Predicting the Dynamics of Atoms in Liquids by a Surrogate Machine-Learned Simulator

Molecular dynamics (MD) is a workhorse of computational material science. However, the inner-loop algorithm of MD (i.e., numerically solving the Newton’s law of motion) is computationally expensive. Here, we introduce a surrogate machine learning simulator that is able to predict the dynamics of liquid systems with no prior knowledge of the interatomic potential or nature of the Newton’s law of motion. The surrogate model consists of a graph neural network (GNN) engine that is trained by observing existing MD-generated trajectories. We demonstrate that the surrogate simulator properly predicts the dynamics of a variety of systems featuring very different interatomic interactions, namely, model binary Lennard-Jones system, silica (which features long-term coulombic interactions), silicon (which comprises three-body interactions), and copper-zirconium alloy (which is governed by many-body interactions). The development of machine-learned surrogate simulators that can effectively replace costly MD simulations could expand the range of space and time scales that are typically accessible to MD simulations.