Learning Smooth and Expressive Interatomic Potentials for Physical Property Prediction

2025-02-17

AI Summary

Despite persistent reductions in energy/force prediction errors, machine-learned interatomic potentials (MLIPs) often yield physically inconsistent property predictions in molecular dynamics (MD) simulations due to poor energy conservation. This work clarifies that low test-set error does not guarantee physical consistency in MD and proposes long-term energy conservation as the core practical metric for MLIP evaluation. To this end, we introduce the eSEN model: a smooth-constrained neural network architecture incorporating a physics-guided, conservation-law-weighted loss function and an MD-trajectory-driven training paradigm. On tasks including material stability, thermal conductivity, and phonon spectra, eSEN achieves state-of-the-art accuracy while significantly strengthening the correlation between test-set error and actual physical property prediction fidelity. Our results uncover a fundamental trade-off between representational capacity and physical consistency—highlighting the necessity of embedding conservation principles directly into MLIP design and training.

Technology Category

• Machine Learning: Ensemble Methods
• Multiagent Systems: Multiagent Learning
• Natural Language Processing: Safety and Robustness

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Machine learning interatomic potentials (MLIPs) have become increasingly effective at approximating quantum mechanical calculations at a fraction of the computational cost. However, lower errors on held out test sets do not always translate to improved results on downstream physical property prediction tasks. In this paper, we propose testing MLIPs on their practical ability to conserve energy during molecular dynamic simulations. If passed, improved correlations are found between test errors and their performance on physical property prediction tasks. We identify choices which may lead to models failing this test, and use these observations to improve upon highly-expressive models. The resulting model, eSEN, provides state-of-the-art results on a range of physical property prediction tasks, including materials stability prediction, thermal conductivity prediction, and phonon calculations.

Problem

• Improve MLIPs for physical property prediction
• Test MLIPs on energy conservation in simulations
• Develop eSEN for superior property prediction

Innovation

• Machine learning interatomic potentials
• Energy conservation in simulations
• Improved physical property prediction