Mapping the catalytic landscape of doped Pd(111) for formic acid synthesis via CO2 hydrogenation using first-principles, microkinetics, and SISSO descriptors

Abstract

Palladium (Pd) is a highly promising catalyst for the carbon dioxide (CO₂) hydrogenation reaction, yet the mechanism by which specific transition metal dopants regulate the catalytic activity and guide the reaction selectivity of the Pd(111) crystal plane remains unresolved. This study integrates density functional theory (DFT) calculations, temperature-dependent microkinetic simulations, and the SISSO machine learning algorithm. DFT was employed to construct pure Pd(111) and eight transition-metal-doped (Ag, Co, Cu, Mn, Ni, Zn, Pt, and Fe) Pd(111) models, for which the adsorption energies of intermediates, electronic structure evolution, and key reaction barriers were calculated. Microkinetic simulations were used to predict reaction rates and product selectivity, whereas SISSO was applied to identify key descriptors and construct quantitative activity models. Fe-doped Pd(111) exhibited the highest intrinsic activity at elevated temperatures, being four orders of magnitude higher than that of pure Pd at 500 K. Pt- and Cu-doped surfaces favored methanol formation, Fe- and Co-doped surfaces were biased toward formic acid production, whereas Mn- and Ni-doped surfaces exhibited over 90% CO selectivity, leading to active-site poisoning. The SISSO-derived descriptors, which incorporate adsorption energies and temperature effects, accurately reproduced the microkinetic trends (R² > 0.98, RMSE < 1.0). This integrated “DFT–microkinetics–SISSO” framework systematically maps the catalytic landscape of doped Pd(111) for CO₂ hydrogenation, elucidates dopant-regulation mechanisms, and offers a quantitative toolkit for rapid Pd-alloy screening, bridging atomic-scale insights with macroscopic catalytic performance.