992
Ind. Eng. Chem. Res. 1995,34, 992-996
Adsorbate Shape Selectivity: Separation of the HW134a Azeotrope over Carbogenic Molecular Sieve Allan Hong, Ravindra K. Mariwala, Michael S. Kane, and Henry C. Foley* Center for Catalytic Science and Technology, Department of Chemical Engineering, University of Delaware, Newark, Delaware 19716
Experimental evidence is provided for adsorptive shape selectivity in the separation of the azeotrope between H F and 1,1,1,2-tetrafluoroethane(134a) over pyrolyzed poly(furfUry1alcohol)derived carbogenic molecular sieve (PPFA-CMS). The separation can be accomplished over coconut charcoal or Carbosieve G on the basis of the differences in the extent of equilibrium adsorption of H F and 134a. On these adsorbents 134a is more strongly bound than HF, thus it elutes much more slowly from the bed. The heat of adsorption for 134a in the vicinity of 200 "C on Carbosieve G is -8.8 kcaymol. I n contrast, when the same azeotropic mixture is separated over PPFA-CMS prepared at 500 "C, 134a is not adsorbed. As a result 134a elutes from the bed first, followed by HF. The reversal is brought about by the narrower pore size and pore size distribution of the PPFA-CMS versus that for Carbosieve G. Thus the separation over PPFACMS is a n example of adsorbate shape selectivity and represents a limiting case of kinetic separation.
Introduction Adsorptive separations in packed beds can be based on the extent of equilibrium achieved either through gas-surface interactions or through the kinetics of adsorption far away from equilibrium. For wide pore carbogenic adsorbents such as activated carbon, the basis for separation of small molecules is thermodynamic since a significant fraction of the pores are in the range of 1.5-4.0 nm. The components of the mixture to be separated diffise intothe activated carbon pores via bulk and Knudsen diffusional transport and then compete for adsorption sites on the surface. The differences in the heats of adsorption of the components give rise to separation when the system reaches a dynamic equilibrium state. In the simplest case carbon provides a highly reticulated surface in a small volume over which the equilibration process can occur. In the case of kinetic separations, "shape selectivity" lies a t the heart of the process. Here the carbon adsorbent is referred to as a molecular sieve because, apart from providing a highly reticulated surface, pores with dimensions similar in size to small molecules (0.40.6nm) are present. In contrast to activated carbons, carbogenic molecular sieves (CMS) can provide shapeselective effects for separation. In the extreme, if the pore size is much smaller than the kinetic diameter of one component in a binary mixture but larger than that of the other component, then the former may be excluded completely from the pore while the latter is not. However, molecules more similar in size also can be separated, if the pore size is sufficiently small to produce differential rates of transport for each species through the CMS network. Small differences in kinetic diameters between two components can be amplified significantly by the entropic barrier presented by either a pore or pore mouth similar in dimension to the dimensions of the molecules. The alternative methods for adsorptive separation of air are classic examples of equilibrium versus kinetic mechanisms (Yang, 1987;Ruthven, 1984;Sircar, 1988). For oxygen recovery by pressure swing adsorption, air
* Author to whom the correspondence should be addressed.
is passed through a packed bed of zeolite, where nitrogen, due to its high quadruple moment, experiences a strong potential interaction with the zeolite surface (Haas et al., 1988;Faroog and Ruthven, 1991). This gives rise to a larger adsorptive equilibrium constant for nitrogen compared to oxygen. As a result oxygen is enriched in the bulk gas as the mass transfer front translates through the bed. To reverse this result, and to recover nitrogen as the primary product of air separation, the mechanism of separation must be significantly altered. First, the polar surface of the zeolite must be replaced with the nonpolar surface of a carbon. This is not sufficient since at equilibrium the extents of adsorption of oxygen and nitrogen are nearly identical (Pilarczyk et al., 1987;Seeman et al., 1988;Juntgen et al., 1981). Hence to recover nitrogen from air the separation must be carried out well away from equilibrium over a carbogenic molecular sieve (CMS), specifically prepared with a pore mouth dimension of 4 h;. The small difference between the molecular dimensions of nitrogen and oxygen, on the order of 0.02 nm, is amplified by the entropic barrier presented by the 4h; CMS pore mouth and results in significantly different kinetics of adsorption (Dominguez et al., 1988;LaCava et al., 1989;Ruthven et al., 1986). This is manifested in a diffusion coefficient for oxygen in CMS which is lo1-lo2 higher than that for nitrogen. As a result the separation can be carried out effectively on the basis of kinetics alone, even though at equilibrium no separation would be achieved. Relatively little has been done to explore the generality of this 2-fold approach to adsorptive separations. Given its potential as a useful alternative to traditional, equilibrium-driven adsorptive separation, further investigation of kinetic separation over carbon adsorbents is justified. To be successful the pore structure of adsorptive carbon must be controlled to match the application. In contrast to zeolites, wherein control of pore size requires a change in the nature of the exchange cation, if not the framework, with carbon the pore structure can be modified either by selective deposition of carbon at the pore mouths of an otherwise non-molecular sieve carbon (Braymer et al., 1994; Cabrera et al., 1994)or, alternatively, by careful control
0 1995 American Chemical Society Q888-5885/95/2634-Q992$09.QQ/Q
Ind. Eng. Chem. Res., Vol. 34, No. 3, 1995 993 Table 1. Properties of Carbogenic Adsorbents Examined for HF/134a Azeotrope Separation' ~~
properties
coconut charcoal
pore volume, cm3/g BET surface area, m21g pore mode, nm particle size, mesh elemental composition, %
0.35 1100 0.65 50 x 250 C = 92.2,O = 7.8 trace metals coconut shells 9 R x 1/4 in.
precursor column dimensions
Carbosieve G
PFA-CMS
0.47 1100 0.56 80 x 100
-0.18 -300 -0.5 80 x 100
silicon WC/PVDC polymer 9 R x 114 in.
poly(furfury1 alcohol) 3 R x 118 in.
-
c = 87.8,O= 11.1
The PFA-CMS was synthesized in our laboratory. Mariwala and Foley (1994a,b) described the synthesis procedure in detail.
of the high temperature stages of pore development (Mariwala and Foley, 1994;Lafyatis et al., 1991). Using the latter method, it has been shown that small molecule diffisivities can be made to vary by more than 2-4 orders of magnitude depending upon the thermal history of the carbon sieve derived from poly(furfury1 alcohol). In this note both thermodynamic and kinetic adsorptive separations of an azeotrope consisting of 97% 1,1,1,2-tetrafluoroethane(134a) and 3%hydrogen fluoride (HF) are examined. This separation is significant because 134a is a replacement for the ozone-depleting Freon 12 refrigerant, and HF must be reduced to less than 50 ppb in the final product. Adsorption offers an alternative to more classical separations, namely cryogenic "cracking" or extractive distillation. The azeotrope can be separated adsorptively by either mechanism depending upon the choice of carbon adsorbent employed. Using commercially available Carbosieve G and coconut charcoal, the separation is equilibrium driven. In contrast, when a carbogenic molecular sieve, derived from poly(furfury1alcohol) by pyrolysis at 500 "C, is used a shape-selective or kinetic separation results. This difference in the mechanism of separation reverses the order of elution of the two components from the carbon bed. These results are placed in the context of the process alternatives they potentially represent.
Experimental Section The adsorptive separation experiments on carbogenic adsorbents were carried out on a Gow-Mac gas chromatograph (Model No. 69-625P) equipped with a Valco sampling valve (Model No. A2C6UWTHC), a gas density detector, and a Spectra Physics integrator (Model No. SP4270). Pulses of the azeotropic mixture were admitted to the bed of carbon adsorbent in flowing helium by means of an automated gas sampling valve. An MDA toxic gas detector, Model No. TLD-1 equipped with Chemcassette No. 705505, was used to monitor gas phase HF level exiting the gas chromatograph. Industrial grade helium and argon gases were supplied by Keen Gas Co. Pure 134a gas and 3%HF/134a azeotropic mixture were provided by DuPont Co. Carbosieve G was obtained from Supelco and was used without further modification. Coconut charcoal was utilized as received from Aldrich Chemical. The poly(furfury1 alcohol)-derivedCMS was prepared at 500 "C according to previously published methods (Mariwala and Foley, 1994a). To demonstrate the two different separation mechanisms, stainless steel columns (9 R. x 114 in. 0.d. and 3 ft. x 118 in. o.d.1 packed with desired adsorbents were used. Stainless steel wool was used to fix the ends of the adsorbent bed in the column. Prior to separation, the packed column was treated inside the gas chromatograph a t 523 K for 16 h with flowing helium at 40 cm3/min. During the separation, the column tempera-
3 ft PFA-CMScolumn.
n
Carrier Gas Flow rate175 cc/min Temperature 473 K 97 % 1,1,1,2-tetrafluoroethane RTm0.3 min
3 % HF RT=W min
97 % 1,1,1,2-tekafIuoroethme RT=12min. 9 ft Cubwieve G column. Carrier Gaa Flow rate175 cc/min Temperature 473 K
1
-
A
I\ 12
Retention T h e minutes
Figure 1. Chromatographic separation of 3% HF/l,l,l,a-tetrafluoroethane over columns packed with different carbogenic adsorbents. The column and GC specifications are provided on the figure.
ture was held fixed at 473 K, the carrier gas flow rate was 75 cm3/min, the detector temperature was 393 K, and the detector current was 200 mA. The column temperature was held at 473 K to ensure the HF was fully dissociated to its monomer unit and to provide well-defined 134a cchromatographic peaks during elution from Carbosieve G. At the gas density detector, the sampling side carrier gas flow rate was maintained a t 75 cm3/min and the reference side flow rate was maintained at 40 cm3/min. For measurement of the 134a adsorption isotherms on Carbosieve G and coconut charcoal, the sampling loop volume was changed for each data point holding all other parameters the same. Downstream, the exit gases from the sampling valve were flowed into a low concentration NaOH solution with bromothymol blue as the pH indicator.
Results Three different types of carbon-based materials were examined for the separation of HF/134a azeotropic mixture. Table 1 summarizes some of the properties of the carbogenic adsorbents. Figure 1 shows the chromatograms of HF/134a on Carbosieve G and poly(furfuryl alcohol)-derivedCMS (PFA-CMS). With Car-
994 Ind. Eng. Chem. Res., Vol. 34, No. 3, 1995 6,
I
I
I
I
-o
0 0
m
I I
IO
20
30
40
-
50
60
70
80
Pressure torr Figure 2. Adsorption isotherms of 1,1,1,2-tetrafluoroethaneon Carbosieve G and coconut charcoal at 473 K.
bosieve G as the adsorbent, HF elutes first ( t =~0.8 min) a t nearly the superficial velocity and 134a elutes much later ( t =~ 12.5 min). The same behavior is observed with a column packed with coconut charcoal. However, with PFA-CMS as the adsorbent, the order of elution is reversed and 134a elutes at the superficial velocity but HF is retained on the adsorbent for a longer time. In a 12 ft column packed with PFA-CMS, the retention time of 134a increases marginally from 0.3 to 0.9 min due to the increase in column length but the HF retention time increases from 2.2 to 12 min. In all the cases the material balance for eluting HF and 134a could be closed to within 5%. These chromatograms demonstrate that HM134a azeotrope can be separated cleanly over carbogenic adsorbents, and that the HF concentration in the 134a is reduced to less than 50 ppb as the azeotrope is broken by adsorptive separation. To assess the strength of 134a interaction on Carbosieve G and coconut charcoal, the adsorption isotherms of 134a were measured at 473 K according to the method suggested by Paryjczak (1986). Isotherm measurements were assessed by varying the injected volume of 134a. The equations (Paryjczak, 1986) used to calculate the isotherms from these data are as follows:
&=- naSads “qleak
P=
naWchhRT Fcspeak
where Q is the amount adsorbed in moVgm of adsorbent, m’ is the mass of the adsorbent, na is the amount of 134a injected, P is the pressure of the injected gas, Wch is the chart speed, h is the peak height, F, is the carrier gas flow rate, Speakis the area of the peak, S a d 8 is the area eluted between the nonadsorbed reference peak and the adsorbate peak, R is the gas constant, and T is the temperature of the column. Figure 2 shows the 134a isotherm on Carbosieve G and coconut charcoal obtained from the 134a chromatograms and eqs 1 and 2 at 473 K. In the pressure range of 0-80 Torr used for the measurement, the isotherm appears to be mostly linear with small curvature in the low pressure range. The adsorption equilibrium constants, & I s ,calculated from the low pressure range were 2.1 x and 1.4 x moWgTorr) for Carbosieve G and coconut charcoal, respectively. Both the isotherms are nearly identical and have the same shape.
3 0.0016
0.0018
0.002
0.0022
m.UK
Figure 3. van’t Hoff plot for adsorption of l,l,l,2-tetrafluoroethane on Carbosieve G. The heat of adsorption from the slope was found to be 8.8 kcdmol. The carrier gas flow rate was 75 cmVmin.
To assess the heat of adsorption of 134a on Carbosieve G, the temperature dependence of the adsorption constant was determined by varying the temperature of the column and measuring the retention time of 134a. Assuming plug flow through the column, a simple material balance shows that, from the retention time as a function of temperature, the adsorption equilibrium constant can be calculated on the basis of the expression
(3) where 2 is the retention time, L is the length of the column, u is the superficial gas velocity, E is the void fraction of the bed, and KHis the Henry’s law constant. Figure 3 shows the temperature dependence Of KH. The Henry‘s law constant was measured over the temperature range of 473-573 K. From the slope of the line, the heat of adsorption for 134a on Carbosieve G is calculated to be 8.8 f 0.8 kcaVmo1.
Discussion The separations over Carbosieve G and coconut charcoal are very interesting and warrant discussion. Despite apparent differences between coconut charcoal and Carbosieve G, no real differences were observed in the nature of their interactions with HF and 134a. Both carbons provide a clean equilibrium type separation of the azeotropic mixture based on nearly linear and favorable adsorption isotherms. The adsorption isotherms of 134a on both carbons are nearly identical in shape and magnitude over the range of pressures examined. The Henry‘s law constants are also similar. The heat of adsorption of 134a on Carbosieve G is 8.8 & 0.8 kcal/mol, and it is expected to be close t o this value on coconut charcoal. The magnitude of the heat of adsorption is comparable to that reported for ethylene (8.5 kcaVmo1, calculated in the temperature range of 335-580 K) adsorbed on molecular sieve carbon 5A (Chihara et al., 1978). In contrast to the strong adsorption of 134a, HF, present as a monomer in the gas phase under these conditions, elutes at a rate consistent with it being virtually nonadsorbed. For this reason the separation is particularly clean. The separation of the HF and 134a azeotropic mixture is completely different over the PFA-CMS. The reversal in elution order over PFA-CMS with 134a eluting first followed by HF is dramatic. Over this adsorbent 134a is a nonadsorbed species but HF is not. This phenom-
Ind. Eng. Chem. Res., Vol. 34, No. 3, 1995 995 enon arises from the difference in the kinetics of adsorption for the two species over PFA-CMS. More specifically, rather than competing for surface sites a t equilibrium, the PFA-carbon pore structure imposes a limitation on diffusive transport of 134a on the basis of its molecular size (Figure 4). The results presented indicate that the 134a molecular dimension is too large for this molecule to effectively transport into the pore structure of PFA-CMS. In contrast the HF monomer is small enough to enter the structure and to diffuse readily through it. Hence even though adsorption of HF on the carbon surface may not be strong a t 473 K, the majority of the HF molecules diffuse into and through the internal pore structure of the particles whereas the 134a molecules do not. Instead 134a translates through the interparticle voids in the bed and elutes at a time corresponding to a nonadsorbed molecule traveling at superficialvelocity with the bulk gases. In this way the mean residence time for HF molecules on the bed is much longer than that of the 134a, and from this arises the separation. If we consider the molecular dimensions of the two components in the azeotrope, we can see why this mechanism is operative. Using CACHE molecular modeling software, models of 134a and HF were constructed and analyzed. For both molecules modeling was done using PM3 parameters, singlet multiplicity, and the Broyden-Fletcher-Goldfarb-Shanno method for optimizing geometry. The geometry corresponding to the minimum heat of formation for each molecule was used to determine the molecular dimensions. Since electron density falls off exponentially,the “edge”of each molecule was established by truncation a t 0.01 electrod A3. This creates an electron density isosurface in threedimensional space. By rotating this model, maximum dimensions for each molecule could be found and measured. On this basis the following values were obtained: for HF, minimum 2.29 A and maximum 2.80 A;for 134A, minimum 4.14 A and maximum 5.61 A. Not surprisingly the maximum dimension of 134a is nearly twice that of HF, and the minimum dimension is also much larger than that of HF. These differences in molecular dimension provide the potential required for shape-selective adsorptive separations. However, these size differences are necessary but not sufficient criteria for the separation to occur on a shapeselective basis. Sufficient criteria include an optimally narrow pore size in the CMS to reject or a t least retard the rate of transport of the larger component. Pore sizes and pore size distributions of molecular sieving carbons are difficult to obtain, but real progress has been made recently (Moyer et al., 1994). Nitrogen adsorption a t 77 K in such narrow pore materials is too slow to reach equilibrium in a reasonable experimental time. Without equilibrium or a t least pseudoequilibrium isotherm data, a pore size distribution cannot be established reliably. Recently, we have found that this problem can be avoided to a large degree by using CH3C1 adsorption at ambient temperature (Mariwala and Foley, 1994b). Adsorption is fast, an equilibrium is established, and the Horvath-Kawazoe model (or some other suitable model) can be used to derive a pore size distribution which is very useful, especially for comparative purposes. Based on this new method, comparative pore size distributions were developed for PFA-CMS and Carbosieve G (Figure 5). As is evident, although similar, the two pore size distributions are not the same. For
Narrow Pore Carbogenic Adsorbent
Wide Pore Carbogenic Adsorbent
Thermodynamic mode of separation
Kinetic mode of separation
Separation by preferential interaction
Separation by size and rate discrimination
r l
0
Not to .%le
1,1,1,2-tetrafluoroethane
Figure 4. Schematic diagram for the thermodynamicand kinetic modes of separation. 2.50
I Carbosieve G
E
-
- - - - PFA-CMS
P
H
’ 1.50
9 c:
4%
1.00
z 0.50
\
I 0.00 J 0.4
I
I
I
0.5
0.6
0.7
0.8
-
--
----\b e
I
I
!
!
0.9
1
1.1
1.2
1.3
r-nm Figure 5. Pore size distribution in Carbosieve G and PFA-CMS. The pore size distribution was calculated from methyl chloride adsorption isotherm and Horvath-Kawazoe model modified for methyl chloride.
PFA-CMS the pore mode is a t 5 A, while for Carbosieve G it is a t 5.6 A. Furthermore, and perhaps more significantly, the pore size distribution for PFA-CMS is narrower with a lower frequency of pores above 6 A.In contrast, pores above 6 A in size constitute a larger fraction of the pore size distribution of Carbosieve G and as a result the distribution is wider overall and tails to large pores considerably more. On this basis we conclude that the PFA-CMS is largely able to reject the 134a molecule from transport and adsorption into its internal structure while the HF molecule is able to diffuse rapidly within its pore structure. Thus a kinetic shape-selective adsorptive separation results. In contrast, because of its larger pores and higher fraction of larger pores, Carbosieve G adsorbs 134a and with highly favorable energetics (-8.8 kcallmol sy 200 “C). This high heat of adsorption may result from the closeness of fit between the 134a molecular dimensions and the pore sizes of the relevant pores in Carbosieve G. Because of this interaction, the separation also is accomplished in Carbosieve G but now the mechanism is equilibrium driven. Hence, just as in the case of adsorptive separation of air over CMS, subtle microscopic changes in the pore size of the carbon lead to dramatic changes in the macroscopic behavior of the adsorbent. The dichotomy of separation mechanisms, equilibrium, or kinetic, offers practical potential for new process
996 Tnd. Eng. Chem. Res., Vol. 34, No. 3, 1995
applications. To purify an HF-rich stream contaminated by 134a, the equilibrium adsorptive separation would be effective since HF elutes first and would be recovered readily. However, in the case of the HF/134a azeotrope, the 134a, which is in great excess, must be recovered, so the equilibrium adsorptive separation would not be a feasible approach for processing. On the other hand, by controlling the pore structure of the carbon using, for example, the PFA-CMS, a kinetic adsorptive separation can be produced which could lead to efficient recovery of 134a.
Conclusions The binary azeotrope of 3% HF/97% 134a can be broken by adsorptive separation operating on the basis of either adsorption equilibrium or kinetics. Significantly, the mechanism of adsorptive separation can be altered by utilizing a suitably structured carbogenic adsorbent. In principle, the mechanism of separation can be varied to fit the needs of the process by controlling the structure of the adsorptive carbon molecular sieve material. Acknowledgment We thank the National Science Foundation (NSF No. CBT-965714) and the DuPont Co. for financial support of this research.
Nomenclature Q = amount adsorbed, moVg m' = mass of adsorbent, g na = amount of adsorbate injected, mol P = pressure of adsorbate, atm W& = chart speed, cdmin h = peak height, cm F, = carrier gas flow rate, cm3/min Speak = area of peak, cm2 S a & = area elution between reference and adsorbate peaks, cm2 R = ideal gas constant, cm3-atd(K.mol) T = absolute temperature, K t = retention time, min v = superficial velocity, cdmin E = void fraction && = adsorption equilibrium constant, moV(gTorr) KH= dimensionless Henry's law constant
drocarbon Deposition with a Single Hydrocarbon. Carbon 1993, 31,969-976. Chihara, K.;Suzuki, M.; Kawazoe, K. Interpretation of Micropore Dfisivities of Gases in Molecular Sieving Carbon. J. Chem. Eng. Jpn. 1978,11,584-587. Dominguez, J. A,; Psaras, D.; LaCawa, A. I. Langmuir Kinetics as an Accurate Simulation of the Rate of Adsorption of Oxygen and Nitrogen Mixtures on Non-Fickian Carbon Molecular Sieves. In AIChE Symposium Series; AIChE: New York, 1988; Vol. 84,pp 73-82. Farooq, S.; Ruthven, D. M. Dynamics of Kinetically Controlled Binary Adsorption in a Fixed Bed. AIChE J. 1991, 37,299301. Haas, 0.W.; Kapoor, A.; Yang, R. T. Confirmation of Heavy RollUp in Diffusion-Limited Fixed Bed Adsorption. AIChE J . 1988, 34,1913-1915. Jiintgen, H.; Knoblauch, K.;Harden, K. Carbon Molecular Sieves: Production from Coal and Application in Gas Separation. Fuel 1981,60, 817-822. LaCava, A. I.; Koss, V. A.; Wickens, D. Non-Fickian Adsorption Rate Behavior of Some Carbon Molecular Sieves. I. SlitPotential Rate Model. Gas Sep. Purif. 1989,3,180-186. Lafyatis, D. S.;Tung, J.; Foley, H. C. Poly(furfury1alcohol)-Derived Carbon Molecular Sieves: Dependence of Adsorptive Properties on Carbonization Temperature, Time, and Poly(ethy1ene glycol) Additives. Ind. End. Chem. Res. 1991,30,865-873. Mariwala, R. K.;Foley, H. C. Evolution of Ultramicroporous Adsorptive Structure in Poly(fix+bryl alcohol)-Derived Carbogenic Molecular Sieves. Ind. Eng. Chem. Res. 1994a,33,607615. Mariwala, R. K.;Foley, H. C. Calculation of Micropore Sizes in Carbogenic Materials from the Methyl Chloride Adsorption Isotherm. Ind. Eng. Chem. Res. 1994b,33,2314-2321. Moyer, J. D.;Gaffhey, T. R.; Armor, J. N.; Coe, C. G. Defining Effective Microporosity in Carbon Molecular Sieves. Microporous Mater. 1994,2,229-236. Paryjczak, T. Gas Chromatography in Adsorption and Catalysis; Ellis Horwood: Chichester, 1986;pp 84-93. Pilarczyk, E.; Knoblauch, K.; Jiintgen, H. Erdgas Aus Biogasen Mittels Druckwechseltechnik. (Natural Gas from Biogases by Means of the Pressure Swing Technique.) GWF, Gas Wasserfach: Gas1Erdgas 1967,128,340-345. Ruthven, D. M. Principles ofAdsorption and Adsorption Processes; John Wiley & Sons: New York, 1984;pp 167-173. Ruthven, D. M.; Raghavan, N. S.; Hassan, M. M. Adsorption and Diffusion of Nitrogen and Oxygen in a Carbon Molecular Sieve. Chem. Eng. Sci. 1986,41,1325-1332. Seemann, A.; Richter, E.; Juentgen, H. Modelling of a Pressure Swing Adsorption for Oxygen Enrichment with Carbon Molecular Sieve. Chem. Eng. Technol. 1988,11, 341-351. Sircar, S. Air Fractionation by Adsorption. Sep. Sci. Technol. 1988, 23, 2379-2396. Yang, R. T. Gas Separation by Adsorption Processes; Butterworths: Boston, 1987.
Received for review September 9, 1994 Accepted December 23, 1994@
Literature Cited Braymer, T. A.; Coe, C. G.; Farris, T. S.; Gaffney, T. R.; Schork, J. M.; Armor, J. N. Granulated Carbon Molecular Sieves. Carbon 1994,32,445-452. Cabrera, A. L.; Zehner, J. E.; Coe, C. G.; Gaffney, T. R.; Armor, J. N. Preparation of Carbon Molecular Sieves, I. Two-step Hy-
IE940536R @
Abstract published in Advance ACS Abstracts, February
1, 1995.
