Accelerated screening for permanent magnetic materials with high Curie temperature using deep learning

Abstract

Permanent magnetic materials are one of the key components in the diverse fields of modern industry. Unfortunately, most of the high-performance permanent magnets heavily rely on rare-earth elements, which suffer from limited and localized supply. When designing high-performance permanent magnet materials, one of the biggest obstacles is to develop high-performance, high Curie temperature permanent magnet materials that are comparable to permanent magnets with rare-earths. Curie temperature (Tc) is also one of the important physical properties of permanent magnetic materials. However, calculating Tc using first-principles calculations is a very difficult task. To solve this problem, we use the Crystal Graph Convolutional Neural Network (CGCNN) to predict the properties of materials with Density Functional Theory (DFT) level accuracy. Applying our model to a large group of material candidates, we designed a new non-rare earth permanent magnet material with high Tc.