Beyond Force Metrics: Pre-Training MLFFs for Stable MD Simulations
This addresses the challenge of reliable molecular dynamics simulations for computational chemistry, though it is incremental as it builds on existing methods.
The study tackled the problem of ensuring stable molecular dynamics simulations with machine-learning force fields, finding that pre-training on a large dataset improved simulation stability by sustaining trajectories up to three times longer than training from scratch, despite similar force errors of 5 meV/A per atom.
Machine-learning force fields (MLFFs) have emerged as a promising solution for speeding up ab initio molecular dynamics (MD) simulations, where accurate force predictions are critical but often computationally expensive. In this work, we employ GemNet-T, a graph neural network model, as an MLFF and investigate two training strategies: (1) direct training on MD17 (10K samples) without pre-training, and (2) pre-training on the large-scale OC20 dataset followed by fine-tuning on MD17 (10K). While both approaches achieve low force mean absolute errors (MAEs), reaching 5 meV/A per atom, we find that lower force errors do not necessarily guarantee stable MD simulations. Notably, the pre-trained GemNet-T model yields significantly improved simulation stability, sustaining trajectories up to three times longer than the model trained from scratch. These findings underscore the value of pre-training on large, diverse datasets to capture complex molecular interactions and highlight that force MAE alone is not always a sufficient metric of MD simulation stability.