Advances in Chemical Engineering and Science, 2012, 2, 309-320

http://dx.doi.org/10.4236/aces.2012.22036 Published Online April 2012 (http://www.SciRP.org/journal/aces)

Comparison of Co-Current and Counter-Current Flow

Fields on Extraction Performance in Micro-Channels
Benny Malengier1, Subramaniam Pushpavanam2*
1
Department of Mathematical Analysis, Ghent University, Ghent, Belgium
2
Department of Chemical Engineering, Indian Institute of Technology Madras (IIT Madras), Chennai, India
Email: [email protected], *[email protected]

Received December 12, 2011; revised January 24, 2012; accepted February 25, 2012

ABSTRACT
Several applications such as liquid-liquid extraction in micro-fluidic devices are concerned with the flow of two immis-
cible liquid phases. The commonly observed flow regimes in these systems are slug-flow and stratified flow. The latter
regime in micro-channels has the inherent advantage that separation of the two liquids at the exit is efficient. Recently
extraction in a stratified counter-current flow has been studied experimentally and it has been shown to be more effi-
cient than co-current flow. An analytical as well as a numerical method to determine the steady-state solution of the
corresponding convection-diffusion equation for the two flow-fields is presented. It is shown that the counter-current
process is superior to the co-current process for the same set of parameters and operating conditions. A simplified
model is proposed to analyse the process when diffusion in the transverse direction is not rate limiting. Different ap-
proaches to determining mass transfer coefficient are compared. The concept of log mean temperature difference used
in design of heat exchangers is extended to describe mass transfer in the system.

Keywords: Plug Flow; Diffusion; Extraction; Co-Current; Counter-Current

1. Introduction behavior, e.g. droplet, slug or stratified flow. These re-

The physical effects which govern the behavior of fluids gimes depend on the relative flow rates of the fluid
flowing at the micro-scale and the macro-scale are dif- phases involved, the resulting interaction between inter-
ferent. Surface tension, viscous effects, energy dissipa- facial and viscous forces and the wetting behavior of the
tion and capillary action begin to dominate system be- channel walls. Different liquid-liquid two-phase flow
havior at the micro-scale. Micro-fluidics studies the be- patterns in micro-channels have been experimentally
havior of the fluids at the micro-scale induced by these analysed, see Dessimoz et al. [1].
effects. It helps exploit the behavior at these scales for In the context of mass transfer across membranes Guo
new applications by improving the efficiency of current and Ho [2] have analysed an analytical solution based on
processes. separation of variables for co-current and counter-current
Micro-fluidics results in Process Intensification through flows in an annulus. The eigen-functions were obtained
miniaturization. These systems are characterized by a using a power series expansion. Here the velocity fields
dominance of viscous forces as compared to inertial in the two chambers were decoupled as they were sepa-
forces; hence, typically low Reynolds numbers are en- rated by a porous membrane. Electro-osmotic flows in
countered. Consequently, the flow regimes observed in micro channels have been frequently proposed as a way
these systems is laminar. Mixing in these micro-channels to overcome dispersion effects. This causes the velocity
occurs primarily by molecular diffusion. The time needed profile in the micro-channel to remain almost uniform
for mixing by molecular diffusion is proportional to the except for a small region near the walls. The flow-field
square of the length of the diff usion path. The marked under these conditions can be approximated as an ideal
shortening of the diffusion path in a micro-channel re- plug-flow in the micro-channel. Liu et al. [3] and Gao et
sults in relatively good mixing. al. [4] have studied the electro-osmotic flow in a rectan-
Multiphase liquid-liquid flows arise when two or more gular channel when one fluid is conducting and the other
partially miscible or completely immiscible fluids are is non conducting as is usually the case in extraction. The
brought in contact and subjected to a pressure gradient. flow-profiles were obtained analytically for the two flu-
The resulting systems display different kinds of flow ids when the non conducting fluid was dragged by the
* viscous force of the conducting fluid. Wang et al. [5]
Corresponding author.

Copyright 2012 SciRes. ACES

310 B. MALENGIER ET AL.

have studied experimentally as well as theoretically how the performance of co-current and counter-current flows
the interface level can be controlled in a two liquid sys- in extraction is being studied with focus on micro-chan-
tem under the influence of a pressure drop and an electric nels. Recently it has been experimentally shown that
field applied to one of the fluids. Their results indicate counter-current operation is possible in micro-channels.
that the flow profiles in the two fluids can be approxi- The primary objective of this work is to establish condi-
mated as plug-flows with a jump discontinuity at the in- tions under which the counter-current operation is supe-
terface for some operating conditions. rior to the co-current operation. To the best of our
Recent studies have focused on liquid-liquid extraction knowledge a theoretical analysis of this system has not
in the stratified flow regime in the micro-channels. Here been carried out. The main motivation is to show that
the two fluids flow side by side. This flow-pattern can be improvements in the extraction performance are possible
exploited to facilitate complete separation at the channel when the flow is counter-current as opposed to co-cur-
exit. The extraction of vanillin dissolved in water using rent. The convective diffusion equation is solved ana-
toluene in micro-structured devices made of Poly Di lytically for the co-current operation. This is a one di-
Methyl Siloxane (PDMS) was studied experimentally by mensional model with diffusion being considered only in
Fries et al. [6]. Here the performance of segmented and the direction transverse to the flow direction. Here axial
stratified flow regimes were compared. LIF and micro- dispersion effects are neglected. The counter-current
PIV measurements showed a laminar profile for stratified system is solved numerically. The algorithm proposed
flow, whereas vortices in the slug were detected for seg- exploits the features of the system. A lumped model is
mented flow. The influence of channel width and there- analysed where the concentration dependency on the
fore, the surface-to-volume ratio was investigated for flow direction alone is considered. It is shown that the
stratified flow. There was a significant enhancement of counter-current flow performs better than the co-current
mass transfer with decrease in the channel cross-sectional flow. Different methods to compute the mass transfer
area for the stratified flow regime. coefficient as proposed in the literature are compared.
Three different fluid-flow patterns in a Y-shaped mi-
cro-channel, contact or stratified flow, segmented flow 2. Co-Current and Counter-Current Flow
and emulsification were investigated in Okubo et al. [7].
We consider three different flow regimes for the analysis
Here a one-dimensional model for extraction assuming
in this work: 1) co-current laminar flow, 2) co-current
the interface to be at the centre of the channel was used
plug flow, and 3) counter-current plug flow. The strati-
to compare the model predictions with the experimental
fied flow (fluids flow side by side as shown in Figure 1)
behavior. A two-dimensional flow-field taking into ac-
of two liquid phases is analysed between two infinite
count the effect of the interface not being at the centre
was analysed numerically by nidari-Plazl and Igor horizontal plates extending to infinity in the z-direction.
Plazl [8]. They compare their model predictions with This assumption on the geometry helps us focus on the
experimental results on steroid extraction. Most of the physics of the problem keeping the mathematics tractable.
research in stratified flow has been when the fluids flow The distance between the plates (along the x-direction) is
co-currently i.e. in the same direction. TeGrotenhuis et al. taken as H and the liquid-liquid interface is at distance hl
[9] has studied the counter-current mass transfer in a mi- (subscript l for laminar) from the lower plate. The flow is
cro-channel when the two fluids are separated by a assumed to be in the y-direction. Figure 1 shows a
membrane. Here the diff usional resistance through the schematic of the system being analyzed.
membrane was incorporated in the analysis. Recently it
has been shown that by suitable modification of the 2.1. Laminar Flow
channel surface it is possible to have counter-current In the case of laminar or Poiseuille flow, the velocity
flow in the micro-channels over a wide range of operat- profile is obtained assuming the flow to be steady, fully
ing conditions, Aota et al. [10]. They found that a maxi- developed and the liquids to be incompressible. The ve-
mum possible theoretical plate number of 4.6 is achiev- locity profiles of the system are governed by the equa-
able in counter-current flow as opposed to co-current tions
flow. The pressure drop characteristics in a counter-flow
micro-channel have been investigated by Hibara et al. p 2v
1 21 for 0 x hl ,
[11]. The velocity profiles in a 100 micron channel with y x
(1)
butylacetate and an aqueous phase flowing in a counter- p 2v
current manner was measured using micro-PIV, Aota et 2 22 for hl x H .
y x
al. [12].
The counter-current flow is known to be more ecient These equations are subject to the conditions of no slip
in the context of heat exchanger networks. In this work at the walls and continuity of velocity and shear stress at

Copyright 2012 SciRes. ACES

 B. MALENGIER ET AL. 311

(a) (b)

Figure 1. Micro-channel layout. 1,2 describe regions occupied by the two fluids. These are separated at the interface h. The
direction of velocity determines the flow type. (a) Co-current; (b) Counter-current.

the interface which is located at hl. So 2.2. Co-Current Plug Flow

v1 0 for x 0, In this case the velocities v1 and v2 have the same sign,
v1 v2 for x hl , and are constant within their phase. It is well established
v1 v2 (2) that plug flow behavior can be achieved in a micro
1 2 for x hl , channel using electro-osmosis [4]. This eliminates axial
x x
dispersion effects which arise from laminar flow in mi-
v2 0 for x H .
cro-channels. The electric field can also be used to con-
In the above the subscript 1 and 2 are used to denote trol the interface height for a fixed combination of flow-
the fluid in the first and second region respectively. The rates.
solutions to the above equation yield the velocity profiles In the case of plug flow, with equal velocity in both
of the two liquids as Equation (3). phases, we denote interface as h = hn. Here
Here the imposed pressure gradient is denoted as P .
Both the fluids are subject to the same pressure drop. The hn Q1 H hn Q2
,
flow behavior is hence similar to the Hagen-Poiseuille H Q1 Q2 H Q1 Q2
flow (parabolic in shape). The velocity field is continu-
When plug flow is assumed viscosity does not play
ous at the interface hl but its derivatives are discontinu-
any role in determining the interface position. If on the
ous. A schematic of the velocity profile in the channel is
other hand we assume the flow to be laminar, then the
shown in Figure 2.
interface position is determined by the viscosity of the
The experimentalist operates the system at fixed flow-
two fluids.
rates Q1 and Q2.
They are given by Equation (4).
2.3. Counter-Current Plug Flow
These equations can be used to determine the pressure
drop P and the height h of the interface for a given In this case the velocities v1 and v2 have an opposite sign.
combination of flow-rates and fluids. Alternatively, if the We take v1 0 and v2 0 . Counter-current flow can
pressure drop and height of interface are specified, the be theoretically simulated using a combination of Poi-
velocity profiles in each liquid layer can be found and seulle flow and a Couette flow. To generate a clear sepa-
from this the flow rates can be determined. ration of the two phases the interface must be located at

v1 x

x hl 2 2 1 xH 1 H 2 1 xhl 2 1 P for 0 x hl ,
2 hl 2 1 H 1 1
(3)
H x ( Hhl 2 1 xH 1 hl 2 2 1 xhl 2 1
v
2 x P for hl x H ,
2 hl 2 1 H 1 1

hl

h 2 h 2 h 2 2 Hhl 1 3H 2 1
Q1 v1 x dx l l 2 l 1
P ,
0 12 2 hl hl 1 H 1 1
(4)
H
4 2 hl H 3 2 hl 4 4hl 1 H 3 hl 4 1 H 4 1 4 H 1hl 3 9hl 2 2 H 2 6hl 2 1 H 2 6 Hhl 3 2
Q2 v2 x dx P
hl 122 2 hl hl 1 H 1

Copyright 2012 SciRes. ACES

312 B. MALENGIER ET AL.

C1 C in for 0 x h, y 0
C2 0 for h x H , y 0, for co-current flow
C2 0 for h x H , y L, for counter-current flow
(7)
The film interface conditions result in a discontinuous
concentration profile, while keeping the mass flux con-
tinuous. If K > 1, C2 remains below the value of C1 at the
interface. When K < 1 the reverse is true and the second
fluid extracts the solute out of the first strongly. Here the
concentration C1 is depleted at the interface and we ob-
Figure 2. Schematic picture of stratified flow system show- tain a larger C2 concentration. In our computations we
ing all variables. The velocity profile shown is for 1 > 2, use Cin = 1 mol/m3.
P 0 . In the co-current flow (superscript co) the concentra-
tions of the outlet streams both tend to an equlibrium and
the point where the velocity is zero. Alternatively com- this limits the extraction. In the counter-current flow
bination of electro osmotic flow with Poiseulle flow can (superscript cc) this limitation does not exist and hence
give a counter-current flow when there are two immis- the performance is much better.
cible liquids as the electric field affects the flow of only For co-current (laminar or plug flow), the conservation
one of the two fluids. The electric field can be manipu- of mass states that
lated to increase or decrease the velocity for a fixed flow
rate. This can be used to control the interface position Q1C in Q2 C2 eq , co Q1C1eq co
h. Experimentally counter-current flow has been achi- under steady-state conditions. This is valid for a long
eved by surface modifications of the micro-channels [10]. channel when the two exiting streams are in equilibrium.
For the sake of mathematical simplicity the velocity in Hence
each of the phases is assumed to be uniform across the
transverse direction in counter-current flow. C in C in
C2eq ,co C2 eq , cc
Q2 Q1 K K
3. Mass Transfer in Extraction
For counter-current flow the overall mass balance
The mass transfer behavior in stratified flow of a liq- gives
uid-liquid extraction system in a micro-channel is now
analysed. Here we consider the flow of a solute in the Q1C1y=0 = Q1C1y=L + Q2 C2y=0
first fluid which is being extracted by the second fluid. This is used to check the numerical solution. The
The concentration in fluid 1, respectively 2 is represented mixed cup average concentration at a particular y is
by C1, respectively C2. Considering steady-state opera- given by Ci y vi Ci x, y dx vi dx .
tion with convection in the y-direction and diffusion in
the x-direction we obtain the equations which govern the 4. Analytical Solution for Co-Current Plug
behavior of the system as Flow
v1 y C1 D1 xx C1 for 0 x h, The convection diffusion equation can be solved analyti-
(5)
v2 y C2 D2 xx C2 for h x H . cally and elegantly under the assumptions of 1) the
co-current Plug Flow Regime (PFR) when the velocity in
Here the expressions of v1, v2 take on distinct values the two fluids is uniform (v1 and v2 are constant), and 2) a
for laminar, co-current and counter-current flows. At the constant transverse diffusion coefficient (D1 and D2). We
interface we have, start with non dimensionalizing the equations with re-
spect to their characteristic lengths and initial concentra-
C1 KC2 for x h,
(6a) tions,
D1 x C1 D2 x C2 for x h,
x y h C
x* , y* , h* , and Ci* ini
at the walls we have H L H C
x C1 0 for x 0, y 0 which gives the dimensionless form as
(6b)
x C2 0 for x H , y 0, v H2
1
y Ci xx Ci where Pei i , i 1, 2. (8)
and at the inlet y = 0 we have Pei Di L

Copyright 2012 SciRes. ACES

 B. MALENGIER ET AL. 313

For simplicity, we drop the superscript * from now on.

We seek the solution Ci (x, y) in the form gi(y) fi(x). Sub-
2n Pe1 h sin 2n Pe2 h
bn

stituting this in (8) gives 4n Pe1

gi y

1 fi x
2
2 Pe1 sin 2 n Pe1 h
gi y Pei fi x K

Pe2 sin n Pe2 1 h
2

Or 1 (15)


2 1 h n Pe2 sin 2n Pe2 1 h 2
fi x 2 Pei fi x 0
(9) 4n Pe2
gi( y ) 2 gi y 0
This results in an eigen value problem in the x direc-
d n bn
Pe1 sin n Pe1 h
tion whose solution is

Pe2 sin n Pe2 1 h

f1 x a sin Pe1 x b cos Pe1 x , The solution for the y dependency is
0 x h,
(10)

g n y kn exp n2 y .
For co-current extraction the initial condition is C1 = 1

f 2 x c sin Pe2 1 x d cos Pe2 1 x for 0 < x < h and C2 = 0 for h < x < 1. The coefficient kn
h x 1 can be obtained from the initial condition as

The boundary conditions at x = 0, 1 yield a = c = 0. At

kn bn

sin n Pe1 h
x = h, the boundary condition C1 = KC2 implies n Pe1

b cos Pe1 h Kd cos Pe2 1 h (11a) Since the boundary conditions are homogeneous
Neumann in the x-direction, = 0 is also an eigen value,
while D1 x C1 D2 x C2 implies which corresponds to n = 1. The eigen-function corre-
sponding to this is the equilibrium solution and is given

bD1 Pe1 sin Pe1 h dD2 Pe2 sin Pe2 1 h by
(11b) 1 0 x h,
h
C eq x, y (16)
We seek b and d to be non-zero. This yields the char- 1 K h
h x 1
1 h
acteristic equation which determines the eigen values as K
the solution to
The complete solution to the convection diffusion

Pe2 cos Pe1 h sin Pe2 1 h (12)
equation is hence

K Pe sin Pe h cos Pe2 1 h 0 1

C x, y
h
1 1
1 K h
1 h
where D1 D 2 . The eigen functions corresponding K
to the nth eigen value is

sin n Pe1 h exp
f1n x bn cos n Pe1 x , 0 x h,
bn
n Pe1
n
2

y f n x
n2
f n x

f 2 x d n cos n Pe2 1 x , h x 1

n
0 x h,
h x 1.
(13)
(17)
It has been shown in [13] that this system is self-ad-
It was found that it was sufficient to take the first fifty
joint in the inner product when the velocities in the two
terms in the summation in the above solution to obtain
fluids are equal i.e., v1 =v2
convergence. This implies that the eigen-value problem
h 1
(14) is solved for the first 50 roots. Care must be taken to
f n , f m f1n f1m dx K f 2 n f 2 m dx (14) ensure that no roots are missed and no roots are calcu-
0 h
lated more than once. This analytical solution is used to
The eigen functions are normalized with respect to this validate the numerical code based on the method of lines
inner product and the constants bn, dn are obtained as with a second order finite difference scheme in the

Copyright 2012 SciRes. ACES

314 B. MALENGIER ET AL.

transverse direction (x). The numerical method was used

to determine the concentration profiles in the laminar
flow regime.
In Figure 3 we show how the analytical solution for
the co-current plug flow based on the separation of vari-
ables (solid line) validates the predictions of the numeri-
cal code (points) based on finite differences. It is seen
that both the average concentration as well as the con-
centration at a fixed y using the two approaches agree
quantitatively.
The numerical code was then used to simulate the be-
havior for the laminar flow profile in co-current mode.
Here the velocity profile obtained in Equation (3) is used
to simulate the laminar behavior. The comparison of the
cup-averaged concentration profiles obtained using the
laminar flow and the plug flow behavior in a micro-
channel is shown in Figure 4. For the plug flow simula- Figure 4. Comparison of the laminar flow versus PFR with
tion the average velocity of the laminar flow is used. It is fixed Q1 and Q2. Full line is laminar flow with h = 0.42 H, K
seen that the extraction performance under laminar flow = 1/4.14, dashed line PFR with same interface position.
conditions is superior to that of the plug flow conditions.
The numerical algorithm we use for solving the
5. Numerical Solution Counter-Current steady-state convection diffusion equation for extraction
Flow under counter-current flow is now described:
1) The channel length is divided into Ny grids in the
For the counter-current plug flow the convective-diffu- y direction. The values of the solute concentrations at
sion equations are solved numerically. Two challenges the interface on the fluid1 side are assumed.
arise in this and need to be addressed. These are, 1) the 2) The values of the concentrations at the interface on
jump discontinuity in concentrations at the interface and the fluid 2 side are obtained using the equilibrium condi-
2) the inlet of the two fluid streams being at the two end tion.
points. The latter renders the system a boundary value 3) Now the convection diffusion equation in each fluid
problem. is solved using the method of lines. This is possible as
we have a Dirichlet boundary condition at one end (the
interface) and a Neuman condition at the other end (wall)
with known inlet conditions. Here a second order scheme
is used to discretise the equations in the transverse direc-
tion and the equations are integrated along the axial di-
rection.
4) After the solutions are obtained the fluxes at the Ny
grid points are calculated in each fluid. The difference in
the fluxes at the interface has to be zero. This condition
is used to iterate on the concentrations at the interface on
fluid 1 till convergence is achieved using a Newton-
Raphson technique.
The above algorithm is implemented in Matlab. The
cup-mixed average concentration profile along the axis
obtained using the above algorithm is shown in Figure 5.

Figure 3. Comparison of the analytical solution and the 6. Simplified Model for Co-Current and
numerical computation for co-current plug-flow. Here H =
4 104 m, h = 2 104 m, Q1 = Q2 = 14.2857 106 m2/s, K
Counter-Current Flow
= 4.14, D1 = 7.4 109 m2/s, D2 = 5.64 108 m2/s. Full line is To obtain a quick physical insight into the behavior ob-
the average concentration for x > h and the dashed line the tained in the two flow-regimes of co-current and counter-
concentration at x = 0.75 H from the analytical computation.
The triangles are the corresponding computed values with
current flow, a simplified model is proposed in this Sec-
the numerical computation, where a grid with 100 points in tion. It is valid under the assumptions of a very small
the transverse direction per phase is considered. height H of the channel (as prevailing in micro-channels),

Copyright 2012 SciRes. ACES

 B. MALENGIER ET AL. 315

dc1
v1h = kl c1 Kc2 (20a)
dy
dc2
v2 H h kl c1 Kc2 (20b)
dy
with initial condition
C1 y 0 1, C2 y L 0,
Introducing kl1 and kl2 as before, the solution is
given by
kl1 Kkl2 kl1 y
C1 y e 1 1 (21a)
kl1 Kkl2 e
l l
Kk 2
k 1
L

kl2 Kkl2 kl1 y Kkl2 kl1 L

C2 y e e (21b)
Figure 5. Concentration profiles of counter-current flow for kl1 Kkl2 e
Kkl2 kl1 L
L = 0.044 m, H = 0.0004 m, Q1 = 14.2857 106 m2/s, Q2 =
2Q1, h = H/3, D1 = 7.4 108 m2/s, D2 = D1/2, K = 1/4.14. We now describe how the mass transfer coefficient kl
can be estimated for a system experimentally.
and large diffusion coefficients Di. Under these condi- For Co-current flow, the simplified Equations (18a)
tions the concentration variation in the direction trans- and (18b) can be rearranged to yield
verse to the flow can be neglected and the evolution of d c1 Kc2 1
K
the average concentration along the axial direction is kl c1 Kc2 (22a)
governed by ordinary differential equations. For sim- dy v1h v2 H h
plicity we assume the velocity profile to follow plug
c Kc2 1 K
flow. ln in1 kl y (22b)
in
c1 Kc2 v1h v2 H h
6.1. Co-Current Flow
The rate at which mass is transferred when the con-
The simplified equations of mass balance are now given centration drops to c1 or c2 in the system is
by
dc
m v1h c1in c1 (23a)
v1h 1 kl c1 Kc2 (18a)
dy m v2 H h c2 c2in (23b)
dc2 Using these equations, we obtain
v2 H h kl c1 Kc2 (18b)
dy 1
K
with y 0, L and initial condition c
c1 K c2 c2in m
in

v1h v2 H h
1

C1 y 0 1, C2 y 0 0 Rearranging and eliminating the terms containing the

Here kl represents an overall mass transfer coeffi- interface position h using (22b) we obtain
cient (between the two phases). We now define
in
kl 2 kl
m kl y

c1 Kc2 c1 Kc2
in
(24)
kl1 , kl
v1h v2 H h
c1in Kc2in

ln
c1 Kc2

The solution to the above two equations is given by
Kkl 2 kl1 kl1 Kkl 2 y
m
C1 y = At y = L, the exit kl =
kl Kkl
1 2
kl Kkl
1 2
e (19a) L clm
The logarithmic mean concentration difference is de-
k2 k2 kl1 Kkl 2 y
C2 y = 1 l 2 1 l 2 e (19b) fined as
kl Kkl kl Kkl
clm

c1in Kc2in c1out Kc2out
(25a)

cin Kc in
6.2. Counter-Current Flow ln out
1 2
The simplified equations are now given by
c1 Kc2out

Copyright 2012 SciRes. ACES

316 B. MALENGIER ET AL.

(a) (b)

(c) (d)

(e)

Figure 6. Co-current and counter-current extraction for L = 0.0044 m, H = 0.0004 m, Q1 = 14.2857 106, Q2 = 2Q1, h = H/3, K
= 1/4.14, D1 = 4 7.4 108 m2/s, D2 = D1/2. (a), (b), (c), (d) Concentration profiles for L = 0.001, 0.002, 0.003, 0.0044 m, full line
and dash-dot counter-current, dashed and dot co-current; (e) Extraction rate, full line counter-current, dashed co-current.

Copyright 2012 SciRes. ACES

 B. MALENGIER ET AL. 317

In a similar manner, kl can be calculated for counter-

current flow, using the simplified Equations (20a) and
(20b). Following the procedure for co-current flow it can
be seen that
m
kl = (25b)
L clm
The mass transfer coefficient kl can also be defined
using the driving force for extraction to be the deviation
from equilibrium value, see Dessimoz [1]. This gives the
mass transfer coefficient as
1 c1eq c1in
kl a = log eq out
(26)
t c1 c1
(a)

7. Results and Discussion

In order to compare the results of our simulations and to
be consistent with the literature, and evaluate the per-
formance of a specific micro-channel set-up, we intro-
duce some characteristic quantities. These are now de-
fined.

Characteristic Quantities
The first is the efficiency E, defined in terms of the
mixed cup concentrations as
C2,out C2,in
E (27)
C2eq C2,in
where C2eq is the concentration of the solute in the sec- (b)
ond region after equilibrium is attained, and typically
C2,in = 0 . E is a measure of how close the exiting
stream is to equilibrium. The overall residence time tres
for co-current flow is defined as
HL
tres = (28)
Q1 Q2
For a given length L, a unique residence time and an
extraction efficiency E(L) is obtained. E = 1 corresponds
to the situation when the exiting streams are in equilib-
rium and no further separation can take place.
The second characteristic which can describe the sys-
tem is the extraction ratio Er. It represents the fraction of
the amount of solute that has been fed to the system
(c)
which is removed by the second fluid. This is defined as
Figure 7. Co-current and counter-current extraction for L =
C2, out Q2 0.0044 m, H = 0.0004 m, Q1 = 14.2857 106, Q2 = 2Q1, h =
Er = (29)
C in Q1 H/3, K = 1/4.14 and large diffusion coefficient D1 = 50 7.4
108 m2/s, D2 = D1/2. Match with same interface height. (a)
Note that for co-current flow Concentration profiles co-current; (b) Concentration pro-
files counter-current, here full line and dash-dot are the 2D
Q1 Q2 numerical computation, and dashed and dot the lumped
E Er K (30)
Q2 Q1 model with matched kl value; (c) Using zero height ap-
proximation for L = 0.044 m, Extraction rate, full line
Figure 6 shows the dependency of concentration along counter-current, dashed co-current.

Copyright 2012 SciRes. ACES

318 B. MALENGIER ET AL.

the length as obtained by solving the convective diffu- is necessary to simulate the behavior for each length and
sion equation for co-current and counter-current extrac- then compute the extraction ratio, since the problem is
tion. Figures 6(a) to (d) shows concentration profiles for now a two point boundary value problem.
different channel lengths. For the choice of parameters it Figure 7 shows the performance of the co-current and
is seen that the extraction is marginally better for the counter-current systems for a different set of parameters.
counter-current operation. It is seen that as we increase The diffusion coefficient has been significantly increased
the channel length the improvement in performance be- here. Consequently here the simplified model is expected
comes more significant. This is verified in Figure 6(e) to give a more accurate representation of the system.
where the extraction ratio of the two models of operation Here again the concentration profiles and the extraction
is compared. In this last curve where we compare the ratios are depicted. For the concentration profiles two
extraction ratio, the curve for the co-current behavior can curves are shown: one is the prediction of the convective
be obtained by a single simulation of the governing diffusion equation and the other is the concentration pro-
equations. However for the counter-current extraction it file based on the simplified model. Here the parameter kl

(a) (b)

(c) (d)

Figure 8. Comparison of PFR numerical simulation (full line) and analytical solution (dotted line) using simplified expression
with kl from logarithmic mean concentration difference for L = 0.044 m, H = 0.0004 m, h = H/3, Q1 = 14.2857 106 m2/s, Q2 =
2Q1, K = 1/4.14 and D1 = 7.4 108 m2/s; D2 = D1/2, (a) Co-current concentration profiles, kl = 7.1976 104 m/s; (b)
Counter-current concentration profiles, kl = 9.9793 104 m/s; (c) With large diffusion coefficient D1 = 4 7.4 108 m2/s,
co-current concentration profiles, kl = 0.0022 m/s; (d) Comparison of kl (Clm ),full line and kl (eqbm), dotted line for
co-current flow.

Copyright 2012 SciRes. ACES

 B. MALENGIER ET AL. 319

is estimated by minimizing the least squares error of the

exit concentration from the two models. It can be seen
that the prediction by the convective diffusion model and
the simplified model agree well throughout the length of
the curve.
We would like to avoid the least squares estimation of
the mass transfer coefficient since this is a statistical
procedure and involves fitting a parameter. For this the
concentration profiles are determined solving the con-
vective diffusion equation. The exit concentration results
are used as experimental results and the log mean con-
centration difference is used to determine the mass
transfer coefficient as described earlier. The predictions
of the concentration profiles of the convective diffusion
(a)
equation and the simplified model are shown in Figures
8(a) and 8(b). It is seen that the concentration profiles are
in better agreement for the counter-current flow as op-
posed to the co-current flow. The prediction of the sim-
plified model using the log mean concentration differ-
ence is better when the diffusion coefficient is higher and
this is seen when we compare Figure 8(c) with Figure
8(a). Figure 8(d) compares the mass transfer coefficient
predicted by the log mean concentration difference and
the deviation from equilibrium (Equation (28)). We see
that the mass transfer coefficient at short lengths of the
channels are high. This drops down sharply first and then
slowly as the length is increased. This is to be expected
since for short lengths the driving force is high and this
results in a larger mass transfer coefficient. As the length
increases the driving force decreases as we approach (b)
equilibrium and this lowers the mass transfer coefficient.
Figure 9(a) shows comparison of co-current concen-
tration profiles obtained from the convection diffusion
model and the simplified model. Here the mass transfer
coefficient is calculated using the driving force as the
deviation from the equilibrium concentration Equation
(28). Here again when the diffusivity is high we see a
much better match between the axial profiles (Figure
9(b)). At the same higher diffusivity, profiles using mass
transfer coefficient calculated from the deviation from
equilibrium matches better than those obtained using the
log mean concentration difference. (compare Figure 9(b)
and 8(c)). Figure 9(c) shows a comparison of kl values
calculated using the above two approaches. It is observed
that mass transfer coefficients from Equation (28) are (c)
much higher than that using log mean difference at all
Figure 9. Comparison of PFR numerical simulation (full line)
the lengths of the channel. This difference is very high at and analytical solution (dotted line) using simplified expres-
the shorter lengths and it decreases as the length in- sion with kl from equilibrium concentration values for L =
creases as observed in Figure 8(d). 0.044 m, H = 0.0004 m, h = H/3, Q1 = 14.2857 106 m2/s, Q2
= 2Q1, K = 1/4.14 and D1 = 7.4 108 m2/s; D2 = D1/2, a =
8. Conclusions 5000 m2/m3 (a) Co-current concentration profiles, kl = 1.2
103 m/s; (b) with high diffusivity, D1 = 4 7.4 108 m2/s,
In this work we have compared the performance of the co-current concentration profiles, kl = 3.5 103 m/s; (c)
co-current and countercurrent modes of operation in an comparison of kl (Clm), full line and kl (eqbm), dotted line.

Copyright 2012 SciRes. ACES

320 B. MALENGIER ET AL.

extraction in a micro channel. This theoretical study ex- doi:10.1016/j.jcis.2004.10.011

amines the convective diffusion equation as well as a [5] C. Wang, Y. Gao, N.-T. Nguyen, T. Wong, C. Yang and
simplified model. An analytical solution is proposed for K. Ooi, Interface Control of Pressure-Driven Two-Fluid
the co-current system based on the method of separation Flow in Microchannels Using Electro-Osmosis, Journal
of Micromechanics and Microengineering, Vol. 15, 2005,
of variables. The counter-current system is solved nu-
pp. 2289-2297. doi:10.1088/0960-1317/15/12/011
merically. For this a novel method which exploits the
feature of the system is proposed. It is based on iterating [6] D. Fries, T. Voitl and P. von Rohr, Liquid Extraction of
Vanillin in Rectangular Microreactors, Chemical Engi-
on the interface values of the concentration. The study is neering & Technology, Vol. 31, No. 8, 2008, pp. 1182-
numerical and shows how concepts from heat exchanger 1187. doi:10.1002/ceat.200800169
design can be extended to mass transfer systems. It is [7] Y. Okubo, T. Maki, N. Aoki, T. H. Khoo, Y. Ohmukai
shown that the counter-current extraction is more effi- and K. Mae, Liquid-Liquid Extraction for Efficient Syn-
cient than the co-current operation. It is found that the thesis and Separation by Utilizing Micro Spaces, Che-
different definitions of the mass transfer coefficient pre- mical Engineering Science, Vol. 63, No. 16, 2008, pp.
dict the behavior of the system accurately. 4070-4077. doi:10.1016/j.ces.2008.05.017
[8] P. Znidarsic Plazl and I. Plazl, Steroid Extraction in a
9. Acknowledgements Microchannel System-Mathematical Modelling and Ex-
periments, Lab on Chip, Vol. 7, 2007, pp. 883-889.
The authors thank Roger Van Keer, coordinator of the doi:10.1039/b704432a
research group NaM2, for his continuous encouragement. [9] W. E. TeGrotenhuis, R. J. Cameron, M. G. Butcher and P.
The second author thanks the MaCKIE project (BOF- M. W. R. S. Martin, Microchannel Devices for Efficient
GOA 01GA0405) of Ghent University which made the Contacting of Liquids in Solvent Extraction, Separation
collaboration possible. Science and Technology, Vol. 34, 1999, pp. 951-974.
doi:10.1081/SS-100100691
[10] A. Aota, M. Nonaka, A. Hibara and T. Kitamori, Coun-
REFERENCES tercurrent Laminar Microflow for Highly Efficient Sol-
[1] A.-L. Dessimoz, L. Cavin, A. Renken and L. Kiwi-Min- vent Extraction, Angewandte Chemie, Vol. 119, No. 6,
sker, Liquid-Liquid Two-Phase Flow Patterns and Mass 2007, pp. 896-898. doi:10.1002/ange.200600122
Transfer Characteristics in Rectangular Micro-Reactors, [11] A. Hibara, M. Nonaka, H. Hisamoto, K. Uchiyama, Y.
Chemical Engineering Science, Vol. 63, No. 16, 2008, pp. Kikutani, M. Tokeshi and T. Kitamori, Stabilization of
4035-4044. doi:10.1016/j.ces.2008.05.005 Liquid Interface and Control of Two-Phase Confluence
[2] J. Guo and C. Ho, Theoretical Study on Membrane Ex- and Separation in Glass Microchips by Utilizing Octade-
traction of Cu2+ with d2ehpa in Laminar Flow Circular cylsilane Modification of Microchannels, Analytical
Tube Modules, Desalination, Vol. 233, No. 1-3, 2008, Chemistry, Vol. 74, No. 7, 2002, pp. 1724-1728.
pp. 247-257. doi:10.1016/j.desal.2007.09.049 doi:10.1021/ac011038c
[3] M. Liu, Y. Linu, Q. Guo and J. Yang, Modeling of Elec- [12] A. Aota, H. Akihide, S. Kyosuke, S. Yasuhiko, O. Koji
trosomotic Pumping of Nonconducting Liquids and Bio and T. Kitamori, Flow Velocity Profile of Micro Coun-
Fluids by a Two-Phase Flow Method, Journal of Elec- ter-Current Flows, Analytical Science, Vol. 23, No. 2,
troanalytical Chemistry, Vol. 636, No. 1-2, 2009, pp. 2007, pp. 131-133. doi:10.2116/analsci.23.131
86-92. doi:10.1016/j.jelechem.2009.09.015 [13] D. Ramakrishna and N. R. Amundson, Transport in
[4] Y. Gao, T. Wong, C. Yang and K. Ooi, Two Fluid Elec- Composite Materials: Reduction to a Self-Adjoint For-
tro Osmotic Flow in Microchannels, Journal of Colloid malism, Chemical Engineering Science, Vol. 29, No. 6,
and Interface Science, Vol. 284, No. 1, 2005, pp. 306-314. 1974, pp. 1457-1464. doi:10.1016/0009-2509(74)80170-3

Copyright 2012 SciRes. ACES