J . Chem. Tech. Biotechnol.

1992, 53, 181-188

Ethanol Dehydration by Extractive Distillation

Antonio Meirelles
Department of Food Engineering, State University of Campinas (UNICAMP), 13081, Campinas-SP,
Brazil

Siegfried Weiss & Herbert Herfurth

Sektion Verfahrenstechnik, Technische Hochschule ' Carl Schorlemmer ', Merseburg 4200, Germany
(Received 4 May 1991 ; accepted 29 August 1991)

Abstract : Experiments and process simulation enable favourable operating

conditions to be determined for the extractive distillation of ethanol-water, with
ethylene glycol as a solvent. The solvent molar rate to feed molar rate ratio
S / F = 0.6 and the reflux ratio R = 0.5 were determined in order to achieve at least
99.5 mole % purity of ethanol. Other parameters examined include feed
concentration, purity of solvent, and number of plates.
Extractive distillation can be used to achieve high purity of ethanol at low
energy consumption and under simple operating conditions.

Key words : ethanol, anhydrous ethanol, distillation, extractive distillation,

azeotropic distillation, ethylene glycol

NOTATION s Solvent
T TOP
ALPHA Non-randomness parameter in the NRTL 1 Water
equation ALPHA = A L P + A L T x t 2 Ethanol
ALP Parameter of the NRTL equation 3 Ethylene glycol
ALT Temperature-dependent parameter of the
NRTL equation ("C-')
A Interaction energies in the NRTL equation
1 INTRODUCTION
(kJ/kmol) A = A A + A T x t
AA Parameter of the NRTL equation (kJ/mol) The petroleum crisis in the 1970s was the starting point
AT Temperature-dependent parameter of the of an increasing interest in renewable sources of energy.
Since that time, ethanol produced by fermentation has
NRTL equation (kJ/kmol "C)
achieved enormous importance, particularly in Brazil, as
F Feed rate of ethanol/water (kmol/s)
Compound indexes an octane enhancer added to gasoline, or as its complete
I,J
n Number of theoretical plates substitute. As a mixture with gasoline, it can be used to
P Top product (anhydrous ethanol) rate operate conventional petrol engines without significantly
(kmol/s) reducing their efficiency.',' However, for this purpose,
R Reflux ratio the ethanol must be free of water.
S Solvent rate (kmol/s) Azeotropic distillation, one of the possible distillation
t Temperature ("C) procedures for producing absolute ethanol, has often
Mole fraction of component i been i n ~ e s t i g a t e d ~and
- ~ is considered to be energetically
x,
favourable.'j By contrast, extractive distillation with
Subscripts ethylene glycol was considered to be very expensive with
respect to energy cons~ mption. '.According ~ to Black,'
B Bottom extractive distillation with ethylene glycol consumes at
F Feed least 0.69 of the enthalpy of combustion of pure ethanol
181
J . Chem. Tech. Biorechnol. 0268-2575/92/$05.00 0 1992 SCI. Printed in Great Britain
182 A . Meirelles, S. Weiss, H . Herfurth

for the purpose of separation, whereas azeotropic in the ethanol depends on the climatic conditions of the
distillation with pentane only requires 0,346. However, country. For example, the average temperatures pre-
Lynn and Hanson* demonstrated, by investigations with vailing in Brazil only require a purity of ethanol of 99.5
the help of an ethanol-water pseudo-equilibrium curve volume YO(at 20°C, 98.4 mole O h ) to prevent the mixture
in the presence of ethylene glycol, that extractive dis- with gasoline from separating.2
tillation may be energetically more favourable. Thermal
integration studies by Knapp and Doherty' confirms
these results. Some industrial plants in the USA already 2 EXPERIMENTAL
use ethylene glycol to produce anhydrous ethano1.l"
However, no experimental studies on the extractive The experiments were first carried out in a laboratory
distillation of ethanol-water have been reported in the column of 65 mm diameter, containing 24 plates with
literature. one bubble cap each. The nearly azeotropic mixture of
The purpose of this paper is to relate the authors' ethanol and water is fed into the column on the 15th
investigations on the extractive distillation of ethanol- plate, and the solvent on the 3rd plate (counted from the
water with ethylene glycol. The reliable data obtained top). Further experiments were then carried out on a
may serve as a base for unequivocal comparison with pilot-scale plant (Fig. I ) consisting of 2 columns of
other procedures concerning energy consumption and in 80 mm diameter; the main extractive column containing
the technological design of plants for producing pure 60 sieve plates with downcomer and the regeneration
ethanol. column containing 40 sieve plates with downcomer. The
A laboratory column and a pilot-scale plant were used water-ethanol mixture was fed into the main extractive
to obtain concentration profiles at different solvent column on plate 31 (counted from the top) and the
molar rate to feed molar rate ratios and reflux ratios, and solvent on plate 3. The bottom product of this column
the effects of these parameters on the purity of the was fed into the regeneration column on plate 19 and
ethanol were examined. The results were extended by processed to give pure ethylene glycol.
process simulation with the help of the NRTL equation Liquid samples required for representing the con-
for describing the phase equilibrium and the relaxation centration profiles were taken every three plates in the
method for calculating the column. laboratory column and about every six plates in the
In these experimental and process simulation studies, pilot-scale plant. Further samples were taken from the
the required purity of ethanol was fixed at a value of top and bottom products and from the two feed products
99.5 mole YO.The maximum permissible content of water in each plant. These samples were analysed by gas

8 m Platform
--

4m Platform
.-

I I
I I
///////// ////// / / / / / / / / /// ///////; 77777777

c, MAlN EXTRACTIVE COLUMN C2 RECOVERY COLUMN

f ROTA METER H CONDENSER
/? CONTAlNER Re REBOILER
s SAMPLE
Fig. 1. Technological scheme of the pilot-scale plant.
Ethanol dehydration by extractive distillation 183

TABLE 1
Experimental operating conditions S -m

Solvent to ,feed ratio Reflux ratio 06 -I

SIF R (L
W
a3
09-1
\WATER
p
\-ETHYLENE
GLYCOL iI-ETHANCL
Laboratory column 0.28 < S / F < 3.1 1 0.94 < R < 4.81 3z 12- I 0

Pilot plant 0.64 < S / F < 1.75 0.71 < R < 2.18 iI
chromatography, and the top product (pure ethanol) was
additionally checked with the help of a densimeter
i
during the experiments. 24
In the experimental studies, the solvent rate to feed
rate ratio S I F and the reflux ratio R were varied as
essential parameters. The ranges investigated are shown
in Table I . Fig. 2. Typical concentration profile from the laboratory
column: S / F = 0.82; R = 1.96.

3 RESULTS AND DISCUSSION greater part of the stripping column is approximately

constant.
3.1 Experimental results The temperature profile (Fig. 3) shows somewhat
unusual behaviour as compared with normal distillation
Figure 2 represents a typical concentration profile from processes, with a maximum temperature between the
the laboratory column. In most of the experiments, the two feed plates which is due to the greater solvent
ethanol concentration at the top amounts to more than concentration existing in this part of the column. The
99.5 mole %, and the ethanol does not contain any absolute value of the temperature difference between the
ethylene glycol. Its concentration changes abruptly on two parts of the column (above and below the feed plate
the two feed plates and reaches very small values at the for the water-ethanol mixture) depends on the SIF ratio
bottom. Starting from the top, the concentration of and reflux ratio and, as does the solvent concentration, it
water increases slowly, reaches a maximum near the decreases with an increased reflux ratio and increases
bottom, and then decreases again. The concentration of with an increased S I F ratio.
ethylene glycol between the two feed plates and in the Table 2 summarizes the most important results of the

TABLE 2
Results of Experiments Conducted on Pilot-scale Plant

Run Feed Solvent to .feed Reflux TOP Bottom

composition ratio ratio composition composition
-‘IF SIF R x2 T
x2B X3B

1 0.1 32 1.75 1.44 0.999 0~0000 0.929

2 0.133 1.51 1.50 0.999 0.0005 0.912
3 0.139 1.35 1.57 0.999 0.002 1 0.862
4 0.156 1.32 1.59 0.999 0~0000 0.877
5 0.132 1.31 1.30 0.999 0.00 I 7 0,865
6 0138 1.12 1.04 0.996 0.0240 0.852
7 0.141 1.08 2.0 1 0.998 0~0000 0.842
8 0.287 1.08 0.7 1 0.998 0.0043 0.798
9 0.152 1.06 1.16 0.998 0.0006 0.864
10 0.124 0.92 0.87 0.997 0.0027 0.855
11 0.151 0.90 2.18 0.996 0~0000 0.823
12 0.141 0.90 1.81 0.996 0.0023 0.832
13 0.151 0.90 0.94 0.997 0.001 1 0.822
14 0.150 0.72 1.29 0.995 0.0430 0.807
15 0.115 068 1.07 0.995 0.006 1 0.836
16 0.151 0.64 1.50 0.996 0.0033 0.806
184 A . Meirelles, S. Weiss, H . Herfilrth

experimental determination of the equilibrium data as

well as for the representation of these data by one of the
known model equations of excess Gibbs energy." This
leads to the fact that the description of the phase
equilibrium is not always satisfactory. Using tempera-
ture-dependent parameters, it was possible to consider-
ably improve the description of each binary mixture,
which increases the accuracy of the prediction of the
phase equilibrium for the overall water-ethanol-ethylene
glycol system.
The parameters for the ethanol-ethylene glycol mix-
ture were obtained by regression of the isobar data (1.013
24-1
B
70 90
L, ,
110
, *
130 150
I bar) according to Thorat and coworker^,'^ and they are
represented with a mean temperature deviation of 1.33"C
and a mean deviation of vapour composition of 0.0217
TEMPERATURE OC mole fraction. Using the NRTL parameters according to
Fig. 3. Temperature profile from the laboratory column : Gmehling and Onken," these values amount to 5.02"C
S / F = 1.30; R = 1.45. and 0.0332 mole fraction.
The parameters for the water+thylene glycol mixture
experiments conducted in the pilot-scale plant. At a S / F were obtained by regression of different isobar and
ratio as small as 0.64, it is still possible to achieve the isothermal data.12.15,1fi They represent the isobar data
required purity of ethanol. For a definite S j F ratio, the (0-996bar) according to Trimble and Potts16 with a mean
purity is guaranteed within a wide range of the reflux temperature deviation of 1.22"C and a mean deviation of
ratio, as is the case in experiments 7 to 9 within the range vapour composition of 0.0200 mole fraction. Using the
R = 0.71-2.01, and in experiments 10 to 13 in the range NRTL parameters according to Gmehling and Onken,17
R = 0.87-2,18. As is evident from experiment 8, pure these values amount to 4.36"C and 0.0488 mole fraction.
ethanol can still be produced at considerably higher The behaviour of the vapour phase was assumed to be
concentrations of water in the feed mixture. At the ideal. The vapour pressure was calculated by the Antoine
bottom, the ethanol concentration is very small, so that equation using the parameters according to Gmehling
almost no ethanol is lost. and Onken.14.17
Process simulation was carried out for a main
3.2 Process simulation extractive column with 24 theoretical plates, with the
solvent being fed into the column on the 2nd plate and
Further investigations of extractive distillation were the feed mixture on the 12th plate. The column efficiency
carried out with the help of process simulation, using a of the pilot-scale plant was assumed to be 40 % (24: 60 =
computer program according to the relaxation method." 0.4). This value was obtained on the basis of investi-
The simulation is based upon rigorous solution of gations on the Murphree efficiency carried out using the
material and enthalpy balances and equilibrium relation- concentration profiles obtained from the laboratory
ships. The phase equilibrium of the water-ethanol- column.
ethylene glycol mixture was described using an NRTL The Murphree efficiency was determined for those
equation with temperature-dependent parameters (see plates in the stripping column which exhibited appre-
Table 3 ) . ciable changes in the concentration of components
Strongly polar systems, such as water-thanol- (cf. Fig. 2), and they were then averaged over the whole
cthylcnc glycol, arc known to present difficulties for the range as a mean value. The results obtained for the

TABLE 3
Parameters of the NRTL Equation

Binury ALP" ALT A A (I,J) A T(I,J) MJ,O AT(J,O

mix turr ("C-I) (kJlkrnol) (kJfkmoPC) (kJlkmol) (kJ/kmoPC)

1-2 0.17 1 23 1 0.005 228 3 293.17 17.047 1 - 44 1.20 18.328 0

1-3 0.185894 - 1383.43 8.040 9 - 1445.97 -9.1 50 6
2-3 0.3704 - 13527.42 - 92.739 1 - 4 35 1.97 53.3769

'I
ALPHA = ALP+ A L T x t.
" A =AA+ATxt.
Ethanol dehydration by extractive distillation 185

a
U
m
5z 0.10 0:15 0.20
W
I- WATER CONCENTRATION IN FEED XIF
a
J
a Fig. 7. Purity of ethanol xPTversus water concentration in
feed x I F .

MOLE FRACTION x i
Fig. 4. Experimental points of experiment 14 compared with S/F=0.6 R = 0.5
the calculated concentration profile. u c I

( 9 9 . i
d 0
0.001 0.002 0.003 0.004
WATER CONCENTRATION IN SOLVENT XIS

Fig. 8. Purity of ethanol x Z T versus water concentration in

solvent xis.

to continue the investigations on the effects of various

I R = 0.5
process parameters by simulation with a high degree of
rr
2
a
II 0 R =?.O
accuracy. In these calculations the top rate was assumed
I I I I I to have the same value as the feed rate of ethanol in the
0.4 0.8 1.2 1.6 2.0
feed mixture.
SOLVENT - FEED RATIO S/F Figure 5 represents the purity of the ethanol over S / F
Fig. 5. Purity of ethanol x Z Tversus ratio SIF. at a constant reflux ratio. Ethanol of a purity of at least
99.5 mole YOcan still be prepared at a S / F ratio of 0.6.
This value agrees well with the experimental value of 0.64
(experiment no. 16).
J
Within a wide range of the reflux ratio (0.5 d R < 2.5)
the purity of ethanol shows small changes (cf. Fig. 6),
I S/F = 0 6 reaches a maximum near the value 1, and decreases
I- abruptly below 0.5. A further increase in reflux ratio
W

0.98
results in decreased purity of ethanol, which has also
> been observed in experiments in the laboratory column.
k i This is due to the dilution of the solvent in the column
LL
and accordingly a smaller effect on the relative volatility
of ethanol-water. The smallest reflux ratio of 071
realised in the experiments can be further decreased to
0.50.
The above results of the simulation were all obtained
for water concentrations of 15 mole % in the feed
efficiency, which range between 0.38 and 0.52, are clearly mixture. The effect of this value is represented in Fig. 7.
below the values found in conventional distillation. They The purity of ethanol depends little on the concentration
show a decreasing tendency with increasing solvent of water. Thus, the required purity can still be achieved
concentration. Similar results and the modelling of the at a water content of 20-0 mole YOwithout changing the
efficiency have been reported in the literature." favourable process parameters SIP = 0.71 and R = 0.5
Figure 4 represents the experimental points of ex- ( S I P = 0.71 corresponds to S / F = 0.6 at xlv= 0.15).
periment 14 in the pilot-scale plant and the concentration In the experiments carried out in the laboratory
profile obtained by simulation. They are in good column, it was found that the purity of ethanol can be
agreement with each other. This provided the possibility considerably decreased when the solvent is contaminated
15 C T B 53
186 A . Meirelles, S . Weiss, H . Herfurth

2.5

2.0
(L

1.5
LL
X
3
14I8 22 26 1.0
w
NUMBER OF THEORETICAL LT
PLATES n
0.5
Fig. 9. Erect of the number of theorctical plates n and the
position of the feed platc n p on the purity of ethanol.
I
I

by water. The simulation showed that for the process I 2 3 4

parameters stated above, a water content as low as SPECIFIC ENERGY f 0 3 k J/kg ETHANOL
0.3 mole % in the solvent does not ensure the required Fig. 10. Specific energy consumption at reboiler and energy
purity (Fig. 8). Since the solvent is fed into the column removed at condenser in dependence on SIF and R .
near the top, its content of water has a direct effect on the
purity of the ethanol. The requirement for anhydrous
ethanol necessarily results in the demand for a high pure affecting the purity of the ethanol. The reduction of the
solvent of about 99.9 mole %. In the separation of number of plates between the two feed positions will
ethylene glycol-water, this degree of purity can be make it impossible to achieve the required purity. Four
achieved at relatively low expense. plates in the stripping column are sufficient for stripping
The effect of the number of theoretical plates and the the ethanol from the bottom product, whereas the purity
position of the feed plate for ethanol-water is evident of the top product can only be achieved by using more
from Fig. 9. Retaining the number of plates located than 10 theoretical plates.
between the two feed positions, the overall number of
plates can be decreased to 16 plates without significantly

TABLE 4
Typical Plants for Producing Absolute Ethanol
(a) Main Column

Plunt Solcent to Reflux Feed Bo t t omr TOP Speclfic

pro& I ratio XI F
~- XZT energy
ratio R XlB X3B consumption
Slp kJlkg ethanol

1 0.7 1 0.5 01500" 0.1968 0.7989 0.9963 1528

2 0.7 1 0.5 0.200 0" 0.257 5 0.737 8 0.9950 1562
3 0.94 0.5 0.1500' 0.1550 0.841 1 0.9955 634

(b) Recovery Column

Plant Reflux Bottom Top Specific

ratio XL< energy
R XI, XPT consumption
kJlkg ethanol

1 0.2 09998 09780 00211 232

2 0.1 0.9998 0.9812 0.0182 302
3 0.3 0.999 8 0.9736 0.025 0 252
~~ ~~~~ ~ ~

Saturated-liquid feed.
" Saturated-vapour feed.
' Feed composition to recovery column.
Ethanol dehydration by extractive distillation 187

3.3 Considerations with respect to energy saving considerations concerning the energy consumption, the
results are modified, taking account of energy-integrated
The specific energy consumption of the main extractive distillation plants. However, extractive distillation is also
column is represented in Fig. 10. As is well known, it is of advantage in this case.
strongly dependent on the reflux ratio, but very little on Extractive distillation is characterised by additional
the S I F ratio; this means that ethanol of almost any favourable properties, such as flexibility and a simple
degree of purity can be produced without significantly mode of operation, which are lacking in azeotropic
changing the energy consumption, by increasing the S l F distillation. Here, the two columns, the azeotropic
ratio and maintaining the reflux ratio at the smallest distillation main column and the recovery column, as
value. well as the separator, are closely connected to each other,
On the basis of the experimental investigations and so that every small disturbance in one of these parts will
process simulation, an SIP ratio = 071 and a reflux directly affect the whole plant. The reflux ratio strongly
ratio R = 0.5 can be fixed as favourable process para- depends on the concentration of the feed stream, so that
meters. Energy consumption in extractive distillation can a small increase in the water concentration, e.g. from 15
be further reduced by feeding a vaporous feed mixture to 17 mole YO,requires a corresponding increase in the
into the main column, since the vapor phase mainly reflux ratio in order to achieve the required degree of
consists of ethanol. However, the vaporous feed reduces ethanol purity. Further, the entrainer and water balances
the purity of the ethanol, so that an ethanol con- have to be strictly checked in order to ensure both the
centration of 99.55 mole YO is only achieved at purity of the ethanol and the mechanical separability of
S I P = 0-94 and R = 0.5. the overhead product. This complex character of the
Table 4 summarizes the most important data obtained operating conditions requires an accurate controlling
for three typical plants for producing pure ethanol. The technique and sometimes leads to low recovery of alcohol
column used for recovering the ethylene glycol contains in the azeotropic column.
10 theoretical plates. Feed plate 5 (counted from the top) In extractive distillation, the main column and the
is operated at a pressure of 20.25 KPa (bottom tem- recovery column are largely independent of each other.
perature of about ISOOC) in order to avoid high The concentration of the feed mixture has little effect on
temperature. The reflux ratio of the recovery column is the operating conditions. Disturbance or changes of
the smallest value that can ensure a pure solvent and an operating conditions, which may occur in the mash
overhead product free of ethylene glycol. High recovery column, e.g. increased water content in the overhead
of ethanol in the extractive column with values above product, can be largely compensated in the main
99.5 YOis reached. Small amounts of ethanol, entrained extractive column, with the process parameters remain-
together with water at the top of the recovery column ing the same, or with a small increase of the SIF ratio.
may be added to the mixture fed into the mash column, The characteristics of the temperature profile may be
which produces the feed mixture for the main extractive used for an efficient control of the column. The
column. temperature difference at the feed plate for the water-
ethanol mixture increases when more water is removed
overhead, and decreases when more ethanol is removed
4 CONCLUSIONS in the bottom. This fact can be used as a possibility of
controlling head product purity and ensuring a bottom
The favourable process parameters established in this product free of ethanol.
paper, particularly those for the reflux ratio, lead to
reduced energy consumption in extractive distillation, as
is evident from a comparison of the energy consumption
per kg ethanol with that of azeotropic distillation in REFERENCES
Table 5. When the mash column is included in the
1. Trindade, S. C. & Yand, V., Brazil’s gasohol program.
Chem. Eng. Progr., 75(4) (1979) 1 1.
TABLE 5 2. Pimentel, L. S., Biotechnology report: The Brazilian
Comparison with Azeotropic Distillation from Energy Require- ethanol program. Biorechnof. Bioeng., 22 (1980) 1989.
ment Point of View 3. Furzer, I. A., Ethanol dehydration column efficiencies
using U N I F A C . AIChE J., 31 (1985) 1389.
4. Prokopakis, G. J. & Seider, W. D., Feasible specifications
Main column Recovery column
in azeotropic distillation. AIChE J., 29 (1983) 49.
(kJfkg ethanor) (kJ/kg ethanol) 5. Meirelles, A., Herfurth, H., Weiss, S. & Schubert, H.,
Azeotropdestillation von ethanol-wasser mit cyclohexan
Proposed process 1528 232 als schleppmittel. Chernische Technik, 39(8) (1987) 331.
Azeotropic distillation 6. Black, C., Distillation modeling of ethanol recovery and
with Pentanefi 2615 696 dehydration processes for ethanol and gasohol. Chem.
Eng. Progr., 76(9) (1980) 78.
188 A , Meirelles, S. Weiss, H . Herjiurth

7. Lee, Fu-Ming & Pahl, R. H., Solvent screening study and thermodynamic properties and isobaric vapour-liquid
conceptual extractive distillation process to produce an- equilibrium data : binary system, ethanolkthylene glycol.
hydrous ethanol from fermentation broth. Ind. Eng. Chem. Znd. Chem. Engineer, 20(3) (1978) 37.
Proc. Des. Deu., 24 (1985) 168. 14. Gmehling, J. & Onken, V., Vapor-liquid equilibrium-
8. Lynn, S. & Hanson, D. N., Multieffect extractive dis- data collection. DECHEMA Chemistry Data Series, Vol.
tillation for separating aqueous azeotropes. Znd. Eng. I, Part 2C-Organic Hydroxy Compounds : Alcohols
Chem. Proc. Des. Deo., 25 (1986) 936. (Supplement l), Frankfurt/Main, Germany, 1982, p. 297.
9. Knapp, J. P. & Doherty, M. F., Thermal integration of 15. Van Ness, H. C., Villamatian, M. A. & Gonzalez, C.,
homogeneous azeotropic distillation sequences. AZChE J., Excess thermodynamic properties for water/ethylene gly-
36 (1990) 969. col. J. Chem. Eng. Data, 29 (1984) 427.
10. Murtagh, J. E., Fuel ethanol production-The US ex- 16. Trimble, H. M. & Potts, W., Glycol-water mixtures: vapor
perience. Process Biochemistry, 21 ( 1986) 61. pressure-boiling point--compositions. Ind. Eng. Chem., 27
1I . King, C. J., Separation Processes. McGraw-Hill, New (1935) 66.
York, 1980. 17. Gmehling, J. & Onken, V., Vapor-liquid equilibrium-
12. Nath, A. & Bender, E., Isothermal vapor-liquid equilibria Data Collection, DECHEMA Chemistry Data Series,
of binary and ternary mixtures containing alcohol, alkanol- Vol. I , Part IA-Aqueous Organic Systems (Supplement
amine, and water with a new static device. J. Chem. Eng. l), Frankfurt/Main, Germany, 1981, p. 173.
Data. 28 (1983) 370. 18. Ark, R. & Weiss, S . , On the modelling of mass transfer in
13. Thorat, R. T., Nageshwar, G . D. & Mene, P. S., Excess extractive distillation. Chem. Eng. Process. 21 (1987) 107.