Journal of Membrane Scrence, 82 (1993) 15-26
15
Elsevrer Science Publishers B V , Amsterdam
Ceramic zeolite composite membranes.
Preparation, characterization and gas permeation Meng-Dong Jia*, Klaus-Viktor Peinemann** and Rolf-Dieter Behling GKSS-Research Center Geesthucht GmbH, Max-Plan&Strasse,
Postfach 1160,2054 Geesthacht (Germany)
(Becemed Apnl29,1992, accepted m revised form February 12,1993)
Abstract Ceramic zeohte composite membranes, comprrsmg a ceramic substrate and a dense pure zeohte thm layer on the surface of the substrate, have been developed by an-sztu synthesis The zeohte thm layer has been shown by X-ray diffractron, SEM and optical microscopy to have a pure zeohte phase, rn which mdlvrdual zeohte crystalhtes have mtergrown m three dimensrons into a polycrystal zeolite thm disc The nitrogen permeabrhty of these composite membranes, after calcmatron, reaches l-4 m3/m2-hr-bar at room temperature. The ideal selectrvrtles are 2 81 for He over N, and 47 7 for N2 over n-butane, which are far beyond the range of Knudsen drffuslon. After improvement of the preparation procedure the membrane even shows a selectrvrty of 6 2 for n-butane over r-butane Combined effects of “shape-selectivity” of the zeohte and caprllary condensation of the non-zeolite pores are considered to be responsible for the high selectrvrtres Key words Keywords. ceranuc membranes, zeolite membranes, srhcahte-1, gas and vapour permeation,
separation, shape-selective
Introduction Ceramic membranes offer several significant advantages over polymeric membranes such as their improved structural, and their chemical and thermal stabilities. Ceramic membranes are the only membranes that can be used when separations have to be conducted at high temperature and in corrosive environments [ 11. Ceramic membranes offer the potential for performing reactions and separations simultaneously in the same vessel. Product yields and se*Present address Center for Separation Using Thm Films, Department of Chemrcal Engmeenng, Umverslty of Colorado, Campus Box #424, Boulder, CO 80309-0424, USA e”ro whom correspondence should be addressed
lectivity may be enhanced by equilibrium displacement obtained by the continuous removal of products and harmful byproducts. Undesirable side reactions may be suppressed by the controlled addition of reactants. Hence, interest m ceramic membranes is increasing and they are expected to play an increasingly important role in separation technology in the near future [ 21. Unlike dense polymeric membranes, all ceramic membranes are porous. Hence they were initially used for microfiltration applications, where pore sizes of l-0.1 pm are required. More recently however ceramic membranes with pore sizes of 50 nm have become commercially available and membranes with pore sizes of 20 A are beginning to be pro-
0376-7388/93/$06 00 0 1993 Elsevler Science Publishers B V All rights reserved
16
duced on a pilot scale. Whilst it is a significant technological challenge to produce ceramic membranes with uniform pores in the lo-30 A range, it is extremely difficult to produce them with pores of molecular dimensions [ 31 which are required for gas separation and pervaporation. Hence ceramic membranes have not yet been applied to these areas. It has been suggested that the combination of zeolite molecular sieves that have a high porosity (about 30% of the total volume) and well defined pore sizes in the range of 3-12 A with ceramic membranes would be suitable for use in gas separation and pervaporation [4,5]. However, the problem remains as to how to produce uniform, defect-free zeolite membranes. In this paper we describe a successful way of preparing ceramic zeolite composite membranes, their characterisation and the unique gas and vapour permeation properties of these membranes. Experimental Preparatwn of the ceramx zeohte composite membranes In-situ synthesis Ceramic discs, with a diameter of 50 mm and thickness of 3 mm or 5 mm, are commercial products of type PSO from KPM-Kiinigliche Porzellan-Manufactur Berlin GmbH, Germany. The discs were polished with fine sandpaper before being used. A gel for synthesis of zeolite silicalite-1 was prepared according to the original patent by Grose and Flanigen [ 61. The gel was composed of silica (Aerosil 130 from Degussa), sodium hydroxide (P.A. ), tetrapropylammonium bromide (TPABr) (P.A. ) and deionized water. A typical molar composition of such a gel is given as follows:
M-D
Jia et al /J Membrane Scr 82 (1993) 15-26
1.5 NaOH-9 0 S102*1.0 TPABr.70 H,O After thorough mixing, the gel was transferred to a Teflon vessel of an autoclave. The ceramic discs were then immersed in the gel. Synthesis was carried out in the autoclave at 180°C under autogenous pressure, with a synthesis time between 3-72 hr. After the synthesis the ceramic discs were taken out, washed thoroughly, dried at 120’ C overnight and subjected to gas permeation test. If the non-calcined membrane had a N, permeability higher than 0.005 m3/m2-hr-bar, the synthesis procedure was repeated. Calcuaution of the ceramic zeolite composste membranes The organic amine (TPABr) in the synthesis gel serves as a template for zeolite growth and remains in the channels of the zeolite at the end of the synthesis. In order to free the zeolite channels the amine must be removed by calcination. In our preparation, calcination was conducted at 400 or 450°C for 8 hr. Charactertsatwn of the ceramzc zeolite composrte membranes Optical microscopy (Olympus B061) , scanning electron microscopy (SEM) (JEOL JSM35C) and X-ray spectroscopy (Philips PW1732/10) were used to characterise the membranes. The micrographs revealed the nature of the surfaces of the membranes while Xray spectroscopy was used to determine whether the synthesised layer of the composite membrane was pure silicalite zeolite or not. Gas permea fan experiments Gas permeabihties of the ceramic zeolite composite membranes were tested usmg two different set-ups. For preliminary measurements and for measurements at high feed pres-
M-D
17
Jta et al /J Membrane Scr 82 (1993) 15-26
sures (greater than 2 bar) a bubble gas permeation test apparatus was used. For measurements at relatively low feed pressures (from 50 to 1800 mbar ) and especially for measurements with organic vapours, an automatic gas permeation test apparatus developed at GKSS was used. Figure 1 shows a schematic diagram of the automatic permeation apparatus. It was possible to perform automated measurements for up to ten different gases or organic vapours sequentially due to the apparatus being computer driven. Each measurement could be conducted at any feed pressure between lo-2000 mbar and could be repeated if necessary. Consequently, it was possible to obtain reliable measurements for the permeability of organic fluids having very low partial pressures. With the bubble test cell, gas volume flux (p) was recorded, and the pressure normalized flux
(P/L) was calculated according to the volume flux (p), membrane area (A) and the pressure difference (Ap) across the membrane: (P/L) =p/A Ap (m3/m2-hr-bar) The experiments were performed at room temperature. Measurements with each gas were repeated at least 5 times after constant flux had been recorded. With the automatic apparatus, the pressure increase in the permeate side (from l-3 mbar ) is recorded by a pressure sensor. Gas permeability was calculated by the computer according to:
(P/L) =
Vx 22.41 x 3600
R TA t
In (::;p,))
where A is membrane area (m” ) , (P/L) is presflux ( m3/m2-hr-bar ) = sure normalized 1.24E - 09 kmol/m2-set-Pa, V is permeate vol-
B
Pressure Sensor Permeate Volume N2
-
He
_
9
_
9
_
Computer
Vacuum Pump
Feed Volume
Tumostated Contamer 0°C - 60” C Fq 1 Schematic diagram of the automatic gas permeation apparatus
18
ume (m3), R is 0.08314 (m3-bar-kmol-‘-K-l), 2’ is temperature (K ) , t is time (set ) , Pf is feed pressure, POis permeate pressure at t = 0, Pp ( t ) is permeate pressure at t = t. The measurements were made at 20°C. Each gas was measured 20 times, with 5 min evacuation between two pressure increase measurements. When switched from one gas to another, both sides of the membrane were evacuated for 10 min, to get rid of any possibly adsorbed molecules. The ideal selectivity, cy,is defined as the pure gas pressure-normalized flux ratio of gases Eand j a= (PIU,I(PIU, Results and discussions
Ceramzc zeolite composrte membranes Our ceramic zeolite composite membrane comprises a layer of pure zeolite grown on the surface of a ceramic substrate. The preparation of the membrane is based on the principle that via crystal growth under hydrothermal conditions, individual zeolite crystals may form po-
M -D J&aet al/J
Membrane Scr 82 (1993) 15-26
lycrystalline clusters that ultimately form a thin dense layer on the ceramic substrate. If crystal intergrowth is perfect, the zeolite layer should comprise a single crystal that is defect-free and contains only the zeolite channels. Figure 2 shows a schematic representation of this idea. Zeolite sihcalite-1 has two types of channels. One type is straight with an elliptical openmg of 5.2 A x 5.8 A; the other is zig-zag with an elliptical opening of 5.4 AX 5.6 A. In the ideal case, a silicalite composite membrane should have pores with such dimensions only. Figure 3 shows a SEM micrograph of synthesised zeolite silica&e. crystals. It can be clearly seen that some of the particles are polycrystals composed of two or three single crystals, indicating the potential for polycrystal formation. Figure 4(a) is an optical micrograph showing a membrane surface after an m-situ synthesis. A piece of hair with a diameter of 75 pm serves as a reference enabling an estimation of the size of the zeolite particles. Zeolite crystals with sizes in the range of 100-150~ are clearly seen on the substrate. In a control test performed in the absence of the ceramic substrate, the maximum zeolite crystal size found was only 5 p diameter. Figure 4 (b) is an optical micro-
Fig 2 Schematlc repreeentatlon of the formatlon of the ceramic zeohte composite membrane
M -D Jza et al fJ Membrane Scl 82 (1993) 15-26
Fig 3 SEM photograph of duxhte-1
19
crystals
graph showing a membrane surface after zn-sztu synthesis and calcination at 400’ C. Cracks can be clearly seen. The cracks indicate, on one hand, that the preparation was not successful; on the other hand, that a new thin layer was created on the surface. Improvements to both the synthesis and calcination conditions overcame these problems, as shown in Fig. 4 (c). This shows a membrane surface after m-situ synthesis and calcination at 400°C. No cracks were detected by optical microscopy. A drop of water that was mtentionally placed on the membrane surface remained intact, rather than being immediately soaked in as would have occurred with a porous substrate, indicating a significant change in the nature of the surface. Scanning electron microscopy was used to gain a greater insight into the nature of the novel layer. Figures 5 (a) and (b ) are SEM micrographs of the same membrane cross-section at two different magnifications. Figure 5 (a) shows an overview of the membrane cross-section. The lower part shows the porous support membrane, while a smooth, dense layer, ca. 5 pm thick, is seen on the top surface. Figure 5 (b) is a view of the cross-section of the top layer at higher magnification. Even at this magnifica-
Fig 4 Optical mlcroscopuzal photographs showmg the surfaces of the ceramic zeohte composde membranes
tlon no individual zeolite particles are visible nor any micropores or interstices inbetween zeolite particles. All that is seen is an integral, dense layer consisting of fused zeolite crystals. The zeolite silicalite, has a typical morphol-
M -D Jga et al/J Membrane Scr 82 (1993) 15-26
A
40
30
20
10
5
Fig 6 X-ray *action c-------1
1p.m
Fig 5 SEM photographs of one membrane profile v&h dflerent magnlficatlons (a) 1200X, (b) 12000X
ogy referred to as “crossed twin” [6]. Under the synthesis conditions used in this study, the crystals produced are uniform and spherical in size and shape, as shown in Fig. 2. It was impossible to change the shape of the crystals by pressing them together so as to eliminate the voids around the crystals. Voids so created would be readily observed by SEM. The observation that no individual zeolite crystals are seen and that no pores are visible in the top layer, which appears to be a uniform dense layer, strongly indicates that the layer on the top of the porous substrate is composed purely
pattern of (A) the ceramic substrate and (B) a ceramic zeohte composite membrane
of a layer of highly intergrown zeolite polycrystals. Confirmation that the layer was pure silicalite was made by X-ray diffraction analysis (Fig. 6). Subtraction of (a), which is the pattern obtained from the ceramic substrate, from that of (b), which is the pattern obtained from the composite, results in the diffraction pattern of pure zeolite silicalite-1. From the evidence shown in Figs. 5 and 6, we conclude that it is possible to form a composite zeolite membrane composed of a pure, thin zeolite layer formed on top of a porous substrate via m-srtu synthesis The next stage of our research was to study
M -D Jla et al/J
21
Membrane Scr 82 (1993) 15-26
the gas permeation membranes.
behaviour
of
these
Gas permeation The transport of gases through microporous membranes is generally classified into: (1) Poiseuille (viscous) flow, (2) Knudsen flow and, (3) surface diffusion on the pore walls. In reality, transport can occur by various combinations of these mechanisms. When the mean-free-path of the gas is much smaller than the pore diameter, gas transport will occur predominantly by Poiseuille flow, which is essentially non-selective. When the mean-free-path of the gas is much larger than the pore diameter and the gas is very weakly adsorbed on the microporous medium, gas transport will occur predominantly by Knudsen flow. For example at atmospheric pressure and room temperature, Knudsen diffusion is predominant in pores with diameter below 50 nm. Under comparable conditions the rates of permeation of different gases will then be inversely proportional to the square root of their molecular weights, i.e. light gases will permeate faster than heavier gases. Gas separation obtained by Knudsen mechanism is small compared to that obtained with nonporous polymer membranes, unless the differences in the molecular weight of the components of a gas mixture are large. Since most types of microporous membranes have a pore-size distribution, gas transport through such membranes may occur by combined Knudsen and Poiseuille flow. Adsorption of a component of a gas mixture on the pore walls results in its transport through the pore also by surface diffusion. The driving force in this case is the concentration gradient of the component in the adsorbed phase. At sufficiently high relative pressures and low temperatures, the rate of permeation by surface diffusion can significantly exceed that by
Knudsen diffusion: a strongly adsorbed component on the pore wall may permeate faster than a weakly adsorbed component, even though the former may have a higher molecular weight than the latter. Surface diffusion also decreases the effective pore size. In the case of multilayer gas adsorption, when the relative pressure of the strongly adsorbed component of a permeating gas mixture is sufficiently high, that component may condense and completely fill the pores of the membrane. Moreover, if the solubility of the other components in the condensed phase is low, their rate of permeation will be markedly reduced and the membrane selectivity will be greatly enhanced. When the pore diameter is reduced to molecular dimensions, the interactions between the gas molecules and the pore wall will be very strong. Small differences in molecular size and affinity between the molecules and the pore wall will lead to large differences in the sorption and diffusion of the gases through the pores. Such behaviour is well known as the “molecular sieving” or “shape selectivity” effect in zeolite studies. Very high selectivities are achievable via this effect. As previously mentioned in the experimental section, an organic amine (TPABr) was used in the zeolite synthesis to act as a template. As this remained in the pores at the end, it was necessary to remove it to free the channels. This was achieved by calcining the composite membrane at high temperature. First the nitrogen permeability of the freshly synthesised membranes was tested. If it was greater than 0.005 m3/mz-hr-bar, this was taken to indicate the presence of defects in the membrane. In the case the synthesis procedure was repeated until the membrane had a lower permeability. Table 1 presents the permeabilities of the ceramic substrate to several gases and their ideal selectivities compared with nitrogen. For comparison, calculated selectivities for Knudsen
Membrane Set 82 (1993) 15-26
M -D J&aet al/J
22 TABLE 1
Permeabditles and selectwltles of the ceramic substrate for various gases at 20°C (IM03 )”
Permeability (m3/m2-hr-bar)
Nz
02
He
HZ
CH,
CO2
n-GH,o
1-GHw
17 2
145
30 1
47 9
239
15 5
229
23 2
Ideal selectwlty =px/pN,
l/l
19
175
Selectivity of Knudsen dlffuslon
l/l
14
2 65
%/N,
2 79
3 74
139
l/l
11
132
l/l 25
133
135
l/l 44
l/l 44
“Feed pressure 1 10 bar, permeate pressure atmosphenc
diffusion are also given. Since the selectivity of hydrogen over nitrogen is only about 75% of that calculated for Knudsen diffusion, the transport through the substrate is believed to be a combination of Poiseuille and Knudsen diffusion. It should be noted that the substrate had slightly higher permeabilities for carbon dioxide, n-butane and i-butane. Table 2 presents the permeabilities of one ceramic zeolite composite membrane after calcination at 400°C. The permeabilities are generally reduced, whilst the selectivities of Hz and He over N2 are increased. Surprisingly, the permeability of n-butane is significantly reduced, resulting in a selectivity of 47.7 for N, over n-butane. The permeability of i-butane is also reduced, but not as much as of n-butane. All these changes can be attributed to the presTABLE 2 Permeabllltlesand the ideal selectlwtles of a ceramic zeohte composite membrane(IM28-1) for some gasesat 20°C N,
H,
He
n-&H,,,
l-C&H,,
Permeablhty (m3/m2-hr-bar)
0612 187 14
00128
0 20
Idealselectlwty (YX,NI= PxfPN%
10
l/477
l/3 06
306 229
‘Feed pressure2 0 bar,permeatepressureatmosphencpressure
ence of the zeolite top layer. For simplicity, we assume that the resistance to gas and vapor transport through the whole membrane is solely determined by the thin zeolite layer and not by the substrate. Based on this assumption, all experimentally determined permeation properties are attributed to the thin zeolite layer. To rule out any possible effects due to vapour condensation at high feed pressure, permeation measurements were subsequently performed on our automatic permeation apparatus at a feed pressure of 100 mbar. Table 3 presents the results obtained for three membranes. These membranes show a high permeability for N, and in the case of membranes IM59-1 and IM71-1 the He/N,-selectivity is greater than 2.65, which is the Knudsen selectivity. For all the membranes the permeabilities decrease with increasing carbon number in the series methane, ethane, propane and n-butane. The selectivities of Nz over n-butane are not as high as that achieved at higher feed pressures (shown in Table 2), but they are higher than that calculated for Knudsen diffusion. The permeabilities of i-butane and cyclohexane were also measured. In view of their larger kinetic diameters [7] compared with those of the other molecules, it was expected that their permeabilities would be lower provided that “shape selectivity” was the predom-
M -D Jla et al/J
Membrane Scr 82 (1993) 15-26
23
TABLE 3 Permeabihtles and selectlvltles of the ceramic zeohte composite membranes, calcmed at 400 ’ C, 8 hr, for various gases, 20 ’ C
(A)
IM58-1 IM60-1 IM71-1 (B)
e
IM58-1 IM60-1 IM71-1
Permeability
(m3/m2-hr-bar)
Nz
He
CO2
CH4
CzH6
C3HB
n-CdH,o
I-CIH,,
Cycle-c,
1 15 2 89 3 85
3 23 692 10 3
1 10 2 65 2 90
148 424 4 75
0 81 2 95
0 19 0444 1 15
0 069 0 383 0 838
0 60 130
0 21 0 905 148
Ideal selectwlty CY~/N. = PJP,, X=He
CO2
CHI
CzHe
C3I-b
n-W-b
~-W-L,
Cycle-c,
2 81 2 39 2 68
l/l l/l l/l
129 146 123
l/l l/l
l/6 05 l/6 51 l/3 35
l/167 l/7 55 l/4 59
l/4 82 l/2 96
l/5 48 l/3 19 l/2 60
05 09 33
42 31
“Feed pressure 100 mbar, permeate pressure l-2 mbar
inant transport mechanism. However, this was not the case. Both the permeabilities of i-butane and cyclohexane are higher than those of propane and n-butane, implying complicated transport mechanisms. Considering that fine cracks might have been formed during calcination, a thin (1 pm) layer of silicone rubber was used to coat and seal the surface of two of the membranes. PermeabiliTABLE 4 Kinetic diameters of various molecules, Lennard-Jones relationship [ 71
based on the
Molecule
Kmetlc diameter (A,
Molecule
Kmetlc diameter (A,
He HZ 0, NZ NO co CG! Hz0 NH, CH,
26 2 89 3 46 3 64 3 17 3 76 33 2 65 26 38
GH, GH,
33 39 43 43 4 23 50 55 62 102 5 85 60
C3H3
n-C,Hlo Cyclopropane GHw SF, Neopentane (&F&N Benzene Cyclohexane
ties were subsequently measured and are presented in Table 5. Membrane IM61-11 still exhibits high permeabilities for NP, He, CO2 and CH4, but low permeabilities for propane and nbutane. Selectivity of N2 over propane therefore increases to 30.6 and selectivity of N2 over n-butane to 144. The effect of silicone coating on membrane IM71-1 designed IM71-11 is not as strong as with membrane IM61-11. It can still be seen that the selectivities for N, over nbutane and i-butane are more enhanced than the selectivities of He, CH, and CO2 over NP. It is well known that silicone rubber is very permeable to hydrocarbons due to their high solubility m the polymer. The present results can only be explained by the assumption that the polymer has plugged the defects and thus prevented direct permeation of the gas or vapour through the defects, while the polymer itself does not contribute to the selectivity. To check the influence of feed pressure on the permeability, the variation of the permeability of various gases through membrane IM6111 was measured as a function of pressure. Figure 7 presents the correlation between the permeabilities of some gases and the mean pressure, which is the sum of the feed and per-
24
M-D
J&aet al /J Membrane SCL 82 (1993) 15-26
TABLE 5 Permeablhtles and selectwitles of the ceramic zeohte composite membranes, coated with a thm layer of s&cone rubber polymer, for vmous gases, 20 oC Permeabdlty (m3/m2-hr-bar)
IM61-1 IM71-11
Nz
He
CO,
CH,
C2HB
C,H,
n-C*H,,
GHH,,
cyclo-cG
101 164
2 60 500
062 157
0 17 2 36
050 150
0 033 0 55
0 0070 0 232
0 324 030
0 040 050
Ideal selectlvlty qNz = PJPNz
IM61-11 IM71-11
X=He
CO2
CHI
C&S
C,H,
n-&HI0
&H,o
Cycle-c,
2 57 2 72
l/l 63 l/l 17
116 129
l/2 02 l/l 23
l/30 6 l/3 35
l/l44 l/7 93
l/3 12 l/6 13
l/25 3 l/3 66
“Feed pressure: 100 mbar, permeate pressure l-2 mbar
meate pressure divided by two [ (P,+P,)/2]. Figure 7 is divided into two parts: data shown on the left hand side where the mean pressure is lower than 1.0, obtained with the automatic
10
1
A.-A
A
a-0
l
+--A-A 01
001
Oool m
A
He
l
N2
A
co2
0
t-butane
V
propane
V
n-butane
Mean Pressure 6
(bar)
Fig 7. Con&&on between the permeability of several gases and the mean pressure
permeation apparatus; and the right hand side where data were obtained for mean pressures between 1.0 and 3.0 using a test cell with bubble meter. Although the permeation experiments were conducted using different apparatus, the data correlate reasonable well. In the total range of mean pressures, permeabilities of He and Nz remain constant, implying the absence of viscous flow. The selectivity of He over Nz lies between 2.5 and 2.7. By contrast, the permeabilities of condensible gases and vapours, such as COa, propane, n-butane and i-butane are strongly pressure dependent. The decrease of the permeabilities of propane and i-butane is more than a factor of ten over the pressure range 0.0549 bar. Within this range of measurement, the permeability of i-butane is higher than those of propane and n-butane. Table 6 presents the permeation properties of two membranes prepared by a modified procedure that did not involve polymer coating. Some striking results are found. First, the selectivity of He over N, is only 1.63, lower than that of the previous membranes. Secondly, the selectivity of CO2 over Nz is greater than 1. Thirdly, the membranes show lower permeability to i-butane than to n-butane, resulting in
M -D Jza et al /J Membrane Set 82 (1993) 15-26
25
TABLE 6 Permeabllltles and ideal selectwtles 20°C Permeablhty
IM84c2 IM6221
of the ceramic zeohte composite membranes, calcmed at 45O”C, 8 hr, for various gases,
(ms/m2-hr-bar)
N2
He
co2
W-b
n-GHlo
1-W-b
neo-C,
n-C&
2 35 102
3 84 166
2 93 115
0 238 -
0 147 0 081
0 068 0 013
0032
0 052 -
Ideal selectlwty (Y,,~~= Px/PNt
IM84c2 IM6221
X=He
(302
163 163
125 123
l/9 87
n-CJL
l-GKo
neo-C,
n-C,
l/16 l/12 6
l/34 6 l/78 5
l/319
l/45 2 -
“Feed pressure 500 mbar, permeate pressure l-2 5 mbar
a selectivity as high as 6.2 for n-butane over ibutane. Such a selectivity is not achievable by Knudsen diffusion. It is unlikely to be due to capillary condensation since no example of isomer separation has been reported. Finally, the permeability of neopentane, which has a kinetic diameter of 6.2 A and is thus larger than the pores of silicalite, is higher than that of ibutane. This clearly indicates that the membranes are not yet defect-free. The transport mechanisms of gases and vapour through our ceramic zeolite composite membranes are complicated. The permeability data presented clearly indicate that the separating layer contains both zeolite channels and larger non-zeolite channels. The ability to separate n-butane and i-butane indicates the presence of the zeolite channels. While the ability of large (i.e. larger than the zeolite pore diameter) molecules to still permeate through, albeit at a low flux, indicates the presence of defects in the membrane structure. The ability to separate n-butane and i-butane to a limited extend indicates that the relative proportion of zeolite pores to defect pores is such that the zeolite pores play a predominating role in de-
termining the permeation mechanism. The observation that the permeabilities of condensible gases and vapours are reduced rather than enhanced indicates to us that the size of the defects in the membrane is possibly less than 2 nm. If the defects were larger the opposite result would have been expected, as demonstrated by Uhlhorn [ 81. The results presented here are only preliminary results in the study of ceramic zeolite composite membranes. More work needs to be done in order to gain a more detailed understanding of the membranes in terms of their structure and transport mechanism. The most important work will be that with binary gas mixtures, as this will enable us to establish the potential for practical separation applications as well as give insight into the transport mechanisms. Conclusions (1) Ceramic zeolite composite membranes, comprising a ceramic substrate and a pure zeolite thin layer on the surface of the substrate, can be prepared by in-situ synthesis.
26
(2) The thin zeolite layer has been confirmed by X-ray diffraction, SEM and optical microscopy to consist of only silicalite crystals that have intergrown into a dense, thin film. (3) The ceramic zeolite composite membranes exhibit high permeabilities to gases after calcinatlon. For instance, permeability to N2 lies between l-4 m3/m2-hr-bar at 20°C. The ideal selectivity, calculated from the ratio of single gas permeabilities, can reach 2.81 for He over N2 and 47.7 for N2 over n-butane. Selectivity of N2 over n-butane increases to 144, when defects in the membrane are blocked with a thin coating of silicone rubber. An improvement to the preparation procedure led to an uncoated membrane demonstrating a selectivity of 6.2 for n-butane over i-butane. (4) The limited permeability of the membranes to molecules which are larger than the zeolite pore diameter indicates that the membranes are not defect-free. Consequently, permeability through the membranes is the result of the combined permeation through both zeolite pores and defects. Acknowledgement The authors would like to thank Gerd Blijcker of GKSS Research Center and Dr. Suzana Per-
M -D Jca et al/J
Membrane SCL 82 (1993) 15-26
eira Nunes, a visiting-scientist on leave from Instituto de Quimica, Universidade E&dual de Campinas, for taking the SEM photographs. We would also like to thank Dr. Tim Naylor from BP Research, Sunbury, for helpful discussions and Marion Aderhold of GKSS for the drawings. References H P Hsleh, Inorganic membrane reactors - A revrew, AIChE Symp Ser ,85 (1989) 268 A Crull, Prospects for the morgamc membrane busrness, Proc 2nd Int Conf on Inorgarnc Membranes ICIM91, July l-4,1991, Montpehier, France, 1991, pp 279-288 R L Goldsmrth, Guest Editorial m the Special Issue on Ceramic Membranes of the Journal of Membrane Scrence, J Membrane Scl ,39 (1988) 197-201 A S Michaels, New vistas for membrane technology, CHEMTECH, March (1989) 162-172 E R Geus, A MuIder, D J Vlschlager, J Schoonman and H van Bekkum, Design of a ceramic zeohte membrane, Proc 2nd Int Conf on Inorgamc Membranes ICIM-91, July l-4,1991, Montpeher, France, 1991, pp 57-63 R W Grose and E.M Flamgen (Union Carbide), Crystalline shca, U S Patent 4,061,724, (1977) L Pa&g, Nature of the Chemical Bond, 3rd edn , Cornell Umv Press, Ithaca, NY, 1960 R J R Ulhorn, Ceramic membranes for gas separation - Syntheses and transport properties, Drssertatron of Umverslty of Twente, Enschede, The Netherlands, Chapters III and VI, 1990