Anda di halaman 1dari 13

Journal of Membrane Science 463 (2014) 113125

Contents lists available at ScienceDirect

Journal of Membrane Science


journal homepage: www.elsevier.com/locate/memsci

SEM imaging of membranes: Importance of sample preparation


and imaging parameters
Syed Z. Abdullah a,n, Pierre R. Brub a, Derrick J. Horne b
a
b

2002 6250 Applied Science Lane, Vancouver, BC, Canada V6T 1Z4
0517 6270 University Blvd, Vancouver, BC, Canada V6T 1Z4

art ic l e i nf o

a b s t r a c t

Article history:
Received 16 August 2013
Received in revised form
18 March 2014
Accepted 19 March 2014
Available online 26 March 2014

SEM imaging has been used by researchers to characterize the membrane properties. However, limited
information is generally reported regarding the sample preparation and imaging parameters used,
making it difcult to compare results from different studies. The present study focuses on the inuences
of sample preparation and imaging parameters on the observed membrane surface properties obtained
from SEM images. Polymeric membranes were prepared with different coating metals (gold, gold
palladium and iridium) and coating thicknesses, and imaged with different imaging parameters
(magnication, detector, sample orientation). Image analysis was done using a custom Matlab application. The effects were observed both qualitatively and quantitatively (pore size/shape and porosity). All
of the sample preparation and imaging parameters affected the observed membrane properties of
interest. In general, the membrane properties were less sensitive to the parameters when membranes
were coated with iridium than with gold or goldpalladium. Iridium should be used as the coating metal
in the studies where coating thickness and/or, one or more imaging parameters need to be varied. The
results suggest the need of a standardized SEM procedure for membrane characterization, and clear
reporting of the parameters used in any SEM study, to enable appropriate comparison of the results
obtained in different studies.
& 2014 Elsevier B.V. All rights reserved.

Keywords:
SEM
Coating
Detector
Magnication
Membrane surface property

1. Introduction
Microltration (MF) and ultraltration (UF) membranes are
increasingly used for water/wastewater treatment applications
because of their ability to effectively remove contaminants of
concern. However, over time, the properties of the membrane
surface can change due to the accumulation of foulants [1] or due
to the weathering of the membrane material [2], both of which
affect the performance of membrane system. Therefore, a proper
understanding of the membrane surface properties is essential to
assess how these can change over time. Scanning Electron Microscopy (SEM) and High Resolution SEM (HRSEM), also known as
Field Emission SEM (FESEM) imaging, have been extensively used
to characterize the membrane surface properties, both qualitatively (e.g. visual observation) and quantitatively (e.g. pore size,
pore shape and porosity). A literature review of all the articles
published in the Journal of Membrane Science in the volumes
51100 (20102012) revealed that a relatively large number of
studies (i.e. 37 in total) have used SEM and/or HRSEM (hereafter

Corresponding author. Tel.: 1 604 783 8137.


E-mail address: zaki@civil.ubc.ca (S.Z. Abdullah).

http://dx.doi.org/10.1016/j.memsci.2014.03.048
0376-7388/& 2014 Elsevier B.V. All rights reserved.

referred simply as SEM unless otherwise indicated) to characterize


the pore size, pore size distribution, pore shape, porosity, surface
roughness, fouling, etc. of PVDF-based polymeric membranes
[1,337]. Although SEM imaging is commonly used to characterize
membrane surface properties, limited information has been
reported in the literature with respect to the sample preparation
(i.e. whether metal coating had been used and identifying the
coating material and coating thickness) and imaging parameters
(i.e. magnication, detector type, sample orientation) used, as
summarized in Table 1. As a result, no standard approach has
been adopted for SEM imaging of membranes.
Varying one or more sample preparation or imaging parameter(s),
especially the imaging parameter(s), is a common practice, and
sometimes necessary, in SEM imaging to obtain a best quality image
of a sample. As an example, taking images at different tilt angles, or
using different detectors, to obtain insight into the 3-dimensional
structure of a surface. As presented in Table 2 the sample preparation
and imaging parameters can signicantly affect the information
obtained from SEM imaging and therefore, the lack of a standard
approach for SEM imaging makes it difcult to compare results from
different studies.
The present study was conducted to quantify the effect of
sample preparation (i.e. coating metal and coating thickness) and

114

S.Z. Abdullah et al. / Journal of Membrane Science 463 (2014) 113125

imaging parameters (i.e. magnication, detector used, sample


orientation) on the observed membrane surface properties of
polymeric membrane obtained from SEM imaging. The effects
were observed both qualitatively (e.g. visual observation) and
quantitatively (e.g. pore size, pore shape and porosity). Although
explaining the detail mechanism of the observed effects was

Table 1
Summary of the information available in reviewed journal articles.
Parameters

Reported
information

Fraction of published
journals providing
information on the
parameters of interesta

Sample
preparation
parameters

Coating material
Coating thickness

28
4

Imaging
parameters

Magnication
Detector type
Sample orientation
(i.e. tilt angle)

30
2
0

beyond the scope of this work, mechanism(s) are suggested to


explain some of the observed effects.

2. Materials and methods


2.1. Membrane
Commercial hollow-ber membranes made from a proprietary
blend of PVDF and additives were used (GE Water and Technologies, Oakville, ON, Canada). Both articially and naturally aged
membranes were considered. Articial ageing was performed by
soaking virgin membranes in a 3600 ppm sodium hypochlorite
solution for 139 h, resulting in an exposure dose of
500,000 ppm h, which corresponds to the maximum lifetime
exposure recommended by the membrane manufacturer. Naturally aged membranes were harvested from a full-scale drinking
water ltration system and been exposed to a sodium hypochlorite
dose of approximately 500,000 ppm h over a 5 year period. The
bers were harvested immediately following chemical cleaning.
2.2. Sample preparation and SEM imaging

Total count based on review of all the journal articles published in the 2010
2012, in which SEM was used as a membrane characterization technique for PVDF
based polymeric membranes (i.e. 37 in total).

The membranes were dried overnight in a vacuum drier at


room temperature and then coated with a metal using a sputter

Table 2
Effect of the different sample preparation and imaging parameters on SEM imaging.
Parameter
Coating material

Effects on the SEM imaging


 Commonly used coating materials are gold (Au), goldplatinum (AuPt), goldpalladium (AuPd), chromium (Cr) and iridium (Ir).
 Metal used affects structure of coating.

Gold and gold composite: often result in coating with uneven thickness, and a relatively thick layer is needed to get a uniform coating.

Iridium/chromium: provides smoother coating and even coating thickness can be archived with a relatively thin layer [38].
 Rate of oxidation of coating metal depends on the metal used.

Samples become unusable if the coating layer becomes oxidized.

Chromium: coating layer becomes oxidized quickly (i.e. within few hours).

Iridium: has a relatively slow oxidation rate.

Coating thickness

Magnication
Detector type







Commonly used range for coating thickness is 25 nm.


Samples with an extremely thin layer of coating (i.e. o 2 nm) can be imaged at moderately high magnication.
However, a thin layer may result in an uneven distribution of metal, which can lead to charging, image drift, and/or beam damage to membrane.
Polymeric membranes can be extremely beam sensitive and can shrink, swell, or rupture when exposed to the electron beam [39].
A thick layer can obscure detail and/or add ne-scale texture. Also a thick layer can ll in nanoscale gaps and also can bridge nanopores, which
will lead to the inaccurate measurements of the membrane surface properties.

 Magnication used to image the membranes has impact on the resolution of the obtained images [40], which may inuence the information on
membrane surface properties obtained from the images.
 Typically SEM comes with two electron detectors.

Two detectors can be used independently (i.e. upper detector UD, and lower detector LD), or

in conjunction with one another (i.e. mixed detector MD).


 The choice of detector also implies a choice of signal [39].

Upper detector has high sensitivity, which results in a greater signal and produce images with high resolution.

Upper detector captures the image of only very top layer of the surface.

When some degree of subsurface imaging is desired, either lower or mixed detector mode is used.

The lower detector's sensitivity is about half that of the upper detector, so a higher beam energy or longer working distance is typically
needed to collect the required signal. This results in a noisier or grainier image with low resolution than is normally produced using the
upper detector.

The mixed detector mode is useful for giving a composite surface and subsurface image, which can take on a 3D-like effect. The
combination of signal may make image interpretation difcult when trying to determine true surface features.

Sample orientation
(i.e. tilt angle)

 Tilting the sample relative to the electron beam increases the signal magnitude, and presents some vertical relief of the membrane surface to
the beam, making the roughness information more evident [39].
 Tilting the sample highlights smaller differences in vertical relief compared to samples that are orthogonal (i.e. 01 tilt angle) to the beam.
 However, the image becomes somewhat foreshortened on the y-axis, while maintaining the x-axis dimensionality.
 Tilting of the membrane samples during SEM imaging is common, as typically membranes have minimal surface topography. Orthogonal
viewing of the membranes results in a at view of the surface, which may be more appropriate for measuring pore diameter, as this provides a
truer representation of the two-dimensional shape of the surface pores. However, this comes at the expense of being able to easily identify pore
structure (e.g. false detection of pores, as a surface depression may appear as a pore).

S.Z. Abdullah et al. / Journal of Membrane Science 463 (2014) 113125

Table 3
Sample preparation and imaging parameters considered.
Sample preparation
parameters

Imaging parameters

Coating
metal

Coating
thickness

Detectora Magnication Tilt angle of the


sample holder
base (deg)

Goldb

2 nm
5 nm

L
M
U

25k
50k

0
30

Iridiumb

2 nm
5 nm

L
M
U

25k
50k

0
30

Goldpalladium

2 nm
5 nm

L
M
U

25k
50k

0
30

Note: All the parameters were used for articially aged membranes.
a
b

converted to gray scale binary images and pixels with less than a
certain value (i.e. threshold value) were considered belonging to
the pores. Threshold values were empirically found by visual
observation of the efciency of the pore identication (i.e. maximizing the detection of the obvious pores and minimizing the
detection of false pores) [41]. Threshold values did not differ more
than 2% for the images obtained with any given parameter
combinations. The pore pixels in contact with each other were
considered belonging to a single pore. The area of a pore was
calculated by multiplying the total number of the pixels in that
pore to the area equivalent of a pixel. The porosity was calculated
by dividing the total cumulative number of the pixels in the pores
by the total number of the pixels in a SEM image. The pore
diameter was the diameter of a circle, which was equal in area to
the estimated area of the pores by image analysis. The pore aspect
ratio was the average ratio of the major axis length to the minor
axis length of an ellipse equivalent in shape to that of a pore. Fig. 1
summarizes the work ow diagram used for the image analysis.

L lower, M mixed, U upper.


Parameter used for naturally aged membranes.

coater (Cressington 208C HR) to achieve a specic coating thickness. A quartz crystal monitor (Ted Pella Inc. MIM 10) with a
resolution of 0.1 nm was used to monitor the coating thickness
during the coating process. A Field Emission Scanning Electron
Microscope was used (Hitachi S-4700) to acquire the images. The
sample preparation (i.e. coating metals and coating thickness) and
imaging parameters (i.e. detector, magnication and tilt angle)
considered for the two types of membranes are presented in
Table 3. These were selected to cover the range of parameters
reported in the reviewed literature.
The upper detector was a through-the-lens (TTL) type detector; the lower detector was an EverhartThornley (ET) type
detector, and the mixed detector was a xed combination of the
upper and lower detectors. Electron energy of 5 kV and recording
time of 80 s were used to acquire the images. The working
distance during the imaging was maintained as close as possible
to 5 mm. However, for only few parameter combinations, a
minimum of 12 mm working distance needed to be used while
imaging with UD. The xed focus technique was used during the
imaging of all the samples.
It should be noted that, the magnication is dependent on the
size of the output medium (e.g. computer screen). The pixel size of
the SEM image is the more appropriate parameter to express the
sizes of the objects in an image. For the present study, the pixel sizes
were 3.968  10  9 m and 1.984  10  9 m for the images obtained
with 25k magnication and 50k, respectively. However, it is common
to express the SEM images in terms of magnication. In the
discussion which follows, images are labeled according to the sample
preparation (i.e. coating metals and coating thickness) and imaging
parameters (i.e. detector type, magnication and sample orientation)
used. As an example, an image labeled Au2nmLD25k30tilt was
obtained with a gold coating, a 2 nm coating thickness, a lower
detector, a 25k magnication and a 301 tilt angle.
2.3. Analysis of the SEM images
SEM imaging was used to quantitatively characterize the
membranes based on the porosity (), the average pore diameter
(dav), the 90th percentile of cumulative pore size distribution (d90),
and the pore aspect ratio (dt/dc). A Matlab application was
developed to obtain the membrane surface characteristics from
the SEM images, using the Matlab and Simulink software package
(MathWorks, Inc.). During the image analysis RGB images were

115

Fig. 1. Flow chart for the SEM image analysis.

116

S.Z. Abdullah et al. / Journal of Membrane Science 463 (2014) 113125

2.4. Parameter sensitivity


It is foreseeable that the experimental design of a study could
require the variation of one of sample preparation or imaging
parameter; for example, taking images at different tilt angles to
elucidate the 3-dimensional structure of a surface. In this case, the
other parameters should be selected to provide the most consistent image quality, when varying the parameter of interest.
To identify the sample preparation and imaging parameters,
which have the least effects (i.e. least sensitivity) on the observed
membrane properties (obtained from the SEM images), a comparative analysis was performed, where all but one of the sample
preparation and imaging parameters were xed, while the
remaining parameter was varied. The effect of changes in the
variable parameters on the membrane properties of interest was
quantied as t statistic. To judge whether the observed mean
values of a membrane property of interest obtained with two
different parameters were signicantly different, Student's t-test
was performed using the following equation:
x1  x2
p
t statistics
Sx1 x2 2=n

where x1 and x2 are the mean values of a membrane property of


interest (e.g. ) obtained with a rst (e.g. iridium) and a second
(e.g. gold) sample preparation or imaging parameter respectively,
while the remaining sample preparation or imaging parameters
were xed; Sx1 x2 is the grand standard deviation; and n is the
number of samples (i.e. 5). As an example, the statistical difference
in the values due to the use of different coating metals (i.e.
iridium and gold) for the images obtained with 2 nm coating, and
using lower detector, 50k magnication and 01 tilt angle was
calculated using Eq. (1) as expressed in the following equation:
t statisticsIr  Au2nmLD50K0tilt

Ir2nmLD50K0tilt  Au2nmLD50K0tilt
p
SIr2nmLD50K0tilt Au2nmLD50K0tilt 2=n

the images was low. Also, the coefcient of variation values for
, dav, d90 and dt/dc were 0.1, 0.016, 0.025 and 0.002, respectively,
which indicated that the measurements were reproducible.
It should be noted that, the smallest pore size detectable was
limited by the size of the pixels (i.e. magnication) used during
imaging. For the present study, the smallest pore diameters were
0.012 m and 0.006 m for the images obtained with 25k and 50k
magnication, respectively.
The values of , dav, d90 and dt/dc obtained from the SEM
images acquired with different parameter combinations, ranged
from 0.04 to 0.22, 0.017 to 0.042 m, 0.023 to 0.066 m, and 1.7 to
2.3, respectively. These values are comparable with the published
values of (i.e. 0.060.15) and dav (i.e. 0.0150.045 m) for the
type of membrane used in the present study [42,43]. This
indicated that the SEM imaging and the image analysis techniques
used in the present study were capable of capturing the actual
membrane surface properties.
3.2. Effect of sample preparation and imaging parameters
on the SEM imaging on articially aged membranes
3.2.1. Qualitative analysis
As done with most of the manuscript reviewed, a preliminary
analysis was rst performed to obtain qualitative information on
the membranes, based on a visual observation of the SEM images.
As presented in Fig. 4, the sample preparation and imaging
parameters had a substantial effect on the appearance of the
SEM images. Images of membranes coated with gold (Fig. 3ac)
suggested that these membranes had greater surface roughness
(i.e. higher density and/or relatively larger height of the mountainous peaks) than membranes coated with iridium (Fig. 3df) or
goldpalladium (Fig. 3gi), for a given combination of imaging

2
3. Results and discussion
3.1. Typical results from SEM imaging and image analysis
A typical image obtained by SEM imaging and the pores
identied by the custom Matlab application are presented in Fig. 2.
The repeatability of the measurement performed was assessed
by quantifying the spread of the results obtained, when a randomly selected (i.e. images were collected from 3 different membrane samples in different collection periods) subset of 5 images
acquired with a given combination of sample preparation and
imaging parameters (e.g. Ir2nmUD50k0tilt) was considered. As
presented in Fig. 3, the spread of the pore size distributions for 5 of

Fig. 3. Typical pore size distribution. (Randomly selected subset of 5 images;


Ir2nmUD50k0tilt.)

Fig. 2. Typical SEM images of articially aged membrane: (a) original image and (b) pores identied with the custom Matlab application; Au2nmUD50k0tilt.

S.Z. Abdullah et al. / Journal of Membrane Science 463 (2014) 113125

117

Fig. 4. Typical HRSEM images of articially aged membranes: (a) Au2nmLD25k30tilt; (b) Au2nmMD25k30tilt; (c) Au2nmUD25k30tilt; (d) Ir2nmLD25k30tilt;
(e) Ir2nmMD25k30tilt; (f) Ir2nmUD25k30tilt; (g) AuPd2nmLD25k30tilt; (h) AuPd2nmMD25k30tilt; (i) AuPd2nmUD25k30tilt; (j) Au5nmUD25k30tilt;
(k) Au5nmUD50k30tilt; and (l) Au5nmUD50k0tilt; insets are magnied (3.6  ) versions of the original images.

parameters. Coating with goldpalladium resulted in images of


membranes that appeared to have the least surface roughness.
The same trend was observed for all of the combinations of
imaging parameters used (result are not shown for all combinations). Images of membranes obtained with the LD appeared to
have the greatest surface roughness (Fig. 3a, d, and g). However,
these images had poorly dened features regardless of the coating
metal used. Images of membranes obtained with the MD (Fig. 3b,
e, and h) and the UD (Fig. 3c, f, and i) were similar, and suggested
that the membranes had lower surface roughness and more
dened features than those imaged with the LD. Images of
membranes obtained with a thinner coating thickness (i.e. 2 nm
gold) (Fig. 3c) appeared to have less surface roughness than those
obtained with a thicker coating thickness (i.e. 5 nm gold) (Fig. 3j),
while those obtained with a lower magnication (i.e. 25k) (Fig. 3j)

appeared to have a lower surface roughness than those obtained


with a higher magnication (i.e. 50k) (Fig. 3k). Also, images of
membranes obtained with a 301 tilt angle (Fig. 3k) suggested that
these membranes had a greater surface roughness and a less
denite pore boundary than those imaged with a 01 tilt angle
(Fig. 3l). These results conclude that the qualitative information
(i.e. surface roughness, well-dened features, pore boundary
sharpness) obtained from SEM images varies based on ALL of the
sample preparation and imaging parameters considered.
3.2.2. Quantitative analysis
To expand the qualitative analysis, a quantitative analysis of the
SEM images was performed. Fig. 5 presents an example of the
extent to which sample preparation parameter can have an effect
on the membrane properties of interest (i.e. , dav, d90 and dt/dc).

118

S.Z. Abdullah et al. / Journal of Membrane Science 463 (2014) 113125

Fig. 5. Average values of the observed membrane properties of interest obtained


from SEM images acquired with 5 nm coating, UD, 50k magnication, and 01 tilt.
(a) Porosity; (b) pore diameter; (c) 90th percentile of cumulative pore size
distribution; (d) pore aspect ratio; error bars correspond to the 95% condence
interval of the repeated measurements.

3.2.2.1. Impact of individual parameters while the combination of the


other parameters were xed. The results of the Student's t-test to
assess the effects of changes in the variable parameters on the
membrane properties of interest (i.e. , dav, d90 and dt/dc) were
presented in Figs. 610. Critical t statistic value for a two-tailed
test, with a signicance level of 0.05, and 8 degrees of freedom (i.e.
2n  2 8) is 2.306. When the calculated t statistic values were
outside of the range of 2.306 to 2.306 (i.e. outside of the shaded
area in Figs. 610), it was considered that the observed mean
values of a membrane property of interest obtained with two
different parameters were statistically signicantly different. The
positive value of the t statistic indicated that the observed mean
value of a membrane property of interest (e.g. ) was larger for
the images obtained with the rst (e.g. iridium) parameter than
that of the images obtained with the second (e.g. gold) parameter.
The negative value of the t statistic indicated the opposite trend.
As an example, the t statistics values for the membrane
properties of interest (i.e. , dav, d90 and dt/dc) for the images
obtained with different coating metals (i.e. iridium vs. gold
(Fig. 6a) and iridium vs. goldpalladium (Fig. 6b)), with the other
parameters (i.e. 5 nm coating, upper detector, 25k magnication
and 301 tilt angle) xed, have been presented inside the dotted box
in Fig. 6a and b. The legend 5nmUD25K0tilt has been used to

represent the xed parameters (i.e. 5 nm coating, upper detector,


25k magnication and 301 tilt angle) for the t statistic values for
both Fig. 6a and b. The positive t statistic values of the top four
bars in Fig. 6a indicated that the images obtained with iridium had
larger values for the membrane properties of interest (i.e. , dav,
d90 and dt/dc) than those of the images obtained with gold, while
the other parameters (i.e. 5 nm coating, lower detector, 25k
magnication and 301 tilt angle) were xed. Also, the t statistic
values for , dav and d90 were outside the range of  2.306 to
2.306 (i.e. outside of the shaded area), indicating that the observed
mean values of , dav and d90 were statistically signicantly different.
As presented in Figs. 610, the sample preparation and imaging
parameters had a substantial effect on the quantitative properties
of interest. The effect of the variation in each parameter on
apparent membrane properties was considered while all other
parameters were kept constant.
The effect of changing the coating material was considered
rst. Presented in Fig. 6 are the differences in the membrane
properties of interest (i.e. , dav, d90 and dt/dc) expressed as the t
statistics values, for each of the sample preparation and imaging
parameter combinations considered when varying the coating
material. Fig. 6a presents the results when comparing iridium
(x1) and gold (x2) coatings, while Fig. 6b presents the results when
comparing iridium (x1) and goldpalladium (x2) coatings. The t
statistic values for , dav and d90 were outside the range of  2.306
to 2.306 (i.e. outside of the shaded area), for most of the sample
preparation and imaging parameter combinations. These results
suggested that the images of membranes coated with iridium had
signicantly different , dav, and d90 values than those coated with
gold (Fig. 6a) or goldpalladium (Fig. 6b), for most of the sample
preparation and imaging parameter combinations. Coating metals
did not affect only the dt/dc values obtained from the images of
membranes for most of the sample preparation and imaging
parameter combinations. These results demonstrated that the
coating metal has a signicant effect on the observed values for
the membrane properties of interest. Hence, the type of metal should
be standardized if results from different studies are to be compared.
It should be noted that the images obtained using several
parameter combinations for goldpalladium coated samples
resulted in images with poorly dened features and, as a result,
no quantitative information could be obtained from those images.
Additional research, beyond the scope of the present study would
be required to determine the causes of this limitation. For this
reason, in the following section (Figs. 710), results from the samples
prepared with goldpalladium coating were not considered.
Presented in Fig. 7 are the t statistic values for the membrane
properties of interest (i.e. , dav, d90 and dt/dc) for each of the
sample preparation and imaging parameter combinations considered when varying the coating thickness. Fig. 7a presents the
results when comparing 2 nm (x1) and 5 nm (x2) gold coatings,
while Fig. 7b presents the results when considering 2 nm (x1) and
5 nm (x2) iridium coatings. Images of the membranes coated with
2 nm of gold suggested that these membranes had statistically
signicant (i.e. values were outside the critical t statistic range of
 2.306 to 2.306, and hence outside of the shaded area) lower
apparent dt/dc values, but higher dav, d90 values than those coated
with 5 nm of gold (Fig. 7a) for most of the sample preparation and
imaging parameter combinations. Also for the images of membranes coated with 2 nm of gold suggested that these membranes
had a variable trend for values (i.e. lower and higher for ve and
two of the 11 sample preparation and imaging parameter combinations, respectively) than those coated with 5 nm of gold
(Fig. 7a). Gold is known to have a tendency to agglomerate during
sputter coating creating uneven surface coverage [38]. This may
have contributed to the observed lower dt/dc values, due to an
increased aspect ratio with narrowing pore proles; the same

S.Z. Abdullah et al. / Journal of Membrane Science 463 (2014) 113125

119

Fig. 6. Effects of coating metal on the average observed membrane properties of interest. (t Statistics values for the membrane properties of interest (i.e. , dav, d90 and dt/dc)
for the images obtained with different coating metals, (a) iridium vs. gold and (b) iridium vs. goldpalladium; n were placed for the parameter combinations for which the
values were missing for at least one of the comparing conditions.)

effect could account for lower dav, d90 for membranes coated with
thicker gold coating as some of the smaller pores could be entirely
covered with thicker gold coating. The tendency of agglomeration may
also have contributed to the observed lower values, and higher dav
and d90 for some sample preparation and imaging parameter combinations, which is mainly due to the detection of the different numbers
of pores per unit membrane area. The t statistic values for the images
of membranes coated with 2 nm and 5 nm of iridium suggested that
these membranes had similar dt/dc, , dav and d90 values (i.e. values
are within the shaded area) (Fig. 7b) for most of the sample
preparation and imaging parameter combinations.
The results demonstrated that the observed membrane properties of interest obtained from the SEM images of the membranes
coated with gold were sensitive to the coating thickness. However,
the sensitivity was statistically insignicant for the membranes

coated with iridium, for most of the sample preparation and


imaging parameter combinations investigated. Hence, if gold is
to be used as a coating metal, the coating thickness should be
standardized if results from different studies are to be compared.
On the other hand, if iridium is used as a coating metal, different
coating thicknesses for different samples can be used during
sample preparation, if necessary. However, when the value is
the major membrane property of interest, care should be taken
before varying iridium coating thickness, as many of the parameter combinations showed sensitivity of the values to the
change in coating thickness (e.g. IrUD25K30tilt and IrLD50K0tilt).
Presented in Fig. 8 are the t statistic values for the membrane
properties of interest (i.e. , dav, d90 and dt/dc) for each of the
sample preparation and imaging parameter combinations considered when varying the detector type. Fig. 8a and b presents the

120

S.Z. Abdullah et al. / Journal of Membrane Science 463 (2014) 113125

Fig. 7. Effects of coating thickness on the average observed membrane properties of interest. (a) t Statistics values for the membrane properties of interest (i.e. , dav, d90 and
dt/dc) for the membranes coated with 2 nm and 5 nm (a) gold and (b) iridium; n was placed for the combination for which the values were missing for at least one of the
comparing conditions.

results when comparing LD (x1) and UD (x2) detectors with gold


and iridium coatings, respectively, while Fig. 8c and d presents the
results when considering either MD (x1) or UD (x2) detector with
gold and iridium coatings, respectively. In general, membranes
coated with either gold or iridium had lower dt/dc values and
higher dav and d90 values when imaged with the LD than those
obtained using the UD (Fig. 8a and b), for most of the sample
preparation and imaging parameter combinations. Although no
consistent trend was observed for the changes in the values, the
images of membranes coated with either gold or iridium, and
imaged with the LD, suggested that these membranes had signicantly different values than those obtained using the UD, for
several parameter combinations (Fig. 8a and b). Images of membranes coated with gold suggested that these membranes had
signicantly different dt/dc, , dav and d90 values when imaged
with the MD than those obtained using the UD, for most of the
sample preparation and imaging parameter combinations (Fig. 8c).
In contrast to that, images of membranes coated with iridium
suggested that these membranes had similar dt/dc, , dav and d90
values for images obtained with the UD and MD (Fig. 8d) for most
of the sample preparation and imaging parameter combinations.
The results demonstrated that the observed membrane properties of interest were sensitive to all the detectors used during
imaging of the membranes coated with gold (i.e. use of LD, MD
and UD all resulted in different values for membrane properties of
interest for any given parameter combinations). Similarly, the
observed membrane properties of interest were sensitive to the
detectors used during imaging of the membranes coated with
iridium. However, the sensitivity was lower for the membranes
coated with iridium than that of membranes coated with gold.
Also, the use of MD and UD during imaging of iridium-coated
membranes resulted in very little sensitivity of membrane properties of interest for any given parameter combination.

Hence, if gold is to be used as a coating metal, the use of detectors


should be standardized if results from different studies are to be
compared. On the other hand, if iridium is used as a coating metal, UD
and MD can be used interchangeably during imaging, if necessary.
Presented in Fig. 9 are the t statistic values for the membrane
properties of interest (i.e. , dav, d90 and dt/dc) for each of the
sample preparation and imaging parameter combinations considered when varying the magnication. Fig. 9a presents the results
when comparing 25k (x1) and 50k (x2) magnication with gold
coating, while Fig. 9b presents the results when considering either
25k (x1) or 50k (x2) magnication with iridium coating. The
magnication did not affect the dt/dc values obtained from the
images of membranes coated with either gold (Fig. 9a) or iridium
(Fig. 9b), for most of the sample preparation and imaging parameter combinations. However, the imaging suggested that membranes had signicantly higher , dav and d90 values with 25k
magnication than with 50k magnication (Fig. 9a and b). The
lower pore diameter values observed with 50k magnication is
likely related to the greater number of smaller pores that can be
detected at a higher magnication and/or to the erroneous grouping of closely spaced pores into a single large pore which was
observed to periodically occur at a lower magnication.
The results demonstrated that the dav and d90 values were very
sensitive to the magnication used during imaging of the membranes coated with either gold or iridium. However, the sensitivity
was higher (i.e. relatively larger t statistic values; results did not
show the full scale) for the membranes coated with gold than that
of membranes coated with iridium for most of the sample
preparation and imaging parameter combinations investigated.
Hence, when the dav and d90 values are the major membrane
properties of interest, magnications should be standardized if
results from different studies are to be compared. This is true for
both gold- and iridium-coated membranes.

S.Z. Abdullah et al. / Journal of Membrane Science 463 (2014) 113125

121

Fig. 8. Effects of detector on the average observed membrane properties of interest. (t Statistics values for the membrane properties of interest (i.e. , dav, d90 and dt/dc) when
imaged with the lower detector than those obtained using the upper detector for the samples coated with (a) gold and (b) iridium; t statistics values for the membrane
properties of interest (i.e. , dav, d90 and dt/dc) when imaged with the mixed detector than those obtained using the upper detector for the samples coated with (c) gold and
(d) iridium; n was placed for the combination for which the values were missing for at least one of the comparing conditions.)

The nal parameter considered was tilt angle. Presented in


Fig. 10 are the t statistic values for the membrane properties of
interest (i.e. , dav, d90 and dt/dc) for each of the sample preparation and imaging parameter combinations considered when varying the tilt angle. Fig. 10a presents the results when considering
either 01 (x1) or 301 (x2) tilt angle with gold coating, while Fig. 10b
presents the results when considering either 01 (x1) or 301 (x2) tilt
angle with iridium coating.
Images of membranes coated with gold suggested that these
membranes had higher dt/dc, , dav and d90 values when imaged
with a 01 tilt angle than those imaged with a 301 tilt angle
(Fig. 10a), for more than half of the sample preparation and
imaging parameter combinations. The effect was more pronounced for the samples coated with 5 nm gold than 2 nm gold.
Shielding of some proportion of the pores, which can occur in
titled positions due to the rough nature of the gold-coated
surfaces, likely contributed to the lower observed values, and
thicker coating appears to worsen this phenomenon. For the
images of membranes coated with iridium, there were changes
in the values of the dt/dc, , dav and d90 for some parameter

combinations with changing tilt angle; however, no consistent


trend was observed. Also the effects of tilt angle on the membrane
properties of interest were signicantly less pronounced for
iridium-coated membranes (Fig. 10b) than for gold-coated membranes (Fig. 10a). The relatively smooth surface coating generated
with iridium likely decreases potential shielding effect.
The results demonstrated that the observed membrane properties of interest were sensitive to the tilt angle used during imaging
of the membranes coated with either gold or iridium. However,
the sensitivity was lower for the membranes coated with iridium
(i.e. relatively lower t statistic values) than that of membranes
coated with gold for most sample preparation and imaging
parameter combinations.
Hence, if gold is to be used as a coating metal, the tilt angle
should be standardized if results from different studies are to be
compared. Special care should be taken for thicker gold coating.
On the other hand, if iridium is used as a coating metal, different
tilt angles for different samples can be used during imaging, for
most of the parameter combinations. However, when the value
is the major membrane properties of interest, care should be taken

122

S.Z. Abdullah et al. / Journal of Membrane Science 463 (2014) 113125

Fig. 9. Effects of magnication on the average observed membrane properties of interest. (t Statistics values for the membrane properties of interest (i.e. , dav, d90 and dt/dc)
when imaged with 25k magnication than those obtained using 50k magnication for the samples coated with (a) gold and (b) iridium; n was placed for the combination for
which the values were missing for at least one of the comparing conditions.)

Fig. 10. Effects of tilt angle on the average observed membrane properties of interest. (t Statistics values for the membrane properties of interest (i.e. , dav, d90 and dt/dc)
when imaged using 01 angle than those obtained using 301 tilt angle for the samples coated with (a) gold and (b) iridium; n was placed for the combination for which the
values were missing for at least one of the comparing conditions.)

S.Z. Abdullah et al. / Journal of Membrane Science 463 (2014) 113125

before using different tilt angles, as several of the parameter


combinations showed relatively large sensitivity of the values
to the change in tilt angle (i.e. Ir5nmUD25K and Ir2nmLD50K).
It should be noted that no manipulation of the images obtained
with 301 tilt angle had been conducted to correct for the tilt.
Correcting the images for the tilt before the image analysis might
affect the outcomes of the analysis. However, one of the main

Table 4
P-values for membrane properties of interest obtained from the multi-way ANOVA
analysis.
Source of variation

Coating metal
Coating thickness
Detector
Magnication
Tilt Angle

Membrane properties of interest

dav

d90

dt/dc

0.0487
0.0006
0.3647
0.0005
0.0186

0.8164
0.4752
o 0.0001
o 0.0001
0.0172

0.3565
0.3956
o 0.0001
o 0.0001
0.0067

0.0155
o 0.0001
o 0.0001
0.0296
o 0.0001

Note: Membrane properties with P-value r 0.05 identied in bold are signicantly
affected by individual or interactive sample preparation or imaging parameters.

123

objectives of the present study was to show that the tilt angle has
signicant impact on the obtained images for both qualitative and
quantitative analyses. To address this research question, no
manipulation of the images for tilt was conducted.

3.2.2.2. Impact of individual parameters regardless of the combination


of the other parameters. A multi-factor multi-way analysis of variance
(ANOVA) was performed to identify the sampling preparation and
imaging parameters that signicantly affected the different
membrane properties of interest, regardless of the combination of
the parameters. Statistics Toolbox anovan was used to conduct
ANOVA (MathWorks, Inc.). As presented in Table 4, the and dt/dc
values were affected by all of the sample preparation and imaging
parameters considered (based on a 95% condence interval).
Similarly, dav and d90 values were affected by majority of the
sample preparation and imaging parameters considered. The only
exceptions were , which was not affected by the detector used, and
dav and d90, which were not affected by either the coating metal or
coating thickness. The results indicate that without a standard
approach for SEM imaging, it is not possible to compare results
from different studies.

Fig. 11. Typical HRSEM images of fouled membranes: (a) Au5nmMD25k30tilt; (b) Ir5nmMD25k30tilt; (c) Ir5nmMD25k0tilt; (d) Au5nmUD25k30tilt; (e) Ir5nmUD25k30tilt;
and (f) Ir5nmUD25k0tilt.

124

S.Z. Abdullah et al. / Journal of Membrane Science 463 (2014) 113125

3.3. Effect of sample preparation and imaging parameters on the


FESEM imaging on naturally aged membranes
The analysis presented in the previous section was performed on
articially aged membranes, which were not fouled. Typical SEM
images of fouled membranes obtained with different sample preparation and imaging parameters are presented in Fig. 11. The images
suggest that the surface roughness of the fouled membranes coated
with gold was greater than that observed when fouled membranes
were coated with iridium, regardless of the sample preparation or
imaging parameters considered. The surface roughness of membranes
imaged with a mixed detector (Fig. 11b and c) or a 301 tilt angle
(Fig. 11b and e) appeared to be greater than for those imaged with an
upper detector (Fig. 11e and f), or a 01tilt angle (Fig. 11c and f). The
above observations are consistent with those for articially aged
membranes and indicate that all of the sample preparation and
imaging parameters considered affect the SEM images and the
conclusions that can be drawn from these images. When imaging
naturally aged membranes with SEM, at least some of the membrane
pore structure generally cannot be observed, as it is typically covered
with an irreversible surface foulant structure, especially when dealing
with relatively old membranes such as those used in the present
study. Although no quantitative information can be obtained from the
images of these heavily irreversibly fouled membranes, similar trends
(i.e. effect of sample preparation and imaging parameters on the
properties of interest) to those obtained when imaging articially aged
membranes are expected when imaging less irreversibly fouled
membranes.

4. Conclusions
Images obtained for both types of membranes (i.e. articially and
naturally aged membranes) were substantially affected by the sample
preparation and imaging parameters considered. The images suggest
that the surface roughness of the membranes coated with gold was
greater than those coated with iridium, regardless of the sample
preparation or imaging parameters considered. The surface roughness
of membranes imaged with a mixed or lower detector, or a 301 tilt
angle also appeared to be greater than for those imaged with an upper
detector or a 01 tilt angle.
All of the sample preparation and imaging parameters signicantly
affect the observed surface properties of interest for membranes. In
general, the apparent membrane surface properties were less sensitive
to the sample preparation or imaging parameters when membranes
were coated with iridium than with gold. Hence, iridium should be
used as the coating metal in the studies where coating thickness and/
or, one or more imaging parameters need to be varied. However,
when the values are the major membrane properties of concern,
care should be taken before using different iridium coating thicknesses
or different tilt angles, as several of the parameter combinations
showed signicant sensitivity of the values to the change in coating
thickness or tilt angle. Also, for iridium-coated membranes, UD and
MD can be used interchangeably during imaging, if necessary.
When the dav and d90 values are the major membrane properties of concern different magnications (i.e. different pixel sizes of
the SEM images) should not be used during imaging, as this will
introduce articial variability among the similar samples and
comparison between different samples will become difcult. This
is true for both the membranes coated with either gold or iridium.
To be able to compare the results obtained in different studies, it is
necessary to conduct the studies with a x set of parameters. This
suggests the need of a standardized SEM procedure for membrane
surface characterization to enable appropriate comparison. Finally, to
facilitate comparison of the SEM results, all studies should clearly
report the sample preparation and imaging parameters used.

Acknowledgments
This work has been supported by the Natural Science and
Engineering Research Council of Canada Collaborative Research and
Development Grant (CRDPJ 413062-11) in Partnership with GE Water
and Process Technologies. The authors are grateful for the assistance
provided by Ms. C. Wang from the University of British Columbia,
Vancouver, for the literature review.

References
[1] C. Regula, E. Carretier, Y. Wyart, M. Sergent, G. Gsan-Guiziou, D. Ferry, et al.,
Ageing of ultraltration membranes in contact with sodium hypochlorite and
commercial oxidant: experimental designs as a new ageing protocol, Sep.
Purif. Technol. 103 (2013) 119138.
[2] X. Yan, M.R. Bilad, R. Gerards, L. Vriens, A. Piasecka, I.F.J. Vankelecom,
Comparison of MBR performance and membrane cleaning in a single-stage
activated sludge system and a two-stage anaerobic/aerobic (A/A) system for
treating synthetic molasses wastewater, J. Membr. Sci. 394395 (2012)
4956.
[3] F. Shi, Y. Ma, J. Ma, P. Wang, W. Sun, Preparation and characterization of PVDF/
TiO2 hybrid membranes with different dosage of nano-TiO2, J. Membr. Sci. 389
(2012) 522531.
[4] M. Hashino, K. Hirami, T. Ishigami, Y. Ohmukai, T. Maruyama, N. Kubota, et al.,
Effect of kinds of membrane materials on membrane fouling with BSA, J.
Membr. Sci. 384 (2011) 157165.
[5] C. Dong, G. He, H. Li, R. Zhao, Y. Han, Y. Deng, Antifouling enhancement of poly
(vinylidene uoride) microltration membrane by adding Mg(OH)2 nanoparticles, J. Membr. Sci. 387388 (2012) 4047.
[6] N. Pezeshk, D. Rana, R.M. Narbaitz, T. Matsuura, Novel modied PVDF
ultraltration at-sheet membranes, J. Membr. Sci. 389 (2012) 280286.
[7] C.-Y. Hung, S.-D. Li, C.-C. Wang, C.-Y. Chen, Inuences of a bipolar membrane
and an ultrasonic eld on alkaline water electrolysis, J. Membr. Sci. 389 (2012)
197204.
[8] S.-K. Mah, C.P. Leo, T.Y. Wu, S.-P. Chai, A feasibility investigation on ultraltration of palm oil and oleic acid removal from glycerin solutions: ux decline,
fouling pattern, rejection and membrane characterizations, J. Membr. Sci. 389
(2012) 245256.
[9] S. Claes, P. Vandezande, S. Mullens, K. De Sitter, R. Peeters, M.K. Van Bael,
Preparation and benchmarking of thin lm supported PTMSP-silica pervaporation membranes, J. Membr. Sci. 389 (2012) 265271.
[10] M.A. Masuelli, M. Grasselli, J. Marchese, N.A. Ochoa, Preparation, structural and
functional characterization of modied porous PVDF membranes by -irradiation, J. Membr. Sci. 389 (2012) 9198.
[11] R. Naim, A.F. Ismail, A. Mansourizadeh, Preparation of microporous PVDF
hollow ber membrane contactors for CO2 stripping from diethanolamine
solution, J. Membr. Sci. 392393 (2012) 2937.
[12] Z. Zhao, J. Zheng, M. Wang, H. Zhang, C.C. Han, High performance ultraltration membrane based on modied chitosan coating and electrospun nanobrous PVDF scaffolds, J. Membr. Sci. 394395 (2012) 209217.
[13] Y. Sui, X. Gao, Z. Wang, C. Gao, Antifouling and antibacterial improvement of
surface-functionalized poly(vinylidene uoride) membrane prepared via
dihydroxyphenylalanine-initiated atom transfer radical graft polymerizations,
J. Membr. Sci. 394395 (2012) 107119.
[14] S. Liang, K. Xiao, Y. Mo, X. Huang, A novel ZnO nanoparticle blended
polyvinylidene uoride membrane for anti-irreversible fouling, J. Membr.
Sci. 394395 (2012) 184192.
[15] M. Khayet, C. Cojocaru, M.C. Garca-Payo, Experimental design and optimization of asymmetric at-sheet membranes prepared for direct contact membrane distillation, J. Membr. Sci. 351 (2010) 234245.
[16] A. Mansourizadeh, A.F. Ismail, M.S. Abdullah, B.C. Ng, Preparation of polyvinylidene uoride hollow ber membranes for CO2 absorption using phaseinversion promoter additives, J. Membr. Sci. 355 (2010) 200207.
[17] M. Gryta, M. Barancewicz, Inuence of morphology of PVDF capillary membranes on the performance of direct contact membrane distillation, J. Membr.
Sci. 358 (2010) 158167.
[18] A. Cui, Z. Liu, C. Xiao, Y. Zhang, Effect of micro-sized SiO2-particle on the performance of PVDF blend membranes via TIPS, J. Membr. Sci. 360 (2010) 259264.
[19] C.-L. Li, D.-M. Wang, A. Deratani, D. Qumener, D. Bouyer, J.-Y. Lai, Insight into
the preparation of poly(vinylidene uoride) membranes by vapor-induced
phase separation, J. Membr. Sci. 361 (2010) 154166.
[20] P. Sukitpaneenit, T.-S. Chung, L.Y. Jiang, Modied pore-ow model for
pervaporation mass transport in PVDF hollow ber membranes for ethanol
water separation, J. Membr. Sci. 362 (2010) 393406.
[21] X. Yang, B. Deng, Z. Liu, L. Shi, X. Bian, M. Yu, et al., Microltration membranes
prepared from acryl amide grafted poly(vinylidene uoride) powder and their
pH sensitive behaviour, J. Membr. Sci. 362 (2010) 298305.
[22] S. Simone, A. Figoli, A. Criscuoli, M.C. Carnevale, A. Rosselli, E. Drioli, Preparation of hollow bre membranes from PVDF/PVP blends and their application
in VMD, J. Membr. Sci. 364 (2010) 219232.

S.Z. Abdullah et al. / Journal of Membrane Science 463 (2014) 113125

[23] A.F. Ismail, A. Mansourizadeh, A comparative study on the structure and


performance of porous polyvinylidene uoride and polysulfone hollow ber
membranes for CO2 absorption, J. Membr. Sci. 365 (2010) 319328.
[24] Y. Ye, L.N. Sim, B. Herulah, V. Chen, A.G. Fane, Effects of operating conditions
on submerged hollow bre membrane systems used as pre-treatment for
seawater reverse osmosis, J. Membr. Sci. 365 (2010) 7888.
[25] F. Liu, M.R.M. Abed, K. Li, Preparation and characterization of poly(vinylidene
uoride) (PVDF) based ultraltration membranes using nano -Al2O3,
J. Membr. Sci. 366 (2011) 97103.
[26] X.-M. Wang, X.-Y. Li, K. Shih, in situ embedment and growth of anhydrous and
hydrated aluminum oxide particles on polyvinylidene uoride (PVDF) membranes, J. Membr. Sci. 368 (2011) 134143.
[27] X. Yang, R. Wang, L. Shi, A.G. Fane, M. Debowski, Performance improvement of
PVDF hollow ber-based membrane distillation process, J. Membr. Sci. 369
(2011) 437447.
[28] S. Wongchitphimon, R. Wang, R. Jiraratananon, L. Shi, C.H. Loh, Effect of
polyethylene glycol (PEG) as an additive on the fabrication of polyvinylidene
uoride-co-hexauropropylene (PVDFHFP) asymmetric microporous hollow
ber membranes, J. Membr. Sci. 369 (2011) 329338.
[29] L. Guo, Y. Liu, C. Zhang, J. Chen, Preparation of PVDF-based polymer inclusion
membrane using ionic liquid plasticizer and Cyphos IL 104 carrier for Cr(VI)
transport, J. Membr. Sci. 372 (2011) 314321.
[30] D.-H. Kim, I.-H. Baek, S.-U. Hong, H.-K. Lee, Study on immobilized liquid
membrane using ionic liquid and PVDF hollow ber as a support for CO2/N2
separation, J. Membr. Sci. 372 (2011) 346354.
[31] P. Sukitpaneenit, T.-S. Chung, Molecular design of the morphology and pore
size of PVDF hollow ber membranes for ethanolwater separation employing
the modied pore-ow concept, J. Membr. Sci. 374 (2011) 6782.
[32] Y.-M. Zheng, S.-W. Zou, K.G.N. Nanayakkara, T. Matsuura, J.P. Chen, Adsorptive
removal of arsenic from aqueous solution by a PVDF/zirconia blend at sheet
membrane, J. Membr. Sci. 374 (2011) 111.
[33] Y.K. Ong, N. Widjojo, T.-S. Chung, Fundamentals of semi-crystalline poly(vinylidene
uoride) membrane formation and its prospects for biofuel (ethanol and acetone)
separation via pervaporation, J. Membr. Sci. 378 (2011) 149162.

125

[34] H. Li, Y.-M. Chen, X.-T. Ma, J.-L. Shi, B.-K. Zhu, L.-P. Zhu, Gel polymer
electrolytes based on active PVDF separator for lithium ion battery. I:
Preparation and property of PVDF/poly(dimethylsiloxane) blending membrane, J. Membr. Sci. 379 (2011) 397402.
[35] R. Halaui, A. Moldavsky, Y. Cohen, R. Semiat, E. Zussman, Development of
micro-scale hollow ber ultraltration membranes, J. Membr. Sci. 379 (2011)
370377.
[36] S.S. Madaeni, S. Zinadini, V. Vatanpour, A new approach to improve antifouling
property of PVDF membrane using in situ polymerization of PAA functionalized TiO2 nanoparticles, J. Membr. Sci. 380 (2011) 155162.
[37] S. Wongchitphimon, R. Wang, R. Jiraratananon, Surface modication of
polyvinylidene uoride-co-hexauoropropylene (PVDFHFP) hollow ber
membrane for membrane gas absorption, J. Membr. Sci. 381 (2011) 183191.
[38] R. Sebring, P. Arendt, J. Panitz, P. Yau, B. Imai, E. Bradbury, J. Gatewood,
Structural Characterization and Comparison of Iridium, Platinum and Gold/
Palladium Ultra-thin Film Coatings for STM of Biomolecules, U.S. Department
of Energy, Ofce of Scientic and Technical Information, 1997.
[39] Jeol USA, JEOL Guide to Scanning Microsocope Observation, JEOL USA, Inc.,
Peabody, MA, 2013.
[40] Y. Wyart, D. Nitsche, D. Chaudanson, D. Glucina, P. Moulin, The use of HRSEM
to characterize new and aged membranes in drinking water production,
Membr. Water Treat. 2 (2013) 251266.
[41] H. Reingruber, A. Zankel, C. Mayrhofer, P. Poelt, A new in situ method for the
characterization of membranes in a wet state in the environmental scanning
electron microscope, J. Membr. Sci. 399400 (2012) 8694.
[42] D.P. Saroj, G. Guglielmi, D. Chiarani, G. Andreottola, Subcritical fouling
behaviour modelling of membrane bioreactors for municipal wastewater
treatment: the prediction of the time to reach critical operating condition,
Desalination 231 (2008) 175181.
[43] A.M. ElHadidy, S. Peldszus, M.I. Van Dyke, Development of a pore construction
data analysis technique for investigating pore size distribution of ultraltration membranes by atomic force microscopy, J. Membr. Sci. 429 (2013)
373383.

Anda mungkin juga menyukai