Material configurations informatics for semiconductor materials research

Even when the crystal structure and alloy composition are fixed, the arrangement of different atoms on crystal sites can significantly affect material properties and formation processes. To address this, we have been developing a research framework called Material Configurations Informatics (MCI), which focuses on the degrees of freedom in atomic configurations through computational science and machine learning approaches. With increasing computational power, large-scale first-principles calculations (DFT) have become feasible. However, the biggest challenge in MCI is the enormous number of possible atomic configurations. We tackle this problem using techniques such as efficient sampling via Bayesian optimization and interaction modeling with approaches like the Ising model. In this presentation, we introduce applications of MCI to issues such as the mixing of different surface reconstructions—an important factor in discussions of crystal growth processes—and the study of semiconductor properties in 2D alloy materials.