Reduction of N2O with CO over FeMFI zeolites: Influence of the preparation method on the iron species and catalytic behavior

Article abstract of DOI:10.1016/j.jcat.2004.01.007

The reduction of N₂O by CO was studied over FeMFI zeolites prepared by differential methods including sublimation, liquid ion exchange, and hydrothermal synthesis followed by steam activation. In the presence of CO, the conversion of N₂

Full text of DOI:10.1016/j.jcat.2004.01.007

Journal of Catalysis 223 (2004) 13–27
www.elsevier.com/locate/jcat
Reduction of N O with CO over FeMFI zeolites: influence of the
2
preparation method on the iron species and catalytic behavior
a,∗
b
b
Javier Pérez-Ramírez, M. Santhosh Kumar, and Angelika Brückner
a
Norsk Hydro, Agri Research Centre, Nitric Acid Technology, PO Box 2560, N-3907, Porsgrunn, Norway
Institute for Applied Chemistry Berlin–Adlershof, Richard-Willstätter-Str.12, D-12484, Berlin, Germany
b
Received 11 November 2003; revised 6 January 2004; accepted 6 January 2004
Abstract
The reduction of N O by CO was investigated over FeMFI zeolites prepared by different methods including sublimation, liquid ion
2
exchange, and hydrothermal synthesis followed by steam activation. This leads to catalysts with different nature and distribution of iron
species, as characterized by HRTEM, UV/vis, and EPR techniques. A common denominator in the FeZSM-5 samples is the heterogeneous
iron constitution, with a significant degree of clustering in the form of iron oxide particles. Iron clustering was suppressed in steam-activated
Fe-silicalite, presenting a remarkable uniform distribution of isolated iron species. In the presence of CO, the conversion of N O over the
2
catalysts is strongly accelerated with respect to direct N O decomposition. The reaction rate increases linearly with the molar CO/N O feed
2
2
ratio and strongly depends on the preparation method applied. A correlation was found between the fraction of isolated Fe(III) species in
the as-prepared catalysts and the activity for N O reduction with CO. Steam-activated Fe-silicalite, containing mostly isolated iron ions in
2
extraframework positions, shows the highest activity per mole of iron, while the highly clustered liquid-ion-exchanged catalyst presents the
lowest activity. In situ UV/vis and EPR studies evidence the participation of mononuclear iron ions in the N O–CO reaction, and also support
2
the involvement of oligonuclear FexOy species. The reaction mechanism is iron site dependent. Over isolated sites, the reduction of N O
2
3
+
3+
2+
with CO occurs via coordinated CO species on Fe ions. The reaction over oligonuclear sites proceeds via a redox Fe /Fe process
−
with intermediate formation of O radicals.

2004 Elsevier Inc. All rights reserved.
Keywords: Fe-zeolites; FeMFI; Preparation method; N O decomposition; N O reduction; CO; Iron species; Active sites; Mechanism; In situ UV/vis; In situ
2
2
EPR
1
. Introduction
were postulated for regeneration of the active site [1,4]: the
reaction of a second N2O molecule with the oxidized site
[Eq. (2)] and the recombination of adsorbed oxygen atoms
Fe-zeolites show a remarkable performance in various
N2O conversions, including its abatement (N2O decompo-
sition and selective catalytic reduction) as well as its utiliza-
tion as a selective oxidant [1–3]. Early studies with this cat-
alytic system aimed at elucidating mechanistic and kinetics
aspects of the direct N2O decomposition [4,5]. This reaction
is initiated by the activation of N2O on a vacant site, leading
[Eq. (3)]. Some of us concluded that O is mainly produced
by recombination of atomic oxygen based on TAP (tempo-
2
ral analysis of products) experiments under high vacuum in
the temperature range of 573–973 K [7,8], while Nobukawa
et al. [9] obtained experimental evidence for Eq. (2) using an
1
8
O-tracer technique at 693 K and atmospheric pressure.
∗
to O species [Eq. (1)]. This atomic oxygen species, other-
∗
∗
N2O + → N2 + O ,
(1)
(2)
(3)
(4)
wise referred to as α-oxygen, has unique properties in selec-
tive oxidation processes involving hydrocarbons, like the di-
rect hydroxylationof benzene to phenol [3] and the oxidative
dehydrogenation of propane to propene [6]. In direct N2O
decomposition over Fe-zeolites, two different mechanisms
∗
∗
N O + O → N + O + ,
2
2
2
∗
∗
2
O → O2 + 2 ,
∗
∗
CO + O → CO2 + .
Although the mechanism of O2 formation is not com-
pletely resolved, it is widely accepted that oxygen desorp-
tion is a rate-determining step during N2O decomposition
*
Corresponding author.
E-mail address: javier.perez.ramirez@hydro.com (J. Pérez-Ramírez).
0021-9517/$ – see front matter  2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2004.01.007

14
J. Pérez-Ramírez et al. / Journal of Catalysis 223 (2004) 13–27
over Fe-zeolites [7–9]. The presence of reducing agents ac-
celerates the removal of atomic oxygen from the catalyst
surface by selective catalytic reduction (SCR), resulting in
a decreased operation temperature. This magnitude ranges
between 50 and 100 K and depends on the nature of the
reductant and its partial feed pressure. The SCR of N2O
over Fe-zeolites has been investigated using various hydro-
carbons, ammonia, and carbon monoxide [10–12]. Delahay
et al. [11] concluded that CO was the most efficient se-
lective reductant for N2O over ion-exchanged Fe-BEA at
low temperatures in the presence of oxygen, showing sig-
nificant activities at 470 K. Previously, Kapteijn et al. [12]
reported the enhanced N2O decomposition activity over
an ion-exchanged FeZSM-5 in the presence of CO around
Table 1
Chemical composition of the catalysts used in this study, as determined by
ICP-OES
Catalyst
Method
Si/Al
Si/Fe
Fe (wt%)
ex-Fe-silicalite
ex-FeZSM-5
Fe/ZSM-5
Steam activation
Steam activation
Sublimation
∞
31
14
37
150
150
14
0.68
0.67
5.0
Fe-ZSM-5
Liquid ion exchange
65
1.4
Details on the preparation and characterization of the cat-
alysts can be found elsewhere [8,20,21]. Ex-FeZSM-5 and
ex-Fe-silicalite were prepared by isomorphous substitution
of iron in the zeolite framework followed by calcination (at
8
23 K for 10 h) and steam activation (partial steam pressure
6
73 K. The proposed microkinetic model assumes that CO
−1
of 300 mbar H2O and 30 ml STP min of N2 flow at 873 K
for 5 h). Fe/ZSM-5 was prepared by sublimation of FeCl3 on
HZSM-5 (Degussa), according to the method described by
Chen and Sachtler [22]. Fe-ZSM-5 was prepared by liquid
removes adsorbed atomic oxygen according to an Eley–
Rideal mechanism [Eq. (4)]. Following the reaction pathway
in Eqs. (1)–(4), in the presence of an efficient oxygen
remover like CO, the overall rate of N2O reduction can ap-
proach that of N2O activation by Eq. (1). Accordingly, CO
can probe the reactivity of oxygen species adsorbed on vari-
ous iron sites of Fe-zeolites, resulting from N2O activation.
Panov and co-workers [13,14] concluded that CO reacts with
α-oxygen in a stoichiometric ratio 1:1, further supporting a
reduction mechanism via Eq. (4).
(
aqueous) ion exchanged with diluted solutions (0.30 mM)
of Fe(NO3)3 · 9H2O. The parent zeolite was provided by Ze-
olyst (NH4-ZSM-5; CBV8020). The exchange was carried
out at room temperature under vigorous stirring for 15 h,
without control of the pH. The ion-exchanged zeolite was
then filtered, washed thoroughly, dried, and finally calcined
in static air at 823 K for 5 h.
To the best of our knowledge, detailed studies of the re-
action between N2O and CO over Fe-zeolites with different
forms of iron species have not been reported. The prepara-
tion method of iron-zeolites is critical for obtaining catalysts
with a desired performance [15–18], and the intrinsic reac-
tion mechanism ultimately determines the optimal structures
of the active site [19]. Hence, fundamental knowledge of the
N2O–CO reaction with respect to the active iron site archi-
tecture(s) and mechanism will have valuable implications for
catalyst design in conversions involving N2O.
The present work was undertaken to examine the cat-
alytic behavior of differently prepared FeMFI zeolites in the
reduction of N2O by CO. The nature and distribution of
iron in the catalysts was determined by HRTEM, UV/vis,
and EPR techniques. Steady-state activity tests at different
molar CO/N2O feed ratios in the temperature range of 475–
2
.2. Catalyst characterization and in situ studies
Transmission electron microscopy (HRTEM) was carried
out on a Philips CM 30 T electron microscope with a LaB6
filament as the source of electrons operated at 300 kV. The
zeolites were amorphized by the electron beam in order to
enhance the visibility of the small iron oxide particles.
UV/vis-DRS (diffuse reflectance spectroscopy) measure-
ments were performed with a Cary 400 spectrometer (Var-
ian) equipped with a diffuse reflectance accessory (Praying
Mantis, Harrick). To reduce light absorption, samples were
diluted by α-Al2O3 (calcined at 1473 K for 4 h) at a ratio
of 1:3. Deconvolution of the UV/vis spectra into subbands
was performed applying Gaussian lineshape functions with
GRAMS/32 (Galactic), which uses a least-square fitting al-
gorithm. In situ UV/vis measurements were carried out in
a heatable reaction chamber equipped with a temperature
programmer and a gas-dosing system with mass-flow con-
trollers. The effect of reductive and oxidative treatments over
some samples was analyzed by treatment with 200 mbar
H2 in N2 and air at 773 K and ambient pressure for 1 h,
respectively. Afterward, the sample was cooled down to
room temperature in the same mixture and the spectrum was
then recorded. In situ experiments with N2O and CO were
performed according to the following sequence: (a) pretreat-
9
00 K evidenced the influence of the iron constitution in
the zeolites on the catalytic performance. The correlation
of steady-state activity data with characterization by ex situ
techniques and in situ UV/vis and EPR spectroscopies en-
ables insights be gained into the nature of the active species
involved in the catalytic process and the associated reaction
mechanism.
2
. Experimental
−1
2
.1. Catalyst preparation
ment in air (15 ml min ) at 773 K for 1 h and cooling down
to 623 K (denoted as the reaction temperature), (b) chang-
ing the flow from air to 20 mbar N2O in He (12.5 ml min
for 1 h, (c) switching to 20 mbar CO in He (12.5 ml min
−
1
1
The chemical composition of the zeolites used in this
study was determined by ICP and is presented in Table 1.
)
)
−

J. Pérez-Ramírez et al. / Journal of Catalysis 223 (2004) 13–27
15
for 1 h, and finally (d) switching back to 20 mbar N2O in He
3. Results
−1
(
12.5 ml min ) at the reaction temperature.
EPR spectra in X band (ν ≈ 9.5 GHz) were recorded with
3.1. Characterization of iron species
the cw spectrometer ELEXSYS 500-10/12 (Bruker). Mea-
surements were typically performed at 298 and 77 K and
for some samples spectra were recorded after evacuation at
Table 1 shows the chemical composition of the catalysts
used in this study. The iron content in the steam-activated
catalysts is very similar (∼0.68 wt% Fe), about three and
eight times lower than in the catalysts prepared by ion ex-
change in aqueous solution (1.4 wt% Fe) and in gas phase
−2
1
0
Pa. The magnetic field was measured with respect to
the standard 2,2-diphenyl-1-picrylhydrazylhydrate (DPPH).
In situ measurements were carried out in a self-constructed
EPR packed-bed reactor [23] using the sequence described
for in situ UV/vis experiments. Additional in situ EPR ex-
periments over Fe/ZSM-5 were performed at 623 K and
atmospheric pressure with mixtures of 10 mbar N2O + 10
(
5 wt% Fe), respectively. The Al content in ex-Fe-silicalite
was below the detection limit of the technique (<100 ppm).
The morphology of the different catalysts was investi-
gated by HRTEM (Fig. 1). The catalysts prepared by liquid
ion exchange and sublimation showed an extensive agglom-
eration of iron species into large iron oxide particles of a
certain size distribution. In general, particles were larger in
Fe-ZSM-5 (7–25 nm) despite the lower iron content com-
pared to Fe/ZSM-5, where the particle size distribution is
in the range of 3–12 nm. Exceptionally, larger particles
were identified in the later sample. Iron clustering in the
steam-activated catalysts, with a relatively lower iron con-
tent, shows important differences compared to the catalysts
prepared by ion-exchange and sublimation methods. The de-
gree of iron clustering in ex-FeZSM-5 is significant, but
relatively small particles with a homogeneous distribution
−
1
mbar CO in He (25 ml min ) and 6.6 mbar N2O + 13.4
−1
mbar CO in He (18.7 ml min ). For on-line analysis of N2
formed during the N2O–CO reaction, the outlet of the EPR
reactor was connected to a capillary gas chromatograph(Shi-
madzu GC 17AAF) equipped with a thermal conductivity
detector and a Molsieve 5A column.
2
.3. Activity tests
Activity measurements were carried out in a parallel-flow
(
1–2 nm) were identified. Remarkably, no iron oxide parti-
reactor system, similar to that described in [24], using 50 mg
cles were detected in ex-Fe-silicalite, indicating the high iron
dispersion in this matrix.
UV/vis-DRS spectroscopy was used to investigate the
nature of Fe(III) species in the catalysts. The measured spec-
tra of as-prepared zeolites without any pretreatment were
converted into Kubelka–Munk functions and deconvoluted
5
of catalyst (125–200 µm) and space times of 3 × 10 and
5
−1
9
× 10 g s mol at atmospheric pressure. The space time
is defined as the ratio W/F(N2O)o, where W is the cata-
lyst mass and F(N2O)o the molar flow of N2O at the reactor
inlet. The catalyst beds were homogeneously diluted with
silicon carbide (125–200 µm) in a volumetric ratio 1:1, in
order to improve heat transfer and minimize axial dispersion
in the bed. At such a degree of dilution the conversion mea-
sured is not affected by the presence of diluent [25]. Feed
mixtures containing N2O (1.5 mbar) and CO (0–1.5 mbar)
in He were applied. Before reaction, the catalysts were pre-
treated in 1.5 mbar N2O in He at 723 K for 1 h and cooled in
that gas flow to the initial reaction temperature. The tem-
perature was increased at intervals of 25 K in the range
3
+
into Gaussian subbands that can be assigned to Fe ← O
charge-transfer (CT) bands (Fig. 2). This deconvolution was
made based on the below assignments and is quantified
in Table 2. It is well known that for Fe3 sites, two CT bands
+
associated to t1 → t2 and t1 → e transitions are expected.
For isolated Fe3 species, they fall in the high-energy range
of the spectrum [26–28]. Thus, the first maximum of the
spectra below 300 nm in Fig. 2 has been deconvoluted into
two subbands, which are attributed to the t1 → t2 and t1 → e
+
3
+
transitions of isolated Fe ions. It is complicated to discern
4
75–900 K. This cycle was followed by a stepwise decrease
whether the isolated Fe3 species are in tetrahedral or higher
coordination. CT bands of tetrahedral iron incorporated in
the framework of silicalite have been observed at 215 and
+
of temperature at 25 K intervals. In this temperature range
deactivation of the catalysts was absent and up and down cy-
cles lead to very similar activity curves. Generally, 1 h after
a change of conditions (temperature and feed composition),
the N2O conversion levels were constant and considered as
the steady state. The product gases were analyzed with a GC
2
41 nm [26], and a band at 278 nm is observed for isolated
3+
Fe species in octahedral coordination in Al2O3 [27].
3+
Octahedral Fe
ions in small oligonuclear clusters
3+
(
Fe Oy) give rise to broad bands between 300 and 450 nm
x
(
Chrompack CP 9001) equipped with a thermal conductiv-
and bands above 450 nm are characteristic for larger Fe2O3
particles [26]. To determine the nuclearity and structure of
these small oligonuclear iron clusters is a very challenging
aspect but is out of the scope of this paper. Analogously to
ity detector, using a Poraplot Q column (for CO2 and N2O
separation) and a Molsieve 5A column (for N2, O2, and CO
separation). During the experiments, the mass balances of N,
C, and O closed at >98%. N2O and CO conversions were
calculated from the amount of N2 and CO2 formed, respec-
tively.
3
+
the case of isolated Fe ions, t → t and t → e transitions
1
2
1
are expected for Fe3 in oligonuclear Fe O species. Light
+
x
y
absorption in the spectra of Fig. 2 above 300 nm occurs in a

16
J. Pérez-Ramírez et al. / Journal of Catalysis 223 (2004) 13–27
Fig. 1. HRTEM micrographs of the FeMFI zeolites investigated in this study.
Table 2
Percentage of the area of the subbands (I at λ < 280 nm, I at 280 <
the extinction coefficient on the wavelength. Despite this fact
and the intrinsic uncertainty of the deconvolution procedure,
the quantification in Table 2 provides a valuable indication of
the relative amount of various iron site structures among the
catalysts investigated. The values in Table 2 do not take into
1
2
λ < 400 nm, and I at λ > 400 nm) estimated by deconvolution of the
3
UV/vis-DRS spectra (I
) of the catalysts (see Fig. 2) and correspond-
total
ing Fe percentage derived from total Fe content (see Table 1) (assumptions
applied in the fitting procedure are described in the text)
account the contribution of eventual Fe2 species present
+
a
b
c
I
Catalyst
I
Fe
1
I
Fe
2
Fe
3
I
total
1
2
3
in the as-prepared samples, although no significant increase
(%) (wt%) (%) (wt%) (%) (wt%)
(a.u.)
2
+
of the UV/vis band intensities, due to oxidation of Fe to
ex-Fe-silicalite
ex-FeZSM-5
Fe/ZSM-5
70
0.47
0.21
1.30
0.21
30
62
67
52
0.20
0.41
3.40
0.75
–
8
6
–
31.0
30.0
85.8
57.6
3
+
Fe , has been observed after air treatment of the catalysts
at 773 K (see below and Fig. 3).
30
26
15
0.05
0.30
0.47
3
+
Fe-ZSM-5
33
In general, d−d transitions of Fe are expected between
350 and 550 nm. However, they are symmetry and spin
forbidden and, therefore, extremely weak. Moreover, they
overlap with the tail of the low-energy CT bands. Hence,
interpretation of the spectra is focused on the intense CT
bands. The spectrum of ex-Fe-silicalite in Fig. 2 and the rel-
ative percentage of the deconvoluted subbands in Table 2
a
3+
Isolated Fe ions in tetrahedral and higher coordination.
b
c
3+
Oligonuclear Fex Oy clusters.
Large Fe O particles.
2
3
very broad range, and therefore certain oligonuclear clusters
and larger Fe2O3 species may overlap in the deconvoluted
bands. Nevertheless, these subbands should be understood
in terms of reflecting a semiquantitative distribution of dif-
ferent cluster geometries and sizes rather than representing
a certain number of different individual clustered species.
The relative percentage of the different iron species de-
termined in Table 2 does not account for the dependence of
indicate that the majority of Fe3 species in the sample is
well isolated. The band at ∼300 nm suggests that a rela-
tively small fraction of iron is present as small oligonuclear
species. No contribution >450 nm was observed, indicating
the absence of large Fe2O3 particles, in good agreement with
HRTEM. In contrast to ex-Fe-silicalite, the percentage of
+

J. Pérez-Ramírez et al. / Journal of Catalysis 223 (2004) 13–27
17
Fig. 2. UV/vis-DRS spectra and deconvoluted subbands of the as-prepared FeMFI zeolites, without pretreatment.
isolated Fe3 species is significantly lower in ex-FeZSM-5,
+
hardly visible in Al-free ex-Fe-silicalite. In agreement with
our results, previous Mössbauer investigations of FeZSM-5
zeolites, prepared by adding Fe to the synthesis gel, have
3
+
with extensive formation of Fex Oy clusters and a small
fraction of Fe2O3 particles (Fig. 2 and Table 2). The latter
can be attributed to the 1- to 2-nm nanoparticles observed
in this sample by electron microscopy (Fig. 1). The fraction
of oligonuclear oxo clusters and iron oxide particles in the
catalysts prepared by liquid ion exchange and sublimation
is even more pronounced. As concluded from the deconvo-
lution in Table 2, Fe-ZSM-5 shows the highest percentage
of large Fe2O3 particles, which is also in line with HRTEM
results.
concluded a more favorable formation of Fe2 in matrices
+
with relatively high Al contents [30].
Previous voltammetric characterization has shown that
the extraction of framework iron in ex-Fe-silicalite is sub-
stantial after steam treatment at 873 K, but not com-
plete [21]. As noted previously, the single application of
UV/vis does not enable a clear discrimination between iso-
lated Fe3 ions in framework or extraframework positions.
+
3
+
Heating ex-FeZSM-5 in a flow of air for 1 h at 773 K leads
to a spectrum with slightly higher intensities (not shown)
However, the influence of Fe reduction by H2 and reox-
idation by air on the UV/vis spectrum of the calcined and
steamed Fe-silicalite samples can be used to further assess
the degree of iron extraction. After hydrothermal synthe-
sis and calcination at 823 K, Fe-silicalite shows typical CT
bands at 215 and 241 nm (Fig. 3), originating from iron in
the lattice. These bands were not sensitive to a reductive
2
+
to of the partial oxidation of Fe , present in the original
sample, to Fe3 . Divalent iron does not contribute to the
UV/vis spectrum, since its bands fall in the near-infrared
range around 1000 nm [29]. A certain oxidation of iron(II)
was also observed in Fe/ZSM-5 and Fe-ZSM-5 but was
+
Fig. 3. UV/vis-DRS spectra of the calcined and steamed Fe-silicalite zeolites without pretreatment, after reduction in 200 mbar H in N for 1 h, and after
2
2
reoxidation in air at 773 K for 1 h.

18
J. Pérez-Ramírez et al. / Journal of Catalysis 223 (2004) 13–27
+
treatment (200 mbar H2 in N2 at 773 K for 1 h). This is ex-
pected for isomorphously substituted iron(III) ions, which
are well shielded in the zeolite framework, being resistant
against reduction. In the calcined sample, a small shoulder
at 300 nm was also observed, which indicates that a small
fraction of (slightly agglomerated) iron was not incorpo-
rated in the framework during the hydrothermal synthesis or,
most likely, was already extracted during calcination [20].
Contrarily, the same reductive treatment over steamed ex-
lated Fe3 ions in positions of high symmetry (zero-field
splitting parameters D, E ≈ 0) also contribute to isotropic
ꢀ
lines at g ∼2 [34,40]. Moreover, it has been demonstrated
that small changes of D and E in the range between zero
and resonance energy induce dramatic changes in the num-
ber and position of observable Fe signals [34,41].
3
+
ꢀ
The intensity of the line at g ∼2 in the EPR spectra of the
iron zeolites containing clusters (ex-FeZSM-5, Fe/ZSM-5,
and FeZSM-5) at room temperature and 77 K is very sim-
ilar (Fig. 4), and does not follow the Curie–Weiss law for
paramagnetic species. This suggests that the iron species re-
sponsible for this signal are coupled by antiferromagnetic
interactions within oxidic clusters, which reduce the num-
ber of unpaired spins contributing to the EPR signal. This
may explain the fact that the total EPR signal intensity of
the clustered zeolites is low as compared to ex-Fe-silicalite,
since a part of these Fe species is probably EPR silent, as
also observed by Kucherov and Shelef [42].
3
+
2+
Fe-silicalite shows that reduction of Fe to Fe caused
a marked decrease of the Fe3 CT band intensity, which
is completely reversible on reoxidation in air flow (Fig. 3).
This redox behavior can be attributed to isolated iron ions in
extraframework positions. Thus, the extraction of iron upon
steaming of Fe-silicalite at 873 K is extensive, leading to iso-
lated extraframework Fe(III) species that can be effectively
reduced by H2.
+
X-band EPR spectra of the catalysts measured at 298 K
and 77 K are shown in Fig. 4. All the samples show typi-
ꢀ
The g ∼2 signal in ex-Fe-silicalite increased upon cool-
ꢀ
cal signals at g values of 2, 4.3, and 6. Spectra simulation
ing from room temperature to 77 K, as expected for para-
ꢀ
ꢀ
3+
has shown that signals at g ∼4.3 and g ∼6 arise from Fe
3+
magnetic Fe species obeying the Curie–Weiss law. This
sites in strong rhombic or axial distortion when the zero-field
splitting is large in comparison to the microwave frequency
behavior, generally observed for isolated species, relates to
the high degree of dispersion of iron in this catalyst. How-
ever, in the EPR spectrum of ex-Fe-silicalite at 77 K, a
rather narrow line and a very broad line are superimposed
ꢀ
1
3
ꢀ
(
g ∼4.3; D ꢀ hν, E = D; g ∼6; D ꢀ hν, E = 0) [31].
The magnitude of the zero-field splitting is a measure of
the distortion of the Fe3 coordination symmetry, but it
+
ꢀ
in the range of g ∼2. The latter is not typical for ideally
does not provide information on the number of coordinating
isolated sites without any mutual magnetic interaction and
suggests the presence, to some extent, of a weak dipolar in-
3
+
ligands. Accordingly, distorted Fe sites in various coordi-
nations can contribute to these signals. In zeolites, the line
teraction between the Fe3 sites contributing to this signal.
+
ꢀ
3+
at g ∼4.3 has been equivocally assigned to Fe sites in
ꢀ
Therefore, the signal at g ∼2 in ex-Fe-silicalite can be ten-
tetrahedral coordination, either in framework [32–34] or ex-
tatively attributed to highly symmetric and rather isolated
ꢀ
traframework positions [35–37], while the line at g ∼6 is
3
+
Fe species, which partly undergo weak dipolar interac-
tions.
Further support for this assignment was gathered by mea-
suring the EPR spectrum of ex-Fe-silicalite after evacuation
frequently assigned to isolated Fe3 species with higher co-
+
ꢀ
ordination numbers [38,39]. The EPR signal at g ∼2 has
been typically assigned to iron oxide clusters. However, iso-
−2
at 10 Pa and ambient temperature for 2 h. This treatment
ꢀ
causes a dramatic decrease of the g ∼2 line, together with
the appearance of additional signals in the low-field range
(
Fig. 5). This can be attributed to the removal, upon evacua-
tion, of water molecules in the pore network of the zeolite
adsorbed on Fe3 species responsible of the EPR signal.
+
3
+
This leads to isolated Fe sites in a lower symmetry (higher
zero-field splitting) that give rise to additional signals in
the low-field range. These spectral changes are slowly re-
versible when exposing the sample to ambient conditions.
ꢀ
A very similar behavior for the g ∼2 signal upon evacuation
and rehydration at room temperature was also observed by
Berlier et al. [43] on iron silicalite (Si/Fe = 90) activated at
ꢀ
7
73–1073 K in vacuum. These authors attributed the g ∼2
3
+
signal to isolated Fe ions, which are solvated by water
molecules under ambient conditions. Following this assign-
ment, the result in Fig. 5 further corroborates that the Fe
species contributing to the g ∼2 line in ex-Fe-silicalite are
well isolated in extraframework positions, being accessible
to gaseous molecules.
3
+
ꢀ
Fig. 4. EPR spectra of FeMFI zeolites (without pretreatment) at room tem-
perature (solid line) and 77 K (dashed line).

J. Pérez-Ramírez et al. / Journal of Catalysis 223 (2004) 13–27
19
Fig. 5. EPR spectra of ex-Fe-silicalite recorded at 298 K before and after
evacuation for 2 h and after re-exposure to ambient atmosphere for 1.5 and
6 h.
1
3
.2. Catalytic activity
Fig. 6 shows the N2O conversion vs temperature re-
sulting from steady-state experiments over ex-FeZSM-5 in
direct N2O decomposition and N2O reduction with CO at
two different molar CO/N2O feed ratios. In the absence of
CO, the catalyst shows a significant N2O conversion above
7
00 K and a complete conversion at 850 K. Addition of 1.5
mbar CO (CO/N2O = 1) dramatically enhances the N2O
decomposition activity over the catalyst, achieving similar
conversions at 100 K lower temperature. The reduction of
N2O by CO also occurs at significantly lower temperatures
at a lower partial CO pressure (CO/N2O = 0.13). However,
when the limiting reactant CO becomes exhausted (∼20%
N2O conversion) the N2O conversion curve shifts to that
of the pure N2O decomposition. The conversion of CO to
CO2, depicted in Fig. 6, was complete in the range of 700–
Fig. 6. Conversions of N O and CO vs T over ex-FeZSM-5 in different feed
2
5
−1
mixtures at W/F (N O)o = 3×10 g s mol and P = 1 bar: (ꢁ) 1.5 mbar
2
N O in He; (ꢂ) 1.5 mbar N O and 0.2 mbar CO in He; (
•
) 1.5 mbar N O
2
2
2
and 1.5 mbar CO in He.
CO is initiated at significantly lower temperatures on the
catalyst prepared by sublimation (500 K) as compared to
ex-FeZSM-5 (>550 K). Complete N2O conversion was
achieved over Fe/ZSM-5 at 625 K using CO/N2O = 1,
while a temperature of 750 K was required over ex-FeZSM-
7
50 K and is effectively coupled with the conversion of N2O
to N2. In the experiment with CO/N2O = 1, the amount of
CO2 formed equals the amount of N2O decomposition (or
N2 formed) at all temperatures, with no formation of O2,
according to Eq. (5). At lower molar CO/N2O feed ratios
5
. The shape of the conversion profiles presents important
differences too. N2O conversion over Fe/ZSM-5 increases
rapidly in the temperature range of 500–600 K. The re-
ducing agent is consumed at 600 K at CO/N2O < 1 and
the N2O conversion remains practically unchanged until
(
CO/N2O = 0.13), oxygen is produced at high temperatures
when CO is consumed and a ratio of N2/O2 = 2 was ob-
tained, corresponding to the direct decomposition reaction
in Eq. (6).
6
75 K. The conversion profiles develop more gradually over
ex-FeZSM-5, not reaching a well-defined plateau at any tem-
perature. The apparent activation energy at CO/N2O = 1
is higher over Fe/ZSM-5 than over ex-FeZSM-5 (190 and
N2O + CO → N2 + CO2,
(5)
(6)
−1
1
28 kJ mol , respectively). These values were estimated
1
N2O → N2 + O2.
2
around the inflection (50% conversion) of the conversion
temperature curves by assuming a plug-flow model and a
first-order reaction in N2O.
The N2O conversion over the catalysts investigated in-
creases linearly with the molar CO/N2O ratio, as shown
in Fig. 8 at 600 K. This temperature was chosen to ensure
the presence of CO. At high temperatures CO becomes ex-
hausted over some catalysts, and such comparison would
not be representative. The different slope of the fitted lines
This general behavior is extrapolable to all the Fe-
zeolites investigated in this study. However, the performance
strongly depends on each particular sample, which is at-
tributed to the impact of the preparation method on the
catalyst constitution with respect to iron. As an exam-
ple, the N2O conversion vs temperature over ex-FeZSM-5
and Fe/ZSM-5 at four different molar CO/N2O feed ra-
tios is compared in Fig. 7. The reduction of N2O with

20
J. Pérez-Ramírez et al. / Journal of Catalysis 223 (2004) 13–27
can be assumed that they are responsible for the observed
performance. This is supported by comparison with the clus-
tered ex-FeZSM-5, prepared using identical methods and
with the same iron content, which shows a much lower activ-
ity. The high activity of Fe/ZSM-5 should not be connected
with the presence of iron clusters, since the relative fraction
of isolated iron species, taking into account the iron con-
tent and the deconvolution in Table 2, is larger than over
ex-Fe-silicalite in the test shown in Fig. 8. From our steady-
state data and characterization results, a correlation between
the fraction of isolated iron sites in the zeolites and the
catalytic activity can be qualitatively derived. This correla-
tion is further discussed in Section 4.2. In situ spectroscopic
investigations are necessary to confirm the essential role
of isolated species and to evaluate the eventual participa-
tion of other iron species in the catalytic reduction of N2O
by CO.
3
.3. In situ UV/vis and EPR spectroscopies
In situ UV/vis and EPR studies were aimed at the identi-
fication of active iron species in the reduction of N2O with
CO over FeMFI zeolites, based on the propensity of the iron
sites to undergo redox processes. To this end, the interac-
tion of CO, N2O, and CO + N2O mixtures with the iron
species in the catalysts has been investigated at 623 K. At
this temperature, all catalysts have shown substantial activ-
ity in steady-state experiments (Section 3.2).
Fig. 7. N O conversion vs T over ex-FeZSM-5 and Fe/ZSM-5 at different
2
5
−1
molar CO/N O feed ratios (in figure) at W/F (N O)o = 9×10 g s mol
2
2
and P = 1 bar. Partial N O pressure was fixed at 1.5 mbar and partial CO
2
pressure was varied from 0 to 1.5 mbar; balance He.
3
.3.1. Steam-activated ex-Fe-silicalite
Fig. 9 shows that Fe3 species in ex-Fe-silicalite are not
+
sensitive to reduction by CO at this temperature, as con-
cluded from the comparison of the UV/vis spectra of the
sample in air and in a CO-containing mixture. The EPR
spectrum of ex-Fe-silicalite changes significantly upon in-
ꢀ
teraction with CO (Fig. 10). The signals at g ∼7.15, 6.05,
and 2, representative for differently distorted isolated sites,
decrease in the presence of CO and a new broad signal at
ꢀ
ꢀ
g ∼3.01 appears. The signal at g ∼4.3 in Fig. 4 is not ob-
served at 623 K, probably due to short relaxation times at
high temperatures. Considering the UV/vis results, the de-
ꢀ
crease of the EPR lines at g of 7.15, 6.05, and 2 should not
be due to reduction of Fe3 . It is plausible to state that CO is
+
3
+
chemisorbed on extraframework Fe isolated ions, causing
changes of the local symmetry and thus altering the posi-
tion of the EPR signals. A similar shift of the Fe signal
Fig. 8. N O conversion at 600 K vs molar CO/N O feed ratio over FeMFI
2
2
5
−1
catalysts at W/F (N O)o = 9 × 10 g s mol and P = 1 bar. Feed compo-
3+
2
sition as in caption of Fig. 7.
upon changing ligands was observed for FeZSM-5 zeolites
prepared by impregnation of HZSM-5 with a FeCl3 solu-
ꢀ
suggests a different kinetic behavior. Fe/ZSM-5 shows the
highest N2O conversions, followed by ex-Fe-silicalite. The
activity of ex-FeZSM-5 and FeZSM-5 is significantly lower.
The iron content in Fe/ZSM-5 is ∼7 times higher than in
ex-Fe-silicalite. Accordingly, the later catalyst is clearly su-
perior if the activity is expressed per mole of iron in the
sample (about 3 times), indicating a better iron utilization.
Since ex-Fe-silicalite mainly contains isolated iron ions, it
tion [42]. In the as-prepared zeolites, a line at g ∼3.65
+
was assigned to isolated (FeCl ) species. Upon calcina-
tion in air, this line disappeared due to the transformation of
(FeCl ) into differently distorted FeO species, giving rise
2
+
+
2
ꢀ
to EPR signals at g ∼6.5, 5.6, and 4.27. For ex-Fe-silicalite,
the EPR changes are reversible upon switching from CO to
N O (compare Figs. 10a–c). The intensity of the band at
240 nm in the UV/vis spectrum of ex-Fe-silicalite is slightly
2

J. Pérez-Ramírez et al. / Journal of Catalysis 223 (2004) 13–27
21
Fig. 9. In situ UV/vis-DRS spectra of FeMFI zeolites at 623 K in different gas mixtures: air, 20 mbar N O in He, 20 mbar CO in He, and back to 20 mbar
2
N O in He. Spectra were measured after treatment of the samples in the respective mixtures at 623 K for 1 h. Catalyst pretreatment in air at 773 K for 1 h.
2
Fig. 10. In situ EPR spectra of FeMFI zeolites at 623 K in different gas mixtures. Conditions as in caption of Fig. 9.
higher after contact with N2O, as compared to the initial
spectrum of the sample in air (Fig. 9). This suggests the pres-
3.3.2. Steam-activated ex-FeZSM-5
The interaction of CO and N2O over ex-FeZSM-5 leads
to important differences compared to ex-Fe-silicalite, which
are associated with the distinct iron constitution. The UV/vis
ence of small amounts of Fe2 ions in the non-pretreated
+
zeolite, which can be oxidized by N2O but not by O2.

22
J. Pérez-Ramírez et al. / Journal of Catalysis 223 (2004) 13–27
spectrum of ex-FeZSM-5 did not change upon switching
from air to N2O/He, but treatment in CO/He decreases the
intensity of the band above 300 nm (Fig. 9), assigned to
oligonuclear clusters. Similar to ex-Fe-silicalite, the band
3.3.3. Sublimed Fe/ZSM-5
As shown in Fig. 9, the UV/vis spectra of Fe/ZSM-5 in
air and N2O were identical. Upon CO treatment, the inten-
sity of the whole spectrum was substantially reduced and
completely recovered by switching back to N2O. The in-
tensity reduction in CO was expected for the bands above
300 nm, arising from clustered iron species. However, the
decrease of the band around 250 nm (isolated Fe3 species)
was not anticipated, since these species were not reduced by
CO in ex-Fe-silicalite and ex-FeZSM-5. It should be noted
at 250 nm, characteristic of isolated Fe3 species, remains
+
3
+
unaltered. This result suggests a different behavior of Fe
species upon internation with carbon monoxide: CO reduces
+
3
+
oligonuclear Fe Oy clusters at typical reaction tempera-
x
3
+
tures, while isolated Fe ions are more stable against CO
reduction. Switching back from CO to N2O restores the band
above 300 nm in ex-FeZSM-5 and surprisingly, enhances the
band at 250 nm above the maximum reached during the first
N2O treatment. This result can be tentatively explained by
that the tail of the CT subbands arising from Fe3 Oy clus-
+
x
ters in Fe/ZSM-5 extends into the low-wavelength range (see
Fig. 2), contributing to the overall intensity below 300 nm.
Thus, the marked reduction of the band above 300 nm influ-
ences the intensity of the signal at lower wavenumbers.
The EPR spectra of Fe/ZSM-5 show a slight decrease of
assuming that the coordination of some of the Fe3+ species
reoxidized by N2O is different from their initial state before
reduction with CO. Phenomena like the partial dissolution
of larger clusters and/or symmetry changes might lead to
an increased absorbance at 250 nm. However, in any case,
the changes are very small and the amount of Fe species in-
volved in these processes is not significant.
ꢀ
the signal at g ∼2 upon switching from air to N2O, while
3
+
the signals of isolated Fe sites at low field remain prac-
tically unchanged (Fig. 10). Subsequent admission of CO
ꢀ
leads to a very large broad singlet at g ∼2.07 in the sample,
which is typical of ferrimagnetic Fe3O4-like species, and the
In the corresponding EPR spectra of ex-FeZSM-5
low-field signals for isolated Fe3 species disappear. This
agrees well with previous observations by Kucherov and
Shelef [42], who found that even a mild reductive treatment
of Fe/HZSM-5 (1 vol% H2 in He at 573 K) leads to (par-
+
(
Fig. 10), switching from air to N2O causes a slight in-
ꢀ
crease of the signals at g ∼6.42, 5.62, and 2.0. This can be
associated with the change from a mixture containing a para-
magnetic gas (O2) to a diamagnetic one (N2O), which causes
tial) irreversible loss of isolated Fe3 sites accompanied by
the formation of ferrimagnetic Fe3O4 clusters. As shown in
Fig. 10, switching back from CO to N2O narrows the line as-
+
3
+
line harrowing since the magnetic interaction between Fe
ꢀ
and O2 is suppressed. Interestingly, the signal at g ∼4.32
is almost not influenced by this change, suggesting that the
sociated with such species. Since the reduced Fe2 sites are
totally reoxidized by N2O, as concluded from UV/vis, the
EPR observation should be related to changes of the intrin-
sic magnetic interactions within the ferrimagnetic clusters.
The spectra c and d of Fe/ZSM-5 in Fig. 10 were reproduced
upon various periodical treatments in N2O/He and CO/He
+
3
+
Fe species contributing to this line are not accessible by
gas-phase components. Alternatively, it cannot be excluded
that traces of iron(II) sites remain in such state after pretreat-
ment in air, but can be oxidized by N2O.
ꢀ
The signals at g ∼6.42, 5.62, and 2.0 in ex-FeZSM-5
decrease upon switching from N2O to CO (Fig. 10). Based
on the various characterizations in Section 2.1 (HRTEM,
UV/vis, and the temperature dependence in the EPR spec-
ꢀ
mixtures. In N2O, a very narrow signal at g ∼2.003 is ob-
served too. Such signal is typical for radical species rather
3+
than for Fe ions. Panov [3] attributed the unique perfor-
ꢀ
tra), the signal at g ∼2 in this sample is assigned to oxidic
mance of iron-containing ZSM-5 in the N2O-mediated oxi-
dation of benzene to phenol to the formation of α-oxygen,
whose charge has been a matter of discussion. The abil-
iron clusters. Accordingly, its decrease upon CO treatment
is associated to the reduction of clustered Fe3 species. As
+
observed in ex-Fe-silicalite, the decrease of the EPR signals
−
ity of N2O to deposit paramagnetic O species on partially
ꢀ
at g ∼6.43 and 5.62 in ex-FeZSM-5 corresponds to iso-
reduced oxide surfaces is well known [44]. Such species
originate narrow EPR signals near the free-electron g value,
having an anisotropic g tensor with g⊥ > gꢁ [44,45]. How-
ever, when they become mobile, e.g., at high temperatures, g
anisotropy averages out [45]. To the best of our knowledge,
neutral oxygen atoms have never been detected by EPR,
due to the short lifetime imposed by the their high reactiv-
ity [46]. In a recent work, Starokon et al. [47] concluded that
3
+
lated Fe species, which suffer a change of symmetry upon
coordination by CO, but it is not accompanied by their re-
duction, as concluded from in situ UV/vis. Different than
ꢀ
ex-Fe-silicalite, no pronounced signal at g ∼3 was observed
in ex-FeZSM-5 upon interaction with CO. Furthermore, EPR
signals in the low-field range appear at slightly different
g-values in the steam-activated Fe-zeolites, suggesting that
the distortion of the local symmetry of the isolated sites is
unequal in the two catalysts. Coordination of CO in such
sites can lead to complexes of slightly different geometry in
both samples, causing a shift of the EPR signal. This could
−
α-oxygen is an anion-radical species, designated as Oα . It
has also been shown that O species, when engaged in the
−
micropores of CaO/Al O crystals with a zeolite-like struc-
2
3
ture, can be rather stable [45]. Based just on the g value, the
ꢀ
also explain the absence of the g ∼3 signal in ex-FeZSM-
narrow isotropic line in Fig. 10 cannot be conclusively as-
ꢀ
−
5
. Finally, the EPR signals at g ∼6.42, 5.62, and 2 were
signed to O species, since different radical species lead to
restored upon switching from CO to N2O (Fig. 10).
similar signals in this region. However, considering that the

J. Pérez-Ramírez et al. / Journal of Catalysis 223 (2004) 13–27
23
signal appears only in the presence of N2O and/or after par-
tial reduction of the sample with CO, i.e., under favorable
−
conditions for formation of O species, it does not seem un-
likely that the signal arises from these radical species, being
mobile within the zeolite pore network.
3
.3.4. Ion-exchanged Fe-ZSM-5
The in situ UV/vis spectra of Fe-ZSM-5 in the low wave-
length range are similar to that of Fe/ZSM-5 (Fig. 9). The
most significant difference was observed in the visible range
upon interaction with CO, leading to an increase of light ab-
sorption above 450 nm in the ion-exchanged catalyst. This
observation is associated to the lower wavelength tail of an
intervalence charge transfer (IVCT) transition, which typi-
cally occurs in the near-IR region and is induced by electron
delocalization between Fe2 and Fe ions. Such a phenom-
enon is characteristic for mixed-valence iron oxides and has
been observed in Fe3O4 nanoparticles [48], where the tail
of the ICVT transition extends to the visible range and con-
tributes to light absorption above 450 nm. Thus, it is very
likely that a similar type of mixed-valence iron oxide par-
ticles is formed upon interaction of CO with Fe-ZSM-5.
This was not observed over Fe/ZSM-5, with a negligible
increase of intensity above 600 nm (Fig. 9). This could
be attributed to the significantly larger Fe2O3 particles in
Fe-ZSM-5, which can more easily undergo reduction to a
defined Fe3O4-like phase. Furthermore, the mutual distribu-
+
3+
Fig. 11. In situ EPR spectra of Fe/ZSM-5 in different gas mixtures: air at
7
73 K, 10 mbar N O + 10 mbar CO in He (CO/N O = 1) at 623 K, and
2
2
6.6 mbar N O + 13.4 mbar CO in He (CO/N O = 2) at T = 623 K and
2
2
P = 1 bar. The N O conversion obtained is shown in the figure.
2
The N2O conversion in the experiment was 82%, in good
agreement with the steady-state tests in Fig. 7. This result
indicates that in the presence of equimolar amounts of N2O
and CO, the average valence of active Fe species is +3 under
steady-state conditions, with no formation of ferrimagnetic
Fe3O4-like particles. This occurs when CO is added in ex-
cess (CO/N2O = 2), while the obtained N2O conversion is
slightly increased to 87%.
tion of Fe3 and Fe species in Fe/ZSM-5 and Fe-ZSM-5
may differ depending on the exact position, intensity, and/or
line width of the IVCT transition.
+
2+
In the EPR spectra of Fe-ZSM-5, signals for isolated
Fe3 species are hardly visible, in agreement with the high
+
4. Discussion
degree of clustering evidenced by this sample. Large iron
ꢀ
oxide particles give rise to the broad EPR signal at g ∼2
4.1. Nature and distribution of iron species
which does not change upon switching from air to N2O. In
the presence of CO, the broad line accounting for ferrimag-
netic Fe3O4-like particles develops. A similar EPR signal
was observed in Fe/ZSM-5 (Fig. 10). However, in the latter
sample, the intensity of this line is five times higher than in
Fe-ZSM-5, although the total Fe content is only 3.5 times
higher (see Table 1). This is likely due to differences in the
microstructural domain of the ferrimagnetic particles and not
In this study, FeMFI catalysts were prepared following
two general preparation strategies to disperse iron species
inside the zeolite channels and cavities: (i) postsynthesis in-
sertion via ion exchange in solution and in the gas phase
and (ii) migration of iron to extraframework positions start-
3
+
ing from MFI containing isomorphously substituted Fe
.
Postsynthesis methods often lead to intrinsic difficulty of the
achievement of a proper cation exchange and highly hetero-
geneous systems are obtained. The extraction of iron from
FeMFI samples exhibiting a unique tetrahedral and isolated
iron species is claimed to represent a simpler and more re-
producible way to disperse iron species into microporous
matrices, although not allowing the insertion of a high per-
centage of iron [49].
The characterization from the different methods applied
(HRTEM, UV/vis, and EPR) to assess the form(s) of iron in
the catalysts is in very good agreement. Our results clearly
show that the distribution of iron species, as isolated ions,
oligonuclear iron species, and iron oxide particles is a func-
tion of the sample genesis. A common denominator of the
Al-containing Fe-zeolites is the heterogeneous constitution
3
+
to a higher degree of Fe reduction by CO in Fe/ZSM-5,
since UV/vis clearly shows a more pronounced reduction
in Fe-ZSM-5 (Fig. 9). Upon contact with N2O, a narrow
ꢀ
EPR signal around g ∼2.003 arises in Fe-ZSM-5. This
−
signal, attributed to O species, is 17 times larger than in
Fe/ZSM-5, despite the lower iron content in the liquid-ion
exchanged sample. These various observations further evi-
dence the markedly different nature of this Fe-oxide phase
in both catalysts.
Finally, an in situ EPR experiment was performed by ex-
posing Fe/ZSM-5 to different N2O–CO mixtures at 623 K
(
Fig. 11). A change from air to a reaction mixture containing
ꢀ
CO/N2O = 1 causes a slight decrease of the signals at g ∼6
ꢀ
ꢀ
and g ∼2, while the line at g ∼4.3 remains unchanged.

24
J. Pérez-Ramírez et al. / Journal of Catalysis 223 (2004) 13–27
with respect to iron and the extensive degree of clustering.
The formation of iron oxide particles at the external surface
of the zeolite crystals and their size are more pronounced in
the samples prepared by postsynthesis ion exchange meth-
ods, especially in the liquid phase. The large particles in
Fe-ZSM-5 (up to 30 nm) originate from two processes: pre-
cipitation and clustering. In the method applied here, the
ion-exchange procedure was performed without pH control,
in order to further promote formation of iron oxide parti-
cles in the final catalyst. The poor catalytic performance and
low redox activity of this sample is in line with the ma-
jor occurrence of iron in such an inactive form. The liquid
ion exchange applied here, to a large extent, mimics the re-
sult obtained with a wet impregnation method. Although
an improved control of the preparation conditions during
liquid ion exchange can lead to a more homogeneous iron
distribution in the zeolite, a lack of reproducibility of ex-
change methods in solution has been frequently reported
which indicates a higher iron dispersion. However, this is
also associated to the lower iron content in the sample (e.g.,
∼7 times lower than in the catalyst prepared by sublima-
tion).
A better-defined distribution of iron species was attained
upon steam treatment of Fe-silicalite. The color of this sam-
ple was nearly white, suggesting the more isolated nature of
the iron species in the catalyst, while ex-FeZSM-5 was light
brownish. Indeed, the majority of iron in ex-Fe-silicalite was
found as isolated Fe(III) ions in extraframework positions.
A very minor degree of iron aggregation is present, as con-
cluded from UV/vis analysis. EPR spectroscopy evidences
the paramagnetic behavior of Fe3 species following the
Curie–Weiss law, as typically observed for highly symmetric
isolated species. However, a certain degree of weak dipolar
+
interactions between Fe3 sites is also identified, indicative
of a certain iron association. The amount of iron in the form
of aggregates is minimal and of very small size, since they
were not detected by HRTEM. Accordingly, the framework
composition of the zeolite plays an essential role in deter-
mining the iron speciation. In particular, the presence of
Al in the MFI structure destabilizes framework iron, pro-
moting extraction and clustering of extraframework species
[19,21,30]. Berlier et al. [56] recently concluded, based on
infrared studies, that Al favors the dispersion of extraframe-
work iron species in isomorphously substituted FeZSM-5
as compared to Fe-silicalite. In that study, both zeolites
were pretreated in vacuum at the same temperature (773 or
973 K). It is well known that the higher stability of iron in
the framework of Al-free ferrisilicate requires more severe
conditions for extensive extraction of framework iron [21],
and thus the term “dispersion” in [56] can be conceptually
understood by “degree of iron extraction,” which is obvi-
ously associated to reactivity. The results presented here and
elsewhere [21], using steam as activation atmosphere (more
effective for iron dislodgment that vacuum treatment), show
that the higher stability of iron in the framework of sili-
calite enables a better control of iron extraction upon steam
activation, since clustering can be largely prevented. Still,
by a proper selection of the steaming temperature (873 K),
a substantial degree of Fe extraction can be accomplished,
although not being complete. This has been demonstrated
by the different reduction characteristics of framework and
extraframework isolated iron species during in situ UV/vis
in H2 at 773 K, and previously by voltammetric response
studies [21]. Furthermore, the extensive iron dislodgment
is expected in view of the high activity of ex-Fe-silicalite
in the reduction of N2O with CO (Section 4.2). Framework
iron ions are catalytically inactive, since activation of N2O
requires coordination of atomic oxygen to the active site, im-
possible for framework iron being saturated by surrounding
oxygen atoms of the lattice [57,58].
+
[
16,50–53]. The use of reducing agents and complexing
ligands was applied to prevent the precipitation of iron hy-
droxides and the formation of Brønsted acid sites during
synthesis of overexchanged FeZSM-5 (Fe/Al > 0.5), but
these attempts proved unsuccessful [50].
Ion exchange in the gas phase (sublimation) is undoubt-
edly a more effective postsynthesis method, since iron clus-
tering is largely suppressed compared to methods in the
liquid phase. Sublimation of FeCl3 in the channels of H-
ZSM-5 enables a high iron loading in the catalyst (Fe/
Al = 1). A complete insertion of iron at ion-exchanged
positions has been claimed, forming diferric (hydr)oxo-
bridge binuclear clusters [22,54,55]. The UV/vis spectrum
of Fe/ZSM-5 indeed shows the largest contribution of Fe(III)
species in the region of the small oligonuclear iron species in
the zeolite channels. However, a uniform distribution of iron
species was certainly not achieved in our Fe/ZSM-5, pre-
pared according to the method in [22]. Isolated iron ions and
large iron oxide particles (up to 15 nm) are also identified in
this sample. The hydrolysis and calcination posttreatments
following the sublimation of FeCl3 have proven to be key
steps in order to prevent formation of large hematite parti-
cles [54,55].
Steam activated ex-FeZSM-5 and ex-Fe-silicalite are
highly diluted samples, with Si/Fe ∼150. The migration
of iron from framework to partial or total extraframework
positions upon steaming is complex and can be envisaged
as a clustering process [21]. An appropriate comparison of
the iron constitution in these two samples can be estab-
lished, since the preparation and activation procedures as
well as the iron content in both catalysts were identical.
The results obtained over ex-FeZSM-5 show that a variety
of iron species originated upon steaming, and even the for-
mation of small iron oxide nanoparticles (1–2 nm). Hence a
remarkable uniformity of iron species and the absence of ex-
tensive clustering are a priori not guaranteed by this method.
Of course, the size of the particles is significantly smaller
than in the iron-zeolites prepared by postsynthesis methods,
A slight increase of the Fe3 signal appears in the in
+
situ UV/vis spectra of the Al-containing zeolites during O2
2
+
treatment at 600 K, indicating a certain amount of Fe
in the fresh catalyst, while hardly any oxidation of ferrous

J. Pérez-Ramírez et al. / Journal of Catalysis 223 (2004) 13–27
25
ions was observed in ex-Fe-silicalite upon oxidative treat-
ment in O2 or N2O. Recent XANES work has shown the
catalytic performance of ion-exchanged Fe-ZSM-5 can be
expected in view of the extensive iron clustering in the form
of large inactive Fe-oxide particles. Iron clustering also oc-
curs in the catalyst prepared by sublimation (Fe/ZSM-5), but
the high iron content together with a relatively high frac-
tion of iron in intrazeolitic positions makes it the most active
formulation in terms of absolute N2O conversions. The ac-
tivity of ex-FeZSM-5 is largely reduced compared to ex-Fe-
silicalite, which correlates with the decreased concentration
of isolated iron species due to clustering. However, ex-
FeZSM-5 was a more effective catalyst than ex-Fe-silicalite
in direct N2O decomposition [19,21]. This has been at-
tributed to the intrinsic mechanism of the N2O conversion
investigated, which determines its optimal iron architecture
[19,63]. Based on our activity and characterization results
with ex-Fe-silicalite and the derived correlation in Fig. 12,
isolated iron ions are very active toward N2O activation
[Eq. (1)] and the removal of the generated atomic oxygen
reduction of Fe3 to Fe in Fe-silicalite (in the form of
isolated, clustered, or grafted species) upon dislodgment to
extraframework positions by thermal treatment in vacuum
+
2+
2
+
or in steam [43,49,59]. These extraframework Fe species
have been characterized by a very high coordinative un-
saturation, being able to form Fe2 (NO)3 complexes upon
+
interaction with NO at room temperature [60]. According
to Dubkov et al. [61], the reduction of Fe3 to Fe upon
steam treatment of FeZSM-5 followed by Mössbauer spec-
troscopy may comprise more than 60% of the total metal
+
2+
content. These Fe2 species, which can be oxidized by N2O
but not by O2, have been associated with binuclear iron com-
plexes. Our characterization results with ex-Fe-silicalite in
Sections 3.1 and 3.3 clearly show the practically sole pres-
+
ence of Fe3 , mainly in extraframework isolated positions.
FTIR studies of NO adsorbed over ex-FeZSM-5 and ex-
Fe-silicalite at 423–573 K have indeed shown absorption
+
*
(O ) can be efficiently done in the presence of CO. However,
bands of Fe2 –NO, but characteristic bands indicative of
more than one nitro group on such iron sites were not iden-
tified [62].
+
in direct N2O decomposition, the isolated nature of iron in
*
ex-Fe-silicalite makes difficult recombination of O accord-
ing to Eq. (3) (rate-determining step), even though the rate of
Eq. (1) is higher than over ex-FeZSM-5. Finally, it should be
noted that the correlation in Fig. 12 assumes that the totality
of isolated sites in the steam-activated samples are in ex-
traframework positions. This is the case for ex-FeZSM-5, as
concluded in [20,21]. For ex-Fe-silicalite, even if our charac-
terization and activity data point to an extensive iron extrac-
tion, a certain (small) fraction of redox-inactive framework
iron cannot be excluded. This fraction was not quantified
here, since framework and extraframework species cannot
be discriminated in UV/vis spectroscopy. In rigorous terms,
the fraction of remaining framework iron in ex-Fe-silicalite
should be deducted from the total fraction of isolated sites in
Fig. 12, since Fe(III) in tetrahedral framework positions do
not contribute to the catalytic activity [58].
4
.2. Activity and mechanism of N2O reduction with CO
In direct N2O decomposition, O2 desorption is a rate-
determining step, which can be accelerated at a high tem-
perature, in the presence of NO [8], and also by addition
of reductants. CO efficiently removes atomic oxygen from
the catalyst surface and leads to a substantially decreased
operation temperature with respect to the direct N2O de-
composition. The reaction of N2O with CO is stoichiometric
[
Eq. (5)] and the N2O conversion increases linearly with the
molar feed CO/N2O ratio (Fig. 8). The method of prepa-
ration and activation of FeMFI zeolites determines the iron
constitution, which strongly influences the catalytic perfor-
mance. Since CO is an efficient oxygen remover, the overall
rate of N2O conversion by SCR can approach that of N2O
activation [Eq. (1)]. This can be used to probe the reactivity
of atomic oxygen species (resulting from N2O activation)
adsorbed on the various iron sites identified in Fe-zeolites.
The heterogeneous nature of iron in the samples often
complicates interpretations concerning the nature of ac-
tive species and their relative contribution in the overall
reaction. The remarkably uniform distribution of iron in ex-
Fe-silicalite as isolated Fe ions connected to its high catalytic
activity is a strong indication of the importance of mononu-
clear iron ions in the reduction of N2O with CO. Fig. 12
shows the derived correlation between the total percentage
of isolated Fe sites estimated from the relative intensity of
the UV/vis subbands below 280 nm (Table 2) and the cat-
alytic activity for N2O conversion at 600 K (Fig. 8). The
The above reasoning does not exclude the participation
of other iron species in the reaction. In situ spectroscopies
3
+
N2O conversion and the percentage of isolated Fe sites
between the catalysts (relative to ex-Fe-silicalite) are in the
same order of magnitude and follow the sequence: Fe/ZSM-
Fig. 12. Correlation between the relative N O conversion at 600 K (from
2
3+
Fig. 8) and the relative fraction of isolated Fe sites (from Table 2) in the
Fe-zeolites investigated. Values are referred to ex-Fe-silicalite.
5
> ex-Fe-silicalite > ex-FeZSM-5 ∼ Fe-ZSM-5. The poor

26
J. Pérez-Ramírez et al. / Journal of Catalysis 223 (2004) 13–27
revealed important differences in the redox behavior of the
iron species in the catalysts upon interaction with O2, N2O,
a complex oxo structure in cationic sites, as similarly found
ꢀ
by us. In our spectra, the signal at g ∼4.3 shows hardly any
3
+
and CO, suggesting that both isolated Fe sites as well as
changes during treatment with N2O or CO, and hence is not
considered as decisive for the catalytic performance. It can
be concluded that the isolated iron ions associated with this
signal are not accessible by reactants and/or have no vacant
coordination sites.
oligonuclear Fe3 Oy species are active in the reaction of
+
x
N2O with CO at 623 K. The reaction mechanism associated
to these species differs substantially. In situ UV/vis analy-
ses have clearly shown that Fe(III) in oligonuclear Fe3 Oy
+
x
clusters can be easily reduced by CO to the divalent state and
can be completely reoxidized by N2O (Fig. 9). In combina-
tion with EPR results, it can be proposed that the reduction
of N2O with CO over these sites occurs according to Eqs. (7)
5. Conclusions
−
and (8), i.e., via intermediate formation of O radicals. This
The reduction of N2O with CO has been investigated over
differently prepared Fe-zeolite catalysts. The preparation
method determines the nature and distribution of iron species
in the final zeolites. Results from the applied techniques to
characterize the iron form(s), including HRTEM, UV/vis,
and EPR spectroscopy, are in excellent agreement. A com-
mon denominator of the Al-containing Fe-zeolites is the
heterogeneous constitution with respect to iron and the ex-
tensive degree of iron clustering, in the form of oligonuclear
iron clusters and large iron oxide particles. Iron association
was largely prevented in steam-activated ex-Fe-silicalite,
which mainly contains isolated iron ions in extraframe-
work positions. CO enhances the N2O decomposition rate
over all the catalysts to lower temperatures with respect
to direct N2O decomposition, and the catalytic activity in-
creases linearly with the molar CO/N2O feed ratio. The
catalytic performance in the N2O reduction with CO is
strongly influenced by the iron constitution in the zeolite.
The high specific activity of ex-Fe-silicalite, with a remark-
able uniform distribution of isolated iron ions, is essential to
conclude the importance of mononuclear iron ions in the re-
duction of N2O by CO. The highly clustered ion-exchanged
FeZSM-5 catalyst shows poor activities in the reaction. Ac-
cordingly, a correlation between the N2O conversion and the
fraction of isolated Fe(III) sites in the catalysts has been
obtained. In situ UV/vis and EPR studies evidenced the par-
ticipation of mononuclear iron ions in the N2O–CO reaction,
mechanism agrees with the classical hypothesis stating that
CO purely acts as an O scavenger leading to CO2 [Eq. (7)]
and regenerating the active site (ꢁ) for subsequent N2O ac-
tivation [12–14,19].
Fe3 −O−Fe + CO → Fe −ꢁ−Fe + CO2,
+
3+
2+
2+
(7)
(8)
Fe2 −ꢁ−Fe + N2O → Fe −O −Fe + N2.
+
2+
3+
−
2+
Differently, the (dominant) isolated Fe3 ions in ex-
traframework positions of ex-Fe-silicalite suffer no reduc-
tion by CO at 623 K (Fig. 9). The relatively weak interaction
between CO and iron ions is commonly accepted [43]. How-
ever, the changes of the local symmetry and altered position
of the signals in the EPR spectra clearly indicate that CO
strongly coordinates to ferric ions in ex-Fe-silicalite. Attend-
ing to these results, the mechanism represented by Eqs. (9)
and (10) can be proposed for the N2O reduction with CO
over isolated iron sites.
+
Fe3 −ꢁ + CO → Fe −CO,
+
3+
(9)
Fe3 −CO + N2O → Fe3+−ꢁ + N2 + CO2.
+
(10)
This mechanism implies that the involved isolated Fe
species must be coordinatively unsaturated in order to
chemisorb CO. This holds for species associated with the
ꢀ
ꢀ
g ∼2 as well as the g > 5 signals in ex-Fe-silicalite (Fig. 10).
The observations related to the latter group of signals are in
excellent agreement with various EPR studies, where the
3
+
and also support the involvement of oligonuclear Fe Oy
x
ꢀ
reactivity of the EPR signals around g ∼6 is elaborated.
species. The interaction of N2O and CO and the reaction
mechanism is iron site dependent. Over isolated sites, the re-
duction of N2O with CO occurs via coordinated CO species
Kucherov et al. [64] determined by in situ EPR that the sig-
ꢀ
nals at g ∼5.6 and 6.5 observed in Fe/ZSM-5 catalysts
+
prepared by sublimation disappear upon contact with NO
or NO2. The corresponding signals were assigned to very
on Fe3 ions, not involving change of oxidation site. The
reaction over oligonuclear iron clusters proceeds via a re-
reactive coordinatively unsaturated Fe3 species. Volodin
et al. [65] also attributed the disappearance of this line (in
vacuum and in the presence of water and NO) as an indi-
cation for the reactivity of an ion-exchanged FeZSM-5 with
+
dox Fe /Fe process via intermediate formation of O
3+
2+
−
radicals. These results confirm the relationship between the
preferred structure of the active iron site and the associ-
ated reaction mechanism and have important implications in
oxidation and reduction conversions involving N2O over Fe-
zeolites.
ꢀ
N2O. Ribera et al. [66] found that the signals at g ∼6.4
and 5.7 in steam-activated FeZSM-5 disappear upon addi-
tion of β-mercaptoethanol, concluding their participation in
redox processes. In a recent study, Kubánek et al. [67] corre-
ꢀ
lated the intensity of the signals at g ∼6.0 and 5.6 and the
Acknowledgment
activity of HZSM-5 zeolites with low Fe concentrations in
the oxidation of benzene to phenol with N2O. These signals
were assigned to distorted T d-coordinated isolated ions with
Dr. P.J. Kooyman is acknowledged for performing the
TEM analysis.

J. Pérez-Ramírez et al. / Journal of Catalysis 223 (2004) 13–27
27
References
[36] A.V. Kucherov, A.A. Slinkin, Zeolites 8 (1988) 110.
37] B. Wichterlová, P. Jiru, React. Kinet. Catal. Lett. 13 (1980) 197.
[
[
1] F. Kapteijn, J. Rodríguez-Mirasol, J.A. Moulijn, Appl. Catal. B 9
1996) 25.
[38] A.F. Ojo, J. Dwyer, R.V. Parish, Stud. Surf. Sci. Catal. 49 (1989) 227.
[39] P. Wenquin, Q. Shilun, K. Zhiyun, P. Shaoyi, Stud. Surf. Sci. Catal. 49
(1989) 281.
[40] G. Catana, J. Pelgrims, R.A. Schoonheydt, Zeolites 15 (1995) 475.
[41] A. Brückner, in: B.M. Weckhuysen, P. Van Der Voort, G. Catana
(Eds.), Spectroscopy of Transition Metal Ions on Surfaces, Leuven
Univ. Press, Leuven, 2000, p. 69.
[42] A.V. Kucherov, M. Shelef, J. Catal. 195 (2000) 106.
[43] G. Berlier, G. Spoto, S. Bordiga, G. Ricchiardi, P. Fisicaro, A.
Zecchina, I. Rosetti, E. Selli, L. Forni, E. Giamello, C. Lamberti,
J. Catal. 208 (2002) 64.
[44] V.A. Shvets, V.B. Kazanski, J. Catal. 25 (1972) 123.
[45] K. Hayashi, M. Hirano, S. Matsuishi, H. Hosono, J. Am. Chem.
Soc. 124 (2002) 738.
(
[2] J. Pérez-Ramírez, F. Kapteijn, K. Schöffel, J.A. Moulijn, Appl. Catal.
B 44 (2003) 117.
[
[
[
[
[
3] G.I. Panov, CATTECH 4 (2000) 18.
4] C.M. Fu, V.N. Korchak, W.K. Hall, J. Catal. 68 (1981) 166.
5] J. Leglise, J.O. Petunchi, W.K. Hall, J. Catal. 86 (1984) 392.
6] J. Pérez-Ramírez, E.V. Kondratenko, Chem. Commun. (2003) 2152.
7] G. Mul, J. Pérez-Ramírez, F. Kapteijn, J.A. Moulijn, Catal. Lett. 7
(
2001) 77.
8] J. Pérez-Ramírez, F. Kapteijn, G. Mul, J.A. Moulijn, J. Catal. 208
2002) 211.
9] T. Nobukawa, S. Tanaka, S. Ito, K. Tomishige, S. Kameoka, K. Kuni-
mori, Catal. Lett. 83 (2002) 5.
[
[
(
[
[
[
[
10] R.W. van den Brink, S. Booneveld, J.R. Pels, D.F. Bakker, M.J.F.M.
[46] M. Che, A.J. Tench, Adv. Catal. 31 (1982) 77.
[47] E.V. Starokon, K.A. Dubkov, L.V. Pirutko, G.I. Panov, Top. Catal. 23
Verhaak, Appl. Catal. B 32 (2001) 73.
11] G. Delahay, M. Mauvezin, A. Guzmán-Vargas, B. Coq, Catal. Com-
mun. 3 (2002) 385.
12] F. Kapteijn, G. Mul, G. Marbán, J. Rodríguez-Mirasol, J.A. Moulijn,
Stud. Surf. Sci. Catal. 101 (1996) 641.
13] G.I. Panov, V.I. Sobolev, A.S. Kharitonov, J. Mol. Catal. A: Chem. 61
(
2003) 137.
48] J. Tang, M. Myers, K.A. Bosnick, L.E. Brus, J. Phys. Chem. B 107
2003) 7501.
49] A. Zecchina, S. Bordiga, G. Spoto, A. Damin, G. Berlier, F. Bonino,
C. Prestipino, C. Lamberti, Top. Catal. 21 (2002) 67.
[
[
[
[
(
(1990) 85.
50] P. Marturano, A. Kogelbauer, R. Prins, Stud. Surf. Sci. Catal. 125
[
[
[
[
14] K.A. Dubkov, E.A. Paukshtis, G.I. Panov, Kinet. Catal. 42 (2001) 230.
15] G. Centi, F. Vazzana, Catal. Today 53 (1999) 683.
16] R. Joyner, M. Stockenhuber, J. Phys. Chem. B 103 (1999) 5963.
17] J. Pérez-Ramírez, F. Kapteijn, G. Mul, J.A. Moulijn, Chem. Commun.
(
1999) 619.
51] P. Marturano, L. Drozdová, A. Kogelbauer, R. Prins, J. Catal. 190
2000) 460.
(
[
[
[
52] X. Feng, W.K. Hall, J. Catal. 166 (1997) 368.
53] W.K. Hall, X. Feng, J. Dumesic, R. Watwe, Catal. Lett. 52 (1998) 13.
54] P. Marturano, L. Drozdová, A. Kogelbauer, R. Prins, J. Catal. 192
(2001) 693.
[18] F. Heinrich, C. Schmidt, E. Löffler, M. Menzel, W. Grünert, J. Ca-
tal. 212 (2002) 157.
(2000) 236.
[
[
19] J. Pérez-Ramírez, F. Kapteijn, A. Brückner, J. Catal. 218 (2003) 234.
20] J. Pérez-Ramírez, G. Mul, F. Kapteijn, J.A. Moulijn, A.R. Overweg,
A. Doménech, A. Ribera, I.W.C.E. Arends, J. Catal. 207 (2002) 113.
21] J. Pérez-Ramírez, F. Kapteijn, J.C. Groen, A. Doménech, G. Mul, J.A.
Moulijn, J. Catal. 214 (2003) 33.
[
55] A.A. Battiston, J.H. Bitter, F.M.F. de Groot, A.R. Overweg, O.
Stephan, J.A. van Bokhoven, P.J. Kooyman, C. van der Spek, G.
Vanko, D.C. Koningsberger, J. Catal. 213 (2003) 251.
[
[
[
[
[
[
[
[
[
56] G. Berlier, A. Zecchina, G. Spoto, G. Ricchiardi, S. Bordiga, C. Lam-
berti, J. Catal. 215 (2003) 264.
57] G.I. Panov, A.K. Uriarte, M.A. Rodkin, V.I. Sobolev, Catal. Today 41
[22] H.-Y. Chen, W.M.H. Sachtler, Catal. Today 42 (1998) 73.
[23] A. Brückner, B. Kubias, B. Lücke, Catal. Today 32 (1996) 215.
[24] J. Pérez-Ramírez, R.J. Berger, G. Mul, F. Kapteijn, J.A. Moulijn,
Catal. Today 60 (2000) 93.
(
1998) 365.
58] J. Pérez-Ramírez, F. Kapteijn, G. Mul, J.A. Moulijn, Catal. Commun. 3
2002) 19.
(
[25] R.J. Berger, J. Pérez-Ramírez, F. Kapteijn, J.A. Moulijn, Appl. Catal.
A 227 (2002) 321.
59] A.M. Ferreti, C. Oliva, L. Forni, G. Berlier, A. Zecchina, C. Lamberti,
J. Catal. 208 (2002) 83.
60] G. Berlier, G. Spoto, G. Ricchiardi, S. Bordiga, C. Lamberti, A.
Zecchina, J. Mol. Catal. A 182–183 (2002) 359.
61] K.A. Dubkov, N.S. Ovansesyan, A.A. Shteinman, E.V. Starokon, G.I.
Panov, J. Catal. 207 (2002) 341.
62] G. Mul, J. Pérez-Ramírez, F. Kapteijn, J.A. Moulijn, Catal. Lett. 80
[26] S. Bordiga, R. Buzzoni, F. Geobaldo, C. Lamberti, E. Giamello, A.
Zecchina, G. Leofanti, G. Petrini, G. Tozzolo, G. Vlaic, J. Catal. 158
(1996).
[
[
[
27] G. Lehmann, Z. Phys. Chem. Neue Folge 72 (1970) 279.
28] J. Patarin, M.H. Tullier, J. Durr, H. Kessler, Zeolites 12 (1992) 70.
29] L.E. Iton, R.B. Beal, D.T. Hodul, J. Mol. Catal. A: Chem. 21 (1983)
(2002) 129.
1
51.
30] A. Brückner, R. Lück, W. Wieker, B. Fahlke, H. Mehner, Zeolites 12
1992) 380.
63] J. Pérez-Ramírez, A. Brückner, S. Kumar, F. Kapteijn, in: S.D. Jack-
son, J. Hargreaves, D. Lennon (Eds.), Catalysis in Application, The
Royal Society of Chemistry, Cambridge, 2003, p. 308.
[
(
[
[
[
31] R. Aasa, J. Chem. Phys. 52 (1983) 3919.
[
[
[
64] A.V. Kucherov, C.N. Montreuil, T.N. Kucherova, M. Shelef, Catal.
Lett. 56 (1998) 173.
65] A.M. Volodin, V.I. Sobolev, G.M. Zhidomirov, Kinet. Catal. 39 (1998)
32] P.N. Joshi, S.V. Awate, V.P. Shiralkar, J. Phys. Chem. 97 (1993) 9749.
33] T. Inui, H. Nagata, T. Takeguchi, S. Iwamoto, H. Matsuda, M. Inoue,
J. Catal. 139 (1993) 482.
775.
[34] A. Brückner, U. Lohse, H. Mehner, Micropor. Mesopor. Mater. 20
66] A. Ribera, I.W.C.E. Arends, S. de Vries, J. Pérez-Ramírez, R.A. Shel-
don, J. Catal. 195 (2000) 287.
(1998) 207.
[35] D. Goldfarb, M. Bernardo, K.G. Strohmaier, D.E.W. Vaughan, H.
Thomann, J. Am. Chem. Soc. 116 (1994) 6344.
[67] P. Kubánek, B. Wichterlová, Z. Sobalik, J. Catal. 211 (2002) 109.