Machine Learning-driven Molecular Dynamics Unveils a Bulk Phase Transformation Driving Ammonia Synthesis on Barium Hydride

The modern view of industrial heterogeneous catalysis is evolving from the traditional static paradigm where the catalyst merely provides active sites, to that of a functional material in which dynamics plays a crucial role. Using machine learning-driven molecular dynamics simulations, we confirm this picture for the ammonia synthesis catalysed by BaH. Recent experiments show that this system acts as a highly efficient catalyst, but only when exposed first to N and then to H in a chemical looping process. Our simulations reveal that when first exposed to N, BaH undergoes a profound change, transforming into a superionic mixed compound, BaH(NH), characterized by a high mobility of both hydrides and imides. This transformation is not limited to the surface but involves the entire catalyst. When this compound is exposed to H in the second step of the looping process, ammonia is readily formed and released, a process greatly facilitated by the high ionic mobility. Once all the nitrogen hydrides are hydrogenated, the system reverts to its initial state, ready for the next looping cycle. Our microscopic analysis underlines the dynamic nature of this heterogeneous catalyst, which does not merely serve as static platform for reactions, rather it is a dynamic entity that evolves under reaction conditions.