Unraveling the effect of single atom catalysts on the charging behavior of nonaqueous Mg–CO2 batteries: a combined density functional theory and machine learning approach

Abstract

This study integrates density functional theory (DFT) and machine learning (ML) methodologies to investigate the charging performance and catalyst design principles of porphyrin-supported single atom catalysts (SACs) based on 3d and 4d transition metals (TMs) in the context of nonaqueous Mg–CO₂ batteries. Specifically, we utilize DFT calculations to elucidate the adsorption energies of the primary discharge product, MgCO₃, on SACs supported on N_xS_y (where x = 4, 3, 2 and y = 0, 1, 2, respectively) moieties of porphyrin. Our analysis unveils the ability of these SACs to effectively bind with MgCO₃, which correlates with enhancing the kinetics of its decomposition, a pivotal factor influencing the charging performance. The results demonstrate that the improved adsorption energies of early TMs are expected to reduce the decomposition barrier for MgCO₃ during battery charging. Furthermore, we leverage a DFT-derived dataset to construct ML models using Gradient Boosting Regression (GBR) and Artificial Neural Network (ANN) algorithms. Employing K-fold cross-validation, both algorithms consistently exhibit remarkable accuracy in their predictions. To unravel the catalyst design principles, we also conduct feature importance analysis, using SHapley Additive exPlanations (SHAP), Permutation Importance, and Mean Decrease Impurity (MDI) techniques to identify the most significant features. This study reveals that the ionization potential of TMs is the most important descriptor for the selection of SACs for cathodes in Mg–CO₂ batteries. Overall, this combined DFT and ML investigation provides insights into both the charging performance of SACs in Mg–CO₂ batteries and the fundamental principles governing catalyst design.