Random matrix theory of integrability-to-chaos transition

Abstract: The statistics of gaps between quantum energy levels is a hallmark criterion in quantum chaos and quantum integrability studies. The relevant distributions corresponding to exactly integrable vs. fully chaotic systems are universal and described by the Poisson vs. Wigner-Dyson curves. In the transitional regime between integrability and chaos, the distributions are much less universal and have not been understood quantitatively until now. We point out that the relevant statistics that controls these distributions is that of the matrix elements of the nonintegrable perturbation Hamiltonian in the energy eigenbasis of the unperturbed integrable system. With this insight, we formulate a simple random matrix ensemble that correctly reproduces the level spacing distributions in a variety of test systems. For the distribution of matrix elements appearing in our construction, we furthermore discover surprising universal features: across a variety of physical systems with diverse degrees of freedom, these distributions are dominated by simple power laws.

• The paper introduces a random matrix model that captures the intermediate spectral statistics between Poisson and Wigner-Dyson limits.
• It employs the full distribution of offdiagonal matrix elements from chaotic perturbations to predict the crossover behavior accurately.
• Quantitative validations on spin chains and quantum resonant systems demonstrate universality and experimental relevance.

Random Matrix Theory Description of the Integrability-to-Chaos Transition

Introduction

The study of quantum systems transitioning from integrability to chaos has centered on spectral properties, particularly the statistics of energy level spacings. It is well-established that the level spacings of integrable quantum systems are uncorrelated and follow Poisson statistics, while chaotic quantum systems exhibit level repulsion described by Wigner-Dyson statistics characteristic of random matrix ensembles. However, the intermediate regime—where an integrable Hamiltonian is perturbed towards chaos—lacks quantitative universality, and no prior construction has captured the intermediate spectral statistics across different physical systems. This paper formulates a random matrix ensemble that accurately describes the crossover in level statistics for a broad class of quantum systems and uncovers striking universality in the underlying matrix element distributions.

Framework for the Crossover Regime

The considered Hamiltonians take the form $H = H_0 + \gamma H_1$, with $H_0$ integrable and $H_1$ a chaotic perturbation. As $\gamma$ is tuned, the system interpolates from integrability to chaos. The traditional universal limits, Poisson and Wigner-Dyson distributions, do not extend into the crossover, leading previous studies to rely on phenomenologically motivated distributions like the Brody curve or on matrix models such as the Rosenzweig-Porter ensemble. While these models interpolate qualitatively, they lack predictive power for the full level spacing distribution in realistic systems.

The pivotal insight of this work is that the relevant statistics governing the crossover distribution are those of the matrix elements of the perturbation $H_1$ expressed in the eigenbasis of $H_0$. Specifically, the full distribution of offdiagonal elements controls the spectral statistics, and this information can be encoded in a random matrix ensemble: $\mathcal{H} = \mathcal{D} + \gamma' \mathcal{M},$ where $\mathcal{D}$ is a diagonal matrix whose entries are drawn from the spectral density of $H_0$, and $\mathcal{M}$ is a symmetric zero-diagonal matrix whose offdiagonal entries are sampled from the empirical distribution of $H_0$0’s offdiagonal elements in the $H_0$1 basis. The interpolation parameter $H_0$2 is fixed using a spectral statistic such as the average $H_0$3-ratio to align with the physical system.

Quantitative Validation Across Physical Models

To validate the random matrix construction, the authors performed direct comparisons between the level spacing distributions generated by the physical Hamiltonians and those generated by the proposed ensemble across the entire crossover region. Two archetypal systems were considered:

• Spin-1/2 Chains: The physical system is a transverse field Ising chain perturbed by a random XY-Heisenberg Hamiltonian with external fields. Across a range of values interpolating from Poisson to Wigner-Dyson, the constructed random matrix ensemble exactly matches the spectral histograms, outperforming the Brody and Rosenzweig-Porter curves.
• Quantum Resonant Systems (QRS): Bosonic many-body models with weakly nonlinear resonant interactions, tuned between integrable and chaotic regimes. Again, the random matrix ensemble produces level spacing distributions in excellent agreement with the quantum spectra. Notably, alternate models do not capture the correct crossover profile, especially the nontrivial behavior at small gap sizes and the emergence of partial level repulsion.

In both cases, comparison is performed at fixed values of the average $H_0$4-ratio, ensuring that the transitions through the spectral regimes are properly aligned.

Universality in Offdiagonal Matrix Element Statistics

A major observation is that the distribution of offdiagonal matrix elements, $H_0$5, in the $H_0$6 eigenbasis, exhibits robust universality: it is dominated by simple power laws over several orders of magnitude across all studied systems (spin chains, QRS, billiards, and perturbed oscillators). In symbolic form,

$H_0$7

with exponents $H_0$8 typically in the range $H_0$9 and an upper cutoff that is exponentially suppressed. The precise tails and deviations at extreme values are irrelevant for the level statistics due to the dominance of the power-law regime. This empirical universality suggests deep, system-independent structure in the nature of perturbations driving the onset of chaos.

This finding relates directly to the eigenstate thermalization hypothesis (ETH), which predicts universal statistics for matrix elements in non-integrable systems but whose implications for the integrability-to-chaos transition had not previously been explored in this level of detail.

Implications and Theoretical Perspectives

The construction establishes a mapping from physical Hamiltonians to an associated random matrix ensemble, with the matrix element statistics serving as the sufficient input. This connection implies that, given crossover spectral data alone, one can infer nontrivial information about the underlying Hamiltonian’s structure—specifically the statistical features of its chaotic component. This is directly relevant to the analysis of experimental spectral data from platforms such as ultracold atomic gases and quantum processors, where the integrability-to-chaos crossover is observed.

The universality of matrix element statistics and the success of the construction suggest that different physical systems possess identical universality classes in the crossover regime, modulo the details of the $H_1$0 power-law exponents. Analytically, the problem reduces to statistically analyzing random matrix ensembles with a structured diagonal and i.i.d. offdiagonal governed by a non-Gaussian, power-law distribution—a route open to advanced field-theoretic techniques.

One outstanding open problem is obtaining analytical expressions for the spectral correlators and the level spacing distributions in these ensembles, which would clarify the precise mechanisms by which offdiagonal statistics shape the crossover in spectral correlations.

Conclusion

By reformulating the problem of integrability-to-chaos transition in terms of the statistics of perturbing matrix elements, this work provides a comprehensive framework for predicting and understanding spectral statistics in the crossover regime. The success across distinct model systems and the discovery of power-law universality in the offdiagonal matrix element distributions signal a substantive advance in the random matrix theory of quantum chaos. These results open multiple avenues for future research, including the analytic treatment of these ensembles and experimental extraction of matrix statistics from empirical spectral data.