Quantitative computational screening of Pd-based intermetallic membranes for hydrogen separation

Powered by DFT: Screening methods that accelerate materials development for hydrogen in metals applications

CONSPECTUS: Not only is hydrogen critical for current chemical and refining processes, it is also projected to be an important energy carrier for future green energy systems such as fuel cell vehicles. Scientists have examined light metal hydrides for this purpose, which need to have both good thermodynamic properties and fast charging/discharging kinetics. The properties of hydrogen in metals are also important in the development of membranes for hydrogen purification. In this Account, we highlight our recent work aimed at the large scale screening of metal-based systems with either favorable hydrogen capacities and thermodynamics for hydrogen storage in metal hydrides for use in onboard fuel cell vehicles or promising hydrogen permeabilities relative to pure Pd for hydrogen separation from high temperature mixed gas streams using dense metal membranes. Previously, chemists have found that the metal hydrides need to hit a stability sweet spot: if the compound is too stable, it will not release enough hydrogen under low temperatures; if the compound is too unstable, the reaction may not be reversible under practical conditions. Fortunately, we can use DFT-based methods to assess this stability via prediction of thermodynamic properties, equilibrium reaction pathways, and phase diagrams for candidate metal hydride systems with reasonable accuracy using only proposed crystal structures and compositions as inputs. We have efficiently screened millions of mixtures of pure metals, metal hydrides, and alloys to identify promising reaction schemes via the grand canonical linear programming method. Pure Pd and Pd-based membranes have ideal hydrogen selectivities over other gases but suffer shortcomings such as sensitivity to sulfur poisoning and hydrogen embrittlement. Using a combination of detailed DFT, Monte Carlo techniques, and simplified models, we are able to accurately predict hydrogen permeabilities of metal membranes and screen large libraries of candidate alloys, selections of which are described in this Account. To further increase the number of membrane materials that can be studied with DFT, computational costs need to be reduced either through methods development to break bottlenecks in the performance prediction algorithm, particularly related to transition state identification, or through screening techniques that take advantage of correlations to bypass constraints.