Combining molecular simulations and machine learning to unlock gas separation performances of MOFs and MOF-based composites

Metal-organic frameworks (MOFs) have become a well-known class of porous materials for solving energy-related gas separation challenges thanks to their high porosities, large surface areas, and easy-to-modify structural properties. Due to the enormous number of synthesized MOFs (>125,000), molecular simulation methods play an important role in assessing the gas separation performances of MOFs and MOF-based composites. In this thesis, we combined high-throughput computational screening (HTCS) and machine learning (ML) approaches to assess the performances of a very large number and type of MOFs and MOF-based composite materials for a variety of gas separation applications. In the first part, we focused on air separation and performed grand canonical Monte Carlo (GCMC) and molecular dynamics (MD) simulations to compute O2 and N2 permeabilities and O2/N2 selectivities of 5629 MOF membranes and 78,806 different types of MOF/polymer mixed matrix membranes (MMMs). Our results showed that many MOF membranes exceed the upper bound established for traditional polymer membranes thanks to the high permeabilities and/or selectivities of MOFs. In the second part, we focused on 11 different gas separation applications to explore 5599 MOF membranes and >180,000 different types of MOF/polymer MMMs. Results showed that many MOFs offer a great opportunity for making MMM applications by improving both the permeability and selectivity of polymers. Since creating all this molecular simulation data needs computationally demanding calculations and analyzing this very large dataset is not practical, in the third part of the thesis, we utilized machine learning (ML) to significantly accelerate the assessment of MOF membranes and MOF/polymer MMMs for six different gas separation applications. Results showed that the ML models that we trained based on the GCMC and MD simulation data accurately predict the adsorption and diffusion properties of He, H2, N2, and CH4 gases in MOFs. In the fourth part, we developed ML models to study 1000 different types of MOFs and ionic liquid (IL)/MOF composites as adsorbents for flue gas separation. The most important features that affect the CO2/N2 selectivity of IL/MOF composites were extracted using the ML results and utilized to computationally generate a new IL/MOF composite, [BMIM][BF4]/UiO-66. Experimentally measured CO2/N2 selectivity of this new composite matched well with the ML-predicted one. In the last part, we focused on water adsorption and performed GCMC and density functional theory (DFT) calculations to explore the effect of framework flexibility on the water adsorption properties of a MOF. Results emphasized the significance of considering the structural flexibility of the MOF for water adsorption. The results of this thesis will provide molecular-level understanding of the gas adsorption and diffusion behavior of MOFs and facilitate the design of new MOFs and MOF-based composites for various gas separation applications at reduced time and cost.