Catalytic Distillation in Structured Packings:

Methyl Acetate Synthesis

´
Andrzej Gorak
Dept. of Chemical Engineering, University of Dortmund, 44221 Dortmund, Germany
Achim Hoffmann
Dept. of Mechanical Engineering, University of Essen, 45117 Essen, Germany

Catalytic Distillation can impro®e process design, the design of column internals re-
quires special attention. The catalytic packing MULTIPAK facilitates effecti®e catalysis,
high separation efficiency, and a wide loading range simultaneously. In this work the
main characteristics of MULTIPAK } pressure drop, loading range, and separation
efficiency } are determined experimentally. The de®eloped nonequilibrium stage model
de®eloped reflects the complexity of catalytic distillation processes and comprises corre-
lations describing pressure drop and separation efficiency of MULTIPAK as a function
of the operating conditions. The model has been implemented in the simulation en®i-
ronment Aspen Custom Modeler, and the simulation results are compared with experi-
mental data for the synthesis of methyl acetate. This model reflects the process beha®ior,
but differences between simulated results and experimental data can still be obser®ed.

Introduction
The combination of heterogeneously catalyzed reaction and tion columns ŽFair, 1998.. However, designing a catalytic dis-
separation in one single process unit, known as catalytic dis- tillation column requires knowledge of device characteristics,
tillation, offers new potential for improved process design. as well as of reaction kinetics, and chemical and vapor-liquid
Even if not a new concept, it still bears a lot of advantages equilibrium. A suitable model is indispensable to reflect the
compared to the conventional approach where reaction and complex behavior of catalytic distillation columns. In recent
separation are carried out sequentially. One of those advan- years, great effort has been made to investigate the relevant
tages is the elimination of equipment for product recovery phenomena occurring in catalytic distillation processes. A de-
and recycling of unconverted reactants, which leads to sav- tailed overview of different aspects of reactive distillation
ings in capital and operation costs. Catalytic distillation may processes has been given recently by Taylor and Krishna
also improve selectivity, mass transfer and process control, Ž2000..
and avoid catalyst fouling. On the other hand, the installation The analysis of reactive distillation from a thermodynamic
of a solid catalyst in a distillation column conflicts with these point of view, as done by Ung and Doherty Ž1995., Nisoli et
benefits since it complicates the design of catalytic distilla- al. Ž1997., or Frey and Stichlmair Ž1999., allows to check if
tion processes. the process is feasible or to identify the presence of reactive
One of the major problems is the choice of suitable cat- azeotropes and attainable regions for product compositions.
alytic column internals. Those devices have to reconcile satis- Okansinski and Doherty Ž1997. outlined the important role
factory separation efficiency and catalytic activity without of uncertainties in model parameters at this early stage of
limiting possible column loads. Catalyst bales, covering a large designing reactive distillation columns. They become even
percentage of the column cross section, have been used suc- more important when the steady-state and dynamic behavior
cessfully in commercial-scale MTBE production, but high- of a reactive distillation column should be predicted. Models
capacity devices can lower capital costs for catalytic distilla- of different complexity have been presented to describe the
behavior of catalytic distillation processes. Bessling et al.
Ž1998. analyzed the process of methyl acetate synthesis via
Correspondence concerning this article should be addressed to . catalytic distillation using a simple model assuming chemical

AIChE Journal May 2001 Vol. 47, No. 5 1067

and vapor-liquid equilibrium on all stages. A steady-state
rate-based model for the production of diacetone alcohol with
catalytic distillation was developed by Podrebarac et al.
Ž1998b.. Higler et al. Ž1999. and Kenig et al. Ž1999. presented
a nonequilibrium model for reactive distillation columns us-
ing the Maxwell-Stefan equations to describe mass transfer
between the vapor and liquid phase. Sundmacher and Hoff-
mann Ž1996. proposed a detailed nonequilibrium model in-
cluding the mass-transfer resistance between the liquid phase
and the porous catalyst for the MTBE process. Bart and
¨
Landschutzer Ž1996. analyzed the influence of axial disper-
sion and varying model parameters on the performance of
catalytic distillation columns. Schneider et al. Ž1999. devel-
oped a dynamic model for the production of methyl-acetate
using reactive batch distillation. A comparison between equi-
librium and nonequilibrium models by Lee and Dudukovic
Ž1998. for reactive distillation columns for the production of
ethyl-acetate showed that the reaction rate is the main factor
that affects the column behavior, a fact that was also pointed
out by Podrebarac et al. Ž1998a.. Subwalla and Fair Ž1999.
used an equilibrium stage model to derive design guidelines
for catalytic distillation systems considering several key de-
sign parameters such as pressure, reactive zone location, re-
actant feed ratio, feed location, amount of catalyst, column
diameter, and the number of equilibrium stages in reactive
and nonreactive zones, but they recommend the use of a
rate-based model to determine the height of the column.
Extensive experimental work has been performed to inves-
tigate the behavior of reactive column internals. The hydro-
dynamics of catalyst bales used for MTBE production have
been determined by Zheng and Xu Ž1992., Xu et al. Ž1997.
Figure 1. Construction of MULTIPAK.
and Subwalla et al. Ž1997., while Huang et al. Ž1998. investi-
gated the mass-transfer behavior of a laboratory-scale pack-
ing for the aldol condensation. Recently, several investiga-
tions concerning the hydrodynamic behavior of the catalytic downwards through the catalyst bags. Sufficient mass transfer
packing Katapak-S have been presented by Moritz et al. between the gas and liquid phase and radial mixing is guar-
Ž1999., Moritz and Hasse Ž1999., as well as Ellenberger and anteed by the segmentation of the catalyst bags and numer-
Krishna Ž1999.. Since the reaction kinetics has been recog- ous contact spots with the wire gauze sheets. The catalyst
nized as a model-parameter of great influence, it has been bags can be filled with different types of particle shape cata-
determined experimentally for different catalytic distillation lyst, and in case of the methyl acetate synthesis, an acidic ion
processes. The synthesis of methyl acetate mentioned in this exchange resin ŽLewatit K2621, Bayer AG. was used. The ap-
study has been investigated by Xu and Chuang Ž1996, 1997. plicability of MULTIPAK in catalytic distillation has already
¨
and Popken et al. Ž1999, 2000.. been proven by a series of catalytic batch distillation experi-
This article presents a concept for the design of catalytic ments with the methyl acetate synthesis chosen as an exam-
distillation processes with the catalytic packing MULTIPAK, ple process ŽKreul et al., 1998; Schneider et al., 1999..
based on a nonequilibrium model. All model parameters Two different types of MULTIPAK can be distinguished.
needed for this approach have been derived from experi- The layers of corrugated wire gauze sheets of MULTIPAK-I
ments, independent of the catalytic distillation process. have an inclination angle of 60 degrees and a specific surface
area of 500 m2rm3, while MULTIPAK-II comprises sheets
with an inclination of 45 degrees and a specific surface area
of 750 m2rm3. Catalyst bags used for MULTIPAK-II are
Structured Catalytic Packing thicker than for MULTIPAK-I and allow a higher catalyst
The catalytic packing MULTIPAK ŽDE 197 01 045 A1, volume fraction in combination with a reduced specific sur-
1997. has been developed in consideration of the qualities of face area. Therefore, specific requirements of the process can
modern structured wire gauze packings and their improve- be taken into account with the choice of the packing type. To
ments on conventional distillation. MULTIPAK is manufac- ensure a similar hydrodynamic behavior of laboratory- and
tured by Julius Montz GmbH. It consists of corrugated wire technical-scale packings, both consist of the same elements.
gauze sheets Ž1. and catalyst bags of the same material Ž2. However, weld joints at the edges of the catalyst bags and the
assembled in alternate sequence, as shown in Figure 1. This gap between the packing and the column wall result in less
construction enables high column loads by providing open catalyst volume fraction, less specific surface area, and a
channels for the gas flow, while most of the liquid is flowing higher void fraction for laboratory-scale packings. This has to

1068 May 2001 Vol. 47, No. 5 AIChE Journal

Table 1. Characteristic Geometrical Data of the Investigated
Packings
A B
Type of packing MULTIPAK-I MULTIPAK-II
Column dia., d wmmx 98 52
Hgt. of packing segment HP wmmx 166 166
Catalyst vol. faction c Ca t 0,250 0,336
Specific surface area a wm2rm3 x 370 270
Void fraction e 0,643 0,580
Inclination angle u wdegx 60 45
Corrugation spacing S wmmx 7,5 5,6

be considered for the scale-up of reactive distillation pro-

cesses with MULTIPAK. The characteristic geometrical data
of the two different laboratory-scale packings which were in-
vestigated in this study are listed in Table 1.
Figure 2b. Pressure drop of packing B.
Packing Characteristics
In order to determine the pressure drop and loading range
while the straight lines have been derived by applying the
of MULTIPAK, experiments were performed for both pack-
hydraulic model by Rocha et al. Ž1993. to MULTIPAK.
ing types. They have been carried out in glass columns with a
The experimental data covers a wide range of possible col-
total packing height of 1 m at ambient pressure. Airrwater
umn loads in both cases, even if the gas load for packing A
has been used as a test system with a circulating liquid phase
was restricted to an F-Factor of 1.7 Pa0.5 due to limitations of
set at a constant temperature of 208C. Liquid and gas flow
the experimental setup. Therefore, flooding of packing A was
was measured with calibrated rotameters. To ensure a com-
only observed at higher liquid loads. Two different flow
plete wetting of the catalytic packing, the column was oper-
regimes similar to those of conventional structured packings
ated at maximum liquid loads before setting the desired gas
can be observed. At column loads below the loading point,
and liquid flow. When steady-state conditions were reached,
the slope of the pressure drop vs. gas load remains the same
the pressure drop over the packed section of the column was
as for the dry packing. A further increase of the gas flow
measured. The results of these experiments are shown in Fig-
results in a steeper slope, since the liquid is dammed up by
ures 2a and 2b, where the specific pressure drop of MULTI-
the gas flow and reduces the hydraulic diameter of the open
PAK is plotted vs. the gas load for different liquid loads. To
channels for the gas flow. Flooding of the packing can be
ensure a comparability with the process conditions in distilla-
observed at pressure drop above 10 3 Parm. The feasible col-
tion columns, the F-Factor
umn loads for MULTIPAK are very similar to those reported
by Moritz and Hasse Ž1999. for Katapak-S.
F s u G s ? rG
' Ž1. A comparison of the experimental data with a mechanisti-
cally-based model for structured packings ŽRocha et al., 1993.
was used for the gas load and the superficial liquid velocity shows inconsistent results for the two packings. While the
for the liquid load. Each dot represents an experimental point, pressure drop of packing A below the loading point can be
described with an accuracy of 30%, the pressure drop pre-
dicted for packing B is about two times higher than the ex-
perimental values. This can be caused by the wall effects in
small diameter columns. The weld joints at the edges of the
catalyst bags provide vertical channels for the gas flow and
therefore enable a bypass with reduced flow resistance. The
only modification to the original model has been made in
assuming a total wetting of the packing surface, due to the
excellent wettability of the wire gauze material used for
MULTIPAK. The hydraulic model of Rocha et al. Ž1993.
seems to be a suitable option to predict the pressure drop of
MULTIPAK in a bigger than laboratory scale. A statement
that has to be proven by further experiments with technical-
scale packings.
Apart from pressure drop and loading range, the separa-
tion efficiency of the packing plays an important role in the
design of catalytic distillation columns. Therefore, distillation
experiments have been carried out with the test system
chlorobenzenerethylbenzene, which has been recommended
Figure 2a. Pressure drop of packing A. by Onken and Arlt Ž1990. as a mixture for testing column

AIChE Journal May 2001 Vol. 47, No. 5 1069

internals at atmospheric pressure or vacuum. The distillation sheets are identical with MULTIPAK-I. Moritz and Hasse
column was equipped with 2 m of the catalytic packing A and Ž1999. determined a value of NTSM s 3 for the laboratory-
operated with total reflux. The column load was changed by scale KATAPAK-S.
varying the heat duty and the operating pressure from 10 to To enable the use of a nonequilibrium stage model for the
60 kPa, to cover the whole loading range of the packing. When design of reactive distillation columns equipped with MUL-
steady-state conditions were reached, liquid samples were TIPAK, mass-transfer coefficients and their dependence on
taken directly above and below the catalytic packing and ana- column loads have to be available. They can be determined
lyzed via gas chromatography. The separation efficiency char- from the total reflux distillation experiments described above.
acterized by the number of theoretical stages can be calcu- Following the concept of transfer units ŽChilton and Colburn,
lated from the Fenske-equation ŽStichlmair and Fair, 1998. 1935., an overall mass-transfer coefficient can be calculated
from
1 x D 1y x B
? ln ?
n th s
ln a ž x B 1y x D / Ž2.
˙
4L 1 x D 1y x B
?
1y x B

The application of Eq. 2 is limited to ideal binary systems; for

k OG s
p c t ,G d aH a y1
2
ln
ž x B 1y x D
qln
/ ž
1y x D / .

the used test system this requirement is only fulfilled for op- Ž5.
erating pressures above 40 kPa. However, for lower operating
pressures, the deviation from a mean relative volatility a is According to the generally adopted two-film theory, the rela-
below 0.2% ŽOnken and Arlt, 1990.. Equation 2 is therefore tionship between the overall height of the gas phase transfer
used for all operating pressures with a mean relative volatility unit HTUOG and the individual film transfer units HTUG and
calculated by HTUL is given by
xD
a sexp Hx ln a Ž x . dx. Ž3. HTUOG s HTUG q mHTUL . Ž6.
B

The number of theoretical stages per meter of the catalytic Even if the mass-transfer resistance in the liquid phase is al-
packing NTSM is given by most negligible in most distillation applications ŽOlujic et al.,
1999., the penetration approach is widely used to calculate
n th the liquid phase mass-transfer coefficient. It can be calcu-
NTSMs . Ž4. lated from
H

Figure 3 shows the NTSM-value of packing A as a function 0.9DL u L,eff

of the column load, given by the F-Factor. For the whole
range of column loads, the separation efficiency is at least
bL s 2 ( pS
Ž7.

four theoretical stages per meter. It remains constant for a

wide loading range of the packing, but increases up to NTSM ŽRocha et al., 1996. with the effective liquid velocity defined
s6 for lower column loads. This phenomenon was already as
reported by Pelkonen et al. Ž1997. for the conventional struc-
ture packing Montzpak A3-500 whose corrugated wire gauze u L, s
u L,eff s . Ž8.
e h L sin u

The liquid holdup h L was calculated according to the hy-

draulic model for structured packings presented by Rocha et
al. Ž1993.. Assuming that this approach is also suitable for
the catalytic packing due to the similar geometry of the gas-
liquid flow channels, the vapor-phase mass-transfer coeffi-
cient can be calculated from

k OG
bG s . Ž9.
k OG u L, s
1y m
bL u G, s

The experimental values of vapor-phase mass-transfer coeffi-

cients have been correlated in the well known form of dimen-
sionless groups. The following correlation

0.96 0.33
Figure 3. Separation efficiency of packing A. Sh G s 0.0064 ReG ScG . Ž 10 .

1070 May 2001 Vol. 47, No. 5 AIChE Journal

 Figure 5. HETP values for packing A.
Figure 4. Vapor-phase mass-transfer coefficients.

as shown in Figure 6 has been built up to investigate experi-

reproduces all experimental data within an accuracy of 15%, mentally the continuous process behavior. The column design
as shown in Figure 4. The dimensionless groups have been is similar to that presented by Krafczyk and Gmehling Ž1994.
defined according to Rocha et al. Ž1996. with or Bessling et al. Ž1998. with an inner column diameter of 50
mm and an effective packing height of 4 m meter comprising
u G, s 4 sections of 1 m each. A reactive section of 2 m is located in
u G,eff s . Ž 11.
e Ž 1y h L . sin u the middle of the column with two nonreactive sections of 1
m each below and above. The reactive section is equipped
A comparison with mass-transfer coefficients calculated from with MULTIPAK-I and the nonreactive section with Montz-
the correlation of Rocha et al. Ž1996. shows that these values Pak A3-500. High conversion can only be achieved if the re-
are considerably higher than the experimental values. This actants can flow countercurrently in the reactive section
ŽAgreda et al., 1990.; thus, methanol is fed to the column
has already been noted by Olujic et al. Ž1999. who explained
the significant difference with the unrealistic effective sur- below the reactive section, while the acetic acid feed is lo-
face area proposed by Rocha et al. Ž1996.. Since we assumed cated above the catalytic packing. Special attention has been
a total wetting of the wire gauze material, the effective sur- paid to allow for a comparison of experimental and theoreti-
face area is considerably higher. To enable a comparison, cal process data. The flow rates of feed-, product- and
HETP Žheight equivalent of a theoretical plate. values have reflux-streams are determined via mass-flow measurements
been calculated with both methods using the equation to ensure the accurate evaluation of process conditions. Liq-
uid samples can be taken from all streams mentioned above,
as well as along the column height. The samples have been
ln m u G, s uL , s
ql analyzed by gas-chromatography to determine the mass-frac-
HETP s
my1 ž bG a bL a / Ž 12.
tion of all organic components and Karl-Fischer titration to
ascertain the amount of water. Resistance thermometers have
and compared with the experimental data. The results for been installed for all inlet and outlet streams and along the
two experimental series are shown in Figure 5. It can be seen column height. This enables together with a heated jacket to
that the correlation presented here reflects the experimental prevent heat loss}a factor of great influence in small diame-
data within an accuracy of 20%, while the HETP values cal- ter columns}to draw an exact energy balance of the column.
culated with the method of Rocha et al. Ž1996. are consider- All data is recorded by a process control system and a data-
ably lower. Due to the stronger dependency of mass-transfer reconciliation ensures that the mass-balance of the column is
efficiency on the column load, the maximum deviation occurs fulfilled for all experimental runs.
at low column loads, while the theoretical curve approaches A series of experiments have been performed with a stoi-
the experimental values at higher loads. chiometric feed ratio of acetic acid and methanol. The pro-
cess conditions characterized by feed and distillate flow rates
and reflux-ratio are listed in Table 2 together with the experi-
Pilot-plant experiments mental conversion and the mass fraction of methyl acetate in
Theoretical and experimental results for the synthesis of the distillate stream.
methyl acetate via catalytic batch distillation with MULTI- Six experiments have been performed with feed mass flows
PAK have already been presented by Kreul et al. Ž1998. and of 1.4 kgrh and 2.0 kgrh. The reflux ratio has been varied
Schneider et al. Ž1999.. To evaluate the influence of the from 1.62 to 2.55 in experiments 1 to 5, while the other oper-
packing characteristics on the process, a pilot-plant column ating conditions Žfeed and distillate mass flow. has been kept

AIChE Journal May 2001 Vol. 47, No. 5 1071

 Table 2. Experimental Results
Total Feed Distillate Reflux Conv. of Purity of
Mass-Flow Mass-Flow ratio Acetic Acid Distillate
No. m˙ F wkgrhx m
˙ D wkgrhx RR jexp wexp wkgrkgx
1 1.39 1.04 1.62 0.761 0.809
2 1.39 1.05 1.99 0.806 0.851
3 1.41 1.06 2.39 0.803 0.827
4 1.41 0.99 2.51 0.806 0.881
5 1.41 1.00 2.55 0.790 0.861
6 2.01 1.65 0.97 0.827 0.782

at constant values. The experimental conversion of acetic acid

is well above the equilibrium value of 0.695 in all experi-
ments, but does not reach the high conversions reported for
the industrial process ŽAgreda et al., 1990.. Conversion can
be improved by increasing the height of the reactive section
as reported by Bessling et al. Ž1998. who compared a pilot-
plant reactive distillation column including 13 reactive stages
with an industrial column design comprising 38 reactive
stages. The reactive section in our experimental setup has
only 8 theoretical stages; thus, conversion is considerably
lower.
For the same reason, the purity of the distillate stream is
low in comparison to the industrial process. Since not all
methanol is converted to methyl acetate, it enters the upper
rectifying section of the column and is enriched towards the
top. However, the objective of the experimental work pre-
sented here was not to produce high-purity methyl acetate,
but providing experimental data of the process behavior to be
compared with theoretical results derived from process simu- Figure 6. Experimental setup.
lation.

Modeling rium of the system methanolracetic acidrrmethyl

A rigorous dynamic model for catalytic distillation pro- acetaterwater. They derived an adsorption-based model for
cesses has been presented by Schneider et al. Ž1999. combin- the heterogeneously catalyzed reaction choosing the ion-ex-
ing the dynamic balance equations required for batch opera- change resin Amberlyst 15 ŽRohm and Haas Co.. as catalyst.
tion and the rate-based approach for the appropriate de-
scription of multicomponent mass transfer. This model has
been modified in view of the availability of model parameter
for the catalytic packing MULTIPAK and experimental in-
vestigation described above. Thus, the time derivatives in the
balance equations have been set equal to zero, since only
steady-state simulations have been performed and compared
with experimental data.
For the nonequilibrium model, the reactive distillation col-
umn was subdivided into a number of segments. The model
for each segment j is based on the two-film theory ŽTaylor
and Krishna, 1993., and therefore consists of an ideally mixed
vapor and liquid bulk phase, as well as two film regions adja-
cent to the interface as shown in Figure 6. The catalyzed re-
action has been considered as quasi-homogeneous with no
mass-transfer resistance between liquid phase and catalyst.
This assumption has already been described as justified for
the synthesis of methyl acetate by several authors ŽKrafczyk
and Gmehling, 1994; Xu and Chuang, 1997.. However, sorp-
tion effects inside the ion-exchange catalyst should not be
neglected ŽKreul et al., 1998..
¨
Popken et al. Ž2000. recently published an extensive exper- Figure 7. Two-film model for a reactive packing seg-
imental study on the reaction kinetics and chemical equilib- ment.

1072 May 2001 Vol. 47, No. 5 AIChE Journal

This model does not only consider thermodynamic nonideali- The linearized theory of Toor, Stewart, and Prober ŽStewart
ties by using liquid-phase activities for the reaction kinetic and Prober, 1964. can be applied to evaluate w R av x, w G av x and
and equilibrium equations, but also sorption and swelling ef- all physical properties, using an average mol-fraction defined
fects. The physical properties of Amberlyst 15 are very simi- as
lar to those of Lewatit K2621 used here. However, the con-
centration of acid sites for dry Lewatit K2621 has been deter- yU
i , j q yi , j
yiav, j s is1 . . . n. Ž 20.
mined experimentally as 4.55 mol Hqrg Cat , close to the value 2
of 4.75 mol Hqrg Cat reported for Amberlyst 15. Therefore, this
model was adopted and, for consistency, the UNIQUAC ap- Phase equilibrium is assumed at the interface with interfacial
proach was used for vapor-liquid equilibrium with binary in- compositions calculated from
¨
teraction parameters also taken from Popken et al. Ž2000..
The component balances for the gas and liquid phase on a
non-equilibrium stage j were written as yU eq U
i, j s K i, j xi, j is1 . . . n Ž 24.

p with vapor-liquid equilibrium constants K i,eqj determined from

˙jy1 yi , jy1 yG˙j yi , j y ˙n i , j a
0 sG d 2D z is1 . . . n Ž 13 .
4 the UNIQUAC equations and the extended Antoine equa-
p tion for vapor pressure ŽReid et al., 1987..
˙ jq1 x i , jq1 y L
0s L ˙ j x i , j q Ž ˙n i , j aq ri , j ccat rcat . d 2D z. Since only n-1 diffusional fluxes are independent, the en-
4
ergy balances for both phases including the conductive heat
is1 . . . n Ž 14. flux across the interface are formulated to relate the diffu-
sional fluxes to the component molar fluxes ŽHigler et al.,
The component balances are completed with the summation 1999.. They are written as follows
conditions for the mol fractions in both phases
ad
˙jy1 h G , jy1 yG˙j h G , j y q˙j® q q˙Uj p dD z
n n
Ý x i , j s Ý yi , j s1. Ž 15 .
0 sG ž 4 /
is 1 is 1
is1 . . . n Ž 25.
The interfacial mass-transfer rates have been calculated with p
0s L ˙ j h L , j q q˙Uj ad 2D z
˙ jq1 h L , jq1 y L is1 . . . n Ž 26.
the Maxwell-Stefan approach to account for diffusional inter- 4
action and thermodynamic nonidealities. To relate the multi-
component mass-transfer rates to the binary mass-transfer The heat transfer across the vapor-liquid interface consist of
experiments described above, the method of Krishna and a convective and conductive part and can be calculated from
Standart Ž1979. was used. The diffusional fluxes can be calcu-
lated from
lav
g
n lav
l
q˙Uj sy Ž TJU yTG, j . q Ý ˙n i , j hG ,i , j sy Ž TL , j yTjU .
Ž J . sy c tav,G w k Gav x Ž yUi , j y yi , j . sy cTav, L w k Lav xw GLav x Ž x i , j y xUi , j . dG is1 dL
Ž 16. n
q Ý ˙n i , j hG ,i , j . Ž27.
is1
with the matrix of mass-transfer coefficients defined as
The film thickness in both phases is needed to calculate the
w k av x s w R av x y1 Ž 17. conductive interfacial heat transfer, but in multicomponent
systems only individual film thicknesses d i j can be derived
with
from binary mass transfer coefficients.
yiav n y kav
R ii s q Ý is1 . . . ny1 Ž 18 . }
Dij
k in k s1 k ik di j s Ž 28.
k/i ki j
1 1
R i j sy yiav
ž ki j
y
k in / i , js1 . . . ny1 Ž 19. Therefore, a mean film thickness has been defined as

1 n n
The binary mass-transfer coefficients k i j can be derived from ds Ý Ýd Ž 29.
suitable mass-transfer correlations for binary systems, using n Ž ny1 . is 1 js 1 i j
the appropriate Maxwell-Stefan diffusion coefficient. For the j/ i

reactive column section, the mass-transfer correlation for

MULTIPAK ŽEqs. 7]10. was used. The mass transfer in the This approximation is justified since all individual film-thick-
nonreactive sections was calculated according to Rocha et al. ness are similar and the conductive heat transfer across the
Ž1996., using the geometrical characteristics of Montz-Pak interface is small in comparison to the convective transfer. It
A3-500. has to be emphasized that the mean film thickness is only

AIChE Journal May 2001 Vol. 47, No. 5 1073

used to calculate the conductive heat transfer and not for the
evaluation of individual mass transfer across the interface.
The segmentation of the catalytic distillation column in ax-
ial direction together with the detailed rate-based approach
results in a complex and highly nonlinear algebraic system of
equations. All model equations have been implemented in
the commercial simulation environment Aspen Custom Mod-
eler. A first guess of the bulk phase compositions and tem-
peratures was provided by the solution of an equilibrium stage
model without reactions as done by Powers et al. Ž1988.. The
first convergent initialization of the rate-based approach was
achieved by implementing effective diffusion coefficients,
which leads to decoupled mass-transfer equations. This sim-
plified model has been extended by adding reaction terms
and the Maxwell-Stefan approach.

Results Figure 9. Liquid-phase composition profiles.

Theoretical and experimental results for the synthesis of
methyl acetate via catalytic batch distillation with MULTI-
PAK have already been presented by Kreul et al. Ž1998. and
Schneider et al. Ž1999.. To ascertain the influence of packing timum also results in lower conversion and purity, but they
characteristics on the process behavior, a continuous process remain almost constant for reflux ratios above 2. A similar
is investigated within this work. The main objective was a behavior was already reported by Bessling et al. Ž1998. for
comparison between theoretical and experimental results to this process, but they determined an optimal reflux ratio of
determine the ability of the model to predict the process be- 1.5 and a constant conversion of 0.66 for reflux ratios above
havior. 10. Unfortunately, a direct comparison of the results is not
Process simulations have been performed for experiments possible, since Bessling et al. Ž1998. did not provide informa-
1 to 5, as listed in Table 2. The feed was stoichiometric with tion about the feed mass flow and the distillate to feed ratio,
a total mass flow of 1.4 kgrh, while the distillate flow rate which are key parameters for the design of reactive distilla-
was held constant at 1.04 kgrh. The process conditions were tion processes. It also has to be considered that their process
varied by changing the reflux ratio from 0.05 up to 4. Figure model does not account for the influence of the reaction ki-
8 shows simulation results for the conversion of acetic acid netics, since they assume chemical equilibrium on all stages,
and the purity of the distillate stream in mass-fraction to- as well as for the hydrodynamic behavior of the column inter-
gether with the experimental results Žexperiments 1 to 5.. nals. The model presented here considers the latter and
The simulation results show a maximum of conversion and therefore the reflux ratio is limited to values below 5, since
purity at a reflux ratio of 1.2 with a strong decrease towards otherwise flooding occurs.
lower reflux ratios. Increasing the reflux ratio beyond the op- The experimental results do not confirm the simulated
process behavior, since the experiments are limited to reflux
ratios between 1 and 2.5. Thus, the strong decrease in con-
version and purity towards lower reflux ratio could not be
observed. The experimental conversion is 4 to 11% lower than
the theoretical values, a phenomenon that can be explained
with a reduced activity of the catalyst inside the column, but
an extended data basis is needed to confirm this assumption.
For the same reason, the experimental purity of the distillate
is 3 to 12% lower than the theoretical values.
Figure 9 shows the composition of the liquid phase along
the column height, with process conditions chosen according
to experiment No. 6. The theoretical values are displayed with
continuous lines, while the experimental results for the sam-
ples taken along the column lines are shown by single data
points.
A maximum concentration of methanol and acetic-acid can
be observed at the feed-locations respectively, while methyl
acetate is enriched towards the top and water towards the
bottom of the column. The experimental data show that the
simulation results reflect the real process behavior of methyl
acetate synthesis, but deviations between simulated and ex-
Figure 8. Theoretical vs. experimental results. perimental composition profiles can still be observed.

1074 May 2001 Vol. 47, No. 5 AIChE Journal

Conclusions Sscorrugation side, m
ScsSchmidt number
The new type of catalytic packing MULTIPAK that shows ShsSherwood number
a wide loading range combined with high separation effi- T stemperature, K
ciency and catalytic activity has been presented in this article. usvelocity, mrs
The main packing characteristics pressure drop and separa- w smass fraction
x sliquid-phase mol fraction
tion efficiency have been determined experimentally for the y sgas-phase mol fraction
whole loading range and compared with data for conven- a srelative volatility
tional structured packings taken from literature. The pres- b smass-transfer coefficient, mrs
sure drop of MULTIPAK can be estimated with the hy- w G x smatrix of thermodynamic correction factors
d sfilm thickness, m
draulic model of Rocha et al. Ž1996., but new models consid- e svoid fraction
ering the special hydrodynamic behavior of catalytic packings h sviscosity, Pa ? s
are necessary to reflect the packing characteristics with good u sinclination angle
accuracy and to provide information about the scale-up be- k sindividual mass-transfer coefficients, mrs
havior. A correlation for mass-transfer coefficients in both l sthermal conductivity, Wrm ? K
r sdensity, kgrm3
phases is presented that represents the experimental data with j sconversion
good accuracy. ccat scatalyst volume fraction
Experiments have been performed in a pilot-scale reactive D z ssegment height, m
distillation column to investigate the behavior of a continu-
ous process. A rigorous dynamic model based on the Subscripts
Maxwell-Stefan approach to describe mass transfer between Bsbottom
phases has been modified and implemented in the simulation cat scatalyst
environment Aspen Custom Modeler. Steady-state simula- Dsdistillate
tions for the process of methyl-acetate synthesis have been eff seffective
expsexperimental
performed and compared with experimental data. It was F sfeed
shown that the model is able to reflect the process behavior, Gsgas phase
but deviation was observed between experimental and theo- iscomponent index
retical conversion and purity. However, by implementing the jssegment index
k scomponent index
reaction kinetics and the hydrodynamics of the catalytic pack- Lsliquid phase
ings in the process simulation, their influence on design as- sssuperficial
pects of reactive distillation columns were taken into ac- t stotal
count.
Superscripts
av saverage
Acknowledgments n sheat loss
The support from Axiva GmbH, BASF AG, Bayer AG, Degussa- U sat the interface
¨ AG, and German Federal Ministry of Education and Research
Huls
Project No. 03C0306 is highly appreciated.
Dimensionless numbers
Reynolds number sReG s wŽ u G ,eff q u L,eff . rG SxrhG
Notation Schmidt number sSc G shGrrG DG
Sherwood number sSh G s bG SrDG
asspecific packing surface, m2rm3
csmolar density, molrm3
dscolumn diameter, m
Dsdiffusivity, m2rs Literature Cited
}
D sMaxwell-Stefan diffusivity, m2rs Agreda, V. H., L. R. Partin, and W. H. Heise, ‘‘High-Purity Methyl
F sF-factor, kg 0.5rm0.5 ? s Acetate via Reactive Distillation,’’ Chem. Eng. Prog., 86 ŽFeb. 1990..
Ž J . svector of diffusion fluxes, molrm2 ? s ¨
Bart, H. J., and H. Landschutzer, ‘‘Heterogene Reaktivdestillation
Ġsgas mol flow, molrs ¨
mit axialer Ruckvermischung,’’ Chem. Ing. Tech., 68, 944 Ž1996.
hsspecific enthalpy Jrmol ¨
Bessling, B., J. M. Loning, ¨
A. Ohligschlager, G. Schembecker, and K.
h L svolumetric holdup Sundmacher, ‘‘Investigations on the Synthesis of Methyl Acetate
H sheight, m in a Heterogeneous Reactive Distillation Process,’’ Chem. Eng.
HETP sheight equivalent of a theoretical plate, m Technol., 21, 393 Ž1998..
HTUsheight of transfer unit, m Chilton, T. H., and A. P. Colburn, ‘‘Distillation and Absorption in
k O G soverall gas-phase mass-transfer coefficient mrs Packed Columns,’’ Ind. Eng. Chem., 27, 255 Ž1935..
K eq svapor liquid equilibrium constant Ellenberger, J., and R. Krishna, ‘‘Counter-Current Operation of
L̇sliquid mol flow, molrs Structured Catalytically Packed Distillation Columns: Pressure
msstripping factor Drop, Holdup and Mixing,’’ Chem. Eng. Sci., 54, 1339 Ž1999..
˙
msmass flow, kgrh Fair, J. R., ‘‘Design Aspects for Reactive Distillation,’’ Chem. Eng.,
nsnumber of components 158 ŽOct. 1998..
˙nsinterfacial molar flux, molrm2 s Frey, T., and J. Stichlmair, ‘‘Thermodynamic Fundamentals of Reac-
n th snumber of theoretical stages tive Distillation,’’ Chem. Eng. Technol., 22, 11 Ž1999..
NTSMsnumber of theoretical stages per meter, Lrm Higler, A. P., R. Taylor, and R. Krishna, ‘‘Nonequilibrium Modelling
˙
qsheat flux, Jrm2 ? s of Reactive Distillation: Multiple States in MTBE Synthesis,’’
r sreaction rate, molrkg ? s Chem. Eng. Sci., 54, 1389 Ž1999..
w R x sinverse mass-transfer coefficient matrix, srm Huang, C., G. G. Podrebarac, F. T. T. Ng, and G. L. Rempel, ‘‘A
ResReynolds number Study of Mass Transfer Behaviour in a Catalytic Distillation Col-

AIChE Journal May 2001 Vol. 47, No. 5 1075

 umn,’’ Can. J. Chem. Eng., 76, 323 Ž1998.. Powers, M. F., D. J. Vickery, A. Arehole, and R. Taylor, ‘‘A
Kenig, E., K. Jakobsson, P. Banik, J. Aittamaa, A. Gorak,´ M. Koski- Nonequilibrium Stage Model of Multicomponent Separation Pro-
nen, and P. Wettmann, ‘‘An Integrated Tool for the Synthesis and cesses: V. Computational Methods for Solving the Model Equa-
Design of Reactive Distillation,’’ Chem. Eng. Sci., 54, 1347 Ž1999.. tions,’’ Comp. Chem. Engng., 12, 1229 Ž1988..
Krafczyk, J., and J. Gmehling, ‘‘Einsatz von Katalysatorpackungen Reid, R. C., J. M. Prausnitz, and B. E. Poling, The Properties of Gases
¨ die Herstellung von Methylacetat durch reaktive Rektifikation,’’
fur & Liquids, 4th ed., McGraw Hill, New York Ž1987..
Chem.-Ing. Tech., 66, 1372 Ž1994.. Rocha, J. A., J. L. Bravo and J. R. Fair, ‘‘Distillation Columns Con-
´
Kreul, L. U., A. Gorak, C. Dittrich and P. I. Barton, ‘‘Dynamic Cat- taining Structured Packings: A Comprehensive Model for Their
alytic Distillation: Advanced Simulation and Experimental Valida- Performance: 1. Hydraulic Models,’’ Ind. Eng. Chem. Res., 32, 641
tion,’’ Comp. Chem. Eng., 22, Suppl., 371 Ž1998.. Ž1993..
Krishna, R., and G. L. Standart, ‘‘Mass and Energy Transfer in Mul- Rocha, J. A., J. L. Bravo, and J. R. Fair, ‘‘Distillation Columns Con-
ticomponent Systems,’’ Chem. Eng. Commun., 3, 201 Ž1979.. taining Structured Packings: A Comprehensive Model for Their
Lee, J. H., and M. P. Dudukovic, ‘‘A Comparison of the Equilibrium Performance: 2. Mass-Transfer Model,’’ Ind. Eng. Chem. Res., 35,
and Nonequilibrium Models for a Multicomponent Reactive Distil- 1660 Ž1996..
lation Column,’’ Comp. Chem. Eng., 23, 159 Ž1998.. Schneider, R., C. Noeres, L. U. Kreul, and A. Gorak, ´ ‘‘Dynamic
Moritz, P., B. Bessling, and G. Schembecker, ‘‘Fluiddynamische Be- Modelling and Simulation of Reactive Batch Distillation,’’ Comp.
¨
trachtungen von Katalysatortragern bei der Reaktivdestillation,’’ Chem. Engng., 23, Suppl., 423 Ž1999..
Chem. Ing. Tech., 71, 131 Ž1999.. Stichlmair, J. G., and J. R. Fair, Distillation: Principles and Practice,
Moritz, P., and H. Hasse, ‘‘Fluid Dynamics in Reactive Distillation Wiley-VCH, New York, Ž1998..
Packing Katapak-S,’’ Chem. Eng. Sci., 54, 1367 Ž1999.. Stewart, W. E., and R. Prober, ‘‘Matrix Calculation of Multicompo-
Nisoli, A., M. F. Malone. and M. F. Doherty, ‘‘Attainable Regions nent Mass Transfer in Isothermal Systems,’’ Ind. Eng. Chem. Fun-
for Reaction with Separation,’’ AIChE J., 43, 374 Ž1997.. dam., 4, 224 Ž1964..
Offenlegungsschrift, Deutsches Patentamt, DE 197 01 045 A1, Ž1997.. Subwalla, H., and J. R. Fair, ‘‘Design Guidelines for Solid-Catalyzed
Okasinski, M. J., and M. F. Doherty, ‘‘Thermodynamic Behavior of Reactive Distillation Systems,’’ Ind. Eng. Chem. Res., 38, 3696
Reactive Azeotropes,’’ AIChE J., 43, 2227 Ž1997.. Ž1999..
Olujic, ˇ A. B. Kamerbeek, and J. de Graauw, ‘‘A Corrugation Ge-
´ Z., Subwalla, H., J. C. Gonzalez, A. F. Seibert and J. R. Fair, ‘‘Capacity
ometry Based Model for Efficiency of Structured Distillation Pack- and Efficiency of Reactive Distillation Bale Packing: Modelling and
ing,’’ Chem. Eng. Proc., 38, 683 Ž1999.. Experimental Validation,’’ Ind. Eng. Chem. Res., 36, 3821 Ž1997..
Onken, U., and W. Arlt, ‘‘Recommendet Test Mixtures for Distilla- Sundmacher, K., and U. Hoffman, ‘‘Development of a New Catalytic
tion Columns,’’ The Institution of Chemical Engineers, Warwick- Distillation Process for Fuel Ethers via a Detailed Nonequilibrium
shire, U.K. Ž1990.. Model,’’ Chem. Eng. Sci., 51, 2359 Ž1996..
Pelkonen S., R. Kaesemann and A. Gorak, ´ ‘‘Distillation Lines for Taylor, R., and R. Krishna, Multicomponent Mass Transfer, Wiley,
Multicomponent Separation in Packed Columns}Theory and New York Ž1993..
Comparison with Experiment,’’ Ind.. Eng. Chem. Res, 36, 5392 Taylor, R., and R. Krishna, ‘‘Modelling Reactive Distillation,’’ Chem.
Ž1997.. Eng. Sci., 55, 5183 Ž2000..
Podrebarac, G. G., F. T. T. Ng, and G. L. Rempel, ‘‘The Production Ung, S., and M. F. Doherty, ‘‘Necessary and Sufficient Conditions
of Diacetone Alcohol with Catalytic Distillation: I. Catalytic Distil- for Reactive Azeotropes in Multireaction Mixtures,’’ AIChE J., 41,
lation Experiments,’’ Chem. Eng. Sci., 53, 1067 Ž1998a.. 2383 Ž1995..
Podrebarac, G. G., F. T. T. Ng, and G. L. Rempel, ‘‘The Production Xu, Z. P., and K. T. Chuang, ‘‘Kinetics of Acetic Acid Esterification
of Diacetone Alcohol with Catalytic Distillation: II. A Rate-Based over Ion Exchange Catalysts,’’ Can. J. Chem. Eng., 74, 493 Ž1996..
Catalytic Distillation Model for the Reaction Zone,’’ Chem. Eng. Xu, Z. P. and K. T. Chuang, ‘‘Effect of Internal Diffusion on Het-
Sci., 53, 1077 Ž1998b.. erogeneous Catalytic Esterification of Acetic Acid,’’ Chem. Eng.
¨
Popken, ¨
T., R. Geisler, L. Gotze, A. Brehm, P. Moritz, and J. Sci., 52, 3011 Ž1997..
Gmehling, ‘‘Reaktionskinetik in der Reaktivrektifikation}Zur Xu X., Z. Zhao, and S. Tian, ‘‘Study on Catalytic Distillation Pro-
Übertragbarkeit von kinetischen Daten aus einer Ruhrzelle¨ auf cesses: III: Prediction of Pressure Drop and Holdup in Catalyst
einen Rieselbettreaktor,’’ Chem.-Ing. Tech., 71, 96 Ž1999.. Bed,’’ Trans IChemE, 75, 625 Ž1997..
¨
Popken, ¨
T., L. Gotze and J. Gmehling, ‘‘Reaction Kinetics and Zheng, Y., and X. Xu, ‘‘Study on Catalytic Distillation Processes:
Chemical Equilibrium of Homogeneously and Heterogeneously Part I. Mass Transfer Characteristics in Catalyst Bed within the
Catalyzed Acetic Acid Esterification with Methanol and Methyl Column,’’ Trans IChemE, 70, 459 Ž1992..
Acetate Hydrolysis,’’ Ind. Eng. Chem. Res., 39, 2601 Ž2000.. Manuscript recei®ed Jan. 18, 2000, and re®ision recei®ed Feb. 12, 2001.

1076 May 2001 Vol. 47, No. 5 AIChE Journal