This article was downloaded by: [University of Southern Queensland]

On: 10 October 2014, At: 14:11

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Chemical Engineering Communications

Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/gcec20

THERMODYNAMIC MODELING OF SOLUTE TRANSPORT

THROUGH REVERSE OSMOSIS MEMBRANE
a
LIANFA SONG
a
School of Engineering, The Hong Kong University of Science and Technology , Clear Water
Bay, Kowloon, Hong Kong
Published online: 31 Oct 2007.

To cite this article: LIANFA SONG (2000) THERMODYNAMIC MODELING OF SOLUTE TRANSPORT THROUGH REVERSE OSMOSIS
MEMBRANE, Chemical Engineering Communications, 180:1, 145-167, DOI: 10.1080/00986440008912206

To link to this article: http://dx.doi.org/10.1080/00986440008912206

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
 Chm. &. c-.. ZWO. Vol. 180, pp. 145-167 0 MOO OPA (Dr- F'ublishen Auodalion) N.V.
RcprinU avnilrble dirstly from the publisher Published by licmre undcr
Photocopying p r m i w d by liscnw only thc Gordon and Brash Sci-
P u b l i i imprint
Rinted in Mslsyda.
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

.THERMODYNAMIC MODELING OF SOLUTE

TRANSPORT THROUGH REVERSE
OSMOSIS MEMBRANE
LIANFA SONG*
School of Engineering. The Hong Kong University of Science and Technology,
Clear Water Bay. Kowloon, Hong Kong

(Received 10 November 1998; In fiat form 16 September 1999)

A new thermodynamic model is developed for water and solute transports through reverse
osmosis membranes. The model is featured with rigorous derivations in theoretical development
and clearly defined parameters for membrane transport properties. The new model can comctly
describe not only the dependence of salt rejection on pressure and salt concentration, but also
the non-linearity between water flux and pressure. Comparisons of model simulations with the
reported revem osmosis experiments demonstrate that the parameters in the new model are
concentration-independent. This study shows that water and salt transports through reverse
osmosis membranes can be satisfactorily described with irreversible thermodynamics.

Keywordr: Membrane transpow salt rejection; irreversible thermodynamics; modeling; reverse

osmosis

1. INTRODUCTION

Reverse osmosis (RO) processes are widely used for separation and concen-
tration of solutes in many fields, such as chemical and biomedical industry,
food and beverage processing, and water and wastewater treatment. Sepa-
ration and concentration in RO processes are achieved with membranes
that have different transport properties for different constituents in the
solutions. However, solute rejection by RO membranes is not determined
solely by the membrane properties. The operational conditions, such as
salt concentration and applied pressure, have tremendous influence on the
 146 L. SONG

performance of membrane process. It is experimentally observed that salt

rejection 'by RO membranes increases monotonically with pressure and
decreases with salt concentration [I -51.
Numerous theoretical attempts have been made to delineate the under-
lying mechanisms of membrane separation and to explain the dependence of
salt rejection on membrane properties and operational conditions [l-41.
Well-known mechanistic models include the solution-diffusion and finely
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

porous models. The solution-diffusion model assumes that both water and
solute diffuse independently across the membrane while a convective flow of
substantial amount of solute is considered in the finely porous model.
Membrane transport is also described with the thermodynamic models that
are developed from the phenomenological linear law of the irreversible
thermodynamics [6,7l. No specific transport mechanisms are required in the
thermodynamic models. Both mechanistic and thermodynamic models
result in similar expressions for water and salt fluxes though interpretations
of the coefficients in these expressions are different [l -41.
These mechanistic and thermodynamic models can satisfactorily correlate
salt rejection with applied pressure or permeate flux if the parameters are
determined directly from the experimental data [5,8- 101. However, many
parameters in these models, which are originally assumed to depend on the
membrane properties only, are found strongly dependent on salt concentra-
tion. At present, the parameters in these models can only be calculated
directly from the experimental datasince there are no intrinsic relations to
determine these parameters from the membrane properties and the
operational conditions. Because of the dependence of the model parameters
on the operational conditions, practical applications of these models are
very limited. The dependence of model parameters on operational condi-
tions is often attributed to the large pressure and concentration gradients
across the membrane in RO processes. Such RO systems are considered
far away from equilibrium so that the irreversible thermodynamics is
not applicable. This argument is questionable if we consider the extremely
small fluxes in most RO processes. The concentration-dependent parameters
can also be introduced from the use of inappropriate assumptions or
transformations in model development.
A new thermodynamic model is developed in this paper with special
attention in handling the concentration-related terms. With concentration-
independent parameters, the new model can satisfactorily describe the
dependence of salt rejection on pressure and salt concentration and the
nonlinearity between permeate flux and pressure. Theoretical predictions of
the new model are compared with the experimental RO data reported by
 THERMODYNAMIC MODELING OF SOLUTE TRANSPORT 147

Pusch [5] and good agreements are obtained. It is demonstrated in this work
that water and salt transports through RO membranes are satisfactorily
described with the principles of irreversible thermodynamics. Because of the
independence of the model parameters on the operational conditions, the
new model possesses the desired capacity to predict membrane transport
under any conditions. The model also provides us a new perspective to
understand transport mechanisms through membranes and an innovative
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

method in membrane characterization.

2. CRITICAL REVIEW OF PREVIOUS IRREVERSIBLE

THERMODYNAMIC MODELS

Irreversible thermodynamics is an extension of the classical thermodynamics

for systems that are not at equilibrium but near equilibrium. One important
principle of the irreversible thermodynamics is that the fluxes of all trans-
port processes are linearly related to the corresponding driving forces in a
system that is not far away from equilibrium. The irreversible thermo-
dynamics is one of the most important methods in describing transports
through the RO membranes. In this section, two important thermodyna-
mic models for membrane transport, the Kedem- Katchalsky and Spiegler-
Kedem models, are critically reviewed.

2.1. Kedem - Katchalsky Model

Kedem and Katchalsky model [6] is the first practical model for solute
transport across the membrane developed upon the irreversible thermo-
dynamics. It starts with the energy dissipation function

dm = J , [ R T A ( ~a,) + VwAp]+J,[RTA(ln a,) + V,Ap] (2.1)

where I$,,, is the dissipation function for transport cross the membrane, R is
the ideal gas constant, T is the absolute temperature, J, and J, are the water
and solute fluxes, a, and a, are the solute and water activities, V, and V, are
the solute and water partial molar volumes, and Ap is the pressure difference
across the membrane. The terms in the square brackets in Eq. (2.1) are the
chemical potential differences for water and solute across the membrane,
respectively. Equation (2.1) is rearranged in the terms of measurable
quantities, the differences of concentration and pressure across the
 148 L. SONG

membrane as

where As is the osmotic pressure difference across the membrane, and (cJln
is the logarithmic average salt concentration. The osmotic pressure s of a
non-electrolyte is given by
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

The logarithmic average concentration (c,)~, is defined as

where c, and c: are the solute concentrations on the feed and permeate sides
of the membrane, respectively. By defining two new fluxes Jv= JwVw+ J,Vs
and JD= JS/(cs),, - J,V,, Eq. (2.2) is transformed to

The above transform looks straightforward but there are hidden

problems. It is well known that the concentration related chemical potential
of a solute is R T Inc,, as used in Eq. (2.1). This is the intrinsic relationship
between chemical potential and solute concentration that cannot be linear-
ized by introducing Eq. (2.4). The linear term AT(= RTc,) in Eq. (2.5) is
not always correct for the concentration-related chemical potential of solute.
With the using of the logarithmic average concentration (c,),, in defining
JD, the concentration-related potential is partially incorporated into the
flux JD.
The application of the linear law of irreversible thermodynamics to the
fluxes and driving forces just defined above results in

where Lp,LPD, LDp and LD are permeability constants. The Kedem-

Katchalsky model is finally written as:
 THERMODYNAMIC MODELING OF SOLUTE TRANSPORT 149

where u is the reflection coefficient and w is the solute permeability at zero

volume flux defined below, respectively
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

Because of the problematic transformation discussed above, the linear

relationships (2.6) and (2.7) between the fluxes and the driving force may not
be held. In other words, the coefficient Lp, LD and LpD can be functions of
the driving forces, rather than constants. At least, the irreversible thermo-
dynamics does not guarantee that these coefficients are independent of pres-
sure and concentration after the transformation. Therefore, it becomes
questionable if the reflection coefficient u,as defined with Eq. (2.10), can
be treated as constant.
The Kedem-Katchalsky model is widely used in correlating experimental
data and the coefficients of the model are shown to be dependent on
pressure and concentration to various extents. This discrepancy with the
irreversible thermodynamics is usually attributed to the large pressure and
concentration differences across the membrane. However, the possibility
cannot be excluded that the discrepancy is brought about by the trans-
formation from the common chemical potentials to the measurable driving
forces, pressure and concentration.

2.2. Spiegler- Kedem Model

It is noted that the linear law of the irreversible thermodynamics may be
inapplicable for the whole membrane thickness because of the large pressure
and concentration differences involved. As an improvement of Kedem-
Katchalsky model, the linear law of the irreversible thermodynamics in
Spiegler and Kedem model is written for an infinitesimal thickness of
membrane as

where L I 1 ,L I 2 ,LZIand Lzz are phenomenological coefficients. Equations

(2.12) and (2.13) are more appropriate than Eqs. (2.6) and (2.7) in applying
 150 L. SONG

the irreversible thermodynamic principle because of the explicit involvement

of water and salt concentrations in the equations.
Rearranging of Eqs. (2.12) and (2.13) results in
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

where

The Onsager reciprocity relation L12 = Lzl is used in reaching Eq. (2.15). It
is obvious that both o and p, defined by Eqs. (2.16) and (2.17) are strong
functions of salt concentration.
However, both p, and o are treated as constants in the integration of Eq.
(2.15) in order to obtain the following expression:

where r is the solute rejection. With the definitions given in Eqs. (2.16) and
(2.17), the assumption of a and p, being constants is inappropriate and
cannot be used. For this reason, the applicability of Eq. (2.18) for salt re-
jection by membranes is doubtful.
In summary, the handling of the concentration-related terms is one of the
toughest challenges in applying thermodynamic principles to membrane
transport. There are spaces left in both Kedem-Katchalsky and Spiegler-
Kedem models for further improvement.

3. THEORETICAL DEVELOPMENT

3.1. .Driving Forces for Species Transport Through Membrane

The driving forces in general are the free energy difference of the species
across the membrane. The species on the side of higher free energy are
 THERMODYNAMIC MODELING OF SOLUTE TRANSPORT IS1

driven to the side of lower free energy. For a solution of single salt that can
completely ionizes in water, the driving forces for water and salt compo-
nents can be written in the following forms:

FW= AEw= X,[RTA(ln a,) +v , ~ ~ ] (3.1)

F, = AE,= X,[vRTA(ln a,) + v,Ap]

Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

(3.2)

where F, and F, are the drivhg forces for water and salt in solution, Xwand
X, are the water and salt molar fractions, Y is the number of ions ionized
from a salt molecule. The expressions in the square brackets are the
chemical potentials. The physical meanings of F, and F, are the changes of
free energy in the feed solution when Xwmole of water and X, mole of salt
are transported through the membrane, respectively. Since salt and water
concentrations are different, the true driving force of any constituent i in a
solution should be the chemical potential corrected by the molar fraction Xi.
This point will become clear in the discussion below. In the following
discussions, X, << X, is generally assumed.
The activity terms in Eqs. (3.1) and (3.2) can be expressed in terms of
concentrations as below

X,RTA(ln a,) = X,RTA

where vT = 1/(c, + c,) is the partial molar volume of the whole solution, 7,
and 7, are the activity coefficients for water and salt, respectively. Both
activity coefficients are assumed unity in Eqs. (3.3) and (3.4).
Substituting Eqs. (3.3) and (3.4) into Eqs. (3.1) and (3.2) results in
 152 L. SONG

When the solute concentration difference across the membrane is small, we

have

Then Eq. (3.6) becomes

Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

Equations (3.5) and (3.8) show that the driving forces for water and salt are
comprised of two parts, namely, the applied pressure and the osmotic
pressure resulted from salt concentration difference. The contributions of
osmotic pressure to the driving forces of salt and water are same in
magnitude but opposite in direction. When there is no applied pressure (i.e.,
A p = O), the driving forces for salt and water in the solution are VTuRTAcS
and - V T v ~ ~ ~respectively.
c,, This has been verified by countless experi-
mental observations. The appearance of applied pressure in Eqs. (3.5) and
(3.8) in terms of partial pressures ( v w c w A pand v,c,Ap) is also reasonable
because the transport of solute through membranes is much less sensitive to
the applied pressure than that of water. .

3.2. Modiied Linear Law of Irreversible Thermodynamics

According to irreversible thermodynamics, the fluxes of water and salt are
linearly related with the driving forces:

where Lid (i,j = 1,2) are the permeability coefficients. The molar fractions
X, and X, appear again in Eqs. (3.9) and (3.10), respectively. Transport of
species through the membrane is different from the transport processes in
unrestricted media (e.g., diffusion in liquid) where the flux is directly
proportional to the driving force. In the current problem, the flux of a
species must be affected by the effective membrane area available for the
transport of the species. In this study, the share of the membrane surface
area by a species is assumed to be proportional to its molar fractions in the
solution. With Xu and X, in Eqs. (3.9) and (3.10), the coupling terms (the
second term in Eq. (3.9) and the first term in Eq. (3.10)) become meaningful.
 THERMODYNAMIC MODELING OF SOLUTE TRANSPORT 153

Otherwise, it is hard to understand that the coupling transport of salt is only

determined by the water driving force but completely unaffected by salt
concentration. In many previous studies on membrane transport, the linear
law of irreversible thermodynamics was used without consideration of the
competition between species for the membrane surface. However, the re-
sultant permeability coefficients from these studies showed dependence on
salt concentration to various extents [I]. This fact suggests that the compe-
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

tition between species for membrane surface has to be considered.

Substituting Eqs. (3.5) and (3.6) into Eqs. (3.9) and (3.10) yields

Let us introduce the coupling coefficients A and A, defined as

then Eqs. (3.11) and (3.12) become

3.3. Salt Rejection

Though membrane separation can be fully describe with Eqs. (3.15) and
(3.16), the performance of membrane processes is more conveniently or
customarily specified with the salt rejection r, defined as
 154 L. SONG

where c: is the solute concentration in the permeate. The terms Ac, and
A l n ( ~ ~ ~inc Eqs.
, ) (3.15) and (3.16) are related to the salt rejection r with
the following equations:
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

Substituting Eqs. (3.18) and (3.19) into Eqs. (3.15) and (3.16) results in

Jw= L ~ + ~- vRTcsr
~ XwvScs)Ap
~ ~ - XwvRTc,In(l
[ ~ -~r)] (3.20)
c ~

J, = L22~&[P,c, + Xvwcw)Ap- XvRTc,r - vRTc,In(l - r)] (3.21)

3.4. Membrane Transport Model

Now let us consider the calculation of salt rejection r with Eq. (3.17). From
the physical meaning, the permeate salt concentration cf can be determined
from the fluxes with

The reciprocal of feed salt concentration c, can be rewritten as

Substituting Eqs. (3.22) and (3.23) into Eq. (3.17) results in

Rearranging Eq. (3.24) yields

 THERMODYNAMIC MODELING OF SOLUTE TRANSPORT 155

Combining Eqs. (3.20), (3.21) and (3.25) gives

where S is the membrane selectivity defined as

Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

- - -
Since cw/cs >> rVs/V,, A, << 1 and V w c , r VTC, 1 for most reverse
osmosis processes, Eqs. (3.20) and (3.26) can be reduced to

It can be seen from Eq. (3.29) that, when pressure is sufficiently high, i.e.,
Ap + m, salt rejection approaches to

where r , is the maximum salt rejection. Equation (3.30) shows that r -P 1 as

S-tm.
In most cases, the volumetric permeate flux, J,, is well approximated with
the volumetric water flux, i.e.,

where R, is the membrane resistance to pure water, defined as

In the above derivation, there are three quantities of concern, namely J,,
Js and r. However, because of the relationships among these three quantities
defined by Eqs. (3.21), (3.29) and (3.31), only two quantities are indepen-
dent. The salt rejection r and the volumetric permeate flux J . are commonly
used as the primary quantities, which are readily calculated with Eqs. (3.29)
and (3.31).
 156 L. SONG

3.5. Determining r with an Iteration Scheme

An explicit expression of r is always preferable but it seems difficult to
obtain at present because r appears on both sides of Eq. (3.29) in quite
complex forms. Further study is needed to find out if there is a possibility to
express r explicitly with some reasonable simplifications. However, this is
not a serious obstacle because r can be easily determined with a simple
iteration scheme as described below.
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

Equation (3.29) can be used directly in the iteration calculation. However,

preliminary calculations show that sometimes it does not work properly. A
more robust and stable procedure can be developed when Eq. (3.29) is
rewritten as

-
where A P (= A p vRTc,r) is the effective applied pressure.
The scheme starts with an initial arbitrary guess of salt rejection, r(O); then
rt0)is substituted into the right hand side of Eq. (3.33) to calculate the first
round estimate, d'); r(') is then substituted into Eq. (3.33) for the next
round. This process is continued until r does not change. In mathematical
language, this iteration scheme can be described as
Initial guess r(O) (0 < r(O) < 1)

-
Stop when Ir("'+') r(")l < E E is the convergence criterion
The author's experience shows that Eq. (3.33) is well behaved with respect
to the above iteration scheme. It only takes a few iterations to converge for
most situations with any initial guess of r(O'(0 < r(O) < 1). The scheme can be
implemented on an MS Excel spreadsheet or more efficiently with a small
FORTRAN program.
 THERMODYNAMIC MODELING OF SOLUTE TRANSPORT 157

3.6. Features of the New Model

Though the new model is developed with the same thermodynamic prin-
ciples as used in Kedem-Katchalsky and Spiegler-Kedem models, the
new model possesses some new features that do not exist in the previous
models. There is no place in the new model for the so-called reflection
coefficient a. This is a fundamental difference between the new model and
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

the previous models for water flux. With a constant u, Kedem-Katchalsky

and Spiegler-Kedem models predict that psrmeate flux has a linear
dependence on the pressure and that the flux is not affected by the actual salt
rejection. The predicted flux 'approaches zero when Ap approaches CAT.
In the new model, the salt rejection r, rather than the reflection coefficient a ,
appears in the similar place. It is rigorously shown that the water flux is
affected by salt rejection r. The use of r instead of a i n determining the effective
driving forces seems more reasonable because it is the part of salt rejected by
the membrane that builds up the osmotic pressure difference. Since r becomes
smaller as the applied pressure decreases, appreciable fluxes can be obtained
under pressures below the osmotic pressure of the solution. Furthermore,
because of the dependence of salt rejection on pressure, there is no longer a
linear relationship between water flux and pressure in the new model.
The membrane selectivity S and coupling coefficient X are needed in the
new model to determine salt rejection and both parameters have clearly
defined physical meanings. The parameter S is the ratio of the conductivity
of water to that of salt in the membrane. The parameter X describes the
coupling between water and salt transports. When X = 0, the transports of
water and salt through the membrane are completely independent. It can be
seen from Eq. (3.21) that in this situation the effect of the applied pressure
on salt transport minimizes and the driving force for salt transport is
strongly affected by concentration difference. On the other side, when X = 1,
salt transport is strongly affected by the applied pressure (viscous flow) and
the concentration difference has less effect on salt transport.

The R O experiments of NaCl solutions reported by Pusch [5] are extensively

used for model testing and verification. In that study, Pusch systematically
investigated the effects of pressure and salt concentration on permeate flux
and salt rejection for various membranes. The RO experimental system was
carefully designed to minimize the effect of concentration polarization. In
 158 L. SONG

this section, the experimental data on CA-90 and CA-85 membranes in

Pusch's study are used to compare with the theoretical predictions of the
new model.
Partial molar volumes of water and salt at different salt concentrations are
needed in the new model. With an assumption of constant partial molar
volume for water (1.8 x 10-'m3/mole) in dilute solutions, the partial molar
volume of salt can be determined from solution density. Let p (g/m3) be the
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

solution density at concentration c, (mole/m3), the partial molar volume of

salt is given by

The partial molar volumes of salt at various salt concentrations are

calculated from the density data reported in Perry's Chemical Engineers'
Handbook [ I l l and listed in Table I. The salt partial molar volume at
vanishing salt concentration can be determined with Eq. (4.1). As c,
approaches zero, p approaches lo6 and the salt partial molar volume
..
avoroaches 5.845 x 10-5m'/mole. The first four data points in Table I. as
well as the point for vanishing concentration (5.845 x ;o-~), are fitted with
an exponential decay curve. The resultant curve is plotted in Figure 1 with
symbols being the data points. From Figure 1, the salt partial molar
volumes of 5.6 x lo-', 4.1 x and 2.4 x 10-'m3/mole are identified
for salt concentrations of 0.01, 0.10, OSOM, respectively.
The parameters (S,X and R,) of CA-90 and CA-85 membranes are
determined with the method of trial-and-error. Model simulations with
different parameters are compared with the experimental data and satis-
factory values can be obtained in a few trials. The parameters thus identi-
fied are listed in Table 11.

TABLE I Density and salt partial molar volume of NaCl solution at different
concentrations
 THERMODYNAMIC MODELING OF SOLUTE TRANSPORT 159
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

Concentration (M)
FIGURE 1 Salt (NaCI) partial molar volume as a function of salt concentration.

TABLE I1 Transport parameters of the new model

Membrane S X R, (Pa-slm)
CA-90 7 0.005 4.1 x 1012
CA-85 3 . 0.01 2.1 x 10l2

Model simulations (lines) are plotted with experimental data (symbols) in

Figures 2 and 3 for CA-90 and CA-85 membranes, respectively. Figures 2a
and 3a demonstrate that salt rejections at three concentrations are well
reproduced by model simulations with constant parameter values for both
membranes. This fact verifies that the selectivity S a n d coupling coefficient X
of CA-90 and CA-85 membranes are thermodynamic properties of the mem-
branes and independent of pressure and salt concentration.
Plotted in Figures 2b and 3b are the theoretical simulations and ex-
perimental measurements of permeate fluxes at different concentrations. The
theoretical values of salt rejection reported in Figures 2a and 3a are used
for the simulations. With a constant membrane resistance R, = 4.1 x 1012or
2.1 x 1 0 ' ~ ~ a - s / mrespectively,
, the model simulations for both membranes
show excellent agreements with the experimental fluxes for salt concentra-
tions of 0.01 and 0.10M. The model also reproduces the nonlinear
relationship between flux and pressure. This nonlinear dependence of
permeate flux on pressure has not been theoretically predicted before.
The experimental fluxes for 0.50M solution are smaller than the model
predictions. There are two possible reasons for the deviation of theoretical
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014
 THERMODYNAMIC MODELING OF SOLUTE TRANSPORT 161
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

Pressure (Pa)

Pressure (Pa)
FIGURE 3 Comparison of model simulations with experimental measurements of Pucsh [5]
on membrane CA-80. (a) Rejection; (b) Flux. S = 3, X = 0.01 and R. = 2.10 x 1 0 ' Pa-s/m
~ are
used in the model simulations.

TABLE 111 Transport parameten of Spiegler-Kedem model [q

c,'(molll) lo6(cm/s..tm) u (PN
.lo6 (cm/4
CA-90
0.01 2.46 0.995 1.619
0.10 2.45 0.988 4.329
0.50 2.11 0.921 6.242
CA-85
0.01 4.89 0.979 5.855
0.10 4.89 0.967 18.69
0.50 3.81 0.793 23.92
 162 L. SONG

5. SIMULATIONS AND DISCUSSION

The comparisons of model simulations with experimental data show that the
parameters in the new model are concentration-independent. This means
that water and salt transports through RO membranes can be correctly
described with irreversible thermodynamics. Therefore, the properly for-
mulated model based on thermodynamics can provide a powerful tool to
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

investigate the effect of membrane properties on the performance of RO

processes. As a demonstration, the effect of membrane selectivity and salt
concentration on the performance of RO processes is investigated with the
new model in this section.

5.1. Effect of Selectivity S on Salt Rejection r and Flux J,

Salt rejection and permeate flux for membranes with different selectivity
are simulated with the new model and plotted in Figures 4a and 4b, respec-
tively. In the simulations, membrane resistance R, = 5.0 x 1 0 ' ~ ~ a - s land
m
salt concentration c,'= 1.00M are used. The values of selectivity are 2 , 5 ,
10,100 and 1000, respectively. As shown in Figure 4a, at a given pressure
the higher selectivity always results in the higher rejection. However, salt
rejection of all membranes increases with increasing pressure.
Figure 4b shows that there is an obvious nonlinear relationship between
the permeate flux and pressure. There are basically two distinguished pres-
sure ranges separated by the osmotic pressure of the solution. Permeate
fluxes increase slowly with pressure when the pressure is lower than the
solution osmotic pressure. As pressure becomes higher than the solution
osmotic pressure, the fluxes approach a Linear relationship with pressure.
This linear dependence of the permeate flux on pressure for high salt
concentration may not be experimentally observed due to concentration
polarization under high pressures. Figure 4b also shows that higher selec-
tivity produces higher curvature in the flux vs. pressure curves.
Furthermore, Figure 4b shows that membrane selectivity has a
tremendous inthence on the magnitude of permeate fluxes. Membranes
with higher selectivity produce lower permeate fluxes. The reason is that the
membranes with higher selectivity have higher salt rejections (as shown in
Fig. 4a). According to Eq. (3.28), the net driving force for water through
-
that membrane, Ap vRTrc,, is smaller. Therefore, the flux through that
membrane will be smaller though all membranes under consideration have
same resistance for pure water.
 THERMODYNAMIC MODELING OF SOLVE TRANSPORT 163
-
-
-
-
-
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

V."
-
0.0 5.0x108 1.0xl0~ 1.5x107

Pressure (Pa)

0.0 5.0~10~ 1.oxl 0' 1 . 5 ~o7

1

Pressure (Pa)
FIGURE 4 Simulations of (a) rejections and (b) fluxes for membranes with different sclcc-
tivities. ( I ) S = 2; (2) S = 5; (3) S = 10; (4) S I()(); and (5) S = 1000. Other parameten are:
A = 0, R,,,= 5 x 10" Pa-slm and c, = 1.00M.

5.2. Effect of Salt Concentration c, on Rejection r and Flux J.

Model simulations are conducted to investigate the effect of salt concentra-
tion on salt rejection and permeate flux. The calculated rejections and fluxes
are plotted in Figures 5a and 5b, respectively. Selectivity of 100 and
membrane resistance of 5.0 x 10" Pa-s/m are used in these simulations. Salt
concentrations under investigation are 0.01,0.10,0.50,1.00 and 2.00M,
respectively. Figure 5a demonstrates the strong dependence of rejection on
salt concentration and the rejection reduces drastically as salt concentra-
tion increases. A membrane that has a very high rejection at low salt
 L. SONG

-
=
-
-
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

-
0.0 5.0~10" 1.0xl0' 1.5x107

Pressure (Pa)

0.0 5.0~10~ 1.oX10' 1.5x107

Pressure (Pa)

c, --
FIGURE 5 Simulations of (a) rejections and (b) fluxes for different salt concentrations. (I)
0.01 M; (2) c, 0.lOM; (3) c, = 0.50M; (4) c, = 1.00M; (5) C, = 2.00M. Other parameters
am: A 0 , R, = 5 x 10" Pa-s/m and S = 100.
'

concentration can show a much lower rejection at high salt concentration.

For example, the membrane shows a near complete rejection at pressure
lower than 5 x 106pa when salt concentration is lower than 0.10M. But the
same membrane rejects salt slightly higher than 90% for salt concentration
of 2.00 M at a pressure as high as 1.5 x 1o7Pa.The simulations indicate that
for desalination of seawater or wastewater of high salt contents, both
membranes of very high quality (high selectivity) and very high pressure
have to be used.
As expected, Figure 5b shows that the permeate flux is strongly affected
by salt concentration. These simulations further demonstrate that only the
part of salt rejected by the membrane works against the driving force Ap. It
 THERMODYNAMIC MODELING OF SOLUTE TRANSPORT 165

is clear in Figure 5b that the permeate flux decreases as salt concentration

increases. However, the differences between the fluxes of different con-
centrations become smaller when pressure decreases. The linearity of flux
with pressure does not exist and the flux of higher salt concentration shows
stronger nonlinearity.

5.3. An Innovative Method in Membrane Characterization

Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

The parameters in the new model are independent of pressure and salt
concentration. Therefore, when the selectivity, coupling coefficient, and
resistance of a membrane are identified, the performance of the membrane
under any operating conditions (pressure and salt concentration) can be
simulated with the new model. Besides, the parameters in the new model
have clearly defined physical meanings and can be easily determined with
experiments. All these features suggest that the transport properties of RO
membranes can be better characterized with the parameters introduced with
this new model. The adaptation of these parameters in membrane
characterization will significantly improve the communications between
manufacturers and users of membrane processes.

6. CONCLUSIONS

Transport through RO membranes is a process that can be well described by

the irreversible thermodynamics. It is evidenced by the parameters in the
new model that are independent on pressure and concentration. The ag-
reement of model simulations with RO experiments demonstrates that the
effects of pressure and salt concentration on the performance of RO mem-
branes are correctly predicted with the new model.
The linear law of irreversible thermodynamics between fluxes and driving
forces is generally valid when the driving forces are given in terms of free
energy differences in solution. The driving forces can be represented in linear
function of pressure and concentration only for special cases where the salt
concentration difference across the membrane is small (Eqs. (3.7) and (3.8)).
Since large salt concentration differences usually exist across most RO
membranes, the intrinsic dependence of chemical potential on the logarithm
of salt concentration has to be considered.
The transport properties of RO membranes can be adequately
characterized with membrane selectivity, coupling coefficient, and pure
water resistance. These parameters are rigorously defined and can be easily
 166 L. SONG

determined with experiments. Because of the independence of these

parameters on pressure and salt concentration, the new model is capable
to predict the performance of RO processes under any operational
conditions.

NOTATIONS
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

Activity of salt component

Activity of water component
Feed solute concentration (mole/m3)
Permeate solute concentration (mole/m3)
Logarithmic average salt concentration (mole/m3)
Water concentration (mole/m3)
Diffusion flux in Kedem-Katchalsky model
Salt flux (mole/m2. s)
Volumetric permeate flux (m3/m2,s or m/s)
Water flux (mole/m2. s)
Permeability coefficients for flux of component i by
driving force j
Permeability coefficients in Kedem-Katchalsky model
Salt diffusion coefficient
Gas constant (= 8.31 JK-' mole-')
Salt rejection
Maximum salt rejection as Ap +0
Membrane resistance to pure water (Pa-s/m)
Selectivity of membrane
Absolute temperature
Partial molar volume of salt (m3/mole)
Partial molar volume of water (m3/mole)
Partial molar volume of solution (m3/mole)
Molar fraction of salt
Molar fraction of water

Greek Letters
Applied pressure (Pa)
Effective applied pressure (Pa)
Convergence criterion
Dissipation function for membrane transport
 THERMODYNAMIC MODELING OF SOLUTE TRANSPORT 167

Activity coefficient of solute

Activity coefficient of water
Coupling coefficient
Number of ions in one salt molecule
Osmotic pressure
Reflection coefficient
Downloaded by [University of Southern Queensland] at 14:11 10 October 2014

References
(11 Soltanieh, M. and Gill. W.N. (1981). Chem. Eng. Commw., 12, 279.
[2] Mason, E. A. and Lonsdale, H. K. (1990). 1.Membr. Sci., 51, 1.
[3] Jonsson, G. (1980). Desalination, 35, 21.
[4] Pusch, W. (1977). Ber. Bunrenges. Phys. Chem., 81,269.
[q Pusch, W. (1977). Ber. Bwrrenges. Phys. Chem., 81, 854.
16) Kedem, 0. and Katchalsky, A. (1958). Biochimica ef Biophysicco Acta, 27, 229.
[A Spieglcr, K. S. and Kedcm, 0. (1966). Desalination, 1, 31 1.
181 Pia, Lipp, Rolf, Gimbel and Fritz H. Frimmel (1994). J. Membr. Sci., 95, 185.
[9] Richard I.. Urama and Benito J. Marinas (1997). 1. Membr. Sci., 123. 267.
[lo] Van Gauwbergcn, D. and Baeyens, J. (1995). Sep. ond PuriJ Tech., 13, 117.
[I I] Perry, R. H., Chilton, C. H. and Kirkpatrick. S. D., Chemicd Engineers' Hana'book,
(McGraw-Hill Book Company Inc., New York, 1963). 4th edn., Chap. 3, p. 70.