N2O decomposition over Fe-zeolites: Structure of the active sites and the origin of the distinct reactivity of Fe-ferrierite, Fe-ZSM-5, and Fe-beta. A combined periodic DFT and multispectral study

Article abstract of DOI:10.1016/j.jcat.2010.04.008

The N₂O decomposition over Fe-ferrierite, Fe-beta, and Fe-ZSM-5 has been recently studied [K. Jisa, J. Novakova, M. Schwarze, A. Vondrova, S. Sklenak, Z. Sobalik, J. Catal. 262 (2009) 27] and a superior activity of Fe-ferrierite with respect to Fe-beta and Fe-ZSM-5 has been shown. In this study, we investigated (1) plausible active sites for the N₂O decomposition over Fe-ferrierite and (2) the origin of the distinct reactivity of Fe-ferrierite, Fe-ZSM-5 and Fe-beta employing a combined theoretical (periodic DFT) and experimental (UV-vis-NIR spectroscopy, IR spectroscopy, ²⁹Si MAS NMR spectroscopy and catalytic batch experiments) approach. We evidenced that two Fe(II) cations accommodated in two adjacent six-membered rings in the eight-membered ring channel (β sites) of Fe-ferrierite (the calculated Fe-Fe distance is 7.4 ) form the active site responsible for the superior activity of this catalyst in the N₂O decomposition in the absence of NO. Similar structures can be formed in Fe-beta. However, the probability of their formation is very low. For Fe-ZSM-5, the geometrical arrangement of the cationic positions is far from that in Fe-ferrierite and it is not suitable for the N₂O decomposition. Therefore, the predicted order of the activity of the Fe(II) exchanged zeolites agrees with our experimental findings and it is: Fe-ferrierite ? Fe-beta > Fe-ZSM-5. We further showed that the accommodation of divalent cations in rings forming cationic sites can lead to significant rearrangements of the local structures of the zeolite framework, and therefore, the precise structure of sites binding a divalent cation cannot be derived from results of X-ray diffraction experiments, but can be inferred from theoretical calculations.

Full text of DOI:10.1016/j.jcat.2010.04.008

Journal of Catalysis 272 (2010) 262–274
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
N2O decomposition over Fe-zeolites: Structure of the active sites and the origin
of the distinct reactivity of Fe-ferrierite, Fe-ZSM-5, and Fe-beta. A combined
periodic DFT and multispectral study
a
a
a
a
Stepan Sklenak a, , Prokopis C. Andrikopoulos , Bundet Boekfa , Bavornpon Jansang , Jana Nováková ,
*
b
b
b
a
a
ˇ
ˇ
ˇ
ˇ
Lubomir Benco , Tomas Bucko , Juergen Hafner , Jirí Dedecek , Zdenek Sobalík
a
´
J. Heyrovsky Institute of Physical Chemistry of the Academy of Sciences of the Czech Republic, Dolejškova 3, 182 23 Prague 8, Czech Republic
b Institut für Materialphysik and Center for Computational Materials Science, Universität Wien, Sensengasse 8, A-1090 Vienna, Austria
a r t i c l e i n f o
a b s t r a c t
Article history:
The N2O decomposition over Fe-ferrierite, Fe-beta, and Fe-ZSM-5 has been recently studied [K. Jisa, J.
Novakova, M. Schwarze, A. Vondrova, S. Sklenak, Z. Sobalik, J. Catal. 262 (2009) 27] and a superior activity
of Fe-ferrierite with respect to Fe-beta and Fe-ZSM-5 has been shown. In this study, we investigated (1)
plausible active sites for the N2O decomposition over Fe-ferrierite and (2) the origin of the distinct reac-
tivity of Fe-ferrierite, Fe-ZSM-5 and Fe-beta employing a combined theoretical (periodic DFT) and exper-
imental (UV–vis–NIR spectroscopy, IR spectroscopy, 29Si MAS NMR spectroscopy and catalytic batch
experiments) approach. We evidenced that two Fe(II) cations accommodated in two adjacent six-mem-
bered rings in the eight-membered ring channel (b sites) of Fe-ferrierite (the calculated Fe–Fe distance is
7.4 Å) form the active site responsible for the superior activity of this catalyst in the N2O decomposition
in the absence of NO. Similar structures can be formed in Fe-beta. However, the probability of their for-
mation is very low. For Fe-ZSM-5, the geometrical arrangement of the cationic positions is far from that in
Fe-ferrierite and it is not suitable for the N2O decomposition. Therefore, the predicted order of the activity
of the Fe(II) exchanged zeolites agrees with our experimental findings and it is: Fe-ferrierite ꢀ Fe-
beta > Fe-ZSM-5. We further showed that the accommodation of divalent cations in rings forming cat-
ionic sites can lead to significant rearrangements of the local structures of the zeolite framework, and
therefore, the precise structure of sites binding a divalent cation cannot be derived from results of X-
ray diffraction experiments, but can be inferred from theoretical calculations.
Received 30 October 2009
Revised 31 March 2010
Accepted 10 April 2010
Available online 15 May 2010
Keywords:
N2O decomposition
Fe in ferrierite
Fe in ZSM-5
Fe in the beta zeolite
DFT
VASP
Ó 2010 Elsevier Inc. All rights reserved.
1. Introduction
accommodated in two adjacent sites located on the opposite sides
of the eight-membered ring channel can explain the superior per-
In the past 10 years, attention was drawn to development of
methods for the abatement of N2O. Transition metal-exchanged
silicon-rich zeolites are excellent candidates for the decomposition
of nitrous oxide to nitrogen and oxygen. Fe-ZSM-5 was studied the
most both experimentally [1–8] and theoretically [9–15] but Fe-
ferrierite [16,17] as well as Fe-beta [17–20] were also investigated.
Our recent study [21] showed that Fe-zeolite samples prepared by
Fe(III) precipitation using organic solvent represent promising cat-
alytic systems with a high and stable activity under conditions of
nitrous acid plants. The catalytic activity of Fe-ferrierite in the ab-
sence of NO was found to be superior with respect to Fe-beta and
Fe-ZSM-5 (ferrierite ꢀ the beta zeolite > ZSM-5). On the other
hand, all the three zeolites have similar activities in the presence
of NO [21]. We suggested that a collaboration of two Fe(II) cations
formance of Fe-ferrierite [21].
In this article, we focus on the investigation and detailed
description of (1) plausible active sites of Fe-ferrierite and their
structures, (2) the origin of the distinct reactivity of Fe-ferrierite,
Fe-ZSM-5, and Fe-beta, (3) the very first steps of the N2O decompo-
sition, i.e. the adsorption of N2O on Fe(II) located in the active sites
and the subsequent release of molecular nitrogen to form FeO(ad).
We will explore the differences between the isolated and two
cooperating adjacent cationic sites. The course of the entire N2O
decomposition on Fe-ferrierite to form N2 and O2 is very complex
(e.g. [8,10,12,13,22–26]) as it involves many reaction steps. It is a
subject of our next study.
2. Experimental
NaK-ferrierite (Tosoh, Japan; Si/Al 8.6), NH4-beta zeolite (Zeo-
lyst; Si/Al 15.5), and Na-ZSM-5 (VURUP, SR; Si/Al 13.4) were used
* Corresponding author.
E-mail address: stepan.sklenak@jh-inst.cas.cz (S. Sklenak).
0021-9517/$ - see front matter Ó 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2010.04.008

S. Sklenak et al. / Journal of Catalysis 272 (2010) 262–274
263
as parent zeolites to prepare the catalysts. The parent samples of
ferrierite and ZSM-5 were converted to a NHþ4 form by three times
repeated equilibration with 1 M NH4NO3. Fe was introduced in the
zeolite samples employing an impregnation of the parent ammonia
zeolite forms by anhydrous FeCl3 in acetylacetone. The method for
introduction of Fe is described in detail elsewhere [27]. The result-
ing Fe/Al ratios were 0.12, 0.13, and 0.11 for ferrierite, the beta zeo-
lite, and ZSM-5, respectively. The samples of the three zeolites
were the same as those used in our recent study [21].
29Si MAS NMR experiment was carried out on the hydrated
NaK-ferrierite sample utilizing a Bruker Avance 500 MHz (11.7 T)
Wide Bore spectrometer using 4 mm o.d. ZrO2 rotor with a rotation
speed of 5 kHz. A 29Si MAS NMR high-power decoupling experi-
ment with a
p/6 (1.7 ls) excitation pulse, a relaxation delay of
30 s, and a 50% ramp cross polarization (CP) pulse sequence was
applied to collect a single pulse and cross polarization spectra.
The 29Si chemical shifts were referenced to Q8M8.
UV–vis–NIR spectra of the hydrated Fe-zeolite samples in the
range from 200 to 2200 nm were collected using a Perkin–Elmer
Lambda 950 spectrometer equipped with Spectralon™ integration
sphere. Spectralon™ served also as the reference. WinLab software
was applied to recalculate spectra using the Schuster–Kubelka–
Munk equation [28].
Fig. 1. 29Si MAS NMR spectrum of NaK-ferrierite. Experimental spectrum (a),
spectrum simulation (b) and individual Gaussian bands of the spectrum simulation
(c).
Si=AlFR ¼ I= 0:25 ꢂ n ꢂ In
R
IR spectra with a resolution of 2 cmꢁ1 of the Fe-ferrierite sample
were measured at ambient temperature after evacuation (10ꢁ3 Pa)
of the sample at 450 °C in a standard glass vacuum cell equipped
with KBr windows using a FTIR spectrometer Nexus 670 (Ther-
moNicolet). The samples were used as self-supported pellets with
the thickness from 5 to 10 mg/cm2. The spectra intensities were
normalized on the sample with the thickness of 5.5 mg/cm2 using
the integral area of the zeolite skeletal bands in the region between
1750 and 2100 cmꢁ1. The positions and relative intensities of the
bands in the ‘‘skeletal transmission window” (the region between
950 and 880 cmꢁ1, where the antisymmetric T–O–T stretching lat-
tice vibrations were observed) were evaluated by a Fourier self-
deconvolution procedure (OMNIC 4.1 software) and spectral fitting
procedure using Origin 7.5 software assuming Gaussian profiles,
respectively, of the spectra. For details of this method, see Refs.
[28–30].
Batch catalytic experiments were carried out in a reaction ves-
sel (175 cm3) at 230 and 280 °C. The samples were pretreated
in situ in vacuum at 450 °C for 2 h and then cooled to the reaction
temperatures. Then nitrous oxide (500 Pa) reacted with the pre-
treated sample for 40 min. Following that the reaction mixture
was analyzed in a Balzers QMG 420 quadrupole mass spectrometer
calibrated using individual reaction components.
where I denotes the total intensity of the 29Si NMR signal in the sin-
gle pulse experiment and In denotes the intensities of the NMR sig-
nals corresponding to the Si(4ꢁnSi, nAl) atoms [34]. Si/AlFR is in
good agreement with the results of chemical analysis (8.6). Other
interpretations of the spectra resulted in unrealistic values of Si/
AlFR. Thus, it can be concluded that Si(2Si, 2Al) atoms, i.e. Al–O–
Si–O–Al sequences, are not present in the investigated ferrierite
sample.
3.2. UV–vis–NIR spectroscopy
Normalized UV–vis–NIR spectra of the hydrated Fe-zeolite sam-
ples with low Fe/Al values (<0.15) (after impregnation, thermal
treatment and rehydration on air) are shown in Fig. 2. Four types
of bands can be resolved in the spectrum. The UV bands around
36,000 and 44,000 cmꢁ1 correspond to Fe(III)O or Fe(II) charge
transfer bands. These bands are unspecific and can correspond to
solvated isolated Fe ions or Fe ion oligomeric species or Fe ions
in oxides. The band around 29,000 cmꢁ1 can be attributed to the
charge transfer of Fe(III)O in binuclear Fe(III) species [28,35]. The
UV bands between 29,000 and 44,000 cmꢁ1 are present in the
spectra of all three samples. The bands between 16,000 and
26,000 cmꢁ1 (easily distinguishable in the spectrum of the Fe-beta
zeolite) can be assigned to pair transitions and charge transfer of
3. Experimental results
3.1. 29Si NMR spectroscopy
29Si MAS NMR spectrum of the parent ferrierite sample and its
analysis is shown in Fig. 1. The spectrum is similar to those already
reported for a ferrierite sample [31]. Six resonances at ꢁ115.3,
ꢁ110.8, ꢁ108.6, ꢁ105.0, ꢁ102.1 and, ꢁ98.9 ppm are present in
the spectrum. A cross-polarization experiment reveals (not shown
in the figure) that the resonance at ꢁ98.9 ppm corresponds to ter-
minal silanol groups (SiOH). The resonances at ꢁ108.6, ꢁ105.0 and
ꢁ102.1 ppm are assigned to Si(3Si,1Al) atoms, while the reso-
nances at ꢁ115.3 and ꢁ110.8 ppm reflect Si(4Si) atoms. A detailed
analysis of 29Si MAS NMR spectra of zeolites is described elsewhere
[32–34]. The assignment of the resonances to the individual types
of Si atoms was confirmed by quantitative analysis of the frame-
work Al content in zeolite. The Si/AlFR value of 8.5 was calculated
using equation
Fig. 2. Normalized UV–vis–NIR spectrum of as prepared Fe-ferrierite (—), Fe-ZSM-5
(— —) and Fe-beta (jjj).

264
S. Sklenak et al. / Journal of Catalysis 272 (2010) 262–274
Fe(III) oxide species. Also the weak band at 12,000 cmꢁ1 is unspe-
cific and can correspond to the d–d transitions of both isolated
Fe(III) ions and Fe(III) ions in oxide [28,36].
The results of the batch experiments at 230 and 280 °C are shown
in Table 1. The N2O decomposition over Fe-ferrierite at 230 °C as
well as Fe-beta and Fe-ZSM-5 at 280 °C yields only negligible
amounts of O2. Therefore, we can assume that the decomposition
involves almost exclusively the adsorption of N2O on Fe(II) to give
FeO(ad) and N2(g). The subsequent reaction steps which would
produce O2(g) can be neglected. Our test at 230 °C clearly evi-
dences a superior activity of Fe-ferrierite over the other two zeo-
lites, while the results obtained at 280 °C reveal a significantly
higher activity of Fe-beta over Fe-ZSM-5.
Binuclear Fe(III) species observed in fully hydrated samples evi-
denced by the band around 29,000 cmꢁ1 are significantly present
in Fe-ferrierite, while their concentration in Fe-ZSM-5 is signifi-
cantly lower and difficult to estimate in Fe-beta due to an overlap
with the bands of Fe pair transitions (from 16,000 to 26,000 cmꢁ1
)
and the strong UV band at 36,000 cmꢁ1. On the other hand, the
presence of Fe(III) oxides in Fe-beta is significant, while only traces
of Fe(III) oxides can be detected in Fe-ZSM-5 and there are no iron
oxides in Fe-ferrierite.
4. Accommodation of divalent cations in pentasil zeolites
3.3. IR spectroscopy
4.1. Cationic sites
FTIR spectrum of the dehydrated Fe-ferrierite sample (Fig. 3)
features a broad band ascribed to shifted antisymmetric T–O–T
stretching lattice modes induced by binding bare Fe(II) ions to
the framework oxygens [37]. Deconvolution of the spectrum with
the subtracted background (Fig. 3) yielded two individual bands
at 940 and 915 cmꢁ1 corresponding to two different local perturba-
tions of the framework induced by the accommodation of Fe(II) in
Fe(II) cations as well as other divalent metal cations can occupy
in pentasil zeolites (e.g. mordenite, ferrierite, ZSM-5, and the beta
zeolite) three types of cationic sites designated as a, b, and c [38–
41]. These three cationic sites for divalent cations were determined
using XRD for mordenite [42,43] and suggested employing UV–vis
spectroscopy of bare Co(II) ions as a probe for ZSM-5 [38], the beta
zeolite [41], and ferrierite [40] for which the sites were later con-
firmed by synchrotron powered XRD [44–46]. The three cationic
sites exhibit significant similarities but they differ in their detailed
geometrical arrangement.
the
a and b cationic sites, respectively [37]. Quantitative analysis of
the FTIR spectrum under the assumption that the extinction coef-
ficients of the individual bands are the same [37] indicates that
Fe(II) ions located in the b site predominate (85% of Fe(II) ions in
the zeolite) in the Fe-ferrierite sample.
The
a cationic site represents an elongated six-membered ring
which is composed of two five-membered rings. The accommo-
dated cation exhibits an open coordination sphere, since it is lo-
cated at the top of a pyramid with the base formed by four
oxygen atoms of the six-membered ring. This site is easily accessi-
ble because it is present on the wall of the main channel in ferrie-
rite, and on the wall of the sinusoidal channel of ZSM-5. But for the
beta zeolite, this site exhibits a low accessibility, since it faces the
inner space of a cage of the beta zeolite.
3.4. Catalytic experiments
Batch catalytic experiments allow investigating the very first
steps of the N2O decomposition on Fe-ferrierite, Fe-beta, and
ZSM-5, i.e. the adsorption of N2O on Fe(II) located in the active sites
and the following release of molecular nitrogen to form FeO(ad).
The b cationic site in ferrierite, ZSM-5, and the beta zeolite cor-
responds to a deformed six-membered ring. The cation is located
near the plane of the six-membered ring, and therefore, the accom-
modated cation exhibits an open coordination sphere. The b cat-
ionic site can be easily reached because it is positioned in (1) an
eight-membered ring channel of ferrierite, (2) channel crossing in
ZSM-5, (3) the wall of a cage of the beta zeolite.
The boat-shaped
c cationic site is a complex location formed by
four five-membered rings (ferrierie) or two five-membered and
two six-membered rings (ZSM-5). For the beta zeolite, this site is
located inside a beta cage. Accommodated cations are assumed
to exhibit a closed coordination sphere with a pseudo-octahedral
coordination. Therefore, only the
sidered in this study.
a and b cationic sites will be con-
The structures of cationic sites accommodating divalent cations
in different zeolite frameworks were subject of several studies [47–
49]. They indicated that the cations prefer to bind to oxygen atoms
of AlO4 tetrahedra. We will employ molecular dynamics (MD) sim-
Fig. 3. FTIR spectrum of Fe-ferrierite featuring shifted antisymmetric T–O–T
stretching lattice vibrations (a), spectra simulation (b), and Gaussian bands
corresponding to the Fe(II) in
a
(940 cmꢁ1) and b (915 cmꢁ1) sites (c).
ulations to investigate the structure and stability of the
and b ꢁ 2 sites accommodating Fe(II).
a
, b ꢁ 1,
Table 1
Amountsa of decomposed N2O and released O2 in
l
mol/g after 40 min at 230 and
4.2. Al distribution in the rings forming cationic sites
280 °C.
Zeolite
230 °C
280 °C
Two Al atoms located in the
a cationic site are required to
accommodate a divalent cation. The location of a divalent bare cat-
ion in the vicinity of only one framework Al with a distant second
framework Al is of very low probability because of the high reactiv-
ity of such a cation [50]. Our 29Si MAS NMR experiments (see Sec-
tion 3.1.) revealed that Al–O–Si–O–Al sequences are not present in
the ferrierite sample used. Therefore, there is only one possible
N2O
O2
N2O
O2
Fe-ferrierite
Fe-beta
Fe-ZSM-5
5.9
1.9
2
0.3
0
0
26
4.6
2.9
6.8
0.6
0.3
a
The starting amount of N2O is 32
result in 16 mol/g of released O2 if there is no capture of oxygen on the active
zeolite centers.
lmol/g. The total decomposition of N2O would
l
arrangement of two Al atoms in the
a site (Fig. 4). On the contrary,

S. Sklenak et al. / Journal of Catalysis 272 (2010) 262–274
265
there are two possible distributions of two Al atoms forming the
b ꢁ 1 and b ꢁ 2 sites (Fig. 4).
the distance between two cations in these sites is about 7.0 Å
(Fig. S2 of the Supplementary material). Two b sites of ZSM-5 at
the channel crossing face each other and the distance of two cat-
ions in these sites is about 4.5 Å (Fig. S3. of the Supplementary
material).
4.3. Cooperation of two adjacent sites accommodating Fe(II) cations
The distance between two cations accommodated in two b sites
of the beta zeolite is about 8.0 Å (Fig. S4. of the Supplementary
material). The b sites face each other and they are located across
the channel of the beta zeolite. The distances between two cations
Besides the cationic site accessibility and the local structure, the
distance between two cations accommodated in individual posi-
tions of the zeolite frameworks plays a significant role if two active
sites are involved in the catalyzed reaction. Rough estimates of the
distances between cations accommodated in two distinct sites
were estimated using models of the zeolite structures based on
the X-ray diffraction data [51,52].
located in two
a sites of the beta zeolite are not taken into account
because these sites are not accessible.
It should be noted that the above discussed distances reflect
only the structures of the zeolite framework and do not include
the probability of the presence of two Al atoms in the six-mem-
bered ring forming the cationic site. Our prior study [40] of the
same ferrierite sample revealed that there was a high probability
of a formation of two adjacent b cationic sites each with two Al
atoms in the ferrierite sample, for details see Section 8.2. Moreover,
binuclear Fe(III)ꢂ ꢂ ꢂFe(III) species are present in the hydrated oxi-
dized Fe-ferrierite sample, as evidenced by the presence of the
band around 29,000 cmꢁ1 in the Fe-ferrierite sample (see Sec-
tion 3.2.). Thus, both single Fe(II) cations located in isolated sites
Two
a cationic sites in ferrierite are located at the opposite sides
of the ferrierite main channel and they face each other. The dis-
tance between the two cations in these sites is about 6.0 Å
(Fig. S1 of the Supplementary material). Two b sites in ferrierite
are present in the eight-membered ring channels and also face
each other. The distance between two cations in these two sites
is calculated to be 7.4 Å (Figs. 5 and 6).
Two
a sites of ZSM-5 are located on the opposite sides of the
main channel but they do not face each other because they are
shifted in the opposite directions along the channel axes. Thus,
Fig. 4. Possible arrangements of two Al atoms in the six-membered rings forming the
atoms in yellow, iron atoms in violet, and oxygen atoms in red. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
a
site (left), b ꢁ 1 (middle), and b ꢁ 2 (right) sites. Silicon atoms are in gray, aluminum
this article.)
Fig. 5. Optimized structures of the b ꢁ 1 site (b ꢁ 1 + b ꢁ 1 model) before (left) and after (right) molecular dynamics simulations. The distances are in Å. Silicon atoms are in
gray, aluminum atoms in yellow, iron atoms in violet, and oxygen atoms in red. (For interpretation of the references to colour in this figure legend, the reader is referred to the
web version of this article.)

266
S. Sklenak et al. / Journal of Catalysis 272 (2010) 262–274
Fig. 6. Optimized structures of the b ꢁ 2 site (b ꢁ 2 + b ꢁ 2 model) before (left) and after (right) molecular dynamics simulations. The distances are in Å. Silicon atoms are in
gray, aluminum atoms in yellow, iron atoms in violet, and oxygen atoms in red. (For interpretation of the references to colour in this figure legend, the reader is referred to the
web version of this article.)
as well as two close Fe(II) cations located in two adjacent b sites are
assumed to exist in the Fe-ferrierite catalyst [21] with low or med-
ium Fe loadings. Therefore, the properties of both former and latter
Fe(II) cations are studied employing computational methods.
to investigate five distinct possible arrangements of the active sites
in ferrierite. The first three models (
, b ꢁ 1, and b ꢁ 2; see Figs. 7–
a
9, respectively) represent isolated cationic sites which do not allow
interactions of the O atom of the adsorbed N2O with another Fe(II)
located in a close cationic position, while the other two models
(b ꢁ 1 + b ꢁ 1 and b ꢁ 2 + b ꢁ 2; see Figs. 5 and 6, respectively) fea-
ture two cooperating adjacent Fe(II) cationic sites permitting inter-
actions of N2O with both Fe(II) cations. Sites accommodating Fe(II)
in the beta zeolite and ZSM-5 are not modeled. Our results ob-
5. Computational details
5.1. Structural models
tained for the
a
, b ꢁ 1, and b ꢁ 2 models of ferrierite will serve as
The starting structures were generated from the experimental
orthorhombic structure of ferrierite determined by neutron diffrac-
tion [53]. Five models possessing the P1 symmetry were employed
estimates of the outcome of calculations of isolated cationic sites
of the beta zeolite and ZSM-5.
Fig. 7. Optimized structures of the
a site (a model) before (left) and after (right) molecular dynamics simulations. The distances are in Å. Silicon atoms are in gray, aluminum
atoms in yellow, iron atoms in violet, and oxygen atoms in red. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
this article.)

S. Sklenak et al. / Journal of Catalysis 272 (2010) 262–274
267
Fig. 8. Optimized structures of the b ꢁ 1 site (b ꢁ 1 model) before (left) and after (right) molecular dynamics simulations. The distances are in Å. Silicon atoms are in gray,
aluminum atoms in yellow, iron atoms in violet, and oxygen atoms in red. (For interpretation of the references to colour in this figure legend, the reader is referred to the web
version of this article.)
Fig. 9. Optimized structures of the b ꢁ 2 site (b ꢁ 2 model) before (left) and after (right) molecular dynamics simulations. The distances are in Å. Silicon atoms are in gray,
aluminum atoms in yellow, iron atoms in violet, and oxygen atoms in red. (For interpretation of the references to colour in this figure legend, the reader is referred to the web
version of this article.)
The
a
model utilizes one unit cell of ferrierite (a = 18.651,
with a magnetic moment of 4 lB suggested by Hund’s rule for
the free Fe(II) cation [50,58]. The Kohn–Sham equations were
solved variationally in a plane-wave basis set using the projec-
tor-augmented wave (PAW) method of Blöchl [59], as adapted by
Kresse and Joubert [60]. The exchange–correlation energy was de-
scribed by the PW91 generalized gradient approximation (GGA)
functional [61,62]. Brillouin zone sampling was restricted to the
b = 14.173, and c = 7.404 Å) with two Al/Si substitutions in the
a
site (Fig. 7) accommodating one Fe(II) cation. The b ꢁ 1, b ꢁ 2,
b ꢁ 1 + b ꢁ 1, and b ꢁ 2 + b ꢁ 2 models (Figs. 8, 9, 5 and 6, respec-
tively) employ a super cell composed of two unit cells along the c
dimension. The b ꢁ 1 and b ꢁ 2 models feature one b ꢁ 1 and b ꢁ 2
site, respectively, (2 Al and 1 Fe(II) per super cell) and the other
unit cell is all-silica, while the b ꢁ 1 + b ꢁ 1 and b ꢁ 2 + b ꢁ 2 mod-
els have two b ꢁ 1 and b ꢁ 2 sites (one per unit cell), respectively,
(4 Al and 2 Fe(II) per super cell).
C
-point. The plane-wave cutoff of 400 eV was utilized for geome-
try optimizations, while a smaller cutoff of 300 eV was used for
the molecular dynamics simulations.
5.2. Electronic structure calculations
5.3. Geometry optimizations
Spin polarized periodic DFT calculations were carried out
employing the VASP code [54–57]. Bonding of the Fe(II) cation to
a zeolite matrix conserves the high-spin configuration Fe d5" d1;
The atomic positions were optimized at constant volume
employing a conjugate-gradient algorithm minimization of ener-
gies and forces while the lattice parameters were fixed at their

268
S. Sklenak et al. / Journal of Catalysis 272 (2010) 262–274
experimental values. Transition structures were identified using
the dimer method [63], as recently improved by Heyden et al. [64].
Surprisingly, the calculated Fe–Fe distance of the active site is
7.404 Å for both b ꢁ 1 + b ꢁ 1 and b ꢁ 2 + b ꢁ 2 models before as
well as after MDs (Figs. 5 and 6).
Our MD simulations of the five models and subsequent optimi-
zations of the selected MD snapshots led to energy stabilizations
relative to the five models which were not relaxed employing
MD simulations but simply optimized using the structure deter-
mined by neutron diffraction [53] as the starting geometry. The
5.4. Molecular dynamics
The molecular dynamics (MD) simulations used the exact Hell-
mann–Feynman forces acting on atoms and applied the statistics of
the canonical ensemble to the motion of the atomic nuclei [65]
using the Verlet velocity algorithm [66,67] to integrate Newton’s
equations of motion. The time step for the integration of the equa-
tions of motion was 1 fs. The simulations were run for 5 ps at 300 K
for each of the five investigated models. The structures of ten dis-
tinct ‘‘snapshots” collected from the last 500 fs of each molecular
dynamics simulation were optimized. The most stable conformer
for each of the five investigated models was then used for subse-
quent calculations.
stabilization is significant for the b ꢁ 2 and
a sites, while it is mod-
erate for the b ꢁ 1 site. The corresponding values for the five mod-
els are listed in Table 2. The largest stabilization is obtained for the
b ꢁ 2 site (ꢁ22.3 and ꢁ12.0 kcal/mol for the b ꢁ 2 + b ꢁ 2 and b ꢁ 2
models, respectively) which also rearranges its structure the most.
The
a
site is similarly stabilized (ꢁ11.0 kcal/mol for the
a model) as
b ꢁ 2, while the energy of the b ꢁ 1 site drops only slightly (ꢁ3.8
and ꢁ4.5 for the b ꢁ 1 + b ꢁ 1 and b ꢁ 1 models, respectively).
6.2. N2O adsorption and N–O cleavage on isolated Fe(II) cations
6. Computational results
The optimized structures of all the calculated species are shown
in Figs. 10 and 11, S5, and S6 of the Supplementary material).
Scheme 1 reveals the adsorption energies of N2O on Fe(II) accom-
modated in isolated cationic sites. N2O coordinates preferably by
the N terminal atom (Fig. 10). The strongest interaction is with
6.1. Structure and stability of the Fe(II) cationic sites
Our MD calculations of the five models revealed the stability of
the
a
, b ꢁ 1, and b ꢁ 2 cationic sites. The optimized structures be-
fore and after molecular dynamics calculations are shown in
Figs. 5–9. The structure of the b ꢁ 1 cationic site in the b ꢁ 1 as well
as b ꢁ 1 + b ꢁ 1 models (Figs. 8 and 5, respectively) is only slightly
Fe(II) in the
a
cationic site (
a
model): ꢁ12.3 and ꢁ9.0 kcal/mol
for N terminal and O atom adsorption, respectively. The interaction
with Fe(II) in the b ꢁ 2 site (b ꢁ 2 model) is weaker (ꢁ10.9 and
ꢁ5.7 kcal/mol for the N terminal and O atom adsorption, respec-
tively). Fe(II) located in the b ꢁ 1 site (b ꢁ 1 model) interacts only
weakly with the N terminal atom of N2O (ꢁ8.4 kcal/mol) and the
Feꢂ ꢂ ꢂONN complex is not formed at all. Scheme 1 further shows
changed, while the geometry of the
a
and b ꢁ 2 cationic sites differ
b ꢁ 2, and
b ꢁ 2 + b ꢁ 2 models (Figs. 7, 9 and 6). The calculated Fe–O bond
lengths range from 2.01 to 2.17 Å for the site, from 2.00 to
significantly after the MD simulations of the
a,
a
2.02 Å for the b ꢁ 1 site, and from 1.92 to 2.20 Å for the b ꢁ 2 site.
that after N2O is adsorbed by the O atom (
a
and b ꢁ 2), the O–N
bond is cleaved to form FeO(ad) and N2(g) (Figs. S5 and S6). The
corresponding barriers and reaction energies are 17.6/19.1 (
b ꢁ 2) and ꢁ7.7/ꢁ10.5 ( /b ꢁ 2) kcal/mol, respectively.
Table 2
a
/
Stabilization energya in kcal/mol of the cationic sites accommodating Fe(II) for the
five models.
a
Model
D
E
D
Eb
6.3. N2O adsorption and N–O cleavage on two cooperating Fe(II)
cations
a
ꢁ11.0
ꢁ4.5
ꢁ11.0
ꢁ4.5
b ꢁ 1
b ꢁ 1 + b ꢁ 1
b ꢁ 2
ꢁ3.8
ꢁ1.9
Figs. 12 and 13, S7, and S8 of the Supplementary material depict
the optimized structures of the calculated species. The energies
and barriers of the adsorption of N2O on Fe(II) accommodated in
two adjacent collaborating b sites and the following reaction steps
leading to FeO(ad) and N2(g) are shown in Scheme 2. N2O interacts
preferably by the N terminal atom (Fig. 12). The adsorption of N2O
is stronger on Fe(II) located in the b ꢁ 2 site (b ꢁ 2 + b ꢁ 2 model):
ꢁ12.0
ꢁ22.3
ꢁ12.0
ꢁ11.2
b ꢁ 2 + b ꢁ 2
a
The difference between the energies of the model relaxed by MD simulations
and subsequently optimized and the model which was not relaxed by MD simu-
lations but simply optimized using the starting structure determined by neutron
diffraction [53].
b
Stabilization energy per cationic site.
Fig. 10. Optimized structures of the Feꢂ ꢂ ꢂNNO complex formed in the
a
site (
a
model) (left), b ꢁ 1 site (b ꢁ 1 model) (middle), and b ꢁ 2 site (b ꢁ 2 model) (right). The
distances are in Å. Silicon atoms are in gray, aluminum atoms in yellow, iron atoms in violet, nitrogen atoms in blue, and oxygen atoms in red. (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of this article.)

S. Sklenak et al. / Journal of Catalysis 272 (2010) 262–274
269
Fig. 11. Optimized structures of the Feꢂ ꢂ ꢂONN complex formed in the
a site (a model) (left) and b ꢁ 2 site (b ꢁ 2 model) (right). The distances are in Å. Silicon atoms are in
gray, aluminum atoms in yellow, iron atoms in violet, nitrogen atoms in blue, and oxygen atoms in red. (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)
Scheme 1. Energies and barriers of the adsorption of N2O on Fe(II) accommodated in isolated
a and b sites and the following reaction steps leading to FeO(ad) and N2(g).
ꢁ9.5 (N terminal atom) and ꢁ3.7 (O atom) kcal/mol than on Fe(II)
accommodated in the b ꢁ 1 site (b ꢁ 1 + b ꢁ 1 model): ꢁ5.0 kcal/
mol (N terminal atom) and no interaction between the O atom of
N2O and Fe(II) is observed. Subsequently, the N–O bond is cleaved
to yield FeO(ad) and N2(ad) (Figs. 13 and S8). The corresponding
barrier is 9.4 and 12.3 kcal/mol for the b ꢁ 1 + b ꢁ 1 and
b ꢁ 2 + b ꢁ 2 models, respectively. The reaction energies of the
N–O bond cleavage to yield FeO(ad) and N2(ad) are ꢁ15.2/ꢁ17.8
(b ꢁ 1 + b ꢁ 1/b ꢁ 2 + b ꢁ 2). The adsorption of N2(ad) on Fe(II) is
weak (ꢁ2.3 (b ꢁ 1 + b ꢁ 1) and ꢁ7.4 (b ꢁ 2 + b ꢁ 2) kcal/mol).
site to ligate a divalent cation, while other rings include only one or
no Al atom (then the rings are not able to accommodate a divalent
cation). These issues result in a significant decrease in the accuracy
of the X-ray diffraction determined structure of the framework
oxygen atoms involved in coordinating the divalent cation. How-
ever, the structure of sites accommodating a divalent cation can
be inferred from theoretical calculations.
The starting structures used in this study were generated from
the experimental orthorhombic structure of ferrierite determined
by neutron diffraction [53]. Divalent metal cations located in sites
prefer to coordinate to oxygen atoms of AlOꢁ4 tetrahedra rather
than SiO4 tetrahedra, since the cations compensate the negative
charge of AlOꢁ4 . However, the starting structures featured this
proper coordination of Fe(II) to four oxygen atoms of two AlOꢁ4 tet-
rahedra only for the b ꢁ 1 site (Figs. 8 and 5), while the Fe(II) cation
is ligated to only three and two oxygen atoms of two AlOꢁ4 tetrahe-
7. Rearrangement of cationic sites accommodating Fe(II)
cations
The structure of sites binding divalent cations were estimated
using X-ray diffraction experiments only for mordenite
[42,43,68,69] and ferrierite frameworks [44–46,70,71]. However,
the obtained structures of the cationic positions represent a super-
position of different arrangements of these sites with and without
accommodated divalent cations. Moreover, some rings forming va-
cant cationic sites contain two Al atoms which are needed for the
dra for the
other one (
rahedra. Therefore, the structure of the
ing Fe(II)
a
a
(Fig. 7) and b ꢁ 2 (Figs. 9 and 6) sites, respectively. The
site) and two (b ꢁ 2) oxygen atoms belong to SiO4 tet-
a
and b ꢁ 2 sites coordinat-
(
a
,
b ꢁ 2, and b ꢁ 2 + b ꢁ 2 models) significantly
rearranged during the MD calculations. The Fe(II) cation correctly
coordinates to four oxygen atoms of two AlOꢁ4 tetrahedra in the

270
S. Sklenak et al. / Journal of Catalysis 272 (2010) 262–274
Fig. 12. Optimized structures of the [Feꢂ ꢂ ꢂNNO Fe] complex formed in the b ꢁ 1 site (b ꢁ 1 + b ꢁ 1 model) (left) and b ꢁ 2 site (b ꢁ 2 + b ꢁ 2 model) (right). The distances are
in Å. Silicon atoms are in gray, aluminum atoms in yellow, iron atoms in violet, nitrogen atoms in blue, and oxygen atoms in red. (For interpretation of the references to colour
in this figure legend, the reader is referred to the web version of this article.)
Fig. 13. Optimized structures of the transition state (N–O cleavage) formed in the b ꢁ 1 site (b ꢁ 1 + b ꢁ 1 model) (left) and b ꢁ 2 site (b ꢁ 2 + b ꢁ 2 model) (right). The
distances are in Å. Silicon atoms are in gray, aluminum atoms in yellow, iron atoms in violet, nitrogen atoms in blue, and oxygen atoms in red. (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of this article.)
resulting structures (Figs. 7, 9 and 6) of the
a
,
b ꢁ 2, and
laxed. The similarity of the structures of the b ꢁ 1 and b ꢁ 2 cat-
ionic sites accommodating Fe(II) is in good agreement with the
observation that only one type of the b site accommodating bare
Co(II) cations was determined by Vis spectroscopy of dehydrated
Co-ferrierite [29,40,72]. The Co(II) coordination is almost identical
b ꢁ 2 + b ꢁ 2 models. On the other hand, the structure of the
b ꢁ 1 site in the b ꢁ 1 and b ꢁ 1 + b ꢁ 1 models (Figs. 8 and 5)
did not noticeably rearrange since the coordination of Fe(II) was
correct in the starting structures, only the frameworks slightly re-

S. Sklenak et al. / Journal of Catalysis 272 (2010) 262–274
271
Scheme 2. Energies and barriers of the adsorption of N2O on Fe(II) accommodated in two adjacent collaborating b sites and the following reaction steps leading to FeO(ad)
and N2(g).
in both b ꢁ 1 and b ꢁ 2 sites, and therefore, they cannot be distin-
two adjacent b ꢁ 2 sites by the O atom to form the [Feꢂ ꢂ ꢂONN Fe]
complex, while there is no such adsorption in two adjacent b ꢁ 1
sites. [Feꢂ ꢂ ꢂONN Fe] very easily isomerizes into [Feꢂ ꢂ ꢂNNO Fe].
Our calculations reveal that the presence of the second Fe(II)
cation coordinated in an adjacent b site has a dramatic effect on
the course of the very first steps of the N2O decomposition. The
adsorption of N2O by the N terminal atom blocks the active site
composed of an isolated Fe(II) cationic site. Then the N–O cleavage
can proceed only when N2O adsorbs by the O atom. This adsorption
is energetically strongly disfavored. The consequence is that the N–
guished by Vis spectroscopy of Co(II) ions as a probe.
The stabilization of the
a
and b ꢁ 2 sites due to the rearrange-
ment is calculated to be about ꢁ11 to ꢁ12 kcal/mol per site while
the relaxation of the framework for the b ꢁ 1 and b ꢁ 1 + b ꢁ 1
models leads to a small drop of the energy by about 4 kcal/mol
regardless of whether the model includes one or two b ꢁ 1 sites
(Table 2).
O cleavage is sluggish for isolated
a
and b ꢁ 2 sites (barriers reach-
8. Discussion
ing 20.9 and 24.3 kcal/mol, respectively) and does not occur for the
isolated b ꢁ 1 site. On the other hand, when two Fe(II) cations
accommodated in two adjacent b sites can collaborate then the
N–O cleavage is facile (barriers are only 9.4 and 12.3 kcal/mol for
the b ꢁ 1 and b ꢁ 2 sites, respectively).
8.1. The effect of the cooperation of two Fe(II) sites on the N2O
adsorption and N–O cleavage
Schemes 1 and 2 show the very first steps of the N2O decompo-
sition, i.e. the adsorption of N2O on Fe(II) accommodated in the ac-
tive sites and the subsequent release of molecular nitrogen to form
FeO(ad). N2O can adsorb on Fe(II) by both the terminal atoms;
however, the adsorption by the terminal N atom is significantly
stronger. Therefore, almost all N2O molecules coordinate to Fe(II)
by the terminal N to give the Feꢂ ꢂ ꢂNNO complex. Further reactions
steps differ depending on whether Fe(II) is accommodated in an
isolated site or two cooperating adjacent sites.
The N2O decomposition to form FeO(ad) and N2(g) is calcu-
lated to be exothermic for all three cationic sites for both isolated
sites as well as two adjacent collaborating b sites, see Schemes 1
and 2.
The structure of the transition state involving two Fe cations
[Fe(1)ꢂ ꢂ ꢂNNO–Fe(2)] (Fig. 13) reveals the geometrical requirements
for the active site. The distance between the O atom of N2O and
Fe(2) is about 1.8 Å, the N–O bond is elongated to around 1.6 Å,
the N–N bond length is about 1.2 Å, and the Fe(1)ꢂ ꢂ ꢂN distance is
about 2.0 Å. The N and O atoms of the NNO moiety pull the Fe cat-
ions out of the plane of the ring forming the b cationic sites by ca
0.2 Å for the complexes (Fig. 12) and more than 0.6 Å for the tran-
sition state (Fig. 13). The Fe–Fe distance decreases from 7.4 Å in Fe-
ferrierite (Figs. 5 and 6) to 7.2 Å in the complexes (Fig. 12) and fur-
ther to 6.2 Å in the transition state (Fig. 13). Since the Fe(II) cations
are pulled out from the plane of the ring at least by 0.2 Å, we as-
sume that the value of 6.6 Å (6.2 + 0.2 + 0.2) is the minimum dis-
tance between two Fe(II) in Fe-zeolite needed for the transition
state to occur. On the other hand, we can estimate the maximum
distance between two collaborating Fe(II) assuming that the tran-
sition state has a linear arrangement, and furthermore, the bond
distances are close to those in the transition state (Fig. 13). Only
the Fe(1)ꢂ ꢂ ꢂN distance might be larger up to about 2.2 Å. In addi-
tion, the cations can be pulled out of the plane of the ring by up
to 0.7 Å. Therefore, we estimate that maximum span between the
two cooperating Fe(II) is 8.2 Å (1.8 + 1.6 + 1.2 + 2.2 + 0.7 + 0.7).
Our theoretical results suggesting that two adjacent collaborat-
ing b sites represent the active center for the N2O decomposition
over the low Fe loaded ferrierite are in good agreement with exper-
imental findings that cooperating Fe(II)ꢂ ꢂ ꢂFe(II) species represent
beside isolated Fe(II) ions a significant fraction of the Fe(II) species
in the investigated ferrierite sample. The formation of Fe oxide was
not observed in the investigated sample.
Fe(II) present in an isolated site (Scheme 1) is blocked, since it
forms the Feꢂ ꢂ ꢂNNO complex and the N2O decomposition cannot
proceed. Only a tiny fraction of N2O molecules ligate to Fe(II) lo-
cated in isolated
a
and b ꢁ 2 sites by the O atom to form the
Feꢂ ꢂ ꢂONN complex while there is no such adsorption for Fe(II) in
an isolated b ꢁ 1 site. The N–O bond of the Feꢂ ꢂ ꢂONN complex
can cleave to yield FeO(ad) and N2(g). The calculated barrier is
17.6 and 19.1 kcal/mol for the isolated
a
and b ꢁ 2 sites, respec-
tively. However, since the Feꢂ ꢂ ꢂONN complex is less stable than
the Feꢂ ꢂ ꢂNNO complex, the barrier of the N–O cleavage calculated
from the more stable complex (Feꢂ ꢂ ꢂNNO) is higher by the differ-
ence between the adsorption energies of the Feꢂ ꢂ ꢂNNO and
Feꢂ ꢂ ꢂONN complexes and thus reaching 20.9 and 24.3 kcal/mol
for isolated
a
and b ꢁ 2 sites, respectively.
The course of the very first steps of the N2O decomposition
notably differs when N2O can interact with two Fe(II) cations
accommodated in two adjacent b sites (Scheme 2). N2O adsorbs
to one of the Fe(II) cations by the terminal N atom to form the
[Feꢂ ꢂ ꢂNNO Fe] complex. The O atom of the adsorbed N2O in [Feꢂ ꢂ ꢂN-
NO Fe] is well positioned for the N–O cleavage, since its distance to
the other Fe(II) cation is about 3 Å and all five atoms (Fe, N, N, O,
and Fe) lie approximately in a line. This unique arrangement
causes the N–O cleavage facile. The calculated barriers are only
9.4 and 12.3 kcal/mol for the b ꢁ 1 and b ꢁ 2 sites, respectively. A
tiny fraction of N2O molecules coordinate with Fe(II) located in

272
S. Sklenak et al. / Journal of Catalysis 272 (2010) 262–274
Binuclear Fe(III)-oxo complexes are evidenced by UV–vis–NIR
rite, the beta zeolite, and ZSM-5. A comparison of the local struc-
tures of the [Feꢂ ꢂ ꢂNNO Fe] complex present in two cooperating b
sites of ferrierite (Fig. 12) and especially the subsequent transition
spectroscopy (band around 29,000 cmꢁ1) (Fig. 2) in as prepared hy-
drated catalyst (after Fe(III) precipitation, calcination and rehydra-
tion). These complexes can be formed in a zeolite with a low Fe
loading (Fe/Al 0.12) under rehydration predominantly from two
Fe ions present in the same section of the ferrierite channel, i.e.
from two Fe(II) cations located after the calcination of the sample
state (Fig. 13) with the arrangements of the
rierite, the and b sites in ZSM-5, and the b sites in the beta zeolite
reveals that (1) the sites in ferrierite and the b sites in ZSM-5 are
a cationic sites in fer-
a
a
too close (less than 6.6 Å; see Section 8.1 and Figs. S1 and S3 of the
in two opposite b or
a sites. The presence of isolated Fe(II)/Fe(III)
Supplementary material) to accommodate the transition state (2)
cations in the sample cannot be excluded due to the strong overlap
of the spectra of these ions by the adsorption of Fe(III)-oxo com-
plexes. Our Mössbauer spectroscopy results showed [73] that a
dehydration (activation) of the Fe-ferrierite sample resulted in
the prevailing presence of Fe(II) cations in the zeolite framework.
Moreover, our analysis of the perturbations of skeletal vibrations
of the ferrierite framework evidences that 85% of Fe(II) cations
are located in the b sites after the dehydration (activation) of the
ferrierite sample (Section 3.3.).
The total concentration of the b sites with two Al atoms in the
ring and capable to accommodate bare Fe(II) cations in our ferrie-
rite sample was estimated elsewhere [29,72] using Co(II) ions as a
probe, for detail see Refs. [74,75]. According to these references,
the concentration of Co(II) ions in the b site is Co/Al 0.22, while
the maximum loading of bare Co(II) ions is Co/Al 0.31. Thus, 44%
of all framework Al atoms are located as pairs in the six-membered
ring of the b site.
two a sites in ZSM-5 do not face each other but they are in an
acceptable distance and could possibly contain the transition state
(Fig. S2 of the Supplementary material); (3) two b sites in the beta
zeolite exhibit a proper arrangement similar to that of the two b
sites in ferrierite with a slightly longer distance (Fig. S4 of the Sup-
plementary material). Thus, they are also appropriate to form the
transition state.
Another important parameter besides the arrangements of the
cationic sites that controls the zeolite activity in the N2O decompo-
sition is the distribution of Al in distinguishable T sites of the zeo-
lite framework as well as the concentration of Al in these T sites. A
formation of the [Feꢂ ꢂ ꢂNNO Fe] complex in two adjacent collabo-
rating
a sites in ZSM-5 or b sites in the beta zeolite can occur only
when two cooperating sites contain two Al atoms each.
The upper limit of the probability of a formation of adjacent col-
laborating cationic sites can be estimated from the analysis of the
concentration of the
a or b sites obtained using bare Co(II) ions as a
The Si/Al ratio of the ferrierite sample is 8.6 meaning that there
are in average 3.75 Al atoms per unit cell. The Fe exchange at Fe/Al
0.12 corresponds to 0.45 Fe atoms per unit cell (3.75 ꢃ 0.12). Our
FTIR experiments (Fig. 3) reveal that 85% of Fe(II) cations is located
probe monitored by UV–vis spectroscopy [74,75].
For the investigated ZSM-5 sample, the maximum Co/Al ratio is
0.4 and thus 80% of the framework Al atoms are located in the six-
membered rings with two Al atoms. 30% of them correspond to the
in the
b
sites. Therefore, 0.38 Fe(II) cations per unit cell
a
site [74]. That means 24% of the framework Al represent the max-
imum concentration of Al atoms in the six-membered rings form-
ing the site for this sample. There are 96 T atoms and four rings
with the geometry of the site in a unit cell of ZSM-5. The Si/Al ra-
tio of 13.4 of the ZSM-5 sample used corresponds in average to 6.7
Al atoms per unit cell of which 1.6 Al atoms create the site
(6.7 ꢃ 0.8 ꢃ 0.3). Thus, 1 site with two Al atoms capable of accom-
modating Fe(II) corresponds to four other structures with the
geometry of the site but with one or no Al atom. Therefore, the
probability of the presence of the site with two Al atoms is 0.2.
Thus, the probability of two adjacent sites accommodating two
Fe(II) cations (the highest occupancy of the site by Fe(II)) is
(0.45 ꢃ 0.85) are located in the b site. This means that more than
every third unit cell contains two Al atoms in the b site balancing
Fe(II). There is one six-membered ring per unit cell, which can form
a b site if occupied by two Al atoms. A random occupation of the
cationic sites by Fe(II) in the zeolite framework then causes that al-
most every third unit cell (2 ꢃ 0.38 ꢃꢃ 2 = 0.29) corresponds to a
pair of unit cells occupied by two Fe(II) ions located in two neigh-
boring opposite b sites. This Fe occupation agrees well with the
observation that binuclear Fe(III) species are significantly present
in the hydrated Fe-ferrierite sample. This significant presence of
the binuclear Fe(III) species indicates that the distribution of Fe
ions in the zeolite channel is not random. Thus, the fraction of Fe
accommodated in two adjacent collaborating b sites might be even
significantly higher.
a
a
a
a
a
a
a
a
0.04 (0.2 ꢃ 0.2). It should be noted that the low probability (0.04)
is not in contradiction with the presence of binuclear Fe complexes
in hydrated Fe-exchanged ZSM-5 evidenced by UV–vis spectros-
copy. These binuclear complexes can be formed in the hydrated
zeolite from Fe ions located in the b sites which concentration is
This arrangement of two cooperating Fe(II) cations is quite un-
ique and exists only in the ferrierite sample used because of two
reasons: the distribution of Al atoms in the ferrierite framework
as well as the optimal structural arrangement of the b sites in
the ferrierite framework.
more than twice higher than that of the
a sites.
For the beta zeolite sample used, the Si/Al ratio is 15.5 (four Al
per unit cell in average), and the maximum Co/Al is 0.45 [41] (90%
of the framework Al atoms are located in the six-membered rings
with two Al atoms). Fifty-five percent of Co(II) cations are in the b
sites [41]. Thus, in average there are 2 (4 ꢃ 0.9 ꢃ 0.55) Al atoms
forming a b site in a unit cell. Since there are eight rings with the
geometry of the b site in a unit cell and only one of them (1/8)
has two Al atoms, the probability of two opposite b sites with Fe(II)
in the cationic sites (the highest occupancy of the b site by Fe(II)) is
0.02 (1/8 ꢃ 1/8). The low probability of the formation of two adja-
cent b sites with two Al atoms is in good agreement with the UV–
vis–NIR spectrum of the hydrated Fe-beta.
The evaluation of the accessibility of cationic sites, distance be-
tween adjacent sites, and the probability of the existence of two
cooperating sites, each with two Al atoms indicates that the ferrie-
rite structure possesses (1) an appropriate distance between two
adjacent b sites, (2) a high probability of a formation of two neigh-
boring b sites each having two Al atoms as well as (3) a linear
arrangement of the [Feꢂ ꢂ ꢂNNO Fe] complex.
8.2. Effect of the Al distribution and framework topology on the
formation of isolated and cooperating cationic sites
The difference between the local geometry of the
a and b sites is
significantly more pronounced than the deviations of the same cat-
ionic sites in different zeolite frameworks (ferrierite, the beta zeo-
lite, and ZSM-5) [38,40,41]. This indicates that the small changes of
the local geometry of the Fe(II) sites are not responsible for the dra-
matic differences in the catalytic activity of Fe-ferrierite, Fe-beta,
and Fe-ZSM-5.
Different framework topologies can significantly affect the
accessibility of Fe(II) ions located in the individual sites of the
investigated zeolites. Fe(II) in the
a site of the beta zeolite is inac-
cessible (see Section 4.1). Moreover, different zeolite framework
topologies result in different distances of the two Fe(II) cations
accommodated in two adjacent collaborating
a or b sites in ferrie-

S. Sklenak et al. / Journal of Catalysis 272 (2010) 262–274
273
The arrangement of two cooperating b sites in the beta zeolite
Acknowledgments
also allows a collaboration of two Fe(II) cations located in these
two sites and thus the [Feꢂ ꢂ ꢂNNO Fe] complex can be formed.
The distance between the two Fe(II) cations is however somewhat
larger than that in the case of two adjacent b sites in ferrierite.
However, the probability of a formation of the [Feꢂ ꢂ ꢂNNO Fe] com-
plex is low due to the low probability of a formation of two adja-
cent b sites each having two Al atoms.
This work was financially supported by the Grant Agency of the
Academy of Sciences of the Czech Republic (Projects Nos.
IAA400400812, IAA400400908, and IAA400400904), the Grant
Agency of the Czech Republic (Project No. 203/09/1627), the
Academy of Sciences of the Czech Republic (Project No.
KAN100400702), and Center of Excellence IDECAT NMP3-CT-
2005-011,730. We would like to thank Dr. Kamil Jisa for helpful
discussions regarding IR spectroscopy of transition metal ex-
changed zeolites.
On the other hand, the likelihood of the existence of two neigh-
boring cationic sites each having two Al atoms in ZSM-5 is higher
than in the beta zeolite but the geometric arrangement of the
a
sites in the zeolite framework is not suitable for a facile formation
of the linear [Feꢂ ꢂ ꢂNNO Fe] complex. Therefore, Fe-ferrierite exhib-
its a superior activity in the N2O decomposition with respect to the
other two zeolites used. The activity of Fe-beta and Fe-ZSM-5 is
estimated to be significantly weaker and negligible, respectively.
This conclusion is in good agreement with the observed activity
of the catalysts for the very first steps of the N2O decomposition,
see Section 3.4.).
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.jcat.2010.04.008.
References
[1] G.I. Panov, V.I. Sobolev, A.S. Kharitonov, J. Mol. Catal. 61 (1990) 85.
[2] V.N. Parmon, G.I. Panov, A. Uriarte, A.S. Noskov, Catal. Today 100 (2005) 115.
[3] G.D. Pirngruber, P.K. Roy, R. Prins, J. Catal. 246 (2007) 147.
[4] P.K. Roy, R. Prins, G.D. Pirngruber, Appl. Catal. B Environ. 80 (2008) 226.
[5] Z. Sobalik, Transworld Research Network, Zeolites: From Model Materials to
Industrial Catalysts 37/661 (2008) 333.
9. Conclusions
[6] J. Perez-Ramirez, J. Catal. 227 (2004) 512.
We provided evidence based on our multiple spectroscopy re-
sults, DFT calculations, and structural models that two Fe(II) cat-
ions coordinated in two adjacent b sites of Fe-ferrierite most
likely form the active site responsible for the superior activity of
this catalyst in the N2O decomposition in the low temperature re-
gion. This conclusion does not imply that this double iron site
which is specific for ferrierite of the optimal Si/Al ratio is the only
possible type of active site for the N2O decomposition. Other types
of iron containing species can be responsible for the catalytic activ-
ity namely in the high temperature region as well as in the pres-
ence of NO.
The calculated Fe–Fe distance of the active site is 7.4 Å. The for-
mation of the active sites results from a combination of (1) a suit-
able topology of the ferrierite framework and (2) an appropriate
distribution of Al in the distinguishable T sites of the ferrierite
framework as well as concentration of Al in these T sites. Both
six-membered rings forming the two neighboring b sites must con-
tain two Al atoms each (four Al atoms in total).
Two adjacent b sites of Fe-beta can also form an active site with
two collaborating Fe(II) cations. However, there are significant dif-
ferences with respect to ferrierite. The Fe–Fe distance is somewhat
longer and, more importantly, the distribution and concentration
of Al in the rings with the geometry of the b site lead to a very
low probability of occurrence of the active sites with two cooper-
ating Fe(II) cations. This is responsible for the superior activity of
Fe-ferrierite relative to Fe-beta.
[7] E.V. Kondratenko, J. Perez-Ramirez, J. Phys. Chem. B 110 (2006) 22586.
[8] L.V. Pirutko, V.S. Chernyavsky, E.V. Starokon, A.A. Ivanov, A.S. Kharitonov, G.I.
Panov, Appl. Catal. B Environ. 91 (2009) 174.
[9] N. Hansen, A. Heyden, A.T. Bell, F.J. Keil, J. Catal. 248 (2007) 213.
[10] N. Hansen, A. Heyden, A.T. Bell, F.J. Keil, J. Phys. Chem. C 111 (2007) 2092.
[11] A. Heyden, A.T. Bell, F.J. Keil, J. Catal. 233 (2005) 26.
[12] A. Heyden, N. Hansen, A.T. Bell, F.J. Keil, J. Phys. Chem. B 110 (2006) 17096.
[13] A. Heyden, B. Peters, A.T. Bell, F.J. Keil, J. Phys. Chem. B 109 (2005) 1857.
[14] B.R. Wood, J.A. Reimer, A.T. Bell, J. Catal. 209 (2002) 151.
[15] M.F. Fellah, I. Onal, Catal. Today 137 (2008) 410.
[16] I. Melian-Cabrera, C. Mentruit, J.A.Z. Pieterse, R.W. van den Brink, G. Mul, F.
Kapteijn, J.A. Moulijn, Catal. Commun. 6 (2005) 301.
[17] A.H. Oygarden, J. Perez-Ramirez, Appl. Catal. B Environ. 65 (2006) 163.
[18] J. Perez-Ramirez, J.C. Groen, A. Bruckner, M.S. Kumar, U. Bentrup, M.N.
Debbagh, L.A. Villaescusa, J. Catal. 232 (2005) 318.
[19] J.A.Z. Pieterse, G.D. Pirngruber, J.A. van Bokhoven, S. Booneveld, Appl. Catal. B
Environ. 71 (2007) 16.
[20] M. Mauvezin, G. Delahay, B. Coq, S. Kieger, J.C. Jumas, J. Olivier-Fourcade, J.
Phys. Chem. B 105 (2001) 928.
[21] K. Jisa, J. Novakova, M. Schwarze, A. Vondrova, S. Sklenak, Z. Sobalik, J. Catal.
262 (2009) 27.
[22] W.B. Tolman, Angew. Chem., Int. Ed. 49 (2010) 1018.
[23] H. Guesmi, D. Berthomieu, B. Bromley, B. Coq, L. Kiwi-Minsker, Phys. Chem.
Chem. Phys. 12 (2010) 2873.
[24] H. Guesmi, D. Berthomieu, L. Kiwi-Minsker, J. Phys. Chem. C 112 (2008) 20319.
[25] M. Rivallan, G. Ricchiardi, S. Bordiga, A. Zecchina, J. Catal. 264 (2009) 104.
[26] Q. Shen, L.D. Li, Z.P. Hao, Z.P. Xu, Appl. Catal. B Environ. 84 (2008) 734.
[27] CZ 293917 B6 (2001).
[28] L. Capek, V. Kreibich, J. Dedecek, T. Grygar, B. Wichterlova, Z. Sobalik, J.A.
Martens, R. Brosius, V. Tokarova, Microporous Mesoporous Mater. 80 (2005)
279.
[29] Z. Sobalik, J. Dedecek, D. Kaucky, B. Wichterlova, L. Drozdova, R. Prins, J. Catal.
194 (2000) 330.
[30] Z. Sobalik, Z. Tvaruzkova, B. Wichterlova, Microporous Mesoporous Mater. 25
(1998) 225.
[31] P. Sarv, B. Wichterlova, J. Cejka, J. Phys. Chem. B 102 (1998) 1372.
[32] J.M. Thomas, J. Klinowski, Adv. Catal. 33 (1985) 199.
[33] C.A. Fyfe, G.C. Gobbi, G.J. Kennedy, J. Phys. Chem. 88 (1984) 3248.
[34] G. Engelhardt, U. Lohse, E. Lippmaa, M. Tarmak, M. Magi, Z. Anorg. Allg. Chem.
482 (1981) 49.
[35] L. Lopes, J. de Laat, B. Legube, Inorg. Chem. 41 (2002) 2505.
[36] A.C. Scheinost, A. Chavernas, V. Barron, J. Torrent, Clay Clay Miner. 46 (1998)
528.
[37] B. Wichtelova, Z. Sobalik, J. Dedecek, Appl. Catal. B Environ. 41 (2003) 97.
[38] J. Dedecek, D. Kaucky, B. Wichterlova, Microporous Mesoporous Mater. 35–36
(2000) 483.
[39] J. Dedecek, B. Wichterlova, J. Phys. Chem. B 103 (1999) 1462.
[40] D. Kaucky, J. Dedecek, B. Wichterlova, Microporous Mesoporous Mater. 31
(1999) 75.
Two close
nating two collaborating Fe(II) cations can compose the active site.
However, the geometrical arrangement of two adjacent sites is
a sites of the Fe(II) exchanged ZSM-5 zeolite coordi-
a
very distant from that of two adjacent b sites in ferrierite. Also
the probability of a formation of these active sites is low due to
an unfavorable Al distribution and Al concentration in the rings
forming the
a sites.
The results of our experimental and theoretical investigation
show that the order of the activity of the Fe(II) exchanged zeolites
is as follows: Fe-ferrierite ꢀ Fe-beta > Fe-ZSM-5.
Our DFT calculations showed that the accommodation of diva-
lent cation in rings forming cationic sites can lead to a significant
rearrangement of the local structure of the zeolite framework. This
is in agreement with the results of previous studies indicating that
the cations prefer to bind to oxygen atoms of AlO4 tetrahedra and
thus change the structure of the cationic site.
[41] J. Dedecek, L. Capek, D. Kaucky, Z. Sobalik, B. Wichterlova, J. Catal. 211 (2002)
198.
[42] W.J. Mortier, J. Phys. Chem. 81 (1977) 1334.
[43] J.L. Schlenker, J.J. Pluth, J.V. Smith, Mater. Res. Bull. 13 (1978) 169.
[44] M.C. Dalconi, A. Alberti, G. Cruciani, J. Phys. Chem. B 107 (2003) 12973.

274
S. Sklenak et al. / Journal of Catalysis 272 (2010) 262–274
[45] M.C. Dalconi, A. Alberti, G. Cruciani, P. Ciambelli, E. Fonda, Microporous
Mesoporous Mater. 62 (2003) 191.
[46] M.C. Dalconi, G. Cruciani, A. Alberti, P. Ciambelli, M.T. Rapacciuolo,
Microporous Mesoporous Mater. 39 (2000) 423.
[59] P.E. Blochl, Phys. Rev. B 50 (1994) 17953.
[60] G. Kresse, D. Joubert, Phys. Rev. B 59 (1999) 1758.
[61] J.P. Perdew, J.A. Chevary, S.H. Vosko, K.A. Jackson, M.R. Pederson, D.J. Singh, C.
Fiolhais, Phys. Rev. B 46 (1992) 6671.
[47] D. Berthomieu, N. Jardillier, G. Delahay, B. Coq, A. Goursot, Catal. Today 110
(2005) 294.
[48] M.H. Groothaert, K. Pierloot, A. Delabie, R.A. Schoonheydt, Phys. Chem. Chem.
Phys. 5 (2003) 2135.
[62] J.P. Perdew, Y. Wang, Phys. Rev. B 45 (1992) 13244.
[63] G. Henkelman, H. Jonsson, J. Chem. Phys. 111 (1999) 7010.
[64] A. Heyden, A.T. Bell, F.J. Keil, J. Chem. Phys. 123 (2005) 224101.
[65] S. Nose, J. Chem. Phys. 81 (1984) 511.
[49] S.A. McMillan, R.Q. Snurr, L.J. Broadbelt, Microporous Mesoporous Mater. 68
(2004) 45.
[66] L. Verlet, Phys. Rev. 159 (1967) 98.
[67] L. Verlet, Phys. Rev. 165 (1968) 201.
[50] L. Benco, T. Bucko, R. Grybos, J. Hafner, Z. Sobalik, J. Dedecek, J. Hrusak, J. Phys.
Chem. C 111 (2007) 586.
[51] G. Vezzalini, S. Quartieri, E. Galli, A. Alberti, G. Cruciani, A. Kvick, Zeolites 19
(1997) 323.
[52] J.M. Newsam, M.M.J. Treacy, W.T. Koetsier, C.B. Degruyter, Proc. Roy. Soc. Lond.
A Math. 420 (1988) 375.
[68] M.P. Attfield, S.J. Weigel, A.K. Cheetham, J. Catal. 170 (1997) 227.
[69] W.J. Mortier, J.J. Pluth, J.V. Smith, Mater. Res. Bull. 10 (1975) 1037.
[70] M.C. Dalconi, G. Cruciani, A. Alberti, P. Ciambelli, Catal. Today (2005) 345.
[71] M.P. Attfield, S.J. Weigel, A.K. Cheetham, J. Catal. 172 (1997) 274.
[72] L. Drozdova, R. Prins, J. Dedecek, Z. Sobalik, B. Wichterlova, J. Phys. Chem. B 106
(2002) 2240.
[53] I.J. Pickering, P.J. Maddox, J.M. Thomas, A.K. Cheetham, J. Catal. 119 (1989) 261.
[54] G. Kresse, J. Furthmuller, Phys. Rev. B 54 (1996) 11169.
[55] G. Kresse, J. Furthmuller, Comput. Mater. Sci. 6 (1996) 15.
[56] G. Kresse, J. Hafner, Phys. Rev. B 48 (1993) 13115.
[57] G. Kresse, J. Hafner, Phys. Rev. B 49 (1994) 14251.
[58] L. Benco, T. Bucko, R. Grybos, J. Hafner, Z. Sobalik, J. Dedecek, S. Sklenak, J.
Hrusak, J. Phys. Chem. C 111 (2007) 9393.
[73] M. Schwarze, Z. Sobalik, E.G. Caspary, D. Niznansky, Czech. J. Phys. 56 (2006)
E147.
[74] J. Dedecek, D. Kaucky, B. Wichterlova, O. Gonsiorova, Phys. Chem. Chem. Phys.
4 (2002) 5406.
[75] J. Dedecek, D. Kaucky, B. Wichterlova, Chem. Commun. (2001) 970.