Machine Learning-assisted Study of Low-, Medium-, and High-Entropy Hydrogen Storage Alloys Validated by the Experimental Data

Advancing of hydrogen and metal hydride energy technologies requires purposeful development of efficient hydrogen storage materials, particularly, tuning their composition towards optimization of hydrogen sorption properties suitable for the end-use applications. This study employed linear regression modelling to analyze hydrogen storage properties of low-, medium- and high-entropy alloys with BCC, C14- and C15-AB2 and AB5 structures found in the literature (>350 entries in total) and to make predictions based on the model further validated by additional reference data and results of own experiments. It was found that the applied model gives a good qualitative correspondence with the reference data on hydrogen sorption capacity and thermodynamics of hydrogen interaction with the alloys but has a limiting predicting capacity allowing only rough quantitative estimations. It was also concluded that the unit cell volume, valence electron concentration, and, to a lesser extent, electronegativity mismatch, exhibit strong effects on the hydrogen sorption properties of the studied alloys while the influence of other factors including the mixing entropy is much less pronounced.