(673b) Computational Discovery of Metal-Organic Frameworks with High Water Uptake Capacity for Next-Generation Membranes

Developing next-generation membranes for ion-selective separation is paramount for applications like desalination, battery recycling, and heavy metal recovery. Still, traditional membranes often have high permeability-selectivity tradeoffs, slow water permeation, complex fabrication routes, and defects. Recent studies have demonstrated that metal-organic frameworks (MOFs) can have more than one order of magnitude higher water flux than current state-of-the-art 2D membranes like graphene and retain almost 100% ion rejection rate. Although some qualitative structure-property relationships exist in the water uptake capacity of MOFs, a quantitative understanding of how MOF linker chemistry and pore geometry affect water uptake capacity is still unknown. Moreover, the diversity of MOF design space is in the order of millions, making it impossible to identify MOFs with high water uptake capacity by performing experiments and computer simulations. Here, we take a computational approach to accelerate discovery of design rules for MOFs capable of high water uptake. We choose a subset of experimentally reported water-stable MOFs and implement an automated functionalization procedure. Next, we perform density functional theory (DFT) calculations to determine the binding affinity of water with a diverse set of functional groups present on the linkers from these MOFs The DFT calculations reveal that the functional groups capable of hydrogen bonding with water molecules result in the strongest binding affinity. Next, we calculate the water uptake capacity of functionalized MOFs from grand canonical Monte Carlo (GCMC) simulations. From the GCMC simulations, we discover that electronegative and small functional groups significantly improve the water uptake capacity of a MOF. Furthermore, we find that the water uptake capacity of MOFs correlates much more strongly with MOF pore size than the water-functional group binding affinity, suggesting MOF pore geometry to be the dominant factor in water uptake capacity. Finally, we develop machine learning models to identify water-stable MOFs in experimental and hypothetical MOF databases by performing virtual high-throughput screening and suggesting design guidelines for selecting functional groups that improve water uptake capacity without compromising the water stability of MOFs.