Optimisation of alkene epoxidation catalysed by polymer supported Mo(VI) complexes and application of artificial neural network for the prediction of catalytic performances

Article abstract of DOI:10.1016/j.apcata.2013.06.055

A greener and efficient alkene epoxidation process using heterogeneous molybdenum (Mo) based catalysts and tert-butyl hydroperoxide (TBHP) as an oxidant has been developed. A polybenzimidazole supported Mo(VI) complex, i.e. PBI.Mo and polystyrene 2-(aminomethyl) pyridine supported Mo(VI) complex, i.e. Ps.AMP.Mo catalysts have been successfully prepared and characterised. The catalytic activities of the polymer supported Mo(VI) catalysts have been tested for epoxidation of 1-hexene and 4-vinyl-1-cyclohexene in a jacketed stirred batch reactor. Batch experiments have been conducted to study the effect of different types of catalysts, catalyst loading, feed mole ratio (FMR) of alkene to TBHP and reaction temperature on the yield of epoxide for both alkenes, i.e. 1-hexene and 4-vinyl-1-cyclohexene. The long-term stability of PBI.Mo and Ps.AMP.Mo catalysts has been evaluated by recycling the catalyst several times for batch experiments using conditions that will form the basis of a continuous epoxidation process. The extent of Mo leaching from each polymer supported catalyst has been investigated by isolating any residue from reaction supernatant solutions after the removal of the heterogeneous catalyst and using the residue as potential catalyst for epoxidation. An artificial neural network (ANN) model has been employed to predict the catalytic performance of PBI.Mo and Ps.AMP.Mo catalysts for all batch experimental results. The ANN predicted values are in good agreement with the batch experimental results. The results obtained from batch experiments and ANN modelling provided useful information for conducting continuous epoxidation experiments in multi-functional reactors such as FlowSyn and reactive distillation column (RDC).

Full text of DOI:10.1016/j.apcata.2013.06.055

Applied Catalysis A: General 466 (2013) 142–152
Contents lists available at SciVerse ScienceDirect
Applied Catalysis A: General
journal homepage: www.elsevier.com/locate/apcata
Optimisation of alkene epoxidation catalysed by polymer supported
Mo(VI) complexes and application of artificial neural network for the
prediction of catalytic performances
Misbahu Ladan Mohammeda, Dipesh Patela, Rene Mbelecka, Debdarsan Niyogib,
David C. Sherringtonc, Basudeb Sahaa,∗
a Centre for Green Process Engineering, Department of Applied Sciences, Faculty of Engineering, Science and The Built Environment, London South Bank
University, 103 Borough Road, London SE1 0AA, UK
b Global Consultancy Practice, Tata Consultancy Services Ltd., UK
c WestCHEM, Department of Pure and Applied Chemistry, University of Strathclyde, Glasgow G1 1XL, UK
a r t i c l e i n f o
a b s t r a c t
Article history:
A greener and efficient alkene epoxidation process using heterogeneous molybdenum (Mo) based cat-
alysts and tert-butyl hydroperoxide (TBHP) as an oxidant has been developed. A polybenzimidazole
supported Mo(VI) complex, i.e. PBI.Mo and polystyrene 2-(aminomethyl) pyridine supported Mo(VI) com-
plex, i.e. Ps.AMP.Mo catalysts have been successfully prepared and characterised. The catalytic activities
of the polymer supported Mo(VI) catalysts have been tested for epoxidation of 1-hexene and 4-vinyl-
1-cyclohexene in a jacketed stirred batch reactor. Batch experiments have been conducted to study the
effect of different types of catalysts, catalyst loading, feed mole ratio (FMR) of alkene to TBHP and reac-
tion temperature on the yield of epoxide for both alkenes, i.e. 1-hexene and 4-vinyl-1-cyclohexene. The
long-term stability of PBI.Mo and Ps.AMP.Mo catalysts has been evaluated by recycling the catalyst sev-
eral times for batch experiments using conditions that will form the basis of a continuous epoxidation
process. The extent of Mo leaching from each polymer supported catalyst has been investigated by iso-
lating any residue from reaction supernatant solutions after the removal of the heterogeneous catalyst
and using the residue as potential catalyst for epoxidation. An artificial neural network (ANN) model
has been employed to predict the catalytic performance of PBI.Mo and Ps.AMP.Mo catalysts for all batch
experimental results. The ANN predicted values are in good agreement with the batch experimental
results. The results obtained from batch experiments and ANN modelling provided useful information
for conducting continuous epoxidation experiments in multi-functional reactors such as FlowSyn and
reactive distillation column (RDC).
Received 7 March 2013
Received in revised form 12 June 2013
Accepted 29 June 2013
Available online 9 July 2013
Keywords:
Alkene epoxidation
Artificial neural network (ANN)
tert-Butyl hydroperoxide (TBHP)
Heterogeneous catalysis
Polymer supported Mo(VI) catalyst
1-Hexene
4-Vinyl-1-cyclohexene
© 2013 Elsevier B.V. All rights reserved.
1. Introduction
decades, soluble compounds of transition metals have been used to
catalyse alkene epoxidation with good activity and product selec-
Alkene epoxidation is an important chemical process as the
resultant epoxide is used as raw materials or intermediates in
the production of commercially important products for flavours,
fragrances, paints and pharmaceuticals [1–4]. The batch produc-
tion of epoxides within industries often uses the stoichiometric
peracids including peracetic acid and m-chloroperbenzoic acid [5].
However, the employment of peracid for alkene epoxidation is
not an environmentally benign method as equivalent amount of
acid waste is produced. In addition, there are safety issues asso-
ciated with handling and storage of peracid. Over the last few
tivity [6–10]. However, homogenous catalysed systems suffer from
numerous industrial problems including corrosion, deposition of
the catalyst on the reactor wall and high chance of product con-
tamination due to difficulties in separation of the catalyst from the
reaction mixture [11,12].
There has been a major concern for environmentally safer
chemical processes in recent years. Consequently, more and more
attention has been focused on developing greener and efficient
epoxidation process employing a heterogeneous catalyst and a
benign oxidant. A number of attempts have been made to develop
suitable heterogeneous catalysts for epoxidation by immobilisa-
tion of catalytically active metal species on organic materials,
such as polymers [13,14] and ion-exchange resins [15,16] or inor-
ganic materials including alumina [17,18], zeolite [19,20] and silica
∗
Corresponding author. Tel.: +44 (0)20 7815 7190; fax: +44 (0)20 7815 7699.
E-mail address: b.saha@lsbu.ac.uk (B. Saha).
0926-860X/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.apcata.2013.06.055

M.L. Mohammed et al. / Applied Catalysis A: General 466 (2013) 142–152
143
optimum reaction conditions obtained from batch experiments
would form the basis for continuous epoxidation in a continu-
ous flow reactor, such as FlowSyn and multi-functional reactor
including reactive distillation column (RDC). FlowSyn and RDC
offer substantial benefits including increased selectivity, scalability
and reproducibility and therefore have enormous potential as pro-
cess alternatives for carrying out liquid phase chemical reactions
[31–33].
Nomenclature
AAS
atomic absorption spectrophotometric
AMP
ANN
ASAP
BET
2-(aminomethyl) pyridine
artificial neural network
accelerated surface area and porosimetry
Brunauer–Emmett–Teller
divinylbenzene
DVB
ECTM-5 Econo-CapTM-5
2. Experimental methods
FID
FMR
GC
MSE
Mo
flame ionisation detector
feed mole ratio
gas chromatography
mean square error
molybdenum
2.1. Materials
1,2-Dichloroethane (anhydrous, 99.8%), 1-hexene (≥99%), 2-
(aminomethyl) pyridine (AMP) (99%), 2-ethylhexanol (≥99.6%),
4-vinyl-1-cyclohexene (97%), acetone (>95%), divinylbenzene
(DVB) (80%), ethanol (≥99.5%), iso-octane (anhydrous, 99.8%),
molybdenyl acetylacetonate (MoO2(acac)2) (99%), styrene (≥99%),
TBHP solution in water (70%, w/w), toluene (anhydrous, 99.8%)
and vinylbenzyl chloride (VBC) (97%) were purchased from
Sigma–Aldrich Co. Ltd., and microporous polybenzimidazole (PBI)
resin beads were supplied by Celanese Corporation. The purity of all
chemicals was verified by gas chromatography (GC). TBHP was ren-
dered anhydrous by Dean-Stark distillation from a toluene solution
using the procedure reported by Sharpless and Verhoeven [34]. All
other chemicals were used without further purification. The con-
centration of TBHP in toluene was found to be 3.57 mol L−1 and was
determined by iodometric titration [34].
MoO2(acac)2 molybdenyl acetylacetonate
NN
PBI
neural network
polybenzimidazole
PBI.Mo polybenzimidazole supported Mo(VI) complex
Ps.AMP polystyrene 2-(aminomethyl) pyridine
Ps.AMP.Mo polystyrene 2-(aminomethyl) pyridine sup-
ported Mo(VI) complex
RDC
TBHP
t
reactive distillation column
tert-butyl hydroperoxide
time [min]
VBC
vinylbenzyl chloride
[21,22]. Polymer supported molybdenum complexes have been
shown to be effective for alkene epoxidation with alkyl hydroper-
oxide as oxygen source [23–27].
2.2. Preparation and characterisation of polymer supported
Mo(VI) catalysts
Due to the complex nature of catalytic reactions, it is challeng-
ing to adopt the conventional method of modelling the kinetic data
that involves assuming an empirical mathematical equation and
finding the unknown parameters by non-linear regression analy-
sis. Neural network (NN) models are very useful in such cases as
they are nonlinear systems and have the ability to predict complex
multi-dimensional problems that are found in catalysis. NN models
such as artificial neural network (ANN) have the capacity to estab-
lish quantitative relationships between the input and output data
without prior knowledge of the correlation between the variables
involved in the system [28,29].
In this work, a novel process in which an efficient and selective
polybenzimidazole supported molybdenum complex (PBI.Mo) and
a polystyrene 2-(aminomethyl) pyridine supported molybdenum
complex (Ps.AMP.Mo) have been used as catalysts for epoxidation
of 1-hexene and 4-vinyl-1-cyclohexene. This process is solvent-free
and uses environmentally benign tert-butyl hydroperoxide (TBHP)
as an oxidant. On the other hand, tert-butanol is also formed as
a co-product during epoxidation. Hence, this process is atom effi-
cient. Oku et al. [30] reported that tert-butanol can be efficiently
recycled in the Sumitomo process through hydrogenolysis and oxi-
dation.
An extensive assessment of PBI.Mo and Ps.AMP.Mo catalysts
has been carried out in a batch reactor to evaluate their stabil-
ity, catalytic activity and reusability. The suitability and efficiency
of both catalysts have been compared by studying the effect of
catalyst loading, feed molar ratio (FMR) of alkene to TBHP and
reaction temperature on the yield of epoxides for optimisation of
the reaction conditions in a batch reactor. A detailed evaluation
of molybdenum (Mo) leaching from the polymer supported cata-
lyst has been conducted by assessing the catalytic activity in the
residue found in the supernatant solutions of the reaction mixture
after the removal of heterogeneous catalyst. All the batch experi-
mental results have been modelled using ANN to predict the trend
in the rate of epoxidation under different reaction conditions. The
PBI has a high degree of thermal stability and chemical resis-
tivity. As a result of these properties, PBI resin is employed in
a wide variety of applications such as ion-exchange, separations,
purifications and support for polymer supported catalysts [35]. On
the other hand, crosslinked polystyrene 2-(aminomethyl) pyridine
(Ps.AMP) was prepared by suspension polymerisation of known
mixtures of divinylbenzene (DVB), vinylbenzyl chloride (VBC) and
styrene in the presence of 2-ethylhexanol as a porogen. This was
followed by amination of the formed resin with an excess of 2-
(aminomethyl) pyridine in ethanol. The crosslinked resin beads
are commonly used as catalyst support due to their high poros-
ity, large surface area and robust spherical particles with uniform
size distribution [36]. PBI and Ps.AMP resins were each loaded
with Mo by reaction with an excess of MoO2(acac)2, relative to
polymer bound ligand under reflux in toluene for four days. The
excess MoO2(acac)2 was removed by exhaustive extraction with
acetone in a Soxhlet. Fresh acetone was replaced at regular time
intervals until the refluxing solvent remained colourless. Molyb-
denum in the prepared catalysts was found to be homogeneously
distributed in the polymer. The particle size distribution of PBI.Mo
and Ps.AMP.Mo catalysts was carried out using a Malvern Mas-
tersizer. BET surface area was determined by nitrogen adsorption
and desorption method using Micromeritics ASAP (accelerated
surface area and porosimetry) 2010. Molybdenum content of the
supported catalysts was assayed using atomic absorption spectro-
photometric (AAS) method. The properties of the prepared PBI.Mo
and Ps.AMP.Mo catalysts are given in Table 1.
2.3. Batch epoxidation studies
Fig. 1 shows reaction scheme for alkene epoxidation. Epoxi-
dation of 1-hexene and 4-vinyl-1-cyclohexene with TBHP as an
oxidant in the presence of polymer supported catalyst was car-
ried out in a jacketed four necked glass reactor of 0.25 L capacity.

144
M.L. Mohammed et al. / Applied Catalysis A: General 466 (2013) 142–152
and the flow rate of helium was maintained at 1 mL min−1. Both
injector and detector temperatures were maintained at 523 K. A
split ratio of 100:1 and injection volume of 0.5 ␮L were selected
as part of the GC method. A ramp method was used to separate
all the compounds present in the sample mixture. In the ramp
method, the initial temperature of the oven was set at 313 K and
the sample was injected by an auto injector. The oven tempera-
ture was programmed to hold at 313 K for 4 min after the sample
injection. The oven temperature was ramped from 313 K to 498 K
at the rate of 20 ◦C min−1. The total run time for each sample
was ∼13 min. After each sample run, the oven temperature was
cooled back to 313 K so that the subsequent sample runs could be
started.
Table 1
Physical and chemical properties of polymer supported Mo(VI) catalysts.
Catalyst properties
Ps.AMP.Mo
PBI.Mo
Colour
Grey
53.5
0.74
Blue
24.8
0.93
2.4
2.58:1
243–335
11.5
BET surface area (m2 g−1
)
Mo loading (mmol Mo g−1 resin)d
Ligand loading (mmol g−1 resin)e
Ligand to Mo ratio
0.5
0.68:1
119–153
6.0
Particle size (␮m)
Average pore diameter (nm)
Total pore volume (cm3 g−1
)
0.08
0.07
d
From AAS analysis of digested resin.
From N% elemental analysis of Mo loaded resins assuming ligand = imidazole or
e
2-(aminomethyl) pyridine as appropriate.
The batch reactor was equipped with agitator, condenser, digital
thermocouple and water bath. Known quantities of alkene and
TBHP were weighed out and introduced into the reactor vessel
(no cosolvent was added). Feed mole ratio (FMR) of alkene to
TBHP of 1:1–10:1 was selected for charging the reactor. Agita-
tion was started at the desired rate, i.e. between 300 and 600 rpm
and heating to the reaction mixture was supplied through water
bath via the reactor jacket and monitored by a digital thermo-
couple. The temperature of the reaction mixture was allowed to
reach the desired value, i.e. 323–353 K and was maintained in
the range of 0.5 K throughout the batch experiment. A known
amount of catalyst (0.15–0.6 mol% Mo) was weighed and added
into the reactor when the reaction mixture achieved a constant
desired temperature. The time at which the catalyst was added
was noted as zero time and a sample was withdrawn, which was
noted as t = 0. The samples were taken at specific time intervals
and were analysed by Shimadzu GC-2014 gas chromatography
(GC). Both PBI.Mo and Ps.AMP.Mo catalysed epoxidations of 1-
hexene and 4-vinyl-1-cyclohexene have found to be selective in
the formation of respective epoxides i.e. 1,2-epoxyhexane and
4-vinyl-1-cyclohexane 1,2-epoxide. The GC analysis of the reac-
tion mixture showed no evidence of either terminal or diepoxides
products. Batch experimental results are discussed in Section 3 in
detail.
3. Results and discussion
Batch experiments were conducted under different reaction
conditions to study the effect of various parameters such as reaction
temperature, feed mole ratio (FMR) of alkene to TBHP and catalyst
loading on the yield of epoxide for epoxidation of 1-hexene and
4-vinyl-1-cyclohexene with TBHP as an oxidant. Reusability and
supernatant studies were carried out to evaluate the long term sta-
bility of each polymer supported catalyst for epoxidation. The batch
experimental results were modelled using artificial neural network
(ANN) and discussed in detail in Section 3.1.
3.1. Artificial neural network (ANN)
Neural models are ideally suited where empirical or pheno-
menological modelling is either difficult or impossible as very
little is known about the physical system studied. ANN has the
ability to represent any arbitrary non-linear system and, because
it is system-agnostic, no insight into the physical process is
required.
The architecture of a neural network (NN) is divided into three
basic layers, i.e. (a) input layer, containing the nodes collecting the
input values, (b) hidden layer, with interconnected nodes and (c)
output layer, having nodes representing output values. The specific
architecture of an ANN implementation (number of input/output
nodes, number of nodes in the hidden layer and number of hidden
layers) depends upon the problem at hand.
The computational algorithm involved in ANN is fundamen-
tally simplistic, despite its surprising potential of making accurate
predictions. A basic programming knowledge is sufficient to imple-
ment an ANN. ANN is never aware of the actual physical system
the training data represent. Therefore, ANN is inherently generic in
nature, i.e. once a solution is developed, it can be used successfully
for any arbitrary physical system.
2.4. Method of analysis
Samples collected from the liquid phase of the reaction mix-
ture were analysed using Shimadzu GC-2014 gas chromatography
(GC). The GC was equipped with flame ionisation detector (FID) and
auto-injector. The GC was fitted with Econo-CapTM-5 (ECTM-5), a
capillary column (purchased from Alltech Associates, Inc., USA) of
length 30 m, internal diameter 320 ␮m and film thickness 0.25 ␮m.
A method was developed and programmed for analysis of all the
species in the sample mixture. Helium was used as a carrier gas
polymer supported
Mo(VI) catalyst
CH3
CH3
OH
CH3
H3C
CH2
H3C
O
OH
H3C
H3C
O
CH3
1-hexene
TBHP
1,2-epoxyhexane
tert-butanol
polymer supported
Mo(VI) catalyst
CH3
O
CH3
O
H3C
OH
H3C
OH
CH3
H2C
H2C
CH3
4-vinyl-1-cyclohexene
TBHP
4-vinyl-1-cyclohexane 1,2-epoxide tert-butanol
Fig. 1. Reaction scheme for epoxidation of 1-hexene and 4-vinyl-1-cyclohexene with TBHP catalysed by polymer supported Mo(VI) catalysts.

M.L. Mohammed et al. / Applied Catalysis A: General 466 (2013) 142–152
145
Table 2
Neural network architecture and parameters.
No. of inputs
4–10
No. of outputs
1
No. of hidden layers
1
No. of nodes in hidden layer
4
Learning parameter
0.40
Initial weight
0.5
Momentum
0.5
Reference
Figs. 2–7
3.2. Implementation steps of ANN
the catalyst was due to slight difference in the measured value of
Mo content obtained from different batches of the prepared cat-
alysts. All batch studies in this work were conducted from one
batch of the prepared catalysts. The effect of catalyst loading on
the yield of epoxide was studied by conducting batch experi-
ments using different catalyst loading. The results are presented
in Fig. 2.
For 1-hexene epoxidation in the presence of PBI.Mo catalyst
(Fig. 2a), catalyst loading of 0.15 mol% Mo has provided sufficient
active site for the reaction. Hence, an increase in catalyst loading
beyond 0.15 mol% Mo (i.e. 0.3–0.6 mol% Mo) has negligible effect
on the rate of formation of 1,2-epoxyhexane. However, in the case
of PBI.Mo catalysed epoxidation of 4-vinyl-1-cyclohexene (Fig. 2c),
the influence of catalyst loading is very clear when the catalyst load-
ing was increased from 0.15 mol% Mo to 0.3 mol% Mo. The yield of
4-vinyl-1-cyclohexane 1,2-epoxide at 0.15 mol% Mo and 0.3 mol%
Mo after 260 min was ∼25% and ∼72%, respectively. However, only
∼10% increase in the yield of 4-vinyl-1-cyclohexane 1,2-epoxide
was found when the catalyst loading was increased from 0.3 mol%
Mo to 0.6 mol% Mo (Fig. 2c).
Epoxidation of 1-hexene and 4-vinyl-1-cyclohexene catalysed
by Ps.AMP.Mo catalyst revealed identical trends in the rate of epox-
idation. Fig. 2b shows that there was a marginal increase in the rate
of formation of epoxides when the catalyst loading was increased
from 0.3 mol% Mo to 0.6 mol% Mo. On the other hand, Fig. 2d
shows that despite the equilibrium was reached within 100 min
for catalyst loading of 0.3 mol% Mo and within 20 min for cata-
lyst loading of 0.6 mol% Mo, reaction rate for the latter catalyst
loading was significantly higher. The yield of 1,2-epoxyhexane and
4-vinyl-1-cyclohexane 1,2-epoxide at 260 min for catalyst loading
of 0.3 mol% Mo and 0.6 mol% Mo was found to be ∼85% (Fig. 2b) and
∼95% (Fig. 2d), respectively. It should be noted that epoxidation of
4-vinyl-1-cyclohexene catalysed by Ps.AMP.Mo catalyst achieved
equilibrium within the first 20 min of the reaction (Fig. 2d). Fig. 2d
illustrates that the experiment conducted at a catalyst loading of
0.15 mol% Mo for 4-vinyl-1-cyclohexene epoxidation catalysed by
Ps.AMP.Mo catalyst resulted in higher yield of epoxide, (i.e. ∼90%)
as compared to other experiments carried out at 0.15 mol% Mo
(Fig. 2a–c). It can be concluded from Fig. 2 that the catalyst load-
ing of 0.6 mol% Mo was found to be the optimum for both reaction
systems.
In this work, ANN model has been employed to predict the effect
of different reaction parameters on the rate of formation of corre-
sponding epoxide. The ANN model has been calibrated with range
of values called training sets. Subsequently, a special fitting proce-
dure, referred to as cross validation has been performed to ensure
that the predicted responses are actually representatives of the
experimental results. In cross validation, part of the known data
have been used to calibrate the parameters of the model, while the
remaining data have been used to confirm the robustness of the
response [37]. The computational parameters used for ANN model
are summarised in Tables 2 and 3.
The ANN model represents very well the batch experimental
results obtained for PBI.Mo and Ps.AMP.Mo catalysed epoxidation
at different reaction conditions (see Figs. 2–7). The application
of ANN to predict the catalytic performance of mesoporous Ti-
silicate catalysts for the optimisation of olefin epoxidation by
taking into account synthetic variables was reported by Corma
et al. [38]. Similarly, ANN model has been successfully applied to
predict the catalytic performance of mono-metallic and bimetal-
lic catalysts for oxidation of ethyl acetate [39] and 2-propanol
[40] in
tions.
a fixed bed reactor under different reaction condi-
3.3. Elimination of mass transfer resistances
In heterogeneous catalysed alkene epoxidation, there exists two
types of mass transfer resistances – one across the solid–liquid
interface, i.e. the influence of external mass transfer resistance and
other in the intraparticle space, i.e. internal mass transfer resis-
tance that is associated with different catalyst particle size and
catalyst internal structure, i.e. chemical structure, pore size dis-
tribution and porosity. The existence of external mass transfer
resistance for alkene epoxidation was studied in a jacketed stirred
batch reactor. It was observed that there was no external mass
transfer resistance for alkene epoxidation when experiments were
carried out using the agitation speed of 300–600 rpm under oth-
erwise identical conditions. On the other hand, the particle size
of both PBI.Mo and Ps.AMP.Mo catalysts is fairly uniform and we
did not notice any change in catalytic performance within this size
range (119–335 ␮m). Hence, internal mass transfer resistance was
absent in this work for both catalysts particles. On the basis of these
investigations, all batch experiments were conducted at an agita-
tion speed of 400 rpm using PBI.Mo and Ps.AMP.Mo catalysts as
prepared.
3.5. Effect of feed mole ratio (FMR) of alkene to TBHP
Epoxidations of 1-hexene/4-vinyl-1-cyclohexene with TBHP are
equilibrium limited reactions. An increase in feed mole ratio (FMR)
of the above mentioned alkenes to TBHP could increase the rate
of formation of the desired epoxides. Hence, a number of batch
experiments were carried out to study the effect of different FMR
of alkene to TBHP on the yield of epoxide.
3.4. Effect of catalyst loading
In this research, catalyst loading is defined as percentage of
mole ratio of Mo to TBHP. The reason for choosing the catalyst
loading based on active Mo component instead of total mass of
In case of Ps.AMP.Mo catalysed epoxidation of 1-hexene and
4-vinyl-1-cyclohexene, an increase in FMR of alkene to TBHP
Table 3
Artificial neural network (ANN) predicted parameters and results.
No. of training cycles
2000
Input presented in random order
Data randomly selected for validation [%]
30%
MSE [original scale]
Training set
0.01–0.45
Validation set
0.01–0.49
Yes

146
M.L. Mohammed et al. / Applied Catalysis A: General 466 (2013) 142–152
100
80
60
40
20
0
100
Catalyst: PBI.Mo
Catalyst: Ps.AMP.Mo
80
60
40
20
0
0
40
80
120 160 200 240 280
0
40
80
120 160 200 240 280
Time [min]
Time [min]
(a)
(b)
100
80
60
40
20
0
100
80
60
40
20
0
Catalyst: PBI.Mo
Catalyst: Ps.AMP.Mo
0
40
80
120 160 200 240 280
0
40
80
120 160 200 240 280
Time [min]
Time [min]
(c)
(d)
• 0.15 mol% Mo ♦ 0.3 mol% Mo
0.6 mol% Mo – ANN predicted
Fig. 2. Effect of catalyst loading (mole ratio of Mo to TBHP) on the yield of epoxide and prediction by artifical neural network (ANN) for epoxidation of 1-hexene catalysed
by (a) PBI.Mo and (b) Ps.AMP.Mo catalysts, and 4-vinyl-1-cyclohexene catalysed by (c) PBI.Mo and (d) Ps.AMP.Mo catalysts at reaction temperature: 343 K; feed mole ratio
(FMR) of alkene to TBHP: 5:1; agitation speed: 400 rpm.
from 2.5:1 to 10:1 gives marked increase in the yields of 1,2-
epoxyhexane (Fig. 3b) and 4-vinyl-1-cyclohexane 1,2-epoxide
(Fig. 3d). Similarly, Fig. 3a and c shows that for PBI.Mo catal-
ysed reactions, an increase in the FMR of alkene to TBHP from
2.5:1 to 5:1, increases the rate of formation of epoxides. Sur-
prisingly, for PBI.Mo catalysed epoxidation of 1-hexene (Fig. 3a)
and 4-vinyl-1-cyclohexene (Fig. 3c), an increase in the FMR of
alkene to TBHP from 5:1 to 10:1 decreases the rate of forma-
tion of corresponding epoxides which is unusual. Ambroziak et al.
[41] reported similar unexpected results for 1-octene epoxidation
catalysed by PBI.Mo catalyst when 1-octene to TBHP ratio was
increased from 5:1 to 10:1. It could be concluded from Fig. 3a and
c that epoxidation of alkenes using PBI.Mo catalyst is dependent
on the alkene concentration. The yield of 1,2-epoxyhexane in the
presence of Ps.AMP.Mo catalyst was ∼100% at 260 min for 10:1
FMR of alkene to TBHP (Fig. 3b), whilst ∼100% yield of 4-vinyl-
1-cyclohexane 1,2-epoxide was achieved within the first 100 min
and at lower FMR of alkene to TBHP (2.5:1) (Fig. 3d). Based on the
results obtained from this study, FMR of 5:1 (alkene to TBHP) at
353 K can be considered as an optimum for PBI.Mo catalysed reac-
tions (Fig. 3a and c), whilst FMR of 10:1 (alkene to TBHP) can be
regarded as the optimum for Ps.AMP.Mo catalysed reactions of
1-hexene (Fig. 3b) and 4-vinyl-1-cyclohexene (Fig. 3d) at 353 K,
respectively.
3.6. Effect of reaction temperature
An increase in reaction temperature increases the rate of reac-
tion. Hence, experiments were conducted to study the effect of
different reaction temperature on the yield of epoxide. Reac-
tion temperature used for this study was 333 K, 343 K and
353 K.
Fig. 4a shows that the yield of 1,2-epoxyhexane using PBI.Mo
catalyst after 260 min was ∼48% and ∼70% at 343 K and 353 K,

M.L. Mohammed et al. / Applied Catalysis A: General 466 (2013) 142–152
147
100
80
60
40
20
0
100
Catalyst: PBI.Mo
80
60
40
20
0
Catalyst: Ps.AMP.Mo
0
40
80
120 160 200 240 280
0
40
80
120 160 200 240 280
Time [min]
Time [min]
(a)
(b)
100
80
60
40
20
0
100
80
60
40
20
0
Catalyst: PBI.Mo
Catalyst: Ps.AMP.Mo
0
40
80
120 160 200 240 280
0
40
80
120 160 200 240 280
Time [min]
(c)
Time [min]
(d)
• FMR of alkene to TBHP = 2.5:1 ♦ FMR of alkene to TBHP = 5:1
FMR of alkene to TBHP = 10:1 – ANN predicted
Fig. 3. Effect of feed mole ratio (FMR) of alkene to TBHP on the yield of epoxide and prediction by artifical neural network (ANN) for epoxidation of 1-hexene catalysed by
(a) PBI.Mo and (b) Ps.AMP.Mo catalysts, and 4-vinyl-1-cyclohexene catalysed by (c) PBI.Mo and (d) Ps.AMP.Mo catalysts at reaction temperature: 343 K; catalyst loading:
0.3 mol% Mo; agitation speed: 400 rpm.
respectively, whilst there was a significant drop in yield of
1,2-epoxyhexane (∼23%) for experiment carried out at 333 K. Sim-
ilar trends for the yield of 1,2-epoxyhexane using Ps.AMP.Mo
catalyst (Fig. 4b) and for the yield of 4-vinyl-1-cyclohexane
1,2-epoxide in the presence of PBI.Mo catalyst (Fig. 4c) were
observed. For Ps.AMP.Mo catalysed epoxidation of 4-vinyl-1-
cyclohexene (Fig. 4d), the reaction reached equilibrium within the
first 5 min for experiments carried out at 353 K. The yield of 4-
vinyl-1-cyclohexane 1,2-epoxide in the presence of Ps.AMP.Mo
catalyst after 5 min at 333 K, 343 K and 353 K was ∼18%, ∼28%
and ∼95%, respectively. On the other hand, the yield of epox-
ide achieved after 260 min was ∼95% for all three temperature
ranges. Hence, it can be concluded that 353 K is the opti-
mum reaction temperature for both PBI.Mo and Ps.AMP.Mo
catalysed epoxidation of 1-hexene and 4-vinyl-1-cyclohexene
(Fig. 4a–d).
3.7. Catalyst reusability studies
The ability to reuse supported transition metal catalyst several
times without significant loss in the activity could be the most
important benefit for commercial application as most of the transi-
tion metal complexesare very expensiveto purchaseand difficultto
prepare. Reusability studies of PBI.Mo and Ps.AMP.Mo catalysts for
epoxidations of 1-hexene and 4-vinyl-1-cyclohexene were inves-
tigated. The results are presented in Fig. 5. In this study, a fresh
catalyst was used for epoxidation plotted as Run 1 in Fig. 5. Once the
reaction experiment for Run 1 was over, the catalyst beads were fil-
tered from the reaction solution, washed with 1,2-dichloroethane,
dried in an oven at 313 K and reused for Run 2. This procedure
was repeated for Run 3–Run 6. Uncatalysed epoxidation experi-
ment was carried out and plotted as Run 7 in Fig. 5 to compare
with the catalysed experimental results.

148
M.L. Mohammed et al. / Applied Catalysis A: General 466 (2013) 142–152
100
80
60
40
20
0
100
Catalyst: PBI.Mo
Catalyst: Ps.AMP.Mo
80
60
40
20
0
0
40
80
120 160 200 240 280
0
40
80
120 160 200 240 280
Time [min]
Time [min]
(a)
(b)
100
80
60
40
20
0
100
80
60
40
20
0
Catalyst: PBI.Mo
Catalyst: Ps.AMP.Mo
0
40
80
120 160 200 240 280
0
40
80
120 160 200 240 280
Time [min]
Time [min]
(c)
(d)
333 K
343 K
♦
353 K – ANN predicted
•
Fig. 4. Effect of reaction temperature on the yield of epoxide and prediction by artifical neural network (ANN) for epoxidation of 1-hexene catalysed by (a) PBI.Mo and (b)
Ps.AMP.Mo catalysts, and 4-vinyl-1-cyclohexene catalysed by (c) PBI.Mo and (d) Ps.AMP.Mo catalysts at catalyst loading: 0.3 mol% Mo; feed mole ratio (FMR) of alkene to
TBHP: 5:1; agitation speed: 400 rpm.
High rate of epoxide formation was observed in Run 1 as
compared to subsequent runs for epoxidation (Fig. 5a–d). This
is due to the availability of sufficient active sites in fresh cat-
alyst that are available for adsorption by the reacting species
as compared to subsequent runs. A significant drop in the cat-
alytic activity and the rate of epoxide formation was observed
catalyst the ligand to Mo ratio is 2.58:1, while it is 0.68:1 for
Ps.AMP.Mo catalyst. It is evident that Ps.AMP.Mo catalyst has an
excess Mo to ligand content as compared to PBI.Mo catalyst. Hence,
Ps.AMP.Mo catalyst continues to lose Mo during catalytic reaction
since even at the outset, a significant proportion of Mo introduced
cannot be coordinated by polymer immobilised ligand. In case of
PBI.Mo catalyst it seems that most of Mo introduced at the out-
set is coordinated by polymer immobilised ligand and Mo species
that become mobile during catalysis are effectively re-captured by
the presence of excess ligand on the support. On the other hand,
Mo species containing microgel are lost from both the polymer
supports as a result of agitation. Hence, it could be concluded
that catalytically active Mo species are lost from each polymer
supported catalyst and this could either be soluble leached com-
plexes or as traces of Mo containing microgel released as a result
of mechanical attrition of the beads or both.
in Run
2 and in the subsequent runs (Run 3–Run 6). For
PBI.Mo catalysed epoxidation, the yield of 1,2-epoxyhexane and
4-vinyl-1-cyclohexane 1,2-epoxide was decreased to ∼11% after
Run 6 as compared to Run 1 (fresh catalyst) (Fig. 5a and c). Sim-
ilarly, for epoxidation in the presence of Ps.AMP.Mo catalyst the
yield of 1,2-epoxyhexane and 4-vinyl-1-cyclohexane 1,2-epoxide
after Run 6 as compared to Run 1 was decreased by ∼36% and ∼22%,
respectively when compared with Run 1 (Fig. 5b and d).
The most distinguish parameter that differ PBI.Mo catalyst from
Ps.AMP.Mo catalyst is the ligand to Mo ratio. In case of PBI.Mo

M.L. Mohammed et al. / Applied Catalysis A: General 466 (2013) 142–152
149
100
80
60
40
20
0
100
Catalyst: PBI.Mo
Catalyst: Ps.AMP.Mo
80
60
40
20
0
0
40
80
120 160 200 240 280
0
40
80
120 160 200 240 280
Time [min]
Time [min]
(a)
(b)
100
80
60
40
20
0
100
80
60
40
20
0
Catalyst: PBI.Mo
•
♦
Catalyst: Ps.AMP.Mo
0
40
80
120 160 200 240 280
0
40
80
120 160 200 240 280
Time [min]
Time [min]
(c)
(d)
Run 1
Run 2
Run 3 ꢀ Run 4
Run 5 + Run 6
Run 7 (uncatalysed) – ANN predicted
Fig. 5. Catalyst reusability studies and prediction by artifical neural network (ANN) for epoxidation of 1-hexene catalysed by (a) PBI.Mo and (b) Ps.AMP.Mo catalysts, and
4-vinyl-1-cyclohexene catalysed by (c) PBI.Mo and (d) Ps.AMP.Mo catalysts at reaction temperature: 343 K; catalyst loading: 0.3 mol% Mo; feed mole ratio (FMR) of alkene
to TBHP: 5:1; agitation speed: 400 rpm.
The industrial use of both PBI.Mo and Ps.AMP.Mo catalysts for
batch alkenes epoxidation is somewhat limited unless their sta-
bility is greatly enhanced. However, in continuous epoxidation
process with FlowSyn and reactive distillation column (RDC), attri-
tion of catalyst particles and Mo leaching would not be a major
problem as the catalyst would be firmly packed inside a fixed col-
umn. On the basis of this study and unpublished pilot plant study
for continuous epoxidation with FlowSyn and RDC using PBI.Mo
and Ps.AMP.Mo catalysts, it could be concluded that PBI.Mo and
Ps.AMP.Mo catalysts could be reused several times for epoxidation
of 1-hexene and 4-vinyl-1-cyclohexene with TBHP.
recover the residue of the catalyst particles. The obtained residue
was used as a potential catalyst for epoxidation known as super-
natant studies. The aim of this study was to investigate the leaching
of catalytically active Mo from the polymer supported catalysts.
The results of supernatant studies are shown in Fig. 6. Uncatalysed
epoxidation experiment was carried out and plotted as Run 7 in
Fig. 6 to compare with other catalytic experiments in the presence
of residue.
Epoxidation of 1-hexene catalysed by the residue of PBI.Mo cat-
alyst revealed the presence of some catalytically active Mo species
after Run 1 as evident by very little catalytic effect displayed
by the reaction (Fig. 6a). However, the residue collected for the
subsequent runs failed to show any catalytic effect (Fig. 6a). On
the other hand, 4-vinyl-1-cyclohexene epoxidation catalysed by
the residue of PBI.Mo shows no trace of Mo leaching (Fig. 6c).
The residue from Ps.AMP.Mo catalysed epoxidation of 1-hexene
and 4-vinyl-1-cyclohexene revealed high content of catalytically
3.8. Supernatant studies
Once each reusability study experiment (see Fig. 5) was com-
pleted, the catalyst particles were separated from the reaction
mixture and the supernatant solutions was vacuum distilled to

150
M.L. Mohammed et al. / Applied Catalysis A: General 466 (2013) 142–152
100
80
60
40
20
0
100
Catalyst: PBI.Mo
Catalyst: Ps.AMP.Mo
80
60
40
20
0
0
40
80
120 160 200 240 280
0
40
80
120 160 200 240 280
Time [min]
Time [min]
(a)
(b)
100
80
60
40
20
0
100
80
60
40
20
0
Catalyst: PBI.Mo
Catalyst: Ps.AMP.Mo
0
40
80
120 160 200 240 280
0
40
80
120 160 200 240 280
Time [min]
Time [min]
(c)
(d)
Run 1
Run 2
Run 3
Run 4
Run 5 + Run 6
Run 7 (uncatalysed) – ANN predicted
Fig. 6. Supernatant studies and prediction by artifical neural network (ANN) for epoxidation of 1-hexene and 4-vinyl-1-cyclohexene catalysed by residues isolated from
supernatant solutions where polymer-supported Mo(VI) catalysts are reused at reaction temperature: 343 K, feed mole ratio (FMR) of alkene to TBHP: 5:1; agitation speed:
400 rpm.
active Mo (Fig. 6b and d). The amount of Mo retained by the cata-
lyst after Run 1 was 0.70 mmol Mo g−1 resin, while it was 0.62 mmol
Mo g−1 resin and 0.56 mmol Mo g−1 resin after Run 2 and Run 3,
respectively (see Fig. 6b and d) as compared to the original Mo
content, i.e. 0.74 mmol Mo g−1 resin (see Table 1). The yield of
epoxide achieved from supernatant studies of 1-hexene and 4-
vinyl-1-cyclohexene after Run 1 was 67% (Fig. 6b) and 47% (Fig. 6d),
respectively. It can be seen from Fig. 6b and d that the yield of epox-
ides decreased steadily for the subsequent runs and becomes negli-
gible after 4–6 runs. Hence, it could be concluded that there was no
Mo leaching after 4–6 runs for Ps.AMP.Mo catalysed epoxidation.
ANN model in order to find the particular set of operating condi-
tions that would produce maximum yield. ANN modelling results
are presented in Fig. 7.
For 1-hexene epoxidation catalysed by PBI.Mo catalyst, it can
be concluded that reaction temperature has the most pronounced
effect on the yield of epoxide, while the influence of catalyst load-
ing or the feed mole ratio of 1-hexene to TBHP is not very distinct
(see Figs. 2a–4a). Simulations of ANN model showed similar trends
(see Fig. 7a). The yield of 1,2-epoxyhexane was found to increase
marginally in a particular regime. For 1-hexene epoxidation catal-
ysed by PBI.Mo catalyst (Fig. 7a), the maximum yield was obtained
at an operating temperature of 353 K, catalyst loading of 0.6 mol%
Mo and feed mole ratio of 1-hexene to TBHP of 5:1, while for 1-
hexene epoxidation catalysed by Ps.AMP.Mo catalyst (Fig. 7b), the
maximum yield was obtained at a reaction temperature of 353 K,
catalyst loading of 0.6 mol% Mo and feed mole ratio of 1-hexene to
TBHP of 10:1. The profile predicted by ANN modelling was found
3.9. Optimisation study of alkene epoxidation using ANN
We have conducted an optimisation study aimed at deriving
a set of optimal conditions to complement the data presented.
Several simulations were carried out using the previously trained

M.L. Mohammed et al. / Applied Catalysis A: General 466 (2013) 142–152
151
100
80
60
40
20
0
100
80
60
40
20
Ps.AMP.Mo
PBI.Mo
Ps.AMP.Mo [ANN predicted]
PBI.Mo [ANN predicted]
0
0
40
80
120 160 200 240 280
0
40
80
120 160 200 240 280
Time [min]
Time [min]
(a)
(b)
Fig. 7. Optimisation studies and prediction by artifical neural network (ANN) for 1-hexene epoxidation catalysed by (a) PBI.Mo catalyst at a feed mole ratio (FMR) of 1-hexene
to TBHP: 5:1, and (b) Ps.AMP.Mo catalyst at a feed mole ratio (FMR) of alkene to TBHP: 10:1 at reaction temperature: 353 K; catalyst loading: 0.6 mol% Mo; agitation speed:
400 rpm.
to be in very good agreement when compared with the experimen-
tal results (Fig. 7).
Brian Mercer Feasibility award. M. L. Mohammed is grateful to
Usmanu Danfodiyo University, Sokoto, Nigeria and the Tertiary
Education Trust Fund (TETFund), Abuja, Nigeria for the PhD scholar-
ship. The authors like to thank Celanese Corporation for supplying
the microporous PBI resin beads.
4. Conclusions
Polymer supported PBI.Mo and Ps.AMP.Mo catalysts were suc-
cessfully prepared and characterised. The performance of these
catalysts was assessed for epoxidation of 1-hexene and 4-vinyl-
1-cyclohexene in a jacketed stirred batch reactor using TBHP as
an oxidant. The experimental results were modelled using ANN
to predict the catalytic activity of PBI.Mo and Ps.AMP.Mo cata-
lysts at various reaction conditions. Ps.AMP.Mo catalyst was found
to be better performed catalyst for epoxidations of 1-hexene and
4-vinyl-1-cyclohexene as compared to PBI.Mo catalyst. Catalyst
reusability and supernatant studies were conducted in a batch reac-
tor to elucidate the efficiency of PBI.Mo and Ps.AMP.Mo catalysts
for continuous epoxidation of 1-hexene and 4-vinyl-1-cyclohexene
using multi-functional reactors such as FlowSyn and RDC. The cat-
alyst reusability studies with Ps.AMP.Mo catalyst showed higher
efficiency as compared to PBI.Mo catalyst for both alkenes. How-
ever, the supernatant studies of Ps.AMP.Mo catalyst indicated the
presence of catalytically active Mo that might have contributed to
catalysis in the Ps.AMP.Mo catalysed epoxidation. The leaching of
Mo from the polymer support was attributed to lower ligand to Mo
ratio that influenced the release of active Mo species under stirring
conditions. Batch experimental results showed that the leaching of
Mo in Ps.AMP.Mo catalysed epoxidation becomes negligible after
4–6 experimental runs. The ANN modelling of the batch experimen-
tal data have successfully predicted the catalytic performances of
PBI.Mo and Ps.AMP.Mo catalysts for epoxidation of 1-hexene and
4-vinyl-1-cyclohexene with a certain probability.
References
[1] G. Sienel, R. Rieth, K.T. Rowbottom, in: B. Elvers (Ed.), Ullmann’s Encyclopedia
of Industrial Chemistry, Wiley-VCH, Weinheim, 2000.
[2] K. Bauer, D. Garbe, H. Surburg, Common Fragrance and Flavour Materials, fourth
ed., Wiley-VCH, Weinheim, 2001, pp. 143–145.
[3] G. Grigoropoulou, J.H. Clark, J.A. Elings, Green Chem. 5 (2003) 1–7.
[4] K.Y. Andrei, Aziridines and Epoxides in Organic Synthesis, Wiley-VCH, Wein-
heim, 2006, pp. 185–186.
[5] D. Swern, Organic Peroxides, Wiley Interscience, New York, 1971.
[6] St.V. Kotov, E. Balbolov, J. Mol. Catal. A: Chem. 176 (2001) 41–48.
[7] R.M. Calvente, J.M. Campos-Martin, J.L.G. Fierro, Catal. Commun. 3 (2002)
247–251.
[8] Q.-H. Xia, H.-Q. Ge, C.-P. Ye, Z.-M. Liu, K.-X. Su, Chem. Rev. 105 (2005)
1603–1662.
[9] G. Blanco-Brieva, M.C. Capel-Sanchez, J.M. Campos-Martin, J.L.G. Fierro, J. Mol.
Catal. A: Chem. 269 (2007) 133–140.
[10] A. Serafimidou, A. Stamatis, M. Louloudi, Catal. Commun. 9 (2008) 35–39.
[11] M. Salavati-Niasari, E. Esmaeili, H. Seyghalkar, M. Bazarganipour, Inorg. Chim.
Acta 375 (2011) 11–19.
[12] Sk.M. Islam, A.S. Roy, P. Mondal, M. Mobarok, B. Roy, N. Salam, S. Paul, S. Mondal,
Monatsh. Chem. 143 (2012) 815–823.
[13] J. Huang, X. Fu, Q. Miao, Appl. Catal. A 407 (2011) 163–172.
[14] E. Angelescu, O.D. Pavel, R. Ionescu, R. Bîrjega, M. Badea, R. Za˘voianu, J. Mol.
Catal. A: Chem. 352 (2012) 21–30.
[15] U. Arnold, W. Habicht, M. Döring, Adv. Synth. Catal. 348 (2006) 142–150.
[16] G.R. Nath, K. Rajesh, Asian J. Chem. 24 (2012) 4548–4550.
[17] J.K. Satyarthi, D. Srinivas, Appl. Catal. A 401 (2011) 189–198.
[18] E. Mikolajska, V. Calvino-Casilda, M.A. Ban˜ares, Appl. Catal. A 421 (2012)
164–171.
[19] S. Hu, D. Liu, C. Wang, Y. Chen, Z. Guo, A. Borgna, Y. Yang, Appl. Catal. A 386
(2010) 74–82.
[20] Q. Jin, J. Bao, H. Sakiyama, N. Tsubaki, Res. Chem. Intermed. 37 (2011) 177–184.
[21] J.K.F. Buijink, J.J.M. van Vlaanderen, M. Crocker, F.G.M. Niele, Catal. Today 93–95
(2004) 199–204.
[22] M. Guidotti, E. Gavrilova, A. Galarneau, B. Coq, R. Psaro, N. Ravasio, Green Chem.
13 (2011) 1806–1811.
Acknowledgements
The authors gratefully acknowledge the financial support pro-
vided by the EPSRC (grant no. EP/H027653/1) and The Royal Society

152
M.L. Mohammed et al. / Applied Catalysis A: General 466 (2013) 142–152
[23] G. Grivani, S. Tangestaninejad, M.H. Habibi, V. Mirkhani, M. Moghadam, Appl.
Catal. A 299 (2006) 131–136.
[33] C.B. Kelly, C. Lee, M.A. Mercadante, N.E. Leadbeater, Org. Process Res. Dev. 15
(2011) 717–720.
[24] S. Tangestaninejad, V. Mirkhani, M. Moghadam, G. Grivani, Catal. Commun. 8
(2007) 839–844.
[25] L. Barrio, J.M. Campos-Martin, M. Pilar de Frutos, L.G. Fierro, Ind. Eng. Chem.
Res. 47 (2008) 8016–8024.
[26] K. Ambroziak, R. Mbeleck, B. Saha, D. Sherrington, Int. J. Chem. React. Eng. 8
(2010) A125.
[27] J. Moreno, J. Iglesias, J.A. Melero, D.C. Sherrington, J. Mater. Chem. 21 (2011)
6725–6735.
[28] K. Huang, F.-Q. Chen, D.-W. Lü, Appl. Catal. A: Gen. 219 (2001) 61–68.
[29] A. Corma, J.M. Serra, Catal. Today 107–108 (2005) 3–11.
[30] N. Oku, T. Seo, US patent 6646138 (November 11, 2003) to Sumitomo Chemical
Company, Limited.
[34] K.B. Sharpless, T.R. Verhoeven, Aldrichim. Acta 12 (1979) 63–74.
[35] P.N. Chen, R.O. Tucker, US patent 4628067 (December 9, 1986) to Celenese
Corporation.
[36] D.C. Sherrington, Chem. Commun. (1998) 2275–2286.
[37] P. Serna, L.A. Baumes, M. Moliner, A. Corma, J. Catal. 258 (2008) 25–34.
[38] A. Corma, J.M. Serra, P. Serna, S. Valero, E. Argente, V. Botti, J. Catal. 229 (2005)
513–524.
[39] A. Jodaei, D. Salari, A. Niaei, M. Khatamian, S.A. Hosseini, J. Environ. Sci. Health
A: Tox. Hazard. Subst. Environ. Eng. 46 (2011) 50–62.
[40] A. Niaei, D. Salari, F. Aghazadeh, N. C¸ aylak, A. Sepehrianazar, J. Envi-
ron. Sci. Health A: Tox. Hazard. Subst. Environ. Eng. 45 (2010)
454–463.
[31] R. Taylor, R. Krishna, Chem. Eng. Sci. 55 (2000) 5183–5229.
[32] C. Mueanmas, K. Prasertsit, C. Tongurai, IJCEA 1 (2010) 77–83.
[41] K. Ambroziak, R. Mbeleck, Y. He, B. Saha, D.C. Sherrington, Ind. Eng. Chem. Res.
48 (2009) 3293–3302.