arXiv:2609.35230
Exact probability distributions of complex spacing ratios in non-Hermitian random matrices
Kohei Kawabata
Abstract
The complex spacing ratio is the complex displacement from a reference eigenvalue to its nearest neighbor divided by the corresponding displacement to its next-to-nearest neighbor. Its statistics provide a useful diagnostic of spectral correlations and nonintegrability in open quantum systems. Here, starting from the exact joint eigenvalue probability densities, we derive finite-$N$ complex-spacing-ratio distributions for the Gaussian ensembles of non-Hermitian random matrices in classes AI$^†$ and AII$^†$, realized by complex symmetric and complex self-dual random matrices, respectively. In class AII$^†$, we obtain an exact algebraic expression for arbitrary $N$ and explicitly evaluate the distributions and representative moments for $N=3, 4, 5, 6$. In class AI$^†$, although the joint density retains a noncompact integral over nonunitary eigenvector degrees of freedom, we analytically derive a normalized one-dimensional integral representation for $N=3$ and determine the asymptotic behavior, including a logarithmic correction to the cubic level repulsion and a nonanalytic contribution to the angular density. We further confirm these analytical results through direct numerical diagonalization of non-Hermitian random matrices.
AI Summary
Complex spacing ratios diagnose correlations in non-Hermitian spectra without explicit unfolding, but analytical distributions are largely unavailable for the transpose-symmetry classes AI$^\dagger$ (complex symmetric matrices) and AII$^\dagger$ (complex self-dual matrices). This paper derives finite-$N$ distributions from recently obtained exact joint eigenvalue densities, retaining the constraints that select nearest and next-to-nearest neighbors. Here $N$ counts distinct eigenvalues, with Kramers pairs counted once in AII$^\dagger$. For AII$^\dagger$, polynomial expansion and Gaussian integration yield an exact algebraic construction for arbitrary $N$, explicitly evaluated through $N=6$. Numerical diagonalization confirms the distributions and moments and reveals nonmonotonic size dependence: the mean ratio modulus increases from approximately $0.7055$ at $N=3$ to $0.7588$ at $N=6$, then approaches approximately $0.749$ for large matrices. For AI$^\dagger$ at $N=3$, the distribution reduces to a normalized one-dimensional integral involving a complete elliptic integral. It gives $p_r(r)=(256/9)r^3\log(1/r)+O(r^3)$ and a nonanalytic $|\theta|^3$ term in the angular density, implying $\langle\cos(n\theta)\rangle\sim-3.28908/n^4$. Larger-$N$ AI$^\dagger$ statistics are sampled from the exact eigenvalue–eigenvector measure and checked against diagonalization. The formulas condition on an eigenvalue at the spectral origin; at finite $N$, they need not describe a generic reference eigenvalue. Explicit AII$^\dagger$ polynomial construction becomes computationally demanding as $N$ grows, while AI$^\dagger$ retains a noncompact eigenvector integral. Universal large-$N$ distributions and conditions for ensemble independence remain open.
Error found: The abstract describes exact finite-N spacing-ratio distributions without stating that they are conditioned on a reference eigenvalue at the spectral origin, although the paper explicitly distinguishes these from distributions for a generally selected reference eigenvalue.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 16 pages, 4 figures, 3 tables