Thermodynamics and Catalytic Activity of the Reduced Cu on a Cu₂O Surface from Machine Learning Atomic Simulation

Electrochemical carbon dioxide reduction, converting CO 2 into high-value-added chemical products, is a key technology to realize a natural carbon cycle. As one of the most effective catalysts, oxide-derived copper (OD-Cu) is widely used in CO 2 reduction because of its ability to produce hydrocarbons. However, the atomic structure of Cu on the Cu 2 O surface and why such a structure shows more catalytic activity than directly synthesized Cu are still unknown. Here, by using stochastic surface walking global optimization combined with a global neural network potential (SSW-NN) method, we explore the phase diagram of Cu x O, the possible atomic structure of the Cu 2 O/Cu interfaces, the Cu 2 O surface reduction, and the Cu catalytic performance. By continuously deleting a certain amount of O atoms to mimic the reduction process, we find that a metastable phase, namely hcp-Cu, is formed on the Cu 2 O (001) surface because of the good atomic position and lattice match between Cu 2 O (001) and hcp-Cu (110), rather than fcc-Cu (111). The total energy barrier for the reduction of CO 2 to methanol is 1.40 eV on the hcp-Cu (110) surface, much lower than the 1.97 eV on the fcc-Cu (111) surface, indicating the excellent catalytic performance of hcp-Cu. This proposed in situ-formed hcp-Cu not only reconciles the longstanding debate regarding the high catalytic activity of OD-Cu but also guides the rational design of electrochemical CO 2 reduction catalysts via phase engineering.