0263–8762/04/$30.00+0.00
# 2004 Institution of Chemical Engineers
www.ingentaselect.com=titles=02638762.htm Trans IChemE, Part A, July 2004
Chemical Engineering Research and Design, 82(A7): 855–864
SEPARATION OF ETHANOL–WATER MIXTURES USING
MOLECULAR SIEVES AND BIOBASED ADSORBENTS
S. AL-ASHEH*,{, F. BANAT and N. AL-LAGTAH
Department of Chemical Engineering, Jordan University of Science and Technology, Irbid, Jordan
T
he potential use of new biobased adsorbents and different types of molecular sieve was
assessed in the separation of the ethanol–water azeotrope. Molecular sieves of type 3A,
type 4A, and type 5A and biobased adsorbents such as natural corncobs, natural and
activated palm stone and oak, were used in this study. Each of these adsorbents was packed in
a column, which was surrounded by a heating jacket. The jacket temperature was maintained
constant in an attempt to maintain the packed-column at constant temperature. The water
concentration in the feed solution was varied from 5% to 12% (weight basis). Although the
initial concentration of ethanol in the flask was constant at the beginning of the experiment, it
varied with time due to boiling. The breakthrough curves of water sorption on these adsorbents
at different water contents showed that among the molecular sieves examined, type 3A
molecular sieves gave the best separation of the ethanol–water system and among biobased
adsorbents examined, natural palm stone was the best.
The Guggenheim, Anderson, and De Boer (GAB) model at different water contents
represented the isotherms for water sorption on molecular sieves and biobased adsorbents. In
addition, the surface area of adsorbents and maximum water uptake were calculated. Type 3A
molecular sieves were determined to have the largest surface area and the highest value of water
uptake compared with the other two types of molecular sieve. However, for the biobased
adsorbents, activated palm stone was found to have the largest surface area and the highest value
of water uptake compared to the other biobased adsorbents. It was found that upon activation, the
adsorption capacity of biobased adsorbents was not enhanced. On the contrary, the activation
gave negative results on some occasions.
Keywords: azeotrope; ethanol–water; molecular sieves; biobased sorbents.
INTRODUCTION
Ethanol is considered as one of the most useful synthetic
oxygen-containing organic chemicals. This is due to its unique
combination of properties as a solvent, a beverage, antifreeze,
a fuel, a depressant, and especially its versatility as a chemical
intermediate for other organic chemicals (Othmer, 1981).
Usually it is produced in an aqueous form either by
synthetic methods or fermentation processes. To obtain
pure ethanol, the conventional separation methods of
water removal include distillation processes. However,
these are inapplicable since at atmospheric pressure ethanol
forms an azeotrope with water at an ethanol composition of
95 wt% and a temperature of 78.1�C (Walas, 1985). There-
fore, special methods are used to break the azeotrope; these
include azeotropic distillation or extractive distillation where
ethylene glycol is added as a solvent (Treybal, 1980). In
addition, a membrane process known as pervaporation has
recently been used to break the ethanol–water azeotrope
using polyvinyl alcohol membranes (Cussler, 1997; Hickey
and Slater, 1990; Strathmann and Gudernatsch, 1991; Choi
et al., 1992; Bruschke, 1990). Pervaporation involves the
preferential passage of ethanol through a dense membrane
matrix. However, these polymeric membranes were not
found to be suitable for applications involving high water
concentrations or to de-water ethanol to very low dryness.
Zeolite-based membranes have been found efficient in
separating the ethanol–water azeotrope (Shah et al., 2000;
Nomura et al., 1998).
One of the methods used in separating water from ethanol
is adsorptive distillation. The adsorbents used are called
desiccants. These include molecular sieves, chloride salts,
silica gel, and biobased desiccants. Type A molecular sieves
are the most used in the application of ethanol–water
azeotrope separation. Sowerby and Crittenden (1998)
studied the recovery of ethanol from the ethanol–water
system at the azeotrope composition by adsorption using
855
*Correspondence to: Dr S. Al-Asheh, Department of Chemical
Engineering, Jordan University of Science and Technology, PO Box 3030,
Irbid 22110, Jordan.
E-mail: alasheh@just.edu.jo
{Currectly on sabbatical leave at Qatar University, Department of Chemical
Engineering, PO Box 2713, Doha, Qatar
E-mail: alasheh@qu.edu.qa
fixed beds packed with various pore sizes of type A
molecular sieves. They found that type 4A are superior to
type 3A molecular sieves, with higher adsorption capacities,
while the type 5A molecular sieves were determined not to
be suitable since there is a great possibility of ethanol
adsorption within their pores. However, it was found that
the use of a type 3A molecular sieve for drying the ethanol–
water azeotrope instead of a type 4A is more efficient in
terms of energy saving since the regeneration energy
requirements for type 3A are lower than those of type 4A
(Trent, 1993).
It has been shown that biobased desiccants, which are
mostly composed of cellulose and starch, are also able to
remove water from a wide range of organics (Ladisch,
1997). Ladisch and Dyck (1979) first demonstrated the
separation of ethanol–water vapour mixtures over cracked
corn, starch and carboxymethyl cellulose, with a positive
energy balance compared with other adsorbents, such as
calcium oxide and sodium hydroxide. The authors found
that the total energy consumed for drying of ethanol by
calcium oxide is 3669 kJ kg�1 ethanol, while the energy
consumed using cellouse is 2873 kJ kg�1 ethanol (Ladisch
and Dyck, 1979). It was also found by the same authors that
the adsorption process using biobased adsorbents is more
energy-efficient than azeotropic distillation. The energy
consumption in azeotropic distillation for the ethanol–
water azeotrope is about 88 kJ m�3 ethanol, which is
higher than the energy consumption needed for the adsorp-
tion process using corn grits by 32 kJ m�3 ethanol (Ladisch
and Dyck, 1979). This indicates that the adsorption process
is less expensive than the azeotropic distillation process (Lee
and Ladisch, 1987). Corn grits are currently being used in
industry to dry over 750 million gallons of ethanol per year,
with 99.8% dryness by weight (Beery and Ladisch, 2001).
However, biobased desiccants are biodegradable and are
derived from renewable resources (Lee and Ladisch, 1987;
Beery and Ladisch, 2001).
In this work, new biobased desiccants that are able to
separate the azeotropic water–ethanol mixture are devel-
oped. Molecular sieves were taken as reference materials for
comparison purposes. The water content in the feed solution
was varied and the breakthrough behaviour at each water
concentration was determined. The isotherms of water
sorption are also presented and discussed at different inlet
water contents.
MATERIALS AND METHODS
Desiccants
Three different types of molecular sieve, namely type 3A,
type 4A, and type 5A, were used in this work. These were
purchased from Scharlau Chemie (Spain) and The Asso-
ciated Cement companies Ltd. (India). The molecular sieves
were dried in an oven at temperatures of 190–210�C for 24
hours. This is to make sure that all humidity within the
crystals of molecular sieves is evaporated (Trent, 1993). The
dried molecular sieves were kept in bottles, which then
were stored in a glass chamber. The glass chamber contained
silica gel to make sure that no humidity would be sorbed on
the molecular sieves.
Palm stones (date pits), oak, and corncobs were used in
this work as natural biobased adsorbents. The palm stone
and oak were washed with distilled water and then dried in
an oven at temperatures of 70–80�C for 24 hours. The dried
natural palm stone and oak were crushed and dried again.
The dried palm stones and oak were crushed, milled and
sieved into different particle sizes. They were then kept in
bottles and stored in a glass chamber, which contained silica
gel. Studies were focused on the size fraction of 0.125–
0.212 mm. Corncobs were cut into small pieces with a knife
and washed with distilled water before being dried in a
packed bed using nitrogen at a temperature of 90�C for six
hours. This was to ensure that no biological degradation
occurred to the polysaccharides within the corncobs
(Westgate and Ladisch, 1993). The dried corncobs grits
were then kept in bottles that are stored in a glass chamber
containing silica gel.
Crushed palm stone and oak also underwent a physical
activation process using a vertical fluidized backed column
oven. The procedure of activation is summarized as follows.
The column was filled with the natural material. The
temperature of the bed was raised to 700�C, using electrical
heaters, in the presence of nitrogen gas at a flow rate of 1.5–
2.0 cm3 min�1. The material was kept at 700�C for two
hours. After that, the temperature was raised to 900�C in the
presence of nitrogen gas. Pyrolysis gases were formed and
these were released to the atmosphere directly. When the
temperature reached 900�C, the nitrogen gas flow was
replaced by carbon dioxide gas for 30 minutes. At the
end, the heaters were switched off and cooling was allowed.
The activated materials were then removed from the column.
They were dried in the oven at temperatures of 190–210�C
for 24 hours to make sure that no humidity would be sorbed
on the activated materials during the cooling process. The
dried activated biobased adsorbents were kept in bottles that
were stored in a glass chamber containing silica gel. Raw
palm stones and activated palm stones have BET surface
areas of 1.56 and 870 m2 g�1, respectively, and micropore
surface areas of 0.19 and 364 m2 g�1, respectively (Guo and
Lua, 2001).
Bench-Scale Fixed Bed Adsorber
A bench-scale fixed bed adsorber apparatus has been
designed and is shown schematically in Figure 1. The
adsorbent was packed in the fixed bed (14.6 mm ID (internal
diameter); 25 cm depth). The column wall temperature was
maintained at the initial bed temperature during the experi-
ment by circulating hot water through the jacket. The jacket
was insulated with heating tape.
In a typical experiment, vapour was boiled up from a
500 ml flask surrounded by an electric heating mantle. Heat
input was controlled by a variable transformer. The tempera-
ture of the mixture in the boiling flask was measured by a
thermocouple. The atmospheric pressure was assumed
constant at 760 mmHg. The temperature of the jacket was
controlled and kept constant during the runs. The exit
stream was condensed using water as a cooling medium.
The condensate was removed every four minutes; a volume
of about 6 ml was collected. Thus, the average flow of the
condensate was about 1.5 ml min�1. This was achieved by
controlling the heat input by manipulating the power supply.
A Karl Fischer titrator measured the water content of
the condensate. At the end of experiment, the adsorbent
was removed from the bed and dried for further use.
Trans IChemE, Part A, Chemical Engineering Research and Design, 2004, 82(A7): 855–864
856 S. AL-ASHEH et al.
The adsorbents were efficient even after 10 cycles of the
adsorption–drying process.
The aqueous ethanol mixtures were prepared at four water
levels, namely, 5 (the azeotropic concentration), 8, 10, and
12 wt%. They were prepared by weighing the desired
amounts of distilled water and ethanol (Frutarom Ltd.,
UK) to make about 350 ml solution.
RESULTS AND DISCUSSION
Breakthrough Curves Using Molecular Sieves
Three molecular sieves of type A were used to study the
breakthrough behaviour, a sudden increase in water concen-
tration in the effluent, of water sorption using four different
water concentrations of ethanol–water mixtures. The break-
through curve of water sorption is plotted for the three
different pore sizes of molecular sieve at the different water
concentrations. The results are shown in Figure 2 for water
contents of 5, 8, 10, and 12 wt%, respectively.
It is seen (Figure 2) that the time required to reach a
plateau value varies as a function of the inlet water content
of ethanol–water mixture. Note that some curves do not
show the plateau behaviour since more time was required for
these curves to attain the plateau value and these data are not
shown. The time needed to reach a plateau value in the case
of 5 wt% was about 144 minutes. This is longer than that of
other water contents, which is about 120 minutes. It is also
noted (Table 1) that the plateau time, time at which the water
content in the exit becomes the same as the inlet, becomes
shorter as the water concentration increases. This behaviour
is attributed to the fact that the 5 wt% water content
represents the azeotropic composition of the mixture; i.e.,
the compositions of ethanol and water in the liquid and
vapour phases are the same. The separation of ethanol from
water is only achieved by the selective sorption of the
molecular sieves used. It is also worth mentioning that
the breakthrough behaviour likely resulted in less than
complete water removal given that the column was relatively
short.
To distinguish between the three types of molecular sieve,
the average water content in the samples of the effluent
stream was considered as well as the calculated breakthrough
time for each type of molecular sieve used. The breakthrough
time is arbitrarily designated to be the time required to reach
2 wt% of water content in the effluent stream. This value was
chosen because it happens that water concentration in the first
droplet of condensate was greater than 1 wt% for some of the
sorbents. It is important to emphasize that the data are plotted
as water content in effluent stream divided by initial water
content (C=C0) versus time, therefore, the breakthough time
varies depending on the initial water content. A 2 wt% was
selected since the sorption of water on type 5A molecular
sieves usually gives outlet water content more than 1 wt%.
The average content of the effluent and the breakthrough time
for the different types of molecular sieve are displayed in
Table 1.
It can be concluded from the data of Table 1 that at a
given water content, type 3A molecular sieves has the best
sorption performance, followed by type 4A, and lastly by
type 5A. The longer the breakthrough time and the lower the
average water content in the outlet, the better is the sorption
performance. The diameter of the ethanol molecule, which
is 4.46 A˚ (Westgate and Ladisch, 1993), is the main deter-
mining factor of the performance of all types of molecular
sieve. Since the size of the ethanol molecule is less than the
pore diameter of type 5A molecular sieves, ethanol
molecules can pass the pores of the molecular sieves and
be sorbed there. However, as the diameter of the water
molecule is 2.8 A˚ (Trent, 1993), which is less than any of the
three types of molecular sieve used in this work, the three
types of the molecular sieve can easily sorb water mole-
cules. The higher adsorption capacity of type 3A molecular
sieves can be attributed to its high surface area compared to
that of type 4A and type 5A molecular sieves (Sowerby and
Crittenden, 1988).
It was expected that when molecular sieves were exposed
to more water vapour, the breakthrough time would be
shorter and the average outlet water content would
be higher. This was verified by the results of Table 1. This
can be related to the fact that as the water content is
increased in the feed, the amount of vapour that passes
through the bed will be higher, thus more water molecules
will pass through the pores. For example, for type 3A
molecular sieves, the breakthrough time was shortened
from 88.9 to 61.2 minutes when the water content was
increased from 5 to 12 wt%. Moreover, the average water
content in the condensate increased to 2.01 mol l�1 at
12 wt%, which is about three times larger than the average
water content of condensate when using 5 wt% inlet water
content.
Figure 1. Schematic diagram of bench-scale fixed bed adsorber.
Trans IChemE, Part A, Chemical Engineering Research and Design, 2004, 82(A7): 855–864
SEPARATION OF ETHANOL–WATER MIXTURES USING MOLECULAR SIEVES 857
The main conclusion of these results is that the best
type of molecular sieve that can be used for drying of
ethanol–water mixtures type 3A. Therefore, the perfor-
mance of type 3A was considered as a datum to which the
performance of the examined biobased adsorbents was
compared.
Breakthrough Curves Using Biobased Adsorbents
Three types of biobased adsorbents were utilized for the
drying of ethanol–water mixtures using four different initial
water contents. These biobased adsorbents were: palm stone,
oak (with and without activation) and natural corncobs.
The breakthrough curves of these biobased adsorbents are
illustrated in Figure 3 using water contents of 5, 8, 10,
and 12 wt%, respectively. The shape of the breakthrough
curves obtained for these biobased adsorbents was similar
to those obtained when type A molecular sieves were used
(Figure 3).
It can be seen that the time needed to reach a plateau
value using the biobased adsorbents is shorter than that for
the molecular sieves (Figure 3). This can be attributed to
the uniform and non-uniform structures of the molecular
sieves and the biobased adsorbents, respectively. As such,
there is a possibility of ethanol being sorbed on the biobased
adsorbents. Lee et al. (1991) studied the sorption pheno-
Figure 2. Breakthrough curves for ethanol–water system during water sorption on different types of molecular sieve using (A) 5 wt%; (B) 8 wt%; (C) 10 wt%;
and (D) 12 wt% inlet water content.
Trans IChemE, Part A, Chemical Engineering Research and Design, 2004, 82(A7): 855–864
858 S. AL-ASHEH et al.
mena of ethanol and water on starch, which is the main
constituent of the biobased adsorbents, and found that
ethanol sorption can be assumed to be negligible compared
to water sorption by the starchy material. This assumption
can also be applied to the results of this work.
The breakthrough time and the outlet average water
content are reported in Table 2. The water concentration
for each outlet-collected sample was determined from the
knowledge of concentration and volume of samples collected
at different time intervals. The average concentration of all
outlet-collected samples, Cw,avg., was then calculated using
the following expression:
Cw,avg: ¼
Pn
i¼1 ViCw,iPn
i¼1 Vi
(1)
where Cw,i is the water concentration in ith sample (mol l
�1),
Vi is the volume of the ith sample (ml), and n is the number of
outlet-collected samples. The data of Table 2 can be used to
compare the performance of the biobased adsorbents to that
of the molecular sieves. It is obvious that natural palm stone,
among the tested biobased adsorbents, is the best one for
ethanol drying; it showed the lowest average exit water
content and the longest breakthrough time compared to the
other tested materials. This is true for all inlet water contents
(Table 2). This would indicate that the surface-active func-
tional groups in palm stone, which are believed to be poly-
saccharides, are more numerous than those in the other
biobased adsorbents. It is also worthy mentioning that acti-
vation of palm stone and oak did not improve their sorption
performance. On the contrary, sorption performance of these
biobased materials deteriorated upon activation (Table 2).
Activation was expected to change the structure of palm
stone and oak, so as to make them more porous materials,
meaning that there would be more possibility of water
molecules being sorbed. However, this is not the case here.
It might happen that the sorption process of water within the
biobased adsorbents is determined by chem