arXiv：cs.LG· Berkay G\"unes, Leif Seute, Jigyasa Nigam, Frauke Gr\"ater·· 6 小时前AI 评分36
grappa-fullFF：从第一性原理一致学习分子力学力场
Learning consistent molecular mechanics force fields from first principles
AI 导读
研究者提出 grappa-fullFF，可同时且一致地从 ab initio 参考数据中学习键合与非键合参数，无需外部指定的非键合参数。该模型通过静电梯度监督引入物理正则化并结合电荷均衡架构，恢复了准确的电响应性质，在几何优化基准上达到 SOTA 精度，并复现了经典力场与已有机器学习力场的构象采样。该工作已被 NeurIPS 2026 的 ML4Molecules Workshop 接收。
正文
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.
| Comments: | Accepted to the ML4Molecules Workshop at NeurIPS 2026 |
| Subjects: | Chemical Physics (physics.chem-ph); Machine Learning (cs.LG); Computational Physics (physics.comp-ph) |
| Cite as: | arXiv:2610.08020 [physics.chem-ph] |
| (or arXiv:2610.08020v1 [physics.chem-ph] for this version) | |
| https://doi.org/10.48550/arXiv.2610.08020 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Berkay Günes [view email]
[v1]
Tue, 6 Oct 2026 09:14:50 UTC (2,988 KB)
来源：arXiv：cs.LG · arxiv.org