Predicting Covalent Organic Framework (COF) Membrane Performance by Mapping Molecular Interactions in Mixed Solvents via Atomistic Modeling

Complex solvent environments continue to limit the widespread adoption of organic solvent nanofiltration (OSN) in many chemical industry applications. Reactive force field (ReaxFF) and nonreactive force field models have been recently developed to molecularly map separation performance of a commercial covalent organic framework (COF), TpPa-1, and a carboxylated COF (C-COF). Specifically, the following factors have been characterized using these atomistic models: layer stacking, effective vs. designed pore size, and solvated solute size in various single organic solvents or solvent pairs. Model predications can be directly compared with experimental filtration results after normalizing model outcomes and filtration data with a common solvent, such as water, to minimize time and length scale mismatch between atomistic modeling and experiments. Model outputs, such as organic solvent permeance and solute rejection rate, matched experimental filtration results well. These findings demonstrate how solvated solute state and effective pore size in mixed solvents cumulatively dictate membrane performance. Additionally, ion effects on selectivity were probed theoretically and experimentally by adding NaOH and HCl to organic solvents, such as DMF and methanol. In sum, force field models can serve as digital twins of COF membranes to simulate separation processes while capturing the effects of COF structure, chemistry, and crystallinity on membrane performance in complex organic solvent environments. This approach will provide insight into future COF design and synthesis for persisting separation challenges.