Machine learning assisted prediction of copper-based catalysts towards carbon dioxide electroreduction into carbon monoxide

Abstract

Copper-based catalyst is very active for electroreduction of carbon dioxide (CO2) to carbon monoxide (CO). However, the Faraday efficiency of copper-based catalysts for CO production can be affected by many complex factors, such as catalyst elemental composition, morphology, supporting substrate, synthesis method, catalyst size, electrolyte concentration, and test potential with unknown correlations, hindering the efficient exploration of active copper-based CO2 reduction catalysts with high CO Faraday efficiency. In this study, a machine learning (ML) model is proposed towards rapid screening of copper-based catalysts with high CO selectivity using experimental database. The significance of different catalyst-related features on the CO Faraday efficiency was evaluated using feature importance and Spearman correlation coefficients. The ML model predicted that the dendrite Pd-doped copper oxides catalyst electroless deposited on copper mesh has the highest CO Faraday efficiency (92.8%). This ML study provides a useful guidance for the screening of copper-based catalysts towards the reduction of CO2 to CO.