock , m. , decin , l. , engelbracht , c. , ogle , p. , avrett , e. , carpenter , j. , cutri , r. m. , armus , l. , gordon , k. , gray , r. o. , hinz , j. , su , k. , & willmer , c. n. a. 2008 , _ aj _ , 135 , 2245 | we present blue optical spectra of 92 members of @xmath0 and @xmath1 per obtained with the wiyn telescope at kitt peak national observatory . from these spectra ,
several stellar parameters were measured for the b type stars , including @xmath2 sin @xmath3 , @xmath4 , log @xmath5 , @xmath6 , and @xmath7 .
strmgren photometry was used to measure @xmath4 and log @xmath5 for the be stars .
we also analyze photometric data of cluster members and discuss the near - to - mid ir excesses of be stars . |
unwanted noise associated with absorption and scattering in high-@xmath0 cavities usually plays a crucial role in experiments in cavity quantum electrodynamics ( cavity qed ) @xcite .
even small values of the corresponding absorption / scattering coefficients may lead to dramatic changes of the quantum properties of the radiation . for typical high-@xmath0 cavities
the unwanted losses can be of the same order of magnitude as the wanted , radiative losses due to the input output coupling @xcite .
in such a case the process of quantum - state extraction from a high-@xmath0 cavity is characterized by efficiency of about 50% , @xcite .
this feature gives a serious restriction for the implementation of many proposals in cavity qed .
particularly , nowadays a lot of schemes for quantum - state engineering of the intracavity field are known .
for example , in ref .
@xcite a scheme for the generation of an arbitrary quantum state of the field is proposed .
also schemes for the generation of entangled states are known @xcite .
unfortunately , due to the small efficiency of the quantum - state extraction , the states of the field may lose essential nonclassical properties after escaping from the cavity . in the framework of quantum noise theories ( qnt ) ,
a high-@xmath0-cavity mode is usually considered as a harmonic oscillator interacting through the coupling mirror with a number of external modes .
this leads to the description of the cavity mode in terms of a quantum langevin equation and input output relation @xcite .
the same result can be obtained in a quantum - field theoretical ( qft ) approach @xcite-@xcite in appropriate limits . for the description of unwanted losses of the intracavity field ,
it is sufficient to suppose the existence of a non - radiative input output channel associated with absorption and scattering processes @xcite .
however this model works properly , with respect to the output field , only in cases when the input ports of the cavity are unused .
indeed , it is clear that an input field can be absorbed or scattered in the coupling mirror .
a detailed analysis shows that for a complete description of the unwanted noise one should take into account also the absorption / scattering losses inside the coupling mirrors @xcite . as we will show below
, the unwanted noise can be modeled by introducing blocks of beam splitters in an appropriately chosen replacement scheme , leading to additional noise terms in the quantum langevin equation and the input
output relation .
such a description allows one to have a clear geometrical interpretation of the operators of unwanted noise as vectors in a unitary space .
the minimum dimension of this space depends on the number of input output ports .
for example , for a one sided - cavity a two - dimensional space is sufficient , two - sided cavities require a three dimensional space and so on .
the requirement of preserving equal - time commutation rules leads to the conclusion that the @xmath1-number coefficients in the quantum langevin equation and the input output relations satisfy several constraints .
in other words , their values belong to a certain multidimensional manifold .
therefore , the problem of consistency and completeness of the corresponding replacement scheme can be solved by applying differential geometry @xcite . in the present paper
we formulate conditions of completeness for replacement schemes modeling unwanted noise in high-@xmath0 cavities and consider several examples .
the paper is organized as follows . in sec .
[ completeness ] the mathematical model of unwanted noise in one - sided cavities is introduced .
examples of replacement schemes are considered in sec .
[ replacementschemes ] and their applicability is discussed .
cavities with two and more input output ports are considered in sec .
[ twosidedcavities ] .
a summary and some concluding remarks are given in sec .
[ conclusions ] .
let us start our consideration by considering a one - sided cavity and reminding the standard qnt model , which does not include channels of unwanted noise @xcite .
in this case the cavity - mode operator @xmath2 obeys the quantum langevin equation @xmath3 where @xmath4 is a resonant frequency of the cavity , @xmath5 is the cavity decay rate , @xmath6 is the complex transmission coefficient describing injection of an input field into the cavity , and @xmath7 is the input - field operator satisfying the commutation rule @xmath8=\delta(t_1-t_2 ) .
\label{incommutator}\ ] ] the output - field operator @xmath9 satisfying the commutation rule @xmath10=\delta(t_1-t_2 ) \label{outcommutator}\ ] ] is connected to the cavity - mode operator @xmath11 and the input - field operator @xmath7 via the input
output relation @xmath12 here @xmath13 is the complex transmission coefficient so that extraction of the cavity field becomes possible ref .
@xcite , and @xmath14 is the complex reflection coefficient at the cavity . the solution of the quantum langevin equation ( [ qlestandard ] ) can be written in the form @xmath15 inserting it into the input
output relation ( [ iorstandard ] ) , one obtains an expression for the output - field operator in terms of the input - field operator and the operator of the cavity mode at the initial time , @xmath16 assuming that the cavity - mode operator obeys the standard bosonic commutation relation at all times @xmath17=1 , \label{cavcommutator}\ ] ] and that the input - field operator commutes with the cavity - mode operator at the initial time @xmath18 , using eqs .
( [ incommutator ] , [ qlestandardsolution ] , [ cavcommutator ] ) one obtains the constraint @xmath19 in a similar way from eqs .
( [ incommutator ] , [ outcommutator ] , [ iorstandardextended ] , [ cavcommutator ] ) we can see that the conditions @xmath20 and @xmath21 are satisfied .
a consequence of these constraints is the fact that the reflection coefficient @xmath14 can be expressed in terms of @xmath13 and @xmath6 as @xmath22 in summary , the qnt model of a cavity without unwanted noise includes the quantum langevin equation ( [ qlestandard ] ) , the input output relation ( [ iorstandard ] ) and the constraints for the @xmath1-number coefficients ( [ constraint1]-[constraint3 ] ) . in particular ,
the constraint ( [ constraint1 ] ) describes the relation between the cavity decay rate and the coefficient @xmath6 .
a consequence of the constraints is the equality of the absolute values of the transmission coefficients @xmath13 and @xmath6 .
as it has already been mentioned in the introduction , unwanted noise can be included in the qnt model through introduction of additional noise terms . indeed , for a description of absorption and scattering of the cavity field
, one can consider the quantum langevin equation @xmath23\hat{a}_\mathrm{cav } + \,{\mathcal{t}}^\mathrm{(c)}\hat{b}_\mathrm{in}\left(t\right ) + \hat{c}^{(c)}\left(t\right ) .
\label{qlenoise}\ ] ] accordingly , the possibility of absorption and scattering of the input field is described by introduction of an additional term into the input
output relation @xmath24 in these equations , the operators of unwanted noise , @xmath25 and @xmath26 , which commute with the input - field operator @xmath7 , the cavity - field operator at the initial time @xmath18 , @xmath27 , satisfy the following commutation rules : @xmath28= \left|{\mathcal{a}}^{(c)}\right|^2\delta(t_1-t_2),\label{commutatorc1}\ ] ] @xmath29= \left|{\mathcal{a}}^{(o)}\right|^2\delta(t_1-t_2),\label{commutatorc2}\ ] ] @xmath30= \xi\,\delta(t_1-t_2).\label{commutatorc3}\ ] ] here , @xmath31 , @xmath32 , and @xmath33 , are @xmath1-number absorption / scattering coefficients .
the set of the coefficients @xmath5 , @xmath4 , @xmath34 , @xmath35 , @xmath36 , @xmath37 , @xmath38 , and @xmath33 characterizes the cavity with unwanted noise .
similar to the case of an idealized cavity , one can show that the requirement of preserving the commutation rules leads to the following constraints : @xmath39 @xmath40 @xmath41 hence , we conclude that the @xmath1-number coefficients describing a realistic cavity should belong to the manifold defined by eqs .
( [ constraintnoise1]-[constraintnoise3 ] ) .
as it is well known from differential geometry , each manifold can be described by means of independent parameters @xcite .
particularly , this means that the @xmath1-number coefficients can be expressed in terms of appropriately chosen parameters .
the corresponding parametrization should cover the whole manifold .
in other words , it should describe all possible cavities .
otherwise , one gets a degenerate model , which describes just a restricted class of the cavities . the easiest way for the parametrization of the considered manifold follows directly from eqs .
( [ constraintnoise1]-[constraintnoise3 ] ) .
indeed , the coefficients @xmath5 , @xmath4 , @xmath34 , @xmath35 , @xmath36 can be considered as independent parameters , and the coefficients @xmath37 , @xmath38 , @xmath33 are expressed in terms of them .
such a parametrization can be convenient in experimental investigations , because the corresponding parameters have a clear physical interpretation .
however , in the theoretical investigation one should use this parametrization very carefully .
in fact , not all values of these independent parameters belong to the manifold .
particularly , their values are necessarily restricted by the inequalities @xmath42 @xmath43 @xmath44 @xmath45 which are derived from eqs .
( [ constraintnoise1]-[constraintnoise3 ] ) .
the fact that the parameters satisfy these inequalities is not a sufficient condition for describing realistic cavities .
however , if they do not satisfy these conditions , one can conclude that the chosen values are not physically consistent ones .
it is worth noting that the operators of unwanted noise , @xmath46 and @xmath47 , can be considered as two vectors in a unitary vector space . in particular , the scalar product of two arbitrary vectors @xmath48 and @xmath49 in this space can be defined by @xmath50.\label{scalarproduct}\ ] ] in this interpretation , @xmath51 and @xmath52 can be considered as the absolute values of the vectors @xmath53 and @xmath54 , respectively , whereas @xmath33 defines the ( complex ) angle between them , i.e. , @xmath55 where @xmath56 and @xmath57 are real .
the vectors @xmath53 and @xmath54 can be expanded in an orthogonal basis , @xmath58 which implies that different representations of the operators of the unwanted noise can be obtained .
it is clear ( in full analogy with usual geometry ) that two vectors always belong to a two - dimensional plane , see fig .
[ figgeom ] .
this plane may be considered as a two - dimensional unitary vector space , with the two basis vectors @xmath59 and @xmath60 playing the role of appropriately chosen operators of the unwanted noise .
consequently , it is sufficient to have two basis operators for a complete description of the unwanted noise of a ( one - sided ) cavity .
however , as we will show in the following , in some cases it is convenient to use representations with a lager number of dimensions , with particular emphasis on the three - dimensional case .
the method of replacement schemes , widely used in quantum optics , can be applied to the formulation of a parametrization which works with all the values of independent parameters .
the main idea is to simulate the channels of unwanted noise by an additional input output port and a block of beam splitters as illustrated in fig .
[ replscheme ] .
the additional input output port models the losses typically responsible for the absorption and scattering of the radiation , which escapes from the cavity , whereas the block of beam splitters describes losses like absorption and scattering of the input field entering the cavity . in fig .
[ replscheme ] , the symmetrical beam - splitters @xmath61 and @xmath62 simulate the unwanted noise inside the coupling mirror .
moreover , due to the requirements of completeness one must include in the scheme the asymmetrical @xmath63 beam - splitter @xmath64 , which simulates feedback . using the quantum langevin equation and the input
output relation for a cavity with two input output ports as well as the input output relations for each beam splitter separately , we obtain eqs .
( [ iorstandard ] , [ qlenoise ] ) , where the operators @xmath53 and @xmath54 have the form @xmath65 @xmath66 the @xmath1-number coefficients in the quantum langevin equation and the input output relation are expressed in terms of the beam - splitter transmission and reflection coefficients @xmath67 and @xmath68 respectively ( @xmath69 ) , the phase factor @xmath70 of the beam splitter @xmath64 , the resonance frequency @xmath71 , the radiation and absorption decay rates of the `` primary '' cavity in the scheme , @xmath72 and @xmath73 respectively , as follows : @xmath74 @xmath75 @xmath76 @xmath77 @xmath78 @xmath79 @xmath80 @xmath81 @xmath82 each complex coefficient @xmath67 and @xmath68 can be expressed in terms of three real independent parameters @xmath83 , @xmath84 and @xmath85 : @xmath86 therefore , one gets the parametrization of the manifold by means of the independent parameters . in order to check the completeness of the proposed parametrization , one should first present eqs .
( [ gamma]-[ao ] ) in the form of real functions of real parameters .
next , one should build the matrix containing the first derivatives of these functions .
the determinant of this function should be non - zero .
the corresponding calculations have been performed by using mathematica 5.1 .
it has been found that the parametrization corresponding to the considered replacement scheme and given by eqs .
( [ gamma]-[ao ] ) completely describes ( one - sided ) cavities with unwanted noise .
let us consider examples of replacement schemes , referred to as degenerate replacement schemes , which do not describe all possible cavities .
as a rule , degenerate schemes can be obtained from the complete scheme by removing one or more elements .
cavities modeled by degenerate replacement schemes usually obey some constraints in addition to eqs .
( [ constraintnoise1]-[constraintnoise3 ] ) . in other words ,
such cavities correspond to points in a certain sub - manifold rather than in the whole manifold .
the first example is the class of cavities obtained from fig .
[ replscheme ] by removing the beam splitters @xmath61 and @xmath62 . for such cavities ,
the input field does not suffer from losses when it enters the cavity .
the operators @xmath53 and @xmath54 in the quantum langevin equation and the input output relation in this case read as @xmath87 @xmath88 it is clear that for such a cavity the noise term associated with absorption and scattering is only included in the quantum langevin equation a model , which can be used for special applications @xcite .
the corresponding parametrization is a rather trivial one , the additional constraint has the form @xmath89 a consequence of this fact is that @xmath90 .
moreover , such a cavity has some properties very close to the idealized cavity without channels of unwanted noise .
the transmission coefficients @xmath13 and @xmath6 are equal and eq .
( [ roexpression ] ) holds true for the reflection coefficient @xmath14 .
another example of a degenerate scheme can be obtained from the complete scheme in fig .
[ replscheme ] by removing the ( non - radiative ) input output channels @xmath91 , @xmath92 and the beam splitter @xmath64 associated with the feedback . in this case , the operators @xmath53 and @xmath54 have the form @xmath93 @xmath94 although both the quantum langevin equation and the input output relation contain noise terms associated with unwanted losses , the scheme is not a complete one .
the corresponding parametrization can be written in the form @xmath95
@xmath96 @xmath97 @xmath98 @xmath99 @xmath100 @xmath101 @xmath102 one can easily prove by direct calculations that along with eqs .
( [ constraintnoise1]-[constraintnoise3 ] ) this parametrization satisfy the following additional constraint @xmath103 it is worth noting that the physics behind this degenerate scheme is closely related to that of a cavity without unwanted noise .
indeed , the unwanted noise can be regarded as noise associated with the transmission channel .
hence , the losses modeled in this way can not affect the decay rate of the intracavity field , but some properties of the external field are changed .
particularly , for such cavities it is impossible to combine a cavity mode and an input mode in an output mode .
so far we have considered one - sided cavities . in various physical applications ,
e.g. , the generation of squeezed light in optical parametric amplification @xcite , it is necessary to consider the problem of unwanted noise in cavities with two ( or even more ) radiative input output ports .
let us generalize the replacement - scheme method developed above to a two - sided cavity , as is sketched in fig .
[ figtwosides ] .
it is straightforward to show that the generalization of eqs .
( [ qlenoise ] ) and ( [ iornoise ] ) is @xmath104\hat{a}_\mathrm{cav } + { \mathcal{t}}^\mathrm{(c)}_\mathrm{(r)}\hat{b}^\mathrm{(r)}_\mathrm{in}(t ) + { \mathcal{t}}^\mathrm{(c)}_\mathrm{(l)}\hat{b}^\mathrm{(l)}_\mathrm{in}(t ) \nonumber\\&+&{\mathcal{a}}^\mathrm{(c)}_{(1)}\hat{c}^{(1)}_\mathrm{in}(t ) + { \mathcal{a}}^\mathrm{(c)}_{(2)}\hat{c}^{(2)}_\mathrm{in}(t ) + { \mathcal{a}}^\mathrm{(c)}_{(3)}\hat{c}^{(3)}_\mathrm{in}(t ) + { \mathcal{a}}^\mathrm{(c)}_{(4)}\hat{c}^{(4)}_\mathrm{in}(t ) + { \mathcal{a}}\hat{c}_\mathrm{in}(t),\label{ts5}\end{aligned}\ ] ] @xmath105 @xmath106 where the coefficients in these equations can be obtained in a similar manner as for the case of a one - sided cavity . as in the case of a one - sided cavity , the @xmath1-number coefficients in eqs .
( [ ts5])([ts7 ] ) are also not independent of each other . from considering the commutation relation for the cavity - mode operator ,
one obtains @xmath107 with regard to the right - hand wall of the cavity , the requirement of preserving the commutation rules implies that @xmath108 @xmath109 finally , for the field outgoing from the left - hand wall of the cavity , one can show that @xmath110 @xmath111 it should be pointed out that the operators of unwanted noise represented by @xmath91 , @xmath112 , @xmath113 , @xmath114 and @xmath115 in eqs .
( [ ts5]-[ts7 ] ) can be also represented in other forms . since the corresponding expressions in eqs .
( [ ts5])([ts7 ] ) contain only three linear combinations of the operators @xmath91 , @xmath112 , @xmath113 , @xmath114 and @xmath115 , one can conclude that there exist equivalent formulations of these equations with three independent operators of unwanted noise .
the concept of replacement schemes is a very helpful tool to study , within the framework of qnt , the effect of unwanted noise associated with absorption and scattering in realistic high-@xmath0 cavities , which leads to the appearance of additional noise terms in both the standard quantum langevin equations and the standard input output relations attributed to them .
an important mathematical feature is the fact that the @xmath1-number coefficients in the quantum langevin equations and in the input output relation are not independent ones .
in particular , the requirement of preserving typical commutation rules leads to the appearance of several constraints .
hence , the corresponding values of the coefficients can be regarded as belonging to a certain manifold .
a consistent physical description of realistic cavities requires to formulate the theory in terms of independent parameters only . in other words
, one must consider the parametrization of the manifold .
so , one can formally express some of the @xmath1-number coefficients in terms of the other ones and simply consider the latter as independent parameters .
another , more physical way is a parametrization on the basis of appropriately chosen replacement schemes .
the method of replacement schemes in fact allows one to distinguish , with respect to the unwanted noise , between qualitatively different cavity models . roughly speaking
, one can distinguish between non - degenerate and degenerate replacement schemes .
in contrast to non - degenerate schemes , where the parametrization completely describes cavities with unwanted losses , degenerate schemes do not describe all possible cavities but only special classes .
99 edited by p. berman ( academic , new york , 1994 ) .
hood , h.j .
kimble , and j. ye , phys .
a * 64 * , 033804 ( 2001 ) .
m. khanbekyan , l. knll , a.a .
semenov , w. vogel , and , phys .
a. * 69 * , 043807 ( 2004 ) .
law and j.h .
eberly , phys .
lett . * 76 * , 1055 ( 1995 ) .
w. lange and h. j. kimble , phys .
a * 61 * , 063817 ( 2000 ) .
collett and c.w .
gardiner , phys .
a * 30 * , 1386 ( 1984 ) ; and m.j .
collett , phys .
a * 31 * , 3761 ( 1985 ) .
l. knll , w. vogel , and d .-
welsch , phys .
a * 43 * , 543 ( 1991 ) .
r. w. f. van der plank and l. g. suttorp , phys .
a * 53 * , 1791 ( 1996 ) .
s. m. dutra and g. nienhuis , phys .
a * 62 * , 063805 ( 2000 ) . c. viviescas , and g. hackenbroich , phys .
a * 67 * , 013805 ( 2003 ) ; j. opt .
b * 6 * , 211 ( 2004 ) . m. khanbekyan , d .-
welsch , a. a. semenov and w. vogel , j. opt .
b * 7 * s689 ( 2005 ) .
dubrovin , a.t .
fomenko , and s.p .
novikov , _ modern geometry methods and applications _
( springer - verlag , new york , 1992 ) .
wu , m. xiao , and h.j .
kimble , j. opt .
b * 4 * , 1465 ( 1987 ) . | the problem of the description of absorption and scattering losses in high-@xmath0 cavities is studied .
the considerations are based on quantum noise theories , hence the unwanted noise associated with scattering and absorption is taken into account by introduction of additional damping and noise terms in the quantum langevin equations and input output relations .
completeness conditions for the description of the cavity models obtained in this way are studied and corresponding replacement schemes are discussed .
pacs : 42.50.lc , 42.50.nn , 42.50.pq |
the principle of successively minimizing upper bounds of the objective function is often called _ majorization - minimization _ @xcite or _ successive upper - bound minimization _ @xcite .
each upper bound is locally tight at the current estimate , and each minimization step decreases the value of the objective function . even though this principle does not provide any theoretical guarantee about the quality of the returned solution
, it has been very popular and widely used because of its simplicity .
various existing approaches can indeed be interpreted from the majorization - minimization point of view .
this is the case of many gradient - based or proximal methods @xcite , expectation - maximization ( em ) algorithms in statistics @xcite , difference - of - convex ( dc ) programming @xcite , boosting @xcite , some variational bayes techniques used in machine learning @xcite , and the mean - shift algorithm for finding modes of a distribution @xcite .
majorizing surrogates have also been used successfully in the signal processing literature about sparse estimation @xcite , linear inverse problems in image processing @xcite , and matrix factorization @xcite . in this paper , we are interested in making the majorization - minimization principle scalable for minimizing a large sum of functions : @xmath0 , \label{eq : prob}\ ] ] where the functions @xmath1 are continuous , and @xmath2 is a convex subset of @xmath3 .
when @xmath4 is non - convex , exactly solving ( [ eq : prob ] ) is intractable in general , and when @xmath4 is also non - smooth , finding a stationary point of ( [ eq : prob ] ) can be difficult .
the problem above when @xmath5 is large can be motivated by machine learning applications , where @xmath6 represents some model parameters and each function @xmath7 measures the adequacy of the parameters @xmath6 to an observed data point indexed by @xmath8 . in this context , minimizing @xmath4 amounts to finding parameters @xmath6 that explain well some observed data . in the last few years
, stochastic optimization techniques have become very popular in machine learning for their empirical ability to deal with a large number @xmath5 of training points @xcite .
even though these methods have inherent sublinear convergence rates for convex and strongly convex problems @xcite , they typically have a cheap computational cost per iteration , enabling them to efficiently find an approximate solution .
recently , incremental algorithms have also been proposed for minimizing finite sums of functions @xcite . at the price of a higher memory cost than stochastic algorithms ,
these incremental methods enjoy faster convergence rates , while also having a cheap per - iteration computational cost .
our paper follows this approach : in order to exploit the particular structure of problem ( [ eq : prob ] ) , we propose an incremental scheme whose cost per iteration is independent of @xmath5 , as soon as the upper bounds of the objective are appropriately chosen .
we call the resulting scheme `` miso '' ( _ minimization by incremental surrogate optimization _ ) . we present convergence results when the upper bounds are chosen among the class of `` first - order surrogate functions '' , which approximate the objective function up to a smooth error that is , differentiable with a lipschitz continuous gradient . for non - convex problems
, we obtain almost sure convergence and asymptotic stationary point guarantees .
in addition , when assuming the surrogates to be strongly convex , we provide convergence rates for the expected value of the objective function . remarkably ,
the convergence rate of miso is linear for minimizing strongly convex composite objective functions , a property shared with two other incremental algorithms for smooth and composite convex optimization : the _ stochastic average gradient _ method ( sag ) of schmidt , le roux and bach @xcite , and the _ stochastic dual coordinate ascent _ method ( sdca ) of shalev - schwartz and zhang @xcite .
our scheme miso is inspired in part by these two works , but yields different update rules than sag or sdca , and is also appropriate for non - convex optimization problems . in the experimental section of this paper ,
we show that miso can be useful for solving large - scale machine learning problems , and that it matches cutting - edge solvers for large - scale logistic regression @xcite .
then , we show that our approach provides an effective incremental dc programming algorithm , which we apply to sparse estimation problems with nonconvex penalties @xcite .
the paper is organized as follows : section [ sec : generic ] introduces the majorization - minimization principle with first - order surrogate functions . section [ sec : incremental ] is devoted to our incremental scheme miso .
section [ sec : exp ] presents some numerical experiments , and section [ sec : ccl ] concludes the paper .
some basic definitions are given in appendix [ appendix : background ] .
in this section , we present the generic majorization - minimization scheme for minimizing a function @xmath4 without exploiting its structure that is , without using the fact that @xmath4 is a sum of functions .
we describe the procedure in algorithm [ alg : generic_batch ] and illustrate its principle in figure [ fig : mm ] . at iteration
@xmath9 , the estimate @xmath10 is obtained by minimizing a surrogate function @xmath11 of @xmath4 . when @xmath11 uniformly majorizes @xmath4 and when @xmath12 , it is clear that the objective function value monotonically decreases .
@xmath13 ( initial estimate ) ; @xmath14 ( number of iterations ) .
compute a surrogate function @xmath11 of @xmath4 near @xmath15 ; minimize the surrogate and update the solution : @xmath16 @xmath17 ( final estimate ) ; ( -5,-0.5)(5,4 ) ( -5,0)(5,0 ) ( 0,-0.5)(0,3 ) ( -4,0.9)(-4,2.5 ) ( 3.84,2.4 ) @xmath18 ( -3.33,1.9 ) @xmath19 ( -5.3,1.6 ) @xmath20 ( 0.62,1.40)(-1.3,0.3 ) ( 0.9,1.2 ) @xmath15 ( -1.4,0.6 ) @xmath21 ( 2.1,1.0 ) @xmath22 for this approach to be effective , intuition tells us that we need functions @xmath11 that are easy to minimize and that approximate well the objective @xmath4 .
therefore , we measure the quality of the approximation through the smoothness of the error @xmath23 , which is a key quantity arising in the convergence analysis . specifically , we require @xmath24 to be @xmath25-smooth for some constant @xmath26 in the following sense : [ def : lsmooth ] a function @xmath27 is called @xmath25-smooth when it is differentiable and when its gradient @xmath28 is @xmath25-lipschitz continuous . with this definition in hand
, we now introduce the class of `` first - order surrogate functions '' , which will be shown to have good enough properties for analyzing the convergence of algorithm [ alg : generic_batch ] and other variants .
[ def : surrogate_batch ] a function @xmath29 is a first - order surrogate function of @xmath4 near @xmath30 in @xmath2 when @xmath31 for all minimizers @xmath32 of @xmath33 over @xmath2 . when the more general condition @xmath34 holds , we say that @xmath33 is a _ majorizing _ surrogate ; the approximation error @xmath35 is @xmath25-smooth , @xmath36 , and @xmath37 .
we denote by @xmath38 the set of first - order surrogate functions and by @xmath39 the subset of @xmath40-strongly convex surrogates .
first - order surrogates are interesting because their approximation error the difference between the surrogate and the objective can be easily controlled .
this is formally stated in the next lemma , which is a building block of our analysis : [ lemma : basic ] let @xmath33 be a surrogate function in @xmath38 for some @xmath30 in @xmath2 . define the approximation error @xmath35 , and let @xmath32 be a minimizer of @xmath33 over @xmath2 .
then , for all @xmath6 in @xmath2 , * @xmath41 ; * @xmath42 .
assume that @xmath33 is @xmath40-strongly convex , i , e . , @xmath33 is in @xmath43 .
then , for all @xmath6 in @xmath2 , * @xmath44 .
the first inequality is a direct application of a classical result ( lemma 1.2.3 of @xcite ) on quadratic upper bounds for @xmath25-smooth functions , when noticing that @xmath36 and @xmath37 .
then , for all @xmath6 in @xmath2 , we have @xmath45 and we obtain the second inequality from the first one .
when @xmath33 is @xmath40-strongly convex , we use the following classical lower bound ( see @xcite ) : @xmath46 since @xmath47 by definition [ def : surrogate_batch ] and @xmath48 , the third inequality follows from the first one .
we now proceed with a convergence analysis including four main results regarding algorithm [ alg : generic_batch ] with first - order surrogate functions @xmath11 .
more precisely , we show in section [ subsec : nonconvex ] that , under simple assumptions , the sequence of iterates asymptotically satisfies a stationary point condition .
then , we present a similar result with relaxed assumptions on the surrogates @xmath11 when @xmath4 is a composition of two functions , which occur in practical situations as shown in section [ subsec : surrogates ] . finally
, we present non - asymptotic convergence rates when @xmath4 is convex in section [ subsec : convex ] . by adapting convergence proofs of proximal gradient methods @xcite to our more general setting , we recover classical sublinear rates @xmath49 and linear convergence rates for strongly convex problems . for general non - convex problems ,
proving convergence to a global ( or local ) minimum is impossible in general , and classical analysis studies instead asymptotic stationary point conditions ( see , _ e.g. _ , @xcite ) .
to do so , we make the following mild assumption when @xmath4 is non - convex : * @xmath4 is bounded below and for all @xmath50 in @xmath2 , the directional derivative @xmath51 of @xmath4 at @xmath6 in the direction @xmath52 exists . the definitions of directional derivatives and stationary points are provided in appendix [ appendix : background ] .
a necessary first - order condition for @xmath6 to be a local minimum of @xmath4 is to have @xmath53 for all @xmath32 in @xmath2 ( see , _
e.g. _ , @xcite ) .
in other words , there is no feasible descent direction @xmath54 and @xmath6 is a stationary point .
thus , we consider the following condition for assessing the quality of a sequence @xmath55 for non - convex problems : under assumption , a sequence @xmath55 satisfies the asymptotic stationary point condition if @xmath56 note that if @xmath4 is differentiable on @xmath3 and @xmath57 , @xmath58 , and the condition ( [ eq : stationary ] ) implies that the sequence @xmath59 converges to @xmath60 .
as noted , we recover the classical definition of critical points for the smooth unconstrained case .
we now give a first convergence result about algorithm [ alg : generic_batch ] .
[ prop : conv1 ] assume that holds and that the surrogates @xmath11 from algorithm [ alg : generic_batch ] are in @xmath61 and are either majorizing @xmath4 or strongly convex .
then , @xmath62 monotonically decreases , and @xmath63 satisfies the asymptotic stationary point condition . the fact that @xmath64 is non - increasing and convergent because bounded below is clear : for all @xmath65 , @xmath66 , where the first inequality and the last equality are obtained from definition [ def : surrogate_batch ] .
the second inequality comes from the definition of @xmath10 .
let us now denote by @xmath67 the limit of the sequence @xmath68 and by @xmath69 the approximation error function at iteration @xmath9 , which is @xmath25-smooth by definition [ def : surrogate_batch ] and such that @xmath70 .
then , @xmath71 , and @xmath72 thus , the non - negative sequence @xmath73 necessarily converges to zero .
then , we have two possibilities ( according to the assumptions made in the proposition ) . *
if the functions @xmath11 are majorizing @xmath4 , we define @xmath74 , and we use the following classical inequality for @xmath25-smooth functions @xcite : @xmath75 therefore , we may use the fact that @xmath76 because @xmath77 , and @xmath78 * if instead the functions @xmath11 are @xmath40-strongly convex , the last inequality of lemma [ lemma : basic ] with @xmath79 and @xmath80 gives us @xmath81 by summing over @xmath9 , we obtain that @xmath82 converges to zero , and @xmath83 since @xmath84 according to definition [ def : surrogate_batch ] .
we now consider the directional derivative of @xmath4 at @xmath10 and a direction @xmath85 , where @xmath86 and @xmath6 is in @xmath2 , @xmath87 note that @xmath10 minimizes @xmath11 on @xmath2 and therefore @xmath88 .
therefore , @xmath89 by cauchy - schwarz s inequality . by minimizing over @xmath6 and taking the infimum limit
, we finally obtain @xmath90 this proposition provides convergence guarantees for a large class of existing algorithms , including cases where @xmath4 is non - smooth . in the next proposition , we relax some of the assumptions for objective functions that are compositions @xmath91 , where @xmath92 is the composition operator .
in other words , @xmath93 for all @xmath6 in @xmath3 . [
prop : conv1_separable ] assume that holds and that the function @xmath4 is a composition @xmath94 , where @xmath95 is @xmath96-lipschitz continuous for some constant @xmath97 , and @xmath98 .
assume that the function @xmath11 in algorithm [ alg : generic_batch ] is defined as @xmath99 , where @xmath100 is a majorizing surrogate in @xmath101 .
then , the conclusions of proposition [ prop : conv1 ] hold .
we follow the same steps as the proof of proposition [ prop : conv1 ] .
first , it is easy to show that @xmath64 monotonically decreases and that @xmath102 converges to zero when @xmath9 grows to infinity .
note that since we have made the assumptions that @xmath103 and that @xmath94 , the function @xmath104 can be written as @xmath105 , where @xmath106 is @xmath25-smooth .
proceeding as in the proof of proposition [ prop : conv1 ] , we can show that @xmath107 converges to zero .
let us now fix @xmath86 and consider @xmath108 such that @xmath109 is in @xmath2 .
we have @xmath110 where @xmath111 is a vector whose @xmath112-norm is bounded by a universal constant @xmath113 because the function @xmath114 is lipschitz continuous .
since @xmath115 is @xmath25-smooth , we also have @xmath116 plugging this simple relation with @xmath117 , for some @xmath118 and @xmath6 in @xmath2 , into the definition of the directional derivative @xmath119 , we obtain the relation @xmath120 and since @xmath121 , and @xmath88 , @xmath122 in this proposition , @xmath11 is an upper bound of @xmath91 , where the part @xmath114 is lipschitz continuous but @xmath123 is not @xmath25-smooth .
this extension of proposition [ prop : conv1 ] is useful since it provides convergence results for classical approaches that will be described later in section [ subsec : surrogates ] .
note that convergence results for non - convex problems are by nature weak , and our non - convex analysis does not provide any convergence rate .
this is not the case when @xmath4 is convex , as shown in the next section .
the next proposition is based on a proof technique from nesterov @xcite , which was originally designed for the proximal gradient method . by adapting it ,
we obtain the same convergence rates as in @xcite .
[ prop : conv2 ] assume that @xmath4 is convex , bounded below , and that there exists a constant @xmath124 such that @xmath125 where @xmath126 is a minimizer of @xmath4 on @xmath2 . when the functions @xmath11 in algorithm [ alg : generic_batch ] are in @xmath127 , we have for all @xmath86 , @xmath128 where @xmath129 .
assume now that @xmath4 is @xmath130-strongly convex .
regardless of condition ( [ eq : bounded ] ) , we have for all @xmath86 , @xmath131 where @xmath132 if @xmath133 or @xmath134 otherwise .
we successively prove the two parts of the proposition .
+ _ non - strongly convex case : _
let us consider the function @xmath135 at iteration @xmath86 .
by lemma [ lemma : basic ] , @xmath136.\ ] ] then , following a similar proof technique as nesterov in @xcite , @xmath137 } \left [ f(\alpha\theta^\star+(1-\alpha)\theta_{n-1 } ) + \frac{l\alpha^2}{2 } \|\theta^\star-\theta_{n-1}\|_2 ^ 2\right ] \\ & \leq \min_{\alpha \in [ 0,1 ] } \left [ \alpha f(\theta^\star ) + ( 1-\alpha)f(\theta_{n-1 } ) + \frac{l\alpha^2}{2 } \|\theta^\star-\theta_{n-1}\|_2 ^ 2\right ] , \end{split } \label{eq : tmp_rate2}\ ] ] where the minimization over @xmath2 is replaced by a minimization over the line segment @xmath138 $ ] .
since the sequence @xmath64 is monotonically decreasing we may use the bounded level set assumption and we obtain @xmath139 } \left [ ( 1-\alpha)(f(\theta_{n-1 } ) - f^\star ) + \frac{lr^2\alpha^2}{2}\right].\label{eq : tmp_rate3}\ ] ] to simplify , we introduce the notation @xmath140 , and we consider two cases : * _ first case : _ if @xmath141 , then the optimal value @xmath142 in ( [ eq : tmp_rate3 ] ) is @xmath143 and we consequently have @xmath144 ; * _ second case : _ otherwise @xmath145 and @xmath146 .
thus , @xmath147 , where the second inequality comes from the convexity inequality @xmath148 for @xmath149 .
we now apply recursively the previous inequalities , starting with @xmath150 .
if @xmath151 , we are in the first case and then @xmath152 ; then , we will subsequently be in the second case for all @xmath153 and thus @xmath154 . otherwise , if @xmath155 , we are always in the second case and @xmath156 , which is sufficient to obtain the first part of the proposition . _
@xmath130-strongly convex case : _
let us now assume that @xmath4 is @xmath130-strongly convex , and let us drop the bounded level sets assumption .
the proof again follows @xcite for computing the convergence rate of proximal gradient methods .
we start from ( [ eq : tmp_rate2 ] ) .
we use the strong convexity of @xmath4 which implies that @xmath157 , and we obtain @xmath158 } 1-\alpha + \frac{l\alpha^2}{\mu}\right)(f(\theta_{n-1})-f^\star).\ ] ] at this point , it is easy to show that if @xmath159 , the previous binomial is minimized for @xmath160 , and if @xmath161 , then we have @xmath162 . this yields the desired result .
the result of proposition [ prop : conv2 ] is interesting because it does not make any strong assumption about the surrogate functions , except the ones from definition [ def : surrogate_batch ] .
the next proposition shows that slightly better rates can be obtained with additional strong convexity assumptions .
[ prop : conva ] assume that @xmath4 is convex , bounded below , and let @xmath126 be a minimizer of @xmath4 on @xmath2 . when the surrogates @xmath11 of algorithm [ alg : generic_batch ] are in @xmath163 with @xmath164 , we have for all @xmath86 , @xmath165 where @xmath129 .
when @xmath4 is @xmath130-strongly convex , we have for all @xmath65 , @xmath166 as before , we successively prove the two parts of the proposition .
_ non - strongly convex case : _ from lemma [ lemma : basic ] ( with @xmath167 , @xmath168 , @xmath169 , @xmath170 ) , we have for all @xmath65 , @xmath171 after summation , @xmath172 where the first inequality comes from the inequalities @xmath173 for all @xmath174 .
this is sufficient to prove the first part .
note that proving convergence rates for first - order methods by finding telescopic sums is a classical technique ( see , _
e.g._,@xcite ) . _
@xmath130-strongly convex case : _
let us now assume that @xmath4 is @xmath130-strongly convex .
the strong convexity implies that @xmath175 for all @xmath9 .
combined with ( [ eq : rate1_proof ] ) , this yields @xmath176 and thus @xmath177 even though the constants obtained in the rates of proposition [ prop : conva ] are slightly better than the ones of proposition [ prop : conv2 ] , the condition @xmath11 in @xmath43 with @xmath164 is much stronger than the simple assumption that @xmath11 is in @xmath178 . it can indeed be shown that @xmath4 is necessarily @xmath179-strongly convex if @xmath180 , and convex if @xmath181 . in the next section ,
we give some examples where such a condition holds .
we now present practical first - order surrogate functions and links between algorithm [ alg : generic_batch ] and existing approaches .
even though our generic analysis does not always bring new results for each specific case , its main asset is to provide a unique theoretical treatment to all of them .
when @xmath4 is @xmath25-smooth , it is natural to consider the following surrogate : @xmath182 the function @xmath33 is an upper bound of @xmath4 , which is a classical result @xcite .
it is then easy to see that @xmath33 is @xmath25-strongly convex and @xmath25-smooth . as a consequence ,
the difference @xmath183 is @xmath184-smooth ( as a sum of two @xmath25-smooth functions ) , and thus @xmath33 is in @xmath185 . when @xmath4 is convex , it is also possible to show by using lemma [ lemma : convexerror ] that @xmath33 is in fact in @xmath186 , and when @xmath4 is @xmath130-strongly convex , @xmath33 is in @xmath187 .
we remark that minimizing @xmath33 amounts to performing a gradient descent step : @xmath188 .
let us now consider a composite optimization problem , meaning that @xmath4 splits into two parts @xmath189 , where @xmath190 is @xmath25-smooth .
then , a natural surrogate of @xmath4 is the following function : @xmath191 the function @xmath33 majorizes @xmath4 and the approximation error @xmath183 is the same as in section [ subsubsec : gradient ] .
thus , @xmath33 in in @xmath192 or in @xmath185 when @xmath193 is convex .
moreover , * when @xmath190 is convex , @xmath33 is in @xmath38 .
if @xmath193 is also convex , @xmath33 is in @xmath186 ; * when @xmath190 is @xmath130-strongly convex , @xmath33 is in @xmath194 .
if @xmath193 is also convex , @xmath33 is in @xmath187 . minimizing @xmath33 amounts to performing one step of the proximal gradient algorithm @xcite .
it is indeed easy to show that the minimum @xmath32 of @xmath33assuming it is unique can be equivalently obtained as follows : @xmath195,\ ] ] which is often written under the form @xmath196 $ ] , where `` @xmath197 '' is called the `` proximal operator '' @xcite . in some cases ,
the proximal operator can be computed efficiently in closed form , for example when @xmath193 is the @xmath198-norm ; it yields the iterative soft - thresholding algorithm for sparse estimation @xcite . for a review of proximal operators and their computations , we refer the reader to @xcite .
assume that @xmath189 , where @xmath193 is concave and @xmath25-smooth .
then , the following function @xmath33 is a majorizing surrogate in @xmath38 : @xmath199 such a surrogate appears in dc ( difference of convex ) programming @xcite .
when @xmath190 is convex , @xmath4 is indeed the difference of two convex functions .
it is also used in sparse estimation for dealing with some non - convex penalties @xcite .
for example , consider a cost function of the form @xmath200| + \varepsilon)$ ] , where @xmath201 $ ] is the @xmath202-th entry in @xmath6 . even though the functions @xmath203|
+ \varepsilon)$ ] are not differentiable , they can be written as the composition of a concave smooth function @xmath204 on @xmath205 , and a lipschitz function @xmath206|$ ] . by upper - bounding the logarithm function by its linear approximation , proposition [ prop : conv1_separable ]
justifies the following surrogate : @xmath207| + \varepsilon ) + \lambda\sum_{j=1}^p \frac{|\theta[j]|-|\kappa[j]|}{|\kappa[j]|+\varepsilon } , \label{eq : upperbounddc}\ ] ] and minimizing @xmath33 amounts to performing one step of a reweighted-@xmath198 algorithm ( see @xcite and references therein ) . similarly , other penalty functions are adapted to this framework .
for instance , the logarithm can be replaced by any smooth concave non - decreasing function , or group - sparsity penalties @xcite can be used , such as @xmath208 , where @xmath209 is a partition of @xmath210 and @xmath211 records the entries of @xmath6 corresponding to the set @xmath33 .
proposition [ prop : conv1_separable ] indeed applies to this setting .
let us now consider a real - valued function @xmath4 defined on @xmath212 .
let @xmath213 and @xmath214 be two convex sets . minimizing @xmath4 over @xmath215 is equivalent to minimizing the function @xmath216 over @xmath217 defined as @xmath218 .
assume now that * @xmath219 is @xmath130-strongly convex for all @xmath220 in @xmath221 ; * @xmath222 is differentiable for all @xmath223 ; * @xmath224 is @xmath225-lipschitz with respect to @xmath220 and @xmath25-lipschitz with respect to @xmath223 .
denotes the gradient with respect to @xmath220 . ]
let us fix @xmath226 in @xmath217 .
then , the following function is a majorizing surrogate in @xmath227 : @xmath228 with @xmath229 .
we can indeed apply lemma [ lemma : danskin ] , which ensures that @xmath216 is differentiable with @xmath230 and @xmath231 for all @xmath220 .
moreover , @xmath33 is @xmath225-smooth and @xmath216 is @xmath232-smooth according to lemma [ lemma : danskin ] , and thus @xmath233 is @xmath234-smooth .
note that a better constant @xmath235 can be obtained when @xmath4 is convex , as noted in the appendix of @xcite .
the surrogate @xmath33 leads to an alternate minimization algorithm ; it is then interesting to note that proposition [ prop : conv2 ] provides similar convergence rates as another recent analysis @xcite , which makes slightly different assumptions on the function @xmath4 .
variational surrogates might also be useful for problems of a single variable @xmath220 .
for instance , consider a regression problem with a huber loss function @xmath236 defined for all @xmath237 in @xmath238 as @xmath239 where @xmath108 is a positive constant .. ] the huber loss can be seen as a smoothed version of the @xmath198-norm when @xmath108 is small , or simply a robust variant of the squared loss @xmath240 that asymptotically grows linearly .
then , it is easy to show that @xmath241 .
\label{eq : huber_var}\ ] ] consider now a regression problem with @xmath242 training data points represented by vectors @xmath243 in @xmath3 , associated to real numbers @xmath244 , for @xmath245 .
the robust regression problem with the huber loss can be formulated as the minimization over @xmath3 of @xmath246 } + \theta_2[i]\right],\ ] ] where @xmath220 is the parameter vector of a linear model .
the conditions described at the beginning of this section can be shown to be satisfied with a lipschitz constant proportional to @xmath247 ; the resulting algorithm is the iterative reweighted least - square method , which appears both in the literature about robust statistics @xcite , and about sparse estimation where the huber loss is used to approximate the @xmath198-norm @xcite .
jensen s inequality also provides a natural mechanism to obtain surrogates for convex functions . following the presentation of lange , hunger and yang @xcite
, we consider a convex function @xmath248 , a vector @xmath249 in @xmath3 , and define @xmath250 as @xmath251 for all @xmath6 .
let @xmath252 be a weight vector in @xmath253 such that @xmath254 and @xmath255 \neq 0 $ ] whenever @xmath256 \!\neq\ ! 0 $ ] .
then , we define for any @xmath30 in @xmath3 : @xmath257 f \left(\frac{\x[i]}{\w[i ] } ( \theta[i]-\kappa[i ] ) + \x^\top \kappa\right),\ ] ] when @xmath4 is @xmath25-smooth , and when @xmath255 \defin |\x[i]|^\nu / \|\x\|_\nu^\nu$ ] , @xmath33 is in @xmath258 with * @xmath259 for @xmath260 ; * @xmath261 for @xmath262 ; * @xmath263 for @xmath264 . to the best of our knowledge ,
non - asymptotic convergence rates have not been studied before for such surrogates , and thus we believe that our analysis may provide new results in the present case .
jensen surrogates are indeed quite uncommon ; they appear nevertheless in a few occasions .
in addition to the few examples given in @xcite , they are used for instance in machine learning by della pietra @xcite for interpreting boosting procedures through the concept of _ auxiliary functions_. jensen s inequality is also used in a different fashion in em algorithms @xcite . consider @xmath5 non - negative functions @xmath265 , and , for some @xmath30 in @xmath3 ,
define some weights @xmath266= f^t(\kappa ) / \sum_{t'=1}^t f^{t'}(\kappa)$ ] . by exploiting the concavity of the logarithm , and assuming hat @xmath266 > 0 $ ] for all @xmath8 to simplify , jensen s inequality yields @xmath267 \log \left(\frac{f^t(\theta)}{\w[t]}\right ) , \label{eq : em}\ ] ] the relation ( [ eq : em ] ) is key to em algorithms minimizing a negative log - likelihood .
the right side of this equation can be interpreted as a majorizing surrogate of the left side since it is easy to show that both terms are equal for @xmath268 .
unfortunately the resulting approximation error functions are not @xmath25-smooth in general and these surrogates do not follow the assumptions of definition [ def : surrogate_batch ] . as a consequence , our analysis may apply to some em algorithms , but not to all of them . when @xmath4 is twice differentiable and admits a matrix @xmath269 such that @xmath270 is always positive definite , the following function is a first - order majorizing surrogate : @xmath271 the lipschitz constant of @xmath272 is the largest eigenvalue of @xmath273 over @xmath2 .
such surrogates appear frequently in the statistics and machine learning literature @xcite .
the goal is to to model the global curvature of the objective function during each iteration , without resorting to the newton method .
even though quadratic surrogates do not necessarily lead to better theoretical convergence rates than simpler lipschitz gradient surrogates , they can be quite effective in practice @xcite .
in this section , we introduce an incremental scheme that exploits the structure ( [ eq : prob ] ) of @xmath4 as a large sum of @xmath5 components . the most popular method for dealing with such a problem
when @xmath4 is smooth and @xmath274 is probably the _
stochastic gradient descent _ algorithm ( sgd ) and its variants ( see @xcite ) .
it consists of drawing at iteration @xmath9 an index @xmath275 and updating the solution as @xmath276 , where the scalar @xmath277 is a step size .
another popular algorithm is the _
stochastic mirror descent _
algorithm ( see @xcite ) for general non - smooth convex problems , a setting we do not consider in this paper since non - smooth functions do not always admit practical first - order surrogates .
recently , linear convergence rates for strongly convex functions @xmath7 have been obtained in @xcite and @xcite by using randomized incremental algorithms whose cost per iteration is independent of @xmath5 .
the method sag @xcite for smooth unconstrained convex optimization is a randomized variant of the incremental gradient descent algorithm of blatt , hero and gauchman @xcite , where an estimate of the gradient @xmath28 is incrementally updated at each iteration .
the method sdca @xcite for strongly convex composite optimization is a dual coordinate ascent algorithm that performs incremental updates in the primal ( [ eq : prob ] ) . unlike sgd , both sag and sdca require storing information about past iterates , which is a key for obtaining fast convergence rates . in a different context
, incremental em algorithms have been proposed by neal and hinton @xcite , where upper bounds of a non - convex negative log - likelihood function are incrementally updated . by using similar ideas
, we introduce the scheme miso in algorithm [ alg : generic_incremental ] . at every iteration , a single function
is observed , and an approximate surrogate of @xmath4 is updated .
note that in the same line of work , ahn et al .
@xcite have proposed a block - coordinate descent majorization - minimization algorithm , which corresponds to miso when the variational surrogates of section [ subsubsec : variational ] are used .
@xmath13 ( initial estimate ) ; @xmath14 ( number of iterations ) .
initialization : choose some surrogates @xmath278 of @xmath7 near @xmath279 for all @xmath8 ; randomly pick up one index @xmath280 and choose a surrogate @xmath281 of @xmath282 near @xmath15 ; set @xmath283 for all @xmath284 .
update the solution : @xmath285 .
@xmath17 ( final estimate ) ; in the next two sections , we study the convergence properties of the scheme miso .
we proceed as in section [ sec : generic ] . specifically , we start with the non - convex case , focusing on stationary point conditions , and we show that similar guarantees as for the batch majorization - minimization algorithm hold .
then , for convex problems , we present convergence rates that essentially apply to the proximal gradient surrogates
. we obtain sublinear rates @xmath286 for the general convex case , and linear ones for strongly convex objective functions .
even though these rates do not show any theoretical advantage over the batch algorithm , we also present a more surprising result in section [ subsec : strong ] ; in a large sample regime @xmath287 , for @xmath130-strongly convex functions @xmath7 , minorizing surrogates may be used and faster rates can be achieved .
we start our analysis with the non - convex case , and make the following assumption : * @xmath4 is bounded below and for all @xmath50 in @xmath2 and all @xmath8 , the directional derivative @xmath288 of @xmath7 at @xmath6 in the direction @xmath52 exists .
then , we obtain a first convergence result .
[ prop : conv13 ] assume that holds and that the surrogates @xmath289 from algorithm [ alg : generic_incremental ] are majorizing @xmath290 and are in @xmath291 .
then , the conclusions of proposition [ prop : conv1 ] hold with probability one .
we proceed in several steps .
_ almost sure convergence of @xmath64 : _
let us define @xmath292 .
we have the following relation for all @xmath86 , @xmath293 where the surrogates and the index @xmath294 are chosen in the algorithm .
then , we obtain the following inequalities , which hold with probability one for all @xmath86 , @xmath295 the first inequality is true by definition of @xmath10 and the second one because @xmath296 is a majorizing surrogate of @xmath297 .
the sequence @xmath298 is thus monotonically decreasing , bounded below with probability one , and thus converges almost surely . by taking the expectation of these previous inequalities , we also obtain that the sequence @xmath299)_{n
\geq 0}$ ] monotonically converges .
thus , the non - positive quantity @xmath300 $ ] is the summand of a converging sum and we have @xmath301 & = \sum_{n=0}^{+\infty } \e[g_{n}^{{\hat t}_{n+1}}(\theta_{n } ) - f^{{\hat t}_{n+1}}(\theta_{n } ) ] \\ & = \sum_{n=0}^{+\infty } \e[\e[g_{n}^{{\hat t}_{n+1}}(\theta_{n } ) - f^{{\hat t}_{n+1}}(\theta_{n})| \ff_{n } ] ] \\ & = \sum_{n=0}^{+\infty } \e[\barg_{n}(\theta_{n } ) - f(\theta_{n } ) ] \\ & = \e\left[\sum_{n=0}^{+\infty } \barg_{n}(\theta_{n } ) - f(\theta_{n})\right ] < + \infty , \\ \end{split}\ ] ] where we use beppo - lvy theorem to interchange the expectation and the sum in front of non - negative quantities , and @xmath302 is the filtration representing all information up to iteration @xmath9 ( including @xmath10 ) . as a result
, the sequence @xmath303 converges almost surely to @xmath60 , implying the almost sure convergence of @xmath64 . _ asymptotic stationary point conditions : _
let us define @xmath304 , which is @xmath25-smooth .
then , for all @xmath6 in @xmath2 and @xmath86 , @xmath305 we have @xmath306 by definition of @xmath10 , and @xmath307 , following similar steps as in the proof of proposition [ prop : conv1 ] .
since we have previously shown that @xmath308 almost surely converges to zero , we conclude as in the proof of proposition [ prop : conv1 ] , replacing @xmath24 by @xmath309 and @xmath11 by @xmath310 .
we also give the counterpart of proposition [ prop : conv1_separable ] for algorithm [ alg : generic_incremental ] . [
prop : conv13b ] assume that is satisfied and that the functions @xmath7 are compositions @xmath311 , where the functions @xmath312 are @xmath96-lipschitz continuous for some @xmath97 .
assume also that the functions @xmath289 in algorithm [ alg : generic_incremental ] are also compositions @xmath313 , where @xmath314 is majorizing @xmath315 and is in @xmath316 .
then , the conclusions of proposition [ prop : conv13 ] hold .
we first remark that the first part of the proof of proposition [ prop : conv13 ] does not exploit the fact that the approximation errors @xmath317 are @xmath25-smooth , but only the fact that @xmath318 is majorizing @xmath7 for all @xmath9 and @xmath8 .
thus , the first part of the proof of proposition [ prop : conv13 ] holds in the present case , such that @xmath64 almost surely converges , and the sequence @xmath319 almost surely converges to zero , where @xmath310 is defined in the proof of proposition [ prop : conv13 ] .
it remains to show that the asymptotic stationary point conditions are satisfied . to that effect
, we follow the proof of proposition [ prop : conv1_separable ] . we first have , for all @xmath86 , @xmath320 with @xmath306 and @xmath321 .
then , following the proof of proposition [ prop : conv1_separable ] , it is easy to show that @xmath322 where @xmath323 , and we conclude as in proposition [ prop : conv1_separable ] . the next lemma provides convergence rates for the convex case , under the assumption that the surrogate functions are @xmath40-strongly convex with @xmath324 .
the result notably applies to the proximal gradient surrogates of section [ subsubsec : proximal ] .
[ prop : conv16 ] assume that @xmath4 is convex and bounded below , let @xmath126 be a minimizer of @xmath4 on @xmath2 , and let us define @xmath325 .
when the surrogates @xmath318 in algorithm [ alg : generic_incremental ] are majorizing @xmath7 and are in @xmath326 with @xmath164 , we have for all @xmath86 , @xmath327 \leq \frac{lt\|\theta^\star-\theta_0\|_2 ^ 2}{2n } , \label{eq : incr : rate}\ ] ] where @xmath328 is the average of the iterates .
assume now that @xmath4 is @xmath130-strongly convex .
for all @xmath86 , @xmath329 \leq \left(1- \frac{2\mu}{t(\rho+\mu)}\right)^{n-1}\frac{l\|\theta^\star-\theta_0\|_2 ^ 2}{2}.\label{eq : incr : ratemu}\ ] ] we proceed in several steps .
_ preliminaries : _ for all @xmath86
, we introduce the point @xmath330 in @xmath2 such that @xmath318 is in @xmath331 .
we remark that such points are drawn recursively according to the following conditional probability distribution : @xmath332 where @xmath333 , @xmath302 is the filtration representing all information up to iteration @xmath9 ( including @xmath10 ) , and @xmath334 for all @xmath8 .
thus we have for all @xmath8 and all @xmath86 , @xmath335= \e[\e[\|\theta^\star-\kappa_{n-1}^t \|_2 ^ 2| \ff_{n-1 } ] ] = \delta\e[\|\theta^\star-\theta_{n-1}\|_2 ^ 2 ] + ( 1-\delta ) \e[\|\theta^\star-\kappa_{n-2}^t\|_2 ^ 2].\label{eq : incr : tmp2}\ ] ] we also need the following extension of lemma [ lemma : basic ] to the incremental setting : for all @xmath6 in @xmath2 and @xmath86 , @xmath336 the proof of this relation is similar to that of lemma [ lemma : basic ] , exploiting the @xmath40-strong convexity of @xmath337 .
we can now study the first part of the proposition .
_ non - strongly convex case ( @xmath338 ) : _ let us define the quantities @xmath339 $ ] and @xmath340 $ ] .
then , we have from ( [ eq : incr : tmp4 ] ) with @xmath170 , and by taking the expectation @xmath341 \leq la_{n-1 } - l \xi_n.\ ] ] it follows from ( [ eq : incr : tmp2 ] ) that @xmath342 and thus , for all @xmath86 , @xmath341 \leq \frac{l}{\delta}(a_{n-1 } - a_n).\ ] ] by summing the above inequalities , and using jensen s inequality , we obtain that @xmath343 \leq \frac{1}{n}\sum_{i=1}^n\e[f(\theta_i)-f^\star ] \leq \frac{l a_0}{\delta},\ ] ] leading to the convergence rate of eq .
( [ eq : incr : rate ] ) , since @xmath344 .
_ @xmath130-strongly convex case : _ assume now that the functions @xmath7 are @xmath130-strongly convex . for all @xmath86 , the strong convexity of @xmath4 and ( [ eq : incr : tmp4 ] ) give us the following inequalities @xmath345 \leq l a_{n-1 } - \rho\xi_n,\ ] ] combining this last inequality with ( [ eq : incr : tmp2 ] ) , we obtain that for all @xmath86 , @xmath346 thus , @xmath347 with @xmath348 .
since @xmath349 , @xmath350 \leq l a_{n-1}$ ] , and @xmath351 , we finally have shown the desired convergence rate ( [ eq : incr : ratemu ] ) .
the convergence rate of the previous proposition in the convex case suggests that the incremental scheme and the batch one of section [ sec : generic ] have the same overall complexity , assuming that each iteration of the batch algorithm is @xmath5 times the one of miso .
for strongly convex functions @xmath7 , we obtain linear convergence rates , a property shared by sag or sdca ; it is thus natural to make a more precise comparison with these other incremental approaches , which we present in the next two sections . in this section
, we assume that the optimization domain is unbounded that is , @xmath274 , and that the functions @xmath7 are @xmath25-smooth . when using the lipschitz gradient surrogates of section [ subsubsec : gradient ] , miso amounts to iteratively using the following update rule : @xmath352 where the vectors @xmath353 are recursively defined for @xmath153 as @xmath354 and @xmath355 for @xmath356 , with @xmath357 for all @xmath8 .
it is then easy to see that the complexity of updating @xmath10 is independent of @xmath5 , by storing the vectors @xmath358 and performing the update @xmath359 . in comparison ,
the approach sag yields a different , but related , update rule : @xmath360 where the value @xmath361 is suggested in @xcite .
even though the rules ( [ eq : miso1 ] ) and ( [ eq : sag1 ] ) seem to be similar to each other at first sight , they behave differently in practice and do not have the same theoretical properties . for non - convex problems
, miso is guaranteed to converge , whereas it is not known whether it is the case for sag or not . for convex problems ,
both methods have a convergence rate of the same nature that is , @xmath286 . for @xmath130-strongly - convex problems ,
however , the convergence rate of sag reported in @xcite is substantially better than ours . whereas the expected objective of sag decreases with the rate @xmath362 with @xmath363 , ours decreases with @xmath364 , which is larger than @xmath365 unless the problem is very well conditioned . by maximizing the convex dual of ( [ eq : prob ] ) when the functions @xmath7 are @xmath130-strongly convex
, the approach sdca yields another update rule that resembles ( [ eq : miso1 ] ) and ( [ eq : sag1 ] ) , and offers similar convergence rates as sag . as part of the procedure ,
sdca involves large primal gradient steps @xmath366 for updating the dual variables .
it is thus appealing to study whether such large gradient steps can be used in ( [ eq : miso1 ] ) in the strongly convex case , regardless of the majorization - minimization principle .
in other words , we want to study the use of the following surrogates within miso : @xmath367 which are lower bounds of the functions @xmath7 instead of upper bounds .
then , minimizing @xmath368 amounts to performing the update ( [ eq : miso1 ] ) when replacing @xmath25 by @xmath130 .
the resulting algorithm is slightly different than sdca , but resembles it .
as shown in the next proposition , the method achieves a fast convergence rate when @xmath287 , but may diverge if @xmath5 is small .
note that at the same time as us , a similar result was independently obtained by defazio et al .
@xcite , where a refined analysis provides a slightly better rate , namely the constant @xmath369 in ( [ eq : incr : ratemu2 ] ) may be replaced by @xmath370 .
[ prop : conv17 ] assume that the functions @xmath7 are @xmath130-strongly convex , @xmath25-smooth , and bounded below .
let @xmath126 be a minimizer of @xmath4 on @xmath2 .
assume that @xmath287 .
when the functions @xmath318 of eq .
( [ eq : lower_surrogates ] ) are used in algorithm [ alg : generic_incremental ] , we have for all @xmath86 , @xmath329 \leq \left(1 - \frac{1}{3t}\right)^{n } \frac{2t}{\mu}\|\nabla f(\theta_0)\|_2 ^ 2.\label{eq : incr : ratemu2}\ ] ] when the functions @xmath7 are lower - bounded by the function @xmath371 , we can use the initialization @xmath372 and @xmath373 for all @xmath8 . then , the quantity @xmath374 in ( [ eq : incr : ratemu2 ] ) can be replaced by @xmath375 . as in the proof of proposition
[ prop : conv13 ] , we introduce the function @xmath376 , which is minimized by @xmath10 for @xmath86 . since @xmath310 is a lower bound on @xmath4 , we have the relation @xmath377 .
inspired by the convergence proof of sdca @xcite , which computes an convergence rate of an expected duality gap , we proceed by studying the convergence of the sequence @xmath378)_{n
\geq 1}$ ] .
on the one hand , we have for all @xmath86 , @xmath379 where @xmath380 .
the first equality is true because @xmath310 is quadratic and is minimized by @xmath10 , and the second one uses the relation ( [ eq : recur ] ) . by definition of @xmath381 , we have that @xmath382 , and by taking the expectation , @xmath383= \e[f^{{\hat t}_n}(\theta_{n-1 } ) ] = \e[\e[f^{{\hat t}_n}(\theta_{n-1})| \ff_{n-1 } ] ] = \e[f(\theta_{n-1})]$ ] , where @xmath302 is the the filtration representing all information up to iteration @xmath9 .
we also have that @xmath384 = \e[\e[g_{n-1}^{{\hat t}_n}(\theta_{n-1 } ) | \ff_{n-1 } ] ] = \e[\barg_{n-1}(\theta_{n-1})]$ ] .
thus , we obtain a first useful relation : @xmath385 = ( 1-\delta)\e[\barg_{n-1}(\theta_{n-1 } ) ] + \delta \e[f(\theta_{n-1 } ) ] - \frac{\mu}{2}\e\left[\|\theta_n-\theta_{n-1}\|_2 ^ 2\right].\label{eq : tmpconv13a}\ ] ] on the other hand , for all @xmath386 , @xmath387 we have used the fact that @xmath388 is a majorizing surrogate of @xmath297 , whereas @xmath389 is minorizing @xmath290 . by adding twice ( [ eq : tmpconv13c ] ) after taking the expectation and once ( [ eq : tmpconv13a ] ) , we obtain that for all @xmath153 , @xmath390 & \geq ( 3 - \delta)\e[\barg_{n-1}(\theta_{n-1 } ) ] + \delta\e [ f(\theta_{n-1 } ) ] + \left(\frac{\mu}{2}-\delta l\right)\e[\|\theta_n-\theta_{n-1}\|_2 ^ 2 ] \\ & \geq ( 3 - \delta)\e[\barg_{n-1}(\theta_{n-1 } ) ] + \delta\e [ f(\theta_{n-1 } ) ] , \end{split } \label{eq : tmpconv13b}\ ] ] where the second inequality comes from the large sample size condition @xmath391 .
since @xmath392 \geq f^\star$ ] , this immediately gives for @xmath386 , @xmath393 \leq \left ( 1- \frac{1}{3 t } \right ) \left(f^\star - \e\left[\barg_{n-1}(\theta_{n-1})\right]\right).\ ] ] to obtain a convergence rate for @xmath394-f^\star$ ] , we use again eq .
( [ eq : tmpconv13b ] ) . for @xmath153 , @xmath395 - f^\star ) & \leq \delta(\e [ f(\theta_{n-1 } ) ] - \e[\barg_{n-1}(\theta_{n-1 } ) ] ) \\ & \leq 3 ( \e[\barg_n(\theta_n)]- \e[\barg_{n-1}(\theta_{n-1 } ) ] ) \\ & \leq 3 ( f^\star- \e[\barg_{n-1}(\theta_{n-1 } ) ] ) \\ & \leq 3 \left ( 1- \frac{1}{3 t } \right)^{n-2}\left(f^\star - \barg_{1}(\theta_{1})\right ) , \end{split } \label{eq : tmpincr1}\ ] ] and we obtain the convergence rate ( [ eq : incr : ratemu2 ] ) by first noticing that @xmath396 where we use the relation @xmath397 and @xmath398 . then , we use the fact that @xmath399 since @xmath400 , such that @xmath401 . to prove the last part of the proposition , we remark that all inequalities we have proved so far for @xmath386 , become true for @xmath150 .
thus , the last inequality in ( [ eq : tmpincr1 ] ) is also true when replacing @xmath402 by @xmath403 and @xmath404 by @xmath405 .
the proof technique is inspired in part by the one of sdca @xcite ; the quantity @xmath406 is indeed a lower bound of @xmath67 , and plays a similar role as the dual value in sdca .
we remark that the convergence rate ( [ eq : incr : ratemu2 ] ) improves significantly upon the original one ( [ eq : incr : ratemu ] ) , and is similar to the one of sag when @xmath5 is larger than @xmath407
. however , proposition [ prop : conv17 ] only applies to strongly convex problems .
in other cases , the more conservative rule ( [ eq : miso1 ] ) should be preferred in theory , even though we present heuristics in section [ subsec : heuristics ] that suggest using larger step sizes than @xmath408 in practice .
when @xmath4 can be written as @xmath409 , where the functions @xmath410 are @xmath25-smooth , we can use the proximal gradient surrogate presented in section [ subsubsec : proximal ] ; it yields the following rule : @xmath411 where the vectors @xmath330 are defined as in section [ subsec : strong ] .
this update is related to sdca , as well as to stochastic methods for composite convex optimization such as the regularized dual averaging algorithm of xiao @xcite . as in the previous section ,
we obtain guarantees for non - convex optimization , but our linear convergence rate for strongly convex problems is not as fast as the one of sdca . even though we do not have a similar result as proposition [ prop : conv17 ] for the composite setting , we have observed that using a smaller value for @xmath25 than the theoretical one could work well in practice .
we detail such an empirical strategy in the next section .
we have found the following strategies to improve the practical performance of miso .
[ [ initialization ] ] initialization + + + + + + + + + + + + + + a first question is how to initialize the surrogates @xmath278 in practice . even though we have suggested the functions @xmath278 to be in @xmath412 in algorithm [ alg : generic_incremental ] , our analysis weakly relies on this assumption .
in fact , most of our results hold when choosing surrogates computed at points @xmath413 that are not necessarily equal to @xmath279 ; at most only constants from the convergence rates would be affected by such a change .
an effective empirical strategy is inspired by the second part of proposition [ prop : conv17 ] : we first define functions @xmath414 , and perform @xmath5 iterations of miso without randomization , selecting the function @xmath7 at iteration @xmath8 , such that each surrogate is updated exactly once .
then , we use these updated surrogates for initializing the regular randomized scheme . [ [ warm - restart - and - continuation ] ] warm restart and continuation + + + + + + + + + + + + + + + + + + + + + + + + + + + + + when available , warm restart can be used for initializing the surrogates .
assume that we are interested in minimizing a composite function @xmath415 , which is parameterized by a scalar @xmath416 , and that we want to obtain a minimizer for several parameter values @xmath417 .
we first solve the problem for @xmath418 , and then use the surrogates obtained at the end of the optimization for initializing the algorithm when addressing the problem with @xmath419 .
we proceed similarly going from larger to smaller values of @xmath416 .
we have empirically observed that the warm restart strategy could be extremely efficient in practice , and would deserve further study in a future work .
[ [ heuristics - for - selecting - step - sizes ] ] heuristics for selecting step sizes + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + choosing proximal gradient surrogates @xmath420 requires choosing some lipschitz constant @xmath25 ( or a strong convexity parameter @xmath130 for proposition [ prop : conv17 ] ) , which leads to a specific step size in ( [ eq : miso_composite ] ) . however , finding an appropriate step size can be difficult in practice for several reasons :
( i ) in some cases , these parameters are unknown ; ( ii ) even though a global lipschitz constant might be available , a local lipschitz constant could be more effective ; ( iii ) the convergence rates of proposition [ prop : conv16 ] can be obtained by choosing a smaller value for @xmath25 than the `` true '' lipschitz constant , as long as the inequality @xmath394
\leq \e[\barg_n(\theta_n)]$ ] is always satisfied , where @xmath421 .
this motivates the following heuristics : * first perform one pass over @xmath422 of the data to select a constant @xmath423 with @xmath424 chosen among positive integers , yielding the smallest objective on the data subset , where @xmath425 is an upper bound of the true lipschitz constant . *
proceed as in miso1 , but choose a more aggressive strategy @xmath426 ; during the optimization , compute the quantities @xmath427 and @xmath428 defined as @xmath429 , @xmath430 if @xmath284 , and otherwise @xmath431 , @xmath432 , where we have parameterized the surrogates @xmath420 by @xmath433 .
every @xmath5 iterations , compare the sums @xmath434 and @xmath435 . if @xmath436 , do nothing ; otherwise , increase the value of @xmath433 until this inequality is satisfied .
the heuristic miso2 is more aggressive than miso1 since it starts with a smaller value for @xmath25 .
after every iteration , this value is possibly increased such that on average , the surrogates `` behave '' as majorizing functions .
even though this heuristic does not come with any theoretical guarantee , it was found to perform slightly better than miso1 for strongly - convex problems . [ [ using - a - different - parameter - l_t - for - every - function - f_t ] ] using a different parameter @xmath437 for every function @xmath438 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + even though our analysis was conducted with a global parameter @xmath25 for simplicity , it is easy to extend the analysis when the parameter @xmath25 is adjusted individually for every surrogate .
this is useful when the functions @xmath438 are heterogeneous .
[ [ parallelization - with - mini - batches ] ] parallelization with mini - batches + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the complexity of miso is often dominated by the cost of updating the surrogates @xmath289 , which typically requires computing the gradient of a function .
a simple extension is to update several surrogates at the same time , when parallel computing facilities are available .
in this section , we evaluate miso on large - scale machine learning problems .
our implementation is coded in c++ interfaced with matlab and is freely available in the open - source software package spams @xcite .
all experiments were conducted on a single core of a 2ghz intel cpu with @xmath439 gb of ram .
[ [ datasets ] ] datasets + + + + + + + + we use six publicly available datasets , which consist of pairs @xmath440 , where the @xmath441 s are labels in @xmath442 and the @xmath443 s are vectors in @xmath3 representing data points . the datasets are described in table [ table : datasets ] .
, , , and are obtained from the 2008 pascal large - scale learning challenge . and are obtained from the libsvm website .
the datasets are pre - processed as follows : all dense datasets are standardized to have zero - mean and unit variance for every feature .
the sparse datasets are normalized such that each @xmath443 has unit @xmath112-norm .
.description of datasets used in our experiments . [ cols="<,^,^,^,^,^",options="header " , ] we consider the @xmath112-regularized logistic regression problem , which can be formulated as follows : @xmath444 where @xmath445 for all @xmath446 . following @xcite , we report some results obtained with different methods with the parameter @xmath447 , which is argued to be of the same order of magnitude as the smallest value that would be used in practice for machine learning problems .
we also performed experiments with the values @xmath448 and @xmath449 to study the impact of the strong convexity parameter ; the output of these two additional experiments is not reported in the present paper for space limitation reasons , but it will be discussed and taken into account in our conclusions . the algorithms included in the comparison are : * the stochastic gradient descent algorithm with a heuristic for choosing the step - size similar to miso1 , and inspired by leon bottou s sgd toolbox for machine learning .
a step - size of the form @xmath450 is automatically adjusted when performing one pass on @xmath451 of the training data .
we obtain consistent results with the performance of sgd reported by schmidt et al .
@xcite when the step - size is chosen from hindsight . based on their findings
, we do not include in our figures other variants of sgd , _
e.g. _ , @xcite . * the accelerated gradient method proposed by beck and teboulle @xcite with a line - search for automatically adjusting the lipschitz constant . * the algorithm of shalev - schwartz and zhang @xcite , efficiently implemented in the language c by mark schmidt . * a fast implementation in c also provided by mark schmidt @xcite .
we use the step - size @xmath408 since it performed similar to their heuristic line search . *
the majorization - minimization algorithm miso , using the trivial upper bound @xmath452 on the lipschitz constant for example @xmath8 . * the majorization - minimization heuristic miso1 described in section [ subsec : heuristics ] . * the heuristic miso2 ,
also described in section [ subsec : heuristics ] .
* the update rule corresponding to proposition [ prop : conv17 ] . for sparse datasets , miso0 , miso1 , and miso2 are not practical since they suffer from a @xmath453 memory cost .
their update rules can indeed be rewritten @xmath454 where @xmath455 .
thus , for every example @xmath8 , the algorithm requires storing the dense vector @xmath456 .
therefore , we use mini - batches of size @xmath457 , where @xmath458 is the density of the dataset ; the resulting algorithms , which we denote by miso0-mb , miso1-mb , and miso2-mb , have a storage cost equal to @xmath459 , which is the same as the dataset . on the other hand
, the update rule miso@xmath130 applied to the @xmath416-strongly convex functions @xmath7 admits a simpler and computationally cheaper form .
since @xmath460 , the update becomes @xmath461 where @xmath462 denotes the derivative of @xmath463 with respect to its second argument .
assuming that the dataset fits into memory , the only extra quantities to store are the scalars @xmath464 , and the resulting memory cost is simply @xmath465 .
we present our comparison of the above methods with @xmath447 on figures [ fig : l2epochs ] and [ fig : l2time ] , where we plot the relative duality gap defined as @xmath466 , where @xmath467 is the best value of the fenchel dual that we have obtained during our experiments .
the conclusions of our empirical study are the following : * _ sag , sdca and miso@xmath130 : _ these methods perform similarly and were consistently the fastest , except in the regime @xmath468 where miso@xmath130 can diverge ; * _ the four variants of miso : _ as predicted by its theoretical convergence rate , miso0 does not perform better than ista @xcite without line - search ( not reported in the figures ) .
miso1 and miso2 perform significantly better .
miso@xmath130 is always better or as good as miso1 and miso2 , except for sparse datasets with @xmath448 where the condition @xmath469 is not satisfied ; * _ influence of mini - batch : _ whereas miso2 performs equally well as sag / sdca for dense datasets , mini - batches for sparse datasets makes it slower ; * _ stochastic gradient descent : _ sgd - h performs always well at the beginning of the procedure , but is not competitive compared to incremental approaches after a few passes over the data . note
that an evaluation of a preliminary version of miso2 is presented in @xcite for the @xmath198-regularized logistic regression problem , where the objective function is not strongly convex .
our experimental findings showed that miso2 was competitive with state - of - the - art solvers based on active - set and coordinate descent algorithms @xcite .
+ + the majorization - minimization principle is appealing for non - convex and non - smooth optimization , where only few algorithms apply . here , we address a sparse estimation problem presented in section [ subsec : dc ] : @xmath470|+\varepsilon ) , \label{eq : expdc}\ ] ] where the scalars @xmath441 and the vectors @xmath443 are the same as in the previous section , and @xmath471 is set to @xmath472 .
the model parameter @xmath416 controls the sparsity of the solution .
even though ( [ eq : expdc ] ) is non - convex and non - smooth , stationary points can be obtained in various ways . in this section
, we consider majorization - minimization approaches where the penalty function @xmath473|+\varepsilon)$ ] is upper - bounded as in eq .
( [ eq : upperbounddc ] ) , whereas the functions @xmath474 are upper - bounded by the lipschitz gradient surrogates of section [ subsubsec : gradient ] .
we compare five approaches : * algorithm [ alg : generic_batch ] with the trivial lipschitz constant @xmath475 . * algorithm [ alg : generic_batch ] with the line - search scheme of ista @xcite for adjusting @xmath25 .
* we compare miso0 , miso1 , and miso2 , as in the previous section .
we choose a parameter @xmath416 for each dataset , such that the solution with the lowest objective function obtained by any of the tested method has approximately a sparsity of @xmath476 for datasets and , 100 for and , and @xmath477 for and .
the methods are initialized with @xmath478 ; indeed , the initialization @xmath372 that was a natural choice in section [ subsec : explog ] appears to be often a bad stationary point of problem ( [ eq : expdc ] ) and thus an inappropriate initial point .
we report the objective function values for different passes over the data in figure [ fig : dcepochs ] , and the sparsity of the solution in figure [ fig : dcspars ] .
our conclusions are the following : * methods with line searches do significantly better than those without , showing that adjusting the constant @xmath25 is important for these datasets ; * miso1 does asymptotically better than mm - ls for five of the datasets after @xmath479 epochs and slightly worse for ; in general , miso1 seems to converge substantially faster than other approaches , both in terms of objective function and in terms of the support of the solution . * the performance of miso2
is mitigated . in one case
, it does better than miso1 , but in some others , it converges to the stationary point @xmath480 . + +
in this paper , we have presented new algorithms based on the majorization - minimization principle for minimizing a large sum of functions .
the main asset of our approach is probably its applicability to a large class of non - convex problems , including non - smooth ones , where we obtain convergence and asymptotic stationary point guarantees . for convex problems
, we also propose new incremental rules for composite optimization , which are competitive with state - of - the - art solvers in the context of large - scale machine learning problems such as logistic regression .
we note that other majorization - minimization algorithms have recently been analyzed , such as block coordinate variants in @xcite and stochastic ones in @xcite .
in particular , we have proposed in @xcite a stochastic majorization - minimization algorithm that does not require to store information about past iterates , when the objective function is an expectation .
since the first version of our work was published in @xcite , miso has also been extended by other authors in @xcite using the alternating direction method of multipliers framework . for future work
, we are currently investigating extensions of the scheme miso@xmath130 for strongly convex objective functions .
we believe that the algorithm can be modified to remove the large sample condition @xmath287 , that the convergence proof can be extended to the proximal setting , and that it is possible to use acceleration techniques in the sense of nesterov @xcite .
another interesting direction of research would be to study the stability of our result to inexact minimization of surrogate functions following for instance the analysis of @xcite for proximal gradient methods .
the author would like to thank zaid harchaoui , francis bach , simon lacoste - julien , mark schmidt , martin jaggi , the associate editor , and the anonymous reviewers for their useful comments .
the following definitions can be found in classical textbooks , e.g , @xcite . for the sake of completeness ,
we briefly introduce them here .
[ def : derivative ] let us consider a function @xmath27 and @xmath481 be in @xmath3 .
when it exists , the following limit is called the directional derivative of @xmath4 at @xmath6 in the direction @xmath52 : @xmath482 when @xmath4 is differentiable at @xmath6 , directional derivatives exist in every direction , and @xmath483 .
[ def : stationary ] let us consider a function @xmath484 , where @xmath2 is a convex set , such that @xmath4 admits a directional derivative @xmath51 for all @xmath50 in @xmath2 .
we say that @xmath6 in @xmath2 is a stationary point if for all @xmath32 in @xmath2 , @xmath485 .
[ def : strong_convexity ] let @xmath2 be a convex set .
a function @xmath484 is called @xmath130-strongly convex when there exists a constant @xmath488 such that for all @xmath32 in @xmath2 , the function @xmath489 is convex .
[ lemma : convexerror ] let @xmath490 be two functions . define @xmath35 . then ,
if @xmath33 is @xmath40-strongly convex and @xmath4 is @xmath25-smooth , with @xmath491 , @xmath492 is @xmath179-strongly convex ; [ step : convexerror ] if @xmath33 and @xmath4 are convex and @xmath25-smooth , @xmath492 is also @xmath25-smooth ; if @xmath33 and @xmath4 are @xmath130-strongly convex and @xmath25-smooth , @xmath492 is @xmath493-smooth .
* @xmath222 is differentiable for all @xmath223 in @xmath495 ; * @xmath496 is @xmath25-lipschitz continuous for all @xmath220 in @xmath221 ; * @xmath219 is @xmath130-strongly convex for all @xmath220 in @xmath221 . also define @xmath497 .
then , @xmath216 is differentiable and @xmath498 , where @xmath499 .
moreover , if @xmath500 is @xmath225-lipschitz continuous for all @xmath220 in @xmath221 , the gradient @xmath501 is @xmath502-lipschitz . | majorization - minimization algorithms consist of successively minimizing a sequence of upper bounds of the objective function .
these upper bounds are tight at the current estimate , and each iteration monotonically drives the objective function downhill .
such a simple principle is widely applicable and has been very popular in various scientific fields , especially in signal processing and statistics .
we propose an incremental majorization - minimization scheme for minimizing a large sum of continuous functions , a problem of utmost importance in machine learning .
we present convergence guarantees for non - convex and convex optimization when the upper bounds approximate the objective up to a smooth error ; we call such upper bounds `` first - order surrogate functions '' .
more precisely , we study asymptotic stationary point guarantees for non - convex problems , and for convex ones , we provide convergence rates for the expected objective function value .
we apply our scheme to composite optimization and obtain a new incremental proximal gradient algorithm with linear convergence rate for strongly convex functions .
our experiments show that our method is competitive with the state of the art for solving machine learning problems such as logistic regression when the number of training samples is large enough , and we demonstrate its usefulness for sparse estimation with non - convex penalties .
non - convex optimization , convex optimization , majorization - minimization . 90c06 , 90c26 , 90c25 |
the ( first part of the ) following problem was suggested by von neumann ( see pp .
60 - 61 , appendix 3 in @xcite ) .
[ von neumann problem ] let @xmath2 if @xmath3 is normal and if @xmath4 does it follow that @xmath5 more generally , if @xmath3 and @xmath6 are normal and if @xmath7 does it follow that @xmath8 if the operators @xmath3 and @xmath9 belong to a finite factor @xmath10 then the first part of the problem was resolved ( in the affirmative ) by von neumann himself . in full generality ,
a problem was resolved by fuglede @xcite .
furthermore , von neumann mentioned that a formal analogue of problem [ von neumann problem ] for unbounded operators can be _ non - rigorously _ answered in the negative due to the fact that a product of @xmath11 unbounded operators does not always exists .
a partial affirmative answer was given by putnam ( see theorem 1.6.2 in @xcite ) .
he proved that if @xmath12 then @xmath13 provided that @xmath9 is _
bounded_. in what follows , we propose a rigorous analogue of the problem [ von neumann problem ] for unbounded operators affiliated with a von neumann algebra @xmath14 we start with a proper framework .
the set of all operators affiliated to a von neumann algebra @xmath15 does not necessarily form an algebra . at the same time
, the class of unital @xmath16-algebras which consist of operators affiliated with @xmath15 is vast .
in particular , it contains all algebras of measurable operators @xcite , and those of @xmath17-measurable operators @xcite . according to @xcite , in this class
, there is a unique maximal element called @xmath18 we call @xmath19 the algebra of all locally measurable operators affiliated with @xmath14 an equivalent constructive definition of @xmath19 is given in section [ prelims ]
. we now properly restate problem
[ von neumann problem ] for unbounded operators affiliated with @xmath14 [ our problem ] let @xmath15 be a von neumann algebra and let @xmath20 if @xmath3 and @xmath6 are normal and if @xmath7 does it follow that @xmath8 theorem 5 in @xcite delivers the positive answer to problem [ our problem ] for the case when @xmath21 @xmath6 and @xmath9 are measurable operators affiliated with a von neumann algebra @xmath15 of type i ( see also @xcite ) . in the case of an arbitrary finite von neumann algebra @xmath10 the problem [ our problem ]
is resolved in the affirmative in @xcite ( see corollary 3.6 there ) .
we answer problem [ our problem ] in the affirmative in full generality .
our methods are stronger than those of @xcite , @xcite , @xcite , @xcite , @xcite and are of independent interest .
the following theorem is the main result of the paper .
[ t3 ] let @xmath15 be an arbitrary von neumann algebra , let @xmath22 be locally measurable operators affiliated with @xmath14 if @xmath3 and @xmath6 are normal and if @xmath7 then @xmath23 the corollary below extends the classical spectral theorem for normal operator ( see e.g. ( * ? ? ?
13 , theorem 13.33 ) ) to the setting of locally measurable operators .
[ t5 ] let @xmath15 be an arbitrary von neumann algebra and let @xmath24 be locally measurable operators affiliated with @xmath14 if @xmath3 is normal and @xmath25 then @xmath26 for every spectral projection @xmath27 of the operator @xmath3 . if @xmath3 and @xmath6 are normal , then the following conditions are equivalent : 1 .
[ cora ] @xmath25 ; 2 .
[ corb ] @xmath28 for every spectral projection @xmath27 of the operator @xmath3 and for every spectral projection @xmath29 of the operator @xmath30 3 .
[ corc ] @xmath31 for every borel complex functions @xmath32 and @xmath33 on @xmath34 , which bounded on compact subsets .
let @xmath1 be a hilbert space , let @xmath0 be the @xmath35-algebra of all bounded linear operators on @xmath1 , and let @xmath36 be the identity operator on @xmath1 . given a von neumann algebra @xmath15 acting on @xmath1 , denote by @xmath37 the centre of @xmath15 and by @xmath38 the lattice of all projections in @xmath15 .
let @xmath39 be the set of all finite projections in @xmath15 . a linear operator @xmath40 , where the domain @xmath41 of @xmath3 is a linear subspace of @xmath1 , is said to be _ affiliated _ with @xmath15 if @xmath42 for all @xmath6 from the commutant @xmath43 of algebra @xmath15 .
a densely - defined closed linear operator @xmath3 ( possibly unbounded ) affiliated with @xmath15 is said to be _
measurable _ with respect to @xmath15 if there exists a sequence @xmath44 such that @xmath45 and @xmath46 for every @xmath47 , where @xmath48 is the set of all natural numbers .
let us denote by @xmath49 the set of all measurable operators .
let @xmath50 .
it is well known that @xmath51 and @xmath52 are densely - defined and preclosed operators . moreover , the closures @xmath53 ( strong sum ) , @xmath54 ( strong product ) and @xmath55 are also measurable , and equipped with this operations @xmath49 is a unital @xmath16-algebra over the field @xmath56 of complex numbers @xcite .
it is clear that @xmath15 is a @xmath16-subalgebra of @xmath49 .
a densely - defined linear operator @xmath3 affiliated with @xmath15 is called _ locally measurable _ with respect to @xmath15 if there is a sequence @xmath57 of central projections in @xmath15 such that @xmath58 and @xmath59 for all @xmath47 .
the set @xmath19 of all locally measurable operators is a unital @xmath16-algebra over the field @xmath56 with respect to the same algebraic operations as in @xmath49 @xcite and @xmath49 is a @xmath16-subalgebra of @xmath19 .
it is clear that if @xmath15 is finite , the algebras @xmath49 and @xmath19 coincide .
if von neumann algebra @xmath15 is of type @xmath60 and @xmath61 , then @xmath62 but @xmath63 . for every subset @xmath64 , the sets of all self - adjoint ( resp .
, positive ) operators in @xmath65 will be denoted by @xmath66 ( resp .
@xmath67 ) .
the partial order in @xmath19 is defined by its cone @xmath68 and is denoted by @xmath69 .
let @xmath3 be a closed operator with dense domain @xmath70 in @xmath1 , let @xmath71 be the polar decomposition of the operator @xmath3 , where @xmath72 and @xmath73 is a partial isometry in @xmath0 such that @xmath74 ( respectively , @xmath75 ) is the right ( left ) support @xmath76 ( respectively , @xmath77 ) of @xmath3 .
it is known that @xmath78 and @xmath79 ( respectively , @xmath80 ) if and only if @xmath81 ( respectively , @xmath82 ) and @xmath83 @xcite . if @xmath3 is a self - adjoint operator affiliated with @xmath15 , then the spectral family of projections @xmath84}(a ) , \
{ \lambda\in \mathbb{r}}$ ] , for @xmath3 belongs to @xmath15 @xcite .
a locally measurable operator @xmath3 is measurable if and only if @xmath85 for some @xmath86 @xcite . in what follows ,
we use the notation @xmath87 for the projection onto the kernel of the operator @xmath3 .
assume now that @xmath15 is a semifinite von neumann algebra equipped with a faithful normal semifinite trace @xmath17 .
a densely - defined closed linear operator @xmath3 affiliated with @xmath15 is called @xmath17-measurable if for each @xmath88 there exists @xmath89 with @xmath90 such that @xmath91 .
let us denote by @xmath92 the set of all @xmath17-measurable operators .
it is well - known @xcite that @xmath92 is a @xmath16-subalgebra of @xmath49 and @xmath93 .
it is clear that if @xmath15 is a semifinite factor , the algebras @xmath92 and @xmath49 coincide .
note also that for every @xmath94 there exists @xmath86 such that @xmath95 ( see @xcite and @xcite ) .
measure topology is defined in @xmath92 by the family @xmath96 @xmath97 of neighborhoods of zero @xmath98 convergence of the sequence @xmath99 in measure topology is called _ convergence in measure_. when equipped with measure topology , @xmath92 is a complete metrizable topological @xmath16-algebra ( see @xcite ) . for basic properties of the measure topology ,
see @xcite .
we remark only that @xmath100 in measure , @xmath101 if and only if @xmath102 let @xmath15 be a von neumann algebra equipped with a faithful normal semifinite trace @xmath103 we set @xmath104 the following property is standard .
[ t1 ] let @xmath15 be a semifinite von neumann algebra and let @xmath17 be a faithful normal semifinite trace on @xmath14 if @xmath105 then @xmath106
the proof of theorem [ t3 ] in full generality is based on its special case for the @xmath16-algebra @xmath107 [ t2 ] let @xmath15 be a semifinite von neumann algebra and let @xmath17 be a faithful normal semifinite trace on @xmath15 .
let @xmath108 and let @xmath3 and @xmath6 be normal .
if @xmath7 then @xmath23 our strategy of the proving theorem [ t2 ] relies on a number of auxiliary lemmas .
whereas some of them look similar to those used in @xcite , the lemmas below appear to be stronger than their counterparts from @xcite .
[ l1 ] if @xmath109 is normal , @xmath110 and @xmath111 with @xmath112 , then @xmath113 denote , for brevity , @xmath114 due to the normality of @xmath3 and using the equality @xmath112 , we have @xmath115 since @xmath111 , it follows that @xmath116 taking the trace and using property [ t1 ] , we conclude that @xmath117 since @xmath17 is faithful , it follows that @xmath118 this completes the proof .
[ l2 ] let @xmath119 are normal , @xmath120 and @xmath121 .
let @xmath86 be such that @xmath122 .
setting @xmath123}(c)$ ] and @xmath124 we obtain @xmath125 set @xmath126 and @xmath127 for @xmath128 * step 1 : * we claim that @xmath129 indeed , use the equality @xmath130 and property [ t1 ] , we have @xmath131 @xmath132 the proof of the second equality in the claim is identical .
* step 2 : * we claim that @xmath133 and @xmath134 since @xmath135 it follows that @xmath136 by assumption , @xmath137 thus , @xmath138 since @xmath139 and since @xmath140 it follows that @xmath141 hence ( see property [ t1 ] ) , @xmath142 using now the inequality @xmath143 , we have that @xmath144 and @xmath145 let @xmath146 and @xmath147 taking the adjoints in the equality @xmath7 we obtain @xmath148 in addition @xmath149 applying to the triple @xmath150 we obtain @xmath151 this proves the claim . *
step 3 : * using steps 1 , 2 , we obtain @xmath152 thus , @xmath153 * step 4 : * we claim that @xmath154 and @xmath155 . by , we have @xmath156 now , using property [ t1 ] , we obtain @xmath157 definition of @xmath158 now yields @xmath159 it follows from step 3 that @xmath160 in other words , @xmath161 hence , @xmath162 since @xmath17 is faithful and since @xmath163 it follows that @xmath164 for every @xmath88 and @xmath165 we have @xmath166 therefore , @xmath167 we now infer from that @xmath168 since @xmath169 in measure as @xmath170 we obtain @xmath171 thus , @xmath172 and @xmath173 hence , we infer from lemma [ l1 ] that @xmath174 similarly , @xmath175 [ l3 ] let @xmath119 be normal , @xmath120 and @xmath121
. then @xmath23 the assumption @xmath120 guarantees that there exists @xmath86 such that @xmath176 .
set @xmath177 and @xmath178 @xmath179 @xmath180 @xmath181 . by lemma [ l2 ] , the operators @xmath3 and @xmath6 commute with @xmath182 , in particular , @xmath183 and @xmath184 are normal @xmath181 . by the same lemma , the operator @xmath3 commutes with projections @xmath185 for all @xmath186 . since finite linear combinations of projections @xmath187 @xmath186 ,
converge to operator @xmath188 in the measure topology and since multiplication in @xmath92 is continuous in that topology , it follows that @xmath189 appealing now to lemma [ l2 ] , we obtain @xmath190 combining now ( [ ch4 ] ) and ( [ ch5 ] ) yields @xmath191 taking into account , we rewrite as @xmath192 combining this with the assumption @xmath7 we infer @xmath193 taking into account that @xmath194 and applying the classical fuglede - putnam theorem we derive that @xmath195 thus , @xmath196 [ ber last lemma ] let @xmath119 be normal and let @xmath197 be such that @xmath198 if @xmath199 or @xmath200 then @xmath23 we only consider the first case ( the second case can be reduced to the first one by considering the triple @xmath201 instead of the triple @xmath202 ) .
let @xmath203 be a polar decomposition of @xmath9 so that @xmath204 and @xmath205 .
let @xmath206 be a partial isometry such that @xmath207 and @xmath208 define an isometry @xmath209 ( that is , @xmath210 ) .
it is immediate that @xmath211 and @xmath212 thus , @xmath213 since @xmath210 and since @xmath3 is normal , it follows that @xmath214 is also normal . applying lemma [ l3 ] to the triple @xmath215 we obtain @xmath216
therefore , @xmath217 this completes the proof .
we give now the proof of theorem [ t2 ] in the case of arbitrary semifinite von neumann algebra @xmath15 with a faithful normal semifinite trace @xmath17 .
let us suppose at first that @xmath218 by ( * ? ? ?
* theorem 2.1.3 ) there exist central projectors @xmath219 such that @xmath220 it is immediate that @xmath221 clearly , @xmath222 and @xmath223 @xmath224 where the left hand side is taken in the algebra @xmath225 applying lemma [ ber last lemma ] to the triples @xmath226 and @xmath227 we obtain @xmath228 summing these equalities , we obtain that @xmath23 this proves the assertion for the case @xmath218 let now @xmath24 be an arbitrary normal operators in @xmath92 and @xmath229 .
let @xmath230 ( respectively , @xmath231 ) be the spectral projection for @xmath3 ( respectively , @xmath6 ) corresponding to the set @xmath232 .
it is clear that @xmath233 and @xmath234 are an increasing sequences of projections with @xmath235 and @xmath236 in addition ( see e.g. ( * ? ? ?
13 , theorems 13.24 , 13.33 ) ) , @xmath237 multiplying the equality @xmath121 by @xmath230 on the left and by @xmath231 on the right , we obtain @xmath238 clearly , @xmath239 and @xmath240 are normal operators for every @xmath241 it follows from the preceding paragraph that @xmath242 thus , @xmath243 since @xmath244 it follows that @xmath245 @xmath246 as @xmath247 thus @xmath248 @xmath249 in measure .
therefore , for every @xmath250 we have @xmath251 in measure as @xmath247 taking @xmath252 we complete the proof .
lemma [ l5 ] below is the key tool used to extend fuglede - putnam theorem from @xmath253measurable operators to measurable ones .
[ l5 ] let @xmath15 be a semifinite von neumann algebra and let @xmath254 be a finite projector .
then there exists partition of unity @xmath255 , such that every von neumann algebra @xmath256 , has a faithful normal semifinite trace @xmath257 with @xmath258 .
it is well known that let @xmath15 a commutative von neumann algebra @xmath37 is @xmath16-isomorphic to the @xmath16-algebra @xmath259 of all essentially bounded measurable complex - valued functions defined on a measure space @xmath260 with the measure @xmath261 satisfying the direct sum property ( we identify functions that are equal almost everywhere ) ( see e.g. ( * ? ? ? * ch .
7 , 7.3 ) ) .
the direct sum property of a measure @xmath261 means that the boolean algebra of all projections of the @xmath35-algebra @xmath259 is order complete , and for any non - zero @xmath262 there exists a non - zero projection @xmath263 such that @xmath264 .
the direct sum property of a measure @xmath261 is equivalent to the fact that the functional @xmath265 is a semi - finite normal faithful trace on the algebra @xmath259
. therefore there exists partition of unity @xmath266 , such that @xmath267 is faithful normal finite trace on @xmath268 for every @xmath269 .
let @xmath270 be a @xmath16-isomorphism from @xmath37 onto the @xmath35-algebra @xmath259 .
denote by @xmath271 the set of all measurable real - valued functions defined on @xmath260 and taking values in the extended half - line @xmath272 $ ] ( functions that are equal almost everywhere are identified ) . by ( * ? ? ?
v , 2 , theor . 2.34 and prop .
2.35 ) there exits a faithful semifinite normal extended center valued trace @xmath273 @xmath274 such that @xmath275 . thus characteristic functions @xmath276 corresponding to sets @xmath277 @xmath278 is partition of the unit element @xmath279 of boolean algebra @xmath280 . in addition
@xmath281 for all @xmath282 .
it is clear that @xmath283 is partition of unity in @xmath284 .
in addition , the functional @xmath285,$ ] given by the formula @xmath286 is a faithful normal finite trace on @xmath287 . in particular , @xmath288 setting @xmath289 and @xmath290 we complete the proof .
choose @xmath300 so large that projections @xmath301 @xmath302 and @xmath303 are finite .
let @xmath304 be a finite projection given by the formula @xmath305 let @xmath306 be the partition of unity constructed in lemma [ l5 ] .
we have @xmath307 it follows from lemma [ l5 ] that , for a given @xmath294 @xmath308 a standard argument yields @xmath309 where the right hand side is taken in the algebra @xmath310 it follows that @xmath311 @xmath312 and @xmath313 are @xmath314measurable operators for every @xmath315 theorem [ t2 ] implies that @xmath316 the assertion follows now from lemma [ central partition ]
. by the ( constructive ) definition of the algebra @xmath317 there exist central projections @xmath318 @xmath319 and @xmath320 such that @xmath321 @xmath322 and @xmath323 and such that @xmath324 denote the triple @xmath325 by @xmath300 and set @xmath326 since @xmath327 it follows from lemma [ semifinite measurable ] that @xmath328 in other words ( here , we let @xmath329 ) , @xmath330 since @xmath331 is a partition of unity which consists of central projections , it follows from lemma [ central partition ] that @xmath332 repeating the argument for @xmath333 and , after that , for @xmath334 we complete the proof .
recall that @xmath335 choose central projections @xmath318 @xmath319 and @xmath320 such that @xmath321 @xmath322 and @xmath323 and such that @xmath336 denote the triple @xmath325 by @xmath300 and let @xmath326 we have @xmath337 by the classical fuglede - putnam theorem , we have @xmath328 the same argument as in lemma [ semifinite locally measurable ] yields the assertion .
it is well known that for every von neumann algebra @xmath15 there exist central projectors @xmath338 such that @xmath339 @xmath340 is the semifinite von neumann algebra and @xmath341 is the purely infinite von neumann algebra ( see , for example , ( * ? ? ?
* ch.2 , 2.2 ) ) .
we have @xmath342 lemmas [ semifinite locally measurable ] and [ purely infinite ] imply that @xmath343 summing these equalities , we complete the proof .
let @xmath365 and @xmath366 by theorem [ t3 ] we have that @xmath367 thus @xmath368 @xmath369 let a borel function @xmath32 be given by the formula @xmath370 @xmath371 and let @xmath372 @xmath369 since @xmath373 and since @xmath374 @xmath371 it follows from spectral theorem that @xmath375 @xmath369
since @xmath368 it follows that @xmath376 that is , @xmath377 theorem 13.33 in @xcite yields that @xmath378 @xmath379 for every spectral projection @xmath27 of the operator @xmath364 thus , @xmath380 @xmath369 summing these equalities , we obtain @xmath381 @xmath382 lemma [ pervoe ] states that @xmath26 for every spectral projection @xmath27 of the operator @xmath364 again applying lemma [ pervoe ] to the couple @xmath383 we obtain that @xmath28 for every spectral projection @xmath27 of the operator @xmath3 and for every spectral projection @xmath29 of the operator @xmath384 @xmath385 let @xmath230 ( respectively , @xmath231 ) be the spectral projection for @xmath3 ( respectively , @xmath6 ) corresponding to the set @xmath386 @xmath241 denote @xmath387 and @xmath388 by spectral theorem , we have @xmath389 bounded operators @xmath390 and @xmath391 are normal and their spectral projections commute . by the spectral theorem for bounded operators , these operators commute and , therefore , @xmath392 thus , @xmath393 @xmath394 taking into account that @xmath395 and using lemma [ l0 ] , we obtain @xmath396 again appealing to lemma [ l0 ] , we obtain . m. v. ahramovich , v.i . chilin and m. a. muratov , fuglede - putnam theorem in the algebra of locally measurable operators , _ proceedings fifth dr .
george bachman memorial conference , indian journal of mathematics _ , 2013 ,
* 55 * , 13 - 20 . | we extend the fuglede - putnam theorem from the algebra @xmath0 of all bounded operators on the hilbert space @xmath1 to the algebra of all locally measurable operators affiliated with a von neumann algebra . |
in two previous works @xcite , we investigated the persistence of extreme value laws ( evls ) whenever a dynamical system is perturbed throughout random transformations .
we considered an i.i.d .
stochastic process @xmath0 with values in the measurable space @xmath1 and with probability distribution @xmath2 . after associating to each @xmath3 a map @xmath4 acting on the measurable space @xmath5 into itself , we considered the random orbit starting from the point @xmath6 and generated by the realization @xmath7 : @xmath8 here , the transformations @xmath4 should be considered close to each other and the suitably rescaled scalar parameter @xmath9 is the strength of such a distance
. we could therefore define a markov process @xmath10 on @xmath5 with transition function @xmath11 where @xmath12 is a measurable set , @xmath13 and @xmath14 is the indicator function of a set @xmath15 .
a probability measures @xmath16 is called a _ stationary measure _ if for any measurable @xmath15 we have : @xmath17 we call it an absolutely continuous stationary measure ( _ acsm _ ) , if it has a density with respect to the lebesgue measure whenever @xmath5 is a metric space .
+ in this work we consider a different type of perturbation , the observational noise , which consists in replacing the orbit of the point @xmath13 at time @xmath18 , namely @xmath19 with @xmath20 there are several physical motivations to investigate the behavior of this kind of perturbation . in fact , as lalley and noble wrote in @xcite : _ `` ... in this model our observations take the form @xmath21 , where @xmath22 are independent , mean zero random vectors .
in contrast with the dynamical noise model ( e.g ; the random transformations ) , the noise does not interact with the dynamics : the deterministic character of the system , and its long range dependence , are preserved beneath the noise .
due in part to this dependence , estimation in the observational noise model has not been broadly addressed by statisticians , though the model captures important features of many experimental situations . '' _
judd @xcite , quoted in @xcite , also pointed out that : _ `` ...
the reality is that many physical systems are indistinguishable from deterministic systems , there is no apparent small dynamic noise , and what is often attributed as such is in fact model error . ''
moreover , a system contaminated by the observational noise raises the natural and practical question whether it would be possible to recover the original signal , in our case the deterministic orbit @xmath23 . in the last years
a few techniques have been proposed for such a noise reduction @xcite : we remind here the remarkable schreiber - lalley method @xcite , which provides a very consistent algorithm to perform the noise reduction when the underlying deterministic dynamical system has strong hyperbolic properties .
another interesting work shows that in the computation of some statistical quantities , the dynamical noise corresponding to random transformations could be considered as an observational noise with the cauchy distribution @xcite .
finally , the paper @xcite proves concentration inequalities for systems perturbed by observational noise .
+ the present work try to re - frame the previous findings in terms of extreme value theory ( evt ) by adding a further motivation driven by the applicability of the whole evt for dynamical systems to experimental data .
it should be a general praxis to check the role of instrument - like - perturbations before applying dynamical systems techniques to experimental datasets . in this sense , the dynamical systems considered in this paper share several properties with observed time series , as the observational noise acts exactly as a physical instrument .
the goal is to exploit the recent advancements of the evt for dynamical systems to define in a more rigorous way the extremes of time series .
a relevant example of a successful application of the theory presented in this paper to experimental datasets is given in @xcite , where temperature data are analyzed with the algorithmic procedure presented in section 4.2 .
more specifically , our interest is to understand which way the results obtained on deterministic dynamical systems are altered by the addition of observational noise and in which cases one can recover classical evls .
we start the discussion by summarizing the main findings of the evt for dynamical systems .
+ the first rigorous mathematical approach to evt in dynamical systems goes back to the pioneer paper by collet @xcite .
important contributions have successively been given in @xcite , @xcite , @xcite and in @xcite .
here we briefly recall the main findings deferring to the previous papers for the full demonstrations .
+ let us consider a dynamical system @xmath24 , where @xmath5 is the invariant set in some manifold , usually @xmath25 , @xmath26 is the borel @xmath27-algebra , @xmath28 is a measurable map and @xmath29 a probability @xmath30-invariant borel measure .
+ in order to adapt the evt to dynamical systems , we follow @xcite , by considering the stationary stochastic process @xmath31 given by : @xmath32 where dist is a distance on the ambient space @xmath5 , @xmath33 is a given point and @xmath34 is a suitable function which will be specified later
. this particular functional form has been introduced first by collet @xcite and allows for a direct connection between recurrence properties around a point of the phase space @xmath33 and the existence of evls .
the object of interest is the distribution of @xmath35 , where @xmath36 ; we say that we have an evl for @xmath37 if there is a non - degenerate distribution function @xmath38 $ ] with @xmath39 and , for every @xmath40 , there exists a sequence of levels @xmath41 , @xmath42 , such that @xmath43 and for which the following limit holds : @xmath44 the motivation for using a normalizing sequence @xmath45 satisfying comes from the case when @xmath46 are independent and identically distributed ( i.i.d ) . in this setting , it is clear that @xmath47 , being @xmath48 the cumulative distribution function for the variable @xmath49 .
hence , condition implies that @xmath50 as @xmath51 . note that in this case @xmath52 is the standard exponential distribution function . by choosing the sequence @xmath53 as one parameter families like @xmath54 , where @xmath55 and @xmath56 for all @xmath57 and @xmath34 as above , whenever the variables @xmath58 are i.i.d .
, if for some constants @xmath59 we have @xmath60 .
when the convergence occurs at continuity points of @xmath61 ( @xmath61 is non - degenerate ) then @xmath62 converges to one of the three evls rewritable in terms of the generalized extreme value ( gev ) distribution as : @xmath63^{-1/{\kappa}}\right\}. \label{cumul}\ ] ] here @xmath64 is the shape parameter also called the tail index : when @xmath65 , the distribution corresponds to a gumbel evl ; when the tail index is positive , it corresponds to a frchet evl ; when @xmath66 is negative , it corresponds to a weibull evl .
the evl obtained depends on the kind of observable chosen .
in particular , in @xcite the authors have shown that , once taken the observable : @xmath67 one gets a gumbel evl , here @xmath68 . in the next section
we prove the existence of gumbel law for the maps perturbed with observational noise .
it is in fact possible to introduce other observables than the one specified above in order to get convergence towards frechet and weibull evls .
however , for any choice different from @xmath69 , the tail index depends on the local property of the measure . for every sequence @xmath70
satisfying we define : @xmath71 when @xmath72 are not independent , the standard exponential law still applies under some conditions on the dependence structure .
these conditions are the following : * condition*[@xmath73[cond : d2 ] we say that @xmath74 holds for the sequence @xmath75 if for all @xmath76 and @xmath77 , @xmath78 where @xmath79 is decreasing in @xmath80 for each @xmath77 and @xmath81 when @xmath82 for some sequence @xmath83 .
now , let @xmath84 be a sequence of integers such that @xmath85 * condition*[@xmath86[cond : d ] we say that @xmath87 holds for the sequence @xmath88 if there exists a sequence @xmath84 satisfying and such that @xmath89 by following freitas and freitas @xcite theorem 1 , if conditions @xmath74 and @xmath87 hold for @xmath90 then there exists an evl for @xmath37 and @xmath91 + in the paper @xcite , freitas , freitas and todd made the interesting observation that the extreme value laws are intimately related to the concept of local recurrence , in particular to the first hitting time function in small sets . their analysis has been brought to systems perturbed with random transformations in @xcite . in section 3 we will show that these kind of results hold also for systems perturbed with observational noise , first by adapting the definition of first hitting time and then by showing that it follows an exponential law tempered by the strength of the perturbation .
+ we remark that the analogy between extreme value laws and local recurrences is possible for particular observables of the type @xmath92 where @xmath33 is a given point .
as explained in @xcite , the additional choice @xmath93 allows to get the gumbel law ( a direct proof of this fact is given after proposition [ p2 ] ) , and moreover it brings information on the local structure of the invariant measure , as we will explain in a moment .
this represents the main motivation for using such observable although a more detailed discussion and other motivations can be found in @xcite .
+ we conclude this introduction by stressing what we believe is an interesting and very general result .
we will see that the probability @xmath94 which we will use to rule out the distribution of the maxima is the product between the invariant measure @xmath29 and the measure of the noise @xmath95 .
whenever one is able to prove the existence of an extreme value law for the process @xmath96 with the observable ( [ g1 ] ) , then proposition [ p2 ] shows how the two sequences @xmath97 and @xmath98 appearing in the affine choice for @xmath45 ( see above ) , are related to the local behavior of the measure @xmath29 at a local scale given by the intensity of the noise .
this local behavior is related to the fine structure of the measure @xmath29 .
we have therefore a useful tool to detect the fine geometric properties of the invariant measure by calibrating the normalizing sequence @xmath45 until we get the gumbel law , and this at different scales for the noise .
we now show how to adapt the evt for orbits of dynamical systems perturbed by instrument - like - error .
each instrument introduces an effect related to the combined accuracy and precision of the measure by replacing the real ( unknown ) value with the biased indicated by the instrument itself . in general ,
if the real dynamics can be represented by the map @xmath30 , what one observes is formally : @xmath99 where @xmath100 is the truncation introduced by the instrument precision and @xmath101 is a random displacement from the real value .
this displacement is what we have defined as observational noise .
here it is rewritten as @xmath102 , with the parameter @xmath103 .
+ although the paper is dedicated to the observational noise , for completeness and in light of applications on time series , we want remark also some relevant properties of truncations .
the role of the truncation , important also for numerical computations , has been discussed in the book by knuth @xcite and then analyzed , among others , by @xcite . on the @xmath104 digit ,
the truncation acts essentially as a random noise of variance @xmath105 .
if @xmath106 the measure underlying @xmath107 will match the one of the original dynamics given by the map @xmath30 , if not the support will appear as a collection of dirac s deltas , precluding the convergence to the gev distribution @xcite .
we defer to section 4.1 for a numerical study on the truncation error which should point out for which values of @xmath104 one should take truncations into account and whether @xmath108 ( the common truncation corresponding to a double precision representation ) is a good choice for representing the properties of a deterministic dynamics .
we thus proceed to show that evls persist for chaotic dynamical systems , whenever they are perturbed with the observational noise @xcite .
the proof we give is reminiscent of that of theorem d in @xcite .
we first point out that , in order to guarantee the stationarity of the random process involved in the distribution of the maxima , we need to evaluate the observable @xmath34 at the point @xmath109 , where @xmath13 and @xmath110 is a random vector .
for this reason and in order to avoid ambiguities , we will choose @xmath5 as a torus .
another alternative would be to take @xmath5 which is strictly sent into itself by @xmath30 , @xmath111 , and with all the components of @xmath110 small enough ( see , for instance , proposition 4.5 in @xcite ) .
this choice is often invoked for random transformation acting of bounded domains of @xmath112 the paragraph below contains the assumptions on the systems which allow us to prove our main results .
+ * assumption m * : we consider maps @xmath30 defined on the torus @xmath113 with norm @xmath114 and satisfying : * there exists a finite partition ( mod-@xmath115 ) of @xmath5 into open sets @xmath116 , namely @xmath117 , such that @xmath30 has a lipschitz extension on the closure of each @xmath118 with a uniform and strictly larger than @xmath119 lipschitz constant @xmath120 , @xmath121 , @xmath122 * @xmath30 preserve a borel probability measure @xmath29 which is also mixing with decay of correlations given by @xmath123 where the constant @xmath124 depends only on the map @xmath30 , @xmath125 denotes the @xmath126 norm with respect to @xmath29 and finally @xmath127 is a banach space included in @xmath128 where @xmath129 denotes the lebesgue ( haar ) measure on @xmath130 : the corresponding norm will be denoted with @xmath131 we will also need @xmath29 to be equivalent to @xmath129 with density in @xmath132 .
* assumption n * : we consider a sequence @xmath133 of i.i.d vector - valued random variables which take values in the hypershpere @xmath134 centered at @xmath115 and of radius @xmath119 , @xmath135 and with common distribution @xmath95 , which we choose absolutely continuous with density @xmath136 , namely @xmath137 , with @xmath138 is a vector with @xmath139 components ; all these components are independent and distributed with common density @xmath140 ; the product of such marginals @xmath140 s gives @xmath141 . ]
+ the paper @xcite contains examples of endomorphisms verifying the assumption m , in particular one - dimensional rychlik maps endowed with bounded variation functions and multidimensional piecewise uniformly expanding maps endowed with quasi - hlder observables . in order to get the decay of correlations ( [ dc ] ) , one needs more regularity for @xmath30 , usually @xmath142 our next proofs crucially depend on the decay against @xmath126 functions ; property for characteristic functions which , for some systems , is easier to show . ]
we believe that one could weaken such assumption and extend the theory to invertible maps , but that would need a different approach : in order to support this claim , section 4 will contain numerical computations on examples which are not covered by our analytical results , but which show similar behaviors .
we will comment further on these issues in section 3 - consequence 3 .
the random orbits @xmath102 generates a new random process when an observable is computed along them .
suppose that @xmath34 is a measurable real function defined on @xmath5 ; we take @xmath143 , where @xmath33 is a given point in @xmath5 .
the process : @xmath144 is endowed with the probability @xmath145 defined on the product space @xmath146 with the product @xmath27-algebra ; a point in this space is the couple @xmath147 where @xmath148 projects onto the @xmath18th component @xmath149 before checking conditions @xmath74 and
@xmath150 we notice that , contrarily to the random setting studied in @xcite , the random variable @xmath151 depends now not only on the initial condition @xmath13 , but also on the random variable @xmath110 and this makes @xmath94 stationary . with the given choice of the observable @xmath34 ,
the set @xmath152 is explicitly given by @xmath153 for convenience , we will also set @xmath154 , the ball of center @xmath33 and radius @xmath155 + [ p1 ] let us suppose that our dynamical systems verifies the assumption m and it is perturbed with observational noise satisfying the assumption n. then conditions @xmath74 and @xmath87 hold for the observable @xmath34
. we will give the proof when @xmath30 is continuous on the torus .
the extension to the piecewise lipschitz case is straightforward and it could be done as explained in the analogous extension proofs of propositions 4.2 and 4.5 in @xcite to which we defer for further details .
we begin to check condition @xmath74 by estimating the contribution given by the first term on the l.h.s of eq .
[ d1 ] : @xmath156 @xmath157 the set of integration variables here being @xmath158 with @xmath159 .
we apply fubini s theorem and factorize the integrals by exploiting the independence of the variables @xmath160 , so that the previous expression becomes : @xmath161 let us introduce the measurable functions : @xmath162 @xmath163 with @xmath164 and @xmath165 then eq . [ e1 ] can be rewritten as @xmath166 . by the decay of correlations assumption , we get @xmath167 @xmath168 where @xmath124 is a constant depending only on @xmath30 . in the previous equation , the second term on the l.h.s corresponds to @xmath169 which is the second term on the l.h.s . of the condition @xmath74 .
let us note , and this will be useful later , that @xmath74 holds with @xmath170 for some @xmath171 and @xmath172 , with @xmath173 in order to deal with condition @xmath87 , we follow the same strategy as in @xcite .
we begin to define the _ approximated first return time of the point _
@xmath6 in @xmath174 in the following way : we fix the couple @xmath175 and we set @xmath176 notice that we keep fixed the variables @xmath177 while iterating the point @xmath6 .
moreover , instead of @xmath6 , we require the initial condition @xmath178 to be in @xmath174 .
then we define the _ approximated first return time of the set _
@xmath174 into itself as : @xmath179 we observe that @xmath180 where @xmath181 is the highest rate of separations for the points .
we use the symbol @xmath114 indifferently to denote _ distance _ and _ diameter_. the notation @xmath182 , where @xmath183 and @xmath184 is a subset of @xmath185 , stands for the set @xmath186 ) .
we now fix some sequences @xmath187 going to infinity and such that @xmath188 , where @xmath189 is the sequence defined in ( [ def : un ] ) .
therefore , whenever @xmath190 @xmath191 , then @xmath192 which in turn implies that @xmath193 .
since @xmath194 we have @xmath195 @xmath196 @xmath197 '' takes into account the multiplicative factor given by the volume @xmath198 of the unit hypersphere @xmath199 .
we now have : @xmath200}(\nu\times\theta^{\mathbb{n}})\{(x,\overline{\xi } ) ; x_0>u_m ; \ x_j > u_m\}=\ ] ] @xmath201}\int d\nu \{\int d\theta(\xi ) { \bf 1}_{\{g(x+\epsilon \xi)>u_m\}}\int d\theta(\xi ' ) { \bf 1}_{\{g(t^jx+\epsilon \xi')>u_m\}}\}+\ ] ] @xmath202 where @xmath203 again , the first term ( i ) on the l.h.s .
can be estimated by using decay of correlations applied to the ( same ) observable @xmath204 we easily get : @xmath205 } \{\mathbb{p}(u_m)^2 + c \mathbb{p}(u_m)j^{-2}\ } \le \frac{(n\mathbb{p}(u_m))^2}{k_m}+mc\mathbb{p}(u_m)\sum_{j=\alpha_m}^{[\frac{m}{k_m}]}j^{-2}= { \mathcal o}(\frac{\tau^2}{k_m}+\tau \sum_{j=\alpha_m}^{[\frac{m}{k_m}]}j^{-2})\underset{m\to+\infty}{\longrightarrow } 0\ ] ] since @xmath206 + for the second term ( ii ) we use holder s inequality and the fact that @xmath29 is equivalent to @xmath129 '' to signify that equivalence , namely there exists a positive constant @xmath207 such that for any measurable set @xmath15 we have that @xmath208 . ] with essentially bounded density and @xmath129 is translationally invariant : @xmath209 @xmath210 and @xmath211 we finally have @xmath212 which goes to zero with the prescribed assumptions for @xmath213 and @xmath214 +
in this section we would like to point out a few interesting properties of the observational noise .
we start by an explicit calculation of the quantity @xmath215 defined in ( [ eq : un ] ) for the observable ( [ g1 ] ) .
the computation is done in @xmath216 , but the generalization to higher dimensions is trivial .
as we have anticipated in section 1 , the following proposition requires only the existence of the gumbel law for the process under investigation , which is of course true for systems verifying proposition [ p1 ] .
[ p2 ] let us suppose that the one - dimensional dynamical systems @xmath217 verifies a gumbel law for the process @xmath218 endowed with the probability @xmath219 and also that @xmath95 is the lebesgue measure measure on @xmath199 .
then the linear sequence @xmath220 , defined by [ eq : un ] , verifies : @xmath221 we begin to observe that @xmath222 @xmath223 since the variable @xmath110 must stay in the ball of center @xmath115 and radius @xmath119 .
+ let us now introduce @xmath224 with @xmath225 ; and observe that @xmath226 this bound is independent from @xmath77 and integrable is _ exactly _ lebesgue , since its translational invariance property allows us to get rid of the variable @xmath6 in the center of the ball . ] .
+ we can apply the theorem of dominated convergence since @xmath227 having passed the limit inside , the integral finally gives @xmath215 : @xmath228 and therefore @xmath229 which is exactly the _ gumbel law_. + whenever @xmath230 a similar computation immediately gives that the linear sequence @xmath220 verifies : @xmath231 the following useful consequences will be exploited in the next section : * _ consequence 1 _ + the scaling parameter @xmath98 depends on the target point @xmath33 via the local density of the invariant measure in a ball of radius given by the @xmath232 .
let us start by considering the case of absolutely continuous invariant measures .
if the point @xmath33 is visited with less frequency , the local density will be of lower order in @xmath232 , which means that one should go to higher values of @xmath77 in order to have a reliable statistics .
this is the case , for instance , for the points @xmath233 for the map introduced by hemmer @xcite : + @xmath234 + and defined on the interval @xmath235 $ ] .
the invariant density @xmath141 can be computed directly by inspection and reads : @xmath236 .
therefore @xmath237 and @xmath238 + a complementary issue will appear whenever the map exhibits a laminar behavior in some regions of the phase space and therefore it will spend there a lot of time .
this happens , for instance , for the well - known map of pomeau - manneville @xcite , which could be written as : + @xmath239 + the origin @xmath115 is a neutral fixed point and , for @xmath240 , the density of the absolutely continuous invariant measure @xmath29 behaves like @xmath241 in the neighborhood of @xmath115 @xcite .
therefore , @xmath242 and @xmath243 we will analyze in the next section which finite size effects arise for these two examples . + _ warning _ : strictly speaking the previous two maps do not fit with the assumptions of proposition [ p1 ] .
in fact , in both cases it is not possible to prove a polynomial decay of correlations against _ all _ @xmath126 functions as it was shown in @xcite . on the other hand
, an inspection of the proof shows that we only require the @xmath126 property for characteristic functions , so that , in principle , such a decay could be obtained .
this was achieved , for instance , in the case of rotations by using a fourier series technique @xcite .
an additional problem concerns the rate of decay . for the pomeau - manneville map it is of order @xmath244 @xcite , so that it fits with our assumption whenever @xmath245 ; for the hemmer s map the situation is worst since the correlations decay as @xmath246 @xcite .
nevertheless , we conjecture that proposition [ p1 ] could be applied to the latter case as well , as we will argue in _ consequence 3 _ , and as we will show numerically in the next section . *
_ consequence 2 _ + let us observe that , if we define the linear scaling factor as above , we have convergence towards the gumbel law @xmath247 for _ any _ @xmath33 . this shows that the extremal index defined and studied in @xcite is @xmath119 everywhere implying that there are no points which behave like unstable fixed points of deterministic dynamical systems . * _ consequence 3 _ + although we were able to prove proposition [ p1 ] whenever the invariant measure is equivalent to lebesgue , we pointed out that proposition [ p2 ] is true for any invariant measure , providing that one can show the existence of an evl .
let us therefore suppose that a given dynamical system with invariant measure @xmath248 , not necessarily equivalent to lebesgue , admits an extreme value law for the observable @xmath34 under the observational noise
. then we can apply proposition [ p2 ] , which only requires that @xmath95 is lebesgue .
remember that in the expression for @xmath98 we are considering a fixed positive size for the noise @xmath232 .
let us suppose that at this scale we have @xmath249 , where @xmath250 is usually an estimation of the geometric and fractal properties of @xmath29 at the point @xmath33 . on @xmath251 is called _ ahlfohrs upper semi - regular _ if there is a constant @xmath252 and a real number @xmath253 such that for all non - empty open intervals
@xmath254 @xmath255 + ahlfors upper semi - regular measures include fractal measures like the measures of maximal dimension of dynamically defined cantor sets , i.e. cantor sets that arise from smooth expanding repellors . more generally , given an invariant measure @xmath248 and whenever the limit @xmath256 exists @xmath6-@xmath248-almost everywhere
, then this limit equals the _ hausdorff dimension of the measure _
@xmath248 , @xmath257 @xcite , also called _
information dimension_. in some cases @xmath258 is given by suitable relations between lyapunov exponents and entropies , formulae better known as _ kaplan - yorke _ and _ ledrappier - young _ : see @xcite for a detailed exposition of these issues .
] in this case , if the ambient space has dimension @xmath139 , the linear scaling parameter @xmath98 has the form : @xmath259 therefore , we have an useful technique to detect the local dimensions of the measure , with a _
finite _ resolution given by the strength of the noise . this will also allow us to compute directly the distribution of the maxima with the linearization , given the explicit expression of the @xmath45
this would be particularly useful whenever the invariant measure is singular and therefore the gev distribution does not admit a probability density function : see section 3.1 in @xcite for a detailed discussion on this point . * _ consequence 4 _ + as we said in the introduction , another interesting property of the observational noise is its direct relationship with the statistics of first hitting times in small sets .
we first define the _ first hitting time of the set @xmath174 _ in a slightly different manner which respect to the quantity @xmath260 introduced above .
given the triple @xmath261 , we set : @xmath262 let us notice that , contrarily to @xmath260 , the error increment changes at each step since we are now dealing with a _ true _ random orbit ; this easily implies that + @xmath263 + by using the expression of @xmath45 found above and setting consequently @xmath264 , in dimension @xmath119 , we could rewrite the previous formula as , for @xmath265 @xmath266 + which shows that the first hitting time follows an exponential law tempered by the strength noise @xmath267
in this section we discuss some important implications connected to the introduction of observational noise in finite time series .
in particular , through numerical experiments devised on low dimensional maps , we show that the influence of truncations and observational noise is related to the intensity of the perturbation applied .
some of the theoretical findings presented in the previous sections present a practical interest in a wide range of applications namely the analysis of the role of truncation errors for instrument with low accuracy , the statistics of points visited sporadically in the analysis of recurrence of time series and the possibility of computing attractor dimension by using eq .
[ superformula ] as an alternative way to other techniques . for each maps , we present and comment numerical experiments and outline possible further applications . + before introducing such examples , let us make a useful comment .
a close inspection to the proof of proposition [ p1 ] , shows that the parameter @xmath232 appears in the denominator of one factor in the r.h.s . of the term ( ii ) at the end of the proof .
this means that the convergence gets better when @xmath232 is large , which is not surprising since a large value of the perturbation implies a more stochastic independence of the process . on the other hand ,
if we want to use the form of the linear scaling parameter @xmath98 to catch the local properties of the invariant measure @xmath29 , we need small values of @xmath268 .
a judicious balance of the value of @xmath232 between these two regimes is therefore necessary when we pursue such numerical analysis .
the experiment consists in producing 30 orbits starting from different initial conditions taken on the support of the truncated measure .
the length of the orbit is fixed according to the results presented in @xcite to be such that @xmath270 . in order to analyze the effect of varying the bin length combined to the order of the truncation
, we consider three different values of @xmath77 .
each series of @xmath34 is therefore divided in @xmath271 bins and in each of them the maxima @xmath272 are extracted and then fitted to the gev model via the l - moments procedure described in @xcite .
note also that the points @xmath33 are chosen after applying the truncation .
the results are shown in fig [ trunc ] where the behavior of the shape parameter @xmath273 is compared to the asymptotic gumbel law @xmath274 . in the deterministic limit @xmath275 the usual gumbel evl
is recovered , whereas for @xmath276 the discretization of the invariant measure becomes relevant and the asymptotic evl appears as a collection of dirac deltas thus producing a divergence of the shape parameter from @xmath115 .
this is exactly what is visible in fig [ trunc ] : the convergence gets worse when @xmath277 and at @xmath278 we are already unable to fit the gev distribution for all the @xmath33 points considered , although by increasing the bin length @xmath77 the convergence improves as one would expect .
these results , as we tested , are reproducible in other maps and gives a more general indication that computing asymptotic properties on truncated series leads to estimation errors and divergence even at high order of truncation .
+ with the introduction of the observational noise , the scaling parameter @xmath98 depends on the target point @xmath33 _ via _ the local density of the invariant measure in a ball whose radius is given by the error @xmath232 .
in _ consequence 1 _ we said that if the point @xmath33 is visited with less frequency , the local density is of lower order with respect to @xmath232 , which means that one should go to higher values of @xmath77 in order to have a reliable statistics . here
we want to test that the order of @xmath77 needed to get convergence to the asymptotic @xmath98 is lower for a highly recurrent point then for a sporadic one .
as highly recurrent point we choose @xmath279 for the pomeau - manneville map in eq .
[ pm ] and as sporadic one the point @xmath119 of the map introduced by hemmer and reported in eq .
[ hemmereq ] .
the experiment consists in computing @xmath280 realizations of the maps perturbed with observational noise .
again , we fit the maxima of the observable @xmath34 to the gev distribution by using the @xmath281-moments procedure and compare the values of @xmath98 obtained experimentally to the theoretical ones stated in proposition [ p2 ] .
we report here the results for three different bin lengths @xmath282 in fig .
[ pmfig ] for the pomeau manneville map and in fig .
[ hemmerfig ] for hemmer map .
the figures show how @xmath98 varies as a function of the noise @xmath283 , in terms of @xmath284 . in both cases
we observe convergence towards the theoretical values ( solid lines ) for high values of @xmath268 ( low orders in @xmath284 ) whereas in the limit of weak noise one must increase the bin lengths to get convergence .
the main result to be highlighted here is the better convergence of highly recurrent points with respect to the ones visited sporadically .
this important property can be used to study time series recurrences and identify extremes as the points visited rarely for which the convergence towards the asymptotic parameters is bad .
the main advantage of studying recurrence properties in this way over applying other techniques is due to the built - in test of convergence of this method : even for a point rarely recurrent there will be a time scale @xmath77 such that the fit converges . for smaller @xmath77
, we can therefore such a @xmath33 as a _ sporadically recurrent _ point of the orbit as explained in @xcite .
there we show how to use this property to define rigorous recurrences in long temperature records collected at several weather stations .
the convergence or divergence of the fit allows us for discriminating between temperatures belonging to the normal variability associated to the time scales defined by the bin length ( e.g. the seasonal cycle ) or as anormal temperature if there are no or few recurrences in @xmath77 .
as we have already said in _ consequence 3 _ , if the invariant measure is not absolutely continuous , one could still perform the previous analysis , but both the ambient space dimension @xmath139 and the local dimension @xmath250 will enter in the computation of @xmath98 via eq .
[ superformula ] .
this formula can be used in principle to test whether a map has a fractal support by comparing the local dimensions with the the ambient space dimension : it is enough to check how the obtained @xmath98 depend on the intensity of the noise @xmath283 .
we should point out that , as we said in the footnote 1 , we are targeting the hausdorff dimension of the measure .
we test this idea on the classical iterated function system @xmath285 used to produce a cantor set : @xmath286 where @xmath287 $ ] , and we set @xmath288 .
therefore , at each time step , we have the same probability to iterate @xmath289 or @xmath290 ; the _ balanced _ invariant measure associated to the map and the way to construct the orbit to detect the maxima are described in @xcite .
the results for different @xmath77 are reported in fig .
[ bm ] for an average among @xmath280 different realizations and three different bin lengths @xmath291 .
experimental data follow the prediction of eq [ superformula ] with @xmath292 as hausdorff dimension and @xmath216 as ambient space dimension but only up to a certain noise intensity @xmath284 beyond which they reach a _
plateau_. a justification for this behavior is that when the noise intensity is very small , the system needs longer trajectories - higher @xmath77 - to explore the ball of radius @xmath268 .
this gives an implicit criterion for the selection of the bin length @xmath77 needed to observe reliable results and it says that one should be careful in applications where the intensity of the observational noise is small compared to the scale of the dynamics .
this analysis is confirmed by the results obtained for the lozi map : @xmath293 for which we consider the classical set of parameter @xmath294 and @xmath295 .
young @xcite proved the existence of the srb measure for the lozi map and found the value @xmath296 for the hausdorff dimension of the measure by computing the lyapunov exponents and using a kaplan - yorke like formula .
the experiments is exactly the same described for the cantor set and the results are presented in fig .
[ lozifig ] for the values of @xmath297 .
agreement with the theoretical behavior of @xmath98 , as expected from eq .
[ superformula ] represented in fig .
[ lozifig ] by the solid straight lines , is found for small values of @xmath284 which means for large @xmath268 .
for the lozi map the convergence is worse then for the cantor case .
this phenomenon has been already observed in @xcite and , up to now , there is not a clear explanation for it .
however , by scanning numerically the ( @xmath298 ) space , one can infer the intervals of such parameters such that the @xmath98 converge towards the asymptotic results : for example , when @xmath299 , an order of noise intensity @xmath300 is needed to get convergence to the predicted theoretical values .
since there is only a limited range of @xmath284 such that the @xmath98 convergence to the prediction of eq .
[ superformula ] , one has to take extremely care on using the results obtained with this method to estimate fractal dimensions . a good strategy to overcome
this problem is to discard the value of @xmath98 which show no dependence on @xmath284 and check that the remaining points are sufficient to perform a linear fit of @xmath98 vs @xmath284 .
+ the possibility of computing fractal dimensions by using a random perturbation of a dynamical system is an interesting fact : when random perturbations are applied , one usually expects the orbit to explore asymptotically the ambient space and this normally hide the fractal property of the measure .
instead , as we said at the beginning of this section , for reasonable choices of @xmath77 and @xmath284 , one gets accurate information on the fractal dimension by fitting the @xmath250 parameter of eq . [ superformula ] .
this finding opens new questions , for instance if it is possible to get analogous formulas in the case of random transformations .
an application can be to adapt this theory for multi - scale systems whose description is usually made through an approximation of a dynamic consisting of a deterministic and a stochastic components .
for example , by tuning the strength of the stochastic components , one can study how the noise affects the structure of the deterministic dynamics ; these aspects will be explored in forthcoming works .
* acknowledgements*. sv was supported by the anr- project _ perturbations _
, by the pics ( projet international de coopration scientifique ) , proprits statistiques des sysmes dynamiques detrministes et alatoires , with the university of houston , n. pics05968 and by the projet mode ter com supported by _
rgion paca , france_. sv thanks support from fct ( portugal ) project ptdc / mat/120346/2010 .
the authors aknowledge jean - ren chazottes , jorge milhazes freitas and paul manneville for useful discussions .
df and sv acknowledge the newton institute in cambridge where this work was completed during the program _
mathematics for the fluid earth_. ayta h , freitas j - m , vaienti s ( 2012 ) laws of rare events for deterministic and random dynamical systems . to appear : _
trans amer .
soc._. lucarini v , faranda d , turchetti g , vaienti s , ( 2012 ) extreme value distributions for singular measures , _ chaos _ , * 22 * , 023135 .
lalley , s.p . ,
nobel , a.b .
( 2006 ) denoising deterministic time series , _ dynamics od pde _ , 3 ( 4 ) , 259 - 279 .
judd , k ( 2003 ) chaotic - time - series reconstruction by the bayesian paradigm : right resultsby wrong methods .
e _ , 67:026212 .
mcgoff k. , mukherjee s. ( 2012 ) pillai n.s .
, statistical inference for dynamical systems : a review , arxiv:1204.6265v1 [ math.st ] kantz h , schreiber t ( 2004 ) , _ nonlinear time series analysis _ , cambridge university press , 2nd edition .
lalley sp ( 2001 ) removing the noise from chaos plus noise , in _ nonlinear dynamics and statistics , ( cambridge , 1998 ) _ , birkhuser boston , ma , 233 - 244 .
lalley sp ( 1999 ) beneath the noise , chaos , _ ann .
_ , 27(2 ) , 461 - 479 .
schreiber t ( 1993 ) , extremely simple nonlinear noise - reduction method , _ phys .
e _ , 47:2401 - 2404 .
schreiber t , grassberger p ( 1991 ) a simple noise - reduction method for real data , _ phys . lett .
, 411 - 418 .
strumik m , macek wm , redaelli s ( 2005 ) , discriminating additive from dynamical noise for chaotic time series , _ phys .
, 72 , 036219 maldonado c ( 2012 ) fluctuation bounds for chaos plus noise in dynamical systems , _ j. stat
_ , 148 , 548 - 564 , collet p ( 2001 ) statistics of closest return for some non - uniformly hyperbolic systems .
syst . _ 21(2 ) : 401 - 420 .
freitas a - c - m , freitas j - m ( 2008 ) on the link between dependence and independence in extreme value theory for dynamical systems , _ stat .
_ , 78 , 9 , 1088 - 1093 freitas a - c - m , freitas j - m , todd m ( 2010 ) hitting time statistics and extreme value theory .
theory and rel .
_ 147(3 - 4 ) : 675 - 710 .
freitas a - c - m , freitas j - m , todd m ( 2011 ) extreme value laws in dynamical systems for non - smooth observations .
_ j. stat .
phys . _ 142(1 ) : 108 - 126 .
freitas a - c - m , freitas j - m , todd m ( 2012 ) the extremal index , hitting time statistics and periodicity .
advances in mathematics 231(5 ) : 2626 - 2665 .
gupta c ( 2010 ) extreme - value distributions for some classes of non - uniformly partially hyperbolic dynamical systems .
30(3 ) : 757 - 771 .
kifer y ( 1986 ) ergodic theory of random transformations , _
birkhuser_. faranda d , freitas , jm , lucarini v , turchetti g , vaienti s , ( 2012 ) extreme value statistics for dynamical systems with noise .
_ nonlinearity _ 26 2597 .
faranda d , vaienti s , ( 2013 ) a recurrence based technique for detecting genuine extremes in instrumental temperature records _ geophysical research letters _ 40.21 : 5782 - 5786 .
knuth d - e ( 2006 ) .
the art of computer programming .
addison - wesley , 2006 .
thiel m , romano m - c , kurths j , meucci r , allaria e , arecchi
f - t ( 2002 ) .
influence of observational noise on the recurrence quantification analysis .
_ physica d : nonlinear phenomena _ , 171(3 ) : 138 - 152 . miyano t ( 1996 ) time series analysis of complex dynamical behavior contaminated with observational noise .
j. bifurcat . chaos _
6(11 ) : 2031 - 2045 .
faranda d , mestre m - f , turchetti g ( 2012 ) analysis of round off errors with reversibility test as a dynamical indicator .
j. bifurcat . chaos _ 22(09 ) .
faranda d , lucarini v , turchetti g , vaienti s ( 2011 ) numerical convergence of the block - maxima approach to the generalized extreme value distribution .
_ j. stat .
phys . _ 145(5 ) 1156 - 1180 . faranda d , lucarini v , manneville p and wouters j ( 2012 ) on using extreme values to detect global stability thresholds in multi - stable systems : the case of transitional plane couette flow .
_ arxiv preprint _ arxiv:1211.0510 .
pomeau y , manneville p ( 1980 ) .
intermittent transition to turbulence in dissipative dynamical systems .
_ communications in mathematical physics _ 74.2 : 189 - 197 .
hemmer p. c. ( 1984 ) the exact invariant density for a cusp - shaped return map .
_ journal of physics a _ , 17.5 : l247 .
young ls ( 1985 ) bowen - ruelle measures for certain piecewise hyperbolic maps .
_ transactions of the american mathematical society _ 41 - 48 .
alves j , freitas j , luzzatto s. , vaienti s. ( 2011 ) from rates of mixing to recurrence times via large deviations , _ advances in mathematics _ ,
* 228 * ( 2011 ) 1203 - 123 liverani , c , saussol b , vaienti , s ( 1999 ) a probabilistic approach to intermittency , _ ergodic theory and dynamical systems _ , * 19 * , 671 - 685 cristadoro g , haydn n , marie ph , vaienti s. ( 2010 ) statistical properties of intermittent maps with unbounded derivative , _ nonlinearity _ , 23,1071 - 1096 young , l - s ( 1982 ) dimension , entropy and lyapunov exponents , _ ergod .
sys_. , * 2 * , 109 - 124 .
eckmann j - p , ruelle d ( 1985 ) , ergodic theory of chaos and strange attractors , _ rev .
phys _ * 57 * , 617656 vs @xmath104 , the digit where the truncation has been applied .
error - bars display the average of @xmath273 over 30 realizations and the standard deviation of the samples for @xmath77=300 ( blue ) , @xmath77=1000 ( magenta ) and @xmath77=3000 ( red ) .
the green line correspond to the gumbel law @xmath301 .
@xmath302 for all the cases considered .
, width=453 ] vs intensity of the noise in terms of @xmath284 ( we recall that @xmath283 ) for the pomeau manneville map ( eq . [ pm ] ) .
we recall that dashed error - bars display the average of @xmath98 over 30 realizations and the standard deviation of the sample .
solid lines the theoretical values .
the blue , red and magenta curves respectively refers to @xmath282 , @xmath279 .
@xmath302 for all the cases considered.,width=377 ] vs intensity of the noise in terms of @xmath284 ( we recall that @xmath283 ) for the hemmer map ( eq . [ hemmereq ] ) .
dashed error - bars display the average of @xmath98 over 30 realizations and the standard deviation of the sample .
solid lines the theoretical values .
the blue , red and magenta curves respectively refers to @xmath282 , @xmath303 .
@xmath302 for all the cases considered.,width=377 ] vs intensity of the noise in terms of @xmath284 ( we recall that @xmath283 ) for the cantor ifs ( eq . [ ifs_c ] ) . dashed errorbars display the average of @xmath98 over 30 realizations and the standard deviation of the sample .
solid lines the theoretical values .
the blue , red and magenta curves respectively refers to @xmath282 , @xmath33s randomly chosen on the attractor .
@xmath302 for all the cases considered.,width=377 ] vs intensity of the noise in terms of @xmath284 ( we recall that @xmath283 ) for the lozi map ( eq . [ lozi ] ) . dashed errorbars display the average of @xmath98 over 30 realizations and the standard deviation of the sample .
solid lines the theoretical values .
the blue , red and magenta curves respectively refers to @xmath282 .
the points @xmath33 are randomly chosen on the attractor .
@xmath302 for all the cases considered , width=377 ] | in this paper we prove the existence of extreme value laws for dynamical systems perturbed by instrument - like - error , also called observational noise .
an orbit perturbed with observational noise mimics the behavior of an instrumentally recorded time series .
instrument characteristics - defined as precision and accuracy - act both by truncating and randomly displacing the real value of a measured observable . here
we analyze both these effects from a theoretical and numerical point of view .
first we show that classical extreme value laws can be found for orbits of dynamical systems perturbed with observational noise
. then we present numerical experiments to support the theoretical findings and give an indication of the order of magnitude of the instrumental perturbations which cause relevant deviations from the extreme value laws observed in deterministic dynamical systems .
finally , we show that the observational noise preserves the structure of the deterministic attractor .
this goes against the common assumption that random transformations cause the orbits asymptotically fill the ambient space with a loss of information about any fractal structures present on the attractor . |
radios ( cr ) @xcite@xcite , having capable of sensing spectrum availability , is considered as a promising technique to alleviate spectrum scarcity due to current static spectrum allotment policy @xcite .
traditional cr link availability is solely determined by the spectrum sensing conducted at the transmitter ( i.e. cr - tx ) .
if the cr - tx with packets to relay senses the selected channel to be available , it precedes this opportunistic transmission . to facilitate the spectrum sensing , at time instant @xmath0
, we usually use a hypothesis testing as follows .
@xmath1 where @xmath2 means the observation at cr - tx ; @xmath3 represents signal from primary system ( ps ) ; @xmath4 is the interference from co - existing multi - radio wireless networks ; @xmath5 is additive white gaussian noise ( awgn ) .
they are all random variables at time @xmath0 .
we can conduct this hypothesis testing in several ways based on different criterions and different assumptions
@xcite-@xcite : 1 .
energy detection @xcite-@xcite : energy detection is widely considered due to its simple complexity and no need of _ a priori _ knowledge of ps .
however , due to noise and interference power uncertainty , the performance of energy detection severely degrades @xcite@xcite , and the detector fails to differentiate ps from the interference .
cyclostationary detection @xcite-@xcite : stationary ps signal can be exploited to achieve a better and more robust detector .
stationary observation coupled with the periodicity of carriers , pulse trains , repeating spreading code , etc .
, results in a cyclostationary model for the signal , which can be exploited in frequency domain by the spectral correlation function @xcite@xcite , with high computational complexity and long observation duration .
3 . locking detection @xcite@xcite : in practical communication systems , pilots and preambles are usually periodically transmitted to facilitate synchronization and channel estimation , etc .
these known signals can be used for locking detection to distinguish ps from noise and the interference . however
, locking detection requires more _ a priori _ information about ps , including frame structure , modulation types , and coding schemes , etc .
covariance - based detection @xcite-@xcite : because of the dispersive channels , the utility of multiple antennas , or even over - sampling , the signals from ps are correlated and can be utilized to differentiate ps from white noise .
the existence of ps can be determined by evaluating the structure of covariance matrix of the received signals .
the detector can be implemented blindly @xcite by singular value decomposition , that is , it requires no _ a priori _ knowledge of ps and noise , but needs good computational complexity .
wavelet - based detection @xcite : the detection is implemented by modeling the entire wideband signals as consecutive frequency subbands where the power spectral characteristic is smooth within each subband but exhibits an abrupt change between neighboring subbands , at the price of high sampling rate due to wideband signals . however , above spectrum sensing mechanisms , focusing on physical layer detection or estimation at cr - tx , ignore the spectrum availability at cr receiver .
we could illustrate the insufficiency of traditional spectrum sensing model , especially to network crs .
due to existence of fading channels and noise uncertainty along with limited sensing duration @xcite , even when there is no detectable transmission of ps during this venerable period , the receiver of this opportunistic transmission ( i.e. cr - rx ) may still suffer from collisions from simultaneous transmission(s ) , as fig .
[ fig_1 ] shows .
the cr - rx locates in the middle of cr - tx and ps - tx and ps activities are hidden to cr - tx , which induces a challenge to spectrum sensing .
we can either develop more powerful sensing techniques such as cooperative sensing @xcite-@xcite to alleviate hidden terminal problem , or a more realistic mathematical model for spectrum sensing what we are going to do hereafter .
the organization of this paper is as follows .
we elaborate a realistic definition of link availability and system model in section ii and present general spectrum sensing with / without cooperation in section iii an iv respectively . in section
v , realistic operation of crn is suggested to investigate impacts of spectrum sensing and cooperative scheme on network operation .
numerical results and examples are illustrated in section vi .
finally , conclusions are made in section vii .
from a viewpoint of information theory , spectrum sensing can be modeled as a binary channel that transmit cr link availability ( one bit information in link layer ) to cr - tx with transition probabilities representing spectrum sensing capability , probability of missing detection and probability of false alarm .
therefore , traditional spectrum sensing mechanisms could be explained by a mathematical structure of defining link availability . cr link availability , between cr - tx and cr - rx , is specified by an indicator function @xmath6 [ d1 ] cr - tx senses the spectrum and determines link availability based on its observation as @xmath7 [ l5 ] traditional spectrum sensing for cr link suggests @xmath8 .
the definition of link availability is pretty much similar to the clear channel assessment ( cca ) of medium access control ( mac ) in the ieee 802.11 wireless local area networks ( wlan ) , or the medium availability indicator in @xcite .
we may have a correspondence between link availability in dynamic channel access of cognitive radio networks ( crn ) , and the cca in mac of wlan . as we explain in figure [ fig_1 ] and/or take interference into testing scenario
, we may note that * lemma [ l5 ] * is not generally true . to generally model spectrum
sensing , including hidden terminal scenarios , we have to reach two simultaneous conditions : ( 1 ) cr - tx senses the link available to transmit ( 2 ) cr - rx can successfully receive packets , which means no ps signal at cr - rx side , nor significant interference to prohibit successful cr packet reception ( i.e. beyond a target sinr )
. in other words , at cr - rx , @xmath9 where @xmath10 is the sinr threshold at cr - rx for outage in reception over fading channels , @xmath11 is the received power from opportunistic transmission from cr - tx , @xmath12 is the power contributed from ps simultaneous operation for general network topology such as ad hoc , @xmath13 is the total interference power from other co - existing radio systems @xcite , and @xmath14 is band - limited noise power , with assuming independence among ps , cr , interference systems , and noise . based on this observation , cr link availability should be composed of localized spectrum availability at cr - tx and cr - rx , which may not be identical in general .
the inconsistency of spectrum availability at cr - tx and cr - rx is rarely noted in current literatures .
however , this factor not only suggests spatial behavior for cr - tx and cr - rx but also is critical to some networking performance such as throughput of crn , etc .
@xcite developed a brilliant two - switch model to capture distributed and dynamic spectrum availability .
however , @xcite focused on capacity from information theory and it is hard to directly extend the model in studying network operation of crn .
actually , two switching functions can be generalized as indicator functions to indicate the activities of ps based on the sensing by cr - tx and cr - rx respectively @xcite .
generalizing the concept of @xcite@xcite to facilitate our study in spectrum sensing and further impacts on network operation , we represent the spectrum availability at cr - rx by an another indicator function .
the true availability for cr - rx can be indicated by @xmath15 please note that the activity of ps estimated at cr - rx in @xcite may not be identical to @xmath16 .
that is , even when cr - rx senses that ps is active , cr - rx may still successfully receive packets from cr - tx if the received power from cr - tx is strong enough to satisfy ( [ e28 ] ) .
we call this rate - distance nature @xcite that is extended from an overlay concept @xcite . therefore , we consider a more realistic mathematical model for cr link availability that can be represented as multiplication ( i.e. and operation ) of the indicator functions of spectrum availability at cr - tx and cr - rx to satisfy two simultaneous conditions for cr link availability . [ p1 ] @xmath17 to obtain the spectrum availability at cr - rx ( i.e. @xmath18 ) and to eliminate hidden terminal problem , a handshake mechanism has been proposed by sending request to send ( rts ) and clear to send ( cts ) frames .
however , the effectiveness of rts / cts degrades in general ad hoc networks @xcite .
furthermore , since crs have lower priority in the co - existing primary / secondary communication model , crs should cherish the venerable duration for transmission and reduce the overhead caused by information exchange and increases spectrum utilization accordingly .
therefore , the next challenge would be that @xmath18 can not be known _ a priori _ at cr - tx , due to no centralized coordination nor information exchange in advance among crs when cr - tx wants to transmit . as a result , general spectrum sensing turns out to be a composite hypothesis testing . in this paper
, we introduce statistical inference that is seldom applied in traditional spectrum sensing to predict / estimate spectrum availability at cr - rx and to regard it as performance lower bound in general spectrum sensing . further examining * proposition [ p1 ] *
, we see that prediction of @xmath18 is necessary when @xmath19 , which is equivalent to prediction of @xmath20 . in this paper
, we model @xmath18 when @xmath19 as a bernoulli process with the probability of spectrum availability at cr - rx @xmath21 .
the value of @xmath22 exhibits spatial behavior of cr - tx and cr - rx and thus impacts of hidden terminal problem .
if @xmath22 is large , cr - rx is expected to be close to cr - tx and hidden terminal problem rarely occurs ( and vise versa ) .
the prediction of @xmath18 at cr - tx can be modeled as a hypothesis testing , that is , detecting @xmath18 with _ a priori _
probability @xmath22 but no observation . to design optimum detection ,
we consider minimum bayesian risk criterion , where bayesian risk is defined by @xmath23 in ( [ e30 ] ) , @xmath24 , @xmath25 , and @xmath26 denotes the normalized weighting factor to evaluate costs of @xmath27 and @xmath28 , where @xmath29 represents prediction of @xmath20 .
we will show that the value of @xmath30 relates to the outage probability of cr link in section v. since @xmath18 is unavailable at cr - tx , we have to develop techniques to `` obtain '' some information of spectrum availability at cr - rx . inspired by the crn tomography @xcite@xcite , we may want to derive the statistical inference of @xmath18 based on earlier observation .
it is reasonable to assume that cr - tx can learn the status of @xmath18 at previous times when @xmath19 , which is indexed by @xmath31 .
that is , at time @xmath31 , cr - tx can learn the value of @xmath32,\mathbf{1}^{rx}[n-2],\ldots$ ] .
in other words , we can statistically infer @xmath33 $ ] from @xmath32 $ ] , @xmath34 $ ] , @xmath35 , @xmath36 $ ] , where @xmath37 is the observation depth .
this leads to a classical problem from bayesian inference .
[ l8 ] through the laplace formula @xcite , the estimated probability of spectrum availability at cr - rx is @xmath38 where @xmath39}$ ] .
inference - based spectrum sensing at cr - tx thus becomes @xmath40 where @xmath41 is in ( [ e1 ] ) .
since the optimum detector under bayesian criterion is the likelihood ratio test @xcite , we have @xmath42 where @xmath43 denotes the cost incurred by determining @xmath44 when @xmath45 . according to bayesian risk in ( [ e30 ] )
, we have @xmath46 and @xmath47 .
rearranging the inequality , we obtain the proposition .
cr - tx believes cr link is available and forwards packets to cr - rx if the probability of spectrum available at cr - rx @xmath22 is high enough .
otherwise , cr - tx is prohibited from using the link even when cr - tx feels free for transmission because it can generate unaffordable cost , that is , intolerable interference to ps or collisions at cr - rx .
spectrum sensing at cooperative node , which can be represented @xmath48 , is to explore more information about @xmath18 and therefore alleviates hidden terminal problem .
we can use fig .
[ fig_2 ] to depict the scenario . in case
the existence of obstacles , @xmath18 is totally orthogonal to @xmath49 .
@xmath48 is useful simply because of more correlation between @xmath48 and @xmath18 . from above observation , we only care about correlation of @xmath18 and @xmath48 when @xmath19 and assume @xmath50 thus the correlation between @xmath48 and @xmath18 , @xmath51 , and corresponding properties become @xmath52 [ l1 ] @xmath51 is a strictly concave function with respect to @xmath53 if @xmath54 but a strictly convex function if @xmath55 .
in addition , @xmath48 and @xmath18 are independent if and only if @xmath56 , i.e , @xmath57 .
let @xmath58 then taking first order and second order differentiation with respect to @xmath22 , we have @xmath59\end{aligned}\ ] ] where @xmath60 for @xmath61 and @xmath62 .
in addition , @xmath63 for @xmath64 .
similarly , we have @xmath65 for @xmath64 .
therefore , combining the second and the third terms , and the fourth and the last terms in the bracket of @xmath66 , we have @xmath67 < 0\end{aligned}\ ] ] if @xmath53 . since @xmath68 , we prove the first statement of the lemma . for the second statement , obviously , @xmath56 if @xmath48 and @xmath18 are independent .
reversely , if @xmath56 , i.e. , @xmath57 , we have @xmath69 similarly , we can show @xmath70 and complete the proof . by statistical inference
, cr - tx can learn statistical characteristic of @xmath18 and @xmath48 , i.e. , @xmath71 , by previous observations . from a viewpoint of hypothesis testing
, we would like to detect @xmath18 with _ a priori _
probability @xmath22 and one observation @xmath48 , which is the detection result at the cooperative node .
in addition , probability of detection and probability of false alarm at the cooperative node are @xmath72 and @xmath73 respectively .
[ p4 ] spectrum sensing with one cooperative node becomes @xmath74 where @xmath75 is the complement of @xmath48 , @xmath76 and @xmath77 .
the likelihood ratio test based on observed signal @xmath48 can be written as follows .
for @xmath78 , @xmath79 for @xmath80 , @xmath81 rearranging the above inequalities , we obtain the proposition .
it is interesting to note that cooperative spectrum sensing is not always helpful , that is , it does not always further decrease bayesian risk .
we see that if @xmath22 is large ( greater than @xmath82 ) , that is , hidden terminal problem rarely occurs because either cr - rx is close to cr - tx or cr - tx adopt cooperative sensing to determine @xmath49 , prediction of @xmath18 is unnecessary at cr - tx . on the other hand ,
if @xmath22 is small ( less than @xmath83 ) , cr - tx is prohibited from forwarding packets to cr - rx even with the aid of cooperative sensing . in the following ,
we adopt minimum error probability criterion ( i.e. , @xmath84 ) and give an insight into the condition that cooperative sensing is helpful .
although we set @xmath84 , we do not lose generality because we can scale _ a priori _
probability @xmath22 to @xmath85 as @xmath86 .
applying * lemma [ l1 ] * and the fact that @xmath87 when @xmath84 , we can reach the following corollary .
[ cor1 ] if we adopt minimum error probability criterion , spectrum sensing with one cooperative node becomes @xmath88}\mathbf{1}^{tx } , & \text{if $ |\rho|\leq\psi$}\\ ( \mathbf{1}_{[\rho>0]}\mathbf{1}^{co}+\mathbf{1}_{[\rho<0]}\bar{\mathbf{1}}^{co})\mathbf{1}^{tx } , & \text{if $ |\rho|>\psi$ } \end{cases}\ ] ] where @xmath89}$ ] is an indicator function , which is equal to 1 if the statement @xmath90 is true else equal to 0 , and @xmath91 the effectiveness of a cooperative node only depends on the correlation of spectrum availability at cr - rx and the cooperative node .
if the correlation is low , information provided by the cooperative node is irrelevant to the spectrum sensing which degenerates to ( [ e13 ] ) . by establishing a simple indicator model in link layer ,
we mathematically demonstrate the limit of a cooperative node in general spectrum sensing .
it is natural to ask what will happen for multiple cooperative nodes and how to compare sensing capability among cooperative nodes . in the following ,
we provide metrics to measure sensing capability of cooperative nodes from link and network perspectives . before exploring multiple cooperative nodes , we introduce notations and properties to systematically construct relation between joint probability mass function ( pmf ) and marginal pmf of spectrum availability at cooperative nodes .
we first define notations in the following . for an @xmath92 matrix @xmath93 and two @xmath94 vectors @xmath95 and @xmath96 , @xmath97:&\text{the } i\text{th row of } \mathbf{a } & \mathbf{a}[\star , j]:&\text{the } j\text{th column of } \mathbf{a}\\ \mathbf{a}^t:&\mathbf{a}^t[i , j]=\mathbf{a}[j , i ] & \mathbf{a}^{rc}:&\mathbf{a}^{rc}[\star , j]=\mathbf{a}[\star , n - j+1]\\ \mathbf{a}^{rr}:&\mathbf{a}^{rr}[i,\star]=\mathbf{a}[m - i+1,\star ] & \mathbf{1}_{m\times n}:&\mathbf{a}[i , j]=1\\ \mathbf{0}_{m\times n}:&\mathbf{a}[i , j]=0 & \mathbf{i}_{n}:&n\times n \text { identity matrix}\\ \mathbf{u}\odot\mathbf{v}:&\mathbf{u}\odot\mathbf{v}[i]=\mathbf{u}[i]\mathbf{v}[i ] & \mathbf{u}\preceq\mathbf{v}:&\mathbf{u}[i]\leq \mathbf{v}[i]\\ \|\mathbf{u}\|_p:&(\sum_i{|\mathbf{u}[i]|^p})^{1/p } & \mathbf{1}_{[\mathbf{u}\geq0]}:&\mathbf{1}_{[\mathbf{u}\geq0]}[i]=\mathbf{1}_{[\mathbf{u}[i]\geq0 ] } \end{matrix}\ ] ] let @xmath98 where @xmath99 is a @xmath100 matrix , @xmath101 , @xmath102 , @xmath103 , and for @xmath104 @xmath105 the role of @xmath106 ( or @xmath107 ) is to specify arrangements of joint pmf and marginal pmf such that their relation can be easily established by @xmath108 . in the following ,
we show properties of @xmath99 and @xmath109 .
[ l2 ] let @xmath110=1,1\leq i\leq k\}$ ] , @xmath111 and we have @xmath112&=\mathbf{1}_{[\mathcal{i}_{m , k}(j)\supseteq\mathcal{i}_{n , k}(i ) ] } \label{e18}\end{aligned}\ ] ] where @xmath113 denotes number of elements in the set @xmath114 .
since @xmath115 is a @xmath116 matrix , from ( [ e17 ] ) , ( [ e22 ] ) , we conclude that @xmath117,1\leq j\leq\binom{k}{m}$ ] contains all possible permutations of @xmath118 ones and @xmath119 zeros .
[ t1 ] let @xmath120 .
let @xmath121 and @xmath122 be @xmath123 matrices , @xmath124 and @xmath125 be @xmath126 matrices , and @xmath127 .
@xmath128 are nonsingular and their inverse matrices become @xmath129 let @xmath130=\pr(\mathbf{1}^{co}_1=\mathbf{a}^1_{m , k}[1,i],\ldots,\mathbf{1}^{co}_k=\mathbf{a}^1_{m , k}[k , i]| \mathbf{1}^{rx}=s,\mathbf{1}^{tx}=1)$ ] , where @xmath131 , @xmath132 and @xmath133 , @xmath134 denotes spectrum availability at the @xmath135th cooperative node , and let @xmath136 therefore , @xmath137 characterizes joint pmf of spectrum availability at @xmath138 cooperative nodes , @xmath139 .
similarly , we divide @xmath137 into two parts .
let @xmath140 and @xmath141 be @xmath142 vectors , @xmath143 and @xmath144 be @xmath145 vectors , and @xmath146 .
in addition , let @xmath147=\pr(\mathbf{1}^{co}_{k_1}=s,\ldots,\mathbf{1}^{co}_{k_m}=s| \mathbf{1}^{rx}=s,\mathbf{1}^{tx}=1,\{k_1,\ldots , k_m\}=\mathcal{i}_{m , k}(j))\ ] ] @xmath131 , @xmath148 and @xmath149 , which specifies the @xmath118th order marginal pmf of @xmath139 .
arrange them into a vector form , we define @xmath150 please note that @xmath151=\mathbf{p}^{(1)}_k[2^k]$ ] and @xmath152=\mathbf{p}^{(0)}_k[1]$ ] . [ t2 ] marginal and joint pmf of spectrum availability at k cooperative nodes satisfy @xmath153 or @xmath154 [ cor3 ] @xmath137 provides equivalent information to @xmath155 , or specifically , @xmath156\mathbf{b}^{(s)}_k , s=0,1\ ] ] where @xmath157 assume there are @xmath158 cooperative nodes with corresponding spectrum availability @xmath159 and their joint pmf conditionally on @xmath160 and @xmath19 , @xmath161 .
[ p2 ] spectrum sensing with multiple cooperative nodes becomes @xmath162\prod_{i=1}^{k } { [ \mathbf{a}^1_k[i , j]\mathbf{1}^{co}_k\oplus(1-\mathbf{a}^1_k[i , j])\bar{\mathbf{1}}^{co}_k]}}\ ] ] and @xmath163,\alpha \mathbf{p}^{(1)}_k[i]\}}\ ] ] where @xmath164}$ ] and @xmath165 denotes or operation . by the likelihood ratio test , we have @xmath166}{\mathbf{p}^{(0)}_k[i ] } { \substack{\overset{\hat{\mathbf{1}}^{link}=1}{\geq}\\ \underset{\hat{\mathbf{1}}^{link}=0 } { < } } } \frac{w(1-\alpha)}{\alpha}\ ] ] therefore , the optimum detector becomes @xmath167=\hat{\mathbf{1}}^{link}|_{\mathbf{1}^{co}_1=\mathbf{a}^1_k[1,i],\cdots,\mathbf{1}^{co}_k=\mathbf{a}^1_k[k , i ] } = \mathbf{1}_{[\alpha\mathbf{p}^{(1)}_k[i]-w(1-\alpha)\mathbf{p}^{(0)}_k[i]\geq0]}\ ] ] with the result and combining binary arithmetic we obtain ( [ e14 ] ) . in terms of bayesian risk ,
@xmath168 \pr(\hat{\mathbf{1}}^{link}=1|\mathbf{1}^{co}_1=\mathbf{a}^1_k[1,i],\cdots,\mathbf{1}^{co}_k=\mathbf{a}^1_k[k , i],\mathbf{1}^{rx}=0,\mathbf{1}^{tx}=1)+}\\ & \quad \alpha\mathbf{p}^{(1)}_k[i ] \pr(\hat{\mathbf{1}}^{link}=0|\mathbf{1}^{co}_1=\mathbf{a}^1_k[1,i],\cdots,\mathbf{1}^{co}_k=\mathbf{a}^1_k[k , i],\mathbf{1}^{rx}=1,\mathbf{1}^{tx}=1)]\\ & = \sum_{i=1}^{2^k}{[w(1-\alpha)\mathbf{p}^{(0)}_k[i]\mathbf{\gamma}[i]+ \alpha\mathbf{p}^{(1)}_k[i](1-\mathbf{\gamma}[i])]}\\ & = \sum_{i=1}^{2^k}{\min\{w(1-\alpha)\mathbf{p}^{(0)}_k[i],\alpha \mathbf{p}^{(1)}_k[i]\}}\end{aligned}\ ] ] we consider two cooperative nodes insightfully to understand how multiple cooperative nodes improve performance of spectrum sensing . for spectrum availability at two cooperative nodes
@xmath169 , let @xmath170 and @xmath171 .
in addition , their joint probability is specified as follows .
when @xmath172 , ps is likely inactive and then @xmath173 and @xmath174 are independent . on the other hand , when @xmath175 , @xmath173 and @xmath174 are correlated with correlation @xmath176 . with the constraints on @xmath177
, we have @xmath178 where @xmath179 . in case @xmath173 and @xmath174 are conditionally independent , it leads to conventional assumption in cooperative spectrum sensing . for two cooperative nodes with independent spectrum availability ,
spectrum sensing becomes @xmath181}\mathbf{1}^{co}_i+\mathbf{1}_{[\rho_i<0]}\bar{\mathbf{1}}^{co}_i ) } , & \text{if $ \alpha_{(3)}\leq\alpha<\alpha_{(4)}$}\\ \mathbf{1}^{tx}(\mathbf{1}_{[\rho_k>0]}\mathbf{1}^{co}_k+\mathbf{1}_{[\rho_k<0]}\bar{\mathbf{1}}^{co}_k ) , & \text{if $ \alpha_{(2)}<\alpha<\alpha_{(3)},k=\arg\max_im_r(i)$ } \\
\mathbf{1}^{tx}\prod_{i=1}^2{(\mathbf{1}_{[\rho_i>0]}\mathbf{1}^{co}_i+\mathbf{1}_{[\rho_i<0]}\bar{\mathbf{1}}^{co}_i ) } , & \text{if $ \alpha_{(1)}<\alpha\leq\alpha_{(2)}$ } \\ 0 , & \text{if $ \alpha\leq\alpha_{(1)}$ } \end{cases}\ ] ] where @xmath182 @xmath183/(\mathbf{p}^{(1)}_2[i]+w\mathbf{p}^{(0)}_2[i])$ ] , @xmath184}\delta^+_{i}+\mathbf{1}_{[\rho_i<0]}\delta^-_{i}$ ] , @xmath185 , @xmath186 , and @xmath187 denotes the correlation between @xmath134 and @xmath18 .
we list all valid orders of @xmath188 in ( [ e36 ] ) and the the spectrum sensing is determined according to the value of @xmath22 by arguing ( [ e14 ] ) .
we observe that @xmath22 is high ( @xmath189 ) , any one of cooperative nodes helps the spectrum sensing , which leads the spectrum sensing to or operation .
however , when @xmath22 is low ( @xmath190 ) , cr - tx requires more evidence to claim available link and the spectrum sensing becomes and operation .
in addition , it is interesting to note that there exists a region of @xmath22 ( @xmath191 ) such that cr - tx only depends on one of two cooperative nodes , which motivates us to define a metric or a measure to evaluate cooperative nodes .
* reliability * of a cooperative node is measured by @xmath192 .
@xmath134 is said to be more or equally * reliable * than ( to ) @xmath193 if @xmath194 , which is denoted by @xmath195 . for @xmath158 cooperative nodes with independent spectrum availability , without loss of generality
, we assume @xmath196 for @xmath197 and @xmath198 for @xmath199 .
spectrum sensing becomes @xmath200}$ ] , where @xmath201 and @xmath202 , @xmath203 .
since @xmath159 are independent , the likelihood ratio test becomes @xmath204}{\mathbf{p}^{(0)}_k[i ] } & = \prod_{i\in\mathcal{c}^+}\frac{\beta_i}{1-\gamma_i } \prod_{i\in\{1,\ldots , n\}\setminus\mathcal{c}^+}\frac{1-\beta_i}{\gamma_i } \prod_{i\in\mathcal{c}^-}\frac{1-\beta_i}{\gamma_i } \prod_{i\in\{n+1,\ldots , k\}\setminus\mathcal{c}^-}\frac{\beta_i}{1-\gamma_i}\\ & = \prod_{i\in\mathcal{c}^+}\frac{\beta_i\gamma_i}{(1-\beta_i)(1-\gamma_i ) } \prod_{i\in\{1,\ldots , n\}}\frac{1-\beta_i}{\gamma_i } \prod_{i\in\mathcal{c}^-}\frac{(1-\beta_i)(1-\gamma_i)}{\beta_i\gamma_i } \prod_{i\in\{n+1,\ldots , k\}}\frac{\beta_i}{1-\gamma_i}\\ & = \prod_{i\in\mathcal{c}^+\cup\mathcal{c}^-}m_r(i ) \prod_{i\in\{1,\ldots , n\}}\frac{1-\beta_i}{\gamma_i } \prod_{i\in\{n+1,\ldots , k\}}\frac{\beta_i}{1-\gamma_i } { \substack{\overset{\hat{\mathbf{1}}^{link}=1}{\geq}\\ \underset{\hat{\mathbf{1}}^{link}=0 } { < } } } \frac{w(1-\alpha)}{\alpha}\end{aligned}\ ] ] taking logarithm at both sides and rearranging the formula , we complete the proof .
we observe that reliability @xmath205 is used to quantify the information of @xmath18 provided by the @xmath135th cooperative node .
please note that if @xmath134 is independent of @xmath18 , @xmath206 and @xmath134 is irrelevant to the spectrum sensing .
therefore , reliability can imply sensing capability , that is , one cooperative node with higher reliability has better sensing capability .
reliability can thus serve a criterion to select cooperative nodes when number of cooperative nodes is limited due to appropriate overhead caused by information exchange .
specially , if there are @xmath158 equally reliable cooperative nodes , each cooperative node provides equal amount of information about @xmath18 and the spectrum sensing rule turns out to be counting rule .
this is a generalization from identically and independently distributed ( i.i.d . )
observations @xcite in conventional distributed detection to equally reliable observations .
when there exists correlation between spectrum availability at cooperative nodes , the joint probabilities @xmath208 are shifted by @xmath209 , as in ( [ e33 ] ) .
for example , if the correlation is positive @xmath210 and @xmath211 increase while @xmath212 and @xmath213 decrease .
if the correlation increases , eventually , the order of @xmath188 will switch and the spectrum sensing in ( [ e32 ] ) will change accordingly . however , the correlated case becomes tedious and we consider a simple but meaningful example , where cooperative nodes have symmetric error rates ( i.e. @xmath214 ) and reliability becomes @xmath215}\beta_i+\mathbf{1}_{[\rho_i<0]}(1-\beta_i)$ ] .
similarly , with the aid of ( [ e14 ] ) and ( [ e33 ] ) , the spectrum sensing can be easily derived .
[ p3 ] for two cooperative nodes with correlated spectrum availability and symmetric error rate satisfying @xmath216 , and @xmath217 , the spectrum sensing would be ( [ e32 ] ) with modifications according to @xmath209 .
@xmath218 where @xmath219 @xmath220 , @xmath221 , and @xmath222 denotes xor operation .
all possible switching orders of @xmath188 according to @xmath209 are listed in ( [ e37 ] ) and the first and the last orders are impossible unless an additional condition is satisfied to make the regions of @xmath209 valid , i.e. @xmath223 and @xmath224 respectively . since @xmath173 and @xmath174
are correlated when @xmath175 ( i.e. @xmath207 or @xmath225 ) , not only @xmath173 and @xmath174 alone but also the identity of @xmath173 and @xmath174 ( i.e. @xmath226 or @xmath227 ) can provide information about @xmath18 and thus can be used to determine cr link availability .
this is actually similar to covariance - based detection .
for example , if @xmath228 , @xmath173 and @xmath174 are probably identical when spectrum is unavailable at cr - rx , i.e. @xmath229 , and @xmath174 in ( [ e32 ] ) is then replaced by @xmath226 . furthermore , when @xmath209 increases and is greater than @xmath230 , the roles of @xmath173 and @xmath226 switch because the identity of @xmath173 and @xmath174 can provide more information about @xmath231 than @xmath173 alone .
alternatively , when @xmath232 , the results can be similarly explained .
in addition , it is interesting to note that even if @xmath174 is independent to @xmath18 ( i.e. @xmath233 ) , @xmath174 may become helpful due to the correlation between @xmath173 and @xmath174 . in the next section
, we will further investigate impacts of correlation between @xmath173 and @xmath174 on network operation . in crn or self - organizing networks , due to lacking of centralized coordination ,
each node in crn can only sense and exchange local information .
in addition , dynamic wireless channels and mobility of nodes make the situation severer and one node can only acquire information within limited sensing duration .
we can either design systems under simplified assumptions , which may result in severe performance degradation , or apply advanced signal processing techniques based on minimax criterion @xcite , robust to outliers in networks , as we are going to do hereafter . to derive the optimum bayesian detection in * proposition [ p2 ] *
, we have to acquire joint pmf of spectrum availability at @xmath158 cooperative nodes , which may require long observation interval to achieve acceptable estimation error .
if we only have up to the @xmath138th order marginal pmf ( related to capability of observation ) , i.e @xmath234 according to * lemma [ cor3 ] * , our design criterion becomes minimax criterion , that is , we find the least - favorable joint pmf @xmath161 such that maximizes bayesian risk and then conduct the optimum bayesian detection under that joint probability .
therefore , the problem can be formulated as follows .
[ robust cooperative sensing ] cooperative spectrum sensing with limited statistical information @xmath234 becomes ( [ e14 ] ) with @xmath235 replaced by @xmath236 , where @xmath237 the last equality in the objective function is based on the fact that @xmath238 and that the sum of probability distribution is equal to one .
the result is reasonable because in order to minimize the objective function , the likelihood ratio @xmath239/\mathbf{p}^{(0)}_k[i]$ ] approaches to the optimum threshold @xmath240 , which induces poor performance of the detector and therefore increases bayesian risk .
furthermore , we could apply * lemma [ t2 ] * to set the constraints on joint pmf .
since vector norm is a convex function , the problem can be solved by well - developed algorithms in convex optimization @xcite .
in last part , we proposed a simple methodology to select cooperative nodes based on reliability under assumption of independent observations .
however , in practice , there exists correlation among spectrum availability at cooperative nodes and spectrum sensing may change as we showed in * proposition [ p3]*. in addition , since the statistical information is limited within reasonable observation interval , cr - tx can select cooperative nodes to minimize maximum bayesian risk by minimax criterion .
in preceding sections , we only considered single cr link in crn . however , crn is not just a link level technology if we want to successfully route packets from source to destination through crs and ps . in the following , we suggest a simple physical layer model for crn and investigate the impacts of spectrum sensing on network operation and the role of a cooperative node playing in crn , which is impossible to be revealed from traditional treatment of spectrum sensing .
since spectrum sensing may not be ideal and there exists hidden terminal problem , we further define the true state for ps .
the true state for ps can be represented by the indicator @xmath241 therefore , with the definition of @xmath22 , we have @xmath242 to connect relations between indicator functions of link availability and realistic operation of crn , we propose a simple received power model .
we model the received power from ps and background noise as log - normal distribution , or @xmath243 and @xmath244 , where @xmath245 and @xmath246 are used to quantify the measurement uncertainty of the received power from ps and noise respectively . in addition , @xmath247 is a constant whereas @xmath248 should be varied according to path loss and shadowing .
more specifically , let @xmath249 , where @xmath250 is a constant , @xmath251 denotes distance from cr ( either cr - tx or cr - rx ) to ps as in fig [ fig_2 ] , @xmath252 means path loss exponent , and @xmath253 represents shadowing effect . when @xmath254 , the received signal only comes from noise . however , when @xmath255 , the received signal is the superposition of signal from ps and noise , which results in addition of two log - normal random variables .
we could simply model the received power as another log - normal random variable with parameters @xmath256 and @xmath257 , and under assumption of @xmath258 , we have @xmath259 by simulation , the distribution of the simplified model , although not exactly identical to , is close to the simulated distribution , especially in terms of mean and variance .
it justifies our simplified model .
recall the conditions that crs can successfully communicate over a link . assume cr - tx adopts an energy detector in the hypothesis testing ( [ e27 ] ) and there is no interference from co - existing systems .
the detector can be represented as @xmath260 where @xmath261 denotes the received power at cr - tx and @xmath262 is a fixed threshold since the detector is designed under a given sinr . on the other hand , to successfully receive packets , the sinr at cr - rx should be greater than minimum value @xmath10 as shown in ( [ e28 ] ) .
similarly , spectrum availability at cr - rx can be represented as @xmath263 where @xmath264 denotes the received power from ps and noise at cr - rx .
different from cr - tx , @xmath265 is varied according to the received power from cr - tx . for simplicity ,
we only consider propagation loss in modeling the received power from cr - tx and have @xmath266 , where @xmath267 is a constant and @xmath268 denotes distance between cr - tx and cr - rx .
we suppose that the measurement uncertainties and hence the received power from ps and noise at cr - tx and cr - rx are independent .
however , to model spatial behavior for cr - tx and cr - rx , we consider the relation of shadowing between cr - tx and cr - rx . intuitively , the relation should depend on locations of cr - tx , cr - rx , ps , along with the obstacle size and we proceed based on a linear model @xmath269 where @xmath270 denotes parameter of obstacle size and @xmath271 is the angle between line segments with starting point at cr - tx and end points at cr - rx and ps , as shown in fig .
[ fig_2 ] for illustration . in this model ,
shadowing at cr - rx @xmath272 linearly decreases with respect to the distance between cr - tx and cr - rx @xmath268 with rate inverse proportional to the obstacle size @xmath270 and is equal to zero when cr - rx is far apart from cr - tx or ps is located in the middle of cr - tx and cr - rx .
additionally , since shadowing parameter achieves maximum at cr - tx , this results in the worst case scenario in spectrum sensing . finally , from log - normal fading distribution , @xmath273 where @xmath274 denotes the right - tail probability of a gaussian random variable with zero mean and unit variance . under above proposed signal model , the analysis can be easily extended to cooperative spectrum sensing . considering a cooperative node conducting an energy detector , we have @xmath275 furthermore
, the correlation due to geography is established similar to ( [ e29 ] ) .
we could therefore calculate @xmath72 and @xmath276 similar to @xmath22 , as @xmath277 with the relation between statistical information @xmath71 and received power model , we can mathematically determine allowable transmission region of cr - tx . in section iii and iv ,
we have developed spectrum sensing under different assumptions and note that spectrum sensing depends on the value of @xmath22 , i.e. , spatial behavior of cr - tx and cr - rx .
especially , there is even a region of @xmath22 that cr - tx is prohibited from forwarding packets to cr - rx and the link from cr - tx to cr - rx is disconnected ( i.e. @xmath278 ) .
this undesirable phenomenon alters crn topology and heavily affects network performance , such as throughput of crn , etc .
therefore , we would like to theoretically study link properties in crn and first define the regions of @xmath22 as follows .
the set @xmath279 is called * prohibitive region * while @xmath280 is called * admissive region*. the boundary between these two sets is called * critical boundary * of @xmath22 and is denoted by @xmath281 . therefore , @xmath282 and @xmath283 .
if @xmath22 lies in the prohibitive region , the link from cr - tx to cr - rx is disconnected .
the property and the engineering meaning of @xmath281 are addressed as follows .
[ l6 ] @xmath281 is a decreasing function with respect to number of cooperative nodes .
it is easy to show that for fixed @xmath30 , @xmath22 decreases as the threshold of the likelihood ratio test @xmath240 increases .
therefore , @xmath281 can be determine by the largest likelihood ratio .
assume the largest likelihood ratio with @xmath284 cooperative nodes occurs at @xmath285 , i.e. , @xmath286/\mathbf{p}^{(0)}_{k-1}[i]\}$ ] .
when the @xmath138th cooperative node enters , let @xmath287,\ldots , \mathbf{1}^{co}_{k-1}=\mathbf{a}^1_{m , k-1}[k-1,i_{max}],\mathbf{1}^{rx}=1,\mathbf{1}^{tx}=1)\\ \tilde{\gamma}_k&=\pr(\mathbf{1}^{co}_k=0|\mathbf{1}^{co}_1=\mathbf{a}^1_{m , k-1}[1,i_{max}],\ldots , \mathbf{1}^{co}_{k-1}=\mathbf{a}^1_{m , k-1}[k-1,i_{max}],\mathbf{1}^{rx}=0,\mathbf{1}^{tx}=1)\end{aligned}\ ] ] then , there are two likelihood ratio with @xmath138 cooperative nodes , say @xmath135th and @xmath288th , becoming @xmath289}{\mathbf{p}^{(0)}_{k}[i]}&= \frac{\mathbf{p}^{(1)}_{k-1}[i_{max}]\tilde{\beta}_k } { \mathbf{p}^{(0)}_{k-1}[i_{max}](1-\tilde{\gamma}_k)}\\ \frac{\mathbf{p}^{(1)}_{k}[j]}{\mathbf{p}^{(0)}_{k}[j]}&= \frac{\mathbf{p}^{(1)}_{k-1}[i_{max}](1-\tilde{\beta}_k ) } { \mathbf{p}^{(0)}_{k-1}[i_{max}]\tilde{\gamma}_k}\end{aligned}\ ] ] since either @xmath290 or @xmath291 , one of the @xmath135th and the @xmath288th likelihood ratio is not less than @xmath292/\mathbf{p}^{(0)}_{k-1}[i]$ ] , which results in lower @xmath281 .
[ l7 ] the following two statements are equivalent : 1 .
@xmath293 2 .
@xmath294 since @xmath293 if and only if @xmath295 , we have @xmath296 the inequality holds because the likelihood ratio is greater than @xmath240 if @xmath297 and @xmath298 is a increasing function with respect to @xmath299 .
reversely , the conditional probability @xmath300 is well - defined if and only if @xmath295 , which implies @xmath293 . in * lemma [ l7 ] *
, @xmath300 could be interpreted as the probability of successful transmission in crn and the weighting factor in bayesian risk ( [ e30 ] ) can be determined by the constraint on the outage probability @xmath301 .
that is , if a crn maintains @xmath302 , @xmath303 .
therefore , the condition that allows cr - tx forwarding packets to cr - rx ( i.e. @xmath22 belongs to admissive region ) guarantees the outage probability of cr link . further considering the proposed physical layer models , we can establish and define a geographic region , where cr - tx is allowed forwarding packets to cr - rx as long as cr - rx lies in the region .
* neighborhood * of cr - tx @xmath304 is @xmath305 or equivalently becomes @xmath306 .
* coverage * of cr - tx is neighborhood of cr - tx without ps .
please not that the coverage of cr - tx is defined without the existence of ps and the neighborhood is the effective area in real operation coexisting with ps .
when we consider a path loss model between cr - tx and cr - rx , coverage becomes a circularly shaped region .
however , due to hidden terminal problem as in fig .
[ fig_1 ] and [ fig_2 ] , where ps is either apart from cr - tx or is blocked by obstacles , the probability of collision at cr - rx could increase and cr - tx may be prohibited from forwarding packets to cr - rx .
therefore , neighborhood of cr - tx shrinks from its coverage and is no longer circular shape .
in addition , hidden terminal problem is location dependent , that is , ps is hidden to cr - tx but not to cr - rx in fig .
[ fig_1 ] and [ fig_2 ] .
thus , cr - rx is possibly allowed forwarding packets to cr - tx . from such observations ,
cr links are directional and can be mathematically characterized as follows .
@xmath307 is said to be * connective * to @xmath308 if @xmath308 is located in the neighborhood of @xmath307 , which is denoted by @xmath309 .
otherwise , @xmath310 if @xmath307 is not * connective * to @xmath308 .
according to above arguments , it is possible that cr - rx is connective to cr - tx but the reserve is not true .
mathematical conclusion is developed in the following , and is numerically verified in fig .
[ fig_4]and [ fig_5 ] in section vi .
[ p6 ] connective relation is asymmetric , that is , for two cognitive radios , @xmath307 is connective to @xmath308 does not imply @xmath308 is connective to @xmath307 , or mathematically , @xmath311 .
we analytically illustrate using fig .
[ fig_1 ] , where @xmath307 lies in the middle of @xmath308 and ps - tx and ps - tx is hidden to @xmath308 but not to @xmath307 . let @xmath312 to guarantee the outage probability of cr link less than @xmath313 and let @xmath314 , i.e. , the spectrum utility of ps is only @xmath315 .
if @xmath307 wants to forward packets to @xmath308 ( i.e. @xmath307 is cr - tx and @xmath308 is cr - rx ) , @xmath307 can successfully detect the activity of ps and @xmath316 and @xmath317 .
therefore , @xmath307 forwards packets to @xmath308 only when @xmath254 .
in addition , since @xmath307 is located in the transmission range of @xmath308 , @xmath308 is located in the transmission range of @xmath307 in a pure path loss model and @xmath318 . applying ( [ e38 ] ) and ( [ e13 ] )
, we have @xmath319 and @xmath309 .
on the other hand , when @xmath308 wants to forward packets to @xmath307 , @xmath308 becomes cr - tx and @xmath307 becomes cr - rx . since ps - tx is hidden to @xmath308 , at @xmath308 , the received signal power from ps is below noise power , and @xmath320 and @xmath321 in ( [ e42 ] ) and ( [ e43 ] ) .
therefore , @xmath316 and @xmath322 .
that is , @xmath308 always feels the spectrum available and intends to forward packets to @xmath307 . however ,
when @xmath255 , collisions occurs at @xmath307 and @xmath323 .
similarly , by ( [ e38 ] ) and ( [ e13 ] ) , we have @xmath324 and @xmath325 .
* proposition [ p6 ] * mathematically suggests that links in crn are generally asymmetric and even unidirectional as the argument in @xcite .
therefore , traditional feedback mechanism such as acknowledgement and automatic repeat request ( arq ) in data link layer may not be supported in general .
this challenge can be alleviated via cooperative schemes .
roles of a cooperative node in cr network operation thus include 1 .
extend neighborhood of cr - tx to its coverage 2 .
ensure bidirectional links in crn ( i.e. enhance probability to maintain bidirectional ) 3 .
enable feedback mechanism for the purpose of upper layers since neighborhood increases as @xmath281 decreases , by * lemma [ l6 ] * , the capability of cooperative schemes to extend neighborhood increases when number of cooperative nodes increases .
therefore , spectrum sensing capability mathematically determine crn topology .
it also suggests the functionality of cooperative nodes in topology control @xcite@xcite and network routing @xcite , which is critical in crn due to asymmetric links and heterogeneous network architecture @xcite . here , we illustrate impacts of correlation between spectrum availability at cooperative nodes on neighborhood .
recall * proposition [ p3 ] * , where we considered two cooperative nodes with @xmath214 , @xmath326 , and @xmath217 . from ( [ e37 ] ) , we have @xmath327 therefore , as @xmath209 increases from @xmath328 , @xmath329 increases from 0 and achieves maximum at @xmath330 . at this point , @xmath331 , which is the critical boundary with node one alone .
if @xmath209 further increases to @xmath332 , @xmath333 decreases to 0 .
we conclude that positive correlation between @xmath173 and @xmath174 shrinks the neighborhood , compared to the independent case ( @xmath180 ) , unless the correlation is high enough , i.e. , @xmath334 by solving @xmath335 according to ( [ e39 ] ) .
if one cr has larger neighborhood area , it is expected to be connective to more crs and to have higher probability to forward packets successfully and higher throughput of crn accordingly .
the result offers a novel dimension to evaluate cooperative nodes .
that is , different from criterions in link level , such as minimum bayesian risk or maximum reliability as we mentioned in last section , maximum neighborhood area is a novel criterion to select the best cooperative node from the viewpoint of network operation .
[ p5 ] ( * optimum selection of cooperative node * ) for a crn with a constraint on the outage probability @xmath302 , there are one cr and @xmath158 cooperative nodes , indexed by @xmath138 .
the best cooperative node for the cr under maximum neighborhood area criterion is @xmath336 where @xmath337 represents neighborhood area of the cr with the aid of the @xmath138th cooperative node . in crn
, crs could act as relay nodes to relay packets to the destination . assume the destination is in the direction @xmath338 of a cr with respect to ps .
it is intuitive for the cr to forward packets to the direction around @xmath338 , say @xmath339 .
let @xmath340 and then the best cooperative node may become ( [ e31 ] ) with @xmath304 replaced by @xmath341 .
the performance of spectrum sensing , measured by bayesian risk ( [ e30 ] ) , is plotted by bayesian risk versus the probability of spectrum availability at cr - rx @xmath22 in fig . [ fig_3 ] .
we set the weighting factor @xmath312 ( @xmath30 is defined in ( [ e41 ] ) ) to guarantee the outage probability of cr link less than @xmath313 .
larger bayerian risk represents worse performance because spectrum sensing induces more possibility of collisions at cr - rx or of losing opportunity to utilize spectrum .
we see that traditional spectrum sensing without considering spectrum availability at cr - rx ( i.e. @xmath18 ) has large bayesian risk when @xmath22 becomes small because collisions usually occur when cr - tx determines link availability only by localized spectrum availability at cr - tx ( i.e. @xmath49 ) . on the other hand , by considering @xmath18 in our general spectrum sensing ( [ e13 ] ) with known @xmath22 , bayesian risk decreases when @xmath22 is less than @xmath342 , which is the critical boundary of @xmath22 , @xmath281 , and cr - tx is prohibited from forwarding packets to cr - rx .
therefore , risk occurs due to loss of opportunity to utilize spectrum .
however , in practice , @xmath22 is unknown and needs to be estimated by * lemma [ l8]*. we set observation depth ( i.e. duration ) @xmath343 and show expected bayesian risk of inference - based spectrum sensing ( [ e13 ] ) with respect to observed sequence @xmath32,\ldots,\mathbf{1}^{rx}[n - l]$ ] .
the performance degrades around @xmath281 and even worse than that of traditional spectrum sensing when @xmath293 because the estimation error may cause the estimated @xmath22 ( [ e1 ] ) to across @xmath281 and results in different sensing rules ; however , it is close to the performance with known @xmath22 .
this verifies the effectiveness of inference - based spectrum sensing .
[ fig_3 ] also shows bayesian risk of cooperative sensing ( [ e15 ] ) under different sensing capability of a cooperative node , i.e. reliability @xmath192 .
we assume statistical information @xmath71 can be perfectly estimated .
the performance curve is composed of three line segments as in ( [ e15 ] ) and shows performance improvement in the middle segment due to the aid of the cooperative node .
however , in the right and the left segments , the cooperative node becomes useless and the performance is equal to that of non - cooperative sensing . in comparison of sensing capability of a cooperative node
, the one with larger reliability is expected to have higher correlation between spectrum availability at cr - rx and the cooperative node ( i.e. @xmath48 ) and to provide more information about @xmath18 ; therefore , it achieves lower bayesian risk and lower @xmath281 ( i.e. larger admissive region ) .
in addition , when @xmath57 ( @xmath344 ) , @xmath18 and @xmath48 are independent and the performance is identical to that of spectrum sensing with known @xmath22 in fig .
[ fig_3 ] .
we next investigate performance of spectrum sensing with two cooperative nodes with respect to the correlation between spectrum availability at these two nodes ( i.e. @xmath173 and @xmath174 ) @xmath176 .
we set @xmath345 , @xmath346 as the scenario in * proposition [ p3 ] * and depict bayesian risk in fig .
[ fig_6 ] .
generally speaking , bayesian risk decreases ( increases ) as @xmath176 increases in high ( low ) @xmath22 .
we also observe that @xmath281 decreases when number of cooperative nodes increases and @xmath281 increases when @xmath176 increases unless the correlation is high enough .
for example , if two cooperative nodes are close in location and @xmath347 , @xmath281 is less than that when @xmath180 .
it is also interesting to note that there exists a region of @xmath22 such that the identity of @xmath173 and @xmath174 instead of @xmath173 determines cr link availability as in ( [ e40 ] ) and bayesian risk is less than that with @xmath173 alone .
the results further suggest trade - off between performance in link layer and network layer when we select cooperative nodes .
that is , for one cr link with @xmath348 ( e.g. cr - rx is close to cr - tx or spectrum utilization of ps is low ) , large @xmath176 is preferred to achieve low risk .
however , from network perspective , to achieve high number of cr links that are admissive to cr - tx ( i.e. to achieve large neighborhood and low @xmath281 ) and thus high throughput of crn , small @xmath176 is preferred . for multiple cooperative nodes , with six nodes in our simulation ,
we show the performance of cooperative spectrum sensing with limited statistical information @xmath234 due to limited sensing duration .
we first find least - favorable joint pmf @xmath349 by ( [ e35 ] ) and then compute corresponding bayesian risk , which is shown in fig .
[ fig_7 ] under different order of known marginal pmf @xmath138 ( i.e. capability of observation ) .
that is , cr - tx only acquires pmf of @xmath138 out of six cooperative nodes .
the risk is compared to that with the optimum sensing rule ( [ e14 ] ) and that with assumption of independent spectrum availability ( [ e34 ] ) .
obviously , bayesian risk decreases and approaches to that in the optimum case as the order of known marginal pmf @xmath138 increases because more information is acquired to generate @xmath349 closer to the true one @xmath350 .
we observe that when the order @xmath138 is greater than 3 , robust spectrum sensing outperforms the case of traditional independence assumption
. therefore , if cr - tx would like to select six cooperative nodes , cr - tx only requires statistical information about spectrum availability among three out of six cooperative nodes , i.e. @xmath351 , to achieve better performance than the case according to reliability criterion . in fig .
[ fig_4 ] , we illustrate neighborhood of cr - tx ( `` @xmath352 '' in the figure ) without blocking .
the neighborhood boundary with / without a cooperative node ( `` @xmath353 '' in the figure ) is depicted by a thick and a thin line respectively .
the parameters are set as follows : @xmath354 , @xmath355 , @xmath356 , @xmath357 , @xmath358 , @xmath359 , @xmath360 , and @xmath361 . in fig .
[ fig_4](a ) , ps ( `` @xmath362 '' in the figure ) is placed near to cr - tx ( @xmath363 ) .
we observe that cr - tx almost perfectly detects the state of ps , i.e. , @xmath316 and @xmath317 , and @xmath364 by ( [ e38 ] ) .
therefore , neighborhood of cr - tx approaches to its coverage and the cooperative node is not necessary in this case from a viewpoint of network operation .
however , when ps is apart from cr - tx ( @xmath365 ) as we shown in fig .
[ fig_1 ] , the neighborhood at ps side shrinks and is no longer circularly shaped because ps is hidden to cr - tx and hence probability of collision at cr - rx increases when @xmath255 .
[ fig_4](b)@xmath366(d ) illustrate the neighborhood under different locations of the cooperative node .
we observe that neighborhood area decreases when the cooperative node moves away from ps and there even exists a region where cooperative sensing can not help .
therefore , the cooperative node in fig .
[ fig_4](b ) is the best among these three nodes according to maximum neighborhood area criterion in * proposition [ p5]*. we present an example of existence of unidirectional link in crn . in fig .
[ fig_4](b ) , assume one cr is located at @xmath367 .
obviously , the cr - tx is not connective to the cr and therefore is prohibited from forwarding packets to the cr .
however , by fig .
[ fig_4](a ) , the cr is connective to cr - tx , which makes the link unidirectional ( only from the cr to cr - tx ) . as * proposition [ p6 ] * , this also shows asymmetric connective relation even under rather ideal radio propagation . with the aid of a cooperative node located at @xmath368 , the link returns to a bidirectional link .
alternatively , we consider effects of shadowing due to blocking , as we illustrated in fig .
[ fig_2 ] .
we set shadowing parameters @xmath369 and parameter of obstacle size @xmath370 and @xmath371 in fig .
[ fig_5](a)(b ) and fig .
[ fig_5](c)(d ) respectively .
we observe that small obstacles size ( i.e. small @xmath270 ) can result in more substantial shrink of the neighborhood , compared to large obstacles size ( i.e. large @xmath270 ) .
the reason is : if @xmath270 is small , only a small region around cr - tx falls in deep shadowing and the state of ps can be successfully detected outside that region .
therefore , this leads to high probability of collision at cr - rx as @xmath255 . on the other hand ,
if @xmath270 is large , crs are likely separated from ps by obstacles , which results in large `` distance '' between cr and ps . here , `` distance '' is measured by received signal power @xcite@xcite . in comparison of the capability of a cooperative node , the one in small @xmath270 has good capability to recover the neighborhood to its coverage even when the node is at opposite side of ps .
however , for large @xmath270 , the cooperative node may also be in deep shadowing and becomes useless to recover neighborhood of cr - tx .
in this paper , we showed that cr link availability should be determined by spectrum availability at both cr - tx and cr - rx , which may not be identical due to hidden terminal problem ( fig .
[ fig_1 ] and [ fig_2 ] ) . in order to fundamentally explore the spectrum sensing at link level and its impacts on network operation , we established an indicator model of cr link availability and applied statistical inference to predict / estimate unknown spectrum availability at cr - tx due to no centralized coordinator nor information exchange between cr - tx and cr - rx in advance .
we therefore expressed conditions for cr - tx to forward packets to cr - rx under guaranteed outage probability . these conditions , along with physical channel models ,
define neighborhood of cr - tx , which is no longer circularly shaped as coverage .
this results in asymmetric or even unidirectional links in crn , as we illustrated in section vi .
the impairment of cr links can be alleviated via cooperative scheme .
therefore , spectrum sensing capability determines network topology and thus throughput of crn .
several factors with impacts on spectrum sensing are analyzed , including : 1 .
correlation of spectrum availability at cooperative nodes 2 .
capability of observation at cr - tx ( i.e. available statistical information at cr - rx ) 3
. locations of cooperative nodes and environment ( i.e. obstacles ) furthermore , limits of cooperative scheme were also addressed in link level and network level .
in addition , to measure sensing capability and then to select cooperative nodes is an important issue because we would like to minimize information exchange to increase spectrum utilization .
criterions from link level ( maximum reliability or minimum bayesian risk ) and network level ( maximum neighborhood area ) perspectives were accordingly proposed .
we numerically demonstrated existence of trade - off in designing systems in different layers .
in addition , robust spectrum sensing was proposed to deal with local and partial information due to no centralized coordination and limited sensing duration in crn .
more useful results in crn extended from this research can be expected in future works .
k. kim , i. a. akbar , k. k. bae , j - s um , c. m. spooner , and j. h. reed , `` cyclostationary approaches to signal detection and classification in cognitive radio , '' _ ieee proc . dyspan _ , pp .
212 - 215 , apr . 2007 .
z. quan , s. cui , and a. h. sayed , `` optimal linear cooperation for spectrum sensing in cognitive radio networks , '' _ ieee journal on selected topics in signal processing _ ,
28 - 40 , feb . 2008 .
z. quan , s. cui , a. h. sayed , and h. v. poor , `` optimal multiband joint detection for spectrum sensing in cognitive radio networks , '' _ ieee transactions on signal processing _ , vol .
1128 - 1140 , mar .
2009 .
s. a. jafar and s. srinivasa , `` capacity limits of cognitive radio with distributed and dynamics spectral activity , '' _ ieee journal on selected areas in communications _ , vol .
529 - 537 , apr .
2007 .
k. c. chen , l. h. kung , d. shiung , r. prasad , and s. chen , `` self - organizing terminal architecture for cognitive radio networks , '' _ the 10th international symposium on wireless personal multimedia communications _ , pp .
926 - 931 , dec . 2007 .
k. c. chen , y. c. peng , b. k. centin , n. prasad , j. wang , and s. y. lee , `` routing of opportunistic links in cognitive radio networks , '' to appear in the _ wiley wireless communicatioins and mobile computing_. , that is , pmf of @xmath138 out of six cooperative nodes is available at cr - tx . the performance is compared to the spectrum sensing with full statistical information and with traditional independence assumption . ] | it is well known that the successful operation of cognitive radio ( cr ) between cr transmitter and cr receiver ( cr link ) relies on reliable spectrum sensing . to network crs requires more information from spectrum sensing beyond traditional techniques , executing at cr transmitter and further information regarding the spectrum availability at cr receiver . redefining the spectrum sensing along with statistical inference suitable for cognitive radio networks ( crn ) , we mathematically derive conditions to allow cr transmitter forwarding packets to cr receiver under guaranteed outage probability , and prove that the correlation of localized spectrum availability between a cooperative node and cr receiver determines effectiveness of the cooperative scheme . applying our novel mathematical model to potential hidden terminals in crn ,
we illustrate that the allowable transmission region of a cr , defined as neighborhood , is no longer circular shape even in a pure path loss channel model .
this results in asymmetric cr links to make bidirectional links generally inappropriate in crn , though this challenge can be alleviated with the aid of cooperative sensing . therefore , spectrum sensing capability determines crn topology . for multiple cooperative nodes , to fully utilize spectrum availability , the selection methodology of cooperative nodes is developed due to limited overhead of information exchange . defining reliability as information of spectrum availability at cr receiver provided by a cooperative node and by applying neighborhood area , we can compare sensing capability of cooperative nodes from both link and network perspectives .
in addition , due to dynamic network topology lack of centralized coordination in crn , crs can only acquire local and partial information in limited sensing duration , robust spectrum sensing is therefore proposed to ensure successful crn operation .
limits of cooperative schemes and their impacts on network operation are also derived .
spectrum sensing , cognitive radio networks , link availability , network tomography , statistical inference , reliability , neighborhood . |
let @xmath2 be a bounded symmetric domain in @xmath3 with bergman kernel @xmath4 we assume that @xmath2 is in its standard representation and the volume measure @xmath5 of @xmath2 is normalised so that @xmath6 for all @xmath7 and @xmath8 in @xmath9 by theorem 5.7 of @xcite and using the polar coordinates representation , there exists a positive number @xmath10 such that for @xmath11 we have @xmath12 for each @xmath13 define @xmath14 then @xmath15 defines a weighted family of probability measures on @xmath9 also , throughout the paper @xmath16 is fixed .
we define the weighted bergman spaces @xmath17 on @xmath2 , as the set of all holomorphic functions @xmath18 on @xmath2 so that @xmath19 note that @xmath20 is a closed subspace of @xmath21 for @xmath22 is just the usual bergman space . for @xmath23
there is an orthogonal projection @xmath24 from @xmath25 onto @xmath26 given by @xmath27 where @xmath28 is the reproducing kernel for @xmath29 suppose , @xmath30 are holomorphic mappings defined on @xmath31 such that @xmath32 then the weighted composition operator @xmath0 is defined as @xmath33 for the study of weighted composition operators one can refer to @xcite and references therein .
recently , smith @xcite has made a nice connection between the brennan s conjecture and weighted composition operators .
he has shown that brennan s conjecture is equivalent to the existence of self - maps of unit disk that make certain weighted composition operators compact .
we know that the bounded symmetric domain @xmath2 in its standard representation with normalised volume measure , the kernel function @xmath34 has the following specialproperties : 1 .
@xmath35 2 .
@xmath36 ( for all @xmath7 in @xmath2 and @xmath37 in @xmath38 ) .
3 . @xmath39 4 .
@xmath40 is a smooth function on @xmath41 here @xmath42 denotes the closure of @xmath2 in @xmath3 and @xmath43 is the topological boundary , see @xcite and @xcite for details . for any @xmath37 in @xmath2 ,
let @xmath44 the @xmath45 are called normalised reproducing kernels for @xmath46 they are unit vectors in @xmath47 moreover , @xmath48 is a unit vector of @xmath26 for any @xmath37 in @xmath49 let @xmath50 be a finite complex borel measure on @xmath49 then the berezin transform of measure @xmath51 denoted by @xmath52 is defined as @xmath53 we will denote by @xmath54 the bergman distance function on @xmath9 for @xmath55 and @xmath56 let @xmath57 we denote by @xmath58 the normalised volume of @xmath59 that is , @xmath60 given a finite complex borel measure @xmath50 on @xmath61 we define a function @xmath62 on @xmath63 by @xmath64 take @xmath65 let @xmath50 be a finite positive borel measure on @xmath9 then the following conditions are equivalent : 1 . the inclusion map @xmath66 is bounded .
the berezin transform @xmath67 is bounded .
@xmath62 is bounded on @xmath2 for all ( or some ) @xmath68 4 .
@xmath69 is a bounded sequence , where @xmath70 is some bounded sequence in @xmath63 independent of @xmath71 a positive borel measure @xmath50 satisfying any one of the conditions of theorem 1.1 is called a carleson measure on the weighted bergman space @xmath72 take @xmath65 let @xmath50 be a finite positive borel measure on @xmath9 then the following conditions are equivalent : 1 .
the inclusion map @xmath73 is compact .
2 . the berezin transform @xmath74 as @xmath75 3 .
@xmath76 as @xmath77 for all ( or some ) @xmath78 4 .
@xmath79 as @xmath80 a positive borel measure @xmath50 satisfying any one of the conditions of theorem 1.2 is called a vanishing carleson measure on the weighted bergman space @xmath81 a positive compact operator @xmath82 on @xmath83 is in the trace class if @xmath84 for some ( or all ) orthonormal basis @xmath85 of @xmath86 take @xmath87 and @xmath82 be a compact linear operator on @xmath88 then we say that @xmath82 belongs to the schatten @xmath1-class @xmath89 if @xmath90 is in the trace class .
also , the @xmath91 norm of @xmath82 is given as @xmath92^{{1}/{p}}.\ ] ] moreover , @xmath93 is a two - sided ideal of the full algebra @xmath94 of bounded linear operators on @xmath86
to make the paper self - contained we state the following lemmas .
-.3pc@xcite .
the bergman kernel @xmath95 is conjugate symmetric .
that is , @xmath96 and it is the function of @xmath7 in @xmath97 -.3pc@xcite .
let @xmath82 be a trace class operator or a positive operator on @xmath98 then @xmath99 where @xmath100 -.3pc@xcite .
let @xmath2 be a bounded symmetric domain in @xmath3 and let @xmath0 be a weighted composition operator on @xmath88 then @xmath101^{\frac{(1 - \lambda)}{2}}.\ ] ] -.3pc@xcite .
suppose @xmath82 is a positive operator on a hilbert space @xmath102 and @xmath103 is a unit vector in @xmath104
. then 1 .
@xmath105 2 .
@xmath106 -.3pc(_theorem _
18.11(_f _ ) , _ p_. 89 _ of _ @xcite ) .
suppose @xmath82 is a bounded linear operator on a hilbert space @xmath104 . if @xmath107 then @xmath108 -.3pc@xcite .
take @xmath65 suppose @xmath109 is a finite borel measure on @xmath9 then @xmath50 is a carleson measure on @xmath20 if and only if @xmath110 is bounded on @xmath111 ( as a function of @xmath7 ) for all ( or some ) @xmath78 moreover , the following quantities are equivalent for any fixed @xmath112 and @xmath113 : @xmath114 and @xmath115 -.3pc@xcite . for any @xmath116 there
exists a constant @xmath117 ( depending only on @xmath118 ) such that @xmath119 for all @xmath120 and @xmath121 note that if we take @xmath122 in the above estimate , then we get @xmath123 for all @xmath124 -.3pc@xcite . for any @xmath125 there
exists @xmath126 constant @xmath127 ( depending on @xmath128 ) such that @xmath129 for all @xmath130 in @xmath2 with @xmath131 -.3pc@xcite . for any @xmath132 there
exists a sequence @xmath133 in @xmath111 satisfying the following two conditions : 1 .
@xmath134 2 .
there is a positive integer @xmath135 such that each point @xmath136 in @xmath2 belongs to at most @xmath135 of the sets @xmath137 -.3pc@xcite . for any @xmath112 and @xmath138 there
exists a constant @xmath127 ( depending only on @xmath118 ) such that @xmath139 for all @xmath18 holomorphic and @xmath37 in @xmath49
by using @xcite , we can prove the following lemma .
take @xmath65 let @xmath2 be a bounded symmetric domain in @xmath140 and @xmath141 suppose @xmath30 be holomorphic functions defined on @xmath2 such that @xmath142 also , suppose that the weighted composition operator @xmath0 is bounded on @xmath143 .
then @xmath144 is compact on @xmath145 if and only if for any norm bounded sequence @xmath146 in @xmath20 such that @xmath147 uniformly on compact subsets of @xmath61 then we have@xmath148 suppose @xmath146 is a sequence in @xmath145 such that @xmath147 weakly in @xmath149 then @xmath146 is a norm bounded sequence and @xmath147 uniformly on each compact subsets of @xmath49 let @xmath150 be a holomorphic function defined on @xmath2 such that @xmath142 suppose @xmath151 and @xmath152 then the nonnegative measure @xmath153 is defined as @xmath154 where @xmath155 is a measurable subset of @xmath9 using lemma 2.1 of @xcite , we can prove the following result .
take @xmath65 let @xmath2 be a bounded symmetric domain in @xmath140 and @xmath141 let @xmath150 be a holomorphic mapping from @xmath63 into @xmath2 and @xmath156 then @xmath157 where @xmath158 is an arbitrary measurable positive function on @xmath9 take @xmath65 let @xmath2 be a bounded symmetric domain in @xmath140 and @xmath159 suppose @xmath30 are holomorphic functions defined on @xmath2 such that @xmath142 then the weighted composition operator @xmath0 is bounded on @xmath20 if and only if the measure @xmath153 is a carleson measure .
suppose @xmath0 is bounded on @xmath149 then there exists a constant @xmath160 such that @xmath161 for all @xmath162 take @xmath163 from lemma 2.1 and using the fact that @xmath164 for any @xmath7 and @xmath8 in @xmath165 we get that @xmath166 thus , we have @xmath167 & \le \int_{\omega}|\psi(\omega ) |^{p}|g(\varphi(\omega ) ) |^{p } { \rm d}v_{\lambda}(\omega)\\[.3pc ] & = \|w_{\varphi,\psi}g\|_{a^{p}_{\lambda}}^{p } < \infty.\end{aligned}\ ] ] by lemma 2.7 , there exists a constant @xmath168 ( depending only on @xmath118 ) such that @xmath169 & \le c \;\;\int_{e(z , r)}|k_{z}(\omega ) |^{2(1 - \lambda)}{\rm d}\mu_{\varphi,\psi , p}(\omega ) < \infty,\end{aligned}\ ] ] for any @xmath170 thus , by using lemma 2.6 , we get that @xmath171 is a carleson measure on @xmath149 conversely , suppose @xmath153 is a carleson measure on @xmath149 then by theorem 1.1 , there exists a constant @xmath160 such that @xmath172 for all @xmath173 hence , using lemma 3.3 , we have @xmath174 & = \int_{\omega}| f(\omega ) |^{p } { \rm d}\mu_{\varphi,\psi , p}(\omega)\\[.4pc ] & \le m^{p } \int_{\omega}| f(\omega)|^{p } { \rm d}v_{\lambda}(\omega ) < \infty,\end{aligned}\ ] ] for all @xmath162 hence , @xmath0 is a bounded operator on @xmath149 take @xmath65 let @xmath2 be a bounded symmetric domain in @xmath140 and @xmath159 suppose @xmath30 are holomorphic functions defined on @xmath2 such that @xmath142 then the weighted composition operator @xmath0 is compact on @xmath20 if and only if the measure @xmath153 is a vanishing carleson measure .
suppose @xmath0 is compact on @xmath149 since , @xmath175 weakly in @xmath145 as @xmath77 , by lemmas 3.1 and 3.2 , we have @xmath176 again , by using lemma 2.7 , we get a constant @xmath177 ( depending only on @xmath118 ) such that @xmath178 & \le c \int_{e(z , r)}|k_{z}(\omega ) |^{2(1 - \lambda)}{\rm d}\mu_{\varphi,\psi , p}(\omega)\\[.4pc ] & \le c \int_{e(z , r)}|\psi(\omega ) |^{2}|k_{z}(\varphi(\omega ) ) |^{2(1 - \lambda)}{\rm d}\mu_{\varphi,\psi , p}(\omega ) \\[.4pc ] & \le c \|w_{\varphi,\psi}k_{z}^{1 - \lambda}\|_{a^{2}_{\lambda}}.\end{aligned}\ ] ] thus , condition ( 3.1 ) implies that @xmath171 is a vanishing carleson measure on @xmath81 conversely , suppose that @xmath153 is a vanishing carleson measure on @xmath149 then , by theorem 1.2 , we have @xmath179 for all ( or some ) @xmath68 so , from theorem 3.4 , we conclude that @xmath0 is bounded on @xmath149 we will prove that @xmath0 is a compact operator .
let @xmath146 be a sequence in @xmath180 and @xmath147 weakly .
by lemma 3.2 , @xmath181 is a norm bounded sequence and @xmath182 uniformly on each compact subsets of @xmath9 by lemmas 2.7 and 2.10 , there exists a constant @xmath177 ( depending only on @xmath118 ) such that @xmath183 & = \frac{c c_{\lambda}}{| e(z , r)| } \int_{e(z , r)}|f(\omega ) |^{p } \frac{{\rm d}v_{\lambda}(\omega)}{k(\omega,\omega)^{\lambda}}\\[.4pc ] & \le \frac{c'}{| e(z , r ) | } \int_{e(z , r)}|f(\omega ) |^{p } \frac{{\rm d}v_{\lambda}(\omega)}{| e(\omega , r ) |^{- \lambda}}\\[.4pc ] & \le c '' \frac{1}{| e(z , r ) |^{1 - \lambda } } \int_{e(z , r)}|f(\omega ) |^{p } { \rm d}v_{\lambda}(\omega),\end{aligned}\ ] ] for all @xmath184 where the last inequality is obtained by using lemma 2.8 . by combining the above inequality with lemma 2.8
, we get @xmath185 where @xmath186 and @xmath187 are constants depending only on @xmath188 and @xmath118 .
hence , we have @xmath189 & = \int_{\omega } |f_{n}(\omega ) |^{p } { \rm d}\mu_{\varphi,\psi , p } \\[.4pc ] & \le \sum_{i = 1}^{\infty } \int_{e(a_{i},r ) } |f_{n}(\omega ) |^{p } { \rm d}\mu_{\varphi,\psi , p } \\[.4pc ] & \le \sum_{i = 1}^{\infty } \mu_{\varphi,\psi , p}((e(a_{i},r ) ) \sup \{| f_{n}(\omega ) |^{p}{\rm : } \
\omega \in e(a_{i},r)\ } \\[.4pc ] & \le c \ ; \sum_{i = 1}^{\infty}\frac{\mu_{\varphi,\psi , p}((e(a_{i},r))}{| e(a_{i},r ) d}v_{\lambda}(\omega),\end{aligned}\ ] ] where the sequence @xmath70 is the same as in lemma 2.9 . from condition
( 3.2 ) , for every @xmath190 there exists a @xmath191 such that @xmath192 whenever @xmath193 take @xmath194 then , @xmath195 is a compact subset of @xmath2 ( as @xmath196 thus , we have @xmath197 \hskip -4pc & \le c \ ; \sum_{i = 1}^{\infty}\frac{\mu_{\varphi,\psi , p}((e(a_{i},r))}{| e(a_{i},r ) |^{1 - \lambda } } \int_{e(a_{i},2r)}|f_{n}(\omega ) |^{p } { \rm d}v_{\lambda}(\omega)\\[.4pc ] \hskip -4pc & \le c \ ; \sum_{d(a , \partial \omega ) \ge \delta } \frac{\mu_{\varphi,\psi , p}((e(a_{i},r))}{| e(a_{i},r ) |^{1 - \lambda } } \int_{e(a_{i},2r)}|f_{n}(\omega ) |^{p } { \rm d}v_{\lambda}(\omega)\\[.4pc ] \hskip -4pc & \quad\ , + c \ ;
\sum_{d(a , \partial \omega ) < \delta}\frac{\mu_{\varphi,\psi , p}((e(a_{i},r))}{| e(a_{i},r ) |^{1 - \lambda } } \int_{e(a_{i},2r)}|f_{n}(\omega ) |^{p } { \rm d}v_{\lambda}(\omega)\ ] ] @xmath198 \hskip -4pc & \quad\ , + c \epsilon \ ; \sum_{d(a , \partial \omega ) < \delta } \int_{e(a_{i},2r)}|f_{n}(\omega ) |^{p } { \rm d}v_{\lambda}(\omega)\\[.4pc ] \hskip -4pc & \le \textit{cnk } \ ; \int_{\omega}|f_{n}(\omega ) |^{p } { \rm d}v_{\lambda}(\omega ) + c \epsilon k \ ; \int_{\omega}|f_{n}(\omega ) |^{p } { \rm d}v_{\lambda}(\omega),\end{aligned}\ ] ] where the constant @xmath135 is the same as in lemma 2.9 . since , @xmath181 is a norm bounded sequence and @xmath181 converges to zero uniformly on compact subsets of @xmath165 we have @xmath199 as @xmath200 the following theorem follows by combining theorems 1.1 , 1.2 , 3.4 and 3.5 .
take @xmath201 and let @xmath159 suppose @xmath30 be holomorphic functions defined on @xmath2 such that @xmath142 then 1 .
the weighted composition operator @xmath0 is bounded on @xmath20 if and only if latexmath:[\[\hskip -1.25pc
\sup_{z \in \omega } \int_{\omega}|k_{z}(\omega ) 2 .
the weighted composition operator @xmath0 is compact on @xmath20 if and only if @xmath203
take @xmath201 and let @xmath159 suppose @xmath30 are holomorphic functions defined on @xmath2 such that @xmath142 let @xmath0 be compact on @xmath204 then @xmath205 if and only if the berezin symbol of measure @xmath153 belongs to @xmath206 where @xmath207 for any @xmath208 we have @xmath209 & = \int_{\omega } f(\varphi(z ) ) \overline{g(\varphi(z))}|\psi(z ) |^{2 } { \rm d}v_{\lambda}(z ) \\[.4pc ] & = \int_{\omega } f(\omega ) \overline{g(\omega)}|\psi(z ) |^{2 } { \rm d}\mu_{\varphi,\psi , p}(\omega).\end{aligned}\ ] ] consider the toeplitz operator @xmath210 then , by using fubini s theorem , we have @xmath211 & = \int_{\omega } f(\omega ) \overline{\int_{\omega } g(z ) k_{\lambda}(\omega , z ) { \rm d}v_{\lambda}(z)}\ ; { \rm d}\mu_{\varphi,\psi , p}(\omega ) \\[.4pc ] & = \int_{\omega } f(\omega ) \overline{g(\omega ) } \ ; { \rm d}\mu_{\varphi,\psi , p}(\omega).\end{aligned}\ ] ] thus , @xmath212 also , by definition an operator @xmath213 if and only if @xmath214 is in the trace class , and this is equivalent to saying that @xmath215 ( lemma 1.4.6 , p. 18
of @xcite ) .
thus @xmath205 if and only if @xmath216 again , by theorem c of @xcite , @xmath217 if and only if the berezin symbol of the measure @xmath153 belongs to @xmath218 by using lemmas 2.22.5 , we get @xmath223 & = \int_{\omega } \langle ( w_{\varphi,\psi}w_{\varphi,\psi}^{*})^{{p}/{2 } } k_{z}^{1 - \lambda } , k_{z}^{1 - \lambda } \rangle_{a^{2}_{\lambda } } { \rm d}\sigma(z ) \\[.4pc ] & \ge \int_{\omega } \langle ( w_{\varphi,\psi}w_{\varphi,\psi}^ { * } ) k_{z}^{1 - \lambda } , k_{z}^{1 - \lambda } \rangle_{a^{2}_{\lambda}}^{{p}/{2 } } { \rm d}\sigma(z ) \\[.4pc ] & = \int_{\omega}\|w_{\varphi,\psi}^ { * } k_{z}^{1 - \lambda}\|_{a^{2}_{\lambda}}^{{p } } { \rm d}\sigma(z)\\[.4pc ] & = \int_{\omega}|\psi(z)|\left[\frac{k(\varphi(z),\varphi(z))}{k(z , z ) } \right]^{\frac{p(1 - \lambda)}{2 } } { \rm d}\sigma(z).\end{aligned}\ ] ] take @xmath224 and let @xmath159 take @xmath0 the weighted composition operator @xmath0 on @xmath204 if @xmath221^{\frac{p(1 - \lambda)}{2 } } { \rm d}\sigma(z ) < \infty,\ ] ] where @xmath225 then @xmath226 by using lemmas 2.22.5 , we get @xmath227 & \ge \int_{\omega } \langle ( w_{\varphi,\psi}w_{\varphi,\psi}^ { * } ) k_{z}^{1 - \lambda } , k_{z}^{1 - \lambda } \rangle_{a^{2}_{\lambda}}^{{p}/{2 } } { \rm d}\sigma(z ) \\[.6pc ] & = \int_{\omega}\|w_{\varphi,\psi}^ { * } k_{z}^{1 - \lambda}\|_{a^{2}_{\lambda}}^{{p } } { \rm d}\sigma(z)\\[.6pc ] & = \int_{\omega}|\psi(z)|\left[\frac{k(\varphi(z),\varphi(z))}{k(z , z ) } \right]^{\frac{p(1 - \lambda)}{2 } } { \rm d}\sigma(z ) < \infty.\end{aligned}\ ] ] thus , @xmath228 | in this paper , we study the weighted compositon operators on weighted bergman spaces of bounded symmetric domains .
the necessary and sufficient conditions for a weighted composition operator @xmath0 to be bounded and compact are studied by using the carleson measure techniques . in the last section
, we study the schatten @xmath1-class weighted composition operators .
[ theore]*theorem * [ theore]lemma |
complex systems science ( css ) studies the behavior of a wide range of phenomena , from physics to social sciences passing through biology just to mention few of them .
the classical approach followed in the css consists first in a decomposition of the system into elementary blocksthat will be successively individually analyzed in details , then the properties determined at micro level are transported to the macro level .
this approach results very fruitful and shaped the css as an highly multidisciplinary field .
recently models of opinion dynamics gathered a considerable amount of interest testified by the production of specialized reviews such as , reinforcing in this way the emergence of the _ sociophysics _
a basic distinction can be done in model of _ continuous opinion with threshold _
@xcite , where opinions can assumed to be well described as continuous quantities ; thus agents update their values because of binary interactions , if their opinions are close enough , i.e. below a given threshold .
the second class consists of models where opinions can be described by _ discrete _
variable , yes / no for instance , and they are updated according to _ local rules , i.e. small group interactions _ as for instance : majority rule @xcite , majority and inflexible rules @xcite , majority and contrarian rules @xcite . in this paper
we introduce a new model of opinion dynamics which naturally sets at the intersection of the former scheme ; in fact individuals have continuous opinions that are updated if they are below some given threshold , once agents belong to a group , whose size evolves dynamically together with the opinion .
moreover each agent possesses an _ affinity _ with respect to any other agent , the higher is the affinity score the more trustable is the relationship .
such affinity evolves in time because of the past interactions , hence the acquaintances group is determined by the underlying evolving social network .
we hereby provide an application of the model to the study of the consensus
polarization transition that can occur in real population when people do agree on the same idea consensus state or they divide into several opinion groups polarization state . the paper is organized as follows . in the next section
we introduce the model and the basic involved parameters , then we present the results concerning the consensus polarization issue and the time needed to reach such asymptotic state .
we will end with some conclusions and perspectives .
the model hereby studied is a generalization of the one proposed in @xcite because now interactions occur in many
agents groups , whose size is not fixed a priori but evolves in time . we are thus considering a fixed population made of @xmath0 agents , i.e. closed group setting , where each agent is characterized by its opinion on a given subject , here represented by a real number @xmath1 $ ] , and moreover each agent possesses an affinity with respect to any other , @xmath2 $ ] : the higher is @xmath3 the more affine , say trustable , the relationships are and consequently agents behave . at each time step a first agent , say _ i _
, is randomly drawn with an uniform probability , from the population ; then , in order to determine its acquaintances group , it computes its _ social distance _ with respect to the whole population : @xmath4 the agents _ j _ whose distance from _
i _ is lesser than a given threshold , @xmath5 , will determine the acquaintances group of _ i _ at time @xmath6 in formula : @xmath7 let us observe that the group changes in time , in size and in composition , because the opinions and/or affinities also evolve .
the rationale in the use of the affinities in the definition of the social metric is to interpret @xcite the affinity as the adjacency matrix of the ( weighted ) social network underlying the population .
we hereby assume a constant threshold @xmath5 for the whole population , but of course one could consider non homogeneous cases as well .
once the agent _ i _ has been selected and the group @xmath8 has been formed , the involved individuals do interact by possibly updating their opinions and/or affinities .
once in the group , all agents are supposed to listen to and speak to all other agents , therefore every one can perceive a personal _ averaged _ by the mutual affinity
_ group opinion _ , @xmath9
, because each agent weights differently opinions of trustable agents from the others . in formula : @xmath10 where we denoted by @xmath11 the size of the subgroup @xmath12 .
the vector @xmath13 , will hereby named _
apopsicentre _ ,
i.e. the barycentre of the opinions ( @xmath14 = opinion ) .
because the affinity is is general not symmetric , some agents could have been included in the group determined by _
i _ , againsttheir advise ,
hence a second relevant variable is the _ averaged affinity _ that each agent perceives of the group itself : @xmath15 once the former two quantities have been computed by each agent , we propose the following update scheme : to belong in the largest size group , each agent , would like to come closer to its perceived apopsicentre if it feels himself affine enough to the group : @xmath16 where @xmath17 $ ] is an activating function , defining the region of trust for effective social interactions , e.g. @xmath18 larger than @xmath19 . moreover sharing a close opinion ,
reinforce the mutual affinity , while too far opinions make the relationship to weak , hence each agent becomes more affine with all the agents in the subgroup that share opinions close enough to its perceived apopsicentre , otherwise their affinities will decrease : @xmath20 where @xmath21 , and @xmath22 \ , , \label{eq : gamma2}\ ] ] that can be considered again as an activating function for the affinity evolution . in the previous relations for @xmath23 , we set the parameters @xmath24 and @xmath25 , large enough to practically replace the hyperbolic tangent with a simpler step function . under these assumptions
@xmath26 takes values either 0 or 1 , while the value of @xmath27 are either -1 or 1 .
the interaction mechanism is schematically represented in fig .
[ fig : cartoon ] .
a typical run of this model is presented in fig .
[ figuranuovaconevolutioneimagescdelleop ] .
first of we can observe that the dynamics is faster than in the similar model presented in , this is because binary interactions are replaced by multi
agents interactions that improve the information spread
. moreover there exists a transient interval of time , where nobody modifies its opinion , but only the mutual affinities ( see insets of fig .
[ figuranuovaconevolutioneimagescdelleop ] and the relative caption ) .
only when the relationships become highly trustable , agents do modify also their opinions .
this behavior is explained by the different time scales of the main processes : evolution of opinions and evolution of affinity , as it clearly emerges again from the insets of fig .
[ figuranuovaconevolutioneimagescdelleop ] , where we show three time - snapshots of the affinity , i.e. the social network , once agents start to modify their opinions . transferring this behavior to similar real social experiment
, we could expect that , initially unknown people first change ( in fact construct ) their affinities relationships ( increasing or decreasing mutual affinities ) and only after that , they will eventually modify their opinions .
namely initially people samplethe group and only after they modify their beliefs . , @xmath28 , @xmath29 , @xmath30 agents
whose opinion are initially uniformly distributed in @xmath31 $ ] , whereas initial affinities are uniformly distributed in @xmath32 $ ] .
three time - snapshot of social networks are reported for increasing times , @xmath33 .
dots represent agents that are possibly linked if they are affine enough .
each network has been characterized by some standard topological indicator ; the diameter and the averaged shortest path ( that are infinite for @xmath34 and respectively have values @xmath35 and @xmath36 for @xmath37 and @xmath38 and @xmath39 for @xmath40 ) , the averaged degree ( that takes values @xmath41 , @xmath42 and @xmath43 respectively for @xmath34 , @xmath37 and @xmath40 ) and the averaged network clustering ( that assumes the values @xmath44 , @xmath45 and @xmath46 respectively for @xmath34 , @xmath37 and @xmath40).,width=264 ] in fig .
[ figuraconconvergenzeedistrofinali ] we report two different outcomes of numerical simulations of the model , for two sets of parameters , in the left panel we found once again a consensus status , where all the population share the same opinion , as reported by the histogram . while in the right panel , the population polarizes , and @xmath5 is not too small , there always exists a finite , though small , probability of selecting in the same acquaintance group individuals belonging to different opinion clusters , hence producing a gradual increase in the mutual affinities , which eventually lead to a merging of the , previously , separated clusters
. this final state will be achieved on extremely long time scales , diverging with the group size : socially relevant dynamics are hence likely to correspond to the metastable regimes . a similar phenomenon has been observed in . ] into clusters of different opinions , here @xmath35 . , @xmath28 ( left panel ) , and @xmath47 , @xmath28 ( right panel ) both with @xmath29 and @xmath30 agents
whose initial opinion are uniformly distributed in @xmath31 $ ] , whereas initial affinities are uniformly distributed in @xmath32$],title="fig:",width=188 ] , @xmath28 ( left panel ) , and @xmath47 , @xmath28 ( right panel ) both with @xmath29 and @xmath30 agents whose initial opinion are uniformly distributed in @xmath31 $ ] , whereas initial affinities are uniformly distributed in @xmath32$],title="fig:",width=188 ] hence one can characterize the final asymptotic state with the number of opinion clusters as a function of the key parameters @xmath5 and @xmath19 .
we observe that often the asymptotic state exhibits _ outliers _ , namely clusters formed by very few agents , and also that , because of the random encounters and initial distributions , the same set of parameters can produce asymptotic state that can exhibit a different number of opinion clusters .
for this reason we define the _ average number _ of opinion clusters , @xmath48 , repeating the simulation a large number of times , here @xmath49 .
a second possibility is to use the _ deridda and flyvbjerg _
number @xcite : @xmath50 where @xmath51 is the total number of clusters obtained in the asymptotic state and @xmath52 is the number of agents in the @xmath53th cluster .
the quadratic dependence on @xmath54 ensures that less weight has been given to small clusters with respect to larger ones . in fig .
[ fig : clumedioederiddavsdeltagc ] we report the results of the analysis of @xmath48 and @xmath55 as a function of @xmath5 , for a fixed value of @xmath19 .
when @xmath56 .
average and standard deviation on @xmath57 simulations .
the distributions of the average number of cluster are presented for @xmath58 ( right inset ) , @xmath59 ( central inset ) and @xmath60 ( left inset),title="fig:",width=207 ] when @xmath56 .
average and standard deviation on @xmath57 simulations .
the distributions of the average number of cluster are presented for @xmath58 ( right inset ) , @xmath59 ( central inset ) and @xmath60 ( left inset),title="fig:",width=207 ] a phase transition from a mono
cluster state , i.e. _ consensus _ , to polarization of the opinions in a population of @xmath61 agents , emerges close to @xmath62 , for smaller values of @xmath5 the distribution of the number of cluster can be well described by a normal distribution ( see left inset fig .
[ fig : clumedioederiddavsdeltagc ] ) , for larger value of @xmath5 , only one cluster is present ( see right inset fig . [
fig : clumedioederiddavsdeltagc ] ) , while for @xmath5 varying around @xmath63 an exponential distribution can be found ( see middle inset fig . [
fig : clumedioederiddavsdeltagc ] ) , reinforcing thus the statement of the existence of a phase transition . data from fig .
[ fig : clumedioederiddavsdeltagc ] suggest an exponential growth of @xmath48 as a function of @xmath5 below the phase transition value , we thus compute a linear fit on log log scale ( see fig .
[ fig : logfig7a ] ) in the region involving small values of @xmath5 , obtaining : @xmath64 when @xmath56 .
a similar power law behavior is still valid also for the derrida and flyvbjerg number .
the existence of a power low seems robust with respect to variations of the parameter @xmath19 ( see fig . [
fig : logfig7a ] ) . the results presented in fig .
[ fig : logfig7a ] allow us to extract also the behavior of the average number of clusters as a function of the second parameter @xmath19 for a fixed @xmath5 .
in fact moving upward on vertical lines , i.e. decreasing @xmath19 , the @xmath48 increases if @xmath5 is below the critical threshold , while above this value the number of clusters is always equal to one .
moreover from these data we could conclude that the phase transition point seems to be independent from the value of @xmath19 .
( log log scale ) .
best linear fits for different values of @xmath19 .
each simulation has been repeated @xmath57 times , width=245 ] another relevant quantity of interest , is the time needed to form an opinion cluster , the _ opinion convergence time _ , @xmath65 , and its dependence of the size of the cluster .
numerical simulations not reported here , emphasize that in the polarization case @xmath65 depends in a highly non trivial way on the total number of clusters and on their sizes , roughly speaking if in consensus state the time needed to form a cluster of say @xmath66 individual is some value @xmath67 , then the time needed to form a cluster of the same size in a polarization case , has nothing to do with @xmath67 and it depends on all the formed clusters . because this paper offers a preliminary analysis
, we decided to consider only the consensus case , hence choosing parameters ensuring the existence of only one cluster and we define the _ convergence time _
@xmath65 to be : @xmath68 where @xmath69 is a small parameter ( hereby @xmath70 ) .
we thus performed dedicated simulations with @xmath0 ranging from few unities to thousand unities .
the results reported in fig . [ fig : opconvdist ] suggest a non linear dependence of @xmath65 on @xmath0 , well approximable by @xmath71 . using a regression analysis on the data
, we can estimate the exponent which results , @xmath72 for @xmath73 and @xmath74 for @xmath75 .
let us observe that as @xmath5 approaches the phase transition value , @xmath65 increases and the curve becomes more noisy ( see for instance the bottom panel of fig .
[ fig : opconvdist ] corresponding to @xmath76 ) , that is because the occurrence of the consensus case becomes lesser and lesser probable .
( top - left ) , @xmath77 ( top - right ) and @xmath78 ( bottom ) .
insets log log plots of @xmath65 as a function of @xmath0.,width=283 ]
in this paper we introduced a new model of opinion dynamics where agents meet in social groups , affinity driven , and possibly update their beliefs as a consequence of these local , i.e. group level , interactions .
the model exhibits a rich phenomenology determined by the interlay between the dynamics of the opinions and the mutual affinities .
we emphasized the role of two parameters , @xmath5 and @xmath19 , which can be qualitatively interpreted respectively as the openness of mind in the formation of the group and as the openness of mind in the intra group dynamics .
we thus studied the behavior of the model as a function of these two parameters .
the formulation of our model has been inspired by the observation of the way in which the formation mechanisms for social interactions do occur in the real world : a large majority of the processes of formation and evolution of the opinions are driven by group based discussions , such groups are determined by the mutual affinity and/or the shared opinion .
the processes of group formation which tends to form clusters of acquaintances ( or collaborators ) are introduced in our model via the selection mechanism based on the mutual trust , i.e. eq . and .
the numerical analysis we performed , shows a dependence of the consensus / polarization state on the degree of mind openness in the creation of the groups of acquaintances , i.e. the parameter @xmath5 , large values corresponding to consensus states while small ones to fragmented groups .
that is the main reason why the model exhibits a phase transition with respect to this variable .
finally the intra
group dynamics , based on the mutual affinity , allows to update the opinions only for agents that perceive the group discussion close enough to its believes .
this phenomenon is modeled by eq . and .
our analysis shows that the stronger is the degree of intra group affinity required to make an interaction effective , i.e. large @xmath19 , the higher will be the degree of polarization of the population .
we can thus conclude that the model here presented , exhibits and well reproduces the two underlying dynamical mechanisms that can drive the opinion formation process in ( relatively ) small groups : exchange of information and mutual trust .
moreover these mechanisms evolve on different times scales as clearly showed previously .
this model represents thus a framework where to study groups interactions with applications to real social systems .
it would be interesting to improve the model by introducing , at least , two factors : vectorial opinions @xcite i.e. agents discuss and exchange information about more than one subject .
second , introduce a limitation in the number of agents with which anyone can be affine with , as usually is the case in the real social networks . | we hereby propose a model of opinion dynamics where individuals update their beliefs because of interactions in acquaintances group .
the model exhibit a non trivial behavior that we discuss as a function of the main involved parameters .
results are reported on the average number of opinion clusters and the time needed to form such clusters . |
except for a few nuclei lying in the vicinity of shell closures , most of the heavy nuclei are difficult to describe in a spherical shell model framework because of the unavoidable problem of dimension explosion .
therefore , the study of nuclear structure in heavy nuclei has relied mainly on the mean - field approximations , in which the concept of spontaneous symmetry breaking is applied @xcite .
however , there has been an increasing number of compelling evidences indicating that the nuclear many - body correlations are important .
thus , the necessity of a proper quantum mechanical treatment for nuclear states has been growing , and we are facing the challenge of understanding the nuclear structure by going beyond the mean - field approximations .
demand for a proper shell model treatment arises also from the nuclear astrophysics . since
heavy elements are made in stellar evolution and explosions , nuclear physics , and in particular nuclear structure far from stability , enters into the stellar modeling in a crucial way .
the nucleosynthesis and the correlated energy generation is not completely understood , and the origin of elements in the cosmos remains one of the biggest unsolved physics puzzles .
nuclear shell models can generate well - defined wave functions in the laboratory frame , allowing us to compute , without further approximations as often assumed in the mean - field approaches , the quantities such as transition probabilities , spectroscopic factors , and @xmath0-decay rates .
these quantities provide valuable structure information to nuclear astrophysics .
in fact , the nuclear shell model calculations could strongly modify the results of nuclear astrophysics , as the recent work of langanke and martinez - pinedo has demonstrated ( see , for example , ref .
@xcite ) .
tremendous efforts have been devoted to extending the shell model capacity from its traditional territory of the @xmath1-shell to heavier shells . over the years
, one has looked for possible solutions in the following two major directions . in the first direction , one employs rapidly growing computer power and sophisticated diagonalization algorithms to improve the traditional shell model code .
the shell model code antoine @xcite is a representative example of the recent developments along this line . using this code the deformed @xmath2 nuclei up to mass @xmath3
can be well explained .
the recent record example performed by this code is the full @xmath4 shell calculation of @xmath5 nuclei @xcite , with the basis dimensions in excess of 100 million . while in principle
, it does not matter how to prepare a shell model basis , it is crucial in practice to use the most efficient one .
moreover , feasibility in computation is not our only concern .
the other important aspect of using an efficient basis is that it may have a good classification scheme , such that a simple configuration in that basis corresponds approximately to a real mode of excitation .
this not only can simplify the calculations , but also make the physical interpretations of results more easy and transparent . in the second direction , which is defined in a much wider scope , one employs various methods in seeking judicious truncation schemes .
such schemes should contain the most significant configurations , each of which can be a complicated combination in terms of the original shell model basis states . in this way
, the basis dimension can be significantly reduced and the final diagonalization is performed in a much smaller space , thus making a shell model calculation for heavy nuclei possible .
the early monster - vampire approach @xcite and the recent monte carlo shell model @xcite are examples along this line
. nevertheless , numerical calculations required by these models are still quite heavy , which may make a systematical application difficult .
there are two other existing models that belong to the second category : the project shell model ( psm ) @xcite , and the fermion dynamical symmetry model ( fdsm ) @xcite . in the psm
, the shell model basis is constructed by choosing a few quasi - particle ( qp ) orbitals near the fermi surfaces and performing angular - momentum and particle - number projection on the chosen configurations . by taking multi - qp states as the building blocks
, the psm has been designed to describe the rotational bands built upon qp excitations @xcite .
the psm has been rather successful in calculating the high - spin states of well - deformed and superdeformed nuclei . for lighter nuclei where the large - scale shell model calculation is feasible
@xcite , studies for the deformed @xmath6cr @xcite and the superdeformed @xmath7ar @xcite have demonstrated that the psm calculation can achieve a similar accuracy in describing the data . in the fdsm , the truncated basis is built by the symmetry detected @xmath8- and @xmath9-pairs , assuming that these are the relevant degrees of freedom for the low - lying collective motions .
having these pairs as the building blocks , the fdsm can provide a unified description for the low - spin collective excitations from the spherical to the well - deformed region @xcite .
it is clear that the psm and the fdsm follow the shell model philosophy and both have their own shell model truncation scheme .
however , the truncations emphasize different excitation modes , which are contained in one model but are absent in the other .
an idea emerges naturally that one may combine the advantages of the two models to construct a new shell model for heavy nuclei .
the question is how .
the psm is a microscopic approach employing the deformed intrinsic states and the projection method , while the fdsm is a fermionic model based on the group theory .
the crucial step that leads us to connect these two different approaches is made through the recent recognition @xcite that the numerical results obtained by the psm exhibit , up to high angular momenta and excitations , a remarkable one - to - one correspondence with the analytical @xmath10 spectrum of the fdsm .
this suggests that the projected deformed - bcs vacuum has at the microscopic level @xmath10-like structures which are very close to the representations of the @xmath10 dynamical symmetry of an @xmath8-@xmath9 fermion - pair system .
this recognition has motivated us to propose a multi - shell shell model for heavy nuclei .
hereafter , we shall call it heavy shell model , or hsm for short . in the following section , the psm and the fdsm will be briefly reviewed .
the emphasis will be given on the discussion of the advantages and deficiencies of each model .
in section 3 , the connection between the two models will be explored . in section 4
, we will discuss in detail how the two models are integrated to form the hsm
. we will give the basis states and the basis truncations for the well - deformed , transitional , and spherical regions .
the effective interactions and the general method for evaluating the projected matrix elements will also be discussed in this section . finally , the paper will be summarized in section 5 .
in this section , we introduce the basic structure of the psm and the fdsm . for each model , we point out the main features and the limitations . for interested readers ,
we refer to the review article of the psm @xcite and of the fdsm @xcite .
the psm begins with the deformed ( e.g. the nilsson - type ) single particle basis , with pairing correlations incorporated into the basis by a bcs calculation for the nilsson states .
the basis truncation is first implemented in the multi - qp states with respect to the deformed - bcs vacuum @xmath11 ; then the angular - momentum and the particle - number projection are performed on the selected qp basis to form a shell model space in the laboratory frame ; finally a shell model hamiltonian is diagonalized in this projected space .
if @xmath12 and @xmath13 are the qp creation operators , with the index @xmath14 ( @xmath15 ) denoting the neutron ( proton ) quantum numbers and running over properly selected single - qp states , the multi - qp bases of the psm are given as follows : @xmath16 in bases ( 1 ) ,
@xmath17 " denotes those configurations that contain more than two like - nucleon quasi - particles .
if one is interested in the low - lying states only , they can practically be ignored because these configurations have higher excitation energies due to mutual blocking of levels .
the bases ( 1 ) can be easily enlarged by including higher order of multi - qp states if necessary . if the configurations denoted by @xmath17 " are completely included , one recovers the full shell model space written in the representation of qp excitation . in the qp basis ,
the truncation can now be easily implemented by simply excluding the states with higher energies .
usually , only a few orbitals around the fermi surfaces are sufficient for a description of the low - lying qp excitations .
the truncation is thus so efficient that dimension never poses a problem even for superdeformed , heavy nuclei .
after the truncation is implemented in the multi - qp basis , the shell model space can be constructed by the projection technique @xcite : @xmath18 where @xmath19 denotes the qp - basis given in ( 1 ) with @xmath20 meaning the multi - qp configuration , and @xmath21 are the angular momentum and particle number projection operators , respectively . in eq.([pn ] ) , @xmath22 is the @xmath9-function @xcite , @xmath23 the rotation operator , @xmath24 the solid angle , @xmath25 the number operator , and @xmath26 the gauge angle .
if one keeps the axial symmetry in the deformed basis , @xmath22 in eq.([pn ] ) reduces to the small @xmath27-function and the three dimensions in @xmath24 reduce to one .
the eigen - values @xmath28 and the corresponding wave functions @xmath29 are then obtained by solving the following eigen - value equation : @xmath30 where @xmath31 and @xmath32 are respectively the matrix elements of the hamiltonian and the norm @xmath33 the psm uses a large size of single - particle ( s.p . ) space , which ensures that the collective motion is defined microscopically by accommodating a sufficiently large number of active nucleons .
it usually includes three ( four ) major shells each for neutrons and protons in a calculation for deformed ( superdeformed ) nuclei .
the effective interactions employed in the psm are the separable forces .
the hamiltonian takes the following form : @xmath34 the first term @xmath35 in @xmath36 of eq.([h ] ) is the spherical single - particle hamiltonian and the remaining terms are residual quadrupole - quadrupole , monopole - pairing , and quadrupole - pairing interactions , respectively .
the strength of the quadrupole - quadrupole force is determined in a self - consistent way that it would give the empirical deformation as predicted in the variation calculations .
the monopole - pairing strength is taken as the form @xmath37 ( @xmath38 is the mass number ) with @xmath39 being adjusted to yield the known odd - even mass differences . the quadrupole - pairing strength @xmath40 is assumed to be about 20@xmath41 of @xmath42 @xcite .
the one - body operators ( for each kind of nucleons ) in eq.([h ] ) are of the standard form @xmath43 where @xmath44 is the one - body matrix element of the quadrupole operator and @xmath45 the nucleon creation operator , with @xmath46 standing for the quantum numbers of a s.p .
state in the spherical basis ( @xmath47 ) .
the time reversal of @xmath45 is defined as @xmath48 . for a description of rotational bands associated with a well - deformed minimum ,
the psm is a highly efficient truncation scheme .
diagonalization for a heavy nucleus can be done almost instantly , yet the results are often satisfactory .
the reason for the success is because the major part of pairing and quadrupole correlations has already been built in the basis through the use of deformed basis and the bcs formalism . therefore , a small configuration space with a few s.p .
orbits around the fermi surface can already span a very good basis for the low - lying excitations .
note that each of the configurations in the basis is a complex mixture of multi - shell configurations of the spherical shell model space .
although the final dimension of the psm is small , it is huge in terms of original shell model configurations . in this sense , the psm is a shell model in a truncated multi - major - shell space .
the pleasant features of the psm make it a frequently - used model in the high - spin physics .
many applications can be found in the review article @xcite .
the recent publications include the study of superdeformed structure in a wide range of different mass regions @xcite , the study of the origin of identical bands @xcite and of the high - k states @xcite .
while the psm is an efficient shell model for deformed systems with rotational behavior , it becomes less valid when going to the transitional region , and eventually loses its applicability for spherical nuclei .
moreover , it can not efficiently describe the @xmath0- and @xmath49-vibrations .
although such collective modes can in principle be obtained by mixing a large amount of excited qp - configurations , it suffers in practice from a similar dimension problem as in conventional shell models .
the main reason for these shortcomings is due to the use of the simple bcs - vacuum which contains only the properties of the ground - state rotational band but not those of the collective vibrations .
if we say that the psm is a shell model in a truncated multi - major - shell space , then the fdsm is a shell model in a truncated one - major - shell space .
the truncation is based on the consideration that the like - nucleons prefer to form coherent @xmath8 ( angular - momentum @xmath50 ) and @xmath9 ( @xmath51 ) pairs .
one may thus assume that a closed subspace built up by these @xmath8-@xmath9 pairs is mainly responsible for nuclear low - lying collective motions .
such an @xmath8-@xmath9 subspace can be carved out by a symmetry requirement that the @xmath8 and @xmath9 creation ( annihilation ) operators together with necessary minimum amount of number conserving operators form a closed lie algebra . to this end , a @xmath52-@xmath53 basis is introduced ( see ref . @xcite and
the references cited therein ) @xmath54 where @xmath52 ( pseudo - orbital angular - momentum ) and @xmath53 ( pseudo - spin ) could be either @xmath55 and @xmath56 any half integer , or @xmath57 and @xmath58 any integer .
this basis must uniquely reproduce the normal - parity levels @xcite in that shell by @xmath52-@xmath53 angular momentum coupling , no more and no less . with this @xmath52-@xmath53 basis ,
the coherent @xmath8 and @xmath9 pairs and the multipole operators @xmath59 , which are necessary in order to form a closed lie algebra , are found to be as follows @xcite : @xmath60^{00}_{00}\\ \mbox{(for any $ k$ and $ i$})\end{array},\
\ \omega_{ki}=\frac{(2k+1)(2i+1)}2 \label{s}\ ] ] @xmath61^{20}_{\mu0 } , \
\ d_\mu^\dagger=\sum_i\sqrt{\frac{\omega_{ki}}2 } \left[\ , b_{ki}^\dagger b_{ki}^\dagger\right]^{02}_{0\mu}\\ \mbox{($k$-active : $ k=1 $ ) } \hspace{64pt}\mbox{($i$-active : $ i=3/2 $ ) } \end{array } \label{d}\ ] ] @xmath62^{r0}_{\mu0},\ p_{r \mu}\hspace{-2pt } = \sum_i\sqrt{2\omega_{ki } } \left[\ , b_{ki}^\dagger \tilde{b}_{ki}\right]^{0r}_{0\mu}\\ \hspace{-6pt}\mbox{($k$-active : $ k=1 $ , $ r\leq 2 $ ) } \hspace{20pt}\mbox{($i$-active : $ i=3/2 $ , $ r\leq 3$)}\hspace{6pt } \end{array}\hspace{12pt } \label{p2}\ ] ] where the symbol @xmath63 $ ] denotes angular - momentum coupling , and the time reversal is defined as @xmath64 .
the operator set @xmath65 forms a closed lie algebra of either @xmath66 ( @xmath52-active ) or @xmath67 ( @xmath53-active ) , depending on the level structure of the valence shell @xcite .
once the @xmath8-@xmath9 subspace is carved out , the form of the effective hamiltonian ( restricted to a two - body interaction ) in this truncated space is uniquely determined : @xmath68 it can be shown that the multipole operators @xmath69 with @xmath70 are proportional to the number operator in normal parity levels @xmath71 and the total angular - momentum @xmath72 , respectively , while @xmath73 is proportional to the effective quadrupole operator in the truncated space .
the term @xmath74 is a quadratic function of valence neutron and proton numbers ( @xmath75 is included in the n - p interaction @xmath76 ) .
this hamiltonian ( eq.([hfds ] ) ) looks formally similar to that of the psm ( eq.([h ] ) ) if we assume that only @xmath77 in the summation .
however , one should bear in mind that the fdsm hamiltonian is written in a truncated one - major shell space and the s.p .
energy splitting within the shell is neglected .
this is the price the fdsm has to pay in order to meet the symmetry requirement so that solutions can be obtained with the aid of group theory . with this hamiltonian
, it can be shown that there exist analytical solutions in various dynamical symmetry limits , each one corresponding to a collective mode known experimentally : the @xmath10-limit in the @xmath52-active shell and the @xmath78-limit in the @xmath53-active shell correspond respectively to rigid and @xmath49-soft rotors for well deformed nuclei , while the @xmath79-limit in the @xmath52-active shell and the @xmath80-limit in the @xmath53-active shell correspond to a vibrator of spherical nuclei .
since the fdsm contains all major collective modes , the general feature of different collective motions arises naturally as the number of valence nucleons varies ; namely , nuclei behave as a spherical vibrator near the closed shell and become a well - deformed rotor around the mid - shell . if the strengths of the hamiltonian are properly chosen and the s.p .
splitting is taken into account as a perturbation , the fdsm can even quantitatively reproduce the low - lying spectra , b(e2 ) s , ground state masses , etc . , in a unified manner from the spherical to well - deformed region . * the @xmath10 intrinsic states for each of the @xmath0-@xmath49 vibrational modes are listed in the first column .
the second and the third column are the phonon excitation energies and the associated quantum numbers according to the particle - rotor model . the 4th and 5th column are the @xmath10 irreps @xmath81 and the corresponding young tablets @xmath82 $ ] , respectively .
the excitation energies obtained from the fdsm are listed in the last column as @xmath83 , where @xmath84 is the changes of the expectation values of the @xmath10 casimir operator with respect to the ground state irrep @xmath85 , @xmath86 is the @xmath87-@xmath88 quadrupole interaction strength , and @xmath89 . here , it may be appropriate to emphasize one remarkable result of the fdsm : there exists an one - to - one correspondence between the @xmath10 irreps and the @xmath0- and @xmath49-vibrations in deformed nuclei . as one can see from table i , the @xmath0-@xmath49 vibrations are microscopically the collective @xmath9-pair excitations ( the @xmath90 excitations are forbidden by the time reversal symmetry ) .
the anharmonic behavior of the @xmath0-@xmath49 vibrations is due to the finite particle number effect . in the large @xmath71 limit ,
ignoring @xmath91 , the fdsm reproduces exactly the particle - rotor model results .
this means that the fdsm has discovered the relevant fermion degrees of freedom of nuclear collective motions , which are the @xmath8 and @xmath9 pairs .
the @xmath8-@xmath9 subspace is so compact that it never suffers from the dimension explosion even for the heaviest nuclear systems . while with the @xmath8-@xmath9 subspace the fdsm is able to provide a microscopic view to the low - lying collective motions , it has difficulties in describing the qp excitations and the high - spin physics for lack of the s.p .
degrees of freedom . in principle
, this can be resolved by allowing a few pairs to break .
the problem is that once the s.p .
degrees of freedom open up , the dimension increases very rapidly .
in addition , inclusion of s.p .
degrees of freedom results in adding many new terms to the effective hamiltonian .
to pin down so many coupling strengths in the hamiltonian is a very difficult task , if not impossible .
therefore , even the number of broken pairs is limited to just one for proton and one for neutron , the model could at best be applied to the @xmath4 shell nuclei , and could not go any further .
let us summarize our main claims in the last section .
both the psm and the fdsm are truncated shell model , aiming at grasping the essential ingredients to describe the low - lying physics . however , the emphasis in each model is different , which is reflected in their different truncation schemes .
the psm emphasizes the high - spin description of rotational states built upon qp - excitations associated with a well - deformed minimum , but it is not an efficient method for the collective vibrations . in contrast , the fdsm has a well - defined classification for all known types of collective vibration , ranging from the spherical , via transitional , to the well - deformed region , but it is lack of the necessary degrees of freedom of qp - excitations .
thus , the main advantages of the psm and the fdsm are mutually complementary to each other .
if we could take the advantages of the psm and the fdsm and combine them into a single model , the deficiencies of each model will be eliminated . at first glance , it is not obvious how to bridge these two different approaches . in this section ,
we show that a realization of the combination is possible .
the assertion is made based on the recent recognition that an @xmath10 symmetry can emerge from the projected deformed - bcs vacuum .
having this as the basis , the full idea of the microscopic classification for the collective vibrations discovered by the fdsm could be adopted by the psm while the latter keeps its original features in building the shell model configuration space through projection .
comparison between the psm ( symbols ) and the fdsm @xmath10 limits ( curves ) assuming @xmath92 and @xmath93 .
the @xmath10 quantum numbers @xmath94 are marked for each band with degeneracy of @xmath95 , and @xmath96 .
the comparison takes the well - deformed nucleus @xmath97er as an example . ]
it is remarkable @xcite that the states numerically obtained by the psm exhibit a beautiful one - to - one correspondence with the analytical @xmath10 spectrum of the fdsm . to show this ,
the original psm was extended in such a way that instead of a single bcs vacuum , the angular momentum projection was performed for separate neutron and proton bcs vacuum with the same deformation , and the projected states were coupled through diagonalization of the hamiltonian @xcite .
the extension is necessary ; otherwise the psm would have no collectively excited bands to compare with the fdsm .
this procedure gives the usual rotational ground - band corresponding to a strongly coupled bcs condensate of neutrons and protons , but also led to a new set of excited bands arising from the vibrations of relative orientation of the neutron and proton cores ( the so - called scissors " mode ) .
as one can see from fig.1 , the spectrum obtained from the psm is , up to high angular momenta and excitation energies , nearly identical to that from the fdsm @xmath10 formulism with @xmath98 and @xmath99 irreps , respectively assigned to the neutron and proton bcs vacuum .
the classification of the spectrum follows exactly the fdsm @xmath10 reduction rules @xmath100 with @xmath101 and @xmath102 not only the spectrum but also the b(e2 ) s exhibit this correspondence @xcite .
note that there are many types of @xmath10 models , and that different @xmath10 models have different permissible irreps and reduction rules due to the different physical input , and therefore can lead to different band structures . here , we emphasize that this @xmath10 symmetry shown in fig .
1 is of the fdsm type .
this is because among the existing fermionic @xmath10 models in nuclear physics , only the fdsm @xmath10 formulism can naturally provide the required irreps and reproduce the band structure of the psm .
other @xmath10 models such as the pseudo-@xmath10 do not have this property . in their recent investigation of the onset of rotational motion ,
zuker _ et .
@xcite introduced a preliminary formulation of an approximate quasi-@xmath10 symmetry .
it would be interesting to study whether the quasi-@xmath10 contains a similar property .
fig.1 presents a highly non - trivial result because the psm , as described above , is not built on any explicit @xmath10 symmetry , and no free parameters have been adjusted to obtain such a symmetry .
nevertheless , the spectra , the electromagnetic transition rates , and the wave functions of the psm agree nearly perfectly with the fdsm @xmath10 results , from the ground - band to states of high spins and high excitations @xcite .
this strongly suggests that the projected deformed - bcs vacuum in the psm could be at the microscopic level close to the @xmath8-@xmath9 core in the fdsm , provided that the @xmath8-@xmath9 pairs must be redefined in a multi - major - shell space , since the one - major - shell is not large enough to accommodate so many active nucleons ( here @xmath103 and @xmath104 ) . in this regard
, there is a conceptual distinction between the @xmath10 symmetry in the one - major - shell fdsm and that emerged from the psm . in the former ,
the @xmath10 symmetry arises entirely from the normal - parity nucleons , and the abnormal parity orbit enters only implicitly through the pauli effect and the renormalization of the parameters . in the latter , the @xmath10 symmetry arises from the explicit dynamical participation of both normal and abnormal - parity nucleons of many shells .
ignoring the direct contribution from the abnormal - parity nucleons in the one - major - shell fdsm is a sacrifice for having an exact symmetry ( which forms a closed lie algebra ) . in practice , this turns out to be not a bad approximation for collective motions because in a single major shell , there is only one single-@xmath105 level with abnormal - parity , which does not have much quadrupole collectivity compared to the normal - parity contributions @xcite .
when going to a multi - major - shell space , the situation will change .
a bunch of abnormal - parity levels , which are located just below the normal - parity levels , will open up .
this means that , when one redefines the fdsm - type coherent @xmath8-@xmath9 pairs in a multi - major - shell space , the conceptual distinction of the @xmath10 symmetry between the fdsm and the psm will be eliminated . to see further the relationship between the two models ,
let us ignore the terms with @xmath106 in the most general fdsm hamiltonian ( [ hfds ] ) .
this is a reasonable approach because from the multipole expansion point of view , there is no multipole interactions with @xmath107 without considering the parity admixture .
one may notice that there is the monopole - monopole interaction ( @xmath108 ) in the fdsm , which is not included in the psm .
however , it is well - known that this interaction only affects the nuclear total binding energy but does not have much influence on the excitations . on the other hand ,
it can be added to the psm if necessary .
thus , we have for both models the interactions of the monopole- and quadrupole - pairing plus the quadrupole - quadrupole type .
the psm hamiltonian contains operators written in the ordinary shell model basis , whereas the fdsm ones in the @xmath52-@xmath53 basis . in order to compare them ,
we transform the fdsm operators in eq.([s][p2 ] ) back to the ordinary shell model basis by rewriting them in terms of @xmath45 and @xmath109 .
we find that @xmath110 with @xmath111 .
\nonumber \end{aligned}\ ] ] comparing eq.(14 ) with eq.([p ] ) , we see that the definitions of @xmath112 , @xmath113 and @xmath114 in the fdsm are similar to that of @xmath115 , @xmath116 and @xmath117 in the psm , except that the model spaces for the two models are different .
however , within one - major - shell , the operator @xmath112 ( @xmath8 ) is exactly the same as @xmath118 ( @xmath119 ) in the psm . for the @xmath9-pair and @xmath120 operators ,
although the coefficients @xmath121 look different from @xmath122 appearing in the corresponding operators @xmath116 and @xmath117 in the psm , their physical meanings are the same .
the fdsm hamiltonian may thus be considered as an one - major - shell version of the psm hamiltonian , with the approximations of @xmath116 and @xmath117 being replaced by @xmath113 and @xmath114 and the s.p .
energy splitting being ignored . in other words ,
if , in the multi - major - shell case , the symmetry constraint is released from the fdsm and the s.p.energy splitting is considered , @xmath112 , @xmath113 and @xmath114 in the fdsm should return back to the version of @xmath115 , @xmath116 and @xmath117 operators in the psm .
so is for the hamiltonian . from the above analysis
we may conclude that the psm and the fdsm are just two approaches to solve an effective hamiltonian of a common form . in order for them to be applicable to heavy nuclei
, approximations have to be made in each model .
the psm is based on a multi - major - shell space so that it can describe the nuclear rotational motion microscopically through a dynamic participation of many particles .
however , it can afford to do so only for a truncated configuration space that includes only a bcs vacuum plus a few qp excitations . as a sacrifice
, this truncation do not include the collective modes such as @xmath0- and @xmath49-vibrations .
in contrast , the fdsm aims at nuclear low - lying collective excitations .
it can afford to do so only when the model space is reduced to one - major shell like in the conventional shell model .
but that is still not enough for heavy nuclei
. additional approximations to further reduce the configuration space down to the symmetry detected @xmath8-@xmath9 subspace are necessary .
these approximations exclude the s.p
. excitations .
having realized that the projected deformed - bcs states exhibit the @xmath10 symmetry , and that the collective excitations may be approached by the @xmath9-pair excitations , with the @xmath9-pair operator defined as the quadrupole - pair operator @xmath116 in a multi - major - shell model space , we propose a multi - shell shell model : the heavy shell model .
the essence of this proposal is to adopt the truncation scheme for the collective modes , which was discovered by the fdsm , into the psm to enrich the shell model basis .
this essentially combines the advantages of both models .
in fact , to incorporate both s.p . and collective excitations , there are in principle two alternatives
: one can either extend the fdsm by adding the psm qp truncation scheme on top of the fdsm collective states , or extend the psm by including the @xmath9-pair collective excitations into the psm vacuum .
however , the fdsm is a severely truncated one - major - shell shell model dictated by symmetry , and in this sense , it is not as microscopic as the psm .
although it is quite successful in the description of low - lying collective motions , the fdsm is just an effective theory . for practical applications , large renormalization effects
must be embedded in the parameters of the fdsm hamiltonian , which have to be determined phenomenologically .
therefore , we choose to construct the heavy shell model based on the extension of the psm .
the main ingredients of the heavy shell model are as follows : 1 .
keep the multi - major - shell basis of the psm as the model space , using the psm @xmath118 and @xmath123 operators to describe the coherent @xmath8- and @xmath9-pairs in the multi - major - shell configurations , and construct the intrinsic collective excitation states by @xmath123 acting on the deformed - bcs vacuum ; 2 .
carry out the shell model truncation by selecting a few single - qp states near the fermi surfaces plus a few @xmath9-pair excitations , and perform angular momentum and particle number projection to obtain a shell - model basis in the laboratory frame ; 3 . keep the psm hamiltonian to be the effective hamiltonian , but allow adding more multipole interactions and/or readjusting the interaction strengths if necessary ; 4 .
utilize the algorithms developed in the psm to carry out calculations for all the necessary matrix elements , and diagonalize the hamiltonian in the truncated shell model space .
let us now discuss each of the items in more detail .
we have demonstrated that the projected deformed - bcs vacuum @xmath124 in the psm is nearly identical to the fdsm-@xmath10 intrinsic ground state ( @xmath125 ) irrep .
furthermore , we have indicated that the fdsm @xmath8- and @xmath9-pair and the @xmath114 operator are , respectively , the symmetry - constraint one - major - shell version of the ordinary monopole pair @xmath119 , quadrupole pair @xmath126 , and quadrupole operator @xmath117 in the psm .
it is therefore natural to believe that in a multi - major - shell space , the fdsm @xmath8- and @xmath9-pairs are nothing but the @xmath119- and @xmath126-pairs if we abandon the symmetry requirement .
so is for the quadrupole operator . in the fdsm
, all known types of low - lying collective excitation can be obtained by acting @xmath127 on the fdsm-@xmath10 intrinsic ground state , @xmath128 ( see table i ) . combining these facts , the collective excitations of the hsm in a multi - major - shell space
may be constructed by replacing , respectively , @xmath128 and @xmath129 in table i with @xmath124 and @xmath130 defined in the psm .
hereafter , we will continue to use the fdsm notations @xmath8 and @xmath9 for the pair operators .
one should bear in mind that they have been redefined as @xmath131 the intrinsic collective states can then be expressed as @xmath132 eq . ( [ sdcore ] ) provides the microscopic meaning of the quantum numbers @xmath133 , @xmath134 and @xmath135 in a very clear manner : the phonon number appearing in phenomenological models is nothing but the total number of @xmath9-pairs @xmath136 the basis of the hsm can be constructed by adding qp - excitations on top of the collective intrinsic states ; the formalism is the same as that used to build the psm bases in eq.(1 ) , but the simple bcs vacuum @xmath124 in eq.(1 ) is now replaced by a more correlated one , @xmath137 .
the general expression of the hsm basis in the laboratory system can be written as @xmath138 where @xmath139 ( @xmath140 ) is the qp number of neutrons ( protons ) , and the indices @xmath20 and @xmath141 stand for the qp and the collective vibrational configurations , respectively .
the hsm basis ( [ newbasis ] ) contains both s.p .
( qp excitations ) and collective ( @xmath9-pair excitations ) degrees of freedom ; the deficiency of lack of collective degrees of freedom in the original psm is redeemed . moreover
, the basis ( [ newbasis ] ) is expected to work also for the transitional ( or weakly deformed ) nuclei , which is beyond the original psm territory .
this is expected because it is known from the fdsm that the collective states of transitional nuclei can be described as a mixture of different @xmath10 irreps @xcite .
the original psm uses only the deformed - bcs vacuum ( the ground state of the @xmath10 irreps ) , and thus does not contain such a mixing mechanism .
this is why the psm becomes less and less valid when going away from the well - deformed region .
the hsm basis ( [ newbasis ] ) now contains all possible @xmath10 irreps , since its labels ( @xmath142 , @xmath134 , @xmath135 ) are in one - to - one correspondence to that of the @xmath10 irreps @xmath81 @xcite . for spherical nuclei
, the basis should be constructed separately since in the spherical case , the rotational symmetry is restored so that no distinction can be made between the intrinsic and the laboratory system . in the spherical limit ,
the bcs vacuum is just the @xmath143 ground state , which , in the fdsm , corresponds to the intrinsic state with @xmath144 ( see table ii ) .
let us denote this bcs vacuum as @xmath145 to distinguish it from the deformed one .
the deformed mean - field ( _ e.g. _ the nilsson scheme ) in the spherical case is also reduced to the spherical shell model level scheme so that a qp state can be labeled by \{@xmath146}. using \{@xmath147 } for short , a qp state can be expressed as @xmath148 ( @xmath149 ) , which is no longer a superposition of all possible angular momentum states but has a definite angular momentum .
likewise , a @xmath9-pair excited state @xmath150 is now a @xmath151 state .
therefore , there is no need for performing angular - momentum projection .
the basis with a given total angular - momentum @xmath152 can be directly obtained by the angular - momentum coupling .
the general expression of the hsm basis in the spherical case can be written as follows : @xmath153^i_m\left|\phi_0\right\rangle \label{sbasis } \end{aligned}\ ] ] with @xmath154^{j_q}_{m_q } \label{aq}\vspace{-12pt}\ ] ] @xmath155$}^{j^\sigma_q}_{m^\sigma_q } \hspace{-4pt}=\hspace{-4pt}\sum_{[m]}c_{[m]}({j_{q}^\sigma m_q^\sigma } ) \prod_{i=1}^{n^\sigma_q}a^\dagger_{j_i m_i}\hspace{4pt } & & \label{aqm}\\ { \bf d}^{\dagger}_{r_cm_c}(n_d)\hspace{-2pt}\equiv\hspace{-2pt } \mbox{\large$[$}\prod_{k=1}^{n_d}d^\dagger\mbox{\large$]$}^{r_c}_{m_c } \hspace{-2pt}=\hspace{-2pt } \sum_{[\mu]}c_{[\mu]}(r_cm_c ) \prod_{k=1}^{n_d}d^\dagger_{\mu_k}\hspace{6pt } & & \label{aqac } \end{aligned}\ ] ] where @xmath156 and @xmath157 are the creation operators of @xmath158 qp s and @xmath159 @xmath9-pairs coupled to angular - momenta @xmath160 and @xmath161 , respectively .
the short - hand notation @xmath162 $ ] represents the configuration @xmath163 of the qp s .
similarly , @xmath164 $ ] is for the @xmath9-pairs .
* the first column lists the fdsm @xmath10 intrinsic states .
the second column is the hsm collective basis for the deformed nuclei , which is obtained by replacing @xmath165 with the bcs vacuum @xmath124 . after projection , each intrinsic state produces a rotational band labeled as @xmath0 , @xmath49 , and @xmath166 listed in the third column .
the fdsm @xmath79 states and the hsm collective basis for spherical nuclei are listed in the 4th and 5th column , respectively .
the last column lists their corresponding spin - parity and excitation energies relative to the bcs vacuum .
note that the @xmath9-pairs in the @xmath79 states have been modified to commute with @xmath112 @xcite .
the amplitudes @xmath167}({j_{q}^\sigma m_q^\sigma})$ ] in eq.([aqm ] ) can be easily calculated by using the standard shell model technique . as a matter of fact
, all spherical nuclei lie very close to the doubly closed shell .
therefore , only a few qp s from the @xmath105-orbits around the fermi surfaces need to be considered .
the amplitudes @xmath168}(r_cm_c)$ ] in eq.([aqac ] ) are difficult to obtain in this way because multi - major shells are involved due to the @xmath9-pair collectivity .
however , they can be easily evaluated through the @xmath27-boson cfp s ( coefficients of fractional parentage ) if the @xmath9-pairs are regarded as bosons .
it should be noted that this is not a boson approximation since the @xmath169 s in eq .
( [ aqac ] ) remain to be fermionic operators .
it is just a symmetry detected truncation to select only those states that are symmetric with respect to interchanging any @xmath9-pairs .
a comparison between the deformed basis and the spherical basis is shown in table ii .
it is interesting to see that the @xmath9-pair excited states , which , in the deformed case , produce the @xmath0-@xmath49 bands after angular - momentum projection , degenerate in the spherical case into a spherical vibration - spectrum .
the spherical vibration - spectrum has much less independent states than the rotational spectrum .
it is so simply because when @xmath144 , most of the states obtained from the angular - momentum projection are not linearly independent due to the spherical symmetry nature .
they are highly over - complete .
for instance , the @xmath0- and @xmath49-bands reduce to a @xmath151 state when @xmath144 .
this is why we should construct the spherical shell - model basis differently .
it can be seen that even without configuration mixing , this collective basis ( as shown in table ii ) can already give the essential features of the low - lying collective modes found in the spherical nuclei .
we therefore feel confident that the hsm basis is going to work well .
of course , having built a shell - model basis is not enough .
the central theme that follows is whether it is possible to find an efficient truncation scheme within the constructed basis so that the calculations for heavy nuclei become feasible . to estimate the size of the deformed intrinsic basis ( [ newbasis ] ) , let us denote the maximum qp number , the number of selected s.p . states , and the maximum number of the excited @xmath9-pairs as @xmath170 , @xmath171 ( @xmath149 ) , and @xmath172 , respectively .
the total dimension is the product of the dimension of neutron qp states , that of proton qp states , and that of the @xmath9 pairs @xmath173 for each of these three terms , it can be shown that @xmath174 where @xmath175 , with @xmath176 or @xmath177 , are the dimensions of the qp states for neutrons or protons , @xmath178 is the dimension of the intrinsic collective states , with @xmath179 when @xmath180 even and @xmath181 when @xmath180 odd . in eq.([dim
] ) , @xmath178 is for separated neutron and proton vacuum @xcite . if we treat the neutron and proton vacuum as a single coupled bcs vacuum , the double - summation in eq.([dim ] ) is reduced to a single one and this can be easily summed up : @xmath182 , \label{dimc}\ ] ] with @xmath1831 or 0 depending on @xmath184 even or odd . according to the experience of the psm calculations , it is sufficient in most cases to take @xmath185 and @xmath186 for both neutrons and protons .
one can then obtain that the dimension of the qp states @xmath187 .
if eq.([dimc ] ) is used for estimating dimension of the collective states , @xmath188 changes from 1 , 3 , 7 , 13 to 161 when the maximum excited @xmath9-pair number @xmath172 varies from 0 , 1 , 2 , 3 , @xmath17 , to 10 .
thus , the total dimension @xmath189 ranges from 49 , 147 , 343 , 637 , @xmath17 , to 7889 . for well - deformed nuclei ,
the @xmath9-pair excitation energy is about 1 mev ( _ i.e. _ the band head of the first @xmath0-@xmath49 band )
. states with more @xmath9-pairs are excited higher in energy . furthermore , well - deformed nuclei are very close to an @xmath10 rotor .
states with different @xmath159 , which corresponds to different @xmath10 irreps , hardly mix with each other although the mixing may not be zero since the symmetry is not perfect .
thus , for the dominant low - lying collective states the number of the excited @xmath9-pairs , @xmath159 , should not be large .
we expect that taking @xmath184 2 to 4 is already good enough to produce all the low - lying collective vibrational states .
thus , the total dimension @xmath189 is of the order of @xmath190 to @xmath191 for well - deformed nuclei . for weakly - deformed nuclei , mixing between different @xmath10 irreps must be stronger , and therefore , it requires a larger @xmath172 to account for such a mixing .
this can increase the basis dimension drastically .
however , transition from the deformed to the spherical region behaves like a phase transition @xcite , which happens suddenly within a small interval of nucleon number .
we therefore expect that as we pass through the weakly - deformed region , nuclei may quickly enter into the spherical region before the @xmath172 number becomes too big ( e.g. @xmath192 ) .
thus , calculations for weakly - deformed nuclei , though harder , are tractable .
when including the scissors mode in the calculation , we can use eq .
( [ dim ] ) for estimating the basis size .
the maximum dimension in this case would rise by one to two orders of magnitude .
namely , the dimension would become 49 , 294 , 1127 , 2646 , @xmath17 , 2@xmath193 for @xmath172 = 0 , 1 , 2 , 3 , @xmath17 , 10 .
moreover , the dimension should be multiplied by a factor of @xmath194 , since , after the neutron and proton states have been built with good angular momentum , there are many ways of coupling these neutron and proton states to a given total angular momentum @xmath152 .
it can be shown that for a given total angular momentum @xmath195 , @xmath196 where @xmath197 is the cut - off angular momentum ( _ i.e._the highest angular momentum that are considered for neutrons and protons ) .
this factor is of an order of @xmath190 for @xmath198 .
thus , the total dimension @xmath189 would be of the order of @xmath199 to @xmath200 for @xmath201 to 4 . however , given the fact that different collective modes usually have different energy scale and different symmetries , there is still room for a further reduction in dimension .
it is known that the excitation energy of the low - lying scissors mode in the well deformed region is about 2 mev higher than the low - lying @xmath0-@xmath49 bands .
furthermore , as shown in refs .
@xcite , the scissors mode is physically caused by the relative motion of separated n- and p - vacuum ( with @xmath202 ) , not by the @xmath9-pair excitation .
therefore , the interplay between these different collective modes should be small and may thus be studied separately .
one can thus set @xmath202 when studying the scissors vibrations .
if we include qp excitation states up to 4-qp and no @xmath9-pair excitations in the basis , the dimension can be reduced to about 5@xmath203 . for spherical nuclei
, the dimension should be estimated differently .
what eqs .
( [ dim ] ) and ( [ dimc ] ) show us is the total @xmath204-scheme dimension , @xmath189 , in the intrinsic frame .
what we are actually interested in is the dimension @xmath205 at a given angular momentum @xmath152 . in the deformed case
, @xmath205 is equal to @xmath189 because after projection , every intrinsic state can in principle contribute one state to an @xmath152 . in the spherical case
, @xmath205 is smaller than @xmath189 since the basis is now constructed by angular - momentum coupling in the laboratory frame .
eq.([totaldim ] ) may be utilized to estimate the @xmath204-scheme dimension , but the formula for counting the collective basis states in the spherical case should be replaced by @xmath206 this formula is obtained because the collective basis states are constructed differently for the spherical case , and we have imposed a constraint that only those basis states that are symmetric with respect to the interchange of @xmath9-pairs are selected . according to eq.([dimboson ] ) , for @xmath207 10 , the dimension @xmath178 is respectively 1 , 6 , 21 , 56 , @xmath208 3003 , which are considerably larger than the dimensions in the deformed case .
an estimation for the maximum dimension of @xmath205 , denoted as @xmath209 , can be made by using the relation @xmath210 : @xmath211 where @xmath212 ( @xmath213 ) are the maximum angular momentum that the neutron qp s , the proton qp s , and the excited @xmath9-pairs can reach .
thus , knowing the @xmath204-scheme dimension @xmath189 one can estimate the maximum dimension @xmath209 .
the reduction factor @xmath214 ^ 2 $ ] makes @xmath209 two to three orders of magnitude smaller than the @xmath204-scheme @xmath189 . on the surface
it seems that the basis for the spherical case is smaller because of the reduction factor @xmath215 in eq.([dimmax ] ) , it is in fact not true .
the problem lies in the fact that in the spherical case , the density of s.p .
states around the fermi surfaces is much larger than that in the deformed case since each @xmath105 has a @xmath216 degeneracy .
the number of qp states @xmath217 must be around 30 instead of 4 adopted in the deformed case .
this increases the @xmath204-scheme dimension on the qp sector enormously . keeping @xmath185
, the dimension of qp basis will increase from 49 in the deformed case ( @xmath218 ) to an order of @xmath199 , leaving almost no room for collective basis .
this is another reason why we can not continue to use @xmath204-scheme basis for the spherical case . using @xmath152-scheme
only @xmath219 should be concerned , which greatly reduces the dimension by a factor of @xmath215 .
the total basis dimension varies from 1125 , 5879 , 17742 , @xmath220 , @xmath17 , to @xmath221 for @xmath222 to 10 , when @xmath223 is assumed .
it is known that the low - lying spherical vibrational energy @xmath224 is about 0.5 mev , which is roughly the @xmath9-pair excitation energy in spherical nuclei .
the s.p .
excitations start to have influence at about @xmath225 and become significant around @xmath226 , where the characteristics of vibrational spectrum are strongly disturbed . thus on the collective sector , taking @xmath201 ( @xmath226 excitation energy ) may be a reasonable choice . if the number of qp s is kept to be @xmath227 ( @xmath228 ) to give the maximum 4-qp states , the dimension one has to deal with is of the order of @xmath229 .
in fact , spherical nuclei lie often near the doubly closed shells and have large binding energies . the 4-qp states may already be rather high and have only little influence to the low - lying states .
if we ignore the 4-qp states , then the maximum shell - model dimension can be reduced to 488 for @xmath201 .
even for @xmath230 the maximum dimension is only 908 .
such a basis is clearly manageable .
the hsm hamiltonian can be generally written in the following form @xmath231 this rotational invariant hamiltonian is the same as that in the psm @xcite ( see also eq.([h ] ) ) , except that the multipole interactions are now extended to include not only the quadrupole but also monopole , octupole , and hexadecupole terms .
all these operators are defined in the multi - major - shell space : @xmath232 the octupole and hexadecupole interactions have been employed in their hamiltonian by chen and gao @xcite in dealing with the actinide nuclei with projection .
this type of schematic interactions ( [ hext ] ) works for the structure calculations surprisingly well despite its simplicity .
in fact , it has been shown by dufour and zuker @xcite that these interactions simulate the essence of the most important correlations in nuclei , so that even the realistic force has to contain at least these basic components implicitly in order to work successfully in the structure calculations .
therefore , we find no compelling reasons for not using these simple interactions .
of course , the model is open to adoption of any realistic forces .
the s.p .
energy term @xmath233 can be simply taken from the nilsson scheme at zero deformation although other schemes such as the woods - saxon may also be adopted if there is an advantage .
the parameterization in the nilsson scheme has been well established for the @xmath0-stable mass regions ( see , for example , ref .
@xcite ) . for the exotic mass regions ,
the standard nilsson parameters may not be valid and improvement may be necessary @xcite . in this
regard , new experimental data that can provide information on the single - particle energies in the exotic mass regions are very much desired . the relativistic mean field theory ( rmf ) @xcite could also be helpful in providing a s.p .
energy scheme for the exotic mass regions where no data are available for determining a phenomenological mean field .
the rmf may have a better extrapolation power than other phenomenological mean fields because a single set of parameters of the rmf is able to fit nuclear ground state properties from light to heavy nuclei reasonably well . in practice
, it may be a convenient approach to use the rmf s.p .
energies as a reference to adjust the nilsson parameters in the hsm . the parameters @xmath234 @xmath235 in eq .
( [ hext ] ) are determined by the self - consistent relation with deformation parameters , in the same way as in the psm @xcite .
the monopole - monopole interactions ( @xmath236=@xmath237 ) includes three terms : @xmath238 ( @xmath149 ) and @xmath239 .
it has been shown that the monopole - monopole interactions are important for the nuclear mass calculations , in particular for those nuclei lying far from the stability line @xcite .
they can be viewed as an average way of accounting for the n - z dependence in s.p .
however , for a given nucleus , the monopole - monopole interactions contribute a constant energy only .
thus , if we are not interested in calculating the absolute energies , these terms can be ignored .
the strengths of monopole - pairing , @xmath240 , and quadrupole - pairing , @xmath241 , depend on the size of single - particle space .
they are inversely proportional to the mass number @xmath38 . for the well - deformed mass regions where
calculations have been extensively performed by the psm , the interaction strengths proposed for the hsm should be similar if the same size of single - particle space is employed . however , when the hsm is applied to other mass regions such as the transitional or spherical region , these strengths may need to be readjusted .
in particular , the self - consistent relation used for determining the multipole interaction strengths will break down when the basis deformation becomes zero .
the strengths for these cases have to be studied separately .
after all , eq.([hext ] ) is an effective hamiltonian in nature , and is subject to vary when the model is applied to different mass regions . nevertheless , a smooth variation in parameterization is expected since the model space is sufficiently large .
having had a tractable basis and a reliable effective hamiltonian , the remaining task is to diagonalize the hamiltonian in the basis to get the eigen - energies and eigen - functions , and then to use the obtained wave functions to calculate the observables .
to do so , one must know how to evaluate the matrix elements in the projected basis .
the projection techniques have already been well developed by the psm based on the pioneering work of hara and iwasaki @xcite .
the extensive discussion about the details of the projection techniques can be found in the psm review article @xcite and the psm computer code @xcite .
here we emphasize only on how to deal with the new ingredient : the @xmath9-pairs and the spherical basis .
let us first discuss the deformed case .
suppose that a one - body tenser operator of @xmath236-rank , @xmath242 , and the eigen - functions @xmath243 are given in the laboratory frame : @xmath244 where @xmath245 is known .
@xmath246 is the intrinsic basis defined in eq.([newbasis ] ) , and the amplitudes @xmath247 are obtained by solving the eigen - value equation , eq.([sheq ] ) .
although we take the tenser operator in eq .
( [ wave ] ) as example , the following discussion applies equally well to the pairing operators if we change @xmath248 in eq .
( [ wave ] ) to @xmath249 or @xmath250 and rewrite @xmath251 or @xmath252 . sandwiched by the wave functions , the matrix element of @xmath242 in the laboratory frame can be expressed as @xmath253 by applying @xmath254=\delta n \hat{t}_{\lambda\mu}$ ] where @xmath255 ( @xmath256 ) for the electromagnetic multipole ( pairing ) operator , it can be shown that @xmath257 inserting eq .
( [ ptp ] ) into eq.([tlm ] ) , we obtain @xmath258 with @xmath259 and @xmath260 here , @xmath261 is a rotation operator in the 4-dimensional ( coordinate plus particle - number ) space . the problem is then reduced to evaluate the matrix elements having the general form @xmath262 .
these are the matrix elements for a one - body tenser operator between the intrinsic state @xmath263 , constructed in ( [ newbasis ] ) , and the rotated one @xmath264 . in order to evaluate the matrix elements , it is convenient to transform the operators @xmath45 ( @xmath109 ) , contained in @xmath242 and in the @xmath9-pairs ( which are embedded in the basis @xmath246 ) , into the qp representation @xmath265 .
the transformation matrices ( of the hartree - fock - bogoliubov type ) are known : @xmath266 where @xmath267 and @xmath268 are the occupation amplitudes in the bcs , and @xmath269 s the nilsson wave functions of the deformed s.p.level @xmath270 .
the evaluation of @xmath271 in eq.([qpmtrix ] ) is equivalent to evaluate the following general expectation quantity with respect to the bcs vacuum state @xmath124 : @xmath272 after applying the transformation from the spherical to the deformed - bcs basis ( eq . ( [ uv ] ) ) , the operators appearing in the above expression , such as @xmath242 and @xmath273 , are generally linear combinations of products of the qp operators @xmath274 and @xmath275 .
the problem eventually becomes the one of performing contractions for a series of qp creation and annihilation operators @xmath276 in performing the contraction calculations , a generalized wick - theorem is used .
the techniques of carrying out the contractions with the rotation operator @xmath23 are available .
it has been shown @xcite that the problem can be reduced th evaluation the following three basic elements @xmath277 a^\dagger_{\nu'}\rangle , \hspace*{4pt } b_{\nu\nu'}\equiv\langle a_\nu a_{\nu ' } [ \omega ] \rangle , \hspace*{4pt } c_{\nu\nu'}\equiv\langle [ \omega ] a^\dagger_{\nu } a^\dagger_{\nu'}\rangle , \label{abc}\ ] ] with @xmath278 \equiv \hat{r}/ \langle\hat{r } \rangle$ ] , and @xmath279 is the short - hand notation of vaccum expectation , for example , @xmath280 .
the contraction can then be calculated by the following theorem @xmath281 a^\dagger_1\cdots a^\dagger_n\rangle= \nonumber \\ & & \hspace{.2in}\sum_{k = m}^{[\frac n2]}\sum_p(\pm)(b)^{k-(n - n')/2}(c)^{n-2k}(a)^k \\ & & ( n+n'=\mbox{even , } m = max\{0,\frac{n - n'}2\ } ) \nonumber \label{abc2 } \end{aligned}\ ] ] where @xmath282 is a permuted sum " with all possible combinations of pairs of indices , and @xmath283 is the permutation parity .
details for these calculations can be found in the psm review article @xcite . for the spherical case , the matrix elements can be evaluated in the same way .
the only difference is that there is no need for doing angular - momentum projection since the basis @xmath284 defined in eq.([sbasis ] ) is constructed in the laboratory frame .
therefore , in the spherical case , the projection operator @xmath285 should be replaced by @xmath286 . as a consequence , eqs .
( [ reduce ] ) and ( [ qpmtrix ] ) become @xmath287 and @xmath288 after transforming @xmath289 and the @xmath9-pairs embedded in @xmath290 and @xmath284 into the qp basis , the evaluation of matrix element @xmath291 is again reduced to the contraction calculations like the type of eq.([contraction ] ) , except that in this case , @xmath23 is replaced by @xmath292 .
in this paper , a multi - shell shell model for heavy nuclei is proposed . for performing a shell model diagonalization involving several major shells in the model space ,
we seek an efficient truncation scheme .
the new heavy shell model can be viewed as an integration of two existing models : the project shell model and the fermion dynamical symmetry model .
the psm is an efficient method for the high - spin description of rotational states built upon qp - excitations , but it is not a practical method for the low - spin collective vibrations . in contrast , the fdsm provides a well - defined truncation scheme for all known types of low - lying collective vibrations , workable from the spherical to the well - deformed region , but it is lack of the necessary degrees of freedom of single particle excitations . the idea proposed in the present paper is to combine the advantages of both models . to construct the shell model basis , we follow the fdsm discovery that the intrinsic collective states can be built by applying @xmath9-pairs onto the bcs vacuum state , and employ the psm qp truncation scheme combined with the projection techniques . in this sense
the model goes beyond the traditional one - major - shell shell model , yet the calculation is tractable for heavy , and even for superheavy nuclei .
given the past success of the psm and the fdsm in their own applicable regimes , which has been documented in the literature , we expect that the new model can work reasonably well for heavy nuclear systems where traditional shell - model calculations are not feasible
. this model should be capable of describing the low - excitation collective vibrations of all known types , including fragmentations due to the quasi - particle mixing .
it should be capable of applying to the high - spin regions where quasi - particle alignments play an important role .
it should also be capable of treating weakly - deformed nuclei across the transitional to the spherical region .
we conclude that the development of the heavy shell model may open possibilities of shell model calculations for heavy nuclei to a much wider range of nuclear structure problems .
research at the university of notre dame is supported by the nsf under contract number phy-0140324 .
research at physics division of national center for theoretical sciences is supported by national science council in taiwan .
ur , d. bucurescu , s.m .
lenzi , g. martinez - pinedo , d.r .
napoli , d. bazzacco , f. brandolini , d.m .
brink , j.a .
cameron , e. caurier , g. de angelis , m. de poli , a. gadea , s. lunardi , n. marginean , m.a .
nagarajan , p. pavan , c. rossi alvarez , and c.e .
svensson phys .
c * 58 * , 3163 ( 1998 ) . the abnormal - parity ( intruder ) level in each major shell is also taken into account in the fdsm , but treated separately . in the context of this paper , we shall ignore it for simplicity , since the collective excitations are mainly determined by the valence nucleons in the normal - parity levels .
the detail discussion about the role played by the intruder level can be found in the review article of the fdsm @xcite . | performing a shell model calculation for heavy nuclei has been a long - standing problem in nuclear physics . here
we propose one possible solution .
the central idea of this proposal is to take the advantages of two existing models , the projected shell model ( psm ) and the fermion dynamical symmetry model ( fdsm ) , to construct a multi - shell shell model .
the psm is an efficient method of coupling quasi - particle excitations to the high - spin rotational motion , whereas the fdsm contains a successful truncation scheme for the low - spin collective modes from the spherical to the well - deformed region .
the new shell model is expected to describe simultaneously the single - particle and the low - lying collective excitations of all known types , yet keeping the model space tractable even for the heaviest nuclear systems . |
let @xmath4 be a finite sequence of positive numbers such that @xmath5 this sequence determines the corner cantor set @xmath6 of generation @xmath7 in @xmath8 , such that the @xmath9-th generation consists of @xmath10 cubes of edge length @xmath11 , each of these cubes contains @xmath12 corner cubes of the @xmath13-th generation , and so on . for brevity , we will call @xmath6 `` a cantor set '' instead of `` a cantor set of generation @xmath7 '' .
there is a number of papers on estimates of various capacities , norms of integral transforms and operators , etc .
, on such cantor sets .
these estimates demonstrate the sharpness of various inequalities where the bounds are attained on cantor sets ; they are also of independent interest . but besides the necessary condition , there are certain additional conditions on @xmath11 in many cases . in the present paper we associate with given numbers @xmath11 satisfying _ only _ the condition , the `` regularized '' sequence @xmath14 such that @xmath15 , @xmath16 , and construct the ( non - corner ) cantor set @xmath6 formed by @xmath17 cubes of edge length @xmath18 .
since the corner and non - corner cantor sets have similar structure , it is unimportant for applications which set to use . for a nonnegative finite borel measure @xmath19 in @xmath8 , @xmath20 , and @xmath21 , @xmath22 , define the @xmath23-truncated @xmath24-riesz transform of @xmath19 by @xmath25 where @xmath26 if the limit @xmath27 exists , we shall call it the @xmath24-riesz transform of @xmath19 at @xmath28 . to consider all finite borel measures and all points @xmath29 , one introduces the quantity that always makes sense , namely the so called maximal @xmath24-riesz transform @xmath30 ( note that @xmath31 and @xmath32 are vectors and @xmath33 is a number ) . besides @xmath34 and @xmath35 , we need the @xmath23-truncated @xmath24-riesz operator defined by @xmath36 for every @xmath22 , the operator @xmath37 is bounded on @xmath38 .
we set @xmath39 later on we denote by @xmath40 ( without indices ) positive constants which may vary from line to line .
let @xmath6 be the corner cantor set generated by a sequence @xmath4 , and consisting of @xmath17 cubes @xmath41 .
let @xmath42 the probability measure uniformly distributed on each cube @xmath41 with @xmath43 .
mateu and tolsa @xcite proved that if @xmath44 , @xmath45 , and @xmath46 with @xmath47 then @xmath48^{1/2}\le\pmb|\mathfrak{r}_{\mu}^s\pmb|\le c\biggl[\sum_{j=1}^n\t_j^2\biggr]^{1/2},\ ] ] where the constants @xmath49 depend only on @xmath50 and @xmath24 .
in fact , mateu and tolsa proved a stronger assertion than the estimate from below : for sufficiently small @xmath23 , @xmath51 this result was refined by tolsa in @xcite , where the condition about monotonicity of densities @xmath52 was dropped .
a more general class of cantor sets for @xmath53 ( again under the condition @xmath46 ) was considered in ( * ? ? ?
* theorem 3.1 ) .
the estimate from above in was also obtained in @xcite by another method .
the arguments in @xcite and especially in @xcite are rather complicated .
we give two independent proofs of for our `` regularized '' cantor set .
the first ( direct ) proof is considerably simpler than in @xcite , @xcite .
the second approach gives the desired inequality as a corollary of the following more delicate result .
we shall prove that the inequality @xmath54 ( and therefore the analogous estimate for @xmath55 ) holds on a `` big '' portion of @xmath6 .
we also consider the related problem in a more general setting and give certain applications .
in particular , we establish the two - sided estimate of the riesz capacity associated with @xmath56 .
this estimate is a refined version ( for non - corner cantor sets ) of the corresponding results in @xcite .
we conclude this section with the construction of `` regularized '' non - corner cantor sets .
let @xmath4 be a finite sequence of positive numbers satisfying , and let two parameters @xmath57 and @xmath58 be given .
( later on @xmath59 will depend on @xmath60 and @xmath24 . )
define the set @xmath61 , @xmath62 , of indices inductively in the following way : @xmath63 ; if @xmath64 , @xmath65 , and @xmath66 , then @xmath67 is the least @xmath68 , such that @xmath69 ; if @xmath64 , @xmath65 , and @xmath70 , then @xmath71 .
thus , @xmath72 we set @xmath73 clearly ( see ) , @xmath74{c } \s_j\le\ell_j<\a^{-1}\s_j,\ 1\le j
< n;\quad \a\s_n\le\ell_n\le\s_n;\\ [ .1 in ] \ell_{j_{p+1}}\le\a2^{-(j_{p+1}-j_p)}\ell_{j_p},\quad p=1,\dots , m-1 .
\end{array}\ ] ] hence , @xmath75 . for @xmath76 and @xmath77 let @xmath78 be the cube @xmath79 construct the cantor set @xmath6 recursively as follows .
for @xmath80 we set @xmath81 .
take @xmath12 closed corner cubes @xmath82 , @xmath83 , of edge length @xmath84 ( i.e. distinct cubes lying inside @xmath85 with edges parallel to the edges of @xmath85 , such that each cube @xmath82 contains a vertex of @xmath85 ) .
suppose that the cubes @xmath86 , @xmath87 , of edge length @xmath88 , @xmath89 , are already defined .
partition each cube @xmath86 into @xmath90 equal subcubes @xmath91 .
( in the figure above @xmath92 , @xmath93 . )
if @xmath94 , we may consider this partition as @xmath95 sequential partitions of @xmath86 , such that on @xmath9-th step , @xmath96 , we split each cube @xmath97 into @xmath12 cubes @xmath98 of edge length @xmath99 .
consider the cubes @xmath100 .
remark that by , @xmath101 take @xmath12 closed corner cubes @xmath102 of edge length @xmath103 in each @xmath104 .
we get @xmath105 cubes @xmath102 , and set @xmath106 for @xmath107 we obtain the desired set @xmath6 .
our first theorem shows that under certain assumptions @xmath3 is comparable with its average value on a set of `` big '' measure , and this property holds not only on cantor sets .
it means that the distribution of values of riesz transform is uniform in a certain sense .
set @xmath108 , and denote by @xmath109 the class of nonnegative borel measures @xmath110 in @xmath2 such that @xmath111 [ th21 ] suppose that @xmath112 , @xmath113 , @xmath114 exists @xmath110-a.e . ,
@xmath115 , and @xmath116 then for every @xmath117 we have @xmath118 on the other hand , obviously implies with @xmath119 instead of @xmath120 .
the analogous statements hold for @xmath121 .
we deduce theorem [ th21 ] in section 3 from a deep result by nazarov , treil and volberg ( theorem 1.1 in @xcite ) . in section 4
we obtain the following estimates for @xmath122 and @xmath123 . as before ,
we denote by @xmath42 the probability measure uniformly distributed on each cube @xmath41 with @xmath43 .
[ th22 ] let an integer @xmath20 and @xmath124 be given .
there are constants @xmath57 , @xmath58 , depending only on @xmath60 , @xmath24 , and such that for any positive numbers @xmath125 , satisfying with @xmath126 , and for the corresponding cantor set @xmath6 , @xmath127^{1/2}&\le\|r_{\mu}^s\|_{l^2(\mu)}\le c\biggl[\sum_{j=1}^n\t_j^2\biggr]^{1/2},\quad \t_j=\frac{2^{-dj}}{\ell_j^s},\label{f24}\\ c\biggl[\sum_{j=1}^n\t_j^2\biggr]^{1/2}&\le\pmb|\mathfrak{r}_{\mu}^s\pmb|\le c\biggl[\sum_{j=1}^n\t_j^2\biggr]^{1/2},\label{f25}\end{aligned}\ ] ] where the positive constants @xmath49 depend only on @xmath60 and @xmath24 . ( we use the same notation @xmath52 for values slightly different from the ones in
. clearly , the corresponding relations in both cases are equivalent . ) set @xmath128^{-1/2}\mu$ ] . if @xmath129 is small enough then @xmath112 , and @xmath115 by the upper bound in . moreover , the first inequality in implies with @xmath130 .
thus , theorem [ th21 ] immediately yields the inequality @xmath131 the existence of cantor - type sets satisfying was established in ( * ? ? ?
* section 7 ) using probabilistic arguments . but a concrete set was not presented .
the particular case @xmath92 , @xmath132 , @xmath133 , was considered in @xcite .
a more general class of plane corner cantor sets was treated in @xcite .
clearly , implies the estimates from below in , ( in fact , the estimates from above were obtained in @xcite see section 4 of the present paper for details ) . in section 5
we give a completely different proof of without the use of theorem 1.1 in @xcite and of theorem [ th22 ] .
this independent approach allows us to consider a more general class of measures and wider range of @xmath24 . in section 6
we consider the capacity @xmath134 of a compact set @xmath135 defined by the equality @xmath136 where @xmath137 is the class of positive radon measures supported on @xmath138 . for @xmath139
, @xmath140 is the analytic capacity @xmath141 , which is comparable with the analytic capacity @xmath142 by the remarkable result of tolsa @xcite . for @xmath143 ,
@xmath144 is comparable with the lipschitz harmonic capacity ( see @xcite and ( * ? ? ?
* section 10 ) for details and references ) .
[ th23 ] let @xmath20 , @xmath124 .
for any finite sequence of positive numbers @xmath125 , satisfying with @xmath126 , @xmath145^{-1/2}\le\g_{s,+}(e_n ) \le c\biggl[\sum_{j=1}^n\biggl(\frac{2^{-dj}}{\s_j^s}\biggr)^2\biggr]^{-1/2},\ ] ] where the positive constants @xmath49 and the parameters @xmath146 , @xmath147 of the corresponding cantor set @xmath6 depend only on @xmath60 and @xmath24 . for corner cantor sets and @xmath139 ( that is for analytic capacity @xmath141 ) ,
the estimates were obtained in @xcite under the additional assumption @xmath148 .
a different proof has been given in @xcite .
the corresponding inequalities for @xmath142 were proved in @xcite ( before the tolsa s result @xcite about comparability of @xmath142 and @xmath141 ) .
the case of the lipschitz harmonic capacity ( i. e. @xmath149 ) was treated in @xcite under the assumptions @xmath45 , and @xmath46 . thus , theorem [ th23 ] is a refined version of these results for regularized " cantor sets .
it is noted at the end of @xcite that holds under the same assumptions for the signed riesz capacity @xmath150 as well .
in fact , our proof of theorem [ th23 ] is a modification of the arguments in @xcite and @xcite , and these arguments also work for @xmath150 and for regularized " cantor sets under the additional assumption @xmath151 , but without monotonicity of @xmath152 .
it is known that @xmath153 for @xmath92 , @xmath132 @xcite , for @xmath154 , @xmath155 ( * ? ? ?
* theorem 1.1 ) , and for @xmath156 , @xmath53 @xcite . as far as we know , the validity of this relation in other cases is an open problem .
the extension to bilipschitz images of corner cantor sets from @xcite is given in @xcite .
in section 7 we give `` the limit case '' of theorem [ th23 ] , when the sequence @xmath157 is infinite .
we use the obtained estimates to demonstrate the sharpness of results in @xcite .
in particular , we consider the problem of comparison of the capacity @xmath144 and hausdorff content .
[ le31 ] if @xmath158 is a non - negative function , @xmath19 is an arbitrary probability measure , and @xmath159 then for @xmath160 we have @xmath161 clearly , @xmath162 assume that @xmath163 then @xmath164 since the left hand side is equal to the same number @xmath165 , we come to a contradiction . by ( * ? ? ?
* theorem 1.1 , p. 467468 ) , the uniform boundedness of the cut - off caldern - zygmund operators @xmath166 on @xmath167 implies the boundedness of @xmath147 and of the corresponding maximal singular operator on @xmath168 for every @xmath169 . applying this theorem for @xmath170 , @xmath171 , we get @xmath172 where the constants @xmath173 depend only on @xmath60 and @xmath24 ( the last statement follows from the proof of theorem 1.1 in @xcite ) .
lemma [ le31 ] with @xmath174 , @xmath175 , @xmath176 , @xmath177 , @xmath178 , yields , since @xmath179 .
the proof of the corresponding statement for @xmath121 is essentially the same .
we need some notation .
let @xmath180 , @xmath181 be the centers of @xmath86 and @xmath182 correspondingly , and let @xmath183 , @xmath184 be the cubes @xmath86 , @xmath182 , containing @xmath28 .
set @xmath185 obviously , @xmath186 [ le41 ] there exists @xmath187 , such that for any @xmath188 and @xmath189 we have @xmath190 with @xmath191 and @xmath192 depending only on @xmath193 .
we have @xmath194 by , the cubes @xmath195 in @xmath183 are separated .
hence , up to a constant , @xmath196 is majorized by the integral over the measure uniformly distributed on @xmath183 with density @xmath197 , that is @xmath198 suppose that @xmath199 ( i. e. @xmath200 ) .
we claim , that for sufficiently big @xmath147 , @xmath201 on the other hand , for @xmath147 big enough we have @xmath202 the lower bound in implies . obviously , @xmath203 whenever @xmath204 and @xmath205 are symmetric with respect to @xmath180 .
hence , for `` half '' of the points @xmath199 , the angle between the vectors @xmath206 and @xmath196 is less then or equal to @xmath207 . for these @xmath28
we have @xmath208 we get the lower bound in .
the upper bound follows directly from and .
[ le42 ] let @xmath209 . then @xmath210 where @xmath192 depends only on @xmath60 and @xmath24 . by symmetry , @xmath211 suppose that @xmath212 .
we have @xmath213\,d\mu(x ) .
\end{split}\ ] ] by , , @xmath214 , we get @xmath215 as in lemma [ le41 ] , we represent the last integral as the sum of the integrals over @xmath216 and @xmath217 .
the second integral is estimated exactly as in .
the first integral , as before , is majorized by the integral with uniformly distributed measure .
thus , the last bound does not exceed @xmath218 < c'\frac{t\ell_{j_p}}{\ell_{j_{q+1}}^{1-\d}}\cdot\frac{2^{-dj_q}}{\ell_{j_q}^{s+\d}}\\ = c't\bigg(\frac{\ell_{j_p}}{\ell_{j_{q+1}}}\bigg)^{1-\d}\bigg(\frac{\ell_{j_p}}{\ell_{j_{q}}}\bigg)^{\d } \frac{2^{-dj_q}}{\ell_{j_q}^{s } } \le c't\a^{\d(p - q)}2^{-\d(j_p - j_q)}\,\frac{2^{-dj_q}}{\ell_{j_q}^s},\quad c'=c'(d , s)\end{gathered}\ ] ] ( in the last inequality we used the obvious relation @xmath219 and ) . since @xmath220 , we obtain the inequality @xmath221 this inequality together with , , and the obvious relation @xmath222 , imply .
set @xmath223 we start from the lower bound in . obviously , @xmath224
^ 2\,d\mu(x)\\ & = \sum_{p=1}^{m}\int_{e_n}\xi_p(x)^2\,d\mu(x)+\sum_{p\ne q}\int_{e_n}\xi_p(x)\xi_q(x)\,d\mu(x)=:\sigma_1+\sigma_2.\end{aligned}\ ] ] from we have @xmath225 enumerate @xmath226 in decreasing order : @xmath227 . from
we derive the estimate @xmath228\\ & < 4ct\a^{\d}\bigg[\sum_{i=1}^{\infty}2^{-\d i}\bigg]\sum_{p=1}^m\t_{p}^2\ , .
\end{split}\ ] ] we can choose @xmath146 and @xmath147 in such a way that the constant @xmath229 in is at least twice as big as the constant before the last sum in .
we have @xmath230 to get the lower bound in , it remains to note that @xmath231 obviously , the estimate from below obtained in implies the lower bound in . to complete the proof of theorem [ th22 ] , it is enough to get the upper bound in .
but the proof of this estimate is literally the same as the proof of the corresponding estimate for corner cantor sets in ( * ? ? ?
* corollary 3.5 ) .
in this section we develop an independent approach to obtaining the estimate , as well as its generalizations and related results .
let a finite sequence @xmath125 , @xmath233 , @xmath234 , and constants @xmath57 , @xmath58 , @xmath235 , be given .
for the corresponding cantor set @xmath6 and for a positive measure @xmath19 supported on @xmath6 , set @xmath236 where @xmath237 is the cube @xmath41 containing @xmath28 .
define the sets @xmath238 , @xmath239 of cubes by the relations @xmath240^{1/2}\bigg\},\quad \t_j=\frac{2^{-dj}}{\ell_j^s},\label{f51}\\ \tilde\ee&=\bigg\{e_{n , k}:|\tilde r_\nu^s(x_{n , k})|>\frac{c_0}{t^s}\biggl[\sum_{j=1}^n\t_j^2\biggr]^{1/2}\bigg\}\label{f52}.\end{aligned}\ ] ] here as before , @xmath241 is the center of @xmath41 ; in we assume that the values @xmath242 exist .
[ th51 ] for every @xmath243 and every integer @xmath20 , there exist constants @xmath244 , @xmath245 , depending only on @xmath193 , with the following properties .
fix some @xmath246 , @xmath247 , and a sequence @xmath125 , @xmath233 , @xmath234 .
let @xmath19 be a measure supported on the corresponding cantor set @xmath6 , equally ( but not necessarily uniformly ) distributed on each cube @xmath41 , @xmath248 , with @xmath249 .
then the number of cubes @xmath41 in @xmath239 is comparable with the number of all cubes in @xmath6 , that is , @xmath250 where @xmath251 is an absolute constant . moreover , if values @xmath242 exist , then @xmath252 the same conclusion holds if we replace @xmath241 in , by any fixed points @xmath253 , such that @xmath254 , @xmath255 .
this theorem implies a useful corollary which will be given after the proof .
the proofs of , , and of the statement for points @xmath256 , are the same . for definiteness , we consider the case .
[ le52 ] for every finite sequence @xmath257 of positive numbers and for given @xmath258 , there is a subsequence @xmath259 , such that @xmath260 where @xmath191 depends only on @xmath261 and @xmath262 .
let the maximal index @xmath263 be the least integer for which @xmath264 ( the value of @xmath265 will be determined by the construction in the sequel ) .
suppose that indices @xmath266 are already defined , and @xmath267 .
then @xmath268 is the least integer satisfying the relation @xmath269 this equality implies that @xmath270 in particular , we get .
clearly , @xmath271 .
using for @xmath272 with @xmath273 , we get the following estimate : @xmath274 which yields .
we use lemma [ le52 ] with @xmath275 , @xmath276 , @xmath277 , in order to extract the future subsequence from @xmath278 ( the numbers @xmath279 are defined by ) . by lemma [ le52 ]
, there exists the set @xmath280 of indices such that @xmath281 where @xmath191 is an absolute constant .
set @xmath282 .
each cube @xmath41 , @xmath248 , can be represented in the form @xmath283 where @xmath284 is the projection of @xmath41 onto the hyperplane @xmath285 , and @xmath286 is the projection of @xmath41 onto the @xmath287-axis . for @xmath288
fixed , we associate the vector @xmath289 with each cube @xmath290 in the following way . the choice of an interval @xmath286 can be viewed as a result of @xmath7 subsequent choices ( steps ) : starting from the interval @xmath291 $ ] , at @xmath9-th step , @xmath292 , we choose the left or the right of two equal subintervals of length @xmath99 in the preceding interval .
we set @xmath293 , if we choose the left subinterval at @xmath9-th step , and @xmath294 in the opposite case .
thus , we get the one - to - one correspondence between the cubes @xmath290 with fixed @xmath288 , and vectors @xmath295 .
[ le53 ] let @xmath246 , @xmath247 , and let the cubes @xmath296 , @xmath297 , be such that @xmath298 suppose that a measure @xmath19 satisfies the conditions of theorem [ th51 ] .
then @xmath299 where @xmath300 is the first component of a vector @xmath32 , and the constants @xmath245 , @xmath301 depend only on @xmath193 .
in order to simplify notation , we set @xmath302 , @xmath303 ( see the figure in section 1 , where a possible location of @xmath120 , @xmath304 is indicated ) . as before ,
we denote by @xmath305 , @xmath306 the corresponding cubes containing @xmath28 .
obviously , @xmath307 hence , @xmath308 ( for @xmath80 the last sums are absent ) .
if @xmath147 is greater than certain @xmath309 then @xmath310 we claim that @xmath311 where @xmath192 is an absolute constant
. indeed , by the the left hand side of does not exceed @xmath312 the inequality yields the estimate @xmath313 continuing the estimation , we get : @xmath314 since @xmath315 in the integrals from @xmath316 and @xmath317 , we have @xmath318 thus , @xmath319 as in the proof of lemma [ le41 ] , we majorize the last integrals by the integrals over the measures uniformly distributed on @xmath320 with densities @xmath321 . we get @xmath322 we consider three cases .
* case 1 . * @xmath323 .
then @xmath324 for @xmath325^{-1}$ ] , where @xmath301 is the constant in .
* case 2 . * @xmath53 . by
and we have @xmath326 if @xmath327^{-1}$ ] .
* @xmath328 .
again by , @xmath329 as concerns @xmath317 , we have @xmath330 the estimates for @xmath316 , @xmath317 and yield .
the next lemma is a particular case of lemma 7.3 in @xcite .
[ le54 ] let @xmath331 , be positive numbers .
let @xmath332 , be independent random variables satisfying @xmath333 such that @xmath334 we need the following lemma 10.2 in @xcite .
[ le55 ] let positive numbers @xmath331 , be given
. for each vector @xmath335 , @xmath336 , with components equal to plus or minus ones we set @xmath337 let @xmath338 be a set of different vectors @xmath339 such that @xmath340 .
then there exists a set @xmath341 with the following properties : \1 ) all the vectors in @xmath342 are distinct ; \2 ) @xmath343 ; \3 ) each vector @xmath344 can be obtained from some vector @xmath345 by replacing some negative components of @xmath346 ( depending on @xmath347 ) by positive components while keeping all the positive components of @xmath346 .
( different vectors @xmath347 can be obtained from the same @xmath346 . )
set @xmath348 , @xmath349 , @xmath350 , @xmath351 .
fix an integer @xmath288 and the components @xmath352 of vectors @xmath289 .
let @xmath353 be the set of cubes @xmath290 with fixed parameters indicated above .
clearly , @xmath353 consists of @xmath354 elements depending only on the choice of the components @xmath355 with @xmath356 .
we can consider this sampling as a choice of a vector @xmath339 .
we introduce the following subsets of @xmath353 ( or equivalently , the subsets of vectors @xmath339 ) : @xmath357 where @xmath358 is the constant in lemma [ le54 ] with @xmath359 , and @xmath301 is the constant in .
we shall consider the choice of a vector @xmath339 , @xmath336 , as a random event with the same probability @xmath360 for all vectors .
this defines a random variable @xmath361 with values @xmath362 .
obviously , we can interpret @xmath361 as the sum of independent random variables @xmath363 , @xmath364 , taking the values @xmath365 and @xmath366 with probability 1/2 .
we now use lemma [ le54 ] . in our case , @xmath367 , @xmath359 . since distribution of @xmath361 is symmetric , yields the inequality @xmath368 hence , @xmath369 consider now the set @xmath370 . by @xmath338
we denote also the set of the corresponding vectors @xmath339 . if @xmath371 , then @xmath372 , and we immediately arrive at the estimates below .
assume that @xmath373 , and let @xmath342 be the set of vectors corresponding to the set @xmath338 by lemma [ le55 ] .
consider an arbitrary vector @xmath344 .
it can be obtained from some @xmath345 by the replacement of some negative components of @xmath346 by positive ones .
let @xmath374 be the set of indices of the replaced components of @xmath346 .
let @xmath375 , @xmath376 , and let @xmath41 , @xmath377 be the cubes associated with @xmath339 and @xmath378 correspondingly . since @xmath379 , it follows that @xmath380 on the other hand , applying lemma [ le53 ] @xmath381 times ( successively for each component in @xmath374 ) we get @xmath382 so that @xmath383 hence @xmath384 . moreover , @xmath385 , since @xmath386 for @xmath387 .
bearing in mind that @xmath343 , we obtain @xmath388 thus , for every set @xmath353 ( that is , for each @xmath288 and for each collection of components @xmath355 , @xmath389 ) , we have @xmath390 since there exist @xmath391 values of @xmath288 , and @xmath392 collection of components @xmath355 with @xmath389 , it follows that @xmath393 it remains to note that @xmath394 theorem [ th51 ] is proved . using theorem [ th51 ]
we show that for the class of measures @xmath19 satisfying the assumptions of this theorem , the riesz transform @xmath395 is large on a `` big '' portion of @xmath6 .
[ co56 ] let an integer @xmath20 and @xmath21 be given
. there are constants @xmath57 , @xmath58 , @xmath396 , depending only on @xmath60 , @xmath24 , and such that for any positive numbers @xmath125 , @xmath233 , @xmath126 , and for the corresponding cantor set @xmath6 , @xmath397^{1/2}\bigg\}>\frac{c_1}2\hh^d(e_n)= \frac{c_1}22^{dn}\ell_n^d\,,\quad \t_j=\frac{2^{-dj}}{\ell_j^s}\,.\ ] ] here @xmath398 is @xmath60-dimensional hausdorff measure , and @xmath19 is a measure satisfying the conditions of theorem [ th51 ] , such that @xmath395 exists ( in the sense of principal values ) @xmath398-a .
e. on @xmath6
. * remark . * for @xmath399
the condition about existence of @xmath395 holds for any borel measure on @xmath6 . generally speaking , it is not correct for @xmath400 .
set @xmath401 , @xmath402 , where @xmath244 and @xmath245 are the constants from theorem [ th51 ] .
choose a sufficiently small @xmath146 which will be specified later , and consider the corresponding cantor set @xmath6 .
let @xmath403 be the point measure with charges equal to @xmath404 and located at the centers @xmath241 of the cubes @xmath41 .
we claim that latexmath:[\[\label{f520 } if @xmath146 is small enough .
indeed , let @xmath205 be any two points lying in the same cube @xmath41 ( that is , @xmath406 ) .
in the same way as we obtain the inequality @xmath407 let @xmath408 $ ] , and let @xmath409 $ ] be such that @xmath410 . by we get @xmath411 ( we recall that @xmath412 ) .
we use with @xmath413 instead of @xmath414 , and with @xmath415 , @xmath416 , @xmath417 .
then @xmath418^{1/2 } \le\frac{c_0}{4t^s}\biggl[\sum_{j=1}^n\t_j^2\biggr]^{1/2 } , \end{split}\ ] ] if @xmath419^{-1},\ c''=c''(d , s)$ ] .
clearly , @xmath146 depends only on @xmath60 and @xmath24 .
the inequality and the same arguments as in yield the estimate @xmath420 let @xmath421 be the set defined by for the measure @xmath403 .
fix some cube @xmath422 , and set @xmath423 ( we consider only the points @xmath28 for which @xmath395 exists ) .
let @xmath424 be the cube symmetric to @xmath41 with respect to the center @xmath425 of the cube @xmath426 .
choose @xmath413 , and let @xmath427 be such that @xmath428 .
clearly , @xmath429 .
moreover , @xmath430 by the symmetry of the measure @xmath403 with respect to @xmath425 .
hence , for at least one couple of vectors @xmath431 , @xmath432 , or @xmath433 , @xmath434 , the angle between these vectors is less or equal to @xmath207 .
so , for the sets @xmath435 for @xmath436 we get @xmath437^{1/2}.\end{aligned}\ ] ] analogous estimates hold for @xmath438 . hence , @xmath439^{1/2}\bigg\ } \ge\hh^d(\ee_{n , k})+\hh^d(\ee_{n , k^\ast})\ge \ell_n^d\ ] ] for every @xmath422 .
since @xmath440 ( see ) , we get with @xmath441^{-1}$ ] . the inequality is a particular case of corollary [ co56 ] . indeed ,
note that @xmath442 , where @xmath443 .
multiplying both parts of by @xmath444 , we get .
as we mentioned in section 2 , our arguments are similar to those in @xcite ( which in turn use the ideas in @xcite ) .
on the other hand , there are certain differences as well .
for instance , in @xcite the harmonicity of the riesz transform ( outside the support ) was used in an essential way . since @xmath24 is not necessarily equal to @xmath445 ,
we can not use harmonicity , and we can not work with the measure supported on the boundaries of certain cubes , our measure will be supported also on the interior of these cubes . moreover , and this is more essential , @xcite uses a certain regularity of their cantor sets .
we have another type of cantor sets , lacking this regularity ( namely , not all our cubes @xmath446 are separated enough ) .
this creates specific difficulties .
there is a number of other differences ( for example , we do not use cotlar s inequality ) .
also , mateu and tolsa @xcite , while claiming the result for all @xmath124 ( and for their cantor sets ) , give the proof only for @xmath53 .
this is why we wish to present a full proof , even though it follows the idea of @xcite .
we need the following characterization of @xmath144 obtained in @xcite , chapter 5 : @xmath447 ( see for definition of @xmath109 ) .
following @xcite , we introduce the capacity @xmath448 where @xmath42 is the probability measure defined in section 1 . using and
, we have @xmath449^{1/2}\approx\tau\biggl[\sum_{p=1}^m\t_p^2\biggr]^{1/2},\quad \t_p=\frac{2^{-dj_p}}{\ell_{j_p}^s}\,.\ ] ] hence , @xmath450^{-1/2}\approx\biggl[\sum_{p=1}^m\t_p^2\biggr]^{-1/2},\ ] ] where the constants of comparison depend only on @xmath60 and @xmath24 .
it is easy to see that the measure @xmath451^{-1/2}\mu$ ] with @xmath452 belongs to @xmath109 . now
the relation and the upper bound in imply the estimate @xmath453^{-1/2}\approx\g_{s,+}^\mu(e_n),\quad c = c(d , s).\ ] ] thus , it is sufficient to prove that @xmath454 let @xmath19 be a positive radon measure supported on a compact set @xmath138 in @xmath2 , for which @xmath455 .
it is shown in @xcite , p. 217
, that the last inequality implies the estimate @xmath456 the arguments in this part of the proof of lemma 4.1 in @xcite are valid not only for @xmath155 , but for @xmath44 as well ( the reference [ p ] , lemma 11 in @xcite should be replaced by [ p ] , lemma 3.1 ) . for @xmath53 , this fact is also noted in @xcite , p. 46 . hence , @xmath457 . by the definition of @xmath144 ( see section 2 ) we have @xmath458 ( the inequality also follows from ( * ? ? ? * lemma 3.2 ) . ) we will prove by induction on @xmath7 .
the induction hypothesis is @xmath459 where the constant @xmath460 will be specified later .
let @xmath461 suppose that @xmath462 .
then by we get @xmath463 .
this inequality together with yield the estimate @xmath464 in particular , we get for @xmath465 , if @xmath466 .
moreover , we can assume without loss of generality that @xmath467 .
hence , there exists @xmath265 , @xmath468 , such that @xmath469 we consider two cases .
* for some constant @xmath470 to be determined below , @xmath471 the set @xmath472 is constructed exactly in the same way as @xmath6 , starting with @xmath473 instead of @xmath474 , and with @xmath475 instead of @xmath7 .
hence , @xmath476^{-1/2 } = a_0^{-1}c_0c\biggl[\sum_{p = k+1}^m\t_p^2\biggr]^{-1/2}\\ & = a_0^{-1}c_0c[s_m - s_k]^{-1/2}\stackrel{\eqref{f68}}{\le}\sqrt2a_0^{-1}c_0cs_m^{-1/2}\\ & \stackrel{\eqref{f63}}{<}a_0^{-1}c_0c'\g_{s,+}^{\mu}(e_n)<c_0\g_{s,+}^{\mu}(e_n),\end{aligned}\ ] ] if @xmath477 .
we get .
* case 2 . * for the constant @xmath478 determined above , @xmath479 as in @xcite , we again distinguish two cases , namely @xmath480 and @xmath481 . if @xmath480 , then @xmath482 and we have @xmath483 thus , holds if @xmath460 is sufficiently big .
suppose now that @xmath481 .
then @xmath484 we consider the measure @xmath485 clearly , @xmath486 .
we will show that @xmath487 .
assuming this fact for a moment , we get @xmath488 and follows . to prove that @xmath489 is bounded , we will use the local @xmath490 theorem of christ @xcite . according to the main theorem 10 in @xcite , p. 605
, it is enough to prove that @xmath110 satisfies the following conditions : \(i ) @xmath491 ; \(ii ) @xmath492 ; \(iii ) for each ball @xmath493 centered at a point in @xmath494 , there exists a function @xmath495 in @xmath496 , supported on @xmath493 , such that @xmath497 and @xmath498 @xmath110-almost everywhere on @xmath494 , and @xmath499 .
first we verify condition ( i ) .
let @xmath500 be a ball centered at @xmath501 .
if @xmath502 , we have @xmath503 suppose that @xmath504 .
let @xmath505 be the least integer for which @xmath506 .
then @xmath493 may intersect at most @xmath507 cubes @xmath508 .
hence , @xmath509 using the same kind of ideas , it is not difficult to verify condition ( ii ) as well .
thus , we only need to check the hypothesis ( iii ) . again , let a ball @xmath510 centered at @xmath511 be given . if @xmath512 , we set @xmath513 .
then @xmath514 suppose that @xmath515 .
let @xmath516 be the least integer for which there is a cube of generation @xmath516 ( i.e. the cube of @xmath517 ) contained in @xmath518 .
clearly , @xmath519 ( since @xmath493 is centered at @xmath511 )
. we will construct the function @xmath495 supported on @xmath493 . by definition of the capacity @xmath144
, there is the positive radon measure @xmath19 supported on @xmath6 such that @xmath455 and @xmath520 .
hence , there is the cube @xmath521 for which @xmath522 we need the localization lemma in @xcite .
let @xmath523 be any cube ( see for the notation @xmath78 ) .
let @xmath524 be an infinitely differentiable function supported on @xmath525 and such that @xmath526 , @xmath527 , @xmath528 , @xmath529 on @xmath530 .
by lemma 3.1 in @xcite , p. 207
, @xmath531 to simplify notation , set @xmath532 for @xmath533 .
at first we define @xmath495 when @xmath534 . in this case ,
we set @xmath536 if @xmath537 , we choose any cube @xmath538 , and define @xmath495 by translation of @xmath539 , namely @xmath540 ( we recall that @xmath541 is the center of @xmath542 ) .
clearly , @xmath543 ( the last inequality follows from the fact that there are at most @xmath544 cubes of @xmath516-th generation in @xmath493 ) . to prove that @xmath495 is bounded , we apply to a cube @xmath545 ( see section 1 for notations ) . by , the cubes @xmath546 are separated .
hence , @xmath547 , and by we have @xmath548 .
thus , @xmath549 since @xmath550 consists of @xmath551 cubes @xmath552 , we get the estimate @xmath553 now implies that @xmath554 to complete the proof we only need to check that @xmath555 . the same estimates as in yield the inequality @xmath556 , @xmath557 . integrating by parts , for every @xmath558
we get @xmath559 thus , it is enough to prove for @xmath560 and @xmath561 .
set @xmath562 for @xmath563 . then supp@xmath564 . as before , let @xmath532 for @xmath533 .
applying to the measure @xmath19 with @xmath565 , and then to the measure @xmath566 instead of @xmath19 and with @xmath567 , we obtain @xmath568 let @xmath569 suppose that @xmath570
. then @xmath571 for which @xmath572 .
hence , @xmath573 therefore , @xmath574 this estimate and imply the inequality @xmath575 in the same way we will show that @xmath576 indeed , @xmath577 for @xmath578 , and hence @xmath579 the same arguments as above together with and the property ( i ) of @xmath110 imply .
it remains to establish the inequality @xmath580 then will follow from and . for every cube @xmath581 we have @xmath582 hence , the left - hand side of does not exceed @xmath583 since @xmath584 .
theorem [ th23 ] is proved .
we start with the extension of theorem [ th23 ] to infinite sequences @xmath157 .
let @xmath585 , @xmath586 be positive numbers such that @xmath587 and let numbers @xmath57 , @xmath58 be given .
we define the set @xmath588 and the `` regularized '' sequence @xmath589 in the same way as in section 1 for @xmath590 .
namely , set @xmath63 .
if @xmath64 , then @xmath67 is the least @xmath68 for which @xmath69 .
possibly , @xmath591 for all @xmath68 . in this case
the sequence @xmath592 is finite : @xmath61 , @xmath593 .
clearly , @xmath594 furthermore , we set @xmath595 if @xmath592 is finite , then @xmath596 with @xmath597 .
the cantor set @xmath138 is defined by the relation @xmath598 where the sets @xmath599 are defined in section 1 . in particular ,
if @xmath592 is finite then @xmath600 .
[ th71 ] let @xmath20 , @xmath124 , and let a sequence @xmath601 satisfies
. then @xmath602^{-1/2}\le\g_{s,+}(e ) \le c\biggl[\sum_{j=1}^\infty\biggl(\frac{2^{-dj}}{\s_j^s}\biggr)^2\biggr]^{-1/2},\ ] ] where the positive constants @xmath49 and the parameters @xmath146 , @xmath147 of the cantor set @xmath138 depend only on @xmath60 and @xmath24 .
assume that @xmath592 is infinite .
since @xmath605^{-1/2},\quad p=1,2,\dots,\ ] ] we get the estimate from above . we also get ( that is @xmath606 ) if the series in diverges .
thus , we may assume that this series converges .
the definition of @xmath144 and theorem [ th23 ] imply the existence of measures @xmath607 , @xmath608 , such that @xmath609^{-1/2},\quad \|r_{\nu_p}^s\|_{l^\infty(\r^d)}\le1.\ ] ] we may extract a weakly convergent subsequence @xmath610 . denote by @xmath19 the weak limit of this subsequence as @xmath611 .
clearly , @xmath612^{-1/2}.\ ] ] for any @xmath613 , we have @xmath614 ( otherwise @xmath615 by continuity of @xmath616 on @xmath617 ) .
hence , @xmath618 by , @xmath619 .
hence , @xmath620 , and theorem [ th71 ] is proved . in @xcite
we obtained estimates for the hausdorff content of the set where the riesz transform @xmath395 is large .
we also obtained certain relations between hausdorff content and the capacity @xmath144 .
we are going to show that these estimates are attained on the cantor sets defined above .
the possibility of considering _ arbitrary _ sequences @xmath157 satisfying enables us to prove the corresponding assertions for _ any _ gauge function @xmath621 ( with the natural assumption that @xmath622 is nonincreasing see the explanation below ) without any additional conditions .
the _ hausdorff content @xmath626 _ of a set @xmath627 is defined by @xmath628 where the infimum is taken over all ( at most countable ) coverings of @xmath629 by balls of radii @xmath630 . later on we assume that @xmath622 is nonincreasing .
this condition , which may seem to be a regularity condition at the first glance , is actually not a restriction at all .
it was proved in @xcite , p. 133 , proposition 5.18 , that for any measure function @xmath621 either @xmath631 for all @xmath627 , or there is another measure function @xmath632 such that @xmath633 is nonincreasing and for which hausdorff contents @xmath634 and @xmath635 coincide up to a constant factor depending only on the dimension @xmath60 . for @xmath636 ,
set @xmath637 clearly , @xmath638 .
let @xmath621 be a measure function , @xmath639 , and let @xmath640 be inverse to @xmath621 .
for a measure @xmath19 consisting of @xmath641 point charges , we have obtained the inequality @xmath642^{1/2},\ ] ] where @xmath643 .
this implicit estimate can be written in a simpler form using a certain auxiliary function ( see @xcite for details ) .
[ pr72 ] for every @xmath644 and @xmath645 , one can find a measure @xmath19 which is a linear combination of @xmath641 dirac point masses , and such that @xmath646 , and @xmath647^{1/2},\quad c = c(d , s)\ge1,\ ] ] for any @xmath648 . in ( * ?
* section 7 ) a certain family of random sets and measures was introduced .
it was proved that this family contains ( random ) measures @xmath19 with the properties indicated in proposition [ pr72 ] .
thus , the existence of a measure with the desired properties was established , but the concrete measure was not presented .
below we give a non - probabilistic construction of such a measure . without loss of generality
we may assume that @xmath649 .
it is sufficient to prove our assertion for @xmath650 . for given @xmath641 , @xmath651
, we introduce the function @xmath652^{1/2},\ \text { where } \ h(a_\s)=n^{-1}h(\s).\ ] ] clearly , @xmath653 as @xmath654 ( we recall that @xmath655 , and @xmath656 ) . since @xmath657 is nonincreasing , @xmath658 in particular , @xmath659 . hence , @xmath660 as @xmath661
thus , there exists @xmath662 such that @xmath663^{1/2}=c_4,\quad \frac1{ph(\s)}\left[\int_{a_{\s}}^{\s}\bigg(\frac{h(t)}{t^s}\bigg)^2\frac{dt}{t}\right]^{1/2}<c_4,\ \s>\s_0,\ ] ] where the constant @xmath664 , depending only on @xmath60 and @xmath24 , will be specified later .
define @xmath11 by the equalities @xmath665 by , @xmath666 .
let @xmath6 be the cantor set from corollary [ co56 ] , and let @xmath19 be the probability measure consisting of @xmath649 equal point masses @xmath667 located at the centers of the cubes @xmath668 .
we will prove that implies .
indeed , @xmath669 if @xmath670 is big enough ; here @xmath671 is the constant from , and @xmath672 .
define the measure @xmath673 by the equality @xmath674 fix a ball @xmath675 .
suppose that @xmath676 . then @xmath677 let @xmath678 , and let @xmath9 be such that @xmath679 ( if @xmath680 , the upper bound is absent )
. then @xmath681 intersects at most @xmath682 cubes @xmath542 of @xmath9-th generation , that is at most @xmath683 cubes @xmath41 .
therefore @xmath684 hence , @xmath685 for every set @xmath629 in @xmath2 .
we have @xmath686^{1/2}\bigg\}\stackrel{\eqref{f519}}{\ge } ch(\s_0).\end{aligned}\ ] ] thus , for any @xmath648 we have @xmath687 .
if @xmath688 , then @xmath689 , and implies with @xmath690 . if @xmath691 , then @xmath692^{1/2}\\ & \ge\frac{c}{c_4p } \left[\int_{h^{-1}(c^{-1}\mathbf m / n)}^{h^{-1}(\mathbf m)}\bigg(\frac{h(t)}{t^s}\bigg)^2\frac{dt}{t}\right]^{1/2},\end{aligned}\ ] ] and we get with @xmath693 and @xmath694 . * remark . * one can see that the relation plays a crucial role in the proof of proposition [ pr72 ] ( as well as in the proof of the assertions below ) .
thus , additional assumptions on @xmath157 imply certain conditions on @xmath621 .
for instance , the assumption @xmath46 gives the unnatural restriction @xmath695 .
for @xmath696 , this means that @xmath697 . moreover , if we assume that @xmath151 , @xmath258 , then @xmath698 , @xmath22 .
our assumption that @xmath699 is nonincreasing , does not provide us with this property .
thus , we need additional conditions on @xmath621 ( for instance , the stronger assumption that @xmath700 is nonincreasing ) .
our next results concern the problem on the comparison of the capacity @xmath144 and hausdorff measure .
it was proved in @xcite ( see theorem 10.1 ) that for each compact set @xmath135 , @xmath701^{-1/2},\quad 0<s < d,\ ] ] where @xmath191 depends only on @xmath60 , @xmath24 , and @xmath702 is defined by the equality @xmath703 .
this relation is sharp in the following sense .
[ pr73 ] for any @xmath704 with @xmath44 , and for any measure function @xmath621 , there is a constant @xmath192 , depending only on @xmath193 , and a compact set @xmath138 , such that @xmath705 , and @xmath706^{-1/2},\ \text { where } \ h(t_2)=m_h(e).\ ] ] fix some @xmath662 , and define the infinite sequence @xmath707 by the equalities .
let @xmath138 be the cantor set from theorem [ th71 ] .
we claim that @xmath708 the upper bound is obvious .
the lower bound for a finite set @xmath592 was proved above .
if @xmath592 is infinite , we consider the weak limit @xmath709 of some weakly convergent subsequence of the sequence @xmath710 of the measures defined in .
clearly , @xmath711 for any ball @xmath681 ( see ) .
hence , @xmath712 .
this statement is a supplement to corollary 10.2 in @xcite : for each compact set @xmath135 , @xmath716^{s/\beta},\quad\text{where}\quad 0<s < d,\quad h(t)=t^\beta,\quad\beta >
s,\ ] ] and @xmath191 depends only on @xmath60 and @xmath24 . * remark . *
it is interesting to compare the condition @xmath718 with the corresponding condition @xmath720 , arising in the analogous problem of the comparison of hausdorff measure and the classical riesz capacity @xmath721 .
the latter is generated by potentials with the _ positive _ kernel @xmath722 ( see @xcite , @xcite , sections 1,2 and the references therein , and @xcite , p. 147
for a more general setting ) . clearly , @xmath723 .
the exponent 2 in @xmath724 reflects the difference between potentials with the signed kernel @xmath725 , and potentials with the positive kernel @xmath722 . | the aim of this paper is to estimate the @xmath0-norms of vector - valued riesz transforms @xmath1 and the norms of riesz operators on cantor sets in @xmath2 , as well as to study the distribution of values of @xmath1 .
namely , we show that this distribution is `` uniform '' in the following sense .
the values of @xmath3 which are comparable with its average value are attended on a `` big '' portion of a cantor set .
we apply these results to give examples demonstrating the sharpness of our previous estimates for the set of points where riesz transform is large , and for the corresponding riesz capacities .
the cantor sets under consideration are different from the usual corner cantor sets .
they are constructed by means a certain process of regularization introduced in the paper . . |
in dense stellar systems , such as globular clusters , galactic nuclei and star forming regions , direct collisions between stars occur quite frequently .
these collisions can modify the stellar populations of the system by creating objects ( e.g. blue stragglers , cataclysmic variables and millisecond pulsars ; * ? ? ?
* ) , or by destroying objects ( e.g. bright giants ; * ? ? ?
these kinds of strong interactions between stars can also modify the overall evolution of the system by changing the dynamics of the system and its energy budget .
the study of stellar collisions has become critical to the study of dense stellar systems @xcite . in order to understand how collisions can modify stellar populations and dynamics ,
we need to understand the collisions themselves .
we need to understand when collisions will destroy the two stars involved , and when a new star will be created .
if a star is created , how much mass has been lost from the system ? what is the composition and structure of the new star ? the answers to these questions depend on the two stars involved in the collision , their impact parameter , and the velocity of the collision .
consider for example , blue straggler stars , which are probable stellar collision products .
these are stars in clusters that are bluer , brighter , and more massive than the main sequence turnoff , and hence must be rejuvenated stars .
we see blue stragglers in all globular clusters that have been observed carefully , and they are also found in many open clusters and other environments .
the number is typically a few 10 s per cluster , but can get as high as a few hundred .
based on their numbers , these stars must have main sequence lifetimes on the order of gyrs , which implies a significant amount of hydrogen in their cores .
is that hydrogen mixed into the core during the stellar collision , or is it there already ?
only a detailed understand of the hydrodynamical mixing processes can answer that question . modelling
the collision itself is a hydrodynamical problem because of the timescales involved .
the merger timescale is typically on the order of hours , over which time the fluid is very close to ideal , with no radiative or diffusive effects of importance .
the first hydrodynamic simulations of stellar collisions were done with low - resolution grid codes , in one or two dimensions @xcite .
these earliest simulations were head - on collisions between equal - mass main sequence stars at a variety of impact velocities .
attention quickly turned to collisions between compact objects ( white dwarfs or neutron stars ) and main sequence stars ( e.g. * ? ? ?
* ; * ? ? ?
* ) , and collisions involving giant branch stars @xcite .
grid methods are used today mostly for collisions involving neutron stars , where general relativity needs to be considered ( e.g. * ? ? ?
. smoothed particle hydrodynamics ( sph ) was invented independently by @xcite and @xcite .
the first sph simulations of stellar collisions were published in 1987 by @xcite , who looked at a variety of collisions involving equal - mass main sequence stars . since then , the sph method has been applied extensively to a wide range of collision scenarios .
a very thorough review of collisions involving main sequence stars can be found in @xcite .
almost all possible combinations of stellar types have been collided , from pre - main sequence stars @xcite through giants ( e.g. * ? ? ?
* ) and even black holes and quark stars @xcite .
the sph method has some intrinsic advantages for studying stellar collisions .
the main advantage is that the sph particles follow the fluid flow .
therefore , all the computational effort is spent on the regions of interest rather than in calculating empty grid cells .
some stellar collisions , particularly those with large impact parameters and low relative velocities , result in a very slow in - spiral of the two stars , with many periastron passages and large apastron distances .
a grid code typically has to sacrifice resolution for coverage of the large physical space traversed by the stars involved in the collision .
however , the sph method also has some limitations which could affect our understanding of the outcome of the collision .
shocks are often not well resolved , in particular the weak ones which are expected in stellar mergers .
artificial viscosity is implemented to prevent unphysical oscillations , but it can cause spurious exchange of angular momentum between shear layers in a flow . a careful choice of the formulation of the artificial visocity is needed to address these concerns .
in addition , there is usually no treatment of turbulent mixing .
since the final structure and subsequent evolution of a collision product can depend quite sensitively on its composition , getting the amount of fluid mixing right is very important . in grid codes
, cells numerically force mixing of all matter that enters .
this will result in overmixing . on the other end of the scale , in sph codes ,
particles have a definite identity and entropy .
mixing in real fluids occurs through a turbulent cascade , where at the end some fluid elements gain entropy and some lose entropy .
this process is not permitted in sph , except in a coarse grained sense . in the presence of gravity
, the residual identity could lead to sedimentation , which can undo the coarse grained mixing . from a physical perspective
, we expect grid codes to overmix , and sph to undermix , with the true answer in between . some comparisons between sph and grid codes have been done for stellar collisions @xcite .
they collided a low - mass unevolved main sequence star ( modelled as a n=3/2 polytrope ) with a white dwarf .
they found that the global values of the simulations were in good agreement , but that some local values ( e.g. the density distribution near the white dwarf ) were slightly different .
they also found that the largest differences between their simulations was caused by different treatments of the potential well of the white dwarf , which was treated as a point mass in both simulations .
the treatment of entropy and its effect on mixing was not investigated in that paper .
comparisons between sph and grid codes for other hydrodynamic problems have also been published .
for example , @xcite looked at resolution requirements for both methods in a molecular cloud collapse calculation @xcite .
they compare both first- and second - order hydrodynamic simulations of a collapse of a protostellar binary system with sph , and find that the second - order codes produce results very similar to the sph results .
the main differences occur in the treatment of rotation .
accretion flows were studied with both sph and tvd methods by @xcite , who found that the general characteristics of the flow were the same , although the details were slightly different .
they found that the location of the shocks were probably better obtained in sph , but that the shocks themselves were better modelled using tvd .
comparisons between sph and eulerian codes were also applied to dynamical bar instabilities in a rotating star by @xcite , and in a variety of cosmological simulations by @xcite .
the overall result of these studies is that both sph and grid codes provide accurate descriptions of the hydrodynamics of most astrophysical situations .
however , different scenarios are better treated with different methods , depending on the science that one wishes to do with the simulation . in this paper
, we compare the results of stellar collision calculations performed using an eulerian tvd grid code @xcite and a lagrangian sph code @xcite . we have simulated the off - axis collision of two equal - mass main sequence stars with @xmath7 and @xmath8 .
each parent main sequence star is chosen to be found near the turnoff , with hydrogen depleted in the core , and is modelled with a @xmath0 realistic stellar model and as a @xmath1 polytrope .
we concentrate on a comparison of the mixing characteristics in the collision products in order to address the question of whether stellar mergers can produce a remnant star where the interior has been replenished with hydrogen due to significant mixing . in sections 2 and 3
we describe the two hydrodynamic methods and in section 4 the numerical setup .
section 5 gives the comparisons and we draw conclusions from those results in section 6 .
a three - dimensional eulerian hydrodynamic code @xcite is used to compute the flux of the conserved fluid variables of mass , momentum , and total energy across a cartesian grid in discrete time steps . for simplicity of notation , we can write the euler equations in vector form as @xmath9 where @xmath10 contains the conserved fluid variables per unit volume , @xmath11 holds the hydrodynamic flux terms , and @xmath12 represents the gravitational source terms .
the conserved cell - averaged quantities @xmath13 are defined at integer grid cell centres @xmath14 while the fluxes @xmath15 are defined at cell boundaries .
the tvd hydro code uses an operator splitting technique @xcite to separately solve the hydrodynamic and gravitational terms .
furthermore , the hydrodynamic step is dimensionally split such that flux updating is done in one dimension at a time . in the hydrodynamic step
, time integration is obtained using a second - order runge - kutta scheme , where a half - step is first performed , @xmath16 and then followed by a full - step , @xmath17 the second - order accurate fluxes @xmath18 at cell boundaries are calculated using the total variation diminishing ( tvd ) condition @xcite .
the hydrodynamic step is conservative in that flux taken out of one cell is added to the neighbouring cells and the conservation of mass , momentum , and energy therefore ensures the correct shock propagation . to address the question of chemical mixing , the hydro code is implemented along with a particle - mesh scheme where tests particles of fixed mass flowing along the fluid velocity field lines can be used to track passively advected quantities like chemical composition .
the particle - mesh scheme is coupled to the eulerian tvd scheme in the following manner .
the hydrodynamic scheme is dimensionally split and in order to synchronize the particles with the fluid , the particle - mesh scheme must also be dimensionally split .
consider solving for the fluid flow in the @xmath19 direction . while the fluid velocity at cell centers can be defined by dividing the cell - averaged momentum by the cell mass
, a more robust measure is that of the velocity at which mass is flowing across cell boundaries .
the fluid velocity at cell boundaries can be calculated as @xmath20 where the fluxes @xmath21 come from solving the mass continuity equation .
both the half - step velocity field @xmath22 and full - step velocity field @xmath23 are needed for the particle - mesh scheme .
the particles are then advected in the @xmath19 direction using a second - order runge - kutta time integration with a half step , @xmath24 followed by a full step , @xmath25 the velocity of the particle at position @xmath26 is interpolated from the fluid velocity field using a ` cloud - in - cell ' ( cic ) technique @xcite .
the cic interpolation is also used to determine local quantities such as density , pressure , entropy , and gravitational potential associated with each particle .
the coupling of the particle - mesh algorithm to the eulerian tvd code has the benefit of being able to track fluid variables like in an sph code , while still being able to solve the fluid equations using the eulerian approach .
the sph code is the parallel version of the code described in detail in @xcite .
it was parallelized using openmp .
our three dimensional code uses a tree to solve for the gravitational forces and to find the nearest neighbours @xcite .
we use an adiabatic equation of state , and the thermodynamic quantities are evolved by following the change in internal energy using the following equation : @xmath27 where @xmath28 is the internal energy of a particle , @xmath29 is its mass , @xmath30 is pressure , @xmath31 is density , @xmath32 is the artificial viscosity , @xmath33 is velocity , and @xmath34 is the sph smoothing kernel . both the smoothing length and the number of neighbours can change in time and space .
we used the balsara @xcite form of artificial viscosity with @xmath35 , following the advice of @xcite .
we ran a number of simulations with other forms of the artificial viscosity and with different values for the coefficients of the balsara form . as expected , the differences between the implementations of the artificial viscosity were slight , except for the completely unphysical case of no artificial viscosity .
standard sph treatments which do not properly account for the change in smoothing length as a function of time or space can either evolve energy or entropy properly , but not the other variable .
since our sph code evolves the energy equation , we are likely to have errors in the determination of entropy and therefore may have unphysical fluid mixing in our simulations . however , earlier simulations of stellar collisions done with this code ( sills et al .
2002 ) have been compared to simulations by lombardi et al .
( 1997 ) performed with an sph code that evolves the entropy equation .
the resulting hydrogen abundance profiles of the collision products were almost identical , and so we feel that the difference between the entropy and energy treatment of the thermodynamics is minimal in these circumstances .
and @xmath8 main sequence star constructed with ( a ) a realistic stellar model ( solid lines ) and ( b ) as an @xmath1 polytrope ( dashed lines ) . the tvd ( blue ) initial conditions exactly match the input models and the sph ( red ) initial conditions are also in very good agreement .
the sph particles have been randomly subsampled by a factor of 5 for clarity .
the entropic variable @xmath36 , where @xmath2 is the ideal gas adiabatic index , is used in place of the entropy .
units are in cgs.,title="fig:",width=307 ] and @xmath8 main sequence star constructed with ( a ) a realistic stellar model ( solid lines ) and ( b ) as an @xmath1 polytrope ( dashed lines ) .
the tvd ( blue ) initial conditions exactly match the input models and the sph ( red ) initial conditions are also in very good agreement .
the sph particles have been randomly subsampled by a factor of 5 for clarity .
the entropic variable @xmath36 , where @xmath2 is the ideal gas adiabatic index , is used in place of the entropy .
units are in cgs.,title="fig:",width=307 ] we have simulated the off - axis collision of two equal - mass main sequence stars with @xmath7 and @xmath8 .
the parent stars are chosen to be found near the turnoff , with hydrogen depleted in their cores , in order to address the question of whether stellar mergers can produce a remnant star where the hydrogen core has been replenished from significant chemical mixing .
the main sequence stars are constructed using two different methods and the initial mass profiles are compared in figure [ fig : initprofile ] .
first , a realistic model is calculated with the yale stellar evolution code ( yrec ; * ? ? ?
the yrec stars are found near the turnoff with an age of 13.5 gyr . in the central @xmath37 of the mass ,
formerly the hydrogen core , the hydrogen has been mostly converted to helium .
second , a polytropic model with index @xmath1 gives a good approximation to main sequence stars near the turnoff that have relatively little mass in their convective envelopes .
note that in contrast , n=3/2 polytropes are good approximations of convective regions , and very good for fully convective or mostly convective unevolved low mass stars . in both of our chosen stellar models , approximately 90% of the total mass
is contained within @xmath38 . however , the realistic model is more centrally condensed with steeper gradients than the polytropic one .
the parent stars are set up on zero - energy parabolic orbits with a pericentre separation equal to @xmath39 and the initial trajectories are calculated assuming point masses .
the orbital plane is taken to coincide with the @xmath40 plane and the stars are initially separated by a distance of @xmath41 .
the dynamical time for the stars , defined to be @xmath42 where @xmath43 is the average density , is approximately 1 physical hour and thus , modelling of the collision is a hydrodynamic problem . in the eulerian simulations ,
the collision is computed on a high resolution cartesian grid with @xmath44 cells .
initially , each parent star has a radius of @xmath45 grid cells and is assigned @xmath46 test particles .
the hydro grid has non - periodic boundaries which allow only one - way transmission and gas exiting the grid can not re - enter .
it is important that the parent stars are chosen to be small enough that the mass loss is still accurately captured on the finite grid .
the fft poisson solver can also be made non - periodic by using a gravity grid with dimensions that are twice as large as the hydro grid and treating the padded cells as vacuum with zero density . however , since the remnant star occupies only a fraction of the hydro grid , a more optimal approach is to make the poisson solver only pseudo non - periodic and use a gravity grid with dimensions that are 1.5 times as large as the hydro grid .
we impose a minimum density of @xmath47 g/@xmath48 whereby gas cells with lower density are considered part of the vacuum and all fluxes are set to zero .
this density threshold is found to reduce spurious shock heating as the stars are advected across the grid @xcite and is still a few orders of magnitude smaller than densities of interest . an ideal fluid approximation with an adiabatic index of @xmath49
is used in the simulations .
the simulations were run at ncsa on a shared - memory sgi altix with itanium 2 processors .
each simulation used 64 processors and 64 gb of memory and required approximately 8000 cpu hours to complete . in the sph simulations
, we used approximately 50 000 particles per star , or 100 000 in total .
the particles were initially distributed on an equally spaced grid , and their masses were varied until the density profile matched that of the stellar model we wished to use . the smoothing length is varied to keep the number of neighbours approximately constant at about 50 .
the particle masses range from @xmath50 to @xmath51 with the highest mass particles at the centre of the original star .
the simulations were run using sharcnet facilities on a 4-cpu shared - memory node of a compaq sc40 alpha with 667 mhz processors .
each simulation used less than 2 gb of memory and required approximately 150 cpu hours . note the substantially smaller resource requirements of the sph method for this kind of simulation compared to the grid method .
thick are taken through the @xmath40 and @xmath52 mid - planes .
the orbital plane coincides with the @xmath40 mid - plane .
both simulations produce a single rotating oblate with a mass of @xmath53.,width=307 ] both tvd and sph simulations of the off - axis collision between @xmath7 main sequence stars and between @xmath1 polytropes produced a single merger remnant with mass @xmath54that establishes hydrostatic equilibrium in approximately @xmath55 .
note that from here on , @xmath56 refers to the mass of bound gas and does not include ejected gas . during the merging process
, the outer envelopes of the parent stars are shock - heated and @xmath57 of gas gets ejected into the interstellar medium ( ism ) .
the resulting remnant is a rotating oblate that shows no indication of having a disc .
the tvd and sph simulations find similar mass loss with each other and also for the two stellar models .
the latter is due to the fact that both stellar models have similar structural profiles at larger radii . we do not attempt to quantify it to higher precision because of the difficulty in precisely determining what gas is bound to the remnant .
we show some images of the remnant in figure [ fig : image ] , taken at time @xmath58 .
all future figures show results also taken at this final time .
main sequence stars ( a ) and between @xmath1 polytropes ( b ) , simulated using a tvd ( blue ) and a sph ( red ) code .
note that ideal gas adiabatic index of @xmath2 is used to define the entropic variable @xmath59 .
units are in cgs.,title="fig:",width=307 ] main sequence stars ( a ) and between @xmath1 polytropes ( b ) , simulated using a tvd ( blue ) and a sph ( red ) code .
note that ideal gas adiabatic index of @xmath2 is used to define the entropic variable @xmath59 .
units are in cgs.,title="fig:",width=307 ] figures [ fig : massprofile]a and [ fig : massprofile]b show the mass profiles of the merger remnant from the collision between main sequence stars and between polytropes , respectively .
the two lines span the 1-sigma spread in the quantities in each mass bin . in this and
all future figures , the red lines give the results of the sph simulations , while the blue lines are the tvd results .
the profiles are constructed by first rank ordering the grid cells or particles in terms of the effective potential and then binning the structural variables .
the entropic variable , @xmath60 where @xmath2 is the ideal gas adiabatic index , is used in place of the entropy .
we also check if the particle - mesh scheme in the eulerian code is working accurately by measuring the mass profiles for the test particles .
the cic interpolation technique is used to assign local gas quantities to the particles and if the particles are indeed tracing the gas , then the mass profiles should be the same .
the particle - mesh algorithm does an accurate job of advecting the test particles along the velocity field lines as we find agreement at the percent level over the entire mass range .
the tvd and sph results for the polytropic case are in good agreement at all mass scales .
the remnant star is a rotating oblate with a mass of @xmath54 and is itself not a polytrope .
the central density and pressure have decreased and the entropic variable floor has risen , relative to the corresponding values in the parent polytropes . these changes can come from a combination of relaxation , shock heating , and mixing . the initial polytropes and the final remnant all have structural profiles that are smooth and featureless , and the modest demand on dynamic range is more than accounted for by the high resolution simulations . numerical relaxation in the core has been found to be only a few percent at these high resolutions . in @xcite , a @xmath1 polytrope with a radius of @xmath61 grid cells , which is lower resolution than that used in this paper , was advected diagonally across the @xmath40 orbital plane in over 1600 times steps to test the tvd code .
for the inner 5% of the enclosed mass , the decrease in density and pressure and the increase in entropy was found to be @xmath62 , while for the rest of the polytrope , negligible change was found .
similarly , a test of the sph code , where an isolated @xmath1 polytrope was evolved for 10 dynamical times , showed that the mixing in the inner 25% of the enclosed mass was @xmath63 . since shock heating is expected to be minimal in the core because of the low mach numbers , the change in entropy suggests that some mixing may have taken place .
we will address the question of mixing in more details in section [ sec : massmixing ] for the main sequence stars , both codes give similar results for the outer two - thirds of the total mass , but there are interesting differences in the interior .
like in the polytropic case , the remnant star is a rotating oblate with a mass of @xmath54 and some structure is retained from that of the parent stars . in the tvd results ,
the central density and pressure are lower and the entropic variable floor is higher compared to the corresponding sph values .
these observed differences are consistent with the differences between the eulerian and lagrangian approaches of the two simulations . in the eulerian case
, there will always be some numerical diffusion between adjacent grid cells and while this is generally alleviated by increasing the resolution , it is difficult to fully remove this effect here . in the hydrogen - depleted core of the main sequence stars ,
the structural variables change rapidly with mass or radius and the demand on dynamic range is challenging for a grid code without refinement meshes . in the sph results
, it appears that the cores of the parent stars have settled to the bottom of the potential well of the remnant star with little change in entropy .
this is consistent , but not necessarily confirmation of the argument that since the weak shocks present in these type of mergers are not well - resolved in sph simulations , particles will simply follow lines of constant entropy and may experience sedimentation . the particular implementation of the sph equations used in this paper was based on non - equal - mass particles .
for this reason , the inner regions of the parent stars consist of a few massive particles , and the density distribution requires that the particles in the inner regions of the realistic stars are even more massive than those in the polytrope . as a result ,
when the particles are binned in the collision product , the innermost bins have very few particles , resulting in the ` teeth ' seen in the top half of figure [ fig : massprofile ] .
the resolution in the inner regions is somewhat worse than that of the grid code simulations .
we should point out , however , that the overall results of the sph simulations are not significantly affected by this lack of resolution .
sph simulations that use equal - mass particles show the same behaviour as the ones presented here , in that the cores of the parent stars do not change their entropy and sink to the centre of the product @xcite .
one of the advantages of lagrangian sph simulations is that the particles of constant mass are directly tracked allowing one to probe the mixing of mass and the change in entropy , and to determine the final chemical composition of the merger remnant .
this is now possible for eulerian hydro codes coupled with a particle - mesh algorithm .
the mixing of mass is closely examined with the following procedure .
the particles are ranked ordered using their effective potential and each particle is assigned a mass coordinate based on its ranking and this mass coordinate is simply the value of the enclosed mass . in figure [ fig : mass ]
the initial and final mass coordinates are compared for both the tvd test particles and the sph particles .
the particles are binned in terms of the final mass @xmath64 in order to see where the mass in a given region of the remnant is coming from . for each bin
the 50% value of @xmath65 is found and the median curve is illustrated by the solid line .
similarly , the long - dashed lines enclose 50% of the scatter , while the short - dashed lines enclose 90% of the scatter .
both simulations of the merger of @xmath1 polytropes produced substantial mixing .
for the inner region of the remnant given by @xmath66 where the replenishing of hydrogen is important , the median curve for @xmath65 deviates from the one - to - one line as mass is mixed into the interior , mainly from larger radii as the mixing from below has a hard limit . considering this region as a whole ,
38% of the contained mass in the tvd case and 32% in the sph one is mixed in from outside .
the average amount of mixing for the entire remnant can be estimated using the average width @xmath67 of the zone enclosing half of the scatter , as bounded by the long - dashed lines . for complete mixing ,
the width is given by @xmath68 .
in the tvd case , the remnant is estimated to be @xmath69 fully mixed while for the sph case , we find that it is @xmath70 fully mixed .
the merger of @xmath7 main sequence stars results in less mixing , particularly in the interior of the remnant and in the sph simulation .
considering the interior region @xmath66 as a whole , 21% of the contained mass in the tvd case , but only 3% in the sph one is mixed in from outside . on average ,
the tvd remnant is estimated to be @xmath71 fully mixed while the sph remnant is only @xmath72 fully mixed .
the weaker mixing seen here in this case compared to the polytrope merger is due to the structural differences between the two stellar models .
the realistic main sequence star is more centrally condensed and it is relatively harder to penetrate the hydrogen - depleted core . during the merging process , the dense interior of each parent star acts like a semi - compact object and remains largely intact until the two cores reach the bottom of the potential well and merge with one another
. the differences in mixing between the realistic and polytropic mergers have previously been found by @xcite for sph simulations .
the stronger mixing seen in the eulerian case is consistent with the tendency of grid codes to overmix , whereas particle codes tend to undermix .
the effect of this mixing on the hydrogen profile is discussed later in section [ sec : hydrogenmixing ] .
@xcite have also studied the mixing of mass in sph simulations of stellar collisions for various stellar models . for comparison , their closest model to ours
involves the collision between @xmath1 polytropes with @xmath73 and @xmath74 .
when considering the interior quarter of the enclosed initial mass @xmath75 , they found a mixing value of @xmath76 .
unlike our simulations , they used equal - mass particles , and so had approximately 7000 particles in the inner 25% of the star . in our simulation
, we have just over 1500 particles in that region of the merger remnant , despite having more particles in our simulations overall .
@xcite show that lower resolution simulations result in more mixing ( see their figure 1 ) , simply because the regions of interest are by necessity more smoothed .
we conclude that our results are in good general agreement with other sph simulations of stellar collisions . .
note that ideal gas adiabatic index of @xmath2 is used to define the entropic variable @xmath36 .
units are in cgs.,title="fig:",width=307 ] .
note that ideal gas adiabatic index of @xmath2 is used to define the entropic variable @xmath36 .
units are in cgs.,title="fig:",width=307 ] in figure [ fig : entropy ] the initial and final values of the entropic variable @xmath36 of the tvd test particles and the sph particles are compared .
note that ideal gas adiabatic index of @xmath2 is used to define the entropic variable . in the absence of shocks , mixing , and relaxation , the fluid elements in the parent stars should be simply sorted by entropy in the remnant star , with @xmath77 . in the polytropic merger , it is not surprising that significant change in the entropic variable is found considering the amount of mixing that has occurred .
the two simulations are also in good agreement , though the small differences suggest that the tvd remnant is more strongly mixed , as seen in the previous section , and more strongly shock heated , which is expected because grid codes are known to provide better capturing of shocks .
note that while strong mixing has been found to occur in the interior of the remnant , it has only raised the entropy floor by a small amount . with main sequence stars ,
the tvd results show an artificial increase in the entropy floor that is caused by numerical relaxation rather than mixing or shock heating . in comparison ,
the lowest entropy sph particles show only relatively small changes with small dispersion .
overall , there is less change in the entropic variable in the realistic merger , but it is difficult to tell from the plots how much of the entropy increase is due to mixing and how much to shock heating . to account for the differences between figure [ fig : entropy]a and figure [ fig : entropy]b , mixing must play a substantial role in changing the entropy profile or there must be significantly more shock heating in the polytropic merger .
note that for some tvd test particles where shocks have not significantly raised the entropic variable , the decrease in local entropy due to mixing is observable .
the hydrogen mass fraction @xmath78 in the merger remnant from the collision between main sequence stars is plotted as a function of the enclosed mass in figure [ fig : hydrogen ] . for comparison
, the short - dashed line represents the hydrogen profile expected in the absence of mixing .
the sph remnant has a hydrogen profile that closely resembles the no mixing case .
the hydrogen profile in the hydrogen - depleted interior is sensitive to mixing , but only weak mixing and small changes of @xmath62 to the hydrogen fraction are found . while stronger mixing
is found at larger radii , it occurs between regions with approximately similar initial hydrogen abundance and this results in only small changes . in the tvd remnant
the stronger , non - local mixing has noticeably changed the hydrogen profile . considering the interior region
@xmath66 as a whole , the hydrogen fraction has on average increased by 24% and this is consistent with the previous finding where 21% of the mass is mixed in from the outside .
we also include the results for the polytropic collision to highlight the difference in mixing , but since this is not a realistic model , the initial chemical composition of the polytropes was taken from that of the @xmath7 main sequence star . in this case , both simulations show stronger , non - local mixing and this has resulted in more changes to the hydrogen mass fraction throughout .
the biggest difference we see between the sph and tvd simulations is in the amount of mixing in the interior of the remnant star . in order for a collision product to be observed as a blue straggler , it must remain as a main sequence star for a reasonably long time .
the sph simulations predict a very small amount of mixing , and we know from stellar evolutionary models of these collision products @xcite that the lifetimes of these collision products are quite short ( @xmath79years ) .
the tvd simulation of the same collision produces a collision product which has a central hydrogen abundance of @xmath80 , which will have a significantly longer lifetime ( @xmath81 years ) .
it is not clear from this work which answer is more correct .
we probably need to turn to the observations of blue stragglers in a variety of clusters . we know already that simple blue straggler population synthesis models ( based on results from sph simulations ) have some trouble reproducing the bright end of the observed distribution in some clusters @xcite , but we also know that eulerian schemes contain some numerical diffusion and hence over - predict the mixing .
the true result is probably somewhere between these two cases .
note that the mixing effect will be strongest in the collision that we have chosen to model : a collision between two stars at the main sequence turnoff .
this is , of course , why we chose to compare this particular collision , but we should caution the reader not to apply these results uniformly to all collisions .
stars that are less evolved and that have flatter composition profiles will show the effect of any mixing less when they collide , since any material that is ` mixed ' will have the same composition .
the particular choice of hydrodynamical method should not produce such a different result as it does for the highly evolved case shown here .
the differences in mixing seen in the merger of main sequence stars and the merger of polytropes are due to the structural differences between the two stellar models .
the realistic main sequence star is more centrally condensed and it is relatively harder to penetrate the hydrogen - depleted interior .
therefore , one expects that the larger the density contrast in the parent star , the weaker the mixing for a fixed collision velocity and geometry .
given the differences in results , it is important to use realistic stellar models rather than approximations like polytropes when simulating stellar collisions @xcite .
the eulerian simulations have the advantage of not requiring any artificial viscosity or the need to adjust any smoothing lengths and therefore , the numerical solutions to the fluid equations , in particular the capturing of shocks , are generally robust . however , they have the disadvantage of limited dynamic range due to the fact that the fluid equations were solved on a single cartesian grid .
one consequence of this is that only a small collisional parameter space can be considered and another is that the simulations often suffer some level of numerical diffusion in regions with large density gradients . in principle , the problems can be rectified by running larger volume and higher resolution simulations , but in practice this is not feasible because of limited computational resources .
one remedy is to use adaptive mesh refinement ( amr ) schemes in codes like art @xcite , enzo @xcite , and flash @xcite .
one caution we make is that it is important to understand any systematic problems in the transport of fluid across refinement mesh boundaries when studying mixing during stellar mergers .
the sph method is the hydrodynamical method of choice for modelling stellar collisions , because of its adaptability to a wide range of impact parameters , velocities , stellar masses , etc .
one does not need to know where the material will go before starting the simulation , since the sph particles are free to travel wherever they would like in space .
the method has its drawbacks in terms of treatment of shocks and spurious diffusion of particles caused by the numerical methods .
in addition , turbulent mixing is generally inhibited by the addition of artificial viscosity @xcite .
we have shown in this paper that for the most part , the two methods produce the same results .
the structure of the resulting collision product is very similar : the density and pressure profiles of the models are almost identical .
however , the amount of fluid mixing is somewhat different between the two methods , which can result in differing stellar lifetimes for the collision product .
we conclude that both eulerian and lagrangian implementations of fluid dynamics can be used to study stellar collisions , and that each code should be applied in a way to maximize its strengths .
care should be taken when interpreting results related to fluid mixing , but all other properties of the system are determined correctly by both approaches .
we thank the anonymous referee for the constructive critique of the paper .
is supported by nserc .
acknowledges nasa grant nng04gc50 g .
this work was made possible by the facilities of the shared hierarchical academic research computing network ( sharcnet : www.sharcnet.ca ) and the national center for supercomputing applications ( ncsa : www.ncsa.uiuc.edu ) .
oshea , b. w. , bryan , g. , bordner , j. , norman , m. l. , abel , t. , harkness , r. , & kritsuk , a. , 2005 , in adaptive mesh refinement : theory and applications , ed .
t. plewa , t. linde , & v. g. weirs ( berlin : springer ) , 341 | an eulerian tvd code and a lagrangian sph code are used to simulate the off - axis collision of equal - mass main sequence stars in order to address the question of whether stellar mergers can produce a remnant star where the interior has been replenished with hydrogen due to significant mixing . each parent main sequence star is chosen to be found near the turnoff , with hydrogen depleted in the core , and is modelled with a @xmath0 realistic stellar model and as a @xmath1 polytrope . an ideal fluid description with adiabatic index @xmath2
is used for all hydrodynamic calculations .
we found good agreement between the simulations for the polytropic case , with the remnant showing strong , non - local mixing throughout . in the interior quarter of the mass ,
@xmath3 is mixed in from larger radii and on average the remnant is @xmath4 fully mixed . for the realistic model , we found less mixing , particularly in the interior and in the sph simulation . in the inner quarter ,
@xmath5 of the contained mass in the tvd case , but only @xmath6 in the sph one is mixed in from outside .
the simulations give consistent results for the overall profile of the merger remnant and the amount of mass loss , but the differences in mixing suggests that the intrinsic difference between grid and particle based schemes remains a possible artifact .
we conclude that both the tvd and sph schemes can be used equally well for problems that are best suited to their strengths and that care should be taken in interpreting results about fluid mixing .
blue stragglers globular clusters : general hydrodynamics methods : numerical stars : evolution stellar dynamics |
the afterglows of gamma - ray bursts ( grbs ) are observed in the x - ray , optical , near - infrared , and radio , and appear to be well - described by the synchrotron blast - wave model in which the source ejects material with a relativistic bulk lorentz factor , driving a relativistic shock into the external medium ( see @xcite and references therein ) .
there is mounting evidence from the observed steepening of afterglow light curves that these ejecta are in many cases mildly to highly collimated , with opening angles @xmath2@xmath3 @xcite .
global fitting of the afterglow light curves over many decades in time and frequency , in the context of this model , can be used to derive constraints on the physical parameters of the model , i.e. , the energy and opening angle of the jet , the external density , the magnetic field strength and the energy distribution of the electrons behind the shock @xcite . in this model , the image of the afterglow is expected to appear highly limb brightened at frequencies above the peak synchrotron frequency @xmath4 , but more uniform at frequencies @xmath5 , especially below the self absorption frequency @xmath6 ( waxman 1997 ; sari 1998 ; panaitescu & mszros 1998 ; granot , piran , & sari 1999a , b ; granot & loeb 2001 )
. a measurement of the surface brightness profile ( sbp ) at several frequencies would thus provide an important test of the model . for typical parameters ,
the afterglow image expands superluminally , and has an angular radius @xmath7 a few days after the grb .
as pointed out by @xcite , this is of the same order as the angular einstein ring radius of a solar mass lens at cosmological distances , @xmath8 where @xmath9 is the lens mass , and @xmath10 , and @xmath11 , @xmath12 and @xmath13 are the angular diameter distances between the observer - source , observer - lens , and lens - source , respectively .
thus lensing by a star along the line of sight will produce a detectable magnification over the course of a few days to weeks .
the probability that any given grb will be microlensed is @xmath14 , where @xmath15 is the angular separation between the grb and lens in units of @xmath16 . since the original suggestion by @xcite
, this application of microlensing has been studied by several authors @xcite .
the well - sampled optical afterglow of grb 000301c exhibited a short time scale , nearly achromatic deviation from the nominal power - law flux decline .
garnavich , loeb & stanek ( 2000 ; hereafter gls ) found that this deviation is well - fit by a model in which the afterglow was microlensed by an intervening star .
however , the model they adopted for the sbp was somewhat crude : they assumed the emission arose solely from an outer ring of fractional width @xmath17 , with @xmath18 providing the best fit to the data .
recently , @xcite argued that , if one adopts realistic sbps , microlensing can not explain the data .
here we revisit the microlensing interpretation of this deviation .
we fit the optical / infrared dataset of grb 000301c to a double - power law plus microlensing model , adopting realistic sbps , and , under the assumption that the observed deviation is indeed due to microlensing , we directly invert the light curve , obtaining a non - parametric measurement of the sbp , and compare this measurement to theoretical expectations .
we will be considering the same dataset as gls , namely 104 photometric data points in the @xmath19-bands distributed as @xmath20 .
these data , shown in figure 1 , are taken from the compilation of @xcite with additional photometry from @xcite .
we will not consider the data in the radio regime @xcite , as the majority is not contemporaneous with the optical deviation . for more details on the grb itself and the optical , infrared and radio data ,
see gls , @xcite , @xcite , @xcite and @xcite .
the magnitude , shape , and duration of the microlensing signal depends sensitively on the sbp of the afterglow @xcite .
the sbp , in turn , depends on the observed frequency and the physical parameters of the shock .
thus detailed measurements of a microlensed afterglow light curve can be used to constrain the physical parameters of the afterglow .
however , the expected sbps in the relativistic blast - wave model can also be calculated _ a priori _ under various assumptions @xcite .
thus one can alternatively use the theoretical sbps to calculate the expected microlensing signature , thereby confirming or refuting the microlensing interpretation .
this was the approach taken by @xcite .
we adopt both approaches here .
we calculate sbps using the method described in @xcite and granot & loeb ( 2001 ) .
briefly , the hydrodynamics are described by the blandford - mckee ( 1976 ; bm hereafter ) self - similar solution , assuming a power - law external density profile , @xmath21 , and a power - law number versus energy distribution of electrons with index @xmath22 , @xmath23 , just behind the shock , which thereafter evolves due to radiative and adiabatic losses .
we integrate the emission over the entire volume behind the shock front .
a few days after the burst the afterglow is typically in the slow cooling regime ( i.e. a typical electron cools on a timescale larger than the dynamical time ) and the ordering of the break frequencies is @xmath24 , where @xmath25 is the cooling frequency @xcite .
we assume that this is the case for grb 000301c , as indicated by @xcite .
we consider only the sbps from frequencies @xmath26 , since after a few days the optical band is typically well above @xmath6 ( and usually also above @xmath27 ) .
the spectrum consists of several power - law segments ( plss ) , where @xmath28 ( see figure 1 in granot & loeb 2001 ) . due to the self - similar nature of the hydrodynamics ,
the normalized sbp within a given pls is independent of time ( though it changes considerably between different plss ) .
thus , for each @xmath29 , there are 3 relevant forms for the sbp , each corresponding to a different pls : @xmath30 ( @xmath31 ) , @xmath32 ( @xmath33 ) and @xmath34 ( @xmath35 ) .
the corresponding values of the temporal index @xmath36 for @xmath37 ( @xmath38 ) are @xmath39 ( @xmath40 ) for @xmath30 , @xmath41 ( @xmath42 ) for @xmath32 and @xmath43 ( for both @xmath37 and @xmath38 ) for @xmath34 . the expected sbps are shown in figure 2 , for a uniform external medium @xmath37 , and a stellar wind environment @xmath38 .
note that the sbp becomes more limb - brightened as the frequency goes to a higher pls , i.e. the spectral slope @xmath44 is reduced , or as the value of @xmath29 is reduced .
we fit the data of the flux as a function of time to the form of two asymptotic power - laws which join smoothly : @xmath45^{-1/s } , \label{eqn : foft}\ ] ] where @xmath46 is the `` break time , '' @xmath47 is the unlensed flux at the point where the asymptotic power laws before and after the break meet , @xmath48 and @xmath49 are the indices of the flux decline , and @xmath50 is a sharpness parameter ; the larger the value of @xmath50 , the sharper the break .
the magnification @xmath51 due to microlensing depends on the impact parameter @xmath15 , the radial sbp of the afterglow image , and the angular size of the afterglow in units of @xmath16 , @xmath52 here @xmath53 , @xmath54 is the time from the grb in days and @xmath55 is the value of @xmath56 after 1 day .
the light curve exhibits an achromatic break at @xmath57 days .
berger et al . ( 2000 ) find that this break can be well explained by jet with @xmath37 , while panaitescu ( 2001 ) argues that a jet alone can not produce the required steepening of the light curve , and invokes a double power law electron distribution , which causes additional ( though chromatic ) steepening of the light curve if the synchrotron frequency corresponding to the break in the electron distribution crosses the optical band around @xmath46 .
either way , we would expect the sbp at @xmath58 to be different than at @xmath59 . however , our best fit models give @xmath60days , while the majority of the constraints on the sbp come from times @xmath61days , where the spherical self - similar bm solution should still be applicable , resulting in the self - similar sbps described in [ sbps ] .
we will therefore adopt the simpler ( and self consistent ) working assumption that the sbp , when normalized to its average value , as a function of radius normalized to @xmath62 , is independent of time within a given pls .
the model in is a function of @xmath63 free parameters : four parameters describe the shape of the broken power - law ( @xmath64 ) ; one parameter provides the normalization @xmath47 for each of @xmath65 frequency bands .
however , extinction from the milky way and possibly the host galaxy itself will induce an ( uncertain ) curvature in the spectrum .
indeed , as noted by @xcite and @xcite , the early optical / infrared data for grb 000301c show evidence for a curved spectrum , which may be due to extinction or is perhaps intrinsic . for simplicity , we will use a free parameter for each @xmath47 .
we have also performed tests where we fit the data assuming the form @xmath66 . in general
we find worse fits , though our conclusions regarding the sbps are essentially unchanged .
as we show , the results for the @xmath67-band data are the same as for the full dataset , indicating that our assumption of independent @xmath47 does not bias our results . ] and two additional parameters describe the microlensing effect ( @xmath68 ) for a given sbp . considering all the data for grb 000301c in figure 1 , there are @xmath69 parameters without microlensing , and @xmath70 free parameters with microlensing using theoretically - calculated sbps . for comparison purposes , we also fit the data to the model adopted by gls : we assume that the source emission is confined to a thin ring of fractional width @xmath17 . however , we extend the model of gls by allowing the image interior to the ring to have a relative surface brightness @xmath71 referenced to the outer ring ( @xmath72 reduces to the gls model ) .
this model introduces two additional free parameters .
finally , we consider a non - parametric fit ( `` direct inversion '' ) to the sbp .
we divide the image into @xmath73 equal - area annuli , and find the fraction @xmath74 of the total flux contributed by each bin @xmath75 . in this case
the magnification is given by @xmath76 , where @xmath77 is the magnification of annulus @xmath75 .
this adds an additional @xmath78 parameters ( due to the constraint that @xmath79 ) .
we find the best - fit solution by minimizing @xmath80 with respect to all of these parameters .
we define the @xmath81 errors on these parameters as the projection of the @xmath82 hypersurface on the parameter axes , where @xmath83 @xmath84 is the minimum @xmath80 for a given model , and @xmath85 is the number of degrees of freedom for the 104 data points . we normalize @xmath86 by the factor @xmath87 because we believe that the errors on the data points are likely underestimated , resulting in inflated values of @xmath80 .
errors on the fit parameters determined using these inflated values of @xmath80 would be significantly underestimated .
figure 1 shows the best fit for the double power - law model with no lensing .
the best - fit parameters and @xmath81 errors for @xmath88 are tabulated in table 1 .
the fit is poor : @xmath89 for @xmath90 dof .
residuals from the broken power - law model are shown in figure 3 ; the systematic deviations from this model are apparent .
when microlensing is included , the fit improves considerably . for the extended gls model ,
we find that the data is best explained by emission solely from an outer ring of relatively small fractional width .
specifically , we find best - fit parameters , @xmath91 , @xmath92 for @xmath37 and @xmath93 , @xmath94 for @xmath38 .
the remaining parameters and the best - fit @xmath80 are shown in table 1 .
the parameters we find differ from those of gls , and we find a @xmath95 smaller than they report . however ,
when we evaluate their model fit , we recover their @xmath95 , indicating that their fit was likely a local , not global , minimum . for the fits with microlensing
, we find that the sharpness parameter @xmath50 is not well constrained , but is always @xmath96 , indicating a sharp break . to avoid excessive covariances with other parameters , we fix @xmath97 for the remaining fits , thus reducing @xmath98 by one .
the results are essentially identical for other values of @xmath96 . for a uniform external medium ( @xmath37 ) ,
for which one generally predicts more limb - brightened sbps , we find @xmath99 , and @xmath100 for @xmath101 , @xmath102 , and @xmath103 , respectively . for the case of a wind - like medium ( @xmath38 ) ,
the fits are generally worse : @xmath104 , and @xmath105 for the same frequency ranges .
the deviations of these fits from the double power - law model are shown in figure 3 .
it is clear that a significantly limb - brightened profile is required to fit the sharp , large deviation at @xmath106 days . for @xmath107 ,
both wind and uniform external media provide satisfactory fits to the data . for @xmath102 ,
the uniform medium provides a marginal fit .
frequencies below the peak synchrotron frequency @xmath27 , or @xmath102 with a wind - like medium ( @xmath38 ) , result in sbps that are too uniform to fit the data . for the non - parametric sbp fit to the data
, we impose the condition that @xmath108 , and we increase the number of bins @xmath109 in the recovered sbp until @xmath80 changes by @xmath110 .
we find that @xmath111 is optimal .
we find @xmath112 for @xmath37 , and @xmath113 for @xmath38 , for 89 dof .
the inferred surface brightnesses are shown in figure 2 , the fit for @xmath37 is shown in figure 1 , and figure 3 shows the deviations of both the @xmath37 and @xmath38 models from the best - fit double power law . the best - fit values and @xmath81 errors for the parameters @xmath48 , @xmath49 , @xmath46 , @xmath55 , and @xmath15 are given in table 1 for all viable ( @xmath114 ) fits .
although the values of @xmath95 for the direct inversion formally indicate poor fits , it is clear from figures 1 and 3 that microlensing reproduces the observed light curve shape quite faithfully , at least for limb - brightened sources .
it is likely that underestimated errors contribute significantly to the inflated @xmath80 .
we have also fit the @xmath67-band data only using both the sbp determined from direct inversion and the theoretically - calculated sbps .
the sbp inferred from direct inversion is consistent with that found from fitting all the data ( with larger errors , of course ) .
the fit is better : @xmath115 for 37 dof .
fits using the theoretical sbp yield the same conclusions as the fits to the entire data set : the sbps for @xmath35 provide fits that are acceptable ( @xmath116 and @xmath117 for @xmath37 and @xmath38 respectively , for 40 dof ) .
the fit for @xmath118 , and @xmath37 is marginal : @xmath119 .
all other fits are considerably worse : @xmath120 .
@xmath121 & @xmath50 & @xmath46 & @xmath15 & @xmath55 & @xmath80/dof + no @xmath51-lensing & @xmath122 & @xmath123 & @xmath124 & @xmath125 & & & 240.7/93 + @xmath37 , @xmath102 & @xmath126 & @xmath127 & 20 & @xmath128 & @xmath129 & @xmath130 & 166.1/92 + @xmath37 , @xmath103 & @xmath131 & @xmath132 & 20 & @xmath133 & @xmath134 & @xmath135 & 154.7/92 + @xmath38 , @xmath103 & @xmath136 & @xmath137 & 20 & @xmath138 & @xmath139 & @xmath140 & 155.5/92 + @xmath37 , di & @xmath131 & @xmath141 & 20 & @xmath142 & @xmath143 & @xmath144 & 148.5/89 + @xmath38 , di & @xmath145 & @xmath146 & 20 & @xmath147 & @xmath148 & @xmath149 & 146.9/89 + @xmath37 , gls & @xmath131 & @xmath150 & @xmath151 & @xmath152 & @xmath153 & @xmath154 & 147.3/89 + @xmath38 , gls & @xmath145 & @xmath155 & @xmath156 & @xmath157 & @xmath158 & @xmath159 & 146.9/89 + fit parameters and @xmath81 errors for grb 000301c .
[ tbl : table1 ]
we find that the early - time ( @xmath160 ) flux decline is well - constrained , @xmath161 .
the slope of the optical spectrum at this time is @xmath162 , taking into account milky way ( mw ) extinction only , but can be as shallow as @xmath163 if smc - type host extinction is included @xcite . for @xmath35 ( the sbps most consistent with the data ) , the value of @xmath48 implies @xmath164 and thus @xmath165 , which is only slightly steeper than the observed spectrum for mw extinction . for @xmath166 and @xmath37
, @xmath167 implies @xmath168 and thus @xmath163 , consistent with observations _ if _ smc - type host extinction is adopted .
if the break at @xmath169 is due to a jet , then simple analytic models predict that the change in temporal flux index @xmath36 should be @xmath170 for @xmath37 and @xmath166 or @xmath171 for @xmath103 and @xmath37 or @xmath172 @xcite .
kumar & panaitescu ( 2000 ) find , using a semi - analytic model , that the jet break is rather smooth , especially for a wind environment ( @xmath38 ) , and therefore inconsistent with the sharp break that we infer .
however , initial results of numerical calculations of the jet break @xcite , based on a 2d hydrodynamic simulation of the jet for @xmath37 , indicate that the break can be rather sharp ( with @xmath173 for an observer along the jet axis ) , and that @xmath49 is smaller by @xmath174 compared to the simple analytic predictions mentioned above , making @xmath175 larger by a similar factor .
nevertheless , the sharpness of the break still presents a serious problem for the model where the jet propagates into a stellar wind ( @xmath38 ) .
also , the sbp is expected to become more uniform at @xmath176 compares to @xmath160 @xcite .
since we assume that the form of the sbp relevant for @xmath160 holds all along , we underestimate the value of @xmath49 , and overestimate the value of @xmath175 accordingly is larger for @xmath102 ( for which the sbp is more homogeneous ) , compared to @xmath103 , for @xmath37 or @xmath172 , ( for which the sbp is more limb brightened ) . ] .
this effect is expected to be larger for @xmath103 , where the sbp is more limb brightened to begin with .
the values we obtain for @xmath175 may therefore serve as upper limits ( especially for @xmath103 ) . for the sbp fit for @xmath37 and @xmath166 , we find @xmath177 whereas based on the electron index derived from @xmath48 , the prediction is @xmath178 . for @xmath35 and @xmath37 , we find @xmath179 , which is reasonably consistent with the prediction @xmath180 .
thus , assuming the break in the light curve is due to jet effects , the only scenario which provides a satisfactory explanation of the data is @xmath103 and @xmath37 . for @xmath181 , the inferred sbp is only marginally consistent with the expected profile , and the observed change in the flux indices is smaller than the theoretical expectation . for @xmath107 and @xmath38 ,
the observed break is too sharp compared to theoretical predictions .
we note that our best fit model parameters , @xmath103 and @xmath37 , are also favored by @xcite , and said to provide a good fit to the data by @xcite .
@xcite argued that the expected sbps are too uniform to reproduce a significant microlensing deviation .
this conclusion appears to be partially due to the fact that he found a best - fit impact parameter of @xmath182 , whereas we find @xmath183@xmath184 ( and @xmath185 for our best fit model ) .
all else equal , more uniform sbps will result in microlensing deviations that are shallower and broader than more limb - brightened sbps .
thus , to reproduce the observed deviation , a more uniform source must have a smaller impact parameter , a trend we find empirically when fitting the theoretical sbps .
thus , it is unclear why @xcite , who argued that the expected sbps should be more uniform than that found by gls , derived a value of @xmath15 that is twice as large ( @xmath182 versus @xmath186 ) .
we conclude that the microlensing of realistic sbps can explain the optical / infrared light curve of grb 000301c . fitting the data with a non - parametric sbp
reveals that the afterglow image must be significantly limb - brightened .
specifically , @xmath187 of the flux must originate from the outer 25% of the area of the image ( with a significance of @xmath188 ) .
the recovered sbp is marginally consistent ( at the @xmath189 level ) with the profile expected for a uniform external density profile ( @xmath37 ) and frequencies
it is consistent at the @xmath190 level with the emission expected from frequencies @xmath107 and both uniform ( @xmath37 ) and wind - like ( @xmath38 ) external media .
we thank k. stanek for providing an updated compilation of the data points . this work was supported in part by nasa through a hubble fellowship grant from the space telescope science institute , which is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas5 - 26555 ( for s.g . ) , by the horowitz foundation and us - israel bsf grant bsf-9800225 ( for j.g . ) and the us - israel bsf grant bsf-9800343 , nsf grant ast-9900877 , and nasa grant nag5 - 7039 ( for a.l . ) . | the optical afterglow of gamma - ray burst ( grb ) 000301c exhibited a significant , short - timescale deviation from the power - law flux decline expected in the standard synchrotron shock model .
garnavich , loeb & stanek found that this deviation was well - fit by an _
ad hoc _ model in which a thin ring of emission is microlensed by an intervening star .
we revisit the microlensing interpretation of this variability , first by testing whether microlensing of afterglow images with realistic surface brightness profiles ( sbps ) can fit the data , and second by directly inverting the observed light curve to obtain a non - parametric measurement of the sbp .
we find that microlensing of realistic sbps can reproduce the observed deviation , provided that the optical emission arises from frequencies above the cooling break .
conversely , if the variability is indeed caused by microlensing , the sbp must be significantly limb - brightened .
specifically , @xmath0 of the flux must originate from the outer @xmath1 of the area of the afterglow image .
the latter requirement is satisfied by the best fit theoretical sbp .
the underlying optical / infrared afterglow lightcurve is consistent with a model in which a jet is propagating into a uniform medium with the cooling break frequency below the optical band .
# 1#23.6pt 0(i / i)_0 |
one of the major challenges of modern astrophysics is to understand the physical processes involved in galaxy formation and evolution .
significant steps in this direction could be made by tracing the build up of stellar mass and metallicity as a function of cosmic time .
one way to address this issue is through cosmologically deep surveys which map how galaxy properties change for samples at different redshifts ( @xmath3 ) . among these properties ,
the relation first observed by lequeux ( 1979 ) between heavy - element nebular abundance and galaxy mass , or its extension , the luminosity - metallicity relation ( e.g. , skillman 1989 ; zaritsky 1994 ) , are being extensively used to probe the metal enrichment along cosmic history .
clear signs of evolution are being revealed by studies of these relations at both intermediate ( savaglio 2005 ; lamareille 2006 ; mouhcine 2006 ) and high @xmath3 ( shapley 2005 ; maier 2006 ; erb 2006 ) , which generally find significant offsets in these relations when compared to their versions in the local universe .
alternatively , one may study galaxy evolution through _ fossil _ methods , which reconstruct the star formation history ( sfh ) of a galaxy from the properties of their stars .
the most detailed studies of this sort are those based on nearby galaxies ( including our own ) , where color - magnitude diagrams and even spectroscopy of individual stars are feasible ( e.g. , smecker - hane 1996 ; hernandez 2000 ; dolphin 2002 ; rizzi 2003 ; skillman 2003 ; monaco 2005 ; koch 2005 ) . comparing such observations with predictions of evolutionary synthesis models then allows one to infer how stars were born as a function of time , as well as to constrain the chemical enrichment histories of the galaxies ( e.g. , lanfranchi , matteucci & cescutti 2006 ; carigi , coln & peimbert 2006 ) . by mapping how stars become richer in metals as a galaxy ages ,
these studies provide useful constraints for chemical evolution models . beyond the local group and its neighborhood
, fossil methods must derive sfhs from integrated galaxy spectra , using either selected spectral features ( kauffmann 2004 ; brinchmann 2004 ; gallazzi 2005 ; nelan 2005 ; thomas 2005 ; bernardi 2006 ) or the full spectrum ( panter , heavens & jimenez 2003 ; heavens 2004 ; cid fernandes 2005 ; mathis , charlot & brinchmann 2006 ; panter 2006 ) .
this approach works with much less information than is available from individual stars in galaxies , so one can not expect similar levels of accuracy .
on the other hand , integrated spectra of galaxies are now available by the hundreds of thousands from modern spectroscopic surveys , so the decrease in accuracy is compensated by orders of magnitude increase in statistics and by the broader range in galaxy types spanned by such surveys .
it is therefore of great interest to explore methods to retrieve information on galaxy evolution from integrated spectroscopic data . in our ongoing series of papers
entitled the semi empirical analysis of sloan digital sky survey ( sdss ) galaxies , ( cid fernandes 2005 ; mateus 2006a ; stasiska 2006 ; mateus 2006b ; seagal i iv , respectively ) we have shown that a decomposition of galaxy spectra in terms of instantaneous bursts of different ages ( @xmath4 ) and metallicities ( @xmath5 ) produces both excellent spectral fits and astrophysically meaningful results .
so far , however , our description of sfhs has been extremely sketchy , based on averages over the @xmath4 and @xmath5 distributions , thus washing out valuable time - dependent information .
this paper shows that we can actually do much better . to illustrate this
we have culled a sample of star - forming ( sf ) galaxies from the sdss ( section [ sec : data ] ) , to which we apply our synthesis method and derive mass - assembly and chemical enrichment histories ( section [ sec : sfh_theory ] ) .
the evolution of stellar mass and metallicity is then studied by grouping galaxies with similar nebular abundances or mass ( section [ sec : sfh_results ] ) .
details of this analysis and cross - checks on the results reported here are presented in a companion paper ( asari , in prep , hereafter a07 ) .
we have recently completed detailed pixel - by - pixel spectral fits of 573141 galaxies from the main galaxy sample of the sdss data release 5 ( adelman - mccarthy 2006 ) with the synthesis code starlight described in seagal i and ii .
starlight decomposes an observed spectrum in terms of a sum of simple stellar populations covering a grid of 25 ages from @xmath6 myr to 18 gyr and 6 metallicities @xmath7 to 0.05 ( ie , from 0.005 to 2.5 @xmath8 ) .
each of these @xmath9 instantaneous bursts is represented by a spectrum extracted from the evolutionary synthesis models of bruzual & charlot ( 2003 ) for a chabrier ( 2003 ) initial mass function , padova 1994 tracks , and stelib library ( le borgne 2003 ) .
illustrative fits are presented in seagal i and a07 . emission lines were measured from the starlight subtracted spectra .
figure [ fig : bpt]a shows the baldwin , phillips & terlevich ( 1981 , bpt ) diagnostic diagram for 175234 galaxies for which h@xmath10 , [ oiii]@xmath115007 , h@xmath12 and [ nii]@xmath116584 are all detected with @xmath13 .
reddening corrected line fluxes are used in this plot and throughout our analysis .
the distribution of points in this diagram resembles a flying seagull , with sf galaxies occupying the left wing and galaxies hosting an agn on the right .
we define sf galaxies as those under the divisory line proposed in seagal iii ( solid line in figure [ fig : bpt]a ) . further requiring a minimum continuum @xmath14 of 10 at @xmath15
leaves a set of 84828 objects , hereafter the sf - sample , shown in figure [ fig : bpt]b
. results for alternative sample definitions ( involving different @xmath14 thresholds , other agn / sf separation criteria , and cuts on galaxy inclination ) are presented in a07 .
note that by selecting sf galaxies we indirectly select spirals over elliptical galaxies .
the sf - wing is essentially a sequence in nebular metallicity ( seagal iii and references therein ) , which we quantify by the oxygen abundance obtained through the strong - line calibration of stasiska ( 2006 ) : @xmath16}\lambda5007}{{\rm [ nii]}\lambda6584}\ ] ] where @xmath17 ( allende prieto , lambert & asplund 2001 ) .
we believe this is the most reliable calibration available in the literature , since it is based on temperature - based abundances in the metal - poor case and on the ( [ oii ] + [ oiii])/h@xmath10 calibration of pilyugin ( 2000 ) in the metal - rich regime , using a large sample of giant hii regions .
@xmath18 is both a physically motivated and mathematically convenient coordinate to map galaxy positions along the left wing in the bpt diagram . from the tip of the sf - wing in fig [ fig : bpt]a to its bottom
, our sf sample spans the @xmath191.6 @xmath8 range . in fig
[ fig : bpt]b this interval is chopped onto 6 logarithmic bins of width @xmath20 dex , except for the first one which is twice as wide to include more sources .
galaxies inside these same bins will be grouped together in the analysis of star - formation and chemical enrichment histories presented in the following .
as thoroughly discussed in the literature on stellar population synthesis , the combination of noise in the data , astrophysical and mathematical degeneracies limits the amount of information about galactic histories that can be extracted from integrated spectra .
clearly , the 150 components of the population vector used in our fits are a greatly over - detailed description , so some _ a posteriori _ compression of the results to a lower dimension is needed .
our previous papers ( seagal i iv ) have taken this compression approach to its extreme by reducing the whole @xmath4 and @xmath5 distributions encoded in the population vector to their first moments , @xmath21 @xmath22 where @xmath23 is the fraction of the total stellar mass ( @xmath24 ) which is in the @xmath25 population .
both @xmath26 and @xmath27 proved to be mathematically robust and astrophysically valuable summaries of the full population vector .
in seagal i , ii and a07 these averages are shown to cleanly reveal important relations , such as those between @xmath26 and @xmath24 ( associated with the galaxy downsizing phenomenon ) , @xmath27 and @xmath24 ( the mass - metallicity relation ) , and @xmath28 and @xmath18 ( which reveals a connection between stellar and nebular chemical enrichment levels ) . given this success , it is natural to explore less drastic compressions of the population vector .
one way to compress the population vector but retain the age - information is to marginalize over the @xmath29 distribution and compute the total mass converted into stars as a function of time : @xmath30 where @xmath31 denotes the fraction of the total mass converted onto stars ( @xmath32 ) that is associated with the @xmath25 base age .
the superscript `` c '' stands for `` converted onto stars '' , and is introduced to distinguish @xmath31 and @xmath33 from @xmath23 and @xmath24 , which refer to the mass still locked inside stars , ie , discounting the mass returned to the inter - stellar medium ( ism ) by stellar evolution .
@xmath34 is thus a monotonic function which grows from 0 to 1 as @xmath4 runs from the time galaxy formation starts to the present , and maps what fraction of @xmath32 was assembled up to a given lookback time .
, such that @xmath4 is both a stellar - age and a lookback time . ] to track the evolution of the _ stellar metallicity _ we compute the total mass in metals inside stars as a function of @xmath4 and divide it by the total mass inside stars at the same time .
this yields the time - dependent mean stellar metallicity : @xmath35 the computation of @xmath36 takes into account that the total mass inside stars changes with time , both due to the sfh and to the time dependence of the returned mass .
accordingly , only populations older than @xmath4 enter the definition of @xmath36 , since younger ones had not formed yet , ie .
, @xmath37 for @xmath38 .
the explicit time - dependence in equations ( [ eq : mah ] ) and ( [ eq : zeta ] ) provides a description of the star formation and chemical histories of galaxies which goes well beyond that obtained with the mere first moments of the age and metallicity distributions .
given the numerous degeneracies which affect population synthesis of galaxy spectra one might be skeptical as to how useful they actually are .
simulations addressing this issue will be presented elsewhere . in what follows we report what is obtained in practice .
our strategy is to chop the sf - wing into bins in @xmath18 and study the evolution of galaxies from the statistics of @xmath34 and @xmath39 in each bin .
the fact that @xmath18 correlates strongly with several physical and observational properties guarantees that galaxies inside each @xmath18 bin are intrinsically similar , which in turn implies that statistics over such bins is meaningful .
fig [ fig : mean_chemevol]a shows the mean mass - assembly functions obtained for each of the six @xmath18 bins illustrated in fig [ fig : bpt]b .
the systematics is evident to the eye .
galaxies with low @xmath18 are slower in assembling their stars than those with high @xmath18 .
this behavior reflects the fact that , although most of the mass is assembled at early times ( @xmath40 gyr ) , galaxies at the tip of the sf - wing have larger star formation rates ( sfr ) per unit mass in the recent past than those at the bottom of the wing .
this trend can be quantified in terms the ratio of the mean sfr in the last 60 myr to that over the galaxy s history , ie , the so called scalo s @xmath41 parameter ( eg , brinchmann 2004 ) .
we find @xmath41 to decrease by a factor of @xmath42 from the lower to the higher @xmath18 bins .
( sfrs are discussed in a07 , where it is also shown that our synthesis based sfrs agree very well with those derived from the h@xmath12 luminosity . )
alternatively , one may quantify this behavior by computing the time at which the mean @xmath34 function reaches 0.75 , ie .
, the time at which the mass involved in star - formation reaches 3/4 of the total .
we find that @xmath43 increases from @xmath44 gyr at @xmath45 ( bin a in fig [ fig : bpt]b ) , to 2 gyr at @xmath46 ( bin d ) , and 8 gyr for the highest @xmath18 bin . given that @xmath18 correlates strongly with @xmath24 ( zaritsky 2004 ; tremonti 2004 ; lee 2006 ) , this behavior is ultimately a signature of galaxy downsizing ( cowie 1996 ) . ) and ( b ) chemical evolution ( @xmath47 ) of stars in sf galaxies for the same six @xmath18-bins defined in fig [ fig : bpt]b .
each line represents a @xmath4-by-@xmath4 average over all galaxies in the bin .
a mild gaussian smoothing with fwhm = 0.2 dex in @xmath48 was applied to the mean @xmath34 and @xmath39 curves , just enough to smooth out discreteness effects associated to the 25 ages in the base.,scaledwidth=50.0% ] the chemical evolution of stars in sf galaxies is shown in fig [ fig : mean_chemevol]b . again , the main message is clear : galaxies with a metal poor ism are low are slow in their stellar chemical enrichment . in the lower @xmath18 bins , @xmath39 only reaches the current stellar metallicity level ( @xmath49 ) as recently as @xmath50 myr .
galaxies with larger @xmath18 , on the other hand , have flatter and systematically higher @xmath39 curves .
these massive systems have essentially completed their mass assembly and chemical evolution long ago , so that recent star - forming activity has a negligible impact upon @xmath39 .
the fact that bins of increasing @xmath18 have increasingly higher present @xmath47 values confirms that nebular and stellar metallicities are related ( seagal i ; gallazzi 2005 ; a07 ) .
we nevertheless warn the reader not to overinterpret the quantitative relation between these two quantities , given that they are derived by completely different and hardly comparable means . in this paper
we use [ oiii]/[nii ] as an indicator of @xmath18 ( equation [ eq : z_neb ] ) , because of its evident relation to the position of the points in the bpt diagram .
however , as emphasized by stasiska ( 2006 ) , strong line methods for deriving nebular abundances have been calibrated using giant hii regions and their relevance for integrated spectra of galaxies has still to be examined , because of the role of metallicity gradients and of the diffuse ionized medium in galaxies .
in addition , there is so far no study comparing nebular abundance calibrations with stellar abundances .
therefore , our @xmath18 values are not necessarily on the same scale as stellar abundances .
insofar as this paper is concerned , this is not a problem , since , irrespective of absolute values , it is the existence of a @xmath18-@xmath47 relation which is responsible for the ordering of the different chemical evolution curves in fig [ fig : mean_chemevol]b . , but binning sf galaxies by their stellar mass , using five 1 dex wide bins , centered at ( from bottom to top ) @xmath51 , 8.5 , 9.5 , 10.5 and 11.5 , which contain 583 , 4750 , 29056 , 46847 and 3550 galaxies , respectively.,scaledwidth=50.0% ] while fig [ fig : mean_chemevol ] illustrates the strong link between the stellar mass assembly and chemical evolution histories of galaxies and the present ionized ism metal content , @xmath18 is not the cause , but the product of galaxy evolution , so its use as an independent variable , although pedagogically useful , is questionable . if one had to chose a single quantity as the main driver of galaxy evolution this would surely be its mass ( eg . , tinsley 1968 ; larson 1974 ; pagel 1997 ) .
we have thus repeated the analysis above binning galaxies by @xmath24 instead of @xmath18 .
results are shown in fig [ fig : mean_chemevol_binnedbymass ] for five 1 dex - wide mass bins centered at @xmath52 . because of the flattening of the @xmath53 relation at high masses and the substantial scatter around the relation , particularly at low @xmath24 and @xmath18 , binning in @xmath24 or @xmath18 samples somewhat different populations of galaxies .
on the whole , however , this alternative grouping scheme leads to the same general scenario outlined above , with low @xmath24 galaxies being slower in their mass and chemical build up than more massive ones .
a caveat is that @xmath24 is not a perfect tracer of the depth of the potential well , since it does not account for the mass in gas nor dark matter . in any case , the signature of galaxy downsizing is evident in fig [ fig : mean_chemevol_binnedbymass ] : massive galaxies today formed most of their stars and metals long time ago , whereas low - mass galaxies are still forming stars and metals actively .
the signal of chemical evolution is present in all curves in figs [ fig : mean_chemevol]b and [ fig : mean_chemevol_binnedbymass]b .
curiously , however , this evolution does not seem to start from very low @xmath47 for massive , high @xmath18 galaxies ( eg , bins e and f ) .
we interpret this apparent lack of old , metal poor populations as due to a combination of lack of age resolution at high @xmath4 and an intrinsically fast evolution ocurring at early times . in low mass systems
( eg , bin a ) , @xmath39 starts is @xmath54 gyr long rise period at @xmath553 gyr , such that @xmath56 is of order unity .
the same process happening at @xmath57 gyr would yield @xmath58 , and consequently be much harder to resolve .
as shown above above , by @xmath59 gyr massive galaxies had already assembled most of their mass , so their chemical evolution must also have occurred early on .
this mixture of populations of widely different metallicities but similar ages results in @xmath47 values at @xmath57 gyr which are dominated by the metal rich stars already present at those times .
such an early and rapid formation period should lead to enhancement of elements synthesized in type ii sne ( worthey , faber & gonzlez 1992 ; thomas 2005 ) . though we can not map elemental abundances with our method
, we do detect a signal of increasing @xmath12 enhancement towards the bottom of the bpt diagram by means of systematically increasing residuals in spectral bands due to @xmath12 elements like mg ( a07 ) , indirectly supporting the scenario outlined above .
our analysis shows that the present ism abundance is intimately connected to the past star - formation and chemical enrichment history of a galaxy . whereas this is intuitively obvious from a physical point of view , it is not obvious at all that one should be able to recover this behavior from population synthesis analysis of integrated galaxy spectra of sdss - like quality .
we emphasize that , unlike previous studies such as that by bica ( 1988 ) , our fits impose no _ a priori _ chemical evolution constraints whatsoever upon the mixture of @xmath60 s and @xmath29 s in the base .
given the absence of constraints and the fact that the fits are completely independent of the emission line data used to compute @xmath18 , the organization of galaxies with different @xmath18 onto the clear and systematically distinct stellar chemical enrichment patterns seen in fig [ fig : mean_chemevol]b is remarkable , revealing that fossil methods based on integrated spectra have reached a level of maturity which would be unthinkable just a few years ago ( eg , cid fernandes 2001 ) . by taking the spectral synthesis output one step further than the trivial first moments of the @xmath4 and @xmath5 distributions , we have been able to uncover the chemical evolution of galaxies with an unprecedented level of detail for such a large and varied sample .
this significant achievement was only possible due to the fabulous statistics of the sdss , combined with state - of - the - art evolutionary synthesis models and reliable sfh recovery techniques .
the sheer number of galaxies allows us to recover an unequivocal chemical evolution signal which would be at best doubtful for individual galaxies or even small samples .
combined with our estimates of the star - formation histories , the empirically derived chemical evolution patterns should provide valuable constraints for galaxy evolution models .
we thank support from the brazilian agencies cnpq and fapesp .
we are greatly in debt with several colleagues and institutions around the globe who have contributed to this project by allowing access to their computers .
the sloan digital sky survey is a joint project of the university of chicago , fermilab , the institute for advanced study , the japan participation group , the johns hopkins university , the los alamos national laboratory , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , princeton university , the united states naval observatory , and the university of washington .
funding for the project has been provided by the alfred p. sloan foundation , the participating institutions , the national aeronautics and space administration , the national science foundation , the u.s .
department of energy , the japanese monbukagakusho , and the max planck society . | we explore the mass - assembly and chemical enrichment histories of star forming galaxies by applying a population synthesis method to a sample of 84828 galaxies from the sloan digital sky survey data release 5 .
our method decomposes the entire observed spectrum in terms of a sum of simple stellar populations spanning a wide range of ages and metallicities , thus allowing the reconstruction of galaxy histories .
a comparative study of galaxy evolution is presented , where galaxies are grouped onto bins of nebular abundances or mass .
we find that galaxies whose warm interstellar medium is poor in heavy elements are slow in forming stars .
their stellar metallicities also rise slowly with time , reaching their current values ( @xmath0 ) in the last @xmath1 myr of evolution .
systems with metal rich nebulae , on the other hand , assembled most of their mass and completed their chemical evolution long ago , reaching @xmath2 already at lookback times of several gyr .
these same trends , which are ultimately a consequence of galaxy downsizing , appear when galaxies are grouped according to their stellar mass .
the reconstruction of galaxy histories to this level of detail out of integrated spectra offers promising prospects in the field of galaxy evolution theories .
galaxies : evolution - galaxies : stellar content - galaxies : statistics |
hidden markov models ( hmms ) are stochastic models for systems with a set of unobserved states between which the system hops stochastically , sometimes emitting a signal from some alphabet , with probabilities that depend upon the current state .
the situation in which we are specifically interested is human mobility , partially observed , _
i.e. _ , occasional signals about a person s location . for example , consider the cells of a mobile phone network , from which a user can make calls . in this case
the states of a hmm are the cells , and the emitted signals are the cell itself , if a call is made by a particular user during each of a sequence of time intervals , or nothing ( 0 ) , if that user does not make a call . in the latter case , the state ( location ) of the user is ` hidden ' , and must be inferred , while in the former case , assuming no errors in the data , the ` hidden ' state is revealed by the call record . since these are data from a _ mobile _ phone network , a user can move from cell to cell .
although many analyses of human mobility have estimated no more than rather crude statistics like the radius of gyration , the fraction of time spent at each location , or the entropy of the timeseries of locations @xcite , others have used hmms to describe partially observed human mobility and have estimated their parameters @xcite . with short time steps , however , a standard hmm ( with time - independent parameters ) is not a plausible model , since human mobility behavior changes according to , for example , the time of day @xcite .
we would like to create , therefore , a hmm with time-_dependent _ parameters .
of course , allowing , for example , arbitrary transition / emission probabilities at each time step , would lead to an extremely underdetermined model .
rather , we need a model with only a few additional parameters to capture the time - dependence of human mobility . since the total numbers of trips @xcite and mobile phone calls @xcite vary with time of day and day of week , we develop a time - dependent hmm in which the non - trivial transition and emission probabilities are proportional to _ activity levels _ , _ i.e. _ , to some given functions modeling how active humans are at different times and places .
since the transition and emission probabilities in our hmm are not constant in time , it is a _ non - stationary _ hmm .
many generalizations of hmms have been considered previously , of course , as more faithful models of various real systems .
some of these are non - stationary : deng , for example , considers a class of models in which the emission probabilities are somewhat non - markovian , depending on a number of previous emissions , and also have polynomial - in - time trend components which are to be estimated @xcite . _ duration _ hmms ( dhmms ) , first suggested by ferguson @xcite , allow a sort of non - stationarity in the state transition process by including a randomly chosen duration each time the state changes , _
i.e. _ , a number of time steps without a transition away from that state
. this kind of model has been generalized to make the transition probabilities functions of the number of steps the system has been in the current state @xcite . in a different direction , since one can think of transitions between the hidden states with different emission probability distributions as a kind of non - stationarity , _ triplet markov chains _ ( tmcs ) include an auxiliary set of underlying states , each of which corresponds to a different stationary regime for a hmm @xcite .
our approach is different than that of dhmms and tmcs in that the time dependence of the transition and emission probabilities is not intrinsic and random , but rather exogenous and deterministic .
furthermore , unlike deng s models @xcite , we take the `` trend '' part of the time dependence to be given , not an additional ( set of ) parameter(s ) to be estimated . formally , our model consists of @xmath0 possible hidden states , and we denote by @xmath1^t$ ] the time series of @xmath2 hidden states ( for any @xmath3 , @xmath4=\{1,\ldots , n\}$ ] ) .
state transitions happen according to a sequence of matrices giving the conditional probabilities of transitions , @xmath5 at each state we observe an emission that takes a value from the set @xmath6\cup\{0\}$ ] , where @xmath7 denotes `` absence of an emission '' .
let @xmath8\cup\{0\})^t$ ] be the series of observed emissions .
the probability of emission @xmath9 at time @xmath10 , from state @xmath11 , is @xmath12 we define time varying _ activity levels _ for state @xmath11 by a pair of functions with non - negative values , @xmath13\to [ 0,1]^2 $ ] , the _ activity functions _ , which modulate transitions and emissions from the state , respectively .
given a state @xmath11 , the transition probabilities from that state are functions of _ transition parameters _
@xmath14 , @xmath15 $ ] , and the activity level : @xmath16 , i\neq j } \tau_{ij } & \text{if } i = j , \end{array } \right.\ ] ] subject to the constraints that for all @xmath17 $ ] and all @xmath18 $ ] , @xmath19 , i\neq j}\tau_{ij } \leq 1.\ ] ] in practice we may have _ a priori _ knowledge that some transitions do not occur , so that @xmath20 ( and @xmath21 ) have some entries set to @xmath7 .
for example , with 10 minute time intervals and states representing cells in a mobile phone network , transitions between sufficiently distant cells are precluded .
similarly , we assign emission parameters @xmath22 , @xmath23 $ ] , to each state @xmath11 .
the emission probabilities are @xmath24}\epsilon_{s j } & \text{if } s=0 , \end{array } \right.\ ] ] subject to the constraints that for all @xmath17 $ ] and all @xmath18 $ ] , @xmath25}\epsilon_{s j}\leq1.\ ] ] denote the initial distribution over states by @xmath26 , subject to the constraints @xmath27 and @xmath28 } \pi_j = 1.\ ] ] were this a typical hidden markov model , we could estimate its parameters using the baum - welch algorithm @xcite . since it is not ,
we develop a novel expectation maximization ( em ) algorithm @xcite to estimate the parameters @xmath29 , given @xmath30 , @xmath31 and @xmath32 , as follows .
the expectation maximization algorithm maximizes , at each iterative step , the ( expected ) log - likelihood function described below .
let @xmath33 be the set of all possible time series of states and let @xmath34 be the estimate of @xmath35 at the @xmath36-th iteration of the algorithm . @xmath37
the algorithm begins by initializing the parameter estimates in the first ( @xmath38 ) iteration .
then the @xmath39 iteration consists of two steps : 1 .
[ itr1 ] compute the _ expectation value _ of the log - likelihood , using the current ( @xmath40 ) estimate for the parameters : @xmath41 \pr(x\mid y;{\hat\theta}^k),\ ] ] where @xmath42 means @xmath43 in a probability distribution parametrized by @xmath35 .
[ itr2 ] find the parameters that _ maximize _ the expected log - likelihood : @xmath44 subject to constraints in the inequalities , and .
as in the regular baum - welch algorithm , we express our computations in terms of certain conditional probabilities based on the parameters estimated at the @xmath40 iteration , @xmath45 some reindexing of eq . yields the following expression for @xmath46 in terms of these probabilities : @xmath47}\log(\pi_j)\gamma^k_j(1 ) + \sum_{i , j\in[n]}\sum_{t=1}^{t-1 } \log\bigl(a_{ij}(t)\bigr)\xi^k_{i j}(t ) + \sum_{j\in[n]}\sum_{t=1}^{t } \log\bigl(b_{y_t j}(t)\bigr)\gamma^k_j(t).\ ] ] we iterate steps [ itr1 ] and [ itr2 ] , for which we compute @xmath48 and @xmath49 in eqs . , and @xmath46 in eq . above .
we continue until some standard of convergence is achieved .
we then output the final @xmath34 as our estimate of @xmath35 .
[ mainthm ] there is a constrained expectation maximization algorithm giving a sequence of estimates @xmath50 that converges to a critical point of the likelihood function , which is the maximum likelihood estimate @xmath51 for the observed sequence @xmath30 when the initial guess is sufficiently close .
further , to achieve a precision @xmath52 in the estimates , the time complexity of the algorithm is @xmath53 . in particular , for a fixed precision @xmath52 , the time complexity is @xmath54 .
suppose we have the estimates @xmath55 defined in eq . from step @xmath36 of the algorithm .
we proceed to compute @xmath56 and @xmath57 in eqs . to begin the next iteration .
just as in the regular baum - welch algorithm , we apply dynamic programming . denote the @xmath40 estimate of the transition matrix by @xmath58 , where @xmath59 similarly , @xmath60 } \hat\epsilon^k_{sj } & \text{if } s=0 .
\end{array } \right .\ ] ] it is convenient to define @xmath61 to be the diagonal matrix with @xmath62 entry @xmath63 . now compute two sequences of ( co)vectors , @xmath64 and @xmath65 , recursively , as follows : @xmath66 then denotes the diagonal matrix whose @xmath62 entry is @xmath67 if @xmath68 and @xmath69 otherwise . ]
@xmath70 the estimates for the initial probabilities @xmath71 are the same as in the normal baum - welch algorithm , as is clear from the expression for @xmath46 in eq . .
thus @xmath72 now notice that all the constraints in and necessary to define @xmath46 are implied by the strongest constraints : for all @xmath17 $ ] , @xmath73 where @xmath74 } \ :
f_{j}(t)$ ] , and @xmath75 } \epsilon_{s j } \leq 1,\ ] ] where @xmath76 } \ : g_{j}(t)$ ] . consider the computation of @xmath77 , for @xmath15 $ ] .
it should lie in the domain @xmath78 defined by constraints and the non - negativity of the parameters .
since these constraints are independent for different @xmath11s , we can consider each @xmath11 separately , and find the optimal parameters @xmath79 by computing the critical points of @xmath46 relative to @xmath80 .
using eqs . and , @xmath81 if the left sum in eq . , @xmath82 , the derivative is nonpositive , so @xmath46 is weakly decreasing and @xmath83 gives its largest value . if the right sum in eq . , @xmath84 , the derivative is nonnegative , so @xmath46 is weakly increasing and takes its maximum value when @xmath79 is as large as possible , _
i.e. _ , when it saturates constraint .
we will show how to handle this situation after discussing the generic case which we do next . assuming then that neither sum in eq .
is @xmath7 , to find the stationary points of @xmath46 we set eq . to @xmath7 and
solve for @xmath79 .
specifically , @xmath79 must satisfy @xmath85 we note that if @xmath86 , which makes the transition probabilities , @xmath87 , time independent , then the solution to eq .
is the familiar baum - welch solution : @xmath88 for all @xmath89 $ ] . for non - constant activity functions
, however , the solution is more complicated . since the right side of eq .
is manifestly independent of @xmath90 , the left side must be , too .
let @xmath91 where the last expression uses the antiderivative convention that for a function of @xmath10 denoted by a letter in lower case , the corresponding upper case letter is upper case @xmath92 ; @xmath93 is upper case @xmath94 ; @xmath95 is upper case @xmath96 ; @xmath0 is upper case @xmath97 ; @xmath98 is upper case @xmath99 . ] represents its sum over its domain of definition ( @xmath100 to @xmath101 in this case ) . now
@xmath102 if we denote the probability of _ moving _ away from state @xmath11 by @xmath103 we can write @xmath104 . substituting in eq . , this gives : @xmath105 or equivalently : @xmath106 we must solve this equation for @xmath107 , whence we can use eq . to solve for each of the @xmath79 .
since @xmath107 is nonnegative and constraint must hold , we recast these conditions in terms of @xmath107 as : @xmath108 let @xmath109\,\mid\,\xi^k_{jj}(t)\neq0 } \ : f_{j}(t)\,\,\leq f^*_{j}.\ ] ] each of the terms in the sum on the right side of eq . is strictly decreasing in @xmath110 when it is well - defined ( @xmath111 ) .
values of @xmath110 just larger than @xmath112 make the sum arbitrarily large , and as @xmath110 increases from that value , the sum decreases monotonically to 0 , so exactly one value of @xmath113 will satisfy eq . .
we can solve the equation numerically to find this value , call it @xmath114 . if @xmath115 , splitting it proportionally to @xmath116 according to eq .
gives the _ unique _ critical point @xmath80 of @xmath46 .
we can compute explicitly the hessian of @xmath117 with respect to the @xmath79 ; its components are : @xmath118 thus , as a matrix the hessian can be written as the sum of two matrices : @xmath119 where @xmath120 is the vector of all @xmath69s .
each of the matrices on the right is negative semi - definite , so the hessian is also .
thus the ( unique ) critical point we found in this case is a global maximum of @xmath117 in @xmath121 and hence the choice for @xmath122 . if the solution does not satisfy the original constraint , _
i.e. _ , @xmath123 , or if the right side of eq . is @xmath7 , the maximum will be on the boundary of @xmath121 .
thus we maximize @xmath117 subject to the boundary constraint @xmath124 } \tau_{ij } = \frac{1}{f^*_j}.\ ] ] this is in the form of the constraint in the regular baum - welch algorithm with 1 replaced by @xmath125 and the self - transition probability set to 0 .
thus the critical @xmath79 can be computed as in the baum - welch algorithm , for all @xmath126 . ] with @xmath127 replaced by @xmath128 and the solution divided by @xmath129 : @xmath130 this is the unique critical point in this case , and the global maximum of @xmath117 in @xmath121 , by the same argument as in the baum - welch algorithm .
this becomes the choice for @xmath122 .
we turn to the computation of @xmath131 , @xmath132 $ ] .
as before , we begin by finding the stationary points of @xmath46 , now relative to @xmath133 defined by constraints and the non - negativity of these parameters . using eqs . and gives : @xmath134}\epsilon_{lj}}\delta_{0,y_t}.\ ] ] as we did for eq .
, we must consider the situations when either of the sums in eq .
vanishes . when the left sum is @xmath7 , a extreme value is given by @xmath135 , and when the right sum is @xmath7 , constraint is saturated . assuming neither of the sums vanishes , we find the stationary points by solving @xmath136}\epsilon_{lj}}\delta_{0,y_t}.\ ] ] solution to the _ emission _ equations , eqs . , follows using the same steps as for the _ transition _ equations , eqs . .
we first denote the probability of emission @xmath132 $ ] from state @xmath11 by : and every state @xmath11 emits either the signal @xmath11 or @xmath7 ; in other words , @xmath137 if @xmath138 $ ] .
this simplifies the following expressions : for @xmath10 such that @xmath139 , @xmath140 ( the state is @xmath11 with certainty if the observed emission is @xmath11 ) , _ i.e. _ , @xmath141 , and @xmath142 if @xmath138 $ ] . ]
@xmath143 in terms of which we rewrite eq . as @xmath144}\epsilon_{lj}}.\
] ] we define ( independent of @xmath9 ) @xmath145 we also define the probability of any _ non - zero _ emission from state @xmath11 , @xmath146}\lambda^k_{lj}(t),\ ] ] which , used in eq . , gives @xmath147 let @xmath148\,\mid\,\lambda^k_{0j}(t)\neq0}\:g_{j}(t)\,\,\leq g^*_{j}.\ ] ] as before , there is exactly one solution @xmath149 to eq . .
we can find it numerically , and if @xmath150 , we use eq . to find all the @xmath151 , which will then satisfy constraint .
again we can compute the hessian explicitly to confirm that this is a global maximum of @xmath117 , now in @xmath152 , and hence the choice for @xmath153 .
if @xmath154 , or if the right side of eq . is @xmath7 , we must find instead the critical point on the boundary of @xmath152 : @xmath155 } \epsilon_{sj } = \frac{1}{g^*_j}.\ ] ] again , this is in the form of the constraint in the regular baum - welch algorithm .
accordingly , we set the critical @xmath156 to the baum - welch estimate , rescaled by @xmath157 : @xmath158 this is the unique critical point and the global maximum of @xmath117 in @xmath152 , and therefore the choice for @xmath131 in this case .
we have shown how to find @xmath159 maximizing @xmath46 in eq . .
this algorithm converges as claimed because it is an instance of expectation maximization . to understand its time complexity we must consider the numerical solution of eqs . and .
to simplify notation we rewrite eq . in terms of @xmath160 , and a function @xmath161 , @xmath162 as @xmath163 .
as an initial estimate for the root , @xmath164 , we can use @xmath165 such that @xmath166 where @xmath167 $ ] satisfies @xmath168 and @xmath169 .
@xmath170 since we found it using only one of the nonnegative terms in the sum in eq . .
now recall that @xmath171 for @xmath172 $ ] , and @xmath173 } \mu^k_j(t ) = \sum_{t\in [ t-1 ] } \sum_{r\neq j}\xi^k_{rj}(t ) \le \sum_{t\in [ t-1 ] } \gamma^k_{j}(t ) \le t-1.\ ] ] thus each term in the sum in eq .
is no more than @xmath174 at @xmath175 this is @xmath176 , so @xmath177 , which implies @xmath178 using newton s method , once we have an initial estimate `` sufficiently close '' to the root of @xmath163 , the time complexity to find it with error less than @xmath52 is @xmath179 , where the @xmath2 comes from the cost of evaluating @xmath161 and @xmath180 at each iteration ; in practice this is how we would find the root .
since the length of the interval in is @xmath181 , however , the bisection method gets us to precision @xmath52 with @xmath182 steps , with total cost @xmath183 ; thus this is the total complexity .
we need to solve eqs .
and @xmath0 and @xmath95 times , respectively , at each iteration , which thus adds @xmath184 to the @xmath185 complexity of the computations for @xmath186 and @xmath187 .
thus the time complexity for the whole algorithm is @xmath188 .
to demonstrate the effect of the activity functions we consider a simple model with @xmath189 states and the same number of possible emissions ( @xmath190 ) . from any state @xmath11
, we only allow an emission to be either its own label @xmath11 or @xmath7 , _
i.e. _ , @xmath191 for @xmath192 $ ] , so a non - zero emission uniquely identifies the state that emits it .
we choose random transition and emission parameters : @xmath193 where the omitted values are the components for which @xmath194 .
we generate sequences of length @xmath195 ( we may think of this as @xmath196 weeks , with an observation every @xmath197 minutes ) .
we consider activity functions with variations that may approximate observed data , _
i.e. _ , periodic variations with a period of @xmath198 ( one day ) .
specifically , our numerical simulations use the following three functions : we generate a random sequence of states , @xmath202 , and resulting emissions , @xmath30 , using the transition and emission parameters above , and a pair of activity functions ( a list of these pairs is shown in table [ table2 ] ) . before computing the sequence of parameter estimates ,
we need to specify how we compute initial estimates to start the iteration .
this can only depend on the observed emission sequence @xmath30 , since in any real scenario @xmath202 is unknown . as a first guess , for this simple model
, we interpolate the state sequence @xmath202 as follows : for every pair of successive non - zero emissions , there is a segment of zeros ( no emission ) separating them .
we divide each such segment into two subsegments : let @xmath203 $ ] be the emission immediately preceding the segment , and @xmath204 $ ] be the emission immediately following the segment .
the second subsegment starts at the first time step after the one where @xmath205 first attains its maximum value on the segment ( _ i.e. _ , a time at which there is the maximum probability of hopping from state @xmath11 to state @xmath90 ) .
we assign state @xmath11 to the time steps in the first subsegment and the state @xmath90 to those in the second .
if the emission sequence @xmath30 starts with a segment of zeros , then that segment is assigned the value of the first non - zero emission ; similarly a terminal sequence of zeros is given the value of the last non - zero emission .
denote the interpolated states by @xmath206 , @xmath18 $ ] . from @xmath207 ,
we compute the estimate @xmath208 for the initial distribution over the states @xmath209 by their frequencies of occurrence . for the initial @xmath79 estimate , @xmath210
, we use the method described in the proof of theorem [ mainthm ] to solve eq . , using @xmath211 , including @xmath194 . for the initial @xmath212 estimate , @xmath213
, we also use the method described in the proof of theorem [ mainthm ] to solve eq . , using @xmath214 . to understand the performance of the algorithm in theorem [ mainthm ]
, we need a measure of the error between the estimates and the real parameter values .
the _ relative entropy _ is one measure for a stationary hmm . in our case
we need to account for the time variation of the transition and emission probabilities , @xmath215 and @xmath216 .
hence we define a modified version of a relative entropy error criterion , the _ averaged relative entropy_. let @xmath217 and @xmath218 , @xmath18 $ ] , be two finite sequences of discrete probability distributions on a finite set @xmath219 .
the _ averaged relative entropy _ ( @xmath220 ) of @xmath221 with respect to @xmath222 is @xmath223}\mathsf{re}(p_t , q_t),\ ] ] where the usual relative entropy ( @xmath224 ) is given by @xmath225 thus the error function that we compute for given @xmath21 and estimate @xmath226 is @xmath227 where @xmath215 and @xmath228 are related through @xmath205 to @xmath79 and @xmath229 by eq .
and eq . , respectively
similarly , for @xmath230 and estimate @xmath231 , the error function is @xmath232 where @xmath216 and @xmath233 are related through @xmath234 to @xmath235 and @xmath236 by eq . and eq . , respectively .
( remember that we are considering the simple case in which the emission @xmath237 when the state @xmath238 . ) to - & a & b & c & d & e & f & g & h + - @xmath205 & @xmath239&@xmath239 & @xmath239 & @xmath240 & @xmath241 & @xmath242 & @xmath240 & @xmath241 + @xmath234&@xmath241&@xmath240&@xmath239&@xmath240&@xmath241&@xmath239&@xmath239&@xmath239 + - for each set of pairs of activity functions in table [ table1 ] we run the algorithm for @xmath243 iterations .
figures [ figtrans ] and [ figemis ] plot the averaged relative entropy for the parameter estimates as a function of iteration step .
the labels ( a)(h ) to the right of each plot appear in the order of the final error values .
we do not provide a plot showing convergence of @xmath244 since the only noticeable trend is that if they converge to an exact state value , it is usually to the initial state of the interpolated sequence @xmath207 . in each case
the error for both the transitions and the emissions decreases to small values . since the @xmath220 depends on the activity functions as well as on the parameters and their estimates , we need to compute a baseline error value for each case . for the parameters @xmath21 and a specific choice of @xmath245
it is : @xmath246,\ ] ] where @xmath215 and @xmath247 are related through @xmath205 to @xmath79 and @xmath248 , respectively , by eq . , and where @xmath249 $ ] denotes the expectation over uniformly random @xmath250 . to estimate this expectation value , we compute the average @xmath220 of the parameters with respect to 1000 independently chosen sets of random parameters ( rather than their estimates from our algorithm ) , for each case ( a)(h ) . for the emission parameters we compute baselines the same way , using 1000 uniformly random values @xmath251 to estimate @xmath252,\ ] ] where
@xmath216 and @xmath253 are related through @xmath234 to @xmath235 and @xmath254 , respectively , by eq . , and where @xmath249 $ ] denotes the expectation value over uniformly random @xmath251 .
the baseline averages thus obtained for function pairs in table [ table1 ] are recorded in table [ table2 ] , where the row labels indicate the parameters being baselined . to - & a & b & c & d & e & f & g & h + - @xmath255&@xmath256 & @xmath257&@xmath258 & @xmath259&@xmath260 & @xmath261&@xmath262 & @xmath263 + @xmath264&@xmath260&@xmath265&@xmath266&@xmath267&@xmath268&@xmath269&@xmath270&@xmath271 + - we plot these baseline errors as horizontal lines in figures [ figtrans ] and [ figemis ] .
most of these are too close to be distinguishable ; indeed they are all @xmath272 , in contrast to the estimation errors plots which are almost all smaller by at least an order of magnitude , and in most cases by 3 or 4 , indicating very good parameter estimates .
furthermore , the relative quality of the estimates can be understood : case ( c ) is the standard hmm , for which our algorithm reduces to the baum - welch algorithm @xcite .
cases ( a ) and ( b ) have greater errors , which is not surprising since they have non - constant emission activity functions , oscillating in value up to 1 .
this means that for each of these cases , non - zero emissions are lower probability events , so there is less information in @xmath30 . possibly surprising is the fact that when the transition activity function is non - constant , cases ( d)(h ) , the errors are _ smaller _ than in the standard hmm case .
but this happens because state changing transitions are reduced , so that each non - zero emission observed provides more information . and among these cases , those with varying emission activity levels have larger errors than those without . , for successive iterations @xmath36 .
labels to the right are of activity function pairs from table [ table1 ] , and are displayed in the order of the final error values .
baseline errors , @xmath255 , from table [ table2 ] are shown as horizontal lines . ] , for successive iterations @xmath36 .
labels to the right are of activity function pairs from table [ table1 ] , and are displayed in the order of the final error values .
baseline errors , @xmath264 , from table [ table2 ] are shown as horizontal lines . ] | we define a hidden markov model ( hmm ) in which each hidden state has time - dependent _ activity levels _ that drive transitions and emissions , and show how to estimate its parameters .
our construction is motivated by the problem of inferring human mobility on sub - daily time scales from , for example , mobile phone records . |
the interaction of a collection of atoms with a radiation field has always been an important topic in quantum optics . the dicke model ( dm ) which describes interaction of @xmath0 identical two level atoms with a single radiation field mode ,
established the importance of collective effects of atom - field interaction , where the intensity of the spontaneously emitted light is proportional to @xmath1 rather than @xmath0 @xcite .
the spatial dimensions of the ensemble of atoms are smaller than the wavelength of the radiation field . as a result , all the atoms experience the same field and this gives rise to the collective and cooperative interaction between light and matter .
the dm exhibits a second - order quantum phase transition ( qpt ) from a non - superradiant normal phase to a superradiant phase when the atom - field coupling constant exceeds a certain critical value @xcite .
the experimental observation of the qpt predicted in the dm required that the collective atom - photon coupling strength to be of the same order of magnitude as the energy separation between the two atomic levels . in conventional atom - cavity setup
this condition was impossible to satisfy until it was observed recently in a trapped bose - einstein condensate ( bec ) in an optical cavity @xcite . in the bec setup ,
the two spin states of the original dm are the two momentum states of the bec which are controlled by the atomic recoil energy and raman pumping schemes .
this approach is similar to a novel scheme proposed by dimer et .
an important aspect of these experimental developments is the possibility to explore exotic phases mediated by the cavity field .
the superradiance phase transition in a bec is accompanied by self - organization of the atoms into a checker board pattern @xcite .
interesting equilibrium and non - equilibrium phases have been predicted in the dm with bec @xcite , including crystallization and frustration @xcite , as well as spin glass phase @xcite .
multimode dm has also been explored recently , revealing interesting physics such as abelian and non - abelian gauge potentials @xcite , spin - orbit induced anomalous hall effect @xcite , and prediction of the nambu - goldstone mode @xcite .
an interesting extension of the bec dicke model is the optomechanical dicke model which has been proposed for detection of weak forces @xcite . in the present paper ,
we investigate the non - equilibrium properties of the two - atom dicke model , which can be realized by two species bec in an optical cavity . apart from the usual non - equilibrium normal and inverted phases , the dynamical phase diagram reveals a new kind of non - equilibrium mixed phase .
this gives rise to a new quantum phase transition from the mixed phase to the superradiant phase by manipulation of the two distinct atom - photon coupling strengths .
in addition , we show that a quantum phase transition from the non - superradiant phase to the superradiant phase is not allowed for certain values of the atom - photon coupling strengths of the two set of atoms .
[ fig1 ] we consider two different ensembles of @xmath2 and @xmath3 atoms coupled simultaneously to the quantized field of an optical cavity mode ( fig.1 ) .
the two sets of atoms have transition frequencies @xmath4 and @xmath5 while the frequency of the cavity mode is @xmath6 .
the cavity is pumped by an transverse external laser with frequency @xmath7 .
the light - matter coupling strengths for the two sets of atoms are @xmath8 and @xmath9 .
these coupling strengths @xmath8 and @xmath9 can be written as @xmath10 ( @xmath11 ) , @xmath12 is the single atom - cavity mode coupling while @xmath13 is the transverse pump beam rabi frequency .
the detuning @xmath14 is considered to be large so as to avoid spontaneous emission .
the effective hamiltonian of the system takes the form of a two - atom dicke model with @xmath15 where @xmath16 is the effective collective spin of length @xmath17 for the two sets of atom and @xmath18 .
we now discuss the non - equilibrium dynamics arising from the above two - atom dicke model.the semi - classical equations of motion for the system are given by @xmath19 @xmath20 @xmath21 @xmath22 @xmath23 here @xmath24 is the decay rate of the cavity photons . in addition
the magnitude of pseudo - angular momentum is conserved , @xmath25 @xmath26 the long time steady state solutions from the equations of motion can be studied with @xmath27 and @xmath28 .
these fixed point solutions can be stable or unstable . separating @xmath29 , @xmath30 @xmath31 ,
one obtains the steady state equations as @xmath32 @xmath33 @xmath34 @xmath35 @xmath36 @xmath37 an analysis of these equations leads us to four types of steady states , namely @xmath38 .
the state @xmath39 is the normal phase while @xmath40 is the inverted phase .
the states @xmath41 and @xmath42 are called mixed phases . as we shall show later that these mixed phases generate interesting non - equilibrium phase diagrams .
the critical coupling strength corresponding to the onset of superradiance starting from the normal , inverted or mixed phase is obtained by putting @xmath43 @xmath31 .
this leads us to the following possible critical constants @xmath44 @xmath45 @xmath46 @xmath47 @xmath48 @xmath49 @xmath50 @xmath51 @xmath52 @xmath53 @xmath54 @xmath55 ) in the @xmath56 plane
. the parameters used are @xmath57 @xmath58 @xmath59 @xmath58 @xmath60 .
the white region is the non - superradiant normal phase while the superradiant phase is indicated by the contours .
the darker region in the contour corresponds to low superradiance while light region corresponds to high superradiance .
the left plot is the @xmath61 root while the right plot is the @xmath62 root . ]
[ fig2 ] these set of expressions reveals one interesting point that the critical coupling strength for one set of atoms depends on the coupling strength of the other set of atoms . for a given set of @xmath63 and @xmath64 , eqns .
[ n1]-[m2 ] determines the boundary between the nonsuperradiant ( normal / inverted / mixed ) and superradiant phase . a trivial manipulation of eqns . [ s1]-[s6 ]
leads us to the following equation for @xmath65 and @xmath66 , @xmath67 now there are two possibilities depending on whether @xmath68 and @xmath69 or @xmath70 and @xmath71 .
the first condition implies that both the set of atoms are in the non - superradiant phase .
the second solution corresponds to the case where the first set of atoms are in the non - superradiant normal phase while the second set of atoms are in the superradiant phase . substituting the second expression for @xmath64 from above in the expression for @xmath72,one obtains @xmath73
this implies that by keeping one coupling strength arbitrarily low , one could enter the superradiant phase by manipulating the second coupling strength alone .
this point would be more clear when we discuss the dynamical phase diagrams in the next section .
in this section , we explore the fluctuation dynamics above the steady state ( fixed points ) .
in particular , we will consider the instability of the normal ( @xmath74 ) , inverted ( @xmath75)and mixed phases ( @xmath76 or @xmath77 ) . to this end
, we write @xmath78 , @xmath79 @xmath31 , where @xmath80 , @xmath81 and @xmath82 @xmath31 . substituting these into eqns.(4)-(6 ) , one obtains the linearized equations @xmath83 @xmath84 @xmath85 ( eqns.[o3 ] ) in the @xmath56 plane .
the parameters used are @xmath57 @xmath58 @xmath59 @xmath58 @xmath60 .
the left plot is the @xmath61 root while the right plot is the @xmath62 root .
the white region is the non - superradiant mixed phase @xmath86 while the superradiant phase is indicated by the contours . ]
[ fig3 ] we now write @xmath87 , @xmath88 and @xmath89 and equate coefficients with the same time dependence to obtain algebraic equations for @xmath90 , @xmath91 , @xmath92 , @xmath93 , @xmath94 and @xmath95 . the corresponding self consistency equations yields a quadratic equation for @xmath96 , whose roots characterize the possible instabilities for @xmath97 .
these instabilities describe the boundaries in the emerging dynamical phase diagrams .
in particular , the various boundaries between exponentially growing and decaying fluctuations are given as : normal phase : @xmath98 inverted phase : @xmath99 mixed phase 1 : @xmath100 mixed phase 2 : @xmath101 note that the `` inverted phase '' is the inversion of the `` normal phase '' around @xmath102 boundary while `` mixed phase @xmath103 '' is the mirror inversion of `` mixed phase @xmath86 '' .
the contour plot of the stable roots @xmath104 of eqns.[o1 ] as a function of @xmath8 and @xmath9 for the normal phase is shown in fig.2 .
fig.2(a ) shows the boundary separating the non - superradiant phase and the superradiant phase for the @xmath61 root .
this boundary is the curve that joins @xmath105 ( with @xmath106 ) and @xmath107 ( with @xmath108 ) .
below this boundary is the non - superradiant normal phase while above this curve is the superradiant phase .
as we move along the @xmath109-axis @xmath110 , we reach the superradiant phase at @xmath111 .
this analysis agrees with our steady state analysis of the previous section . thus along the @xmath112 or the @xmath109 axis
, the system behaves as if only one set of atoms are present . in any other direction ,
both set of atoms contribute to the dynamics .
note that white region is the non - superradiant normal phase while the superradiant phase is indicated by the contours .
the darker region in the contour corresponds to low superradiance while light region corresponds to high superradiance .
fig2b shows the plot of @xmath62 root .
the combination of @xmath61 and @xmath62 determine the complete boundary between the non - superradiant phase and superradiant phase .
the phase diagrams of the inverted phase ( not shown ) is the mirror inversion of the normal phase . in a similar manner
, one can determine the dynamical phase diagrams for the mixed phases . in fig.3(a ) and 3(b ) , we demonstrate this for the mixed phase @xmath86 .
a new kind of dynamical phase diagram emerges for the mixed phase .
the phase diagram now splits into two distinct superradiant regimes separated by the non - superradiant phase .
the two critical points @xmath105 @xmath113 and @xmath107 @xmath110 along the @xmath112 and @xmath109 axis are still the same . in fig.3a ,
@xmath61 root is shown and on moving along the @xmath112 axis ( increasing @xmath8 ) , we encounter the usual superradiant phase with increasing energy . on the other hand ,
moving along the @xmath109 axis ( increasing @xmath9 ) , we get a superradiant phase of constant low energy .
there are regions in the phase diagram , where even when @xmath114 and @xmath115 , the system stays in the non superradiant phase .
interestingly for @xmath116 , the superradiant phase can never be reached .
infact the energy landscape in the @xmath56 plane gives an impression of anti - crossing of energy levels .
fig.3(b ) shows the plot of @xmath62 root whose behavior is opposite to that of the @xmath61 root . a superradiant phase with decreasing energy is encountered along the @xmath109 axis while a constant high energy phase is encountered along the @xmath112 axis .
note that if we choose @xmath117 , then the energy plots of fig.2 and fig.3 becomes asymmetric ( figure not shown ) .
the current predictions can be tested in an experiment similar to that of ref.@xcite but with two species condensate .
in addition one has to look into the long duration of these experiments beyond the @xmath118 @xmath119 time scale as noted in ref.@xcite
in summary , we have investigated the non - equilibrium quantum phases of a two - atom dicke model , which is realized in a collection two set of cold atoms coupling simultaneously to a single quantized cavity mode . within the framework of the non - equilibrium two - atom dicke model
, we reveal a rich and new set of phase diagrams .
we have shown the existence of a new kind of quantum phase transition from the non - superradiant mixed phase ( where one set of atoms are in the normal phase while the other set is in the inverted phase ) to the superradiant phase .
in addition , we have demonstrated that in the quantum phase diagram of the mixed phase , there are regions where the superradiant phase can not exist even if the light - matter coupling constants of both set of atoms are above the critical value .
these predictions can be realized in a two species cold atoms in an optical cavity .
a. bhattacherjee acknowledges financial support from the department of science and technology , new delhi for financial assistance vide grant sr / s2/lop-0034/2010 .
dicke , phys . rev . * 93 * , 99 ( 1954 ) .
k. hepp and e.h .
lieb , ann .
phys.(ny ) , * 76*,360 ( 1973 ) .
wang and f. t. hioes , phys .
a , * 7 * , 831 ( 1973 ) .
hioes , phys .
a , * 8 * , 1440 ( 1973 ) .
clive emary and tobias brandes , phys .
e * 67 * , 066203 ( 2003 ) .
k. baumann , c. guerlin , f. brennecke and t. esslinger , nature ( london ) , * 464 * , 1301 ( 2010 ) . k. baumann , r. mottl , f. brennecke and t. esslinger , phys . rev
. lett . * 107 * , 140402 ( 2011 ) .
f. brennecke , r. mottl , k. baumann , r. landig , t. donner and t. esslinger , proc .
sci.usa , * 110 * , 11763 ( 2013 ) .
h. ritsch , p. domokos , f. brennecke , and t. esslinger , rev .
phys . * 85 * , 553 ( 2013 ) .
f. dimer , b.estienne , a. s. parkins and h.j .
carmichael , phys .
a , * 75 * , 013804 ( 2007 ) .
d. nagy , g. konya , g. szirmai and p. domokos , phys .
lett . * 104 * , 130401 ( 2010 ) . m.j .
bhaseen , j. mayoh , b.d .
simons and j. keeling , phys .
a , * 85 * , 013817 ( 2012 ) .
ni liu , jinling lian , jie ma , liantuan xiao , gang chen , j .- q .
liang and suotang jia , phys .
rev . a , * 83 * , 033601 ( 2011 ) .
s. gopalakrishnan , b.l . lev and p.m. goldbart , nat
. phys . * 5 * , 845 ( 2009 ) .
s. gopalakrishnan , b.l .
lev and p.m. goldbart , phys .
lett . , * 107 * , 277201 ( 2011 ) .
p. strack , s. sachdev , phys .
lett . * 107 * , 277202 ( 2011 ) .
m. buchhold , p. strack , s. sachdev and s. diehl , phys .
a , * 87 * , 063622 ( 2013 ) .
a. andreanov and m. mller , phys .
lett . , * 109 * , 177201 ( 2012 ) .
j. larson and s. levin , phys .
lett . * 103 * , 013602 ( 2009 ) .
j. larson , phys .
a , * 81 * , 051803 ( 2010 ) .
jingtao fan , zhiwei yang , yuanwei zhang , jie ma , gang chen and suotang jia , phys .
a , * 89 * , 023812 ( 2014 ) .
neha aggarwal , sonam mahajan and a. bhattacherjee , jour . of mod .
, * 60 * , 1263 ( 2013 ) .
priyanka verma , a. bhattacherjee and man mohan , canad .
* 90 * , 1223 ( 2012 ) . | in this paper , we investigate the non - equilibrium quantum phases of the two - atom dicke model , which can be realized in a two species bose - einstein condensate interacting with a single light mode in an optical cavity . apart from the usual non - equilibrium normal and inverted phases , a non - equilibrium mixed phase is possible which is a combination of normal and inverted phase . a new kind of quantum phase transition is predicted from non - superradiant mixed phase to the superradiant phase which can be achieved by tuning the two different atom - photon couplings .
we also show that a quantum phase transition from the non - superradiant mixed phase to the superradiant phase is forbidden for certain values of the two atom - photon coupling strengths .
* keywords : * non - equilibrium dicke model , quantum phase transition . |
quantum thermodynamics is the study of heat and work dynamics in quantum mechanical systems @xcite . in the extreme limit of small systems with only a few degrees of freedom , both the finite - size effect and quantum effects
influence the thermodynamic properties of the system dramatically @xcite . the traditional thermodynamic theory based on classical systems of macroscopic size
does not apply any more , and the quantum mechanical generalization of thermodynamics becomes necessary .
the interplay between thermodynamics and quantum physics has been an interesting research topic since 1950s @xcite . in recent years , with the developments of nanotechnology and quantum information processing , the study of the interface between quantum physics and thermodynamics begins to attract more and more attention @xcite .
studies of quantum thermodynamics not only promise important potential applications in nanotechnology and quantum information processing , but also bring new insights to some fundamental problems of thermodynamics , such as maxwell s demon and the universality of the second law @xcite . among all the studies about quantum thermodynamics ,
a central concern is to make quantum mechanical extension of classical thermodynamic processes and cycles @xcite .
it is well know that in classical thermodynamics there are four basic thermodynamic processes : adiabatic process , isothermal processes , isochoric process , and isobaric process @xcite .
these four processes correspond to constant entropy , constant temperature , constant volume , and constant pressure , respectively . from these four basic thermodynamic processes
, we can construct all kinds of thermodynamic cycles , such as carnot cycle , otto cycle , brayton cycle , et al @xcite .
among all the four kinds of basic thermodynamic processes , adiabatic process has been extended to quantum domain , and has been extensively studied ever since the born of quantum mechanics .
nevertheless no attention was paid to the quantum mechanical generalization of the remaining three basic thermodynamic processes until most recently . in a recent paper @xcite , along with our collaborators , we systematically study the quantum mechanical generalization of the isothermal and the isochoric process .
base on these studies , the properties of quantum carnot cycle and quantum otto cycle are clarified .
meanwhile in recent years , numerous studies on other quantum thermodynamic cycles are also reported @xcite .
however , as to our best knowledge , the quantum mechanical generalization of isobaric process ( constant pressure ) has not been studied systematically so far .
possibly the lack of the consideration of quantum isobaric process is due to the fact that pressure " ( force ) @xcite is not a well defined observable in a quantum mechanical system .
because of the short of a well defined pressure " ( force ) and thus the quantum isobaric process , the properties of quantum thermodynamic cycles that consist of quantum isobaric process , such as quantum brayton cycle and quantum diesel cycle @xcite can not be clarified .
we notice that some discussions about quantum brayton cycle have been reported @xcite
. nevertheless , their definitions of quantum isobaric process and quantum brayton cycle are ambiguous , and even in contradiction sometimes . as a result , in their studies they can not bridge the quantum and classical thermodynamic cycles . in this paper , along with our previous effort @xcite , we will focus on the study of the quantum isobaric process @xcite and its related quantum thermodynamic cycles .
we begin with the definition of pressure " for an arbitrary quantum system , and then generalize the isobaric process to quantum mechanical systems . based on this and our previous @xcite generalizations of thermodynamic processes
, we are able to study an arbitrary quantum thermodynamic cycle constructed by any of these four quantum thermodynamic processes . as an example
, we will discuss the quantum brayton cycle and the quantum diesel cycle and compare their properties with their classical counterpart .
comparisons between these quantum thermodynamic cycles and their classical counterparts enable us to extend our understanding about the thermodynamics at the interface of classical and quantum physics .
this paper is organized as follows : in sec .
ii , we define microscopically pressure " for an arbitrary quantum mechanical system and study the quantum mechanical generalization of isobaric process ; in sec .
iii , we study quantum brayton cycle and study how the efficiency of brayton cycle bridges quantum and classical thermodynamic cycles ; in sec .
iv we study quantum diesel cycle in comparison with their classical counter part ; sec .
v is the remarks and conclusion .
in order to study the quantum isobaric process , we must first study pressure in an arbitrary quantum mechanical system .
let us recall that in some previous work @xcite , heat and work have been extended to quantum mechanical systems and expressed as functions of the eigenenergies @xmath0 and probability distributions @xmath1 .
the first law of thermodynamics has also been generalized to quantum mechanical systems : @xmath2 where @xmath3 is the @xmath4th eigenenergy of the quantum mechanical system with the hamiltonian @xmath5 under consideration ; @xmath6 is the occupation probability in the @xmath4th eigenstate ; @xmath7 is the @xmath4th eigenstate of the hamiltonian . the density matrix of the system can be written as @xmath8 .
@xmath9 and @xmath10 depict the heat exchange and work done respectively during a thermodynamic process . from classical thermodynamics
we know that the first law can be expressed as @xmath11 . here ,
@xmath12 and @xmath13 refer to temperature and the thermodynamic entropy ; @xmath14 is the generalized force , and @xmath15 is generalized coordinate corresponding to @xmath16 ( @xmath17 is the generalized displacement ) @xcite .
inversely , the generalized force conjugate to the generalized coordinate @xmath18 can be expressed as @xcite @xmath19 for example , when the generalized coordinate is chosen to be the volume @xmath20 , we have its corresponding generalized force pressure @xmath21 . motivated by the definition of the generalized force for a classical system ,
we define analogously the force ( for 1d system , force is the same as pressure ) for a quantum mechanical system @xmath22 where @xmath23 is the generalized coordinate corresponding to the force @xmath24 . in obtaining eq .
( 3 ) , we have used the expression of work for a quantum system @xmath25 . for a single particle in a 1d box ( 1db ) @xcite ( see fig .
1 ) , the generalized coordinate is the width of the potential , and the eigenenergies for such a system depend on the generalized coordinate @xmath26 . here
@xmath27 is the plank s constant ; @xmath4 is the quantum number ; @xmath28 is the mass of the particle .
we obtain the derivative of @xmath29 over @xmath23 straightforwardly @xmath30 .
when the system is in thermal equilibrium with a heat bath at a inverse temperature @xmath31 , the force exerting on either wall of the potential can be calculated by substituting @xmath32 and the gibbs distribution @xmath33 into eq .
( [ 3 ] ) . here
@xmath34 is the boltzmann s constant , and @xmath35 is the partition function .
alternatively , the expression of force ( 3 ) in a quantum mechanical system can be obtained in a statistical - mechanical way @xcite .
@xmath36 where @xmath37 is the free energy of the quantum system .
it should be pointed out that eq .
( [ 3 ] ) is more general than eq .
( 4 ) because eq .
( [ 3 ] ) stands no matter the system is in equilibrium or not .
when @xmath1 in eq .
( [ 3 ] ) satisfies gibbs distribution @xmath33 , or the system is in thermal equilibrium , the expectation value of force @xmath24 of eq .
( 3 ) reproduce the usual force in classical thermodynamics .
another model example is the single particle in a 1d harmonic oscillator potential ( 1dh ) ( see fig .
its hamiltonian is the same as the hamiltonian of a single - mode radiation field in a cavity @xcite .
for such 1dh , we will see later that the definition of force ( [ 3 ] ) for 1dh agrees with the radiation pressure ( 4 ) of a single mode radiation field .
we would like to mention that the definition of force in ( 3 ) is a further step in quantum thermodynamics after the definitions of heat and work ( [ 1 ] ) .
we will see all these definitions of work , heat , entropy and pressure for a quantum mechanical system are self - consistent , and consistent with classical thermodynamics . ) , width=302 ] .
the @xmath4th eigenenergy of the single mode radiation field with the potential width @xmath38 is given by @xmath39 , @xmath40 .
this pressure is equivalent to the pressure in a quantum mechanical system single particle in 1d harmonic oscillator .
hence quantum isobaric process based on single - mode radiation field is equivalent to that based on a 1d quantum harmonic oscillator.,width=302 ] having clarified force for a quantum mechanical system , in the following we study how to extend classical isobaric process to a quantum mechanical system .
classical isobaric process is a quasi - static thermodynamic process , in which the pressure of the system remains a constant @xcite .
the time scale of relaxation of the system with the heat bath is much shorter than the time scale of controlling the volume of the system @xcite . in a classical isobaric process , in order to achieve a constant pressure , we must carefully control the temperature of the system , i.e. , carefully control the temperature of the heat bath , when we change the volume of the classical system @xcite .
for example , for the classical idea gas with the equation of state @xmath41 , the temperature of the system in the isobaric process is required to be proportional to the volume@xmath42 of the gas , so that the pressure can remains a constant .
for a quantum mechanical system , however , the change of the temperature of the heat bath with the generalized coordinate may not be so obvious as the classical ideal gas . because we usually do not know the equation of state of a quantum mechanical system .
let us consider the quantum isobaric process based on 1db ( see fig .
1 ) . for such a quantum mechanical system
, the pressure on the wall can be obtained from eq .
( 3 ) @xmath43}{z(l)}\frac { de_n(l)}{dl}\\ & = -\sum_n \frac{\exp[-\beta(l)\frac{\pi^2\hbar^2n^2}{2ml^2}]}{\frac{1}{2}\sqrt{\frac{2ml^2}{\pi\hbar^2\beta(l)}}}\times \frac{(-2)}{l}\times \frac{\pi^2\hbar^2n^2}{2ml^2 } \\ & = \frac{4}{l } \sqrt{\frac{\pi\hbar^2\beta(l)}{2ml^2 } } \left[-\frac{\partial}{\partial \beta(l)}\sum_n \exp[-\beta(l)\frac{\pi^2\hbar^2n^2}{2ml^2 } ] \right ] \\ & = \frac{4}{l } \sqrt{\frac{\pi\hbar^2\beta(l)}{2ml^2 } } \left[-\frac{\partial}{\partial \beta(l ) } \frac{1}{2 } \sqrt{\frac{2ml^2}{\pi\hbar^2\beta(l ) } } \right ] \\ & = \frac{1}{l\beta(l)}. \end{split}\ ] ] eq . ( 5 ) can be regarded as the equation of state @xmath44 for 1db obtained from eq .
( 3 ) , and it means that if we want to keep the pressure @xmath24 as a constant , we must control the temperature of the system to be proportional to the width of the potential well @xmath45 when the system inside the box pushes one of the walls to perform work .
this property of 1db is the same as the classical ideal gas .
we will see more analogues between them later .
it should be mentioned that the temperature function @xmath46 of the volume " in a quantum isobaric process is system - dependent .
i.e. , for different quantum systems , the function of the temperature over the volume " in the quantum isobaric process differs from one to another . in the following we consider the quantum isobaric process based on a single mode radiation field in a cavity , which was first proposed as the working substance for a quantum heat engine in ref .
we assume that the cavity of length @xmath23 and cross - section @xmath47 can support only a single mode of the field @xmath48 , where @xmath49 is an integer , and @xmath50 is the speed of light .
the hamiltonian reads @xmath51 where @xmath52 is the fock state of the radiation field . from eq .
( [ 3 ] ) we obtain the radiation force @xmath24 as a function of the temperature @xmath53 and the length of the cavity @xmath23
@xcite @xmath54\frac{1}{l},\label { 22}\\ & = \left[\frac{\hbar\frac{s\pi c}{l}}{e^{\beta(l)\hbar\frac{s\pi c}{l}}-1}+\frac{1}{2}\hbar\frac{s\pi c}{l } \right]\frac{1}{l}. \end{split}\ ] ] from eq .
( 7 ) it can be inferred that in order to achieve a constant force , we must carefully control the temperature of the heat bath in the following subtle way @xmath55 it can be seen that in a quantum isobaric process , the temperature function ( 8) for the single - mode radiation field is much more complicated than that @xmath56 of 1db . for the convenience of later analysis
, we would also like to calculate the entropy and the internal energy of the two systems in a quantum isobaric process .
first we consider the 1db .
the entropy expression can be obtained from the above eq .
( [ 1 ] ) @xcite . @xmath57 . \label { 4.5 } \end{split}\ ] ] through the comparison with the entropy of classical ideal gas @xcite , we find that the entropy of classical idea gas reproduces the entropy ( 9 ) of 1db if we choose the molecule number of the classical ideal gas to be @xmath58 .
we plot the entropy - temperature curve ( [ 4.5 ] ) of a quantum isobaric process in fig .
the internal energy of the 1db during the isobaric process can also be obtained analytically ( the temperature of the heat bath is time - dependent ) .
@xmath59 this expression of internal energy verifies the equipartition theorem @xcite , and justifies the result in ref .
@xcite again : the internal energy of the 1db depends only on the temperature . from eqs .
( 9 ) and ( 10 ) we see that both the entropy and the internal energy of 1db have the same form as that of the classical ideal gas @xcite if we choose the molecule number of the classical ideal gas to be @xmath58 . moreover , from eq .
( 5 ) we know that 1db has the same equation of state @xmath44 as that of the classical ideal gas @xmath60 except the difference of the particle number .
thus we conclude that 1db is the quantum mechanical counterpart of the classical ideal gas .
we would like to mention that in the study of the single - molecule engine by szilard and zurek @xcite , they simply employ the equation of state for the classical ideal gas @xmath61 and choose the particle number to be @xmath58 .
nevertheless , this treatment may be questionable because the equation of state @xmath61 stands in the macroscopic and classical case .
when we come to the extreme limit of small system with only a few degrees of freedom , we must use the quantum mechanical treatment as we present here .
fortunately , all the treatments of the single molecule engine @xcite by szilard and zurek is in accordance with our quantum mechanical treatments .
thus we say that our discussions lay the foundation for szilard - zurek single molecule engine @xcite . as to the single - mode radiation field
, the entropy and the internal energy can be calculated as that in ref .
@xcite @xmath62 @xmath63 where @xmath64^{-1}$ ] is the mean photon number .
it is easy to see that the entropy ( 11 ) and the internal energy ( 12 ) of a single mode radiation field have different forms from that of 1db ( 9),(10 ) , and thus from the classical ideal gas .
the internal energy ( 12 ) of single mode radiation field depends on both the temperature @xmath53 and the width @xmath23 of the potential well , while the internal energy of 1db ( 10 ) depends on @xmath53 only .
in addition , the equation of state ( 7 ) of the single - mode radiation field differs from from that ( 5 ) of 1db , and thus from the classical ideal gas .
based on these observations , we say that the single mode radiation field has totally different thermodynamic properties from that of classical ideal gas . it can be inferred that quantum heat engine based on single mode radiation field can give us new results beyond that of classical ideal gas . as we mentioned before , the hamiltonian of the single - mode radiation field is the same as that of 1dh .
thus all the results about single mode radiation field are the same as that for 1dh .
alternatively , we can say that 1dh is the counterpart of single - mode photon gas , in analogy to the fact that 1db is the counterpart of classical ideal gas .
but it should be mentioned that single - mode photon gas are still quantum mechanical system , while classical ideal gas are classical system . in the following
we will alternatively use 1dh and single - mode photon gas .
in addition to our previous studies @xcite , up to now we have extended all four basic thermodynamic processes to quantum mechanical domain . for a comparison of quantum thermodynamics processes and
their classical counterparts see table i. [ cols="^,^,^,^,^ " , ]
diagram of a quantum brayton cycle ( see fig .
4 ) based on 1db . in two adiabatic processes , @xmath65 and @xmath66 the entropy remains a constant .
, width=302 ] in the preceding section , we extend the classical isobaric process to quantum mechanical systems based on the definition of pressure ( 3 ) . in this section and the next section ,
we study two kinds of thermodynamic cycles consisting of the quantum isobaric process , and compare them with their classical counterparts .
we first consider the quantum brayton cycle based on 1db .
a quantum brayton cycle is a quantum mechanical analogue of the classical brayton cycle @xcite , which consists of two quantum isobaric processes and two quantum aidabatic processes .
similar to our discussion in ref .
@xcite , the counterpart of classical adiabatic plus quasistatic process is quantum adiabatic process @xcite . in constructing quantum brayton cycle , we also requires that i ) all the energy level spacing of the work substance change by the same ratio in the quantum adiabatic process , and ii ) this ratio be equal to the ratio of the temperatures of the two heat baths just before and after the quantum adiabatic process .
it should be mentioned that in the isothermal process of a carnot cycle , the temperature of the heat bath is fixed . however
, this is not the case in the isobaric process of a bayton cycle .
hence , we can not simply say that the ratio of the change of the energy level spacings should be equal to the ratio of the temperatures of the two heat baths @xcite . for the current example
, it can be checked that the change of the energy level spacings in quantum adiabatic process ( see fig .
3 ) @xmath67 and @xmath68 should be equal to the ratio of temperatures of the heat baths at @xmath69 and @xmath70 , or at @xmath71 and @xmath47 ( because @xmath72 ) .
fortunately , the above condition i ) can be satisfied by some quantum mechanical system , such as 1db , 1dh , two - level system , et al , and our study will focus on these systems whose energy level spacings change in the same ratio in the quantum adiabatic process .
otherwise , the irreversibility will arise @xcite .
we give a temperature - entropy @xmath73 diagram of the quantum brayton cycle ( see fig .
3 ) . through a standard procedure ,
we obtain ( see appendix a ) the efficiency of the quantum brayton cycle based on the 1db @xmath74 where @xmath75 and @xmath76 are the pressures of the system during two quantum isobaric processes ( see fig .
4 ) . diagram of a quantum brayton cycle based on a single particle in 1db .
@xmath77 represents an isobaric expansion process with a constant force @xmath78 ; @xmath65 represents an adiabatic expansion process with constant entropy @xmath79 ; @xmath80 represents an isobaric compression process with constant pressure @xmath81 ; @xmath82 is another adiabatic compression process with constant entropy @xmath83.,width=302 ] we would like to compare this efficiency of the quantum brayton cycle ( 13 ) with its classical counterpart . from eq .
( 9 ) we know that in the quantum adiabatic process ( @xmath84 ) , we have @xmath85 . as a result the adiabatic exponent @xmath86 is obtained through the comparison with @xmath87 for the adiabatic process .
let us recall that the efficiency of a classical brayton cycle is @xmath88 @xcite , where @xmath89 is the classical adiabatic exponent .
thus our result ( 13 ) bridges the quantum brayton cycle and classical brayton cycle .
hence this justifies that the definition of pressure ( 3 ) for a quantum mechanical system is self - consistent .
similarly we obtain we obtain the efficiency of a quantum brayton cycle based on 1dh ( see appendix a ) @xmath90 from eq . ( 11 ) we know that in a quantum adiabatic process @xmath91 .
thus @xmath92 for 1dh is obtained .
it can be seen that the efficiency of a quantum brayton cycle obtained here ( 14 ) is the same as that of a classical brayton cycle . through the discussion of quantum brayton cycles based on two model systems 1dh and 1db
, we see that the definition of pressure ( 3 ) for quantum systems has clear physical implication , and our study bridges the thermodynamic cycles based on quantum and classical systems .
except for the thermodynamic cycles consisting of two pairs of basic thermodynamic processes , such as carnot cycle , otto cycle @xcite , and brayton cycle , there are some interesting thermodynamic cycles consisting of more than two kinds of thermodynamic processes , such as diesel cycle .
the diesel cycle consists of two adiabatic processes , one isobaric processes and one isochcoric process @xcite ( see fig . 5 ) . in order to construct such a quantum diesel cycle , we require 1 ) the quantum adiabatic conditions are satisfied , and 2 ) all energy level spacings change in the same ratio in the thermally isolated process @xcite . because this is the quantum counterpart of classical adiabatic process ( thermally isolated plus quasistatic process ) @xcite . besides
, the ratio of the change of the energy level spacings in the quantum adiabatic process @xmath93 should be equal to the ratio @xmath94 of the temperatures of the heat bath at @xmath47 and at @xmath71 ( see fig .
5 ) ; the energy level spacing at @xmath70 should be equal to that at point @xmath71 ( see fig . 5 ) .
diagram of a quantum diesel cycle based on 1db and single mode radiation field .
@xmath77 represents an isobaric expansion process with a constant pressure @xmath78 ; @xmath65 represents an adiabatic expansion process with constant entropy ; @xmath80 represents an isochoric compression process with constant volume @xmath95 ; @xmath82 is another adiabatic compression process.,width=302 ] in the following we will consider implementing the quantum diesel cycle in 1db and in 1dh .
first we consider the 1db .
the efficiency of a quantum diesel cycle based on 1db can be obtained through a straightforward calculation ( see appendix b ) @xmath96 here @xmath97 ( see fig . 5 ) and @xmath98 ( see fig .
5 ) are the compression and expansion ratios of the volumes .
this efficiency for a quantum diesel cycle based on 1db agrees with that of a classical diesel cycle , too . through a similar analysis
we obtain the efficiency for a quantum diesel cycle based on 1dh with the only change of @xmath89 from 3 in eq .
( 15 ) to 2 @xmath99 c|c|c|c|c|c & & & & + & @xmath100 & @xmath101 & @xmath102 & @xmath103 + & monoatomic classical idea gas ( @xmath104 ) & @xmath105 & @xmath106 & @xmath107 & @xmath108 + * classical * & diatomic classical idea gas ( @xmath109 ) & @xmath100 & @xmath110 & @xmath111 & @xmath112 + & polyatomic classical idea gas ( @xmath113 ) & @xmath105 & @xmath114 & @xmath115 & @xmath116 + & single particle in 1d box ( @xmath86 ) & @xmath105 & @xmath117 & @xmath118&@xmath119 + & single particle in 2d box ( @xmath120 ) & @xmath100 & @xmath121 & @xmath122 & @xmath123 + & single particle in 3d box ( @xmath104 ) & @xmath100 & @xmath106 & @xmath107 & @xmath108 + & 1d single mode photon field ( @xmath92 ) & @xmath100 & @xmath124 & @xmath125&@xmath126 + * quantum * & 3d black body radiation field ( @xmath113 ) & @xmath100 & @xmath114 & @xmath115 & @xmath116 + & 1d harmonic oscillator ( @xmath127 ) & @xmath100 & @xmath124 & @xmath125&@xmath126 + & 2d harmonic oscillator ( @xmath128 ) & @xmath100 & @xmath129 & @xmath130 & @xmath131 + & 3d harmonic oscillator ( @xmath113 ) & @xmath100 & @xmath114 & @xmath115 & @xmath116 + & spin-1/2 ( 2-level system ) ( @xmath92 ) & @xmath100 & @xmath124 & @xmath125 & @xmath126 + before concluding this section , we would like to mention that we can also discuss the quantum brayton cycle and the quantum diesel cycle based on an arbitrary quantum system , such as the 3d black body radiation field or a spin-1/2 system in an external magnetic field with the hamiltonian @xmath132 . here
@xmath133 is the pauli matrix and @xmath69 is the external magnetic field .
it can be seen from the table ii that the efficiencies for both quantum carnot cycle and classical carnot cycle are always equal to the carnot efficiency @xmath134 irrespective of the properties of the working substance .
different from the carnot cycle , the efficiencies of otto cycle , brayton cycle and diesel cycle are working substance - dependent ( see table ii ) . more specifically
, they depend on the adiabatic exponent @xmath89 of the working substance .
as long as we get the adiabatic exponent @xmath89 of the quantum system , we obtain the explicit expression of the efficiencies of the quantum thermodynamic cycles by substituting @xmath89 into the expression of the efficiencies of the classical thermodynamics with their adiabatic exponent .
for example , for a spin-1/2 system in an external magnetic field , we choose the inverse of the magnetic field strength as the generalized coordinate @xmath135 .
then it can be found that the adiabatic exponent for such a system is @xmath92 . as a result ,
the efficiencies of a quantum brayton cycle and a quantum diesel cycle based on a spin-1/2 system in an external magnetic field are the same as that based on 1dh ( 14 ) , ( 16 ) .
similarly , the efficiencies for a brayton cycle and a diesel cycle based on 3d radiation field can be obtained straightforwardly by substituting @xmath89 with the adiabatic exponent @xmath136 @xcite for 3d radiation field . in table
ii we list the working efficiencies for several typical thermodynamic cycles based on different kinds of classical and quantum working substances ( based on their adiabatic exponent ) .
in summary , in this paper , we study the quantum mechanical analogy of the classical isobaric process based on a microscopic definition of force . in studying the thermodynamic properties of a small quantum system
, we use a new pair of conjugate variables @xmath1 and @xmath0 instead of the usual thermodynamic variables @xmath137 and @xmath20 or @xmath12 and @xmath13 @xcite .
the general expression of force for an arbitrary quantum system @xmath138 is found .
it can be checked that this expression is in accordance with the force @xmath139 in statistical mechanics if the quantum system is in thermal equilibrium with a heat bath .
in addition we clarify the relation between adiabatic process ( thermally isolated plus quasistatic process ) in classical systems and quantum adiabatic process in quantum systems , and we find that all energy level spacings change in the same ratio in the quantum adiabatic process is essential in simulating the classical adiabatic process .
otherwise irreversibility will arise @xcite .
based on quantum isobaric processes , we make quantum mechanical extension of some typical thermodynamic cycles .
the properties of these quantum thermodynamic processes and cycles are clarified , and we bridges the quantum thermodynamic cycles and their classical counterpart . the quantum heat engines and their classical counterparts
have the same efficiencies as long as their working substance has the same adiabatic exponent .
the definitions of force and work for a single - particle quantum system may have important application in the experimental exploration of nonequilibrium thermodynamics in small quantum systems , such as quantum jarzynski equality and quantum crooks fluctuation theorem @xcite .
though the working substance of quantum heat engines deviates from thermodynamic limit , we reproduce the efficiency of classical heat engines .
hence our study lay the concrete foundation for szilard - zurek single molecule engine .
moreover , we found the close relation between classical ideal gas and 1db , and between single - mode photon gas and 1dh . before concluding this paper , we would like to mention that in our current study we focus on the quantum single - particle system , and its related quantum mechanical generalization of heat , work , pressure , and we regain the results of classical thermodynamic processes and cycles
. we also notice some studies about quantum heat engines with quantum many body system as the working substance @xcite . for quantum many body system ,
e.g. , ideal bosonic gas or ideal fermionic gas , the mechanical variable , such as heat work , pressure , are well defined and their equation of state as well as their expression of internal energy @xcite deviate from that of the classical ideal gas . as a result , the properties of quantum thermodynamic cycles based on the quantum many - body system deviate from that of classical ideal gas due to quantum degeneracy . finally ,
similar to the discussion about finite - power carnot engine @xcite , we can discuss about finite - time quantum brayton cycle and quantum diesel cycle .
finite - power analysis of brayton cycle and diesel cycle will be given later .
this work is supported by u.s .
department of energy through the lanl / ldrd program .
the author thanks prof .
p. w. milonni for helpful discussions and prof .
j. q. you for the hospitality extended to him during his visit to fudan university .
according to the definition of heat exchange ( [ 1 ] ) in the quantum mechanical system , we obtain the heat absorbed by the system from a time - dependent heat bath during the quantum isobaric expansion process @xmath77 ( see fig . 3 and fig .
4 ) @xmath140dl\\ = & \sum_n\int_{l_a}^{l_b}\left [ \left[e_n(l)p_n(l)\right]^{\prime}-\frac{de_n(l)}{dl}p_n(l)\right ] dl \label { 14}\\ = & \sum_n[e_n(l_b)p_n(l_b)-e_n(l_a)p_n(l_a)]\\ & + \int_{l_a}^{l_b}f(l)dl\\ = & \frac{1}{2}[f_1l_b - f_1l_a]+f_1(l_b - l_a)\\ = & \frac{3}{2}f_1(l_b - l_a ) .
\end{split}\ ] ] in obtaining the above result we have used eq .
( 5 ) and eq . ( 10 ) .
similarly , we obtain the heat released to the time - dependent entropy sink @xmath141 hence , the efficiency of the quantum brayton cycle based on a 1db can be expressed as @xmath142 due to the equation of motion ( 5 ) and the expression of the internal energy ( 10 ) , we have @xmath143 , @xmath144 .
in addition to the relation of the internal energies in the quantum adiabatic process @xmath65 @xmath145 we have @xmath146 for the quantum adiabatic process @xmath65 . through a similar analysis we obtain @xmath147 for another quantum adiabatic process @xmath66 . based on all the above results ( a3 ) , ( a5 ) , and ( a6 )
, we obtain the efficiency of the quantum brayton cycle based on the 1db @xmath148 in the following we consider a quantum brayton cycle based on 1dh .
similar to the above analysis , we calculate the heat absorbed by the system during the quantum isobaric expansion process @xmath149 ( see fig .
3 ) @xmath140dl \label { 24}\\ = & [ u(l_b)-u(l_a)]+\int_{l_a}^{l_b}f d(l)\\ = & \left(\frac{\hbar\omega_b}{e^{\beta(l_b ) \hbar\omega_b}-1}+\frac{\hbar\omega_b}{2}\right)\\ & -\left(\frac{\hbar\omega_a}{e^{\beta(l_a ) \hbar\omega_a}-1}+\frac{\hbar\omega_a}{2}\right)+f_h ( l_b - l_a)\\ = & f_1 ( l_b - l_a ) , \end{split}\ ] ] where we have used the relations ( 7 ) and ( 12 ) in the quantum isobaric process ( @xmath150 ) .
similarly , we obtain the heat released to the entropy sink in another quantum isobaric process @xmath151 @xmath152 the efficiency of the quantum brayton cycle based on a 1dh can be expressed as @xmath153 from eqs .
( 7 ) and ( 12 ) we have @xmath154 and @xmath155 .
in addition to the relation of internal energy in the quantum adiabatic process @xmath156 we have @xmath157 hence , from eqs .
( a10 ) and ( a12 )
we obtain the efficiency of a quantum brayton cycle based on 1dh @xmath90
for a quantum diesel cycle ( see fig .
5 ) , the input energy in the quantum isobaric process @xmath158 and the output energy in the quantum isochoric process @xmath159 can be calculated as @xmath160 where @xmath161 and @xmath162 are the heat capacity at constant pressure and constant volume respectively ; @xmath163 , @xmath164 , @xmath165 , and @xmath166 are the temperatures of the system at different points of the diesel cycle ( see fig . 5 )
thus the efficiency of the quantum diesel cycle can be expressed in terms of heat capacities and temperatures @xmath167 it is convenient to express this efficiency ( b2 ) in terms of compression ration @xmath97 ( see fig . 5 ) and the expansion ratio @xmath98 ( see fig .
5 ) of the volumes . now using the equation of state @xmath44 ( 5 ) and @xmath168 for 1db , the efficiency ( b2 ) can be rewritten as @xmath169 by utilizing the facts @xmath170 and @xmath171 ( see fig .
5 ) , we further simplify the eq . ( b3 ) to @xmath172 finally by making use of the adiabatic condition @xmath173 for 1db in the quantum adiabatic process , we obtain @xmath174 substituting eq .
( b5 ) into eq .
( b4 ) , the efficiency of a quantum diesel cycle based on 1db can be written as @xmath96 m. o. scully , phys .
_ _ _ _ * * 88 * * , 050602 ( 2002 ) ; m. o. scully , m. suhail zubairy , g. s. agarwal , and h. walther , science * 299 * , 862 ( 2003 ) ; m. o. scully , in _ quantum limits to the second law : first international conference _ ,
sheehan ed .
( american institute of physics , melville , ny , 2002 ) , pp .
83 - 91 ; m. s. zubairy , in _
quantum limits to the second law : first international conference _ , d.p . sheehan ed .
( american institute of physics , melville , ny , 2002 ) , pp .
92 - 97 .
m. o. scully phys .
lett . * 87 * , 220601 ( 2001 ) ; m. o.
scully , y. rostovtsev , z. e. sariyanni , and m. s. zubairy , physica e * 29 * , 29 ( 2005 ) ; y. rostovtsev , z. e. sariyanni , m. s. zubairy , and m. o. scully , _ ibid . _ * 29 * , 40 ( 2005 ) ; z. e. sariyanni , y. rostovtsev , m. s. zubairy , and m. o. scully , _ ibid . _ * 29 * , 47 ( 2005 ) ; h. t. quan , y. d. wang , yu - xi liu , c. p. sun , and f. nori , phys .
* 97 , * 180402 ( 2006 ) .
e. geva and r. kosloff , j. chem . phys . *
96 * , 3054 ( 1992 ) ; e. geva and r. kosloff , j. chem
. phys . * 97 * , 4398 ( 1992 ) ; y. rezek and r. kosloff , new j. phys .
* 8 * , 83 ( 2006 ) ; t. feldmann , e. geva , and r. kosloff , am . j. phys .
* 64 * , 485 ( 1996 ) ; c. m. bender , d. c. brody , b. k. meister , proc .
a * 458 * , 1519 ( 2002 ) ; c. m. bender , d. c. brody , and b. k. meister , j. phys . a * 33 * , 4427 ( 2000 ) ; k. bhattacharyya and s. mukhopadhyay , j. phys .
a * 34 * , 1529 ( 2001 ) ; m. h. lee , am . j. phys . * 69 * , 874 ( 2001 ) ; y. v. rostovtsev , a. b. matsko , n. nayak , m. s.zubairy , and m. o. scully , phys . rev .
a * 67 * , 053811 ( 2003 ) ; t. opatrny and m. o. scully , fortschr . physik .
* 50 * , 657 ( 2002 ) ; s. li , h. wang , y. d. sun , x. x. yi , j. phys .
a * 40 * , 8655 , ( 2007 ) ; a. o. niskanen , y. nakamura , j. p. pekola , phys .
b , * 76 * , 174532 ( 2007 ) ; k. maruyam , f. nori , and v. verdral , rev .
, in press ( 2008 ) ; m. grajcar , s. ashhab , j. r. johansson , and f. nori , phys .
b , * 78 * , 035406 ( 2008 ) ; t. jahnke , j. birjukov , g. mahler , eur . phys
. j. special topics , * 151 * , 167 ( 2007 ) ; j. birjukov , t. jahnke , g. mahler , eur . phys . j. b , * 64 * , 105 ( 2008 ) .
a. e. allahverdyan , r. s. gracia , and t. nieuwenhuizen , phys .
e * 71 * , 046106 ( 2005 ) ; j. arnaud , l. chusseau , and f. philippe , eur . j. phys .
* 23 * , 489 ( 2002 ) ; j. arnaud , l. chusseau , and f. philippe , phys . rev .
e. * 77 * , 061102 ( 2008 ) ; f. tonner , g. mahler , fortschr . physik . * 54 * , 939 ( 2006 ) ; f. tonner , g. mahler , phys . rev .
e , * 72 * , 066118 ( 2005 ) ; t. feldmann and r. kosloff , phys . rev .
e * 68 * , 106101 ( 2003 ) ; h. t. quan , p. zhang , and c. p. sun , phys .
e. * 72 * , 056110 ( 2005 ) ; t. zhang , w .- t .
liu , p .- x .
chen , c .- z .
li , phys .
a. * 75 * , 062102 ( 2007 ) ; d. janzing , j. stat .
phys . , * 122 * , 531 ( 2006 ) ; a. e. allahverdyan , r. s. johal , and g. mahler , phys .
e * 77 * , 041118 ( 2008 ) ; j. arnaud , l. chusseau , and f. philippe , arxiv:0812.2757 ; e. a. ivanchenko , arxiv:0901.3637 ; d. segal , phys
* 101 * , 260601 ( 2008 ) .
bihong lin , and jincan chen , j. phys . a : math .
gen . , * 38 * , 69 ( 2005 ) ; f. wu , l. chen , f. sun , c. wu , and q. li , phys .
e , * 73 * , 016103 ( 2006 ) ; h. wang , s. liu , and j. he , physica b * 403 * , 3867 ( 2008 ) .
the generalzied force in thermodynamics was originally studied by lars onsager in phys .
, * 37 * , 405 ( 1931 ) .
the usual generalized force includes the pressure @xmath137 , the chemical potential @xmath175 , and their conjugate variables ( generalized coordinate ) are the volume @xmath20 and the particle number @xmath176 respectively .
here we consider the zero point energy of the specific mode @xmath177 of the radiation field inside the cavity ( 6 ) . to be more accurate
we should also consider the zero point energies of the remaining permissible discrete modes inside the cavity and the continuous modes outside the cavity as well .
the net effect of all these zero point energies is the casimir force @xcite , and it is a tiny correction to the usual radiation pressure . when taking into account of all those modes , the value of the correction should be @xmath178 @xcite instead of @xmath179 in eq .
( 7 ) . here
the negative and positive signs mean attractive and repulsive force respectively .
nevertheless the casimir effect is not our focus and it does not affect our main result .
so we do not include this effect in our discussion . in the classical thermodynamic cycles , the adiabatic process implies 1 ) thermally isolated and 2 ) quaistatic , or very slow @xcite .
if the thermally isolated process proceeds fast , the quasistatic conditions are not satisfied . as a result
, there will be internal entropy increase , and irreversibility arises . in order to construct a reversible cycle , such as carnot cycle and brayton cycle , quasistatic condition must be satisfied in the thermally isolated process . similarly .
for a quantum thermodynamic cycle , the quantum mechanical counterpart of classical adiabatic process implies 1 ) thermally isolated and 2 ) no interstate excitations @xcite and 3 ) all energy level spacings change in the same ratio .
alternatively , the quantum mechanical counterpart of adiabatic process is quantum adiabatic process given that all energy level spacings change in the same ratio .
this is because if one starts from an gibbs density operator , when the quantum adiabatic conditions are satisfied and all energy level spacings change in the same ratio as we change a parameter of the hamiltonian , it will remain gibbsian .
so there is no source of irreversibility . on the contrary , if either 1 ) the quantum adiabatic conditions are not satisfied , or 2 ) the energy level spacings do not change in the same ratio in the thermally isolated process , the density operator will not remain gibbsian , and irreversibility arises . hence , in order to construct a reversible quantum thermodynamic cycles ( quantum mechanical counterpart of reversible classical thermodynamic cycles ) , we must focus on those systems whose energy level spacings change in the same ratio and use quantum adiabatic process to replace the classical adiabatic plus quasistatic process .
there are many other types of conjugate pair of variables , for example , the electromotive force and the amount of charge , the magnetization and the magnetic field , the surface tension and the surface area , the elastic force and the length stretched , and the gravitational potential and the mass .
g. g. potter , g. muller , and m. karbach , phys .
e , * 75 * , 061120 ( 2007 ) ; a. sisman , and h. saygin , appl .
energy , * 68 * , 367 ( 2001 ) ; a. sisman , and h. saygin , phys .
scr . * 64 * , 108 ( 2001 ) ; f. wu , l. chen , f. sun , c. wu , f. guo , and q. li , energy convers .
47 * , 3008 ( 2006 ) ; open sys .
information dyn . ,
* 13 * , 55 ( 2006 ) ; f. wu , l. chen , f. sun , c. wu , and y. zhu , energy convers .
* 39 * , 773 ( 1998 ) ; a. sisman , and h. saygin , phys .
* 63 * , 263 ( 2001 ) ; h. saygin , and a. sisman , j. appl .
phys . , * 90 * , 3086 ( 2001 ) ; h. saygin , and a. sisman , appl .
energy , * 69 * , 77 ( 2001 ) ; f. wu , l. chen , f. sun , c. wu , and f. guo , phys .
* 73 * , 452 ( 2006 ) ; a. sisman , and h. saygin , j. phys .
* 32 * , 664 ( 1999 ) ; bihong lin , jizhou he , and jincan chen , j. non - equilib .
thermodyn . , * 28 * , 221 ( 2003 ) ; b. lin , and j. chen , phys .
, * 77 * , 055005 ( 2008 ) . | we study the quantum mechanical generalization of force or pressure , and then we extend the classical thermodynamic isobaric process to quantum mechanical systems . based on these efforts , we are able to study the quantum version of thermodynamic cycles that consist of quantum isobaric process , such as quantum brayton cycle and quantum diesel cycle .
we also consider the implementation of quantum brayton cycle and quantum diesel cycle with some model systems , such as single particle in 1d box and single - mode radiation field in a cavity .
these studies lay the microscopic ( quantum mechanical ) foundation for szilard - zurek single molecule engine . |
multi - dimensional generalization of classical soliton equations has been one of the most exciting topic in the field of integrable systems . among other things ,
calogero @xcite proposed an interesting example that is a ( 2 + 1)-dimensional extension of the korteweg - de vries equation , @xmath1 yu et al .
@xcite obtained multi - soliton solutions of the ( 2 + 1)-dimensional kdv equation by using the hirota s bilinear method .
let us consider the following hirota - type equations , @xmath2 where we have used the @xmath3-operators of hirota defined as @xmath4 we remark that we have introduced auxiliary variables @xmath5 that is a hidden parameter in .
if we set @xmath6 and use to eliminate @xmath7 , then one can show that @xmath8 solves .
bogoyavlensky @xcite showed that there is a hierarchy of higher - order integrable equations associated with . in the paper @xcite , ikeda and takasaki generalized the bogoyavlensky s hierarchy from the viewpoint of the sato s theory of the kp hierarchy @xcite , and discussed the relationship to toroidal lie algebras .
we note that the relation between integrable hierarchy and toroidal algebras has been discussed also by billig @xcite , iohara , saito and wakimoto @xcite by using vertex operator representations . in the present paper , we propose the following differential - difference system that has the same symmetry : @xmath9 where @xmath10 denotes the backward difference operator @xmath11 ( @xmath12 ) . we also show that this system has soliton - type solutions .
here we briefly review the lie algebraic derivation of the bilinear identity of the bogoyavlensky s hierarchy @xcite , which is a generating function of hirota - type differential equations .
we remark that the lie algebra considered in @xcite is bigger than that of present article . here
we do nt include the derivations to @xmath0 since those are not needed for our purpose .
owing to this difference , the proof given below may be simpler than that of @xcite .
the 2-toroidal lie algebra @xmath0 @xcite is the universal central extension of the double loop algebra @xmath13 $ ] , while the affine lie algebra @xmath14 is the central extension of @xmath15 $ ] .
let @xmath16 be the ring of laurent polynomials of two variables @xmath17 and @xmath18 . as a vector space
, @xmath0 is isomorphic to @xmath13 \oplus \omega_a/{\mathrm{d}}a$ ] , where @xmath19 denotes the module of khler differentials of @xmath16 defined with the canonical derivation @xmath20 .
we define the lie algebra structure of @xmath0 by @xmath21 = [ x , y]\otimes ab + ( x|y ) \overline{({\mathrm{d}}a)b } \quad ( x , y\in\mathfrak{sl}_2,\ ; a , b\in a ) , \label{crofsl2tor}\\ & & [ { \mathfrak{sl}^{\mathrm{tor}}}_2 , \omega_a/{\mathrm{d}}a ] = 0 , \end{aligned}\ ] ] where @xmath22 denotes the killing form and @xmath23 the canonical projection . in terms of generating series @xmath24 ( @xmath25 , @xmath26 ) , @xmath27 and @xmath28 , defined by @xmath29 the relation
can be expressed as @xmath30_{m+n}(w ) } \nonumber\\ & & \;+\ ; \frac{1}{(z - w)^2}(x|y)k_{m+n}^s(w ) \;+\ ; \frac{m}{z - w}(x|y)k_{m+n}^{t}(w ) \label{torope}\\ & & \;+\ ; \mbox { regular as } \;z\rightarrow w. \nonumber\end{aligned}\ ] ] there exists a class of representations of @xmath0 , which comes directly from that of @xmath14 .
we consider the space of polynomials , @xmath31\otimes{\mathbb{c}}[\exp(\pm y_0)],\ ] ] and define the generating series @xmath32 and @xmath33 by @xmath34.\ ] ] ( cf .
@xcite ) [ prop : tor_ext ] let @xmath35 be a representation of @xmath14 such that @xmath36 for @xmath37
. then we can define the representation @xmath38 of @xmath0 on @xmath39 such that @xmath40 where @xmath25 , @xmath26 and @xmath41 . using the operator product expansion for @xmath14 , @xmath42(w ) + \frac{1}{(z - w)^2}(x|y)k\\ & & \;+\ ;
\mbox { regular as } \;z\rightarrow w , \end{aligned}\ ] ] and the property @xmath43 , it is straightforward to show that @xmath24 above satisfies .
the remaining relations can be checked by direct calculations . to see the relationship to soliton theory
, we shall consider the representation of @xmath14 on the fermionic fock space @xcite .
let @xmath44 , @xmath45 ( @xmath46 ) be free fermions with the canonical anti - commutation relation . in terms of the generating series defined as @xmath47
the canonical anti - commutation relation is written as @xmath48_+ = \delta(\lambda/\mu ) , \qquad \left[\psi(\lambda ) , \psi(\mu)\right]_+ = \left[\psi^*(\lambda ) , \psi^*(\mu)\right]_+=0,\ ] ] where @xmath49 is the formal delta - function .
consider the fermionic fock space @xmath50 with the vacuum vector @xmath51 satisfying @xmath52 and the dual fock space @xmath53 with the dual vacuum vector @xmath54 satisfying @xmath55 as mentioned in @xcite , a level-1 representation of @xmath14 is given by the elements , @xmath56 where @xmath57 denotes the fermionic normal ordering , @xmath58 . applying proposition [ prop : tor_ext ]
, we can construct a representation of @xmath0 on the space @xmath59 with the vacuum vector @xmath60 .
we now introduce the following operator acting on @xmath61 : @xmath62 using the anti - commutation relation and the relation @xmath63 , we can obtain the following identity by direct calculations : @xmath64 = 0,\ ] ] which means the action of @xmath0 on @xmath61 commutes with @xmath65 .
then it is straightforward to show that @xmath66 for @xmath67 , @xmath68 . to rewrite into bosonic language ,
we prepare two lemmas : @xcite [ lem : bfcorr ] for any @xmath69 , we have the following formulas , @xmath70))|\nu\rangle,\\ \langle\mathrm{vac}|\psi_{-1}\exp(h({\textnormal{\mathversion{bold}$x$}}))\psi^*(\lambda)|\nu\rangle & = \lambda\exp(-\xi({\textnormal{\mathversion{bold}$x$}},\lambda ) ) \langle\mathrm{vac}|\exp(h({\textnormal{\mathversion{bold}$x$}}+[\lambda^{-1}]))|\nu\rangle,\\ \end{aligned}\ ] ] where we have used the following notation , @xmath71 { \stackrel{\mathrm{def}}{=}}\left(\ , 1/\lambda , 1/2\lambda^2 , 1/3\lambda^3,\ldots \,\right ) .
\end{aligned}\ ] ] @xcite [ lem : billig ] let @xmath72 , where @xmath73 diff@xmath74 are differential operators that may not depend on @xmath75 .
if @xmath76 for some formal series @xmath77 , then @xmath78 as a polynomial in @xmath79 . define the @xmath80-function as @xmath81 from the relation , together with lemma [ lem : bfcorr ] and lemma [ lem : billig ] , we have the following bilinear identity : @xmath82 , y_0+\eta(\check{{\textnormal{\mathversion{bold}$b$}}},\lambda^2),\check{{\textnormal{\mathversion{bold}$y$}}}-\check{{\textnormal{\mathversion{bold}$b$ } } } ) \nonumber\\ \times & \tau({\textnormal{\mathversion{bold}$x$}}'+[\lambda^{-1 } ] , y_0-\eta(\check{{\textnormal{\mathversion{bold}$b$}}},\lambda^2),\check{{\textnormal{\mathversion{bold}$y$}}}+\check{{\textnormal{\mathversion{bold}$b$ } } } ) = 0 , \label{bi}\end{aligned}\ ] ] where @xmath83 and @xmath84 .
expanding , we can obtain hirota - type differential equations including , ( @xmath85 , @xmath86 , @xmath87 , @xmath88 ) . in this sense ,
the bilinear identity is a generating function of hirota - type differential equations of the bogoyavlensky s hierarchy .
the @xmath89-soliton solution of is obtained as follows @xcite : @xmath90
we now apply the miwa transformation @xcite , @xmath91 + m [ \beta ] + n [ \gamma ] , \\ { \textnormal{\mathversion{bold}$x$ } } & = & ( l+1 ) [ \alpha ] + ( m+1 ) [ \beta ] + ( n+1 ) [ \gamma ] , \end{array}\ ] ] to the bilinear identity . here
we have used the notation @xmath92 = ( l\alpha , l\alpha^2/2,l\alpha^3/3,\ldots)$ ] .
we first consider the case @xmath93 . in this case , the bilinear identity is reduced to that of the ordinary kp hierarchy .
thus we have the hirota - miwa equation ( or the discrete kp equation ) , @xmath94 where @xmath95 denotes @xmath96+m[\beta]+n[\gamma],{\textnormal{\mathversion{bold}$y$}}).\ ] ] we then consider the time - evolutions with respect to @xmath97 , @xmath98 . collecting the coefficient of @xmath99 in the bilinear identity , we have @xmath100,{\textnormal{\mathversion{bold}$y$}})\cdot \tau({\textnormal{\mathversion{bold}$x$}}-[\lambda^{-1}],{\textnormal{\mathversion{bold}$y$}})= 0 . \end{aligned}\ ] ] applying the miwa transformation ,
we obtain @xmath101 we further impose the condition @xmath102 . then the @xmath80-function @xmath95 depends only on @xmath103 , @xmath104 and @xmath105 . in this sense , we rewrite @xmath106 under this condition , the equations and are reduced to @xmath107 \alpha\beta^2 d_{y_2}&\left ( 2\alpha\;\tau(l+1,k;{\textnormal{\mathversion{bold}$y$}})\cdot\tau(l , k;{\textnormal{\mathversion{bold}$y$ } } ) \right.\nonumber\\ \,&\quad -(\alpha+\beta)\ ; \tau(l , k+1;{\textnormal{\mathversion{bold}$y$}})\cdot\tau(l+1,k-1;{\textnormal{\mathversion{bold}$y$}})\nonumber\\ \,&\quad \left . -(\alpha-\beta)\ ; \tau(l , k-1;{\textnormal{\mathversion{bold}$y$}})\cdot\tau(l+1,k+1;{\textnormal{\mathversion{bold}$y$}})\right)\nonumber\\ = d_{y_0}&\left(2(2\beta^2-\alpha^2)\ ; \tau(l+1,k;{\textnormal{\mathversion{bold}$y$}})\cdot\tau(l , k;{\textnormal{\mathversion{bold}$y$ } } ) \right.\nonumber\\ \,&\quad -\alpha(\alpha+\beta)\ ; \tau(l , k+1;{\textnormal{\mathversion{bold}$y$}})\cdot\tau(l+1,k-1;{\textnormal{\mathversion{bold}$y$}})\nonumber\\ \,&\quad \left .
-\alpha(\alpha-\beta)\ ; \tau(l , k-1;{\textnormal{\mathversion{bold}$y$}})\cdot\tau(l+1,k+1;{\textnormal{\mathversion{bold}$y$}})\right ) . \label{difdif2}\end{aligned}\ ] ] furthermore , we can construct @xmath89-soliton solution by applying to : @xmath108 where @xmath109 is the same as the continuum one .
we remark that the @xmath89-soliton @xmath80-function can be written as wronskian determinant . using the determinant expression
, we can show that both and are reduced to the plcker relations . introducing the variables as @xmath110 , \\ & & v_k(t , x ) { \stackrel{\mathrm{def}}{=}}\partial_x \log\frac{\tau(l , k+2)}{\tau(l , k ) } , \end{aligned}\ ] ] we have the differential - difference equations and , which have the @xmath89-soliton solution corresponding to the @xmath80-function .
in this paper , we have introduced the differential - difference system and , which is related to toroidal lie algebra @xmath0 .
since the symmetry of the toroidal lie algebra allows us to introduce extra parameters of wave numbers in the soliton solution ( i.e. , @xmath111 in and ) , it might be possible to construct some interesting solutions . in particular , we can obtain a class of traveling - wave solutions that has the shape of the character `` v '' ( figure [ fig : vsoliton ] ) , which is a special case of the two - soliton solutions .
the existence of the v - soliton type solution is one of the features of this class of equations .
we note that there exist solutions of the same shape for the ( 2 + 1)-dimensional kdv equation , and for a ( 2 + 1)-dimensional generalization of the nonlinear schrdinger ( nls ) equation @xcite that also has the symmetry of the toroidal lie algebra @xmath0 @xcite
. we also remark that oikawa et al .
@xcite discussed the propagation of the v - soliton in a two - layer fluid , which is governed by a equation similar to the ( 2 + 1)-dimensional nls equation .
the authors acknowledge doctor takeshi ikeda and professor kanehisa takasaki for discussions . the first author is partially supported by the grant - in - aid for scientific research ( no .
12740115 ) from the ministry of education , culture , sports , science and technology . e. date , m. jimbo , m. kashiwara and t. miwa , in proceedings of rims symposium _ nonlinear integtable systems classical theory and quantum theory _ ( m. jimbo and t. miwa , eds . ) 39 - 120 , world scientific , 1983 . | we present a novel differential - difference system in ( 2 + 1)-dimensional space - time ( one discrete , two continuum ) , arisen from the bogoyavlensky s ( 2 + 1)-dimensional kdv hierarchy .
our method is based on the bilinear identity of the hierarchy , which is related to the vertex operator representation of the toroidal lie algebra @xmath0 . |
in recent years , uv and optical photometry and spectroscopy of nearby elliptical galaxies has suggested that these galaxies , which have a reputation for being old , red , and dead , may not be quite as dead as previously assumed .
some 15% to 30% of local ellipticals seem to be experiencing small amounts of present day star formation activity ( schawinski et al . 2007a , 2007b ; kaviraj et al .
the star formation is not intense enough to cause serious problems for the galaxies morphological classification , as it only amounts to a few percent of the total stellar mass .
however , this current day disk growth inside spheroidal galaxies may be a faint remnant of a process which was more vigorous in the past and may have played a role in establishing the spectrum of galaxy morphologies we observe today .
star formation of course requires cold gas , so interpreting the uv and optical data in terms of star formation activity has important implications both for the early - type galaxies and for a general understanding of the star formation process .
it is not obvious that star formation should work " the same way inside spheroidal galaxies as it does inside disks , with the same efficiency or the same dependence on the gas surface density .
for example , it has been hypothesized that even if there is a molecular disk inside an elliptical or lenticular galaxy , the disk would probably be stabilized by the galaxy s steep gravitational potential ( e.g. kennicutt 1989 ; okuda et al .
2005 ; kawata et al .
thus it is of interest to probe the relationships between molecular gas and star formation activity in early - type galaxies .
it is not as straightforward to measure star formation rates in early - type galaxies as it is in spirals , however .
optical line emission is common in ellipticals ( shields 1991 ; goudfrooij et al .
1994 ; sarzi et al .
2006 ) , but usually its widespread and/or filamentary distribution and its line ratios indicate that it is more closely related to agn activity or to cooling from hot gas than to star formation .
far - ir and cm - wave radio continuum emission are also commonly used as tracers of star formation activity in gas - rich spirals , but the mid - ir and far - ir emission in ellipticals is usually attributed to agn activity and/or the evolved stellar population rather than to star formation ( temi et al .
2007 , 2008 ) . here
we investigate some evidence for and against star formation activity in a sample of elliptical and lenticular galaxies that have unusually large molecular gas contents ( table 1 ) .
we make use of matched - resolution images of the molecular gas distribution , the cm - wave radio continuum and the 24 intensity .
ccccccc ugc 1503 & e & 67 & 10.2 & 9.4 & 25 & field + ngc 0807 & e & 63 & 10.5 & 9.2 & 19 & field + ngc 2320 & e & 77 & 10.8 & 9.2 & 43 & group dominant + ngc 3032 & s0 & 21 & 9.7 & 9.0 & 5.0 & field + ngc 3656 & i0 & 42 & 10.2 & 9.7 & 64 & merger remnant + ngc 4459 & s0 & 16 & 10.2 & 8.7 & 1.6 & virgo cluster + ngc 4476 & e / s0 & 17 & 9.5 & 8.3 & 1.5 & virgo cluster + ngc 4526 & s0 & 16 & 10.5 & 9.2 & 5.7 & virgo cluster + ngc 5666 & ec / s ? & 34 & 9.8 & 9.4 & 7.8 & field +
early surveys for molecular gas in early - type galaxies were strongly biased towards fir - bright targets , but more recent work is not biased in this way and still finds significant molecular gas contents .
welch & sage ( 2003 ) found a surprisingly high co detection rate of @xmath0 in a volume - limited sample of nearby field lenticular galaxies , and sage et al .
( 2007 ) detected co emission in @xmath1 of a similar sample of field ellipticals
. combes et al .
( 2007 ) also detected co emission in @xmath2 of the early - type galaxies in the sauron survey ( de zeeuw et al .
2002 ) , a representative sample which uniformly fills an optical magnitude apparent axis ratio space .
thus , the co detection rates may be high enough to support the uv - inferred incidence of star formation activity .
the cold gas masses are highly variable in these detections , with @xmath3 in the range @xmath4 to @xmath5 and lower . since we are interested in morphology as a means of distinguishing the origin of the radio and ir emission , we have selected for this project elliptical and lenticular galaxies with maps resolving their molecular gas distribution .
if the 24emission arises in star formation activity , we expect it to trace the molecular gas .
if the 24 emission is related to an agn we expect it to be a point source , and if it is circumstellar dust it should trace the stellar distribution .
thus , we also required the targets to have molecular gas which is extended enough that it should be resolved in the 24 images .
the molecular gas maps are published by young ( 2002 , 2005 ) and young , bureau , & cappellari ( 2008 ) ; 1.4 ghz ( 20 cm ) radio continuum maps are available from the vla first survey and from lucero & young ( 2007 ) .
a handful of elliptical and lenticular galaxies were observed as part of the sings survey and their morphologies are discussed by bendo et al .
( 2007 ) . with a couple of exceptions , such as ngc 1316 ( which contains an extended arc of emission ) and ngc 5866 ( which contains an edge - on disk )
, these early - type galaxies tended to have highly compact , symmetric nuclear pointlike or nearly - point sources at 24 .
at least three of these are detected in co emission but co maps are not yet available .
mips observations of elliptical galaxies have also been published by temi et al.(2007 , 2008 ) and kaneda et al .
( 2007 ) . of the 19 galaxies discussed by temi et al.(2008 ) , 13 have been searched for co emission and none have been detected . in these works
the 24 emission from the ellipticals follows the near - ir surface brightness profiles very closely , so the 24 surface brightness is close to a @xmath6 profile with the same effective radius as in the @xmath7-band .
in addition , temi et al .
( 2007 ) have shown that the 24 emission globally tracks the optical luminosity in elliptical galaxies as there is a tight linear correlation between the 24 flux density and the @xmath8-band flux density .
this 24emission is interpreted to be circumstellar dust from the mass loss of post main sequence stars .
thus , in the majority of the elliptical galaxies that are not co - rich the 24 emission seems to either follow the stellar photospheric emission or a nuclear source .
simple model fits are made in order to provide some parametrization of the 24 morphologies .
we first constructed an empirical psf from archival observations of 3c 273 , 3c 279 , and bl lac .
point sources , exponential disks , rings , and de vaucouleurs @xmath6 profiles ( and combinations of these ) are convolved with the psf and fit to the images , just as was done for the sombrero galaxy by bendo et al . (
none of the co - rich early - type galaxies are pure point sources at 24 ; all are resolved , but are still significantly less extended than the stellar distributions in the nir and optical .
the 24 intensity is a function of both the dust surface density and the illuminating radiation field .
possible dust heating sources include the post - main sequence stellar population , star formation regions , and agn .
thus , the 24 emission by itself may not necessarily indicate the presence of star formation .
however , star formation is expected to be accompanied by spatially resolved cm - wave radio continuum emission ( condon 1992 ) , whereas dust heated by the radiation from evolved stars would not be .
therefore the comparisons with the distribution of the molecular gas ( the raw material for star formation ) and the radio continuum provide constraints on the origin of the 24 emission .
figures [ u1503 ] and [ n2320 ] show some of the variety of morphologies in these co - rich early - type galaxies .
for example , ugc 1503 shows a regularly rotating molecular gas ( and dust ) disk of diameter 30 = 9.7 kpc , with a similarly sized ring of radio continuum emission .
the 24 emission also shows a ring with a central dip , and both radio continuum and 24 have morphologies consistent with star formation in the molecular gas .
most of the other cases in the sample also have closely matching radio , 24 and co morphologies , and they are interpreted in a similar manner .
optical , ir , co and radio continuum morphology of ugc 1503 .
optical contours are spaced by a factor of two .
the top row , center panel is an unsharp - masked @xmath9 image and the left panel is the ratio of the @xmath9 to a smooth multi - gaussian expansion model ( cappellari 2002 ) .
contour levels in the 24 image are 2 , 5 , 10 , 20 , 30 , 50 , 70 , and 75 percent of the peak ( 6.1 ) .
contour levels in the co image are @xmath10 , 10 , 20 , 30 , 50 , 70 , and 90 percent of the peak ( 6.3 ) .
contours in the 1.4 ghz radio continuum image are 0.09 , 0.12 , 0.15 , 0.18 , and 0.24 .
green circles show the spatial resolution . ]
optical , ir , co and radio continuum morphology of ngc 2320 .
optical contours are spaced by a factor of two .
contours in the 24 image are 2 , 5 , 10 , 20 , 30 , 50 , 70 , and 90 percent of the peak ( 5.6 ) .
contours in the co image are @xmath11 , 5 , 10 , 20 , 30 , 50 , 70 , and 90 percent of the peak ( 29.1 ) .
contours in the 1.4 ghz radio continuum image are @xmath12 , 3 , 10 , 20 , 50 , and 90 percent of the peak ( 13.1 ) .
green circles show the beam sizes ( fwhm ) . ] ( left ) comparison of the 24 luminosity and optical luminosity for the co - rich early - type galaxies ( circles ) and the generally co - poor ellipticals of temi , brighenti , & mathews 2008 ( crosses ) .
( right ) fir - to - radio flux ratio @xmath13 .
the circles and the arrow are the co - rich early - type galaxies studied here and crosses are the data of yun , reddy , & condon ( 2001 ) . as defined in yun et al .
, the flux ratio @xmath13 has a mean value of 2.34 for star - forming galaxies ( mostly spirals ) ; the lines at @xmath14 and @xmath15 indicate the ir excess and radio excess boundaries , respectively , at roughly @xmath16 from the mean .
, title="fig : " ] ( left ) comparison of the 24 luminosity and optical luminosity for the co - rich early - type galaxies ( circles ) and the generally co - poor ellipticals of temi , brighenti , & mathews 2008 ( crosses ) .
( right ) fir - to - radio flux ratio @xmath13 .
the circles and the arrow are the co - rich early - type galaxies studied here and crosses are the data of yun , reddy , & condon ( 2001 ) . as defined in yun et al .
, the flux ratio @xmath13 has a mean value of 2.34 for star - forming galaxies ( mostly spirals ) ; the lines at @xmath14 and @xmath15 indicate the ir excess and radio excess boundaries , respectively , at roughly @xmath16 from the mean . , title="fig : " ] in the case of ngc 2320 it is not at all clear whether the 24 emission is driven by star formation .
the galaxy s radio continuum emission is more likely due to agn activity than to star formation ( young 2005 ) .
the 24 emission is well fit by a smooth @xmath6 model with half - light semimajor and semiminor axes of 4@xmath17 2 , but this is five times smaller than the effective radius of the stellar distribution .
this modestly extended 24 emission certainly could be driven by star formation in the inner part of the molecular gas disk , but star formation activity is not required .
figure [ temifig ] compares the 24 and optical luminosities of co - rich and co - poor early - type galaxies .
the ones rich in molecular gas are typically a factor of 10 more luminous at 24 , for a given optical luminosity , than would be expected if their 24emission were entirely circumstellar in origin .
our morphological analysis has also shown that the 24 morphology is not that of point sources and is a much better match to the molecular gas distributions than to the stellar light profiles .
in addition , figure [ temifig ] shows that the fir / radio continuum flux density ratios are ( with the exception of ngc 2320 ) consistent with star formation activity . based on this evidence
we believe it is reasonable to attribute the bulk of the 24 emission in the co - rich early - type galaxies to star formation . under the assumption that the same is true for the fir emission in the iras 60and 100 bands ,
combes et al .
( 2007 ) found that the implied star formation efficiencies in the sauron early - type galaxies are similar to those in normal spirals ( combes et al .
the iras data gave no useful morphological information , though , so our analysis provides the validation for this star formation assumption .
for the majority of the co - rich early - type galaxies the close agreements between co , 24 , and radio continuum morphologies suggest that the bulk of the 24 emission should be attributed to star formation activity .
radio and fir flux density ratios are consistent with this interpretation , as are the increased @xmath18 ratios of the co - rich over the co - poor early - type galaxies .
the co , radio , and mips data are thus roughly consistent with the uv results implying star formation activity in a few tens of percent of the nearby early - type galaxies .
the necessary raw material is present , more or less , and the molecular gas is often being turned into stars .
detailed comparisons with the uv morphologies should be made as well .
future work should also make more quantitative tests of theoretical and phenomenological models of the star formation process .
for example , with a model of the gravitational potential ( from the stellar distribution and a mass - to - light ratio ) one can test whether the toomre - type local gravitational instability is consistent with the locations of star formation activity , as discussed by kawata et al .
the gas - rich early - type galaxies could also provide useful perspective on the workings of the kennicutt - schmidt relationship between the star formation rate and the gas surface density . if these models have truly captured some underlying physics of the star formation process they ought to work in the ellipticals and lenticulars as well as in the spirals .
where the molecular gas came from to begin with is , in general , still an open question . in this
regard comparisons of the kinematics of the gas and stars will be useful , as shown in young et al .
( 2008 ) , as will correlations between the gas content and the stellar structural and dynamical indicators of the galaxy s history .
the gas - to - dust ratios might also help to distinguish whether the molecular gas was acquired from a low metallicity source such as a dwarf galaxy or from a progenitor of roughly solar metallicity .
this work is based on observations made with the _
spitzer space telescope _ , which is operated by the jet propulsion laboratory ( jpl ) , california institute of technology under nasa contract 1407 .
support for this work was provided by nasa and through jpl contract 1277572 .
bendo , g. j. , calzetti , d. , et al .
2007 , mnras , 380 , 1313 bendo , g. j. , buckalew , b. a. , dale , d. a. , et al .
2006 , apj , 645 , 134 cappellari , m. , 2002 , mnras , 333 , 400 combes , f. , young , l. m. , & bureau , m. 2007 , mnras , 377 , 1795 condon , j. j. 1992 , ara&a , 30 , 575 de zeeuw , p. t. , bureau , m. , emsellem , e. , et al .
2002 , mnras , 329 , 513 goudfrooij , p. , hansen , l. , et al .
1994 , a&as , 105 , 341 kaneda , h. , onaka , t. , kitayama , t. , okada , y. , & sakon , i. 2007 , pasj , 59 , 107 kaviraj , s. , schawinski , d. , devriendt , j. e. g. , et al .
2007 , apjs , 173 , 619 kawata , d. , cen , r. , & ho , l. c. 2007 , apj , 669 , 232 kennicutt , r. c. jr .
1989 , apj , 344 , 685 lucero , d. m. , & young , l. m. 2007 , aj , 134 , 2148 okuda , t. , kohno , k. , iguchi , s. , & nakanishi , k. 2005 , apj , 620 , 673 sage , l. j. , welch , g. a. , & young , l. m. 2007 , apj , 657 , 232 sarzi , m. , et al .
2006 , mnras , 366 , 1151 schawinski , k. , kaviraj , s. , khochfar , s. , et al.2007a , apjs , 173 , 512 schawinski , k. , thomas , d. , sarzi , m. , et al.2007b , mnras , 382 , 1415 shields , j. c. 1991 , aj 102 , 1314 temi , p. , brighenti , f. , & mathews , w. g. 2008 , apj 672 , 244 temi , p. , brighenti , f. , & mathews , w. g. 2007 , apj 660 1215 welch , g. a. , & sage , l. j. 2003 , apj , 584 , 260 young , l. m. 2002 , aj 124 , 788 young , l. m. 2005 , apj 634 , 258 young , l. m. , bureau , m. , & cappellari , m. 2008 , apj , in press yun , m. s. , reddy , n. a. , & condon , j. j. 2001 , apj , 554 , 803 | many early - type galaxies are detected at 24 to 160 but the emission is usually dominated by an agn or heating from the evolved stellar population . here
we present mips observations of a sample of elliptical and lenticular galaxies which are rich in cold molecular gas , and we investigate how much of the mir to fir emission could be due to star formation activity .
the 24 images show a rich variety of structures , including nuclear point sources , rings , disks , and smooth extended emission , and comparisons to matched - resolution co and radio continuum images suggest that the bulk of the 24 emission could be traced to star formation .
the star formation efficiencies are comparable to those found in normal spirals .
some future directions for progress are also mentioned . |
our understanding of the properties of neutrinos in a plasma has recently undergone some appreciable theoretical progress @xcite . the interaction of neutrinos with a plasma particles , the creation of @xmath4 pairs , the emission of neutrinos due to the collapse of a star are of primary interests @xcite in the description of some astrophysical events such as supernova explosion .
one of the key processes upon the explosions are the large - scale hydrodynamic instabilities as well as @xmath5 driven plasma instabilities .
these processes are also believed to have occurred during the lepton stage of the early universe . during the formation of a neutron star
the collapsed core of the supernova is so dense and hot that @xmath5 and @xmath6 are trapped and are thus unable to leave the core region of the neutron star .
the rates of escape of @xmath5 and @xmath6 are very small , inside the star an equilibrium state is reached , which includes the @xmath4 concentration .
recently , the neutrino transport phenomenon during the kelvin - helmholtz phase of birth of a neutron star in the diffusion approximation was investigated by pons et al .
@xcite . a detailed knowledge of transport properties of the neutrinos in these extreme environments must include the neutrino - electron fermi weak interaction coupling .
there are different mechanisms involved in the creation of the @xmath5 , as well as pairs .
the collective effects of the stellar plasmas can significantly alter the production rate of neutrinos .
it is widely thought that the most dramatic plasma process is the decay of photons as plasmons into neutrino pairs .
it was pointed out by adams et al .
@xcite that the neutrino pair radiation can be the dominant energy loss mechanism for the plasma of the very dense stars , as well as white dwarfs , red giants and supernovas .
it has been also shown @xcite that a @xmath7 can undergo radiative decay into a photon and a @xmath2 ( @xmath8 ) in the presence of a strong magnetic field , with the strength greater than the critical value given by @xmath9 gauss .
the increase ( decrease ) of the @xmath5 ( @xmath6 ) energy within the plasma was also demonstrated @xcite .
this may have a significant and potentially detectable effect .
observationally , the recent results from the kamiokande group provide strong evidence for the existence of neutrino oscillations .
these results , together with the increased confidence in ability to produce and manipulate intense muon beams makes feasible the future neutrino factories based on muon storage rings .
a muon storage ring , as presently envisioned , would have energy in the 10 - 50 gev range , and produce directed beams of intense neutrinos , and naturally such an intense stream of neutrinos can exist in astrophysical and cosmological plasmas .
thus , it is of interest to examine if new physical processes that are caused by the intense flux of neutrinos , can appear in such a plasma , i.e. , of particular conceptual interest are effects which have no counterpart in vacuum . in this paper
we consider a medium of neutrino gas and plasma in semitransparent regions , where collective process plays an important role . in our consideration
we assume that the interaction between neutrinos is weak in comparison with the neutrino - electron interaction , i.e. , the neutrino gas is ideal .
we can also ignore the spin of neutrinos in an isotropic plasma , since in this case the energy of neutrinos is independent of the spin operator . with this modeling , here we show that there exists the opposite effect that has been proposed by others @xcite : namely that the neutrinos production does not generally lead to energy losses from a hot and dense system .
in the following we shall demonstrate the excitation of longitudinal oscillations in neutrino flux - plasma medium .
for this purpose , we make use of hydrodynamic equations for the neutrino gas , which have been derived by tsintsadze et al .
these equations for the neutrino density @xmath10 and the fluid velocity @xmath11 take the following forms @xmath12 where @xmath13 describes the universal fermi weak interaction , @xmath14 is the fermi coupling constant , @xmath15 , and @xmath16 is the electron plasma density .
the electron continuity equation takes the form : @xmath17 for the equation of motion of electrons in the presence of a neutrino gas , we have @xmath18 where we have neglected the pressure term in comparison with the first term on the right hand side under assumption that @xmath19 , where @xmath20 is the debye length of the electrons , and
l is the characteristic length of the system .
note that the last term in eq.([motion e ] ) is the neutrino ponderomotive force . a more elaborate discussion on the neutrino ponderomotive force is given in @xcite .
the electrostatic potential @xmath21 is determined from poisson equation @xmath22 we first consider a situation in which the ions are in an equilibrium state , i.e. @xmath23 , where @xmath24 is the equilibrium value of the ion density . in this consideration
the range of frequency of small perturbation of quantities of the medium is much less than the langmuir frequency of the electrons .
thus , we can neglect the inertial term in eq.([motion e ] ) , and define @xmath21 from it .
finally , we obtain from the poisson equation the expression for the electron density @xmath25 where @xmath26 is the equilibrium value of the neutrino density . substituting expression ( [ eldens ] ) into eq.([momen ] ) we get the following equation , which except quantities of the neutrinos contains only the electron charge @xmath27 where @xmath28
. equations ( [ cont n ] ) and ( [ mom1 ] ) form a closed set of equations that describes the linear as well as nonlinear waves in the neutrino gas .
let us consider the propagation of small perturbations in a homogeneous plasma - neutrino beam . to this end , we linearize eqs.([cont n ] ) and ( [ mom1 ] ) with respect to the perturbations , which are represented as @xmath29 where the suffix @xmath30 denotes the constant equilibrium value , and @xmath31 are small variations in the wave .
after linearization of eqs.([cont n ] ) and ( [ mom1 ] ) , we will seek plane wave solutions proportional to @xmath32 .
we can then derive the following dispersion relation @xmath33 this leads to a solution of the form @xmath34 such that : @xmath35 and the growth rate for the unstable branch becomes : @xmath36 this solution clearly describes the emission of the low - frequency neutrinos inside a resonance cone ( @xmath37 ) , similar to the well - known cherenkov emission of electromagnetic waves by charged particles moving in uniform medium with a velocity larger than the phase velocities of the emitted waves . on the other hand
, strictly speaking the effect considered here is physically quite different from the usual cherenkov emission by charged particles .
the point is that the neutrino flux can no longer excite langmuir waves and energetic ( heavy ) neutrino scatters of the electron density perturbation ( caused by the ponderomotive force of neutrinos ) and emits low ( light ) energy neutrino .
we specifically note here that such decay process @xmath3 is forbidden without a plasma .
let us estimate the growth rate ( [ i m ] ) for some typical neutrino parameters .
accordingly , we take the @xmath2 energy to be about 100 kev , the @xmath7 energy to be about 1 - 10 mev , and @xmath38 .
it is important to emphasize that the growth rate does not depend on the density and temperature of a plasma .
then for the growth rate ( [ i m ] ) we obtain @xmath39 . comparing this with the collision frequency @xmath40 of neutrinos
, it can be seen the latter to be much less than the former .
indeed , for the given mean free path @xmath41 cm @xcite , the collision frequency is @xmath42 , i.e. under these circumstances , the medium for the neutrino is `` collisionless '' .
we now discuss the nonlinear features of evolution of the unstable waves .
if we consider the case , when the period of oscillations , @xmath43 , is much less than the wave amplitude growth time , @xmath44 , then the dispersion relation can be written in a one - dimensional perturbation as @xmath45 .
this corresponds to the condition for the neglect of the third derivatives in eq.([mom1 ] ) .
doing so , we obtain @xmath46 which describes a beam of noninteracting neutrinos . as was shown in @xcite , the equation type of eq.([mom2 ] ) has solutions with the characteristic properties of nonlinear waves .
it has been verified that the initial perturbation of the velocity @xmath47 in the phase plane @xmath48 after some time leads the wave @xmath49 to break .
thus , our results indicate that the neutrino beam of noninteracting particles , similar to a beam of plasma particles , has many properties of nonlinear system .
namely , it is subject to wave breaking and to generation of higher harmonics .
if we take into account the temperature @xmath50 of neutrinos and suppose that the neutrino gas is an ideal ( @xmath51 ) , then the equation of motion ( [ mom1 ] ) modifies as @xmath52 where @xmath53 , @xmath54 is the velocity of sound in the neutrino gas , and @xmath55 is the effective adiabatic index .
equations ( [ cont n ] ) and ( [ momt ] ) are so called boussinesq equations .
we can see that the character of nonlinear processes strongly depends on the dispersion , i.e. , the dependence of phase velocity on the wave number .
the most important nonlinear effect here , as it was shown above , is that the wave front steepens . at this point
the dispersion becomes significant .
waves having different wave number have different phase velocities , and the nonlinear steepening of front can be compensated by the dispersive spread . therefore ,
in this case the stationary waves can exist , which propagate with constant velocities without changing their shape ( e.g. , solitons ) . to note , eqs.([cont n ] ) and ( [ momt ] ) describe also the nonstationary waves @xcite .
equations type of eqs.([cont n ] ) and ( [ momt ] ) have been thoroughly studied , and it has been shown that these kind of equations have many periodic as well as solitary solutions . if we use the common method for the weak dispersive medium , we obtain from eqs.([cont n ] ) and ( [ momt ] ) the korteweg - de vries equation for the neutrino velocity @xmath56 where @xmath57 , and @xmath58 . this equation is well studied and we do not repeat all solutions and conservation laws of it . our aim was to prove that the neutrino gas exhibits the same features as the waves in plasmas ( e.g. electromagnetic waves , waves on the surface of a heavy liquid , waves in biology , etc . ) .
we next consider the excitation of an ion - sound waves by the neutrino flux .
above we studied the high - frequency oscillations neglecting the ion dynamics . however , the ion impact turns out to be extremely important in the low - frequency waves which may be excited in a strong non - isothermal plasma with hot electrons and cold ions . as for the electrons we may assume that the electrons are in equilibrium under the conditions of low - frequency oscillation and that their density is determined by the boltzmann formula :
@xmath59 the ions can be described hydrodynamically by means of the hd equations ( @xmath60 ) : @xmath61 and the poisson equation in this case is @xmath62 after linearization of the set of eqs.([cont n]),([momen]),([ion ] ) and ( [ pois2 ] ) we obtain the dispersion relation for the ion - sound oscillation in the presence of the monoenergetic neutrino beam as @xmath63 where @xmath64 , and @xmath65 is simply the ion - sound frequency , @xmath66 here @xmath67 is the ion - sound velocity .
we examine the dispersion relation ( [ disi ] ) for two interesting cases .
let us first discuss the small wavelength case , @xmath68 and @xmath69 . in this case eq.([disi ] ) reduces to @xmath70 which has a solution of the form @xmath34 with @xmath71 hence , this solution describes the emission of low - frequency neutrinos inside a resonance cone ( @xmath72 ) .
next , for the long wavelengths , @xmath73 , the dispersion relation ( [ disi ] ) casts in to the form @xmath74 here we look for the solution with coincide roots , i.e. @xmath75 and @xmath76 , and get for the growth rate of the oscillatory instabilities the following expression @xmath77
let us now investigate the nonlinear low - frequency waves . to this end
, we add the ion pressure to the equation of motion of the ions , eq.([ion ] ) , and since @xmath69 neglect the inertial term . then solution of this equation is @xmath78 if we further impose the quasi - neutrality , @xmath79 , we obtain @xmath80 substituting ( [ phi ] ) into the boltzmann distribution function of the electrons ( [ bol ] ) , we get @xmath81 for the neutrino gas the natural requirement is @xmath82 , so that we have @xmath83 this equation when substituted into eq.([momen ] ) yields the equation of motion of the neutrinos @xmath84 where @xmath85 .
if we linearize this equation and the equation of density ( [ cont n ] ) , then we obtain the dispersion equation ( [ disi1 ] ) for @xmath60 .
we now consider the one - dimensional nonlinear waves , for this case equations are @xmath86 these equations have exact solutions for the initial profile of the neutrino density @xmath87 and in nondimensional quantities are written as @xmath88 where @xmath89 , @xmath90 , @xmath91 , @xmath92 , with @xmath93 being the langmuir frequency of ions . from eqs.([ro ] ) and ( [ v ] ) , we can readily derive @xmath94 examining eq.([xi ] ) , we now discuss the question how the maximum of the initial density changes in time . for this purpose
, we take a derivative of eq.([xi ] ) by @xmath95 , i.e. @xmath96 , and allow it to be infinite . from this condition
, we find the expression for the maximum density @xmath97 one can see that at @xmath98 , @xmath99 , @xmath100 .
then with increase of time ( @xmath101 ) the shape of the density changes , namely becomes narrow .
also , the center of the density ( @xmath102 ) increases with rise of @xmath101 and reaches another maximum , @xmath103 at @xmath104 . after
a certain value @xmath105 , the shock waves are formed , and derivatives are @xmath106 @xmath107 however , the density itself remains continuous .
using these conditions from eq.([xi ] ) follows the relation @xmath108 , where @xmath109 and @xmath110 .
3d numerical simulation which underscores the above picture is carried out .
namely , we consider 3d problem , when the initial profile of density of neutrinos is the gaussian distribution @xmath111 numerical studies have shown that the dynamics of non - linearity is determined by the dimensionality of problem and of the initial profile of the neutrino density . in order to confirm the validity of 1d analytical calculation
, we first discuss 3d numerical analysis of eqs.([cont n ] ) and ( [ mom1n ] ) for pancake - shaped beams , i.e. , @xmath112 , and @xmath113 . figures 1a and 1b show that the width of the density becomes shorter in time along the propagation direction ( z - axis ) .
the density of neutrinos is concentrated in the small region ( on the z - axis ) , and at @xmath114 starts to form the shock wave . whereas figures 1c and 1d show the insignificant expansion of the pancake beams in the transverse direction .
therefore , we can conclude that the 1d analytical calculation is very good approximation for the 3d pancake - shaped beams .
quite different processes develop when the initial distribution of neutrinos has a form of the bullet , i.e. , when @xmath115 and @xmath116 .
we can see that there are two special competitive regimes : one , which is due to the self - focusing , is the compression of the neutrino density in the transverse direction ( see figures 2a and 2b ) , and the second is the expansion of the density along the propagation direction ( see figures 2b and 2c ) .
the redistribution of neutrinos in the transverse direction is accompanied by the formation of sharp maximum of density near the z - axis ( @xmath117 ) , and as figure 2a shows , the formation of the shock wave takes place at t=2.8 .
we have studied the problem of weak fermi interaction of a neutrino beam with a plasma .
novel fluid instabilities driven by the neutrino flux in a plasma are observed .
the range of wavelengths corresponding to several instabilities , discussed in this paper , for relevant parameters of the neutrino gas and plasma is @xmath118 .
we have found that for the neutrino beam there exists a cherenkov type of emission of low energy neutrinos ; this is different from the usual cherenkov effect , though .
it should be emphasized that in our case the physical mechanism is that the high energy neutrino does not just excite the plasma wave but also can scatter from density perturbations of electrons and can emit the low energy neutrino .
we also found that in the usual hot plasma environments such as supernova explosions , the effect considered above prevents neutrino escape .
since the low energy neutrinos can not stream away from the central regions of the plasma one can have an accumulation of such neutrinos in a dense stellar medium .
thus , the above discussed processes can considerably modify certain phases of evolution in a stellar model .
it should be emphasized that unlike the previous models @xcite,@xcite on neutrino pair emission by a stellar plasma , the emission of low energy neutrinos by high energy neutrinos , as discussed in this paper , does not depend on the density and temperature of a plasma .
the growth rate found for some typical neutrino parameters is @xmath119 . which is much larger than the collision frequency of neutrinos , @xmath120 .
thus in this case the medium for the neutrino is `` collisionless '' .
the nonlinear stationary as well as nonstationary waves in the neutrino gas are also discussed . the dispersive effect , which is responsible for the existence of the nonlinear stationary waves , e.g. solitons , in the neutrino gas is due to the electron density modulation by the neutrino ponderomotive force @xcite .
our results indicate that a neutrino beam is subject to wave breaking and shock wave formation .
thus , high energy neutrinos will lose energy also by wave breaking in addition to cherenkov emission , leading to plasma heating ( and neutrino cooling ) . whereas , the shock waves can produce a relativistic energy flow .
in addition , the solitons can also be a potential candidates for the generation of relativistic particles .
these particles then release the energy and produce the observed radiation in gamma - ray bursts ( grbs ) .
hence , one has a origin of radiation in the modeling of burst sources , which requires a discussion of particle acceleration processes .
since the simplest , most conventional , and practically inevitable interpretation of the observations of grbs is that grbs result from the conversion of the kinetic energy of ultra - relativistic particles to radiation .
it is well accepted that many of them originate in the very distant , early universe . in the early universe , the processes discussed in the paper
can also lead to formation of nonlinear structures , contributing to the formation of the large scale structure of the universe @xcite . which , is believed , grew gravitationally out of small density fluctuations @xcite .
the effect of this density variation in the early universe was left on the cosmic microwave background radiation in the form of spatial temperature fluctuations .
it should be emphasized that the gravity , however , can not produce these fluctuations , but increase alone .
therefore a discussion of the physical models of generation of the initial matter density fluctuations is very crucial .
there is no comprehensive model at present that can explain their origin .
the weak fermi interaction in a plasma , as discussed in this paper , can be an alternative and a new source for the required density perturbations . as demonstrated , a long lived nonlinear structures , which carry large amount of mass and energy , are generated in such a system .
since an initial localization of mass and energy is exactly that the gravity needs for eventual structure formation , weak fermi interaction may have provided a decisive element in the formation of a large scale map of the observable universe .
the wigner - moyal equation in quantum kinetic theory is the analogy of the one - particle liouville equation in kinetic theory . in order to derive this equation , we start with a dispersion relation of a single neutrino , taking into account the universal weak fermi interaction between plasma electrons and neutrinos ( @xmath121 ) @xmath122 where e is the energy , p is the momentum , @xmath123 is the neutrino rest mass .
this equation has two distinct energy solutions @xmath124 the physical meaning of this result is best understood when we are in the rest frame of the background medium , which creates the potential energy u. the positive sign in eq.([ensol ] ) corresponds to the neutrino solution and it shows that , for a given magnitude of the particle momentum , the energy of the neutrino is increased by an amount u with respect to its value in vacuum .
the other solution in eq.([ensol ] ) , corresponding to the negative sign , is that of an antineutrino solution . as mentioned in the introduction ,
in an isotropic plasma one can ignore the spin of neutrinos , and therefore we can consider a scalar wave function @xmath125 , associated with the neutrino gas , and we can write in equilibrium @xmath126 where @xmath127 , @xmath128 , @xmath129 and @xmath130 is the fermi frequency .
on the other hand , if the neutrino gas is not in equilibrium , the amplitude of @xmath131 will slowly change in space and time due to the interaction of neutrinos with the background plasma fluctuations .
we can then use the geometrical optic approximation , @xmath132 @xmath133 and @xmath134 ( @xmath135 , @xmath136 ) . doing expansion @xmath137 and neglecting higher order derivatives greater than @xmath138 in eq.([expen ] )
, we arrive from eq.([sfan ] ) to @xmath139 where use was made of the relation @xmath140 , and the following notations were introduced @xmath141 we now write the product of two wave functions @xmath125 , which are defined at two distinct points and instants of time @xmath142 where we have introduced new space and time variables defined by @xmath143 let us make a fourier transform of ( [ twops ] ) on the variables @xmath144 in order to introduce the wigner function @xmath145 @xmath146 this is the generalized wigner distribution function .
integration of ( [ wiggen ] ) by the frequency @xmath147 leads to the ordinary expression of the wigner distribution function @xcite as @xmath148 from ( [ wiggen ] ) also follows important relation @xmath149 assuming that the neutrinos verify the dispersion relation ( [ ensol ] ) , we can say that for a given value @xmath150 we always have a well - defined value of the frequency , @xmath151 .
this means that we can write @xmath152 integration over the wave vector spectrum leads to the density of neutrinos @xmath153 following the procedure described in previous works @xcite , we can derive an equation for the function ( [ twops ] ) , the result is @xmath154 now making the fourier transformation of eq.([func ] ) , we obtain an evolution equation for function ( [ wiggen ] ) in the form @xmath155 where @xmath156 if all derivatives of @xmath157 exist , we can expand @xmath158 around @xmath159 , i.e. @xmath160\delta n(\vec{r},t)\end{aligned}\ ] ] and obtain the generalized wigner - moyal equation @xmath161 integration of eq.([moyalg ] ) by the frequency leads to the wigner - moyal equation for the wigner function ( [ wigor ] ) without the last term on the right hand side of eq.([moyalg ] ) . if we just retain the first term on the right hand side of eq.([moyalg ] ) , i.e. , the term corresponding to @xmath162 , we obtain the well - known liouville - vlasov kinetic equation for the distribution function ( [ wiggen ] ) , but with an additional time derivative term @xmath163 the stationary solution of eq.([vlas ] ) is the fermi distribution function @xmath164 from eq.([vlas ] ) we can now derive a set of fluid equations for the neutrino gas . to this end
we shall introduce the principal definitions of quantities which describe the state of the neutrinos in a plasma .
namely , the total number of neutrinos is defined as @xmath165 the total energy of the neutrino gas @xmath166 and the neutrino mean velocity @xmath167 having definitions ( [ numb]-[mean ] ) and using the usual momentum procedure , from eq.([vlas ] ) we obtain a set of fluid equations @xmath168 @xmath169 @xmath170 where we used the definitions of the momentum density @xmath171 and @xmath172 the quantity @xmath173 plays role of the neutrino pressure and is defined as @xmath174 equations ( [ apcon]-[apen ] ) together with equations of the plasma , given in this paper , describe the collective process , which plays a more important role than the pair collision effect .
n.l.tsintsadze , j.t.mendonca , p.k.shukla , phys .
a * 249 * , 110 ( 1998 ) ; j.t.mendonca , et al . ,
ibid , * 239 * , 373 ( 1998 ) ; n.l.tsintsadze , p.k.shukla , j.t.mendonca , and l.n.tsintsadze , phys .
scripta * t82 * , 128 ( 1999 )
. l.n.tsintsadze , and n.l.tsintsadze , _ proceedings of the international conference on superstrong fields in plasmas _
varenna , 1997 , ed . by m.lontano , ( american institute of physics , new york , 1998 ) , 170 . | the system of neutrino - antineutrino @xmath0 - plasma is considered taking into account their weak fermi interaction .
new fluid instabilities driven by strong neutrino flux in a plasma are found .
the nonlinear stationary as well as nonstationary waves in the neutrino gas are discussed .
it is shown that a bunch of neutrinos , drifting with a constant velocity across a homogeneous plasma , can also induce emission of lower energy neutrinos due to scattering , i.e. the decay of a heavy neutrino @xmath1 into a heavy and a light neutrino @xmath2 ( @xmath3 ) in a plasma .
furthermore we find that the neutrino production in stars does not lead in general to energy losses from the neutron stars . |
the use of probability to reason about uncertainty is key to modern science and engineering , and the operation of _ conditioning _ , used to perform bayesian inductive reasoning in probabilistic models , directly raises many of its most important computational problems .
faced with probabilistic models of increasingly complex phenomena that stretch or exceed the limitations of existing representations and algorithms , researchers have proposed new representations and formal languages for describing joint distributions on large collections of random variables , and have developed new algorithms for performing automated probabilistic inference .
what are the limits of this endeavor ?
can we hope to automate probabilistic reasoning via a general inference algorithm that can compute conditional probabilities for an _ arbitrary _ computable joint distribution ?
we demonstrate that there are computable joint distributions with noncomputable conditional distributions . of course
, the fact that generic algorithms can not exist for computing conditional probabilities does not rule out the possibility that large classes of distributions may be amenable to automated inference .
the challenge for mathematical theory is to explain the widespread success of probabilistic methods and characterize the circumstances when conditioning is possible . in this vein , we describe broadly applicable conditions under which conditional probabilities are computable . within probabilistic artificial intelligence ( ai ) and machine learning ,
the study of formal languages and algorithms for describing and computing answers from probabilistic models is the subject of _
probabilistic programming_. probabilistic programming languages themselves build on modern programming languages and their facilities for recursion , abstraction , modularity , etc . , to enable practitioners to define intricate , in some cases infinite - dimensional , models by implementing a generative process that produces an exact sample from the model s joint distribution .
( see , e.g. , ibal @xcite , @xmath0@xcite , church @xcite , and @xcite . for related and earlier efforts , see , e.g. , pha @xcite , infer.net @xcite , markov logic @xcite .
probabilistic programming languages have been the focus of a long tradition of research within programming languages , model checking and formal methods . ) in many of these languages , one can easily represent the higher - order stochastic processes ( e.g. , distributions on data structures , distributions on functions , and distributions on distributions ) that are essential building blocks in modern nonparametric bayesian statistics .
in fact , the most expressive such languages are each capable of describing the same robust class as the others the class of _ computable distributions _ , which delineates those from which a probabilistic turing machine can sample to arbitrary accuracy .
traditionally , inference algorithms for probabilistic models have been derived and implemented by hand .
in contrast , probabilistic programming systems have introduced varying degrees of support for computing conditional distributions . given the rate of progress toward broadening the scope of these algorithms
, one might hope that there would eventually be a generic algorithm supporting the entire class of computable distributions . despite recent progress towards a general such algorithm , support for conditioning with respect to continuous random variables
has remained ad - hoc and incomplete .
our results explain why this is necessarily the case . in order to characterize the computational limits of probabilistic inference
, we work within the framework of _ computable probability theory _ , which pertains to the computability of distributions , random variables , and probabilistic operations ; and builds on the classical computability theory of deterministic functions . just as the notion of a turing machine allows one to prove results about discrete computations performed using an arbitrary ( sufficiently powerful ) programming language , the notion of a probabilistic turing machine provides a basis for precisely describing the operations that probabilistic programming languages are capable of performing .
the tools for describing computability in this setting are drawn from the theory of _ computable metric spaces _ , within the subject of _
computable analysis_. this theory gives us the ability to study distributions on arbitrary computable metric spaces , including , e.g. , distributions on distributions . in section [ compprob ]
we present the necessary definitions and results from computable probability theory . for an experiment with a discrete set of outcomes , computing conditional probabilities
is , in principle , straightforward as it is simply a ratio of probabilities .
however , in the case of conditioning on the value of a continuous random variable , this ratio is undefined .
furthermore , in modern bayesian statistics , and especially the probabilistic programming setting , it is common to place distributions on higher - order objects , and so one is already in a situation where elementary notions of conditional probability are insufficient and more sophisticated measure - theoretic notions are necessary .
kolmogorov @xcite gave an axiomatic characterization of conditional probabilities , but this definition provides no recipe for their calculation .
other issues also arise : in this setting , conditional probabilities are formalized as measurable functions that are defined only up to measure zero sets . therefore , without additional assumptions , a conditional probability is not necessarily well - defined for any particular value of the conditioning random variable .
this has long been understood as a challenge for statistical applications , in which one wants to evaluate conditional probabilities given particular values for observed random variables . in this paper , we are therefore especially interested in situations where it makes sense to ask for the conditional distribution given a particular point .
in particular , we focus on the case when conditional distributions are everywhere or almost everywhere continuous , and thus can be given a unique definition for individual points in the support of the underlying measure .
as we will argue , this is necessary if there is to be any hope of conditioning being computable . under certain conditions , such as
when conditional densities exist , conditioning can proceed using the classic bayes rule ; however , it may not be possible to compute the density of a computable distribution ( if the density exists at all ) .
the probability and statistics literature contains many ad - hoc techniques for calculating conditional probabilities in special circumstances , and this state of affairs motivated much work on constructive definitions ( such as those due to tjur @xcite , @xcite , @xcite , pfanzagl @xcite , and rao @xcite , @xcite ) , but this work has often not been sensitive to issues of computability .
we recall the basics of the measure - theoretic approach to conditional distributions in section [ conddist ] , and in section [ ccd2 ] we use notions from computable probability theory to consider the sense in which conditioning could be potentially computable .
conditional probabilities for computable distributions on finite , discrete sets are clearly computable , but may not be efficiently so . in this
finite discrete setting , there are already interesting questions of computational complexity , which have been explored by a number of authors through extensions of levin s theory of average - case complexity @xcite .
for example , under cryptographic assumptions , it is difficult to sample from the conditional distribution of a uniformly - distributed binary string of length @xmath1 given its image under a one - way function . this can be seen to follow from the work of ben - david , chor , goldreich , and luby @xcite in their theory of polynomial - time samplable distributions , which has since been extended by yamakami @xcite and others . extending these complexity results to the more general setting considered here could bear on the practice of statistical ai and machine learning .
osherson , stob , and weinstein @xcite study learning theory in the setting of _ identifiability in the limit _
( see @xcite and @xcite for more details on this setting ) and prove that a certain type of `` computable bayesian '' learner fails to identify the index of a ( computably enumerable ) set that is `` computably identifiable '' in the limit .
more specifically , a so - called `` bayesian learner '' is required to return an index for a set with the highest conditional probability given a finite prefix of an infinite sequence of random draws from the unknown set .
an analysis by roy @xcite of their construction reveals that the conditional distribution of the index given the infinite sequence is an everywhere discontinuous function ( on every measure one set ) , hence noncomputable for much the same reason as our elementary construction involving a mixture of measures concentrated on the rationals and on the irrationals ( see section [ sec : discontinuous ] ) . as we argue , it is more appropriate to study the operator when it is restricted to those random variables whose conditional distributions admit versions that are continuous everywhere , or at least on a measure one set . our work is distinct from the study of conditional distributions with respect to priors that are universal for partial computable functions ( as defined using kolmogorov complexity ) by solomonoff @xcite , zvonkin and levin @xcite , and hutter @xcite .
the computability of conditional distributions also has a rather different character in takahashi s work on the algorithmic randomness of points defined using universal martin - lf tests @xcite .
the objects with respect to which one is conditioning in these settings are typically _ computably enumerable _ , but not computable . in the present paper , we are interested in the problem of computing conditional distributions of random variables that are _ computable _ , even though the conditional distribution may itself be noncomputable . in the most abstract setting ,
conditional probabilities can be constructed as radon - nikodym derivatives .
in work motivated by questions in algorithmic randomness , hoyrup and rojas @xcite study notions of computability for absolute continuity and for radon - nikodym derivatives as elements in @xmath2 , i.e. , the space of integrable functions . they demonstrate that there are computable measures whose radon - nikodym derivatives are not computable as points in @xmath2 , but these counterexamples do not correspond with conditional probabilities of computable random variables .
hoyrup , rojas and weihrauch @xcite show an equivalence between the problem of computing general radon - nikodym derivatives as elements in @xmath2 and computing the characteristic function of computably enumerable sets .
however , conditional probabilities are a special case of radon - nikodym derivatives , and moreover , a computable element in @xmath2 is not well - defined at points , and so is not ideal for statistical purposes . using their machinery , we demonstrate the non-@xmath2-computability of our main construction .
but the main goal of our paper is to provide a detailed analysis of the situation where it makes sense to ask for the conditional probability at points , which is the more relevant scenario for statistical inference .
following our presentation of computable probability theory and conditional probability in sections [ compprob ] through [ ccd2 ] , we provide our main positive and negative results about the computability of conditional probability , which we now summarize . in proposition [ discontinuouscond ]
, we construct a pair @xmath3 of computable random variables such that every version of the conditional probability @xmath4 $ ] is discontinuous everywhere , even when restricted to a @xmath5-measure one subset .
( we make these notions precise in section [ ccd2 ] . )
the construction makes use of the elementary fact that the indicator function for the rationals in the unit interval
the so - called dirichlet function is itself nowhere continuous . because every function computable on a domain @xmath6 is continuous on @xmath6
, discontinuity is a fundamental barrier to computability , and so this construction rules out the possibility of a completely general algorithm for conditioning .
a natural question is whether conditioning is a computable operation when we restrict the operator to random variables for which _ some _ version of the conditional distribution is continuous everywhere , or at least on a measure one set .
our central result , theorem [ mainnegativeresult ] , states that conditioning is not a computable operation on computable random variables , even in this restricted setting .
we construct a pair @xmath7 of computable random variables such that there is a version of the conditional distribution @xmath8 $ ] that is continuous on a measure one set , but no version of @xmath8 $ ] is computable .
( indeed , every individual conditional probability fails to be even lower semicomputable on any set of sufficient measure . )
moreover , the noncomputability of @xmath8 $ ] is at least as hard as the halting problem , in that if some oracle @xmath9 computes @xmath8 $ ] , then @xmath9 computes the halting problem .
the construction involves encoding the halting times of all turing machines into the conditional distribution @xmath10 $ ] , while ensuring that the joint distribution remains computable .
in theorem [ newmainnegativeresult ] we strengthen our central result by constructing a pair of computable random variables whose conditional distribution is noncomputable but has an everywhere continuous version with infinitely differentiable conditional probabilities .
this construction proceeds by smoothing out the distribution constructed in theorem [ mainnegativeresult ] , but in such a way that one can still compute the halting problem relative to the conditional distribution . despite the noncomputability of conditioning in general , conditional distributions are often computable in practice .
we provide some explanation of this phenomenon by characterizing several circumstances in which conditioning _ is _ a computable operation . under suitable computability hypotheses , conditioning is computable in the discrete setting ( lemma [ discreteconditioning ] ) and where there is a conditional density ( corollary [ independentdensity ] ) .
we also characterize a situation in which conditioning is possible in the presence of noisy data , capturing many natural models in science and engineering .
let @xmath11 , @xmath12 and @xmath13 be computable random variables , and suppose that @xmath14 is absolutely continuous with a bounded computable density @xmath15 and @xmath13 is independent of @xmath11 and @xmath12 .
we can think of @xmath16 as the corruption of an idealized measurement @xmath11 by independent source of additive error @xmath13 . in corollary [ indnoise ]
, we show that the conditional distribution @xmath17 $ ] is computable ( even if @xmath18 $ ] is not ) .
finally , we discuss how symmetry can contribute to the computability of conditional distributions .
we now give some background on computable probability theory , which will enable us to formulate our results .
the foundations of the theory include notions of computability for probability measures developed by edalat @xcite , weihrauch @xcite , schroeder @xcite , and gcs @xcite .
computable probability theory itself builds off notions and results in computable analysis . for a general introduction to this approach to real computation ,
see weihrauch @xcite , braverman @xcite or braverman and cook @xcite .
we first recall some elementary definitions from computability theory ( see , e.g. rogers ( * ? ? ?
* ch . 5 ) ) .
we say that a set of natural numbers ( potentially in some correspondence with , e.g. , rationals , integers , or other finitely describable objects with an implicit enumeration ) is _ computably enumerable _
( c.e . ) when there is a computer program that outputs every element of the set eventually .
we say that a sequence of sets @xmath19 is c.e .
_ uniformly in @xmath1 _ when there is a computer program that , on input @xmath1 , outputs every element of @xmath20 eventually .
a set is co - c.e . when its complement is c.e .
( and so the ( uniformly ) computable sets are precisely those that are both ( uniformly ) c.e . and co - c.e ) .
we now recall basic notions of computability for real numbers ( see , e.g. , ( * ? ? ?
4.2 ) or ( * ? ? ?
* ch . 1.8 ) ) .
we say that a real @xmath21 is a _
real _ when the set of rationals @xmath22 is c.e .
similarly , a _
co - c.e .
real _ is one for which @xmath23 is c.e .
( c.e . and co - c.e .
reals are sometimes called _ _ left - c.e.__and _ right - c.e .
_ reals , respectively . )
a real @xmath21 is _ computable _ when it is both c.e . and co - c.e .
equivalently , a real is computable when there is a program that approximates it to any given accuracy ( e.g. , given an integer @xmath24 as input , the program reports a rational that is within @xmath25 of the real ) .
a function @xmath26 is lower ( upper ) semicomputable when @xmath27 is a c.e .
( co - c.e . )
real , uniformly in @xmath1 ( or more precisely , when the rational lower ( upper ) bounds of @xmath27 are c.e .
uniformly in @xmath1 ) .
the function @xmath28 is computable if and only if it is both lower and upper semicomputable .
computable metric spaces , as developed in computable analysis @xcite , @xcite and effective domain theory @xcite , @xcite , provide a convenient framework for formulating results in computable probability theory . for consistency , we largely use definitions from @xcite and @xcite
. additional details about computable metric spaces can also be found in ( * ? ? ?
8.1 ) and ( * ? ? ?
* b.3 ) . a * computable metric space * is a triple @xmath29 for which @xmath30 is a metric on the set @xmath31 satisfying 1 .
@xmath32 is a complete separable metric space ; 2 .
@xmath33 is an enumeration of a dense subset of @xmath31 , called * ideal points * ; and , 3 .
the real numbers @xmath34 are computable , uniformly in @xmath35 and @xmath36 .
let @xmath37 denote the ball of radius @xmath38 centered at @xmath39 .
we call @xmath40 the * ideal balls of @xmath31 * , and fix the canonical enumeration of them induced by that of @xmath41 and @xmath42 . for example
, the set @xmath43 is a computable metric space under the discrete metric , characterized by @xmath44 .
cantor space , the set @xmath45 of infinite binary sequences , is a computable metric space under its usual metric and the dense set of eventually constant strings ( under a standard enumeration of finite strings ) .
the set @xmath46 of real numbers is a computable metric space under the euclidean metric with the dense set @xmath42 of rationals ( under its standard enumeration ) .
we let @xmath47 denote the borel @xmath48-algebra on a metric space @xmath31 , i.e. , the @xmath48-algebra generated by the open balls of @xmath31 . in this paper , measurable functions will always be with respect to the borel @xmath48-algebra of a metric space .
let @xmath29 be a computable metric space .
a point @xmath49 is * computable * when there is a program that enumerates a sequence @xmath50 in @xmath41 where @xmath51 for all @xmath35 .
we call such a sequence @xmath50 a * representation * of the point @xmath52
. a real @xmath53 is computable ( as in section [ cereal ] ) if and only if @xmath54 is a computable point of @xmath46 ( as a computable metric space ) .
although most of the familiar reals are computable , there are only countably many computable reals , and so almost every real is not computable . the notion of a c.e .
open set ( or @xmath55 class ) is fundamental in classical computability theory , and admits a simple definition in an arbitrary computable metric space .
let @xmath29 be a computable metric space with the corresponding enumeration @xmath56 of the ideal open balls @xmath57 .
we say that @xmath58 is a * c.e .
open set * when there is some c.e
. set @xmath59 such that @xmath60 .
note that the class of c.e .
open sets is closed under computable unions and finite intersections
. a computable function can be thought of as a continuous function whose local modulus of continuity is witnessed by a program .
it is important to consider the computability of _ partial _ functions , because many natural and important random variables are continuous only on a measure one subset of their domain .
[ comp - partial - func ]
let @xmath61 and @xmath62 be computable metric spaces , the latter with the corresponding enumeration @xmath63 of the ideal open balls @xmath64 .
a function @xmath65 is said to be * computable on @xmath66 * when there is a computable sequence @xmath67 of c.e .
open sets @xmath68 such that @xmath69 \cap r = u_n \cap r$ ] for all @xmath70 .
we call such a sequence @xmath71 a * witness * to the computability of @xmath28 . in particular , if @xmath28 is computable on @xmath72 , then the inverse image of c.e .
open sets are c.e .
open ( in @xmath72 ) sets , and so we can see computability as a natural restriction on continuity .
[ comppointremark ] let @xmath31 and @xmath73 be computable metric spaces .
if @xmath74 is computable on some subset @xmath75 , then for every _ computable _ point @xmath76 , the point @xmath77 is also computable .
one can show that @xmath28 is computable on @xmath72 when there is a program that uniformly transforms representations of points in @xmath72 to representations of points in @xmath31 .
( for more details , see ( * ? ? ?
3.3.2 ) . )
suppose that @xmath78 is computable on @xmath75 with @xmath71 a witness to the computability of @xmath28 .
one can show that there is an effective @xmath79 set @xmath80 and a function @xmath81 such that @xmath82 is computable on @xmath83 , the restriction of @xmath82 to @xmath72 and @xmath28 are equal as functions , and @xmath71 is a witness to the computability of @xmath82 .
furthermore , a @xmath79-code for @xmath83 can be chosen uniformly in the witness @xmath84 .
one could consider such an @xmath82 to be a canonical representative of the computable partial function @xmath28 with witness @xmath71 .
note , however , that the @xmath79-set chosen depends not just on @xmath28 , but also on the witness @xmath71 . in particular , it is possible that two distinct witnesses to the computability of @xmath28 could result in distinct @xmath79-sets . intuitively , a random variable maps an input source of randomness to an output , inducing a distribution on the output space . here
we will use a sequence of independent fair coin flips as our source of randomness .
we formalize this via the probability space @xmath85 , where @xmath45 is the product space of infinite binary sequences , @xmath86 is its borel @xmath48-algebra ( generated by the set of basic clopen cylinders extending each finite binary sequence ) , and @xmath87 is the product measure of the uniform distribution on @xmath43 .
henceforth we will take @xmath85 to be the basic probability space , unless otherwise stated . for a measure space @xmath88 ,
a set @xmath89 is a * @xmath90-null set * when @xmath91 . more generally ,
for @xmath92 $ ] , we say that @xmath93 is a * @xmath90-measure @xmath94 set * when @xmath95 .
a relation between functions on @xmath96 is said to hold * @xmath90-almost everywhere * ( abbreviated * @xmath90-a.e . * ) if it holds for all @xmath97 outside of a @xmath90-null set .
when @xmath90 is a probability measure , then we may instead say that the relation holds for * @xmath90-almost all * @xmath98 ( abbreviated * @xmath90-a.a . * ) .
we say that an event @xmath99 occurs * @xmath90-almost surely * ( abbreviated * @xmath90-a.s . * ) when @xmath100 .
in each case , we may drop the prefix @xmath90- when it is clear from context ( in particular , when it holds of @xmath87 ) .
we will use a @xmath101 font for random variables .
let @xmath31 be a computable metric space .
a * random variable in @xmath31 * is a function @xmath102 that is measurable with respect to the borel @xmath48-algebras of @xmath45 and @xmath31 . for a measurable subset @xmath103 , we let @xmath104 denote the inverse image @xmath105 = \ { \omega \in { \{0,1\}^\infty } : { \mathsf{x}}(\omega )
\in a \},\ ] ] and for @xmath49 we similarly define the event @xmath106 .
we will write @xmath5 for the * distribution of @xmath107 * , which is the measure on @xmath31 defined by @xmath108 .
[ compranvardef ] let @xmath31 be a computable metric space .
then a random variable @xmath107 in @xmath31 is a * computable random variable * when @xmath107 is computable on some @xmath87-measure one subset of @xmath45 . more generally , for a probability measure
@xmath90 and function @xmath28 between computable metric spaces , we say that @xmath28 is * @xmath90-almost computable * when it is computable on a @xmath90-measure one set .
( see @xcite for further development of the theory of almost computable functions . ) intuitively , @xmath107 is a computable random variable when there is a program that , given access to an oracle bit tape @xmath109 , outputs a representation of the point @xmath110 ( i.e. , enumerates a sequence @xmath50 in @xmath41 where @xmath111 for all @xmath35 ) , for all but a measure zero subset of bit tapes @xmath109 . even though the source of randomness is a sequence of discrete bits ,
there are computable random variables with _ continuous _ distributions , such as a uniform random variable ( gotten by subdividing the interval according to the random bittape ) or an i.i.d .- uniform sequence ( by splitting up the given element of @xmath45 into countably many disjoint subsequences and dovetailing the constructions ) .
( for details , see ( * ? ? ?
* ex . 3 , 4 ) .
) all of the standard distributions ( standard normal , uniform , geometric , exponential , etc . ) found in probability textbooks , as well the transformations of these distributions by computable ( or almost computable ) functions , are easily shown to be computable distributions .
it is crucial that we consider random variables that are computable only on a @xmath87-measure one subset of @xmath45 .
consider the following example : for a real @xmath112 $ ] , we say that a binary random variable @xmath113 is a * bernoulli*(@xmath54 ) random variable when @xmath114 .
there is a @xmath115 random variable that is computable on all of @xmath45 , given by the program that simply outputs the first bit of the input sequence .
likewise , when @xmath54 is * dyadic * ( i.e. , a rational with denominator a power of 2 ) , there is a @xmath116 random variable that is computable on all of @xmath45 .
however , this is not possible for any other choices of @xmath54 ( e.g. , @xmath117 ) .
[ bernoulliproof ] let @xmath112 $ ] be a nondyadic real .
every @xmath116 random variable @xmath118 is discontinuous , hence not computable on all of @xmath45 .
assume @xmath107 is continuous .
let @xmath119 and @xmath120 .
then @xmath121 , and so both are closed ( as well as open ) .
the compactness of @xmath45 implies that these closed subspaces are also compact , and so @xmath122 and @xmath123 can each be written as the finite disjoint union of clopen basis elements .
but each of these elements has dyadic measure , hence their sum can not be either @xmath54 or @xmath124 , contradicting the fact that @xmath125 .
on the other hand , for an arbitrary computable @xmath112 $ ] , consider the random variable @xmath126 given by @xmath127 if @xmath128 and @xmath129 otherwise .
this construction , due to @xcite , is a @xmath116 random variable and is computable on every point of @xmath45 other than a binary expansion of @xmath54 .
not only are these random variables computable , but they can be shown to be optimal in their use of input bits , via the classic analysis of rational - weight coins by knuth and yao @xcite . hence it is natural to admit as computable random variables those measurable functions that are computable only on a @xmath87-measure one subset of @xmath45 , as we have done . in this section , we introduce the class of computable probability measures and describe their connection with computable random variables .
let @xmath61 be a computable metric space , and let @xmath130 denote its borel sets .
we will denote by @xmath131 the set of ( borel ) measures on @xmath31 and by @xmath132 the subset which are probability measures .
consider the subset @xmath133 comprised of those probability measures that are concentrated on a finite subset of @xmath134 and where the measure of each atom is rational , i.e. , @xmath135 if and only if @xmath136 for some rationals @xmath137 such that @xmath138 and some points @xmath139 , where @xmath140 denotes the dirac delta on @xmath141 .
gcs ( * ? ? ?
* b.6.2 ) shows that @xmath142 is dense in the prokhorov metric @xmath143 given by @xmath144 where @xmath145 is the @xmath146-neighborhood of @xmath9 and @xmath147 is the open ball of radius @xmath146 about @xmath94 .
moreover , @xmath148 is a computable metric space .
( see also ( * ? ? ?
4.1.1 ) . )
we say that @xmath149 is a computable ( borel ) probability measure when @xmath90 is a computable point in @xmath132 as a computable metric space .
note that the measure @xmath87 on @xmath45 is a computable probability measure .
we can characterize the class of computable probability measures in terms of the computability of the measure of open sets .
[ lowerbounds ]
let @xmath61 be a computable metric space .
a probability measure @xmath149 is computable if and only if the measure @xmath150 of a c.e . open set @xmath151 is a c.e . real , uniformly in @xmath9 .
a * computable probability space * is a pair @xmath152 where @xmath31 is a computable metric space and @xmath90 is a computable probability measure on @xmath31 .
let @xmath152 be a computable probability space .
we know that the measure of a c.e .
open set @xmath9 is a c.e .
real , but is not in general a computable real . on the other hand ,
if @xmath9 is a decidable subset ( i.e. , @xmath153 is c.e .
open ) then @xmath154 a c.e .
real , and therefore , by the identity @xmath155 , we have that @xmath150 is a computable real . in connected spaces ,
the only decidable subsets are the empty set and the whole space .
however , there exists a useful surrogate when dealing with measure spaces .
let @xmath156 be a computable probability space .
a ( borel ) measurable subset @xmath151 is said to be * @xmath90-almost decidable * when there are two c.e .
open sets @xmath157 and @xmath158 such that @xmath159 and @xmath160 and @xmath161 the following is immediate .
[ almostdec - comp ] let @xmath152 be a computable probability space , and let @xmath9 be @xmath90-almost decidable .
then @xmath150 is a computable real . while we may not be able to compute the probability measure of ideal balls
, we can compute a new basis of ideal balls for which we can .
( see also bosserhoff ( * ? ? ?
* lem . 2.15 ) . ) [ almdecbasis ] let @xmath156 be a computable probability space , and let @xmath134 be the ideal points of @xmath31 with standard enumeration @xmath162 .
there is a computable sequence @xmath163 of reals , dense in the positive reals , such that the balls @xmath164 form a basis of @xmath90-almost decidable sets .
we now show that every c.e .
open set is the union of a computable sequence of almost decidable subsets .
[ comppair ] let @xmath156 be a computable probability space and let @xmath158 be a c.e .
then , uniformly in @xmath158 , we can compute a sequence of @xmath90-almost decidable sets @xmath165 such that , for each @xmath24 , @xmath166 , and @xmath167 .
let @xmath168 be a standard enumeration of the ideal balls of @xmath31 where @xmath169 , and let @xmath59 be a c.e .
set such that @xmath170 .
let @xmath164 form a basis of @xmath90-almost decidable sets , as shown to be computable by lemma [ almdecbasis ] .
consider the c.e .
set @xmath171 because @xmath172 is dense in @xmath31 and @xmath173 is dense in the positive reals we have for each @xmath174 that @xmath175 .
in particular this implies that the set @xmath176 is a c.e . set with @xmath177 .
let @xmath178 be a c.e .
enumeration of @xmath179 and let @xmath180 , which is almost decidable . by construction , for each @xmath24 , @xmath166 , and @xmath167 . using the notion of an almost decidable set
, we have the following characterization of computable measures .
[ almostcomplementary - to - compmeas ] let @xmath31 be a computable metric space and let @xmath149 be a probability measure on @xmath31
. then @xmath90 is computable if the measure @xmath150 of every @xmath90-almost decidable set @xmath9 is a computable real , uniformly in @xmath9 .
let @xmath158 be a c.e .
open set of @xmath31 . by theorem [ lowerbounds ]
, it suffices to show that @xmath181 is a c.e .
real , uniformly in @xmath158 . by lemma [ comppair ]
, we can compute a nested sequence @xmath165 of @xmath90-almost decidable sets whose union is @xmath158 . because @xmath158 is open ,
@xmath182 . by hypothesis , @xmath183 is a computable real for each @xmath24 , and so the supremum is a c.e .
real , as desired .
computable random variables have computable distributions .
[ rvtodist ] let @xmath107 be a computable random variable in a computable metric space @xmath31 .
then its distribution is a computable point in the computable metric space @xmath132 . on the other hand
, one can show that given a computable point @xmath90 in @xmath132 , one can construct an i.i.d.-@xmath90 sequence of computable random variables in @xmath31 .
informally , the _ conditional probability _ of an event @xmath184 given an event @xmath9 is the likelihood that the event @xmath184 occurs , given the knowledge that the event @xmath9 has occurred .
[ def : conditional - prob - set ] let @xmath31 be a measurable space and let @xmath185 be a probability measure on @xmath31 .
let @xmath186 be measurable sets , and suppose that @xmath187 .
then the * conditional probability of @xmath184 given @xmath9 * , written @xmath188 , is defined by @xmath189 note that for any fixed measurable set @xmath190 with @xmath187 , the function @xmath191 is a probability measure .
this notion of conditioning is well - defined precisely when @xmath192 , and so is insufficient for defining the conditional probability given the event that a _ continuous _ random variable takes a particular value , as such an event has measure zero
. we will often be interested in the case where @xmath184 and @xmath9 are measurable sets of the form @xmath193 and @xmath194 . in this case
, we define the abbreviation @xmath195 again , this is well - defined when @xmath196 . as a special case , when @xmath197 is an atom , we obtain the notation @xmath198 the modern formulation of conditional probability is due to kolmogorov @xcite , and gives a consistent solution to the problem of conditioning on the value of general ( and in particular , continuous ) random variables .
( see kallenberg ( * ? ? ?
* chp . 6 ) for a rigorous treatment . )
let @xmath107 and @xmath199 be random variables in measurable spaces @xmath31 and @xmath73 , respectively , and let @xmath200 be a measurable set .
then a measurable function @xmath201 $ ] from @xmath31 to @xmath202 $ ] is ( a version of ) the * conditional probability of @xmath203 given @xmath107 * when @xmath204\ , d{\mathbf{p}}_{{\mathsf{x}}}\ ] ] for all measurable sets @xmath190 .
@xmath201 $ ] is uniquely defined up to a @xmath5-null set ( i.e. , almost surely unique ) and so we may sensibly refer to _ the _ conditional probability when we mean a generic element of the equivalence class of versions .
[ rnremark ] it is also common to define a conditional probability given @xmath107 as a @xmath205-measurable random variable @xmath206 that is the radon - nikodym derivative of the measure @xmath207 with respect to @xmath87 , both viewed as measures on the @xmath48-algebra @xmath205 .
( see kallenberg ( * ? ? ?
* chp . 6 ) for such a construction . )
however there is a close relationship between these two definitions : in particular , there exists a measure function @xmath208 from @xmath31 to @xmath202 $ ] such that @xmath209 a.s .
we have taken @xmath208 to be our definition of the conditional probability as it is more natural to have a function from @xmath31 in the statistical setting .
( we take the same tact when defining conditional distributions . ) in propositions [ prevmainnegativeresult ] and [ condprobpropasrv ] , we demonstrate that both definitions of conditional probability admit noncomputable versions .
it is natural to want to consider not just individual conditional probabilities but the entire conditional distribution @xmath210 $ ] of @xmath199 given @xmath107 . in order to define conditional distributions ,
we first recall the notion of a probability kernel .
( for more details , see , e.g. , ( * ? ? ?
* ch . 3 , 6 ) .
) let @xmath31 and @xmath73 be measurable spaces .
a function @xmath211 $ ] is called a * probability kernel ( from @xmath31 to @xmath73 ) * when 1 . for every @xmath212 ,
the function @xmath213 is a probability measure on @xmath73 ; and 2 . for every @xmath214 ,
the function @xmath215 is measurable .
it can be shown that @xmath216 is a probability kernel from @xmath31 to @xmath73 if and only if @xmath217 is a measurable map from @xmath31 to @xmath218 ( * ? ? ?
1.40 ) .
[ conddef ] let @xmath107 and @xmath199 be random variables in measurable spaces @xmath31 and @xmath73 , respectively .
a probability kernel @xmath216 is called a * ( regular ) version of the conditional distribution @xmath219 $ ] * when it satisfies @xmath220 for all measurable sets @xmath151 and @xmath200 .
we will simply write @xmath221 $ ] in place of @xmath219 $ ] . let @xmath90 be a measure on a topological space @xmath31 with open sets @xmath222 .
then the * support of @xmath90 * , written @xmath223 , is defined to be the set of points @xmath224 such that all open neighborhoods of @xmath52 have positive measure , i.e. , @xmath225 given any two versions of a conditional distribution , they need only agree almost everywhere .
however , they will agree at points of continuity in the support : [ supportlemma ] let @xmath107 and @xmath199 be random variables in topological spaces @xmath31 and @xmath73 , respectively , and suppose that @xmath226 are versions of the conditional distribution @xmath221 $ ] .
let @xmath49 be a point of continuity of both of the maps @xmath227 for @xmath228 .
if @xmath229 , then @xmath230 . fix a measurable set @xmath231 and define @xmath232 .
we know that @xmath233 @xmath5-a.e . , and also that @xmath234 is continuous at @xmath52 .
assume , for the purpose of contradiction , that @xmath235 . by continuity
, there is an open neighborhood @xmath184 of @xmath52 , such that @xmath236 .
but @xmath229 , hence @xmath237 , contradicting @xmath238 @xmath5-a.e .
when conditioning on a discrete random variable , i.e. one whose image is a discrete set , it is well known that a version of the conditional distribution can be built by elementary conditioning with respect to single events .
[ discrete - ex ] let @xmath107 and @xmath199 be random variables on measurable spaces @xmath31 and @xmath73 , respectively .
suppose that @xmath107 is a discrete random variable with support @xmath66 , and let @xmath239 be an arbitrary probability measure on @xmath73 .
define the function @xmath240 $ ] by @xmath241 for all @xmath242 and @xmath243 for @xmath244 .
then @xmath216 is a version of the conditional distribution @xmath245 $ ] .
the function @xmath216 , given by @xmath246 for all @xmath242 and @xmath243 for @xmath244 , is well - defined because @xmath247 for all @xmath242 , and so the right hand side of equation is well - defined .
furthermore , @xmath248 and so @xmath216 is characterized by equation for @xmath5-almost all @xmath52 . finally , for all measurable sets @xmath151 and @xmath200 , we have @xmath249 and so @xmath216 is a version of the conditional distribution @xmath245 $ ] .
we begin by demonstrating the computability of elementary conditional probability given positive - measure events that are almost decidable .
we then return to the abstract setting and lay the foundations for the remainder of the paper .
[ condcontset ] let @xmath152 be a computable probability space and let @xmath9 be an almost decidable subset of @xmath31 satisfying @xmath187 . then @xmath191 is a computable probability measure . by corollary
[ almostcomplementary - to - compmeas ] , it suffices to show that @xmath250 is computable for an almost decidable set @xmath184 .
but then @xmath251 is almost decidable and so its measure , the numerator , is a computable real .
the denominator is likewise the measure of an almost decidable set , hence a computable real . finally , the ratio of two computable reals is computable .
conditioning in the abstract setting is more involved .
having defined the abstract notion of conditional probabilities and conditional distributions in section [ conddist ] , we now define notions of computability for these objects , starting with conditional distributions .
[ compkern ] let @xmath31 and @xmath73 be computable metric spaces and let @xmath252 $ ] be a probability kernel from @xmath31 to @xmath73 .
then we say that @xmath216 is a * computable ( probability ) kernel * when the map @xmath253 given by @xmath254 is a computable function .
similarly , we say that @xmath216 is computable on a subset @xmath255 when @xmath256 is computable on @xmath6
. the following correspondence will allow us to derive other characterizations of computable probability kernels .
the proof , however , will use that fact that given a sequence of ideal balls @xmath257 and an ideal ball @xmath258 we can semi - decide when @xmath259 ( uniformly in the indexes of the ideal balls ) .
henceforth , we will make the assumption that our computable metric spaces have the property .
this holds for all the specific spaces ( @xmath260,@xmath43 , @xmath45 , @xmath261 , etc . )
that we consider .
[ equiv - weak - top - and - metric - for - prokhorov ] let @xmath262 be a computable metric space . let @xmath263 be the collection of sets of the form @xmath264 where @xmath9 is a c.e
. open subset of @xmath73 and @xmath265 is a rational .
then the elements of @xmath263 are c.e .
open subsets of @xmath218 uniformly in @xmath9 and @xmath265 . note that @xmath263 is a subbasis for the weak topology induced by the prokhorov metric ( see ( * ? ? ?
3.2 ) ) .
let @xmath266 for a rational @xmath265 and c.e .
open subset @xmath9 .
we can write @xmath267 for a sequence ( computable uniformly in @xmath9 ) of ideal balls in @xmath73 with centers @xmath268 and rational radii @xmath269 .
. then @xmath271 and @xmath272 .
writing @xmath273 we have @xmath274 . in order to show that @xmath275 is c.e .
open uniformly in @xmath265 and @xmath9 , it suffices to show that @xmath276 is c.e .
open , uniformly in @xmath265 , @xmath277 and @xmath9 .
let @xmath142 be the ideal points of @xmath218 ( see section [ sec : compdists ] ) , let @xmath135 , and let @xmath72 be the finite set on which it concentrates .
gcs ( * ? ? ?
b.17 ) characterizes the ideal ball @xmath93 centered at @xmath239 @xmath278 for all subsets @xmath279 , where @xmath280 .
it is straightforward to show that @xmath281 if and only if @xmath282 , where @xmath283 note that @xmath284 is a decidable subset of @xmath72 ( uniformly in @xmath277 , @xmath9 , and @xmath93 ) and that @xmath285 is a rational and so we can decide whether @xmath281 , showing that @xmath276 is c.e .
open , uniformly in @xmath265 , @xmath277 , and @xmath9 .
recall that a lower semicomputable function from a computable metric space to @xmath202 $ ] is one for which the preimage of @xmath286 $ ] is c.e .
open , uniformly in rationals @xmath265 .
furthermore , we say that a function @xmath28 from a computable metric space @xmath31 to @xmath202 $ ] is _ lower semicomputable on @xmath287 _ when there is a uniformly computable sequence @xmath288 of c.e .
open sets such that @xmath289\bigr ] \cap d = u_q \cap d.\ ] ] we can also interpret a computable probability kernel @xmath216 as a computable map sending each c.e
. open set @xmath290 to a lower semicomputable function @xmath291 .
[ phikappa ] let @xmath31 and @xmath73 be computable metric spaces , let @xmath216 be a probability kernel from @xmath31 to @xmath73 , and let @xmath292 .
if @xmath256 is computable on @xmath6 then @xmath291 is lower semicomputable on @xmath6 uniformly in the c.e .
open set @xmath9 .
let @xmath293 be a rational , and let @xmath9 be a c.e.open set . define @xmath294 $ ] .
then @xmath295 = \phi_\kappa^{-1}[p ] , $ ] where @xmath296 this is an open set in the weak topology induced by the prokhorov metric ( see ( * ? ? ?
3.2 ) ) , and by lemma [ equiv - weak - top - and - metric - for - prokhorov ] , @xmath275 is c.e .
open . by the computability of @xmath256
, there is a c.e .
open set @xmath158 , uniformly computable in @xmath265 and @xmath9 such that @xmath297 \cap d = \phi_\kappa^{-1}[p ] \cap d = v \cap d,\ ] ] and so @xmath291 is lower semicomputable on @xmath6 , uniformly in @xmath9 .
in fact , when @xmath298 is a decidable set ( i.e. , when @xmath9 and @xmath299 are both c.e .
open ) , @xmath291 is a computable function .
[ decidable ] let @xmath31 and @xmath73 be computable metric spaces , let @xmath216 be a probability kernel from @xmath31 to @xmath73 computable on a subset @xmath292 , and let @xmath290 be a decidable set .
then @xmath300 $ ] is computable on @xmath6 . by lemma [ phikappa ] , @xmath291 and @xmath301 are lower semicomputable on @xmath6 . but
@xmath302 for all @xmath303 , and so @xmath291 is upper semicomputable , and therefore computable , on @xmath6 .
although a conditional distribution may have many different versions , their computability as probability kernels does not differ ( up to a change in domain by a null set ) .
let @xmath107 and @xmath199 be computable random variables on computable metric spaces @xmath31 and @xmath73 , respectively , and let @xmath216 be a version of a conditional distribution @xmath221 $ ] that is computable on some @xmath5-measure one set
. then any version of @xmath221 $ ] is also computable on some @xmath5-measure one set .
let @xmath216 be a version that is computable on a @xmath5-measure one set @xmath6 , and let @xmath304 be any other version .
then @xmath305 is a @xmath5-null set , and @xmath306 on @xmath307 .
hence @xmath304 is computable on the @xmath5-measure one set @xmath307 .
this observation motivates the following definition of computability for conditional distributions .
let @xmath107 and @xmath199 be computable random variables on computable metric spaces @xmath31 and @xmath73 , respectively , and let @xmath216 be a version of the conditional distribution @xmath245 $ ] .
we say that @xmath245 $ ] is computable when @xmath216 is computable on a @xmath5-measure one subset of @xmath31 .
note that this definition is analogous to our definition [ compranvardef ] of a _
computable random variable_. in fact , if @xmath216 is a version of a computable conditional distribution @xmath221 $ ] , then @xmath308 is a ( @xmath205-measurable ) computable random probability measure ( i.e. , a probability - measure - valued random variable ) .
intuitively , a conditional distribution is computable when for some ( and hence for any ) version @xmath216 there is a program that , given as input a representation of a point @xmath212 , outputs a representation of the measure @xmath309 for @xmath310 all inputs @xmath311 .
suppose that @xmath245 $ ] is computable , i.e. , there is a version @xmath216 for which the map @xmath256 is computable on some @xmath5-measure one set @xmath312 .
as noted in definition [ compkern ] , we will often abuse notation and say that @xmath216 is computable on @xmath313 .
the restriction of @xmath256 to @xmath313 is necessarily continuous ( under the subspace topology on @xmath313 ) .
we say that @xmath216 is * @xmath5-almost continuous * when the restriction of @xmath256 to some @xmath5-measure one set is continuous .
thus when @xmath245 $ ] is computable , there is some @xmath5-almost continuous version .
we will need the following lemma in the proof of lemma [ lem : effectiveversion ] , but we postpone the proof until section [ cds ] . [ denstokern ] let @xmath107 and @xmath199 be random variables on metric spaces @xmath31 and @xmath73 , respectively , and let @xmath66 . if a conditional density @xmath314 of @xmath107 given @xmath199 is continuous on @xmath315 , positive , and bounded , then there is a version of the conditional distribution @xmath245 $ ] that is continuous on @xmath72 . in particular , if @xmath72 is a @xmath5-measure one subset , then there is a @xmath5-almost continuous version .
we now define computability for conditional probability .
let @xmath107 and @xmath199 be computable random variables in computable metric spaces @xmath31 and @xmath73 , respectively , and let @xmath200 be a @xmath316-almost decidable set .
we say that the conditional probability @xmath201 $ ] is computable when it is computable ( when viewed as a function from @xmath31 to @xmath202 $ ] ) on a @xmath5-measure one set . in section [ sec :
discontinuous ] we describe a pair of computable random variables @xmath317 for which @xmath245 $ ] is not computable , by virtue of no version being @xmath5-almost continuous . in section [ sec : c.e .
neg ] we describe a pair of computable random variables @xmath317 for which there is a @xmath5-almost continuous version of @xmath245 $ ] , but still no version that is computable on a @xmath5-measure one set .
our study of the computability of conditional distributions begins with a description of the following roadblock : a conditional distribution need not have _ any _ version that is continuous or even almost continuous ( in the sense described in section [ ccd2 ] ) .
this rules out computability .
we will work with the standard effective presentations of the spaces @xmath46 , @xmath318 , @xmath43 , as well as product spaces thereof , as computable metric spaces .
for example , we will use @xmath46 under the euclidean metric , along with the ideal points @xmath42 under their standard enumeration .
recall that a random variable @xmath319 is a * bernoulli*(@xmath94 ) random variable , or equivalently , a * @xmath94-coin * , when @xmath320 .
we call a @xmath321-coin a * fair coin*. a random variable @xmath322 is * geometric * when it takes values in @xmath323 and satisfies @xmath324 a random variable that takes values in a discrete set is a * uniform * random variable when it assigns equal probability to each element .
a continuous random variable @xmath11 on the unit interval is * uniform * when the probability that it falls in the subinterval @xmath325 $ ] is @xmath326 .
it is easy to show that the distributions of these random variables are computable .
let @xmath319 , @xmath11 , and @xmath322 be independent computable random variables , where @xmath319 is a fair coin , @xmath11 is a uniform random variable on @xmath202 $ ] , and @xmath322 is a geometric random variable .
fix a computable enumeration @xmath327 of the rational numbers ( without repetition ) in @xmath328 , and consider the random variable @xmath329 it is easy to verify that @xmath107 is a computable random variable .
[ discontinuouscond ] every version of the conditional distribution @xmath330 $ ] is discontinuous everywhere on every @xmath5-measure one set . in particular , every version of the conditional probability @xmath331 $ ] is discontinuous everywhere on every @xmath5-measure one set . [ discontinuouscond - proof ] note that @xmath332 and , furthermore , @xmath333 therefore , any two versions of the conditional distribution @xmath334 $ ] must agree on _ all _ rationals in @xmath202 $ ] . in addition , because @xmath335 , i.e. , @xmath336 for all measurable sets @xmath337 $ ] , any two versions must agree on a lebesgue - measure one set of the irrationals in @xmath202 $ ] .
an elementary calculation shows that @xmath338 while @xmath339 therefore , all versions @xmath216 of @xmath340 $ ] satisfy , for @xmath5-almost all @xmath52 , @xmath341 where the right hand side , considered as a function of @xmath52 , is the so - called dirichlet function , a _ nowhere continuous _ function .
suppose some version of the conditional probability @xmath342 were continuous at a point @xmath343 on a @xmath5-measure one set @xmath72 .
then there would exist an open interval @xmath344 containing @xmath343 such that the image of @xmath345 contains 0 or 1 , but not both .
however , @xmath72 must contain all rationals in @xmath344 and almost all irrationals in @xmath344 .
furthermore , the image of every rational in @xmath345 is 1 , and the image of almost every irrational in @xmath345 is @xmath129 , a contradiction .
discontinuity is a fundamental obstacle to computability , but many conditional probabilities do admit continuous versions , and we can focus our attention on such settings , to rule out this objection . in particular , we might hope to be able to compute a conditional distribution of a pair of computable random variables when there is _ some _ version that is almost continuous or even continuous .
however we will show that even this is not possible in general .
in this section , we construct a pair of random variables @xmath346 that is computable , yet whose conditional distribution @xmath10 $ ] is not computable , despite the existence of a @xmath5-almost continuous version .
let @xmath347 denote the @xmath1th turing machine , under a standard enumeration , and let @xmath348 be the map given by @xmath349 if @xmath347 does not halt ( on input 0 ) and @xmath350 if @xmath347 halts ( on input 0 ) at the @xmath24th step . let @xmath351 be the indices of those turing machines that halt ( on input 0 ) .
we now use @xmath208 to define a pair @xmath352 of computable random variables such that @xmath353 is computable from the conditional distribution of @xmath322 given @xmath107 .
let @xmath322 be a computable geometric random variable , @xmath319 a computable @xmath117-coin , and @xmath11 and @xmath12 both computable uniform random variables on @xmath202 $ ] , all mutually independent .
let @xmath354 denote the greatest integer @xmath355 .
note that @xmath356 is uniformly distributed on @xmath357 . consider the derived random variables
@xmath358 for @xmath359 . almost surely , the limit @xmath360 exists and satisfies @xmath361 finally , we define @xmath362 the random variable @xmath107 is computable .
[ thmcomp ] because @xmath11 and @xmath12 are computable and a.s .
nondyadic , their ( a.s .
unique ) binary expansions @xmath363 and @xmath364 are themselves computable random variables in @xmath43 , uniformly in @xmath1 . for each @xmath365 , define the random variable @xmath366 by simulating @xmath367 for @xmath24 steps , we can decide whether @xmath368 is less than , equal to , or greater than @xmath24 . therefore the random variables @xmath369 are computable , uniformly in @xmath24 . we now show that , with probability one , @xmath369 is the binary expansion of @xmath107 , thus showing that @xmath107 is itself a computable random variable . there are two cases to consider : first , conditioned on @xmath370 , we have that @xmath371 for all @xmath372 .
in fact , @xmath373 when @xmath370 , and so the binary expansions match . condition on @xmath374 and let @xmath375 denote the computable random real whose binary expansion is @xmath376 .
we must then show that @xmath377 a.s .
note that @xmath378 and thus the binary expansions agree for the first @xmath277 digits .
finally , notice that @xmath379 , and so the next binary digit of @xmath380 is @xmath319 , followed by the binary expansion of @xmath11 , thus agreeing with @xmath375 for all @xmath372 . , where @xmath199 is uniformly distributed and @xmath381 .
regions that appear ( at low resolution ) to be uniform can suddenly be revealed ( at higher resolutions ) to be patterned .
deciding whether the pattern is in fact uniform ( or below the resolution of this printer / display ) is tantamount to solving the halting problem , but it is possible to sample from this distribution nonetheless .
note that this is not a plot of the density , but instead a plot where the darkness of each pixel is proportional to its measure .
, title="fig : " ] , where @xmath199 is uniformly distributed and @xmath381 .
regions that appear ( at low resolution ) to be uniform can suddenly be revealed ( at higher resolutions ) to be patterned .
deciding whether the pattern is in fact uniform ( or below the resolution of this printer / display ) is tantamount to solving the halting problem , but it is possible to sample from this distribution nonetheless . note that this is not a plot of the density , but instead a plot where the darkness of each pixel is proportional to its measure .
, title="fig : " ] we now show that @xmath8 $ ] is not computable , despite the existence of a @xmath5-almost continuous version of @xmath8 $ ] .
we begin by characterizing the conditional density of @xmath107 given @xmath322 .
[ densitylemma ] for each @xmath382 , the distribution of @xmath383 admits a density @xmath384 with respect to lebesgue measure on @xmath202 $ ] , where , for all @xmath385 and @xmath5-almost all @xmath52 , @xmath386 and @xmath387 .
we have @xmath388 a.s .
and so the constant function taking the value @xmath389 is the density of @xmath390 with respect to lebesgue measure on @xmath202 $ ] .
let @xmath391 . with probability one ,
the integer part of @xmath392 is @xmath393 while the fractional part is @xmath11 .
therefore , the distribution of @xmath394 ( and hence @xmath383 ) admits a piecewise constant density with respect to lebesgue measure .
in particular , @xmath395 almost surely and @xmath396 is independent of @xmath319 and uniformly distributed on @xmath397 .
therefore , @xmath398 for every @xmath399 .
it follows immediately that the density @xmath94 of @xmath394 with respect to lebesgue measure on @xmath400 $ ] is given by @xmath401 and so the density of @xmath383 is obtained by rescaling : @xmath402 .
as @xmath383 admits a density with respect to lebesgue measure on @xmath202 $ ] for all @xmath403 , it follows that the conditional distribution of @xmath107 given @xmath322 admits a conditional density @xmath404 ( with respect to lebesgue measure on @xmath202 $ ] ) given by @xmath405 each of these densities is positive , bounded , and continuous on the nondyadic reals , and so they can be combined to form a @xmath5-almost continuous version of the conditional distribution .
[ lem : effectiveversion ] there is a @xmath5-almost continuous version of @xmath10 $ ] . by bayes rule ( lemma [ bayesrule ] )
, the probability kernel @xmath216 given by @xmath406 is a version of the conditional distribution @xmath407 $ ] .
moreover , @xmath404 is positive , bounded and continuous on @xmath408 , where @xmath72 is the @xmath5-measure one set of nondyadic reals in the unit interval .
it follows from lemma [ denstokern ] that @xmath216 is @xmath5-almost continuous .
[ propcomp ] let @xmath216 be a version of @xmath409 $ ]
. for all @xmath410 and @xmath5-almost all @xmath52 , @xmath411 let @xmath412 . by and then , for @xmath5-almost all @xmath52 , @xmath413 by lemma [ lem : effectiveversion ] , we have @xmath387 and @xmath414 for all @xmath385 and @xmath5-almost all @xmath52 .
the result follows by considering all possible combinations of values for each regime of values for @xmath415 and @xmath416 . despite the computability of @xmath322 and @xmath107 , conditioning on @xmath107
leads to noncomputability .
[ prevmainnegativeresult ] for all @xmath24 , the conditional probability @xmath417 $ ] is not lower semicomputable ( hence not computable ) on any measurable subset @xmath418 $ ] where @xmath419 .
choose @xmath1 to be the index of some turing machine @xmath347 that halts ( on input 0 ) , i.e. , for which @xmath420 , let @xmath421 , let @xmath216 be a version of the conditional distribution , let @xmath422 be as in lemma [ propcomp ] , and let @xmath423 and @xmath424 be disjoint c.e .
open sets that contain @xmath425 and @xmath426 , respectively . by lemma [ propcomp ] ,
when @xmath427 , we have @xmath428 ) = 1 $ ] . likewise ,
when @xmath429 , we have @xmath430 ) = 1 $ ] .
suppose that for some @xmath24 that @xmath417 $ ] were lower semicomputable on some set @xmath72 satisfying @xmath419 . note that the numerator of equation is computable on the nondyadics , uniformly in @xmath431 , and the denominator is independent of @xmath184 .
therefore , because @xmath417 = \kappa(\cdot,\{k\})$ ] @xmath5-a.e . , it follows that @xmath432 $ ] is lower semicomputable on @xmath72 , _ uniformly _ in @xmath24 .
but then , uniformly in @xmath24 , we have that @xmath417)$ ] is computable on @xmath72 because @xmath433 is a decidable subset of @xmath261 and @xmath434 = 1- \sum_{j\neq k}{\mathbf{p}}[{\mathsf{n}}=j|{\mathsf{x } } ] \quad \text{a.s.}\ ] ] therefore , @xmath435 is computable on @xmath72 , uniformly in @xmath277 .
it follows that there exist c.e .
open sets @xmath436 and @xmath437 , uniformly in @xmath277 , such that @xmath438 \cap r = u_{<\infty } \cap r$ ] and thus @xmath439 \cap r)$ ] , and similarly for @xmath440 and @xmath437 .
if @xmath427 , then @xmath428 ) = 1 $ ] and thus @xmath441 \cap r ) = { \mathbf{p}}_{{\mathsf{x}}}(r ) > \frac 2 3.\ ] ] similarly , @xmath429 implies that @xmath442 .
therefore , at least one of @xmath443 holds . as @xmath444
, we have @xmath445 and so @xmath446 .
thus , @xmath447 and so we have that at most one ( and hence exactly one ) of _ ( i ) _ @xmath448 and @xmath427 or _
( ii ) _ @xmath442 and @xmath429 . but @xmath449 and @xmath450 are c.e .
reals , uniformly in @xmath277 , and so we can computably distinguish between cases _ ( i ) _ and _ ( ii ) _ , and thus decide whether or not @xmath427 , or equivalently , whether or not @xmath451 , uniformly in @xmath277 .
but @xmath353 is not computable and so we have a contradiction . the conditional distribution @xmath10 $ ] is not computable .
as @xmath452 is a decidable subset of @xmath261 , uniformly in @xmath1 , it follows from corollary [ decidable ] that the conditional probability @xmath453 $ ] is computable on a measure one set when the conditional distribution is computable . by proposition [ prevmainnegativeresult ] , the former is not computable , and so the latter is not computable .
note that these proofs show that not only is the conditional distribution @xmath8 $ ] noncomputable , but in fact it computes the halting set @xmath353 in the following sense .
although we have not defined the notion of an oracle that encodes @xmath8 $ ] , one could make this concept precise using , e.g. , infinite strings in the type-2 theory of effectivity ( tte ) .
however , despite not having a definition of computability _ from _ this distribution , we can easily relativize the notion of computability _ for _ the distribution .
in particular , an analysis of the above proof shows that if @xmath8 $ ] is @xmath9-computable for an oracle @xmath454 , then @xmath9 computes the halting set , i.e. , @xmath455 .
computable operations map computable points to computable points , and so we obtain the following consequence .
[ mainnegativeresult ] the operation @xmath456 $ ] of conditioning a pair of real - valued random variables is not computable , even when restricted to pairs for which there exists a @xmath5-almost continuous version of the conditional distribution .
conditional probabilities are often thought about as generic elements of @xmath457 equivalence classes ( i.e. , functions that are equivalent up to @xmath5-null sets . )
however , proposition [ prevmainnegativeresult ] also rules out the computability of conditional probabilities in the weaker sense of so - called @xmath457-computability .
let @xmath90 be a computable measure on @xmath202 $ ] .
then a function @xmath458 is said to be * @xmath459-computable * if there is a computable sequence of rational polynomials @xmath460 such that @xmath461 if we let @xmath462 then we can turn @xmath459 into a computable metric space by quotienting out by the equivalence relation @xmath463 and letting the dense @xmath464 consist of those equivalence classes containing a polynomial with rational coefficients .
a function @xmath28 is then @xmath459-computable if and only if its equivalence class is a computable point in @xmath459 considered as a computable metric space .
the following result is an easy consequence of the fact that @xmath459-computable functions are computable `` in probability '' .
for all @xmath465 , the conditional probability @xmath466 $ ] is not @xmath457-computable .
let @xmath465 . by proposition [ prevmainnegativeresult
] , the conditional probability @xmath466 $ ] is not computable on any borel set @xmath72 such that @xmath419 . on the other hand , a result by hoyrup and rojas ( * ?
3.1.1 ) implies that a function @xmath28 is @xmath457-computable only if @xmath28 is computable on some @xmath5-measure @xmath467 set , uniformly in @xmath21 .
therefore , @xmath466 $ ] is not @xmath457-computable .
in fact , this result can be strengthened to a statement about the noncomputability of each conditional probability @xmath466 $ ] when viewed as @xmath205-measurable random variables ( remark [ rnremark ] ) , or , equivalently , @xmath205-measurable radon - nikodym derivatives .
( see kallenberg ( * ? ? ?
* chp . 6 ) for a discussion of the relationship between conditional expectation , radon - nikodym derivatives and conditional probability . ) in this form , we tighten a result of hoyrup and rojas ( * ? ? ?
[ condprobpropasrv ] let @xmath465 , and then consider the conditional probability @xmath466 $ ] as a @xmath468-measurable radon - nikodym derivative @xmath469 $ ] of @xmath470 with respect to @xmath87 .
then @xmath471 is not lower semicomputable ( and hence not computable ) on any measurable subset @xmath472 where @xmath473 . in particular , @xmath471 is not @xmath474-computable
. let @xmath216 be a version of the conditional distribution @xmath475 $ ] , and , for @xmath476 , let @xmath471 be defined as above . in particular , note that , for all @xmath477 and @xmath87-almost all @xmath98 , @xmath478 in particular , this implies that , for all @xmath479 and @xmath87-almost all @xmath98 , @xmath480 using equations , , and , one can express @xmath471 as a ratio whose numerator is computable on a @xmath87-measure one set , uniformly in @xmath1 , and whose denominator is positive for @xmath87-almost all @xmath98 and independent of @xmath1 .
one can then show , along the same lines as proposition [ prevmainnegativeresult ] , that @xmath471 is not lower semicomputable on any set @xmath31 as defined above , for otherwise we could decide @xmath353 .
this then rules out the @xmath474-computability of @xmath471 .
it is natural to ask whether this construction can be extended to produce a pair of computable random variables whose conditional distribution is noncomputable but has an _ everywhere continuous _ version .
we provide such a strengthening in section [ sec : c.e .
neg cont ] . despite these results ,
many important questions remain : how badly noncomputable is conditioning , even restricted to these continuous settings ? in what restricted settings is conditioning _ computable _ ? in section [ positiveresults ] , we begin to address the latter of these .
as we saw in section [ sec : discontinuous ] , discontinuity poses a fundamental obstacle to the computability of conditional probabilities . as such
, it is natural to ask whether we can construct a pair of random variables @xmath481 that is computable and admits an _ everywhere _ continuous version of the conditional distribution @xmath482 $ ] , which is itself nonetheless not computable .
in fact , this is possible using a construction similar to that of @xmath7 in section [ sec : c.e .
in particular , if we think of the construction of the @xmath24th bit of @xmath107 as an iterative process , we see that there are two distinct stages . during the first stage , which occurs so long as @xmath483 ,
the bits of @xmath107 simply mimic those of the uniform random variable @xmath12 .
then during the second stage , once @xmath484 , the bits mimic that of @xmath485 .
our construction of @xmath486 will differ in the second stage , where the bits of @xmath486 will instead mimic those of a random variable @xmath487 specially designed to smooth out the rough edges caused by the biased coin @xmath319 , while still allowing us to encode the halting set .
in particular , @xmath487 will be absolutely continuous and will have an infinitely differentiable density .
we now make the construction precise .
let @xmath322 , @xmath11 , @xmath12 and @xmath319 be as in the first construction .
we begin by defining several random variables from which we will construct @xmath487 , and then @xmath486 .
[ c - infinity - step ] there is a random variable @xmath488 in @xmath202 $ ] with the following properties : 1 .
@xmath488 is computable .
2 . @xmath489 admits a density @xmath490 with respect to lebesgue measure ( on @xmath202)$ ] that is infinitely differentiable everywhere .
3 . @xmath491 and @xmath492 .
4 . @xmath493 , for all @xmath494 ( where @xmath495 and @xmath496 are the left and right derivatives respectively ) .
( see figure [ stepdensity ] for one such random variable . )
let @xmath488 be as in lemma [ c - infinity - step ] , and independent of all earlier random variables mentioned .
note that @xmath488 is almost surely nondyadic and so the @xmath21-th bit @xmath497 of @xmath488 is a computable random variable , uniformly in @xmath21 .
, for @xmath498 , and @xmath129 otherwise , a @xmath499 bump function whose derivatives at @xmath500 are all 0 .
( right ) a density @xmath501 for @xmath502 , of a random variable satisfying lemma [ c - infinity - step ] , where @xmath503 is the integral of the bump function . , title="fig:",height=190 ] , for @xmath498 , and @xmath129 otherwise , a @xmath499 bump function whose derivatives at @xmath500 are all 0 .
( right ) a density @xmath501 for @xmath502 , of a random variable satisfying lemma [ c - infinity - step ] , where @xmath503 is the integral of the bump function .
, title="fig:",height=190 ] let @xmath375 be a computable random variable , independent of all earlier random variables mentioned , and uniformly distributed on @xmath504 .
consider @xmath505 it is clear that @xmath487 is itself a computable random variable , and straightforward to show that * @xmath506 admits an infinitely differentiable density @xmath507 with respect to lebesgue measure on @xmath202 $ ] ; and * for all @xmath508 , we have @xmath509 .
( for a visualization of the density @xmath507 see figure [ smooth - dens ] . )
we say a real @xmath510 $ ] is * valid for @xmath506 * if @xmath511 .
in particular ,
when @xmath512 , then @xmath487 is valid for @xmath506 .
the following are then straightforward consequences of the construction of @xmath487 and the definition of valid points : * if @xmath52 is valid for @xmath506 then @xmath513 . *
the @xmath506-measure of points valid for @xmath506 is @xmath514 .
next we define , for every @xmath515 , the random variables @xmath516 mimicking the construction of @xmath383 .
specifically , for @xmath515 , define @xmath517 and let @xmath518 a.s .
then the @xmath1th bit of @xmath516 is @xmath519 for @xmath520 , we say that @xmath510 $ ] is * valid for @xmath521 * if the fractional part of @xmath522 is valid for @xmath506 , and we say that @xmath52 is * valid for @xmath523 * for all @xmath52 .
let @xmath524 be the collection of @xmath52 valid for @xmath525 , and let @xmath526 $ ] .
it is straightforward to verify that @xmath527 for all @xmath520 , and @xmath389 for @xmath528 .
it is also straightforward to show from _ (
i ) _ and _ ( ii ) _ above that @xmath525 admits an infinitely differentiable density @xmath529 with respect to lebesgue measure on @xmath202 $ ] .
, when constructed from @xmath488 as given in figure [ stepdensity ] . , title="fig:",height=190 ]
+ to complete the construction , we define @xmath530 .
the following results are analogous to those in the almost continuous construction : the random variable @xmath486 is computable .
there is an everywhere continuous version of @xmath531 $ ] . by construction ,
the conditional density of @xmath486 is everywhere continuous , bounded , and positive .
the result follows from lemma [ denstokern ] for @xmath532 $ ] .
[ taucont ] let @xmath216 be a version of the conditional distribution @xmath482 $ ] . for all @xmath533 and @xmath534-almost all @xmath52 ,
if @xmath52 is valid for @xmath535 and @xmath536 then @xmath411 for all @xmath24 , the conditional probability @xmath537 $ ] is not lower semicomputable ( hence not computable ) on any measurable subset @xmath418 $ ] where @xmath538 . choose @xmath1 to be the index of a turing machine @xmath347 that does not halt ( on input 0 ) so that @xmath349 , let @xmath421 , let @xmath216 be a version of the conditional distribution , let @xmath422 be defined as in lemma [ taucont ] , and let @xmath423 and @xmath440 be disjoint c.e .
open sets containing the points @xmath539 and @xmath540 , respectively .
notice that all @xmath541 $ ] are valid for @xmath542 and so @xmath543 .
therefore , by lemma [ taucont ] , when @xmath427 , we have @xmath544 ) \ge { \mathbf{p}}_{{\mathsf{z}}}(a_{h(m ) } ) \ge \frac { 9}{16}$ ] . on the other hand , when @xmath429 , we have @xmath545 )
= { \mathbf{p}}_{{\mathsf{z}}}(a_\infty ) = 1 $ ] .
the remainder of the proof is similar to that of proposition [ prevmainnegativeresult ] , replacing @xmath107 with @xmath486 , replacing @xmath546 with @xmath547 , and noting , e.g. , that @xmath544 \cap r ) > \frac 9 { 16 } - \frac 1 { 24}$ ] when @xmath427 . in particular , were @xmath548 $ ] lower semicomputable on @xmath72 , we could computably distinguish whether or not @xmath549 , uniformly in @xmath277 , which is a contradiction .
as before , it follows immediately that @xmath482 $ ] is not computable .
it is possible to carry on the same development , showing the non - computability of the conditional probabilities as elements in @xmath457 and @xmath474 for simplicity , we state the following strengthening of theorem [ mainnegativeresult ] .
[ newmainnegativeresult ] let @xmath107 and @xmath199 be computable real - valued random variables .
then operation @xmath550 $ ] of conditioning a pair of real - valued random variables is not computable , even when restricted to pairs for which there exists an everywhere continuous version of the conditional distribution .
we now consider situations in which we can compute conditional distributions .
probabilistic methods have been widely successful in many settings , and so it is important to understand situations in which conditional inference is possible .
we begin with the setting of discrete random variables .
a very common situation is that in which we condition with respect to a random variable taking values in a discrete set .
as we will see , conditioning is always possible in this setting , as it reduces to the elementary notion of conditional probability with respect to single events .
[ discreteconditioning ] let @xmath107 and @xmath199 be computable random variables in computable metric spaces @xmath31 and @xmath73 , respectively , where @xmath31 is a finite or countable and discrete set .
then the conditional distribution @xmath245 $ ] is computable , uniformly in @xmath107 and @xmath199 .
let @xmath216 be a version of the conditional distribution @xmath245 $ ] , and define @xmath551 .
we have @xmath552 , and so , by lemma [ discrete - ex ] , @xmath216 is characterized by the equations @xmath553 let @xmath200 be an @xmath316-almost decidable . because @xmath554 is decidable , @xmath555 is @xmath556-almost decidable , and
so the numerator is a computable real , uniformly in @xmath557 and @xmath184 .
taking @xmath558 , we see that the denominator , and hence the ratio is also computable , uniformly in @xmath557 and @xmath184 .
thus , by corollary [ almostcomplementary - to - compmeas ] , @xmath559 is computable on a @xmath5-measure one set .
note that one could provide an alternative proof using the `` rejection sampling '' method , which is used to define the semantics of conditioning on discrete random variables in several probabilistic programming languages .
this could , for example , be formalized by using least fixed points or unbounded search .
a related situation is one where a computable random variable @xmath107 concentrates on a finite or countable subset of a computable metric space . in this case ,
equation is well - defined and characterizes every version of the conditional distribution , but the event @xmath560 , for @xmath229 , is not decidable .
however , if the countable subset is discrete , then each such event is almost decidable . in order to characterize the computability of conditioning in this setting ,
we first define a notion of computability for discrete subsets of computable metric spaces .
[ compdisc2 ] let @xmath31 be a computable metric space .
we say that a ( finite or countably infinite ) subset @xmath292 is a * computably discrete subset * when there exists a function @xmath561 that is computable and injective on @xmath6 .
we call @xmath28 the * witness * to the computable discreteness of @xmath6 .
the following result is an easy consequence of the definition of computably discrete subsets and the computability of conditional distributions given discrete observations .
[ compdiscsubsetcond ] let @xmath107 and @xmath199 be computable random variables in computable metric spaces @xmath31 and @xmath73 , respectively , let @xmath292 be a computable discrete subset with witness @xmath28 , and assume that @xmath562 a.s .
then the conditional distribution @xmath245 $ ] is computable , uniformly in @xmath107 , @xmath199 and ( the witness for ) @xmath28 .
let @xmath216 be a version of the conditional distribution of @xmath245 $ ] and let @xmath563 be a version of the conditional distribution of @xmath564 $ ] .
then , for @xmath565 , @xmath566 and these equations completely characterize @xmath216 as @xmath567 .
note that @xmath568 is a computable random variable in @xmath261 , and so by lemma [ discreteconditioning ] , @xmath563 is computable on @xmath569 , and so @xmath216 is computable on @xmath6 .
the most common way to calculate conditional probabilities is to use bayes rule , which requires the existence of a conditional density .
( within statistics , a probability model is said to be _ dominated _ when there is a conditional density . )
we first recall some basic definitions .
let @xmath570 be a measure space and let @xmath571 be an @xmath239-integrable function .
then the function @xmath90 on @xmath572 given by @xmath573 for @xmath574 is a ( finite ) measure on @xmath575 and @xmath28 is called a * density of @xmath90 ( with respect to @xmath239)*. note that @xmath234 is a density of @xmath90 with respect to @xmath239 if and only if @xmath576 @xmath239-a.e .
let @xmath107 and @xmath199 be random variables in ( complete separable ) metric spaces , let @xmath577 be a version of @xmath578 $ ] , and assume that there is a measure @xmath579 and measurable function @xmath580 such that @xmath581 is a density of @xmath582 with respect to @xmath239 for @xmath316-almost all @xmath343 .
that is , @xmath583 for measurable sets @xmath151 and @xmath316-almost all @xmath343
. then @xmath314 is called a * conditional density of @xmath107 given @xmath199 ( with respect to @xmath239)*. common parametric families of distributions ( e.g. , exponential families like gaussian , gamma , etc . )
admit conditional densities , and in these cases , the well - known bayes rule gives a formula for expressing the conditional distribution .
we give a proof of this classic result for completeness .
[ bayesrule ] let @xmath107 and @xmath199 be random variables as in definition [ conddef ] , let @xmath577 be a version of the conditional distribution @xmath578 $ ] , and assume that there exists a _ conditional density _ @xmath584 .
then a function @xmath211 $ ] satisfying @xmath585 for those points @xmath52 for which the denominator is positive and finite , is a version of the conditional distribution @xmath586 $ ] .
by definition [ conddef ] and fubini s theorem , for borel sets @xmath151 and @xmath200 , we have that @xmath587 taking @xmath588 , we have @xmath589 because @xmath590 , this implies that the set of points @xmath52 for which the denominator of is infinite has @xmath239-measure zero , and thus @xmath5-measure zero . taking @xmath9 to be the set of points @xmath52 for which the denominator is zero , we see that @xmath591 .
it follows that characterizes @xmath216 up to a @xmath5-null set . also by
, we see that the denominator is a density of @xmath5 with respect to @xmath239 , and so we have @xmath592 finally , by the definition of @xmath593 , equation , and the fact that the denominator is positive and finite for @xmath5-almost all @xmath52 , we see that @xmath593 is a version of the conditional distribution @xmath221 $ ] .
comparing bayes rule to the definition of conditional density , we see that any conditional density of @xmath199 given @xmath107 ( with respect to @xmath316 ) satisfies @xmath594 for @xmath595-almost all @xmath596
. we can now give the proof of lemma [ denstokern ] .
let @xmath597 be given by , and let @xmath200 be an open set . by hypothesis , the map @xmath598 given by @xmath599 is continuous on @xmath72 , while the indicator function @xmath600 is lower semicontinuous .
integration of a lower semicontinuous function with respect to a probability measure is a lower semicontinuous operation , and so the map @xmath601 is lower semicontinuous on @xmath72 .
note that for every @xmath76 , the function @xmath602 is positive and bounded by hypothesis .
integration of a bounded continuous function with respect to a probability measure is a continuous operation , and so the map @xmath603 is positive and continuous on @xmath72 .
therefore the ratio in is a lower semicontinuous function of @xmath242 for fixed @xmath184 , completing the proof . using the following well - known result about integration of computable functions
, we can study when the conditional distribution characterized by bayes rule is computable .
[ intcomp ] let @xmath31 be a computable metric space , and @xmath90 a computable probability measure on @xmath31 .
let @xmath604 be a bounded computable function .
then @xmath605 is a computable real , uniformly in @xmath28 .
[ independentdensity ] let @xmath11 , @xmath12 , and @xmath199 be computable random variables ( in computable metric spaces ) , where @xmath199 is independent of @xmath12 given @xmath11 .
assume that there exists a conditional density @xmath606 of @xmath199 given @xmath11 ( with respect to @xmath239 ) that is bounded and computable .
then the conditional distribution @xmath607 $ ] is computable .
let @xmath608 .
then @xmath609 is the conditional density of @xmath199 given @xmath107 ( with respect to @xmath239 ) .
therefore , the computability of the integrand and the existence of a bound imply , by proposition [ intcomp ] , that @xmath607 $ ] is computable . as an immediate consequence of corollary [ independentdensity
] , we obtain the computability of the following common situation in probabilistic modeling : where the observed random variable has been corrupted by independent absolutely continuous noise .
[ indnoise ] let @xmath11 be a computable random variable in a computable metric space and let @xmath12 and @xmath13 be computable random variables in @xmath46 .
define @xmath610 .
if @xmath14 is absolutely continuous with a bounded computable density @xmath15 and @xmath13 is independent of @xmath11 and @xmath12 then the conditional distribution @xmath611 $ ] is computable .
we have that @xmath612 is the conditional density of @xmath199 given @xmath11 ( with respect to lebesgue measure ) .
the result then follows from corollary [ independentdensity ] .
pour - el and richards ( * ? ? ?
* ch . 1 ,
2 ) show that a twice continuously differentiable computable function has a computable derivative ( despite the fact that myhill @xcite exhibits a computable function @xmath202\to{\ensuremath{\mathbb{r}}}$ ] whose derivative is continuous , but not computable ) .
therefore , noise with a sufficiently smooth distribution has a computable density , and by corollary [ indnoise ] , a computable random variable corrupted by such noise still admits a computable conditional distribution .
furthermore , corollary [ indnoise ] implies that noiseless observations can not always be computably approximated by noisy ones .
for example , even though an observation corrupted with zero mean gaussian noise with standard deviation @xmath48 may recover the original condition as @xmath613 , by our main noncomputability result ( theorem [ mainnegativeresult ] ) one can not , in general , compute how small @xmath48 must be in order to bound the error introduced by noise .
this result is also analogous to a classical theorem of information theory .
hartley @xcite and shannon @xcite show that the capacity of a continuous real - valued channel without noise is infinite , yet the addition of , e.g. , gaussian noise with @xmath614 variance causes the channel capacity to drop to a finite amount .
the gaussian noise prevents all but a finite amount of information from being encoded in the bits of the real number .
similarly , the amount of information in a continuous observation is too much in general for a computer to be able to update a probabilistic model .
however , the addition of noise , as above , is sufficient for making conditioning possible on a computer .
the computability of conditioning with noise , coupled with the noncomputability of conditioning in general , has significant implications for our ability to recover a signal when noise is added , and suggests several interesting questions .
for example , suppose we have a uniformly computable sequence of noise @xmath615 with absolutely continuous , uniformly computable densities such that the magnitude of the densities goes to @xmath129 in some sufficiently nice way , and consider @xmath616 .
such a situation could arise , e.g. , when we have a signal with noise but some way to reduce the noise over time .
when there is a continuous version of @xmath617 $ ] , we have @xmath618 = { \mathbf{p}}[({\mathsf{u } } , { \mathsf{v}})|{\mathsf{y } } ] \quad \text{a.s.}\ ] ] however , we know that the right side is , in general , noncomputable , despite the fact that each term in the limit on the left side is computable .
this raises several questions , such as : what do bounds on how fast the sequence @xmath619\}_{n\in{{\mathbb{n}}}}$ ] converges to @xmath617 $ ] tell us about the computability of @xmath617 $ ] ?
what conditions on the relationship between @xmath11 and the sequence @xmath615 will allow us to recover information about @xmath620 $ ] from individual distributions @xmath621 $ ] ?
freer and roy @xcite show how to compute conditional distributions in the setting of _ exchangeable sequences_. a classic result by de finetti shows that exchangeable sequences of random variables are in fact conditionally i.i.d .
sequences , conditioned on a random measure , often called the _ directing random measure_. freer and roy describe how to transform an algorithm for sampling an _ exchangeable sequence _ into a rule for computing the posterior distribution of the directing random measure given observations .
the result is a corollary of a computable version of de finetti s theorem @xcite , and covers a wide range of common scenarios in nonparametric bayesian statistics ( often where no conditional density exists ) .
a preliminary version of this article appeared as `` noncomputable conditional distributions '' in _ proceedings of the 26th annual ieee symposium on logic in computer science _ , 107116 @xcite .
cef was partially supported by nsf grants dms-0901020 and dms-0800198 .
his work on this publication was made possible through the support of a grant from the john templeton foundation .
the opinions expressed in this publication are those of the authors and do not necessarily reflect the views of the john templeton foundation .
the authors would like to thank jeremy avigad , henry cohn , leslie kaelbling , vikash mansinghka , hartley rogers , michael sipser , and joshua tenenbaum for helpful discussions , bill thurston for a useful comment regarding lemma [ c - infinity - step ] , martn escard for a helpful discussion regarding computable random variables , and quinn culver , noah goodman , leslie kaelbling , bjrn kjos - hanssen , vikash mansinghka , timothy odonnell , and geoff patterson for comments on earlier drafts .
a. edalat , _ the scott topology induces the weak topology _ , proc . of the 11th ann .
ieee symp . on logic in comput .
( new brunswick , nj , 1996 ) , ieee comput .
press , los alamitos , ca , 1996 , pp .
372381 .
c. e. freer and d. m. roy , _ computable exchangeable sequences have computable de finetti measures _ , mathematical theory and computational practice ( cie 2009 ) , proc . of the 5th conf .
on computability in europe ( k. ambos - spies , b. lwe , and w. merkle , eds . ) , lecture notes in comput .
5635 , springer , 2009 , pp .
218231 .
, _ posterior distributions are computable from predictive distributions _ , proc . of the 13th int .
conf . on artificial intelligence and statistics
( aistats 2010 ) ( y. w. teh and m. titterington , eds . ) , jmlr : w&cp 9 , 2010 , pp .
233240 .
m. hoyrup and c. rojas , _ applications of effective probability theory to martin - lf randomness _ , automata , languages and programming ( s. albers , a. marchetti - spaccamela , y. matias , s. nikoletseas , and w. thomas , eds . ) , lecture notes in comput .
sci , vol .
5555 , springer , 2009 , pp .
549561 .
m. hoyrup , c. rojas , and k. weihrauch , _ computability of the radon - nikodym derivative _ , models of computation in context , proc . of the 7th conf .
on computability in europe ( cie 2011 ) ( b. lwe , d. normann , i. n. soskov , and a. a. soskova , eds . ) , lecture notes in comput .
6735 , springer , 2011 , pp .
132141 .
d. e. knuth and a. c. yao , _ the complexity of nonuniform random number generation _ ,
algorithms and complexity ( proc .
carnegie - mellon univ . ,
pittsburgh , pa . , 1976 ) , academic press , new york , 1976 , pp .
357428 .
, _ admissible representations of probability measures _ , proceedings of the third international conference on computability and complexity in analysis ( cca 2006 ) , electron .
notes theor .
167 , elsevier sci .
b. v. , amsterdam , 2007 , pp .
6178 ( electronic ) .
a. k. zvonkin and l. a. levin , _ the complexity of finite objects and the basing of the concepts of information and randomness on the theory of algorithms _ , uspehi mat .
nauk * 25 * ( 1970 ) , no . 6 ( 156 ) , 85127 . | as inductive inference and machine learning methods in computer science see continued success , researchers are aiming to describe even more complex probabilistic models and inference algorithms .
what are the limits of mechanizing probabilistic inference ?
we investigate the computability of conditional probability , a fundamental notion in probability theory and a cornerstone of bayesian statistics , and show that there are computable joint distributions with noncomputable conditional distributions , ruling out the prospect of general inference algorithms , even inefficient ones .
specifically , we construct a pair of computable random variables in the unit interval such that the conditional distribution of the first variable given the second encodes the halting problem . nevertheless , probabilistic inference is possible in many common modeling settings , and we prove several results giving broadly applicable conditions under which conditional distributions are computable .
in particular , conditional distributions become computable when measurements are corrupted by independent computable noise with a sufficiently smooth density . [ multiblock footnote omitted ] [ multiblock footnote omitted ] [ multiblock footnote omitted ] |
the periodic anderson model ( pam ) is one of the basic models largely used in the study of strongly correlated systems whose properties can be described at the level of two effective bands , like heavy - fermion systems @xcite , intermediate - valence compounds @xcite , or even high critical temperature superconductors @xcite .
the model contains a free @xmath4 band hybridized with a correlated system of @xmath2 electrons for which the one - site coulomb repulsion in the form of the hubbard interaction is locally present .
seen from the theoretical side , pam has the peculiarity that even its one dimensional hamiltonian is sufficiently complicated to not allow the knowledge of its exact solutions even in 1d . as a consequence ,
taking into account that the exact description possibilities increase in difficulty with the increase of the dimensionality of the system in the physical region @xmath5 , the physics provided by pam is almost exclusively interpreted based on approximations .
this situation enhance the difficulty of a good quality theoretical analysis , since exact bench - marks in testing the approximations or numerical simulations are almost completely missing .
because of this fact , even the starting point of the theoretical description , the knowledge of the ground - state is poorly developing . given by this , efforts have been made for the construction of exact ground - states at least in restricted regions of the parameter space . in this frame ,
based on the observation that the infinitely repulsive case in relative terms is easier to treat , the first exact ground - states have been deduced at @xmath6 , in restricted regions of the parameter space , based on a work of several years @xcite . the first exact ground - states at finite value of the interaction have been published recently in 1d @xcite , and 2d @xcite , respectively .
these ground - states emerge on continuous but restricted regions of the @xmath7 phase diagram of the system which extend from the low @xmath1 limit up to the high @xmath1 limit as well .
these solutions have been obtained by a decomposition of the hamiltonian in positive semidefinite operators ( pso ) as follows ( i ) the interaction term has been transformed into a pso requiring at least one @xmath2 electron on every lattice site @xcite , and ( ii ) the remaining parts of the interaction term together with the one - particle components of the hamiltonian have been transformed in pso based on cell operators , using bonds in 1d @xcite or elementary plaquette operators in 2d @xcite .
two type of ground - states have been obtained in this manner : localized and itinerant once .
the itinerant solution was found to emerge for imaginary hybridization matrix elements ( @xmath8 ) , while the localized solution for real @xmath8 .
besides , the itinerant solution has been obtained only in the presence of anisotropic or distorted unit cells .
the cell operators used in the transformation of the hamiltonian had always the extension of an unit cell ( elementary plaquette in 2d ) , the @xmath2-creation operators being , , uniformly considered , acting on all lattice sites of the elementary plaquette .
we must note , that all exact ground - states deduced up today for pam at finite value of the interaction @xmath1 , based on the cell operators mentioned above , require the presence of the direct hopping @xmath9 as well in the correlated @xmath2 band of the hamiltonian , since the products of these cell operators leading to pso generate always @xmath9 contributions .
the presence of the direct @xmath2 hopping can be argued based on experimental data in the case of @xmath10 @xcite , or some heavy - fermion compounds @xcite , but it remains in fact an extension term to the generic pam hamiltonian which does not contain such type of contributions . particularly for 2d , since for the @xmath11 limit the results deduced in refs.[@xcite ] are no more valid , the case of the generic pam hamiltonian treated in exact terms remains still a completely open problem .
driven by the challenge to obtain exact solutions for the generic pam in 2d , the first questions which have to be clarified in these conditions are the following : can we consider the ground - states deduced for pam at finite @xmath1 in the presence of direct @xmath2 hopping in the hamiltonian also potential ground - states for the generic pam at finite @xmath1 ( and @xmath11 ) ?
the unit cell distortions and imaginary hybridization matrix elements are essential for the emergence of itinerant phases ?
this questions are important , since are connected to main problems related to pam , for example the localized vs. itinerant behaviour of @xmath2-electrons , and of particles in general in the system @xcite , or the interpretation of the pam behaviour based exclusively on the local @xmath2-moment and its compensation ( analyses made based on kondo physics ) . starting from this background , in this paper we are reporting the first exact ground - states for generic pam in 2d at finite value of the interaction . as a consequence
, the deduction is made without the presence of the direct hopping terms for @xmath2 electrons in the hamiltonian of the model ( @xmath12 . the obtained exact ground - state is itinerant , emerges for real values of the hamiltonian parameters ( including the hybridization matrix elements as well ) , is not necessarily connected to distorted unit cells and presents clear similarities in its physical properties with the itinerant solution obtained in the presence of direct @xmath2 hopping terms in @xmath13 @xcite .
since the momentum distribution function is continuous without any non - regularities in its derivatives of any order , the presented ground - state is a non - fermi liquid state in the normal phase , possesses a large spin degeneracy , and globally is paramagnetic . in order to deduce the reported results ,
major developments have been applied in comparison with the plaquette operator method used in ref.(@xcite ) and described in detail in ref.(@xcite ) , which has been generalised to a block operator procedure .
the introduced block is qualitatively different from the elementary plaquette previously used since a ) it has an extension greater than an unit cell and b ) it contains @xmath2 creation operators acting only on one unique central site of the block .
this choice allows us to represent the pam hamiltonian in term of pso block products even in the absence of the direct @xmath2 hopping terms and to obtain non - localized ground - states in the presence of real hybridization matrix elements and absence of lattice distortions .
all these are not possible to obtain in the frame of refs.[@xcite ] .
the consequences of the presented results are multiple .
( i ) at the level of exact solutions in 2d , the difference between the case @xmath9 and @xmath14 in @xmath13 , ( at least at the level of known exact ground - states ) seems to not be extremely significant .
physically this can be understood based on the hybridization term which allows the movement of @xmath2-electrons even if direct hopping in the correlated band is not present in @xmath13 , and , in fact for the behaviour of the system ( and also the ground - state ) , not the bare bands are important but the diagonalized once . for the presented case the @xmath15 value seems only to shift the position in the parameter space of the emerging phase in comparison with the @xmath11 situation , without changing essentially its physical properties .
( ii ) the fact that similar phases are obtained in exact terms for @xmath9 and @xmath11 as well shows that pam can provide a behaviour different from kondo physics .
the statement is underlined as well by the absence of the exponential contributions characteristic for a kondo type of behaviour in the exactly deduced ground - state energy values ( see for example ref .
this possibility in fact exceeds the vicinity of a metal - insulator transition mentioned from this point of view in ref.(@xcite ) since the presented phase diagram region extends continuously from the low @xmath1 limit to the high @xmath1 limit as well , up to @xmath16 .
( iii ) combining the here obtained result with the results previously deduced , we observe that the delocalization of the electrons in pam can emerge simply because in this manner the system decreases its ground - state energy . there are situations ( depending on the parameters of the starting hamiltonian)@xcite when long - range density - density correlations are developing creating a localized phase , maintaining localized as well the @xmath2 electrons within the system . when the delocalization occurs , the long - range density - density correlation is lost .
( iv ) we learn that using different type of blocks in constructing block operators in the process of the decomposition of the hamiltonian into products of positive semidefinite terms , different classes of hamiltonians can be analysed .
the block form , its extension , the uniform or non - uniform nature of the operator action inside the block , the type of dependence of the numerical coefficients entering in the block operator on the physical parameters ( like spin ) , all are important in this description process . based on this degree of freedom
, the method can be applied as well in the absence of direct @xmath2-hopping terms as well .
( v ) it becomes clear that in conditions in which the hamiltonian remains hermitic , the real or imaginary ( or even complex ) hybridization matrix elements could provide similar physical consequences .
the nature of these matrix elements is given by the symmetry @xcite , and in fact imaginary hybridization matrix elements has been used previously as well ( see for example @xcite ) . ( vi )
the non - fermi liquid properties present in rigorous terms in the pam phase diagram , at least in 2d are not essentially connected to imaginary hybridization matrix elements , nor direct @xmath2 hopping in the hamiltonian , nor the presence of distorted unit cells .
the remaining part of the paper has been constructed as follows .
section ii . presents the model and its description with block operators , sect.iii . describes the deduced exact ground - state
, sect.iv presents the conclusions , and appendix a. containing mathematical details closes the presentation .
we start with a generic pam hamiltonian taken for a 2d lattice in the form @xmath17 where , the first term gives the kinetic energy for @xmath4-electrons , the second term is the on - site @xmath2-electron energy , the third term represents the hybridization , and the interaction term @xmath18 describes the on - site hubbard interaction written for @xmath2 electrons @xmath19 , @xmath20 , @xmath21 being considered during this paper . as it can be seen , direct @xmath2-band is not present in the starting hamiltonian . for technical reasons ,
we transcribe @xmath13 in @xmath22 space , using for the operators @xmath23 the fourier sum @xmath24 . for this
, we take into consideration at the level of the hybridization term , local ( @xmath25 ) and non - local ( @xmath26 ) nearest - neighbour contributions as well .
the @xmath4-electron dispersion is taken into account including contributions up to next nearest - neighbour hoppings , which is not unusual in the case of the study of real materials @xcite .
the kinetic energy of @xmath4-electrons becomes @xmath27 \ : , \label{h2}\end{aligned}\ ] ] where @xmath28 represent the versors of the unit cell ( see fig.1 ) , and the dispersion relation for @xmath4 electrons presented in eq.([h1 ] ) becomes @xmath29 for the hybridization term we consider @xmath30 \ : , \label{h4}\end{aligned}\ ] ] from where , the @xmath31 hybridization matrix element from eq.([h1 ] ) can be expressed as @xmath32 denoting by @xmath33 the on - site @xmath2-electron energy and using eqs.([h2]-[h5 ] ) , for the starting hamiltonian presented in eq.([h1 ] ) we find @xmath34 the interaction term during this paper is exactly transformed in the form @xcite @xmath35 where , the positive semidefinite operator @xmath36 defined by eq.([h7 ] ) requires for its lowest zero eigenvalue at least one @xmath2-electron on every lattice site . as will be clarified further on , the representation presented in eq.([h7 ] ) is a key feature from the point of view of the interaction term in the deduction of exact ground - states at @xmath21 presented here .
let us introduce connected to every site @xmath37 of the lattice the block @xmath38 as presented in fig.1 .
it contains the lattice sites @xmath39 , being centred on the site @xmath37 .
the numbering of the sites inside the block is block independent and given by the numbers @xmath40 in fig .
1 . the site - numbering inside the block is considered block - independent given by the translational symmetry of the system .
starting from the block @xmath38 we are introducing block operators as a sum of fermionic operators acting on the sites contained in @xmath38 as follows @xmath41 where the @xmath42 and @xmath43 prefactors are numerical coefficients . taking a block operator centred on another site @xmath44 , the indices of the fermionic operators follow the indices of the new sites @xmath45 , but for the numerical prefactors of @xmath46 we are keeping the same strategy in notation : @xmath47 represents the centre of the block , the index @xmath48 is of the site at the bottom of the block , the notation inside the block continuing anti clock - wise at the border of the block from @xmath49 up to the index @xmath50 .
the block operator @xmath51 used in the present description has significant differences in comparison to the plaquette operators used previously @xcite .
its novelty is twofold : ( i ) the here introduced block operator contains @xmath2-fermionic operators only in its unique central site , so contains a non - homogeneous f - operator action inside the block , and ( ii ) the number of lattice sites per block being @xmath52 , the used block has an extension greater than an unit cell .
these are key features which allow the study of the pam hamiltonian without the presence of the direct hopping terms at the level of @xmath2-electrons . summing up now @xmath53 over all lattice sites and taking periodic boundary conditions into account in both directions ( for the result see appendix a. ) ,
the following relation is obtained @xmath54 \ : , \label{h9}\end{aligned}\ ] ] if the following equalities are present between the parameters of @xmath13 and the numerical prefactors of the block operators @xmath51 @xmath55 taking into account that @xmath56 , where @xmath57 represents the number of lattice sites , the hamiltonian from eq.([h6 ] ) becomes @xmath58 where @xmath59 is given by @xmath60 the hamiltonian contained in eq.([h11 ] ) will be analysed in detail below .
we mention that the transformation of the starting @xmath13 from eq.([h6 ] ) to the studied @xmath13 from eq.([h11 ] ) is possible only if the hamiltonian parameters ( considered known variables ) satisfy the equations eqs.([h10 ] , [ h12 ] ) . on their turn , eqs.([h10],[h12 ] ) determine the unknown block operator parameters @xmath61 in term of hamilton operator parameters that must be considered known variables .
this works only for the case in which @xmath13 from eq.([h6 ] ) can be written in the form of @xmath13 presented in eq.([h11 ] ) .
since the number of unknown variables is less than the number of hamiltonian parameters , solutions for the block operator parameters will be present only if some inter - dependences between @xmath62 are present .
these inter - dependences fix the parameter space region @xmath63 in which the here obtained results are valid ( see fig
the explications from sect.iii.b . ) .
we are studying the hamiltonian from eq.([h11 ] ) at a fixed number of particles @xmath64 in the system . as a consequence ,
since @xmath64 is a constant of motion , @xmath13 becomes @xmath65 , where the positive semidefinite operator @xmath66 has zero minimum eigenvalue , and the constant @xmath67 is given by @xmath68 . in these conditions , the ground - state wave function of the model inside @xmath69 is defined via @xmath70 . to find @xmath71
, we have to keep in mind that @xmath72 requires for its minimum ( and zero ) eigenvalue at least one @xmath2-electron on every lattice site , and that the introduced block operators satisfies the following properties @xmath73 starting from eq.([h13 ] ) we observe that the block operator part of eq.([h11 ] ) applied to @xmath74 gives zero .
furthermore , given by the presence of @xmath72 in @xmath13 , we add to the ground - state the contribution @xmath75 , where @xmath76 are arbitrary coefficients . as a consequence ,
the ground - state wave - function with the property @xmath70 becomes @xmath77 |0 \rangle \ : , \label{h14}\end{aligned}\ ] ] where , @xmath78 is the bare vacuum with no fermions present . the wave - function presented in eq.([h14 ] )
is the first exact ground - state obtained for generic pam at finite @xmath1 . the product in eq.([h14 ] ) must be taken over all lattice sites . because of this reason , the product of the creation operators in eq.([h14 ] ) introduces @xmath79 particles into the system , so @xmath71 corresponds to @xmath80 filling .
all degeneration possibilities of the ground - state are contained in eq.([h14 ] ) , since the wave function with the property @xmath81 at @xmath80 filling always can be written in the presented @xmath71 form .
we however underline that pam contains two hybridized bands , and @xmath80 filling for a two - band system means in fact half filled upper hybridized band ( the lower band being completely filled up ) .
we mention , that @xmath71 describes rigorously only the @xmath21 case , since the presence of the @xmath82 operator into the ground - state is just required by the non - zero @xmath1 value . as a consequence ,
the ground - state at @xmath83 can not be expressed in the form presented in eq.([h14 ] ) .
we are now interested to find the @xmath7 phase diagram region where the solution from eq.([h14 ] ) is valid .
for this reason the system of equations eqs.([h10],[h12 ] ) must be solved for the block operator parameters . solving the problem
, we are considering all @xmath13 parameters real . from the @xmath84 components of eq.([h10 ] ) we find @xmath85 , and introducing the anisotropy parameter @xmath86 ( which must be real since @xmath87 is real ) , we realize that all @xmath88 components of @xmath13 parameters can be expressed via @xmath89 components and @xmath90 as follows @xmath91 solutions are obtained for @xmath92 , @xmath93 , @xmath94 , and the remaining equations for the @xmath89 components of the @xmath13 parameters provide the solutions @xmath95 all @xmath96 being considered real . introducing the notations @xmath97 , and @xmath98 , the solutions require @xmath99 and are situated in the parameter space on the surface @xmath100 \ : .
\label{h17}\end{aligned}\ ] ] this surface is presented ( for @xmath101 ) in fig .
as can be seen , it extends from the small @xmath1 domain continuously to the high @xmath1 domain up to @xmath16 in the @xmath102 limit .
modifying @xmath90 , the general shape of the obtained phase diagram region will not be changed .
to be situated inside the phase diagram region @xmath69 where the reported ground - state occurs for example in the isotropic case @xmath103 , ( which means @xmath104 ) , the parameters @xmath105 can be arbitrarily chosen , and @xmath106 must hold together with eq.([h17 ] ) which determine @xmath107 . as can be observed ,
@xmath69 can be reached by quite physical @xmath13 parameter values .
the magnetic properties of the wave - function of the form presented in eq.([h14 ] ) have been analysed in detail previously @xcite . here
the expression of @xmath51 is completely new , but the described techniques can be well applied .
given by the arbitrary nature of the @xmath76 coefficients , @xmath71 possesses a large spin degeneracy in the total spin of the system , being globally paramagnetic . studying the particle number distribution on different sites created by the @xmath108 product over the creation operators in @xmath71 from eq.([h14 ] ) together with the concrete block operator presented in eq.([h8 ] )
, it turns out that the obtained ground - state wave - function contains different contributions with one , two and three particles per site in the lattice . as a consequence
, the system described by @xmath71 is not characterised by an uniform particle distribution , the expectation value of the hopping matrix elements and non - local hybridizations is non - zero , so the system is not localized and the electrons in the ground - state are itinerant .
all these information show that the deduced ground - state is an itinerant paramagnet .
the deduced ground - state being itinerant , its properties can be easier described using a @xmath109-space representation . starting this , for the fourier transform of the block operators we find @xmath110 where @xmath111 .
using now the definition of @xmath112 and @xmath31 from eqs.([h3],[h5 ] ) , we observe that exactly when the presented solution holds ( i.e. eqs.([h10],[h12 ] ) are satisfied ) , we obtain @xmath113 using the notation @xmath114 , we introduce new canonical fermi operators @xmath115 which satisfy standard fermionic anti - commutation rules . using now eqs .
( [ a2],[a3 ] ) , we realize that @xmath13 from eq.([h1 ] ) becomes @xmath116 where @xmath117 constant , and @xmath118 , @xmath119 .
since for the ground - state @xmath120 , we obtain @xmath121 , the hamiltonian exactly diagonalized for the ground - state , in the form @xmath122 so , for the deduced ground - state ( i.e.in the parameter space inside @xmath69 ) , the hamiltonian can be mapped into a two - band hamiltonian with separated bands ( determined also by @xmath1 ) whose upper band is completely flat ( note that the starting @xmath4-band in eq.([h1 ] ) is with dispersion and there is no hopping present for @xmath2-electrons in the starting hamiltonian ) . because of @xmath79 , the upper flat band ( in which fermions are created by @xmath123 ) is half filled , and the lower band ( in which @xmath124 creates particles ) is completely filled up . as a consequence , denoting the ground - state expectation values by @xmath125 , we have ( see also @xcite ) @xmath126 we also mention that since the lower band is completely filled up , @xmath127 .
the second relation from eq.([a6 ] ) is trivial from physical point of view since the lower band is completely filled up .
contrary to this , the first relation ( see fig .
3 . ) shows a momentum distribution function for the upper band without non - regularities in its derivatives of any order , signalling a clear non - fermi liquid type of behaviour in 2d deduced in exact terms .
the deduced phase is present also in the isotropic case ( @xmath128 ) , and we underline that the result has clear physical signification even in the case in which @xmath69 behaves complete repulsively from rg point of view @xcite . the novelty of this result in comparison with the behaviour reported in @xcite is threefold : ( i ) here we are situated in generic pam ( i.e. @xmath14 ) ; ( ii ) the hybridization matrix element is real ; and ( iii ) distorted unit cell is not necessary for the emergence of the itinerant phase . as a consequence ,
the presence of the deduced behaviour is much more general then suggested by our previous work . expressing @xmath129 from eq.([a3 ] ) , and using eq.([a6 ] ) , all needed ground - state expectation values connected to @xmath13 in term of the starting fermionic operators can be deduced . introducing the notation @xmath130 based on eq.([a7 ] ) it can be checked that @xmath131 and @xmath132 as well are free from non - regularities in the whole first brillouin zone .
the ground - state energy becomes @xmath133 .
we are presenting for the first time exact ground - states for the generic periodic anderson model ( pam ) at finite on - site repulsion for @xmath2-electrons @xmath1 , in @xmath0 dimensions , the hamiltonian not containing direct hopping terms for @xmath2-electrons @xmath3 .
for this reason , and on this line a ) we generalised the previously used elementary plaquette operators @xcite to a block operator containing non - uniform @xmath2-operator contributions ( the @xmath2-operators acting only on one site of the block ) ; b ) the block itself has been chosen to have an extension higher than an unit cell , so it can not be considered elementary plaquette operator as used in previous studies ; c ) based on the presented developments it was possible to analyse for the first time for pam the @xmath14 generic case at finite @xmath1 and 2d in exact terms ( in restricted regions of the parameter space ) ; d ) again based on points a),b ) an itinerant state holding non - fermi liquid properties and presenting similarities with the itinerant state deduced in ref.[@xcite ] has been obtained without distortions in the system and presence of real hybridization matrix elements ( we underline that results of the type mentioned in the points c),d ) can not be deduced in the frame of refs .
( @xcite ) ) ; e ) the deduced exact results allow to present the first exact phase diagram region for the generic pam at finite @xmath1 and 2d , which is not included in the previously deduced phase diagram regions ; f ) starting from the presented technique and results we learn that using different type of blocks in constructing block operators , different classes of hamiltonians can be analysed . the block form ,
its extension , the uniform or non - uniform nature of the operator action inside the block , the spin dependence or non - dependence of the numerical coefficients of the block operator , all are important in this description process ; g ) the fact that at least in some regions of the parameter space ( which however extend from the low @xmath1 to the high @xmath1 limit ) , similar phases are obtained in exact terms at @xmath9 and @xmath11 for pam , underlines that this model can provide a behaviour differently from kondo physics ( where only the local f - moments and its compensation play the main role ) .
this can happen not only in the region of a potential metal - insulator transition but also elsewhere in the phase diagram .
the research has been supported by contract otka - t-037212 and fkfp-0471 of hungarian founds for scientific research .
the author kindly acknowledge extremely valuable discussions on the subject with dieter vollhardt .
he also would like to thank for the kind hospitality of the department of theoretical physics iii .
, university augsburg in autumn 2001 , 4 months of working period relating this field spent there , and supported by alexander von humboldt foundation .
the expression @xmath53 summed up over the whole lattice considered with periodic boundary conditions in both directions is presented below .
@xmath134 + [ \hat d^{\dagger}_{{\bf i},\sigma } \hat
d_{{\bf i } + { \bf y } , \sigma } ( a^{*}_{1,d } a_{4,d } + a^{*}_{2,d } a_{1,d } ) + h.c . ]
+ \nonumber\\ % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % & & [ \hat d^{\dagger}_{{\bf i},\sigma } \hat d_{{\bf i } + ( { \bf x}+{\bf y } ) , \sigma } ( a^{*}_{5,d } a_{4,d } + a^{*}_{2,d } a_{3,d } ) + h.c . ]
+ [ \hat d^{\dagger}_{{\bf i},\sigma } \hat d_{{\bf i } + ( { \bf y}-{\bf x } ) , \sigma } ( a^{*}_{2,d } a_{5,d } + a^{*}_{3,d } a_{4,d } ) + h.c . ] + \nonumber\\ % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % & & [ \hat d^{\dagger}_{{\bf i},\sigma } \hat d_{{\bf i } + 2{\bf y},\sigma } a^{*}_{2,d } a_{4,d } + \hat d^{\dagger}_{{\bf i},\sigma } \hat d_{{\bf i } + 2{\bf x } , \sigma } a^{*}_{5,d } a_{3,d } + h.c . ] + \nonumber\\ % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % & & [ \hat f^{\dagger}_{{\bf i},\sigma } \hat d_{{\bf i } + { \bf x},\sigma } a^{*}_{1,f } a_{3,d } + \hat d^{\dagger}_{{\bf i},\sigma } \hat f_{{\bf i } + { \bf x } , \sigma } a^{*}_{5,d } a_{1,f } + h.c . ]
+ \nonumber\\ % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % & & [ \hat f^{\dagger}_{{\bf i},\sigma } \hat d_{{\bf i } + { \bf y},\sigma } a^{*}_{1,f } a_{4,d } + \hat d^{\dagger}_{{\bf i},\sigma } \hat f_{{\bf i } + { \bf y } , \sigma } a^{*}_{2,d } a_{1,f } + h.c . ] + \nonumber\\ % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % & &
\hat f^{\dagger}_{{\bf i},\sigma } \hat f_{{\bf i } , \sigma } |a_{1,f}|^2 + \hat d^{\dagger}_{{\bf i},\sigma } \hat d_{{\bf i } , \sigma } ( \sum_{n=1}^5 |a^{*}_{n , d}|^2 ) + [ \hat d^{\dagger}_{{\bf i},\sigma } \hat f_{{\bf i},\sigma } a^{*}_{1,d } a_{1,f } + h.c . ] \ } \ : . % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % % \label{b1}\end{aligned}\ ] ] | we report for the first time exact ground - states deduced for the @xmath0 dimensional generic periodic anderson model at finite @xmath1 , the hamiltonian of the model not containing direct hopping terms for @xmath2-electrons @xmath3 .
the deduced itinerant phase presents non - fermi liquid properties in the normal phase , emerges for real hybridization matrix elements , and not requires anisotropic unit cell . in order to deduce these results ,
the plaquette operator procedure has been generalised to a block operator technique which uses blocks higher than an unit cell and contains @xmath2-operator contributions acting only on a single central site of the block . |
a lot of techniques and concepts of statistical mechanics of disordered spin systems , in particular , the replica method originally used to analyze the thermodynamics of spin glass model by sherrington and kirkpatrick @xcite , have been applied to various research fields beyond conventional physics , i.e. information processing @xcite , game theory @xcite and so on . the exactly solvable mathematical model , which describes these problems , is categorized as mean - field class @xcite . on the other hand , as another exactly tractable model , in 1907 , paul and tatiana ehrenfest published a paper corroborating boltzmann s view of thermodynamics @xcite .
their urn model has been defined by kac @xcite as an exactly solvable example in statistical physics . while it has also been criticized as a marvelous exercise too far removed from reality ,
their urn model has been applied to modern problems such as complex networks @xcite or econophysics @xcite , etc .
for instance , based on extensive simulations of the lennard - jones fluid requiring in part a parallel computer in juelich , an italian - german team has shown that the prediction of the ehrenfest urn effectively describes the behavior of the gas phase @xcite .
moreover , it has been revealed that the mathematical structure of equilibrium state of the urn model @xcite is similar to the zero - range process , which has been widely investigated in research fields of non - equilibrium statistical physics @xcite .
recently , in the research field of complex networks @xcite , ohkubo et al .
@xcite proposed a network model based on ` ehrenfest class urn model ' to explain how the complex network gets scale - free - like properties , where ` ehrenfest class ' means that each urn has distinguishable balls . in the model , each urn corresponds to a node in graph ( network ) and the number of distinguishable balls , @xmath1 , in each urn is regarded as degree of nodes .
for this model system , they succeeded in deriving the scale - free - like properties @xmath3 in the probability of the degree of nodes by the usage of the replica symmetric theory @xcite .
in addition , the similarity between the disordered urn model and the random field ising model @xcite , and the condensation phenomena in the disordered urn model have been investigated @xcite .
we here note that there are a lot of works in which the power - law behavior and the condensation phenomena in urn models have been studied @xcite . for example , in the zeta urn model @xcite , the power - law behavior in the probability of the number of balls , i.e. , the occupation distribution , stems from a power - law form of the boltzmann weight .
however , when we attempt to describe various problems in the real world , we should take into account the disorder and treat the urns as a heterogeneous system .
mainly , the previous models which cause power - law behavior in the occupation distribution do not contain any disorder , and hence it would be important to investigate ` disordered ' urn models which cause the power - law distribution . in this paper , we propose two disordered urn models in which quenched randomness is important for generating the power - law behavior .
one of them belongs to the ehrenfest class , and the other corresponds to the monkey class , in which each urn has indistinguishable balls . in particular , for the monkey class urn model , we investigate a real - space condensation phenomenon in which macroscopic number of balls are condensed into only one urn .
the occupation probability @xmath0 with which an urn has @xmath1 balls is calculated analytically , and furthermore , the critical density @xmath4 for a given temperature is evaluated . as the result , it is shown that the occupation distribution function @xmath0 changes its scaling behavior from the exponential @xmath5-law to the @xmath6 power - law in large @xmath1 regime .
this paper is organized as follows . in the next section [ sec : general ]
, we introduce the general formalism for the urn model with an arbitrary energy function .
although there are several analytical treatments for disordered urn models @xcite , we give the formalism for the disordered urn model with an arbitrary energy function with a different point of view , and additionally , the analytical treatment makes this paper self - contained .
this analytical treatment contains both ehrenfest and monkey classes as its special cases .
we explain the relation between the saddle point , that determines the thermodynamic properties of the system , and the chemical potential . with the assistance of this general formalism
, we provide an analysis for a special choice of the energy function , which is an example of the ehrenfest class in section [ sec : ehrenfest ] .
we discuss the condition on which the power - law appears in the tail of the occupation probability for the model . in section [ sec : be ] , we show that the condensation occurs for the special case of the monkey class with disorder , and a heavy tailed power - law emerges in the occupation probability . in section [ sec : econophysics ] , we provide a possible link between our results and macro economy , in particular , wealth differentials .
last section is a summary .
let us prepare @xmath7 urns and @xmath8 balls ( @xmath9 ) and consider the situation in which the @xmath7 urns share the @xmath8 balls .
then , we start our argument from the ehrenfest class urn model @xcite in which each ball in urns is distinguishable . for the mathematical model categorized in the ehrenfest class , the boltzmann weight @xmath10 that @xmath11-th urn possesses @xmath12 balls is given by @xmath13}{n_{i}!},\end{aligned}\ ] ] where @xmath14 is an energy function , @xmath15 a disorder parameter for urn @xmath11 , and @xmath16 the inverse temperature of the system .
the factorial @xmath17 stems from the property of the ehrenfest class @xcite .
the only point of our analysis which is different from @xcite is quenched disorder @xmath18 appearing in the energy function . for the ehrenfest class ,
the probability that an urn specified by the disorder parameter @xmath19 possesses @xmath1 balls is given by @xmath20 \oint \frac{dz}{2\pi i } { \exp } \left [ \left(k+\sum_{i=2}^{n } n_{i}-m-1 \right ) \log z \right ] \nonumber \\
\fl \mbox { } = \frac{e^{-\beta e(\epsilon_1 , k)}}{k ! } \left ( \frac{z_{2}}{z_{1 } } \right ) \label{eq : ehren},\end{aligned}\ ] ] where @xmath21 and we defined @xmath22 as @xmath23 \oint \frac{dz}{2\pi i } { \exp } \left [ \left ( k+\sum_{i=2}^{n } n_{i}-m-1
\right ) \log z \right],\end{aligned}\ ] ] and used the fourier transform of the kronecker - delta : @xmath24 to introduce the conservation of the total balls : @xmath25 into the system . in order to calculate @xmath0
, we take the configuration average of @xmath26 .
while one can calculate the configuration average by means of the replica method @xcite , it has been revealed that the mathematical structure of the disordered urn model is related to that of a random field ising model @xcite .
hence , we here use the law of large numbers and simplify the calculation . to calculate the average of the quantity @xmath27 $ ] over the configuration @xmath28 , we consider the tayler - expansion : @xmath29 \rangle_{\{1,2,\cdots , n\ } } & = & 1 + \langle \log f_{k}^{(\epsilon_{1 } ) } ( \{\epsilon_{i/1}\ } ) \rangle_{\{1,2,\cdots , n\ } } \nonumber \\ \mbox { } & + & \frac{1}{2 } \langle ( \log f_{k}^{(\epsilon_{1 } ) } ( \{\epsilon_{i/1}\}))^{2 } \rangle_{\{1,2,\cdots , n\}}+\cdots \label{eq : tayler1}\end{aligned}\ ] ] where @xmath30 means the configuration average over @xmath31 . here
we assume that the observable @xmath32 for a given realization of configuration @xmath28 is almost identical to the average @xmath33 .
in other words , the deviation is vanishingly small as @xmath34 in the thermodynamics limit . by using the assumption ,
we rewrite ( [ eq : tayler1 ] ) as follows .
@xmath29 \rangle_{\{1,2,\cdots , n\ } } & \simeq & 1 + \langle \log f_{k}^{(\epsilon_{1 } ) } ( \{\epsilon_{i/1}\ } ) \rangle_{\{1,2,\cdots , n\ } } + \frac{1}{2 } \langle \log f_{k}^{(\epsilon_{1 } ) } ( \{\epsilon_{i/1}\ } ) \rangle_{\{1,2,\cdots , n\}}^{2 } \nonumber \\
\mbox { } & + & \cdots = \exp [ \langle \log f_{k}^{(\epsilon_{1 } ) } ( \{\epsilon_{i/1}\ } ) \rangle_{\{1,2,\cdots , n\ } } ] \label{eq : tayler2}\end{aligned}\ ] ] this replacement of the configuration average reads @xmath35 + \langle \log z_2 \rangle_{\ { 2,3,\cdots , n\ } } - \langle \log z_1 \rangle_{\{2,3,\cdots , n\ } } \right\rangle_{\{1\ } } \right ) .
\end{aligned}\ ] ] in the thermodynamic limit @xmath36 , @xmath37 is evaluated as @xmath38 \right ) \right\rangle_{\{2,3,\cdots , n\ } } \nonumber \\
\fl \simeq \log \left ( \oint \frac{dz}{2\pi i } \exp\left [ ( k-\rho n-1 ) \log z + ( n-1)\left\langle \log \sum_{n=0}^{\infty } \frac{{\rm e}^{-\beta e(\epsilon , n ) } } { n ! } z^{n } \right\rangle \right ] \right ) \nonumber \\
\fl \simeq k\log z_{s1 } - ( \rho n + 1 ) \log z_{s1 } + ( n-1 ) { \biggr \langle } \log \sum_{n=0}^{\infty } \frac{{\rm e}^{-\beta e(\epsilon , n ) } } { n!}z_{s1}^{n } { \biggr \rangle } \label{eq : genez2},\end{aligned}\ ] ] where we used the law of large numbers , and in the final line the saddle point method was used .
@xmath39 means the average over only @xmath40 , namely , @xmath41 . in the next two sections , we consider specific choices of @xmath42 to evaluate the occupation probability distribution concretely . using the same way as the @xmath22 , @xmath43 , which is obtained by the normalization condition of @xmath44 ,
namely , @xmath45 with equation ( [ eq : ehren ] ) , is rewritten as @xmath46 we easily find @xmath47 because the first terms for each saddle point equation ( [ eq : genez2 ] ) or ( [ eq : genez1 ] ) are vanishingly smaller than the other two terms in the limit of @xmath36 .
thus , we obtain the saddle point equation with respect to @xmath48 and the occupation probability @xmath49 that an arbitrary urn with the energy function @xmath50 at inverse temperature @xmath16 has @xmath1 balls are given by @xmath51 and @xmath52 respectively .
it should be noticed that the above saddle point equation for the ehrenfest class urn model ( [ eq : saddle ] ) is now rewritten in terms of chemical potential @xmath53 as @xmath54 } } { n ! } } { \sum_{n=0}^{\infty } \frac{{\rm e}^{-\beta [ e(\epsilon , n)-n\mu ] } } { n ! } } { \biggr \rangle}. \label{eq : def_rho}\end{aligned}\ ] ] then , we define the probability @xmath55 that an arbitrary ehrenfest class urn with energy @xmath50 has @xmath56 balls by @xmath57 } } { n!}. \label{eq : dist_n}\end{aligned}\ ] ] from this formula of the probability @xmath55 with the effective boltzmann factor @xmath58 , the equation ( [ eq : def_rho ] ) means that the ratio @xmath59 corresponds to the average number of balls put in an arbitrary urn : @xmath60 , and its value is controlled by the chemical potential @xmath61 through the equation ( [ eq : def_rho ] ) . then , the chemical potential @xmath61 and the saddle point @xmath62 are related through the equation ( [ eq : chemi ] ) .
therefore , when we construct the system so as to have a density @xmath63 , the corresponding saddle point @xmath62 is given by ( [ eq : def_rho ] ) . as the result , the chemical potential @xmath61 that gives @xmath63 is determined by the relation ( [ eq : chemi ] ) .
thus , our problem is now to solve the saddle point equation @xmath64 and to calculate the following averaged occupation probability for the solution @xmath65 of the equation ( [ eq : gene_saddle ] ) : @xmath66 now it is time for us to stress that the ehrenfest or monkey class is recovered if we choose the effective boltzmann factor @xmath67 as follows @xcite
. @xmath68 & ( \mbox{ehrenfest class } ) , \\ { \exp } \left [ -\beta ( e(\epsilon , n)-n\mu ) \right ] & ( \mbox{monkey class } ) . \end{array } \right .
\label{eq : monkey}\end{aligned}\ ] ] it should be noted that in our formalism , the distinction between two models only comes from the difference of the effective boltzmann factor ( [ eq : monkey ] ) .
we here also comment on the effect of the disorder to the dynamics of urn models . in the uniform case without disorder ,
all urns are equivalent , and we do not distinguish each urn in the monkey class . on the other hand , in the disordered urn model , each urn @xmath11 has own label @xmath15 , and hence they are distinguishable at least in principle .
however , we assume that this heterogeneity of urns does not change the `` monkey '' nature in the monkey class ; we can not see the disorder parameter assigned to each urn from outside , and the dynamics ( so called ` box - to - box choice ' ) is not changed .
as an demonstration of the ehrenfest urn model whose thermodynamic properties are specified by equations ( [ eq : saddle ] ) and ( [ eq : dist ] ) , we introduce a kind of disordered ehrenfest urn models and consider the condition on which the power - law appears . to this end
, we choose the energy function @xmath69 as @xmath70 where @xmath40 means an urn - dependent disorder of the system , and we here assume that @xmath40 takes a value in the range @xmath71 $ ] randomly , that is , @xmath72 with the step function @xmath73 .
of course , we might choose the other distribution @xmath42 , however , the main issue of this section is to examine whether the power - law appears in the occupation probability distribution when we introduce the disorder , and for this purpose , the simplest choice is enough .
the tendency of this energy function to force each urn of the system to gather balls as much as possible results in the fact that _ the rich get richer _ as its collective behavior . we should mention that in @xcite , the co - called backgammon model @xcite described by the cost function @xmath74 for each urn was studied . in the model ,
the cost decreases if and only if each urn is empty . in this sense ,
the model we deal with in this section is regarded as an opposite situation of our model .
therefore , it is interesting to investigate whether there exists any significant difference or the similarity between two models from the view point of the occupation probability distribution .
the extensive studies concerning this issue will be our future studies . for this choice of the energy function ( [ eq : ene9 ] ) , the saddle point equation ( [ eq : saddle ] ) leads to @xmath75 from equation ( [ eq : dist ] ) , the occupation probability for the choice ( [ eq : ene9 ] ) , @xmath76 , is given by @xmath77 in following , we evaluate the above occupation probability .
we first show the phase diagram that indicates the area of the power - law behavior .
-@xmath1 phase diagrams concerning the heavy tailedness of the occupation probability . in the phase diagram for @xmath78 ( left panel ) , the occupation probability follows non - power , exponential - law .
the power - law regime exists for @xmath79 . in the right panel
, we plot the phase boundaries @xmath80 as a function of density @xmath63 for several values of inverse temperature , namely , @xmath81 and @xmath82 .
, title="fig:",width=291 ] -@xmath1 phase diagrams concerning the heavy tailedness of the occupation probability . in the phase diagram for @xmath78 (
left panel ) , the occupation probability follows non - power , exponential - law .
the power - law regime exists for @xmath79 . in the right panel
, we plot the phase boundaries @xmath80 as a function of density @xmath63 for several values of inverse temperature , namely , @xmath81 and @xmath82 .
, title="fig:",width=291 ] in figure [ fig : fgjpa2 ] ( left ) , we show the phase diagram for the case of @xmath78 .
the shaded area in this figure means non - power , exponential - law region . from this figure
, we find that while the upper bound ( the cut - off ) @xmath83 and the lower bound @xmath84 increase as the density @xmath63 increases , the cut - off @xmath83 increases much more quickly than the lower bound @xmath84 . as the result , the heavy tailed power - law @xmath85-region is broadened by increase of the density @xmath63 . in the right panel of figure [ fig : fgjpa2 ] , we display the inverse temperature dependence of the area of the power - law . as temperature increases ,
the area of the power - law shrinks to zero .
then , we have a poisson - law in whole region of the phase diagram in the high - temperature limit of @xmath86 .
we explain the detail of the evaluations as follows .
we first consider the high - temperature limit @xmath87 . for this case
, the saddle point ( [ eq : saddle2 ] ) leads to @xmath88 and we obtain @xmath89 which is nothing but a poisson distribution .
for finite temperature @xmath90 , the occupation probability @xmath76 is rewritten as @xmath91 \label{eq : henkan},\end{aligned}\ ] ] where we defined the incomplete gamma function by @xmath92 . in figure
[ fig : fgjpa1 ] , we plot the occupation probability @xmath0 for several values of @xmath63 for @xmath78 , @xmath93 and @xmath94 , namely , @xmath95 and @xmath96 . as a function of @xmath1 for several values of @xmath62 , namely , @xmath93 and @xmath97 , namely , @xmath95 and @xmath96 .
we set the inverse temperature @xmath98 .
the right panel shows the behavior of @xmath99 .
, title="fig:",width=291 ] as a function of @xmath1 for several values of @xmath62 , namely , @xmath93 and @xmath97 , namely , @xmath95 and @xmath96 .
we set the inverse temperature @xmath98 .
the right panel shows the behavior of @xmath99 . , title="fig:",width=291 ] from this figure , we find that @xmath76 obeys a power - law @xmath100 in the intermediate regime of @xmath1 and there exists a cut - off value from which the distribution decays exponentially .
to see the existence of the cut - off value explicitly , we rewrite the distribution @xmath76 as @xmath101.\end{aligned}\ ] ] in figure [ fig : fgjpa1 ] ( right ) , we plot the function @xmath99 for several values of @xmath62 at @xmath102 .
we easily find that in the range of @xmath103 $ ] defined in terms of the function @xmath99 as @xmath104 , @xmath105 , the occupation probability @xmath0 follows a power - law distribution @xmath100 .
thus , we specified the region @xmath106 in which a power - law heavy tail appears in the occupation distribution as shown in figure [ fig : fgjpa2 ] . as shown in this section , for the ehrenfest class disordered urn model
, we could clarify the control parameters of the system for which the heavy tail power - law emerges in the occupation probability . in the next section ,
we consider a monkey class urn model with disorders .
in the previous section , we evaluated the asymptotic form of the occupation probability @xmath76 for the urn model of the ehrenfest class with energy function @xmath107 .
obviously , from the view point of the energy cost , it is a suitable strategy for each urn to gather balls as much as possible . in that sense , this case should be referred by the concept _ the rich get richer _ in the context of social networks . however , by using the general definition of the problem , we freely choose the energy function for both the ehrenfest and monkey classes . in this section , for the monkey class urn model
whose thermodynamic properties are defined by equations ( [ eq : gene_saddle ] ) , ( [ eq : gene_dist_k ] ) and ( [ eq : monkey ] ) , we evaluate @xmath76 for a specific choice of energy function @xmath69 .
we choose the energy @xmath69 as @xmath108 we should notice that for this simple choice of the energy function , the urn labeled by @xmath109 is hard to gather the balls . on the other hand , the urn with @xmath110 energy level
easily gathers the balls . the urn model having this type of energy function does not agree with the concept _ the rich get richer_. nevertheless , we use the energy function ( [ eq : energy_en ] )
because as we shall see below , a kind of condensation with respect to the urns occurs for this choice of energy function , and as the result , the power - law in the tail of the occupation probability emerges . for a given choice of @xmath42 as the density of state ,
namely , degeneracy of the energy level of the urn , we rewrite the saddle point equation ( [ eq : gene_saddle ] ) as follows : @xmath111 to proceed to the next stage of the calculation , we choose the density of state @xmath42 explicitly as @xmath112 where @xmath113 is a constant . although we chose the above form , more general setup of the argument by choosing @xmath114 is possible .
we shall discuss the result for this kind of generalization later on .
then , the equation ( [ eq : rho_be1 ] ) is rewritten by @xmath115 where @xmath116 means the density of balls in the urn labeled by the zero - energy level @xmath110 .
we should notice that the second term appearing in the right hand side of equation ( [ eq : rho_be2 ] ) , namely , @xmath116 vanishes in the thermodynamic limit @xmath36 when a condensation does not occur .
in other words , when the condensation arises , the second term , @xmath116 , becomes from zero to a finite value ; this means that an urn with @xmath117 becomes to have a macroscopic number of balls . in following ,
we show the system undergoes a condensation and investigate the behavior of the system when the density @xmath63 increases beyond the critical point @xmath4 for a given finite inverse - temperature @xmath16 . * before condensation : @xmath118 + by a simple transformation @xmath119 , the equation ( [ eq : rho_be2 ] ) is rewritten in terms of the so - called appeli function ( see e.g. @xcite ) @xmath120 as follows : @xmath121 where the appeli function ( see e.g. @xcite ) @xmath120 is defined by means of the gamma function @xmath122 as @xmath123 we should keep in mind that @xmath124 is satisfied ( @xmath125 ) .
the solution of the saddle point equation ( [ eq : saddle_before ] ) possesses a solution @xmath126 .
* at the critical point : @xmath127 + the critical point at which the condensation occurs is determined by the radius of convergence for the following partition function : @xmath128 namely , @xmath129 for @xmath110 gives the critical point .
therefore , substituting @xmath129 for a given density level @xmath63 , the critical point @xmath4 , above which the condensation occurs , is obtained by @xmath130 * after condensation : @xmath131 + for @xmath131 , the saddle equation of ( [ eq : saddle_before ] ) no longer has any solution as @xmath132 .
obviously , for the solution @xmath133 , the partition function diverges .
then , we should bear in mind that the term @xmath116 in ( [ eq : rho_be2 ] ) , which was omitted before the condensation , becomes @xmath134 object and the saddle point equation we should deal with is not ( [ eq : saddle_before ] ) but ( [ eq : rho_be2 ] ) . as the result , the equation ( [ eq : rho_be2 ] ) has a solution @xmath129 even for @xmath131 and the number of balls @xmath135 in the condensation state increases linearly in @xmath63 as @xmath136 whereas the number of balls in excited states reaches @xmath137 .
thus , we obtained the saddle point @xmath62 for a given density and inverse - temperature .
we found that the condensation is specified by the solution @xmath129 .
we next investigate the density dependence of the occupation probability through the saddle point .
for the solution of the saddle point equation @xmath62 , the occupation probability at inverse temperature @xmath16 is evaluated as follows .
@xmath138 the above occupation probability is valid for an arbitrary integer value of @xmath1 for @xmath139 . ) in non - asymptotic regime .
we set @xmath140 and @xmath62 as @xmath141 , and the inverse temperate is @xmath142 .
the inset of the figure shows the asymptotic behavior of the occupation probability @xmath76 as log - log plots for the cases of @xmath143 and @xmath144 .
, width=377 ] we plot the behavior of the occupation probability @xmath76 in finite @xmath1 regime in figure [ fig : fg3 ] . in this plot , we set @xmath140 and @xmath62 as @xmath141 , and the inverse temperate is @xmath142 . in the inset of the same figure , we also show the same data in log - log scale for the asymptotic behavior of the probability @xmath76 for several values of @xmath62 , namely , @xmath143 and @xmath144 . from this figure
, we find that the power - law @xmath145 emerges when the condensation is taken place for @xmath131 .
the numerical analysis of the occupation probability ( [ eq : be_dist2 ] ) in the limit of @xmath146 is easily confirmed by asymptotic analysis of equation ( [ eq : be_dist2 ] ) .
we easily find that the asymptotic form of the wealth distribution @xmath76 behaves as @xmath147 we also should notice that a macroscopic number of balls @xmath135 is gathered to a specific urn with energy level @xmath110 when the condensation occurs .
as the result , the term such as @xmath148 should be added to the occupation probability .
let us summarize the results as follows : @xmath149 & ( \rho < \rho_{\rm c } : z_{s } < 1 ) , \\
% % % % % % % \frac{3 \varepsilon_{0 } \gamma ( 3/2 ) } { 2\beta^{3/2}}\ , k^{-5/2 } + \frac{1}{n } \delta ( k - k _ { * } ) & ( \rho \geq \rho_{\rm c } : z_{s}=1 ) . \end{array } \right .
\label{eq : result3}\end{aligned}\ ] ] the scenario of the condensation is as follows . for a given @xmath118 , one can find a solution @xmath126 for the saddle point equation ( [ eq : rho_be2 ] ) , and hence the second term of equation ( [ eq : rho_be2 ] ) is zero in the thermodynamic limit .
then , for non - condensed @xmath150 balls , the occupation probability follows @xmath151-law .
namely , urns possessing a large number of balls do not appear due to the repulsive force as @xmath152 . when @xmath131 , the saddle point @xmath62 is fixed as @xmath153 ; if @xmath154 , the first term of the saddle point equation ( [ eq : rho_be2 ] ) has a singularity .
therefore , in order to avoid the singularity , the second term of the saddle point equation ( [ eq : rho_be2 ] ) becomes from zero to a finite value . as the result
, the occupation probability is described by the @xmath145-law with a delta peak which corresponds to an urn of @xmath110 gathering the condensed @xmath155 balls .
this corresponds to the condensation phenomena in the disordered urn model . in particular ,
the occurrence of the condensation in the disordered urn model treated in the present paper is characterized by the transition from the exponential - law to the heavy tailed power - law .
we also mention the effect of disorder on the power - law behavior of the occupation probability .
we easily find that the power - law behavior disappears when one cancels the disorder of the system by choosing the density of the energy such as @xmath156 ( @xmath157 is a constant ) .
this fact means that the disorder appealing in the system possesses a central role to make the occupation probability to have a power - law behavior .
we should notice that in the above argument , the solution @xmath129 that indicates the condensation does not change even if we choose the density as @xmath158 .
for this choice , the @xmath4 is given by @xmath159 then , one obtains the following occupation probability @xmath160 at the end of this section , we should mention the result for the uniform distribution of @xmath40 , that is , the case of @xmath161 leading to @xmath162 . for this choice
, we have the following occupation probability @xmath163 then , beyond the critical density @xmath164 , the occupation probability @xmath0 behaves as @xmath165 therefore , after the condensation , the crossover from the @xmath166-law to the @xmath167-law is observed around @xmath168 and as the result , the power - law heavy tail appears .
in this section , we reconsider the results obtained in the previous sections from a view point of macro economics .
it is easy for us to regard the occupation probability as wealth distribution when we notice the relations : balls - _ money _ and urns - _ people in a society_. in following , we attempt to find an interpretation of the condensation and the emergence of the pareto - law @xcite in terms of wealth differentials @xcite . in section [ sec : ehrenfest ] , we devoted our analysis to extremely large income regimes ( the tail of the wealth distribution ) , however , it is quite important for us to consider the whole range of the wealth . as reported in @xcite , the wealth distribution for small income regime follows the gibbs / log - normal law and a kind of transition to the pareto - law phase is observed . for the whole range distribution of the wealth , the so - called lorentz curve @xcite is obtained .
the lorentz curve is given in terms of the relation between the cumulative distribution of wealth and the fraction of the total wealth .
then , the so - called gini index @xcite , which is a traditional , popular and one of the most basic measures for wealth differentials , could be calculated .
the index could be changed from @xmath169 ( no differentials ) to @xmath144 ( the largest differentials ) .
for the energy function ( [ eq : energy_en ] ) in the previous section , we derived the wealth distribution for the whole range of incomes @xmath1 . in this section
, we evaluate the gini index analytically .
as we mentioned above , the lorentz curve is determined by the relation between the cumulative distribution of wealth @xmath170 and the fraction of the total wealth @xmath171 for a given wealth distribution @xmath0 .
for instance , the lorentz curve for the exponential distribution @xmath172 is given by @xmath173 we should notice that the lorentz curve for the exponential distribution is independent of @xmath174 . for the power - law distribution @xmath175 , we have @xmath176 as the lorentz curve .
this curve depends on the exponent @xmath174 . in figure
[ fig : fg4 ] , we plot the lorentz curve for the exponential distribution ( [ eq : l_exp ] ) and the power - law distribution ( [ eq : l_pow ] ) with several values of @xmath174 . and the lorentz curve ( e.g. @xmath177 for exponential distributions ) .
the right panel shows the lorentz curves for the exponential distribution ( [ eq : l_exp ] ) and the power - law distributing ( [ eq : l_pow ] ) with several values of @xmath178 and @xmath94 .
, title="fig:",width=291 ] and the lorentz curve ( e.g. @xmath177 for exponential distributions ) .
the right panel shows the lorentz curves for the exponential distribution ( [ eq : l_exp ] ) and the power - law distributing ( [ eq : l_pow ] ) with several values of @xmath178 and @xmath94 .
, title="fig:",width=291 ] then , as shown in the left panel of figure [ fig : fg4 ] , the gini index @xmath179 is defined as an area between the perfect equality line @xmath180 and the lorentz curve .
this quantity explicitly reads @xmath181 and we have @xmath182 @xcite for the exponential distribution and @xmath183 for the power - law distribution . as the occupation probability distribution ( [ eq : be_dist2 ] ) is defined for @xmath184 , one can evaluate the gini index as a function of the saddle point @xmath62 . in figure
[ fig : fg5 ] , we plot the lorentz curve ( left ) for several values of @xmath62 . in the right panel ,
the gini index @xmath185 is shown .
we find that the index approaches to @xmath186 as @xmath187 . from the argument in the previous section
, we easily find that the occupation distribution for @xmath188 non - condensed balls beyond the critical point is modified such as @xmath189 by choosing the density of the energy @xmath190 .
namely , for the pareto - law distribution @xmath191 , the gini index leads to @xmath192 .
therefore , the condensation is specified by the change of the gini index from @xmath182 to @xmath193 .
however , we should keep in mind that the gini index itself has less information about the differentials than the wealth distribution .
for example , the gini index for @xmath194 of the pareto - law @xmath195 gives the same gini index as the exponential distribution .
this fact stems from the definition of the gini index @xmath179 , that is , @xmath179 is defined as an area between @xmath180 and the lorentz curve .
it could be possible to draw lots of the lorentz curves that give the same area ( the same gini index ) . as we explained above ,
it should be noted that actually the gini index is one of the measure for the earning differentials , however , the wealth distribution is much more informative than the gini index . ) .
the right panel shows the gini index for several values of @xmath62 . ,
title="fig:",width=291 ] ) .
the right panel shows the gini index for several values of @xmath62 .
, title="fig:",width=291 ] although the gini index is less informative than the distribution , however , for a given real ( empirical ) data @xmath196 ( @xmath197 denotes the amount of money for example .
such empirical data are not always massive enough for us to specify the distribution ) , one can evaluate it as a statistics by @xmath198 with @xmath199 @xcite .
therefore , it is helpful for us to use the gini index to compare the earning differentials between different countries ( the population of each country should be different and of course it is finite @xmath200 ) . in our analysis of this paper ,
the distribution was analytically obtained in the thermo - dynamic limit because we treated an ideal case as a society .
nevertheless , even if we encounter more realistic situation for which the analytical evaluation of wealth distribution is very tough , one can evaluate the earning differentials via the gini index by computer simulations for finite population @xmath7 .
then , one can investigate the earning differentials by comparing the numerical results with the analytical expressions obtained in this paper .
in this paper , we investigated equilibrium properties of disordered urn model and discuss the condition on which the heavy tailed power - law appears in the occupation probability by using statistical physics of disordered spin systems .
we applied our formalism to two urn models of both the ehrenfest and monkey classes .
in particular , for the choice of the energy function as @xmath201 with density of state @xmath202 for the monkey class urn model , we found that above the critical density @xmath131 for a temperature , the condensation phenomenon is taken place , and most of the balls falls in an urn with the lowest energy level . as the result , the occupation probability changes its scaling behavior from the exponential @xmath5-law to the @xmath6 power - law in large @xmath1 regime .
we also provided a possible link between our results and macro economy , in particular , wealth differentials .
of course , there might exist the other urn models showing the power - law behavior after the condensation .
in fact , we find such a case in a recent study on the ehrenfest urn model @xcite , in which the occupation probability follows a poisson - law when the condensation occurs . although we provided a piece of evidence to show that the power - law behavior in the occupation probability distribution takes place after the condensation for several restricted cases of the cost function , it is not yet clear whether the condensation always causes the power - law or not . the nature of the link between them will be a central problem to be clarified in future .
therefore , as one of our future studies , it might be important to investigate the universality class of urn models that shows the power - law behavior in the occupation probability beyond the critical point .
we hope that various versions and extensions of the disordered urn model , including backgammon model @xcite , could be applied to research area beyond conventional statistical physics .
one of the authors ( j.i . ) was financially supported by _ grant - in - aid scientific research on priority areas deepening and expansion of statistical mechanical informatics ( dex - smi ) " of the ministry of education , culture , sports , science and technology ( mext ) _ no .
we would like to thank enrico scalas for introducing us their very recent studies @xcite concerning the ehrenfest urn model . | we investigate equilibrium statistical properties of urn models with disorder .
two urn models are proposed ; one belongs to the ehrenfest class , and the other corresponds to the monkey class .
these models are introduced from the view point of the power - law behavior and randomness ; it is clarified that quenched random parameters play an important role in generating power - law behavior .
we evaluate the occupation probability @xmath0 with which an urn has @xmath1 balls by using the concept of statistical physics of disordered systems . in the disordered urn model belonging to the monkey class
, we find that above critical density @xmath2 for a given temperature , condensation phenomenon occurs and the occupation probability changes its scaling behavior from an exponential - law to a heavy tailed power - law in large @xmath1 regime .
we also discuss an interpretation of our results for explaining of macro - economy , in particular , emergence of wealth differentials . |
vaccination programs are an efficient and cost effective method to improve public health . with sufficiently many people vaccinated the population gains herd immunity , meaning the disease can not spread .
timely actions to avoid drops in vaccination coverage are therefore of great importance .
many countries have no registries of timely vaccination uptake information , but rely for example on yearly surveys . in such countries estimations of near real - time vaccination uptake based solely on web data are valuable .
we extend prior work in this area @xcite , which showed that vaccination uptake can be estimated sufficiently accurately from web search data .
our extension consists of a new estimation method that adapts dynamically to temporal fluctuations in the signal ( web search queries in our case ) instead of assuming temporal stationarity as in @xcite .
this contribution is novel within vaccination uptake estimation .
linear models have been used previously to estimate health events , for instance by combining data from multiple sources with an ensemble of decision trees @xcite , or , closer to our work , by using query frequencies for influenza like illness @xcite or vaccination uptake estimation @xcite .
these approaches are designed for stationary time - series analysis , i.e. they assume data is generated by a stationary stochastic process .
our motivation is that vaccination uptake often does not follow stationary seasonal patterns .
external events such as disease outbreaks , suspicion of adverse effects , or temporary vaccine shortages can alter uptake patterns for shorter or longer periods of time .
hence , while historical data is a good estimator in stable periods , as shown in @xcite , we reason that adapting the estimation to any unstability can reduce estimation error .
we experimentally confirm this on all official children vaccines data used in denmark between 2011 - 2016 .
to account for seasonal non - stationarity , we use an online learning method , aggregation algorithm ( aa ) @xcite , designed to automatically reduce estimation error in a changing environment . aa was recently used in time series prediction combined with an ensemble of arima models @xcite .
however , we reason that arima models , or other traditional time series models , are not likely to be sufficient for vaccination uptake estimation in cases where : ( i ) there is more than one data source , e.g. vaccine uptake data and search frequency data , and ( ii ) when the time series data to be estimated are assumed to be unavailable ( near real - time ) . to address these challenges ,
we combine aa with regression trees , motivated by recent research showing that random forests outperform arima models on avian influenza prediction @xcite .
a random forest , i.e. an ensemble of decision trees , is well suited for our problem since it is easy to extend to multiple data sources .
our method works as follows : we initially generate a set of regression trees . for each time
step the ensemble of regression trees is retrained based on the initial set of trees and a weighted sum is used to make the estimation .
aa is used to continuously update the weights of each tree .
each regression tree is trained based on a set of features and training samples .
for each tree a feature set is drawn with replacement from the complete feature set .
training samples are selected based on time - relative indices , where index 0 corresponds to the current time step .
the indices are uniformly drawn with replacement from the interval @xmath0 $ ] , where @xmath1 is a window size .
we use trees with different window sizes to account for stationarity and non - stationarity of the signal .
our adaptive vaccination estimation algorithm is shown in algorithm [ alg1 ] , where @xmath2 is the learning rate , rt a set of @xmath3 regression trees , @xmath4 the amount of initial training data and @xmath5 the vaccination uptake .
rt , @xmath2 , @xmath4 @xmath6 list with weights , initialize to be uniform @xmath7 list with the first @xmath4 training samples @xmath8 list with the first @xmath4 observations of @xmath5 @xmath9 empty list of estimations @xmath10 current time step , starting at @xmath11 @xmath12 receive new observation from data stream train rt@xmath13 $ ] using @xmath14 and @xmath15 @xmath16 \leftarrow$ ] estimation of rt@xmath13 $ ] given @xmath17 @xmath18 \leftarrow$ ] @xmath19\cdot\hat{y}_{temp}[n]$ ] @xmath20 \leftarrow $ ] observed @xmath5 at time @xmath21 @xmath22 \leftarrow w[n]\cdot\exp(-\eta\cdot(\hat{y}_{temp}[n]-y[t])^2))$ ] @xmath6 normalize @xmath23 @xmath24 \leftarrow x_t$ ] @xmath25
to facilitate direct comparison , we evaluate our method on the same data as @xcite : monthly vaccination uptake of all official children vaccines in denmark from january 2011 - june 2016 .
vaccination uptake is defined as the total number of people vaccinated in a month divided by the birth cohort for that month . to estimate vaccination uptake , we use frequencies of web search queries extracted from google trends .
we use the exact same frequencies of single term queries provided by @xcite .
we compare to two baselines : ( 1 ) linear regression with lasso regularization where the hyper - parameter is found using three fold cross - validation on the training data ; ( 2 ) linear regression with elastic net regularization where the two hyper - parameters are also selected using three fold cross - validation .
we also include for reference two upper bounds corresponding to the best score reported in @xcite when using ( i ) only web data , and ( ii ) web data combined with clinical data .
these scores are not theoretical upper bounds , but just the best scores across all methods evaluated in @xcite .
we treat them as performance upper bounds because they do not correspond to any individual method , but to the best score per vaccine across all methods reported in @xcite .
neither baselines or upper bounds account for time - series adaptation , i.e. they all assume data stationarity .
the initial number of training samples , @xmath4 , is set to 24 .
all algorithms are evaluated in a leave - one - out fashion , where all data prior to the data point being estimated is used for training . for our algorithm ( atse )
a parameter search is performed by randomly sampling from the following intervals : window size interval 1 - 46 , number of features derived from vaccination data 0 - 45 , number of features derived from web data 0 - 30 , number of regression trees 500 - 10000 , @xmath2 between 0.001 - 0.25 .
table [ tab : results ] displays the root mean squared error ( rmse ) between the estimated vaccination uptake and the real vaccination uptake for all methods .
our method yields the overall best performance compared to the baselines ( it outperforms all baselines for 8 out of 12 vaccines ) .
our method also outperforms the upper bounds of @xcite ( any of the two ) for 6 vaccines .
this supports our reasoning that adapting the estimation to temporal fluctuations is a better strategy than assuming data stationarity .
our method yields the strongest performance improvements for hpv-1 , hpv-2 , mmr-2(12 ) and ditekipol .
all of these vaccines have temporally irregular uptake patterns , as explained next .
hpv-2 have been subject to a heavy media debate in denmark , with a subsequent drop in vaccinations .
mmr-2(12 ) denotes the second mmr vaccine targeted 12 year - olds .
as children grow , parents are less likely to follow the recommended vaccination schedule and fluctuations correlated with measles outbreaks are observed , thus making the time series less stationary .
lastly , in recent years there have been problems obtaining a sufficient supply of certain ditekipol vaccines in denmark , which might have forced people to postpone the initial vaccination , hence introducing irregularities in the signal . for pcv vaccines
there have been no noted irregularities in their uptake patterns , which explains the slight drop in performance by our method compared to the baselines .
we presented an automatic method for near real time estimation of health events using web search query data .
our method combines an aggregation algorithm ( aa ) to automatically reduce estimation error in changing environments with regression trees .
we applied our method to estimate vaccination uptake in all official danish children vaccines , following @xcite , and showed that our approach overall outperformed strong baselines that assumed data to be temporally regular .
our method was particularly strong estimating uptake for vaccines with known irregularities in their usage , such as hpv-1 , hpv-2 , mmr-2(12 ) and ditekipol .
n. dalum hansen , c. lioma , and k. mlbak .
ensemble learned vaccination uptake prediction using web search queries . in _ proceedings of the 25th acm international on conference on information and knowledge management _ , cikm 16 , pages 19531956 , 2016 .
w. jamil , y. kalnishkan , and h. bouchachia .
aggregation algorithm vs. average for time series prediction . in _ proceedings of the ecml pkdd
2016 workshop on large - scale learning from data streams in evolving environments , streamevolv-2016 _ , 9 , 2016 .
m. santillana , a. t. nguyen , m. dredze , m. j. paul , e. o. nsoesie , and j. s. brownstein . combining search , social media , and traditional data sources to improve influenza surveillance .
, 11(10):e1004513 , 2015 . | estimating vaccination uptake is an integral part of ensuring public health . it was recently shown that vaccination uptake can be estimated automatically from web data , instead of slowly collected clinical records or population surveys @xcite .
all prior work in this area assumes that features of vaccination uptake collected from the web are temporally regular .
we present the first ever method to remove this assumption from vaccination uptake estimation : our method dynamically adapts to temporal fluctuations in time series web data used to estimate vaccination uptake .
we show our method to outperform the state of the art compared to competitive baselines that use not only web data but also curated clinical data .
this performance improvement is more pronounced for vaccines whose uptake has been irregular due to negative media attention ( hpv-1 and hpv-2 ) , problems in vaccine supply ( ditekipol ) , and targeted at children of 12 years old ( whose vaccination is more irregular compared to younger children ) . |
the phenomenon of liquid helium ( lhe ) scintillation due to passage of charged particles was discovered in the late 1950 s @xcite .
since then rather extensive studies have been conducted , motivated both by its intrinsic interest ( including an interest in illuminating the behavior of ions and neutrals in superfluid helium ) and by the application of liquid helium as a particle detector @xcite .
( for a brief review of early work , see ref .
@xcite . a more recent review
can be found in the introduction of ref .
@xcite . )
the following picture has emerged from these studies as the process for liquid helium scintillation production ( see e.g. refs .
@xcite ) : when a charged particle passes through liquid helium , it deposits energy to the medium , part of which goes to ionization , creating electron - ion pairs along its track .
the deposited energy also goes to promoting atoms to excited states .
the @xmath2 value , the average energy required to produce one electron - ion pair , in liquid helium is @xmath3 ev @xcite .
electrons and ions then thermalize with the liquid helium . the electron subsequently forms a `` bubble '' in the liquid , pushing away surrounding helium atoms as a consequence of pauli exclusion . the he@xmath4 ion , on the other hand , forms a `` helium snowball '' attracting surrounding helium atoms . the bubbles and snowballs then recombine quickly ( @xmath5 s ) forming excited helium molecules ( excimers ) .
the molecules are formed in the triplet and singlet states .
the lowest - energy singlet state molecule , he@xmath6(a@xmath7 ) , radiatively decays in less than 10@xmath8 s to the ( unbound ) ground state , emitting a @xmath9 ev extreme - ultraviolet ( euv ) photon , contributing to the prompt component of scintillation .
excited atoms in singlet states also contribute to the prompt scintillation signal .
the triplet state molecule , he@xmath6(a@xmath10 ) , on the other hand , has a lifetime of @xmath11 s in liquid helium @xcite . in a high excitation density environment , however , the triplet state excimers can also be destroyed through the penning ionization process @xmath12 or @xmath13 if a singlet excimer is formed as a result of penning ionization after the first @xmath14 s , then it makes a contribution to the delayed scintillation component .
the study of liquid helium scintillation has been conducted using radioactive sources , such as @xmath15 , conversion electron , and @xmath0 sources as well as electron beams .
there were also experiments that studied liquid helium scintillation induced by the products of the @xmath16he(@xmath17,@xmath18)@xmath16h reaction @xcite .
various characteristics of the scintillation light depend on the type of ionizing particles employed . when a @xmath15 or conversion electron source is used , the scintillation light yield is known not to show a strong temperature dependence , whereas when the scintillation is induced by @xmath0 particles , the scintillation yield does exhibit a temperature dependence .
in addition , a difference in time dependence of the scintillation signal has been observed between @xmath0-induced scintillation and @xmath15-induced scintillation @xcite : @xmath0-induced scintillation light shows a long lived component that decays with a @xmath19 dependence following the prompt pulse .
this slow component shows up as a series of single photoelectron ( pe ) pulses following the prompt pulse in an experiment in which the euv scintillation is wavelength - shifted and detected using a photomultiplier ( pmt ) .
( thus these single pe pulses are called `` afterpulses '' . ) a typical pmt oscilloscope trace from our experiment described in this paper is shown in fig . [
fig : pmtsignal ] .
one distinct difference between @xmath0 tracks and electron tracks is the ionization density .
for example , a 5.5-mev @xmath0 particle deposits energy into superfluid liquid helium ( @xmath20 g / cm@xmath16 ) at a rate of @xmath22 ev/@xmath23 m on average along its 0.27 mm long track ev/@xmath23 m and occurs when the remaining kinetic energy is 0.65 mev . ] , whereas for a 500 kev electron , the average energy deposition rate is 40 ev/@xmath23 m @xcite . with @xmath24
ev , it follows that ionization events are separated by an average distance of @xmath25 nm and @xmath26 nm in @xmath0 and electron tracks , respectively . with an average separation of several tens of nanometers between the electron and its parent ion @xcite , ionization events in @xmath0 tracks overlap with each other whereas electron - ion pairs are well separated in electron tracks . as we will discuss later ,
many of the differences in the characteristics between @xmath0-induced scintillation and electron - induced scintillation can be attributed to the difference in the ionization density . although the amount of data on lhe scintillation induced by protons and tritons from the @xmath16he(@xmath17,@xmath18)@xmath16h reaction is rather limited
, it shows qualitatively similar characteristics to @xmath0-induced scintillation because of the high ionization density of the proton and triton tracks . in establishing the picture described above , a study of scintillation in the presence of an electric field played an important role .
the effect of an electric field on the lhe scintillation was first studied by hereford and moss @xcite . in this study
, the authors measured both the intensity of @xmath0-particle - induced scintillation and the ionization current extracted from the @xmath0 tracks in the presence of an electric field ( 9 to 43 kv / cm ) in the temperature range of 1.23 to 4.2 k. the scintillation signal was integrated over @xmath27 @xmath23s .
they observed a decrease in the scintillation intensity and a corresponding increase in the ionization current with an increasing electric field particle . ] .
we have extended their work and have studied the effect of an application of an electric field on both the prompt scintillation and afterpulses in the temperature range of 0.2 k to 1.1 k with an electric field up to @xmath28 kv / cm .
we measured the time dependence of the scintillation signal up to 14 @xmath23s .
the research reported here is directed primarily towards an understanding of the physical processes underlying the electric field dependence of the scintillation yield of liquid helium .
it provides information on the dynamics of the charged particles generated by the passage of an alpha particle through helium and the interaction of molecular species produced on recombination .
however , the work is motivated by the desire to use helium scintillation in the presence of a high electric field in the nedm experiment @xcite .
the nedm experiment , currently being developed to be mounted at the spallation neutron source at the oak ridge national laboratory , will search for the neutron electric dipole moment ( edm ) using a method proposed by golub and lamoreaux @xcite . in this experiment ,
spin polarized ultra cold neutrons will be produced and stored in a volume filled with superfluid helium , to which spin polarized @xmath16he atoms will be added as cohabiting magnetometer .
lhe scintillation produced by the reaction products of the spin - dependent @xmath16he(@xmath17,@xmath18)@xmath16h reaction will be used to analyze the spin of the neutrons .
the signature of a non - zero edm is a shift in the neutron precession frequency upon application of an electric field .
therefore , the effects of an electric field on lhe scintillation produced by heavy particles such as protons , tritons , and @xmath0 particles are of particular interest . an earlier study performed by part of the nedm collaboration on lhe scintillation produced by the products of neutron capture on @xmath16he
was reported in ref .
the experiment was performed using a leiden cryogenics model cf-600 cryogen - free dilution refrigerator @xcite , which had a measured cooling power of 1.8 mw at 0.2 k. a measurement cell , which housed the electrodes immersed in lhe , was mounted on the mixing chamber of the dilution refrigerator ( dr ) .
an @xmath29am @xmath0 source was electroplated on the ground electrode .
@xmath0 particles from the source induced scintillation light in the volume of lhe between the electrodes .
the euv scintillation light was converted to blue light by a wavelength shifter , then was guided through a light guide to the viewport at the bottom of the measurement cell and to a pmt mounted outside the measurement cell in vacuum . a description of each component is given below .
a schematic of the measurement cell and the pmt is shown in fig .
[ fig : apparatus ] .
the measurement cell was made of a stainless steel cross with 4 - 5/8 and 4 - 3/4 conflat flanges attached to its ends .
the volume of the lhe inside the cell was approximately 600 ml . at the top of the cell was a heat exchanger made of stack of gold - coated copper plates , which provided cooling from the mixing chamber to the lhe inside the cell .
the electrodes were made of stainless steel .
the ground electrode was 19.05 mm in diameter and 6.35 mm in thickness , and had an edge rounded with a radius of 3.175 mm .
the high voltage ( hv ) electrode was 19.05 mm in diameter and 12.7 mm in thickness , and had an edge rounded with a radius of 6.35 mm .
the gap between the electrodes was approximately 4 mm .
the hv electrode was mounted on a commercially made hv vacuum feedthrough .
the hv was supplied by a glassman high voltage inc .
model eh30r3 power supply .
a thin wall stainless steel tubing was used as the hv wire inside the cryostat to reduce the heat load . a hv resistor ( @xmath30 g@xmath31 or @xmath30 m@xmath31 ) needed to be inserted in series in the hv cable , in order to reduce the noise on the pmt signal induced by the ripple of the hv power supply for the hv electrode .
the capacitance of the hv cable and this resistor formed a low pass filter , which reduced the effect of the ripple to a sufficiently low level .
approximately 300 bq of @xmath29am was electroplated on the ground electrode .
the activity had a diameter of 6 mm . because the range of 5.5 mev @xmath0 particles in superfluid liquid helium is only @xmath32 mm , the ionization due to @xmath0 particles occurred in a region very close to the surface of the ground electrode .
the electric field at the location of the ionization was uniform to 5% .
larger electrodes and a smaller gap would have given a more uniform electric field at the location of the ionization .
however , that would have , at the same time , significantly reduced the fraction of euv photons detected and hence the detection efficiency by limiting the solid angle subtended at the location of the ionization by the wavelength - shifter - coated surface .
the size and shape of the electrodes were chosen as a compromise between electric field uniformity and the detection efficiency .
the electrodes were enclosed in a sleeve made of g10 .
this was to prevent possible hv breakdowns due to the hv electrode seeing structures on the inner walls of the stainless steel measurement cell .
a sapphire viewport with a view diameter of 49.3 mm was mounted at the bottom of the measurement cell .
an acrylic light guide with a diameter of 49.3 mm and a length of 30.5 mm was mounted between the viewport and the g10 sleeve .
the g10 sleeve had a 48 mm diameter hole on the side to allow the scintillation light to reach the light guide .
the top surface of the light guide was coated with polystyrene doped with tetraphenyl butadiene ( tpb ) , which is known to have good euv - visible conversion properties ( see ref .
@xcite and references therein ) .
a hamamatsu r7725 pmt , modified for cryogenic use with a pt underlayer @xcite , was mounted underneath the viewport in vacuum .
a fraction of the @xmath0 particle induced euv light emitted from the region between the electrodes was converted to blue light at the top surface of the light guide .
the blue light was then guided towards the pmt through the light guide , transmitted through the sapphire viewport and then detected by the pmt .
the fraction of the euv light that reached the tpb coated surface of the light guide was about 5% ( determined by the solid angle subtended by the coated surface of the light guide at the location of scintillation events ) .
the overall detection efficiency of the apparatus (= the number of detected pes per emitted euv photon ) was estimated to be @xmath33 from the solid angle mentioned above ( 5% ) , the euv to visible conversion efficiency ( @xmath34% @xcite ) , the estimated transport efficiency of the visible light ( @xmath35% ) , and the pmt s quantum efficiency ( 18% @xcite ) . the observed number of pes in the prompt pulse in the absence of an electric field @xmath36 ( see sec . [
sec : adcanalysis ] ) was consistent with the expectation based on the known scintillation efficiency (= the number of euv photons in the prompt pulse per energy deposition ) @xcite and the estimated detection efficiency .
the pmt was thermally anchored to the 3 k plate of the dr , which was cooled by a pulse - tube refrigerator , in order to reduce heat load , both radiative and conductive , to the measurement cell .
the base circuit adopted the split design described in ref .
@xcite with some modifications .
the resistor chain was heat - sunk on the 3 k plate . in order to reduce the heat load from the joule heating of the resistor chain , a resistor value 10 times larger than the manufacturer recommended value was used .
because of the low event rate ( @xmath37 s@xmath38 ) , this did not affect the performance of the pmt .
the rest of the base circuit , which consisted mostly of capacitors , was mounted at the back of the pmt .
previously , this pmt model had been demonstrated to function properly at temperatures down to @xmath39 k @xcite .
the present work showed that it functions properly at @xmath40 k. there were two ruo@xmath6 temperature sensors mounted on the g10 sleeve to monitor the temperature of the lhe inside the measurement cell .
there were two lhe level sensors inside the measurement cell .
they were each made of two concentric cylinders that formed a capacitor .
one was mounted in such way that it surrounded the heat exchanger and the other surrounded the light guide , monitoring the lhe level at the top and the bottom of the measurement cell .
in addition there was a lhe pressure sensor monitoring the lhe pressure inside the measurement cell .
a silicon diode sensor was mounted on the pmt as well as on the pmt heat shield to monitor the temperature of the pmt and the heat shield .
a schematic of the helium gas / liquid flow in the system is shown in fig .
[ fig : flowdiagram ] . in filling the system , clean helium gas taken from boiloff from a dewar was further cleaned by a ln@xmath6 cold trap and then was introduced into the system .
the helium was first condensed by a liquefier ( a box filled with sintered metal ) mounted at the 3 k plate , then flowed through a capillary tube with an i d of 0.6 mm and a length of 1.8 m which was thermally anchored at the still , the 50 mk plate , and at the mixing chamber plate to the measurement cell .
the cell was filled with the dr running , which kept the cell temperature at @xmath41 k during the fill .
it took approximately 24 hours to fill the 600 ml cell in this way .
data on the pmt signal were collected using a camac and nim based data acquisition ( daq ) system .
the schematic of the daq system is shown in fig .
[ fig : daqdiagram ] .
the trigger was generated by a discriminator whose threshold was set to a level corresponding to a fraction of a pe , which allowed us to monitor the pmt gain using single pe events from the dark noise . for each event , the size of the prompt signal was recorded using a lecroy 2249a charge - sensitive analog - to - digital converter ( adc ) with a 80 ns - long gate .
the length of the gate was chosen to capture all of the prompt part of the signal ( see fig . [
fig : pmtsignal ] ) .
the pmt signal was amplified and then was split into two , one of which was further amplified with a gain of @xmath42 before being fed to the adc . the high - gain signal ( adc ch1 in fig .
[ fig : daqdiagram ] ) was used to monitor the pmt gain using the single pe events from the dark noise . the low - gain signal ( adc ch0 in fig .
[ fig : daqdiagram ] ) was used to measure the size of the signal generated by @xmath0 particles .
randomly generated pulses were mixed into the trigger in order to monitor the adc pedestal . a typical low - gain adc spectrum is shown in fig . [ fig : adc ] .
the intensity of the delayed component was measured by recording the timing of each afterpulse ( see fig .
[ fig : pmtsignal ] ) with respect to the prompt pulse using a lecroy 2277 multi - hit tdc .
this tdc can record up to 16 hits within a 16 @xmath23s time window .
typical time spectra thus obtained are shown in fig .
[ fig : tdc ] for two temperatures .
also a typical distribution of the number of afterpulses per event is shown in fig .
[ fig : tdc_mult ] .
s is due to afterpulses caused by helium contamination in the pmt.,width=302 ] the cell was filled with lhe following the procedure described in sec .
[ sec : gashandling ] .
once the cell was filled and a desirable measurement cell temperature was reached , the prompt scintillation yield and the time spectrum of the afterpulses were measured as the function of the applied hv .
when increasing or decreasing the hv , the pmt was kept turned off to protect the pmt from being exposed to bright flashes of light due to hv breakdowns , which occurred occasionally when changing the hv setting ; we had found that exposing the pmt to flashes of light from hv breakdowns decreased the pmt gain and the change appeared to be permanent .
the hv electrode was biased with a negative voltage . for each hv and temperature
setting , data were acquired for approximately 5 minutes , which corresponds to @xmath43 prompt scintillation events .
each of such a continuous data taking period is called a `` run '' .
data were taken at five different temperature settings .
the average temperature and the 1-@xmath44 variation in temperature for each of these settings were @xmath45 mk , @xmath46 mk , @xmath47 mk , @xmath48 mk , and @xmath49 mk . in the remainder of this paper , for conciseness
we refer to these temperatures by their `` nominal '' values : 0.2 k , 0.3 k , 0.4 k , 1.0 k , and 1.1 k. in addition , data were taken at temperatures up to 2 k with no hv applied .
the gap in the data between 0.4 k and 1.0 k was due to a thermal short that developed in the system after the data were taken for 0.4 k , 0.2 k , and 0.3 k. the thermal short did not allow the system to run below 1.0 k. as mentioned earlier , a hv resistor was inserted in the hv supply line to suppress noise due to the ripple on the hv .
most of the data were taken with a 1 g@xmath31 resistor .
the results did not change when a 1 m@xmath31 resistor was used , confirming that the resistance between the two electrodes was significantly larger than 1 g@xmath31 and the voltage across the electrode gap was indeed what was supplied by the hv power supply .
the effect of the electric field on the prompt scintillation yield manifests itself as a shift in the @xmath0 peak in the low - gain adc spectrum . in order to remove effects due to possible changes in the pmt gain
, the horizontal scale of the spectrum needs to be properly calibrated in terms of the number of pes . as noted earlier , the gain of the pmt was continuously monitored throughout the experiment by the high - gain adc that recorded single pe events from the dark noise . in order to calibrate the horizontal scale of the low - gain adc spectrum using the pmt gain information obtained by the high - gain adc , it is necessary to determine the gain of the amplifier .
this was achieved by determining the slope of the correlation between the high - gain and low - gain adc spectra .
determination of the pmt gain and the amplifier gain was performed for each run .
the resulting low - gain adc spectrum was fit to determine the location of the @xmath0 peak using the following model function ( a poisson distribution representing the distribution of the number of pes convoluted with a gaussian distribution representing the pmt response ) : @xmath50 where @xmath51 is the overall normalization , @xmath52 a variable corresponding to the adc channel , @xmath23 the mean number of photoelectrons , and @xmath53 the gain of the system ( the adc channel corresponding to one pe ) .
@xmath54 is given by @xmath55 where @xmath56 is the width of the response function of the pmt for single pe events and @xmath57 is the width of the adc pedestal . @xmath56 and
@xmath57 were determined by fitting to the single pe and pedestal peaks respectively , prior to fitting eq .
( [ eq : adcfit ] ) to the @xmath0 peak . @xmath51 and
@xmath23 were varied in the @xmath0 peak fit .
the shape of the @xmath0 peak in the low - gain adc spectrum is well described by the model function in eq .
( [ eq : adcfit ] ) ( see fig . [
fig : adcfit ] ) .
peak in the low - gain adc spectrum fit with the model function in eq .
( [ eq : adcfit]).,width=340 ] figure [ fig : pevstemp ] shows the dependence of the mean number of pes observed in the prompt pulse ( @xmath58 ) on the temperature with no electric field . shown in fig .
[ fig : pevshv ] is @xmath58 plotted against the strength of the electric field for different temperatures .
the prompt scintillation yield decreases by about 15% with an electric field of 45 kv / cm .
the effect of the electric field on the prompt scintillation has little temperature dependence . ) vs the temperature.,width=340 ] ) vs the electric field strength.,width=340 ] our data shown in figs .
[ fig : pevstemp ] and [ fig : pevshv ] are qualitatively consistent with what was observed in the earlier work by hereford and coworkers @xcite .
( compare , for example , fig .
[ fig : pevstemp ] of this work with fig . 2 of ref .
@xcite and fig .
[ fig : pevshv ] of this work with fig . 3 of ref .
@xcite . ) note , however , that a direct comparison between our data and the data in refs .
@xcite and @xcite can not be made because of the difference in the integration time for the prompt pulse .
the mean number of afterpulses per prompt pulse observed in the first 14 @xmath23s was determined by fitting a poisson distribution to the observed distribution of the number of afterpulses per event ( fig .
[ fig : tdc_mult ] ) .
the mean number of afterpulses per prompt pulse thus obtained ( @xmath59 ) is plotted against the temperature in fig .
[ fig : apvst ] . our results , showing a reduction in the afterpulse intensity at lowered temperatures , are consistent with refs .
@xcite and @xcite .
( although ref .
@xcite only shows a plot of the `` pulse height '' which is the signal integrated for the first 1 @xmath23s , and the total intensity plotted against the temperature , the temperature dependence of the afterpulse intensity can be inferred from the data . )
s ( @xmath59 ) plotted as a function of the temperature for zero electric field.,width=340 ] figures [ fig : apvse ] and [ fig : apperpe ] show the mean number of the afterpulses ( @xmath59 ) and the mean number of the afterpulses normalized to the mean number of pes observed in the prompt pulse ( @xmath60 ) plotted against the strength of the electric field , respectively .
we see from the figures that the afterpulse intensity decreases with an increasing electric field strength ( fig . [ fig : apvse ] ) and that the afterpulse intensity is more strongly affected by the electric field than the prompt scintillation ( fig .
[ fig : apperpe ] ) . also , in contrast to the prompt scintillation , the effect of the electric field on the afterpulse intensity exhibits some temperature dependence .
s ( @xmath59 ) plotted against the strength of the electric field.,width=340 ] ) plotted against the strength of the electric field.,width=340 ]
what we have observed from our data can be summarized as follows : + 1 ) the prompt scintillation yield is reduced by 15% at @xmath28 kv / cm .
+ 2 ) the reduction in the prompt scintillation yield is approximately linear in the strength of the electric field .
+ 3 ) the effect of the electric field on the prompt scintillation yield has a very weak temperature dependence in the temperature range of 0.2 k to 1.1 k. + 4 ) the electric field has a stronger effect on the intensity of the delayed component than it has on the intensity of the prompt pulse .
+ 5 ) the observed temperature dependence of the prompt and afterpulse intensities at zero electric field is in qualitative agreement with previous work .
+ below we discuss the implication of these observations .
the prompt scintillation signal is the result of the radiative decay of singlet he@xmath61 atoms in excited states and of he@xmath62 excimers also in singlet states .
an @xmath0 particle produces these excimers and atoms either by ionization followed by recombination or by promoting an atom directly to an excited state without ionization .
the ratio of the number of direct excitations to ionizations has been predicted in helium to be 0.45:1 @xcite .
( although this calculation was performed with 100 kev electrons as the primary particle , the expectation is that it holds for @xmath0 particles as well @xcite . ) among the atoms promoted directly to excited states , the ratio of singlets to triplets was calculated to be 5:1 . on the other hand
, the ratio of singlets to triplets formed by recombination of ionized atoms is expected to be the statistical value of 1:3 for @xmath0 particles ( not the case when electrons are the primary ionizing particle @xcite ) .
scintillation from atoms in excited states that have previously not undergone ionization is not affected by an electric field whereas the number of those that have been ionized is decreased in an electric field because of charge separation .
were there no other complicating factors , the electric field - dependent fraction , @xmath52 , of the prompt scintillation would be @xmath63 .
however , a number of phenomena can influence this fraction , most likely increasing its value .
\1 ) the prompt signal for @xmath0 particles is substantially quenched @xcite by the nonradiative destruction of singlet excimers by the penning process [ eq . ( [ eq : penning2 ] ) ] but by the same mechanism it may be slightly enhanced with the destruction of triplet excimers and the subsequent recombination of the resulting ions into singlet states .
excited state singlet atoms may also nonradiatively decay upon interaction with other species of excitations .
( nonradiative destruction of singlet species is discussed in sec .
[ sec : quenching ] . )
\2 ) singlet helium atoms in excited states with principal quantum number , @xmath17 , of 3 or greater can autoionize by the hornbeck - molnar process @xcite @xmath64 since the binding energy of he@xmath65 is @xmath25 ev , which is greater than the energy to ionize a he(@xmath66 ) atom .
based on the oscillator strengths for the transitions between the ground state and the various excited states of helium @xcite slightly more than one third of the atoms promoted to excited states will have a principal quantum number of 3 or greater , the other two thirds having @xmath67 .
if all atoms with @xmath66 were to undergo autoionization , then the electric field - dependent fraction of the prompt scintillation would be increased to @xmath68 .
various theoretical models have been developed that describe electron - ion recombination in an ionization track in fluid media : jaffe s columnar theory of recombination @xcite describes the case in which a dense plasma of positive and negative ions is formed along the ionization tracks while onsager s theory @xcite describes the case where each electron - ion pair is spatially separated and recombination occurs between an electron and its parent ion ( geminate recombination ) . for the highly ionizing track of an @xmath0 particle in lhe , the jaffe model would appear the more applicable . in the columnar theory ,
the rate of recombination is governed by the following equations : @xmath69 and @xmath70 where @xmath71 and @xmath72 are the densities of the snowballs and bubbles , @xmath73 and @xmath74 are the mobilities , @xmath75 and @xmath76 are the diffusion coefficients , and @xmath0 is the recombination coefficient .
what this model describes are clouds of negative and position ions being pulled away from each other by the electric field while at the same time recombining at a rate proportional to the product of their densities .
also , the clouds expand radially due to the diffusive motion of the charges . both the reduction in the scintillation yield and the generation of an ionization current result from the suppression of recombination due to the presence of an electric field .
the motion and recombination of electrons and ions described by these equations occur on a time scale of @xmath5 s after an @xmath0 particle is stopped in helium .
the processes associated with the decay and destruction of excimers created on recombination , processes that are not influenced by the electric field and occur on time scales of @xmath77 to @xmath78 s , are discussed separately in later sections .
in looking for a solution to these rate equations jaffe assumed the diffusion terms were larger than those associated with recombination and hence , along with the electric field term , were the principal cause of the change in densities of the charges .
this is a valid approach for ionization in gases . on the other hand , kramers @xcite , in attempting to explain the electric field dependence of the ionization currents from @xmath0 particles stopped in liquid helium measured by gerritsen @xcite , pointed out that in a dense fluid the diffusivity of the ions is small and that recombination has a much larger influence on the time dependence of the charge density .
gerritsen s data for the electric field dependence of the ionization current can be fit approximately by the kramers theory with a cylindrical gaussian charge distribution having the same density and spatial distribution for @xmath71 and @xmath72 @xmath79 where @xmath80 is the initial number of positive ion snowballs or electron bubbles per unit length along the track .
the diffusion coefficient was assumed to be zero .
the relationship between the recombination coefficient and the mobilities is given by the langevin relation ( see below ) .
the best fit of kramers theory to gerritsen s data with the one adjustable parameter , @xmath81 nm , is shown in fig .
[ fig : gerritsen ] .
the ionization current is normalized to the saturation current.,width=340 ] several comments regarding this result are relevant to understanding the scintillation and ionization currents of lhe .
\1 ) the ionization current data contains contributions that are not included in the jaffe - kramers theory , which describes the ionization current due to charge carriers that escaped the initial recombination particle . ] .
however , at electric field strengths below that corresponding to the saturation current , charge carriers that initially recombine can contribute to the ionization current through processes such as penning ionization [ eqs .
( [ eq : penning1 ] ) and ( [ eq : penning2 ] ) ] .
a crude estimate indicates that such contributions could be larger than 10% .
\2 ) the langevin relation between recombination coefficient and mobility is @xmath82 . in the approximation that the mobilities of the two species are the same
, the electric field dependence of the current is independent of mobility , as it enters the recombination and electric field terms in the same manner . at the temperature within the track of approximately 2 k ( see below ) , the mobilities of the positive and negative ions are approximately the same @xcite , having the value of @xmath83 m@xmath84/(v@xmath85s ) .
hence the recombination coefficient is @xmath86 m@xmath16/s .
\3 ) from the einstein relation @xmath87 , the diffusion coefficient is calculated to be @xmath88 m@xmath84/s . since the initial charge densities along the track are the order of @xmath89 m@xmath90 , diffusion plays little role in affecting the density distribution at early times when recombination is dominant . at longer times
when the concentrations have decreased substantially , diffusion can influence the charge separation .
\4 ) the assumption that the positive ions and the electron bubbles initially have the same spatial distribution is unlikely to be correct .
the ionization events from the @xmath0 particle or from secondary electrons with energy sufficient to create further ionizations occur in a small cylinder about the track of less than 10 nm .
he ions do not move appreciably in the time it takes for the electrons to thermalize and form a bubble .
however , once the energy of secondary electrons drops below 20 ev , the first excited state of he , the only process by which they can lose energy is elastic scattering from helium atoms . since the fraction of energy an electron loses in such a collision is very small , being dependent on the ratio of the mass of the electron to that of the atom , the order of @xmath91 collisions are required to decrease an initial energy of 10 ev to below 0.1 ev thought to be necessary for bubble formation . from the cross section for elastic scattering , which varies from @xmath92 cm@xmath84 between 1 and 10 ev , and the liquid density ,
the mean free path is roughly @xmath94 nm and the mean distance for diffusion , proportional to the square root of the number of scatters , is the order of 100 nm @xcite .
but as mu@xmath95 oz _ et al . _
@xcite point out , if the initial positive and negative charge distributions are gaussian with very different radii , then the radial field arising from the different charge distributions can be large , easily in the range of 100 to 1000 kv / cm for an @xmath0 track in helium assuming 10 nm for the radius of the positive ions and 100 nm for the electrons .
the two distinct distributions will therefore rapidly merge to form a more uniform distribution assumed in the jaffe and kramers theories .
the positive ion snowballs having a mobility comparable to that of the electron bubble at the initial temperature of @xmath25 k along the track ( see below ) , will expand outward and the electrons contract inward .
\5 ) a significant fraction of the secondary electron produced by an @xmath0 particle have energies above the he ionization energy of 24.6 ev and are therefore capable of creating further ionization . at low energies the probability that a secondary electron will have a particular energy @xmath96 decreases relatively slowly with increasing energy , varying as @xmath97 @xcite .
since the cross section for ionization or excitation by an electron in the range of 25 to several hundred ev is the order of @xmath98 cm@xmath84 , these ionizations / excitations would occur a considerable distance from the track were it not for elastic scattering . the elastic scattering cross section for an 100 ev electron is @xmath99 cm@xmath84 so that energetic electrons will undergo many random scatters , on average remaining in the vicinity of the @xmath0 track .
such electron may have the effect of modifying the charge distribution but not to the extent that a gaussian distribution is not a reasonable approximation .
the measured electric field dependence of the prompt scintillation can be fit well using the kramers theory for a range of different gaussian widths , @xmath100 .
the value of @xmath100 depends on the choice of the fraction , @xmath52 , of scintillation resulting from species created by ionization , which can be affected by the electric field .
the results are shown in fig .
[ fig : promptfit ] .
figure [ fig : b_vs_x ] shows the @xmath101 per degree of freedom of the fit of kramers theory to the prompt scintillation data plotted against @xmath52 and @xmath100 . if @xmath52 is taken to be 0.95 then the best fit is for @xmath102 nm , while for @xmath103 , @xmath104 nm . in the kramers theory
@xcite , the fraction of ions that recombine depends on a single parameter @xmath105 . that is , increasing the radius of the track @xmath100 has the same effect as increasing the electric field @xmath96 . from this
it is expected that a smaller @xmath52 would require a larger @xmath100 .
our results ( figs . [
fig : promptfit ] and [ fig : b_vs_x ] ) are consistent with this expectation . .
the curves are calculated using eq.(9 ) of ref .
@xcite.,width=340 ] per degree of freedom of the fit of kramers theory to the prompt scintillation data plotted against @xmath52 and @xmath100.,width=302 ] a few comments about the relationship between ionization current and the effect of an electric field on the prompt scintillation intensity are appropriate .
these two types of data reflect two different aspects of the same phenomenon , and as such it should be possible to construct a model that describe both data consistently .
however , such an attempt is complicated by a number of factors .
1 ) as mentioned earlier , the ionization current has a contribution from charge carriers that initially recombine .
2 ) there is uncertainty in the value of @xmath52 .
3 ) even with a short integration time ( in our case @xmath106 ns ) , there is inevitably a contribution to the prompt signal from the delayed component , which has a different electric field dependence ( see sec iv.c ) .
because of these complications we have not attempted a unified analysis of our data and gerritsen s ionization current data . while we note that , as seen in figs .
[ fig : promptfit ] and [ fig : b_vs_x ] , there is considerable range in the values of @xmath52 and @xmath100 that fit the data , the choice of @xmath107 and @xmath108 nm is consistent both with our estimate of the fraction @xmath52 and the best fit of kramers theory to gerritsen s data . as seen in fig .
[ fig : promptfit ] , the effect of the electric field on the prompt scintillation has little temperature dependence .
this is attributed to three causes : 1 ) the effect of the electric field on recombination is nearly independent of the mobility of the ions as discussed earlier .
2 ) the initial temperature of the track is in the vicinity of 2 k , irrespective of the temperature of the bulk liquid ( see below ) , resulting in the same mobility of ions in the track irrespective of the bulk liquid temperature .
3 ) the characteristic time of recombination @xmath109 , where @xmath110 is the initial ionization density , is shorter than the time it takes the @xmath0 track to thermalize with the bulk liquid @xmath111 , where @xmath112 is the sound velocity ( @xmath113 m / s ) . indeed , with @xmath114 nm , which gives @xmath115 m@xmath90 , the recombination time is @xmath116 s as compared to the thermalization time @xmath117
s. the fraction of energy of an @xmath0 particle that is quickly down converted to excitations of the superfluid , phonons and rotons , is sufficient to heat a column of liquid a few tens of nanometers radius along the track to the vicinity of 2 k irrespective of how low the temperature of the bulk liquid @xcite . since the density of liquid helium changes by less that 0.5% below 2 k , the initial environment in which the electrons , ions , excimers and excited state atoms reside , on their creation , is independent of the temperature of the bulk liquid
. therefore , the temperature dependence of the prompt scintillation , illustrated in fig .
[ fig : pevstemp ] , must result from the manner in which those entities responsible for the scintillation are influenced by the expansion of the roton / phonon cloud ( with velocity of @xmath118 m / s ) taking place at times the order of @xmath119 s. below a temperature of 0.6 k the density of thermal excitations in the bulk liquid is so low that the hot columnar cloud expands into essentially a vacuum .
but at higher temperatures the thermal excitations in the bulk become sufficiently dense to scatter and slow the radially expanding cloud . as discussed below
, the @xmath19 component of the afterpulse intensity also decreases with decreasing temperature .
the temperature dependence of both the prompt and afterpulse signals is attributed to the more rapid decrease of the triplet excimer density due to expansion , leading to a reduced production of singlets via the penning process .
hereford and coworkers @xcite , in modeling the temperature dependence of the scintillation they observed , argued that the increased diffusion at low temperatures results increased quenching , that is , the enhanced nonradiative destruction of the singlet species .
however , their argument was not informed by the fact that the roton and phonon densities in the vicinity of the track is essentially independent of the temperature of the bulk liquid . following the work of mckinsey _ et al . _
@xcite the time dependence of the afterpulse scintillation for times between 0.4 @xmath23s and 14 @xmath23s has been characterized using the function @xmath120 @xmath121 , @xmath122 , @xmath123 , and @xmath124 were varied in the fit . for convenience , we define @xmath125 , @xmath126 , and @xmath127 to be the contributions to the number of afterpulses from the first , second , and third terms in eq .
[ eq : apspectrum ] , respectively , that is @xmath128 @xmath129 and @xmath130 @xmath125 , @xmath126 , and @xmath127 are plotted as a function of the electric field for different temperatures in figs .
[ fig : apexp ] , [ fig : aptinv ] , and [ fig : apconst ] respectively . in figs .
[ fig : apexp ] and [ fig : aptinv ] , the 1.0 k data points for electric fields between 10 kv / cm and 20 kv / cm seem to show a larger scatter .
we are not able to offer an explanation .
the value of @xmath121 that we obtained was @xmath131 @xmath23s , slightly shorter than the value quoted in ref . @xcite .
our results show that it is nearly independent of the temperature and the electric field . also in fig .
[ fig : atinvvst ] , @xmath126 is plotted as a function of the temperature for zero electric field .
, the number of afterpulses from the exponential component ( @xmath132 ) in the afterpulse time spectrum plotted against the strength of the electric field.,width=340 ] , the number of afterpulses from the @xmath133 component in the afterpulse time spectrum plotted against the strength of the electric field.,width=340 ] , the number of afterpulses from the constant component in the afterpulse time spectrum plotted against the strength of the electric field.,width=340 ] , the number of afterpulses from the @xmath133 component in the afterpulse time spectrum plotted as a function of the temperature for zero electric field.,width=340 ] at times longer than @xmath14 s all singlet state excimers and excited atoms ( except , perhaps , for the first excited state he(2@xmath134s ) , which in vacuum has a lifetime of 19 ms @xcite ) , will have long decayed .
any photons produced at a later time must result from triplet states .
mckinsey _ et al . _
@xcite hypothesized that the afterpulse term dependent exponentially on time was the result of he(2@xmath134s ) atoms forming excimers in a@xmath7 states from which they radiatively decay to the dissociated ground state of the helium pair .
given that the lifetime of the triplet state he(2@xmath16s ) has been determined in the liquid to be 15 @xmath23s @xcite , this appears to be a reasonable hypothesis .
furthermore , there appears to be little or no temperature dependence to the exponentially time - dependent term , as expected for such a source . on the other hand
, one would not expect that a signal arising from a neutral atom to exhibit a dependence on electric field as we have observed ( see fig .
[ fig : apexp ] ) , unless it originated through recombination . since the probability that ions form excimers prior to recombination is high , atomic he(2@xmath134s ) appears more likely to be produced by decay from a higher lying he(n@xmath134p ) state created de novo by electromagnetic excitation from the ground state than through a cascade from an atom formed by recombination .
we are unable to offer an explanation of this exponentially time - dependent component of the afterpulse scintillation .
keto _ et al . _
@xcite , using a beam of 160 kev electrons as the primary ionizing source , measured the long - time afterglow in the infrared spectra associated with a number of transitions between excited states of excimers and atoms to have a time dependence that varied as @xmath19 .
they ascribed this time dependence to the production of the excited states by the penning ionization of triplet excimers , since the bimolecular process , @xmath135 results in @xmath17 proportional to @xmath19 for @xmath136 .
a more detailed description of the time dependence of the singlet excimer density is expressed by eq .
( [ eq : quench ] ) in the next section .
a similar relation can be given for the triplet density , the essentials of which for the purposes of this discussion are contained in eq .
( [ eq : birate ] ) . the problem of describing the time dependence of the afterpulse scintillation from an @xmath0 particle differs from that in the work of keto _ et al .
_ @xcite since in one case the density of excimers is uniform whereas in the other the excimer cloud is expanding about the @xmath0 track .
king and voltz @xcite developed a theoretical model of afterpulses that includes the spatial evolution of the excimer density with time .
the description involves the kinetics of the diffusion of the triplets at times well after the prompt scintillation has decayed . if the triplets maintain a gaussian distribution as they expand by diffusion , then at long times the afterpulsing decreases approximately inversely with time .
et al . _
@xcite cited this model by way of explaining the origin of the @xmath19 component of the afterpulses . at shorter times
the king model predicts a time variation that is considerably faster than @xmath19 , but it does not provide an explanation of the exponential dependence on time , as experimentally observed .
the amplitude of the @xmath19 afterpulse component exhibits interesting temperature and electric field dependences .
firstly , the number of afterpulses varies much more with temperature than the number of photons in the prompt scintillation .
secondly , the number of afterpulses decreases more rapidly with increasing electric field than the prompt signal . as with the slight variation of the prompt signal with temperature below 1 k
, the rapid variation of the @xmath19 afterpulse component with temperature can be understood qualitatively as being the result of the change in propagation of the triplet excimers away from the @xmath0 track .
while the temperature within the core of the track is not sensitive to the ambient temperature of the liquid , the thermal excitations in the surrounding liquid can affect the rate at which the excimers expand away from the track . at high temperatures , the excimers diffuse slowly because of scattering from phonons and rotons . at low temperatures
the excimers propagate ballistically into the cold liquid in the absence of scattering .
the excimer density drops much more rapidly at low temperatures and as a consequence the rate of generation of singlet species by the penning process is strongly temperature dependent .
the strong electric field dependence of the @xmath19 component of the afterpulse results from the two separate ionization steps that are involved in the production of these photons .
firstly , the initial creation of triplet excimers by recombination is affected by the electric field in the same manner as is the production of singlet excimers responsible for the prompt signal .
secondly , the recombination of the electron - ion pair produced by the penning process involving the triplet excimers is also influenced by the electric field .
those recombinations that result in singlet species are responsible for the @xmath19 component .
since the delayed generation of the electron - ion pair occurs in isolation from other charges , the recombination is geminate .
the onsager theory of geminate recombination @xcite is not an appropriate approach to calculating the effect of electric field on the separation of charges created via penning ionization in liquid helium at low temperature .
that theory assumes diffusion to be the dominant process affecting the charge motion in the presence of the field . at low field and high temperatures where the ion mobilities are low
the motion of the charges is governed by viscous drag . in this case
the charges move along field lines , and whether two charges recombine or not is dependent only on the magnitude and direction of their initial separation with respect to an applied electric field . if @xmath137 the initial separation and @xmath138 is the angle between @xmath137 and the applied electric field @xmath96 , the charges will recombine if @xmath139 geminate recombination in lhe under this condition was studied in ref .
@xcite using a @xmath15 source at 1.5 k. however , at 1 k where the mobility of ions is the order of @xmath140 m@xmath84/v@xmath85s the inertial term in the equation of motion becomes important at modest fields . in the limit of high mobility charges
will recombine if @xmath141 independent of orientation of the charges with respect to direction of the field .
if this expression were used in conjunction with a gaussian distribution of charge separations to fit the field dependence of the @xmath19 component of the afterpulse signal , the width @xmath100 of the distribution would be unreasonably small , much less than the 60 nm determined for the distribution of electrons about an alpha track .
since the average energy of electrons created by the penning process responsible for the @xmath19 afterpulse component is comparable to the average energy of secondary electrons from an alpha , the inclusion of the inertial term in describing the ion motion does not provide a satisfactory explanation to the field dependence . at low temperatures and at zero pressure both positive ion snowballs and electron bubbles moving in liquid helium create quantized vortex rings @xcite when their velocities reach the order of 60 m / s .
since ion mobilities are the order of @xmath140 m@xmath142v@xmath85s , at 1 k , fields of less than 1 kv / cm are sufficient for vortex creation . at low fields the charges remain attached to the rings that they have produced , but in very high fields , greater than 25 kv / cm at low temperatures , it has been found @xcite that electrons no longer remain attached to vortices but create and shed vortex rings as they move through the superfluid . we have not attempted to apply the theory of vortex creation by moving ions @xcite to explain the field dependence of the @xmath19 component of the afterpulses . the energy a charge acquires in moving in the field is transferred to the vortex , which increases its diameter and decreases its velocity .
qualitatively , the vortices introduce an enormous drag on the motion of the ions , which has the effect of retarding their separation by the applied field .
presumably vortices are the reason the field dependence of the afterpulses produced by penning ionization is not stronger than observed .
the constant component of the afterpulses , independent of time , is most naturally associated with triplet excimers whose density is constant and are uniformly distributed throughout the volume of the cell , the rate of destruction being balanced by the production from the 300 bq @xmath0 source .
destruction occurs in part by the spontaneous radiative decay of triplet excimers to the dissociated ground state and in part by the penning process .
the most notable features of the data seen in fig .
[ fig : apconst ] are the apparent weak field dependence of this afterpulse component at 1 k and the substantial decrease in intensity as the temperature is lower from 1 k to 0.4 k. these two properties are naturally explained by the scintillation being principally due to the spontaneous radiative decay of the excimers , the @xmath143 state having a lifetime of 13 s in the bulk @xcite . while the electric field is not expected to affect this process
, it does lead to a decreased recombination and density of triplets with increasing field .
below about 0.6 k the mean free path of the excimers becomes long , comparable to the dimensions of the helium cell , so that nonradiative destruction of the excimers on solid surfaces decreases the photon yield at low temperatures .
one observation remains unexplained .
hereford and coworkers @xcite noted that the total scintillation below 0.5 k depended on the geometry of their cell , the smaller cell having somewhat greater light output .
the scintillation intensity decreased with decreasing temperature to a minimum at 0.4 k and then increased on going from 0.4 k to 0.2 k. we have observed similar behavior of the time - independent component of the afterpulse signal at low temperatures ( see figs .
[ fig : apvst ] and [ fig : atinvvst ] ) .
as discussed in the introduction , in the nedm experiment , lhe scintillation produced by the products of the neutron capture reaction on @xmath16he will be used as the neutron spin analyzer . in this reaction , the @xmath144 value of 760 kev is shared between the two reaction products , a proton and a triton ; the proton and the triton are emitted with kinetic energies of 570 kev and 190 kev respectively .
the range of a 570-kev proton in superfluid helium is 0.06 mm , corresponding to an average energy deposition rate of @xmath145 ev/@xmath23 m @xcite .
a 190-kev triton travels 0.018 mm in superfluid helium depositing energy at a rate of @xmath146 ev/@xmath23 m before it comes to rest @xcite . in both cases ,
the average energy deposition rate and hence the number of ionizations per unit length along the track ( @xmath80 ) are almost exactly half of that for a 5.5-mev @xmath0 particle ( @xmath147 ev/@xmath23 m ) .
the radius of the ionization track @xmath100 is expected to be the same for @xmath0 particles and the products of neutron capture reaction on @xmath16he because the track radius is determined by the diffusion of the electrons .
therefore , the ionization density inside the tracks created by the reaction products of the @xmath16he(@xmath17,@xmath18)@xmath16h reaction is also almost exactly half of that in the track created by a 5.5-mev @xmath0 particle .
the lower number of ionizations per unit length along the track ( and hence the lower ionization density ) causes lhe scintillation produced by the products of the @xmath16he(@xmath17,@xmath18)@xmath16h reaction to differ from @xmath0-induced lhe scintillation in two aspects : 1 ) the lower ionization density results in reduced quenching , thereby a larger fraction of the deposited energy being emitted as prompt scintillation compared to @xmath0-induced scintillation .
2 ) as noted earlier in sec .
[ sec : promptdiscussion ] , in the kramers theory the fraction of ions that recombine depends on a single parameter @xmath105 .
therefore , the lower number of ionizations per unit length along the track means that the effect of an electric field is larger .
the effect of the quenching by the nonradiative destruction of singlet species by the penning process can be described by the following equation : @xmath148 where @xmath149 and @xmath150 are the densities of the singlet and triplet species , respectively , @xmath151 is the coefficient for bimolecular decay involving singlet species , @xmath152 is the coefficient for bimolecular decay of the triplet species , @xmath153 is the diffusion coefficient of the singlet species , and @xmath121 is the radiative lifetime of the singlet species .
@xmath154 , @xmath155 , and @xmath156 are numerical factors that come from the fact that in the penning process for every two excimers ( or excited atoms ) destroyed , a new one is formed , one - quarter of the time in the singlet state and three - quarters of the time in the triplet state .
the diffusion term can be neglected for the conditions we are interested in , namely , the first @xmath78 s of the hot track . also the term describing the penning ionization of two triplet states can be neglected since @xmath151 is sufficiently larger than @xmath152 ( see below ) . noting that @xmath150 changes with time more slowly than @xmath149
, we solve eq .
( [ eq : quench ] ) for two limiting cases : 1 ) @xmath150 changes with time keeping the @xmath157 ratio constant , that is @xmath158 , where @xmath159 is a time independent constant .
2 ) @xmath150 does not vary with time , that is @xmath160 , where @xmath161 is the initial triplet density . solving eq .
( [ eq : quench ] ) yields an analytical solution of the same form for both cases . integrating the last term of the right hand side of eq .
( [ eq : quench ] ) over time gives the fraction , @xmath162 , of the singlets that contribute to the prompt scintillation escaping the bimolecular annihilation .
@xmath163 where @xmath164 is the initial singlet density and @xmath165 . for the case in which @xmath158 , @xmath166 and @xmath167 are given by @xmath168 and @xmath169 . for the case
where @xmath170 , @xmath166 and @xmath167 are given by @xmath171 and @xmath172 .
for @xmath0-induced lhe scintillation , the fraction of deposited energy that is emitted as prompt scintillation was found to be 10% @xcite . on the other hand , following a similar discussion to the one in the beginning of sec .
[ sec : promptdiscussion ] , we expect 23% of the deposited energy to be emitted as prompt scintillation in the absence of quenching , giving @xmath173 for @xmath0-induced lhe scintillation .
solving eq .
( [ eq : fs ] ) for @xmath174 gives @xmath175 for @xmath0-induced scintillation .
the numerical value of @xmath176 thus obtained allows us to make a crude estimate of the coefficient for bimolecular decay of the singlet species . for @xmath177 , with @xmath178 cm@xmath90 and @xmath180 s , @xmath175 gives @xmath181 cm@xmath16/s for the singlet excimers .
this value for @xmath151 is about an order of magnitude larger than the value measured for the coefficient for bimolecular decay of the triplet excimers @xcite .
the difference could be attributed to the suppression of this process for the triplets states due to the electron spin flip required for the triplet excimer to go to the dissociated ground state .
similar differences have been observed in the deexcitation rates of singlet and triplet states of helium atoms when encountered by other atoms and compounds @xcite .
assuming that the quenching fraction is the same for excimers and excited atoms , setting @xmath182 in eq .
( [ eq : fs ] ) yields @xmath183 for lhe scintillation induced by the @xmath16he(@xmath17,@xmath18)@xmath16h reaction .
it follows that the fraction of deposited energy emitted as prompt scintillation is expected to be 13% for lhe scintillation produced by the products of the @xmath16he(@xmath17,@xmath18)@xmath16h .
note that this result is somewhat independent of the details of the model .
as long as the differential equation for @xmath149 has the form @xmath184 , eq .
( [ eq : fs ] ) holds and we obtain the same value for @xmath162 for lhe scintillation induced by the @xmath16he(@xmath17,@xmath18)@xmath16h reaction . with the model for quenching discussed above and kramers theory ,
we can make a prediction on the number of the prompt euv photons emitted when a neutron is captured by a @xmath16he atom in superfluid lhe . in fig .
[ fig : neuv_3he_capture ] the predicted number of the prompt euv photons for @xmath16he(@xmath17,@xmath18)@xmath16h is plotted as a function of the electric field .
he(@xmath17,@xmath18)@xmath16h reaction with @xmath185 and @xmath104 nm.,width=340 ] habicht @xcite in his thesis reports measurements of the scintillation resulting from a number of different ionization sources including the @xmath16he(@xmath17,@xmath18)@xmath16h reaction in liquid helium at zero field and 1.8 k. however , it is difficult to make a comparison with his results given the lack of specificity of the experimental parameters such as geometry , solid angle , etc .
the prompt scintillation signal from @xmath0 particles stopped in helium exhibits a 15% reduction in an electric field of 45 kv / cm .
this field dependence is consistent with the current versus field measurements of gerritsen @xcite and kramers analysis of columnar recombination .
we conclude using kramers theory that roughly 40% of the scintillation results from species formed from atoms originally promoted to excited states by the @xmath0 particle and 60% from excimers created by ionization and subsequent recombination , with the electrons initially having a cylindrical gaussian distribution about the @xmath0 track of 60 nm .
the delayed scintillation signal , the time dependence of which is decomposed in the manner suggested by mckinsey _
et al . _
@xcite , exhibits stronger field and temperature dependences than does the prompt scintillation .
the stronger field dependence is the consequence of the fact that the slow component of afterpulses are from triplet excimers undergoing the penning ionization process followed by recombination forming singlet excimers . as such
, it receives the effect of the electric field twice , once at the initial recombination producing triplet excimers , and once at the recombination following the penning process .
the temperature dependence involves the diffusion of the excimers away from the @xmath0 track into the surrounding bulk liquid .
the authors are grateful for the help and advice provided by the following individuals at various stages of this work : e. bond , m. fanning , s. currie , d. g. haase , g. frossati , r. golub , w. guo , e. korobkina , h .- o .
meyer , c. nelson , j. self , b. lozowski , t. rinckel , p. childress , d. baxter , w. m. snow , and a. edwards .
this work was supported by the us department of energy , the national science foundation , and the doe mie no.71re .
g. archibald , j. boissevain , r. golub , c. r. gould , m. e. mayden , e. korobkina , w. s. wilburn , and j. zhou , in _ proceedings of the 24th international conference on low temperature physics ( lt24 ) _ , edited by y. takano , s. p. hershfield , s. o. hill , p. j. hirschfeld , and a. m. goldman , aip conf .
proc . * 850 * , 143 ( 2006 ) .
m. j. berger , j. s. coursey , m. a. zucker , and j. chang , _
estar , pstar , and astar : computer program for calculating stopping - power and range tables for electrons , protons and helium ions _ , national institute of standards and technology , gaithersburg ,
md ( 2005 ) , http://physics.nist.gov/star [ originally published as : m. j. berger , nistir 4999 , national institute of standards and technology , gaithersburg md ( 1993 ) ] . | we report a study of the intensity and time dependence of scintillation produced by weak @xmath0-particle sources in superfluid helium in the presence of an electric field ( @xmath1 kv / cm ) in the temperature range of 0.2 k to 1.1 k at the saturated vapor pressure .
both the prompt and the delayed components of the scintillation exhibit a reduction in intensity with the application of an electric field .
the reduction in the intensity of the prompt component is well approximated by a linear dependence on the electric field strength with a reduction of 15% at 45 kv / cm .
when analyzed using the kramers theory of columnar recombination , this electric field dependence leads to the conclusion that roughly 40% of the scintillation results from species formed from atoms originally promoted to excited states and 60% from excimers created by ionization and subsequent recombination with the charges initially having a cylindrical gaussian distribution about the @xmath0 track of 60 nm radius .
the intensity of the delayed component of the scintillation has a stronger dependence on the electric field strength and on temperature .
the implications of these data on the mechanisms affecting scintillation in liquid helium are discussed . |
knowledge of the mass distribution within clusters , ( also in relation to the distributions of the different cluster components ) , gives important clues about the formation process of the clusters and of the galaxies in them , as well as on the nature of dark matter .
there have been many studies of the mass distribution in galaxy systems over the last decade , stimulated by the discovery of the universal nfw mass density profile of dark matter haloes by navarro et al .
( 1996 , 1997 ) .
a cluster mass distribution can be derived in three ways : 1 ) through the gravitational lensing of distant objects , 2 ) using the spatial distribution and temperature profile of the x - ray emitting , intra - cluster ( ic hereafter ) gas , and 3 ) through the kinematics and spatial distribution of tracer particles moving in the cluster potential . lensing mostly works for clusters at intermediate and large redshifts , and only few nearby cluster lenses are known ( see e.g. cypriano et al .
x - ray observations only sample the inner ( see , e.g. , pratt & arnaud 2002 ) or , at best , the virialized ( neumann 2005 ) cluster regions . the third method is particularly suited for studying the mass profiles of relatively nearby galaxy clusters , out to large radii , well beyond the virialized region ( see , e.g. , reisenegger et al .
2000 ) . additionally , using galaxy distributions it is also possible to constrain the velocity anisotropies , and hence the orbits , of cluster galaxies ( see , e.g. , biviano & katgert 2004 , bk04 hereafter ) .
in this paper , i review recent results on cluster mass profiles as obtained from the analysis of the spatial and velocity distributions of cluster member galaxies ( see biviano 2002 for another recent review on this topic ) .
in order to determine a cluster mass profile , @xmath0 , using the projected phase - space distribution of its member galaxies we can use the jeans analysis ( see , e.g. , binney & tremaine 1987 , bt87 hereafter ) , or the caustic method recently introduced by diaferio & geller ( 1997 ; see also diaferio 1999 ) . in the jeans analysis the observable projected phase - space distribution of galaxies is related to the cluster @xmath0 , using the abel and jeans equations ( eqs . 4 - 55 , 4 - 57 , and 4 - 58 in bt97 , for the simplified case of a stable , non - rotating , spherically symmetric system ) , through knowledge of the velocity anisotropy profile , @xmath1 , which characterizes the orbits of cluster galaxies . in the caustic method , one infers @xmath0 of a given cluster from the amplitude of the caustics in the plane of line - of - sight velocities vs. projected clustercentric distances .
formally , the caustic amplitude is related to the gravitational potential through a function @xmath2 of the potential itself , and of @xmath1 , ( see eqs . 9 and 10 in diaferio 1999 ) .
numerical simulations indicate @xmath3 , but only at radii larger than the cluster virial radius , @xmath4 ( see fig . 2 in diaferio 1999 ) .
hence , for @xmath5 the caustic mass estimate is not very accurate . on the other hand ,
since the method does not rely on the assumption of dynamical equilibrium , it is a very powerful tool to constrain @xmath0 at large radii .
note that both with the jeans and the caustic method one samples the _ total , _ not the _ dark _ mass of a cluster .
this must be taken into consideration when observational results are compared with results from numerical simulations of collsionless dark matter haloes . using x - ray data it is possible to subtract the baryonic component of the total mass and infer the dark mass distribution ( see okas & mamon 2003 , lm03 hereafter ) , but
this has not been done very often .
a fundamental problem of the jeans analysis ( and , to a lesser extent , also of the caustic method ) is the mass orbit degeneracy , i.e. the solution obtained for @xmath6 is degenerate with respect to the solution obtained for @xmath1 . in order to break the degeneracy
, @xmath1 must be constrained independently from @xmath0 .
this can be achieved via the analysis of the _ shape _ of the galaxy velocity distribution , that contains the required information about the orbital anisotropy of cluster galaxies ( merritt 1987 ; van der marel et al .
2000 , vdm00 hereafter ; lm03 ) . alternatively ,
if several tracers of the gravitational potential are available , the jeans equation can be solved for @xmath0 independently for each of the tracers , thus restricting the range of possible solutions ( bk04 ) .
another relevant problem in the jeans analysis occurs if the cluster is not in steady state and dynamical equilibrium .
since clusters grow by accretion of field galaxies ( e.g. moss & dickens 1977 ) , they are not steady - state systems . formally , inclusion of the time derivative in the jeans equation ( bt87 , eq .
4 - 29c ) is then needed . on the other hand ,
the fractional mass infall rate is estimated to be quite negligible for nearby clusters ( ellingson et al .
moreover , most of the mass is accreted in big , discrete clumps ( zabludoff & franx 1993 ) . hence
, those clusters undergoing substantial mass accretion can be identified through the presence of substructures in their galaxies distribution , and excluded from the analysis ( vdm00 ; biviano & girardi 2003 ) . the problem gets tougher when one is dealing with small galaxy systems ( galaxy groups ) most of which are probably still in a pre - virialized collapse phase ( giuricin et al .
1988 ) .
interlopers are another potential problem in cluster mass estimates .
cluster velocity dispersion estimates have been shown to be robust vs. modifications of the method of interlopers removal ( girardi et al .
1993 ) , but estimates of the kurtosis of the velocity distribution ( needed to constrain @xmath7 , see lm03 ) are not .
hence , it is advisable to use other , more robust , estimators of the shape of a cluster velocity distribution ( e.g. the gauss - hermite moments , vdm00 ) .
as the number of available redshifts for cluster galaxies increases , the interloper selection procedure becomes more robust .
since in practice spectroscopic samples of more than 100 member galaxies per cluster are rare , quite often a composite cluster is built by stacking together the data for several clusters ( see , e.g. , carlberg et al .
1997 ; vdm00 ; katgert et al .
the procedure is justified if clusters form a homologous family , as suggested by the existence of a fundamental plane of cluster properties ( schaeffer et al .
1993 ; adami et al .
1998 ) . in the jeans analysis it is generally assumed that clusters are spherical and
do not rotate .
the composite cluster is spherically symmetric by construction , and deviation from spherical symmetry is unlikely to be a major problem for individual clusters either ( see vdm00 ; sanchis et al .
2004 ) . while evidence for cluster rotation has been claimed for a couple of clusters ( see , e.g. , biviano et al . 1996
; dupke & bregman 2001 ) , the energy content in the rotational component is marginal .
finally , dynamical friction and galaxy mergers could in principle invalidate the use of the _ collisionless _ jeans equation .
however , the cluster velocity dispersion is too high for galaxy mergers to take place , and the dynamical friction timescale is too long for most cluster galaxies , except for very massive galaxies ( e.g. biviano et al .
these can be removed from the sample ( katgert et al .
2004 ) prior to application of the jeans analysis .
galay mergers in low - velocity dispersion systems , like groups , can however be a critical issue .
since groups contain only a small number of galaxies each , their mass profile can only be inferred by stacking several of them together .
results so far are contradictory .
mahdavi et al .
( 1999 ) used 588 galaxies in 20 groups to conclude that their mass density profile is consistent with a hernquist ( 1990 ) model , and that the mass - to - galaxy number - density profiles ratio is constant .
mahdavi & geller ( 2004 ) used the rasscals sample ( 893 galaxies in 41 nearby groups ) to determine a single power - law mass density profile , @xmath8 over a wide radial range ( 0@xmath9 ) .
finally , carlberg et al . ( 2001 ) analysed @xmath10 galaxies in @xmath11 groups from the cnoc2 survey , at redshifts @xmath120.55 .
they found a cored mass density profile near the groups centres , decreasing outside with a shallow slope , hence implying a steeply increasing mass - to - light ratio with radius . despite these different results ,
all these studies conclude that galaxies in groups have , on average , nearly isotropic orbits , except late - type galaxies that are on mildly radial orbits ( mahdavi et al .
1999 ) .
the coma cluster , with its nearly 900 spectroscopically confirmed cluster members ( adami et al .
2005 ) , is probably the best studied cluster in the universe . leaving apart historical studies ( see biviano 1998 and references therein ) ,
the first modern analysis of the coma cluster mass profile was done by merritt & saha ( 1993 ) .
they found that the density profile is cuspy near the centre , @xmath13 , and decreases as @xmath14 , at large radii , if the orbits are isotropic .
more recently , lm03 found that the nfw profile provides a good fit to the derived mass distribution , but other models are acceptable , and even a cored profile near the centre can not be excluded . at variance with merritt & saha , and unlike most other studies , lm03 analysed the _ dark , _ not the total , mass profile of the cluster . using the kurtosis profile of the galaxies velocity distribution
, they obtained @xmath15 for early - type galaxies , i.e nearly isotropic orbits .
the above results are in agreement with those obtained through the caustic method ( geller et al .
1999 ; rines et al . 2001 ) .
these studies also constrained the behaviour of the mass density profile at very large radii , @xmath16 with @xmath174 .
the isothermal sphere model ( @xmath18 ) is rejected , while both a nfw and a hernquist ( 1990 ) model are acceptable .
the mass - to - light ratio in the @xmath19-band is found to be nearly constant out to @xmath20 . in a series of papers ,
rines et al .
( 2000 , 2003 , 2004 ) have applied the caustic technique to another 8 nearby clusters of the cairns survey .
the results are not very different from those found for the coma cluster .
typically , @xmath21 near the centre , and the best - fit asymptotic slope of the mass density profile at large @xmath22 is either @xmath23 ( nfw ) or @xmath24 ( hernquist 1990 ) .
when the nfw profile is adopted , the best - fit values of the concentration parameter @xmath25 vary between 5 and 17 . using the mass profiles obtained from the caustic technique in the jeans analysis ,
@xmath26 is obtained .
the mass - to - light ratio in the @xmath19 band is approximately constant out to @xmath4 , then decreases by a factor @xmath27 out to the turnaround radius . following up a previous study by carlberg et al .
( 1997 ) , vdm00 stacked together 16 clusters at @xmath280.55 from the cnoc survey , to build a composite cluster with @xmath29 galaxies . through the analysis of the shape of the galaxies velocity distribution , vdm00 were able to constrain @xmath30 . from the jeans analysis they then determined a mass profile which is fully consistent with a @xmath31 nfw model .
the ratio between the mass density profile and the galaxies number density profile was found to be nearly constant .
biviano & girardi ( 2003 ) analysed a composite cluster of 1345 member galaxies , obtained from the stacking of 43 clusters from the 2dfgrs .
they performed a joint jeans and caustic analysis , using the former method to determine the mass profile in the virialized region ( @xmath32 ) , and the latter to extend @xmath0 to @xmath33 .
i.e. they used each of the two methods in the cluster region where it is expected to perform best .
they found that a @xmath34 nfw profile fits the data over the whole radial range explored .
cored profiles are also acceptable but only for core radii @xmath35 .
the ratio of the mass density to the galaxy number density profiles was found to be constant out to @xmath4 , and to decrease beyond that , mostly because of the contribution of late - type galaxies in the external regions . , of 59 clusters from the enacs ( points with 1-@xmath36 error bars ) .
the solid , dot - dashed , and dashed lines are the best - fit nfw , burkert ( 1995 ) , and softened isothermal sphere models , respectively , width=453 ] katgert et al .
( 2004 ) analysed a composite of 1129 galaxies from 59 nearby clusters from the enacs .
@xmath0 was determined in a fully non - parametric way .
only early - type galaxies were used in the analysis .
their velocity distribution was showed to imply nearly isotropic orbits , hence @xmath37 was adopted in the jeans analysis .
the resulting mass density profile has a slope of @xmath38 at @xmath39 , fully consistent with an asymptotic slope of @xmath23 at larger radii .
both a @xmath40 nfw profile , and a burkert ( 1995 ) profile with a rather small core ( @xmath41 ) are acceptable ( see figure 1 ) .
the ratio of the mass - to - luminosity density profiles ( in the @xmath42 band ) is nearly constant out to @xmath43 , when both the cd - like galaxies ( that give an excess luminosity near the centre ) , and the late - type galaxies ( that give an excess luminosity in the outer regions ) are excluded .
the @xmath0 solution obtained for the early - type galaxies was later confirmed by using a different tracer of the gravitational potential , early - type _ spirals _ ( sa sb ; bk04 ) . inverting the jeans equation ( via the method of solanes & salvador - sol 1990 , developed from binney & mamon 1982 )
, bk04 found that also early - type spirals have nearly isotropic orbits in the cluster potential . on the other hand the late spirals ( sbc sd ) plus
irregulars have @xmath26 only near the centre , and increasingly radial orbits outside .
galaxies in substructures were found to have tangential orbits at all radii .
work is in progress ( biviano & salucci in preparation ) to determine the relative distributions of the different mass components , namely the baryonic matter ( in galaxies and in the intra - cluster gas ) , the dark matter in subhaloes , and the diffuse dark matter .
preliminary results indicate that once the baryons and the dark matter subhaloes are subtracted from the total mass profile , the remaining diffuse dark matter profile is still consistent with both a burkert ( 1995 ) and a nfw model , but the concentration is a factor of two higher than for the _ total _ mass profile .
the galaxy baryonic mass profile is found to be similar to the total mass profile , but the total baryonic mass is more widely distributed , because of the dominant contribution by the ic gas .
superclusters are not virialized systems .
it is then impossible to determine their mass profile other than by the use of the caustic method .
reisenegger et al .
( 2000 ) and rines et al .
( 2002 ) applied this method to the shapley and the a2197/a2199 supercluster , respectively .
the superclusters mass profiles are well fit by both a nfw and a hernquist ( 1990 ) model .
useful constraints can be put on the mass distribution within clusters from the analysis of the projected phase - space distribution of their member galaxies .
the constraints are poor near the centre , @xmath44 with @xmath45 , because of an intrinsic finite resolution ( the size of the cd galaxy ) .
cored profiles are allowed , but only for core sizes below this resolution size , @xmath46 .
stronger constraints are obtained at large radii , and they are stronger than those obtained with any other methods , @xmath47 with @xmath48 . in summary , the isothermal sphere is ruled out , while the nfw , burkert ( 1995 ) and hernquist ( 1990 ) models are acceptable .
when comparing a cluster mass density profile with the galaxy number ( or luminosity ) density profiles , it must be realized that the result of this comparison depends upon which cluster galaxy population ( or , almost equivalently , which photometric band ) is considered .
more relevant is the comparison of the total mass distribution with the baryonic mass distribution .
the galaxy baryonic mass profile is similar to the total mass profile , but the total baryonic mass distribution ( including the ic gas ) is more extended . as fas
as the orbits are concerned , ellipticals , lenticulars , and early - type spirals ( sa sb ) are all found to be on nearly isotropic orbits .
late - type spirals ( sbc - sd ) and irregulars have increasingly radial velocity anisotropy with increasing radius , and galaxies in substructures seem to be characterized by tangential orbits .
the implications of these results for the evolution of galaxies in clusters is discussed in bk04 .
very little is known about the evolution of cluster dynamics with redshift . the average mass profile and the orbital characteristics of galaxies of nearby clusters ( enacs , cairns , 2dfgrs ) are similar to those of medium - distant clusters ( cnoc ) .
hence , little or no evolution over @xmath490.5 is implied ( see also girardi & mezzetti 2001 ) . in the near future
it will be possible to use the sdss products ( see e.g. goto 2005 ) to reduce the current uncertainties on the dynamics of nearby clusters .
it is also urgent to improve our understanding of the dynamics of small galaxy systems ( groups ) .
a promising data - set in this respect is the gems groups sample of osmond & ponman ( 2004 ) , for which also x - ray temperatures are available .
these can provide robust scaling parameters for the build - up of a composite group sample . determining the average mass profile ( and galaxy orbits ) of distant clusters ( @xmath50 )
is a more demanding task .
distant clusters are likely to be in a more turbulent phase of their evolution than nearby clusters , making the dynamical analysis even more difficult . as a consequence ,
a deeper sampling of the projected phase - space distribution of member galaxies is needed , with a strong investment in terms of observing time ( e.g. demarco et al . 2005 ) .
adami , c. , biviano , a. , durret , f. , & mazure , a. 2005 , astro - ph/0507542 adami , c. , mazure , a. , biviano , a. , katgert , p. , & rhee , g. 1998 , a&a , 331 , 493 binney , j. , & mamon , g. 1982 , mnras , 200 , 361 binney , j. , & tremaine , s. 1987 , galactic dynamics , princeton univ . press ( bt87 ) biviano , a. 1998 , in untangling coma berenices : a new vision of an old cluster , eds .
mazure , a. , casoli , f. , durret , f. , & gerbal , d. , world scientific , also astro - ph/9711251 , and http://nedwww.ipac.caltech.edu/level5/biviano/frames.html biviano , a. 2002 , in tracing cosmic evolution with galaxy clusters , eds .
borgani , s. , mezzetti , m. , valdarnini , r. , asp conf .
series 268 ; also astro - ph/0110053 biviano , a. , durret , f. , gerbal , d. et al .
1996 , a&a , 311 , 95 biviano , a. & girardi , m. 2003 , apj , 585 , 205 biviano , a. , girardi , m. , giuricin , g. , et al .
1992 , apj , 396 , 35 biviano , a. , & katgert , p. 2004 ,
a&a , 424 , 779 ( bk04 ) burkert , a. 1995 , apj , 447 , l25 carlberg , r.g . ,
yee , h.k.c .
, ellingson , e. , et al .
1997 , apj , 485 , l13 carlberg , r.g . ,
yee , h.k.c .
, morris , s.l .
2001 , apj , 552 , 427 cypriano , e.s . ,
sodr , l.jr . ,
kneib , j .-
p . , & campusano , l.e .
2004 , apj , 613 , 95 demarco , r. , rosati , p. , lidman , c. , et al . 2005 , a&a , 432 , 381 diaferio , a. 1999 , mnras , 309 , 610 diaferio , a. & geller , m.j .
1997 , apj , 481 , 633 dupke , r.a . , &
bregman , j.n .
2001 , apj , 547 , 705 ellingson , e. , lin , h. , yee , h.k.c . & carlberg , r.g .
2001 , apj , 547 , 609 geller , m.j . , diaferio , a. , & kurtz , m.j . 1999 ,
apj , 517 , l23 girardi , m. , biviano , a. , giuricin , g. , et al .
1993 , apj , 404 , 38 girardi , m. & mezzetti , m. 2001 , apj , 548 , 79 giuricin , g. , gondolo , p. , mardirossian , f. , et al .
1988 , a&a , 199 , 85 goto , t. 2005 , mnras , 359 , 1415 hernquist , l. 1990 , apj , 356 , 359 katgert , p. , biviano , a. , & mazure , a. 2004 , apj , 600 , 657 okas , e.l . , & mamon , g.a .
2003 , mnras , 343 , 401 ( lm03 ) mahdavi , a. , & geller , m.j .
2004 , apj , 607 , 202 mahdavi , a. , geller , m.j . , bhringer , h. , et al .
1999 , apj , 518 , 69 merritt , d. 1987 , apj , 313 , 121 merritt , d. & saha , p. 1993
, apj , 409 , 75 moss , c. & dickens , r.j .
1977 , mnras , 178 , 701 navarro , j.f . ,
frenk , c.s .
, & white , s.d.m .
1996 , apj , 462 , 536 navarro , j.f . ,
frenk , c.s .
, & white , s.d.m .
1997 , apj , 490 , 493 neumann , d.m .
2005 , a&a , 439 , 465 osmond , j.p.f . & ponman , t.j .
2004 , mnras , 350 , 1511 pratt , g.w . , & arnaud , m. 2002 , a&a , 394 , 375 reisenegger , a. , quintana , h. , carrasco , e.r . , & maze , j. 2000 , aj , 120 , 523 rines , k. , geller , m.j . ,
diaferio , a. , et al . 2000 ,
aj , 120 , 2338 rines , k. , geller , m.j .
, diaferio , a. , et al .
2002 , aj , 124 , 1266 rines , k. , geller , m.j . , diaferio , a. , et al .
2004 , aj , 128 , 1078 rines , k. , geller , m.j . ,
kurtz , m.j .
, & diaferio , a. 2003 , aj , 126 , 2152 rines , k. , geller , m.j . , kurtz , m.j . , et al . 2001 , apj , 561 , l41 sanchis , t. , okas , e.l . , & mamon , g.a .
2004 , mnras , 347 , 1198 schaeffer , r. , maurogordato , s. , cappi , a. , & bernardeau , f. 1993 , mnras , 263 , l21 solanes , j.m . &
salvador - sol 1990 , a&a , 234 , 93 van der marel , r. , magorrian , j. , carlberg , r.g . , et al .
2000 , aj , 119 , 2038 ( vdm00 ) zabludoff , a.i .
, & franx , m. 1993 , aj , 106 1314 | the mass distribution of galaxy clusters can be determined from the study of the projected phase - space distribution of cluster galaxies .
the main advantage of this method as compared to others , is that it allows determination of cluster mass profiles out to very large radii . here
i review recent analyses and results on this topic .
in particular , i briefly describe the jeans and caustic methods , and the problems one has to face in applying these methods to galaxy systems .
then , i summarize the most recent and important results on the mass distributions of galaxy groups , clusters , and superclusters .
additional covered topics are the relative distributions of the dark and baryonic components , and the orbits of galaxies in clusters . |
the 455 nm laser is frequency - stabilized with a blue detuning of @xmath59 ghz to the @xmath60 transition ( see fig . [
fig : fig2 ] ) .
the 1070 nm laser is scanned over the two - photon resonance to the rydberg state @xmath61 .
both lasers have an estimated linewidth below 5 mhz .
the frequency of the 1070 nm laser is calibrated for each measurement using a fabry - prot interferometer and additionally by an eit - signal @xcite to fix the origin of the frequency axis .
the 455 nm laser typically has a power of 3 mw .
the 1070 nm laser has a power of 15 w and passes through a pockels cell , allowing to switch the power of this laser for the experiment as fast as @xmath62 ns with a repetition rate of 10 khz .
the two - photon rabi frequency is @xmath63 ghz , the detuning to the rydberg state is @xmath64 ghz .
the glass cell is home - made and consists of two 1 mm - thick quartz optical flats of @xmath65 @xmath66 separated by a 220 @xmath3 spacer and sealed at the edge .
a glass tube was connected to the cell , filled with cesium under vacuum and sealed off .
this glass tube serves as a reservoir .
the temperature of the cell is kept constant at @xmath67c to prevent the metallic cesium from condensing , whereas the temperature of the reservoir is varied between @xmath68c and @xmath69c to tune the atom number density . the atom number density in the cell
is determined for each measurement by performing absorption spectroscopy on the d2-line of cesium @xcite .
both beams are linearly polarized , overlapped in a counter - propagating configuration and focused inside the cell to a beam waist of approx .
15 @xmath3 . after passing through the cell the blue beam
is focused on a pinhole in order to select the central part of radius 6.25 @xmath3 inside the cell .
we verified that the resulting imaged volume is indeed almost cylindrical inside the cell .
the change in transmission is detected using a fast amplified silicon photodiode ( femto hsa - x - s-1g4-si ) . during the measurement ,
the transmission change of the 455 nm laser is monitored and averaged 300 times by a fast oscilloscope .
we ensured that we operate in the linear response regime of the photodiode and measured the conversion efficiency of the detection to be approx .
850 v / w .
this allows us to convert the transmission change into the real number of rydberg atoms that are excited , assuming that no decays and losses of rydberg atoms occur .
this assumption is valid for short times as the time spent by an atom inside the excitation volume is on the order of 20 ns . and @xmath70 is driven with the rabi frequency @xmath71 .
the detuning to the intermediate state is @xmath27 .
right panel : approximation to a 2-level system after adiabatic elimination of the intermediate state @xmath72 .
the effective 2-level rabi frequency is @xmath73 . for simplicity reasons
the detuning to state @xmath74 and the arising light shifts are not shown . ]
we reduce our 3-level system to a 2-level system by making use of the adiabatic elimination of the intermediate state .
the basic principles of this approximation and the relevant parameters are shown in fig .
[ fig : figs1 ] .
the density matrix for the 3-level system is @xmath75 , and we define the density matrix for the effective 2-level system as @xmath76 .
the populations of states @xmath77 and @xmath74 coincide between both descriptions , i.e. @xmath78 and @xmath79 .
the approximation relies on assuming that the intermediate state remains unpopulated ( @xmath80 and @xmath81 ) .
the resulting identity from the optical bloch equations without decays is @xmath82 using @xmath83 one obtains @xmath84 which means that the absorption on the lower transition is proportional to the excitation rate to the rydberg state @xmath74 . for the 2-level system we also obtain from the optical bloch equations @xmath85 and therefore @xmath86
this last relation shows the link between the 2-level model and what is experimentally measured . at short inter - atomic distances , i.e. below 1 @xmath3 ,
the pair - state interaction potentials for rydberg s - states become very complex , as shown in fig .
[ fig : figs2 ] for the 32s state .
we will focus on this 32s state , but the situation is similar for other principal quantum numbers .
neighboring @xmath87 pair - states interact with the @xmath88 state with weak but resonant dipole - quadrupole interaction , leading to avoided crossings and state mixing .
this means that the @xmath87 pair - states , which are dipole - forbidden for laser excitation from the @xmath89 state in the non - interacting case , carry some admixture @xmath90 of the @xmath88 state . therefore it is possible to excite pair - states at negative detunings , where the detuning is defined with respect to the unperturbed @xmath88 state , which would not be the case with purely repulsive van - der - waals interaction . versus the inter - atomic distance .
the molecular quantum number here is @xmath91 , and interactions up to the quadrupole - quadrupole order are included @xcite . the green ( resp .
red ) line depicts the extrapolated van - der - waals ( resp .
dipole - dipole ) pair - state potential .
bottom panel : 32s-32s admixture @xmath90 versus the inter - atomic distance at a @xmath92 ghz potential energy ( depicted as the blue dash - dot line in the upper panel ) . ]
these pair state potentials are actually only valid for an excitation close to an atom in the 32s state .
once such a pair is created , it is mainly a @xmath87 pair , with an amplitude of @xmath93 .
for any subsequent excitation around this pair , one therefore has to consider the pair - state interaction potentials for the @xmath94 and @xmath95 pair states . for an s - p pair state the interaction potential
is purely of the form @xmath96 , since the dipole - dipole interaction of an s - p pair with its permutation p - s is resonant , and the symmetric ( resp .
antisymmetric ) linear combination exhibits attractive ( resp .
repulsive ) interaction .
the corresponding interaction strength is @xmath97 @xmath44 , which is in excellent agreement with the value of @xmath98 @xmath44 extracted from the experimental results .
for an s - d pair state the interaction is of van - der - waals character and repulsive ( @xmath99 @xmath42 ) .
the @xmath100 component therefore plays no role in our case of aggregation with red detunings .
the admixture @xmath90 is the rescaling factor for the rabi frequency when exciting a @xmath87 pair . because of the nature of the aggregates , which enclose previously excited rydberg atoms , at most every second rydberg excitation
is created via this process .
the other excitations occur either as a direct off - resonant excitation from the ground state , or close to a @xmath87 pair .
the rabi frequency does not need rescaling in both cases .
an estimate for the overall rescaling factor is therefore @xmath101 , which is close to the rescaling factors of 0.38 ( for van - der - waals interaction ) and 0.23 ( for dipole - dipole interaction ) needed to match the experimental data to the results from the simulation . in the whole treatment
, we neglected the distance dependency of @xmath90 . in order to justify this approximation ,
let us consider just two pair states @xmath102 and @xmath103 with energies @xmath104 and @xmath105 , and a small dipole quadrupole interaction @xmath106 between the two pair states .
the resulting admixture of pair state @xmath102 at small @xmath107 is @xmath108 if we consider @xmath109 to be small , and the pair state energy is @xmath110 .
therefore @xmath111 under these rather crude approximations .
using the same argument as before we obtain @xmath112 which is a sufficiently weak dependence to be neglected . the first contribution to dephasing in the experiment is the doppler effect , characterized by the two - photon doppler width . here
@xmath113 is the atomic mass of cesium , @xmath114 is the temperature and @xmath115 and @xmath116 the wavenumbers of the two lasers , from which we consider the difference because the lasers are counter - propagating . at @xmath67c
we have @xmath117 mhz .
the other important source of dephasing arises from the short transit time during which the atoms are in the resonant shell ( or facilitation region ) of width @xmath118 . in first approximation ,
only the velocity component perpendicular to the shell @xmath119 determines the transit time , as depicted in fig.[fig : figs3 ] .
therefore we define this motional dephasing rate as @xmath120 where @xmath121 is the one dimensional mean atomic velocity .
for a temperature of @xmath67c , this 1d mean velocity is @xmath122 @xmath123 .
since both dephasing mechanisms are gaussian , the total dephasing rate is defined as @xmath124 .
moreover @xmath118 is related to the total dephasing rate and on the interaction potential ( see fig.1 ) .
first assuming rydberg interaction of the van - der - waals ( vdw ) type , we rename the relevant quantities as @xmath125 , @xmath126 ( motional dephasing rate ) and @xmath127 ( total dephasing rate ) . the width of the resonant shell is also given by @xcite @xmath128 by combining equations and we obtain the following self consistent equation for the motional dephasing rate @xmath126 : @xmath129 solving equation yields , so that the total dephasing rate is .
for the case of pure dipole - dipole ( dd superscript ) interaction , equation changes to @xmath130 and the dephasing rate can be estimated to with . @xmath127 and @xmath131 are the values that were used in the simulations . for each atom , within which the excitation is facilitated . the velocity component @xmath132 perpendicular to the resonant shell
is shown for an atom inside the shell .
bottom : probabilities @xmath133 of creating an @xmath51-atom aggregate in an interval @xmath134 and typical number of @xmath51-atom aggregates @xmath135 for @xmath136 facilitating atoms , similarly to the experimental situation . ]
since we are not in the frozen gas regime , aggregates are only spatially correlated like the ones shown in the inset of fig .
1(b ) for a short period of time . at longer times
the atomic motion will destroy the spatial correlations between the atoms . this will , however , have no impact on the fact that rydberg excitations are facilitated by already existing one . in the following
we will estimate the size of the frozen aggregates [ fig .
the time scale over which the ensemble can be considered frozen is given the transit time of the atoms through the resonant shell of width @xmath118 , given by @xmath137 ns . during this time interval
the ensemble can be considered as a frozen gas for our purposes .
we estimate the size of the aggregates that can be formed during this time in a very simplified model .
the parameters are the same as in fig .
2(c ) with the extracted values of @xmath1 and @xmath138 : @xmath28 mhz , @xmath139 mhz , @xmath140 @xmath6 , @xmath41 @xmath42 and @xmath141 mhz .
@xmath142 is the rescaling factor for @xmath24 discussed in the main text .
let us first consider one individual rydberg atom which acts as the seed for an aggregate . in the resonant shell ( sphere of radius @xmath143 @xmath3 and width @xmath144 nm , see eq . ) around this first atom reside @xmath145 atoms in average .
moreover the size of the resonant shell grows with the number of rydberg atoms . for this
we approximate an @xmath51-atom aggregate to a sphere ( whose volume is occupied by @xmath51 rydberg atoms separated by the facilitation radius ) .
then the entire resonant shell contains @xmath146 atoms . for each ground state atom in the shell , the time constant for the excitation is , and therefore the excitation rate of the @xmath147-th atom is @xmath148 . if we simplify the problem and consider that the atoms are excited sequentially , the probabilities @xmath149 that an aggregate formed by @xmath51 rydberg atoms at the time @xmath150 per pre - existing rydberg atom obeys the following rate equation : @xmath151 in the table in figure [ fig : figs3 ] we show the solutions of these equations for aggregates consisting of up to 5 atoms at the time @xmath134 .
the number of @xmath51-atom aggregates that are excited in the whole excitation volume during @xmath134 is given by @xmath152 , where @xmath153 is the number of rydberg atoms which can act as a nucleation grains . as the excitation occur at the boundary of the already excited rydberg ensemble @xmath153 is only a fraction of the @xmath154 rydberg atoms .
the values of @xmath135 are shown in figure [ fig : figs3 ] , with @xmath136 ( corresponding to the outer shell of a sphere with @xmath155 rydberg atoms ) .
for this rydberg atom number the largest spatially correlated aggregates that we create in our experiment can thus be estimated to consist of approximately 4 atoms . | the observation of strongly interacting many - body phenomena in atomic gases typically requires ultracold samples . here
we show that the strong interaction potentials between rydberg atoms enable the observation of many - body effects in an atomic vapor , even at room temperature .
we excite rydberg atoms in cesium vapor and observe in real - time an out - of - equilibrium excitation dynamics that is consistent with an aggregation mechanism .
the experimental observations show qualitative and quantitative agreement with a microscopic theoretical model .
numerical simulations reveal that the strongly correlated growth of the emerging aggregates is reminiscent of soft - matter type systems .
due to their exaggerated properties , rydberg atoms find applications in various research fields ranging from cavity qed @xcite , quantum information @xcite , quantum optics @xcite , microwave sensing @xcite to molecular physics @xcite .
one particular area of research employs the strong interactions between rydberg atoms to create strongly interacting many - body quantum systems for quantum simulation @xcite , quantum phase transitions @xcite and the realization of correlated or spatially ordered states @xcite . in any system ,
gaseous , liquid , glass or solid , spatial correlations can only arise if interactions are present . in rydberg gases these correlations
have recently been revealed by the direct imaging of the resonant excitation blockade effect @xcite . in our experiment an initially nearly ideal gas of thermal atoms at room temperature
is excited into a strongly interacting rydberg state .
the rydberg excitations show correlated many - body dynamics that shares similarities with that of soft - matter systems .
, the atoms are exactly shifted in resonance with the excitation laser , red - detuned by @xmath0 .
the grey shaded area symbolizes the excitation bandwidth , a gaussian whose width is given by the dephasing rate @xmath1 .
( b ) left : typical time evolution of the rydberg population and of the transmission change of the exciting laser [ experimental signal , which is proportional to the derivative of the rydberg density , see text and fig .
[ fig : fig2 ] for details ] during the aggregation process .
right : snapshots of the aggregation at times a , b and c , illustrated for a frozen gas in 2d .
red dots correspond to rydberg atoms .
blue dots correspond to resonant atoms .
the dark grey line shows the ` resonant shell ' in which the excitation of rydberg atoms is facilitated .
inset : sketch of the density - density correlation function expected for the steady state [ corresponding to ( c ) ] . ]
we consider the situation of a dense atomic gas , where dense means that the mean inter - particle distance is smaller than the typical interaction distance .
the interplay between these strong interactions and an off - resonant excitation laser results in a non - equilibrium dynamics with pronounced spatial correlations ( related is the antiblockade in ultracold atoms @xcite ) . at a certain distance @xmath2 from a rydberg atom the interaction shifts the rydberg states in resonance with the detuned excitation light fields , as depicted in fig .
[ fig : fig1](a ) .
the subsequent resonant excitations are restricted to specific distances and therefore result in the formation of so - called aggregates that feature non - trivial spatial correlations @xcite .
this so - called facilitated excitation occurs predominantly at the boundary of aggregates [ see fig . [
fig : fig1](b ) ] , similarly to seeded nucleation .
the number of rydberg excitations grows increasingly faster as the size of the aggregates gets larger .
after a certain time a ( quasi ) steady state is reached and the excitation saturates because the sample is maximally filled with rydberg atoms .
the spatial correlations expected for this steady state are depicted in fig .
[ fig : fig1](b ) .
the density - density correlation function sketched here shows an enhanced probability for finding two atoms separated by the facilitation radius @xmath2 .
there is no apparent long - range order since the isotropic facilitation mechanism does not favor the creation of crystalline arrangements .
the existence of such rydberg aggregates has recently been observed in ultracold systems @xcite . in these studies
the rydberg number distributions were extracted and matched to the results of empirical rate equation models , proving the existence of small aggregates with up to 8 excitations . in this letter
we investigate a complementary regime and study the dynamics of the aggregation process in an atomic vapor at room temperature .
central parameters such as the dephasing rate and the atomic density are up to 2 orders of magnitude larger than in previous work @xcite . to quantify the non - equilibrium dynamics
we perform a systematic study of the characteristic time scale that underlies the aggregation process .
we find that the functional dependence of this time scale on various system parameters can be approximated by power laws .
the corresponding exponents agree well with those obtained from a theoretical aggregation model .
further evidence for the aggregation dynamics is found through a quantitative evaluation of the number of rydberg excitations and the resulting mean inter - particle distance which is consistent with the existence of a facilitation radius .
we perform our experiments in a gas of cesium atoms at room temperature or above .
the cesium atoms are confined in a 220 @xmath3 thick glass cell [ see fig . [
fig : fig2](b ) ] . the atomic density @xmath4 ranges from @xmath5 @xmath6 to @xmath7 @xmath6 , which is large compared to previous experiments on interacting rydberg atoms in thermal vapor @xcite and non - bec cold atomic gases @xcite where @xmath8 @xmath6 .
we excite cesium atoms from the ground state @xmath9 to a rydberg state @xmath10 off - resonantly via the intermediate state @xmath11 [ see fig . [
fig : fig2](a ) ] .
the two - photon transition is red - detuned by @xmath0 with respect to the unperturbed rydberg state .
the lower transition is driven by a cw laser at 455 nm , which also serves as the probing field .
for the upper transition a laser at @xmath12 nm is pulsed on the nanosecond scale using a fast pockels cell .
the effective rabi frequency of the two - photon transition is then given by @xmath13 , where @xmath14 and @xmath15 are the respective single transition rabi frequencies and @xmath16 ghz is the detuning to the intermediate state @xcite .
we measure the transmission change of the lower 455 nm laser after suddenly switching on the upper 1070 nm laser on .
note that this signal is directly proportional to the time derivative of the rydberg population @xcite .
typical transmission traces are shown in fig .
[ fig : fig2](c ) for various atom number densities . immediately after the 1070 nm laser
is switched on ( at @xmath17 ) atoms are transferred to the rydberg state , and for each excited rydberg atom one photon is removed from the probe resulting in a decrease of the transmission .
mhz .
the two - photon rabi frequency is @xmath18 mhz .
the grey - shaded area in the background represents the temporal pulse shape of the infrared laser .
the time delay @xmath19 used for the quantitative analysis is also depicted .
the black curve ( upper panel ) is the expected transmission signal ( not to scale ) for non - interacting atoms ( following the approach from ref .
@xcite ) . ]
one striking feature is the large red detuning of up to @xmath5 ghz at which these signals are observed , in contrast to previous work on coherent excitation in a more dilute gas @xcite .
line broadenings and shifts to the blue have been observed with s - states in cold atoms @xcite , and rydberg aggregates have been observed only at blue detunings @xcite , as expected from the repulsive rydberg - rydberg interaction potential for most rydberg s - states @xcite .
however at large atomic densities , the mean inter - particle distance is so small that the perturbative character of the van - der - waals potential is not valid anymore .
level crossings between pair states now have to be considered . for rydberg s - states in cesium
a strong resonant dipole - quadrupole interaction with neighboring pair - states occurs at @xmath20 @xmath3 for @xmath21 @xcite .
initially dipole - forbidden attractive pair states acquire some admixture at short inter - atomic distances due to the dipole - quadrupole interaction .
therefore the rydberg - rydberg interaction for s - states in cesium also features an effective attractive component , allowing for us to observe an excitation signal at large red detunings . a first indication for the collective nature of the excitation dynamics is provided by the data in fig .
[ fig : fig2 ] . here
we consider the characteristic time scale of the excitation @xmath19 , i.e. the time at which the transmission reaches its minimum [ as shown in fig .
[ fig : fig2](c ) ] . in the non - interacting case
this time scale is given by @xmath22 ns ( black line ) .
in the experiment however the excitation signal is slow ( @xmath23 ) and density dependent , which is only possible if many - body effects are included .
pulse propagation effects can be excluded due to the small optical depth at the large detunings applied . to analyze the dynamics ,
we systematically varied all accessible parameters independently : @xmath0 , @xmath24 , @xmath4 and @xmath25 , where @xmath26 is the effective principal quantum number of the rydberg state , @xmath27 being the quantum defect . only the dephasing rate @xmath1 , determined by the doppler shifts , the velocity of the atoms and the steep slope of the pair - state potentials , can not easily be tuned over a significant range ( for the derivation of the actual dephasing rate , see supplemental material @xcite ) .
the basic experimental settings are the following : @xmath28 mhz , @xmath18 mhz , @xmath29 @xmath6 and @xmath30 ( 32s state ) , and are all but one kept constant .
first , by fitting the values of the time scale of the excitation @xmath19 against @xmath0 we obtain a power law as @xmath31 with @xmath32 . in the same way
, we find that @xmath33 with @xmath34 .
the power law for @xmath4 is @xmath35 with @xmath36 .
finally we performed density scans at three different principal quantum numbers ( 32s , 34s and 36s states ) and were able to extract a scaling behavior as @xmath37 with @xmath38 .
these results are summarized in table [ tab : tab1 ] . by integrating the transmission signal over time
we obtain an estimate of the actual number of rydberg atoms present in the sample . at the time when the rydberg excitation starts to saturate
the rydberg number in the excitation volume reaches approximately 50000 atoms .
the corresponding mean inter - atomic distance is @xmath39 @xmath3 , leading to approximative values for the effective van - der - waals or dipole - dipole strength of interaction as @xmath40 @xcite with @xmath41 @xmath42 and @xmath43 @xmath44 ( at @xmath45 mhz ) .
we use these values as an input for simulations described in the following .
l d d d d & & & n_g ] & n^ * + exp & 1.99(14 ) & -2.10(5 ) & -1.09(6 ) & -4.6(5 ) + vdw & 2.15(1 ) & -2 & -0.86(1 ) & -4.72(5 ) + dd & 2.38(0 ) & -2 & -0.74(1 ) & -2.95(2 ) + we compared these findings to the results of the rydberg aggregation model adapted from ref .
@xcite .
this model describes the dynamics of rydberg aggregation within a system of two - level atoms in the presence of strong dephasing .
this central assumption of strong dephasing applies to the experimental situation discussed here since the motion - induced dephasing is larger than the rabi frequency ( @xmath46 ghz @xcite ) .
the reduction to a two - level system is well justified in our case , although the commonly used adiabatic elimination of the intermediate state typically breaks down for quantitative considerations of the coherent evolution @xcite . here however , because of the large dephasing rates due to the atomic motion we do not observe coherent evolution , and because of the large detunings we expect only less than @xmath47 population of the intermediate state .
we performed simulations for an ensemble of randomly distributed atoms in 3d , assuming a pure van - der - waals ( @xmath48 ) or pure dipole - dipole ( @xmath49 ) attractive interaction potential .
in fact the actual experimental potential curves have a more complex spatial dependence , due to strong state mixing @xcite .
the latter makes the actual microscopic description of the excitation dynamics rather intricate since a two - level description of an atom might not be sufficient for covering all possible pair excitation channels .
therefore the van - der - waals and dipole - dipole potential that are used in our two - level model should be regarded as the limiting approximations of the actual experimental situation .
we evaluate the time scale of the excitation @xmath19 defined by the transmission minimum , as in the experiment .
all results are listed in table [ tab : tab1 ] .
we find an excellent agreement between the experimental and theoretical exponents .
in particular the dependence of @xmath19 on approximately @xmath50 indicates that a strong dephasing effect is present in our system @xcite .
moreover the power laws vary only slightly , independently of whether pure van - der - waals or pure dipole - dipole interaction are used .
our experimental results are compatible with both potentials , suggesting a rather weak dependence on the actual shape of the potential and that the basic mechanism of aggregation only relies on the existence of a facilitation radius . in fig .
[ fig : fig3](a ) [ resp .
( b ) ] we show the transmission change extracted from the model with van - der - waals [ resp .
dipole - dipole ] interaction and in fig .
[ fig : fig3](c ) from the experiment .
the quantitative agreement for the time scale @xmath19 between theory and experiment is excellent . due to the complex pair state potentials and strong state - mixing that are crucial here
, an adaptation of the rabi frequency has to be performed in order to use the two - level model ( i.e. a model with a single rydberg state ) . in our simulations the rabi frequency is reduced by a factor of 0.38 ( for van - der - waals interaction ) and 0.23 ( for dipole - dipole interaction ) compared to the experiment .
these factors are compatible with the @xmath51s-@xmath51s admixture that one can extract from the pair - state potentials @xcite . note that with dipole - dipole interaction there is a dependence on the system size due to the long - range character of the @xmath52 potential , hence the different rescaling factors applied in fig .
[ fig : fig3](a ) and [ fig : fig3](b ) . .
( a ) simulated data with pure van - der - waals interaction .
the rabi frequency was rescaled with a factor @xmath53 .
( b ) simulated data with pure dipole - dipole interaction ( for @xmath54 atoms ) .
the rabi frequency was rescaled with a factor @xmath55 .
( c ) experimental data .
the blue ( resp .
green ) line shows the position of the absorption maximum @xmath19 in the simulated data with van - der - waals ( resp .
dipole - dipole ) interaction . ]
the comparison of the experimental results to those of the model show that facilitated excitation is occurring , which implies that an aggregation dynamics takes place .
the frozen gas approximation which is used to model the aggregation in @xcite holds strictly for time intervals of about 0.3 ns , the transit time of the atoms through the ` resonant shell ' ( see figure [ fig : fig1 ] ) . during this time aggregates consisting of up to 4 atoms are formed @xcite .
their spatial correlations will then quickly decay to those of a non - interacting ideal gas .
note , however , that even after that time the aggregation of excitations continues at the boundary of the uncorrelated aggregates . despite this effect
the agreement between the experimental values of @xmath19 and those obtained from the simple theoretical model is in general good [ see fig . [
fig : fig3 ] ] . yet
the time evolution differs as the experimental signal appears much more narrow in time than in the simulations .
we attribute this discrepancy to the inherent differences between the microscopics of the experimental system and the theoretical model ( e.g. the complex potentials and the atomic motion ) .
interestingly , when considering the temporal width of the signal , the simulations with dipole - dipole interaction seem to agree best with the experiment , which is realistic considering that one component of the actual interaction potentials is of pure dipole - dipole nature .
moreover , the model does not account for the fact that the excitation of isolated , i.e. not facilitated , atoms is in fact coherent .
another feature of the experimental data that is not reproduced by the theory is the small overshoot in transmission just before the system reaches the steady state [ for instance at 20 ns for @xmath56 @xmath6 in fig .
[ fig : fig2](a ) ] .
remnants of coherent rabi oscillations can be ruled out given the large dephasing rates and large detunings with respect to the single - atom resonance .
moreover the time scale of the overshoot , e.g. the time between the two maxima , does not show a specific dependency on the detuning , as is the case for rabi flopping . explaining this feature will require further study and a more sophisticated theory .
note that ion effects can be neglected .
there are no ions in the sample prior to the excitation , otherwise they would act as aggregation seeds @xcite at @xmath57 and change the signal dramatically . moreover plasma formation would occur only once rydberg atoms have been excited @xcite and therefore can not explain our rydberg excitation signal . in spite of the low collisional ionization rate ( @xmath58 mhz @xcite )
it might happen that in the course of the experiment single ions are created .
however single ions will not disturb the aggregation dynamics , because of the similarity between the ion - rydberg and the rydberg - rydberg potentials @xcite . in conclusion ,
we have shown a facilitated excitation process of rydberg atoms in thermal vapor .
we have characterized the dynamics by measuring power laws for all relevant experimental parameters .
these power laws , as well as absolute numbers for the dynamics and rydberg density are in excellent agreement with those from a model for rydberg aggregation .
these results were obtained in the framework of thermal vapors , but the general ideas also apply to experiments in ultracold atoms . in the ultracold regime
it would be interesting to study how quantum effects influence the aggregation time scale when the strength of dephasing noise is systematically lowered .
the authors would like to thank d. peter , s. weber , s. hofferberth and j.p .
garrahan for fruitful discussions , as well as h. kbler for proof reading .
this project was supported by the carl - zeiss - stiftung and the erc under contract number 267100 .
i.l . acknowledges funding from the european research council under the european union s seventh framework programme ( fp/2007 - 2013 ) / erc grant agreement n. 335266 ( escquma ) and the eu - fet grant no .
512862 ( hairs ) .
d.b . and j.p.s .
acknowledge financial support from grants nsf(phy-1205392 ) and darpa(60181-ph - drp ) for this work .
36ifxundefined [ 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.85.1083 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.82.2313 [ * * , ( ) ] link:\doibase 10.1126/science.1217901 [ * * , ( ) ] link:\doibase 10.1038/nature11361 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.110.103001 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.113.053601 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.113.053602 [ * * , ( ) ] link:\doibase 10.1038/nphys2423 [ * * , ( ) ] link:\doibase 10.1038/nature07945 [ * * , ( ) ] link:\doibase 10.1038/nphys1614 [ * * , ( ) ] link:\doibase 10.1126/science.1244843 [ * * , ( ) ] link:\doibase 10.1038/nphys3119 [ * * , ( ) ] link:\doibase 10.1103/physreva.80.033422 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.104.043002 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.105.135301 [ * * , ( ) ] link:\doibase 10.1088/0953 - 4075/44/18/184008 [ * * , ( ) ] link:\doibase 10.1038/nature11596 [ * * , ( ) ] http://arxiv.org/abs/1404.0980 [ ] link:\doibase 10.1103/physreva.88.061406 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.98.023002 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.104.013001 [ * * , ( ) ] link:\doibase 10.1103/physreva.90.011603 [ * * , ( ) ] link:\doibase 10.1088/1751 - 8113/47/48/482001 [ * * , ( ) ] link:\doibase 10.1103/physreva.88.043410 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.112.013002 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.113.023006 [ * * , ( ) ] link:\doibase 10.1088/0022 - 3700/14/21/003 [ *
* , ( ) ] link:\doibase 10.1103/physrevlett.110.123001 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.111.113901 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.99.163601 [ * * , ( ) ] @noop link:\doibase 10.1103/physrevlett.107.243001 [ * * , ( ) ] link:\doibase 10.1088/0953 - 4075/38/2/021 [ * * , ( ) ] link:\doibase 10.1103/physreva.74.020701 [ * * , ( ) ] link:\doibase 10.1088/0022 - 3700/15/2/004 [ * * , ( ) ] @noop link:\doibase 10.1088/0953 - 4075/46/24/245001 [ * * , ( ) ] link:\doibase 10.1088/0953 - 4075/41/15/155004 [ * * , ( ) ] |
galaxy populations were first studied by hubble ( 1926 ) , who developed the familiar tuning fork diagram of ellipticals , spirals and barred spirals .
most bright galaxies can be morphologically classified by their hubble type .
however , many types of galaxy have been found that do nt fit the tuning fork .
these occur both at low redshift and at high redshift where the galaxies can be intrinsically different due to evolution .
some of these galaxies are shown in fig .
1 . the tuning fork can be extended to include these new types of galaxy as shown in fig .
however , morphological classification only tells us what type of galaxies there are .
it does not tell us what proportion are of each type or whether this varies over time .
galaxy formation and evolution is a relatively new field and so far very little is known about what produced the galaxy population we see today . to understand this process
it is first necessary to have some quantitative information about the local galaxy population .
the luminosity is an easy to measure quantity which can be used to classify galaxies .
the luminosity function ( lf ) ( peebles & hauser 1974 ) measures the space density of galaxies as a function of luminosity .
this can be convolved with different evolutionary models and compared to number - magnitude counts ( driver et al .
however , recent surveys have produced a large range in the measured lf , fig .
3 , with the variation between the surveys much greater than random errors .
the systematic errors causing this wide variation must be understood before any progress is going to be made .
disney ( 1976 ) pointed out that surface brightness selection effects are important to take into account when measuring the luminosity function .
phillipps & disney ( 1986 ) and phillipps , davies & disney ( 1990 ) went on to calculate the effects of surface brightness .
these calculations take into account light lost below the isophote , malmquist bias and surface brightness dimming due to cosmological expansion . here
, we will discuss how we have used the `` two degree field galaxy redshift survey '' ( 2dfgrs ) to produce a bivariate brightness distribution ( bbd ) , which corrects for these effects .
then we will describe some of the results and some of the selection effects that we have not corrected for .
we will then briefly describe the millennium galaxy catalogue ( mgc ) , one of the projects in the isaac newton telescope wide field survey ( int - wfs ) .
the mgc is a deep , imaging survey designed to remove some of the remaining selection effects .
the selection effects are then modelled by convolving the bbd with different functions based on the visibility theory of phillipps , davies & disney 1990 ) adopting parameters appropriate for the eso slice project ( esp ) ( zucca et al . 1997 ) , the las campanas redshift survey ( lcrs ) ( lin et al .
1996 ) and the second southern sky redshift survey ( ssrs2 ) ( marzke et al . 1998 ) .
the 2dfgrs is a @xmath0 redshift survey of the north galactic and south galactic polar ( ngp , sgp ) regions .
when it is complete it will have redshifts for 250,000 galaxies with @xmath1 . at present
more than 100,000 redshifts have been measured , making this the largest survey to date by a factor of 4 .
the input catalogue is the apm catalogue ( maddox et al .
1990a , b ) . for more details on
the 2dfgrs see the article in these proceedings by bridges et al .
one method to remove surface brightness selection effects is to construct a bbd .
this is a plot of the space density of galaxies versus the absolute magnitude , and the effective surface brightness .
this is a simple extension of the lf , whereby galaxies of different surface brightnesses are treated separately . in cross
et al . ( 2000 ) , we describe how to construct the bbd for the 2dfgrs .
throughout this work , we make corrections based upon the data as much as possible . the data has been corrected for light lost below the isophote , malmquist bias , redshift incompleteness and clustering . in each case
these corrections were functions of both absolute magnitude and effective surface brightness .
the bbd was produced from 50,000 galaxies in the sgp .
4 shows the final malmquist corrected space density of galaxies as a function of @xmath2 and @xmath3 .
there is a strong luminosity - surface brightness relation .
l-@xmath4 relations are also seen in driver ( 1999 ) , binggeli ( 1993 ) and de jong and lacey ( 2000 ) .
the shaded region on fig .
4b shows where the volume is less than 10,000mpc@xmath5 .
this plot suggests that luminous lsbgs such as malin 1 are extremely rare .
faint hsbgs such as m32 are likely to be rare as they are far from the l-@xmath4 axis .
the luminosity density @xmath6mpc@xmath7 .
the luminosity density is plotted as a function of absolute magnitude and surface brightness in fig .
this shows clearly that the luminosity density is decreasing well before the selection boundaries imposed by the survey .
this implies that we are indeed seeing the majority of the light emitted by local galaxies .
the input catalogue completeness is a complex function of both surface brightness and magnitude .
the curve on fig .
4b shows where the volume , over which a galaxy can be detected , is equal to 10,000mpc@xmath5 . outside this region ,
the detection of galaxies becomes increasingly difficult .
the input catalogue is only complete in regions determined by the detection isophote ( @xmath8mag arcsec@xmath9 ) , the apparent magnitude limit ( @xmath10mag ) and the smallest isophotal radius ( @xmath11 ) .
the smallest isophotal radius depends on the seeing ( @xmath12 for the apm ) .
the apm suffers from less than ideal photometry which is only accurate to @xmath13mag .
the data does not contain information on the light profiles of the galaxies such as disc - bulge separation parameters .
to remove these problems we have undertaken a deep equatorial ccd survey over part of the 2dfgrs northern galactic polar region .
this survey is described in more detail in these proceedings by lemon et al .
it will push the boundaries of the @xmath14 plane that can be studied , giving us a quantitative measure of the population that we may have missed with the 2dfgrs .
it will also improve the photometry of the galaxies we have already observed and allow us to do morphological classification and disc - bulge separation .
fig . 6 shows the schechter function fits to three lfs shown in fig .
these are the lcrs ( lin et al . 1996 ) , the ssrs2 ( marzke et al . 1998 ) and the esp ( zucca et al . 1997 ) .
the lcrs and esp show the largest variation amongst the surveys .
also plotted are three lines showing the lfs produced when the 2dfgrs bbd is convolved with three different visibility functions .
the parameters used for the lcrs are @xmath15b@xmath16 , m@xmath17b , and d@xmath18 , calculated using a b - r colour of 1.5 ( driver et al . 1994 ) .
the parameters used for the esp are @xmath19b@xmath16 , m@xmath20b , and d@xmath18 .
the parameters used for the ssrs2 are @xmath19b@xmath16 , m@xmath21b , and d@xmath18 .
this plot shows that the lfs can be reproduced well at the faint end .
this demonstrates that surface brightness selection effects are a major systematic error in determining luminosity functions .
to understand the effects of evolution , it is essential to have a good quantitative knowledge of the local galaxy population .
the new technologies in this `` new era of wide field astronomy '' allow us to detect larger numbers of galaxies with higher precision .
these new technologies are the multi - fibre spectrometers such as 2df and the large format ccds such as those in the int wide field camera .
these new surveys allow us to properly tackle the selection effects that have dogged our surveys and therefore will allow us to finally pin down the space and luminosity densities of galaxies . | we discuss the quantification of the local galaxy population and the impact of the `` new era of wide - field astronomy '' on this field , and , in particular , systematic errors in the measurement of the luminosity function .
new results from the 2dfgrs are shown in which some of these selection effects have been removed .
we introduce an int - wfs project which will further reduce the selection biases .
we show that there is a correlation between the surface brightness and the luminosity of galaxies and that new technologies are having a big impact on this field .
finally selection criteria from different surveys are modelled and it is shown that some of the major selection effects are surface brightness selection effects . |
in this paper we consider smooth higher dimensional knots , that is , spheres @xmath5 smoothly embedded in @xmath6 . in @xmath6 we have the canonical cubulation @xmath1 by translates of the unit @xmath7-dimensional cube .
we will call the @xmath0-skeleton @xmath3 of this cubulation the _ canonical scaffolding _ of @xmath6 ( see section 2 for precise definitions ) .
we consider the question of whether it is possible to continuously deform the smooth knot by an ambient isotopy so that the deformed knot is contained in the scaffolding .
in particular , a positive answer to this question implies that knots can be embedded as cubic sub - complexes of @xmath6 , which in turn implies the well - known fact that smooth knots can be triangulated by a pl triangulation ( @xcite ) .
the problem of embedding an abstract cubic complex into some skeleton of the canonical cubulation can be traced back to s.p .
a considerable amount of work has been done regarding this problem ( see , for example @xcite ) .
the question is non - trivial ; for instance , among cubic manifolds there are non - combinatorial ones and therefore non - smoothable ones .
also there is a series of very interesting papers by louis funar regarding cubulations of manifolds ( @xcite , @xcite ) .
the possibility of considering a knot as a cubic submanifold contained in the @xmath0-skeleton of the canonical cubulation of @xmath6 has many advantages .
for instance ,
in the important case of classical knots @xmath8 , matveev and polyak @xcite begin the exposition of finite type invariants from the `` cubic '' point of view and show how one can clearly describe invariants such as polynomial invariants , vassiliev - goussarov invariants and finite type invariants of three - dimensional integer homology spheres ( in this regard see also the unpublished important paper by fenn , rourke and sanderson @xcite ) .
cubic complexes may play a role in extending these invariants to higher dimensional knots .
+ in this paper we prove that any smooth , compact , closed , @xmath0-dimensional submanifold of @xmath6 with trivial normal bundle can be continuously isotoped by an ambient isotopy of @xmath6 onto a cubic submanifold contained in @xmath3 .
in particular , any knot can be isotoped onto a cubic knot contained in @xmath3 .
a _ cubulation _ of @xmath2 is a decomposition of @xmath2 into a collection @xmath1 of @xmath7-dimensional cubes such that any two of its hypercubes are either disjoint or meet in one common face of some dimension .
this provides @xmath2 with the structure of a cubic complex . + in general ,
the category of cubic complexes and cubic maps is similar to the simplicial category .
the only difference consists in considering cubes of different dimensions instead of simplexes . in this context
, a cubulation of a manifold is specified by a cubical complex pl homeomorphic to the manifold ( see @xcite , @xcite , @xcite ) . + the _ canonical cubulation _ @xmath1 of @xmath2 is the decomposition into hypercubes which are the images of the unit cube @xmath9 by translations by vectors with integer coefficients .
consider the homothetic transformation @xmath10 given by @xmath11 , where @xmath12 is an integer .
the set @xmath13 is called a _ subcubulation _ or _ cubical subdivision _ of @xmath1 .
the @xmath0-skeleton of @xmath1 , denoted by @xmath14 , consists of the union of the @xmath0-skeletons of the cubes in @xmath1 , _
i.e. , _ the union of all cubes of dimension @xmath0 contained in the faces of the @xmath7-cubes in @xmath1 .
we will call @xmath14 the _ canonical scaffolding _ of @xmath2 .
any cubulation of @xmath2 is obtained by applying a conformal transformation @xmath15 to the canonical cubulation .
+ in this section , we will prove that the union of the cubes of the canonical cubulation which intersect a fixed hyperplane is a closed tubular neighborhood of the hyperplane ( _ i.e. , _ a bicollar ) .
hence the boundary of the union of these cubes has two connected components and therefore the hyperplane can be isotoped to any of them .
this isotopy can be realized using normal segments to the hyperplane .
the boundaries of this bicollar are contained in the @xmath16-skeleton of the canonical cubulation .
if @xmath17 is a hyperplane of @xmath2 and @xmath18 is an @xmath0-dimensional affine subspace we use the same sort of ideas to show that the couple @xmath19 can be isotopically deformed to the boundary of the bicollar above in such a way that @xmath20 is deformed into the @xmath0-skeleton of the canonical cubulation .
+ these linear cases already contain the main ingredients of our proofs which in particular include convexity properties .
+ later we will generalize these ideas to the case of any smooth , compact , closed codimension one submanifold @xmath21 of @xmath2 : the union of the cubes of a small cubulation ( _ i.e. , _ with cubes of sufficiently small diameter ) which intersect @xmath21 is a bicollar neighborhood of @xmath21 .
we can deform @xmath21 to any of the boundary components using an adapted flow which is nonsingular in the tubular neighborhood and is transverse to @xmath21 . finally using the same type of ideas we prove at the end of the section that any smooth codimension two submanifold @xmath22 can be deformed into the @xmath0-skeleton of a sufficiently small cubulation .
let @xmath17 be a hyperplane in @xmath2 .
let @xmath23 be the canonical cubulation of @xmath2 and @xmath24 .
then : 1 .
@xmath17 is contained in the interior of @xmath25 .
2 . let @xmath26 such that the distance @xmath27 .
then the set @xmath28 is a cubic simplex , of some dimension , of the boundary of @xmath29 .
3 . @xmath30 is the union of @xmath16-dimensional cubes which are faces of cubes in @xmath25 .
_ let @xmath31 and @xmath32 be the canonical cubulation of @xmath33 .
let us consider the set @xmath34 .
let us first show by induction that @xmath35 .
if @xmath8 the result is obvious . let @xmath36 .
if @xmath37 for some cube @xmath29 then the result follows immediately . if @xmath38 for some cube @xmath29 then @xmath39 belongs to at least one @xmath16-face @xmath40 of @xmath29
the cubulation @xmath32 induces a cubulation @xmath41 in the hyperplane which contains the face @xmath40 . by induction hypothesis @xmath39
is in the interior ( relative to the hyperplane ) of the union of the @xmath16-cubes of @xmath41 which contain @xmath39 .
each of these @xmath16-cubes is a face of exactly two @xmath7-cubes of @xmath32 which contain @xmath39 .
then @xmath39 is in the interior of the union of these cubes .
therefore , by induction , @xmath35 . to prove 1 it is sufficient to observe that , by the preceding argument , @xmath35 , since @xmath42 . + let @xmath43 be the hyperplane parallel to @xmath17 at distance @xmath44 of @xmath17 which intersects @xmath29 .
then @xmath43 is a support plane of @xmath29 and therefore there exists a linear functional @xmath45 such that @xmath46 and @xmath47 for all @xmath48 .
since @xmath29 is a convex polytope , it follows from standard facts of the geometry of convex sets and linear programming that @xmath49 is a face of @xmath29 ( cubic simplex ) of some dimension since it is the set where the linear function @xmath50 , restricted to @xmath29 , achieves its minimum .
+ we have that @xmath51 .
therefore @xmath30 is contained in a union of @xmath16-faces .
each face @xmath40 of a cube in @xmath1 is a face of exactly two cubes of @xmath1 .
furthermore @xmath30 consists of faces @xmath40 of @xmath7-cubes in @xmath52 with the property that @xmath40 is also the face of a cube not belonging to @xmath25 .
this is true since such type of faces belong to @xmath30 and every point in @xmath30 is contained in one of those faces by the proof of 1 in proposition 2.2 .
@xmath53 let @xmath17 be a hyperplane in @xmath2 .
let @xmath23 be the canonical cubulation of @xmath2 and @xmath24 .
let @xmath54 be a point in @xmath17 and @xmath55 the normal line to @xmath17 at @xmath54
. then @xmath56 is connected . by proposition 2.2 , @xmath17 is contained in the interior of @xmath25 and therefore @xmath57 is a non - trivial compact interval .
_ @xmath17 divides @xmath2 in two open connected components @xmath58 and @xmath59 .
let us consider one of these components , for instance @xmath58 .
let @xmath60 and suppose that @xmath61
. then @xmath62 and therefore @xmath27 .
by proposition 2.2 the set @xmath28 is a cubic simplex , of some dimension contained in the boundary of @xmath29 . for each @xmath16-hyperface @xmath40 of @xmath29 which intersects @xmath63
, there is a hyperplane @xmath64 which supports @xmath40 .
only one of the closed halfspaces determined by @xmath64 contains @xmath29 .
we define @xmath65 as the intersection of all such closed halfspaces ( see figure 1 ) .
= 2.2 in the set @xmath65 is a convex unbounded set which contains @xmath29 and it is a union of cubes of @xmath1 .
we will show that @xmath66 is a support hyperplane of @xmath65 and @xmath67 .
+ first we will prove : + * claim 1 . * @xmath68 .
+ _ proof of claim 1_. if @xmath69 is a @xmath16-dimensional face of @xmath29 , the result follows easily .
if @xmath70 , then @xmath69 is contained in the intersection of two hyperplanes @xmath71 and @xmath72 .
one has that @xmath73 is of dimension @xmath0 , and therefore divides @xmath71 into two components , only one of which contains @xmath74 .
therefore @xmath65 lies in the halfspace determined by @xmath43 which contains @xmath29 and this halfspace does not contain @xmath17 .
this proves claim 1 .
+ to continue with the proof let us describe @xmath17 , @xmath66 and @xmath65 in terms of linear equalities and linear inequalities .
+ let us consider the set of hyperplanes @xmath75 defining @xmath65 which are not perpendicular to @xmath17 .
then the intersection of the halfspaces corresponding to these hyperplanes and containing @xmath29 , is defined by the set of points @xmath76 which satisfy the following set of inequalities : @xmath77 where @xmath78 is the standard inner product and @xmath79 and @xmath80 .
if none of the hyperplanes defining @xmath65 is orthogonal to @xmath17 then @xmath81 . +
* definition : * the vector @xmath82 is called the _
exterior normal vector _ to the face @xmath83 contained in the hyperplane @xmath84 . + if a hyperplane @xmath85 is perpendicular to @xmath17 the hyperplane corresponding to the opposite face to @xmath86 is also perpendicular to @xmath17 .
let @xmath87 be the closed halfspace determined by @xmath85 which contains @xmath29 .
+ * obvious remark . * if @xmath88 is a line which is perpendicular to @xmath17 then either @xmath89 or @xmath90 .
+ using a translation , if necessary , we can assume without loss of generality that the hyperplane @xmath17 is given by the linear equation : @xmath91 where @xmath92 is chosen in such a way that @xmath93 ( see figure 2 ) .
= 1.7 in the hyperplane @xmath43 divides @xmath2 into two closed halfspaces , one of which contains @xmath29 and the other contains @xmath17 .
+ then the hyperplane @xmath43 is given by the equation : @xmath94 and by hypothesis we have that @xmath95 . + for each @xmath96 , let @xmath97 be the normal line to @xmath17 which contains @xmath39 .
then the line @xmath97 can be parametrized by the function @xmath98 defined as follows : @xmath99 * claim 2 . * @xmath100 is either empty ( it happens only if @xmath29 has a hyperface , of dimension @xmath101 , which is contained in a hyperplane perpendicular to @xmath17 ) or @xmath100 is a ray _ i.e. , _ a noncompact closed interval contained in @xmath97 .
+ _ proof of claim 2 .
_ by the obvious remark above we need to show that if @xmath102 then @xmath103 where @xmath104 .
but this fact is immediate since , by ( 1 ) and ( 4 ) , it follows that if @xmath105 then since @xmath95 we have that @xmath106 .
we can choose @xmath107 .
since @xmath65 is contained in the halfspace determined by @xmath66 that contains @xmath29 we must have @xmath108 .
this finishes the proof of claim 2 .
+ claim 2 implies that for each @xmath109 , @xmath110 is a ray .
let @xmath111 be the union of all such rays .
by the above we have the following + * corollary of claim 2*. if @xmath112 is a cube which intersects @xmath113 then @xmath114 .
+ now we are able to finish the proof of lemma 2.3 . for @xmath115 ,
let @xmath116 with the order induced by the distance to @xmath17 .
let @xmath117 be the first point on @xmath118 such that @xmath119 . by proposition 2.2 the boundary
@xmath120 is the union of @xmath16-dimensional cubes , each of which is a hyperface of exactly two hypercubes in @xmath23 : one intersecting @xmath17 and one completely contained in @xmath58 .
let @xmath121 be the cube containing @xmath117 which does not intersect @xmath17 . then , by the above corollary , for all @xmath122 , all cubes containing @xmath123 do not intersect @xmath17 .
hence , @xmath117 is the only point on @xmath118 which lies on @xmath120 .
therefore @xmath124 is connected .
@xmath53 + lemma 2.3 implies the following corollary : the lengths of @xmath125 and @xmath126 vary continuously with @xmath54 .
these intervals have a natural order which provides them with a continuous orientation .
we denote @xmath57 and @xmath118 by @xmath127 $ ] and @xmath128 $ ] , respectively .
furthermore , the functions @xmath129 and @xmath130 are homeomorphisms from @xmath17 to @xmath131 and @xmath132 , respectively . as a consequence @xmath25 is a closed tubular neighborhood ( or bi - collar neighborhood ) of @xmath17 .
_ by lemma 2.3 , @xmath133 is connected .
we will prove that @xmath133 is in fact one point .
since the cubes in the cubulation @xmath23 are right - angled , if @xmath40 is a face of a cube orthogonal to @xmath17 then @xmath17 intersects @xmath40 .
this means that @xmath40 can not be contained in the boundary of @xmath25 .
therefore @xmath118 , which is orthogonal to @xmath17 , must intersect each @xmath16-cube in @xmath120 transversally .
+ we have proved that @xmath134 is connected and in particular , @xmath135 consists of two points @xmath136 and @xmath137 .
let @xmath137 be the corresponding point in @xmath138 .
we define the map @xmath139 such that @xmath140 is sent to @xmath141 . without loss of generality
we can assume that @xmath142 thus can identify @xmath17 with @xmath143 .
in terms of this identification we have @xmath144 , and @xmath145 .
consequently @xmath146 can be thought of as the graph of the function @xmath147 and since this graph is closed , it follows that @xmath148 is continuous .
this implies that the length of @xmath149 varies continuously .
hence @xmath150 is a homeomorphism .
obviously all the preceding arguments remain valid if we had used @xmath151 , @xmath152 and @xmath153 .
@xmath53 since @xmath25 is a bi - collared neighborhood of @xmath17 we obtain that @xmath17 can be continuously isotoped onto any of the two connected components of the boundary @xmath30 .
this can be done by an ambient isotopy of @xmath6 @xcite , @xcite .
we will use the following lemma to prove our main theorem .
let @xmath154 and @xmath155 be two orthogonal codimension one hyperplanes in @xmath2
. then @xmath156 can be cubulated ; more precisely , there is an ambient isotopy which takes @xmath156 onto a cubic complex , contained in the scaffolding @xmath14 of the canonical cubulation @xmath157 of @xmath2 .
_ let @xmath158 be the cubic complex formed by the union of elements of @xmath157 intersecting @xmath154 . by the above corollary
, @xmath158 is a tubular neighborhood ( or bi - collar in this case ) of @xmath154 .
let @xmath159 be a connected component of the boundary of @xmath158 .
recall the isotopy @xmath160 constructed in corollary 2.4 from @xmath154 to @xmath159 , which assigns to each point @xmath117 in @xmath154 the unique point on @xmath159 lying on the line normal to @xmath154 at @xmath117 .
the inverse of @xmath161 in this case is in fact the canonical projection @xmath162 of @xmath2 onto @xmath154 .
+ since @xmath154 and @xmath155 are orthogonal , if @xmath117 is a point in @xmath163 then the normal line at @xmath117 to @xmath154 will lie in @xmath155 .
therefore , @xmath161 restricted to @xmath156 is a homeomorphism which can be extended to an ambient isotopy ( @xcite ) between @xmath156 and @xmath164 .
now by proposition 2.2 @xmath159 is a cubic complex of dimension @xmath101 .
we will repeat the idea of lemma 2.3 for @xmath165 .
take the union @xmath166 of all @xmath16-dimensional cubes in @xmath159 that intersect @xmath167 .
we will prove that @xmath166 is a bi - collar of @xmath167 in @xmath159 and therefore one of its boundary components is ambient isotopic to @xmath168 .
this boundary component is obviously contained in the @xmath0-skeleton .
this would prove the lemma .
+ the intersection @xmath156 is an @xmath0-dimensional hyperplane , and hence of codimension one in @xmath154 .
consider the foliation of @xmath154 given by lines orthogonal to @xmath156 .
notice that the image of each of these lines under @xmath161 is a polygonal curve @xmath169 in @xmath159 , intersecting @xmath170 in one unique point .
+ as mentioned above , @xmath159 is cubulated as a union of @xmath16-faces of cubes of @xmath1 .
the projection @xmath162 of this @xmath16-dimensional cubulation gives a decomposition @xmath171 of @xmath154 into parallelepipeds .
observe that , in general , this decomposition is not a cubulation .
we will prove that a result analogous to the lemma 2.3 holds for @xmath171 . by this
we mean the following : let @xmath29 be a parallelepiped in @xmath171 which does not intersect @xmath156 , @xmath117 be a point in @xmath29 and @xmath172 be the set of points @xmath173 on the line in @xmath174 normal to @xmath156 which passes through @xmath117 such that @xmath175 , with the order induced by the distance to @xmath176 .
we will prove that any @xmath177 which intersects @xmath172 does not intersect @xmath156 .
+ the parallelepiped @xmath29 is the image under @xmath162 of an @xmath16-cube in @xmath159 .
this @xmath16-cube is a face of two @xmath7-cubes in @xmath2 , at least one of which does not intersect @xmath155 .
let @xmath121 be the @xmath7-cube not intersecting @xmath155 .
then , the set @xmath178 defined in the proof of the lemma 2.3 does not intersect @xmath155 .
thus @xmath179 does not intersect @xmath156 .
moreover , @xmath172 is contained in @xmath179 and , any parallelepiped in @xmath171 which intersects the interior of @xmath179 is contained in @xmath179 , and therefore does not intersect @xmath156 .
hence , no cube in @xmath180 intersects @xmath164 .
that is , no @xmath16-cube in @xmath159 which intersects the polygonal ray @xmath181 intersects @xmath164 .
+ proceeding as in corollaries 2.4 and 2.5 , choosing one of these points for each @xmath182 yields a continuous function from @xmath164 to the boundary of @xmath166 .
hence , @xmath166 is a bi - collar of @xmath164 in @xmath159 and @xmath164 can be deformed by an ambient isotopy into one boundary component of @xmath166 .
observe that this isotopic copy of @xmath164 is contained in the @xmath0-dimensional skeleton of the cubulation @xmath157 .
@xmath183 in lemmas 2.3 and 2.6 we only need to consider the subset of those cubes of @xmath184 whose distance to the corresponding hyperplanes @xmath17 , @xmath174 and @xmath176 is sufficiently small , for instance , less or equal than @xmath185 , _
i.e. , _ the cubes between two parallel hyperplanes at distance @xmath186 . the number @xmath187 appears because it is the diameter of the unit cubes in @xmath6 . a modification of the methods of the preceding results for hyperplanes can extend lemma 2.3 to the more general case given by following theorem : let @xmath188 be a smooth , compact and closed manifold .
let @xmath189 be a closed tubular neighborhood of @xmath21 .
we can assume that @xmath190 is the union of linear segments of equal length @xmath191 , centred at points of @xmath21 , and orthogonal to @xmath21 .
let @xmath23 be the canonical cubulation of @xmath2 .
for @xmath192 , let @xmath193 denote the corresponding subcubulation .
let @xmath194 .
then , we can choose @xmath44 big enough such that @xmath195 is a closed tubular neighborhood of @xmath21 and @xmath21 can be deformed by an ambient isotopy onto any of the two boundary components of this tubular neighborhood . _ proof .
_ the idea of the proof is to `` blow up '' the manifold and its tubular neighborhood by a homothetic transformation so that inside balls of large ( but fixed ) radius the manifold is almost flat and the normal segments are almost parallel .
then , locally , the manifold is approximately a hyperplane and we can apply a modification of the methods of the previous lemmas and then we rescale back to the original size by the inverse homothetic transformation .
the modification consists in replacing the normal segments to a hyperplane by segments of the flowlines of a nonsingular vector field which is very close to these normal segments in the tubular neighborhood of @xmath21 .
+ more precisely , let @xmath196 be the homothetic transformation @xmath197 where @xmath44 is a positive integer .
let @xmath198 and @xmath199 .
the homothetic transformation @xmath200 is isotopic to the identity , hence @xmath201 is isotopic to @xmath21 and @xmath202 is isotopic to @xmath190 . given @xmath203 we can choose @xmath44 large enough such that at each point of @xmath201 its sectional curvature is less than @xmath204 .
this means we can choose @xmath44 large enough such that for every point @xmath205 the pair @xmath206 is @xmath207-close to the pair @xmath208 , where @xmath209 denotes the closed ball centred at @xmath39 of radius sufficiently large , for instance of radius @xmath210 , and @xmath211 is the tangent space of @xmath201 at @xmath39 .
+ let @xmath212 , @xmath213 be the unit normal vector field of @xmath214 with respect to an orientation of @xmath214 .
it follows from standard facts of differential geometry ( see @xcite ) that the normal map @xmath215 , for @xmath216 has no focal points in @xmath202 .
this implies that the map @xmath217\rightarrow v_{m}$ ] given by @xmath218 is a diffeomorphism onto its image .
furthermore , there exists a constant @xmath219 such that if @xmath220 is a unit vector satisfying @xmath221 then the map @xmath222 is still a diffeomorphism from @xmath223 $ ] onto its image and @xmath224 for any normal vector to @xmath201 at a point @xmath216 .
+ the map @xmath225\to{v_{m}}\subset{\mathbb r}^{n+2}$ ] given by @xmath226 is a parametrization of @xmath227 . let @xmath228)$ ] be the smaller neighborhood of width @xmath229 .
the boundary of @xmath230 has two connected components @xmath231
@xmath232 let @xmath233 .
its boundary has also two connected components @xmath234 and @xmath235 .
+ * claim .
* there exists a finite family of diffeomorphisms @xmath236\to{v_{m}}\}_{i=1}^{k}\ ] ] such that if we set @xmath237)$ ] then 1 .
@xmath238 2 .
@xmath239 3 . for each @xmath240 the curves @xmath241 are parallel to the unit vector @xmath242 4 . for every cube @xmath29 of @xmath1 such that @xmath243 , @xmath244 and @xmath245 we have that @xmath246 for every @xmath247 exterior normal to a face of @xmath29 which is contained in @xmath248 ( see definition in proof of lemma 2.3 ) . 5
we define the set @xmath65 as in proof of lemma 2.3 , where the set @xmath69 is replaced by @xmath249 ( see also equation 1 ) .
we consider the convex set @xmath250 .
then there exists @xmath251 such that @xmath252 , @xmath253 .
_ proof of claim . _ since @xmath201 is compact , there exist a finite open subcovering @xmath254 of the open covering @xmath255 , where @xmath256 and vectors @xmath257 such that @xmath258 and the maps @xmath259\rightarrow v_{m}$ ] given by @xmath260 have the property that @xmath261)$ ] .
furthermore , @xmath257 can be chosen such that @xmath246 for every @xmath247 exterior normal to a face of @xmath29 which is contained in @xmath248 and intersects @xmath262)$ ] ( see figure 3 ) .
notice that this can be done because @xmath29 is a cube , so its faces meet at right angles .
using the fact that @xmath263 is homeomorphic to @xmath264 , the claim follows .
* remark : * the above claim can be modified to prove a similar result for @xmath265 . + for each @xmath266 , let @xmath267 $ ] be a smooth function such that @xmath268 if @xmath269 and @xmath270 if @xmath271 .
let @xmath272 be the global constant vector field equal to @xmath242 at every point @xmath273 .
let @xmath274 .
+ then @xmath275 is nonzero in @xmath230 and it has compact support inside @xmath190 , hence it defines a global flow @xmath276 . +
* claim . *
let @xmath277 and @xmath278 be the two boundary components of @xmath230
. then 1 . if @xmath279 there exits @xmath280 such that @xmath281 .
2 . the flowlines of @xmath282 are transversal to @xmath283 $ ] .
the flowlines of @xmath282 meet both @xmath277 and @xmath278 in one point and furthermore the flowlines meet both @xmath234 and @xmath235 in one point .
_ proof of claim .
_ first we remark that the flow is defined for all @xmath284 because the vector field @xmath285 has compact support .
every point of @xmath202 can be written uniquely in the form @xmath286 $ ] .
let @xmath287 $ ] be the map @xmath288 . by the properties of @xmath289 , we have that for fixed @xmath173 the function @xmath290 is strictly increasing for @xmath291 $ ] since the derivative of the function is positive .
this implies that the flowlines meet both @xmath278 and @xmath277 in one point .
furthermore , if @xmath29 is a cube which meets @xmath292 but does not meet @xmath201 we have that @xmath65 , defined above , is positively invariant under the flow , in fact @xmath293 .
therefore the flowlines meet both @xmath248 and @xmath265 in exactly one point .
this proves the claim .
+ we can rescale the flow by multiplying the vector field @xmath289 by a positive smooth function to obtain a new flow @xmath294 such that for every @xmath295 we have @xmath296 and @xmath297 .
+ then the map @xmath298\to{\mathbb r}^{n+2}$ ] defined by @xmath299 has the properties : 1 .
@xmath300)={\cal { q}}_{m_m}$ ] and therefore @xmath301 is a closed tubular neighborhood of @xmath21 .
2 . @xmath302 3 .
@xmath303 we have concluded that the set of cubes of the _ canonical _ cubulation of @xmath304 that touch @xmath201 is a closed tubular neighborhood of @xmath201 and @xmath201 can be deformed by an ambient isotopy onto any of the two boundary components of this tubular neighborhood .
to finish the proof we now rescale our construction back to its original size using the inverse homothetic transformation @xmath305 .
this transformation transforms @xmath184 onto the subcubulation @xmath306 .
@xmath53 + let @xmath307 , @xmath308 , be compact , closed and smooth submanifolds of @xmath6 such that @xmath309 and @xmath310 . since @xmath21 is codimension one in @xmath6 it is oriented and we will assume that @xmath22 has a trivial normal bundle in @xmath21 ( _ i.e. , _ @xmath22 is a two - sided hypersurface of @xmath21 )
. then under these hypotheses we have the following theorem for pairs @xmath311 of smoothly embedded compact submanifolds of @xmath6 : there exists an ambient isotopy of @xmath6 which takes @xmath21 into the @xmath16-skeleton of the canonical cubulation @xmath1 of @xmath6 and @xmath22 into the @xmath0-skeleton of @xmath1 . in particular , @xmath22 can be deformed by an ambient isotopy into a cubical manifold contained in the canonical scaffolding of @xmath6 . _ proof . _
the proof is very similar to that of the previous theorem and lemma 2.6 .
as before , given @xmath203 there exists @xmath192 , large enough , such that if we consider the homothetic transformation @xmath197 then the sectional curvatures of _ both _ @xmath312 and @xmath313 are less than @xmath204 .
more precisely , for every @xmath314 the triple @xmath315 is @xmath207-close to the triple @xmath316 , where @xmath209 denotes the closed ball centred at @xmath39 of radius sufficiently large , for instance of radius @xmath210 , and @xmath211 is the tangent space of @xmath201 at @xmath39 and @xmath317 is the tangent space to @xmath318 at @xmath39 .
+ let @xmath319 , @xmath202 , @xmath230 , @xmath294 , and @xmath320 be as in the previous theorem .
let @xmath321 be given by @xmath322 .
let @xmath323 . from above we know that @xmath324 is isotopic to @xmath201 and it is cubulated since it is a union of faces of cubes of @xmath1 .
we have that @xmath318 can be deformed to @xmath325 by a global isotopy of @xmath33 . since @xmath325 is contained in @xmath324 to prove the theorem we only need to find an isotopy @xmath326}$ ] such that @xmath327 , @xmath328 and @xmath329 is contained in the @xmath0-skeleton of @xmath1 .
+ let @xmath330 be the union of all faces of @xmath292 which intersect @xmath325 . let @xmath331 be the boundary of @xmath330 as a subset of @xmath332 . then @xmath331 is a union of @xmath0-cubes . +
* claim . *
@xmath330 is a closed tubular neighborhood of @xmath325 in @xmath292 .
+ _ proof of claim . _
the proof is similar to the proof of lemma 2.6 .
let @xmath333\to{m_m}$ ] be the parametrization of a tubular neighborhood of @xmath318 in @xmath201 given by @xmath334 , where @xmath335 is the geodesic in @xmath201 ( with respect to the induced metric of @xmath33 in @xmath201 ) such that @xmath336 and @xmath337 where @xmath338 is a unit vector at @xmath117 tangent to @xmath201 and transversal to @xmath339 .
in other words we use a sort of fermi coordinates for the tubular neighborhood . given @xmath340 we can choose @xmath44 large enough so that the curvature of both @xmath201 and @xmath318 is very small and the geodesics @xmath341 are very close ( in the @xmath342-topology ) to linear segments for @xmath343 $ ] .
we can take @xmath344 .
let @xmath345 , @xmath346 and let @xmath347\to \partial{\cal { q}}_{m_m}^+$ ] the function @xmath348 and @xmath349\}$ ] . then just as in the proofs of lemma 2.6 and theorem 2.8 we can prove that we can choose the vector field @xmath350 in such a way that @xmath351 is homeomorphic to a non - trivial segment and intersects @xmath331 in exactly one point ( see figure 4 ) .
the curves @xmath351 are rectifiable and the function @xmath352 is continuous .
hence , @xmath330 is a bi - collar of @xmath325 in @xmath292 .
this proves the claim . by the above
, @xmath325 can be deformed to one connected component of @xmath353 by an isotopy in @xmath324 . by standard theorems ( see @xcite , @xcite ) , this isotopy can be extended to a global isotopy @xmath326}$ ] .
observe that this isotopic copy of @xmath325 is contained in the @xmath0-dimensional skeleton of the cubulation @xmath157 .
to finish the proof we now rescale our construction back to its original size using the inverse homothetic transformation @xmath354 .
this transformation transforms @xmath1 onto the subcubulation @xmath355 . @xmath53
in classical knot theory , a subset @xmath356 of a space @xmath357 is a _ knot _ if @xmath356 is homeomorphic to a sphere @xmath358 .
two knots @xmath356 , @xmath359 are _ equivalent _ if there is a homeomorphism @xmath360 such that @xmath361 ; in other words @xmath362 . however , a knot @xmath356 is sometimes defined to be an embedding @xmath363 or @xmath364 ( see @xcite , @xcite ) .
we are mostly interested in the codimension two smooth case @xmath365 .
+ in this section we will prove the main theorem of this paper .
let @xmath23 be the canonical cubulation of @xmath2 .
let @xmath366 be a smooth knot of dimension @xmath0 .
given any closed tubular neighborhood @xmath367 of @xmath356 there exists an ambient isotopy @xmath368 with support in @xmath367 and @xmath369 $ ] such that @xmath370 and @xmath371 is contained in the @xmath0-skeleton of the subcubulation @xmath355 for some integer @xmath44 .
in fact @xmath372 is contained in the boundary of the cubes of @xmath355 which are contained in @xmath367 and intersect @xmath356 . in particular , using the homothetic transformation @xmath373 we see that there exists a knot @xmath374 isotopic to @xmath356 , which is contained in the scaffolding ( @xmath0-skeleton ) of the canonical cubulation @xmath23 of @xmath2 . _
proof_. let @xmath375 be a smoothly embedded knotted @xmath0-sphere in @xmath2 .
we denote @xmath376 and endow it with the riemannian metric induced by the standard riemannian metric of @xmath2 .
+ let @xmath367 be a closed tubular neighborhood of @xmath356 . by a theorem of whitney ( @xcite )
any embedded @xmath377 in @xmath378 has trivial normal bundle . since @xmath379 for @xmath380 , any embedding of @xmath381 in @xmath2 has trivial normal bundle .
hence @xmath367 is diffeomorphic to @xmath382 .
let @xmath383 be a diffeomorphism and @xmath384 be the projection .
we can assume that the fibers of @xmath39 are euclidean disks of radius @xmath385 .
if @xmath386 , the restriction of @xmath387 to @xmath388 ( where @xmath389 is the closed disk of radius @xmath390 ) is a closed tubular neighborhood @xmath391 of @xmath356 . then @xmath392
take a section in @xmath393 , _
i.e. _ for a fixed @xmath394 consider @xmath395 .
observe that @xmath396 is isotopic to the knot @xmath356 . now consider the pair @xmath397 . by the above
@xmath396 has normal bundle in @xmath393 and applying theorem 2.9 yields the result .
@xmath53 e. artin .
_ zur isotopie zweidimensionalen flchen i m @xmath398 _ abh .
math . sem .
hamburg ( 1926 ) , 174 - 177 .
s. cairns .
_ a simple triangulation method for smooth manifolds_. bull .
389 - 390 , 1961 .
a. v. ernavski .
_ isotopies in euclidean spaces_. ( russian ) uspehi mat .
nauk 19 1964 no .
6 ( 120 ) , 7173 .
54.78 n. p.
dolbilin , m. a. shtan ko , m. i. shtogrin .
_ cubic manifolds in lattices .
58 ( 1994 ) , no .
2 , 93107 ; translation in russian acad .
44 ( 1995 ) , no .
2 , 301313 . r. h. fox . _ a quick trip through knot theory_. topology of 3-manifolds and related topics .
prentice - hall , inc .
roger fenn , colin rourke , brian sanderson .
_ james bundles and applications ( 1996 ) .
_ http://www.maths.warwick.ac.uk/ cpr / ftp / james.ps l. funar .
_ cubulations , immersions , mappability and a problem of habegger . _
, 4e srie , t. 32 , 1999 , pp . 681 - 700
. l. funar .
_ cubulations mod bubble moves . low - dimensional topology _
( funchal , 1998 ) , contemp .
, 233 , amer . math .
soc . , 29 - 43
. w. hirsch .
_ smooth regular neighborhoods_. annals of mathematics vol .
76 , no.3 ( 1962 ) , 524 - 530 .
_ the definition of equivalence of combinatorial imbeddings .
_ publications mathmatiques de li.h..s .
, tome 3 ( 1959 ) p.5 - 17 .
s. matveev , m. polyak .
_ finite - type invariants of cubic complexes_. acta applicandae mathematicae 75 , pp . 125132 , 2003 . j. milnor .
_ morse theory_. princeton university press , 1969 .
r. palais .
_ local triviality of the restriction map for embeddings_. comment .
34 , 1960 , 305 - 312 .
d. rolfsen .
_ knots and links_. publish or perish , inc .
b. rushing .
_ topological embeddings_. academic press , 1973 , vol 52
. h. whitney .
_ on the topology of differentiable manifolds .
_ lectures in topology , pp
. 101 - 141 .
university of michigan press , ann arbor , mich . , 1941 .
_ e - mail address : _
.3 cm a. verjovsky .
instituto de matemticas , unidad cuernavaca .
universidad nacional autnoma de mxico .
/ n , col .
lomas de chamilpa .
cuernavaca , morelos , mxico , 62209 . | the @xmath0-skeleton of the canonical cubulation @xmath1 of @xmath2 into unit cubes is called the _ canonical scaffolding _
@xmath3 . in this paper , we prove that any smooth , compact , closed , @xmath0-dimensional submanifold of @xmath2 with trivial normal bundle can be continuously isotoped by an ambient isotopy to a cubic submanifold contained in @xmath3 .
in particular , any smooth knot @xmath4 can be continuously isotoped to a knot contained in @xmath3 . |
today , the question of life in the universe is a compelling goal of scientific research .
finding such life would have vast implications for the human mind , consequences that seem real to everyone , even if difficult to express .
down through the ages , the uniqueness of earth has been a stimulating issue for philosophers and ordinary people alike .
today , inspired by the discovery of many large planets around nearby stars , and propelled by progress in utilizing broad wavefronts of astronomical light , we can at last prepare to test the earth s exceptionalism by means of large optical systems in space .
an important , specific goal is to search for evidence of biologically significant molecules , particularly free oxygen , in spectra of the atmospheres of earth - size planets located in the habitable zones of nearby stars
planets with likely surface temperatures in the range compatible with water occurring in liquid form .
recently , the exoplanet task force of the astronomy and astrophysics advisory committee ( lunine et al .
2008 ) recommended as a precursor to such direct detection and study a space astrometry mission with `` the capability to measure convincingly wobble semi - amplitudes down to 0.2 @xmath0as integrated over the mission lifetime , '' where the astrometric semi - amplitude for zero eccentricity is @xmath2 here , the range of @xmath3 , which is the planetary semimajor axis in au , is the habitable zone , or 0.71.5 au for a star of @xmath4 solar luminosity , and otherwise varying as @xmath5 ; @xmath6 is the planetary mass in earth units ; @xmath7 is the stellar mass in solar units ; and @xmath8 is the distance of the star in parsecs . for the earth - sun system observed at @xmath9 pc , @xmath10as .
there are several good reasons for the exoptf s recommendation on space astrometry .
nevertheless , there has been no detailed explanation of what the numerical goal `` 0.2 @xmath0as '' implies , either in terms of expected scientific performance or regarding the true requirements on a mission to achieve it .
is it a good number ?
is it achievable by any currently planned astrometric mission ?
these questions are sharpened by the bias in least - squares estimation of keplerian parameters near the sensitivity limit , most recently discussed by shen & turner ( 2008 ) for radial - velocity studies , and documented here for astrometry .
therefore , the immediate purpose of this paper is to expose some performance issues that attend reflex astrometry with noisy data using the least - squares estimator in most common use .
( alternative estimators might be less biased or unbiased .
bayesian estimation is identical to least - squares when the errors are normally distributed and the priors are flat , as in the experiments described here . ) the ultimate goal is to find , if possible , a coherent program of space astrometry and follow - on coronagraphic spectroscopy of earth - like planets that reasonably fulfills the expectations of the community , as well as the public . in the first instance , our expectations are that astrometry can overcome three technical difficulties that have been identified with the direct - detection approach : ( 1 ) low search completeness per observation , implying many long exposures with no results , ( 2 ) low final search completeness for many high - priority stars , due to a high fraction of permanently obscured habitable - zone orbits , and ( 3 ) low probability of recovering in the future any planets that are discovered , because of the difficulty in estimating the planetary orbit from the limited range of the few direct astrometric measurements that might be obtained in the first observing season the season of discovery before the target is lost in the sun ( brown 2005 ; brown , shaklan , & hunyadi 2007 ; pravdo et al .
2007 ) . of no less importance , at least qualitatively , our expectations are also scientific , namely that astrometry will provide important physical information that direct detection can not provide , particularly estimates of the planetary mass , which controls atmospheric structure and composition , and orbit size , which determines the planet s thermal regime .
such information is basic for understanding the planet s physical conditions and , indeed , its very identity .
it may also provide clues about a planet s origins and dynamical evolution .
planets can not hide from astrometry , except in the noise . in principle
, only astrometry can measure the true mass and determine all the orbital elements . in principle , only astrometry can ensure that follow - on spectroscopy is planned with accurate predictions of the separation , position angle , and brightness of the planet at future times .
_ in practice _
, however , as we show here using the completeness metric , these expectations may not be satisfied by the design parameters currently associated with the observing system most strongly advocated for the task , the _ space astrometry mission _ , light version , or _ sim lite_.
we assume an idealized astrometric instrument , which measures the two - dimensional position of the star relative to a local astrometric reference frame on the sky .
an idealized astrometric data set comprises @xmath11 measurements of the form ( @xmath12 , @xmath13 , @xmath14 , @xmath15 , @xmath16 ) : the @xmath17@xmath18 position at epochs @xmath19 spread over duration @xmath20 , longer than one planetary period , where the exposure time is @xmath21 for each of the two quasi - orthogonal directions , and where @xmath22 is the positional uncertainty in either direction .
we assume the astrometric errors are normally distributed . assuming that @xmath8 is known from the annual parallax and that @xmath7 is known from stellar spectrophotometry , the theoretical apparent position of the star is fully described by seven keplerian parameters , of which three are physical properties of the system ( @xmath3 , eccentricity @xmath23 , and @xmath6 ) , three are eulerian rotations of the orbit with respect to the line of sight ( argument of periastron @xmath24 , inclination angle @xmath25 , and longitude of the ascending node @xmath26 ) , and one specifies the orbital phase at a moment in time , such as at the epoch of the first data point ( initial mean anomaly @xmath27 ) .
currently , we consider only planetary orbits with periods significantly shorter than @xmath20 .
we consider a subset of the universe of astrometric reflex signals due to planets on randomized habitable - zone orbits for which @xmath3 is uniformly distributed in the habitable - zone range @xmath28 , @xmath23 is uniformly distributed over 0 @xmath29 ; and @xmath24 , @xmath30 and @xmath26 are randomized to distribute the orbit pole uniformly on the sphere .
the subset of signals we consider is made up of those for which a key characteristic , the astrometric signal - to - noise ratio _ snr _ , @xmath31 is sufficiently small to approach the limits of useful scientific performance . here , the mission error " for a star is @xmath32 where @xmath33 is the total exposure time .
the noise reduction in eq . 3 with @xmath33 and @xmath11 is valid only down to the noise floor , or for @xmath34 , where @xmath35 is determined by systematic effects .
_ snr _ can be small due to any combination of the several instrumental , observational , and physical parameters appearing in eqs . 1 , 2 , and 3 .
nevertheless , our interest is focused on @xmath36as , which is the exoptf s recommended goal , as well as 2/3 the iconic value for the earth - sun system at @xmath9 pc .
( due to orbital eccentricity and the arbitrary alignment of the orbital major axis with respect to the line of sight , @xmath37 may be larger or smaller than the semi - major axis of the apparent orbital ellipse on the plane of the sky . )
we use the completeness metric @xmath38(_snr _ ) to estimate scientific performance .
we define @xmath38 as the fraction of possible planets with the characteristic _ snr _ that are successfully observed , assuming the planet is present .
`` successfully observed '' variously means ( 1 ) that the planet is detected by periodogram analysis of the data , or if it is detected , ( 2 ) that the planetary mass is , or ( 3 ) future position is , usefully estimated from the orbital solution produced by a least - squares fit of the parameters of the keplerian theory to the noisy data set .
traub et al . ( 2009 ) is the most recent discussion of the scientific performance of space reflex astrometry on earth - like planets , assuming the design parameters of _ sim lite_. that paper reports the preliminary findings of a jpl - sponsored study in which one `` goal was to see what accuracy of _ sim lite _ is needed to detect earth - like planets . ''
the study builds on earlier analyses by the _ sim _ science team of variants of the _ sim _ mission , including catanzarite et al .
( 2006 ) and unwin et al .
the findings have been reiterated by the astrometry committee of the jpl - sponsored 2008 exoplanet forum ( lawson , traub , & unwin 2009 ; muterspaugh & tanner 2009 ) .
one purpose of that forum was to inform the decadal survey of astronomy and astrophysics ( astro2010 ) `` concerning technology development and technology readiness of proposed exoplanet missions , '' in this case , _
sim _ and its variants .
the forum report urges that a _ sim_-type astrometry mission be implemented forthwith .
it cites the exoptf finding `` that for the 1115 year ( 20192023 ) timeframe , ` assuming the space - borne astrometric mission described above is fielded in the second time epoch , no additional major space - based astrometric effort is envisioned in this time frame . '
the committee agrees with this conclusion . ''
we infer the following common denominators of the studies and reports mentioned in the previous paragraph . 1 .
for _ sim lite _ , the instrumental parameters are @xmath39as , @xmath40 sec , and @xmath41as , which from eq
. 3 implies @xmath42 and @xmath43 sec . for @xmath44as ( exoptf ) or @xmath45as ( earth - sun at 10 pc ) , @xmath46 or 8 .
we accept these numbers as given .
2 . for detection via periodogram ( power spectrum ) , _ snr _ = 5.8 is required for search completeness @xmath47 with a false - alarm probability _ fap _ = 0.01 . we agree with this , finding the value is closer to _ snr _ = 6.0 for our adopted planetary population , as shown below .
in any case , the _ sim lite _ search completeness for the exoptf s @xmath44as is @xmath48 .
estimates of planetary mass are achievable with accuracy `` close to the theoretically expected value '' from the minimum - variance bound or cramr - rao limit ( traub et al .
2009 , gould 2008 ) . using the least - squares estimator ,
we differ , but actually it depends on the meaning of `` close . ''
we show below that , because the least - squares estimator of mass is biased , the completeness @xmath38 of even roughly accurate estimates of mass for earth - like planets is significantly degraded in the _ sim lite _ domain of @xmath49as .
the estimates of the orbital period approach the minimum - variance bound .
just as shen & turner ( 2008 ) found for radial - velocity data sets , for example , we find no sign of bias for period estimation via least - squares fitting of noisy astrometric data sets .
5 . other than our studies reported below
, we are not aware of any previous tests of the ability to accurately predict future planetary positions from the analysis of noisy astrometric data sets .
nevertheless , the literature contains fulsome assertions about the adequacy of _ sim_-based positional predictions , even if their basis has not been explained .
unwin et al .
( 2007 ) : `` for many of these stars , _ sim _
s orbital solution will be precise enough to predict the best timing for a direct observation .
this information is crucial for direct imaging , since a planet in the habitable zone can spend much of its time hidden in the glare of the planet star . ''
lunine et al . (
2008 ) : `` because the astrometric mission would have already determined the locations ( ` addresses ' ) and orbits of the candidate planets , the spaceborne direct detection system is relieved of the burden of searching ; its mission is to point at those stars and study the planets .
this greatly simplifies a direct detection mission in several ways : the target list can be ordered and prioritized ahead of time to maximize efficiency in moving from one target to another and in observing those targets whose orbital phases make them most observable ''
muterspaugh & tanner ( 2009 ) : `` astrometric observations of stars hosting earth - like planets allow the orbits to be determined .
the position of the planet can be predicted as a function of time .
this provides a solid basis for future planning [ of ] direct imaging programs , as the observing schedule can be set to look at the targets when the star and planet are most optimally configured for isolating light from the planet . '' in the current study , we do not confirm these assertions using the least - squares estimator of the keplerian parameters . in the _ sim lite _ performance domain
, we find low @xmath38 for even roughly accurate predictions of the future positions of earth - like planets .
currently , the most popular algorithm to search noisy data for faint periodic signals is the periodogram , @xmath50 , where @xmath51 is the angular frequency . in this paper
, we use the most popular normalization of the periodogram , which subtracts the mean value of the measurements and divides by the standard deviation of the data set ( lomb 1976 ; scargle 1982 ; horne & baliunas 1986 ) .
recently , the theoretical basis for closed - form estimates of periodogram performance has collapsed spectacularly ( frescura , engelbrecht , & frank 2008 ) .
a new calibration approach , based wholly on monte carlo methods , is now emerging ( catanzarite et al .
2006 ) , and is followed here . for two - dimensional astrometry
, we use the `` joint '' periodogram defined by catanzarite et al .
( 2006 ) . instead of selecting the maximum peak value @xmath52 at a finite list of `` independent frequencies''a now - discredited theoretical concept , particularly for arbitrarily spaced data
we locate the true peak in the continuous range of frequencies between the fundamental , 2@xmath53 , and the nyquist frequency , @xmath54 .
( orbital frequencies in the habitable zone are always near the short end of this range . ) at its performance limits , discovery via periodogram is based on selecting a value of _ fap _ , then estimating the threshold value for the maximum peak , @xmath55 , of @xmath50 such that data sets of pure noise with no periodic signal present
produce @xmath56 with a probability equal to _
fap_. the value of @xmath55 depends on @xmath11 , but not on the timing of the data points .
for all the periodogram analysis report here , we adopted _ fap _ = 0.01 , and we estimated the values of @xmath55 for given @xmath11 from 100,000 data sets of pure noise . for discovery , we are interested in the search completeness @xmath57 , which is the fraction of random planets with given _ snr _ that are detected by the criterion @xmath56 . for this purpose , we prepared 880,000 random data sets at spaced values of _ snr _ in the range @xmath58 , then subjected the data sets to periodogram analysis .
the results are shown in fig . 1 .
we confirm that @xmath47 is achieved at _ snr _ @xmath59 .
this result indeed the value of @xmath38 at any _
snr_depends somewhat on the assumed planetary population .
for example , face - on orbits achieve the highest @xmath38 at any _ snr _ , and high eccentricity degrades @xmath38 , due to the spectral power distributed into harmonics .
the shaded region of fig .
1 is the range of _ snr _ that is excluded for _ sim lite _ on earth - like planets at 10 pc , @xmath60 . at _ _
snr__@xmath61 , @xmath62 for discovery . in the process of conducting these periodogram investigations
, we confirmed that search completeness @xmath38 depends on the parameters @xmath37 , @xmath63 , @xmath33 , and @xmath11 only as they are combined in the _ snr _ characteristic according to eqs . 2 and 3 , as expected from the literature .
at spaced values of _ snr _ in the range @xmath58 , we found orbital solutions for 265,000 data sets in which periodogram analysis had discovered a planet .
we compared the true value of the planetary mass , @xmath64(true)the value with which the data set was prepared with the estimated value , @xmath64(est ) , found by a least - squares orbital fit to the data . in this application , @xmath38(_snr _ ) means the fraction of data sets with the value of _ snr _ for which the fractional error , ( @xmath64(est)@xmath64(true))/@xmath65(true ) , occurs in a specified range centered on zero .
2 shows the results for the ranges @xmath6610% and @xmath6625% .
we see that for the mass - estimation accuracy @xmath6625% , @xmath67 for @xmath68 , and for @xmath6610% accuracy , @xmath69 . the dashed lines in fig .
2 show the performance associated with the minimum - variance bound ( gould 2008 ) . at this bound , the expected fractional error in mass is @xmath70__snr _ _ , and the normalized deviations observed fractional errors in mass times _ _ snr__@xmath71should be normally distributed with mean zero and unit variance ( traub & gould 2008 ) .
therefore , for mass estimation accuracy @xmath72 , which equals 0.1 or 0.25 for the cases shown , the completeness associated with the minimum - variance bound is @xmath73 where _ erf _ is the error function . comparing the dashed and solid curves ,
it is clear that the performance estimated by these monte carlo simulations , using least - squares estimation , falls significantly short of that expected from the minimum - variance bound for @xmath68
. a closer look at the normalized deviations ( fig .
3 ) reveals the reason for the reduction of @xmath38 for mass estimation from the minimum - variance bound : the least - squares estimator produces biased estimates of mass from noisy astrometric data sets , just as shen and turner ( 2008 ) found for noisy radial - velocity data sets . in the process of conducting these mass investigations
, we confirmed that @xmath38 for accurate mass estimation depends on the parameters @xmath37 , @xmath63 , @xmath33 , and @xmath11 only as they are combined in _ snr _ according to eqs . 2 and 3 .
habitable - zone orbits of nearby stars are highly obscured , with many possible planets peeking out only for short times from behind a central field obscuration .
also , suppression of starlight may be easier to achieve in a smaller region of the focal plane . in response to these considerations
, we suggest that a positional accuracy of 30% the semi - major axis@xmath74is a reasonable first estimate of a possible requirement on astrometry to support the planning of direct observations of the planet .
we analyzed the same sample of 265,000 pairs of true and fitted orbital solutions discussed in the previous section to estimate performance in predicting the planetary position .
the six solid curves in fig . 4
show @xmath38(_snr _ ) for the positional metric better than @xmath660.3 @xmath3 after 020 years have elapsed after the end of the 5-year duration of astrometric observations .
the positional metric is defined as the norm of the vector deviation between the true and estimated planetary positions on the plane of the sky . for @xmath68 , we find @xmath75 after more than one year has elapsed , and @xmath76 after ten years . for a further , more elaborate , metric of recoverability , we considered the time interval of one ( estimated ) orbital period following the end of astrometric observations .
we computed the true position of the planet for each day from the true orbital elements , and computed the estimated position from the estimated orbital elements .
next , we identified the subset of days for which the planet was predicted ( from the estimated elements ) to have a greater separation from the star than 0.7 @xmath3 , which seems like a reasonable definition of _ when _ follow - up spectroscopy should be planned .
next , we checked on those days whether the norm of the deviation between the true and estimated positions of the planet was less than 0.3 @xmath3 , which seems like a reasonable working size of the dark hole that a coronagraph might create around the entrance aperture of a spectrometer . at each _ snr _ , we counted the number of days for which both these criteria were satisfied , and divided it by the number of days only the first was met .
this recoverability metric tends to unity in the limit of excellent predictions , at high _
snr_. if the metric is @xmath770.95 , say , it means that the planet is recovered ( within 0.3 @xmath3 of the estimated location ) for @xmath7795% of the days that the data predicts the planet to be observable ( outside 0.7 @xmath3 ) .
preliminarily , we define a planet as recovered " if the metric is @xmath770.95 . the dashed red line in fig
. 4 shows @xmath38(_snr _ ) , the recovery completeness by this test , @xmath78 for @xmath68 . in the process of conducting these positional investigations
, we confirmed that @xmath38 depends on the parameters @xmath3 , @xmath63 , @xmath33 , and @xmath11 only as they are combined in the _ snr _ characteristic according to eq . 2 .
our preliminary analysis suggests that the scientific performance on earth - like planets of a space astrometry mission characterized by @xmath79as may not satisfy expectations , particularly , the exoptf recommendation for `` the capability to measure convincingly wobble semi - amplitudes down to 0.2 @xmath0as integrated over the mission lifetime , '' particularly in estimating the planetary mass and future planetary position .
the reason for this underperformance is that the maximum astrometric signal - to - noise ratio , @xmath80 for @xmath44as , which is the _ snr _ required for @xmath8150% discovery completeness , is well below the _ _ snr__s required for high completeness of accurate estimates of planetary mass and future positions . in the case of mass estimation ,
the degradation from the minimum - variance bound is only quantitative , and perhaps not a serious problem for science .
for example , we expect that meaningful and useful _
upper limits _ to the planetary mass can still be set by reflex astrometry , as is customary for the radial - velocity technique . in the case of predicting the planetary position ,
the problem is qualitative and more serious . knowing where and when to place the entrance aperture of the spectrometer is a huge science - operational issue for the follow - on mission , due to the very long exposure times .
we had hoped that _
sim_-class astrometry would be a solution , but now that appears not to be the case , at least for the most interesting planets in terms of life , which are the earth - like planets at typical distances of 520 pc .
this paper reports only a preliminary analysis , not yet taking into account the characteristics of individual stars , which affect both astrometric performance and follow - up coronagraphic spectroscopy . also , we treat the idealized case of a single planet orbiting a single star , and assume there are no non - keplerian complications or uncertainties such as due to multiplicity , proper motion , parallax , stellar activity , reference - frame errors , etc .
the real case can only be more difficult .
we hope and expect that these results will motivate a critical reexamination of exoplanetary program strategies involving space astrometry .
we must ensure either that an adequate technical solution for reflex astrometry is part of the plan , or we must define a programmatic pathway toward spectroscopy on earth - like planets that is less dependent on reflex astrometry .
an important unsolved , strategic problem in the exoplanetary science program is figuring out how to predict the future position of an earth - like planet with accuracy sufficient to ensure the success of the science operations of a follow - on spectroscopic mission , which would search for biologically significant molecules in the atmosphere .
i appreciate jeremy kasdin s raising the issue of the accuracy of the planetary ephemeris , and rmi soummer s calling this to my attention .
i thank christopher burrows for interesting discussions , both directly and indirectly related to this work .
i thank christian lallo , for making the computations possible , and sharon toolan for her excellent help preparing the manuscript .
brown , r. a. 2004 , apj , 610 , 1079 .
brown , r. a. 2005 , apj , 624 , 1010 .
brown , r. a. , shaklan , s. b. & hunyadi , s. l. 2007 , in jpl 07 - 02 7/07 , _ proceedings of tpf - c workshop 9/2829/2007_. catanzarite , j. , shao , m. , tanner , a. , unwin , s. , & yu , j. 2006 , pasp , 118 , 1319 .
frescura , f. a. m. , engelbrecht , c. a. , & frank , b. s. 2008 , mnras , 388 , 1693 .
gould , a. , 2009 , apj submitted ; arxiv:0807.4323v1 .
horne , j. h. & baliunas , s. l. 1986 , apj , 302 , 757 .
lawson , p. r. , traub , w. a. , & unwin , s. c. , eds .
, _ exoplanet forum 2008 report _ , jpl xx - xx mm / yy , in preparation .
lomb , r. n. 1976 , ap .
space sci . , 39 , 447 .
lunine , j. , et al .
2008 , _ report of the exoplanet task force : worlds beyond : a strategy for the detection and characterization of exoplanets_,http://www.nsf.gov / mps / ast / aaac / exoplanet_task_force / reports / exoptf_final_report.pdf/. muterspaugh , m. & tanner , a. 2009 , in the proceedings of the _ exoplanet forum 2008 _ , eds .
lawson , p. r. , traub , w. a. , & unwin , s. c. , jpl xx - xx mm / yy , in preparation .
pravdo , s. h. , shaklan , s. , brown , r. , & hunyadi , s. l. 2007 , baas , 38 , 333 .
scargle , j. d. 1982 , apj , 263 , 835 .
shen , y. & turner , e. l. 2008 , apj , 685 , 553 .
traub , w. a. , beichman , c. , boden , a. f. , boss , a. p. , casertano , s. , catanzariate , j. , fischer , d. , ford , e. p. , gould , a. s. , halverson , s. , howard , a. , kasdin , n. j. , laughlin , g. p. , levison , h. f. , lin , d. , makarov , v. , marr , j. , muterspaugh , m. , raymond , s. n. , savransky , d. , shao , m. , sozzetti , a. , & zhai , c. , 2009 , in `` extrasolar planets in multi - body systems : theory and observations , '' ed .
krzysztof gozdziewski , european astronomical society publication series , in preparation .
unwin , s. c. , shao , m. , tanner , a. m. , allen , r. j. , beichman , c. a. , boboltz , d. , catanzarite , j. h. , chaboyer , b. c. , ciardi , d. r. , edberg , s. j. , fey , a. l. , fischer , d. a. , gelino , c. r. , gould , a. p. , grillmair , c. , henry , t. j. , johnston , k. v. , johnston , k. j. , jones , d. l. , kulkarni , s. r. , law , n. m. , majewski , s. r. , makarov , v. v. , marcy , g. w. , meier , d. l. , olling , r. p. , pan , x. , patterson , r. j. , pitesky , j. e. , quirrenbach , a. , shaklan , s. b. , shaya , e. j. , strigari , l. e. , tomsick , j. a. , wehrle , a. e. , & worthey , g. 2008 , pasp , 120 , 38 . | we provide a preliminary estimate of the performance of reflex astrometry on earth - like planets in the habitable zones of nearby stars . in monte carlo experiments ,
we analyze large samples of astrometric data sets with low to moderate signal - to - noise ratios .
we treat the idealized case of a single planet orbiting a single star , and assume there are no non - keplerian complications or uncertainties
. the real case can only be more difficult .
we use periodograms for discovery and least - squares fits for estimating the keplerian parameters .
we find a completeness for detection compatible with estimates in the literature .
we find mass estimation by least squares to be biased , as has been found for noisy radial - velocity data sets ; this bias degrades the completeness of accurate mass estimation .
when we compare the true planetary position with the position predicted from the fitted orbital parameters , at future times , we find low completeness for an accuracy goal of 0.3 times the semimajor axis of the planet , even with no delay following the end of astrometric observations .
our findings suggest that the recommendation of the exoplanet task force ( lunine et al .
2008 ) for `` the capability to measure convincingly wobble semi - amplitudes down to 0.2 @xmath0as integrated over the mission lifetime , '' may not be satisfied by an instrument characterized by the noise floor of the _ space interferometry mission _
, @xmath1as .
an important , unsolved , strategic challenge for the exoplanetary science program is figuring out how to predict the future position of an earth - like planet with accuracy sufficient to ensure the efficiency and success of the science operations for follow - on spectroscopy , which would search for biologically significant molecules in the atmosphere . |
looking at the diverging uniform ( @xmath210 ) susceptibility @xmath6 as @xmath14 ( fig . [ chi ] ) intuitively suggests large low@xmath30 response and in turn increasing contribution of @xmath211 to the spin relaxation rate @xmath29 as @xmath14 .
this is not what is observed , since we see no increase of @xmath94 ( fig .
4b in the main text ) as @xmath14 , but instead @xmath94 decreases with decreasing @xmath2 , which is in agreement also with experimental data ( ref .
@xcite , fig .
3a ) . this dichotomy can be partly understood by exploring the connection between static uniform spin susceptibility @xmath6 with the static spin structure factor ( equal - time correlation ) @xmath115 , representing also the frequency integral of dynamical spin structure factor @xmath212 .
the connection @xmath213 together with the low@xmath2 rg results ( see fig . [ chi ] ) @xmath214 $ ] leads to @xmath215 $ ] .
this shows that @xmath118 goes to 0 as @xmath14 and is not diverging , rather its slow logarithmic approach to @xmath216 ( in contrast to linear in @xmath2 decrease for homogeneous system ) .
this is in agreement with results in fig . 15 in ref
. @xcite , which show that @xmath115 at @xmath139 goes to 0 as @xmath63 and is only slightly increased by randomness for @xmath217 .
another remarkable property of rhc can be seen in the difference between dynamic ( @xmath218 ) properties , e.g. @xmath219 , and strictly @xmath114 contribution .
this is shown in the lower panel of fig .
[ s_fig_sqw ] , which shows that @xmath220 is singular at @xmath114 since it consists besides the continuous background ( regular part ) also of a distinct delta peak at @xmath114 ( see fig .
[ s_fig_sqw]b ) .
this peak is non dispersive and is the signature of non ergodicity in the random system ( absence of diffusion ) at least for low@xmath2 .
similar peak is observed even in a random non interacting electron system at finite@xmath2 ( not presented ) . in our analysis of spin relaxation for which @xmath33 is relevant , the @xmath114 peak was excluded ( see fig .
[ s_fig_sqw]c and also section above ) .
distinction between strictly @xmath114 and @xmath221 properties can be traced back to the difference between diagonal elements , e.g. @xmath222 determining @xmath5 with only total spin @xmath223 ( triplet ) states contributing , and non diagonal elements describing the transitions between , e.g. singlet and triplet states , which are relevant for dynamical ( @xmath221 ) properties . for @xmath224 and @xmath225 showing much larger weight at low@xmath15 at @xmath226 .
panel ( b ) show @xmath220 for low @xmath15 where the non dispersive @xmath114 delta peak is seen .
the width of the peak is due to thew broadening used in the presentation .
panel ( c ) show low@xmath15 @xmath220 with removed @xmath114 delta peak . ] by approximating nmr relaxation rate @xmath29 with local @xmath227 , we neglected the effect of form factors ( @xmath228 , eq . 2 in the main text ) , which is a good approximation since in the regime of our calculations the main contribution comes only from @xmath226 .
to show this , we present @xmath220 in fig .
[ s_fig_sqw ] , where it is evident that the main contribution at @xmath229 comes from @xmath226 .
this stays valid even in the low@xmath2 regime , where @xmath6 is already increased due to renormalization of @xmath3-s as observed in the rg flow .
randomness does in fact slightly reduces the contribution of @xmath230 and slightly increases the contribution of @xmath217 , but the transfer of weight is much too small to make @xmath217 dominant .
this is in agreement with finding for @xmath115 shown in fig . 15 in ref .
@xcite , where even for very large randomness and @xmath139 the main contribution stays at @xmath226 similarly to the homogeneous system @xcite
. it also agrees with experimental observation and direct statement of the authors @xcite , that there in no indication of important @xmath217 contributions as , e.g. , the d.c . field dependence of @xmath29 , being indication of the absence of ( anomalous ) low@xmath30 ( diffusion ) contribution .
our analysis of local @xmath18 was based on assumption that there is no singular contribution emerging from long
wavelength @xmath199 physics .
indeed , all our available data for @xmath231 for afm rhc confirm that the dominant regime at low@xmath2 is @xmath34 .
still , @xmath199 regime needs further attention since it can lead at @xmath232 to a divergent @xmath233 either from the propagation ( prevented by randomness in the rhc ) in the homogeneous xx chain @xcite ( with @xmath234 ) or even more as the consequence of the spin diffusion @xcite ( @xmath235 ) .
the latter can be realized at @xmath12 but vanishes at @xmath84 within the rhc @xcite .
our results so far indicate that in spite of possible @xmath12 diffusion its contribution to @xmath8 is unresolvable for reachable systems , as follows also from nmr experiments @xcite where it can be directly tested via the magnetic field dependence of @xmath7 .
concrete realization of a random system in ref .
@xcite , namely , has binary disorder distribution with two exchange couplings , @xmath236 $ ] and @xmath237 $ ] , which is in contrast with our continuous disorder distribution model , motivated by a usual theoretical reference .
therefore the question arises , how different are the results for the binary distribution from our results . here
we argue that the results are qualitatively and even quantitatively very similar for both distributions .
this was realized already by j. e. hirsch @xcite , who showed that arbitrary disorder distribution lead to the similar low@xmath2 behaviour .
axis is not rescaled and both distributions would therefore give almost the same @xmath94 .
plots are for @xmath41 , @xmath238 and obtained with @xmath239 realizations . ] to demonstrate the effect of binary distribution we show in fig .
[ s_fig_bin ] the comparison of relaxation rate cdf - s for continuous disorder distribution with the ones for binary disorder distribution with the same effective width .
it is seen that the difference is small and largest for strongest disorder , where it still remains only quantitative , while for low disorder cdf - s are essentially the same for both disorder distributions .
therefore our results obtained with continuous distribution can easily be compared with measurements and they indeed agree qualitatively and to some extend even quantitatively with them ( see main text ) .
using phenomenological stretched exponential form to fit experimental data on magnetization relaxation seems to be a common practice , which can be attributed to the fact that stretched exponential form can capture anomalously long tails in the distribution of the relaxation rates ( normal distribution can not ) and is at the same time very convenient for the fitting procedure .
this immediately raises the question , how good this form really is for the description of experimental data and can it be motivated by some microscopic picture , e.g. model hamiltonian . in fig .
[ s_fig_pdf ] we show our rhc model results for relaxation time distributions ( pdf - s ) , which shows several important features when compared to the experimentally suggested stretched exponential forms ( see fig . 4 in ref .
first one can see that the @xmath2 evolution is similar to the experimental one and more importantly at low@xmath2 anomalously long tails ( or large spread ) in pdf appear , which can be captured with stretched exponential form and not with , e.g. , normal ( gaussian ) distribution .
this could be the reason for the success of the stretched exponential form in the fitting procedures and its phenomenological description of experimental data . ) for several temperatures and one randomness @xmath240 .
these results should be compared with the shape of experimental ones in fig . 4 in ref .
temperature evolution of the distribution agrees with the experiment , as well as appearance of anomalously long tails of the distribution at low@xmath2 .
points correspond to fitted @xmath89 which shows similar @xmath2dependence as experiment , while dashed vertical line corresponds to homogeneous ( @xmath51 ) model . ]
however , the description of our rhc model results with stretched exponential is not perfect as can be expected , and the most obvious deviations can be found in the long tails .
e.g. , for few specific value of @xmath88 analytical form of pdf - s is know @xcite and for @xmath241 has a form @xmath242}{\sqrt{4\pi\tau_0s^3}}\,.\label{pdfan}\ ] ] on fig .
[ s_fig_sea ] we compare our numerical result with @xmath243 ( for @xmath225 and @xmath68 ) with stretched exponential cdf obtained from eq . ,
@xmath244 $ ] .
we observe that the rhc model predicts longer ( shorter ) tails in the pdf for smaller ( larger ) @xmath18 than stretched exponential form .
this is in turn reflected in the corresponding time dependent magnetization relaxation function ( being laplace transform of pdf ) directly probed by experiment .
one could , for example , from our pdf - s propose a new form ( instead of stretched exponential ) by approximating pdf - s with some function and performing its laplace transform .
this is however not trivial and we leave it as a motivation for future work . in this way
obtained form is expected to describe experimental data better than stretched exponential , although differences might be small and experimental data with higher resolution might be needed .
24 s. r. white , phys .
* 69 * , 2863 ( 1992 ) .
u. schollwck , rev .
phys . * 77 * , 259 ( 2005 ) .
j. kokalj and p. prelovek , phys .
b * 80 * , 205117 ( 2009 ) .
j. kokalj and p. prelovek , phys .
b * 82 * , 060406(r ) ( 2010 ) . for a recent review ,
see p. prelovek and j. bona , in _ strongly correlated systems - numerical methods _ , edited by a. avella and f. mancini ( springer series in solid
state sciences 176 , berlin , 2013 ) , pp .
j. jakli and p. prelovek , adv .
phys . * 49 * , 1 ( 2000 ) .
d. fisher , phys .
b * 50 * , 3799 ( 1994 ) .
j. e. hirsch , phys . rev .
b * 22 * , 5355 ( 1980 ) .
a. zheludev , t. masuda , g. dhalenne , a. revcolevschi , c. frost , and t. perring , phys .
b * 75 * , 054409 ( 2007 ) .
t. shiroka , f. casola , v. glazkov , a. zheludev , k. pra , h.r .
ott , and j. mesot , phys .
lett . * 106 * , 137202 ( 2011 ) .
t. masuda , a. zheludev , k. uchinokura , j .- h . chung , and s. park , phys .
93 * , 077206 ( 2004 ) .
h. mori , prog .
phys . * 34 * , 399 ( 1965 ) .
t. n. tommet and d. l. huber , phys .
b * 11 * , 1971 ( 1975 ) .
j. oitmaa , m. plischke , and t. a. winchester , phys .
b * 29 * , 1321 ( 1984 ) .
o. a. starykh , a. w. sandvik , and r. r. p. singh , phys .
b * 55 * , 14953 ( 1997 ) .
hoyos , a.p .
vieira , n. laflorencie and e. miranda , phys .
b * 76 * , 174425 ( 2007 ) .
a. w. sandvik , phys .
b * 52 * , r9831 ( 1995 ) .
f. naef , x. wang , x. zotos , and w. von der linden , phys .
rev b * 60 * , 359 ( 1999 ) .
j. sirker , r. g. pereira , and i. affleck , phys . rev
* 103 * , 216602 ( 2009 ) .
g. theodorou and m. h. cohen , phys .
b * 13 * , 4597 ( 1976 ) .
o. motrunich , k. damle , and d. a. huse , phys .
b * 63 * , 134424 ( 2001 ) .
o. motrunich , k. damle , and d. a. huse , phys .
b * 63 * , 134424 ( 2001 ) . c. p. lindsey and g. d. patterson , j. chem . phys . * 73 * , 3348 ( 1980 ) .
d. johnston , phys .
rev b * 74 * , 184430 ( 2006 ) . | finite temperature local dynamical spin correlations @xmath0 are studied numerically within the random spin@xmath1 antiferromagnetic heisenberg chain .
the aim is to explain measured nmr spin
lattice relaxation times in , which is the realization of a random spin chain . in agreement with experiments
we find that the distribution of relaxation times within the model shows a very large span similar to the stretched exponential form .
the distribution is strongly reduced with increasing @xmath2 , but stays finite also in the high@xmath2 limit .
anomalous dynamical correlations can be associated to the random singlet concept but not directly to static quantities .
our results also reveal the crucial role of the spin anisotropy ( interaction ) , since the behavior is in contrast with the ones for xx model , where we do not find any significant @xmath2 dependence of the distribution .
one dimensional ( 1d ) quantum spin systems with random exchange couplings reveal interesting phenomena fundamentally different from the behavior of ordered chains .
since the seminal studies of antiferromagnetic ( afm ) random heisenberg chains ( rhc ) by dasgupta and ma @xcite using the renormalization group approach and further development by fisher @xcite , it has been recognized that the quenched disorder of exchange couplings @xmath3 leads at lowest energies to the formation of random singlets with vanishing effective @xmath4 at large distances .
the consequence for the uniform static susceptibility @xmath5 is the singular curie type temperature ( @xmath2 ) dependence , dominated by nearly uncoupled spins at low@xmath2 and confirmed by numerical studies of model systems @xcite , as well by measurements of @xmath6 on the class of materials being the realizations of rhc physics , in particular the mixed system @xcite .
recent measurements of nmr spin lattice relaxation times @xmath7 in @xcite reveal a broad distribution of different @xmath7 resulting in a nonexponential magnetization decay being rather of a stretched exponential form . in connection to this
the most remarkable is the strong @xmath2 dependence of the @xmath7 span becoming progressively large and the corresponding distribution non gaussian at low@xmath2 .
it is evident that in a random system @xmath7 , which is predominantly testing the local spin correlation function @xmath8 , becomes site @xmath9 dependent and we are therefore dealing with the distribution of @xmath10 leading to a nonexponential magnetization decay .
theoretically the behavior of dynamical spin correlations in rhc has not been adequately addressed so far .
there is ( to our knowledge ) no established model result and moreover no clear prediction for the behavior of dynamical ( @xmath11 ) spin correlations at @xmath12 in rhc .
it seems plausible that the low@xmath2 behavior should follow from the random singlet concept and its scaling properties , discussed within the framework of the renormalization
group approaches @xcite .
still , the relation to singular static correlations as evidenced , e.g. , by @xmath13 diverging at @xmath14 , and low- @xmath15 dynamical correlations is far from clear .
one open question is also the qualitative similarity to the behavior of the random anisotropic xx chain invoked in several studies @xcite .
the latter system is equivalent to more elaborated problem of noninteracting ( ni ) spinless fermions with the off diagonal ( hopping ) disorder @xcite . in the following we present results for the dynamical local spin correlation function @xmath0 ,
in particular for its limit @xmath16 relevant for the nmr @xmath7 , within the afm rhc model for @xmath12 , obtained using the numerical method based on the density matrix renormalization group ( dmrg ) approach @xcite . at high
@xmath17 , distribution of @xmath18 reveals a modest but finite width qualitatively similar both for the isotropic and the xx chain . on the other hand ,
the low@xmath2 variation established numerically is essentially different . while for the xx chain there is no significant @xmath2 dependence , results for the isotropic case reveal at low @xmath19 a very large span of @xmath18 values and corresponding @xmath10 , qualitatively and even quantitatively consistent with nmr experiments @xcite .
we study in the following the 1d spin@xmath1 model representing the afm rhc , @xmath20 where @xmath21 are random and we will assume their distribution as uncorrelated and uniform in the interval @xmath22 , with the width @xmath23 as the parameter .
in the following we will consider predominantly the isotropic case @xmath24 , but as well the anisotropic xx case with @xmath25 .
the chain is of the length @xmath26 with open boundary conditions ( o.b.c . ) as useful for the dmrg method .
we further on use @xmath27 as the unit of energy as well as @xmath28 . our aim is to analyse the local spin dynamics in connection with the nmr spin
lattice relaxation @xcite . in a homogeneous system
the corresponding relaxation rate @xmath29 is expressed in terms of the @xmath30dependent spin correlation function , @xmath31 where @xmath32 involve hyperfine interactions and nmr form factors @xcite . in the supplement
@xcite we show that the dominant dynamical @xmath33 contribution at low@xmath2 is coming from the regime @xmath34 . therefore the variation @xmath32 is not essential and the rate depends only on the local spin correlation function @xmath35 . in a system with
quenched disorder the relaxation time becomes site dependent , i.e. @xmath10 , hence we study in the following the local correlations @xmath0 and the distribution of local limits @xmath36 and related relaxation times @xmath37 where @xmath38 in order to reduce finite size effects
we study large systems employing the finite temperature dynamical dmrg ( ftd dmrg ) @xcite method to evaluate the dynamical @xmath0 , eq . .
to reduce edge effects we choose the local site @xmath9 to be in the middle of the chain , @xmath39 .
the distribution of @xmath18 is then calculated with @xmath40 different realizations of the system with random @xmath21 .
more technical detail on the calculation can be found in the supplement @xcite .
we start the presentation of results with typical examples of @xmath0 . in fig .
[ spect ] we show calculated spectra for system with @xmath41 sites , @xmath42 , @xmath24 , and three different realizations of @xmath21 , i.e. the homogeneous system with @xmath43 and two configurations with @xmath44 .
spectra for the uniform system are broad and regular at @xmath45 agreeing with those obtained with other methods @xcite , while @xmath0 for random case strongly depend apart from @xmath46 also on the local @xmath47 .
in particular , spectra with both @xmath48 and @xmath49 small have large amplitude at the relevant @xmath50 , while spectra with one large @xmath48 or @xmath49 have most of the weight at high@xmath15 and small amplitude at @xmath50 ( elaborated further in the conclusions ) . for the following analysis
it is important that @xmath16 can be extracted reliably . for different configurations of @xmath21 .
shown are spectra for the homogeneous case @xmath51 and two configurations with @xmath52 , calculated for @xmath42 and @xmath41 sites . ]
_ results for @xmath17 : _ before displaying results for most interesting @xmath53 regime , we note that even at @xmath54 one can not expect a well defined @xmath55 but rather a distribution of values .
one can understand this by studying analytically local frequency moments within the high@xmath2 expansion and using the mori s continued fraction representation @xcite with the gaussian type truncation at the level of @xmath56 @xcite ( see @xcite for more details ) . in the inset of fig .
[ momex ] we present the high@xmath2 result for @xmath57 and compare it with the numerical results evaluated for @xmath58 .
several conclusions can be drawn from results presented on fig .
[ momex ] : ( a ) the agreement of pdf@xmath59 obtained via the analytical approach and numerical ftd
dmrg method is satisfactory having the origin in quite broad and featureless spectra @xmath0 at @xmath17 .
still we note that median value of @xmath18 ( @xmath60 ) differ between both approaches and that for @xmath61 ( unlike @xmath62 ) contribution of @xmath63 can become essential @xcite .
( b ) pdf@xmath59 becomes quite asymmetric and broad for @xmath64 .
( c ) consequently , also the distribution of local relaxation times pdf@xmath65 has finite but modest width for @xmath66 .
this seems in a qualitative agreement with nmr data for , where the width was hardly detected at high@xmath2 @xcite .
at @xmath67 evaluated using the moment expansion for different @xmath46 .
inset : comparison of analytical and ftd dmrg result for @xmath68 , @xmath69 with full basis and averaged over @xmath70 realizations . ] _ results for @xmath71 : _ more challenging is the low@xmath2 regime which we study using the ftd
dmrg method for typically @xmath41 and @xmath72 .
besides the isotropic case @xmath73 , we investigate for comparison also the xx model ( @xmath25 ) . as the model of ni fermions with the off diagonal disorder @xcite
it can be easily studied via full diagonalization on much longer chains with @xmath74 .
pdf for @xmath53 can become very broad and asymmetric .
hence , we rather present results as the cumulative distribution @xmath75 .
further we rescale @xmath76 values to the median defined as @xmath77 .
results for cdf@xmath59 are presented in fig .
[ cdf ] . note that @xmath78 .
panels in fig .
[ cdf ] represent results for the isotropic case @xmath24 with ( a ) fixed @xmath79 and varying @xmath80 , while in ( b ) @xmath81 is fixed and @xmath82 .
inset of fig .
[ cdf]b displays the @xmath2 dependence ( for fixed @xmath52 ) of cdf for the xx chain . .
shown are ftd dmrg results for @xmath24 : ( a ) for fixed @xmath79 and various @xmath46 , ( b ) for fixed @xmath52 and various @xmath83 .
inset of ( b):full diagonalization results for @xmath25 , @xmath52 and various @xmath2 . ]
we first note that within the xx chain cdf@xmath59 are essentially @xmath2 independent .
this appears as quite a contrast to , e.g. , static @xmath6 which exhibits a divergence at @xmath84 @xcite .
results for the isotropic case @xmath24 in figs .
[ cdf]a , b are evidently different .
the span in cdf becomes very large ( note the logarithmic scale ) either by increasing @xmath85 at fixed @xmath2 or even more by decreasing @xmath2 at fixed @xmath85 . from the corresponding pdf
one can calculate the relaxation function @xmath86 , which is in fact the quantity measured in the nmr as a time
dependent magnetization recovery @xcite . as in experiment
the large span in our results for low@xmath2 can be captured by a stretched exponential form , @xmath87 $ ] , where @xmath88 and @xmath89 are parameters to be fitted for particular pdf@xmath59 and corresponding @xmath90 .
it is evident that @xmath91 means large deviations from the gaussian like form , and in particular very pronounced tails in pdf@xmath59 , both for @xmath92 as well as a singular variation for @xmath93 . in the latter regime @xmath94 can deviate substantially from average of local @xmath95 .
it should be also noted that stretched exponential form , is the simplest one capturing the large span of @xmath18 values .
it is also used in the experimental analysis @xcite , but the corresponding pdf@xmath59 reveal somewhat enhanced tails for @xmath96 relative to calculated ones in fig .
[ cdf]a , b , and the opposite trend for @xmath97 .
this suggests possible improvements and description beyond stretched exponential form , which we leave as a future challenge .
more details can be found in the supplement @xcite .
results for the fitted exponent @xmath98 for @xmath24 as extracted from numerical pdf@xmath59 for various @xmath46 are shown in fig .
[ fit]a .
they confirm experimental observation @xcite of increasing deviations from simple exponential variation ( @xmath99 ) for @xmath19 . while for @xmath100 , @xmath101 for modest @xmath102 , low@xmath2 values can reach even @xmath103 at lowest reachable @xmath104 .
note that in such a case values of @xmath18 are distributed over several orders of magnitude . of interest for the comparison with experiment
is also the @xmath2 variation of fitted @xmath94 .
results are again essentially different for @xmath25 and @xmath24 .
@xmath89 ( as well @xmath60 ) for @xmath25 follows well the korringa law @xmath105 for @xmath106 , as usual for the system of ni fermions with a constant density of states ( dos ) ( divergent dos at @xmath107 could induce a logarithmic correction ) . on the other hand , for the isotropic ( @xmath24 ) chain with no randomness
@xmath108 it should follow @xmath109 for @xmath110 @xcite .
similar behavior is observed for weak disorder @xmath111 as shown in fig .
[ fit]b .
however , with increasing randomness @xmath46 , @xmath94 becomes more @xmath2 dependent and increases with @xmath2 .
such @xmath2 dependence in the rhc of @xmath94 is , although in agreement with experiment , in apparent contrast with diverging @xmath112 .
this remarkable dichotomy between static and dynamical @xmath33 behaviour can be reconciled by the observation that in a random system @xmath113 reveals besides the regular part also a delta peak at @xmath114 ( not entering @xmath29 ) , which can be traced back to diagonal matrix elements @xcite being an indication of a nonergodic behaviour ( at least at low@xmath2 ) .
note that more frequently studied static @xmath115 ( equal - time correlation ) @xcite represents a sum rule containing both parts .
also , the relation @xmath116 in spite of divergent @xmath117 leads to vanishing @xmath118 at @xmath14 only slower than linearly @xcite . vs. @xmath2 obtained from pdf@xmath59 data for different @xmath46 and isotropic case @xmath24 .
( b ) @xmath2 dependence of fitted @xmath94 for @xmath24 and different @xmath46 . ] as a partial summary of our results , we comment on the relation to the experiment on @xcite .
the spin chain is in this case assumed to be random mixture of two different values @xmath119 k , @xmath120 k , which correspond roughly to our @xmath121 ( fixing the same effective width ) and @xmath122 k. taking these values , our results for @xmath98 as well as @xmath123 agree well with experiment . in particular we note that at lowest @xmath124 our calculated @xmath125 for @xmath126 matches the measured one .
some discrepancy appears to be a steeper increase of measured @xmath98 towards the limiting @xmath99 coinciding with observed very narrow pdf@xmath65 which remains of finite width in our results even for @xmath66 as seen in fig .
[ momex ] .
as far as calculated @xmath123 vs. nmr experiment is concerned we note that taken into account the normalization of average @xmath3 disordered system reveals at @xmath84 smaller @xmath94 than a pure one consistent with the experiment @xcite . in agreement with the experimental analysis
is also strong @xmath2 variation of @xmath94 at low@xmath2 in disordered system in contrast to a pure one . our results on the local spin relaxation @xmath0 and
in particular its @xmath2 dependence can not be directly explained within the framework of existing theoretical studies and scaling approaches to rhc @xcite .
our study clearly shows the qualitative difference in the behavior of the xx chain and the isotropic rhc . while in the former model mapped on ni electrons
, @xmath2 does not play any significant role on pdf@xmath59 as seen in inset of fig .
[ cdf]b , @xmath24 case shows strong variation with @xmath124 .
it is plausible that the difference comes from the interaction and many body character involved in the isotropic rhc . to account for that we design in the following a simple qualitative argument .
the behavior of @xmath127 at low@xmath2 is dominated by transitions between low lying singlet and triplet states which become in a rhc nearly degenerate following the scaling arguments with effective coupling @xmath128 for more distant spins and reflected in diverging @xmath129 @xcite .
such transitions are relevant at @xmath33 behavior as presented in fig .
[ spect ] .
moreover , local @xmath130 exhibit large spread due to the variations in the local environment .
let us for simplicity consider the symmetric heisenberg model on four sites ( with o.b.c . ) with a stronger central bond @xmath131 and @xmath132 .
it is then straightforward to show that the lowest singlet
triplet splitting is strongly reduced , i.e. @xmath133 where @xmath134 .
within the same model one can evaluate also the ratio between two different amplitudes of @xmath135 , on sites @xmath136 neighboring the weak and strong bond , @xmath137 the relation shows that the span between largest and smallest amplitudes increases as @xmath138 .
continuing in the same manner the scaling procedure for afm rhc @xcite for a long chain the smallest effective coupling between further spins @xmath4 vanishes at @xmath139 and @xmath140 , so that one expects @xmath141 for @xmath142 .
on the other hand , for @xmath12 the scaling should be cut off at @xmath143 at least for @xmath24 , finally leading to the strong @xmath144 dependence ( @xmath145 ) . in the summary
, we have reproduced qualitatively main experimental nmr results on mixed system including anomalously wide distribution of relaxation rates , together with @xmath2 dependencies of experimental parameters ( @xmath94 , @xmath88 ) and provide microscopic explanation with the help of the random - singlet framework .
our qualitative conclusions on the rhc do not change by changing @xmath146 ( adding finite field in the fermionic language ) or even reducing @xmath147 provided that @xmath148 ( see supplement @xcite ) .
we also comment on striking difference between static and dynamic quantities and observed deviations from stretched exponential phenomenology .
23 s .- k .
ma , c. dasgupta , and c .- k .
hu , phys .
rev .
lett . * 43 * , 1434 ( 1979 ) . c. dasgupta and s .- k .
ma , phys .
rev .
b * 22 * , 1305 ( 1980 ) . d. s. fisher , phys . rev .
b * 50 * , 3799 ( 1994 ) .
j. e. hirsch , phys .
rev .
b * 22 * , 5355 ( 1980 ) .
a. zheludev , t. masuda , g. dhalenne , a. revcolevschi , c. frost , and t. perring , phys .
rev .
b * 75 * , 054409 ( 2007 ) .
t. shiroka , f. casola , v. glazkov , a. zheludev , k. pra , h.r .
ott , and j. mesot , phys .
rev .
lett . * 106 * , 137202 ( 2011 ) .
f. casola , t. shiroka , v. glazkov , a. feiguin , g. dhalenne , a. revcolevschi , a. zheludev , h .-
r .
ott , and j. mesot , phys .
rev .
b * 86 * , 165111 ( 2012 ) .
e. westerberg , a. furusaki , m. sigrist , and p. a. lee , phys .
rev .
b * 55 * , 12578 ( 1997 ) .
o. motrunich , k. damle , and d. a. huse , phys .
rev .
b * 63 * , 134424 ( 2001 ) .
l. n. bulaevskii , zh .
eksp .
. fiz . * 62 * , 725 ( 1972 ) .
g. theodorou and m. h. cohen , phys .
rev .
b * 13 * , 4597 ( 1976 ) .
t. p. eggarter and r. riedinger , phys .
rev .
b * 18 * , 569 ( 1978 ) .
j. kokalj and p. prelovek , phys .
rev .
b * 80 * , 205117 ( 2009 ) .
see suplemetary material for more details .
for a recent review , see p. prelovek and j. bona , in _ strongly correlated systems - numerical methods _
, edited by a. avella and f. mancini ( springer series in solid
state sciences 176 , berlin , 2013 ) , pp .
129 .
u. schollwck , rev .
mod .
phys . * 77 * , 259 ( 2005 ) .
f. naef , x. wang , x. zotos , and w. von der linden , phys .
rev b * 60 * , 359 ( 1999 ) .
h. mori , prog .
theor .
phys . * 34 * , 399 ( 1965 ) .
j. oitmaa , m. plischke , and t. a. winchester , phys .
rev .
b * 29 * , 1321 ( 1984 ) .
t. n. tommet and d. l. huber , phys .
rev .
b * 11 * , 1971 ( 1975 ) . o. a. starykh , a. w. sandvik , and r. r. p. singh , phys . rev .
b * 55 * , 14953 ( 1997 )
. a. w. sandvik , phys .
rev .
b * 52 * , r9831 ( 1995 ) .
j. a. hoyos , a. p. vieira , n. laflorencie and e. miranda , phys .
rev .
b * 76 * , 174425 ( 2007 ) . * for `` local spin relaxation within random heisenberg chain '' * [ [ section ] ] - in this section we present in more detail the numerical method , finite temperature dynamical dmrg ( ftd dmrg ) . the method is a variation of a zero temperature ( @xmath139 ) dmrg @xcite , with targeting of the ground state or ground state density matrix @xmath149 generalized to targeting of the finite@xmath2 density matrix @xmath150 , @xcite .
similar generalization is applied to targeting of the operator on the ground state . from such targets ,
the reduced density matrix is calculated and then truncated in the standard dmrg like manner for basis optimization .
all quantities , that need to be evaluated at finite@xmath2 , are calculated with the use of finite temperature lanczos method ( ftlm ) @xcite , which in ftd dmrg replaces @xmath139 lanczos method used in the standard dmrg algorithm .
the method is most efficient at low@xmath2 and for low frequencies , where basis can be efficiently truncated and only small portion ( @xmath151 basis states ) of the whole basis for block can be kept . in this regime large system sizes can be reached .
the truncation error becomes larger at higher@xmath2 , and one needs to either use larger @xmath151 , or reduce system size , which is legitimate approach , since finite size effects are smaller at higher@xmath2 due to reduced correlation lengths .
we typically keep @xmath152 basis states in the dmrg block and use systems with length @xmath153 at low @xmath154 , while for @xmath155 smaller systems are employed down to @xmath156 , for which full basis can be used .
we stress that randomness of @xmath21 reduces the truncation error since some larger values of @xmath21 induce strong tension for formation of a local singlet and therefore in turn reduces the entanglement on larger distances .
also the local operator , acting on the middle of the chain , where the local one site basis is not truncated , helps in this respect .
the quenched random @xmath21 are introduced into the dmrg procedure at the beginning of _ finite _ algorithm .
_ infinite _ algorithm is preformed for homogeneous system @xmath157 and the randomness of @xmath21 is introduced in the first sweep ( see fig .
[ sche ] for schematic presentation ) . in this way
the preparation of the basis in the _ infinite _ algorithm is performed just once and for all realizations of @xmath21s , while larger number of sweeps ( usually @xmath158 ) is needed to converge the basis within the _ finite _ algorithm for random @xmath21 .
after _
finite _ algorithm local dynamical spin structure factor @xmath0 at desired @xmath2 is calculated for the site in the middle of the chain within _ measurements _ part of dmrg procedure . )
representation of the beginning of the sweeping method in _ finite_dmrg algorithm , in which randomness is introduced .
open circle represents site of a local operator used to calculated local spin correlation function in the _ measurement _ part of the dmrg method . ] any spectra on finite system consists of separate @xmath159peaks , which we broaden by changing them into gaussian with small broadening @xmath160 and in this way obtain a smooth spectra .
since nmr relaxation rate is related to @xmath8 , we are interested in the limit @xmath33 , which should be contrasted with the singular @xmath161 . in order to avoid the problem of diagonal elements and keeping @xmath162
we perform the evaluation of @xmath163 in the magnetization sector @xmath164 , which in terms of spinless fermions corresponds to the canonical ensemble ( in the thermodynamic limit canonical and grand canonical give the same result ) and we remove @xmath165 peak ( see also section ) .
diagonal elements are however essential when evaluating static uniform susceptibility @xmath166 , which we show in fig .
[ chi ] .
it has been argued @xcite , that in 1d random heisenberg chain the density of low lying excitation is strongly increased , which is observed in diverging @xmath6 for @xmath167 .
our numerical results show similar behavior ( see fig .
[ chi ] ) , which agrees also with experiment @xcite . vs. temperature @xmath2 for various randomness @xmath46 .
for random system ( @xmath168 ) @xmath6 is strongly increased at low@xmath2 and agrees with the random singlet @xcite prediction ( black dotted line ) .
sudden drop of @xmath6 at low@xmath2 shown for @xmath51 and @xmath169 represents opening of finite size gap , which is strongly reduced in the random ( shown for @xmath170 ) system .
results are obtained with the finite temperature lanczos method @xcite on @xmath171 sites . ]
increased number of low lying excitations ( see fig .
[ chi ] ) also reduces the finite size effect , since , e.g. , finite size gap is reduced , and in this way also the temperature @xmath172 , below which the finite size effects become important .
therefore , smaller @xmath2 can be numerically reached in a random system .
[ [ section-1 ] ] - the local spin correlation function can be related to the ( local ) dynamical spin susceptibility by relation @xmath173=\chi_{nn}''(\omega)\ , , \label{fdt}\ ] ] with @xmath174\rangle\,.\ ] ] taking the high@xmath2 limit ( @xmath175 ) of eq .
one gets @xmath176 , which is so called relaxation function - symmetric with respect to @xmath114 , non negative function . note that due to symmetric
form of relaxation function all odd frequency moments , @xmath177 , are equal to zero .
the local spin correlation function can by expressed by the mori s continued fraction representation @xcite : @xmath178 where coefficient @xmath179 are cumulants of @xmath0 , i.e. @xmath180 , @xmath181 , @xmath182 .
@xmath177 are frequency moments of the local spectra , @xmath183 . for @xmath56 we chose a truncation @xmath184 , which assumes @xcite a gaussian like decay of correlation function , i.e. @xmath185 . the @xmath0 can be recovered from eq . by the relation @xmath186/\pi$ ] , leading to @xmath187 ^ 2 + \left(\omega^{2}-\delta_{1n}\right)^2}\ , .
\label{hislo}\ ] ] note that eq .
gives the first three nonzero ( @xmath188 ) frequency moments @xmath177 correctly , independent of a choice of @xmath189 .
frequency moments @xmath177 of @xmath0 can be evaluated analytically for @xmath190 , e.g. , @xmath191 , @xmath192[h , s^{z}_{n}]\rangle$ ] , etc . for zero magnetization , @xmath193 , where @xmath194 ( @xmath195 ) is number of up ( down ) spins , the first three nonzero moments of the order of @xmath196 are : @xmath197 between analytical high@xmath2 expansion result and numerical ftd
dmrg result ( @xmath69 , @xmath198 , full basis ) for three realization of @xmath21 . ] in fig .
[ specht ] we present comparison of high@xmath2 expansion result and ftd dmrg result ( @xmath69 , @xmath198 , full basis ) for @xmath0 and three realizations of @xmath21 .
one can see that the agreement is good for actual finite size system .
it should be , however , noted that @xmath199 contribution ( leading to finite size corrections ) can become essential for @xmath61 @xcite .
as a final remark of this section we comment on the probability distribution function ( pdf ) of @xmath36 presented in fig . 2 in the main text .
assuming the uniform distribution of @xmath21 , @xmath200 the pdf@xmath59 can be can be generated from expression @xmath201 the pdf - s presented in fig . 2 ( main text ) where obtained from @xmath202 realizations of @xmath21 .
[ [ section-2 ] ] - and times @xmath37 . shown
are ftd
dmrg results for various @xmath203 and : ( a ) for @xmath204 , @xmath164 and @xmath52 , ( b ) for @xmath24 , @xmath205 and @xmath52 , while ( c ) are diagonalization results for @xmath25 , @xmath52 and @xmath206 . ] in fig .
[ cdf_supp ] we show that our main conclusions stay valid also in a more general case , such as for @xmath207 and for finite magnetization ( @xmath208 ) , where considerable @xmath2 dependence of distribution with large spread is observed . in the last panel of fig .
[ cdf_supp ] we show that the distribution for noninteracting case ( xx model ) stays @xmath2 independent even in a finite magnetic field @xmath209 or for finite magnetization . |
in the past decade in four independent articles it was observed that the billiard orbit of any point which begins perpendicular to a side of a polygon and at a later instance hits some side perpendicularly retraces its path infinitely often in both senses between the two perpendicular collisions and thus is periodic .
the earliest of these articles is a numerical work of ruijgrok which conjectures that every triangle has perpendicular periodic orbits @xcite . in 1992 boshernitzan
@xcite and independently galperin , stepin and vorobets @xcite proved that for any rational polygon , for every side of the polygon , the billiard orbit which begins perpendicular to that side is periodic for all but finitely many starting points on the side .
finally for an irrational right triangle cipra , hansen and kolan have considered points which are perpendicular to one of the legs of the triangle .
they showed that for almost every such point the billiard orbit is periodic @xcite . here
the almost everywhere statement is with respect to the length measure on the side considered .
the method of their proof in fact implies a stronger result , namely that in any ( generalized ) parallelogram or right triangle , for each direction @xmath0 , the set of points @xmath1 whose orbit starts in the direction @xmath0 and never returns parallel to itself has lebesgue measure 0 @xcite .
the result of cipra _ et .
_ on periodic orbits follows from the above observation .
their computer simulation indicated that perhaps one can improve the almost everywhere and that is the starting point of this research . in theorem
[ thm1 ] we show that in any ( generalized ) parallelogram , for each direction @xmath0 , the set of points whose orbit starts in the direction @xmath0 and never returns parallel to itself has lower box dimension at most one half . a corollary to this theorem
is that the set of points whose orbit starts perpendicular to a leg of a right triangle whose orbit is not periodic has lower box dimension at most one half .
we then turn to the question , whether there are directions @xmath0 for which we can improve the constant 1/2 in theorem [ thm1 ] . in theorem [ thm2 ]
we show that if @xmath2 is the approximation order of @xmath0 by @xmath3 , then in fact the lower box dimension of @xmath1 is at most @xmath4 .
we have two applications of theorem [ thm2 ] .
we consider the set @xmath5 of directions @xmath0 for which the set of points which never return parallel to themselves have lower box dimension at most @xmath6 $ ] . in theorem [ thmcs ]
we prove that the hausdorff dimension of @xmath5 is at least @xmath7 , the set @xmath5 is residual and has box dimension 1 . fix a direction @xmath8 and consider the set @xmath9 of angles @xmath3 such that for any generalized parallelogram with angle @xmath3 the set of points in direction @xmath8 which never return parallel to themselves has dimension at most @xmath10 $ ] .
we also consider the set @xmath11 of right triangles for which the set of non - periodic points which are perpendicular to a fixed leg of the triangle has dimension at most @xmath12 $ ] . in theorem [ thmds ]
we prove that the hausdorff dimension of @xmath13 and @xmath11 are at least @xmath14 , these sets are residual and have box dimension 1 . the proofs of theorems [ thmcs ] and [ thmds ] come from purely number theoretic arguments .
the approximation of @xmath15 by @xmath16 is a classical area of research in number theory which is referred to as inhomogeneous diophantine approximation .
a classical result in this direction is the theorem of minkowski @xcite which states that if @xmath15 is not in the orbit of @xmath16 then @xmath17 has infinitely many integer solutions @xmath18 and the constant 1/4 can not be improved in general . here
@xmath19 is the standard distance on @xmath20 . in the spirit of minkowski s
theorem we consider the set @xmath21 for any @xmath22 in theorem [ thmam ] we prove that the hausdorff dimension of this set is @xmath23 .
theorem [ thmcs ] follows easily from a similar statement for approximations along an arithmetic subsequence .
another classical result in number theory is jarnik s theorem on the hausdorff dimension of well approximable irrational numbers @xcite . in the spirit of jarnik s
theorem we consider the set @xmath24 for any @xmath2 levesley has proven that the hausdorff dimension of this set is @xmath25 @xcite .
the proof of theorem [ thmds ] uses a similar inhomogeneous jarnik result generalized to arithmetic subsequences .
the structure of the paper is as follows . in section [ section2 ]
we describe the needed background results on dimension theory and billiards in section [ section3 ] we state the purely number theoretic results and prove the new results . in section [ section4 ]
we state and prove the new billiard results . finally we remark that a generalization in a different direction of cipra _ et .
s results was obtained by one of us in @xcite .
let @xmath26 be a subset of @xmath27 .
let @xmath28 denote the minimal number of @xmath29 balls needed to cover @xmath26 . for a subset @xmath26 of @xmath27 , the lower box dimension of @xmath26 , denoted by @xmath30
is given by @xmath31 the upper box dimension @xmath32 is defined similarly , replacing the @xmath33 by @xmath34 . if @xmath35 and @xmath30 both exist and are equal , we define the box dimension of @xmath26 to be this value , and write @xmath36 = @xmath35 = @xmath30 . for a subset @xmath37 of @xmath27 , we let @xmath38 denote the diameter of the set @xmath37 .
let @xmath39 $ ] .
the @xmath40-dimensional hausdorff measure @xmath41 of a subset @xmath26 of @xmath27 is defined by the following limit of covering sums : @xmath42 it is easy to see that there exists a unique @xmath43 such that @xmath44 the unique number @xmath45 given by equation is defined to be the hausdorff dimension of @xmath26 and is denoted by @xmath46 .
standard arguments give that for a subset @xmath26 of @xmath27 , @xmath47 there are examples which show that these inequalities may be strict . the box dimension can also be defined in terms of covering sums .
the only change being that the covering sets all have equal diameter .
we note that in order to estimate the box dimension , it suffices that the diameters of the covering sets tend to @xmath48 along a geometric sequence .
lastly , we define the hausdorff dimension of a measure : let @xmath49 be a borel probability measure on @xmath50
. then the hausdorff dimension of the measure @xmath49 is defined by @xmath51 we remark that the dimension of a measure is clearly always less than or equal to the dimension of its ( borel ) support .
there is a well known method of computing lower bounds on the dimension of a measure or a set .
let @xmath52 denote the ball of radius @xmath53 centered at @xmath54 .
[ lsy ] if for some finite measure @xmath49 @xmath55 then the dimension of the measure is at least @xmath40 .
a survey of the methods and results in dimension theory can be found in @xcite . consider a polygon @xmath56 . a billiard ball , i.e. a point mass , moves inside @xmath57 with unit speed along a straight line until it reaches the boundary @xmath58 , then instantaneously changes direction according to the mirror law : `` the angle of incidence is equal to the angle of reflection , '' and continues along the new line .
for the moment we assume that @xmath57 is convex
. we will describe the billiard map in @xmath57 as a transformation of the set @xmath50 of rays which intersect @xmath57 .
we parameterize @xmath50 via two parameters .
the first parameter is the angle @xmath0 between the ray and the positive @xmath54axis . to define the second parameter
consider the perpendicular cross section @xmath59 to the set of rays whose angle is @xmath0 .
the set @xmath59 is simply an interval , thus the second parameter is then the arc - length on @xmath59 for a fixed orientation . for a non - convex polygon we must differentiate the portion of rays which enter and leave the polygon several times .
the above construction can be done locally , yielding the set @xmath59 which consists of a finite union of intervals .
let @xmath60 be the unnormalized length measure on @xmath59 .
the billiard map @xmath61 takes the ray , which contains a segment of a billiard trajectory oriented by the direction of its motion , to the ray which contains the next segment of this trajectory after the reflection in the boundary .
a polygon @xmath57 is called a generalized parallelogram if every side of @xmath57 is parallel to one of two fixed directions .
we suppose that one of these directions is the direction of the positive @xmath54-axis .
then there is a unique @xmath62 such that the angle between any pair of sides is @xmath63 or @xmath64 .
we suppose throughout that @xmath65 , that is that @xmath57 is not a rational polygon .
we remark that the billiard in a right triangle is equivalent to the billiard in a rhombus consisting of four copies of the triangle via the process of unfolding .
the billiard map in a generalized parallelogram has the following special properties ( see figure [ fig1 ] ) .
the interval @xmath59 can be partitioned into three sets @xmath66 and @xmath67 which consist of a finite union of intervals such that @xmath68 and @xmath69 and each interval in @xmath66 and @xmath67 is mapped isometrically onto its image .
the length measure @xmath60 is @xmath70invariant .
furthermore there is a constant @xmath71 such that @xmath72 we call the endpoints of the intervals of @xmath73 and @xmath67 the singularities of @xmath74 .
let @xmath75 and @xmath76 . the set @xmath77 is called the @xmath78th level of @xmath79 .
the set @xmath79 is @xmath74-invariant with infinite invariant measure @xmath60 .
there are several good survey articles on billiards @xcite .
the structure discussed here is described in more detail in @xcite .
in this section we work with the circle @xmath80 of unit length rather than the circle of length @xmath81 . the role of @xmath0 and @xmath3 in the billiard sections
is played by @xmath82 and @xmath83 .
let @xmath84 let @xmath85 be fixed such that @xmath86 and let @xmath87 note that @xmath88 and @xmath89 .
minkowski has shown that for @xmath15 not in the orbit of @xmath16 the inequality @xmath90 has infinitely many solutions and in general the constant @xmath91 is optimal @xcite . a simple borel
cantelli argument tells us that directions which can be approximated better than in the statement of minkowski s theorem have zero measure .
[ borelcantelli ] suppose that @xmath2 and @xmath92 .
the lebesgue measure of @xmath93 is 0 while @xmath94 .
the sets @xmath95 are residual . hence , the set @xmath96 is also residual and has box dimension 1 . for simplicity ,
if @xmath97 we denote the interval @xmath98 by @xmath99 .
let @xmath100 , this yields @xmath101 and @xmath102 thus since @xmath103 and @xmath104 are dense @xmath95 is residual .
thus the set @xmath105 is dense , hence it s lower and upper box dimension are 1 . to see that the lebesgue measure of @xmath93 is 0 note that the sequence @xmath106 has finite sum for @xmath2 .
hence , by the borel
cantelli lemma almost no point @xmath15 is contained in more than a finite number of intervals @xmath107 . due to this proposition
we consider the dimension of the set @xmath93 of @xmath15 s with better approximations .
[ thmam ] for any @xmath2 and @xmath108 @xmath109 we begin by proving the upper bound . for any @xmath110 equation implies @xmath111 we turn to the lower bound . for @xmath112 let @xmath113 $ ] and
@xmath114 given any integer sequence @xmath115 which meets all residue classes infinitely often for both positive and negative @xmath116 we have that @xmath117 to prove the lower bound we will use induction to construct a sequence @xmath116 which is simply a subsequence of the denominators @xmath118 of the continued fraction approximants . given the sequence @xmath115 we will construct a measure @xmath119 supported on the set @xmath120 whose dimension is @xmath23 .
the idea of our construction is as follows : the faster the sequence @xmath116 increases the better the points @xmath121 are distributed along the circle . if a point @xmath54 belongs to @xmath93 then it has to be infinitely often in intervals of the form @xmath122 .
so for some large @xmath123 it has to fall into @xmath124 for some @xmath125 .
the set @xmath126 forces the point to be in the `` right half '' of the collection of intervals @xmath124 , where @xmath127 runs from 0 to @xmath128 .
if the sequence @xmath115 is sufficiently sparse this restriction turns out to be mild enough to maintain the dimension of @xmath93 .
let @xmath129 and assume that the sequence @xmath130 is constructed up to @xmath131 .
then the set @xmath132 consists of a finite number of open intervals @xmath133 satisfying @xmath134 with equality if and only if they are completely contained in @xmath135 .
we will drop the hat from the notation for any interval for which equality holds .
we consider the set @xmath136 and @xmath137 clearly @xmath138 .
assume that inductively @xmath139 is already constructed for @xmath140 .
it is well known that for any @xmath141 the sequence @xmath142 is well distributed in @xmath20 , i.e. for any continuous function @xmath143 and any @xmath110 there is a number @xmath144 such for any @xmath145 @xmath146 for any @xmath147 .
thus if @xmath148 we can choose @xmath116 to be the least positive @xmath149 such that the following two conditions hold : @xmath150 and @xmath151 otherwise if @xmath152 we choose @xmath153 satisfying conditions and .
this finishes the construction of the sequence @xmath115 .
clearly , the sequence @xmath115 meets all residue classes infinitely often for positive and negative @xmath116 .
we next construct a measure on the set @xmath126 with the desired dimension .
we assume that @xmath154 consists of the whole circle and thus has length 1 .
we begin by a recursive definition of an outer measure @xmath155 .
we put @xmath156 and @xmath157 we will now compute an upper bound on the outer measure . using equations and we have @xmath158 the fast growth rate on @xmath116 assumed by equation implies that @xmath159 as @xmath160 .
thus it is clear that @xmath155 satisfies kolmogorov s compatibility conditions and can be extended to a measure @xmath119 on @xmath126 .
we are now ready to prove that the dimension of the measure is at least @xmath161 .
equation implies @xmath162 where the @xmath163 are all positive constants .
equation implies that @xmath164 to use lemma [ lsy ] to conclude our result we must evaluate the ratio @xmath165 for the `` intermediate '' intervals @xmath166 , that is when the interval @xmath166 satisfies @xmath167 let @xmath168 be the neighborhood of radius @xmath169 around the set @xmath170 and let @xmath171 if @xmath172 then we denote by @xmath173 the unique such interval which is contained in @xmath166
. then @xmath174 now suppose @xmath175 . since @xmath176 is some continued fraction convergent @xmath177 we have @xmath178 we remark that this is the only place where we use that the @xmath116 is a subsequence of the continued fraction convergents .
equations and imply ( assuming that @xmath179 ) that @xmath180 for any @xmath110 , using inequality we can choose @xmath181 sufficiently large such that for any @xmath182 we have that @xmath183 thus @xmath184 where the last inequality holds for any sufficiently small @xmath185 . now lemma [ lsy ] implies our result . let @xmath186 let @xmath85 be fixed such that @xmath187 and let @xmath188 then @xmath189 and @xmath190 .
[ borelcantelli1 ] suppose that @xmath2 , and @xmath92 .
the lebesgue measure of @xmath191 is 0 while @xmath192 .
the sets @xmath191 are residual . hence , the set @xmath193 is also residual and has box dimension 1 .
let @xmath194 , this yields @xmath195 hence @xmath191 is residual .
the set @xmath191 is dense , thus it s lower and upper box dimension are 1 .
we have @xmath196 , hence , by the borel
cantelli lemma almost no point @xmath16 is contained in more than a finite number of intervals @xmath197 , i.e. the lebesgue measure of @xmath191 is 0 .
the following generalization of jarnik s classical theorem @xcite is a special case of a more general result which has been proven by levesley : [ thmbm ] @xcite for any @xmath2 @xmath198 we are going to improve levesley s theorem in this special setting .
the proof essentially follows his ideas and we do not claim any originality .
[ general ] for any @xmath2 and @xmath199 @xmath200 we will need the ( adapted to our case ) notion of an _ ubiquitous _ system .
for fixed @xmath201 let @xmath202 .
let @xmath203 with @xmath204 be given .
then the family of point sets @xmath205 is said to be _ ubiquitous _ with respect to @xmath206 if @xmath207 we can express the sets @xmath191 in the following way @xmath208 @xmath209 and @xmath210 we are going to use the following special case of the more general theorem 2 in @xcite proved by dodson , rynne and vickers [ dpv ] suppose that for each @xmath187 the family @xmath211 is ubiquitous with respect to @xmath206 . then @xmath212 levesley proved by using discrepancy estimates that the system @xmath213 is ubiquitous with respect to the function @xmath214 . noting that the sequence @xmath215 has the same discrepancy as the sequence @xmath216 we can follow the arguments of levesley and obtain that @xmath211 is ubiquitous with respect to the function @xmath217 . now theorem [ dpv ] implies theorem [ general ] .
unfolding a right triangle with angle @xmath221 around its right angle yields a rhombus with angle @xmath3 .
we consider a direction @xmath0 in the rhombus which is perpendicular to one of the legs of the triangle .
the orbits which start in @xmath59 and return to @xmath59 , when considered as orbits in the triangle are twice perpendicular to a side . as explained in the introduction
, such orbits must be periodic . thus applying theorem [ thm1 ] to this direction yields the following improvement of the theorem of cipra , hansen and kolan @xcite : [ lemma1 ]
fix an interval @xmath59 and a positive integer @xmath144 .
then there is a partition of @xmath73 into @xmath222 intervals for some positive constant @xmath223 such the forward orbit of each interval of the partition is an isometric mapping until one of the following happens 1 ) the orbit of the interval reaches level @xmath144 before returning to @xmath59 , or 2 ) the orbit of the interval returns to @xmath59 without having reached level @xmath144 .
the orbits of all these intervals are mutually disjoint until they possibly return to @xmath59 .
consider the set @xmath225 of the singularities of @xmath74 on levels 0,1 to @xmath144 .
the set @xmath225 has cardinality at most @xmath226 ( here @xmath223 can be taken to be twice the number of vertices of the polygon ) .
for each @xmath227 consider the first ( if any ) preimage which is on level 0 .
we remark the length of time for @xmath227 to be mapped to level 0 is not necessarily bounded .
this forms the indicated partition of @xmath73 .
the mapping @xmath228 is continuous on each element of the partition until it is mapped into level @xmath229 or returns to level 0 .
the disjointness follows from the invertability of @xmath74 . to see this
suppose that @xmath230 , @xmath231 but @xmath232 with @xmath233 . we can not have @xmath234 since then the point @xmath235 has two preimages . since @xmath236 the orbit of @xmath54 has returned to @xmath59 at time @xmath237 and
its orbit is disjoint from @xmath238 s before this time as claimed .
if one of the intervals ( call it @xmath239 ) never returned to @xmath59 or reached level @xmath144 then the set @xmath240 would consists of an infinite number of disjoint intervals of the same length , yet it would be contained in @xmath241 .
this later set consists of a finite union of intervals of finite length yielding the desired contradiction .
fix an arbitrary @xmath0 .
we will argue about the dimension of @xmath242 .
the argument for @xmath243 is similar with the proviso that since we use equation @xmath244 must be close to 0 for @xmath67 while for @xmath73 , @xmath245 must be close to @xmath3 . because of this proviso , since we need both @xmath242 and @xmath243 to be simultaneously small , rather than using the approximants by @xmath246 we will use the approximants by @xmath3 , using the even iterates for the set @xmath242 and the odd iterates for the set @xmath243 .
consider the set @xmath247 of those points @xmath248 whose forward orbit reaches @xmath249 before returning to @xmath59 .
we call this event @xmath54 exiting from level @xmath250 ( necessarily to level @xmath251 ) .
when the orbit of @xmath54 reaches level @xmath250 and is in the process of exiting level @xmath250 it must pass through the `` gate '' @xmath252 on level @xmath250 . using lemma [ lemma1 ] clearly at most @xmath226 of the intervals exit before returning to @xmath59 .
hence the set @xmath247 consists of at most @xmath226 intervals which we call @xmath253 .
when they exit level @xmath250 before returning to level 0 the sum of there lengths can not be more than the width of the gate on level @xmath250 through which they exit .
here we emphasize that we use the statement in lemma [ lemma1 ] that the exiting orbits are all disjoint .
thus the total length of the intervals exiting level @xmath250 is at most the total length of the gate of level @xmath250 , i.e. at most @xmath254 .
we use @xmath255 $ ] to denote the integer part of @xmath54 .
let @xmath256 be a sequence of positive numbers and let @xmath257 . a simple estimate yields ( see @xcite ) : @xmath258 + 1 \right ) \leq 3n \ .\ ] ] using inequality and the fact that @xmath222
, we have @xmath262 + 1 \right ) \left(\frac{k(|\sin(\t - 2n\a)|)}{j_n}\right)^{1/2+\e } \nonumber\\ & \leq \frac{3j_n}{(j_n)^{1/2+\e } } \left ( k(|\sin(\t - 2n\a)| \right ) ^{1/2+\e } \\ & \le ln^{1/2-\e}\left((\t-2n\a)\mod \pi \right)^{1/2+\e}\nonumber \end{aligned}\ ] ] for some constant @xmath263 . to get an effective cover we just need to find a sequence @xmath264 such that the inhomogeneous diophantine approximation @xmath265 is small .
it is well known that for any @xmath266 the inequality @xmath267 is solvable for infinitely many @xmath268 .
we remark that this is a much weaker statement than minkowski s theorem @xcite .
hence , we can find sequences @xmath269 such that @xmath270 for @xmath271 the right hand
side tends to zero as @xmath272 . to estimate the dimension of @xmath243 we need to use the second part of equation in equation . with this change
it is useful to consider the variable @xmath274 . noticing that @xmath275 and remarking that @xmath276 has infinitely many integer solutions we conclude that @xmath277 and thus that @xmath278 . in the previous section we proved a recurrence theorem for all directions @xmath0 . for our first results in this section
we consider special directions @xmath0 which are extremely well approximable .
for these directions we can conclude a stronger recurrence result .
[ thm2 ] consider a generalized parallelogram with angle @xmath3 and let @xmath0 be a direction such that @xmath279 for infinitely many @xmath280 and @xmath281 for infinitely many @xmath282 .
then the set @xmath1 has lower box counting dimension at most @xmath283 .
fix @xmath2 and consider a right triangle with angle @xmath221 , such that the approximations of @xmath284 by the orbit of @xmath3 in the sense of the previous theorem is of order @xmath49 .
then the set of points which are perpendicular to the base of the triangle and whose orbit is not periodic has lower box counting dimension at most @xmath285 .
next we will fix the direction and vary the generalized polygons . to do this
we assume that one of the sides of the polygon is fixed and we measure the direction @xmath0 with respect to this fixed direction .
we remark that our estimates on @xmath1 will depend only on the angle @xmath3 and no other parameters of the generalized parallelogram .
fix a direction @xmath8 .
consider the sets @xmath289 where perp is a direction perpendicular to one of the legs of the triangle .
[ thmds ] ( a ) for @xmath6 $ ] we have @xmath290 the set @xmath11 is residual and has box dimension 1 .
+ ( b ) for @xmath8 fixed and @xmath6 $ ] we have @xmath291 .
the set @xmath13 is residual and has box dimension 1 .
the results on hausdorff dimension are immediate consequences of theorem [ thm2 ] and theorem [ general ] with @xmath288 .
the residuality and box dimension results are immediate corollaries of proposition [ borelcantelli1 ] . | for a given rotation number we compute the hausdorff dimension of the set of well approximable numbers .
we use this result and an inhomogeneous version of jarnik s theorem to show strong recurrence properties of the billiard flow in certain polygons . |
the similarity of amino acids(@xmath0 ) is the basis of protein sequence alignment , protein design and protein structure prediction .
several scoring schemes have been proposed based on amino acid similarity .
the mutation data matrices of dayhoff @xcite and the substitution matrices of henikoff @xcite are standard choices of scores for sequence alignment and amino acid similarity evaluation
. however , these matrices , focusing on the whole protein database , pay little attention on protein secondary structures(@xmath1 ) .
how the amino acid similarity is influenced by different secondary structures is an interesting question .
furthermore , understanding the differences can help us in protein sequence analysis . despite efforts in uncovering the information encoded in the primary structure , we still can not read the language describing the final 3d fold of an active biological macromolecule . compared with the dna sequence , a protein sequence is generally much shorter , but the size of the alphabet is five times larger .
a proper coarse - graining of the 20 amino acids into fewer clusters for different conformation is important for improving the signal - to - noise ratio when extracting information by statistical means .
it is our purpose to propose a simple scheme to study amino acid similarity from amino acid string statistics .
information about the environment for an amino acid at a certain conformation state may be provided by statistics of residue strings or windows centered at the amino acid .
the success of window - based approaches such as gor @xcite for secondary structure prediction validates the use of such statistics .
we shall derive a measure for the difference of amino acid pairs based on the distance of probability distributions , and investigate how the difference is dependent on conformations .
our discussion will be heavily based on the distance between two probability distributions .
a well defined measure of the distance is the kullback - leibler ( kl ) distance or relative entropy @xcite , which , for two distributions @xmath2 and @xmath3 , is given by @xmath4 it corresponds a likelihood ratio , and , if @xmath5 is expanded around @xmath6 , its leading term is the @xmath7 distance : @xmath8 it is often to use the following symmetrized form for the kl distance @xmath9 .
\label{kl}\ ] ] the distributions to be considered here come from window statistics . for a given amino acid residue @xmath10 at the conformation state @xmath11 in a sequence @xmath12 , we take the string @xmath13 of width @xmath14 as a window .
denote by @xmath15 the count of residue @xmath16 at the @xmath17-th site from the center of such windows . as in gor
, only the conformation of the central residue is concerned .
a quantity derived from @xmath18 is @xmath19 which , as the total count of residue @xmath20 at the conformation @xmath11 , is independent of @xmath17 .
the conditional probability distribution @xmath21 is estimated as @xmath22 the weight matrix @xmath23 with its entries being @xmath21 is the so - called residue profile of @xmath20 at @xmath11 .
such profiles are used in window - based approaches , e.g. gor and artificial neural network algorithm @xcite .
we expect that on average the correlation between the central residue and an outer site decays when they become far apart in sequence . to examine the correlation
, we consider a large window width of 21 , i.e. @xmath24 , and take the ` noise ' background to be the following average : @xmath25.\ ] ] the kl distance @xmath26 provides a measure of the correlation between the central site and site @xmath17 .
as we shall see , for our purpose of amino acid comparison a narrow window of a strong correlation with width of 7 is used to describe amino acid enviroment . using distribution @xmath21 from window statistics to characterize amino acid residues
, we define the distance of residue pair @xmath20 and @xmath16 at the same conformation @xmath11 as the following sum of kl distances @xmath27 similarly , to explore the difference of the same residue @xmath20 at different conformations @xmath11 and @xmath28 , we may define the distance @xmath29 by means of the residue pair distances we can further study the classification of amino acids . with the kl distance , we may define the cluster distance in a way consistent with that for residue pairs .
for example , we characterize the cluster consisting of residues @xmath20 and @xmath16 by the ` coarse - grained ' probability @xmath30 we then may define the distance between this cluster and some other residues or clusters . with cluster distance defined , the cluster analysis can be used to reduce amino acid alphabets .
our analysis is performed on a data set taken from the database pdb_select@xcite of nonredundant protein sequences with known structures .
the sequences share amino acid identity less than 25% .
we keep only the non - membrane sequences with their lengths between 80 and 420 .
the secondary structure assignment is taken from the dssp database @xcite .
as in gor , we use the following reduction of the 8 dssp states to 4 states of helix(h ) , sheet(e ) , coil(c ) and turn(t ) : @xmath31 , @xmath32 , @xmath33 and @xmath34 .
the counts of each amino acid at the reduced four different conformation states are given in table 1 .
we first estimate probability distributions of residues for each central residue at a given conformation . at this step ,
the window width is 21 .
we then calculate distances @xmath35 of these distributions to their corresponding noise distributions .
the results are shown in figs . 1 to 4 ,
each of which is for one conformation of the central residue .
the 20 curves in each figure correspond to 20 central amino acids . due to the sample size difference , curves are not directly comparable . ( roughly speaking , under the null hypothesis of identical distribution the @xmath7 distance should be scaled with the sample size , so a small sample size would give a relatively large distance . ) however , a decay is clearly seen when the site @xmath17 become far away from the center . for more discussions on correlations we refer reader to @xcite . as seen from most curves of the figures , distances at the 6 sites nearest to the center are significantly larger than those at window border sites .
we shall use window width of 7 for further comparison of amino acids .
it is natural to expect that similar residues would have similar window statistics .
thus , the kl distance between two residue profiles provides a measure of their similarity , i.e. a small kl distance implies a large similarity .
we calculate the kl distance matrices @xmath36 for residue pairs at different conformations with formula ( [ dxy ] ) .
the results are given in tables 2 and 3 , where entries have been multiplied by a factor 200 .
with the distributions ( [ x+y ] ) defined for clusters , we further perform the simplest bottom - up approach of hierachical clustering for residues , by starting from 20 clusters of single residues , and then joining two nearest clusters step by step until a single cluster is obtained .
the results of clustering are given in tables 4 to 7 .
since the dendritic trees returned from clustering are less informative , for visualization we introduce graphs where vertices are the 20 amino acids , and an edge exists between a pair of amino acids if and only if their distance is below some preset threshold .
graphs obtained from the distance matrices are shown in figs . 5 to 8 , where vertices with no connecting edges are neglected .
in sequence pair alignment we often do not have structure information of both sequences .
with the structure information ignored , we have the mixed counts @xmath37 from which we calculate the residue pair distances averaged over conformations .
the distance matrix obtained is given in table 8 .
we have also calculated distances ( [ dab ] ) to compare different conformations .
distances between any two conformations for various residues are listed in table 9 .
figures 1 to 4 illustrate the dependence of outer sites in a window on the center .
although in the kl distance we sum up effects on individual residues from the center , we still can see the tendency that the center is generally more strongly correlated with the c - terminal sites than n - terminal sites .
furthermore , we may divide the 20 amino acids into two groups with m , i , l , v , f , y and w in one , and the remainders in the other .
they roughly correspond to hydrophobic and hydrophilic groups .
it is seen that for the coil and turn conformations a hydrophobic center exhibits a stronger correlation with outer sites than a hydrophilic center , while for the sheet conformation a hydrophilic center exhibits a stronger correlation .
it is interesting to make a comparison between the distance matrices obtained here with the commonly used blosum62 similarity score matrix .
a small distance implies a large similarity score .
there are many evidences showing the consistency between the distances and scores .
for example , residue pairs vi , il , vl and st have positive blosum scores and at the same time small distances .
the graphs in figs . 5 to 8 contain two connected subgraphs : one consists of i , l , v , f , y , and the other consists of s , t. this is another evidence of the consistency .
generally , the averaged distance matrix is closer to blosum62 than the conformation specific ones .
however , there do exist some remarkable differences .
for example , residue pairs gt , qa , fv with negative scores have rather small distances in either the conformation helix , or sheet or coil , while pairs yh and nh with positive scores have rather large distances in the helix conformation .
moreover , yh has a large distance in all the four conformations .
blosum matrices are derived from conserved amino acid patterns called blocks .
it is expected that for most score entries we should see the consistency in at least one conformation specific distance matrix . for a given residue pair ,
if residue profiles of an amino acid center are very dissimilar for different conformations , after averaging over conformations the pair distance would generally become smaller . in this case ,
blosum scores and conformation specific distance need not be consistent since the former contains no structure information . our results show some strong dependence of residue behavior on conformations .
for example , the distances of pairs cd and si in helix are about twice higher than in sheet .
there are many residue pairs displaying strong dependence of distances on conformations .
table 9 views the conformation dependence from conformation pair comparison .
indeed , the table indicates that for any conformation pairs there are certain residues which behave very differently in the two conformations
. however , generally speaking , coil and turn are quite similar . in comparison of physicochemical properties of amino acids
, the abundance of amino acids is not taken into consideration .
this is also the case for the above defined distances .
other statistical variables including the effect of sample size may be introduced .
one candidate is the @xmath7 statistic for identical distributions .
the analysis using this new statistic is under study .
we expect that algorithms using multiple conformation specific matrices should work better in sequence alignment .
the popular needleman - wunsch algorithm can be modified to include putative conformation for each residue .
this will be discussed elsewhere .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ this work was supported in part by the special funds for major national basic research projects and the national natural science foundation of china . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ s. henikoff and j.g .
henikoff , proc .
( usa ) , * 89 * , 10915 ( 1992 ) .
j. garnier , d. osguthorpe and b. robson , j.mol.biol .
* 120 * , 97 ( 1978 ) .
u. hobohm and c. sander , protein science * 3*,522 ( 1994 ) .
ftp.embl-heidelberg.de/pub/databases/protein_extras/pdb_select/old/2001_sep.25 w. kabsch and c. sander , biopolymers , * 22 * , 2577 ( 1983 ) . m. o. dayhoff and r. v. eck , atlas of protein sequence and structure ( natl . biolmed .
res . found . , silver spring , md ) , * 3 * , 33 ( 1968 ) .
s. kullback , j.c .
keegel and j.h .
kullback , information theory and statistics , wiley , new york ( 1959 ) .
s. kullback , topics in statistical information theory , springer , berlin ( 1987 ) .
t.sakamoto , m. ishiguro and g.kitagawa , akaike information criterion statistics , ktk scientific , tokyo ( 1986 ) .
o. weiss and h. herzel , j. theor .
biol , * 190 * , 341 ( 1998 ) .
o. weiss , m.a .
jimnez - montano and h. herzel , j. theor .
biol , * 206 * , 379 ( 2000 ) .
p. stolorz , a. lapedes and y. xia , j.mol.biol , * 225 * , 363 ( 1992 ) .
.... 19 a d e k q r s t n g h c f i lv m y w p 18 a d e k q r s t n g h c f ilv m y
w p 17 a d e k q r s t n g h c filv m y w p 16 a d e k q r st n g h c filv m y w p 15 a d e k qr st n g h c filv m y w p 14 a d e kqr st
n g h c filv m y w p 13 a d e kqrst n g h c filv m y w p 12 a d e kqrstn g h c filv m y w p 11 a d ekqrstn g h c filv m y w p 10 a dekqrstn g h c filv m y w p 9 a dekqrstn g h c filvm y w p 8 adekqrstn g h c filvm y w p 7 adekqrstn g h c filvmy w p 6 adekqrstng h c filvmy w p 5 adekqrstngh c filvmy w p 4 adekqrstngh c filvmyw p 3 adekqrstngh cfilvmyw p 2 adekqrstnghcfilvmyw p .... .... 19 a g f il v y m d e q s t r k h n c w p 18 a g f ilv y m d e q s t r k h n c w p 17 a g filv y m d e q s t r k h n c w p 16 a g filvy m d e q s t r k h n c w p 15 a g filvy m d e q st r k h n c w p 14 a g filvy m d e qst r k h n c w p 13 a g filvy m d eqst r k h n c w p 12 a g filvy m d eqstr k h n c w p 11 a g filvy m d eqstrk h n c w p 10 ag filvy m d eqstrk h n c w p 9 agfilvy m d eqstrk h n c w p 8 agfilvym d eqstrk h n c w p 7 agfilvym d eqstrkh n c w p 6 agfilvym d eqstrkhn c w p 5 agfilvym deqstrkhn c
w p 4 agfilvymdeqstrkhn c w p 3 agfilvymdeqstrkhnc w p 2 agfilvymdeqstrkhncw p .... .... 19 a e k q r st n g d f
l v i y m h p w c 18 a e k q r st n g d f lv i y m h p w c 17 a e k q r st n g d flv i y m h p w c 16 a e k q r st n g d flvi y m h p w c 15 a ek q r st n g d flvi y m h p w c 14 a ekq r st n g d flvi y m h p w c 13 a ekqr st n g d flvi y m h p w c 12 a ekqrst n g d flvi y m h p w c 11 aekqrst n g d flvi y m h p w c 10 aekqrstn g d flvi y m h p w c 9 aekqrstng d flvi y m h p w c 8 aekqrstng d flviy m h p w c 7 aekqrstngd flviy m h p w c 6 aekqrstngdflviy m h p w c 5 aekqrstngdflviym h p w c 4 aekqrstngdflviymh p w c 3 aekqrstngdflviymhp w c 2 aekqrstngdflviymhpw c .... .... 19 a dn e k s t r q l y f v h g i p m w c 18 a dn e k st r q l y f v h g i p m w c 17 a dn ek st r q l y f v h g i p m w c 16 a dnek st r q
l y f v h g i p m w c 15 a dnekst r q
l y f v h g i p m w c 14 a dnekstr q l y f v h g i p m w c 13 adnekstr q l y f v h g i p m w c 12 adnekstrq l y f v h g i p m w c 11 adnekstrql y f v h g i p m w c 10 adnekstrqly f v h g i p m w c 9 adnekstrqlyf v h g i p m w c 8 adnekstrqlyfv h g i p m w
c 7 adnekstrqlyfvh g i p m w c 6 adnekstrqlyfvhg i p m w c 5 adnekstrqlyfvhgi p m w c 4 adnekstrqlyfvhgip m w c 3 adnekstrqlyfvhgipm w c 2 adnekstrqlyfvhgipmw c .... | window profiles of amino acids in protein sequences are taken as a description of the amino acid environment .
the relative entropy or kullback - leibler distance derived from profiles is used as a measure of dissimilarity for comparison of amino acids and secondary structure conformations .
distance matrices of amino acid pairs at different conformations are obtained , which display a non - negligible dependence of amino acid similarity on conformations .
based on the conformation specific distances clustering analysis for amino acids is conducted .
= -2 cm = -1 cm = 16.5 cm |
typical wavelength of gravitational waves from astrophysical compact objects such as bh(black hole)-bh binaries is in some cases very long so that wave optics must be used instead of geometrical optics when we discuss gravitational lensing .
more precisely , if the wavelength becomes comparable or longer than the schwarzschild radius of the lens object , the diffraction effect becomes important and as a result the magnification factor approaches unity @xcite .
mainly due to the possibility that the wave effects could be observed by future gravitational wave observations , several authors @xcite have studied wave effects in gravitational lensing in recent years . in most of the works which studied gravitational lensing phenomenon in the framework of wave optics , isolated and
normal astronomical objects such as galaxies are concerned as lens objects . recently
yamamoto and tsunoda@xcite studied wave effects in gravitational lensing by an infinite straight cosmic string .
the metric around a cosmic string is completely different from that around a usual massive object .
cosmic strings generically arise as solitons in a grand unified theory and could be produced in the early universe as a result of symmetry breaking phase transition@xcite . if symmetry breaking occurred after inflation , the strings might survive until the present universe . recently , cosmic strings attract a renewed interest partly because a variant of their formation mechanism was proposed in the context of the brane inflation scenario@xcite . in this scenario inflation
is driven by the attractive force between parallel d - branes and parallel anti d - branes in a higher dimensional spacetime . when those brane - anti - brane pairs collide and annihilate at the end of inflation , lower - dimensional d - branes , which behave like monopoles , cosmic strings or domain walls from the view point of four - dimensional observers , are formed generically @xcite .
for some time , cosmic string was a candidate for the seed of structure formation of our universe , but this possibility was ruled out by the measurements of the spectrum of cosmic microwave background ( cmb ) anisotropies@xcite .
the current upper bound on the dimensionless string tension @xmath4 is around @xmath5 , which comes from the observations of cmb@xcite and/or the pulsar timing @xcite .
although cosmic string can not occupy dominant fraction of the energy density of the universe , its non - negligible population is still allowed observationally@xcite .
in fact , sazhin et al.@xcite reported that csl-1 , which is a double image of elliptical galaxies with angular separation @xmath6 , could be the first case of the gravitational lensing by a cosmic string with @xmath7 .
we study in detail wave effects in the gravitational lensing by an infinite straight cosmic string . in ref .
@xcite , wave propagation around a cosmic string was studied but they put the waveform around the string by hand . compared with that corresponding to the geometrical optics .
we show that the importance of the diffraction effects are determined by the combination of three parameters , @xmath4 , the distance from the string to the observer and the wavelength and that the relative amplitude of the diffracted wave can be @xmath8 for realistic astrophysical situations . ]
their prescription is correct only in the limit of geometrical optics , which breaks down when the wavelength becomes longer than a certain characteristic length . in this paper
, we present exact solutions of the ( scalar ) wave equation in a spacetime with a cosmic string .
we analytically show that our solutions reduce to the results of the geometrical optics in the short wavelength limit .
we derive a simple analytic formula of the leading order corrections to the geometrical optics due to the finite wavelength effects and also an expression for the long wavelength limit .
interference caused by the lensing remains due to the diffraction effects even when only a single image can be seen in the geometrical optics .
this fact increases the lensing probability by cosmic strings .
this paper is organized as follows . in section
ii , we construct a solution of the wave equation on a background spacetime with an infinite straight cosmic string in the case that a source of the wave is located infinitely far .
an extension to the case in which a point source is located at a finite distance is given in appendix b. in section iii , we study properties of the solution obtained in sec .
ii in detail . in section iv ,
we focus on compact binaries as the sources of gravitational waves and discuss the possible effects due to finiteness of the lifetime and the frequency evolution of the binaries on the detection of the gravitational waves which pass near a cosmic string .
we also give a rough estimate for the event rate of the lensing of gravitational waves from ns - ns mergers assuming decigo / bbo .
section v is devoted to summary .
a solution of einstein equations around an infinite straight cosmic string to first order in @xmath9 is given by @xcite @xmath10 where @xmath11 is a cylindrical coordinate(@xmath12 ) and @xmath13 is the deficit angle around the cosmic string
. spatial part of the above metric describes the euclidean space with a wedge of angular size @xmath14 removed .
due to the deficit angle around a string , double images of the source are observed with an angular separation @xmath15 when a source is located behind the string in the limit of geometrical optics . in general for a wave with a finite wavelength , some interference pattern appears .
an exact solution of einstein equations around a finite thickness string has been already obtained @xcite , but we use the metric ( [ 1.1 ] ) as a background since the string thickness is negligibly small compared with the einstein radius , @xmath16 , where @xmath17 is the distance from the observer to the string . throughout the paper ,
we consider waves of a massless scalar field instead of gravitational waves for simplicity , but the wave equations are essentially the same in these two cases .
an extension to the cosmological setup is straightforwardly done by adding an overall scale factor . in that case
the time coordinate @xmath18 is to be understood as the conformal time .
the wave equation remains unchanged if we consider a conformally coupled field , but it is modified for the other cases due to curvature scattering .
the correction due to curvature scattering of the friedmann universe is suppressed by the square of the ratio between the wavelength and the hubble length , which can be neglected in any situations of our interest .
our goal of this section is to construct a solution of the wave equation which corresponds to a plane wave injected perpendicularly to and scattered by the cosmic string .
this situation occurs if the distance between the source and the string is infinitely large . in order to construct such a solution
, we introduce a monochromatic source uniformly extended in the @xmath19-direction and localized in @xmath20 plane , @xmath21 where @xmath22 is the frequency and we have introduced @xmath23 , a constant independent of @xmath24 , to adjust the overall normalization when we later take the limit @xmath25 .
the factor @xmath26 appears because @xmath27-coordinate used in the metric ( [ 1.1 ] ) differs from the usual angle @xmath28 here we consider a uniformly extended source instead of a point source since the former is easier to handle . when the limit @xmath25 is taken , the answers are identical in these two cases .
the case with a point - like source at a finite distance is more complicated .
this case is treated in appendix b. now the wave equation that we are to solve is @xmath29 since @xmath30 satisfies the same equation ( [ 1.3 ] ) as @xmath31 does , @xmath31 is even in @xmath27 .
thus , it can be expanded as @xmath32 from eqs .
( [ 1.3 ] ) and ( [ 1.4 ] ) , the equations for @xmath33 are @xmath34 where @xmath35 and @xmath36 .
the solution of eq .
( [ 1.5 ] ) except for @xmath37 is a linear combination of bessel function and hankel function .
we impose that the wave @xmath38 is regular at @xmath39 and pure out - going at infinity .
further , imposing that the wave is continuous at @xmath37 , @xmath33 becomes @xmath40 where @xmath41 is the heaviside step function .
substituting eq .
( [ 1.6 ] ) into eq .
( [ 1.5 ] ) , the normalization factor @xmath42 is determined as @xmath43^{-1}\cr & = & { br_0 \epsilon_m(-1)^m\over 4i(1-\delta ) } , \label{1.7}\end{aligned}\ ] ] where @xmath44 denotes a differentiation with respect to the argument . from eqs .
( [ 1.6 ] ) and ( [ 1.7 ] ) with the aid of the asymptotic formulae of the bessel and hankel functions , @xmath31 for @xmath45 can be written as @xmath46 we determine the overall normalization of the source amplitude @xmath23 , independently of @xmath4 , so that eq .
( [ 1.8 ] ) becomes a plane wave @xmath47 when @xmath48 .
this condition leads to @xmath49 .
then , finally @xmath38 becomes @xmath50
the solution ( [ 1.9 ] ) describes the waveform propagating around a cosmic string .
but it is not easy to understand the behavior of the solution because it is given by a series .
in fact , it takes much time to perform the summation in eq .
( [ 1.9 ] ) numerically for a realistic value of tension of the string , say , @xmath51 because of slow convergence of the series .
in particular it is not manifest whether the amplification of the solution in the short wavelength limit coincides with the one which is obtained by the geometrical optics approximation
. therefore it will be quite useful if one can derive a simpler analytic expression .
here we reduce the formula by assuming that the distance between the string and the observer is much larger than the wave length , @xmath52 which is valid in almost all interesting cases . using an integral representation of the bessel function , @xmath53 where the contour of the integral @xmath54 is such as shown in fig .
[ contour ] , eq .
( [ 1.9 ] ) can be written as @xmath55 when @xmath18 is in the segment of the integration contour @xmath54 along the imaginary axis , the summation over @xmath56 does not converge because the absolute value of each term in the summation is all unity . in order to make the series to converge , we need to think that the integration contour @xmath54 is not exactly on the imaginary axis but @xmath18 always has a positive real part . for bookkeeping purpose
, we multiply each term in the sum by a factor @xmath57 ( @xmath58 is an infinitesimally small positive real number ) .
then eq .
( [ 3.2 ] ) becomes @xmath59 where @xmath60 is defined by @xmath61 with @xmath62 and @xmath63 .
now we find that all we need to evaluate is @xmath60 in order to obtain an approximate formula for @xmath64 .
this integral will not be expressed by simple known functions in general , but the integration can be performed by using the method of steepest descent in the limit @xmath65 . the integrand of eq .
( [ 3.3 ] ) has two saddle points located at @xmath66 and @xmath67 in the vicinity of the integration contour @xmath54 .
we should also notice that the integrand has a pole at @xmath68 , which is also infinitesimally close to the contour of the integral @xmath54 .
this pole is located near the saddle point at @xmath69 as far as @xmath24 and @xmath27 are small .
hence the treatment of the saddle point at @xmath69 is much more delicate than that of the saddle point at @xmath70 .
we only discuss the saddle point at @xmath69 , then the case at @xmath70 is a trivial extension . when @xmath71 or @xmath72 , which corresponds to shaded regions in fig .
[ contour ] , @xmath73 diverges in the limit @xmath74 . if @xmath75 , the pole at @xmath68 is in the bottom - left unshaded region . in this case
we can not deform the contour to the direction of the steepest descent at @xmath69 without crossing the pole at @xmath68 .
the deformed contour which is convenient to apply the method of the steepest descent is such that is shown as @xmath76 in fig .
[ contour ] .
when we deform the integration contour from @xmath54 to @xmath77 , there arises an additional contribution corresponding to the residue at @xmath78 when @xmath75 . on the other hand , if @xmath79 , the pole is in the top - left shaded region . in this case
, we can deform the contour of the integral to the direction of the steepest descent without crossing the pole @xmath80 . hence no additional term arises . from these observations , we find that it is necessary to evaluate the integral ( [ 3.3 ] ) separately depending on the signature of @xmath81 . though the calculation itself can be done straightforwardly
, it is somewhat complicated because the saddle point and the pole are close to each other .
when the pole is located inside the region around the saddle point that contributes dominantly to the integral , a simple gaussian integral does not give a good approximation .
detailed discussions about this point are given in appendix a. here we only quote the final result which keeps terms up to @xmath82 , @xmath83 where @xmath84 , \label{3.3b1}\\ \sigma(\theta ) & : = & { \rm sign } ( { \alpha}(\theta ) ) , \label{3.3c}\end{aligned}\ ] ] and @xmath85 we are mostly interested in the cases with @xmath86 . then , we have @xmath87 , and therefore @xmath88 reduces to @xmath89 . the second term in eq .
( [ 3.3bb ] ) is the contribution from the integral around the saddle point at @xmath69 along the contour @xmath90 .
this term is not manifestly suppressed by @xmath91 . as far as @xmath89 is fixed
, this term does not vanish in the limit @xmath92 .
of course , if we fix @xmath24 and @xmath27 first , and take the limit @xmath92 , the argument of the error function goes to @xmath93 and the function itself vanishes .
however , @xmath89 vanishes at @xmath94 .
hence even for a very large value of @xmath95 there is always a region of @xmath27 in which this second term can not be neglected .
however , for @xmath27 in such a region , @xmath89 can not be very large . therefore , we can safely drop the second term in the exponent . on the other hand , the last term in eq .
( [ 3.3bb ] ) , which is the contribution from the saddle point at @xmath70 , is always suppressed by @xmath91 .
hence , this term does not give any significant contribution for @xmath65 . the first term in eq .
( [ 3.31 ] ) can be dropped in the same manner for @xmath65 . keeping only the terms which possibly remain in the limit @xmath92
, we finally obtain @xmath96 for illustrative purpose , we compared the estimate given in eq .
( [ 3.3bb ] ) with the exact solution eq .
( [ 1.9 ] ) in fig .
[ approx ] .
they agree quite well at @xmath65 .
the deficit angle and the observer s direction are chosen to be @xmath97 and @xmath98 , respectively . , @xmath76 and @xmath99 , respectively .
@xmath100 are the saddle points of @xmath73 . ]
geometrical optics limit corresponds to the limit @xmath74 with @xmath24 and @xmath27 fixed . in this limit @xmath89
also goes to @xmath93 , and hence the error function in eq .
( [ 3.3b ] ) vanishes .
hence the waveform in the geometrical optics limit , which we denote as @xmath101 , becomes @xmath102 where @xmath103 is defined by eq .
( [ 1.25 ] ) . since @xmath38 and
hence @xmath101 are even in @xmath27 , it is sufficient to consider the case with @xmath104 . in fig .
[ configuration ] , the configuration of the source , the lens and the observer is drawn in the coordinates in which the deficit angle @xmath105 is manifest , i.e. , the wedge aob is removed from the spacetime . both points
a and b indicate the location of the source .
the lines oa and ob are to be identified .
the angle made by these two lines is the deficit angle .
the locations of the string and the observer are represented by o and p , respectively . in our current setup
the distance between o and a ( @xmath106 ) is taken to be infinite .
when @xmath107 , only the source a can be seen from the observer .
this corresponds to the fact that only the first term remains for @xmath108 in eq .
( [ go1 ] ) . for @xmath107 , we have @xmath109 this is a plane wave whose traveling direction is @xmath110 , which is the direction of @xmath111 in fig . [ configuration ] in the limit @xmath112 . for @xmath113 ,
@xmath101 is @xmath114 this is the superposition of two plane waves whose traveling directions are different by the deficit angle @xmath105 .
hence amplification of the images and interference occur for @xmath113 as expected . as we shall explain below , eq .
( [ go1 ] ) coincides with the one derived under the geometrical optics . in geometrical optics , wave form is given by @xcite @xmath115 , \label{go6}\ ] ] where @xmath116 represents a two - dimensional vector on the lens plane and @xmath117 represents the summation of time of flight of the light ray from the source to the point @xmath118 on the lens plane and that from the point @xmath118 to the observer .
@xmath119 is a stationary point of @xmath117 , and @xmath120 when @xmath119 is a minimum , saddle and maximum point of @xmath117 , respectively .
the amplitude ratio @xmath121 is written as @xmath122 , \label{go7}\ ] ] where @xmath123 in eq .
( [ go7 ] ) is the deflection potential @xcite which is the integral of the gravitational potential of the lens along the trajectory between the source and the observer .
( [ go6 ] ) represents that the wave form is obtained by taking the sum of the amplitude ratio @xmath124 of each images with the phase factor @xmath125 .
if the lens is the straight string , the spacetime is locally flat everywhere except for right on the string .
this means that the deflection potential @xmath126 is zero and hence the amplitude ratio is unity for all images @xcite and the trajectory where the time of flight @xmath127 takes the extremal value is a geodesic in the conical space , and @xmath127 of any geodesic takes minimum , which means @xmath128 .
there are two geodesics if the observer is in the shaded region in fig .
[ configuration ] .
the time of flight along the trajectory ap is @xmath129 where @xmath130 .
the time of flight along the trajectory bp is obtained by just replacing @xmath103 with @xmath131 . hence , substituting ( [ go8 ] ) into ( [ go6 ] ) , we find that the waveform in the geometrical optics is the same as eq .
( [ go3 ] ) except for an overall phase @xmath132 .
this factor has been already absorbed in the choice of the normalization factor @xmath23 in our formula ( [ 1.9 ] ) .
we define the amplification factor @xmath133 where @xmath134 is the unlensed waveform . using eq .
( [ go3 ] ) , the amplification factor of @xmath101 for @xmath135 is given by @xmath136 where we have assumed @xmath103 and @xmath24 are small and dropped terms higher than quadratic order .
it might be more suggestive to rewrite the above formula into @xmath137 where @xmath138 .
the distance from a node to the next of when the observer is moved in @xmath139-direction is @xmath140 , where @xmath3 is a wavelength .
this oscillation is seen in the right panel of fig .
[ approx ] . in the previous subsection
, we have derived the waveform in the limit @xmath141 which corresponds to the geometrical optics approximation .
here we expand the waveform ( [ 3.3b ] ) to the lowest order in @xmath142 .
this includes the leading order corrections to the geometrical optics approximation due to the finite wavelength effects . for the same reason as we explained in the previous subsection , we assume that @xmath24 and @xmath103 are small .
using the asymptotic formula for the error function eq .
( [ ap.5a ] ) , the leading order correction due to the finite wavelength , which we denote as @xmath143 , is obtained as @xmath144 as is expected , the correction blows up for @xmath145 , where @xmath81 or @xmath146 vanishes , irrespectively of the value of @xmath95 . in such cases
, we have to evaluate the error function directly , going back to eq .
( [ 3.3b ] ) . the expression on the first line in eq.([qgo1 ] ) manifestly depends only on @xmath147 aside from the common phase factor @xmath148 .
this feature remains true even if we consider a small value of @xmath147 .
this can be seen by rewriting eq .
( [ 3.3b ] ) as @xmath149 the common phase @xmath148 does not affect the absolute magnitude of the wave .
except for this unimportant overall phase , the waveform is completely determined by @xmath150 .
the geometrical meaning of these parameters @xmath150 is the ratio of two length scales defined on the lens plane . to explain this ,
let us take the picture that a wave is composed of a superposition of waves which go through various points on the lens plane
. in the geometrical optics limit the paths passing through stationary points of @xmath127 , which we call the image points , contribute to the waveform .
the first length scale is @xmath151 which is defined as the separation between an image point and the string on the lens plane . in this picture
we expect that paths whose pathlength is longer or shorter than the value at an image point by about one wavelength will not give a significant contribution because of the phase cancellation .
namely , only the paths which pass within a certain radius from an image point need to be taken into account .
then such a radius will be given by @xmath152 , which we call fresnel radius .
namely , a wave with a finite wavelength can be recognized as an extended beam whose transverse size is given by @xmath153 .
the ratio of these two scales gives @xmath150 : @xmath154 when @xmath155 , i.e. , @xmath156 , the beam width is smaller than the separation . in this case
the beam image is not shadowed by the string , and therefore the geometrical optics becomes a good approximation . when @xmath157 , i.e. , @xmath158 we can not see the whole image of the beam , truncated at the location of the string
. then the diffraction effect becomes important .
the ratio of the beam image eclipsed by the string determines the phase shift and the amplification of the wave coming from each image .
if we substitute @xmath159 as a typical value , we obtain a rough criterion that the diffraction effect becomes important when @xmath160 or @xmath161 in terms of @xmath95 .
the same logic applies for a usual compact lens object . in this case
the fresnel radius does not change but the typical separation of the image from the lens is given by the einstein radius @xmath162 , where @xmath163 is the mass of the lens .
then the ratio between @xmath164 and @xmath153 is given by @xmath165 , which leads to the usual criterion that the diffraction effect becomes important when @xmath166@xcite . from the above formula ( [ qgo1 ] )
, we can read that the leading order corrections scales like @xmath167 .
this dependencies on @xmath3 and @xmath168 differ from the cases that the lens is composed of a normal localized object , in which the leading order correction due to the finite wavelength is @xmath169 @xcite .
the condition for the diffraction effect to be important ( [ qgo4 ] ) can be also derived directly from eq .
( [ qgo1 ] ) . in order
that the current expansion is a good approximation , @xmath170 must be smaller than @xmath101 .
this requires that @xmath171 , which is identical to ( [ qgo4 ] ) .
we plot the absolute value of the amplification factor under the quasi - geometrical optics approximation as dashed line in fig .
[ quasi1 ] .
we find that the quasi - geometrical optics approximation is a good approximation for @xmath172 .
for @xmath173 , the quasi - geometrical optics approximation gives a larger amplification factor than the exact one . in the quasi - geometrical optics approximation
, we find from eqs .
( [ go3 ] ) and ( [ qgo1 ] ) the absolute value of the amplification factor for @xmath174 is @xmath175^{1/2}. \label{qgo5}\ ] ] from this expression , we find that the position of the first peak of the amplification factor lies at @xmath176 , which can be also verified from fig .
[ quasi1 ] . for @xmath173
the present approximation is not valid , but we know that the amplification factor should converge to unity in the limit @xmath177 , where @xmath153 is much larger than @xmath178 . for @xmath98 .
black line and dashed one correspond to eq .
( [ 3.3b ] ) and the quasi - geometrical optics approximation , respectively .
the string tension is chosen to be @xmath179 . ]
we show in fig .
[ quasi2 ] the absolute value of the amplification factor as a function of @xmath103 for four cases of @xmath95 around @xmath180 .
top left , top right , bottom left and bottom right panels correspond to @xmath181 and @xmath182 , respectively .
black curves are plots for eq .
( [ 3.3b ] ) and the dotted ones are plots for the quasi - geometrical optics approximation . as is expected , the error of the quasi - geometrical optics approximation becomes very large near @xmath183 , where @xmath89 vanishes . as the value of @xmath95 increases , the angular region in which the quasi - geometrical optics breaks down is reduced .
interestingly , the absolute value of the amplification factor deviates from unity even for @xmath184 which is not observed in the geometrical optics limit .
this is a consequence of diffraction of waves , the amplitude of oscillation of the interference pattern becomes smaller as @xmath27 becomes larger , which is a typical diffraction pattern formed when a wave passes through a single slit .
the broadening of the interference pattern due to the diffraction effect means that the observers even in the region @xmath185 can detect signatures of the presence of a cosmic string .
but the deviation of the amplification from unity outside the wedge @xmath186 is rather small except for the special case @xmath187 : for @xmath188 the magnification is inefficient and for @xmath189 the magnification itself does not occur .
hence the increase of the event rates of lensing by cosmic strings compared with the estimate under the geometrical optics approximation could be important only when the relation @xmath187 is satisfied .
if we take @xmath190 and @xmath191 which is in the frequency band of lisa(laser interferometer space antenna)@xcite , we find that the typical value of @xmath4 is @xmath192 .
so far , we have considered the stringy source rather than a point source .
extension to a point source can be done in a similar manner to the case of the stringy source and is treated in appendix b. the result is @xmath193 where @xmath194 is the distance between the source and the observer .
@xmath195 , which is defined by eq .
( [ b1 ] ) , is related to @xmath196 as @xmath197 { \big|}_{\xi \to x}+(\theta \to -\theta).\ ] ] hence @xmath38 for the point source is similar to that for the stringy source . in particular , assuming that @xmath198 , and keeping terms which could remain for @xmath199 , we have @xmath200 we have derived an approximate waveform ( [ qgo1.5 ] ) which is valid in the wave zone from the exact solution of the wave equation eq .
( [ 1.9 ] ) . here
we show that eq .
( [ qgo1.5 ] ) can be obtained by a more intuitive and simpler method . in the path integral formalism @xcite
, the wave form is given by the sum of the amplitude @xmath201 for all possible paths which connect the source and the observer . here
@xmath202 is the time of flight along the path @xmath203 .
if the cosmic string resides between the source and the observer , the wave form will be given by the sum of two terms one of which is obtained by the path integral over the paths which pass through the upper side of the string ( @xmath204 ) in fig .
[ configuration ] , and the other through the lower side of it ( @xmath205 ) .
the waveform coming from the former contribution will be given by @xmath206 where @xmath207 is a point on the lens plane specified by @xmath208 .
one can determine the normalization constant @xmath209 by a little more detailed analysis , but we do not pursue it further here . by integrating eq .
( [ sd1 ] ) , we recover the first term in eq .
( [ qgo1.5 ] ) . for completeness
, we consider the case in which the wavelength is longer than the distance from the string @xmath210 . in this limit , the first few terms in eq .
( [ 1.9 ] ) dominate , and we find @xmath211 in particular , for @xmath212 eq .
( [ 3.17 ] ) becomes @xmath26 which is larger than unity .
this differs from the cases of gravitational lensing by a normal compact object , where the amplification becomes unity in the long wavelength limit .
the reason why the amplification differs from unity even in the long wavelength limit is that the space has a deficit angle and hence the structure at the spatial infinity is different from the usual euclidean space .
waves with very long wavelengths do not feel the local structure of string . however , uniform amplification of waves should occur as a result of total energy flux conservation because the area of the asymptotic region at a constant distance from the source is reduced due to the deficit angle . in this sense
such modes feel the existence of a string .
in this section , we consider compact binaries as sources of gravitational waves . gravitational waves from compact binaries are clean in the sense that the waves are almost monochromatic : the time scale for the frequency to change is much longer than the orbital period of the binary except for the phase just before plunge . hence interference between two waves coming from both sides of the cosmic string could be observed by future detectors .
since each compact binary has a finite lifetime , lensing events can be classified roughly into two cases .
if the difference between the times of flight along two geodesics is larger than the lifetime of the binary , we will observe two independent waves separately at different times . on the other hand ,
if the time delay is shorter than the lifetime , what we observe is the superposition of two waves .
the remaining lifetime of the binary @xmath213 when the period of the gravitational waves measured by an observer is @xmath214 is estimated as @xmath215 where @xmath216 is the mass ratio of the binary ( @xmath217 ) , @xmath163 is the mass of the more massive star in the binary and @xmath218 is the source redshift .
the time delay @xmath219 is @xmath220 taking the typical values of parameters as @xmath221 and @xmath222 , the condition @xmath223 gives the upper bound on the mass @xmath163 , @xmath224 the time scale for the orbital frequency of the binary to change is the same order as @xmath213 .
hence the condition @xmath223 implies that the frequencies of two waves are almost the same .
the left and right panels in fig .
[ lisa - decigo ] which correspond to different frequencies of gravitational waves show the region where the condition eq .
( [ com3 ] ) is satisfied for three different values of string parameter @xmath24 .
the shaded area represent the parameter region beyond the detector s sensitivities . in the left and right panels we assumed , respectively , that the threshold value for detection in strain amplitude for lisa and decigo(decihertz interferometer gravitational wave observatory)@xcite
/ bbo(big bang observer)@xcite , which are given by @xmath225 and @xmath226 .
we find that both cases @xmath223 and @xmath227 can occur both for lisa and bbo / decigo .
we can easily extend our waveform ( [ 1.9 ] ) to the case that the frequency of the source changes in time .
let us write the source as @xmath228 .
the fourier transformation of @xmath229 is defined by @xmath230 denoting the solution @xmath231 for a monochromatic source obtained in the previous sections by @xmath232 , @xmath38 can be written as @xmath233 where we assumed that @xmath229 is real .
substituting eq .
( [ qgo7 ] ) to the above expression , we have @xmath234 eq .
( [ com4.1 ] ) is a general formula which applies to any time dependent source .
here we consider the special case in which @xmath229 takes the form @xmath235 with @xmath236 where @xmath237 and @xmath238 are assumed .
this represents a quasi - monochromatic source with its frequency slowly changing .
then @xmath239 is @xmath240 substituting eqs .
( [ qgo8 ] ) and ( [ com4.3 ] ) into eq .
( [ com4.1 ] ) , and using the method of the steepest descent , we have @xmath241 with @xmath242 this represents a superposition of two waves coming from both sides of the string whose arrival times differ by @xmath243 . in the preceding subsections , we study the waveforms
observed in two cases with @xmath244 and @xmath245 . as we have explained in the preceding subsection ,
what we observe is a superposition of two waves in this case .
because the relative phase difference of these waves slowly increases or decreases in time due to the frequency change of the binary source and the optical path difference between two geodesics , we will observe the beat if the amplitude of the integrated relative phase difference over observation time is larger than @xmath8 .
the condition that the beat is observed can be derived as follows .
if we denote the total observation period by @xmath246 , then from eq .
( [ com4.4 ] ) the integrated relative phase difference is @xmath247 , where both @xmath168 and @xmath248 are assumed to be @xmath249 . hence we can observe the beat if @xmath250 because @xmath213 is roughly the same as the time scale for the frequency of the binary to change , i.e. @xmath251 , eq .
( [ com10 ] ) can be written as @xmath252 if @xmath246 is fixed , e.g. @xmath253 for lisa , eq .
( [ com12 ] ) is written as an lower bound on @xmath163 .
for @xmath254 and @xmath255 , eq .
( [ com12 ] ) becomes @xmath256 we show in fig .
[ lisa2 ] the region where eq .
( [ com13 ] ) is satisfied for lisa with @xmath254 .
we find that if @xmath257 which is about one order of magnitude below the current upper bound , lisa will detect the beat of gravitational waves for all observable ranges in @xmath258 space as long as @xmath259 . )
is satisfied .
the frequency of the gravitational waves is assumed to be @xmath260.,width=302 ] if @xmath261 , we observe the waveform of either the first term or the second one in eq .
( [ com4.4 ] ) at a given time .
we show in fig .
[ oneside ] the amplification of the wave corresponding to the first term in eq .
( [ com4.4 ] ) as a function of @xmath262 normalized by @xmath263 , which is nothing but @xmath264 in the case discussed in sec.[sec : behaviorsofsolution ] .
we find that the amplification approaches zero more slowly for @xmath265 and oscillates around unity for @xmath266 and the angular size in which non - trivial oscillations due to the diffraction effect can be observed is given by @xmath263 . since @xmath267 in the present case , we have @xmath268 .
therefore this angular size of oscillation is much smaller than @xmath269 .
hence it will be very difficult to detect a lensing event in which this diffraction effect is relevant . in this section
, we estimate the detection rate of the gravitational lensing caused by cosmic strings for planned gravitational wave detectors such as lisa , decigo and bbo .
it is well known that string network obeys the scaling solution where the appearance of the string network at any time looks alike if it is scaled by the horizon size .
there are a few dozen strings spread crossing the horizon volume and a number of string loops @xcite .
since the horizon scale increases in the comoving coordinates as time goes , the number of strings increase if there is no interaction between them . however , since strings are typically moving at a relativistic speed , they frequently intersect with each other . as a result reconnection between strings occurs , reducing the number of long strings which extend over the horizon scale . during the process of reduction of the number of long strings a large number of string loops are formed , but they shrink and decay via gravitational radiation . due to the balance of two effects , the number of long strings in a horizon volume remains almost constant in time .
the reconnection probability @xmath270 is essentially @xmath271 for gauge theory solitons @xcite because reconnection allows the flux inside the string to take an energetically favorable shortcut . for f - strings ,
the reconnection is a quantum process and its probability is roughly estimated as @xmath272 , where @xmath273 is the string coupling and is predicted in @xcite that @xmath274 for d - strings , the reconnection probability might be @xmath275 @xcite .
if the reconnection probability is less than @xmath271 , the number of long strings is expected to be @xmath276 times larger than that in the case with @xmath277 .
therefore it is expected that in the context of cosmic strings motivated by superstring theory the number of long strings in a horizon volume can be @xmath278 or more . to estimate the event rate for the gravitational lensing , here we consider a compact binary ( such as binary neutron stars and/or black holes ) as a source of gravitational waves .
there are large uncertainties about the event rate of mbh ( massive black hole ) merger detected by lisa or decigo / bbo .
several authors @xcite employed a model in which mbh mergers are associated with the mergers of host dark matter halos to estimate the event rate of mbh - mbh mergers . in this model , the event rate is dominated by halos with the minimum mass @xmath279 above which halos have a central mbh and some scenario predict that the event rate could reach @xmath280 events / yr . for decigo / bbo , the binary neuron stars will be observed @xmath281 events / yr . the probability of lensing for a single source by an infinite straight cosmic string both at cosmological distances is @xmath282 eq . ( [ es3 ] ) is derived under the geometrical optics approximation . in section iii , we found that the signal of lensing by cosmic strings ( the interference pattern of gravitational waves at detectors ) extends over an angular scales larger than the deficit angle @xmath283 when the diffraction effect is marginally important .
this is a well known fact for the gravitational lensing by usual stellar objects @xcite . as we estimated in section iii
, the critical distance @xmath284 below which the diffraction effect becomes important is @xmath285 therefore the probability of lensing by cosmic strings may be enhanced due to the diffraction effect for @xmath286 at lisa band ( @xmath287sec ) and for @xmath288 at decigo / bbo band ( @xmath289sec ) .
assuming the prospective values of the parameters that determines the rate of lensing events @xmath290 , we obtain @xmath291 where @xmath292(@xmath293 ) denotes the numerical factor arising from the enhancement of the lensing probability due to the diffraction effect .
@xmath294 is almost upper bound on the total event rate of neutron star mergers detectable by decigo / bbo .
if the event rate is even higher , the number of events becomes comparable to or larger than the number of frequency bins .
then we will not be able to distinguish each event , and undistinguishable signals become confusion noise . in the case of lisa ,
this bound on @xmath295 is even lower .
unfortunately , a large number of lensing events by cosmic strings can be expected only for marginally large @xmath296 with a small reconnection probability @xmath270 .
finally we briefly comment on the validity of the assumption that most of cosmic strings can be treated as straight ones in studying gravitational lensing by them . in geometrical optics approximation ,
only light paths which satisfy the fermat s principle contribute to the amplification factor .
if we take into account the finiteness of the wavelength , the trajectories whose optical path differences are less than a few times of its wavelength will dominantly contribute to the amplification factor . in terms of the distance on the lens plane ( @xmath208-plane in fig .
[ configuration ] ) , the optical paths within @xmath297 from the intersection of the geodesic will give a dominant part of the amplification factor .
in the standard literature , the typical size of small - scale structure of a long string is given by the gravitational back - reaction scale @xmath298 , where @xmath18 is a cosmic time @xcite .
but this is not an established argument and some recent studies suggest that the smallest size of the wiggles could be much smaller than @xmath299 @xcite . if we assume here that the smallest size of the wiggles is @xmath299 , then the condition that the straight string approximation is good is @xmath300 . substituting the appropriate values of the parameters
, it gives the condition , @xmath301 hence approximating a cosmic string by a straight one is good for wide range of possible values of the parameters .
we have constructed a solution of the klein - gordon equation for a massless scalar field in the flat spacetime with a deficit angle @xmath13 caused by an infinite straight cosmic string .
we showed analytically that the solution in the short wavelength limit reduces to the geometrical optics limit .
we have also derived the correction to the amplification factor obtained in the geometrical optics approximation due to the finite wavelength effect and the expression in the long wavelength limit .
the waveform is characterized by a ratio of two different length scales .
one length scale @xmath178 is defined as the separation between the image position on the lens plane in the geometrical optics and the string .
we have two @xmath178 since there are two images corresponding to which side of the string the ray travels .
( when the image can not be seen directly , we assign a negative number to @xmath178 . )
the other length scale @xmath153 , which is called fresnel radius , is the geometrical mean of the wavelength and the typical separation among the source , the lens and the observer .
the waveform is characterized by the ratios between @xmath178 and @xmath153 . if @xmath302 , the diffraction effect becomes important and
the interference patterns are formed .
even when the image in the geometrical optics is not directly seen by the observer , the interference patterns remain .
in contrast , in the geometrical optics magnification and interference occur only when the observer can see two images which travel both sides of the string .
namely , the angular range where lensing signals exist is broadened by the diffraction effect .
this broadening may increase the lensing probability by an order of magnitude compared with that estimated by using the geometrical optics when the distance to the source is around the critical distance @xmath284 given in eq .
( [ es4 ] ) .
we finally estimated the rate of lensing events which can be detected by lisa and decigo / bbo assuming bh - bh or ns - ns mergers as a source of gravitational waves .
for possible values of the parameters that determines the event rate such as string reconnection rate , string tension and the event rate of the unlensed mergers , the lensing event rate could reach several per yr .
thanks kunihito ioka , takashi nakamura and hiroyuki tashiro for useful comments .
this work is supported in part by grant - in - aid for scientific research , nos .
14047212 and 16740141 , and by that for the 21st century coe `` center for diversity and universality in physics '' at kyoto university , both from the ministry of education , culture , sports , science and technology of japan .
here we derive a formula eq .
( [ 3.3b ] ) from the integral representation of the solution eqs .
( [ 3.31 ] ) and ( [ 3.3 ] ) . as we explained in the sec.iii
, we have to calculate the integral for @xmath303 and @xmath304 separately . in this case
, there is no contributions from the pole @xmath80 . since @xmath305 , the integral @xmath306
is dominated from the two regions @xmath307 , where @xmath308 is the saddle points of @xmath73 .
let us first calculate the integral around @xmath309 .
we can not apply the method of steepest descent where the denominator of the integrated function is replaced with the value at @xmath310 because the pole @xmath80 of the integrand can lie in the region @xmath311 and the denominator is no longer constant around @xmath309 .
fortunately the integral can be approximated written by the special function which can be evaluated easily .
we first do the transformation of variable such that @xmath312 ( u : real number ) which corresponds to the deformation of the contour of the integral from @xmath54 to @xmath76 as shown in fig .
[ contour ] . expanding @xmath73 around @xmath309 to second order in @xmath313 and the denominator of the integral to the first order in @xmath313
gives the integral @xmath314 where @xmath315 is defined by eq .
( [ 3.3b1 ] ) .
hence we need to evaluate the integral @xmath316 where @xmath58 was introduced to remember that the imaginary part of @xmath317 is positive when @xmath318 .
this integral is given by an error function as @xmath319 this can be derived by solving a differential equation @xmath320 which follows from the definition of @xmath321 , with the boundary condition that @xmath322 .
asymptotic formulas for the error function are @xmath323 using eq .
( [ ap.3 ] ) , the integral eq .
( [ ap.2 ] ) becomes @xmath324 next let us calculate the integral eq .
( [ ap.1 ] ) around @xmath325 . since the pole @xmath80 is far from @xmath326 , we can approximate the denominator of the integrated function as a constant and apply the usual saddle point method .
this gives @xmath327 the sum of eqs .
( [ ap.6 ] ) and ( [ ap.7 ] ) gives @xmath328 for @xmath304 . in this case
, there is a contribution from the pole @xmath80 . hence the integral
is divided into the integral around the pole and the one whose circuit of integration is @xmath76 .
the integral around the pole gives @xmath329 the integral around @xmath330 along the trajectory @xmath76 is the same as for @xmath304 and is given by eq .
( [ ap.2 ] ) .
the only difference is the signature of @xmath331 . by changing the integration variable from @xmath313 to @xmath332 ,
@xmath88 is replaced with @xmath333 and the overall signature flips . as a result
we find that the integration along @xmath77 gives @xmath334 integral eq .
( [ ap.1 ] ) around @xmath326 is also given by eq .
( [ ap.7 ] ) . combining the results of subsections @xmath335 and @xmath336 , adding the similar terms @xmath337 , and also using the asymptotic form of @xmath338 , we have eq .
( [ 3.3b ] ) .
here we consider a point source at a finite distance . for a point source , @xmath339 where @xmath340 .
we consider a solution written in the form of the following expansion , @xmath341 the solution for @xmath342 is the same as @xmath33 in ( [ 1.6 ] ) but @xmath22 contained in @xmath95 and @xmath343 are here replaced with @xmath344 , and @xmath345 first we compute @xmath346 for @xmath347 . as in the case of bessel function
, we also use the integral representation for hankel function @xmath348 here the integration is to be performed along the path @xmath349 presented in fig .
[ contour ] . using the above formula and ( [ 3.1 ] ) , we have @xmath350 we introduce a new variable @xmath351 . under the assumption that @xmath352 , the integration over @xmath203 is dominated by the contribution around @xmath353 .
hence , the integration contour for @xmath18 is unaltered even if we change the integration variable from @xmath18 to @xmath354 .
after this change of the variable , we have @xmath355 , \cr & & \end{aligned}\ ] ] where @xmath356 we expand the exponent around a zero of its derivative .
the derivative vanishes at @xmath357 , and @xmath358 is given by @xmath359 taylor expansion of @xmath360 around @xmath357 becomes @xmath361 we truncate this expansion at the quadratic order because the higher order terms are suppressed by @xmath362 or @xmath363 .
performing gaussian integral , we obtain @xmath364 further , we expand @xmath365 around an approximate stationary point at @xmath366
. then we have @xmath367 again we truncate this expansion at the quadratic order for the same reason as before .
then one finds that @xmath368 is approximately given by @xmath369 where @xmath370 and @xmath371 the function @xmath372 is almost identical to @xmath31 discussed in sec.iii , except that @xmath148 and other @xmath95 are replaced with @xmath373 and @xmath374 , respectively .
finally , we perform the integration over @xmath375 . from ( [ b1 ] ) , we have @xmath376 since @xmath377 , we can invoke the saddle point method again to perform @xmath375-integral when @xmath378 is large . evaluating the contribution from the saddle point at @xmath379 with @xmath380 ,
we obtain @xmath381 the calculation for @xmath382 can be done in a completely parallel way , and the final result becomes identical to the case with @xmath347 .
s. deguchi and w. d. watson , astrophys .
j. * 307 * , 30 ( 1986 ) .
r. takahashi and t. nakamura , astrophys .
j. * 595 * , 1039 ( 2003 ) [ arxiv : astro - ph/0305055 ] .
n. seto , phys .
d * 69 * , 022002 ( 2004 ) [ arxiv : astro - ph/0305605 ] .
t. t. nakamura , phys .
lett . * 80 * , 1138 ( 1998 ) .
k. yamamoto , arxiv : astro - ph/0309696 . c. baraldo , a. hosoya and t. t. nakamura , d * 59 * , 083001 ( 1999 ) . t. t. nakamura and s. deguchi prog .
. suppl . * 133 * , 137 ( 1999 ) . k. yamamoto and k. tsunoda , phys .
d * 68 * , 041302 ( 2003 ) [ arxiv : astro - ph/0309694 ] .
t. suyama , r. takahashi and s. michikoshi , arxiv : astro - ph/0505023 .
r. takahashi , t. suyama and s. michikoshi , astron .
astrophys . * 438 * , l5 ( 2005 ) [ arxiv : astro - ph/0503343 ] .
r. takahashi , astron .
astrophys . * 423 * , 787 ( 2004 ) [ arxiv : astro - ph/0402165 ] .
m. b. hindmarsh and t. w. b. kibble , rept .
. phys .
* 58 * , 477 ( 1995 ) [ arxiv : hep - ph/9411342 ] .
a. vilenkin and e. p. s. shellard , cosmic strings and other topological defects , cambridge university press ( cambridge , 2000 ) . g. r. dvali and s. h. h. tye , arxiv : hep - ph/9812483 .
g. r. dvali , b * 459 * , 489 ( 1999 ) [ arxiv : hep - ph/9905204 ] .
c. p. burgess , m. majumdar , d. nolte , f. quevedo , g. rajesh and r. j. zhang , * 0107 * , 047 ( 2001 ) [ arxiv : hep - th/0105204 ] .
s. h. s. alexander , phys .
d * 65 * , 023507 ( 2002 ) [ arxiv : hep - th/0105032 ] .
g. r. dvali , q. shafi and s. solganik , arxiv : hep - th/0105203 .
n. jones , h. stoica and s. h. h. tye , * 0207 * , 051 ( 2002 ) [ arxiv : hep - th/0203163 ] . g. shiu and s. h. h. tye , phys .
b * 516 * , 421 ( 2001 ) [ arxiv : hep - th/0106274 ] .
m. majumdar and a. christine - davis , jhep * 0203 * , 056 ( 2002 ) [ arxiv : hep - th/0202148 ] .
g. dvali and a. vilenkin , phys .
d * 67 * , 046002 ( 2003 ) [ arxiv : hep - th/0209217 ] .
n. t. jones , h. stoica and s. h. h. tye , phys .
b * 563 * , 6 ( 2003 ) [ arxiv : hep - th/0303269 ] .
g. dvali and a. vilenkin , jcap * 0403 * , 010 ( 2004 ) [ arxiv : hep - th/0312007 ] .
e. j. copeland , r. c. myers and j. polchinski , jhep * 0406 * , 013 ( 2004 ) [ arxiv : hep - th/0312067 ] .
d. n. spergel _ et al .
_ [ wmap collaboration ] , astrophys .
j. suppl .
* 148 * , 175 ( 2003 ) [ arxiv : astro - ph/0302209 ] .
w. j. percival _ et al . _
[ the 2dfgrs team collaboration ] , mon . not .
soc . * 337 * , 1068 ( 2002 ) [ arxiv : astro - ph/0206256 ] .
e. jeong and g. f. smoot , astrophys . j. * 624 * , 21 ( 2005 ) [ arxiv : astro - ph/0406432 ] .
l. pogosian , s. h. h. tye , i. wasserman and m. wyman , phys .
d * 68 * , 023506 ( 2003 ) [ arxiv : hep - th/0304188 ] .
l. pogosian , m. c. wyman and i. wasserman , arxiv : astro - ph/0403268 .
m. wyman , l. pogosian and i. wasserman , phys .
d * 72 * , 023513 ( 2005 ) [ arxiv : astro - ph/0503364 ] . v. m. kaspi , j. h. taylor and m. f. ryba , astrophys .
j. * 428 * ( 1994 ) 713 .
s. e. thorsett and r. j. dewey , phys . rev .
d * 53 * , 3468 ( 1996 ) . m. p. mchugh , g. zalamansky , f. vernotte and e. lantz , phys .
d * 54 * , 5993 ( 1996 ) .
a. n. lommen , arxiv : astro - ph/0208572 . f. r. bouchet , p. peter , a. riazuelo and m. sakellariadou , phys .
d * 65 * , 021301 ( 2002 ) [ arxiv : astro - ph/0005022 ] .
j. rocher and m. sakellariadou , phys .
lett . * 94 * , 011303 ( 2005 ) [ arxiv : hep - ph/0412143 ] .
m. sazhin , m. capaccioli , g. longo , m. paolillo and o. kovanskaya , arxiv : astro - ph/0506400
. m. sazhin _ et al .
_ , mon . not .
soc . * 343 * , 353 ( 2003 ) [ arxiv : astro - ph/0302547 ] .
b. linet , annales poincare phys .
* 45 * , 249 ( 1986 ) .
a. vilenkin , phys .
d * 23 * , 852 ( 1981 ) .
j. r. i. gott , astrophys .
j. * 288 * , 422 ( 1985 ) .
p. schneider , j. ehlers and e. e. falco , gravitational lenses , springer , new york , ( 1992 ) c. cutler and e. e. flanagan , phys .
d * 49 * , 2658 ( 1994 ) [ arxiv : gr - qc/9402014 ] .
a. albrecht and n. turok , phys .
d * 40 * , 973 ( 1989 ) .
d. p. bennett and f. r. bouchet , phys .
d * 41 * , 2408 ( 1990 ) . b. allen and e. p. s. shellard , phys .
lett . * 64 * , 119 ( 1990 ) .
r. a. matzner , comput .
* 2 * , 51 ( 1989 ) .
m. g. jackson , n. t. jones and j. polchinski , arxiv : hep - th/0405229 .
m. g. haehnelt , mon . not .
astron . soc .
* 269 * , 199 ( 1994 ) [ arxiv : astro - ph/9405032 ] .
r. r. islam , j. e. taylor and j. silk , mon . not .
* 354 * , 629 ( 2004 ) [ arxiv : astro - ph/0309559 ] .
k. ioka and p. meszaros , arxiv : astro - ph/0502437 .
a. a. ruffa , astrophys . j. * 517 * , l31 ( 1999 ) a. vilenkin , arxiv : hep - th/0508135 .
x. siemens and k. d. olum , nucl .
b * 611 * , 125 ( 2001 ) [ erratum - ibid .
b * 645 * , 367 ( 2002 ) ] [ arxiv : gr - qc/0104085 ] .
x. siemens , k. d. olum and a. vilenkin , phys .
d * 66 * , 043501 ( 2002 ) [ arxiv : gr - qc/0203006 ] . | we present exact solutions of the massless klein - gordon equation in a spacetime in which an infinite straight cosmic string resides .
the first solution represents a plane wave entering perpendicular to the string direction .
we also present and analyze a solution with a static point - like source . in the short wavelength limit
these solutions approach the results obtained by using the geometrical optics approximation : magnification occurs if the observer lies in front of the string within a strip of angular width @xmath0 , where @xmath1 is the string tension . we find that when the distance from the observer to the string is less than @xmath2 , where @xmath3 is the wave length , the magnification is significantly reduced compared with the estimate based on the geometrical optics due to the diffraction effect . for gravitational waves from neutron star(ns)-ns mergers the several lensing events per year may be detected by decigo / bbo . |
angle - resolved photo - emission spectroscopy ( arpes ) directly probes the dispersion of electrons ; identifying a kink associated with the active optical phonon in the cuprates @xcite .
neutron scattering has shown an anomalous change in the phonon spectrum at the superconducting transition , indicating an interesting role for phonons in the cuprates @xcite .
estimates of the magnitude of the electron - phonon coupling , and the energy of the phonon mode put the problem outside the limited region of applicability for bcs theory , so schemes to cope with a wider range of parameters need to be developed .
moreover , since there a node in the superconducting gap consistent with d - wave symmetry @xcite , any theory implicating phonons as the mechanism must be able to explain the unconventional order .
electron - phonon ( e - ph ) interactions can be described by the following generic model , @xmath3 here , @xmath4 and @xmath5 with @xmath6 and @xmath7 ( representing a quasi-2d system and taming the van - hove singularities ) .
the holstein model is a special case where @xmath8 and @xmath9 , for which @xmath10 is the dimensionless e - ph coupling ( @xmath11 is the half - band width ) .
the momentum - independent phonon dispersion approximates an optical phonon , and the momentum - independent e - ph coupling corresponds to a completely local interaction
. a coulomb pseudo - potential , @xmath12 , may be added ( i.e. hubbard model at the hartree fock level ) .
( a ) overview of the diagrammatic representations of the self - energy used in this work .
@xmath13 is the electron , and @xmath14 the phonon self energy .
the row labeled me contains the vertex neglected and vc the corrected diagrams .
( b ) schematic of the dca ( quasi - local ) formalism for cluster size @xmath15 . within the squares @xmath16
where @xmath17 are represented by solid circles .
equivalently , @xmath18 . a partial density of states ( dos )
is associated with each sub - zone .
dca enables computations in the thermodynamic limit , via an equivalent cluster impurity problem.,scaledwidth=46.0% ] some treatments of the arpes kink used a standard eliashberg theory with anisotropic coupling and an assumed @xmath0-wave order parameter for fitting purposes @xcite .
the extended eliashberg approach that i have developed differs significantly from previous work in the sense that it is fully self - consistent , with no prior assumptions about the form of the order parameter ( i.e. any @xmath0-wave order arises directly from the self - consistency of the theory ) @xcite .
in this section , i discuss how quasi - local eliashberg equations can be constructed , and demonstrate how the lowest - order fock diagram does not contribute to the gap equations .
the calculation is intended to be demonstrative , and shows that standard eliashberg equations fail for an optical phonon mediated @xmath0-wave state .
the eliashberg equations are computed from the fock contribution to the self - energy : @xmath19 and equate @xmath20 with the approximate form , @xmath21 normally the eliashberg equations are computed using a local ( momentum - independent ) approximation
. however , to examine @xmath0-wave states , the local approximation is not sufficient . substituting the green function @xmath22 in eq .
[ eqn : se ] ( @xmath23 , @xmath24 are pauli matrices and @xmath25 is the chemical potential ) , and treating the gap function as constant in sub - zones of @xmath26-space in the manner of dca , one obtains a quasi - local approximation for the eliashberg equations , @xmath27 when gap functions have @xmath0-wave symmetry , @xmath28 and @xmath29 , and the partial dos @xmath30 has the symmetry of the lattice , it is immediately clear that the @xmath31 contribution to the quasi - local eliashberg equations vanishes , and 2nd order terms must be considered .
when there is some weak momentum dependence in the e - ph coupling and phonon frequency , this cancellation will not be exact , but the contribution from 2nd - order terms can still be expected to be the largest in the perturbation series .
thus a fully extended set of eliashberg equations including vertex corrections needs to be constructed .
currently , it is not possible to construct a similar approximation to eq .
[ eqn : qlapprox ] for the vertex corrected theory . however
a numerical approach using the full dca equations is available to enlighten some aspects of the e - ph problem .
the rest of this article concerns results from an extended eliashberg theory with momentum dependence and vertex corrections determined numerically using dca .
no form for the order parameter is assumed in advance and full forms for the green functions and self - energies are used .
full details of the translation of the diagrams in fig .
[ fig : overview](a ) may be found in ref .
figure [ fig : one ] shows image plots of the electron spectral functions @xcite .
vertex corrections are most pronounced for @xmath32 , as expected from analysis of the parameter space @xcite . in particular ,
the vertex corrected theory shows a kink at the energy scale relating to the phonon frequency * measured * at the @xmath33 point ( i.e. renormalised zone edge phonon frequency ) in agreement with the arpes measurements in ref .
@xcite .
phonon spectral functions are shown in fig .
[ fig : two ] .
the phonon renormalisation is extremely small for @xmath34 .
in contrast , modes at the @xmath35 point in the brillouin zone are strongly softened for @xmath36 when vertex corrections are not taken into account .
the second order terms act against the softening and reinforce the theory .
a small reduction in spectral weight is seen mid way along the @xmath37-x and @xmath37-w lines when @xmath32 .
this might be a precursor to the strong softening seen in ref .
@xcite on passing through the transition temperature , but further analysis of the superconducting state is necessary to be sure .
@xmath0-wave superconductivity is one of the remarkable results from the theory extended with vertex corrections .
figure [ fig : three ] shows the anomalous density as the chemical potential is varied first upwards from half - filling and then back from large filling showing that two solutions are stable .
a small @xmath38 is applied . @xmath39 and @xmath40 . starting with a half - filled solution and increasing , then decreasing the chemical potential , first @xmath0-wave and then ( on decreasing from high electron density ) @xmath41-wave solutions to the self - consistent equations are found ( as detailed in the caption of fig . [ fig : three ] ) .
highly anisotropic solutions form as the fermi - surface approaches the van - hove points . in the @xmath0-wave state ,
the first order terms are not the leading order of the perturbation theory in @xmath2 .
it was argued in ref .
@xcite that this was in part due to the nearly flat phonon spectrum , so that the phonon propagator @xmath42 and the 1st order contribution to the anomalous self - energy @xmath43 is small ( @xmath44 is the anomalous green function ) .
the anomalous contribution to the pseudopotential term is @xmath45 .
since @xmath44 is modulated in the @xmath0-wave state , then the contribution of the pseudopotential term is zero . in the @xmath41-wave state ,
@xmath44 is not modulated , and the resulting finite contribution destroys the @xmath41-wave order for large enough @xmath46 , thus selecting a dominant @xmath0-wave contribution . the phonon spectra in fig .
[ fig : two ] demonstrate that so long as @xmath47 , then this condition is met , and 2nd order terms are essential .
2nd order terms remain the leading order correction to the anomalous self - energy , even for the larger @xmath2 used here .
it is clear that the 2nd order terms are similar for both repulsive and attractive models , and thus it is perhaps not surprising that both attractive and repulsive models show a @xmath0-wave state .
i have described spectroscopic results and order parameters from an extended eliashberg - style theory with vertex and momentum corrections .
the extended theory is essential for the investigation of d - wave superconductivity mediated by optical phonons , since the vertex corrections are the lowest order terms in the perturbation expansion in electron - phonon coupling @xmath2 .
indeed , d - wave superconductivity is immediately predicted with such an extension .
there are two remaining areas which require further investigation .
the first involves a better treatment of the coulomb repulsion , which is currently only treated as a simple hartree - fock pseudopotential . in particular ,
the ability to treat large repulsion and antiferromagnetism is essential for high-@xmath48 materials .
treatment of momentum dependent e - ph coupling would also be of interest .
however , the overriding challenge at this stage is to convert the numerical results into a coherent analytic theory with vertex corrections . | i present results from an approach that extends the eliashberg theory by systematic expansion in the vertex function ; an essential extension at large phonon frequencies , even for weak coupling . in order to deal with computationally expensive double sums over momenta , a dynamical cluster approximation ( dca ) approach
is used to incorporate momentum dependence into the eliashberg equations .
first , i consider the effects of introducing partial momentum dependence on the standard eliashberg theory using a quasi - local approximation ; which i use to demonstrate that it is essential to include corrections beyond the standard theory when investigating @xmath0-wave states .
using the extended theory with vertex corrections , i compute electron and phonon spectral functions .
a kink in the electronic dispersion is found in the normal state along the major symmetry directions , similar to that found in photo - emission from cuprates .
the phonon spectral function shows that for weak coupling @xmath1 , the dispersion for phonons has weak momentum dependence , with consequences for the theory of optical phonon mediated d - wave superconductivity , which is shown to be 2nd order in @xmath2 .
in particular , examination of the order parameter vs. filling shows that vertex corrections lead to @xmath0-wave superconductivity mediated via simple optical phonons .
i map out the order parameters in detail , showing that there is significant induced anisotropy in the superconducting pairing in quasi-2d systems . *
[ published as : journal of physics and chemistry of solids , vol .
69 , 2982 - 2985 ( 2008 ) ] * extended eliashberg theory , superconductivity , spectroscopy , unconventional pairing |
mass loss is inexorably linked to evolution for high - mass stars .
in fact , it has a _
deterministic _ influence , which in turn has tremendous impact on other areas of astronomy influenced by stellar feedback ( regulating star formation , galaxy evolution , chemical evolution , reionization , etc . ) . for most of their lives ,
massive stars above @xmath020 @xmath1 shed mass in fast winds that affect their subsequent evolution , but in post - main sequence ( post - ms ) phases the mass loss becomes critical in determining the type of resulting supernova ( sn ) .
the most dramatic mass loss in post - ms evolution is during the luminous blue variable ( lbv ) phase .
some of the most pressing issues in massive star research concern resolving problems with mass - loss rates and how they are incorporated into evolutionary models , as well as the treatment ( or neglect ) of close binaries in these models ( see recent reviews by @xcite ) .
a majority of massive stars are in binary systems whose separation is small enough that they will exchange mass @xcite . with current uncertainties in mass loss and mass transfer , connecting massive stars to their end fates as various types of sn explosions
is still very challenging .
lbvs and their eruptive mass loss have emerged as the linchpin in our paradigm for the evolution of single massive stars . in light of recent estimates of lower mass - loss rates for clumpy winds ,
eruptive mass loss of lbvs would need to be the primary agent responsible for turning most h - rich o stars ( above @xmath030 - 35 @xmath1 , the upper limit for red supergiants ) into h - free wolf - rayet ( wr ) stars before they die as sne ibc @xcite .
however , our understanding of the physics and evolutionary states of lbvs is still poor , and their connection to ( or overlap with ) the evolution of close binary systems leaves many open questions . when ideas about lbvs were taking shape , @xcite discussed mass - transferring binaries and how they may provide suitable explanations for lbvs , but for whatever reason , these ideas did not appear to dominate the interpretation of lbvs and so they were mostly discussed as very massive single - stars .
later , @xcite mentioned difficulties in explaining the formation of the nebula around @xmath2 carinae with a single star , although that paper did not discuss evolutionary aspects of lbvs .
the role of binarity and initial mass in evolution is a central question we address here from a new perspective , by studying lbv environments .
lbvs were recognized early - on as the brightest blue irregular variables in nearby spiral galaxies like m31 , m33 , and ngc 2403 @xcite , originally referred to as the `` hubble - sandage variables '' .
famous galactic objects like p cygni and @xmath2 carinae had spectacular outbursts in the 17th and 19th centuries , respectively , and appeared to share many of the same properties , so they were grouped together and called `` lbvs '' by @xcite .
over a dozen true lbvs are now identified in the milky way ( mw ) and the large and small magellanic clouds ( lmc / smc ) , and a similar number resides in other local group galaxies @xcite . stars that spectroscopically resemble lbvs with similar luminosity and color , but which have not ( as yet ) been observed to show the signature eruptive variability of lbvs , are often called `` lbv candidates '' ; these are presumed to be temporarily dormant lbvs .
the traditional view of lbvs , which emerged in the 1980s and 1990s , is that they correspond to a very brief _ transitional _ phase of evolution of the most massive single stars , when the star moves from core h burning , through the onset of h shell burning , to the core he burning phase @xcite .
the motivation for a very brief phase comes from the fact that lbvs are extremely rare compared to o stars : the duration of the lbv phase is thought to be only a few 10@xmath3 yr @xcite based on this line of reasoning .
10@xmath4 yr @xcite , but it is unclear if this longer lbv lifetime ( and shorter wr lifetime ) agrees with observed statistics . ]
a typical and often quoted monotonic evolutionary scheme for a star of 60 - 100 @xmath1 initially is : o star @xmath5 of / wnh @xmath5 lbv @xmath5 wn @xmath5 wc @xmath5 sn ibc . in this scenario ,
the strong mass - loss experienced by lbvs is a crucial component of the central paradigm of massive star evolution , which is that a star s own mass loss determines its evolution .
high luminosity powers strong radiatively driven winds , and this wind mass loss propels the evolution that converts the most massive h - rich main - sequence stars to become h - free wr stars .
this is the co - called `` conti scenario '' @xcite .
lbvs are crucial to this picture because of the recent downward revisions of o - star mass - loss rates , as noted above ( see review by @xcite and many references therein ) .
thus , single massive stars of 30 - 80 @xmath1 do not have strong enough winds to make wr stars on their own ; lbv eruptions are needed @xcite .
( especially wnh stars ) , might be strong enough to make wr stars without an intervening lbv phase @xcite . in this case
, the remaining he cores would be more massive than any observed h - free wr stars @xcite .
such massive he stars might be rare , so this is not yet a solved issue . ]
the lbv phase is still generally presumed to arise in the evolution of single stars @xcite , and it is _ required _ in the standard scenario for single massive - star evolution .
the origin of the instability is not fully understood , but it is expected that the core luminosity goes up with time as the growing he core contracts , while winds steadily reduce the stellar mass .
thus , the ratio @xmath6 goes up until the star confronts the classical eddington limit and somehow erupts with catastrophic mass loss @xcite . as seen in the case of @xmath2 car , these eruptions can eject 10 - 20 @xmath1 in a single event @xcite .
while this traditional view of lbvs persisted through the 1990s @xcite , in the past decade or so , several problems have emerged that threaten it .
one issue has to do with the interpretation of lbv variability and the behavior of their winds .
the proposed interpretation was that during a standard s dor eruption , the star brightens while staying at constant bolometric luminosity because its mass - loss rate is assumed to increase , which in turn initiates a cooler and brighter `` pseudo photosphere '' @xcite . however , quantitative spectroscopy has since revealed that mass - loss rates in s dor outbursts do not increase enough to cause true pseudo photospheres @xcite , and the changes do not really occur at constant bolometric luminosity @xcite .
instead , the brightening in an s dor event is more akin to a pulsation or temporary inflation of the envelope , perhaps driven by the sub - surface fe opacity peak @xcite . similarly , the traditional explanation for lbv giant eruptions was that they experience a substantial _ increase _ in their bolometric radiative luminosity ( for reasons unexplained ) that temporarily pushes them well above the classical eddington limit @xcite ; this necessarily initiates a super - eddington continuum - driven wind @xcite .
however , several lines of evidence now point to an explosive mechanism driving lbv giant eruptions , instead of ( or in addition to ) a wind @xcite .
moreover , spectra of the light echoes from @xmath2 car s 19th century great eruption seem inconsistent with a wind pseudo - photosphere @xcite .
these considerations point to problems with our interpretation of the physics behind lbv eruptions , but even more serious problems have arisen with the traditional role played by lbvs in evolution .
the first clear sign of something very fishy came from the recognition that eruptive lbvs are the most likely progenitors of a particular class of luminous narrow - lined sne with type iin spectra ( sne iin ) . the dense h - rich circumstellar material ( csm ) of some sne
iin requires eruptive pre - sn mass loss akin to lbvs , and in some cases very large csm mass budgets ( see review by @xcite ) .
other aspects of the csm around sne iin point to wind variability reminiscent of lbvs @xcite .
moreover , we now have four direct detections of lbv - like progenitors of sne iin , including sn 1961v , sn 2005gl , sn 2010jl , and sn 2009ip ( @xcite ; see review by @xcite ) . in the traditional paradigm of single - star evolution ,
massive lbvs are prohibited from exploding as sne iin , because lbvs represent merely a brief intermediate phase after core - h burning and before core - he burning .
they should still have 0.5 - 1 myr to live as a wr star , and should die without their h envelopes .
another clue that something was seriously amiss came from `` sn impostors '' , which are thought to be lbv giant eruptions associated with very massive stars in external galaxies @xcite .
for some nearby sn impostors , detections of their relatively low - luminosity dust - enshrouded progenitors , as well as the ages of surrounding stars , suggest either progenitors that are 10 - 15 @xmath1 bsgs / rsgs or even @xmath08 @xmath1 super - agb stars ( e.g. , @xcite ) .
this is much lower than the initial masses attributed to eruptive lbvs , which are generally above 25 @xmath1 @xcite .
the reason this is physically significant is because these lower - mass stars ( @xmath08 @xmath1 ) never approach the eddington limit in their normal evolution , but they appear to suffer lbv - like eruptions anyway .
if they can do this , then the physical cause of the outbursts might not be related exclusively to high initial mass , but perhaps some other exotic mechanism ( collisions , mergers , etc . ) .
additionally , anderson and collaborators @xcite examined the spatial correlation between various types of extragalactic sne or sn impostors , as compared to presumed indicators of youth such as h@xmath7 emission in their host galaxies . on the one hand
, they found the expected result that sne ii - p ( arising from 8 - 20 @xmath1 stars ; @xcite ) are less correlated with h@xmath7 than stripped - envelope sne .
however , they also found the very surprising result that _ sn impostors and sne iin are apparently even less correlated with star formation than sne ii - p_. this was curious , and motivated us to take another look at lbvs in the mw and lmc / smc .
in this paper , we will argue that the likely solution to many of these problems with lbvs resides in binary evolution , and that the resulting sn kicks may be quite important in explaining their isolation .
the idea that binary interaction influences the outcomes of stellar evolution is not a new one @xcite , and various aspects including sn kicks have been studied in theoretical work by several groups @xcite .
uncertainty persists in the difficult problems of mass transfer and the subsequent evolution of the mass gainer .
the way that lbvs may fit into this picture is disussed in more detail later in the paper .
we begin with a discussion of the projected locations on the sky of classic galactic and magellanic cloud lbvs as compared to the locations of o - type stars ( 2 ) .
this includes a qualitative examination of the sites of individual lbvs , as well as a quantitative statistical comparison of the relative isolation of early , mid , and late - type o stars , wr stars , lbvs , b[e ] supergiants , and rsgs . for this statistical comparison ,
we only consider stars in the lmc and smc , due to large distance and reddening uncertainties in the milky way . in
3 we discuss how these results are fundamentally incompatible with the standard monotonic evolutionary view of lbvs as the descendants of very massive single stars , and in 4 we argue for an alternative view of bifurcated evolutionary paths for massive stars ( where the `` bifurcation '' corresponds to mass donors and mass gainers ) . in 5 we discuss the environments of extragalactic sn impostors and sne .
we end ( 6 ) with an epilogue about why something so simple as the avoidance of ob - star clusters by lbvs was not emphasized before . , but for the two confirmed lbvs in the smc.,width=268 ]
in figures [ fig : mw ] , [ fig : lmc ] , and [ fig : smc ] we present maps of o - type stars in the surrounding environments for several individual cases of well - studied lbvs .
the positions of o - type stars are taken from the simbad databse .
this depends on the spectral type listed in simbad , which in some cases contains errors . as a check
, we also made similar plots using the very recently revised galactic o - star catalog @xcite . using this newer catalog ( not shown ) , some spectral types changed slightly ( for example , some b0 stars changed to late o - types and vice versa , and some o stars shifted from our early - type bin to late - type or vice versa ) , but this did not change the overall result described below where lbvs generally avoid clusters of o stars .
it did not , for example , produce o - star clusters where there were none previously .
for galactic lbvs we plot a projected size on the sky of @xmath83.8 deg , and for stars in the magellanic clouds we plot @xmath80.3 deg . in each case
this represents ( very roughly ) the surrounding several 10@xmath9 pc of each lbv .
this is the 2d location projected on the sky , which gives the lbvs the `` benefit of the doubt '' that they are actually at the same distance as the o - type stars seen near them on the sky .
we did not attempt to correct for distances along the line - of - sight to further cull the sample of o - type stars , because that may introduce a bias .
for example , in many cases , such as milky way lbvs whose distances are based on radial velocities , the distance adopted might be substantially wrong if the star has a peculiar velocity .
since our main finding is that lbvs are relatively isolated , treating all stars projected on the sky as being at the same distance is conservative .
[ cols="<,<,^,^,^,^,^ " , ] notes : + p1 = ks p - value for separations to the nearest o star .
+ p2 = ks p - value for separations to the second - nearest o star .
+ [ tab : ks ] the large fraction of lbvs that appear isolated should sound alarm bells for anyone familiar with the traditional view of lbvs as `` massive stars in transition '' , just having finished core - h burning and soon moving on to he burning as wr stars . upon comparing positions of o - type stars to the coordinates of known lbvs
, the result is astonishing . basically , _
lbvs systematically avoid clusters of o - type stars , and they are rarely associated with o - type stars of similar ( presumed ) initial mass_. this is the opposite of what is expected in the standard paradigm of massive star evolution where lbvs are the immediate descendants of the most massive o stars , and the predecessors of luminous wr stars .
[ fig : cum ] and [ fig : cum2 ] indicate that the vast majority of o - type stars are indeed in highly clustered environments .
wr stars are significantly more dispersed , consistent with their more evolved status .
lbvs should be intermediate between o stars and wr stars , but in fact they are more isolated than wr stars .
this requires that they are older , on average , than wr stars , which in turn implies that they arise either from lower initial masses ( longer core - h burning lifetimes ) , that they are a more advanced evolutionary state than wr stars ( improbable based on chemical composition ) , or that some mechanism has systematically removed them from clusters without disturbing other massive stars .
initial masses implied by lbv environments differ considerably from other inferences about their likely initial masses based primarily on turnoff masses and ages of nearby stars .
( actually , there have been very few studies of the ages and masses of lbvs based on their environments , perhaps reflecting the fact that many of them are not in clusters where this is usually done . )
@xmath2 car is located in the tr16 cluster .
tr16 is inferred to have an age of @xmath03 myr , which is , however , based largely on the fact that @xmath2 car itself has not yet gone sn . excluding @xmath2 car
, tr16 has a turnoff mass of roughly 80 - 100 @xmath1 , which is significantly below the presumed initial mass of @xmath2 car itself although perhaps consistent enough that it does not immediately raise questions , due to the similar main - sequence lifetimes of very massive stars at @xmath10100 @xmath1 .
based on studying ob associations in the lmc / smc and adopting characteristic turnoff masses for the earliest o - type stars in each , @xcite inferred initial masses of lbvs and wr stars .
they favored an initial mass of @xmath1185 @xmath1 for the two lbvs in their study ( s dor and r85 ) , and high initial masses for wr stars as well ( @xmath1170 @xmath1 in the smc , and 30 - 60 @xmath1 in the lmc ) .
there are , however , two key effects that compromise this method of determining initial masses : ( 1 ) it requires that the cluster or association is coeval , but many of the regions studied have large age spreads compared to the lifetimes of very massive stars . for example , the association lh41 in which s dor and r85 reside has a very large age spread ; while its most luminous members do suggest a turnoff mass around 85 @xmath1 by comparison to single - star evolutionary tracks , this same association also contains cool supergiants with initial masses of only 10 - 15 @xmath1 @xcite . ( 2 ) this method ignores important effects of binary evolution that corrupt the inferred turnoff mass ( see , e.g. , @xcite )
. both binary mass transfer and binary mergers will produce luminous stars that repopulate the upper main sequence ( e.g. , @xcite ) . when compared to isochrones of single - star evolutionary models
, these massive blue stragglers will imply an age younger than the true age of the cluster , and will overestimate the main - sequence turnoff mass .
convolved with an age spread due to star formation that may be comparable to teh lifetimes of massive stars , this makes it difficult to use cluster turnoffs to estimate the initial masses of the most luminous evolved stars .
looking at the spatial relationship between lbvs and other massive stars gives an independent and quantitative check on relative ages that does not have the same pitfalls as cluster turnoffs .
if we assume that the vast majority of massive stars form in clusters ( indeed they do , as o - type stars are observed to be highly clustered in figure [ fig : cum ] ) , then as the most massive stars explode early and the cluster spreads out due to random motions , the distribution of distances to a nearest neighboring o - type star must slowly and steadily increase with age .
this method has no reliance on uncertain input physics of stellar evolution models or their neglect of binaries .
it does not give a precise absolute age , but it does provide a reliable measure of the _ relative _ ages of different populations of massive stars .
this method produces a very different result from cluster turnoffs @xcite .
the relative isolation of lbvs is apparent in the mw as well as the lmc / smc , although only the lmc / smc produce a meaningful statistical analysis due to distance and reddening uncertainties in the mw .
if there were only a few isolated lbvs , it would nt be so troubling because we might expect a few cases where the lbv was initially the most massive star in a small cluster that contains no other o - type stars .
indeed , a few percent of o stars ( mostly late o - type stars ) appear to have been born in relative isolation ( e.g. , fig .
[ fig : cum ] and @xcite ) but @xmath090% of early o - type stars _ are _ in clusters .
lbvs are the opposite : only rare exceptions are found in clusters ( @xmath2 car and r127 ) and most of them are either in loose associations with large age spreads , or they are found 10s or even 100s of pc from other o - type stars .
this isolation requires that lbvs are either very massive stars that for some reason were preferentially born well outside clusters ( hard to justify physically ) , or that they have lived longer than we expect based on single - star evolution and have systematically moved away from their birth environments .
this can not be reconciled with the traditional view of lbvs in single - star evolution .
as noted in the introduction , the current standard paradigm of massive - star evolution is that single massive o - type stars evolve into h - deficient wr stars by virtue of their own radiation - driven mass loss . for all but the most luminous and most massive stars
( i.e. @xmath10100 @xmath1 ) that pass through a strong - winded wnh phase , accounting for clumping effects indicates that line - driven stellar winds are too weak to accomplish this ( see @xcite and references therein ) .
thus , if lbvs are presumed to fit into the evolutionary sequence of single stars : o star @xmath5 of / wnh @xmath5 lbv @xmath5 wn @xmath5 wc @xmath5 sn ibc , then enhanced eruptive lbv mass loss would be essential in order for single massive o - type stars to become wr stars @xcite . in this scenario ,
the lbv phase is a very brief , fleeting transitional phase ( see , e.g. , @xcite ) . in evolutionary models ,
there is little time ( @xmath010@xmath4 yr ) after the end of core - h burning and before the onset of the wr phase when a star can be an lbv . in this short time interval
, stars can not move very far , and so lbvs are expected to have a spatial distribution on the sky that is very similar to the o stars that are supposedly their immediate progenitors . for a typical velocity dispersion in a cluster of a few km s@xmath12 , a single star will move less than @xmath010 pc in 3 myr , and much less than 1 pc in 10@xmath13 yr . therefore , if early o - type stars are mostly in clusters ( they are ; see figs .
[ fig : cum ] and [ fig : cum2 ] ) , then their immediate single - star descendants must be as well . in fig .
[ fig : cum ] , the cumulative distribution of lbv separations should therefore be in between o - type stars and wn stars , and lbvs should be much more clustered than wc stars . in this paper
we have shown that the opposite is true that lbvs are surprisingly isolated from o stars .
most critically , observations indicate that in terms of their separation from o - type stars , _ lbvs are even more isolated than wr stars ( both wn and wc ) .
_ the observed population of lbvs therefore can not evolve into the observed population of wr stars , because lbvs would need to turn around and systematically move back toward clusters and associations in order to match the locations of wr stars .
this can not be .
the standard view of lbvs as single `` massive stars in transition '' is incompatible with observations of their locations , and must no longer be considered as a viable evolutionary scenario for the majority of massive stars .
this rules out the monotonic evolutionary paradigm outlined above , making it difficult for single massive stars to become h - poor wr stars and sne ibc ( because without lbv mass loss , their winds are nt strong enough ) .
most often , the lbv phase must play a different role .
a straightforward alternative is discussed next .
the fact that lbvs are more isolated than wr stars , plus the fact that some stars appear to remain in an lbv phase until death as sne iin ( see @xcite for a review ) , suggests an entirely different picture . instead of the monotonic evolutionary scheme for single stars discussed above , the spatial distributions discussed in this paper suggest that massive star populations and especially lbvs are dominated by bifurcated evolutionary trajectories : o star @xmath5 @xmath14 } \rightarrow { \rm sniin } & { \rm ( gainer ) } \end{cases}$ ] . in this scenario ,
o - type stars evolve off the main sequence , and through binary interaction the majority of massive stars either ( 1 ) lose their h envelope through mass transfer to a companion , become wr stars and die as stripped - envelope sne , or ( 2 ) gain mass from a companion , become bsgs , sgb[e]s , and lbvs , and then retain their h envelopes until they die as sne iin .
after the mass transfer phase , the subsequent evolution of the mass gainer may be quite varied , and so not all mass gainers will necessarily be lbvs and sne iin . in an insightful early paper ,
@xcite discussed the fact that a few of the hubble - sandage variables in m31 appeared relatively isolated , and they suggested a similar binary mass - transfer scenario for these , while still favoring a single - star evolutionary scheme for most lbvs . here , we show that the environments of _ most _ lbvs violate the single - star picture , not just a few exceptions . instead of lbvs
being a brief transitional phase for all very massive stars , they become the dominant late evolutionary phase for a subset of massive binaries .
lbvs are rare enough that the fraction of stars which do this and for how long they remain in the lbv phase is poorly constrained observationally .
( note that when the oversimplified , monotonic evolutionary scheme is abandoned , the number ratio of lbvs to o - type stars offers no meaningful constraint on the lbv duration without additional information . )
[ fig : hrd ] shows the locations of lbvs on the hr diagram , mostly taken from @xcite , except as noted in the caption .
the locations of lbvs are compared to standard evolutionary tracks for single stars , and also to an example of an evolutionary track for an interacting binary system from @xcite . using single - star evolution tracks as a reference
, the initial masses of lbvs would appear to range from 30 - 40 @xmath1 for the lower - l lbvs , and from 50 - 250 @xmath1 or more for the classical high - luminosity lbvs .
we infer from figure [ fig : hrd ] are approximate .
values of the luminosity depend on the analysis technique used for each star and the distance assumed , and there are multiple values available in the literature for several lbvs ( see , e.g. , @xcite ) . ] a main point of our paper is that the isolated environments of lbvs are incompatible with the short lifetimes experienced by stars with such high initial mass , and they are incompatible with the key idea that most lbvs continue their evolution to wr stars . instead , ascribing lbvs as the mass gainers in interacting binaries provides an attractive and plausible alternative . as noted earlier ,
the plausibility of this basic idea was mentioned long ago @xcite , but did not become the dominant view . through the process of rlof , the mass gainer can substantially increase its mass and luminosity to resemble a star that had an initially much higher mass .
one example is shown in fig . [ fig : hrd ] , where a star with an initial mass of 14 @xmath1 accretes mass in a binary and ends up with a luminosity commensurate with a @xmath026 @xmath1 star @xcite .
it is easy to see that such an evolutionary path could , in principle , give rise to the low - luminosity group of lbvs .
similarly , somewhat more massive stars in binaries ( initial masses of 30 - 40 @xmath1 ) , could accrete mass to appear as 60 - 80 @xmath1 classical lbvs .
one can imagine higher mass analogs of this that could correspond to the classical lbvs like ag car and r127 .
a crucial point is that this mass accretion and corresponding increase in luminosity may occur after a long delay , after both stars in the binary system have lived through much or all of their h - core burning main - sequence lifetimes . as initially lower mass stars
, their core - h burning lifetimes may be much longer than the main - sequence lifetime expected for the more massive star that they become . as such
, a mass gainer may appear younger than the population of stars around it . in cases where the discrepancy between the initial mass and @xmath15 is close to a factor of 2 ,
the lifetime can be doubled for moderately massive stars ( 15 - 30 @xmath1 ) .
for example , the approximate lifetimes of single stars of 15 and 30 @xmath1 initial mass are about 14 and 6.5 myr , respectively @xcite .
if lbvs fit this role , then they are essentially evolved massive blue stragglers , not massive stars in transition .
lbvs originating this way may help solve a number of issues that have been problematic for a single - star scenario : \1 .
as noted above , the mass gainer will have a longer age than expected for its current mass and luminosity , helping to rectify the isolation of lbvs discussed in this paper and the initial mass discrepancy .
in addition to becoming more massive and more luminous , the mass gainer will also gain angular momentum during rlof . this may cause lbvs to appear as rapid rotators late in life .
a number of lbvs do exhibit properties consistent with rapid rotation @xcite .
moreover , rotation has been inferred to be quite important in some ideas about lbv instability @xcite . in a single - star scenario , it is difficult to understand how they can shed large amounts of mass in steady winds while retaining their angular momentum .
single - star models generally predict negligible rotational speeds for the lbv phase @xcite .
this scenario also implies that a large fraction of the fast runaway stars that are kicked out of clusters by their companion s sn will be these rapidly rotating mass gainers ( i.e. there must be a fair number of them to account for the observed distribution of lbvs ) .
it is not clear yet if this is the case , but further study of this may provide an important test
rlof or other binary interaction may produce very asymmetric csm , as seen around many lbvs with bipolar or elliptical nebulae ( @xmath2 car , ag car , hr car , hd 168625 , etc . ) .
using spectropolarimetry , equatorial distributions of csm around sne iin have also been inferred @xcite .
some of the more extreme observed asymmetries in the csm may be difficult to achieve with single stars .
\4 . removing the restriction that lbvs must transition into h - poor wr stars before they die
, mass gainers might retain some of their h envelopes until death .
this would reconcile the problem of lbvs exploding as sne iin .
the mass gained from a more evolved companion may be significantly enriched in nitrogen .
this by itself does not necessarily argue against a single - star scenario if efficient mixing can bring n to the star s
surface @xcite , but the n - enriched nebulae around lbvs are consistent with n enrichment in a binary mass - gainer scenario .
similarly , the accretion of enriched material from a companion could help explain why lbv relative h / he abundances are similar to those of wn stars with h ( actually between those of very luminous wnh stars and other wn stars with h ; @xcite ) , even though their locations on the sky are very different .
when their stripped - envelope companion explodes as a sn ibc or iib , the mass - gainer star is likely to receive a kick .
this provides an attractive explanation for why lbvs would preferentially and systematically seem to avoid star clusters , and adds to their longer lifetimes in explaining their observed isolation .
even if they are born in clusters and live in associations for several myr , they may get kicked out of the cluster . because the initially more massive star will usually explode first
, the kicked mass gainer may have a few more myr to live after that sn event .
traveling at 50 - 100 km s@xmath12 , the star can move about 50 - 100 pc in 1 myr .
such motion for only 1 - 2 myr is therefore sufficient to explain the relative isolation of most of the lbvs discussed here .
it is important to recognize that the mass gainer may not exhibit the lbv instability immediately upon accreting mass , but is likely to do so in its own time when it finishes core - h burning and evolves to become a supergiant .
( in the mean time , it will be a much hotter o - type star that is fainter in visual light ; thus , we would not expect bright lbvs to be seen at the positions of most sne ibc , because it may take them another 10@xmath16 yr to evolve into their own supergiant phase . ) the delayed onset of the lbv phase may lead them to preferentially appear outside clusters . depending on how long it takes them to die , this may also lead sn impostor eruptions and the eventual sne iin to appear isolated .
even if they are a product of binary evolution , lbvs may appear as single stars if their companion has already exploded . in some cases ,
of course , the close companion may not have exploded yet .
if they were born in triple systems , they may still be in binaries .
\9 . in the mw , lmc / smc , and in other nearby galaxies
, there is a large population of massive stars that resemble lbvs but that have not yet exhibited the eruptive lbv instability .
these are usually called `` lbv candidates '' as noted earlier .
@xcite find an order of magnitude more lbv candidates than bona fide lbvs in nearby galaxies , and massive dusty shells around massive stars in the mw suggest a similarly large number here @xcite . counting only lbvs ,
their small number compared to o - type stars has been used to justify a very brief transitional phase ( e.g. , @xcite ) before becoming a wr star . however ,
if lbv candidates are included in the count , the implied lbv lifetime becomes much longer ( see @xcite and references therein ) . in the bifurcated evolutionary scheme discussed above
, the post main - sequence lifetime of the h - rich mass gainer is much longer than the lbv phase envisioned in single - star models , allowing lbv candidates to be the same stars as lbvs , if the lbv eruptive instability is only active for part of that time or in a subset of circumstances .
in addition to mass accretion through rlof , stellar mergers or more exotic systems such as blue thorne - zytkow - like objects ( tzos ) might also yield some of the same results that are compatible with observed properties of lbvs ( overluminous or young compared to surrounding stars , rapid rotators , enriched and asymmetric csm , etc).10@xmath16 @xmath17 ) , rsgs do not exist .
very luminous shell burning stars that would otherwise be luminous rsgs are thought to reside in the blue because of mass loss and instability in their outer envelopes .
the same might apply to the envelopes of tzos in this high - luminosity range , if they exist .
this is still speculative , but we mention it because blue tzos can not be ruled out for lbvs without more detailed study . ] the role of mergers in producing asymmetric csm , in particular , has been discussed extensively ( see review by @xcite and references therein ) .
theoretical predictions for tzos are highly uncertain , and it remains difficult to rule out the possibility that some lbvs are merger products or blue tzos , especially those that may still be in clusters or associations .
however , lacking a significant kick from a companion s sn , stellar mergers do not offer a compelling explanation for why the larger population of lbvs seem to systematically avoid clusters .
so , what about the fates of single stars or binaries that merge early on the main sequence ? if lbvs are exclusively or mostly a binary phenomenon , then it becomes very unclear if wr stars can arise from single stars .
perhaps rsg mass loss becomes critical .
uncertainties ( in most cases overestimates ) in mass loss , plus errors in models that have arisen from comparing single - star models to binary populations , prohibit existing evolution models from making unique predictions ( again , see @xcite ) . with all known stellar mass black holes in binaries , there are few observational constraints on the end fates ( back hole or neutron star ; successful exoplosion or quiet collapse ) for truly single high - mass stars .
a key conclusion from lbv environments as compared to other types of stars is that while lbvs are indeed very massive stars based on their current high luminosity , they did nt necessarily begin their lives that way .
although this paper advocates a different origin for lbvs than is usually assumed , much of the traditionally discussed phenomena associated with lbvs may still apply .
they are still stars with extreme mass loss and instability that have eruptive super - eddington winds or explosive mass loss ( see , e.g. , @xcite ) .
what is very different in the evolutionary scheme advocated above is the path by which the stars have come to be in their current state .
hence , their initial masses and ages are very different than previously assumed .
lbvs have been discussed as a normal transitional state for the most massive stars that have evolved to be in close proximity to the classical eddington limit .
instead , we argue that in most cases they are stars that have become more luminous , rapidly rotating , and unstable as a consequence of mass accretion in a binary system ( or perhaps in some cases something more exotic , such as a merger or blue tzo ) . although the scenario suggested here does not solve the overarching mystery of the physical trigger of lbv eruptions , the evolutionary path to arrive at this point may be a critical part of the puzzle ( especially where angular momentum and stellar structure are concerned ) . this remains a challenge for future theoretical work .
a clear message with broader significance is that cluster turnoffs analyzed with single - star model isochrones will systematically underestimate the age and overestimate the turnoff mass of clusters and associations , because the apparent turnoffs are contaminated by massive blue stragglers .
this method is especially unreliable for very massive stars , where age spreads in clusters are comparable to their absolute age .
similarly , we must be mindful of sn progenitors and their `` initial mass '' , as discussed next .
evidence presented in recent years has begun to shift our central paradigm of massive star evolution . on the one hand ,
wind mass - loss rates are lower than we used to think , while the close binary fraction among o - type stars is shown to be quite high ( roughly 2/3 are interacting , as noted in the introduction ) .
the isolation of lbvs now seems to require a major shift .
the interaction process of rlof or mergers has a profound impact on the types of evolved stars that are observed , and binarity must therefore also have a strong ( or dominant ) impact on the distribution of sn subtypes that mark the end fates of these massive stars perhaps even moreso than metallicity dependent winds and initial mass , as has generally been assumed
. observed statistics of sn subtypes and other considerations already argue that binary rlof and not stellar winds must dominate the removal of the h envelope for most stripped - envelope sn progenitors @xcite , and binary population synthesis studies have suggested this for some time @xcite .
there are too many sne ibc and iib than there are stars initially massive enough to remove their h envelopes via their own winds . including sne iib , the fraction of core - collapse sne that have stripped - envelope progenitors is @xmath036% , whereas stars that are luminous enough to shed their h envelopes through their own winds comprise only 10 - 15% of a normal sn - producing imf @xcite , or perhaps even less if some massive stars collapse to black holes without producing bright sne .
interestingly , from the fraction of o stars that will interact and exchange mass , @xcite estimate that 1/3 is the approximate fraction of massive stars that we might expect to lose their h envelope in rlof . comparing this 33% to the observed value of 36% of ccsne that have stripped - envelope progenitors
, there appears to be little wiggle room to allow massive single stars to die as sne ibc .
similar arguments in favor of binary progenitors have been made based on the mass and composition of the ejecta in stripped - envelope sne @xcite .
a natural question arises .
if binary rlof strips the h envelope to make most sne iib , ib , and ic , then what happens to the mass - gainer companions of these stripped - envelope stars ?
what do those mass - gainer stars look like , and which sne do they yield ? in a theoretical study , @xcite assumed that the spun - up mass gainers would experience enhanced rotational mixing and would undergo quasi - homogeneous chemical evolution , suggesting that they would die as h - free wr stars and possibly grbs .
@xcite therefore inferred that there may be a population of relatively isolated wr stars and grbs . in this paper
we have argued something different : that lbvs might be the direct products of mass transfer in binaries , and in particular , the mass gainers .
( if this is correct , it may suggest that quasi - homogeneous chemical evolution may not be appropriate for these stars , but a further discussion is beyond the scope of this paper . )
we have discussed above that this binary scenario alleviates the apparent paradox of lbvs exploding as sne iin , which is prohibited in the standard single - star evolution framework ( although see @xcite for a possible caveat ) . if lbvs do indeed explode as sne iin , then we should expect the relative isolation of lbvs discussed herein to be reflected in their resulting sne as well . as noted above ,
if these stars receive a kick when a companion in the binary system explodes , they may travel 100 pc or more from their birthsites by the time they die .
they should die at locations that are preferentially outside clusters , causing them to systematically avoid bright h ii regions or blue star clusters . even without a kick , the extended lifetime of an lbv ( because its true initial mass was much lower than its apparent @xmath15 )
would allow it to live much longer then the h ii region that it might have been born in , since the massive stars that ionized the region may have long since died away .
evidence consistent with this from observations of sn host environments has already been published in the literature , although the results were interpreted somewhat differently . as noted in the introduction , anderson and collaborators @xcite
have found that sn iin show a weaker association with h@xmath7 emission in host galaxies , as compared to sne ibc , and their h@xmath7 association is more like that of sne ii - p .
those authors interpreted these results in a context where initial mass dominates sn types ; they attributed the correlation of sne ibc with h@xmath7 to indicate the highest range of initial mass , and they interpreted the lack of correlation between sne iin and h@xmath7 as indicating lower initial masses .
they therefore concluded that lbvs are a less likely progenitor channel for sne iin . as we noted earlier , this result motivated us to examine the environments of nearby lbvs .
we found that despite their high luminosities and initial masses higher than those of most rsgs , lbvs do actually appear to be very isolated from ob star clusters ( moreso than wr stars , and almost as isolated as rsgs ) .
we have argued that this is most likely due to receiving a kick from a companion s sn that preferentially sends the mass gainers flying out of clusters .
binary evolution will have an impact not only on the stars that are kicked ( in this case the lbvs and resulting sne iin ) , but also on the mass donors that explode first .
they ( in this case sne ibc or iib ) will occur preferentially in young clusters as compared to other sn types that are found anywhere , and will be anticorrelated with those that are preferentially kicked out of clusters .
thus , assuming that both sne ibc and sne iin come from a wide range of overlapping initial masses appears to be consistent with the data . because of the important influence of binary evolution , interpreting sn host invironments only in terms of progenitor initial mass is overly simple .
some stars and certain types of sne can selectively move from their birthsites ; a theoretical study of the delay - time distributions and correlations ( or not ) with h ii regions is needed to quantify this effect
. moreover , there are reasons why a correlation with h@xmath7 as adopted by anderson and collaborators is not necessarily a reliable indicator of extreme youth and the highest intial masses .
that interpretation was criticized by @xcite , who pointed out that the brightest ( and therefore the most easily detected h ii regions in distant galaxies ) are giant h ii regions that are actually quite long lived ( @xmath020 myr ) , with ages comparable to the ages of stars with initial masses of only @xmath012 @xmath1 .
anderson and collaborators have based their metric of association with h@xmath7 , and hence youth , as flux - depenent , so that a sn landing on a bright source of h@xmath7 is interpreted as younger than a sn landing on pixels with fainter h@xmath7 .
in some sense this is backwards the brightest h ii region complexes are the longest lived ones with median ages @xmath1110 myr , whereas more isolated , smaller , and fainter h ii regions are fleeting , and have shorter ages of only @xmath03 - 4 myr .
therefore , finding a sn in a smaller isolated h ii region is more likely to indicate a very high initial mass above 60 @xmath1 than if a sn is seen in a very bright and long - lived giant h ii region complex @xcite .
when these fainter and more isolated h ii regions go preferentially undetected in ground - based imaging of distant galaxies , their associated sne may be mistakenly assumed to come from an older population .
note that sn 1987a , with an initial mass of @xmath018 @xmath1 would have a high correlation with star formation as seen from a distant galaxy because of its proximity to 30 dor ( only @xmath01 at a distance of 50 mpc ) , whereas @xmath2 carinae would explode in the fainter carina nebula , which in some cases would not be detected .
anderson et al . have discussed potential biases in detecting various sn types and found that it does not strongly impact the trends in their observations , but there may be other selection effects besides those that govern the discovery of various sn types .
for example , a metallicity bias may cause there to be more of some type of sn in the inner regions of a galaxy as compared to outer regions . in a typical spiral galaxy ,
inner regions are more densely packed with star formation , and they have a larger fraction of the observed surface area covered by very bright h ii regions than in the much sparser outer regions .
indeed , @xcite found that sne ibc in their sample were systematically found at smaller galactic radii than sne ii .
thus , even if sn types have the same range of progenitor initial mass , a type of sn that has a preference for smaller galactocentric radii because of metallicity may have a higher probably of coincidence with a very bright h ii region .
the potential influence of this bias deserves further attention .
the sensitivity and potential bias in the flux - dependent h@xmath7 diagnostic used by anderson and collaborators @xcite may explain why their conclusions differed from other studies of sn host environments .
earlier studies of spatial proximity to h ii regions ( rather than the strength of h@xmath7 flux ) found no statistically significant preference between sne ii and ibc @xcite
. some of these might be chance coincidence with long - lived giant h ii regions at relatively course ground - based angular resolution . in general
, one expects no direct correlation between a sn and its own h ii region for any but the most massive stars above 75 @xmath1 @xcite .
indeed , @xcite found that high resolution _
images for nearby sne with good constraints on the progenitor stars do not support a correlation with h ii regions , and @xcite argued against the monotonic increasing progenitor mass from type ii to iib to ibc based on this and a number of other considerations ( see also @xcite ) . when examining the sn host color at the location of sne rather than host h@xmath7 , @xcite found that sne ibc , sne ii , and sne iin show basically the same distribution of @xmath18 colors .
they did , however , find bluer colors for broad - lined sne ic and sne iib ; it is unlikely that this is due to initial mass , since progenitor detections suggest rather low masses for sne iib .
one may ask , if lbvs are the mass gainers that get a kick from their companion s sn ibc in a cluster or association , then why do we not detect bright lbvs at the positions of sne ibc ?
the most likely answer is that the mass gainer will appear as a more luminous but hot o - type star that is still quite difficult to detect ( see figure [ fig : hrd ] ) , and may not become a visually bright lbv until 1 - 2 myr later , when its own core evolution drives it off the main sequence to become a cooler b[e]/lbv supergiant . moreover , in the binary scenario , the majority of sne ibc will be from relatively low initial masses , compared to the high - mass wr progenitors envisioned in the single - star scenario .
overall , we find it likely that the sequential order of explosion in a binary system and subsequent kicks out of clusters ( and to a somewhat lesser extent , metallicity and star cluster density ) , plus the extended main sequence lifetime before mass accretion , will have a more important influence on the observed environments of core - collapse sne than monotonic differences in the range of initial mass .
this may be especially true for more extreme or rare explosions that may require special evolutionary paths in binaries , such as long grbs @xcite .
of course , the observed isolation of sne iin might also be influenced to some degree by sn iin progenitor contamination from non - lbv objects .
indeed , there is strong evidence for this already , although the fraction of contamination is poorly constrained . in principle
, any type of explosion can yield a type iin event , since the iin designation depends on csm interaction and not the explosion mechanism . while lbv mass loss fits the bill @xcite , the case for lbvs is strongest for the most luminous sne iin where the csm mass budget requires extreme parameters @xcite .
b[e ] supergiants are also suitable for providing moderate - luminosity sne iin , and their spatial distribution is even more isolated than lbvs , comparable to rsgs ( figures [ fig : cum ] and [ fig : cum2 ] ) .
moreover , the concentration of eruptive mass loss associated with the final nuclear burning sequences in the last few years before core collapse @xcite means that the progenitor need not have been in a dense - wind phase for very long . besides massive lbvs exploding as sne iin , there are also examples where sne iin may result from : ( 1 ) extreme rsgs with strong clumpy winds @xcite , ( 2 ) electron - capture sne ( ecsne ) from 8 - 10 @xmath1 super - agb stars ( such as the class of sne iin - p and possibly the crab nebula ; @xcite ) , and ( 3 ) some hybrid type ia / iin events that are apparently thermonuclear sne ia exploding in a dense h - rich csm , like sn 2002ic and several similar objects @xcite .
when these other objects having much lower mass progenitors are included with sne iin , it would make the mixed population appear statistically even more isolated .
this is the likely explanation for why @xcite found that sne iin had a spatial distribution similar to sne ii - p , even though we find lbvs have isolation intermediate between wr stars and rsgs .
it is worth reiterating the point that not all mass gainers will become lbvs , since this likely depends on the mass accreted and final luminosity ( many may become b[e ] supergiants instead , or they may evolve to become rsgs , etc . ) , and not all mass gainers will necessarily become sne iin , since these sne require special conditions ( pre - sn mass ejections ) . comparing the observed fraction of sne iin ( 89% of core collapse sne ; @xcite ) to theoretical expectations
requires continued modeling of the evolution of mass gainers , which still depends on uncertain assumptions .
in addition to core - collapse sne , caution about inferring an initial mass also applies to sn impostors and related transients .
some eruptive transients have been discussed where the apparent age of surrounding stellar population indicates a surprisingly low initial mass .
the most representative cases are sn 2008s and the 2008 transient in ngc 300 @xcite . if lbvs or lower - mass analogs preferentially receive a kick from an exploded companion , then this needs to be included in the uncertainty when inferring an age and initial mass from their surrounding stellar populations .
one last important consequence to mention is that stars with initial masses below 8 @xmath1 may , if in a binary system , accrete mass and explode as a ccsn or an electron - capture sn , even if their initial mass would have precluded that fate in a single - star scenario . due to the slope of the imf , these initially lower - mass stars
may potentially contribute a substantial number of observed sne .
we encourage a detailed study of this effect , including the delay time distribution and its contribution to sn statistics .
this paper suggests a major reversal in our interpretation of the nature of lbvs . instead of being the result of instability in very massive single stars that have evolved to become unstable through their own core evolution and mass loss , they must preferentially or exclusively be the products of binary evolution .
they are most likely the mass gainers in binary rlof or mergers because their ages and initial masses ( inferred from single - star evolution tracks ) systematically disgree with the stars around them . while counter arguments can point to one or two possible exceptions
, this must apply to most of the lbvs .
why was this seemingly obvious result missed ?
the answer may be partly sociological : lbvs are very few in number , and @xmath2 car garners disproportionate attention , so its location in the carina nebula did not arouse suspicion .
also , interest in lbvs has concentrated more on driving their eruptive mass loss , as opposed to the evolutionary history that led them to be unstable in the first place .
last , lbvs seem to fit nicely into the picture of single - star evolution ( and indeed they are needed for it ) , so the question of whether or not their environments contradict models was not pursued with adequate vigor . actually , a suggestive result along similar lines was found more than a decade ago by n.l .
king and collaborators ( 2000 ; see also @xcite , as well as @xcite ) . based on environments of some candidate lbvs in m31
, it was found that they seem to reside outside ob associations and clusters , implying stellar ages of several myr .
that work did not instigate a major rethinking of our standard paradigm of massive star evolution because the paper was not accepted for publication in the refereed literature ( the first author has since left the field ) . in hindsight
, the main result of that work appears to have pointed in the correct direction after all .
ns thanks j.s .
gallagher for discussions regarding binary evolution and lbvs , and in particular for drawing our attention to previous work by n. king .
we benefitted from discussion about sn host environments with p. crowther , j. anderson , and p. kelly , and from further discussions concerning binary evolution with s. de mink .
we also thank the referee ( j. eldridge ) for a careful reading and thoughtful comments that improved the paper .
partial support was provided by nsf award ast-1312221 .
this research has made use of the simbad database , operated at cds , strasbourg , france .
smith n , li w , silverman jm , ganeshalingam m , filippenko av .
2011a , mnras , 415 , 773 smith
n , li w , filippenko av , chornock r. 2011b , mnras , 412 , 1522 smith n , li w , miller aa , silverman jm , filippenko av , cuillandre jc , cooper mc , matheson t , van dyk sd .
2011c , apj , 732 , 63 | based on their relatively isolated environments , we argue that luminous blue variables ( lbvs ) must be primarily the product of binary evolution , challenging the traditional single - star view wherein lbvs mark a brief transition between massive o - type stars and wolf - rayet ( wr ) stars .
if the latter were true , then lbvs should be concentrated in young massive clusters like early o - type stars .
this is decidedly not the case .
examining locations of lbvs in our galaxy and the magellanic clouds reveals that , with only a few exceptions , lbvs systematically avoid clusters of o - type stars . in the large magellanic cloud ,
lbvs are statistically much more isolated than o - type stars , and ( perhaps most surprisingly ) even more isolated than wr stars .
this makes it impossible for lbvs to be single `` massive stars in transition '' to wr stars .
instead , we propose that massive stars and supernova ( sn ) subtypes are dominated by bifurcated evolutionary paths in interacting binaries , wherein most wr stars and sne ibc correspond to the mass donors , while lbvs ( and their lower - mass analogs like b[e ] supergiants , which are even more isolated ) are the mass gainers . in this view , lbvs are evolved massive blue stragglers . through binary mass transfer ,
rejuvinated mass gainers get enriched , spun up , and sometimes kicked far from their clustered birthsites by their companion s sn .
this scenario agrees better with lbvs exploding as type iin sne in isolation , and it predicts that many massive runaway stars may be rapid rotators .
mergers or blue thorne - zykow - like objects might also give rise to lbvs , but these scenarios may have a harder time explaining why lbvs avoid clusters .
[ firstpage ] binaries : general stars : evolution stars : winds , outflows |
open quantum systems are at the heart of many physical phenomena from nuclear physics to quantum information theory @xcite . in fact , all `` real '' quantum systems are , to some extent , open systems .
interactions with the environment cause decoherence , resulting in non - equilibrium dynamics .
it is often simpler to design experiments that probe non - equilibrium physics than it is to design experiments that probe equilibrium physics .
conversely , the theoretical toolkit for describing systems in equilibrium is generally much farther developed than that for describing systems in non - equilibrium .
ultracold atom systems provide a platform for realizing clean and tunable quantum systems @xcite . over the past few years
, much effort has gone into describing non - equilibrium experiments that are accessible , within approximate or exact frameworks , to theory .
notable experiments are the equilibration dynamics of one - dimensional bose gases @xcite , the spin dynamics of dipolar molecules in optical lattices with low filling factor @xcite , and the tunneling dynamics of effectively one - dimensional few - fermion systems @xcite .
this paper focuses on the latter set of experiments .
specifically , the goal of the present work is to describe the tunneling dynamics of few - fermion systems , which are prepared in a well defined quasi - eigenstate ( metastable state ) , into free space .
we consider small systems and directly solve the time - dependent schrdinger equation in coordinate space .
as we will show , this approach provides a means to quantify the importance of the particle - particle interaction , covering time scales from a fraction of the trap scale to thousands times the trap scale .
alternatively , one could adopt a quantum optics perspective and pursue a master equation approach .
tunneling is arguably the most quantum phenomenon there is : if the system was behaving classically , tunneling would be absent @xcite .
tunneling plays an important role across physics , chemistry and technology .
the scanning tunneling microscope @xcite , for example , nicely illustrates how a physics phenomenon , the tunneling of electrons , has been turned into a powerful practical tool ( the imaging of materials ) .
the @xmath1-decay , i.e. , the decay of a @xmath2he nucleus from a heavy nucleus , is an example discussed in most undergraduate physics texts ( see , for example , ref .
the typical picture is to identify an effective reaction coordinate and to obtain the tunneling rate from a wkb analysis .
while powerful , such treatments completely neglect the effect of interactions .
interactions also play a crucial role in sorting out under which conditions electrons in light atoms tunnel sequentially or simultaneously @xcite . the two - particle system considered in this work
has been realized experimentally and is the possibly simplest scenario that deals with a truly open quantum system ( the atoms can escape to infinity ) in which interactions ( short - range atom - atom interactions ) play a crucial role .
as we will show , even for this relatively simple set - up , matching theory and experiment is a non - trivial task .
of course , two - particle tunneling has been investigated previously in this and related contexts @xcite .
the remainder of this paper is organized as follows .
section [ sec_system ] introduces the hamiltonian , the heidelberg experiment and selected simulation details .
sections [ sec_molecular ] and [ sec_upper ] discuss the molecular and upper branch tunneling dynamics . for both cases , it is argued that the trapping potential needs to be reparameterized . using the reparameterized trapping potential , numerical simulations for the tunneling dynamics of two distinguishable @xmath0li atoms on the molecular branch and the upper branch
are discussed .
comparisons with the experimentally measured tunneling rates are presented .
finally , sec .
[ sec_summary ] summarizes and provides an outlook .
simulation details and some technical aspects are relegated to appendices [ app_br][app_flux - anal ] .
this section considers a single @xmath0li atom with mass @xmath3 .
the atom is assumed to be in the hyperfine state @xmath4 .
we consider the three lowest hyperfine states of the @xmath0li atom , referred to as @xmath5 , @xmath6 , and @xmath7 .
figure [ fig_br - energy ] li as a function of the magnetic field strength @xmath8 .
solid , dashed and dotted lines correspond to states @xmath9 , @xmath10 , and @xmath11 , respectively ( see text for details ) .
states @xmath9 and @xmath10 are used in the upper branch experiments @xcite , while states @xmath9 and @xmath11 are used in the molecular branch experiments @xcite .
the higher - lying energy states shown by dash - dotted lines are not relevant for the present paper .
, scaledwidth=40.0% ] shows the dependence of the hyperfine energy levels on the magnetic field strength @xmath8 .
the atom with coordinates @xmath12 is trapped optically in a non - separable potential that is much tighter in the @xmath13-direction ( @xmath14 ) than in the @xmath15-direction @xcite . throughout this work
, we do not simulate the motion in the tight transverse confining direction .
the transverse trapping frequency does , however , enter into the calculation of the renormalized one - dimensional coupling constant ( see sec .
[ subsec_two - body_ham ] ) . evaluating the confinement created by the gaussian laser beam at @xmath16 , the effective one - dimensional single - particle hamiltonian @xmath17 reads @xcite @xmath18 where the trapping potential @xmath19 along the @xmath15-direction depends implicitly on the internal or hyperfine state @xmath20 of the atom through the coefficient @xmath21 , @xmath22 the first term on the right hand side of eq .
( [ vtrap ] ) accounts for the optical confinement .
@xmath23 denotes the maximum depth of the trap , @xmath24 a time - dependent parameter [ @xmath25 , and @xmath26 the rayleigh range of the laser beam that produces the confinement . the second term on the right hand side of eq .
( [ vtrap ] ) is linear in @xmath15 and makes the tunneling possible .
@xmath27 is the bohr magneton and @xmath28 depends on the hyperfine state , magnetic field strength and magnetic field gradient @xmath29 , @xmath30 here , @xmath31 is a dimensionless parameter close to @xmath32 ( see below for details ) . table [ table - exp ] .parameters from refs .
@xcite that define the trapping potential . since the energy of the two - particle system on the molecular branch is smaller than the energy of the two - particle system on the upper branch , the @xmath33 value for the molecular branch is chosen to be smaller than that for the upper branch
; this guarantees that the tunneling rates for the two experiments have roughly comparable orders of magnitude .
the harmonic oscillator units are defined in terms of @xmath34hz , corresponding to @xmath35j , @xmath36 m , and @xmath37s , or @xmath38 , @xmath39 , and @xmath40 . in an alternative levitation measurement ,
the magnetic field gradient was found to be @xmath41g / m @xcite . [
cols="^,^",options="header " , ] due to the dependence of @xmath42 on the state ( through the wkb energy ) , the wkb rates @xmath43 for the three states vary by about a factor of 6 for cases ( a ) and ( b ) . for the parameters considered in table [ table_app_wkb ] and in the main text , the wkb rate for the ground state is smaller than that obtained through the full time propagation , with the ratio @xmath44 depending on the exact shape of the trap . for the excited states , in contrast , the wkb rates are larger than those obtained through the full time propagation . 49ifxundefined [ 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.com/books?id=dkcjpwaacaaj[__ ] ( , ) https://books.google.com/books?id=znjvhah8qa4c[__ ] ( , ) link:\doibase 10.1103/revmodphys.80.885 [ * * , ( ) ] http://stacks.iop.org/0034-4885/75/i=4/a=046401 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.80.1215 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.82.1225 [ * * , ( ) ] link:\doibase 10.1126/science.1257026 [ * * , ( ) ] link:\doibase 10.1038/nature12483 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.108.075303 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.111.175302 [ * * , ( ) ] https://books.google.com/books?id=acj0nqeacaaj[__ ] ( , ) link:\doibase 10.1103/revmodphys.86.1127 [ * * , ( ) ] https://books.google.com/books?id=yfo3rnt3bkec[__ ] ( , ) link:\doibase 10.1126/science.1163439 [ * * , ( ) ] link:\doibase 10.1073/pnas.1201345109 [ * * , ( ) ] link:\doibase 10.1103/physreva.87.043626 [ * * , ( ) ] http://stacks.iop.org/0953-4075/44/i=19/a=195301 [ * * , ( ) ] http://stacks.iop.org/0953-4075/42/i=4/a=044018 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.108.115302 [ * * , ( ) ] link:\doibase 10.1103/physreva.88.043633 [ * * , ( ) ] link:\doibase 10.1103/physreva.91.041601 [ * * , ( ) ] https://books [ _ _ ] , ( , ) link:\doibase 10.1103/physrev.38.2082.2 [ * * , ( ) ] @noop ( ) link:\doibase 10.1103/physrevlett.110.135301 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.81.938 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.110.203202 [ * * , ( ) ] link:\doibase 10.1103/physreva.84.043619 [ * * , ( ) ] http://stacks.iop.org/1367-2630/7/i=1/a=192 [ * * , ( ) ] http://stacks.iop.org/1367-2630/11/i=7/a=073031 [ * * , ( ) ] link:\doibase 10.1103/physreva.89.023603 [ * * , ( ) ] @noop ( ) .
link:\doibase 10.1103/physreva.70.042709 [ * * , ( ) ] m. l. wall , k. r. a. hazzard , and a. m. rey , " effective many - body parameters for atoms in nonseparable gaussian optical potentials , phys .
a * 92 * , 013610 ( 2015 ) .
https://books.google.com/books?id=i-40vaxqrj0c[__ ] , graduate texts in contemporary physics ( , ) https://books.google.com/books?id=t_7hg08x7cmc[__ ] , international series of monographs on physics ( , ) @noop * * ,
( ) \doibase http://dx.doi.org/10.1063/1.448136
[ * * , ( ) ] \doibase http://dx.doi.org/10.1016/0021-9991(91)90137-a [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/0375-9474(80)90509-6 [ * * , ( ) ] \doibase http://dx.doi.org/10.1016/0009-2614(86)80262-7 [ * * , ( ) ] link:\doibase 10.1103/physreva.37.973 [ * * , ( ) ] link:\doibase 10.1103/physreva.43.68 [ * * , ( ) ] link:\doibase 10.1103/physreva.91.043607 [ * * , ( ) ] @noop * * , ( ) \doibase http://dx.doi.org/10.1063/1.3126363 [ * * , ( ) ] | we present one - dimensional simulation results for the cold atom tunneling experiments by the heidelberg group [ g. zrn _ et al .
_ , phys . rev
. lett . * 108 * , 075303 ( 2012 ) and g. zrn _
et al . _ ,
phys .
rev . lett . * 111 * , 175302 ( 2013 ) ] on one or two @xmath0li atoms confined by a potential that consists of an approximately harmonic optical trap plus a linear magnetic field gradient . at the non - interacting particle level , we find that the wkb ( wentzel - kramers - brillouin ) approximation may not be used as a reliable tool to extract the trapping potential parameters from the experimentally measured tunneling data .
we use our numerical calculations along with the experimental tunneling rates for the non - interacting system to reparameterize the trapping potential . the reparameterized trapping potentials serve as input for our simulations of two interacting particles . for two interacting ( distinguishable ) atoms on the upper branch ,
we reproduce the experimentally measured tunneling rates , which vary over several orders of magnitude , fairly well . for infinitely strong interaction strength
, we compare the time dynamics with that of two identical fermions and discuss the implications of fermionization on the dynamics . for two attractively - interacting atoms on the molecular branch
, we find that single - particle tunneling dominates for weakly - attractive interactions while pair tunneling dominates for strongly - attractive interactions .
our first set of calculations yields qualitative but not quantitative agreement with the experimentally measured tunneling rates .
we obtain quantitative agreement with the experimentally measured tunneling rates if we allow for a weakened radial confinement . |
the determination of @xmath7 based on the small magellanic cloud can have at least four significant advantages and one disadvantage with respect to those based on distant h region complexes : a ) no underlying absorption correction for the helium lines is needed because the ionizing stars can be excluded from the observing slit , b ) the determination of the helium ionization correction factor can be estimated by observing different lines of sight of a given h region , c ) the accuracy of the determination can be estimated by comparing the results derived from different points in a given h region , d ) the electron temperature is generally smaller than those of metal poorer h regions reducing the effect of collisional excitation from the metastable 2 @xmath8s level of he , and e ) the disadvantage is that the correction due to the chemical evolution of the smc is in general larger than for the other systems .
the determination of the pregalactic , or primordial , helium abundance by mass @xmath7 is paramount for the study of cosmology , the physics of elementary particles , and the chemical evolution of galaxies ( e. g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* and references therein ) . in this paper
we present a new determination of @xmath7 based on observations of the smc .
this determination is compared with those carried out earlier based on extremely metal poor extragalactic h regions .
long slit spectra were obtained at ctio during two observing runs , in august and september 1990 , with the 4-m telescope equipped with the r - c spectrograph and a coated gec ccd detector . using three different gratings ( at first order ) the spectral ranges between @xmath0 3440 - 5110 ,
4220 - 7360 , and 5800 - 7370 were covered .
the slit , oriented e - w , was 4.7 arcmin long and 1.6 arcsec wide ; the scale along the dispersion axis was 0.73" / pixel .
the resolution was 7 for the blue and red wavelength ranges and 14 for the intermediate one .
the slit was placed at two different positions of the nebula , five extraction windows were defined in one slit position ( regions 1 - 5 ) and eight in the other ( regions 11 - 18 ) , four of the extraction windows included the brightest ionizing stars ( m@xmath914 ) , while the other 9 positions avoided stars brighter than m = 17 to minimize the stellar contamination of the nebular spectra .
table [ positions ] presents the positions and sizes of the extraction windows . in table
[ positions ] we also present the observed h@xmath10 fluxes , @xmath11(h@xmath10 ) , before correction for extinction .
@xcite present excellent pictures of the stellar cluster of ngc 346 where the ionizing stars included in table [ positions ] are indicated , and the observed regions can be located .
@xcite also present excellent pictures of ngc 346 where the filamentary structure of the nebula can be appreciated .
also in table [ positions ] we define regions a and b. region a is the sum of regions 2 , 3 , 5 , 12 , 13 , 17 , and 18 .
region a was defined to minimize the errors of the emission lines and the effect of stellar underlying absorption ; the main results of this paper will be based on it .
regions 1 and 11 were not included in region a because they are fainter than the others ( see table [ positions ] ) and the equivalent widths of their emission lines are also smaller ( see the last paragraph of this section ) , both effects could increase the systematic errors .
region b is the sum of all observed regions and will be used to show the effect of the underlying absorption on the measured line intensities . to flux calibrate the spectra five or six spectrophotometric standards , from @xcite ,
were observed each night with the slit widened to 6.4 " .
a he - ne - ar lamp was used to perform the wavelength calibration .
dome - flats and sky - flats were obtained to flatten the red frames while a quartz lamp flat and a sky flat were used for the blue range frames .
several bias frames were taken each night .
data reduction was performed using the iraf reduction package .
sky subtraction was made from observations taken one degree away from the nebula before and after each nebular observation .
short exposure frames at each position were used to remove cosmic rays and to measure fluxes of strong lines saturated in the longer exposures . in figures [ blue ] , [ red ] and [ red - detail ]
we show spectra of region a at different wavelength ranges . in figure
[ blue - detail ] we compare regions a and b near the balmer limit to show the effect of the underlying absorption on the balmer lines and on @xmath12 4026 of he . in tables [ int - a ] , [ int-355 ] , and [ int-324 ] we present the intrinsic line intensities , @xmath13 , given by @xmath14 where @xmath15 is the observed line flux corrected for atmospheric extinction and @xmath16(h@xmath10 ) is the logarithmic reddening correction at h@xmath10 , and @xmath17 is the reddening function .
for @xmath17 we adopted the normal extinction law @xcite .
@xmath16(h@xmath10 ) was obtained by fitting the observed balmer decrement with that computed by @xcite for @xmath18 = 12 000 k , and @xmath19 = 100 @xmath20 ; the balmer decrement is almost insensitive to the expected variations in @xmath18 and @xmath19 over the observed volumes .
table [ int - a ] presents the line intensities for regions a and b that were obtained after adding the spectra of all their components , while tables [ int-355 ] and [ int-324 ] present the line intensities for regions 1 - 5 and 11 - 18 respectively . the @xmath16(h@xmath10 ) value for region a amounts to @xmath21 , a value in good agreement with the values derived from the stellar data of the cluster by @xcite and @xcite that amount to 0.20 and 0.18 respectively .
we have adopted the @xmath16(h@xmath10 ) value of region a for all the observed regions .
the differences between the adopted @xmath17 and the reddening function in the visual region in the direction of ngc 346 are small , moreover their effect on the determination of the line intensity ratios is negligible because with the adopted @xmath16(h@xmath10 ) value we have recovered the theoretical balmer decrement of the four brightest lines that are not expected to be affected by underlying absorption .
the smc reddening law in the uv is very different to the normal galactic one , but fortunately in the visual both laws are similar @xcite ; according to @xcite @xmath22 amounts to @xmath23 for the smc . from the use of various reddening laws with @xmath24 in the 3.0 to 3.2 range
@xcite we estimate that the error introduced by the adopted reddening law is about 0.002 dex for all line ratios and smaller than 0.001 dex for ratios of lines closer than 500 . in tables
[ int - a ] , [ int-355 ] , and [ int-324 ] one-@xmath25 errors are included , the errors were estimated by comparing the results derived from the two different observing seasons .
the total number of photons received by each of the he lines of region a is in the @xmath26 to @xmath27 range , therefore the errors presented in table [ int - a ] are about a factor of five larger than those given by photon statistics .
the errors presented for region a are larger because they include other sources of error present in the reduction procedure .
the line intensities in tables [ int - a ] , [ int-355 ] , and [ int-324 ] were not corrected for underlying stellar absorption with the exception of the h@xmath28 lines of regions b , 4 , 14 , 15 , and 16 , where we adopted an underlying absorption of 2 ; for these regions we adopted the theoretical @xmath29(h@xmath28)/@xmath29(h@xmath10 ) value to normalize all the lines to h@xmath28 , notice that h@xmath10 has an intensity smaller than unity because it has not been corrected for underlying absorption .
the @xmath12 3727 line intensity was corrected for the contribution due to h13 and h14 and to @xmath12 3724 of [ s ] , these contributions were estimated from the other balmer lines and from the @xmath12 6312 of [ s ] ; @xmath29(4711 ) of [ ar ] was obtained after subtracting the expected contribution of @xmath12 4713 of he based on the work by @xcite ; @xmath29(3889 ) of he was obtained after subtracting the expected contribution of h8 based on the work by @xcite . in table [ e - w ] we present the h and he equivalent widths for all the observed regions .
the equivalent widths of regions b , 4 , 14 , 15 , and 16 are strongly affected by the stellar underlying absorption . no correction due to underlying absorption has been made for any line in this table . after correcting region
a for extinction , based on the four brightest balmer lines , it is found that the weaker balmer lines ( h9 to h12 ) are not affected by stellar underlying absorption ( see figure [ blue - detail ] ) , therefore the he lines are not expected to be affected by underlying absorption . from similar arguments
we expect each of the regions included in region a ( 2 , 3 , 5 , 12 , 13 , 17 , and 18 ) to be unaffected by underlying absorption .
we derived the temperatures and densities presented in tables [ tandd ] and [ tandda ] based on the program of @xcite for forbidden lines .
@xmath30(o ) , @xmath30(o ) , and @xmath19(s ) values were derived from the 4363/5007 , 3727/7325 and 6716/6731 ratios respectively ; to derive @xmath30(o ) we considered the contribution by recombination to the @xmath0 7320,7330 line intensities @xcite .
we will define another temperature given by @xmath31 where @xmath32(o@xmath33 ) and @xmath32(o@xmath34 ) are obtained in section [ ica ] ( see table [ tandda ] ) . to derive the root mean square density , @xmath19(rms )
, we adopted the following equation @xmath35,\ ] ] where @xmath36(h@xmath28 ) is the effective recombination coefficient ( e. g. * ? ? ?
* ) , @xmath37 is the distance to the smc ( 64 kpc * ? ? ? * and references therein ) , and @xmath38 is the radius of the adopted homogeneous sphere ( 150 arcsec ) . from the he@xmath33/h@xmath33 and he@xmath34/h@xmath33 values for region a presented in the next section , the @xmath36(h@xmath28 ) value for @xmath30 = 11950 k , and the @xmath29(h@xmath28 ) measured by @xcite we find that @xmath39 ( see table [ tandda ] ) .
it can be shown that , in the presence of density fluctuations ( which is always the case ) , @xmath19(rms ) provides us with a lower limit for the local density , @xmath19(local ) . to derive the abundance ratios we need to use @xmath19(local ) values ; in the presence of density fluctuations @xmath40(rms ) = @xmath41(local ) , where @xmath42 is the filling factor . from our @xmath19(rms ) value and the @xmath19(he ) @xmath43 present in table [ tandda ] ( defined in the last paragraph of this section ) it follows that 0.01 is a representative value for @xmath42 . from the ratio of the balmer continuum flux to a balmer line
flux it is possible to derive the temperature @xmath18(bac ) , from the balmer line emissivities @xcite and the continuum emissivities for the he and h continuum @xcite we find the @xmath18(bac ) value presented in table [ tandda ] . by combining @xmath30(bac ) , @xmath30(o ) , and @xmath30(o ) , together with @xmath32(o@xmath33 ) and @xmath32(o@xmath34 ) ( see section [ ica ] ) , and assuming that @xmath44(o ) = @xmath44(o ) , it is possible to determine the mean square temperature variation @xmath44(h ) @xmath45 @xmath44 , the average temperature @xmath46(h ) @xmath45 @xmath46 , and @xmath44(o ) over the observed volume since @xcite @xmath47 @xmath48 @xmath49 @xmath50,\ ] ] @xmath51,\ ] ] @xmath52 in region a @xmath30(o ) @xmath53 @xmath30(o ) , therefore @xmath46(o ) @xmath54 @xmath46(o ) , and from equation ( [ t^2 ] ) it follows that @xmath55(o ) . in table
[ tandda ] , we present the @xmath44(bac , o + o ) value for region a. in the low density and low optical depth limit it follows that the emissivities of the helium and hydrogen lines are proportional to powers of the temperature and consequently that @xmath18(he ) is given by @xcite @xmath56 \\ & = & t_0(1 - 1.43t^2),\end{aligned}\ ] ] where @xmath57 is the average value of the power of the temperature for the helium lines and @xmath10 for h@xmath10 .
the @xmath28 powers in the low density limit for the @xmath0 3889 , 4026 , 4388 , 4471 , 4922 , 5876 , 6678 , 7065 , and 7281 lines are -0.72 , -0.98 , -1.00 , -1.02 , -1.04 , -1.12 , -1.14 , -0.55 , and -0.60 , respectively @xcite , and @xmath58 ( e. g. * ? ? ?
* ) . in the low density limit and weighted
according to the observational errors we obtain @xmath59 ; for @xmath19(he ) = 143 @xmath20 we obtain @xmath60 . from @xmath44(bac , o + o ) and @xmath46(bac , o + o ) derived from equations ( [ t0bac ] - [ t0o ] ) together with equation ( [ t0he ] ) we obtain @xmath18(he ) @xmath61 , which is a representative temperature for the he lines .
based on nine emission line ratios in the next section we derive @xmath32(he@xmath33)/@xmath32(h@xmath33 ) , @xmath18(he ) and @xmath19(he ) self - consistently , hereinafter @xmath19(he ) @xmath43 and @xmath18(he ) @xmath43 ( which are presented in table [ tandda ] ) . in table
[ tandda ] we also include @xmath44(he , o + o ) derived from equations ( [ t0oiii ] ) , ( [ t0oii ] ) , ( [ t0o ] ) , and ( [ t0he ] ) ; and @xmath44(bac , he , o + o ) , an average of the two @xmath44 determinations ; @xmath62(he ) @xmath63 which is the average of @xmath18(he ) @xmath64 and @xmath18(he ) @xmath43 ; and @xmath65(he ) @xmath63 which is the density that corresponds to @xmath62(he ) @xmath63 . from this discussion and the values in table [ tandda ] we conclude that @xmath66 and that to a very good approximation @xmath67 .
to determine the abundances from collisionally excited lines many authors adopt a two temperature scheme , with @xmath30(o ) for the high degree of ionization zones and @xmath30(o ) or @xmath30(n ) for the low degree of ionization zones .
therefore to compare with the abundances determined from collisionally excited lines by other authors we will assume also that that the temperature within the o and o zones is constant , that is @xmath44(o ) = @xmath44(o ) = 0.000 .
two points should be stated here : ( 1 ) under the assumption that @xmath44(o ) = @xmath44(o ) = 0.000 and since @xmath30(o ) @xmath68(o ) , from equation ( [ t^2 ] ) , it follows that @xmath69 ; from the @xmath30(o ) and @xmath30(o ) values for region a ( see table [ tandd ] ) and equations ( [ t^2 ] ) and ( [ t0o ] ) , we find that @xmath70 ; ( 2 ) the @xmath70 abundances are a lower limit to the real abundances since the assumption that @xmath44(o ) = @xmath44(o ) = 0.000 implies constant temperature within the o and o zones , which is not the case ( see section [ t&d ] ) .
the total abundances for @xmath44(h ) = 0.022 , our preferred @xmath44 value , will be discussed in section [ ta ] .
therefore we have determined the ionic abundances of the heavy elements for all the regions observed using @xmath71 , the abundances are presented in table [ regionions ] .
the abundances were computed with the program presented by @xcite . to determine the o@xmath34 , ne@xmath34 , s@xmath34 , ar@xmath34 , and ar@xmath72 abundances we used @xmath30(o ) , while for n@xmath33 , o@xmath33 , and s@xmath33 we used @xmath30(o ) . for @xmath30(o ) we adopted the value derived for each region ( see table [ tandd ] ) . on the other hand ,
it can be seen in table [ tandd ] that the @xmath30(o ) temperatures are systematically higher for those regions that include bright stars and have relatively smaller @xmath73(h@xmath10 ) values ( regions 4 , 14 , 15 , 16 , and b ) than for the regions that have larger @xmath73(h@xmath10 ) values ; however , this effect is not real and has to do with the difficulty of establishing a proper continuum baseline owing to the presence of underlying balmer lines in absorption in the 37003750 region , where the @xmath123727 lines originate .
this effect , together with large @xmath30(o ) errors presented in table [ tandd ] , which are due to other causes , led us to assume for all regions that @xmath30(o ) @xmath74(o ) , the value determined for region a. to estimate the h@xmath10 emissivity we adopted @xmath75 @xmath76 where @xmath29(h@xmath77o@xmath33 ) and @xmath29(h@xmath77o@xmath34 ) are the intensities of h@xmath10 in the o@xmath33 and o@xmath34 regions respectively .
we consider our procedure adequate since the total o abundances derived for all the regions show a small dispersion despite their different ionization degree ( compare for example the total o abundances of regions 16 and 18 that are practically the same ) .
to obtain he@xmath33/h@xmath33 values we need a set of effective recombination coefficients for the he and h lines , the contribution due to collisional excitation to the helium line intensities , and an estimate of the optical depth effects for the helium lines .
the recombination coefficients that we used were those by @xcite for h , and @xcite for he .
the collisional contribution was estimated from @xcite and @xcite . from the intensity of @xmath123614 and the computations of @xcite
it was found that the he singlet lines were produced under case b. the optical depth effects in the triplet lines were estimated from the computations by @xcite . to derive the he@xmath33/h@xmath33 value of region a
, in addition to the balmer lines , we made use of nine he lines , @xmath0 3889 , 4026 , 4387 , 4471 , 4922 , 5876 , 6678 , 7065 , and 7281 to determine @xmath19(he ) and @xmath18(he ) self - consistently .
each of the 9 he /h
line ratios depends on @xmath18 , @xmath19 , @xmath78 , and the optical depth of 3889 ( @xmath79 ) , and each dependence is unique .
therefore we have a system of 9 equations and 4 unknowns .
we decided to obtain the best value for the 4 unknowns by minimizing @xmath80 .
the @xmath80 value is given by : @xmath81 where @xmath82 is the absolute error in the measurement that can be obtained from table [ int - a ] , and @xmath83 is the abundance derived from each line for those parameters .
the best @xmath79 value is slightly negative which is unphysical , moreover from cloudy models @xcite it is found that @xmath79 is close to zero , therefore we decided to adopt @xmath84 and to use equation ( [ eq - chi ] ) to derive @xmath18 , @xmath19 , and @xmath85 self - consistently . for a system with 9 independent determinations and 3 unknowns we have 6 degrees of freedom and we expect the minimum @xmath80 to be in the range @xmath86 at the 90% confidence level .
the value of @xmath87 found in table [ chi ] is in excellent agreement with this range . in table
[ regionhe ] we present he@xmath33/h@xmath33 values for different temperatures and densities ; the temperatures were selected to include @xmath30(o ) , @xmath30(bac ) , @xmath30(he ) @xmath43 , and two representative temperatures ; the densities were selected to include the minimum @xmath80 at each one of the five temperatures . the temperature with the minimum @xmath80 is the self - consistent @xmath30(he ) and amounts to @xmath88 k ; this temperature is in excellent agreement with the temperature derived from the balmer continuum that amounts to @xmath89 k , alternatively @xmath30(o ) amounts to @xmath90 k. notice that the @xmath80 test requires a higher density for a lower temperature , increasing the dependence on the temperature of the he@xmath33/h@xmath33 ratio .
as mentioned above , the values in table [ chi ] correspond to the case where @xmath79 equals zero , for higher values of @xmath79 the @xmath80 values increase . in table [ chi ] we present he@xmath33/h@xmath33 values for a set of temperatures and densities . from table [ chi ]
we obtain that @xmath18(he ) @xmath91 k and @xmath19(he ) @xmath92 @xmath20 , which correspond to he@xmath33/h@xmath93 . from the surface of @xmath18 , @xmath19 and he@xmath33/h@xmath33
values defined by the condition @xmath94 the one-@xmath25 errors presented in table [ tandda ] were computed . in table [ regionhe ] we present the he@xmath33/h@xmath33 values for all the observed regions without underlying stellar absorption .
we have adopted the @xmath19(he ) @xmath43 and @xmath18(he ) @xmath43 values of region a to determine the he@xmath33/h@xmath33 values for all the other regions in table [ regionhe ] because their observational errors are higher than for region a and do not permit to obtain @xmath19(he ) and @xmath18(he ) self - consistently for each region .
in table [ totabun ] we present the total abundances of region a for @xmath44(o ) = @xmath44(o ) = 0.000 and @xmath71 .
the gaseous abundances were obtained from the following equations @xcite @xmath95 the abundances derived for all the other regions based on these equations agree within the errors with those derived from region a indicating the reliability of the equations . to obtain the total o abundances we assumed a correction of 0.04 dex due to the fraction of o tied up in dust grains ,
this fraction was estimated from the si / o values derived for the orion nebula @xcite and for h regions in irregular galaxies @xcite . to obtain the total s abundance we adopted the following equation @xmath96 from table [ regionions ]
it can be seen that the higher the o ionization degree the lower the @xmath97 / n({\rm h^+})$ ] ratio indicating , as expected , the increase of s@xmath72 with the o ionization degree . therefore to obtain the s abundance we decided to take regions 11 and 12 as representative of ngc 346 since they are expected to have the smallest @xmath98(s ) values .
therefore from the @xmath98(s ) values computed by @xcite , the data for regions 11 and 12 in table [ regionions ] , and equation ( [ sulphur ] ) we derived the value presented in table [ totabun ] . to obtain the total ar abundance we adopted the following equation @xmath99 where @xmath98(ar ) includes the ar@xmath33/h@xmath33 contribution and according to @xcite can be approximated by @xmath100 based on the @xmath18 values presented in table [ tandda ] we also include in table [ totabun ] the total abundances for @xmath101 , our preferred @xmath44 value , following the procedure outlined by @xcite .
the relevant equations are : @xmath102_{\lambda/\lambda'}\end{aligned}\ ] ] where @xmath103 is the energy difference between the ground and the excited levels , @xmath12 is the wavelength of the nebular line ( 5007 for o ) , @xmath104 is the wavelength of the auroral line ( 4363 for o ) , @xmath105 is the @xmath106 value in the region where the x@xmath107 ion is present and is given by @xmath108 @xmath109 is the representative temperature of the nebular line which can be obtained from @xmath110\frac{t^2({\rm x}^{i+})}{2}\right);\end{aligned}\ ] ] and [ @xmath32(x@xmath107)/@xmath32(h@xmath33)]@xmath111 is the abundance derived from @xmath112 ( @xmath30(o ) for o ) .
it is possible to derive the abundances for other @xmath44 values by interpolating or extrapolating the abundances presented in table [ totabun ] . in table
[ totabun ] we compare the abundances of ngc 346 with an average of the h regions in the smc derived by @xcite and those of the sun and m17 .
notice that the abundances derived by dufour should be compared with those for @xmath71 , and that no correction for the fraction of o tied up in dust grains was included by dufour .
alternatively the m17 values are for @xmath113 and 0.08 dex has been added to the o abundance to correct for the fraction of this element tied up in dust grains , consequently the m17 values can be compared directly with the ngc 346 values for @xmath114 .
the total he / h value is given by : @xmath115 the he@xmath34/h@xmath33 ratio can be obtained directly from the 4686/h@xmath10 intensity ratio . in objects of low degree of ionization
the presence of neutral helium inside the h region is important and @xmath98(he ) becomes larger than 1 .
the @xmath98(he ) can be estimated by observing a given nebula at different lines of sight since he@xmath116 is expected to be located in the outer regions .
another way to deal with this problem is to observe h regions of high degree of ionization where the he@xmath116 amount is expected to be negligible .
* see also @xcite ) defined a radiation softness parameter given by @xmath117 for large values of @xmath118 the amount of neutral helium is significant , while for low values of @xmath118 it is negligible , where the critical value is around 8 . in table
[ regionhe ] we present the @xmath118 values for all the observed regions , the @xmath118 values indicate that the amount of he@xmath116 inside the h@xmath33 region is negligible .
on the other hand , for ionization bounded objects of very high degree of ionization the amount of h@xmath116 inside the he@xmath33 strmgren sphere becomes significant and the @xmath98(he ) can become smaller than 1 .
this possibility was firstly mentioned by @xcite and studied extensively by @xcite , @xcite , and @xcite . to study this possibility
we have estimated the @xmath98(he ) from three different methods .
( 1 ) we have divided the regions in table [ regionhe ] in three groups : h ( regions 5 , 17 , and 18 ) , i ( regions 1 , 2 , 3 , and 13 ) and l ( regions 11 and 12 ) , where h , i , and l stand for high , intermediate and low degree of ionization regions ; within the errors the three groups yield the same he / h ratio , indicating that the @xmath98(he ) is very close to unity . ( 2 ) @xcite have defined the following cutoff @xmath119\lambda5007/{\rm h}\beta)_{\rm cutoff}= ( 1.139 \pm 0.306 ) + ( 2.5 \pm 0.4 ) { \rm o / h } \times 10 ^ 4.\ ] ] for observed values higher than the cutoff @xmath98(he ) is very close to unity ; from table [ totabun ] , and without considering the fraction of o embedded in dust grains , we have that @xmath120(o)/@xmath32(h)@xmath121 and , consequently , a cutoff of @xmath122 .
alternatively the observed value is @xmath123 .
( 3 ) @xcite also find that for [ o ] @xmath124/[o ] @xmath125 , the @xmath98(he ) becomes very close to unity ; from table [ int - a ] it is found that @xmath126 .
( it should be noted that the [ o ] lines present in figure [ red ] are blends of telluric and nebular lines even though we subtracted the sky contribution .
unfortunatelly the [ o ] line intensities in the sky varied in a scale of minutes leaving a telluric remnant present , which becomes apparent because the centroids of the [ o ] lines are blue shifted by about 3 from the centroid of the other lines . ) from these three methods we conclude that the amount of h@xmath116 inside the he@xmath33 strmgren sphere is negligible and in what follows we will adopt an @xmath98(he ) = 1.000 . in table
[ ysmc ] we present the helium abundance by mass @xmath1(smc ) , derived from region a. the @xmath1(smc ) values were derived from table [ chi ] , the he@xmath34/h@xmath33 value ( that amounts to @xmath127 for the @xmath18 range present in table [ ysmc ] ) , and the @xmath128 values presented in table [ composition ] .
the @xmath1(smc ) error presented in table [ composition ] is based only on the results of the self consistent method ; by considering the @xmath18(bac ) measurement the @xmath1(smc ) error diminishes from 0.0018 to 0.0013 ( and @xmath1(smc ) diminishes by 0.0001 ) . in table
[ composition ] we present the helium , heavy elements , and oxygen abundance by mass of ngc 346 .
the estimated error for @xmath3 amounts to 0.06 dex .
the @xmath128 value was obtained by assuming that @xmath3 comprises 54% of the heavy elements by mass , this fraction was estimated by @xcite from a group of ten irregular galaxies that includes the smc .
the estimated error for the @xmath128 value amounts to 0.08 dex .
to determine the @xmath7 value from the smc it is necessary to estimate the fraction of helium present in the interstellar medium produced by galactic chemical evolution .
we will assume that @xmath129 in a recent review @xcite derive @xmath7 for the smc from a similar discussion but using @xmath128 instead of @xmath3 in equation ( [ deltao ] ) ; since the error in the @xmath3 determination is smaller than in the @xmath128 determination ( see table [ composition ] ) it is better to use @xmath3 to determine @xmath7 . to estimate @xmath130 we will consider three observational determinations and a few determinations predicted by chemical evolution models .
@xcite and @xcite found that @xmath131 and @xmath132 for the galactic h region m17 , where we have added 0.08 dex to the oxygen gaseous abundance to take into account the fraction of these elements embedded in dust grains @xcite . by comparing the @xmath1 and @xmath3 values of m17 with those of region a we obtain @xmath133 .
m17 is the best h region to determine the helium abundance because among the brightest galactic h regions it is the one with the highest degree of ionization and consequently with the smallest correction for the presence of he@xmath116 ( i.e. @xmath98(he ) is very close to unity e. g. * ? ? ?
* ; * ? ? ?
it can be argued that the m17 @xmath130 value is not representative of irregular galaxies , because the yields are heavy element dependent and the @xmath3 value is considerably higher than that of the smc . from a group of 10 irregular and blue compact galaxies , that includes the lmc and the smc , @xcite found @xmath134 , where they added 0.2 dex to the o / h abundance ratios derived from the nebular data to take into account the temperature structure of the h regions and the fraction of o embedded in dust ; moreover they also estimated that @xmath3 constitutes 54% of the @xmath128 value .
@xcite from a group of 45 supergiant h regions of low metalicity derived a @xmath135 ; we find from their data that @xmath136 by adding 0.2 dex to the o abundances to take into account the temperature structure of the h regions and the fraction of o embedded in dust ; furthermore from their data we also find that @xmath137 by assuming that @xmath3 constitutes 54% of the @xmath128 value .
based on their two - infall model for the chemical evolution of the galaxy @xcite find @xmath138 for the solar vicinity .
@xcite computed chemical evolution models for the galactic disk , under an inside - out formation scenario , based on different combinations of seven sets of stellar yields by different authors ; the @xmath130 spread predicted by her models is in the 2.9 to 4.6 range for the galactocentric distance of m17 ( 5.9 kpc ) .
@xcite , based on yields by @xcite , computed closed box models adequate for irregular galaxies , like the smc , and obtained @xmath139 .
they also computed models with galactic outflows of well mixed material , that yielded @xmath130 values similar to those of the closed box models , and models with galactic outflows of o - rich material that yielded values higher than 2.95 .
the maximum @xmath130 value that can be obtained with models of o - rich outflows , without entering into contradiction with the c / o and @xmath140 observational constraints , amounts to 3.5 .
@xcite , based on yields by @xcite and @xcite , computed chemical evolution models for irregular galaxies also , like the smc , and found very similar values for closed box models with bursting star formation and constant star formation rates that amounted to @xmath141 .
the models with o - rich outflows can increase @xmath142 , but they predict higher c / o ratios than observed . from the previous discussion it follows that @xmath143 is a representative value for models and observations of irregular galaxies .
the @xmath7 values in table [ ypsmc ] were computed by adopting @xmath144 .
the differences between tables [ ysmc ] and [ ypsmc ] depend on @xmath18 because the lower the @xmath18 value the higher the @xmath3 value for the smc . in figure [ eta ]
we present our @xmath7 value as well as the theoretical @xmath7 value derived from big bang nucleosynthesis computations by @xcite for three neutrino species as a function of @xmath145 , the baryon to photon ratio .
the @xmath7 value derived by us is significantly smaller than the value derived by @xcite from the @xmath1 versus o / h linear regression for a sample of 45 bcgs , and by @xcite from the average for the two most metal deficient galaxies known ( i zw 18 and sbs 0335052 ) , that amount to @xmath146 and @xmath147 respectively ( see figure [ eta ] )
. the difference could be due to systematic effects in the abundance determinations .
there are two systematic effects not considered by izotov and collaborators that we did take into account , the possible presence of h@xmath116 inside the he@xmath33 region and the use of a lower temperature than that provided by the [ o ] lines .
we consider the first effect to be a minor one and the second to be a mayor one but both should be estimated for each object . for further discussion of the first effect see the papers by @xcite and @xcite .
the second effect was first mentioned by ( * ? ? ?
* see also @xcite ) . from constant density chemically homogeneous models computed with cloudy @xcite
we estimate that the maximum temperature that should be used to determine the helium abundance should be 5% smaller than @xmath18(o ) . moreover ,
if in addition to photoionization there is additional energy injected to the h region @xmath18(he ) should be even smaller .
@xcite produced a detailed photoionized model of ngc 2363 .
for the slit used by @xcite they find an @xmath98(he ) = 0.993 ; moreover they also find that the @xmath18(o ) predicted by the model is considerably smaller than observed . from the data of @xcite for ngc 2363 , adopting a @xmath18(he ) 10% smaller than @xmath18(o ) and @xmath148 we find that @xmath149 .
similarly , @xcite produced a detailed model of i zw 18 and find that the photoionized model predicts a @xmath18(o ) value 15% smaller than observed , on the other hand their model predicts an @xmath98(he ) = 1.00 . from the observations of @xmath0 5876 and 6678 by @xcite of i zw 18 , and adopting
a @xmath18(he ) 10% smaller than @xmath18(o ) we obtain @xmath150 ; for a @xmath18(he ) 15% smaller than @xmath18(o ) we obtain @xmath151 , both results in good agreement with our determination based on the smc .
further discussion of these issues is presented elsewhere @xcite .
the primordial helium abundance by mass of @xmath152 based on the smc combined with the big bang nucleosynthesis computations by @xcite for three light neutrino species implies that , at the one-@xmath25 confidence level , @xmath153 is in the 0.0065 to 0.0091 range .
for @xmath154 the @xmath7 value corresponds to @xmath155 , a value considerably smaller than that derived from the pregalactic deuterium abundance , @xmath156 , determined by @xcite that corresponds to @xmath157 for @xmath154 .
our @xmath158 value is in very good agreement with the low redshift estimate of the global budget of baryons by @xcite who find @xmath159 for @xmath154 and is consistent with their minimum to maximum range for redshift @xmath160 that amounts to @xmath161 .
the discrepancy between @xmath7 and @xmath156 needs to be studied further .
in addition to the relevance for cosmology an accurate @xmath7 value permits to determine @xmath162 , ratio that provides a strong constraint for the models of chemical evolution of galaxies ( see for example * ? ? ?
* ; * ? ? ?
* ) . to increase the accuracy of the @xmath7 determinations we need observations of very high quality of as many he lines as possible to derive @xmath18(he ) , @xmath19(he ) , @xmath79 , and @xmath32(he@xmath33)/@xmath32(h@xmath33 ) self - consistently .
we also need observations with high spatial resolution to estimate the @xmath98(he ) along different lines of sight .
we are grateful to s. torres - peimbert for her participation on the observations presented in this paper .
it is a pleasure to acknowledge several fruitful discussions on this subject with : l. carigi , v. luridiana , b. e. j. pagel , m. pea , e. skillman , g. steigman , s. torres - peimbert , and s. viegas .
we are also grateful to the referee for some excellent suggestions .
mp received partial support from conacyt grant 25451-e , mtr received partial support from a ctedra presidencial and fondecyt grant 1980659 .
l [email protected] lr [email protected] clll 1 & 102&2@xmath163 w & 0.0@xmath163 & & 34&1@xmath163 & 6.23 - 14 + 2 & 58&0 w & 0.0 & & 22&4 & 1.14 - 13 + 3 & 24&9 w & 0.0 & & 13&6 & 1.63 - 13 + 4 & 2&1 e & 0.0 & & 18&6 & 2.37 - 13 & 355 & o3v((f@xmath164 ) ) + 5 & 86&4 e & 0.0 & & 43&6 & 2.54 - 13 + + 11 & 123&9 w & 5.5 s & & 24&8 & 4.33 - 14 + 12 & 77&6 w & 5.5 s & & 32&9 & 9.48 - 14 + 13 & 42&5 w & 5.5 s & & 25&8 & 2.52 - 13 + 14 & 15&5 w & 5.5 s & & 5&6 & 4.51 - 14 & 324 & o4v((f ) ) + 15 & 7&8 e & 5.5 s & & 9&0 & 1.08 - 13 & 396 & o7v + 16 & 22&6 e & 5.5 s & & 9&3 & 1.39 - 13 & 470 + 476 & o8-o8.5iii+ ? + 17 & 66&9 e & 5.5 s & & 29&2 & 2.22 - 13 + 18 & 92&4 e & 5.5 s & & 13&1 & 5.70 - 14 + + a & .&. & & & .&. & 11.6 - 13 + b & .&. & & & .&. & 17.9 - 13 + 3614 & he & @xmath1652.411 & @xmath1652.361 @xmath166 0.040 & ... + 3634 & he & @xmath1652.475 & @xmath1652.425 @xmath166 0.040 & ... + 3726 + 3729 & [ o ] & @xmath1650.049 & @xmath1650.003 @xmath166 0.005 & + 0.058 @xmath166 0.005 + 3750 & h12 & @xmath1651.546 & @xmath1651.501 @xmath166 0.010 & ... + 3771 & h11 & @xmath1651.421 & @xmath1651.376 @xmath166 0.010 & ... + 3798 & h10 & @xmath1651.307 & @xmath1651.264 @xmath166 0.010 & ... + 3835 & h9 & @xmath1651.176 & @xmath1651.134 @xmath166 0.007 & @xmath1651.677 @xmath166 0.015 + 3869 & [ ne ] & @xmath1650.404 & @xmath1650.364 @xmath166 0.003 & @xmath1650.386 @xmath166 0.003 + 3889 + 3889 & he + h8 & @xmath1650.740 & @xmath1650.700 @xmath166 0.004 & @xmath1650.915 @xmath166 0.006 + 3889 & he & ... & @xmath1651.027 @xmath166 0.008 & ... + + 3967 + 3970 & [ ne ] + h7 & @xmath1650.568 & @xmath1650.533 @xmath166 0.005 & @xmath1650.719 @xmath166 0.006 + 4026 & he & @xmath1651.767 & @xmath1651.733 @xmath166 0.015 & ... + 4069 + 4076 & [ s ] & @xmath1652.060 & @xmath1652.029 @xmath166 0.020 & ... + 4102 & h@xmath167 & @xmath1650.621 & @xmath1650.591 @xmath166 0.003 & @xmath1650.702 @xmath166 0.005 + 4340 & h@xmath168 & @xmath1650.342 & @xmath1650.322 @xmath166 0.003 & @xmath1650.400 @xmath166 0.004 + 4363 & [ o ] & @xmath1651.139 & @xmath1651.119 @xmath166 0.005 & @xmath1651.175 @xmath166 0.006 + 4388 & he & @xmath1652.351 & @xmath1652.332 @xmath166 0.020 & ... + 4471 & he & @xmath1651.432 & @xmath1651.416 @xmath166 0.006 & @xmath1651.657 @xmath166 0.008 + 4658 & [ fe ] & @xmath1652.688 & @xmath1652.681 @xmath166 0.050 & ... + 4686 & he & @xmath1652.581 & @xmath1652.574 @xmath166 0.040 & ...
+ + 4711 + 4713 & [ ar ] + he & @xmath1651.856 & @xmath1651.851 @xmath166 0.015 & @xmath1652.028 @xmath166 0.020 + 4711 & [ ar ] & ... & @xmath1652.029 @xmath166 0.025 & ... + 4740 & [ ar ] & @xmath1652.242 & @xmath1652.238 @xmath166 0.025 & @xmath1652.426 @xmath166 0.030 + 4861 & h@xmath10 & 0.000 & 0.000 @xmath166 0.000 & @xmath1650.023 @xmath166 0.003 + 4922 & he & @xmath1651.998 & @xmath1652.000 @xmath166 0.008 & @xmath1652.310 @xmath166 0.030 + 4959 & [ o ] & + 0.267 & + 0.264 @xmath166 0.002 & + 0.238 @xmath166 0.003 + 5007 & [ o ] & + 0.739 & + 0.735 @xmath166 0.002 & + 0.707 @xmath166 0.003 + 5876 & he & @xmath1650.942 & @xmath1650.973 @xmath166 0.005 & @xmath1651.022 @xmath166 0.006 + 6300 & [ o ] & @xmath1692.016 & @xmath1692.060 & @xmath1692.083 + 6312 & [ s ] & @xmath1651.727 & @xmath1651.771 @xmath166 0.010 & @xmath1651.870 @xmath166 0.012 + + 6563 & h@xmath28 & + 0.501 & + 0.451 @xmath166 0.002 & + 0.451 @xmath166 0.000 + 6584 &
[ n ] & @xmath1651.371 & @xmath1651.422 @xmath166 0.005 & @xmath1651.365 @xmath166 0.005 + 6678 & he & @xmath1651.474 & @xmath1651.528 @xmath166 0.003 & @xmath1651.582 @xmath166 0.005 + 6716 & [ s ] & @xmath1651.094 & @xmath1651.149 @xmath166 0.003 & @xmath1651.142 @xmath166 0.004 + 6731 & [ s ] & @xmath1651.237 & @xmath1651.293 @xmath166 0.004 & @xmath1651.274 @xmath166 0.004 + 7065 & he & @xmath1651.615 & @xmath1651.675 @xmath166 0.004 & @xmath1651.696 @xmath166 0.005 + 7136 & [ ar ] & @xmath1651.065 & @xmath1651.127 @xmath166 0.003 & @xmath1651.129 @xmath166 0.003 + 7281 & he & @xmath1652.135 & @xmath1652.199 @xmath166 0.020 & @xmath1652.295 @xmath166 0.025 + 7320 + 7330 & [ o ] & @xmath1651.549 & @xmath1651.614 @xmath166 0.005 & @xmath1651.511 @xmath166 0.007 + l [email protected] [email protected] [email protected] [email protected] [email protected] 3726 + 3729 & @xmath1700&225 @xmath166 0.008 & @xmath1700&215 @xmath166
0.007 & @xmath1650&246 @xmath166 0.009 & @xmath1650&042
@xmath166 0.007 & @xmath1650&225 @xmath166 0.007 + 3835 & @xmath1651&203 @xmath166
0.040 & @xmath1651&125 @xmath166 0.025 & @xmath1651&127 @xmath166 0.020 & .&. & @xmath1651&127 @xmath166 0.015 + 3869 & @xmath1650&472 @xmath166 0.012 & @xmath1650&419 @xmath166 0.008 & @xmath1650&403 @xmath166 0.006 & @xmath1650&394 @xmath166
0.005 & @xmath1650&323 @xmath166 0.005 + 3889 + 3889 & @xmath1650&776 @xmath166 0.020 & @xmath1650&702 @xmath166 0.012 & @xmath1650&695 @xmath166 0.009 & @xmath1651&154 @xmath166 0.015 & @xmath1650&705 @xmath166 0.007 + 3889 & .&. &
@xmath1651&031 @xmath166 0.025 & @xmath1651&017 @xmath166 0.015 & .&. & @xmath1651&038 @xmath166 0.012 + 3967 + 3970 & @xmath1650&695 @xmath166 0.020 & @xmath1650&574
@xmath166 0.010 & @xmath1650&567 @xmath166 0.008 & @xmath1650&886 @xmath166 0.008 & @xmath1650&552 @xmath166 0.006 + 4026 & .&. & .&. &
@xmath1651&735 @xmath166 0.040 & .&. & .&. + 4102 & @xmath1650&656 @xmath166 0.015 & @xmath1650&590 @xmath166 0.010 & @xmath1650&577 @xmath166 0.008 & @xmath1650&844 @xmath166 0.008 & @xmath1650&594 @xmath166 0.006 + 4340 & @xmath1650&354 @xmath166 0.008 & @xmath1650&319 @xmath166 0.004 & @xmath1650&323
@xmath166 0.003 & @xmath1650&434 @xmath166 0.004 & @xmath1650&320 @xmath166
0.004 + 4363 & @xmath1651&155 @xmath166 0.020 & @xmath1651&152
@xmath166 0.015 & @xmath1651&172 @xmath166 0.012 & @xmath1651&171
@xmath166 0.010 & @xmath1651&032 @xmath166 0.007 + 4471 & @xmath1651&384 @xmath166 0.030 & @xmath1651&404 @xmath166 0.020 & @xmath1651&409
@xmath166 0.015 & @xmath1651&469 @xmath166 0.012 & @xmath1651&422 @xmath166
0.015 + 4686 & @xmath1692&230 & @xmath1692&700 & @xmath1692&873 & @xmath1691&974 & @xmath1652&199 @xmath166 0.040 + 4711 + 4713 & .&. & @xmath1652&275 @xmath166 0.070 & @xmath1651&950 @xmath166 0.035 & .&. & @xmath1651&697 @xmath166 0.020 + 4711 & .&. & .&. & @xmath1652&188 @xmath166 0.050 & .&. & @xmath1651&814 @xmath166 0.030 + 4740 & .&. & .&. & @xmath1652&498 @xmath166 0.080 & .&. & @xmath1652&035 @xmath166 0.035 + 4861 & 0&000 @xmath166 0.000 & 0&000 @xmath166 0.000 & 0&000 @xmath166 0.000 & @xmath1650&045 @xmath166 0.003 & 0&000 @xmath166 0.000 + 4922 & @xmath1652&019 @xmath166
0.035 & @xmath1652&004 @xmath166 0.025 & @xmath1651&968 @xmath166 0.020 & .&. & @xmath1652&005 @xmath166 0.015 + 4959 & @xmath1700&173 @xmath166 0.004 & @xmath1700&209
@xmath166 0.003 & @xmath1700&264 @xmath166 0.002 & @xmath1700&232
@xmath166 0.003 & @xmath1700&311 @xmath166 0.002 + 5007 & @xmath1700&646
@xmath166 0.004 & @xmath1700&679 @xmath166 0.003 & @xmath1700&736 @xmath166 0.002 & @xmath1700&704 @xmath166 0.003 & @xmath1700&783 @xmath166 0.002 + 5876 & .&. &
@xmath1650&998 @xmath166 0.020 & @xmath1650&962 @xmath166 0.009 & @xmath1651&029 @xmath166 0.007 & @xmath1650&970 @xmath166 0.007 + 6312 & @xmath1651&736 @xmath166 0.050 & @xmath1651&694 @xmath166 0.030 & @xmath1651&823 @xmath166 0.025 & @xmath1651&997 @xmath166 0.030 & @xmath1651&970 @xmath166 0.030 + 6563 & @xmath1700&453 @xmath166 0.004 & @xmath1700&453 @xmath166 0.003 & @xmath1700&453 @xmath166 0.002 & @xmath1700&451 @xmath166 0.000 & @xmath1700&447 @xmath166
0.002 + 6584 & @xmath1651&227 @xmath166 0.015 & @xmath1651&211 @xmath166 0.010 & @xmath1651&705 @xmath166 0.015 & @xmath1651&485 @xmath166 0.007 & @xmath1651&574 @xmath166
0.008 + 6678 & @xmath1651&537 @xmath166 0.015 & @xmath1651&509
@xmath166 0.012 & @xmath1651&532 @xmath166 0.009 & @xmath1651&576 @xmath166 0.007 & @xmath1651&537 @xmath166
0.006 + 6716 & @xmath1650&961 @xmath166 0.008 & @xmath1650&966 @xmath166
0.006 & @xmath1651&526 @xmath166 0.009 & @xmath1651&274
@xmath166 0.005 & @xmath1651&177 @xmath166 0.005 + 6731 & @xmath1651&108 @xmath166 0.010 & @xmath1651&110
@xmath166 0.007 & @xmath1651&667 @xmath166 0.012 & @xmath1651&430
@xmath166 0.006 & @xmath1651&319 @xmath166 0.005 + 7065 & @xmath1651&645 @xmath166 0.020 & @xmath1651&643
@xmath166 0.015 & @xmath1651&679 @xmath166 0.012 & @xmath1651&688
@xmath166 0.009 & @xmath1651&673 @xmath166 0.008 + 7136 & @xmath1651&152
@xmath166 0.010 & @xmath1651&090 @xmath166 0.007 & @xmath1651&115
@xmath166 0.005 & @xmath1651&160 @xmath166 0.004 & @xmath1651&195 @xmath166 0.005
+ 7320 + 7330 & @xmath1651&410 @xmath166 0.020 & @xmath1651&444 @xmath166 0.015 & @xmath1651&798
@xmath166 0.015 & @xmath1651&580 @xmath166 0.010 & @xmath1651&873 @xmath166 0.015 + l [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected]
3726 + 3729 & @xmath1700&129 @xmath166 0.010 & @xmath1700&231 @xmath166 0.007 & @xmath1700&167 @xmath166 0.006 & @xmath1650&115 @xmath166 0.020 & @xmath1700&051 @xmath166 0.012 & @xmath1700&328 @xmath166 0.008 & @xmath1650&224 @xmath166 0.008 & @xmath1650&430 @xmath166
0.020 + 3835 & @xmath1651&174 @xmath166 0.040 & @xmath1651&155 @xmath166 0.030 & @xmath1651&163 @xmath166 0.015 & .&. & .&. & .&. &
@xmath1651&155 @xmath166 0.015 & @xmath1651&116 @xmath166 0.040 + 3869 & @xmath1650&488 @xmath166 0.012 & @xmath1650&453 @xmath166 0.008 & @xmath1650&422 @xmath166 0.005 & @xmath1650&354 @xmath166 0.020 & @xmath1650&494 @xmath166 0.012 & @xmath1650&485
@xmath166 0.012 & @xmath1650&326 @xmath166 0.005 & @xmath1650&319 @xmath166 0.012
+ 3889 + 3889 & @xmath1650&718 @xmath166 0.015 & @xmath1650&707 @xmath166 0.012 &
@xmath1650&704 @xmath166 0.007 & .&. & .&. & .&. & @xmath1650&688 @xmath166 0.010 & @xmath1650&698 @xmath166 0.020 + 3889 & @xmath1651&069 @xmath166 0.030 & @xmath1651&043 @xmath166 0.020 & @xmath1651&037 @xmath166 0.012 & .&. & .&. & .&. & @xmath1651&002 @xmath166 0.012 & @xmath1651&024 @xmath166 0.035 + 3967 + 3970 & @xmath1650&609 @xmath166 0.015 & @xmath1650&555 @xmath166 0.010 & @xmath1650&573 @xmath166 0.006 & .&.
& .&. & .&. & @xmath1650&539 @xmath166 0.006 & @xmath1650&537 @xmath166 0.015 + 4069 + 4076 & .&. & @xmath1651&930 @xmath166 0.080 & @xmath1651&988 @xmath166 0.040 & .&. & .&. & .&. & @xmath1652&131 @xmath166 0.060 & .&. + 4102 & @xmath1650&597 @xmath166 0.015 & @xmath1650&593 @xmath166 0.012 & @xmath1650&595 @xmath166 0.006 & .&. & @xmath1650&982 @xmath166 0.025 & @xmath1650&953 @xmath166 0.025 & @xmath1650&589 @xmath166 0.007 & @xmath1650&583 @xmath166 0.015 + 4340 & @xmath1650&323 @xmath166 0.008 & @xmath1650&332 @xmath166 0.005 & @xmath1650&329 @xmath166 0.003 & @xmath1650&509 @xmath166 0.015 & @xmath1650&496
@xmath166 0.009 & @xmath1650&405 @xmath166 0.007 & @xmath1650&328 @xmath166 0.003 & @xmath1650&325 @xmath166 0.006 + 4363 & @xmath1651&259 @xmath166 0.025 & @xmath1651&264 @xmath166 0.015 & @xmath1651&237 @xmath166 0.010 & @xmath1651&141 @xmath166 0.040 & @xmath1651&250 @xmath166 0.025 & @xmath1651&272 @xmath166 0.025 & @xmath1651&085 @xmath166 0.010 & @xmath1651&022 @xmath166 0.020 + 4471 & .&. & @xmath1651&410 @xmath166 0.020 & @xmath1651&419 @xmath166 0.012 & .&. & .&. & .&. & @xmath1651&415 @xmath166 0.015 & @xmath1651&428 @xmath166
0.035 + 4686 & @xmath1692&181 & @xmath1692&662 & @xmath1692&815 & .&. & .&. & .&. & @xmath1692&433 & @xmath1692&609 + 4711 + 4713 & .&. & .&.
& @xmath1652&152 @xmath166 0.035 & .&. & .&. & .&. & @xmath1651&847 @xmath166 0.025 & @xmath1651&694 @xmath166 0.050 + 4711 & .&. & .&. & @xmath1652&637 @xmath166 0.050 & .&. & .&. & .&. & @xmath1652&023 @xmath166 0.035 & @xmath1651&810 @xmath166 0.060 + 4740 & .&. & .&. & .&. & .&. & .&. & .&. & @xmath1652&239 @xmath166 0.040 & @xmath1651&965 @xmath166 0.070 + 4861 & 0&000 @xmath166 0.000 & 0&000 @xmath166 0.000 & 0&000 @xmath166 0.000 & @xmath1650&077 @xmath166 0.008 & @xmath1650&071 @xmath166 0.005 & @xmath1650&025 @xmath166 0.004 & 0&000 @xmath166 0.000 & 0&000 @xmath166 0.000 + 4922 & .&. & @xmath1651&995 @xmath166 0.025 & @xmath1651&982 @xmath166 0.012 & .&. & .&. & .&. & @xmath1652&023 @xmath166 0.020 & .&. + 4959 &
@xmath1700&205 @xmath166 0.004 & @xmath1700&188 @xmath166 0.003 & @xmath1700&210 @xmath166 0.002 & @xmath1700&282 @xmath166 0.008 & @xmath1700&151 @xmath166 0.005 & @xmath1700&135 @xmath166 0.004 & @xmath1700&294 @xmath166 0.002 & @xmath1700&332 @xmath166
0.004 + 5007 & @xmath1700&672 @xmath166 0.004 & @xmath1700&660 @xmath166 0.003 & @xmath1700&683 @xmath166 0.002 & @xmath1700&757 @xmath166 0.008 & @xmath1700&628 @xmath166 0.005 & @xmath1700&597 @xmath166 0.004 & @xmath1700&767 @xmath166 0.002 & @xmath1700&804 @xmath166
0.004 + 5876 & @xmath1650&977 @xmath166 0.020 & @xmath1650&988 @xmath166 0.012 & @xmath1650&987 @xmath166 0.008 & @xmath1651&074 @xmath166 0.040 & @xmath1651&047 @xmath166 0.020 & @xmath1651&205 @xmath166 0.020 & @xmath1650&983 @xmath166 0.008 & @xmath1650&981 @xmath166
0.015 + 6312 & @xmath1651&623 @xmath166 0.040 & @xmath1651&672 @xmath166 0.035 & @xmath1651&761 @xmath166 0.015 & @xmath1651&953 @xmath166 0.100 & @xmath1651&931 @xmath166
0.060 & @xmath1651&961 @xmath166 0.070 & @xmath1651&869 @xmath166 0.025 & @xmath1651&980 @xmath166 0.040 + 6563 & @xmath1700&445 @xmath166 0.004 & @xmath1700&446 @xmath166 0.003 & @xmath1700&446 @xmath166 0.002 & @xmath1700&451 @xmath166 0.000 & @xmath1700&451 @xmath166 0.000 & @xmath1700&451
@xmath166 0.000 & @xmath1700&452 @xmath166 0.002 & @xmath1700&443 @xmath166
0.004 + 6584 & @xmath1651&381 @xmath166 0.020 & @xmath1651&228 @xmath166 0.012 & @xmath1651&273
@xmath166 0.008 & @xmath1651&546 @xmath166 0.040 & @xmath1651&379 @xmath166 0.020 & @xmath1651&137 @xmath166 0.012 & @xmath1651&732 @xmath166 0.015 & @xmath1651&769 @xmath166 0.025 + 6678 &
@xmath1651&546 @xmath166 0.020 & @xmath1651&542 @xmath166 0.012 & @xmath1651&532 @xmath166 0.007 & @xmath1651&672 @xmath166 0.040 & @xmath1651&600 @xmath166 0.020 & @xmath1651&719 @xmath166 0.020 & @xmath1651&526 @xmath166 0.008 & @xmath1651&546 @xmath166
0.020 + 6716 & @xmath1651&118 @xmath166 0.012 & @xmath1650&966 @xmath166 0.006 & @xmath1651&041 @xmath166 0.004 & @xmath1651&396 @xmath166 0.030 & @xmath1651&140 @xmath166 0.012 & @xmath1650&985 @xmath166 0.008 & @xmath1651&246 @xmath166 0.005 & @xmath1651&512 @xmath166
0.020 + 6731 & @xmath1651&306 @xmath166 0.015 & @xmath1651&124 @xmath166 0.007 & @xmath1651&184 @xmath166 0.005 & @xmath1651&548 @xmath166 0.035 & @xmath1651&251 @xmath166 0.015 & @xmath1651&075 @xmath166 0.009 & @xmath1651&389 @xmath166 0.006 & @xmath1651&677 @xmath166
0.020 + 7065 & @xmath1651&627 @xmath166 0.025 & @xmath1651&690 @xmath166 0.015 & @xmath1651&687 @xmath166 0.010 & @xmath1651&815 @xmath166 0.050 & @xmath1651&759 @xmath166 0.025 & @xmath1651&907 @xmath166 0.030 & @xmath1651&669 @xmath166 0.010 & @xmath1651&679 @xmath166 0.020 + 7136 & @xmath1651&097
@xmath166 0.012 & @xmath1651&098 @xmath166 0.007 & @xmath1651&096
@xmath166 0.004 & @xmath1651&168 @xmath166 0.020 & @xmath1651&174 @xmath166 0.012 & @xmath1651&175 @xmath166 0.010 & @xmath1651&145 @xmath166 0.005 & @xmath1651&202 @xmath166 0.012 + 7320 + 7330 & @xmath1651&585 @xmath166 0.025 & @xmath1651&385 @xmath166 0.015 & @xmath1651&406 @xmath166 0.012 & @xmath1651&645 @xmath166 0.040 & @xmath1651&486 @xmath166 0.020 & @xmath1651&189 @xmath166 0.020 & @xmath1651&864 @xmath166 0.015 & @xmath1651&945 @xmath166 0.030 + lrrrrrrrrrr 1 & 520 & 85 & 25 & 11 & 2.2 & 2.5 & 0.9 & 18 & 7.0 & 6.5 + 2 & 720 & 190 & 75 & 38 & 10 & 5.5 & 2.0 & 28 & 9.5 & 8.5 + 3 & 1650 & 380 & 140 & 65 & 17 & 12 & 4.5 & 60 & 20 & 16 + 4 & 115 & 19 & 5.0 & 1.5 & -0.65 & 0.55 & 0.35 & 2.8 & 1.2 & 1.1 + 5 & 785 & 210 & 90 & 45 & 11 & 7.5 & 2.2 & 28 & 8.5 & 7.0 + + 11 & 580 & 140 & 60 & 30 & 7.5 & 5.0 & ... & 22 & 7.0 & 5.5 + 12 & 790 & 200 & 75 & 38 & 9.5 & 7.2 & 1.4 & 33 & 8.5 & 7.0 + 13 & 1760 & 350 & 115 & 60 & 13 & 11 & 4.0 & 65 & 21 & 16 + 14 & 55 & 7.0 & 2.0 & ... & ... & ... & ... & 1.4 & 0.50 & 0.45 + 15 & 200 & 22 & 4.8 & ... & ... & ... & ... & 4.5 & 2.0 & 1.9 + 16 & 69 & 8.0 & 2.5 & ... & ... & -0.55 & ... & 1.0 & 0.29 & 0.16 + 17 & 1950 & 340 & 120 & 55 & 14 & 10 & 3.5 & 70 & 21 & 19 + 18 & 1800 & 300 & 100 & 60 & 13 & 9.5 & 4.0 & 65 & 20 & 18 + + a & 1250 & 250 & 99 & 44 & 12.5 & 8.5 & 2.5 & 46 & 14.7 & 11.2 + b & 230 & 35 & 11.1 & 4.7 & 0.43 & 0.7 & 0.18 & 6.5 & 2.7 & 2.2 + cr@ @xmath166 rr@ @xmath166 rr@ @xmath166 r 1 & 13 730 & 250 & 11 650 & 450 & 40 & 35 + 2 & 13 350 & 190 & 11 150 & 400 & 50 & 25 + 3 & 12 500 & 130 & 12 600 & 350 & 55 & 45 + 4 & 12 840 & 130 & 13 600 & 350 & 15 & 20 + 5 & 13 540 & 100 & 10 950 & 250 & 55 & 20 + + 11 & 12 230 & 250 & 10 500 & 600 & 14 & 50 + 12 & 12 320 & 170 & 12 100 & 350 & 15 & 25 + 13 & 12 360 & 130 & 12 600 & 300 & 50 & 20 + 14 & 12 580 & 390 & 13 550 & 1150 & 25 & 150 + 15 & 12 800 & 270 & 12 550 & 550 & 150 & 60 + 16 & 12 890 & 310 & 12 550 & 600 & 220 & 45 + 17 & 13 120 & 140 & 11 100 & 300 & 55 & 20 + 18 & 13 440 & 280 & 13 500 & 950 & 14 & 65 + + a & 13 070 & 50 & 11 810 & 160 & 50 & 15 + b & 12 755 & 90 & 12 360 & 220 & 85 & 20 + cr@=lcr@=lcr@=rc & ( o ) & @xmath171 & & & & ( rms ) & @xmath172 + & ( o ) & @xmath173 & & & & ( s ) & @xmath174 + & ( o + o i ) & @xmath175 & & & & + & ( bac ) & @xmath176 & & ( bac , o + o ) & @xmath177 & & + & ( he ) @xmath64 & @xmath178 & & & & + & ( he ) @xmath43 & @xmath179 & & ( he , o + o ) & @xmath180 & & ( he ) @xmath43 & @xmath181 + & @xmath182(he ) @xmath183 & @xmath184 & & ( bac , he , o + o ) & @xmath185 & & @xmath182(he ) @xmath183 & @xmath186 + ccccccccc 1 & 7.44 & 7.79 & 5.87 & 6.95 & 5.42 & 6.11 & 5.52 & .... + 2 & 7.47 & 7.86 & 5.91 & 7.04 & 5.44 & 6.19 & 5.61 & .... + 3 & 7.10 & 7.99 & 5.45 & 7.15 & 4.93 & 6.16 & 5.64 & 4.65 + 4 & 7.27 & 7.93 & 5.67 & 7.12 & 5.15 & 5.95 & 5.57 & .... + 5 & 7.01 & 7.94 & 5.53 & 7.11 & 5.21 & 5.89 & 5.49 & 4.98 + + 11 & 7.51 & 7.96 & 5.81 & 7.10 & 5.35 & 6.40 & 5.68 & .... + 12 & 7.60 & 7.94 & 5.96 & 7.12 & 5.50 & 6.34 & 5.67 & .... + 13 & 7.53 & 7.95 & 5.91 & 7.15 & 5.42 & 6.25 & 5.66 & 4.56 + 14 & 7.22 & 8.01 & 5.62 & 7.19 & 5.05 & 6.02 & 5.58 & .... + 15 & 7.38 & 7.86 & 5.77 & 7.03 & 5.32 & 6.02 & 5.56 & .... + 16 & 7.65 & 7.82 & 6.01 & 7.03 & 5.49 & 5.98 & 5.55 & .... + 17 & 7.06 & 7.97 & 5.40 & 7.16 & 5.17 & 6.05 & 5.56 & 4.81 + 18 & 6.83 & 7.97 & 5.34 & 7.13 & 4.88 & 5.90 & 5.49 & 5.07 + + a & 7.28 & 7.94 & 5.69 & 7.13 & 5.27 & 6.15 & 5.59 & 4.81 + b & 7.41 & 7.97 & 5.81 & 7.17 & 5.33 & 6.11 & 5.63 & 4.64 + ccccccc 11200 & & 805 & 798 & 793 & 791 & 781 + & & ( 83.2 ) & ( 47.7 ) & ( 26.4 ) & ( 20.0 ) & ( 8.24 ) + + 11800 & & 806 & 799 & 793 & 790 & 780 + & & ( 38.6 ) & ( 15.9 ) & ( 7.37 ) & ( 6.59 ) & ( 20.4 ) + + 11950 & & 806 & 799 & 793 & 790 & 779 + & & ( 30.8 ) & ( 11.7 ) & ( 6.53 ) & ( 7.25 ) & ( 27.7 ) + + 12400 & & 807 & 799 & 793 & 790 & 778 + & & ( 15.0 ) & ( 7.17 ) & ( 12.5 ) & ( 17.9 ) & ( 58.6 ) + + 13000 & & 809 & 800 & 793 & 790 & 777 + & & ( 9.72 ) & ( 18.2 ) & ( 38.4 ) & ( 50.2 ) & ( 118 ) + ccccccccl 1 & ... & 838 & 739 & ... & 778 & 849 & 802 & 2.07 + 2 & 802 & 800 & 764 & 755 & 830 & 853 & 815 & 2.21 + 3 & 828 & 791 & 831 & 820 & 787 & 784 & 804 & 2.08 + 5 & 789 & 768 & 763 & 806 & 778 & 796 & 788 & 0.525 + 11 & 735 & ... & ... & 793 & 762 & 885 & 794 & 4.03 + 12 & 780 & 789 & 780 & 763 & 769 & 765 & 773 & 3.10 + 13 & 791 & 773 & 804 & 775 & 787 & 771 & 783 & 2.36 + 17 & 850 & 783 & 735 & 786 & 805 & 788 & 797 & 0.873 + 18 & 815 & 757 & ... & 785 & 763 & 785 & 780 & 0.719 + + a & 809 & 779 & 772 & 800 & 795 & 792 & 793 & 1.57 + cccccc n & 6.46 & 6.44 & 6.51 & 7.92 & 7.90 + o & 8.02 & 8.07 & 8.15 & 8.83 & 8.87 + ne & 7.22 & 7.22 & 7.30 & 8.08 & 8.02 + s & 6.49 & 6.50 & 6.59 & 7.33 & 7.31 + ar & 5.78 & 5.74 & 5.82 & 6.40 & 6.60 + ccccccc 11200 & & 0.2431 & 0.2416 & 0.2404 & 0.2399 & 0.2377 + 11800 & & 0.2435 & 0.2419 & 0.2405 & 0.2399 & 0.2375 + 11950 & & 0.2436 & 0.2420 & 0.2405 & 0.2399 & 0.2374 + 12400 & & 0.2439 & 0.2421 & 0.2406 & 0.2399 &
0.2372 + 13000 & & 0.2443 & 0.2423 & 0.2407 & 0.2400 & 0.2370 + ccr@@xmath166lr@@xmath166l @xmath1 & 0.2445 & 0.24050 & 0.00180 & 0.2797 & 0.0060 + @xmath128 & 0.00263 & 0.00315 & 0.00063 & 0.0212 & 0.0030 + @xmath3 & 0.00142 & 0.00171 & 0.00025 & 0.0083 & 0.0012 + + @xmath187 & ... & 1.9 & 0.5 & 2.13 & 0.5 + @xmath130 & ... & 3.5 & 0.9 & 5.45 & 1.1 + ccccccc 11200 & & 0.2363 & 0.2348 & 0.2336 & 0.2331 & 0.2309 + 11800 & & 0.2373 & 0.2357 & 0.2343 & 0.2337 & 0.2313 + 11950 & & 0.2376 & 0.2360 & 0.2345 & 0.2339 & 0.2314 + 12400 & & 0.2384 & 0.2366 & 0.2351 & 0.2344 & 0.2317 + 13000 & & 0.2395 & 0.2375 & 0.2359 & 0.2352 & 0.2322 + | spectrophotometry in the @xmath0 3400 - 7400 range is presented for 13 areas of the brightest h region in the smc : ngc 346 .
the observations were obtained at ctio with the 4-m telescope . based on these observations
its chemical composition is derived .
the helium and oxygen abundances by mass are given by : @xmath1(smc)@xmath2 and @xmath3(smc)@xmath4 . from models and observations of irregular and blue compact galaxies
it is found that @xmath5 and consequently that the primordial helium abundance by mass is given by : @xmath6 .
this result is compared with values derived from big bang nucleosynthesis , and with other determinations of @xmath7 . |
entangled paired photons or biphotons provide an unprecedented tools for research in fundamental physics such as quantum information processing @xcite and tests of fundamentals of quantum mechanics @xcite .
biphotons produced from either spontaneous parametric down conversion ( spdc ) @xcite or atomic cascade transitions have already been used in optical experimental tests of fundamentals of quantum theory , and have demonstrated violations of bell s inequalities @xcite .
most of experimental tests of the inequalities have involved the polarization entanglement .
rarity and tapster @xcite have reported experiments with momentum entanglement of the beams .
the experiment proposed by franson @xcite concerning a bell inequality for nonpolarization variables , relies on the entanglement of a continuous variable , energy .
there are number of experimental realizations @xcite of the franson interferometer .
et al_. @xcite have achieved sinusoidal fringes of the two - photon interference with visibility greater than @xmath1 .
such high visibility is required to confirm violation of a bell - type inequality . in this paper
, we propose a new type of two - photon beating experiment by using biphotons generated from a two - level system operating in the resonant - pumping case @xcite .
recently , the coincidence counting rate of paired photons generated from a two - level atomic system exhibits a damped rabi oscillation and photon anti - bunching - like effect @xcite .
these features are caused by the destructive interference between two types of biphoton generation [ or four - wave mixings ( fwms ) ] sideband - sideband and central - central in the
_ mollow triplet_. the two - photon state is therefore a coherent superposition of these two fwms .
we have also emphasized that the coupling coefficients in the coupled field operators equations for paired - photon generation are different from those in the classical fwm equations in a two - level system @xcite .
the two - photon wavepacket created in such a two - level system is generally a convolution of the phase matching and the third - order nonlinear susceptibility , as in the three - level @xcite and four - level @xcite double-@xmath2 systems .
a further analysis shows that these two kinds of biphotons created in a two - level system are associated with different refractive indices @xcite .
the linear optical response of the paired photons from two sidebands in the mollow triplet @xcite implies that these biphotons propagate with tunable refractive indices and vanishing absorption , while biphotons from the central component propagate with unity refraction of index .
these different refractive indices induce a relative phase difference between the twin - photon propagation through the two - level medium , and play the same role as the pathway - length difference between the short - short ( s - s ) and long - long ( l - l ) in the original franson s experiment .
it should be noted that two - photon interference disappears , if the detection system can spectrally distinguish these two fwm processes .
such a filtering would destroy the biphoton generated by the sidebands ( or central components ) . on the other hand , the proposed experiment may be an interesting way of probing the quantum nature of the detection process .
the interference should disappear when the separation of the mollow peaks approaches the fundamental timescales for photon absorption in the detector .
the proposed two - photon beating experiment depends on the interference between the probability amplitudes of two types of biphotons emitted at various time from a two - level system . in the counter - propagating geometry
, the two - photon interference has a symmetrically damped rabi oscillation with the oscillation period inversely proportional to the effective rabi frequency .
moreover , due to the relative phase difference caused by the refractive indices of two types of biphotons , the visibility of the center part of the two - photon coincidence pattern can be altered from 0 to @xmath3 in the ideal case , by subtracting the linear rayleigh scattering of the input pump beam .
this change indicates a switch between photon anti - bunching - like and photon bunching - like effects .
alternatively , the interference between two types of fwms can be switched from destructive to constructive by changing the pump rabi frequency ( or the pump power ) , the length of the medium , and the atomic density .
furthermore , because the proposed beating experiment has direct similarity with the franson interferometer , such a high visibility ( ideally , 100@xmath4 ) of the center part in the two - photon beating pattern , greater than 70.7@xmath4 as predicted by quantum mechanics , implies a violation of bell s inequality .
the paper is organized as follows . in sec .
ii , we describe the linear and nonlinear optical responses for different photon pairs generated in the mollow triplet . in sec .
iii , we present the detailed analysis of the two - photon beatings using biphotons created from a two - level system .
we emphasize that the two - photon state is a coherent superposition of two fwm processes , and paired photons produced from these processes experience different phases .
the analog between the proposed experiment and the franson interferometer is built by analyzing the two - photon state .
finally , we give our conclusions in sec . iv .
consider a medium of identical two - level _ atoms _ ( or _ molecules _ ) initially prepared in their ground level @xmath5 .
we configure our system with presence of a retro - reflected pump beam . in this configuration , the generated paired photons
are spontaneously emitted in opposite directions and are detected by two photodetectors d@xmath6 and d@xmath7 , as illustrated in fig .
the strong pump field with angular frequency @xmath8 and rabi frequency @xmath9 is applied to the quantum transition @xmath10 with a detuning @xmath11 , where @xmath12 is the atomic transition frequency .
the atomic oscillators are driven into the nonlinear regime where the absorbed two counter - propagating pump photons are re - radiated at frequencies @xmath13 and @xmath14 as seen in fig .
it is found that the linear optical response to the @xmath15 field and the third - order nonlinear susceptibility of the @xmath15 field are @xcite @xmath16,\label{eq : linear}\end{aligned}\ ] ] @xmath17 , \label{eq : nonlinearity}\end{aligned}\ ] ] with @xmath18 in eqs .
( [ eq : linear])-([eq : d ] ) , @xmath19 is the decay rate of the ground state , @xmath20 is the dephasing rate , @xmath21 and @xmath22 are the linewidths of biphotons produced from the central component and two sidebands in the mollow triplet , @xmath23 is the atomic density , and @xmath24 is the dipole matrix element , respectively . in the resonant pumping case , @xmath21 reduces to @xmath20 and @xmath22 becomes @xmath25 . to emphasize two fwm processes occurring in the system
, we have decomposed @xmath26 into two terms as shown in the second line of eq .
( [ eq : nonlinearity ] ) .
the mollow triplet resonances embedded in @xmath27 is depicted by the third - order nonlinear susceptibility @xmath28 shown in fig .
1(c ) . to look at the optical response of the medium , we begin with eqs .
( [ eq : linear ] ) and ( [ eq : nonlinearity ] ) .
it is imperative to examine the linear optical response of paired photons generated from the central or two sidebands in the mollow triplet and ascertain their propagation properties . in fig .
1(d ) , we plot real ( red ) and imaginary ( blue ) parts of the linear susceptibility of the @xmath15 field [ eq . ( [ eq : linear ] ) ] for the resonant pumping case .
the slopes of the real part of @xmath29 directly govern the group velocities of paired photons crossing the medium while the imaginary part dictates types of spectral transmission windows . as shown in fig .
1(d ) , biphotons originating from the central component experience an anomalous dispersion inside the medium .
consequently , photons generated from the central component undergo a superluminal propagation ( group velocity greater than @xmath30 , the speed of light in vacuum ) .
this can be understood from the gain - assisted superluminal light propagation mechanism @xcite where the anomalous dispersion plays the dominant role .
also shown in fig .
1(d ) , near the region of the sideband frequencies @xmath31 the indices of refraction reach a maximum ( or minimum ) value and the absorption of the medium vanishes .
in other words , paired photons originating from the two sidebands propagate with tunable refractive indices accompanied by vanishing absorption ( of the medium ) . for the photons originating from the central component
, they propagate with unity index of refraction accompanied by ( strong ) absorption .
this difference in refractive indices between the two types of biphotons generated from either central or sidebands serves as the pathway - length difference in the franson s experiment .
in the case of spontaneous emission , the two - photon amplitude at the two photodetectors can in general be represented as @xmath32 where @xmath33 is the timing difference between two detectors , @xmath34 is the effective rabi frequency ( in the resonant pumping case as discussed in this paper , it is only determined by the pump rabi frequency ) , and @xmath35 is the annihilation operator of the free - space propagating electric field at detector d@xmath36 ( @xmath37 ) , respectively . in the second line of eq .
( [ eq : state ] ) , the first term in the bracket implies that biphotons are produced from the central component @xmath38 ( which have the same frequency as the pump ) , and the second term stands for biphotons generated from two correlated sidebands @xmath39 ( whose central frequencies are peaked away from the pump frequency by @xmath40 and @xmath41 ) , respectively .
the spectral separation between these two fwm processes is equal to @xmath40 .
the two - photon amplitude shown in eq .
( [ eq : state ] ) is a coherent superposition of two fwm processes .
the - " sign in eq .
( [ eq : state ] ) indicates a @xmath42 phase shift between two fwm processes .
the phase term @xmath43 arises from the rabi oscillation due to the dynamic stark shift of the atomic levels .
@xmath44 is the relative phase difference between two types of biphotons generated as they pass through the medium with length @xmath45 , @xmath46l}{2},\label{eq : phi}\end{aligned}\ ] ] where @xmath47 and @xmath48 are central wave numbers of two sidebands inside the material and @xmath49 the central wave number of the central component .
the dispersion within the medium is @xmath50 , and the refractive index @xmath51 where @xmath52 represents the real part of the linear susceptibility @xmath53 . in the resonant pumping case ,
@xmath52 at two sideband frequencies is opposite in sign : positive at @xmath54 but negative at @xmath55 [ fig .
this property sets a useful region of the refractive index for two sidebands , i.e. , @xmath56 and @xmath57 .
we note that @xmath58 can be enhanced without absorption @xcite . if @xmath59 , however , @xmath60 becomes imaginary , indicating an evanescent exponentially decreasing field and destroying the efficiency of biphoton generation .
( [ eq : state ] ) also shows that the two types of biphotons have different spectral widths @xmath19 and @xmath20 .
@xmath61 and @xmath44 together determine the pattern of the two - photon temporal correlation , as shown below . to compare our result with the franson interferometer , we know that the two - photon amplitude presented in eq .
( 4 ) can also be visualized as a path - entangled biphoton wavepacket , as paired photons from central component in the mollow triplet traverse the short - short pathways while biphotons from correlated sidebands traverse the long - long pathways to cross over the medium . even though the amplitude given in eq .
( [ eq : state ] ) is physically different from that required in the franson experiment , the similarity between these two states offers another way to look at the path - entangled problem .
one consequence is to examine the violation of bell - type inequality by studying the two - photon beatings .
, weak fields at frequencies @xmath62 and @xmath63 are generated , shown by wavy arrows .
( c ) spectrum of the third - order nonlinear susceptibility @xmath64 shows a feature of the mollow triplet .
( d ) linear optical response in the resonant pumping case : the red line is the real part of @xmath29 and the blue one is the imaginary part . the central component will experience anomalous dispersion . ]
the two - photon temporal coincidence counting rate @xmath65 between detectors d@xmath6 and d@xmath7 is @xmath66 which describes the second - order quantum coherence . in eq .
( [ eq : definition ] ) @xmath67 and @xmath68 are the quantum detection efficiencies of two detectors , and @xmath69 is the two - photon amplitude given in eq .
( [ eq : state ] ) , respectively .
now we concentrate on the experimental setup shown in fig .
setting @xmath70 ideally and making use of eq .
( [ eq : state ] ) give @xmath71.\label{eq : coin1}\end{aligned}\ ] ] the physics of eq .
( [ eq : coin1 ] ) is easy to be understood as follows : the first term corresponds to the correlation of biphotons from the central component in the mollow triplet , the second term to the correlation of biphotons from two sidebands , and the third term to an interference between two previous correlations .
considering the beating experiment proposed here , an alternative explanation of eq .
( [ eq : coin1 ] ) can therefore be visualized as the first term gives the intensity of the biphotons following the short - short pathway in the franson experiment , the second term gives the intensity of the biphotons following the long - long pathway , and the third term is the interference between the two biphotons .
equation ( [ eq : coin1 ] ) shows that the pattern of the two - photon interference is a damped rabi oscillation where the oscillation frequency is the effective rabi frequency @xmath40 and the damping rate is determined by the linewidths in the mollow triplet . by subtracting accidental coincidences from the pump
, the visibility in the center part of @xmath72 can be changed between 0 and 1 by manipulating @xmath44 .
this change indicates a switch occurrence from photon anti - bunching - like effect to bunching - like effect .
alternatively , the interference occurring between two types of fwm processes can be switched from destructive to constructive , similar as the discussions presented in the case of a three - level double-@xmath2 system @xcite .
equation ( [ eq : coin1 ] ) can be also written as @xmath73-\cos(\omega_e\tau-\phi)\}e^{-(\gamma_g+\gamma_2)\tau},\label{eq : coin2}\end{aligned}\ ] ] where @xmath74 is a hyperbolic function . if @xmath75 , eq .
( [ eq : coin2 ] ) reduces to a simple form @xmath76e^{-2\gamma\tau}.\label{eq : coin3}\end{aligned}\ ] ] without the background noise , the two - photon temporal correlation is shown in fig . 2 .
by adjusting the pump rabi frequency , the atomic density , and the length of the medium , one can manipulate the center part in the two - photon interference ( or beating ) pattern .
the ( red ) dotted line represents result using a weaker pump while the ( blue ) solid line represents a stronger pump case .
the oscillation period using a weaker pump power is slower than that with a stronger pump power . the central parts ( which are highlighted by a green rectangular ) of the beating interference patterns are different : the ( red ) dotted curve almost goes to zero , indicating that the relative phase difference @xmath44 in this case is negligible ; while the ( blue ) solid line is above zero , indicating that @xmath44 plays a major role in such a case .
hence , by changing the pump power , the range of the visibility in the center part is ranged from one to zero . as discussed in many papers @xcite
, 50@xmath4 is the maximum visibility possible in a classical field approach to this sort of experiment . before proceeding the discussion , we note that a narrower filter would destroy the interference effect here .
however , since the mollow triplet separation is tunable , this experiment may provide an interesting way of probing the quantum nature of the detection process .
the interference will disappear when the separation of the mollow peaks approaches the fundamental timescales for photon absorption in the detector .
to realize this beating experiment , the natural bandwidth of biphotons from the phase matching conditions should be made larger than the linewidths of the mollow triplet so that the intrinsic mechanism of biphoton generation is not erased by the phase matching .
experimentally , the optical depth of the medium should be small such that the two - photon wavepacket is determined by @xmath77 and not by the phase matching @xcite .
realistically , two potential candidate configurations may be chosen as candidates for this type of experiment : ( 1 ) the system of cavity enhanced quantum dots , and ( 2 ) the laser cooled and trapped atomic ensemble .
however , when operating in the resonant pumping case , ( 2 ) imposes a tougher challenge because atoms are often kicked out " of the trapping volume and this leads to the absorption dominant in the process . to avoid such problems
, it may be advantageous to work with blue - detuned and red - detuned pumping arrangements .
in summary , a new beating experiment is proposed by using paired photons generated from a two - level system in the resonant pumping configuration .
the mollow triplet in the third - order nonlinear susceptibility indicates that two types of biphotons are created in the parametric interaction .
the linear response to two types of biphotons allows tunable refractive indices with vanishing absorption to two sidebands in the mollow triplet while as a unity index of refraction to the central component .
the relative phase difference caused by different refractive indices is analogous to the interferometer proposed by franson . by manipulating the pump power , the length of the material , and the atomic density , the visibility of the center part in the two - photon beating interference can be changed for 0 to 1 by subtracting background coincidences from the linear rayleigh scattering of the pump beam .
the observed interference pattern shows a switch from the destructive interference to the constructive in the fwm processes .
although the experiment is difficult , it is definitely possible to observe the nonzero effect in the center part of the beating pattern in the current labs .
in addition , the experiment may be a way of probing the quantum nature of the detection process , since the interference will disappear when the separation of the mollow peaks approaches the fundamental timescales for photon absorption in the detector .
the authors are grateful to yanhua shih , james d. franson , and todd b. pittman for fruitful discussions .
we also thank the referees for insightful suggestions . j .-
w. and m. h. r. were supported in part by the u.s .
army research office under muri grant w911nf-05 - 1 - 0197 .
s. d. acknowledges the start - up support from the department of physics at the hong kong university of science and technology e. o. was supported by the office of naval research .
s. j. freedman and j. f. clauser , phys .
* 28 * , 938 - 941 ( 1972 ) ; a. aspect , p. grangier , and g. roger , phys .
lett . * 49 * , 91 - 94 ( 1982 ) ; a. aspect , j. dalibard , and g. roger , phys .
lett . * 49 * , 1804 - 1807 ( 1982 ) . z. y. ou , x. y. zou , l. j. wang , and l. mandel , phys
lett . * 65 * , 321 - 324 ( 1990 ) ; j. brendel , e. mohler , and w. martiensen , europhys . lett . * 20 * , 575 - 580 ( 1992 ) ; y .- h .
shih , a. v. sergienko , and m. h. rubin , phys .
a * 47 * , 1288 - 1293 ( 1993 ) . | we propose a two - photon beating experiment based upon biphotons generated from a resonant pumping two - level system operating in a backward geometry . on the one hand
, the linear optical - response leads biphotons produced from two sidebands in the mollow triplet to propagate with tunable refractive indices , while the central - component propagates with unity refractive index .
the relative phase difference due to different refractive indices is analogous to the pathway - length difference between long - long and short - short in the original franson interferometer . by subtracting the linear rayleigh scattering of the pump
, the visibility in the center part of the two - photon beating interference can be ideally manipulated among [ 0 , 100@xmath0 by varying the pump power , the material length , and the atomic density , which indicates a bell - type inequality violation . on the other hand , the proposed experiment may be an interesting way of probing the quantum nature of the detection process .
the interference will disappear when the separation of the mollow peaks approaches the fundamental timescales for photon absorption in the detector . |
since 2009 an increasingly significant deviation from background expectations has been identified in the data of the large area telescope ( lat ) on board the fermi gamma ray space telescope satellite @xcite .
the deviation appears around 2 gev in the energy spectrum of gamma ray flux originating from an extended region centered in the galactic center .
the source of the excess photons is unknown .
their origin can be dark matter ( dm ) annihilation , a population of millisecond pulsars or supernova remnants @xcite , or cosmic rays injected in a burst - like or continuous event at the galactic center @xcite .
it is , however , challenging to explain the excess with millisecond pulsars @xcite based on their luminosity function .
recently , several groups including daylan et al .
@xcite , calore et al .
@xcite , and the fermi collaboration @xcite re - analyzed data from the fermi - lat @xcite and concluded that the 13 gev gamma ray signal is statistically significant and appears to originate from dark matter particles annihilating rather than standard astrophysical sources .
the peak in the energy distribution is broadly consistent with gamma rays originating from self - annihilation of dark matter particles @xcite .
the intensity of the signal suggests a dark matter annihilation cross section at thermal freeze out @xcite . the diffuse nature and morphology of the gamma ray excess
is consistent with a navarro - frenk - white - like galactic distribution of dark matter @xcite .
this gamma ray excess thus drew the attention of a number of particle model builders and phenomenologists @xcite . the conclusion that we have discovered dark matter particles , however ,
can not be drawn yet .
first , we have to be able to exclude the possibility of a standard astrophysical explanation .
second , we need to demonstrate that a dark matter particle that explains the gamma ray excess ( with a given mass , spin , and interaction strength to the standard sector ) is consistent with a large number of other observations .
the latter concerns our paper .
we aim to determine the microscopic properties of the dark matter particle from the gamma ray excess and check that these properties comply with limits from other experiments .
we use dark matter abundance and direct detection data , measurements of the gamma ray flux from the galactic center , near earth positron and anti - proton flux data , cosmic microwave background ( cmb ) observations , and measurements of galactic radio emission as experimental constraints . amongst the above listed experimental bounds the constraining role of radio emission has been debated in the literature .
bringmann et al .
have shown that radio emission imposes severe constraints on dark matter annihilation in the galactic center @xcite .
radio emission , however , could be induced by various processes including synchrotron radiation , inverse compton scattering , ionization , and bremsstrahlung .
most studies of the radio constraint on dark matter , including that of bringmann et al .
, ignore energy loss processes other than synchrotron radiation . however , as pointed out by cholis et al . in ref .
@xcite , there are several reasons why the other processes could be important .
cholis et al .
have shown that after considering inverse compton scattering induced by high densities of radiation in the inner milky way the radio constraint on dark matter is weakened by about three orders of magnitude @xcite . as a result dark matter annihilating at the thermal rate remains compatible with the radio data .
after considering the effect of diffusion the constraint will be further weakened . due to this , we will exclude the radio data point from our combined fit . as theoretical description of dark matter we use the simplified model framework . within this ansatz
we make minimal and general theoretical assumptions .
we consider a single dark matter particle that couples to various standard fermions via a mediator .
our dark matter particle thus annihilates to several final states which all contribute to the observables mentioned above .
this paper is organized as follows . in sec .
[ sec : models ] we introduce the simplified dark matter model we use . in sec .
[ sec : constraints ] , we describe the observables of dark matter abundance and scattering on nuclei , gamma , @xmath2 , and @xmath3 cosmic ray fluxes , and the effect of dark matter annihilations on the cmb . our numerical results are given in sec .
[ sec : results ] . finally in sec .
[ sec : concl ] we summarize our main results .
we collect the formulae of bayesian inference and likelihood functions in the appendix .
in this section we motivate and describe the theoretical hypothesis we test . in ref . @xcite we compared bayesian evidences for three leading simplified models to explain the gamma ray excess from the galactic center .
we found that the experimental data , especially dark matter direct detection , clearly preferred a majorana fermion dark matter particle coupled to standard model ( sm ) fermions via a real scalar .
motivated by this we assume that the dark matter particle is a majorana fermion , which we denote with @xmath4
. inspired by the higgs portal mechanism @xcite , we use a simplified model to describe interactions between @xmath4 and sm matter .
we assume that the dark - standard mediator is a real scalar field , @xmath5 , and the form of the dark matter to mediator coupling is @xmath6 the presence of @xmath7 is essential since it is lifting the velocity suppression that one otherwise encounters in the indirect detection cross section , thus making this operator capable of explaining the gamma ray excess .
the interaction between the mediator and sm fermions @xmath8 is assumed to be @xmath9 in line with minimal flavor violation @xcite , we only consider the third generation fermions , i.e. @xmath10 . for simplicity we assume that mediator pair final states are not present in the dark matter annihilation and only consider s - channel annihilation diagrams .
according to power counting of the dark matter transfer momentum or velocity @xcite , with the bi - linears in eqs .
( [ eq : interaction0 ] ) and ( [ eq : intercation1 ] ) the annihilation cross section of the fermionic dark matter candidate is not velocity suppressed , that is @xmath11 . the dark matter - nucleon elastic scattering cross section is spin - independent ( si ) and momentum suppressed .
in this section we describe the calculation of the observables that we use to constrain the parameter space of our hypothesis .
table [ tab : observables ] summarizes these observables .
.summary of observables we use to constrain our dark matter scenario .
the expressions in the second column are defined in the text of this section . [ cols="^,^,^,^,^",options="header " , ] in our numerical calculation we fix the dark matter to mediator coupling as @xmath12 , and we scan the following free parameters : @xmath13 here @xmath14 is the mass of the dark matter particle , @xmath15 is the mass of the scalar mediator , and @xmath16 is the coupling of the mediator to the sm fermion pair @xmath17 as defined in eqs .
( [ eq : interaction0 ] ) and ( [ eq : intercation1 ] ) .
the range of our scan over the above parameters and the type of prior we use is given in table [ tab : scanrange ] .
and @xmath18 credible regions , respectively.,title="fig:",width=302 ] and @xmath18 credible regions , respectively.,title="fig:",width=302 ] to build some intuition , first we examine the constraining effect of each observable one by one . to this end
we plot the posterior probability distributions marginalized to the scanned model parameters such that the likelihood function only contains the dark matter abundance and one of the other observables . in fig .
[ fig : gammaray ] we show marginalized posterior probability distributions taking into account the dark matter abundance and the gamma ray data . the first frame of fig .
[ fig : gammaray ] confirms that the gamma ray data restrict the range of the dark matter mass close to 3560 gev @xcite .
it is less appreciated , however , that uncertainties still allow a 40100 ( 25160 ) gev dark matter mass range at the @xmath1 ( @xmath18 ) credibility level .
the gamma ray data , coupled with the dark matter relic density , allows the whole mass range of the scalar mediator .
it is also interesting to note that the preferred dark matter mass region is dissected by a diagonal band with a lower posterior around the on - shell resonance region @xmath19 . in this valley ,
dark matter resonantly annihilates via the @xmath20-channel mediator , depleting its abundance .
thus , it is harder for the model to match the planck constraint .
the relevant interaction strengths also remain virtually unconstrained as shown by the right frame of fig .
[ fig : gammaray ] .
planck and the anomalous fermi - lat gamma ray data only restrict these coupling in the @xmath21 @xmath22 ( @xmath23 1 ) and @xmath24 @xmath22 ( @xmath25 1 ) ranges at the @xmath1 ( @xmath18 ) credibility level .
simultaneous order 1 couplings are marginally allowed and appear in the part of the parameter space where the annihilation cross section is suppressed by sizable @xmath15 . in fig .
[ e+ ] we show marginalized posterior probability distributions with the likelihood function containing only the dark matter abundance and the ams-02 positron flux data .
the ams-02 measurement of the positron flux features a small upward kink , a sudden change of slope , around 35 gev .
a smooth background prediction has a hard time to reproduce this kink and systematically falls below the experimental points in the 3550 gev region .
positrons originating from the annihilation of a 35 - 50 gev dark matter particle can fill the gap between the background and the data .
hence the ams-02 data show a mild preference toward a dark matter candidate with 1665 ( 10160 ) gev mass at @xmath1 ( @xmath18 ) credibility level .
ams-02 also restricts the dominant decay to @xmath26 leptons with a @xmath27 coupling around @xmath28 and @xmath29 coupling below about @xmath30 at the @xmath1 credibility level .
+ fig .
[ pbar ] shows marginalized posterior probability distributions with the likelihood function including the planck and pamela anti - proton cosmic ray flux data only .
the pamela data in itself does not prefer any particular parameter region .
dark matter and mediator masses are both allowed in the full scanned range at the @xmath18 credibility level .
this happens with the exception of a small island around @xmath31 gev where the combined planck and pamela constraints are harder to satisfy .
the reasons for this are that this island falls on the @xmath19 resonant annihilation corridor and the pamela data around 10 gev leave very little room for dark matter .
this situation improves for lower dark matter masses .
since dark matter masses above the top quark mass are allowed by the combination of planck and pamela , the @xmath32 coupling comes into play . these data , however , are not sufficient to constrain @xmath32 .
it is interesting to note that planck and pamela allow fairly large values of @xmath32 , @xmath29 , and @xmath27 for heavier @xmath14 and @xmath15 in order to accommodate the correct relic abundance . )
.,title="fig:",width=302 ] + ) .,title="fig:",width=302 ] ) .,title="fig:",width=302 ] in fig .
[ cmb ] we show marginalized posterior distributions for the dark matter abundance and cmb likelihood function , defined in eq .
( [ eq : cmbl ] ) .
the diagonal depletion of the likelihood function due to dark matter resonant annihilation is apparent in the @xmath14 vs @xmath15 frame .
the cmb likelihood function suppresses the posterior around @xmath33 gev providing more constraint on low mass dark matter .
the posterior probability distribution projected to the @xmath29 vs. @xmath27 couplings shows a peculiar pattern .
this pattern is the combined result of two relatively simple sets of constraints .
dark matter abundance is responsible for the low likelihood values at low @xmath29 and @xmath27 .
it turns out that the planck constraint on the amount of relic dark matter is hard to respect unless one of these couplings is sizable , that is @xmath29 or @xmath34 at the @xmath1 credibility level .
if both of these couplings are small then annihilation is slow and dark matter is overproduced in the early universe . in the large coupling region , on the other hand , the cmb constrains @xmath29 and @xmath27 from above .
if any of these couplings are larger than about 0.1 then dark matter tends to become under - produced and the cmb receives too much modification from dark matter .
the @xmath32 coupling is hardly constrained by the cmb at the @xmath18 credibility level . + in fig .
[ dd ] we show the marginalized posterior distribution with the likelihood function containing only the planck and lux data .
due to momentum suppression of the nucleon-@xmath4 elastic scattering , the lux data very weakly constrain the dark matter or mediator mass . in the @xmath14 vs. @xmath15 plane
the diagonal resonant annihilation valley is visible , but no other structure is present .
the posterior probability distribution for the couplings is very similar to that in fig .
similarly to the case of the cmb , planck and lux only impose a constraint on the order 1 couplings . as discussed in the introduction ,
the radio signal potentially very strictly constrains dark matter @xcite . assuming that dark matter contributes to the radio signal only by synchrotron radiation we find the radio flux upper limit of jodrell bank at 408 mhz @xcite excludes the dark matter hypothesis we consider by two orders of magnitude .
our finding fully confirms that of ref .
this exclusion , on the other hand , is lifted if inverse compton scattering , ionization , and bremsstrahlung are also considered as potential dark matter energy loss mechanisms leading to radio emission @xcite .
as shown by ref .
@xcite the bound from the radio data is weakened by about three orders of magnitude if inverse compton scattering is considered and is expected to pose no constraint after including galactic diffusion effects . due to this
, we do not include the radio emission data point in our combined likelihood function .
the summary of all constraints is presented in fig . [ fig : combined ] .
the constraint on the dark matter mass is dominated by the gamma ray data and the final combination restricts @xmath14 to the 10100 ( 7125 ) gev region with @xmath1 ( @xmath18 ) credibility .
less of the low mediator mass region survives the scrutiny of the combined constraints , leaving the 31000 gev @xmath15 region preferred at the @xmath1 credibility level .
the combined constraints prefer a somewhat correlated pair of @xmath27 and @xmath29 couplings in the intermediate @xmath01 region at the @xmath1 credibility level .
small ( @xmath35 ) and large ( @xmath36 ) values of couplings are disfavored mostly by planck at the @xmath18 credibility . comparing the second frame of fig .
[ fig : combined ] to those showing the individual constraints it is clear that simultaneous order 1 couplings are mildly under stress from almost all the data .
in this work we perform a comprehensive statistical analysis of the gamma ray excess from the galactic center in a simplified dark matter model framework .
according to our previous study , majorana fermion dark matter interacting with standard model fermions via a scalar mediator is the most favoured explanation of the galactic center excess when characterised by bayesian evidence .
we locate the most plausible parameter regions of this theoretical hypothesis using experimental data on the dark matter abundance and direct detection interactions , the gamma ray flux from the galactic center , near earth positron and anti - proton fluxes , the cosmic microwave background , and galactic radio emission .
we find that the radio data excludes the model if we include synchrotron radiation as the only energy loss channel .
since it was shown that inclusion of other types of energy losses lifts this exclusion we discard the single radio data point from our combined likelihood @xcite .
the rest of the data prefers a dark matter ( mediator ) mass in the 10100 ( 31000 ) gev region and weakly correlated couplings to bottom quarks and tau leptons with values of @xmath01 at the @xmath1 credibility level .
we thank pat scott for helping with the cmb constraint and alejandro ibarra for providing positron flux data points in ref . @xcite . this work in part was supported by the arc centre of excellence for particle physics at the terascale .
the use of monash university sun grid , a high - performance computing facility , and the national computational infrastructure ( nci ) , the southern hemisphere s fastest supercomputer , is also gratefully acknowledged .
mjw is supported by the australian research council future fellowship ft140100244 .
cs thanks the school of physics at monash university , where part of this work was performed , for its hospitality .
cs was partially supported by nsf award 10681111 .
in this section we summarize the statistical background of our analysis .
let @xmath37 and @xmath38 denote the plausibility of two non - exclusive propositions , @xmath39 and @xmath40 , in light of some prior information , @xmath41 .
the probability that both @xmath39 and @xmath40 are correct is given by the conditional expression @xmath42 bayes theorem follows from the symmetry of the conditional probability under the exchange of @xmath39 and @xmath40 : @xmath43 in this context @xmath37 is typically called the prior probability and represents the plausibility of our hypothesis given information prior the observation @xmath40 .
the likelihood function @xmath44 indicates how accurately the hypothesis can replicate the data .
the posterior probability @xmath45 quantifies the plausibility of the hypothesis @xmath39 given the data @xmath40 .
the evidence @xmath38 serves to normalize the posterior . for theoretical models with a continuous parameter @xmath46 bayes theorem can be recast in the form @xmath47 the posterior distribution can be used to estimate the most likely region of @xmath46 .
the evidence is calculated via an integral over the full parameter space @xmath48 for more than one continuous parameters , @xmath49 , marginalization is performed by integrating the posterior over various parameters in the higher dimensional parameter space @xmath50 100 l. goodenough and d. hooper , arxiv:0910.2998 [ hep - ph ] .
d. hooper and l. goodenough , phys .
b * 697 * , 412 ( 2011 ) [ arxiv:1010.2752 [ hep - ph ] ] .
a. boyarsky , d. malyshev and o. ruchayskiy , phys .
b * 705 * , 165 ( 2011 ) [ arxiv:1012.5839 [ hep - ph ] ] .
k. n. abazajian and m. kaplinghat , phys . rev .
d * 86 * , 083511 ( 2012 ) [ arxiv:1207.6047 [ astro-ph.he ] ] .
d. hooper and t. r. slatyer , phys .
dark univ .
* 2 * , 118 ( 2013 ) [ arxiv:1302.6589 [ astro-ph.he ] ] . c. gordon and o. macias , phys .
d * 88 * , 083521 ( 2013 ) [ arxiv:1306.5725 [ astro-ph.he ] ] .
t. daylan , d. p. finkbeiner , d. hooper , t. linden , s. k. n. portillo , n. l. rodd and t. r. slatyer , arxiv:1402.6703 [ astro-ph.he ] . f. calore , i. cholis and c. weniger , jcap * 1503 * , 038 ( 2015 ) [ arxiv:1409.0042 [ astro-ph.co ] ] .
s. murgia , presentation at fifth fermi symposium , 20 - 24 oct .
e. carlson and s. profumo , phys .
d * 90 * , 023015 ( 2014 ) [ arxiv:1405.7685 [ astro-ph.he ] ] .
r. m. oleary , m. d. kistler , m. kerr and j. dexter , arxiv:1504.02477 [ astro-ph.he ] .
j. petrovic , p. d. serpico and g. zaharijas , jcap * 1502 * , no .
02 , 023 ( 2015 ) [ arxiv:1411.2980 [ astro-ph.he ] ] .
q. yuan and k. ioka , astrophys . j. * 802 * , no . 2 , 124 ( 2015 ) [ arxiv:1411.4363 [ astro-ph.he ] ] . j. petrovic , p. d. serpico and g. zaharijas , arxiv:1405.7928 [ astro-ph.he ] . i. cholis , d. hooper and t. linden , arxiv:1407.5625 [ astro-ph.he ] . i. cholis , d. hooper and t. linden , arxiv:1407.5583 [ astro-ph.he ] .
w. b. atwood _ et al . _
[ lat collaboration ] , astrophys .
j. * 697 * , 1071 ( 2009 ) [ arxiv:0902.1089 [ astro-ph.im ] ] . g. a. gmez - vargas , m. a. snchez - conde , j. h. huh , m. peir , f. prada , a. morselli , a. klypin and d. g. cerdeo _ et al .
_ , jcap * 1310 * , 029 ( 2013 ) [ arxiv:1308.3515 [ astro-ph.he ] ] . k. n. abazajian , n. canac , s. horiuchi and m. kaplinghat , phys .
d * 90 * , 023526 ( 2014 ) [ arxiv:1402.4090 [ astro-ph.he ] ] .
s. ipek , d. mckeen and a. e. nelson , arxiv:1404.3716 [ hep - ph ] .
ko , w. i. park and y. tang , arxiv:1404.5257 [ hep - ph ] .
j. m. cline , g. dupuis , z. liu and w. xue , jhep * 1408 * , 131 ( 2014 ) [ arxiv:1405.7691 [ hep - ph ] ] .
ko and y. tang , arxiv:1407.5492 [ hep - ph ] .
a. martin , j. shelton and j. unwin , arxiv:1405.0272 [ hep - ph ] ; k. kong and j. c. park , arxiv:1404.3741 [ hep - ph ] . c. boehm , m. j. dolan and c. mccabe , phys .
d * 90 * , 023531 ( 2014 ) [ arxiv:1404.4977 [ hep - ph ] ] .
d. k. ghosh , s. mondal and i. saha , arxiv:1405.0206 [ hep - ph ] .
l. wang , arxiv:1406.3598 [ hep - ph ] .
b. d. fields , s. l. shapiro and j. shelton , phys .
lett . *
113 * , 151302 ( 2014 ) [ arxiv:1406.4856 [ astro-ph.he ] ] .
a. berlin , d. hooper and s. d. mcdermott , arxiv:1404.0022 [ hep - ph ] .
k. n. abazajian , n. canac , s. horiuchi and m. kaplinghat , arxiv:1402.4090 [ astro-ph.he ] ; e. hardy , r. lasenby and j. unwin , arxiv:1402.4500 [ hep - ph ] ; d. p. finkbeiner and n. weiner , arxiv:1402.6671 [ hep - ph ] ; t. lacroix , c. boehm and j. silk , arxiv:1403.1987 [ astro-ph.he ] ; a. hektor and l. marzola , arxiv:1403.3401 [ hep - ph ] ; a. alves , s. profumo , f. s. queiroz and w. shepherd , arxiv:1403.5027 [ hep - ph ] ; p. agrawal , b. batell , d. hooper and t. lin , arxiv:1404.1373 [ hep - ph ] ; e. izaguirre , g. krnjaic and b. shuve , arxiv:1404.2018 [ hep - ph ] ; v. gammaldi , j. a. r. cembranos , a. de la cruz - dombriz , r. a. lineros and a. l. maroto , arxiv:1404.2067 [ hep - ph ] ; q. yuan and b. zhang , arxiv:1404.2318 [ astro-ph.he ] ; d. g. cerdeno , m. peiro and s. robles , arxiv:1404.2572 [ hep - ph ] ; m. demianski and a. doroshkevich , arxiv:1404.3362 [ astro-ph.co ] ; s. ipek , d. mckeen and a. e. nelson , arxiv:1404.3716 [ hep - ph ] ; k. kong and j. -c .
park , arxiv:1404.3741 [ hep - ph ] ; c. boehm , m. j. dolan and c. mccabe , arxiv:1404.4977 [ hep - ph ] ; p. ko , w. -i .
park and y. tang , arxiv:1404.5257 [ hep - ph ] ; m. shirasaki , s. horiuchi and n. yoshida , arxiv:1404.5503 [ astro-ph.co ] ; m. abdullah , a. difranzo , a. rajaraman , t. m. p. tait , p. tanedo and a. m. wijangco , arxiv:1404.6528 [ hep - ph ] ; d. k. ghosh , s. mondal and i. saha , arxiv:1405.0206 [ hep - ph ] ; a. drlica - wagner , g. a. gomez - vargas , j. w. hewitt , t. linden and l. tibaldo , arxiv:1405.1030 [ astro-ph.he ] ; j. bramante and t. linden , arxiv:1405.1031 [ astro-ph.he ] ; a. berlin , p. gratia , d. hooper and s. d. mcdermott , arxiv:1405.5204 [ hep - ph ] ; t. basak and t. mondal , arxiv:1405.4877 [ hep - ph ] ; n. bernal , j. e. forero - romero , r. garani and s. palomares - ruiz , arxiv:1405.6240 [ astro-ph.co ] ; p. agrawal , m. blanke and k. gemmler , arxiv:1405.6709 [ hep - ph ] ; t. m. yoast - hull , j. s. gallagher and e. g. zweibel , arxiv:1405.7059 [ astro-ph.he ] ; k. agashe , y. cui , l. necib and j. thaler , arxiv:1405.7370 [ hep - ph ] ; e. carlson and s. profumo , arxiv:1405.7685 [ astro-ph.he ] ; j. m. cline , g. dupuis , z. liu and w. xue , arxiv:1405.7691 [ hep - ph ] ; j. petrovic , p. d. serpico and g. zaharijas , arxiv:1405.7928 [ astro-ph.he ] ; s. k. n. portillo and d. p. finkbeiner , arxiv:1406.0507 [ astro-ph.im ] ; t. han , z. liu and s. su , arxiv:1406.1181 [ hep - ph ] ; w. detmold , m. mccullough and a. pochinsky , arxiv:1406.2276 [ hep - ph ] ; l. wang , arxiv:1406.3598 [ hep - ph ] ; c. boehm , p. gondolo , p. jean , t. lacroix , c. norman and j. silk , arxiv:1406.4683 [ astro-ph.he ] ; b. d. fields , s. l. shapiro and j. shelton , arxiv:1406.4856 [ astro-ph.he ] ; a. askew , s. chauhan , b. penning , w. shepherd and m. tripathi , international journal of modern physics a , vol . 29 ( 2014 ) 1430041 [ arxiv:1406.5662 [ hep - ph ] ] ; c. cheung , m. papucci , d. sanford , n. r. shah and k. m. zurek , arxiv:1406.6372 [ hep - ph ] ; s. d. mcdermott , arxiv:1406.6408 [ hep - ph ] .
p. ko and y. tang , jcap * 1501 * , 023 ( 2015 ) [ arxiv:1407.5492 [ hep - ph ] ] ; d. borah and a. dasgupta , phys .
b * 741 * , 103 ( 2015 ) [ arxiv:1409.1406 [ hep - ph ] ] ; j. guo , j. li , t. li and a. g. williams , phys .
d * 91 * , no .
9 , 095003 ( 2015 ) [ arxiv:1409.7864 [ hep - ph ] ] ; p. agrawal , b. batell , p. j. fox and r. harnik , jcap * 1505 * , no .
05 , 011 ( 2015 ) [ arxiv:1411.2592 [ hep - ph ] ] ; k. cheung , w. c. huang and y. l. s. tsai , arxiv:1411.2619 [ hep - ph ] ; f. calore , i. cholis , c. mccabe and c. weniger , phys .
d * 91 * , no .
6 , 063003 ( 2015 ) [ arxiv:1411.4647 [ hep - ph ] ] ; k. ghorbani and h. ghorbani , arxiv:1501.00206 [ hep - ph ] ; c. h. chen and t. nomura , arxiv:1501.07413 [ hep - ph ] ; k. p. modak and d. majumdar , arxiv:1502.05682 [ hep - ph ] ; a. achterberg , s. caron , l. hendriks , r. ruiz de austri and c. weniger , arxiv:1502.05703 [ hep - ph ] ; j. conrad , j. cohen - tanugi and l. e. strigari , arxiv:1503.06348 [ astro-ph.co ] ; j. m. cline , g. dupuis , z. liu and w. xue , arxiv:1503.08213 [ hep - ph ] ; e. c. f. s. fortes , v. pleitez and f. w. stecker , arxiv:1503.08220 [ hep - ph ] ; p. ko and y. tang , arxiv:1504.03908 [ hep - ph ] ; j. kim , j. c. park and s. c. park , arxiv:1505.04620 [ hep - ph ] ; s. dado and a. dar , arxiv:1505.04988 [ astro-ph.he ] . t. bringmann , m. vollmann and c. weniger , phys . rev .
d * 90 * , no .
12 , 123001 ( 2014 ) [ arxiv:1406.6027 [ astro-ph.he ] ] .
i. cholis , d. hooper and t. linden , phys .
d * 91 * , no .
8 , 083507 ( 2015 ) [ arxiv:1408.6224 [ astro-ph.he ] ] . c. balzs and t. li , phys
. rev .
d * 90 * , no . 5 , 055026 ( 2014 ) [ arxiv:1407.0174 [ hep - ph ] ] . for recent discussions , see j. m. cline , k. kainulainen , p. scott and c. weniger , phys .
d * 88 * , 055025 ( 2013 ) [ arxiv:1306.4710 [ hep - ph ] ] and references therein .
g. dambrosio , g. f. giudice , g. isidori and a. strumia , nucl .
b * 645 * , 155 ( 2002 ) [ hep - ph/0207036 ] .
p. a. r. ade _ et al .
_ [ planck collaboration ] , arxiv:1303.5076 [ astro-ph.co ] .
m. aguilar [ ams collaboration ] , phys .
lett . * 113 * , 121102 ( 2014 ) .
o. adriani , g. a. bazilevskaya , g. c. barbarino , r. bellotti , m. boezio , e. a. bogomolov , v. bonvicini and m. bongi _ et al .
_ , jetp lett .
* 96 * , 621 ( 2013 ) [ pisma zh .
fiz . * 96 * , 693 ( 2012 ) ] . j. m. cline and p. scott , jcap * 1303 * , 044 ( 2013 ) [ erratum - ibid . * 1305 * , e01 ( 2013 ) ] [ arxiv:1301.5908 [ astro-ph.co ] ] .
d. s. akerib _ et al . _
[ lux collaboration ] , phys .
* 112 * , 091303 ( 2014 ) [ arxiv:1310.8214 [ astro-ph.co ] ] .
r. d. davies , d. walsh , and r. s. booth , mon . not .
r. astron
. soc . * 177 * , 319 ( 1976 ) .
g. belanger , f. boudjema , a. pukhov and a. semenov , comput .
commun . * 185 * , 960 ( 2014 ) [ arxiv:1305.0237 [ hep - ph ] ] .
j. f. navarro , c. s. frenk and s. d. m. white , astrophys . j. * 462 * , 563 ( 1996 ) [ astro - ph/9508025 ] ;
j. f. navarro , c. s. frenk and s. d. m. white , astrophys . j. * 490 * , 493 ( 1997 ) [ astro - ph/9611107 ] .
g. belanger , f. boudjema , p. brun , a. pukhov , s. rosier - lees , p. salati and a. semenov , comput .
commun . * 182 * , 842 ( 2011 ) [ arxiv:1004.1092 [ hep - ph ] ] .
a. ibarra , a. s. lamperstorfer and j. silk , phys .
d * 89 * , 063539 ( 2014 ) [ arxiv:1309.2570 [ hep - ph ] ] .
h. b. jin , y. l. wu and y. f. zhou , arxiv:1410.0171 [ hep - ph ] .
h. b. jin , y. l. wu and y. f. zhou , arxiv:1504.04604 [ hep - ph ] .
p. junnarkar and a. walker - loud , phys .
d * 87 * , no .
11 , 114510 ( 2013 ) [ arxiv:1301.1114 [ hep - lat ] ] .
a. crivellin , m. hoferichter and m. procura , phys .
d * 89 * , 054021 ( 2014 ) [ arxiv:1312.4951 [ hep - ph ] ] .
m. szydagis , n. barry , k. kazkaz , j. mock , d. stolp , m. sweany , m. tripathi and s. uvarov _ et al .
_ , jinst * 6 * , p10002 ( 2011 ) [ arxiv:1106.1613 [ physics.ins-det ] ]
. m. szydagis , a. fyhrie , d. thorngren and m. tripathi , jinst * 8 * , c10003 ( 2013 ) [ arxiv:1307.6601 [ physics.ins-det ] ] .
a. alloul , n. d. christensen , c. degrande , c. duhr and b. fuks , comput .
commun . * 185 * , 2250 ( 2014 ) [ arxiv:1310.1921 [ hep - ph ] ] .
f. feroz , m. p. hobson and m. bridges , mon . not .
soc . * 398 * , 1601 ( 2009 ) [ arxiv:0809.3437 [ astro - ph ] ] .
j. skilling , bayesian analysis 1 ( 4 ) 833860 ( 2006 ) . | we test the plausibility that a majorana fermion dark matter candidate with a scalar mediator explains the gamma ray excess from the galactic center . assuming that the mediator couples to all third generation fermions we calculate observables for dark matter abundance and scattering on nuclei , gamma , positron , and anti - proton cosmic ray fluxes , radio emission from dark matter annihilation , and the effect of dark matter annihilations on the cmb . after discarding the controversial radio observation
the rest of the data prefers a dark matter ( mediator ) mass in the 10100 ( 31000 ) gev region and weakly correlated couplings to bottom quarks and tau leptons with values of @xmath01 at the @xmath1 credibility level . |
the central challenge for particle physics nowadays is the origin of mass . in the standard model
both fermions and gauge boson masses are generated through interactions with the same scalar particle , the higgs boson , h. if it exists , the higgs boson will certainly be discovered by the time a photon collider is constructed .
the aim of this machine will be then a precise measurement of the higgs properties . the photon scattering can be used to produce the higgs particles singly in the s - channel of the colliding photons .
this facility permits a high precision measurement of the h @xmath4 partial width , which is sensitive to new charged particles .
the measurement is significantly important .
if we find a deviation of the two photon width from the standard model prediction it means that an additional contribution from unknown particles is present , and so it is a signature of physics beyond the sm .
for example , the minimal extension of the sm predicts the ratio of the two photon width @xmath5/ @xmath6 1.2 @xcite for a higgs boson with a mass of 120 gev , assuming a supersymmetry scale of 1 tev and the chargino mass parameters @xmath7 and @xmath8 of 300 and 100 gev , respectively . a photon collider can measure the product @xmath9@xmath10br@xmath11 . to obtain the two - photon partial width independent of the branching ratio one has to combine with an accurate measurement of the br(@xmath12 ) from another machine .
this study investigates the accuracy of the measurement of the two photon decay width times the branching ratio for a higgs boson with the mass of 120 gev , the preferred mass region by recent electroweak data @xcite .
the signal and background processes studied are specified in section 2 .
event selection is described in section 3 .
results are summarized in section 4 .
the feasibility of the measurement of the two photon decay width of the higgs boson in this mass region has also been reported by @xcite .
high energy photon beams can be produced at a high rate in compton backscattering of laser photons off high energy electrons @xcite . setting opposite helicities for the laser photons and the beam electrons the energy spectrum of the backscattered photons
is peaked at 80@xmath2 of the @xmath13 beam energy .
the backscattered photons are highly polarized in this high energy region . with an integrated luminosity of 80 fb@xmath3 per year for @xmath14 @xcite ,
about 20000 higgs bosons with standard model coupling and a mass of 120 gev can be produced in the @xmath0h process . in this mass region
the higgs particle will decay dominantly into a b@xmath1 pair .
the beam spectra at @xmath15 = 210 gev are simulated using the compaz @xcite , a fast parameterization which includes multiple interactions and non - linearity effects .
the response of the detector has been simulated with simdet 4 @xcite , a parametric monte carlo for the tesla @xmath16 detector .
it includes a tracking and calorimeter simulation and a reconstruction of energy - flow - objects ( efo ) .
only efos with a polar angle above 7@xmath17 can be taken for the higgs reconstruction simulating the acceptance of the photon collider detector as the only deference to the @xmath16 detector @xcite .
the considered backgrounds are the direct continuum @xmath18 and @xmath19 production . due to helicity conservation
, the continuum background production proceeds mainly through states of opposite photon helicities , making the states @xmath20 .
choosing equal helicity photon polarizations the cross section of the continuum background is suppressed by a factor @xmath21 , with @xmath22 being the quark mass .
unfortunately , this suppression does not apply to the process @xmath23 , because after the gluon radiation the @xmath24 system is not necessarily in a @xmath25 state .
the surviving background is large and overwhelms the signal .
signal @xmath26 events are generated using pythia 6.2 @xcite .
background processes @xmath27(g ) are generated with the sherpa ( the simulation for high energy reactions of particles ) @xcite generator .
higher order qcd effects are simulated in pythia by evolving the hard process event using the parton shower , which allows partons to split into pairs of other partons .
this technique is most effective when the emitted gluons are soft or collinear , while the region of high transverse momentum is poorly described . for a reliable background estimation
correct nlo corrections are needed .
combining the hard process with its higher order correction in pythia is not trivial .
a fraction of @xmath28 events are included in the @xmath29 process via gluon radiation in the parton shower . combining the two processes without special procedures
amounts in double counting of some phase space regions .
sherpa @xcite is a generator which matches correctly the exact matrix elements with showering .
a comparison between the distributions of jet multiplicity in the @xmath30 process obtained with pythia and sherpa is shown in figure [ fig_jets ] .
however , since virtual diagrams are not included , to obtain the correct rates the sherpa results for the spin @xmath31 states have to be scaled by the correct nlo cross section .
the cross sections at the nlo for the @xmath32 process have been calculated @xcite , @xcite .
a cross section comparison is shown in figure [ fig_cs ] .
.cross sections and the number of expected and generated events for the signal and background processes [ cols="<,^ " , ] [ tab_cs ]
the analysis aims to select events with two or three jets from the higgs boson decay .
two of these jets contain bottom quarks .
the invariant mass of the jets has to be consistent with the higgs mass .
high multiplicity ( feo ) events are selected and their visible energy is required to be greater than 95 gev .
events with longitudinal imbalance greater than 10@xmath2 of the visible energy are rejected .
finally the cosine of the thrust angle has to be less than 0.7 . in the remaining event sample jets
are reconstructed using the durham clustering scheme @xcite with the resolution parameter y@xmath33=0.02 .
the cross section for the continuum production of the charm quark is 16 times larger than for bottom quarks , therefore b - quark tagging is crucial for this analysis .
the b - tagging algorithm combines several discriminating variables , as for example , the impact parameter joint probability tag introduced by aleph @xcite , the @xmath34 corrected vertex invariant mass obtained with the zvtop algorithm written for the sld experiment @xcite and a one - prong charm tag , into a feed forward neural network with 12 inputs and 3 output nodes , described in ref .
@xcite . the distribution of the neural network output @xmath35 to discriminate b - quark jets from u- , d- , s- and c - quark jets is presented in figure [ fig_btag]a .
the performance of the neural network b - tag in z@xmath17 decays is shown in figure [ fig_btag]b .
the b - tagging efficiency is 70@xmath2 and corresponds to a purity of 98@xmath2 .
the b - quarks coming from the decay of the higgs boson are highly energetic , whereas in the case of the background processes the gluon and one b - quark jet are the most energetic . in order to reduce the background
further we look at the two fastest jets in the event and require the @xmath35 to be greater than 0.95 for one jet and greater than 0.2 for the second one
. + the expected number of signal and background events are summarized in table [ tab_cs ] .
a total signal efficiency is estimated to be 36@xmath2 .
the hadronic cross section for @xmath0hadrons events , within the energy range above 2 gev , is about 400 nb , so that about 1 such event is produced per bunch crossing ( pileup ) .
these events obscure the interesting physics processes described in the previous sections . for this reason this class of events needs to be included in the pythia simulation for overlap in the next step of this analysis .
the hades @xcite program will be used for this purpose .
a large fraction of this background is distributed at small angles and we believe that it can be reduced cutting on the polar angle of the tracks @xcite .
a careful study of this hadronic background is currently being performed .
the reconstructed invariant mass for the selected signal and background events is shown in figure [ fig_mass ] . to enhance the signal a cut on the invariant mass
is tuned such that the statistical significance of the signal over background is maximized .
events in the mass region of 114 gev @xmath36 126 gev are selected .
the number of estimated signal and background events in this window are 4505 and 1698 , respectively .
the two photon decay width of the higgs boson is proportional to the event rates of the higgs signal .
the statistical error of the number of signal events , @xmath37 , corresponds to the statistical error of this measurement .
here @xmath38 is the number of observed events , while @xmath39 is the number of expected background events .
we conclude that for a higgs boson with a mass @xmath41=120 gev we can measure the product @xmath42 with an accuracy of 1.7@xmath2 using an integrated luminosity corresponding to one year of data taking at the tesla photon collider .
t. ohgaki , t. takahashi and i. watanabe , hep - ph/9703301 ; g. jikia and s. sldner - rembold , nucl . inst . and meth a472 ( 2001 ) 133 ; p. niezurawski , a.f .
zarnecki and m. krawczyk , acta physica polonica b34 ( 2003 ) 177 .
p. niezurawski , a.f .
zarnecki and m. krawczyk , hep - ph/0307183 . | this study investigates the potential of a photon collider for measuring the two photon partial width times the branching ratio of a light higgs boson .
the analysis is based on the reconstruction of the higgs events produced in the @xmath0h process , followed by higgs decay into a b@xmath1 pair . a statistical error of the measurement of the two - photon width times the b@xmath1 branching ratio of the higgs boson
is found to be 1.7 @xmath2 with an integrated luminosity of 80 fb@xmath3 in the high energy part of the spectrum . |
the challenges faced by companies working in nowadays complex it environments pose the need for comprehensive and dynamic systems to cope with the information flow requirements @xcite , @xcite , @xcite .
planning can not answer all questions : we must take a step further and discuss a model for application management .
one of the possible approaches to deal with this problem , is to use the decision support system that is capable of supporting decision - making activities . in @xcite , we proposed the foundations of our decision support system for complex it environments . developing our framework
, we examined the time , energy usage , qop , finance and carbon dioxide emissions .
regarding financial and economic analyzes , we considered only _
variable _ costs .
however , calculating total operating cost of a data center , one needs to take into account both _ fixed _ and _ variable _ costs , which are affected by complex and interrelated factors . in this paper , performing a financial analysis of security - based data flow , we present an improved method for measuring the total cost of its maintenance , exploring trade - offs offered by different security configurations , performance variability and economic expenditures . with the proposed approach it is possible to reduce the data center costs without compromising data security or quality of service .
the main contributions of this paper are summarized as follows : * we enhanced previous studies on security management presented in @xcite , extending it with the analysis of fixed costs , * we proposed a full cost model for data centers , being an economic method of their evaluation , in which all costs of operating , maintaining , and disposing are considered important , * we prepared a case study , in which we : * * applied the developed , financial model to an example data center , in order to show how the model actually works , * * we evaluated the proposed economic scheme and analyzed the distribution of data center maintenance costs over five years , taking into account different security levels , and comparing them with reference to data center total costs , * * based on the results gathered with the introduced method , we calculated possible profits and return on investment values for over five years , in order to choose the best option ( considering the security of the information , along with high company incomes ) .
the total cost of the data center is somewhat elusive to project accurately .
there are many subtleties which can be overlooked or are simply unaccounted for ( or perhaps underestimated ) , over the operational life of a data center .
looking to involve all existing elements of a modern data center into cost calculations , several approaches have been proposed in the literature ( @xcite , @xcite , @xcite ) .
in @xcite the authors presented a way of predicting the total budget required to build a new data center .
they distinguish three primary construction cost drivers , namely : power and cooling capacity and density , tier of functionality and the size of the computer room floor .
they utilize proposed cost model , providing calculations for the example data center .
researchers state that their cost model is intended as a quick tool that can be applied very early in the planning cycle to accurately reflect the primary construction cost drivers .
however , proposed model makes some rigid assumptions about the data center ( such as the minimum floor space or the type of the utilized rack ) , making it quite inflexible .
moreover , as the authors themselves admit , some significant costs were not included in this cost model ( for instance , the operational costs ) .
an interesting approach to assessing and optimizing the cost of computing at the data center level is presented in @xcite . here ,
scientists consider five main components , that should be taken into account , while evaluating the total cost of a data center : the construction of the data center building itself , the power and cooling infrastructure , the cost of electricity to power ( and cool ) the servers and the cost of managing those servers . besides creating a cost model for the data center ,
the authors examine the influence of server utilization on total cost of a data center , and state that an effective way to decrease the cost of computing is to increase server utilization .
another method for assessing the total cost of a data center is proposed in @xcite .
an approach examined in @xcite is a part of research which seeks to understand and design next generation servers for emerging warehouse computing environments .
the authors developed cost models and evaluation metrics , including an overall metric of performance per unit total cost of ownership .
they identify four key areas for improvement , and study initial solutions that provide significant benefits . unlike our approach
, @xcite focuses mainly on performance and calculates data center costs only on its basis .
the method proposed in our paper considers different aspects at once , which makes it flexible and suitable for heterogeneous environments .
although existing cost models for the data center include many components , none of them mention security as the significant factor . however , security influences data center costs as well : proper security management translates into better utilization of central resources as well as reduced systems management and administration . in this section , we present and describe the formulas utilized in the proposed analysis process - in particular , in economic and financial analyzes , extending them with the calculation of _ fixed _ costs .
introduced equations are used to evaluate the financial aspect of the data center maintenance . performing the financial analysis
, we took under consideration cost of power delivery , cost of cooling , software and hardware expenditures as well as personnel salaries ( both _ fixed _ and _ variable _ costs ) .
proposed method of data center cost evaluation is the sum of the present values of all costs over the lifetime of a data center ( including investment costs , capital costs , financing costs , installation costs , energy costs , operating costs , maintenance costs and security assurance costs ) .
as shown in figure [ fig : extended_financial_analysis ] , in the introduced financial and economic analyzes , we distinguished @xmath0 main components : _ cost of power delivery _
, _ cost of cooling infrastructure utilization _ and _ operational costs_. each of them can be further specified .
firstly , we introduce the general formula used for calculating total energy costs .
the following equation is further detailed in terms of cpu and server utilization : @xmath1 where : * @xmath2 - is the total amount of the utilized kilowatt - hours * @xmath3 - is the total amount of hours when the server was busy * @xmath4 - is the total amount of days when the server was busy * @xmath5 - is the cost of a one kwh in us dollars the above formula is the crucial point in computing the total cash outlay for the considered system .
original equation is elaborated in the following sections .
the design of the electrical power system must ensure that adequate , high - quality power is provided to each machine at all times . except the back - up system operational expenditures
, data center spends a lot of money on current power consumption - utilized both for the compute , network and storage resources . at this point one
should pay special attention to the relationship between the cpu utilization and energy consumed by other components of a working server .
since the amount of the power consumed by a single machine consists of the energy usage of all its elements , it is obvious that a decrease in the energy consumed by the cpu will result in a lower energy consumption in total .
thus , total cost of both power delivery and utilization can be summarized as follows : @xmath6 where : * @xmath7 - is the total amount of the utilized kilowatt - hours by the server * @xmath3 - is the total amount of hours when the server was busy * @xmath4 - is the total amount of days when the server was busy * @xmath5 - is the cost of a one kwh being aware of the price of one kwh , and knowing that cpu worked @xmath3 hours through @xmath8days , utilizing @xmath2 kilowatt - hours , it is fairly straightforward to calculate the total financial cost ( @xmath9 ) of its work , using the equation . before we start further evaluation of the energy consumed by the cpu , we need to make some assumptions about its utilization .
let us introduce the simplified cpu utilization formula : @xmath10 where : * @xmath11 - stands for the cpu utilization , expressed in percentage * @xmath12 - defines our requirements , the actual busy time of the cpu ( seconds ) * @xmath13 - is the cpu capacity , the total time spent on analysis ( seconds ) usually , the cpu utilization is measured in percentage .
the requirements specified in the above formula refer to the time we require from the cpu to perform an action .
this time is also known as the _ busy _ time .
cpu capacity can be expressed as the sum of the _ busy _ and _ idle _ time ( that is , the _ total _ time available for the cpu ) .
going simple , one can say that over a 1 minute interval , the cpu can provide a maximum of @xmath14 of its seconds ( power ) . the cpu _ capacity _ can then be understood as _ busy time _
+ _ idle time _ ( the time which was used plus the one which was left over ) . using the above simplifications , when going multi - core , cpu capacity should be multiplied by the number of the cpu cores ( @xmath15 ) .
the presented equation can be further detailed as follows : @xmath16 = \frac{time_{_{session } } \ , \cdot \ , users}{time_{_{total}}}\ ; , \ ] ] where @xmath17 is expressed as @xmath18 .
+ supposing a specified cpu load and assuming the server is able to handle a defined number of users within a given time , we can calculate the _ idle _ time using the above equation .
regarding the _ busy _ time , one should use the results obtained for the prepared , example model .
this simple formula can be used for calculating the total energy consumed by the cpu . knowing the amount of energy utilized by the cpu , it is quite straightforward to assess costs incurred for the consumed energy . as the cooling infrastructure absorbs energy to fulfill its function
, the cost of the cooling needs to be included in the total cost of the server maintenance . to obtain an approximate amount of the power consumed by the cooling
, one can use the equipment heat dissipation specifications , most often expressed in british thermal units ( btu ) .
specifications generally state how many btu are generated in each hour by the individual machine .
therefore , the formula for calculating the cooling cost to keep the equipment in normal operating conditions , is given as follows ( values per server ) : @xmath19 where : * @xmath20 - is the amount of the btus generated by the cooling system * @xmath21 - is the total amount of the utilized kilowatt - hours by the cooling system * @xmath3 - is the total amount of hours when the cooling system was busy * @xmath4 - is the total amount of days when the cooling system was busy * @xmath5 - is the cost of a one kwh operational costs of the data center management depend on miscellaneous factors - among them one can enumerate salaries of the employees responsible for managing servers ( along with the number of employees needed to adequately maintain and operate the data center ) , prices of equipment , the amount of a reduction in the value of servers with the passage of time and finally on software and licensing costs . in order to determine the complete cost of a data center
, software and licensing costs should be analyzed as well .
the server one purchases may , or may not include an operating system .
when it comes to selecting a server os , high - end server operating systems can be quite expensive . except the operating system ,
one also needs to budget for the software applications the server will need in order to perform its tasks .
the dollar amounts can add up quite quickly in this area , depending on the role of the data center and the server itself .
it is very common in high - end server applications to offer per core licensing for some editions of the software .
dealing with hardware costs , one should be aware of server depreciation as well .
hence , the annual hardware cost is in fact an amortization cost , calculated as follows : @xmath22 where : * @xmath23 - is the purchase cost of a server ( in us dollars ) * @xmath24 - is the average life - time of a single server ( in years ) when it comes to the brand new equipment , besides the amortization expenditures ( the overall costs associated with installing , maintaining , upgrading and supporting a server ) , one should consider purchasing costs as well .
saying so , the total , annual hardware and software cost of a single machine can be estimated using the formula below : @xmath25 the day may come when data centers are self - maintaining , but until then , one will need personnel to operate and maintain server rooms .
internal personnel of the data center usually consist of it staff , data center security personnel , the data center managers , facilities maintenance personnel and housekeeping personnel .
if the data center contains tens - if not hundreds - of thousands of working machines , it is common to have more than one employee dealing with a given equipment .
discussing personnel costs , it can be calculated as the total number of employees multiplied by the salary of the particular staff member ( to simplify our evaluation , we can consider the average salary for every employee in the enterprise ) : @xmath26 where : * @xmath27 - is the total number of it personnel * @xmath28 - is the total number of ordinary workers * @xmath29 - is the total number of the housekeeping and facilities maintenance personnel * @xmath30 - is the average salary in the enterprise ( per month ) once the data center is built , it still requires financial investment to ensure a high - quality , competitive services with guaranteed levels of availability , protection and support continuously for 24 hours a week .
key elements in data center budgets are the power delivery system , the networking equipment and the cooling infrastructure . besides the above most - crucial factors , there exist additional costs associated with data center operation , such as personnel and software expenses .
therefore , the real operating cost of the data center can be expressed as : @xmath31 where each of the defined components consists of further operational expenditures .
this concludes the discussion on calculating the total cost of a data center maintenance , resulting in the following formula ( being the combination of all the above equations ) : @xmath32 it is significant to remember that @xmath33 refers to the annual hardware and software costs .
therefore , when calculating the total cost of a data center maintenance , one needs to adjust this value individually , depending on the considered analysis time interval .
to demonstrate the use of the proposed analysis scheme , we used the role - based access control approach , prepared an example data center scenario and analyzed it with the help of the introduced method .
we made use of qop - ml @xcite , @xcite and created by its means the role based access control model to examine the quality of chosen security mechanisms in terms of financial impact of data center maintenance .
before we perform the actual estimation of the data center maintenance cost , let us give some assumptions about the examined environment .
consider a call center company located in nevada , usa , managing a typical it environment of @xmath34u server racks ( @xmath35 physical servers in total , @xmath36 physical servers per rack ) . given a specified load capacity , servers handle enterprise traffic continuously for @xmath37 hours . in our analysis , we assume that all the utilized applications are tunnelled by the tls protocol . in the considered access control method ,
users are assigned to specific roles , and permissions are granted to each role based on the users job requirements .
users can be assigned any number of roles in order to conduct day - to - day tasks ( figure [ fig : callcenter ] ) . in order to emphasize and prove role s influence on data flow management and system s performance , we prepared and analyzed a simple scenario .
this scenario refers to the real business situation and possible role assignment in the actual enterprise environment .
given the example enterprise network infrastructure , consider having three roles : _
role1 _ , _
role2 _ and _ role3 _ with corresponding security levels : _ low _ , _ medium _ and _ high_. each server in the example call center is equipped with the intel xeon @xmath38 processor , being able to handle the required number of employees connections , regardless of the assigned rbac role .
prepared scenarios are listed in table [ table : rbac_scenarios ] .
qop - ml s
security models used in our case study can be downloaded from qop - ml s project webpage @xcite . after introducing the environment , we present an overview of predicting the total budget required to manage an example data center , focusing on an introduced method for measuring its total cost and indicating possible gains . by the analysis of an example scenario
, we try to confirm our thesis about the influence of security management on the total cost of the data center maintenance . 1 . _ cost of power delivery _ to calculate the total cost of power delivery , we performed both experimental and theoretical investigations .
regarding experiments , we utilized dell ups to measure the current power consumption of a single dell poweredge r@xmath39 server performing operations defined in our scenario . in order to obtain the most accurate results of the cpu power consumption , we made use of our model , and performed analysis using the aqopa tool @xcite . to ensure the accuracy of gathered results , in our simulation
, we utilized real hardware metrics , provided by the cmt @xcite for dell poweredge r@xmath39 server . 1 . _ server power consumption _
according to dell ups in the laboratory , poweredge r@xmath39 performing defined operations consumes on average @xmath40w . since the server handles enterprise s traffic continuously for @xmath37 hours , its annual power consumption is equal to @xmath41 kwhs . as stated by http://www.electricitylocal.com/states/nevada/las-vegas/ ,
the average industrial electricity rate in las vegas , nevada is @xmath42 cents @xmath43 ( 0.0756 $ ) per kwh .
if the server works @xmath44 days a year , it will cost the company about @xmath45 $ . 2 .
_ cpu power consumption _
when it comes to the cpu power consumption , we assumed its load to be equal to @xmath46 .
we performed a simulation using prepared model , considering @xmath0 levels of users permissions ( which differ in security level ) .
+ in table [ table : cost_cpu_server ] we collected power consumption costs , both for the cpu and the server in total ( rounded up to the nearest dollar ) .
as evidenced by table [ table : cost_cpu_server ] , considering only cost savings related to the consumed energy , it is possible to handle about @xmath47 ( when switching between the first and the third role ) and @xmath48 ( by changing the role from third to second ) more users with the exact cpu load .
those figures , when put in the context of a large data center environment , quickly become very significant .
cost of cooling infrastructure utilization _ in addition to the power delivered to the compute hardware , power is also consumed by the cooling resources .
the load on the cooling equipment is directly proportional to the power consumed by the compute hardware . in such case ,
the cost of the cooling infrastructure is equal to the cost of the energy consumed by the server and its cpu .
3 . _ operational costs _ apart from electricity and cooling , calculating the total maintenance cost of a data center , one should take under consideration also the cost of its physical infrastructure , such as hardware amortization and the actual price of physical machines .
besides the equipment cost , operational expenditures must be covered as well .
determining the approximate , total cost of a whole data center , we assumed to use dell power edge r@xmath39 servers ( @xmath49 each ) .
however , we have include neither the network , nor the storage footprint ( nor its equipment ) .
regarding utilized software , we assumed that working machines have windows server data center edition installed ( whose price is equal to about @xmath50 dollars ) .
except the os , workstations use some proprietary software , which can cost about @xmath51 $ on average ( @xcite , @xcite ) .
( we assume the cost of the software for a single machine to be equal to @xmath52 $ . )
the number of employees of the whole data center consists of security managers , system operators , call center employees and housekeeping and facility maintenance personnel , resulting in @xmath53 employees in total .
as stated by http://swz.salary.com , the median salary of security manager in us is equal to about @xmath54 dollars per month ( at the time of writing ) . in our estimation
, we used this value as the average salary in our call center company . in order to prove that a proper data flow management has a significant impact on data center maintenance costs
, we tried to estimate them over @xmath55 years .
although it might not be easily noticed at first glance , it turned out that our approach can bring meaningful savings and influence rapid _ return on investment _ ( _ roi _ ) increase .
( table [ table : r1r2compareme ] explores this concept in more detail . ) in our approach , economic profits come from the number of handled users - the more customers ( served clients ) , the higher company profits . with the exact cpu load ,
the same number of working machines is capable of handling a greater number of users . by switching between the _ strongest _ and _ weakest _ security mechanisms ,
the example call center company can achieve actual roi growth .
our analysis showed , that it is possible to provide effective services and keep the utilization of hardware resources at a certain level .
since we can accomplish given goal using _ weaker _ security mechanisms , in many situations it is wasteful to assign too many hardware resources to perform the given task . applying the proposed solution to the existing it environment
, one can observe a serious increase in company incomes , while preserving the efficiency , utilization and flexibility of the existing computer hardware .
one of the main pricing models of a call center company concerns _ cost per contact _
, where all costs are combined into an unit price .
the total income is then based on the number of served clients , such as calls , emails or chat sessions .
for the example call center , we assumed that on average , single served customer brings the income of about @xmath0 us dollars . as it was proven by the time analysis presented in @xcite , server working with _
permissions is able to handle about @xmath56 users within an hour having @xmath57% of cpu load .
since we assumed that the number of users grows linearly , within @xmath37 hours , it gives us @xmath58 employees a day , resulting in @xmath59 connections a year per server .
if we assume that we have at our disposal the whole data center , it will turn out that we can serve roughly @xmath60 users assigned _
permissions a year . with the exact cpu load , using _ role1 _
permissions , server is capable of dealing with @xmath61 users within an hour , which results in @xmath62 served customers a year per single machine and @xmath63 clients per whole data center . when we translate the above calculations to the incomes and outcomes of the company , we see that the proper information management brings a variety of economic advantages .
according to our previous assumptions , given the incomes equal to @xmath64 $ , @xmath65 $ and @xmath66 $ and outcomes of @xmath67 $ , @xmath68 $ and @xmath69 $ for roles @xmath70 , @xmath71 and @xmath0 respectively , in table [ table : r1r2compareme ] we calculated actual profits and roi values of the company for over @xmath55 years .
as it is summarized in table [ table : r1r2compareme ] , considering the first working year of the example call center , the efficiency of an investment is much bigger when we handle users using role s @xmath70 security mechanisms ( comparing to the role @xmath0 ) , and about @xmath72 times greater , considering roles @xmath71 and @xmath0 .
real profits can be observed after the @xmath55 years of call center business activity .
the analysis shows , that if the company used the first role permissions instead of those from role @xmath71 , it could gain about @xmath55 times more money .
what is more , if we consider the third and the first role , profits grows rapidly , resulting in about @xmath14 times greater gain . since high roi values
means that the investment gains compare favorably to investment cost , and the primary goal of the call center is a fast return on the investment , company should re - think implemented information flow mechanisms .
as proved by our analyzes , the main drivers of data center cost are power and cooling .
in contrast to operational costs , they represent _ variable _ costs , which vary over time and depend on many factors .
as power consumption and electricity prices rise , energy costs are receiving more scrutiny from senior - level executives seeking to manage dollars .
however , focusing on the financial aspect of the data center , one can not forget about the proper data management . in this paper , we utilized qop - ml to increase company incomes , without compromising data security or quality of service .
the proposed analysis scheme provides new opportunities and possibilities , not only for measuring data center costs , but also for increasing incomes .
optimization of the available computational power can be accomplished in many different ways : by modifying system configurations , switching between utilized security mechanisms , by the suitable selection of used applications and services and adequate application management .
k.mazur , b.ksiezopolski and a.wierzbicki `` on security management : improving energy efficiency , decreasing negative environmental impact and reducing financial costs for data centers , '' _ mathematical problems in engineering _ , 418535 , 119 .
kevin lim , parthasarathy ranganathan , jichuam chang , chadrakant patel , trevor mudge and steven reinhardt , understanding and designing new server architectures for emerging warehouse - computing environments , _ proceeding of the acm international symposium on computer architecture _ ,
beijing , china , june 2008 .
d.rusinek , b.ksiezopolski and a.wierzbicki , `` aqopa : automated quality of protection analysis framework for complex systems '' , _ 14th international conference on computer information systems and industrial management applications _ , warsaw , springer : lncs , v.9339 , 475486 , 2015 . | information management is one of the most significant issues in nowadays data centers .
selection of appropriate software , security mechanisms and effective energy consumption management together with caring for the environment enforces a profound analysis of the considered system . besides these factors ,
financial analysis of data center maintenance is another important aspect that needs to be considered .
data centers are mission - critical components of all large enterprises and frequently cost hundreds of millions of dollars to build , yet few high - level executives understand the true cost of operating such facilities .
costs are typically spread across the it , networking , and facilities , which makes management of these costs and assessment of alternatives difficult .
this paper deals with a research on multilevel analysis of data center management and presents an approach to estimate the true total costs of operating data center physical facilities , taking into account the proper management of the information flow . |
the computational task of detecting maximum weight matchings is one of the most fundamental problems in discrete optimization , attracting plenty of attention from the operations research , computer science , and mathematics communities .
( for a wealth of references on matching problems see @xcite . ) in such settings , we are given an undirected graph @xmath2 whose edges are associated with non - negative weights specified by @xmath3 .
a set of edges @xmath4 is a _ matching _ if no two of the edges share a common vertex , that is , the degree of any vertex in @xmath5 is at most @xmath6 .
the weight @xmath7 of a matching @xmath8 is defined as the combined weight of its edges , i.e. , @xmath9 .
the objective is to compute a matching of maximum weight .
we study this problem in two related computational models : the _ semi - streaming _ model and the _ preemptive online _ model .
[ [ the - semi - streaming - model . ] ] the semi - streaming model .
+ + + + + + + + + + + + + + + + + + + + + + + + + even though these settings appear to be rather simple as first glance , it is worth noting that matching problems have an abundance of flavors , usually depending on how the input is specified . in this paper , we investigate weighted matchings in the _ semi - streaming _ model , was first suggested by muthukrishnan @xcite .
specifically , a _ graph stream _ is a sequence @xmath10 of distinct edges , where @xmath10 is an arbitrary permutation of @xmath11 .
when an algorithm is processing the stream , edges are revealed sequentially , one at a time . letting @xmath12 and
@xmath13 a graph algorithm uses , the time @xmath14 it requires to process each edge , and the number of passes @xmath15 it makes over the input stream .
the main restriction is that the space @xmath16 is limited to @xmath17 bits of memory .
we refer the reader to a number of recent papers @xcite and to the references therein for a detailed literature review .
[ [ online - graph - problems . ] ] online graph problems .
+ + + + + + + + + + + + + + + + + + + + + + unlike the semi - streaming model , in online problems the size of the underlying graph is not known in advance .
the online matching problem has previously been modeled as follows .
edges are presented one by one to the algorithm , along with their weight .
once an edge is presented , we must make an irrevocable decision , whether to accept it or not .
an edge may be accepted only if its addition to the set of previously accepted edges forms a feasible matching .
in other words , an algorithm must keep a matching at all times , and its final output consists of all edges which were ever accepted . in this model , it is easy to verify that the competitive ratio of any ( deterministic or randomized ) algorithm exceeds any function of the number of vertices , meaning that no competitive algorithm exists .
however , if all weights are equal , a greedy approach which accepts an edge whenever possible , has a competitive ratio of @xmath18 , which is best possible for deterministic algorithms @xcite .
similarly to other online settings ( such as call control problems @xcite ) , a preemptive model can be defined , allowing us to remove a previously accepted edge from the current matching at any point in time ; this event is called _
preemption_. nevertheless , an edge which was either rejected or preempted can not be inserted to the matching later on .
we point out that other types of online matching problems were studied as well @xcite . [ [ comparison - between - the - models . ] ] comparison between the models .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + both semi - streaming algorithms and online algorithms perform a single pass over the input . however , unlike semi - streaming algorithms , online algorithms are allowed to concurrently utilize memory for two different purposes .
the first purpose is obviously to maintain the current solution , which must always be a feasible matching , implying that the memory size of this nature is bounded by the maximal size of a matching .
the second purpose is to keep track of arbitrary information regarding the past , without any concrete bound on the size of memory used .
therefore , in theory , online algorithms are allowed to use much larger memory than is allowed in the semi - streaming model . moreover , although this possibility is rarely used , online algorithms may perform exponential time computations whenever a new piece of input is revealed . on the other hand , a semi - streaming algorithm may re - insert an edge the current solution , even if it has been temporarily removed , as long as this edge was kept in memory .
this extra power is not allowed for online ( preemptive ) algorithms , making them inferior in this sense in comparison to their semi - streaming counterparts .
[ [ previous - work . ] ] previous work .
+ + + + + + + + + + + + + + feigenbaum et al .
@xcite were the first to study matching problems under similar assumptions .
their main results in this context were a semi - streaming algorithm that computes a @xmath19-approximation in @xmath20 passes for maximum cardinality matching in bipartite graphs , as well as a one - pass @xmath21-approximation for maximum weighted matching in arbitrary graphs .
later on , mcgregor @xcite improved on these findings , to obtain performance guarantees of @xmath22 and @xmath23 for the maximum cardinality and maximum weight versions , respectively , being able to handle arbitrary graphs with only a constant number of passes ( depending on @xmath24 ) .
in addition , mcgregor @xcite tweaked the one - pass algorithm of feigenbaum et al . into achieving a ratio of @xmath25 .
finally , zelke @xcite has recently attained an improved approximation factor of @xmath26 , which stands as the currently best one - pass algorithm .
note that the @xmath21-approximation algorithm in @xcite and the @xmath25-approximation algorithm in @xcite are preemptive online algorithms . on the other hand ,
the algorithm of zelke @xcite uses the notion of shadow - edges , which may be re - inserted into the matching , and hence it is not an online algorithm .
[ [ main - result - i . ] ] main result i. + + + + + + + + + + + + + + the first contribution of this paper is to improve on the above - mentioned results , by devising a deterministic one - pass algorithm in the semi - streaming model , whose performance guarantee is @xmath27 . in a nutshell ,
our approach is based on partitioning the edge set into @xmath28 weight classes , and computing a separate maximal matching for each such class in online fashion , using @xmath17 memory bits overall .
the crux lies in proving that the union of these matchings contains a single matching whose weight compares favorably to the optimal one .
the specifics of this algorithm are presented in section [ sec : app ] .
[ [ main - result - ii . ] ] main result ii .
+ + + + + + + + + + + + + + + our second contribution is motivated by the relation between semi - streaming algorithms and _ preemptive _ online algorithms , which must maintain a feasible matching at any point in time . to our knowledge
, there are currently no lower bounds on the competitive ratio that can be achieved by incorporating preemption .
thus , we also provide a lower bound of @xmath1 on the performance guarantee of any such deterministic algorithm . as a result , we show that improved one pass algorithms for this problem must store more than just a matching in memory .
further details are provided in section [ sec : lower_bound ] .
this section is devoted to obtaining main result i , that is , an improved one - pass algorithm for the weighted matching problem in the semi - streaming model .
we begin by presenting a simple deterministic algorithm with a performance guarantee of @xmath29 .
we then show how to randomize its parameters , still within the semi - streaming framework , and obtain an expected approximation ratio of @xmath30 .
finally , we de - randomize the algorithm by showing how to emulate the required randomness using multiple copies ( constant number ) of the deterministic algorithm , while paying an additional additive factor of at most @xmath31 , for any fixed @xmath32 . [
[ preliminaries . ] ] preliminaries .
+ + + + + + + + + + + + + + we maintain the maximum weight of any edge @xmath33 seen so far in the input stream .
clearly , the maximum weight matching of the edges seen so far has weight in the interval @xmath34 $ ] . note that if we disregard all edges with weight at most @xmath35 , the weight of the maximum weight matching in the resulting instance decreases by an additive term of at most @xmath36 .
our algorithm has a parameter @xmath37 , and a value @xmath38 .
we define weight classes of edges in the following way . for every @xmath39
, we let the class @xmath40 be the collection of edges whose weight is in the interval @xmath41 .
we note that by our initial assumption , the weight of each edge is in the interval @xmath42 $ ] , and we say that a weight class @xmath40 is _ under consideration _ if its weight interval @xmath41 intersects @xmath42 $ ] .
the number of classes which are under consideration at any point in time is @xmath43 .
[ [ the - algorithm . ] ] the algorithm .
+ + + + + + + + + + + + + + our algorithm simply maintains the list of classes under consideration and maintains a maximal ( unweighted ) matching for each such class .
in other words , when the value of @xmath33 changes , we delete from the memory some of these matchings , corresponding to the classes which stop being under consideration . note that to maintain a maximal matching in a given subgraph , we only need to check if the two endpoints of the new edge are not covered by existing edges of the matching . to conclude , for every new edge @xmath44 we proceed as follows .
we first check if @xmath45 is greater than the current value of @xmath33 .
if so , we update @xmath33 and the list of weight classes under consideration accordingly .
then , we find the weight class of @xmath45 , and try to extend its corresponding matching , i.e. , @xmath46 will be added to this matching if it remains a matching after doing so .
note that at each point the content of the memory is the value @xmath33 and a collection of @xmath43 matchings , consisting of @xmath47 edges overall .
therefore , our algorithm indeed falls in the semi - streaming model . at the conclusion of the input sequence ,
we need to return a single matching rather than a collection of matchings . to this end
, we could compute a maximum weighted matching of the edges in the current memory .
however , for the specific purposes of our analysis , we use the following faster algorithm .
we sort the edges in memory in decreasing order of weight classes , such that the edges in @xmath40 appear before those in @xmath48 , for every @xmath49 . using this sorted list of edges ,
we apply a greedy algorithm for selecting a maximal matching , in which the current edge is added to this matching if it remains a matching after doing so .
then , the post - processing time needed is linear in the size of the memory used , that is , @xmath47 .
this concludes the presentation of the algorithm and its implementation as a semi - streaming algorithm .
[ [ analysis . ] ] analysis .
+ + + + + + + + + for purposes of analysis , we round down the weight of each edge @xmath46 such that @xmath50 to be @xmath51 . this way
, we obtain _ rounded _ edge weights .
now fix an optimal solution @xmath52 and denote by @xmath52 its weight , and by @xmath53 its rounded weight .
the next claim immediately follows from the definition of @xmath40 .
@xmath54 . as an intermediate step ,
we analyze an improved algorithm which keeps all weight classes .
that is , for each @xmath49 , we use @xmath55 to denote the maximal matching of class @xmath40 at the end of the input , and denote by @xmath8 the solution obtained by this algorithm , if we would have applied it .
similarly , we denote by @xmath56 the set of edges in @xmath52 which belong to @xmath40 . for every @xmath49
, we define the set of vertices @xmath57 , associated with @xmath40 , to be the set of endpoints of edges in @xmath55 that are not associated with higher weight classes : @xmath58 for a vertex @xmath59 , we define its associated weight to be @xmath60 . for vertices which do not belong to any @xmath57
, we let their associated weight be zero . we next bound the total associated weight of all the vertices .
the total associated weight of all the vertices is at most @xmath61 . consider a vertex @xmath62 and let @xmath63 be the edge in @xmath55 adjacent to @xmath64 .
if @xmath65 then we charge the weight associated with @xmath64 to the edge @xmath63 .
thus , an edge @xmath66 is charged at most twice from vertices associated with its own weight class . otherwise , if @xmath67 then there must be some other edge @xmath68 , for some @xmath69 , that prevented us from adding @xmath63 to @xmath8 , in which case we charge the weight associated with @xmath64 to @xmath46 .
notice that @xmath70 , for otherwise , @xmath64 would not be associated with @xmath40 .
thus , the edge @xmath71 must be of the form @xmath72 and can only be charged twice from vertices in weight class @xmath49 , once through @xmath73 and once through @xmath74 . to bound the ratio between @xmath7 and the total associated weight of the vertices , it suffices to bound the ratio between the weight of an edge @xmath75 and the total associated weight of the vertices which are charged to @xmath46 .
assume that @xmath76 , then there are at most two vertices which are charged to @xmath46 and class @xmath49 for all @xmath77 , and no vertex is associated to @xmath46 and class @xmath49 for @xmath78 .
hence , the total associated weight of these vertices is at most @xmath79 and the claim follows since @xmath80 .
it remains to bound @xmath53 with respect to the total associated weight .
@xmath53 is at most the total weight associated with all vertices .
it suffices to show that for every edge @xmath81 the maximum of the associated weights of @xmath74 and @xmath82 is at least the rounded weight of @xmath46 .
suppose that this claim does not hold , then @xmath74 and @xmath82 are not covered by @xmath55 , as otherwise their associated weight would be at least @xmath51 .
hence , when the algorithm considered @xmath46 , we would have added @xmath46 to @xmath55 , contradicting our assumption that @xmath74 and @xmath82 are not covered by @xmath55 .
using the above sequence of lemmas , and recalling that we lose another @xmath31 in the approximation ratio due to disregarding edges of weight at most @xmath35 , we obtain the following inequality : @xmath83 therefore , we establish the following theorem . [ simple_det_theorem ] our simple deterministic algorithm has an approximation ratio of @xmath84 .
this ratio can be optimized to @xmath85 by picking @xmath86 .
the next example demonstrates that the analysis leading to theorem [ simple_det_theorem ] is tight .
let @xmath87 be some large enough integer and @xmath32 be sufficiently small .
consider the instance depicted in figure [ fig : tight - example-8 ] , where @xmath88 consists of a single edge @xmath89 with weight @xmath90 .
for every @xmath91 , the matching @xmath92 consists of exactly two edges @xmath93 and @xmath94 each of weight @xmath95 , and @xmath56 consists of two edges @xmath96 and @xmath97 each of weight @xmath98 .
in addition , there are two edges @xmath99 and @xmath100 whose weight is @xmath101 .
it is easy to see that each @xmath55 is indeed maximal in its own weight class .
given these matchings , our greedy selection rule will output a single edge @xmath89 with total weight @xmath90 ( notice that computing a maximum weight matching in @xmath102 does not help when @xmath103 ) .
moreover , the value of the optimal solution matches our upper bound up to an additive @xmath104 term .
= [ label distance=2pt , fill , inner sep=1.25pt ] = [ fill = white , inner sep=3pt ] ( 2,2 ) node[label = below:@xmath74 ] ( x )
node[weight ] @xmath90 + ( 1,0 ) node[label = below:@xmath82 ] ( y ) ; //in 1/0/1,3/k-1/k ( 0 , ) node[label = above:@xmath107 node[weight ] @xmath108 ( 1 , ) node[label = above:@xmath109 node[weight ] @xmath110 ( x ) ; ( y )
node[weight ] @xmath110 ( 4 , ) node[label = above:@xmath111 node[weight ] @xmath108 ( 5 , ) node[label = above:@xmath112 ; ; ( 0.5,1.75 ) ( 0.5,2.25 ) ; ( 4.5,1.75 ) ( 4.5,2.25 ) ; ( x )
node[weight ] @xmath113 + ( 0.2,1.3 ) node[label = above:@xmath114 ; ( y )
node[weight ] @xmath113 + ( -0.2,1.3 ) node[label = above:@xmath115 ; in what follows , we analyze a randomized variant of the deterministic algorithm which was presented in the previous subsection . in general
, this variant sets the value of @xmath116 to be @xmath117 where @xmath118 is a random variable .
this method is commonly referred to as _
randomized geometric grouping_. formally , let @xmath118 be a continuous random variable which is uniformly distributed on the interval @xmath119 .
we define the weight class @xmath120 , and run the algorithm as in the previous subsection .
note that this algorithm uses only the partition of the edges into classes and not the precise values of their weights .
in addition , we denote by @xmath121 the resulting matching obtained by the algorithm , and by @xmath122 the total associated weight of the vertices , where for a vertex @xmath59 we define its associated weight to be @xmath123 ( i.e. , the minimal value in the interval @xmath124 ) .
we also denote by @xmath125 the value of @xmath53 for this particular @xmath118 . for any fixed value of @xmath118 , inequality ( [ eq1 ] )
immediately implies @xmath126 note that @xmath125 and @xmath127 are random variables , such that for each realization of @xmath118 the above inequality holds .
hence , this inequality holds also for their expected values .
that is , we have established the following lemma where @xmath128}$ ] represents expectation with respect to the random variable @xmath118 .
[ lem2prime ] @xmath129 } \leq ( { 2\gamma\over \gamma-1 } + { \varepsilon})\cdot { \mathrm{e } _ { \delta } [ w(m(\delta ) ) ] } $ ] .
we next lower bound @xmath52 in terms of @xmath129}$ ] .
[ lem:3prime ] @xmath130 } \geq { \mbox{\textsc{opt}}}$ ] .
we will show the corresponding inequality for each edge @xmath131 .
we denote by @xmath132 the rounded weight of @xmath46 for a specific value of @xmath118 .
then , it suffices to show that @xmath133 } \geq w(e)$ ] .
let @xmath134 be an integer , and let @xmath135 be the value that satisfies @xmath136 .
then , for @xmath137 , @xmath138 , and for @xmath139 , @xmath140 , thus the expected rounded weight of @xmath46 over the choices of @xmath118 is @xmath141 } = \int_0^\alpha \gamma^{p+\delta}d\delta+\int_\alpha^1 \gamma^{p-1+\delta}d\delta={1\over \ln \gamma}\cdot \left ( \gamma^p(\gamma^\alpha-1)+\gamma^{p-1}(\gamma-\gamma^\alpha)\right)=w(e)\cdot \left(1-\frac 1{\gamma}\right){1\over \ln \gamma},\ ] ] and the claim follows . combining the above two lemmas
we obtain that the expected weight of the resulting solution is at least @xmath142 .
this approximation ratio is optimized for @xmath143 , where it is roughly @xmath144 .
hence , we have established the following theorem .
the randomized algorithm has an approximation ratio of roughly @xmath145 .
prior to presenting our de - randomization , we slightly modify the randomized algorithm of the previous subsection . in this variation , instead of picking @xmath118 uniformly at random from the interval @xmath119 we pick @xmath146 uniformly at random from the discrete set @xmath147 .
we apply the same method as in the previous section where we replace @xmath118 by @xmath146 .
then , using lemma [ lem2prime ] , we obtain @xmath148 } \leq ( { 2\gamma\over \gamma-1 } + { \varepsilon})\cdot { \mathrm{e } _ { \delta ' } [ w(m(\delta ' ) ) ] } $ ] .
to extend lemma [ lem:3prime ] to this new setting , we note that @xmath146 can be obtained by first picking @xmath118 and then rounding it down to the largest number in @xmath149 which is at most @xmath118 . in this way , we couple the distributions of @xmath118 and @xmath146 .
now consider the rounded weight of an edge @xmath46 in @xmath52 in the two distinct values of @xmath118 and @xmath146 .
the ratio between the two rounded weight is at most @xmath150 .
therefore , we establish that @xmath151 } \geq { \mbox{\textsc{opt}}}$ ] .
therefore , the resulting approximation ratio of the new variation is @xmath152 . by settinf @xmath153 to be large enough ( picking @xmath154 is enough ) , the resulting approximation ratio is bounded by @xmath155 . de - randomizing the new variation in the semi - streaming model is straightforward .
we simply run in parallel all @xmath153 possible outcomes of the algorithm , one for each possible value of @xmath146 , and pick the best solution among the @xmath153 solutions we obtained . since @xmath153 is a constant ( for fixed values of @xmath31 ) , the resulting algorithm is still a semi - streaming algorithm whose performance guarantee is @xmath156 . by scaling @xmath31 prior to applying the algorithm , we establish the following result . for any fixed @xmath32 ,
there is a deterministic one - pass semi - streaming @xmath157-approximation algorithm for the weighted matching problem .
this algorithm processes each input edge in constant time and required @xmath158 time at the end of the input to compute the final output .
in this section , we established the following theorem .
let @xmath162 for some @xmath163 and assume that a deterministic online algorithm achieves a competitive ratio of at most @xmath164 .
we construct an input graph iteratively , and show that after a finite number of steps , the competitive ratio is violated . in the construction of the input ,
all edge weights come from two weight sequences .
the main weight sequence is @xmath165 , and an additional weight function is @xmath166 .
these sequences are defined as follows : the first sequence is defined for @xmath171 only as long as @xmath172 .
as soon as @xmath173 , the sequence stops with @xmath174 , and the length of the sequence @xmath175 is @xmath176 .
we later show that such a value @xmath87 must exist .
let @xmath177 ( and @xmath178 ) . by definition , since @xmath179 , if @xmath173 , then @xmath180 holds as well .
note that @xmath181 for all @xmath182 , by definition , since @xmath183 , but @xmath184 .
in addition , we have the following : @xmath185 this equality holds for @xmath186 since @xmath187 where the first equality holds by definition of @xmath188 , the second equality holds by definition of @xmath189 , and the third one by simple algebra .
in addition , @xmath190 the last equality holds for @xmath191 since @xmath192 where the first equality holds by definition of @xmath188 , the second by definition of @xmath189 , the third by simple algebra , the fourth by definition of @xmath193 and @xmath194 , and the last one by definition of @xmath195 . to better understand our construction
, we advice the reader to consult figure [ fig : preemptive - lowerbound ] .
the input is created in @xmath196 steps . in the initial step ,
two edges @xmath197 and @xmath198 , each of weight @xmath199 , are introduced .
assume that after both edges have arrived , the online algorithm holds the edge @xmath197 .
all future edges either have endpoints which are new vertices , or in the set @xmath200 ( i.e. , they do not contain @xmath201 as an endpoint ) .
an optimal solution keeps @xmath198 .
node[round ] @xmath206 ; + + ( 0,-2 ) node[vertex , label = below:@xmath201 ] ( b1 ) + + ( 0,1 )
node[vertex , label = above:@xmath204 ( x1 ) + + ( 1,0 ) node[vertex , label = above : ] ( a1 ) ; ( a1 ) + ( 0,-1 )
node[vertex , label = below:@xmath207 ; ( a1 ) + ( 1,0 ) node[vertex , label = above:@xmath208 ( a2 ) ; ( a1 ) ( a2 ) ; node[round ] @xmath209 ; + + ( 0,-2 ) node[vertex , label = below:@xmath201 ] ( b1 ) + + ( 0,1 )
node[vertex , label = above:@xmath204 ( x1 ) + + ( 1,0 ) node[vertex , label = above : ] ( a1 ) ; ( a1 ) + ( 0,-1 )
node[vertex , label = below:@xmath207 ; ( a1 ) + ( 1,0 )
node[vertex , label = above : ] ( a2 ) ; ( a2 )
+ ( 0,-1 ) node[vertex , label = below:@xmath210 ; ( a2 ) + ( 1,0 ) node[vertex , label = above:@xmath211 ( a3 ) ; ( a2 ) + ( 2,0 ) node[vertex , label = above:@xmath212 ( c3 ) ; ( a1 ) edge [ decorate , bend right ] ( c3 ) ; node[round ] @xmath213 ; + + ( 0,-2 ) node[vertex , label = below:@xmath201 ] ( b1 ) + + ( 0,1 ) node[vertex , label = above:@xmath204 ( x1 ) + + ( 1,0 ) node[vertex , label = above : ] ( a1 ) ; ( a1 ) + ( 0,-1 ) node[vertex , label = below:@xmath207 ; ( a1 ) + ( 1,0 ) node[vertex , label = above : ] ( a2 ) ; ( a2 )
+ ( 0,-1 ) node[vertex , label = below:@xmath210 ; ( a2 ) + ( 1,0 ) node[vertex , label = above:@xmath211 ( a3 ) ; ( a2 ) + ( 2,0 ) node[vertex , label = above : ] ( c3 ) ; ( a1 ) edge [ bend right ] ( c3 ) ; ( c3 )
+ ( 0,-1 ) node[vertex , label = below:@xmath214 ; ( c3 ) + ( 1,0 )
node[vertex , label = above:@xmath215 ( a4 ) ; ( a2 ) + ( 4,0 ) node[vertex , label = above:@xmath216 ( c4 ) ; ( a1 ) edge [ decorate , bend right ] ( c4 ) ; node[round ] @xmath217 ; + + ( 0,-2 ) node[vertex , label = below:@xmath201 ] ( b1 ) + + ( 0,1 ) node[vertex , label = above:@xmath204 ( x1 ) + + ( 1,0 ) node[vertex , label = above : ] ( a1 ) ; ( a1 )
+ ( 0,-1 ) node[vertex , label = below:@xmath207 ; ( a1 ) + ( 1,0 ) node[vertex , label = above : ] ( a2 ) ; ( a2 ) + ( 0,-1 ) node[vertex , label = below:@xmath210 ; ( a2 ) + ( 1,0 )
node[vertex , label = above:@xmath211 ( a3 ) ; ( a2 ) + ( 2,0 ) node[vertex , label = above : ] ( c3 ) ; ( a1 ) edge [ bend right ] ( c3 ) ; ( c3 ) + ( 0,-1 ) node[vertex , label = below:@xmath214 ; ( c3 ) + ( 1,0 ) node[vertex , label = above:@xmath215 ( a4 ) ; ( a2 ) + ( 4,0 ) node[vertex , label = above : ] ( c4 ) ; ( a1 ) edge [ bend right ] ( c4 ) ; ( c4 ) + ( 0,-1 ) node[vertex , label = below:@xmath218 ; ( c4 ) + ( 1,0 ) node[vertex , label = above:@xmath219 ( a5 ) ; ( a5 ) edge [ decorate ] ( c4 ) ; every future step can be of two distinct types , which will be described later on . among the edges introduced below
, vertices called @xmath220 denote endpoints which occur each on a single edge .
after step @xmath49 , the following invariants are maintained .
the algorithm keeps a single edge denoted by @xmath221 . if @xmath202 , then @xmath222 . if @xmath223 , then this edge can be one of two edges , @xmath224 or @xmath225 . if @xmath222 , then its weight is @xmath175 , and an optimal solution has one edge of each weight @xmath226 .
no future edges will have common endpoints with these @xmath49 edges , except , possibly , with the endpoint @xmath227 of the edge of weight @xmath175 ( the edge of this weight which this optimal solution keeps is always @xmath228 ) .
otherwise , @xmath229 , and its weight is @xmath195 , in which case an optimal solution can have edges of weights @xmath230 , except for one weight @xmath231 for some @xmath232 . this index @xmath233 is used in the definition of the next step , and the properties of the current step .
in addition to these @xmath234 edges , the optimal solution also has the edge @xmath225 .
future edges will have endpoints which are new vertices , or in the set @xmath235 . in the last case
, the vertex @xmath236 is equal to the vertex @xmath237 .
the invariants clearly hold after the first step .
we next define all other steps and show that the invariants hold for each option . if @xmath239 , the last step consists of an edge of weight @xmath240 .
let @xmath241 , if @xmath242 and otherwise @xmath243 .
the new edge is @xmath244 , where @xmath245 is a new vertex .
this edge has a common endpoint with the edge that the algorithm has .
in fact , the algorithm has an edge of weight at least @xmath246 , and thus we assume that it does not preempt it . if the algorithm has an edge of weight @xmath247 , the edge @xmath244 does not have @xmath227 as an endpoint , so adding the new edge to the optimal solution does not require the removal of any edges , and the profit of the optimal solution is @xmath248 . if the algorithm has an edge of weight @xmath249 , the new edge is @xmath250 .
we replace the edge @xmath251 of the optimal solution by the new edge .
in addition , the edge @xmath252 ( where @xmath233 is the index such that the optimal solution before the modification of the current step does not have an edge of weight @xmath231 ) is added to the optimal solution , since the endpoint @xmath253 became free , and the endpoint @xmath220 only has degree 1 . the profit of the optimal solution is @xmath248 again .
recall that @xmath254 , and hence the algorithm earns ( in both cases ) at most @xmath247 .
note also that the optimal solution has value of @xmath248 and if @xmath255 then we can drop the edge of this weight from the optimal solution and get a solution of value @xmath256 .
therefore , we will use @xmath256 as a lower bound on the value of the optimal solution in this case .
thus we will show later that @xmath257 .
[ [ input - construction - step - boldsymboli1-for - boldsymboli1n . ] ] input construction , step @xmath258 , for @xmath259 .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we next show how to construct the edges of step @xmath260 , for the case @xmath261 .
we introduce two new edges of weight @xmath189 .
let @xmath262 , if @xmath222 and otherwise @xmath263 .
the new edges are @xmath264 , and @xmath265 , where @xmath266 and @xmath267 are new vertices .
both these edges have a common endpoint with the edge that the algorithm has , and the algorithm can either preempt the edge it has , in which case we assume ( without loss of generality ) that it now has @xmath265 , or else it keeps the previous edge . if the algorithm keeps the previous edge ,
let @xmath268 , if @xmath222 and otherwise @xmath269 . in this case
a third edge , @xmath270 , which has a weight of @xmath188 , is introduced .
the vertex @xmath271 is new .
there are four cases to consider . in the first case , if the algorithm replaces the edge @xmath224 with the edge @xmath272 , then an optimal solution can add the edge @xmath264 to its edges , since the endpoint @xmath267 is new , and the endpoint @xmath273 was introduced in the previous step , in which the optimal solution obtained the edge @xmath228 . if the algorithm replaces the edge @xmath225 with the edge @xmath274 , an optimal solution can remove the edge @xmath225 from its solution and add the two edges @xmath275 and @xmath276 ( where @xmath233 is the index such that the optimal solution before the modification of the current step does not have an edge of weight @xmath231 ) .
this is possible since the endpoints @xmath267 and @xmath220 do not have other edges , and the endpoints @xmath277 and @xmath236 become free . in the last two cases , the invariants hold . for the remaining two cases
note that if @xmath278 or @xmath279 and the algorithm has a single edge of weight @xmath195 or @xmath175 , respectively , then the optimal solution is strictly positive and the value of the algorithm is non - positive , and hence the resulting approximation ratio in this case is unbounded .
hence , we can assume without loss of generality that if the algorithm has a single edge at the end of step @xmath49 , then its weight is strictly positive .
if the algorithm does not replace the edge @xmath224 with the edge @xmath272 , we show that it must replace it with the edge @xmath280 .
assume that this is not the case .
then the profit of the algorithm is @xmath175 and the optimal solution can omit its edge @xmath228 and add the edges @xmath281 and @xmath282 ( since all these endpoints are introduced in steps @xmath49 and @xmath260 , except for @xmath227 , which becomes free ) .
thus the profit of the optimal algorithm is @xmath283 , while the profit of the online algorithm is @xmath284 .
thus , the algorithm must switch to the edge @xmath265 , and the structure of the optimal solution is according to the invariants .
if the algorithm does not replace the edge @xmath225 with the edge @xmath274 , we show that it must replace it with the edge @xmath285 .
assume that this is not the case .
then the profit of the algorithm is @xmath195 and the optimal solution can omit its edge @xmath225 and add the edges @xmath286 and @xmath287 ( since @xmath277 and @xmath236 become free , and the other two endpoints are introduced in step @xmath260 ) .
thus the profit of the optimal algorithm is @xmath288 , where @xmath289 and @xmath290 , since @xmath291 as @xmath292 , we get that the optimal profit is at least @xmath293 , while the profit of the online algorithm is @xmath294 .
thus , the algorithm must switch to the edge @xmath265 , and the structure of the optimal solution is according to the invariants .
we next define a recursive formula for @xmath295 . by the definition of the sequence @xmath175
, we have @xmath296 we first use this recurrence to show that if @xmath297 then @xmath298 . to see this note that by assumption @xmath299 , hence using the recurrence formula we conclude that @xmath300 that is , @xmath301 which is equivalent to @xmath302 , so @xmath303 , and we conclude that @xmath304 , as we argued .
therefore , it remains to show that there is a value of @xmath196 such that @xmath305 .
to establish this claim , it suffices to show that there is a value of @xmath233 for which @xmath306 ( since @xmath307 ) . to prove this last claim
, we will show that there is a value of @xmath87 such that @xmath308 . finally , to show the existence of such @xmath87
, we will solve the linear homogeneous recurrence formula , and use the explicit form of @xmath309 to show that there is a value of @xmath87 such that @xmath308 . to solve the recurrence formula ( [ rec ] ) ,
we guess solutions of the form @xmath310 for all @xmath87 , and get the following quadratic equation for @xmath74 : @xmath311we solve this quadratic equation and get its solutions @xmath312 note that using @xmath313 , and recalling that @xmath160 is the unique real solution of the equation @xmath161 , we conclude that @xmath314 and hence the two solutions are complex numbers whose imaginary parts are not zero . since we got two distinct solutions of @xmath74 , it is known that the recurrence formula ( [ rec ] ) is solved by a formula of the form @xmath315 where @xmath316 and @xmath317 are constants .
we find the value of @xmath316 and @xmath317 using the conditions @xmath178 and @xmath318 .
so we get the following set of two equations : @xmath319 ( corresponding to @xmath178 ) , and @xmath320 ( corresponding to @xmath318 ) . from the first equation
we conclude that @xmath321 , and using this we obtain @xmath322 .
hence , the closed form solution of @xmath323 for values of @xmath324 is as follows .
@xmath325 we use the notation @xmath326 , and let @xmath327 . as noted above @xmath314 , and hence @xmath328 is a real number .
we also define @xmath329 and @xmath330 such that @xmath331 , and also @xmath332 , then we get the following formula for @xmath323 .
@xmath333 note that @xmath334 for all @xmath233 , and hence to show that the sequence @xmath335 changes its sign as we required , it suffices to show that the sequence @xmath336 changes its sign , but this last claim holds because @xmath337 ( as the solutions @xmath338 and @xmath339 are not real numbers ) .
hence , the claim follows .
n. bansal , n. buchbinder , a. gupta , and j. naor .
an @xmath340-competitive algorithm for metric bipartite matching . in _ proceedings of the 15th annual european symposium on algorithms _ , pages 522533 , 2007 . | we study the maximum weight matching problem in the semi - streaming model , and improve on the currently best one - pass algorithm due to zelke ( proc .
stacs 08 , pages 669680 ) by devising a deterministic approach whose performance guarantee is @xmath0 . in addition , we study _ preemptive _ online algorithms , a sub - class of one - pass algorithms where we are only allowed to maintain a feasible matching in memory at any point in time .
all known results prior to zelke s belong to this sub - class .
we provide a lower bound of @xmath1 on the competitive ratio of any such deterministic algorithm , and hence show that future improvements will have to store in memory a set of edges which is not necessarily a feasible matching . |
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.