Journal of Membrane Scrence, 16 (1983) 269-284 269

Elsevier Science Publishers B.V.. Amsterdam - Printed in The Netherlands

ETHANOL-WATER SEPARATION BY PERVAPORATION*

M H.V. MULDER, J. OUDE HENDRIKMAN, H. HEGEMAN and C.A. SMOLDERS

Department of Chemical Technology, Twente Unwersrty of Technology, P 0 Box 217,
7500 AE Enschede (The Netherlands)

(Received January 24, 1983; accepted in revised form July 13, 1983)

Summary

The separation of ethanol-water mixtures is of great importance for the production of

ethanol from biomass. Both ultrafiltration and pervaporation processes can be used for the
continuous processing of fermentation and separation. The removal of ethanol from the
ultrafiltration permeate can be accomplished by pervaporation Separation of ethanol-
water mixtures by the pervaporation process has been investigated. Results are presented
for membranes which are preferentially permeable for ethanol and for others which are
preferentially water permeable. Details on the preparation of several membrane types
(homogeneous, asymmetric and composite) are given. A schematic process diagram is
given in which the fermentation of sugars to ethanol is membrane-controlled

Introduction

The application of pervaporation in biotechnology

The application of membrane separation processes in biotechnology is
rapidly growing. Conventional separation techniques such as distillation,
adsorption, liquid-liquid extraction and crystallization are often inefficient
and uneconomic. Contemporary membrane technology can save in process
costs because energy consumption is low, raw materials and nutrients can be
recovered and reused, fermentation processes can be carried out continuously
and disposal problems can be reduced or eliminated.
Membrane processes can be applied to one of the oldest and most famous
fermentation processes: the production of ethanol from the fermentation
of biomass. In the past decade this process has become of renewed interest
because of the impending petroleum shortage. One of the main advantages
of this fermentation process is that fuels are produced from renewable
resources.
Ethanol fermentation can be accomplished batchwise or continuously.
A substantial reduction in costs can be realized by a continuous operation [l]

*Paper presented at the Symposium on Membranes and Membrane Processes, May 19.-22,
1982, Peru&a, Italy.

0376-7388/83/$03.00 0 1983 Elsevier Science Publishers B.V.

270

According to Gregor [2,3] more cost saving can be achieved by the use of
various membrane processes, such as ultrafiltration, reverse osmosis and
electrodialysis in combination with distillation.
In principle, a combination of ultrafiltration and pervaporation makes it
possible to remove and concentrate ethanol during a continuous fermentation
process. It was demonstrated by Lee [ 31 that an ethanol-water mixture can
be removed and that the rejection of suspended solids is complete when beer
from the fermentor is ultrafiltrated.
Figure 1 gives a schematic presentation of a membrane-controlled contin-
uous fermentation of sugars to ethanol. Before the sugars enter the fermentor
reverse osmosis can be applied to concentrate the feed. In the fermentor
glucose is converted to ethanol. The ethanol productivity is limited by
ethanol inhibition and a low cell-mass concentration. Ultrafiltration can be
used as a cell recycle system, where the rejected cells are returned to the
fermentor and the ethanol is removed. Thus, the rate of ethanol production
remains high. The ultrafiltration permeate contains components such as salts,
glucose, ethanol and other low molecular weight substances, The ethanol
concentration in the ultrafiltration permeate will be about 5 to 10% by
weight. Ethanol can be purified to 99% or more by different separation tech-
niques such as distillation, adsorption or extraction. At this moment, the most

Fig. 1. Schematic presentation of membrane-controlled continuous fermentation process

for the production of pure ethanol.
 271

important process is distillation. The disadvantage of this process is that energy

consumption is rather high, especially when the azeotropic composition is
reached. Membrane processes are in general less energy consuming and a mem-
brane process which can be used to separate ethanol-water of any composition
is pervaporation.
In order to keep the energy consumption of this process as low as possible,
for low ethanol concentrations in the feed pervaporation should be performed
with ethanol-permeable membranes, and for high ethanol concentrations in
the feed water should be preferentially removed. Therefore, Fig. 1 shows a
two-stage pervaporation process with ethanol-selective membranes in the first’
stage and water-selective membranes in the second stage. It represents a con-
ceptual diagram showing a possible application of different membrane pro
cesses in biotechnology. The value of 40% for the permeate concentration in
the first stage is more or less arbitrary. From a commercial point of view, the
purification of ethanol from 5% to 39% by pervaporation alone is not
attractive, at least not at this moment.

Pervaporation in ethanol-water separation

Pervaporation involves selective sorption of a liquid mixture into the mem-
brane, diffusion through the membrane, and desorption into a vapour phase
on the permeate side, Until recently, pervaporation was not commercially
available despite the prospect of interesting potential applications such as
the separation of isomeric and azeotropic mixtures and the separation of
aqueous organic mixtures. During the European Workshop on Pervaparation
in Nancy a commercial pervaporation unit, in combination with distillation
for the purification of biomass ethanol was presented [ 41. A general dis-
advantage of the pervaporation process is the relatively high energy consump-
tion in comparison to pressure-driven membrane processes such as reverse
osmosis and ultrafiltration where no phase transition occurs. The pervapora-
tion process consumes an amount of energy which is at least equal to the heat
of vaporization of the complete pure product that has to be separated. An-
other disadvantage, the low permeation rate, can be compensated by mem-
brane configurations with a large area to volume ratio such as hollow fiber
systems.

Objectives of this research work

In this paper we will describe the separation of ethanol-water mixtures
by pervaporation. In the past this has been carried out primarily using dense
homogeneous membranes. Table 1 gives a summary of the selectivities and
permeation rates found in the literature; from this table, it can be seen that
the best results so far have been obtained with cellulose acetate and cellophane
membranes.
In a search for new membranes, the type of membrane structure to be
used has to be considered, because the membrane structure determines the
272

TABLE 1

Separation of ethanol- --water mixtures by pervaporation

Feed (% by weight) Perm. rate Temp. Type of Ref.

(cm/hr) (“C) membrane
Water (A) Ethanol (B)
-
45 55 8.5 0.20 80 CA [51
4 96 2.9 0.24 25 PTFE-PVP [6]
4 96 6.6 0.08 60 PTFE-PVP [7]
4 96 6.2 0.01 20 cellulose [71
4 96 5.9 0.02 20 CA [71
50 50 9.0 0.13 30 cellophane [f31
50 50 0.9 0.10 25 cellophane 191
50 50 2.0 0.46 45 cellophane [91
50 50 5.0 0.19 25 PTFE-PS [lOI
99.99 0.01 0.6b < 0.01 25 PE 1111
99.99 0.01 < 0.0006b <O.Ol 25 PETF [1J-1
99.99 0.01 0.4b 0.05 25 PVA [Ill
99.99 0.01 7 < 0.01 25 PDMS 1111
4 96 11 0.05 60 CA [I21
9.9 90.1 8 0.3 60 cellophane 1121
50 50 4.6b <O.Ol 25 PDMS [I31
-
a,AjB: Concentration A/B (weight %) in the permeate divided by the concentration AIB
in the feed.
b
“B/A.

separation mechanism. Pervaporation occurs according to a solution-diffusion

mechanism [ 14-171, transport taking place only by diffusion and not by
convective flow. This implies that very dense homogeneous membranes with-
out permanent pores are necessary. There are several types of membranes
possible with such a dense structure:
. homogeneous membranes;
o asymmetric membranes;
l composite membranes.
With symmetric homogeneous membranes the whole membrane thickness
contributes to the resistance to mass transfer, while for asymmetric and
composite membranes the major part of the resistance is situated in the thin
dense toplayer. We have developed all three types of membranes for ethanol-
water separation by pervaporation.

Experimental

Materials
Cellulose derivatives were obtained from Eastman Chemicals except for
cellulose tripropionate which was obtained from Aldrich. Polysulfone
(P 3500) was obtained from Union Carbide, polyvjnylidenefluoride (Solef
 273

2008) from Solvay, polydimethylsiloxane (RTV 670) and polydimethyl-

phenyleneoxide from General Electric, polyacrylonitrile from DuPont and
polyamide (Akulon M2) from Akzo. The solvents used were of analytical
grade.

Membrane preparation
Homogeneous membranes
Polymer solutions were prepared by dissolving the polymer in a suitable
solvent. The membranes were prepared by casting the polymer solution on a
glass plate and allowing the solvent to evaporate in a nitrogen atmosphere.
The membranes obtained were completely transparent.

Asymmetric membranes
Asymmetric membranes with a dense toplayer and a porous sublayer were
prepared by phase inversion. The polymer was dissolved in a solvent to form
a solution containing 10 to 30% by weight of polymer. The polymer solution
was cast on a glass plate and, after immersion in a nonsolvent bath, the mem-
brane was obtained.

Composite membranes
Composite membranes were prepared by means of dip-coating. A suitable
air-dried support layer was immersed in a dilute polymer solution and, after
evaporation of the solvent, a composite membrane was obtained with a thin
homogeneous polymer toplayer on the support layer, As support layers, poly-
vinylidenefluoride (PVDF) and Nylon-6 membranes have been used.
PVDF membranes were prepared by casting a polymer solution containing
25% PVDF, 60% dimethylacetamide and 5% dioxane by weight on a glass
plate; after immersion in a water bath at 20°C the membrane was obtained,
Nylon-6 membranes were prepared by casting a polymer solution containing
9% Nylon-6, 18% calcium chloride and 73% methanol by weight on a glass
plate; after an evaporation time of 60 seconds the film was immersed in a
water bath at 0°C.

Pervaporation
The pervaporation experiments were carried out as described previously
[18]. Vacuum at the downstream side was maintained at a pressure of 13.3 Pa
(0.1 mmHg) by a Crompton Parkinson vacuum pump. The pressure was
measured by an Edwards piranhi. Permeation experiments were carried out
for eight hours. After about three hours steady state condit,ions were reached.
A product sample was taken at least every hour. In most experiments a
50-50% by weight ethanol-water mixture was used as feed at a temperature
of 20°C. The asymmetric and composite membranes were installed with the
toplayer facing the feed.
274

Product analysis
Analysis of binary ethanol-water solutions was performed on a Varian
model 3700 gas chromatograph fitted with a Chromosorb SO/SO column and
equipped with a thermal conductivity detector. Low ethanol concentrations
(O--5%) were determined with a flame ionization detector. D-glucose was
determined as D-glucose phenylosazone by spectrophotomrtry at 390 nm.
Sodium chloride was measured by conductometry.

Results and discussion

Homogeneous membranes
Homogeneous membranes have been prepared using different polymers..
These are given in the first column of Table 2; membrane thickness is givcvl
in the second column.

TABLE 2

Pervaporation results using homogeneous membranes; feed: ethanol-water 50 --50% by

weight; temperature: 20°C
---__. -
Polymer Thickness aH,O Weight % H,O Permeation rate
(pm) in permeate (cm/hr) x 10’

CA 383a 10 2.0 66.9 11.3

CA 398 20 4.2 80.7 6.8
CTAb 10 2.7 73.1 8.9
CTPC 20 2.6 72.0 5.5
CAB 171d 30 4.0 80.0 4.2
CAB 272 20 4.1 80.5 3.3
CAB 381 30 3.2 76.3 2.3
PAN@- 25 70 98.6 0.15
PVDF’ 20 1.0 50 4.5
Psfs 20 332 99.7 0.04
PDMS b 10 0.3 21 1.7
____.
a CA : cellulose acetate, e PAN: polyacrylonitrile.
bCTA: cellulose triacetate. f PVDF : polyvinylidenefluoride.
c CTP: cellulose tripropionate. gPSf: polysulfone.
dCAB: cellulose acetate butyrate. hPDMS : polydimethylsiloxane.

In pervaporation literature selectivity is usually expressed by a selectivity

factor, O! (a = (yA/yB)j(xA/xB), where yA and yB are concentsations nf com-
ponents A and B in the permeate and XA and xa are concentrations of com-
ponents A and B in the feed). Because 01 may depend on the feed concentra-
tion, both selectivity factor and concentration in the permeate are given in
Table 2 (third and fourth column). In the last column the permeation rates
are given.
From Table 2 it can be concluded that in all cases, except that of poly-
dimethylsiIoxane (PDMS), water permeates preferentially through the
membrane. Polysulfone (PSf) and polyacrylonitrile (PAN) show the best
selectivities, but the permeation rates, especially for PSf, are very low. The
cellulose derivatives give moderate selectivities and rather high permeation
rates.
From reverse osmosis experiments it is known that annealing strongly
affects the performance of cellulose ester membranes. Homogeneous cellulose
acetate (CA 398), cellulose triacetate (CTA) and cellulose acetate butyrate
(CAB 171) membranes were annealed in water at 95°C for IO minutes. The
pervaporation results for these membranes are given in Table 3, which shows
that annealing hardly affects CA 398 membranes but that the selectivity of
CTA and CAB 171 membranes increases drastically in comparison with the
unannealed membranes. All the above-mentioned experiments were carried
out with a 50-5070 by weight mixture of ethanol-water.

TABLE3

Pervaporation result.6 of some homogeneous membranes with and without heat treatment;
feed: ethanol-water 50 -50% by weight; temperature: 20°C ; heat treatment: 10 min at
95°C

Polymer Without heat treatment With heat treatment

aHzO Weight % H,O Perm. rate aNzO Weight % H,O Perm. rate
in permeate (cm/hr) X 10” in permeate (cm/hr) x 10"

CA 398 4.2 80.7 6.8 3.9 79.5 6.5

CTA 2.6 72.0 5.5 36.0 97.3 2.3
CAB 171 4.0 79.8 4.2 9.6 90.6 2.9

Because selectivity may depend on liquid feed composition, the permeabil-

ity and selectivity of different compositions of ethanol-water were investi-
gated with homogeneous CA 398 (unannealed), PAN and CTA (annealed)
membranes. The results are given in Figs, 2 and 3. PSf has not been considered
further because of its extremely low permeation rate. In Fig. 2, the vapour-
liquid equilibrium of ethanol-water at 20°C and 1 bar [19] is also given
(dashed curve); from this figure can be seen that very high selectivities are
achieved with simple (bulk) polymers. This result is very interest,ing from a
commercial point of view.
Another objective of this research work was to find membranes which are
preferentially permeable to ethanol in order to remove ethanol from the
ultrafiltration permeate in the first step (see Fig. 1). From the results given
in Table 2, it is important to note that if either a hydrophobic glassy polymer
(PSf) or a more hydrophilic glassy polymer (CA) is used, water permeates
preferentially in both cases. This means that for hydrophobic glassy polymers
such as polysulfone (PSf) and polydimethylphenyleneoxide (PPO), which
exhibit no significant water sorption, water permeates preferentially because
of the presence of ethanol. With pure water, no permeability has been
276

I-O-

O-8

,_-’

O- I---- I 7- i
02 04 06 08 AIO
weight fraction Hz0 In feed
Fig, 2. Water concentration in the permeate as function of water concentration in the
feed for different homogeneous membranes. The dashed line is the vapour--liquid equilib-
rium of ethanol-water at 20°C.

0
l
l
l l l l l I

0-w 3 y c ‘;: 0 (
0 o-2 04 O-6- 0.8 10
weight fraction H,O in feed
Fig. 3. Permeation rate as function of the concentration of water in the feed for different
membranes.
 277

observed at all with PSf membranes. We conclude from these examples that it
would not be possible to predict selectivity for ethanol-water mixtures from
pure component permeability measurements. In a forthcoming article [ 1’71
we will discuss these phenomena in more detail.
Another interesting point to be learned from Table 2 is that when using
a hydrophobic rubber instead of a hydrophobic glassy polymer, ethanol
permeates preferentially. Comparing this result with the investigations of
Eustache [ll] on the removal by pervaporation of a large number of traces
of organic contaminants from water using polyethylene (PE), polyethylene-
terephtalate (PETF), polyvinyl alcohol (PVA) and polydimethylsiloxane
(PDMS), the rubber membranes (PE and PDMS) show a preferential permeabil-
ity to all organic solutes while the glassy polymers (PVA and PETF) ahow a
selectivity to water. From their extended investigations the only exception
found was for the system polyethylene-water-ethanol. For glassy polymers
diffusion through the membrane is rate determining. In the case of rubbers,
where the chains between the crosslink points are much more flexible and
segmental motions are less restricted than in glassy polymers, diffusion is
much faster and therefore sorption on the feed/membrane interface will
become important. Investigations by Hwang [ 291 on the permeation of
oxygen through a silicone rubber membrane did show that, besides the mem-
brane itself, interfacial resistance also contributes to the tot,al resistance.
Because ethanol permeates preferentially through polydimethylsiloxane
(PDMS) membranes, it is possible to apply pervaporation to remove ethanol
from the ultrafiltration permeate. The ethanol content of the ultrafiltration
permeate is low (5-10% by weight). The permeate also contains other low
molecular weight substances such as salts, non-converted glucose and other
organic components. To check the membrane selectivity for the various com-
ponents present, we performed an experiment using a PDMS membrane and
a 5% by weight solution of ethanol in water with D-glucose and sodium
chloride added to the feed. Since the exact composition of t.he ultrafiltration
permeate is difficult to establish (see Ref. [ 3]), we used the mixture of
D-glucose, sodium chloride, ethanol and water to get an indication of the
feasibility of the pervaporation process. The results are given in Table 4, from
which it can be seen that the rejection of D-glucose and sodium chloride is

TABLE 4

Pervaporation results with a PDMS membrane and as feed a mixture of’ ethanol, D-glucose,
sodium chloride and water; membrane thickness: 30 pm; temperature: 30°C; permeation
rate: 0.019 cm/hr

Feed @pm) Permeate (ppm)

Ethanol 50,000 190,000

D-glucose 10,000 < 200
NaCl 5,000 < 200
278

complete. However, the selectivity is rather low for ethanol-water mixtures,

lower than the vapour-liquid equilibrium.
The experiments performed with homogeneous membranes in.dicate that
for all membranes tested except the PDMS rubber membranes, water
permeates preferentially. Very high selectivities to water can be achieved
with simple homogeneous membranes but the selectivity to ethanol in case
of PDMS membranes is rather poor.

Composite membranes
The results obtained with homogeneous PAN and PSf membranes are very
promising as regards their selectivity, but they show low permeability. IBecause
the permeation rate through a homogeneous membrane is roughly inversely
proportional to membrane thickness, a reduction of the thickness will improve
the permeability. One can achieve this by preparing asymmetric or composite
membranes. These membranes consist of a thin dense toplayer and a porous
sublayer with a much smaller resistance to mass transfer, In tne case of com-
posite membranes, toplayer and sublayer originate from different polymeric
materials. Each of the two layers can be optimized independently for
maximum performance. Different methods can be used to apply the ultrathin
toplayer upon the support: dip-coating [ 201, interfacial polymerization [ 211
and plasma polymerization [ 221. The most simple method is dip-coating.
Hence, if it were possible to put a very thin homogeneous PAN or PSf layer,
by dip-coating, upon a suitable support, we would expect an increase in
permeability while the selectivity would remain the same or lessen only
slightly, The choice of the support depends on the solvent used for preparing
the polymer toplayer; this solvent should not attack the microporous struc-
ture of the support.
PSf is soluble in chlorinated hydrocarbons like methylene chloride,
trichloroethylene and tetrachloroethylene, Polyvinylidenefluoride (PVDF)?
however, is able to resist these solvents. A PVDF membrane has been used as
support for a composite membrane with a PSf toplayer. This PVDF mem-
brane itself has no selectivity towards an ethanol-water mixture and the
permeation rate is high (see Table 5).
PAN is soluble in amides like dimethylformamide (DMF), dimethyl-
acetamide (DMAc) and NUmethylpyrrolidone (NMP), Although there are few
polymers which are resistant to these solvents, aliphatic polyamides can be
used as support materials for a PAN toplayer, We have used for this purpose a
Nylon-6 membrane, which, uncoated, also shows no selectivity for ethanol--
water mixtures, but still has a very high permeation rate.
The coating procedure has been described in the experimental section. ‘The
following factors have been varied:
(1) Polymer concentration in the coating solution; 0.5~-8% by weight of
polymer.
(2) Coating time; generally about 2 minutes, variation from 1 to 60 minutes.
(3) Prewetting of the support.
 279

(4) Kind of solvent used in the coating solution; dichloromethane (b,p. 40°C)
and tetrachloroethylene (b.p. 146°C) have been used as solvents for PSf.
In the case of PAN, only dimethylformamide has been used as solvent.
(5) Drying of the support; careful drying was carried out at elevated tempera-
tures (90°C) and reduced pressure (0.6 kPa).
(6) Coating in a nitrogen atmosphere.
The best results obtained are given in Table 5. These membranes are ob-
tained by dip-coating a supporting PVDF membrane in a 6% PSf/CH,C&
solution and a Nylon-6 supporting membrane in a 8% PAN/DMF solution.
The permeation rates of the composite PSI and PAN membranes are one
order of magnitude higher than those of the homogeneous membranes; we had.
anticipated a larger increase. The reason for this low permeability is the
resistance which the support layer also exerts on mass transfer, and the rather
thick toplayers (6 pm) which were obtained by using the high (6-8%)
polymer concentration during dip-coating. When lower polymer concentra-
tions were used, the permeability did increase but the selectivity decreased
drastically. According to Cadotte [ 211) the possible reason for this
phenomenon is that dilute low-viscous polymer soIutions tend to migrate
upon drying to produce defective toplayers.

TABLE 5

Pervaporation results of composite membranes obtained by dip-coating; feed: ethanolk

water 50-50% by weight; temperature: 20°C

Polymer aH,O Weight % H,O Permeation rate

(toplayer/sublayer) in permeate (cm/hr) X 10’

PSf/PVDF 19.0 95.0 0.5

PAN/Nylon-6 9.0 90.0 2.0
PVDF support 1.0 50.0 20
Nylon-6 support 1.0 50.0 _a

aPermeability too high.

One can conclude from these results that, in principle, it is possible to

prepare composite membranes for ethanol-water separation by means of
dip-coating. However, up to now we have not been able to prepare very thin
(less than 1 ym) dense toplayers upon a support by this method without loss
of selectivity. Even the best composite membranes did show a loss of selectiv-
ity with respect to homogeneous membranes.

Asymmetric membranes
Another type of membrane with a very dense toplayer and a porous sub-
layer is formed by the asymmetric membranes. Most asymmetric membranes
are prepared by phase inversion. Morphology and performance of these mem-
branes largely depend on the choice of the ternary system polymer/solvent/
280

nonsolvent. Two different demixing processes determine the formation of

asymmetric membranes [ 23-2’71 :
(1) (Micro)crystallization or gelation for the formation of the toplayer.
(2) Liquid-liquid phase separation followed by gelation of the concentrated
polymer phase for the formation of the porous sublayer,
As already mentioned, during pervaporation transport takes place by dif-
fusion. This requires asymmetric membranes with a very dense toplayer.
Different factors can favour the formation of a dense toplayer 125,271 :
l A high initial concentration of the polymer solution.
l A lower tendency of the nonsolvent to induce liquid-liquid pbase separa-
tion. This means a low tendency of mixing of solvent and nonsolvent.
o A low temperature of the coagulation bath.
e An evaporation step before entering the coagulation bath _
Most of the commercially available asymmetric membranes are applied in
hyperfiltration processes. The membranes generally used in the pervaporation
process are of the homogeneous type. Only Aptel [7] reported on the
separation of water-dioxane mixtures with asymmetric polytetrafluoro-
ethylene membranes. In general, it can be said that asymmetric hyperfiltration
membranes are not suitable for pervaporation because the toplayer is not
dense enough.
Experiments performed with a Loeb-Manjikian type of CA reverse osmosis
membrane [ 301 obtained from a polymer solution consisting of 25% cellulose
acetate (E 39%3), 30% formamide and 45% acetone, did not show any selec-
tivity towards ethanol-water mixtures. Bijddeker [ 311 showed that reverse
osmosis membranes (polybenzimidazolone (PBIL) and RC-100), which
exhibit very high salt rejection (R > 99%) and can be used for single-pass
seawater desalination, have hardly any selectivity to ethanol-water mixtures
in pervaporation processes.
With the procedure given above to obtain a denser toplayer, we developed
asymmetric pervaporation membranes for ethanol-water separation using
different polymers. The results are given in Table 6.
The asymmetric CA membranes prepared from acetone-containing solutions
show a better performance than the homogeneous ones: the permeability is
slightly less but the selectivity is much better. The reverse is the case with PSP
where the asymmetric membranes have a much higher permeability but a
lower selectivity.
Polydimethylphenyleneoxide (PPO) membranes were obtained with a high
selectivity but with a low permeability. CA, PSf and PPO membranes were
also prepared from a more dilute polymer solution and the results are also
given in Table 6. These latter membranes exhibit a much lower selectivity. It
seems clear from these results that polymer concentration is a very important
factor for obtaining dense-skinned asymmetric membranes suitable for per-
vaporation. Another important factor in obtaining dense toplayers is the
solvent-nonsolvent interaction. If the solvent-nonsolvent interaction is
low (i.e., a high value of the excess free energy of mixing), the binodal de-
 281

TABLE 6

Pervaporation results of asymmetric membranes obtained by phase inversion, feed:

ethanol-water 50-50% by weight; temperature: 20°C

Polymer Solvent Nonsolvent Polym. Cont. aYHo Weight % H,O Perm. rate
*
(S) in permeate (em/hr) x 1 O2

CA Acetone Water 25 12.3 92.5 2.7

CA Acetone Water 18 5.9 85.5 4.2
CA DMSO Water 25 1.0 50.0 32.5
CA DMSO Water 18 -. _b
PSf DMAc Water 35 3.0 75.0 1.4
b
PSI DMAc Water 15
PPO TCEa Methanol 20 9.3 90.3 0.2
PPO TCE Methanol 10 1.0 50.0 19.2

a TCE : trichloroethylene.
bPermeability too high,

mixing gap for liquid-liquid phase separation in the phase diagram is shifted
to higher water concentrations. This is the case for CA as polymer, acetone
as solvent and water as nonsolvent. Acetone-water solutions have a very
high excess free energy of mixing whereas DMSO-water solutions show a
negative excess free energy of mixing which means a high mutual affinity
[28]. The binodals of these systems are given in Fig. 4. For solutions of CA in
DMSO, where the location of the liquid-liquid phase separation demixing
gap is near the polymersolvent axis, addition of a small amount of water is
sufficient to induce liquid-liquid phase separation. As a result, membranes
are obtained with a less dense toplayer [ 27]. Also, membranes which were
prepared from both dilute and concentrated solutions of CA in DJSO Jid
not show any selectivity towards ethanol-water mixtures because the top-
layer was not dense enough.
For solutions of CA in acetone, a relatively large amount of nonsolvent is
necessary before liquid-liquid phase separation occurs. Consequently, a
denser toplayer is the result.

/ ~_.I_ ---\ water

DMSO acetone
Fig, 4. Schematic presentation of the liquid--liquid separation demixing gap for CA/solvent/
water systems with acetone and DMSO as solvent. (From Ref. [ 281.)
282

It is obvious that in preparing asymmetric membranes for pervaporation

purposes two factors are very important: the polymer concentration in the
casting solution should be high and the solvent-nonsolvent interaction
should be low. Besides the above-mentioned thermodynamic aspects, kinetic
factors such as nonsolvent flow into the nascent membrane and solvent flow
into the coagulation bath, together with the gelation/crystallization kinetic‘s,
also play an important role in membrane formation.

Conclusions

With pervaporation in combination with ultrafiltration, it is possible to

remove ethanol preferentially and continuously from a fermentation reactor.
The ethanol-water mixture obtained after ultrafiltration can be purified by a
two-stage pervaporation process with homogeneous, composite or asymmetric
membranes.
The problem with the comp0sit.e membranes prepared by dip-coating is the
inadequate homogeneity and/or thickness of the selective toplayer.. We did
not succeed in preparing very thin toplayers (less than 1 pm) by this t,echnique
without loss of selectivity.
Both homogeneous and asymmetric membranes prepared from various
polymers showed very high selectivities to ethanol-water mixtures. Annealing
of the homogeneous cellulose ester membranes had a positive effect on the
selectivity.

Acknowledgement

This paper is based. upon work financially supported by the Ministeries van
Economische Zaken en Wetenschapsbeleid, in The Netherlands.

References

1 G.R. Cysewski and C.R. Wilke, Process design and economic studies of alternative
fermentation methods for the production of ethanol, Biotechnol. Bioeng., 20 (1978)
1444.
2 H.P. Gregor and T.W. Jeffries, Ethanolic fuels from renewable resources in the solar
age, Ann. N.Y. Acad. Sci., 326 (1979) 273.
3 T.S. Lee, D. Omstead, N.H. Lu and H.P, Gregor, Membrane separations in alcohol
production, Ann. N.Y. Acad. Sci., 369 (1981) 367.
4 H.E.A. Briischke, W.H. Schneider and G.F. Tusel, Lecture presented at the European
Workshop on Pervaporation, September 21-22, 1982, Nancy, France.
5 R.C. Binning and R.J. Lee, Separation of azeotropic mixtures, U.S. Patent 2,953,502
(September 20, 1960).
6 P. Aptel, N. Challard, J. Cuny and J. Neel, Application of the pervaporation process
to separate azeotropic mixtures, J. Membrane Sci., 1 (1976) 27 1.
7 P. Aptel, J. Cuny, J. Jozefonvicz, G. Morel and J. Neel, Liquid transport through
membranes prepared by grafting of polar monomers onto poly(tetrafluoroethylene)
films. I. Some fractionations of liquid mixtures by pervaporation, J. Appl. Polym.
Sci., 16 (1972) 1061.
 283

8 R.Y.M. Huang and N.R. Jarvis, Separation of liquid mixtures by using polymer
membranes. II. Permeation of aqueous alcohol solutions through cellophane and
polyjvinylalcohol), J. Appl. Polym. Sci., 14 (1970) 2341.
9 E.G. Heisler, A.S. Hunter, J. Sicilian0 and R.H. Treadway, Solute and temperature
effects in the pervaporation of aqueous alcoholic solutions, Science, 124 (1956) 77
10 G.C. Tealdo, P. Canepa and S. Munari, Water-ethanol permeat.ion through grafted
PTFE membranes, J. Membrane Sci., 9 (1981) 191.
11 H. Eustache and G. Histi, Separation of aqueous organic mixtures by pervaporation
and analysis by mass spectrometry or a coupled gas chromatograph-mass spectrometer,
J, Membrane Sci., 8 (1981) 105.
12 E, Nagy, 0. Borlai and A. Ujhidy, Membrane pervaporation of water~malcohoi binary
mixtures, J. Membrane Sci., 7 (1980) 109.
13 K.C. Hoover and S.T. Hwang, Pervaporation by a continuous membrane column,
J. Membrane Sci., 10 (1982) 253.
14 R.C. Binning, R.J. Lee, J.F. Jennings and E,C. Martin, Separation of liquid mixtures,
Ind, Eng. Chem., 53 (1961) 45.
15 D.R. Paul and J.D. Paciotti, Driving force for hydraulic and pervaporative transport
in homogeneous membranes, J. Polym. Sci., A-2, 13 (1975) 1201,
16 C.H. Lee, Theory of reverse osmosis and some other membrane permeation operations,
J. Appl. Polym. Sci., 19 (1975) 83,
17 M.H.V. Mulder and C.A. Smolders, On the mechanism of separation of ethanol/water
mixtures by pervaporation. I. Calculations of concentration profiles, J. Membrane Sci.,
in press.
18 M.H.V. hlulder, F. Kruitz and C.A. Smolders, Separation of isomeric xylenes by
pervaporation through cellulose ester membranes, J. Membrane Sci., 11 (1982) 319.
19 J, Gmehling and U. Onken, Vapour-liquid Equilibrium Data Collection, Vol I,
Part 1, DECHEMA, Frankfurt am Main, F.R.G., 19’i7.
20 H.K. Lonsdale, R.L. Riley, C.R. Lyons and D.P. Carosella, Transport in composite
reverse osmosis membranes, in: M. Bier (Ed.), Membrane Processes in Tndust,ry and
Biomedicine, Plenum Press, New York, 1971, p. 101.
21 J.E. Cadotte and R.J. Petersen, Thin-;‘ilm composite reverse-osmosis membranes:
origin, development and recent advances, in: A.F. Turbak (Ed.), Synthetic Membranes,
Vol. 1, ACS Symp. Ser. 153, ACS, Washington, D.C., 1981, Chap. 21.
22 H. Yasuda, Composite reverse osmosis -membranes prepared by plasma polymer-ization,
in: 8. Sourirajan iEd.), Reverse Osmosis and Synthetic Membranes, National Research
Council Canada, Ott.awa, 1977, Chap. 13.
23 D.M. Koenhen, M.H.V. Mulder and C.A. Smolders, Phase separation phenomrna during
the formation of asymmetric membranes, J Appl. Polym. Sci., 21 (1977) 199
24 L. Broens, D.M. Koenhen and C.A. Smolders, On the mechanism of formation ot
asymmetric ultra- and hyperfiltration membranes, Desalination, 22 (1977) 205
25 L. Broens, P.W. Altena, C.A. Smolders and D.-M. Koenhen, Asymmetric membrane
structure as a result of phase separation phenomena, Desalination, 32 (1980) 33
26 F.W I Altena and C,A. Smolders, Phase separation phenomena in solutions of cellulose
acetate. I. Differential scanning calorimetry of cellulose acetate in mixture’: of dioaane

and water, J. Polym. Sci., Polym. Symp., 69 (1981 ) 1.

27 C.A. Smolders, Morphology of skinned membranes: A rationale from phase sepal-aiion
phenomena, in: A.R. Cooper (Ed.), Ultrafiltration Membranes and Application, Polym.
Sci. Technol., J-3 (1980) 57, Plenum Press, New York.
28 F.W. Altena and C.A. Smolders, Calculation of liquid-liquid phase separation ill a
ternary system of a polymer in a mixture of a solvent and a nonsolvent> Llacromolecules,
15 (1982) 1491.
29 S.T. Hwang, T.E.S. Tang and K. Kammermeyer, Transport of dissolved clxygen through
silicone rubber membrane, Amer. Chem. Sot. Polym. ?repr.> 10 (1969) 978
284

30 S, Manjikian, S. Loeb and J.W. McCutchan, Improvement in fabrication techniques for

reverse osmosis desalination membranes, Proceedings of the First International
Symposium on Water Desalination, U.S. Department of the Interior, Washington, D.C.,
1965, p. 159.
31 K.W. Boddeker, A. Wenzlaff and D. Cavigelli, Membranverfahren in der Rioalkohol-
gewinnung. Versuche iiber Verdampfungspermeation, Report GKSS 81 /E/62, CKSS-
Forschungszentrum Geesthacht GmbH, Geesthacht, F.R.G., 1982.