Machine Learning,\nDensity Functional Theory, and Experiments\nto Understand the Photocatalytic Reduction of CO<sub>2</sub> on CuPt/TiO<sub>2</sub>

The photoconversion of CO<sub>2</sub> to hydrocarbons\nis a sustainable\nroute for its transformation into value-added compounds, which is\ncrucial to mitigating energy and climate crises. CuPt nanoparticles\non TiO<sub>2</sub> surfaces have been reported to show promising photoconversion\nefficiencies. For further progress, a mechanistic understanding of\nthe catalytic properties of these CuPt/TiO<sub>2</sub> systems is\nvital. Here, we employ <i>ab initio</i> calculations, machine\nlearning, and photocatalysis experiments to understand the photocatalytic\nreduction of CO<sub>2</sub> on CuPt/TiO<sub>2</sub>. We explore the\nconfigurational space of the CO<sub>2</sub>@CuPt/TiO<sub>2</sub> systems\nand examine their structures and energetics. We find that the CuPt/TiO<sub>2</sub> interface plays a key role in determining CO<sub>2</sub> activation\nand, thus, the conversion to hydrocarbons. The interface stabilizes\n*CO and other intermediates containing CH groups, thus facilitating\na higher activity and selectivity for methane. A bias-corrected machine-learning\ninteratomic potential trained on density functional theory data enables\nthe efficient exploration of the potential energy surfaces of numerous\nCO<sub>2</sub>@CuPt/TiO<sub>2</sub> configurations using basin-hopping\nMonte Carlo simulations, greatly accelerating the study of these photocatalyst\nsystems. Our simulations show that CO<sub>2</sub> preferentially adsorbs\nat the interface, with a C atom bonded to a Pt site and one O atom\noccupying an O-vacancy site. The interface also promotes the formation\nof *CH and *CH<sub>2</sub> intermediates. For confirmation, we synthesize\nCuPt/TiO<sub>2</sub> samples with various compositions, analyze their\nmorphologies and compositions using scanning electron microscopy and\nenergy-dispersive X-ray spectroscopy, and measure their photocatalytic\nactivity. Our computational and experimental findings qualitatively\nagree and highlight the importance of the interface design for the\nselective conversion of CO<sub>2</sub> to hydrocarbons.