As lower quality gas feeds increase in economic viability, energy-efficient membrane processes for removing associated sour gases (e.g., CO2 and H2S) are becoming increasingly important. For this application, intrinsically microporous CANAL polymers have attractive property sets, including high selectivities and resistance to plasticization. We build upon previous experimental work by developing a computational system for an archetypal CANAL polymer, CANAL-Me-Me2F, to help address experimental challenges including extensive testing time and toxic gas hazards. Pore size distribution (PSD) analysis of the simulated system reveals micropores ranging from approximately 1.5–11 Å, showing better agreement with experimentally observed size-sieving separations compared to PSD results from non-local density functional theory. High-pressure sorption isotherms of H2, N2, O2, CH4, CO2, H2S, C3H8, and C3H6 were computed via grand canonical Monte Carlo and iterated Monte Carlo-molecular dynamics simulations and showed strong alignment with experimentally gathered results at 35 °C. Isotherms of H2S, C3H8, and C3H6 were further computed at 55 °C, 120 °C, and 190 °C. Dual mode sorption parameters were fit using the constrained optimization approach of Wu et al. (2021) using a Python package under development in our lab to improve the reproducibility of DMS fitting. Computation of sorption energetics from DMS parameters revealed greater Langmuir sorption affinity for hydrocarbons than for H2S, potentially due to CANAL-Me-Me2F lacking heteroatoms. Heats of Henry sorption were less exothermic for all three gases of interest in CANAL-Me-Me2F compared to PIM-1, potentially indicating less favorable intra-chain “gap” openings in CANAL-Me-Me2F to accommodate diluents. The mixed-gas selectivity of H2S/(C3H8+C3H6) computed via the mixed-gas DMS model showed minimal selectivity boost despite the presence of competitive sorption, which we attribute to the stronger pressure dependence on the hydrocarbon sorption coefficients compared to H2S. Overall, this study demonstrates the benefit of combining computational and experimental insights to probe the unique nanoscale behavior of CANAL polymer membranes while also highlighting how computation can be used to generate data that is difficult to obtain experimentally.
