Query-and-Learn Machine Learning framework to model the stability mechanism of REFe12 magnets

2023 Fiscal Year Final Research Report

• Project/Area Number: 21K14396
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 26010:Metallic material properties-related
• Research Institution: Japan Advanced Institute of Science and Technology
• Principal Investigator: NGUYEN DuongNguyen 北陸先端科学技術大学院大学, 先端科学技術研究科, 助教 (20879978)
• Project Period (FY): 2021-04-01 – 2024-03-31
• Keywords: materials discovery / structure prediction / rare-earths magnets / Dempster-Shafer theory / Sm-Fe magnets

Outline of Final Research Achievements

This research aims to build a query-and-learn Machine Learning frameworks to investigate the stability mechanism of REFe magnets with RE as rare-earth element. There are three main goals were achieved. Firstly, this research developed an active learning-based framework to optimize the quantum data querying cost in materials discovery. Quantum calculations for SmFe12 with ThMn12-type magnets with various substitution elements were systematically investigated. Secondly, this research developed an evolutionary algorithm-based framework to optimize multi-objective crystal structure search. The crystal structure-stability relationship of multiple Sm-Fe families was investigated. Lastly, this research developed an evidence-based similarity measure for materials regarding physical property. The method incorporates measurement uncertainty into the similarity measure under the Dempster-Shafer evidence theory

Free Research Field

Materials Informatics

Academic Significance and Societal Importance of the Research Achievements

This work contributes to the research of rare-earth iron-based magnetic materials by (1) finding optimal materials composition of high-performance magnet discovery and (2) introducing Machine Learning frameworks with standardized materials discovery process and results interpretation.