Xiaozhou Feng, Matteo Ippoliti (Mar 15 2024).
Abstract: The dynamics of quantum entanglement plays a central role in explaining the emergence of thermal equilibrium in isolated many-body systems. However, entanglement is notoriously hard to measure, and can in fact be forged: recent works have introduced a notion of pseudoentanglement describing ensembles of many-body states that, while only weakly entangled, cannot be efficiently distinguished from states with much higher entanglement, such as random states in the Hilbert space. In this work we initiate the study of the dynamical generation and propagation of pseudoentanglement. As generic quantum dynamics tends to maximize actual entanglement, we consider constrained models of time evolution: automaton (i.e. reversible classical) circuits that, when fed suitable input states, provably produce the “standard models” of pseudoentangled ensembles–uniformly random subset(-phase) states–at late times, a phenomenon we name ‘pseudothermalization’. We examine (i) how a pseudoentangled ensemble on a small subsystem spreads to the whole system as a function of time, and (ii) how a pseudoentangled ensemble is generated from an initial product state. We map the above problems onto a family of classical Markov chains on subsets of the computational basis. The mixing times of such Markov chains are related to the time scales at which the states produced from the dynamics become indistinguishable from Haar-random states at the level of each statistical moment (or number of copies). Based on a combination of rigorous bounds and conjectures supported by numerics, we argue that each Markov chain’s relaxation time and mixing time have different asymptotic behavior in the limit of large system size. This is a necessary condition for a cutoff phenomenon: an abrupt dynamical transition to equilibrium. We thus conjecture that our random circuits give rise to asymptotically sharp pseudothermalization transitions.