(178g) Combining Molecular Simulation and Machine Learning to Find Optimal Conditions for Hydrogen Storage and Hydrogen/Nitrogen/Ammonia Separation

Nanoporous materials have found numerous applications, including gas storage, adsorptive separations, and membrane separations. However, the dimensionality of the materials space with regard to pore size distribution, topology, and chemical functionalization, and the state space (temperature, pressure, and composition of adsorbate mixture) poses a challenge for finding optimal materials and conditions. Here we incorporate high-throughput molecular simulations and machine learning into a platform to find optimal conditions for hydrogen storage and hydrogen/nitrogen/ammonia separation. Efficiently identifying the optimal hydrogen storage temperature for a specific nanoporous material and pressure swing requires modeling hydrogen loading as a continuous function of pressure and temperature. Using data obtained from high-throughput Monte Carlo simulations for zeolites, metal–organic frameworks, and hyper-cross-linked polymers, we develop a meta-learning model which jointly predicts the adsorption loading for multiple materials over wide ranges of pressure and temperature. Meta-learning gives higher accuracy and improved generalization compared to fitting a model separately to each material. Here, we apply the meta-learning model to identify the optimal hydrogen storage temperature with the highest working capacity for a given pressure difference. Materials with high optimal temperatures are found closer in the fingerprint space and exhibit high isosteric heats of adsorption. In a second application, we use high-throughput Monte Carlo and molecular dynamics simulations to generate data for adsorption and transport selectivities for ammonia/nitrogen/hydrogen mixtures covering a wide range of temperatures and pressures. Again, a machine learning model is trained on these simulation data and used to find the condition space that yields selectivities above a specific threshold.