Reactivity of binuclear Fe complexes in over-exchanged Fe/ZSM5, studied by in situ XAFS spectroscopy 2. Selective catalytic reduction of NO with isobutane

Article abstract of DOI:10.1016/S0021-9517(03)00120-9

Over-exchanged Fe/ZSM5 has exhibited a remarkable activity for the decomposition of N₂O and for the SCR of NO_x using hydrocarbons (HC-SCR), even in the presence of water. The catalytically active sites in Fe/ZSM5 were determined using in situ XAFS spectroscopy. The number of (inactive) spectators in this catalyst was minimized using a specially dedicated calcination procedure. Binuclear Fe complexes with a Fe-O-Fe core were catalytically active species in this material. During heat treatment in He to 350°C, the Fe complexes underwent auto-reduction, due to the removal of oxygen from Fe-O-Fe bridges (closest Fe-O-coordination sphere) and the formation of Fe-□-Fe vacancies. Treatment with isobutane led to a further slight average reduction of iron, accompanied by an additional removal of oxygen from the Fe-O-Fe bridges. Fe in the binuclear complexes was readily oxidized when NO, NO + O₂, or a typical HC-SCR mixture (NO, isobutane, O₂) was fed to the Fe/ZSM5 catalyst. The binuclear complexes remained stably ligated to the framework and that no agglomeration of iron toward larger clusters occurred in the presence of NO and NO + O₂.

Full text of DOI:10.1016/S0021-9517(03)00120-9

Journal of Catalysis 218 (2003) 163–177
www.elsevier.com/locate/jcat
Reactivity of binuclear Fe complexes in over-exchanged Fe/ZSM5,
studied by in situ XAFS spectroscopy
2. Selective catalytic reduction of NO with isobutane
A.A. Battiston, J.H. Bitter, and D.C. Koningsberger ∗
Department of Inorganic Chemistry and Catalysis, Debye Institute, Utrecht University, Sorbonnelaan 16, 3584 CA Utrecht, The Netherlands
Received 11 November 2002; revised 27 February 2003; accepted 3 March 2003
Abstract
In situ XAFS spectroscopy was applied in order to determine the catalytically active sites in Fe/ZSM5, prepared by FeCl sublimation. The
3
number of (inactive) spectators in this catalyst was minimized by using a specially dedicated calcination procedure. The catalytic activity of
Fe/ZSM5 during the collection of the XAFS data was monitored via chemiluminescence analysis of the gas outlet. Binuclear Fe complexes
◦
with a Fe–O–Fe core were found to be catalytically active species in this material. During heat treatment in He to 350 C the Fe complexes
undergo auto-reduction, ascribed to the removal of oxygen from Fe–O–Fe bridges (closest Fe–O-coordination sphere) and the formation of
Fe–ꢀ–Fe vacancies. Treatment with isobutane results in a further slight average reduction of iron, accompanied by an additional removal of
oxygen from the Fe–O–Fe bridges. Fe in the binuclear complexes is readily oxidized when NO, NO +O , or a typical HC-SCR mixture (NO,
2
i-C H , O ) is fed to the Fe/ZSM5 catalyst. Reoxidation is accompanied by filling of the Fe–ꢀ–Fe vacancies. Under HC-SCR working
4
10
2
conditions the average oxidation state of iron is 3+.
2003 Elsevier Inc. All rights reserved.
Keywords: Fe/ZSM5; Binuclear; Fe; Reactivity; He; O ; NO; Isobutane; deNO ; HC-SCR; XAFS; In situ
x
2
1. Introduction
O2, and i-C4H10 (or alternatively, a different alkane). Based
on these results, and on catalytic tests performed with la-
beled reactants, a model for the HC-SCR reaction pathway
has been proposed. According to this model, the first HC-
SCR steps are believed to be the oxidation of NO by O2
to form NO2, which is then released to the gas phase, and
the formation and adsorption of nitro (NO2)− and nitrate
(NO3)− complexes. These (NOy)− complexes react further
with the hydrocarbon, forming organic nitro and nitroso
compounds. The nitro/nitroso deposits may evolve through
different pathways, depending on the nature of the hydrocar-
bon. Independent of the type of hydrocarbon employed, the
final step in the pathway is proposed to be the reaction of the
organic compound with NO2 from the gas phase, leading to
the formation of a N–N bond and the desorption of N2.
The active species responsible for the peculiar catalytic
properties of over-exchanged Fe/ZSM5 are believed to
consist of extraframework binuclear Fe-oxo/hydroxo com-
plexes, with a [OH–Fe–O–Fe–OH]2+ core [1–3,7,8], ligated
to the Brønsted aluminum sites of the zeolite. Several spec-
troscopic studies have indeed demonstrated that binuclear
Fe-oxo/hydroxo complexes are generated by the reaction
Over-exchanged Fe/ZSM5, obtained by chemical vapor
deposition (CVD) of FeCl3, has shown a remarkable activ-
ity for the decomposition of N2O [1] and for the selective
catalytic reduction of nitrogen oxides using hydrocarbons
(HC-SCR), even in the presence of water [2–5]. Catalytic
tests have demonstrated that the HC-SCR catalytic activity
of Fe/ZSM5 depends on the concentration of oxygen and
on the type of hydrocarbon employed in the feed. The high-
est N2 yields have been obtained in the 2–10 vol% O2 range
by using n-butane or isobutane as reductant [3].
In order to unravel the HC-SCR reaction pathway on
Fe/ZSM5, several studies have been performed by apply-
ing in situ FTIR spectroscopy [3,4,6]. Intermediate adsorp-
tion species have been monitored while feeding (separately,
or at the same time) typical HC-SCR reagents, i.e., NO,
*
Corresponding author.
E-mail address: d.c.koningsberger@chem.uu.nl
(D.C. Koningsberger).
0021-9517/03/$ – see front matter
doi:10.1016/S0021-9517(03)00120-9
2003 Elsevier Inc. All rights reserved.

164
A.A. Battiston et al. / Journal of Catalysis 218 (2003) 163–177
of sublimated FeCl3 with the Brønsted protons of the ze-
olite and by the subsequent hydrolysis processes occur-
ring during washing [9–12]. The Fe-binuclear species in
washed Fe/ZSM5 consist of two octahedrally coordinated
FeIII oxo/hydroxo ions, with a structure similar that of the
closest Fe-binuclear building unit in α-goethite [10,11].
The final treatment in the synthesis of Fe/ZSM5 is cal-
cination. Calcination, applied to stabilize and activate the
Fe phase, is a delicate step for the further evolution of the
Fe-binuclear complexes. Our group has shown that, if the
calcination procedure is not carefully controlled, this treat-
ment can result in the removal of a significant fraction of
iron from the zeolite micropores, finally resulting in the for-
mation of large catalytically inactive [8] goethite/hematite
crystals [11,13]. On the contrary, by adequately tuning the
conditions applied during calcination a high concentration
of Fe-binuclear species in Fe/ZSM5 can be preserved. From
our EXAFS study [11] binuclear Fe clusters in “mildly cal-
cined” Fe/ZSM5 (Si/Al = 17) are estimated to account for
at least 70% of the total iron amount. This material is there-
fore the best suited for characterizing structure and reactivity
of the Fe-binuclear complexes, minimizing the interference
of (inactive) spectators.
In a previous paper [14] we have studied the structure of
the Fe-binuclear complexes in Fe/ZSM5, measured at 30 ◦C
in He, upon exposure to air. Furthermore, we have focused
our attention on the changes occurring in the Fe–O coor-
dination of the iron complexes during heating treatments
in He and in a He/O2 (50/50) mixture. The results ob-
tained, discussed in Ref. [14], are summarized in the model
presented in Fig. 1. Upon exposure to air, six Fe–O neigh-
bors have been identified in the as-synthesized Fe/ZSM5.
In the model proposed the oxygen atoms have been lumped
together in three different Fe–O shells. The closest Fe–O
shell (Fe–O1, 1.85–1.89 Å) has been assigned to a termi-
nal hydroxyl group and to an oxygen atom bridging the Fe
ions, according to the [HO–Fe–O–Fe–OH]2+ core model
proposed by Chen and Sachtler [2]. The Fe–O shell at in-
termediate distance (Fe–O2, 1.97–1.99 Å) has been assigned
to two oxygen atoms ligating the Fe atoms to the aluminum-
centered tetrahedra of the zeolite [14]. The most distant shell
(Fe–O3, 2.02–2.09 Å) has been attributed to an oxygen atom
of the zeolite and to an adsorbed water molecule, respec-
tively.
The assignment of the different Fe–O shells is based
on the results obtained during the heat treatments. Heat-
ing Fe/ZSM5 to a moderate temperature (up to ∼ 150 ◦C),
in He as well as in the presence of O2, results in the re-
moval of approximately one oxygen atom (O3) from the
most distant Fe–O shell. Since this is not accompanied by
changes in the oxidation state of iron, the oxygen removal is
consistent with desorption of weakly bound H2O. The fur-
ther evolution of the Fe complexes at higher temperature
(ꢀT = 150–350◦C) is strongly influenced by the compo-
sition of the gas phase. In He a significant fraction of iron
undergoes auto-reduction, a phenomenon already known in
the literature for over-exchanged Fe/ZSM5 [9,13,15]. This
phenomenon coincides with the removal of only one oxygen
atom (O1) from the closest Fe–O shell. The most plausi-
ble interpretation appears to be the removal of the Fe–O–Fe
bridging atom, as discussed in Ref. [14]. The high reactiv-
ity of the bridging oxygen is claimed to be responsible for
the capability of Fe/ZSM5 to oxidize NO to NO2 and to de-
compose N2O [1,8]. Noticeably, the presence of oxygen in
the gas phase completely suppresses both the reduction of
iron and the removal of the Fe–O–Fe bridging oxygen. The
core structure of the binuclear complexes at 350 ◦C in oxy-
gen (PO = 0.5 bar) matches the model previously proposed
2
by Chen and Sachtler [2].
In this paper the structure and the reactivity of the Fe-
binuclear complexes in mildly calcined over-exchanged
Fe/ZSM5 are studied under HC-SCR working conditions.
This has been achieved by monitoring the local environment
of the Fe-binuclear complexes at 350 ◦C: (i) in the presence
of single HC-SCR reagents, i.e., NO, O2, or i-C4H10, fed
separately and (ii) during the HC-SCR reaction, i.e., in the
contemporary presence of the three reagents. Information on
the coordination geometry of Fe, obtained by the analysis of
the Fe K-preedge features (centroid position and integrated
intensity), has been used as input information for the fit-
ting of the recorded EXAFS data. The catalytic activity of
Fe/ZSM5 during the XAFS measurements has been mon-
itored by analyzing the composition of the gas phase via a
NOx chemiluminescence analyzer.
2. Experimental
2.1. Catalyst preparation
NH4/ZSM5 (Si/Al = 17.0, determined by ICP), ob-
tained from Zeolyst, was converted into the acidic form
by calcination in oxygen at 550 ◦C for 3 h. The result-
ing H/ZSM5 was used as host for the synthesis of over-
exchanged Fe/ZSM5, by applying the FeCl3 chemical vapor
deposition technique [2]. The procedure applied has been de-
scribed in detail previously [11,14]. First, anhydrous FeCl3
Fig. 1. Proposed structure and reactivity of the binuclear oxo/hydroxo Fe
complexes in Fe/ZSM5, obtained by FeCl CVD, during heating treatments
3
◦
to 350 C in helium and in the presence of oxygen [14].

A.A. Battiston et al. / Journal of Catalysis 218 (2003) 163–177
165
was sublimated into the zeolite micropores. The sample was
2.3. In situ Fe K-edge XAFS
subsequently washed under stirring in doubly demineralized
water for hydrolyzation and removal of excess chlorine. Af-
ter drying overnight at 70 ◦C, the sample was calcined. The
procedure applied (mild calcination) was identical to the
one proposed in our previous work [11]. The sample was
heated under a He flow of 800 ml/min with a moderate tem-
perature ramp (0.5 ◦C/min) to 200 ◦C. At this temperature,
200 ml/min of O2 were added to the He flow while, un-
der the same temperature ramp, heating was continued to
550 ◦C. After 3 h at 550 ◦C, the temperature was decreased
to 30 ◦C. The sample after calcination will be further de-
noted as Fe/ZSM5.
2.3.1. XAFS data collection
The XAFS data were collected at the XI.1 measuring
station at Hasylab (Hamburg, Germany). The experimen-
tal setup used is described in detail in Ref. [11]. Fe/ZSM5
was pressed to obtain self-supporting wafers, calculated to
have a total absorption (µ × x) of 2.5, and placed in a
controlled atmosphere cell [16]. Spectra were recorded af-
ter specific pretreatments. The complete list of the XAFS
measurements, including detailed information on pretreat-
ments, measuring conditions, and sample codes, is collected
in Table 2. Fe/ZSM5 was heated to 350 ◦C under 50 ml/min
flowing He (with the exception of measurement Fe/ZSM5-
O2 350/350, heated in the presence of oxygen). After 10 min
stabilization at 350 ◦C, specific gas mixtures were fed to
the cell, while the temperature was maintained at 350 ◦C.
The composition of the different feeds is presented in Ta-
ble 3. After 30 min on stream, three spectra were recorded
in situ at 350 ◦C. After closing the in situ cell and cooling to
77 K, three additional scans were recorded. XANES spectra
of hematite and Fe-titanate were also measured, upon purg-
ing with He at RT and cooling to 77 K. They were used,
respectively, as internal FeIII and FeII references. The three
scans recorded for each measurement were compared and
averaged, with the exception of the measurement Fe/ZSM5-
He 350/350. As extensively discussed in Ref. [14], due to
differences between the scans, caused by a partial reoxida-
tion of the sample with time on stream, the three spectra of
this measurement were not averaged. In this case, only data
from the first scan were used.
2.2. Catalytic testing
Fe/ZSM5 was tested for the selective catalytic reduc-
tion of NO with isobutane under plug-flow conditions. The
catalyst (powder) was pressed, sieved to 90–212 µm, and
charged in the reactor upon dilution with quartz. The sam-
ple was preheated to 350 ◦C with a rate of 5 ◦C/min under a
He flow of 100 ml/min. The reaction feed was obtained by
blending four flows, i.e., 1% NO/He, 1% i-C4H10/He, O2,
and He. Each flow rate was regulated by a digital Brooks
mass-flow controller. The resulting inlet composition was
2000 vppm NO, 2000 vppm i-C4H10, 3 vol% O2, balance
He, with an overall flow of 140 ml/min. Catalytic tests were
performed using 100 or 20 mg catalyst. Based on an apparent
zeolite density of 0.5 g/cm3, the calculated gas hourly space
velocity (GHSV) was respectively 42,000 and 210,000 h−1
.
The operating conditions applied during the activity tests are
summarized in Table 1.
2.3.2. Analysis of the Fe K preedge
A Perkin-Elmer Autosystem XL gas chromatograph
equipped with a TCD detector was used to monitor the ef-
fluent N2, i-C4H10, N2O, CO, and CO2. The gases were
separated using a Haysep Q and a Molsieve 13 X column
(80–100 mesh). To measure NO and NO2, a Thermo En-
vironmental Instruments 42C NOx chemiluminescence an-
alyzer was connected in parallel to the gas chromatograph.
The GC and the NOx analyzer could be operated simulta-
neously. Catalytic data reported were measured after 60 min
stabilization at each temperature.
The oxidation state of iron was determined by the analy-
sis of the centroid position of the background subtracted Fe
K preedge and by the position of the Fe K absorption edge.
In addition, information on the local coordination geometry
of iron was derived from the analysis of the preedge cen-
troid position and of the background subtracted preedge in-
tegrated intensity [17]. The Fe K preedge was isolated from
the XAFS spectra by using a cubic spline function, obtained
by interpolating the data several electron volts before and
after the preedge [14,17]. Centroid position and integrated
intensity of the background-subtracted preedge were calcu-
lated with the software Grams. The estimated accuracy in
the determination of the centroid position is 0.1 eV. The
error in the integrated intensity is approximately 5%.
Table 1
Description of the reactor and of the operating conditions during HC-SCR
catalytic tests with isobutane on Fe/ZSM5
Parameters description
Test 1
Test 2
Catalyst amount (Fe/ZSM5) (mg)
Quartz amount (g)
Sieve fraction (µm)
100
4.0
90–212
12
36
140
20
0.45
90–212
4
25
140
2.3.3. XAFS data processing
XAFS data were extracted from the measured absorp-
tion spectra by means of the XDAP code [18]. The preedge
was subtracted using a modified Victoreen curve [19]. Back-
ground subtraction from the absorption XAFS spectra was
performed by employing cubic spline routines, with a con-
tinuously adjustable smooth parameter [20,21]. Normaliza-
Reactor internal diameter (mm)
Total bed length (mm)
3
Total feed rate (cm /min)
−1
Calculated GHSV (h
)
42,000
210,000
Feed composition (vol%): NO (0.2), i-C H (0.2), O (3.0), balance He.
4
10
2

166
A.A. Battiston et al. / Journal of Catalysis 218 (2003) 163–177
Table 2
List of XAFS spectra recorded from Fe/ZSM5: pretreatments and measuring conditions
Pretreatments (flowing atmosphere)
Environmental conditions during measurement
Measurements code
◦
◦
↑; He, 30 min at 350 C
Flowing He, 350 C
Fe/ZSM5-He 350/350
◦
◦
⇑; O /He, 60 min at 350 C
Flowing O /He, 350 C
Fe/ZSM5-O 350/350
2
2
2
◦
◦
↑; i-C H /He, 60 min at 350 C
Flowing i-C H /He, 350 C
Fe/ZSM5-CH 350/350
x
4
10
4 10
◦
↑; i-C H /He, 90 min at 350 C
i-C H /He, 77 K
Flowing NO/He, 350 C
NO/He, 77 K
Fe/ZSM5-CH 350/LN
x
4
10
4 10
◦
◦
↑; NO/He, 60 min at 350 C
Fe/ZSM5-NO 350/350
Fe/ZSM5-NO 350/LN
◦
↑; NO/He, 90 min at 350 C
◦
◦
↑; NO + O /He, 60 min at 350 C
Flowing NO + O /He, 350 C
Fe/ZSM5-NO + O 350/350
2
2
2
◦
↑; NO + O /He, 90 min at 350 C
NO + O /He, 77 K
Fe/ZSM5-NO + O 350/LN
2
2
2
◦
◦
↑; deNO/He, 60 min at 350 C
Flowing deNO/He, 350 C
Fe/ZSM5-deNO 350/350
Fe/ZSM5-deNO 350/LN
◦
↑; deNO/He, 90 min at 350 C
deNO/He, 77 K
◦
◦
◦
◦
↑, heating treatment to 350 C (5 C/min) performed in He. ⇑, heating treatment to 350 C (5 C/min) performed in O /He.
2
Table 3
Errors in the numerical values obtained by the EXAFS
data analysis are estimated to be 10% in the coordination
number (N), 1% in the distance (R), 5% in the Debye
Waller factor (ꢀσ2), and 10% in the inner potential cor-
rection (ꢀE0) [26].
Inlet gas composition during in situ XAFS measurements of Fe/ZSM5
Measurements code
Inlet gas composition (vol%)
He (100)
Fe/ZSM5-He 350/350
Fe/ZSM5-O 350/350
Fe/ZSM5-CH 350/350
O (50), balance He
2
2
i-C H (0.2), balance He
4 10
x
Fe/ZSM5-NO 350/350
NO (0.2), balance He
2.3.5. Catalytic activity during XAFS data collection
In order to measure the catalytic activity of Fe/ZSM5
during the collection of XAFS spectra Fe/ZSM5-NO 350/
350, Fe/ZSM5-NO + O2 350/350, and Fe/ZSM5-deNO
350/350 (see Table 2), a Thermo EnvironmentalInstruments
42C NOx chemiluminescence analyzer was connected in
parallel to the outlet of the cell. The analyzer was calibrated
independently for the detection of NO and NO2. Concentra-
tions of NO and NO2 were measured at a time interval of
1 min, while recording the XAFS spectra. Stabilization of
the NO and NO2 concentration in the outlet was obtained
for all the three measurements within 60 min from the be-
ginning of the treatments.
Possible oxidation of NO to NO2 occurring already at
room temperature in the lines was monitored prior to and
after the collection of the XAFS spectra Fe/ZSM5-NO
350/350, Fe/ZSM5-NO + O2 350/350, and Fe/ZSM5-
deNO 350/350. This was obtained by directly feeding the
inlet gas to the NOx analyzer via a bypass connected in par-
allel to the cell. The bypass was connected to the main line
by means of two three-way valves, located at the inlet and at
the outlet of the cell, respectively, and operated simultane-
ously. In all cases conversion of NO to NO2 in the line was
found to be below 0.4% with respect to the inlet concentra-
tion.
Fe/ZSM5-NO + O 350/350
NO (0.2), O (5), balance He
2
2
Fe/ZSM5-deNO 350/350
NO (0.2), i-C H (0.2), O (5),
4 10 2
balance He
tion was performed by dividing the data by the intensity of
the absorption spectrum at 50 eV above the Fe K edge.
2.3.4. EXAFS data analysis, phase shifts, backscattering
amplitudes
The local environment around iron, i.e., number and dis-
tances of Fe-neighboring atoms, was determined by EX-
AFS [20]. The information on the coordination geometry of
Fe, obtained from the analysis of the Fe K preedge, was used
as input for the EXAFS analysis.
The Fe–O reference was obtained from experimental EX-
AFS data of ferric acetylacetonate. The procedure followed
for the creation of the Fe–O reference file is described in
Ref. [11]. Fe–Fe phase shifts and backscattering amplitudes
were calculated using the software FEFF7 [22]. The Fe–Fe
reference was calibrated on EXAFS data obtained from
hematite at 77 K [11], by fitting in R space. The crystal-
lographic parameters used for the fit of the calibration refer-
ence [23,24] can be found in Ref. [11].
The EXAFS data analysis was performed by applying
multiple-shell fitting in R space (ꢀR = 0.7–4.2 Å; ꢀk =
2.7–13.2 Å−1). The difference file technique was applied
together with phase-corrected Fourier transforms to resolve
the different contributions in the EXAFS data [20]. The EX-
AFS fits were checked by applying k1, k2, and k3 weight-
ings.
The number of independent fit parameters (Nindp) was de-
termined as outlined in Reports on Standard and Criteria in
XAFS Spectroscopy [25]. For the measurements performed
in this work it was calculated to be 25.4.
3. Results
3.1. HC-SCR catalytic testing
The results of the catalytic test performed on Fe/ZSM5
for the selective catalytic reduction of NO with isobutane
under excess oxygen at 42,000 h−1 GHSV are depicted in
Fig. 2a. Feed composition and GHSV for this test were cho-
sen in order to be able to compare our results to the ones

A.A. Battiston et al. / Journal of Catalysis 218 (2003) 163–177
167
50/50 O2/He mixture (Fe/ZSM5-O2 350/350) (dotted line)
are presented in Fig. 3a. The corresponding background-
subtracted preedges are depicted in the upper-left part of
the figure. XANES spectra measured from Fe-titanate and
hematite are reported as internal references. As can be seen,
the position of the Fe/ZSM5 Fe K edges are separated by
approximately 4.0 eV. While the edge of the sample treated
in oxygen corresponds to that of hematite (FeIII), the edge
of the sample measured in He appears to be close to that of
Fe-titanate (FeII). The features of the preedges appear also to
be different. The preedge measured in He is clearly shifted
toward lower energy and displays a lower intensity.
Centroid position and integrated intensity of the preedge
in Fe/ZSM5, measured upon the heat treatments, are plot-
ted in Fig. 3b. In the same figure the Fe K preedge features
of crystalline single-phase iron-containing references (large
gray circles), obtained from the work of Wilke et al. [17], are
reported for comparison. The Fe references differ for oxida-
tion state (II or III) and for coordination geometry (from per-
fectly octahedral(6) to perfectly tetrahedral(4)). The number
in the upper brackets indicates the number of neighboring
atoms in the Fe-coordination sphere. In the upper x axis of
the figure the average oxidation state of iron (a.o.s.) is plotted
as a function of the preedge centroid energy (lower x axis).
This relation has been obtained by XAFS measurements on
(6)FeIII/(4)FeII physical mixtures [17]. By comparison with
the references, it can be concluded that the average oxida-
tion state of iron in Fe/ZSM5 measured at 350 ◦C in He is
close to 2.3+. Although only qualitative conclusions can be
derived by the present data [14], the average Fe coordina-
tion in Fe/ZSM5 at 350 ◦C in helium appears to be close
to that of fivefold or highly distorted tetrahedrally coordi-
nated iron. Reduction of iron and oxygen removal from the
Fe-coordination sphere, as already discussed in our previous
work [14], are the result of auto-reduction phenomena occur-
ring to Fe/ZSM5 during the heat treatment in the absence of
oxygen. Iron in untreated Fe/ZSM5, indeed, is octahedrally
coordinated, with an oxidation state of 3+ [11,14].
Fig. 2. Selective catalytic reduction of NO by isobutane over Fe/ZSM5
as a function of the reaction temperature: (Q) N yield, (!) i-C H
2
4 10
conversion, (2) i-C H conversion to CO, (1) i-C H conversion
4
10
4 10
to CO . Inlet composition: 0.2% i-C H + 0.2% NO + 3% O , bal-
2
4
10
2
3
−1
−1
;
ance He. Total flow rate = 140 cm /min . (a) GHSV = 42,000 h
−1
(b) GHSV = 210,000 h . GHSV are based on an apparent zeolite den-
−3
sity of 0.5 g/cm
.
already available in the literature [2]. In agreement with
those results, a maximum in the conversion of NO to N2 was
obtained at a temperature of about 360 ◦C. The maximum
N2 yield was around 78%, confirming the high activity of
Fe/ZSM5 toward the HC-SCR reaction, when using isobu-
tane as reductant. Contemporarily, a considerable amount
of CO (above 50% yield at maximum NO conversion) was
formed, as a result of the incomplete i-C4H10 combustion.
Both N2O and NO2 were absent from the products, the se-
lectivity toward N2 being complete in the whole temperature
range.
Reducing the residence time of the reactants (Fig. 2b,
GHSV = 210,000 h−1) caused a shift in the maximum N2
yield toward higher temperature (to approximately 400 ◦C).
Nevertheless, even under this extreme flow rate, the catalyst
maintained a considerably high activity (around 36% max-
imum N2 yield), while no N-containing by-products were
detected in the gas phase.
The oxidation state of iron in Fe/ZSM5 measured in the
presence of oxygen at 350 ◦C is 3+. The Fe-coordination
geometry corresponds to that of a fivefold coordinated refer-
ence. This shows that heating in oxygen influences the Fe–O
coordination, decreased from six (untreated sample [14]) to
five oxygen neighbors, but does not change the oxidation
state of iron.
The differences in the Fe–O coordination sphere of
Fe/ZSM5-He 350/350 and Fe/ZSM5-O2 350/350 can be
visualized by comparing the Fourier transforms of their EX-
AFS data (see Fig. 3c). The most distant peaks (Fe–Fe1 and
Fe–Fe2) are due to Fe–Fe neighbors. As already discussed
in our previous work [14], they are ascribed to the presence
of binuclear Fe complexes, together with a minor fraction of
Fe species with a higher Fe–Fe coordination. The main peak
(Fe–O) is due to oxygen. As can be seen, this Fe–O peak in
Fe/ZSM5-O2 350/350 displays a higher intensity (real part)
than in Fe/ZSM5-He 350/350. Furthermore the imaginary
3.2. Heat treatment in He and O2
The normalized XANES spectra of Fe/ZSM5 recorded at
350 ◦C in He (Fe/ZSM5-He 350/350) (solid line) and in a

168
A.A. Battiston et al. / Journal of Catalysis 218 (2003) 163–177
(a)
(b)
(c)
(d)
Fig. 3. (a) Normalized XANES spectra and background-subtracted preedges (enlarged in the upper-left part of the figure) of Fe/ZSM5-He 350/350 (solid
III
line) and Fe/ZSM5-O 350/350 (dotted line). Normalized XANES spectra of hematite and Fe-titanate are also reported, respectively as internal Fe and
2
II
Fe references. (b) Preedge characteristics (centroid position and integrated intensity) of Fe/ZSM5-He 350/350 (!) and Fe/ZSM5-O 350/350 ("). Preedge
Fe , Fe : (!)), obtained from the literature [17], are
reported for comparison. The average oxidation number (a.o.s.), reported in the upper x axis against the corresponding energy of the centroid positions (lower
2
(6) III (5) III (4) III (6) II (4) II
characteristics of iron with known coordination and oxidation state ( Fe
,
Fe
,
Fe
,
∗
(6) III (4) II
1
−1
)
x axis), is based on experiments performed on Fe
/
Fe binary mixtures with predetermined ratios. (c) Fourier transform (k , ꢀk = 2.7–13.2 Å
of EXAFS data of Fe/ZSM5-He 350/350 (solid line) and Fe/ZSM5-O 350/350 (dotted line). (d) Fourier transform of the Fe–O contribution obtained by
2
1
1
−1
R-space analysis (k , ꢀk = 2.7–13.2 Å ; ꢀR = 0.7–4.2 Å) of EXAFS data of Fe/ZSM5-He 350/350 (solid line) and Fe/ZSM5-O 350/350 (dotted line).
2
part appears to be clearly shifted toward lower distance. This
indicates that in Fe/ZSM5-O2 350/350 additional oxygen is
present, located at low R.
Fe–O1 was attributed to a Fe–O–Fe bridging oxygen and
to a terminal OH group. As can be seen, the most signifi-
cant differences in the fit results for Fe/ZSM5-He 350/350
and Fe/ZSM5-O2 350/350 are in shell Fe–O1. Approxi-
mately two oxygen atoms (1.9) are present for Fe/ZSM5-
O2 350/350, while only 1.2 oxygen atoms are present for
Fe/ZSM5-He 350/350. This difference can be explained by
the removal of the bridging Fe–O–Fe oxygen, as a result of
auto-reduction of iron at 350 ◦C in He. The clear differences
in the Fe–O1 shell as a function of the gas-phase composi-
tion can also be visualized qualitatively in Fig. 3d, where, for
both measurements, the EXAFS Fourier transform of only
contribution Fe–O1 is plotted. As can be seen, the intensity
of the shell in He appears to be almost halved. The presence
of O2 in the gas phase inhibits removal of oxygen.
The coordination parameters obtained by R-space analy-
sis of the EXAFS data are presented in Table 4. For the
fit, a distribution of oxygen in three different sub-shells
was chosen. The assignment of three different Fe–O sub-
shells was based on the results obtained during the heat
treatments [14]. The model proposed for the evolution of
the binuclear complexes upon heat treatments is shown in
Fig. 1. Shell Fe–O2, which remained unchanged, was as-
cribed to two oxygen atoms, anchoring the Fe atoms of
the binuclear complexes to the Al-centered tetrahedra of
the zeolite. The most distant O shell (Fe–O3) could be at-
tributed to a semibridging oxygen from the zeolite. Shell

A.A. Battiston et al. / Journal of Catalysis 218 (2003) 163–177
169
Table 4
sented for measurements Fe/ZSM5-CHx 350/350 (dashed
line), Fe/ZSM5-He 350/350 (solid line) and Fe/ZSM5-O2
350/350 (dotted line). As expected, in Fe/ZSM5 measured
under isobutane this shell is significantly less intense than
in the presence of oxygen. The intensity of the shell ap-
pears to be slightly lower than in the sample measured
under He. The results confirm that the removal in a re-
ducing atmosphere of oxygen is ascribable to the Fe–O–Fe
bridges. They also demonstrate that almost all reactive oxy-
gen from the sample can be removed by merely heating.
A slight increase is visible in the coordination number of
shell Fe–O3 (from 1.2 to 1.4) when comparing the fit results
of Fe/ZSM5-He 350/350 (Table 4) with those of Fe/ZSM5-
CHx 350/350 (Table 5). Since EXAFS in not able to dis-
tinguish between light-scattering atoms (in this study O, C,
N), a precise assignment to the scattering neighbors is not
possible. Nevertheless, it appears reasonable to ascribe the
increase of the coordination number of shell Fe–O3 to ab-
sorbed C-containing species.
1
Coordination parameters obtained by R-space analysis (k , ꢀk = 2.7–
−1
13.2 Å ; ꢀR = 0.7–4.2 Å) of EXAFS data of Fe/ZSM5-He 350/350
and Fe/ZSM5-O 350/350; N
= 25.4; N
R (Å) ꢀσ (10 Å ) ꢀE (eV)
= 22
2
indp
2
free (fit)
−3
2
1
k
Shells
N
variance (%)
0
10% 10%
5%
10% Im. part Abs. part
Fe/ZSM5-He 350/350
0.55
1.26
0.30
0.50
Fe–O
Fe–O
Fe–O
Fe–Fe
Fe–Fe
Fe–O
1.2
2.0
1.2
1.1
0.4
1.6
1.88
1.97
2.08
2.99
3.40
4.09
2.6
7.2
8.8
13.0
17.2
15.0
−5.5
12.6
−9.3
0.4
0.1
−8.8
1
2
3
1
2
Z
Fe/ZSM5-O 350/350
2
Fe–O
Fe–O
Fe–O
Fe–Fe
Fe–Fe
Fe–O
1.9
2.0
1.0
1.2
0.4
4.4
1.86
1.99
2.02
2.99
3.40
3.81
2.0
5.0
1.6
9.8
15.1
15.0
−2.5
13.0
−13.3
−1.6
−3.7
1.0
1
2
3
1
2
Z
Table 6 collects the coordinationparameters of Fe/ZSM5,
measured upon cooling the in situ cell to 77 K in the pres-
ence of the different reactants. A more severe increase in
the coordination number of the Fe–O3 shell (from N = 1.4
to 1.8), accompanied by a shift in its distance (from 2.08 to
2.14 Å), was detected upon cooling the sample under isobu-
tane to 77 K. This can be explained by a further adsorption
of C-containing species from isobutane, occurring during the
cooling process.
The XANES spectra, the preedge features and the EX-
AFS results of the heat treatments in He and O2 will be
used as reference for the analysis of the XAFS data obtained
in situ, during the separate feeding of i-C4H10, NO, and
NO + O2 and during the HC-SCR reaction with isobutane
(see Tables 2 and 3).
3.3. Treatment with isobutane
The normalized XANES spectrum and the background-
subtracted preedge of Fe/ZSM5 recorded at 350 ◦C in the
presence of isobutane (Fe/ZSM5-CHx 350/350, dashed
line) is presented in Fig. 4a. The XANES spectra and
preedges of Fe/ZSM5-He 350/350 (solid line) and of
Fe/ZSM5-O2 350/350 (dotted line) are also plotted, for
comparison. Both the Fe K edge and the preedge of
Fe/ZSM5-CHx 350/350 appear to be similar to those mea-
sured at 350 ◦C in helium and only slightly shifted toward
lower energy. This is ascribed to a minor additional reduc-
tion of Fe. The additional Fe reduction is confirmed by the
analysis of the preedge features (Fig. 4b). The centroid po-
sition corresponds to an average iron oxidation state of 2.2.
In Fig. 4c the Fourier transform of the EXAFS data of
measurement Fe/ZSM5-CHx 350/350 (dotted line) is com-
pared to that of Fe/ZSM5-He 350/350 (solid line). Differ-
ences are visible only in the 1.3–2.0 Å R range, where the
real part of the sample treated with isobutane appears to be
slightly enhanced. No differences are visible in the Fe–Fe
coordination, showing no further aggregation of iron dur-
ing the treatment. The Fourier transform of the EXAFS data
of Fe/ZSM5-CHx 350/350 (solid line) and the R-space fit
(dotted line) are presented in Fig. 4d. It can be seen that the
quality of the R-space fit is excellent. The coordination pa-
rameters obtained by the R-space analysis of the EXAFS
data of Fe/ZSM5-CHx 350/350 are presented in Table 5.
In Fig. 4e the Fourier transform of shell Fe–O1 is pre-
3.4. Treatment with NO and NO + O2
The catalytic activity of Fe/ZSM5 measured while record-
ing the XAFS spectra of Fe/ZSM5-NO 350/350 and Fe/
ZSM5-NO + O2 350/350 is reported in Fig. 5. The con-
version of NO is depicted by the black bars, while the cor-
responding NO2 yield is depicted in white. Mass balance,
calculated on nitrogen, showed that no measurable quanti-
ties of N2 or N2O were formed during the treatments. The
concentration of the outlet gases reached a steady state after
around 60 min from the beginning of the treatments and re-
mained constant during the whole measurement. The delay
in the stabilization is the result of back-mixing processes in
the volume of the cell. As can be seen in Fig. 5 almost no ac-
tivity was detected while flushing NO (2000 ppm) at 350 ◦C
over Fe/ZSM5. A NO2 yield lower than 1% was measured.
This could be the result of oxidation of NO by traces of
oxygen in the gas phase but is most probably due to inaccu-
racy in the NO2 measurement. Clear oxidation of NO, on the
other hand, was detected when feeding NO (2000 ppm) to
the sample in the presence of oxygen (5%). A NO2 yield of
approximately 35% was obtained, as a result of the catalytic
oxidation of NO by Fe/ZSM5.
The normalized XANES spectrum and the background-
subtracted preedge of Fe/ZSM5-NO 350/350 (dashed line)
and of Fe/ZSM5-NO + O2 350/350 (dashed-dotted line)
are presented in Fig. 6a. The XANES spectra and preedges

170
A.A. Battiston et al. / Journal of Catalysis 218 (2003) 163–177
(a)
(b)
(c)
(d)
(e)
Fig. 4. (a) Normalized XANES spectra and background-subtracted preedges of Fe/ZSM5-CH 350/350 (dashed line), Fe/ZSM5-He 350/350 (solid line),
x
and Fe/ZSM5-O 350/350 (dotted line). (b) Preedge characteristics (centroid position and integrated intensity) of Fe/ZSM5-CH 350/350 (1) and of
x
2
1
−1
Fe/ZSM5-He 350/350 (!). For details on Fe references, see caption Fig. 3b. (c) Fourier transform (k , ꢀk = 2.7–13.2 Å ) of EXAFS data of Fe/ZSM5-CH
x
1
−1
350/350 (dotted line) and Fe/ZSM5-He 350/350 (solid line). (d) Fourier transform (k , ꢀk = 2.7–13.2 Å ) of EXAFS data of Fe/ZSM5-CH 350/350
x
1
(solid line) and R-space fit (dotted line; ꢀR = 0.7–4.2 Å). (e) Fourier transform of the Fe–O contribution obtained from the R-space analysis (k ,
1
−1
ꢀk = 2.7–13.2 Å ; ꢀR = 0.7–4.2 Å) of EXAFS data of Fe/ZSM5-CH 350/350 (dashed line), Fe/ZSM5-He 350/350 (solid line), and Fe/ZSM5-O
x
2
350/350 (dotted line).

A.A. Battiston et al. / Journal of Catalysis 218 (2003) 163–177
171
Table 5
Table 6
1
1
Coordination parameters obtained by R-space analysis (k , ꢀk = 2.7–
Coordination parameters obtained by R-space analysis (k , ꢀk = 2.7–
−1
−1
13.2 Å ; ꢀR = 0.7–4.2 Å) of EXAFS data of Fe/ZSM5-CH 350/350,
13.2 Å ; ꢀR = 0.7–4.2 Å) of EXAFS data of Fe/ZSM5-CH 350/LN,
x
x
Fe/ZSM5-NO 350/350, Fe/ZSM5-NO + O 350/350, and Fe/ZSM5-
Fe/ZSM5-NO 350/LN, Fe/ZSM5-NO + O 350/LN, Fe/ZSM5-deNO
2
2
deNO 350/350; N
= 25.4; N
= 22
350/LN; N
= 25.4; N
= 22
−3
indp
free (fit)
−3
indp
free (fit)
2
2
1
2
2
1
k
Shells
N
R (Å) ꢀσ (10 Å ) ꢀE (eV)
k
variance (%)
Shells
N
R (Å) ꢀσ (10 Å ) ꢀE (eV)
variance (%)
0
0
10% 10%
5%
10% Im. part Abs. part
10% 10%
5%
10% Im. part Abs. part
Fe/ZSM5-CH 350/350
0.36
1.18
1.55
1.37
0.27
0.45
0.51
0.60
Fe/ZSM5-CH 350/LN
0.45
0.88
0.66
0.80
0.22
0.52
0.30
0.60
x
x
Fe–O
Fe–O
Fe–O
Fe–Fe
Fe–Fe
Fe–O
1.0
2.0
1.4
1.2
0.4
1.7
1.88
1.97
2.08
2.99
3.40
4.09
0.8
8.4
5.7
14.5
17.2
15.0
−5.5
12.6
−9.2
0.4
0.1
−8.8
Fe–O
Fe–O
Fe–O
Fe–Fe
Fe–Fe
Fe–O
1.0
2.0
1.8
1.2
0.4
2.7
1.85
1.97
2.14
2.99
3.40
4.02
−1.4
1.0
9.5
3.8
15.1
15.0
−0.4
12.6
−12.5
1.1
1
2
3
1
2
3
1
2
1
2
−6.6
3.1
Z
Z
Fe/ZSM5-NO 350/350
Fe/ZSM5-NO 350/LN
Fe–O
Fe–O
Fe–O
1.7
2.0
1.0
1.2
0.4
4.4
1.86
1.99
2.02
3.00
3.40
3.81
1.6
8.5
0.5
8.9
15.0
15.0
−1.5
9.5
−10.2
−2.7
−3.7
1.0
Fe–O
Fe–O
Fe–O
Fe–Fe
Fe–Fe
Fe–O
1.7
2.0
1.2
1.2
0.4
4.4
1.86
1.99
2.09
3.00
3.40
3.81
−3.5
−3.8
8.7
1.6
15.0
15.0
−3.2
9.7
−12.6
−0.4
−3.7
1.0
1
2
3
1
2
3
Fe–Fe
Fe–Fe
Fe–O
1
2
1
2
Z
Z
Fe/ZSM5-NO + O 350/350
Fe/ZSM5-NO + O 350/LN
2
2
Fe–O
Fe–O
Fe–O
Fe–Fe
Fe–Fe
Fe–O
2.0
2.0
1.0
1.2
0.4
4.1
1.86
1.97
2.02
2.96
3.40
3.91
1.0
3.2
1.0
10.5
15.0
15.0
−6.9
13.0
−6.0
1.9
0.1
−2.2
Fe–O
Fe–O
Fe–O
Fe–Fe
Fe–Fe
Fe–O
2.0
2.0
1.2
1.2
0.4
4.2
1.85
1.97
2.05
2.98
3.40
3.91
−4.0
−5.0
−1.2
3.1
15.0
15.0
−4.0
13.0
−6.2
2.0
0.0
−2.2
1
2
3
1
2
3
1
2
1
2
Z
Z
Fe/ZSM5-deNO 350/350
Fe/ZSM5-deNO 350/LN
Fe–O
Fe–O
Fe–O
1.8
2.0
1.2
1.2
0.4
5.1
1.85
1.97
2.02
2.97
3.40
3.85
3.2
6.0
9.1
9.5
15.0
15.0
0.3
13.0
−12.1
−0.2
0.0
Fe–O
Fe–O
Fe–O
Fe–Fe
Fe–Fe
Fe–O
2.0
2.0
1.4
1.2
0.4
5.3
1.85
1.97
2.09
3.02
3.40
3.91
−4.5
−5.0
0.2
3.2
15.0
15.0
−0.7
13.0
−6.3
−2.7
0.0
1
2
3
1
2
3
Fe–Fe
Fe–Fe
Fe–O
1
2
1
2
−0.4
−2.2
Z
Z
gen (FeIII). They are separated by 0.4 eV, accounting for a
slight average reduction of iron when NO was fed alone.
No differences are visible when comparing the preedges of
Fe/ZSM5-NO + O2 350/350 and Fe/ZSM5-O2 350/350.
The preedge of Fe/ZSM5-NO 350/350, on the contrary, ap-
pears to be slightly less intense and shifted to lower energy.
The preedge features are visualized in Fig. 6b. When NO
is fed together with O2, the features of iron in Fe/ZSM5
are typical of fivefold-coordinated FeIII. A similar coordina-
tion is measured in the absence of oxygen (Fe/ZSM5-NO
350/350), together with an average oxidation number of
2.8–2.9.
The Fourier transform of the EXAFS data of Fe/ZSM5-
NO 350/350 and Fe/ZSM5-NO + O2 350/350 is reported
in Fig. 6c. As can be seen, the total Fe–O coordination
of both measurements overlaps with that of Fe/ZSM5-O2
350/350. The imaginary part of Fe/ZSM5-NO 350/350 in
the R range 1.0–2.0 Å appears only slightly shifted to higher
distance when compared to that of Fe/ZSM5-NO + O2
350/350. The structural parameters obtained by the analysis
of the EXAFS data, measured at 350 ◦C and LN temperature,
are presented in Tables 5 and 6, respectively. Measurements
Fig. 5. NO conversion (black bars) and NO yield (white bars), as de-
2
tected by NO chemiluminescence analysis of the gas outlet, during in situ
x
XAFS measurements of Fe/ZSM5-NO 350/350 and Fe/ZSM5-NO + O
2
350/350.
of Fe/ZSM5-He 350/350 (solid line) and of Fe/ZSM5-O2
350/350 (dotted line) are included as references. The Fe K
edge in the presence of NO and NO + O2 appears to be
positioned close to that measured in the presence of oxy-

172
A.A. Battiston et al. / Journal of Catalysis 218 (2003) 163–177
(a)
(b)
(c)
(d)
Fig. 6. (a) Normalized XANES spectra and background-subtracted preedges of Fe/ZSM5-NO 350/350 (dashed line), Fe/ZSM5-NO + O 350/350
2
(dashed-dotted line), Fe/ZSM5-He 350/350 (solid line), and Fe/ZSM5-O 350/350 (dotted line). (b) Preedge characteristics (centroid position and integrated
2
intensity) of Fe/ZSM5-NO 350/350 (E) and Fe/ZSM5-NO + O 350/350 (F). For details on Fe references, see caption to Fig. 3b. (c) Fourier transform
2
1
−1
(k , ꢀk = 2.7–13.2 Å ) of EXAFS data of Fe/ZSM5-NO 350/350 (dashed line), Fe/ZSM5-NO+O 350/350 (dashed-dotted line), Fe/ZSM5-He 350/350
2
1
−1
(solid line), and Fe/ZSM5-O 350/350 (dotted line). (d) Fourier transform of the Fe–O contribution obtained by R-space analysis (k , ꢀk = 2.7–13.2 Å
;
2
1
ꢀR = 0.7–4.2 Å) of EXAFS data of Fe/ZSM5-NO 350/350 (dashed line), Fe/ZSM5-NO + O 350/350 (dashed-dotted line), Fe/ZSM5-He 350/350 (solid
2
line), and Fe/ZSM5-O 350/350 (dotted line).
2
Fe/ZSM5-NO 350/350 and Fe/ZSM5-NO + O2 350/350
show a total Fe–O coordination of 4.7 and 5.0, respectively.
Differences between the two measurements are visible in the
coordination number of the closest shell (Fe–O1), showing
the presence on average of 1.7 (Fe/ZSM5-NO 350/350) or
2.0 (Fe/ZSM5-NO + O2 350/350) neighbors. Similar coor-
dination numbers have been obtained for the same measure-
ments upon cooling to LN temperature.
The Fourier transforms of the EXAFS function of the
closest Fe–O shell (Fe–O1) are shown in Fig. 6d. Differ-
ences are visible between the intensities of the real part
of the Fourier transforms. The intensities are in the order
Fe/ZSM5-NO 350/350 < Fe/ZSM5-O2 350/350 < Fe/
ZSM5-NO + O2 350/350.
3.5. SCR of NO with isobutane
The catalytic activity of the Fe/ZSM5 sample towards
the HC-SCR reaction of NO with isobutane was monitored
during the collection of the XAFS data of measurement
Fe/ZSM5-deNO 350/350. A complete stabilization of the
composition of the outlet gas was obtained upon approxi-
mately 60 min under reaction. The outlet composition re-
mained stable during the recording of the XAFS spectra. The
catalytic results are shown in Fig. 7. The conversion of NO
is indicated by the black bar, while the NO2 yield is depicted
in white. The N2 yield (gray bar) was calculated by mass
balance on nitrogen, by assuming no formation of N2O. In-
deed, as shown in Figs. 2a and 2b, N2O is not a by-product

A.A. Battiston et al. / Journal of Catalysis 218 (2003) 163–177
173
indeed a slight decrease in the intensity of the real part. This
may indicate a slight removal of oxygen from the Fe–O–Fe
bridges. Due to the contemporary presence of chemically
different light scatterers, attribution to specific neighbors is
not possible. It should furthermore be noted that the slight
decrease in intensity in the Fe–O1 contribution is not accom-
panied by a reduction of iron.
Cooling to LN temperature resulted in an average in-
crease in the number of neighbors in the closest Fe coordina-
tion (Fe/ZSM5-deNO 350/LN, Table 6), ascribed to further
adsorption at low temperature. The coordination number of
shell Fe–O1 appeared to be slightly increased, from 1.8 to
2.0 neighbors. Also the coordination of shell Fe–O3, pre-
viously attributed to adsorbed water [14], appeared to be
enhanced (from 1.2 to 1.4). This may be the result of adsorp-
tion, upon cooling, of water produced during the HC-SCR
reaction, but could as well be caused by the presence of
C/N-containing reaction-intermediates.
Fig. 7. NO conversion (black bar), NO yield (white bar), and N yield
2
2
(dark-gray bar), as detected by NO chemiluminescence analysis of the gas
x
outlet, during in situ XAFS measurements of Fe/ZSM5-deNO 350/350.
∗
N
yield is calculated from mass balance on nitrogen, assuming no forma-
2
tion of N-containing products different from NO and NO .
2
of HC-SCR with isobutane on Fe/ZSM5. Furthermore, the
reproducibility of the catalytic results was confirmed by sub-
sequent catalytic tests performed in the in situ XAFS cell,
by measuring N2 directly by gas chromatography. As can be
see in Fig. 7, a NO conversion of 32% could be obtained.
This result demonstrates that the XAFS spectra were ob-
tained from the catalyst under deNOx working conditions.
Together with N2 (28% yield) also the formation of a minor
amount of NO2 was detected (4% yield). Since NO2 is not a
by-product of the HC-SCR reaction catalyzed by Fe/ZSM5,
its formation was most probably the result of oxidation of
NO by oxygen, occurring on the stainless-steel walls of the
cell or in the gas phase.
The normalized XANES spectrum and the background-
subtracted preedge of Fe/ZSM5 recorded at 350 ◦C dur-
ing the SCR reaction of NO by isobutane (Fe/ZSM5-deNO
350/350, dashed line) are presented in Fig. 8a and com-
pared with those of Fe/ZSM5-He 350/350 (solid line) and
of Fe/ZSM5-O2 350/350 (dotted line). The position of the
Fe K edge shows that the average oxidation state of iron
in Fe/ZSM5 during reaction is 3+. The intensity of the
preedge of Fe/ZSM5-deNO 350/350 appears to be lower
than that measured at 350 ◦C in oxygen. This can be ascribed
to the presence of reaction intermediates in the coordination
of iron. The analysis of the preedge features, presented in
Fig. 8b, confirms that the oxidation state of iron during re-
action is 3+. The integrated preedge intensity is consistent
with the presence on average of 5 to 6 neighboring atoms.
The Fourier transform of the EXAFS data of Fe/ZSM5-
deNO 350/350 is shown in Fig. 8c. As in the case of separate
feeding of NO or NO + O2 (Fig. 6c), the Fourier trans-
form appears similar to that of measurement Fe/ZSM5-O2
350/350. The EXAFS data analysis (Table 5) shows that the
coordination number of the Fe–O1 shell is slightly reduced,
from 2.0 to 1.8. The Fourier transform of the EXAFS func-
tion of the Fe–O1 contribution (plotted in Fig. 8d) shows
4. Discussion
4.1. Catalytic activity of Fe/ZSM5 toward SCR of NO with
isobutane
The catalytic tests performed on the Fe/ZSM5 sample,
further used for the in situ XAFS characterization, confirm
the high activity and selectivity of this catalyst for the selec-
tive catalytic reduction of NO to N2 with isobutane (Figs. 2a
and 2b). The performance of our sample appears to be iden-
tical to that shown by previous studies [2–5]. A maximum
N2 yield can be obtained at around 360 ◦C (around 78%
at a GHSV of 42,000 h−1), the selectivity toward N2 be-
ing complete. Reducing the contact time of the reactants by
five times (GHSV = 210,000 h−1) does not result in a cor-
responding decrease in activity. Even under this extremely
high flow rate, Fe/ZSM5 still provides a N2 yield of around
36% (Fig. 2b), coupled by a complete selectivity toward
N2. While these results appear very promising for practical
applications of Fe/ZSM5, further tests are needed. In partic-
ular, the stability of the catalyst should be established under
more realistic conditions, i.e., in the presence of water. As
shown in our previous work [11], the presence of water at
high temperature can be responsible for aggregation of iron
and formation of unreactive hematite crystals on the exter-
nal surface of the zeolite. This phenomenon could result in
severe deactivation of the catalyst with time on stream.
4.2. Structure of the Fe-binuclear complexes and reactivity
during heat treatment in He and O2
XAFS spectroscopy allowed us to describe the structure
of the Fe-binuclear complexes in Fe/ZSM5, prepared by
sublimation of FeCl3. This was achieved by studying a sam-
ple calcined with a controlled procedure, aimed at minimiz-
ing the formation of (inactive) spectators [11,14]. It was also

174
A.A. Battiston et al. / Journal of Catalysis 218 (2003) 163–177
(a)
(b)
(c)
(d)
Fig. 8. (a) Normalized XANES spectra and background-subtracted preedges of Fe/ZSM5-deNO 350/350 (dashed line), Fe/ZSM5-He 350/350 (solid
line), and Fe/ZSM5-O 350/350 (dotted line). (b) Preedge characteristics (centroid position and integrated intensity) of Fe/ZSM5-deNO 350/350 (a),
2
1
Fe/ZSM5-He 350/350 (!), and Fe/ZSM5-O 350/350 ("). For details on Fe references, see caption to Fig. 3b. (c) Fourier transform (k ,
2
−1
ꢀk = 2.7–13.2 Å ) of EXAFS data of Fe/ZSM5-deNO 350/350 (dashed line), Fe/ZSM5-He 350/350 (solid line), and Fe/ZSM5-O 350/350 (dotted line).
2
1
−1
(d) Fourier transform of the Fe–O contribution obtained by R-space analysis (k , ꢀk = 2.7–13.2 Å ; ꢀR = 0.7–4.2 Å) of EXAFS data of Fe/ZSM5-deNO
1
350/350 (dashed line), Fe/ZSM5-He 350/350 (solid line), and Fe/ZSM5-O 350/350 (dotted line).
2
possible, as shown in a previous paper [14], to rationalize the
reactivity of the Fe-binuclear clusters during heat treatments
in He and O2 (see Fig. 1). The reactivity of oxygen in the
complexes has been identified in the Fe–O1 and Fe–O3 shells
located in the R range (1.85–1.89 Å) and (2.02–2.09 Å),
respectively. Shell Fe–O2 (1.97–1.99 Å), which remained
unmodified during the treatments, has been ascribed to oxy-
gen from the aluminum-centered tetrahedra, anchoring the
Fe atoms to the framework of the zeolite. Removal of one
oxygen atom from shell Fe–O3 occurs already at low tem-
perature (∼ 150 ◦C) and does not affect the oxidation state
of iron. This has therefore been assigned to desorption of
water. Shell Fe–O1, on the contrary, appears to be affected
only at higher temperatures (above 250 ◦C) and is strongly
influenced by the composition of the gas phase. Approxi-
mately one oxygen atom can be removed, and only in the
complete absence of oxygen from the gas phase. Removal
of this oxygen results, in reduction of iron. The reactivity of
this shell has been ascribed to the presence of highly reac-
tive Fe–O–Fe bridges in the Fe-binuclear complexes. This
background information will be used as a basis for a further
investigation of the active coordination site(s) of the binu-
clear Fe complexes. The influence of the different reactants
on these coordination sites will be discussed.
4.3. Treatment with isobutane
Fe/ZSM5 was treated with isobutane upon heating to
350 ◦C in He. The influence of isobutane on the binuclear
Fe complexes can therefore be tracked by comparing the
results obtained by measurements Fe/ZSM5-CHx 350/350
and Fe/ZSM5-He 350/350. The average oxidation state of
iron for the two measurements appear similar, being respec-
tively around 2.2 and 2.3 (Figs. 4a and 4b). It can there-

A.A. Battiston et al. / Journal of Catalysis 218 (2003) 163–177
175
fore be concluded that treating with isobutane results in
only a slight further reduction of iron, when compared to
that caused by heating in He to 350 ◦C (auto-reduction).
Thus, the majority of reactive oxygen can be activated and
desorbed from the Fe-binuclear complexes by merely heat-
ing. The slight increase in the reduction upon reaction with
isobutane corresponds to a further removal of oxygen from
the closest Fe–O shell (Fe–O1), as shown by the decrease of
its coordination number (from 1.2 in Fe/ZSM5-He 350/350
to 1.0 in Fe/ZSM5-CHx 350/350, Tables 4 and 5). It is
therefore ascribed to the extraction of oxygen from the
Fe–O–Fe bridges. The slight decrease in the preedge inten-
sity (Fig. 4b), together with the increase in the coordination
number of shell Fe–O3, suggests the presence on iron of C-
containing adsorbed species from isobutane. After cooling
to LN temperature, the coordination number of shell Fe–O3
increased significantly (Tables 5 and 6), as a result of further
adsorption resulting from the cooling. The treatment in i-
C4H10 did not cause any change in shell Fe–O2, attributed to
oxygen in Fe–O–Al bridges (see Fig. 1). This shows that the
Fe complexes remained firmly ligated to the zeolite frame-
work. Indeed, no increase was visible in the Fe–Fe coordina-
tion, demonstrating that no further aggregation phenomena
of iron occurred during the treatment.
where ꢀ stands for a vacancy. The N2O formed, present
in negligible concentration in the gas outlet, could not have
been detected by our gas-analysis system.
Contemporarily with the reoxidation of iron, an increase
was visible in the number of Fe-neighboring atoms in the
Fe–O1 shell (from 1.2 in Fe/ZSM5-He 350/350 to 1.7 in
Fe/ZSM5-NO 350/350). When compared to the measure-
ment performed in the presence of oxygen, the shell was not
completely regenerated (1.7 neighbors for Fe/ZSM5-NO
350/350 against 1.9 for Fe/ZSM5-O2 350/350). As previ-
ously discussed, EXAFS is not able to distinguish between
different light-scattering atoms (O or N). The neighboring
atoms in shell Fe–O1 could therefore be attributed to oxygen
(partial regeneration of Fe–O–Fe bridges) or to the presence
of N-containing adsorbed species. Several FTIR studies have
been performed to identify the species formed by the interac-
tion of NO with the active sites of iron catalysts [3,4,6,8,13,
29–33]. Chen et al. [8], in particular, have studied the evolu-
tion of the N-containing adsorbed species generated by NO
on over-exchanged Fe/ZSM5, obtained by chemical vapor
deposition of FeCl3. From their study different species have
been identified as nitrosyl Fen+(NO)1,2 and Fen+(NO2), the
latter being formed with increasing time on stream.
These species, which tend to be formed more easily on
reduced iron (Fe2+) (2), may exist as well on Fe3+ [34] (3):
4.4. Treatment with NO and NO + O2
ꢀ
ꢁ
2NO + HO–Fe2+–ꢀ–Fe2+–OH
The catalytic measurements performed during the collec-
tion of the XAFS spectra of Fe/ZSM5-NO 350/350 do not
show any detectable activity for the catalytic decomposition
of NO by Fe/ZSM5 (Fig. 5).
O
O
|
|
N
N
|
|
ꢀ
ꢁ
→ HO–Fe2+–ꢀ–Fe2+–OH ,
(2)
(3)
Flushing NO results in a clear increase in the oxidation
state of iron, from 2.3 (Fe/ZSM5-He 350/350, Fig. 3b) to
approximately 2.8–2.9 (Fig. 6b). The reason for the reoxida-
tion could be manifold. As shown in our previous work [14]
reduced Fe-binuclear complexes are highly unstable. Indeed,
partial reoxidation of the complexes was detected with time
on stream under He. Voskoboinikov et al. [7] suggested the
possibility of oxygen exchangeability between the Fe com-
plexes and the zeolite matrix. Reoxidation could have there-
fore been caused by extraction of oxygen from the zeolite
lattice. Traces of oxygen in the gas phase may also have
been responsible for the reoxidation. Indeed, as previously
demonstrated [14], the oxygen vacancy between the two
iron ions is readily regenerated by molecular oxygen from
the gas phase. Nevertheless, oxidation by oxygen is not the
only possible explanation. Fu et al. [27] have shown that NO
can act as an oxidant in Fe/Y, where also the presence of
binuclear Fe complexes has been suggested. Lei et al. [28]
have demonstrated that on [Cu–O–Cu]2+ ions in Cu/ZSM5
the same phenomenon occurs by the interaction of two NO
molecules and the generation of one molecule of N2O. The
same pathway may occur on the Fe-binuclear complexes in
Fe/ZSM5:
ꢀ
ꢁ
2NO + HO–Fe3+–O–Fe3+–OH
O
O
|
|
N
N
|
|
ꢀ
ꢁ
→ HO–Fe3+–O–Fe3+–OH .
Fe3+, as previously discussed, may have been gener-
ated by reoxidation with oxygen, or by NO, following path-
way (1).
Together with the partial regeneration of the Fe–O–Fe
bridges, via the pathway proposed by Lei et al. [28], nitrosyl
species appear to be the most probable responsible for the
changes affecting the closest coordination of iron (Fe–O1)
under NO at 350 ◦C.
While Fe/ZSM5 is not able to decompose NO to N2 and
O2, it is an active catalyst for the oxidation of NO to NO2
in the presence of excess oxygen (see Fig. 5). The capability
of Fe/ZSM5 to oxidize NO to NO2 is well known. Catalytic
tests performed by Chen et al. [8], using a feed composition
similar to the one used in this work, showed a maximum
conversion of NO to NO2 of around 75% at 300 ◦C under
a GHSV of 42,000 h−1. As suggested by the authors, NO2
is most likely formed by the reaction of NO with the bridg-
ing oxygen in Fe–O–Fe. The vacancy generated should be
reoccupied by oxygen from the gas phase, or possibly by
ꢀ
ꢁ
2NO + HO–Fe2+–ꢀ–Fe2+–OH
ꢀ
ꢁ
→ HO–Fe3+–O–Fe3+–OH + N2O↑,
(1)

176
A.A. Battiston et al. / Journal of Catalysis 218 (2003) 163–177
NO, following pathway (1). FTIR studies [8,31] have shown
that prevalently nitro (NO2)− and nitrate (NO3)− species
are formed when NO and O2 are fed together on Fe/ZSM5.
Different pathways have been suggested for the formation
of these species. Lobree et al. [31] have proposed, as the
first step in this process, the formation of a superoxide ion
(O2)−, generated by the reaction of O2 with the Fe com-
plexes, followed by a disproportionation of the iron atoms in
the Fe complexes, with the formation of a [HO–Fe2+] and a
[O2−–Fe3+–OH] group. Nitrate and nitro species should be
formed by the subsequent reaction of NO and NO2 from the
gas phase, respectively with the [O2−–Fe3+–OH] and the
[HO–Fe2+] groups. Chen et al. [8] have proposed a differ-
ent pathway, via the formation of a peroxo bridge between
the two Fe atoms in the complexes. The reaction of NO and
NO2 with the highly unstable peroxo group would generate
the nitro and nitrate species.
The immediate reoxidation of iron (Fig. 6a) occurred
upon contemporarily purging of NO and O2 shows that, in-
deed, the Fe-binuclear complexes in Fe/ZSM5 are highly
reactivity towards this mixture. The oxidation state of iron
appears to be somewhat higher than that measured in the
presence of NO alone. This observation, coupled with the
increase in the coordination number of shell Fe–O1, which
appears almost identical to that measured in the presence of
oxygen alone (Fe/ZSM5-O2 350/350), suggests a complete
regeneration of the Fe–O–Fe bridges. The possible forma-
tion of highly unstable peroxo bridges, suggested by Chen
et al. [8] could not have been detected by the EXAFS setup
used in this study. It should be noted that the increment in the
coordination number of shell Fe–O1 could also be assigned
to the presence of absorbed nitro and nitrate species. A clear
assignment with the present data is not possible.
companied by a complete reoxidation of iron (Fe3+). This
result confirms that the reactivity of the Fe complexes orig-
inates in the oxygen vacancies formed upon heating. Due
to the inability of EXAFS to distinguish between light scat-
terers of different nature (C, N, O), it is not possible to
assign the changes in the Fe–O1 coordination to an inter-
action with a specific reactant. As previously observed, the
increase in the coordination number of shell Fe–O1 can in
principle be ascribed to the regeneration of the Fe–O–Fe
bridges, operated by O2 or by NO/NO2 from the gas phase,
following pathway (2). Nevertheless, the reaction pathway
for the HC-SCR reaction, elucidated by Chen et al. [3,4]
and Gao et al. [6], and presented in the introduction, shows
that an extremely wide range of reaction intermediates may
coexist on the binuclear Fe complexes during reaction. The
most plausible candidates consist of organic nitro and ni-
troso compounds. The presence during the SCR reaction of
different adsorbed species in the coordination sphere of iron
is confirmed by the large increase detected in the static dis-
order in the most distant Fe–O shell of the Fe complexes
(Fe–O3), and by the lowered intensity of the Fe preedge
when compared to that measured in the presence of oxygen
alone (Fig. 8b). The intensity of the preedge appears to be
consistent with the presence of five to six atomic neighbors.
Coherently, 5.0 neighbors were found by the EXAFS analy-
sis. It should nevertheless be noted that, due to the extremely
high static disorder originated by the different reaction in-
termediates, this number could be slightly underestimated.
Indeed, upon cooling to 77 K, a slight increase was detected
in the coordination number of shells Fe–O1 and Fe–O3.
The increase in the coordination number of shell Fe–O3, in
particular, could be attributed as well to adsorption at low
temperature of water produced by combustion of isobutane
during reaction.
As in the case of the treatment with isobutane and NO,
also in the contemporary presence of NO and O2 no sig-
nificant changes were detected in shell Fe–O2, attributed to
oxygen binding the Fe complexes to the zeolite framework,
nor in the Fe–Fe coordination. It can be concluded that also
in the presence of NO and NO +O2 the binuclear complexes
remained stably ligated to the framework and that no ag-
glomeration of iron toward larger clusters occurred.
5. Conclusions
The majority of iron in Fe/ZSM5 obtained by sublima-
tion of FeCl3, and calcined by using a special procedure
aimed at minimizing the number of (inactive) Fe specta-
tors, consists of oxo/hydroxo binuclear Fe complexes with
a Fe–O–Fe core.
4.5. SCR of NO with isobutane
During heat treatment in He to 350 ◦C the Fe complexes
undergo auto-reduction. This is ascribed to removal of oxy-
gen from the Fe–O–Fe bridges (closest Fe–O coordination
sphere), with the consequent formation of Fe–ꢀ–Fe vacan-
cies.
The majority of reactive oxygen can be activated and
desorbed from the Fe-binuclear complexes by merely heat-
ing. Treatment with isobutane results in only a further slight
(average) reduction of iron, accompanied by an additional
removal of oxygen from the Fe–O–Fe bridges.
The activity of Fe/ZSM5 towards the selective catalytic
reduction of NO to N2 with isobutane was confirmed by the
analysis of the outlet gases (Fig. 7), performed during the
collection of the XAFS spectra Fe/ZSM5-deNO 350/350.
The gas analysis shows that the XAFS spectra were recorded
on Fe/ZSM5 under deNO working conditions.
As in the case of the treatments with NO or NO + O2,
changes occurring upon reaction with the HC-SCR mix-
ture were identified in the closest Fe–O coordination sphere
of the complexes. The coordination number of shell Fe–O1
(N = 1.8) appeared to be significantly increased when com-
pared to the sample heated in He (N = 1.2). This was ac-
The active sites of the Fe complexes are located in the va-
cancies generated in the closest Fe–O coordination sphere
by the oxygen desorption. Upon reaction with NO and

A.A. Battiston et al. / Journal of Catalysis 218 (2003) 163–177
177
NO + O2 these vacancies are promptly reoccupied and iron
is oxidized. This can be ascribed to the regeneration of the
Fe–O–Fe bridges in the complexes, possibly accompanied
by adsorption of N-containing species.
Also during HC-SCR conditions the closest Fe–O shell
is reoccupied, leading to reoxidation of Fe (Fe3+). Contem-
porarily, a large increase is detected in the static disorder
in the most distant Fe–O shell of the Fe complexes. This
suggests that adsorbates (HC-SCR reaction intermediates)
are formed on these sites. No conclusions can nevertheless
be derived from the EXAFS analysis on the nature of these
species.
[9] A.V. Kucherov, M. Shelef, J. Catal. 195 (2000) 112.
[10] P. Marturano, L. Drozdová, G.D. Pirngruber, A. Kogelbauer, R. Prins,
Phys. Chem. Chem. Phys. 3 (2001) 5585.
[11] A.A. Battiston, J.H. Bitter, F.M.F. de Groot, A.R. Overweg, O. Ste-
phan, J.A. van Bokhoven, P.J. Kooyman, C. van der Spek, G. Vankó,
D.C. Koningsberger, J. Catal. 213 (2003) 251.
[12] P. Marturano, L. Drozdová, A. Kogelbauer, R. Prins, J. Catal. 192
(2000) 236.
[13] L.J. Lobree, I.-C. Hwang, J.A. Reimer, A.T. Bell, J. Catal. 186 (1999)
242.
[14] A.A. Battiston, J.H. Bitter, W.M. Heijboer, F.M.F. de Groot, D.C. Ko-
ningsberger, J. Catal., in press.
[15] T.V. Voskoboinikov, H.-Y. Chen, W.M.H. Sachtler, Appl. Catal. B 19
(1998) 279.
[16] M. Vaarkamp, B.L. Mojet, F.S. Modica, J.T. Miller, D.C. Konings-
berger, J. Phys. Chem. 99 (1995) 16067.
[17] M. Wilke, F. Farges, P.-E. Petit, G.E. Brown Jr., F. Martin, Am.
Miner. 86 (2001) 714.
Acknowledgments
[18] M. Vaarkamp, J.C. Linders, D.C. Koningsberger, Phys. B 209 (1995).
[19] M. Vaarkamp, I. Dring, R.J. Oldman, E.A. Stern, D.C. Koningsberger,
Phys. Rev. B 50 (1994) 7872.
[20] D.C. Koningsberger, B.L. Mojet, G.E. van Dorssen, D.E. Ramaker,
Top. Catal. 10 (2000) 143.
[21] J.W. Cook, D.E. Sayers, J. Appl. Phys. 52 (1981) 5024.
[22] A.L. Ankudinov, J.J. Rehr, Phys. Rev. B 56 (1997) R1712.
[23] J. Iball, C.H. Morgan, Acta Crystallogr. 23 (1967) 239.
[24] R.L. Blake, R.E. Hessevick, T. Zoltai, L.W. Finger, Am. Miner. 51
(1966) 123.
We kindly thank N. Haack and M. Hermann (station X1.1
at Hasylab, Hamburg) for the help provided during the X-ray
absorption measurements. The European Union program for
Large Scale Facilities (Contract No. ERBFMGECT950059)
and NWO (Contract No. 326-710) are acknowledged for fi-
nancial support.
[25] D.C. Koningsberger, Jpn. J. Appl. Phys. 32 (1993) 877.
[26] G.G. Li, F. Bridges, C.H. Booth, Phys. Rev. B 52 (1995) 6332.
[27] C.M. Fu, M. Deeba, W.K. Hall, Ind. Eng. Chem. Prod. R&D 19 (1990)
299.
[28] G.D. Lei, B.J. Adelman, J. Sárkány, W.M.H. Sachtler, Appl. Catal. B 5
(1995) 245.
[29] G. Busca, V. Lorenzelli, J. Catal. 72 (1981) 303.
[30] M.D. Amiridis, F. Puglisi, J.A. Dumesic, W.S. Millman, J. Topsøe,
J. Catal. 142 (1993) 572.
[31] L.J. Lobree, I.-C. Hwang, J.A. Reimer, A.T. Bell, Catal. Lett. 63
(1999) 233.
References
[1] E.M. El-Malki, R.A. van Santen, W.M.H. Sachtler, J. Catal. 196 (2000)
212.
[2] H.-Y. Chen, W.M.H. Sachtler, Catal. Today 42 (1998) 73.
[3] H.-Y. Chen, T. Voskoboinikov, W.M.H. Sachtler, J. Catal. 180 (1998)
171.
[4] H.-Y. Chen, T. Voskoboinikov, W.M.H. Sachtler, Catal. Today 54
(1999) 483.
[5] H.-Y. Chen, X. Wang, W.M.H. Sachtler, Appl. Catal. A 194–195
(2000) 159.
[32] K. Hadjiivanov, H. Knözinger, B. Tsynstsarski, L. Dimitrov, Catal.
Lett. 62 (1999) 35.
[33] K. Hadjiivanov, J. Saussey, J.L. Freysz, J.C. Lavalley, Catal. Lett. 52
(1998) 103.
[6] Z.-X. Gao, Q. Sun, W.M.H. Sachtler, Appl. Catal. B 33 (2001) 9.
[7] T. Voskoboinikov, H.-Y. Chen, W.M.H. Sachtler, J. Molec. Catal. A
155 (2000) 155.
[8] H.-Y. Chen, E.M. El-Malki, X. Wang, R.A. van Santen, W.M.H.
Sachtler, J. Molec. Catal. A 162 (2000) 159.
[34] K.-I. Segawa, Y. Chen, J.E. Kubsh, W.N. Delgass, J.A. Dumesic, W.K.
Hall, J. Catal. 76 (1982) 112.