[SIS14] DX in PM

[16E-SIS14-01] Keynote: Informatics Approaches to Materials Science and Engineering

*I. Tanaka^{1, 2} (1.Kyoto University, Japan, 2.Japan Fine Ceramics Center, Japan)

キーワード：first principles calculations, recommender system, interatomic potentials

Materials Informatics (MI) is highly effective in accelerating the solution of certain materials problems. Materials data are of critical importance in MI. Such data can be obtained from databases/literature of experimental data, from parallelized/combinatorial experiments, or from systematically performed first-principles calculations. Two types of approaches can be used to exploit the data: One is the principle-driven approach based on physical models, thermodynamics laws, etc. The other is the data-driven approach, which is useful when the origin of the materials problems is unknown or complicated multi-scale/multi-physics phenomenon. Here I show some examples of successful applications of MI. One is the development of highly accurate machine learning potentials. The other is the recommender system useful for efficient discovery of currently unknown inorganic ionic compounds. The recommender system for predicting successful processing conditions for the new inorganic compounds based on our parallel experiment data set is also described.