Bioresource Technology 101 (2010) 1459–1465

Contents lists available at ScienceDirect

Bioresource Technology
journal homepage: www.elsevier.com/locate/biortech

The formation and characterisation of an asymmetric nanofiltration membrane

for ammonia–nitrogen removal: Effect of shear rate
Nora’aini Ali a,*, N. Syazana A. Halim a, A. Jusoh a, Azizah Endut b
a
Faculty of Science and Technology, University of Malaysia Terengganu (UMT), 21030 Kuala Terengganu, Terengganu, Malaysia
b
Faculty of Innovative Design and Technology, Universiti Darul Iman Malaysia, 21030 Kuala Terengganu, Terengganu, Malaysia

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

Article history: The focus of this research is to study the potential of nanofiltration membrane technology in removing
Received 28 April 2009 ammonia–nitrogen from the aquaculture system. One of the major fabrication parameters that directly
Received in revised form 15 August 2009 affect the separation performance is shear rate or casting rate during membrane fabrication. In this study,
Accepted 18 August 2009
asymmetric polyethersulfone (PES) nanofiltration membranes were prepared at five different shear rates
Available online 27 September 2009
within the range of 67–400 s1. Membrane productivity and separation performance were assessed via
pure water, salt and ammonia–nitrogen permeation experiments, and their structural properties were
Keywords:
determined by employing the combination of the irreversible thermodynamic (IT) model, solution diffu-
Theoretical models
Shear rate
sion model, steric hindrance pore (SHP) model and Teorell–Meyers (TMS) model. The study reveals that
Ammonia–nitrogen removal the alteration of shear rate enormously affects the membrane morphology and structural parameters,
Nanofiltration hence subsequently significantly influencing the membrane performance. It was found that, membrane
Polyethersulfone produced at the shear rate 200 s1 or equivalent to 10 s of casting speed during membrane fabrications
managed to remove about 68% of ammonia–nitrogen, in which its separation performance is the most
favourable by means of highest flux and rejection ability towards unwanted solutes. Besides, from the
research findings, nano-membrane technology is a potential candidate for the treatment of aquaculture
wastewater.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction cost effective compared to RO membranes in certain applications.

Ammonia, nitrate and nitrite are labelled as major pollutants in
aquaculture wastewater. Among these, ammonia is the main toxic
waste produced by fishes through their gills and faeces as a by-
product of protein metabolism. Ammonia is excreted from fish as
they assimilate feed. An increase in feeding rates will increase
the production of ammonia. In addition, bacterial decomposition
of organic waste solids such as uneaten feed and dead algae further
contribute to the production of ammonia.
Recently, nanofiltration (NF) membranes have become an ad-
vanced technology and are more preferable compared to MF, UF
and RO membranes due to high retention of multivalent anion salts
and organic molecular above 300 (Lu et al., 2002). Besides, NF
membranes can operate at low operating pressure and are more
Abbreviations: CA, cellulose asetate; Cp, concentration of permeate; Cr, concen-
tration of retentate; NF, nanofiltration; NaCl, sodium chloride; NMP, N-methyl-2-
pyrrolidone; MF, microfiltration; PES, polyethersulfone; PSf, polysulfone; PWP, pure
water permeation; RO, reverse osmosis; SEM, scanning electrone microscopy; SHP,
steric hindrance pore; TMS, Teorell–Meyer–Sievers; UF, ultrafiltration.
* Corresponding author. Tel.: +60 96683254; fax: +60 96694660.
E-mail address: noraaini@umt.edu.my (N. Ali).
Having a pore size which is smaller than UF and larger than RO,
NF membranes can widely separate solvent, monovalent salts,
small organics from divalent ions and larger species (Van Reis
and Zydney, 2007). Thus, a NF membrane is considered to be the
most appropriate method for water treatment separation pro-
cesses (Bessarabov and Twardowski, 2002).
It is well known that the properties of the casting solution such
as choices of polymer, solvent and non-solvent and the polymer
concentration substantially affect the final membrane microstruc-
ture and performance (Zheng et al., 2006). By changing one or
more of these variables, the thermodynamic and kinetic process
will fluctuate and thus produce different kinds of membrane struc-
tures that ultimately affect the membrane performance. Shear rate
is one of the phase inversion process parameters that commonly
influence the general formation of membrane thickness, mem-
brane porosity and pore radius. These morphology structures will
eventually affect the membrane separation performance in terms
of its selectivity and productivity. By varying the casting speed,
the shear rate induced and the molecular orientation will be al-
tered to more align structures. Shilton et al. (1997) had measured
the molecular orientation in the active layer of polysulfone mem-

0960-8524/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.biortech.2009.08.070
1460 N. Ali et al. / Bioresource Technology 101 (2010) 1459–1465

Nomenclature

Symbols r pore size (m)

Ak Membrane porosity rs solute radius (m)
Ak/Dx ratio of membrane porosity to membrane thickness rp pore radius (m)
c concentration (mol/m3) SF, SD distribution coefficients of solute by steric hindrance ef-
cp,i concentration of component i in the permeate (mol/l) fect under convection and diffusion condition, respec-
cr,i concentration of component i in the rejection (mol/l) tively
DS solute diffusivity for neutral molecule, or generalized Dx membrane thickness (m)
diffusivity for 1–1 type of electrolyte defined DP applied pressure (Pa)
D1,2 diffusivity of ion types 1 and 2 Dd effective membrane charge density (mol/m3)
F Faraday constant (=96487) (C/mol)
HF, HD steric parameters related to wall correction factors un- Greek
der convection and diffusion conditions, respectively e membrane porosity (dimensionless)
Js average solute flux over membrane surface (mol/(m2s)) g ratio of solute radius to membrane pore radius
Jv averaged volume flux over membrane surface (m/s) s tortuosity (dimensionless)
P permeability (m/s) r reflection coefficient (%)
Ps solute permeability (m/s) n ratio of fixed charge density to salt concentration
R rejection, or gas constant (8.314) (J/(mol3K))
Ri rejection of component i (%)

branes using polarised reflection infrared spectroscopy and gas
permeation tests. They found that there was an increase in molec-
ular orientation in the high sheared membranes which enhanced
the gas selectivity.
Moreover, shear during casting and spinning has been shown to
affect the flux rate of cellulose acetate (CA) hollow fibre mem-
branes (Idris et al., 2002). They found that as shear rate increased,
the flux and selectivity of the membrane also increased. The same
observation was found by Chung et al. (2000) who indicated that
selectivity increased with increasing of dope extrusion rate in hol-
low fibre spinning while the gas permeation rate reduced. Mean-
while, Ismail et al. (2002) who demonstrated the effect of shear
rate on the CA flat sheet membrane for water desalination found
that the percentage of rejection and fluxes were found to increase
with increasing of shear rate until a critical level of shear rate was
achieved. Critical shear is an optimum shear rate that can induce a
certain degree of molecular orientation to yield membrane mor-
phology with optimum separation performance (Idris et al.,
2002). Besides, as the shear rate increased, there is transformation
from thick skin to thin skin. As shear rate increased further, the
surface pores became detrimental to the rejection rates. In a study
conducted by Ismail and Hassan (2006), the optimum shear rate of
different polysulfone (PSf) concentration ranging from 19.6% to
32.8% was found at 203.20 s1. At this point, the membrane prop-
erties such as salt rejection, flux, pore radius and membrane charge
exhibited optimum values.
Most of the researchers who developed their own membranes
have not reported on the casting speed or the shear rate of the
fabricated membranes. Moreover, only a few studies reported
about the optimum shear rate that had been achieved and most
of them were conducted for gas separation and hollow fibre mem-
branes. Therefore, the present work will not only study the effect of
shear rate on the separation performance, morphological structure
and fine structural details of the flat sheet, asymmetric PES mem-
branes but also determine the best shear rate for the membrane
preparation.
OHG, Germany) was used as a solvent because of its high solvency
and low volatility.
2.1. Membrane preparation
The PES dope solution was prepared as described by Ismail et al.
(2002) where PES (18% wt.) was dissolved in NMP (82% wt.) and
the mixture was stirred at 350 rpm for about 8 h or more until
the solution became homogenous. Then, the dope solution was
put into an ultrasonic bath for about 3 h to remove bubbles and
kept at room temperature for 24 h. The asymmetric NF membranes
were fabricated via a phase inversion process using a semi-auto-
mated casting machine with the dope solution prepared earlier
being poured onto a glass plate of 30 cm  21 cm. The dope solu-
tion was spread by a casting knife with a slit height of 150 lm
across the glass plate. The membranes were cast at various casting
speeds and hence varied the shear rates were varied, as summa-
rised in Table 1. The shear rate was calculated using Eq. (1). Then,
the membrane film was immediately immersed in a water bath
and kept there for 24 h. The next day, the membrane was placed
in methanol for at least 8 h as a post-treatment to remove excess
solvent and it was then dried for one day before use.
Velocity of casting knife ðms=sÞ
Shear rate; s1 ¼ ð1Þ
Membrane thickness ðmÞ
2.2. Membrane performance measurement
Membrane performance was measured in terms of pure water
permeation (PWP) and 0.01 M sodium chloride rejection. The per-
meation test was performed using Sterlitech dead-end stirred cell.
A circular membrane with an effective membrane area of 14.6 cm2
Table 1
The casting speed used in varying asymmetric membranes properties.

Membrane ID Casting speed (s) Shear rate (s1)

2. Methods PES 1 5 400
PES 2 10 200
Polyethersulfone (PES) Radel A-300 purchased from Amoco was PES 3 15 133
preferred as a membrane material due to its high thermoplastic PES 4 20 100
PES 5 25 80
performance, excellent thermal stability and lower hydrophobic-
PES 6 30 67
ity. N-Methyl-2-pyrrolidone (NMP) from MERCK (Schuchardt
 N. Ali et al. / Bioresource Technology 101 (2010) 1459–1465 1461

was cut and mounted in a SterlitechTM HP4750 stainless steel stir membrane properties and membrane performance which have to
cell. Membranes were initially subjected to de-ionised water of be determined as in Eqs. (6)–(9).
5 bar for about 2 h before being tested. Feed pressure was con-
HD ¼ 1 ð6Þ
trolled from 2 to 10 bar using compressed nitrogen. Then, the pure
water fluxes versus pressures were plotted. Prior to testing, the 16 2
pure water flux was measured to ensure that the membranes used HF ¼ 1 þ g ð7Þ
9
were stable. For the pure water permeations (PWPs) and NaCl solu-
tions, the flux rate of the solutions, Jv, is calculated as in Eq. (2) SD ¼ ð1  gÞ2 ð8Þ
where Jv is the permeate flux (m3/m2 s), A is effective area of mem-
brane (m2), v is volume of permeate solution collected (m3) and t is
SF ¼ ð1  gÞ2 b2  ð1  gÞ2 c ð9Þ
time (s).
HF and HD are convection and diffusion steric parameters, respec-
volume permeation rate; v =t ðm3 =sÞ
Flux; J v ¼ ð2Þ tively, which are related to wall correction factors whereas SF and
membraneareas; Aðm2 Þ SD are distribution coefficients of solute by steric hindrance effect
Meanwhile, the selectivity of a membrane for a given solute is usu- under convection and diffusion conditions, respectively. Then, the
ally expressed by the percentage rejection as in Eq. (3), which is de- numerical values obtained were applied in Eq. (5) to determine
fined as the concentration of permeate Cp relative to the the reflection coefficient in order to estimate the solute permeabil-
concentration of solute in feed, Cf. ity (Ps), of membrane as in Eq. (10). By using the value of solute per-
  meability, the membrane thickness (Dx), was determined using Eq.
Cp (11).
Percentage rejection; Rð%Þ ¼ 1  100% ð3Þ
Cf  
1r
F ¼ exp  Jv ð10Þ
Ps
2.3. Analysis of ammonia–nitrogen removal
Ds
Ps ¼ ð11Þ
This test was conducted to evaluate the potential of the pro- Dx
duced PES membranes in removing ammonia–nitrogen in aquacul-
ture production. The permeation test was conducted similarly to 2.6. Pore radius
PWP and salt rejection tests. Both permeate and retentate were
collected and analysed using Hach Method 10031 based on the Pore radius (rp), was determined using Eq. (12) where the solute
salicylate method for ammonia–nitrogen detection. radius (rs), was calculated using Stokes–Einstein equation as
shown in Eq. (13) below. Based on the ionic data shown in Table
2.4. Determination of membrane structural details 2, the value of solute diffusivity (Ds), was obtained as in Eq. (14).
rs
Membrane structural details such as pore radius, membrane g¼ ð12Þ
thickness, membrane porosity, ratio of effective membrane thick-
rp
ness to membrane porosity and effective membrane charge density
can be determined using theoretical models. There were two mod-
rs ¼ kT=6plDs ð13Þ
els involved in this calculation process; the Steric Hindrance Pore
Ds ¼ 2ðD1 D2 Þ=ðD1 þ D2 Þ ð14Þ
(SHP) model and the Teorell–Meyers–Siever (TMS) model. By using
these phenomenological equations, transportation of solute inside where Ds is the diffusivity of solute molecule in a dilute solution
the membrane can be described. Sets of experimental data ob- (m2 s1), k is the Boltzmann constant, rs the Stokes radius of the sol-
tained from permeation test of sodium chloride were used in this ute (m), and l is the solvent viscosity. It is assumed that the diffu-
study. The ratio of solute radius to the membrane pore radius sivity of solute in the membrane matrix is equal to that in bulk
(f(g)), was estimated using Spiegler–Kedem equation (Van der solutions approximately.
Bruggen and Vandecasteele, 2002) as in Eq. (4).
2.7. Membrane porosity and ratio of effective membrane thickness to
Cp 2  2R
r¼1 ¼1 ¼ f ðgÞ ð4Þ membrane porosity
Cr 2R
This equation is commonly used to describe the rejection of the un- The ratio of effective membrane thickness to membrane poros-
charged molecules in aqueous solution, combining both diffusion ity (Dx/Ak) was first obtained by rearranging Eq. (15) in SHP model
and convection effects. The value of r is refers to the fraction of to get another equation as described in Eq (16). Since membrane
membrane pores that are smaller than molecules in feed solution porosity (Ak) depends on the solute permeability, effective mem-
which allow the estimation of membrane performance. Rejection, brane thickness and distribution coefficient of solute, all the
R can be determined directly from the results of sodium chloride numerical values obtained before were used to calculate mem-
rejection data. brane porosity based on Eq. (17).
Ps ¼ HD SD DS ðAk =DxÞ ð15Þ
2.5. Solute permeability and membrane thickness

In order to calculate the solute permeability (Ps), using SHP

Table 2
model, the reflection coefficient must be obtained first using Eq. Ions, ion atomic weights, ion diffusivities and solute radius data (Ismail and Hassan,
(5). 2006).

r ¼ 1  H F SF ð5Þ Ionic type Atomic weight Ds [109] (m2/s) rs [1010] (m)

Na+ 23.99 1.33 1.84
In this model, the steric hindrance factors (HF, HD, SF and SD) were
Cl 35.45 2.03 1.21
accounted as for convection and diffusion steric parameters to the
1462 N. Ali et al. / Bioresource Technology 101 (2010) 1459–1465

Dx H D SD D S the membranes were stable and the membranes were acceptable

¼ ð16Þ
Ak Ps and suitable to be used for the membrane-testing process. Based
on the PWP results, we can postulate that membranes cast at high-
Dx=Ak er shear rates would form thinner skins that result in higher flux.
Ak ¼ ð17Þ
Dx This finding is supported by the membrane cross-section and the
modelling results, which will be explained later.
2.8. Effective membrane charge density
3.2. Sodium chloride permeation
The membrane charge density Xd, is an electrical property of the
Membrane separation performance was measured in terms of
membrane where the electrostatic effects n are presented as a ratio
its selectivity and productivity towards the sodium chloride ions.
of the fixed charge density Xd to the bulk concentration of electro-
The selectivity is related to the percentage rejection of chloride
lyte Cb. Using TMS model, the electrical effects were obtained as
ions whereas productivity is the volume flowing through the mem-
rewritten in the Eqs. (18)–(20) below (Ismail and Hassan, 2006).
brane per unit area and time, which implies the permeation rate.
n The salt rejection performance was determined with 0.01 M NaCl
Xd ¼ ð18Þ
Cb solution at operating pressures ranging from 2 to 10 bar. Fig. 1
shows the rejection and flux versus pressure for salt. Rejection
2 ability of PES membranes towards chloride ions (Cl) is shown in
rsalt ¼ 1  ð19Þ
Fig. 1a and follows the sequence of: PES 2 > PES 3 > PES 4 > PES
ð2a  1Þg þ ðg2 þ 4Þ1=2
5 > PES 1 > PES 6. This trend shows that an increasing of shear rate
  increased the percentage rejection of Cl until an optimum shear
Ak
Psalt ¼ DS ð1  rsalt Þ ð20Þ rate is achieved. Beyond the optimum shear rate, the rejection
Dx
abruptly decreased. Optimum shear rate is also known as critical
shear rate, where the separation performance of membrane is
2.9. Scanning electron microscopy (SEM) the most favourable. Further increment in shear rate will cause
the molecular chains in the PES dope solution to separate from
Cross-sections of the prepared membranes were inspected with each other, which contributes to a loose and porous structure with
SEM (Model JSM 630LA) at 700  magnification. For this purpose, a lack of mechanical strength. As a result, the selectivity and the
the membrane samples were fractured in liquid nitrogen. After separation performance decreased. The consequence is clearly seen
sputtering with gold, the samples were transferred into the as the shear rate was increased to 400 s1 where the rejection of
microscope. PES 1 suddenly decreased to 14% at 10 bar applied pressure. Thus,
it could be postulated that the optimum shear rate is at 200 s1 (or
PES 2).
3. Results and discussion
Fig. 1b shows the salt fluxes of PES membranes by which the
flux increased with increasing of shear rate. However, as shear rate
3.1. Membrane permeability

The pure water permeation was first measured in order to PES 1 PES 2 PES 3 PES 4
determine the permeability coefficient of the PES NF membrane
used. The value of the permeability coefficient is directly propor- a 70 PES 5 PES 6

tional to the flux, and it is used to estimate the degree of mem- 60

brane porosity. The high permeability coefficient indicates the
Rejection (%)

50
high porosity of the membrane. As has been observed previously,
40
the pure water permeation increased as the pressure increased.
These findings corroborated with the Hagen–Poiseuille equation, 30
which has been reported in many cases (Nakao and Kimura, 20
1981) where the increase in water flux is proportional to the in-
10
crease in applied pressure ðJ v aDPÞ as expressed in Eq. (21).
0
er DP 0 2 4 6 8 10
Jv ¼ ð21Þ
8 g s Dx Pressure (bar)

The permeability coefficients of the PES membranes were calcu-

PES 1 PES 2 PES 3 PES 4
lated and tabulated in Table 3. It was shown that the permeability
coefficient increased with the increasing of shear rate and decreas- b 30 PES 5 PES 6

ing of casting speed. These results indicate that the water fluxes of 25
Flux (L/m 2.h)

20
Table 3
15
Permeability coefficient of PES membranes at different shear rate.

Shear rate/ Casting Permeability coefficient Regression 10

membrane ID (s1) speed (s) (L/m2 h bar) line R2
5
400 (PES 1) 5 11.04 0.997
200 (PES 2) 10 9.239 0.994 0
0 2 4 6 8 10
133 (PES 3) 15 7.038 0.999
100 (PES 4) 20 5.667 0.994 Pressure (bar)
80 (PES 5) 25 2.697 0.994
67 (PES 6) 30 1.592 0.991 Fig. 1. (a) Salt rejection of PES membranes at various shear rates and (b) fluxes of
sodium chloride for various shear rates of PES membranes.
 N. Ali et al. / Bioresource Technology 101 (2010) 1459–1465 1463

Table 4 Table 6
Numerical values of membrane parameter and steric hindrance factors for PES Comparison of the characteristics of 29 commercial NF membranes with the in house
membranes at various shear rates. fabricated PES membrane (Bowen and Mohammad, 1998).

Membrane ID/shear rate (s1) g r HF SF SD Parameters rp (nm) n

PES 1 (400) 0.103 0.020 1.019 0.962 0.805 Commercial membrane Minimum 0.39 1.5
PES 2 (200) 0.520 0.405 1.495 0.412 0.239 Mean 0.66 9.2
PES 3 (133) 0.386 0.226 1.265 0.612 0.377 Maximum 1.59 44.5
PES 4 (100) 0.304 0.145 1.165 0.734 0.484
Locally fabricated membrane PES 2 0.31 1.92
PES 5 (80) 0.158 0.044 1.045 0.915 0.708
PES 6 (67) 0.081 0.013 1.011 0.976 0.845

goes beyond the optimum point (>200 s1), the flux decreased. By
PES 2 PES 3 PES 4
altering the shear rate during membrane casting, the membrane 80
PES 5 PES 6
structure and morphology will change. At higher shear rate, the 70
pore size of the membrane skin decreases. The decrement in pore 60

Rejection (%)
size resulted in better rejection of solute solution. This finding
50
agrees with those reported by Ismail et al. (2002), which claimed
40
that increment in shear rate increased the membrane selectivity.
30
Moreover, casting membrane with a higher shear rate will increase
the molecular orientation. It is believed that the molecular chains 20
of the dope solution experienced at higher shear rate tend to align 10
themselves much better than those experiencing at lower shear 0
rate. Enhancing the molecular orientation will cause the polymer 0 2 4 6 8 10
molecules to become closely packed (Chung et al., 1998) thus Pressure (bar)
enhancing the separation performance with better rejection. How-
ever, those findings only take place until optimum shear rate is 20 PES 2 PES 3 PES 4
achieved and beyond the optimum point the inverted conditions 18 PES 5 PES 6
occur. 16
14
Flux (L/m 2.h)

3.3. Characterization of structural parameter details using Steric
Hindrance Pore (SHP) Model
Tables 4 and 5 show the modelling result values that represent
the steric hindrance parameters and the fine structural details of
PES membranes produced. In order to obtain the pore radius of
the membranes, the values of r, HF, SF, HD and SD were first deter-
mined based on Eqs. (5)–(9) as described before. Based on this the-
oretical approach, it was found that the value of the reflection
coefficient, r, strongly affected the rejection rates. As r values in-
creased, the rejection of PES membranes increased. It was revealed
that increasing of shear rate will increase the r values and thus in-
crease the rejection rates.
Generally, the effective membrane thickness, Dx increased as
the shear rate decreased, as shown in Table 5. The PES membranes
developed show the following sequence of membrane thicknesses:
PES 6 (20.7  103 m) > PES 5 (14.1  103 m) > PES 4 (8.06 
103 m) > PES 3 (7.80  103 m) > PES 2 (5.67  103 m) < PES 1
(5.78  103 m). A slight increment of membrane thickness was
observed for PES 1 as compared to PES 2. The rate of permeation
or fluxes will directly affected by membrane thickness, as shown
in Fig. 1b by which flux deteriorated as shear rate decreased. How-
ever, the formation of pore radius shows an abrupt decrement of
pore radius values for PES 1 (rp = 1.50 nm) and PES 2 (rp = 0.31 nm),
as shown in Table 5. Beyond the shear rate of 200 s1, pore radius
12
10
8
6
4
2
0
0 2 4 6 8 10
Pressure (bar)
Fig. 2. (a) Fluxes of ammonia–nitrogen at various shear rates and (b) removal of
ammonia–nitrogen at various shear rates for PES membranes.
of the membrane for PES 3, PES 4, PES 5 and PES 6 grows bigger. A
similar trend was shown by the rejection of Cl as shown in Fig. 1a
whereby PES2 exhibited the highest rejection performance. Apart
from sieving effects, the high retention of salt was also influenced
by the Donnan effect, where ions with the same charge as the
membrane are repelled whereas ions with the opposite charge will
drag passing through the membrane. In this case, the effective
membrane charge density of PES membranes was estimated using
the TMS model, and the values obtained are presented in Table 5.
Based on the data, all the membranes produced were negatively
charged at the surface layer, with PES 2 the most negatively
charged. As reported by Wang et al. (1995), the charge density in-
creased as pore radius decreased, which resulted in high rejection.

Table 5
Modelling results of PES membranes at different shear rate.

Membrane ID/shear rate (s1) Ps Dx (103)m rp (nm) Ak Dx/Ak (103) xd (ve) n (ve)
7 9
PES 1 (400) 3.21  10 5.78 1.50  10 1.17 4.93 0.0005 0.052
PES 2 (200) 3.10  107 5.67 0.31  109 2.76 2.21 0.0191 1.914
PES 3 (133) 2.37  107 7.80 0.40  109 2.08 3.82 0.0077 0.774
PES 4 (100) 1.97  107 8.06 0.51  109 1.77 4.81 0.0048 0.476
PES 5 (80) 1.45  107 14.1 0.93  109 1.33 10.8 0.0012 0.123
PES 6 (67) 9.48  108 20.7 1.59  109 1.19 17.8 0.0003 0.034
1464 N. Ali et al. / Bioresource Technology 101 (2010) 1459–1465

This statement was in line with the results obtained in this study
where PES 2 obtained the highest rejection of chloride ions with
the smallest pore size. It also revealed that the optimum shear rate
was found at 200 s1 or PES2 membrane, by which its thickness
and pore radius was found to be preeminent. Furthermore, the
modelling results revealed that PES 2 in this study has a similar po-
tential with the 29 commercial NF membranes reported by Bowen
and Mohammad (1998), which is based on the value of pore radius
and ratio of effective charge density to bulk electrolyte concentra-
tion, as summarised in Table 6.
3.4. Removal of ammonia–nitrogen: effect of shear rate
The analysis results of ammonia–nitrogen are shown in Fig. 2a
and b. It was clearly observed that, as shear rate increased, the flux
and rejection of total ammonia–nitrogen (TAN) increased. As the
shear rate reached the optimum point (PES 2), the rejection of
TAN achieved was 68% and the flux was unexpectedly higher than
anticipated. Shear rate is one of the phase inversion process
parameters that generally influenced the morphology of the mem-
brane such as membrane thickness and membrane porosity. As the

Fig. 3. SEM image of PES membranes at various shear rate. (a) Shear rate 67 s1 (PES 6), (b) shear rate 80 s1, (PES 5), (c) shear rate 100 s1, (PES 4), (d) 133 s1 (PES 3), (e)
200 s1 (PES 2), (f) 400 s1 (PES 1).
 N. Ali et al. / Bioresource Technology 101 (2010) 1459–1465
shear rate increased, the pore size of the membrane decreased due
to the effect of molecular orientation. It was believed that, at high-
er shear rates, the molecular chains of the dope solution tend to
align themselves much better than those experiencing a lower
shear rate. Enhancing the molecular orientation will cause the
polymer molecules to become closely packed. As a result, the rejec-
tion of solute will increase. This finding is supported by the calcu-
lated values of its characteristic properties using theoretical
approach and morphology inspection observed from SEM micro-
graph images, whereby membranes with higher shear rates show
denser substructure.
3.5. Morphology inspection using SEM
The SEM electron micrographs of cross-sections of the PES
membranes at various shear rates are shown in Fig. 3. As illus-
trated, all PES membranes developed in this study comprised fin-
ger-like and sponge-like substructures. According to Brown et al.
(2002), casting a membrane from a polymer concentration of 15–
18% will give a regular structure of finger voids. Moreover, mem-
branes fabricated using a phase inversion process often produced
sponge-like and finger-like structures in the membrane substruc-
ture (Zhou et al., 2008). It was clearly seen that, as shear rate in-
creased, the morphological structure of the PES membranes
changed from spongy to a more compact structure. Generally, all
of the PES membranes comprise a finger-like structure with mac-
rovoids at the bottom of the layer. A relatively thick transition
layer was observed in the membrane at the lowest shear rate
and the skin became thinner as the shear rate increased to
200 s1. A thinner skin layer resulted in higher flux rate as previ-
ously reported. Moreover, the finger-like structure becomes nar-
rower as the shear rate increased and the macrovoid structure
suppressed, which enhanced the rejection ability. The morphology
image also revealed that at optimum shear, more uniform finger-
like structures were found to penetrate through the film. At this
point, both rejection and permeability are at the optimum level
as reported before. However, as shear rate increased to 400 s1,
the molecular chains seem to be pulled away and cause slight
imperfections in the skin layer as seen in Fig. 3f. As a result, the
selectivity decreased. This is probably due to the early demixing
and precipitation of PES 1, which resulted in a porous layer.
4. Conclusion
Varying the shear rate during membrane casting has significant
effects on the morphology and structural details of the membranes.
1465
By changing the shear rate, the effective pore radius, membrane
thickness and charge density will vary and thus change the mem-
brane performance. It was revealed, that as shear rate increased,
the pore radius was reduced while the membrane thickness in-
creased, causing an increment in selectivity but tending to lower
the selectivity. In this study, a shear rate of 200 s1 at a 10 s casting
speed during membrane fabrication is the most recommended be-
cause its separation performance was the most favourable. Also,
the PES 2 obtained the highest retention of salt and ammonia–
nitrogen at 61% and 68%, respectively.
References
Bessarabov, D., Twardowski, Z., 2002. Industrial Application of Nanofiltration – New
Perspectives, Memb. Tech. (Chapter 2).
Bowen, W.R., Mohammad, A.W., 1998. Characterization and prediction of
nanofiltration membrane performance – a general assessment. Trans. Ichem.
E 76, 885–893.
Brown, P.J., Ying, S., Yang, J., 2002. Morphological structure of polyetherketone
membranes for gas separation prepared by phase inversion. AUTEX Res. J. 2,
101–108.
Chung, T.S., Teoh, S.K., Lau, W.W.Y., Srinivasan, M.P., 1998. Effect of shear stress
within the spinneret on hollow fibre membrane morphology and separation
performance. Ind. Eng. Chem. Res. 37, 3930–3938.
Chung, T.S., Lin, W.H., Vora, R.H., 2000. The effect of shear rates on gas separation
performance of 6FDA-durene polyimide hollow fibre. J. Membr. Sci. 167, 55–66.
Idris, A., Ismail, A.F., Noorhayati, M., Shilton, S.J., 2002. Measurement of
rheologically induced molecular orientation using attenuated total reflection
infrared dichroism in reverse osmosis hollow fibre cellulose acetate membrane
and influence on separation performance. J. Membr. Sci. 213, 45–54.
Ismail, A.F., Hassan, A.R., 2006. Formation and characterization of asymmetric
nanofiltration membrane: effect of shear rate and polymer concentration. J.
Membr. Sci. 270, 57–72.
Ismail, A.F., Hassan, A.R., Cheer, N.B., 2002. Effect of shear rate on the performance of
nanofiltration membrane for water desalination. Songklanakrin J. Sci. Technol.
24, 879–889.
Lu, X., Bian, X., Shi, L., 2002. Preparation and characterization of NF composite
membrane. J. Membr. Sci. 210, 3–11.
Nakao, S., Kimura, S., 1981. Analysis of solutes rejection in ultraÒ ltration. J. Chem.
Eng. Jpn. 14, 32–37.
Shilton, S.J., Ismail, A.F., Gough, P.J., 1997. Molecular orientation and the
performance of synthetic polymeric membranes for gas separation. Polymer
38, 2215–2220.
Van der Bruggen, B., Vandecasteele, C., 2002. Modelling of the retention of
uncharged molecules with nanofiltration. J. Water Res. 36, 1360–1368.
Van Reis, R., Zydney, A.L., 2007. Bioprocess membrane technology. J. Membr. Sci.
297, 16.
Wang, X.L., Tsuru, T., Togoh, M., Nakao, S.I., Kimura, S., 1995. Evaluation of pore
structure and electrical properties of nanofiltration membranes. J. Chem. Eng.
Jpn. 28 (2), 186.
Zheng, Q.Z., Wang, P., Yang, Y.N., 2006. Rheological and thermodynamic variation in
polysulfone solution by PEG introduction and its effect on kinetics of membrane
formation via phase-inversion process. J. Membr. Sci. 279, 230.
Zhou, J., Ren, J., Lin, L., Deng, M., 2008. Morphology evolution of thickness-gradient
membranes prepared by wet phase-inversion process. Sep. Purif. Technol. 63,
484–486.