Dual‐Metal Co‐Doping Strategy Enhances the Hydrogen Dissociation and Hydrogen Release Performance of LiBH ₄ Assisted by Ensemble Learning

LiBH 4 has drawn significant attention due to its high theoretical hydrogen storage capacity of 18.5 wt%, yet its practical application is hindered by the harsh dehydrogenation conditions. This paper proposes a strategy integrating ensemble learning with density functional theory (EL–DFT) for rapidly screening modified LiBH 4 systems that exhibit both structural stability and low dehydrogenation temperatures within an acceptable capacity loss range. The results indicate that, in dual‐metal‐doped LiBH 4 structures (TM 1 TM 2 @LiBH 4 ) at different doping sites, the synergistic interaction between the crystal structure and the local electronic environment collectively reduces the dissociation energy ( E d ) of H in the BH 4 − group. Using the substitution energy ( E sub ) and dissociation energy ( E d ) of 60 randomly selected TM 1 TM 2 @LiBH 4 structures as target values, a dual ensemble model AdaBoost.R2 was trained, yielding excellent predictive models: AdaBoost‐SVR( E sub ) and AdaBoost‐GBR( E d ), with R 2 values reaching 0.970 and 0.927, respectively. SHAP analysis reveals that structures with TM 1 anchored at the S1 and S2 sites exhibit lower hydrogen dissociation energy. In particular, the ScSc (S1) @LiBH 4 structure shows E sub = −0.977 eV and E d = 0.549 eV. DFT calculations further demonstrate that dual Sc doping weakens the stability of B–H bonds, significantly lowering the dissociation energy barrier. AIMD simulations confirm that this structure initiates hydrogen release at 500 K. The EL–DFT strategy successfully identifies materials with mild dehydrogenation conditions and stable structures, providing theoretical guidance for further exploration of solid‐state hydrogen storage materials.