Rath, Y; Booth, G H (2025) Interpolating numerically exact many-body wave functions for accelerated molecular dynamics. Nature Communications, 16. 2005

Preview

Text eid10212.pdf - Published Version Available under License Creative Commons Attribution. Download (1MB) | Preview

Abstract

While there have been many developments in computational probes of both strongly-correlated molecular systems and machine-learning accelerated molecular dynamics, there remains a significant gap in capabilities between them, where it is necessary to describe the accurate electronic structure over timescales in which atoms move. We describe a practical approach to bridge these fields by interpolating the correlated many-electron state through chemical space, whilst avoiding the exponential complexity of these underlying states. With a small number of accurate correlated wave function calculations as a training set, we demonstrate provable convergence to near-exact potential energy surfaces for subsequent dynamics with propagation of a valid many-body wave function and inference of its variational energy at all points, whilst retaining a mean-field computational scaling. This represents a profoundly different paradigm to the direct interpolation of properties through chemical space in established machine-learning approaches. It benefits from access to all electronic properties of interest from the same model without relying on local descriptors, and demonstrates improved performance compared to the direct training on energies themselves. We combine this with modern systematically-improvable electronic structure methods to resolve the molecular dynamics for a number of correlated electron problems, including the proton dynamics of a Zundel cation trajectory, where we highlight the qualitative improvement from traditional machine learning or ab initio dynamics on mean-field surfaces.

Item Type:	Article
Keywords:	numerical techniques, molecular dynamics, eigenvector continuation, machine learning
Subjects:	Mathematics and Scientific Computing > Numerical Computation
Divisions:	Quantum Technologies
Identification number/DOI:	10.1038/s41467-025-57134-9
Last Modified:	30 Jul 2025 13:27
URI:	https://eprintspublications.npl.co.uk/id/eprint/10212