Computational exploration of strong permanent magnet compounds

Abstract

Modern strong magnets are rare-earth magnets in which high saturation magnetization andhigh Curie temperature come from transition-metal 3d electrons, and strongmagnetorcystalline anisotropy originates from rare-earth 4f electrons [1]. There are varioustypes of crystal structures and chemical composition, and exploration of a new magnetcompound is a hot topic. Among them, RFe12-type compounds with the ThMn12 structure areattracting renewed interest because of their high iron content. Recently synthesizedNdFe12Nx film has higher saturation magnetization and anisotropy field than Nd2Fe14B,although its bulk phase is thermodynamically unstable. I will present a first-principles study onthe effect of element substitution on magnetism and structural stability. I will also discuss howmachine learning accelerates magnetic-materials discovery. Application to the Curietemperature of RFe12-type compounds shows that Bayesian optimization offers an efficientway to optimize chemical composition of magnet compounds. Kernel ridge regression usingorbital-field matrix as a descriptor reproduces the magnetic moment and formation energy ofthousands of transition-metal compounds in reasonable accuracy [2], which can be utilized forvirtual screening of new magnetic compounds. Bayesian optimization approach to crystalstructure prediction is also presented [3].
[1] Takashi Miyake and Hisazumi Akai, J. Phys. Soc. Jpn. 87, 041009 (2018).
[2] T.L. Pham et al., Sci. Tech. Adv. Mater. 18, 756 (2017).
[3] T. Yamashita et al., Phys. Rev. Mater. 2, 013803 (2018).