Journal of the Taiwan Institute of Chemical Engineers 59 (2016) 8690

Contents lists available at ScienceDirect

Journal of the Taiwan Institute of Chemical Engineers

journal homepage: www.elsevier.com/locate/jtice

Short communication

Dehydration of methanol to dimethyl ether in a dual-catalyst system

catalytic distillation tower
Wei-Bin Su, Jyh-Haur Hwang, Hsun-Yi Huang, Tse-Kuei Chang
Rening and Manufacturing Research Institute, CPC Corporation, Taiwan, No.217, Min-Sheng South Road, Chia-Yi 60051, Taiwan

a r t i c l e i n f o a b s t r a c t

Article history: A novel design of a dimethyl ether (DME) catalytic distillation (CD) tower is explored. Due to the wide boiling
Received 28 January 2015 range among the reaction mixture components: water, methanol and DME, a dual-catalyst system CD tower
Revised 27 July 2015
is designed to minimize the tower temperature and catalyst deactivation rate. A DME CD tower model was
Accepted 28 July 2015
developed and validated with both VLE data and pilot test data, and the model so obtained agrees with the
Available online 17 August 2015
results published by An et al. The dual-catalyst system CD tower, wherein the upper section of bed is loaded
Keywords: with low temperature catalysts and the lower section of bed is loaded with high temperature catalysts, are
Dimethyl ether arranged such that both catalysts are used at optimum dehydration temperatures and the product DME can
Catalytic distillation still be condensed with normally available cooling water.
Dehydration 2015 Taiwan Institute of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
Tower design

1. Introduction
Methanol dehydration to DME as shown in Eq. (1) is a reversible
reaction [1]. The coproduct water is strongly adsorbed on the cata-
lyst surface and thus inhibits the dehydration rate [2]. The process
is exothermic and the heat of reaction should be removed from the
reactor. The CD tower [35] can replace one dehydration reactor and
two distillation towers used in the conventional DME process [6]. Ap-
plication of CD technology in DME synthesis offers many advantages,
such as higher conversion, energy saving and capital cost reduction.
However, only a few reports were published for this process [35].
CD tower pressure selection has been included in this study because
it was never been explored [35].
Starting with the vaporliquid equilibrium data for H2 OMeOH
DME system [710], the binary parameters for VLE methods were re-
gressed and validated. Combining available kinetic data [3] and the
VLE method, a pilot CD model was formulated and validated with pi-
lot test data. The activity and aging tests were performed in a bench-
scale xed bed to evaluate the commercial catalysts life. Instead of a
single-catalyst system, a dual-catalyst system CD tower is designed
and a model is setup to explore the relationship between operating
pressure and catalyst bed locations.

2CH3 OHCH3 OCH3 +H2 O (1) 2. Experimental

The concept to design a CD tower for undergoing methanol de-
hydration seems clear. The reactant methanol has an intermediate
boiling point among three species in the reaction mixture so that
methanol can be concentrated in the middle of the CD tower, where
catalyst is located. DME and water can be withdrawn respectively
from the top and bottom of the CD tower. Due to a large difference
in the boiling point between DME and water, the design pressure of
the tower shall enable the use of the available low cost heat sink and
heat source. For example, the dew point of DME is 40.3 C when the
CD tower pressure is 9 bar and therefore using 30 C cooling water as
heat sink to condense DME is impractical. Another consideration for
a CD tower is that higher tower pressure results in higher tempera-
ture in the reaction zone, thus the catalyst may be deactivated. The
Corresponding Author. Tel.: +886 5 2224171, fax: +886 5 2285590.
E-mail address: 078221@cpc.com.tw (W.-B. Su).
2.1. Pilot tower and test procedure
The pilot system consists of a fully instrumented CD column. The
inside diameter of the column was 76.2 mm and the height was
5.79 m. The reaction section was loaded with 949 g of dry Amberlyst
35 (A35) (Rohm and Haas). The distillation sections were packed with
20 mm Raschig rings. The column conguration is shown in Fig. 1.
Real-time temperature was monitored via on-line instrumentation.
Test data was collected under steady-state conditions.
Before start-up, the system was blanketed with nitrogen at
1.36 bar. After the column reached the steady state at total reux
for one hour, a 99.5% methanol was continuously fed to the reaction
zone. The control for reux ratio and liquid level across reboiler was
turned on. After the column was operated for six hours, six samples
were taken. The sampling points were S-1, S-2, S-3 and S-4, which
corresponded to 5.182, 3.962, 1.524 and 0.305 m height above the

http://dx.doi.org/10.1016/j.jtice.2015.07.037
1876-1070/ 2015 Taiwan Institute of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
 W.-B. Su et al. / Journal of the Taiwan Institute of Chemical Engineers 59 (2016) 8690 87

ID=76.2 mm 40

S-top 30

Pressure, bar
overhead
reflux
T1 20
1.524 m

Rectification T2 S-1
zone T3
feed
10
T4 S-2
1.829 m

Reaction
T5
zone 0
T6
0 0.2 0.4 0.6 0.8 1
H2 O mole fraction
T7
feed
Fig. 2. The vaporliquidliquid data of DME-H2 O at 100.11 C. xH2O , o yH2O , ----------
2.438 m

Stripping T8 S-3 PRWS-UNIFAC, UNIFAC.

zone T9

T10 S-4
S-btm
bottom
Fig. 1. The pilot CD tower conguration.
reboiler shown in Fig. 1. Other two samples, S-top and S-btm, were
taken at the outlet of the condenser and reboiler.
2.2. Aging and residual activity test
The catalyst aging test was performed with a feed of 50/50% by
weight methanol/soften water in a continuousdownow xed-bed
reactor. The experiments with 99.5% methanol feed were performed
to observe the residual catalysts activity after a period of operation.
A gradient packing method was used so that the catalyst bed would
have a nearly uniform temperature and minimum wall effects. Mass
balance was 99% and the loss material can be attributed to the
fugitive dimethyl ether in the overhead sample.
3. VLE method and parameters
Generally, a cubic equation of state (EOS) coupling with an activ-
ity model for liquid phase e.g., NRTLRK method can predict reliable
phase behavior. The WongSandler (WS) mixing rule was developed
to implement the EOS, allowing VLE data obtained at low pressures
to make good VLE predictions for non-ideal mixtures at high pres-
sures [11]. To simplify VLE method, programming parameter regres-
sion was performed on a commercial simulator, Aspen V8. The VLE
data include three sets of T-x-y data at 5, 8, and 11.2 atm for H2 O-
MeOH [7] and fourteen sets of P-x-y data for MeOHDME [8] and
H2 ODME [9], measured at temperatures up to 160 C. These pres-
sures and temperatures are within the range of operating conditions
of the CD tower, as shown later in this paper. The ternary data for VLE
method evaluation are four sets of P-x-y data at 60, 80, 100 and 120 C
[10].
The recommended PRWSUNIFAC [10], PengRobinson (PR) EOS
[8], RKSWSUNIQUAC [12], and NRTLRK [13] methods were evalu-
ated. To understand the differences among various property meth-
ods, RKSWSUNIFAC and UNIFAC method were included in the eval-
uation. The average absolute, maximum and minimum relative error
between the calculated and experimental VLE data for the ternary
system showed that the PRWSUNIFAC and RKSWSUNIFAC meth-
ods were comparable and superior to others. The VLE prediction by
UNIFAC method was reasonable but its deviation interval was quite
different. According to Pozo and Streeett [9], there existed VLL equi-
libria for H2 ODME system. Carefully checked with the VLLE data, the
theoretical UNIFAC method cant give a good prediction. This is shown
in Fig. 2. The regressed interaction parameters ki j for PRWSUNIFAC
method are 0.1113, 0.3654, and 0.0992 for H2 OMeOH, H2 ODME and
MeOHDME system, respectively. These results were used in the CD
tower simulation.
It should be noted that the VLE predictions derived from vari-
ous property methods via different commercial software, such as As-
pen Plus, Hysys or ChemCAD, arent always consistent. Even with
the same software, different VLE predictions may result from dif-
ferent versions. The parameters for PRWSUNIFAC method are only
suitable for use in Aspen V8. It is important that parameters for the
VLE method should be regressed with VLE data, so that the selected
method can give a good VLE prediction.
4. Kinetic model
The past studies for DME synthesis showed that suitable catalysts
are acidic ion exchange resins, e.g. A35 [35]. The reaction mecha-
nism was reported to t with either LangmuirHinshelwood (LH) or
EleyRideal (ER) kinetic model, and the LH model was found to be su-
perior [2]. The coproduct water competes with methanol for adsorp-
tion and inhibits the methanol dehydration rate [2,3,5]. So the water
concentration plays a key role in rate of reaction. The published ki-
netic data were conducted at temperatures in the range of 110135
[2], 70130 [3] C. Because the data from the above references cover
the reaction temperature in our column, the kinetic data reported in
[3] were used in this study.
5. Results and discussion
5.1. Pilot CD model validation
The pilot CD tower shown in Fig. 1 comprised ten stainless steel
sections, three for rectication zone, three for reaction zone, and four
for stripping zone. The length per section was 0.61 m. A thermocou-
ple was positioned at the middle of each section. So there are twelve
measurement points, including the condenser and reboiler. The sim-
ulated vapor compositions were compared with the four sampling
points located at the middle of each section. Including the condenser
and the reboiler, the total stage number was 18. The reactive stages
were number 7, 8 and 9. Table 1 summarizes the operating conditions
and outlet streams analysis of the pilot tests.
From tower conguration (Fig. 1), operating conditions (Table 1),
VLE method and kinetic data, the pilot CD model can be formulated
as a RadFrac block in Aspen V8. In the model, a parameter overall
88 W.-B. Su et al. / Journal of the Taiwan Institute of Chemical Engineers 59 (2016) 8690

140 1 40

120 0.8

MeOH Conversion, %
vapor weight fraction
30
Temperature, oC

100 0.6

20
80 0.4

60 0.2 10

40 0
1 2 3 4 5 6 7 8 9 10 11 12 0
Measurements 0 100 200 300
Days on Stream
Fig. 3. Temperature and mass concentration proles for pilot CD Run3. Experimental
data: o temperature,  MeOH,  H2 O, simulation data: dashed line and solid line. Fig. 4. Activity and aging test for A35 and A70. 99.5%MeOH-A35, o 50%MeOH-A35,
 99.5%MeOH-A70,  50%MeOH-A70.
column eciency is a needed input such that simulation results can
match the experimental data. The criterion is to minimize the objec-
tive function (OBJ) which can be a function of temperature, compo- possible reason is that the rst run didnt reach steady state. Although
nent concentration, or heat duty. Four temperature points along the the tower temperatures and owrate were stable, the initial charge in
pilot tower as shown in Fig. 1 were used to dene the OBJ in Eq. (5). the reboiler for start-up wasnt replaced completely after 6 hr of con-
The calculated eciency for Run2 and Run3 were 0.275 and 0.4, re- tinuous operation for Run1. The modeling results in Fig. 3 are fairly
spectively. However, the obtained eciency for Run1 was larger than consistent with the pilot test data. This model was then used to ex-
1.5. The reason for this discrepancy is explained later. The average plore the performance of a commercial CD tower.
eciency for the pilot tower, including three CD trays was found to
be 0.3375. The HETP of rectication and stripping zones packed with 5.2. Catalyst aging tests
Raschig rings equals 0.903 m.
Ob jective F unction The methanol conversion is dened as:
 2  2  2 MeOH
T2cal T2exp T3cal T3exp T9cal T9exp XMeOH = 100% (6)
= + + FMeOH
T2exp T3exp T9exp
 2 where MeOH and FMeOH are outlet and inlet methanol molar ow
T10cal T10exp rates. Fig. 4 shows the methanol conversion variation with different
+ (5)
T10exp feed mixtures in the bench test for both catalysts. It is known that
acidic ion exchange resins are made of divinyl-benzene cross-linked
After the eciency of 0.3375 was introduced to the pilot CD
sulfonated polystyrene. These catalysts can be damaged or poisoned
model, a typical recalculated result for Run3 is shown in Fig. 3. Also,
under some conditions such as exposure to excessive temperature
a similar prole for Run2 was obtained and not shown here. Because
or impurities. For example, A35 is recommended to be used below
the tower pressure was at 6.44 bar, the temperatures of reaction zone
150 C. If reaction temperature is close to the limitation, catalyst ac-
fell in the range of 118120 C. The vapor mass fraction proles of
tive sites would be lost and catalyst life shortened.
water and methanol along the CD tower are also shown in Fig. 3. The
For A35, the time on stream totaled 180 days and accumulated
simulated vapor mass fractions agree with the component analysis
aging test was 120 days. The initial conversion for the feed with 99.5%
data.
methanol was about 35.8%. After 50 day aging operation at 120 C the
In Table 1, the mass concentration deviation of tower outlet liq-
conversion decreased to 32.3%. The conversion further decreased to
uid streams, S-top and S-btm seems smaller from Run1 to Run3. The
21% at the end of 120 aging days. These results ensure that the activity
of A35 is sucient to catalyze methanol dehydration at 120 C and the
Table 1 catalyst remained active for more than 120 days.
Operating conditions and outlet streams analysis of the pilot tests. Lower activity but less deactivation rate for Amberlyst 70 (A70)
Run number Run1 Run2 Run3 was observed at the similar conditions (Fig. 4). Although accumulated
aging test in the latter run was only 30 days, the results still indicate
Operating conditions
A70 could replace A35 at reaction temperature higher than 120 C.
Feed rate, kg/hr 1.84 1.103 1.393
Feed stage 6 11 11
The recommended operating temperature for A70 is up to 190 C. Its
Tower pressure, bar 6.44 6.457 6.436 thermal stability has also been reported to be better than that of A35
Tower P, bar 0.0038 0.0065 0.0141 by Siril et al. [14]. Although A70 has lower acid site concentration, its
Reux ratio 3.056 2.437 3.923 activity can be enhanced by increasing reaction temperature so that
Distillate to feed ratio (wt) 0.369 0.548 0.486
methanol conversion is acceptable for industrial applications.
Condenser temp. C 30.0 30.0 29.7

Sample analysis 6. Dual-catalyst system CD tower

Component, wt% Exp. Calc. Exp. Calc. Exp. Calc.

H2 O 0.0 0.8 0.0 2.3 0.0 1.7

Incorporating the PRWSUNIFAC method, the CD tower model
S-top MeOH 67.5 40.0 42.2 53.0 42.6 47.5 proposed by An et al. [3] was rebuilt and simulated with the optimum
DME 32.5 59.1 57.8 44.7 57.4 50.8 input parameters of a CD tower. The calculated DME and water con-
H2 O 56.2 19.0 47.5 26.5 31.7 25.1 tent in overhead and bottom product was 97.2 wt% and 93.2 wt%, re-
S-btm MeOH 43.7 81.0 52.4 73.5 66.6 74.9
spectively. Its slightly different from the original work. If the property
DME 0.0 0.0 0.0 0.0 1.6 0.0
method was changed to the default NRTLRK, the calculated DME and
 W.-B. Su et al. / Journal of the Taiwan Institute of Chemical Engineers 59 (2016) 8690 89

Table 2
Comparison of the single-catalyst (A35) and dual-catalyst (A35/A70) systems CD tower at 9 and 11 bar.

Pressure, bar 9 9 9 11 11 11
Single-catalyst Dual-catalyst Dual-catalyst Single-catalyst Dual-catalyst Dual-catalyst

Total stage 30 30 44 30 30 42
Reactive stages 820 820 834 820 820 832
In upper section 812 812 812 811 811 810
Catalyst A35 A35 A35 A35 A35 A35
Temp. C 126130 129133 126130 58124 135139 75128
In lower section 1320 1320 1334 1220 1220 1132
Catalyst A35 A70 A70 A35 A70 A70
Temp. C 131137 133136 131138 133149 140145 133148
DME wt% 97.2 79.8 97.6 99.999 92.3 99.999
H2 O wt% 93.2 60.7 94.1 99.999 82.5 99.999

water content in overhead and bottom product was 99.98 wt% and
99.96 wt%. The different results between them may be caused by the
160
use of different VLE methods. Another reason is that different meth-
ods for molar volume estimation were used to calculate the reaction
rate. The method for computing the molar volume might be different
in the user-supplied subroutine. The third reason is different versions 140
of commercial software were used, because the work by An et al. was
performed in Aspen Plus 10.10.
The temperature proles in the reaction zone fell in the range
120

Temperature, oC
of 126137 C at tower pressure of 9 bar. In this case methanol
concentration below the reaction zone is about 80 wt%, with 20%
water. Fig. 4 indicates that A35 deactivated under the conditions of
120 C. The deactivation rate shall be faster at 130 C. This means 100
that a single catalyst system CD tower loaded with A35 cant provide
a satisfactory operation cycle. However A70 is much stable and has
lower deactivation rate. It could be a catalyst candidate for the CD 80
tower, although it shows lower activity than A35 [2]. The combina-
tion, dual-catalyst system design would optimize the catalyst activity
and result in a superior overall eciency.
60
6.1. Effect of bed design on pressure

Followings are some option conditions for DME production via the
dual-catalyst system CD tower. Table 2 shows the modeling results of
the dual-catalyst system CD tower at 9 and 11 bar.
It starts at the production of 97.2 wt% DME and 93.2 wt% water
from the overhead and bottom of a CD tower. Total stages are 30
and reactive stages are 8 to 20. The temperatures at reaction zone
are 126.3137.2 C. Then the reaction zone is divided into the upper
section of bed, staged 812 and 126.3130.3 C, for low temperature
dehydration and the lower section of bed, staged 1320 and 131.2
137.2 C, for high temperature dehydration. To lower the rate of cata-
lyst deactivation, attempts were made to keep the upper bed temper-
ature below 130 C and A35 in the lower section of bed was replaced
with A70. The average methanol conversion of A70 during activity
and aging test is 35% of A35, as shown in Fig. 4. The overall dehy-
dration rate of A70 was assumed to be 35% of A35. Since A70 is less
active, stage numbers across the lower section of bed are increased in
the dual-catalyst system tower model such that the same methanol
conversion could be obtained. In order to compare with the base case
model, other input parameters were not varied. Fig. 5 is the temper-
ature prole of the nal design of the dual-catalyst system CD tower
for producing 97.2 wt% DME at 9 bar. It requires an extra 14 reactive
stages in the lower section to compensate the less active A70 such
that the dual-catalyst system CD tower can give similar performance
as the single-catalyst system CD tower. The temperatures across the
lower section of bed are 130.7137.6 C, almost the same as that in a
single-catalyst system CD tower. However, its prole is atter.
The dew point of DME is 48.3 C at pressure of 11 bar. In this case,
33 C cooling water that is normally available in the industry can be
used. The single-catalyst system temperature is in the range of 57.7
40
0 10 20 30 40 50
Stage Number
Fig. 5. The temperature prole of a CD tower at 9 bar. ---------- Single-catalyst system,
Dual-catalyst system, Low temp reaction, High temp. zone with A35,
High temp. zone with A70.
148.6 C (Table 2). No organic catalyst can provide good and stable
performance with such a wide range of reaction temperature. One
solution is to use a dual-catalyst system design in the reaction zone.
Table 2 shows the temperature across the upper section of bed is in
the range of 74.8128.1 C so A35 can be used here. On the other hand,
the temperature across the lower section of bed is 133.2147.9 C. The
use of A70 is required.
7. Conclusions
Like all process models, the simulation results are dependent on
the kinetic data, VLE method, and physical property calculations. The
best VLE method for the systems composed of H2 OMeOHDME was
found to be PRWS or RKSWS property method in Aspen V8. The pi-
lot test data showed that methanol dehydration to DME can be car-
ried out in a CD tower equipped with A35 and A70. Therefore, a CD
tower model coupling with the VLE method and the rate equation
was developed and validated with our test data. The catalyst aging
tests demonstrated that ion exchange resins, e.g. A35 and A70, can be
deactivated by sodium ion in the 50/50% methanol/water feed mix-
ture. Besides catalyst deactivation, the tower pressure of 11 bar was
90 W.-B. Su et al. / Journal of the Taiwan Institute of Chemical Engineers 59 (2016) 8690
designed to use the low cost cooling water and steam. A dual-catalyst
system design was proposed and could be arranged and installed
with two different catalysts, the upper section of bed with A35 and
the lower section of bed with A70. In this way longer catalyst life can
be expected. The results presented in this study are valuable to the
development of a DME production process.
References
[1] Spivey JJ. Review: dehydration catalysts for the methanol/dimethyl ether reaction.
Chem Eng Comm 1991;110:12342.
[2] Hosseininejad S, Afacan A, Hayes RE. Catalytic and kinetic study of methanol de-
hydration to dimethyl ether. Chem Eng Res Des 2012;90:82533.
[3] An W, Chuang KT, Sanger AR. Dehydration of methanol to dimethyl ether by cat-
alytic distillation. Can J Chem Eng 2004;82(5):94855.
[4] Di Stanislao M, Malandrino A, Patrini R, Pirovano C, Viva A, Brunazzi E. DME syn-
thesis via catalytic distillation: experiments and simulation. Comput Aid Chem
Eng 2007;24:107782.
[5] Lei Z, Zou Z, Dai C, Li Q, Chen B. Synthesis of dimethyl ether (DME) by catalytic
distillation. Chem Eng Sci 2011;66:3195203.
[6] Mii T, Hirotani K. Economic Evaluation of a Jumbo DME Plant. WPC Asia Regional
Meeting, Shanghai, China, Sep. 1720; 2001.
[7] Hirata M, Suda S. Vaporliquid equilibrium under pressurized conditions:
experimental apparatus and the system methanol-water. Kagaku Kogaku
1967;31(8):75966.
[8] Chang E, Calado JCG, Streett WB. Vaporliquid equilibrium in the system dimethyl
ether/methanol from 0 to 180 C and at pressures to 6.7 Mpa. J Chem Eng Data
1982;27(3):2938.
[9] Pozo ME, Streett WB. Fluid phase equilibria for the system dimethyl ether/water
from 50 to 220 C and pressures to 50.9 Mpa. J Chem Eng Data 1984;29(3):3249.
[10] Ye K, Freund H, Sundmacher K. Modeling (vapor + liquid) and (vapor + liquid +
liquid) equilibria of {water (H2 O) + methanol (MeOH) + dimethyl ether (DME)
+ carbon dioxide (CO2 )} quaternary system using the Peng-Robinson EoS with
WongSandler mixing rule. J Chem Thermodyn 2011;43(12):200214.
[11] Huang H, Sandler SI. Prediction of vaporliquid equilibria at high pressures using
activity coecient parameters obtained from low-pressure data: a comparison of
two equation of state mixing rules. Ind Eng Chem Res 1993;32:1498503.
[12] Teodorescu M, Rasmussen P. High-pressure vaporliquid equilibria in the systems
nitrogen + dimethyl ether, methanol + dimethyl ether, carbon dioxide + dimethyl
ether + methanol, and nitrogen + dimethyl ether + methanol. J Chem Eng Data
2001;46(3):6406.
[13] Song H, Zhang H, Ying W, Fang DJ. Measurement and calculation of vaporliquid
equilibrium for dimethyl ethermethanolwater ternary system. Chem Ind Eng
(China) 2006;57:18716.
[14] Siril PF, Cross HE, Brown DR. New polystyrene sulfonic acid resin catalysts with
enhanced acidic and catalytic properties. J Mol Catal A: Chem 2008;279:638.