Black-box optimization technique for investigation of surface phase diagram

Abstract

Surface phase diagrams are useful in material design for understanding catalytic reactions and deposition processes and are usually obtained by numerical calculations. However, a large number of calculations are required, and a strategy to reduce the computation time is necessary. In this study, we proposed a black-box optimization strategy to investigate the surface phase diagram with the smallest possible number of calculations. Our method was tested to examine the phase diagram in which two types of adsorbates, i.e., oxygen and carbon monoxide, were adsorbed onto a palladium surface. In comparison with a random calculation without using machine learning, we confirmed that the proposed method obtained a surface phase diagram with a small number of calculations. In conclusion, our strategy is a general-purpose method that can contribute to the rapid study of various types of surface phase diagrams.