2022-Sustainable Industrial Processing Summit
SIPS2022 Volume 9 Mizutani Intl. Symp. Science of Intelligent & Advanced Materials (SISAM) and Quasi-crystals
| Editors: | F. Kongoli, J. Dubois, E. Gaudry, T. Homma, V. Fournee |
| Publisher: | Flogen Star OUTREACH |
| Publication Year: | 2022 |
| Pages: | 116 pages |
| ISBN: | 978-1-989820-50-6(CD) |
| ISSN: | 2291-1227 (Metals and Materials Processing in a Clean Environment Series) |
CD shopping page
Alloy design based on first-principles calculations from deductive to inductive approach
Ryoji Asahi1;1NAGOYA UNIVERSITY, Chikusa-ku, Nagoya, Japan;
Type of Paper: Invited
Id Paper: 414
Topic: 42
Abstract:
The first-principles calculations based on density functional theory (DFT) have succeeded in a broad range of systems thanks to accurate description of electronic structure and their transferability. Here we present some studies on alloys using the first-principles calculations to understand deductively mechanism of properties such as stability and superelasticity. Regarding the stability of the gamma-phase alloys, the Hume-Rother electron concentration rule was revisited in terms of the first-principles calculations. The detailed analysis elucidates an interaction between the Fermi surface and the Brillouin zone that results in pseudogap formation and stability of the system with a certain electron density [1]. The electron density also plays a significant role in Ti-Nb-Ta-Zr-O alloys called “gum metal” which shows high strength, low Young's modulus and high elastic deformability, simultaneously. These unusual properties can be understood by softening with a particular electron concentration and Zr-O nano-clusters to be obstacles for dislocation movement [2].
Despite great success of the first-principles approach, it often faces at computationally-accessible simulation size typically within 1000 atoms and 100 ps. In recent year, machine learning potential (MLP) approach has been developed. Here the use of the first-principles calculations is extremely effective for data generation to construct MLP inductively. We demonstrate the design of RhAu alloy nanoparticles for NO decomposition catalysis using machine-learning approach [3]. A local similarity kernel based on the local atomic configuration is employed as descriptors which allow interrogation of catalytic activities. With data of the first-principles calculations on single crystals and their surfaces, MLP provides size- and composition-dependent catalytic activities of the nanoparticles.
Keywords:
Computational simulation; Electronic structures; machine learningReferences:
[1] Asahi, Sato, Takeuchi, Mizutani, Phys. Rev. B 71, 165103 (2005).[2] Nagasako, Asahi, Isheim, Seidman, Kuramoto, Furuta, Acta Mater. 105, 347 (2016).
[3] Jinnouchi, Asahi, J. Phys. Chem. Lett. 8, 4279 (2017).