Data-driven search for magnetocaloric materials with help of crystral structure data and first principles calculations

2023 Fiscal Year Final Research Report

• Project/Area Number: 20K05070
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Review Section: Basic Section 26010:Metallic material properties-related
• Research Institution: National Institute for Materials Science
• Principal Investigator: TERASHIMA Kensei 国立研究開発法人物質・材料研究機構, ナノアーキテクトニクス材料研究センター, 主任研究員 (20551518)
• Project Period (FY): 2020-04-01 – 2024-03-31
• Keywords: 磁気冷凍材料 / データ駆動型物質探索

Outline of Final Research Achievements

Using a data-driven approach, we identified HoB2 as a material exhibiting significant magnetocaloric effects near hydrogen liquefaction temperatures. Additionally, we clarified the impact of elemental substitution on this material. The machine learning model we developed not only predicts properties of individual materials but also provides insights through model analysis that can guide the search for new materials. To identify synthesized materials, we developed and released software that automates the analysis of X-ray diffraction patterns. This software connects to an external crystal structure database to efficiently analyze samples containing multiple impurities. Moreover, we leveraged neural networks to accurately interpolate preliminary data exhibiting nonlinear changes. This capability allows for real-time optimization of measurement conditions during experiments. We have published these softwares as well.

Free Research Field

データ駆動型機能性材料探索

Academic Significance and Societal Importance of the Research Achievements

データ駆動型の物質探索が、ただの提案にとどまらず合成および実験検証まで行い、磁気冷凍材料という機能性材料を実際に見出すに至ることを示す例となった。機械学習モデルの知見を探索指針に活かす具体的方法と、合成物をX線回折から自動で特定する手段、および予備データを非線形補間することで本測定をシミュレートし効率測定行う手段を開発提示するなど、合成探索を行うユーザー側の視点からのツール開発を行い公開した。