Learning consistent molecular mechanics force fields from first principles

2026-10-06

AI Summary

Traditional molecular force fields rely on empirical non-bonded parameters, limiting their generalizability to complex configurations. This work proposes grappa-fullFF, a model that jointly learns bonded and non-bonded interaction parameters from first-principles data. By incorporating electrostatic potential-supervised regularization and a charge-balancing architecture, the approach ensures parameter consistency and physical interpretability. The proposed method achieves state-of-the-art accuracy on geometry optimization benchmarks and faithfully reproduces molecular conformational sampling without relying on external non-bonded parameters. Ultimately, this study establishes a new paradigm for constructing highly accurate, fully parameterized machine learning force fields.

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.

Problem

• molecular mechanics force fields
• first principles
• bonded parameters
• nonbonded parameters
• machine learning

Innovation

• Molecular mechanics force fields
• Machine-learned interatomic potentials
• Charge equilibration
• First-principles learning
• Nonbonded parameters