ZSM-5 zeolite membranes prepared from a cleartemplate-free solution

Gang Li, Eiichi Kikuchi, Masahiko Matsukata *

Department of Applied Chemistry, Waseda University, 3-4-1 Okubo, Shinjuku-ku, Tokyo 169-8555, Japan

Received 17 October 2002; received in revised form 28 March 2003; accepted 8 April 2003

Abstract

Template-free nanosized ZSM-5 seeds were prepared from commercially available ZSM-5 powder. By use of these

seeds, thin and hydrophilic ZSM-5 zeolite membranes were prepared on the outer surface of a porous a-alumina tube ina clear solution free from organic template. The membranes showed high thermal stability to withstand pretreatment at

400 �C. At 270 �C, the ideal selectivities for H2/i-butane and N2/SF6 were 61 and 20, respectively. The membrane was

effective to separate butane isomers at high temperatures. The maximum of the n/i-butane ideal selectivity was 36 at 300

�C. The separation factors for a 50/50 n/i-butane mixture were as high as 13 in the temperature range from 300 to

400 �C.� 2003 Elsevier Science Inc. All rights reserved.

Keywords: Template-free; Clear solution; ZSM-5 membrane; Thermal stability; Butane isomers

1. Introduction

One of the advantages of inorganic membranes

is that they can be used at high temperatures,

whereas polymer membranes cannot due to their

low thermal stability. Zeolite membranes are ex-

pected to have great potential for separation and

reactor applications at high temperatures due totheir molecular sieving properties. Most of the

research on zeolite membranes has been focused

on MFI zeolite membranes (i.e. silicalite-1 and

ZSM-5) [1].

MFI-type zeolite membranes are generally

synthesized in the presence of organic templates.

These organic templates are incorporated in the

zeolite channels during synthesis. They have to be

removed by calcination after synthesis to open up

the zeolite channels. While MFI-type membranes

showing high performance after calcination at

high temperatures have been reported [2], several

studies have shown that cracks tend to form in the

zeolite membranes during calcination due to thethermal stress, which is caused by the difference in

thermal expansion between zeolite layer and sup-

port and/or by changes in lattice parameters of

zeolite crystals as a consequence of the removal of

templates [3,4]. The presence of cracks in the

membrane will significantly deteriorate the sepa-

ration performance of the membrane, in particular

at high temperatures where differences in adsorp-tion and mobility become less significant for the

gas to be separated.

*Corresponding author. Tel./fax: +81-3-5286-3850.

E-mail address: mmatsu@waseda.jp (M. Matsukata).

1387-1811/03/$ - see front matter � 2003 Elsevier Science Inc. All rights reserved.

doi:10.1016/S1387-1811(03)00380-9

www.elsevier.com/locate/micromeso

Microporous and Mesoporous Materials 60 (2003) 225–235

One way to avoid the calcination step is to

synthesize MFI zeolite membranes without the use

of organic templates. Such attempts have recently

been reported [5–9]. Hedlund et al. [5] reported the

preparation of 1.5 lm thick ZSM-5 membranes onseeded porous a-alumina disks using a synthesissolution free from organic templates. Gas perme-

ation experiments performed in the temperature

range of 25–145 �C after treated at 120 �C for 16 hunder vacuum showed that the permeances were

lower and the separation factors for H2/i-butane

mixtures were higher than those through similar

membranes reported by other groups. In addition,

their membrane could not separate mixtures ofgases larger than methane such as n/i-butane

mixtures. Noack et al. [6] reported similar results

that the separation factors were <2 for two dif-

ferent n/i-butane mixtures (90:10 mol% and 10:90

mol%) in the temperature range of 25–130 �Cusing a similar template-free ZSM-5 membrane as

used by Hedlund et al. [5].

Lai and Gavalas [7] also reported the prepara-tion of ZSM-5 zeolite membranes using a template-

free synthesis gel on seeded asymmetric a-aluminatubes. Gas permeation properties were investigated

in a temperature range of 25–150 �C after the

membrane was treated at 160 �C overnight undervacuum. They reported that the ideal selectivities

of H2 over n-butane were above 104 and those of

O2 over N2 were 9–10 at room temperature. How-ever, their membranes also failed to separate bu-

tane isomers.

In another publication, Pan and Lin [8] re-

ported the preparation of ZSM-5 zeolite mem-

branes on porous a-alumina disks without organictemplate. The membrane was used to separate an

eight-component hydrogen-hydrocarbon (C1–C4)

mixture at 25, 100, and 200 �C after dried in adesiccator at 40 �C for 2 days. As the membranewas impermeable to propane, propylene, n-butane

and i-butane, it also could not separate butane

isomers.

In a more recent publication, Lassinantti et al.

[9] also reported the synthesis of ZSM-5 mem-

branes from a gel free from organic templates on

porous a-alumina disks. They studied the separa-tion of a 50/50 n/i-butane mixture as a function of

temperature. Being different from the studies

mentioned above, a maximum of separation factor

of 16.7 was observed at 220 �C for a 50/50 n/

i-butanemixture. However, operation above 270 �Ccaused the formation of cracks, which severely

deteriorated the performance of the membrane,

and the separation factor decreased to 3.8 at400 �C.The present paper reports on the preparation of

ZSM-5 zeolite membranes with high thermal sta-

bilities and good separation factors for butane

isomers. These membranes were prepared on the

outer surface of a porous a-alumina tube using aclear solution free from organic templates. Being

different from above-mentioned studies, template-free nanosized ZSM-5 crystals prepared from

commercial powder were used as seeds herein. The

membrane was characterized by permeation mea-

surements using single gases and an n/i-butane

mixture as a function of temperature.

2. Experimental section

2.1. Preparation of ZSM-5 zeolite membranes

Porous a-Al2O3 tubes (NGK Insulators Co.,

Japan, length¼ 30 mm, i.d.¼ 10 mm, o.d.¼ 8 mm,average pore diameter in the outer top layer¼ 100nm) were used as supports for zeolite membrane

preparation. The supports were cleaned succes-sively in acetone and deionized water each for 20

min under ultrasonication and dried. Prior to hy-

drothermal treatment, the outer surface of the

support was seeded by means of dip coating with a

colloidal ZSM-5 crystal suspension. The colloidal

suspension was prepared from commercially avail-

able template-free ZSM-5 powder (SiO2/Al2O3¼23.8, Tosoh Co., Japan). A typical experiment wasperformed as follows. ZSM-5 powder weighing 2.5

g was first ground for 2 h in an agate mortar. The

ground powder was then mixed with 50 g of de-

ionized water in a beaker to form a slurry. The

slurry was treated in an ultrasonic bath for 20 min

at room temperature. Finally the slurry was kept

at room temperature for about two weeks. Due to

the gravitational effect, bigger particles were set-tled down to the bottom of the beaker while a

milky suspension dispersed by ultra-small particles

226 G. Li et al. / Microporous and Mesoporous Materials 60 (2003) 225–235

was formed in the upper part of the slurry. This

milky suspension was carefully transferred to

another clean beaker. The concentration of the

crystals in the suspension can be adjusted by the

addition of deionized water. In this study, a col-

loidal suspension with the solids content of 3.8 g/land pH 7.4 was used for seeding. Fig. 1 shows the

typical FE-SEM image of the crystals in the sus-

pension. The average particle size is about 100 nm.

The seeding was carried out as follows. The

inside of the tube support was plugged with a

Teflon rod and then dipped in the colloidal sus-

pension for 2 min, withdrawn vertically at a rate of

approximately 0.5 cm s�1, and then dried for 30min at 20 �C and subsequently for 30 min at 100�C. This process was run twice.The seeded tube was immersed vertically in a 50

ml Teflon-lined stainless steel autoclave and filled

with a clear template-free solution. The clear so-

lution with the molar composition of 10Na2O:

0.15Al2O3:36SiO2:960H2O was prepared as fol-

lows. Sodium aluminate (31.0–35.0 wt.% Na2O,34.0–39.0 wt.% Al2O3, Kanto Chemical Co. Inc.)

weighing 0.21 g was added to an aqueous sodium

hydroxide solution prepared by dissolving 3.93 g

of sodium hydroxide in 64.23 g of distilled water

while stirring. Then, 36.77 g of colloidal silica ST-S

(30–31 wt.% SiO2, <0.6 wt.% Na2O, Nissan

Chemical Ind. LTD) was slowly added to the

above solution under stirring. The resultant mix-ture became a clear solution after it was stirred at

50 �C for 4 h. Prior to being transferred to a

Teflon-lined autoclave, the clear solution was fil-

tered through a film (Millipore, / 0.8 lm). Hy-drothermal treatment was carried out at 180 �C for6–18 h. Then, the samples were recovered, washed

with deionized water, and dried at ambient con-

ditions.

2.2. General characterization

The crystalline phase of the zeolite layer formed

on the support surface was determined using a

Rigaku RINT 2000 X-ray diffractometer (XRD)

equipped with a CuKa radiation source at 40 kVand 20 mA at a scanning speed of 2� (2h) min�1.The nanosized particles in the colloidal suspensionand the seeded support were characterized using a

Hitachi S4500S field emission scanning electron

microscope (FE-SEM) operated at 15 keV. The

morphology of the zeolite layer was examined by

means of a Hitachi S-2150 scanning electron mi-

croscope (SEM) operated at 15 keV.

2.3. Pervaporation test

The pervaporation experiments were performed

using a stainless steel tube module cell. The

membrane tube was placed inside the module cell

and sealed using graphite cylindrical rings at the

both ends. The feed was circulated at a flow rate of

26 cm3 min�1 from a feed tank of 1000 cm3

through the outer side of the membrane tube,while the inside was evacuated by use of an oil

rotary vacuum pump. The stainless steel module

cell was heated with a ribbon heater. A thermo-

couple was inserted into the module cell and

placed close to the outer surface of the membrane.

The feed in the tank was heated to about 2 �Chigher than the membrane surface.

The permeate was condensed in a cold trapcooled by liquid nitrogen. The effective membrane

area was 6.28 cm2. The compositions of feed and

permeate were determined by means of a gas

chromatograph (Shimadzu GC-8A) equipped with

a 2-m long stainless steel column packed with

Porapak QS and a thermal conductivity detec-

tor (TCD). The separation factor, aW=A, was de-

fined by

aW=A ¼ ðYW=YAÞ=ðXW=XAÞFig. 1. Typical FE-SEM image of the crystals in the suspen-

sion.

G. Li et al. / Microporous and Mesoporous Materials 60 (2003) 225–235 227

where YW=YA is the weight ratio of water to aceticacid in permeate, and XW=XA is that in feed.

2.4. Gas permeation test

Gas permeation measurements were performed

in the same stainless steel tube module cell as used

in pervaporation experiments. Prior to gas per-

meation measurements, the membrane was set in

the permeation cell and heated up to 270 �C at arate of 1 �Cmin�1. During drying, feed and per-meate sides of the membrane were swept by a flow

of N2 and He, respectively. After keeping thetemperature at 270 �C for approximately 2 h,

single gas permeances of hydrogen (H2), nitrogen

(N2), n-butane (n-C4H10), i-butane (i-C4H10), and

sulfur hexafluorides (SF6) were measured at this

temperature. The transmembrane pressure differ-

ence was 100 kPa and the permeate side was at-

mospheric pressure. No sweep gas was used. To

eliminate the effect of the adsorbed gas in theprevious experiment on the subsequent experiment

when the gas type was changed, H2 permeances

were measured until it returned to the original

value.

To investigate the effect of thermal treatment at

different temperatures and the membrane thermal

stability, the membrane was further heated up to

400 �C at a rate of 1 �Cmin�1. After holding themembrane at 400 �C for approximately 2 h, the

temperature was decreased to 270 �C and single

gas permeances of H2, N2, n-C4H10, i-C4H10, and

SF6 were measured again at 270 �C. After themembrane was confirmed to be stable during this

heating and cooling cycle, single gas permeation

measurements of n-butane and i-butane were per-

formed as a function of temperature (170–400 �C).Besides single gas permeationmeasurements, the

separation of a 50/50 kPa n/i-butane mixture was

also investigated based on the Wicke–Kallenbachmethod as a function of temperature (170–400 �C).No absolute pressure difference was applied over

the membrane in this case. The mixture was fed to

the outer side of the membrane tube at a flow rate

of 20 mlmin�1. Helium was used as sweep gas that

was introduced to the inside of the membrane tube

at a flow rate of 20 mlmin�1. The flow rate was

measured using soap film flowmeters. The perme-ate and retentate were analyzed with a gas chro-

matograph (Shimadzu, GC-8A) equipped with

a 2-m long stainless steel column packed with

Porapak QS and a hydrogen flame ionization de-

tector (FID).

3. Results and discussion

3.1. Membrane preparation and characterization

Fig. 2 shows the typical FE-SEM images of the

a-alumina support used in this study before andafter seeding. It can be seen that some seeds are

located in the voids between the a-alumina parti-cles and others are attached to the surface of thesupport particles. It is obvious that the seed par-

ticles did not form a continuous layer. Xomeritakis

et al. [10] have reported that the final film quality

depends on the quality of the seed layer. The ab-

Fig. 2. Typical FE-SEM images of an a-alumina support before (A) and after (B) seeding.

228 G. Li et al. / Microporous and Mesoporous Materials 60 (2003) 225–235

sence of close-packed precursor seeds throughout

the area of the substrate can contribute to devel-

opment of dome-like defects. On the other hand, it

will be shown later in this study that continuous

films can be produced even with a lower seeding

density.The effect of crystallization time on the mem-

brane quality was first studied. The membrane

quality was evaluated using the pervaporation re-

sults for a 50 wt.% acetic acid aqueous solution

at 70 �C. Table 1 shows the pervaporation re-sults for four membranes prepared after different

crystallization times. CS-1 and CS-2 membranes,which were prepared under the same conditions

Table 1

Pervaporation results of ZSM-5 zeolite membranes for a 50/50

(w/w) water–acetic acid mixture at 70 �C

Sample Crystallization

time (h)

Water flux

(kgm�2 h�1)

a (w/a)

CS-1 6 3.96 216

CS-2 6 2.07 227

CS-3 12 0.21 17

CS-4 18 0.71 2

Fig. 3. Top and side view SEM images of ZSM-5 zeolite membranes prepared in a clear solution free from organic templates con-

taining 10Na2O:0.15Al2O3:36SiO2:960H2O at 180 �C after: (A, B) 6 h; (C, D) 12 h, and (E, F) 18 h.

G. Li et al. / Microporous and Mesoporous Materials 60 (2003) 225–235 229

(180 �C, 6 h), showed quite high water flux andseparation factors. However, the membranes pre-

pared for a longer crystallization time gave poor

pervaporation performance. In addition, the sep-

aration factor decreased with crystallization time.

To find out the possible reason, the morphologiesof these membranes were checked.

Fig. 3 shows the top and side view SEM images

of the membranes obtained after 6, 12, and 18 h of

crystallization, respectively. A remarkable differ-

ence is that larger crystals were embedded and

stacked in the growing zeolite layer in the samples

prepared for 12 and 18 h, while not in the mem-

brane crystallized for 6 h. It seems that those largercrystals are grown homogeneously in the bulk

solution and subsequently attached to the zeolite

layer on the support. As the number of the crystals

in the bulk solution increases with crystallization

time, it can be expected that more and more these

homogeneously grown crystals will be attached to

the growing zeolite layer, probably imposing a

negative effect on the zeolite layer and causing a

decrease in the separation factor observed in per-

vaporation tests.

Figs. 4 and 5 show the X-ray diffraction pat-

terns of the zeolite layer grown on the support and

the powder formed in the bulk solution after thecrystallization for 6, 12 and 18 h at 180 �C. For acomparison, the XRD patterns recorded using

commercial ZSM-5 and mordenite powder were

also incorporated in these figures. After 6 h, crys-

tals with an MFI structure were formed on the

support, whereas only amorphous species were

formed in the bulk solution. These results are in

consistent with the SEM observation that no lar-ger crystals were embedded in the zeolite layer.

After 12 and 18 h, the crystals grown on the sup-

port surface still have MFI structure, whereas

mordenite crystals were formed in the bulk solu-

tion. It is thus indicated that the larger crystals

5 10 15 20 25 30 35 40

6 h

12 h

18 h

2 θ (Cu Kα) [degree]

Inte

nsit

y[a

.u.]

Commercial ZSM-5 powder

*

**

Fig. 4. X-ray diffraction patterns of ZSM-5 zeolite membranes

grown on a seeded a-alumina support in a clear solution freefrom organic templates containing 10Na2O:0.15Al2O3:36SiO2:

960H2O at 180 �C. (*: alumina support peak).

12 h

18 h

5 10 15 20 25 30 35 40

6 h

2 θ (Cu Kα) [degree]

Inte

nsit

y[a

.u.]

Commercialmordenite powder

Fig. 5. X-ray diffraction patterns of powder formed in the bulk

solution containing 10Na2O:0.15Al2O3:36SiO2:960H2O in the

presence of an a-alumina support coated with ZSM-5 crystalseeds.

230 G. Li et al. / Microporous and Mesoporous Materials 60 (2003) 225–235

embedded in the zeolite layer, as observed in SEM

images, are mordenite crystals instead of ZSM-5

crystals.

3.2. Gas permeation properties

3.2.1. Effect of thermal treatment on single gas

permeation

Fig. 6 shows the single gas permeances of N2,

H2, SF6, n-butane, and i-butane at 270 �C as a

function of kinetic diameter. The single gas per-

meance was measured after the membrane was

dried under different conditions. After the mem-

brane was treated at 270 �C for 2 h, the measuredsingle gas permeance decreased with increasing

kinetic diameter. After the membrane was treated

at 400 �C for 2 h, however, n-butane gave a larger

permeance than N2 and SF6 gave a larger perme-

ance than i-butane. These results are consistent

with those reported by other investigators [11–13].

From the results shown in Fig. 6, three con-

clusions could be drawn. First, the membrane isstable during heating-cooling cycles in this tem-

perature range. This can be judged from the fact

that i-butane permeance hardly changed after

these two thermal treatments. Second, thermal

treatment at 270 �C for 2 h could not remove allthe adsorbed species from pores in the zeolite

membrane since all single gas permeances except

for i-butane increased when the membrane was

further dried at 400 �C for 2 h. Third, the mem-brane has a low concentration of defects. This is

clearly illustrated by the fact that i-butane and SF6permeances were significantly lower than the per-

meances of smaller molecules. Note that i-butane

(0.5 nm) and SF6 (0.55 nm) are close to the poresize of ZSM-5 crystals (0.55 nm).

After the second thermal treatment, single gas

permeances of H2, N2, n-butane, and SF6 in-

creased while i-butane permeance did not. It in-

dicates that i-butane may mainly permeate through

the crystal boundaries. It can be expected that

adsorbed species in the crystal boundaries would

be easily removed than those in the zeolitic pores.It seems that adsorbed species in the crystal

boundaries had completely been removed during

the first thermal treatment. The second thermal

treatment had little influence on the size of the

crystal boundaries. This may explain why i-butane

permeances were almost unchanged when mea-

sured after the second thermal treatment. On the

other hand, due to the removal of adsorbed speciesfrom zeolitic pores, single gas permeances of small

molecules increased after the second thermal

treatment.

Another point that needs to explain is why

n-butane permeance became larger than N2 per-

meance after the second thermal treatment. The

permeance of one component is expected to be

proportional to both its concentration in the ad-sorbed phase and its diffusion rate. n-Butane is

adsorbed more strongly than N2 on MFI crystals,

while N2 diffuses faster than n-butane. When the

membrane was treated at 270 �C, the effective dia-meter of zeolitic pores was smaller than expected

for ZSM-5 due to partial blocking of the zeolite

channels. It will impose more transport barrier on

larger molecules (n-butane) than smaller molecules(N2). Thus, n-butane may diffuse much more

slowly than N2. In this case, the diffusion rate

dominates over sorption and N2 permeance is thus

larger than n-butane permeance. After the second

thermal treatment, the effective pore size becomes

large due to the removal of adsorbed species. In

this case, the diffusion of n-butane may be com-

parable to that of N2. n-Butane permeance be-comes larger than N2 permeance due to its high

occupancy in the membrane surface.

Fig. 6. Single gas permeances at 270 �C as a function of kineticdiameter measured after different thermal treatments. CS-2

membrane.

G. Li et al. / Microporous and Mesoporous Materials 60 (2003) 225–235 231

Fig. 7 shows the molecular size dependence of

H2/N2, H2/n-butane, H2/i-butane and H2/SF6 ideal

selectivities measured at 270 �C after the thermaltreatment at 400 �C for 2 h. For a comparison, thecorresponding values calculated from Knudsen

diffusion were also shown in Fig. 7. The ideal se-

lectivity of H2/N2 was close to the value calculated

from Knudsen diffusion. This can be expected since

both H2 (0.289 nm) and N2 (0.364 nm) are smaller

than the pore diameter of zeolite MFI (0.55 nm).

The ideal selectivity of H2/n-butane was a littlesmaller than the value calculated from Knudsen

diffusion. This is probably due to surface diffusion

of n-butane, which increases its permeance. On the

other hand, the ideal selectivities of H2/i-butane

and H2/SF6 were 61 and 47, respectively. The H2

permeance was 2.85� 10�7 mol s�1 m�2 Pa�1. These

data are significantly larger than the values (5.4 and

8.5, respectively) calculated from Knudsen diffu-sion. It can be considered as a result of molecular

sieving effect. A ZSM-5 zeolite membrane, reported

by Coronas et al. [14], showed H2/i-butane ideal

selectivity was 2820 at room temperature but de-

creased to less than 5 at 267 �C.

3.2.2. Permeation of butane isomers

Fig. 8 shows the n-butane permeances forsingle-component and for a 50/50 n/i-butane mix-

ture as a function of temperature. The single gas

permeance increased with increasing temperature

and then showed a slight decrease above 300 �C. Ingeneral, maxima in the n-butane permeance versus

temperature curve are always observed at ap-

proximately 150 �C for silicalite-1 membranes

[11,15–17]. It has been attributed to a result of

opposite effects of sorption and diffusion. At low

temperatures, strong adsorption dominates, the

concentration in the adsorbed phase is constant,

and the permeance increases with temperature dueto the increase in mobility of the molecule. At a

certain temperature the decrease in occupancy

supersedes the increase in mobility and the per-

meance decreases. This maximum occurs at higher

temperatures for more strongly adsorbed mole-

cules [15]. The maximum in the current study

appears at a temperature as high as 300 �C, indi-cating that n-butane molecules are adsorbedstrongly on ZSM-5 membrane prepared by us.

Over the entire temperature range studied, n-

butane permeances in the presence of i-butane are

smaller than the corresponding single gas perme-

ances. The lower n-butane permeance is probably

due to the inhibition by i-butane and/or back

permeation of helium. The temperature depen-

dency of n-butane permeances in the presence of i-butane was somewhat different from that of single

gas permeances. At low and high temperatures, n-

butane permeances increased slightly with tem-

perature. Between 250 and 320 �C, the permeancesincreased much with temperature. It can be ex-

0

10

20

30

40

50

60

70

3 4 5 6

H2/N2

H2/n-C4H10

H2/i-C4H10

H2/SF6

Knudsen selectivities

Idea

lsel

ectiv

ities

[-]

Kinetic diameter [Å]

Fig. 7. Ideal selectivities as a function of kinetic diameter at

270 �C. CS-2 membrane.

0

50

100

150

200

100 200 300 400 500

Temperature [oC]

Single component

A 50/50 n/i-butane mixture

Perm

eanc

e×10

9[m

ols-1

m-2

Pa-1

]

Fig. 8. n-Butane permeances for single-component and for a

50/50 n/i-butane mixture as a function of temperature.

232 G. Li et al. / Microporous and Mesoporous Materials 60 (2003) 225–235

plained as follows. At low temperatures, the ad-

sorption of i-butane is strong, which will hinder

the adsorption and thus the permeance of n-

butane. As the temperature increases, the increase

in the permeance of n-butane could be due to the

less significant hindering by i-butane and the in-crease in mobility of n-butane molecules. The

permeance of n-butane increases less at high tem-

perature due to the decrease in occupancy.

Fig. 9 shows the i-butane permeances for single

component and for a 50/50 n/i-butane mixture as a

function of temperature. It is interesting to note

that the temperature dependency of single gas

permeances of i-butane is quite different from thatof n-butane. The single gas permeances of i-butane

increased with temperature and had a maximum at

200 �C. A minimum was also observed around 300�C. Above 300 �C, the permeances increasedmonotonously with temperature. This behavior

disagrees with other reports [14–16], where the

permeance of i-butane increased monotonously

with temperature. The maximum in the i-butanepermeance that was observed for our membrane

can also be attributed to a result of opposite effects

of sorption and diffusion, as discussed earlier. As

the adsorption capacity of i-butane is smaller than

that of n-butane, the maximum in the permeance

versus temperature of i-butane occurs at a lower

temperature than n-butane. The increase in the i-

butane permeance at high temperature is probably

due to the increase in zeolite and/or non-zeolite

pore size with temperature [13].

The permeances of i-butane in the presence of

n-butane are also smaller than single gas perme-

ances over the entire temperature range studied,

indicating that n-butane also inhibits the per-meation of i-butane. At low temperatures, the

permeance of i-butane hardly changes with tem-

perature, as observed for the permeance of n-

butane. Above 300 �C, it increases monotonouslywith temperature. It may also be due to the in-

crease in zeolite and/or non-zeolite pore size with

temperature [13].

Fig. 10 shows the temperature dependency of n/i-butane ideal selectivities calculated from the

single gas permeances and separation factors for a

50/50 mixture. Over the entire temperature range

studied, the separation factors are smaller than the

ideal selectivities. Vroon et al. [15,18] observed

that the separation factors were smaller than the

ideal selectivities at 25 �C, while both were close at200 �C. A similar trend was also observed by vande Graaf et al. [19]. On the other hand, Coronas

et al. [14] reported that n/i-butane separation fac-

tors were higher than ideal selectivities, which was

attributed to preferential adsorption of n-butane.

Fig. 10 shows that the ideal selectivities first

increase and then decrease with temperature. It

reaches a maximum of 36 at approximately 300

�C. Tuan et al. [20] reported similar behavior ofideal selectivities versus temperature for silicalite-1

membranes on c-alumina and stainless steel

0

2

4

6

8

10

100 200 300 400 500

Temperature [°C]

Perm

eanc

e×10

9[m

ols-1

m-2

Pa-1

]

Single-component

A 50/50 n/i-butane mixture

Fig. 9. i-Butane permeances for single-component and for a 50/

50 n/i-butane mixture as a function of temperature.

0

10

20

30

40

100 200 300 400 500

Temperature [°C]

Idea

lsel

ecti

vity

/sep

arat

ion

fact

or[-

]

Ideal selectivities

Separation factors

for the mixture

Fig. 10. Ideal selectivities and separation factors for a 50/50 n/i-

butane mixture as a function of temperature.

G. Li et al. / Microporous and Mesoporous Materials 60 (2003) 225–235 233

supports, whereas the ideal selectivities of the a-alumina-supported membrane decreased with in-

creasing temperature. Bakker et al. [16], Bai et al.

[21], and Geus et al. [22] reported that n/i-butane

ideal selectivities decreased with temperature,

whereas the opposite trends were reported byGiroir-Fendler et al. [17], Yan et al. [23], and

Kusakabe et al. [24].

Fig. 10 shows that the temperature dependency

of separation factors is different from that of ideal

selectivities. While being close to unity below 250

�C, the separation factor increased with tempera-ture between 250 and 320 �C. Above 320 �C, theseparation factor showed a slight decrease but stillas high as 13 at 400 �C.Maxima of separation factor as a function of

temperature were also reported by Tuan et al.

[13,20]. As the temperature increases, the n-butane

permeance increases because of activated diffusion,

but i-butane is blocked by n-butane so the i-butane

permeance does not increase so fast. Thus, the

separation factor increases. At high temperatures,the n-butane permeance increases less due to the

decrease in occupancy. However, the i-butane

permeance increases at high temperature, and this

decreases the separation factor. The increase in i-

butane permeance may be due to a slight increase

in zeolite pore size with temperature, which would

affect the permeance of the larger molecule more.

It may also be due to an increased size for non-zeolite pores at high temperature [13].

3.3. Comparison with MFI membranes in the

literature

The n/i-butane ideal selectivities and separation

factors of a 50/50 mixture have been established as

a criterion for the quality of an MFI membrane[11,12,16,18,19,25]. Burggraaf et al. [11] consid-

ered a membrane to be ‘‘defect-free’’ if the n/i-

butane ideal selectivity was >10 at 200 �C. On theother hand, van de Graaf et al. [19] considered

their membranes to be of good quality if the ideal

selectivity was >10 at room temperature. Vroon

et al. [18] reported that a compact membrane

showed 11 of the n/i-butane ideal selectivity at200 �C. Our membrane showed n/i-butane ideal

selectivities of 10–35 at 200–400 �C, which are

comparable or better than the above membranes.

This might be due to the difference in the effective

zeolitic pore size. ZSM-5 membranes grown from

organic-free solutions may have smaller effective

pore size, possibly due to incorporation of Na ions

and OH groups in the zeolitic channels.Though the permeation of butane isomers

through MFI zeolite membranes has been studied

by many investigators, however, data are quite few

at temperatures above 200 �C. van de Graaf et al.[19] reported that the n/i-butane separation factor

with a stainless steel-supported silicalite-1 disk

membrane was 23.3 at 30 �C, but it decreased tounity at 402 �C. Coronas et al. [25] reported thatthe separation factor of one ZSM-5 zeolite mem-

brane had a maximum of 55 at 117 �C for a 50/50n/i-butane mixture, but the selectivity decreased to

unity at 213 �C. In contrast, the n/i-butane sepa-

ration factor of our membrane is as high as 13 at

400 �C.Another characteristic of our membrane is its

high thermal stability. This membrane has experi-enced several heating and cooling cycles (100–400

�C).Nodeterioration of themembrane performancehas been observed after thermal cycles, suggesting

that the membrane prepared in the current study is

very attractive for applications in catalytic mem-

brane reactors.

4. Conclusions

Template-free nanosized ZSM-5 seeds were

prepared from commercially available powder. By

use of these seeds, thin and hydrophilic ZSM-5

zeolite membranes were prepared on the outer

surface of a porous a-alumina tube in a clear so-lution free from organic templates.The pores in the zeolite membrane were occu-

pied by adsorbed species. Thermal pretreatment at

400 �C is necessary to remove these species. Themembranes had excellent thermal stability to

withstand high temperature treatment. The ideal

selectivities of H2/i-butane and N2/SF6 were 61 and

20 at 270 �C, respectively.The membrane is effective to separate butane

isomers at high temperatures. The n/i-butane ideal

selectivity showed a maximum of 36 at 300 �C. The

234 G. Li et al. / Microporous and Mesoporous Materials 60 (2003) 225–235

separation factors for a 50/50 n/i-butane mixture

were smaller than the corresponding ideal selec-

tivities over a wide temperature range (170–400

�C). However, the membrane still showed separa-tion factors as high as 13 in the temperature range

from 300 to 400 �C.

References

[1] J. Caro, M. Noack, P. Kolsch, R. Schafer, Micropor.

Mesopor. Mater. 38 (2000) 3.

[2] J. Hedlund, J. Sterte, M. Anthonis, A.-J. Bons, B. Carsten-

sen, N. Corcoran, D. Cox, H. Deckman, W.D. Gijnst, P.-P.

de Moor, F. Lai, J. McHenry, W. Mortier, J. Reinoso, J.

Peters, Micropor. Mesopor. Mater. 52 (2002) 179.

[3] M.J. Exter, H. Bekkum, C.J.M. Rijn, F. Kapteijn, J.A.

Moulijn, H. Schellevis, C.I.N. Beenakker, Zeolites 19

(1997) 13.

[4] J. Dong, Y.S. Lin, M.Z. Hu, R.A. Peascoe, E.A. Payzant,

Micropor. Mesopor. Mater. 34 (2000) 241.

[5] J. Hedlund, M. Noack, P. Kolsch, D. Creaser, J. Caro, J.

Sterte, J. Membr. Sci. 159 (1999) 263.

[6] M. Noack, P. Kolsch, J. Caro, M. Schneider, P. Toussaint,

I. Sieber, Micropor. Mesopor. Mater. 35–36 (2000) 253.

[7] R. Lai, G.R. Gavalas, Micropor. Mesopor. Mater. 38

(2000) 239.

[8] M. Pan, Y.S. Lin, Micropor. Mesopor. Mater. 43 (2001)

319.

[9] M. Lassinantti, F. Jareman, J. Hedlund, D. Creaser, J.

Sterte, Catal. Today 67 (2001) 109.

[10] G. Xomeritakis, A. Gouzinis, S. Nair, T. Okubo, M. He,

R.M. Overney, M. Tsapatsis, Chem. Eng. Sci. 54 (1999)

3521.

[11] A.J. Burggraaf, Z.A.E.P. Vroon, K. Keizer, H. Verweij, J.

Membr. Sci. 144 (1998) 77.

[12] K. Kusakabe, A. Murata, T. Kuroda, S. Morooka, J.

Chem. Eng. Jpn. 30 (1997) 72.

[13] Vu A. Tuan, J.L. Falconer, R.D. Noble, Micropor.

Mesopor. Mater. 41 (2000) 269.

[14] J. Coronas, J.L. Falconer, R.D. Noble, AIChE J. 43 (1997)

1797.

[15] Z.A.E.P. Vroon, K. Keizer, M.J. Gilde, H. Verweij, A.J.

Burggraaf, J. Membr. Sci. 113 (1996) 293.

[16] W.J.W. Bakker, F. Kapteijn, J. Poppe, J.A. Moulijn, J.

Membr. Sci. 117 (1996) 57.

[17] A. Giroir-Fendler, J. Peureux, H. Mozzanega, J.A. Dal-

mon, Stud. Surf. Sci. Catal. 111 (1996) 127.

[18] Z.A.E.P. Vroon, K. Keizer, M.J. Gilde, H. Verweij, J.

Membr. Sci. 144 (1998) 65.

[19] J.M. van de Graaf, E. van der Bijl, A. Stol, K. Kapteijn,

J.A. Moulijn, Ind. Eng. Chem. Res. 37 (1998) 4071.

[20] Vu A. Tuan, J.L. Falconer, R.D. Noble, Ind. Eng. Chem.

Res. 38 (1999) 3635.

[21] C. Bai, M.D. Jia, J.L. Falconer, R.D. Noble, J. Membr.

Sci. 105 (1995) 79.

[22] E. R Geus, H. van Bekkum, W.J.W. Bakker, J.A. Moulijn,

Micropor. Mater. 1 (1993) 131.

[23] Y. Yan, M.E. Davis, G.R. Gavalas, Ind. Eng. Chem. Res.

34 (1995) 1652.

[24] K. Kusakabe, S. Yoneshige, A. Murata, S. Morooka, J.

Membr. Sci. 116 (1996) 39.

[25] J. Coronas, R.D. Noble, J.L. Falconer, Ind. Eng. Chem.

Res. 37 (1998) 166.

G. Li et al. / Microporous and Mesoporous Materials 60 (2003) 225–235 235