Harnessing Pore Size in COF Membranes: A Concentration Gradient-Driven Molecular Dynamics Study on Enhanced H₂/CH₄ Separation

This work presents a novel approach for accurately predicting the gas transport properties of covalent organic framework (COF) membranes using a nonequilibrium molecular dynamics (NEMD) methodology called concentration gradient-driven molecular dynamics (CGD-MD). We first simulated the flux of hydrogen (H₂) and methane (CH₄) across two distinct COF membranes, COF-300 and COF-320, for which experimental data are available in the literature. Our CGD-MD simulation results aligned closely with the experimentally measured gas permeability and selectivity of these COF membranes. Leveraging the same methodology, we discovered promising COF candidates for H₂/CH₄ separation, including NPN-1, NPN-2, NPN-3, TPE-COF-I, COF-303, DMTA-TPB2, 3D-Por-COF, COF-921, COF-IM AA, TfpBDH, and PCOF-2. We then compared our findings with simulations utilizing the well-known approach that merges grand canonical Monte Carlo (GCMC) and equilibrium molecular dynamics (EMD) to predict gas adsorption and diffusion parameters in COFs. Our results showed that when the pore sizes of COF membranes are below 10 Å, the choice of the method plays a significant role in determining the performance of the membranes. The GCMC+EMD approach suggested that COFs tend to exhibit CH₄ selectivity when their pore limiting diameters are below 10 Å, whereas the CGD-MD results reveal a preference for H₂. Density functional theory calculations indicate that H₂ has a lower affinity for three promising COFs, NPN-1, NPN-2, and NPN-3, compared to CH₄, which results in H₂ remaining unbound, while CH₄ occupies all of the adsorption sites, thereby facilitating the selective recovery of H₂ at the end of the separation process. We proposed a relationship between adsorption time and diffusion time, highlighting the critical role of selecting an appropriate simulation method. This relationship underscores how adsorption and diffusion processes interplay, impacting material performance. Overall, these insights not only improve the accuracy of predictive models but also guide the development of more efficient COF-based membrane applications for future research and industrial applications.