Single-atom catalysts design for oxygen reduction reaction using machine learning

Abstract

The oxygen reduction reaction is a critical reaction that determines the performance of a fuel cell because of its slow kinetics and large overpotential. In addition, as noble metals are used as catalysts, it is essential to develop inexpensive new catalysts. In this study, we proposed single-atom catalysts(SACs) that can replace platinum through wide range screening. Using Density Functional Theory(DFT), descriptors were defined to express the activity of oxygen reduction reactions occurring on the catalyst, and design SACs. In addition, a methodology to save the computational cost and time cost required for designing by machine learning is presented, and a more effective prediction method(accuracy of 94.6%) than the previous property prediction model(accuracy of 71.4%) is proposed.