Speaker
Description
High-entropy alloys (HEAs) have attracted considerable interest as potential hydrogen storage
materials due to the diversity of their local chemical environments and interstitial sites. In par-
ticular, the equiatomic TiVZrNbHf alloy, with a body-centered cubic (BCC) structure, exhibits a
high hydrogen absorption capacity, associated with lattice distortions and the presence of different
interstitial sites [1, 2]. This work investigates interstitial hydrogen stability in TiVZrNbHf using
a methodology that combines density functional theory (DFT) calculations and MACE machine
learning interatomic potentials. Five special quasirandom structures (SQS) are considered to
represent the chemical disorder of the alloy. Within these structures, interstitial hydrogen config-
urations, including octahedral, tetrahedral, and randomly selected sites, are evaluated to explore
the influence of local atomic environments on their stability. DFT results are used to fine-tune the
pre-trained MACE-MP-0 model. Its predictive capability is assessed through an intra-configuration
generalization strategy, in which the training, validation, and test sets contain distinct interstitial
sites belonging to the same five SQS configurations. The energies and atomic forces predicted by
MACE are compared with DFT reference values to determine its ability to reproduce previously
unseen interstitial environments within known chemical configurations and to identify potentially
favorable sites for hydrogen trapping.