MATHEMATICAL

COMPUTER
MODELLING
PERGAMON Mathematical and Computer Modelling 32 (2000) 465-480
www.elsevier.nl/locate/mcm

Mass Transfer in
Supported Liquid Membranes:
A Rigorous Model
A. ALHUSSEINI*AND A. AJBAR
Department of Chemical Engineering, King Saud University
P.O. Box 800, Riyadh 11421, Saudi Arabia
amalik@ksu.edu.sa

(Received December 1998; revised and accepted November 1999)

Abstract-A rigorous steady-state model for the facilitated solute transport across ceramic-tube
supported liquid membranes is presented. The proposed permeation model accounts for the effect
of the variation of the distribution coefficient, ss well as nonconstant solute concentration in the
strip phase and nonequilibrium interfacial reactions. The resulting convection-diffusion transport
equations form a set of boundary-coupled partial differential equations. As a case system taken from
hydrometallurgy, the extraction of copper using ceramic tubes impregnated with LIX-84 as a carrier
is considered. It is shown that the variation of the distribution coefficient is a factor that must be
considered when designing or analyzing the performance of supported liquid membrane modules. The
model is also used to examine the effect of key operating conditions on the module’s performance.
@ 2000 Elsevier Science Ltd. All rights reserved.

Keywords-Membranes, Supported, Liquid, Tubular, Modeling.

NOMENCLATURE

bj Dimensionless concentration of cH,j’ Concentration of hydrogen ions

the carrier in the interface of the in the interface of the j-stream
j-stream (equations (33),(37)) ( mol/m3)

Bl Dimensionless model parameter cH,j,c Initial concentration of hydrogen

(equation (43)) ion in the j-stream (mol/m3)

B2 Dimensionless model parameter cMM,j Concentration of metal ions in the

(equation (43)) j-stream (mol/m3)

83 Dimensionless model parameter cMM,j,o Initial concentration of metal ion in

(equation (43)) the j-stream (mol/m3)

B4 Dimensionless model parameter cM,j * Concentration of metal ions in the

(equation (43)) interface of j-stream (mol/m3)

cH,j
Concentration of hydrogen ions in CRH Concentration of the carrier
the j-stream (mol/m3) ( mol/m3)

*Author to whom all correspondence should be addressed.

0895-7177/00/$ - see front matter @ 2000 Elsevier Science Ltd. All rights reserved. Typeset by .AA&-W
PII: SO895-7177(00)00145-X
 466 A. ALHWEINI AND A. AJBAR

CRH,O Initial concentration of the carrier X Dimensionless concentration of

(mol/m3) metal ion in feed stream (equation
,._-\.
cRH,j l
Concentration of the free carrier (J5))
in the interface of the j-stream f a” Dimensionless mixing cup concen-
(mol/m3) tration of metal ion in the feed
Concentration of metal complex stream (equation (44))
CRM
(mol/m3) 1/ Dimensionless radial distance
(equation (27))
d Geometry parameter (equation
(42)) Y Dimensionless concentration of
Diffusive coefficient of metal M metal ion in the strip stream
DM
(equation (25))
BM Distribution coefficient of metal M
Dimensionless mixing cup concen-
(equation (20))
tration of metal ion in the strip
DRM Diffusion coefficient of metal carrier stream (equation (45))
DRH Diffusive coefficient of complex z Axial distance (m)
gj
Dimensionless concentrations of Greek Symbols
the complex at the interface of the
Metal-to-carrier ratio (equation
j-stream (equation (26))
(42))
HF Dimensionless concentrations of
Dimensionless model parameter
hydrogen ions in the feed stream
(equation (42))
(equation (25))
Dimensionless initial concentrations Applied pressure drop (N/m2)
HF,O
(equation (24))
of hydrogen ions in the feed stream
Dimensionless concentration of Porosity of the ceramic tube
HS
hydrogen ions in the strip side Dimensionless parameter (equation
(equation (25)) (42))
Hs,o Dimensionless initial concentrations Geometry parameter of the module
of hydrogen ions in the strip side (Figure 1)

K=l Equilibrium extraction constant Geometry parameter of the module

(equation (19)) (Figure 1)

kf Hate constant of the forward viscosity of the feed stream

extraction reaction (kg/m.s)
kb Hate constant of the backward Dimensionless axial distance
extraction reaction (equation (27))
L Contactor length Turtosity of the ceramic tube
T Hadial distance (m) Subscripts
R Radius of the module (m) 0 Feed condition
RE Hate of extraction reaction j Feed (F) or strip (S)
(mol/m2s)
F Feed stream
Rs Hate of strip reaction (mol/m%)
H Hydrogen ion
“j Fluid velocity in the j-stream
M Metal ion
(m/s)
RH Free carrier
aj
Average fluid velocity in the
j-stream (m/s) RM Complex of metal ion M
s Strip stream
* Interfacial conditions

INTRODUCTION
The increasingly stringent environmental regulations over the discharge of hazardous species
have driven the process industry to seek new and cost-effective technologies for treatment of
toxic wastes. Supported liquid membrane (SLM) processes are being recognized as promising
separation technologies for the selective separation and concentration of valuable or toxic metals
from dilute aqueous solutions [l-6]. In this process, the membrane is often formed by immobilizing
a suitable extractant in a porous polymeric support. The SLM is then used to physically partition
 Supported Liquid Membranes 467

the stream containing the metal to be removed from the strip stream; thus, the processes of metal
ion extraction, stripping, and extractant regeneration are simultaneously combined in a single
contactor. The liquid membrane process can, therefore, be considered as an advanced variant of
the traditional solvent extraction process. However, the particular arrangement in SLM eliminates
the equiiibrium iimitation inherent to soivent extraction and makes it economicaiiy feasibie to
use tailor-made expensive extractants because of the small extractant hold up requirements.
In spite of these advantages, there are no major industrial applications of supported liquid
membranes. This is largely due to stability problems that cause a gradual loss of the membrane
phase making the useful lifetime of the module far too short to ensure reliable operation (7-91.
In a recent experimental study, Yi and Tavlarides [lo] showed that it was possible to make fairly
stable SLMs using ceramic disks as supports. However, planar membrane geometry is of little
practical use since it is difficult to build a module with sufficient mass transfer area per volume.
A shell-and-tube configuration represents, on the other hand, a viable solution since it can offer
a larger mass transfer area.
Modeling mass transfer in tubular modules is important for design, optimization, and control
purposes. Numerous models have been developed which can be classified into two major groups
according to how they describe mass transfer to the membrane. The majority of the developed
models ‘assume plug flow in the feed and strip streams, and do lump mass transfer effects into a
film-type mass transfer coefficient [9,11-141. These models are simple to solve and are particularly
useful if mass transfer parameters are to be estimated from experimental data [15,16]. The
drawback of these models is that mass transfer coefficients must be known in order to obtain
predictions,
The second class of models is based on the more general approach of solving the convection-
diffusion equations in the feed, membrane, and strip phases. Such models have been used by a
number of investigators for the limiting cases of equilibrium reactions [17] and negligible solute
concentration in the strip stream [18,19]. However, these limiting cases are rather restrictive.
Experimental results by Juang and Jiang [20] and Marchese et al. [21], for instance, showed
that the interfacial chemical reactions could appreciably contribute to the total mass transfer
resistance. In addition, the solute concentration in the strip stream is not always negligible
especially for the practicaiiy important case of counter-transport of ions (an important class of
Type-II facilitated transport) where one of the objectives is to concentrate a target ion in the
strip stream. In this case, the equations governing the concentrations in the feed, membrane,
and strip phases are coupled by the interfacial boundary conditions, and the model is a coupled
boundary problem.
Aware of these shortcomings, Yi and Tavlarides [22] proposed a modified model that accounts
for nonequilibrium effects and for the presence of the solute in the strip phase. However, the
model assumes constant pH in the feed and strip streams, and excess carrier. These assumptions
restrict the model to cases in which the distribution coefficient is constant. But as was shown
by Urtiaga et al. [23] for Type-I facilitated transport, the effect of variation in the distribution
coefficient is a factor that must be considered when predicting the performance of supported liquid
membranes. This effect is expected to be even more pronounced in counter-transport systems
where the distribution coefficient depends in a nonlinear form on the carrier and hydrogen ion
concentrations.
In this paper, we treat the problem of counter-transport in tubular modules by using a rigorous
permeation model that accounts for variations in distribution coefficient. As a case system, the
extraction of copper from acidic stream by ceramic tubes impregnated with LIX-84 is studied.
This ketoxime carrier is a mixture of predominantly 2-hydroxy-5 nonylacto-phenone oxime with a
small amount of 5-dodecylsalicylal doxime in a high flash point diluent. Predictions are generated
for various conditions and compared with the special case of constant distribution coefficient.
The effect of various operating conditions on the performance of the module is simulated and
discussed.
 A. ALHUSSEINI AND A. AJBAR
,............,.......

I Strip
Stream
Feed !
Stream !
’ J
CH,S !,H_

1 L-J-HkiR:
Figure 1. Schematic diagram of the permeation mechanism across the supported
liquid membrane.

PERMEATION PROCESS DESCRIPTION

A schematic diagram of the shell and tube membrane module under study is shown in Figure 1.
The feed stream flows inside the tubes concurrently with a strip stream that flows in the shell
side. As the two streams flow along the membrane module, they exchange ions according to
the permeation mechanism depicted in the figure. The ion exchange process is facilitated by an
active carrier immobilized in the pores of the ceramic tube. The permeation mechanism consists
following steps.
Diffusion of metal ions from the bulk of the feed stream to the inner surface of the .tube.
Diffusion of hydrogen ions from the inner surface of the tube to the bulk of the feed stream.
Upon reaching the membranefeed stream interface, the metal ions, assumed to be divalent
in this work, complex with the carrier according t.o the overall reaction

M2+ + 2HR w MR2 + 2H+. (1)

Diffusion of the complex from the inner to the outer surface of the tube.
Upon reaching the membrane-strip stream interface, the strip reaction takes place, regen-
erating the carrier and liberating the metal ions.
Diffusion of the regenerated carrier back to the inner surface of the tube.
Diffusion of the liberated metal ions from the outer surface of the tube to the strip stream
bulk.
Diffusion of hydrogen ions from the bulk of the strip stream to the outer surface of the
tubes.
 Supported Liquid Membranes 469

MODEL DEVELOPMENT
In developing the steady state permeation model, a number of simplifying assumptions are
introduced below:
l fully developed laminar flow;
a Fickian diffusion;
l negligible axial dispersion in the feed and strip stream;
l negligible axial diffusion in the membrane phase.
A mathematical presentation of the permeation process is given by the following mass conserva-
tion equations with the appropriate boundary conditions.

Tube-Side (Feed-Side) Mass Transfer (0 5 T < AR)

Balance on metal

(2)

Balance on hydrogen

!3?
Subject to the inlet conditions,

‘?M,&o = ‘%,F,‘J and CH,Fl,,o = cH,F& (4)

where the subscript (0) denotes initial conditions. The equations are also subject to the following
four boundary conditions:

acM,F =HF
=A =o, (5)
& r=O ar r=o

(6)

The first boundary conditions (equation (5)) are symmetry conditions. The second boundary
conditions (equation (6)) indicate that the metal diffusive flux and half the diffusive flux of
hydrogen are equal to the extraction reaction which takes place at the pores mouth only.

Membrane Phase Mass Transfer (AR 5 T _< rcR)

Balance on the complex

Balance on the free carrier

 470 A. ALHWSEINIAND A. AJBAR

subject to the following four boundary conditions:

f acRM ~5acRH
-DRM; - = ill
2 RH-- = ERE, (9)
dr T-AR 7 dr T=XR

f acRM 6 =RH
~DRH -- = cRs. (10)
-DRM7 dr r=nR= 2 7 ar r=tc,R

The four boundary conditions indicate that the diffusive flux at the interfaces are equal to the
reaction rates.

Shell-Side (Strip-Side) Mass Transfer (& 5 r < R)

Balance on metal

(11)

Balance on hydrogen

acH S
US(r)* = D”f$ (12)
Subject to the inlet conditions,

c~,sI,=o = c&W and cH,SI,,o = cH,S,O (13)

and the following boundary conditions:

dCM,S _ dCH,S = 0,
04

= cRS. (15)
r=RR

The four boundary conditions are similar to those of the tube side.
In the above equations, cM,j and cH,j denote the metal and hydrogen ions concentrations in
the j-stream (J’ = F for the feed and j = S for the strip). CRM and CRH are the concentrations
of complex and carrier in the membrane phase, respectively. DM, DH, DRM, and DRH are the
diffusion coefficients of the metal ion, hydrogen ion, complex, and carrier, respectively. RE refers
to the rate of extraction reaction taking place at the tube side of the membrane while Rs is the
rate of the strip reaction taking place at strip side.
In the present study, we consider copper transport through a liquid membrane made of LIX-84
8s a carrier and n-heptane as a diluent. The extraction reaction proceeds as follows (24,251:

Cu++ + 2HR t+ CuRz + 2H+. (1’3)

According to Komasawa et al. [13] and DeHaan et al. [ll], the rate of the extraction reaction is
given by
CM,F* CRH,F’ CRM,F’CH,F’
RE=kj - h (17)
CH,F’ CRH,F’ ’
where the symbol (*) denotes interfacial conditions.
 Supported Liquid Membranes 471

The rate Rs for the strip reaction is given by the same authors as

RS = k&RM,S’CH,S’ - kf
cnn,s*
Cl&S’
cH,S’ *

The equilibrium constant Kes for this system is given by

K
eq
= CR&f (19)
CM c:H

The distribution coefficient Dn/l is defined as

(20)
As was mentioned in the introduction, a number of authors have used simplified rate expressions to
either analyze limiting cases or to obtain analytical solutions for the model. Yi and Tavlarides [22],
for instance, simplified the extraction and stripping reaction rates to the linear form,
RF = k;CM,p - kiCRM,p (21)
and
Rs = k;&,+_i,s* - k;cM,p. (22)
From the rate expressions (equations (17),(18)), it can be seen that this case corresponds to the
reaction taking place at high carrier-to-metal concentration ratio with negligible change in pH.
This case is equivalent to setting the distribution coefficient D)M (equation (20)) constant and
equal to its value at the entrance. While this assumption simplifies the resulting mathematical
problem, it is rather restrictive and may not hold as demonstrated by the work of Urtiaga et
al. (231.
Finally, an important note should be made about the useful life of the membrane. In order
to prevent membrane expelling from the pores, the transmembrane pressure difference should be
minimized. This condition can be satisfied by the proper choice of the feed and strip stream flow
rates. ,To maintain near zero transmembrane pressure difference, the iniet pressures and axiai
pressure drops should be the same for both streams. Assuming that the inlet pressures are the
same, the local axial pressures can have the same value by adjusting the area average velocities
in the tube and shell sides to satisfy the ratio [22],
fiF x2
-=
(1+ G) + (1 - n2)/ln(l/6)’ (23)
%S

tiF = $AR)2. (24)

In deriving this relation, the shell side of the module was treated as an annular region. This
condition is imposed in all the calculations presented in this study.
It is more convenient to express the model equations in dimensionless form using the metal
concentration in the entering feed stream CM,F,~ and the initial carrier concentration CRH,~ as
reference concentrations. The different dimensionless concentrations are
CMF CMS
x= ---1 y=---- HF = CH,F Hs = cH,s
(25)
cM,F,O CM* F,O ’ CM,F,O ’ cM,F,O ’
and
gj _ CRMj’ , CRH j*
bj = -----, (j = F,S), (26)
CRH,O CRH,O

whereas the dimensionless radial and axial distances are defined as

y=$ +;. (27)

The governing equations in dimensionless form along with their boundary conditions are given
by the following.
 472 A. ALHUSSEINI AND A. AJBAR

Tube-Side (Feed-Side)

with
Xl&o = I and HF&=O = HF,O (30)

$I,=, = ~~,=, = 0 and $fivzA = -7 %I,=, = -EB& (31)

and

The dimensionless extraction reaction rate fin is given by

(32)

with
bF = 1 - 2&J’. (33)
The derivation of this equation is given in the Appendix.

Membrane

Integrating the dimensional equations (equations (7)-(10)) and putting them in dimensionless
form yields

SF -g.9
Xln(K/X)
= 7B2fiE, (34
gF--gS
= rB&. (35)
tcln(K/X)

The dimensionless strip reaction rate & is give by

b2
Rs = +sH; - Y;-s, (36)
eq Hi
with
bs = 1 - 2pgs. (37)

The derivation of this equation is given in the Appendix.

Shell-Side (Strip-Side)

1- y2+ h(y)) g = -$$(yg) , (38)

Y$(1 - y2+ din(y)) T (y$$),

=CH~$ (39)

with
Y(+, = yo= GCMSO, I and Hsl+,, = Hs,o = A
CHSo
cM,F,O
(40)
 Supported Liquid Membranes 473

(41)

The rest of the dimensionless parameters that appear in the model are given by

1- K2 CMFO
d=- cy=A p=- DRM EH=--,
DH
(42)
Zn( l/K) ’ CRH,O ’ DRH' DIVI

b l b B = kbKeqCRH,o B = kbKe&RH,o 1
" = (DM/R) &' B2 = (DRM/R)’ 3 (DRMIR) ’ 4 (DM/R) zy’ (43)

It can be seen that the governing equations are strongly coupled through the nonlinear interfacial
reactions which make them not amenable to analytical solutions.

NUMERICAL PROCEDURE AND

PRESENTATION TECHNIQUES
The resulting distributed parameter model was solved using the method of lines. A finite dif-
ference discretization was used in the radial direction and the resulting set of ordinary differential
equations at each point T were integrated using Dassl (261; a solver suitable for stiff algebraic-
differential equations. Grid independent solutions were obtained when the number of nodes in
the radial direction was about 50 or more. A tolerance of 10m6 was maintained in all calculations.
To evaluate the membrane performance, mixing cup concentrations were defined for both feed
and strip sides. The feed and strip mixing cups concentrations were defined as

fav= $ IAX (1- $) dy

and
J,’ Y (I- y2 + din(y)) y dy
U @Y = (45)
rl 17 _ g” + &n(g)) g&
Jn 11
From its definition, the strip mixing cup concentration u,, also represents the enrichment factor
for the module.

Model Parameters

The reaction kinetic rate and equilibrium parameters as well as the diffusion coefficients of the
L-l__- L-- #-i..l__ rnv1 _____11--
various species -were W%Kt?Il IIuIII QUIItL ,‘,I, mid aiT :istec: iii T&k 1. For ihe module wyeralJ11g

conditions, the following baseline case values were assumed: the metal-to-carrier ratio (o) was
taken 0.1, the dimensionless hydrogen feed concentration HF,O = 0.1, the dimensionless hydrogen
strip concentration Hs,c = 100 and a metal free strip steam was assumed, i.e., Yc = 0. These
values reflect conditions that are likely to be encountered when treating dilute streams. For
example, for a feed stream with a relatively low copper concentration of O.OOlgmol/l (about
65ppm), this would correspond to a slightly acidic feed stream (pH = 4), a highly acidic strip
stream (pH = l), and an initial carrier concentration of 0.01 gmol/l.

RESULTS AND DISCUSSION

One of the key points addressed in this study is to assess the validity of the constant distribution
coefficient assumption for Type-II facilitated transport in supported liquid membranes. Using
the baseline values cited before, Figure 2 shows the mixing cup concentration (faV) and (u,)
versus the dimensionless axial distance (E), for the two cases of constant and variable distribution
coefficient. The values of the constant distribution coefficients on the feed and strip sides are 5882
 474 A. ALHUSSEINI
AND A. AJBAR

Table 1. Nominal model parameters.

Parameter Value

Kes 1.7

kf ($> 4.510-E

7.210-6

1.4610-6

1.4610-”

610-5

e 0.42

7 1.73

a = 0.1 -

6
f av

0.4

0.2 L

0.0
t
Case1 -&-
Case2

-'
+

uA'
-y
--O.-
-_I

,'
-

,,*

I
,,,,;J(y

,I"

I
\ \
j

I
4

0
U av

0.001 0.01 0.1 1

Fieure
- -o--- 2.
-- Dimensionlm f 1 and
feed t,,W,
____._.__.__._____.~~ ~ __, concentrations vs. dimensionless
~~ strio1 (‘u,,)
length for the baseline case. Cases 1 and 2 refer to predictions with constant and
variable distribution coefficients, respectively.

and 0.0059, respectively. The variable distribution coefficients were calculated by equation (20)
using interfacial concentrations. As expected, the feed concentration (faV) decreases while the
enrichment factor (u,) increases with distance, for both cases. It can be seen that the predicted
average concentration for the variable distribution coefficient case is always smaller than that
for the constant distribution case, indicating a slower permeation rate. In fact, for the latter
case, an uphill transport occurs when the value of the dimensionless length is larger than 0.07,
whereas for the variable distribution case the uphill transport occurs when the dimensionless
length exceeds 0.1. Consequently, a membrane module designed without accounting for the
variation of the distribution coefficient would not be large enough to meet the desired level of
purification. For instance, removing 50 percent of the metal in the feed stream would require
a dimensionless length of around 0.15. However, the predicted length would only be 0.05 if the
assumption of constant distribution is applied, i.e., almost one third of that required.
 Supported Liquid Membranes 475

0.8 8

0.8 8
f U av
av
0.4 -Case2 +--- 4

Figure 3. Performance of the module for a low metal-to-carrier ratio (Y = 0.05.

Cases 1 and 2 refer to predictions with constant and variable distribution coefficients,
respectively.

0.8 8

0.8 8
f av U av
0.4 4

0.2 ,,’ / 2
t , -;;
.A
1
5
Figure 4. Performances of the module for a high hydrogen feed concentration HF,~
= 10. Cases 1 and 2 refer to predictions with constant and variable distribution
coefficients, respectively.

To further assess the validity of the constant distribution coefficient assumption, more analysis
was performed for conditions that promote smaller changes in the distribution coefficient. Such
conditions exist when excess carrier or highly acidic feed streams are encountered. In Figure 3,
the excess carrier condition is examined by decreasing the ratio of metal-to-carrier, to Q = 0.05.
As expected, the difference between the constant and variable distribution cases is smaller than
that for the baseline case. However, the difference is still large enough to warrant accounting for
the variation of the distribution coefficient. For instance, a 50 percent purification level would
require a dimensionless length of 0.05 for the constant distribution case against a real value of
0.07.
Figure 4 shows the predictions for a highly acidic feed stream case. This condition also promotes
small changes in the distribution coefficient. The metal-to-carrier ratio is kept constant at its
baseline value of Q = 0.1, while the hydrogen feed concentration is increased to HF,~ = 10,
i.e., 100 times its nominal value. It can be seen that, for both cases, the permeation rates
are considerably smaller than the baseline case. This is due to the suppression of the forward
extraction reaction by the high concentration of hydrogen ions in the feed stream. Figure 4 also
shows that the model predicts a smaller difference between the permeation rates of constant
 476 A. ALHUSSEINI
ANDA. AJBAR

f
av

k-_____--+--_---4
0.2 - /

,’
0.0 I I I ,_d-f,,,l I I I111111 I I ,““C
0.01 0.1 1 10 100

H FO

Figure 5. Rate determining steps ss function of hydrogen feed concentration HF,~.

,,I’
-- RDSrBLs

f av

_-______

0.2 -
/

,,,I, , ,( ,,,,,
0.0
0.01 0.1 1 IO

Figure 6. Rate determining steps ss function of metal-to-carrier ratio CL

and variable distribution coefficient, cases than the baseline case. However, the difference is still
significant. A 50 percent purification level, for instance, would require a module length of 0.7 for
constant distribution case against a value of 0.9 for the variable case.
The results presented in Figures 2-4 show the importance of accounting for the variations in
the distribution coefficient in the design or analysis of SLM modules. In the following sections,
the rigorous model is used to examine the effect, of key operating conditions on the module’s
performance.

Rate Determining Steps (RDS)

The permeation process of copper ions from the feed to strip stream consists of a number of
serial steps, as outlined in the theoretical part of this paper. Identifying which of these steps
is the rate determining step (RDS) is quite important for optimal design and operation of SLM
modules. In Figure 5, the mixing cup concentration (faV) at the exit of a module of length < = 0.1
is plotted versus the hydrogen feed concentration (HF,o), for the following four cases: permeation
 Supported Liquid Membranes 477

rate controlled by all steps (i.e., competitive case), membrane diffusion is RDS, interfacial reaction
is RDS, and boundary layers diffusion is RDS. As expected, the competitive case is the slowest of
the four cases. The membrane diffusion step appears to be RDS over a wide range of HF,a while
the resistance offered by the boundary layers is appreciable only for HF,~ less than 1. On the other
hand, the contribution of the interfacial reactions to the totai resistance becomes appreciable at
high values of hydrogen concentrations, i.e., Hp,a larger than 10, since at these large values the
extraction reaction is suppressed.
Figure 6 examines the effect of the metal-to-carrier ratio cr on the rate determining step. At
small values of (Y, (CY I 0.05), the carrier molecules are present in excess, which makes the
interfacial reactions and membrane diffusion proceed at a fast rate. The permeation process in
this region is, therefore, controlled by the resistance of the boundary layers. As o increases,
the permeation rate decreases since the membrane diffusion resistances become more important
as the available carrier molecules become progressively fewer. It can be seen that membrane
diffusion becomes RDS over a wide range of a (0.1 5 0 I 10).
The results presented in Figures 5 and 6 suggest that the most effective way to improve the
performance of the module is to minimize the membrane diffusion resistance, which can be
accomplished by a number of techniques. One such technique involves the minimization of the
diffusion path by using ceramic tubes with larger porosity or thinner walls. Another possible
approach is to use diluents with smaller viscosity. On the other hand, the use of turbulence
promoters to enhance mass transfer in the feed and strip stream is not recommended as it will,
at best, generate modest improvements in the permeation rate except when the metal-to-carrier
ratio is very small. Similarly, the use of another extractant with faster kinetics than LIX84 is
also not recommend especially if the pH of the feed stream is high.

Effect of Acid Concentration in the Stripping Solution

The concentration of hydrogen ions in the strip stream is an important variable that controls
the permeation driving force. The effect of this variable in dimensionless form (Hs,c) on the
nnrfnrma.nw
,~____‘________of
-- the module
1.-- ____ is shown in -Fkwrp
--_a --L _o-__, 7 _--
for ___-
metal-to-carrier
_..- 1.. __.__._I_ratios of 6:65 and Q;l, The
mixing-cup concentrations were calculated for a dimensionles distance value of c = 0.1. It can
be seen that the permeation of copper is negligible when Hs,s is smaller than unity (Hs,c I 1)
for both values of cr. In this region, the permeation process is controlled by the strip reaction,
which proceeds at a very slow rate since it is difficult to strip copper from the organic complex

1.0 10

0.8 8

0.8 8
f av U av
4

Figure 7. Effect of hydrogen concentration HS,O in the strip on the module perfor-
mances for low and high values of metal-to-carrier ratio a.
 478 A. ALHUSSEIN~
ANDA. AJBAR

1.0

_ anO. -@-- -0 -
0.6
a-0.05 -)_ * -

0.6
f IV - u av
0.4 -4

0.2

0.0
0.01 0.1 1 10 100
H FO
Figure 8. Effect of hydrogen concentration HF,,J in the feed on the module perfor-
mance for low and high values of metal-to-carrier ratio Q.

at such low values of hydrogen concentrations. As (Hs,o) increases beyond unity, copper is
more easily stripped, which reduces the resistance contributed by the strip reaction, resulting in
improved permeation rates. The permeation process becomes insensitive to HS,O when its value
exceeds 100. At such large values of Hs,o, the resistance offered by the strip reaction becomes
negligible and the permeation process is mainly controlled by diffusion across the membrane.
Hence, a significant improvement in performance can be achieved by decreasing the value of (Y,
i.e., increasing the carrier concentration.

Effect of Hydrogen Ion Concentration in the Feed Stream

‘I’he PI-I of the feed stream is an important parameter that is often dictated by the process
upstream of the membrane module. The effect of the hydrogen ion concentration in the feed
stream (HF,~) on the module performance for different metal-to-carrier ratios is shown in Figure 8.
The mixing-cup concentrations were calculated for a dimensionles distance value of 5 = 0.1. At
low HF,~ values, it can be seen that the module performance is almost insensitive to the feed
acidity. In this region, HF,~ is too small to affect the concentrations profile along the module.
This leads to length-averaged distribution coefficients that are virtually independent of HF,o.
As NF,O increases beyond the value of 1, the module performance progressively deteriorates due
to equilibrium limitations of the extraction reaction. In fact, at very high values of HF,~, the
extraction reaction is almost completely suppressed resulting in near zero removal of copper. It
should be pointed out that it is not possible to increase copper removal for highly acidic feed
streams by increasing the hydrogen concentration in the strip stream because of the equilibrium
limitations on the feed-side of the membrane. Therefore, it might be necessary to add buffers to
increase the pH of highly acidic feed streams to acceptable levels. Furthermore, the performance
of the module can be improved for any feed pH by increasing initial carrier concentration, i.e.,
lowering the value of (Y.

Effect of Copper Concentration in the Stripping Solution

One of the interesting features of Type-II facilitated transport is the ability to pump the target
solute against its concentration gradient and highly concentrate it in the strip stream. Therefore,
it is of interest to explore the effect of copper concentration in the entering strip stream on the
performance of the module. This is illustrated in Figure 9 showing the mixing cup concentration
faV versus the metal initial concentration in the strip stream Yo for HS,O values of 100 and 1000.
The mixing-cup concentrations were calculated for a dimensionles distance value of t = 0.1. The
 Supported Liquid Membranes 479

0 H&00
t 1
0.8
I H,,=lOOO
t

0.66 -I
f av
I

0.4 -

0.2 -

I I , , llltll I I , llltll I , ,I,,,,_

0.0
0.1 1 10 100 1000

YO
Figure 9. Effect of metal concentration Yo in the strip on the module performances
for low and high values of hydrogen strip concentration Hs,n.

ability of the module to move copper against its concentration gradient and enrich it in the strip
stream is demonstrated for cases in which Ys is greater than unity. Depending on the value of
Hs,s, copper permeation is unaffected by its concentration in the strip stream up to a certain
point where it begins to deteriorate because excess copper increases the concentration of the
complex on the strip-side of the membrane which reduces the permeation driving force. To get
highly concentrated copper at the maximum possible recovery, highly acidic strip streams should
be used as demonstrated for the case with Hs,0=1000, where the module can concentrate copper
up to 1000 times its concentration in the feed without any adverse effects on the permeation rate.

CONCLUSIONS

membrane modules. The permeation model takes into account the resistances offered by diffu-
sion across the boundary layers of the feed and strip stream, diffusion across the membrane phase,
and nonequilibrium extraction and strip reactions. In addition, the model allows the distribution
coefficient to vary with axial distance. An application of the model to the practically relevant
case of copper ion extraction using LIX-84 impregnated in porous ceramic tubes confirmed the
importance of accounting for the effect of the variation of the distribution coefficient. Neglecting
this effect would lead to undersized tubuiar moduie. it was shown that the permeation process
is mainly controlled by diffusion across the membrane phase. Under such conditions, the per-
formance of the module can be improved by increasing the feed stream pH, decreasing the strip
stream pH, or decreasing the metal-to-carrier ratio.

APPENDIX
DERIVATION OF EQUATIONS (33),(37)
The carrier and its copper complex are nearly insoluble in both aqueous feed and strip streams.
It is, therefore, conserved in the membrane phase and the mass balance equation yields,
acRH
DRH- dr + ODRM% = 0.

The integration yields

DRHCRH+~DRMCRM = COIlStaIlt = DRHCRH,~.

 480 A. ALHUSSEIN~AND A. AJBAR

Dividing by DRHCRH,Q yields

CRH DRMCRM
-+2--= 1,
cRH,O DRH CRH

or equivalently, using the dimensionless variables in equation (26),(42).

b+2/3g= 1.
Applying this equation to the feed and strip sides of the membrane yields

bi+ = 1 - 2@gF and bs = 1 - Zpg,.

REFERENCES
1. N.N. Li, Separating hydrocarbons with liquid membranes, U.S. Patent 3,410,749, (1968).
2. R.D. Noble and P.R. Dane& Liquid Membranes, Theory and Applications, lst Edition, p. 110, ACS Symp.
Ser. 347, Washington, (1987).
3. RD. Noble, C.A. Koval and J.J. Pallegrino, Facilitated transport membrane systems, Chem. Eng. Prog. 85
(3), 58, (1989).
4. J.D. Way, RD. Noble, T.M. Flynn and E.D. Sloan, Liquid membrane transport: A survey, J. Memb. Sci.
12 (2), 239, (1982).
5. J.S. Schultz, Synthetic Membranes: Science, Engineering, and Applications, (Edited by P.M. Bungay, H.K.
Lonsdale and M.N. DePinho), D. Reidel, Dordrecht, Netherlands, (1983).
6. Y.C. Huang and S.S. Koseoglu, Separation of heavy metals from industrial waste streams by membrane
separation technology, Waste Management 13, 481, (1993).
7. R.D. Noble, An overview of membrane separations, Sep. Sci. Tech. 22, 731, (1987).
8. M. Sugiura, Transport of lanthanide ions through cellulose triacetate membranes containing hinokitol and
flavanol as carriers, Sep. Sci. Tech. 26, (1990).
9. M. Teramoto and H. Tanimoto, Mechanism of copper permeation through hollow fiber liquid membranes,
Sep. Sci. Tech. 18, 871, (1983).
10. J. Yi and L.L. Tavlarides, Chemically active liquid membranes in inorganic supports for metal ion separations,
AIChE J. 38, 1957, (1992).
11. A.B. DeHaan, P.V. Bartels and J. DeGrauw, Extraction of metal ions from wastewater: Modeling of the
mass transfer in a supported liquid membrane process, J. Memb. Sci. 45, 281, (1989).
12. R.S. Juang, Permeation and separation of zinc and copper by supported liquid membranes using bis(2-
ethylhexyl) phosphoric acid ss a mobile carrier, Ind. Eng. Chem. Rea. 32, 911, (1993).
13. I. Kornasawa, T. Otake and T. Yamsshita, Mechanism and kinetics of copper permeation through a supported
liquid membrane containing a hydroxyoxime ss a mobile carrier, Ind. Eng. Chem. find. 22, 127, (1983).
14. R.S. Juang and R.H. Lo, Kinetics of the coupled transport of vanadium (IV) from sulfate solutions through
supported liquid membranes, Ind. Eng. Chem. find. 33, 1011, (1994).
15. R. Prasad and K.K. Sirkar, Dispersion-free solvent extraction with microporous hollow-fiber modules,
AIChE J. 34, 177, (1988):
16. C.H. Yun, R. Prssad, A.K. Guha and K.K. Sirkar, Hollow fiber solvent extraction removal of toxic heavy
metals from aqueous waste streams, Ind. Eng. Chem. Res. 32, 1186, (1993).
17. J.I. Kim and P. Stroeve, Mass transfer in separation devices with reactive hollow fibers, Chem. Eng. Sci. 43,
247, (1988).
18. E.N. Rudisill and M.D. Levan, Analytical approach to msss transfer in laminar flow in reactive hollow fibers
and membrane devices with nonlinear kinetics, Chem. Eng. Sci. 45, 2991, (1990).
19. A.M. Urtiaga, M.I. Ortiz, E. Salazar and J.A. Irabien, Supported liquid membranes for the separation-
concentration of phenol. 2. Mass-transfer evaluation according to fundamental equations, Ind. Eng. Chem.
Res. 31, 1745, (1992).
20. R.S. Juang and J.D. Jiang, Rate-controlling mechanism of cobalt transport through supported liquid mem-
branes containing di(2-ethylhexyl) phosphoric acid, Sep. Sci. Tech. 29, 223, (1994).
21. J. Marchese, E.C. Campderros and A. Costa, A mechanistic study of cobalt, nickel, and copper transfer
across a supported liquid membrane, J. Chem., Tech. Biotechnol. 67, 37, (1993).
22. J. Yi and L.L. Tavlarides, Modelling chemically sctive liquid membranes in tubular organic supports,
AIChE J. 41, 1403, (1995).
23. A.M. Urtiaga, J.A. Irabien and P. Stroeve, Effect of a variable solute distribution coefficient on msss separa-
tion in hollow fibres, Ind. Eng. Chem. Res. 31, 1362, (1992).
24. K. Lee, D.F. Evans and E.L. Cussler, Selective copper recovery with two types of liquid membranes, AIChE J.
24, 860, (1978).
25. D. Pearson, Supported liquid membranes for metal extraction from dilute solutions, In Ion Exchange Mem-
branes, (Edited by D.S. Flett), Chapter 4, Ellis Horwood Limited, Chichester, (1983).
26. L.R. Petzold, In Scientific Computing, (Edited by R.S. Stepleman), North-Holland, Amsterdam, (1983).
27. A.K. Guha, C.H. Yun, R. Bssu and K.K. Sirkar, Heavy metal removal and recovery by contained liquid
membrane permeator, AIChE J. 40, 1223, (1994).