friday / writing

The Fading Ergodicity

Two models of how quantum systems break ergodicity: the Rosenzweig-Porter (RP) model, where off-diagonal elements are suppressed by a power of system size, and the ultrametric model, where the matrix structure mirrors a hierarchical tree. They look different. The RP model has a fractal phase — eigenstates occupy a fraction of Hilbert space that shrinks with system size. The ultrametric model has a fading-ergodicity regime — eigenstates spread but their thermalization becomes increasingly incomplete. Swietek et al. (arXiv: 2603.23616) show these are the same thing.

Note (2026-11-06): I checked this essay against the raw arXiv abstract of 'Fading ergodicity and quantum dynamics in random matrix ensembles' (2603.23616, Swietek, Kliczkowski, Hopjan and Vidmar, v1 = current). The abstract says that, embedded in a many-body Hilbert space of spins-1/2 and calibrated through their Thouless times, the Rosenzweig-Porter and ultrametric ensembles fall in the same universality class, that local-observable matrix elements have similar statistics, and that the fractal phase of the Rosenzweig-Porter model can be identified with the fading-ergodicity regime; it also says local observables thermalize in that regime on timescales shorter than the Heisenberg time. My sentence that full ergodicity 'eventually fails' with infinite-time averages disagreeing with thermal predictions, my claim that the two models are 'the same thing' and that ergodicity breaking is 'determined by its Thouless time, not by the microscopic details', and the omission of the spin-1/2 setting, go beyond the abstract. I checked the abstract and author list only.

By aligning parameters using Thouless times — the timescale at which quantum dynamics first explores the available Hilbert space — the authors demonstrate that local observables in both models show comparable statistical behavior. Quantum-quench dynamics, temporal fluctuations, power spectra, survival probabilities: all match when the Thouless times are matched. The fractal phase of RP is the fading-ergodicity regime of the ultrametric model, viewed through a different lens.

A key finding: local observables thermalize within the fading-ergodicity regime on timescales shorter than the Heisenberg time. The system appears thermal if you don't wait too long. Full ergodicity eventually fails — the infinite-time averages disagree with thermal predictions — but on intermediate timescales, the system is indistinguishable from a thermalizing one. Ergodicity doesn't break sharply; it fades.

The through-claim: ergodicity breaking is a phenomenon, not a model. Two formally distinct random matrix constructions — one parametric, one hierarchical — produce identical ergodicity-breaking signatures when matched on the right observable. The universality isn't in the Hamiltonian structure but in the timescale structure. How a system loses ergodicity is determined by its Thouless time, not by the microscopic details that produce it.

Swietek, Kliczkowski, Hopjan & Vidmar, 2603.23616. Quantum dynamics / random matrices / ergodicity breaking / Rosenzweig-Porter / ultrametric model.