Nanoporous carbon membranes for separation of gas mixtures by selective surface flow

Nanoporous carbon membranes for separation of gas mixtures by selective surface flow

Journal of Membrane Science, 85 (1993) 253-264 253 Elsevier Science Publishers B.V., Amsterdam Nanoporous carbon membranes for separation of gas mi...

1MB Sizes 0 Downloads 6 Views

Journal of Membrane Science, 85 (1993) 253-264


Elsevier Science Publishers B.V., Amsterdam

Nanoporous carbon membranes for separation of gas mixtures by selective surface flow M.B. Kao and S. Sircar* Air Products and Chemicaki, Inc., 7201 Hamilton Boulevard, Allentown, PA 18195 (USA)

(Received April 27,1993; accepted in revised form July 12,1993)

Abstract A novel nanoporous carbon membrane produced by carbonization of poly(vinylidene chloride) is described. The membrane separates gas mixtures by selective adsorption and surface diffusion of the more strongly adsorbed components. In particular, the membrane can very efficiently separate hydrocarbons from hydrogen from low-pressure gas streams. The method of preparation of the membrane and its characterization, pure and mixed gas permeabilities through the membrane, as well as an industrial application are reported. Key words: gas separations; inorganic membranes; microporous and porous membranes; carbon

Introduction There has been an exponential growth in the research and development of porous and nonporous membrane materials for separation of gas mixtures during the last 25 years. Numerous non-porous thin-film solid (polymeric and inorganic) or liquid membranes with or without the incorporation of a facilitated transport agent have been synthesized and evaluated. Three excellent reviews by Matson, Lopez and Quinn [ 11, Kesting [ 21, and Way and Noble [ 31 provide comprehensive descriptions of the state of the art in this area. The synthesis and evaluation of porous membranes for gas separation has also made remarkable progress, albeit somewhat limited, in comparison with the non-porous category. Ulhom and Burggraaf [ 41

and Hsieh [ 51 have recently reviewed the state of the art in this area. It is generally recognized that non-porous membranes for gas separation can exhibit very high selectivity of separation for certain components of a gas mixture, but the rates of transport of the gases through these membranes are usually low due to the solution-diffusion or solution-reaction-diffusion mechanisms of transport. Porous membranes, on the other hand, can generally provide very high rates of transport for gases, but they exhibit relatively lower separation selectivity, although the situation is changing [ 41. Four different mechanisms for separation of a gas mixture through a porous membrane can be identified: (a) Separation based on the differences in the molecular weights of the components of a

l’o whom correspondence should be addressed.


0 1993 Elsevier Science Publishers B.V. All rights reserved.


gas mixture due to Knudsen diffusion through the pores. (b) Separation based on molecular sieving caused by passage of smaller molecules of a gas mixture through the pores while the larger molecules are obstructed. (c) Partial condensation of some components of a gas mixture in the pores with exclusion of others and subsequent transport of the condensed molecules across the pore. (d) Selective adsorption of the more strongly adsorbed components of a gas mixture onto the pore surface followed by surface diffusion of the adsorbed molecules across the pore. A concentration gradient for the diffusing species must be imposed across the porous membrane in order to provide the driving force for transport by all four mechanisms. The selectivity of separation achievable by mechanism (a) is generally very low and not practical except in very special cases. Mechanism (b ) has been successfully used by Koresh and Soffer [ 61 for separating gas mixtures using molecular sieving carbon membranes. These membranes exhibit high selectivity and permeability for the smaller components of a gas mixture but they require a very fine control of the pore sizes (diameter < 4 A) which are equal to or slightly bigger than the size of the diffusing molecules. Furthermore, since these membranes have thicknesses of several microns, they may also require operation at an elevated temperature in order to provide practically acceptable flux for the smaller molecules. Mechanism (c) requires that the pores of the membrane be in the mesoporous size range (diameter > 30 A) so that condensation of the components of a gas mixture can take place. A very high selectivity of separation of the condensable component can be achieved by this mechanism, but the extent of removal of that component from the gas mixture is limited by the condensation partial pressure of that component at the system temperature (which is

M.B. Rao and S. Sircar / J. Membrane Sci. 85 (1993) 253-264

dictated by the Kelvin equation) and the pore size and geometry of the membrane [ 71. Mechanism (d), in principle, provides the most flexible and attractive choice for practical separation of gas mixtures because the separation selectivity is determined by preferential adsorption of certain components of the gas mixture on the surface of the membrane pores, as well as by selective diffusion of the adsorbed molecules. Thus, both the pore size and the physicochemical nature of the pore surface play key roles in determining the separation efficiency of these membranes. Consequently, the properties of these membranes can be altered by appropriate molecular engineering of the surface chemistry. We have developed membranes which rely on this mechanism and call them “Selective Surface Flow ( SSFTM) Membranes”. The SSFTM membranes offer the following key advantages for separation of gas mixtures: (i) A significant amount of the more strongly adsorbed components of a gas mixture can be adsorbed into the pores at the high-pressure side of the membrane even when the partial pressure of that component is relatively low. Thus, the imposed partial pressure gradient of the diffusing species across the membrane need not be very large to obtain practical separation. This is in contrast with thin-film solid membranes which require a large partial pressure gradient across them for sufficient dissolution of the transporting components. (ii) The larger or more polar molecules of a gas mixture can be selectively adsorbed and preferentially diffused across the membrane pore surface instead of the smaller molecules, as in mechanism (b), or as in most thin-film solid polymeric membranes. Thus the smaller molecules, which often constitute the desired product in a practical separation, remain on the high-pressure side of the membrane (at essentially the feed gas pressure) as the enriched product gas. This eliminates the need for re-

M.B. Rao and S. Sircar / J. Membrane Sci. 85 (1993) 253-264

compression of the product gas for many applications. (iii) The driving force for mass transfer across the membrane is the difference in the adsorbed phase concentration of the diffusing species. Thus, a large driving force can be achieved (in view of (i) ) even when the partial pressure of the diffusing component is low at the high-pressure side of the membrane. Furthermore, the activation energy for surface diffusion can be significantly lower than that for transport through a membrane using mechanism (b) or a polymeric membrane [ 21. As a result, surface flow membranes can provide very high fluxes for the diffusing species without requiring an ultra-thin configuration or a large pressure differential across them. (iv) Adsorption capacity and selectivity increase with decreasing temperature. Thus, the membrane performance can be improved by operation at near ambient or a lower temperature. The reverse is true for efficient operation of a molecular sieve membrane or many polymeric membranes. (v ) The adsorbed molecules on the pore surface can effectively reduce the open void space across the pores. This can eliminate or significantly hinder the diffusion (mostly activated or Knudsen) of non-adsorbed molecules through the void space, which enhances the selectivity of separation [ 81. All of these conceptual advantages of a surface flow membrane for gas separation have been experimentally demonstrated by Barrer and coworkers [ 3-111. The original paper describing the phenomenon was published in 1955 [9 1. They used highly compressed plugs of nonporous graphitized carbon (3.3-4.8 cm long) as the membrane. Many other researchers [ 4,1215] have demonstrated the existence of surface flow through various porous substrates such as Vycor glass, alumina, silica, zeolites, etc. The purpose of this paper is to describe the preparation, the properties and the application

of a novel nanoporous carbon membrane which can be used to separate hydrogen-hydrocarbon mixtures by the surface diffusion mechanism. The technology has been patented by Air Products and Chemicals [ 161. Requirements for a practical surface flow membrane

There are four critical requirements for a practical surface flow membrane: (a) The diameter of the pores in the membrane must be larger than the diameter of the adsorbing molecules but should not exceed three to four times the diameter of the largest molecule of the feed gas mixture to be separated. Otherwise, the diffusion of the less strongly adsorbed components through the void space in the membrane would obliterate the selectivity of separation. A typical acceptable pore diameter range may be between 4 and 15 A. (b) The pores must be continuous across the thickness of the membrane and the surface density of these pores (the ratio of pore area to geometric surface area of the membrane) should be greater than 0.2. (c) Even though the flux of the adsorbed species through the membrane can be large, the membrane thickness should not exceed 5 pm for obtaining a practically acceptable permeance.

Fig. 1. Separation mechanism in surface flow membrane.


(d) The thin membrane has to be supported on a meso- or macroporous support for structural integrity. Figure 1 shows a schematic description of a practical membrane structure. The membrane can be formed as a thin layer on the surface of a porous support or it can be formed within the pores of the support. The figure also shows a schematic of the transport mechanism through the membrane for hydrogen-hydrocarbon mixture separation. The mixture is passed over the membrane surface at the high-pressure side. The hydrocarbon molecules are selectively adsorbed over hydrogen onto the pores. The adsorbed molecules diffuse on the surface towards the low-pressure side of the membrane where they desorb to form the permeate stream. Consequently, a hydrogen-rich product gas at the feed-gas pressure is produced on the highpressure side of the membrane and a hydrocar-

M.B. Rao and S. Sircar / J. Membrane

Sci. 85 (1993) 253-264

bon-rich waste gas is produced sure side.

at the low-pres-

Membrane preparation We developed a novel technique for preparing nanoporous carbon membranes which satisfies all of the requirements of a practical membrane discussed earlier. A disk of macroporous graphite (average pore diameter 0.7 pm) was used as the support. It was coated with a thin uniform layer of a poly (vinylidene chloride )-acrylate terpolymer latex which contained 0.1-0.14 pm polymer beads in aqueous suspension (55 wt% solid). The coated graphite sheet was spun to remove excess latex and then dried at room temperature. It was then further dried under N, at 150” C for a few minutes followed by heating under a N, purge to 1000°C. The temperature

Fig. 2. Scanning electron micrograph of a nanoporous five-coated carbon membrane.


M.B. Rao and S. Sircar / J. Membrane Sci. 85 (1993) 253-264

was ramped at a rate of approximately 1 “C/ min during the heating period, followed by holding the temperature constant at 1000 ’ C for three hours. The temperature was then reduced to ambient at a rate of lO”C/min while maintaining the inert purge. This produced the single-coated nanoporous carbon membrane. It was formed by crosslinking and carbonization of the polymer. The procedure was repeated several times to produce a multicoated membrane. The results reported in this paper are for four- and five-coated carbon membranes. Figure 2 is a scanning electron micrograph of a five-coated membrane. It was measured using a magnification of 20,000. It may be seen that the membrane consists of five uniform layers of carbon without any visible cracks (1000 A resolution). The thickness of each layer is approximately 0.5 pm giving a total membrane thickness of 2.5 pm. Permeation characteristics Pure gas permeabilities The pure gas permeabilities through the carbon membrane were measured using a volumetric membrane apparatus. Figure 3 shows a schematic drawing of the apparatus. It consisted of a thermostated cylindrical membrane holder (B ) consisting of a high-pressure and a low-pressure side. The carbon membrane (M ) could be mounted in the holder using O-rings, which provided the gas-tight seal between the high-pressure (H) and low-pressure (L) sides. Both sides were fitted with an inlet and an outlet port and corresponding regulating valves (Hl, H2, Ll and L2). In addition, the low-pressure side of the membrane was connected to a gas holder ( G ) through another regulating valve (L3 ) . The gas holder was equipped with a Heise pressure gauge (PL). The gas pressure in the high-pressure side was monitored by pressure gauge ( PH ) . Both sites of the membrane could

r---------1 CONSTANT







L @&TO


Fig. 3. Schematic of membrane apparatus for measuring pure and multicomponent gas permeabilities through carbon membrane.

be connected to a vacuum pump through appropriate valves (H3 and L4). The volume ( VL) of the low-pressure side of the membrane including the gas holder was measured by sealing off the high-pressure side using a metal disc having a thickness equal to that of the carbon membrane, and expanding helium into that side from a tank of known pressure. The experiment for measuring pure gas permeabilities (Pp ) through the carbon membrane consisted of evacuating both sides of the membrane to 10 -’ atm pressure and then flowing the pure gas through valves Hl and H2 at a constant pressure of P& and monitoring the pressure change in the gas holder, PL (t), as a function of time (t). Valves H3, Ll, L2 and L4 were closed and valve L3 was opened during that time. A certain period of time (t*) was needed to attain a constant pressure in the highpressure side (typically 5-10 set). The pressure in the gas holder rose to PE during that time. (Pp ) was defined by:







M.B. Rao and S. Sircar / J. Membrane Sci. 85 (1993) 253-264

Jp (t) is the transient molar rate of transport of pure gas i across the membrane at time t. A is the surface area of the membrane and 2 is its thickness. Pn and PL are, respectively, the gas pressures in the high- and low-pressure sides of the membrane at time t. R is the gas constant and T is the temperature in the gas holder. According to the experiment described earlier, eqn. (1) can be integrated to obtain a relationship between PL (t) and t after a time period oft*, as

five-coated carbon membrane at 295 K. The P& for both cases were 1.85 atm. It should be noted here that the Pp values measured by the above-described technique represent overall permeabilities of the pure gases through the carbon membrane under a pressure gradient across the membrane. No attempt was made to deconvolute the surface diffusion and void-gas diffusion contributions through the membrane because of the uncertainty in such procedures [ 171. Mixed gas permeabilities





Equation (2) shows that a plot of the left-handside term against ( t - t * ) should give a straight line with a slope of (+A*$*$).


quently, Pp can be estimated by knowing A, 1, VLandT. It was assumed in the derivation of eqn. (2 ) that Pp is independent of pressure. Figure 4 shows two typical plots according to eqn. (2) for permeation of pure Hz and CzHs through a









A = 13.203 VL=


1’ = 0.15

cm2 cm3







6 (t - t’),







Fig. 4. Examples of pressure changes in the low-pressure side of membrane apparatus for pure gas permeation through a five-coated membrane.

The mixed gas permeabilities through the carbon membrane were measured using the same apparatus described by Fig. 3. In this case, valves L3 and H3 were closed and a stream of a binary or multicomponent gas mixture of constant composition was passed across the high-pressure side of the membrane at a constant pressure of P&. The gas was introduced to the system through valve Hl and the effluent was withdrawn through valve H2. Simultaneously, a stream of pure helium (sweep gas) was introduced into the low-pressure side of the membrane through valve Ll at a constant pressure of Pt (
M.B. Rao and S. Sircar / J. Membrane Sci. 85 (1993) 253-264

ponent i through the membrane from a mixture (pi ) was then estimated using steady-state values of flow rates and compositions and the following definition: ~~ =A_E.



1 ln{ (dpi)l/(dpi)u}=Q

L fL ‘.


where Ji is the steady-state molar rate of transport of component i through the membrane. (dpi)r is the difference in the partial pressures of component i between the high- and lowpressure sides of the membrane at the highpressure gas entrance end (I). (Api)” is the difference in the partial pressures of component i between the high- and low-pressure sides of the membrane at the high-pressure gas exit end (II). QL is the steady-state effluent gas flow rate at the low-pressure gas exit end (I), and yiL is its steady-state mole fraction of component i. This log-mean partial pressure driving force for the transport of the components across the membrane was chosen for eqn. (3) because of the uncertain flow patterns of the gases within the membrane holder between the inlet and the outlet ports due to varying cross-sectional area of the flow path. Equation (3) shows that pi can be calculated using experimentally measured values of QL, yii, (dpi)‘, (dpi)“, A and 1. Equation (3) reduces to eqn. (1) for a pure gas. It should be mentioned here that pi values calculated by assuming CSTR behavior on both sides of the membrane or by assuming a constant arithmetic average partial pressures of the components (between the inlet and outlet) on both sides of the membrane, were very close to those obtained by eqn. (3) due to small changes in compositions between inlet and outlet streams on both sides. Another interesting observation was that helium did not permeate from the low- to the highpressure side during the experiments due to hindrance created by more strongly adsorbed molecules within the membrane pores even


though a driving force for helium transport existed. Permeation of pure hydrogen and helium Pure gas permeabilities of helium and hydrogen through a four-coated carbon membrane were measured at 298.5,273.1 and 255.4 K using a Pk value of 1.85 atm. Table 1 reports these data. The table also shows the effective diffusivities (0: ) for these pure gases defined by J;(t) =A.?.

[CH(t) -C,(t)].r,


Cn and CL are the gas densities at the high- and low-pressure sides of the membrane. A, 1 and Jp (t) are the same quantities as defined for eqn. (1) . tM is the void fraction in the membrane. Assuming ideal gas behavior (C=P/RT), eqns. ( 1) and (4) can be compared to obtain Dp =Pp RT/cM


Table 1 shows that the ratios of permeabilities (or diffusivities) of H2 to He vary between 4.0 and 5.5 for the membrane in the temperature range of the data. For Knudsen diffusion, the ratio of permeabilities for these gases should be equal to 1.414 and be independent of temperature. The molecular diameters of Hz (2.9 A) and He (2.6 A) are very close and it is not expected that the tortuosity of the diffusional path for these two gases across the membrane can be very different. Furthermore, substantial adsorption of these gases on the carbon surface at the temperature and pressure of the experiments is not expected. Consequently, we concluded that the mechanism of transport of these non-adsorbing gases through the membrane was activated permeation described by p; =PT e--EIRT


The activation energies (E) for He and Hz were calculated to be respectively 1.63 and 0.72 kcal/ mol. PT is the value of Spat the limit of T-tm.


M.B. Rao and S. Sircar / J. Membrane

Sci. 85 (1993) 253-264

TABLE 1 Permeabilities and diffusivities of pure hydrogen and helium through four-coated carbon membrane T

Permeabilities (Barrers )


256.1 273.1 298.1

Diffusivity (cm2/sec X 105)






280 340 440

1520 1700 1850

0.55 0.72 1.01

3.00 3.59 4.26

5.4 5.0 4.2

I’& = 1.85 atm; e,=O.36.

Figure 5 shows the calculated diffusivity (circles) of N2 (D, cm2/sec) as a function of pore diameter (d, A) at 296 K. Bulk diffusion dominates when the pore diameter is greater than 1000 A. Knudsen diffusion controls the N2 flow in the pore diameter range of 10-1000 A. Activated diffusion is predominant when d < 10 A. The activated diffusivities for N, were obtained from experimental diffusivities in 3A, 4A and 5A zeolites [ 181. We measured pure N2 permeation through the five-coated membrane at 295.1 K which is reported in Table 2. It may be seen by comparing Fig. 5 and the measured N2 diffusivity (cal-

d, 8 2345




*030405D II,,










-fl/ N2 AT 296


,, 3



Permeabilities of pure hydrogen, hydrocarbons and nitrogen through five-coated carbon membrane at 295.1 K Gas

Permeabilities (Barrers)

Nz* Hz GH, ‘%H, GsHs G,H,,

75 130 660 850 290 155

“Nitrogen diftirsivity= 1.71 x 10e6 cm’/sec; P& = 1.17 atm.

culated by eqn. 5) that the average pore diameters of the membrane are between 5 and 6 A. This may not be an exact pore size since N2 is weakly adsorbed. Thus, the carbon membranes produced by the technique described in this paper contain pores in the nanometer size range which are continuous across the membrane thickness. The membranes are < 5 pm thick and they are supported on a mesoporous structure. Consequently, all requirements of a practical membrane described earlier are satisfied. Permeation of pure hydrogen and hydrocarbons





= ii)

Fig. 5. Nitrogen diffusivity as a function of pore diameter at 295 K.

The permeabilities of pure H2 and C,-C, hydrocarbons were measured on a five-coated membrane at 295.1 K using a P& value of 1.85


M.B. Rao and S. Sircar / J. Membrane Sci. 85 (1993) 253-264

atm. Table 2 shows the results. The Hz diffusivity decreased by an order of magnitude on the five-coated membrane compared to that for the four-coated membrane. We calculated an effective Knudsen pore diameter corresponding to these diffusivity values. They were, respectively, 0.08 and 0.006 A for the four- and five-coated membranes, which are well below the molecular diameter of HP Thus, these numbers are fictitious and Knudsen flow is not the mechanism of transport. This also indicates that there are no defects in either of the membranes. The difference of Hz diffusivity between the four- and five-coated membranes is caused by a decrease in the size of the nanopores. A very small change ( N 0.5 A) in the pore diameter in the 3-10 A range reduces the diffusivity by an order of magnitude as shown in Fig. 5 for Nz diffusion. All hydrocarbons exhibited larger pure gas permeabilities than Hz because of their preferential adsorption on the carbon surface compared to Hz. The permeabilities were in the order H2 < CHI < C&He> C&H8> C!,H,,. The permeabilities increased from Hz to CH, to C&He.Then they progressively decreased as the molecular weight of the hydrocarbon chain increased. This behavior is consistent with the observation that the specific amount of the hydrocarbons adsorbed in the carbon pores at any given pressure and temperature may increase as the hydrocarbon molecular weight increases, but their strength of adsorption also increases, causing the higher hydrocarbons to be less mobile on the surface. Permeability of hydrogen-hydrocarbon mixtures It is clear from the data of Table 2 that the five-coated carbon membrane can be used to separate H, from mixtures with hydrocarbons by selective adsorption and permeation of the hydrocarbons through the pores. However, the

hydrocarbons will compete with each other due to selective adsorption. Thus, permeabilities from a mixture will be different than those measured for pure components. We measured the permeabilities of the components of a multicomponent gas mixture containing 41.0% HP, 20.2% CHI, 9.5% C2H6, 9.4% C&H8and 19.9% &HI0 (mol% ) through a five-coated carbon membrane using a P& of 4.4 atm at 295.1 K. A helium sweep at 1.07 atm ( = PE ) was used on the low-pressure side of the membrane. The steady-state permeabilities of the components are given in Table 3. The table also gives the mixture selectivity of separation for the hydrocarbons over hydrogen. It may be seen that the permeabilities of the most weakly adsorbed components (H, and CH,) are drastically reduced in the presence of the higher hydrocarbons compared to their pure gas permeabilities (Table 2). The most selectively adsorbed component, C4H1,,,permeates through the membrane at a very high rate followed by C&H8and CzHG. In other words, the order of permeabilities of the hydrocarbons follow the order of their strength of adsorption on the carbon. Consequently, the selectivity of separation of C,H,, and C3HBover Hz is very high, being, respectively, 94.4 and 21.3. The selectivity of separation of C&H, over Hz is 6.5, TABLE 3 Permeabilities of components of a gas mixture’ through five-coated membrane at 295.1 K Gas


cZ& C&B CJ-L

Permeability (Barrer)

1.2 1.3


25.4 112.3

Selectivity over Hz from mixture

Pure gas permeability ratio

1.1 6.5 21.3 94.4

5.1 6.6 2.3 1.2



“41.0% Ha, 20.2% CHI, 9.5% C2H8, 9.4% C3Hs, 19.9% C,H,,. Pi?, =4.4 atm.


M.B. Rae and S. Sir-car/J.

while CH, shows practically no selectivity of separation over HP. These selectivities are far different from those obtained by ratios of pure gas permeabilities which demonstrates the role of competitive adsorption and surface diffusion of the components within the carbon pores. It should be recognized that C,H, will be the most selectively adsorbed component in absence of C&H,, and its selectivity of separation over H2 will be much larger than the value given in Table 3. Similarly, C&H, will exhibit a much higher selectivity of separation over H, than that exhibited from the above-described mixture if higher hydrocarbons were absent. A key observation from the data of Table 3 is that the adsorption of higher hydrocarbons effectively blocks the permeation of Hz through the void space in the pores. Otherwise, the H2 permeability from the gas mixture could not be reduced by approximately two orders of magnitude compared to the pure gas permeability. This demonstrates the “hindrance effect”, which can be a major advantage of a nanoporous membrane relying on surface diffusion.


Sci. 85 (1993) 253-264

Practical application of SSFTM membrane Many refinery waste gases contain H2 and hydrocarbon mixtures which are lean in H2 ( < 50.0 mol% ) and are available at low pressures ( < 50 psig). These gases are usually burned as fuel because the cost of recovery of H, by conventional methods is prohibitive. A carbon SSFTM membrane may be used to recover a good portion of the H2 from such mixtures without further compression of the feed gas while rejecting a substantial portion of the hydrocarbons. The Hz-enriched stream, which leaves the membrane at essentially the feed pressure, can then be compressed and separated in a conventional pressure swing adsorption (PSA) process to produce ultra-pure Hz. The waste gas from the PSA system containing some H, and lower hydrocarbons can be used to provide the low-pressure purge stream for the membrane. Figure 6 shows a schematic flow sheet for such a membrane-PSA hybrid scheme. We built a plate-and-frame membrane mod-


40.9% H2 20.2% CH4 19.6% C2H6 9.2% C3 Hg



100.0% C4 Hi o REJECTION 9 1.1% C3 H, REJECTION 67.5% C2 H6 REJECTION 36.0% CH4REJECTION H2 RECOVERY = 63.0%

Fig. 6. Schematic flow sheet for SSF TMmembrane-adsorption hybrid system for production of Hz from refinery waste gas.


M.B. Rae and S. Sircar / J. Membrane Sci. 85 (1993) 253-264 , CARBON HIGH PRESSURE LOW PRESSURE









Fig. 7. Schematic of plate-and-frame membrane module for continuous testing.

ule unit for continuous testing of membrane performance for separation of a multicomponent Hz-hydrocarbon mixture. Figure 7 shows a diagram of the module unit. It consisted of six flat sheets of the five-coated membrane connected in series. The purge gas was passed through the module in a countercurrent direction to the feed gas flow. The module was instrumented to measure all inlet and outlet flow rates and gas compositions. A gas mixture containing40.9% Hz, 20.2% CH4, 19.8% C2Hs, 9.2% C3Hs and 9.9% &Hi, (mol%) at 4.4 atm was used as the feed gas. This is a typical composition in a refinery waste stream. The module was operated at - 11.0 ’ C using a purge gas consisting of Hz, CH, and CzHGmixtures (typical PSA waste of Fig. 6). It was found that the membrane could be used to produce a Hz-enriched gas containing 56.0% Hz while rejecting 100.0% C4H10,92.0% C&He,67.5% C&H6and 36.0% CH,. The Hz recovery in the membrane was 63.0%. The Hz-rich gas was produced at - 4.3 atm from the membrane unit. Although the module was run at - 11“C, the mixture selectivity data reported in Table 3 shows that the above-described separation can also be run efficiently at ambient temperature.

The hydrocarbon-rich membrane reject gas could be used as fuel. The recovered Hz-rich gas from the membrane could be further compressed to a pressure of 18.0 atm and fed to a conventional PSA system in order to produce a 99.99 + % Hz product with an overall H, recovery of -43.0% from the waste feed gas. Thus, the nanoporous membrane can be used to recover a valuable chemical (H,) from a waste gas which may not, otherwise, be practical. References S.L. Matson, J. Lopez and J.A. Quinn, Separation of gases with synthetic membranes, Chem. Eng. Sci., 38 (1983) 503. R.E. Kesting, Synthetic Polymeric Membranes, A Structural Perspective, 2nd ed., John Wiley and Sons, New York, 1985, pp. 1-81. J.D. Way and R.D. Noble, Facilitated transport, in: W. Ho (Ed.), Membrane Handbook, Van Nostrand Reinhold, New York, 1991, pp. 833-865. R.J.R. Ulhom and A.J. Burggraaf, Gas separation with inorganic membranes, in: R.R. Bhave (Ed.), Inorganic Membranes: Synthesis, Characterization, and Application, Van Nostrand Reinhold, New York, 1991, pp. 155-176. H.P. Hsieh, Inorganic membranes, AIChE Symp. Ser., 84 (1988) 1.

M.B. Rao and S. Sircar /J. Membrane

264 6

7 8





J.E. Koresh and A. Soffer, Molecular sieve carbon permselective membrane: Part 1. Presentation of a new device for gas mixture separation, Sep. Sci. Technol., 18 (1983) 723. J. Sakata and M. Yamamoto, Apparatus for separating condensable gas, U.S. Patent 4,583,996 (1986). R. Ash, R.M. Barrer and C.G. Pope, Flow of adsorbable gases and vapours in microporous medium: binary mixtures, Proc. R. Sot. London, Ser. A, 271 (1963) 19. R.M. Barrer and E. Strachan, Sorption and diffusion in microporous carbon cylinders, Proc. R. Sot. London, Ser. A, 231 (1955) 52. R. Ash, R.M. Barrer and C.G. Pope, Flow of adsorbable gases and vapours in a microporous medium: single sorbates, Proc. R. Sot. London, Ser. A, 271 (1963) 1. R. Ash, R.M. Barrer and R.T. Lowson, Transport of single gases and binary gas mixtures in a microporous carbon membrane, J. Chem. Sot., Faraday Trans. 1, 69 (1973) 266. A. Yamasaki and H. Inoue, Surface diffusion of organic vapor mixtures through porous glass, J. Membrane Sci., 59 (1991) 233.





17 18

Sci. 85 (1993) 253-264

R.J.R. Ulhorn, M.H.B.J. Huis In’t Veld, K. Keizer and A.J. Burggraaf, High permselectivities of microporous silica-modified y-alumina membranes, J. Mater. Sci. Lett., 8 (1989) 1135. K. Keizer, R.J.R. Ulhorn, R.J. Van Vuren and A.J. Burggraaf, Gas separation mechanisms in microporous modified y-A&O, membranes, J. Membrane Sci., 39 (1988) 25. E.R. Geus, M.J. den Exter and H. van Bekkum, Synthesis and characterization of zeolite (MFI) membranes on porous ceramic supports, J. Chem. Sot., Faraday Trans., 88 (1992) 301. M.B. Rao, S. Sircar and T.C. Golden, Gas separation by adsorbent membranes, U.S. Patent 5,104,425 (1992). S. Sircar and M.B. Rao, Estimation of surface diffusion through porous media, AIChE J., 36 (1990) 1249. R.M. Barrer, Zeolites and Clay Minerals as Sorbents and Molecular Sieves, Academic Press, New York, 1978, p. 292.