arXiv:2610.01538
Efficient learning of quantum interactions from thermal metastable states
Bingrun Wang, Qi Ye, Chi-Fang Chen
Abstract
Learning quantum interactions from finite-temperature many-body systems is a central task in emerging quantum platforms. Recently, the problem of learning from lattice quantum Gibbs states has found rigorous, efficient protocols. Nevertheless, exact Gibbs states, as the input premise, are in fact computationally intractable to prepare and may not faithfully represent generic finite-temperature quantum systems. In contrast, a system coupled to a heat bath can be stuck at an approximate stationary state (metastable state) long before it truly equilibrates. Here, we formulate a physically and algorithmically consistent alternative: learning from such metastable states of detailed-balanced master equations (Lindbladians) arising from system-bath interactions. We distill the algorithmic mechanism and structural condition underlying Gibbs-state learning and extend it in full to metastable states, attaining nearly optimal sample and computational complexity (in the system size and the precision). More broadly, we sharpen notions of metastability and develop a unified framework for finite-temperature learning.
AI Summary
Can local quantum interactions be learned from thermal states that have not globally equilibrated? This paper replaces exact Gibbs-state inputs with approximate stationary states of a specified quasi-local, KMS-detailed-balanced Lindbladian, including Hamiltonian evolution. The setting is an $n$-qubit, fixed-dimensional lattice Hamiltonian with known bounded-range Pauli terms, unknown coefficients in $[-1,1]$, and known inverse temperature $\beta$. Given independent, possibly nonidentical states satisfying $\|\mathcal L(\sigma_i)\|_1\leq\epsilon$, the protocol estimates every coefficient to additive error $\eta$ with failure probability $\delta$, using $\widetilde O(e^{\operatorname{poly}(\beta^{\pm1})}\eta^{-2}\log(n/\delta))$ samples and $n$ times that computational cost. The guarantee requires $\eta\geq C_\beta\epsilon^{\alpha_1}n^{\alpha_2}$, with temperature-dependent positive exponents. Assuming instead $\max_a\|\mathcal L_a(\sigma_i)\|_1\leq\epsilon$ removes the system-size factor from this precision floor. Random-time averaging of Lindbladian evolution supplies suitable inputs in polynomial evolution time at fixed temperature. The proof converts metastability into approximate detailed balance, then derives approximate KMS identities, local faithfulness, and tested commutator bounds. These support measurable local identifiability tests and iterative min–max coefficient refinement. Additional results relate entropy production, detailed-balance defects, and local/global stationarity, including regularization arguments accommodating rank-deficient states. The guarantees use a particular Gaussian-filtered thermal generator with single-site Pauli jumps throughout the lattice; extensions to other generators remain unproved. Temperature dependence and error exponents are not optimized. Whether global metastability necessitates a system-size-dependent precision floor remains open. The physical model assumes weak-coupling, Markovian bath dynamics.
LLMs also assisted with polishing the writing and figure preparation.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.