Interpolating non-Hermitian universality classes A and AI † : eigenvalue density and transition regime Preprint: APS/123-QED

Mark J. Crumpton 1,2, Francesco Mezzadri 1 · 2026

Read Interpolating non-Hermitian universality classes A and AI † : eigenvalue density and transition regime Preprint: APS/123-QED by Mark J. Crumpton 1,2, Francesco Mezzadri 1 (2026) 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-2606-03447.txt.

Introduction

Crumpton and Mezzadri study random matrices that blend two non-Hermitian classes. One end of the blend is the complex Ginibre class, labeled A. The other end is the complex symmetric class, labeled AI dagger. A parameter sigma slides between them. The team uses the new Kac-Rice formalism for such matrices.

They derive the joint law of one eigenvalue and its right eigenvector at finite size N. The result holds for every sigma between zero and one. The eigenvalue density is then worked out, at finite N and in the large limit. In the bulk, the standard circular law returns for all sigma. At the edge, fixed sigma still gives the class A density. A new regime appears when sigma nears one at a set rate. There the edge density blends the two known results smoothly. Numeric tests back the claim that this blend is universal.

Catalog record

CREATORMark J. Crumpton 1,2, Francesco Mezzadri 1
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE47 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2606.03447

Browse the full library · About the Tower of Babel

Full text

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

Interpolating non-Hermitian universality classes A and AI † : eigenvalue density and transition regime Preprint: APS/123-QED Mark J. Crumpton 1,2, Francesco Mezzadri 1 Abstract We employ the recently developed Kac-Rice formalism for non-Hermitian random matrices to derive the joint distribution of an eigenvalue and its associated normalised right eigenvector in a Gaussian ensemble that interpolates between complex Ginibre (Class A) and complex symmetric matrices (Class AI † ). This distribution is valid at finite matrix size, N N , for any value of the interpolation parameter σ ∈ [ 0 , 1 ] \sigma\in[0,1] , with 0 0 and 1 1 corresponding to classes A and AI † respectively. The marginal distribution for the density of the eigenvalues is derived at finite N N and then considered asymptotically as N → ∞ N\to\infty . When considering bulk eigenvalues, we recover the standard circular law for all σ \sigma . Furthermore, for edge eigenvalues we find that for fixed σ \sigma , the eigenvalues follow the edge density associated with matrices in Class A. However, a transitional regime is discovered for the interpolation parameter being scaled as σ = 1 − κ N − 1 / 2 \sigma=1-\kappa N^{-1/2} , where new edge behaviour is observed for the density of eigenvalues - smoothly interpolating two previously known results. This transitional regime and the associated density of eigenvalues is conjectured to be universal for non-Gaussian matrices and we provide numerical evidence in support of this. 1 Introduction We introduce a new random matrix ensemble that models the breaking of time reversal symmetry (TRS) in non-Hermitian matrices. This ensemble is constructed such that it interpolates between two non-Hermitian universality classes, represented by complex Ginibre matrices (no TRS) and complex symmetric matrices (positive TRS). Specifically, we consider a Gaussian ensemble of N\times N random matrices constructed as X_{N}\equiv\sqrt{\frac{1+\tau}{2}}G_{N}+\sqrt{\frac{1-\tau}{2}}G_{N}^{T}\ , (1.1) where G_{N} is a Ginibre matrix, T denotes the matrix transpose and \tau\in[0,1] is an interpolation parameter, such that \tau=1(0) yields complex Ginibre (symmetric) matrices. It is useful to view the interpolation parameter in terms \sigma\equiv\sqrt{1-\tau^{2}} , which we call the symmetry parameter - see Section 2 for more details on this ensemble. Within this ensemble, we will present a result for the JPDF of an eigenvalue and the associated eigenvector, which is valid at finite N - Theorem 1 . From this JPDF, we will obtain the associated density of eigenvalues at finite N , see Corollary 2 , then assess its asymptotic limit as N\to\infty . This analysis shows that the leading order contribution to the density in the bulk is the same across the entire interpolating ensemble, which simply yields the celebrated circular law. On the other hand, in the edge regime we will prove that, for fixed values of the symmetry parameter, the density always follows the behaviour associated with Ginibre matrices (Corollary 4 ). We refer to this as the strong asymmetry regime. Whereas, for values of the symmetry parameter asymptotically close to unity, we will present a result that smoothly interpolates the well-known edge density in the GinUE and the recently derived edge density in the SymOE AFS25 . This proves the presence of a previously unknown regime of weak asymmetry , where time reversal symmetry is broken in non-Hermitian matrices. We conjecture that this regime is universal beyond Gaussian matrices and we provide numerical evidence in support of this. These new results constitute the first analytical description of the transition between non-Hermitian universality classes and also serve as the non-Hermitian extension of the works of Mehta & Pandey PM83 ; MP83 , who studied the breaking of TRS in Hermitian matrices in the 1980s. Background : The question of universality has been at the heart of Random Matrix Theory (RMT) for many decades now, see KuijlaarsUniversality for a review of results. Central to this topic is finding a way of classifying matrices according to mathematical/ physical symmetries and assigning such matrices to universality classes. Celebrated studies in this direction include the early works of Freeman Dyson, who classified closed quantum systems according to their symmetry under time reversal, using real symmetric, complex Hermitian and quaternion self-dual matrices DysonThreefold - belonging to the well-known GOE, GUE and GSE respectively. This so called threefold way was later generalised by Altland & Zirnbauer to the tenfold way of classifying Hermitian matrices, now applicable to a wider range of physical systems AZ97 . However, strikingly from an RMT perspective, it turns out that for the ten different symmetry classes, the bulk spectral statistics fall into the same three universality classes associated with the threefold way, with different statistics at the edge KW17 . The spectral statistics of Hermitian matrices in the different universality classes are well-established, with a range of important results available in Wigner57Surm ; MehtaBook ; TW ; Soshnikov99 ; PS1997 ; TaoVuUniversality ; Johansson12 . On the other hand, the associated picture of universality classes for non-Hermitian matrices, which describe open quantum mechanical systems GHS88 ; KS , is not as well-analytically developed, but has received a recent boost in attention, partly thanks to the results of KSUS19 ; HKKU20 . In KSUS19 , the authors establish that for non-Hermitian matrices there are 38 possible symmetry classes according to which matrices and quantum systems can be classified. Then, through intensive numerical simulation of 9 of these 38 symmetry classes, a conjecture is proposed in HKKU20 that only three types of bulk statistics exist for the eigenvalues of non-Hermitian matrices. This conjecture was also recently extended to include edge behaviour - see the works of Akemann et al. AAKP25 , which also contains a thorough and illuminating discussion of the topics alluded to here. It should be noted at this point that these works build upon earlier theoretical findings, where a classification of non-Hermitian matrices based on symmetry properties was proposed by Bernard & LeClair BL02 and then was given associated matrix representations in the work of Magnea Magnea08 . Consequently, it was proposed in HKKU20 that these 3 non-Hermitian universality classes can be represented using matrices from the complex Ginibre, the complex symmetric and the complex self-dual ensembles, where classes as denoted as A, AI † and AII † respectively - with the naming system due to Bernard & LeClair BL02 . Matrices within these classes are distinguished according to their symmetry under time-reversal. Specifically, matrices in Class A have no time-reversal symmetry, whereas matrices in Classes AI † and AII † are invariant under time-reversal with sign +1 and -1 respectively. Analytic results for spectral correlations are well-established in the complex Ginibre ensemble, originally studied in Ginibre , which have been demonstrated to hold for a wider class of non-Hermitian Wigner matrices, up to a four moment matching condition, in the work of Tao & Vu TaoVu2015 . Furthermore, this work has been generalised to prove universality of the correlation kernels at the edge and in the bulk in CES21 ; MO24 respectively. Whereas, in comparison, associated results in classes AI † and AII † are much less developed. For example, there currently does not exist a closed formula for the density of eigenvalues in class AII † and the corresponding equation was only very recently found

Read the complete text