Journal of Membrane Science 201 (2002) 203212

Modeling of cross-ow osmotic pressure controlled membrane

separation processes under turbulent ow conditions
Kosaraju Sreenivas, P. Ragesh, Sunando DasGupta, Sirshendu De

Department of Chemical Engineering, Indian Institute of Technology, Kharagpur 721302, India

Received 30 March 2001; received in revised form 30 March 2001; accepted 2 November 2001
Abstract
A comprehensive predictive model for the osmotic pressure controlled turbulent cross-ow membrane separation process
is presented. A thin rectangular channel is selected as the system geometry. The model can handle solutes with wide range of
diffusivities (from sodium chloride to dextran) under different modes of membrane separations (RO, UF, etc.). It is not limited
to the condition that the mass transfer boundary layer lies within the viscous sublayer of the hydrodynamic boundary layer. The
governing equations are solved using an integral method. The performance characteristics of the system in terms of permeate
ux and permeate concentration can be predicted simultaneously. The experimental data available in the literature under
turbulent conditions [Ind. Eng. Chem. Fundam. 3 (1964) 210; J. Membr. Sci. 22 (1985) 117] are successfully compared with
the model predicted results. Aparametric study has been carried out to observe the effects of various operating conditions on the
permeate ux and permeate concentration (in terms of observed retention). 2002 Elsevier Science B.V. All rights reserved.
Keywords: Cross-ow; Diffusivities; Hydrodynamic boundary layer; Turbulent ow; Permeate ux
1. Introduction
One of the major problems associated with mem-
brane separation processes, which restricts the
widespread application of this process in the industry
is the decline in ux. This occurs due to the build up
of the solute concentration near the membrane sur-
face, which is dened as concentration polarization
[1]. The transport of species towards the membrane
is strongly inuenced by the conditions inside the
hydrodynamic and mass transfer boundary layers.
Therefore, the control of the boundary layers is cru-
cial for the optimization of the process. The conven-
tional method to reduce concentration polarization

Corresponding author. Tel.: 91-3222-83926 (Off)/83927 (Res);

fax: 91-3222-55303.
E-mail address: sde@che.iitkgp.ernet.in (S. De).
is to increase turbulence e.g. stirring the solution or
cross-ow of the feed. An accurate quantication
of concentration polarization as a function of pro-
cess conditions is necessary to estimate the system
performance satisfactorily.
Many researchers attempted modeling of UF/RO
systems in the past [25]. Initial models are based
on lm theory [1,6] where concentration boundary
layer is assumed to be fully developed. This led to
the under-prediction of ux. The models which con-
sider developing concentration boundary layer contain
assumptions like no axial variation of permeate ux
[7,8], surface concentration [9,10], and simplied hy-
drodynamics, etc. [11,12].
In order to solve the governing mass transfer equa-
tion one needs to evaluate the detailed velocity proles
in the particular system. There has been a considerable
effort towards developing a detailed transport model
0376-7388/02/$ see front matter 2002 Elsevier Science B.V. All rights reserved.
PII: S0376- 7388( 01) 00730- X
204 K. Sreenivas et al. / Journal of Membrane Science 201 (2002) 203212
Nomenclature
A damping length constant
(26(
w
/)
1/2
)
B
1
, B
2
, B
3
viral coefcients
c solute concentration (kg/m
3
)
D diffusivity (m
2
/s)
d
e
equivalent diameter (=4 h) (m)
h Channel half height (m)
k Von Karmans constant (0.4)
l Prandtl mixing length (m)
L channel length (m)
L
p
membrane permeability (m/(Pa s))
P transmembrane pressure (Pa)
Re Reynolds number (u
o
d
e
/)
R
o
observed rejection
R
r
intrinsic rejection
u axial velocity (m/s)
u

friction velocity (

w
/) (m/s)
u

dimensionless axial velocity (u/u

)
u
o
time averaged axial velocity (m/s)
u
/
v
/
time averaged velocity uctuation
(m
2
/s
2
)
v vertical velocity (m/s)
v
w
permeate ux (m/s)
!v
w
) average permeate ux (m/s)
v

w
dimensionless suction velocity (v
w
/u

)
x axial coordinate (m)
y transverse coordinate (m)
y

dimensionless transverse coordinate

(yu

/)
Greek letters

1
,
2
,

3
,
4
coefcients in Eqs. (A.1)(A.4)
concentration boundary layer
thickness (m)

dimensionless concentration
boundary layer thickness, dened
as

= (/)u

kinematic viscosity (m
2
/s)
osmotic pressure (Pa)
osmotic pressure difference across
the membrane (Pa)
density (kg/m
3
)
total shear stress (N/m
2
)

w
wall shear stress (N/m
2
)
coefcient in Eq. (15)
Subscripts
m membrane surface
o feed bulk
p permeate
in a variety of channel geometries. Belfort and Klein-
streuer [13] have reviewed much of the analytical and
numerical work in this area. These provide insight
into the uid ow problems in the membrane systems.
Most of the work in modeling of UF/RO systems
has been undertaken in the laminar range [2,3,14,15].
This severely restricts their applicability, as these sys-
tems often operate under turbulent ow conditions or
employ turbulence promoters. Silva-freire [16], Sucec
and Oljaca [17] have undertaken detailed calculation
for turbulent transpiration ows. However, these so-
lutions are for the case of external ows and when
the free stream velocity is known as a function of the
axial distance. The transpiration considered in their
solution is injection and it is taken to be constant with
axial distance. The most common approach for mod-
eling turbulence in membrane separation processes is
to use empirical correlations for the mass transfer co-
efcients [18]. However, these correlations are system
specic and valid in certain ranges of the operating
conditions thereby limiting their applicability.
Recently, more fundamental modeling of turbulent
cross-ow UF was attempted [11,12]. These analyses
are valid for low diffusivity (e.g. dextran T70) solutes.
For these solutes, the concentration boundary layer
lies within the viscous sub-layer (of the hydrodynamic
boundary layer). Another simplied assumption re-
garding the hydrodynamics is that the x-component
velocity remains unaffected by membrane perme-
ation. Moreover, the turbulent mass transport terms
(Reynolds stresses) are neglected, which may not be
true outside the viscous sub-layer (of the hydrody-
namic boundary layer). Hence, such analysis cannot
account for the ltration of the high diffusivity so-
lutes (sodium chloride, sucrose, etc.) for which the
concentration boundary layer may grow beyond the
viscous sub-layer in to the buffer zone or even in to
the turbulent core.
In the present work, most of the above draw-
backs for the osmotic pressure controlled turbulent
cross-ow UF/RO have been taken care of. The axial
K. Sreenivas et al. / Journal of Membrane Science 201 (2002) 203212 205
Fig. 1. Schematic of the system.
velocity prole is developed using the van Driest mix-
ing length hypothesis. This is then coupled to the mass
conservation equation. The numerical solution of the
model equations predicts the system performance in
terms of permeate ux and permeate concentration.
The model predictions are compared with the experi-
mental data available in the literature [19,20].
1.1. Theory
The ow conguration is a rectangular channel as
shown in the Fig. 1. The uid ows axially over the
membrane surface. The process is assumed to be in
steady state with fully developed turbulent hydrody-
namic boundary layer. Physical properties of the com-
ponents are assumed to be constant.
1.1.1. Hydrodynamic considerations
The momentum equation for a two dimensional, in-
compressible, turbulent boundary layer with zero pres-
sure can be written as
u
u
x
v
u
y
=
1

y
(1)
where is the total shear stress and is expressed as
=
l

t
=
u
y
u
/
v
/
(2)
where
l
is the laminar and
t
the turbulent shear stress
[21].
The boundary conditions for Eq. (1) are
u = 0, v = v
w
: =
w
at y = 0 (3)
For the fully developed hydrodynamic boundary layer,
the Eq. (1) reduces to
v
w
u
y
=

y
(4)
Integrating the above equation yields the following
expression for the shear stress
=
w
uv
w
(5)
The turbulent shear stress term can be expressed ac-
cording to the Prandtl mixing length theory as [22,23]

t
= l
2

u
y

u
y
(6)
where l is the Prandtl mixing length. The Prandtl mix-
ing length can be expressed using the van Driest hy-
pothesis in the following way [23]:
l = ky

1 exp

y
A

(7)
where A is the damping length constant, dened as
26(
w
/)
1/2
. The modied value of A for suction
is given as A = 26(
w
/2)
1/2
exp(5.9v

w
) [23], with
k as Von Karmans constant (equal to 0.4) [23]. It
may be noted that the entire domain of the turbulent
hydrodynamic boundary layer is represented by the
expression of the mixing length given by Eq. (7).
On substitution of Eq. (7) into Eq. (6), the following
expression for the turbulent shear stress is obtained:

t
=

ky

1 exp

y
A

2

du
dy

2
(8)
Using the above expression of
t
, Eq. (5) may be writ-
ten as
=
l

t
=
du
dy

ky

1 exp

y
A

du
dy

2
=
w
uv
w
(9)
Simplifying the above expression, the velocity gradi-
ent may be expressed as
du
dy
=

2
4(
w
uv
w
)B
2B
(10)
where
B =

ky

1 exp

y
A

2
206 K. Sreenivas et al. / Journal of Membrane Science 201 (2002) 203212
The wall shear stress in turbulent ow can be ex-
pressed in terms of the time smoothed average veloc-
ity (u
0
) [24],

w
= 0.03325 u
2
o

2
d
e
u
o

0.25
(11)
Using the above expression for wall shear stress,
Eq. (10) can be expressed in an implicit form as
du
dy
= E(v
w
, y) (12)
The boundary condition for Eq. (12) is u = 0 at y = 0.
1.2. Mass transfer boundary layer
The steady state mass balance near the membrane
surface (within the concentration boundary layer) is
described by the governing species conservation equa-
tion:
u
c
x

c
y
= D

2
c
y
2



J
y
y
(13)
where

J
y
is a term representing turbulent mass trans-
port and is expressed using Prandtl analogy as [21]

J
y
= l
2

du
dy

c
y
(14)
Assuming = l
2
[du/dy[, Eq. (13) is modied as
u
c
x

v

y

c
y
= (D )

2
c
y
2
(15)
The boundary conditions for Eq. (15) are
c(0, y) = c
0
(16)
c(x, ) = c
0
(17)
The solute mass balance at the membrane surface, i.e.
at y = 0 gives the third boundary condition:
v(c c
p
) D

c
y

= 0 (18)
Since, the mass transfer boundary layer is very thin
compared to the half channel height, the transverse
velocity prole may be expressed as [11,12,25]
v = v
w
An integral method is employed for solving the gov-
erning mass balance equation (Eq. (15)), by assuming
the concentration prole is of the form
c = a
1
a
2
a
3

2
where =
y

The coefcients a
1
, a
2
, a
3
are evaluated using the
boundary conditions on the mass transfer governing
Eqs. (16) and (17) along with, the condition at the
edge of the boundary layer, (c/y)
y=
= 0. Evalu-
ating the constants, the concentration prole becomes
c
c
0
=
c
m
c
0

1
y

2
y

(19)
Substituting the above concentration prole into
Eq. (15), the following equation is obtained:
f
d
dx
g
dc
m
dx
= k
1
(20)
where
f = u

c
o
2
y

2c
m

1
y

g =

1
y

2
k
1
=2(D )
(c
m
c
0
)

v
w


y

2c
m

1
y

2
y

y
= l
2
d
2
u
dy
2
2l
du
dy

ky l
A

l
y

The expression for (d

2
u/dy
2
) can be obtained from
Eq. (10) (given in Appendix A).
1.3. Osmotic pressure model
Solvent ow through the membrane is quantied by
Darcys law for ow through a porous medium.
v
w
= L
p
(P ) (21)
Osmotic pressure of any macromolecular solution can
be expressed in the form of a viral expansion:
= B
1
c B
2
c
2
B
3
c
3
(22)
K. Sreenivas et al. / Journal of Membrane Science 201 (2002) 203212 207
Osmotic pressure difference can be estimated if the
concentration of the solute at the membrane surface
(c
m
) and the permeate concentration (c
p
) can be eval-
uated. These two concentrations are related by in-
trinsic retention (R
r
) which is constant for a specic
membranesolute system [26] dened as
R
r
= 1
c
p
c
m
(23)
The quality of permeate stream is characterized by the
observed retention (R
o
)
R
o
= 1
c
p
c
o
(24)
Using Eqs. (22) and (23), the osmotic pressure differ-
ence across the membrane can be expressed in terms
of the membrane surface concentration as
=B
1
c
m
R
r
B
2
c
2
m
(1 (1 R
r
)
2
)
B
3
c
3
m
(1 (1 R
r
)
3
) (25)
On substituting Eq. (25) into Eq. (21), the permeate
ux can be obtained as a function of the membrane
surface concentration:
v
w
=L
p
[P B
1
c
m
R
r
B
2
c
2
m
(1 (1 R
r
)
2
)
B
3
c
3
m
(1 (1 R
r
)
3
)] (26)
Combining Eqs. (18) and (19) the following expres-
sion for ux can be obtained:
v
w
=
2D(c
m
c
o
)
c
m
R
r
(27)
Combining Eqs. (26) and (27), we get an algebraic
equation containing surface concentration and concen-
tration boundary layer thickness, which can be given
as
(c

m
1)
2(/d
e
)
= [
1
c

m

2
c

2
m

3
c

3
m

4
c

4
m
] (28)
where c

m
= c
m
/c
o
and
1
,
2
,
3
,
4
are given in
Appendix A. Differentiating the above expression with
respect to x gives
dc
m
dx
=
2(c
m
c
o
)
F
d
dx
= f
1
d
dx
(29)
where F is a polynomial given in the Appendix A.
Substituting Eq. (29) in Eq. (20), results in a nonlin-
ear rst-order differential equation for the mass trans-
fer boundary layer:
(f gf
1
)
d
dx
= k
1
(30)
Integrating the above equation over the concentration
boundary layer thickness, the following equation is
obtained:

0
(f gf
1
)
d
dx
dy =

0
k
1
dy (31)
P
d
dx
= T (32)
where P and T are the functions of x, v
w
, and c
m
:
P =

0
(f gf
1
) dy (33)
T =

0
k
1
dy (34)
Eq. (32) has to be solved along with the Eqs. (12) and
(28) to obtain the prole of the concentration boundary
layer and in turn the axial variation of c
m
and v
w
.
The average permeate ux is obtained using
!v
w
) =
1
L

L
0
v
w
dx

(35)
Once the membrane surface concentration prole is
determined by the above method, the prole of ob-
served retention is obtained.
2. Numerical solution
The coupled differential and algebraic equations
(Eqs. (12), (28), (29) and (32)) are solved using a
fourth-order RungeKutta method. The governing hy-
drodynamic equation (Eq. (12)) becomes indetermi-
nate at y = 0. This problem is circumvented by the
following approach. As the velocity prole very close
to the wall is linear in nature and is given by u

= y

,
a very small value of y (10
9
m) and the correspond-
ing value of u (from the viscous sub-layer relation) are
chosen as the initial condition for the solution of the
velocity prole (Eq. (12)). The lattice spacing in the
transverse direction is taken as 10
8
m. No apprecia-
ble changes in the calculated values of the dependent
208 K. Sreenivas et al. / Journal of Membrane Science 201 (2002) 203212
variables are noticed by changing the y spacing and
the starting value of y.
The mass transfer governing equation (Eq. (32))
has been coupled with the differential Eq. (12) and is
solved using the fourth order RungeKutta method.
The coefcients P and T (Eq. (32)) are evaluated nu-
merically at each step of integration using Simpsons
one-thirds rule. It may be noted that the algebraic equa-
tion (Eq. (28)) is coupled with Eqs. (12) and (32) and
is solved using NewtonRaphsons method.
The initial condition for the solution of Eq. (32) is
= 0 at x = 0. This will make the solution inde-
terminate. Therefore, an asymptotic solution is sought
to the limit x 0, 0. Under these conditions,
the developing concentration boundary layer (near the
entrance) lies well within the viscous sub-layer. The
form of the asymptotic solution of Eq. (32) will be as
follows [11]:
d
dx
=
8h
3
Z
2
(36)
where
Z =
u
2

d
3
e
64Dv
(37)
The solution of Eq. (36) is
= h

24x
Z

1/3
(38)
The axial starting point for integration is taken as x =
10
10
m. Any further decrease in the initial point did
not result in any appreciable increase in accuracy. The
axial lattice spacing is reduced until the solution was
effectively independent of spacing. Variable step size
is employed with an initial axial spacing of 10
12
m
and it has been increased up to 10
4
m at the end of
the channel.
3. Results and discussion
3.1. Comparison of model predictions with the
experimental data
The model Eqs. (12), (28), (29) and (32) are solved
as outlined in the numerical solution section. The
solution provides the proles of the permeate ux
and the membrane surface concentration along the
Fig. 2. Comparison with the experimental data for NaCl [19]. The
predictions are the solid lines whereas the experimental data are
presented by the symbols.
channel length. The length averaged permeate ux is
obtained using Simpsons one-thirds rule. The exper-
imental data for two different solutessodium chlo-
ride (high diffusivity, order of magnitude 10
9
m
2
/s,
[19]) and dextran-70,000 (low diffusivity, order of
magnitude 10
11
m
2
/s [20]) are taken for compari-
son. The comparison of sodium chloride data under
turbulent ow conditions with the predicted values
is presented in Fig. 2. The predictions are the solid
lines whereas the experimental data are presented by
the symbols. The experiments were conducted [19]
in a cell of dimensions 75 mm 25 mm 2.5 mm.
The feed concentration was 4.2 wt.% and the op-
erating pressure was 103 atm. Cross-ow velocities
were varied from 0.08 to 0.5 m/s. Eastman E-398-3
cellulose acetate membrane was used with a mem-
brane permeability of 7.6 10
6
g/(cm
2
s atm). The
comparison for dextran-70,000 is presented in Fig. 3.
The ultra-ltration experiments [20] were carried out
in a thin channel of dimensions 100 mm 60 mm
5.9 mm. Kalle polysulphone membranes were used
with a permeability of 1.44 10
10
m
3
/(Ns). The
feed concentrations were varied from 0.43, 0.935
and 1.42 kg/m
3
; the operating pressure differences
were 200, 400 and 600 kPa; cross-ow velocities
were 1.06, 1.84, 2.75 m/s. It is clear from these two
gures that the predicted results are in good agree-
K. Sreenivas et al. / Journal of Membrane Science 201 (2002) 203212 209
Fig. 3. Comparison with the experimental ux (10
5
m
3
/(m
2
s))
data for dextran [20].
ment with the experimental data. This indicates that
the model presented herein is suitable for solutes of
widely varying diffusivities. The major advantage
of this model is that it can be successfully used,
with the assumptions inherent in the development
of the model (as described in the Section 1.1), for
high diffusivity solutes for which the concentration
boundary layer will denitely cross the viscous sub-
layer.
Fig. 4 represents the dimensionless concentration
boundary layer development along the channel length
for different solutes, namely, sodium chloride, sucrose
and dextran. The solutes are chosen in the decreas-
ing order of diffusivity (orders of magnitudes of 10
9
,
10
10
, 10
11
m
2
/s, respectively). Physical properties
of these solutes in aqueous solution are presented in
the Appendix B. As is evident from the gure, that
for the high diffusivity solutes (NaCl and sucrose,
curves 1 and 2), the concentration boundary layer de-
velops and goes beyond the viscous sub-layer into
the buffer region of the hydrodynamic boundary layer.
Only for dextran (curve 3), which is having the small-
est diffusivity (the highest Schmidt number) of the
three solutes, the concentration boundary layer lies
well within the viscous sub-layer. Thus, the present
model is more generalized compared to the existing
fundamental models of turbulent cross-ow ltration
[11,12].
Fig. 4. Variation of concentration boundary layer thickness along
the channel length. c
o
= 1 kg/m
3
, R
r
= 1, L = 1000 mm,
P = 800 kPa, Re = 4600. Curve: 1, NaCl; 2, sucrose; 3, dextran.
The axial velocity prole after incorporation of suc-
tion is expressed in Eq. (12). Fig. 5 represents a plot
of the x-component of the velocity prole as a func-
tion of the distance from the wall when the average
velocity in the channel is u
0
= 1 m/s. Curve 1 rep-
resents the velocity prole when there is no suction
(i.e. v
w
= 0). The effect of suction (v
w
= 0.001 m/s)
on the velocity prole is presented in curve 2. It can
be clearly observed from the gure that the velocity
decreases across the channel due to suction.
Fig. 5. Variation of x-component of velocity with suction. Curve:
1, v
w
= 0.0 m/s; 2, v
w
= 0.001 m/s.
210 K. Sreenivas et al. / Journal of Membrane Science 201 (2002) 203212
Fig. 6. Variation of membrane surface concentration along the
channel length for NaCl solution. c
o
= 42 kg/m
3
, P = 103 atm.
Curve: 1, Re = 4600; 2, Re = 9200; 3, Re = 13,800.
Since the major thrust of this work is to develop a
model that would be valid even for high diffusivity so-
lutes, a parametric study is undertaken with NaCl as
the solute. The aim of the study is to quantify the ef-
fects of various operating conditions (e.g. transmem-
brane pressure, R
r
, u
0
) on the permeate ux and the
observed retention.
Fig. 6 represents the variation of dimensionless
membrane surface concentration along the length of
the channel for various cross-ow velocities. Curve 1
is for the lowest and curve 3 is for the highest
cross-ow velocities. It is clear from the gure that in-
creased turbulence (high cross-ow velocity) impedes
the accumulation of the solutes on the membrane sur-
face. This would result in a lower value of membrane
surface concentration.
The relative decrease in membrane surface concen-
tration results in a reduction in the build up of osmotic
pressure, thereby, increasing the effective driving force
across the membrane. The consequent increases in the
permeate ux are presented in Fig. 7, where curves
13 are arranged in increasing order of cross-ow ve-
locity.
Fig. 8 presents the variation of quality in the form
of observed retention of the permeate along the chan-
nel for different cross-ow velocities. Curve 3 is for
the highest and curve 1 is for the lowest cross-ow ve-
locity. It can be observed that the quality deteriorates
Fig. 7. Variation of permeate ux along the channel length for
NaCl solution. Operating conditions are same as in Fig. 6.
along the channel due to increase in the membrane sur-
face concentration (Fig. 6). As explained earlier, an in-
crease in cross-ow velocity decreases the membrane
surface concentration, thereby decreasing the chem-
ical potential difference across the membrane. This
would result in less solute transport through the mem-
brane (lower permeate concentration) and hence, the
observed retention will increase.
Fig. 8. Effect of cross-ow velocity on the prole of observed
retention along the channel length. Operating conditions are same
as in Fig. 6.
K. Sreenivas et al. / Journal of Membrane Science 201 (2002) 203212 211
Fig. 9. Effect of intrinsic retention on the prole of observed
retention along the channel length. c
o
= 42 kg/m
3
, P = 103 atm;
Re = 4600. Curve: 1, R
r
= 0.98; 2, R
r
= 0.95; 3, R
r
= 0.90.
Fig. 9 represents the effects of R
r
on the observed
retention. For a xed R
r
, R
o
decreases along the chan-
nel. This decrease appears less than the previous g-
ure (Fig. 8) due to the length scale for R
o
chosen here
to accommodate larger variation in R
o
. The axial vari-
ation (reduction) in R
o
is consistent with the relative
increases in C
m
(Fig. 6) for a xed value of R
r
. At a
xed location in the channel, R
o
is lower for smaller
values of R
r
. Low values of R
r
indicate that the mem-
brane is more porous and therefore, the solute concen-
tration in the permeate stream tends to increase. This
leads to a decrease in the values of R
o
with smaller R
r
.
4. Conclusions
A transport model for the osmotic pressure con-
trolled turbulent cross-ow membrane ltration pro-
cess is developed. The limitations of the previous
models, specially, for high diffusivity solutes have
been overcome in this work. Thus, the model is ap-
plicable for most of the solutes typically encountered
in membrane separation processes (from RO to UF).
An integral method is employed to solve the coupled
governing hydrodynamic and mass transfer equa-
tions. The solution simultaneously predicts the per-
meate ux and permeate concentration as a function
of the operating conditions. The results from the
model are successfully compared with the data avail-
able from literature for sodium chloride and dextran
under turbulent ow conditions.
Appendix A

1
= R
r
(A.1)

2
=
R
2
r
c
0
B
1
(P)
(A.2)

3
=
R
r
c
2
o
B
2
(P)
[1 (1 R
r
)
2
] (A.3)

4
=
R
r
c
3
o
B
3
(P)
[1 (1 R
r
)
3
] (A.4)
where
=
L
p
d
e
(P)
4D
(A.5)
F =2
1

2
1

(c

m
) 3
1

(c

m
)
2
4
1

(c

m
)
3
(A.6)
where
c

m
=
c
m
c
o
and

=

d
e
d
2
u
dy
2
=

2B
2
dB
dy

v
w
[B (dB/dy)]
MB

M
2B
2
dB
dy
(A.7)
where
M =

2
4(
w
uv
w
)B
B =

ky

1 exp

y
A

2
dB
dy
=

k
2
y

1 exp

y
A

2y exp(y/A)
A

1 exp

y
A

where
A = 26

1/2
exp(5.9v

w
)
212 K. Sreenivas et al. / Journal of Membrane Science 201 (2002) 203212
Appendix B. Physical properties of the solutes
(i) Diffusivity [19,20,27]: NaCl, 1.5 10
9
m
2
/s;
sucrose, 5.24 10
10
m
2
/s; dextran, 4
10
11
m
2
/s.
(ii) Viscosity: Viscosity of the solutions have been
taken as 0.001 Pa s.
(iii) Density: Density of the solutions have been taken
as 1 g/cm
3
.
(iv) Osmotic pressure: Osmotic pressure of macro-
molecular solutions can be expressed in a viral
form:
= B
1
C B
2
C
2
B
3
C
3
where is in Pa and C is in kg/m
3
.
For dextran T70 [20]: B
1
= 37.5; B
2
= 0.752;
B
3
= 0.00764.
For NaCl [19]: B
1
= 83086.5; B
2
= 0.0; B
3
= 0.0.
For sucrose [27]: B
1
= 3770; B
2
= 38.79; B
3
=
0.04.
References
[1] W.F. Blatt, A. Dravid, A.S. Michaels, L. Nelson, Solute
polarization and cake formation in membrane ultra-ltration:
causes, consequences and control techniques, in: J.E. Flinn
(Ed.), Membrane Science and Technology, Plenum Press, New
York, 1970.
[2] S. Srinivasan, C. Tien, W.N. Gill, simultaneous development
of velocity and concentration proles in reverse osmosis
systems, Chem. Eng. Sci. 22 (1967) 417.
[3] J.S. Shen, R.F. Probstein, On the prediction of limiting ux in
laminar ultra-ltration of macromolecular solutes, Ind. Eng.
Chem. Fundam. 16 (1977) 459.
[4] D.R. Trettin, M.R. Doshi, Ultra-ltration in an unstirred batch
cell, Ind. Eng. Chem. Fundam. 19 (1980) 189.
[5] J.G. Wijmans, S. Nakao, C.A. Smolders, Flux limitations in
ultra-ltration: osmotic pressure model and gel layer model,
J. Membr. Sci. 20 (1984) 115.
[6] R. Rautenbach, R. Albrecht, Membrane Processes, Academic
Press, New York, 1986.
[7] T.K. Sherwood, P.L.T. Brian, Salt concentration at phase
boundaries in desalination by reverse osmosis, Ind. Eng.
Chem. Fundam. 4 (1965) 113.
[8] L. Dresner, Boundary layer build-up in the demineralization of
salt water by reverse osmosis, Oak Ridge National Laboratory
Report, 1964, p. 3621.
[9] D.R. Trettin, M.R. Doshi, Pressure independent ultra-ltration
is it gel limited or osmotic pressure limited? in: A.F. Turbak
(Ed.), Synthetic Membranes. Part II. Hyper and Ultra-ltration
Uses, ACS Symposium Series, 1981, p. 154.
[10] P.L.T. Brian, Concentration polarization in reverse osmosis
desalination with variable ux and incomplete salt rejection,
Ind. Eng. Chem. Fundam. 4 (1965) 439.
[11] R. Prabhakar, S. DasGupta, S. De, Simultaneous prediction
of ux and retention for osmotic pressure controlled turbulent
cross-ow ultra-ltration, Sep. Purif. Technol. 18 (2000) 13.
[12] V.S. Minnikanti, S. DasGupta, S. De, Prediction of mass
transfer with suction for turbulent ow in cross-ow
ultra-ltration, J. Membr. Sci. 4079 (1999) 1.
[13] G. Belfort, C. Kleinstreuer, Mathematical modeling of uid
ow and solute distributions in pressure driven membrane
modules, in: G. Belfort (Ed.), Synthetic Membrane Processes:
Fundamentals and Water Applications, Academic Press, New
York, 1984.
[14] S. Ganguly, P.K. Bhattacharya, Development of concentration
prole and prediction of ux for ultra-ltration in a radial
cross-ow cell, J. Membr. Sci. 97 (1994) 185.
[15] L. Huang, M.T. Morrisey, Finite element analysis as a tool
for cross-ow membrane lter simulation, J. Membr. Sci. 155
(1999) 19.
[16] A.P. Silva-freire, An asymptotic solution for transpired
incompressible turbulent boundary layers, Int. J. Heat Mass
Transfer 31 (5) (1988) 1011.
[17] J. Sucec, M. Oljaca, Calculation of turbulent boundary layers
with transpiration and pressure gradient effects, Int. J. Heat
Mass Transfer 38 (15) (1995) 2855.
[18] V. Gekas, B. Hallstrom, Mass transfer in the membrane
concentration polarization layer under turbulent cross-ow.
Part I. Critical literature review and adaptation of existing
Sherwood correlations to membrane operations, J. Membr.
Sci. 80 (1987) 153.
[19] U. Merten, H.K. Lonsdale, R.L. Riley, Boundary layer effects
in reverse osmosis, Ind. Eng. Chem. Fundam. 3 (1964)
210.
[20] J.G. Wijmans, S. Nakao, J.W.A. van den Berg, F.R. Trolestra,
C.A. Smolders, Hydrodynamic resistance of concentration
polarization boundary layers in ultra-ltration, J. Membr. Sci.
22 (1985) 117.
[21] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport
Phenomena, Wiley, New York, 1960.
[22] P. Bradshaw, T. Cebeci, Momentum Transfer in Boundary
Layers, Hemisphere, Washington, DC, 1977.
[23] T. Cebeci, A.M.O. Smith, Analysis of Turbulent Boundary
Layers, Academic Press, New York, 1975.
[24] R.W. Fox, A.T. McDonald, Introduction to Fluid Mechanics,
Wiley, New York, 1985.
[25] S. De, S. Bhattacharya, P.K. Bhattacharya, A. Sharma,
Generalized integral and similarity solutions for concentration
proles for ultra-ltration, J. Membr. Sci. 130 (1997) 99.
[26] W.S. Opong, A.L. Zydney, Diffusive and convective protein
transport through asymmetric membranes, AIChE J. 37 (1991)
1497.
[27] S. Sourirajan, Reverse Osmosis, Academic Press, New York,
1970.