Applied Catalysis A: General 371 (2009) 31–42

Contents lists available at ScienceDirect

Applied Catalysis A: General

journal homepage: www.elsevier.com/locate/apcata

Transient kinetic modeling of the oxidative dehydrogenation of propane over a

vanadia-based catalyst in the absence of O2
V. Balcaen, I. Sack, M. Olea 1, G.B. Marin *
Laboratory for Chemical Technology, Department of Chemical Engineering, Ghent University, Krijgslaan 281, B-9000 Ghent, Belgium

A R T I C L E I N F O A B S T R A C T

Article history: The oxidative dehydrogenation of propane in the absence of O2 over the vanadia-based EL10V1 Eurocat
Received 16 April 2009 catalyst is investigated in a Temporal Analysis of Products (TAP) reactor over a completely oxidized
Received in revised form 22 July 2009 catalyst and at reduction degrees up to 0.47. Only CO2 and propene are detected as reaction products at
Accepted 11 September 2009
temperatures from 723 to 823 K. The reduction of the catalyst with propane occurs through the
Available online 17 September 2009
formation of propene, the consecutive oxidation of propene to CO2 and the parallel total oxidation of
propane. Four elementary reactions are considered as kinetically relevant: (1) the methyl C–H bond
Keywords:
dissociation in propane during the formation of propene, (2) the methylene C–H bond breaking in
Vanadia-based catalyst
Reduction
propane to form surface isopropoxide, which finally transforms into CO2, (3) the C5 5C double bond
Propane breaking in propene in the formation of surface formate species and (4) the oxidation of these surface
TAP formate species to CO2. The formation of propene via (1) is favored at the investigated conditions, while
Kinetic modeling CO2 is largely produced by the sequential oxidation of propene (3–4), and to a lesser extent by the
parallel route of the direct total oxidation of propane (2). Both the oxidative dehydrogenation and the
direct total oxidation of propane reaction paths involve only one kinetically relevant step, (1)
respectively (2), with an activation energy of 36 and 74 kJ mol1 over a completely oxidized catalyst and
of 15 and 40 kJ mol1 over a partially reduced catalyst. The further oxidation of propene involves two
kinetically relevant steps (3–4) of which only the double bond breaking (3) is slightly activated
(5 kJ mol1) over the completely oxidized catalyst, while both steps are non-activated when the catalyst
is partially reduced.
ß 2009 Elsevier B.V. All rights reserved.

1. Introduction being reduced to a certain extent while the hydrocarbon is

oxidized by the oxygen from the catalyst lattice. After separation of
Vanadia-based catalysts show high activity in various partial the oxidation products, the reduced catalyst is regenerated with
oxidation and oxidative dehydrogenation reactions of hydrocar- oxygen or air in a separate reactor. Since in this moving bed
bons [1]. If the oxidation process is carried out in the presence of technique, the reduction and reoxidation are separated, a
oxygen in a fixed bed reactor, the production capacity is limited comprehensive kinetic study of both reactions is necessary.
because of the explosion limits. Moreover, the selectivity is limited The influence of the partially reduced state of the catalyst on the
by total oxidation side reactions. A particular way to increase the kinetics of the reoxidation of the EL10V1 Eurocat catalyst and of
production capacity and to suppress or limit the total oxidation some other vanadia-based catalysts [3,4] has already been studied.
reactions consists of performing the oxidation reaction in the In this paper, a TAP reactor set-up is employed to quantify the
absence of gas-phase oxygen. The latter can only be accomplished influence of a change in the catalyst state from the completely
if the oxidation reaction occurs through the Mars–van Krevelen oxidized to a partially reduced one, on the kinetics of the propane
mechanism [2], i.e. through the cyclic reduction and reoxidation of oxidative dehydrogenation reaction over the vanadia-based
the catalyst. In the so-called moving bed reactor, the catalyst and EL10V1 Eurocat catalyst.
the hydrocarbon move together through the reactor, the catalyst
2. Experimental

2.1. EL10V1 Eurocat catalyst

* Corresponding author. Tel.: +32 9 264 4516; fax: +32 9 264 4999.
E-mail address: [email protected] (G.B. Marin).
1
Present address: School of Science and Technology, University of Teesside, The EL10V1 Eurocat catalyst, i.e. 1% V2O5/TiO2 calcined, was
Borough Road, Middlesbrough TS1 3BA, UK. prepared by Rhône-Poulenc and exhaustively characterized by 25

0926-860X/$ – see front matter ß 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.apcata.2009.09.014
32 V. Balcaen et al. / Applied Catalysis A: General 371 (2009) 31–42

Nomenclature
Greek letters
3 2 1
Di shifted residence time of gas i [s]
A pre-exponential factor [mg molV mol s ] or ec void fraction of the catalyst packing [mg3 mr3]
[molV mol1 s1] ei void fraction of the inert packing [mg3 mr3]
Ci gas-phase concentration of component i rB density of the catalyst packing [kgc mr3]
[mol mg3] si stoichiometric number of an elementary reaction
C i;k ðt j Þ gas-phase concentration of component i corre- in global reaction i
sponding to observation k at sampling time tj sii stoichiometric number of an elementary reaction
[mol mg3] or a global reaction in reaction path ii
C i surface concentration of active sites occupied with t time constant of the inlet pulse [s]
species i [mol kgc1] ti,diff residence time of gas i of a purely diffusional
C O surface concentration of free oxidized active sites process [s]
[mol kgc1] K objective function
C* surface concentration of free reduced active sites
[mol kgc1] Superscripts
Ct total concentration of active VOx sites [mol kgc1] 0 initial condition
Dceff;i effective Knudsen diffusivity in the catalyst bed of ^ calculated value
component i [mg3 mr1 s1]
Dieff;i effective Knudsen diffusivity in the inert bed of Subscripts
component i [mg3 mr1 s1] k kth observation
Ea activation energy [kJ mol1] ref reference
F tabulated value to account for the significance of
the regression
Fi flow rate of gas i [mol s1] European laboratories. A short summary of the catalyst prepara-
F i;k ðt j Þ flow rate of component i corresponding to obser- tion and its properties is presented by Garcin et al. [5]. The BET
vation k at sampling time tj [mol s1] surface area amounts to 10.1 (0.5) m2 g1 [6]. The V content
i index of infinite series reported by Nogier et al. [7] (0.107  0.006 mol V kgc1) is in good
j index of infinite series agreement with the value obtained by Sack through ICP measure-
k index of infinite series ments, who reported 0.108 mol V kgc1 [8].
kn reaction rate coefficient in reaction n during the Busca et al. [9] concluded from IR and Raman spectroscopic
studies that EL10V1 consists of a monolayer of surface vanadyl
reduction of the catalyst with propane
complexes spread on the TiO2 support surface, indicating that all
[mg3 molV mol2 s1] or [molV mol1 s1]
vanadia is present at the surface of the catalyst. Such a monolayer
Lc length of the catalyst bed [mr]
contains both Lewis and Brønsted acid sites, while the nucleophilic
Li length of the inert bed at the inlet [mr] sites of the support surface are suppressed.
Lo length of the inert bed at the outlet [mr] In the vanadia layer, different kinds of oxygen atoms are present
L length of the reactor [mr] (see Fig. 1). An oxygen atom can be attached to the vanadium with
m number of gas-phase components a double bond, giving rise to vanadyl oxygen, O1, which is present
M0,i zeroth moment of the outlet flow rate of gas i [mol] on top of the vanadia layer. Alternatively, oxygen can be bound to
M1,i first moment of the outlet flow rate of gas i [mol s] two vanadium atoms, O2, or to one vanadium and one support
Np,i number of moles i in the pulse [mol] atom, O3. Based on TAP results, it is not possible to distinguish
n number of observations between the different lattice oxygens of the VOx site. Therefore, the
active VOx site, represented between brackets in Fig. 1, is
nt number of samples per pulse response
considered to consist of one vanadium, V, one vanadyl oxygen,
R universal gas constant [8.314 J mol1 K1]
O1, one bridging V–O–V oxygen, O2, and one V–O–Ti oxygen, O3,
R reduction degree assuming that only one oxygen atom is active per vanadium atom,
Rg,i specific net production rate of gas-phase compo-
nent i [mol kgc1 s1]
Rs,i* specific net production rate of surface species i
[mol kgc1 s1]
S cross-sectional surface area of the reactor [mr2]
t time [s]
tj time between the pulse and the measurement of
sample j, i.e. sampling time [s]
Dt sampling interval [s]
T temperature of packed bed [K]
V0 volume of the gas in the inlet section of the reactor
[mg3]
wi weight factor of component i
Fig. 1. Schematic representation of polymeric V2O5. For one V atom, three different
z axial coordinate in the reactor [mr] oxygen atoms in the same relative amounts are present: vanadyl oxygen, O1,
bridging V–O–V oxygen, O2, and bridging V–O–support oxygen, O3. The structure
between brackets is considered as the oxidized active site, O*.
 V. Balcaen et al. / Applied Catalysis A: General 371 (2009) 31–42 33

irrespective of its nature. The concentration of the active VOx sites, experiments will be used in the modeling to quantify the influence
Ct, is then equal to the concentration of the surface vanadium of the change in the catalyst state on the kinetics of the propane
atoms and can be determined as the vanadium loading is known. oxidative dehydrogenation. The different reduction degrees are set
by performing multi-pulse experiments with respectively 1.5, 3.0,
2.2. TAP set-up, procedures and conditions 5.0 and 7.5  1018 propane molecules at the five temperatures.
Since an identical amount of reducing molecules is pulsed at
The experiments are carried out in an original TAP set-up [10] different temperatures, various reduction degrees are obtained at
using the high-speed pulse valves of the TAP-2 set-up [11]. The TAP each temperature. At each reduction degree, single-pulse experi-
reactor consists of three vacuum chambers in line. The first ments with propane are carried out in order to determine the
chamber is the reactor chamber and contains a quartz micro- mechanism and the reaction kinetics. During the analysis of the
reactor, 45 mm long and 5 mm in diameter. The entrance of this single-pulse experiments, the outlet flow rate of gas A can be
reactor is connected via a small inlet volume with two high-speed expressed in a generalized form by dividing this flow rate by the
pulse valves and two continuous feed valves. The pulse valves number of moles in the inlet pulse of gas A. This generalized form is
allow rapid injection of small amounts of reactants, typically 1014 called the normalized flow rate. After each reduction treatment,
to 1016 molecules/pulse, with a pulse time less than 100 ms. The the catalyst is completely reoxidized by admitting multi-pulses of
background pressure in the reactor chamber varies between 104 oxygen until a constant oxygen level is reached. Additionally,
and 105 Pa, and consists mainly of N2 and H2O. The second propane/oxygen alternating pulse experiments are performed in
chamber is the differential chamber, where the pressure amounts order to examine the presence of possible weakly adsorbed
to 106 Pa acting as pre-vacuum chamber for the analysis chamber. intermediates. Nine different pump-probe intervals ranging from
The responses of the reactants and products at the exit of the 0.01 to 2 s are applied, pulsing propane first (pump molecule)
reactor are measured with a UTI 100C quadrupole mass spectro- followed by oxygen (probe molecule) at 723, 773 and 823 K. More
meter located in the analysis chamber, where the pressure is only information on how a TAP alternating pulse experiment is
107 Pa. This ensures that the signal detected with the mass performed, can be found in Yablonsky et al. [12].
spectrometer is proportional to the flow rate at the outlet of the Since the amount of propene formed during the experiments
reactor. The time resolution at which the evolution of the flow rate with propane is rather small, not all details of the further oxidation
at the outlet of the reactor can be measured is less than 1 ms. of propene can be elucidated using propane as reductant.
TAP experiments are generally performed in the Knudsen Therefore, next to the experiments with propane, a limited
diffusion regime in which transport is well-defined and the outlet number of experiments with propene is also performed. However,
pulse shape is independent of pulse intensity. Therefore, experi- these experiments are not considered during the kinetic modeling
ments over an inert bed, consisting of quartz particles with the as they only aim at obtaining more qualitative information about
same particle size as the catalyst, are performed to determine the the part in the reaction mechanism, starting from propane, where
maximum pulse intensity for which the Knudsen diffusion regime propene reacts further to CO2. Single-pulse experiments with pure
is still valid. Since this maximum pulse intensity amounts to 1016 propene at 723 K over a completely oxidized catalyst surface are
admitted molecules in one pulse, an upper limit of 1015 molecules/ used to acquire a detailed product distribution. Also, propene/
pulse is used in the reported experiments. oxygen alternating pulse experiments in which propene is pulsed
The experiments are typically carried out over 100 mg of first are carried out at the same temperature in order to identify
catalyst, diluted with 100 mg of quartz. This amount of catalyst possible weakly adsorbed species in the consecutive reaction of
corresponds to 6.5  1018 V atoms, i.e. the total number of active propene to CO2. These measurements are performed for the same
sites in the reactor. Since a maximum inlet pulse intensity of nine time intervals as for the propane/oxygen alternating pulse
1015 molecules/pulse is applied in all experiments, the number of experiments.
reactant molecules in one pulse is always three orders of The different gases are monitored at their traditional atomic
magnitude lower than the number of active sites in the reactor. mass unit (AMU) except for CO, which is monitored at AMU 12
If then a limited amount of pulses (25–50) is admitted in a single- during the single-pulse experiments with propane. This can be
pulse experiment, the catalyst state is not significantly altered. The explained by the large fragmentation of propane (59%) detected on
catalyst bed has a density, rB, of 540 kgc mr3 and is placed AMU 28, next to a smaller fragmentation of CO2 (15%), which
between two beds filled with inert quartz (three-zone configura- makes an accurate determination of CO more difficult. On the
tion) using a comparable particle size of 250–500 mm for the contrary, during single-pulse experiments with propene, which
catalyst as for the quartz. The void fraction of the catalyst packing, has a negligible fragmentation on AMU 28, CO is measured at its
ec, equals the void fraction of the inert packing, ei, and amounts to traditional AMU, correcting for the interference of other products
0.25. The fresh catalyst samples are first heated to reaction formed from propene.
temperature with a ramp of 5 K/min under vacuum. Subsequently,
the catalyst is pretreated with multi-pulses of oxygen until a 2.3. Symmetrical thin-zone
constant oxygen response is obtained, meaning that the catalyst is
oxidized. The completely oxidized state of the catalyst is verified For a better understanding of the reaction mechanism, the
by performing single-pulse experiments with a mixture of oxygen oxidative dehydrogenation of propane is also studied on a
and argon after every pretreatment. symmetrical thin-zone TAP configuration. In a thin-zone TAP
The influence of the reduction degree is studied by multi- and reactor, the concentration gradients across the catalyst bed can be
single-pulse experiments with propane over the pretreated neglected. The current configuration has the following character-
catalyst with temperatures varying in steps of 25 K between istics: the first and third zones are filled with inert quartz and are
723 and 823 K. The multi-pulse experiments are performed in 13 mm in length; the second zone is thin (2.78 mm) and filled with
order to change the catalyst state from the completely oxidized to a 61.7 mg of catalyst. For a thin-zone TAP configuration, Constales
partially reduced one, without using these experiments during the et al. [13] reported criteria for instantaneousness of the reaction
kinetic modeling. The single-pulse experiments, on the contrary, in based on shifted residence times of reactants (A) and products (B).
which the catalyst state is not significantly perturbed, are These shifted residence times are equal to the difference between
performed to study the interaction of reactants and products with the residence times of the substances in the presence of reaction,
the catalyst at a well-defined state of the latter. These single-pulse experimentally determined as M1,A/M0,A and M1,B/M0,B, and the
34 V. Balcaen et al. / Applied Catalysis A: General 371 (2009) 31–42

residence times corresponding to a purely diffusional process, i.e. and on the catalyst surface:
tA,diff and tB,diff, multiplied by a factor:
@C A
¼ Rs;A ; Li < z < Li þ Lc (7)
M M @t
DA ¼ 1;A  cA t A;diff DB ¼ 1;B  cB t B;diff (1)
M 0;A M 0;B The boundary condition at the inlet of the reactor, Eq. (8), assumes
where cA = 1/12 and cB = 3/4 are numerical coefficients that are that the inlet section is perfectly mixed and that the gas-phase
functions of the TAP-system configuration and procedure. The concentration in the inlet volume V0 equals the gas-phase
shifted residence times are a measure for the time the substances concentration at the inlet of the first inert bed:
need for the reaction over the catalyst. Intermediates should be    
@C A @C A
considered in the modeling of the reaction sequence where A V0 ¼ SDieff;A þ FA; t  0; z ¼ 0 (8)
@t @z
reacts to B, if the shifted residence times of both substances are
significantly different, i.e. DA < DB. If both substances have the where
same shifted residence time, i.e. DA = DB, the generation of the  
N p;A t
product B is considered instantaneous. FA ¼ t exp  (9)
t 2 t
Initially, single-pulse experiments with argon at 723 K are
performed to determine the diffusion characteristics of this represents the flow rate at the inlet of the reactor with a time
configuration. Next, single-pulse experiments with propane over constant, t, of 1 ms.
the oxidized catalyst are carried out at 723 and 823 K, while single- The boundary condition at the outlet of the reactor, Eq. (10),
pulse experiments with propene over the oxidized catalyst are states that the gas-phase concentration at the outlet is negligible
performed at 723 K only. compared to the concentration in the bed:

2.4. Reduction degree C A ¼ 0; t  0; z ¼ Li þ Lc þ Lo (10)

The boundary conditions between the beds, Eqs. (11)–(14), express

The reduction degree, R, defined as the ratio of the concentra-
that the concentration profiles and the flux profiles are continuous
tion of free reduced active sites, C*, to the total concentration of
functions of z:
active sites, Ct:
ðC A Þz¼Li ¼ ðC A Þz¼Liþ (11)
C
R¼ (2)
Ct    
@C A @C A
is determined as the ratio of the number of O atoms consumed Dieff;A ¼ Dceff;A (12)
from the catalyst during reduction to the number of surface
@z z¼Li @z z¼Liþ

vanadium atoms as found by ICP [8]. The number of O atoms

consumed from the catalyst during reduction is assumed to be ðC A Þz¼ðLi þLc Þ ¼ ðC A Þz¼ðLi þLc Þþ (13)
equal to the number of O atoms consumed during subsequent    
oxygen uptake experiments. The latter is experimentally deter- @C A @C A
Dceff;A ¼ Dieff;A (14)
mined by the titration of the reduced vanadia species with oxygen, @z z¼ðLi þLc Þ @z z¼ðLi þLc Þþ
assuming a dioxygen to vanadium chemisorption stoichiometry of
Eq. (15) describes the initial conditions:
1 to 2. This will only lead to correct values, provided that oxygen is
not used for the oxidation of possibly adsorbed carbonaceous C A ¼ 0; 0
C A ¼ CA ; t ¼ 0; 0  z  Li þ Lc þ Lo (15)
species or adsorbed hydroxyl species. This condition is verified by
monitoring the responses of CO2 and H2O during the multi-pulse The specific net production rates of the gas-phase and the surface
experiments with oxygen. Since neither CO2 nor H2O could be components, respectively Rg,A and Rs,A*, contain the rate coeffi-
observed, it can be concluded that no oxygen reacts with cients to be estimated. Since these coefficients obey the Arrhenius
carbonaceous or adsorbed hydroxyl species and is thus exclusively equation, a pre-exponential factor (A) and activation energy (Ea)
used to reoxidize the surface [3,4]. Hence, the reduction degree can can be estimated for each coefficient. A reparametrization is
be obtained from the following equation: applied to minimize the correlation between the pre-exponential
factor and the activation energy, according to Eq. (16):
mol O2 =2   
R¼ (3) Ea 1 1
mol V kn ¼ kn;T ref exp   (16)
R T T ref
2.5. Kinetic parameter estimation
The parameter estimates are obtained using a combination of the
Rosenbrock [14] and Levenberg-Marquardt [15] methods for the
To propose an appropriate reaction network and to determine
minimization of the objective function, F, Eq. (17):
the corresponding kinetic parameters as a function of the catalyst’s
reduction degree, a non-linear, multivariate multiple regression X
m X
n X
nt
2
analysis is applied. The outlet flow rate of component A is F¼ wi ðF i;k ðt j Þ  F̂i;k ðt j ÞÞ (17)
i¼1 k¼1 j¼1
calculated by integrating the following partial differential equa-
tions for A in the gas phase: where F i;k ðt j Þ ¼ Dieff;i;k Sðð@C i;k ðt j ÞÞ=@zÞ is the observed outlet flow
rate and F̂i;k ðt j Þ ¼ Dieff;i;k Sðð@Ĉi;k ðt j ÞÞ=@zÞ is the calculated flow rate
@C A @2 C A of component i, i.e. of propane, propene and CO2. The flow rates are
ei ¼ Dieff;A ; 0 < z < Li (4)
@t @z2 recorded by the mass spectrometer by taking different samples at
equidistant times: tj = j Dt where Dt is called the sampling interval.
@C A @2 C A
ec ¼ Dceff;A þ rB Rg;A ; Li < z < Li þ Lc (5)
@t @z2 2.6. Numerical integration

@C A @2 C A To minimize the objective function, F, a numerical integration

ei ¼ Dieff;A ; Li þ Lc < z < Li þ Lc þ Lo (6)
@t @z2 of the reactor model equations is necessary. The integration of the
 V. Balcaen et al. / Applied Catalysis A: General 371 (2009) 31–42 35

partial differential equations in this reactor model is performed in

the time domain using a line method. Therefore, the spatial
derivatives are approximated while the time derivatives are
retained. Since the approximation of the spatial derivatives is done
by applying the finite differences method, a grid of points is defined
where the continuous variables are replaced by discrete ones. The
result is a set of ordinary differential equations (ODE) in each grid
point which is solved using LSODA (Livermore Solver for Ordinary
Differential equations), which can automatically switch between
stiff and non-stiff problems [16]. Simultaneously solving the
ordinary differential equations returns the ‘solution lines’ in each
point.
The significance of the regression is measured by means of the
so-called F-test, which is based on the comparison of the calculated
sum of squares and the residual sum of squares. A high F value Fig. 2. Height normalized C3H8 (—), C3H6 (—) and CO2 (. . .) responses corresponding
corresponds to a high significance of the regression. For each to a single-pulse experiment with pure C3H8 over 100 mg completely oxidized
EL10V1 at 773 K.
parameter an individual confidence interval is determined. If the
confidence interval encloses zero, the parameter cannot be
determined significantly different from zero. Moreover, the can be assumed that CO2 either desorbs more slowly than propene
correlation between two parameters, i and j, is determined. If or is formed consecutively. Since the height normalized response
this correlation is higher than 95%, the parameters cannot of propene presented in Fig. 2 decays more rapidly than the CO2
independently be determined and an independent estimate of response, propene is certainly converted to CO2. The direct
one of them should be provided to be able to estimate the other. formation of CO2 from propane however cannot be ruled out.
From a single-pulse experiment with a gas mixture of 52%
3. Results and discussion argon and 48% propane presented in Fig. 3, it can be observed
that the normalized propane response is completely situated
3.1. Determination of the oxidation state of the pretreated catalyst within the response of argon. This indicates that the interaction
of propane with the oxidized catalyst at 773 K occurs through
The oxidation treatment implying multi-pulses of oxygen at the irreversible adsorption or reaction. This type of interaction does
beginning of each reduction–reoxidation cycle is performed for not change with temperature. Therefore, interaction of propane
several reasons. Due to the heating of the fresh catalyst to reaction with free oxidized sites leading to weakly, reversibly adsorbed
temperature and/or to the vacuum environment, the oxidation propane as described by Chen et al. [18] should not be
state of the catalyst can be lower than V5+. Moreover, adsorbed considered. Moreover, in propane/oxygen alternating pulse
species can be present on the fresh catalyst, which can be removed experiments, no weakly adsorbed intermediate is detected,
by pretreatment in oxygen. Therefore, by oxidizing the catalyst confirming the above.
sample before starting the actual reduction experiment, the V5+
state is ensured. The latter can be verified by performing a single- 3.3. Interaction of propene with the oxidized catalyst
pulse experiment with a mixture of oxygen and argon. If there is no
reversible adsorption or reaction of oxygen with the oxidized Fig. 4 shows the formation of acrolein (C3H4O), formaldehyde
sample, the difference between the height normalized responses of (CH2O), CO and CO2 during a single-pulse experiment with pure
oxygen and argon can only be ascribed to the difference in propene over a completely oxidized catalyst at 723 K. Note that in
diffusivity. For all samples investigated, no oxygen interaction is this experiment CO is detected because propene itself is admitted
observed after the pretreatment and hence the reduction experi- over an oxidized catalyst. If instead of propene, propane is
ments always start from a completely oxidized catalyst. introduced into the reactor, practically no CO is detected under the
reaction conditions used in this work (see Section 3.2). Because the
3.2. Interaction of propane with the oxidized catalyst amounts of acrolein and formaldehyde that leave the reactor are
very small, the outlet flow rates show a high noise level, especially
The outlet flow rates corresponding to a single-pulse experi- the one for acrolein. A careful analysis also shows the formation of
ment with pure propane over a completely oxidized catalyst at
773 K are shown in Fig. 2. Propene and CO2 are detected at all
temperatures, while H2O is only observed at the highest
temperature. The amount of CO that can be detected on AMU
12 is very small which results in a low signal-to-noise ratio. It is
verified that this small amount of CO is not due to its measurement
at AMU 12 by monitoring at least once the CO response at AMU 28,
giving rise to a negligible amount of CO. Since the carbon balance is
closed if only the formation of CO2 and propene is taken into
account, the formation of CO can be omitted from the reaction
mechanism under the reaction conditions used in this work.
Creaser et al. [17] also reported that no CO is formed in the absence
of gas-phase O2, i.e. if only lattice oxygen is available.
The peak maxima of the products are obtained in the following
time-increasing order: non-converted propane < propene < CO2.
For the other temperatures that are studied, i.e. 723, 748, 798 and Fig. 3. Dimensionless normalized flow rate of C3H8 (—) and Ar (—) corresponding to
823 K, no changes are observed and also the order of the mean a single-pulse experiment with a mixture of 48% C3H8 and 52% Ar over 100 mg
residence times of the products remains the same. Based on this, it completely oxidized EL10V1 at 773 K.
36 V. Balcaen et al. / Applied Catalysis A: General 371 (2009) 31–42

hand, practically no CO or CO2 is detected on the oxygen pulse and

therefore, the surface is not covered with a carbon-like deposition.
The reaction products acrolein (C3H4O) and acetaldehyde (C2H4O)
are not obtained on the oxygen but only on the propene pulse,
indicating that the intermediates of these reaction products
directly desorb from the catalyst surface. As formaldehyde is the
only product detected on the oxygen pulse, the influence of the
pump-probe interval should be considered upon the absolute
amount of formaldehyde. From the responses presented in Fig. 5,
the amount of formaldehyde formed from interaction of oxygen
with the adsorbed oxygenate can be calculated. At 723 K,
1.6  1013 molecules of formaldehyde are formed on the oxygen
pulse, which is 5 orders of magnitude lower than the number of
active sites in the reactor. This amount remains constant
Fig. 4. Outlet flow rates of propene (C3H6), acrolein (C3H4O), formaldehyde (CH2O), irrespective of the pump-probe interval, indicating that at least
CO and CO2 corresponding to a single-pulse experiment with pure C3H6 over 1.6  1013 surface species derived from propene with a life time
100 mg completely oxidized EL10V1 at 723 K.
>2 s are present as strongly adsorbed oxygenates during alter-
nating pulse experiments.

3.4. Intermediates in the reaction sequences

From the single-pulse experiments with argon at 723 K over a

symmetrical thin-zone TAP reactor, the average residence time of
argon is determined as M1,Ar/M0,Ar and equals 79 ms. Based on this
residence time, tA,diff and tB,diff can be calculated correcting for the
differences in molecular mass and temperature. Next, the
residence times of propane, propene and CO2 are determined
from the single-pulse experiments with propane and propene.
Finally, the shifted residence times Di can be calculated, which are
listed in Table 1 together with the residence times at two
temperatures.
The difference between the shifted residence times of propane
Fig. 5. Outlet flow rate of formaldehyde (CH2O) corresponding to four propene/
and propene is around 6–8 ms. Therefore, the reaction of propane
oxygen alternating pulse experiments over 100 mg completely oxidized EL10V1 at to propene can be considered instantaneous on the time scale of a
723 K. The time interval between the propene (pump molecule) and the oxygen TAP single-pulse experiment, i.e. without considering an inter-
(probe molecule) pulse is respectively 0.25, 0.5, 1 and 2 s. mediate in the modeling. The difference between the shifted
residence times of propane and CO2 is 133 and 36 ms at
very small amounts of acetaldehyde (C2H4O), but this response respectively 723 and 823 K. It is clear that the difference is
disappears in the background due to the extremely high noise significant at 723 K and becomes less important at 823 K. It is thus
level. possible that an intermediate should be considered explicitly in the
In the alternating pulse experiments in which propene is pulsed formation of CO2 from propane, although no unambiguous
first, followed by an oxygen pulse at different time intervals, no evidence is found based on the criteria for instantaneousness.
propene is observed on the oxygen pulse. The presence of weakly The difference between the shifted residence times of propene and
adsorbed propene species can thus be ruled out. However, from CO2 equals 79 ms at 723 K. This significant difference indicates that
Fig. 5, it is clear that a significant amount of formaldehyde is CO2 is not instantaneously generated from propene in this reaction.
formed on the oxygen pulse, indicating the strong adsorption of an The adequate modeling of the total oxidation of propene to CO2
oxygenate that leads to formaldehyde. This observation is in thus requires to consider at least one intermediate.
agreement with the results of Grzybowska-Swierkosz [19], who
suggested that oxygen is needed to facilitate the desorption of the 3.5. Reaction path analysis and rate equations
strongly bound oxygenated products. The desorption then occurs
simultaneously with the reoxidation of the surface, i.e. a so-called In Table 2 a reaction scheme for the reduction of the catalyst
conjugate oxidation mechanism. Since the reoxidation of the with propane is presented, which is compatible with the
reduced catalyst, involving oxidation of V4+ and V3+ to V5+ leads to conclusions of Sections 3.2–3.4. The elementary reactions in this
an increase of the catalyst acidity and a decrease of the basicity, the scheme are denoted by a number from (R1) to (R13). These
desorption of acidic products becomes more easy. On the other elementary reactions are combined, in a way defined by the

Table 1
Residence times M1,i/M0,i and shifted residence times Di of C3H8, C3H6 and CO2 during propane oxidative dehydrogenation experiments at 723 and 823 K and of C3H6 and CO2
during propene oxidation experiments at 723 K.

Residence time M1,i/M0,i (ms) Shifted residence time Di (ms)

C3H8 experiments C3H6 experiments C3H8 experiments C3H6 experiments

T (K) 723 823 723 723 823 723

C3H8 98 84 – 91 77 –
C3H6 158 126 69 97 69 62
CO2 286 171 203 224 113 141
 V. Balcaen et al. / Applied Catalysis A: General 371 (2009) 31–42 37

Table 2
Elementary reactions ((R1)–(R13)) considered during the reduction of EL10V1 with propane with corresponding global reactions (a, b and g) and reaction paths (aa, bb and gg).
Kinetically significant steps are underlined and shown in bold. A step is called kinetically significant, if the value of the corresponding rate coefficient affects the product yields.

sa sb sg
1 0 0 (R1)

0 1 0 (R2)

1 0 0 (R3)
C3 H7 O ! C3 H6 þ HO

0 0 1 (R4)

0 1 0 (R5)
C2 H4 O CH3 þ O ! C2 H4 O CH2 þ HO

0 1 0 (R6)
C2 H4 O CH2 þ O ! C2 H4 O þ CH2 O

0 1 1 (R7)
C2 H4 O þ O ! C2 H4 ðO Þ2

0 1 1 (R8)
C2 H4 ðO Þ2 þ 2O ! C2 H3 ðO Þ3 þ HO

0 1 1 (R9)
C2 H3 ðO Þ3 þ 2O ! C2 H2 ðO Þ4 þ HO

0 1 1 (R10)
C2 H2 ðO Þ4 ! 2CHðO Þ2

0 1 1 (R11)
CH2 O þ O ! CH2 ðO Þ2

0 1 1 (R12)
CH2 ðO Þ2 þ O ! CHðO Þ2 þ HO

saa sbb sgg

1 0 0 (a)
C3 H8 þ 2O ! C3 H6 þ 2HO

0 1 0 (b)
C3 H8 þ 11O ! 3CHðO Þ2 þ 5HO

0 0 1 (g)
C3 H6 þ 9O ! 3CHðO Þ2 þ 3HO

0 3 3 (R13)

(aa)
C3 H8 þ 2O ! C3 H6 þ 2HO

(bb)
C3 H8 þ 14O ! 3CO2 þ 8HO þ 6

(gg)
C3 H6 þ 12O ! 3CO2 þ 6HO þ 6
38 V. Balcaen et al. / Applied Catalysis A: General 371 (2009) 31–42

columns with stoichiometric numbers si, into global reactions kinetically relevant while the desorption of propene by hydride
presented by the Greek symbols a, b and g. The latter can be elimination, reaction (R3), is considered to proceed potentially fast
further combined with elementary reactions, in a way defined by compared to the methyl C–H bond dissociation. Therefore, no
the columns with stoichiometric numbers sii, yielding the reaction intermediate species in propene formation should be taken into
paths designated by the symbols aa, bb and gg. account, in line with the conclusions based on the instantaneous-
Three reactions paths are taken into account: the oxidative ness analysis (see Section 3.4). This is also reported by Busca et al.
dehydrogenation of propane to propene (aa), the total oxidation of [20] who performed IR studies on the activation of C–H
propane to CO2 (bb), and the consecutive total oxidation of hydrocarbon bonds on oxidation catalysts. During reduction of
propene (gg). Since the amount of propene formed during reaction the catalyst, it is assumed that one O atom per active site interacts
with propane is rather small and since the products of the with the hydrocarbon without specifying the nature of the
oxidation of propene can only be detected over an oxidized catalyst interacting oxygen (see Section 2.1). Since it is generally accepted
if propene itself is fed (see Section 3.3), the only two products that that two O atoms from two different VOx sites are necessary for the
should be considered during the reaction with propane, are initial C–H bond breaking during propane oxidative dehydrogena-
propene and CO2. Therefore, in the reaction mechanism described tion, the interaction of propane with the catalyst is assumed to
in Table 2, the reaction paths leading to gas-phase acrolein, occur with two, non-occupied, oxidized active sites.
acetaldehyde, formaldehyde and CO are omitted. Since the
desorption of H2O according to: 2HO* ! H2O + O* + *, is very slow 3.5.2. Total oxidation of propane to CO2
compared to the time scale of a TAP single-pulse experiment, this The rate of the global reaction (b) is given by the rate of the
reaction is also omitted from Table 2. This is justified as during a methylene C–H bond dissociation, elementary reaction (R2). The
single-pulse experiment the catalyst state and thus the HO- subsequent decomposition and oxidation reactions (R5)–(R12) are
coverage is not significantly altered. As such, strictly speaking, the assumed to occur potentially fast compared to reaction (R2).
catalyst is not reduced during a single-pulse experiment. Together with the elementary reaction (R13), this global reaction
Regression of the experimental data was performed by constitutes the reaction path (bb). In this sequence, the oxidation
integrating model equations (4)–(7), with different expressions of the formate species (CH(O*)2), reaction (R13), is considered to be
for the net production rates, Rg,A and Rs,A*, corresponding to several, potentially fast compared to global reaction (b). Busca et al. [20]
a priori possible, kinetically significant steps. A step is called also reported that CO2 arises from the combustion of the
kinetically significant, if the value of the corresponding rate carboxylate species (acetates and formates) formed by the reactive
coefficient affects the product yields. The most adequate descrip- adsorption of propane. Since no CO could be detected, the reaction
tion of the data was obtained by considering the four steps shown that describes the formation of CO is omitted. This is in line with
in bold in Table 2 and schematically presented in Fig. 6. In the next Novakova et al. [21] who reported that the side-product in the
paragraphs, the final expressions for the net production rates will formation of acetic acid is CO2.
be derived based on the analysis of the three reaction paths,
considering these four steps as kinetically significant. 3.5.3. Total oxidation of propene to CO2
The global reaction (g) describes the consecutive oxidation of
3.5.1. Oxidative dehydrogenation of propane to propene propene through the C5 5C bond breaking, elementary reaction (R4),
Reaction path (aa) solely consists of the global reaction (a) in which is assumed to be a kinetically relevant step. The reverse of
which the methyl C–H bond dissociation (R1) is considered to be reaction (R3) is much less probable compared to reaction (R4), since

Fig. 6. Proposed kinetically relevant steps in the three reaction paths aa, bb and gg of the oxidation of propane with indication of the four corresponding kinetically
significant parameters: k1, k2, k4 and k13. In the reaction path (gg) of the consecutive oxidation of propene to CO2, the oxidation of the formate species (CH(O*)2) with
parameter k13 is kinetically relevant while this step is not significant in the pathway (bb) of the direct total oxidation of propane to CO2.
 V. Balcaen et al. / Applied Catalysis A: General 371 (2009) 31–42 39

in the former propene should interact with an active site occupied These specific net production rates are substituted in the mass
with hydrogen, while in the latter the reduction of the catalyst balances for propane, propene and CO2, Eqs. (4)–(6), and the mass
occurs through the interaction of two, non-occupied, oxidized active balances for adsorbed formate species (CH(O*)2) and free oxidized
sites as mentioned in literature. Moreover, according to Grabowski sites (O*), Eq. (7). The boundary conditions are described by
[22], the reaction products quickly desorb and do not block the Eqs. (8)–(14), while the initial condition for propane, propene, CO2,
surface, implying that no propene will be present on the surface, but surface formate species and free oxidized sites is given by Eq. (15).
will have to readsorb to give CO2. The other elementary oxidation For the surface formate species, this initial condition equals:
reactions (R7)–(R12) are considered to proceed potentially fast
0
compared to the C5 5C double bond breaking, so that the rate of the CCHðOÞ2
¼ 0; t ¼ 0; 0  z  Li þ Lc þ Lo (23)
global reaction (g) is determined by the rate of this double bond
The initial condition for the free oxidized sites follows from the
breaking. From the experimental part (see Section 3.3), it is clear that
initial reduction degree, R0 and the total concentration of active
the reduction of the catalyst with propene occurs through the
sites, Ct:
formation of acrolein, formaldehyde, CO, CO2 and very small
amounts of acetaldehyde. Since the criteria for instantaneousness 0
CO ¼ ð1  R0 ÞC t ; t ¼ 0; 0  z  Li þ Lc þ Lo (24)
expressed the need of at least one intermediate in the modeling of
the consecutive total oxidation of propene, the question can be
raised which intermediate(s) should be considered. As no acrolein is 3.6. Parameter estimation for different reduction degrees
detected on the oxygen pulse during the propene/oxygen alternat-
ing pulse experiments, the pathway through adsorbed acrolein is The above model is applied to regress the experimental data
omitted. Moreover, since the experimental data are adequately and to determine the kinetics of the interaction of propane with the
described if only one intermediate is taken into account and since catalyst. First, an isothermal regression yields the four rate
both formaldehyde and acetaldehyde react to surface formate coefficients of the kinetically relevant steps in the reaction
species (CH(O*)2), the latter is considered as the intermediate in the mechanism (see Fig. 6) at different temperatures and for different
reaction pathway of propene to CO2. Thus, together with the global reduction degrees. The influence of the reduction degree upon
reaction (g), the elementary reaction (R13) completes the formation these isothermal parameter estimates can be evaluated. Secondly,
of CO2 from propene, i.e. reaction path (gg). Global reaction (g) is the influence of temperature can be examined and initial estimates
now considered to be potentially fast compared to the desorption of for the pre-exponential factor and activation energy of each
CO2. As such, the oxidation of adsorbed formate species, reaction coefficient can be obtained. Finally, these initial estimates will be
(R13) must also be included in the reaction scheme as kinetically applied in the non-isothermal regression which provides the final
relevant step. parameter estimates for each rate coefficient. Both the isothermal
If the oxidation of the adsorbed formate species, reaction (R13), as well as the non-isothermal regression are performed by the
which is a common step in the two routes to CO2, i.e. in the total minimization of the objective function F (Eq. (17)).
oxidation of propane to CO2 (path bb) and in the total oxidation of Fig. 7 shows the rate coefficients, k1 and k2, for different
propene to CO2 (path gg), is assumed to be kinetically significant in temperatures as a function of the reduction degree. The rate
both routes, the calculated CO2 response appears too late coefficient of the total oxidation of propane to CO2, k2, shows a
compared to the experimental response. Clearly, considering the pronounced decrease from the fully oxidized state to the lowest
formate species as significant intermediates in both routes to CO2 finite reduction degree after which it practically remains
causes a delay on this response, leading to the conclusion that only unchanged. Apparently, except for the oxidized state, the reduction
in one route the formate species should be considered as degree of the catalyst has little influence on the rate of this reaction
significant intermediates. Because the criteria for instantaneous- path. For the rate coefficient of the oxidative dehydrogenation of
ness (see Section 3.4) clearly indicated that CO2 is not instanta- propane, k1, a same small influence of the reduction degree is
neously generated from propene while no unambiguous observed. The rate coefficients k4 and k13, corresponding to the
conclusion could be drawn concerning the significance of an total oxidation of propene, do not vary to a great extent as soon as
intermediate in the formation of CO2 from propane, only in the the catalyst is partially reduced (results not shown). The significant
total oxidation of propene to CO2 an intermediate is considered. higher values of the rate coefficients, k1 and k2, over the fully
The above reaction path analysis allows to formulate the oxidized catalyst and the small dependence, if any, of all the rate
following specific net production rates of the gas-phase, Rg,i, and coefficients on the reduction degree as soon as it is different from
the surface components, Rs,i*: zero, can be explained as follows. In the fully oxidized state, the
catalyst mainly consists of V5+ and according to Wachs [23] the
k1 k2 following order is valid: kn(V5+) > kn(V4+)  kn(V3+). During the
Rg;C3 H8 ¼  CC H C2  CC H C2 (18)
C t 3 8 O C t 3 8 O reduction of the catalyst with propane, electrons are transferred
from lattice oxygen atoms to the vanadium metal centers, leading
k1 k4 to a V5+–V4+ reduction or, alternatively a V5+–V3+ reduction. As the
Rg;C3 H6 ¼ CC H C2  CC H C2 (19)
C t 3 8 O C t 3 6 O former reduction process requires a lower activation energy,
confirmed by molecular simulations performed by Gilardoni et al.
k2 k13 [24], it is assumed that the reduced states of the catalyst consist of
Rg;CO2 ¼ 3 CC H C2 þ 3 C CHðOÞ2 C O (20) a mixture of V4+ and V3+ with the former in large excess. The small
C t 3 8 O Ct
dependence of the rate coefficients on the reduction degree as soon
as it is finite can then be understood by the dominant presence of
k4 k13
Rs;CHðOÞ2 ¼ 3 CC H C2  3 C CHðOÞ2 C O (21) V4+ on the partly reduced catalyst surface.
C t 3 6 O Ct
Since the reduction degree of the catalyst only slightly affects
the four rate coefficients, it is attempted to simultaneously regress
k1 k2 k4 all experimental data obtained on a partially reduced catalyst,
Rs;O ¼ 2 C C H C 2  14 C C3 H8 CO
2 2
 9 C C3 H6 CO
C t 3 8 O Ct Ct accounting for the different reduction degrees through the initial
k13 0
concentration of free oxidized sites, CO (Eq. (24)). Fig. 8 illustrates
3 C CHðOÞ2 C O (22)
Ct the correspondence between the experimental and the calculated
40 V. Balcaen et al. / Applied Catalysis A: General 371 (2009) 31–42

Fig. 7. Kinetically significant rate coefficients k1 (empty symbols) and k2 (full

symbols) for the oxidative dehydrogenation respectively the total oxidation of
propane at 723 (&), 773 (~) and 823 K (*) for different reduction degrees.

responses over the catalyst with a reduction degree of 0.12 at 773 K

after isothermal regression analysis, considering all finite reduc-
tion degrees separately or simultaneously. While the correspon-
dence between experimental and calculated propane responses is
very good and shows practically no influence of the used
parameter value, the calculated CO2 and propene responses show
a clear difference dependent on the parameter value. However, the
correspondence between experimental and calculated responses
can still be considered adequate in view of the noise-to-signal
ratio. This indicates that the experimental data can be described
with a single rate coefficient for all reduction degrees, except for
the oxidized state, which should be considered separately.
Apart from the reduction degree, also the temperature has a
rather small influence on the rate coefficients k1 and k2 once the
catalyst is partially reduced, which is illustrated in Fig. 7. On the
contrary, over a fully oxidized catalyst, these two coefficients are
clearly dependent on temperature since k1 is doubled and k2
quadrupled as temperature is increased from 723 to 823 K. Rate
coefficients k4 and k13 are practically independent of temperature
over a fully oxidized as well as over a partially reduced catalyst.
This different temperature dependency of the rate coefficients will
be reflected in the values of the activation energies during the non-
isothermal parameter estimation.
Table 3 shows the parameters obtained after non-isothermal
regression. Since the oxidized state should be considered
separately, non-isothermal regression is performed either on the
oxidized state or on the reduced states, which can all be considered Fig. 8. C3H8 (a), CO2 (b) and C3H6 (c) reactor exit flow rates at 773 K after reduction to
a degree of 0.12. (~) Experimental C3H8 response; (*) experimental CO2 response;
simultaneously. This leads to two estimates for each parameter:
(*) experimental C3H6 response; . . . responses calculated with parameter estimates
one for the fully oxidized state and one for the partially reduced obtained by the isothermal regression and by considering the reduction degrees
states. In Fig. 8 the correspondence between the experimental and separately; — responses calculated with parameters obtained by the isothermal
the calculated responses, obtained from the non-isothermal regression and by considering the finite reduction degrees simultaneously; —
regression (Table 3) can be compared with the good correspon- responses calculated with parameters shown in Table 3, obtained by the non-
isothermal regression and by considering all finite reduction degrees
dence obtained from the isothermal regression, considering all simultaneously. All calculated responses are obtained by integration of Eqs. (4)–
reduction degrees simultaneously. As the responses from both (7) with the corresponding net production rates Eqs. (18)–(22).
regression analyses more or less coincide, it can be concluded that
the proposed model can quantitatively describe the interaction of
propane with the catalyst at different reduction degrees, and at found by Dai et al. [25], Wolf et al. [26], Grabowski [22], Genser and
different temperatures. Pietrzyk [27] and Grabowski et al. [28]. This result is in agreement
After estimating the four kinetically significant rate coefficients with kinetic studies for n-butane oxidation to maleic anhydride
of the three reaction paths in the oxidation of propane, as over VPO catalysts performed by Huang et al. [29] who showed
presented in Fig. 6, the relative importance of the steps in the that the activation energy of the complete oxidation reaction is
reaction scheme can be addressed at different reduction degrees of higher than that of the partial oxidation in a selective oxidation
the catalyst. From Table 3 it is clear that the kinetically significant reaction of the hydrocarbon. Although the activation energies of
step in the total oxidation of propane (step R2 in reaction path bb) both reaction paths decrease as soon as the catalyst is reduced, the
requires a higher activation energy than the kinetically significant direct oxidation of propane still requires higher activation energy.
step in the oxidative dehydrogenation (step R1 in reaction path Over the fully oxidized catalyst, the propane adsorption in the
aa), leading to a rate coefficient k2 that is one order of magnitude oxidative dehydrogenation (k1) requires higher activation energy
lower than k1 (see Fig. 7). This indicates that the route to propene is compared to the propene readsorption in the total oxidation of
more important than the direct total oxidation to CO2, as was also propene (k4). This originates from the lower reactivity of propane
 V. Balcaen et al. / Applied Catalysis A: General 371 (2009) 31–42 41

Table 3
Estimates of the pre-exponential factors and the activation energies with their 95% confidence intervals and the F value obtained by the non-isothermal regression of propane
single-pulse experiments between 723 and 823 K over oxidized EL10V1 and after reduction at the same temperatures to different reduction degrees.

R0 = 0 R0 > 0
1
A Ea (kJ mol ) A Ea (kJ mol1)

k1 (mg3 molV mol2 s1) 18.60  0.12 36.35  0.55 (3.32  0.02)10 1
15.21  0.60
k2 (mg3 molV mol2 s1) (1.00  0.02)103 74.29  1.07 1.70  0.04 40.11  2.11
k4 (101 mg3 molV mol2 s1) 7.56  0.16 4.57  1.77 4.23  0.07 –a
k13 (102 molV mol1 s1) 1.56  0.05 –a 6.27  0.14 –a
5
F (10 ) 4.52 21.15
a
Not significantly different from 0 at 95% probability level.

compared to propene, which ensures that propene is activated at a 4. Conclusions

lower reaction temperature. The difference of reactivity between
propane and propene is obvious as it is linked to the strengths of The oxidative dehydrogenation of propane in the absence of
the involved C–H and C5 5C bonds and to the stability of the gas-phase oxygen is shown to occur with lattice oxygen. Only CO2
corresponding transition states formed in the kinetically relevant and propene are detected as reaction products. No CO is formed
steps [30]. Apart from the readsorption of propene, also the further while H2O is strongly adsorbed on the catalyst surface.
oxidation (k13) of adsorbed formate, (CH(O*)2), is a kinetically The reduction of the catalyst with propane occurs through a
relevant step in the global reaction path (gg) of propene to CO2. As, parallel-consecutive mechanism. In the reaction scheme, four
however the latter is independent of temperature over a fully elementary reactions are considered as kinetically relevant: (1)
oxidized catalyst, indicated by a non-significantly estimated the methyl C–H bond dissociation in propane in the formation of
activation energy (see Table 3), the activation energy of the total propene, (2) the methylene C–H bond breaking in propane to surface
oxidation of propene is solely determined by the adsorption of isopropoxide (C2H4O*CH3) and finally to CO2, (3) the C5 5C double
propene. The higher activation energy for propane oxidative bond breaking in propene in the formation of surface formate species
dehydrogenation compared to the total oxidation of propene was (CH(O*)2) and (4) the oxidation of the surface formate species to CO2.
also found by Chen et al. [18], Bottino et al. [31] and Dai et al. [25]. Based on criteria for instantaneousness, it can be concluded that in
Moreover, if the catalyst is reduced, the readsorption of propene is the reaction path of propane to propene no intermediate should be
not activated anymore. This is due to an increasing bond strength considered during the kinetic modeling. On the other hand, adequate
between the olefin and the surface as described by Sloczynski [32]. modeling of the total oxidation of propene to CO2 requires
He assumes that a p-complex of propene associated with the consideration of an intermediate, in casu a surface formate species.
vanadium cation is formed, as a result of an electron transfer from The proposed kinetic model with four kinetically relevant rate
the p bonding orbital of propene to the empty d orbital of the metal coefficients can quantitatively describe the interaction of propane
ion (i.e. p donation), and from the metal d orbital to the empty p* with the catalyst at different reduction degrees, and at different
anti-bonding orbital of propene (i.e. p* back-donation). Both temperatures. The rate coefficient for the propane activation in the
electron transfer mechanisms activate propene, weakening the oxidative dehydrogenation (k1) as well as in the total oxidation (k2)
C55C double bond and increasing the interaction with the catalyst is higher over the fully oxidized catalyst compared to a partially
surface. As the valence of the vanadium cation decreases upon reduced catalyst, because the oxidized state mainly consists of V5+,
reduction of the catalyst, the strength of propene bonding to the considering the fully oxidized surface VOx site as the catalytic
surface increases, due to the enhanced p* back-donation. active site. As soon as the catalyst is reduced, mainly V4+ is present,
In order to compare the two reaction paths leading to CO2 (path leading to a small, statistically not significant dependency of the
bb from propane and gg from propene), the rate coefficient k2 four kinetically relevant rate coefficients on the reduction degree.
should be compared with a combination of the rate coefficients k4 Consequently, it is sufficient to account for the initial reduction
0
and k13. However, because several steps starting from the adsorbed degree through the initial concentration of free oxidized sites, CO ,
acetaldehyde (C2H4O*) and formaldehyde (CH2O*) species are to obtain a good correspondence between the experimental and
identical in these two reaction paths, leading to a common branch the calculated data on the partially reduced catalyst with a single
in the reaction scheme as presented in Fig. 6, the kinetics of the two set of significant parameter estimates.
paths can be compared by weighing k2 against k4. As already For any reduction degree, i.e. either a fully oxidized or a
stated, propene is more reactive than propane, leading to a rate partially reduced catalyst, the selective route to propene is favored
coefficient k4 that is substantially higher than k2 over a fully above the direct total oxidation of propane to CO2. However, the
oxidized as well as over a partially reduced catalyst. Hence, the formed propene reacts relatively easy over the catalyst, due to the
contribution of the consecutive reaction of propene to the higher reactivity of propene compared to propane. Hence, the
formation of CO2 is significant, as also reported by Grabowski contribution in the formation of CO2 originating from propene is
[22]. The observed significant difference in the shifted residence higher than the contribution originating from propane.
time between propane and CO2 at 723 K can now be explained by
the importance of the formation of CO2 via propene. As this Acknowledgement
difference was only significant at 723 K and not at higher
temperatures, the criteria for instantaneousness could not This work was performed in the framework of a Concerted
unambiguously point to the need of an intermediate in the Research Action (GOA) financed by Ghent University.
reaction path of propane to CO2. Moreover, in the kinetic model
used in this work, such an intermediate appeared unnecessary. References
Therefore, since practically all CO2 comes from the readsorption of
propene, propene and adsorbed formate species should be [1] B. Grzybowska, F. Trifiro, J.C. Védrine, Appl. Catal. A 157 (1997) 1–425.
considered as the intermediates that influence the outlet flow [2] P. Mars, D.W. van Krevelen, Chem. Eng. Sci. 3 (1954) 41–58.
[3] I. Sack, M. Olea, H. Poelman, K. Eufinger, R. De Gryse, G.B. Marin, Catal. Today 91–
rate and hence the shifted residence time of CO2 during the 92 (2004) 149–153.
oxidation with propane. [4] I. Sack, V. Balcaen, M. Olea, H. Poelman, G.B. Marin, Catal. Today 112 (2006) 68–72.
42 V. Balcaen et al. / Applied Catalysis A: General 371 (2009) 31–42

[5] E. Garcin, F. Luck, R. Surantyn, Catal. Today 20 (1994) 7–10. [16] L. Petzold, SIAM J. Sci. Stat. Comput. 4 (1983) 136–148.
[6] J. Haber, A. Baiker, M. Nobbenhuis, J. Kiwi, R. Thampi, M. Gratzel, L. Lucarrelli, P. [17] D. Creaser, B. Andersson, R.R. Hudgins, P.L. Silveston, Appl. Catal. A 187 (1999)
Albers, K. Seibold, M. Baerns, F. Majunke, M. del Arco, C. Martin, V. Rives, E. Garcin, 147–160.
R. Surantyn, J.-C. Volta, J.G. van Ommen, M. de Boer, A.J. van Dillen, J.W. Geus, J. [18] K. Chen, E. Iglesia, A.T. Bell, J. Catal. 192 (2000) 197–203.
Haber, J. Sloczynski, A. Andersson, L.T. Andersson, M. Sanati, B. Svensson, B. [19] B. Grzybowska-Swierkosz, Appl. Catal. A 157 (1997) 409–420.
Rebenstrof, R. Wallenberg, Catal. Today 20 (1994) 11–16. [20] G. Busca, E. Finocchio, V. Lorenzelli, G. Ramis, M. Baldi, Catal. Today 49 (1999)
[7] J.Ph. Nogier, E. Garcin, R. Surantyn, A. Chambosse, A. Foucault, G. Coudurier, J.-C. 453–465.
Volta, A.M. De Kersabiec, F. Cavani, J. van Ommen, B. Grzyvbowska, M. Czervenka, [21] E.K. Novakova, J.C. Védrine, E.G. Derouane, J. Catal. 211 (2002) 226–234.
K. Malec, A. Andersson, Catal. Today 20 (1994) 23–34. [22] R. Grabowski, Appl. Catal. A 270 (2004) 37–47.
[8] I. Sack, A kinetic study of the reduction and the reoxidation of vanadia-based [23] I.E. Wachs, Catal. Today 100 (2005) 79–94.
catalysts by Temporal Analysis of Products, PhD thesis, Ghent University, 2006. [24] F. Gilardoni, A.T. Bell, A. Chakraborty, P. Boulet, J. Phys. Chem. B 104 (51) (2000)
[9] G. Busca, A. Zecchina, A. Andersson, B. Rebensdorf, M. Baerns, S. Trautman, J. Billy, 12250–12255.
G. Coudurier, J.-C. Volta, G.C. Bond, C.R. Dias, V. Lorenzelli, G. Ramis, S. Coluccia, F. [25] H. Dai, A.T. Bell, E. Iglesia, J. Catal. 221 (2004) 491–499.
Geobaldo, E. Giamello, L. Marchese, M. del Arco, C. Martin, V. Rives, J. Krysciak, F. [26] D. Wolf, N. Dropka, Q. Smejkal, O. Buyevskaya, Chem. Eng. Sci. 56 (2001) 713–719.
Trifiro, Catal. Today 20 (1994) 61–76. [27] F. Genser, S. Pietrzyk, Chem. Eng. Sci. 54 (1999) 4315–4325.
[10] J.T. Gleaves, J.R. Ebner, T.C. Kuechler, Catal. Rev. Sci. Eng. 30 (1988) 49–116. [28] R. Grabowski, S. Pietrzyk, J. Sloczynski, F. Genser, K. Wcislo, B. Grzybowska-
[11] J.T. Gleaves, G.S. Yablonsky, P. Phanawadee, Y. Schuurman, Appl. Catal. A 160 Swierkosz, Appl. Catal. A 232 (2002) 277–288.
(1997) 55–88. [29] X.-F. Huang, C.-Y. Li, B.-H. Chen, P.L. Silveston, AIChE J. 48 (4) (2002) 846–
[12] G.S. Yablonsky, M. Olea, G.B. Marin, J. Catal. 216 (2003) 120–134. 855.
[13] D. Constales, G.S. Yablonsky, G.B. Marin, J.T. Gleaves, Chem. Eng. Sci. 59 (2004) [30] M.M. Bettahar, G. Costentin, L. Savary, J.C. Lavalley, Appl. Catal. A 145 (1996) 1–48.
3725–3736. [31] A. Bottino, G. Capannelli, A. Comite, S. Storace, R. Di Felice, Chem. Eng. J. 94 (2003)
[14] H.H. Rosenbrock, Comput. J. 3 (1961) 175–184. 11–18.
[15] D.W. Marquardt, J. Soc. Ind. Appl. Math. 11 (1963) 431–443. [32] J. Sloczynski, Appl. Catal. A 146 (1996) 401–423.