We present an approach for a probabilistic and unbiased discovery of selective and active catalysts for the carbon dioxide (CO2) and carbon monoxide (CO) reduction reactions on high-entropy alloys (HEAs). By combining density functional theory (DFT) with supervised machine learning, we predict the CO and hydrogen (H) adsorption energies of all surface sites on the (1 1 1) surfaces of the disordered CoCuGaNiZn and AgAuCuPdPt HEAs.
This allows an optimization for the HEA compositions with increased likelihood for sites with weak hydrogen adsorption to suppress the formation of molecular hydrogen and with strong CO adsorption to favor the reduction of CO. As opposed to the construction of specific arrangements of surface atoms, our approach makes the desired surface sites more frequent through an increase in their probability. This leads to the unbiased discovery of several catalyst candidates for which selectivity toward highly reduced carbon compounds is expected and of which some have been verified in the literature.
Jack K. Pedersen, Thomas A. A. Batchelor, Alexander Bagger and Jan Rossmeisl
ACS Catal. 2020, 10, 3