of

BRANE
SCIENCE

Mrnal
ELSEVIER

Journal of Membrane Science 128 (1997) 119-131

Prediction of mass-transfer coefficient with suction in the

applications of reverse osmosis and ultrafiltration
S. D e , P.K. B h a t t a c h a r y a *
Department of Chemical Engineering, Indian Institute of Technology at Kanpur, Kanpur- 208016, India
Received 22 April 1996; received in revised form 5 August 1996; accepted 11 October 1996

Abstract
Sherwood-number relations for prediction of the mass-transfer coefficient for developing concentration boundary-layer
have been obtained for laminar flow-regime from first principles. The common flow-modules, namely, rectangular channel,
tubular and radial cross-flow are considered. The relationships developed include the effect of suction through the membrane.
Relevant relations for estimation of mass-transfer coefficient for cross-flow reverse osmosis and ultrafiltration are formulated.
The Sherwood-number relations developed are compared with the standard correlations to quantify the effect of the suction.
The proposed Sherwood relations are used in conjunction with the osmotic-pressure model to predict the permeate flux in
reverse osmosis and osmotic-pressure governed ultrafiltration.

Keywords: Mass-transfer coefficient; Suction; Laminar flow; Osmotic pressure: Cross flow; Reverse osmosis; Ultrafiltration

1. Introduction
The design of pressure driven membrane-separation
processes, like reverse osmosis (RO) and ultrafiltration
(UF), are generally based on the mass-transfer coefficient (k) for the relevant flow-configuration and flowregime. The mass-transfer coefficients used for such
purposes are usually derived from the correlations
obtained from heat-mass-transfer analogies. The maj or
drawbacks of the use of such Sherwood-number correlations with regard to RO/UF are: (a) they are derived for flow through a non-porous conduit; hence,
the effect of suction cannot be considered; (b) changes
in properties like viscosity and density due to concentration polarization near the membrane surface cannot
*Corresponding author,

0376-7388/97/$17.00
1997 Elsevier Science B.V. All rights reserved.
PII S0376-7388(96)00313-4

be taken into consideration; (c) it is tacitly assumed

that the concentration boundary-layer is fully developed over most of the channel length, which may not
be the case for RO/UF; (d) the osmotic pressure, built
up near the membrane surface, cannot be considered
in such approaches; and, finally, (e) the mass-transfer
coefficient is assumed to be independent of pressure,
which may not be valid for RO/UF operations.
Therefore, the use of standard correlations lead to an
inaccurate estimation of the mass-transfer coefficient
and hence, an incorrect prediction of the permeate
flux. One way to avoid this is to perform a detailed
simulation, solving relevant momentum and solute
mass-balance equations with pertinent boundary conditions [1-4]. But such methods may not be very attractive from a designer's point of view owing to their
extensive computational effort and complications.

120

S. De. PK. Bhattacharya/Journal of Membrane Science 128 (1997) 119-131

The mass-transfer correlations for membraneseparation processes were reviewed in detail [5,6].
It was concluded, in both of the reviews, that the
present correlations need to be modified in light of
their limitations discussed earlier. In fact, it was
suggested that mass-transfer correlations should be
developed, based on experimental techniques, namely,
the velocity-variation technique or osmotic-pressure
model [5-7]. Each technique has its own disadvantages which were discussed in detail [5,6].
Another alternative method, which was quite successfully employed in our earlier work [8-11] on UE
includes the development of a second correlation for
concentration polarization in terms of polarized layerresistance along with a standard mass-transfer correlation like the Leveque correlation for laminar flow in
a channel or Colton's correlation in a stirred cell [3].
But such approaches are solute and system specific. In
addition, it is difficult to work with two correlations
simultaneously.
The role of suction in mass transfer through porous
membranes is very important. It has been identified
earlier [5,6] that the effect of suction on mass-transfer
coefficient is two-fold. First, it enhances the mass
transfer from the surface to the bulk; and, second, it
stabilizes the laminar-flow condition in the conduit by
delaying the laminar-to-turbulent transition (typically,
critical Reynolds number is shifted from 2100 to 4000
in the presence of suction [6]).
Therefore, it seemed possible that a generalized
mass-transfer relation may be obtained theoretically
for laminar flow from first principles. The present
work aims to develop a generalized mass-transfer
relation, including the effects of suction over a developing concentration boundary-layer. Such relations
can be coupled with the osmotic-pressure model to
predict permeate flux for osmotic-pressure governed
UF and also for RO. Further, the theoretical work is
extended to include all the flow modules, usually
encountered in membrane-separation processes,
namely, rectangular channel, tubular and radial
cross-flow configurations.

the basic approach of solving simultaneously, the

governing solute-mass and momentum-balance equations along with the boundary conditions.
The flow configuration in a closed conduit is shown
in Fig. la. Fig. lb depicts the flow geometry of a
radial cross-flow cell. The fluid is allowed to flow
tangentially over the membrane surface. The upper
boundary of the channel is impervious for the rectangular and radial cells. The permeate flux is a function
of the channel length for the rectangular and tubular
modules; and for a radial module, it is a function of the
radius of the cell. The concentration boundary-layer
develops over the effective length of the membrane.
Assumptions made in this model are: (a) the flow is
steady; (b) the diffusion along the membrane is negligible, compared to the convection in the same direction; (c) the flow is laminar and fully developed; (d)

I
.

I
.

. . . . . .

Vw (x)

Non-permeating section

Top Disk

Feed

y=2h
2h
Membrane

B o t t o m Disk

Permeate

2. Theory
In this section, an attempt has been made to develop
a generalized mass-transfer-coefficient relation from

I
h

x=0, y=0

Fig. 1. Schematic diagram of (a) the flow configuration in a

rectangular conduit, and (b) the flow configuration in a radial
cross-flow cell.

S. De. P.K. Bhattacharya/Journal of Membrane Science 128 (1997) 119-131

the permeate velocity is small enough, compared to

the feed velocity, keeping the parabolic velocity profile in the channel undistorted for rectangular and
tubular modules; (e) the concentration at the membrane surface is constant; and (f) the physical properties of the solution are constant.
With the above-mentioned assumptions, the solute
mass-balance equation over a differential element in
the conduit gives

V V c = V(DVc)

(1)

where the velocity vector, V, is the resultant of the

axial velocity, u, and the transverse-velocity component, v. Assuming no permanent (irreversible) adsorption of the solute on the membrane surface, which may
be possible with an efficient cross-flow system and a
non-adsorbing membrane,
v ~ -Vw

(2)

within the thin concentration-boundary-layer.

The boundary conditions of Eq. (1) are as follows:
for rectangular and tubular modules,
c=c0 atx=0

where

Rr

(8)

From this point onwards, estimation of the masstransfer coefficient for different flow modules is presented separately.

2.1. Flow through a rectangular cell

The solute mass-balance equation, for flow through
a rectangular channel under the assumptions stated
earlier, may be written in the following form.

Oc
Oc
02c
u Ox - Vw~y -= D Or~.

(9)

r/= y

(5)

(6)

c*(rl) = c/co

is the intrinsic rejection of the solute by the

(11)

In terms of c ~ and ~, Eq. (9) becomes an ordinary

differential equation of second order, as follows:
(12)
The axial velocity profile may be taken as
u=~u0

1-

(13)

Within the thin concentration-boundary-layer, when y

is small compared to the half-channel height, Eq. (13)
can be approximated as (by neglecting y2/h:)
u --

(7)

(10)

and dimensionless concentration can be expressed by

or,

Oc
VwCmRr + D ~ y = 0 at y = 0

~c

(4)

At the membrane surface, the concentration is greater

than the bulk; back diffusion occurs from the surface
to the bulk solution due to the concentration gradient.
This is opposite to the convective movement of the
solute particles towards the membrane. At steady
state, the net result of these two opposing fluxes is
equal to the convective flux of the permeating solution.
Therefore, one has the following boundary condition
at the membrane surface:

Oc
VwCm + D-~y = VwCp at y = 0

c=c0aty=

A similarity solution for Eq. (9) is obtained by defining a dimensionless variable (lumped parameter),

The assumption of constant solute concentration at the

membrane surface results in the following boundary
condition:
C=Cm a t y = 0

membrane, which may be assumed to be constant for a

membrane-solute system [2,12].
Finally, in the bulk, the solute concentration is
constant beyond the concentration boundary-layer,
the thickness of which is negligible compared to the
half-channel height. Hence, a common way to represent the boundary condition for the bulk solution is
[131

(3)

and, for the radial cross-flow cell,

c = c 0 at r = 0

121

3uoy
h

(14)

For constant physical properties and channel geometry, VwX1/3

=constant. Thus, a non-dimensional form of

122

S. De. PK. Bhattacharya/Journal of Membrane Science 128 (1997) 119-131

can be rewritten as

this can be defined as:

Vw

= A1

(15)

O.42Alrlld~

dc*
(r?2 + A 1 ) dr/

(16)

k(cm - CO) = -D(O~yy)

and the transformed boundary conditions become
c* = 1 a t ~ / = ~

(17)

and
dc*

d---~+A1Rrc* =

(26)

Now, the mass-transfer coefficient, k, is defined and

obtained from a solute mass-balance at the membrane
surface, as

Therefore, Eq. (12) becomes

d2c *
d~72 -

Ii = f o ~ e X p [ - ~ -

0 at ~7 = 0

(18)

The solution of Eq. (16), along with the boundary

conditions (Eqs. (17) and (18)), can be written as

In terms of non-dimensional forms of c and y, the

above equation can be represented by:

k(c m -

(19)

1) = - D

( U0 ) 1 / 3 ( d c * ~
hxD
\d~/n= 0

(28)

Substituting values of c m and dc*/d~ at 77=0 from

Eq. (19), we have

k(K2 c*07)=Klfo~lexp(-~-A97)d~+K2

(27)

y=0

_ / u0 ,~1/3
l) = - t g ~ , ~ )
K1

(29)

or

where
:

A1Rr

K1 -/(2 --

(20)

1 -- A1RrI1
1

(21)

1 - A 1RrI1

(30)

Expressing the mass-transfer coefficient, k, in terms of

the Sherwood number (kdffD) as a function of dimensionless channel length, x*(x/L), we can write

41/3
Sh(x*) = ~
(Re Sc de/L)U3(x*) -1/3

and,
11 = f0 e x p ( - ~T]3 - Alr/)dr/

(22)

and the average Sherwood number over length, L, can

be expressed as

The average flux over the membrane length can be

obtained from Eq. (15) as:

S-h =

Vw : -1fo L Vw(X)dx :

where I1 is given by Eq. (26).

f uoD2) l/3 a

1.5 ~ , ~ - - )

(23)

Now, for a rectangular channel, the equivalent hydraulic diameter can be defined by:
de = 4h

(24)

Therefore, A1 can be expressed from Eqs. (23) and

(24), in terms of Pew = Vwde/D, as
A1 = 0.42A1
where

(25)

Al=Pew/(Re Sc de~L)1/3, Re=puodff# and

Hence, the integral, I1, given by Eq. (22)

Sc=#/pD.

(31)

/0'

Sh(x*)dx* = 2.381 ( R e S c d e / L ) l / 3
Ii

(32)

2.2. Flow through a tubular module

For steady flow through a tubular module under the
previously mentioned assumptions, the solute massbalance equation can be written as:

Oc
oc
O (rOC
UOx- Vwor -- r Or \ Or/

(33)

Considering a thin concentration-boundary-layer

adjacent to the wall, the curvature effects may be

123

S. De. P.K. Bhattacharya/Journal of Membrane Science 128 (1997) 119-131

neglected and the problem may be treated as though

the wall were flat. If the distance from the wall is
denoted as y=R-r, the fluid may be regarded as being
confined between a flat mass-transfer surface extending from y = 0 to y=cx~. Therefore, the solute massbalance equation (Eq. (33)) can again be expressed by
Eq. (9).
The fully developed velocity profile in a tube can be
described by
u=2uo

1-

(34)

In the concentration boundary-layer near the wall, the

above expression for the velocity profile may be
expressed as:
u=2u0

[ ( )21
1-

(35)

Neglecting higher order terms (yZ/R2) for the thin

concentration-boundary-layer, the axial-velocity profile may be approximated as:
V
u = 4u0 R

(36)

profile in the tubular module, as follows:

c* (6) = K3

fo '~exp (

84'3
A~O)d6+K4
9
"'

(41)

where
K3-

(42)

A2Rr

1 - A2Rrl2
1

K 4 - 1 - A2Rrl2

(43)

and
12 =

exp

~A?0 dO

(44)

Proceeding exactly as for rectangular channel, the

Sherwood-number profile as a function of the module
length can be obtained from
Sh(x*) = 1 (Re Sc d/L)

1/3(x*)-1/3

(45)

and the average Sherwood number over the module

length can be expressed as
= __1"5(ReScd/L)l/3

(46)

The similarity parameter chosen in this case is:

( U0 ,~1/3
4' = Y\ x ~ /

(37)

In terms of c* and 4', the solute mass-balance equation

(Eq. (33)) can be expressed in the following form:
d2c *
(~
)dr*
d02 _ _
4'2 + A24' dr/

(38)

where

( xD ~1/3
A2 = Vw\uod---Sj

(39)

The average flux over the length of the module can be

defined by Eq. (23), Expressing the average flux in
terms of Pew(Vwd/D), A2 can be written as:
9
A2 = ~,~2
(40)
where A2=Pew/(Re Sc d/L) ~/3. The initial and boundary conditions in terms of c* and 4' remain the same, as
Eqs. (3), (5), (7) and (8). The solution of Eq. (38) with
the boundary conditions provides the concentration

2.3. Flow through a radial cross-flow cell

The geometry of a radial cross-flow cell is best
represented by an axisymmetric cylindrical-coordinate system, where r is the radial and y is the transverse direction (Fig. lb). However, for convenience,
we assume a two-dimensional cartesian-coordinate
system, where the radial direction, r, is not a radial
coordinate but a cartesian axis. Such simplifications
make the model equations simpler and do not alter the
results to any significant extent. Therefore, the steadystate solute mass-balance equation within the concentration boundary-layer can be written in the following
form:
OC

OC

02C

U~rr - Vw~y = D -Oy

- 2

(47)

The radial-velocity profile within the boundary layer

can be expressed [4] as:

3Qy
u(r, y) - 47rrh2

(48)

S. De. P.K. Bhattacharya/Journal of Membrane Science 128 (1997) 119-131

124

where Q is the average volumetric flow-rate, given by

Q=47rRhuo, and Uo is the average velocity in the
conduit.
The similarity parameter in this case may be chosen

of channel radius can be obtained as:

as~

and the average Sherwood number over the channel

radius can be expressed by

) 1/3

= y ~

41/3
Sh(r*) = ~ (Re Sc h/R)l/3(r*) -2/3

(58)

(49)

In terms of c* and ~, Eq. (47) can be written in the

following form:
d2c*
(~)dc*
d~2--+A3~

d~

(50)

where
V w ( 7 ( h ; D ) 1/3
r2/3
A3 = ~- - -

(51)

The average flux over the membrane radius can be

defined as

f0"

Vw = - ~

Vw(r)rdr

(52)

Expressing the average flux in terms of Pew(Vwh/D),

A3 c a n be written as:
A3 = 0.42~3

(53)

where )~3 = Pew/(Re Sc h[R) 1/3, Re=puoh/# and Sc=#/

(pD). The boundary conditions in terms of c* and
remain the same as Eqs. (4), (5), (7) and (8). The
solution of Eq. (50) results in the following concentration profile:

~3

c*(~) = K, f ~ e x p ( - ~

-- A3~) d~ + g 6

(54)

Sh = __2'381(Re Sc h/R) 1/3

/3

(59)

In the next section, the integrals, 11-13, are examined

for different domains of suction (Pew).
2.3.1. Case 1." No Suction; Pew=O
For the rectangular cross-flow cell, the integral 11
takes the following form when there is no suction:

/'1 = JO exp(--r/3/3) dr/

= 3-2/3F(1/3)
= 1.2879

(60)

The corresponding expression for Sh is:

g~ = 1.85(Re Sc ae/L) 1/3

(61)

which is identical to the Leveque solution [5,6] for

heat transfer in a non-porous channel. Similarly, for a
tubular module, the average Sherwood number may be
obtained from Eq. (46) for Pew=0, as follows:
S---h= 1.62(Re Sc d/L)1/3

(62)

This expression for average Sherwood number is

again identical to the Leveque solution [14] for heat
transfer in a non-porous tube.
For a radial cross-flow cell, the average Sherwood
number may be obtained from Eq. (59) as:
S---h= 1.47(Re Sc h/R)1/3

(63)

where
K5 -K6 --

A3Rr
1 - A3Rr13
1

(55)
(56)

1 - A3Rr/'3

and
/3=f0

ec

~3 _ 0'42"X3~) d~

exp ( - ~-

(57)

Proceeding exactly as in the case of rectangular

channel, the Sherwood-number profile as a function

2.3.2. Case 2: RO/UF system

For a typical RO/UF system, ~1.2,3 vary from very
low values up to 10. The behaviour of the integrals 1/
I1,2,3 in this range of -~1,z,3 dictates the dependence of
the average Sherwood number on the suction (i.e.
Pew). To visualize the variation of 1/11,2,3 (given by
Eqs. (26), (44) and (57)) for this range, 1//'1,2, 3 is
evaluated for ,~1,2,3 varying in the range 0-10 by
numerical integration and is presented in Fig. 2. In
Fig. 2, the symbols represent the numerically integrated values and the lines represent the best-fit data.

S. De, P.K. Bhanacha~a/Journal of Membrane Science 128 (1997) I19-131

125

expressed by the phenomenological equation,

Vw = L p ( A p - ATr)

I
lg

where
z_~7r :

//
t/

71-m - - 71-p

(68)

i3

The osmotic pressure of the solution can be expressed

as a function of solute concentration, as:

i2/"

~4

(67)

-)

7r = C~zc+ (~2c- + ~3c 3

13"" 223"
[3"

In terms of Pe~,,, Eq. (67) may be written as

Pew = Bi (1.0 - ATr/~P)

0 ~rTwn~n~;
0
2

10

/~ 1,2,3

Fig. 2. Variationof 1Hi,2,3 with A1.2,3.Solid line is for rectangular

cell (Eq. (64)), dotted line is for tubular module (Eq. (65)) and long
dashed line is for radial cross-flow cell (Eq. (66)). Symbols
represent the numerically integrated values, obtained from Eq. (22)
(circle), Eq. (44) (box) and Eq. (57) (triangle).
From this figure, it may be observed that the variation
of 1/I1,2, 3 with AI.2, 3 is not linear. In all the cases, the
correlation coefficients are greater than 0.9999. Therefore, the average Sherwood numbers for different
flow-geometries can be represented by the following
equations:
For a rectangular cross-flow cell:

Sh : 1.85(ReScde/L) '/3 [1.0 + 0.32A, + 0.02A~

--8.05 10-4A~]

(64)

For a tubular module:

Sh = 1.62(ReScd/L) 1/3 [1.0 + 0.37A2 + 0.03A~

- 1 . 0 5 > 10-3A~]

(69)

(65)

(70)

where B~=LpAPde/D; it may be noted that for a

tubular and radial cross-flow module, de in the relation
of Bj should be replaced by d and h, respectively.
Now, the average solvent-flux through the membrane is given as:
Vw(cm

Cp) :

_o(c)
\~'J,:0

(71)

In terms of average Sherwood number and non-dimensional flux (Pew) from Eqs. (27) and (71), one can
obtain
Pe,~.

S~h(1
= Rr
- co/cm)

(72)

The expressions of Sh for different flow configurations

are presented in Eqs. (64)-(66). Therefore, a simultaneous solution of Eqs. (68), (70) and (72) provides the
predicted value of Pew and hence, permeate flux.
However, it may be noted here that apart from the
operating conditions (Ap, Uo, Co), the characteristic
retention for solute-membrane system (Rr) is required
to predict the permeate flux. Generally, this characteristic retention parameter is obtained from a separate
set of experiments.

For a radial cross-flow cell:

Sh = 1.467(Re Sc h/R)1/3 [1.0 q- 0.41A3 + 0.03A 2
- 1 . 2 5 10 3A~]

(66)

2.4. Application of Sherwood-number relations m

RO and UF for prediction of flux
In an osmotic-pressure controlled membraneseparation process, permeate flux can simply be

3. Results and discussion

In this section, several implications of the Sherwood-number relations, developed here, are examined. The typical value of the group (Re Sc dffL)
for the rectangular cell (or Re Sc d/L, for the tubular
and Re Sc h/R, for radial cross-flow cell) varies in the
range 103-106 , in membrane separation processes.
Corresponding permeate fluxes, in terms of the suction

126

s. De. P.K. Bhattacharya/Journal of Membrane Science 128 (1997) 119-131

parameter (Pew), can be in the range 1-500. Sherw o o d - n u m b e r profiles along the c h a n n e l length, for a
rectangular cell, for different operating conditions, are
shown in Fig. 3a. Similar profiles for tubular and
radial m o d u l e s are depicted in Fig. 3b and Fig. 3c,
respectively. Such profiles were generated from

600 ~1

400

Eqs. (31), (45) and (58), where the integrals Ii, /2

and 13 were evaluated numerically. In these figures, the
suction parameter, Pew, varies in the range 0 - 3 0 0 ; the
solid lines are for Re Sc de/L (or Re Sc d / L or Re Sc h~
R) =-103 and the dashed lines are for Re Sc de/L (or
Re Sc d / L or Re Sc h/R) =105. It can be observed from

500

400

300

i,\

~I~~\

r/?

20O
20O

-----2222

.....

100

2
0

0
0.0

0.2

0.4

0.6

0.8

1.0

"'llll''''l"ll'J''tlrl'll'T''lll''ll'''l'''li'i~''l''''~llll

0.0

1.2

0.2

0.4

x/L

O~
X

0.8

1.0

1.2

600

400

200

Ij I P i l l

0.0

II I~ [ I , , , , 1 1

0.2

I[11 JIIIIl~[,llla

0.4

0,6
r

II I I I '

0.8

IIt III tt [III

1.0

I 1 ~ 1 el

1.2

Fig. 3. (a) Variation of local Sherwood number along the channel length for different values of suction, for a rectangular cross-flow cell. 1:
Pew=0; 2: Pe,,=50; 3: Pew=100; 4: Pew=200; and 5: Pew=300. Solid lines are for Re Sc dJL=lO 3 and dashed lines are for Re Sc de/L=-lO5.
(b) Variation of local Sherwood number along the module length for different values of suction, for a tubular module. 1: Pew=0; 2: Pew=50; 3:
Pew=lO0; 4: Pew=200; and 5: Pew=300. Solid lines are for Re Sc d/L=lO 3 and dashed lines are for Re Scd/L=lO 5. (c) Variation of local
Sherwood number along the channel radius for different values of suction, for a radial cross-flow cell. 1: Pew=0; 2: Pew=50; 3: Pew= 100; 4:
Pew=200; and 5: Pew=300. Solid lines are for Re Sc h/R-lO 3 and dashed lines are for Re Sc h/R=lO 5.

S. De. P.K. Bhattacharya /Journal of Membrane Science 128 (1997) 119-131

the same solute and system geometry, an increase in

Re increases forced convection and, consequently, the
growth of the concentration boundary-layer is minimized so that the Shelwood number increases.
Now, for the description of a realistic mass-transfer
operation in a conduit, it is convenient to work with an
average mass-transfer coefficient and an average Sherwood-number relationship. The effects of suction on
the average Sherwood number, as estimated from

the figures that the Sherwood number decreases sharply near the entrance and then gradually for the rest of
the conduit. Local Sherwood number increases with
an increase in the suction (as Pew increases). This
leads to an increased mass transfer from the surface to
the bulk, in agreement with the qualitative description
of the effects of suction on mass-transfer coefficient by
Gekas and Hallstrom [6]. At higher Re Sc d J L , the
Sherwood number is larger for the same suction. For

127

5
q 4

100

200
300
Pe~
C

400

500

100

200
300
Pew

400

500

2
3

i00

200
300
Pew

400

500

Fig. 4. (a) Variation of ~/S'hno suction with Pe,~ for different Re Sc de~L, in a rectangular cell. 1: Re Sc de/L=103; 2: Re Sc de/L--104; 3:
Re Sc dJL=105; and 4: Re Sc dJL=lO 6. (b) Variation of Sh/Sh,o suctionwith Pew for different Re Sc dJL in a tubular module. 1: Re Sc d~
L=103: 2: Re Sc d/L=104; 3: Re Sc dlL=105; and 4: Re Sc d/L=lO 6. (c) Variation of Sh/Shno suctionwith Pe,~, for different Re Sc dJL, in a
radial cross-flow cell. 1: Re Sc h/R=103; 2: Re Sc hIR=104; 3: Re Sc h/R=105; and 4: Re Sc h/R=lO 6.

128

S. De. P.K. Bhattacharya/Journal of Membrane Science 128 (1997) 119-131

Eqs. (64)-(66), can be compared with the average

Sherwood number with no suction (the standard Leveque equation (Eqs. (61)-(63)) for the different operating conditions and flow geometries. The ratio of Sh
with suction to that without suction versus Pew, for
different values of Re Sc de/L for a rectangular cell, is
plotted in Fig. 4a. For a realistic UF system, Re Sc de/
L is of the order of 105-106. When Re Sc de/L is 105, a
Pew of the order of 500 results in about a 6-fold
increase in the average Sherwood number compared
to that given by Leveque equation, where the enhancement factor is ~3 for Re Sc de/L=lO 6. The effects of
suction on the average Sherwood number for a tubular
module is presented in Fig. 4b. In this case, the effects
of suction are even greater. When Re Sc d/L= 105, the
enhancement factor is ~7, while for Re Sc d/L= 106, it
is ,-~3.5, for Pew=500. The ratio of Sherwood number
with and without suction for a radial cross-flow cell is
presented in Fig. 4c. From the figure, it can be
observed that the enhancement ratio is 7.7 for
Re Sc h/R=lO 5 and 3.8 for Re Sc h/R~-lO 6, for
Pew=500. Hence, the effect of suction on Sherwood
number increases for rectangular, tubular and radial
modules. Therefore, it can be concluded that there
exists a significant effect of suction on Sherwood
number and, hence, on mass transfer.
The proposed Sherwood-number relations can also
be used to predict the permeate flux in RO and UF. For
RO in a rectangular channel, Eq. (64) mass-transfer
coefficient along with Eq. (72) for the osmotic pressure and Eq. (70) were solved iteratively using the
Newton-Raphson technique. The experimental data of
Merten et al. [15] were used for this purpose and the
simulations were carried out for different channel
dimensions (L/de). The predicted change in Pew with
Re is shown in Fig. 5. The figure indicates a very close
agreement between the predicted and experimental
Pew (experimental data correspond to L/de=16.56).
Interestingly, the agreement is excellent up to
Re=2620. It was expected that laminar-to-turbulence
transition would occur in the range 2000-2200. But
suction has stabilized the boundary layer, leading to a
delay in the onset of turbulent flow [6].
The prediction of flux in UF was carried out for
dextran and PEG solutions in the rectangular crossflow cell. The mass-transfer coefficient in this case
was evaluated using Eq. (64). This equation, along
with the osmotic-pressure model (Eqs. (70) and (72))

15

10
09

'

2000

I000

3000

Re
Fig. 5. Variation of dimensionless flux (Pew) with Re for RO
system. Solid lines are predicted flux and symbols are the
experimental data of Merten et al. [15]. l: L/de=5.0; 2: L/
de=16.56; 3: L/de=30.O; 4: L/de=60.0; 5: L/de=150.0; and 6: L/
de=300.0.

was solved, as described earlier, to obtain the permeate

flux. The experimental data of our earlier work [ 16]
are considered for comparison with the predicted
values. Intrinsic retention (Rr) for dextran was taken
as 1.0 and that for PEG, 0.9 [16]. The experimental
and predicted values of the permeate flux for dextran,
for all the operating conditions, are shown in Fig. 6.
The figure indicates an excellent match for the two.
For PEG, at all operating conditions, the predicted and
experimental flux values are plotted in Fig. 7. The
figure reveals a close agreement between the experimental and calculated flux values. The maximum
deviation between the values is 10%.
The comparison of predicted and experimental
permeate fluxes for ultrafiltration of PEG 6K in a
radial cross-flow cell was also carried out. In this case,
the average mass-transfer coefficient was evaluated
from Eq. (66). As mentioned earlier, the osmoticpressure model Eq. (70), along with Eqs. (66) and
(72), was solved iteratively. The experimental data for
UF of PEG are obtained from Ganguly [17], for
comparison with the predicted results. The value of
Rr was taken as 0.88 [17]. The comparison between
experimental and predicted permeate fluxes is presented in the non-dimensional (i.e. in terms of Pew)

S. De. P.K. Bhanacharya/Journal of Membrane Science 128 ~1997) 119-131

129

35

~ . . .9

Water/~Flux

i i

//

"

~
0 /'/O//

25

q)
o

/ , ~ ->, .s-a-5.
. ~ ".,
.........
3 :::fl::::::
.:A
.:!.S
:.:.

-~

, ~
!:'~ . ::::::{:%%...A. . . . . . .
/.S:j'cb . .......

/ / /

//(~

.'.:~l
.,',-::--'"

i 3

':~:.

,,* ,,q?.-'

"

is_,

+ 1 0 ~ / / ~

I i

300

150

450

i,

600

p i J i

25

35
Pe,~

Fig. 8. Fitting between predicted and experimental permeate flux

for UF of PEG 6K in a radial cross-llow cell. Dashed lines are for
10% deviations.

include the effects of suction and can be used for an

accurate prediction of permeate flux in both cross-flow
RO and UF.

18
/

[ i

Experimental

Fig. 6. Variation of permeate flux with pressure for UF of dextran

(T-20). l : c o - 1 0 k g m 3; 2:eo-30kgm 3; and 3:Co--50kgm 3
Symbols represent: open circle, ao=0.44ms 1 box,
Uo=0.38m s i; and triangle, uo=O.30ms -l. Curves are the
predicted values of the permeate flux.

+10~/

15

AP (kPa)

"~"

I i

4. Conclusions
v

/ /~///105g
C

"U

,//

I I I I I I t l t l l r l l l I ' l l l l l l t

I I 1 1 1

0
4
e~
Experimental Flux x lO m3/m2.s)

12

Fig. 7. Fitting between predicted and experimental permeate flux

for UF of PEG 6K in a rectangular channel. Dashed lines are for
10% deviations.
form, in Fig. 8, for all the experimental conditions.
Most of the predicted values lie within 4-10% of the
experimental data.
Therefore, the Sherwood-number relationships
developed in this work for different flow geometries

General Sherwood-number relations for cross-flow

RO and UF, including the effects of suction for
different flow geometries, were obtained from first
principles. Suction through the porous membrane had
a significant effect on the mass-transfer coefficient.
The proposed Sherwood-number relations were used
to predict mass-transfer coefficient and, in turn, the
permeate flux for both RO and UE The simple relations developed in this work to quantify the effects of
suction on mass-transfer coefficient should be of
immense help to the process and design engineers.

5. List of symbols
A1,2.3

B1
c
C*

Constants defined in Eqs. (15), (39) and

(51 ), respectively
Term defined in Eq. (70)
Solute concentration, kg m ~
Dimensionless solute concentration (C/Co)

S. De. P.K. Bhattacharya/Journal of Membrane Science 128 (1997) 119-131

130

de
d
h
11.2.3

k
K1,2,3,4,5,6

L
Lp
PEG
Pew
Q
F
/-

R
Re
RO
Rr
Sc
Sh
Sh
U

Uo

Vw
!)w
V

x
x

Y
UF

Diffusivity, m 2 s -1
Equivalent diameter, m
Diameter of the tube, m
Half-channel height, m
Integral defined by Eqs. (22), (44) and
(57), respectively
Mass-transfer coefficient, m s -1
Constants defined by Eqs. (20), (21),
(42), (43), (55) and (56), respectively
Channel length, m
Membrane permeability, m 3 N-1 S-1
Polyethylene glycol
Wall peclet number
Volumetric flow-rate, m 3 s - t
Radial coordinate, m
Dimensionless radial distance, (r/R)
Cell radius, m
Reynolds number
Reverse osmosis
Real rejection, (1 - Cp/Cm)
Schimdt number
Local Sherwood number
Average Sherwood number
Axial velocity, m s -1
Average bulk velocity, m s-1
Velocity, m s -1
Average permeate flux, m 3 m -2 s -1
Local permeate flux, m 3 m 2 s-1
Transverse velocity, m s-1
Axial distance, m
Dimensionless axial distance, (x/L)
Normal distance, m
Ultrafiltration

Greek symbols
AP
An
71"

P
#
A1,2,3

Pressure differential, Pa
Osmotic pressure differential, Pa
Osmotic pressure, Pa
Parameter defined by Eq. (10)
Parameter defined by Eq. (37)
Density, kg m -3
Viscosity, Pa s
Parameter defined by Eq. (49)
Pew/(Re Sc de[L)1/3, Pew/(Re Sc d/L) 1/3
and Pew/(Re S c h/R) 1/3, respectively

Subscripts
m
o
p

Membrane surface condition

Bulk condition
Permeate condition

Appendix

6. Physical properties of the solutes

The diffusion coefficient of dextran (T-20) was
taken as 6.75 x 10- l t m e s- 1 [ 18]. The diffusion coefficient of PEG (in m2s 1) was obtained from the
empirical equation, for a polymeric solution [16]:
DpEG = 1.50788 10 -1

(A1)

Osmotic pressure for dextran is obtained from the

correlation developed by Wijmans et al. [19], which is
given as:
7r -- (0.375c + 7.52c 2 + 76.4c 3) 105

(A2)

where 7r is in Pa and c is in g ml i.
The osmotic pressure for PEG was calculated from
Flory's equation [4,20].

References
[1] C. Kleinstreuer and M.S. Paller, Laminar dilute suspension
flows in plate and frame ultrafiltration units, AIChE J., 29
(1983) 529.
[2] C.R. Bouchard, P.J. Carreau, T. Matsuuara and S. Sourirajan,
Modeling of ultrafiltration: predictions of concentration
polarization effects, J. Membrane Sci., 97 (1994) 215.
[3] S. Bhattacharjee, A. Sharma and P.K. Bhattacharya, Surface
interactions in osmotic pressure controlled flux decline during
ultrafiltration, Langmuir, 10 (1994) 4710.
[4] S. Ganguly and P.K. Bhattacharya, Development of concentration profile and prediction of flux for ultrafiltration in a
radial cross flow cell, J. Membrane Sci., 97 (1994) 287.
[5] G.B. van den Berg, I.G. Racz and C.A. Smolders, Mass
transfer coefficients in cross flow ultrafiltration, J. Membrane
Sci., 47 (1989) 25.
[6] V. Gekas and B. Hallstrom, Mass transfer in the membrane
concentration polarization layer under turbulent cross flow. I.
Critical literature review and adaptation of existing Sherwood
correlations to membrane operations, J. Membrane Sci., 80
(1987) 153.

S. De. P.K. Bhattacharya/Journal of Membrane Science 128 (1997) 119-131

[7] S. De, S. Bhattacharya and EK, Bhattacharya, Development
of correlations for mass transfer coefficient in ultrafiltration
systems, Dev. Chem. & Min. Processes, 3 (1995) 187.
[8] T.K. Poddar, R.P. Singh and EK. Bhattacharya, Ultrafiltration
flux and rejection characteristics of black liquor and
polyethylene glycol, Chem. Eng. Commun., 75 (1989) 39.
[9] S. Dasgupta and EK. Bhattacharya, Comparative limiting flux
analysis of black liquor and polyethylene glycol in
ultrafiltration, Chem. Eng. Commun., 93 (1990) 193.
[10] S. Sridhar and P.K. Bhattacharya, Limiting flux phenomena in
ultrafiltration of kraft black liquor, J. Membrane Sci., 57
(1991) 187.
[11] C. Bhattacharjee and EK. Bhattacharya, Flux decline analysis
in ultrafiltration of kraft black liquor, J. Membrane Sci., 82
(1993) 1.
[12] W.S. Opong and A.L. Zydney, Diffusive and convective
protein transport through asymmetric membranes, AIChE J.,
37 (1991) 1497.
[13] J.S Shen and R.E Probstein, On the predictions of limiting
flux in laminar ultrafiltration of macromolecular solutes, Ind.
Eng. Chem. Fundam., 16 (1977) 459.

131

[14] EB. Bungay, H.K. Lonsdale and M.N. de Pinho (Eds),

Synthetic Membranes: Science, Engineering and Applications, D. Reidel Publishing Co., Dodrecht, 1983.
[15] U. Merten, H.K. Lonsdale and R.L. Riley, Boundary layer
effects in reverse osmosis, Ind. Eng. Chem. Fundam., 3
(1964) 210.
[16] S. De and EK. Bhattacharya, Flux prediction of black liquor
in cross flov, ultrafiltration using low and high rejecting
membranes, J. Membrane Sci., 109 (1996) 1(19.
[17] S. Ganguly, Development of concentration profile in cross
flow ultrafiltration using kraft black liquor in comparison to
polyethylene glycol, M. Tech. Thesis, IIT, Kanpur, 1991.
[18] L. Lebrun and G. Junter, Diffusion of dextran through
microporous membrane filters, J. Membrane Sci., 88 (1994)
253.
[19] J.G. Wijmans. S. Nakao, J.W.A. van den Berg, ER. Troelstra
and C.A. Smolders, Hydrodynamic resistance of concentration polarization boundary layers in ultrafiltration, J.
Membrane Sci., 22 (1985) 117.
[20] E J, Flory, Principles of Polymer Chemistry. Comell University Press. Ithaca, NY, 1953.