Machine-learning-accelerated density functional theory screening of Cu-based high-entropy alloys for carbon dioxide reduction to ethylene

Abstract

Computational screening of high-entropy alloy (HEA) catalysts as alternatives to the typical Cu electrocatalyst for CO2 reduction reaction (CO2RR) has been extensively focused on C1 products, but C2+ products have received significantly less attention. This work optimized CuZnPdAgAu HEA catalyst composition for CO2RR to ethylene via density functional theory and supervised machine learning regression techniques. Candidates were identified from 106,045 HEA data for enthalpy of adsorption of *CO2, *H, *HOCCOH, and *C2H4 species, and the Gibbs free energy of *H. The electrocatalytic properties during the reaction were examined on the surface of the optimized HEA candidate – Cu0.36Zn0.18Pd0.10Ag0.18Au0.18 benchmarked to Cu (111). The Pd site of such a candidate functions as the active site for the CO2 activation step. In terms of catalytic activity, it showed lower Gibbs free energy for the potential determining step – the *OCCOH formation step compared to that on the Cu (111). Insight into electronic properties demonstrated that the candidate reduces the uphill reaction energy for *HOCCOH production pathway due to increased electron density in the C–C bond, donated from two Cu sites. It is shown that the HEA catalyst candidate has the potential for CO2RR targeting ethylene, an alternative to a common Cu catalyst.