Microporous and Mesoporous Materials 100 (2007) 173–182

www.elsevier.com/locate/micromeso

Preparation of templated mesoporous silica membranes

on macroporous a-alumina supports via direct coating
of thixotropic polymeric sols
V. Boffa, J.E. ten Elshof *, D.H.A. Blank
Inorganic Materials Science, MESA+ Institute for Nanotechnology and Faculty of Science and Technology, University of Twente,
P.O. Box 217, 7500 AE Enschede, The Netherlands

Received 19 September 2006; received in revised form 20 October 2006; accepted 24 October 2006
Available online 8 December 2006

Abstract

A simple method for depositing mesoporous silica films directly on macroporous a-alumina supports is reported. A polymeric silica
sol was prepared by hydrolysis of tetraethylorthosilicate in acid propanol in presence of hexadecyltrimethylammonium bromide as tem-
plate. The rheology of the sol was changed by a modified urea-based thixotropic agent, which was used in a concentration range between
2% and 10% on volume. The dynamic viscosity of the sol was measured as a function of shear rate. The rheology modifier was found to
increase the viscosity of the sol and confer at the same time a thixotropic behavior to it. Long range order in the unmodified and modified
materials was analyzed by XRD on powders after calcination at 500 C. It was found that the pore size increased with increasing con-
centration of additive, while the degree of order decreased until eventually, at high concentrations, the ordered porous structure collapsed
completely. The gradual pore size increase at lower concentration is thought to be caused by swelling of the surfactant micelles by the
rheology modifier.
The modified sols were deposited via dip-coating on a-alumina disks. The morphology of the silica top layers was characterized by
SEM. Pore size distributions and permeances of the prepared membranes were assessed by permporometry and single gas permeation
measurements, respectively. The membranes had narrow pore size distributions, with an average diameter of about 2 nm. Permeabilities
of H2, He, CH4 and O2 were compared. Hydrogen permeances up to 6 · 107 mol s1 m2 Pa1 were measured at 473 K. The gas flux
across the membrane follows a Knudsen-type mechanism, as shown by the permselectivities.
 2006 Elsevier Inc. All rights reserved.

Keywords: Mesoporous silica; Membranes; Thixotropy; Rheology modifier; Sol–gel

1. Introduction non-ionic or cationic surfactants as templates [3,4]. The

interest in this new class of materials grew when it was dis-
Templated silica is one of the most attractive materials covered that glasses with regular porous structure were
for those applications in which a high specific surface area formed in the presence of quaternary ammonium surfac-
and a narrow pore size distribution in the mesoporous tants (M41S materials) [5–7]. It was directly recognized
range are required [1,2]. Among all the organic molecules that these materials may be applicable in a large number
that can be used to tailor the pore structure of the silica of fields beside heterogeneous catalysis; for example as
matrix, surfactants are the most flexible ones. Silicon oxi- membranes, sensors and in nano-electronics [8,9]. The
des with lamellar, hexagonal or cubic pore structure can two main requirements for a membrane, for instance, are
be obtained under different synthetic conditions by using high selectivity and high permeate flux [10]. Thus, tem-
plated glasses, which have a high porosity and a regular
*
Corresponding author. Tel.: +31 53 4892695. pore structure, are really attractive materials for the prep-
E-mail address: [email protected] (J.E. ten Elshof). aration of selective ceramic membranes [11].

1387-1811/$ - see front matter  2006 Elsevier Inc. All rights reserved.
doi:10.1016/j.micromeso.2006.10.035
174 V. Boffa et al. / Microporous and Mesoporous Materials 100 (2007) 173–182

Nomenclature

g dynamic viscosity [mPa Æ s] cs gas-solid interfacial tension [J Æ m2]

s torque [%] U flux [mol Æ s1 Æ m2]
u cylinder velocity h contact angle of the condensed phase on the
R gas constant [J Æ mol1 Æ K1] pore wall
T temperature [K] e porosity
rk kelvin radius [nm] s tortuosity
lt t-layer thickness [nm] Atot total area [m2]
Vmol molar volume [m3 Æ mol1] M molecular mass [g Æ mol1]

In the first report in which a surfactant templated silica ter cases the sol is formed by dense particles that have a size
layer was applied in a stacked membrane, it was applied as between 20 and 50 nm. Large quantities of binder such as
a spacer [12] between the mesoporous sub-layer and the polyvinyl alcohol (PVA) are added to change the rheolog-
microporous top layer of a ceramic gas separation mem- ical properties of the sols and the viscoelastic properties of
brane. The spacer prevented penetration of the top layer dried uncalcined films. The coating of polymeric silica sols
into the sub-layer during deposition. In later works tem- onto a-alumina supports is complicated by the fact that the
plated silica has also been used to replace the conventional polymeric silica sol is formed by particles that are far smal-
mesoporous intermediate layers [13–15]. However, the ler (only a few nm [24]) than the pores of the support,
application of templated silica films directly on a macropo- which are often larger than 80–100 nm diameter. Further-
rous support is not trivial. Indeed both sol synthesis and more, binders can be used only in small concentrations in
membrane coating procedure are critical steps. order not to affect the supramolecular ordering of the sur-
In the first years after the discovery of ordered mesopor- factant micelles. The necessity to work in a restricted range
ous oxides the focus was mainly on the preparation of pow- of synthetic conditions, the difficulty to coat a polymeric
ders and thin films on dense substrates [5,16]. However, the sol onto a macroporous support, and the impossibility to
recipes that were developed for these cases are often not use ordinary binders makes the coating of mesoporous sil-
suitable for preparing membranes. Only a stable silica ica films on porous substrates a challenge.
sol, with polymeric particles in the nanometer-size region, Templated silica membranes applied directly on a-alu-
can yield a gas-selective membrane. Highly ordered M41S mina substrates were successfully made by Roy Chowdh-
powders can be prepared easily at basic pH via strong ury et al. via dip coating [13,14], and by Huang et al. via
electrostatic interaction between the cationic ammonium casting [15]. Solvent transport through membranes pre-
salt (S+) and the negatively charged silica particles (I) pared by Roy Chowdhury et al. showed behavior which
[17,18]. The S+I type interaction is commonly used for is typical for truly mesoporous systems. On the other hand,
making powders. However, in order to synthesize a stable the XPS depth profiles implied the absence of a real silica
polymeric sol, a high degree of control over the condensa- layer after calcination [13]. The concentration of alumin-
tion reaction is required. At basic pH the rapid condensa- ium at the surface of the membrane was larger than that
tion of silica particles would lead to fast precipitation or of silicon, which suggests that the membrane was consti-
gelation. But when the pH is slightly below the silica iso- tuted by silica penetrated into the substrate [14,25].
electric point at pH 2–2.5 [19], the condensation rate is at In recent years two solutions to this problem have been
a minimum and it is possible to obtain transparent sols that proposed. The most common way for preparing templated
are stable for many days even at room temperature. Never- silica membranes is via hydrothermal growth of the mate-
theless, under such conditions the interaction between the rial on top of a substrate. The substrate is immersed in a
cationic surfactant and the slightly positively charged silica mixture containing surfactant, solvent and an organic or
particles (I+) is weaker and it is mediated by the surfactant inorganic silica precursor [26,27]. Water and a base are
counterion (X). Under such conditions, which are gener- present to promote the hydrolysis and condensation reac-
ally described as S+XI+ interactions, it is more difficult tions that lead to the formation of silica. Membranes
to obtain an ordered structure [17]. Furthermore, coating obtained with this procedure were gas selective and appli-
the sol onto a porous support is far more difficult because cable in nanofiltration processes [28]. In spite of that, silica
of the perturbing effect of the substrate on the drying film. grew both inside and on top of the support, forming a
Substrate roughness and capillary forces exerted by the resistive thick layer. An alternative procedure has been
substrate pores can easily result in film cracking or distor- introduced by Boissiere et al. [29]. A tubular support was
tion of the ordered mesostructure [20]. The coating of a impregnated by an aqueous solution of NaF. NaF cata-
templated silica gel on a macroporous substrate such as lyzes the condensation reaction of the silica precursor. In
a-alumina differs from the coating of traditional colloidal this way the growth of the selective silica layer was
sols like boemite [21], titania [22] or zirconia [23]. In the lat- restricted to the area on top of the support. In order to
 V. Boffa et al. / Microporous and Mesoporous Materials 100 (2007) 173–182 175

be reproducible this procedure requires a high degree of Table 1

control over all the parameters involved in the preparation. Coating solutions (S) used in this work for the preparation of powders (P)
and membranes (M)
In this work, we propose a different approach to deposit
templated mesoporous silica membranes directly on mac- Coating BYK-420 Name of Number Name of
solution volume unsupported of layers membrane
roporous supports. It involves tuning of the rheological fraction sample
properties of the sol to render it a thixotropic behavior,
S00 0.00 P00 * *
so that it can be coated directly on a macroporous sub- S02 0.02 P02 2 M02
strate. We modified the rheological properties of a poly- S06 0.06 P06 2 M06
meric silica sol by adding BYK-420 (kindly provided by S10 0.10 P10 2 M10
BYK-Chemie, Wesel, Germany), which consists of a mod- *No membranes were obtained without BYK-420.
ified urea solution. BYK-420 is a commercial additive used
in aqueous coatings to improve anti-sagging properties or
as anti-settling agent for pigments. This product is able then fired at 500 C for 3 h (heating/cooling rates of
to increase the viscosity of a suspension, conferring thixot- 0.5 C/min). After deposition of the first layer macropores
ropy at the same time. The thixotropy inhibits penetration in the support could still be detected by permporometry in
into the pores of the a-alumina substrate during the drying all samples, irrespective of BYK-420 concentration. In
stage of the silica film. This behavior can simplify the pro- order to cover these defects a second coating layer was
cessing of polymeric sols in the preparation of silica applied, followed by calcination.
membranes. Powders were prepared by drying the same sols as used
for thin films in plastic Petri-dishes. Table 1 lists the differ-
2. Experimental part ent concentrations of additive used for preparing powders
and thin films. For the sake of simplicity we will refer to the
2.1. Sol synthesis samples following the designations listed in Table 1.

Aqueous nitric acid (0.05 N) was dropped under vigor- 2.3. Characterization
ous stirring in a conical flask containing a solution of
tetraethylorthosilicate (TEOS, 98% purity, Aldrich), in The viscosity of the coating sols were measured by a
1-propanol (analytical grade, Acros). After 5 min the tem- Brookfield DV-II+ Programmable Viscometer (Brookfield
plating agent cetyltrimethylammonium bromide (CTAB, Engineering Laboratories, Middleboro, MA, USA). The
P99% purity, Aldrich) was added. The molar ratio of instrument was calibrated using Brookfield fluid 5 (Brook-
the final mixture was TEOS/CTAB/1-propanol/H2O/ field Engineering Laboratories, Middleboro, MA, USA) as
HNO3 = 1:0.13:15:15.5:0.14. All reagents were mixed standard. The viscosity of this fluid is 4.8 MPa at 25 C.
under nitrogen atmosphere in order to avoid the hydrolysis The temperature was kept constant at 30 C by a thermo-
of TEOS before the reaction started. static water bath. The viscosity was measured using a
The mixture was heated at boiling temperature for 2 h 90 mm long cylinder with a diameter equal to 25 mm and
under reflux; then it was allowed to cool down. The trans- a weight of 78.5 g. The cylinder was stirred in 25 ml of sam-
parent sol was kept at 5 C for a few days until it was used ple contained in a metallic tube. Data were recorded with
for sample preparation. increasing angular speed, ranging from 1.0 to 10.5 rad s1,
followed by stepwise decreasing the speed to 1.0 rad s1.
2.2. Thin film deposition and powder preparation All data on powders and films reported in this paper con-
cern calcined samples. XRD patterns of powders were
a-Alumina disks were used as macroporous supports for recorded using a Philips SR5056 with Cu Ka radiation.
film deposition. They were prepared by colloidal filtration Nitrogen sorption measurements were performed at 77 K
of a-alumina powder (AKP30, Sumitomo, Tokyo, Japan) on calcined silica, using a porosimeter (Micromeretics,
dispersed in acidic water, as described elsewhere [30]. After ASAP 2400). Pore size distributions of templated silica
drying at room temperature, the green supports were membranes were measured by permporometry using the
calcined at 1100 C (heating/cooling rates of 2 C/min). type of set-up described by Cao et al. [32,33]. In the perm-
Flat disks of 39 mm diameter and 2.0 mm thickness were porometer the flux of oxygen across the membrane was
obtained after polishing. The final porosity of these sup- measured by a Varian 3300 gas chromatograph (Varian,
ports was about 30% and the average pore size was in Palo Alto, CA, USA). Scanning Electron Microscopy
the range of 80–120 nm [31]. (SEM) images were taken on a LEO 1550 FEG. Single
Varying amounts of BYK-420 (BYK-Chemie, Wesel, gas permeation measurements were carried out in a dead-
Germany) were added to the aged silica sol and the mixture end mode permeation set-up [34]. The gas flow was
was stirred for a few hours. The alumina macroporous sup- measured with a soap film flow meter. Hydrogen, helium,
ports were then dip-coated in the solution under class-1000 methane and oxygen permeances were measured on differ-
clean room conditions. The samples were dried at 40 C in ent samples at 200 C, with a pressure of 4 bar on the side
air (60% relative humidity) for 3 h. The membranes were of the feed and atmospheric pressure on the retentate side.
176 V. Boffa et al. / Microporous and Mesoporous Materials 100 (2007) 173–182

Since atmospheric water can condense in the pores of the highly non-polar systems the active component precipitates
silica matrix, all membranes were dried at 200 C in a in the form of white powder.
hydrogen flow for at least 16 h before measurements were The rheological properties of the modified sols are
done. In a typical series of measurements, the permeation shown in Fig. 1. Prior to measurement, the coating solu-
rate of different gases was determined starting from the tions were prepared by stirring 50 ml of silica sol with an
one with the lower molecular mass, and equilibrating at adequate amount of additive for 2 h. Fig. 1A shows the
least 2 h each time a new gas was introduced into the torque versus angular speed of different solutions. The vis-
set-up. cosities of solvent and sols shown in Fig. 1B were calcu-
lated from these data according to the formula [35]:
3. Results and discussion s
g ¼ K1  : ð1Þ
u
3.1. Sol modification
K1 is a constant that depends on the geometry, weight and
BYK-420 is a pale yellow solution and 52% of the total dimensions of the spindle stirring the fluid. As can be seen
weight is non-volatile material. It is constituted by modi- in Fig. 1B, the viscosity of both solvent and unmodified sol
fied urea, which is stabilized in solution by N-methylpyrr- is independent of shear rate, which indicates that they be-
olidone. Once the product has been added to the final have as Newtonian fluids. The viscosity of 1-propanol
coating solution the modified urea crystallizes in the etha- was measured to be 1.76 mPa s, which is in agreement with
nol-based solution, yielding thin fibers which result in a tri- the literature value of 1.72 mPa s at 30 C [36]. The viscos-
dimensional network. This network confers pseudo-plastic ity of the unmodified coating sol S00 was measured to be
behavior to the coating mixture. BYK-420, like other prod- 2.71 mPa s.
ucts with similar composition, is effective at 1–3% in vol- After addition of the modifier the viscosity increased
ume, but cannot be applied to all sols. In very polar or over the whole range of shear rates. The increasing viscos-
ities at lower shear rates shown in Fig. 1B indicate the
occurrence of pseudo-plasticity in the sols. After leaving
the sols undisturbed for a while it was seen that the viscos-
ity at the onset of a measurement increased to values higher
than 5 MPa s for sample S02. The sol also became thixotro-
pic after addition of BYK-420, since it was observed that
the high viscosity of the coating solution is partially
reduced upon prolonged stirring. This phenomenon is vis-
ible in Fig. 1A and B, where hysteretic behavior can be
observed in the curves of the urea-containing sols. Higher
concentrations of BYK-420 yielded sols with higher viscos-
ities. These observations show that BYK-420 can be a suit-
able rheology modifier to tune the visco-elastic properties
of silica sols. The appearance of S10 was turbid. Turbidity
usually implies that it is not suitable for preparation of a
defect-free membrane coating. A membrane was however
prepared with this sol and its permporometry profile will
be shown below. Concentrations of additive higher than
10 vol% on total volume resulted in unstable silica sols in
which precipitation occurred rapidly.

3.2. Powder analysis

Templated silicon oxides were prepared as powder with

the aim to analyse the intrinsic properties of the material
and to study the effect of the rheological additive on the
mesoporous structure. The XRD profile of P00 in Fig. 2
after calcination at 500 C shows a broad peak at 2.7 2h,
indicating the presence of long range order.
Fig. 2 also shows the XRD patterns of three powders
Fig. 1. Rheological properties of solvent and silica sols, before and after
prepared with different concentrations of BYK-420, and
being modified with BYK-420. (A) Torque stress against the cylinder
angular speed; (B) viscosity versus angular speed. The names of the calcined at 500 C. The XRD peak of P02 is broader but
samples are indicated on the right side of the graphs. Angular speed is more intense than the one obtained from P00. The
expressed in rad s1. increased peak intensity can probably be explained by the
 V. Boffa et al. / Microporous and Mesoporous Materials 100 (2007) 173–182 177

Fig. 2. XRD patterns of powders obtained with different concentrations Fig. 3. Nitrogen sorption isotherms at 77 K for four samples prepared
of BYK-420 after calcination at 500 C. with various BYK-420 concentrations: adsorption (black bullets) and
desorption (white bullets) isotherms are shown.

longer drying time of sol S02, which is due to the presence

of the slow evaporating cosolvent N-methylpyrrolidone. As Table 2
a result the micelles have more time to form an ordered BET surface area, pores fraction and average pore width as obtained from
mesostructure. In general, we found that the rheological nitrogen sorption data
additive had a perturbing effect on the degree of order in Sample BET Surface Porosity BET average pore
the templated microstructure: from P00 to P06 the peak area (m2/g) width (Å)
became progressively wider and in the case of P10 it had P00 1380 0.61 18
disappeared completely. We may suppose that at least P02 1220 0.60 20
one of the components of BYK-420 is strongly interacting P06 790 0.40 21
P10 420 0.44 42a
with the surfactant and perturbing the formation of
a
micelles. Both solvent [37] and organic modifier [38,39] Average value of pores smaller than 650 Å. This is an average value of
a broad distribution.
may cause the micelles either to swell or shrink, leading
to larger or smaller pores in the final oxide. However, it
has been noticed that the degree of long range order is desorption cycle. The specific surface area was reduced
higher when no perturbing agents are present [3]. Anderson from 1220 m2 g1 in P02 to 423 m2 g1 in P10, while the
et al. [40], for instance, monitored by XRD the long range average pore diameter of P10 was 4.2 nm, more than dou-
order in templated silica obtained by hydrolysis of tetra- ble of that measured for P00. It is stressed that the average
methoxysilane in aqueous methanol. The long-range order pore diameter was calculated from the BET pore volume
of the material decreased as the methanol concentration and specific surface area. While the pore size distributions
increased, while at concentrations higher than 70 wt% a of P00, P02, and P06 were narrow, the pore size distribu-
disordered arrangement of pores was observed. In Fig. 2, tion of P10 was broad. Hence, although P10 has an average
the shift of the reflection signal from 2.7 for P00 to about pore diameter of about 3.9 nm, this value is not so relevant
2 for P06 suggests that BYK-420 causes an enlargement of in view of the broadness of the distribution.
the surfactant micelles. Indeed, similar XRD results have
been obtained, for instance, by adding decane as swelling
agent to a silica sol in the preparation of MCM-41 [41]. 3.3. SEM images
These considerations are supported by the sorption
curves in Fig. 3 and the data listed in Table 2. The increase Fig. 4A and B shows SEM pictures of doubly coated
of the concentration of rheology modifier led to the forma- and calcined M02 at different magnifications. The tem-
tion of a denser material, with lower specific surface area plated silica layer is seen to completely penetrate the
and porosity, and larger pores. Comparison of the sorption support, without forming a continuous layer on top. Pene-
curves of P00 and P02 indicate that a small amount of tration into the support was inhomogeneous and was at
additive BYK-420 does not change the total porosity of least a few micrometers in depth. Upon increasing the con-
the material considerably, but it does increase the degree centration of the rheological additive to 6 vol%, a continu-
of mesoporosity, as illustrated by the BJH pore sizes in ous flat layer was obtained as shown in Fig. 4C. The layer
Table 2. At higher concentrations of BYK-420 the specific covered the rough support, yielding a thin film with a long
surface area decreased and the pores became larger. In the range modulation of 1 lm period. In Fig. 4D a close-up
case of P10 a hysteresis loop appeared in the sorption– of an a-alumina grain covered by the silica film is shown.
178 V. Boffa et al. / Microporous and Mesoporous Materials 100 (2007) 173–182

Fig. 4. SEM pictures of membranes prepared with different binder concentrations. (A) M02 twice coated: cross section and surface. (B) M02: penetration
of the silica sol inside the pores of the support. (C and D) M06 after the first coating. (E and F) M06 after the second coating.

The picture shows how the grains of the support are oxygen) are injected at the two sides of the membrane.
responsible for the modulation of the silica film. After During the measurement the pores are selectively blocked
the second coating a smoother layer was obtained, as by the condensation of a vapour (cyclohexane). The pore
shown in Fig. 4E. Thickness and morphology of the first size in which capillary condensation can still occur is
and the second coated layer after calcination can be com- expressed as function of the relative cyclohexane pressure
pared in Fig. 4F. In the figure the different layers are by the Kelvin equation [44]:
marked by white bars. The thickness of the first layer is cs V mol 1
inhomogeneous along the membrane surface, depending ln P r ¼  cos h; ð2Þ
RT rK
on whether it covers voids or a-alumina grains. In contrast,
the second layer is about 170 nm thick over the entire sur- where h is normally considered equal to 0. For each relative
face area of the membrane. The surprising difference in pressure all pores with radii smaller than the threshold va-
morphology between the first and second coated silica lay- lue rK are blocked by the condensed vapour. At a relative
ers is noted. The second coated layer looks more compact pressure of 1 all pores of the membrane are filled and gas
and is more regular than the first one. Indeed it is well transport trough the membrane is not possible when no
known from the literature that the morphology of the sup- macroscopically large defects and cracks are present. When
port plays an important role in the development of mor- the vapour pressure is reduced, pores with larger size than
phology of templated silica films [42,43]. rK are emptied and become available for gas transport.
During the experiment the pressure at both sides of the
membrane is kept equal to 1 bar, so that no pressure gradi-
3.4. Permporometry ents are present in the system. rK is the Kelvin radius of the
pore, which does not correspond to the actual pore radius.
Permporometry is an effective technique for determining To obtain the real pore radius it must to be corrected by
the pore size distribution of a mesoporous layer on a mac- adding the thickness of the cyclohexane monolayer, which
roporous support, and to check if defects in the layer are forms on the surface of the pores before capillary conden-
absent. In short, two incondensable gases (nitrogen and sation occurs [45]. This layer is generally called t-layer and
 V. Boffa et al. / Microporous and Mesoporous Materials 100 (2007) 173–182 179

here it is indicated as lt. The lt thickness can be estimated Eq. (4) was used to calculate the pore size distribution of
from the oxygen permeation at low cyclohexane relative M06 in the different stages of its preparation, as shown
pressure and is generally about 0.4 nm thick [45]. In the in Fig. 5B. The distributions were derived assuming a cylin-
case of templated silica, which has pores in the region be- drical pore shape and normalized to unity. The support has
tween truly microporous and mesoporous it was not possi- pores with Kelvin radii larger than 40 nm. After the first
ble to calculate its value. Therefore, in this paragraph we deposition of S06 two different kinds of pores were detected
refer to the Kelvin radius of the pores only. (see Fig. 5B): a distribution of pores with main radius equal
The results of permporometry measurements are to 1.3 nm, which is a typical value for CTAB-templated sil-
reported in Fig. 5A. In order to calculate the pore size dis- icas [26], and larger pores, which may be macropores of the
tribution a few assumptions were made. First of all we support that are only partially occluded by the deposited
assumed that gas diffusion occurred in the mesoporous layer. After the second coating the average pore Kelvin ra-
regime; therefore the transport across the membrane can dius was about 1 nm and no defects were detected. As we
be expressed by the Knudsen equation [46]: already mentioned in Section 2, we were not able to pre-
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pare a defect-free membrane based on just one dip-coated
er 2 dP 8 layer. For this reason a second coating was applied on all
UA ¼  : ð3Þ
s 3 dz pM A RT samples presented here.
As shown in Fig. 5A, M02 and M06 showed similar
This assumption is justified by the fact that in literature permporometry profiles, but M02 was more permeable to
Knudsen flow has been reported for silica membranes that oxygen. Despite the fact that no layer was observed with
had been prepared with CTAB micelles as templates [47]. SEM (see Fig. 4A), in M02 the deposited templated silica
The geometry of the pores also plays a role. Assuming film appeared to form a selective barrier with pores smaller
that all pores are cylindrical and parallel, we can express than 5 nm. Therefore, we may conclude that the membrane
the density function with the following equation: is defect-free. In Fig. 5C the pore size distributions of the
rffiffiffiffiffiffiffiffiffiffi  powder and the membrane obtained from S02 are
3 slAtot RTM DF acc compared. In the case of the powder the distribution was
nðrK Þ ¼  : ð4Þ
4 r3K 2p Dr calculated by nitrogen desorption at 77 K using the BJH

Fig. 5. Permporometry analysis of support and membranes: (A) oxygen permeance as function of the cyclohexane relative pressure; (B) evolution of the
pore size distribution of M06 in the different stages of the membrane preparation, calculated according to the Eq. (4); (C) comparison between the pore
size distributions of powder (N2 desorption, BJH method) and supported film (permporometry, desorption mode) obtained from the sol S02.
180 V. Boffa et al. / Microporous and Mesoporous Materials 100 (2007) 173–182

method. Although pores with radius smaller than 2 nm

cannot be assessed with this method, the assumptions on
which the calculation is based are similar to those that have
been used above for the determination of a distribution
from the permporometry measurement, namely (1) pores
in the mesoporous regime, (2) validity of the modified Kel-
vin equation and (3) cylindrical pore shape. Hence direct
comparison of the two curves, which are derived from
materials obtained from the same sol, is allowed. The aver-
age pore diameter of supported silica M02 was calculated
from the distribution in Fig. 5C and was found to be
1.8 nm. This value is comparable with the one obtained
for P02, which was 2.0 nm. Despite the fact that the main
pore width is almost the same for the two samples, the pore
size distribution of the supported film seems to be broader
than that one of the powder, and it also has a consistent
number of larger pores. This difference can be attributed Fig. 6. Single gas permeance of templated silica membranes M02 and M06
to the lower shrinkage of the supported material during as function of the molecular mass of the probe molecules. The Knudsen
calcination as a result of being clamped to a rigid support. behavior of M02 and M06 is stressed respectively by a continuous line and
Furthermore, the drying rate of thin films is much faster a dotted line.
because of the higher surface/volume ratio and the effect
of solvent drainage by the porous support. The rate of dry-
ing can affect the structure of the sol since longer drying Permselectivities below the Knudsen limit (Eq. (5)) indi-
times provides more time for the sol to condense into a cates a defective mesoporous layer, i.e., a membrane with
more regular mesostructure. a significant contribution of non-selective viscous flow on
At a cyclohexane pressure equal to zero, M10 was more the overall rate of molecular transport. Therefore, gas per-
permeable to oxygen than M02 (Fig. 5A). Nevertheless meation measurements not only give a direct indication of
oxygen permeated through the membrane already at a rel- the performance of a membrane in terms of flux and selec-
ative pressure of about 0.75, which corresponds to pores tivity, but may also provide a proof of either the presence
with a Kelvin radius of about 8 nm. This observation is or absence of macroscopic defects in the silica layer. As re-
in agreement with what was measured by nitrogen sorption ported in Fig. 6, Knudsen-type selectivity was observed for
on P10, a measurement in which pores larger than in the M02 and M06. This means that all support macropores are
other samples were observed. S10 yielded more permeable filled or covered with a templated silica layer, and the layer
membranes, but these are less selective in processes in does not have measurable defects.
which particles/molecules with sizes below 10 nm are A ‘‘defect free’’ templated silica membrane can be
involved. defined as a film with the selectivity of a mesoporous layer.
This definition is often used in membrane technology in
cases in which membrane selectivity is more important than
3.5. Gas permeation
the intrinsic structural quality of the separating layer. Gas
permeation and permporometry, which are methods based
As a general rule the flow of a gas permeating through
on the measurement of membrane performance, have been
macroporous media can be described in terms of viscous
used as a proof for the absence of defects in the top-layer.
flow. Knudsen diffusion is typical for transport of gases
This is certainly true for mesoporous layers obtained by
in mesopores, and for micropores the transport rate is gen-
deposition of colloidal sols, where penetration is negligible
erally described in terms of activated transport [46]. We
and membrane performance is determined by the quality of
suppose that Knudsen transport is dominant in our mate-
the top layer. On the other hand, in the case of membranes
rial; indeed the sorption data in Table 2 show that the pre-
obtained by polymeric sols the penetrated material also
pared materials are mesoporous.
plays a role. The most representative case is membrane
From Eq. (3) it follows that the ideal permselectivity for
M02: although Knudsen separation was observed for this
a binary mixture consisting of gases A and B (where
membrane and its pore size distribution was in the meso-
MB > MA) that are permeating through a mesoporous
porous range, Fig. 4A clearly demonstrates that a top layer
membrane is:
that is as such recognizable is completely absent. Hence,
rffiffiffiffiffiffiffiffi
UA MB gas permeation measurements are not an exhaustive proof
aA=B ¼ ¼ : ð5Þ of the existence of a continuous porous film obtained from
UB MA
a polymeric sol. The high resistance to gas transport of
In the system under consideration here, a mesoporous thin M06 suggests a comparable degree of penetration of silica
film is supported by a macroporous a-alumina support. into the a-alumina support. In the case of M06 some degree
 V. Boffa et al. / Microporous and Mesoporous Materials 100 (2007) 173–182 181

of penetration is visible in Fig. 4C. BYK-420 is an effective References

additive in order to obtain silica membrane on a macropo-
rous a-alumina. Nevertheless it is not possible with this [1] S. Mann, S.L. Burkett, S.A. Davis, C.E. Flower, N.H. Medelson,
method to avoid penetration in the support completely. S.D. Sims, D.W. Walsh, N.T. Whilton, Chem. Mater. 9 (1997) 2300.
[2] V.V. Guliants, M.A. Carreon, Y.S. Lin, J. Membrane Sci. 235 (2004)
In Fig. 4C silica on the a-alumina grains is still visible in 53.
a region deeper than 1 lm. [3] N.K. Raman, M.T. Anderson, C.J. Brinker, Chem. Mater. 8 (1996)
The permeance values reported in Fig. 6. are higher than 1682.
those obtained by Nishiyama et al. [26] with membranes [4] I. Honma, H.S. Zhou, D. Kundu, A. Endo, Adv. Mater. 12 (2000)
that had been prepared by hydrothermal treatment on 1529.
[5] C.T. Kresge, M.E. Leonowicz, W.J. Roth, J.C. Vartulli, J.S. Beck,
commercial alumina plates (NGK Insulators, Ltd.). The Nature 359 (1992) 710.
hydrogen permanence of those membranes was about [6] J.S. Beck, J.C. Vartulli, W.J. Roth, M.E. Leonowicz, C.T. Kresge,
2.7 · 107 mol Pa1 m2 s1 at 295 K, while that of the K.D. Schmitt, C.T.W. Chu, D.U. Olson, E.W. Sheppard, S.B.
asymmetric supports was 1.1 · 104 mol Pa1 m2 s1 at McCullen, J.B. Higgins, J.L. Schlenker, J. Am. Chem. Soc. 114 (1992)
10834.
a temperature of 295 K and an overpressure of 2.7 bar.
[7] C.T. Kresge, M.E. Leonowicz, W.J. Roth, K.D. Schmitt, J.C. Vartuli,
These commercial substrates had on the top a pore size US Patent 5, 198, 203, 1993.
of about 0.1 lm, which is similar to that of the home-made [8] H. Yang, N. Coombs, I. Sokolova, G.A. Ozin, J. Mater. Chem. 7
supports utilized by us. The supports that we used were (1997) 1285.
prepared by colloidal filtration from a suspension of alu- [9] D.M. Antonelli, J.Y. Ying, Curr. Opin. Colloid Interf. Sci. 1 (1996)
mina powder in water. Commercial supports are usually 523.
[10] M. Mulder, Basic Principles of Membrane Technology, second ed.,
asymmetric systems with smaller pores located on the side Kluwer Academic Publisher, Dordrecht, 1996.
where the coating is deposited, and pores of a few microns [11] L.G.A. van de Watera, T. Maschmeyera, Top. Catal. 29 (2004) 67.
width on the opposite side. Commercial supports have gas [12] C.-Y. Tsai, S.-Y. Tam, Y. Lu, C.J. Brinker, J. Membrane Sci. 169
resistances smaller than 105 Pa m2 s mol1 (expressed as (2000) 255.
the reciprocal of permeance), but often a high number of [13] S.R. Chowdhury, R. Schmuhl, K. Keizer, J.E. ten Elshof, D.H.A.
Blank, J. Membrane Sci. 225 (2003) 177.
defects is present on their surface. The symmetric supports [14] S.R. Chowdhury, K. Keizer, J.E. ten Elshof, D.H.A. Blank, Lang-
prepared by us have a ten times larger resistance. On the muir 11 (2004) 4548.
other hand their surface is very well defined. The roughness [15] L. Huang, S. Kawi, K. Hidajat, S.C. Ng, Micropor. Mesopor. Mater.
depends only on the alumina grains, which have dimen- 88 (2006) 254.
[16] M. Ogawa, J. Am. Chem. Soc. 116 (1994) 7941.
sions of a few hundred nanometers. Defects like scratches
[17] P.T. Tanev, T.J. Pinnavia, Chem. Mater. 8 (1996) 2068.
or holes are not detectable by optical analysis. For this [18] J.M. Kim, S.K. Kim, R. Ryoo, Chem. Commun. 2 (1998) 259.
reason we chose them as support for lab testing and preli- [19] A.C. Pierre, Introduction to Sol–Gel Processing, Kluwer Academic
minary studies. The next step of this research is to test this Publisher, Dordrecht, 1998.
procedure on commercial tubular supports, which will [20] A.J. Burggraaf, L. Cot, Fundamentals of Inorganic Membrane
eventually yield membranes with enhanced permeances. Science and Technology, Elsevier, Amsterdam, 1996.
[21] A.F.M. Leenaars, K. Keizer, A.J. Burggraaf, J. Mater. Sci. 19 (1984)
1077.
[22] J. Sekulic, A. Magraso, J.E. ten Elshof, D.H.A. Blank, Micropor.
4. Conclusions
Mesopor. Mater. 79 (2004) 49.
[23] X.S. Ju, P. Huang, N.P. Xu, J. Shi, J. Membrane Sci. 166 (2000) 41.
Templated silica membranes were successfully prepared [24] B.N. Nair, J.W. Elferink, K. Keizer, H. Verweij, J. Sol–Gel Sci.
by direct coating of a thixotropic polymeric sol onto a mac- Technol. 8 (1997) 471.
roporous support. BYK-420 was able to confer a thixotro- [25] S.R. Chowdhury, A.M. Peters, D.H.A. Blank, J.E. ten Elshof, J.
Membrane Sci. 279 (2006) 276.
pic behavior to the sol at relatively low concentrations. As
[26] N. Nishiyama, D. Huy Park, A. Koide, Y. Egashira, K. Ueyama, J.
result the penetration of the sol into the support during the Membrane Sci. 182 (2001) 235.
coating procedure was strongly inhibited. Doubly coated [27] D.H. Park, N. Nishiyama, Y. Egashira, K. Ueyama, Micropor.
membranes showed no indications of defects and pores lar- Mesopor. Mater. 66 (2003) 69.
ger than 5 nm were not detected, while the main pore width [28] N. Nishiyama, H. Saputra, D. Huy Park, Y. Egashira, K. Ueyama, J.
Membrane Sci. 218 (2003) 165.
was about 1.8 nm. Membranes with such pore size distribu-
[29] C. Boissiere, M.U. Martines, A. Larbot, E. Prouzet, J. Membrane Sci.
tion can be applied as molecular sieves in many industrially 251 (2005) 17.
relevant processes. For instance they can be use in order to [30] A. Nijmeijer, H. Kruidhof, R. Bredesen, H. Verweij, J. Am. Ceram
separate homogenous catalysts [48] or drugs [49,50] from Soc. 84 (2001) 136.
their reaction mixtures. [31] P.M. Biesheuvel, H. Verweij, J. Membrane Sci. 156 (1996) 250.
[32] G.-Z. Cao, J. Meijerink, H.W. Brinkman, A.J. Burggraaf, J.
Membrane Sci. 85 (1993) 221.
Acknowledgment [33] G.-Z. Cao, H.W. Brinkman, J. Meijerink, K.J. de Vries, A.J.
Burggraaf, J. Am. Ceram. Soc. 76 (1993) 2201.
[34] R.M. de Vos, H. Verweij, J. Membrane Sci. 143 (1998) 37.
Financial support from the Netherlands Technology [35] A.G. Fredrickson, Principles and Applications of Rheology, Prentice-
Foundation (STW) is gratefully acknowledged. Hall, Englewood Cliffs, N.J, 1964.
182 V. Boffa et al. / Microporous and Mesoporous Materials 100 (2007) 173–182

[36] Handbook of Chemistry and Physics, 55th ed., CRC Press, Cleveland, [44] D.D. DoAdsorption Analysis: Equilibria and Kinetics, vol. 1,
Ohio. Imperial College Press, London, 1998.
[37] M.T. Anderson, J.E. Martin, J.G. Odinek, P.P. Newcomer, Chem. [45] F.P. Cuperus, D. Bargeman, C.A. Smolders, J. Membrane Sci. 71
Mater. 10 (1998) 311. (1992) 57.
[38] Y. Di, X. Meng, S.G. Li, F.S. Xiao, Micropor. Mesopor. Mater. 82 [46] R.R. Bhave, Inorganic Membranes: Synthesis, Characteristics and
(2005) 121. Applications, Van Nostrand Reinhold, New York, 1991.
[39] C. Boissiere, M.A.U. Martines, M. Tokumoto, A. Larbot, E. Prouzet, [47] N. Nishiyama, D.H. Park, Y. Egashira, K. Ueyama, Separ. Purif.
Chem. Mater. 15 (2003) 509. Technol. 32 (2003) 127.
[40] M.T. Anderson, J.E. Martin, J.G. Odinek, P.P. Newcomer, Chem. [48] P.T. Witte, S. Roy Chowdhury, J.E. ten Elshof, D. Sloboda-Rozner,
Mater. 10 (1998) 1490. R. Neumann, P.L. Alsters, Chem. Commun. (2005) 1206.
[41] J.L. Blin, C. Otjacques, G. Herrier, B.-L. Su, Langmuir 16 (2000) [49] D.L. Oatley, B. Cassey, P. Jones, W.R. Bowen, Chem. Eng. Sci. 60
4229. (2005) 1953.
[42] H. Miyata, K.J. Kuroda, Am. Chem. Soc. 121 (1999) 7618. [50] W.R. Bowen, B. Casseyb, P. Jonesc, D.L. Oatleya, J. Membrane Sci.
[43] R. Bandyopadhyaya, E. Nativ-Roth, R. Yerushalmi-Rozen, O. 242 (2004) 211.
Regev, Chem. Mater. 15 (2003) 3619.