UC Davis

In this era driven by data science and artificial intelligence, traditional computational catalysis has undergone a profound revolution facilitated by a variety of machine learning and deep learning techniques. For the new generation of computational theorists and experts in the field of catalysis and surface science, it is crucial to grasp how data-driven models and workflows can seamlessly integrate into scientific research domain. This dissertation explores two primary categories of studies in modern catalysis informatics. Type I studies are centered around predicting specific physical properties related to and playpivotal roles in governing various aspects of catalyst performance, including reactivity, selectivity, and activity. Type II studies, on the other hand, focus on the development of Machine-learning Potentials (Force Fields). These are meticulously engineered to extend the applicability of computationally intensive ab-initio methods (density functional theory, conventional molecular dynamics, and metadynamics) to larger catalytic systems and longer simulation timeframes for enhanced transferability and scalability. The overarching objective of both studies is to extract valuable structural, thermodynamic, and kinetic insights from catalytic systems that were previously inaccessible or inefficient using traditional simulation methods. This dissertation is organized into 5 chapters. Chapter 1 serves as an extensive review, akin to a concise textbook, covering data-driven research methods in catalysis informatics. It places particular emphasis on multiscale modeling techniques, descriptor (fingerprint) design for complex molecular systems, the foundational theory and typical algorithms of machine learning and deep learning, and the workflow for developing Machine-learning Potentials. Chapters 2 - 4 comprise the papers where I served as the first (or co-first) author, delving into accelerating the research progress in catalysis informatics. The final chapter 5 offers a summary and outlines potential avenues for future research in the field of catalysis informatics.