Metal-organic frameworks are ultraporous materials that can exhibit selective and cooperative CO2 adsorption chemistry, with potential for future reversible carbon capture applications. While CO2 adsorption enthalpies are relatively well documented, many temperature-dependent and chemical dynamical phenomena related to CO2 and other competing molecular species remain poorly understood. Here, we summarize the development and use of ab initio neural network potentials, constructed with an active machine learning approach from density functional theory calculations, for studies of thermal and dynamical properties of amine-appended metal-organic frameworks. Using these potentials, we predict adsorption energies, mechanical properties, vibrational, thermal conductivity, and adsorption kinetics with and without CO2, reaching ab initio accuracy at a fraction of the computational cost. We further discuss the extent to which this approach can be used to predict adsorption isotherms and can be combined with simulated annealing approaches to identify novel carbon capture framework materials. In all cases, we compare closely with prior and ongoing experiments.
Location:
373 Cory Hall