Investigation of crystalline interface using machine learning

Abstract

Owing to the recent trend of "Materials Informatics", we can easily access excellent databases that provide a variety of information on materials. In contrast, such databases provide only single crystal bulk information, and do not provide information on lattice defects. The group of the present authors succeeded in dramatically improving the process of determining the crystal interface structures using machine learning techniques such as virtual screening, Bayesian optimization, and transfer learning. This paper will introduce the results of our studies.