20th International CODATA Conference (Times New Roman, 11pt)
Session: Computational informatics: integrating data science with materials modeling

Integrating Data and Modeling in Materials Design

Ying Chen¹, Yasunori Kaneta¹, Shuichi Iwata²

¹Department of Quantum Engineering and System Science

School of Engineering, The University of Tokyo

ying@q.t.u-tokyo.ac.jp

²Graduate School of Frontier Sciences

The University of Tokyo

With the spectacular development of information technology and computational techniques, people have been dreaming to design materials with specific properties from all possible atomic combinations by using powerful computational physics. However, the huge amount of possibilities to be treated is still out of the manageable scale. Nowadays, a combination of two types of approach, the data-driven approach based on the comprehensive materials database and the model-driven approach based on the theoretical calculation corresponding to various physical models, is proposed as an efficient way to accelerate the speed of finding target atomic configuration from huge number of candidates.

Mapping approach is utilized by mining the data in materials databases for discovering the regularities and correlations on the compound formation, constitution and crystal structures, which directly provide hints on candidate materials in preliminary stage. On the basis of regularities revealed by data mining, various theoretical approaches are applied to the target substances for investigating further their structural stability, phase equlibrium, and physical properties in order to estimate the possibility to be new materials and to understand the insight into the origin of those regularities.

Several examples are presented to show how this combined approach is attempted towards materials design.