Micro-kinetic modelling of the CO reduction reaction on single atom catalysts accelerated by machine learning.

Electrochemical CO transformation to fuels and chemicals is an effective strategy for conversion of renewable electric energy into storable chemical energy in combination with reducing green-house gas emission. Metal-nitrogen-carbon (M-N-C) single atom catalysts (SAC) have shown great potential in the electrochemical CO reduction reaction (CORR). However, exploring advanced SACs with simultaneously high catalytic activity and high product selectivity remains a great challenge. In this study, density functional theory (DFT) calculations are combined with machine learning (ML) for rapid and high-throughput screening of high performance CO reduction catalysts. Firstly, the electrochemical properties of 99 M-N-C SACs were calculated by DFT and used as a database. By using different machine learning models with simple features, the investigated SACs were expanded from 99 to 297. Through several effective indicators of catalyst stability, inhibition of the hydrogen evolution reaction, and CO adsorption strength, 33 SACs were finally selected. The catalytic activity and selectivity of the remaining 33 SACs were explored by micro-kinetic simulation based on Marcus theory. Among all the studied SACs, Mn-NC, Pt-NC, and Au-NC deliver the best catalytic performance and can be used as potential catalysts for CO/CO conversion to hydrocarbons with high energy density. This effective screening method using a machine learning algorithm can promote the exploration of CORR catalysts and significantly reduce the simulation cost.