A Strategy for Proving the Strong Eigenstate Thermalization Hypothesis : Chaotic Systems and Holography
Taishi Kawamoto
TL;DR
This work provides sufficient conditions for the strong ETH in highly chaotic theories by linking long-time real-time thermal correlators to probabilistic bounds that decay doubly exponentially with entropy. The key mechanism relies on clustering of Wightman functions and quantum mixing of two-point functions, together enabling sub-Gaussian tail bounds and, in turn, double-exponential suppression of diagonal and off-diagonal ETH violations. The authors illustrate the approach with bottom-up holographic toy models, including large-$N$ generalized free fields and controlled $1/N$ corrections, and discuss how wormhole contributions and back-reaction may affect late-time behavior. They also connect these results to the standard ETH ansatz and discuss energy-window considerations, offering a framework to understand thermalization in chaotic quantum many-body systems and their gravity duals. The findings advance a quantitative pathway to proving strong ETH in holographic contexts and suggest robust thermalization mechanisms in theories with gravity duals.
Abstract
The strong eigenstate thermalization hypothesis (ETH) provides a sufficient condition for thermalization and equilibration. Although it is expected to be hold in a wide class of highly chaotic theories, there are only a few analytic examples demonstrating the strong ETH in special cases, often through methods related to integrability. In this paper, we explore sufficient conditions for the strong ETH that may apply to a broad range of chaotic theories. These conditions are expressed as inequalities involving the long-time averages of real-time thermal correlators. Specifically, as an illustration, we consider simple toy examples which satisfy these conditions under certain technical assumptions. This toy models have same properties as holographic theories at least in the perturbation in large $N$. We give a few comments for more realistic holographic models.
