Learning Long-Range Representations with Equivariant Messages

TL;DR

This paper introduces LOREM, a graph neural network architecture that uses equivariant charges for long-range interactions, enabling consistent and accurate modeling of non-local physical effects in atomic systems.

Contribution

The paper proposes a novel equivariant message-passing neural network, LOREM, that effectively captures long-range physical effects without extensive hyperparameter tuning.

Findings

LOREM outperforms existing methods on datasets emphasizing non-local effects.

Equivariant long-range message passing improves accuracy over scalar approaches.

LOREM demonstrates consistent performance across various datasets.

Abstract

Machine learning interatomic potentials trained on first-principles reference data are becoming valuable tools for computational physics, biology, and chemistry. Equivariant message-passing neural networks, including transformers, achieve state-of-the-art accuracy but rely on cutoff-based graphs, limiting their ability to capture long-range effects such as electrostatics or dispersion, as well as electron delocalization. While long-range correction schemes based on inverse power laws of interatomic distances have been proposed, they are unable to communicate higher-order geometric information and are thus limited in applicability. To address this shortcoming, we propose the use of equivariant, rather than scalar, charges for long-range interactions, and design a graph neural network architecture, LOREM, around this long-range message passing mechanism. We consider several datasets specifically designed to highlight non-local physical effects, and compare short-range message passing with different receptive fields to invariant and equivariant long-range message passing. Even though most approaches work for careful dataset-specific choices of their model hyperparameters, LOREM works consistently without such changes, with excellent benchmark performance.