Solving Newton's Equations of Motion with Large Timesteps using Recurrent Neural Networks based Operators

TL;DR

This paper introduces neural network-based operators that solve Newton's equations of motion with significantly larger timesteps, enabling faster molecular dynamics simulations while conserving energy.

Contribution

The authors develop recurrent neural network operators that accurately solve Newton's equations with large timesteps, surpassing traditional integrators in speed and efficiency.

Findings

Achieved up to 4000 times larger timesteps than Verlet integrator.

Demonstrated speedup in 3D systems of up to 16 particles.

Produced energy-conserving particle trajectories using neural operators.

Abstract

Classical molecular dynamics simulations are based on solving Newton's equations of motion. Using a small timestep, numerical integrators such as Verlet generate trajectories of particles as solutions to Newton's equations. We introduce operators derived using recurrent neural networks that accurately solve Newton's equations utilizing sequences of past trajectory data, and produce energy-conserving dynamics of particles using timesteps up to 4000 times larger compared to the Verlet timestep. We demonstrate significant speedup in many example problems including 3D systems of up to 16 particles.