Advances in Colloid and Interface Science 164 (2011) 12–20

Contents lists available at ScienceDirect

Advances in Colloid and Interface Science

j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / c i s

Positively charged nanofiltration membranes: Review of current fabrication methods

and introduction of a novel approach
Shuying Cheng, Darren L. Oatley, Paul M. Williams, Chris J. Wright ⁎
Centre for Complex Fluids Processing, Multidisciplinary Nanotechnology Centre, School of Engineering, Swansea University, Singleton Park, Swansea SA2 8PP, UK

a r t i c l e i n f o a b s t r a c t

Available online 20 January 2011 A review of the fabrication processes currently available to produce positively charged nanofiltration
membranes has been conducted. The review highlights that there are few membranes and studies currently
Keywords: available. The preparation of a novel positively charged nanofiltration membrane is also described. This
Nanofiltration membrane was fabricated by surface modification of a prepared base membrane using polyethyleneimine
Positive charge followed by cross linking with butanedioldiglycidylether. The fabrication process uses standard organic
Surface modification
solvents and avoids the need for hazardous materials, such as concentrated sulphuric acid, which significantly
Atomic force microscopy
benefits the scale up potential of any future commercial manufacturing process.
The new membrane was characterised using a number of state-of-the-art techniques, including a novel use of
atomic force microscopy to determine pore size. Streaming potential measurements confirmed that this new
membrane is indeed positively charged in the pH range below pH 9, which covers the majority of normal
operating conditions. The performance characteristics for the new membrane were very favourable, with a
pure water flux determined to be 20 LMH bar−1 and a rejection of MgCl of 96%. Thus, this new membrane both
adds to and complements the existing short supply of positively charged NF membranes and is suitable for
applications such as the recovery of valuable cationic macromolecules in the bioprocess and pharmaceutical
industries or removal of multi-valent cations such as dyes and heavy metals in the paper and pulp, textiles,
nuclear, and automotive industries.
© 2011 Elsevier B.V. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2. Experimental . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.1. Substrate membrane formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2. Surface modification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3. Membrane characterisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.1. Zeta potential measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.2. Frontal filtration experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3.3. Atomic force microscopy — membrane structural characterisation and topography . . . . . . . . . . . . . . . . . . . . . . 15
2.3.4. Scanning electron microscopy — membrane cross section analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3. Results and discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.1. Zeta potential measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2. Water permeability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.3. Frontal filtration experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.4. Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
4. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

⁎ Corresponding author. Tel.: +44 1792 295 200.

E-mail address: C.Wright@swansea.ac.uk (C.J. Wright).

0001-8686/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.cis.2010.12.010
 S. Cheng et al. / Advances in Colloid and Interface Science 164 (2011) 12–20 13

1. Introduction hydroxypropyltrimethyl ammonium chloride. This membrane exhib-

ited a reduced molecular weight cut off of 560 Da, a rejection of MgCl2
Nanofiltration (NF) is a pressure driven membrane separation approaching 97%, and an improved membrane flux of 14–18 LMH at
process with characteristics between reverse osmosis and ultrafiltra- 12 bar. Tongwen and Weihua [9] prepared a new membrane using
tion. The nominal molecular weight cut off of a NF membrane is in the 2,6-dimethyl-1,4-phenylene oxide dissolved in chlorobenzene and
range 100–1000 Da, indicating that NF membranes have an approx- then brominated, the resulting solution was precipitated by addition
imate pore size of 1 nm. Separation of solutes in the NF range is of methanol and dried to form the primary membrane polymer. This
dependent upon the micro-hydrodynamics and interfacial events brominated primary polymer was dissolved in chloroform and cast
occurring at the membrane surface and inside the membrane pore. onto a substrate membrane (prepared from similar materials) and
Rejection may be attributed to a combination of both steric and charge then cross linked in situ using a mixture of trimethylamine and
effects. ethylenediamine. The resulting membrane best case performance
Nanofiltration was introduced in the early 1980s and has since achieved a very favourable flux (63 LMH at 2.5 bar) but exhibited
gained in popularity due to improved selectivity for mono and multi a low rejection value of 73% for MgCl2. Each of the membranes
valent ions, low operating pressures and relatively low capital and described above was prepared by interfacial polymerisation upon a
operating costs. To date, as a direct result of the fabrication materials substrate micro or ultra filter to produce a composite NF membrane.
traditionally employed, most NF membranes are either neutral or The production of composite membranes is generally considered to
negatively charged in an aqueous processing environment. These be complex. Each of the fabrication processes described above uses
polymeric membranes are often amphoteric in nature and can exhibit hazardous or carcinogenic materials, involves several preparation
a net positive charge at low pH, typically less than pH 4 [1,2]. This has stages, is time consuming, and offers little potential to be expanded
led to some confusion over the actual charge nature of some to an economical full scale production process. Integrally skinned
negatively charged membranes and is the result of much confusion asymmetric membranes offer a direct alternative and are fabricated
between membrane manufacturers, end users and the literature [3]. using the more simplistic phase inversion technique [10].
Positively charged NF membranes can have excellent hydrophilicity Tang et al. [11–13] produced a membrane using the same
and exhibit high rejections for multi valent cations. These properties materials as Tongwen and Weihua [9]. In this case, the membrane
are particularly important for recovery of valuable cationic macro- was formed by phase inversion using an ethanol bath rather than cast
molecules in the bioprocess and pharmaceutical industries or removal onto a substrate membrane. The authors studied the influence of
of multi valent cations such as dyes and heavy metals from effluents in several fabrication parameters on the membrane performance and
the paper and pulp, textiles, nuclear, and automotive industries. The morphology. However, characterisation of the resulting membrane
distinct lack of positively charged NF membranes is true of both the demonstrated a best pore size distribution in the range 7 to 10 nm
commercial and laboratory environment alike and very limited which is in the ultrafiltration range. Su et al. [14] produced an
studies are currently available that utilize positively charged NF asymmetric base membrane using the phase inversion technique
membranes. from chloromethylated poly(phthalazinone ether sulfone ketone)
Du and Zhao [4,5] were the first to produce a positively charged NF dissolved in N-methyl-2-pyrrolidine and cast onto glass sheets at
membrane using poly(N,N-dimethylaminoethyl methacrylate). The 60 °C prior to immersion in water. The base membrane was then
poly(N,N-dimethylaminoethyl methacrylate) was first purified by immersed into a solution containing trimethylamine to induct
ultrafiltration and then cast as an aqueous solution onto a polysulfone quaternary nitrogen groups into the membrane. Thus, applying a
microfilter that acted as a mechanical support plate. Following slight surface modification to the original asymmetric membrane and
drying, the new membrane was cross linked by immersion in p- introducing a positively charged quaternized surface layer. The effects
xylylene dichloride/heptane solution. The resulting NF membrane of evaporation time, coagulation temperature and additives in the
exhibited a best rejection of 90% for MgSO4, with a membrane flux of fabrication bath were studied and membrane performance was
10–20 LMH at 8 bar. Wang and co-workers [6] investigated the effects optimised. The resulting membranes exhibited excellent pure water
of pH and NaHCO3 concentration in the casting solution of a similar flux performance, around 154 LMH at 2 bar, whilst maintaining
membrane. They determined that a uniform top layer is obtained almost 100% rejection of the dyes Clayton Yellow (MW = 696 Da) and
through stabilising the pH of the casting solution during the pre- Methyl Orange (MW = 327 Da). The rejection of MgCl2 for this same
drying phase of the fabrication process. Stabilisation at different pH membrane was low at a value of 20%. Interestingly, when a series of
values had a profound impact on the membrane separation properties rejection experiments using different dyes were conducted for this
with a maximum NaCl rejection (93%) and flux (24 LMH at 8 bar) at membrane, the rejections of dyes X-6G, X-8B and Methyl Orange (all
pH 7. The relationship between rejection and flux was counter negatively charged) were all higher than the rejections of dyes
intuitive, i.e. one would normally expect lower rejection at high flux, Methylene Blue, Centian Violet and Malachite Green (all positively
and the workers acknowledged this fact but had no explanation. charged). The authors explain this phenomenon by assuming that the
A similar effect was noted when varying the salt concentration increased rejection of the negative dyes are the result of steric
added to the casting solution, producing a maximum rejection for partitioning caused by increased aggregation of these dye species and
MgSO4 of 91% (no flux data was provided). This value is similar to not the result of Donnan effects. If this were to be the case, then the
the performance of the membrane produced by Du and Zhao [4,5] fact that the new membrane is positively charged has no bearing on
described previously. Huang et al. [7] produced a positive NF mem- this specific separation. In addition, the possibility arises that this
brane using a copolymer graft of trimethylallyl ammonium chloride membrane could in fact exhibit a net negative charge as there is no
onto chitosan as the active top layer. Quaternation was applied to direct evidence presented confirming that this membrane is indeed
modify the chitosan in order to improve hydrophilicity and introduce positively charged (i.e. via streaming potential measurement or
the positively charged groups. This solution was cast onto a polyac- equivalent) and the positive nature of this membrane is only assumed
rylonitrile ultrafiltration membrane as the support and then cross as a result of the quaternization procedure. Yan and co-workers
linked with toluene diisocyante. The molecular weight cut off of [15,16] produced an NF membrane using similar methods and
the resulting membrane was determined to be 930 Da and the materials as Su et al. [14]. Poly(phthalazinone ether sulfone ketone)
membrane performance was reasonable, with a rejection of MgCl2 was dissolved in concentrated sulphuric acid and reacted with
greater than 90% in some cases and a modest membrane flux of 8.5 chloromethyl octyl ether to form the base membrane material
LMH at 12 bar. In an analogous study, the same workers [8] produced which was prepared by phase inversion. The base membrane was
a membrane by substituting trimethylallyl ammonium chloride for 2- then soaked in trimethylamine solution to allow chemical adsorption
14 S. Cheng et al. / Advances in Colloid and Interface Science 164 (2011) 12–20

CH2CH 2NH
and self assembly formation of the positively charged membrane. The
resultant membrane exhibited similar performance as those previous H 2N (CH 2CH2N ) (CH 2CH2NH) CH2CH 2NH CH2CHCH 2(CH 2)4OR
[4–8], with a MgCl2 rejection of 84% and a membrane flux of 49 LMH at
4 bar. The observation was also made that the membrane flux could
be significantly improved by increasing the processing temperature
(from 30 LMH at 15 °C to 120 LMH at 85 °C) with little detriment to O
O
the rejection performance (82% to 75% respectively). As for Su et al.
CH 2 CHCH 2O(CH2)4OCH 2CH CH 2
[14], Yan and co-workers [15,16] failed to directly measure the
membrane charge properties. However, the rejection characteristics
for a series of salts were provided, where R(MgCl2) N R(MgSO4) N R
(NaCl) N R(Na2SO4), inferring that the membrane formed was indeed
positively charged. CH2O(CH2)2OR
In the following sections of this paper, the fabrication of a novel
positively charged NF membrane using the phase inversion technique CHOH
followed by chemical adsorption is demonstrated, although avoiding
the need to use hazardous materials such as concentrated sulphuric CH2 CH2CH 2NH OH
acid in the production process. Furthermore, the influence of
N (CH 2CH2N) (CH 2CH2N) CH2CH 2NH CH2CHCH 2(CH 2)4OR
fabrication conditions in order to produce a membrane of superior
processing performance is investigated. The modified membrane is CH2 CH2
also characterised using an array of state-of-the-art techniques,
including a novel use of atomic force microscopy (AFM), in order to CHOH
CHOH
clearly demonstrate both the physical properties and separation
CH2O(CH2)4OR
performance. CH2O(CH2)4OR

R= CH2CHOHCH 2OH, or CH2CHOHCH 2

2. Experimental
poly ethylenimine
2.1. Substrate membrane formation
Fig. 1. Cross-link reaction for polyethylenimine with BUDGE.
The membranes fabricated and characterised in this study were
prepared in the Swansea laboratory. A flat sheet polyetherimide/
sulfonated polyether ether ketone blend membrane (A6DT) was 2.3. Membrane characterisation
made using the traditional phase inversion method from a N-
methylpyrrolidone (NMP)/tetrahydrofuran (THF)/dioxane solution 2.3.1. Zeta potential measurement
of the polymers using a doctor blade (CAMAG), described previously The zeta potential (ζ-potential) was determined with an electro-
[17]. kinetic analyser (EKA, Anton Paar KG, Graz, Austria) based on the
streaming potential method [2,3]. Flat sheets of the membranes were
used as samples to study the influence of pH of the solution on
2.2. Surface modification the membrane ζ-potential. The streaming potential was measured
by forcing the electrolyte solutions through a thin slit channel
The negatively charged A6DT blend membranes were immersed in (90 × 10 × 0.27 mm) formed by a PVTF spacer between two mem-
an aqueous solution of 5 g l−1 polyethyleneimine (pH 10.5, which is branes facing each other, for continuously increasing values of
the original pH value of the polyethyleneimine solution) for 8 h. pressure (up to 200–240 mbar). The ζ-potential was calculated
Polyethyleneimine was deposited on the substrate membrane via using the Smoluchowski equation. The pH and conductivity of the
electrostatic interactions, to fabricate a new positively charged solutions were also measured. The ζ-potential was measured in
membrane PA6DT. The modified membranes were washed with 1.0 mol m−3 KCl over a pH range of 3.5 to 11.0, altered by addition of
deionised (DI) water three times and then stored in DI water. 0.1 M HCl or 0.1 M NaOH, at 25 ± 0.5 °C.
Polyethylenimine (50% in water) and butanedioldiglycidylether
(BUDGE) were purchased from Sigma-Aldrich, UK. Polyethyleneimine 2.3.2. Frontal filtration experiments
is a water-soluble polymer. In order to improve the stability of the Frontal filtration was used to measure the water permeability and
PA6DT modified membranes, BUDGE was used as a cross-linking rejection of the modified membranes using constant feed conditions
agent on the membrane surface [18]. The modified membranes were over a range of pressures, as described in previous studies [17]. All
immersed in 50 ml of a 5% (V/V) BUDGE in methanol solution and experiments were conducted at room temperature (20 ± 2 °C) and a
allowed to react overnight to obtain a cross-linked positive mem-
brane. The reaction of cross-linking is shown in Fig. 1. The membranes
were then washed with methanol and DI water three times Table 1
respectively. The cross-linked modified membrane was denoted as Modification conditions for membrane A6DT.
PA6DT-C and was stored in DI water until characterisation.
Membrane pH NaCl (mol m−3)
Membrane modification at lower pH values was also completed
to obtain other types of modified membrane, using 0.1 M HCl to PA6DT-C (6) 6 –
PA6DT-C (8) 8 –
change the pH value of polyethyleneimine solution from pH 10.5 to
PA6DT-C 10.5 –
pH 8 and pH 6. The substrate membrane was placed in a mixed PA6DT-S-C 10.5 100
solution of 5 g l−1 polyethyleneimine and 100.0 mol m−3 NaCl, which
P - the polyethyleneimine modified membrane,
was then used to study the effect of salt in the casting solution on C - membrane cross-linking, and
the characteristics of the resulting membrane. The modification con- S - the addition of salt in the modification solution.
ditions are provided in Table 1. The numbers represent pH values of the modification solution.
 S. Cheng et al. / Advances in Colloid and Interface Science 164 (2011) 12–20 15

pH of 5.8–6.0 (which is the pH of the pure water used in the study), represents the phase angle between the applied and the actual
except when indicated otherwise. All filtration experiments were vibration of the tapping tip. Thus, the phase signal is a function of the
carried out in an Amicon 8050 cell with 50 ml of fresh feed solu- mechanical properties of the material under the tip, the interaction
tion and an effective membrane area of 13.4 × 10−4 m2. The water forces, and the area of contact between tip and surface (i.e. any change
permeability of each membrane was measured in a pressure range in topography).
of 100 and 500 kPa. Solute rejection was measured using 0.1 g l−1
Vitamin B12, 0.1 g l−1 PEG 1500, and 1.0 mol m−3 salts (NaCl, Na2SO4, 2.3.4. Scanning electron microscopy — membrane cross section analysis
MgCl2 and MgSO4) over the same pressure range. Vitamin B12 Scanning electron microscopy (SEM) was used to image the cross-
(MW = 1355 Da) and PEG1500 were supplied by Sigma-Aldrich, UK. section of the asymmetric membranes formed. The membranes were
All the salts were supplied by Fisher-Scientific UK. Rejection was immersed in ethanol followed by cyclohexane for 2 h respectively
based on measurement of 15 ml of permeate after the first 4 ml and then dried in air. The dried membranes were then cryogenically
permeate was discarded. The membrane was washed with DI water fractured in liquid nitrogen and coated with Pt/Pd. During the prepa-
after each experimental run to thoroughly remove the remains of ration of the membrane sample, care was taken to avoid damage or
the previous solute. Data from filtration experiments was used to the creation of artefacts in the membrane structure. The instrument
calculate the water permeability and effective pore radius; this used was a Philips XL30 CP SEM (FEI, The Netherlands).
method has been extensively discussed in previous work [17,19,20].
These parameters were used to compare the process performance of
3. Results and discussion
the unmodified and modified membranes. Rejection of solute was
calculated using the following equation,
3.1. Zeta potential measurements
Cp
Robs = 1− ð1Þ Fig. 2 presents the pH dependence of ζ-potential of membrane
Cb
A6DT and modified membranes PA6DT and PA6DT-C in 1.0 mol m−3
where Cp and Cb are the concentration of permeate solution and feed KCl. The charge reversal of the new membranes and significant shift
solution respectively. As a result of concentration polarization, the in the isoelectric point (IEP) from that of the substrate membrane
concentration of a solute at the membrane wall is higher than that of (A6DT) due to the adsorption of polyethyleneimine followed by
the bulk. Therefore, real rejection is defined as, BUDGE is clearly observed. The ζ-potential of the membrane A6DT
had a near zero value at pH 3.5, which is the IEP of SPEEK. The IEP
Cp determined for the new membranes PA6DT and PA6DT-C was at pH
Rreal = 1− ð2Þ values 10.2 and 9.0 respectively. Therefore, at low pH values (below
Cw
IEP), the new membranes are positively charged and will become
where Cw is the solute concentration at the membrane surface. The negatively charged above the IEP. The IEP of the cross linked mem-
value of Cw may be calculated indirectly from the bulk feed solute brane PA6DT-C is lower than that of membrane PA6DT as some of the
concentration (Cb) using a mass transfer correlation [21]. As the exposed amino groups at the membrane surface have formed ether
sample size is significant in comparison to the total volume of the bonds in the reaction with BUDGE. The shift in the IEP value for the
Amicon 8050 cell, the feed concentration was taken as the average new membranes represents approximately 7 pH units and indicates
from the start and end of the experiment, i.e. that both new membranes will be positively charged under the
majority of practical processing conditions. Interestingly, the magni-
1  tude of the ζ-potential for membrane PA6DT is greater than that of
Cb = C + Cb;end : ð3Þ
2 b;start PA6DT-C over the entire range studied, thus indicating that there is a
net trade in the magnitude of membrane fixed charge for an improved
Membrane flux (F) was calculated as membrane stability.

W
F= ð4Þ 3.2. Water permeability
At

where W is the mass of pure water or solution permeated during the The pure DI water flux was measured for each membrane using an
experiment, A is the membrane area and t is the experimental time. applied pressure range of 100–500 kPa. Water permeability can be
determined from the slope of the straight line obtained when plotting
2.3.3. Atomic force microscopy — membrane structural characterisation the variation of the water flux versus the applied pressure. Table 2
and topography
AFM of the membrane surface was conducted using a Dimension
3100 AFM (Veeco Instruments, Santa Barbara, CA) in tapping mode to 20 A6DT
image (in air) the surface topography by probing the surface with an PA6DT
PA6DT-C
oscillating tip. Standard tapping mode cantilevers (Olympus, Japan)
with a spring constant of 42 N/m and a typical tip apex of 5 nm were 10
zeta potential / mV

used. Phase images captured simultaneously with topographic images

were used to study the properties of the membrane. Phase imaging is
the mapping of the phase lag between the periodic signal that drives 0
the cantilever and the oscillations of the cantilever [22]. Using tapping 0 2 4 6 8 10 12
pH
mode AFM, the phase shift is derived from the difference in phase
angle between the freely oscillating cantilever in air and the cantilever -10
oscillation during scanning. Changes in the phase lag during tapping
mode operation often indicate changes in the properties of the sample
surface [23,24]. The phase shift is zero when there is no interaction -20
between the tip or the cantilever and the sample surface. The phase
signal is collected alongside the tapping mode topography data and Fig. 2. Zeta potential as a function of pH (background electrolyte 1.0 mol m−3 KCl).
16 S. Cheng et al. / Advances in Colloid and Interface Science 164 (2011) 12–20

Table 2
Characteristics of modified and basic membranes.

Membranes Pm m3s−1N−1 × 1011 Rreal a

PEG1500 rp a
nm Rreal b
B12 rp b
nm Rreal c
NaCl Xd mol m−3

A6DT 27.5 ± 1.9 0.692 2.07 – – 0.5790 −45.8

PA6DT-C (6) 16.2 ± 0.02 0.8243 1.74 0.5424 1.65 0.4553 3.8
PA6DT-C (8) 13.6 ± 0.23 0.9151 1.51 0.5790 1.49 0.5937 8.3
PA6DT-C 5.34 ± 0.48 0.9274 1.47 0.6408 1.46 0.8249 12.5
PA6DT-S-C 3.70 ± 1.1 0.9316 1.46 0.6690 1.42 0.8524 13.8

For rejection data. 100 kNm− 2 for A6DT, and 400 kNm− 2 for all other membranes.
a
From rejection of the 0.1 g/l PEG1500 solution.
b
From rejection of the 0.1 g/l Vitamin B12 solution.
c
From rejection of the 0.001 M NaCl solution.

summarises the permeability of the original and modified mem- for membrane PA6DT-C(6) at ca. 59 LMH bar−1 and the flux order was
branes. Note that from this point forward only the cross linked PA6DT-C(6) N PA6DT-C(8) N PA6DT-C N PA6DT-S-C. The flux value
membrane is considered due to the poor stability of polyethylenei- obtained for the membrane PA6DT, prepared under standard
mine in water for the PA6DT membrane. As expected, the highest pure conditions, was ca. 19.4 LMH bar−1. This value is almost twice that
water flux was achieved by the original A6DT membrane with a value achieved by Yan and co-workers [15,16] for their asymmetric
of ca. 100 LMH bar−1. The deposition of polyethyleneimine on the membrane and is an order of magnitude greater than the flux values
substrate membrane surface resulted in a remarkable decrease in the obtained using the composite membranes [4–9].
water flux due to the formation of a more condensed surface layer.
The maximum pure water flux achieved for the other membranes was 3.3. Frontal filtration experiments

The neutral solute rejection data is presented in Table 2. Each of

the new membranes studied exhibited high rejection of the PEG1500,
a with rejection greater than 90% in all cases except for membrane
6
PA6DT-C(6), which had a lower rejection of 82%. This indicates that
PA6DT-C(6)
PA6DT-C(8)
membrane PA6DT-C(6) has a more open pore structure than the other
5 PA6DT-C membranes and is confirmed by the increased water permeability
PA6DT-S-C
best fit
observed. The rejection of vitamin B12 follows a similar pattern as
Jv / 10-5 m3 m-2 s-1

4 that for PEG1500 but with lower rejection values. The neutral solute
rejections were then used to calculate the effective pore size for the
membranes. Membranes PA6DT-C, PA6DT-C(8), and PA6DT-S-C all
3 had a calculated pore size of ca. 1.5 nm, with the PA6DT-C(6)
membrane having a slightly larger pore size of around 1.7 nm. Again,
2 this is consistent with the membrane flux and rejection trends
observed for these membranes.
The membrane rejection and flux determined over a range of
1
effective pressures for a solution of NaCl using the new membranes
are presented in Fig. 3. The highest membrane flux was obtained for
0 membrane PA6DT-S-C; however, this membrane also exhibited the
0 100 200 300 400 500 600
lowest rejection values. The flux of membrane PA6DT-C(8) was only
ΔPe / kN m-2 slightly lower than membrane PA6DT-S-C but the rejection perfor-
mance was improved, by approximately 20%. Membranes PA6DT-C
b
1.0
1.0

0.8
0.8

0.6
Rejection

0.6
Rejection

0.4
0.4
PA6DT-C(6)
PA6DT-C(8)
0.2 PA6DT-C
PA6DT-S-C
0.2

0.0
0 100 200 300 400 500 600
0.0
ΔPe / kN m-2 0 2 4 6 8 10 12
pH
Fig. 3. The effect of effective pressure on the rejection of NaCl solution with concen-
tration of 1 mol m−3 at membranes PA6DT-C(6), PA6DT-C(8), PA6DT-C, PA6DT-S-C, Fig. 4. Determination of the isoelectric point of the modified membrane (PA6DT-C) by
and PAA- PA6DT-C. (a) Flux. (b) Rejection. rejection of 2.9 mol m−3 NaCl.
 S. Cheng et al. / Advances in Colloid and Interface Science 164 (2011) 12–20 17

1.0 Table 3
NaCl
AFM characteristics of the membranes.
MgCl2
0.8 AFM characterisation A6DT PA6DT-C
Na2SO4
MgSO4 rAFM (nm) 4.08 ± 2.64 3.52 ± 1.90
Roughness (nm) 1.39 ± 0.12 1.56 ± 0.10
Rejection

0.6
ε (%) 1.42 0.84

0.4
pH during the modification process has the effect of increasing the
pore size and reducing the magnitude of charge for the membrane
0.2
produced. The addition of salt during the modification process has
little effect on the properties of the membrane produced as ionic
0.0 strength does not significantly affect the adsorption of weak poly-
0 100 200 300 400 500 600 electrolytes on a highly charged surface. The slight decrease in
ΔPe / kN m-2 membrane flux observed for PA6DT-S-C in comparison to PA6DT-C
may be the result of the NaCl screening the electrostatic charges of
Fig. 5. Salt rejection as a function of effective pressures for membrane PA6DT-C. both polyethyleneimine and the A6DT substrate membrane allowing
the formation of a slightly more dense modified surface layer.
The rejection of NaCl from membrane PA6DT-C over a range of pH
and PA6DT-S-C exhibited very similar performance. The flux values is presented in Fig. 4. The data clearly indicates a minimum rejection
obtained for these membranes were significantly lower than those for at pH 9 and confirms the isoelectric point previously evaluated from
membranes PA6DT-C(6) and PA6DT-C(8), with the rejection values streaming potential measurements. Rejection experiments for mem-
significantly higher. The NaCl rejection data was also used to calculate brane PA6DT-C using various single salt solutions were conducted
the volumetric membrane charge density (Xd), this information is over a range of pressures from 200 kPa to 500 kPa and the results are
shown in Table 2. The calculated Xd value for membrane PA6DT-C(6) detailed in Fig. 5. The salt rejection order from high to low is
was modest at 3.8 mol m−3, the value for membrane PA6DT-C(8) was MgCl2 N NaCl N MgSO4 N Na2SO4 and is as expected for a positively
double at 8.3 mol m−3, and the values for both membranes PA6DT-C
and PA6DT-S-C were triple at 12.5 and 13.8 mol m−3 respectively.
This indicates that the low rejection and high flux obtained for
membrane PA6DT-C(6) are directly attributable to the more open
a 60
pore size and the lower membrane charge. In comparison, mem-
branes PA6DT-C and PA6DT-S-C both have a higher membrane charge
50
and reduced pore size, which provides explanation of the significantly
improved rejection with reduced membrane flux. Membrane PA6DT-
C(8) is essentially a compromise between these two extremes, with 40
% of pores

both a pore size and membrane charge between the values obtained
for the other membranes, thus providing explanation for the rejection 30
and flux performance observed. Therefore, a reduction in the solution
20

10

0
0 2 4 6
Pore radius / nm

b 60

50

40
% of pores

30

20

10

0
0 2 4 6
pore radius / nm

Fig. 6. The three dimensional images of membranes by AFM. (a) A6DT and (b) PA6DT-C. Fig. 7. Pore size distribution for (a) A6DT and (b) PA6DT-C by AFM.
18 S. Cheng et al. / Advances in Colloid and Interface Science 164 (2011) 12–20

charged membrane. The limiting rejection of MgCl2 was R = 96%. The and the hypothesis is made that the reduction in pore size and
salt concentration used in this study was less than that used in other porosity of the PA6DT-C membrane was a direct result of the depo-
published works [15,16], thus a meaningful comparison is not sition of polyethyleneimine layers on the membrane surface and
available (this will be addressed in a future study). Nonetheless, a potentially within the pore walls.
rejection of 96% is considerable. A pore size of 1.47 nm for this Pore size and distribution were quantitatively determined by
membrane indicates that the steric contribution to overall rejection analysis of the images in conjunction with digitally stored line pro-
will be very low and the single salt rejections will be dominated by files and are reported in Fig. 7. The modified membrane exhibited
Donnan exclusion [25]. The Donnan exclusion mechanism dictates relatively narrow pore size distributions. The mean pore size values
that a high valence co-ion would be more highly rejected than mono obtained by the AFM technique were found to be 2.0–2.5 times
valent co-ion, thus explaining why MgCl2 rejection is higher than larger than those calculated from the solute transport data (Table 2).
NaCl. Also, a high valence counter-ion will actively permeate due to AFM has been employed to measure the pore size of NF filtration
electrostatic attraction, thus explaining why the sulphate salts have membranes previously [26–28]; however, the technique tends to
lower rejection than the chloride salts, with the magnesium salt being provide a larger pore size than that determined from experimental
more highly rejected for the previously explained reason. data due to the limit of AFM resolution. Singh et al. [28] and Bessières
et al. [29] also observed that AFM gave 2–4 times larger pore
3.4. Microscopy diameters than those obtained from solute transport. Bessières et al.
[29] postulated that the pore sizes obtained from solute rejection
AFM topographic images of the substrate membrane (A6DT) and correspond to a minimal size of the pore constriction experienced by
new membrane (PA6DT-C) are presented in Fig. 6. The AFM image the solute while passing through the pore, whereas pore sizes
analysis program was used to measure surface pore size. Pore sizes measured by AFM correspond to the pore entrances on the membrane
were measured by visually inspecting line profiles of different pores. surface. In the present study pore size was measured at the half way
The pore size, roughness and porosity of the substrate membrane and point, as the AFM line trace descended into the pore. A great deal of
the modified membrane, as measured using AFM, are summarised in care must be taken when interpreting images of NF membranes, as the
Table 3. The average pore size of the A6DT membrane is 4.08 nm. size of the pores are at the limits of the microscopic technique. In
After the deposition of polyethyleneimine and cross linking, the Table 3, the pore radius of A6DT was 4.08 nm, but was previously
average pore size of PA6DT-C decreased to 3.52 nm. The porosity also determined to be 1.19 nm [17]. The previous AFM work used an older
decreased with deposition of polyelelctrolyte layers, from 1.42% AFM (Auto probe) in contact mode in contrast to the work presented
for A6DT to 0.84% for PA6DT-C. Note that PA6DT-C was made by in this paper which used a state of the art instrument (Dimension
depositing a thin layer of polyethyleneimine on the A6DT membrane 3100) in tapping mode. Comparison between pore size measurements

Fig. 8. Determination of the pore size (diameter) of the membrane PA6DT-C from topography and phase images using line profiles. (a) topography, (b) phase image,
(c) determination of pore size from 1 μm topography, and (d) determination of pore size from 1 μm phase image.
 S. Cheng et al. / Advances in Colloid and Interface Science 164 (2011) 12–20 19

from different studies is difficult, and strengthens the argument that modified membranes, varying from 1.39 nm to 2.54 nm for a scanned
supporting techniques such as pore size calculation from rejection area of 1 × 1 μm2.
data should be used in combination with AFM techniques. The membrane cross-sectional morphology and internal structure
Fig. 8 presents the topography and phase image for the PA6DT-C were imaged using SEM and are presented as Fig. 9. Each of the
membrane. In membrane studies phase imaging can be used to membranes shown has a similar asymmetric structure with a sponge-
detect heterogeneous mixing of polymers within a blend membrane; like top layer and finger-like support structure. The cross-section was
however, there is also the potential to aid in the study of membrane imaged at 1000× magnification and the top layer was imaged at
pore structure. The morphology of the polymer membrane in the 10,000× magnification. The high magnification image of the top layer
phase image is much sharper than in the topographical image. indicates very little difference, if any, between the two membranes.
Unfortunately, there is no height data available in the phase image as Therefore, the polyethyleneimine modification layer is thin and
the data is measured as a phase angle. However, X–Y distances can be certainly less than 15–20 μm which is typical of that for composite
measured and compared to the same profile from the topographical membranes. This thin top layer offers little extra hydraulic resistance
image. From the topography line profile illustrated in Fig. 8c, the pore to flow and explains the significant increase in flux for these mem-
radius was determined to be 5.61 nm. Evaluation of the same pore branes when compared to the composite membranes. The resolution
using the phase image (Fig. 8d) gives a pore radius of 4.68 nm. This is of this technique is not sufficient to determine if the modified layer is
arguably due to the fact that as the tapping tip passes over the void of evenly distributed across the surface or penetrates the pore structure.
the pore, the interactions between the walls of the pore and the sharp
tip are minimum. Thus, AFM phase imaging could provide a more 4. Conclusions
accurate definition and determination of pore size.
Surface roughness was also measured using AFM. At least three A review of the fabrication of positively charged NF membranes
AFM images of different parts of the same membrane were analysed has been completed. Currently there are a limited number of studies
and mean values of roughness are reported in Table 3. The general available, thus further research in this area is justified. The
assumption is made that the surfaces of membranes prepared from preparation of a novel positively charged NF membrane has been
polymeric materials and used for various membrane separation pro- demonstrated and will add to the limited number of membranes
cesses have to be smooth. However, in the process of the preparation currently available. This new membrane designated PA6DT-C was
of polymeric membranes, the membrane surfaces increase in rough- fabricated using the phase inversion technique followed by surface
ness. The roughness values are different for the basic membrane and modification. The base membrane materials of construction were a

Fig. 9. SEM images of membrane cross-section [left] and top layer [right]. (a) A6DT, and (b) PA6DT-C.
20 S. Cheng et al. / Advances in Colloid and Interface Science 164 (2011) 12–20

blend of polyetherimide and sulfonated poly(ether ether ketone) [4] Du R, Zhao J. J Appl Polym Sci 2004;91:2721.
[5] Du R, Zhao J. J Membr Sci 2004;239:183.
modified with polyethyleneimine to achieve the positive charge and [6] Wang W, Du Q, Li G. Nanoscience 2006;11:38.
cross linked with BUDGE to improve stability. The fabrication process [7] Huang R, Chen G, Sun M, Gao C. Desalination 2009;239:38.
used standard industrial solvents and avoided hazardous materials [8] Huang R, Chen G, Sun M, Hu Y, Gao C. J Membr Sci 2006;286:237.
[9] Tongwen X, Weihua Y. J Membr Sci 2003;215:25.
such as concentrated sulphuric acid, which significantly benefits the [10] Mulder M. Basic principles of membrane technology. London: Kluwer Academic
scale up potential for a commercial manufacturing process. Publishers; 1996.
The new membrane was characterised using an array of state-of- [11] Tang B, Xu T, Gong M, Yang W. J Membr Sci 2005;248:119.
[12] Tang B, Xu T, Yang W. J Membr Sci 2006;268:123.
the-art techniques, including a novel use of AFM. ζ-potential [13] Tang B, Xu T, Sun P, Yang W. J Membr Sci 2006;279:192.
measurements and frontal filtration experimentation demonstrated [14] Su Y, Jian X, Zhang S, Wang G. J Membr Sci 2004;241:225.
that the new membrane was indeed positively charged. The effect of [15] Yan C, Zhang S, Yang D, Jian X. J Appl Polym Sci 2008;107:1809.
[16] Yan C, Zhang S, Liu C, Yang D, Yang F, Jian X. J Appl Polym Sci 2009;113:1389.
pH and salt in the modification solution of polyethyleneimine was
[17] Bowen WR, Cheng SY, Doneva TA, Oatley DL. J Membr Sci 2005;250:1.
also studied. Lowering the pH was found to produce a membrane of [18] Kennedy LA, Kopaciewicz W, Regnier FE. J Chrom 1986;359:73.
larger pore size and lower surface charge, thus lowering rejection and [19] Oatley DL, Cassey B, Jones P, Bowen WR. Chem Eng Sci 2005;60:1953.
increasing membrane flux. Addition of salt had little effect on the [20] Bowen WR, Welfoot JS. Chem Eng Sci 2002;57:1121.
[21] Bowen WR, Mohammad AW. Desalination 1998;117:257.
membrane performance characteristics as ionic strength does not [22] Boussu K, Van der Bruggen B, Volodin A, Snauwaert J, Van Haesendonck C,
significantly affect the adsorption of weak poly-electrolytes on a Vandecasteele C. J Colloid Interface Sci 2005;286:632.
highly charged surface. The performance characteristics evaluated for [23] Huang HS, Chen CY, Lo SC, Lin CJ, Chen SJ, Lin SJ. Appl Surf Sci 2006;253:2685.
[24] Ghassemi H, McGrath JE, Zawodzinski Jr TA. Polymer 2006;47:4132.
this new membrane were very favourable, with a membrane flux of [25] Donnan GF. J Membr Sci 1995;100:45.
approximately 20 LMH bar−1 and a MgCl2 (divalent cation) rejection [26] Bowen WR, Welfoot JS. Chem Eng Sci 2002;57:1393.
of 96%. Thus, this new membrane is suitable for applications such as [27] Hilal N, Al-Zoubi H, Darwish NA, Mohammad AW. Desalination 2005;177:187.
[28] Singh S, Khulbe KC, Matsuura T, Ramamurthy P. J Membr Sci 1998;142:111.
the recovery of valuable cationic macromolecules in the bioprocess [29] Bessieres A, Meireles M, Coratger R, Beauvillain J, Sanchez V. J Membr Sci
and pharmaceutical industries or removal of multi-valent cations such 1996;109:271.
as dyes and heavy metals from effluents in the paper and pulp,
textiles, nuclear, and automotive industries.

References
[1] Tsuru T, Takezoe H, Asaeda M. AIChE 1998;44:765.
[2] Childress AE, Elimelech M. J Membr Sci 1996;119:253.
[3] Schaep J, Vandecasteele C. J Membr Sci 2001;118:129.