Machine learning speeds up molecular simulations by trimming small interactions
Truncated automatic sparse differentiation for machine learning interatomic potentials
Machine Learning
Summary
Predicting forces between atoms is critical for understanding materials, but calculating how these forces change together can be very slow for big systems. The authors found that many of these detailed calculations can be skipped because distant atoms barely affect each other. By focusing only on the important nearby interactions, their method speeds up these calculations a lot without losing accuracy. This approach makes it easier to simulate and predict materials' behavior at the atomic level.
What this means in practice
- •For materials simulation teams: Compute accurate atomic motion patterns faster by ignoring negligible long-distance atomic interactions in machine-learned potentials.
- •For computational chemistry groups: Speed up Hessian calculations to predict experimental observables more efficiently for large molecular systems using truncated differentiation.
Authors
Marcel F. Langer, Adrian Hill, Michele Ceriotti
Abstract
Machine learning interatomic potentials (MLIPs) learn the mapping from atomic positions to potential energy. The forces, the negative gradient of this energy, drive molecular dynamics and are readily obtained using automatic differentiation. Higher-order derivatives, most notably the Hessian, describe collective motion and allow the direct prediction of experimental observables, but are considered computationally inaccessible for large systems. We suggest a solution: in physical systems, interactions decay with distance, and most MLIPs build on this locality through message passing up to a finite receptive field. This implies both sparsity of higher-order derivatives and their decay with distance. This structure can be exploited using automatic sparse differentiation (ASD). We explain how to compute the sparsity pattern for MLIP derivatives and demonstrate that, for multiple foundation MLIPs, ASD computes full Hessians of large porous materials exactly, but with modest speedups at best. The larger gains come from truncated ASD: discarding small, but nonzero, Hessian entries between distant atoms yields order-of-magnitude speedups with negligible impact on predicted observables.