Four-Dimensional-Spacetime Atomistic Artificial Intelligence Models

TL;DR

This paper introduces a 4D-spacetime AI model that predicts molecular dynamics continuously over time, enabling efficient long-term simulations without stepwise force calculations.

Contribution

The novel 4D-spacetime GICnet model allows for continuous, order-independent predictions of molecular trajectories, surpassing traditional stepwise molecular dynamics methods.

Findings

Accurately predicts nuclear positions and velocities over long times.

Enables high-resolution molecular dynamics simulations efficiently.

Provides insights into nuclear motions and vibrational spectra.

Abstract

We demonstrate that AI can learn atomistic systems in the four-dimensional (4D) spacetime. For this, we introduce the 4D-spacetime GICnet model which for the given initial conditions - nuclear positions and velocities at time zero - can predict nuclear positions and velocities as a continuous function of time up to the distant future. Such models of molecules can be unrolled in the time dimension to yield long-time high-resolution molecular dynamics trajectories with high efficiency and accuracy. 4D-spacetime models can make predictions for different times in any order and do not need a stepwise evaluation of forces and integration of the equations of motions at discretized time steps, which is a major advance over the traditional, cost-inefficient molecular dynamics. These models can be used to speed up dynamics, simulate vibrational spectra, and obtain deeper insight into nuclear motions as we demonstrate for a series of organic molecules.