"Heusler Alloys as Catalysts for Hydrogen Production by Ammonia Decomposition: Data-Driven Screening Via Graph Neural Networks"

Hydrogen as an alternate fuel in the transport and energy sector holds significant potential in mitigating the deleterious impact of anthropogenic greenhouse gas emissions. However, its low volumetric energy density, storage and transportation difficulties have been a major detriment in the advancement of a hydrogen economy. Ammonia has recently emerged as a promising hydrogen vector that can be transported easily and chemically transformed to hydrogen at the point of use. A major bottleneck in the realization of this technology is the availability of active and economical catalysts for the said conversion. Recently, Heusler alloys (HAs) (formula: X2YZ) which are known for their magnetic properties, have emerged as a class of promising catalysts for diverse chemical conversions. In this work, state of the art GEMNET-OC, a graph based deep learning model, was used to screen potentially active HA catalysts. Using adsorption energies computed by GEMNET-OC and a quasi-equilibrium kinetic model, 280 HAs were evaluated. 15 HAs showed better performance than state-of-the-art Ru catalyst. Based on the Herfindahl–Hirschman index (HHI) and scarcity of screened alloys, three alloys were further shortlisted. Finally, DFT validation was used to recommend two of these alloys for synthesis and validation in experiments.