Separation and Purification Technology 47 (2005) 80–87

Analysis of the membrane thickness effect on the pervaporation

separation of methanol/methyl tertiary butyl ether mixtures
J.P.G. Villaluenga ∗ , M. Khayet, P. Godino, B. Seoane, J.I. Mengual
Department of Applied Physics I, Faculty of Physics, University Complutense of Madrid, 28040 Madrid, Spain

Received 11 March 2005; received in revised form 1 June 2005; accepted 2 June 2005

Abstract

The effect of the membrane thickness on the pervaporation separation of methanol and methyl tertiary butyl ether mixtures through
membranes was studied. Membranes of a wide range of thicknesses were prepared from two different polymers: cellulose acetate and
poly(2,6-dimethyl-1,4-phenylene oxide). For each membrane, the experiments were performed at the same feed pressure, feed temperature
and permeate pressure. The results showed that the permeate flux through both membrane types decreased markedly with increasing the
membrane thickness, while the separation factor remained nearly constant. This behavior was discussed in terms of a resistance-in-series
model.
© 2005 Published by Elsevier B.V.

Keywords: Mass transfer; Membranes; Modeling; Pervaporation; Separations

1. Introduction (50/50 vol.%) through a plastic membrane was proportional

Liquid mixtures can be separated by partial vaporization
through a dense permselective membrane. This separation
technique has been termed pervaporation in order to
emphasize the fact that the permeate undergoes a phase
change, from liquid to vapor, during its transport through
the membrane matrix. In this process, the feed mixture is
maintained in direct contact with one side of the membrane,
and the permeate is evolved, in the vapor state, from
the opposite side of the membrane, which is kept at low
pressure. The permeate is collected, in the liquid state after
condensation, on a cooled wall [1–5].
The characterization of a pervaporation process is usually
considered with regard to the variations of different operation
variables, such as the feed composition, the temperature, the
permeate pressure, the membrane thickness and the feed flow
velocity. The influence of membrane thickness on selectivity
and flux has been studied by a few workers. Binning et al. [6]
observed that the flux of a mixture of n-heptane and isooctane
∗ Corresponding author. Tel.: +34 91 394 4454; fax: +34 91 394 5191.
E-mail address: juanpgv@fis.ucm.es (J.P.G. Villaluenga).
to the reciprocal membrane thickness and the selectivity was
independent of the thickness for membrane thicknesses in the
range of 20–50 ␮m.
Brun et al. [7], who studied the influence of the membrane
thickness on the selectivity using nitrile rubber membranes
and a mixture of butadiene and isobutene (60/40 vol.%), con-
cluded that the selectivity was constant above a membrane
thickness of 100 ␮m, and a lower selectivity was found when
using membranes of 17 ␮m. The selectivity lowering for
thin membranes was explained by assuming the existence
of micropores in the membrane matrix, which allowed the
diffusion of molecules through them.
Spitzen et al. [8] studied the influence of the membrane
thickness on the separation of water/ethanol mixtures using
polyacrylonitrile membranes, and they found that the selec-
tivity decreased drastically when using membranes with a
thickness below 20 ␮m. This behavior was also attributed to
the existence of the same artifacts in the membrane.
Aptel et al. [9] observed a selectivity lowering with a
decrease of the membrane thickness, for grafted polyte-
trafluoroethylene membranes, using water/dioxane mixtures.
Koops et al. [10] also investigated the effect of the membrane

1383-5866/$ – see front matter © 2005 Published by Elsevier B.V.

doi:10.1016/j.seppur.2005.06.006
 J.P.G. Villaluenga et al. / Separation and Purification Technology 47 (2005) 80–87
thickness on the separation of water/acetic acid mixtures
using polysulfone, poly(vinyl chloride) and polyacrylonitrile
membranes. It was observed that the selectivity was indepen-
dent of the membrane thickness above 15 ␮m, but below this
limiting thickness, the selectivity decreased with decreasing
membrane thickness. This dependence, which could not
be explained by differences in the polymer morphology,
or by flow coupling, was attributed to the formation of
induced defects in the membrane during the pervaporation
process.
Qunhui et al. [11] studied the influence of the membrane
thickness on the permselectivity of chitosan membranes, used
for the separation of water/ethanol mixtures. For membranes
with thicknesses lower than 30 ␮m, it was found that the
selectivity increased with the membrane thickness, whereas
for membranes with thickness higher than 50 ␮m, the mem-
branes exhibited constant selectivities. It was also found that
the flux was proportional to the reciprocal of the membrane
thickness.
Recently, Kanti et al. [12] studied the effect of the mem-
brane thickness on the permselectivity for the dehydration
of a mixture of 95.4 wt.% ethanol and 4.6 wt.% water, using
blended chitosan/sodium alginate membranes. It was found
that the flux decreased significantly with the increase in
the membrane thickness from 25 to 190 ␮m, whereas the
selectivity increased to a lesser extent. The variation of the
selectivity with the membrane thickness was related to the
existence of a dry layer in the membrane on the permeate side,
which was responsible of the permselective properties of the
membrane. It was explained that the thickness of this dry
layer increased with the total membrane thickness, causing
a rise in the membrane mass transfer resistance. Therefore,
the selectivity increased when the membrane thickness was
increased. Sridhar et al. [13] observed the same behavior
for chitosan membranes using mixtures of water and
acetamide.
From the above-cited studies, it can be concluded that
an optimal membrane thickness is required in order to
obtain a constant selectivity, and below which the selectivity
decreases with membrane thickness decreasing. Moreover,
the flux through the membrane decreases with its thickness
to a variable extent.
On the other hand, different classes of models for perva-
poration can be found in the literature. Reviews on perva-
poration models were published by Feng and Huang [4] and
Lipnizki and Trägårdh [14]. One of the models developed to
study the mass transport in membranes is the resistance-in-
series approach, which has been used by several researchers
in pervaporation [15–18]. Basically, this approach considers
that the mass transfer from the liquid feed to the permeate
vapor involves four successive steps: (i) a mass transfer from
the feed bulk to the feed–membrane interface boundary layer,
(ii) a sorption into the membrane, (iii) a mass transport in the
membrane matrix and (iv) a desorption to the permeate. Each
step is modeled with different approaches, and various fun-
damental assumptions have been considered.
81
The main objective of the current study is to investi-
gate the membrane thickness influence on the pervaporation
separation of methanol/methyl tertiary butyl ether mixtures.
Dense membranes with a wide range of thicknesses were pre-
pared by using two different polymers: cellulose acetate and
poly(2,6-dimethyl-1,4-phenylene oxide). The pervaporation
behavior for the separation of methanol and methyl tertiary
butyl ether from their mixtures was studied in terms of the
permeation rate and the separation factor. The analysis of the
influence of the membrane thickness on the pervaporation
performance was carried out by means of a resistance-in-
series model.
2. Experimental
2.1. Materials
Cellulose acetate (CA) polymer of a molecular weight of
37,000, with a 39.8% degree of acetylation, was purchased
from Aldrich Chemicals. Poly(2,6-dimethyl-1,4-phenylene
oxide) (PPO) powder of an intrinsic viscosity of 1.57 dL/g
and a density of 1.04 g/cm3 was supplied by General Electric.
Two solvents were used in this study to prepare the casting
solutions: acetone for CA and chloroform for PPO. To carry
out the pervaporation experiments, methanol and methyl ter-
tiary butyl ether (MTBE) of analytical purity grade (97–99%)
were used without further purification.
2.2. Membrane preparation
Casting solutions were first prepared by dissolving
3.3 wt.% of CA polymer in DMF. The membrane samples
were prepared by pouring a predetermined amount of poly-
mer solution over a mirror-polished glass plate with a circular
edge. The solution was then dried in three steps. First, the bulk
of the solvent was removed by slow evaporation inside a fume
hood at an average humidity of 30% and at a temperature
of 23 ◦ C until the membrane was formed. Subsequently, the
membrane was kept in an oven at 70 ◦ C during 8 h. Finally,
the formed membrane was left overnight in a vacuum oven
at 70 ◦ C to remove the traces of the solvent.
The PPO polymer solution was prepared by using 4 wt.%
in chloroform. Different volumes of PPO solution were
spread smoothly over a leveled glass plate inside a stain-
less steel-made O-ring of about 10 cm inner diameter. The
casting ring was then covered with a filter paper to keep out
of dust. After 24 h at a temperature of 25 ◦ C, the membranes
were removed very cautiously by immersing the glass plate
in a water bath. The membrane was then dried at 25 ◦ C for
24 h in a fume hood, and for 72 h in vacuum to remove the
last traces of solvent.
The thickness of each membrane was measured by a
Millitron micrometer (Mahr Feinpruf 1202 IC) over at least
20 different spots, and the mean value was used in this
study.
82 J.P.G. Villaluenga et al. / Separation and Purification Technology 47 (2005) 80–87
2.3. Pervaporation experiments
Pervaporation experiments were performed by using
the system described elsewhere [19,20]. It consists of a
separation cell, a circulation pump, two permeate traps,
two vacuum pumps and a pressure transducer. The effective
membrane pervaporation surface area was 28 cm2 . The feed
was circulated over the membrane sample. The feed pressure
was kept near 105 Pa. The permeate stream was evacuated by
one of the vacuum pumps, and the permeate was collected in
one of the traps immersed in a filled liquid nitrogen flask. In
all the experiments, the downstream pressure was maintained
less than 133 Pa. After the completion of each experiment,
the permeate collected inside the cold trap was warmed up to
room temperature, and then weighed. The feed and permeate
compositions were determined by measuring their refractive
index with an Abbey-type refractometer Model 60/ED. The
pervaporation selectivity of the membranes was studied in
terms of the separation factor, α, which is defined as:
wp,i /wp,j
α= (1)
wb,i /wb,j
where wb and wp are the weight fractions of the components
i and j in the bulk feed and permeate, respectively. Indexes
i and j refer to the more permeable component and the less
permeable one, respectively.
In each experimental run, it was observed that the con-
centration in the feed solution after the completion of the
experiment was almost the same as that of the initial feed solu-
tion. Therefore, the feed concentrations could be considered
constant throughout the experiment. In each experimental
run, at least three samples were collected to determine the
total flux and the composition of the permeate. The alternative
use of two cold traps allowed a continuous sampling in order
to avoid interrupting the experiment. It must be mentioned
that both membranes exhibited initially high fluxes, which
then gradually reached steady-state values. This behavior is
due to the conditioning process, which is a time-dependent
adaptation of the membrane transport properties because of
the rapidly changed process conditions. This conditioning
time, which can last from minutes to several hours, could be
explained in terms of the polymer relaxation. All data given in
the present paper correspond to the final steady-state values.
In this study, the pervaporation experiments were con-
ducted at a feed temperature of 25 ◦ C using; as feed, binary
mixtures of methanol/MTBE with a 21 wt.% of methanol, in
the case of the CA membranes and 47 wt.% methanol, in the
case of the PPO membranes, were employed.
3. Theory
As previously mentioned, the resistance-in-series
approach considers that the mass transfer of a mixture,
from the feed bulk to the permeate, takes place in four
successive steps. In order to make the model more useful,
some simplifying assumptions have been made: absence
of flow coupling effects, constancy of the diffusivity of the
components in the membrane, equilibrium at the membrane
interfaces and steady-state conditions. Based on the results
published in previous papers [19,20], the two first assump-
tions may seem not to be reasonable in the case of the
pervaporation of methanol and MTBE mixtures when PPO
and CA membranes are used; however, the consistency of
the results and the model predictions will be used in order to
justify the above assumptions.
The model considers that the transport across the system
is produced by a gradient of the chemical potential of the feed
mixture components. Thus, the flux, Ji , of component i can
be written as
dµi
Ji = −Li (2)
dz
where dµi /dz is the gradient of chemical potential of compo-
nent i and Li is a phenomenological coefficient.
The mass transport of component i from the bulk feed to
the feed membrane interface can be written as:
Ji = Lbl,i (µb,i − µbl,i ) (3)
where Lbl,i is the mass transfer coefficient in the boundary
layer formed at the membrane feed side, µb,i is the chemical
potential of component i at the bulk and µbl,i is the chemical
potential of component i at the feed–membrane interface.
The transport through the membrane matrix can be
expressed as:
(µmb,i − µmp,i )
Ji = Lm,i (4)
l
where Lm,i is the mass transfer coefficient of component i in
the membrane, µmb,i and µmp,i are the chemical potentials of
i, in the membrane, at the feed–membrane interface and at
the permeate–membrane interface, respectively and l is the
membrane thickness.
The transport of component i can also be written as fol-
lows:
(µb,i − µp,i )
Ji = Lov,i (5)
l
where Lov,i is the overall mass transfer coefficient of compo-
nent i, and µb,i and µp,i are the chemical potentials at the bulk
and at the permeate, respectively. If equilibrium conditions
are assumed at both membrane interfaces, the following rela-
tionship between the transport coefficients can be obtained
from Eqs. (3) to (5):
1 1 1
= + (6)
Lov,i lLbl,i Lm,i
Following Raghunath and Hwang [15], it is of practical
interest to write Eq. (6) in terms of typical transport coef-
ficients often employed in pervaporation. In Eq. (2), the
chemical potential difference may be related to the differ-
ences in pressure, temperature, concentration and electrical
 J.P.G. Villaluenga et al. / Separation and Purification Technology 47 (2005) 80–87 83

potential. In the case of pervaporation, where the driving force In various pervaporation studies [15,21], it was shown that
is generated by concentration and pressure gradients, the fol- second term of the right-hand side is negligible.
lowing equation is applied: In pervaporation, the overall flux is expressed in terms of
the fugacity difference between the bulk feed and the perme-
dµi = RT d ln(γi xi ) + vi dp (7) ate as follows [22]:
where xi is the mole fraction of i, γ i the activity coefficient,
fb,i − fp,i (psat γi xb,i − p2 xp,i )
vi the molar volume of component i and p is the pressure. In Ji = Qi = Qi i (16)
the case of a liquid, the integration of Eq. (7) gives l l
where Qi is the overall or apparent permeability of component
µi = µoi + RT ln(γi xi ) + vi (p − psat
i ) (8)
i, and fb,i and fp,i are the fugacities of component i at the bulk
where µoi is the chemical potential of pure i at a reference and at the permeate, respectively.
pressure, which is defined as the saturation vapor pressure, From Eqs. (15) and (16), under the assumptions discussed
psat
i . The following equation may be obtained from Eqs. (2)
above, it can be obtained that
and (8),
pi Qi
Lov,i = (17)
Li RT dxi dp RT
Ji = − − Li v i (9)
xi dz dz
where
If the solution–diffusion model [21] is considered, this
equation can be applied within the membrane limits, and the i γi x
(psat b,i − pp xp,i )

pi = (18)
γx
following expression is obtained: ln xi p,ib,i


Lm,i RT xmb,i
Ji = ln (10) Finally, Eq. (6) can be rewritten as follows:
l xmp,i
1 pi pi
where xmb,i and xmp,i are the mole fractions of i in the = + (19)
membrane, at the feed–membrane interface and at the Qi RTLbl,i l Di cm,i
membrane–permeate interface, respectively.
This equation can be expressed in a more practical form
If the diffusion of compounds in the membrane is Fickian
with a diffusion coefficient independent of the concentration, 1 1 1
it can be written = + (20)
Qi ki l D i S i
(xmb,i − xmp,i )
Ji = ρm Di (11) where
l
where Di is the diffusion coefficient of component i in the RTLbl,i
ki = (21)
membrane and ρm is the membrane density. The comparison pi
of Eqs. (10) and (11) gives cm,i
Si = (22)
cm,i Di pi
Lm,i = (12)
RT
where ki is the mass transfer coefficient in the boundary layer
where with respect to the fugacity difference and Si is the solu-
ρm (xmb,i − xmp,i ) bility coefficient of component i in the membrane. Eq. (20)
cm,i = (13) shows that the apparent permeability varies with the mem-
ln(xmb,i /xmp,i )
brane thickness when the boundary layer contributes to the
On the other hand, by using Eq. (8) the difference in the mass transport of compounds. In addition, by plotting the
chemical potential between the bulk feed and permeate may inverse permeability as a function of the inverse membrane
be expressed as follows: thickness, a straight line should be obtained. From the slope

 of the line ki can be calculated, and from the intercept Di Si
γi xb,i
µb,i − µp,i = RT ln + vi (p1 − p2 ) (14) can be estimated. It is worth quoting that Di Si is the intrin-
xp,i sic permeability of component i in the membrane (Pm,i ). An
where xb,i and xp,i are the mole fractions of i at the bulk and interesting additional finding can be obtained from Eq. (20)
at the permeate, respectively, and p1 and p2 are, respectively, by considering that the overall mass transfer resistance of
the pressure at the feed and permeate. Using Eq. (14), Eq. (5) component i (Rov,i = l/Qi ) consists of the sum of the resistance
can be rewritten as follows: in the liquid boundary layer (Rbl,i = 1/ki ) and a membrane

  resistance (Rm,i = l/Pm,i ). Thus, it can be noticed that the so-
Lov,i RT γi xb,i vi
Ji = ln + (p1 − p2 ) (15) defined overall mass resistance coefficient depends on the
l xp,i RT membrane thickness.
84 J.P.G. Villaluenga et al. / Separation and Purification Technology 47 (2005) 80–87
Fig. 1. Overall and partial fluxes of methanol and MTBE as a function of
the membrane thickness for PPO membranes.
4. Results and discussion
Fig. 1 shows the data on the overall flux, and the methanol
and MTBE fluxes through PPO membranes, as a function of
the membrane thickness. The composition of the feed solu-
tion was the same in all cases, i.e. a 47 wt.% methanol in a
methanol/MTBE mixture. Both the overall flux and the partial
fluxes of methanol and MTBE decreased with the membrane
thickness. For example, the overall flux obtained when using
the thinnest membrane (28 ␮m) is approximately twice that
of the thickest membrane (126 ␮m). It was also observed that
plots of fluxes versus the reciprocal of the membranes thick-
ness gave straight lines with correlation coefficients higher
than 0.99. Fig. 2 presents the results of the overall, methanol
and MTBE fluxes, obtained using CA membranes of differ-
ent thicknesses. The feed was a mixture of methanol and
MTBE with 21 wt.% of methanol. Both the overall flux and
the component fluxes decreased with membrane thickness.
The overall flux, as an example, is about twice that of the
Fig. 2. Overall and partial fluxes of methanol and MTBE as a function of
the membrane thickness for CA membranes.
thinnest membrane (20 ␮m) than for the thickest one (96 ␮m).
In addition, as observed for PPO membranes, it was also
checked that the fluxes were proportional to the inverse of
the membrane thickness with correlation coefficients higher
than 0.98.
It can be observed in Figs. 1 and 2 that the overall perme-
ation flux through the PPO membranes is greater than that of
the CA membranes. Based on the solution–diffusion model,
the permeability of the membranes is determined by the sol-
ubility and diffusivity of permeants in the membranes. It was
shown in previous papers [19,20] that the liquid sorption of
PPO was larger than that of CA due to the higher affinity
of the PPO membranes towards the liquid mixture than that
of CA. This makes PPO membrane to swell more than CA
membrane. Because of the higher swelling, the diffusivity of
permeants in the PPO membrane was larger than in the CA
membrane. Consequently, the higher flux observed for the
PPO membrane is due to the greater solubility and diffusiv-
ity of the permeants. On the other hand, in the present study, it
is also observed that the MTBE flux of CA membranes is two
orders of magnitude lower than that of methanol. This indi-
cates that the selectivity of CA membranes is higher than that
of PPO membranes, although both PPO and CA membranes
are methanol selective. In Fig. 3, the separation factors of both
membrane types are presented as a function of the membrane
thickness. There is no clear tendency between the separation
factor and the membrane thickness within experimental error.
It seems that the separation factor is about 2.1 for PPO mem-
branes and around 192 for CA membranes. It was found [20]
that the higher separation factor of CA membranes was due
to the combination of both high sorption and diffusion selec-
tivities. In contrast, the lower separation factor exhibited by
the PPO membranes was due to a low sorption selectivity as
it was reported elsewhere [19].
As stated earlier, the data given in Figs. 1–3 were ana-
lyzed by using the resistance-in-series approach developed
in the preceding section. First, the values of Qi were calcu-
lated from Eq. (16). Data on the mole fractions of methanol
Fig. 3. Separation factor in pervaporation of methanol/MTBE mixtures
through PPO and CA membranes.
 J.P.G. Villaluenga et al. / Separation and Purification Technology 47 (2005) 80–87 85

Table 1
Constant values in Eq. (23) for methanol and MTBE
a1 a2 a3 a4
Methanol −8.54796 0.76982 −3.1085 1.54481
MTBE −7.82516 2.95493 −6.94079 12.17416

and MTBE, at the bulk and at the permeate, were obtained

experimentally, as well as the pressure at the permeate side.
The considered thicknesses of the membranes were those of
the dry membranes. The saturation vapor pressure values for
methanol and MTBE were calculated by using the following
equation [23]:

sat 
pi a1 x + a2 x1.5 + a3 x3 + a4 x6
ln = (23)
pc 1−x
Fig. 5. Inverse of the overall permeability of methanol and MTBE as a
where pc is the critical pressure of component i, x is equal function of the reciprocal membrane thickness for CA membranes.
to 1 − T/Tc , where Tc is the critical temperature. Values for
dimensionless constants a1 , a2 , a3 and a4 are given in Table 1.
Experimental data of the activity coefficients of methanol lines with reasonably good correlation coefficients, about
and MTBE, as a function of temperature and concentration 0.99 for PPO membranes and about 0.98 for CA membranes.
in their mixtures, were reported by Coto et al. [24]. Based Based on these results, both the intrinsic permeability in
on these data, the following expression can be written for the membranes, Pm,i , and the mass transfer coefficients in
the dependence of the activity coefficients of methanol and the boundary layer, ki , were calculated for methanol and
MTBE on their mole fractions at 25 ◦ C: MTBE. In the case of PPO, the permeability coefficients
of methanol and MTBE are, respectively, 3.5 × 10−13 and
ln γi = b1 xb,i + b2 xb,i + b3 xb,i + b4 xb,i + b5 1.4 × 10−13 (kg m/m2 s Pa). In the case of CA membrane,
4 3 2
(24)
the permeability of methanol and MTBE are, respectively,
Values for dimensionless constants b1 , b2 , b3 , b4 and b5 are
2.1 × 10−13 and 1.2 × 10−15 (kg m/m2 s Pa). These results
given in Table 2.
agree well with those presented earlier by other authors for
Figs. 4 and 5 give the inverse of the calculated Qi val-
pervaporation of methanol/MTBE mixture using PPO and
ues of methanol and MTBE as a function of the inverse
CA membranes [20,25]. For both membranes, methanol per-
of membrane thickness, for the PPO and CA membranes,
meability is greater than MTBE one. This may be attributed
respectively. It can be observed that such plots yield straight
not only to the methanol preferential sorption but also to the
Table 2 diffusion selectivity, because the diffusional cross section of
Constant values in Eq. (24) for methanol and MTBE methanol is much lower than that of MTBE [19,20].
b1 b2 b3 b4 b5 Moreover, an ideal separation factor can be calculated
Methanol 0.8554 −2.5627 3.8749 −3.4947 1.3236
as the ratio between the intrinsic membrane permeability of
MTBE 0.8326 −2.1289 2.937 −2.8673 1.2239 methanol and MTBE in each membrane. The obtained values
are 2.5 and 175 for PPO and CA membranes, respectively.
By comparing these results with those reported in Fig. 3, it
can be seen that both sets of data are nearly the same. This
confirms the validity of the theoretical approach used in this
study.
Furthermore, the mass transfer coefficient, Ri , of
methanol and MTBE in the liquid boundary layer, of
the PPO membrane are, respectively, 7.8 × 10−9 and
3.6 × 10−9 (kg/m2 s Pa). The mass transfer coefficient in the
boundary layer of the CA membrane of methanol and MTBE
are, respectively, 6.2 × 10−9 and 5.2 × 10−11 (kg/m2 s Pa).
For methanol the difference between the mass transfer coef-
ficients obtained in PPO and CA membranes is about 11%,
whereas for MTBE the difference goes up to 36%. This may
be attributed to the different membranes and feed solutions
used.
Fig. 4. Inverse of the overall permeability of methanol and MTBE as a On the basis of the theoretical framework developed in
function of the reciprocal membrane thickness for PPO membranes. the preceding section, the mass resistance coefficients of
86 J.P.G. Villaluenga et al. / Separation and Purification Technology 47 (2005) 80–87

Table 3 as the polydimethylsiloxane ones, the mass transfer resis-

Mass resistance coefficients of methanol and MTBE in PPO membranes as tance at the liquid boundary layer contributes significantly
a function of the membrane thickness
to the overall transport resistance, being the rate-determining
l (␮m) Rm,methanol (108 m2 s Pa/kg) Rm,MTBE (108 m2 s Pa/kg) step depending on the existing hydrodynamic conditions. On
28 0.88 2.0 the contrary, for less permeable membranes, such as those
38 1.2 2.7 prepared with ethylene propylene rubber, the overall mass
73 2.3 5.2
81 2.6 5.8
transfer resistance is dominated by the membrane resistance,
118 3.7 8.4 which becomes significant with increasing the membrane
126 4.0 9.0 thickness. Raghunath and Hwang [15] conducted pervapora-
tion experiments with polydimethylsiloxane and polyether-
block-polyamide membranes using dilute phenol and toluene
methanol and MTBE, in the boundary layer and in the mem-
aqueous solutions. It was investigated the effect of the mem-
branes were estimated from Eq. (20). In the case of PPO,
brane thickness on the permeability of the organic compound,
the mass resistance coefficients of methanol and MTBE in
which was the preferentially permeated component. It was
the liquid boundary layer are, respectively, 1.3 × 108 and
concluded that the organic transfer resistance in the bound-
2.8 × 108 (m2 s Pa/kg). In the case of CA membrane, the val-
ary layer was significant, by limiting the overall mass transfer
ues of methanol and MTBE are, respectively, 1.6 × 108 and
through the membrane, when the membrane resistance to the
1.9 × 1010 (m2 s Pa/kg). The values obtained of the resistance
transport of organic was low. In the present study, it is found
coefficients in the membranes are given in Tables 3 and 4.
that for highly selective membranes, such as those prepared
It can be observed that the mass resistance coefficients of
with CA, and for reasonably good permeable membranes,
methanol are the same order of magnitude in both mem-
such as PPO membranes, both the boundary layer resistance
branes, whereas the values for MTBE are noticeably larger in
and the membrane resistance contribute to the overall trans-
the CA membranes than in the PPO membranes. Moreover, it
port resistance.
is important to note that there is a limiting membrane thick-
As previously mentioned, the selectivity was found to be
ness, below which the contribution of the boundary layer
almost independent of the membrane thickness when both
resistance to the overall resistance is larger than the mem-
PPO and CA membranes were used (Fig. 3). Based on this
brane resistance. As the membrane thickness is increased,
experimental observation and on others [15,26,27], it can
the contribution of the membrane resistance becomes greater
be stated that for pervaporation systems in which the over-
than the resistance in the boundary layer. So, in the case of
all mass transfer resistance is dominated by the membrane
PPO membranes, the resistance in the boundary layer is larger
resistance or the liquid boundary layer and the membrane
than in the membrane matrix, when membranes with a thick-
resistances are comparable, i.e. the present study, the selec-
ness of lower than 38 ␮m approximately are used. Above this
tivity remains nearly independent of the membrane thickness;
thickness, the resistance in the PPO membranes becomes
on the contrary, when the overall mass transfer resistance is
larger than in the liquid boundary layer. In the case of CA
dominated by the boundary layer resistance, the selectivity
membranes, this limiting thickness is about 23 ␮m for MTBE
depends on the membrane thickness.
and 33 ␮m for methanol.
The extent to which the boundary layer affects the over-
all transport resistance depends not only on the membrane
5. Conclusions
thickness, but also on the permeability of the membrane. For
example, Nijhuis et al. [26] studied the effect of the membrane
Cellulose acetate and poly(2,6-dimethyl-1,4-phenylene
thickness on the performance of polydimethylsiloxane, ethy-
oxide) membranes were used for the separation of methanol
lene propylene rubber and polyoctenamer membranes, used
from methyl tertiary butyl ether by pervaporation. Mem-
for the removal of toluene and trichloroethylene from water.
branes of different thicknesses were prepared in order to
It was found that boundary layer at the liquid/membrane
evaluate the influence of the variation of membrane thick-
interface affected the transport of the water and the organic
ness on the process performance.
compounds. In fact, for highly permeable membranes, such
It was found that the transmembrane flux decreased
Table 4 markedly with the membrane thickness for all the mem-
Mass resistance coefficients of methanol and MTBE in CA membranes as a branes used. Moreover, the overall flux was proportional to
function of the membrane thickness the reciprocal of the membrane thickness. On the other hand,
l (␮m) Rm,methanol (108 m2 s Pa/kg) Rm,MTBE (1010 m2 s Pa/kg) both membrane types are methanol selective, and it was not
20 0.95 1.7 observed, within the experimental accuracy, any effect of the
23 1.1 1.9 variation of membrane thickness on the pervaporation sepa-
33 1.6 2.8 ration factor.
46 2.2 3.8 The pervaporation results, which were analyzed by apply-
72 3.4 6.0
ing a resistance-in-series model, indicate that, besides a resis-
96 4.6 8.0
tance to the transport of methanol and methyl tertiary butyl
 J.P.G. Villaluenga et al. / Separation and Purification Technology 47 (2005) 80–87
ether in the membranes, there is an additional resistance at the
liquid feed/membrane interface. It was found that when thin-
ner membranes were used, the boundary layer resistance was
larger than the membrane resistance. As the membrane thick-
ness is increased, the membrane resistance becomes greater
than the resistance in the liquid boundary layer. The turn-
ing point has been localized for cellulose acetate membrane
between 23 and 33 ␮m of membrane thickness, while for
poly(2,6-dimethyl-1,4-phenylene oxide) membrane is near
38 ␮m.
Acknowledgements
The authors acknowledge the financial support from
the Spanish Ministry of Science and Technology (MCYT)
through its project PPQ2003-03299. Thanks are due to Prof.
T. Matsuura from the Industrial Membrane Research Institute
(IMRI) for supplying PPO polymer.
References
[1] F. Lipnizki, R.W. Field, P.K. Ten, J. Membr. Sci. 153 (1999) 183.
[2] N. Wynn, Chem. Eng. Prog. 97 (2001) 66.
[3] W.S.W. Ho, K.K. Sirkar, Membrane Handbook, Chapman and Hall,
New York, 1992.
[4] X. Feng, R.Y.M. Huang, Ind. Eng. Chem. Res. 36 (1997) 1048.
[5] S.P. Nunes, K.V. Peinemann, Membrane Technology in the Chemical
Industry, VCH, Weinheim, 2001.
[6] R.C. Binning, R.J. Lee, J.F. Jennings, E.C. Martin, Ind. Eng. Chem.
53 (1961) 45.
87
[7] J.P. Brun, G. Bulvestre, A. Kergreis, M. Guillou, J. Appl. Polym.
Sci. 18 (1974) 1663.
[8] J.W.F. Spitzen, G.H. Koops, M.H.V. Mulder, C.A. Smolders, in: R.
Bakish (Ed.), Proceedings of third International Conference on Per-
vaporation Processes in Chemical Industry Nancy France, Bakish
Materials Corp., Englewood, NJ, 1988, p. 252.
[9] P. Aptel, J. Cuny, J. Jozefonvicz, G. Morel, J. Neel, J. Appl. Polym.
Sci. 18 (1974) 351.
[10] G.H. Koops, J.A.M. Nolten, M.H.V. Mulder, C.A. Smolders, J. Appl.
Polym. Sci. 53 (1994) 1639.
[11] G. Qunhui, H. Ohya, Y. Negishi, J. Membr. Sci. 98 (1995) 223.
[12] P. Kanti, K. Srigowri, J. Madhuri, B. Smitha, S. Sridhar, Sep. Purif.
Technol. 40 (2004) 259.
[13] S. Sridhar, T. Srinivasan, U. Virendra, A.A. Khan, Chem. Eng. J. 94
(2003) 51.
[14] F. Lipnizki, G. Trägårdh, Sep. Purif. Method 30 (2001) 49.
[15] B. Raghunath, S.T. Hwang, J. Membr. Sci. 65 (1992) 147.
[16] M.G. Liu, J.M. Dickson, P. Côté, J. Membr. Sci. 111 (1996)
227.
[17] J.G. Wijmans, A.L. Athayde, R. Daniels, J. Ly, H.D. Kamaruddin,
I. Pinnau, J. Membr. Sci. 109 (1996) 135.
[18] J.S. Jiang, D.B. Greenberg, J.R. Fried, J. Membr. Sci. 132 (1997)
263.
[19] J.P.G. Villaluenga, P. Godino, M. Khayet, B. Seoane, J.I. Mengual,
Ind. Eng. Chem. Res. 43 (2004) 2548.
[20] A. Tabe-Mohammadi, J.P.G. Villaluenga, H.J. Kim, T. Chan, V.
Rauw, J. Appl. Polym. Sci. 82 (2001) 2882.
[21] J.G. Wijmans, R.W. Baker, J. Membr. Sci. 107 (1995) 1.
[22] J.G. Wijmans, J. Membr. Sci. 79 (1993) 101.
[23] H.D. Kamaruddin, Ph.D. Thesis, University of Texas, Austin, 1997.
[24] B. Coto, R. Wiesenberg, C. Pando, R.G. Rubio, J.A.R. Renuncio,
Ber. Bunsenges. Phys. Chem. 100 (1996) 482.
[25] S. Cao, Y. Shi, G. Chen, J. Membr. Sci. 165 (2000) 89.
[26] N.H. Nijhuis, M.H.V. Mulder, C.A. Smolders, J. Membr. Sci. 61
(1991) 99.
[27] W. Ji, S.K. Sikdar, S.T. Hwang, J. Membr. Sci. 93 (1994) 1.