Searcharxiv⌕ Search

arXiv subjects

Florent Benaych-Georges

Publications and source records attributed to Florent Benaych-Georges.

52 records · Page 3Linked to original sources
arXiv2011

Rectangular R-transform as the limit of rectangular spherical integrals

In this paper, we connect rectangular free probability theory and spherical integrals. In this way, we prove the analogue, for rectangular or square non-Hermitian matrices, of a result that Guionnet and Maida proved for Hermitian matrices in 2005. More specifically, we study the limit, as $n,m$ tend to infinity, of the logarithm (divided by $n$) of the expectation of $\exp[\sqrt{nm}θX_n]$, where $X_n$ is the real part of an entry of $U_n M_n V_m$, $θ$ is a real number, $M_n$ is a certain $n\times m$ deterministic matrix and $U_n, V_m$ are independent Haar-distributed orthogonal or unitary matrices with respective sizes $n\times n$, $m\times m$. We prove that when the singular law of $M_n$ converges to a probability measure $μ$, for $θ$ small enough, this limit actually exists and can be expressed with the rectangular R-transform of $μ$. This gives an interpretation of this transform, which linearizes the rectangular free convolution, as the limit of a sequence of log-Laplace transforms.

math.PR↗
arXiv2010

The eigenvalues and eigenvectors of finite, low rank perturbations of large random matrices

We consider the eigenvalues and eigenvectors of finite, low rank perturbations of random matrices. Specifically, we prove almost sure convergence of the extreme eigenvalues and appropriate projections of the corresponding eigenvectors of the perturbed matrix for additive and multiplicative perturbation models. The limiting non-random value is shown to depend explicitly on the limiting eigenvalue distribution of the unperturbed random matrix and the assumed perturbation model via integral transforms that correspond to very well known objects in free probability theory that linearize non-commutative free additive and multiplicative convolution. Furthermore, we uncover a phase transition phenomenon whereby the large matrix limit of the extreme eigenvalues of the perturbed matrix differs from that of the original matrix if and only if the eigenvalues of the perturbing matrix are above a certain critical threshold. Square root decay of the eigenvalue density at the edge is sufficient to ensure that this threshold is finite. This critical threshold is intimately related to the same aforementioned integral transforms and our proof techniques bring this connection and the origin of the phase transition into focus. Consequently, our results extend the class of `spiked' random matrix models about which such predictions (called the BBP phase transition) can be made well beyond the Wigner, Wishart and Jacobi random ensembles found in the literature. We examine the impact of this eigenvalue phase transition on the associated eigenvectors and observe an analogous phase transition in the eigenvectors. Various extensions of our results to the problem of non-extreme eigenvalues are discussed.

math.PR↗
arXiv2010

Cycles of free words in several independent random permutations with restricted cycle lengths

In this text, we consider random permutations which can be written as free words in several independent random permutations: firstly, we fix a non trivial word $w$ in letters $g_1,g_1^{-1},..., g_k,g_k^{-1}$, secondly, for all $n$, we introduce a $k$-tuple $s_1(n),..., s_k(n)$ of independent random permutations of $\{1,..., n\}$, and the random permutation $σ_n$ we are going to consider is the one obtained by replacing each letter $g_i$ in $w$ by $s_i(n)$. For example, for $w=g_1g_2g_3g_2^{-1}$, $σ_n=s_1(n)\circ s_2(n)\circ s_3(n)\circ s_2(n)^{-1}$. Moreover, we restrict the set of possible lengths of the cycles of the $s_i(n)$'s: we fix sets $A_1,..., A_k$ of positive integers and suppose that for all $n$, for all $i$, $s_i(n)$ is uniformly distributed on the set of permutations of $\{1,..., n\}$ which have all their cycle lengths in $A_i$. For all positive integer $l$, we are going to give asymptotics, as $n$ goes to infinity, on the number $N_l(σ_n)$ of cycles of length $l$ of $σ_n$. We shall also consider the joint distribution of the random vectors $(N_1(σ_n),..., N_l(σ_n))$. We first prove that the order of $w$ in a certain quotient of the free group with generators $g_1,..., g_k$ determines the rate of growth of the random variables $N_l(σ_n)$ as $n$ goes to infinity. We also prove that in many cases, the distribution of $N_l(σ_n)$ converges to a Poisson law with parameter $1/l$ and that the random variables $N_1(σ_n),N_2(σ_n), ...$ are asymptotically independent. We notice the surprising fact that from this point of view, many things happen as if $σ_n$ were uniformly distributed on the $n$-th symmetric group.

math.PR↗
arXiv2009

On a surprising relation between the Marchenko-Pastur law, rectangular and square free convolutions

n this paper, we prove a result linking the square and the rectangular R-transforms, the consequence of which is a surprising relation between the square and rectangular versions the free additive convolutions, involving the Marchenko-Pastur law. Consequences on random matrices, on infinite divisibility and on the arithmetics of the square versions of the free additive and multiplicative convolutions are given.

math.PR↗
arXiv2009

A matrix interpolation between classical and free max operations: I. The univariate case

Recently, Ben Arous and Voiculescu considered taking the maximum of two free random variables and brought to light a deep analogy with the operation of taking the maximum of two independent random variables. We present here a new insight on this analogy: its concrete realization based on random matrices giving an interpolation between classical and free settings.

math.PR↗
arXiv2009

A permutation model for free random variables and its classical analogue

In this paper, we generalize a permutation model for free random variables which was first proposed by Biane in \cite{biane}. We also construct its classical probability analogue, by replacing the group of permutations with the group of subsets of a finite set endowed with the symmetric difference operation. These constructions provide new discrete approximations of the respective free and classical Wiener chaos. As a consequence, we obtain explicit examples of non random matrices which are asymptotically free or independent. The moments and the free (resp. classical) cumulants of the limiting distributions are expressed in terms of a special subset of (noncrossing) pairings. At the end of the paper we present some combinatorial applications of our results.

math.PR↗
arXiv2009

Cycles of random permutations with restricted cycle lengths

We prove some general results about the asymptotics of the distribution of the number of cycles of given length of a random permutation whose distribution is invariant under conjugation. These results were first established to be applied in a forthcoming paper (Cycles of free words in several random permutations with restricted cycles lengths), where we prove results about cycles of random permutations which can be written as free words in several independent random permutations. However, we also apply them here to prove asymptotic results about random permutations with restricted cycle lengths. More specifically, for $A$ a set of positive integers, we consider a random permutation chosen uniformly among the permutations of $\{1,..., n\}$ which have all their cycle lengths in $A$, and then let $n$ tend to infinity. Improving slightly a recent result of Yakymiv (Random A-Permutations: Convergence to a Poisson Process), we prove that under a general hypothesis on $A$, the numbers of cycles with fixed lengths of this random permutation are asymptotically independent and distributed according to Poisson distributions. In the case where $A$ is finite, we prove that the behavior of these random variables is completely different: cycles with length $\max A$ are predominant.

math.PR↗
arXiv2008

A continuous semigroup of notions of independence between the classical and the free one

In this paper, we investigate a continuous family of notions of independence which interpolates between the classical and free ones for non-commutative random variables. These notions are related to the liberation process introduced by D. Voiculescu. To each notion of independence correspond new convolutions of probability measures, for which we establish formulae and of which we compute simple examples. We prove that there exists no reasonable analogue of classical and free cumulants associated to these notions of independence.

math.PR↗
arXiv2008

On a surprising relation between rectangular and square free convolutions

Debbah and Ryan have recently proved a result about the limit empirical singular distribution of the sum of two rectangular random matrices whose dimensions tend to infinity. In this paper, we reformulate it in terms of the rectangular free convolution introduced in a previous paper and then we give a new, shorter, proof of this result under weaker hypothesis: we do not suppose the \pro measure in question in this result to be compactly supported anymore. At last, we discuss the inclusion of this result in the family of relations between rectangular and square random matrices.

math.PR↗
arXiv2008

Regularization by free additive convolution, square and rectangular cases

The free convolution is the binary operation on the set of probability measures on the real line which allows to deduce, from the individual spectral distributions, the spectral distribution of a sum of independent unitarily invariant square random matrices or of a sum of free operators in a non commutative probability space. In the same way, the rectangular free convolution allows to deduce, from the individual singular distributions, the singular distribution of a sum of independent unitarily invariant rectangular random matrices. In this paper, we consider the regularization properties of these free convolutions on the whole real line. More specifically, we try to find continuous semigroups $(μ_t)$ of probability measures such that $μ_0$ is the Dirac mass at zero and such that for all positive $t$ and all probability measure $ν$, the free convolution of $μ_t$ with $ν$ (or, in the rectangular context, the rectangular free convolution of $μ_t$ with $ν$) is absolutely continuous with respect to the Lebesgue measure, with a positive analytic density on the whole real line. In the square case, we prove that in semigroups satisfying this property, no measure can have a finite second moment, and we give a sufficient condition on semigroups to satisfy this property, with examples. In the rectangular case, we prove that in most cases, for $μ$ in a continuous rectangular-convolution-semigroup, the rectangular convolution of $μ$ with $ν$ either has an atom at the origin or doesn't put any mass in a neighborhood of the origin, thus the expected property does not hold. However, we give sufficient conditions for analyticity of the density of the rectangular convolution of $μ$ with $ν$ except on a negligible set of points, as well as existence and continuity of a density everywhere.

math.PR↗
arXiv2008

Rectangular random matrices. Related convolution

We characterize asymptotic collective behaviour of rectangular random matrices, the sizes of which tend to infinity at different rates: when embedded in a space of larger square matrices, independent rectangular random matrices are asymtotically free with amalgamation over a subalgebra. Therefore we can define a "rectangular free convolution", linearized by cumulants and by an analytic integral transform, called the "rectangular R-transform".

math.OA↗
arXiv2006

Infinitely divisible distributions for rectangular free convolution: classification and matricial interpretation

In a previous paper (called "Rectangular random matrices. Related covolution"), we defined, for $λ\in [0,1]$, the rectangular free convolution with ratio $λ$. Here, we investigate the related notion of infinite divisiblity, which happens to be closely related to the classical infinite divisibility: there exists a bijection between the set of classical symmetric infinitely divisible distributions and the set of distributions infinitely divisible with respect to this convolution, which preserves limit theorems. We give an interpretation of this correspondance in term of random matrices: we construct distributions on sets of complex rectangular matrices which give rise to random matrices with singular laws (i.e. uniform distributions on their singular values) going from the symmetric classical infinitely divisible distributions to their images by the previously mentioned bijection when the dimensions go from one to infinity in a ratio $λ$.

math.OA↗
arXiv2006

Taylor expansions of R-transforms, application to supports and moments

We prove that a probability measure on the real line has a moment of order p (even integer), if and only if its R-transform admits a Taylor expansion with p terms. We also prove a weaker version of this result when p is odd. Then, we apply this to prove that a probability measure whose R-transform extends analytically to a ball with center zero is compactly supported, and that a free infinitely divisible distribution has a moment of order p even, if and only if its Levy measure does so. We also prove a weaker version of the last result when p is odd.

math.PR↗
arXiv2005

Rectangular random matrices, related free entropy and free Fisher's information

We prove that independent rectangular random matrices, when embedded in a space of larger square matrices, are asymptotically free with amalgamation over a commutative finite dimensional subalgebra $D$ (under an hypothesis of unitary invariance). Then we consider elements of a finite von Neumann algebra containing $D$, which have kernel and range projection in $D$. We associate them a free entropy with the microstates approach, and a free Fisher's information with the conjugate variables approach. Both give rise to optimization problems whose solutions involve freeness with amalgamation over $D$. It could be a first proposition for the study of operators between different Hilbert spaces with the tools of free probability. As an application, we prove a result of freeness with amalgamation between the two parts of the polar decomposition of $R$-diagonal elements with non trivial kernel.

math.OA↗
arXiv2005

Classical and free infinitely divisible distributions and random matrices

We construct a random matrix model for the bijection Ψbetween clas- sical and free infinitely divisible distributions: for every d\geq1, we associate in a quite natural way to each *-infinitely divisible distribution μa distribution P_d^μ on the space of d\times d Hermitian matrices such that P_d^μP_d^ν=P_d^{μ*ν}. The spectral distribution of a random matrix with distribution P_d^μ converges in probability to Ψ(μ) when d tends to +\infty. It gives, among other things, a new proof of the almost sure convergence of the spectral distribution of a matrix of the GUE and a projection model for the Marchenko-Pastur distribution. In an analogous way, for every d\geq1, we associate to each *-infinitely divisible distribution μ, a distribution L_d^μ on the space of complex (non-Hermitian) d\times d random matrices. If μis symmetric, the symmetrization of the spectral distribution of |M_d|, when M_d is L_d^μ-distributed, converges in probability to Ψ(μ).

math.PR↗