Speaker
Description
MXenes show broad potential as metal-ion battery electrodes due to their tunable surface chemistry and open layered structures. However, traditional density functional theory (DFT) is computationally expensive for systematically screening metal-atom adsorption across diverse MXenes. To overcome this, we propose an efficient machine-learning (ML) framework to predict adsorption energies. Using DFT data, we constructed a cross-element dataset of 90 MXene substrates and 5 metal atoms. We introduce a novel feature-engineering approach that divides MXene surface atoms into three spatial layers, extracting key descriptors like electron affinity and atomic radii. Among four tested ML models, eXtreme Gradient Boosting (XGB) achieved the best predictive performance (MAE = 0.21 eV, R2 = 0.93) with significantly higher efficiency. SHAP analysis identified the electron affinities of the first-layer atoms and the adsorbed metal as the primary factors dictating adsorption energy. Furthermore, we analyzed prototypical high-adsorption MXenes (Sc2NO2 and Y2NO2). Their d-band centers (2.0325 eV and -1.8770 eV) lie closer to the Fermi level, fundamentally explaining their strong adsorption. This study provides a robust, high-throughput screening methodology and critical insights into cross-element adsorption mechanisms for advanced two-dimensional battery materials.