Accelerated Data-Driven Discovery and Screening of Two-Dimensional Magnets Using Graph Neural Networks

Abstract

In this study, we employ graph neural networks (GNNs) to accelerate the discovery of novel 2D magnetic materials which have transformative potential in spintronic applications. Using data from the Materials Project database and the Computational 2D materials database, we train three GNN architectures on a dataset of 1190 magnetic monolayers with energy above the convex hull (E hull ) less than 0.3 eV/atom. Our Crystal Diffusion Variational Autoencoder generates 11,100 candidate crystals. Subsequent training on two Atomistic Line GNNs achieves 93% accuracy in predicting magnetic monolayers and a mean average error of 0.039 eV/atom for E hull predictions. After narrowing down candidates based on magnetic likelihood and predicted energy, constraining the atom count in the monolayers to five or fewer, and performing dimensionality checks, we identified 190 candidates. These are validated using density-functional theory to confirm their magnetic and energetic favorability, resulting in 167 magnetic monolayers with E hull < 0.3 eV/atom and a total magnetization of ≥0.5 μ B . Our methodology offers a way to accelerate the exploration and prediction of potential 2D magnetic materials, contributing to the ongoing computational and experimental efforts aimed at the discovery of new 2D magnets.

Virtual Special Issue on Machine Learning in Physical Chemistry Volume 2

“…In JPC C , the special issue papers describe ML and other data science techniques utilized in the scope of nanoparticles and nanostructures; surface and interface processes; electron, ion, and thermal transport; optic, electronic, and optoelectronic materials; and catalysts and catalysis; as well as energy conversion and storage materials and processes. As in the Parts A and B, a number of articles directly address either the development of new methods or the use of ML methods in new ways. − Several contributions relate to methods in the use or development of so-called machine learning potentials (MLP) or machine learning interaction potentials (MLIP), including a Perspective on improving these models authored by Maxson et al, as well as other contributions. − Interfaces and related phenomena are addressed in several articles. − Nanomaterials and their properties are also prominently featured. − Another large category in JPC C includes studies that address the calculation of properties of materials for wide-ranging applications. − …”