High-Throughput Computation and Machine Learning Prediction Accelerating the Design of Cathode Catalysts for Li–CO2 Batteries

Developing new cathode catalysts to facilitate the electrochemical reaction toward the product of lithium oxalate would enhance the performance of Li–CO2 batteries. Herein, high-throughput calculations and machine learning technology were synergistically utilized to effectively screen new transition metal-based cathode catalysts. Gibbs free energies of CO2 reduction on the surfaces of 11 transition metal compounds were explored by the high-throughput first-principles calculations, and MoN was forecasted to be a promising cathode catalyst material, promoting the reaction product of lithium oxalate and possessing the minimum overpotential. On the other hand, based on the data of first-principles calculations, machine learning models with respect to overpotential were built to evaluate the electrochemical performances of transition metal nitrides. The origin of catalytic activity was revealed to be the adsorption energy by the feature importance analysis of overpotential. Our work not only puts forward a paradigm of the highly effective screening of cathode catalysts for Li–CO2 batteries but also provides a new understanding of the overpotential of CO2 reduction from the perspective of molecule adsorption.