Simulation and Machine Learning-Integrated Investigation of Double-Linker MOFs for the Separation of CF₄ from CH₄ and N₂

One of the most crucial problems of the modern world is to combat the emission of greenhouse gases as these lead to extreme weather conditions, droughts, etc., and endanger the global ecosystem. Among the various greenhouse gases, the removal of carbon tetrafluoride, CF4, from the atmosphere is of particular interest as its global warming potential is several orders of magnitude higher than that of many other greenhouse gases, including carbon dioxide, CO2. The separation of CF4 from methane, CH4, and nitrogen, N-2, which are present in the atmosphere alongside CF4, can be efficiently accomplished by adsorption- or membrane-based processes. Metal-organic frameworks (MOFs) offer a variety of possibilities for gas separation due to their large surface areas and tailored chemical and physical structures. In this work, the separation of CF4/CH4, CF4/N-2, and CF4/CH4/N-2 in a subset of MOFs, double-linker MOFs (DL-MOFs), is investigated by using molecular simulations and machine learning (ML) methods. Among similar to 500 MOFs, the best MOFs for adsorption-based CF4/CH4, CF4/N-2, and CF4/CH4/N-2 separation are determined by considering the adsorption selectivity, working capacity, and regenerability. The ML models developed for adsorption-based CF4/CH4, CF4/N-2, and CF4/CH4/N-2 separation have revealed that only one or two features (Henry's constant and/or surface area) are sufficient to establish models that can make accurate predictions of gas uptake. The membrane-based CF4/CH4 and CF4/N-2 separation performances of DL-MOFs have also been studied, and it was found that inverse selective MOFs preferring CH4 or N-2 over CF4 can exist. Overall, this study will open avenues for the design and development of novel MOF adsorbents and membranes for the separation of CF4 from CH4 and N-2 by providing atomistic insights into the adsorption and diffusion properties of DL-MOFs.