Data-driven Exploration of New Two-Dimensional Magnets Using Graph Neural Networks

Two-dimensional (2D) magnets have a transformative potential in spintronics applications. In this study, we employ Graph Neural Networks (GNNs) to discover novel 2D magnetic materials. Using data from the Materials Project database and the Computational 2D materials database (C2DB), we train three GNN architectures on a dataset of approximately 1,200 magnetic monolayers with energy above hull less than 0.3 eV. Our Crystal Diffusion Variational Auto Encoder (CDVAE) generates around 11,000 material candidates. Subsequent training on two Atomistic Line Graph Neural Networks (ALIGNN) achieves a 93% accuracy in predicting magnetic monolayers and a mean average error of 0.04 eV for energy above hull prediction. After narrowing down candidates based on magnetic likelihood and predicted energy above hull, and constraining the atom count in the monolayer to four or fewer, we identified 160 potential materials. These are validated using density-functional theory (DFT) to confirm their magnetic and energetic favorability. Our approach offers a methodical way to explore and predict potential 2D magnetic materials, contributing to ongoing computational and experimental efforts aimed at the discovery of new 2D magnets.