Thermalization and Eigenstate Thermalization

The eigenstate thermalization hypothesis (ETH) explains how individual many-body energy eigenstates can reproduce thermal values of a restricted class of observables. The research frontier is to establish its scope, exceptions, time scales, and finite-size corrections. ETH is a hypothesis about a Hamiltonian, an energy region, and an observable class; it is not a universal theorem that every isolated quantum system thermalizes.

Helpful background. ETH owns the formal statement; relaxation distinguishes equilibration from thermal agreement.

Matrix elements make the question testable

A common ETH ansatz is

$$O_{mn}=\overline O(\bar E)\delta_{mn} +e^{-S(\bar E)/2}f_O(\bar E,\omega)R_{mn},$$

where $\bar E=(E_m+E_n)/2$, $\omega=E_m-E_n$ is an energy difference, and $S$ is a dimensionless microcanonical entropy. The ansatz is evaluated within the appropriate conserved-quantity sectors. Smooth diagonal matrix elements and the statistical structure of off-diagonal elements are separate claims.

For an initial state with coefficients $c_n$, the dephased expectation is $\sum_n|c_n|^2O_{nn}$ when the relevant degeneracies are handled correctly. Thermal agreement requires a sufficiently narrow energy distribution and suitable behavior of the participating eigenstates. Spectral level repulsion alone cannot establish these conditions.

Known results and real boundaries

Established: many nonintegrable finite models exhibit ETH behavior in appropriate spectral regions; the D’Alessio review assembles this evidence and its thermodynamic implications. Integrable models, localization regimes, and special nonthermal states show why qualifications matter. Active: identify the scale of deviations and the relation between matrix-element statistics, conservation laws, and relaxation. Conjectural: strong ETH for broad classes of realistic interacting Hamiltonians without a model-specific argument.

Weak ETH allows a vanishing fraction of exceptional eigenstates, whereas strong ETH excludes exceptions in the stated energy window. A small outlier fraction can still matter to a specially prepared experiment. Conversely, an outlier at a spectral edge does not disprove a claim restricted to finite energy density in the bulk.

Starter project: typical states versus worst cases

Diagonalize a spin chain with conserved total magnetization, resolving spatial symmetries when using level statistics. Compare an integrable point with a nonintegrable perturbation. For a local or few-site observable, plot $O_{nn}$ against energy and evaluate both the variance and the largest deviation within a fixed energy-density window.

Use several sizes that fit a workstation. Deliver raw sector definitions, an energy-window sensitivity study, and separate fits for typical and extremal deviations. Check the diagonal ensemble for two initial states: a generic product state and a deliberately chosen state with unusually large overlap on an outlier.

A failure test is apparent ETH violation that disappears after separating symmetry sectors or narrowing an energy window consistently. Another is a claimed exponential trend determined by only two system sizes. Extend the project by relating a particular class of outliers to a structural mechanism, or by testing how weak symmetry breaking changes their lifetimes.

Open questions that support a thesis

How do hydrodynamic modes constrain $f_O$ near small energy differences? What finite-size information can bound thermalization for a restricted family of initial states? Which experimentally measurable observables distinguish an ETH exception from slow equilibration? A rigorous theorem project and a dynamical experiment require different success criteria; choose one before selecting numerical diagnostics.

The connection to scrambling is valuable but not an equivalence. Likewise, many-body localization demands an additional analysis of disorder and asymptotic limits.

Entry points and evidence horizon

Start with the 2016 review and reproduce one finite-system test before studying specialized exceptions. Literature links were checked on 2026-10-11; the cited review is a foundational entry point, not a claim that all later ETH questions have been settled.

References

• D’Alessio, L., Kafri, Y., Polkovnikov, A., and Rigol, M., “From quantum chaos and eigenstate thermalization to statistical mechanics and thermodynamics,” Advances in Physics 65, 239–362 (2016), arXiv:1509.06411 — ETH formulation, evidence, and boundaries.