Investigation of the selective sites on graphitic carbons for oxidative dehydrogenation of isobutane

Article abstract of DOI:10.1016/j.jcat.2009.08.005

Reaction network analysis of the oxidative dehydrogenation (ODH) reaction of isobutane over model carbon catalysts with tailored open-edge graphitic structure and quinone-type oxygenated functionalities was used to identify the selective pathways for the

Full text of DOI:10.1016/j.jcat.2009.08.005

Journal of Catalysis 267 (2009) 158–166
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Investigation of the selective sites on graphitic carbons for oxidative
dehydrogenation of isobutane
*
*
Hong Xie, Zili Wu, Steven H. Overbury, Chengdu Liang , Viviane Schwartz
Center for Nanophase Materials Sciences, Oak Ridge National Laboratory, Oak Ridge, TN 37831, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
Reaction network analysis of the oxidative dehydrogenation (ODH) reaction of isobutane over model car-
bon catalysts with tailored open-edge graphitic structure and quinone-type oxygenated functionalities
was used to identify the selective pathways for the formation of isobutene. Carbon-based materials have
been widely used in catalysis, but the active sites are not well-understood due to the complexity of the
carbon structure. Correct identification of these sites is essential for learning how to manipulate material
structure to achieve high catalytic yields of the desired products. In this study, we created model catalysts
with controllable surface concentration of oxygen based on graphitized mesoporous carbon (GMC). Our
studies reveal that the ODH reaction of isobutane on carbon catalysts is a parallel-consecutive pathway
with partial oxidative dehydrogenation for the formation of isobutene and deep oxidation pathway for
the direct formation of CO and CO2 from isobutane. These two pathways show different dependence
on the quinone-type oxygen sites: the rate constant leading to the desired partial oxidation product does
not show a strong correlation to the density of the oxygen sites, whereas the rate constant leading to the
unselective COx products increases continuously with the density of oxygen sites.
Received 26 June 2009
Revised 10 August 2009
Accepted 10 August 2009
Available online 10 September 2009
Keywords:
Carbon catalysts
Oxidative dehydrogenation
Isobutane
Isobutene
Graphitized mesoporous carbon
Edge sites
Quinone sites
Ó 2009 Elsevier Inc. All rights reserved.
1. Introduction
leading to their activity can also be detrimental to the elucidation
of the active sites responsible for high selectivity.
Isobutene is the most versatile chemical intermediate among all
C4 olefins. The demand for isobutene has been increasing continu-
ously over the years due to its importance as a monomer in the
synthesis of various organic products for chemical and polymer
industries. Conventional dehydrogenation route using metal cata-
lysts for light olefin synthesis usually involves a high temperature
endothermic process which has several disadvantages such as cat-
alyst deactivation by coke formation and the consequent need for
periodic catalyst regeneration [1]. Alternatively, oxidative dehy-
drogenation (ODH) reaction of paraffins to their corresponding ole-
fins has received increasing attention since, as an exothermic
process, ODH reaction requires much less energy consumption
than dehydrogenation process which is commonly adopted in cur-
rent industrial operations.
Most of the studies on ODH reactions have been done using
transition metal oxide catalysts such as Mg–V–Sb-oxide, vanadium
oxide, and manganese molybdate [1–3]. One of the major chal-
lenges of this kind of reaction is to be able to minimize the com-
bustion side reactions leading to the undesired CO and CO2
products while increasing the limited alkene selectivity achieved
by these catalysts. The chemical complexity of the oxide catalysts
Carbon-based catalysts have proven to be attractive alterna-
tives to the conventional metal-based ones. Carbon materials
have significant advantages due to their unparalleled flexibility
as their physical and chemical properties can be tailored for
specific needs. So far, activated carbon, carbon nanofibers, and
carbon nanotubes have been reported to catalyze ODH reactions
with high selectivity results. Delgado et al. [4] reported catalytic
activity of carbon nanofibers immobilized on carbon felt support
for the ODH reaction of ethyl benzene. Díaz Velásquez et al. [5]
tested the surface-modified activated carbon for the isobutane
ODH reaction, and considered the carbonyl/quinone groups on
the surface as the active sites, where the isobutene yields were
lower with lower concentrations of these groups. Due to the
diverse nature of the structure and functionalities present in dif-
ferent carbon materials including the traditional activated car-
bons and the carbon nanotubes, the active sites of ODH
reactions on carbon materials leading to the formation of the de-
sired unsaturated products still have not been clarified. The pre-
cise mechanism of the ODH reaction over carbon surface
remains controversial due to undetermined metal impurities as
found in activated carbons and carbon nanotubes, dangling bonds,
mixed sp2 and sp3 hybridization, surface functional groups, and
the defect/edge sites, which could all contribute to the reactivity
with different selectivity.
* Corresponding authors.
E-mail addresses: liangcn@ornl.gov (C.D. Liang), schwartzv@ornl.gov (V. Schwartz).
0021-9517/$ - see front matter Ó 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2009.08.005

H. Xie et al. / Journal of Catalysis 267 (2009) 158–166
159
In order to identify the active sites of the ODH reaction on car-
bon catalysts, model compounds have been used to simplify the
catalytic system and thereby leading to better understanding of
the catalytic properties of carbon catalysts. A recent publication
by Zhang et al. [6] used surface-modified carbon nanotubes as
model catalysts and discovered that the nature of the surface oxy-
gen species has a strong correlation with the selectivity of the cat-
alysts: electrophilic oxygen species (converted from oxygen
molecules on the defect/edge sites) led to unselective total com-
bustion of n-butane and nucleophilic-oxygenated groups led to
the preferential ODH reaction that formed butaldiene. More re-
cently, we demonstrated that the graphitized mesoporous carbon
(GMC) with controllable openness of the graphite edge sites is an
excellent model catalyst for the study of active sites in the ODH
reaction on carbon catalysts [7]. The open graphitic edges in the
GMC catalyst have been identified as the active sites for the ODH
of isobutane. Interestingly, the initial existence of the oxygenated
functional groups, mainly carbonyl groups, on the open graphitic
edges of GMC catalyst shows no effect on its catalytic performance.
The GMC catalysts with or without initial oxygen functional groups
show negligible differences in the selectivity and reactivity. A plau-
sible reason for this observation is that the open edge sites on the
graphitic carbon are very active and the oxygenated functional
groups are formed promptly when the reactant oxygen is intro-
duced into the system. Indeed, oxygenated functional groups were
observed after the ODH reaction even for the GMC catalyst without
initially attached oxygenated groups. Although the edge sites are
unambiguously the active sites for the ODH reaction, the role of
the oxygenated functionalities that are either initially attached
on the carbon catalysts surface or generated during the reaction
has not been systematically studied on model catalysts. Revealing
the role of these oxygenated functionalities in the ODH reaction is
crucial for the fundamental understanding of the ODH reaction
mechanism of hydrocarbons over carbon catalysts. Reaction kinetic
analysis on carbon catalysts has been rarely studied, particularly
on designed model catalysts. Therefore, a detailed kinetic study
of the ODH reaction on model catalysts with intentionally gener-
ated functionalities will be valuable for gaining such a fundamental
understanding of ODH reaction on carbon catalysts.
37 wt% aqueous solution (10.0 g) were mixed in an round-bot-
tomed flask and dissolved in 1300 ml of anhydrous ethanol. The
resulting mixture was then heated with vigorous stirring until re-
flux. An aqueous solution of 26.0 g of formaldehyde 37 wt% was
added to the reaction mixture. The heating and stirring were con-
tinued for additional 2 h. Orange-colored particles were collected
after the reaction and were further pyrolysized at 850 °C in a tube
furnace (Thermolyne, model: 79300) in the presence of nitrogen.
The resulting carbon materials were then treated in helium at
2600 °C in a graphite furnace (Thermal Technologies, model:
1000-2560-FP20) to reach the starting material of pristine gra-
phitic mesoporous carbon denoted as sample SK0.
In our previous study we found out that our GMC pristine pre-
cursor can be homogeneously oxidized by air at 500 °C [7]. At this
condition, we ensured the integrity of the graphitic structure while
opening the fullerene-like cavities and creating open edges on its
structure. Additionally, oxygen functional groups were created
during the oxidation in air. In this study, the burn-off (B.O.) level,
defined as the weight loss percentage after the oxygen treatment,
and the amount of oxygenated groups were systematically varied
by changing the duration of the oxidation treatment. The GMC
samples prepared by this systematic approach were labeled as
SK1, SK2, SK3, and SK4 corresponding to the oxidation treatment
of 24 h, 30 h, 36 h, and 48 h, respectively.
2.2. Characterization of the physical structure of GMC catalysts
The nanometer-sized graphite crystals were characterized by
Raman spectra, which were collected via fiber optics connected di-
rectly to the spectrographic stage of a triple spectrometer (Prince-
ton Instruments Acton Trivista 555). Edge filter (Semrock) was
used in front of the UV–Vis fiber optic bundle (Princeton Instru-
ments) to block the laser irradiation. The 532 nm excitation was
emitted from a solid state laser (Princeton Scientific, MSL 532-
50) and the power was about 20 mW at the sample. The sample
sat on a XY stage (Prior Scientific, OptiScan XY system) and trans-
lated in raster mode while collecting the spectrum in order to elim-
inate/minimize any laser damage of the samples. Cyclohexane
solution was used as a standard for the calibration of the Raman
shifts.
We report herein a detailed study on the reaction kinetics on
carbon catalysts with tightly controlled surface functionality to
gain insights on the reaction mechanism of ODH reactions. The
GMC is an appealing model catalyst because of its simplicity in
structure and surface functionality. In this study, we created syn-
thetic carbon catalysts with controlled surface concentration of
oxygenated functionalities based on GMC. These compounds have
been investigated as model carbon catalysts for the ODH reaction
of isobutane to reveal the reaction network at differential reaction
conditions. We validate that the ODH reaction on carbon catalysts
is a multistep parallel-sequential reaction path which has been
thoroughly demonstrated over oxide catalysts [1,3,8,9]. Further-
more, the derived reaction rate constants for each step clarified
the factors leading to superior selectivity to isobutene on these
model carbon catalysts.
The porous structures of the model catalysts were characterized
by physisorption. N2 adsorption and desorption isotherms were
measured in liquid N2 at 77 K using an AUTOSORB-1C instrument
(Quantachrome Corporation, USA). The samples were outgassed
at 200 °C for 2 h prior to the isotherm measurement. The Bru-
nauer–Emmett–Teller (BET) specific area, SBET, was calculated from
P
the relative pressure 0:05 < P < 0:3 interval; and the total pore
0
P
P0
volume, Vtotal, at ¼ 0:95. The pore size distribution was calcu-
lated based on Barrett–Joyner–Halender (BJH) method using
P
0
0:05 < P < 0:94 interval of the adsorption branch of the isotherm.
2.3. Probing surface functionalities
The surface-oxygenated functionalities before and after reac-
tion were characterized by Temperature-Programmed Desorption
(TPD) using a U-tube reactor (Altamira AMI-200) in flowing helium
(20 ml/min) with a heating rate of 10 K/min from 298 K to 1300 K.
The amounts of CO and CO2 desorbed were quantified by a quadru-
pole mass spectrometer equipped with a 1 m long gas sampling
capillary (Pfeiffer-Balzer Omnistar).
2. Experimental
2.1. Synthesis of model catalysts
Pluronic surfactant F127, phloroglucinol, hydrochloric acid
(37 wt%), and formaldehyde (37 wt%) were obtained from Sigma–
Aldrich, and anhydrous ethanol was purchased from Pharmco Aa-
per. The preparation of mesoporous carbon precursors was carried
out following a procedure established by our group [7,10]. Briefly,
F127 (50.4 g), phloroglucinol (25.2 g), and hydrochloric acid
2.4. Catalytic measurements
The catalytic performance test for isobutane ODH reaction was
carried out at atmospheric pressure in a packed bed stainless steel
autoclave reactor (PID Eng & Tech, Spain). The feed consisted of iso-

160
H. Xie et al. / Journal of Catalysis 267 (2009) 158–166
butane and oxygen at 2.02 and 1.01 kPa, respectively, with balance
N2. The gases 2% O2 (balance N2) and 4% i-C4H10 (balance N2) were
purchased from Air Liquid.
D
An amount of 200 mg catalyst was used to yield a gas hourly
space velocity (GHSV) of 5100 hꢀ1. The reaction temperature was
set to 400 °C and controlled by a thermocouple located inside the
reactor bed. The reactants and products were analyzed using an
on-line gas chromatography conducted on an Agilent 6890N Gas
Chromatography system with a Hayesep N column and a molecu-
lar sieve column. CO and CO2 were converted to CH4 through a
methanizer before they passed through a flame ionization detector
for analysis. Isobutene, CO, and CO2 were the only products de-
tected. In all tests, carbon mass balances were within 100 1%.
The reaction network analysis was performed at differential
reaction conditions. Two different sets of experiments were con-
ducted at 400 °C: in the first set, for reaction rate dependence on
SK0
SK1
SK2
SK3
SK4
G
D'
1200
1300
1400
1500
1600
1700
1800
Raman Shift, cm-1
ꢀ
ꢁ
O2 partial pressure study, the isobutane partial pressure
P
i-C4H10
was kept at 0.45 kPa while the oxygen partial pressure (P ) was
O2
Fig. 1. Raman spectra of pristine GMC (SK0) and GMCs (SK1, SK2, SK3, and SK4)
ꢂ
ꢃ
PO2
PO2 þPi-C
with increasing burn-off levels.
varied to yield relative oxygen mole fraction
ꢁ 100%
H
4
10
between 20% and 80%. In the second set, for initial reaction rate
and selectivity study, Pi-C
and PO were kept constant at 2.02
4H10
oxidative gasification as a method for the creation of functionalities
on GMC in a controllable manner: (1) the integrity of the GMC
should be conserved to ensure that the gasification does not cause
the collapse of the porous structure; and (2) the nature of the sur-
face oxygen (type of functionality) should be independent of the
burn-off level of the GMC.
2
and 1.01 kPa, respectively. Isobutane conversion was varied by
changing the feed flow rate from 40 to 160 sccm with 200 mg of
carbon catalyst. The reaction rate and selectivity were extrapolated
to zero contact time in order to determine the rate constants of pri-
mary and secondary reactions. In all cases, the feed was diluted
with N2 and He, and the total isobutane conversion was controlled
below 5% to ensure differential conditions.
N2 physisorption confirms that the structural integrity of the
GMC is conserved with a burn-off level up to 70%. N2 adsorp-
tion–desorption isotherms of GMCs are shown in Fig. 2. These iso-
therms are type IV isotherms with H1 hysteresis that corresponds
to the uniform large mesopores formed through the polymer tem-
plates [7,10,13–15]. The specific surface area and pore volume of
the GMCs with different burn-off levels are given in Table 1. A sig-
nificant increase of the pore volume and surface area of SK1 was
observed when the pristine GMC (SK0) was slightly oxidized. This
abrupt increase of surface area and pore volume is mainly attrib-
uted to the opening of the voids in the fullerene-like cavities
caused by oxidative pitting at the defect sites as discussed in our
previous studies [7]. Further increase of the burn-off level gives a
steady and less sharp increase in the surface area and pore volume
of the GMCs due to the continuous decrease of the graphite nano-
crystal size caused by the oxidative process. The decrease of the
graphite nanocrystal size was verified in our previous work by X-
ray diffraction [7]. The XRD patterns before and after gasification
(not shown) presented similar features characteristic of graphitic
3. Results and discussion
3.1. Structure of and functional groups on the GMC model catalysts
As we discovered in our previous research, the secondary poros-
ity with a pore size of 2.4 nm in the GMC can be opened through air
oxidation [7]. In this process, a portion of the carbon, most likely the
defects on the GMC, is gasified. This gasification process results in
open graphite edges that can be covered with oxygen-containing
groups. These open edges are more active than the basal plane of
the graphite; therefore, higher burn-off level continues the gasifica-
tion of the edge sites and leads to the shrinking of graphite nanod-
omains in the a, b plane. Consequently, the La value, defined as the
size of the a, b plane (basal plane), is expected to decrease in the
GMC with high burn-off level. Tuinstra and Koening studied the size
dependence of the graphite nanocrystals to the ratio of the D- and
G-band intensities in the Raman spectrum of carbon [11]. The TK
equation describes that the ratio of intensities of the D- to G-band
is proportional to the inverse of La. The Raman spectra of the GMCs
with various burn-off levels are shown in Fig. 1. The spectra are nor-
malized to the G-band intensity to highlight the trend of the D-band
intensity. Obviously, the D-band of the GMC increases with the in-
crease of the burn-off level. Thus, the La value of the GMC decreases
with the burn-off level. The gasification of the GMC is along the
direction of the a, b plane, namely, the basal plane. This observation
indicates that the ratio of the edge sites to the basal plane sites pro-
gressively increases with the oxidative gasification of the GMC. In-
crease of the edge sites with burn-off level is also validated by the
work derived by Ferrari et al., [12] in which a relationship was
found between the ratio of intensities of the D- to G-band and sp3
content in amorphous carbon. In our previous study, we found that
the basal plane is inert for ODH reaction; and the surface oxygen is
attached to the edge plane. Hence, the oxidation of GMC with con-
trolled burn-off level could possibly generate surface oxygen at
continuously adjustable concentration on the edge surface of
GMC. Two additional conditions have to be met to validate the
SK0
SK1
800
SK2
SK3
SK4
600
400
200
0
0.0
0.2
0.4
0.6
0.8
1.0
Relative Pressure, P/P
0
Fig. 2. N2 adsorption–desorption isotherms for GMCs at 77 K.

H. Xie et al. / Journal of Catalysis 267 (2009) 158–166
161
Table 1
6.0
BET surface area and pore volume for GMCs.
GMC
B.O. (%)
SBET (m2/g)
Vtotal (cm3/g)
5.0
SK1
SK2
SK3
SK4
SK0
SK1
SK2
SK3
SK4
0
13
20
33
70
311
629
717
784
832
0.54
0.83
0.96
1.22
1.44
4.0
3.0
2.0
material [7], confirming the integrity of the GMC structure
throughout the oxidation process.
1.0
The pore size distributions of the GMCs (Fig. 3) further confirm
that the oxidative gasification process, although destructive, has no
effect on the integrity of the framework of the GMCs. All samples
show bimodal distributions of mesopores with a lower peak cen-
tered at 2.4 nm for all carbons and a higher peak centered at about
7.6 nm for SK0 and SK1 carbons, 8.1 nm for SK2 carbon, and 8.8 nm
for SK3 and SK4 carbons, respectively. While the small pores had a
narrow distribution between 2.5 and 4.5 nm for all these carbons,
the large pore size distribution became broader with increasing
B.O. (Fig. 3), which suggests that further treatment in air for ex-
tended time at 500 °C slightly expands the large mesopores with
less influence over the small mesopores in these carbons.
The surface oxygen functional groups were examined by TPD.
The similar curve profiles of all the carbon catalysts suggest that
the oxidation of the GMCs developed the same type of oxygenated
functional groups on the surface of GMCs. While their surface con-
centration increases with the increase of the burn-off level, the
composition of the oxygenated groups remains the same for all lev-
els of burn-off. The CO desorption profile for the GMCs is shown in
Fig. 4a. A pronounced peak at about 1023 K was observed, which
can be associated to phenol, quinone or semiquinone groups
formed on zig-zag edges [16–18]. Quinone-type surface groups
have been reported as the active sites for ODH reaction [19]. For
the CO2 desorption shown in Fig. 4b, two pronounced peaks at
about 548 K and 923 K were observed, which should correspond
to the lactone groups [17]. In addition, a shoulder at ꢂ833 K was
also observed, which corresponds to anhydride groups [20]. No
CO or CO2 desorption from the untreated precursor material, SK0
carbon, was detected, which confirms a clean and inert surface
for the pristine graphitic carbon.
0.0
700
800
900
1000
1100
1200
1300
Temperature, K
Fig. 4a. CO desorption spectra during TPD of fresh GMCs.
4.0
3.0
2.0
1.0
0.0
SK1
SK2
SK3
SK4
400
500
600
700
800
900 1000 1100 1200 1300
Temperature, K
Fig. 4b. CO2 desorption spectra during TPD of fresh GMCs.
Table 2
Calculated amounts of CO and CO2 desorbed during TPD (N:
l
mol/g).
P
Table 2 summarizes the amounts of CO and CO2 species evolv-
ing from GMCs through TPD. The data show that the desorbed CO
amount was one order of magnitude greater than that of CO2 for all
samples. The ratios CO/CO2 were essentially constant for these four
GMC
NCO
NCO
NO
NCO=NCO
2
2
SK1
SK2
SK3
SK4
923
96
1115
1343
1882
2261
892
9.6
10.5
8.8
10.6
10.8
10.5
9.9
1129
1534
1901
752
980
1129
1426
107
174
180
70
93
114
153
SK1a
SK2a
SK3a
SK4a
1166
1357
1731
0.014
SK0
9.3
0.012
0.010
0.008
0.006
0.004
0.002
0.000
SK1
SK2
SK3
SK4
a
Used carbon after reaction.
carbons averaging around 9.9, corroborating that the distribution
of surface-oxygenated functional groups is comparable on all
GMCs. Similar trends were also observed for used carbon catalysts
after ODH reaction (Figs. 5a and 5b); the ratios of CO/CO2 stayed
the same at around 9.9. The disappearance of the low temperature
CO2 desorption peak after reaction (Fig. 5b) is consistent with the
fact that these carbons were exposed to a temperature of 400 °C
during ODH tests. However, considering that the amount of CO2
desorbed was an order of magnitude smaller compared to that of
the CO, the contribution from these lactones to the ODH reaction
should be secondary.
0
50
100
150
Pore width,
200
250
300
The amount of desorbed COx decreased about 20% for each car-
bon after reaction. This could be due to the formation of unstable
Fig. 3. Pore size distribution for SK0, SK1, SK2, SK3, and SK4 carbons.

162
H. Xie et al. / Journal of Catalysis 267 (2009) 158–166
6.0
5.0
4.0
3.0
2.0
1.0
0.0
conclude that mainly quinone-type groups should be forming at
the surface based on the high temperature dry air oxidation pro-
cess, the pristine nature of the carbon precursor, the peak temper-
ature of CO elution, and the high value obtained for the CO/CO2
ratio. Although phenol groups could also desorb as CO at the ob-
served temperature range, the oxidation of GMC at 500 °C in dry
air excludes the formation of functional groups containing hydro-
xyl groups since there was no source of hydrogen from the precur-
sor synthetic carbon. Moreover, no H2O desorption was detected.
These quinone-type functional groups are continuously generated
on the GMC surface at all burn-off levels. As confirmed by the Ra-
man and physisorption measurements, these model catalysts share
the same framework structure with differences in the ratio of the
edge- to the basal-planes, which also results in the differences in
the concentration of the oxygenated functionalities. Therefore,
the controlled air oxidation of GMCs at 500 °C is a valid approach
for the synthesis of model catalysts with tunable concentration
of surface quinone-type functionalities.
SK1
SK2
SK3
SK4
700
800
900
1000
1100
1200
1300
Temperature, K
Fig. 5a. CO desorption spectra during TPD of used GMCs.
3.2. Catalytic performance and reaction pathway
4.0
A blank gas phase homogeneous reaction of isobutane with O2
was measured in the reactor packed with quartz beads and no con-
version of isobutane was detected under reaction conditions con-
firming the absence of such homogeneous reaction. To verify the
absence of mass transfer limitations, Thiele Modulus [22,23] were
calculated and values significantly inferior to 0.4 were obtained,
confirming the absence of mass transfer limitation on the meso-
porous carbons under reaction conditions. Under the presence of
the GMC catalysts, the main reaction products were isobutene,
CO, and CO2. Figs. 6 and 7 show comparisons of the specific rate
of isobutane consumed and isobutene selectivity versus time-on-
stream (TOS) for all the functionalized GMC catalysts, respectively.
After a short period of slight deactivation, all the carbons reached
steady-state and showed exceptional stability during 20 h of test.
Similar initial deactivation behavior has also been observed else-
where [5].
3.0
2.0
1.0
0.0
SK1
SK2
SK3
SK4
400
500
600
700
800
900 1000 1100 1200 1300
Temperature, K
During the 20 h TOS, the isobutane conversion increased and
the selectivity decreased with increasing B.O. level. The increase
in overall isobutane conversion suggests that more active sites
are available for GMCs with higher B.O. for the ODH reaction.
The higher consumption rate of isobutane was accompanied by
lower isobutene selectivity, which suggests a parallel-consecutive
reaction network as will be discussed later. Both CO and CO2 can
be formed by deep oxidation of isobutane and isobutene, and
Fig. 5b. CO2 desorption spectra during TPD of used GMCs.
O-containing surface groups on the carbons during the oxidation
treatment, that desorbed in the presence of isobutane during reac-
tion. It is also consistent with the observation that the reactivity of
ODH of isobutane in the first 10 h showed a slightly higher reactiv-
ity before it reached steady-state as discussed below. Nevertheless,
these GMCs had similar and well-developed surface oxygen-con-
taining groups which were considerably stable during the course
of ODH reaction.
1.2
The amount and the nature of the oxygen-containing surface
functional groups can vary over a wide range, and both depend
on the type of the oxidizing agent used, on the conditions of the
modification process, and on the nature of the carbon precursor
material. A variety of surface functional groups could be originated
from the precursor material and it is particularly true in the case of
carbons originated from oxygen-rich raw materials followed by
incomplete carbonization [20,21]. In our case, the use of a highly
pure carbon synthesized using an ultrahigh temperature process
can be very advantageous as it narrows down significantly the
range of possible functional groups present on the carbon surface
that undergoes further oxidative treatment. Certainly, the assign-
ment of the TPD peaks can be questionable as it is affected by many
factors such as carbon texture, heating rate, and experimental set-
up. However, by coupling our knowledge of the carbon material
and the synthesis process with the observed CO and CO2 peak tem-
perature profiles, we can limit even further the possible oxygen-
containing species being formed at the carbon surface. We
SK1
SK2
SK3
SK4
1.0
0.8
0.6
0
5
10
TOS, h
15
20
Fig. 6. Rate of isobutane consumption versus time-on-stream (TOS). Reaction
conditions: Pi-C
¼ 2:02 kPa, PO ¼ 1:0 kPa, T = 400 °C, catalyst weight: 200 mg.
4H10
2

H. Xie et al. / Journal of Catalysis 267 (2009) 158–166
163
100.0
95.0
90.0
85.0
80.0
75.0
900
3.0
2.0
1.0
0.0
SK1
SK2
SK3
SK4
SK4
800
SK2
SK3
SK1
700
600
surface area
500
oxygen density
SK0
400
300
0
5
10
15
20
0
10
20
30
40
50
60
70
TOS, h
B.O., %
Fig. 7. Isobutene selectivity versus time-on-stream (TOS). Reaction conditions:
Fig. 8a. Surface area and oxygen density as a function of B.O. for GMCs. The density
of oxygen sites was calculated taking into account the total amount of oxygen
desorbed during TPD (Table 2) and the total surface area of each carbon (Table 1).
Pi-C
¼ 2:02 kPa, PO ¼ 1:01 kPa, T = 400 °C, catalyst weight: 200 mg.
4H10
2
isobutene is a dehydrogenated intermediate product that can
either desorb as isobutene from carbon surface or be captured by
neighboring dissociated oxygen species to form CO or CO2.
1.0
0.9
0.8
0.7
0.6
0.5
0.4
100.0
90.0
80.0
70.0
60.0
50.0
40.0
The progressive changes in the ratio of edge to basal-planes and
the amount of surface oxygen groups deposited on the surface due
to the systematic burn-off process clearly causes changes in activ-
ity and selectivity. The effect of burn-off on the surface chemistry
of the carbons and their catalytic performance are given in Fig. 8
and Table 3. Fig. 8a shows how both the surface area and oxygen
density systematically varied by increasing the burn-off, whereas
Fig. 8b shows the changes in activity and selectivity with burn-
off. Table 3 presents the isobutane steady-state rate (defined at
15 h TOS) normalized on the basis of catalyst weight, total surface
area, and total oxygen content. It can be clearly observed that iso-
butane conversion, specific rate, and areal rate increase with burn-
off. However, there is a 1.9-fold increase of conversion and specific
rate from the lower burn-off catalyst (SK1) to the highest burn-off
catalyst (SK4), whereas the areal rate only presents a 1.45-fold in-
crease. It is important to emphasize that areal rate is based on the
total surface area, which includes both the inactive basal-planes
and the catalytic active edge planes. The isobutane consumption
normalized by oxygen content increases for the first two samples
and then decreases as the burn-off progresses. The small variation
of the rate with oxygen content could indicate that the total oxy-
gen content measured by TPD would be a proper basis for calcula-
tion of turnover rates. However, care should be taken as there are
most likely uncovered edge sites that can originate a different type
of active site. Indeed, it is well-known that not only the oxygen
functionality, but the defect edge sites of graphitic carbons contain
various active sites such as arm-chair and zig-zag sites with differ-
ent reactivities that can create sites for different oxygen species.
For instance, arm-chair sites have been reported to be less active
i-C4H10 consumption
i-C4H8 selectivity
10
20
30
40
50
60
70
B.O., %
Fig. 8b. Isobutane consumption rate and isobutene selectivity after 15 h time-on-
stream (TOS). Reaction conditions: Pi-C
catalyst weight: 200 mg.
¼ 2:02 kPa, PO ¼ 1:01 kPa, T = 400 °C,
4 H10
2
than zig-zag ones [18]. Therefore, a more detailed analysis of reac-
tion pathways and kinetics is necessary in order to draw conclu-
sions about the selective active sites for ODH reaction to see
whether they are related to the oxygen functionalities deposited
on the carbon edge surface or the dissociatively adsorbed O created
on the open edges and defect sites during reaction.
Oxidative dehydrogenation reaction of alkanes have been
extensively studied over oxide catalysts [1,3,8,9] and the reaction
sequence has been mostly described to involve parallel-consecu-
tive steps as illustrated for the case of ODH of isobutane in Scheme
Table 3
Conversion and reaction rates at steady-state (15 h TOS).
GMC catalyst
Conversiona (%)
Specific rateb (mol gꢀ1 sꢀ1
)
Areal ratec (mol mꢀ2 sꢀ1
)
Normalized rated (sꢀ1
)
SK1
SK2
SK3
SK4
1.8
2.5
3.2
3.4
4.83Eꢀ08
6.81Eꢀ08
8.57Eꢀ08
9.29Eꢀ08
7.68Eꢀ11
9.50Eꢀ11
1.09Eꢀ10
1.12Eꢀ10
4.33Eꢀ05
5.07Eꢀ05
4.55Eꢀ05
4.11Eꢀ05
a
b
c
Conversion: based on isobutane conversion.
Specific rate: based on isobutane consumption.
Areal rate: based on BET surface area.
d
Normalized Rate: based on the total oxygen content.

164
H. Xie et al. / Journal of Catalysis 267 (2009) 158–166
1. In this scheme, isobutene is a primary product of the oxidative
dehydrogenation whereas carbon oxides can form via either subse-
quent oxidation of isobutene or direct catalytic combustion of iso-
butane. This sequence is believed to occur also in carbon-based
catalysts during ODH of n-butane [6]. In order to verify the reaction
network occurring in our GMCs, a simple method based on an anal-
ysis of selectivity-conversion was carried out [9,24]. Primary prod-
ucts of a reaction can be discriminated from secondary or even
higher-order products by checking the zero-conversion intercept
of their selectivities. In this method, products with non-zero inter-
cepts are from the primary reactions.
ODH of ethylbenzene [19,27,28]. These studies suggest that the
activity increases with the degree of oxidation and that the qui-
none/carbonyl functional groups are the active sites for such reac-
tion. In more recent work, Zhang et al. [6] examined the ODH
reaction of butane on carbon nanotubes and, likewise, demon-
strated the same parallel-consecutive pathway as we observed.
Previously [7], we reported that the open edges of the fullerene-
like cavities are crucial for the catalytic activity of our highly stable
GMCs. If indeed the partial oxidation takes place over an oxygen
functional group, the formation rate of isobutene should present
a zero-order dependence on the O2 partial pressure PO as in a
2
Fig. 9 shows the selectivity-conversion plot for the SK2 catalyst
measured at 400 °C with an oxygen:isobutane ratio of 1:2. The
selectivity to isobutene increased when isobutane conversion de-
creased but did not reach 100% when extrapolated to zero conver-
sion. Selectivity analyses of CO and CO2 confirmed that both
species were also primary products of ODH of isobutene, as non-
zero intercepts were found for the selectivities at zero conversion.
Similar trends were also observed for the other carbons, SK1, SK3,
and SK4 (not shown), which further confirmed a parallel-consecu-
tive reaction network.
MVK-type mechanism. Fig. 10 shows the rate dependence for the
SK2 catalyst. As opposed to a zero-order dependence on PO , we
2
found that isobutene formation rate increased with PO and the
2
dependence could be fitted using a reaction order of 0.22 in PO
.
2
Partial order dependence is a strong indication that dissociative
adsorption of O2 is a step of the reaction mechanism leading to
the partial oxidation product.
3.3. Reaction rates: a selectivity study
The presence of two parallel-consecutive pathways is of partic-
ular interest as it suggests the existence of different surface sites,
responsible for the partial oxidation or deep oxidation products,
separately. Indeed, the involvement of distinct forms of reactive
oxygen species has been investigated especially in the case of oxide
catalysts [2,25]. In that case, it is generally believed that lattice
oxygen sites could be responsible for the partial oxidation follow-
ing a Mars–Van Krevelen (MVK) type of mechanism whereas the
deep oxidation could be carried out over the dissociatively ad-
sorbed oxygen species following a Langmuir–Hinshelwood (LH)
mechanism. However, little is known regarding the oxygen active
sites of carbon-based catalysts [26]. Most of the studies so far have
been carried out using complex activated carbons as catalysts for
A more detailed kinetic analysis can be performed by variation
of reactants flow and working at differential conversions. Follow-
ing the method described by Khodakov et al. [3], reaction rate
and selectivity for isobutene can be extrapolated to zero conver-
sion in order to determine the relative rates of the primary dehy-
drogenation and the parallel-consecutive oxidation reactions.
This method is based on assuming a first-order kinetic analysis of
the network and treating the concentrations of isobutane and oxy-
gen as constant at low conversion levels. The rates of each pathway
shown in Scheme 1 can be then simplified as follows:
0
r1 ¼ k1Pa
PbO ¼ k1Pi-C
;
;
ð1Þ
ð2Þ
ð3Þ
1
1
4H10
i-C4H10
2
0
r2 ¼ k2Pa
PbO ¼ k2Pi-C
2
2
4H10
i-C4H10
2
0
r3 ¼ k3Pa
PbO ¼ k3Pi-C
;
3
3
4H8
i-C4H8
2
where rn is the reaction rate per gram of carbon and kn is the first-
order rate coefficient for reaction n, n = 1, 2, 3.
The rate for isobutene formation can be written for a differential
weight (dW) and total volumetric rate (F) as
k1
i-C4H10
i-C4H8
k3
k2
F dPi-C
RT dW
4H8
¼ k1Pi-C
ꢀ k3Pi-C
:
ð4Þ
4H10
4H8
CO/CO2
Scheme 1. Reaction network in oxidative dehydrogenation of isobutane.
10.0
100.0
i-C H
4
CO
CO
8
2
96.0
92.0
88.0
1.0
6.0
4.0
2.0
0.0
Rate of isobutene formation
Fitting Function (y=axb b=0.22)
0.1
0.1
1
10
0.0
0.5
1.0
1.5
2.0
2.5
P
, kPa
Isobutane conversion, %
oxygen
Fig. 9. Isobutene, CO, and CO2 selectivity profiles in ODH of isobutane over SK2
Fig. 10. Isobutene formation rate as a function of PO over SK2 catalyst, reaction
2
catalyst at 400 °C, Pi-C
:PO ¼ 2 : 1, catalyst weight: 200 mg.
conditions: T = 400 °C Pi-C
¼ 0:45 kPa, catalyst weight: 200 mg.
4H10
4H10
2

H. Xie et al. / Journal of Catalysis 267 (2009) 158–166
165
100.0
Fig. 11a shows the experimental dependence between the rate of
isobutene formation and weight time (W/F). From the model de-
scribed by Khodakov et al. [3], it can be shown that at small W/F,
the rate can be expanded as a quadratic dependence upon W/F.
The value of k1 can be obtained from quadratic extrapolation for
W/F approaching zero (zero conversion) as shown in Fig. 11a. In this
case, k1 is calculated from
SK1
SK2
SK3
SK4
96.0
92.0
88.0
84.0
80.0
r
ðW=F¼0Þ
¼
k1
;
ð5Þ
0
P
i-C4H10
where r(W/F=0) is the extrapolated rate as W/F goes to zero and
P0i-C
is the initial partial pressure of isobutane.
4H10
Following the same methodology described by Khodakov et al.
[3], isobutene selectivity can be expressed by Eq. (6), as a linear
function of W/F.
0.00
0.05
0.10
0.15
0.20
0.25
0.30
-3
ꢄ
ꢅ
W/F, g.s.cm
k1
k1 þ k2
k1
k1 þ k2
k1 þ k2 þ k
W
F
Si-C
¼
ꢀ
3 RT
:
ð6Þ
4H8
2
Fig. 11b. Isobutene selectivity as
a
function of W/F:Pi-C
PO ¼ 1:01 kPa, catalyst weight = 200 mg. Conversions in all cases are less than 4%.
¼ 2:02 kPa,
4H10
2
Fig. 11b shows the linear fit of our experimental selectivity for
isobutene at variable weight time. Once k1 is known from the
above-mentioned equation, the rate constant k2 can be calculated
10.0
8.0
6.0
4.0
2.0
0.0
1.0
0.8
0.6
0.4
0.2
0.0
by the intercept of isobutene selectivity at zero weight time,
k
k1
k
k
1
2
3
which is
. Then k3 can be calculated from the slope of the fit-
ted line aks1þekx2 pressed by Eq. (6). The computed values of the three
rate constants are given in Fig. 12 for all the functionalized GMC
carbons.
The calculated reaction rate constants were plotted as a func-
tion of surface oxygen (Fig. 12). The rate constant leading to the
partial oxidation product does not show a strong correlation to
the density of the oxygen sites and, indeed, it remains fairly con-
stant above a certain level. On the other hand, the rate constant
leading to the total combustion increases continuously with the
density of oxygen sites. These differences appear to suggest that
the two reactions take place on different sites. However, different
from reported elsewhere [5,6,19,20], quinone-type functional
groups show only correlation to the total combustion pathway.
Additionally, the direct combustion pathway should play a less
important role since rate constant analyses show that COx were
formed mainly by the consecutive deep oxidation of isobutene.
It is still very hard to decouple the effect of the oxygen func-
tional sites to the number of defect edge sites. The air treatment
process utilized to create the oxygen sites on the mesoporous car-
bon surface was also responsible for the opening of the fullerene-
like cavities and the exposition of the edge sites. Evidently, more
1.4
1.6
1.8
2.0
2.2
2.4
2.6
-2
2.8
3.0
Oxygen surface density, μmol.m
Fig. 12. Dependence of the calculated rate constants of k1, k2, and k3, assuming a
first-order analysis, on the density of surface oxygen for different carbons.
than one type of sites are involved in the different pathways, which
is not surprising giving that others [29,30] have shown that on
oxide catalysts, the partial oxidation pathway is structure insensi-
tive and the total combustion is structure sensitive. Our studies
indicate that our highly stable GMCs modified by mild oxidation
suppress the direct total combustion pathway and, that, no clear
correlation was observed between the quinone-type functional
sites and the selective oxidation products.
1.5
SK1
SK2
SK3
SK4
1.3
1.0
0.8
0.5
4. Conclusions
In this study, we demonstrated that the controlled air oxidation
is a facile method for the preparation of model carbon catalysts
that contain a tunable amount of quinone-type oxygen functional-
ity on the surface of GMCs. The concentration of the oxygenated
functional groups on the surface of the GMCs is tunable in a contin-
uous manner through a simple control of the burn-off level. Sys-
tematically varying the concentration of the oxygenated
functional groups opens an avenue to investigate the function of
oxygenated functional groups in the ODH reactions with least
influence from the complexity of traditional carbon materials
which vary in composition of surface functional groups, the impu-
0.00
0.05
0.10
0.15
0.20
-3
0.25
0.30
W/F, g.s.cm
Fig. 11a. Rate of isobutene formation as a function of W/F: Pi-C
¼ 2:02 kPa,
4H10
PO ¼ 1:01 kPa, catalyst weight = 200 mg. Conversions in all cases are less than 4%.
2

166
H. Xie et al. / Journal of Catalysis 267 (2009) 158–166
rity of minerals, the sp2 and sp3 hybridization carbon atoms struc-
ture, and porous structures. Kinetic studies using these GMC cata-
lysts show the concurrence of a consecutive pathway for the partial
oxidative dehydrogenation for the formation of isobutene and sub-
sequent CO and CO2, and a parallel deep oxidation pathway for di-
rect formation of CO and CO2 from isobutane. Initial rate constant
analyses suggest that two different active sites appear to be
responsible for partial oxidative dehydrogenation and deep oxida-
tion processes separately. No clear correlation was observed be-
tween the quinone-type functional sites and the selective
oxidation product, which is different from the current commonly
accepted concept that only the surface oxygen-containing groups
are responsible for the ODH reactions on carbon catalysts. Given
the complexity of the carbon surface, although these model cata-
lysts have a simple surface structure, the edge planes are not com-
pletely covered by the oxygenated functionalities. The possible
involvement of the uncovered edge sites and defect sites in the
ODH reaction yields to uncertainties in probing the origin of the
selectivity of carbon catalysts for ODH reactions; further identifica-
tion of the uncovered edge sites and defects on the carbon catalysts
may lead to a thorough understanding of the ODH reaction mech-
anism and therefore enables the rational design of catalysts with
desirable properties.
References
[1] Y.A. Agafonov, N.V. Nekrasov, N.A. Gaidai, A.L. Lapidus, Kinet. Catal. 48 (2007)
255–264.
[2] J.N. Michaels, D.L. Stern, R.K. Grasselli, Catal. Lett. 42 (1996) 139–148.
[3] A. Khodakov, J. Yang, S. Su, E. Iglesia, A.T. Bell, J. Catal. 177 (1998) 343–351.
[4] J.J. Delgado, D.S. Su, G. Rebmann, N. Keller, A. Gajovic, R. Schlogl, J. Catal. 244
(2006) 126–129.
[5] J.D.D. Velasquez, L.A.C. Suarez, J.L. Figueiredo, Appl. Catal. A – Gen. 311 (2006)
51–57.
[6] J. Zhang, X. Liu, R. Blume, A.H. Zhang, R. Schlogl, D.S. Su, Science 322 (2008) 73–
77.
[7] C.D. Liang, H. Xie, V. Schwartz, J. Howe, S. Dai, S.H. Overbury, J. Am. Chem. Soc.
131 (2009) 7735–7741.
[8] S. Albonetti, F. Cavani, F. Trifiro, Catal. Rev. – Sci. Eng. 38 (1996) 413–438.
[9] J.N. Michaels, D.L. Stern, R.K. Grasselli, Catal. Lett. 42 (1996) 135–137.
[10] C.D. Liang, S. Dai, J. Am. Chem. Soc. 128 (2006) 5316–5317.
[11] F. Tuinstra, J.L. Koenig, J. Chem. Phys. 53 (1970) 1126.
[12] A.C. Ferrari, J. Robertson, Phys. Rev. B 61 (2000) 14095–14107.
[13] C.D. Liang, Z.J. Li, S. Dai, Angew. Chem., Int. Ed. 47 (2008) 3696–3717.
[14] S. Lowell, J. Shields, M. Thomas, TM, Characterization of Porous Solids and
Powders: Surface Area, Pore size, and Density, Springer, Netherland, 2004.
347p.
[15] M. Steinhart, C.D. Liang, G.W. Lynn, U. Gosele, S. Dai, Chem. Mater. 19 (2007)
2383–2385.
[16] U. Zielke, K.J. Huttinger, W.P. Hoffman, Carbon 34 (1996) 983–998.
[17] B. Marchon, W.T. Tysoe, J. Carrazza, H. Heinemann, G.A. Somorjai, J. Phys.
Chem. 92 (1988) 5744–5749.
[18] S.E. Stein, R.L. Brown, Carbon 23 (1985) 105–109.
[19] M.F.R. Pereira, J.J.M. Orfao, J.L. Figueiredo, Appl. Catal. A – Gen. 184 (1999)
153–160.
[20] J.L. Figueiredo, M.F.R. Pereira, M.M.A. Freitas, J.J.M. Orfao, Carbon 37 (1999)
1379–1389.
Acknowledgments
[21] V. Strelko, D.J. Malik, M. Streat, Carbon 40 (2002) 95–104.
[22] G.W. Roberts, C.N. Satterfield, Indus. Eng. Chem. Fundam. 4 (1965) 288–293.
[23] A. Dufour, A. Celzard, B. Ouartassi, F. Broust, V. Fierro, A. Zoulaljan, Appl. Catal.
A – Gen. 360 (2009) 120–125.
[24] N.A. Bhore, M.T. Klein, K.B. Bischoff, Ind. Eng. Chem. Res. 29 (1990) 313–316.
[25] R. Grabowski, Catal. Rev. 48 (2006) 199–268.
[26] F. Rodriguez-Reinoso, Carbon 36 (1997) 159–175.
[27] A. Guerreroruiz, I. Rodriguezramos, Carbon 32 (1994) 23–29.
[28] M.F.R. Pereira, J.J.M. Orfao, J.L. Figueiredo, Colloids Surf. A – Physicochem. Eng.
Aspects 241 (2004) 165–171.
This research was supported by the Center for Nanophase Mate-
rials Sciences, which is sponsored at Oak Ridge National Laboratory
by the Scientific User Facilities Division, Office of Basic Energy Sci-
ences, U.S. Department of Energy. The research was supported in
part by an appointment to the ORNL Postdoctoral Research Associ-
ates Program administered jointly by Oak Ridge Institute for Sci-
ence and Education (ORISE) and ORNL. James W. Klett and Jim
Kiggans at ORNL are gratefully acknowledged for their assistance
with the graphitization process.
[29] S.T. Oyama, J. Catal. 128 (1991) 210–217.
[30] A. Baiker, D. Monti, J. Catal. 91 (1985) 361–365.