Adsorption and reaction in the transesterification of ethyl acetate with methanol on Lewatit K1221

Article abstract of DOI:10.1016/j.molcata.2012.03.021

The reaction kinetics of the liquid-phase transesterification of ethyl acetate with methanol to methyl acetate and ethanol have been investigated in a temperature range from 303.15 K to 333.15 K as a model reaction for the transesterification of triglycerides in the production of biodiesel. The reaction has been catalyzed by the acidic ion-exchange resin Lewatit K1221. The effect of the initial reactant molar ratio and the temperature on the reaction kinetics was investigated and kinetic models, based on pseudo-homogeneous (PH), Eley-Rideal (ER) and Langmuir-Hinshelwood (LH) mechanisms, were used to describe the reaction rate. Because of the pronounced non-ideality of the reaction mixture, the kinetics were expressed in terms of activities. Additional experiments, based on a D-optimum design of experiments, were performed to obtain more precise parameter estimates as required for final model discrimination. The kinetic model with the surface reaction of adsorbed methanol with ethyl acetate from the bulk as the rate-determining step according to an Eley-Rideal mechanism was found to best describe the observed kinetics. The corresponding rate equation agrees with a reaction mechanism in which physically adsorbed methanol reacts with protonated ethyl acetate.

Full text of DOI:10.1016/j.molcata.2012.03.021

Journal of Molecular Catalysis A: Chemical 359 (2012) 57–68
Contents lists available at SciVerse ScienceDirect
Journal of Molecular Catalysis A: Chemical
journal homepage: www.elsevier.com/locate/molcata
Adsorption and reaction in the transesterification of ethyl acetate with methanol
on Lewatit K1221
Evelien Van de Steenea,b, Jeriffa De Clercq , Joris W. Thybautb,∗
a
a
Faculty of Applied Engineering Sciences, University College Ghent, Schoonmeersstraat 52, B-9000 Ghent, Belgium
Laboratory for Chemical Technology, Ghent University, Krijgslaan 281, S5, B-9000 Ghent, Belgium
b
a r t i c l e i n f o
a b s t r a c t
Article history:
The reaction kinetics of the liquid-phase transesterification of ethyl acetate with methanol to methyl
acetate and ethanol have been investigated in a temperature range from 303.15 K to 333.15 K as a
model reaction for the transesterification of triglycerides in the production of biodiesel. The reac-
tion has been catalyzed by the acidic ion-exchange resin Lewatit K1221. The effect of the initial
reactant molar ratio and the temperature on the reaction kinetics was investigated and kinetic
models, based on pseudo-homogeneous (PH), Eley–Rideal (ER) and Langmuir–Hinshelwood (LH) mech-
anisms, were used to describe the reaction rate. Because of the pronounced non-ideality of the
reaction mixture, the kinetics were expressed in terms of activities. Additional experiments, based
on a D-optimum design of experiments, were performed to obtain more precise parameter esti-
mates as required for final model discrimination. The kinetic model with the surface reaction of
adsorbed methanol with ethyl acetate from the bulk as the rate-determining step according to an
Eley–Rideal mechanism was found to best describe the observed kinetics. The corresponding rate equa-
tion agrees with a reaction mechanism in which physically adsorbed methanol reacts with protonated
ethyl acetate.
Received 27 January 2012
Received in revised form 20 March 2012
Accepted 21 March 2012
Available online 30 March 2012
Keywords:
Heterogeneous catalysis
Transesterification
Ion-exchange resin
Adsorption
Kinetics
Eley–Rideal
©
2012 Elsevier B.V. All rights reserved.
1
. Introduction
and neutralization, such as in acidic or alkaline catalyzed reaction
technologies. Increasing waste disposal costs are adding to the
environmental and societal costs of an increasingly critical public
toward chemical waste.
Catalytic technologies have played a vital role in the economic
development of the chemical industry during the 20th century,
with a total estimated contribution of catalytic processes amount-
ing to about 20% of the world’s GNP. In the 21st century, there
is a drive toward cleaner technologies brought about by public,
legislative and corporate pressure to provide new and exciting
opportunities for catalysis and catalytic processes [1].
The well-known disadvantages of homogeneous catalysis are
reinforced by environmental policies and result in heterogeneous
catalysis as a rapidly growing area [2,3]. Although heterogeneously
catalyzed processes are widely used in large-scale petrochemical
processes, the majority of fine, speciality, and pharmaceutical
chemicals manufacturing processes still relies on homogeneous
catalysts as, e.g., in the (trans)esterification. Many of these pro-
cesses were developed about a half century ago [4,5]. While
aiming at maximizing product yields, the environmental impact
of inorganic waste and toxic by-products formed during the
reaction was disregarded. Most of the waste is generated during
the separation stage of the process by a typical water quench
Esters are of great significance in various industrial prod-
ucts including fragrances, flavors, solvents, plasticizers, medicinal
and surface-active agents [6]. Esterification and transesterification
plays an important industrial role with numerous applications.
Large scale applications of transesterification are, e.g., the produc-
tion of biodiesel, polyesters or PET in the polymer industry, while
small scale fine chemical production includes synthesis of inter-
mediates in the pharmaceutical industry, the production of food
additives or surfactants and the curing of resins in the paint industry
[7].
Transesterification reactions can be performed using base cat-
alysts, such as metal hydroxides, metal alkoxides, alkaline-earth
oxides or hydrotalcites or using acid catalysts, such as sulfuric,
sulfonic, phosporic and hydrochloric acids. Base catalysts are typ-
ically preferred because of the higher reaction rates and the mild
reaction conditions as compared to acid-catalyzed transesterifica-
tions. Nowadays, biodiesel is conventionally produced through a
batch or continuous transesterification of highly refined vegetable
oils with methanol by using homogeneous alkaline catalysts such
as sodium or potassium hydroxides or methoxides. This conven-
tional, industrial technology is not compatible with oils which have
∗ Corresponding author. Tel.: +32 9 264 55 63; fax: +32 9 264 49 99.
E-mail address: Joris.Thybaut@UGent.be (J.W. Thybaut).
1
381-1169/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.molcata.2012.03.021

5
8
E. Van de Steene et al. / Journal of Molecular Catalysis A: Chemical 359 (2012) 57–68
process for their high content in free fatty acids, would, hence,
Nomenclature
offer an enormous strategic advantage. Innovative biodiesel
production processes may be based on supercritical methanol or
a two-stage process including esterification prior to transesterifi-
cation. The esterification pre-treatment aimed at the abatement
of FFAs is generally promoted by homogeneous acids or by solid
acid ion-exchange resins. Strong sulfonic acid ion-exchange resins
are promising candidates for catalyzing the complete esterifica-
tion of FFAs and a first part of the transesterification [2,8–11].
Zielinska-Nadolska et al. [9], e.g., proved that the strong acid ion-
exchange resin Lewatit K1221, modified potassium carbonate and
the superacid ion-exchange resin Nafion SAC-13, gave the highest
ethyl methyl carbonate yield in the transesterification of dimethyl
carbonate with ethanol. Therefore, Lewatit K1221 catalyst was
selected in this work.
Pseudo-homogeneous (PH) and adsorption-based mechanisms
have been used to describe heterogeneously catalyzed transester-
ification. Adsorption-based mechanisms include the Eley–Rideal
(ER) and the Langmuir–Hinshelwood (LH) mechanisms. Saha and
Streat [12] have evaluated different heterogeneous acid catalysts
including ion-exchange resins as catalysts for the transesterifi-
cation of cyclohexyl acrylate with n-butanol and 2-ethylhexanol.
The experimental results were best described by a kinetic model
based on a Langmuir–Hinshelwood–Hougen–Watson mechanism.
Jimenez et al. [13] studied the transesterification of n-butyl acetate
and methanol catalyzed by ion-exchange resin Amberlyst 15 and
found the pseudo-homogeneous model as the best description
for the experimental data. Bo z˙ ek-Winkler and Gmehling [14]
and Steinigeweg and Gmehling [15] studied the kinetic behav-
ior of the reaction of methyl acetate and butanol catalyzed by
the ion-exchange resin Amberlyst 15. Two different kinetic mod-
els, pseudo-homogenous and Langmuir–Hinshelwood, were used
to describe the reaction rate. The simpler pseudo-homogeneous
model provided similar results as the Langmuir–Hinshelwood
model [14].
ꢀ
specific surface area particle (m2
−1
a
kg
)
)
particle
cat
ai
activity of component i (mol m−3)
3
−1
−1 −1
A
pre-exponential factor (m kg mol s
cat
−
1
Apeak,i
surface peak area of component i (V s
)
b
parameter estimate
−3
)
gas
cb
bulk fluid concentration (mol m
−
)
gas
3
cs
Ca
CFi
concentration at external particle surface (mol m
Carberry number
calibration factor of component i with respect to n-
octane
−
−3
)
3
Ci
observed concentration of component i (mol m
predicted concentration of component i (mol m
)
ˆ
C
i
2
−1
De
EA
k
effective diffusion coefficient (m s
activation energy (kJ mol
reaction rate coefficient (m kg mol
)
−1
)
3
−1
−1 −1
s
)
cat
−1
kg
Keq
Ki
mass transfer coefficient (m s
equilibrium coefficient of the overall reaction
adsorption equilibrium coefficient of component i
)
3
−1
(
m mol
)
L
characteristic particle size (m)
likelihood function
mass of component i (kg)
Lmax
mi
n
order of the reaction
ni
N
p
number of moles of component i (mol)
total number of experimental points or components
number of parameters
−
1
−1
s )
cat
r
reaction rate (mol kg
−
3
−1
s
robs
R
observed volumetric reaction rate (mol m
gas constant (J mol
)
−
1
−1
K )
−
1
−1
Ri
net production rate of component i (mol kg
regression sum of squares
residual sum of squares
objective function
s
)
cat
REG
RSSQ
S
The objective of this work is to investigate the transesterifi-
cation kinetics of ethyl acetate (EtOAc) with methanol (MeOH)
to form methyl acetate (MeOAc) and ethanol (EtOH) catalyzed
t
time (s)
T
temperature (K)
W
Xi
weight of catalyst (kgcat)
conversion of component i
by Lewatit K1221. Experiments in a perfectly mixed batch
reactor have been performed. These experimental data have
been modeled to obtain more insight in the reaction mecha-
nism and a potential rate-determining step. Pseudo-homogeneous
and different adsorption-based models, such as Eley–Rideal and
Langmuir–Hinshelwood, have been used. This work fits into
an overall strategy for the adequate simulation of a two-stage
biodiesel manufacturing process. Base catalyzed transesterification
kinetics have been previously studied [3], acid catalyzed transes-
terification kinetics are the subject of the present study, while the
acid catalyzed esterification kinetics will be studied in the future.
Greek symbols
ˇ
ꢀi
ꢁ
ꢂi
ꢃ
ꢄi
˚
ꢅHr
parameter
activity coefficient of component i
effectiveness factor
fractional coverage of catalyst surface
likelihood ratio
stoichiometric coefficient for component i
Weisz modulus
reaction enthalpy (kJ mol−1)
Subscripts and superscripts
2. Procedures
0
*
A
initial conditions
adsorption site
model A
2.1. Materials
B
model B
catalyst
at equilibrium
component i
Methanol (Fiers, purity >99.85%) and ethyl acetate (Fiers, purity
≥99.5%) were used as reagents. n-Octane (Acros Organics, purity
>99%) was used as internal standard. Lewatit K1221 (Lanxess) is
a strongly acidic, dark brown translucent, polymer-based resins
in spherical bead form, with sulfonic acid groups. This gel-type
resin is moderately crosslinked (4%) with polystyrene. The bead
size varies between 0.4 and 1.2 mm. Lewatit K1221 contains mini-
cat
eq
i
+
a free fatty acids (FFAs) content exceeding a threshold value of
about 0.5 wt%. The development of new technologies, enabling to
employ waste raw materials such as fried oils or mixtures of oils
from various sources that cannot be treated in the conventional
mal 1.2 mol H /L. The resin is heat-sensitive and experiences a loss
of activity above 398 K. Prior to transesterification, the catalyst was
dried under vacuum at 233 K for 24 h to completely remove any
moisture.

E. Van de Steene et al. / Journal of Molecular Catalysis A: Chemical 359 (2012) 57–68
59
Table 1
Range of experimental conditions.
Temperature (K)
Pressure (MPa)
303.15–333.15
0.1
Initial MeOH:EtOAc molar ratio
1:1 to 10:1
0.5–5.0
12.6–126
−
3
Mass of catalyst (10 kg)
a
Batch time (kg s)
a
Defined as the product of experiment time and the mass of catalyst.
2.3. Sampling and analysis
Samples of the reaction mixture were withdrawn through the
sampling port every 1800 s. The samples were analyzed by gas
chromatography using a Focus GC, equipped with an AS3000 auto
sampler, a Stabilwax capillary column (30 m × 0.32 mm, 0.25 ␮m
thickness) and a flame ionization detector. The injector and detec-
tor temperatures were both set at 523 K. The oven temperature
program started at 313 K for 300 s, then increased to 343 K at a
−
1
−1
rate of 6.0 K s , further increased to 473 K at a rate of 50.0 K s
and finally held constant at 473 K. The Chromchard 2.3.2 software
was used to analyze the gas chromatography data. No by-product
formation was observed. No conversion was observed when the
reaction was performed in the absence of the catalyst. No catalyst
activity decay, e.g., due to leaching, was observed in successive runs
on the same catalyst batch.
Quantification of the reaction components was performed by
relating the peak surface areas to the mass of n-octane, the internal
standard:
Apeak,ref
mref
mi
=
CFi
(1)
Apeak,i
where Apeak,i and Apeak,ref, are the peak surface area of component i
and n-octane, and CF the calibration factor of the component i with
Fig. 1. Experimental setup.
i
respect to n-octane.
Conversions were calculated as follows:
Ci,0 − Ci,t
Xi =
(2)
C
i,0
2.2. Experimental setup
where X is the conversion of reactant i, C the initial concentration
i
i,0
The transesterification of ethyl acetate with methanol was car-
of i and Ci,t the concentration of i at time t [16].
Only experiments with a mass balance deviation below 5% were
used in the kinetic modeling. Prior to any regression, the corre-
sponding experimental results were scaled to obey a 100% mass
balance.
ried out in a three-necked glass flask of 180 mL capacity equipped
with a reflux condenser, a thermocouple and a sampling port,
see Fig. 1. The temperature in the reactor was maintained within
0
.5 K from the set point using a PID-controller (Lauda Proline
RP845). The reaction mixture was stirred with a magnetic stirrer
and set at a constant speed of 500 rpm throughout the experi-
ment. The reactor was first loaded with methanol and catalyst and
then heated to the reaction temperature. The gas volume of the
reactor was minimal, hence the loss to the vapor phase could be
neglected.
When the reactor mixture reached the desired temperature,
preheated ethyl acetate and n-octane were added through the
sampling port in order to have a well determined starting point
of the experiment without disturbing the reactor temperature. n-
Octane was used as internal standard for analytical purposes and
for obtaining a constant reaction volume. The n-octane content
ranged from 56 to 3 mol%, depending on the molar ratio 1:1 to
2.4. Intrinsic kinetics measurements
In order to ascertain that the obtained data represent intrin-
sic kinetics, they should be free from deviations due to non-ideal
reactor behavior and interference with transport phenomena. Two
types of mass-transfer resistance have been verified, one across
the solid-liquid interface and the other in the intraparticle space.
Quantitative criteria have been developed to verify the absence of
mass transfer limitations, such as the Carberry number Ca for exter-
nal mass transfer and the Weisz modulus ˚ for internal diffusion.
These parameters were calculated using the appropriate correla-
tions (Table 2) [16,17].
1
0:1. All experiments have been performed at atmospheric pres-
In addition to these quantitative criteria, also some qualitative
tests have been performed to verify the importance of transfer phe-
nomena and the ideality of the reactor flow pattern. To evaluate the
significance of external mass transport phenomena, transesterifi-
cation experiments were performed at different agitation speeds.
When external mass transport is not rate limiting, there is no
effect of the agitation speed [16]. It is clear that the external mass-
transfer resistance is negligible between 200 and 1000 rpm, see
Fig. 2. Hence, all further experiments were conducted at a stirrer
sure. Each experiment was performed at least quadruple with
an experimental error below 5%. The range of experimental con-
ditions is given in Table 1. A reference experimental data set
consisting of 1012 points from 68 experiments has been used
for initial model regression. It has subsequently been extended
with 270 data points according to a D-optimal design, see Section
2
.6, for more precise parameter estimation and conclusive model
discrimination.

6
0
E. Van de Steene et al. / Journal of Molecular Catalysis A: Chemical 359 (2012) 57–68
Table 2
2
.5. Thermodynamic calculations
Criteria for the absence of transport limitations for a catalyst particle under steady-
state operation.
The reaction enthalpy was calculated based on thermodynamic
Transport phenomenon
Criterion
Calculation
data obtained form the tabulated values of the ASPEN engineer-
ing suite
r
TM
obs
11.1, and was found to vary from −4.20 at 303 K to
Extraparticle mass transport
Ca =
ꢀ
< 0.05
Ca = 7.7E − 03
kg a c
b
−
1
(
‘Carberry number’)
Intraparticle mass transport
‘Wheeler–Weisz criterion’)
−0.50 kJ mol
at 333 K, showing a slightly exothermic reaction.
ꢀ
ꢁ
2
r
L
n+1
obs
˚ = ꢁ
< 0.1
˚ = 7.5E − 02
The corresponding equilibrium coefficient Keq:
Decs
2
in
(
CMeOAcCEtOH ꢀMeOAcꢀEtOH
Keq = K Kꢀ =
(3)
C
CMeOHCEtOAc ꢀMeOHꢀEtOAc
8
6
4
2
0
0
0
0
0
ranges from 1.59 at 303 K to 1.46 at 333 K, which is similar to values
reported in previous transesterification studies [7]. The equilibrium
conversion of ethyl acetate Xeq varies from 58 to 94% depending on
the initial molar ratio of methanol to ethyl acetate (MeOH:EtOAc),
i.e., ranging from 1:1 to 10:1.
The non-ideal character of the reaction mixture was assessed
and activity coefficients ꢀi for each component i were calculated at
temperatures ranging from 303.15 K to 333.15 K with the UNIFAC
group contribution method [19]. The molecular activity coeffi-
cient, see Eq. (4), is separated into two parts: the combinatorial
part provides the contribution due to differences in molecular size
and shape, and the residual part provides the contribution due
to energy interactions, functional group sizes and interactions of
surface areas:
0
rpm
200 rpm
500 rpm
1000 rpm
2
4
6
8
10
12
14
R
ln ꢀ =
ln ꢀc
combinatorial
+ ln ꢀ
(4)
i
i
i
Batch time (kg s)
residual
Fig. 2. Effect of the agitation speed (legend) on the conversion of EtOAc (T = 333 K,
W = 0.58 × 10 kg, initial MeOH:EtOAc molar ratio = 10:1).
Table 3 shows the activity coefficients of each component for all the
experimental conditions, at the start of the reaction and at thermo-
dynamic equilibrium. Since the activity coefficients were different
from 1.0, the reaction mixture clearly exhibits non-ideal behavior.
Deviations from ideality were found to be most pronounced in the
experiments with a high initial molar methanol to ethyl acetate
ratio.
−
3
speed of 500 rpm to ensure that the reaction rate was not limited by
external diffusion. This was also observed by Sanz et al. [18]. These
authors indicate that external diffusion does not usually control
the overall rate in the reactions catalyzed by ion-exchange resin
unless the agitation speed is very low or the reaction mixture is
very viscous.
The possible effect of internal mass-transfer was verified by
using different catalyst particle sizes [16]. The commercial Lewatit
K1221 resin was screened into four different size ranges. As shown
in Fig. 3, no significant differences were found in the reaction rate
for the different catalyst sizes, i.e., the effect of internal mass trans-
fer on the reaction kinetics could be considered negligible. Hence,
all experiments were conducted with the ion-exchange resin as
supplied by the manufacturer without any size screening.
2.6. Modeling and regression analysis
The mass balance for species i in a batch reactor, which is con-
sidered to be spatially uniform in composition and temperature, is
given by
1
dni
= R = ꢄ r
(5)
i
i
W dt
with R the net production rate of component i, W the catalyst mass
i
in the reactor, v the stoichiometric coefficient of component i, n the
i
i
number of moles of component i in the reaction mixture, and t the
time. In Eq. (5), the net production rate is calculated according to
equations derived from the various mechanisms that are proposed,
i.e., pseudo-homogeneous (PH), Langmuir–Hinshelwood (LH) and
Eley–Rideal (ER).
The net production rate is a function of the reaction rate coef-
ficient, temperature, activities and equilibrium coefficients. The
reaction rate coefficient is described by the Arrhenius law, which is
reparameterized according to Kittrell [20] in order to avoid strong
binary correlation between the Arrhenius parameters:
80
70
60
50
40
30
20
ꢂ
ꢃ
ꢄꢅ
EA
R
1
T
1
k = kTref exp
−
− T
(6)
ref
4
6
00 - 500 µm 500 - 600 µm
00 - 720 µm 720 - 1000 µm
with EA the activation energy, R the universal gas constant and kTref
the reaction rate coefficient at the reference temperature. The acti-
vation energy and the rate coefficient at the reference temperature
as well as the adsorption equilibrium coefficient for the adsorption-
based mechanisms have been determined from regression, see
2
4
6
8
10
12
14
Batch time (kg s)
Fig. 3. Effect of the particle size (legend) on the conversion of EtOAc (T = 333 K,
−
3
W = 0.58 × 10 kg, initial MeOH:EtOAc molar ratio = 10:1).

E. Van de Steene et al. / Journal of Molecular Catalysis A: Chemical 359 (2012) 57–68
61
Table 3
Activity coefficients of MeOH, EtOH, MeOAc, EtOAc and n-octane at different temperatures and different initial MeOH:EtOAc molar ratios, at the start of the reaction and at
equilibrium. CEtOAc,0=1.79 M except at 333 K and initial MeOH:EtOAc molar ratio = 1:1 where CEtOAc,0=5.79 M.
Temperature
303 K
1:1
0
313 K
10:1
0
323 K
10:1
0
333 K
1:1
0
MeOH:EtOAc ratio
Conversion (%)
10:1
0
5:1
0
10:1
0
56
94
1.02
1.19
2.12
2.27
17.22
94
1.02
1.18
2.08
2.24
16.42
94
1.02
1.18
2.05
2.20
15.67
56
89
94
1.02
1.17
2.02
2.17
14.96
MeOH
EtOH
MeOAc
EtOAc
1.47
–
–
1.25
4.84
1.52
1.41
1.27
1.25
4.57
1.02
–
–
2.27
17.39
1.02
–
–
2.24
16.58
1.02
–
–
2.21
15.82
1.45
–
–
1.23
4.51
1.51
1.39
1.25
1.23
4.29
1.29
–
–
1.59
4.02
1.31
1.28
1.62
1.57
4.01
1.02
–
–
2.17
15.10
n-Octane
Table 5. The latter was performed by minimizing the following
objective function S:
between rival models can also be performed based on the so-called
likelihood ratio. The likelihood ratio ꢃ is the ratio of the maxi-
mum in the likelihood function, Lmax obtained with the rival models
A, with pA parameters and RSSQ , and B, with p parameters and
N
ꢆ
ꢆ
b
2
A
B
S =
(CEtOAc,k − CEtOAc,k) −→ minimum
(7)
RSSQB:
k=1
N/2
ꢇ
ꢉ
(
N − p)
N − p
with CEtOAc,k the experimentally observed concentration of ethyl
Lmax =
exp
N
−
(11)
ꢆ
ꢇꢈ
ꢉ
2
acetate in the kth experiment and CEtOAc,k the corresponding model
calculated value, N the total number of experimental measure-
ments and b the parameter vector [14,16]. The minimization was
achieved by a single response Levenberg–Marquardt algorithm. The
residual sum of squares (RSSQ) equals the final value of S of Eq. (7).
The regression sum of squares (REG) is given by
2
ꢇRSSQ
ꢊ
ꢋ
N/2
ꢇ
ꢉꢊ
ꢋ
N/2
ꢃ = (
L )
=
N − pA
N − pB
exp
pA − pB
RSSQB
RSSQA
(12)
A
max
(
LB)max
2
For a high ratio, model A has a higher probability to adequately
describe the experimental data than model B and vice versa for a
small ratio.
N
ꢆ
2
ˆ
C
REG =
(8)
In case the likelihood ratio does not lead to a decisive discrimi-
nation, a design of experiments (DOE) can be performed to obtain
more precise parameter estimates and, hence, to allow a better
discrimination. In this work, a D-optimal experimental design is
used, which minimizes the joint confidence interval of the model
parameters and, hence, maximizes the determinant of the param-
eter estimation covariance matrix [16,23,24].
k=1
The estimated parameters were subsequently evaluated based on
their physicochemical and statistical significance. The reaction rate
coefficients have to increase with temperature, i.e., the activation
energy and the rate coefficient at the reference temperature have
to be positive and, based on the Van’t Hoff equation, the adsorp-
tion equilibrium coefficients have to decrease with temperature.
The statistical significance of the parameter estimates was assessed
with the Student t-test:
3. Kinetic model
b − ˇ
The kinetic model for the transesterification of ethyl acetate
i
i
t =
> ttab(N − p, 95%)
(9)
s(b )
with methanol can be expressed according to either a pseudo-
homogeneous or an adsorption reaction based reaction mech-
anism such as Langmuir–Hinshelwood (LH) or Eley–Rideal (ER)
i
with b the estimated parameter value, ˇ , the proposed value, c.q.,
i
i
zero, for parameter i, and s(b ) an estimate of the standard deviation
16,21].
The global significance of the regression was assessed with the
i
[
14,25–27].
[
The pseudo-homogeneous kinetic model is given by
ꢃ
ꢄ
F-test:
1
Keq
r = kPH aMeOHaEtOAc
−
aEtOHaMeOAc
(16)
REG/p
F =
> Ftab(p, N − p, 95%)
(10)
RSSQ/(N − p)
with kPH the reaction rate coefficient, a the activity of component
i
with p the number of model parameters [16,22].
Apart from the statistical and physical significance of the regres-
sion and the corresponding model parameters, discrimination
i and Keq the equilibrium coefficient of the overall reaction.
For the adsorption-based models, the various LH mechanisms
consist of five elementary steps, while the ER mechanisms consist
Table 4
Elementary steps and reaction mechanism for the kinetic modeling of transesterification of ethyl acetate with methanol. * = active site.
Elementary reactions
Reaction mechanism
PH
LH
ER
Alcohol adsorption
1
Ester adsorption
1
CH3OH + * ꢀ CH3OH*
1
1
CH3CH2COOCH3 + * ꢀ CH3CH2COOCH3*
CH3OH + CH3CH2COOCH3 ꢀ CH3COOCH3 + CH3CH2OH
CH3OH* + CH3CH2COOCH3* ꢀ CH3COOCH3* + CH3CH2OH*
CH3OH + CH3CH2COOCH3* ꢀ CH3COOCH3* + CH3CH2OH
CH3OH* + CH3CH2COOCH3 ꢀ CH3COOCH3 + CH3CH2OH*
CH3CH2OH* ꢀ CH3CH2OH + *
1
1
1
1
1
1
1
1
CH3COOCH3* ꢀ CH3COOCH3 + *

6
2
E. Van de Steene et al. / Journal of Molecular Catalysis A: Chemical 359 (2012) 57–68
Table 5
ꢀ
Reaction rates based on the mechanism and rate-determining step (with K = Kiai).
i
Abbreviation
Rate-determining step
Rate equation
LH mechanism
kMeOH(aMeOH−(1/Keq)(aMeOAcaEtOH/aEtOAc))
LH-MeOH
CH3COOH + * ꢀ CH3COOH*
r = 1
ꢀ
ꢀ
ꢀ
+(KMeOH/Keq)(aMeOAcaEtOH/aEtOAc)+K
+K
+K
EtOAc
MeOAc
EtOH
kEtOAc(aEtOAc−(1/Keq)(aMeOAcaEtOH/aMeOH))
LH-EtOAc
LH-SR
CH3CH2COOCH3 + * ꢀ CH3CH2COOCH3*
CH3COOH* + CH3CH2COOCH3* ꢀ CH3COOCH3* + CH3CH2OH*
CH3CH2OH* ꢀ CH3CH2OH + *
r = 1+K
ꢀ
ꢀ
ꢀ
+(KEtOAc/Keq)(aMeOAcaEtOH/aMeOH)+K
+K
MeOH
MeOAc
EtOH
kSRKMeOHKEtOAc(aMeOHaEtOAc−(1/Keq)aMeOAcaEtOH)
r =
ꢀ
ꢀ
ꢀ
ꢀ
2
)
(
1+K
+K
+K
+K
MeOH
EtOAc
MeOAc
EtOH
kEtOHKeq((aMeOHaEtOAc/aMeOAc)−(1/Keq)aEtOH)
LH-EtOH
LH-MeOAc
r = 1+K
r = 1+K
ꢀ
ꢀ
ꢀ
+K
+K
+KEtOHKeq(aMeOHaEtOAc/aMeOAc)
MeOH
EtOAc
MeOAc
kMeOAcKeq((aMeOHaEtOAc/aEtOH)−(1/Keq)aMeOAc)
CH3COOCH3* ꢀ CH3COOCH3 + *
ꢀ
ꢀ
ꢀ
+K
+K
+KMeOAcKeq(aMeOHaEtOAc/aEtOH)
MeOH
EtOAc
EtOH
ER mechanism with acetate adsorption
kEtOAc(aEtOAc−(1/Keq)(aMeOAcaEtOH/aMeOH))
ER-EtOAc
CH3CH2COOCH3 + * ꢀ CH3CH2COOCH3*
r = 1
ꢀ
+(KEtOAc/Keq)(aMeOAcaEtOH/aMeOH)+K
MeOAc
kSRKEtOAc(aMeOHaEtOAc−(1/Keq)aMeOAcaEtOH)
ER-EtOAc-SR
ER-MeOAc
CH3CH2COOCH3* + CH3OH ꢀ CH3CH2OH + CH3COOCH3*
CH3COOCH3* ꢀ CH3COOCH3 + *
r =
r =
ꢀ
ꢀ
1
+K
+K
EtOAc
MeOAc
kMeOAcKeq((aMeOHaEtOAc/aEtOH)−(1/Keq)aMeOAc)
ꢀ
1
+K
+KMeOAcKeq(aMeOHaEtOAc/aEtOH)
EtOAc
ER mechanism with alcohol adsorption
kMeOH(aMeOH−(1/Keq)(aMeOAcaEtOH/aEtOAc))
ER-MeOH
ER-MeOH-SR
ER-EtOH
CH3COOH + * ꢀ CH3COOH*
r = 1
ꢀ
+(KMeOH/Keq)(aMeOAcaEtOH/aEtOAc)+K
EtOH
kSRKMeOH(aMeOHaEtOAc−(1/Keq)aMeOAcaEtOH)
CH3COOH* + CH3CH2COOCH3 ꢀ CH3CH2OH* + CH3COOCH3
CH3CH2OH* ꢀ CH3CH2OH + *
r =
r =
ꢀ
ꢀ
1
+K
+K
MeOH
EtOH
kEtOHKeq((aMeOHaEtOAc/aMeOAc)−(1/Keq)aEtOH)
ꢀ
1
+K
+KEtOHKeq(aMeOHaEtOAc/aMeOAc)
MeOH
of three elementary steps. The steps considered according to the
rival mechanisms are summarized in Table 4. For the LH mecha-
nisms, both reactants are adsorbed on a catalytically active site.
Adsorbed ethyl acetate reacts with adsorbed methanol to form
adsorbed methyl acetate and adsorbed ethanol. Methyl acetate and
ethanol finally desorbs from the active sites.
For the ER mechanisms, ethyl acetate or methanol first adsorbs
on a catalytically active site. Then the adsorbed reactant reacts with
the other reactant from the bulk phase to form methyl acetate and
ethanol. One of these products is adsorbed. In the third step, the
adsorbed product, methyl acetate or ethanol, desorbs.
Each step in the LH and ER mechanism can be considered as
potentially rate determining, leading to 11 rival expressions for the
corresponding reaction rates, see Table 5. The concentrations of the
surface species are obtained assuming quasi equilibration of the
other elementary steps in the reaction mechanism. In combina-
tion with a balance over the active sites, analytical expressions are
derived for the reaction rate equations as a function of the activities
of the reaction components, the reaction rate coefficient and the
adsorption and surface reaction equilibrium coefficients (Table 5).
molar ratios result in higher ethyl acetate conversions. Because
the EtOAc/catalyst amount ratio was kept constant in these exper-
iments, this also corresponds to higher reaction rates. Also, the
equilibrium conversion increases with increasing initial reactant
molar ratio, see Fig. 5 and Table 3. For lower MeOH:EtOAc ratios,
longer batch times are needed to reach thermodynamic equilib-
rium.
4.2. Model discrimination
The model parameters of the rival models reported in Table 5
have been estimated by regression. Model discrimination is per-
formed in order to determine the best performing model and,
correspondingly, the most likely reaction mechanism, including
the rate-determining step. Prior to regression and discrimination,
initial parameter estimates have been determined.
8
6
4
2
0
0
0
0
4
4
4
. Results and discussion
.1. Experimental
.1.1. Temperature effect
The temperature effect on the transesterification reaction rate
was studied by conducting experiments between 303 and 333 K. In
accordance with the Arrhenius law, a higher temperature results
in a higher transesterification rate and a correspondingly higher
ethyl acetate conversion at the same batch time, see Fig. 4. The
final, equilibrium conversion is practically temperature indepen-
dent, which is a logic consequence of the limited reaction enthalpy
of the investigated transesterification reaction (not shown) [14].
3
03 K
313 K
333 K
2
4
6
8
10
12
14
Batch time (kg s)
4
.1.2. Initial reactant molar ratio effect
The effect of the initial molar methanol to ethyl acetate ratio
Fig. 4. Simulated (lines) and experimental (symbols) conversion of ethyl acetate
versus batch time at different temperatures (legend) (initial MeOH:EtOAc molar
ratio = 1:1, simulation model = ER-MeOH-SR).
in the range of 1:1 to 10:1 is shown in Fig. 5. Higher initial

E. Van de Steene et al. / Journal of Molecular Catalysis A: Chemical 359 (2012) 57–68
63
Table 6
1
00
Statistical evaluation of all models (1012 experimental points, experimental condi-
tions as reported in Table 1).
80
60
40
20
0
Model
Number of statistically
significantly estimated
RSSQ
F
(
total) parameters
ER-EtOAc/LH-EtOAca
ER-MeOH-SR
ER-EtOAc-SR
PH
ER-EtOH/LH-EtOH
ER-MeOAc/LH-MeOAc
3 (4)/3 (6)
4 (4)
4 (4)
5
5
7
8
9
9
25
59
31,000
21,000
15,000
41,000
20,000
18,000
8000
2 (2)
a
3 (4)/3 (6)
3 (4)/3 (6)
3 (4)/3 (6)
4 (6)
a
a
ER-MeOH/LH-MeOH
LH-SR
1000
a
Models have become mathematically equivalent because of the not statistically
significantly estimated parameters
1
:1
5:1
10:1
0
10
20
30
regression range from 1008 to 40,650. Typically, only one or two of
the adsorption equilibrium coefficients could be estimated signif-
icantly different from zero. As a result, various ER and LH models,
such as ER-EtOAc and LH-EtOAc, LH-EtOH and ER-EtOH, and ER-
MeOH and LH-MeOH, became mathematically equivalent to each
other.
batch time (kg s)
Fig. 5. Simulated (lines) and experimental (symbols) conversion of ethyl acetate
versus batch time at different initial MeOH:EtOAc molar ratios (legend) (tempera-
ture = 333 K, simulation model = ER-MeOH-SR).
The best performing models, i.e., those with the lowest resid-
ual sum of squares and the higher F value (31,000), correspond to
an Eley–Rideal or a Langmuir–Hinshelwood mechanism with ethyl
acetate adsorption as the rate-determining step (ER-EtOAc/LH-
EtOAc). Model ER-MeOH-SR, corresponding to an Eley–Rideal
mechanism with methanol adsorption and with the reaction of
ethyl acetate from the bulk with adsorbed methanol on the cata-
lyst surface as rate-determining step, has also a good performance
in terms of the residual sum of squares. Because it contains an
additional, significantly estimated parameter compared to the two
previously discussed models, the corresponding F value is limited
to about 20,000, however. Similarly, the ER-EtOAc-SR model has
a lower F value for the global significance of the regression, and
moreover, the corresponding residual sum of squares is about 50%
higher than that obtained with the above discussed models. The
PH-model has an even higher residual sum of squares, but because
this model has only two adjustable parameters, it has the highest
F value for the global significance of the regression. The next four
models that still perform reasonably well, i.e., ER-EtOH/LH-EtOH,
ER-MeOAc/LH-MeOAc, all contain 3 significantly estimated param-
eters and correspond to Eley–Rideal mechanisms with ethanol
desorption (LH-EtOH is mathematically equivalent to ER-EtOH),
resp. methyl acetate desorption as rate-determining steps (LH-
MeOAc is mathematically equivalent to ER-MeOAc). However, the
combination of (1) F values for the global significance of the regres-
sion below 20,000 and (2) RSSQ exceeding 8, is clearly inferior
compared to the just described models. The likelihood ratios of
these 4 models compared to the best performing models exceed
18 and, hence, they can be rejected with a probability of 90%. The
other models, i.e., ER-MeOH/LH-MeOH, LH-SR, all have significantly
higher residual sum of squares and corresponding F values below
4
.2.1. Initial parameter estimates
Values for the rate coefficient at each investigated tempera-
ture were estimated by isothermal regressions using the PH model,
see Fig. 6. A linear regression of the obtained values to the Arrhe-
nius relationship resulted in initial estimates of the activation
−1
energy EA amounting to 50.3 kJ mol
and of the rate coeffi-
cient at reference temperature of 328.38 K amounting to 2.62 ×
−
6
3
−1
−1 −1
1
0
m kg mol s .
cat
Initial values for the different adsorption equilibrium coeffi-
3
−1
cients were taken from the literature and set equal to 0.1 m mol
28]. From initial regression efforts, it was clear that, in the rather
[
narrow temperature range that has been investigated, the tem-
perature dependence of the adsorption coefficients K would be
practically impossible to estimate significantly. Hence, in an effort
to reduce the number of adjustable parameters, the determination
of adsorption coefficient values, K , has been limited to an average
i
i
value over the investigated temperature range.
4.2.2. Performance evaluation between all rival models
The model regression and discrimination results are shown in
Table 6. The obtained F values for the global significance of the
-5.0
-5.5
-6.0
-6.5
-7.0
-7.5
-8.0
ln k= -6061.9/T + 12.515
1
0,000 and, hence, will not be further considered either.
Table 7
Statistical evaluation of the 4 best performing models (1282 experimental points,
experimental conditions as reported in Table 1).
Model
Number of statistically
significantly estimated
RSSQ
F
(
total) parameters
3
.00
3.05
3.10
3.15
000/T (K )
3.20
3.25
3.30
ER-MeOH-SR
ER-EtOAc-SR
PH
4 (4)
3 (4)
2 (2)
3 (4)/3 (6)
27
35
36
48
62,000
-
1
78,000
149,000
47,000
1
a
ER-EtOAc/LH-EtOAc
Fig. 6. Arrhenius diagram for the transesterification reaction of methanol and ethyl
acetate using Lewatit K1221 catalyst (symbols obtained by isothermal regression
with the PH model, line obtained by linear regression).
a
Models have become mathematically equivalent because of the not statistically
significantly estimated parameters

6
4
E. Van de Steene et al. / Journal of Molecular Catalysis A: Chemical 359 (2012) 57–68
Table 8
Parameter estimates with their 95% confidence interval, obtained by regression of 1282 experimental points (Tref = 328.38 K).
ER-EtOAc/LH-EtOAc
ER-MeOH-SR
ER-EtOAc-SR
PH
kTref (10 3 m3 kg mol
−
−1
−1 −1
s
)
0.053 ± 0.002
52.538 ± 1.542
2.643 ± 0.194
0.259 ± 0.052
50.190 ± 0.940
67.130 ± 0.719
49.328 ± 0.995
n.s.
0.003 ± 0.00003
cat
EA (kJ mol 1)
−
49.405 ± 0.967
−
3
3
−1
−1
−1
a
KEtOAc(10
KEtOH (10
KMeOH (10
m
m
mol
mol
mol
)
)
n.a.
n.a.
n.a.
n.a.
n.a.
−
3
3
a
b
n.a. /n.s.
0.233 ± 0.040
n.a.
n.a.
−
3
3
m
)
)
n.a./n.s.
n.s./n.s.
0.015 ± 0.004
KMeOAc (10 m3 mol
−
3
−1
n.a.
a
0.052 ± 0.019
a
not applicable
not significantly estimated
b
The physical meaning of the remaining models (ER-EtOAc/LH-
EtOAc, ER-MeOH-SR, ER-EtOAc-SR and PH) is explained below.
ER-EtOAc corresponds to an Eley–Rideal mechanism with ethyl
acetate adsorption as rate-determining step. Ethyl acetate adsorp-
tion is then followed by reaction with methanol from the bulk
which are both estimated significantly. This results in the highest
F value among the considered models, despite the RSSQ which
is about one third bigger of the lowest RSSQ, obtained with
ER-MeOH-SR. Due to the absence of any adsorption related terms
in the PH-model, it cannot account for any surface coverage effects,
however. ER-EtOAc/LH-EtOAc, with only 3 parameters that are
estimated significantly different from zero, has an even higher
RSSQ. Hence, this model is considered to describe the experimental
data set not adequately.
(
Table 5). The 4 adjustable parameters in this model are the acti-
vation energy and the rate coefficient kEtOAcTref at the reference
temperature, the adsorption equilibrium coefficient of respectively
ethyl acetate and methyl acetate, KEtOAc and KMeOAc. The first three
model parameters are estimated statistically significantly different
from zero, while the adsorption equilibrium coefficient of methyl
acetate is not. The final rate expression, hence, becomes:
The likelihood ratio of ER-MeOH-SR and ER-EtOAc-SR (Eq. (12))
is higher than 18, showing that there is a higher probability that
the ER-MeOH-SR-model describes the data set better than the ER-
EtOAc-SR-model.
kEtOAc(aEtOAcaMeOH − (1/Keq)(aMeOAcaEtOH/aMeOH))
−1
Activation energies of about 50 kJ mol are obtained, irrespec-
r =
(17)
1
+ (KEtOAc/Keq)(a
MeOAc
aEtOH/aMeOH
)
tive of the model used. For a transesterification catalyzed with an
acid ion-exchange resin similar results were published in litera-
ture [14,27–29], e.g., López et al. [28] found an activation energy
ER-MeOH-SR also corresponds to an Eley–Rideal mechanism but
with the reaction of ethyl acetate from the bulk with adsorbed
methanol on the catalyst surface as rate-determining step, see
Table 5. The 4 model parameters are the activation energy and the
rate coefficient ksr at reference temperature, the adsorption equi-
librium coefficients of respectively ethanol and methanol, KEtOH
and KMeOH. All these model parameters are estimated significantly
different from zero. The corresponding rate expression is given in
Table 5.
ER-EtOAc-SR also corresponds to an Eley–Rideal mechanism
with the reaction of adsorbed ethyl acetate with methanol from the
bulk as rate-determining step (Table 5). The 4 model parameters are
the activation energy and the rate coefficient ksr at reference tem-
perature, the adsorption equilibrium coefficients of respectively
ethyl acetate and methyl acetate KEtOAc and KMeOAc and are all
estimated significantly different from zero. The corresponding rate
expression is given in Table 5.
−
1
of 48.5 kJ mol for the transesterification of triacetin to diacetin
on a Nafion® SAC-13 ion exchange resin, which was compara-
−
1
ble with that for H SO (46.1 kJ mol ). An activation energy of
2
4
−
1
2
0 kJ mol
was determined by Dossin et al. [3] for the trans-
esterification of ethyl acetate with methanol on a base MgO
catalyst.
−
3
The kTref at 328.38 K varies between 0.003 × 10 and 67.130 ×
−
3
3
−1
−1 −1
10 m kg mol s , depending on the model used. This value
cat
is similar to the one published by Bo z˙ ek-Winkler and Gmehling
[14] for the transesterification of methyl acetate and n-butanol
catalyzed by Amberlyst 15.
The estimated values of the adsorption equilibrium coefficients
for the transesterification with acid ion-exchange resins are quite
similar with the published ones [14,28,29]. Based on the structural
similarity and on the acid dissociation coefficient of MeOH and
EtOH, which are rather close to each other, Dossin et al. [7] decided
to estimate a single adsorption equilibrium coefficient for both
alcohols. Results published by Bo z˙ ek-Winkler and Gmehling [14]
and Tesser et al. [2] indicate that the alcohol adsorption equilib-
rium coefficient values are indeed relatively close to each other,
but that the one corresponding to methanol has the lowest value.
The ratio of the adsorption coefficients as determined from the ER-
MeOH-SR model amounts to 15. In particular the value obtained
for KEtOH significantly exceed the one reported in literature, i.e.,
The PH model does not take any adsorption into account. Ethyl
acetate and methanol react from the bulk to form ethanol and
methyl acetate in the bulk. The corresponding rate expression and
the explanation of the parameters is given in Eq. (16) in Section 3.
4.2.3. Discrimination between best performing models using an
experimental design
A D-optimum design of experiments has been performed aim-
ing at a more precise parameter estimation in the remaining 4 rival
models. This design is expected to simultaneously allow further
model discrimination. It proposes experiments in the outer range
of the operating conditions, in particular at a molar ratio of 1:1, a
−
3
3
−1
−3
3
−1
[14]. As a
0.233 × 10 m mol
versus 0.0289 × 10 m mol
result, a refined version of this model has been tested, in which this
ratio has been fixed to 2. The remaining 3 adjustable parameters
have been determined by regression. With an RSSQ of 31 and an
F value amounting to 80,000 with only 3 adjustable parameters,
this refined version performs statistically better than the original
ER-MeOH-SR model but is still inferior to the PH model. Hence, an
ultimate model refinement consisted of fixing the adsorption coef-
ficients for methanol and ethanol at their literature determined
−
3
temperature of 303 K and a catalyst amount of 4.0 × 10 kg. These
experiments were performed. Subsequently, regression has been
performed for the rival models using the enlarged dataset. The sta-
tistical performance of the models is shown in Table 7, while the
corresponding parameter estimates are reported in Table 8.
The obtained F values range from 62,000 to 149,000.
−
3
3
−1
−3
3
−1
ER-MeOH-SR has the lowest residual sum of squares (27) with
all four adjustable parameters being estimated significantly. With
one parameter less, ER-EtOAc-SR has a higher F value but also a
higher RSSQ (35). The PH-model only has 2 adjustable parameters,
values, i.e., 0.0140 × 10 m mol
and 0.0289 × 10 m mol
,
respectively [14]. With an RSSQ amounting to 36 and a correspond-
ing F value of 142,000, this ultimate version of the ER-MeOH-SR
model is statistically practically identical to the PH model.

E. Van de Steene et al. / Journal of Molecular Catalysis A: Chemical 359 (2012) 57–68
65
Besides the statistical significance of the models, the models are
also evaluated on their physicochemical significance. Sulfonic acid
ion-exchange resins show a higher affinity for alcohols than for
esters [14,18,30–33]. This indicates that methanol will adsorb more
strongly than ethyl acetate, which is in line with the assumptions
made in ER-MeOH-SR but in clear contrast with the assumptions
made in ER-EtOAc/LH-EtOAc and ER-EtOAc-SR. These physico-
chemical considerations, combined with the higher RSSQ of the
latter models, see Table 8, lead to the elimination of ER-EtOAc/LH-
EtOAc and ER-EtOAc-SR from the list of rival models. The PH-model
lacks any adsorption effect. The good statistical performance of the
PH-model, in terms of F value for the significance of the regression
is an indication that the partial replacement of adsorbed methanol
reactant by adsorbed ethanol product has only a limited effect on
the simulated ethyl acetate concentration at the investigated oper-
0
0
0
.4
.2
.0
-
0.2
0.4
-
0
1
2
3
4
5
6
ating conditions. Fixing the adsorption coefficient values KMeOH
,
simulated concentration EtOAc (mol m-3
)
KEtOH in ER-MeOH-SR at literature reported values, the correspond-
ing F value and RSSQ of ER-MeOH-SR become comparable with the
statistical results of the PH model. Hence, because of the addi-
tional physicochemical phenomena accounted for in the former
model based on literature reported parameter values, ER-MeOH-SR
is selected from the rival models as the most adequate model.
Fig. 7 shows the residual diagram of the experimental and calcu-
lated ethyl acetate concentration using ER-MeOH-SR. The diagram
shows a good agreement between experimental and simulated val-
ues and is not indicative of any systematic deviation. The agreement
between ER-MeOH-SR model calculations and the experimental
data is also evident from Figs. 4 and 5.
Fig. 7. Residual figure for the concentration of ethyl acetate for the complete set
of 1282 data points. The simulated values are obtained using Eq. (16) with the
estimated model parameters of ER-MeOH-SR (Table 8). Range of experimental con-
ditions (Table 1).
acetate, step 1, followed by a nucleophilic attack of the methanol on
this protonated ethyl acetate, step 2, yielding a tetrahedral inter-
mediate. In step 3, the proton migrates from the reacting methanol
oxygen atom to the product ethanol oxygen atom after which the
tetrahedral intermediate decomposes into a protonated methyl
acetate and ethanol, step 4. The fifth and last step in the reaction
mechanism is the regeneration of the acid site [5,28,34–39].
According to the selected ER-MeOH-SR model, the transester-
ification mechanism on Lewatit K1221, would be clearly distinct
from the mechanism presented in Fig. 8. Methanol should be
adsorbed on the catalyst’s active site with ethyl acetate reacting
from the bulk, see Table 4 and Section 4.2.3, while, according
to the homogeneously catalyzed mechanism, it should be ethyl
acetate which is adsorbed, c.q., protonated on the acid sites with
4
4
.3. Selected model assessment
.3.1. Further considerations on the actual reaction mechanism
Homogeneous acid catalyzed transesterification has a mech-
anism that is equivalent to hydrolysis [5]. The mechanism, see
Fig. 8, starts with the protonation of the carboxyl group of the ethyl
Fig. 8. Homogeneous acid-catalyzed reaction mechanism for the transesterification of ethyl acetate with methanol.

6
6
E. Van de Steene et al. / Journal of Molecular Catalysis A: Chemical 359 (2012) 57–68
Fig. 9. Heterogeneous acid ion-exchange resin catalyzed reaction mechanism for the transesterification of ethyl acetate with methanol.
methanol reacting from the bulk. This apparent contradiction
can be explained by invoking a physical adsorption step in the
mechanism, which is preceding any chemical elementary step
on the acid sites. In this mechanism including reactant physical
adsorption, both methanol and ethyl acetate are physisorbed
in the pores of the resin. The physisorption of methanol being
more pronounced than that of ethyl acetate, the physisorption of
methanol is more likely to experience saturation effects, while that
of ethyl acetate is situated in the Henry regime. Starting from these
physically adsorbed reactants, the homogeneous acid catalyzed
mechanism can occur within the resin’s pores, i.e., ethyl acetate
protonation followed by reaction with methanol. As a result, an
overall rate equation for transesterification on Lewatit K1221 will
require an adsorption term for methanol but not for ethyl acetate,
if in addition to the physisorption in the Henry regime, also the
chemisorption/protonation of ethyl acetate is not suffering from
saturation effects. Given the pKa values of methanol and ethyl
acetate, i.e., 16 vs. 21, it is indeed the protonation of ethyl acetate,
which is expected to be more pronounced than that of methanol.
The above considerations about the reaction mechanism are in
agreement with the results obtained by López et al. [28]. These
authors identified a methanol partial reaction order, which tends
to zero at high methanol concentrations [40]. It indicates the
importance of accounting for methanol adsorption in the reaction
mechanism, even if methanol is unlikely to interact directly with
the acid sites. As a result, Lopez et al. [28] ultimately derived a
rate equation starting from a classical Eley–Rideal mechanism with
ester adsorption on the active sites and reaction with methanol
from the bulk that was extended with a methanol adsorption term
KMeOHCMeOH. An adsorption term corresponding to the product
ester was not included. The adsorption coefficients for methanol
and the reacting ester, KMeOH and KTG, were small and similar. How-
ever, due to the significantly higher methanol concentration than
ester concentration, only the methanol adsorption term was math-
ematically significant, as it is the case in the model developed as
part of the present work.
The view on the acid catalyzed transesterification by ion
exchange resins developed in this work also agrees with the inter-
pretation by Alonso et al. [34]. Their principal peculiarities between
heterogeneous and homogeneous mechanisms refer to the follow-
ing two aspects: firstly, the activation of the carbonyl group of
the ester comes with the chemisorption of the ester molecule on
the Brønsted active site. The second aspect refers to the methanol
involved in the reaction, which is the rate-determining step in the
Eley–Rideal model. The methanol is not chemisorbed, although
methanol can also be chemisorbed on the Brønsted acid sites, but
the methanol is available in the liquid medium present in the pores
of the solid catalyst [34] (Fig. 9).

E. Van de Steene et al. / Journal of Molecular Catalysis A: Chemical 359 (2012) 57–68
67
Fig. 10. Calculated physisorbed fractions as a function of the batch time at 333 K, and with initial MeOH:EtOAc molar ratio = 10:1 and 1:1 (MeOH* and EtOH* physisorbed
fractions of methanol and ethanol, respectively; (*) fraction of free physisorption sites).
4
.3.2. Evolution of the physisorbed fractions in the catalyst pores
The evolution of the physisorbed methanol and ethanol frac-
and product adsorption, resulted in the selection of an Eley–Rideal
mechanism considering methanol adsorption with subsequent
reaction with ethyl acetate from the bulk as rate-determining step
as the most adequate model. No adsorption term related to ethyl
acetate protonation had to be incorporated, indicating that the
resin’s acid sites were mainly unoccupied at the investigated oper-
ating conditions.
tions, i.e., ꢂMeOH
physisorption sites, ꢂ , as a function of batch time at different
experimental conditions has been calculated using the site balance
and the quasi-equilibrium assumption for methanol and ethanol
adsorption and are shown in Fig. 10:
and ꢂEtOH , as well as the fraction of free
∗ ∗
*
The model discrimination as performed in this work demon-
strates the need of an assessment of the physical significance of
the model and the corresponding parameters. This can be achieved
by comparison with literature reported information on the vari-
ous elementary steps in the reaction mechanism. The combined
information about the homogeneously acid catalyzed transesterifi-
cation reaction mechanism and physical adsorption measurements
of alcohols and esters on resins have allowed an advanced inter-
pretation of the selected model based on typical discrimination
activities.
1
= ꢂ∗ + ꢂMeOH
∗
+ ꢂEtOH
∗
(20)
(21)
1
ꢂ∗ =
1
+ KMeOHaMeOH + KEtOHaEtOH
KMeOHaMeOH
ꢂMeOH
∗
= 1
(22)
(23)
+ KMeOHaMeOH + KEtOHaEtOH
KEtOHaEtOH
ꢂEtOH
∗
= 1
+ KMeOHaMeOH + KEtOHaEtOH
At 333 K and an initial molar methanol to ethyl acetate ratio of
0:1, about 80% of the physisorption sites is free while 20% is occu-
1
pied by methanol at the start of the experiment. Near completion
of the experiment, the fraction of free sites is reduced to about
Acknowledgments
6
0%, while 25% is occupied by ethanol and 15% by methanol. The
The Research Fund of the University College Ghent and the
faculty of Engineering and Architecture of Ghent University are
acknowledged for their financial support. Lanxess is acknowledged
for providing the ion-exchange resin.
fraction of free physisorption sites on the catalyst decreases with
increasing batch time, because ethanol physisorbs more strongly
in the catalyst’s pores than methanol. As a result, rather than satu-
ration effects by methanol physisorption, some product inhibition
effect will occur, because the product ethanol is occupying about
one fourth of the available pore volume of the catalyst. At lower ini-
tial molar methanol to ethyl acetate ratios, this product inhibition
effect will, accordingly, be more pronounced, because of the higher
ester and, hence, product ethanol concentration. Given the domi-
nance of the fraction of free sites, the above described inhibition
effects are only having a moderate effect on the simulated reaction
rates. This is in agreement with the model discrimination on a sta-
tistical basis, in which the PH model was considered equivalent to
the ER-MeOH-SR model.
References
[1] J.H. Clark, Acc. Chem. Res. 35 (2002) 791–797.
[
2] R. Tesser, L. Casale, D. Verde, M. Di Serio, E. Santacesaria, Chem. Eng. J. 157
2010) 539–550.
3] T.F. Dossin, M.F. Reyniers, G.B. Marin, Appl. Catal. B-Environ. 62 (2006) 35–45.
(
[
[4] F. Ullmann, M. Bohnet, Ullmann’s Encyclopedia of Industrial Chemistry, Wiley-
VCH, 2003.
[
5] M. Smith, J. March, March’s Advanced Organic Chemistry: Reactions, Mecha-
nisms, and Structure, Wiley-Interscience, 2007.
[6] S.H. Ali, Int. J. Chem. Kinet. 41 (2009) 432–448.
[
7] T.F. Dossin, M.F. Reyniers, R.J. Berger, G.B. Marin, Appl. Catal. B-Environ. 67
2006) 136–148.
8] B.M.E. Russbueldt, W.F. Hoelderich, Appl. Catal. A: Gen. 362 (2009) 47–57.
(
[
5
. Conclusions
[9] I. Zielinska-Nadolska, K. Warmuzinski, J. Richter, Catal. Today 114 (2006)
26–230.
2
[
[
10] N. Ozbay, N. Oktar, N.A. Tapan, Fuel 87 (2008) 1789–1798.
11] K.W. Chan, Y.T. Tsai, H.M. Lin, M.J. Lee, J. Taiwan Inst. Chem. Eng. 41 (2010)
414–420.
Acid catalyzed transesterification reactions occur via a mech-
anism involving ester protonation followed by reaction with the
alcohol. The use of sulfonic acid ion-exchange resins such as
Lewatit K1221, induces a physisorption step prior to the chem-
ical elementary steps of the reaction mechanism. Because the
physisorption of alcohols is more pronounced than that of esters,
a model discrimination between a pseudo-homogeneous model
and Langmuir–Hinshelwood and Eley–Rideal models with vary-
ing assumptions related to the rate-determining step and reactant
[12] B. Saha, M. Streat, React. Funct. Polym. 40 (1999) 13–27.
[13] L. Jimenez, A. Garvin, J. Costa-Lopez, Ind. Eng. Chem. Res. 41 (2002) 6663–6669.
[14] E. Bozek-Winkler, J. Gmehling, Ind. Eng. Chem. Res. 45 (2006) 6648–6654.
[15] S. Steinigeweg, J. Gmehling, Chem. Eng. Process. 43 (2004) 447–456.
[16] G.F. Froment, K.B. Bischoff, Chemical Reactor Analysis and Design, second ed.,
Wiley, 1990.
[
17] R.J. Berger, F. Kapteijn, J.A. Moulijn, G.B. Marin, J. De Wilde, M. Olea, D. Chen,
A. Holmen, L. Lietti, E. Tronconi, Y. Schuurman, Appl. Catal. A: Gen. 342 (2008)
3–28.

6
8
E. Van de Steene et al. / Journal of Molecular Catalysis A: Chemical 359 (2012) 57–68
[
18] M.T. Sanz, R. Murga, S. Beltran, J.L. Cabezas, J. Coca, Ind. Eng. Chem. Res. 41
2002) 512–517.
[30] Y.X. Qu, S.J. Peng, S. Wang, Z.Q. Zhang, J.D. Wang, Chin. J. Chem. Eng. 17 (2009)
773–780.
(
[19] B.E. Poling, J.M. Prausnitz, J.P. O’Connell, The Properties of Gases and Liquids,
McGraw-Hill, 2001.
[31] M.T. Sanz, R. Murga, S. Beltran, J.L. Cabezas, J. Coca, Ind. Eng. Chem. Res. 43
(2004) 2049–2053.
[
[
[
20] M. Schwaab, J.C. Pinto, Chem. Eng. Sci. 62 (2007) 2750–2764.
21] N.R. Draper, H. Smith, Applied Regression Analysis, Wiley, 1981.
22] D.M. Himmelblau, Basic Principles and Calculations in Chemical Engineering,
Prentice Hall, 1989.
[32] Y. Zhang, L. Ma, J.C. Yang, React. Funct. Polym. 61 (2004) 101–114.
[33] P. Delgado, M.T. Sanz, S. Beltran, Chem. Eng. J. 126 (2007) 111–118.
[34] D.M. Alonso, M.L. Granados, R. Mariscal, A. Douhal, J. Catal. 262 (2009) 18–26.
[35] E. Lotero, Y.J. Liu, D.E. Lopez, K. Suwannakarn, D.A. Bruce, J.G. Goodwin, Ind.
Eng. Chem. Res. 44 (2005) 5353–5363.
[23] A.C. Atkinson, B. Bogacka, M.B. Bogacki, Chemometr. Intell. Lab. Syst. 43 (1998)
1
85–198.
[36] M. Di Serio, R. Tesser, M. Dimiccoli, F. Cammarota, M. Nastasi, E. Santacesaria,
J. Mol. Catal. A-Chem. 239 (2005) 111–115.
[
[
[
[
24] Y. Zhang, T.F. Edgar, Ind. Eng. Chem. Res. 47 (2008) 7772–7783.
25] H.T.R. Teo, B. Saha, J. Catal. 228 (2004) 174–182.
26] B. Saha, M.M. Sharma, React. Funct. Polym. 34 (1997) 161–173.
27] B.Y. Xu, W.J. Zhang, X.M. Zhang, C.F. Zhou, Int. J. Chem. Kinet. 41 (2009) 101–
[37] M. Di Serio, R. Tesser, L. Pengmei, E. Santacesaria, Energ. Fuel 22 (2008) 207–217.
[38] E. Lotero, J.G. Goodwin, D.A. Bruce, K. Suwannakarn, Y.J. Liu, D.E. Lopez, in: J.J.
Spivey, K.M. Dooley (Eds.), Catalysis, vol. 19, Royal Soc Chemistry, Cambridge,
2006, pp. 41–83.
1
06.
[
[
28] D.E. Lopez, J.G. Goodwin, D.A. Bruce, J. Catal. 245 (2007) 381–391.
29] S.H. Ali, O. Al-Rashed, F.A. Azeez, S.Q. Merchant, Bioresour. Technol. 102 (2011)
[39] U. Schuchardt, R. Sercheli, R.M. Vargas, J. Braz, Chem. Soc. 9 (1998) 199–210.
[40] M. Mazzotti, B. Neri, D. Gelosa, A. Kruglov, M. Morbidelli, Ind. Eng. Chem. Res.
36 (1997) 3–10.
1
0094–10103.