CTGNN: Crystal Transformer Graph Neural Network for Crystal Material Property Prediction

• DOI: 10.48550/arxiv.2405.11502
• Publication Date: 2024-05-19
• Authors: Zijian Du; Luozhijie Jin; Le Shu; Yan Cen; Yuanfeng Xu; Yongfeng Mei; Hao Zhang
• Keywords: Crystal (programming language)

The combination of deep learning algorithm and materials science has made significant progress in predicting novel understanding various behaviours materials. Here, we introduced a new model called as the Crystal Transformer Graph Neural Network (CTGNN), which combines advantages graph neural networks to address complexity structure-properties relation material data. Compared state-of-the-art models, CTGNN incorporates network structure for capturing local atomic interactions dual-Transformer structures intra-crystal inter-atomic relationships comprehensively. benchmark carried on by proposed indicates that significantly outperforms existing models like CGCNN MEGNET prediction formation energy bandgap properties. Our work highlights potential enhance performance properties accelerates discovery materials, particularly perovskite