arXiv:2609.30248
Deep thermalization and Hilbert space ergodicity
Daniel K. Mark, Manuel Endres, Matteo Ippoliti, Wen Wei Ho, Soonwon Choi
Abstract
Recent advances in quantum simulation have enabled the discovery of novel forms of universality in quantum many-body dynamics. In this Review, we discuss deep thermalization and Hilbert space ergodicity--two recently discovered phenomena characterized by the emergence of distributions of quantum states that are "maximally random" in a precise sense. They offer a new perspective on how irreversible statistical mechanics arises from reversible unitary quantum dynamics, going beyond conventional theories of quantum thermalization and equilibration. We review a unifying framework, rooted in quantum information theory and principles of maximum entropy, that explains the different forms of ergodicity which arise under various physical constraints. We discuss open questions and active research directions, including generalizations beyond ensembles of pure quantum states, phase transitions in deep thermalization tied to deeper forms of ergodicity-breaking, connections to broader topics in quantum thermalization and ergodicity, and applications in quantum information science, such as for benchmarking or tomography.
AI Summary
This Review examines how quantum many-body dynamics generates universal distributions of states beyond the density matrices described by conventional thermalization. It connects deep thermalization—the distribution of subsystem states conditioned on measurements of a bath—with Hilbert-space ergodicity, which concerns the distribution of states visited along a temporal trajectory. The organizing framework is the hierarchy of ensemble moments $\rho_{\mathcal E}^{(k)}=\mathbb E_{\psi\sim\mathcal E}[(|\psi\rangle\langle\psi|)^{\otimes k}]$. Agreement with Haar moments defines state designs; distinguishability is quantified by $\Delta^{(k)}=\tfrac12\|\rho_{\mathcal E}^{(k)}-\rho_{\mathcal E'}^{(k)}\|_1$. The review surveys experimental evidence from neutral atoms, superconducting qubits and NV centers, including probability-of-probabilities statistics and tomographic moment reconstruction, alongside rigorous results for random circuits, dual-unitary circuits and deterministic drives. These are results from the literature rather than new experiments or proofs. A central synthesis concerns randomness under constraints. Scrooge ensembles minimize accessible classical information at fixed mean density matrix, providing a constrained counterpart to Haar randomness. Bath measurements that reveal conserved quantities can instead produce generalized Scrooge ensembles. The discussion also covers Gaussian systems and deep-ergodicity-breaking transitions, where conventional thermalization persists but higher ensemble statistics change. Open issues include when ordinary thermalization implies deep thermalization, convergence timescales, classification of exceptions and transitions, and extensions to mixed states and processes. Direct ensemble reconstruction requires exponentially many samples; computationally assisted diagnostics shift costs to classical simulation. Efficient characterization beyond classical simulability and applications specific to Scrooge randomness remain unresolved.
Error found: The review identifies the temporal ensemble of time-independent Hamiltonian evolution with a Scrooge ensemble under no-resonance conditions alone, although their second moments generally differ because Hamiltonian evolution fixes all energy-basis populations.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 11 pages, 5 figures. Comments welcome!