CONCENTRATION POLARIZATION IN REVERSE

OSMOSIS AND ULTRAFILTRATION

Mark C. Porter

Consultant
3449 Byron Court,
Pleasanton, California 94566

The major limiting factor in pressure-driven membrane processes such

as ultrafiltration and reverse osmosis is concentration polarization. This
paper discusses its effect on membrane flux and solute ~etention both the-
oretically and experimentally. Fluid management techniques effective in
reducing concentration polarization are reviewed.

1. INTRODUCTION

2. REVERSE OSMOSIS
2.1. Theory
2.2. Effect of Concentration Polarization on Salt Retention
2.3. Effect of Concentration Polarization on Membrane Flux
2.4. Fluid Management Techniques to Reduce Concentration Polarization

3. ULTRAFILTRATION
3.1. Effect of Concentration Polarization on Retention
3.2. Effect of Concentration Polarization on Membrane Flux
3.3. Evaluation of the Mass Transfer Coefficient

367

P. M. Bungay et af. (eds.), Synthetic Membranes: Science, Engineering and Applications, 367-388.
© 1986 by D. Reidel Publishing Company.
368 M. C. Porter

1. INTRODUCTION

The development of the anisotropic membrane in the late fifties by Loeb

and Sourirajan (1) was the major breakthrough that made reverse osmosis
(RO) and ultrafiltration (UF) commercially feasible . The membranes were
said to be "skinned," having a very thin retentive film integrally supported by
a finely porous substrate (see Figure 1). This gave the membrane sufficient
strength while the thin skin reduced the resistance to flow. Further, solutes
that penetrate the skin usually pass through the membrane with the filtrate,
resulting in virtually no entrapment of solutes within the pore network.
However, the development of anisotropic membranes is only half the
battle. The high-flux characteristics of these membranes result in rapid
convection of retained solutes to the membrane surface, in turn resulting in
the well-known phenomenon of concentration polarization. This accumula-
tion of solute at the membrane interface can severely limit the flux, leading to
an ~pparent fouling of the membrane, which some workers have interpreted
as pore blockage. That this is not the case is demonstrated by the easy
restoration ofthe initial flux simply by washing the surface ofthe membrane.
In RO, where the solutes retained can have a significant osmotic pres-
sure, concentration polarization can result in osmotic pressures considerably
higher than those represented by the bulk stream concentration. Higher
pressures are required to overcome the osmotic pressure and produce reason-
able flux values.

FIGURE I.-Cross-sectional view of a "skinned" membrane.

Concentration Polarization 369

For UF, where macromolecular solutes and colloidal species are involved,
the concentration at the membrane surface can rise to the point of incipient
gel precipitation, forming a dynamic secondary membrane on top of the
primary structure. This secondary membrane can offer the major resistance
to flow.
Therefore, the second major hurdle to be overcome in the development of
a practical industrial unit operation is concentration polarization.

2. REVERSE OSMOSIS

2.1. Theory

Figure 2 is a schematic of the boundary layer formation in RO. The

transport of water through the membrane with a volumetric flux J v carries
solute with it to the membrane surface at a transport rate of J i = J v c; where c;
is the concentration of solute in feed solution. Because the solute is mostly
retained by the membrane, the concentration at the wall (c:' int) increases until
the concentration-dependent back diffusion, D, ~~, plus the solute leakage
through the membrane B, (c; - c7), isjust equal to the transport of solute to the
membrane.

Jv
--Jv

C'i,bulk J-----
Bulk Feed Steam Laminar Sub-Laver

FIGURE 2.-Concentration polarization in reverse osmosis.

370 M. C. Porter

In RO, the solvent flux is represented by:

(1)

where J v volumetric flux of solvent

Av solvent permeation coefficient
I1p differential pressure across the membrane
11'IT differential osmotic pressure across the membrane

On the other hand, the solute flux is represented by:

(2)

solute flux,
permeation membrane constant for a specified solute (takes
into account solubility of solute in membrane and diffusibility
across membrane).

Therefore, a steady state mass balance for the laminar sub-layer results
in the following equations (2):

Jv c: - D::: = Bi (c: - c7), (3)

where D! solute diffusion coefficient in the feed solution,

x - distance parameter in solution perpendicular to membrane
surface,

and

C!,int
,
Ci,bulk
_
- exp
(JvD,'.0) ' (4)

where C!, bulk "bulk" solute concentration far from membrane surface,
cLint interfacial or "wall" concentration,
characteristic boundary layer distance in solution.

The ratio, c!,int1c!,bulk> is often referred to as the "polarization modulus;' It

is increased by a higher water flux and decreased by a reduction in the
boundary layer thickness, o. Because we want to maximize J v and we have
little control over D!, fluid management techniques designed to reduce 0 and
consequently increase the concentration gradient for back diffusion seem to
be the most practical approach.
Concentration Polarization 371

Using the mass transfer-heat transfer analogies, 8 may be calculated for

turbulent flow in tubular membranes by the following equations:

2(DD 1I3 V 2/3 Re 1l4

(5)
0.08 it

35.3(DD 1I3 VO. 42 dl14 , (6)

where v = kinematic viscosity,

dt tube diameter,
it mean velocity of the feed stream parallel to the membrane,
- d
Re Reynolds number = ~
v

Substituting equation 6 into equation 4 gives

C[.int _ [35.3Jv VO. 42 dl 14] (7)

C~
l,bulk
- exp (D')2/3 it3/4
l
•

Thus, the polarization modulus may be reduced by decreasing the tube

diameter or increasing the cross-flow velocity u. The latter is more important
as it appears to the 0.75 power; there is a weak dependence on tube diameter.
In l~minar flow, the solution is more complex because the boundary layer
is developing all the way down the tube. The results do not appear in closed
form and finite difference methods have been used (3). The essential dif-
ference is that tube diameter is much more important in laminar flow. This is
demonstrated in Figure 3, which shows the polarization modulus for tubular
membranes with a 50% water recovery and a flux of 10 gals/(ft2 • day). Flow
was assumed to be laminar for values ofthe Reynolds number at the channel
inlet less than 2100 and turbulent for values at the channel outlet greater
than 2100. Because half the water is removed through the membrane, there is
a range representing a factor of 2 in inlet velocity which separates these two
regions; in this transition region, a dashed line is drawn to connect the
laminar and turbulent flow solutions.
Note that since the water recovery is fixed at 50% in Figure 3, the
polarization modulus is independent of the inlet channel velocity UIn for
laminar flow but decreases with inlet channel velocity in turbulent flow. This
curious result is explained by the fact that even though in laminar flow the
feed velocity UIn affects the polarization at each point along the length ofthe
372 M. C. Porter

d, =1 in
\
\
\
\
10 in \
\
\
\
\
0.1 in

\ \

_·_0_.0_1_i_n -~----=_\~:~S:~:f.......::::::::;;-=--_.2_.0_1_in
1~__~~~~~~__~__~~~~______~~~~
0.1 10 100
Uln • ft Is

FIGURE 3.-Theoretical average concentration polarization modulus for tu-

bular reverse osmosis membranes with a 50% water recovery and an average
flux of 10 gal/(ft2) (day) as a function of inlet channel velocity.

channel, the length of the channel required for a given recovery increases
with increasing velocity, resulting in a constant average polarization
modulus.
Also note in Figure 3 that the polarization modulus decreases as the tube
diameter decreases in either laminar or turbulent flow, but the effect is
generally more pronounced in the laminar-flow regime.

2.2. Effect of Concentration Polarization on Salt Retention

Equation 2 indicates that the solute flux will increase with a higher
differential solute concentration across the membrane. This is demonstrated
in Figure 4, which shows the effect of brine velocities on Cl- rejection. Higher
velocities increase the rejection by virtue of decreasing the polarization
modulus. Higher flux membranes show the most marked improvement.
Concentration Polarization 373

100

1.24 GPD/FTZ I
so 7.8

:z MEMBRANE
0
~ 60 HEAT
......,
0 MODULE TREATMENT
...
II:
o II-I!l- I
• 10- 14 - I
(80·C, 30 MIN)
(80·C, 30 MIN)
, 40 11 12 -18 - 2 (SO·C, 30 MIN)
U c 8- 26- I (7'·C, 3 MIN)
~ • 12 -18 - 2 (6!l·C, 30 MIN)

20

0
0 2 !l
• 7 8 ,
AVERAGE BRINE VELOCITY (CM ISEC)

FIGURE 4.-Effect of brine velocities on Cl- rejection by heat-treated

membranes.

2.3. Effect of Concentration Polarization on Membrane Flux

Equation 1 indicates that the solvent fluxJv is decreased by an increased

osmotic pressure differential. The osmotic pressure 7T due to solute species i
may be approximated for dilute solutions by the well-known van't Hoff
equation

7T = Vi ciRT (8)

where Ci molar concentration of the solute

R gas constant
T absolute temperature
number of ions formed ifthe solute dissociates (e.g., for NaCl, V
= 2; for BaCI2 , v = 3)

Inasmuch as the osmotic pressure difference .17T is associated with the

concentration of solutes at the membrane surface, an increased polarization
modulus will result in a higher .17T and a reduced solvent flux J v . This is
demonstrated in Figure 5, which shows the effects of velocity (feed flow rate)
on membrane flux. Higher velocities decrease the polarization modulus,
decreasing the osmotic pressure difference and increasing the flux.
374 M. C. Porter

100
KP·98
P'"""'" KP·96
~
z
o ~
{3 Kp·90
w
-, .0
~
/
w
a:
o~

eo

-
KP·90
30

~
~
ell
"0
KP·96

-ro
------
OJ

~
20
())
"",-
X
:3
u.
Kp·98
!--'""
10
",-
-----
o
500 1000 1500
FEED FLOW RATE, eel min

FIGURE 5.-Effect of feed velocity on performance of Eastman Membrane

(typical values in Eastman's Test Cell at 600 psi and 78°F)
Concentration Polarization 375

2.4. Fluid Management Techniques to Reduce Concentration

Polarization
As already mentioned, maintenance of a high axial velocity is important
in reducing the polarization modulus. As water is removed from the tube, for a
given recovery, the velocity will drop, resulting in increased salt passage and
decreased water flux (see Figure 6). To alleviate this effect and maintain a
high axial velocity, a "Christmas tree" configuration or "tapered system" is
sometimes used (see Figure 15 of chapter on reverse osmosis).
In addition to maintaining high-sweep velocities, turbulence promoters
can also be effective in reducing the polarization modulus. Figure 7 shows the
effect of a promoter in increasing salt rejection.
Other techniques such as pulsed flow and secondary flow effects have also
resulted in improvements.

ppm
2000 ~ 6 6. --
~2000
0 , 1800
0..
(!)

:z: 1600
...
0

U - 50
III
~
~
::) PRODUCTION en
0
0
ex:
- - - SALT , 40 U)
PASSAGE 30 ~
0..
;;.~
,.~"1000 !J
. . , . . ..4--
__ ,:;" _6
20

--- --
4(

II:.:: ~... __ --
U)
2000 10
1000 0 it
0 30 50 75 90
-.4 RECOVERY

FIGURE 6.-Influence of recovery level on transmembrane flux of water and

passage of salt in reverse osmosis.
376 M. C. Porter

10 10

s 1
II
j
Z5
i/ V
j '/ #-
50~

,,
'1

----
60~

•
()
w
--
U)

iil
D
0 1-
§=; U)
D
0 75 a:
./. it .........
ow
I §=;
L
,;, ,-' v·
Z
r fi:
w
- 90~
CJ)

10"
.'
• ,//,
f
-

t
o IiO TURBUlENCE PfIOMOTEIt
• DtAC'jEO T~RB\""rC£ ,Pft0'1TU
r o I 4 I • 10 II M 'I Ie 20
Y.. (Re) 1/4 x 104
a

FIGURE 7.-Effect of twisted tape turbulence promoters.

3. ULTRAFILTRATION

For UF, the macromolecular solutes and colloidal species usually have
insignificant osmotic pressures. Nevertheless, flux is affected along with
retention by concentration polarization and the formation of a gel-layer or
secondary membrane. In RD, the precipitation of macromolecules or salts on
the membrane surface would be called "fouling?' In UF, the higher flux often
drives the concentration at the wall above the solubility limit and we are
accustomed to operating in the gel-polarized regime (see Figure 8).
Concentration Polarization 377

Membrane

II
I
I
I n' int ~ n'bulk

Jv---~ :
I
I
I
I - - -•• JVe"i
e'i, bulk

e"i« e'i
Laminar ke (sludge)
Sublayer
(Non Newtonian)

FIGURE 8.-Boundary layer formation in ultrafiltration.

3.1. Effect of Concentration Polarization on Retention

For dilute solutions without gel formation, the rejection of solutes is not
unlike the mechanism discussed for ROo As shown in Figure 9, increases in
transmembrane pressure increase the flux and the polarization modulus,
with a resultant decrease in rejection.
For more concentrated solutions, after incipient gel precipitation, the
concentration at the membrane surface is fixed and solute passage occurs at a
constant rate irrespective of changes in the feed concentration.
The biggest disappointment is what happens with binary solutes. For
example, the pharmaceutical industry is keenly interested in having a UF
membrane capable offractionating albumin/globulin mixtures. In fact, there
are membranes available that offer better than 95% retention for gamma
globulin and less than 10% retention for albumin. However, as shown in
Figure 10, the presence of even small amounts of gamma globulin forms a gel
layer on the membrane that is retentive for the smaller albumin molecule.
This limits our ability to fractionate various proteins. On a laboratory basis,
one can dilute the mixture, perform the fractionation,"and then reconcentrate
the two fractions. However, this is too cumbersome for a production process.
 ...;r
'"
CXl

1.0 i ....:::
(22)
Figures in brackets indicate order of runs
0.8

g 0.6
+-
u
Q>
Q>
0::: 0.4

0.2 (14)

0' '.
o 20 40 60 80 100
Transmembrane pressure drop, Ib/in 2
~
o
FIGURE 9.-Decrease in dextran rejection with increasing pressure difference across ultrafiltration membrane. ;c
~
 Concentration Polarization 379

REJECTION
OF HUMAN
SERUM
ALBUMIN

20%

1~L."~~~ ________ ~ ______ ~ ____ ~ __ ~ ____. .

0.01% 0.1%
CONCENTRATION OF "I-GLOBULIN (WT. %)

FIGURE 10.-Rejection of albumin by gamma-globulin polarization layer.

3.2. Effect of Concentration Polarization on Membrane Flux

Figure 11 shows flux data as a function oftransmembrane pressure drop.

The curious thing about these data is that at high enough protein con-
centratIons and pressures the flux is invariant with pressure. Apparently the
gel layer resistance Rc increases to offset any increases in /1p

(9)

where Rm is the membrane resistance.

At steady state and with no leakage of solute, equation 3 applies within
the laminar sublayer

J v c/ - D-' d c;'
, dx
O. (10)

For any increase in /1p, J v increases, carrying more solute to the wall.
However, the back-diffusive transport is fixed. Therefore, the gel layer in-
creases in thickness or is compacted by the increase in /1p such that the
resistance to flow increases, reducing J v to a level balancing the back-
diffusive transport.
380 M. C. Porter

0.10 r--.,-----,,-------,---.,--,---.----.------.
• 0.9% SALINE
UM-IO MEMBRANE
M-50 CELL
'"E 0.08
0.65% PROTEIN 0830 RPM)
~
]
-?
x 0.06
3
"-
z
o
f= 0.04
«
Q:
~
iL
«
:: 0.02
...J
:::>

10 20 30
TRANSMEMBRANE PRESSURE (psi)

FIGURE H.-Effect oftransmembrane pressure difference on ultrafiltration

flux for bovine serum albumin solutions.

Note in Figure 11 that lower protein concentrations have a higher

asymptotic flux level and the threshold pressure to achieve pressure indepen-
dence is higher. This is because the driving force to remove material from the
wall is higher. Assuming the gel concentration, cg ' = c;. geb is fixed, lower bulk
concentrations, Cb ' = C;, bulk, provide a higher concentration gradient dc;' Idx.
Likewise, note that higher stirrer speeds also result in higher asymptotie
flux levels and threshold pressures. This is because the boundary layer
thickness has been decreased, again providing a higher concentration gra-
dient for back-diffusive transport.
Figure 12 demonstrates that the gel layer is indeed the limiting resis-
tance to flow. Three membranes of widely different water permeabilities have
essentially the same flux with proteins ifthe transmembrane pressure drop is
sufficiently high. The decrease in flux with the XM-100 membrane may be
due to membrane compaction.
The gel polarization model enables us to fix the boundary conditions and
solve the differential equation 10

(H)
Concentration Polarization 381

o.o3o,-----,-----r---,------,---,-----r----.----,
_ _ _ _ _ _ .A... _ _ ...A._
i 0.025
_...."'='".---~-;:- -----7:"
.",...-'" _ . " -0--" ---o-.. ---Oa_
N
E
~
/'.. ./'.
I 0
! 0.020 o o
~ I ./.
1 r
0

~
I ./'
... 0.015
~ PURE H20 FLUX AT 10 psi

i ~ (ml/cm 2 - min)

.
~ 0.010
;;: I ,7 0

"
UM-IO
PM-30
0.04
0.15

/
It: XM-IOO 0.25

~
0
!:i
::>
0.005

.j'
~L---~5---L10---415--~20~-~2~5--~3~0---3~5---J40
TRANSMEMBRANE PRESSURE (psi)

FIGURE 12.-Flux-pressure relationships for human plasma in stirred cell by

use of membranes of differing water permeabilities.

where k i is the solute mass transfer coefficient,

ki = D;'/B. (12)

Equation 11 indicates that under conditions where the gel polarization

model holds, the membrane flux is invariant with respect to the trans-
membrane pressure drop or membrane permeability and is dependent only
on the solute characteristics (D;' and cg ') and the boundary layer thickness B.
Thus fluid management techniques must be directed toward decreasing the
boundary layer thickness, i.e., toward increasing the mass transfer
coefficient ki'
The validity of equation 11 has been demonstrated for a large number of
macromolecular solutes and colloidal species. The data of Figure 13 show the
semi-logarithmic variation of flux with concentration for two proteins and
two colloidal suspensions. Note that the gel concentration cg ' may be deter-
mined from these plots as the intercept with the abscissa (where the flux
drops to zero and ci,' = cg '); it is higher for colloidal suspensions (60-70%)
than for protein solutions (25-45%). This is not without significance: in the
case of colloidal suspensions, the gel layer would be expected to resemble a
layer of close-packed spheres having 65-75% solids by volume.
The data of Figure 14 show the increase in the slope of the J v us In ci, plot
(i.e., the mass transfer coefficient kJ for higher recirculation rates (tangential
velocities). Higher velocities would be expected to decrease the boundary
layer thickness-increasing ki'
382 M. C. Porter

COD':
<1> "l"crRo O£POSl rION PRIMER
A ~"""'R£Ne ~G/rAOI£/VE LAr£x
o "OMAN ALSUMIN
Q'ZZO o GELArlN (@ 70' C)

"
~ zoo
~ ISO
~ 160

~ 140
«
~ IZO
::!
Ii 100
~
~ SO
~
60

40

20

o
Z 3 4 5 0; 78 910 20 30 40 5060 70BO
CONC£NrRAr/ON (wr·... SOUOS)

FIGURE 13.-Variation of flux with concentration for several solutions and

suspensions by use of thin-channel ultrafiltration.

140
Flow Rate
0
LL 0--0 3000 ee/min
en 120
~ 0--0 2000 eel min
x 100 6--6 1000 ee/min
::>
...J
LL
0---0 500 ee/min
w 80
TC-1C PM 30 MEMBRANE
~
a: 30 MIL CHANNEL
60
.....
....J
LL
<{ 40
a:
.....
...J
::> 20

2 3 4 5 678910 20 30 40 50 60
PROTEIN CONCENTRATION (WGT %)
FIGURE14.-Cross plot of flux vs. recirculation rate data (TC-IC, PM30
membrane, 30-mil channel).
Concentration Polarization 383

3.3. Evaluation of the Mass-Transfer Coefficient

The mass transferlheat transfer analogies well-known in the chemical
engineering literature make possible an evaluation of the mass-transfer
coefficient k i and provide insight into how membrane geometry and fluid-flow
conditions can be specified to optimize flux. A detailed derivation, using the
Graetz and Leveque solutions for convective heat transfer in laminar flow and
the Dittus-Boelter correlation for fully developed turbulent flow, is presented
in reference (4). The final solutions are given here:

For laminar flow:

_ [11 (D;,)2J 0.33 (13)

ki - 1.62 dh L '

where d h = equivalent hydraulic diameter, (d h 2>.. for flat rectangular

channels of height >")
L = channel length,

or

k = 0816 [6 Q (D;,)2JO.33 (14)

, • >..2 wL '

where Q feed stream volumetric flow rate,

w channel width.

Figure 15 shows the 0.33 power dependence on the wall shear rate
(proportional to I1ldh ) for a variety of channel configurations and dimensions.

For turbulent flow:

110.8 (D;')0.67
k i = 0.023 df.2 V0.47 (15)

or

(16)

Equations 13-16 show that for both laminar and turbulent flow, the mass
transfer coefficient may be increased by increasing the channel velocity and
decreasing the channel height. The flux is dependent on channel length in
laminar flow because the boundary layer is still developing even at the end of
the channel. Naturally, in turbulent flow there is no dependence on channel
length because the boundary layer is developed rapidly at the channel inlet.
384 M. C. Porter

0.3 140
0.2 70
c
52.5
E
N
0.1 35
E 0.08 28
u
~ 0.06 21
E
0.04 14
..,>
x "0
3 0.02 7 C'
IJ.. ..,>
z
Q 0.01 3.5
~ 0.008 2.8
~ 0.006 2.1
:...J
41 0.010
i<: 0.004 0 6.5 0.005 1.4
<t 6.5 0.017
0: 65 0.035
~ 13 0.005
::> 0.002 13
13
0.017
0.035
0.7
40 0.008

0.001 .. I • 0.35
• .-'--'..J.• ..J.......I'--_--'---.J.--'--'--~..L.u
10 100 1000 10,000
WALL SHEAR RATE -,
CHANNEL LENGTH' (sec-em)

FIGURE 15.-Flux dependence on wall shear rate in laminar flow.

Equations 13-16 can be used satisfactorily to calculate quantitatively

the flux for macromolecular solutions, but they are inaccurate for colloidal
suspensions. Figure 16 shows excellent agreement between theory and ex-
periment for human albumin solutions, but Figure 17 shows calculated flux
values for latex that are more than an order of magnitude too low (4). An
evaluation of more than 40 different colloidal suspensions in the author's
laboratory have indicated that the diffusion coefficient calculated from the
flux is generally from one to three orders of magnitude higher than the
theoretical Stokes-Einstein diffusivity for colloidal suspensions.
It was first suggested by the author (4) in 1972 that the augmented
colloidal particle mass transfer coefficients could be explained by the so-
called "tubular-pinch effect". The lateral movement of suspended particles
flowing down a tube toward the center-line of the tube was first observed in
1836, when Poiseuille noted that the region immediately adjacent to the walls
of a capillary tended to be free of blood corpuscles. In 1961 and 1962, Segre
and Silberberg (5) noted that rigid, spherical, neutrally buoyant particles
carried along in Poiseuille flow also migrate radially. The lateral migration of
rigid particles arises from fluid inertial effects (6).
Concentration Polarization 385

RECIRCULATION RATE (ccJHIN)

100 1000

FIGURE l6.-Ultrafiltration of human albumin (TC-le, PM30 membrane).

200

100
80
>. 60
0
:g
N
....
....
40
.:'1. '" '"
'"
"'"
0 20
0,,,,"0"
:i.'v" ~ ...~ .1....
'"
o~"',,~"'~ ,\~"
,; ,\~'l; • 0'" \0"
..:! " ...'",,~'QYi !\,c;" ... '\
10 ( " ~"'"'O~'"~
'\~"'o'"
....'" 8 I ",/...,,~
~
;;::
6 I ... "
I I
4
....~ I I
::; . \ \0'0_.) I
~\eo_- I
2 ~~eole_- ~,o'll 'Z.,~
_ -,;"(!\i(lOI O,S e(!\ I
"?> '/.. \ 0·'·.)
\) ~ '2.. ~ieO' "--
~~~I!. ~ ;; \,0'11
"",'" \otf\\(\

100 200 4006001000 2000 4000

Recirculation rate (mL/min)

FIGURE l7.-Ultrafiltration of styrene butadiene latex suspension in a mod-

ule with a 30-mil channel and PM30 membrane at an average trans-
membrane pressure difference of 40 psi.
386 M. C. Porter

The lateral migration velocity, VL, is given by an expression of the form:

(17)

where b particle radius,

rt tube radius,
n positive integer,
r radial position of the particle in the tube,
F [rlr t ] = function of the ratio, rlrh particle buoyancy and tube
orientation.

Thus for colloidal suspensions, equation 10 must be modified to read:

Jv C,·' -
d c/
D·, -
I
- - VL Ci
I
=
0, (18)
dx

Integrating the above using equations 12 and 17 gives

J
v
= F u2
v
dt (!!.-)n
rt
+ k. In
'
(C g
Cb '
I
).
(19)

Belfort (7) has obtained agreement between theory and experiment

using the data of Porter (4) and taking into account the tubular-pinch effect.
These results suggest another technique besides fluid management and
membrane geometry to reduce concentration polarization. The introduction
of particulates to a solution may tend to transport adsorbed macromolecules
away from the membrane surface thereby enhancing the flux. Dorr-Oliver
has done this in commercial plants and Bixler obtained a patent (8) for
improving the ultrafiltrate flux by introducing solid particulate materials
into the process stream (see Figure 18).
A review of equations 13-16 miglit lead one to surmise that hollow fibers
(see Figure 19) represent the optimum membrane geometry. Inside diame-
ters of 0.2 to 2 mm can be spun easily and there is no parasite drag-Le., there
is no hydraulic friction against an inactive surface. Unfortunately, hollow
fibers seldom operate in the gel-polarized regime because of the inherent
limitation of the burst strength. This limits the inlet pressure to the fiber and
thereby limits the maximum attainable velocity. Flux values are often below
those obtainable in gel-polarized UF and are pressure dependent. New tech-
niques for supporting and/or making stronger hollow fibers should overcome
this problem.
Concentration Polarization 387

2.

2.

2.

Ii
2. 7(1. rtI'VI BEADS
UIH "E"BRANE

""
0
l-
V
~
!2;
~
"-'
1;
!""
'~

0.8
0 10 15 20 25 30 35 40
PARTICLE LOADING (VOL. %)

FIGURE 18.~Stirred cell flux improvement factor for various loadings of glass
and MMA beads.

HOllow fiber fluid flow

FIGURE 19.-Hollow-fiber fluid flow. Type A hollow-fiber cr.oss-section.

388 M. C. Porter

REF'ERENCES
1. Loeb, S.: 1981,. "The Loeb-Sourirajan membrane: How it came about:' Synthetic Membranes:
Desalination (A.T. Turbak, ed.) Vall, ACS Symposium Series 153, ACS, Washington, D.C.
2. Lonsdale, H.K.: 1972, "Theory and practice of reverse osmosis and ultrafiltration;' Industrial
Processing with Membranes, R.E. Lacey and S. Loeb lEds), John Wiley, New York, pp.
123-178.
3. Porter, M.e.: 1979, "Membrane filtration:' section 2.1, Handbook of Separation Techniques for
Chemical Engineers, (Ed. P.A. Schweitzer), McGraw-Hili, N.¥.
4. Porter, M.C.: 1972, "Concentration polarization with membrane ultrafiltrati.on;' I&EC Prod·
uct Research and Development, vol 11, pp. 234--248.
5. Segre,. G., and Silberberg, A.: 1962, "Behavior of macroscopic rigid spheres in PoiseuiHe flow:'
~ J. Fluid Mech. 14, pp, U5-157.
6. Brenner,. H.: 1966, "Hydrodynamic resistance .of particles at small Reynolds numbers;' AdJ-
vanees in Chemical Engineering 6, pp. 287-438.
7. Belfort, G., and Green, G.: 1980, "Fouling of ultrafiltration membranes; Lateral migration and
the particle trajectory model;' Desalination 35, pp. 129--147.
8. Bixter, H.J., Rappe, G.C.: (Nov. 17, 1970). U.S. Patent 3,541,006.

SYMBOLS
F particle radial migration function, equation 17
L channel length
Ii feed stream mean velocity parallel to membrane
VL lateral migration velocity

Subscripts
b, bulk interior of stream outside of boundary layer
g, gel gel layer