Dual-Site Photocatalysts for Selective CO2 to C2+ Conversion: Pathway Regulation and Machine Learning Guided Design.

Abstract

Photocatalytic CO2 reduction to C2+ products is an attractive route for solar to chemical conversion, yet its development is still limited by inefficient charge utilization, sluggish C─C coupling, competing C1 pathways, and the complex adsorption behavior of carbonaceous intermediates. These challenges become more difficult in dual-site photocatalysts, where metal combinations, coordination structures, intersite distances, support effects, and reaction conditions jointly influence CO2 activation, intermediate evolution, and product selectivity. Machine learning offers a practical way to deal with this complexity by connecting catalyst structures with key descriptors and catalytic performance. In this review, we summarize recent progress in machine-learning-assisted design of dual-site photocatalysts for selective CO2 to C2+ conversion. We first discuss the reaction network of photocatalytic CO2 reduction, especially proton-coupled electron transfer and C─C coupling pathways involving *CO, *CHO, *OCCO, and *CH3 intermediates. We then classify representative dual-site systems, including dual-atom catalysts, single atom-nanocluster/nanoparticle composites, and alloy/interface-based catalysts. Particular attention is given to how machine learning supports descriptor identification, adsorption-energy prediction, catalyst screening, and reaction-pathway analysis, providing useful guidance for developing pathway-oriented dual-site photocatalysts.