Many-body localization in disorder-free systems: The importance of finite-size constraints

Z. Papić, E. Miles Stoudenmire, Dmitry A. Abanin

Research output: Contribution to journalArticlepeer-review

91 Scopus citations

Abstract

Recently it has been suggested that many-body localization (MBL) can occur in translation-invariant systems, and candidate 1D models have been proposed. We find that such models, in contrast to MBL systems with quenched disorder, typically exhibit much more severe finite-size effects due to the presence of two or more vastly different energy scales. In a finite system, this can artificially split the density of states (DOS) into bands separated by large gaps. We argue for such models to faithfully represent the thermodynamic limit behavior, the ratio of relevant coupling must exceed a certain system-size depedent cutoff, chosen such that various bands in the DOS overlap one another. Setting the parameters this way to minimize finite-size effects, we study several translation-invariant MBL candidate models using exact diagonalization. Based on diagnostics including entanglement and local observables, we observe thermal (ergodic), rather than MBL-like behavior. Our results suggest that MBL in translation-invariant systems with two or more very different energy scales is less robust than perturbative arguments suggest, possibly pointing to the importance of non-perturbative effects which induce delocalization in the thermodynamic limit.

Original languageEnglish (US)
Pages (from-to)714-725
Number of pages12
JournalAnnals of Physics
Volume362
DOIs
StatePublished - Nov 1 2015
Externally publishedYes

All Science Journal Classification (ASJC) codes

  • General Physics and Astronomy

Keywords

  • Entanglement
  • Ergodicity breaking
  • Exact diagonalization
  • Hubbard model
  • Many-body localization
  • Non-equilibrium quantum dynamics

Fingerprint

Dive into the research topics of 'Many-body localization in disorder-free systems: The importance of finite-size constraints'. Together they form a unique fingerprint.

Cite this