Professional Documents
Culture Documents
Environmental and Clean Technology Laboratory, Suzhou Institute of Sichuan University, Suzhou, China
School of Civil and Environmental Engineering, Nanyang Technological University, Singapore
Singapore Membrane Technology Centre, Nanyang Technological University, Singapore
d
Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, Nanjing, China
b
c
a r t i c l e
i n f o
Article history:
Received 22 July 2009
Received in revised form 8 November 2009
Accepted 9 November 2009
Available online 13 November 2009
Keywords:
Forward osmosis (FO)
Pressure retarded osmosis (PRO)
Reverse osmosis (RO)
Internal concentration polarization (ICP)
Osmotic driving force
a b s t r a c t
Forward osmosis (FO) and pressure retarded osmosis (PRO) are concentration-driven membrane processes. While they can be potentially used in water, wastewater, and energy applications, these processes
suffer from the concentration polarization inside the porous membrane support resulting in severe ux
decrease, a phenomenon known as internal concentration polarization (ICP). Researchers have investigated the effect of ICP both in theoretical and experimental studies. The current study extends the
existing ICP model to include the effect of draw solution dilution by membrane permeate ow in a spiral
wound FO module (SWFO). FO and PRO experiments were performed using a Hydrowell SWFO under
both submerged and cross-ow conditions. The effect of draw solution concentration, draw solution ow
rate, feed water ow rate, and membrane orientation on FO and PRO water ux performance was systematically investigated. Permeate ow increased with greater draw solution concentration in both FO and
PRO modes. ICP was found to drastically limit the available membrane ux in the concentration-driven
membrane processes, and its adverse effect was more severe at greater draw solution concentration.
Membrane ux was also affected by the dilution of draw solution when the permeate ow rate was comparable or greater than the draw solution ow rate. The submerged FO conguration performed nearly
as good as the cross-ow conguration with feed water circulating outside of the membrane envelope
(shorter ow path). In this case, the feed water ow rate only had limited effect on membrane ux likely
due to its low mass transfer resistance. In contrary, the membrane ux can be adversely affected at low
feed water ow rate when it was circulated inside of the membrane envelope (longer ow path).
2009 Elsevier B.V. All rights reserved.
1. Introduction
Membrane separation has gained increasing popularity in
water, wastewater, and many other industrial applications.
Traditionally, pressure-driven membrane processes, such as microltration (MF), ultraltration (UF), nanoltration (NF), and reverse
osmosis (RO), have received much attention. For example, lowpressure porous MF and UF membranes are widely used for surface
water treatment [1,2], in membrane bioreactors (MBRs) [3,4], as
well as for pretreatment of RO processes [5]. In parallel, the
application of reverse osmosis (RO) is fuelled by the increasing
needs for seawater desalination, water treatment, and wastewater reclamation [5,6], along with the signicant improvements of
RO membrane properties and module design [6,7]. On the other
hand, although their full-scale implementation is still lacking, there
is also increased interest in concentration-driven forward osmosis
(FO) and pressure retarded osmosis (PRO) processes for water and
energy applications due to their potential for energy production (or
saving) [810].
During FO, a high concentration solution (draw solution) is separated from a low concentration solution by a semi-permeable
(water permeable but not solute permeable) membrane (Fig. 1(a)).
Due to the osmotic pressure difference across the membrane, water
ows spontaneously through the membrane from the low concentration side to the draw solution side. This water permeating
through the FO membrane is of high quality, with nearly complete
retention of organic matter, particulates, and microorganisms as
well as signicant retention of dissolved salts [1013]. Where a natural source of osmotic energy is available (e.g., seawater) [8,9,14],
FO can be highly attractive since the process requires very little additional energy input (except a small amount of energy is
299
Fig. 1. Illustration of FO, PRO, and RO processes. (a) FO process where no pressure is applied on the high concentration solution. Water ows from low concentration side
to high concentration side. (b) PRO process where applied pressure on the high concentration solution is less than the osmotic pressure difference across the membrane.
Water ows from low concentration side to high concentration side. (c) RO process where applied pressure on the high concentration solution is greater than the osmotic
pressure difference across the membrane. Water ows from high concentration side to low concentration side. (d) Classication of FO, PRO, and RO in a ux versus pressure
plot. Adapted from Refs. [8,10].
Fig. 2. Modeling internal concentration polarization. (a) Active layer is facing feed water. Draw solution concentration in the support layer is diluted by the water permeate
ux, causing the available driving force (the concentration difference across the rejection layer) to be much lesser than the apparent driving force (concentration difference
between the draw solution and the feed water). (b) Active layer is facing draw solution. The solute concentration at the rejection-support layer interface is much higher than
the feed concentration, leading to reduced available driving force. Adapted from [8,10].
300
of the dense rejection layer and the porous support layer (Csupport )
is signicantly lower than the bulk draw solution concentration
(Cdraw ) due to the convective transport of solute away from the
support back to the draw solution (dilutive internal concentration polarization (DICP) [23]). Consequently, the available driving
force (i.e., the concentration difference across the dense membrane
rejection layer) is much smaller than the apparent concentration
difference between the draw solution and the feed [8,10]. Similar
ICP effect has also been reported when the draw solution is oriented towards the dense rejection layer (Fig. 2(b)), where solute
from the feed solution and that transmitted through the membrane from the draw solution are accumulated and concentrated in
the support layer [12,22,27]. In this case, the concentration in the
support (Csupport ) is signicantly higher than the feed water concentration (Cfeed ), which in turn reduces the available driving force
(concentrative internal concentration polarization (CICP) [23]).
The effect of internal concentration polarization can be modeled
by adopting the classical solution-diffusion theory for the dense
rejection layer coupled with convection and diffusion transport of
the solute in the porous support layer [8,22,24]. Following Lee et
al. [8] and Leob et al. [24] (refer to Appendix A), the water ux in
FO can be expressed as
Jv = Km ln
Adraw + B
Afeed + Jv + B
Jv = Km ln
Adraw Jv + B
Afeed + B
(2)
D
S
(3)
Jv = Km ln
2. Theory
Adraw + B((AP/Jv ) + 1)
(Afeed + Jv + AP) + B((AP/Jv ) + 1)
(1)
Jv = Km ln
(Adraw AP Jv ) + B((AP/Jv ) + 1)
Afeed + B((AP/Jv ) + 1)
(4)
(5)
301
Fig. 3. Experimental setup. (a) An illustration of an open Hydrowell spiral wound module. (b) A cross-ow ltration test setup where the spiral wound membrane module
was housed inside a pressure vessel. Cross-ows can be maintained both inside and outside of the membrane envelope. (c) Different test congurations adopted in the
current study.
For the special case where P = 0, Eqs. (4) and (5) become identical
to Eqs. (1) and (2), respectively.
Eqs. (1), (2), (4), and (5) are valid for small membrane coupons
where the membrane permeate ow (Qp ) is small compared to the
ow rate of draw solution (i.e., the cross-ow of draw solution, Qcf ),
e.g., in batch experiments with low recoveries. In a spiral wound
module, however, the permeate ow can signicantly dilute the
bulk draw solution concentration, which results in reduced ux
performance compared to small coupon tests. Such dilution effect
needs to be explicitly accounted for in a spiral wound module. Consider a module with a draw solution concentration Cdraw,0 and a
cross-ow of Qcf,0 at the module inlet. The bulk draw solution concentration Cdraw,x anywhere inside the module can be determined
by
Cdraw,x = Cdraw,0
Qcf,0
Qcf,0 +
x
(6)
J dAm
0 v
x
x
via a volumetric dilution factor (Qcf,0 /(Qcf,0 + 0 Jv dAm )). At the outlet of the module, the draw solution is diluted to a concentration
Cdraw,exit :
Cdraw,exit = Cdraw,0
Qcf,0
Qcf,0 + Qp
(7)
where Qp is the total permeate ow rate through the entire membrane module, i.e.:
Qp =
exit
Jv dAm
(8)
Qp
Am
(9)
Eq. (6) can be used to determine the bulk draw solution concentration Cdraw anywhere in the module, and Cdraw can then be used
to determine the local membrane ux Jv via the FO and/or PRO ux
models (Eqs. (1), (2), (4), or (5)). Thus, Eq. (6), coupled with the
ICP ux models, can be used for ux performance modeling of spiral wound FO/PRO modules. A numerical example is provided in
Appendix A.
For PRO applications, the power output is of interest. The total
power available from the module (PRO ) can be determined as a
302
(10)
The power density, that is, the total power normalized by the
membrane area in the module, is given by
PRO = PJv
(11)
cross-ow rate at the feed solution side. A separate pump was used
to circulate the draw solution from a 40-L stainless steel pressure
tank. The pressure inside the draw solution tank was controlled
by adjusting a needle valve located at the top of the tank, and it
was measured by a mechanical pressure gauge (Fig. 3(b)). Where
no back pressure was required in the FO mode, the needle valve
was fully open so that the pressure inside the draw solution tank
reached atmospheric pressure.
303
Fig. 4. Effect of draw solution concentration on permeate ow in s-FO-in conguration. The error bars shown in the gure represent one standard deviation based
on at least three different tests. Note: 1 L/m2 h = 2.78 107 m/s.
304
Fig. 6. Effect of draw solution cross-ow on permeate ow in s-FO-in conguration. Both experimental (solid symbols) and simulation results (continuous lines)
were shown. The s-FO-in simulation results were obtained from Eq. (6) coupled
with ICP models (Section 2 and Appendix A). The error bars shown in the gure represent one standard deviation based on at least three different tests. Note:
1 L/m2 h = 2.78 107 m/s.
(Cdraw,0 > 0.5 M). In both congurations, the draw solution was circulated at a rate of 75 ml/min. While a cross-ow of 140 ml/min
pure water was circulated on the feed water side for the x-FO-in
conguration, no forced circulation was available for submerged
one. Thus, the slightly better ux in x-FO-in conguration at greater
Cdraw,0 was likely due to the reduced mass transfer resistance and
ECP on the feed water side, which is consistent with our earlier
discussion in Section 4.1. For comparison purpose, simulation was
performed based on Eqs. (1) and (6). It is worthwhile to note that
the simple theoretical model in the current study does not take into
consideration the feed water cross-ow by assuming that the membrane ux is not limited by the mass transfer and ECP on the feed
water side. Therefore, identical simulation results were obtained
for both x-FO-in and s-FO-in congurations. A comparison between
the simulation results and the experimental data shows (1) that
the model prediction agreed better with the experimental results
at lower draw solution concentration; and (2) that the model simulation worked better for the x-FO-in conguration. Both trends can
be explained by the lower mass transfer resistance and ECP on the
feed water side under these conditions.
The permeate ow of the x-FO-out conguration is also presented in Fig. 7. This conguration (where the dense active layer
was facing the draw solution) had higher ux compared to the
s-FO-in and x-FO-in congurations where active layer was facing the feed water. Similar results have been reported by various
research groups [12,13,22]. Such difference was likely due to the
more severe ICP in the s-FO-in and x-FO-in congurations, where a
dilutive concentration polarization occurred in the support (i.e., the
draw solution was diluted inside the porous support layer to cause
a drastic loss of driving force, Fig. 2(a)). This explanation was also
conrmed by the simulation results based on Eqs. (2) and (6)the
simulated permeate ow for x-FO-out is clearly superior to those
for x-FO-in and s-FO-in (Fig. 7).
Surprisingly, the experimental results in the x-FO-out conguration were signicantly lower compared to the simulated results
(Fig. 7). It is hypothesized that the mass transfer resistance on the
feed water side was much greater in this conguration due to its
much longer ow path. The ow path of feed water in the x-FOout conguration was 4.5 m (twice the length of the membrane
leaf (Fig. 3(a)) as compared to a 0.3 m ow path length for x-FO-in
where feed water owed outside of the membrane envelope. The x-
305
FO-out mode could result in more severe salt accumulation and ECP
on the feed water side inside the membrane envelope. The effect
of feed water ow rate on module performance in x-FO-out conguration is better illustrated in Fig. 8 for different feed and draw
ow rates. At a draw solution cross-ow of 75 ml/min, the permeate ow at a feed ow rate of 60 ml/min was drastically lower than
that at 140 ml/min feed water ow. Indeed, the maximum permeate ow rate (60 ml/min) was identical to the feed ow rate at the
spiral wound module inlet. When this maximum permeate ow
was achieved (which corresponds to 0.5 and 0.8 M draw solutions),
the feed water ow at the module outlet was almost zero. Under
this condition, nearly all the available feed water passed through
the membrane resulting in a limitation of the maximum permeate ow to this amount. The same was true for the combination
of a 140 ml/min draw solution and a 140 ml/min feed water. Fig. 8
clearly suggests that the permeate ow in the x-FO-out conguration was constrained by the feed water ow rate in addition to
the ICP effect and the draw solution dilution effect. In contrast, the
effect of feed water ow rate in the x-FO-in conguration was much
less important. The ux performance of s-FO-in (where there was
no forced circulation of feed water) was nearly identical to that of
x-FO-in, as a result of signicantly shorter ow path of feed water.
4.3. x-PRO congurations
The PRO performance of the Hydrowell spiral wound module was evaluated in both x-PRO-in and x-PRO-out congurations
(Fig. 9). While the permeate ow in the x-PRO-out conguration
seems to be more stable, the ux performance in x-PRO-in deteriorated drastically upon applying a back pressure. In the x-PRO-in
conguration, the draw solution that owed inside the membrane
envelope was pressurized. Major leakage of draw solution was
identied at a pressure of 3 bar which forced the experiment to
be stopped. The leakage was due to the failure of the glue between
the two pieces of membranes (Fig. 3(a)) under the applied pressure.
Similar failures were observed when additional Hydrowell modules were tested (data not reported here). It is concluded that the
Hydrowell module is not suitable for the x-PRO-in conguration,
except where applied pressure is relatively low. Where it is desirable to have the dense rejection layer facing the feed water (e.g.,
for fouling control [12,13]), the module may need to be redesigned
such that the rejection layers are inside of the membrane envelope and pressure is applied outside of the envelope to ensure the
integrity of the module.
306
Fig. 9. Comparison of x-PRO-out and x-PRO-in congurations. Draw solution concentration at 0.5 M. The open symbols represent the respective uxes obtained in
the x-FO-out and x-FO-in congurations. Note: 1 L/m2 h = 2.78 107 m/s.
Fig. 10. Effect of applied pressure on permeate ow in the x-PRO-out conguration at different draw solution concentrations. Cross-ow rate of draw solution
at 75 ml/min. Simulation results were obtained from Eq. (6) coupled with Eq. (5)
(Section 2 and Appendix A). Note: 1 L/m2 h = 2.78 107 m/s.
Fig. 11. Available power density from the Hydrowell spiral wound module in xPRO-out conguration. A 0.5 M NaCl was used as the draw solution.
307
Acknowledgements
This research received nancial support from Suzhou Industrial
Park Technology Innovation Zone and Suzhou Institute of Sichuan
University. Xu and Peng were supported by Suzhou Institute of
Sichuan University.
Appendix A.
A.1. Derivation of ICP models
Based on the classical solution-diffusion model for the rejection
layer, we have
Jv = A(draw support P)
(A1)
(A2)
D dC
dx
(A3)
(A4)
and
C = Cfeed at x = t
(A5)
Jv
Km
(A6)
where
Km =
S=
D
D
=
t
S
(A7)
t
(A8)
Jv
Km
(A9)
Jv = Km ln
(Adraw AP Jv ) + B((AP/Jv ) + 1)
Afeed + B((AP/Jv ) + 1)
(5)
Eq. (2) can be obtained by setting P = 0 in Eq. (5). Eqs. (1) and (4) can
be derived in a similar fashion.
A.2. An example for modeling FO/PRO permeate ow in a spiral
wound module
The local membrane ux anywhere inside the spiral wound
module can be modeled by ICP ux models presented in Section 2
(Eqs. (1)(5)). Based on the at coupon FO test results and additional
independent RO tests by She [12], the HTI membrane used in the
current studies had the following properties: A = 2.2 1012 m/s Pa
308
Table A1
An example for simulating spiral wound FO permeate uxa .
Stepb
Incremental membrane
area, Am (m2 )b
Incremental permeate
ow, Qp,x (ml/min)e
Accumulative
permeate ow, Qp,x
(ml/min)f
1
2
3
4
100
0.500
0.492
0.484
0.477
0.237
7.90
7.79
7.68
7.58
4.04
0.0094
0.0094
0.0094
0.0094
0.0094
1.24
1.22
1.20
1.19
0.63
1.24
2.46
3.66
4.85
83.87
Notes:
a
Simulation conditions: s-FO-in/x-FO-in congurations; total effective membrane area in the module of 0.94 m2 ; a 0.5 M draw solution with a ow rate of 75 ml/min at
the module inlet.
b
The total membrane area was divided into 100 equal small areas to compute the local incremental permeate ow.
c
The local draw solution was determined from the inlet draw solution concentration according to Eq. (6).
d
The local ux was obtained as a function of local draw solution concentration from Fig. A.1. Unit conversion factor: 1 L/m2 h = 2.78 107 m/s.
e
The incremental permeate ow was determined from Qp,x = Jv,x Am .
f
and B = 1.7 107 m/s. The mass transfer coefcient Km inside the
membrane support was 3.3 106 m/s when the dense rejection
layer faced the draw solution, and it was 4.2 106 m/s in the alternative membrane orientation. These values agreed reasonably well
with Gray et al. [22]. Based on Eqs. (1) and (2), the local FO ux can
be determined as a function of local draw solution concentration
(Fig. A.1). Similarly, ux in PRO can be determined by Eqs. (4) and
(5) (results not shown here).
The FO/PRO permeate ow rate of a spiral wound module can be
modeled using both the ICP equations (e.g., the ux-concentration
equations in Fig. A.1 for FO mode) together with Eq. (6) which
accounts for the dilution effect. A worked example is shown in
Table A1. With a 0.5 M NaCl draw solution at a ow rate of
75 ml/min at the module inlet, the local membrane ux is as high as
7.9 L/m2 h according to Fig. A.1. As the draw solution ows through
the module, its concentration becomes increasingly diluted by the
permeate ow. A total of 84 ml/min permeate ow is available from
the entire membrane module. Consequently, the concentration
at the module outlet is determined by 0.5 75/(75 + 84) = 0.24 M,
which is much lower than the inlet concentration. The corresponding local ux at the outlet is only 4.0 L/m2 h. Clearly, the ux
x
0
Jv dAm .
draw
feed
PRO
PRO
References
[1] E. Aoustin, A.I. Schafer, A.G. Fane, T.D. Waite, Ultraltration of natural organic
matter, Separation and Purication Technology 22 (3) (2001) 63.
Fig. A.1. Local permeate ux in FO mode as a function of local draw solution concentration. Data obtained from Ref. [12]. The permeate ux was approximated
by the respective equations in the gure for both membrane orientations. Note:
1 L/m2 h = 2.78 107 m/s.
309
[25] J.R. McCutcheon, M. Elimelech, Inuence of membrane support layer hydrophobicity on water ux in osmotically driven membrane processes, Journal of
Membrane Science 318 (2008) 458466.
[26] W. Tang, H.Y. Ng, Concentration of brine by forward osmosis: performance and
inuence of membrane structure, Desalination 224 (2008) 143153.
[27] T.Y. Cath, D. Adams, A.E. Childress, Membrane contactor processes for wastewater reclamation in space: II. Combined direct osmosis, osmotic distillation,
and membrane distillation for treatment of metabolic wastewater, Journal of
Membrane Science 257 (2005) 111119.
[28] A. Achilli, T.Y. Cath, A.E. Childress, Power generation with pressure retarded
osmosis: an experimental and theoretical investigation, Journal of Membrane
Science 343 (2009) 4252.
[29] http://www.htiwater.com/.
[30] C.Y. Tang, Y.-N. Kwon, J.O. Leckie, Effect of membrane chemistry and coating
layer on physiochemical properties of thin lm composite polyamide RO and
NF membranes. I. FTIR and XPS characterization of polyamide and coating layer
chemistry, Desalination 242 (2009) 149167.
[31] C.Y. Tang, Y.-N. Kwon, J.O. Leckie, Effect of membrane chemistry and coating layer on physiochemical properties of thin lm composite polyamide
RO and NF membranes II. Membrane physiochemical properties and their
dependence on polyamide and coating layers, Desalination 242 (2009)
168182.
[32] S. Loeb, Large-scale power production by pressure-retarded osmosis, using
river water and sea water passing through spiral modules, Desalination 143
(2002) 115122.
[33] K.B. Petrotos, P. Quantick, H. Petropakis, A study of the direct osmotic concentration of tomato juice in tubular membranemodule conguration. I. The
effect of certain basic process parameters on the process performance, Journal
of Membrane Science 150 (1998) 99110.
[34] J.R. McCutcheon, R.L. McGinnis, M. Elimelech, Desalination by
ammoniacarbon dioxide forward osmosis: inuence of draw and feed
solution concentrations on process performance, Journal of Membrane Science
278 (2006) 114123.
[35] R.L. McGinnis, M. Elimelech, Energy requirements of ammoniacarbon dioxide
forward osmosis desalination, Desalination 207 (2007) 370382.
[36] P. Bacchin, D. Si-Hassen, V. Starov, M.J. Clifton, P. Aimar, A unifying model
for concentration polarization, gel-layer formation and particle deposition in
cross-ow membrane ltration of colloidal suspensions, Chemical Engineering
Science 57 (2002) 77.
[37] T.H. Chong, F.S. Wong, A.G. Fane, Implications of critical ux and cake
enhanced osmotic pressure (CEOP) on colloidal fouling in reverse osmosis: experimental observations, Journal of Membrane Science 314 (2008)
101111.
[38] E.M.V. Hoek, M. Elimelech, Cake-enhanced concentration polarization: a new
fouling mechanism for salt-rejecting membranes, Environmental Science &
Technology 37 (2003) 5581.
[39] T.H. Chong, F.S. Wong, A.G. Fane, Enhanced concentration polarization by
unstirred fouling layers in reverse osmosis: detection by sodium chloride tracer response technique, Journal of Membrane Science 287 (2007)
198210.
[40] R.J. Aaberg, Osmotic power: a new and powerful renewable energy source?
Refocus 4 (12) (2003) 4850.
[41] W. Ludwig, A. Seppala, M.J. Lampinen, Experimental study of the osmotic
behaviour of reverse osmosis membranes for different NaCl solutions and
hydrostatic pressure differences, Experimental Thermal and Fluid Science 26
(2002) 963969.