Rare earth modified carbon-based catalysts for oxygen electrode reactions: A machine learning assisted density functional theory investigation

Abstract

The oxygen electrode reactions (oxygen reduction reaction, ORR and oxygen evolution reaction, OER) are two key reactions in applications such as metal-air batteries, however, slow kinetics have a significant impact on the overall reaction efficiency of the batteries, thus emphasizing the profound significance of catalyst development. In this study, we systematically investigated the catalytic activity of rare-earth-doped graphene (RENxC4-x) as electrocatalysts using a combination of density functional theory (DFT) and machine learning (ML). Furthermore, we successfully screened and identified one ORR catalyst, four OER catalysts, and one bifunctional electrocatalyst from candidate materials. The origins of activity were elucidated in two dimensions using the SHAP (SHapley Additive exPlanation) analysis framework and DFT calculations, revealing that atomic (covalent) radius and ΔG*OH are an important characteristics for describing ORR electrocatalysts, while Pauling electronegativity is crucial for describing OER. Finally, the explicit relationship expression between properties and activity was obtained using the SISSO method, and its generalizability was verified. The interdisciplinary approach of DFT-ML provides insights into the complex origins of activity, offering a new pathway for the discovery and design of high-performance single-atom catalysts (SACs).