Computational Modeling of Activation and Deactivation of Supported Metal Catalysts

Supported metal catalysts are widely used in heterogeneous catalysis applications such as automobile exhaust gas treatment, petrochemical processes, and biomass conversion. The transition metals that are often used in catalysts are either deposited on support as metal (or oxide) particles or exchanged on or into the support in their cationic form. While both cationic and particle forms could be active or different reactions, their catalytic performances are widely different. Further, cationic species exchanged into zeolite supports can interconvert between multiple types of monomeric, dimeric, and metallic species under reaction conditions. We used computational methods such as Density Functional Theory (DFT), first principles-based thermodynamic calculations, and Monte Carlo (MC) simulations to interrogate the Cu cation speciation in five zeolite frameworks (CHA, MOR, BEA, AFX, and FER). Our thermodynamic analysis shows that frameworks such as MOR possess more exchange site locations that prefer binuclear Cu species, leading to higher populations of Cu dimers than in zeolites such as CHA. Cu exchange free energies calculated for these five zeolites were then extrapolated to 200 other zeolite frameworks using a machine learning model trained on our calculations.
Secondly, we modeled interconversion between metal (oxide) nanoparticles and metal cations observed for Pd exchanged zeolites under high-temperature (593–973 K) oxidizing conditions. Kinetic measurements show that the initial Pd particle size distribution strongly determines the rates of Pd nanoparticle redispersion into Pd²⁺ cations. We explained this strong size dependency quantitatively by combining classical particle growth (Ostwald ripening) theory with an atom trapping reaction. The same mathematical model was then extended to explain Pt metal redispersion (into Pt²⁺) reported for Pt-CHA, and our kinetic model shows an excellent agreement with experimental data. Moreover, the model reveals the metal particles encapsulated in zeolites have low surface energy, decreasing the driving force for Ostwald ripening and leading to longer catalyst lifetimes.
Thirdly, we developed a model for atom trapping processes using a statistical mechanics approach to evaluate the equilibrium state of atom trapping systems. Our results show that the configurational entropy of trapping sites can drive metal particles into the trapping sites on the lattice, even with thermodynamically uphill trapping reaction energies. These findings are consistent with experimental observations and help design catalytic materials with stable performance.
Next, we studied the sintering of Pd supported on Al2O3 under oxidative (10 kPa O₂, 10 kPa H₂O, balance N₂ at T= 1073 K) and reductive (2 kPa H₂, 10 kPa H₂O, balance N₂ at T= 1073 K) gas conditions and initial size distributions. The growth of particle sizes over time (known as sintering) is experimentally quantified using O₂ chemisorption, TEM imaging, and C₂H₄ hydrogenation. Then, we used an Ostwald ripening sintering model to extract kinetic parameters from the experimental sintering data (size vs time). The extracted kinetic parameters revealed that under oxidative conditions, the decreased surface energy of particles (PdO) causes slower sintering, consistent with previous experimental and theoretical observations. Taken together, the mathematical models we developed help estimate the population of cationic and metallic species as a function of material composition and reacting gas environment to help design catalytic materials with improved performance.