Oxidative dehydrogenation of cyclohexane over alumina-supported vanadium oxide nanoliths

Article abstract of DOI:10.1016/j.jcat.2009.11.026

Featuring highly ordered one-dimensional nanopores, anodic aluminum oxide (AAO) makes an ideal substrate for fabrication of catalysts by atomic layer deposition (ALD). Vanadium oxide (VOx) catalysts supported on AAOs and prepared by ALD and incipient wetness impregnation are characterized by X-ray fluorescence and ultraviolet-visible (UV-Vis) spectroscopy. At low loadings (<4 V/nm²) the supported VOx are mostly isolated monomers; polyvanadate domains are gradually formed as the surface vanadium content increases. The catalytic performance at a series of loadings (<3-32 V/nm²), and hence different forms of VOx, for the oxidative dehydrogenation (ODH) of cyclohexane are investigated. Compared to the catalysts prepared by incipient wetness impregnation, the ALD VOx catalysts show specific activities that are between 2 and 7.5 times higher. This reflects a better dispersion of the catalytic species on the surface as synthesized by ALD. In the cyclohexane ODH reaction with the supported ALD VOx, the kinetic orders and activation energies are comparable to kinetics data reported previously for the supported VOx. The results indicate that the ALD technique can be applied as an alternative approach to synthesize the supported VOx catalysts and achieves very good dispersion even at loadings above one monolayer (8 V/nm²). In the ODH reaction, polyvanadate sites are shown to be more active, overall, than monovanadate sites. However, numerical modeling of the reaction pathways indicates that the olefin formation rate is ～3 times faster on monomeric VOx sites than on polymeric VOx. By comparing the ODH of cyclohexane and the oxidations of cyclohexene and benzene, we find that both the sequential path and the direct path (the direct conversion from cyclohexane to benzene) are important in the oxidation process of cyclohexane.

Full text of DOI:10.1016/j.jcat.2009.11.026

Journal of Catalysis 269 (2010) 421–431
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Oxidative dehydrogenation of cyclohexane over alumina-supported
vanadium oxide nanoliths
d
b
b
c
a,d,*
H. Feng , J.W. Elam , J.A. Libera , M.J. Pellin , P.C. Stair
a
Department of Chemistry and Center for Catalysis and Surface Science, Northwestern University, Evanston, IL 60208, USA
Energy Systems Division, Argonne National Laboratory, Argonne, IL 60439, USA
Materials Science Division, Argonne National Laboratory, Argonne, IL 60439, USA
b
c
d
Chemical Sciences and Engineering Division, Argonne National Laboratory, Argonne, IL 60439, USA
a r t i c l e i n f o
a b s t r a c t
Article history:
Featuring highly ordered one-dimensional nanopores, anodic aluminum oxide (AAO) makes an ideal sub-
strate for fabrication of catalysts by atomic layer deposition (ALD). Vanadium oxide (VOx) catalysts sup-
ported on AAOs and prepared by ALD and incipient wetness impregnation are characterized by X-ray
Received 24 November 2009
Accepted 27 November 2009
Available online 28 December 2009
2
fluorescence and ultraviolet–visible (UV–Vis) spectroscopy. At low loadings (<4 V/nm ) the supported
VOx are mostly isolated monomers; polyvanadate domains are gradually formed as the surface vanadium
content increases. The catalytic performance at a series of loadings (<3–32 V/nm ), and hence different
Keywords:
Oxidative dehydrogenation (ODH)
Cyclohexane
Vanadium oxides (VOx)
Anodic aluminum oxide (AAO)
Atomic layer deposition (ALD)
2
forms of VOx, for the oxidative dehydrogenation (ODH) of cyclohexane are investigated. Compared to
the catalysts prepared by incipient wetness impregnation, the ALD VOx catalysts show specific activities
that are between 2 and 7.5 times higher. This reflects a better dispersion of the catalytic species on the
surface as synthesized by ALD. In the cyclohexane ODH reaction with the supported ALD VOx, the kinetic
orders and activation energies are comparable to kinetics data reported previously for the supported VOx.
The results indicate that the ALD technique can be applied as an alternative approach to synthesize the
supported VOx catalysts and achieves very good dispersion even at loadings above one monolayer (8 V/
2
nm ). In the ODH reaction, polyvanadate sites are shown to be more active, overall, than monovanadate
sites. However, numerical modeling of the reaction pathways indicates that the olefin formation rate is
ꢀ
3 times faster on monomeric VOx sites than on polymeric VOx. By comparing the ODH of cyclohexane
and the oxidations of cyclohexene and benzene, we find that both the sequential path and the direct path
the direct conversion from cyclohexane to benzene) are important in the oxidation process of
cyclohexane.
(
Ó 2009 Elsevier Inc. All rights reserved.
1
. Introduction
Due to its favorable thermodynamics and resistance to catalyst
ciated electronic and chemical properties of the supported vanadia
species change progressively from isolated monomers, to two-
dimensional polymers, to three-dimensional crystals with increas-
ing surface coverage [6], the corresponding performance in pro-
deactivation from coke formation, the heterogeneous oxidative
dehydrogenation (ODH) of alkanes would be highly desirable for
the production of olefins [1,2]. Supported vanadium oxide (VOx)
catalysts have been intensively investigated for the ODH of ethane,
propane, and butane because of their high activity and selectivity
pane ODH is an example of
a catalytic structure-function
relationship. Due to the inhomogeneity of oxide supports and the
catalysts they carry, a detailed, atomic-level understanding of this
relationship is still lacking. For a given surface coverage the nature
of the supported vanadia species, and hence their properties, de-
pends strongly on the method of preparation. For example, the for-
[
2–5]. The type of support has been identified as the most impor-
tant factor in determining the catalytic performance of the sup-
ported VOx [5–8]. However, the activity and selectivity for
propane ODH were also shown to depend strongly on the vanadia
surface coverage [3,5]. Since the molecular structure and the asso-
mation of V
loadings of 3.6 V/nm and 8 V/nm in the preparation of VOx sup-
ported on -Al by impregnation using an aqueous NH VO [9]
and by the grafting of OV(OC dissolved in methanol [10],
2 5
O crystals was observed by Raman spectroscopy at
2
2
c
2
O
3
4
3
3 5 3
H )
respectively, due to the differences in the degree of dispersion
produced by the two preparation methods. At loadings higher than
one monolayer (ca. 8 V/nm on alumina) characteristic peaks of
*
Corresponding author. Address: Department of Chemistry and Center for
Catalysis and Surface Science, Northwestern University, Evanston, IL 60208, USA.
Fax: +1 847 467 1018.
2
2 5
crystallized V O appear in the Raman spectra of both the
E-mail address: pstair@northwestern.edu (P.C. Stair).
0
021-9517/$ - see front matter Ó 2009 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2009.11.026

4
22
H. Feng et al. / Journal of Catalysis 269 (2010) 421–431
impregnated and the grafted VOx [11]. The formation of V
2
O
5
crys-
breaking one C–H bond on each methylene group. Consequently,
a catalyst that is very active only for breaking C–H bonds will favor
benzene formation, while the same catalyst will only make COx
from propane. COx formation from cyclohexane requires breaking
of multiple C–H bonds on a single methylene, C–C scission, O-
insertion, or a radical process. The ability to form two different
hydrocarbon products in this reaction also makes it possible to
evaluate the contributions from a reaction path where the forma-
tion of benzene occurs during a single adsorption of the cyclohex-
ane reactant on the catalyst compared to a true sequential reaction
path that involves cyclohexene as the intermediate. These issues
tals dramatically (and trivially) lowers the specific activity of the
catalyst since, depending on the crystal size, a fraction of the vana-
dium is in the interior of the crystal and unable to participate in
surface reactions [12].
In recent years atomic layer deposition (ALD) has been shown
as an alternative method to synthesize supported catalysts [13–
1
5]. In the ALD process two types of gaseous precursors are pulsed
alternately, each reacting with the surface functional groups gener-
ated in the previous pulse to produce a deposition cycle. Since the
surface functional groups are regenerated at the end of each depo-
sition cycle, this technique is capable of coating as many layers of
catalytic material as desired. The precise layer-by-layer deposition
can effectively prevent aggregation of the catalytic species so that
the catalyst has 100% dispersion on the surface. The uniformity of
the catalytic materials synthesized by ALD reduces the typical
inhomogeneity of heterogeneous catalysts and thereby has the
potential to enhance our fundamental understanding of composi-
tion/structure/function relationships in supported heterogeneous
catalysts.
The work described here makes use of a structure which pro-
vides an exceptionally uniform platform for catalytic studies that
we have called the catalytic nanolith [16]. The nanolith is an array
of uniform, straight channels with controllable dimensions in the
range 10–100 nm. It can be coated with catalytic layers and sealed
in a flow reactor so that the reactants must pass through the cata-
lytic pores during reaction, and all the pores have the same dimen-
sions. An extremely uniform coating throughout the entire length
of the nanolith channels can be realized by the ALD technique
will be discussed further in terms of
relationship.
a structure–function
2. Experimental
The nanolith supports were prepared using the method previ-
ously described [16]. A two-step anodization of high purity alumi-
num disks (Alfa Aesar, 99.99%) was carried out in 0.3 M oxalic acid
(Fisher 99.9%) at a constant potential of 40 V, resulting in 40 nm
pores with an average pore to pore separation of 110 nm [22,23].
The thickness (pore length) of the nanolith, ꢀ70
lm in this case,
is determined by the duration of the second-step anodization.
The aluminum disks were masked during fabrication to preserve
an aluminum border around the perimeter of the nanolith. This en-
abled the nanoliths to be sealed in the reactor using commercial
compression fittings. Before any catalyst was loaded, the nanolith
2 3
pores were coated with 1 nm of Al O to completely cover electro-
[
17]. Thus, each channel has identical catalytic layers. Moreover,
lyte impurities incorporated into the oxide material during anod-
ization. Detailed descriptions about the introduction of coatings
in nanoliths can be found elsewhere [17]. In brief, alumina ALD
was carried out at 200 °C in a viscous flow reactor [24]. Ultra high
purity nitrogen (Airgas, 99.999%) was used as the purge gas. The
layer-by-layer growth of alumina was achieved by alternately
exposing the substrates to trimethyl aluminum (Aldrich, 97%)
and water for 8-cycles using a quasi-static reactor operation mode
in which the reactor is pumped down to 0.1 Torr after each dose
and purge.
there is no change in surface area due to pore blocking and negli-
gible change in pore diameter even at high catalyst loadings unlike
some previous studies [18]. Therefore, surfaces with well-dis-
persed catalytic species, at loadings up to one monolayer, and no
pore blockage are made possible by the combination of the nano-
lith and the ALD technique. Initial studies also indicate that the
nanolith structure minimizes the contribution of homogeneous,
radical reactions under ODH conditions and produces a higher
activity and selectivity to olefin than a comparable catalyst in the
form of a powder [16]. Finally, the nanolith structures used in
The VO
wet impregnation and ALD. For wet impregnation an aqueous
solution of NH VO (Aldrich, 99.99%) was dropped onto one side
x
catalysts were introduced onto the nanoliths by both
the present study are only 70
lm thick, making it possible to per-
form UV–Vis absorption measurements in transmission rather
4
3
than in diffuse reflectance.
of the nanolith and at the same time a slight vacuum was ap-
plied from the other side to draw the solution through the pores
and to assist the evaporation of liquid. To ensure a complete
The reaction studied here is the oxidative dehydrogenation of
cyclohexane. In contrast to ethane, propane, and butane, ODH of
cyclohexane has received much less attention. Lopez and cowork-
ers studied the performance of different groups of catalysts for this
reaction, including silica and niobia supported vanadia, V, Fe, and
Ce phosphates, and protonic and transition metal exchanged
ZSM-5 zeolites, finding that the major product of this reaction is
wetting of the pore walls by the liquid,
a dilute solution
ꢂ3
(3.2 ꢁ 10 mol/l) was used so that the volume of the solution
was much greater than that of the nanolith pores. The precursor
solution was applied from both sides of the nanolith alternately,
and the application and drying procedure was repeated several
times until the desired amount of vanadium was introduced.
The samples were first dried at 150 °C for 1 h and then calcined
in oxygen at 500 °C for 5 h.
2
either benzene or COx (CO and CO ) [19]. Usually benzene is not
considered to be a useful product in this reaction since the hydro-
genation of benzene is an industrial process for cyclohexane pro-
duction. The valuable product from this reaction is cyclohexene,
which is an intermediate in the production of adipic acid, the
monomer for Nylon 6, 6 [20]. A traditional way to produce cyclo-
hexene is by the dehydration of cyclohexanol, which is obtained
from the oxidation of cyclohexane in liquid phase [21]. A direct
conversion from cyclohexane to cyclohexene through a heteroge-
neously catalyzed process is highly desirable but no catalytic sys-
tem has shown a meaningful yield of cyclohexene. From a
fundamental point of view the study of cyclohexane ODH provides
more information about the catalytic reaction pathways than with
comparable studies using propane. With propane the formation of
COx is the inevitable result of breaking three or more C–H bonds.
This is not the case for cyclohexane where benzene is formed by
Using the ALD procedure, VOx was deposited in an integral
number of sub-monolayer cycles using vanadium oxytriisopropox-
ide (Aldrich, 99.98%)/(hydrogen peroxide + water) as the precur-
sors. A dose-purge-dose-purge timing sequence of 45–30–45–
30 s was used in each cycle to deposit layer(s) of VOx onto the
nanolith at 100 °C, 1.1 Torr, with
a nitrogen flow rate of
240 sccm/min [16]. Different numbers of cycles (1–18) of VOx
were deposited to make samples with a series of loadings. After
coating the catalytic materials, the samples were tested in the
catalysis experiments without further treatment. After initial catal-
ysis testing, the ALD samples were calcined in oxygen at 500 °C for
2 h and then re-tested under the same conditions to probe for
reproducibility and catalyst aging.

H. Feng et al. / Journal of Catalysis 269 (2010) 421–431
423
X-ray Fluorescence Spectroscopy (XRF) was used as a non-
destructive method to characterize the amount of vanadium sup-
ported in the nanolith. The spectra were measured with an Oxford
ED2000 instrument. The XRF data were calibrated by the results
from inductively coupled plasma (ICP) measurements taken with
a Varian Vista-MPX instrument. To prepare the samples for ICP,
the nanoliths were removed from the aluminum mounting ring
and dissolved in 10% phosphoric acid solution. The primary atomic
emission wavelength of vanadium at 292.4 nm was used for
analysis.
UV–Vis spectra of the catalysts were measured with a Cary 500
UV–Vis-NIR spectrometer under ambient conditions. Since the
nanoliths are transparent, the spectra were measured in transmis-
sion using a double beam configuration with a blank nanolith as
the reference. For each nanolith sample, UV–Vis spectra were re-
corded both before and after catalysis testing and after subsequent
calcinations.
nanolith, measured from both the pore geometry and by BET, is
ꢀ0.05 m [16]. The base material that forms the nanolith structure,
2
prior to ALD, is an amorphous alumina with impurity electrolyte
anions, such as oxalates and sulfates, incorporated into the alu-
mina framework during the anodization process [26]. TGA mea-
surements show that the decomposition temperatures of these
incorporated impurities are as high as 890–950 °C. Since these
temperatures exceed the melting point of the supporting alumi-
num ring, calcination is not a feasible way to remove the impuri-
ties. To avoid the possible interference from these impurities in
the catalytic reactions, the nanoliths are coated with 1 nm of pure
2 3
Al O by ALD to completely cover the incorporated anions. In blank
catalytic tests without VOx the ALD alumina-coated nanoliths
exhibited low, stable background reaction rates (0.1–0.5% conver-
sion from 400 °C to 480 °C). The diameter of the pores is reduced
from 40 nm to 38 nm by the ALD alumina coating. The correspond-
ing reduction in the total surface area is ꢀ5%.
The catalytic experiments were carried out in a flow reactor with
the nanolith sealed in a modified Swagelok VCR fitting [16]. Prior to
performing the catalytic experiments, the reactor was coated with
The aqueous impregnation of vanadium containing precursors
and subsequent calcination can form a two-dimensional mono-
layer structure of supported VOx on the surface of alumina, titania,
and niobia [3,4]. The monolayer loading of VOx depends on the
5
2 3
0 nm of Al O by ALD to reduce interference from reactions cata-
lyzed by the metal reactor itself. The nanolith seal was carefully
tightened to a force where the upstream gas pressure was stable
and no leak was detected. Quartz chips (14–30 mesh) were used
to fill the empty space in the reactor and minimize possible gas
phase reactions [2]. Cyclohexane vapor was introduced into the
reaction system using a helium gas bubbler. Oxygen (Airgas,
type of support. On c-alumina, this loading is reported to be
13 lmol/m or approximately 8 V/nm [3]. On the surface of Si
(0 0 1), the structure of the coated ALD alumina has been shown
2
2
to be amorphous [27]. Since the surface of amorphous alumina is
structurally similar to
c
-alumina [28], the value of 8 V/nm2 is
adopted as the monolayer loading when introducing VOx onto
9
9.99%) was used as the oxidant, and a mixture of methane and he-
the nanolith by wet impregnation.
lium (Matheson, 1.00:99.00) was used as the balance gas. With the
exception of experiments designed to determine reaction orders,
cyclohexane ODH reactions were performed with cyclohexane:oxy-
gen:balance gas compositions of 1:2:97 and 1:8:91. Reaction gas
mixtures of 1:8:91 were employed for comparing the ODH of cyclo-
hexane, cyclohexene, and benzene to ensure a low conversion
The amount of VOx loaded by ALD was first measured by the
non-destructive XRF technique. An accurate calibration of the
XRF signals using known standards was not available since it was
not possible to fabricate a series of homogeneous powder support
samples that had the same composition as the nanolith and a
known amount of VOx. Instead, ICP measurements of the nanolith
samples were used to provide a calibration for the XRF data. In
obtaining the XRF results, the vanadium surface concentrations
were calculated based on the measured amount of the VOx species,
and the total weight and surface area of the nanolith. The calcu-
lated vanadium surface concentrations are listed in Table 1.
The quantity of vanadium coated by each cycle of ALD is much
(
<10%) of oxygen and consequently to eliminate the possible influ-
ence of oxygen concentration on reaction rates. Catalytic reactions
were carried out in the temperature range 400–480 °C with a reac-
tant flow rate of 10 sccm/min at atmospheric pressure. Reaction
products were analyzed by a HP 6890 gas chromatograph equipped
with flame ionization (FID) and thermal conductivity (TCD) detec-
tors for probing the organic and inorganic products, respectively.
Methane was used as an internal standard to convert signals be-
tween the two detectors. The measured conversion of methane
was zero under the conditions of the catalysis experiments. The
published values of the response factors for the FID and TCD were
used to correct the peak areas obtained from the two detectors
2
less than the monolayer loading (8 V/nm ), which means that a
single cycle of VOx ALD does not completely cover the surface. This
is likely due to steric constraints imposed by the rather bulky li-
gands on the vanadia precursor. The amount of vanadium loaded
by ALD does not increase linearly for the first few ALD cycles. In Ta-
ble 1 the XRF/ICP results show that the increase in vanadium for
the first few ALD cycles was larger than that for the subsequent cy-
cles. Since in the ALD of VOx the vanadium precursor attachment is
a self-limiting process determined by the available surface func-
tional (hydroxyl) groups, the change in vanadium loading with
each cycle may indicate a gradual decrease in the density of surface
hydroxyls as more cycles of VOx are coated [29].
[
25]. In our reports the selectivity is defined as the percentage of
the reactant that is converted to a specific product:
Moles of Reactant to Form Product
SelectivityðProductÞ ¼
Moles of Reactant Consumed
A Hitachi S4700 scanning electron microscope (SEM) was used
to study the surface topology and structure of the nanoliths. The
aluminum oxide material was removed from the aluminum sealing
ring for thermal gravimetric analysis (TGA). TGA was performed
using a Q50 TGA instrument. During the measurement the temper-
ature was increased from room temperature up to 1000 °C at a rate
of 20 °C/min in a nitrogen atmosphere.
Table 1
The vanadium surface density and UV–Vis edge energies of wet impregnated (W.I.)
and ALD VOx measured by XRF and ICP.
VOx sample
Surface density (V/nm2)
g
E (eV)
W.I. 1 L
W.I. 2 L
W.I. 4 L
ALD 1-cycle
ALD 2-cycles
ALD 4-cycles
ALD 8-cycles
ALD 12-cycles
ALD 18-cycles
8
16
32
2.8
3.7
6.2
9.0
11.7
14.1
3.69
3.52
3.37
3.64
3.48
3.33
3.17
3.11
3.04
3
. Results and discussion
3.1. Nanolith-supported VOx
A typical nanolith consists of 1011 straight pores. Each pore is
4
0 nm in diameter and 70 lm long. The total surface area of the

4
24
H. Feng et al. / Journal of Catalysis 269 (2010) 421–431
An intense, broad absorption band is typically observed in UV–
Vis spectra of the supported VOx catalysts due to oxygen to tetra-
hedral V charge transfer [30,31]. The absorption edge energy (E )
g
The UV–Vis absorption edge from both the ALD and wet
impregnated VOx samples exhibits a red shift as the loading in-
creases. Gao and Wachs reported an approximately linear correla-
tion between the edge energy and the number of V–O–V covalent
bonds around the central V cations based on the measurements
of standard compounds [32]. The edge energy values of the stan-
dard compounds with 0–5 V–O–V bonds range between 3.48 and
2.35 eV. Their results are plotted in Fig. 2 and are used as the ref-
erence for comparing with the measured edge energies of VOx sup-
ported on the nanolith.
5
+
can be used to determine the state of the VOx species since it is re-
lated to the number of V–O–V bonds or the domain size of vanadia
polymers on the surface [32,33]. Fig. 1 shows the transmission UV–
Vis spectra of the nanoliths loaded by ALD and wet impregnation.
The spectra measured before and after catalytic reaction and sub-
sequent calcination are essentially identical. In these spectra the
peaks at ꢀ225 nm (5.52 eV) and ꢀ275 nm (4.51 eV) are assigned
to the isolated monomeric tetrahedral vanadium species [34,35].
In the UV–Vis spectra of the ALD samples, a peak at ꢀ315 nm
5
+
The edge energy of 1-cycle ALD VOx sample is 3.64 eV, which
corresponds to an isolated, tetrahedrally coordinated VO
4
species
(
3.94 eV) becomes pronounced at loadings P6.2 V/nm2 (4-cycle
according to the E values in the reference compounds. This result
g
ALD VOx), indicating the presence of oligomeric distorted tetrahe-
dral VOx [36,37]. The edge energies, E , were determined from the
x-intercept of the linearized near edge region of absorption vs. h
i.e. using the method of Tandon and Gupta [38]. E values from the
wet impregnated and ALD-loaded samples are listed in Table 1.
for the first cycle of ALD VOx is consistent with the expected chem-
istry between surface hydroxyl groups (Al–OH) and the isopropox-
ide ligands where only V–O–Al bonds are formed in the first part of
the cycle. Removal of the remaining isopropoxide ligands by
hydrogen peroxide/water in the second part of the cycle is ex-
pected to produce V–OH species (see Scheme 1). As more cycles
of VOx are deposited, V–O–V bonds will be formed in a reaction be-
tween the V–OH groups and the vanadyl precursor, resulting in a
gradual increase in the average VOx domain size and a correspond-
ing decrease in the edge energy. The lowest edge energy is ob-
tained on the nanolith with the highest VOx loading. However,
g
m,
g
even at the highest loading (18-cycles of VOx), the E
3.04 eV) is still much higher than that of the V crystal, indicat-
ing that at this loading the ALD VOx has fewer V–O–V bonds than
; a two-dimensional structure is still the predominant form of
g
value
(
2 5
O
2 5
V O
VOx on the surface.
Interestingly, despite the large difference in VOx loadings be-
tween the impregnated and the ALD VOx samples, the edge ener-
gies of the 1-cycle, 2-cycle, and 4-cycle ALD samples appear to
be close to their 1-layer, 2-layer, and 4-layer wet impregnated
counterparts, indicating that the predominant forms of the VOx
sites are nearly the same for these pairs of ALD and impregnated
samples. These results reflect the different qualities of the two
types of catalyst loading methods. The ALD technique enables an
extremely uniform dispersion of species on the surface [39]. In
contrast, incipient wetness impregnation and the following calci-
nations cannot guarantee complete spreading of the catalytic
Fig. 1. The UV–Vis absorption spectra of ALD-synthesized and the wet impregnated
VOx supported on the catalytic nanolith.
Fig. 2. Number of V–O–V bonds plotted against the edge energies of vanadium oxide reference compounds and the ALD-synthesized, nanolith-supported VOx edge energy
positions with correlated V–O–V bond numbers. The solid line represents the correlation between the number of V–O–V bonds and the edge energies of the reference
compounds reported by Gao and Wachs [31].

H. Feng et al. / Journal of Catalysis 269 (2010) 421–431
425
Scheme 1. Possible molecular structures of surface vanadyl species formed by
atomic layer deposition of vanadyl triisopropoxide.
2 5
species; particles of V O may form even at loadings significantly
below a theoretical monolayer [11]. Consequently, a relatively
smaller vanadium density is expected on some areas of the surface.
This effect appears to be exacerbated during impregnation of the
nanoliths, as evidenced by an observed accumulation of the pre-
cursor solution near the boundary of the AAO and the aluminum
2 5
ring so that the layers of crystalline V O are very likely to be
formed in these regions. The UV–Vis absorption measurements
probe only the central portion (ꢀ30%) of the AAO material where
the vanadium surface density is likely to be lowest. The results of
reaction studies reported below suggest that the central portion
of the nanolith primarily determines the catalytic performance.
3
3
.2. Catalytic properties of VOx supported on alumina nanoliths
.2.1. Comparison between the ALD VOx and wet impregnated VOx
In all the catalysis tests the absolute conversions measured for
Fig. 3. Conversion (a) and selectivity (b) vs. absorption edge energy for ALD-
synthesized (solid symbols) and wet impregnated (hollow symbols) catalysts at
450 °C.
duplicate samples prepared with the same composition vary less
than 5%. The catalytic layers appear to be stable under our calcina-
tion and reaction conditions. Fresh samples, tested immediately
after the ALD VOx loading, and those re-tested after calcination
in oxygen show little difference in catalytic performance.
Fig. 3a and b depict the catalytic conversion and selectivity re-
sults for ALD (1–4 cycles) and wet impregnated (1–4 layers) VOx at
4
tion of this method in cyclohexane ODH is technically very difficult
and currently beyond our capabilities.
For the present discussion an ‘‘effective TOF” is calculated by
normalizing the measured rates by the total vanadium content
determined by XRF/ICP. As we shall see, the effective TOF and
the selectivities observed for ALD- and W.I.-prepared catalysts
are sensitive to the nature of the catalytic species. Since the ALD
technique produces uniformly dispersed surface VOx species, it is
anticipated that all vanadium atoms contribute to the activity at
50 °C, respectively. The catalysis data are plotted against the
absorption edge energies in the UV–Vis spectra. The conversion/
selectivity performance is similar for each pair of W.I. and ALD
samples having comparable absorption edge energies. Since the
catalytic performance (especially the selectivity) is dependent on
the structure of the VOx species, the similarities in the catalytic
performance provide additional support for the statement made
earlier that the structures of the catalytic sites are essentially the
same for the pairs of ALD and W.I. samples with comparable edge
energies.
To quantitatively compare the specific activities or turnover fre-
quencies (TOF’s) between samples, knowledge of the number of ac-
tive vanadium oxide sites is required. A number of methods for
counting redox-active sites have been proposed in the literature.
Prominent among these are methods that make use of methanol
chemisorption and reaction [40–44], but it has not been estab-
lished that methanol is necessarily a suitable probe molecule for
counting the active sites in alkane ODH reactions. More recently
a method based on the anaerobic oxidation of ethanol has been re-
ported that could be generalized to count sites for the ODH of
cyclohexane [45]; however, due to the low loading of vanadia in
the nanoliths and the multitude of products formed, the applica-
2
loadings 69 V/nm and that the effective TOF and the ‘‘true” TOF
are approximately equal. For the W.I. samples the UV–Vis data sug-
gest that the vanadia species are poorly dispersed.
The effective TOF’s are compared as Arrhenius plots of ln(TOF)
vs. 1/T in Fig. 4. Effective TOF’s for the wet impregnated samples
are substantially lower than the ALD samples due to poor disper-
sion. The percentage of active VOx species on the wet impregnated
nanoliths can be estimated by comparing with the TOF of the ALD
samples that have similar vanadia surface species (by UV–Vis)
using:
TOFW:I:
TOFALD
Active VOx % ¼
ꢁ 100%
ð1Þ
Using Eq. (1) we find that 45 ± 1% of the VOx is active on the nano-
lith at the one monolayer wet impregnation loading; this percent-
age drops to 21 ± 2% and 13 ± 1% as the loading increases to 2-
layers and 4-layers by impregnation. These percentages are similar

4
26
H. Feng et al. / Journal of Catalysis 269 (2010) 421–431
with cyclohexane concentration, consistent with the flow rate
experiments, and remains almost constant at all oxygen concentra-
tions (unless the O
partial pressure is very close to 0). Fig. 5 pre-
2
sents a log–log plot of reaction rate vs. cyclohexane and oxygen
concentrations measured at 450 °C. The reaction orders deter-
mined from the slopes of these plots are 0.9 ± 0.1 for cyclohexane
and 0.1 ± 0.1 for oxygen. The uncertainties are due to variations
in the measured rates. This shows that the ODH reaction is essen-
tially first-order in cyclohexane and zero order in oxygen. This re-
sult is consistent with the reported reaction orders in the ODH of
propane and butane over the supported VOx [46–48]. The resultant
rate equation signifies a typical Mars–van Krevelen redox mecha-
nism, in which lattice oxygen in the catalytic site is responsible
for the dehydrogenation process, and gas phase oxygen only serves
as a reservoir of the oxidant that continuously replenishes lattice
oxygen on the surface.
The TOF of cyclohexane measured using the ALD catalysts (see
Fig. 6) exhibits some interesting trends. The TOF increases dramat-
ically as the loading increases from 2.8 to 6.2 V/nm or from 1 to 4
Fig. 4. Plots of log(TOF) vs. 1/T for ALD-synthesized and wet impregnated
vanadium oxide catalysts.
2
cycles of VOx ALD. It is relatively constant between 6.2 and 11.7 V/
2
nm (4–12 cycles of VOx ALD) and finally decreases at higher load-
to the values reported in reference [45] for ethanol oxidation by
ing. Normally at loadings higher than one monolayer a continuous
decline in the TOF is expected since part of the vanadia will be bur-
ied and will be unable to participate in the reaction. At loadings
vanadia supported on
impregnation.
c-alumina also prepared by wet
The reaction activation energies can be obtained from the
slopes of the plots in Fig. 4. The activation energies range from
2
lower than 11.7 V/nm (1–12 cycles of VOx ALD), the VOx species
are in a highly dispersed form so that nearly all vanadia species are
available for catalysis. The initial increase in TOF with loading re-
flects changes in the intrinsic catalytic activity of the vanadia sites.
From Table 2 we see that the activation energy decreases as the
VOx loading increases. This trend indicates that the surface VOx
species formed at higher loadings are more active for the ODH of
cyclohexane.
Analysis of the UV–Vis spectra shows that the VOx species
formed by 1-cycle of ALD are isolated monomers, consistent with
Scheme 1. The V–O–V bond connecting two adjacent V centers be-
gins to form and gradually prevails on the surface as more cycles of
VOx are deposited. Apparently the presence of these V–O–V bonds
makes two major changes to the catalyst: first, they replace V–O–
Al bonds so that the related chemical properties, such as the
1
9
07 to 81 kJ/mol for 1–4 layers of wet impregnated VOx and from
8 to 63 kJ/mol for the 1–18 cycles of ALD VOx (see Table 2). The
activation energies measured in our system are consistent with
those reported by Lopez and coworkers in studies of the same reac-
tion (60–100 kJ/mol depending on the substrates) [19]. Therefore,
neither the AAO scaffold nor the ALD loading technique seems to
impart unusual chemical properties in the supported VOx other
than enhanced specific activity. This implies that the ALD tech-
nique can be used as an alternative method to synthesize conven-
tional supported catalysts. Further discussion of the VOx catalytic
properties in the ODH of cyclohexane focuses on the ALD samples.
3.3. Activity analysis
The observed rate of a heterogeneous catalytic reaction can be
influenced by the surface kinetics, pore diffusion resistance, and
film mass transfer. In the nanolith catalytic system where the reac-
tants are forced through the nanopores, diffusion resistance will
not significantly affect the reaction rate. Conversions are low (less
than 10%) in most experiments, so that mass transfer and product
concentrations will not significantly influence reaction rates. Our
previous analysis of the dependence of conversion on flow rate in
the nanolith catalytic system showed that the reaction is first-or-
der in cyclohexane [16]. The reaction order with respect to cyclo-
hexane and oxygen was also measured by varying the
concentration of one reactant while holding that of the other con-
stant. The measured cyclohexane reaction rate increases linearly
Table 2
The activation energies of wet impregnated and ALD VOx supported on the nanoliths.
Sample
a
E (kJ/mol)
W.I. 1-layer
W.I. 2-layer
W.I. 4-layer
ALD 1-cycle
ALD 2-cycle
ALD 4-cycle
ALD 8-cycle
ALD 12-cycle
ALD 18-cycle
106.6
80.6
82.4
98.1
92.1
83.5
80.3
65.8
63.3
Fig. 5. The logarithm of cyclohexane reaction rate vs. the logarithm of cyclohexane
and oxygen concentrations measured using the ALD-synthesized catalyst with a
2
vanadium loading of 3.7 V/nm . The reaction rate is first-order (0.9 ± 0.1) in
cyclohexane and approximately zero order (0.1 ± 0.1) in oxygen.

H. Feng et al. / Journal of Catalysis 269 (2010) 421–431
427
sibly affect the reaction path by promoting multiple C–H bond
breaking on the same catalytic site. This would lead to direct con-
version from cyclohexane to benzene. As we see below, a kinetic
analysis of reaction pathways indicates that such a direct conver-
sion is favored at monolayer vanadium surface density (ꢀ9.0 V/
2
nm ).
In contrast to the results presented here, Tian and Wachs re-
ported that for the ODH of propane the specific activity of surface
monomeric and polymeric vanadia species is the same but depends
strongly on the nature of the support cations, analogous to a ligand
effect [33]. While, it would seem that such a ligand effect should
also apply to the formation of V–O–V bonds, it may be that the
ODH of cyclohexane is more sensitive to vanadia species structure
because of its larger size and the possibility of multiple bonds to
the vanadia. The presence of polymeric vanadia clearly plays an
important role in the direct reaction path from cyclohexane to ben-
zene, which accounts for a large portion of the cyclohexane conver-
sion. In the case of propane, a comparable direct reaction path does
not exist since the only deep oxidation products are COx, which re-
quires C–O insertion or C–C breaking in addition to C–H bond
activation.
Fig. 6. The turnover frequency vs. Vanadium loading for ALD-synthesized catalysts
at 450 °C.
reducibility of the vanadia, are enhanced; second, the adjacent V
centers are brought closer and consequently a concerted reaction
mechanism that involves the attendance of more than one V center
becomes possible. The increased TOF at high VOx loadings is due to
one or both of these effects.
3.4. Analysis of selectivity
The conversion and product selectivities for the reaction run at
450 °C over the series of VOx catalysts are illustrated by the data in
Fig. 7. Consistent with the low reaction order in oxygen partial
pressure shown in Fig. 5, the results from experiments using cyclo-
hexane:oxygen mixtures of 1:2 and 1:8 are identical. As the con-
version increases with vanadium loading, the selectivities to CO
The dependence of turnover frequency on the domain size of
VOx is not unique to the cyclohexane ODH reaction. A similar cor-
relation has been found by Iglesia and coworkers for the ODH of
propane [49]. They attribute the change in TOF to changes in the
reducibility of the VOx species. Since oxidative dehydrogenation
and CO
has been reported for the ODH of propane catalyzed by vanadia
supported on -alumina [9] and explained by sequential oxidation
2
increase slightly with values that are very similar to what
5
+
of the hydrocarbon involves the reduction–oxidation of the V
c
center, the ease of reduction of the catalytic site will largely deter-
mine the activity of the catalyst. Therefore, polyvanadate domains,
which are more easily reduced by hydrogen than vanadate mono-
mers, are more active in the ODH reactions. Because the ODH of
cyclohexane and propane share many common features, especially
the hydrocarbon activation step, a common explanation for the
dependence of activity on loading is expected.
reactions that typically accompany increases in conversion in the
range of low conversion. The selectivities to cyclohexene and ben-
zene change much more dramatically. Cyclohexene is the most
2
abundant product (55%) at a vanadium density of 2.8 V/nm but
2
much less abundant (10%) at 6.2 V/nm . In contrast the selectivity
to benzene increases from 26% to 68%. The formation of cyclohex-
ene and benzene is due to the breaking of multiple C–H bonds but
no more than one on each methylene group. In contrast COx for-
mation involves fundamentally different reactions such as C–C
bond breaking, oxygen insertion, or the breaking of both C–H
The fact that the V centers on the surface are brought closer at
higher loadings may also be a factor in their increased catalytic
activity. The proximity of V centers will enable the formation of
multiply bonded reaction intermediates, which are more strongly
adsorbed to the catalytic sites and consequently less likely to des-
orb from the surface before reaction. If the adsorption of cyclohex-
ane onto the catalytic site is due to interaction between a C atom
and a V center, the spatial distance between adjacent V centers will
largely determine whether the formation of multiply bonded spe-
cies is possible. With the V–O–V bonding geometry, the shortest
distance between two V centers is approximately 0.32 nm, based
on crystallographic data of di-vanadium compounds [50]. By com-
parison, the average distance between two isolated V centers sit-
ting next to each other is ꢀ0.58 nm at an average surface density
2
of 3 V/nm (which corresponds to the surface vanadium density
of a 1-cycle ALD sample). The C–C bond length in the cyclohexane
molecule is 0.154 nm, therefore the distances between C–C atoms
in the 1,2, 1,3, and 1, 4 positions are 0.15, 0.24, and 0.29 nm,
respectively. The estimated V–C bond length is approximately
.14 nm based on the reported ionic radius of V5 (0.072 nm) and
+
0
the covalent radius of C (0.07 nm) [51]. Using these estimates a
cyclohexane molecule can bond to the two vanadium centers in
a V–O–V bond via 1,2, 1,3, or 1,4 diadsorbed intermediates while
this is impossible with the two isolated V centers. Multiply bonded
intermediates will facilitate the dehydrogenation process and pos-
Fig. 7. Conversion and selectivity vs. vanadium loading for ALD-synthesized
catalysts at 450 °C. The lines are not fits but intended to make the trends clearer.

4
28
H. Feng et al. / Journal of Catalysis 269 (2010) 421–431
bonds on a methylene group. We argue below that the pronounced
change in selectivities toward cyclohexene and benzene is due to a
variation in the kinetics of competing reaction pathways with
increasing vanadia loading.
In the ODH of alkanes, the conversion is of primary importance
in determining the selectivity because the partial oxidation prod-
uct is typically easier to oxidize than the reactant so that the selec-
tivity drops significantly with increasing conversion. To evaluate
the intrinsic selectivity of different catalyst materials, the preferred
method is to compare experiments at constant reaction tempera-
ture and conversion by adjusting the space velocity. This is not fea-
sible for the catalytic nanoliths because the amount of catalyst in
the reactor is fixed by the nanolith and the range of flow rates is
limited by the mechanical stability of the nanolith material. The
conversion can be adjusted by changing the reaction temperature,
and for this reaction it appears that the temperature and space
velocity have a very similar influence on the selectivity/yield to
cyclohexene in the tested temperature range. Fig. 8 displays the
Scheme 2. Cyclohexane ODH reaction pathways. The numbers represent the rate
coefficients of the different reaction paths at 450 °C of the two samples with 3.7 and
2
9
.0 V/nm . The numbers with underscore represent the rate coefficients of the
2
sample with 9.0 V/nm .
6 6
C H10 yield vs. C H12 conversion curves of the ALD samples with
2
vanadium densities of 2.8 and 9.0 V/nm measured at different
reaction temperatures (and a fixed space velocity) or at different
space velocities (and a fixed temperature). For each sample the
curves measured at different temperatures or at different space
velocities overlap very well. This is also true for the yields of ben-
of rates k
route. In the sequential reaction route, increasing the rate of cyclo-
hexene dehydrogenation (k ) will lower the selectivity to the olefin
and consequently raise the selectivity to benzene. COx can be
formed by the oxidation of benzene (the k route), cyclohexene
the k route), or even directly from cyclohexane (the k route).
To study the roles of the sequential reaction route and the direct
2 1
/k will lead to more benzene through the direct reaction
3
2
zene, CO and CO . Furthermore, it is clear that the yield of cyclo-
5
(
4
6
hexene is much higher on the sample with a low VOx loading
than with a high VOx loading. Taken together the results depicted
in Figs. 7 and 8 point to a strong dependence of selectivity on the
nature of the vanadia species that is distinct from the differences
in conversion.
The selectivity in cyclohexane ODH is determined by the rela-
tive importance of the reaction pathways shown in Scheme 2.
The selectivity to cyclohexene and benzene is determined by the
relative rates of converting cyclohexane to these two products
routes in the ODH of cyclohexane as a function of temperature and
their relationship to the nature of the VOx catalyst, oxidation reac-
tions using cyclohexane, cyclohexene, and benzene as reactants
were studied on catalysts with VOx loadings of 3.7 and 9.0 V/
2
nm . The different reactants were studied using the same reaction
conditions (the same hydrocarbon space velocity, temperature
range and an oxygen/hydrocarbon ratio of ꢀ8/1 to ensure a very
low oxygen conversion and thus pseudo-first-order conditions for
all the ODH reactions).
Comparing the conversion/selectivity patterns of cyclohexene
and benzene as reactants to those of cyclohexane, we can conclude
(
k
1
and k
). The conversion from cyclohexane to benzene via the forma-
? k route) is a
sequential reaction path while the direct conversion from cyclo-
hexane to benzene (k ) is considered an independent direct reac-
2
), and the rate of the further oxidation of cyclohexene
(k
3
tion and further oxidation of cyclohexene (the k
1
3
that the direct conversion from cyclohexane to benzene (the k
2
2
route in Scheme 2) plays a significant role in benzene formation
from cyclohexane. For example, at 400 °C where the benzene con-
version to COx is very low, the yield of benzene from cyclohexane
tion. Both the sequential path and the direct path likely share a
partially dehydrogenated cyclohexane as a common reaction inter-
mediate, in a structure that resembles an adsorbed olefin. This
intermediate can either desorb from the surface to produce cyclo-
1 3
ODH is determined by the sum of the consecutive (k ? k ) and di-
rect (k ) routes. Assuming that the percent conversion of cyclohex-
2
hexene (the k
zene (the k route). A change in the catalyst that increases the ratio
1
route), or lose another C–H bond and head for ben-
ene is the same regardless of whether it is fed as a reactant or
produced by cyclohexane ODH (equivalent to pseudo-first-order
kinetics, differential reaction conditions, and a well-mixed reac-
tor), the yield of benzene from the consecutive route (Y1?3(ben-
zene)) can be estimated by Equation (2) and compared to the
measured yield of benzene from cyclohexane ODH:
2
Y
1
ðcyclohexeneÞ
Y
1!3ðbenzeneÞ ¼ Y
3
ðbenzeneÞ ꢁ
ð2Þ
Y
3þ4ðcyclohexeneÞ
where Y
3
(benzene) is the yield of benzene via the direct reaction
(
reaction 3), Y (cyclohexene) is the yield of cyclohexene from cyclo-
1
hexane ODH (reactions 1), and Y3+4(cyclohexene) is the percentage
of cyclohexene remaining when cyclohexene is fed as a reactant.
Fig. 9 presents the product selectivity patterns in cyclohexane
ODH at 400 °C for the catalysts with VOx loadings of 3.7 and
2
9
.0 V/nm . The hatched area represents the contribution from the
consecutive path estimated from Eq. (2). For both catalysts the ole-
fin formation and reaction rates cannot account for the observed
yields of benzene. In fact, the direct reaction appears to be the dom-
inant route to benzene. Moreover, the branching between consecu-
tive and direct routes appears to change significantly with the
structure of the catalyst. As shown in Fig. 9, a much larger portion
Fig. 8. The C
6 6
H10 yield vs. C H12 conversion curves of the 1-cycle and 8-cycle ALD
VOx samples measured at different temperatures or at different space velocities.

H. Feng et al. / Journal of Catalysis 269 (2010) 421–431
429
d½C
d½C
6
H
dt
10
ꢃ
ꢂ
ꢂ
¼ k
3
½C
6
H
10ꢃ þ k
4
½C
6
H
10
6
ꢃ
ꢃ
ð7Þ
6
H
dt
6
ꢃ
¼ ꢂk
3
½C
6
H
10ꢃ þ k
5
½C
6
H
ð8Þ
Finally, the parameters for reactions 1, 2, and 6 are determined from
the experiment with reactant cyclohexane. Altogether there are 12
parameters obtained from fitting 30, independent, data points. The
solid curves obtained from the model are compared to the experi-
mental results in Fig. 10. The fits between the experiments and
Fig. 9. The product distribution patterns of the ALD catalysts with VOx loadings of
.7 and 9.0 V/nm2 in the ODH of cyclohexane at 400 °C. The hatched areas in the
3
charts represent the contributions from the consecutive paths.
of benzene is produced from the direct path on the catalyst with a V
2
loading of 9.0 V/nm .
To provide a more quantitative comparison of sequential and
direct reaction paths and the role of other reactions, a first-order
kinetic analysis of Scheme 2 was performed, analogous to that of
reference [52]. This simple analysis ignores the effects of inhibition
by any of the intermediates and products. Fig. 5 confirms that
cyclohexane ODH is nearly first-order for cyclohexane and zero or-
der for oxygen. The product selectivities were the same for cyclo-
hexane:oxygen ratios of 1:2 and 1:8, consistent with near zero
order oxygen dependence in the reactions of cyclohexene and ben-
zene. Indeed, the ODH of propene and butene were reported to fol-
low the Mars–van Krevelen mechanism and were first-order in
olefin and low order in oxygen [47]. Thus, the assumption of
first-order kinetics should be adequate at the low conversions em-
ployed here.
Since gas flow in the nanolith pores conforms to the mixed flow
regime (diffusion + plug flow) [16], the nanolith system is similar
to a continuous-feed, batch reactor. Thus, the conversion of cyclo-
hexane is given by:
d½C
6
H
dt
12
ꢃ
ꢂ
¼ k
1
½C
6
H
12ꢃ þ k
2
½C
6
H
12ꢃ þ k
6
½C
6
H
12
ꢃ
ð3Þ
Corresponding equations are applied to describe the conversions of
cyclohexene and benzene:
d½C
6
H
dt
10
ꢃ
ꢂ
¼ ꢂk
1
½C
6
H
12ꢃ þ k
3
½C
6
H
10ꢃ þ k
4
½C
6
H
10
ꢃ
ð4Þ
d½C
6
H
dt
6
ꢃ
ꢂ
¼ ꢂk
2
½C
6
H
12ꢃ ꢂ k
3
½C
6
H
10ꢃ þ k
5
½C
6
H
6
ꢃ
ð5Þ
The pseudo-first-order rate coefficients are written as
ꢀ
ꢁ
E
RT
i
k
i
¼ A
i
exp
The coupled differential equations are solved
numerically, and the rate coefficient parameters are obtained from
experiments with benzene, cyclohexene, and cyclohexane as the
reactants by least squares fitting the product yields [53]. The rate
5 5
coefficient parameters for reaction 5 in Scheme 2 (A and E ) are
first determined from the experiment with reactant benzene, in
which the benzene consumption is expressed as:
d½C
6
H
dt
6
ꢃ
ꢂ
¼ k
5
½C
6
H
6
ꢃ
ð6Þ
Fig. 10. Comparison of experimental and fitted yields vs. temperature using the
ALD-synthesized catalyst with vanadium loadings of 3.7 and 9.0 V/nm2 for reactant
benzene (a), cyclohexene (b), and cyclohexane (c). For the reaction of cyclohexene
with the higher loading sample only the three data points at lower temperatures
are used in the modeling because at higher reaction temperatures the conversion of
cyclohexene is 100%.
The resulting rate coefficient parameters are used in the determina-
tion of the parameters for reactions 3 and 4 (A , E , and A , E ) from
the experiment with reactant cyclohexene, where
3
3
4
4

4
30
H. Feng et al. / Journal of Catalysis 269 (2010) 421–431
the model are very good for both catalysts. Correlation coefficients
cyclohexene product but the direct surface reaction (the k2 reac-
tion path in Scheme 2). It is likely that the same process applies
to the formation of carbon oxides in the ODH of other alkanes, such
as propane, by supported vanadium oxide, and that this is con-
trolled, in part, by the structure of the vanadium oxide species.
are all above 0.99, and the largest percent standard deviation for the
0
calculated k s is 5% with typical values nearer to 1%.
i
What is of interest here is how the partitioning along reaction
paths changes between high and low loading of vanadia. This can
be seen by examining the values of the rate coefficients given in
1
Table 3 relative to k for the low loading catalyst. We see that on
2
the sample with the low catalyst loading (3.7 V/nm ), a large frac-
tion (56–48%) of the reacted cyclohexane follows the sequential
reaction path and a smaller fraction (33–38%) is directly converted
to benzene at the lower reaction temperatures; at higher temper-
atures more cyclohexane (45–51%) is directly converted to ben-
4. Conclusions
VOx catalysts of different structures were synthesized on the
surface of catalytic nanoliths either by performing cycles of ALD
using vanadium oxytriisopropoxide /hydrogen peroxide + water
as the precursors or by incipient wetness impregnation. XRF/ICP
measurements indicate that each ALD cycle enables sub-mono-
layer loading (1/3 of a monolayer) of VOx onto the surface. In the
UV–Vis spectra of the nanolith-supported catalysts, the absorption
edge energies reveal that the predominant form of the supported
VOx transforms from monovanadate sites to polyvanadate do-
2 2 1
zene through the k route. The k /k ratio ranges between 0.6
and 1.3 depending on the reaction temperature. On the high load-
2
ing sample (9.0 V/nm ), at all reaction temperatures, nearly 70% of
the reacted cyclohexane is directly converted to benzene through
the direct path and in this case the k /k ratio is 5.2–6.8. These
2 1
trends show that the form or structure of the catalytic site is a ma-
jor factor determining the reaction path in the ODH of cyclohexane.
As discussed above, during the cyclohexane activation process, the
multi-adsorbed intermediate can be formed on polyvanadate do-
mains containing the V–O–V bond, suggesting that this configura-
tion facilitates the direct conversion from cyclohexane to benzene.
Additionally, the larger VOx polymeric domains possess more lat-
tice oxygen and are more easily reduced; therefore they are more
likely to be capable of multiple C–H breaking steps on the same
catalytic site as required for the direct reaction mechanism.
Scheme 2 presents the rate coefficients of the different reaction
paths at 450 °C from the two samples. Comparing these rate coef-
ficients, it is interesting to note that although the overall reaction
2
mains as the loading increases from ꢀ3 to 14 V/nm .
The nanoliths loaded with the ALD VOx show ꢀ2 to 5 times
higher specific activities than those with the impregnated VOx,
which reflects the better dispersion of the catalytic species synthe-
sized by ALD. The activation energies of the ALD and impregnated
VOx were in the range 60–100 kJ/mol; these values are comparable
to those reported in the literature. These results imply that the
chemical properties of the supported VOx are not altered by the
nanolith scaffold or the ALD loading technique. Therefore ALD
can be applied as an alternative method to prepare supported
catalysts.
The cyclohexane ODH reaction over ALD VOx is found to be
approximately first-order in hydrocarbon and zero order in oxy-
gen. The observed reaction orders are in agreement with the
Mars–Van Krevelen redox mechanism, which is typical for the
ODH of alkanes over supported metal oxide catalysts. In the ODH
reaction of cyclohexane, a rise in the catalytic activity (TOF) to-
gether with a decrease in the activation energy is observed with
the increasing loading of the catalyst, implying that the polyvana-
date domains are more active than the monovanadate sites. How-
ever, the selectivity to cyclohexene drops dramatically as the VOx
loading increases. Even when compared at the same conversions,
the samples with higher VOx loadings always show lower selectiv-
ity to cyclohexene. The reaction network for cyclohexane ODH is
analyzed by numerical simulation of the oxidation of cyclohexane,
cyclohexene, and benzene under the same experimental condi-
tions. It is found that both sequential and direct paths exist in
the cyclohexane oxidation process. The formation of the deep oxi-
dation products, benzene and COx, can proceed through either the
sequential path or the direct paths, depending on the structure of
the supported catalysts. When the catalyst loading is low (3.7 V/
2
rate of the sample with 9.0 V/nm is higher than the lower loading
1
sample, the latter has a higher rate of cyclohexene formation (k ) in
accordance with the yields shown in Fig. 8. The same trend can also
be observed at all other temperatures. This suggests that cyclohex-
ene is formed mainly (if not exclusively) on monomeric vanadyl
centers. It is possible that at some point the conversion from cyclo-
hexane to cyclohexene and to benzene share a common reaction
intermediate, which then desorbs from the monomer sites to form
cyclohexene but undergoes further C–H breaking to form benzene
on polymeric sites.
These results demonstrate that the primary route to nonselec-
tive products from cyclohexane is via breaking of multiple C–H
bonds to form benzene and ultimately carbon oxides. The model-
ing also shows that the most important pathway at high loadings
of vanadium oxide is not the sequential reaction of a primary
Table 3
Rate coefficients for the reaction paths shown in Scheme 2.
2
T (°C)
400
420
450
465
480
nm ), 56–36%, depending on the temperature, of the reacted cyclo-
(
a) The rate coefficients of different reactions routes (shown in Scheme 2)
hexane follows the sequential reaction path and 33–51% is directly
converted to benzene. At high loading (9.0 V/nm ), regardless of
2
obtained from the simulation on the 2-cycle ALD VOx sample. The units are (V-
ꢂ1
s) ꢁ 1000
the reaction temperature, over 70% of the reacted cyclohexane is
directly converted to benzene through the direct path. Although
the overall reaction rate of the sample with 9.0 V/nm2 is much
higher than that of the lower loading sample, the lower loading
sample has a higher (ꢀ3 times) rate of cyclohexene formation. This
suggests that cyclohexene is formed mainly on monomeric vanadyl
centers.
k
k
k
k
k
k
1
2
3
4
5
6
1.5
0.9
22
6.5
1.7
0.3
2.1
1.7
36
10
2.3
0.6
3.5
3.8
72
19
3.6
1.1
4.4
5.4
99
27
4.5
1.5
5.4
7.8
133
37
5.6
2.1
(
b) The rate coefficients of different reactions routes (shown in Scheme 2)
obtained from the simulation on the 8-cycle ALD VOx sample. The units are (V-
ꢂ1
s) ꢁ 1000
k
k
k
k
k
k
1
2
3
4
5
6
0.5
2.6
36
10
1.4
0.8
0.6
3.9
70
21
2.1
1.2
1.1
7.2
180
54
3.9
2.4
1.4
9.5
285
82
5.3
3.2
1.8
12
420
126
6.9
4.2
Acknowledgment
This work was supported by the US Department of Energy, Of-
fice of Basic Energy Science, under Contracts DE-AC02-
0
6CH11357 and DE-FG02-03-ER15457.

H. Feng et al. / Journal of Catalysis 269 (2010) 421–431
431
[
27] R. Katamreddy, R. Inman, G. Jursich, A. Soulet, C. Takoudis, J. Electrochem. Soc.
53 (2006) C701–C706.
References
1
[
[
28] G. Gutierrez, B. Johansson, Phys. Rev. B 65 (2002) 104202–104211.
29] J.C. Badot, S. Ribes, E.B. Yousfi, V. Vivier, J.P. Pereira-Ramos, N. Baffier, D. Lincot,
Electrochem. Solid-State Lett. 3 (2000) 485–488.
30] B.M. Weckhuysen, D.E. Keller, Catal. Today 78 (2003) 25–46.
31] G. Centi, S. Perathoner, F. Trifiro, A. Aboukais, C.F. Aissi, M. Guelton, J. Phys.
Chem. 96 (1992) 2617–2629.
[
[
[
[
1] E.A. Mamedov, V. Cortes Corberan, Appl. Catal. A 127 (1995) 1–40.
2] H.H. Kung, Adv. Catal. 40 (1994) 1–38.
3] A. Khodakov, B. Olthof, A.T. Bell, E. Iglesia, J. Catal. 181 (1999) 205–216.
4] L.E. Briand, O.P. Tkachenko, M. Guraya, X. Gao, I.E. Wachs, W. Gruenert, J. Phys.
Chem. B 108 (2004) 4823–4830.
5] I.E. Wachs, B.M. Weckhuysen, Appl. Catal. A 157 (1997) 67–90.
6] N. Ballarini, F. Cavani, A. Cericola, C. Cortelli, M. Ferrari, F. Trifiro, G. Capannelli,
A. Comite, R. Catani, U. Cornaro, Catal. Today 91–92 (2004) 99–104.
7] G. Deo, I.E. Wachs, J. Catal. 146 (1994) 323–334.
8] A.A. Lemonidou, L. Nalbandian, I.A. Vasalos, Catal. Today 61 (2000) 333–341.
9] M.D. Argyle, K.D. Chen, A.T. Bell, E. Iglesia, J. Catal. 208 (2002) 139–149.
[
[
[
[
[
[
32] X. Gao, I.E. Wachs, J. Phys. Chem. B 104 (2000) 1261–1268.
33] H. Tian, I. Ross Elizabeth, E. Wachs Israel, J. Phys. Chem. B 110 (2006) 9593–
9600.
[
[
[
[
[
34] P. Knotek, L. Capek, R. Bulanek, J. Adam, Top. Catal. 45 (2007) 51–55.
35] P. Van Der Voort, M.G. White, M.B. Mitchell, A.A. Verberckmoes, E.F. Vansant,
Spectrochim. Acta Part A 53A (1997) 2181–2187.
36] M.L. Pena, A. Dejoz, V. Fornes, F. Rey, M.I. Vazquez, J.M. Lopez Nieto, Appl.
Catal. A 209 (2001) 155–164.
[
[
[
[
10] M.A. Vuurman, I.E. Wachs, J. Phys. Chem. 96 (1992) 5008–5016.
11] G.T. Went, S.T. Oyama, A.T. Bell, J. Phys. Chem. 94 (1990) 4240–4246.
12] F. Cavani, N. Ballarini, A. Cericola, Catal. Today 127 (2007) 113–131.
13] J. Keranen, A. Auroux, S. Ek, L. Niinisto, Appl. Catal. A: General 228 (2002) 213–
25.
14] J. Keranen, C. Guimon, E. Iiskola, A. Auroux, L. Niinisto, Catal. Today 78 (2003)
49–157.
15] R.J. Silvennoinen, O.J.T. Jylha, M. Lindblad, H. Osterholm, A.O.I. Krause, Catal.
Lett. 114 (2007) 135–144.
16] H. Feng, J.W. Elam, J.A. Libera, M.J. Pellin, P.C. Stair, Chem. Eng. Sci. 64 (2009)
[
[
37] M. Baltes, K. Cassiers, P. Van Der Voort, B.M. Weckhuysen, R.A. Schoonheydt,
E.F. Vansant, J. Catal. 197 (2001) 160–171.
2
[
[
38] S.P. Tandon, J.P. Gupta, Phys. Status Solidi 38 (1970) 363–367.
39] J.W. Elam, G. Xiong, C.Y. Han, H.H. Wang, J.P. Birrell, U. Welp, J.N. Hryn, M.J.
Pellin, T.F. Baumann, J.F. Poco, J.H. Satcher Jr., J. Nanomater. (2006) 1–5.
40] I.E. Wachs, T. Kim, E.I. Ross, Catal. Today 116 (2006) 162–168.
41] I.E. Wachs, J.-M. Jehng, W. Ueda, J. Phys. Chem. B 109 (2005) 2275–2284.
42] L.J. Burcham, L.E. Briand, I.E. Wachs, Langmuir 17 (2001) 6164–6174.
43] L.J. Burcham, L.E. Briand, I.E. Wachs, Langmuir 17 (2001) 6175–6184.
44] M. Badlani, I.E. Wachs, Catal. Lett. 75 (2001) 137–149.
[
[
[
[
1
[
[
[
[
[
[
[
[
5
60–567.
17] P.C. Stair, C. Marshall, G. Xiong, H. Feng, M.J. Pellin, J.W. Elam, L. Curtiss, L. Iton,
H. Kung, M. Kung, H.-H. Wang, Top. Catal. 39 (2006) 181–186.
18] E.V. Kondratenko, M. Baerns, Appl. Catal. A 222 (2001) 133–143.
19] M. Panizza, C. Resini, G. Busca, E. Fernandez Lopez, V. Sanchez Escribano, Catal.
Lett. 89 (2003) 199–205.
20] K. Sato, M. Aoki, R. Noyori, Science (Washington DC) 281 (1998) 1646–1647.
21] W. Yao, Y. Chen, L. Min, H. Fang, Z. Yan, H. Wang, J. Wang, J. Mol. Catal. A:
Chem. 246 (2006) 162–166.
45] H. Nair, C.D. Baertsch, J. Catal. 258 (2008) 1–4.
46] K. Chen, A.T. Bell, E. Iglesia, J. Phys. Chem. B 104 (2000) 1292–1299.
47] A. Dejoz, J.M. Lopez Nieto, F. Melo, I. Vazquez, Ind. Eng. Chem. Res. 36 (1997)
[
[
2
588–2596.
48] K. Inumaru, M. Misono, T. Okuhara, Appl. Catal. A: General 149 (1997) 133–
49.
[
[
[
1
[
[
[
[
49] K. Chen, A. Khodakov, J. Yang, A.T. Bell, E. Iglesia, J. Catal. 186 (1999) 325–333.
50] R. Gopal, C. Calvo, Acta Crystallogr. Sect. B 30B (1974) 2491–2493.
51] R.D. Shannon, Acta Crystallogr. Sect. A A32 (1976) 751–767.
[
22] T. Gao, G. Meng, J. Zhang, S. Sun, L. Zhang, Appl. Phys. A: Mater. Sci. Process. 74
(
2002) 403–406.
[
[
[
[
23] H. Masuda, K. Fukuda, Science 268 (1995) 1466–1468.
24] J.W. Elam, M.D. Groner, S.M. George, Rev. Sci. Instrum. 73 (2002) 2981–2987.
25] W.A. Dietz, J. Gas Chromatogr. 5 (1967) 68–71.
26] G. Xiong, J.W. Elam, H. Feng, C.Y. Han, H.-H. Wang, L.E. Iton, L.A. Curtiss, M.J.
Pellin, M. Kung, H. Kung, P.C. Stair, J. Phys. Chem. B 109 (2005) 14059–14063.
52] S.T. Oyama, A.M. Middlebrook, G.A. Somorjai, J. Phys. Chem. 94 (1990) 5029–
5033.
[
53] Scientist by Micromath Inc.