A method is presented to prepare high-density, vertically aligned carbon nanotube (VA-CNT) membranes. The CNT arrays were prepared by chemical vapor deposition (CVD), and the arrays were collapsed into dense membranes by capillary-forces due to solvent evaporation. The average space between the CNTs after shrinkage was approximately 3 nm, which is comparable to the pore size of the CNTs. Thus, the interstitial pores between CNTs were not sealed, and gas permeated through both CNTs and interstitial pores. Nanofiltration of gold nanoparticles and N(2) adsorption indicated the pore diameters were approximately 3 nm. Gas permeances, based on total membrane area, were 1-4 orders of magnitude higher than VA-CNT membranes in the literature, and gas permeabilities were 4-7 orders of magnitude higher than literature values. Gas permeances were approximately 450 times those predicted for Knudsen diffusion, and ideal selectivities were similar to or higher than Knudsen selectivities. These membranes separated a larger molecule (triisopropyl orthoformate (TIPO)) from a smaller molecule (n-hexane) during pervaporation, possibly due to the preferential adsorption, which indicates separation potential for liquid mixtures.

Abstract

Binary and ternary mixtures of organic vapors were separated at elevated temperatures with a silicalite zeolite membrane on a porous, tubular, γ-alumina support. Linear alkanes (C5−C9), branched alkanes, aromatics, and saturated ring compounds were used as feeds, and permeances of pure compounds and mixtures were measured between ∼360 and 510 K. Pure compound permeances of the linear alkanes strongly decrease with increasing chain length, whereas the branched and cyclic compounds permeate at rates similar to those of n-hexane and n-heptane. Almost all permeances increase with increasing temperature. Mixtures of branched or cyclic molecules with small linear alkanes were readily separated with high selectivities (over 200 for n-hexane/benzene), even though the ratios of pure component permeances were small. The separation behavior is not due to molecular sieving but instead appears to be due to preferential adsorption (adsorption on external surface, pore entering, adsorption in pores) of one species, which prevents the other organics from adsorbing and transporting through the membrane. Mixtures of cyclic or branched molecules showed small or no separations. For all systems, separations factors decrease as temperature increases apparently because preferential adsorption becomes less important at elevated temperatures. For mixtures of benzene or methylyclohexane with 2,2,4-trimethylpentane and for mixtures of 2,2-dimethylbutane with 3-methylpentane, both compounds permeated at similar rates and no separations were obtained. Single-file transport in the zeolite channels is suspected to limit transport. The membranes have intercrystalline regions in parallel with the zeolite pores that may also permeate the organics.

The permeation behavior of n-octane, isooctane, and n-hexane vapors through continuous, silicalite zeolite membranes on porous alumina supports was investigated between 383 and 523 K. For binary and ternary mixtures, n-octane permeated through the membrane faster than either n-hexane or isooctane, and a selectivity for n-octane over isooctane of 40 was obtained in a ternary mixture at 413 K. Selectivity was a function of temperature, however, and lower selectivities were obtained above and below 413 K. The n-octane/n-hexane selectivity was 9 at 413 K and decreased with increasing temperature. Permeances were a strong function of other organics present in the feed. The permeance of n-octane increased a factor of 16 in the presence of isooctane and n-hexane. The pure component permeances thus could not be used to predict separations for mixtures. Whereas n-octane was always the faster permeating compound in mixtures, pure isooctane permeated up to 5 times faster than pure n-octane, and pure n-hexane permeated faster than pure isooctane. That is, the largest molecule did not have the lowest permeance for pure components, and the smallest molecule did not have the highest permeance in mixtures. Thus relative permeances could not be predicted on the basis of size or shape alone. The permeances of the pure compounds were activated with activation energies between 18 and 45 kJ/mol. The mixed feed permeances also increased with temperature but did not follow an exponential relationship. Transport limitations caused by membrane saturation, diffusion through boundary layers on the feed side, and diffusion through the alumina support were not important.

No abstract