Learning the Action for Long-Time-Step Simulations of Molecular Dynamics
Physical Review Letters · 09.06.2026 · 136 (23)
DOI: 10.1103/shj2-hyw8 ↗Kurz gefasst
Fachartikel in der Zeitschrift Physical Review Letters, online erschienen am 09.06.2026. Verfasst von Filippo Bigi und 2 weiteren Autor:innen. Beteiligte Einrichtung: École Polytechnique Fédérale de Lausanne, Laboratory of Computational Science and Modeling, Institut des Matériaux, 1015 Lausanne, Switzerland. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 30 Literaturangaben. Gefördert von: Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung, National Center of Competence in Research Affective Sciences - Emotions in Individual Behaviour and Social Processes, Swiss Platform for Advanced Scientific Computing.
Abstract (Originalsprache)
The equations of classical mechanics can be used to model the time evolution of countless physical systems, from the astrophysical to the atomic scale. Accurate numerical integration requires small time steps, which limits the computational efficiency-especially in cases such as molecular dynamics that span wildly different timescales. Using machine-learning (ML) algorithms to predict trajectories allows one to greatly extend the integration time step, at the cost of introducing artifacts such as lack of energy conservation and loss of equipartition between different degrees of freedom of a system. We propose learning data-driven structure-preserving (symplectic and time-reversible) maps to generate long-time-step classical dynamics and show that this method is equivalent to learning the mechanical action of the system of interest. These models can be learned based on short reference trajectories and be transferred across thermodynamic conditions and chemical composition. We show that an action-derived ML integrator eliminates the pathological behavior of non-structure-preserving ML predictors and that the method can be applied iteratively, serving as a correction to computationally cheaper direct predictors.
Autor:innen und Institutionen
- École Polytechnique Fédérale de Lausanne, Laboratory of Computational Science and Modeling, Institut des Matériaux, 1015 Lausanne, Switzerland.
Förderung
- Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung
- National Center of Competence in Research Affective Sciences - Emotions in Individual Behaviour and Social Processes
- Swiss Platform for Advanced Scientific Computing
Zitieren
APA
Bigi, F., Spies, J., & Ceriotti, M. (2026). Learning the Action for Long-Time-Step Simulations of Molecular Dynamics. Physical Review Letters, 136(23). https://doi.org/10.1103/shj2-hyw8
Vancouver
Bigi F, Spies J, Ceriotti M. Learning the Action for Long-Time-Step Simulations of Molecular Dynamics. Physical Review Letters. 2026;136(23). doi:10.1103/shj2-hyw8
BibTeX
@article{Bigi2026,
author = {Bigi, Filippo and Spies, Johannes and Ceriotti, Michele},
title = {Learning the Action for Long-Time-Step Simulations of Molecular Dynamics},
journal = {Physical Review Letters},
year = {2026},
volume = {136},
number = {23},
doi = {10.1103/shj2-hyw8}
}
EUVIMED