Toward the inverse design of MOF membranes for efficient D2/H2 separation by combination of physics-based and data-driven modeling

Abstract Hydrogen isotopes are useful for scientific research, energy generation and medical treatment. However, their industrial production is expensive because conventional processes for separation of hydrogen isotopologues are mostly based on energy-intensive macroscopic procedures with extremely low separation efficiency. Metal-organic frameworks (MOFs) provide a promising route to D2/H2 separation by leveraging their well-defined chemical and structural features. In this work, we report high-throughput screening of 12,723 experimentally synthesizable MOF membranes for D2/H2 separation by predicting gas adsorption and transport properties underpinning the separation efficiency. A membrane performance score is introduced to identify top ranked MOFs with the best selectivity and capacity. The extensive data generated from the physics-based modeling enables application of machine learning methods to predict desirable features of novel nanoporous materials for more efficient separation of hydrogen isotopes.