Corrections to Classical Matrix Ensemble Moments, Non-Crossing Annular Pairings, and Ribbon Graphs

Anas A. Rahman, Daniel Munoz George, and James A. Mingo · 2025

Read Corrections to Classical Matrix Ensemble Moments, Non-Crossing Annular Pairings, and Ribbon Graphs by Anas A. Rahman, Daniel Munoz George, and James A. Mingo (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-22308.txt.

Introduction

Anas A. Rahman, Daniel Munoz George, and James A. Mingo draw pictures to correct the statistics of random matrices. The classical ensembles of random matrix theory have spectral moments, averages of powers of eigenvalues, that expand in corrections shrinking with matrix size. Each correction term hides a combinatorial pattern that combinatorics can expose.

The team establishes a bijection, an exact pairing, between ribbon graphs drawn on the real projective plane and non-crossing annular pairings, the arc diagrams that organize free probability calculations. That pairing explains the first correction term for the orthogonal and Laguerre orthogonal ensembles. They build the analogous objects for the next correction and for the unitary families as well. Ribbon graphs, maps drawn on surfaces, turn out to be the right bookkeeping device for how finite matrices deviate from their infinite-size limits.

Catalog record

CREATORAnas A. Rahman, Daniel Munoz George, and James A. Mingo
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE83 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-22308

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Corrections to Classical Matrix Ensemble Moments, Non-Crossing Annular Pairings, and Ribbon Graphs Anas A. Rahman, Daniel Munoz George, and James A. Mingo Abstract We elucidate a bijection between ribbon graphs on the real projective plane and non-crossing annular pairings that relate to the 1 / N 1/N correction term of the GOE and LOE spectral moments. We also derive analogous objects for the 1 / N 2 1/N^{2} correction terms of said moments and their equivalents for the GUE and LUE. 1 Introduction Since its inception, random matrix theory has steadily gained attention due to its unreasonable effectiveness in studying heavy atom spectra Wig55 ; Dys62b ; Wig67 , wireless communication channels TV04 , random quantum states CN16 , log-gases Fo10 , and quantum transport Be97 (to give a non-exhaustive list), with a recent surge in interest stemming from advances in the fields of high-dimensional data and machine learning. The most famous result in the field is undoubtedly due to the pioneering work Wig55 ; Wig58 of Wigner, who showed that in the limit of large matrix size and with appropriate scaling, the limiting distribution of the eigenvalues of so-called Wigner matrices is the celebrated Wigner semicircle law , \rho(x)=\frac{1}{2\pi}\sqrt{4-x^{2}}. (1) To be precise, let X be a real symmetric or complex Hermitian N\times N random matrix with independently distributed entries X_{ij} ( 1\leq i\leq j\leq N ) of mean zero, identical variance for i=j , variance \mathbb{E}|X_{ij}|^{2}=1 for i\neq j , and all other moments having upper bound independent of i,j . Then, X is said to be a Wigner matrix and it is precisely the eigenvalues of \widetilde{X}:=\frac{1}{\sqrt{N}}X that follow the above semicircular distribution in the large N limit. The seminal result of Wigner recounted above was obtained through the method of moments , where one uses the fact that the normalised moments \widetilde{m}_{n}=\int_{\mathbb{R}}x^{n}\rho_{N}(x)\,\mathrm{d}x\quad(n=0,1,2,\ldots) of the density \rho_{N}(x) of the eigenvalues of \widetilde{X} fully characterise \rho_{N}(x) when Carleman’s condition ( Wal48 , Sec. 88) is satisfied. Wigner studied said moments by noting that they are are given by the averages with respect to the distribution of \widetilde{X} , \widetilde{m}_{n}=\frac{1}{N}\mathbb{E}\Tr\widetilde{X}^{n}, which he then showed via combinatorial arguments to be given in the large N limit by the (n/2)\textsuperscript{th} Catalan number when n is even (the moments vanish for odd n by symmetry) — these characterise the Wigner semicircle law, hence \rho_{N}(x) converges weakly to \rho(x) . Since Wigner’s application of the method of moments, moments of random matrices of the form \mathbb{E}\Tr M^{n} have been interpreted as counts of graphs in a variety of ways, with the physics community largely using diagrammatic theories à la Feynman to simplify matrix integrals Hoo74 ; BIPZ78 ; BIZ80 ; YK83 ; VWZ84 , while combinatorialists and algebraic geometers have dually used matrix integrals to enumerate graphs HZ86 ; Kon92 ; GJ97 ; GHJ01 . In particular, when matrix entries are simple sums and products of Gaussian variables, one has access to the Isserlis–Wick theorem Iss18 ; Wic50 , which can be used to show that the moments \mathbb{E}\Tr M^{n} are given by genus expansions in 1/N whose coefficients are counts of ribbon graphs. We review this connection in § 1.1 for the Gaussian and Laguerre orthogonal and unitary ensembles. Moving onto another perspective, it has been shown in recent years that the infinitesimal moments, equivalently 1/N correction to the spectral moments, of the Gaussian orthogonal ensemble (GOE) enumerates a certain class of non-crossing annular pairings mingo2019 . In this paper we show this can be extended to the N^{-2} terms in the genus expansion of the GOE and the GUE, despite the difficulty in enumeration of genus 1 permutations (see Cori and Hetyei ch14 ). We expect this extension to be relevant to the problem of finding the correction to Fevrier’s definition of higher order infinitesimal freeness; according to the definition of fev12 , independent GUEs are not asymptotically infinitesimally free of higher order as it can be seen from the tables of Harer and Zagier HZ86 that \kappa_{4}^{\prime\prime}=1 , whereas infinitesimal freeness of second order requires that \kappa_{4}^{\prime\prime}=0 . Moreover, the recursion of HZ86 still lacks an elementary topological rationale. There is now evidence that higher order freeness collins2007second is necessary for both said rationale and the correction to Fevrier’s higher order infinitesimal freeness, and this theory is based on non-crossing permutations on a multi-annulus. These are exactly the objects of consideration in the present paper. The goal of this paper is to highlight the relationship between ribbon graphs and non-crossing annular pairings in the context of low order corrections to the spectral moments of the Gaussian and Laguerre orthogonal and unitary ensembles. Thus, in the remainder of this introduction, we review the relevant theory of ribbon graphs and non-crossing annular pairings. Then, in § 2.1 , we give a bijection between the ribbon graphs and non-crossing annular pairings enumerated by the infinitesimal moments of the GOE, while in § 2.2 and § 2.3 , we derive non-crossing annular pairings corresponding to the 1/N^{2} corrections to the spectral moments of the Gaussian unitary ensemble (GUE) and GOE, respectively. We give analogous arguments for the corresponding Laguerre ensembles (LOE and LUE) in Section 3 . From the topological viewpoint, one motivation for studying these new classes of non-crossing annular pairings is that they give a tractable, planar classification of the corresponding ribbon graphs. In particular, we give planar descriptions of the ribbon graphs that can be embedded into the torus, real projective plane, and Klein bottle. These ribbon graphs are special because, by the classification theorem for closed surfaces, every other non-planar ribbon graph involved in the computation of the moments of interest can be constructed by gluing a finite number of torii to them — we leave it to future work to understand how gluing torii to ribbon graphs should be interpreted in the setting of non-crossing annular pairings. 1.1 Matrix Moments and Ribbon Graphs We begin by introducing the random matrix ensembles studied in this paper, which are constructed from the real and complex Ginibre ensembles. Definition 1 (GOE, GUE, LOE, and LUE) The M\times N real and complex Ginibre ensembles Gin65 are respectively the sets of M\times N real ( \beta=1 ) and complex ( \beta=2 ) matrices with probability density functions (p.d.f.s) P(G)=\pi^{-MN\beta/2}\exp\left(-\Tr\,G^{\dagger}G\right)=\prod_{i=1}^{M}\prod_{j=1}^{N}\pi^{-\beta/2}\exp\left(-|G_{i,j}|^{2}\right) defined with respect to the Lebesgue measure \mathrm{d}G=\prod_{i=1}^{M}\prod_{j=1}^{N}\prod_{s=1}^{\beta}\mathrm{d}G_{i,j}^{(s)}, where we write G_{i,j}^{(1)}=\mathrm{Re}(G_{i,j}) and G_{i,j}^{(2)}=\mathrm{Im}(G_{i,j}) . We say that • H=\tfrac{1}{2}(G^{\intercal}+G) belongs to the N\times N Gaussian orthogonal ensemble when G is drawn from the N\times N real Ginibre ensemble; • H=\tfrac{1}{2}(G^{\dagger}+G) belongs to the N\times N Gaussian unitary ensemble when G is drawn from the N\times N complex Ginibre ensemble; • W=G^{\intercal}G belongs to the (M,N) Laguerre orthogonal ensemble when G is drawn from the M\times N real Ginibre ensemble; • W=G^{\dagger}G belongs to the (M,N) Laguerre unitary ensemble when G is drawn from the M\times N complex Ginibre ensemble. Remark 1 The entries of real Ginibre matrices are independent centred normal variables of variance 1/2 . For G a complex Ginibre matrix, the entries are independent complex normal variables such that \mathbb{E}[G_{i,j}^{2}]=0 , but \mathbb{E}[G_{i,j}\overline{G}_{i,j}]=1 . The key property of the ensembles defined above is that the entries of Ginibre matrices are Gaussian, so

Read the complete text