Research Projects

Sub-nanometer metal clusters supported on amorphous oxides are active catalysts for several industrial reactions, such as propane dehydrogenation, naphtha reforming, and CO oxidation. These clusters can adopt a variety of different shapes and sizes. Furthermore, the amorphous support introduces a quenched disordered environment, where different cluster shapes and sizes can stabilize at different sites. Because of this two-fold structural disorder their spectra do not produce sharp features, precluding precise characterization and mechanistic studies. As a result, experimental studies rely on trial-and-error approaches to discover new catalysts. Furthermore, the large configurational space of clusters and the disordered support make expensive ab initio computational methods infeasible. Consequently, computational studies continue to make ad hoc assumptions about active site-structure and mechanisms, often representing cluster distributions by single-site models. As a result, these materials remain poorly understood, and we do not have general design principles for catalyst discovery. Here we propose the development of integrated multiscale simulations, machine learning tools, and experiments to characterize the structure and activity of sub-nanometer metal cluster catalysts on amorphous supports. As a first example, we will model the distribution of amorphous SiO₂ supported Pt clusters and their propane dehydrogenation activity. We will develop actively trained machine learned potentials (MLPs) with basin-hopping (a global-minima finding method algorithm) to efficiently discover stable Pt/SiO₂ cluster configurations. We will synthesize Pt/SiO₂ clusters via three different synthesis protocols: ligand-controlled synthesis, single atom bottom-up synthesis, and atomic layer deposition (ALD). The synthesized clusters will be characterized via X-ray absorption (XAS), nuclear magnetic resonance (NMR), and tunneling electron microscopy (TEM) measurements. These measurement will help determine the effect of the synthesis protocol on catalyst structure. Coordination numbers and oxidation states from XAS and NMR, along with particle size distributions from TEM, will inform the computationally discovered ensemble of clusters to develop accurate Pt/SiO₂ models for different synthesis protocols. Furthermore, the discovered ensemble of clusters will be used to calculate site-dependent propane dehydrogenation kinetics and develop structure-activity models with kernel regression. Site-averaged kinetics will be calculated by combining the structure-activity model with TEM-measured particle size distributions and basin-hopping discovered cluster ensembles. Finally, site-averaged kinetic predictions will be compared with the experimentally measured propane dehydrogenation turnover frequency (TOF) of the catalyst. Our investigation will reveal the effect of synthesis protocol on the propane dehydrogenation activity and selectivity of the catalyst. Our framework will, for the first time, enable rigorous in silico modeling of sub-nanometer cluster catalysts on amorphous supports. This approach will provide insights into features of stable and active sites and reveal how synthesis-dependent particle size distributions affect kinetics. Moreover, our framework will be readily extendable to several metal-support combinations, potentially helping design active and selective supported cluster catalysts for various applications.