First-principles thermodynamics for optimal design of atomic structure and properties of grain boundaries in ceramic materials

2022 Fiscal Year Final Research Report

• Project/Area Number: 21K14405
• Research Category: Grant-in-Aid for Early-Career Scientists
• Allocation Type: Multi-year Fund
• Review Section: Basic Section 26020:Inorganic materials and properties-related
• Research Institution: Nagoya University
• Principal Investigator: Yokoi Tatsuya 名古屋大学, 工学研究科, 講師 (70791581)
• Project Period (FY): 2021-04-01 – 2023-03-31
• Keywords: 粒界 / 第一原理計算 / 機械学習 / 熱力学

Outline of Final Research Achievements

This work attempted to integrate density-functional-theory (DFT) calculations and machine-learning models and to efficiently determine free energies of grain boundaries, in order to understand their thermodynamic stability in ceramics materials. For this purpose, artificial-neural-network potentials were implemented and then combined with lattice dynamics and molecular dynamics simulations.In order to evaluate the predictive power,grain boundaries of Al were chosen as model systems,and their lattice vibrational modes and energetics were predicted by the ANN potential. As a result,our ANN potential was found to accurately predict the finite-temperature properties with the DFT level and practical computational cost, even for grain boundaries absent in training datasets.

Free Research Field

材料科学

Academic Significance and Societal Importance of the Research Achievements

粒界の熱力学的安定性の微視的理解は、有限温度における多結晶材料の材料組織や巨視的特性を緻密制御する上で必須である。しかし理論解析では莫大な計算コストを要するため、系統的な知見は無く、有効な解析手法も確立されていなかった。本研究では第一原理計算と機械学習を組み合わせ、粒界特性を予測する機械学習型原子間ポテンシャルを構築し、その有用性を実証した。これにより、粒界の熱力学的安定性を高効率かつ高精度で予測する技術基盤が確立できた。今後、この手法を酸化物セラミックスを含む種々の材料に展開していくことで、粒界研究の発展に大きく貢献することが期待される。