Simple slow operators and quantum thermalization

Published in arXiv preprint, 2026

Understanding how thermalization happens from microscopic first principles is one of the holy grails of theoretical physics. For quantum thermalization, the canonical theory is the eigenstate thermalization hypothesis (ETH). Despite being widely used for three decades, as its name suggests, it remains a hypothesis that is not proven. In fact, it is not even a hypothesis awaiting a proof – it is an ansatz, and it did not specify under what conditions (on the Hamiltonian and the observable) should this ansatz hold.

In this work, we develop a rigorous theory of thermalization without relying on the ETH ansatz, in which we explicitly prescribe physical conditions under which thermalization will happen. The condition is a formalization of the intuitive notion that non-thermal systems are characterized by the presence of integrals of motion / conserved quantitites. We formalize this by defining the concept of simple slow operators (SSOs), which are roughly operators that are both (a) integrals of motion or spectrum generating algebra operators and (b) have significant small-sized components. We prove that the absence of SSOs implies the thermalization of typical physical initial states (cf. Huang, Pilatowsky-Cameo & Choi). Importantly, this result can describe both infinite-time thermalization and finite-time thermalization: for the finite-time case, one just relax the condition in (a) to include approximate IOM/SGA. Our key technical apparatus is what we call the ensemble variance norm, a notion related to the shadow norm (Huang et al.), that connects the thermalization of typical physical initial states to operator growth.

Download paper here