Bridging the Pressure Gap in Hydrogenation of CO 2 to Formate on Cu(100) by Machine Learning Molecular Dynamics and First-Principles Microkinetic Modeling

The discrepancy in the catalytic behaviors in ultrahigh vacuum vs realistic gas environments, known as “pressure gap”, has been a long-standing challenge in the field of surface science. The challenge lies in the limited knowledge of the surface morphology that might undergo reconstruction as a result of exposure to gas and adsorbates. In this study, we address the challenge by applying a set of multiscale simulations that included density functional theory (DFT), machine learning-based molecular dynamics (MLMD), and microkinetic modeling (MKM) simulations, taking the case of hydrogenation of CO2 into formate on the Cu(100) surface. MLMD simulations allow direct observation of hydrogen-induced reconstruction of the Cu(100) surface in the form of a rotated and line-shifted structure even at room temperature. This insight into the surface dynamics is used to provide more realistic structures of the active sites under experimental conditions. Our calculations reveal that this reconstruction significantly influences the CO2 hydrogenation rate on the Cu(100) surface. In particular, the line-shifted reconstruction on the Cu(100) surface brings the energy barrier and the hydrogenation rate much closer to those on the Cu(111) surface, matching the experimentally observed insensitivity of the Cu surface facets to the CO2 hydrogenation rate. On the other hand, the hydrogenation on the pristine Cu(100) surface exhibits a lower barrier compared to the Cu(111) surface, resulting in significant overestimation of the reaction rate compared to the experiment. This study demonstrates that adopting more realistic active sites from in situ observation of simulated catalytic reactions paves the way for a better connection between theory and experiment in the catalysis field.