Learning consistent molecular mechanics force fields from first principles
Abstract
Classical force fields (FFs) remain the workhorse for large-scale simulations even as machine-learned interatomic potentials (MLIPs) approach ab initio accuracy. They decompose total configuration energies into simple effective interactions whose parameters are traditionally assigned based on atom or bond types, enabling efficient simulations but also limiting their ability to adapt across configurations. Recent machine learning approaches have improved the accuracy and transferability of bonded parameters in these FFs by inferring them as functions of local atomic environments, but still rely on empirical nonbonded parameters for practical simulations. In this work, we introduce a unified approach, \texttt{grappa-fullFF}, which learns both bonded and nonbonded parameters \emph{consistently} and simultaneously from ab initio reference data. By incorporating physically inspired regularization via supervision of the electrostatic potential and an architecture that facilitates charge equilibration, our model recovers accurate electric response properties, achieves state-of-the-art accuracy on geometry optimization benchmarks, and reproduces the conformational sampling of both classical and existing machine-learned FFs, without relying on externally assigned nonbonded parameters.
Related
- Ammonia-Hydrogen Phase Equilibria in Planetary Envelopes from First-Principles-Based Machine-Learning Molecular Dynamics
- Accuracy of Vibrational Properties of Machine-Learning Force Fields for Molecular Crystal Transformations
- Neural network self-consistent fields for density functional theory
- Molecular augmented dynamics: Generating experimentally consistent atomistic structures by design
- Molecular Mechanics-Based Symmetric Two-State Energy-Gap Estimation of Binding Free Energies
BibTeX
@misc{gunes2026learning,
title = {{Learning consistent molecular mechanics force fields from first principles}},
author = {Berkay Günes and Leif Seute and Jigyasa Nigam and Frauke Gräter},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.08020},
eprint = {2610.08020},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.08020},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Learning consistent molecular mechanics force fields from first principles
AU - Berkay Günes
AU - Leif Seute
AU - Jigyasa Nigam
AU - Frauke Gräter
PY - 2026
DA - 2026/10/06
PB - arXiv
DO - 10.48550/arxiv.2610.08020
UR - https://arxiv.org/abs/2610.08020
AB - Classical force fields (FFs) remain the workhorse for large-scale simulations even as machine-learned interatomic potentials (MLIPs) approach ab initio accuracy. They decompose total configuration energies into simple effective interactions whose parameters are traditionally assigned based on atom or bond types, enabling efficient simulations but also limiting their ability to adapt across configurations. Recent machine learning approaches have improved the accuracy and transferability of bonded parameters in these FFs by inferring them as functions of local atomic environments, but still rely on empirical nonbonded parameters for practical simulations. In this work, we introduce a unified approach, \texttt{grappa-fullFF}, which learns both bonded and nonbonded parameters \emph{consistently} and simultaneously from ab initio reference data. By incorporating physically inspired regularization via supervision of the electrostatic potential and an architecture that facilitates charge equilibration, our model recovers accurate electric response properties, achieves state-of-the-art accuracy on geometry optimization benchmarks, and reproduces the conformational sampling of both classical and existing machine-learned FFs, without relying on externally assigned nonbonded parameters.
ER -