Optimal statistical ensembles for quantum thermal state preparation within the quantum singular value transformation framework

Yasushi Yoneta

Published 2025-05-09Version 1

Preparing thermal equilibrium states is an essential task for finite-temperature quantum simulations. In statistical mechanics, microstates in thermal equilibrium can be obtained from statistical ensembles. To date, a variety of ensembles have been devised, including not only the Gibbs ensembles, such as the canonical and microcanonical ensembles, but also various generalized ensembles. Since these ensembles yield equivalent thermodynamic predictions, one can freely choose an ensemble for computational convenience. In this paper, we exploit this flexibility to develop an efficient quantum algorithm for preparing thermal equilibrium states. We first present a quantum algorithm for implementing generalized ensembles within the framework of quantum singular value transformation. We then perform a detailed analysis of the computational cost and elucidate its dependence on the choice of the ensemble. We show that selecting an appropriate ensemble can significantly mitigate ensemble-dependent overhead and yield improved scaling of the computational cost with system size compared to existing methods based on the canonical ensemble. Moreover, we numerically demonstrate that our approach achieves a significant reduction in the computational cost even for finite-size systems. Our algorithm applies to arbitrary thermodynamic systems at any temperature and is thus expected to offer a practical and versatile method for computing finite-temperature properties of quantum many-body systems. These results highlight the potential of ensemble design as a powerful tool for enhancing the efficiency of a broad class of quantum algorithms.