Accelerating Molecular Dynamics Simulations with Foundation Neural Network Models using Multiple Time-Step and Distillation

Abstract: We present a strategy to accelerate molecular dynamics simulations using foundation neural network models. To do so, we apply a dual-level neural network multi-time-step (MTS) strategy where the target accurate potential is coupled to a simpler but faster model obtained via a distillation process. Thus, the 3.5 \r{A}-cutoff distilled model is sufficient to capture the fast varying forces, i.e. mainly bonded interactions, from the accurate potential allowing its use in a reversible reference system propagator algorithms (RESPA)-like formalism. The approach conserves accuracy, preserving both static and dynamical properties, while enabling to evaluate the costly model only every 3 to 6 fs depending on the system. Consequently, large simulation speedups over standard 1 fs integration are observed: 4-fold in homogeneous systems and 2.3-fold in large solvated proteins. Such a strategy is applicable to any neural network potential and reduces their performance gap with classical force fields.