Optimisation of Ru-promoted Ir-catalysed methanol carbonylation utilising response surface methodology

Article abstract of DOI:10.1016/j.apcata.2010.12.036

In this study, central composite design (CCD) at five levels (-1.63, -1, 0, +1, +1.63) combined with response surface methodology (RSM) have been applied to optimise methanol carbonylation using a ruthenium-promoted iridium catalyst in a homogenous phase. The effect of seven process variables, including temperature, pressure, iridium, ruthenium, methyl iodide, methyl acetate and water concentrations, as well as their binary interactions, were modelled. The determined R² values greater than 0.9 for the rate and methane formation data confirmed that the quadratic equation properly fitted the obtained experimental data. The optimum conditions for maximum rate and minimum methane formation were obtained via the RSM to be: temperature of 191 °C, pressure of 32.48 barg, iridium concentration of 939.40 ppm, ruthenium concentration of 2099.22 ppm, methyl iodide concentration of 14.46 wt.%, methyl acetate concentration of 17.55 wt.% and water content of 7.60 wt.%. The results predicted by the developed correlation at the optimum determined conditions, 28.63 mol/l h for the reaction rate and 1.97 mol% CH₄, were reasonably compared with the experimental data obtained for the reaction rate and methane formation of 27.10 mol/l h and 3.06 mol% CH₄, respectively. Crown Copyright

Full text of DOI:10.1016/j.apcata.2010.12.036

Applied Catalysis A: General 394 (2011) 166–175
Contents lists available at ScienceDirect
Applied Catalysis A: General
journal homepage: www.elsevier.com/locate/apcata
Optimisation of Ru-promoted Ir-catalysed methanol carbonylation utilising
response surface methodology
Vahid Hosseinpoura, Mohammad Kazemeinia,∗, Alireza Mohammadrezaeeb
a Department of Chemical and Petroleum Engineering, Sharif University of Technology, Azadi Avenue, P.O. Box 11365-9465, Tehran, Iran
b Petrochemical Research and Technology Company, National Petrochemical Company, P.O. Box 14965-115, Tehran, Iran
a r t i c l e i n f o
a b s t r a c t
Article history:
In this study, central composite design (CCD) at five levels (−1.63, −1, 0, +1, +1.63) combined with
response surface methodology (RSM) have been applied to optimise methanol carbonylation using a
ruthenium-promoted iridium catalyst in a homogenous phase. The effect of seven process variables,
including temperature, pressure, iridium, ruthenium, methyl iodide, methyl acetate and water concen-
trations, as well as their binary interactions, were modelled. The determined R2 values greater than
0.9 for the rate and methane formation data confirmed that the quadratic equation properly fitted the
obtained experimental data. The optimum conditions for maximum rate and minimum methane forma-
tion were obtained via the RSM to be: temperature of 191 ◦C, pressure of 32.48 barg, iridium concentration
of 939.40 ppm, ruthenium concentration of 2099.22 ppm, methyl iodide concentration of 14.46 wt.%,
methyl acetate concentration of 17.55 wt.% and water content of 7.60 wt.%. The results predicted by the
developed correlation at the optimum determined conditions, 28.63 mol/l h for the reaction rate and
1.97 mol% CH4, were reasonably compared with the experimental data obtained for the reaction rate and
methane formation of 27.10 mol/l h and 3.06 mol% CH4, respectively.
Received 4 November 2010
Received in revised form
15 December 2010
Accepted 27 December 2010
Available online 4 January 2011
Keywords:
Acetic acid
Carbonylation of methanol
Iridium/ruthenium
Optimisation
RSM
Crown Copyright © 2011 Published by Elsevier B.V. All rights reserved.
1. Introduction
cess utilised iodide as a promoter and required high temperature
(250 ◦C) and pressure (600 bar) [10,11]. Today, this process is found
Acetic acid is a very important petrochemical product with
worldwide production of over 10 million tons per year [1]. Acetic
acid is widely used for the production of vinyl acetate monomer
(VAM) and the synthesis of acetic anhydride. It is also used as a sol-
vent for producing purified terephthalicacid (PTA). The traditional
route for the production of acetic acid is aerobic fermentation of
ethanol. The first commercial process which synthesised acetic acid
used an organomercury compound as a catalyst for the oxidation
of acetaldehyde [2]. Today, several processes are used to produce
acetic acid. These include methanol carbonylation [3], methyl for-
mate isomerisation [4], synthesis of gas to acetic acid [2], oxidation
of ethylene in the gas phase [5], ethane oxidation [6], methane
carbonylation [7] and the use of methane and carbon dioxide [8].
Amongst these processes, methanol carbonylation has been used
extensively in the worldwide production of acetic acid. Recently,
improved methanol carbonylation in the vapour phase has been
attempted [9]; however, the homogenous phase process is still the
major route for the production of this species. Nickel, and later
cobalt, was the first catalytic compound used in the carbonylation
process of methanol to acetic acid developed by BASF. This pro-
only in books and in the literature. The utilisation of a rhodium
catalyst and iodide as a co-catalyst beginning in 1973 led to the
development of a new process for methanol carbonylation known
as the Monsanto process which took place under relatively mild
operating conditions (i.e., temperature of 180–200 ◦C and pressure
of 30–40 bar) [12,13]. In 1996, the BP Chemical Co. developed a
new process, named CativaTM, based on an iridium/iodide catalyst
system for methanol carbonylation [14,15]. Due to a low partial
pressure of CO in the purification stage in the Monsanto process,
the expensive rhodium catalyst precipitated; thus, to establish cat-
alyst stability and a high reaction rate, a high concentration of water
in the reactor composition was required. This restriction necessi-
tated more distillation columns in the purification stage to remove
considerable amounts of water in the production stream. Stronger
metal–ligand bonding with iridium relative to rhodium inhibited
iridium loss. Due to greater stability of the iridium complex in the
CativaTM process, the water concentration in the reactor was low,
thus making the capital cost of this process relatively lower than the
Monsanto process [10,12,14]. Using promoters for improving cat-
alytic activity was a key feature of the CativaTM process for which
ruthenium was utilised [16–19]. Due to process limitations accom-
panied by the Monsanto process and the significant price difference
between iridium and rhodium, CativaTM is the preferred route for
this process. The catalytic cycle of iridium catalysed methanol car-
∗
Corresponding author. Tel.: +98 21 6600 5819; fax: +98 21 6602 2853.
E-mail address: kazemini@sharif.edu (M. Kazemeini).
0926-860X/$ – see front matter. Crown Copyright © 2011 Published by Elsevier B.V. All rights reserved.
doi:10.1016/j.apcata.2010.12.036

V. Hosseinpour et al. / Applied Catalysis A: General 394 (2011) 166–175
167
it is formed according to the following reaction:
CH3OH + CO → CH4 + CO2
(1)
During methanol carbonylation utilising an iridium/iodide cata-
lyst, methane is formed as a result of the reaction of [CH3Ir(CO)2I3]−
with acetic acid through hydrogenolysis [26].
In the current research, the effects of seven factors and
their interactions on methanol carbonylation were experimentally
investigated. Then, the empirical relations for the reaction rate as
well as methane formation were developed. To the best of our
knowledge, no such study has previously been reported on the
application of the RSM technique for optimising the methanol car-
bonylation process.
2. Materials and methods
2.1. Materials
Methyl acetate, methyl iodide and acetic acid were obtained
from the Merck Chemical Company. The iridium catalyst
(IrCl3·xH2O, 52.88% Ir) and the promoter (Ru3(CO)12, 47.2% Ru)
were supplied by the Heraeus and Strem Companies, respectively.
Carbon monoxide (99.95%) was purchased from the Technical Gas
Service Company and was used as feedstock for the carbonylation
reaction.
Fig. 1. Proposed mechanism by Forster for carbonylation of methanol using iridium
catalyst [15].
bonylation involved key steps similar to those in the well-known
rhodium carbonylation process which included oxidative addition
of methyl iodide to iridium complex and the migratory insertion
of CO to produce an acetyl complex being the precursor to the
acetic acid. Nevertheless, the iridium catalytic cycle is considerably
more complex than rhodium. Forster [15] proposed two different
catalytic cycles for iridium catalysed carbonylation of methanol
including; an anionic cycle which involved predominantly anionic
iridium intermediate species and a neutral cycle that involved neu-
tral species of iridium complex interactions both of which shown in
Fig. 1. Furthermore, it is noteworthy that, the water–gas shift reac-
tion for anionic and neutral cycles also proposed by the Forster’s
scheme.
2.2. Experimental procedures
The carbonylation reaction was performed in a 450 ml Parr
Hastelloy B2 autoclave, equipped with a magnetically-driven stir-
rer, as well as a liquid injection facility and water-fed cooling coils.
To maintain the autoclave at constant pressure, carbon monox-
ide was supplied to it. Carbon monoxide consumption throughout
the reaction was recorded with a data logger. The carbonylation
rate was calculated on the basis of the consumption rate of car-
bon monoxide. Due to the esterification of methanol by acetic acid
where it was present as a solvent in the reactor, methyl acetate was
used as the substrate in batch studies. The overall stoichiometry of
the reaction considered may be represented as follows:
The dependence of the carbonylation rate upon process vari-
ables in the iridium system is more complicated than that of
rhodium. In addition, in the CativaTM route, process variables have
complex interactions between them as well. In other words, the
carbonylation rate and selectivity in the process depend upon
temperature, partial pressure of CO, and the concentrations of
water, methyl iodide, methyl acetate, the catalyst and the pro-
moter. Therefore, to optimise reaction conditions, investigations
into the effects of individual and interaction of process variables
on the carbonylation rate and byproduct formation are required.
In previous studies [10,12], only the individual effects of process
variables were studied. In such experiments, one process vari-
able was varied while the other factors were held constant. Such
experimental methodology usually ignored interactions between
variables and led to unrealistic optimal conditions, but the use of
response surface methodology (RSM) overcame this problem. RSM
is a combination of mathematical and statistical techniques for
designing experiments which lead to an empirical model. Through
the experimental design, it considers the effect of factors and their
interactions involved for optimising the process [20–22]. The main
advantage of the RSM method is a reduction in the number of
experiments needed to evaluate the effect of variables. Recently,
optimisations utilising the RSM method have been applied to a
variety of processes [23–25].
CH3COOCH3 + H2O + CO → 2CH3COOH
(2)
Thus, consumption of 1 mol each of carbon monoxide, methyl
acetate and water led to the carbonylation of 1 mol of methanol. To
calculate the reaction rate, the cold degassed volume was utilised.
For each carbonylation reaction run, the desired amounts of methyl
acetate, methyl iodide, ruthenium, water and acetic acid were
charged into the autoclave, then sealed and pressure tested to
35 barg with nitrogen and flushed three times with CO to 5 barg
to completely purge the remaining nitrogen. The autoclave was
then heated to the reaction temperature under slow stirring at
150 rpm. Once the reaction temperature was reached, the catalyst
solution (i.e., IrCl3, acetic acid and water) was injected into the auto-
clave to initiate the reaction. Autoclave pressure was increased to
the desired level and stirred at 1300 rpm. The experimental setup
designed and constructed for this purpose of this research is shown
in Fig. 2.
In the present work, the autoclave pressure was constantly
maintained ( 5 psig) by feeding CO to the reactor. The reaction
temperature also was held constant ( 1 ◦C) by means of a heating
mantle connected to the temperature control system. Furthermore,
the excess heat generated was removed by a cooling coil. At the end
of the reaction, the autoclave was detached from the CO feed and
quenched to room temperature by a cooling coil.
Methane formation is
a major problem associated with
methanol carbonylation using an iridium/iodide system. Usually,

168
V. Hosseinpour et al. / Applied Catalysis A: General 394 (2011) 166–175
Fig. 2. Schematic diagram of the experimental setup used to perform the carbonylation reaction in this work. A: gas cylinders, B: regulator, C: on–off valve, D: mass flow
controller (MFC), E: check valve, F: stirrer, G: thermocouple, H: pressure sensor, I: catalyst injection facility, J: mantle heater, K: cooling coil, L: solenoid valve, M: condenser,
N: three-way valve and O: gas chromatography.
parameter estimations for such a model are reported elsewhere
[20,22].
2.3. Analysis of gas phase
The most popular second order experimental design is central
composite design (CCD) which was utilised in this study. CCD is
an efficient way of providing a sufficient amount of information to
test the fitness of a model. Furthermore, the CCD approach does
not require a large number of design points; thus, it reduces the
cost and time needed for performing experiments. A number of
experiments using CCD contain three sets, including: (1) fractional
factorial runs (2k−1, where k is the number of factors) studying fac-
tors at −1 (minimum) and +1 (maximum) level, (2) centre point
runs examining factors at the centre point of the design space, help-
ing to understand the curvature and replicating the data to evaluate
pure errors and (3) axial or star point runs (2k, where k is the num-
ber of factors) setting all factors to 0 (i.e., the centre point) except
one, which has the value +˛ and –˛ [20].
Analysis of gas phase was carried out on an Agilent refinery gas
analyser (RGA) 6890N Gas Chromatograph (GC) equipped with a
flame ionisation detector (FID) and thermal conductivity detector
(TCD). Samples were injected to an oven at a temperature of 50 ◦C,
which was then heated up to 180 ◦C. Helium was used as the carrier
gas and the injector, FID and TCD temperatures were 200 ◦C, 250 ◦C
and 200 ◦C, respectively. Methane content in the gas phase was
indicated as a mole percent.
2.4. Experimental design and optimisation method
RSM is a practical, economical and an easy way to optimise
a given process. The first step in a successful RSM optimisation
is to properly design the experiments for evaluating the model
parameters efficiently. The second step is to develop a polynomial
model fitted to the experimental data through regression and then
to check the model’s suitability by applying a statistical test (e.g.,
Lack-of-fit, F-test) [20–22]. The last step in the RSM is to determine
the value of factors satisfactory for optimum conditions. Usually, a
first or second order polynomial is used for the RSM analysis; fur-
thermore, for responses which include a curvature, a second order
polynomial is preferred. The general form of such polynomial is as
follows:
In this work, Design-Expert® software (Ver. 8.0.1) was utilised
to design the experiments. Here, a CCD, with ˛ = k1/4, where k
is the number of factors, was used to investigate the effect of
operating parameters on ruthenium-promoted methanol carbony-
lation utilising an iridium catalyst. The results were incorporated
to determine a satisfying empirical equation capable of predict-
ing the optimum operating conditions. To reduce the number of
experiments, resolution V was applied to the experimental design.
In resolution V, the main effects are aliased with four-factor inter-
actions, and two-factor interactions are aliased with three-factor
interactions [20]. The operation factors picked out to investigate the
undertaken reaction were: temperature, pressure, iridium concen-
tration, ruthenium concentration, methyl acetate concentration,
methyl iodide concentration and water content. The actual and
coded values for each factor investigated in this work are presented
in Table 1. The satisfaction degree of the polynomial equation devel-
oped through regression of Eq. (3) was evaluated by R2 as a measure
of the amount of variation around the mean determined by the
model through Eq. (4) and RA2dj which is a measure of the amount
k
k
k
k
ꢀ
ꢀ
ꢀꢀ
y = a0
+
aixi +
aiixi2
+
aijxixj + ε
(3)
i=1
i=1
i
j
i<j
where y is the predicted response, a0 is a constant, aij is the ith
linear coefficient, aii is the ith quadratic coefficient, aij is the ith
interaction coefficient, xi is the independent variable, k is number
of factors and ε is the associated error. In addition, coefficients
of the model are predicted through regression. Details of the

V. Hosseinpour et al. / Applied Catalysis A: General 394 (2011) 166–175
169
Table 1
Levels for process variables in actual and coded values.
Independent variable
Unit
Level
−˛ (−1.63)
−1
0
+1
191
+˛ (+1.63)
A: Temperature
B: Pressure
C: Iridium
D: Ruthenium
E: Methyl iodide
F: Methyl acetate
G: Water
◦C
bar
ppm
ppm
wt.%
wt.%
wt.%
175
20
500
0
4
12
6
179
23
185
29
1250
1300
11
195
38
2000
2600
18
35
788.91
500.78
6.7
17.39
7.93
1711.01
2099.22
15.3
34.61
14. 07
26
11
40
16
of variation around the mean determined by experiments and
adjusted for the number of terms in the model through Eq. (5), i.e.,
Y2(CH4mole %) = 5.9 + 1.29A − 3.14B − 0.92C + 1.17D − 0.61E
+ 1.05F + 1.69G − 1.42AB + 1.13AC + 0.98AF
SSresidual
R2 = 1 −
(4)
(5)
− 0.95BE + 0.56BF − 1.57BG + 0.89CD + 1.30CF
− 0.63CG − 0.61DE − 1.49DG − 0.83EF + 1.22EG
− 0.73FG + 1.22A2 + 1.76B2 + 0.70C2
SSmodel + SSresidual
SSresidual/DFresidual
RA2dj = 1 −
(SSmodel + SSresidual)/(DFmodel + DFresidual
)
(9)
Here, SS is the sum of squares and DF is the degrees of freedom.
The statistical importance of the model was checked with adequate
precision through Eqs. (6) and (7), which basically determine the
signal-to-noise ratio; i.e.,
The statistical importance of the generated model evaluated by
the Fisher test (i.e., F-test) was quantified by dividing the Model
Mean Square by its Residual Mean Square for analysis of variance
(ANOVA). The results of ANOVA for the reaction rate and CH4 mol%
are presented in Table 3.
max(yˆ) − min(yˆ)
Adequate precision =
(6)
ꢁ
¯
V(yˆ)
The ANOVA results displayed in Table 3 confirmed that these
models might be applied to the considered design space. The F-
value for Y1 and Y2 were 21.05 and 28.81, respectively. For these
two responses, there was only a 0.01% chance that a “Model F-
Value” having the same great magnitude could occur due to noise.
A very low probability value for both models (p-value < 0.00001)
implied that they were significant for the 95% confidence inter-
val (i.e., p-value less than 0.05 indicate significance). Insignificant
terms in the model had p-values greater than 0.100, which could
be dropped out manually from the model to enhance the regres-
sion quality and optimisation results. For the rate response, Eq.
(8), the AB, AD, BC, BF, BG, CF, CE, DF, DG, EF, FG, A2, C2, D2, E2, F2
and E terms were determined to be trivial in the model regard-
ing the design space and were removed from it. It is noteworthy
that, although the E term had a p-value greater than 0.100, it was
kept in the respective correlation since some interaction effects
involving E had significant co-effects on the rate response. Simi-
larly, for the CH4 mol% response, Eq. (9), AD, AE, AG, BC, BD, CE, DF,
D2, E2, F2 and G2 were determined to be insignificant and were
eliminated.
n
ꢀ
1
n
Pꢀ2
¯
V(yˆ) =
V(yˆ) =
(7)
n
i=1
Here, yˆ is the predicted response, P is the number of model
parameters, ꢀ2 is the residual mean square and n is the number
of experiments. This compares the range of the predicted values
at the design points to the average prediction error. Ratios greater
than 4 indicate adequate model discrimination. Other criteria to
check the statistical significance are the F- and P-values.
3. Results and discussions
To evaluate the effect of process variables on the reaction rate
and methane formation during the carbonylation reaction, experi-
ments were performed on the basis of the design matrix of central
composite design with six replicated points. Design points and
experimental results of reaction rate mol/l h and CH4 mol% are
shown in Table 2. To minimise the effects of uncontrolled factors,
experiments were performed in random sequence.
The comparison between residual error and pure error was
made through the “lack of fit test”. As may be seen from Table 3
for both responses Y1 and Y2, the p-value for the lack of fit test was
greater than 0.05; thus, as for the former, the latter was insignif-
icant. Adequate precision for Y1 and Y2 were 17.585 and 22.966,
respectively, which denoted that the values were greater than 4.
These statistical criterions indicated that these were significant for
describing the considered process. Goodness of fit of the predicated
values for response by the model was measured by the predicted
R2. For an adequate model, the predicted R2 was within 0.2 of the
adjusted R2. Table 3 reveals that both responses passed this crite-
rion successfully.
Predicted versus actual plots for Y1 and Y2 are shown in Fig. 3.
As may be seen from the figure; the values predicted by the
model (from Eqs. (8) and (9)) and the results obtained from the
experiments were distributed evenly around a 45◦ line. The coef-
ficients of determination for both responses were close to unity
(i.e., for Y1 was 0.93 and for Y2 was 0.96). This provided further
support of a good correlation between the actual and predicted
values.
3.1. Data analysis
After the evaluation of experimental results, the quadratic func-
tion for the reaction rate and CH4 mol% was obtained by utilising
Design-Expert® 8.0.1 software. To estimate the coefficients of the
polynomial, a least square fit procedure was applied, and then
based upon the fitted surface response, analysis was performed.
Analysis of the fitted surface nearly corresponded to that of the
actual system if the fitted surface was an adequate approximation
of the true response function [22]. Quadratic equations based upon
the coded values, for reaction rate and CH4 mol% are presented in
Eqs. (8) and (9); respectively, i.e.,
Y1(Rate) = 22.48 + 3.52A + 3.36B + 2.49C + 3.20D − 8.432
× 10−3E + 1.52F − 2.83G − 1.87AC − 1.56AE − 1.96AF
+ 1.5AG + 2.15BD + 1.75BE − 1.04CD − 1.89CG
− 1.03DE − 1.47EG − 3.30B2 − 2.95G2
(8)

170
V. Hosseinpour et al. / Applied Catalysis A: General 394 (2011) 166–175
Table 2
Central composite design experiment matrix and experimental results for this study.
Run
Variables
Results
(A) Temperature
(◦C)
(B) Pressure
(bar)
(C) Iridium
(ppm)
(D) Ruthenium
(ppm)
(E) MeIa
(wt.%)
(F) MeOAcb
(wt.%)
(G) Water
(wt.%)
Rate
(mol/l h)
CH4 %mol
1
2
3
4
5
6
7
8
9
179
185
191
191
185
185
191
185
179
191
179
179
191
179
185
179
191
195
179
185
179
179
191
179
185
179
175
179
191
185
185
191
191
185
185
179
185
179
191
185
185
191
191
185
191
191
185
179
185
185
35
29
35
35
29
20
23
29
23
23
23
35
35
35
29
23
35
29
35
29
23
23
35
23
29
23
29
35
23
29
29
23
23
29
38
35
29
35
35
29
29
23
35
29
23
35
29
23
29
29
788.91
1250
788.91
788.91
1250
1250
788.91
1250
788.91
1711.09
788.91
1711.09
1711.09
1711.09
1250
1711.09
788.91
1250
500.78
1300
500.78
500.78
1300
1300
2099.22
1300
2099.22
500.78
2099.22
500.78
2099.22
2099.22
1300
500.78
500.78
1300
6.7
11
6.7
15.3
4
11
15.3
18
15.3
15.3
6.7
15.3
15.3
6.7
11
6.7
6.7
11
34.61
26
34.61
34.61
26
26
17.39
26
34.61
17.39
34.61
17.39
17.39
17.39
26
17.39
17.39
26
14.07
16
7.93
14.07
11
11
7.93
11
14.07
7.93
7.93
14.07
7.93
14.07
11
14.07
14.07
11
3.05
8.79
12.67
16.82
20.52
5.70
14.81
23.73
7.63
18.22
13.10
6.84
7.83
8.54
6.84
5.02
6.20
14.34
9.37
5.93
10.59
7.63
9.14
5.39
5.51
5.84
6.04
8.52
3.44
11.17
1.97
7.49
3.55
13.67
10.59
13.49
6.27
6.80
5.48
2.66
23.63
3.97
2.94
17.13
19.53
6.45
5.16
4.35
7.11
3.04
6.44
6.90
4.52
19.23
5.57
4.16
16.47
12.13
7.45
8.68
5.73
7.34
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
31.96
14.51
24.07
7.80
18.19
27.00
12.03
30.67
12.77
5.43
20.50
12.09
23.80
20.65
12.05
13.59
12.13
20.98
19.53
18.58
16.66
11.83
20.76
26.58
24.78
32.01
29.42
20.89
28.69
8.15
788.91
1250
500.78
2600
15.3
11
17.39
26
7.93
11
788.91
788.91
1711.09
1711.09
1250
1711.09
1250
1711.09
788.91
1250
500.78
2099.22
2099.22
2099.22
1300
500.78
1300
500.78
500.78
1300
15.3
6.7
15.3
6.7
11
6.7
11
6.7
15.3
11
34.61
17.39
34.61
34.61
26
34.61
26
17.39
17.39
12
7.93
14.07
14.07
14.07
11
7.93
11
7.93
14.07
11
1250
1300
11
26
6
788.91
1711.09
500
1250
788.91
1250
1711.09
788.91
1250
500.78
2099.22
1300
1300
2099.22
1300
500.78
2099.22
1300
6.7
15.3
11
11
15.3
11
15.3
6.7
11
34.61
34.61
26
26
34.61
26
34.61
34.61
26
14.07
7.93
11
11
7.93
11
7.93
14.07
11
1250
1300
11
40
11
1711.09
788.91
1250
1711.09
1711.09
2000
1711.09
1250
1250
2099.22
2099.22
0
2099.22
500.78
1300
2099.22
1300
1300
15.3
6.7
11
6.7
6.7
11
15.3
11
11
17.39
17.39
26
17.39
34.61
26
17.39
26
26
14.07
7.93
11
7.93
14.07
11
7.93
11
11
29.23
13.73
23.13
15.70
29.83
14.82
21.67
24.76
a
MeI: methyl iodide.
MeOAc: methyl acetate.
b
Table 3
Analysis of variance (ANOVA) for two responses.
Source
Sum of squares
DF
Mean square
F value
p-Value probability > F
For Y1
Model
Residual
Lack of fit
Pure error
2685.25
201.42
187.76
13.65
19
30
25
5
141.33
6.71
7.51
21.05
2.75
<0.0001
0.1314
2.73
Correlation total
2886.67
49
R
2 = 0.9302, Adj-R2 = 0.8860, Pre-R2 = 0.8308, Adequate precision = 17.585
For Y2
Model
Residual
Lack of fit
Pure error
1055.70
38.17
36.11
24
25
20
5
43.99
1.53
1.81
0.41
28.81
4.38
<0.0001
0.0543
2.06
Correlation total
1093.87
49
R
2 = 0.9651, Adj-R2 = 0.9316, Pre-R2 = 0.8290, Adequate precision = 22.966

V. Hosseinpour et al. / Applied Catalysis A: General 394 (2011) 166–175
171
Fig. 3. Predicted versus actual plot of (a) reaction rate and (b) CH4 mol%.
are varied while other five remain constant at the centre point in
the design space. Fig. 5a shows the joint effect of temperature and
iridium concentration on the carbonylation rate. It was observed
that increased temperature at high iridium concentartion had no
substantial effect, while temperature enhancement at low iridium
content had a more pronounced effect. The interaction between
methyl iodide and temperature shown in Fig. 5b indicated that an
increase in methyl iodide concentration had no significant effect
on the reaction rate; however, temperature had a more consid-
erable effect at a low methyl iodide concentration. One possible
reason for less dependence of the reaction rate on methyl iodide
at high concentrations might be the formation of inactive cataly-
sis species ([Ir(CO)2I4]−). The high reaction rate region was located
in the high temperature zone and with low methyl iodide concen-
tration. Both methyl acetate concentration and temperature had a
positive effect on the carbonylation rate, but the temperature effect
was greater. At a lower methyl acetate concentration, an increase
in temperature had a more considerable effect relative to a higher
concentration of this material (see Fig. 5c). The interaction between
temperature and water concentration is shown in Fig. 5d. This plot
implies that the reaction rate had a maximum in the centre of the
water range and at a high temperature level. The possible reason
for the reaction rate passing through a maximum might be the fact
that changing a reaction mechanism from an anionic catalytic cycle
(i.e., the reaction rate increases with a decrease in water concen-
tration) to a neutral catalytic cycle (i.e., the reaction rate decreases
as the water concentration is reduced) when the water concentra-
tion was decreased. The temperature effect on the reaction rate at
a high water concentration was greater than its effect at low water
content.
Fig. 6a shows the interaction between pressure and ruthe-
nium on the carbonylation rate. It was observed that after the
total pressure reached 29 barg, the reaction rate essentially became
independent of pressure. Due to the insufficient partial pressure of
CO at low total pressure, the addition of ruthenium did not show
any considerable effect on the reaction rate at low total pressure,
while at high pressure, ruthenium addition led to a steep change
in the reaction rate. The effect of pressure at high concentrations
of ruthenium was considerably more pronounced than at lower
ruthenium concentrations. The effect of pressure and methyl iodide
is shown in Fig. 6b. This figure implies that the maximum rate
was located at a high methyl iodide concentration and pressure.
At low CO pressure, an increase in methyl iodide concentration
caused a lowering of the reaction rate, but at high CO pressure, an
increase in methyl iodide concentration led to an increased reac-
tion rate. This was due to the reason that a decrease in CO partial
pressure, inactive form of catalyst ([Ir(CO)2I4]−) formed, hence the
Fig. 4. Perturbation plot for rate response (for A: temperature, B: pressure, C: irid-
ium, D: ruthenium, E: methyl iodide, F: methyl acetate and G: water).
3.2. Effect of parameters on the carbonylation reaction rate
Perturbation plots show the comparison between all factors at
a selected point in the considered design space. The perturbation
plot for the carbonylation rate is shown in Fig. 4. The rate response
was drawn by changing only one factor over its range while the
other factors were held constant. The plot demonstrates the effect
of all factors at a central point in the design space (e.g., temperature
185 ◦C, pressure 29 barg, iridium 1250 ppm, ruthenium 1300 ppm,
methyl iodide 11 wt.%, methyl acetate 26 wt.% and water 11 wt.%).
The comparatively flat line of methyl iodide showed an insignifi-
cant effect of this factor on the reaction rate in the design space.
In other words, the coefficient of methyl iodide was very low in
Eq. (8). All factors, except water concentration, indicated a positive
effect on the reaction rate. On the other hand, through the com-
parison of coefficients in Eq. (8), the most significant parameter
was determined. In this manner, the order of positive influence of
the individual terms on the obtained rate response were temper-
ature, pressure, ruthenium, iridium and methyl acetate. It can be
seen from Eq. (8) and the perturbation plot that pressure and water
concentration had curvature effects. The reaction rate approxi-
mately increased linearly with temperature, iridium, ruthenium
and methyl acetate in the considered design space.
One of the best ways to study interactions between factors is
to consider the three dimonsional plot in which only two factors

172
V. Hosseinpour et al. / Applied Catalysis A: General 394 (2011) 166–175
Fig. 5. Response surface plot showing the interaction effects of the process variables on the carbonylation rate. (a) Temperature and iridium concentration, (b) temperature
and methyl iodide concentration, (c) temperature and methyl acetate concentration and (d) temperature and water content.
reaction rate dropped. Fig. 6c shows the interaction effect of water
and iridium concentration. It was observed that increased irid-
ium at a low water concentration had a considerable effect on the
reaction rate; however, as a result of dominant negative effect of
higher water concentration, no considerable change was observed
in the predicted rate response while iridium concentration was
enhanced. The reason was that at high water concentrations, a
steady state concentration of HI increased which meant that the
effect of HI on the reaction rate was negative as well, i.e., it reacted
with ([Ir(CO)2I2]−) and generated the inactive type of iridium com-
plex ([Ir(CO)2I4]−) (see also Fig. 1). Furthermore, this species might
act as a source of ionic iodide which could prevent the migratory
insertion step in the catalytic cycle of carbonylation. The maximum
response rate was placed at 9.5 wt.% of water and a high iridium
concentration.
mation while pressure increased. In addition to the carbonylation
reaction, iridium catalysed the water–gas shift reaction [27]. Pro-
motion of the water–gas shift reaction resulted in the production
of more hydrogen, enhancing methane formation. Furthermore,
methyl iodide had the least effect on methane formation (see Eq.
(9)). An increase in water concentration enhanced the steady state
concentration of HI which had a significant effect on water–gas
shift and methane formation reactions [15]. Methane formation
increased with increased methyl acetate and ruthenium concentra-
tions, but an enhancement in methyl iodide concentration reduced
the amount of methane formed. Similar results have been reported
elsewhere [15,28,29]. To study the interaction between factors,
three dimensional plots were generated. Fig. 8a shows the pressure
and temperature effect on methane formation. The effect of pres-
sure on methane formation was more considerable than the effect
of temperature; minimum methane formation was associated with
mid-range temperature and high pressure. The change in methane
formation due to a temperature change was greater at low pres-
sure than at a higher one. Fig. 8b shows the interaction effect of
pressure and water on methane formation. The plot shows that to
minimise methane formation, high pressure and low water con-
centration are required, although the pressure effect is permanent.
High CO pressure reduces the active site on the metal, and then
prevents hydrogenation of the iridium complex. High water con-
centration increases the steady state concentration of HI, which is a
key species in the water–gas shift reaction. The interaction effects of
ruthenium and iridium on methane formation are shown in Fig. 8c.
Ruthenium increased methane formation in the whole range of
iridium concentration, but its effect at high iridium concentration
was considerable. Therefore, to reduce methane formation in the
carbonylation reaction, low iridium and ruthenium concentrations
3.3. The effect of parameters on methane formation
Fig. 7 shows the perturbation plot for methane formation. The
steep curvature in pressure behaviour demonstrated the response
of CH4 mol% was very rapid to this factor (see Eq. (9)). Methane for-
mation during the carbonylation reaction increased with all factors
except for an increase in pressure and methyl iodide (Fig. 7 and
Eq. (9)). Temperature, pressure (linear and quadratic) and water
(linear) were the most significant factors in the response (Eq. (9)).
As discussed earlier, the reaction of [CH3Ir(CO)2I3]− with hydro-
gen produced by the in situ water-gas shift reaction stood for
methane formation. For activated and coordinated hydrogen, the
active site on the metal was essential [26], but due to the presence
of CO, these active sites were occupied and methane formation was
held back. In other words, CO supported a decrease in methane for-

V. Hosseinpour et al. / Applied Catalysis A: General 394 (2011) 166–175
173
Fig. 6. Response surface plot showing the interaction effects of the process variables on the carbonylation rate. (a) Pressure and ruthenium concentration, (b) pressure and
methyl iodide concentration and (c) iridium concentration and water content.
are suggested. Fig. 8d shows the interaction effect of iridium and
water concentrations on methane formation. It was shown that an
enhancement in the concentration of both iridium and water led
to an increase in methane formation; however, the water concen-
tration effect was more pronounced than that of iridium. On the
other hand, at low iridium concentrations, water affected methane
formation more than it did at higher iridium concentrations.
3.4. Optimisation and validation
To determine the optimum conditions for the carbonylation pro-
cess, the optimisation tool of Design-Expert® 8.0.1 was utilised.
The strategy of the programme is to optimise multiple responses
for the maximisation of the desirability function ranging between
zero and one. All individual desirability functions became one, and
then the programme searched for the conditions in which desirabil-
ity reached a maximum level. To achieve the optimum conditions,
all factors were selected ‘within the range’, while the reaction rate
and methane formation response were defined as a maximum and
minimum, respectively. Ultimately, the optimum process variables
and related responses were obtained and are presented in Table 4.
These were selected from several optimum conditions suggested
Fig. 7. Perturbation plot for the methane formation response (A: temperature, B:
pressure, C: iridium, D: ruthenium, E: methyl iodide, F: methyl acetate and G: water).
Table 4
Observed and predicted results of the reaction rate and methane formation at the optimum point predicted by the RSM in this study.
Optimum condition
Rate (mol/l h)
Experiment
CH4 mol%
Temperature
(◦C)
Pressure
(bar)
Iridium
(ppm)
Ruthenium
(ppm)
MeI
(wt.%)
MeOAc
(wt.%)
Water
(wt.%)
Predicted
28.63
Experiment
Predicted
1.9707
191
32.48
939.4
2099.22
14.46
17.55
7.65
27.1
3.063

174
V. Hosseinpour et al. / Applied Catalysis A: General 394 (2011) 166–175
Fig. 8. Response surface plot showing the interaction effects of the process variables on methane formation in the carbonylation reaction. (a) Temperature and pressure, (b)
pressure and water, (c) iridium and ruthenium and (d) water and iridium.
by the programme. The selected optimum conditions with 0.95 for
the desirability values were predicted to be 28.63 mol/l h for the
reaction rate and 1.97 mol% for methane formation. Under these
optimum conditions, low amounts of water and more expensive
iridium were considered.
To confirm the quadratic models developed by CCD, a verifica-
tion experiment was also performed under the predicted optimum
conditions. The results shown in Table 4 established a good agree-
ment between the predicted and the experimental data.
pressure and methyl iodide content would decrease methane for-
mation while an increase in all the other factors would promote
methane formation. Water, iridium concentration and pressure
had square effects on methane formation. Pressure, water con-
centration, temperature, ruthenium, methyl acetate, iridium and
methyl iodide concentrations were sequentially the most impor-
tant factors affecting methane formation. Ultimately, the optimum
conditions led to a maximum rate and minimum methane forma-
tion simultaneously, and were found to be: temperature of 191 ◦C,
pressure of 32.48 barg, iridium content of 939.4 ppm, ruthenium
concentration of 2099.22 ppm, methyl iodide content of 14.46 wt.%,
methyl acetate concentration of 17.55 wt.% and water content of
7.6 wt.%. This investigation helps considering complicated optimi-
sations involving the interaction of several parameters in an orderly
theoretical fashion which might be applied to catalytic reactions.
4. Conclusion
This study showed that the response surface methodology can
be successfully applied to optimise the carbonylation process.
To investigate the effects of parameters like iridium, ruthenium,
methyl iodide, methyl acetate and water concentration, as well as
temperature and pressure on the carbonylation rate and methane
formation, the central composite design was utilised. It was
shown that an increase in all factors except water and methyl
iodide enhanced the reaction rate. The influence order of fac-
tors on the reaction rate was determined to be: temperature,
pressure, ruthenium, water, iridium and methyl acetate concen-
tration. Furthermore, it was observed that methyl iodide had no
significant effect on the reaction rate. In addition, in the reac-
tion rate correlation, the square effects of water content and
pressure were significant. On the other hand, the correlation devel-
oped for methane formation suggested only an enhancement in
Acknowledgement
Authors are very thankful to the Iranian National Petrochemical
Company’s Research and Technology Division (NPC-RT) for sup-
porting this research under the grant number of 870138705.
References
[1] Consulting SRI, Process Economic Program, Methanol and Derivatives, vol. 3,
2009.
[2] N. Yoneda, S. Kusano, M. Yasui, P. Pujado, S. Wilcher, Appl. Catal. A: Gen. 221
(2001) 253–265.

V. Hosseinpour et al. / Applied Catalysis A: General 394 (2011) 166–175
175
[3] N. Kutenpov, W. Himmele, H. Hohenshutz, Hydrocarbon Proc. 45 (1966) 141.
[17] A.J. Miller, S.J. Smith, US. Pat. Pub. 2009/0177006 A1 (2009).
[18] A.J. Miller, M.J. Payne, Eur. Pat. Pub. 1979302 B1 (2009).
[19] A.J. Miller, M.J. Payne, US. Pat. Pub. 2008/0319223 A1 (2008).
[20] R.L. Mason, R.F. Gunst, J.L. Hess, Statistical Design and Analysis of Experiments
with Applications to Engineering and Science, 2nd ed., John Wiley and Sons,
USA, 2003.
[4] D.J. Schreck, D.C. Busby, R.W. Wegman, J. Mol. Catal. 47 (1988) 117–121.
[5] W. Chua, Y. Ooka, Y. Kamiya, T. Okuhara, Catal. Lett. 101 (2005) 225–228.
[6] D. Linke, D. Wolf, M. Baerns, U. Dingerdissen, S. Zeyß, Stud. Surf. Sci. Catal. 136
(2001) 477–482.
[7] M. Zerella, A.T. Bell, J. Mol. Catal. A: Chem. 259 (2006) 296–301.
[8] W. Huang, K.-C. Xie, J.-P. Wang, Z.-H. Gao, L.-H. Yin, Q.-M. Zhu, J. Catal. 201
(2001) 100–104.
[21] Z.R. Lazic, Design of Experiments in Chemical Engineering, WILEY-VCHVerlag
GmbH and Co. KGaA, Germany, 2004.
[9] H. Jiang, Z. Liu, P. Pan, G. Yuan, J. Mol. Catal. A: Chem. 148 (1999) 215–225.
[10] G.J. Sunley, D.J. Watson, Catal. Today 58 (2000) 293–307.
[11] C.M. Thomas, G. Suss-Fink, Coord. Chem. Rev. 243 (2003) 125–142.
[12] J.H. Jones, Platinum Met. Rev. 44 (2000) 94–105.
[22] D.C. Montgomery, Design Analysis of Experiments, 4th ed., John Wiley and Sons,
USA, 1996.
[23] M. Amid, C.P. Tan, H. Mirhosseini, N. Ab Aziz, T.C. Ling, Food Chem. 124 (2011)
666–671.
[13] D. Forster, T.C. Singleton, J. Mol. Catal. 17 (1982) 299–314.
[14] A. Haynes, P.M. Maitlis, G.E. Morris, G.J. Sunley, H. Adams, P.W. Badger, C.M.
Bowers, D.B. Cook, P.I.P. Elliott, T. Ghaffar, H. Green, T.R. Griffin, M. Payne,
J.M. Pearson, M.J. Taylor, P.W. Vickers, R.J. Watt, J. Am. Chem. Soc. 126 (2004)
2847–2861.
[24] M.A. Olutoye, B.H. Hameed, Appl. Catal. A: Gen. 371 (2009) 191–198.
[25] A. Ruggieroa, L. Tricaricoa, A.G. Olabib, K.Y. Benyounis, Opt. Laser Technol. 43
(2011) 82–90.
[26] T. Ghaffar, J.P.H. Charmant, G.J. Sunley, G.E. Morris, A. Haynes, P.M. Maitlis,
Inorg. Chem. Commun. 3 (2000) 11–12.
[15] D. Forster, J. Chem. Soc. Dalton Trans. (1979) 1639–1645.
[16] M.J. Baker, M.F. Giles, C.S. Garland, G. Rafeletos, Eur. Pat. Pub. 0749948 A1
(1996).
[27] T.C.Singleton, US. Pat. Pub 4151107 (1979).
[28] B.M. James, G.M. Francis, M.M. James, Eur. Pat. Pub. 0752406 A1 (1997).
[29] E.A. Ditzel, J.G. Sunely, R.J. Watt, US. Pat. Pub. 5877348 (1997).