ホーム > 研究活動 > 口頭発表(2018) > Search for magnet compounds using Bayesian optimization

研究活動

Search for magnet compounds using Bayesian optimization

NIMS学術シンポジウム「磁性材料イノベーション」、東京国際フォーラム

2018年10月15日(月)

深澤太郎(産総研)、原嶋庸介(産総研)、三宅隆(産総研)

Abstract

Computer-aided materials-search is attracting much attention due to growing demands for high performance functional materials. Those based on informatics and massive data are especially considered to improve search efficiency significantly. In the realization, we need to obtain accurate data to describe magnetic properties of magnet compounds.

To provide a data set that represents a search space of practical interest, we have performed first-principles calculation of (R, Z) (Fe, Co, Ti)12 with the ThMn12 structure for 3630 systems (R=Y, Nd, Sm; Z=Zr, Dy). Then, we obtained theoretical values of magnetization, total energy, and Curie temperature within mean-field approximation.

In this paper, we show results of the Bayesian optimization using the data for (R, Z) (Fe, Co,Ti)12 mentioned above, and discuss how it has advantage in search efficiency over the random search. We also propose an interpolation technique for calibrating calculated quantities with values from experiments or other theoretical calculations to provide more accurate data to the optimization scheme.


研究活動

文部科学省

文部科学省
元素戦略プロジェクト(活動紹介)
NIMS磁石パートナーシップ

元素戦略拠点

触媒・電池元素戦略拠点
触媒・電池元素戦略研究拠点 (京都大学)
東工大元素戦略拠点
東工大元素戦略拠点 (東京工業大学)
構造材料元素戦略研究拠点
構造材料元素戦略研究拠点 (京都大学)
高効率モーター用磁性材料技術研究組合
高効率モーター用 磁性材料技術研究組合