Assessing CO2 separation performances of IL/ZIF-8 composites using molecular features of ILs

Given the vast number and diversity of metal-organic frameworks (MOFs) and ionic liquids (ILs), it is impractical to experimentally test the gas adsorption and separation potential of each one of the possible IL/MOF composites formed by the different combinations of these two components. In this study, we developed a comprehensive computational approach integrating Conductor-like Screening Model for Realistic Solvents (COSMO-RS) calculations, density functional theory (DFT) calculations, Grand Canonical Monte Carlo (GCMC) simulations, and machine learning (ML) algorithms to evaluate a wide variety of IL-incorporated ZIF-8 composites for CO₂ separations. We examined 1322 different types of IL/ZIF-8 composites, covering the largest variety of ILs studied to date (8 cations and 35 anions) at various loadings, for flue gas separation and natural gas purification. We simulated CO₂, CH₄, and N₂ adsorption properties of these composites and used this high-quality molecular simulation data to develop ML models that can predict gas uptakes of any IL/ZIF-8 composite when chemical and structural features of the IL are given. The accurate prediction power of these ML models was shown by comparing their estimates with the experimental and simulation data. Our approach significantly accelerates the assessment of a very large number of IL/ZIF-8 composites and reveals the key molecular features of ILs to make composites for achieving superior gas separation performance.