arXiv:2611.00233
Distance to integrability and complexity
Sean Casey, Marius de Leeuw, Tristan McLoughlin, Sergiy Zhuk
Abstract
We consider nearest-neighbour interactions on spin-1/2 chains. We find the full set of integrable models of this form and apply a numerical algorithm to determine the "closest" integrable model for a given Hamiltonian. This allows us to define a model-independent distance to integrability. We show that standard measures of chaos interpolate between integrable and chaotic values with increasing values of this distance. However, there is a large variation across models and the distance to integrability provides additional information not directly provided by these diagnostics.
AI Summary
The paper asks how to quantify proximity to integrability without choosing a particular integrable Hamiltonian and perturbation direction. For translationally invariant nearest-neighbour spin-$1/2$ chains, it reports 12 integrable families obtained by solving $[Q_2,Q_3]=0$ within the regular Yang–Baxter/boost framework. It defines distance by minimising the full-chain Frobenius norm over these families and local $\mathrm{SL}(2,\mathbb R)$ similarity transformations. The normalisation $d_{\min}^2=\min_H\|M-H\|_F^2/(L2^L)$ removes chain-length dependence for traceless densities. Linear parameters are eliminated by least squares, leaving a three-parameter non-convex optimisation. A released dataset contains 200,000 random real-symmetric densities, normalised after removing identity and telescoping terms. For $L=16$, spectral-spacing ratios and eigenstate Shannon entropies show average crossovers towards chaotic values as distance increases, but substantial model-to-model variation. At $d_{\min}\gtrsim0.07$, more than half the sampled models have mean adjacent-gap ratio above $0.459$, halfway between Poisson and GOE values. A minimised charge-commutator residual correlates with distance but is not a reliable proxy. Optimising the first three Lanczos coefficients over one- and two-site seed operators shows that proximity to integrability permits both slower and faster early-time operator growth, rather than uniformly suppressing it. The numerical study is restricted to real-symmetric inputs; the closest integrable model need not be symmetric. The spectral crossover depends on system size, distance depends on the chosen norm, and the higher-Lanczos-coefficient optimisations use heuristic global searches. Complex-Hermitian inputs and family-resolved analysis remain open extensions.
Central claim may not hold: The 12 listed families and real local similarities omit the real symmetric integrable XYZ density XX + 3YY + 2ZZ, invalidating completeness even within the sampled real domain. The search is restricted to real densities and real local similarities, so it does not establish the abstract's unrestricted classification. Because non-symmetric candidates are allowed, the reported distance is not established as a distance to the nearest physical Hermitian integrable Hamiltonian. The numerical optimisation lacks a demonstrated global-optimality guarantee, so the reported minima remain uncertified even over the listed search space.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 8 pages, 4 figures