First-principles approach to screening multi-component metal alloys for hydrogen purification membranes
Semidey Flecha, Lymarie
MetadataShow full item record
Metal membranes play a vital role in hydrogen purification. Defect-free membranes can exhibit effectively infinite selectivity for hydrogen. Membranes must meet multiple objectives, including providing high fluxes, resistance to poisoning, long operational standards, and be cost effective. Alloys offer an alternate route in improving upon membranes based on pure metal such as Pd. Development of new membranes is hampered by the large effort and time required not only to experimentally develop these membranes but also to properly test these materials. We show how first principle calculations combined with coarse-grained modeling can accurately predict H2 fluxes through binary and ternary alloy membranes as a function of alloy composition, temperature and hydrogen pressures. Our methods require no experimental input apart from the knowledge of the bulk crystal structure. Our approach is demonstrated for pure Pd, Pd-rich binary alloys, PdCu binary alloys, and PdCu-based ternary alloys. PdCu alloys have experimentally shown to have potential for resistance to sulfur poisoning. First, we used plane wave Density Functional Theory to study the binding and local motion of hydrogen for the alloys of interest. This data was used in combination with a Cluster Expansion Method along with the Leave-One-Out analysis to generate comprehensive models to predict hydrogen behavior in the interstitial binding sites within the bulk of the alloys of interest. These models not only were required to correctly fit our calculated data, but they were also required to properly predict behaviors for local conditions for which we had not collected information. These models were then used to predict hydrogen solubility and diffusivity at elevated temperatures. Although we are capable of combining first principle theory calculations with coarse grain modeling, we have explored a pre-screening method in order to determine which a particular material are worth performing additional calculations. Our heuristic lattice model is a simplified model involving as few factors as possible. It is by no means intended to predict the exact macroscopic H properties in the bulk of fcc materials, but it is intended as a guide in determining which materials merit additional characterization.