Metal-Supported Carbogenic Molecular Sieve Membranes - American

Jun 15, 1997 - Central Research and Development, E. I. du Pont de Nemours and Company, Inc.,. Wilmington ... within the surface pores of the support. ...
26 downloads 9 Views 403KB Size
2924

Ind. Eng. Chem. Res. 1997, 36, 2924-2930

Metal-Supported Carbogenic Molecular Sieve Membranes: Synthesis and Applications Madhav Acharya, Brenda A. Raich, and Henry C. Foley* Center for Catalytic Science and Technology, Department of Chemical Engineering, University of Delaware, Newark, Delaware 19716

Michael P. Harold and Jan J. Lerou Central Research and Development, E. I. du Pont de Nemours and Company, Inc., Wilmington, Delaware 19880

Carbogenic molecular sieve membranes supported on macroporous sintered stainless steel flat plates have been synthesized by pyrolysis of a (poly)furfuryl alcohol-acetone solution imbibed within the surface pores of the support. Permeances of H2, He, Ar, O2, N2, and SF6 were measured to determine the permselectivity of these membranes. Experiments performed on flat-plate membranes revealed molecular sieving behavior with the permeance decreasing with increasing molecular size. Separation factors of 2-3 for O2/N2 and up to 30 for H2/N2 were obtained from single gas permeation experiments at 293 K. Higher temperature experiments demonstrated the activated nature of transport of various molecules. A 1:1 O2/N2 mixture was partially separated over a range of pressures at steady state. The lack of pressure dependence of the permeances indicated that shape- and size-selective effects dominated the separation. An EDAX analysis of the flat-plate surface reveals a defect-free CMS layer. Introduction Carbogenic Molecular Sieves. Carbogenic molecular sieves (CMS) have gained considerable importance for their role in gas separations, adsorption, and various catalytic applications. CMS are glass-like materials which are typically synthesized by the high-temperature pyrolysis of organic polymers. This treatment leads to a complex nanostructure that is thought to consist of a network of aromatic and amorphous carbon domains. Figure 1 shows a high-resolution transmission electron microscopy (HRTEM) image of a CMS sample synthesized at 800 °C, and the network of graphite-like layers is clearly visible. However, only polymers that do not undergo transformation to the thermodynamically preferred graphite phase at high temperatures can be used for CMS synthesis. This “non-graphitizing” character, which is present in materials like poly(acrylonitrile) (PAN), poly(furfuryl alcohol) (PFA), and poly(vinylidene chloride) (PVDC), can be attributed to the presence of heteroatoms such as oxygen and nitrogen, as well as to the formation of cross-links between the polymer chains when pyrolyzed. This “frozen” cross-linked structure, which provides the molecular sieving nature of CMS, has been discussed in the literature (Foley, 1995; Lafyatis et al., 1991; Mariwala and Foley, 1994). Franklin was the first to characterize the nanostructure of polymer-derived CMS with small-angle scattering of X-rays (Franklin, 1950). The X-ray diffraction studies then, as now, did not show a distinct diffraction pattern even on the length scale of 25 Å. This revealed the globally amorphous nature of CMS. Recently, HRTEM studies of the structure, combined with FFT analysis, have been used to determine the spacing between the aromatic domains (Kane et al., 1996), and an analogy with fullerene carbons has been proposed. Most interestingly, and somewhat paradoxically, al* To whom correspondence should be addressed. Phone: (302) 831-6856. Fax: (302) 831-2085. E-mail: foley@ che.udel.edu. S0888-5885(96)00769-5 CCC: $14.00

Figure 1. HRTEM image of the nanoporous structure of CMS showing aromatic and graphitic microdomains: (a) normal gray scale; (b) image enhanced for clarity.

though the CMS do not have a unique long-range structure, they exhibit a narrow distribution of pores in the region of 4.5 ( 0.5 Å. Furthermore, the evolution of the nanostructure depends on the polymer precursor as well as the pyrolysis time and temperature and leads to predictable, reproducible changes in apparent pore size (Lafyatis et al., 1991; Mariwala and Foley, 1994). The pore size is referred to as “apparent” since it is an indirect measurement based on gas-adsorption experiments. Investigations have shown that for most precursors, high-temperature sintering leads to a diminution in porosity, with concomitant shrinkage of pores. Eventually, depending on the polymer precursor, a collapse of the structure and creation of macroporous defects occurs above a certain temperature, leading to a loss in sieving capability (Lamond et al., 1965; Mariwala and Foley, 1994). CMS Membranes. Pioneering work on gas transport through a carbon membrane was carried out by Barrer et al. (Ash et al., 1967, 1973). The study focused on the surface flow of gases under different conditions, and it was found that adsorbing gases such as argon exhibited strongly pressure-dependent diffusivities. In the case of gas mixtures, ammonia was found to be strongly © 1997 American Chemical Society

Ind. Eng. Chem. Res., Vol. 36, No. 8, 1997 2925

sorbed in the pores and, consequently, hindered the transport of helium and other light gases. Bird and Trimm (1983) pyrolyzed PFA on various support materials including silica frits, sintered bronze and copper, and iron gauzes. The researchers encountered the problem of being unable to create a uniform, defect-free layer on any support surface, with the exception of silica frits. However, there was some evidence for activated diffusion as well, and activation energies were obtained for different gas-support material pairs. Both planar and hollow fiber CMS membranes were developed by Koresh and Soffer (1983) by pyrolysis of polymer hollow fibers, probably consisting of poly(acrylonitrile). They found that the final apparent pore size of the CMS membrane was a function of its synthesis temperature and it could be enlarged by oxidative treatment. Interestingly, Koresh and Soffer (1986) reported findings similar to those of Ash et al.snonadsorbing gas permeances were independent of pressure, while adsorbing gases showed a decrease in permeance at higher pressures due to the nature of the adsorption isotherm. This provided further evidence for the significant effects of surface adsorption and transport in these membranes. Recently, the surface flow mechanism has been successfully used by Air Products and Chemicals (Rao and Sircar, 1993; Rao et al., 1992). In contrast to the membrane used by Barrer et al., which consisted simply of carbon powder compressed into membrane form, the membrane used by Rao and Sircar was synthesized by pyrolysis of Saran latex upon a porous graphite support. This distinction is crucial in recognizing the advance that the Air Products work represents. The membrane has been patented (Rao et al., 1992) and is used for recovery of waste hydrogen in plants. However, the rather low temperature at which the reported experiments were performed (-11 °C) may be a limitation that needs to be overcome. Graphite supports also were used by Chen and Yang (1994) to synthesize membranes from PFA. Again, the carbon layer was found to be crack free, and its thickness was 15 µm. Diffusivities of hydrocarbon mixtures in the membrane were found to be concentration dependent. The experimental data were explained quite well by the binary diffusivity theory developed by the authors. Jones and Koros carried out a series of experiments to study the effect of humidity and organic compounds on the performance of hollow fiber membranes (Jones and Koros, 1994, 1995a,b). They found a severe reduction in membrane fluxes at high humidity levels, due to the tendency of CMS to adsorb large amounts of water. A similar effect was seen with hydrocarbons, and the extent of fouling was considered to be dependent on the initial sorption rate on the CMS. To counter these detrimental effects, the authors developed a carbon composite membrane with a protective polymer layer on top of the hollow fiber membrane. Linkov et al. (1994a,b) used blends of different polymers to synthesize large-pore-size carbons. Mercury porosimetry analysis confirmed that their samples contained pores ranging from 100 to 1000 Å in size. These asymmetric carbon membranes were synthesized from acrylonitrile and methyl methacrylate and were observed to have three distinct regions: an outer layer of dense skin, a network of channels, and the lower lying macropores. Here we report on an alternative to the previous approaches in which we use a rigid, durable support

Figure 2. Typical structure of a supported CMS membrane.

material, namely, porous stainless steel, to provide mechanical strength to the membrane. The desired structure of such a membrane is shown in Figure 2. It consists of a thin CMS layer on top of a macroporous, nonselective support. Bonding between the two materials is achieved by allowing a small amount of CMS to penetrate the support pore structure. An ideal support should be an inexpensive material, available in various different geometries such as flat plates and tubes (to be used as per the requirements of the application), and it should allow for relatively easy module formation for implementation in current chemical processes. This might not provide a packing density as high as is possible with hollow fibers, but the trade-off would be practicality. The aim of this study was 2-foldsfirstly, to prepare metal-supported carbogenic molecular sieve membranes using established CMS synthesis techniques and, secondly, to characterize the morphology of these membranes and study their transport behavior for pure component gases and mixtures. Experimental Section Several supported CMS membranes were synthesized using furfuryl alcohol resin (Monomer Polymer & Dajac Laboratories Inc., Lot A-1-143) as the precursor material. Due to the high viscosity of the polymer, it was diluted in acetone to form a solution of approximately 60% by weight resin. The acetone was used as a thinner to allow easier coating on the support. Stainless steel supports of different pore sizes were obtained from Mott Metallurgical Corp. (Farmington, CT). Synthesis. Flat stainless steel supports (of diameter 1.875 in., thickness 0.039 in., and 0.2-µm pore size) were used as the support material. Prior to the coating operation, the flat-plate supports were cleaned with chloroform and allowed to dry in air. The precursor solution was brush-coated by hand onto the support to form a layer of uniform thickness. The coated sample was allowed to dry in air for a period of 6-12 h to allow complete evaporation of acetone. The samples were placed on a glass boat and pyrolyzed in a quartz tubular reactor placed inside a Lindberg single-zone furnace fitted with an Omega CN2041 temperature controller. The temperature inside the reactor was monitored by a J-type thermocouple, and an inert purge of helium was maintained during the pyrolysis. The pyrolysis protocol followed established methods used for the synthesis of

2926 Ind. Eng. Chem. Res., Vol. 36, No. 8, 1997 Table 1. Pure Component Gas Permances through Flat-Plate Membrane Synthesized at 600 °Ca temp, sulfur K hydrogen helium oxygen argon nitrogen hexafluoride 293 328 353 a

42.5

19.3 29.1 29.9

5.48

3.57 24.0 35.9

1.58 19.8 25.6

0.39 15.2 15.3

Units ) mol/(m2 s Pa) × 10-10.

Figure 3. Continuous flow membrane separation system.

unsupported CMS (Lafyatis et al., 1991; Mariwala and Foley, 1994). The flat plates were weighed before and after coating and then once again after pyrolysis. It was observed that pyrolysis led to a loss of around 65-70 wt % of the fresh coat, thus resulting in a 30-35 wt % yield of carbon on the support. Gas Separation Experiments. The membrane was sealed with Viton gaskets inside a module consisting of two double-sided flanges (MDC P/N 140013). On either side of these were single-sided flanges (MDC P/N 110008) that had welded stainless steel tubes provided for flow of gases. The rise time experiment was performed on a flat-plate sample synthesized at 600 °C (5 coats, 2 h soak time, 10 °C/min ramp rate) to determine gas permeances through it. Both sides of the membrane were initially at atmospheric pressure. A probe gas was introduced on the top side at 30 psig, and the downside pressure was monitored. Experiments were also performed at elevated temperatures to determine whether the transport was of an activated nature. The membrane module was purged with helium after each run to maintain identical operating conditions for each gas. Unsteady-State Experiment Analysis. A fairly simple model can be derived to describe the unsteadystate experiments. Writing a mass balance for the permeating species on the top side, we have

dM ) JMw dt

(1)

where m is the mass gas, J is the molar flux (mol/m2) across the membrane, and Mw is the gas molecular weight. The flux across the membrane can be expressed as

J)

π′ (P - PSS) L TS

(2)

where π′ is the gas permeability (mol/(m s Pa)) and L is the thickness of the CMS layer (m). Using the ideal gas law, the mass of the gas can be expressed in terms of PSS, and the final expression is

dPSS ART π′ ) (P - PSS) dt VSS L TS

[

]

(3)

Figure 4. Rise-time plot (293 K) for flat-plate membrane synthesized at 600 °C.

where VSS is the downside volume and A is the membrane surface area. Integrating form t ) 0 (PSS ) 0 psig), we get

VSS PTS π′ ln ) t ART PTS - PSS L

(4)

Thus, a plot of the left-hand side expression vs time (t) gives the permeance π0 ()π′/L) of the gas (mol/(m2 s Pa)). Table 1 provides the permeances of the various probe molecules at different temperatures. Separation of gas mixtures under steady-state conditions was also studied using the setup shown in Figure 3. Multicomponent mixtures could be fed on either side of the membrane, and the streams entering and exiting the membrane module could be analyzed using TCD/ FID. A series of experiments were performed on a feed stream (F1) containing 46.5% oxygen and 53.5% nitrogen. An inert sweep was maintained on the permeate side (F2) of the membrane. The flux of both components was calculated from the following equation:

Ji )

M2XiC ART

where Ji is the flux of species “i” (either oxygen or nitrogen), M2 is the permeate flow rate in cm3/s, Xi is the mole fraction of species i given by the GC calibrated with a known sample of oxygen and nitrogen, C is the net combined fraction of oxygen and nitrogen in the permeate stream (again based on calibrated GC areas), and A is the net surface area of the membrane available for gas permeation. The permeance of the gases was calculated using both CSTR and PFR assumptions. The average composition of the inlet and outlet streams was considered in the PFR case. Due to the very small change in composition, both assumptions gave nearly identical values for the

Ind. Eng. Chem. Res., Vol. 36, No. 8, 1997 2927 Table 2. Data for Steady-State Oxygen-Nitrogen Separation Experiment Performed on 600 °C Flat-Plate Membrane at 293 K top-side pressure (Pt1), atm

down-side pressure (P2), atm

2.06 3.16 4.31 6.27

1.06 1.07 1.07 1.07

feed comp (F1),

permeate comp (M2)

% O2

% N2

% O2

% N2

permeate concn in He, % (O2 + N2)

45.67 46.5 46.55 46.47

54.33 53.5 53.45 53.53

60.2 65.02 62.21 62.52

39.8 34.98 37.79 37.48

1.184 1.875 2.663 3.889

Figure 5. Temperature dependence of nitrogen permeance through flat-plate membrane synthesized at 600 °C.

retentate comp (M1) % O2

% N2

45.42 46.48 46.33 46.36

54.58 53.52 53.67 53.64

Figure 6. Steady-state permeance of oxygen and nitrogen through 600 °C CMS membrane as a function of mean pressure across membrane.

permeance. The expression for permeance of species i is

πi )

Ji Pi,1 - Pi,2

where Pi,1 and Pi,2 are the partial pressures of species i on the top and down side of the membrane, respectively. The runs were carried out for up to 12 h, and the compositions and flow rates were constant. No fouling of the membrane based on reduced separation factors was observed. Finally, SEM images of a few membranes were obtained, and EDAX analysis was carried out to determine the surface elemental composition. Results The ambient temperature rise-time plot reveals clear evidence of molecular sieving through the flat-plate membrane (Figure 4). Hydrogen and helium were transported at a much faster rate than molecules like argon, oxygen, and nitrogen. In fact, the rise-time curves are arranged quite nicely in order of increasing molecular size. The high sensitivity of the membrane to molecular size is evident from the fact that there is noticeable separation between oxygen, argon, and nitrogen, which differ by only 0.2 Å in size. The temperature-dependent data for nitrogen are shown in Figure 5. The data show a rapid increase in the rise time curve for a small rise in temperature. The increase was more pronounced for larger diameter molecules. As a result, the separation factors at elevated temperatures were reduced considerably. The permeate stream composition in the steady-state experiments (M2 in Figure 3) was found to vary from 60 to 65% oxygen over the range of pressures studied (Table 2). The permeances of both gases were fairly independent of pressure (Figure 6). Permeation of helium to the top side of the membrane was also observed. In order to determine whether this affected

Figure 7. SEM micrograph of CMS layer edge on 0.2-µm stainless steel support.

the separation or not, a final experiment was performed in the absence of a helium sweep. The top-side pressure was maintained at 6 atm, and the flow rate on the top side of the membrane was very low. The same (46.5% oxygen) composition mixture was fed on the top side, and the permeate composition was analyzed. Due to very low flow rates, the GC sampling valve load time was increased to 2 h to obtain a meaningful analysis. The result is consistent with the experiments performed with helium sweep. A steady composition of 61% oxygen was observed downstream after 22 h. After completion of the experiments described, the flat plate was bent and a small portion of the CMS layer was peeled off the surface. Shown in Figure 7 is an SEM image of the boundary between the remaining CMS layer and the underlying stainless steel support. The fact that a substantial part of the film was not destroyed suggests that there is a strong bonding interface with the support. An EDAX surface analysis (Figure 8) of the support (area A) and film (area B) confirms that the layer is indeed continuous and defect-

2928 Ind. Eng. Chem. Res., Vol. 36, No. 8, 1997

supports had thinner layers under the same synthesis conditions, due to penetration of the precursor into the matrix. Discussion

Figure 8. EDAX analysis of membrane imaged in Figure 7.

free. These images were obtained on a Hitachi S-4000 instrument. An STM image of the CMS layer revealed a very low surface roughness of the order of 1 nm (Figure 9). However, identification of distinct pores was not possible through this image. An SEM micrograph of a flat-plate membrane is shown in Figure 10. The sample was coated 10 times with 60 wt % PFA in acetone solution and pyrolyzed at 600 °C. The macroporous support had a pore size of 0.2 µm. The micrograph clearly reveals a layer of pyrolyzed carbon of average thickness 10 µm on the support surface. Samples synthesized on larger pore

Figure 9. STM image of CMS membrane synthesized at 600 °C.

The rise-time data obtained on the flat-plate membrane are indicative of molecular sieving, with activated transport of molecular species. A plot of ln(π0T) vs 1/T was used to regress the activation energies for the different molecules (Figure 11). However, we noticed a deviation from the straight line fit in going to the highest temperature (353 K). Hence, we also considered a temperature-dependent activation factor of the form A ) BTn and tried different values of the parameter n to fit the datashowever, the fit was still not very satisfactory. We used both two and three data points to calculate the activation energies and obtained a range for the different molecules. The linear dependence of activation energy on molecular diameter is similar to that for established materials like Zeolite 4A and MSC 5A. The relative location and slope of the line for the CMS membrane suggests that it has pores smaller than 4 Å; however, a direct comparison cannot be made with confidence, since the activation energies here are for permeance rather than diffusivity, as in the case of the zeolites. Even so, the result is suggestive of what we expected the pore size to be and appears to be consistent with the nature of the rise-time plots, where an imaginary line can be traced between the curves for helium (2.6 Å) and oxygen (3.46 Å). The slight deviation of the activation energy plot in Figure 12 from a straight line at higher temperatures could be an indication of a possible change in transport mechanism through the membrane. The pores in CMS are generally thought to consist of narrow mouths, followed by a larger cavity where molecules can be potentially trapped. This “ink-bottle-like” picture of the pore results in there being two different types of transportsone with a high activation energy at the pore mouth and the other with a lower activation energy in the larger channel. Walker et al. (1965) first proposed this picture of the structure of CMS in their analysis of transport within them. This analysis, which involved transition-state theory of diffusion in zeolites as well as CMS, was well ahead of its time. At low temperatures, the limiting step was taken to be at the pore

Ind. Eng. Chem. Res., Vol. 36, No. 8, 1997 2929

Figure 10. SEM micrograph of flat-plate membrane synthesized at 600 °C on a 0.2-µm stainless steel support.

Knudsen value of 0.94. The separation factors are also independent of pressure, which is of considerable benefit, since this implies that the membrane can be operated at high pressure drops without any effect on its performance. An interesting observation in these experiments was the backflow of helium from the low to the high pressure side of the membrane. Rao and Sircar (1993) comment on this phenomenon in their paper on SSF membranessin their case, the absence of this backflow (despite the presence of a driving force for helium) was attributed to blockage of the membrane by large hydrocarbon molecules exhibiting surface flow. Since nitrogen and oxygen are not strongly adsorbing molecules, helium backflow would be expected in our experiments. The experiment performed in the absence of helium flow confirms that the presence of a sweep gas does not interfere with the permeation and separation of oxygen and nitrogen. Conclusions

Figure 11. Regression of activation energy.

Figure 12. Activation energy range for permeances of different gas molecules as a function of molecular diameter measured on a flat-plate membrane synthesized at 600 °C vs activation energies for diffusion through Zeolite 4A and MSC 5A.

mouthsbut at higher temperatures, this barrier was overcome, and the overall activation energy in the large channel model was very low and nearly inconsequential. This seems to be the case based on our experiments with CMS membranes. The pressure independence of the permeances in the steady-state experiment with oxygen and nitrogen suggests that the membrane has very few cracks and is predominantly molecular sieving in nature. The separation factor for oxygen and nitrogen is found to be around 2 at steady state, a value well above the

In this study, carbogenic membranes on metal supports were synthesized using the ramp and soak pyrolysis technique. The samples were brush coated several times with a PFA-acetone precursor solution. Probe gas permeance experiments indicated some form of sieving by the membrane, with lighter gases experiencing higher transport rates. For the flat-plate membranes synthesized in this way, ideal separation factors based on single component permeation experiments were much higher than predicted by Knudsen transport. Furthermore, there was clear evidence for activated transport. The partial separation of an oxygen/nitrogen mixture, although not commercially significant at this stage, did illustrate that shape- and size-selective diffusion dominated at steady state. EDAX analysis of a flat plate revealed a continuous, defect-free CMS layer on the surface. With these data, the feasibility of a porous metalsupported CMS membrane, with all its advantage over other systems, has been demonstrated. The composite CMS membranes were durable and showed very little change in behavior over a period of 6 months. The design, construction, and use of a flange module unit were also very convenient, and its scale-up to commercial applications should not pose much of a problem, provided issues of forming the CMS layer can be overcome. Viton gaskets used in these experiments with the flat-plate membranes limit the maximum temperature in the flange module to 150 °C, and this needs to be improved if we are to construct a membrane reactor capable of high-temperature reaction. The metal support was able to withstand the high temperatures of pyrolysissin a control experiment, an uncoated support was heat treated at 800 °C, and there was negligible change in its behavior. Of course, the permeance of the uncoated flat plate was orders of magnitude higher than the CMS membranes. Acknowledgment We thank the DuPont for providing funding of this study and help in SEM/EDAX analysis of membrane samples. The help of Mahmoud Kaba and Prof. Mark Barteau in performing STM imaging of membrane samples is also greatly appreciated.

2930 Ind. Eng. Chem. Res., Vol. 36, No. 8, 1997

Nomenclature A ) membrane surface area, m2 C ) combined fraction of oxygen and nitrogen in permeate stream F1 ) top-side feed stream to membrane module, as shown in Figure 3 F2 ) down-side inert feed to membrane module, as shown in Figure 3 J ) molar flux, mol/(m2 s) L ) membrane thickness, m M1 ) retentate stream from membrane module as shown in Figure 3, cm3/s M2 ) permeate stream from membrane module as shown in Figure 3, cm3/s m ) mass of gas, kg Mw ) molecular weight, kg/kmol P ) pressure, Pa R ) gas constant, cm3 Pa/(mol K) T ) temperature, K VSS ) volume of down side of flange module, cm3 X ) mole fraction Greek Symbols πo ) gas permeance, mol/(m2 s Pa) π′ ) gas permeability, mol/(m s Pa) Subscripts TS ) topside of flange module SS ) downside of flange module i ) molecular species

Literature Cited Ash, R.; Baker, R. W.; Barrer, R. M. Sorption and surface flow in graphitized carbon membranes. Proc. R. Soc. London, Ser. A 1967, 299, 434. Ash, R.; Barrer, R. M.; Lowson, R. T. Transport of Single Gases and of Binary Gas Mixtures in a Microporous Carbon Membrane. J. Chem. Soc., Faraday Trans. 1 1973, 69, 2166-2178. Bird, A. J.; Trimm, D. L. Carbon Molecular Sieves used in Gas Separation Membranes. Carbon 1983, 21, 177. Chen, Y. D.; Yang, R. T. Preparation of Carbon Molecular Sieve Membrane and Diffusion of Binary Mixtures in the Membrane. Ind. Eng. Chem. Res. 1994, 33, 3146-3153. Foley, H. C. Carbogenic molecular sieves: synthesis, properties and applications. Microporous Mater. 1995, 4, 407-433. Franklin, R. E. The Interpretation of Diffuse X-ray Diagrams of Carbon. Acta Crystallogr. 1950, 3, 107. Jones, C. W.; Koros, W. J. Carbon Molecular Sieve Gas Separation Membranes IsPreparation and Characterization based on Polyimide Precursors. Carbon 1994, 32, 1419-1425.

Jones, C. W.; Koros, W. J. Carbon-composite membranes: A Solution to adverse humidity effects. Ind. Eng. Chem. Res. 1995a, 34, 164-167. Jones, C. W.; Koros, W. J. Characterization of Ultramicroporous Carbon Membranes with Humidified feeds. Ind. Eng. Chem. Res. 1995b, 34, 158-163. Kane, M. S.; Goellner, J. F.; Foley, H. C.; DiFrancesco, R.; Billinge, S. J. L.; Allard, L. F. Symmetry Breaking in Nanostructure Development of Carbogenic Molecular Sieves: Effects of Morphological Pattern Formation on Oxygen and Nitrogen Transport. Chem. Mater. 1996, 8, 2159-2171. Koresh, J. E.; Soffer, A. Molecular Sieve Carbon Permselective Membrane Part I. Presentation of a New Device for Gas Mixture Separation. Sep. Sci. Technol. 1983, 18, 723-734. Koresh, J. E.; Soffer, A. Mechanism of Permeation through Molecular Sieve Carbon Membrane. J. Chem. Soc., Faraday Trans. 1 1986, 82, 2057-2063. Lafyatis, D. S.; Tung, J.; Foley, H. C. Poly(furfuryl alcohol)-Derived Carbon Molecular Sieves: Dependence of Adsorptive Properties on Carbonization Temperature, Time and Poly(ethylene glycol) Additives. Ind. Eng. Chem. Res. 1991, 30, 865-873. Lamond, T. G.; Metcalf, J. E., III.; Walker, P. L., Jr. Molecular Sieve properties of Saran type Carbons. Carbon 1965, 3, 5963. Linkov, V. M.; Sanderson, R. D.; Jacobs, E. P. Carbon Membranes from Precursors containing low-carbon residual polymers. Polym. Int. 1994a, 35, 239-242. Linkov, V. M.; Sanderson, R. D.; Jacobs, E. P. Highly asymmetrical carbon membranes. J. Membr. Sci. 1994b, 95, 93-99. Mariwala, R. K.; Foley, H. C. Evolution of Ultramicroporous Adsorptive Structure in Poly(furfuryl alcohol)-Derived Carbogenic Molecular Sieves. Ind. Eng. Chem. Res. 1994, 33, 607615. Rao, M. B.; Sircar, S. Nanoporous carbon membranes for separation of gas mixtures by surface selective flow. J. Membr. Sci. 1993, 85, 253-264. Rao, M. B.; Sircar, S.; Golden, T. C. Gas separation by Adsorbent Membranes. U.S. Patent 5,104,425, 1992. Walker, P. L., Jr.; Austin, L. G.; Nandi, S. P. Activated Diffusion of Gases in Molecular-Sieve Materials. In Chemistry and Physics of Carbon; Walker, P. L., Jr., Ed.; Marcel Dekker: New York, 1965.

Received for review December 2, 1996 Revised manuscript received February 3, 1997 Accepted February 4, 1997X IE960769D

X Abstract published in Advance ACS Abstracts, June 15, 1997.