Machine Learning-Guided Discovery of Ce-based Ternary Intermetallics

Cerium-based intermetallics, as potential light rare earth element substitutes for permanent magnets, have drawn significant research attention. We present an integrated ML approach with first-principles calculations to efficiently explore low-energy ternary Ce-Co-Cu compounds. Our study reveals several structures are energetically as well as dynamically stable, along with a number of metastable ones. Notably, two Co-rich metastable compounds exhibit high magnetization, suggesting their potential as doped Ce2Co17-based permanent magnets.