Ind. Eng. Chem. Res. 2002, 41, 3145-3150
3145
MATERIALS AND INTERFACES Intrapore Synthesis of Silicalite Membranes at Temperatures below 100 °C Beom Seok Kang and George R. Gavalas* Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, California 91125
Growth of silicalite membranes inside the pores of R-Al2O3 support tubes was carried out at 95 °C. The support tubes had a 1-µm mean pore diameter and were seeded by impregnation in a suspension of 0.1-µm silicalite particles prior to membrane growth. The membranes were weighed to determine the total reaction product and examined by SEM and EDS to estimate the spatial distribution. Reaction product was found at a depth of 100 µm or more inside the pores of the support. The membranes were tested in single gas permeations of hydrogen, nitrogen, n-butane, and isobutane and in the separation of n-butane/isobutane mixtures. After three growth periods, mixture measurements gave n-butane permeances of 10-17 MPU [1 MPU ) 10-8 mol/(m2 s Pa)] and n-butane/isobutane selectivities of 30-40 at room temperature. The selectivity results are discussed in terms of gaps between the intrapore crystals. Introduction Of the different zeolite membranes, ZSM-5 and silicalite membranes have been the most extensively investigated because of their robust preparation methods and high separation selectivities for various gas and liquid mixtures.1-8 Membrane permeation and transport properties have been the main focus of these studies, but structural properties such as crystal orientation4,9 and intercrystal voids that influence the transport properties10,11 have also been investigated. A structural feature that has not been adequately explored is the effect of zeolite grown inside the pores of the support on the membrane properties. Virtually all membranes grown on alumina supports contain zeolite inside as well as outside the pores. Coronas et al.1 prepared ZSM-5 membranes by two techniques, one resulting in an external layer with very little pore infiltration and the other resulting in a patchy external layer and a deep internal layer. The second type had lower permeances and selectivities for hydrocarbon separations. Other membranes reported in the literature contained a dense, continuous external layer [at scanning electron microscope (SEM) resolution] and an internal layer extending as far as 150 µm from the support surface.6,12,13 The pore volume fraction occupied by the reaction product (crystalline or amorphous) generally declined with distance from the support surface on account of pore diffusion limitations during the hydrothermal reaction. Kusakabe et al.6 observed good selectivities only when the external layer was fully developed and continuous, but this observation does not necessarily imply that separation is localized at the external layer. It might simply signify that, by the time the external layer becomes fully * Corresponding author. E-mail:
[email protected]. Tel.: 626-395-4152. Fax: 626-568-8743. Address: Caltech 21041, Pasadena, CA 91125.
continuous (as observed by SEM), the internal layer is also sufficiently developed to provide the separation function either by itself or in combination with the external layer. Masuda et al.,14 on the other hand, reported that their membranes contained an external ZSM-5 layer only; they did not report any permeation measurements. Internally grown silicalite membranes with high permeances and selectivities for the separation of straight-chain from branched hydrocarbons were prepared by Giroir-Fendler et al.15,16 These authors used asymmetric alumina supports with a γ-Al2O3 layer of 4.5-nm mean pore diameter at the tube i.d. supported on consecutive R-Al2O3 layers with pore diameters increasing toward the tube o.d. The synthesis solution was applied at the o.d., and silicalite growth was a maximum at an intermediate position about 40 µm from the tube i.d. Given the thickness of the internal layer and the gradually declining zeolite occupancy of the pores with distance from the surface, it is possible that only the first few tens of microns from the surface provide useful separation properties. The material deposited deeper into the pores reduces the selectivity as well as the permeance because it adds a less selective resistance in series to the composite membrane. It would be useful, therefore, to examine more closely the properties of membranes intentionally grown within the pores of the support. This paper reports the preparation, characterization, and permeation properties of internally grown membranes. Intrapore growth of zeolite membranes has some practical advantages, the most significant of which is the ability to use supports of relatively large pore size without concern for surface smoothness. Departing from standard practice, the membrane preparations in this study were carried out at temperatures below 100 °C.
10.1021/ie010918e CCC: $22.00 © 2002 American Chemical Society Published on Web 05/08/2002
3146
Ind. Eng. Chem. Res., Vol. 41, No. 13, 2002
Figure 1. Schematic diagram of the permeation cell.
Figure 3. EDS profiles of silicon over the cross section of membranes at different stages of preparation: (a) seeded before growth, (b) after seeding and three growth cycles, and (c) unseeded after three growth cycles.
Figure 2. Scanning electron micrograph of seed particles.
Experimental Section Supports. The supports used in this study were R-Al2O3 tubes (99.5%, Dongsuh Company, Korea) of 1.0µm mean pore diameter, 7.2-mm o.d., 6.0-mm i.d., and ca. 40% porosity. The tubes were cut into 2.8-cm-long segments, ultrasonicated in deionized (DI) water for 30 min, and dried at 200 °C before use. Seeding. Silicalite seed particles of colloidal size were grown using the protocol of Persson et al.17 by heating a mixture of tetraethyl orthosilicalite (TEOS, 98%, Aldrich), tetrapropylammonium hydroxide (TPAOH, 1 M aqueous solution, Aldrich), NaOH (1 M aqueous solution, Aldrich), and DI water at 95 °C for 16-23 h. The molar composition of the mixture was TEOS:0.2 TPAOH:0.08 NaOH:60 H2O. The resulting silicalite particles were separated by filtration, washed with DI water, and dried. Before use, the particles were redispersed in water with the help of ultrasonication, forming a suspension of pH about 8. Seeding Procedure. Seeding of the supports was carried out by immersing the tubes for about 10 min in a suspension of 1 wt % particle content. After withdrawal, the tubes were briefly washed with DI water, dried in air, and heated for 1 h at 200 °C to bond the particles to the support pore surface. Seeding was repeated once to increase the particle loading. A somewhat more favorable spatial distribution of the seeds
Figure 4. Weight gain vs growth cycle.
in the support can be obtained if the suspension influx is restricted to the o.d. or the i.d. side of the tube, but this mode of seeding was not used in this study. Membrane Growth. The growth solution was prepared by stirring at room temperature for 1 day DI water, NaOH, tetrapropylammonium bromide (TPABr, 98%, Aldrich), and TEOS in the molar ratio TEOS:0.1 TPABr:0.1-0.7 NaOH:98 H2O. The seeded tubes were secured vertically in 14-mL Teflon vials (with screw cap) with Teflon holders. A closely fit Teflon rod was inserted into the bore of the tubes so that the membranes could grow only in the pore region adjacent to the outer cylindrical surface. The vials were then filled with 10 mL of growth solution, sealed, and attached to a rotating shaft inside a forced convection oven preheated to 95 °C. The shaft carrying the vials was tumbled at 40-50
Ind. Eng. Chem. Res., Vol. 41, No. 13, 2002 3147
Figure 5. Scanning electron micrographs of a bare support and an intrapore grown silicalite membrane: (a) support surface, (b) support cross section, (c) membrane surface, and (d) membrane cross section.
rpm to enhance mass transfer. After 20 h, the vials were removed from the oven, cooled, and opened, and the membranes were thoroughly washed with water and dried first at room temperature and then at 200 °C. The growth procedure was repeated two or three times before calcination. After completion of growth, calcination was carried out to remove the TPA molecules occluding the zeolitic channels. For this purpose, the membranes were heated in a stream of air first to 350 °C at 10 °C/min and, after that temperature was maintained for 1 h, to 500 °C at 5 °C/min. They were maintained at the final temperature for 3 h and then cooled to ambient temperature by turning off the furnace power. In a few preparations, calcination was carried out after the second growth period and repeated after the final growth period. Membrane Characterization. After each growth and drying period, the membranes were weighed to determine the amount of product deposited. The external o.d. surfaces and cross sections of selected membranes were examined using a scanning electron microscope (SEM, Camscan), and a few cross sections were subjected to energy dispersive analysis (EDS) to trace the product spatial distribution. Nitrogen adsorption (ASAP 2010, Micromeritics) was carried out on selected
membranes at 77 K to obtain information about the mesopore and micropore size distributions. Permeation Measurements. Membrane permeance and separation factor were measured using a permeation cell that can accommodate up to four tubes (Figure 1) and operates at temperatures up to 200 °C. After being mounted on the permeation cell, the membranes were tested with single gases or gas mixtures. In each set of measurements, the feed gas was passed continuously through the shell (space surrounding the tubes), while the permeate gas was collected through the tube bores. Single-gas measurements were carried out using the pressure rise technique in which the feed gas passes continuously through the shell. The permeate side was evacuated and then shut off from vacuum, after which the pressure rise was measured by a sensitive capacitance gauge. The rate of pressure rise gave the membrane permeance by means of a suitable calibration. Mixture permeation properties were obtained using the Wicke-Kallenbach (WK) type of measurements where both feed and permeate sides were kept at 1 bar while the permeate side was continuously swept by a stream of He. The mixture of sweep gas and permeate gas was analyzed downstream using a gas chromatograph (HP 5890) with a packed column (0.19% picric acid on
3148
Ind. Eng. Chem. Res., Vol. 41, No. 13, 2002
Graphpac GC, Alltech) and a thermal conductivity detector. The known sweep gas flow rate and the outlet gas composition measured with the chromatograph were used to calculate the transmembrane flux of each component. Results and Discussion Seeding. Considerable effort was devoted to finding a range of particle sizes suitable for internal seeding of alumina supports of 1.0-µm mean pore diameter. The particles finally selected were those produced by a 16-h reaction at 95 °C, which had about a 0.1-µm diameter, as estimated by SEM. Figure 2 shows particles of that size coated onto an alumina tube of 0.15-µm pore size. After 23 h, the particles had grown to about 0.2 µm, but longer reactions did not result in substantial additional growth, perhaps because of depletion of silica in the solution. Each support tube was seeded twice by impregnation, resulting in an average weight gain of 0.084% ( 0.045%. The large variation was because of the small weight gain and water adsorption during handling and weighing. A suspension volume equal to the pore volume carries particles that make up 0.168 wt % of the support. Hence, about one-half of the particles were not retained during impregnation. Figure 3 shows an EDS trace of silicon on the cross section of a tube that had been seeded twice. The concentration of silicon decreases with distance from the surface, and after 50 µm, it falls below the measurement error. Hence, the seed particles are preferentially retained within the first 50 µm from the external surface. Membrane Characterization. Figure 4 shows the cumulative amount of reaction product for an unseeded tube and a tube seeded by impregnation versus the number of growth cycles at 95 °C. Product formation in the tube seeded by impregnation was rapid in the first and second periods and leveled off thereafter. For the unseeded tube, product formation was low in the first period but accelerated thereafter so that, in effect, the first growth period provided seeding for subsequent growth. A 5% weight gain is equivalent to a 97-µm region of pores fully occupied by silicalite. An 8% weight gain corresponds to 157 µm of pores fully occupied with product. The spatial distribution of reaction product is shown in Figures 3 and 5. Figure 5 shows SEM images of the surfaces and cross sections of an untreated tube and a tube seeded by impregnation and subjected to four growth periods. Reaction product is evident in the support 20-30 µm from the external surface. There is also a 2-3-µm-thick external film, but it does not appear to be fully continuous. A more quantitative view of the product distribution is provided in Figure 3, which shows EDS traces of silicon over the cross sections of a tube seeded by impregnation and an unseeded tube, both after three growth periods. The product concentration in the seeded tube is high near the surface, and after about 20 µm, it drops to a lower level that is maintained until at least 80 µm. Judging from the mass gain shown in Figure 4, the lower level of silicon must extend for several hundred microns from the surface. The product concentration in the unseeded tube remains at a low level throughout the 80-µm region, including the first 5 or 10 µm from the surface. Figure 6 shows the nitrogen adsorption-desorption isotherm of a membrane after three growth periods followed by calcination. The isotherm has two hysteresis loops, the first at P/Po below 0.3 and the second at P/Po
Figure 6. Nitrogen adsorption-desorption isotherm at 77 K of a membrane after three growth cycles and calcination.
Figure 7. Differential pore size distribution of the membrane of Figure 6, calcualted from the desorption isotherm by the BJH procedure.
above 0.5. The first loop is in the microporous region and is very similar to that measured for zeolite powder collected under the same synthesis conditions. Its origin is still under some debate but is of no concern here. The loop at higher P/Po derives from pores above 1.5 nm in diameter. The distribution of these pores calculated using the BJH procedure (based on the desorption branch) and shown in Figure 7 has a peak centered at less than 2 nm, the descending part of which is beyond the applicability of Kelvin’s equation used in the BJH procedure. Figure 7 also shows a smaller peak at about 4 nm. The pores of about 2-nm diameter seem to be nonzeolitic pathways of the type discussed by previous authors.18 Permeation Properties. After the second or third growth period, the uncalcined membranes were impermeable to nitrogen, but after calcination, they became permeable to nitrogen and other gases. In a few samples,
Ind. Eng. Chem. Res., Vol. 41, No. 13, 2002 3149
Figure 8. Single-gas permeance versus temperature for a membrane prepared using three growth cycles. Table 1. Permeation Results of ZSM-5/Silicalite Membranes from the Literature and from the Present Study
membrane/support
SFa
silicalite/stainless steel MFI/Al2O3 silicalite/Al2O3 ZSM-5/Al2O3 silicalite/Al2O3 silicalite/ Al2O3 silicalite/Al2O3
27 5.5-33 1.5-12c 7.7-110 28-53 NA 37 39 32 17
n-butane permeance (MPUb) 1.5 5.0-9.0 30d 2-10 1.6-7.6 75d 9.9 15 17 29
temp (°C) 25 90-140 100 25-200 25 150 25 25 25 150
ref 7 1 6 3 5 15 this work
a Separation factor. b 1 MPU ) 10-8 mol/(m2 s Pa). c Ideal selectivity. d Single-gas measurements.
the second or third growth period was followed by calcination, after which an additional growth period and calcination were carried out. Permeation measurements for single gases conducted by the pressure rise technique are shown in Figure 8. The membrane has a high n-butane permeance and a high n-butane/isobutane ideal selectivity but very low ideal selectivity among hydrogen, nitrogen, and n-butane. The hydrogen-tonitrogen permeance ratio is well below the Knudsen value and rises with temperature, suggesting that the bulk of transport takes place through zeolitic pores or fine nonzeolitic pores, below 2 nm in diameter. A small component of transport could take place by viscous flow through larger defects. Measurements were also carried out by the WK technique for a 50:50 mixture of n-butane and isobutane. Table 1 lists room-temperature results for several membranes as well as results from the literature. The n-butane/isobutane separation factor increases with each additional growth period, and after three or four periods, it reaches 30-40 but remains well below the ideal selectivity calculated from the permeances in Figure 8. A permeance of n-butane in the range 10-17 MPU for the 50:50 mixture is in the high range of previously reported values, but the n-butane/isobutane
Figure 9. Temperature dependence of permeances and separation factor (SF) of a silicalilte membrane prepared using three growth cycles.
separation factor is lower than some of the best values recently reported in the literature. Figure 9 shows the butane separation factor to decline with temperature, which is opposite to the trend observed for the ideal selectivity in Figure 8. We have evidence from membrane modification experiments to be reported in a future communication that the internal membranes contain a small number of gaps in the tens of nanometers range between crystals or between crystals and the pore surface. Because the membranes dried at 200 °C were impermeable before calcination, these gaps were generated during calcination. Their relatively large size but small number suggests that they are associated with the larger pores at the tail end of the pore size distribution. One possible source of these gaps is the dimensional changes and stresses accompanying calcination, as discussed in refs 19 and 20. Another possible source of gaps is intrinsic to the chemistry of zeolite growth, especially at relatively low alkalinity and temperature as in the present experiments. Under most conditions, crystallization takes place within a gel layer formed on the alumina surface (externally and internally) soon after immersion in the synthesis solution.13 Crystal growth subsequently proceeds until the gel is consumed or its composition can no longer support growth. Enrichment of the gel with aluminum leached from the support might inactivate the gel after some length of time. Although additional growth components can be supplied from the bulk of the synthesis mixture, the transport of these components to the gaps might be hindered by the residual inactive gel trapped there. Shrinkage of the residual gel by loss of water and TPA during calcination at ca. 500 °C would tend to generate gaps of low permselectivity. Both possibilities are consistent with our observation that a second or third growth period is more effective in increasing membrane selectivity if preceded by calcination rather than by mere drying. Conclusions Internal seeding of alumina supports of 1-µm mean pore diameter allows for the growth of silicalite layers
3150
Ind. Eng. Chem. Res., Vol. 41, No. 13, 2002
inside the pores of the support at a depth of 100 µm or more below the external surface. Zeolite penetration is even deeper in membranes grown under the same conditions on unseeded supports. Despite the large thickness of the internal zeolite layers, membranes grown on internally seeded supports have relatively high permeances. Their selectivities are low after a single growth period but improve with successive growth periods without a sizable drop of the permeance. After three growth periods, the n-butane-to-isobutane separation factor reaches the 30-40 range. The ability to use supports of relatively large pore diameter is the main advantage of internally grown membranes, but the need for several growth periods is a disadvantage. Calcination of internally grown membranes can be safely conducted at heating rates as high as 5-10 °C/min, in contrast with externally grown membranes, which need rates of 1 °C/min or less to avoid the formation of large cracks. The internally grown membranes reported here might have a distribution of defects, including a small number that are tens of nanometers in size. Acknowledgment This work was funded by the Department of Energy, National Energy Technology Laboratory, Grant DEFG26-00NT40817. Literature Cited (1) Coronas, J.; Falconer, J. L.; Noble, R. D. Characterization and Permeation Properties of ZSM-5 Tubular Membranes. AIChE J. 1997, 43, 1797. (2) Funke, H. H.; Argo, A. M.; Falconer, J. L.; Noble, R. D. Separations of Cyclic, Branched, and Linear Hydrocarbon Mixtures through Silicalite Membranes. Ind. Eng. Chem. Res. 1997, 36, 137. (3) Tuan, V. A.; Falconer, J. L.; Noble, R. D. Alkali-Free ZSM-5 Membranes: Preparation Conditions and Separation Performance. Ind. Eng. Chem. Res. 1999, 38, 3635. (4) Xomeritakis, G.; Gouzinis, A.; Nair, S.; Okubo, T.; He, M.; Overney, R. M.; Tsapatsis, M. Growth, Microstructure, and Permeation Properties of Supported Zeolite (MFI) Films and Membranes Prepared by Secondary Growth. Chem. Eng. Sci. 1999, 54, 3521. (5) Bernal, M. P.; Xomeritakis, G.; Tsapatsis, M. Tubular MFI Zeolite Membranes Made by Secondary (Seeded) Growth. Catal. Today 2001, 67, 101.
(6) Kusakabe, K.; Murata, A.; Kuroda, T.; Morooka, S. Preparation of MFI-Type Membranes and Their Use in Separating n-Butane and i-Butane. J. Chem. Eng. Jpn. 1997, 30, 72. (7) Bakker, W. J. W.; Kapteijn, F.; Poppe, J.; Moulijn, J. A.; Permeation Characteristics of a Metal-Supported Silicalite-1 Zeolite Membrane. J. Membr. Sci. 1996, 117, 57. (8) Lai, R.; Gavalas, G. R. ZSM-5 Membrane Synthesis with Organic-Free Mixtures. Microporous Mesoporous Mater. 2000, 38, 239. (9) Lovallo, M. C.; Tsapatsis, M. Preferentially Oriented Submicron Silicalite Membranes. AIChE J. 1996, 42, 3020. (10) Lin, X.; Falconer, J. L.; Noble, R. D. Parallel Pathways for Transport in ZSM-5 Zeolite Membranes. Chem. Mater. 1998, 10, 3716. (11) Gump, C. J.; Noble, R. D.; Falconer, J. L. Separation of Hexane Isomers through Nonzeolite Pores in ZSM-5 Zeolite Membranes. Ind. Eng. Chem. Res. 1999, 38, 2775. (12) Sano, T.; Hasegawa, Y.; Kiyozumi, Y.; Yanagishita, H.; Kitamoto, D.; Mizukami, F. Potentials of Silicalite Membranes for the Separation of Alcohol/Water Mixtures. Stud. Surf. Sci. Catal. 1994, 84, 1175. (13) Lai, R.; Gavalas, G. R. Surface Seeding in ZSM-5 Membrane Preparation. Ind. Eng. Chem. Res. 1998, 37, 4275. (14) Masuda, T.; Sato, A.; Hara, H.; Kouno, M.; Hashimoto, K.; Preparation of a Dense ZSM-5 Zeolite Film on the Outer Surface of an Alumina Ceramic Filter. Appl. Catal. 1994, 111, 143. (15) Giroir-Fendler, A.; Peureux, J.; Mozzanenga, H.; Dalmon, J.-A. Characterization of a Zeolite Membrane for Catalytic Membrane Reactor Application. Stud. Surf. Sci. Catal. 1994, 101, 127. (16) Ramsay, J.; Giroir-Fendler, A.; Julbe, A.;, Dalmon, J.-A. French Patent 9405562, 1994. (17) Persson, A. E.; Schoeman, B. J.; Sterte, J.; Otterstedt, J. E. The Synthesis of Discrete Colloidal Particles of TPA-Silicalite1. Zeolites 1994, 14, 557. (18) Lin, X.; Falconer, J. L.; Noble, R. D. Parallel Pathways for Transport in ZSM-5 Zeolite Membranes. Chem. Mater. 1998, 10, 3716. (19) Geus, E. R.; van Bekkum, H. Calcination of Large MFItype Single Crystals, Part 2: Crack Formation and Thermomechanical Properties in View of the Preparation of Zeolite Membranes. Zeolites 1995, 15, 333. (20) Dong, J.; Lin, Y. S.; Hu, M. Z•C.; Peascoe, R. A.; Payzant, E. A. Template Removal-Associated Microstructural Development of Porous Ceramic-Supported MFI Zeolite Membranes. Microporous Mesoporous Mater. 2000, 34, 241.
Received for review November 13, 2001 Revised manuscript received March 28, 2002 Accepted April 2, 2002 IE010918E
