Machine Learning Potential for Modelling H$_2$ Adsorption/Diffusion in MOF with Open Metal Sites

Abstract

Metal-organic frameworks (MOFs) incorporating open metal sites (OMS) have been identified as promising sorbents for many societally relevant-adsorption applications including CO$_2$ capture, natural gas purification and H$_2$ storage. It is critical to derive generic interatomic potential to achieve accurate and effective evaluation of MOFs for H$_2$ adsorption. On this path, as a proof-of-concept, the Al-soc-MOF containing Al-OMS, previously envisaged as a potential candidate for H$_2$ adsorption, was selected and a machine learning potential (MLP) was derived from a dataset initially generated by ab-initio molecular dynamics (AIMD) simulations. This MLP was further implemented in MD simulations to explore the binding modes of H$_2$ as well as its temperature dependence distribution in the MOFs pores from 10K to 90K. MLP-Grand Canonical Monte Carlo (GCMC) simulations were further performed to predict the H$_2$ sorption isotherm of Al-soc-MOF at 77K that was further confirmed by gravimetric sorption measurements. As a further step, MLP-based MD simulations were conducted to anticipate the kinetics of H$_2$ in this MOF. This work delivers the first MLP able to describe accurately the interactions between the challenging H$_2$ guest molecule and MOFs containing OMS. This innovative strategy applied to one of the most complex molecules owing to its highly polarizable nature alongside its quantum-mechanical effects that are only accurately described by quantum calculations, paves the way towards a more systematic accurate and efficient in silico assessment of the MOFs containing OMS for H$_2$ adsorption and beyond to the low-pressure capture/sensing of diverse molecules.