Detecting Ergodic Bubbles at the Crossover to Many-Body Localization using Neural Networks
October 4th, 2021 TOMASZ SZOLDRA Jagiellonian University, Krakow

Several theories of ergodic to many-body localized transition suggest the existence of an avalanche mechanism, in which ergodic bubbles (local, thermal fluctuations of the system properties) thermalize their surroundings, leading to delocalization of the entire system, unless the disorder is sufficiently strong to suppress this process. In this talk we present a tool based on neural networks that allows one to directly identify the ergodic bubbles using experimentally accessible two-site correlation functions. Studying time evolutions of the disordered Heisenberg spin chain, we observe a logarithmic in time growth of the ergodic bubbles in the MBL phase. Investigating the distributions of the bubble sizes, we find an exponential decay in the MBL regime and a power-law distribution with a thermal peak in the critical regime, confirming the presence of the avalanche mechanism. We also find quantitative differences in time evolution of chains with random and quasiperiodic disorder, as well as detect rare (Griffiths) events. These results open new pathways in the research of the mechanisms of thermalization in disordered many-body systems.

Seminar, October 4, 2021, 10:15. Blue Lecture Room

Hosted by Maciej Lewenstein