Diffusion, mechanical and thermal properties of sT hydrogen hydrate by machine learning potential

Newly-synthesized structure T (sT) hydrate show promising practical applications in hydrogen storage and transport, yet the properties remain poorly understood. Here, we develop a machine learning potential (MLP) of sT hydrogen hydrate derived from quantum-mechanical molecular dynamics (MD) simulations. Using this MLP forcefield, the structural, hydrogen diffusion, mechanical and thermal properties of sT hydrogen hydrate are extensively explored. It is revealed that the translational and rotational mobilities of hydrogen molecule in sT hydrate are limited due to unique shape and finite cavities, and tiny windows of neighboring cavities. sT hydrogen hydrate exhibits unique uniaxial tension stress-strain response, with first nonlinear increase to GPa-level but followed by deep drop in the stretching stress, indicating brittle failure, similar to that by DFT and empirical forcefields. Moreover, temperature-dependent thermal conductivity in sT hydrogen hydrate is mainly contributed by hydrogen-bonded network formed by host water molecules, while hydrogen guest molecules play an insignificant role in the thermal transport.