Modeling Complex Materials with Chemical Bond Hierarchy: A Machine-Learning Interatomic Potential Approach
Prof. Wen-Qing Zhang
Southern University of Science and Technology

Modeling complex materials by machine learning approach has attracted much attention in recent years. Developing robust and reliable machine-learning-based interatomic potentials (ML-IPs) has become increasingly important in studying physical property such as lattice thermal transport in materials with complex structures and chemical bondings. In this work, ML-IPs based on dual adaptive sampling or even multiple adaptive sampling method is developed, with the automatically generated training dataset covering a wide spectra of temperatures and varying thermodynamic conditions. In comparison with the random approach in settinging training set, the above strategy could avoid the trial-and-error uncertainty and lead to high-quality ML-IPs. The Green-Kubo simulations with the above ML-IPs (GK-MLIP) predicted the lattice thermal conductivities and their temperature dependence for CoSb3 with up to 3rd-order phonon scattering, BAs and diamond with very high lattice thermal conductivities, Mg3Sb2 with up to 4th-order phonon scattering, liquid-like materials with infinity phonon nonlinearity, and TE alloys with different composition ratios. All results show good agreement with experiments. We also reveal that the GK-MLIP approach fails in the evaluation of lattice thermal conductivity of liquid-like materials such as β-Cu2-xSe with severe atomic migration as exacerbated by the Cu diffusion at elevated temperatures. This could be attributed to intrinsic problem of the ambiguous projection of local atomic potential energy in ML-IPs. General understanding for the temperature dependence of lattice thermal conductivity of different materials with extremely strong phonon nonlinearity is provided. Discussion on modeling complex materials with chemical bond hierarchy will also be briefly discussed in this talk.

About the Speaker

张文清, 南方科技大学教授, 国家杰出青年科学基金获得者, 2014年当选美国物理学会会士(APS Fellow)。 主要从事计算材料物理和材料设计、电热输运和高性能热电能量转化材料研究等。中国材料研究学会-热电材料及应用分会主任理事、中国材料研究学会-计算材料学分会副主任理事; Appl. Phys. Lett.杂志副编辑(Associate editor), Chinese Physics Letters, J. Materiomics, npj Comput. Mater.等杂志编委。

2023-10-20 10:00 AM
Room: A403 Meeting Room
CSRC 新闻 CSRC News CSRC Events CSRC Seminars CSRC Divisions