Journal of Membrane Science 243 (2004) 177–187

Composite PDMS membrane with high flux for the separation

of organics from water by pervaporation
Lei Lia,∗ , Zeyi Xiaob , Shujuan Tana , Liang Pua , Zhibing Zhanga
a Department of Chemical Engineering, School of Chemistry and Chemical Engineering, Nanjing University, 210093 Nanjing, China
b Department of Chemical Engineering Machinery, Sichuan University, Chengdu 610065, China

Received 20 February 2004; accepted 12 June 2004

Available online 18 August 2004

Abstract

The composite polydimethylsiloxane (PDMS) membranes supported by Cellulose acetate (CA) were prepared by pre-wetting method
for the separation of methanol, ethanol, n-propanol and acetone from water. The experiments were carried out to investigate the effects of
operating parameters on the pervaporation performance and the results showed that the membranes exhibited high total permeation flux. A
resistance-in-series model was applied to analyze the transport of the permeants. It was found that boundary-layer mass transfer coefficient
was proportional to Re0.5 and increased exponentially with temperature. The membrane mass transfer coefficient conformed to Arrhenius
correlation and was independent of flow status. The estimated diffusivities of alcohols in membrane had a magnitude order of 10−10 m2 ·s−1
under a wider range of temperature, which was identical with those reported in the literature.
© 2004 Elsevier B.V. All rights reserved.

Keywords: Composite membrane; Pervaporation; Transport; PDMS; Pre-wetting

1. Introduction For organic compound separation from water, many

hydrophobic materials used for volatile organic compounds
Among the major applications of pervaporation mem- (VOCs) removal have been studied. The polydimethylsilox-
brane processes, organic separation from organic/water mix- ane (PDMS) is among the most interesting and promising
tures is becoming more and more important. Especially there membranes and has been extensively investigated [5,6]. The
has been growing research interest in the application of perva- preparation and pervaporation performance of composite
poration to biotechnology in areas such as recovery of ethanol PDMS membranes, consisting of a thin PDMS toplayer
from fermentation broths [1–4]. However, a major hurdle adhered on a porous support, has mainly been described in
limits pervaporation commercialization, namely, the lack of a variety of technical papers. The support materials mainly
proper membrane materials for this application. O’Brien et used were polyethersulfones [7–9], polyetherimides [10–12],
al. [2], after analyzing the fermentation–pervaporation pro- polyimides [13], polyacrylonitriles [14,15], polyesters [16],
cesses of a commercial-scale fuel ethanol plant, concluded and ceramics [17]. In addition, Several techniques have
that such a coupling system could be cost-competitive if the already been developed to prevent dilute PDMS solution
performance of membranes was improved modestly so as to from intruding pores in the support which may in a indirect
exhibit either the pervaporation total flux of 0.15 kg·m−2 ·h−1 way lead to high mass transfer resistance. The support
or selectivity of 10.3 for ethanol to water. pores could be filled with a nonsolvent [18,19] for the
coating polymer, or with a solvent [16], before applying the
∗ Corresponding author. Tel.: +86 25 8359 3772/6665; coating solution. However, in the case of alcohols or ketone
fax: +86 25 3317761. separation from water, it was recognized that composite
E-mail address: [email protected] (L. Li). PDMS membrane did not exhibit sufficient permeation

0376-7388/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.memsci.2004.06.015
178 L. Li et al. / Journal of Membrane Science 243 (2004) 177–187

flux due to the following reasons: (1) alcohols and ketone

compounds, compared with aroma compounds or VOCs,
are generally not very hydrophobic; (2) the proper choice of
support materials which are perfect matches for the PDMS
toplayer, has been not yet studied thoroughly and practically.
In this study, by a pre-wetting method the PDMS was cast
on the cellulose acetate (CA) microfiltration membranes, a
type of highly hydrophilic material, to prepare the composite
membranes. The pervaporation experiments were carried out
to investigate the separation of methanol, ethanol, n-propanol
and acetone from water. In addition, the mass transport of the
alcohol and ketone compounds through the PDMS membrane
was investigated and correlated using a resistance-in-series Fig. 1. Scheme of resistance-in-series model.
model.

where ko is the overall mass transfer coefficient, defined as:

2. Theory
1 1 1 1 L
= + = + (4)
Based on the solution-diffusion model, the flux of the com- ko kl km kl Dm K
ponent i through membrane (from membrane surface to per-
meate side) can be expressed as: where km denotes the membrane mass transfer coefficient
through the membrane. Eq. (4), known as the resistance-in-
Dm Ci,m (1 − pi,d /pei ) series model, has been widely used in the study of boundary
Ji = (1)
L layer mass transfer resistance effect in pervaporation of dilute
organic–water binary mixtures [21–23].
where Dm is the diffusion coefficient in the membrane, For a sufficient low downstream pressure, the second term
L the membrane thickness, Ci,m the permeate concentra- on the right hand of Eq. (3) becomes negligible. Then Eq. (3)
tion in membrane at the liquid–membrane interface, pi,d reduces to:
the partial downstream pressure of the component i, and
pei the equilibrium vapor pressure of the component i at Ji = ko Ci,b (5)
aqueous–membrane interface. Meanwhile the general equa-
tion of mass transfer in liquid boundary layer can be generally
given by: Eq. (5) can be used to derive the overall mass transfer
coefficient from the experimentally measured flux Ji and feed
concentration Ci,b .
Ji = kl (Ci,b − Ci,w ) (2)
From Eq. (4), the boundary-layer mass transfer resistance
(the intercept of the plot) and the permeability of the permeant
where kl is the boundary layer mass transfer coefficient, Ci,b
(=Dm K, the reciprocal slope) in addition to the membrane
the concentration of component i in liquid bulk and Ci,w the
resistance can be obtained form a plot of the permeation flux
concentration in liquid at the liquid–membrane interface.
data versus the membrane thickness.
It is necessary to note that Eq. (1) should be regarded as an
approximation of the mass transfer across PDMS membrane
by pervaporation, provided that the diffusion of permeants is
assumed to be independent of their concentrations in mem- 3. Experimental
brane. This assumption is reasonable since the concentration
of alcohols or ketone aqueous solution in this study is so low 3.1. Materials
that the swelling of PDMS is regarded nearly negligible. Fur-
thermore, Ci,m in Eq. (1) and Ci,w in Eq. (2) can conveniently ␣,␻-Dihydroxypolydimethylsiloxane (PDMS) with an av-
be correlated with the partition coefficient of the permeant in erage molecular weight of 5000 was purchased from Shang-
both liquid and membrane: K = Ci,m /Ci,w [20]. hai synthetic resin Company, China. Tetraethylorthosilicate
In accordance with the resistance-in-series model as (TAOS), dibutyltin dilaurate, n-heptane, methanol, ethanol,
shown in Fig. 1, combining Eqs. (1) and (2) together gen- n-propanol and acetone were obtained as analytical reagents
erates the following expression for the average permeation from Shanghai Chemical Reagent Company, China. Distilled
flux: and deionized water was used. Cellulose acetate (CA) micro-
filtration membranes, with an average pore size of 0.5 ␮m,
Ci,b − Ci,w pi,d /pei were from Shanghai Filter Company, China and used as sup-
Ji = = ko (Ci,b − Ci,w pi,d /pei ) (3)
1/kl + L/Dm K ports.
 L. Li et al. / Journal of Membrane Science 243 (2004) 177–187 179

3.2. Preparation of composite PDMS membranes

PDMS, crosslinking agent TAOS, and catalyst dibutyltin

dilaurate were mixed according to a 10/1/0.2 weight ratio
in n-heptane. Prior to coating, the CA support was laid and
spread out on the surface of water in a basin. Excess water on
the CA support surface was wiped off quickly with a filter pa-
per. Directly afterwards, the PDMS solution was poured over
the surface of support and the basin was put under a hood.
The membrane system containing some crosslinked PDMS,
after kept under ambient temperature for 2 h, was introduced
into an vacuum oven at 60 ◦ C for 4 h to complete the cross-
linking. With this technique, mass transfer resistance due to
the intrusion of the PDMS solution into the porous substrate Fig. 2. Cross-section of pervaporation cell.
during fabrication of the composite membrane could be re-
duced.
The composite membranes with skin layers of variable The experimental conditions were chosen in the following
thickness could be achieved by controlling the concentra- two series. One series of experiments was performed under
tion of PDMS solution or the coating amount. In this way, constant feed temperature of 313 K and variable feed flow
two composite membranes with skin layer thickness of 8 ␮m rate in the range of 5 to 120 L·h−1 , to investigate the effect
(1# membrane), and 16 ␮m (2# membrane) respectively were of the liquid flow over the surface of membrane on the mass
prepared for this study. transfer. This enabled the determination of correlations of
ko , kl and km to liquid flow status on the membrane. The
3.3. Scanning electron microscopy (SEM) other series of experiments was conducted under variable
feed temperatures in the range of 298–323 K and fixed feed
Scanning electron microscopy was used to study the cross- flow rate of 120 L·h−1 , to evaluate the effect of the tem-
sectional morphology of the composite PDMS membrane and perature on the mass transfer with constant hydrodynamic
to measure the skin layer thicknesses. The composite PDMS conditions. This established the relationship between ko , kl ,
membranes were fractured in liquid nitrogen. The fractured km and the feed temperature. Furthermore, the effect of the
section was coated with a conductive layer of sputtered gold. physicochemical properties of different compounds on mass
The cross-section of the composite membrane was investi- transfer coefficients could also be elucidated by experimental
gated using a Hitachi S-800 SEM. measurements with the same methods for various diffusing
compounds under constant feed temperature and Reynolds
3.4. Pervaporation number.
The permeation flux (J) at steady state was determined
Binary aqueous solutions containing methanol (5 wt.%), from the weight (W) of the collected permeant by using the
ethanol (5 wt.%), n-propanol (2 wt.%) and acetone (2 wt.%) following equation:
were prepared as feed solutions for the experiments, respec- W
tively. J= (6)
At
Pervaporation experiments were carried out using a con-
tinuous set-up reported by Li et al. [24]. The composite mem- where t is the experimental time interval for the pervapora-
branes were fixed into a plate and frame pervaporation test tion, and A the effective membrane surface area.
cell, as shown in Fig. 2. The system provided an effective A densimeter (DMA500, Anton Paar, Austria) was
mass transfer area of 0.024 m2 . The distance between the used to measure the liquid densities (the accuracy of
parallel plates was 0.01 m. The liquid entered the feed com- 0.000001 g·ml−1 ) and determine the organic concentrations
partment at the center of the cell and exited at a circular (the accuracy of 0.001 wt.%) by means of a standard curve
channel on the perimeter of the cell. The liquid flow rate was of density versus concentration attached to the densimeter.
measured by means of rotameters. After a steady state was The permselectivity of the membrane was calculated via the
obtained, the permeating vapor was collected by two stages separation factor (α) defined as
of cold traps. The first cold trap was provided with a refrig-
erant stream of −5 to −10 ◦ C, while the freezing medium in Yo /Yw
α= (7)
the second trap was liquid nitrogen. Xo /Xw
The downstream vacuity of membrane was kept at 266 Pa
in all experiments. To keep the feed concentration almost where X and Y are the weight fractions of species in the feed
constant, the collected permeant was added into the reservoir and permeate, respectively. Subscript “o” denotes organic
with 5 L feed solution in. compound and “w” water.
180 L. Li et al. / Journal of Membrane Science 243 (2004) 177–187

4. Results

4.1. SEM micrograph of composite PDMS-CA

membrane

The morphology of the composite PDMS-CA membrane

used in this study is presented in Fig. 3. It is evident from
the picture that the PDMS top layer is tightly and properly
cast on the top of the CA substrate. It can also be found that
less intrusion of PDMS into the micropores of the substrate
occurs as a result of the fabrication method used in this study. Fig. 5. Plot of total flux as a function of feed flow rate (T = 313 K); data for
the following membranes: (solid line)–1# , (dashed line)—2# ; (♦) acetone.

4.2. Pervaporation of the organic–water mixtures

4.2.1. Effect of feed flow rate

The effect of feed flow rate on the total flux and selec-
tivity for methanol (ethanol, n-propanol, acetone)–water
mixtures are presented in Figs. 4–7, respectively. With
increasing flow rate, both total flux and selectivity increase.
In addition, the effect of flow rate on the pervaporation
performance of acetone–water mixture is more evident than
three alcohol–water systems.
Fig. 6. Plot of selectivity as a function of feed flow rate (T = 313 K);
data for the following membranes: (solid line)—1# , (dashed line)—2# ; ()
methanol; (♦) ethanol.

4.2.2. Effect of temperature

The effect of feed temperature on the total flux and selec-
tivity for methanol (ethanol, n-propanol, acetone)–water mix-
tures are revealed in Figs. 8–10, respectively. The total flux of
four systems all increase monotonically with increasing tem-
perature because the mobility of permeating molecules are
enhanced both by the temperature and by the higher mobility
of the polymer segments. However, the effects of tempera-
ture on the pervaporation performance of the four mixtures
are distinctly various for different systems. The selectivity
of ethanol to water increases with increasing temperature. In
contrast, those of acetone and methanol to water descend. Be-
sides, that of n-propanol to water remains nearly invariable.
Our results are evidently different from those in the literature.
Fig. 3. Cross-section of composite PDMS membrane by SEM.

Fig. 4. Plot of total flux as a function of feed flow rate (T = 313 K); data for the Fig. 7. Plot of selectivity as a function of feed flow rate (T = 313 K); data
following membranes: (solid line)—1# , (dashed line)—2# ; () methanol; for the following membranes: (solid line)—1# , (dashed line)—2# ; () n-
(
) n-propanol; (♦) ethanol. propanol; (♦) acetone.
 L. Li et al. / Journal of Membrane Science 243 (2004) 177–187 181

4.2.3. Comparison of pervaporation performance

For comparison purpose, the pervaporation performances
of different PDMS membrane reported by other research
groups for separating ethanol–water mixture are listed in
Table 1. It can be seen that the total flux of the composite
PDMS-CA membranes is remarkably higher than those of
most PDMS supported and unsupported membranes reported
in the literature while the composite PDMS-CA membranes
in this study keep reasonably and acceptably selective for
ethanol. In the case of the separation of ethanol from 5 wt.%
ethanol solution at feed temperature of 313 K under 266 Pa,
the composite PDMS-CA membrane with skin layer thick-
ness of 8 ␮m in this study has a total flux of 1300 g·m−2 ·h−1
and a selectivity of 8.5 for ethanol to water, which has in prac-
Fig. 8. Plot of total flux as a function of temperature (F = 120 L·h−1 );
tice reached and exceeded the requirement of the membranes
data for the following membranes: (solid line)—1# , (dashed line)—2# ; ()
methanol; (
) acetone; (+) n-propanol; (♦) ethanol. for the use in cost-competitive fermentation–pervaporation
processes, assumed that thermodynamic effects of other com-
ponents in fermentation broths on activity coefficients and
coupling effects will not severely affect pervaporation per-
formance. Thus, it is evident that the composite PDMS-CA
membrane prepared in this study has excellent practical in-
dustrial application future.
Based on the comparison, it is believed that the excel-
lent pervaporation performance of the composite PDMS-CA
membrane depends on both the material type of the substrate
and the preparation method. Brief qualitative explanation is
given as follows. On one hand, CA is a type of highly hy-
Fig. 9. Plot of selectivity as a function of temperature (F = 120 L·h−1 ); drophilic material. By the pre-wetting method, the pores of
data for the following membranes: (solid line)—1# , (dashed line)—2# ; ()
n-propanol; (♦) acetone.
the CA substrate prior to coating are full of water due to
the capillary effect. This can mitigate mass transfer resis-
tance caused by the intrusion of the PDMS solution into
Molina et al. [25] found that separation index of methanol,
the porous substrate during the fabrication of the compos-
ethanol, iso-propanol increased by increasing temperature
ite membrane. Meanwhile thin and defect-free dense PDMS
on CMG-OM-010 and 1060-SULZER PDMS membranes.
skin layer can be assured to coat on CA support. On the other
Vankelecom et al. [26] investigated from their experiments
hand, hydrogen-bonding force exists between the O–H in the
that selectivity of ethanol tended to decrease with increasing
CA and oxygen atom in the PDMS, which causes the toplayer
temperature using PDMS membranes. The rational explana-
and substrate to adhere tightly. These effects both lead to a
tion of those phenomena is as follows: diffusion and solu-
perfect match between CA support materials and the PDMS
bility of penetrating components changes significantly with
toplayer, which consequentially gives rise to high total flux
temperature, which depends on many factors such as differ-
and acceptable selectivity of PDMS composite membranes.
ent organics, different membrane-preparation method, and
different supports of composite membranes and so on. The
further study is being carried on. 4.3. Analysis of transport

4.3.1. Definition of Reynolds number

In the present work, the liquid on the surface of the mem-
brane underwent a circular dipole flow pattern, as shown in
Fig. 11. Reynolds number is redefined for this special pattern
as follows:
Fρ
Re = (8)
πbµ

where F denotes volumetric flow rate of the feed, ρ the den-

sity, b the thickness of the cell and µ the dynamic viscosity.
Fig. 10. Plot of selectivity as a function of temperature (F = 120 L·h−1 );
data for the following membranes: (solid line)—1# , (dashed line)—2# ; () This definition of Re is similar to that in the work of Bandini
methanol; (♦) ethanol. et al. [37].
182 L. Li et al. / Journal of Membrane Science 243 (2004) 177–187

Table 1
Prevaporation performance of different PDMS supported and unsupported membranes
Membrane Membrane Ehtanol concentration Temperature J (g·m−2 ·h−1 ) α Ref.
thickness (␮m) in the feed (wt.%) (◦ C)
PDMS-PS graft copolymer supported on a PES 20 10 60 130 6.2 [27]
PDMS-PS block copolymer 39 10 25 27 6.2 [28]
PDMS-PPP graft copolymer 30 7.0 30 19 40 [29]
PDMS 40 11.9 25 14 7.1 [30]
PDMS – 8 30 25 10.8 [31]
PDMS-PI graft copolymer 20 6.6 48 32 6.6 [32]
PDMS-PS IPN supported membrane 15 10 60 160 5.5 [33]
PDMS composite membrane 2 5 30 1130 4.2 [34]
PDMS coated on silicate membrane 30 5 30 110 37 [35]
PDMS/PEI/PPP composite membrane 40 5 40 270 3.7 [36]
1060 Sulzer PDMS commercial membrane – 13 40 800 2.1 [25]
PDMS supported on a CA 8 5 40 1300 8.5 This study

coefficients are distinctly different for four organics. The ef-

fect is much significant for acetone while it is negligible for
methanol. As for alcohol homologues (methanol, ethanol and
n-propanol), the effect of Re on ko is becoming more signif-
icant with increasing number of carbon atoms. This seems
owing to the increase of hydrophobicity of the membrane
and molar volume of alcohols with increasing carbon atoms.
The curve of ko versus feed temperature under fixed
flowrate is illustrated in Fig. 13. It can be seen that ko is
linearly dependent on feed temperature.
Fig. 11. Dipole flow on the surface of membrane.
4.3.3. Liquid boundary and membrane mass transfer
There have existed several semi-empirical mass trans-
4.3.2. Overall mass transfer coefficient
fer correlations [40,41] to estimate the boundary layer mass
A plot of ko , obtained from the Eq. (5), against Re at con-
transfer coefficient as a function of the Reynolds number in
stant feed temperature is shown in Fig. 12. A primarily fit-
pervaporation. For laminar flow along a flat-channel cell, a
ting to experimental data shows that the increasing Reynolds
general correlation can be expressed directly in terms of the
numbers produced higher permeating rates of organics. It im-
boundary layer mass transfer coefficient as
plies that the mass transfer is influenced by organic diffusion

through the liquid boundary layer adjacent to the membrane. b c D
kl = aRe Sc = a × 10−6 Reb (8)
These results are consistent with those from pervious studies dh
of Urtiaga et al. [38], Psaume et al. [39], and Lipski and Côte
[21] using PDMS membranes for the removal of VOC from where a, b, c and a are constants particular to each module
aqueous solution. and to the flow regime, D is diffusivity related to properties
Furthermore, Fig. 12 also demonstrates that the effects of the fluid and dh is the hydraulic diameter of the membrane
of liquid flow in the module on the overall mass transfer cell. Various values for constants a, b and c can be found in
the literature. For fixed module geometry and dilute solution,

Fig. 12. Plot of overall mass transfer coefficient as a function of Re (T Fig. 13. Overall mass transfer coefficient versus feed temperature (F =
= 313 K); data for the following membranes: (solid line)—1# , (dashed 120 L·h−1 ); data for the following membranes: (solid line)—1# , (dashed
line)—2# ; () methanol; (
) ethanol; (+) n-propanol; (♦) acetone. line)—2# ; (+) methanol; (
) ethanol; (♦) n-propanol; () acetone.
 L. Li et al. / Journal of Membrane Science 243 (2004) 177–187 183

For the involved organic compounds, the relative contribu-

tion of the boundary layer and membrane mass transfer coef-
ficient to the overall transfer coefficient is obviously different
for various organics. For example, in the range of Reynolds
number investigated, methanol transfer is limited by the
membrane mass transfer resistance whereas acetone transfer
under the same hydrodynamic conditions depends on both the
mass transfer resistance. The effect of membrane thickness is
particularly important for methanol. Our experimental results
are similar with those of Brookes and Livingston [20]. They
Fig. 14. Liquid film and membrane mass transfer coefficient as a function argued that the percentage of each mass transfer resistance in
of Re for membranes of different skin layer thickness (T = 313 K; data for overall resistance is highly dependent on partition coefficient
acetone) (♦) kl ; (
) km (1# ); () km (2# ). K, one of the important thermodynamic property of organic
compound in a binary phase system of liquid and membrane.
It is demonstrated clearly that the analytical method based
on the resistance-in-series model is easy of access to deter-
mining the dominant resistance for various organic systems
and any given set of conditions. The method also provides an
optimizing tool for the practical pervaporation process.
The data for the membrane mass transfer coefficient rep-
resent two horizontal straight lines in Figs. 14 and 15, which
show the independence of the diffusion in membrane of flow
status on membrane surface. It is in accordance with the def-
inition of km in Eq. (4).
Fig. 15. Liquid film and membrane mass transfer coefficient as a function Estimation of the boundary layer thickness for all organ-
of Re for membranes of different skin layer thickness (T = 313 K; data for ics in this work can be obtained from the definition: kl =
methanol) (♦) kl ; (
) km (1# ); () km (2# ).
Di,b /Lb , where Di,b denotes the diffusion coefficient of or-
ganic compound in the boundary layer and Lb is the thick-
the exponent b should be independent of type of the particular ness of the boundary layer. The values Di,b can be obtained
VOC. by Wilke–Chang correlation [46]. A plot of Lb against Re is
Figs. 14 and 15 show the curves for the boundary layer shown in Fig. 16. It can be seen that Lb gradually decreases
and membrane mass transfer coefficients against Re (data for to a stable state with a rise of Re.
methanol and acetone). The boundary layer mass transfer The liquid film mass transfer on membrane surface is
data are well-fitted by the power law relationship as shown substantially dependent on the diffusion of ethanol in water.
in both figures. Values for exponent b of 0.5161 and 0.5142 According to Wilke–Chang correlation, the viscosity of a
are obtained for methanol and acetone, respectively. liquid is an exponentially decreasing function of temperature,
To test the validity of the method further, the experiments causing the diffusivity to increase exponentially with tem-
were repeated with two other organics, i.e., ethanol and n- perature. Fig. 17 gives the dependence of the boundary layer
propanol. The results with all four organics are consistent mass transfer coefficient on temperature. As can be seen, the
as shown in Table 2, yielding an average b value of 0.50. boundary layer mass transfer coefficient does increase expo-
The b value obtained in this work is identical with those in nentially with temperature at constant flow status. This ob-
works of Urtiaga et al. [42], Jiang et al. [43] and Smart et al. servation explains well the dependence of kl on temperature.
[44]. In all cases the observed behaviors clearly indicate that
there is an appreciable concentration polarization confined in
a thin layer close to the membrane. Consequently the bound-
ary layer effect at a given membrane thickness becomes more
significant at lower Reynolds numbers. This is in agreement
with the result from Urtiaga et al. [45].

Table 2
Values for constants in Eq. (8) by data fitting
Compound a b
Methanol 0.3171 0.5161
Ethanol 0.3294 0.4903
n-Propanol 0.2685 0.4765
Acetone 0.4932 0.5142 Fig. 16. Liquid boundary layer thickness versus Re (T = 313 K) (♦)
methanol; (
) ethanol; (×) n-propanol; () acetone.
184 L. Li et al. / Journal of Membrane Science 243 (2004) 177–187

Table 4
Values for constants in Eq. (10) by data fitting
Compound e f R2
Methanol 49297 3166.4 0.9124
Ethanol 1E+10 6948.1 0.9529
n-Propanol 1E+7 4550.3 0.9878
Acetone 2E+7 4409.5 0.9783

Values of e and f in Eq. (10) for four organics investigated

are given in Table 4, by fitting to the data published in our
previous work [49]. It is necessary to point out that the values
Fig. 17. kl vs. feed temperature (F = 120 L·h−1 ) (
) methanol; (+) ethanol;
for constants shown in Table 4 are only the results from data
() n-propanol; (♦) acetone.
fitting that can not be assumed to have any concrete physi-
Table 3 cal meaning. However, it is advantageous for us to find the
Values for constants in Eq. (9) by data fitting correlation above from the experimental results quickly.
Compound c d R2 The boundary layer and membrane mass transfer coeffi-
Methanol 1E−08 0.0669 0.9141 cients take on such temperature-dependent behaviors that the
Ethanol 0.0016 0.0267 0.9576 analytical method based on the resistance-in-series model
n-Propanol 1E−08 0.0647 0.9857 seems to be thought of a reasonable and promising technique
Acetone 0.0005 0.0336 0.9844 to conveniently determine both two coefficients. A compre-
hensive experimental project is being conducted so as to fur-
A fitting with much high correlation coefficient is carried ther validate this technique.
out for formulating the relationship of kl and T:
4.3.4. Diffusion coefficient in silicone rubber
kl = 10−6 c exp(dT ) (9)
There have been several studies involving the measure-
Under the present experiment, values of c and d in Eq. (9) ment of the diffusion coefficient of ethanol in PDMS mem-
are obtained from fitting data for four organics and shown in branes. Watson and Payne [47] measured the diffusion
Table 3. coefficient of ethanol in PDMS membrane to be 1.5 ×
According to Eq. (4), the membrane mass transfer 10−10 m2 ·s−1 at 1 wt.% ethanol concentration and feed tem-
coefficient km is essentially dependent on the diffusivity Dm perature of 80 ◦ C, while the result obtain by LaPack et al.
in membrane and the partition coefficient K in liquid–mem- [50] in PDMS membrane filled with fumed silica was 0.4
brane bi-phase system. The conventionally accepted view- × 10−10 m2 ·s−1 at concentration lower than 0.1% and feed
point is that both the diffusivity and the partition coefficient temperature of 25 ◦ C. Doig et al. [51] gave diffusivity for
conform to Arrhenius-type correlations, such as Scatchard– ethanol in PDMS membrane of 6 × 10−10 m2 ·s−1 at feed
Hildebrand activity-solubility equation for partition coeffi- temperature of 25 ◦ C.
cient and the equation used by Watson et al. [47,48] for the In terms of Eq. (4), the parameter Dm K can be cal-
diffusivity. Hence product of Dm and K should certainly be culated by product of km and δ, which characterizes the
an Arrhenius equation as well. Fig. 18 reveals the Arrhenius- solubility–diffusion of solutes in membrane. Thus, by the
type curves of the splitter membrane transfer coefficient. estimation of K, the diffusion coefficient of ethanol is de-
The fitting generates a quite typical Arrhenius equation: rived from the experiments to compare with the valves in the

 literature.
−6 f Silicone rubber is a type of polymer in the rubbery state
km = 10 e exp − (10)
T and so this enables the equilibrium partitioning of ethanol
between the liquid and membrane phases to be modeled us-
ing Hildebrand solubility parameter theory. Brooks and Liv-
ingston [20] gave the following equation for low organic com-
pound activity (less than 0.4):
vi vi
ln K = ((δi − δw )2 − (δi − δm )2 ) − (11)
RT vw
where δm , δi and δw denote solubility parameter for mem-
brane, organic compound and water, respectively; ␯i and νw
are molar volume of organic compound and water, respec-
tively.
Fig. 18. Arrhenius curves of the membrane mass transfer coefficient vs. 1/T; The required parameters for the experimental systems
(
) methanol; (×) n-propanol; () ethanol; (♦) acetone. can be obtained in the literature [52]. They are: νe =
 L. Li et al. / Journal of Membrane Science 243 (2004) 177–187 185

58.5 × 10−6 m3 ·mol−1 , νw = 18 × 10−6 m3 ·mol−1 , δw

= 36.6 MPa0.5 , δe = 26.6 MPa0.5 , δm = 17 MPa0.5 . Thus, Ci,m organic compound concentration in membrane
Eq. (11) can be solved to give a value of K of 0.04 at 25 ◦ C phase at equilibrium (g·ml−1 )
for ethanol. So the diffusion coefficient of ethanol in PDMS Ci,w organic compound concentration in aqueous
membrane at 5 wt.% ethanol concentration and feed temper- phase at equilibrium (g·ml−1 )
ature of 25 ◦ C is 2.875 × 10−10 m2 ·s−1 , which is at the same D diffusivity related to properties of the fluid
order of magnitude as values reported in the literature. In (m2 ·s−1 )
analogous way, values for Dm of 5.72 × 10−10 m2 ·s−1 and Di,b aqueous phase diffusion coefficient of organic
1.312 × 10−10 m2 ·s−1 can be obtained for methanol (298 K) compound (m2 ·s−1 )
and n-propanol (303 K), respectively. Dm diffusion coefficient of organic compound in
silicone membrane (m2 ·s−1 )
dh hydraulic diameter of the membrane cell (m)
F volumetric flow rate of the feed (m3 ·h−1 )
5. Conclusions
J permeation flux at steady state (g·m−2 ·h−1 )
Ji permeation flux of component i (g·m−2 ·h−1 )
Composite PDMS-CA membranes with thin-film toplayer
W weight of the permeant collected (g)
were prepared by a pre-wetting method to separate methanol,
K membrane/aqueous phase partition coefficient
ethanol, n-propanol and acetone from water. The effects
kl boundary layer mass transfer coefficient
of operating parameters on the pervaporation performance
(m·s−1 )
were investigated. It was found that the composite PDMS-
km membrane mass transfer coefficient (m2 ·s−1 )
CA membranes had remarkably higher permeation flux than
ko overall mass transfer coefficient (m·s−1 )
those PDMS supported and unsupported membranes reported
pi,d partial downstream pressure of the component
in literature and was reasonably and acceptably selective for
i (Pa)
ethanol.
pei equilibrium vapor pressure of the component i
A resistance-in-series model was used to obtain the mass
at aqueous-membrane interface (Pa)
transfer coefficients in the separation of four organics above-
L membrane thickness (␮m)
mentioned from water. It showed that the overall mass transfer
Lb boundary layer thickness (␮m)
coefficient increased linearly with temperature, and is expo-
T temperature (K)
nentially related to Reynolds number. The boundary layer
Xo weight fractions of organic in the feed
mass transfer coefficient appeared in a power law relation-
Xw weight fractions of water in the feed
ship to Reynolds number, and changed linearly with temper-
Yo weight fractions of organic in the permeate
ature. The membrane mass transfer coefficient was revealed
Yw weight fractions of organic in the permeate
to conform to an Arrhenius correlation and to be indepen-
dent of flow status. In addition, the relative importance of
Greek letters
the contributory mass transfer resistances for the overall per-
α separation factor
vaporation rates of organics through the composite PDMS
δe solubility parameter for ethanol (MPa0.5 )
membranes was investigated clearly in this study. The pre-
δm solubility parameter for membrane (MPa0.5 )
diction of K led to a rough estimation of diffusivity for alco-
δi solubility parameter for organic compound
hols in membrane, which gave values of Dm for alcohols at
(MPa0.5 )
a magnitude order of 10−10 m2 ·s−1 under a wider range of
δw solubility parameter for water (MPa0.5 )
temperature.
ρ density of the feed (kg·m−3 )
µ dynamic viscosity of the feed (Pa·s)
νe molar volume of ethanol (m3 ·mol−1 )
Acknowledgements νi molar volume of organic compound
(m3 ·mol−1 )
The present work was supported financially by National νw molar volume of water (m3 ·mol−1 )
985 Project of PR China (No. 985XK-015).

References
Nomenclature
[1] K. Kargupta, S. Datta, S.K. Sanyal, Analysis of the performance of
A effective membrane surface area (m2 ) a continuous membrane bioreactor with cell recycling during ethanol
b thickness of the membrane cell (m) fermentation, Biochem. Eng. J. 1 (1998) 31.
Ci,b organic compound concentration in bulk feed [2] D.J. O’Brien, L.H. Roth, A.J. McAloon, Ethanol production by con-
solution (g·ml−1 ) tinuous fermentation–pervaporation: a preliminary economic analy-
sis, J. Membr. Sci. 166 (2000) 105.
186 L. Li et al. / Journal of Membrane Science 243 (2004) 177–187

[3] M. Di Luccio, C.P. Borges, T.L.M. Alves, Economic analysis of [25] J.M. Molina, G. Vatai, E.B. Molnar, Comparison of pervaporation
ethanol and fructose production by selective fermentation coupled of different alcohols from water on CMG-OM-010 and SULZER
to pervaporation: effect of membrane costs on process economics, membranes, Desalination 149 (2002) 89.
Desalination 147 (2002) 161. [26] I.F.J. Vankelecom, D. Depre, S.D. Beukelaer, J.B. Uytterhoeven,
[4] D.J. O’Brien, G.E. Senske, M.J. Kurantz, J.C. Craig, Ethanol re- Influence of zeolites in PDMS membranes: pervaporation of wa-
covery from corn fiber hydrolysate fermentations by pervaporation, ter/alcohol mixtures, J. Phys. Chem. 99 (1995) 13193.
Bioresource Technol. 92 (2004) 15. [27] S. Takegemi, H. Yamada, S. Tusujii, Pervaporation of ethanol/water
[5] C.B. Almquist, S.-T. Hwang, The permeation of organophosphorus mixture using novel hydrophobic membrane containing poly-
compounds in silicone rubber membranes, J. Membr. Sci. 153 (1999) dimethylsiloxane, J. Membr. Sci. 75 (1992) 93.
57. [28] K. Okamoto, A. Butsuen, S. Nishioka, K. Tanaka, H. Kita, S.
[6] C.K. Yeom, H.K. Kim, J.W. Rhim, Removal of trace VOCs from Asakawa, Pervaporation of water–ethanol mixtures through polydi-
water through PDMS membranes and analysis of their permeation methylsiloxane block-copolymer membranes, Polym. J. 19 (1987)
behaviors, J. Appl. Polym. Sci. 73 (1999) 601. 734.
[7] Y. Chen, T. Miyano, A. Fouda, T. Matsuura, Preparation and gas [29] Y. Nagase, S. Mori, K. Matsui, Chemical modification of poly
permeation properties of silicone-coated dry polyethersulfone mem- (substituted-acetylene) IV. Pervaporation of organic liquid–water
branes, J. Membr. Sci. 48 (1990). mixture through poly(1-phenyl-1-propyne)/polydimethylsiloxane
[8] J.D. Le Roux, D.R. Paul, Preparation of composite membranes by a graft copolymer membrane, J. Appl. Polym. Sci. 37 (1989) 1259.
spin coating process, J. Membr. Sci. 74 (1992) 233. [30] Z. Changliu, L. Moe, X. We, Separation of ethanol–water mix-
[9] A. Fouda, Y. Chen, J. Bai, T. Matsuura, Wheatstone bridge model tures by pervaporation-membrane separation process, Desalination
for the laminated polydimethylsiloxane/polyethersulfone membrane 62 (1987) 299.
for gas separation, J. Membr. Sci. 64 (1991) 263. [31] K. Ishibara, K. Matsui, Pervaporation of ethanol–water mixture
[10] C.A. Page, A.E. Fouda, R. Tyagi, T. Matsuura, Pervaporation per- through composite membrane composed of styrene-fluoroalkyl acry-
formance of polyetherimide membranes spin- and dip coated with late graft copolymers and cross-linked polydimethylsiloxane mem-
polydimethylsiloxane, J. Appl. Polym. Sci. 54 (1994) 975. brane, J. Appl. Polym. Sci. 34 (1987) 437.
[11] S.-G. Li, K.V. Peinemann, A novel method for the preparation of [32] T. Kashiwagi, K. Okabe, K. Okita, Separation of ethanol from
composite membranes for gas separation, in: W.R. Bowen, R.W. ethanol/water mixtures by plasma-polymerized membranes from sil-
Field, J.A. Howell (Eds.), Euromembrane (Bath) Proceedings, 1995, icone compounds, J. Membr. Sci. 36 (1988) 353.
I 353. [33] L. Liang, E. Ruckenstein, Pervaporation of ethanol–water mixtures
[12] J. Bai, A.E. Fouda, T. Matsuura, J.D. Hazlett, A study on the prepara- through polydimethylsiloxane–polystyrene interpenetrating polymer
tion and performance of polydimethylsiloxane-coated polyetherimide network supported membranes, J. Membr. Sci. 114 (1996) 227.
membranes in pervaporation, J. Membr. Sci. 88 (1992) 100. [34] I. Blume, J.G. Wijmans, R.W. Baker, The separation of dissolved
[13] A. Fouda, J. Bai, S.Q. Zhang, O. Kutowy, T. Matsuura, Membrane organics from water by pervaporation, J. Membr. Sci. 49 (1990)
separation of low volatile organic compounds by pervaporation and 253–286.
vapor permeation, Desalination 90 (1993) 209. [35] T. Ikegami, H. Yanagishita, D. Kitamoto, H. Negishi, K. Haraya,
[14] K. Ebert, A. Bezjak, K. Nijmeijer, M.H.V. Mulder, H. Strathmann, T. Sano, Concentration of fermented ethanol by pervaporation using
The preparation of composite membranes with a glassy top layer, in: silicalite membranes coated with silicone rubber, Desalination 149
W.R. Bowen, R.W. Field, J.A. Howell (Eds.), Euromembrane (Bath) (2002) 49.
Proceedings, 1995, I 237. [36] M.O. Galindo, A.I. Clar, I.A. Miranda, A.R. Greus, Characterization
[15] J.M.S. Henis, M.K. Tripodi, Composite hollow fiber membrane for of poly(dimethylsiloxane)–poly(methyl hydrogen siloxane) compos-
gas separation: the resistance model approach, J. Membr. Sci. 8 ite membrane for organic water pervaporation separation, J. Appl.
(1981) 233. Poly. Sci. 81 (2001) 546.
[16] F.-J. Tsai, D. Kang, M. Anand, Thin-film-composite gas separation [37] S. Bandini, A. Saavedra, G.C. Sarti, Vacuum membrane distillation:
membranes: on the dynamics of thin film formation mechanism on experiments and modeling, AIChE J. 43 (1997) 398.
porous substrates, Sep. Sci. Tech. 30 (7–9) (1995) 1639. [38] A.M. Urtiaga, E.D. Gorri, J.K. Beasley, I. Ortiz, Mass transfer
[17] M.E. Rezac, W.J. Koros, Preparation of polymer–ceramic composite analysis of the pervaporation separation of chloroform from aque-
membranes with thin defect-free separating layers, J. Appl. Polym. ous solutions in hollow fiber devices, J. Membr. Sci. 156 (1999)
Sci. 46 (1992) 1927. 275.
[18] B. Bikson, J.K. Nelson, Composite membranes and their manufacture [39] R. Psaume, P. Aptel, Y. Aurelle, J.C. Mora, J.L. Bersillon, Per-
and use, AP 4826599 (1989). vaporation: importance of concentration polarization in the ex-
[19] S.E. Williams, B. Bikson, J.K. Nelson, R.D. Burchesky, Composite traction of trace organics from water, J. Membr. Sci. 36 (1988)
membranes for enhanced fluid separation, EP 0,286,091 B1 (1994). 373.
[20] P.R. Brookes, A.G. Livingston, Aqueous–aqueous extraction of or- [40] H.O.E. Karlsson, G. Tragardh, Pervaporation of dilute organic-waters
ganic pollutants through tubular silicone rubber membranes, J. mixtures. A literature review on modeling studies and application to
Membr. Sci. 104 (1995) 119. aroma compound recovery, J. Membr. Sci. 76 (1993) 121.
[21] C. Lipski, P. Côte, The use of pervaporation for the removal of [41] G. Charbit, F. Charbit, C. Molina, Study of mass transfer limitations
organic contaminants from water, Environ. Prog. 9 (1990) 254. in the determination of waste waters by pervaporation, J. Chem. Eng.
[22] B. Raghunath, S.-T. Hwang, General treatment of liquid-phase Jpn. 30 (1997) 382.
boundary layer resistance in the pervaporation of dilute aqueous [42] A.M. Urtiaga, E.D. Gorri, I. Ortiz, Modeling of the concentra-
organics through tubular membranes, J. Membr. Sci. 75 (1992) 29. tion–polarization effects in a pervaporation cell with radial flow,
[23] J.G. Wijmans, A.L. Athayde, R. Aaniels, J.H. Ly, H.D. Kamaruddin, Sep. Purif. Technol. 17 (1999) 41.
I. Pinnau, The role of boundary layers in the removal of volatile [43] J.-S. Jiang, D.B. Greenberg, J.R. Fried, Pervaporation of
organic compounds from water by pervaporation, J. Membr. Sci. methanol from a triglyme solution using a Nafion membrane. 2.
109 (1996) 135. Concentration–polarization, J. Membr. Sci. 132 (1997) 263.
[24] L. Li, Z.Y. Xiao, Z.B. Zhang, S.J. Tan, Mass transfer kinetics of [44] J. Smart, V.M. Starov, R.C. Schucker, D.R. Lloyd, Pervaporation
pervaporation by using a composite silicone rubber membrane: (I) extraction of volatile organic compounds from aqueous system with
the convective transport on membrane surface, J. Chem. Ind. Eng. use of a tubular transverse flow module. Part II. Experimental results,
53 (11) (2002) 1169 (in Chinese). J. Membr. Sci. 143 (1998) 159.
 L. Li et al. / Journal of Membrane Science 243 (2004) 177–187 187

[45] A. Urtiaga, E.D. Gorri, I. Ortiz, Mass-transfer modeling in the per- The convective transport on membrane surface, J. Chem. Ind. Eng.
vaporation of VOCs from diluted solutions, AIChE J. 48 (2002) 53 (11) (2002) 1169 (in Chinese).
572. [50] M.A. LaPack, J.C. Tou, V.L. McGuffin, C.G. Enke, The correlation
[46] C.R. Wilke, P.C. Chang, Correlation of diffusion coefficients in dilute of membrane permselectivity with Hildebrand solubility parameters,
solutions, AIChE J. 1 (1955) 264. J. Membr. Sci. 86 (1994) 263.
[47] J.M. Watson, P.A. Payne, A study of organic compound pervapora- [51] S.G. Doig, A.T. Boam, A.G. Livingston, D.C. Stuckey, Mass trans-
tion through silicone rubber, J. Membr. Sci. 49 (1990) 171. fer of hydrophobic solutes in solvent swollen silicone rubber mem-
[48] J.M. Watson, G.S. Zhang, P.A. Payne, The diffusion mechanism in branes, J. Membr. Sci. 154 (1999) 127.
silicone rubber, J. Membr. Sci. 73 (1992) 55. [52] D.R. Lide, Handbook of Chemistry and Physics, 71st ed., CRC, Boca
[49] L. Li, Z.Y. Xiao, Z.B. Zhang, S.J. Tan, Mass transfer kinetics of Raton, FL, 1991.
pervaporation by using a composite silicone rubber membrane. (I)