Critical Points in The Analysis of Membrane Pore Structures by Thermoporometry

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

45

Journal of Membrane Science, 66 (1992) 45-53


Elsevier Science Publishers B.V., Amsterdam

Critical points in the analysis of membrane pore


structures by thermoporometry
F.P. Cuperus, D. Bargeman

and C.A. Smolders

University of Twente, Faculty Chemical Technology, P.O. Box 217, NL-7500 AE Enschede (Netherlands)
(Received January 24,199l;

accepted in revised form August 26,199l)

Abstract
Several ultrafiltration membranes of the anisotropic and isotropic type were characterized by means
of the thermoporometry. Successive cooling and heating runs were performed in order to investigate the
effects of the water-ice and ice-water phase transitions on the structure of the membranes. The results
found for membranes having different casting thicknesses indicate that, in some cases, pores in the
sublayer of anisotropic UF membranes frustrate the measurement of the, for the separation, relevant
pores present in the top layer.
Keywords: membrane preparation and structure; microporous and porous membranes; ultrafiltration;
thermoporometry; membrane characterization

Introduction
Since the introduction
of thermoporometry
by Brun et al. in 1973, various porous structures have been examined by means of this
technique, mainly by Brun et al. [l-7].
Most
of these structures were of inorganic, ceramic
nature although in the eighties a few organic
swollen and non-swollen resins were investigated [5,6]. All the examined structures were
of the isotropic type and the presence of a relatively high pore volume makes the pore size
analysis by thermoporometry
fairly simply. The
success of thermoporometry
in the systems used
by Brun, brought several investigators to apply
the technique for the characterization
of poCorrespondence to: F.P. Cuperus, AT0 Agrotechnologie,
Agrotechnological Research Institute, P.O. Box 17, 6700
AA Wageningen, The Netherlands.

0376-7388/92/$05.00

rous structures in the (wet) environment


in
which they are actually used. Desbrieres et al.
[ 81 measured pore size distributions
of hemodialysis membranes, whereas Zeman and Tkacik [ 91 and Smolders and Vugteveen [lo] used
the method for the characterization
of ultrafiltration (UF) membranes.
Ultrafiltration
membranes are generally prepared from polymers; this enables one to manufacture porous membranes in a large area per
volume ratio, i.e. in the form of flat sheets or
hollow fibres. The main difference between
these polymer systems and the porous ceramic
media studied by Brun is the superior mechanical and chemical stability and the better delinition of the morphological structure (in terms
of pore size distribution)
of the ceramic materials. Also, polymeric UF membranes are anisotropic: they consist of a thin skin layer with
very small pores determining the performance

0 1992 Elsevier Science Publishers B.V. All rights reserved.

46

of the membrane and a thicker macroporous


sublayer which provides the membrane with
mechanical solidity. As the skin layer is very
thin and its (surface) porosity generally is in
the order of a few percent [11,X?], the pore volume is low in comparison with ceramic membranes or catalysts. These aspects can have a
large impact on the experimental procedure to
be used and on the interpretation of the data
found with thermoporometry.
Normally in the thermoporometry measurements of polymeric membranes, water is used
as the liquid imbibed inside the pores, as this
system of water filled pores has the closest resemblance with the practical situation. During
the measurements, the water inside the pores
will undergo a liquid-solid transition at least
one time, a change of state which is accompanied by a volume expansion of the water. The
question arises whether this phase transition
changes the structure of the membrane in a reversible or in an irreversible way. For ceramics
such a possible effect has never been mentioned, which confirms the assumed morphological stability of these materials. On the other
hand, Brun suggests [4,5] that in a swollen
resin in which the pore size in principle can
change reversibly, different pore size distributions are found depending on the degree of
swelling. The swollen membranes show broad
distributions with sizes spreading over one order of magnitude, whereas the unswollen dry
membranes analysed by the widely used method
based on gas adsorption-desorption measurements, exhibit a very sharp pore size
distribution.
Brun et al. [6] compared different characterization techniques by applying them to one
and the same ceramic system. It appeared that
results found with thermoporometry and gas
adsorption-desorption hysteresis are consistent with one another. Similar agreement was
found by Zeman and Tkacik [ 91 and Smolders
and Vugteveen [lo] when both methods were

F.P. CUPERUS ET AL.

applied to polymeric ultrafiltration membranes. On the other hand, the thermoporometry results found by Zeman did not match
with his SEM observations and could hardly
explain the low cut-off values found for the
same membranes, even when concentrationpolarization effects were taken into account
[ 131. Consequently, Zeman suggests that by
thermoporometry and gas adsorption-desorption the pores in the sublayer of the membrane
are measured and therefore these methods do
not give relevant information on the pores
present in the skin. Smolders and Vugteveen
found that the retention of components with
molecular weights in the order of 10,000 Da was
very low, which was not in agreement with the
small pores found with thermoporometry. The
conclusion must be that, although in some cases
thermoporometry is a simple and effective
method to determine the pore structure of certain porous media, for a number of particular
systems, especially anisotropic polymeric
membranes, doubtful results are found.
Theory
The basic principle of thermoporometry is
the freezing (or melting) point depression
which is due to the strong curvature of the solidliquid interface present within small pores. A
full thermodynamic description of this phenomenon is given by Brun et al. [ 11. According
to his treatise the size of a confined ice crystal
is inversely proportional to the degree of undercooling (AT). Finely dispersed ice as present in a porous matrix therefore melts at temperatures below the ambient melting point of
ice. The smallest size of the ice crystals or the
smallest pore size that can be described properly is set by the validity of the assumptions
made in the thermodynamic description of
curved surfaces. For instance, the approximation of several properties of dispersed water by
using values of normal, bulk water is limited to

47

ANALYSIS OF MEMBRANE PORE STRUCTURES BY THERMOPOROMETRY

volumes with characteristic radii larger than l2 nm. According to Defay and Prigogine [ 151
dispersed water at temperatures below - 50 C
cannot be described anymore by using normal
equilibrium thermodynamics. At such low temperatures any form of liquid water is instable
and will become solid. This phenomenon can
interfere with the normal equilibrium freezing
of water in very small pores ( < 2 nm). Monitoring cooling effects can also be disturbed by
delayed nucleation of the ice phase inside the
pores. This will be a minor effect when hydrophilic samples are analysed because the wetted
membrane walls can act as nuclei and induce a
fast liquid-solid transition [ 151. However, for
extremely hydrophobic materials delayed nucleation can influence the pore size analysis.
Therefore the melting transition is preferred
above the solidification for the pore size analysis, although the latter in principle should
render identical information.
For the analysis of a porous substrate, the
melting diagram (Fig. 1) can be monitored in
a differential scanning calorimeter (DSC ) and
the pore size and pore volume are calculated
from the equations derived by Brun. For the
solid-liquid transition of water inside a cylindrical pore with radius r eqn. (1) holds. From
the heat effect occurring during the transition,
the pore volume at a certain undercooling (i.e.
pore radius) is calculated. Equation (2 ) gives
the heat of melting as a function of temperature, used in this procedure.

o T(C)
Fig. 1. DSC thermogram as found for a narrow pore size
distribution (A) and for a broad pore size distribution (B ) .

r=0.68-

(32.33/47)

W,= -0.155AT2-11.39AT-332

(1)
(2)

Here r represents the pore radius (nm), AT


the characteristic undercooling ( C), and W,
apparent transition energy (J/g).
Similar equations can be derived for the solidification process. These are not given here
because only the melting process was used in
the pore size analysis described in this paper.
As long as the pores have sizes between 2 nm
and 25 nm, the shift of the freezing or melting
points is large enough to be measured accurately. A typical example of a melting curve is
given in Fig.1, curve A. Here, water inside the
pores and free water outside the membrane
give rise to separate peaks and consequently the
pore size distribution can be calculated easily.
For some membrane types a deviant curve is
found (see Fig. 1, curve B) and distributions
calculated from such a curve remain questionable. This problem is discussed in more detail
later-on in this paper (see: results and
discussion).
Experimental
Both ceramic and polymeric membranes were
investigated by means of thermoporometry.
The ceramic y-alumina systems were supplied
by the inorganics group of our department and
are described in the literature. According to gas
adsorption-desorption and transmission electron microscopy these membranes have a welldefined pore structure and a porosity of ca. 50%
[ 141. Furthermore hollow fibre CA membranes
used in blood filtration were analysed. According to literature these isotropic membranes exhibit a high porosity [ 8,161 which should make
analysis by thermoporometry fairly simple.
Lab-made anisotropic ultrafiltration membranes were made of poly( 2,6-dimethyl-1,4
phenylene oxide) (PPO ) , polysulfone (PSf)

F.P. CUPERUS ET AL.

48

and cellulose acetate (CA). PPO membranes


were prepared from a 10 wt.% polymer concentration in a mixture of trichloroethylene and loctanol in a weight ratio of 78/22. PSf membranes were cast from a solution composed of
15 wt.% PSf (P-3500 Union Carbide) in DMF.
Flat CA membranes were made out of a 20 wt.%
CA solution in 59 wt.% aceton, 19 wt.% demiwater and 2 wt.% magnesiumperchlorate. The
solutions were cast on a glass plate at room
temperature at thicknesses of 0.2 and 0.03 mm,
respectively. In case of PPO, methanol was used
as the coagulation bath, PSf and CA membranes were immersed in a water bath.
The calorimetric experiments were performed using a Perkin Elmer DSC 2 or DSC 4.
First the sample with a total mass of ca. 50 mg
(including water) was cooled at maximum
speed to - 45 C. After equilibrium was reached
the heat effects during a controlled heating run
at a scanning rate of 1 C/min were measured.
Pure water flux measurements were performed using freshly filtered RO water and a
dead-end Amicon cell with a membrane area of
35 cm, at a pressure of 300 kPa. The steady
state flux was determined after 3 hr of
permeation.

0.3 ,

ads.-des.

I-

1I\

therm0

-z
6 0.2 3
0
z

O.l-

,J!I
0.0

r (W

Fig. 2. Pore size distributions found for alumina membranes using thermoporometry and gas adsorption-desorption measurements (dv/dr: differential pore volume per
g membrane ) .

10

r (nm)

Fig. 3. Change in pore size distribution for hollow fibre CA


membranes after two successive runs.

Results and discussion


The effect of successive heating runs
y-Alumina membranes and hollow fibre CA
membranes exhibit a sharp pore size distribution (Figs. 2 and 3 respectively). The results
found for the alumina membranes are in agreement with the data found with gas adsorptiondesorption [ 141. Alumina samples were analysed during several successive heating runs (up
to 8 runs), but differences between the runs
were not found. This means that this type of
membrane is perfectly stable; the structure does
not change during the analysis.
In Fig. 3 pore size distributions for hollow
fibre CA membranes, calculated from several

consecutive runs carried out with the same


sample of membrane material, are shown. It
appears that between the first and second run
the pore size distribution shifts to slightly larger
pore values. Differences between the following
experiments are not significant. Obviously, the
pore structure of the sample changes during or
directly after the first two experiments but is
stable thereafter. Probably this alteration has
occurred during the cooling of the sample. Since
the liquid-solid transition is accompanied by a
volume expansion of 8-20% (dependent on the
temperature and the pressure there exist different ice phases) [17,18], the nuclei deform
the porous matrix to some extent. Because the

ANALYSIS

OF MEMBRANE

PORE STRUCTURES

49

BY THERMOPOROMETRY

nuclei are induced within the original restricting pore space, the ultimate size of the crystal
will only be a little larger than the nascent pore
size. During the second cooling run the nuclei
again deform the membrane matrix and enlarge the pore. Consequently the pore size measured in the second heating run is larger than
that measured in the first run. From the volume expansion of the water the increase in pore
size (for an originai) pore of 5 nm) is calculated to range from 0.1 nm (0 C and lo5 Pa) to
0.3 nm ( - 35 C and 2 x 10 Pa). Especially the
latter value corresponds quite well with the experimental pore size increase. The fact that the
volume expansion has to proceed in a confined
pore space makes a transition accompanied by
such a high pressure comprehensible. Nevertheless, the deformation appears to have
reached its ultimate extent already after the
second cooling step, since from the third run no
change in the pore size distribution could be observed anymore. Probably the affine deformation caused by the water-ice transition
stretches the pore walls to such an extent
(reached after 2 transitions) that under the
prevailing forces no further deformation is possible. The deformation of the original pore is
not reversible because a relaxation phenomenon was not found; even after three days relaxation at 20C the pore size distribution was
identical with the previous (second) analysis
run.
The pore size distribution of the thick PPO
membranes determined by thermoporometry
appeared to be very sharp (Fig. 4). All the pores have sizes between 1.5 and 4 nm and a maximum is found at ca. 2 nm. Figure 4 shows that
the cumulative pore volume found in the first
run is significantly lower than the value found
(with the same sample) in the following runs.
This remarkable difference suggests that not all
the water present in the pores has been crystallized upon cooling during this first run but
that at -45C part of the water is present in

200

200

150

150

,:

NE
2
u
L
9
5

E
p
z
E

100

100

u)
z
50

2
(D
0

50

0
0

r (W

Fig. 4. Change in total pore volume (V,,)


for PPO membranes after several successive runs (casting thickness 0.2
mm; pore volume expressed as volume per unit membrane
area).

the glassy state. This so-called vitrification occurs when the cooling rate has been so fast that
nuclei could not be formed. However the extreme conditions needed for vitrification (cooling rates> lo6 C/set) are certainly not present in the equipment used here, hence the
presence of glassified water is not probable
[ 191. Since PPO is a hydrophobic material,
heterogeneous nucleation induced by, e.g. the
polymer wall is less probable than for hydrophilic materials. In the absence of other nuclei
which could induce crystallization, homogeneous nucleation may be important. When homogeneous nucleation indeed should induce
crystallization, the water (at -45C) may be
present as an undercooled liquid. According to
the analysis of Defay and Prigogine [ 151, the
homogeneous nucleation process at low temperatures is extremely slow and it might need
several months to induce crystallization in all
the pores of the membrane. This means that
after cooling to - 45 C, water can be present
as an undercooled liquid instead of ice. Hence,
phase transition (accompanied by a measurable heat effect) does not occur. The question is
why during the second cooling run more ice is
formed and crystallization appears to be less
delayed. It should be expected that after the

50

heating run, all the ice has become water and


the nucleation process then has to start all over
again. Apparently, after the first heating run,
conditions are permanently changed in a way
that heterogeneous nucleation can take place
in all the pores.
Another possible explanation for the observed increase in pore volume after the first
heating run might be the presence of non-wetted pores, which become wetted after the first
run. However, the origin of such non-wetted
pores inside the membranes is hard to understand. Pores in UF membranes are formed by
liquid-liquid demixing, which for PPO means
that, in the native state, all the pores in the
membrane are filled with methanol from the
coagulation bath. The pores are filled with water
by exchanging all the methanol, so in this way
dry pores or de-wetting of the membranes
cannot be present. It may be that a trace of
methanol is still present in some pores and
could hinder crystallization, but also then it is
not clear why during the second cooling run
crystallization is more complete.
The structure of the top- and sublayer
The thermogram of a heating run executed
with PPO membranes with a casting thickness
of 0.20 mm shows two distinctly separated
peaks (Fig. 1, curve A) and a simple calculation
of the pore size distribution is possible. Other
UF membranes, like the thick PSf and flat CA
membranes, exhibit deviant melting curves like
depicted in Fig. 1, curve B. Although in these
latter cases the calculation of the pore size is
possible, the resulting pore size distribution will
be more questionable. Not only the impossibility to discriminate fully between pore peak and
free water peak gives rise to doubtful results
but also the precarious drawing of the base line.
Furthermore it is not sure whether the larger
pores, related to the higher melting temperatures, are present in the skin or in the sublayer.
Especially in case of anisotropic membranes

F.P. CUPERUS ET AL.

which have relatively thick sublayers it can be


imagined that going from the top to the bottom
of the membrane a pore size gradient is present
which leads to a melting curve with a shape as
given in Fig. 1, curve B.
From the distribution obtained for PPO
membranes (Fig. 4)) the cumulative pore volume was calculated. Here, the total pore volume is expressed as volume per unit area of the
top layer and for the thicker PPO membranes
the values ranged from 5 x 10m5till 20 x lop5
cm3/cm2 and appeared to be strongly dependent on the casting thickness [lo]. Because a
PPO membrane is an anisotropic membrane,
this pore volume is expected to be related to
pores in the top layer. Consequently, the top
layer thickness of the membrane can be calculated when a pore model is assumed. Recently,
the top layer thickness of these PPO membranes, directly determined by a new method
[20], was found to be ca. 0.2 pm. This result
makes it questionable whether the pore volume
measured with thermoporometry is only present in the skin. A simple calculation shows that
a porosity value of more than 100% is needed
to account for this volume to be present in a
skin layer with a thickness of 0.2 pm. Hence,
there are also small pores present in the sublayer of the membrane.
It has already been mentioned that the calculation of the pore size distributions of CA and
PSf membranes from thermograms of the type
shown in Fig. 1 (curve B ) is cumbersome. From
the anisotropic membrane concept one should
expect that the small pores of the PSf and CA
membranes are present in the skin. If this is
really the case it is mainly because of the very
low porosity of the skin [ 11,121 that the number of the pores or the cumulative volume in the
skin cannot be measured. The smallest pore size
can be found from the lowest temperature at
which the thermogram (Fig. 1) deviates from
the base line, but the larger pores in the sublayer make the complete analysis troublesome.

51

ANALYSIS OF MEMBRANE PORE STRUCTURES BY THERMOPOROMETRY

In order to diminish the influence of the sublayer pores, the ratio of the pore volume present in the skin to that in the sublayer was increased by preparing ultra thin PPO, CA and
PSf membranes. For membranes with a casting
thickness of 0.03 mm two separate peaks were
found again and a pore size distribution could
be calculated. The resulting distributions for CA
and PSf membranes are given in Figs. 5 and 6,
respectively. The distributions are considerably broader than the pore size distributions
found for the thicker PPO membranes (Fig. 4).
The broad pore size distribution found for the
thin (0.03 mm) PPO membranes (Fig. 7) completely deviates from the characteristic sharp

40

4
g
E3
s
e
B
2
rc,
a

30

g
E
a

20

_
0

10

0
0

10

30

20
r (nm)

Fig. 7. Comparison of PPO membranes with casting thicknesses 0.03 mm and 0.20 mm. (scale dv/dr 0.20 mm membranes should be multiplied by 10, see also Fig. 4).

TABLE 1
Flux and pore volume as determined by thermoporometry
for various membranes
Membrane type
Casting
thickness (mm)
Flux (cm/hrbar) (& 1)
lo5 V,,, (cm3/
cm) (*30%)
0

10

PPO
0.2

0.03

5-20

PSf
0.2

CA

0.03

0.2

0.03

30

15

r (W

Fig. 5. Pore size distribution of thin CA membranes (casting thickness 0.03 mm, pore volume expressed as volume
per unit membrane area).

20
z

16-

E
p
E
+

12-

9
a
-

, 200

160%
4
- 120rnE
2
5
.6O

4-

-40

10

20

2
0
-

30

r (nm)

Fig. 6. Pore size distribution of thin PSf membranes (casting thickness 0.03 mm, pore volume expressed as volume
per unit membrane area).

peak exhibited by the thick (0.20 mm) PPO


membranes. This difference in pore size distribution between the thin and the thick PPO
membranes suggests a change in membrane
structure caused by a change in casting thickness. Now the question arises whether the results found for the PSf and CA membranes are
biased by the change of the casting thickness.
To investigate this possibility, the pure water
flux of the thin and thick membranes (of CA,
PSf and PPO) was measured. The results are
given in Table 1.
The pure water fluxes of both PPO membrane types are surprisingly close to each other.
This does not only imply that the pore structure of both types is similar but also supports
the idea of the predominant role of the top layer

F.P.CUPERUS
ETAL.

52

in this case. On the other hand, it can be argued


that the small pores found by thermoporometry in thicker PPO membranes are not all related to pores in the skin. The number of pores
that do determine the performance is so small
that they cannot be measured by thermoporometry [ 11,121. The thin membranes used are
assumed to have a higher ratio of pore volume
in the skin to pore volume in the sublayer than
the thick membranes, but this is only due to a
thicker sublayer. A DSC-sample of a thin membrane therefore contains about 10 times more
skin than a sample of a thick membrane, which
makes the detection of the pores in the skin
easier. Consequently the pore size distribution
found for the thin membranes is expected to be
a better representation of the actual distribution of the pores in the skin. Or in other words:
the real pore size distribution in the skin of thick
PPO membranes resembles the pore size distribution found for the thin membranes. For the
CA and PSf membranes significant differences
in flux values are found for different membrane
thicknesses.
The increase in the flux value for thin PSf
membranes might imply that the structure of
the membrane has been changed, e.g. the average pore size has increased or the skin thickness has decreased or a combination of these
two effects has occurred. It can also be argued
that for these membranes the sublayer has a
substantial resistance for transport. The existence of a pore gradient from the skin to the
bottom of the membrane, which could contribute to this resistance was indicated by other
methods too [ 10,20,21]. When such a gradient
does exist in both thin and thick membranes,
the effect of using thin membranes in thermoporometry might be that the larger pores ( > 30
nm) which are present at the bottom of the
thick membranes are missing in the thin membrane. The tenfold increase of the water flux,
however, can hardly be explained in this way
and it must be concluded that the decrease in

membrane thickness resulted in a change of


pore size as well.
The flux of the 0.20 mm CA membranes is
higher than that of the 0.03 mm membranes.
This can only be explained by a change in pore
structure. Consequently, CA membranes of different casting thickness cannot be compared.
Conclusions
It has been shown that thermoporometry can
be a very effective method for the characterization of porous media. However, for anisotropic polymeric membranes the method cannot be used without certain restrictions.
Because the whole membrane, i.e. top layer and
sublayer structure which both can contain pores of comparable size, is analysed, it is sometimes difficult to decide whether a measured
pore size distribution corresponds to pores in
the sublayer or to pores in the skin [21,22].
Furthermore, the very low porosity of the skin
frustrates the accurate determination of the
pore size distribution.
The smallest pore that can be analysed properly by thermoporometry is ca. 2 nm. The limits are set by the assumptions made in the thermodynamic description of the process which are
no longer valid for temperatures lower than
- 40 C. The upper limit, or the largest pore that
can be determined successfully by thermoporometry, is set by the effect of the curvature on
the freezing point depression. For pores of 30
nm or larger the shift from the ambient fusing
point is so small that the effect cannot be separated from the free water peak and accurate
determination of pore radius and pore volume
is not possible anymore.
References
1

M. Brun, A. Lallemand, J.F. Quinson and Ch. Eyraud,


Changements d&at liquide solide dans les milieux poreux I-III, J. Chim. Phys., 70 (1973) 973.

ANALYSIS OF MEMBRANE PORE STRUCTURES BY THERMOPOROMETRY

M. Brun, A. Lallemand, J.F. Quinson and Ch. Eyraud,


A new method for the simultaneous determination of
the size and the shape of the pores: the thermoporometry, Thermochim. Acta, 21 (1977) 59.
3 M. Brun, J.F. Quinson and L. Benoist, Determination
of pore size distributions by DSC, Thermochim. Acta,
49 (1981) 49.
4 M. Brun, J.F. Quinson and R. Spitz, Caracterisation
texturale par thermoporometrie et par adsorption-desorption dazote de resin d&tat gonfle, Macromol.
Chem., 183 (1982) 1523.
5 M. Brun, J.F. Quinson and R. Blanc, Caracterisation
texturale des resins en milieu reactionel par thermoporometrie, Macromol. Chem., 181 (1981) 873.
6 Ch. Eyraud, J.F. Quinson and M. Brun, The role of
thermoporometry in the characterization of porous
solids, in: K.K. Unger, J. Rouquerol, K.S.W. Sing and
H. Kral (Eds.), Characterization of Porous Solids,
Elsevier, Amsterdam, 1988, p. 295.
7 Ch. Eyraud, J.F. Quinson and M. Brun, Progress in
thermoporometry, in: K.K. Unger, J. Rouquerol,
K.S.W. Sing and H. Kral (Eds.), Characterization of
Porous Solids, Elsevier, Amsterdam, 1988, p. 307.
8 J. Desbrieres, M. Rinando and M. Brun, Relation entre
le taux de retention et la distribution des pores dans
une membrane dUF, J. Chim. Phys., 78 (1981) 187.
9 L. Zeman and G. Tkacik, Pore volume distribution in
UF membranes, in: D.R. Lloyd (Ed.), Materials Science of Synthetic Membranes, ACS Symp. Ser. No.
269, American Chemical Society, Washington, DC,
1985, p. 339.
10 C.A. Smolders and E. Vugteveen, New characterization methods for asymmetric UF membranes, in: D.R.
Lloyd (Ed. ), Materials Science of Synthetic Membranes, ACS Symp. Ser. No. 269, American Chemical
Society, Washington, DC, 1985, p. 329.

11

53

A.G. Fane, C.J.D. Fell and A.G. Waters, The relationship between surface pore characteristics and flux for
UF membranes, J. Membrane, Sci., 9 (1981) 245.
12 U. Merin and M. Cheryan, A study of the foulingphenomena during ultrafiltration of cottage cheese whey,
J. Appl. Pol. Sci., 2 (1980) 2139.
13 L. Zeman and G. Tkacik, Characterization of porous
sublayers by thermoporometry, J. Membrane Sci., 32
(1987) 329.
14 A. Leenaars, Ceramic membranes, Thesis University
of Twente, The Netherlands, 1984.
15 R. Defay and I. Prigogine, Surface Tension and Adsorption: The Effect of Curvature on the Triple Point,
Longmans, New York, NY, 1966, p. 244.
16 T. Kobayashi, M. Todoki, M. Kimura, Y. Fujii, T.
Takeyama and H. Tanzawa, Characterization of UF
membranes, in: E. Drioli and M. Nakayaki (Eds.),
Membranes and Membrane Processes, Plenum Press,
New York, NY, 1986, pp. 25-33.
17 Handbook of Chemistry and Physics, 58th edn., R.C.
Weast (Ed.), CRC Press, Boca Raton, FL, 1977.
18 International Critical Tables, 1st edn., Vol. iv, E.W.
Washburn (Ed.), MC Graw Hill, New York, NY, 1928.
19 R.H. Lange and J. Blodorn, Das Elektronenmikroskop, Thieme Verlag, Stuttgart, 1981, Chap. 9, pp. 192201.
20 F.P. Cuperus, D. Bargeman and C.A. Smolders, A new
method to determine the skin thickness of asymmetric UF membranes using colloidal particles, J. Colloid
Interface Sci., 135 (1990) 486.
21 F.P. Cuperus, Characterization of UF membranes:
Pore structure and top layer thickness, Thesis University of Twente, The Netherlands, 1990.
22 F.P. Cuperus and C.A. Smolders, Characterization of
UF membranes: membrane characteristics and characterization techniques, Adv. Colloid Interface Sci.,
34 (1991) 135.

You might also like