Selective reduction of NO with Fe-ZSM-5 catalysts of low Fe content: Part II. Assessing the function of different Fe sites by spectroscopic in situ studies

Article abstract of DOI:10.1016/j.jcat.2006.01.024

A series of Fe-ZSM-5 catalysts prepared by improved liquid ion exchange (see part I [J. Catal. 231 (2005) 314]) containing 0.2-1.2 wt% Fe, with a systematically changing nature of Fe sites, was studied during the selective catalytic reduction (SCR) of NO with NH₃ or isobutane or during the interaction with feed components by various in situ methods (EPR, UV-vis, and FTIR spectroscopy). The results were related to the catalytic behavior. Several types of isolated Fe³⁺ sites with different reducibility were identified. FTIR results revealed that reduction of NO proceeds via intermediate formation of adsorbed nitrato species, which are subsequently reduced. Their formation requires the presence of Fe³⁺. Those Fe³⁺ ions that are permanently reduced to Fe²⁺ under reaction conditions probably do not contribute to catalytic activity. In general, the degree of steady-state Fe site reduction during NH₃ SCR is markedly lower than that during isobutane SCR, which may be a reason for the lower activity of Fe-ZSM-5 in the latter reaction. UV-vis and EPR results suggest that isolated Fe³⁺ sites in octahedral coordination are more easily reduced than tetrahedral Fe³⁺. In contrast to some types of isolated Fe ³⁺ species, Fe_xO_y clusters reoxidize much more rapidly than they reduce, and thus remain essentially trivalent under reaction conditions. Due to their higher oxidation potential, they cause undesired total oxidation of the reductant, particularly in the case of isobutane.

Full text of DOI:10.1016/j.jcat.2006.01.024

Journal of Catalysis 239 (2006) 173–186
www.elsevier.com/locate/jcat
Selective reduction of NO with Fe-ZSM-5 catalysts of low Fe content:
Part II. Assessing the function of different Fe sites by spectroscopic
in situ studies
a
b
b
a
a,∗,1
M. Santhosh Kumar , M. Schwidder , W. Grünert , U. Bentrup , A. Brückner
a
Institut für Angewandte Chemie Berlin-Adlershof e. V., Richard-Willstätter-Str. 12, D-12489 Berlin, Germany
b
Lehrstuhl für Technische Chemie, Ruhr-Universität Bochum, D-44780 Bochum, Germany
Received 18 November 2005; revised 18 January 2006; accepted 20 January 2006
Abstract
A series of Fe-ZSM-5 catalysts prepared by improved liquid ion exchange (see part I [J. Catal. 231 (2005) 314]) containing 0.2–1.2 wt% Fe,
with a systematically changing nature of Fe sites, was studied during the selective catalytic reduction (SCR) of NO with NH or isobutane or
3
during the interaction with feed components by various in situ methods (EPR, UV–vis, and FTIR spectroscopy). The results were related to the
3+
catalytic behavior. Several types of isolated Fe sites with different reducibility were identified. FTIR results revealed that reduction of NO
3
+
proceeds via intermediate formation of adsorbed nitrato species, which are subsequently reduced. Their formation requires the presence of Fe
.
3+
2+
Those Fe ions that are permanently reduced to Fe under reaction conditions probably do not contribute to catalytic activity. In general, the
degree of steady-state Fe site reduction during NH SCR is markedly lower than that during isobutane SCR, which may be a reason for the lower
activity of Fe-ZSM-5 in the latter reaction. UV–vis and EPR results suggest that isolated Fe sites in octahedral coordination are more easily
reduced than tetrahedral Fe . In contrast to some types of isolated Fe species, FexOy clusters reoxidize much more rapidly than they reduce,
and thus remain essentially trivalent under reaction conditions. Due to their higher oxidation potential, they cause undesired total oxidation of the
reductant, particularly in the case of isobutane.
3
3+
3+
3+
©
2006 Elsevier Inc. All rights reserved.
Keywords: DeNOx; Fe-ZSM-5; Isobutane; Ammonia; In situ-EPR; In situ-UV/VIS spectroscopy; In situ-FTIR spectroscopy; Active sites
1
. Introduction
preparation method and the Fe content, various coexisting Fe
species are created, ranging from isolated Fe ions via dimers
and small oligonuclear FexOy clusters inside the pores to large
Fe2O3 particles on the external surface.
Recently Fe-ZSM-5 zeolites containing nonframework iron
species have become recognized as highly efficient catalysts for
a number of reactions, including selective oxidation of ben-
zene to phenol [1,2], N O decomposition [3], and selective
catalytic reduction (SCR) of NOx by NH3 [4] and hydrocar-
bons [5,6]. But despite extensive studies of structure–reactivity
relationships by various characterization techniques, debate on
the nature of active Fe sites in these catalysts is ongoing. This
is due mainly to the fact that in most cases, depending on the
The inhomogeneous distribution of Fe species is a major
problem hindering a definitive identification of active sites.
A possible way to overcome this difficulty would be the ded-
icated synthesis of Fe-ZSM-5 materials with Fe species of ho-
mogeneous and defined nature. With Fe content as low as 0.5%,
it has been possible to make some significant strides toward
this goal. Fe silicalite prepared by isomorphous substitution
and subsequent steam treatment [7] and Fe-ZSM-5 obtained
by a mechanochemical route [8] have been shown to contain
predominantly isolated Fe species and to have high catalytic
2
*
Corresponding author. Fax. +49 30 6392 4454.
E-mail address: brueckner@aca-berlin.de (A. Brückner).
1
performance in the SCR of N O by CO and in the SCR of NO
Present address: Leibniz-Institut für Katalyse an der Universität Rostock
e. V., Branch Berlin (former ACA), P.O. Box 96 11 56, D-12474 Berlin.
2
by both isobutane and NH3. Based on these findings, it has been
0021-9517/$ – see front matter © 2006 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2006.01.024

174
M. Santhosh Kumar et al. / Journal of Catalysis 239 (2006) 173–186
proposed that isolated Fe sites may play a major role in the cat-
alytic cycle.
just spectator species, because their disappearance has not been
reported to lead to decreased catalytic activity.
Selective analysis of the role of FexOy clusters requires a
catalyst without isolated Fe sites, the synthesis of which once
seemed impossible. However, recently we succeeded in prepar-
ing a series of Fe-ZSM-5 catalysts with a systematically de-
creasing dispersion of Fe sites [9,10] using a variation of the im-
proved liquid ion exchange (ILIE) method introduced by Long
and Yang [11]. The nature of the Fe sites in these catalysts has
been studied by several techniques, including X-ray absorption
spectroscopy (XAFS), electron paramagnetic resonance (EPR),
UV–vis spectroscopy, temperature-programmed reduction (H2-
TPR), and transmission electron microscopy (TEM); the results
have been described in detail in part I of this study [10]. It was
found that in samples with 0.2–0.3% Fe, >95% of the total
iron was present in form of isolated sites. At intermediate Fe
content (0.6–0.7%), the percentage of additional oligonuclear
intrazeolitic clusters became substantial, whereas the amount
of large oxide particles remained negligible. The latter species
were formed in reasonable amounts only at the highest Fe con-
tent of 1.2%. These conclusions are based mainly on a com-
bined evaluation of deconvoluted UV–vis spectra showing a
red shift of subbands with increasing degree of Fe site ag-
glomeration and of temperature-dependent EPR measurements
Moreover, little is known about whether the different coor-
dination states of Fe species in Fe-ZSM-5 gives rise to changes
in reduction and reoxidation behavior, which should be a cru-
cial property in view of these species’ participation in the cat-
alytic redox cycle. According to Kucherov et al. [11], there is
no doubt that the formation of ferrimagnetic oxidic agglom-
erates, which is favored by reducing conditions at high tem-
perature, is detrimental to catalytic activity [11]. Wichterlova
and co-workers [14] identified three different sites for isolated
metal ions in ZSM-5: the straight channels (α), the intersection
between the straight and sinusoidal channels (β), and a boat-
shaped site in the sinusoidal channel (γ ) [14]. Although this
information was originally derived for Co2 ions from UV–vis
data, it has been analogously adopted for Fe ions as well [15].
It was found that Cu and Co ions are most easily reduced when
located in α positions and are highly resistant to reduction when
in γ positions [16]. Whether the same is true for Fe ions has not
been explicitly stated. Moreover, how the observed EPR signals
relate to any of these positions remains unknown.
In this work we studied a series of Fe-ZSM-5 catalysts
prepared by ILIE and found them to contain almost only iso-
lated Fe sites in tetrahedral and octahedral coordination for low
Fe content, small oligonuclear FexOy clusters besides isolated
sites for medium Fe content, and additional Fe2O3 particles for
the highest Fe content [10]. The goal of our study was to eluci-
date how the differently coordinated isolated sites, as well as the
clusters of different sizes, interact with reactants of the NH3 and
isobutane SCR feeds, using in situ methods able to monitor the
Fe species (EPR, UV/vis-DRS) as well as surface adsorbates
formed on reaction (FTIR spectroscopy). In particular, we ex-
plored the reactivity of the various Fe species in both NH3 SCR
and isobutane SCR to identify reasons for previously detected
differences in activity and selectivity [10]. Beneficial and detri-
mental types of Fe species are discriminated as a basis for future
catalyst design.
+
(
77–773 K), allowing differentiation between isolated and ag-
3+
glomerated Fe species by their magnetic behavior [10]. The
formation of Fe2O3 particles in the sample with 1.2% Fe was
also confirmed by TEM [10]. This series of samples made it
possible to derive a unified concept of the active sites in the
SCR of NO demonstrating that isolated Fe sites play major
roles in both reactions and that oligomers also contribute, but
in differing ways depending on the reducing agent. A detailed
discussion of the nature of Fe sites and their relationship to cat-
alytic performance can be found in part I of this work [10].
It must be mentioned, however, that even in samples with
almost exclusively isolated Fe ions there exist at least three
different kinds of these sites, which have been discriminated
by EPR spectroscopy on the basis of their effective g val-
2
. Experimental
ꢀ
ues (g ≈ 6, 4.3, and 2). It remains an open question as to
whether all of these participate equally in the catalytic cycle.
Spectroscopic in situ studies at elevated temperature and in
ꢀ
Based on the observation that signals at g ≈ 5.6 and 6.5 in
the presence of reactants were performed with Fe-ZSM-5 sam-
ples described in part I of this study [10] labeled Fe-Z(0.2),
Fe-Z(0.3), Fe-Z(0.7), and Fe-Z(1.2) (with the Fe iron content
in wt% in parentheses). The catalysts were prepared by ILIE as
described in part I [10]. In brief, a commercial Na-ZSM-5 and
Fe powder was stirred in 0.1 M HCl under a protective gas at-
mosphere for 5 days, washed with deionized water (separating
residual iron with a magnet), dried, and calcined in air at 873 K
for 2 h.
In situ EPR spectra in X-band (ν ≈ 9.5 GHz) were recorded
with a cw-spectrometer (ELEXSYS 500-10/12; Bruker) us-
ing a microwave power of 6.3 mW, modulation frequency of
100 kHz, and modulation amplitude of 0.5 mT. The magnetic
field was measured with respect to the standard 2,2-diphenyl-1-
picrylhydrazyl hydrate (DPPH). A homemade quartz plug-flow
reactor connected to a gas-dosing system containing mass flow
Fe-ZSM-5 prepared by sublimation lose intensity on treatment
with NO/O2/C3H6 mixtures of increasing propene percentage
at 773 K, Kucherov et al. [11] suggested that Fe ions reflected
by these lines might be coordinatively unsaturated active sites.
Likewise, Ribera et al. [12] suggested that Fe species evidenced
ꢀ
by signals at g ≈ 6.4 and 5.7 in steam-activated Fe-ZSM-5
participate in the redox process, because they disappeared on
contact with β-mercaptoethanol at room temperature. In a re-
cent study, Kubánek et al. [13] correlated the sum intensity of
the signals at g ≈ 6.0 and 5.6 in fresh Fe-ZSM-5 with the rate
of phenol formation, observed when these samples were used
as catalysts in the direct oxidation of benzene. In the aforemen-
ꢀ
tioned studies it was assumed that signals around g ≈ 6 were
active sites, due to the fact that they decrease more or less in
the presence of reactants. However, it is possible that they are

M. Santhosh Kumar et al. / Journal of Catalysis 239 (2006) 173–186
175
controllers (Bronkhorst) was implemented in the rectangular
cavity of the spectrometer. For all in situ experiments, 50 mg
of catalyst particles (125–200 µm) were used and exposed to
the following sequence of experimental steps: (a) pretreatment
in air (20 ml/min) at 773 K and cooling to 623 K; (b) SCR us-
ing 0.1% NO, 0.1% reducing agent (NH3 or isobutane), and 2%
O2/He for 1 h (GHSV = 30,000 h for both NH3 and isobu-
tane); (c) reoxidation in air (20 ml/min) at 773 K and cooling
to 623 K; (d) treatment in 0.1% NH3/He (34.7 ml/min) or 0.1%
isobutane/He (34.7 ml/min) for 1 h; and (f) treatment in 0.1%
NO/He (34.7 ml/min) for 1 h.
In situ UV/vis-DRS measurements were performed with a
Cary 400 spectrometer (Varian) equipped with a diffuse re-
flectance accessory (Praying Mantis, Harrick) and a heatable
reaction chamber (Harrick). The samples were used as pow-
ders and placed in a sample cup of 5 mm diameter and 3 mm
depth, at the bottom of which is a small sieve. The temperature
was controlled by a thermocouple located on the bottom of the
sample cup and connected to a temperature programmer (Eu-
rotherm). Reactant gases were provided by a gas-dosing system
as mentioned above and led from top to bottom through the
sample layer. The same sequence of experimental steps as in
the in situ EPR experiments was used.
(30 ml/min) for 1 h, with spectra recorded at 623 K at different
time intervals before and after evacuation; and (c) reoxidation
in air (30 ml/min) at 623 K for 1 h.
3. Results and discussion
−
1
3.1. In situ studies in the presence of NH SCR feed
3
components
3.1.1. In situ UV–vis studies
UV–vis spectra of the catalysts used in this study were given
in part I [10] and are summarized here in Fig. 1, to facilitate
discussion of the following data. These spectra are character-
ized by broad, partially overlapping Fe3 ← O charge-transfer
bands (Fig. 1). To facilitate the identification of different Fe
species, the experimental spectra were deconvoluted into sev-
eral subbands, the assignment of which was discussed previ-
ously [17,18] based on earlier proposals [19,20]. In brief, sub-
+
3+
bands around 240 and 290 nm arise from isolated Fe sites in
tetrahedral and higher coordination (five or six oxygen ligands),
respectively, whereas bands at 300 < λ < 400 nm are assigned
to oligomeric clusters and subbands >400 nm are assigned
to large Fe O particles (Table 1). The spectra deconvolution
2
3
Reduction and reoxidation kinetics of the Fe sites were de-
termined by following the absorbance at 241 and 291 nm for
shown in Fig. 1 indicates that the apparent broadening of the
experimental spectrum above 290 nm with increasing Fe con-
tent comes from the contribution of subbands assigned to small
oligomeric clusters and Fe O particles that partially overlap
isolated Fe3 ions in Fe-Z(0.3) and at 350 nm for oligonuclear
FexOy clusters in Fe-Z(1.2) at 673 K as a function of time. Af-
ter pretreatment in air (10 ml/min) at 773 K for 1 h, reduction
was performed at 673 K in a flow of 1% NH3/He (10 ml/min)
for 2 h, followed by reoxidation in air (10 ml/min) for 2 h at the
same temperature. The experimental curves thus obtained were
fitted by a pseudo-first-order rate law using
+
2
3
with the high-wavelength tail of the subband at 290 nm assigned
to isolated Fe3 species in fivefold and/or sixfold coordination.
+
The dividing line between oligomeric clusters and Fe O parti-
2
3
cles may be blurred. Theoretical calculations of UV–vis spectra
of trimeric Fe3 hydroxo complexes in solution revealed the
appearance of CT bands above 400 nm for even this low degree
of aggregation [21]. Nevertheless, there is no doubt that light
absorbance increases in the visible range, as does the tendency
to form FexOy clusters.
+
Abst = Abst=∞ + (Abst=0 − Abst=∞)e−
kt
,
(1)
in which Abst=0, Abst , and Abst=∞ are the absorbance values
at the start of the experiment, at time t, and after reaching steady
state, respectively.
Fig. 1 clearly shows that the Fe-Z(0.2) and Fe-Z(0.3) sam-
ples contained almost only isolated Fe sites in both tetrahe-
dral (241 nm) and higher (probably octahedral) coordination
(291 nm). Exposure of these catalysts to the feed mixture of
NH3 SCR reduced the intensity of both bands due to the re-
−
1
FTIR spectra (2 cm resolution, 100 scans) were recorded
by a Bruker IFS 66 spectrometer equipped with a heatable and
evacuable reaction cell with CaF2 windows and connected to
a gas dosing and evacuation system. The zeolite powder was
pressed into self-supporting wafers with a diameter of 20 mm
and a weight of 50 mg. NO adsorption experiments were per-
formed according to the following sequence: (a) oxidative pre-
treatment in air (30 ml/min) at 673 K for 1 h, followed by
cooling to 293 K; (b) adsorption of 1% NO/He (40 ml/min) for
duction of isolated Fe3 (Fig. 2); this effect was even more
pronounced for Fe species with more than four oxygen ligands
(291 nm). Treatment in NH3 slightly enhanced the reduction
process even more, whereas subsequent treatment in NO flow
led to reoxidation, which was roughly complete for tetrahedral
Fe sites (241 nm) but only partial for octahedrally coordinated
Fe (291 nm).
+
1
h, followed by evacuation for 30 min, with spectra recorded
before and after evacuation; (c) heating in air at 673 K for
h; (d) reductive pretreatment in 2% NH3/He (40 ml/min)
for 1 h, followed by evacuation at the same temperature for
0 min and cooling to 293 K; and (e) adsorption of 1% NO/He
40 ml/min) for 1 h followed by evacuation for 30 min. Spectra
1
With an increasing percentage of oligonuclear clusters and
even Fe2O3 particles in the Fe-Z(0.7) and Fe-Z(1.2) samples,
the effects on treatment in the NH3-SCR feed, NH3, and NO
became less pronounced. In sample Fe-Z(1.2), which contains
more than half of its iron in agglomerated Fe species [10], only
a weak intensity decrease was observed for the CT band around
350 nm. This indicates a markedly different redox behavior of
FexOy clusters and particles, but also suggests that redox prop-
3
(
were recorded before and after evacuation.
In situ FTIR experiments during isobutane SCR used the fol-
lowing treatment sequence: (a) oxidation in air (30 ml/min) at
6
1
73 K for 1 h; (b) cooling to 623 K and flushing by a mixture of
% NO/He (30 ml/min), 1% isobutane/He (30 ml/min), and air
erties of isolated Fe3 sites (reflected by the band at 241 nm)
+

176
M. Santhosh Kumar et al. / Journal of Catalysis 239 (2006) 173–186
Fig. 1. Experimental UV–vis spectra measured at ambient temperature and atmosphere (—) including deconvoluted sub-bands, which are assigned to the following
3+
Fe species: isolated Fe (- - -), oligonuclear FexOy clusters inside pores (—), Fe oxide particles (—).
Table 1
Assignments of spectroscopic signals
UV–vis subbands
λ (nm)
Assignment of Fe bands in as-received samples
Ref.
3
+
≈
240
290
Fe
Fe
O
O
[17–20]
[17–20]
[17–20]
[17–20]
4
4
3+
≈
(x = 1 or 2)
+x
300–400
small oligonuclear FexOy clusters inside pores
Fe O particles in the nm range
>
400 nm
400
2
3
<
C/N-containing deposits in Fe-Z(0.2) during
isobutane-SCR (Fig. 7)
[44,45]
[46,47]
[8,36]
>
400
500
carbon deposits in Fe-Z(0.2) during treatment with isobutane (Fig. 7)
2+
3+
>
Fe → Fe ICVT in Fe-Z(1.2) during treatment with isobutane (Fig. 7)
EPR signals
ꢀ
g
Assignment
Ref.
⎫
⎪
3
+
≈
≈
≈
≈
4.3
6
2
2
isolated Fe O in strong rhombic distortion
4
⎪
⎬
3
+
isolated Fe O in strong axial distortion
6
[8,10,17,41–43]
3
+
isolated Fe Ox in high symmetry
⎪
⎪
⎭
3
+
Fex Oy clusters
FTIR bands
−
1
Wavenumber (cm
)
Assignment
Ref.
+
2
2
1
1
1
1
1
2
1
193
134
745
880, 1575
601
[NO ][N O ]
[23]
[25,28]
[23]
[23,30]
[23]
[26,29]
[29]
2
4
+
NO
N O
2
4
N O
2
3
adsorbed NO
differently bound nitrato species
adsorbed nitrite and ammonium species
C/N-containing species (nitrile, cyanate and/or isocyanate)
carbonyl-containing species of different structure
2
608, 1628, 1618, 1576
437
240
[29]
[29]
690, 1723, 1765

M. Santhosh Kumar et al. / Journal of Catalysis 239 (2006) 173–186
177
Fig. 2. In situ UV–vis spectra measured at 623 K during sequential treatment in a flow of (1) air, (2) 0.1% NO, 0.1% NH , 2% O /He, (3) 0.1% NH /He (after
3
2
3
reoxidation at 773 K) and (4) 0.1% NO/He. Each spectrum measured after 1 h exposure time.
might change when they coexist with dominating clusters in
the same sample. Kucherov et al. [11] has demonstrated that
reason for the smaller changes of the UV band at 241 nm in the
Fe-Z(0.7) and Fe-Z(1.2) samples.
isolated Fe3 can migrate out of position and be incorporated
into FexOy clusters during calcination. Others have shown that
agglomeration of isolated Fe3 into clusters is particularly fa-
vored for easily reducible Fe species [22,23]. Thus, it can be
+
To further analyze the differences in the reduction/reoxida-
tion behavior of the various Fe species, absorbance in the band
maximum representing isolated tetrahedral (241 nm) and iso-
+
lated octahedral Fe3 (291 nm) was followed as a function
of time for sample Fe-Z(0.3), and the absorbance at 350 nm
in sample Fe-Z(1.2) was used to elucidate redox kinetics for
oligonuclear FexOy clusters. The resulting time dependencies
were fitted using the pseudo–first-order rate law of Eq. (1).
As examples, Fig. 3 shows the reduction/reoxidation curves
for the 241 and 350 nm positions in Fe-Z(0.3), and Table 2
summarizes the corresponding rate constants. Note that the ex-
perimental kinetic curves for isolated sites could not be fitted
assuming only one single process. Two processes with differ-
ent rate constants, a faster one (I) and a slower one (II), had
to be superimposed for a satisfactory fit (Table 2), whereby the
difference of the rate constants was larger for reduction. This
+
assumed that in samples with higher Fe content, those Fe3
+
sites may preferentially withstand the tendency to form clusters
that are rather resistant against reduction. This could explain
why the band at 241 nm in Fe-Z(1.2) was by far less sensitive
against redox treatment than those in Fe-Z(0.2) and Fe-Z(0.3).
It also means that different isolated tetrahedral Fe3 sites coex-
ist in Fe-ZSM-5 and contribute to the 241 nm band depending
on the Fe loading, which cannot be distinguished by UV–vis-
DRS. However, as discussed in part I [10], oxide clusters and/or
particles may also contribute to a certain extent to the UV–vis
absorbance below 300 nm, particularly when they show a short-
range order similar to that of γ -Fe O . EXAFS data for these
+
2
3
3+
suggests that at least two kinds of isolated Fe sites (I and II)
samples suggested that the γ -Fe2O3 structure is not typical for
these clusters, but rather the short-range order of the agglomer-
ated species in Fe-Z(1.2) resembles that of α-Fe2O3, for which
a negligible contribution to UV–vis absorbance below 300 nm
was shown [10]. Nevertheless, the possibility that oxide clus-
ters contribute to a certain extent to the band at 241 nm cannot
be completely excluded. Given that the clusters are more dif-
with different reducibility contribute to each of the two bands
below 300 nm. Considering the negligible change of the 241-
nm band in sample Fe-Z(1.2) (Fig. 2), it is likely that this band
arises only from the hardly reducible species II. Generally, it
appears that isolated octahedral Fe3 species I (band at 291
+
nm) are significantly more rapidly reduced but more slowly
reoxidized than isolated tetrahedral Fe3 species I (band at
+
3+
ficult to reduce than isolated Fe , this could be an additional
241 nm) (Table 2). Comparing the rate constants for oligonu-

178
M. Santhosh Kumar et al. / Journal of Catalysis 239 (2006) 173–186
Fig. 3. Time dependence of (A) the absorbance at 241 nm in sample Fe-Z(0.3) and (B) at 350 nm in sample Fe-Z(1.2) during reduction in 1% NH /N and during
3
2
reoxidation in air at 673 K. Fits are superimposed as dashed black lines (reduction) and solid black lines (reoxidation). UV–vis spectra at the start and end of the
reduction as well as at the end of reoxidation at 673 K are also shown.
Table 2
symmetry was not possible. In samples containing FexOy clus-
ꢀ
Pseudo-first order rate constants derived from the time dependence of the UV–
vis absorbance during reduction in 1% NH /N and reoxidation in air
ters and particles, the latter species also contribute to the g ≈ 2
3
2
signal (Table 1) [10,17].
Sample
Species
Treatment
λ
t
k
Switching from air to the SCR feed mixture and further to
NH3 flow in the Fe-Z(0.2) and Fe-Z(0.3) samples gave rise to a
−
2
−1
(nm)
(min)
(10 min )
Fe-Z(0.3)
I
II
I
NH
3
241
0–25
25–120
0–10
10.3
1.2
16.6
15.9
3+
ꢀ
marked decrease in the signals of octahedral Fe (g ≈ 6) and
a lesser decrease in the signals of tetrahedral Fe3 at g ≈ 4.3.
+
ꢀ
Air
241
ꢀ
II
10–120
In contrast, almost no change was observed for the g ≈ 2 line.
Only when the experiment was performed at 773 K did the EPR
Fe-Z(0.3)
Fe-Z(1.2)
I
II
I
NH
3
291
291
0–25
25–120
0–45
17.5
2.4
8.4
8.8
ꢀ
signal at g ≈ 2 decrease, and then only slightly (not shown).
Air
This agrees very well with the UV–vis results described above
(Fig. 2, Table 2), in which a more rapid reduction of isolated oc-
II
45–120
tahedral Fe3 in comparison to tetrahedral Fe was observed.
This finding also indicates differing reduction stabilities for the
various kinds of isolated Fe ions, with those with the highest
+
3+
NH
Air
350
350
0–120
0–120
1.2
3
65.0
ꢀ
reduction resistance reflected by the EPR signal at g ≈ 2. Fur-
clear FexOy clusters (350 nm) with those of isolated Fe3
+
thermore, the fact that the same trend was observed for the
ions clearly shows that the former are markedly more slowly
reduced but much more rapidly reoxidized. Interestingly, the
reduction/reoxidation curves could be fitted assuming one sin-
gle process, suggesting a more uniform nature and/or position
of the clusters compared with isolated Fe sites.
ꢀ
behavior of UV–vis and EPR signals at 241 nm/g ≈ 4.3 and
ꢀ
2
91 nm/g ≈ 6 further supports assignment of the EPR sig-
nals to the aforementioned coordination, which would not be
possible solely from their positions in the magnetic field. On
subsequent treatment in NO/He flow, the changes observed dur-
ing reduction were only partly reversible, whereas complete
restoration of both EPR and UV/vis signals could be achieved in
NO/He at 773 K (not shown). This means that under SCR con-
ditions in the low-temperature range, only a certain percentage
of isolated Fe species remained trivalent, mainly those giving
3
.1.2. In situ EPR studies
EPR spectra of the Fe-Z(0.2) and Fe-Z(0.3) samples show
ꢀ
signals at effective g-values of g ≈ 6, 4.3, and 2, which are
known from earlier studies (Fig. 4) [8,10,17]. Based on their
behavior during dehydration/rehydration experiments [17], the
ꢀ
rise to the g ≈ 2 EPR line. Comparing these with the UV–vis
ꢀ
3+
g ≈ 4.3 and 6 signals were assigned to isolated Fe
in
spectra recorded under the same conditions (Fig. 2) reveals that
these Fe3 sites are reflected by the (less-intense) bands at 241
+
strongly distorted tetrahedral and octahedral coordination, re-
ꢀ
spectively, whereas the g ≈ 2 line in samples containing only
and 290 nm remaining under SCR conditions. This suggests
3
+
ꢀ
isolated Fe was attributed to Fe in sites of high symmetry,
for which a discrimination between tetrahedral and octahedral
that the reduction-resistant EPR signal at g ≈ 2 reflects iso-
lated Fe3 sites in both tetrahedral and higher coordination.
+

M. Santhosh Kumar et al. / Journal of Catalysis 239 (2006) 173–186
179
Fig. 4. In situ EPR spectra measured at 623 K during sequential treatment in a flow of (1) air, (2) 0.1% NO, 0.1% NH , 2% O /He, (3) 0.1% NH /He (after reoxi-
3
2
3
dation at 773 K) and (4) 0.1% NO/He. Each spectrum measured after 1 h exposure time.
Comparing the EPR spectra of sample Fe-Z(0.7) with those
of samples Fe-Z(0.2) and Fe-Z(0.3) recorded under the same
conditions reveals rather similar behavior for the signals at low
magnetic field (Fig. 4). However, it must be noted that the to-
clusters and particles unchanged at +3. This agrees nicely with
the UV–vis results showing slow reduction and rapid reoxi-
dation of agglomerated oxide species (Fig. 2, Table 2). On a
further change to NH /He flow, the intensity of the narrow line
3
ꢀ
ꢀ
tal intensity of the lines at g ≈ 6 and 4.3 in sample Fe-Z(0.7) is
in the g ≈ 2 range decreases, whereas that of the broad line in-
similar to that in samples Fe-Z(0.2) and Fe-Z(0.3), even though
creases. The latter effect is characteristic for the formation of
the iron content is more than twice as high. This clearly shows
that the relative percentage of (more easily reducible) isolated
Fe O -like areas with ferrimagnetic ordering of spins [8,11].
3 4
Because the EPR intensity of ferrimagnetic species is usually
larger than that of paramagnetic species by orders of magni-
3+
ꢀ
Fe species represented by the signals at g ≈ 6 and 4.3 de-
creases with rising iron content. This effect is even more pro-
nounced for sample Fe-Z(1.2), in which these signals are hardly
visible (Fig. 4). This result supports the conclusion derived
from the UV–vis measurements (Fig. 2) that with increasing
ꢀ
tude, the rather small increase of the broad g ≈ 2 subsignal
of sample Fe-Z(1.2) in Fig. 4 suggests that only a few of the
Fe species that form agglomerates (probably those located on
the outer surface of the particles) contribute to the ferrimag-
netic signal. This may be why only a very slight reduction of
the band intensity around 350 nm in the UV–vis spectrum was
observed (Fig. 2). In sample Fe-Z(0.7), no increase, but rather
Fe content only reduction-resistant isolated Fe3 sites may sur-
+
vive.
As mentioned above, oligonuclear Fe O clusters and par-
x
y
ꢀ
ꢀ
ticles also contribute to the EPR signal at g ≈ 2. This is most
a slight decrease, in the intensity at g ≈ 2 occurred on partial
ꢀ
significant for sample Fe-Z(1.2), in which the g ≈ 2 signal is
reduction, probably because this sample contains only oligonu-
much more intense than in the other samples and represents
a superposition of at least two contributions, an intense broad
line and a less-intense narrower line. The two signals in the
clear FexOy clusters inside the pores that are partially reduced
ꢀ
on NH /He treatment (intensity decrease at g ≈ 2). Obviously,
3
these species are not large enough to demonstrate ferrimagnetic
ordering of sufficient extension.
ꢀ
g ≈ 2 range may arise from oxidic clusters of different sizes.
3+
ꢀ
Isolated Fe sites may contribute to the line at g ≈ 2 as well,
Switching from NH3/He to NO/He flow causes a marked in-
ꢀ
but these cannot be distinguished from the cluster signals in
crease of the g ≈ 2 line in sample Fe-Z(0.7) that may arise from
this case. Switching from air to NH SCR feed has virtually
ferrimagnetic ordering of clusters grown on reduction. Based
on the knowledge that reductive treatment can favor the ag-
glomeration of Fe sites [11,22,23], this signal likely arises from
3
no effect on the EPR spectra, suggesting that cyclic reduc-
tion/reoxidation keeps the mean valence state of the iron oxide

180
M. Santhosh Kumar et al. / Journal of Catalysis 239 (2006) 173–186
Fig. 5. In situ FTIR spectra after adsorption of NO at 293 K on (a) H-ZSM-5 and (b) Fe-Z(1.2) after preoxidation (1 h at 673 K in air, —) and prereduction (1 h at
6
73 K in 2% NH /He, ·····) before evacuation of the IR cell. Difference spectra are shown.
3
partially reoxidized Fe3O4-like species that have agglomerated
during reduction in NH3/He flow. A similar explanation can be
and
+
+
NO + NO2 + 2H → 2NO + H2O.
Because no bands for purely ionic nitrate (around 1380 cm
[31]) were detected in this study, NO is most likely formed via
Eq. (3) by the reaction of NO and NO2 with Brønsted sites [32].
This is also supported by changes in the OH region (not shown).
(3)
ꢀ
given for the increase of the narrow g ≈ 2 subsignal in sam-
−1
ple Fe-Z(1.2) during NO/He treatment even above its original
value (Fig. 4).
+
The observation that Fe species reflected by EPR signals
ꢀ
around g ≈ 6 are preferentially reduced was previously made
−
1
by others [11,12] and taken as an indication that these sites de-
The OH bands around 3610 cm disappeared after NO adsorp-
tion, whereas the Si–OH bands around 3740 cm remained
termine the catalytic activity. When this is true, Fe2 should
+
−1
3+
interact with NO equally well or even better than Fe . In
contrast to Cu-ZSM-5, in which adsorption of NO on reduced
diamagnetic Cu centers could be detected by EPR [24], this
would not be expected for either Fe or Fe , because both of
these ions are paramagnetic. Dipolar interaction with paramag-
netic NO would render the EPR signal undetectable. Therefore,
the adsorption of NO was studied by in situ FTIR spectroscopy
unaffected.
The intense band at 1880 cm 1 (Fig. 5) can, in principle,
arise from two sources: (a) NO adsorbed on an Fe site, whereby
higher band intensities have been found for prereduced Fe-MFI
[22,25] and (b) N2O3 [32]. In the latter case, another equally
−
+
3+
2+
−
1
intense band at around 1575 cm must be found that also be-
longs to N2O3. This is true for the spectra shown in Fig. 5.
On evacuation, both bands disappear completely in H-ZSM-5,
whereas a certain rest of intensity remains at around 1574 cm
after sample Fe-Z(1.2) is evacuated (cf. Fig. 6). This suggests
on the best NH -SCR catalyst of the ILIE series, Fe-Z(1.2) [10],
3
−1
after preoxidation in air and after prereduction in NH3.
−
1
3
.1.3. In situ FTIR measurements
that the bands around 1880 and 1575 cm belong to the same
compound; therefore, they are assigned to N2O3 that may arise
from dimerization of NO and NO2. As mentioned above, NO2
is contained as an impurity in the NO used for this study. How-
ever, the markedly higher intensity of the bands of N2O3 and
N2O4 in sample Fe-Z(1.2) compared with H-ZSM-5 indicates
that the main source of NO2 (which contributes to N2O3 and
N2O4 formation) is oxidation of NO by the Fe sites of the cata-
lyst, and that the NO2 impurity may play a minor role because
of its low concentration. Moreover, it must be taken into ac-
count that the commercial H-ZSM-5 used in this study contains
traces of Fe as an impurity (ca. 500 ppm), which could also re-
act with NO.
FTIR spectra of sample Fe-Z(1.2) and of the bare H-ZSM-
matrix after oxidative pretreatment and 30 min adsorption of
% NO/He showed bands at around 2193, 2134, 1880, 1744,
5
1
1
−
1
601–1628, and 1576 cm (Fig. 5). Others have observed sim-
ilar bands on adsorption of NO (Table 1) [25–28]. Based on
−
1
these studies, bands around 2193 and 2134 cm are assigned
+
+
to [NO ][N2O4] and NO , respectively, whereas the band at
745 cm is ascribed to adsorbed N2O4. The latter can form
−1
1
easily already at room temperature on oxidation of NO to NO2
followed by subsequent dimerization [29]. The NO used for the
experiments in the present work contained traces of NO2 and
was not purified further. This is one possible explanation for
Surprisingly, no bands below 1700 cm 1 were detected on
NO adsorption when thoroughly purified NO was used [22]. In
this case only bands in the range of 1800–1900 cm were ob-
served, and these were assigned to mononitrosyl and dinitrosyl
complexes of iron. This is in good agreement with results of
−
the formation of N O , besides oxidation of NO by the Fe sites.
2
4
+
For NO , different formation routes must be considered [30],
including
−
1
N2O4 → NO3− + NO+
(2)

M. Santhosh Kumar et al. / Journal of Catalysis 239 (2006) 173–186
181
Fig. 6. In situ FTIR spectra after adsorption of NO at 293 K on (a) H-ZSM-5 and (b) Fe-Z(1.2) after preoxidation (1 h at 673 K in air, —) and prereduction (1 h at
6
73 K in 2% NH /He, ·····) after evacuation of the IR cell. Difference spectra are shown.
3
Long and Yang [33], who observed that Fe-ZSM-5 reacts much
more strongly with a mixture of NO and NO2 than with NO
alone.
Fe-Z(1.2) with a markedly higher Fe content, although not as
pronounced as on prereduced H-ZSM-5. That the severity of the
reductive pretreatment was only moderate must also be taken
into account. In situ UV–vis and EPR experiments revealed that
Bands above 1700 cm 1 disappear on evacuation for 30 min
−
complete reduction of Fe3 was not possible even after 1 h at
+
(cf. Figs. 5 and 6). This suggests that the aforementioned ni-
trogen oxide species to which they have been assigned are
773 K in NH /He flow. Thus, two superimposing effects can
3
−
1
only weakly adsorbed. In contrast, bands below 1700 cm re-
main after evacuation, although there appear to be differences
account for the presence of NO and its secondary products re-
2
flected by the aforementioned bands: (a) reaction of the NO2
−
1
in intensity. Thus, the band at 1601 cm is no longer ob-
served, probably due to weakly adsorbed NO2, because it is
close to the asymmetric stretching frequency of gaseous NO2
impurity in the NO gas, which should not depend on the pre-
treatment and should be comparable over both samples, and (b)
oxidation of NO to NO2 by the Fe3 that persisted the reductive
pretreatment. Effect (a) may have been dominant on prereduced
H-ZSM-5 (Fig. 5); however, because it is not completely Fe-
+
−1
−1
(
1610 cm ) [25]. Bands at 1634, 1625, and 1570 cm have
been assigned to adsorbed nitrato species [25–28]. It is well
−
−
known that NO3 and NO2 anions give rise to bands at 1380
3+
free, and it is not known whether all of the Fe sites were
−1
and 1260 cm , respectively [29]. For adsorbed nitrate and
nitrite species that are not purely ionic, an increase in bond
strength and change in symmetry cause a blue shift of the bands.
This is in agreement with the results of Davydov [31], who
reported bands of differently bound nitrate species at 1480–
reduced by the pretreatment, effect (b) cannot be completely
excluded. Certainly, effect (b) is more important over reduced
Fe-Z(1.2) and dominant over preoxidized samples. In general,
the decreased band intensities in the in situ FTIR spectra af-
ter reductive pretreatment are obvious, at least for the weakly
adsorbed nitrogen oxide species. The stronger interaction of
−1
−1
1
650 cm and those of nitrite species at 1205–1520 cm .
Based on these findings, bands at 1628, 1618, and 1576 cm
−
1
3+
Fe with NO suggests that trivalent iron is required to oxidize
in the spectra of evacuated samples (Fig. 5) can be assigned
to differently bound nitrato species. Interestingly, the band at
NO to NO2 and further to adsorbed nitrate that accumulates on
2+
the surface. The more weakly interacting Fe is probably less
−1
1
576 cm is missing on the bare H-ZSM-5 and observed only
active (if at all) for this reaction; a redox cycle with this ion
would involve low-valent Fe species that should be highly un-
stable.
on Fe-Z(1.2). This suggests that nitrato species reflected by this
band are preferably adsorbed on Fe species. The fact that a
weak band of nitrato species is also seen on the bare H-ZSM-5
can be explained by the presence of traces of Fe (ca. 500 ppm)
as impurities.
3
.2. In situ studies in the presence of isobutane-SCR feed
components
To learn more about the influence of the Fe valence state on
the interaction with NO, we performed the same experiments
with reductively pretreated samples. Similar bands were ob-
served, although with lower intensity (Fig. 5). Over H-ZSM-5,
Spectroscopic in situ studies during NH3-SCR revealed ob-
vious differences between the samples with the highest and
lowest Fe content. Therefore, samples Fe-Z(1.2) and Fe-Z(0.2),
respectively, Fe-Z(0.3) were selected for in situ studies in
isobutane-SCR to elucidate the influence of the different reduc-
ing agents.
+
−1
+
−1
the bands for [NO ][N O ] (2204 cm ), NO (2136 cm ),
2
4
−
1
and N2O3 (1875 and 1577 cm ) were very weak, whereas
the band for N2O4 (1745 cm ) can hardly be seen (Fig. 5).
The decrease in these bands was also obvious on prereduced
−
1

182
M. Santhosh Kumar et al. / Journal of Catalysis 239 (2006) 173–186
Fig. 7. In situ UV–vis spectra measured at 623 K during sequential treatment in a flow of (1) air, (2) 0.1% NO, 0.1% isobutane, 2% O /He, (3) 0.1% isobutane/He
2
(after reoxidation at 773 K) and (4) 0.1% NO/He. Each spectrum measured after 1 h exposure time.
+
3
.2.1. In situ UV–vis studies
In contrast to the NH3-SCR feed mixture (Fig. 2), treatment
be due to the strong oxidizing power of the Fex3 Oy clusters
that are abundant in this sample, which prevents C/N deposit
formation. Even in a flow of isobutane/He containing no ox-
idizing agent, no carbon deposits are formed, in contrast to
sample Fe-Z(0.2) (Fig. 7), suggesting that isobutane is oxidized
to COx by the iron oxide clusters. The related partial reduction
of sample Fe-Z(0.2) under isobutane-SCR conditions caused
not a decrease, but rather a marked increase in UV–vis ab-
sorbance below 600 nm (Fig. 7). This effect may be due to
the formation of C/N-containing organic deposits in the pores,
which has been discussed for the SCR of NO by hydrocar-
bons [26,34–37]. Based on in situ FTIR results, the formation
of several C/N-containing organic species, including alkyl ni-
trites, nitrates and nitriles, cyanates, and isocyanates, has been
proposed. Almost all of these species are UV–vis active and
give rise to bands below 400 nm; thus, it is not possible to
identify particular species from UV–vis spectra alone. There-
fore, additional in situ FTIR measurements were performed, as
discussed below. When the preoxidized sample Fe-Z(0.2) was
exposed to a stream of only isobutane/He, a rather different be-
havior was observed (Fig. 7): The intensity of the bands below
of Fe3 is reflected in a slight decrease of the CT band intensity
around 350 nm and a concomitant increase in absorbance above
500 nm. The latter originates from the low-wavelength tail of
+
2+
an intervalence charge-transfer transition (IVCT) between Fe
and Fe3 that is characteristic for mixed-valence iron oxides
and has been observed in Fe3O4 nanoparticles in the NIR range
[8,38]. The spectrum is restored on NO/He treatment, indicat-
ing that NO is able to reoxidize the clusters.
+
3.2.2. In situ EPR studies
In the EPR spectrum of sample Fe-Z(0.2), the signal at
ꢀ
ꢀ
3
4
00 nm decreased, whereas light absorption increased above
00 nm. The former effect is due to the reduction of isolated Fe
g ≈ 6 disappeared and the signal at g ≈ 4.3 lost much of its in-
tensity as soon as the preoxidized catalyst came in contact with
the isobutane-SCR feed mixture (Fig. 8), indicating rapid re-
species, whereas the enhanced absorption in the visible range
arises from the formation of carbonaceous deposits. This may
occur preferentially on acidic sites and is due to incomplete
oxidation of isobutane. Most probably, those carbonaceous de-
posits also contribute to the UV–vis absorbance below 300 nm
and partly compensate for the loss of intensity caused by reduc-
3+
duction of the isolated Fe species represented by these lines.
ꢀ
3+
,
In contrast, the line at g ≈ 2, which is also due to isolated Fe
decreased only marginally. This behavior is comparable to that
under NH -SCR conditions (Fig. 4), although the reduction of
3
the Fe3 sites reflected by the low-field signals is much more
+
tion of Fe3 . This is suggested by comparing the total intensity
losses during treatment in NH3/He (Fig. 2) and in isobutane/He
+
pronounced in isobutane-SCR. In addition, a narrow singlet is
ꢀ
superimposed in the g ≈ 2 range, indicating the formation of
(Fig. 7). On subsequent treatment in NO/He flow, the UV–vis
some kind of carbon radical that may be connected with the
formation of C/N deposits, as evidenced by UV–vis results de-
scribed above.
band intensity again increased below 400 nm but decreased
above 400 nm. This points to a reaction of NO with the carbona-
ceous deposits on which C/N-containing species are formed, as
was also observed in the total isobutane-SCR feed mixture. In
addition, the reoxidation of Fe to Fe by NO should also
contribute to the increased absorption below 400 nm.
ꢀ
In the presence of isobutane/He, the g ≈ 4.3 line decreased
even more, and the intensity loss is almost nonreversible on
2+
3+
ꢀ
treatment with NO/He. In addition, the g ≈ 6 line does not
reappear in NO/He, whereas the Fe3 signal at g ≈ 2 remains
roughly unchanged during all steps of this sequential treatment.
In general, the EPR spectrum of sample Fe-Z(1.2) (Fig. 8)
showed changes similar to those observed during sequential
+
ꢀ
In sample Fe-Z(1.2), which is dominated by Fe O clusters
x
y
and particles of different sizes, treatment with the isobutane-
SCR feed mixture had virtually no effect on the intensity of
the UV–vis bands (Fig. 7). As for NH SCR, this suggests that
treatment with NH -SCR feed components (Fig. 4). Under the
3
3
most of the Fe species remain trivalent under isobutane-SCR
conditions and also that insufficient C/N-containing deposits
detectable by UV–vis-DRS are formed on Fe-Z(1.2). This might
total isobutane-SCR mixture, virtually no changes were ob-
ꢀ
served for the broad signal at g ≈ 2 due mainly to iron ox-
ide clusters and particles in this sample, because the reduc-

M. Santhosh Kumar et al. / Journal of Catalysis 239 (2006) 173–186
183
Fig. 8. In situ EPR spectra measured at 623 K during sequential treatment in a flow of (1) air, (2) 0.1% NO, 0.1% isobutane, 2% O /He, (3) 0.1% isobutane/He
2
(after reoxidation at 773 K) and (4) 0.1% NO/He. Each spectrum measured after 1 h exposure time.
Fig. 9. In situ FTIR spectra measured at 623 K. Treatment: (1) 1 h in air at 673 K + 1 h in a flow of 1% NO/He (30 ml/min), 1% isobutane/He (30 ml/min) and air
30 ml/min) at 623 K (—); (2) subsequent air treatment at 673 K for 1 h and cooling to 623 K (- - -). Difference spectra after evacuation are shown.
(
tion/reoxidation process keeps the average steady-state Fe va-
lence at +3. However, due to the stronger reducing power of
isobutane compared with NH3, partial reduction to ferrimag-
netically ordered iron oxide clusters and/or particles was more
pronounced in isobutane/He (Fig. 8) than in NH3/He (Fig. 4).
This is reflected by the much higher intensity of the EPR signal
in isobutane/He (Fig. 8). On subsequent treatment by NO/He,
the line shape was modified, but the intensity gain was only
partially reversible, suggesting a change in the ferrimagnetic
ordering but not complete reoxidation. Interestingly, no carbon
radical signal was observed in the EPR spectra of Fe-Z(1.2) as
was the case for Fe-Z(0.2). This suggests that such radicals are
indeed part of the C/N-containing deposits formed in significant
amounts on the latter catalyst but not on the former catalyst.
isobutane-SCR feed at 623 K followed by evacuation and af-
ter subsequent 1 h of reoxidation at 673 K. For comparison, the
same experiment was also performed with the bare H-ZSM-5
(Fig. 9).
It can be seen that the FTIR spectra after SCR feed treat-
ment are rather similar for H-ZSM-5 and Fe-Z(0.3). The strong
−
1
band at 1608 cm clearly indicates the formation of adsorbed
−1
nitrate species [28,31], and bands around 1437 cm point to
the formation of adsorbed nitrite species [31]. However, it must
be mentioned that bands in the latter range can also arise from
deformation vibrations of adsorbed ammonium species. In par-
ticular, this possibility cannot be excluded for the Fe-ZSM-5
samples, because an additional weak and broad band was ob-
served at 3340/3132 cm (not shown) in the range for ν(N–H)
vibrations, which is not visible for H-ZSM-5. Moreover, it is
known that the intermediate formation of NH3 can occur dur-
ing SCR of NO with hydrocarbons [39,40].
−
1
3
.2.3. In situ FTIR studies
To gain more information about the nature of adsorbates de-
tected during the in situ UV–vis experiment described above,
in situ FTIR spectra were recorded after 1 h of treatment in
−
1
The band at 2240 cm is certainly due to the C–N stretching
vibration of nitrile, cyanate, and/or isocyanate species, which

184
M. Santhosh Kumar et al. / Journal of Catalysis 239 (2006) 173–186
can be formed only in the presence of both NO and i-butane.
Interestingly, this band was not observed over H-ZSM-5, sug-
gesting that this reaction requires the presence of a sufficient
amount of Fe species, probably for the intermediate oxidation
of NO to NO2 and/or adsorbed NOx species, which then react
with i-butane to C/N-containing intermediates. Obviously, the
amount of adsorbed nitrates formed by the very low level of Fe
impurities (0.05%) present in the commercial H-ZSM-5 is not
sufficient to produce C/N-containing deposits with detectable
surface concentrations.
and strongly adsorbed nitrate intermediates are readily formed
(Figs. 5 and 6). They were present in lower amounts on prere-
duced Fe-ZSM-5, suggesting that nitrogen oxides interact more
strongly with trivalent iron. Hence, active sites for NO oxida-
tion should be Fe3 species. Fe–NO adsorption complexes, as
detected in the complete absence of NO2 and O2 [22,25], were
not found. This suggests that under SCR conditions in the com-
mon presence of NO, NO2, and O2, the reaction proceeds via
the intermediate formation of adsorbed nitrate intermediates,
which might then be reduced by NH3 to N2. This is in agree-
ment with the reaction mechanism proposed by Long and Yang
[33] involving initial oxidation of NO to adsorbed NOx (x ꢀ 2)
+
After subsequent treatment in air at 673 K, the band at
−1
2
240 cm decreased due to the oxidative degradation of ni-
+
trile, cyanate and/or isocyanate species. The bands around
and further reduction to N2 by NH4 ions. Clearly, the inter-
−1
3+
1
600 and 1430 cm were markedly diminished in samples
mediate oxidation of NO may be favored by Fe but not by
2+
Fe-Z(0.3) and Fe-Z(1.2) but remained almost unchanged in H-
ZSM-5. This suggests that in the former two samples those
bands may arise in large part from oxidizable organic (alkyl)
nitrates and nitrites, whereas in H-ZSM-5 they may originate
from inorganic nitrate and nitrite species, the thermal degrada-
tion of which is probably not favored at this temperature.
Fe . Thus, it seems likely that, among the isolated Fe species,
those that are not permanently reduced under steady-state SCR
conditions are active sites in the SCR of NO by NH3. This is
also supported by the fact that in used Fe-ZSM-5 catalysts, the
ꢀ
g ≈ 6 signal disappears completely, although no deactivation is
observed during the SCR reaction [17]. Because the amount of
barely reducible isolated Fe increases with increasing Fe con-
tent, it is not surprising that the activity of the Fe-Z samples has
been observed to increase as well [10].
−
1
Moreover, a band around 1690 cm was observed in all
−
1
samples along with additional bands at 1723 and 1765 cm
occurring only in sample Fe-Z(1.2), characterized by abundant
FexOy clusters. These bands were all in the range of ν(C=O)
and diminished on reoxidation. They may have arised from
carbonyl-containing species of different structures formed on
Despite these findings, however, attempts to correlate the
rate of the SCR reaction with the number of Fe sites in the
Fe-Z catalysts [10] were most successful when assuming that
not only isolated Fe ions, but also oligonuclear FexOy clus-
ters and probably even Fe ions accessible on the surface of
oxide particles, are redox-active species in NH3-SCR of NO.
In situ UV–vis studies, including those of the redox kinetics
partial oxidation of i-butane by Fe3 species and burned to COx
+
in air flow.
3
.3. Active sites and catalytic behaviour
(Table 2), have shown that FexOy agglomerates reoxidize much
Comparing the results of in situ UV–vis and EPR mea-
more rapidly than isolated Fe sites and thus can immediately en-
ter into another redox cycle. Therefore, it is plausible to assume
that they contribute to the selective catalytic process at lower re-
action temperatures as well. At higher temperatures (>700 K),
these agglomerates, due to their higher oxidation potential in
surements during sequential treatment with components of the
NH3-SCR feed reveals some crucial differences (Figs. 2 and 4).
Isolated Fe3 ions in samples free of oxidic clusters [Fe-Z(0.2)
+
and Fe-Z(0.3)] were partially reduced under steady-state SCR
3+
3+
conditions, albeit to different extents. Octahedral Fe , re-
comparison to isolated Fe species, give rise to unselective to-
tal oxidation of NH3, thus limiting NO conversion [10].
ꢀ
flected by EPR signals around g ≈ 6 and a UV–vis band at 291
nm, were the most sensitive to reduction, followed by tetrahe-
In general, the behavior of isolated Fe sites during isobutane-
SCR, as evidenced by in situ EPR results, is similar to that
under NH3-SCR conditions (Fig. 8). This agrees with the ob-
servation that at low reaction temperature, catalytic activity
increases with increasing Fe content [10]. A major difference
arises from the fact that isolated Fe sites are more strongly re-
duced under isobutane-SCR conditions (Fig. 8). Note that the
3
+
ꢀ
3+
dral Fe (g ≈ 4.3 and 241 nm), whereas Fe ions indicated
ꢀ
by the EPR signal at g ≈ 2 were hardly reduced. For these
Fe species, the coordination geometry cannot be easily speci-
fied; however, with respect to the UV–vis spectra under steady-
state SCR conditions showing bands of both reduction-resistant
3+
tetrahedral and octahedral Fe , it is likely that the EPR sig-
ꢀ
ꢀ
nal at g ≈ 2 also reflects Fe sites of both coordinations. The
EPR signal at g ≈ 6 is no longer present and that the one at
3+
ꢀ
amounts of these barely reducible isolated Fe sites increased
with increasing Fe content, as can be seen by comparing the
UV–vis intensity below 300 nm under steady-state conditions
g ≈ 4.3 is much smaller than during NH3-SCR. However, the
exposed Fe sites of oxidic clusters are more strongly reduced,
too, giving rise to more pronounced formation of ferrimagnetic
species; compare the EPR spectra of Fe-Z(1.2) in Figs. 4 and 8.
This stronger reduction in steady state can explain the fact that
the overall activity of the Fe-Z samples in isobutane-SCR is
much lower than in NH3-SCR under the same reaction con-
ditions [10], because the intermediate oxidation of NO is not
as effectively catalyzed due to the lower amount of accessible
trivalent Fe species. Another obvious difference that depends
on the nature of the reductant is evident from the behavior of
(
Fig. 2). Regarding the relative intensities of the different EPR
signals (Fig. 4), it appears that the barely reducible sites com-
pose the majority of the isolated Fe species and are reflected
ꢀ
mainly by the signal at g ≈ 2. In contrast, easily reducible sites
ꢀ
give rise to the line at g ≈ 6 and also partly to the signal at
ꢀ
g ≈ 4.3 and seem to be much less abundant.
In situ FTIR spectra suggest that, in the presence of even
traces of NO2 and/or oxygen, weakly adsorbed N2O3, N2O4,

M. Santhosh Kumar et al. / Journal of Catalysis 239 (2006) 173–186
185
the oxide clusters. In contrast to NH3-SCR, where they are se-
lective up to temperatures of about 700 K, FexOy agglomerates
play a detrimental role in isobutane-SCR because they oxidize
the isobutane reductant to COx at temperatures as low as 623 K.
This is evident from carbonyl-containing species detected by
in situ FTIR preferentially on Fe-Z(1.2) but not on Fe-Z(0.3)
content ꢁ0.3 wt%) represents isolated reduction-resistant and
3
+
thus probably active Fe sites. Permanent reduction of isolated
3+
Fe was more pronounced during SCR with i-butane than with
NH , which may be why the catalysts were markedly less active
3
in the former case.
In contrast to certain isolated Fe3 sites, oxidic clusters of
+
(
Fig. 9). Recently we showed that the commercial H-ZSM-5
different sizes, reflected by UV–vis signals above 300 nm and
ꢀ
matrix used for the preparation of the Fe-Z samples contains
about 500 ppm of Fe impurities in the form of clusters rather
than as isolated sites [17], demonstrating why a carbonyl band
an EPR line at g ≈ 2 (in samples with Fe content ꢁ0.6 wt%),
remained essentially trivalent during SCR of NO with both NH
and i-butane, due to their slow reduction and very rapid re-
oxidation. Consequently, these species have a higher oxidizing
3
−1
at 1690 cm is also observed on this sample (Fig. 9).
power than isolated Fe3 sites. This leads to undesired total ox-
idation of the reducing agent at higher temperatures, an effect
that is most detrimental for i-butane but not as crucial for NH3.
As a rule for knowledge-based catalyst design, the results
+
Carbonyl-containing adsorbates are considered as intermedi-
ates in the total oxidation of isobutane. As a result of this unse-
lective oxidation behavior of the FexOy agglomerates, NO con-
version decreases dramatically above 600 K, and this undesired
effect becomes more and more pronounced as the content of
oxide clusters increases with increasing Fe content [10]. In con-
trast, the formation of C/N-containing species, such as nitriles,
cyanates, and isocyanates, seems to proceed preferentially on
isolated Fe sites, the amount of which increases from sample
Fe-Z(0.3) to sample Fe-Z(1.2) (Fig. 9). Some have claimed that
these species are active intermediates in NOx reduction [35,38–
of this study indicate that for both NH - and i-butane-SCR,
3
the concentration of accessible Fe3 that can participate in a
reversible redox cycle should be maximized, whereas the for-
mation of Fe O clusters must be completely (i-butane-SCR)
+
x
y
or largely (NH -SCR) avoided.
3
Acknowledgments
4
0], which can be considered another reason for the superior
catalytic performance of isolated Fe sites in isobutane-SCR.
We thank the Deutsche Forschungsgemeinschaft for finan-
cial support (grants BR 1380/7-2 and Gr 1447/7-2).
4
. Conclusion
References
A series of Fe-ZSM-5 catalysts with a tailored dispersion
of Fe sites ranging from almost exclusively isolated Fe ions
[
1] G.I. Panov, G.A. Sheveleva, A.S. Kharitonov, V.N. Romannikov, L.A.
Vostrikova, Appl. Catal. A: Gen. 82 (1992) 31.
(
Fe content ꢁ0.3 wt%) via some additional oligonuclear FexOy
clusters (0.6–0.7 wt% Fe) to abundant oligonuclear FexOy clus-
ters and large Fe2O3 particles were studied by in situ UV–vis,
EPR, and FTIR spectroscopy during interaction with feed com-
ponents of the SCR of NO by NH3 and i-butane, respectively.
[2] A. Ribera, I.W.C.E. Arends, S. de Vries, J. Perez-Ramirez, R. Sheldon,
J. Catal. 195 (2000) 287.
[3] J. Perez-Ramirez, F. Kapteijn, G. Mul, X. Xu, J.A. Moulijn, Catal. To-
day 76 (2002) 55.
[4] A.-Z. Ma, W. Grünert, Chem. Commun. (1999) 71.
3+
A certain part of the isolated Fe species reflected by UV and
[5] X. Feng, W.K. Hall, Catal. Lett. 41 (1996) 45.
[6] H.-Y. Chen, W.M.H. Sachtler, Catal. Today 42 (1998) 73.
[7] F. Heinrich, C. Schmidt, E. Löffler, W. Grünert, Catal. Commun. 2 (2001)
ꢀ
3+
EPR signals at 241 nm and g ≈ 4.3 (Fe O4), respectively, at
ꢀ
3+
2
90 nm and g ≈ 6 (Fe O4+x, x = 1–2) was irreversibly re-
3
17.
8] J. Pérez-Ramirez, M. Santhosh Kumar, A. Brückner, J. Catal. 223 (2004)
3.
9] M. Schwidder, M. Santhosh Kumar, A. Brückner, W. Grünert, Chem.
Commun. (2005) 805.
duced under steady-state SCR conditions. For both NH3- and
i-butane-SCR, this reduction was more pronounced for Fe sites
coordinating to more than four oxygen ligands (290 nm and
[
[
1
ꢀ
g ≈ 6), and the total extent of permanent reduction was high-
[
[
[
10] M. Schwidder, M. Santhosh Kumar, K. Klementiev, M.-M. Pohl, A. Brüc-
est in i-butane-SCR.
kner, W. Grünert, J. Catal. 231 (2005) 314.
In situ FTIR results suggest that the reduction of NO to N2
in the presence of NO2 and/or O2 proceeds via the formation of
higher oxidized intermediates, such as adsorbed N2O4, N2O3,
and nitrates, which are subsequently reduced. In i-butane-SCR,
this reduction proceeds via the formation of C/N-containing
intermediates and possibly organic nitrates detected by FTIR
spectroscopy.
11] A.V. Kucherov, C.N. Montreuil, T.N. Kucherova, M. Shelef, Catal. Lett.
56 (1998) 173.
12] A. Ribera, I.W.C.E. Arends, S. de Vries, J. Pérez-Ramírez, R.A. Sheldon,
J. Catal. 195 (2000) 287.
[13] P. Kubánek, B. Wichterlová, Z. Sobalik, J. Catal. 211 (2002) 109.
14] J. D eˇ de cˇ ek, D. Kaucky, B. Wichterlová, Microporous Mesoporous
Mater. 35–36 (2000) 483.
15] L. Capek, V. Kreibich, J. D eˇ de cˇ ek, T. Grygar, B. Wichterlová, Z. Sobalík,
[
ˇ
[
The formation of nitrogen oxide intermediates in higher ox-
idation states requires the presence of trivalent active Fe sites,
J.A. Martens, R. Brosius, V. Tokarová, Microporous Mesoporous Mater.
80 (2005) 279.
which must undergo a periodic Fe3 /Fe redox cycle. Fe sites
+
2+
[16] B. Wichterlová, J. D eˇ de cˇ ek, Z. Sobalík, Proceedings of the 12th Interna-
tional Zeolite Conference, vol. II, Mater. Res. Soc., Warrendale, PA, 1999,
p. 941.
17] M. Santhosh Kumar, M. Schwidder, W. Grünert, A. Brückner, J. Catal. 227
2004) 384.
that are permanently reduced to Fe2 under SCR conditions and
+
2+
+
cannot be reoxidized are likely not active, because a Fe /Fe
[
redox cycle should be energetically unfavorable. These Fe sites
(
ꢀ
are reflected by EPR signals at g ≈ 6 and partially also at
[
18] J. Pérez-Ramirez, J.C. Groen, A. Brückner, M. Santosh Kumar, U. Ben-
trup, M.N. Debbagh, L.A. Villaescusa, J. Catal. 232 (2005) 318.
ꢀ
ꢀ
g ≈ 4.3, whereas the EPR signal at g ≈ 2 (in samples with Fe

186
M. Santhosh Kumar et al. / Journal of Catalysis 239 (2006) 173–186
[
19] S. Bordiga, R. Buzzoni, F. Geobaldo, C. Lamberti, E. Giamello, A. Zec-
[33] R.Q. Long, R.T. Yang, J. Catal. 207 (2002) 224.
[34] Y. Ukisu, S. Sato, G. Muramatu, K. Yoshida, Catal. Lett. 11 (1991) 177.
[35] N.W. Hayes, W. Grünert, G.J. Hutchings, R.W. Joyner, E.S. Shpiro,
J. Chem. Soc. Chem. Commun. (1994) 531.
[36] K.A. Bethke, C. Li, M.C. Kung, B. Yang, H.H. Kung, Catal. Lett. 31
(1995) 287.
china, G. Leofanti, G. Petrini, G. Tozzolo, G. Vlaic, J. Catal. 158 (1996)
486.
[
[
[
20] G. Lehmann, Z. Phys. Chem. Neue Folge 72 (1970) 279.
21] L. Lopes, J. de Laat, B. Legube, Inorg. Chem. 41 (2002) 2505.
22] L.J. Lobree, I.-C. Hwang, A. Jeffrey, A. Reimer, A.T. Bell, J. Catal. 186
(
1999) 242.
23] J. Dedecek, D. Kaucky, B. Wichterlova, Microporous Mesoporous Mater.
5–36 (2000) 483.
24] V. Umamaheswari, M. Hartmann, A. Pöppl, J. Phys. Chem. B 109 (2005)
537.
[37] F. Radtke, R.A. Koeppel, E.G. Minardi, A. Baiker, J. Catal. 167 (1997)
127.
[38] J. Tang, M. Myers, K.A. Bosnick, L.E. Brus, J. Phys. Chem. B 107 (2003)
7501.
[39] F. Poignant, J. Saussey, J.-C. Lavalley, G. Mabilon, J. Chem. Soc. Chem.
Commun. (1995) 89.
[40] H.-Y. Chen, T. Voskoboinikov, W.M.H. Sachtler, J. Catal. 186 (1999) 91.
[41] D. Goldfarb, M. Bernardo, K.G. Strohmaier, D.E.W. Vaughan, H. Tho-
mann, J. Am. Chem. Soc. 116 (1994) 6344.
[
[
[
3
1
25] H.-Y. Chen, El-M. El-Malki, X. Wang, R.A. van Santen, W.M.H. Sachtler,
J. Mol. Catal. A: Chem. 162 (2000) 159.
26] H.-Y. Chen, T. Voskoboinikov, W.M.H. Sachtler, J. Catal. 180 (1998) 171.
27] L.J. Lobree, I.-C. Hwang, A.R. Jeffrey, A.T. Bell, Catal. Lett. 63 (1999)
[
[
2
33.
28] K. Hadjiivanov, H. Knözinger, B. Tsyntsarski, L. Dimitrov, Catal. Lett. 62
1999) 35.
[42] P. Wenquin, Q. Shilun, K. Zhiyun, P. Shaoyi, Stud. Surf. Sci. Catal. 49
(1989) 281.
[43] A.F. Ojo, J. Dwyer, R.V. Parish, Stud. Surf. Sci. Catal. 49 (1989) 227.
[44] Z. Kónya, I. Hannus, Á. Molnár, I. Kiricsi, Appl. Catal. A: Gen. 146
(1996) 323.
[45] R.K. Talukdar, J.B. Burkholder, M. Hunter, M.K. Gilles, J.M. Roberts,
A.R. Ravishankara, J. Chem. Soc., Faraday Trans. 93 (1997) 2797.
[46] A. Brückner, Chem. Commun. (2001) 2122.
[47] A.M. Beale, A.M.J. van der Eerden, K. Kervinen, M.A. Newton, B.M.
Weckhuysen, Chem. Commun. (2005) 3015.
[
[
[
[
[
(
29] U. Bentrup, A. Brückner, M. Richter, R. Fricke, Appl. Catal. B: Envi-
ron. 32 (2001) 229.
30] K. Hadjiivanov, J. Saussey, J.L. Freysz, J.C. Lavalley, Catal. Lett. 52
(1998) 103.
31] A. Davydov, Molecular Spectroscopy of Oxide Catalyst Surfaces, Wiley,
Chichester, 2003.
32] K. Hadjiivanov, Catal. Rev.—Sci. Eng. 42 (2000) 103.