Laves type intermetallic compounds as hydrogen storage materials: A review

Towards Pareto optimal high entropy hydrides via data-driven materials discovery

“…For example, Δ H 's dependence on Δ H̄ b and are qualitatively similar to Δ S 's, indicative of the enthalpy/entropy compensation effect that generally constrains materials design. 25,58 The similar feature dependencies of the H/M and Δ H models ( e.g. , SHAP values for mean electronegativity, , and mean binary hydride formation enthalpies, ΔH̄ b ) indicate the limited ability to independently tune hydride stability and capacity, a critical optimization problem that must be overcome, as discussed next.…”

Large Destabilization of (TiVNb)-Based Hydrides via (Al, Mo) Addition: Insights from Experiments and Data-Driven Models

“…Trained on experimental metal hydride thermodynamic data and accompanied by experimental validation, data-driven models can qualitatively and quantitatively predict the effects of alloy composition tuning on hydride reaction enthalpies. ,− Because the accuracy of such methods is typically improved by including more training data, we created the ML-ready HydPARK Zenodo repository, which we continually update as more experimental data become available. For this study, we released v0.0.5 of the ML-ready HydPARK database, which augments v0.0.4 with the comprehensive AB 2 metal hydride thermodynamic data recently presented in ref . Gradient boosting tree models were trained to predict the enthalpy of the hydriding reaction, Δ H , using the identical strategy described in ref , with scripts to reproduce them provided here ().…”

Towards Pareto optimal high entropy hydrides via data-driven materials discovery

“…For example, note how ∆H's dependence on ∆ Hb and SG# are inversely correlated with ∆S's, indicative of the enthalpy/entropy compensation effect that generally constrains materials design. 23,55 Figure 1c breaks down the K-fold test set errors against the underlying data distribution. Higher accuracy is directly correlated with the density of training data for a given target property; thus, the highly non-uniform distribution of all measured properties in ML-HydPARK presents a challenge but also a significant opportunity for model improvement with the collection of additional data.…”