Journal of Membrane Science 207 (2002) 73–89

Alumina and titania multilayer membranes for nanofiltration:

preparation, characterization and chemical stability
Tim Van Gestel a,b,∗ , Carlo Vandecasteele a , Anita Buekenhoudt b , Chris Dotremont b ,
Jan Luyten b , Roger Leysen b , Bart Van der Bruggen a , Guido Maes c
a K.U. Leuven, Department of Chemical Engineering, W. de Croylaan 46, B-3001 Heverlee, Belgium
b Vito (Flemish Institute for Technological Research), Process Technology, Boeretang 200, B-2400 Mol, Belgium
c K.U. Leuven, Department of Chemistry, Celestijnenlaan 200, B-3001 Heverlee, Belgium

Received 2 July 2001; received in revised form 11 February 2002; accepted 12 February 2002

Abstract
The preparation and characterization of porous ceramic multilayer nanofiltration (NF) membranes is described. During
preparation, special care was given to each sub-layer that forms a part of the multilayer configuration: the macroporous
substrate, the membrane interlayers and the NF toplayers. High-quality macroporous supports are prepared from ␣-Al2 O3 .
Three types of colloidal sol–gel derived mesoporous interlayers are considered: Al2 O3 , TiO2 and mixed Al2 O3 –TiO2 . The
active NF toplayer is a very thin and fine textured polymeric TiO2 toplayer.
Optimized ␣-Al2 O3 /␥-Al2 O3 /anatase and ␣-Al2 O3 /anatase/anatase multilayer configurations show high retentions for
relatively small organic molecules (molecular weight cut-off <200). Different membrane layers have very different crystal-
lographic properties resulting in a considerably different chemical stability. Corrosion measurements showed that application
of a multilayer configuration including weakly crystallized ␥-Al2 O3 layers is restricted to mild aqueous media (pH 3–11)
or non-aqueous media (organic solvents). For NF applications in aqueous media with a lower or higher pH, the multilayer
membrane composed of anatase on a ␣-Al2 O3 support is to be preferred. © 2002 Elsevier Science B.V. All rights reserved.
Keywords: Ceramic membranes; Sol–gel; Characterization; Chemical stability; Nanofiltration

1. Introduction merous applications in the area of process, waste and

Nanofiltration (NF) membranes can be gener-
ally classified into two major groups according to
their material properties: organic polymeric mem-
branes and inorganic ceramic membranes. Polymeric
NF-membranes are today commercially available in a
large number of different polymeric materials (poly-
sulfone, cellulose acetate, polyamide, . . . ). Membrane
filtration based on such membranes has found nu-
∗ Corresponding author. Tel.: +32-1433-5615;
fax: +32-1432-1186.
E-mail address: vgestelt@vito.be (T. Van Gestel).
drinking water treatment. However, the major draw-
back is that the chemical stability of the conventional
polymeric membranes is limited with respect to cor-
rosive media like strong acids and organic solvents
[1,2]. Therefore, much research to date focuses on
ceramic NF-membranes. Since it is generally recog-
nized that inorganic materials are inherently more
stable than polymers, they are expected to constitute
a promising alternative for polymeric membranes [3].
Ceramic NF-membranes with high performance
parameters such as low cut-off values or high fluxes
can only be obtained in an asymmetric multilayer
configuration. The development of such a multilayer

0376-7388/02/$ – see front matter © 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 3 7 6 - 7 3 8 8 ( 0 2 ) 0 0 0 5 3 - 4
74 T. Van Gestel et al. / Journal of Membrane Science 207 (2002) 73–89
configuration includes: shaping of an appropriate sup-
port material; formation of mesoporous interlayers;
and synthesis of a microporous toplayer with a cut-off
value below 1000. Alumina, titania, zirconia or silica
are considered as the main ceramic materials for the
formation of the multilayer structures [4–6].
The preparation route starts with the production
of a high-quality support system, as the effectiveness
of the developed NF membrane is greatly influenced
by the structural properties of the membrane support.
Based on earlier studies in our group, alumina sup-
ports produced by conventional slip-casting and reac-
tion bonded Al2 O3 (RBAO) manufacturing routes are
used as substrate material for the development of a
multilayer membrane [7–9].
The second part deals with the synthesis of several
intermediate membrane layers. The aim of this modifi-
cation is twofold: fine-grained mesoporous membrane
layers are necessary in order to reduce the coarse pore
structure of the support material and large surface
irregularities or defects can only be covered properly
by a multiple dip-coating procedure. In this study,
colloidal sol–gel derived membrane layers consisting
of alumina or titania are considered. Firstly, ␥-Al2 O3
and anatase/TiO2 membrane interlayers were pre-
pared based on earlier studies of Burggraaf et al. [10],
Zaspalis et al. [11], Kumar et al. [12] and Lin et al.
[13]. Additionally, different combined Al2 O3 –TiO2
membrane interlayers were developed in order to
combine the specific properties of both materials,
namely the possibility of obtaining membrane layers
with a high thickness and without defects on one hand
and a good chemical stability, on the other hand.
The final step in the multilayer preparation route
involves the synthesis of a thin separation toplayer.
Formation of such a layer with very small pores re-
quires an appropriate sol consisting of very small
nanometer-sized particles, which is typically obtained
by the so-called ‘polymeric’ sol–gel technique. In
this work, TiO2 NF toplayers are formed by a simi-
lar sol–gel procedure as previously described by de
Lange et al. [14] and Keizer et al. [15]. They syn-
thesized microporous amorphous SiO2 toplayers for
the modification of mesoporous ␥-Al2 O3 membrane
layers in order to obtain a multilayer gas separation
membrane. Recently, Voigt et al. [16,17] prepared
also microporous toplayers, consisting of amor-
phous TiO2 , on anatase/TiO2 membrane layers and
proved their feasibility for NF applications (MWCO
<500).
All the authors mentioned mainly considered on
the structural properties of the developed membrane
materials. Therefore, only limited data on the chemi-
cal stability of meso- or microporous membranes are
available. Züter et al. [18] reported acid corrosion
tests on mesoporous ␥-Al2 O3 , anatase/TiO2 and ZrO2
membranes. In our group, Schaep et al. [19] tested the
corrosion properties of mesoporous ␥-Al2 O3 mem-
branes. These contributions have however mainly
a descriptive character and the membrane stability
was not investigated systematically by taking into
account important parameters like the structural char-
acteristics of the ceramic material (phase behaviour,
crystallinity, surface area). To our knowledge, only
Hollstein et al. [20] addressed in detail the corrosion
properties of ceramic materials in aqueous media, but
did not describe the stability of porous materials like
sol–gel membrane materials.
The aim of this paper is to provide results on the
chemical stability of different sol–gel derived mem-
brane materials. The corrosion behaviour of meso-
porous interlayers and microporous toplayers, along
with the membrane characterization in terms of pore
size, surface area, pore volume and crystallite size is
described. For the determination of the corrosion rate,
preference was given to simple static corrosion exper-
iments whereby the amount of dissolved membrane
material was determined in acid or alkaline solutions
with a pH ranging from 1 to 13. With the experimen-
tal information, materials could be selected specifi-
cally for each membrane layer in order to develop an
optimal multilayer configuration for particular pro-
cess conditions (non-aqueous solvent, acid, alkali).
The performance characteristics of the multilayer
membranes are also described to indicate possible
application fields of the developed NF-membranes.
2. Experimental
2.1. Membrane preparation
2.1.1. Preparation of support materials
For the preparation of the membrane support, two
routes were followed. In the first route, Al2 O3 supports
were prepared by a slip-casting technique, starting
 T. Van Gestel et al. / Journal of Membrane Science 207 (2002) 73–89
from an aqueous suspension of commercially avail-
able Al2 O3 powder (AKP-30, Sumitomo) with organic
additives (Vanisperse CNH, Borregaard) [7]. In the
second route, Al2 O3 supports were prepared by the
reaction bonded Al2 O3 (RBAO) manufacturing pro-
cess [8]. In this process Al2 O3 powder (HC Starck)
and Al particles (Baudier) are milled in a weight ra-
tio of 40/60. After shaping of the support material by
isostatic pressing, oxidation is performed at 1100 ◦ C,
followed by a final thermal treatment at 1300 ◦ C.
According to both preparation methods, either
disk-shaped or tubular membrane supports can be
produced. The final support disks were 25 mm in di-
ameter and 2 mm in thickness. The tubular supports
were 120 mm in length with an i.d. of 7 mm and an
o.d. of 10 mm.
2.1.2. Preparation of the membrane interlayers by
the colloidal sol–gel process
Precursors for the colloidal sol–gel process were
common aluminium and titanium based organometal-
lic compounds, which were hydrolyzed by the addition
of an excess H2 O. The ␥-AlOOH sol was prepared
by hydrolysis of Al-tri-sec-butoxide (Al(OC4 H9 )3 )
(ASB, Merck). The sol was peptized at 90 ◦ C by addi-
tion of 0.07 mol HNO3 per mole ASB. The TiO2 sol
was obtained by hydrolysis of Ti-tetra-isopropoxide
(Ti(OC3 H7 )4 ) (TTI, Janssen Chimica). Peptization
was obtained at 50 ◦ C with 0.5 mol of HNO3 per mole
of TTI. Binary alumina/titania sols were prepared
using two different sol–gel routes. The first route in-
volved mixing of two separately peptized ␥-AlOOH
and TiO2 sols in a certain ratio. In the second method,
homogeneous mixed sols were produced in one step
by co-hydrolysis of ASB–TTI precursor mixtures
followed by peptization with HNO3 .
For alumina and mixed membranes, a solution
of polyvinyl alcohol (PVA, Erkol) with an aver-
age molecular weight of 72,000 (3.5 g/100 ml H2 O)
was added to the sols before membrane formation
as drying chemical controlling additive (DCCA).
For the preparation of titania membranes, a solu-
tion of hydroxypropyl cellulose (HPC, Aldrich) with
MW = 100,000 (1 g/100 ml H2 O) and PVA with
MW = 72,000 was used (0.1 g/100 ml H2 O).
Supported gel layers were formed on the mem-
brane support by dip-coating the supports with
freshly prepared sols. Non-supported gel layers were
75
produced by pouring the remaining sol–gel solutions
in petri-dishes. In both cases, the obtained gel layers
were dried at room temperature for 24 h. Finally, the
membranes were calcinated for 1 h at 400 ◦ C in a
temperature programmable furnace (Scandia SK-15).
The heating and cooling rates for Al2 O3 , TiO2 and
mixed membranes were 60, 10 and 15 ◦ C/h, respec-
tively. Then, depending on the type of membrane
material, the first calcination step was followed by a
thermal treatment for 1 h at a temperature between
400 and 1200 ◦ C.
2.1.3. Preparation of the membrane toplayers
by the polymeric sol–gel process
A polymeric TiO2 sol was produced starting from
the common TTI precursor, which was partially hy-
drolyzed with a less than equivalent amount of water
in order to obtain a precipitate free TiO2 sol. In con-
trast with the colloidal route, for the preparation of the
very fine polymeric membrane layer no high molecular
weight organic additives could be added. Therefore,
the freshly prepared titania sol was aged for 3 days in
order to increase the viscosity of the dipping solution
and to prevent penetration of the sol into the pores of
the intermediate support layer during dip-coating. At
the same time, unsupported membranes were prepared
by evaporating the sol in a petri-dish. After drying at
room temperature, thermal treatment was performed
for 1 h with heating and cooling rates of 15 ◦ C/h in the
temperature range 200–700 ◦ C.
2.2. Characterization
2.2.1. Static characterization
Several methods were used for the characterization
of the membrane support. The structure was ana-
lyzed by X-ray diffraction (XRD, Siemens D500) and
Hg-porosimetry (Autoscan-33, Porosimeter, Quan-
tachrome). The pore size and the presence of possible
defects in the supports were determined with field
emission scanning electron microscopy (JEOL JSM
6340 F). The surface roughness was measured with
a laser profilometer (Micro Focus Measurement Sys-
tem, UBM). The mechanical properties were charac-
terized by four-point bend tests (Chevron-notch).
The particle sizes in ␥-AlOOH, TiO2 and mixed sols
were determined by the dynamic laser beam scattering
technique (PCS, N4 plus Coulter).
76 T. Van Gestel et al. / Journal of Membrane Science 207 (2002) 73–89
The morphology, surface quality and thickness
of intermediate and toplayer membranes was ex-
amined with optical microscopy (Zeiss) and field
emission scanning electron microscopy (JEOL JSM
6340 F). The structural properties of the membrane
materials were characterized by X-ray diffrac-
tion, N2 -adsorption/desorption (Autosorb 1, Quan-
tachrome) and thermogravimetric analysis (TGA,
Setaram). XRD measurements were used for the de-
termination of the phase composition and the average
crystallite size in the membrane material. The crys-
tallite size was determined with the Debye–Scherrer
equation by using the full-width at half-maximum
value (FWHM) of the main diffraction peaks.
N2 -adsorption/desorption measurements were per-
formed to determine the pore size distribution, BET
surface area and pore volume. All structural measure-
ments were carried out on unsupported membrane in-
terlayers and toplayers, assuming that their properties
were similar to those of supported membrane layers.
2.2.2. Static corrosion tests
Static corrosion tests were performed in aque-
ous solutions at a pH ranging from 1 to 3. These
solutions were prepared using demineralized water
(MilliQ, Millipore) and nitric acid (Merck, pro anal-
ysis ([Al] = 0.05 ppm, [Ti] = 0.02 ppm)). Then,
100 mg calcinated membrane material was immersed
in 100 ml of the acid solution. The solution was shaken
continuously for 4 days on a shaking table (Gerhardt
laboshake). After the exposure time, liquid samples
were taken, filtered over a Whatman 42 filter and ana-
lyzed for their Al and Ti concentration by inductively
coupled plasma-mass spectroscopy (VG-PlasmaQuad
PQ-2 Plus). Additionally, corrosion experiments were
also conducted in NaOH solutions (Merck, pro anal-
ysis (Al content = 0.0005%)) at a pH ranging from
11 to 13. All corrosion experiments were performed
at ambient temperature. The pH measurements were
carried out with a digital pH-meter (ATI Orion 310).
Table 1
Characteristics of the support materials
Type Thermal Phase Pore size
treatment (◦ C) (␮m)
AKP-30 1130 ␣-Al2 O3 0.1
RBAO 1300 ␣-Al2 O3 0.2
2.2.3. Dynamic characterization
Filtration experiments were carried out with a
commercial laboratory scale installation (Amafilter
PSS-2TZ). The molecular weight cut-off (MWCO)
of the membranes was determined by using a solu-
tion containing polyethylene glycols (PEG, Merck)
of different molecular weight: PEG 200, PEG 600,
1500, 5600, 10000 and 15000. To determine the PEG
retention as a function of the molecular weight, feed
and permeate samples were taken and analyzed by
gel permeation chromatography (GPC, Waters, ultra-
hydrogel). During the filtration experiments, the flux
was measured in order to determine the membrane
permeability. All the experiments were carried out at
a transmembrane pressure of 5 bar.
3. Results and discussion
3.1. Support materials
For the development of high-quality supports the
following properties are of major importance: pore
size distribution; porosity; surface quality with the ab-
sence of large defects or large pores; mechanical and
chemical stability.
Structural characteristics of the final Al2 O3 mem-
brane supports prepared according to the two produc-
tion routes are summarized in Table 1. Basically, the
support material consisted of ␣-Al2 O3 . The pore size
depended on the preparation method: the fine-grained
AKP-30 support had an average pore size around
0.1 ␮m, while the RBAO support contained larger
grains and showed also a larger pore size. The poros-
ity was typically ca. 40% and surface roughness was
ca. 0.3 ␮m. Based on four-point bend tests, it is shown
that the RBAO derived support has the advantage of
being mechanically much stronger than the conven-
tional AKP-30 support. According to Luyten et al.,
the higher mechanical stability of the RBAO material
Porosity (%) Roughness (␮m) Bend-strength Corrosion pH 1
(mPa) (␮g Al/l)
40 0.3 40 571
40 0.3 100 –
 T. Van Gestel et al. / Journal of Membrane Science 207 (2002) 73–89
is due to the presence of fine ␣-Al2 O3 grains—formed
during the oxidation of the Al powder fraction—in
the large grained primary ␣-Al2 O3 structure [8]. Long
term corrosion measurements on ␣-Al2 O3 support ma-
terial in nitric acid solutions showed also a high chem-
ical stability. For a corrosion test at pH 1 only small
amounts of aluminium were found even after an ex-
posure time of 6 months.
3.2. Sols
Both the colloidal and polymeric sols were prepared
by hydrolysis of alkoxide precursors to induce con-
densation and inorganic polymerization reactions. In
the different routes, the essential parameter to control
was the degree of hydrolysis.
Al(OC4 H9 )3 + xH2 O → Al(OC4 H9 )3−x (OH)x
+xC4 H9 OH
Ti(OC3 H7 )4 + xH2 O → Ti(OC3 H7 )4−x (OH)x
+xC3 H7 OH
For the formation of mesoporous membrane inter-
layers, suspensions of relatively large nanometer-sized
particles were aimed at. This was obtained by react-
ing the alkoxide precursors with a large excess water
([H2 O]/[ASB] > 3, [H2 O]/[TTI] > 4) followed by a
refluxing step overnight under the conditions shown
in Table 2.
The final products after peptization were stable fine
colloidal ␥-AlOOH, TiO2 and mixed sols. According
to PCS measurements, the ␥-AlOOH sols contained
plate-shaped particles with an average diameter of
35 nm. Stable TiO2 sols consisted of cylindrical par-
ticles with a mean size of about 37 nm. Due to the
Table 2
Experimental conditions for the preparation of stable sols
Stable sols [H2 O]/[M]a [HNO3 ]/[M]a
␥-AlOOH 100 0.07
TiO2 b 25 0.50
␥-AlOOH–TiO2
Mixed 150 0.50
TiO2 c 1 80
a [M] = [ASB] or [TTI].
b Colloidal sol.
c Polymeric sol.
77
presence of relatively large particles these sols were
semi-transparent and looked light blue. Binary sols,
obtained by mixing separately peptized ␥-AlOOH
and TiO2 sols, showed both typical characteristics of
the individual sols, namely a light blue colour and
an average particle size distribution around 40 nm.
Stable mixed sols, prepared by co-hydrolysis of the
precursors, were completely transparent (colourless)
after peptization. In this case smaller sol particle sizes
were obtained with a mean diameter of ca. 25 nm.
For the synthesis of the TiO2 toplayer, a precipitate-
free sol containing small nanometer-sized polymeric
structures (fractals) was prepared. The essential
features of the polymeric sol preparation were a care-
fully controlled hydrolysis with a less than stoechio-
metric amount of water ([H2 O]/[Ti] < 4) followed by
an appropriate aging. The final aged polymeric TiO2
sol was completely transparent. Average fractal sizes
were smaller than 5 nm, as no particle sizes above
5 nm were observed by laser scattering experiments.
3.3. Membrane interlayers
3.3.1. Morphology
For the development of high-quality multilayer
NF-membranes, we aimed at the formation of rather
thick intermediate layers, which were required to
overcome inevitable surface irregularities of the sup-
port.
Alumina layers could easily be produced with an
average thickness above 2 ␮m, thus properly smooth-
ing the roughness of the support [8]. Titania layers
obtained by a single dip-coating procedure showed an
average thickness of ca. 0.5 ␮m. Therefore, a second
dipping procedure had to be used to increase the mem-
brane thickness up to ca. 1 ␮m (Fig. 1a). By applying
pH Temperature (◦ C) Particle size (nm) Characteristics
4 90 35 Opaque blue
1 80 37 Opaque blue
2 40 Opaque blue
1 85 25 Transparent
<5 Transparent
78 T. Van Gestel et al. / Journal of Membrane Science 207 (2002) 73–89

Fig. 1. FESEM cross-sections (20,000×) pictures of titania (a) and mixed (b) colloidal sol–gel membrane layers on ␣-Al2 O3 (a) and
RBAO (b) supports.

an Al2 O3 –TiO2 combination as intermediate layer a
thickness of ca. 1 ␮m was obtained (Fig. 1b).
In practice, in all the membrane formation proce-
dures a second dip-coating step was used after calcina-
tion of the first layer. This was done to repair defects
such as pinholes or regions with non-uniform layer
formation, which often appeared in the first sol–gel
intermediate layer. Each final support thus produced
consisted of the main support with two sol–gel derived
membrane layers.
3.3.2. Structural properties
3.3.2.1. Standard thermal treatment. After thermal
treatment at 400 ◦ C, alumina and titania membranes
appeared in the crystallographic phases ␥-AlOOH
(boehmite) and anatase. X-ray diffraction patterns
of binary alumina/titania membranes prepared by
mixing separately peptized sols, show also typical
␥-AlOOH and anatase diffraction peaks (Fig. 2). It
may be concluded that this route did not yield a new
membrane material, but rather resulted in a binary
␥-AlOOH/anatase composite membrane material.
Pore structure characteristics of this membrane were
comparable with those of ␥-AlOOH membranes. This
suggested that the final calcinated membrane con-
sisted of a ␥-alumina-like structure with titania grains
incorporated [11]. Pore size characterization showed
in each case a narrow pore size distribution in the
3–4 nm range. The shape of the adsorption/desorption

Fig. 2. XRD patterns of binary ␥-AlOOH/anatase composite membrane material.

 T. Van Gestel et al. / Journal of Membrane Science 207 (2002) 73–89 79

Table 3
Structural characteristics of colloidal alumina, titania and mixed membranes
Phase Firing (◦ C) Mean pore size (nm) Surface area (m2 g−1 ) Pore volume (ml g−1 ) Isotherm (type)

␥-AlOOH 400 3.7 318 0.27 IV

␥-Al2 O3 600 3.8 239 0.27 IV
␥-Al2 O3 800 4.6 154 0.22 IV
␦, ␪-Al2 O3 1000 6.3 87 0.18 IV
␣-Al2 O3 1200 9.2 10 0.09 II
Anatase 400 3.4 136 0.14 IV
Anatase 500 4.8 68 0.11 IV
Anatase/rutile 600 7.8 28 0.05 II, IV
Rutile 700 12.0 12 0.04 II
␥-AlOOH/anatasea 400 3.6 294 0.26 IV
Amorphousb 400 1.6 318 0.17 I
Amorphous/anataseb 800 4.8 120 0.14 IV
Amorphous/anataseb 900 7.6 53 0.14 IV
␣-Al2 O3 /anataseb 920 7.6 63 0.15 IV
␣-Al2 O3 /anatase/rutileb 950 8.0 13 0.07 IV, II
␣-Al2 O3 /anatase/rutileb 1000 8.9 5 0.05 II
a Composite membrane material.
b Mixed membrane material.

isotherms was characterized as type IV, which con-
firmed the fine mesoporous characteristics of these
membrane materials. The comparable average sizes
of ␥-AlOOH and titania particles in the sol state
(35–40 nm) (Table 2) resulted in similar final mem-
brane pore sizes (Table 3).
The structure of mixed membranes prepared from
mixed sols differed considerably from those reported
above. In this case, the membrane material showed
a completely amorphous structure, which was re-
flected in a microstructure with an average pore size
below 2 nm. Additionally, the corresponding adsorp-
tion/desorption isotherms were found to be type I,
typical for a porous material with a mainly microp-
orous texture. Membrane preparation from a mixed
sol in which smaller particles were present (25 nm),
resulted in a pore structure with a smaller mean pore
size.
3.3.2.2. Thermal treatment at high temperatures.
The effect of increasing firing temperature during
thermal treatment on structural properties of the dif-
ferent membrane materials is summarized in Table 3
and Fig. 3. Initially, the relatively lower tempera-
ture regions before phase transformations are char-
acterized by small crystallite growth and relatively
minor or gradual changes in the crystallinity of the
membrane material. In these stages, surface diffusion
accompanied with neck-formation activity between
the ceramic grains played a major role, so that thermal
treatment involved mainly a gradual decrease in sur-
face area, while pore size and pore volume changed
only little [13].
For alumina membranes, collapse of the fine meso-
porous structure started as conversion to the stable
␣-Al2 O3 phase took place. At 1200 ◦ C, after ␣-Al2 O3
transformation, grain size grew up to about 50 nm,
which was accompanied by strong surface area reduc-
tion (Fig. 3a). The shape of the adsorption/desorption
isotherm was characterized as type II, indicating the
presence of larger macropores [21].
Measurements on titania membranes, showed a sim-
ilar behaviour in the temperature region 500–700 ◦ C.
Up to 500 ◦ C, the material was present in the
anatase phase and average pore size was 3.5–5 nm.
From 500 ◦ C on, the anatase transformed into rutile
and the fine membrane structure changed drasti-
cally. Fig. 3b shows that at 600–700 ◦ C accelerated
crystallite growth took place causing strong sur-
face area reduction and large changes in the pore
volume.
For mixed membranes prepared from a homo-
geneous mixed sol, the anatase/rutile transforma-
tion occurred at higher temperature resulting in
80 T. Van Gestel et al. / Journal of Membrane Science 207 (2002) 73–89
Fig. 3. Structural development as a function of temperature for alumina (a), titania (b) and mixed (c) membranes.
an extended temperature region where only minor
structural changes occur. Between 400 and 800 ◦ C,
the membrane material remained in an amorphous
structure. Average pore diameters of 1.5–5 nm were
obtained and the resulting pore size distributions
were rather narrow. Measurements from 800 ◦ C on
indicated the formation of an anatase phase. After
thermal treatment at 920 ◦ C, this phase consisted
of crystalline grains with an average size of 20 nm.
Fig. 3c shows that at higher firing temperatures,
the formation of larger rutile grains resulted in a
strong increase in crystallite size and decrease of
surface area and pore volume. From 950 ◦ C on, the
shape of the adsorption/desorption isotherm changed
also from type IV to II, indicating serious effects
on the membrane pore structure during transforma-
tion. Meanwhile, the resulting membranes started to
show a broad pore size distribution with considerable
tailing.
3.3.3. Chemical stability
3.3.3.1. Acid corrosion tests.
3.3.3.1.1. Standard thermal treatment. Fig. 4
shows the results of acid corrosion tests on colloidal
sol–gel derived unsupported membrane layers ob-
tained after thermal treatment at 400 ◦ C. The amount
of dissolved aluminium is given as a function of pH
for ␥-AlOOH, ␥-AlOOH/anatase and mixed mem-
branes, for anatase the amount of dissolved titanium
is given.
The results illustrate that all membrane materials
showed a rather low corrosion at a pH above 3. At pH 2
and 1, the membrane materials showed, depending on
the type of material, a different behaviour. The titania
corrosion at pH 2 and 1 was low, so that the anatase
membrane material can be considered as chemically
stable down to pH 1. Contrarely, ␥-AlOOH and
 T. Van Gestel et al. / Journal of Membrane Science 207 (2002) 73–89
Fig. 4. Corrosion behaviour of ␥-AlOOH, anatase, composite and mixed membranes (*pH 2:Ti concentration = 15 ␮g/l: pH 1:Ti
concentration = 80 ␮g/l).
alumina/titania membranes underwent drastic cor-
rosion effects after exposure to pH 2 and 1, as the
nitric acid solutions contained very large amounts of
dissolved Al.
For ␥-AlOOH membranes, the results are in agree-
ment with earlier findings of Schaep et al. [19], who
already demonstrated the poor chemical stability of
␥-Al2 O3 after exposure to HCl solutions with a pH
below 3. The chemical stability of the different alu-
mina/titania membrane materials prepared from bi-
nary sols (␥-AlOOH/anatase) or from homogeneously
mixed sols is even worse than those of pure ␥-alumina
and therefore membrane applications are restricted to
a pH above 3. Obviously, below pH 3, anatase is the
best membrane material.
From these measurements, it appeared that the
crystalline properties of the ceramic base material
were determing for the membrane corrosion be-
haviour in acid solutions. For titania, the influence
of the degree of crystallinity on the properties of
the membrane material was apparent. Anatase tita-
nia showed a high degree of crystallinity and very
small corrosion effects. Contrarely, membrane ma-
terials that contain aluminium were characterized as
semi-crystalline or completely amorphous and high
amounts of Al were released. The order of degree
of corrosion was: ␥-AlOOH < ␥-AlOOH/anatase <
mixed Al2 O3 –TiO2 . The degree of crystallinity fol-
lowed the opposite order.
81
3.3.3.1.2. Thermal treatment at high temperatures.
Al2 O3 : experimental results of corrosion tests with
alumina samples treated at different firing tempera-
tures are summarized in Fig. 5 along with the phase
structure.
It is clear that corrosion of alumina was very sen-
sitive to the applied firing temperature. Between 400
and 600 ◦ C, increase of the calcination temperature
resulted in a significant increase of the corrosion. In
this temperature range, the membrane microstruc-
ture changed as a result of the first alumina phase
transformation involving the transition from an alu-
mina phase with semi-crystallinity (␥-AlOOH) to a
mainly amorphous phase containing ␥-Al2 O3 . The
increase of corrosion with increasing firing temper-
ature can be explained as a result of the change in
microstructure and the decrease of crystallinity in
the membrane material. Further increase of the firing
temperature between 600 and 1000 ◦ C, resulted in a
gradual decrease of the corrosion. As in this temper-
ature region no phase transformations occurred, the
most important change in the membrane microstruc-
ture involved a reduction of the specific surface area.
At around 1000 ◦ C, transformation into the crystal-
lographic transition phases ␦-Al2 O3 and ␪-Al2 O3
was observed. Although the corrosion resistance im-
proved substantially, such a thermal treatment still
did not yield a type of alumina membrane material
with satisfactory corrosion resistance. After thermal
82 T. Van Gestel et al. / Journal of Membrane Science 207 (2002) 73–89
Fig. 5. Corrosion of alumina as a function of temperature.
treatment above 1000 ◦ C, transformation to the final
crystallographic ␣-Al2 O3 phase started and from this
point on the material properties changed seriously:
the mainly amorphous and semi-crystalline alumina
membrane materials were transformed into a strong
crystalline structure. XRD spectra showed very nar-
row diffraction peaks after firing at 1200 ◦ C and the
estimated degree of crystallinity was very high. At the
same time, these structural changes were accompa-
nied by serious improvements in the material stability.
The corrosion, both at pH 2 and 1, decreased: the
concentration of aluminium was 200 and 330 ␮g/l
respectively, for a pH 2 and 1 corrosion test.
As a result, in the case of alumina, pore size and
corrosion resistance behave oppositely with increas-
ing firing temperatures. Fig. 5 shows that alumina
membranes prepared at temperatures below 1000 ◦ C
(␥-AlOOH, ␥-Al2 O3 , ␦-Al2 O3 , ␪-Al2 O3 ), with meso-
porous properties, showed high corrosion in strong
acid solutions. After thermal treatment at 1200 ◦ C, a
very good chemical stability was obtained, but the cor-
responding pore size distribution showed a transition
of a fine mesoporous texture into a partially macrop-
orous structure. Therefore, ␣-Al2 O3 can be used e.g.
as macroporous support material, but not as membrane
interlayer with a desired pore size distribution in the
range 5–10 nm.
TiO2 : titania membranes were not subjected to
higher firing temperatures, because the anatase mem-
branes prepared after thermal treatment at 400 ◦ C
already showed a very good corrosion resistance.
Al2 O3 –TiO2 : the mixed preparation method yielded
amorphous membranes with micropores after heating
at 400 ◦ C and mesoporous semi-amorphous mem-
branes after heating up to 800 ◦ C. At the same time,
for all these membranes a low corrosion resistance
was observed. Between 400 and 800 ◦ C, the most im-
portant change was a reduction in surface area, along
with a gradual decrease of the corrosion. After thermal
treatment above 800 ◦ C, transformation to a crystalline
structure occurred, resulting in serious changes in the
structural and corrosion properties of the membrane
material. This transformation involved the formation
of three chemically stable phases: ␣-Al2 O3 , anatase
and rutile. Between 900 and 1000 ◦ C, the transfor-
mation of the membrane material resulted also in a
significant change in the corrosion behaviour (Fig. 6).
At 920 ◦ C, anatase and ␣-Al2 O3 were the dominant
phases and rutile occured in small quantities. As
a result of the presence of two chemically stable
phases, the corrosion decreased significantly and the
pore structure remained relatively stable and kept its
mesoporous properties. With increasing temperature,
the degree of crystallinity for the different phases in-
creased sharply and a continuous improvement in the
corrosion behaviour was observed. A small increase
of the firing temperature from 920 to 950 ◦ C led to a
three times lower corrosion rate. Unfortunately, as a
result of drastic structural changes in this region, such
as macropore formation and pore volume collapse,
membranes treated at temperatures above 920 ◦ C were
unsuitable for mesoporous membrane applications.
 T. Van Gestel et al. / Journal of Membrane Science 207 (2002) 73–89 83

Fig. 6. Corrosion of mixed membranes as a function of temperature.

3.3.3.2. Alkaline corrosion tests. Additionally, me- 3.4. Membrane toplayers

mbrane materials with mesoporous properties were
tested for their corrosion resistance in alkaline media.
Table 4 shows that the corrosion tendencies in NaOH
solutions were largely comparable to those in HNO3
solutions.
For ␥-Al2 O3 membranes, low amounts of Al were
released up to pH 11; above this pH, very large
amounts of Al were found. In this pH range, mixed
␣-Al2 O3 /anatase membranes showed a much lower
corrosion rate. For this reason we propose these mixed
membranes as a good alternative for the commonly
used mesoporous intermediate ␥-Al2 O3 membrane
layers. The price to pay for the improved corrosion
resistance is an increase of the average pore diameter,
7.5 nm (Al2 O3 –TiO2 ) compared to 4 nm (␥-Al2 O3 ).
Therefore, when a smaller average pore size is re-
quired in combination with an improved corrosion
resistance, membrane interlayers of anatase are to be
preferred.
3.4.1. Morphology
The finally obtained ceramic membranes can be cat-
egorized as typical asymmetric multilayer membranes.
Fig. 7 shows the cross-section of such a membrane
consisting of a macroporous ␣-Al2 O3 support, two
mesoporous colloidal sol–gel derived anatase mem-
brane layers with an overall thickness of ca. 1 ␮m, and
a polymeric sol–gel derived anatase toplayer with a
thickness of ca. 100 nm. It appears that uniform and
discrete layers were formed and penetration of the
toplayer into the interlayer pores was succesfully pre-
vented.
3.4.2. Structural properties
Fig. 8 shows the phase behaviour of the toplayer
material as a function of heating temperature. Ini-
tially, polymeric sol–gel derived titania showed a
completely amorphous structure. Above 200 ◦ C,

Table 4
Corrosion of mesoporous interlayer membrane materials in NaOH solutions
pH ␥-Al2 O3 (600 ◦ C), ␣-Al2 O3 /anatase (920 ◦ C), Anatase (400 ◦ C),
concentration Al (␮g/l) concentration Al (␮g/l) concentration Ti (␮g/l)
11 930 570 <10
12 48000 630 35
13 201000 730 50
84 T. Van Gestel et al. / Journal of Membrane Science 207 (2002) 73–89

Fig. 7. FESEM cross-section (50,000×) of a multilayer TiO2 membrane.

transformation into a crystalline structure took place.
This transformation produced the anatase phase,
which was present up to 600 ◦ C. Pore structure char-
acterization of fine-grained polymeric sol–gel derived
membranes, at the boundary of the meso- and the
microporous region, appeared difficult because of the
lack of a reliable method. Therefore, pore size char-
acterization was based on two models. In order to de-
termine the pore size distribution in the microporous
region, the Horvath–Kawazoe model (HK) was se-
lected. At the same time, the Barrett–Joyner–Hallenda
model (BJH) was applied for pore size analysis in
the mesoporous region [22]. Before transformation,
characterization of the membrane structure indicated
mainly a microporous character. Isotherm shape cor-
responded to type I, representative for microporous
structures. With increasing temperature, crystalliza-
tion resulted in a decrease of the micropore volume

Fig. 8. XRD patterns of polymeric sol–gel titania.

 T. Van Gestel et al. / Journal of Membrane Science 207 (2002) 73–89
Fig. 9. Micropore (a) and mesopore (b) distribution of unsupported polymeric sol–gel derived titania membranes.
and a gradual increase of the average pore size
(Fig. 9). After thermal treatment at 400 ◦ C, mainly a
mesoporous anatase structure was formed with simi-
lar properties as the anatase membranes obtained via
the collodial route (Table 5).
Structural evolution of the membrane toplayer dur-
ing calcination was evaluated by thermogravimetric
analysis (TGA). In the polymeric sol–gel method, cal-
cination at an appropriate temperature was especially
important in order to remove organic side products
and to obtain a stable final inorganic structure. In TGA
studies, anatase membrane formed via a polymeric
sol–gel route (Fig. 10a) appeared rather similar to that
formed by the colloidal route (Fig. 10b). No weight
losses were detected above 300 ◦ C, as all organics,
nitric acid and water were completely eliminated at
this temperature. This implies also that processing
of a stable polymeric sol–gel derived toplayer was
perfectly possible at low thermal treatment with
Table 5
Pore structure characteristics of polymeric titania membranes
Phase Firing (◦ C) Mean pore size (nm)a
Amorphous 200 1.6
Anatase 300 2.4
Anatase 400 3.4
a BJH-model.
85
preservation of the fine pore properties and limited
risk of cracking.
3.4.3. Chemical stability
Results of corrosion tests with polymeric sol–gel
derived titania membrane materials are summarized
in Fig. 11. After thermal treatment at 200 ◦ C, titania
corrosion was rather low at pH 2, but high amounts of
Ti were released during acid exposure at pH 1.
Between 200 and 300 ◦ C, the microstructure
changed significantly as a result of the transformation
from an amorphous into a crystalline structure. As the
transformation involved the formation of the chemi-
cally stable anatase phase, this resulted in a significant
decrease of the corrosion. With increasing tempera-
ture, at 300 and 350 ◦ C, the degree of crystallinity in-
creased and a good corrosion resistance was obtained
over the complete acid range down to pH 1. Additional
tests on the resistance against alkaline corrosion of the
Surface area (m2 g−1 ) Pore volume (ml g−1 ) Isotherm (type)
384 0.21 I
307 0.19 I, IV
130 0.13 IV
86 T. Van Gestel et al. / Journal of Membrane Science 207 (2002) 73–89

Fig. 10. TGA of polymeric (a) and colloidal sol–gel derived titania (b).

Fig. 11. Corrosion of polymeric titania as a function of temperature.

same membrane material proved also a high stability 4. Dynamic characterization of the multilayer
in NaOH solutions with a pH up to 13 (Table 6). membranes
From the results of all characterizations (pore struc-
ture, TGA, chemical stability), it was concluded that The retention behaviour of supported multilayer
an anatase toplayer membrane structure with microp- membranes was expressed by the MWCO, the molec-
ores and ultrafine mesopores, as obtained by a thermal ular weight of the component from the PEG mixture
treatment at 300 ◦ C, appeared to be the best toplayer that was retained for 90%. Membrane permeability
material. was determined according to the Hagen–Poisseuille
equation from flux measurements. As shown in
Table 6 Fig. 12, flux measurements as a function of pressure
Corrosion of polymeric sol–gel derived anatase (300 ◦ C) in HNO3 showed a linear dependency of flux on transmembrane
and NaOH solutions pressure.
pH Table 7 summarizes the MWCO and the mem-
1 2 3 11 12 13
brane permeability determined for membrane support
systems consisting of the main Al2 O3 support and
Ti concentration (␮g/l) 610 16 <10 <10 <10 11 two mesoporous colloidal sol–gel derived interlayers.
 T. Van Gestel et al. / Journal of Membrane Science 207 (2002) 73–89 87

Fig. 12. Flux as a function of transmembrane pressure (TiO2 multilayer NF-membrane).

Table 7 rived anatase toplayer yielded multilayer membranes

Performance mesoporous UF-membrane supports with NF-characteristics. Firstly, with very smooth
Multilayer configuration MWCO Permeability high-quality ␥-Al2 O3 supporting layers, high-quality
(support = ␣-Al2 O3 ) (l h−1 m−2 bar−1 ) NF-membranes could be rather easily obtained. It
(1) ␥-Al2 O3 5000 5 appeared that the smooth surface of these support
(2) ␣-Al2 O3 /anatase 15000–20000 27 membrane layers was helpful for modification with a
(3) Anatase 10000–15000 22 high-quality crack-free anatase toplayer. Membrane
system (4) has a potential for applications in the low
NF-region (MWCO <200), but shows a low corro-
Membrane system (1), prepared by coating colloidal
sion resistance compared to the TiO2 membranes.
␥-Al2 O3 layers, showed the properties of a tight
This membrane configuration is especially interest-
UF-membrane with a MWCO of ca. 5000. Membrane
ing for NF-applications in neutral or non-aqueous
systems (2) and (3) showed both typical UF-properties
media. Performance measurements on the multi-
with a MWCO of ca. 15000 in combination with a
layer membrane configuration (5), consisting of an
higher permeability. Type II was developed as an alter-
anatase toplayer on an anatase UF-membrane support,
native for ␥-Al2 O3 membranes, but showed a smaller
showed a MWCO <1000 and a high permeability
thickness and larger pore size which resulted in a
(±20 l h−1 m−2 bar−1 ). This multilayer membrane,
higher permeability and cut-off value. The high per-
consisting of three anatase sol–gel layers, is there-
meability of the anatase support system (3) can also be
fore a convenient configuration for NF-applications
explained by the small layer thickness. The intermedi-
in aqueous corrosive media. A lower MWCO in
ate cut-off is probably due to the (inevitable) presence
the range of 200 is also posssible for a comparable
of small defects in these thin membrane layers.
multilayer membrane (6), consisting of an anatase
According to the results in Table 8, an addi-
intermediate UF-membrane support modified with an
tional modification with a polymeric sol–gel de-
optimized anatase toplayer obtained by a very slow
drying process, but for such a tighter NF-membrane
Table 8
Membrane performance disk-shaped multilayer membranes
the permeability is strongly reduced.

Multilayer configuration MWCO Permeability

(support = ␣-Al2 O3 ) (l h−1 m−2 bar−1 ) 5. Conclusions
(4) ␥-Al2 O3 /anatase <200 4
(5) Anatase/anatase 500–1000 19 The membranes obtained consist of a macrop-
(6) Anatase/anatase <200 2
orous ␣-Al2 O3 support, two mesoporous membrane
88 T. Van Gestel et al. / Journal of Membrane Science 207 (2002) 73–89
layers with an overall thickness ranging from
1 ␮m for titania to 4 ␮m for alumina, and a fi-
nal polymeric titania toplayer with a thickness of
ca. 100 nm. Depending on the thermal treatment,
␥-AlOOH, ␥-Al2 O3 , anatase, ␥-AlOOH/anatase,
amorphous mixed and ␣-Al2 O3 /anatase mixed mem-
brane layers were formed with narrow pore size
distributions ranging from the higher microporous
region (<2 nm) to the lower mesoporous region
(2–8 nm).
In mild aqueous (pH 3–11) or non-aqueous liquids,
all membrane materials studied are chemically stable.
In our NF-membrane development, strong preference
was given to a multilayer configuration, consisting
of mesoporous ␥-Al2 O3 layers and an anatase NF
toplayer due to the relatively easy processability of
this membrane configuration.
In strong acid (pH <3) or alkaline solutions
(pH >11), membrane materials consisting of weakly
crystallized or amorphous phases have to be avoided
in the multilayer configuration. Although the finer
pore structure and the high-quality layer forma-
tion of such materials is very advantageous, tests
on the corrosion behaviour indicated that a transi-
tion to crystalline membrane structures is strictly
needed. For the development of membrane interlay-
ers, mesoporous anatase or mixed ␣-Al2 O3 /anatase
were considered. With a second dip-coating pro-
cedure included, preference was given to anatase
as interlayer material due to the higher chemi-
cal stability, smaller pore size and narrow pore
size distribution, while a sufficient overall inter-
layer thickness of ca. 1 ␮m was obtained. For the
formation of the NF toplayer with a very fine mi-
crostructure, only anatase is suitable as membrane
material.
The ␣-Al2 O3 support, the mesoporous ␥-Al2 O3 ,
anatase and ␣-Al2 O3 /anatase interlayers and the
anatase toplayer were characterized by a MWCO in
the micro-, ultra- and NF range, respectively. For
the finally optimized ␣-Al2 O3 /␥-Al2 O3 /anatase and
␣-Al2 O3 /anatase/anatase NF-membranes, a MWCO
below 200 was obtained. Due to their chemical sta-
bility with respect to non-aqueous organic liquids
or aqueous solutions with acid or alkaline pH, our
membrane development has provided ceramic mem-
branes applicable in a wide range of NF appli-
cations.
Acknowledgements
Herman Cooreman is gratefully acknowledged
for performing the ICP-Mass Spectroscopy mea-
surements and Frans Servaes for his kind help and
technical support.
References
[1] M. Mulder, Basic principles of membrane technology, 2nd
Edition, Kluwer Academic, Dordrecht, 1996.
[2] J.A. Hestekin, C.N. Smothers, D. Bhattacharyya, in: Procee-
dings of the 1999 AIChE Annual Meeting,Nanofiltration for
Removal of Organics from Aqueous and Organic Solvent
Streams, cd-ROM Edition, Dallas, USA, 31 October–5
November, 1999.
[3] A.J. Burggraaf, L. Cot, Fundamentals of inorganic membrane
science and technology, Elsevier science and Technology
Series 4, Elsevier, Amsterdam, 1996.
[4] J. Luyten, T. Van Gestel, J. Cooymans, C. Smolders, Proce-
ssing of multilayer ceramic nanofiltration membranes, Inno-
vative Processing/Synthesis: Ceramics, Glasses, Composites
IV, Ceramic Transactions, Vol. 115, 2000.
[5] A. Larbot, S. Alami-Younssi, M. Persin, J. Sarrazin, L.
Cot, Preparation of a ␥-alumina nanofiltration membrane, J.
Membr. Sci. 97 (1994) 167–173.
[6] R. Vacassy, C. Guizard, V. Thoraval, L. Cot, Synthesis and
characterisation of microporous zirconia powders. Application
in nanofiltration characteristics, J. Membr. Sci. 132 (1997)
109–118.
[7] S. Vercauteren, Clay and pillared clay membranes: synthesis,
characterization and transport properties, PhD thesis,
University of Antwerp, Belgium, 1997.
[8] J. Luyten, J. Cooymans, C. Smolders, S. Vercauteren,
E.F. Vansant, R. Leysen, Shaping of Multilayer Ceramic
Membranes by Dip-Coating, J. Eur. Ceram. Soc. 17 (1997)
273–279.
[9] J. Luyten, J. Cooymans, W. Adriansen, R. Leysen,
in: Proceedings of the 9th CIMTEC—World Ceramic
Congress on Alternative process routes for better ceramic
membrane supports, Florence, Italy, 14–19 June, 1999,
pp. 449–456.
[10] A.J. Burggraaf, K. Keizer, in: R.R. Bhave (Ed.), Synthesis of
Inorganic membranes, Inorganic membranes: Characterization
and Applications, van Nostrand Rheinhold, New York, 1991,
pp. 10–63.
[11] V.T. Zaspalis, W. Van Praag, K. Keizer, J.R.H. Ross, A.J.
Burggraaf, Synthesis and characterization of primary alumina,
titania and binary membranes, J. Mater. Sci. 27 (1992) 1023–
1035.
[12] K.N. Kumar, Nanostructured ceramic membranes: layer
and texture formation, PhD thesis, University of Twente,
Enschede, The Netherlands, 1993.
[13] C.-H. Chang, R. Gopalan, Y.S. Lin, A comparative
study on thermal and hydrothermal stability of alumina,
 T. Van Gestel et al. / Journal of Membrane Science 207 (2002) 73–89
titania and zirconia membranes, J. Membr. Sci. 91 (1994)
27–45.
[14] R.S.A. de Lange, J.H.A. Hekkink, K. Keizer, A.J. Burggraaf,
Formation and characterisation of supported microporous
ceramic membranes prepared by sol–gel modification
techniques, J. Membr. Sci. 99 (1995) 57–75.
[15] S. Vercauteren, K. Keizer, E.F. Vansant, J. Luyten, R.
Leysen, Porous ceramic membranes: preparation, transport
properties and applications, J. Porous Mater. 5 (1998) 241–
258.
[16] I. Voigt, G. Fischer, P. Puhlfürb, D. Seifert, in: Procee-
dings of the 5th International Conference on Inorganic
Membranes ICIM, New Filtration Ceramics—From Support
to NF-Membrane Completely of TiO2 , Nagoya, Japan, 22–26
June, 1998, pp. 42–45.
[17] P. Puhlfürb, A. Voigt, R. Weber, M. Morbé, Microporous
TiO2 membranes with a cut-off <500 Da, J. Membr. Sci.
174 (2000) 123–133.
89
[18] J.M. Hofman-Züter, Chemical and thermal stability of
mesoporous ceramic membranes, PhD thesis, University of
Twente, Enschede, The Netherlands, 1995.
[19] J. Schaep, C. Vandecasteele, B. Peeters, J. Luyten, C.
Dotremont, D. Roels, Characteristics and retention properties
of a mesoporous ␥-Al2 O3 membrane for nanofiltration, J.
Membr. Sci. 163 (1999) 229–237.
[20] T. Hollstein, in: Proceedings of the 9th CIMTEC—World
Ceramic Congress on Corrosion of Advanced Ceramics by
Liquid Media: a Review, Florence, Italy, 14–19 June, 1999,
pp. 433–444.
[21] S. Lowell, J.E. Shields, Powder surface area and porosity,
3rd edition, Powder technology series, Delft University of
Technology, The Netherlands, 1991.
[22] H. Richter, G. Tomandl, C. Siewert, A. Piorra, in: Proceedings
of the 5th International Conference on Inorganic Membranes
ICIM, Ceramic nanofiltration membranes made of ZrO2 and
TiO2 , Nagoya, Japan, 22–26 June, 1998, pp. 30–33.