Catalysed ethylbenzene dehydrogenation in CO2 or N2 - Carbon deposits as the active phase

Article abstract of DOI:10.1016/j.apcata.2011.12.037

Bare alumina support transforms into an active catalyst for the dehydrogenation of ethylbenzene to styrene in CO₂ or N₂. During the first 15 h time on stream in CO₂, or the first 10 h time on stream in N₂, the a

Full text of DOI:10.1016/j.apcata.2011.12.037

Applied Catalysis A: General 417–418 (2012) 163–173
Contents lists available at SciVerse ScienceDirect
Applied Catalysis A: General
journal homepage: www.elsevier.com/locate/apcata
Catalysed ethylbenzene dehydrogenation in CO2 or N2—Carbon deposits as the
active phase
Christian Nederlof, Freek Kapteijn, Michiel Makkee∗
Catalysis Engineering, Department of Chemical Engineering, Faculty of Applied Sciences, Delft University of Technology, Julianalaan 136, NL-2628 BL, Delft, The Netherlands
a r t i c l e i n f o
a b s t r a c t
Article history:
Bare alumina support transforms into an active catalyst for the dehydrogenation of ethylbenzene to
styrene in CO2 or N2. During the first 15 h time on stream in CO2, or the first 10 h time on stream in N2,
the alumina shows an increase in ethylbenzene conversion and styrene selectivity from 15% to 60% and
from 60% to 92%, respectively, under industrially relevant conditions of 600 ◦C and 10 vol% ethylbenzene.
Thereafter, the system slowly deactivates, but remains highly selective. TGA analysis shows an increase
in coke content. The specific surface area and pore volume show a decrease with time on stream. TEM-
imaging reveals that the spent catalyst surface is completely covered by several layers of coke. These
results combined suggest that the carbon deposits on the alumina are responsible for the increase in
activity and selectivity, and also are the cause of deactivation once a monolayer of carbon is deposited on
the support surface. Similar trends are observed for zirconia support. Supported vanadium and chromium
oxides on alumina all give similar results, but after a faster activity development. Also for these supported
catalysts and even carbon samples, deposited coke is the active and selective phase.
Received 23 November 2011
Received in revised form
21 December 2011
Accepted 22 December 2011
Available online 31 December 2011
Keywords:
Dehydrogenation
Carbon dioxide
Nitrogen
Ethylbenzene
Styrene
Carbon deposits
Coke
© 2012 Elsevier B.V. All rights reserved.
Alumina
Zirconia
1. Introduction
temperatures CO2 is considered to be a mild oxidant that relaxes the
equilibrium limitations of the current steam-aided dehydrogena-
Styrene monomer is one of the largest bulk monomers at this
moment and will probably be in this position for a long time to
come. In industry it is mainly produced by steam-aided dehydro-
genation of ethylbenzene over a promoted potassium-iron catalyst.
This process suffers from high energy requirements and low per-
pass conversions due to its endothermic nature and equilibrium
limitations. A large part of the energy is lost in the steam conden-
sation, which cannot be recovered due to practical reasons [1,2].
To overcome these issues, oxidative dehydrogenation with pure
oxygen has been explored. This solves the issues related to the
endothermic nature of dehydrogenation and its equilibrium lim-
itations, but brings its own issues. It has a poor selectivity due to
secondary oxidation, safety issues like risk of explosion, and may
lead to hot-spots in the catalyst bed due to the highly exothermic
nature of this type of reaction. Due to these issues and the higher
operating costs oxidative dehydrogenation has not been commer-
cialised yet [2].
tion process. Also the energy losses due to steam condensation can
be circumvented, making the overall process energy saving [3–7].
The CO2 oxidative dehydrogenation (ODH) process is, however,
still equilibrium limited [4]. The equilibrium conversion at 600 ◦C
and 1:10 ratio of EB:CO2 is 91%, compared to 76% for the steam-
aided dehydrogenation process under the same conditions. The CO2
oxidative dehydrogenation reaction (3) is a result of a two-step
mechanism consisting of direct dehydrogenation (1) followed by
the reverse water-gas-shift (RWGS) (2) that removes the hydrogen
[4,8].
C8H10 ꢀ C8H8 + H2
(1)
(2)
(3)
CO2 + H2 ꢀ CO + H2O
C8H10 + CO2 ꢀ C8H8 + CO + H2O
A lot of research effort has already been put into the develop-
ment of a CO2 ODH process for styrene production. Many catalysts
have been tested, using many types of supports, and quite active
and selective catalysts have been claimed [4–7,9–35]. An often
used catalyst support is alumina [6,7,12–17]. The overview in
Table 1 shows some examples of alumina supported catalysts with
chromium or vanadium. Most of the recent publications on alumina
supported catalysts reported that the alumina support itself had a
CO2 has been investigated as an alternative oxidant for this pro-
cess. The use of CO2 is safer than using oxygen and at the used
∗
Corresponding author. Tel.: +31 15 2781391; fax: +31 15 2785006.
E-mail address: m.makkee@tudelft.nl (M. Makkee).
0926-860X/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
doi:10.1016/j.apcata.2011.12.037

164
C. Nederlof et al. / Applied Catalysis A: General 417–418 (2012) 163–173
Table 1
Overview of performance and conditions of supported Al2O3 catalysts with chromium or vanadium.
Source
Catalyst [mmol/g]
X [%]
S [%]
T [◦C]
CO2:EB
Wt [g]
WHSV [g/g/h]
Fedorov et al. [7]
Ye et al. [13]
Sakurai et al. [5]
Vislovskiy et al. [12]
Sun et al. [14]
0.65 Cr
2.6 Cr
1 V
2.2 V
0.57 V
1.5 V
24
76.9
59
74.8
45.5
67.4
79
550
550
550
595
550
550
1
19
50
1
11
20
10
0.65
0.42
1.5
1.0
2.16
0.58
98.6
84.9
94.5
97.8
96
0.10
0.10
1.0
0.20
0.30
Chen et al. [16]
X = conversion; S = selectivity; T = reaction temperature; EB = ethylbenzene; Wt = catalyst weight; WHSV = [(g EB)/(g cat)/h].
poor activity. Additional metal oxides were required to make a good
catalyst. Only Sato et al. found alumina and sodium on alumina to
be active under slightly different conditions, but with a low selec-
tivity. An overview of reported alumina performances and applied
conditions is presented in Table 2.
Generally, the CO2 oxidative dehydrogenation is believed to
proceed through a redox type mechanism over a transition metal
oxide [4,5,8–35]. Sato et al. [6] attributed the activity of alumina
to the combined action of acid and basic sites on the surface of the
alumina. Ethylbenzene preferably adsorbs on the acidic sites and
carbon dioxide on the basic sites. The dehydrogenation could then
proceed via acidic protolysis, or directly by the reaction with car-
bon dioxide. The activity of the alumina was improved by adding
alkaline metals, increasing also the conversion of hydrogen with
CO2 via the RWGS reaction.
The major issue with CO2 ODH is the deactivation of the cata-
lysts due to excessive coking [36]. Vislovskiy et al. [12] compared
conversion data from 1 h and 4 h time on stream at 595 ◦C, where
up to 46% of activity was lost. Chen et al. [16] compared conversion
data at 1 h and 20 h operation at 550 ◦C, where up to 21% of the
activity was lost. In all cases a large amount of coke is found on the
spent catalysts, the cause of this deactivation. In the steam-aided
process one of the functions of steam is to counteract this reversible
deactivation by coke gasification (4). A similar role is envisaged for
CO2 through a reverse Boudouard reaction (5), but this has not been
proven successful yet.
90 ␮m was used for the dehydrogenation experiments without any
further pre-treatment.
Six supported catalysts were made by co-impregnation using
the incipient wetness method, using the same ␥-Al2O3 support
as above. Three had vanadium oxide as the main component
(1 mmol/g loading), promoted by potassium oxide (0.5 mmol/g
loading) or magnesium oxide (0.5 mmol/g loading). These will
be denoted by V, KV, and MgV. The other three were a sim-
ilar series of chromium oxide, denoted by Cr, KCr, and MgCr.
The support was dried at 120 ◦C before impregnation. The
required amounts of Cr(NO3)3·9H2O, Mg(NO3)2·6H2O, KNO3 or
NH4VO3 + 2C2H2O4·2H2O were dissolved in demineralised water,
after which the gamma alumina support was impregnated with
the solution. The wetted support was shaken until a homogeneous
colour was obtained visually. The wetted support was then dried
at 120 ◦C in air for 4 h, followed by calcination in air at 600 ◦C for
4 h in a static air calcination oven.
Additional experiments with CO2 and N2 were performed
with another ␥-Al2O3 sample (Ketjen CK300-000-1.5E, 270 m2/g,
0.65 ml/g). These extrudates were crushed and sieved, the frac-
tion between 212 and 425 ␮m was used for the dehydrogenation
experiments without any further pre-treatment.
2.2. Catalyst testing
The dehydrogenation experiments were carried out in a quartz
fixed bed reactor with inner diameter of 8 mm. The reaction tem-
perature was 600 ◦C and the pressure at the reactor outlet was
atmospheric. In a typical experiment 400 mg of catalyst was used.
On top of the catalyst bed was a 1 cm layer of SiC to preheat the feed.
The catalyst bed and SiC were fixed in the quartz tube by quartz
wool plugs. Carbon dioxide was fed through a mass-flow controller
set at 15 ml (NTP)/min (20 ◦C, 1 atm). Ethylbenzene (EB) was fed to
the reactor by passing the gas feed over an ethylbenzene saturator
vessel at a temperature of 85 ◦C. This resulted in feed mixture of
ethylbenzene and CO2 with a molar ratio of 1:10, as confirmed by
gas chromatography (GC). The ethylbenzene weight hourly space
velocity (WHSV) was 1 g/g/h (wt EB/wt cat). The total GHSV was
1500 l/l/h (vol. gas/vol. cat). These conditions are representative
for industrial operation. Before starting the experiment the reac-
tor would be heated up with 10 ◦C/min and allowed to stabilise at
reaction temperature for half an hour under CO2 atmosphere. The
experiment was started by feeding ethylbenzene to the reactor.
C + H2O → CO + H2
C + CO2 → 2CO
(4)
(5)
In this paper we demonstrate the transformation of ␥-alumina
and zirconia into a very active and selective CO2 ODH catalyst by
deposition of coke under reaction conditions. The activity and selec-
tivity are attributed to this coke.
2. Experimental
2.1. Catalyst preparation
The ␥-Al2O3 (Engelhard Al-3972 P, 222 m2/g, 0.53 ml/g), ZrO2
(AlfaAesar 043815, 104 m2/g, 0.38 ml/g), activated coal (Norit
RB-II, 691 m2/g, 1.06 ml/g), carbon nanotubes (Hyperion Catalysis
CS-02C-063-XD, 425 m2/g, 1.02 ml/g) and SiC (120 ␮m) samples
were crushed if needed and sieved, the fraction between 40 and
Table 2
Overview of reported Al2O3 data [6,7,11,16].
Source
X [%]
S [%]
T [◦C]
CO2:EB
Wt [g]
WHSV [h−1
]
Fedorov et al. [7]
Sato et al. [6]
Mimura and Saito [11]
Chen et al. [16]
1
58
100
72
550
610
550
550
1
3.4
11
20
10
2
1.4
0.3
0.65
1.72
2.26
0.58
7a/11b
5
80a/87b
88
X = conversion; S = selectivity; T = reaction temperature; EB = ethylbenzene; Wt = catalyst weight; WHSV = [(g EB)/(g cat)/h].
a
0.42 h reaction time.
6 h reaction time.
b

C. Nederlof et al. / Applied Catalysis A: General 417–418 (2012) 163–173
165
For the comparative experiments with CO2 and N2, a parallel
flow reactor setup was used with quartz fixed bed reactors with
inner diameter of 4 mm. In a typical experiment 1000 mg of catalyst
was used. Carbon dioxide and ethylbenzene flow were adjusted
accordingly (36 ml CO2 (NTP)/min and 1.00 g EB/h) to reach the
same ethylbenzene weight hourly space velocity (WHSV) of 1 g/g/h
(wt EB/wt cat). Other reaction conditions and the reactor loading
procedure were the same.
The concentrations in the reactor effluent were measured online
by a two channel gas chromatograph with a TCD (columns: 0.3 m
Hayesep Q 80-100 mesh with back-flush option, 25 m × 0.53 mm
Porabond Q, and 15 m × 0.53 mm molsieve 5 A with bypass option
for CO2 and H2O, all in series) for permanent gas analysis (CO2, H2,
N2, O2, CO) and a FID (column: 30 m × 0.53 mm, Df = 3 ␮m, RTX-1)
for the hydrocarbons analysis (methane, ethane, ethene, benzene,
toluene, ethylbenzene, styrene, and tars). A sample was analysed
every 15 min in order to follow the catalyst performance over time.
A constant flow of nitrogen was used as internal standard. Con-
versions and selectivities are based on moles of ethylbenzene, and
calculated according to Eqs. (6) and (7). Conversion of CO2 was cal-
culated according to Eq. (8). The overall carbon balance is >99%,
only carbon that was deposited on the catalyst as coke is lacking
from the carbon balance by GC.
80% and increases to a steady value of 92% for all catalysts after
about 10 h. Similar to the ethylbenzene conversion (Fig. 1C), the
CO2 conversion starts high at 10–19% and is reduced to 2–5% in 20 h.
The increase in styrene selectivity is due to a decrease of cracking
products (benzene, toluene) and a decrease of the surplus hydro-
gen and CO. The surplus of hydrogen and CO is the amount of these
products that is above the expected amount based on the dehydro-
genation coupled with the RWGS reaction. Differences between the
catalysts are small and disappear with time on stream. The initial
differences in ethylbenzene conversion are related to the CO2 con-
version. The potassium promoted vanadium catalyst (KV) has the
highest initial conversion of ethylbenzene and CO2.
Five other bare samples were tested under CO2 oxidative
dehydrogenation conditions: ␥-alumina (Al2O3), high surface area
zirconia (ZrO2), activated carbon (AC), carbon nanotubes (CNT), and
the silicon carbide (SiC) that was used as diluent material. The con-
version and selectivity data for these samples are presented in Fig. 2
for 20 h time on stream. The SiC diluent material showed a con-
stant 3% conversion and 60–70% selectivity during the 20 h time
on stream. Therefore, silicon carbide is considered to be inert in
this reaction. The alumina sample started at 20% conversion and
70% selectivity, but both increased steadily over time. After 10 h
on stream the conversion and selectivity increase levelled off, the
conversion stabilised around 60% with a selectivity of 92%. Zirconia
was initially more active than alumina. ZrO2 shows an optimum
after 2 h and then deactivates. The selectivity to styrene over zirco-
nia increased initially, until reaching a stationary value of 92%. The
selectivity plateau was reached at the same time as the maximum
conversion was reached. The CNT sample behaved differently from
alumina and zirconia. It deactivated at first, after which the con-
version stabilised around 13%. Its selectivity was constant during
20 h of reaction at 76%. The activated carbon material was initially
very active, but it also deactivated very fast to end up at a conver-
sion of 13% after 20 h. The initial selectivity of activated carbon was
poor, but very quickly increased to its highest value of 83% after
2 h. After this optimum the styrene selectivity slowly decreased to
78% after 20 h on stream at 13% conversion. The conversion of CO2
is much lower for the bare samples than for the alumina supported
samples. Over alumina just 2% is reached, over the AC sample it sta-
bilises around 1%, whilst the CNT sample displays almost no CO2
conversion. Only ZrO2 exhibits somewhat higher CO2 conversions,
decreasing from 7% to 3% in 20 h.
In order to see if alumina reached an optimum in performance or
a stable performance after 20 h, a longer time on stream was eval-
uated. The results are shown in Fig. 3 (left), together with data on
coke loadings. An optimum in activity was obtained after 15 h, with
a conversion of 62% and with 92% selectivity, after which the cata-
lyst started deactivating. The deactivation progressed at an almost
constant rate until around 80 h. After this time the catalyst seemed
to deactivate at a slower rate. The selectivity to styrene increased
in the first 15 h, after which it reached a stationary level of 92%.
This experiment with alumina shows large similarity with zirconia
(Fig. 3, right). For both samples a selectivity plateau was reached
at the same time as the maximum conversion was reached, and for
both supports these levels were nearly identical. However, the ZrO2
sample is activated and deactivated about ten times faster than the
Al2O3 sample.
ethylbenzenein − ethylbenzeneout
conversion =
selectivity =
(6)
(7)
ethylbenzenein
styreneout
ethylbenzenein − ethylbenzeneout
CO2,in − CO2,out
CO2 conversion =
(8)
CO2,in
2.3. Catalyst characterization
After an experiment, a sample of the catalyst was analysed
to determine the amount of coke by a TGA (MettlerToledo
TGA/SDTA851e, 20 mg sample of spent catalyst, 100 ml/min air,
50 ml/min He flow) using a ramp of 3 ◦C/min from RT to 723 ◦C.
Surface area, pore volume, and pore size distribution were deter-
mined of both fresh and spent catalyst samples by N2 adsorption at
−196 ◦C (BET-method, Quantachrome Autosorb 6B). The samples
were pre-treated in nitrogen at 250 ◦C. In Raman spectroscopy, the
fresh and spent catalyst samples were analysed without any pre-
treatment. Spectra were recorded using a Renishaw Raman Imaging
Microscope, system 2000, using a green 514 nm laser of 20 mW. A
Leica DMLM optical microscope with a 50× objective lens was used
to determine the analysed part of the sample. The Ramascope was
calibrated using a silicon wafer. The power output of the laser was
set at 20% to avoid influencing the sample.
Transmission electron microscopy (TEM) was performed using
a FEI Tecnai TF20 electron microscope with a field-emission gun as
the source of electrons operated at 200 kV. Samples were mounted
on Quantifoil® carbon polymer supported on a copper grid by plac-
ing a few droplets of a suspension of ground sample in ethanol on
the grid, followed by drying at ambient conditions.
3. Results
Comparative experiments with alumina and CO2 or N2 were
done to observe if carbon dioxide could have an effect on the forma-
tion of coke and the rate thereof, or on the activity of the deposited
coke. The comparative performance data in Fig. 4 shows that the
alumina in N2 becomes active faster and selective faster than in CO2.
In N2 the optimal performance is reached after 10 h, where 15 h are
required in CO2. Also the coke build-up under nitrogen atmosphere
is faster than that in CO2. When comparing the data for N2 after
10 h and for CO2 after 15 h, the ethylbenzene conversion, styrene
3.1. Catalyst performance
All alumina supported catalysts (V, KV, MgV, and Cr, KCr, MgCr)
showed a very similar performance. The ethylbenzene conversion,
styrene selectivity, and CO2 conversion in the CO2 ODH of ethylben-
zene to styrene over these six catalysts is shown in Fig. 1. Under the
applied conditions the ethylbenzene conversion starts between 92
and 85% but decreases to 53–56% in 20 h. Selectivity starts below

166
C. Nederlof et al. / Applied Catalysis A: General 417–418 (2012) 163–173
100%
90%
80%
70%
60%
50%
100%
A
B
90%
80%
70%
60%
V
Cr
V
Cr
KV
MgV
KCr
KV
MgV
KCr
MgCr
MgCr
50%
0
5
10
15
20
0
5
10
15
20
Time [h]
Time [h]
20%
16%
12%
8%
C
V
Cr
KV
MgV
KCr
MgCr
4%
0%
0
5
10
15
20
Time [h]
Fig. 1. Ethylbenzene conversion (A), styrene selectivity (B), and CO2 conversion (C) as a function of time on stream with CO2 as diluent for six Al2O3 supported catalysts: V
(᭹), KV (ꢁ), MgV (ꢂ), Cr (ꢀ), KCr (ꢁ), MgCr (ꢃ). Conditions as described in Section 2.
selectivity and coke build-up are identical. The distribution of by-
products during these experiments are shown in Fig. 5. The main
byproducts are benzene and toluene, followed by the surplus of
CO + H2. Tars are produced the least. The toluene formation follows
the same trend as the EB conversion, whilst benzene is formed in
an earlier stage of time on stream, indicating a decreasing cracking
severity with increasing coke deposition.
The styrene selectivity increases with ethylbenzene conversion
in these experiments (Fig. 6) indicating that an improved selectivity
is not due to a decreasing activity, as would be the case for consec-
utive reactions, but should be attributed to the development of a
more selective system. The performance for both cases is identical
for styrene selectivities above 90%.
To check if the selected diluent really did not have any effect
on the performance of the alumina samples, the other diluent was
fed over the reactor, a few hours after their optimal performance in
either CO2 or N2 flow. The results are presented in Fig. 7. Switching
from CO2 to N2 diluent after 20 h gives almost an almost identical
performance with the CO2 experiment. Similarly, switching from
N2 to CO2 diluent after 15 h gives an almost identical performance
with the N2 experiment. The minor differences can be attributed
to the dilution ratio that was measured to be a little higher for
N2 than that for CO2 (10.3 and 9.9), having a small effect on the
conversion.
3.2. Catalyst characterization
In replicated experiments with alumina and zirconia, catalyst
samples were produced after different hours on stream for further
characterization. All these experiments showed the same charac-
teristic changes in selectivity and conversion as the experiments
shown in Fig. 3 (left) and Fig. 4 for alumina and in Fig. 3 (right) for
zirconia. Thermo gravimetric analysis (TGA) of the samples in air
showed an increase in the amount of coke on the catalyst with time,
reaching 36 wt% of coke after 96 h with alumina and 13.5 wt% coke
after 16 h with zirconia. The TGA profiles of CNT, AC, Al2O3, and
zirconia samples are presented in Fig. 8. The oxidation profiles of
spent CNT and AC samples are shifted to lower temperatures than
those of the pristine carbon material. Coke oxidised the easiest on
zirconia, followed by alumina, AC, and CNT, in that order.
An overview of the characterization data is given in Table 3.
The six supported vanadia and chromia catalysts all showed high
loadings of coke in TG analysis. The promoted samples had slightly
reduced coke loadings. The KV sample had the highest CO2 conver-
sion and RWGS activity. From the other samples, only ZrO2 showed
some RWGS activity. All samples exhibited a large drop in surface
area and pore volume due to the coke build-up. The coke load-
ing on alumina used in nitrogen is higher than the coke loading
on alumina used in carbon dioxide. The characterization data for

C. Nederlof et al. / Applied Catalysis A: General 417–418 (2012) 163–173
167
70%
60%
50%
40%
30%
20%
10%
0%
100%
B
ZrO2
A
Al2O3
90%
ZrO2
Al2O3
AC
80%
70%
60%
50%
40%
CNT
SiC
AC
CNT
SiC
0
5
10
15
20
0
5
10
15
20
Time [h]
Time [h]
10%
8%
6%
4%
2%
0%
C
ZrO2
Al2O3
AC
CNT
0
5
10
15
20
Time [h]
Fig. 2. Ethylbenzene conversion (A), styrene selectivity (B), and CO2 conversion (C) as a function of time on stream with CO2 as diluent for five bare samples: ZrO2 (ꢁ), Al2O3
(᭹), activated carbon (×), carbon nanotubes (ꢄ), SiC (ꢂ). Conditions as described in Section 2.
100%
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
30%
25%
20%
15%
10%
5%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
selectivity
selectivity
conversion
conversion
coke
coke
Al2O3
80 100
ZrO2
0%
0
20
40
60
0
5
10
15
20
Time [h]
Time [h]
Fig. 3. Styrene selectivity, ethylbenzene conversion, and coke deposited as a function of time with CO2 as diluent using Al2O3 or ZrO2. Conditions as described in Section 2.

168
C. Nederlof et al. / Applied Catalysis A: General 417–418 (2012) 163–173
Table 3
Characterization data of fresh and spent catalyst samples (after 20 h time on stream).
Sample
V
KV
25.7
201
95
0.46
0.15
MgV
Cr
26.3
205
98
0.47
0.15
KCr
22.2
199
110
0.45
0.18
MgCr
Al2O3
ZrO2
AC
93.4
691
n.d.
0.52
n.d.
CNT
Al2O3-N2
Coke [wt%]
SA [m2/g]
28.5
202
96
25.5
193
89
0.46
0.14
25.0
202
94
0.46
0.15
20.8
222
142
0.53
0.25
13.5
104
68
0.31
0.18
96.7
24.6
n.d.
n.d.
n.d.
n.d.
F
S
F
S
425a
317
0.47
0.15
1.02a
1.06
Vp [ml/g]
F = fresh samples; S = spent sample; AS = specific surface area from BET analysis; Vp = pore volume from BET analysis.
a
Data from the manufacturer.
Table 4
Characterization data of the duplicated Al2O3 experiments in CO2 after different times on stream (TOS).
TOS [h]
0
5
10
15
22
48
80
0.12
32.8
96
SA [m2/g]
Vp [ml/g]
Coke [wt%]
222
178
174
0.37
11.4
155
0.30
18.1
142
0.25
21.6
41
0.07
35.7
0.53
–
0.42
6.0
Table 5
Characterization data of the duplicated Al2O3 experiments in N2 after different times on stream (TOS).
TOS [h]
0
2
6
8
14
178
0.34
21.0
20
SA [m2/g]
Vp [ml/g]
Coke [wt%]
270
237
223
209
152
0.27
28.0
0.65
–
0.58
6.2
0.47
13.6
0.43
15.8
Table 6
Characterization data of the duplicated ZrO2 experiments in CO2 after different times on stream (TOS).
TOS [h]
0
0.5
70
0.24
4.0
1
2
4
8
16
SA [m2/g]
Vp [ml/g]
Coke [wt%]
101
72
0.23
6.3
70
0.23
7.1
71
0.21
8.5
71
0.21
10.3
72
0.18
13.5
0.29
–
the duplicated experiments are given in Tables 4–6 for alumina in
CO2, alumina in N2, and zirconia, respectively. These analyses show
that initially there is a fast drop in surface area and pore volume,
for longer times the changes are smaller. Coke build-up is also fast
initially and slows down at longer times on stream.
Raman spectra from several of the spent catalysts (Fig. 9)
showed the same peaks with the same intensity ratio between
1330 cm−1 (amorphous carbon [37]) and 1590 cm−1 (graphitic car-
bon [37]). These normalised spectra show no qualitative differences
between the different catalysts, including the supported catalysts,
or the different times on stream.
HR-TEM micrographs in Figs. 10 and 11 show both fresh and
spent catalysts. The pristine ZrO2 support samples have a clear
structure of crystalline spheres (Fig. 10) whilst crystalline plates are
visible for Al2O3 (Fig. 11). The spent samples in Fig. 10B and Fig. 11B
show the presence of irregular, layered structures of coke that are
deposited on the external surface of the particles. At least three
layers of carbon are present after 20 h on stream on both supports.
In the highly deactivated and coked Al2O3 sample in Fig. 11C, all
voids between the crystalline structures are completely filled with
coke. Eight and even more layers of carbon can be counted. The
structure of the parent material remains unchanged in the spent
catalyst samples.
100%
selectivity
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
conversion
N2
4. Discussion
CO2
All six supported catalysts (V, KV, MgV, and Cr, KCr, MgCr) show
a very similar behaviour in time (Fig. 1). They are initially very
active, but unselective. Selectivities rise to a steady, high value of
92% after 5 h, whilst conversion continuously decreases. A possi-
ble explanation for the increasing selectivity in the first hours is
that the catalyst is reduced to the appropriate oxidation state and
the most acidic sites on the catalyst are changed or have become
inaccessible. Cracking of ethylbenzene into benzene or toluene is
reduced, as well as the surplus of hydrogen and CO. Highly active
and unselective sites disappear by coke coverage, first forming
on the non-selective sites, decreasing activity but increasing the
coke
0
5
10
15
20
Time [h]
Fig. 4. Styrene selectivity (ꢁꢁ), ethylbenzene conversion(᭹ꢀ), and coke build-up
(ꢂꢃ) as a function of time on stream for Al2O3 under CO2 (ꢁ᭹ꢂ) or N2 (ꢁꢀꢃ)
atmosphere. Conditions as described in Section 2.

C. Nederlof et al. / Applied Catalysis A: General 417–418 (2012) 163–173
169
3%
3%
2%
1%
0%
surplus CO+ H2
toluene
2%
1%
0%
benzene
tars
CO2
N2
CO2
N2
0
5
10
15
20
0
5
10
15
20
Time [h]
Time [h]
Fig. 5. Selectivities of benzene (ꢄ♦), toluene (ꢂꢃ), surplus CO + H2 (ꢁꢁ), and tars (᭹ꢀ) as a function of time on stream for Al2O3 under CO2 (ꢁ᭹ꢂꢄ) or N2 (ꢁꢀꢃ♦) atmosphere.
Conditions as described in Section 2.
selectivity at the same time. The formation of coke does not stop
when there are no non-selective sites left, but it continues, deac-
tivating the catalysts further. The catalysts become less accessible
by this fouling and differences disappear as the original catalyst
surface is no longer accessible. Coke initially has a positive influ-
ence on the selectivity, but later a negative effect on the activity.
With decreasing dehydrogenation activity, the CO2 conversion also
decreases (Fig. 1C). These seem to be directly related to each other.
The observed activity of alumina in the CO2 ODH of ethyl-
benzene is consistent with previous literature reports [6,7,11,16].
Apparent discrepancies can be attributed to differences in the time
on stream, molar ratio of CO2 and ethylbenzene, ethylbenzene par-
tial pressure, and space velocity of the reported activity data. For
short times on stream the EB conversion over alumina is below
20%, representing a poor activity, but for longer times on stream
conversion may develop to above 60%, which represents a reason-
able activity and is similar to the maximum conversion of ZrO2
under the same conditions (Fig. 2). These examples clearly show
that initial catalyst performance data, at short times on stream, are
a wrong performance indicator. In the presented case, dehydro-
genation experiments should be run for at least 20 h in order to
get a good and relevant impression of the catalyst performance in
terms of conversion, selectivity, and stability. The experiment with
alumina is a prime example, since at 20 h on stream the catalyst
has barely started to deactivate even though >20 wt% coke has been
deposited. This indicates that some catalysts like alumina have to
run longer (e.g. for 96 h, Fig. 3) to fully appreciate their performance.
The combined information on Al2O3 from TGA, N2-
adsorption, and HR-TEM micrographs (Fig. 11), confirm that
a
growing, irregular layer of coke is being deposited on
the catalyst surface. At first this increases both activity and
selectivity. Secondly, with increasing time, the coke is filling up
the catalyst pores, decreasing the specific surface area and pore
volume (Tables 4 and 5). After about 15 h in CO2 diluent or 10 h
in N2 diluent, the loss of available surface area causes activity to
decrease, but the catalytic system remains highly selective. Only
selective poisoning of unselective active sites, that would increase
selectivity, as is the case with the supported catalysts samples,
does not apply here. The EB dehydrogenation activity of alumina
increases as well. Coke clearly contributes to the catalyst activity
and selectivity and is likely to be a catalytic species.
Similar reasoning can be followed for the case of ZrO2. A fast
initial coke build-up in the first 2 h is observed by TG analysis, as
is displayed in Fig. 3, whilst conversion and selectivity are also
both rising quickly. Zirconia has been reported active in the CO2
oxidative dehydrogenation [18]. The active site was identified to
be tetragonal phase zirconia [18]. As this phase is not present in
alumina, and both supports exhibit a very similar EB dehydrogena-
tion behaviour, this points to another active site that both materials
have in common, viz. the carbon deposits. The formation of coke
proceeds over acidic sites [38]. Zirconia is a much more acidic
support than alumina [39], hence the formation of coke and its
activation for dehydrogenation is also much faster.
100%
95%
90%
85%
80%
75%
70%
CO2
N2
0%
20%
40%
60%
80%
Conversion
Interestingly, the amount of coke at which these systems reach
their optimum conversion and selectivity, corresponds with the
Fig. 6. Styrene selectivity at different ethylbenzene conversions over bare Al2O3 in
CO2 or N2 as diluent. Conditions as described in Section 2.

170
C. Nederlof et al. / Applied Catalysis A: General 417–418 (2012) 163–173
100%
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
selectivity
selectivity
90%
80%
CO2
N2
N2
CO2
70%
60%
50%
40%
30%
20%
10%
0%
conversion
conversion
N2
CO2
0
5
10
15
20
25
0
5
10
15
20
Time [h]
Time [h]
Fig. 7. Performance of the Al2O3 sample, switching from CO2 to N2 (left) and from N2 to CO2 (right), compared with the performance in CO2 (left) or N2 diluent (right).
Conditions as described in Section 2.
amount of a theoretical monolayer of coke on the surface. Calcu-
lated from the unit cell dimensions, graphitic carbon has a single
side surface area of about 1300 m2/g. The fresh ␥-Al2O3 has a sur-
face area of 222 m2/g, so 0.17 g/g (15 wt%) of graphitic carbon is
needed to cover the surface with a single layer. For ZrO2 with a
specific surface area of 104 m2/g this amounts to 0.08 g/g (7.5 wt%).
Monolayer coverage of the initial material appears to be the optimal
coke loading.
Proper analysis of the gas stream by GC is an important aspect
of this work. Besides the common side products such as benzene
and toluene, we were also able to detect many higher aromatics
(naphthalenes, biphenyl, benzylbenzene, stilbene, and many more
unidentified) that we named tars, and surplus of hydrogen/carbon
monoxide. For instance for catalyst KV, omitting the selectivity to
tars (4%) and surplus H2 and CO (1.5%), would give a selectivity to
styrene of 97.5%. This clearly shows the importance of taking these
side products into account. The molar balance for surplus hydrogen,
based on the formation of coke and hydrogen from ethylbenzene, is
approximated by applying Eq. (9). Hydrogen is easily converted into
CO by the RWGS reaction, so these were always taken into account
together. It is assumed that the carbon from this reaction ends up
as coke on the catalyst. By integrating the amount of coke from
the surplus of H2 and CO, we obtain the same order of magnitude
of coke as what is determined by TGA. This demonstrates that the
carbon balance in this study seems well closed.
C8H10 → 8C + 5H2
(9)
Assuming that coke is the active species of the catalyst, other
carbon materials could also exhibit a good performance, from the
initial stage onwards. The experiments with carbon based materials
such as AC and CNT, shown in Fig. 2, also point in the direction that
carbon is responsible for the dehydrogenation activity and selec-
tivity. Although AC en CNT deactivate much faster than Al2O3 or
ZrO2, they show a constant high selectivity right from the start. The
1.2
AC
fresh
1
0.8
0.6
0.4
0.2
0
CNT
fresh
AC
spent
Al2O3 - 96h
Al2O3 - 20h
V-Al2O3 - 20h
CNT
spent
ZrO2- 16h
ZrO2
spent
Al2O3
spent
ZrO2- 2h
ZrO2- 0.5h
500
1000
1500
2000
300
400
500
600
700
Wave number [cm-1]
Temp [°C]
Fig. 8. Weight loss rate for the oxidation in air of the fresh and spent CNT and AC
samples compared with spent Al2O3 and ZrO2 samples as a function of temperature.
Fig. 9. Raman spectra of several spent samples obtained after different times on
stream.

C. Nederlof et al. / Applied Catalysis A: General 417–418 (2012) 163–173
171
Fig. 10. HR-TEM micrographs of ZrO2: A – fresh, B – spent 20 h.
difference in selectivity between samples might be explained as the
result of impurities in these materials, that were used as received.
The reason for faster deactivation is, however, unclear, as the mea-
sured specific surface area of the spent samples is still high. The
BET surface area, however, is not a measure for the accessible sur-
face area for ethylbenzene due to the presence of microporosity.
The large difference between the surface areas and pore volumes
of fresh and first spent sample for the alumina and zirconia series
in Tables 4–6, is attributed to easy blockage of the micropores by
the initial coke deposition and thereby becoming inaccessible.
The TGA profiles of the carbon materials in Fig. 8 show a clear
change in reactivity, showing that the type of carbon material
deposited is more reactive. The amount of material removed by TGA
in case of AC was 90.5% and 93.4% for the fresh and spent material,
respectively. This indicates that more carbon material was present
on the spent sample. With the CNT samples there was no significant
change in weight loss, the ash content is very low. Also the spent
CNT is more reactive in TGA, indicating large coke deposits.
Fig. 11. HR-TEM micrographs of Al2O3: A – fresh, B – spent 22 h, C – spent 96 h.
The behaviour of the carbon deposits on alumina in CO2 and
N2 diluent shows similarities with oxidative dehydrogenation cat-
alysts that use molecular oxygen. In the presence of oxygen,
carbon deposits and other forms of structured carbons like car-
bon nanotubes and onion-like carbon have been identified as active
materials. The actual catalytic site under molecular oxygen condi-
tions is believed to be oxygen-surface groups like quinones and
hydroxyls [2,40–51]. However, it is reasoned that in the absence of
molecular oxygen these groups cannot be formed and that hence
coke cannot be the active phase [2]. This reasoning seems to break
down as we clearly see a catalytic effect of the carbon deposits
under CO2 and N2 flow (Fig. 4), although a higher temperature is
required. Even in the absence of any oxidant, as the N2-experiment
shows, carbon deposits catalyse the ethylbenzene dehydrogena-
tion reaction.
More examples of the beneficial effects of coke deposition in
catalysed reactions exist, including butane skeletal isomerisation,

172
C. Nederlof et al. / Applied Catalysis A: General 417–418 (2012) 163–173
alkylation of toluene with methanol to form xylenes, catalytic
cracking, hydrotreating, and aromatization [38,52–55]. Even other
direct dehydrogenation reactions are known to proceed over car-
bon deposits, like with cyclohexene [56] and butane [57]. Going
back even further in time also direct ethylbenzene dehydrogena-
tion was already known to be catalysed by coke [58–62]. Sato et al.
[6] published comparable results showing increasing conversion
and selectivity with time on stream, but carbon was not identified
as the key in the performance.
redox function on the catalyst is needed for RWGS activity, but this
function will also influence the coke formation and, thereby, the
catalyst deactivation.
The gasification of coke with CO2, the reverse Boudouard reac-
tion, was not observed in our experiments. The carbon monoxide
product that was measured can be explained by the RWGS alone,
where hydrogen is produced from dehydrogenation or coke forma-
tion. A simple TGA experiment with a coked sample such as a spent
alumina support and CO2 as the oxidant showed that temperatures
higher than 650 ◦C are required for gasification of the coke by CO2.
The selective dehydrogenation activity of a catalyst under CO2
or N2 atmosphere is clearly correlated to the formation of carbon
deposits on alumina and zirconia. The actual activity of the cata-
lyst is a result of the sensitive balance between the amount of coke,
the characteristics of the parent material, and to a lesser extent
the available pore volume and surface area. For a stabilised catalyst
system, a stationary amount of coke has to be present. This can be
achieved by avoiding the excessive formation of coke, preferably
by reducing the rate of formation or otherwise by promoted coke
removal/gasification. The latter could be possible by addition of a
more oxidative gas to the reaction mixture, removing excess carbon
and creating the radical edge sites needed for the dehydrogenation.
The reactivity of the oxidant dictates its concentration and the tem-
perature level on which a stable steady state is reached. Without an
activating oxidant mainly thermal activation occurs, but inherently
a faster deactivation takes place.
As to the active site or mechanism of dehydrogenation on carbon
deposits, this is still not resolved. Good characterization of the car-
bon deposits is a challenge. Also the characterizations presented
in this paper do not provide a unambiguous explanation of the
catalytic activity. The Raman spectra do not provide any indica-
tion (Fig. 9). The ratio between the ‘graphitic’ carbon (∼1600 cm−1
)
and the ‘amorphous’ carbon (∼1350 cm−1) is apparently not a mea-
sure. In the range of 100–2000 cm−1 no other peaks were observed
than those attributed to carbon deposits, even though vanadium
oxides and zirconia have very distinct peaks in this region of Raman
spectroscopy. Electron spin resonance and THz-time domain spec-
troscopy have been used to characterise the carbon [57,58,61], but
information from these techniques is also limited. The intensity
of the signal increases with carbon content and activity of the
catalyst, leading to believe that unpaired electrons play a role in
the dehydrogenation reaction. This is also observed with oxida-
tive dehydrogenation [48]. Such free radicals can perform hydrogen
atom abstraction, and successive hydrogen atom abstraction will
lead to dehydrogenation [56–59,61]. The difference between direct
dehydrogenation, CO2 oxidative dehydrogenation, and oxidative
dehydrogenation may be only limited by the way of hydrogen rad-
ical removal: by formation of dihydrogen, or water formation via
RWGS, or oxidation, respectively. In the case of CO2, a less acidic
support can improve adsorption properties and, thereby, RWGS
reactions and increase dehydrogenation activity, as was reported
by Sato et al. [6].
5. Conclusion
This study with ZrO2 and Al2O3 supports for the dehydrogena-
tion reaction of ethylbenzene to styrene in CO2 shows that with
both systems carbon deposits are responsible for the activity and
selectivity. These bare catalysts show an optimal activity after 1 h
and 15 h of time on stream respectively, which coincide with reach-
ing a stable styrene selectivity plateau of 92%. Also with Al2O3
support and N2 as diluent the same phenomenon is observed. The
optimum performance in this case is at 10 h of time on stream. Char-
acterizations of the Al2O3 samples after different times on stream
show a clear correlation between the formed coke and the activity
and selectivity.
The vanadium and chromium supported catalysts act in the
same manner as the bare Al2O3 and ZrO2 supports, only the rates
of activation and deactivation are higher. The amounts of coke on
these catalysts support that carbon deposits are the active and
selective ingredient in those catalyst formulations. The advantage
of the supported catalysts is their RWGS activity that gives a minor
increase in styrene yield.
The coke that is initially formed during reaction is responsible
for the increased activity and selectivity for the dehydrogena-
tion reaction. After a certain time on stream the increase in the
amount of coke, however, reduces the activity due to a loss in avail-
able surface area. Oxygen groups on the coke seem not essential
for dehydrogenation, but rather defects and edges of the carbon
deposits play a role in the reaction mechanism. The overall activity
and selectivity of the catalyst depends on its amount of coke, parent
material, and accessible surface area.
The results with alumina and the use of nitrogen as diluent
shows that carbon dioxide, or another oxidant, is not necessary
to achieve a good dehydrogenation performance. Only a certain
amount of carbon deposit, enough to create a surface monolayer,
seems to be required. Oxygen groups on the coke surface are also
not necessary. Defects and edges of the coke deposits are more
likely to be part of the active site. Differences in the initial per-
formance development (the first 10–15 h on stream, up to the
optimum) between CO2 and N2 diluent can be explained by com-
petitive adsorption in the case of CO2. Here CO2 could block sites
that otherwise participate in the formation of coke or other side
reactions. This causes slower coke formation in CO2, and as coke
(below monolayer) makes the catalyst more active and more selec-
tive, the reduction of by-product formation is also slower. In the
experiment where we switched from CO2 to N2 diluent or the other
way around during dehydrogenation (Fig. 7), the conversion and
selectivity are identical after initial monolayer coke laydown. This
shows that the coke is active and of similar composition, indifferent
from the used diluent.
From the perspective of CO2 ODH, where the RWGS reaction is
used to elevate the equilibrium conversion, carbon and bare alu-
mina samples exhibit hardly any RWGS activity. From the bare
supports, only ZrO2 shows some RWGS activity, because of its redox
properties [63]. These redox properties are also not active enough
to reach the conversion levels that would make this process com-
mercially interesting. The (formed) carbon is inactive for the RWGS
reaction. The six supported catalysts show a better RWGS activity,
the best is KV with a CO2 conversion of 15%, decreasing to 5% as sta-
bility is an issue with these catalysts. Carbon dioxide is fed in large
excess. Initial CO2 conversions are higher than those expected due
to high hydrogen formation rates from the coking reaction (9). A
Acknowledgements
Dr. Patricia J. Kooyman is greatly acknowledged for the HR-TEM
analysis. This work was granted by STW (Foundation for Applied
Sciences) in the Green & Smart Process Technologies Program
under number 07983; entitled: CO2 oxidative dehydrogenation for
hydrocarbons to olefins and styrene production. The authors want

C. Nederlof et al. / Applied Catalysis A: General 417–418 (2012) 163–173
173
to thank Lummus Technology, a CB&I company, for their financial
contribution.
[30] N. Mimura, I. Takahara, M. Saito, Y. Sasaki, K. Murata, Catal. Lett. 78 (2002)
125–128.
[31] G. Carja, R. Nakamura, T. Aida, H. Niiyama, J. Catal. 218 (2003) 104–110.
[32] X.N. Ye, N. Ma, W.M. Hua, Y.H. Yue, C.X. Miao, Z.K. Xie, Z. Gao, J. Mol. Catal. A:
Chem. 217 (2004) 103–108.
[33] D.R. Burri, K.M. Choi, J.H. Lee, D.S. Han, S.E. Park, Catal. Commun. 8 (2007) 43–48.
[34] B.S. Liu, G. Rui, R.Z. Chang, C.T. Au, Appl. Catal. A: Gen. 335 (2008) 88–94.
[35] Y.Y. Qiao, C.X. Miao, Y.H. Yue, Z.K. Xie, W.M. Yang, W.M. Hua, Z. Gao, Microp-
orous Mesoporous Mater. 119 (2009) 150–157.
[36] G.R. Meima, P.G. Menon, Appl. Catal. A: Gen. 212 (2001) 239–245.
[37] D. Espinat, H. Dexpert, E. Freund, G. Martino, M. Couzi, P. Lespade, F. Cruege,
Appl. Catal. 16 (1985) 343–354.
[38] P.G. Menon, J. Mol. Catal. A: Chem. 59 (1990) 207–220.
[39] T. Klimova, M.L. Rojas, P. Castillo, R. Cuevas, J. Ramirez, Microporous Meso-
porous Mater. 20 (1998) 293–306.
[40] A. Schraut, G. Emig, H.G. Sockel, Appl. Catal. 29 (1987) 311–326.
[41] G. Mestl, N.I. Maksimova, N. Keller, V.V. Roddatis, R. Schlogl, Angew. Chem. Int.
Ed. 40 (2001) 2066–2068.
[42] J.J. Delgado, R. Vieira, G. Rebmann, D.S. Su, N. Keller, M.J. Ledoux, R. Schlogl,
Carbon 44 (2006) 809–812.
[43] J.J. Delgado, D.S. Su, G. Rebmann, N. Keller, A. Gajovic, R. Schlogl, J. Catal. 244
(2006) 126–129.
References
[1] D.H. James, W.M. Castor, Ullmann’s Encyclopedia of Industrial Chemistry,
Wiley-VCH Verlag GmbH & Co. KGaA, 2002.
[2] F. Cavani, F. Trifiro, Appl. Catal. A: Gen. 133 (1995) 219–239.
[3] S.E. Park, S.C. Han, J. Ind. Eng. Chem. 10 (2004) 1257–1269.
[4] A.L. Sun, Z.F. Qin, J.G. Wang, Appl. Catal. A: Gen. 234 (2002) 179–189.
[5] Y. Sakurai, T. Suzaki, N. Ikenaga, T. Suzuki, Appl. Catal. A: Gen. 192 (2000)
281–288.
[6] S. Sato, M. Ohhara, T. Sodesawa, F. Nozaki, Appl. Catal. 37 (1988) 207–215.
[7] G.I. Fedorov, S.G. Sibgatullin, N.L. Solodova, R.I. Izmailov, Pet. Chem. 16 (1976)
157–162.
[8] S.B. Wang, Z.H. Zhu, Energy Fuels 18 (2004) 1126–1139.
[9] M.O. Sugino, H. Shimada, T. Turuda, H. Miura, N. Ikenaga, T. Suzuki, Appl. Catal.
A: Gen. 121 (1995) 125–137.
[10] N. Ikenaga, T. Tsuruda, K. Senma, T. Yamaguchi, Y. Sakurai, T. Suzuki, Ind. Eng.
Chem. Res. 39 (2000) 1228–1234.
[11] N. Mimura, M. Saito, Catal. Today 55 (2000) 173–178.
[44] D.S. Su, N. Maksimova, J.J. Delgado, N. Keller, G. Mestl, M.J. Ledoux, R. Schlogl,
Catal. Today 102 (2005) 110–114.
[45] D.S. Su, N.I. Maksimova, G. Mestl, V.L. Kuznetsov, V. Keller, R. Schlogl, N. Keller,
Carbon 45 (2007) 2145–2151.
[12] V.P. Vislovskiy, J.S. Chang, M.S. Park, S.E. Park, Catal. Commun.
227–231.
3 (2002)
[13] X.N. Ye, W.M. Hua, Y.H. Yue, W.L. Dai, C.X. Miao, Z.K. Xie, Z. Gao, New J. Chem.
28 (2004) 373–378.
[46] N. Keller, N.I. Maksimova, V.V. Roddatis, M. Schur, G. Mestl, Y.V. Butenko, V.L.
Kuznetsov, R. Schlogl, Angew. Chem. Int. Ed. 41 (2002) 1885–1888.
[47] A.E. Lisovskii, C. Aharoni, Catal. Rev. – Sci. Eng. 36 (1994) 25–74.
[48] R. Fiedorow, W. Przystajko, M. Sopa, J. Catal. 68 (1981) 33–41.
[49] L.E. Cadus, L.A. Arrua, O.F. Gorriz, J.B. Rivarola, Ind. Eng. Chem. Res. 27 (1988)
2241–2246.
[50] L.E. Cadus, O.F. Gorriz, J.B. Rivarola, React. Kinet. Catal. Lett. 42 (1990) 101–106.
[51] L.E. Cadus, O.F. Gorriz, J.B. Rivarola, Ind. Eng. Chem. Res. 29 (1990) 1143–1146.
[52] R. Le Van Mao, L.A. Dufresne, J.H. Yao, Y.L. Yu, Appl. Catal. A: Gen. 164 (1997)
81–89.
[14] A.L. Sun, Z.F. Qin, S.W. Chen, J.G. Wang, J. Mol. Catal. A: Chem. 210 (2004)
189–195.
[15] X.N. Ye, Y.H. Yue, C.X. Miao, Z.K. Xie, W.M. Hua, Z. Gao, Green Chem. 7 (2005)
524–528.
[16] S.W. Chen, Z.F. Qin, X.F. Xu, J.G. Wang, Appl. Catal. A: Gen. 302 (2006) 185–192.
[17] J.S. Chang, D.Y. Hong, V.P. Vislovskiy, S.E. Park, Catal. Surv. Asia 11 (2007) 59–69.
[18] J.N. Park, J. Noh, J.S. Chang, S.E. Park, Catal. Lett. 65 (2000) 75–78.
[19] J. Noh, J.S. Chang, J.N. Park, K.Y. Lee, S.E. Park, Appl. Organomet. Chem. 14 (2000)
815–818.
[20] J. Ren, W.Y. Li, K.C. Xie, Catal. Lett. 93 (2004) 31–35.
[21] D.R. Burri, K.M. Choi, D.S. Han, J.B. Koo, S.E. Park, Catal. Today 115 (2006)
242–247.
[53] G.C. Bond, Appl. Catal. A: Gen. 149 (1997) 3–25.
[54] S. van Donk, J.H. Bitter, K.P. de Jong, Appl. Catal. A: Gen. 212 (2001) 97–116.
[55] C. Glasson, C. Geantet, M. Lacroix, F. Labruyere, P. Dufresne, J. Catal. 212 (2002)
76–85.
[22] D.R. Burri, K.M. Choi, S.C. Han, A. Burri, S.E. Park, J. Mol. Catal. A: Chem. 269
(2007) 58–63.
[56] H. Amano, S. Sato, R. Takahashi, T. Sodesawa, PCCP 3 (2001) 873–879.
[57] J. McGregor, Z.Y. Huang, E.P.J. Parrott, J.A. Zeitler, K.L. Nguyen, J.M. Rawson, A.
Carley, T.W. Hansen, J.P. Tessonnier, D.S. Su, D. Teschner, E.M. Vass, A. Knop-
Gericke, R. Schlogl, L.F. Gladden, J. Catal. 269 (2010) 329–339.
[58] P.A. Berger, J.F. Roth, J. Phys. Chem. 72 (1968) 3186.
[59] J.B. Donnet, Carbon 6 (1968) 161.
[60] J.F. Roth, A.R. Schaefer, Catalyst and Catalytic Process, United States Patent
3,446,865 (1969).
[61] J.F. Roth, J.B. Abell, L.W. Fannin, A.R. Schaefer, Adv. Chem. Ser. (1970) 193–203.
[62] J.F. Roth, J.B. Abell, L.W. Fannin, A.R. Schaefer, Abstracts of Papers of the Amer-
ican Chemical Society, 1969, p. 90.
[23] D.R. Burri, K.M. Choi, S.E. Park, Adv. Nanomater. Proc. 124–126 (Pts 1–2) (2007)
1737–1740.
[24] D.R. Burri, K.M. Choi, D.S. Han, Sujandi, N. Jiang, A. Burri, S.E. Park, Catal. Today
131 (2008) 173–178.
[25] B.M. Reddy, H. Jin, D.S. Han, S.E. Park, Catal. Lett. 124 (2008) 357–363.
[26] B.M. Reddy, D.S. Han, N. Jiang, S.E. Park, Catal. Surv. Asia 12 (2008) 56–69.
[27] B.M. Reddy, S.-C. Lee, D.-S. Han, S.-E. Park, Appl. Catal. B: Environ. 87 (2009)
230–238.
[28] H.Y. Li, Y.H. Yue, C.K. Miao, Z.K. Xie, W.M. Hua, Z. Gao, Catal. Commun. 8 (2007)
1317–1322.
[29] Y. Sakurai, T. Suzaki, K. Nakagawa, N. Ikenaga, H. Aota, T. Suzuki, J. Catal. 209
(2002) 16–24.
[63] K. Tanabe, Mater. Chem. Phys. 13 (1985) 347–364.