Abstract:

The transition to sustainable chemical processes relies on the development of efficient electrocatalysts for energy conversion reactions. Traditional electrochemical analysis methods are labor-intensive and limit throughput. By integrating data-driven techniques with physics-based modeling, it is possible to accelerate catalyst discovery and mechanistic understanding.  This dissertation develops a suite of computational tools to enhance the analysis, interpretation, and predictive modeling of electrocatalytic systems, with an emphasis on energy conversion reactions. First, supervised machine learning models are applied to identify kinetic trends in the nitrogen reduction reaction. Next, a novel computer vision pipeline is introduced to extract electrochemical parameters from cyclic voltammograms, enabling faster, automated analysis of experimental data. Finally, machine learning models trained on density functional theory data are used to screen oxygen reduction reaction catalysts, culminating in the synthesis and validation of novel catalysts. Collectively, these approaches illuminate catalyst behavior, reduce analysis bottlenecks, and demonstrate practical integration of machine learning in experimental electrocatalysis. This work expands the computational electrochemistry toolbox and establishes methods for data-driven catalyst discovery and autonomous experimentation.