Computational Studies of Pd-based Hydride Catalysts for CO₂ Reduction

Emission of the greenhouse gas carbon dioxide (CO₂) has increased rapidly with the development of industrialization in the past decades. The electrochemical CO₂ reduction reaction (CO₂RR) is considered as a promising strategy to convert CO₂ into valuable chemicals. Palladium (Pd)-based hydride catalysts hold promise for producing syngas via both CO₂RR to CO and hydrogen evolution reaction (HER) to H₂, which can be effectively synthesized into valuable chemicals. In this thesis, Pd-based hydride catalysts are systematically studied to improve the CO₂RR performance of pure Pd hydride (PdH) by doping and alloying methods with various theoretical approaches.

First, density functional theory (DFT) is used to systematically screen for stability, activity, and selectivity of transition metal dopants in PdH(111). Sc, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, Y, Zr, Nb, Mo, Ru, Rh, Ag, Cd, Hf, Ta, W, and Re are doped into PdH surface with six different doping configurations: single, dimer, triangle, parallelogram, island, and overlayer. We find that several dopants, such as Ti and Nb, have excellent predicted catalytic activity and selectivity towards CO₂RR compared to the pure PdH.

Second, using DFT calculations in combination with active learning cluster expansion (ALCE) and Monte Carlo simulated annealing (MCSA), we identify 12 stable PdH_x(111) configurations on the DFT convex hull and investigate the binding energies of intermediates during the CO₂RR and the competing HER. Through analysis of intermediate binding energies and a microkinetic model, we identify the atomic structures of the PdH_x phase most likely to produce syngas. The high activity of the PdH_0.6 surface can be attributed to the fact that the H segregation in the PdH_x(111) surface breaks the linear relation between HOCO* and CO* adsorbates.

Third, an ALCE surrogate model equipped with MCSA, CO* filter and the kinetic model are used to screen out excellent Pd_xTi_{1 – x}H_y and Pd_xNb_{1 – x}H_y catalysts with both high stablity and superior activity. Since the calculations of the convex hull are finally verified by DFT and all binding energy calculations of the limited candidates are also calculated by DFT, the calculations of all enregies are more reliable compared to those only using the surrogate model with uncertainty. Finally, the stable and active 24 candidates of Pd_xTi_{1 – x}H_y and 6 active candidates of Pd_xNb_{1 – x}H_y are found according to our approach.

Finally, a deep learning-assisted multitasking genetic algorithm is used to screen Pd_xTi_{1 – x}H_y surfaces containing multiple adsorbates for CO₂RR under different reaction conditions. The ensemble deep learning model can greatly speed up the structure relaxations and keep a high accuracy as well as low uncertainty of energies and forces. The multitasking genetic algorithm is used to simultaneously globally find stable surface structures at each reaction condition. Finally, 23 stable structures are screened out under different reaction conditions. Among them, Pd_0.56Ti_0.44H_1.06+25%CO, Pd_0.31Ti_0.69H_1.25+50%CO, Pd_0.31Ti_0.69H_1.25+25%CO, and Pd_0.88Ti_0.12H_1.06+25%CO are found to be very active for CO₂RR and suitable to generate syngas.