Machine Learning-Assisted Discovery of Unconventional Molybdenum Nitride Nano-Facets as Hydrogen Evolution Electrocatalysts

Transition-metal nitrides (TMNs) are promising electrocatalysts owing to their unique electronic structures and surface properties. Yet, the formation of metastable phases during particle growth is unavoidable in the traditional synthesis of TMNs. These metastable phases cannot be holistically depicted by conventional phase diagrams. Here, we use density functional theory (DFT) calculations in conjunction with thermodynamic analysis to establish particle-size-dependent phase diagrams for Mo nitrides (Mo_xN_y), including both metastable and thermodynamically stable phases. To accelerate the identification of energetically favourable high-Miller-index facets of Mo nitrides, we complement our DFT approach with a machine learning (ML) model trained on computed results for low-Miller-index facets. Furthermore, we compute hydrogen evolution reaction (HER) free energy diagrams using DFT for both low- and high-Miller-index facets, uncovering that several previously underexplored high-Miller-index facets of Mo nitrides are not only stable but also catalytically active for HER. Our study elucidates the correlation between nanoparticle dimension and phase stability, predicts unconventional facets of Mo nitrides that serve as good catalysts in HER, and ultimately presents a computationally guided strategy to efficiently navigate the material discovery landscape and further discover robust and efficient catalysts for green energy conversion schemes.