arXiv:2610.30070
A random sampling method to test the weak Eigenstate Thermalization Hypothesis
Nicolaus A. Chlanda, John Keim, Isabel Martinez-Robles, Sarah E. Spielman, Thomas J. Carroll, Michael W. Noel
Abstract
We introduce a random sampling method for testing the weak Eigenstate Thermalization Hypothesis (wETH) that is particularly useful for large, sparse Hamiltonians. We test the method on a well-studied Rydberg atom system that exhibits both two- and three-body dipole-dipole interactions. Our results show that our method allows for the study of system sizes significantly beyond the limits of exact diagonalization. We find that the fluctuations in eigenstate expectation values decay as the Hilbert space dimension $D^{-1/2}$, consistent with wETH and in agreement with previous numerical results.
AI Summary
The paper addresses how to test weak eigenstate thermalization (wETH) in sparse many-body Hamiltonians without full diagonalization. Its diagnostic is the standard deviation $\sigma_{\Delta S}$ of eigenstate expectation values in the middle 5% of the spectrum, examined as the Hilbert-space dimension $D$ increases. The method combines a kernel polynomial estimate of the density of states with density-weighted random sampling of target energies. Chebyshev filter diagonalization then extracts eigenpairs in adaptive brackets around those targets. Statistics are assembled from 2,000 sampled expectation values per Hamiltonian. The application is a Rydberg-atom chain with resonant two- or three-body dipole-dipole exchange and hopping interactions; the three-body regime was previously associated with many-body scars. The reported results extend from exact-diagonalization limits of 12 atoms ($D\approx7.38\times10^4$) to 14 atoms ($D\approx6.13\times10^5$) for two-body interactions, and from 11 atoms ($D\approx1.35\times10^5$) to 13 atoms ($D\approx1.8\times10^6$) for three-body interactions. Sampling agrees with exact diagonalization where both are available. Fits give $\sigma_{\Delta S}\sim D^{-1/2}$, interpreted as evidence consistent with wETH; exact diagonalization alone is inconclusive in the three-body case. Stated limitations include substantial small-system effects, possible correlations among eigenstates collected within a bracket, and memory and convergence constraints. The observed scaling does not resolve the contrast with an analytical inverse-system-size prediction, which may require larger systems. The parameter table lists additional, larger calculations beyond the sizes discussed in the results.
Central claim may not hold: The observable underlying every reported fluctuation is never specified, preventing reproduction and leaving the scope of the wETH test undefined. The parameter table lists larger completed CFD system sizes than the results text reports, without explaining whether those sizes enter the scaling analysis. The D^{-1/2} fits lack numerical exponents, uncertainties, and sampling errors, and the text and caption disagree about the three-body fit range. The sampling procedure does not quantify batch correlations or convergence-dependent selection, so its effective sample size and uniform eigenstate inclusion are not established. The duplicate-removal rule uses eigenvalue proximity alone, which can discard distinct eigenstates; its effect on the reported fluctuations is not tested.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- initial version