Tuning Properties of Heterogeneous Catalysts using First Principles and Machine Learning

Abstract

The surface-to-volume ratio plays a crucial role in determining the catalytic performance of heterogeneous catalysts by influencing surface energy, reactivity, and stability. This thesis systematically explores how tuning intrinsic properties and external parameters affects both the activity and stability of catalysts, aiming to provide a comprehensive framework for designing efficient and versatile catalytic systems. The research employs density functional theory (DFT), ab initio molecular dynamics, and machine learning to analyze various catalytic materials. Studies on OH-functionalized Ti₂C MXenes reveal distinct pathways for formic acid and formate formation during CO2 reduction, while metal doping and functional group modifications are shown to steer product selectivity. Investigations on CoFeNiCuMn high-entropy alloy nanoclusters supported on CeO2 elucidate CH4 dry reforming mechanisms, supported by d-band center and orbital analyses. SnPt bimetallic nanoparticles on oxide supports are evaluated for stability improvements using Random Forest models, providing design guidelines for optimal nanoparticle-support configurations. Additionally, the thesis examines C2 product formation on Cu-based intermetallic surfaces, with a 3,000-point adsorption database analyzed through LightGBM modeling (R2 = 0.89) to identify active sites and enhance C–C coupling efficiency in CO2 reduction. This work bridges computational modeling with predictive analytics, offering strategies to optimize both activity and stability in heterogeneous catalysts. The findings contribute to the rational design of next-generation catalysts for sustainable energy and chemical applications.