Data-Driven Design of MX/ O enes@TiFeH 2 Heterostructures for Regulating Hydrogen Release Performance

Abstract

To address the issues of high activation energy barrier and sluggish hydrogen release kinetics in the hydrogen storage process of TiFe alloys, this study proposes an ML-DFT screening strategy combining machine learning with first-principles calculations to explore the regulatory mechanism of MX/Oenes two-dimensional materials on the hydrogen release performance of TiFeH2. First, 71 MX/Oenes@TiFeH2 heterostructures were constructed, and 12 representative configurations were selected through stability analysis. On this basis, a physical feature set PPF containing 10 intrinsic physical parameters was established. Combined with the AdaBoost-XGB model, high-accuracy prediction of the hydrogen dissociation energy was achieved. SHAP analysis revealed that the ionization energy of the M1 layer, the boiling point of the X/O layer, and the electronegativity of the M2 layer are key descriptors affecting hydrogen release performance. The finally screened WYSe@TiFeH2 weakens the Ti-H bond through a long-range electron-withdrawing effect, reducing the transition-state barrier from 1.04 eV to 0.85 eV. This work provides a new theoretical basis for the rapid screening and rational design of high-performance solid-state hydrogen storage materials.