Machine learning assisted hierarchical filtering: a strategy for designing magnets with large moment and anisotropy energy

Abstract

Machine learning models are developed to hierarchically filter and select stable magnetic materials with large magnetization and magnetic anisotropy energy. Starting from an initial set of 278 materials, 10 are identified by the models to satisfy the desired target properties. Subsequent first principles calculations find 7 stable compounds with large moment and high anisotropy energy. This machine learning assisted filtering procedure reduces computational cost by more than an order of magnitude. Over an expanded search space, it is expected to lead to even more dramatic reduction in computational time. This method can find practical use in screening candidate materials for 2D magnets with high magnetization, and also in design of permanent magnets without rare earth elements.