Prediction of Large Magnetic Moment Materials With Graph Neural Networks and Random Forests

TL;DR

This paper employs advanced machine learning techniques, including graph neural networks and random forests, to identify potential large magnetic moment materials from extensive crystal structure databases, aiding the discovery of new magnetic materials.

Contribution

It introduces a combined approach using GNNs and random forests trained on high-throughput DFT data to predict large magnetic moment materials from the ICSD database.

Findings

Identified 15 candidate materials with large magnetic moments not yet experimentally studied.

Achieved comparable prediction accuracy between neural networks and random forests.

Provided estimates of error margins for different machine learning models.

Abstract

Magnetic materials are crucial components of many technologies that could drive the ecological transition, including electric motors, wind turbine generators and magnetic refrigeration systems. Discovering materials with large magnetic moments is therefore an increasing priority. Here, using state-of-the-art machine learning methods, we scan the Inorganic Crystal Structure Database (ICSD) of hundreds of thousands of existing materials to find those that are ferromagnetic and have large magnetic moments. Crystal graph convolutional neural networks (CGCNN), materials graph network (MEGNet) and random forests are trained on the Materials Project database that contains the results of high-throughput DFT predictions. For random forests, we use a stochastic method to select nearly one hundred relevant descriptors based on chemical composition and crystal structure. This gives results that are comparable to those of neural networks. The comparison between these different machine learning approaches gives an estimate of the errors for our predictions on the ICSD database. Validating our final predictions by comparisons with available experimental data, we found 15 materials that are likely to have large magnetic moments and have not been yet studied experimentally.