Transient studies on the mechanism of N2O activation and reaction with CO and C3H8 over Fe-silicalite

Article abstract of DOI:10.1016/j.jcat.2005.05.014

The mechanism of the reaction of N₂O and ¹³CO over Fe-silicalite was investigated with the use of the temporal analysis of products (TAP) reactor and compared with that of the reaction of N₂O and C₃H₈ previously reported (Appl. Catal. A 267 (2004) 181). Upon direct N₂O decomposition at 523-573 K, Fe-silicalite stored ca. 10¹⁸ atoms of oxygen per gram, with a ratio of 1 O atom per each 30-60 Fe atoms in the sample. Only a small fraction of the deposited oxygen was reactive for CO oxidation. Pump-probe experiments at different time delays (0-2 s) between the pulses of nitrous oxide and the reducing agent indicated the markedly different mechanisms of the N₂O-¹³CO and N ₂O-C₃H₈ reactions in the temperature range of 623-673 K. Fe-silicalite is active for propane oxidation in the presence of short-lived oxygen species, that are produced when N₂O and C ₃H₈ are pulsed simultaneously. Time delays between the N₂O and C₃H₈ pulses greater than 0.1 s are sufficient to transform these active oxygen species for hydrocarbon conversion into inactive ones. In contrast, the oxidation of CO by N₂O does not depend on the lifetime of the oxygen species in the range of time delays investigated. The mechanisms of the N₂O-mediated ¹³CO and C₃H₈ oxidations differ as a consequence of the different interactions of the two reducing agents with iron species in the zeolite. Pulse experiments support the occurrence of the scavenging mechanism with both propane and carbon monoxide. In this mechanism, short-lived oxygen deposited by N ₂O is efficiently eliminated by the reductant. Distinctive to propane, the strikingly high affinity of carbon monoxide for isolated Fe ³⁺ ions in the zeolite gives rise to an additional pathway for N ₂O reduction in the presence of chemisorbed CO species. These particular Fe³⁺-CO species were identified by in situ UV/vis and EPR spectroscopies (J. Catal. 223 (2004) 13).

Full text of DOI:10.1016/j.jcat.2005.05.014

Journal of Catalysis 233 (2005) 442–452
www.elsevier.com/locate/jcat
Transient studies on the mechanism of N O activation and reaction
2
with CO and C H over Fe-silicalite
3
8
a,∗
b
c,1
Javier Pérez-Ramírez , Evgenii V. Kondratenko , M. Naoufal Debbagh
a
Laboratory for Heterogeneous Catalysis, Catalan Institution for Research and Advanced Studies (ICREA)
and Institute of Chemical Research of Catalonia (ICIQ), Av. Països Catalans 16, E-43007, Tarragona, Spain
b
Institute for Applied Chemistry Berlin-Adlershof, Richard-Willstätter-Str. 12, D-12489 Berlin, Germany
Department of Inorganic Chemistry, University of Alicante, P.O. Box 99, E-03080 Alicante, Spain
c
Received 16 January 2005; revised 15 May 2005; accepted 20 May 2005
Available online 16 June 2005
Abstract
The mechanism of the reaction of N O and CO over Fe-silicalite was investigated with the use of the temporal analysis of products
13
2
(
TAP) reactor and compared with that of the reaction of N O and C H previously reported (Appl. Catal. A 267 (2004) 181). Upon direct
2 3 8
18
N O decomposition at 523–573 K, Fe-silicalite stored ca. 10 atoms of oxygen per gram, with a ratio of 1 O atom per each 30–60 Fe
2
atoms in the sample. Only a small fraction of the deposited oxygen was reactive for CO oxidation. Pump-probe experiments at different time
13
delays (0–2 s) between the pulses of nitrous oxide and the reducing agent indicated the markedly different mechanisms of the N O– CO
2
and N O–C H reactions in the temperature range of 623–673 K. Fe-silicalite is active for propane oxidation in the presence of short-lived
2
3 8
oxygen species, that are produced when N O and C H are pulsed simultaneously. Time delays between the N O and C H pulses greater
2
3
8
2
3 8
than 0.1 s are sufficient to transform these active oxygen species for hydrocarbon conversion into inactive ones. In contrast, the oxidation of
CO by N O does not depend on the lifetime of the oxygen species in the range of time delays investigated. The mechanisms of the N O-
2
2
13
mediated CO and C H oxidations differ as a consequence of the different interactions of the two reducing agents with iron species in the
3
8
zeolite. Pulse experiments support the occurrence of the scavenging mechanism with both propane and carbon monoxide. In this mechanism,
short-lived oxygen deposited by N O is efficiently eliminated by the reductant. Distinctive to propane, the strikingly high affinity of carbon
2
3
+
monoxide for isolated Fe ions in the zeolite gives rise to an additional pathway for N O reduction in the presence of chemisorbed CO
species. These particular Fe –CO species were identified by in situ UV/vis and EPR spectroscopies (J. Catal. 223 (2004) 13).
2
3
+

2005 Elsevier Inc. All rights reserved.
Keywords: N O decomposition; N O reduction; Fe-silicalite; Iron zeolites; Oxygen species; CO; C H ; Mechanism; TAP reactor
2
2
3 8
1
. Introduction
ties for direct N2O decomposition in tail gases at temper-
atures lower than 675 K, because of the slow desorption of
oxygen and the inhibiting effect of water vapor [4]. The pres-
ence of reducing agents accelerates the removal of oxygen
species deposited by N2O from the catalyst surface by selec-
tive catalytic reduction (SCR), resulting in a decreased op-
eration temperature. Light hydrocarbons (C1–C3) have been
mostly applied for SCR of N2O over iron zeolites [4–12].
Delahay et al. [13] compared the influence of the reducing
agent (H2, NH3, CO, C3H6, and n-C10H22) on the perfor-
mance of ion-exchanged Fe-beta zeolite for SCR of N2O in
Fe-MFI zeolites are attractive catalysts for N2O abate-
ment in tail gases from chemical production and stationary
combustion processes, where N2O coexists with other com-
ponents such as O2, NOx, H2O, and SO2 [1,2]. In these
sources, the temperature determines the type of catalytic
after-treatment [3,4]. Iron zeolites show insufficient activi-
*
Corresponding author. Fax: +34-977-920-224.
E-mail address: jperez@iciq.es (J. Pérez-Ramírez).
1
the presence of excess O . CO was found to be the most
On leave from Département de Chimie, Université Abdelmalek Es-
saâdi, Tétouan, Morocco.
2
effective selective reductant in the low-temperature range,
0021-9517/$ – see front matter  2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2005.05.014

J. Pérez-Ramírez et al. / Journal of Catalysis 233 (2005) 442–452
443
leading to significant N2O conversion at 470 K. A more re-
cent study by Nobukawa et al. [12] has concluded that CH4
gives a higher N2O conversion than CO and H2 over Fe-MFI
catalysts in the absence of O2.
The considerable number of works reporting catalytic
data for iron zeolites in SCR of N2O contrasts with a
few publications assessing the intrinsic reaction mecha-
nism of the process. Exceptionally, the reaction of N2O and
CH4 over Fe-beta has been investigated in detail by the
group of Kunimori, who used in situ infrared spectroscopy,
temperature-programmed desorption, and pulse techniques
A recent study using in situ UV/vis and EPR spectro-
scopies [23] has demonstrated that the N O reduction by
2
CO over differently prepared FeMFI zeolites is more com-
plex than originally outlined. This involved not only the
classic scavenging mechanism in Eqs. (1) and (3) occur-
3+
2+
ring over oligonuclear iron clusters via a redox Fe /Fe
process, but also the reduction of N O with coordinated CO
2
species over isolated Fe3 ions [Eqs. (4) and (5)]. Steam-
activated Fe-silicalite, containing mostly isolated iron ions
in extraframework positions, showed the highest activity per
mole of iron [23]. Based on this, it was proposed that the
pathways in Eqs. (4) and (5) contribute significantly to the
overall reaction rate. Gaining further insight into the mecha-
+
[
14–16]. These studies concluded that short-lived oxygen
∗
species (denoted hereafter as O ), present on the catalyst sur-
face when N2O and CH4 are fed together, efficiently activate
methane at low temperature (>473 K). This leads to reactive
methoxide species (Fe–OCH3), which are further oxidized
by N2O, with the possible formation of intermediate for-
mate species (Fe–OOCH), finally leading to N2, CO2, and
H2O, and the regeneration of the active iron site (Fe–OH).
In contrast, the reaction of CH4 with the N2O-pretreated
catalyst yielded very low activities even at 623 K. Based
on earlier work by Au and Roberts [17] with the NH3 +
N2O reaction on a Mg(0001) surface, Kunimori et al. [14–
nism of the N O + CO reaction on this particular catalyst is
2
of fundamental and practical interest, especially if a detailed
understanding of the differences between the interactions of
carbon monoxide and light hydrocarbons with nitrous oxide
could be achieved.
In this work, we have applied a transient pulse technique,
the TAP reactor, with isotopic tracers in order to quantita-
tively investigate the process of N O activation over steam-
2
activated Fe-silicalite and the interaction of the generated
13
oxygen species with CO. The mechanism of the N2O +
1
3
1
6] have distinguished two types of oxygen species derived
CO reaction has been comparatively discussed with that
from N2O activation: (highly reactive) nascent oxygen tran-
of the N2O + C3H8 reaction analyzed previously [18]. The
mechanistic aspects elucidated here are relevant for deriving
rational microkinetic models to predict steady-state catalytic
performances for N2O abatement over iron-containing ze-
olites, a process of prominent importance in environmental
catalysis.
∗
sient species (O ) and (nonreactive) thermally accommo-
dated oxygen species (O). In good agreement, the different
reactivity of oxygen species from N2O, depending on their
lifetime, was also concluded from our recent study of the
N2O + C3H8 reaction over steam-activated Fe-MFI cata-
lysts in the temporal analysis of products (TAP) reactor [18].
Accordingly, several pieces of evidence concur on the as-
sumption that short-lived oxygen species formed upon N2O
decomposition [Eq. (1)] are responsible for the hydrocarbon
activation and subsequent conversion [Eq. (2)].
2
. Experimental
2
.1. Catalyst
N2O + ∗ → N2 + O∗
(1)
(2)
(3)
(4)
(5)
Details of the hydrothermal synthesis of isomorphously
∗
∗
HC + O → CO2 + H2O +
substituted Fe-silicalite have been described elsewhere [24].
The as-synthesized sample was calcined in air at 823 K for
10 h and converted into the H-form by three consecutive ex-
changes with a NH4NO3 solution (0.1 M) for 12 h and sub-
sequent air calcination at 823 K for 5 h. Finally the catalyst
was activated in a flow of nitrogen and steam at 1 bar, with
the use of a partial steam pressure of 300 mbar and 30 ml
∗
∗
CO + O → CO2 +
CO + ∗ → CO∗
N2O + CO → CO2 + N2 +
∗
∗
For the N2O + CO reaction over FeZSM-5, it has been
also proposed that CO scavenges adsorbed atomic oxygen
species deposited by N2O, leading to N2 and CO2 [Eqs. (1)
and (3)] [19]. Panov et al. [20,21] determined that α-oxygen,
formed by decomposition of N2O at temperatures less than
−1
STP N2 min at 873 K for 5 h. Characterization studies of
the steam-activated Fe-silicalite (Si/Al ≈ ∞ and 0.68 wt%
Fe) indicated a rather uniform distribution of iron species
dominated by isolated iron ions in extraframework positions.
A certain fraction of oligonuclear iron-oxo species in the ze-
olite channels was also identified, but clustering into large
iron oxide particles was totally prevented [23,24].
2+
5
73 K over certain Fe sites in iron-containing zeolites,
reacts with CO in a stoichiometric ratio of 1:1. However,
Kiwi-Minsker et al. [22] observed that only a fraction of the
oxygen atoms stored on an iron-containing H-ZSM-5 upon
N2O decomposition participates in CO oxidation. This re-
sult indicates a parallel between carbon monoxide and light
hydrocarbons as N2O reductants, strongly connoting the co-
existence of various forms of adsorbed oxygen with distinct
reactivity, rather than a single type of species.
2.2. Transient studies
Transient experiments were performed in the TAP reac-
tor, a pulse technique with a time resolution in the sub-

444
J. Pérez-Ramírez et al. / Journal of Catalysis 233 (2005) 442–452
millisecond range. The TAP-2 reactor system has been de-
scribed in detail elsewhere [25,26]. The catalyst (50 mg,
3. Results and discussion
2
50–355 µm) was packed between two layers of quartz
3.1. Direct N O decomposition versus N O reduction
2
2
spheres of the same size fraction in the quartz reactor (6 mm
id). Before the experiments, Fe-silicalite was pretreated in
In order to understand primary mechanistic differences
between direct N2O decomposition and N2O reduction, sin-
gle pulsing of N2O and simultaneous pulsing of N2O– CO
and N2O–C3H8 were conducted in the TAP reactor at differ-
ent temperatures (Fig. 1). The transient responses in this fig-
ure were normalized for a better comparison of pulse shapes,
which provide valuable qualitative information about the re-
action mechanisms. During direct N2O decomposition at
−1
flowing He (30 ml STP min ) at 773 K and atmospheric
pressure for 2 h. The pretreated sample was then exposed
13
−
5
to vacuum (10 Pa), and pulse experiments were subse-
14
15
quently performed. Pulse sizes in the range of 10 –10
molecules were dosed by means of high-speed pulse valves.
This pulse size range corresponded to the Knudsen diffu-
sion regime, where the interaction of molecules in the gas
phase is minimized. Accordingly, purely heterogeneous re-
action steps were accounted for. Various types of transient
experiments were conducted:
7
73 K (Fig. 1a), the response of N2 directly follows that of
N2O, whereas the response of O2 is broader and consider-
ably shifted to longer times. This indicates that N2 formation
is faster than O2 formation, in conformity with the well-
–
The concentration of oxygen atoms adsorbed on the zeo-
lite by N2O decomposition was determined from multi-
pulse experiments. In these, a mixture of N2O/Ne = 1/1
was repetitively pulsed over the He-pretreated catalyst at
accepted fact that the overall reaction rate of direct N O
decomposition is limited by oxygen desorption from the cat-
2
alyst surface [12,27–29]. The uncoupling of the N and O
2
2
signals in Fig. 1a indicates that in this temperature range,
5
23 and 573 K until no N2O conversion was observed.
2
O should be formed mainly by the slow recombination and
Approximately 1 min after the last N2O pulse, multi-
pulse experiments with a mixture of 13CO/Ne = 1/1
were carried out at the same temperatures in order to
probe the deposited oxygen species via CO oxidation to
CO2.
desorption of surface oxygen atoms according to Eq. (6),
regenerating the active site. A recent work has correlated
the shape of the oxygen response during single pulsing of
N O in the TAP reactor with the steady-state performance
2
of steam-activated FeMFI zeolites in direct N O decompo-
2
–
–
Direct N2O decomposition was studied at 673–773 K by
single pulsing of a gas mixture of N2O/Ne = 1/1 over
the He-pretreated catalyst.
2 2
sition [30]. The sharper the O signal, the higher the N O
conversion due to an improved oxygen desorption, which is
essentially determined by the forms of iron species in the
catalyst.
The reaction of N2O and 13CO (or C3H8) was investi-
gated at 623 and 673 K by means of pump-probe experi-
∗
∗
∗
13
O + O → O2 + 2
(6)
ments. Mixtures of N2O/Xe = 1/1 and CO/Ne = 1/1
(
or C3H8/Ne = 1/1) were sequentially pulsed in the re-
Active site regeneration can be strongly accelerated by
the addition of reducing agents, leading to a reduced oper-
actor, with time delays (ꢀt) between the two mixtures
in the range of 0–2 s. Experiments at ꢀt = 0 s involved
simultaneous pulsing of both mixtures.
∗
ation temperature for N2O abatement by facilitated O re-
moval [Eqs. (2) and (3)] [4]. As a matter of fact, the N2O
conversion in the presence of CO or C3H8 (Figs. 1b–c) was
The reactants Ne (4.5), Xe (4.0), N2O (2.0), CO (4.7),
CO (99 at% C, Aldrich), and C3H8 (3.5) were used with-
9
0 and 83% at 673 K, respectively, as compared with the
N2O conversion of 70% at 773 K in direct decomposition
Fig. 1a). The effect of CO and C3H8 can be mechanisti-
1
3
13
out additional purification. A quadrupole mass spectrome-
ter (Hiden Analytical) was used for quantitative analysis of
reactants and reaction products. The transient responses at
the reactor outlet were monitored at the following atomic
(
cally explained, with attention to the shape of the transient
responses in Fig. 1. For example, the reaction products of
N2O reduction and 13CO oxidation, N2 and CO2, appear
at very similar times (Fig. 1b), contrasting with the delay
of the O response with respect to that of N in direct N O
13
13
mass units (AMUs): 132 (Xe), 45 ( CO2), 44 (N2O, CO2,
C3H8), 42 (C3H6, C3H8), 32 (O2), 30 (N2O), 29 (C3H8,
2
2
2
1
3
∗
CO), 28 (N2, CO, N2O, CO2, C3H8, C3H6), 20 (Ne), and
8 (H2O). The individual AMUs in multi-pulse experiments
decomposition. Accordingly, the removal of O species by
1
CO, yielding CO , is much faster than the recombination
2
∗
were recorded without averaging. In other experiments, 10
pulses were recorded and averaged for each AMU in or-
der to improve the signal-to-noise ratio. This was done once
stable transient responses of the chemical species involved
were obtained. The variations in feed components and re-
action products were determined from the respective AMUs
with the use of standard fragmentation patterns and sensitiv-
ity factors.
of O species in direct N O decomposition. In this study,
isotopically labeled CO and nonlabeled N O were used
2
13
14
2
to uncouple the analysis of N O–CO and CO–N in mass
2
2
2
spectrometry. The use of the most abundant isotopes for dini-
trogen oxide and carbon monoxide makes it impossible to
discriminate between them because of the identical main
1
4
12
14
masses of N O and CO (AMU 44) and of N and
2
2
2
12
CO (AMU 28). In a manner very similar to that of the

J. Pérez-Ramírez et al. / Journal of Catalysis 233 (2005) 442–452
445
13
Fig. 1. Transient responses of reactants and products during (a) single pulsing of N O at 773 K, (b) simultaneous pulsing of N O and CO at 673 K, and
2
2
(
c) simultaneous pulsing of N O and C H at 673 K over Fe-silicalite. Pulse size of N O, CO, and C H ≈ 2 × 1014 molecules.
2
3
8
2
3 8
1
3
N2O– CO system, the responses of the N2 and C3H6 prod-
ucts upon simultaneous pulsing of N2O–C3H8 at 623 K are
also nicely coupled (Fig. 1c). It should be pointed out that
propylene is a primary product of the oxidative dehydro-
genation of propane with N2O, where steam-activated iron
zeolites have demonstrated superior performance [18,31,32].
CO2 was also formed in the reaction as a product of the total
propane oxidation (cf. Section 3.4). This would be the main
C-product of the reaction when O2 is in the feed, as shown
in steady-state studies dealing with the C3H8-SCR of N2O
in the presence of excess oxygen [6,7].
the two reactants for N2O reduction. It is worth noting in
particular the pronounced tailing of the 13CO and CO2 re-
sponses over Fe-silicalite (up to 6 s) as compared with those
of C3H8 and C3H6 (ca. 1 s). This indicates a strong interac-
tion of carbon monoxide with the catalyst, accompanied by a
very slow desorption process. As elaborated in Section 3.4,
this aspect turns out to be of paramount importance to the
elucidation of the principal mechanistic differences between
CO and C3H8 in the catalytic reduction of N2O over Fe-
silicalite.
13
The type of pulse experiments presented in Figs. 1b–c
further confirms that both CO and C3H8 are efficient O
3.2. Active sites for N2O activation
∗
scavengers according to Eqs. (2) and (3) but does not demon-
strate conclusive differences between the modus operandi of
2 2
The O /N ratio at the reactor outlet during the single
N2O pulse experiment in Fig. 1a was ca. 1:3, indicating a

446
J. Pérez-Ramírez et al. / Journal of Catalysis 233 (2005) 442–452
Fig. 3. Amount of atomic oxygen irreversible adsorbed on Fe-silicalite upon
direct N2O decomposition and amount of atomic oxygen removed by prob-
ing the resulting oxidized surface by CO at different temperatures.
Fig. 2. N O conversion over Fe-silicalite in multi-pulse experiments at
2
15
13
5
73 K. Pulse size of N O ≈ 3 × 10 molecules.
2
certain consumption of oxygen by the zeolite. Indeed, no O2
was produced over the iron-containing zeolite at T < 673 K;
N2 was the only decomposition product detected. Taking
into account this experimental observation, we carried out
multi-pulse experiments with N2O in the TAP reactor over
Fe-silicalite (pretreated in He at 723 K) in order to determine
the capacity of the sample to store oxygen species at 523 and
ture from 523 to 603 K. Higher concentrations of α-sites, on
the order of 1019 sites g , have been determined by Dubkov
−1
cat
et al. [35,36] over FeZSM-5 with 0.8 wt% Fe2O3; they used
their well-established static method at low partial N2O pres-
sures (0.2–0.4 Torr) at 523 K and isotopic oxygen exchange
at 323 K. The different nature of the experimental techniques
applied may explain the discrepancies in the determined con-
centration of active sites for N2O decomposition. However,
the dissimilar genesis of the zeolite catalysts in the various
studies with respect to the iron content, synthesis method,
and activation procedure, which strongly influence the forms
of iron in the final material, will also affect the obtained con-
centrations of active sites for the reaction. For example, it has
recently been reported that the ratio between the amount of
oxygen deposited by N2O decomposition at 473 K and the
iron content (Odep/Fe) in Fe-ferrierite zeolites with Fe/Al =
0.05–0.25 increased from 0.5 to 1 as the sample pretreatment
temperature was increased from 723 to 973 K [37].
5
73 K. This should provide a measure of the number of sites
for N2O decomposition according to Eq. (1). Following the
original nomenclature by Panov, these sites are referred to
2
+
in the literature as α-sites and have been defined as Fe
species in the zeolite, which are reversibly oxidized to Fe3
+
species by N2O, generating α-oxygen [20,33]. This form of
active surface oxygen readily exchanges with O2 and oxi-
dizes CO and CH4 already at room temperature, and benzene
to phenol at moderate temperatures [34].
A typical curve resulting from the multi-pulse experi-
ments with N2O at 573 K is shown in Fig. 2. It should be
noted that the peak N2O pressure in these experiments at the
In our case, the amount of atomic oxygen deposited on
the catalyst as determined from the multi-pulse experiments
represents only 1.5% (at 523 K) and 3.7% (at 573 K) of the
15
pulse size applied (3 × 10 molecules) is ca. 0.2 Torr and
3
that the number of N2O molecules in one pulse is 10 times
−
1
smaller than the total number of iron atoms in the catalyst.
The N2O conversion in the first pulse (ca. 30%) decreases
with the pulse number as a consequence of the gradual sat-
uration of the active sites in the catalyst. The decrease is
sharper during the first 200 pulses and gradually approaches
the value of zero conversion after 500 pulses. The amount of
atomic oxygen deposited in the zeolite upon N2O activation
is shown as empty bars in Fig. 3, increasing from 1.8 µmol
total iron content in Fe-silicalite (122 µmol Fe g ). This in-
cat
dicates that a very large fraction of the iron in the zeolite
(>95%) is unable to activate the N2O molecule under TAP
conditions at these temperatures.
3.3. Titration of adsorbed oxygen species with 13CO
Approximately 1 min after saturation of Fe-silicalite by
−
1
−1
O g at 523 K to 4.5 µmol O gcat at 573 K. These values can
atomic oxygen species resulting from N O decomposition,
cat
2
be expressed as a concentration of sites, assuming that a sin-
gle oxygen atom is chemisorbed on each active site, resulting
multi-pulses of 13CO were introduced at the same temper-
ature. The aim of this experiment was to probe the oxygen
1
8
18
−1
in 1.1 × 10 and 2.7 × 10 sites g at 523 and 573 K, re-
species [deposited by N O, Eq. (1)] that are active for CO
cat
2
spectively. Similarly, Kiwi-Minsker et al. [22] determined
oxidation, according to Eq. (3). The results obtained were
13
concentrations of sites (referred to as Cad) in the range of
surprising, since the amount of oxygen reacting with CO
was 4 and 10 times lower than the amount of oxygen de-
17
18
−1
1
0 –10 sites g with a transient step-change method at
cat
a partial N2O pressure of 15 Torr over H-ZSM-5 with iron
contents in the range of 70–1000 ppm and experienced a
slight increase in Cad upon increasing the reaction tempera-
posited by N O and was not temperature dependent (solid
bars in Fig. 3). Since no O2 desorption was observed in the
period of time between the last N2O pulse in Fig. 2 and the
2

J. Pérez-Ramírez et al. / Journal of Catalysis 233 (2005) 442–452
447
first 13CO pulse in the titration experiments, a large fraction
of oxygen species deposited over Fe-silicalite during N2O
decomposition was concluded to be irreversibly adsorbed
in the zeolite and unable to oxidize CO. Our results are in
line with a recent work by Kiwi-Minsker et al. [22], who
investigated the oxidation of CO at 523 K over iron-doped
H-ZSM-5, which was pre-covered by oxygen species (Oad)
resulting from N2O decomposition at the same temperature.
These authors concluded that the amount of Oad consumed
limited redox capacity of ion-exchanged iron mordenite for
16
oxygen isotopic exchange during N2 O decomposition ex-
periments over 18O-treated catalysts, which did not conform
to the authors’ expectations.
The results derived from our multi-pulse experiments in
the TAP reactor and from previous works [14–16,18,22,
28,38–40] produce unquestionable evidence that not all of
the atomic oxygen deposited by N2O on active iron sites
in the zeolite exhibit the described reactivity properties for
α-oxygen [20,21,33–36]. The deposited O atoms include
(at least) more than one type of adsorbed oxygen species
with low activity for CH4 oxidation [14–16], C3H8 oxida-
tion [18], CO oxidation [22], O2 desorption during N2O
decomposition [28], and isotopic oxygen exchange [38,39].
(or CO2 formed) corresponds to ca. 65% of the total amount
of atomic oxygen chemisorbed from N2O, whereas in our
case the percentages of oxygen species reacted with CO
amount to 24% at 523 K and 10% at 573 K of the total oxy-
gen species deposited by N2O. To explain the reactivity of a
fraction of adsorbed oxygen species for CO oxidation, Kiwi-
Minsker et al. [22] proposed the presence of two oxygen
pools: one with the oxygen capable of reacting with CO and
the other with adsorbed oxygen, which was not active. These
results are in contradiction to those reported by the group
of Panov [20,21], establishing that the stoichiometry of the
reaction between α-oxygen and carbon monoxide over Fe-
containing ZSM-5 zeolite is 1:1, leading to CO2 in amounts
equal to those of α-oxygen on the surface. It should be noted
that in contrast to our experiments, Panov et al. determined
the CO:Oα stoichiometry from the reaction between Oα (de-
posited by N2O decomposition at 523 K) and CO at 373 K.
The hypothesis of two oxygen pools in the N2O + CO
reaction [22] fits nicely with the mechanistic description of
the N2O + C3H8 reaction [18] and the N2O + CH4 reac-
tion [14–16]. In the latter, Kunimori et al. discerned two
types of adsorbed oxygen species on Fe-beta: “nascent oxy-
gen transient” able to activate CH4 at 473 K and “thermally
accommodated oxygen” that is hardly able to activate CH4
even at 623 K. These studies are further discussed in Sec-
tion 3.4 in relation to our sequential pulse experiments be-
tween N2O and C3H8. The “oxygen pools” concept was also
proposed by Pirngruber [28] to account for the two different
O2 desorption processes during direct N2O decomposition
over Fe-zeolites prepared by sublimation: one oxygen pool
desorbed relatively fast from a small fraction of sites (re-
sponsible for catalytic activity), and a second oxygen pool
desorbed very slowly from the catalyst.
3.4. Interaction of O-species from N2O with reducing
agents: 13CO versus C3H8
The reaction of adsorbed oxygen species generated upon
N2O decomposition with 13CO and C3H8 was investigated at
623 and 673 K by means of pump-probe experiments in the
TAP reactor. To this end, N2O and 13CO (or N2O and C3H8)
were sequentially pulsed, with a time delay between them
in the range of 0–2 s. The application of short and variable
time delays in sequential pulsing represents a unique feature
of the TAP reactor, particularly for the analysis of the influ-
ence of the lifetime of the oxygen species deposited by N2O
on their reactivity for the oxidation of carbon monoxide or
propane.
Fig. 4 shows the conversion of reactants and yield
of products during these experiments over Fe-silicalite,
which indicate the markedly different performance and re-
action mechanism, depending on the reductant applied. At
13
ꢀt = 0 s, that is, when N2O and CO (or N2O and C3H8)
were simultaneously pulsed, the conversion of 13CO (ca.
80%) was significantly higher than the conversion of C3H8
(ca. 50%). Associated with this, the N2O conversion was
also higher with 13CO than with C3H8 (e.g., 70% vs. 50% at
623 K). From this result, it can be concluded that N2O ox-
idizes carbon monoxide more easily than it does propane,
or in other words, that CO is a better reducing agent for the
conversion of N2O to N2 over our catalyst. Furthermore, an-
other revealing feature originally reported in [18] and shown
for comparative purposes in Fig. 4 is the strong decrease in
propane conversion that occurs with an increasing time de-
lay between the N2O and C3H8 pulses, approaching zero at
ꢀt = 2 s. The same behavior was observed in the yield of
the main products, C3H6 and CO2. In contrast, pump-probe
Related studies in agreement with the observations here
and in [14–16,22,28] have been also reported by Jia et
al. [38]. These authors investigated the exchange of 18O2
at 523 K with a Fe/MFI prepared by chemical vapor depo-
sition, which was previously treated in N2 16O at the same
temperature. The experiments unequivocally concluded that
the number of exchangeable oxygen atoms was smaller than
the number of oxygen atoms deposited by N2O. Using Fe-
ferrierite zeolites with Fe/Al = 0.03–0.6, Nováková et al.
13
13
experiments with N2O– CO show that the CO conver-
sion and 1 CO2 yield were not affected by the time delay
between the N2O and 13CO pulses.
3
[
39] also concluded that the amount of oxygen participating
The latter result apparently indicates that the reactivity
18
13
in O isotopic exchange at room temperature is 30–50%
lower than the amount of oxygen captured by the catalysts
upon N2O decomposition at 473 K. The results in [38,39]
agree well with earlier work by Leglise et al. [40] on the
of the oxygen species deposited by N2O toward CO ox-
idation does not change in the range of 0–2 s. In contrast,
the dependencies observed in the N2O–C3H8 system high-
light the vital importance of the lifetime of atomic oxygen

448
J. Pérez-Ramírez et al. / Journal of Catalysis 233 (2005) 442–452
13
Fig. 4. Conversion of reactants and yield of products vs time delay during sequential pulse experiments of N O–C H and N O– CO at 623 K (open
2
3
8
2
13
14
14
symbols) and 673 K (solid symbols). Pulse sizes of N O, C H , and CO were in the range of 2 × 10 –5 × 10 molecules.
2
3 8
∗
species deposited by nitrous oxide for propane conversion:
short-lived oxygen species at ꢀt = 0 s effectively activate
propane [Eq. (7)], which cannot be accomplished by long-
lived oxygen species at ꢀt > 0.1 s. Once again, this result
draws a remarkable parallel with previous work on the N2O
of Hall, it can be proposed that the number of active O
species, decreased because of their limited lifetime, may
involve their incorporation into vacancies of the zeolite ma-
trix. This implies the dynamic nature of the adsorbed oxy-
gen in the zeolite and the presence of oxygen vacancies in
the zeolite. The mobility of oxygen species (in both lat-
tice ions and adsorbed species) was elegantly demonstrated
by Valyon et al. [41] with step-switching experiments from
+
CH4 reaction over Fe-beta, where “short-lived oxygen” is
referred to as “nascent or hot oxygen transients,” and “long-
lived oxygen” is referred to as “thermally accommodated
oxygen.” From our experiments, a decrease in the concen-
tration of adsorbed atomic oxygen induced by its desorption
can be categorically excluded as a reason for the decreased
propane conversion at increased time delays, since no gas-
phase O2 was ever observed. Accordingly, it is reasonable
16
18
N2 O to N2 O over Fe-mordenite. In that work, it was
18
also found that ∼ 7 times more O could be accommo-
dated than Fe present in the sample. Accordingly, Fe was
proposed as a “porthole” through which this 18O was in-
corporated into the vicinity of the zeolite matrix. Further
support for this model comes from works by Nováková et
al. [39,42]. In these, the labilization of framework oxygen in
to postulate a rapid transformation of highly reactive species
∗
(
O , active pool) into highly stabilized species (Oꢀ, inactive
Fe-ferrierites was demonstrated by 18O isotopic exchange
pool), according to Eq. (8).
2
16
∗
∗
followed by TPD experiments over N O-pretreated sam-
2
C3H8 + O → C3H6, CO2 + H2O +
(7)
(8)
ples, since mostly 16O2 was analyzed at the reactor outlet
∗
∗
O + ꢀ → Oꢀ +
At this stage, no definitive experimental evidence has
been gathered to assess the nature and location of the pro-
posed Oꢀ species. Based on pioneering work by the group
above 473 K.
We have observed that the propane conversion and the
corresponding yield of products at ꢀt = 0.1 s and 0.5 s, for
example, did not change after ca. 200 cycles in sequential

J. Pérez-Ramírez et al. / Journal of Catalysis 233 (2005) 442–452
449
pulse experiments. This result indicates that the amount of
previous step by N2O decomposition, implying competitive
adsorption of CO and N2O for the same active iron sites.
Accordingly, this type of experiment merely examines the
elementary step represented by Eq. (2). The number of ad-
active sites for the N2O + C3H8 does not decrease, thus
∗
suggesting that active sites for C3H8 oxidation by O are
not eventually saturated by the presence of inactive Oꢀ
species. In agreement with this conclusion, the practically
unchanged N2O conversion with an increased time delay in
Fig. 4 suggests a rather constant number of active sites for
3
sorption sites during pump-probe experiments is 10 times
∗
higher than the amount of reactants pulsed, so that both O
∗
and CO species derived from N2O decomposition and CO
∗
the generation of highly reactive O species from nitrous
adsorption, respectively, may coexist. The validity hypothe-
sis has been proved in Section 3.5.
oxide decomposition. This can be tentatively explained by
∗
a certain balance between (i) the removal of O by propane
[
Eq. (7)], which prevails at ꢀt = 0 s, and (ii) the transforma-
3.5. Mechanism of N2O reduction in the presence of
adsorbed CO species
tion of reactive oxygen species into inactive ones [Eq. (8)],
which is favored by long time delays. According to this, the
slight decrease in N2O conversion that occurs with an in-
As expressed in Section 3.1, simultaneous pulsing of
∗
13
creased time delay is indicative of the faster removal of O
N2O–C3H8 and N2O– CO provided an indication of the
by the hydrocarbon as compared with the self-diffusion of
O into vacancies of the zeolite framework.
The transformation in Eq. (8) can also account for the
low ratio between the amount of CO2 produced in the titra-
tion studies with CO multi-pulses and the amount of atomic
oxygen deposited upon saturation of Fe-silicalite with N2O
different interactions of the reducing agents with Fe-silicalite.
The marked tailing of the transient response of carbon
monoxide and carbon dioxide as compared with that of
propane (compare Figs. 1b–c) denotes a strong interaction
of the former gases with the catalyst, inducing a very slow
∗
desorption process of both 1 CO and CO2. This feature
may be related to the constant 13CO conversion at different
time delays in Fig. 4. In the same manner, the weaker inter-
action of C3H8 with the catalyst (sharper responses, Fig. 1c)
could be associated with the fast decay of C3H8 conversion
as the time delay between the N2O and C3H8 pulses in Fig. 4
is increased.
3
13
(Fig. 3), and in related works assessing the reactivity of ad-
sorbed oxygen with CO [22]. It should be stressed that the
time between the last N2O pulse and the first CO pulse
in the multi-pulse experiments described in Sections 3.2
and 3.3 was ca. 1 min. Similarly, Kiwi-Minsker et al. [22]
deposited O species on H-ZSM-5 by N2O pretreatment at
5
23 K, followed by a 15-min purge with He, and the sub-
sequent admission of CO for 6 min at 523 K. Panov et al.
20,21] investigated the reaction of carbon monoxide with
These suppositions were substantiated by analyzing the
transient responses of the various species implicated during
the pump-probe experiments. The results in Fig. 5 are illus-
trative of the different affinities of the two reducing agents
for the iron species in Fe-silicalite. In the case of N2O–C3H8
experiments (Fig. 5a), the response of propane returns to
the baseline ca. 1 s after being pulsed at t = 2 s. In addi-
tion, as expected, no C-containing products were identified
in the N2O pulse at t = 0 s. In contrast, the transient response
[
α-oxygen according to a similar procedure: deposition of
Oα by N2O decomposition at 523 K, followed by cooling
to 373 K, and subsequent reaction with CO. With this ex-
perimental methodology and based on our results and [22],
the process in Eq. (8) is expected to take place extensively,
and the CO:Oα stoichiometry of 1:1 obtained by Panov et al.
cannot be substantiated by our observations.
Several causes can be put forward to explain the remark-
ably constant conversion of carbon monoxide in sequential
N2O– CO pulsing experiments in Fig. 4. In contrast to the
case of propane, it could be inferred that the adsorbed oxy-
gen species generated from N2O have the same reactivity
13
13
of CO in N2O– CO experiments (Fig. 5b) reaches the
baseline ca. 5 s after being pulsed at t = 1 s, indicating the
strong adsorption and slow desorption of carbon monoxide.
This statement is reinforced by the observation of substan-
tial amounts of 13CO desorbing from the catalyst at the time
of the N2O pulse at t = 0 s. Blank experiments excluded this
signal to be caused by an impurity with AMU 29 in the N2O-
containing mixture. This 13CO stayed on the catalyst surface
for at least 5 s (period of time from the 13CO pulse at t = 1 s
to the next N2O pulse) under vacuum conditions. This result
indicates that N2O decomposes and deposits oxygen species
on the same iron sites where CO was adsorbed, inducing its
desorption according to Eq. (9).
13
13
for CO oxidation in the range of time delays investigated.
Based on our previous discussion, this suggests that the
eventual transformation of the short-lived oxygen species
does not lead to an inactive form of adsorbed oxygen for
13
CO oxidation. This is rather unlikely, however, based on
the low reactivity of oxygen species deposited by N2O in
the titration experiments with 13CO multi-pulses, even if the
1
-min interval between the last N2O pulse and the first CO
CO + N2O → N2 + CO + O∗
∗
(9)
pulse largely exceeds the maximum time delay of 2 s in
pump-probe experiments. Another explanation for the ob-
served behavior in Fig. 4 with carbon monoxide is that this
reactant would catalyze, in an adsorbed form, the reduction
of N2O. CO adsorption in the titration experiment is adverse
because of the blockage by oxygen species deposited in the
Additional insights into the mechanism of the N2O +
3
1
CO reaction can be obtained from an analysis of the CO2
response in pump-probe experiments at 673 K and its de-
pendence on the pulse size. This is illustrated in Fig. 6 for
ꢀt = 1 s, using CO pulse sizes of 2×10 (black line) and
13
14

450
J. Pérez-Ramírez et al. / Journal of Catalysis 233 (2005) 442–452
1
3
13
Fig. 7. Transient responses of CO and CO during sequential pulsing
2
13
of N O and CO with ꢀt = 1 s over Fe-silicalite at 673 K. Pulse size of
2
14
13
15
N O = 2 × 10 molecules and pulse size of CO ≈ 1 × 10 molecules.
2
sions of 80 and 73% were obtained, respectively. Increasing
13
the pulse size of CO by one order of magnitude led to
complete N2O conversion, whereas the 13CO conversion de-
creased to 20%. This is expected, since the lower amount of
N2O limits the degree of conversion of the reducing agent
[
23,43]. An important observation from this experiment is
13
the comparable intensity of the transient response of CO2
in both N2O and 13CO pulses, which unequivocally indicates
two reaction routes for carbon dioxide production, occurring
to a similar extent. This feature was never observed in the
N2O–C3H8 system; that is, no residual propane or carbon-
containing reaction products were observed in the pulse of
N2O, regardless of the pulse size of propane. It should also
Fig. 5. Transient responses during sequential pulsing of (a) N O and C H
3 8
2
13
1
3
at ꢀt = 2 s and (b) N O and CO at ꢀt = 1 s over Fe-silicalite at 673 K.
be pointed out that the CO2 response at the time of the
2
1
3
Pulse sizes as in caption of Fig. 4.
CO pulse presented a more extensive tailing (up to 9 s)
when larger pulse sizes of carbon monoxide were applied.
This is rather logical, because of the higher surface concen-
tration of C-containing species.
In view of the order of the pulses in our sequential N2O–
CO experiments, it can be proposed that the classic scav-
13
enging mechanism proposed in Eqs. (1) and (3) accounts
for the 13CO2 formed at the time of the CO pulse. This
pathway principally describes the N2O reduction by light
hydrocarbons like CH4 [14–16] and C3H8 [18]. However, in
13
∗
the case of the N2O + CO reaction, the previous O -assisted
CO oxidation is complemented by the N2O conversion in the
∗
presence of adsorbed CO species (denoted as CO -assisted
13
N2O reduction, Eqs. (4) and (5)). The appearance of CO
and CO2 as reaction products in the N2O pulse (cf. Figs. 5b
and 7) suggests that strongly adsorbed CO species can un-
13
1
3
13
Fig. 6. Influence of the CO pulse size on the transient responses of CO
2
1
3
and attained degree of CO conversion during sequential pulsing of N O
2
13
13
dergo two processes in the presence of N O: reaction of
and CO at ꢀt = 1 s over Fe-silicalite at 673 K. Pulse size of CO was
2
4
15
2
× 101 molecules (black line) or 1 × 10 molecules (gray line), keeping
N O with adsorbed CO to yield CO [Eq. (5)] and desorp-
2
2
1
4
constant the pulse size of N O at 2 × 10 molecules.
2
2
tion of CO species by N O [Eq. (9)]. As shown in Fig. 7, the
13
relatively higher intensity of the CO2 signal as compared
with the 13CO signal at the time of the N2O pulse during
sequential pulsing at 673 K indicates that Eq. (5) prevails
over Eq. (9). It can be tentatively suggested that the step
in Eq. (5) alternatively occurs in the presence of adsorbed
species [Eq. (10)]. However, assessing a detailed molecular
1
5
1
×10 (gray line) molecules. The N2O pulse size was kept
14 13
constant at 2×10 molecules. At equal pulse sizes of CO
13
and N2O, an intense CO2 response appeared at the time of
13
the CO pulse, and only a minor signal was identified at the
time of the N2O pulse. Furthermore, CO and N2O conver-
13

J. Pérez-Ramírez et al. / Journal of Catalysis 233 (2005) 442–452
451
∗
description of the CO -assisted N2O reduction requires fur-
This agrees with our previous findings and observations
by other authors for the reaction between N2O and CH4,
suggesting a certain regularity of the activation of light
hydrocarbons by short-lived oxygen species from N2O,
which rapidly transform and lose their unique reactive
character. In contrast to the N2O–C3H8 system, the life-
time of the adsorbed oxygen species did not influence
the activation of 13CO over Fe-silicalite in the range of
time delays investigated (0–2 s).
ther research, since the nature of the exact adsorption site(s)
of the participative CO and O species is not completely
clear at this stage.
∗
∗
∗
∗
∗
CO + O → CO2 + 2
In any case, we believe that the reduction of N2O in the
presence of adsorbed CO species is responsible for the in-
variable conversion of 13CO and stable yield of CO2 at
different time delays during sequential N2O– CO pulse ex-
periments in Fig. 4. Accordingly, the pathway represented
by Eqs. (4) and (5) seems to overcome the transformation
(10)
13
13
– Both the N2O-mediated 13CO and C3H8 oxidations
proceed via the classic scavenging mechanism, where
the reducing agent is activated by short-lived oxygen
species deposited by N2O on oligonuclear iron species
in the zeolite. Furthermore, the reaction of nitrous oxide
with carbon monoxide incorporates an additional path-
way for CO2 production, involving the reaction of N2O
with chemisorbed CO species. This CO-mediated N2O
reduction occurs in view of the strong interaction of CO
on isolated Fe(III) ions in extraframework positions of
the zeolite, as supported by previous in situ UV/vis and
EPR spectroscopic studies. This unique feature of CO
may account for the remarkably high efficiency of CO
for the low-temperature SCR of N2O.
∗
of O species into inactive forms [Eq. (8)] in the ꢀt range
investigated, which is detrimental to the extent of the path-
way represented by Eqs. (1)–(3) (as shown in sequential
N2O–C3H8 pulse experiments in Section 3.4 and in titration
experiments with CO in Section 3.3).
The identification of two reaction pathways in the reac-
tion of N2O and CO over Fe-silicalite with a transient pulse
technique, the temporal analysis of products reactor, agrees
well with recent in situ EPR and UV/vis spectroscopic in-
vestigations by some of us [23]. These concluded that the
∗
above processes are site-dependent: the O -assisted CO ox-
idation [Eqs. (1) and (3)] is favored over oligonuclear iron-
oxo species and is able to coordinate highly reactive oxygen
species by N2O decomposition [Eq. (1)]. The newly re-
Acknowledgments
∗
ported CO -assisted N2O reduction mainly involved isolated
Fe(III) ions in extraframework positions. Steady-state exper-
iments in [23] showed that the highest activity per mole of
iron site was obtained for the Fe-silicalite investigated here,
which has the highest concentration of such isolated sites
among the different materials investigated. Accordingly, the
M.N.D. thanks AECI for a fellowship to complete his
doctoral studies.
References
created Fe3 –CO species can be considered to have a very
high activity for N2O reduction. This may be responsible for
the high efficiency of CO in the low-temperature catalytic
N2O reduction over iron-containing zeolites.
+
[
[
[
[
[
1] J. Pérez-Ramírez, F. Kapteijn, G. Mul, J.A. Moulijn, Chem. Commun.
2001) 693.
2] J. Pérez-Ramírez, F. Kapteijn, G. Mul, J.A. Moulijn, Appl. Catal. B 35
2002) 227.
3] J. Pérez-Ramírez, F. Kapteijn, G. Mul, X. Xu, J.A. Moulijn, Catal.
Today 76 (2002) 55.
4] J. Pérez-Ramírez, F. Kapteijn, K. Schöffel, J.A. Moulijn, Appl. Catal.
B 44 (2003) 117.
5] R.W. van den Brink, S. Booneveld, J.R. Pels, D.F. Bakker, M.J.F.M.
Verhaak, Appl. Catal. B 32 (2001) 73.
(
(
4
. Conclusions
Our TAP studies have elucidated novel mechanistic fea-
tures of the reaction between N2O and 13CO or C3H8 over
Fe-silicalite:
[6] G. Centi, F. Vazzana, Catal. Today 53 (1999) 683.
[7] J. Pérez-Ramírez, F. Kapteijn, Appl. Catal. B 47 (2004) 177.
[8] K. Yamada, S. Kondo, K. Segawa, Micropor. Mesopor. Mater. 35–36
(
2000) 227.
–
–
Multi-pulse experiments concluded that N2O deposits
ca. 1018 atoms of oxygen per gram of catalyst upon
decomposition at 523–573 K, although subsequent titra-
tion showed that a small fraction of the stored oxygen
participates in CO oxidation. This suggests the presence
of different pools of oxygen species on the catalyst with
a markedly distinct oxidizing power.
In line with this, pump-probe experiments at 623–673 K
showed that Fe-silicalite effectively activates propane
when N2O and C3H8 are pulsed together. However,
these oxygen species become inactive for propane ox-
idation if the pulses are separated by more than 0.1 s.
[
9] T. Chaki, M. Arai, T. Ebina, M. Shimokawabe, J. Catal. 218 (2003)
220.
[10] S. Kameoka, K. Kita, T. Takeda, S. Tanaka, S. Ito, K. Yuzaki, T.
Miyadera, K. Kunimori, Catal. Lett. 69 (2000) 169.
11] M. Yoshida, T. Nobukawa, S. Ito, K. Tomishige, K. Kunimori,
J. Catal. 223 (2004) 454.
12] T. Nobukawa, M. Yoshida, K. Okumura, K. Tomishige, K. Kunimori,
J. Catal. 229 (2005) 374.
13] G. Delahay, M. Mauvezin, A. Guzmán-Vargas, B. Coq, Catal. Com-
mun. 3 (2002) 385.
14] S. Kameoka, T. Nobukawa, S. Tanaka, S. Ito, K. Tomishige, K. Kuni-
mori, Phys. Chem. Chem. Phys. 5 (2003) 3328.
[
[
[
[
[
15] T. Nobukawa, M. Yoshida, S. Kameoka, S. Ito, K. Tomishige, K. Ku-
nimori, J. Phys. Chem. B 108 (2004) 4071.

452
J. Pérez-Ramírez et al. / Journal of Catalysis 233 (2005) 442–452
[
16] T. Nobukawa, M. Yoshida, S. Kameoka, S. Ito, K. Tomishige, K. Ku-
[29] D.A. Bulushev, L. Kiwi-Minsker, A. Renken, J. Catal. 222 (2004) 389.
[30] J. Pérez-Ramírez, J. Catal. 227 (2004) 512.
nimori, Catal. Today 93–95 (2004) 791.
[
[
[
17] C.T. Au, M.W. Roberts, Nature 319 (1986) 206.
[31] J. Pérez-Ramírez, E.V. Kondratenko, Chem. Commun. (2003) 2152.
[32] J. Pérez-Ramírez, A. Gallardo-Llamas, J. Catal. 223 (2004) 382.
[33] G.I. Panov, A.K. Uriarte, M.A. Rodkin, V.I. Sobolev, Catal. Today 41
(1998) 365.
18] E.V. Kondratenko, J. Pérez-Ramírez, Appl. Catal. A 267 (2004) 181.
19] F. Kapteijn, G. Mul, G. Marbán, J. Rodríguez-Mirasol, J.A. Moulijn,
Stud. Surf. Sci. Catal. 101 (1996) 641.
[20] G.I. Panov, V.I. Sobolev, A.S. Kharitonov, J. Mol. Catal. A. Chem. 61
[34] G.I. Panov, Cattech 4 (2000) 18.
(
1990) 85.
[35] K.A. Dubkov, V.I. Sobolev, G.I. Panov, Kinet. Catal. 39 (1998) 79.
[36] K.A. Dubkov, N.S. Ovanesyan, A.A. Shteinman, E.V. Stakoron, G.I.
Panov, J. Catal. 207 (2002) 341.
[37] J. Nováková, M. Schwarze, Z. Tvar u˚ zková, Z. Sobalík, Catal. Lett. 98
(2004) 123.
[
[
[
21] K.A. Dubkov, E.A. Paukshtis, G.I. Panov, Kinet. Catal. 42 (2001) 230.
22] L. Kiwi-Minsker, D.A. Bulushev, A. Renken, J. Catal. 219 (2003) 273.
23] J. Pérez-Ramírez, M. Santosh Kumar, A. Brückner, J. Catal. 223
(2004) 13.
[
[
[
[
[
24] J. Pérez-Ramírez, F. Kapteijn, J.C. Groen, A. Doménech, G. Mul, J.A.
Moulijn, J. Catal. 214 (2003) 33.
25] J.T. Gleaves, J.R. Ebner, T.C. Kuechler, Catal. Rev.-Sci. Eng. 30
[38] J. Jia, B. Wen, W.M.H. Sachtler, J. Catal. 210 (2002) 453.
[39] J. Nováková, M. Lhotka, Z. Tvar u˚ zková, Z. Sobalík, Catal. Lett. 83
(2002) 215.
[40] J. Leglise, J.O. Petunchi, W.K. Hall, J. Catal. 86 (1984) 392.
[41] J. Valyon, W.S. Millman, W.K. Hall, Catal. Lett. 24 (1994) 215.
[42] J. Nováková, Z. Sobalík, Catal. Lett. 89 (2003) 243.
[43] M.N. Debbagh Boutarbouch, J.M. García Cortés, M. Soussi El Be-
grani, C. Salinas Martínez de Lecea, J. Pérez-Ramírez, Appl. Catal.
B 54 (2004) 115.
(1988) 49.
26] J.T. Gleaves, G.S. Yablonsky, P. Phanawadee, Y. Schuurman, Appl.
Catal. A 160 (1997) 55.
27] J. Pérez-Ramírez, G. Mul, F. Kapteijn, J.A. Moulijn, J. Catal. 208
(2002) 211.
28] G.D. Pirngruber, J. Catal. 219 (2003) 456.