Manipulating Local Atomic Environment on Solid Surfaces for Catalysis: From Single Atoms to Clusters and High-Entropy Oxides

Abstract

Understanding and engineering local atomic environment of an active site is critical for optimizing catalytic performance. This dissertation investigates how electronic structure, atomic ensembles, lattice strain, and compositional effects influence catalysis at the atomic scale. Using density functional theory (DFT) and machine learning, several catalytic systems are explored. First, Pt single atoms on Gd-doped CeO₂ are shown to enhance CO oxidation due to a new reaction channel opened from the interaction of the Pt single atom with the oxygen vacancy induced by Gd-doping and the CO reactant. Second, Pd dimers on CeO₂ surfaces are optimized for methane combustion through precise control of the Pd–Pd distance on the surface and in the oxygen lattice of CeO2. Third, a PdO trimer on CeO₂ enables efficient CO₂ hydrogenation and C–C coupling, leading valuable C₂ products. Finally, the complex local environment on the surfaces of a medium-entropy oxide is investigated for methane activation using hydrogen adsorption energy (HAE) as a descriptor, revealing key local factors governing adsorption energies; state-of-the-art graph neural network models are compared and DimeNet++ is shown to provide strong predictive performance and transferability for HAEs. Together, these studies provide atomic-level insights into how local environment can be manipulated to design more efficient catalysts for energy and environmental applications.