清華大學材料科學與工程研究院《材料科學論壇》學術報告

報告題目：Accelerating Nanomaterial Simulations Using Machine Learning Techniques
報告人：Satoshi Watanabe教授（東京大學）

報告時間：2019年11月20日（周三） 下午15:30-16:30

報告地點：清華大學逸夫技術科學樓205報告廳

聯系人：潘偉教授 62772858

歡迎廣大師生踴躍參加！

報告摘要：
Accelerating Nanomaterial Simulations Using
Machine Learning Techniques
Satoshi Watanabe
 
Department of Materials Engineering, Graduate School of Engineering,
The University of Tokyo, Tokyo, Japan, and
Research and Services Division of Materials Data and Integrated System,
NIMS, Tsukuba, Japan.
 
Keywords: Neural Network, Interatomic potential, density functional theory,
ion diffusion, thermal transport
 
Progress in computer and methodologies enables reliable simulations on various topics concerning structures and properties of nanomaterials and phenomena occurring in nanomaterials using first-principles methods such as density functional theory (DFT). However, some of important topics still demands practically too heavy computations, such as ion diffusion in amorphous materials, thermal transport in defective materials, atomic structures of complex interfaces, etc.
Recently, machine learning techniques have attracted much attention because of their capability to achieve computational efficiency and reliability simultaneously in nanomaterial simulations. In this talk, I will demonstrate their capability with showing our recent results on the construction and applications of interatomic potentials optimized using DFT calculation data and one of machine-learning techniques, neural network. I will mainly discuss Li ion diffusion in amorphous Li3PO4 [1] and thermal conductivities of wuitize GaN and silicon crystals [2], and will also touch on the Born effective charges [3], atomic structure of metal/solid-electrolyte interface [4], and atomic energy mapping [5].
 
The works of Refs. [1] to [4] were done in collaboration with some of present/previous members in my group, K. Shimizu, T. Moriya, M. Ogura, W. Liu, W. Li, Y. Ando, E. Minamitani, and that of Ref. [5] with Prof. S. Han and some of members in his group, D. Yoo, K. Lee, W. Jeong. The works were supported in part by MI2I project of the Support Program for Starting Up Innovation Hub from JST, CREST-JST, PRESTO-JST and JSPS KAKENHI, Japan.
 
REFERENCES
[1] W. Li, et al.: J. Chem. Phys., 147, 214106 (2017).
[2] E. Minamitani, et al.: Appl. Phys. Express, 12, 095001 (2019).
[3] T. Moriya: Master thesis, The University of Tokyo (2019) [in Japanese].
[4] K. Shimizu, et al.: in preparation.
[5] D. Yoo, et al.: Phys. Rev. Mater., 3, 093802 (2019).