A Universal Graph Deep Learning Interatomic Potential for the Periodic Table

TL;DR

This paper introduces M3GNet, a universal graph neural network-based interatomic potential trained on a large database, enabling accurate and broad applications in material simulations and discovery of stable, synthesizable materials.

Contribution

The paper presents M3GNet, a novel universal interatomic potential that covers the entire periodic table, trained on extensive data, and validated for stability and property prediction.

Findings

Identified 1.8 million potentially stable materials from 31 million candidates.

Verified stability of 1578 materials using DFT calculations.

Demonstrated M3GNet's effectiveness in structural relaxation and property prediction.

Abstract

Interatomic potentials (IAPs), which describe the potential energy surface of atoms, are a fundamental input for atomistic simulations. However, existing IAPs are either fitted to narrow chemistries or too inaccurate for general applications. Here, we report a universal IAP for materials based on graph neural networks with three-body interactions (M3GNet). The M3GNet IAP was trained on the massive database of structural relaxations performed by the Materials Project over the past 10 years and has broad applications in structural relaxation, dynamic simulations and property prediction of materials across diverse chemical spaces. About 1.8 million materials were identified from a screening of 31 million hypothetical crystal structures to be potentially stable against existing Materials Project crystals based on M3GNet energies. Of the top 2000 materials with the lowest energies above hull, 1578 were verified to be stable using DFT calculations. These results demonstrate a machine learning-accelerated pathway to the discovery of synthesizable materials with exceptional properties.