A Carbon Molecular Sieve (CMS) membrane, in order to test separation of propylene and propane gases, was subjected to increases in feed pressure, varied propylene feed composition, and temperature changes to determine how these factors affect permeance and selectivity. This membrane was prepared on a support made of α-alumina and γ-alumina supports. Feed pressure was increased between 30 and 100psi, and propylene and propane permeance decreased, as did the selectivity. Propylene feed composition increases resulted in an order of magnitude increase in propane permeance, but a decrease in propylene permeance, and a decrease in selectivity. Increases in temperature resulted in increases of propylene and propane permeance, but a decrease in selectivity.

While the solution diffusion model and pore flow model dominate pervaporation transport mechanism modeling, a new model combining the solution diffusion and viscous flow models is validated using membranes with large scale defects exceeding 2 nm in diameter. A range of membranes was characterized using scanning electron microscopy and x-ray diffraction (XRD) to determine quality and phase characteristics. MFI zeolite membranes of He/SF6 pure gas permeation ideal selectivities of 25, 15, and 3 for good, medium, and poor quality membranes were subjected to liquid pervaporations with a 5% ethanol in water feed, by weight. Feed pressure was increased from 1 to 5 atm, to validate existence of viscous flow in the defects. Component molar flux is modeled using the solution diffusion model and the viscous flow model, via J_i=F_i (γ_i x_i P_i^sat )+(ρ )/M_W ∅/μ_ij x_i P_h. A negative coefficient of thermal expansion is observed as permeances drop as a function of temperature in all three membranes, where ϕ=((ϵr_p^2)/τ∆x). Experimental parameter ϕ increased as a function of temperature, and increased with decreasing membrane quality. This further proves that zeolitic pores are shrinking in one direction, and pulling intercrystalline voids larger, increasing the (ϵ/τ) ratio. Permiabilities of the bad, medium, and good quality membrane also decreased over time for both ethanol and water, meaning that fundamental membrane characteristics changed as a function of temperature. To conclude, the model reasonably fits empirical data reasonably well.