English

Information Compression at Criticality

Statistical Mechanics 2026-07-20 v1 Disordered Systems and Neural Networks Quantum Physics

Abstract

Highly excited quantum states at the critical boundary of ergodicity are known to deviate from thermal behavior, yet their dynamical properties remain poorly understood. Here, we uncover the complexity of quantum dynamics at criticality through the lens of intrinsic information compression in energy space. We show that the Hamiltonian spectrum can be systematically truncated, yielding a simplified description of the dynamics while preserving its essential features. Specifically, for both interacting and noninteracting systems, we demonstrate that a vanishing fraction of Hamiltonian eigenlevels suffices to reproduce the power-law decay of the survival probability. The resulting truncated spectrum exhibits a fractal structure characterized by a level-spacing distribution with a power-law tail, while its spectral form factor displays the same asymptotic power-law decay as the survival probability.

Cite

@article{arxiv.2607.18388,
  title  = {Information Compression at Criticality},
  author = {Simon Jiricek and Miroslav Hopjan and Boris Altshuler and Vladimir Kravtsov and Lev Vidmar},
  journal= {arXiv preprint arXiv:2607.18388},
  year   = {2026}
}

Comments

8 pages, 7 figures