Published April 2, 2024 | Version v1
Journal article

Diagnosing non-Hermitian many-body localization and quantum chaos via singular value decomposition

  • 1. Department of Physics and Materials Science, University of Luxembourg, L-1511 Luxembourg

Description

Strong local disorder in interacting quantum spin chains can turn delocalized eigenmodes into localized eigenstates, giving rise to many-body localized phases. This is accompanied by distinct spectral statistics: chaotic for the delocalized phase and integrable for the localized phase. In isolated systems, localization and chaos are defined through a web of relations among eigenvalues, eigenvectors, and real-time dynamics. These may change as the system is made open. We ask whether random dissipation (without random disorder) can induce chaotic or localized behavior in an otherwise integrable system. The dissipation is described using non-Hermitian Hamiltonians, which can effectively be obtained from Markovian dynamics conditioned on null measurement. In this non-Hermitian setting, we argue in favor of the use of the singular value decomposition. We complement the singular value statistics with different diagnostic tools, namely, the singular form factor and the inverse participation ratio and entanglement entropy of singular vectors. We thus identify a crossover of the singular values from chaotic to integrable spectral features and of the singular vectors from delocalization to localization. Our method is illustrated in an XXZ Hamiltonian with random local dissipation.

Additional details

Identifiers

DOI
10.1103/PhysRevB.109.L140201;
arXiv
arXiv:2311.16229;
Crossref Funder ID
10.13039/100000925; 10.13039/501100007601; 10.13039/501100001866; 10.13039/100008665;

Publishing Information

Journal Title
Physical Review B
Journal Volume
109
Journal Issue
14
Journal Page Range
6 pgs.
ISSN
1550-235X

Optional Information

Copyright
©2024 American Physical Society
Contract/Grant/Project number
62171; 16434093; 15382998
Notes
These authors contributed equally to this work.; Record automatically processed
Funding organization
John Templeton Foundation; Horizon 2020; Fonds National de la Recherche Luxembourg; Université du Luxembourg