Leveraging molecular simulations and machine learning to assess CO2, O2, and N2 adsorption and separation performances of diverse MOF databases

• Combined molecular simulations with ML to predict CO 2 , O 2 , and N 2 adsorption in MOFs. • Predicted gas uptakes of ∼130,000 MOFs in minutes using simple features. • Identified structural and chemical properties for high CO 2 /N 2 and O 2 /N 2 selectivity. We integrated molecular simulations and machine learning (ML) to comprehensively explore the gas adsorption and separation performances of both synthesized and hypothetical metal-organic frameworks (MOFs) available in five different MOF databases. Following the generation of CO 2 , O 2 , and N 2 adsorption data for synthesized MOFs at varying pressures through grand canonical Monte Carlo (GCMC) simulations, we developed ML models that can swiftly and accurately predict the gas adsorption properties of any MOF based on its structural, chemical, and energetic characteristics. These ML models were then transferred to four distinct hypothetical MOF databases consisting of nearly 130,000 structures to assess their CO 2 , O 2 , and N 2 adsorption properties in addition to CO 2 /N 2 and O 2 /N 2 separation performances as a very efficient alternative to computationally time and resource demanding molecular simulations. We identified the top-performing materials from each database to uncover their structural, chemical, and topological properties leading to high selectivities and concluded that synthesized MOFs with narrow pores, lanthanide metals, and linkers featuring oxalate, pyridine dicarboxylate, and fumarate offer the highest CO 2 /N 2 selectivities. Our work presents the most extensive dataset produced for CO 2 , O 2 , and N 2 gas adsorption in MOFs, composed of ∼3.9 million data points for materials’ structural, chemical, and energetic features, gas adsorption properties, and selectivities computed at different pressures to accelerate the materials design and discovery for CO 2 , O 2 , and N 2 adsorption and separation.