Machine Learning in Metal Alloy Catalyst Design with Controlled Structure Interfaces

Traditionally, the discovery of catalysts relied on trial-and-error in situ methods and later on in silico methods due to ever-growing advancements in computational methods and computing infrastructure. However, a vast chemical space of catalysts makes it highly cumbersome and time-consuming. In recent years, machine learning (ML) has emerged as a promising approach to discovering and designing new catalysts by learning from existing data. Consequently, ML approaches have increasingly improved catalytic processes and are constantly applied in catalysis to unravel their challenging scientific problems. Metal alloying remains a useful and promising design strategy for adjusting the electronic structure of metal catalysts due to the benefits of alloy effects. As a result, ML techniques have been applied in developing and designing new metal alloy catalysts and to expedite the results of testing new catalysts. Besides handling large datasets, ML can provide insights into the stability, activity, and selectivity of metal alloy heterogeneous catalysts. This chapter presents recent developments in the application of ML in metal alloy catalyst design, as well as opportunities and challenges. In addition, an illustration of the application of ML methods to design optimal catalysts for water gas shift is presented.