Effect of VOx species and support on coke formation and catalyst stability in nonoxidative propane dehydrogenation

Article abstract of DOI:10.1002/cctc.201500151

VO_x/SiO₂-Al₂O₃ catalysts were prepared by grafting vanadyl acetylacetonate onto the supports with a SiO₂ content between 0 and 100wt. %. The degree of polymerization of VO_x species and acidity both of pristine supports and the catalysts were evaluated. To determine their on-stream stability and carbon deposition activity in nonoxidative propane dehydrogenation, continuous-flow tests and insitu thermogravimetric measurements were performed. The rate constants of catalyst deactivation and carbon deposition were derived from kinetic evaluation of these experiments. Gathered experimental evidence pointed out that VO_x species were significantly more active for coke formation than acid sites of the supports. The rate constant of carbon formation was found to increase with the degree of polymerization of VO_x species, whereas no correlation between catalyst acidity and the rate constants of coking or deactivation could be drawn.

Full text of DOI:10.1002/cctc.201500151

DOI: 10.1002/cctc.201500151
Full Papers
Effect of VOx Species and Support on Coke Formation and
Catalyst Stability in Nonoxidative Propane
Dehydrogenation
Sergey Sokolov,[a] Victor Yu. Bychkov,[b] Mariana Stoyanova,[a] Uwe Rodemerck,[a]
Ursula Bentrup,[a] David Linke,[a] Yurij P. Tyulenin,[b] Vladimir N. Korchak,[b] and
Evgenii V. Kondratenko*[a]
VOx/SiO2–Al2O3 catalysts were prepared by grafting vanadyl
acetylacetonate onto the supports with a SiO2 content be-
tween 0 and 100 wt.%. The degree of polymerization of VOx
species and acidity both of pristine supports and the catalysts
were evaluated. To determine their on-stream stability and
carbon deposition activity in nonoxidative propane dehydro-
genation, continuous-flow tests and in situ thermogravimetric
measurements were performed. The rate constants of catalyst
deactivation and carbon deposition were derived from kinetic
evaluation of these experiments. Gathered experimental evi-
dence pointed out that VOx species were significantly more
active for coke formation than acid sites of the supports. The
rate constant of carbon formation was found to increase with
the degree of polymerization of VOx species, whereas no corre-
lation between catalyst acidity and the rate constants of
coking or deactivation could be drawn.
Introduction
Propylene is the second important building block of the chem-
ical industry after ethylene and finds its application in produc-
tion of polymers, fibers, resins, and solvents. It is industrially
produced as a co-product of steam cracking of naphtha and
fluid catalytic cracking of heavy oil fractions.[1,2] Metathesis of
ethylene with 2-butenes[3] and nonoxidative dehydrogenation
(DH) of propane[2] are two commercially available on-purpose
propylene production alternatives. Growing extraction of shale
gas drives the price of propane down thus making the latter
process more commercially attractive despite high energy
demand caused by strong endothermicity of the reaction. The
existing industrial DH processes employ catalysts based on
chromium oxide or metallic platinum.[2,4] Supported VOx-con-
taining catalysts were extensively studied for their activity and
selectivity in the conversion of C3–C4 hydrocarbons into corre-
sponding olefins in the absence of gas-phase O2. From a mech-
anistic viewpoint, this reaction can run oxidatively or nonoxida-
tively with participation of lattice oxygen of oxidized VOx spe-
cies[5–14] or their reduced counterparts,[15,16] respectively. Al-
though the former approach is thermodynamically not limited,
it suffers from low on-stream stability caused by a reduced ca-
pacity of VOx species to provide their active oxygen. Recently,
VOx/SiO2–Al2O3 catalysts were shown to have commercially at-
tractive activity, propylene selectivity and durability in propane
DH.[15,16] The latter is an important process parameter, because
all known DH catalysts suffer from deactivation caused by coke
deposits formed on their surface.[17,18] To remove coke and re-
store the initial activity and selectivity, catalysts must be oxida-
tively regenerated. Such regeneration results in restructuring
of the active Pt and CrOx species and, consequently, in a gradu-
al loss of the catalytic performance with a growing number of
DH/regeneration cycles,[19] while VOx-based catalysts remain
stable.[15,16] It was also demonstrated that VOx/MCM-41 main-
tains higher stability than its CrOx counterpart in a given DH
cycle,[15] which, combined with VOx better durability, sets
a case for further investigation of vanadium oxide catalysts as
a potential alternative to commercial chromium oxide analogs.
Thorough understanding of the process(es) responsible for
coke deposition and removal is highly important for further
development of novel or improving of the existing catalysts.
The analytical methods typically employed to study the phe-
nomenon include temperature-programmed oxidation of
spent catalysts, as well as in situ spectroscopic and thermo-
gravimetric techniques. The latter in situ methods allow kinetic
analysis of coke formation and thus make deeper fundamental
understanding of coke formation possible. They were instru-
mental in elucidating relations between coking behavior and
catalytic activity in several reactions[20–23] including propane de-
hydrogenation on CrOx-[24,25] and Pt-based[26–28] catalysts. To our
knowledge, there are no mechanistic and kinetic studies of
coke formation in propane DH over VOx-based catalysts.
[a] Dr. S. Sokolov, Dr. M. Stoyanova, Dr. U. Rodemerck, Dr. U. Bentrup,
Dr. D. Linke, Dr. E. V. Kondratenko
Leibniz-Institut für Katalyse e.V. an der Universität Rostock
Albert-Einstein-Str. 29 a,d-18059 Rostock (Germany)
E-mail: Evgenii.Kondratenko@catalysis.de
[b] Dr. V. Y. Bychkov, Dr. Y. P. Tyulenin, Prof. V. N. Korchak
Semenov Institute of Chemical Physics
Russian Academy of Sciences
Kosygina str. 4, GSP-1, Moscow 119991 (Russia)
Supporting Information for this article is available on the WWW under
http://dx.doi.org/10.1002/cctc.201500151.
ChemCatChem 2015, 7, 1691 – 1700
1691
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
The purpose of the current work was to elucidate physico–
chemical properties of VOx-based catalysts responsible for coke
deposition in propane DH and hence detrimental for their on-
stream stability. To this end, we prepared catalysts comprising
highly dispersed or polymerized VOx species on the supports
of different acidity [SiO2, Al2O3, and SiO2–Al2O3 (Siral)] as well as
nanocrystalline V2O5 on a single selected Siral. Their on-stream
activity and selectivity in propane DH were determined under
industrially relevant conditions, and the kinetics of coke forma-
tion was derived from time-resolved in situ thermogravimetric
experiments.
Results and Discussion
Determination of VOx molecular structures
The degree of polymerization of surface VOx species was evalu-
ated by in situ UV/Vis spectroscopy at 5508C in air to exclude
the effect of adsorbed water. This technique has been success-
fully employed by many researchers to distinguish between
isolated, polymerized VOx species or crystalline V2O5 in sup-
ported vanadium-containing catalysts.[29–40] In Figure 1, the
UV/Vis spectra of 5V/MCM-41, 4V/S10, 6V/S10, 13V/S10 and
3V/Al2O3 are shown (for abbreviations see Experimental Sec-
tion). Briefly, the number in front gives the rounded vanadium
weight content, the index in the Siral formula stands for
a weight percentage of SiO2. The spectra of other V/Siral cata-
lysts with V loading of 4–5 wt% were similar to that of 4V/S10.
A broad adsorption maximum at approximately 300 nm in
the spectrum of 5V/MCM-41 stems from O2À to V5+ charge-
transfer transitions and is characteristic of highly dispersed de-
hydrated VO4 species.[33,36,37,41] The red shift and broadening of
the maximum take place if such species become connected via
oxygen bridges or if three-dimensional V2O5 is formed.[33,37,41]
For a qualitative comparison of the catalysts with respect to
Figure 1. UV/Vis spectra of selected catalysts collected at 5508C in air.
favors formation of higher polymerized VOx species. Contrarily,
for the catalysts with high V surface density, for example, 13V/
S10 and, to a lesser extent, 6V/S10, VOx particles were larger
than follows from the trend. This deviation suggests that sup-
port acidity plays less or no role in determining the size of VOx
particles if the support surface coverage with V approaches or
exceeds a monolayer.
On-stream catalytic performance
For all catalysts tested, propylene was the main product, and
methane, ethane, ethylene, and C4 and C5 hydrocarbons were
gas-phase side products. In Figure 3 propane conversion is
shown as a function of time on stream. The corresponding
data for selectivity to propylene and carbon are given in Sup-
porting Information, Figure S1. For all catalysts, the initial pro-
pane conversion was below the equilibrium conversion of 0.43
calculated for the current experimental conditions. Clearly, the
catalysts differed substantially in their on-stream stability and
selectivity to carbon. 5V/MCM-41 and 5V/S70 showed the
highest stability and the lowest carbon selectivity with the
latter being between 1 and 5% over the duration of DH test.
In contrast, this selectivity over other tested catalysts dropped
the kind of VOx species, we calculated the edge energy (EEdge
)
from the UV/Vis spectra according to the procedure described
in Ref. [42]. The results are given in Table 1. As expected, EEdge
decreased with increasing apparent V surface density (Fig-
ure 2a) and reached the lowest value for 13V/S10. This catalyst
possessed crystalline V2O5 [PDF#00-074-1595] with an average
crystallite size of 65 nm as determined by XRD analysis, where-
as all other samples were free of XRD-detectable vanadia.
However, the loading and
hence the surface V density was
Table 1. Physico–chemical properties of catalysts and rate constants of deactivation (kdeac) and carbon deposi-
tion over the catalyst surface (kcoke(m)) and over already existing carbon layer (kcoke(M)).
not the only factor determining
the degree of polymerization of
VOx species. Acidity of the sup-
ports on which VO(acac)2 was
grafted also influenced the final
distribution of VOx species, as
shown in Figure 2b. Clearly, for
the catalysts with low V surface
coverage, EEdge decreases with
an increase in acidity of the cor-
responding pristine supports. In
other words, higher acidity
[a]
Catalyst
V
SBET
EEdge
[eV]
kDH
kdeac
[minÀ1
k
coke(m)
k
coke(M)
[wt.%]
[m2 gÀ1
]
[molkgÀ1 PaÀ1 sÀ1
]
]
[minÀ1
]
[minÀ1
]
3V/Al2O3
4V/S1
2.7
4.3
4.2
6.4
13.2
5.1
74 (89)
1.87
2.08
2.11
1.98
1.72
2.13
2.24
2.76
3.8”10À8
6.1”10À8
6.3”10À8
6.8”10À8
4.0”10À8
5.4”10À8
2.4”10À8
4.4”10À8
17.8”10À4
11.3”10À4
10.5”10À4
14.0”10À4
18.3”10À4
10.0”10À4
4.3”10À4
3.9”10À4
5.5”10À3
0.9”10À3
1.1”10À3
1.8”10À3
7.1”10À3
0.6”10À3
0.5”10À3
0.5”10À3
1.5”10À4
0
235 (255)
280 (350)
276 (350)
150 (350)
353 (475)
266 (360)
810 (901)
4V/S10
6V/S10
13V/S10
5V/S40
5V/S70
5V/MCM41
0
0
2.1”10À4
0
0
0
4.6
4.9
[a] BET surface area of pristine supports is given in parentheses.
ChemCatChem 2015, 7, 1691 – 1700
www.chemcatchem.org
1692
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Figure 2. EEdge of VOx species versus a) apparent V surface density and
b) specific support acidity measured by in situ FTIR pyridine adsorption at
1508C: Numbers in the squares: 1) 3V/Al2O3, 2) 4V/S1, 3) 4V/S10, 4) 6V/S10,
5) 13V/S10, 6) 5V/S40, 7) 5V/S70, and 8) 5V/MCM-41. Apparent V surface
density was calculated as a ratio of V atoms to catalyst surface area.
from approximately 15% at the beginning to zero towards the
end of the test. The decrease in Scarbon was concomitant with
the activity loss, which was especially rapid on 3V/Al2O3 and
13V/S10, making these two catalysts the least stable.
To quantitatively compare the catalysts in terms of their on-
stream stability, we estimated the constant of deactivation rate
by fitting of the experimental data in Figure 3 to the kinetic
model expressed in Equation (1). The deactivation rate was as-
sumed to be of the first order and independent of the concen-
tration of gas-phase species. The reactor was modeled as iso-
thermal plug–flow. Both the constant of deactivation rate
(kdeac) and the constant of propane dehydrogenation (kDH) were
the fitting parameters.
Figure 3. On-stream propane conversion over a) V/Sirals with V loading 4–
5 wt.%; b) 5V/MCM-41 and 3V/Al2O3; and c) V/S10 with V loading of 4.2, 6.4,
and 13.2 wt.%. Test conditions: 5508C, C3H8/N2 =40:60, WHSV=0.94 hÀ1
.
&
&
*
Dashed lines were calculated by using Equation (1). , 3V/Al2O3; , 4V/S1;
,
~
!
^
^
*
4V/S10; , 6V/S10; , 13V/S10; , 5V/S40; , 5V/S70; and , 5V/MCM-41.
r ¼ expðÀkdeac  tÞ Â ½kDH  pðC3H8Þ À kDH=Keq  pðH2Þ Â pðC3H6ފ
ð1Þ
ChemCatChem 2015, 7, 1691 – 1700
www.chemcatchem.org
1693
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
where p(C3H8), p(C3H6), and p(H2) are the partial pressures of
the respective gases. Keq is the equilibrium constant of propane
DH (1.36”104 Pa at 5508C) and t is time.
In Figure 3, the experimental data points and the “deactiva-
tion” curves obtained through modeling are displayed. Appa-
rently, the model correctly predicts evolution of propane con-
version with time on stream over all catalysts. The deactivation
rate constant (kdeac) obtained from the fitting is given in
Table 1.
If kdeac values are related to the corresponding EEdge (Table 1,
Figure 6d), it becomes evident that kdeac increases as EEdge de-
creases, that is, smaller VOx species (high EEdge) deactivate
slower. To gain further fundamental insight into this relation-
ship, we analyzed the kinetics of carbon formation under pro-
pane DH conditions by means of thermogravimetric analysis
(TGA) with a simultaneous analysis of gas-phase products.
In situ monitoring of catalyst reduction and carbon
deposition
In Figure 4, the weight-change curves are shown obtained in
the in situ TGA experiments on the catalysts in a flow of C3H8/
He=50:50. For all catalysts, the curves feature an initial drop
followed by a rise, yet their temporal evolution strongly varies
between the catalysts. The initial weight loss commences in
two steps occurring in the temperature ranges of 90–450 and
450–5508C (Figure S2). The mass spectrometry (MS) measure-
ments performed simultaneously with TGA revealed evolution
of water on all catalysts also proceeding in two steps with the
maxima falling into the 170–2708C and 500–5508C ranges (Fig-
ure S3).
The first step was ascribed to desorption of physically ad-
sorbed water, and the second to the loss of lattice oxygen
from VOx species upon their reduction by propane and/or reac-
tion products. The latter phenomenon was reported in several
previous studies.[43–49] Dehydroxylation of the surface may
occur in a wider range of temperatures and thus contributes
to both weight-loss steps. Noteworthy is the difference be-
tween 13V/S10 and two other V/S10 catalysts (Figure S3c). On
the former, the low-temperature peak of water desorption (at
ꢀ1908C) is negligible compared to the one related to water
formation through reduction of VOx species by propane in the
300–5508C range. On the latter two catalysts, both peaks are
of comparable height. Recalling that Siral10 surface in 13V/
S10 is covered by more than a monolayer of VOx, it can be hy-
pothesized that water adsorption on VOx and V2O5 is hindered
compared to that on Siral10 resulting in a relatively small “de-
sorption” peak. On other hand, the “reduction” peak of 13V/
S10 has two maxima caused by participation of surface and
bulk lattice oxygen of V2O5 in water formation.
Figure 4. TGA profiles recorded during propane DH over the catalysts with
a) V loading of 3–5 wt.% and b) over V/S10 with V loading of 4.2, 6.4, and
13.2 wt.%. Test conditions: 5508C, C3H8/He=50:50. Catalyst numbers are de-
fined in Figure 2.
tion likewise oxidative propane dehydrogenation are largely
completed. The intensity of H2 keeps growing indicating that
the DH reaction proceeds further via a nonoxidative route.
Once water is removed from the catalysts through the pro-
cesses discussed above, the weight curves start ascending be-
cause of continuous accumulation of coke on the catalysts. It
is evident that the trends in weight increase in Figure 4 varied
dramatically between the catalysts. 4V/S1 and 4V/S10 showed
the strongest weight gain that lasted over entire duration of
the experiment. Contrarily, the weight curves of 3V/Al2O3 and
13V/S10 are characterized by fast increase in the first 6 h on
stream followed by a moderate growth phase that flattened
into a plateau by the end of the experiment. Specifically,
during the first 6 h on stream, 3V/Al2O3 and 13V/S10 added re-
spectively 73 and 65% of their total weight gain. Further, 5V/
S40 sustained a slow increase through the whole test, whereas
5V/S70 and 5V/MCM-41 gained very little weight.
Besides H2O, H2 was observed during the second step of
weight loss indicating beginning of nonoxidative DH reaction.
Simultaneous evolution of water and hydrogen stems from the
fact that oxidative and nonoxidative dehydrogenation run con-
currently at the earliest stage of the reaction over initially oxi-
dized catalysts. After approximately 20 min on stream, the in-
tensity of H2O signal drops sharply suggesting that VOx reduc-
To compare the rates of carbon deposition on different cata-
lysts quantitatively, we kinetically evaluated the weight change
ChemCatChem 2015, 7, 1691 – 1700
www.chemcatchem.org
1694
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
curves starting from the second step of water formation (oxi-
dative dehydrogenation of propane). To this end, we used
Equation (2) for describing both oxygen removal from VOx and
coke deposition (descending and ascending parts of the TGA
curves in Figure 4 respectively). The first term in this equation
accounts for coke formation as originally suggested in
Refs.[24–26] for evaluation of similar experiments of propane de-
hydrogenation over Pt- and CrOx-based catalysts. In these stud-
ies, coke formation with its own kinetics was assumed to take
place i) over coke-free surface of the catalysts to yield “carbon”
monolayer as well as ii) over formed coke species resulting in
multilayer deposits. The second term in Equation (2) describes
removal of lattice oxygen from oxidized (+5 or +4) VOx spe-
cies in form of water and carbon oxides.
Dm ¼ Dmcoke  ð1 À expðÀkcokeðmÞ Â tÞ þ kcokeðMÞ Â tÞÀ
DmO  ð1 À expðÀkO  tÞÞ
ð2Þ
~
where mcoke is a weight gain from a monolayer of carbon,
kcoke(m) and kcoke(M) respectively are the rate constants of coke
formation over carbon-free catalyst surface and over already
existing carbon deposits. DmO and kO stand for a weight loss
upon reduction of oxidized VOx species and the rate constant
of this process.
The kinetic model assumed here provides a good descrip-
tion of the weight curves. The experimental TGA data and
their fitting are shown in Figure 5 for 5V/MCM-41 and 4V/S10
selected as examples of the catalysts with very different coking
behavior. The data fitting overlays for other catalysts are given
in Figure S4. Importantly, the goodness of fit of the curves for
all catalysts with the exception of 3V/Al2O3 and 13V/S10 was
not sensitive to the kcoke(M) value. The latter therefore could be
set to zero suggesting that not enough carbon was deposited
to form a monolayer. In the case of 3V/Al2O3 and 13V/S10, in-
clusion of kcoke(M) in the fitting procedure improved the good-
ness of fit significantly indicating that carbon monolayer was
formed on these two catalysts and further coke deposition
took place over that monolayer. Apparently, one reason
behind formation of the monolayer on 3V/Al2O3 and 13V/S10
is their low specific surface area.
Figure 5. Comparison between calculated (line) and experimental (open cir-
cles) TGA profiles recorded during propane DH over a) 5V/MCM-41 and
b)4 V/S10. Experimental conditions as in Figure 4.
7.07·10À3 minÀ1 as V loading increases from 4.2 to 13.2 wt% for
the V/S10 catalysts.
Factors governing coke formation and on-stream stability
As seen from Table 1, the obtained values of kcoke(M) are in
general significantly lower than those of kcoke(m) meaning that
formation of coke on already deposited carbon layer proceeds
much slower than on carbon-free surfaces of the catalysts. It is
in agreement with the results previously obtained for coke for-
mation in propane DH over Cr2O3/Al2O3 described by the same
kinetic model.[25] Two important observations regarding the
rate constant of coke formation over carbon-free surfaces
should be emphasized:
Catalyst acidity is often reported to be one of the important
parameters determining coke formation and deactivation in
nonoxidative dehydrogenation of C2–C4 alkanes.[50–55] To test
this hypothesis, we constructed plots showing the relationship
between surface-normalized catalyst acidity and the rate con-
stants of coke formation kcoke(m) (Figure 6a) and catalyst deac-
tivation kdeac (Figure 6b). Despite the fact that most acidic 3V/
Al2O3 and 13V/S10 deactivated faster than all other catalysts,
no general correlation for all tested catalysts could be dis-
cerned. Yet, if kcoke(m) and kdeac are plotted versus EEdge of VOx
species, one can see trends holding for all catalysts (Figure 6c
and d). Both constants increase as EEdge decreases suggesting
that the degree of oligomerization of surface VOx species af-
fects coke formation and catalyst deactivation, that is, the
For the catalysts with V loading of 3–5 wt%, kcoke(m) increas-
es with Al content in the support from the minimum on 5V/
MCM-41 (0.5”10À3 minÀ1) to the maximum on 3 V/Al2O3 (5.49”
10À3 minÀ1).
Vanadium loading and consequently its surface density also
influence the kcoke(m) constant. It rises from 1.96”10À3 to
ChemCatChem 2015, 7, 1691 – 1700
www.chemcatchem.org
1695
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
Figure 6. Rate constants of a, c) coke formation (kcoke(m)) and b, d) catalyst deactivation (kdeac) versus a, b) surface-normalized catalyst acidity and c, d) edge
energy of VOx species. Catalyst numbers are defined in Figure 2.
larger the species, the faster these processes. Remarkably, loss
of activity accelerates as carbon deposition mounts, which be-
comes especially visible if kdeac is plotted versus kcoke(m) as in
Figure 7. This correlation suggests that carbon deposits directly
on the surface of VOx sites shielding them from propane and
thus resulting in catalyst deactivation.
ly, and the respective pristine supports were investigated. After
exposing the above catalysts and supports to the C3H6/N2 feed
for 3 h, the spent materials were analyzed by TPO to deter-
mine the amount of carbon deposits. The amount of deposited
carbon is found in Table 2.
Zero propylene conversion was observed over pristine MCM-
41, which remained snow-white after 3 h on stream. TPO anal-
ysis did not reveal any carbon-containing species on this mate-
rial. Contrarily, propylene conversion and carbon deposits were
observed on 5V/MCM-41. Therefore, we can safely conclude
that coke formation on 5V/MCM-41 during the catalytic tests
with propylene and propane occurred exclusively on VOx spe-
cies. In the case of 4V/S10, coke was formed on both the sup-
To deepen our understanding of the effect of VOx species
size on coke formation, we performed catalytic tests with
a feed containing propylene (11.4 vol.% C3H6 in N2) as it is
known to be a coke precursor. To distinguish between the
coke formed on the supports and on VOx, the catalysts with
highly dispersed, moderately, and highly polymerized VOx spe-
cies, that is, 5V/MCM-41, 4V/S10 and 3V/Al2O3, corresponding-
ChemCatChem 2015, 7, 1691 – 1700
www.chemcatchem.org
1696
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
tween the values of 3V/Al2O3 and 5V/MCM-41, which matches
the arrangement of the catalysts’ VOx species according to
their size.
Based on the fact that coke deposition takes place on VOx
species, the effect of their size on coking rate can be explained
as follows. From a mechanistic viewpoint, adsorbed propylene
molecules interact with each other to initially form small aro-
matic structure followed by further oligomerization and con-
densation to large graphitic structures.[9] As for isolated VOx
sites, they should be located in close proximity to each other
to oligomerize. Alternatively, adsorbed propylene or its frag-
ments may migrate from one VOx species to another, on which
they can further transform into coke. Upon increasing apparent
V surface density and the size of VOx species, a probability of
interaction between several adsorbed propylene molecules in-
creases resulting in a higher rate of carbon deposition. The
highest rate will be achieved if the support surface is com-
pletely covered by highly polymerized VOx species or crystal-
line V2O5 as in 3V/Al2O3. Exactly this trend was observed exper-
imentally.
Figure 7. Catalyst deactivation constant kdeac versus rate constant kcoke of
coke formation. Catalyst numbers are defined in Figure 2.
Conclusions
Table 2. Initial rate of carbon formation (rV(carbon)) and overall amount
of carbon (m(carbon)) formed after 3 h in 11.5 vol.% C3H6/N2 flow at
Nonoxidative propane dehydrogenation to propylene over VOx
species supported on Al2O3–SiO2 with SiO2 content varying
from 0 to 100 wt.% was studied in conventional catalytic tests
and by in situ thermogravimetric analysis (TGA). Kinetic evalua-
tion of the obtained data combined with characterization of
VOx species allowed novel insight into coke formation. Irre-
spective of the kind of the supported VOx species, they are
more active in coke formation from propylene than acidic sites
of the supports. The kinetic analysis of TGA profiles enabled us
to conclude that carbon monolayer–multilayer model quite ac-
curately described this process. A clear correlation between
the rate constant of carbon deposition and that of catalyst de-
activation was found thus verifying that resistance to coking
determines catalysts on-stream stability. Importantly, the rate
constant of carbon deposition strongly increased with the
degree of VOx polymerization reaching the maximum on nano-
crystalline V2O5. This tendency was rationalized through a gen-
eral mechanism of coke formation. To form carbon precursors
from adsorbed propylene or its fragments, they should be lo-
cated in close proximity to each other. Such situation is easily
realized on large VOx species, whereas on the smaller ones it is
statistically less probable. As coke formation proceeds faster
on larger VOx particles, the catalysts designed for long-term
stability in propane dehydrogenation shall have high coverage
of the support with small or even isolated VOx species.
5508C and WHSV of 2.6 hÀ1
.
Catalyst
m(carbon)
rV(carbon) per weight V
[mggÀ1]
[mg(C)g(V)À1 minÀ1
]
Pure support
Catalyst
3V/Al2O3
4V/S10
5V/MCM-41
8.4
43.7
0
74.0
199.3
128.2
1150
800
650
port and VOx. However, the amount deposited on 4V/S10 was
4.2 times higher than that on pristine S10. This evidences that
VOx species on this catalyst are significantly more active in
coke formation than the support, especially given that the sup-
port is partially covered by VOx. In the case of 3V/Al2O3, for
which V coverage exceeds one monolayer, coke formation
occurs entirely on VOx; the amount of carbon on the catalyst
was 8.8 times higher than on pristine Al2O3. These comparisons
certainly verify that VOx species function as the sites for coke
formation and are substantially more active in this process
than acidic sites on S10 and g-Al2O3. The fact that the amount
of carbon deposited on 3V/Al2O3 and Al2O3 is lower than on
other catalysts and pristine S10 stems from their fast deactiva-
tion.
Returning to the correlation between size and activity of VOx
species in coke formation, we calculated the initial rate (rV) of
carbon formation per weight of V [Eq. (9)] for estimating intrin-
sic activity of differently sized VOx species. The activity of acidic
sites on S10 was neglected for the reasons discussed above.
The rate values found in Table 2 indicate that carbon deposi-
tion on 3V/Al2O3 proceeds much faster than on 5V/MCM-41.
Recalling that a certain fraction of V atoms on the former cata-
lyst is buried in multilayers, the difference between the cata-
lysts must be even larger. The rV obtained on 4V/S10 falls be-
Experimental Section
Preparation of catalytic materials
To prepare supported catalysts we used g-Al2O3 (NorPro Saint-
Gobain), SiO2 (MCM-41) and SiO2–Al2O3 (Siral1, 10, 40, and 70, pro-
vided by Sasol) as supports. The index in the Siral formula stands
for a weight percentage of SiO2. MCM-41 was synthesized as de-
scribed in Ref. [56]. The support densities starting from Al2O3 and
ChemCatChem 2015, 7, 1691 – 1700
www.chemcatchem.org
1697
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
going in the direction of increasing SiO2 content were 0.641, 0.591,
0.517, 0.352, 0.313, and finally 0.303 gcmÀ3 for MCM-41. The proce-
dure from Ref. [16] was used for deposition of VOx species on the
supports. Briefly, vanadyl acetylacetonate (VO(acac)2, 2.6 g) was dis-
solved in dry toluene (1 L). Hereafter, a dried support (10 g) was
added to this solution, which was then stirred for 24 h. Then the
solid was isolated, washed in toluene twice, dried, and calcined in
air at 5508C for 12 h. To deposit higher loadings of VOx on Siral10,
we used amounts of VO(acac)2 of 9.4 or 18.8 g to prepare 1 L tolu-
ene solutions. The catalysts are abbreviated as 3V/Al2O3, 4V/S1,
4V/S10, 6V/S10, 13V/S10, 5V/S40, 5V/S70, and 5V/MCM-41 with
vanadium weight content rounded to the whole number shown in
front of the abbreviation. The exact content is given in Table 1.
pyridine. The IR spectra of adsorbed pyridine were recorded at
1508C.
Continuous-flow catalytic tests
Catalytic propane DH tests were performed in a multichannel
setup equipped with 15 plug–flow fixed-bed quartz tube reactors
(Øid =3.8 mm) operated in parallel. Each reactor was filled with
300 mg (315–710 mm grain fraction) of the fresh catalyst. The cata-
lysts were heated to 5508C (5 KminÀ1) in synthetic air (10 mLminÀ1
per reactor), held in the air flow for 1 h and then purged with N2
for 20 min. Hereafter a mixture of 40 vol% C3H8 in N2 was fed at
a rate of 6 mLminÀ1 per reactor. All catalytic experiments were per-
formed at a weight hourly space velocity (WHSV) of 0.94 hÀ1 and
a pressure of 1.2 bar. The experiments were run for 25 h followed
by cooling the reactors to RT in N2. Catalytic experiments aimed to
study coking behavior of selected catalysts and their respective
supports were performed in a feed containing 11.5 vol% of propyl-
ene in nitrogen at 5508C. This propylene concentration was deter-
mined to be an average in propane DH tests. The experiments de-
signed to quantify the amount of carbon deposited on catalysts
and supports were performed at WHSV of 0.26 hÀ1 and lasted for
3 h. To find the initial propylene conversion rates, 5V/MCM-41, 4V/
S10 and 3V/Al2O3 were tested at WHSV of 2.6 hÀ1 for 1 h. The initial
rates of propylene conversion were calculated as mol(C3H6) con-
verted per mol(V) per hour at zero time on stream. The initial con-
version values were obtained from fitting the X(C3H6) curves by
Equation (1).
Catalyst characterization
Nitrogen adsorption isotherms collected at 77 K on BELSORP-mini
II (BEL Japan) were used to calculate the specific surface area of
the calcined catalysts applying the Brunauer, Emmet and Teller
(BET) equation for N2 relative pressure range of 0.05<P/P0 <0.30.
The BET values are provided in Table 1.
The vanadium content of the calcined catalysts was determined by
inductively coupled plasma optical emission spectroscopy (ICP–
OES, Varian 715-ES).
X-ray diffraction was performed on the theta/theta diffractometer
X’PertPro from Panalytical (Almelo; The Netherlands) equipped
with the X’celerator RTMS detector system using Cu Ka 1/2 radia-
tion (l=1.5418 Š, 40 kV, 40 mA). The alignment was checked
against a silicon standard. The data were collected in the 2q range
of 5–708 with a total measurement time of 12 min. The phase com-
position of the samples was determined using the program suite
WINXPow by STOE&CIE with inclusion of the Powder Diffraction
File PDF2 of the ICDD (International Centre of Diffraction Data).
The Scherrer equation was applied to calculate the V2O5 [PDF#00–
074–1595] average crystallite size from the reflection at 2q of
26.138.
The feed components and the reaction products were analyzed by
an on-line gas chromatograph (Agilent 6890N) equipped with
PLOT/Q (for CO2), HP-PLOT Al2O3 “KCl” (for hydrocarbons) and Mol-
sieve5 (for H2, O2, N2, and CO) columns, flame ionization and ther-
mal conductivity detectors. The analyses were performed succes-
sively, reactor by reactor. The propane conversion (X(C3H8)) was cal-
inlet
3H8
culated from the inlet (nC ) and outlet (noCutlet) propane molar
_
_
3H8
flows [Eq. (4)]. The outlet flow was corrected with respect to the
flow of nitrogen to take into account reaction-induced changes in
the number of moles of gas-phase components. The propylene
yield (Y(C3H6)) and selectivity to propylene (S(C3H6)) and carbon
(S(carbon) were calculated according to Equations (5), (6), and (7),
respectively. Yield and selectivity of the side products methane,
ethane, ethylene, C4 and C5 hydrocarbons, CO, and CO2 were calcu-
lated accordingly taking into account their different numbers of C
atoms.
The UV/Vis spectra of fresh catalysts were collected at 5508C using
an in-house developed setup described in Ref. [57]. Prior to the
analysis, the samples were conditioned at the same temperature in
a flow of synthetic air for 1 h. The AVASPEC UV/Vis spectrometer
(Avantes) used in the study was equipped with a DH-2000 deuteri-
um-halogen light source and a CCD array detector. BaSO4 was
used as a white reference standard. All UV/Vis spectra are present-
ed as Kubelka-Munk function F(R ) calculated according to Equa-
tion (3),
1
inlet
C3H8
outlet
C3H8
_
_
n
À n
ð4Þ
ð5Þ
ð6Þ
XðC3H8Þ ¼
YðC3H6Þ ¼
SðC3H6Þ ¼
inlet
3H8
_
nC
2
ð1 À R Þ
1
outlet
C3H6
ð3Þ
FðR Þ ¼
_
n
1
2 Â R
1
inlet
_
nC
3H8
where R is reflection.
outlet
C3H6
1
_
n
inlet
outlet
3H8
_
_
For determining catalyst acidity, we performed IR measurements of
adsorbed pyridine. They were performed in transmission mode on
a Bruker Tensor 27 FT IR spectrometer equipped with a heated and
evacuated custom made reaction cell with CaF2 windows connect-
ed to a gas dosing and evacuation system. The powdery samples
(50 mg) were pressed into self-supporting wafers with a diameter
of 20 mm. Before pyridine adsorption, the samples were pretreated
at 4008C for 10 min in synthetic air followed by cooling to 1508C
and evacuation. Pyridine was adsorbed at 1508C until saturation.
Then the reaction cell was evacuated for removing physisorbed
nC À nC
3H8
P
XðC3H8Þ À Yi
XðC3H8Þi
ð7Þ
SðcarbonÞ ¼
X
YðcarbonÞ ¼ XðC3H6Þ À
Yi
ð8Þ
ð9Þ
i
_
YðcarbonÞ Â mðC3H6Þ
rV ðcarbonÞ ¼
mV
ChemCatChem 2015, 7, 1691 – 1700
www.chemcatchem.org
1698
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
outlet
3H6
[10] J. McGregor, Z. Huang, G. Shiko, L. F. Gladden, R. S. Stein, M. J. Duer, Z.
Wu, P. C. Stair, S. Rugmini, S. D. Jackson, Catal. Today 2009, 142, 143–
151.
_
_
Where nC , mðC3H6Þand mV are molar flow of propylene at the re-
actor outlet, weight flow of propylene at the reactor inlet, and
weigh of vanadium in reactor respectively. Subscript i stands for
any gas-phase products measured.
[11] Z. Wu, P. C. Stair, J. Catal. 2006, 237, 220–229.
[12] M. Volpe, G. Tonetto, H. de Lasa, Appl. Catal. A 2004, 272, 69–78.
[13] N. Ballarini, F. Cavani, M. Ferrari, R. Catani, U. Cornaro, J. Catal. 2003,
213, 95–102.
[14] N. Ballarini, F. Cavani, C. Cortelli, C. Giunchi, P. Nobili, F. Trifirò, R. Catani,
U. Cornaro, Catal. Today 2003, 78, 353–364.
In situ thermogravimetric and temperature-programmed
tests
[15] S. Sokolov, M. Stoyanova, U. Rodemerck, D. Linke, E. V. Kondratenko, J.
Catal. 2012, 293, 67–75.
[16] S. Sokolov, M. Stoyanova, U. Rodemerck, D. Linke, E. V. Kondratenko,
Catal. Sci. Technol. 2014, 4, 1323–1332.
[17] E. Daniel, E. Resasco, G. L. Haller in Catalysis, Vol. 11 (Eds.: J. J. Spivey,
S. K. Agarwal), Royal Society of Chemistry, Cambridge, 1994, pp. 379–
411.
[18] O. A. Bariås, A. Holmen, E. A. Blekkan, J. Catal. 1996, 158, 1–12.
[19] A. Iglesias-Juez, A. M. Beale, K. Maaijen, T. C. Weng, P. Glatzel, B. M.
Weckhuysen, J. Catal. 2010, 276, 268–279.
The thermogravimetric setup described in Ref. [20] consisted of
a thermobalance SETSYS Evolution (Setaram) and an on-line mass
spectrometer OmniStar GSD 301 (Pfeiffer). A catalyst sample
(50 mg) was loaded into a quartz cap suspended inside a thermo-
balance furnace. The sample was heated to 5508C at 10 KminÀ1 in
a flow of 7.7% O2 in He and held at this temperature for 12 h to
oxidize VOx species. After cooling to RT, a C3H8:He=50:50 mixture
was fed at 20 mLminÀ1 for approximately 30 min. Hereafter, the re-
actor was heated to 5508C at 30 KminÀ1 in the same flow and held
at this temperature for 24 h. Weight of the sample was continuous-
ly measured over the entire experiment. The reaction products and
propane were monitored by mass spectrometry.
[20] V. Y. Bychkov, Y. P. Tyulenin, V. N. Korchak, E. L. Aptekar, Appl. Catal. A
2006, 304, 21–29.
[21] V. Y. Bychkov, Y. P. Tyulenin, A. A. Firsova, E. A. Shafranovsky, A. Y. Goren-
berg, V. N. Korchak, Appl. Catal. A 2013, 453, 71–79.
[22] S. Chen, G. Manos, J. Catal. 2004, 226, 343–350.
[23] A. Platon, H.-S. Roh, D. King, Y. Wang, Top. Catal. 2007, 46, 374–379.
[24] J. A. PeÇa, A. Monzón, J. Santamaría, J. L. G. Fierro, Appl. Catal. A 1993,
101, 185–198.
[25] J. Gascón, C. TØllez, J. Herguido, M. MenØndez, Appl. Catal. A 2003, 248,
105–116.
[26] M. van Sint Annaland, J. A. M. Kuipers, W. P. M. van Swaaij, Catal. Today
2001, 66, 427–436.
[27] M. Santhosh Kumar, A. Holmen, D. Chen, Microporous Mesoporous
Mater. 2009, 126, 152–158.
[28] M. Santhosh Kumar, D. Chen, A. Holmen, J. C. Walmsley, Catal. Today
2009, 142, 17–23.
[29] U. Scharf, M. Schraml-Marth, A. Wokaun, A. Baiker, J. Chem. Soc. Faraday
Trans. 1991, 87, 3299–3307.
[30] F. Arena, F. Frusteri, G. Martra, S. Coluccia, A. Parmaliana, J. Chem. Soc.
Faraday Trans. 1997, 93, 3849–3854.
The amount of carbon deposits formed during catalytic experi-
ments was determined from the amount of CO and CO2 evolved
during temperature-programmed oxidation (TPO) of spent cata-
lysts. TPO measurements were performed in a setup equipped
with eight individually heated plug–flow fixed-bed quartz tube re-
actors of 6.9 mm inner diameter. The used catalysts were heated to
9008C at 10 KminÀ1 in a flow (10 mLminÀ1) of 10 vol.% O2 in Ar.
Oxygen consumption and formation of reaction products were
monitored by a quadrupole mass spectrometer (Pfeiffer Vacuum
OmniStar 200). The following atomic mass units (AMUs) were ana-
lyzed: 44 (CO2), 40 (Ar), 32 (O2), 28 (CO, CO2), and 18 (H2O). The
concentrations of O2, CO, and CO2 were determined from the re-
spective AMUs using standard fragmentation patterns and sensitiv-
ity factors determined by analyzing calibration gas mixtures.
[31] G. Catana, R. R. Rao, B. M. Weckhuysen, P. Van Der Voort, E. Vansant,
R. A. Schoonheydt, J. Phys. Chem. B 1998, 102, 8005–8012.
[32] X. Gao, I. E. Wachs, J. Phys. Chem. B 2000, 104, 1261–1268.
[33] M. Baltes, K. Cassiers, P. Van Der Voort, B. M. Weckhuysen, R. A. Schoon-
heydt, E. F. Vansant, J. Catal. 2001, 197, 160–171.
[34] E. V. Kondratenko, M. Baerns, Appl. Catal. A 2001, 222, 133–143.
[35] N. Steinfeldt, D. Muller, H. Berndt, Appl. Catal. A 2004, 272, 201–213.
[36] E. V. Kondratenko, M. Cherian, M. Baerns, X. Su, R. Schlçgl, X. Wang, I. E.
Wachs, J. Catal. 2005, 234, 131–142.
Acknowledgements
The authors are thankful to A. Simmula, M. Schneider, C. Rauten-
berg for assistance with sample characterization.
Keywords: carbon · dehydrogenation · kinetics · monolayers ·
vanadium
[37] E. L. Lee, I. E. Wachs, J. Phys. Chem. C 2007, 111, 14410–14425.
[38] Z. Wu, H.-S. Kim, P. C. Stair, S. Rugmini, S. D. Jackson, J. Phys. Chem. B
2005, 109, 2793–2800.
[39] B. M. Weckhuysen, D. E. Keller, Catal. Today 2003, 78, 25–46.
[1] W.-C. Cheng, E. T. Habib, Jr., K. Rajagopalan, T. G. Roberie, R. F. Worms-
becher, M. S. Ziebarth in Handbook of Heterogeneous catalysis, Vol. 6
(Eds.: G. Ertl, H. Knçzinger, F. Schüth, J. Weitkamp), Wiley-VCH, Wein-
heim, 2008, pp. 2741–2777.
[2] K. J. Caspary, H. Gehrke, M. Heinritz-Adrian, M. Schwefer in Handbook of
Heterogeneous Catalysis, Vol. 7 (Eds.: G. Ertl, H. Knçzinger, F. Schüth, J.
Weitkamp), Wiley-VCH, Weinheim, 2008, pp. 3206–3228.
[3] J. C. Mol, P. W. N. M. van Leeuwen in Handbook of Heterogeneous Catal-
ysis, Vol. 7 (Eds.: G. Ertl, H. Knçzinger, F. Schüth, J. Weitkamp), Wiley-
VCH, Weiheim, 2008, p. 3240.
[4] A. Aitani in Encyclopedia of Chemical Processing, Vol. 4 (Ed.: S. Lee), Tay-
lor&Francis, New York, 2006, p. 2461.
[5] M. E. Harlin, V. M. Niemi, A. O. I. Krause, J. Catal. 2000, 195, 67–78.
[6] M. E. Harlin, V. M. Niemi, A. O. I. Krause, B. M. Weckhuysen, J. Catal.
2001, 203, 242–252.
´
[40] J. Słoczynski, R. Grabowski, A. Kozłowska, R. Tokarz-Sobieraj, M. Witko, J.
Mol. Catal. A 2007, 277, 27–34.
ˇ
ˇ
ˇ
ˇ
[41] R. Bulµnek, L. Capek, M. Setnicka, P. Cicmanec, J. Phys. Chem. C 2011,
115, 12430–12438.
[42] W. N. Delgass, G. L. Haller, R. Kellerman, J. H. Lunsford, Spectroscopy in
Heterogeneous Catalysis, Academic Press, New York, 1979.
[43] A. Pantazidis, S. A. Bucholz, H. W. Zanthoff, Y. Schuurman, C. Mirodatos,
Catal. Today 1998, 40, 207–214.
[44] D. Creaser, B. Andersson, R. R. Hudgins, P. L. Silveston, J. Catal. 1999,
182, 264–269.
[45] K. Chen, E. Iglesia, A. T. Bell, J. Catal. 2000, 192, 197–203.
[46] K. Chen, A. T. Bell, E. Iglesia, J. Catal. 2002, 209, 35–42.
[47] R. Grabowski, J. SloczyÇski, N. M. Grzesik, Appl. Catal. A 2002, 232, 277–
288.
[7] H. Kwon, L. T. Thompson, J. Eng, Jr., J. G. Chen, J. Catal. 2000, 190, 60–
68.
[8] S. D. Jackson, S. Rugmini, J. Catal. 2007, 251, 59–68.
[9] J. J. H. B. Sattler, J. Ruiz-Martinez, E. Santillan-Jimenez, B. M. Weckhuy-
sen, Chem. Rev. 2014, 114, 10613–10653.
[48] N. Ballarini, F. Cavani, A. Cericola, C. Cortelli, M. Ferrari, F. Trifirò, G. Capa-
nnelli, A. Comite, R. Catani, U. Cornaro, Catal. Today 2004, 91–92, 99–
104.
[49] E. V. Kondratenko, N. Steinfeldt, M. Baerns, Phys. Chem. Chem. Phys.
2006, 8, 1624–1633.
ChemCatChem 2015, 7, 1691 – 1700
www.chemcatchem.org
1699
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Papers
[50] M. M. Bhasin, J. H. McCain, B. V. Vora, T. Imai, P. R. Pujadó, Appl. Catal. A
2001, 221, 397–419.
[55] S. A. Bocanegra, A. Guerrero-Ruiz, S. R. de Miguel, O. A. Scelza, Appl.
Catal. A 2004, 277, 11–22.
[51] J. Salmones, J.-A. Wang, J. A. Galicia, L. Balderas, G. Aguilar-Rios, J. Mol.
Catal. A 2002, 184, 203–213.
[56] G. Grubert, J. Rathousky, G. Schulz-Ekloff, M. Wark, A. Zukal, Microporous
Mesoporous Mater. 1998, 22, 225–236.
[52] H. Armendµriz, A. Guzmµn, J. A. Toledo, M. E. Llanos, A. Vµzquez, G.
Aguilar-Ríos, Appl. Catal. A 2001, 211, 69–80.
[57] E. V. Kondratenko, Catal. Today 2010, 157, 16–23.
[53] J. Llorca, N. Homs, J. León, J. Sales, J. L. G. Fierro, P. Ramirez de La Pisci-
na, Appl. Catal. A 1999, 189, 77–86.
[54] A. D. Ballarini, S. A. Bocanegra, A. A. Castro, S. R. de Miguel, O. A. Scelza,
Catal. Lett. 2009, 129, 293–302.
Received: February 16, 2015
Revised: March 25, 2015
Published online on May 12, 2015
ChemCatChem 2015, 7, 1691 – 1700
www.chemcatchem.org
1700
ꢀ 2015 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim