Metal-organic frameworks (MOFs) are highly tuneable, extended-network, crystalline, nanoporous materials with applications in gas storage, separations, and sensing. We review how molecular models and simulations of gas adsorption in MOFs have informed the discovery of performant MOFs for methane, hydrogen, and oxygen storage, xenon, carbon dioxide, and chemical warfare agent capture, and xylene enrichment. Particularly, we highlight how large, open databases of MOF crystal structures, post-processed to enable molecular simulations, are a platform for computational materials discovery. We discuss how to orient research efforts to routinise the computational discovery of MOFs for adsorption-based engineering applications.

Metal-organic frameworks (MOFs) are highly tunable, extended-network, crystalline, nanoporous materials with applications in gas storage, separations, and sensing. We review how molecular models and simulations of gas adsorption in MOFs have informed the discovery of performant MOFs for methane, hydrogen, and oxygen storage, xenon, carbon dioxide, and chemical warfare agent capture, and xylene enrichment. Particularly, we highlight how large, open databases of MOF crystal structures, post-processed to enable molecular simulations, are a platform for computational materials discovery. We discuss how to orient research efforts to routinize the computational discovery of MOFs for adsorption-based engineering applications.

Designing better porous materials for gas storage or separations applications frequently leverages known structure-property relationships. Reliable structure-property relationships, however, only reveal themselves when adsorption data on many porous materials are aggregated and compared. Gathering enough data experimentally is prohibitively time consuming, and even approaches based on large-scale computer simulations face challenges. Brute force computational screening approaches that do not efficiently sample the space of porous materials may be ineffective when the number of possible materials is too large. Here we describe a general and efficient computational method for mapping structure-property spaces of porous materials that can be useful for adsorption related applications. We describe an algorithm that generates random porous "pseudomaterials", for which we calculate structural characteristics (e.g., surface area, pore size and void fraction) and also gas adsorption properties via molecular simulations. Here we chose to focus on void fraction and Xe adsorption at 1 bar, 5 bar, and 10 bar. The algorithm then identifies pseudomaterials with rare combinations of void fraction and Xe adsorption and mutates them to generate new pseudomaterials, thereby selectively adding data only to those parts of the structure-property map that are the least explored. Use of this method can help guide the design of new porous materials for gas storage and separations applications in the future.

Decades of research have yet to yield porous adsorbents that meet the U.S. Department of Energy's methane storage targets. To better understand why, we calculated high-pressure methane adsorption in 600 000 randomly generated porous crystals, or "pseudomaterials," using atomistic grand canonical Monte Carlo (GCMC) simulations. These pseudomaterials were periodic configurations of Lennard-Jones spheres whose coordinates in space, along with corresponding well depths and radii, were all chosen at random. GCMC simulations were performed for pressures of 35 and 65 bar at a temperature of 298 K. Methane adsorption was compared for all materials against a range of other properties: average well depths and radii, number density, helium void fraction, and volumetric surface area. The results reveal structure-property relationships that resemble those previously observed for metal-organic frameworks and other porous materials. We contend that our computational methodology can be useful for discovering useful structure-property relationships related to gas adsorption without requiring experimentally accessible structural data.