Embracing data science in catalysis research

Abstract

Accelerating catalyst discovery and development is of paramount importance in addressing the global energy, sustainability and healthcare demands. The past decade has witnessed significant momentum in harnessing data science concepts in catalysis research to aid the aforementioned cause. Here we comprehensively review how catalysis practitioners have leveraged data-driven strategies to solve complex challenges across heterogeneous, homogeneous and enzymatic catalysis. We delineate all studies into deductive or inductive modes, and statistically infer the prevalence of catalytic tasks, model reactions, data representations and choice of algorithms. We highlight frontiers in the field and knowledge transfer opportunities among the catalysis subdisciplines. Our critical assessment reveals a glaring gap in data science exploration in experimental catalysis, which we bridge by elaborating on four pillars of data science, namely descriptive, predictive, causal and prescriptive analytics. We advocate their adoption into routine experimental workflows and underscore the importance of data standardization to spur future research in digital catalysis. The use of data science tools in catalysis research has experienced a surge in the past 10–15 years. This Review provides a holistic overview and categorization of the field across the various approaches and subdisciplines in catalysis.