Combined High-Throughput DFT and ML Screening of Transition Metal Nitrides for Electrochemical CO2 Reduction

Abstract

The electrochemical reduction of CO2 (CO2RR) using renewable electricity has the potential to reduce atmospheric CO2 levels while producing valuable chemicals and fuels. However, the practical implementation of this technology is limited by the activity, selectivity, and stability of catalyst materials. In this study, we employ high-throughput density functional theory (DFT) calculations to screen ∼800 transition metal nitrides and identify potential catalysts for CO2RR. The stability and activity of the screened materials were thoroughly evaluated via thermodynamic analysis, revealing Co, Cr, and Ti transition metal nitrides as the most promising candidates. Additionally, we conduct a feature importance analysis using machine learning (ML) regression models for binding energy prediction and determine the primary factors influencing the stability of catalysts. We show that the group number of metals has a significant impact on the binding energy of *OH and thus on the stability of the catalysts. We anticipate that this combined approach of high-throughput DFT screening and design strategy derived from ML regression analysis could effectively lead to the discovery of improved energy materials.