Enhanced CO ₂ Electroreduction to Ethylene and Acetamide: Modulating the Microenvironment of CuAg by Imidazolium Salts via Modeling and Machine Learning

Electrocatalytic CO₂ reduction (eCO₂R) to high-value multicarbon (C₂₊) hydrocarbons such as ethylene and acetamide via C-C/N coupling is an attractive and effective technique for achieving zero carbon emissions and advancing renewable energy. Recent studies report the use of engineered microenvironments formed via imidazolium salts (ionic liquids) to establish an electric double-layer (EDL) interfacial Helmholtz layer at the Cu-based catalyst interface. However, monometallic Cu catalysts exhibit low activity and poor Faradaic efficiency (selectivity) for hydrocarbon products, limiting their commercial application. Herein, we demonstrated targeted delivery of CuAg nanoparticles (NPs) and Ag single atoms (SA) tandem on the Cu(111) facet. The improved chemical properties of bimetallic nanocrystals arise from the synergistic interaction between Cu and Ag metals, enabling this tandem catalyst system (with Ag active sites) to catalyze CO₂ to CO and subsequently convert CO into (C₂₊) hydrocarbon intermediates via C-C coupling on Cu sites. Furthermore, the reduction of CO₂ to CCO on Cu sites and subsequent conversion to acetamide via C-N coupling between CCO and NH₃ on Ag sites are achieved. Our results suggest that C-C couplings between *CO and *CO or *CH and *CH are the most favorable for the formation of ethylene. Specifically, 1-butyl-3-methylimidazolium tetrafluoroborate (B2195) and 1-butyl-3-methylimidazolium hexafluorophosphate (B2320) salts decrease the maximum limiting potential (<i>U</i>_max(η)) to -0.84 and -1.00 V, respectively, positioning them as promising ionic liquids for eCO₂R. To understand EDL effects in eCO₂R at the molecular scale, we employed ab initio molecular dynamics simulation, focusing on hydrogen-bond networks and cation effects through a multiscale approach. Furthermore, this design strategy incorporated regression machine learning (ML) using the extreme gradient boosting regression model and the sure independence screening and sparsifying operator approach to identify key features influencing the target property <i>U</i>_max(η) serving as the ML input data. Results show that the coupling energy (<i>E</i>_cplg) and the average deviation in ground-state band gaps of constituent elements (AvgDev_GSgap) are the most important features for both ethylene and acetamide synthesis, with B2195 and B2320 imidazolium salts efficiently activating CO₂ and driving electroreduction to ethylene with optimized <i>U</i>_max(η).