Nature and catalytic role of active silver species in the lean NO x reduction with C3H6 in the presence of water

Article abstract of DOI:10.1016/S0021-9517(03)00055-1

A study of the lean NO_x reduction activity with propene in the presence of water over Ag/Al₂O₃ catalysts with different silver loadings (1.5-6 wt%) has been done using X-ray diffraction, ultraviolet-visible spectroscopy, transmission electron microscopy, and in situ diffuse reflectance infrared and X-ray absorption spectroscopies under reaction conditions. The catalysts were prepared by an impregnation method employing EDTA complexes that allow highly dispersed silver phases to be obtained, which are stabilized under reaction conditions by strong interactions with the support. It is shown that the active species corresponds to silver aluminate-like phases with tetrahedral local symmetry. The role of silver in the reaction mechanism is shown to be mainly in the activation of NO _x and propene species. In particular, the silver entities have been found to offer a new reaction path for propene activation which involves generation of acrylate species as a partially oxidized active intermediate. Differences between two active catalysts containing 1.5 and 4.5 wt% of Ag suggest that optimization of the SCR activity can be related to the oxygen lability of the tetrahedral silver aluminate-like phase present in the catalyst. As postulated previously, the high nonselective propene oxidation activity of the highest loaded sample (with 6 wt% Ag) appears to be related to formation of metallic silver surface states at low reaction temperatures which are active for NO dissociation.

Full text of DOI:10.1016/S0021-9517(03)00055-1

Journal of Catalysis 217 (2003) 310–323
www.elsevier.com/locate/jcat
Nature and catalytic role of active silver species in the lean NOx reduction
with C3H6 in the presence of water
A. Iglesias-Juez,a A.B. Hungría,a A. Martínez-Arias,a A. Fuerte,a M. Fernández-García,a,∗
J.A. Anderson,b J.C. Conesa,a and J. Soria a
a
Instituto de Catálisis y Petroleoquímica, CSIC, Campus Cantoblanco, 28049 Madrid, Spain
b
Division of Physical and Inorganic Chemistry, University of Dundee, Dundee DD1 4HN, Scotland, UK
Received 24 September 2002; revised 23 January 2003; accepted 3 February 2003
Abstract
A study of the lean NO reduction activity with propene in the presence of water over Ag/Al O catalysts with different silver loadings
x
2
3
(1.5–6 wt%) has been done using X-ray diffraction, ultraviolet–visible spectroscopy, transmission electron microscopy, and in situ diffuse
reflectance infrared and X-ray absorption spectroscopies under reaction conditions. The catalysts were prepared by an impregnation method
employing EDTA complexes that allow highly dispersed silver phases to be obtained, which are stabilized under reaction conditions by
strong interactions with the support. It is shown that the active species corresponds to silver aluminate-like phases with tetrahedral local
symmetry. The role of silver in the reaction mechanism is shown to be mainly in the activation of NO and propene species. In particular,
x
the silver entities have been found to offer a new reaction path for propene activation which involves generation of acrylate species as a
partially oxidized active intermediate. Differences between two active catalysts containing 1.5 and 4.5 wt% of Ag suggest that optimization
of the SCR activity can be related to the oxygen lability of the tetrahedral silver aluminate-like phase present in the catalyst. As postulated
previously, the high nonselective propene oxidation activity of the highest loaded sample (with 6 wt% Ag) appears to be related to formation
of metallic silver surface states at low reaction temperatures which are active for NO dissociation.
2003 Elsevier Science (USA). All rights reserved.
Keywords: NO -propene SCR; Ag/Al O catalysts; In situ DRIFTS; XAFS; UV–vis; XRD; TEM
x
2
3
1. Introduction
although at a higher temperature window than noble metal
systems, for the lean NO reduction with C3H6 in the pres-
ence of water and all of these apparently operate through a
very similar reaction mechanism [4,8,9]. The catalytic activ-
ity of these materials is maintained at a relatively high level
even in the presence of H2O and SO2 [8,10–13]. Higher NO
conversions at lower temperatures are generally observed
when using oxygenated hydrocarbons, such as ethanol and
acetone, as reductants instead of propene [8]. Unfortunately,
harmful by-products, such as hydrogen cyanide, can be
formed when using such reductants [14].
The effectiveness of the Ag/Al2O3 catalyst depends
greatly on the Ag loading, the best catalytic performance
being achieved with intermediate loadings (2–3%, for sys-
tems employing typical high surface area γ -Al2O3 mate-
rials as support) [8,15–18]. Higher silver loadings produce
higher rates of C3H6 oxidation with O2, at the expense of
its reaction with NO. This has been attributed to the detri-
mental effect of increasing the particle size of silver oxide
Selective catalytic reduction (SCR) of NOx in the oxy-
gen-rich exhaust streams of lean-burn engines remains one
of the major challenges for environmental catalysis. Cu-
ZSM-5 was the first catalyst displaying good lean NOx re-
duction activity [1]; however, zeolite-based systems lack sta-
bility in the presence of water or sulfur dioxide [2]. The
best results in this respect have been observed when using
alumina-supported systems [3–5]. Among these, Pt and Pd
catalysts have shown high activity for the reaction, present-
ing NOx conversion windows at much lower temperatures
than zeolite-based catalysts [6]. They are, however, highly
selective toward N2O, an undesirable greenhouse gas [7].
Alumina-supported nonnoble metal oxide systems (among
others, those of tin, cobalt, and silver) are effective catalysts,
*
Corresponding author.
E-mail address: m.fernandez@icp.csic.es (M. Fernández-García).
0021-9517/03/$ – see front matter
doi:10.1016/S0021-9517(03)00055-1
2003 Elsevier Science (USA). All rights reserved.

A. Iglesias-Juez et al. / Journal of Catalysis 217 (2003) 310–323
311
species, yielding reducible entities (leading to Ag0) more
readily with a higher activity for nonselective propene ox-
idation [15–18]. It has been proposed that for size-limited
clusters, silver can be stabilized in an oxidized (Ag+) state,
as a consequence of the interaction with the alumina sup-
port, which would constitute the active centers for the NOx
reduction reaction [15]. This theory, however, has recently
been challenged [17]. In this respect, an Ag/Al2O3 catalyst
has been reported to show markedly improved performance
after an XRD-detectable silver aluminate phase was gener-
ated by hydrothermaltreatment at 1073 K [19]. The presence
of silver aluminates has also been reported as part of the ac-
tive metal chemical distribution in other alumina-supported
samples [17,19]. It may also be recalled that, when present
as exchanged cations in zeolites, i.e., a situation where large
particles of oxide or metal are less readily formed, NOx con-
version (using CH4 as reductant) does not decrease with Ag
loading below ca. 10 wt% [5,20].
The aim of the present work is to study the catalytic
activity for lean NO reduction with C3H6 in the presence of
water over Ag/Al2O3 catalysts prepared by an impregnation
method designed to maximize the amount of active silver
phases. Emphasis will be placed on the structural analysis of
the silver phases of most active samples. For these purposes,
a multi-technique approach has been followed using X-
ray diffraction (XRD), transmission electron microscopy
(TEM), UV–vis spectroscopy for catalyst characterization,
and employing in situ diffuse reflectance infrared Fourier
transform (DRIFTS), and X-ray absorption spectroscopies
(XAFS) to gain insight into the redox and chemical changes
occurring during the course of the reaction as well to
gain key information concerning the activation of reactant
molecules and the interaction with adsorbed species.
showed values similar to that of the parent alumina for all
the samples.
Powder XRD patterns were recorded on a Siemens D-500
diffractometer using nickel-filtered Cu-Kα radiation operat-
ing at 40 kV and 25 mA and with a 0.02◦ step size. Sam-
ples for transmission electron microscopy were prepared by
crushing in an agate mortar, dispersing in isobutanol, and
depositing on perforated carbon films supported on copper
grids. TEM data were obtained on a JEOL 2000 FX II sys-
tem (with 3.1 Å point resolution) equipped with a LINK
probe for energy-dispersive analysis (EDX). Around 10 ag-
gregates and 2 to 4 zones per aggregate were analyzed for
each sample. A Varian UV–vis spectrophotometer, Model
2300, was used for the recording UV–vis absorption spectra
(transmittance) of bulk samples of squared shape and 0.5 cm
of path length.
Catalytic activity tests were performed in a Pyrex flow
reactor system using a feed stream consisting of 0.1%
NO, 0.1% C3H6, 5% O2, and 3% H2O (N2 balance; ca.
1 × 103 cm3 min−1 g−1 at a fix space velocity of 4 ×
104 h−1). The effluent from the reactor, with the exception
of H2O which was (mostly) removed by a cold trap (NaCl–
ice mixture), was analyzed on line using a Perkin-Elmer
1725X FTIR spectrometer coupled with a multiple reflection
transmission cell (Infrared Analysis Inc.). In all cases the
samples were subjected to an in situ calcination pretreatment
under dry air at 773 K. Typical tests were performed in the
light-off mode by increasing the temperature at 5 K min−1
from 298 to 823 K.
In situ DRIFTS experiments were performed using a
Perkin-Elmer 1750 FTIR fitted with an MCT detector. The
DRIFTS cell (Harrick) was equipped with CaF2 windows
and a heating cartridge that allowed samples to be heated up
to 773 K. The spectral resolution employed was 4 cm−1
.
Samples of ca. 80 mg were calcined in situ in dry air
at 773 K. The spectra were taken by accumulating 25
scans, except where noted otherwise. The composition of
the inlet gases was manipul−at1ed by a computer-controlled
gas blender with 80 cm3 min passing through the catalyst
bed, and with the NOx gases in the outlet from the DRIFTS
cell monitored by a chemiluminiscence analyzer (Thermo
Environmental Instruments 42C).
In situ X-ray absorption experiments at the Ag K-edge
were performed at the BM29 line of the ESRF synchrotron
at Grenoble. A Si(111) monochromator, detuned at 50%
to minimize harmonic content of the beam, and ionization
chambers filled with Kr/Ar were used to obtain the trans-
mission XAFS data. An Ag foil between the second and
a third ionization chamber was measured simultaneously
to calibrate the energy scale. Self-supported wafers (ab-
sorbance below 0.5) were placed in a controlled-atmosphere
cell for treatment under controlled atmosphere and temper-
ature. EXAFS spectra were obtained using standard proce-
dures; preedge substraction was performed by using a Vic-
toreen formula, postedge substraction using a cubic spline
procedure, and normalization by fixing to one the value of a
2. Experimental
Alumina-supported silver catalysts of different Ag load-
ings were prepared using AgNO3 (supplied by Aldrich) as
silver precursor and Al2O3 (Condea, SBET = 200 m2 g−1
)
as support. Aqueous solutions containing different AgNO3
concentrations (in order to obtain final nominal silver load-
ings of 1.5, 4.5, and 6 wt%) were first mixed with EDTA
in 1:2.2 molar ratio (aiming to produce negatively charged
silver complexes, which may interact with the positively
charged surface alumina at pH about 6.3—obtained by using
tetramethylammonium hydroxide (TMAH) as base of the
buffer system) and subsequently used to impregnate the alu-
mina support by the incipient wetness technique. The solids
were then dried for a short period at room temperature and
then at 383 K for 24 h and finally calcined in air at 773 K
for 2 h. Ag analyses of the samples by ICP-AES yielded val-
ues which coincide, within the error limits, with the nomi-
nal values. Samples are denoted as 1.5Ag, 4.5Ag, and 6Ag,
in accordance with the weight loadings. BET surface areas

312
A. Iglesias-Juez et al. / Journal of Catalysis 217 (2003) 310–323
postedge background point at about 50 eV from the edge.
Phase and amplitude functions from the Ag2O reference
EXAFS spectra were used to analyze Ag–O contributions in
used and freshly calcined Ag samples. Analysis of EXAFS
signals was performed over a K range from 1.95 to 8.2 Å−1
and an R range from 0.7 to 3.22 Å. This implies a num-
ber of free parameters, according to the Nyquist theorem,
of 12 [21].
3. Results
3.1. Catalytic activity tests
Fig. 1 shows the propene (A) and NOx (NO+NO2) to N2
(B) conversions, along with the yields of CO (C), NO2 (D),
and N2O (E) for the C3H6-SCR of NO in the presence of
water. No carbon-containing products other than CO2 and
Fig. 1. Conversion of propene (A) and NO (B) to nitrogen for Ag samples and alumina support. Yields to CO (C), NO (D), and N O (E) are also displayed.
x
2
2
Circles, alumina; squares, 1.5Ag; triangles, 4.5Ag; stars, 6Ag.

A. Iglesias-Juez et al. / Journal of Catalysis 217 (2003) 310–323
313
CO (e.g., oxygenated hydrocarbons) were detected in the
evolved gases, with a marked trend to decrease selectivity
to the latter product with increasing the Ag content of the
catalyst. NO, on the other hand, was partially converted to
NO2 at low temperatures (where absence of propene con-
version can be noted), mostly through the homogeneous re-
action with oxygen (Fig. 1D). Parallel to hydrocarbon con-
version, NO reduction products are detected; because N2O
production is modest (Fig. 1E) and no other N-containing
molecule (e.g., NH3) is detected, it can be said that NOx
were essentially converted to N2 in all cases except for 6Ag.
The alumina support showed low SCR activity, which can be
attributed to deactivation of the system induced by the pres-
ence of water [8], a fact never observed in Ag-containing
systems (see Ref. [18] and references therein). Certain dif-
ferences in terms of catalytic performance were noted be-
tween these catalysts and other Ag/Al2O3 systems reported
in the literature. Thus, CO selectivity has not been typi-
cally reported for catalysts prepared by impregnation with
aqueous silver nitrate solutions [7,22] while C3H6 conver-
sion levels are generally higher here than those achieved
in our laboratory with catalysts prepared by a microemul-
sion method [18]. Production of significant amounts of car-
bon monoxide, here detected for the alumina carrier and the
1.5Ag sample, appears to be a characteristic of acidic cat-
alysts [23]. Comparison of the NOx reduction activity of
these samples reveals only subtle differences at T < 600 K
at which low NOx conversions are observed; in view of
the absence of C3H6 oxidation under these conditions, the
NOx consumption can be mostly attributed to adsorption
phenomena.
Fig. 2. UV–vis spectra of Ag samples and reference compounds: (a)
α-AgAlO , (b) Ag O, (c) β-AgAlO , (d) 1.5Ag, (e) 4.5Ag, and (f) 6Ag.
2
2
2
Full line, calcined materials; dashed line, postreaction materials.
UV–vis spectra of calcined and postreaction samples and
reference compounds are shown in Fig. 2. Only minor
differences were noted among calcined samples indicating
the similar nature of the initial, supported silver species;
larger differences were, however, encountered between the
6Ag sample and the others after reaction. In the latter case,
a broad band centered at ca. 350 nm was detected, which
can be attributed to the presence of small Ag(0) clusters
[16,17]. This indicates a reasonable degree of stability of the
Ag phase in the 1.5Ag and 4.5Ag samples and the evolution
under reaction conditions for the 6Ag to yield metallic silver
particles.
The results indicate the promoting effect of silver at
intermediate loadings on the SCR reaction, in agreement
with previous reports [8,15–18]. Thus, the 1.5Ag and 4.5Ag
samples are active for NOx reduction, showing similar
NOx and C3H6 conversion profiles but shifted to lower
temperatures for the 4.5Ag. On the other hand, the catalytic
behavior of the sample with the highest silver loading
(6Ag) shows similarities with previous studies [7,16,18,
22], indicating that low NOx conversion levels are achieved
over alumina-supported samples with relatively high silver
loadings, while a relatively higher rate for the nonselective
C3H6 combustion occurs.
DRIFT spectra of the alumina and fresh 1.5Ag, 4.5Ag,
and 6Ag samples calcined in dry air at 773 K are shown in
Fig. 3. The samples show similarities in the hydroxyl stretch-
ing region with bands at 3754, 3720, 3678, and 3631 cm−1
,
3.2. Sample characterization
Samples in their calcined state were analyzed by several
techniques. ED and XRD data (not shown) of calcined
and postreaction samples only displayed alumina-related
features without significant differences between samples.
Only a limited number of amorphous (i.e., not detected
by diffraction techniques) Ag particles were visible in
some TEM micrographs. However, the number of silver
particles detected during sampling was very small compared
with previous experience of silver–alumina systems [17,18],
a fact that will be discussed below in light of XAFS results.
Fig. 3. IR spectra corresponding to the OH stretching region for samples in
their calcined state: (a) alumina, (b) 1.5Ag, (c) 4.5Ag, and (d) 6Ag.

314
A. Iglesias-Juez et al. / Journal of Catalysis 217 (2003) 310–323
and a broader one below 3600 cm−1 showing a maximum
around 3575 cm−1. These bands can all be attributed to di-
verse surface Al–OH groups. Bands in the 3800–3600 cm−1
region correspond to different isolated hydroxyl species,
their frequency being related to the nature of the Al cation
and/or to the coordination number of the corresponding hy-
droxyl group [24]. On the other hand, the broad band peak-
ing at 3575 cm−1 can be assigned to associated (mutually
interacting by hydrogen bonds) hydroxyl groups [23]. De-
creased intensity was detected upon increasing the amount
of silver deposited on alumina, which mainly affected the
1305 cm−1 (type II), that can be attributed to different
(monodentate and/or bidentate) nitrate species adsorbed on
the alumina surface [18,22,25,26]. In the case of 6Ag, bulk-
like nitrite (NO2−) species giving a band at 1300 cm−1
[22] might also be formed (Fig. 6Da,b). Similar adsorbed
NOx species have previously been observed for Ag/Al2O3
[18,22,25,27].Small differences between catalysts regarding
band frequencies for nitrate species can be ascribed to slight
differences in the local environment of adsorption sites.
Certain differences were additionally observed between
the alumina and the silver-containing samples concerning
involvement of hydroxyl species during nitrate formation
most basic (3754 cm−1) and acidic (3631 and 3575 cm−1
)
hydroxyls for the 1.5Ag sample and progressively decreased
the intensity of the remaining terminal OH groups, with
particular effect in the 6Ag case. It was noted, however,
that a general decrease in intensity for the 6Ag spectrum
occurred over the whole 3800–2800 cm−1 region, Fig. 3,
suggesting that the reflectance from the solid was different
from the other samples. However, by analyzing the 1250–
1000 cm−1 region (result not shown), characteristic of alu-
mina skeletal vibrations and the OH deformation modes
(1000–1050 cm−1), we can conclude that a change in sur-
face hydration of the samples does contribute to the (OH
region) intensity behavior noted.
OH(ad) + NO2(g) → HNO3(ad) ↔ H+(ad) + NO3
.
−
Thus, while a net consumption of acidic hydroxyls (bands
at 3632 and 3542 cm−1) was apparent for the silver-free
alumina, the hydroxyl species were less perturbed during
nitrate formation for the silver-containing samples.
Clearer differences due to the presence of silver were ob-
served in the presence of species formed upon partial oxi-
dation of the hydrocarbon, as detected following subsequent
addition of C3H6 to the flowing gas mixture (Figs. 4Ac,d
to 4Dc,d). Thus, while a group of bands at ca. 1642, ≈ 1575,
1454–1458, 1378–1392, and 1296–1298 cm−1, attributed to
vibrations of adsorbed acrylate species [18,22,25,28], were
observed for Ag-containing samples, no acrylate was appar-
ent for the alumina sample, in which case propene oxida-
tion appeared to proceed mainly to formate species (bands at
3000, 2907, 1589, 1392, and 1375 cm−1), probably via in-
teraction of CO or, less likely, of CO2 with the alumina sur-
face. Additionally, a further carboxylate-type species (giv-
ing rise to bands at ca. 1540 and 1455 cm−1) [25,26] was
observed for all the samples. It can be noted that bicarbonate
ad-species, formed from CO2, can contribute to the spectra
with bands at 1648, 1453, and 1230 cm−1; however, absence
of intensity at 3605 cm−1 eliminates such possibility. In gen-
eral terms, introduction of propene to the reaction mixture
induced concomitant decreases or complete disappearance
of nitrate groups, particularly those of type II. Changes were
noted in the hydroxyl stretching region upon introduction of
propene, which mainly involved a depletion of the more ba-
sic hydroxyl species (at 3769–3747 cm−1) and generation
of acidic hydroxyls (at 3567–3542 cm−1). In general terms,
a net production of hydroxyl species was apparent follow-
ing exposure of the NOx covered surface to propene, except
for the 6Ag sample which showed significantly greater con-
sumption of basic hydroxyls during the process.
3.3. In situ DRIFTS experiments
In order to identify the different species formed by
chemisorption of the different reactants, which might allow
a degree of rationalization in terms of the catalytic behavior
of these systems, a series of infrared experiments were per-
formed by subjecting the samples to specific reactant mix-
tures starting with simple mixtures and gradually introduc-
ing the other reactant components. Following calcination in
dry air at 773 K, the samples were cooled under the dry air
flow to temperatures ranging from 573 to 723 K (see below)
before switching to inert gas (N2) for 15 min and record-
ing a background spectrum under these conditions. The de-
sired reactant gases (diluted in N2) were then introduced at a
fixed temperature (leaving 15–45 min contact with the cor-
responding reactant gases before recording the spectrum in
each case, ensuring the stability of band intensity) and the
changes in the infrared spectra recorded (100 scans being ac-
cumulated for these experiments). Different recording tem-
peratures, calculated to attain conversions of NOx (Fig. 4)
and C3H6 (Fig. 5) of about 20% with the full mixture, are
included in these figures. These are intended to provide a
representative (though approximate) picture of the surface
species which might be involved in the reaction mechanism.
Introduction of NO alone induced only minor changes
in the IR spectra (Figs. 4Aa–Da), whereas more drastic
changes occurred when a NO + O2 mixture was allowed
to contact the sample (Figs. 4Ab–Db). The latter treatment
led to the appearance of two groups of bands in the spectra
of the alumina and supported Ag samples (Figs. 4Ab–Db)
at ca. 1550 and 1245 cm−1 (type I) and 1580–1560 and
For the 1.5Ag and 4.5Ag systems, experiments performed
by introducing the samples initially to a C3H6 + O2 mixture
(Figs. 5A and 5B) confirm the formation of acrylate species
(bands at 1642–1643, 1572–1575, 1456–1461, 1383–1372,
and ≈ 1300 cm−1) and in addition, the presence of the
other carboxylate-type species (bands at ca. 1540 and 1460–
1455 cm−1). The concentration of surface acrylate increased
following the subsequent addition of NO, which in turn led
to the generation of nitrate species for the 4.5Ag sample.

A. Iglesias-Juez et al. / Journal of Catalysis 217 (2003) 310–323
315
(A)
(B)
Fig. 4. DRIFTS spectra at temperatures corresponding to 20% NO conversion for (A) alumina, (B) 1.5Ag, (C) 4.5Ag, and (D) 6Ag after consecutive treatment
with 1800 ppm of NO (a), and subsequent addition of 5% O (b), 900 ppm of propene (c), and up to 1800 ppm of propene (d). Complete gas mixture after
2
1 h (e).
Hydroxyl groups evolved in a similar way as described
in the former paragraph; thus, contact with the C3H6–O2
mixture generated acidic hydroxyls (3553 cm−1) while the
the experiment may also contribute to reduce intensity
of “isolated” hydroxyls and enhance that of associated
groups.
DRIFT spectra of the samples recorded under reac-
tion conditions at increasing temperatures are shown in
Figs. 6A–D. Analysis of these results essentially confirms
the points noted for experiments performed using sequential
admission of adsorbates (Figs. 4 and 5). Thus, for the alu-
more basic ones (3770–3760, 3725, and 3680–3660 cm−1
)
were consumed. As in the experiments of Fig. 4, analysis
of the overall intensity suggests that a net formation of
hydroxyl species was already produced during propene
oxidation with oxygen. Of course, formation of water during

316
A. Iglesias-Juez et al. / Journal of Catalysis 217 (2003) 310–323
(C)
(D)
Fig. 4. Continued.
mina alone, (Figs. 6A), propene oxidation appears to lead to
the formation of adsorbed formate species (bands at 3000,
2907, 1590, 1391, and 1376 cm−1), and, likely, of a higher
carbon carboxylate-type compound (evidenced by the pres-
ence of the band at 1460 cm−1). In contrast, acrylate species
were somewhat detected for the active silver-containing
(1.5Ag and 4.5Ag) samples, as characterized by the detec-
(according to the intensity of the band at 1460–1455 cm−1
)
decreased with increasing silver loading, which is also in
accordance with the evolution of the CO2 gas contribution
(bands at 2400–2200 cm−1), which, in turn, roughly fol-
lows the propene conversion profiles obtained during the
catalytic tests (Fig. 1A). Other effects induced by the pres-
ence of silver concern the formation of mono- and/or biden-
tate nitrates (1585–1590 and 1305 cm−1, and 1545–1550,
and 1250 cm−1) and bridging nitrates or adsorbed NO2
(1611–1616 cm−1) [18,22,25,28,29] immediately following
tion of the weak bands at ≈ 1645, 1580, and 1390 cm−1
.
In general terms, the temperature at which maximum inten-
sity of bands due to carboxylate-type species was observed

A. Iglesias-Juez et al. / Journal of Catalysis 217 (2003) 310–323
317
(A)
(B)
Fig. 5. DRIFTS spectra at temperatures corresponding to 20% propene conversion for (A) 1.5Ag, and (B) 4.5Ag, after consecutive treatment with 900 ppm of
propene +5% O (a), and subsequent addition up to 1800 ppm of propene (b), 900 ppm of NO (c), and up to 1800 ppm of NO (d).
2
exposure to the reactive mixture. These species rapidly de-
pleted on raising the temperature, particularly those giving
bands at ca. 1590 and 1305 cm−1 (type II nitrate). In gen-
eral, a small amount of the type I nitrate (bands at ca. 1550
and 1250 cm−1) was retained even at the highest sample
temperature. A significant difference between sample 6Ag
and the other two silver-containing samples (or the alumina
alone) was the identification of cyanide species (band at
2170 cm−1—Fig. 6) for the former, appearing initially at
573 K. This appears to be a particular characteristic of this
sample as indicated by the detection of a small amount of
such cyanide species upon introduction of C3H6 in the ex-
periments of Fig. 4D. The absence of type II nitrates was
also noticeable. These points taken together suggest that a
distinct pathway for NO dissociation/activation exists over
sample 6Ag. Other features appearing in the spectra under
reaction conditions (Fig. 6) include those giving a broad
band between 2311 and 2237 cm−1. These bands are prob-

318
A. Iglesias-Juez et al. / Journal of Catalysis 217 (2003) 310–323
(A)
(B)
Fig. 6. DRIFTS spectra under reaction conditions for (A) alumina, (B) 1.5Ag, (C) 4.5Ag, and (D) 6Ag samples. Samples under dry air at 300 K (a) and under
reaction conditions at 473 (b), 573 (c), 673 (d), 723 (e), 758 (f), 773 (g), and 798 K (h).
ably related to the nitrosonium (NO+) species, which are
possible (positively charged) intermediates in formation of
nitrate species upon interaction of NO or NO2 with sur-
face hydroxyl or are concomitantly produced with nitrate-
precursors molecules by disproportionation reaction of N-
dimers [18,22,25,26]. On the other hand, the evolution of
hydroxyl groups over the course of these experiments shows
a strong dependence on the reaction temperature. Thus,
whereas only limited consumption of basic hydroxyls and
an accompanying increase of the more acidic groups is pro-

A. Iglesias-Juez et al. / Journal of Catalysis 217 (2003) 310–323
319
(C)
(D)
Fig. 6. Continued.
duced at low reaction temperatures (corresponding to rela-
tively low conversion levels and in agreement with results
displayed in Figs. 4 and 5), a more pronounced decrease of
the latter is produced at higher reaction temperatures, and
this is acute in the case of the 6Ag sample. As noted, pro-
duction of water can concur in such phenomenon but seems
not enough (band at ca. 1625 cm−1) to justify the described
changes in the hydroxyl region (Figs. 4–6).
3.4. In situ XAFS experiments
XANES and EXAFS spectra (Fig. 7 and 8, respectively)
are shown for the most active catalysts (1.5Ag and 4.5Ag)

320
A. Iglesias-Juez et al. / Journal of Catalysis 217 (2003) 310–323
Fig. 7. XANES spectra of calcined Ag systems and reference materials: (a)
Ag O, (b) α-AgAlO , (c) 1.5Ag, (d) 4.5Ag, and (e) β-AgAlO .
−1
Fig. 8. FT of EXAFS spectra (1.95–8.2 Å ) for 1.5Ag and 4.5Ag samples
and reference compounds: (a) 1.5Ag, (b) 4.5Ag, (c) β-AgAlO , and
2
2
2
2
(d) Ag O. Black line, calcined materials; gray line, postreaction materials;
2
full line, modulus; dashed line, imaginary part.
along with several reference compounds. The references
(Fig. 7) may be classified in two groups, those which have a
linear local silver geometry, as is the case for Ag2O and α-
AgAlO2, and those with tetrahedral (or subgroup) symmetry
β-AgAlO2 [30]. The two active systems have XANES spec-
tra which do not change under reaction conditions and dis-
play a high degree of similarity with the β-AgAlO2 phase,
giving strong support to an assignment of the active silver
species to small silver aluminate particles but possessing
tetrahedral symmetry. Small differences in intensity and in
the position of continuum resonances (CRs) are a conse-
quence of the small particle size of the aggregates and/or
the interactions with the alumina support, as has been re-
ported for many other supported-metal systems [31]. In fact,
the shift to lower energy in the CRs at ca. 25525 eV, due to
the 5sp electronic density, provides evidence that the Ag lo-
cal coordination distances are modified with respect to the
reference compound [31]. Further proof that the major part
of silver is stable under reaction conditions is given in Fig. 8,
where the Fourier Transform (FT) of EXAFS spectra of both
systems is presented for samples measured in situ after calci-
nation and after reaction at 773 K. From this figure it is clear
that only a minimal part (less than 10%) of silver atoms un-
dergo sintering. The comparison with the FTs of reference
compounds displayed (Fig. 8) clearly indicates the presence
of a contribution to the first Ag–O shell with shorter dis-
tance than those present in the β-AgAlO2 phase. The Ag-
cation distances which must contribute to the second peak of
the FTs appear of a similar nature to those present in the β-
AgAlO2 reference and are located at shorter distances than
the Ag–Ag contribution of the Ag2O reference.
Analysis of the FTs presented in Fig. 8 gives evidence
for three different Ag–O contributions, located at 2.15, 2.4,
and 2.75 Å with coordination numbers of 0.30, 0.10, and
0.9 and Debye–Waller factors of 0.03, 0.015, and 0.065 Å2,
respectively, for 1.5Ag and 4.5Ag calcined materials (within
a maximum deviation of 0.04 Å for the first and 15%
for the second and third parameters). The high disorder
(probably non-Gaussian) displayed by the solids strongly
limits the accuracy of the EXAFS fitting results. Moreover,
they are additionally limited by the fact that both O anions
and Ag/Al cations are expected as second nearest neighbors
at distances close to 2.75 Å [30]. A correct fit may need
to include Ag–Ag and Ag–Al contributions at the second
coordination shell, a fact precluded by the absence of a
confident Ag–Al phase and amplitude functions as well as
for the limited number of fitting free parameters. In any
case, the analysis performed is certainly indicative of the
strong structural disorder at a local level displayed by the
Ag samples.
Additional XANES experiments were performed to fol-
low the reducibility of the Ag species present for the two

A. Iglesias-Juez et al. / Journal of Catalysis 217 (2003) 310–323
321
average Ag–O first shell coordination distance with respect
to those present in the more regular, tetrahedral-like local
structure of the β-AgAlO2 material. Obviously, this local
structure has notable differences with the linear structure
typical of the Ag(I) oxide or alpha aluminate, which may
be present in less active/selective systems. It is interesting
to note that the linear, β-Al2O3-type AgAl11O17 aluminate
has previously been detected in calcined alumina-supported
silver samples [17]. This phase usually contains excess Ag
cations between the characteristic alumina layers of the beta
phase (which may be described as a Ag(Ag2O)Al11O17
phase), and its catalytic performance may be poor as
segregation of the Ag2O entities with subsequent formation
of Ag(0) particles could occur readily upon contact with
a reducing agent (hydrocarbon) at reaction temperatures
[16,17].
The stability of the species as indicated by the simi-
larity of the XANES shape of the two active samples un-
der reaction conditions, in agreement with UV–vis results
(Fig. 2), indicates the absence of silver phase transforma-
tion processes, indicating a low probability for stabilization
of other oxidation states (Ag(0) and AgO) or, as noted, of
linear Ag+ species during the course of the reaction. Ad-
ditionally, the EXAFS results (Fig. 9) demonstrate that no
significant particle size growth occurs under reaction up to
823 K. It is important to stress the absence of significant
amounts of other silver redox states/phases, as this would
hinder interpretation of DRIFTS results concerning the SCR
mechanism and intermediates. Several TEM/EDS studies of
Ag/Al2O3 systems in the 1–2 wt% range [5,17,18] have de-
tected, particularly after reaction, the presence of large clus-
ters of amorphous or poorly crystallized silver phases, which
are very scarce in our micrographs and, as noted previously,
do not influence the EXAFS spectra. Although the contribu-
tion of these aggregates to the exposed silver surface under
reaction conditions may be argued, it is not clear if some of
the intermediates and/or spectator species detected in previ-
ous studies may be associated with these phases.
Fig. 9. Concentration profiles of silver species during a thermo-programmed
run under 0.1% propene. Triangles, 4.5Ag; squares, 1.5Ag. Open symbols,
initial, oxidized species; closed symbols, final, reduced species. See text for
details.
most active catalysts. Analysis of data obtained during a
temperature ramp in the presence of 0.1% propene using fac-
tor analysis [31] shows the presence of two silver-containing
species, the initial beta silver aluminate and metallic sil-
ver at high temperature, with no clear differences between
the 1.5Ag and 4.5Ag samples in terms of the shape of the
corresponding XANES spectra. The important differences
between the two samples concerns the onset temperature
for reduction, indicated by the appearance of reduced sil-
ver species in Fig. 9, which is indicative of the enthalpy of
formation of the oxygenated intermediates. Although this
is unspecific for the different intermediates which may be
formed, it indicates a more facile extraction of oxygen for
the 4.5Ag sample, despite the similarity in the chemical na-
ture of the Ag+ species present initially in the calcined ma-
terials.
4. Discussion
In contrast to the behavior of the “low-loaded” samples,
UV–vis results for the 6Ag sample (Fig. 2) indicate that
changes in the nature of the silver species are produced in
this sample during the course of the reaction. These can be
related to the greater ease by which metallic silver states are
generated in this sample, which would facilitate silver sin-
tering processes above a certain temperature. In accordance
with this hypothesis, formation of surface metallic silver
states for this sample under reaction conditions is indicated
by the ease by which NO dissociation takes place as inferred
by the detection of adsorbed cyanide species at relatively low
temperatures in the DRIFTS experiments (Fig. 6D). This is
also in agreement with the formation of significant amounts
of N2O which is specific to this sample (Fig. 1), as NO
dissociation/coupling reactive paths leading to N2O forma-
tion can be related to the presence of metallic particles at the
sample surface [32].
The active silver species involved in the SCR of NOx
has been the subject of much debate although it is generally
believed that highly dispersed Ag+ cations are involved
[5,15–20]. However, the local structure of such species
which define the active phase has not been ascertained; it
is only known that a strong interaction with the support is
required to stabilize such species under reaction conditions
and preclude the formation of zero-valent silver which favors
the nonselective combustion of the hydrocarbon [5,15–20].
Comparison of the XANES shapes for the different samples
(Fig. 8) and the Ag+-containing reference compounds which
may exist in the Ag/Al2O3 system clearly indicates that
such cations possess (nearly) tetrahedral local coordination
which closely resembles that of the β-AgAlO2 reference.
The EXAFS results suggest a significant local disorder
around Ag centers as well as a noticeable reduction of the

322
A. Iglesias-Juez et al. / Journal of Catalysis 217 (2003) 310–323
The mechanism of the NO-SCR with hydrocarbons
Hydrocarbon activation appears to be a very important
step for silver systems [17,34] and, in fact, the reaction of
an intermediate oxygenate hydrocarbon with the alumina
surface is proposed as the rate-limiting step for supported
tin catalysts [35]. Here, the hydrocarbon is activated in the
presence of oxygen (Fig. 5, 6) initially at the methyl rather
than the vinyl end of the molecule leading to a partial oxida-
tion product with retention of the C=C bond and giving ad-
sorbed acrylate. Oxidation via activation of the vinyl part of
the molecule may give rise to the other carboxylate species,
such as propanoate or ethanoate species. An alternative to
the latter, concurrent scheme would be a consecutive route,
where acrylate interacts with the appropriate acid–base cen-
ters of the alumina surface (consistent with modifications
observed in the hydroxyl region of the infrared spectra in
Figs. 3–5) and evolves in the presence of water with forma-
tion of acetaldehyde (plus formaldehyde) which, in turn, is
subsequently attacked by NO-derived species after the rate-
limiting step [35]. The proximity of surface Al sites (which
act as anchoring sites for acrylate species) and Ag centers
may be another reason for the high activity of the aluminate
phase. As previously noted, consumption of surface hydrox-
yls is particularly significant for 6Ag, a fact which must in-
fluence N-activation (absence of type II nitrates) and should
be considered, together with the presence of Ag(0) under re-
action conditions, as a possible cause for the low selective
reduction of propene with NOx observed for this sample. On
the other hand, formate, acetate, or other carboxylate-related
compounds are fairly unreactive with either NO or O2 alone
but can be removed in NO+O2 gas mixtures [16]. However,
they do not seem to favor or have a negligible contribution
to the coupling of N-species with the subsequent formation
of N2. DRIFTS results show that silver phases are involved
in opening up alternative reaction pathways for hydrocarbon
activation which lead to the generation of acrylate species,
consistent with previous results from our laboratories [18].
Differences between the activities of the alumina alone and
the supported silver samples are most likely related to some
extent to the way in which the hydrocarbon is activated in
each case which, as postulated in studies of partial oxidation
of propene, can be related to the nature of the allyl interme-
diate formed in each case [26]. The silver-related phase may
be involved in supplying the active oxygen species for the
initial partial oxidation step and as shown in Fig. 9 the silver
oxide anion species are in fact more labile for 4.5Ag than
the 1.5Ag, which could then explain the higher NOx reduc-
tion activity of the former (Fig. 1). This may favorably alter
the rate of formation of the hydrocarbon intermediate. It can
be noted, however, that the initial Ag–O distance and coor-
dination numbers of our two active samples are very similar
in the calcined state (as suggested by the EXAFS FTs dis-
played in Fig. 8) and only a very small variation after reac-
tion is detected for the more active 4.5Ag sample. This small
variation in size (and/or shape) could be the origin of the
difference between the oxygen labilities detected for these
two samples. Other parameters such as the number of active
has been extensively studied for silver [5,18,22,25,33,34],
cobalt, and tin [9,35] on alumina samples, and a degree of
similarity in terms of behavior and mechanism is apparent.
It is generally accepted that both the hydrocarbon (propene)
and NO molecules must first be activated by formation of
some oxidized, adsorbed intermediates, typically carboxy-
lates such as acrylate for the hydrocarbon and NO2 or nitrite
and nitrate compounds for NO [5,18,22,25,33–35]. Some of
these activated, intermediate compounds may be present for
the alumina support alone (Figs. 4 and 5), indicating that
the support itself is, as well known, able to promote the SCR
of NOx using hydrocarbons as reducing agents. Silver is
thought to play an important role in modifying the surface
concentration and selectivity among these intermediates, en-
hancing the activity of the carrier. However, the coupling be-
tween these intermediates may occur primarily on the alu-
mina surface and the role of silver in such coupling steps
may be secondary [18,27,35]. It appears, therefore, that the
main role of the metal may be restricted to the initial steps
of the reaction and is mainly related to activation of the reac-
tant molecules, either directly at the silver surface and/or by
the modification of the acid/base (hydroxyls) properties of
alumina (Fig. 3). DRIFTS results (Figs. 3–6) strongly sup-
port a clear role of active silver entities in modifying both
the NO and the propene activation paths.
For NO, there is some evidence (Figs. 4 and 5), that
NO may be activated by interaction with acidic hydroxyls,
thus generating nitrosonium ion (NO+) in a process involv-
ing H+ (originating from the hydroxyls) consumption, as
has been previously postulated for alumina [18] or zeolite-
based systems [36,37]. Some of these NO-derived adsor-
bates are detected at high temperatures under reaction condi-
tions (Figs. 6) and display maximum intensity for the most
active (4.5Ag) sample. As observed in Fig. 4, such acidic
hydroxyls are recovered by interaction with the hydrocar-
bon (or derived products), thus closing the catalytic cycle
and regenerating the active surface species. Alternatively,
nitrosonium ions plus nitrate precursor molecules (nega-
tively charged N-containing molecules) can be simultane-
ously formed by disproportionation reaction of N-dimers
(N2O3, N2O4) [25,37]. An additional pathway for nitrate
formation is, as noted, reaction of NO2(g) and hydroxyls
groups. In any case, in our samples, two types of nitrates
giving bands at ca. 1545–1550 and 1245 cm−1 (type I) and
1580–1590 and 1305 cm−1 (type II) are detected. Both mea-
surements under reaction conditions during a light-off ex-
periment (Fig. 6) and at low, steady-state conversion (Fig. 4)
show the greater activity of type II nitrates. Type II nitrates
are usually ascribed to bidentate species [22,25,26], which
must therefore be considered as the NO-derived active en-
tities which preferentially react with C-containing interme-
diates. Differences between the two active samples in this
respect mainly concern the greater number of type II nitrates
for 4.5Ag, but with no clear difference concerning the ratio
of the N-containing ad-species was detected.

A. Iglesias-Juez et al. / Journal of Catalysis 217 (2003) 310–323
323
sites for propene activation and or acrylate vs other types
References
of carboxylate-related compound selectivity would appear
to be of lesser importance. This comes from the fact that al-
though the 4.5Ag system has a significantly greater number
of silver sites for the activation of propene (IR intensity—
Fig. 4), in the presence of NO this can be roughly balanced
by the greater selectivity to acrylate species displayed by the
1.5Ag sample.
[1] M. Iwamoto, S. Yokoo, K. Saaki, S. Kagawa, J. Chem. Soc. Faraday
Trans. 1 77 (1981) 1629.
[2] V.I. Parvulescu, P. Grange, B. Delmon, Catal. Today 46 (1998) 233.
[3] H.-W. Jen, Catal. Today 42 (1998) 37.
[4] T. Miyadera, K. Yoshida, Chem. Lett. (1993) 1483.
[5] T. Furisawa, K. Seshan, J.A. Lercher, L. Leffterts, K.-I. Aika, Appl.
Catal. B 37 (2002) 205.
[6] R. Burch, P.J. Millington, A.P. Walker, Appl. Catal. B 4 (1994) 65.
[7] R. Burch, T.C. Watling, Appl. Catal. B 11 (1997) 207.
[8] T. Miyadera, Appl. Catal. B 2 (1993) 199.
5. Conclusions
[9] M.C. Kung, H.H. Kung, Topics Catal. 10 (2000) 21.
[10] S. Sumiya, M. Saito, H. He, Q.-C. Feng, N. Takezawa, K. Yoshida,
Catal. Lett. 50 (1998) 87.
A series of Ag/Al2O3 catalysts with silver loadings in the
1.5–6 wt% range was prepared by an impregnation method
employing EDTA complexes to promote a high dispersion
of the silver phase. This was successful in obtaining highly
active systems for silver loadings of 4.5 wt%. Structural
analysis of the active silver phase gives evidence of a
tetrahedral-like local symmetry for the active silver phase,
with significant disorder and a Ag–O distance contribution
to the first shell which is significantly shorter than the
corresponding values for the symmetry-related (at a local
level) β-AgAlO2 phase. The high degree of stabilization of
Ag(I) cations inserted into this structure contrasts with the
more labile nature typically displayed by linear Ag species
present either in oxide or aluminate phases [15–20].
[11] T.N. Angelidis, N. Kruse, Appl. Catal. B 34 (2001) 201.
[12] F.C. Meunier, R. Ukropec, C. Stapleton, J.R.H. Ross, Appl. Catal. B 30
(2001) 163.
[13] A. Abe, N. Aoyama, S. Sumiya, N. Kakuta, K. Yoshida, Catal. Lett. 51
(1998) 5.
[14] T. Miyadera, Appl. Catal. B 16 (1998) 155.
[15] K.A. Bethke, H.H. Kung, J. Catal. 172 (1997) 93.
[16] K.-I. Shimizu, J. Shibata, H. Yoshida, A. Satsuma, T. Hattori, Appl.
Catal. B 30 (2001) 151.
[17] N. Bogdanchikova, F.C. Meunier, M. Avalos-Borja, J.P. Breen,
A. Prstryakov, Appl. Catal. B 36 (2002) 287.
[18] A. Martínez-Arias, M. Fernández-García, A. Iglesias-Juez, J.A. An-
derson, J.C. Conesa, J. Soria, Appl. Catal. B 28 (2000) 29.
[19] T. Nakatsuji, R. Yasukawa, K. Tabata, K. Ueda, M. Niwa, Appl. Catal.
B 17 (1998) 333.
[20] Z. Li, M. Flytzani-Stephanopoulos, Appl. Catal. A 165 (1997) 15.
[21] E.A. Stern, Phys. Rev. B 48 (1993) 9825.
The silver active species influences the reaction mech-
anism mainly by acting, in conjunction with alumina hy-
droxyls, in the activation steps of both NO and propene. For
NO, positively charged N-containing ions and surface type II
nitrates may evolve from N-containing dimers; the type II
nitrate species being a potential intermediate which reacts
with C-containing intermediates. Additionally, active silver
species are involved in the opening of a route leading to ad-
sorbed acrylate species as the active partially oxidized inter-
mediate responsible for hydrocarbon activation. A possible
relationship between the oxygen lability of the beta silver
aluminate-like active phase and acrylate formation is postu-
lated as being of importance in the catalytic performance of
active silver systems.
[22] F.C. Meunier, J.P. Breen, V. Zuzaniuk, M. Olsson, J.R.H. Ross, J. Ca-
tal. 187 (1999) 93.
[23] T. Gerlach, U. Illgen, M. Bartoszek, M. Baerns, Appl. Catal. B 22
(1999) 269.
[24] (a) H. Knözinger, P. Ratnasamy, Catal. Rev. Sci. Eng. 17 (1978) 31;
(b) J.A. Anderson, C.H. Rochester, in: A. Hubbard (Ed.), Encyclope-
dia of Surface and Colloid Science, Dekker, New York, 2002, p. 2528.
[25] K. Hadjiivanov, Catal. Rev. Sci. Eng. 42 (2000) 71.
[26] A.A. Davydov, Infrared Spectroscopy of Adsorbed Species on the
Surface of Transition Metal Oxides, Wiley, New York, 1984.
[27] V. Zuzaniuk, F.C. Meunier, J.R.H. Ross, J. Catal. 202 (2001) 340.
[28] K.-I. Shimizu, H. Kawabata, A. Satsuma, T. Hattori, J. Phys. Chem.
B 103 (1999) 5240.
[29] G.M. Underwood, T.M. Miller, V.H. Grassian, J. Phys. Chem. A 103
(1999) 6184.
[30] W. Gessner, Z. Anorg. Allg. Chem. 352 (1967) 145.
[31] (a) M. Fernández-García, C. Márquez, G.L. Haller, J. Phys. Chem. 99
(1995) 12565;
(b) M. Fernández-García, Catal. Rev. Sci. Eng. 44 (2002) 59.
[32] M. Hirsimaki, M. Valden, J. Chem. Phys. 114 (2001) 2345.
[33] S. Sumiya, H. He, A. Abe, N. Takezawa, K. Yoshida, J. Chem. Soc.,
Faraday Trans. 94 (1998) 2217.
[34] K.-I. Shimizu, J. Shibata, A. Satsuma, T. Hattori, Phys. Chem. Chem.
Phys. 3 (2001) 880.
[35] J.-H. Lee, A. Yezerets, M.C. Kung, H.H. Kung, Chem. Commun.
(2001) 1404.
[36] T. Gerlach, F.-W. Schütze, M. Baerns, J. Catal. 185 (1999) 131.
[37] K. Hadjiivanov, J. Saussey, J.L. Freysz, J.C. Lavalley, Catal. Lett. 52
(1998) 103.
Acknowledgments
A.I.-J. and A.B.H. thank the Comunidad Autónoma de
Madrid for a predoctoral grant. Thanks are also due to
the technical staff of BM29 line at ESRF Synchrotron for
their hep during XAFS measurements. Dr. F. Del Monte
for its kind support during the recording of UV–vis spectra.
Support from CAYCIT (Project No. MAT 2000-1467) is
greatly appreciated.