Resolving the "CO puzzle": Disentangling electronic structure and dynamic effects via an operando dynamics framework

The adsorption and activation of carbon monoxide (CO) on catalytic surfaces are fundamental steps in heterogeneous catalytic processes relevant to energy conversion, environmental remediation, and fine chemical synthesis. A notable theoretical challenge in this context is the "CO puzzle," wherein conventional density functional theory methods incorrectly predict a hollow-site preference for CO adsorption on Pt(111), contrary to experimental observations favoring the top site. This discrepancy has been primarily attributed to limitations in describing the electronic structure; however, recent studies have highlighted the potential role of dynamical effects. The respective contributions of electronic structure and dynamics to site preference remain unresolved. In this study, we systematically examine the interplay between electronic structure and dynamics by employing high-dimensional neural network potential energy surfaces (NN-PESs) integrated with molecular dynamics simulations, using both PBE and vdW-DF functionals. Our results demonstrate that dynamic effects, particularly entropic contributions, are crucial for accurately capturing temperature-dependent adsorption behavior. PBE-based PES simulations predict a preference for hollow-site adsorption at low temperatures but shift toward top-site adsorption at elevated temperatures, aligning with experimental trends. In comparison, vdW-DF-based simulations consistently favor the top site across all temperatures. In addition, comparison with lower-dimensional (6D) rigid PES simulations underscores the importance of adsorbate-surface atoms coupling in accurately modeling thermodynamic properties and site preferences. These findings disentangle the roles of electronic structure and dynamics in resolving the CO puzzle, establishing a robust framework for modeling catalytic processes under operando conditions.