The role of NO2 in the selective catalytic reduction of nitrogen oxides over Fe-ZSM-5 catalysts: Active sites for the conversion of NO and of NO/NO2 mixtures

Article abstract of DOI:10.1016/j.jcat.2008.07.014

The selective catalytic reduction by NH₃ of NO ( standard SCR ) and NO/NO₂ mixtures ( fast SCR ) were compared over Fe-ZSM-5 catalysts of different Fe content ranging from 0.2 to 5 wt%. The Fe site structure was monitored by UV-vis spectroscopy. From the response of the reaction rates to the variation of the site structure, it was concluded that standard and fast SCR require different Fe sites; standard SCR proceeds both on isolated and oligomeric Fe oxo sites, whereas a small concentration of isolated Fe sites is sufficient to sustain the high reaction rates of fast SCR, with no discernible contribution of different sites. The response of both reactions to hydrothermal stress (with or without SO₂ present) differs; strong deactivation in standard SCR accompanied by increasing NH₃ oxidation activity being opposed to stable activity in fast SCR, suggesting that those isolated sites, which are responsible for fast SCR are very resistive toward clustering under the conditions of automobile exhaust catalysts. The potential of preparation procedures to improve the activity by increasing the Fe content is very different. The best performance was achieved using solid-state ion exchange with FeCl₃, with NO_x conversions >90% between 473 and 773 K obtained over the best catalyst. But if a concomitant increase of the aggregated site abundance cannot be avoided, then NH₃ oxidation will effectively compete with fast SCR already at relatively low temperatures. The role of free NO₂ in the reaction mechanism of standard SCR was assessed by comparing the SCR reaction rate with that of NO oxidation for all catalysts used. NO oxidation was significantly slower than NO reduction in all cases, which excludes free NO₂ produced by NO oxidation as an intermediate of the standard SCR reaction.

Full text of DOI:10.1016/j.jcat.2008.07.014

Journal of Catalysis 259 (2008) 96–103
Contents lists available at ScienceDirect
Journal of Catalysis
www.elsevier.com/locate/jcat
The role of NO2 in the selective catalytic reduction of nitrogen oxides over
Fe-ZSM-5 catalysts: Active sites for the conversion of NO and of NO/NO2 mixtures
Michael Schwidder a, Sascha Heikens a, Andrea De Toni a, Simone Geisler b, Malte Berndt b, Angelika Brückner c,
Wolfgang Grünert a,∗
a
Lehrstuhl für Technische Chemie, Ruhr-Universität Bochum, PO Box 102148, D-44780 Bochum, Germany
Interkat Catalysts GmbH, D-53639 Königswinter, Germany
Leibniz Institute for Catalysis at Rostock University, Berlin Branch, Richard-Willstätter-Str. 12, D-12489 Berlin, Germany
b
c
a r t i c l e i n f o
a b s t r a c t
Article history:
The selective catalytic reduction by NH3 of NO (“standard SCR”) and NO/NO2 mixtures (“fast SCR”)
were compared over Fe-ZSM-5 catalysts of different Fe content ranging from 0.2 to 5 wt%. The Fe site
structure was monitored by UV–vis spectroscopy. From the response of the reaction rates to the variation
of the site structure, it was concluded that standard and fast SCR require different Fe sites; standard
SCR proceeds both on isolated and oligomeric Fe oxo sites, whereas a small concentration of isolated
Fe sites is sufficient to sustain the high reaction rates of fast SCR, with no discernible contribution of
different sites. The response of both reactions to hydrothermal stress (with or without SO2 present)
differs; strong deactivation in standard SCR accompanied by increasing NH3 oxidation activity being
opposed to stable activity in fast SCR, suggesting that those isolated sites, which are responsible for
fast SCR are very resistive toward clustering under the conditions of automobile exhaust catalysts. The
potential of preparation procedures to improve the activity by increasing the Fe content is very different.
The best performance was achieved using solid-state ion exchange with FeCl3, with NOx conversions
>90% between 473 and 773 K obtained over the best catalyst. But if a concomitant increase of the
aggregated site abundance cannot be avoided, then NH3 oxidation will effectively compete with fast SCR
already at relatively low temperatures. The role of free NO2 in the reaction mechanism of standard SCR
was assessed by comparing the SCR reaction rate with that of NO oxidation for all catalysts used. NO
oxidation was significantly slower than NO reduction in all cases, which excludes free NO2 produced by
NO oxidation as an intermediate of the standard SCR reaction.
Received 1 April 2008
Revised 19 July 2008
Accepted 26 July 2008
Available online 23 August 2008
Keywords:
DeNOx
Fe-ZSM-5
NH3-SCR
Fast SCR
Active sites
Reaction mechanism
© 2008 Elsevier Inc. All rights reserved.
1. Introduction
The reason for the acceleration of NO reduction by gas-phase
NO2 is a matter of debate, which is closely related to the discus-
sion about the mechanism of “standard” SCR, that is, the SCR of NO
The selective catalytic reduction (SCR) of nitrogen oxides has
been in the focus of NOx abatement research for mobile sources
over the past two decades. The discovery of catalysts capable of
reducing NO with fuel components [1,2] drew much effort on the
hydrocarbon-SCR in the early 1990s, but the failure to identify
catalysts with sufficient activity and durability meanwhile has re-
verted the attention to the ammonia reductant, which is fed into
the exhaust gas stream as urea. The technological development
takes advantage of the fact that equimolar mixtures of NO and
NO2 are reduced by NH3 at much lower temperatures than NO
or NO2 alone [3]. This “fast SCR” was first observed on Fe2O3/TiO2
catalysts [4], but it proceeds also over V/W–TiO2 [5] and over Fe-
modified zeolites [6–8].
alone [9]. The presence of NO2 may provide a shortcut in a mecha-
nism where its formation is rate-limiting, or it may open a reaction
channel alternative to a route where it is not included. Such routes
have been discussed in the literature, including activation of NH3
by a redox reaction with Fe(III) [4], V(V) [10], or Cu(II) [11], the
amide-nitrosamide mechanism of Ramis et al. [12], which has also
been discussed for Fe catalysts [13,14], or a route in which ni-
trogen in NO is first oxidized, but not beyond the +3 oxidation
state [6]. To accelerate the redox routes by NO2, the reoxidation
must be rate-limiting and more rapid with NO2 than with O2. For
the V/W–TiO2 system, the former was rejected by Tuenter et al. [5]
with reference to the low reaction order in oxygen above 0.5 vol%,
although a favorable influence of facilitated reoxidation by surface
nitrates was recently reported by Tronconi et al. [15]. Contrary to
expectations, Delahay et al. reported that Cu(I) in Faujasite was re-
oxidized less effectively by NO2 than by O2 [16].
Corresponding author. Fax: +49 234 321 4115.
*
E-mail address: w.gruenert@techem.rub.de (W. Grünert).
0021-9517/$ – see front matter © 2008 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2008.07.014

M. Schwidder et al. / Journal of Catalysis 259 (2008) 96–103
97
Free NO2 is a crucial intermediate in the mechanism proposed
previously [17], in which the catalytic role of the iron is confined
to the oxidation of NO and the subsequent fast SCR proceeds ex-
clusively on zeolite Brønsted sites. This mechanism also was sup-
ported by Devadas et al. [7], although they noted a catalytic rele-
vance of the Fe sites in the fast SCR step as well. In recent work
with iron-free catalysts, Sachtler et al. [18] emphasized the forma-
tion of NH4NO2, which is easily formed from NO, NO2, NH3, and
H2O [18]. Its decomposition to nitrogen and water is very rapid
with no catalyst at low temperatures, but enhancement by acidic
sites also has been observed [19].
In our recent work, we investigated the relevance of different
Fe sites in Fe-ZSM-5 for the SCR of NO by ammonia in some
detail, combining structural information from UV–vis, EPR, and
X-ray absorption spectroscopy with catalytic data [20–22]. Accord-
ing to these results, all iron sites present, including isolated Fe
oxo species, can catalyze this reaction, and at low temperatures
(523 K), the reaction rate can even be correlated with the total Fe
content for catalysts with well-dispersed iron. At the same time,
a particular activity contribution of oligomeric clusters was noted,
which would have impeded such correlation at higher tempera-
tures [21].
The present paper reports an attempt to extend this approach
on the fast SCR by studying this reaction over Fe-ZSM-5 catalysts of
different Fe site structures, characterized by UV–vis spectroscopy.
Striking differences in site requirements of fast and standard SCR,
as well as in catalyst durability, indicate that the fast SCR pro-
ceeded on isolated Fe oxo sites, which is further supported by a
successful optimization effort for Fe-ZSM-5 inspired by this con-
clusion. A clear discrepancy between the rates of oxidation and
the selective reduction of NO over the catalysts used suggests no
mechanistic relation between the seemingly similar selective cat-
alytic reduction of NO and of NO/NO2 mixtures.
was redissolved in 1 ml nitric acid, and the liquid was washed
into a volumetric flask containing demineralized water. The UV–
vis spectra of the catalysts in calcined state (heating in synthetic
air to 423 K with 2 K min−1, after a 15 min isothermal period heat-
ing to 873 K with 5 K min−1, followed by 1 h at this temperature)
were recorded with a Cary 400 spectrometer (Varian) equipped
with a diffuse reflectance accessory (Harrick). For light absorption
reduction, samples were diluted with α-Al2O3 (calcined for 4 h at
1473 K) at a ratio of 1:10.
The catalytic data were measured in a catalytic microflow reac-
tor (6 mm i.d.) in the temperature range 423–873 K. In standard
SCR, the feed gas consisted of 1000 ppm NO, 1000 ppm NH3, and
2% oxygen with helium as balance. For fast SCR, 500 ppm NO2
and 500 ppm NO were used instead of 1000 ppm NO. NO2 was
introduced into the feed gas mixture directly before the reactor in-
let, to prevent the formation of ammonium nitrate within the gas
lines. For the NO oxidation reaction, only 1000 ppm and 2% oxy-
gen with helium as balance were used. In all reactions, the space
−1
velocity applied was 750,000 h
(catalyst mass, 10 mg; particle
size, 250–350 μm; total flow rate, 183.3 ml min−1). Conversions
were determined using calibrated mass spectrometry (Balzers QMS
200). The only nitrogen-containing components detected were NO,
NO2, N2, and NH3, N2O was not found in measurable quantities. In
fact, N2O detection is not very sensitive with our analytical scheme
because of rather intense fragmentation, but the N2O selectivity is
known to be low over Fe-ZSM-5 under the conditions applied here
[7,24]. Cross-sensitivities between ammonia and water fragments
(m/z = 17) and between NO and NO2 fragments (m/z = 30) were
eliminated by calibrated fragmentation ratios.
This experimental scheme has two problems. First, at tempera-
tures below 500 K, the measurements may have been influenced
by the formation of ammonium nitrate. The role of ammonium
nitrate in fast SCR is presently under debate. Its formation was
first considered as a side reaction [25], and recently its role as an
intermediate in the fast-SCR reaction sequence over V/W–TiO2 cat-
alysts has been proposed [26]. We discuss the extent of this effect
over our catalysts later in the paper. The observation of white (am-
monium nitrate) powder deposition at inadequately heated reactor
walls near the reactor entrance suggests, however, that this reac-
tion can proceed noncatalyzed and parallel to the main reaction.
Second, mass spectrometry analysis of coexisting NO and NO2 is
prone to experimental errors in the individual concentration of the
oxides, whereas their sum is less affected. Due to the strong frag-
mentation of NO2 in the mass spectrometer, measured conversions
of both NO and NO2 depend critically on the correct intensity mea-
surement of the weak NO2 signal at m/e = 46. Underestimation of
the NO2 signal will result in insufficient correction of the NO in-
tensity, that is, overestimation of residual NO. Thus, errors in NO
and NO2 conversion are complementary, leaving the NOx conver-
sion largely unaffected. The problem disappears when the residual
NO2 concentration tends toward zero, obviously at high tempera-
tures in our experiments. Although evidence also exists suggesting
that the divergence between NO and NO2 conversion may be real
in some cases, for now we assign it to undiscovered changes in
the mass spectrometry conditions, leading to the aforementioned
errors, and confine ourselves to the discussion of total NOx conver-
sion in what follows.
2. Experimental
The ZSM-5 zeolites used had a Si/Al ratio of 14 and were pro-
vided by Chemiewerk Bad Köstritz (Germany; Na-ZSM-5, used for
ILIE and CVD preparations, see below) and by Tricat Zeolites GmbH
Bitterfeld, now Südchemie Bitterfeld GmbH (Germany; NH4-ZSM-
5, for SSIE samples and study with parent zeolite). H-ZSM-5 was
made from these zeolites through standard procedures (i.e., ex-
change with HCl, calcination at 873 K).
The Fe-ZSM-5 catalysts were prepared through various routes.
The improved liquid ion-exchange (ILIE) samples were prepared ac-
cording to a procedure described in detail elsewhere [21]. In brief,
the zeolite was exchanged with Fe2+ generated in situ from iron
powder under an inert gas atmosphere. The chemical vapor de-
position (CVD) sample was prepared as described previously [23],
with the zeolite in the H form exchanged with vaporized FeCl3
under an inert gas atmosphere. The solid-state ion exchange (SSIE)
samples were prepared by heating a mixture of the zeolite in the H
form with a suitable amount of FeCl3 in flowing nitrogen according
to the following protocol. First, the mixture was heated to 423 K
−1
with a temperature ramp of 2 K min
to protect the pore sys-
tem of the zeolite from being destroyed by vaporizing water. Then
−1
the temperature was increased to 573 K at a rate of 5 K min
and maintained there for 1 h. Finally, the sample was washed with
deionized water and dried at ambient atmosphere. In what follows,
we designate the samples based on their preparation method and
their iron content; that is ILIE-0.2 designates a Fe-ZSM-5 prepared
by the ILIE method and containing 0.2 wt% iron.
3. Results and discussion
3.1. Characterization of Fe-ZSM-5 samples by UV–vis spectroscopy
The Fe content of the samples was analyzed by atomic absorp-
tion spectroscopy. For digestion, 50 mg of each sample was dis-
solved in a mixture of concentrated acids: HNO3 (5 ml), HCl (1 ml),
and HF (1 ml). After evaporation of the acid mixture, the residue
Fig. 1 presents the UV–vis spectra of the Fe-ZSM-5 catalysts
used in this study. The Kubelka–Munk function is plotted in nor-
malized form to facilitate comparison of signal shapes. The spec-
tra differ strongly and indicate a widely varying degree of clus-

98
M. Schwidder et al. / Journal of Catalysis 259 (2008) 96–103
Fig. 1. UV–vis spectra (normalized Kubelka–Munk function) of Fe-ZSM-5 catalysts used in the present study.
tering of the Fe sites present. The signals arise from Fe3+ ← O
charge-transfer transitions. It has been proposed that bands be-
low 300 nm indicate isolated Fe ions in tetrahedral and octahedral
coordinations, whereas the t1 → t2 and t1 → e transitions remain
unresolved [27,28]. Signals between 300 and 400 nm have been
assigned to oligomeric Fe oxo entities, and bands above 400 nm
have been assigned to Fe oxide aggregates, which may be strongly
disordered. The smallest aggregate size that would cause a shift of
the band into the region >400 nm is unknown.
Spectra as shown in Fig. 1 may be deconvoluted into sub-
bands that fall into the aforementioned wavelength regions, as has
been shown previously [20,21,28]. For our purposes here, a qual-
itative discussion of this is sufficient. Among the ILIE catalysts,
that with the lowest Fe content (ILIE-0.2, Fig. 1a) gave rise to
a very narrow spectrum consisting almost completely of bands
peaking below 300 nm. In a quantitative analysis of such a spec-
trum, 95% of the intensity was found to arise from signals below
300 nm [21]. Thus, when changes of the extinction coefficients
over the wavelength range are neglected, one may suggest that
almost all iron is present in isolated sites, with the remainder
present as oligomers [21]. It was recently shown that signals be-
low 300 nm also may arise from binuclear Fe oxo clusters provided
that the oxo bridge between the Fe ions is hydroxylated [29]. Such
hydroxylation may be questioned for the calcined state in which
the spectra shown in Fig. 1 were measured. Moreover, in that pre-
vious study [29], only 30% of the iron turned out to be actually
paired at a Fe content of 0.4 wt% in magnetic measurements made
to differentiate sites identified as isolated by UV–vis. Therefore, we
believe that the vast majority of Fe sites in ILIE-0.2 indeed may be
considered isolated.
The remaining spectra clearly extended into the range typical of
clustered species, but they differed significantly in detail. The spec-
trum of ILIE-0.6 extended into the 300–400 nm range, assigned
to oligomeric clusters, whereas all other samples exhibited signifi-
cant intensity also beyond 400 nm. In the ILIE series, the intensity
near 300 nm increased with increasing Fe content and provided
the total intensity maximum in the spectrum of ILIE-1.2. This is a
clear difference from the spectrum of SSIE-1.1, which peaked be-
low 250 nm, with a well-separated signal in the oligomer range
300–400 nm and increased intensity above 400 nm. Earlier anal-
yses suggested a distribution of ca. 50% isolated, 25% oligomeric,
and 25% particulate species for ILIE-1.2 [21]. For SSIE-1.1, a some-
what higher weight of isolated and of particulate sites the signal
shape suggests, with a smaller contribution of oligomeric sites. The
spectrum of CVD-5 (Fig. 1a) was clearly dominated by species ab-
sorbing above 300 nm; that is, clustering of the Fe oxo species was
greatest in this catalyst. But signals below 300 nm contributed to
the spectrum as well; thus, even a highly loaded Fe-ZSM-5 con-
tained isolated Fe sites as confirmed previously [29] by magnetic
measurements. But one should keep in mind that only relative
contributions of species have been considered to date. Due to the
increasing Fe content, the absolute abundance of isolated sites still
grew from ILIE-0.2 via ILIE-1.2/SSIE-1.1 to CVD-5, although the pro-
portion of total Fe present decreased.
3.2. Fast and standard SCR: Activities
Fig. 2 compares conversions measured during fast SCR and
standard SCR for the three catalysts of the ILIE series. As men-
tioned in the Experimental section, fast SCR conversion data below
500 K may have been affected to some extent by the formation of
NH4NO3. Above 500 K, where deposited NH4NO3 should have dis-
appeared due to melting and decomposition, the time to achieve
steady state was extended. To evaluate the influence of ammonium
nitrate formation, a full nitrogen balance, including N2 formed, was
conducted in a specific run. In the entire temperature range, the
N balance closed between 94% and 104%, with the lowest value
of 94% occurring at 423 K. Thus, it may be concluded that even
at temperatures below 500 K, the error due to the side reaction
was hardly beyond the limits of experimental error. As an addi-
tional check, the fast SCR over ILIE-0.2 (cf. Fig. 1a) was measured
both with increasing and decreasing sequences of reaction temper-
atures. The results were very close, confirming an only minor effect
of ammonium nitrate formation on the data even below 500 K.
Fig. 2a shows the data for the run with decreasing temperature se-
quence due to some outliers in the increasing temperature series.
Fig. 2 confirms the favorable effect of NO2 on the NOx conver-
sion over Fe-ZSM-5. At the same time, it indicates that the reac-
tions in presence and absence of NO2 were very different and pro-
ceeded on different sites. A first indication of this is the dramatic
shift in the difference of light-off temperatures (T50) between fast
and normal SCR. Over ILIE-0.2, the light-off temperature of normal
SCR was as much as 240 K higher than that of fast SCR. Over the
remaining catalysts, this difference was only 135 K.
Obviously, the performance in fast SCR is not correlated with
the iron content; even 0.2 wt% Fe was sufficient to make a cata-
lyst with very respectable conversions over the entire temperature
range. In fact, the catalysts behaved very similarly, with a strong
increase in all conversions at low temperatures passed over in
a gradual growth around 520 K. Above 700 K, NOx conversion
tended to decrease, due mainly to decreasing NO conversion. This
tendency is known from standard SCR [21,28] where it has been
ascribed to oxidation of the NH3 reductant, particularly over clus-
tered sites. The NOx conversion curves of all three ILIE catalysts
were identical within the limits of experimental error below 773 K,
with a light-off temperature around 480 K. Above 773 K, NOx con-
version remained somewhat higher with ILIE-0.2, which is almost
void of clustered iron species. It should be noted that the diver-

M. Schwidder et al. / Journal of Catalysis 259 (2008) 96–103
99
Fig. 2. Comparison of standard SCR and fast SCR over different Fe-ZSM-5 catalysts: (a) ILIE-0.2, ꢀT50 = 240 K; (b) ILIE-0.6, ꢀT50 = 135 K; (c) ILIE-1.2, ꢀT50 = 135 K.
Conditions: 1000 ppm NOx, 1000 ppm NH3, 2% O2, at 750,000 h−1, NOx being NO (standard SCR) or an equimolar NO/NO2 mixture (fast SCR). Divergence of NO and NO2
conversions below 650 K in fast SCR possibly measurement artifact (cf. text).
gence between NO and NO2 conversions at 520–670 K may be
an experimental artifact (see Section 2); thus, the corresponding
curves are dashed in this region. The overall NOx conversion data
are reliable, however. Indeed, the NOx conversions were always
close to the NH3 conversions in this temperature range, where NH3
oxidation is known to be negligible over these catalysts [21,28].
In the standard SCR of NO, the differences between the cata-
lysts were more pronounced (Fig. 2). At higher Fe content (e.g.,
ILIE-1.2), the NO conversions increased at much lower tempera-
tures than with ILIE-0.2; thus, the difference between the light-off
temperatures of ILIE-0.2 and ILIE-1.2 was as much as 100 K. No-
tably, NO and NH3 conversions were identical over ILIE-0.2 up to
823 K, only at 873 K, ammonia oxidation was observed over the
largely isolated Fe sites of this catalyst. Over the remaining ILIE
catalysts, NH3 conversions exceeded those of NO (i.e., NH3 was ox-
idized to N2) already at 673 K. Analogously, the NO conversions
began to decline at much lower temperatures over the latter cata-
lysts than over ILIE-0.2.
These observations confirm earlier conclusions that standard
SCR is supported both by isolated and oligomeric Fe oxo species,
but with a greater intrinsic activity of the latter [21,22]. Indeed,
standard SCR did proceed over ILIE-0.2 (Fig. 2a), which contains
almost exclusively isolated sites, but it was much more rapid over
the remaining catalysts, which contain oligomeric clusters as well.
On the other hand, isolated Fe oxo species appeared to be most ef-
fective for fast SCR; the addition of more iron sites to the isolated
entities in ILIE-0.2 had no effect on this reaction, except for a de-
terioration at higher temperatures due to ammonia oxidation over
clustered species (Fig. 2). At the same time, it is obvious that not
all isolated sites were involved in the fast SCR reaction. Their abso-
lute abundance was greater In ILIE-0.6 and ILIE-1.2 compared with
in ILIE-0.2 (vide supra), which had no effect on NOx conversion at
low temperatures. Isolated Fe sites are known to exist in different
coordinations, which seem to contribute differently to the catalytic
behavior of Fe-ZSM-5 in standard SCR [21,22]. Similar differences
between isolated sites most likely apply for the fast SCR reaction
as well.
competitive with ILIE-0.2. Its light-off temperature was 75 K higher
than that of the latter, and its peak NOx conversion was 83%, com-
pared with 96% for the latter. A similar observation was reported
previously [7]. Because the H-ZSM-5 used here had a nonnegligible
Fe content (ca. 0.05 wt%), one could doubt whether the fast SCR
conversions shown in Fig. 3a were caused by reactions on acidic
sites, as proposed previously [17,19]. On the other hand, there ex-
ists a noncatalytic fast SCR reaction path (over nonacidic Ba-Y [18]),
and earlier work by our group found that the Fe species in our H-
ZSM-5 were clustered rather than isolated [28] and thus should be
poor sites for fast SCR. Irrespective of the sites involved in fast SCR
over H-ZSM-5, their contribution can be considered minor com-
pared with the activity of isolated Fe sites deliberately introduced
into the zeolite.
Over H-ZSM-5, ammonia conversion equaled NOx conversion up
to 723 K, at which point the declining NOx conversion indicated
the onset of ammonia oxidation. This decline of NOx conversion
above 723 K most likely was due to the considerable oxidation
activity of the Fe oxo clusters contained in the zeolite [28], as
confirmed by the steep onset of NH3 conversion in standard SCR
(Fig. 3a). Significantly, NH3 conversion did not reach 100% at high
temperatures despite the increasing NH3 oxidation. A weaker tem-
perature dependence of the fast SCR proceeding on acidic sites
(compared with the Fe-catalyzed reaction) may explain this find-
ing.
Fig. 3 also reports activity data for two catalysts prepared via
dry interaction of FeCl3 with H-ZSM-5. Although the general course
of the conversion-temperature curve in fast SCR was very similar
to those for the ILIE samples (Fig. 2), the NOx conversions achieved
at low temperatures were somewhat higher. Over SSIE-1.1 (Fig. 3b),
NOx conversion exceeded 70% already at 473 K, whereas with
ILIE-1.2, it remained well below 40% at this temperature (Fig. 2c).
CVD-5 achieved 68% NOx conversion at 473 K; however, at higher
temperatures, the NOx conversion curves became rather similar
to those of the ILIE samples and remained at 80%–95%, whereas
above 800 K, the NOx conversion over SSIE-1.1 and CVD-5 dropped
below that of ILIE-1.2. In standard SCR, SSIE-1.1 was less active
than ILIE-1.2, apparently due to its higher content of isolated Fe
sites (Fig. 1b). Its light-off temperature was 640 K (vs 620 K), it
achieved a peak conversion of 77% (vs 82%), and it had a signifi-
cantly greater tendency toward ammonia oxidation. Therefore, the
difference in light-off temperatures between standard and fast SCR
(195 K) was rather high with SSIE-1.1, possibly reflecting the im-
The surprising activity of a catalyst with only 0.2 wt% Fe may
suggest that the fast SCR proceeds just on acidic sites without in-
volving iron, which is inferred by some SCR mechanisms proposed
in the literature [17,19]. Catalytic data for the parent H-ZSM-5 are
given in Fig. 3a. H-ZSM-5, which was very poor in standard SCR,
exhibited a remarkable performance in fast SCR but without being

100
M. Schwidder et al. / Journal of Catalysis 259 (2008) 96–103
Fig. 3. Comparison of standard SCR and fast SCR over H-ZSM-5 and over Fe-ZSM-5 made by dry interaction of FeCl3 with H-ZSM-5: (a) H-ZSM-5; (b) SSIE-1.1, ꢀT50 = 195 K;
(c) CVD-5, ꢀT50 = 170 K. Conditions: 1000 ppm NOx, 1000 ppm NH3, 2% O2, at 750,000 h−1, NOx being NO (standard SCR) or an equimolar NO/NO2 mixture (fast SCR).
Divergence of NO and NO2 conversions below 650 K in fast SCR possibly measurement artifact (cf. text).
portance of the isolated sites to its catalytic behavior. The activity
of CVD-5 in standard SCR was of comparable to that of ILIE-1.2.
Whereas the latter observations confirm that fast SCR and stan-
dard SCR rely on different actives sites, the improved fast SCR
activity through the use of a different preparation procedure is an
encouraging result. From the ILIE series, one might conclude that
the amount of active sites for fast SCR cannot be increased above
the level in ILIE-0.2; however, this applies apparently only to this
specific preparation route. The preparations involving FeCl3 vapors
seemed to achieve a higher population of the active site or to cre-
ate another site that was not accessible via ILIE. At present, the
nature of this site can only be speculated. From the UV–vis spec-
tra shown in Fig. 1, it is obvious that in the ILIE series, the relative
contribution of the signal at ≈240 nm (which most likely arises
Fig. 4. UV–vis spectra (absolute Kubelka–Munk function) of Fe-ZSM-5 catalysts pre-
pared by solid-state ion exchange.
from tetrahedrally coordinated sites [27,28]) decreased with in-
creasing Fe content, whereas the intensity in all other wavelength
regions increased. Thus, because the abundance of active sites ap-
of isolated sites indeed parallels the total Fe content. On the other
peared to remain constant in this series, one might be tempted to
hand, extensive clustering of Fe sites also is obvious in these sam-
search for the active site among these isolated tetrahedral sites. In-
ples. A quite unexpected finding is that this occurred at Fe content
deed, SSIE-1.1, which is more active than ILIE-1.2, exhibited more
intensity in the 240 nm region (Fig. 1b). Other sites cannot be
ruled out, however. Thus, the prominent band at 360 nm, assigned
to Fe oxo trimers by Pirngruber et al. [29], seems to be completely
+
as low as 0.2 wt% where the route via mononuclear FeCl2 in-
termediates should have led to a site structure similar to that of
ILIE-0.2 (Fig. 1b). Instead, significant intensity was seen even be-
yond 400 nm, and the 360-nm band of Fe oxo oligomers (trimers?)
missing in the ILIE preparations and thus might arise from a can-
was well developed. A tentative explanation for this finding refers
didate species.
to an observation of a study in which CVD was performed in a flu-
idized bed instead of a fixed bed to prepare partially exchanged
Fe-ZSM-5 [31]; the materials thus obtained were strongly surface-
3.3. Optimization of Fe-ZSM-5 for fast SCR
enriched in Fe even at low Fe content, as detected by XPS, sug-
The results presented so far suggest that dry interaction with
gesting that the reaction between H-ZSM-5 and FeCl3 proceeded
in zones not only in a fixed bed, but also in a crystallite. There-
fore, local Fe concentrations, and thus clustering tendencies, might
be high in all cases.
Fig. 5 illustrates the temperature dependence of total NOx con-
version measured over the catalysts described above. The low-
temperature conversion data show that the activity increased
monotonously with the Fe content in the samples. NOx conver-
sions of almost 90% were achieved at temperatures as low as
423 K (SSIE-3.2), and the very low conversions measured with
SSIE-0.2 in the same type of experiment proves without the need
for further explanation that noncatalytic NH4NO3 formation did
not contribute to this conversion to any significant extent. On the
FeCl3 may be a promising preparation route, resulting in activi-
ties that exceeded those obtained with the ILIE series already at
1.1 wt% Fe. Although the example of CVD-5 may subdue the opti-
mism, because this catalyst is similar to SSIE-1.1 despite its much
higher Fe content, a series with Fe concentrations ranging from 0.2
to 3.2 wt% was prepared by SSIE. According to the widely accepted
concepts of gas-phase preparation of Fe-ZSM-5 from FeCl3 and H-
ZSM-5, mononuclear Z FeCl2 species are formed in the first step
and are hydrolyzed on washing. Although these isolated species
may become mobile and aggregate as in other preparations [30],
the greater abundance of isolated sites in SSIE-1.1 compared with
ILIE-1.2 (Fig. 1b) was the rationale behind this attempt.
+
−
Fig. 4 displays the UV–vis spectra of these catalysts (SSIE-1.1
included, Kubelka–Munk function not normalized). From the in-
tensity trend below 300 nm, it is obvious that the concentration
other hand, NH3
oxidation activity increased dramatically above
2 wt% Fe, and the useful temperature range became very narrow

M. Schwidder et al. / Journal of Catalysis 259 (2008) 96–103
101
ther work which is needed to identify the active sites for fast SCR
more precisely, should concentrate on the various isolated (extra-
framework) Fe sites in the Fe zeolites.
3.4. Fast and standard SCR: durability
Fig. 6 characterizes the affects of aging on the two reactions
under study, showing the catalytic behavior of ILIE-1.2 after two
different aging treatments. The first of these treatments involved
aging for 2 days in moist air (10% H2O) at 923 K (Fig. 6a). This
clearly led to a significant loss of activity in standard SCR, although
the slightly increased conversions below 473 K are somewhat sus-
picious. The NH3 conversions exceeded the NO conversions at a
lower temperature than occurred in the fresh state, indicating
clustering of isolated and/or growth of oligomeric species in the
sample. Fast SCR was hardly affected by this hydrothermal stress,
which was found for other Fe-ZSM-5 preparations (e.g., SSIE-1.1).
Fig. 6b shows the result of an aging experiment in the presence
of SO2 (5 wt% H2O, 200 ppm SO2 at 823 K for 65 h). Here the
damage in standard SCR activity was much greater. At low tem-
peratures, NO conversion increased as in the fresh state, but at
623 K, the NH3 conversion curve dropped sharply and even fell
below zero due to drastically increased ammonia oxidation ac-
tivity. Apparently, SO2 accelerated the aforementioned clustering
processes significantly. Remarkably, in fast SCR, the impact of this
harsh treatment was not seen at low temperatures. Only between
550 and 670 K were the NO conversions somewhat short of those
measured with the fresh sample.
The very different response of standard and fast SCR to the ag-
ing treatments confirms that both reactions proceeded on different
sites. The strong changes in ammonia oxidation activity seen in the
standard SCR measurements hardly affected the fast SCR reaction,
which was much more rapid and apparently won the competition
even with the accelerated NH3 oxidation reaction. Isolated sites
are known to be inactive in ammonia oxidation [21] [see also the
parallelism of NO and NH3 activities in ILIE-0.2 (Fig. 2a)]; there-
fore, the increased ammonia oxidation activity indicates clustering
of isolated sites and/or restructuring of clusters into a size favor-
ing the oxidation activity at the expense of standard SCR activity.
Fast SCR was little affected by these changes, in agreement with
our earlier conclusion that it proceeds in isolated sites. Moreover,
Fig. 5. Fast SCR over Fe-ZSM-5 catalysts prepared by solid-state ion exchange: tem-
perature dependence of total NOx conversion.
(SSIE-2.5). However, with the optimum Fe content of 2.0 wt%, NOx
conversions >90% were achieved in a broad temperature range of
473–773 K.
The enhanced activity of the SSIE catalysts (Fig. 5) correlates
well with the increasing amount of isolated sites in these catalysts
(Fig. 4); however, the abundance of other sites also grew with in-
creasing iron content. Although clustered sites are undesirable due
to their ammonia oxidation activity, they might catalyze fast SCR
as well as they do standard SCR [21]; however, comparing the cat-
alytic behavior and UV–vis spectra of ILIE-0.2 (Figs. 1a and 2a) and
SSIE-0.2 (Figs. 4 and 5) supports the notion that fast SCR required
only isolated sites. SSIE-0.2 clearly contained a significant amount
of clustered species, among them the oligomer giving rise to the
360-nm UV–vis band. Consequently, ILIE-0.2 had a greater abun-
dance of isolated sites. If the new site that caused the superior
activity of the SSIE preparations had been a clustered site (e.g.,
the site absorbing at 360 nm), then SSIE-0.2 should have been
more active than ILIE-0.2. This clearly was not the case, however;
instead, SSIE-0.2 was slightly inferior to ILIE-0.2. Consequently, fur-
Fig. 6. Deactivation of Fe-ZSM-5 (ILIE-1.2) in standard and in fast SCR: (a) ageing in moist atmosphere, 10 vol% H2O in air, 923 K, 48 h; (b) ageing in presence of water and
SO2, 5% H2O, 200 ppm SO2 in He, at 823 K over 65 h. Measurement conditions as in Figs. 1 and 2.

102
M. Schwidder et al. / Journal of Catalysis 259 (2008) 96–103
Fig. 7. Conversion of NO in NO oxidation and in standard SCR: (a) Temperature dependence of NO conversion (standard SCR data taken from Figs. 2 and 3), 1000 ppm NO,
2% O2 in He at 750,000 h−1, (") ILIE-0.2, ( ) ILIE-1.2, ( ) SSIE-1.1, (!) CVD-5. (b) Isothermal comparison between NO oxidation and standard SCR over SSIE-1.1, T = 573 K,
initial feed 1000 ppm NO, 2% O2 in He at 750,000 h−1, first switch 1000 ppm NH3 added, second switch NH3 removed, third switch NO removed, (d) isothermal comparison
between NO oxidation and fast SCR over SSIE-1.1, T = 573 K, initial feed 500 ppm NO, 500 ppm NO2, 2% O2 in He at 750,000 h−1, first switch 1000 ppm NH3 added, second
switch NH3 removed. Concentrations measured by non-dispersive IR photometry given without corrections for cross-sensitivities.
those isolated sites that catalyze fast SCR appeared to be very sta-
ble under realistic conditions of flue gas catalysis.
which would lead to an acceleration of the reaction at moderate
decreases in the ammonia surface concentration. In the analogous
fast SCR experiment (Fig. 7c), the NO/NO2 mixture left the cata-
lyst equimolar in the absence of NH3, again indicating the minor
NO oxidation conversion. The apparent NH3 effluent concentration
of almost 300 ppm resulted from a cross-sensitivity effect by NO2.
Adding NH3 to the feed removed the NOx from the effluent almost
completely, with the concentrations of NO and NO2 matching over
the entire period. When NH3 was switched off, the NOx effluent
concentrations returned to the feed values without deviating from
the 1:1 NO:NO2 ratio.
3.5. Mechanistic considerations
As mentioned in Section 1, NO oxidation has been considered
a crucial step in the reaction mechanism of NO reduction. There-
fore, we measured NO conversions to NO2 over our catalysts at the
same gas-phase concentrations and space velocity as the SCR data
presented earlier, that is, 1000 ppm NO and 2% O2 in He (NH3
missing) at 750,000 h−1. The results are shown in Fig. 7a. For
comparison, the NO conversion curves obtained for the standard
SCR are repeated as thin lines. The NO oxidation measurements
were not very accurate; NO conversions remained very low at the
high space velocity applied, as confirmed by the small NO2 yields
found. It appears that samples with a greater degree of clustering
(CVD-5, ILIE-1.2) were somewhat more active than those with pre-
dominantly isolated Fe sites, but the most important result is that
the rates of NO oxidation were inferior to those of standard SCR
for all catalysts studied.
For the SSIE-1.1 sample, this finding was confirmed by direct
comparison using a different measurement technique (nondisper-
sive IR photometry). In these runs, conducted at 573 K, the am-
monia reductant was removed from and switched into the feed
without changing the GHSV. For standard SCR (Fig. 7b), this is
equivalent to a change from standard SCR feed to NO oxidation
feed and vice versa. For fast SCR, the feed without NH3 is an
oxygen-containing NO2/NO (1:1) mixture in which the NO2/NO
ratio should increase over a catalyst of sufficient NO oxidation
activity (cf. equilibrium curve in Fig. 7a). Fig. 7b shows that the
NO oxidation conversion indeed was below the detection limit,
with an apparent NO2 effluent concentration of ca. 20 ppm due to
cross-sensitivity from NO. The NO outlet concentration decreased
markedly after addition of the ammonia reductant, demonstrat-
ing the superiority of the SCR rate over the NO oxidation rate.
When NH3 was removed from the feed, the NO conversion in-
creased temporarily before gradually returning to the low value
characteristic of NO oxidation. The transient increase in NO con-
version may be due to a negative reaction order of NH3 [7,9,24],
Fig. 7 raises serious doubt that NO oxidation was the rate-
limiting step avoided by adding NO2 to the gas phase. This reaction
was significantly slower than NO reduction over the catalysts stud-
ied, and is therefore not a candidate for an SCR reaction scheme.
Our findings do not rule out adsorbed higher oxidized NOx species
as SCR intermediates, but the participation of free NO2 in standard
SCR is unlikely. It has been argued that NO2 desorption might be
the rate-limiting step in NO oxidation over Fe ions. But, in terms
of the NO oxidation mechanism over Fe3+/Fe2+ ions summarized
by Brandenberger et al. [9] this means that NO2 would accumulate
at Fe2+ sites in the stationary state, which seems not very likely
for the typical reaction conditions for fast SCR applications.
Consequently, we believe that other mechanisms that operate
without free NO2 (e.g., the amide-nitrosamide mechanism [12,13],
as well as routes that avoid the N4+ oxidation state [6]), deserve
more attention. The effect of an NO2 admixture on the gas phase
likely is not so much to shift the rate-determining step in a given
reaction sequence, but rather to open up a completely new reac-
tion mechanism. A less efficient SCR mechanism, possibly via Fe
amide species, is outperformed by a highly efficient route involv-
ing both NO and NO2. The fast SCR mechanisms proposed for Fe
catalysts in the literature do not involve a catalytic role for iron,
but differ in terms of the role of Brønsted sites. Their direct par-
ticipation has been suggested previously [17], where a reaction
+
between NH4 and NO2 opens the reaction sequence. According to
another report [18], NH4NO2 is formed and rapidly decomposed
noncatalytically, but a participation of Brønsted sites in the decom-
position has been proposed later [19]. Thus, these mechanisms can

M. Schwidder et al. / Journal of Catalysis 259 (2008) 96–103
103
explain only our data for H-ZSM-5 (Fig. 3a). The clear evidence of
References
the participation of Fe sites (Fig. 2a compared with Fig. 3a) demon-
strates that more work is needed to identify the actual reaction
mechanism.
[1] M. Iwamoto, H. Yahiro, Y. Mine, S. Kagawa, Chem. Lett. (1989) 213.
[2] W. Held, A. König, T. Richter, L. Puppe, SAE Paper 900496 (1990).
[3] M. Koebel, M. Elsener, M. Kleemann, Catal. Today 59 (2000) 335.
[4] A. Kato, S. Matsuda, F. Nakajima, H. Kuroda, T. Narita, J. Phys. Chem. 85 (1981)
4099.
4. Conclusion
[5] G. Tuenter, W. Leeuwen, L. Snepvangers, Ind. Eng. Chem. Prod. Res. Dev. 25
(1986) 633.
[6] Q. Sun, Z.X. Gao, H.-Y. Chen, W.M.H. Sachtler, J. Catal. 201 (2001) 89.
[7] M. Devadas, O. Kröcher, M. Elsener, A. Wokaun, N. Söger, M. Pfeifer, Y. Demel,
L. Mussmann, Appl. Catal. B 67 (2006) 187.
[8] K. Rahkamaa-Tolonen, T. Maunula, M. Lomma, M. Huuhtanen, R.L. Keiski, Catal.
Today 100 (2005) 217.
[9] S. Brandenberger, O. Kröcher, A. Tissler, R. Althoff, Catal. Rev. Sci. Eng. (2008),
in press.
[10] N.-Y. Topsoe, J.A. Dumesic, H. Topsoe, J. Catal. 151 (1995) 241.
[11] B. Coq, G. Delahay, R. Durand, D. Berthomieu, E. Ayala-Villagomez, J. Phys.
Chem. B 108 (2004) 11062.
[12] G. Ramis, G. Busca, F. Bregani, P. Forzatti, Appl. Catal. 64 (1990) 2259.
[13] G. Ramis, L. Yi, G. Busca, M. Turco, E. Kotur, R.J. Willey, J. Catal. 157 (1995)
523.
[14] N. Apostolescu, B. Geiger, K. Hizbullah, M.T. Jan, S. Kureti, D. Reichert, F. Schott,
W. Weisweiler, Appl. Catal. B 62 (2006) 104.
[15] E. Tronconi, I. Nova, C. Ciardelli, D. Chatterjee, M. Weibel, J. Catal. 245 (2007) 1.
[16] G. Delahay, S. Kieger, N. Tanchoux, P. Trens, B. Coq, Appl. Catal. B 52 (2004) 2.
[17] R.Q. Long, R.T. Yang, J. Catal. 207 (2002) 224.
[18] Y.H. Yeom, J. Henao, M.J. Li, W.M.H. Sachtler, E. Weitz, J. Catal. 231 (2005)
181.
[19] M.J. Li, Y. Yeom, E. Weitz, W.M.H. Sachtler, Catal. Lett. 112 (2006) 129.
[20] M. Schwidder, M. Santhosh Kumar, A. Brückner, W. Grünert, Chem. Commun.
(2005) 805.
[21] M. Schwidder, M. Santhosh Kumar, K.V. Klementiev, M.M. Pohl, A. Brückner,
W. Grünert, J. Catal. 231 (2005) 314.
[22] M. Santhosh Kumar, M. Schwidder, W. Grünert, U. Bentrup, A. Brückner,
J. Catal. 239 (2006) 173.
[23] F. Heinrich, C. Schmidt, E. Löffler, M. Menzel, W. Grünert, J. Catal. 212 (2002)
157.
[24] O. Kröcher, M. Devadas, M. Elsener, A. Wokaun, N. Söger, M. Pfeifer, Y. Demel,
L. Mussmann, Appl. Catal. B 66 (2006) 208.
[25] M. Koebel, G. Madia, M. Elsener, Catal. Today 73 (2002) 239.
[26] I. Nova, C. Ciardelli, E. Tronconi, D. Chatterjee, B. Bandl-Konrad, Catal. Today 114
(2006) 3.
The SCR by NH3 of NO (standard SCR) and of NO/NO2 mix-
tures (fast SCR) over Fe-ZSM-5 was found to proceed over differ-
ent Fe sites. Whereas standard SCR proceeded on both isolated
and oligomeric Fe oxo sites, fast SCR appeared to require only
isolated Fe sites, with a small site concentration rendering very
high reaction rates. Therefore, the response of Fe-ZSM-5 catalysts
to conditions typical of flue gas catalysis (i.e., moist feeds, high
temperatures) differed as well. Fast SCR was hardly affected un-
der conditions that strongly deactivated the catalyst for standard
SCR and caused significant enhancement of NH3 oxidation activity,
indicating the high stability of those isolated sites responsible for
fast SCR. The potential of preparation procedures to improve the
activity by increasing the Fe content varied widely. Whereas aque-
ous exchange H-ZSM-5 with Fe2+ ions rendered the best activity
at only 0.2 wt% Fe, much better performance could be achieved by
increasing the Fe content via solid-state ion exchange with FeCl3,
the best catalyst, giving NOx conversions >90% at 473–773 K. At
high Fe content, clustered sites, which could not be avoided in this
preparation, led to decreased NOx conversion at rather low tem-
peratures, due to competing ammonia oxidation activity.
The standard SCR reaction was significantly faster than NO ox-
idation. Thus, free NO2 produced by NO oxidation was not an
intermediate of standard SCR. The acceleration of standard SCR by
addition of NO2 to the feed most likely can be interpreted as a
switch between different reaction channels, the slower one with-
out participation of NO2 and the rapid one (which likely is possible
on Brønsted sites alone but is strongly accelerated over isolated Fe
oxo sites) involving NO2 and Fe sites.
[27] S. Bordiga, R. Buzzoni, F. Geobaldo, C. Lamberti, E. Giamello, A. Zecchina, G. Le-
ofanti, G. Petrini, G. Tozzola, G. Vlaic, J. Catal. 158 (1996) 486.
[28] M. Santhosh Kumar, M. Schwidder, W. Grünert, A. Brückner, J. Catal. 227 (2004)
384.
[29] G.D. Pirngruber, P.K. Roy, R. Prins, Phys. Chem. Chem. Phys. 8 (2006) 3939.
[30] H.-Y. Chen, W.M.H. Sachtler, Catal. Today 42 (1998) 73.
[31] M. Schwidder, Ph.D. thesis, Bochum, 2004.
Acknowledgments
The authors thank Ms. M. Holz and Ms. A. Rottmann for assis-
tance with catalyst preparation, and Drs. E. Tronconi and O. Krö-
cher for valuable discussions, in particular for access to the as-yet
unpublished review [9] by Dr. Kröcher.