Journal of Membrane Science 297 (2007) 5–9

Rapid communication

A hybrid processing method for high performance

hydrogen-selective silica membranes
Suraj Gopalakrishnan a,∗ , Yasushi Yoshino b , Mikihiro Nomura a ,
Balagopal N. Nair b , Shin-Ichi Nakao a
a Department of Chemical System Engineering, University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan
b R&D Center, Noritake Co. Limited, Aichi 470-0293, Japan

Received 8 February 2007; received in revised form 16 March 2007; accepted 17 March 2007
Available online 23 March 2007

Abstract
Development of high performance hydrogen-selective inorganic membrane has become an important requisite for the production of hydrogen,
an energy carrier that could cater the ongoing energy revolution. This paper describes a novel, hybrid method to process high performance hydrogen
perm-selective membranes by the rapid CVD modification of a sol–gel silica layer. The initial N2 permeance values through a membrane were
brought down by four orders of magnitude to 2.74 × 10−10 mol m−2 s−1 Pa−1 within 5 min of CVD, while maintaining the H2 permeance values
as high as 6.43 × 10−7 mol m−2 s−1 Pa−1 (H2 /N2 = 2300). It is likely that the reduction in the CVD zone thickness achieved by the presence of
pre-existing sol–gel silica layer is the reason for the rapid modification of the pore structure to high performance membranes.
© 2007 Elsevier B.V. All rights reserved.

Keywords: Hydrogen separation; CVD; Sol–gel; Membrane reactor; Silica membranes

1. Introduction thermal and chemical stability while offering high permeability

and selectivity for hydrogen [4].
Hydrogen perm-selective membranes have attracted much Several techniques have been reported to process silica mem-
interest in membrane separation applications, particularly for branes, among which sol–gel synthesis [5–10] and chemical
applications in fuel cells and membrane reactors, because vapor deposition (CVD) [11–19] attract most attention. Sol–gel
of the importance of hydrogen as an energy carrier for the method has been shown to precisely control the pore size and
future [1,2]. Membrane reactors for the production of hydro- pore structure; though the retention of unreacted or partially
gen require thermally and mechanically stable membranes reacted materials in the sol–gel matrix have shown to consid-
with excellent hydrogen permeability and perm-selectivity. erably affect the stability of the membranes prepared by this
Metal composite membranes, especially thin palladium-based method. CVD methods derive chemically homogenous deposits,
films/membranes on porous supports, show excellent hydrogen especially inside the mesopores of the supports. The difficulties
permeance and separation. However cost concerns and durabil- in processing multiple tubes in one step and the long durations
ity problems associated with their reaction with sulphur, carbon of chemical vapor deposition required for membrane formation
monoxide and catalyst materials limit their applications [3]. restricted the use of CVD in commercial applications so far.
Membranes made from porous ceramics, such as microporous Moreover, the gas permeance values were generally lower than
silica membranes, offer alternatives to palladium membranes. the sol–gel counter parts, even though the stability and durability
These membranes have long been considered as replacements of the films were considerably better. In this paper we describe,
to the conventional polymeric membranes for their physical, a novel hybrid coating method that could drastically reduce the
membrane processing effort, at the same time achieve superior
∗ performance than the membranes made with the base technolo-
Corresponding author. Tel.: +61 7 3346 7829; fax: +61 7 3365 4199.
E-mail addresses: [email protected] (S. Gopalakrishnan), gies’ when used alone. These new membranes have shown high
[email protected] (B.N. Nair). gas permeance and selectivity.

0376-7388/$ – see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.memsci.2007.03.034
6 S. Gopalakrishnan et al. / Journal of Membrane Science 297 (2007) 5–9

A good membrane is determined by (a) precise control of time and hence the total time required for fabricating high per-
the pore size in the molecular size region and hence having the formance silica membranes, where mostly CVD is the slowest
ability to selectively filter gas and vapor molecules, (b) only processing step involved.
minute amounts of pinholes and cracks and thereby avoiding
non-selective molecular diffusion and (c) minimum membrane 2. Experimental
thickness to maximize the flux of permeating molecules. Pro-
cessing an ideal membrane is in practical difficult due to the Pore size of the commercially available capillary substrates
competing nature of some of these requirements. (NOK, Japan) was brought down to 5 nm range by applying a thin
Generally, sol–gel silica membranes are prepared by dip- ␥-alumina layer. A silica layer was coated over this ␥-alumina
coating a mesoporous substrate (usually ␥-alumina) with a silica layer using sol–gel processing. The alumina and silica sol prepa-
sol made using alkoxide precursor, followed by controlled dry- ration and coating conditions were based on our previous studies
ing and firing at high temperatures between 673 and 1073 K. In [9,21,24]. This process distributed a thin layer of silica net-
most cases, the process has to be repeated several times to ensure work over the ␥-alumina layer with some infiltration into the
a defect-free membrane surface. Here, the membrane quality is ␥-alumina pores. This sol–gel coated substrates were subjected
greatly dependant on the sol-chemistry [15,20,21]. CVD mem- to CVD using tetramethylorthosilicate (TMOS) and O2 as reac-
branes are prepared by the reaction of gas phase precursors inside tants at 873 K. The CVD process parameters were selected and
or around the mesoporous substrate pores. The reaction took modified from our previous studies of counter diffusion CVD in
place inside the substrate pores, the selectivity values reported ␥-alumina layer [19,25,26]. Chemical vapor deposition times of
were much higher than sol–gel membranes, though gas perme- 2 h or more were necessary in those studies to prepare a mem-
ance values were lower. It was observed that a strong dependency brane giving H2 /N2 separation factor of ∼3000. Just as in the
with the preparation conditions (deposition gas composition, previous studies, TMOS was supplied through the shell side and
temperature, pressure, etc.) determines membrane quality and O2 was supplied through the inner side so that the reaction takes
long CVD times were generally necessary to process high qual- place at the point of intersection of the reactants and forms a
ity membranes. Earlier reports suggested a denser structure for deposit in the region. Gas permeance measurements were per-
these membranes compared to sol–gel membranes based on the formed by the dead-end method and perm-selectivity values
high activation energy values measured for the permeance of were calculated from the ratio of individual gas permeance val-
small gas molecules [22,23]. As mentioned, even though the ues. XPS measurements were performed by ULVAC-PHI, Inc.,
H2 selectivity measured with these membrane were quite high, Kanagawa, Japan.
the H2 permeance were lower (∼10−8 mol m−2 s−1 Pa−1 ) when
compared to sol–gel membranes. Fig. 1 displays the relation of
3. Results and discussion
H2 permeability with that of selectivity for various silica mem-
branes at high temperatures as reported in the literature. In order
The gas separation performance of a membrane prepared by
to simultaneously improve the membrane gas permeance and
5 min of CVD is shown in Fig. 2. As shown, H2 and N2 gas
separation performance, we have devised a new hybrid process-
molecules displayed activated diffusion through the membrane.
ing method involving both sol–gel and CVD procedures. This
The activation energy values calculated for H2 and N2 through
method should also assist the large scale processing of mem-
the membrane were 15.56 and 11.86 kJ mol−1 , respectively. This
branes and membrane bundles, by drastically reducing the CVD
data is quite different from sol–gel silica membranes as activa-
tion energy for the poorly permeating N2 molecule is seldom

Fig. 1. H2 permeance and ideal selectivity values of H2 with other gases reported
for silica membranes at high temperatures (>673 K). Numbers represents corre- Fig. 2. Membrane performance after chemical vapor depositing silica for 5 min
sponding references. at 873 K.
 S. Gopalakrishnan et al. / Journal of Membrane Science 297 (2007) 5–9 7

measured or reported for sol–gel silica membranes [21]. We have

previously reported that membranes prepared using the counter
diffusion CVD method for 2 h usually could show very low N2
permeance through them [25–27]. Obviously, the present mem-
brane could be presumed to have a structure formed within a
very short span of CVD time, though the ideal selectivity values
(H2 /N2 = 2300) and hydrogen permeance values achieved were
in par with other reported CVD silica membranes and sol–gel
silica membranes, respectively (see Fig. 1).
The initial N2 permeance through the silica membrane layer,
before CVD reaction, was rather similar to that through the ␥-
alumina layer without silica sol–gel coating (1.17 × 10−6 and
1.99 × 10−6 mol m−2 s−1 Pa−1 , respectively, at 873 K). These
values also did not vary much with feed gas pressure, suggesting
a defect-less substrate structure. The similarity in permeance val-
ues indicated that the sol–gel silica coating failed to completely
block the ␥-alumina pore structure or form a separation layer
under the experimental conditions employed. Rather, it formed
a loose network of silica particles over and into the ␥-alumina
pore structure.
The structural features of the sol–gel layer and the changes
occurred in the membrane within the 5 min of CVD could be
explained as in scheme given in Fig. 3. The silica sol formed a
thin web-like network of silica particles around the ␥-alumina
pores and pore-mouth. TMOS-O2 CVD reaction deposited silica
in this sol–gel silica pre-filled region, effectively concentrating
the active reaction zone within this thin layer. Based on measured
thickness values of similar ␥-alumina membranes (2–5 ␮m) and
reported values of silica membrane layer thickness (10–50 nm)
[9], it could be assumed that the thickness of the reaction zone
could be reduced by a factor of 10–100 by this procedure. Five
minutes of reaction should be sufficient for depositing enough
silica to cover the already modified ␥-alumina pore surface in
this thin region.
The CVD reaction time on this membrane was further
increased up to 2 h to evaluate the change in membrane per-
formance. The H2 /N2 separation values almost doubled to 5000.
But, even after 2 h of CVD reaction, N2 gas maintained activated
transport through this membrane; though with slightly lower
activation energy of 8.54 kJ mol−1 compared to the initial case.
The variation in membrane performance with increased CVD
time is given in Table 1. Increase in the H2 /N2 selectivity values
together with the EaH2 values are typical of changes in mem-
brane structure with two types of pores, as in the case of sol–gel
membranes [21], where the relative flux through the smaller
pores become increasingly dominant. In the present case, the
membrane structure transformed from a discontinuous network
of sol–gel silica particles to a continuous and selective barrier Fig. 3. Membrane formation: (a) ␥-alumina layer; (b) silica sol forms a network
within the first 5 min of CVD. With additional CVD, the mem- over and into the ␥-alumina layer; (c) TMOS and oxygen are supplied through
brane structure improved further with decreased non-selective either sides of the ␥-alumina layer where they react at the intersection region or
sol–gel silica pre-filled region (silica coating effectively reduces the thickness of
diffusion, denoting a gradual transformation to a fully compact
the zone of CVD reaction); (d) continuing deposition completes the membrane
layer of silica membrane. architecture.
XPS analyses of the silica sol-coated ␥-alumina samples,
before and after CVD reaction have shown that the modifica-
tion of layer (deposition of CVD silica) occurred mostly within
the existing silica infiltrated region of ␥-alumina layer (see
supplementary information attached at the end of the paper for
8 S. Gopalakrishnan et al. / Journal of Membrane Science 297 (2007) 5–9

Table 1
Single gas permeance values of H2 and N2 through the silica membrane at 600 ◦ C
CVD reaction H2 (×10−7 mol m−2 s−1 Pa−1 ) N2 (×10−11 mol m−2 s−1 Pa−1 ) H2 /N2 (ratio) EaH2 (kJ/mol) EaN2 (kJ/mol)
time (min)

120 3.27 6.51 5027 19.40 8.54

30 3.02 7.62 3970 18.45 12.31
5 6.43 27.4 2343 15.36 11.88

Activation energy values were calculated for the temperature range 100–600 ◦ C. Ideal selectivity values were calculated from the single gas permeance data.

the XPS plots). It was found that, the Si 2p peak taken from The H2 /N2 separation values almost doubled to 5000, after 2 h
the surface of sol–gel silica layer remained intact even after of CVD reaction. Permeance and separation results coupled with
deposition, while the peaks collected after sputtering-out 100 nm XPS analysis have shown that, a reduction in vapor phase deposi-
have grown significantly. Si 2p collected from 200 nm depth has tion zone area (thickness) by the pre-existing silica sol–gel layer
shown marginal growth. This showed that most of the CVD has effectively contributed to the quick membrane formation.
reaction took place inside the ␥-alumina layer, within the fist
few 100 nm’s range from the surface, which was also the region Acknowledgement
where silica sol particles presumed to have infiltrated into the
alumina layer. This result supports the previous hypothesis that This work was performed as a part of the R&D Project for
the vapor phase deposition process zone area (thickness) was High Efficiency Hydrogen Production/Separation System using
restricted by the silica sol–gel layer and has effectively con- Ceramic Membranes funded by NEDO, Japan.
tributed to the quick membrane formation. The high values of
hydrogen permeance measured with these membranes also sup- Appendix A. Supplementary data
ported the fact that silica deposition due to CVD occurred only
in a very thin layer. Thus this novel hybrid processing method Supplementary data associated with this article can be found,
could effectively be utilized for preparing membranes with much in the online version, at doi:10.1016/j.memsci.2007.03.034.
better performances than the conventional sol–gel or CVD silica
membranes in a rapid manner. References
Based on the XPS data and the permeance through the mem-
brane, the structural formation process could be assumed as [1] T.N. Veziroglu, Hydrogen energy system as a permanent solution to global
follows: during CVD, silica particles derived from the vapor energy—environmental problems, Chem. Ind. 53 (1999) 383.
phase will get deposited as particles/islands, whose number den- [2] S. Ahmed, M. Krumplet, Hydrogen from hydrocarbon fuels for fuel cells,
sity will grow with deposition time, thus forming a layer of Int. J. Hydrogen Energy 26 (2001) 291.
[3] R. Bredesen, K. Jordal, O. Bolland, High-temperature membranes in power
percolating silica particles along with the sol–gel deposited silica generation with CO2 capture, Chem. Eng. Process. 43 (2004) 1129.
particles/islands. The presence of sol–gel layer accelerates the [4] A.J. Burgraaf, L. Cot, Fundamentals of Inorganic Membrane Science and
process of continuous layer formation by restricting the thick- Technology, Elsevier, Amsterdam, 1996.
ness of active CVD zone to the sol–gel silica modified region of [5] J.P. Collins, R.W. Schwartz, R. Sehgal, T.L. Ward, C.J. Brinker, G.P. Hagen,
␥-alumina membrane. The resultant membrane would be much C.A. Udovich, Catalytic dehydrogenation of propane in hydrogen perms-
elective membrane reactors, Ind. Eng. Chem. Res. 35 (1996) 4398.
thinner, thus reducing the membrane formation time as well as [6] Z. Li, K. Kusakabe, S. Morooka, Pore structure and permeance of amor-
increasing gas permeance through it. Previous studies with big- phous Si–C–O membranes with high durability at elevated temperature,
ger pore sized ␥-alumina membrane also supports this theory as Sep. Sci. Technol. 32 (1997) 1233.
the gas permeance after CVD reportedly decreased with increase [7] M. Asaeda, M. Kashimoto, Sol–gel derived silica membranes for separa-
in the pore size of the substrate [27]. tion of hydrogen at high temperature-separation performance and stability
against steam, in: Proceedings of the International Conference on Inorganic
Recently, one of our groups have also scaled-up the method Membranes, Nagoya, Japan, 1998, pp. 172–175.
to make multiple membrane tube bundles in one-step and hence [8] K. Yoshida, Y. Hirano, H. Fuji, T. Tsuru, M. Asaeda, Hydrothermal stability
proved its potential as a technology ready for commercial appli- and performance of silica–zirconia membranes for hydrogen separation in
cations. hydrothermal conditions, J. Chem. Eng. Jpn. 34 (2001) 523.
[9] Y. Yoshino, T. Suzuki, B.N. Nair, H. Taguchi, N. Itoh, Development of
tubular substrates, silica based membranes and membrane modules for
4. Conclusions hydrogen separation at high temperature, J. Membr. Sci. 267 (2005) 8.
[10] T. Tsuru, K. Yamaguchi, T. Yoshioka, M. Asaeda, Methane steam reforming
by microporous catalytic membrane reactors, AIChE J. 50 (2004) 2794.
Hydrogen perm-selective membranes were prepared using a [11] J.C.S. Wu, H. Sabol, G.W. Smith, D.L. Flowers, P.K.T. Liu, Character-
hybrid process involving the chemical vapor deposition of silica ization of hydrogen-permselective microporous ceramic membranes, J.
on a sol–gel silica layer coated alumina substrates. The initial Membr. Sci. 96 (1994) 275.
N2 permeance values through the membrane were brought down [12] S.S. Kim, B.K. Sea, Gas permeance characteristics of silica/alumina com-
posite membranes prepared by chemical vapour deposition, Korean J.
by four orders of magnitude to 2.74 × 10−10 mol m−2 s−1 Pa−1 Chem. Eng. 18 (2001) 322.
within 5 min of CVD, while maintaining the H2 permeance val- [13] B.-K. Sea, M. Watanabe, K. Kusakabe, S. Morooka, S.-S. Kim, Forma-
ues as high as 6.43 × 10−7 mol m−2 s−1 Pa−1 (H2 /N2 = 2300). tion of hydrogen permselective silica membrane for elevated temperature
 S. Gopalakrishnan et al. / Journal of Membrane Science 297 (2007) 5–9 9

hydrogen recovery from a mixture containing steam, Gas Sep. Purif. 10 [21] B.N. Nair, K. Keizer, T. Okubo, S.-I. Nakao, Evolution of pore structure
(1996) 187. in microporous silica membranes: sol–gel procedures and strategies, Adv.
[14] S. Morooka, S.S. Kim, S. Yan, K. Kusakabe, M. Watanabe, Separation of Mater. 10 (1998) 249.
hydrogen from an H2 –H2 O–HBr system with an SiO2 membrane formed [22] G.R. Gavalas, C.E. Megris, S.W. Nam, Deposition of H2 -permselective
in macropores of an ␣-alumina support tube, Int. J. Hydrogen Energy 21 SiO2 films, Chem. Eng. Sci. 44 (1989) 1829.
(1996) 183. [23] S.-I. Nakao, T. Suzuki, T. Sugawara, T. Tsuru, S. Kimura, Preparation
[15] S. Morooka, S. Yan, K. Kusakabe, Y. Akiyama, Formation of hydrogen- of microporous membranes by TEOS/O3 CVD in the opposing reactants
permselective SiO2 membrane in macropores of ␣-alumina support tube geometry, Microporous Mesoporous Mater. 37 (2000) 145.
by thermal decomposition of TEOS, J. Membr. Sci. 101 (1995) 89. [24] B.N. Nair, K. Keizer, H. Verweij, A.J. Burggraaf, Surface selective
[16] T. Ionannides, X.E. Verykios, Application of a dense silica membrane reac- separation of propylene to propane, Sep. Sci. Technol. 31 (1996)
tor in the reactions of dry reforming and partial oxidation of methane, Catal. 1907.
Lett. 36 (1996) 165. [25] G. Suraj, K. Ono, M. Nomura, T. Sugawara, S.-I. Nakao, Microporous silica
[17] M. Tsapatsis, G. Gavalas, Structure and aging characteristics of H2 - membrane prepared using TMOS/O3 CVD in opposing reactants geometry,
permselective SiO2 -Vycor membranes, J. Member. Sci. 87 (1994) 281. Trans. Mater. Res. Soc. Jpn. 29 (2004) 3263.
[18] A.K. Prabhu, S.T. Oyama, Highly hydrogen selective ceramic membranes: [26] S. Gopalakrishnan, B.N. Nair, S.-I. Nakao, High performance hydro-
application to the transformation of greenhouse gases, J. Membr. Sci. 176 gen selective membranes prepared using rapid hybrid processing
(2000) 233. method, in: Proceedings of the International Conference on Nanoscience
[19] M. Nomura, K. Ono, S. Gopalakrishnan, T. Sugawara, S.-I. Nakao, Prepa- and Nanotechnology (ICONN 2006), Brisbane, Australia, July 3–7,
ration of a stable silica membrane by a counter diffusion chemical vapor 2006.
deposition method, J. Membr. Sci. 251 (2005) 151. [27] M. Nomura, H. Aida, S. Gopalakrishnan, T. Sugawara, S. Nakao, S.
[20] B.N. Nair, W.J. Elferink, K. Keizer, H. Verweij, Preparation and structure Yamasaki, T. Inada, Y. Iwamoto, Steam stability of a silica membrane pre-
of microporous silica membranes, J. Sol-Gel Sci. Technol. 8 (1–3) (1997) pared by counter diffusion chemical vapour deposition, Desalination 193
471. (2006) 1.