Accelerated Discovery of Energy Materials via Graph Neural Network

Abstract

Graph neural networks (GNNs) have rapidly matured into a unifying, end-to-end framework for energy-materials discovery. By operating directly on atomistic graphs, modern angle-aware and equivariant architectures achieve formation-energy errors near 10 meV atom−1, sub-0.1 V voltage predictions, and quantum-level force fidelity—enabling nanosecond molecular dynamics at classical cost. In this review, we provide an overview of the basic principles of GNNs, widely used datasets, and state-of-the-art architectures, including multi-GPU training, calibrated ensembles, and multimodal fusion with large language models, followed by a discussion of a wide range of recent applications of GNNs in the rapid screening of battery electrodes, solid electrolytes, perovskites, thermoelectrics, and heterogeneous catalysts.