Accelerated Optimization of Compositions and Chemical Ordering for Bimetallic Alloy Catalysts Using Bayesian Learning

Abstract

Alloy materials are crucial to various applications, including catalysis and energy storage, due to their superior performance, cost-efficiency, and tunable properties. However, the vast compositional space and complex chemical ordering of alloys pose significant challenges in identifying the optimal material designs. We present an active learning framework utilizing Bayesian optimization to streamline the discovery of high-performance alloy materials. Applying this framework to PtNi oxygen reduction reaction (ORR) catalysts, we successfully identified the global optimal structures featuring a Pt shell and a PtNi core. Our approach was further extended to explore different morphologies and compositions, revealing the most favorable chemical orderings for ORR. This work provides a comprehensive strategy for the accelerated design of multicomponent alloy materials and highlights the critical role of chemical ordering in optimizing the structure–performance relationship, facilitating the development of high-performance catalysts for energy applications.