Crystal Graph

Crystal graph representations are transforming materials science by encoding crystal structures as graphs for machine learning (ML) based property prediction. Current research focuses on improving these representations to capture both local atomic arrangements and global periodic information, employing graph neural networks (GNNs) with architectures like message-passing networks and incorporating multimodal data (e.g., textual descriptions). These advancements enable more accurate and efficient prediction of material properties, accelerating materials discovery and design for various applications, including alloy development and the prediction of material behavior under extreme conditions.