Study by in situ FTIR spectroscopy of the SCR of NOx by ethanol on Ag/Al2O3 - Evidence of the role of isocyanate species

Article abstract of DOI:10.1016/S0021-9517(03)00035-6

The selective catalytic reduction (SCR) of NO_x in excess of oxygen using ethanol as reducing agent on silver/alumina was studied by coupling in situ FTIR spectroscopy and mass spectrometry. A comparison with the same reaction on bare alumina permitted us to show the connection between the presence of silver and isocyanate species on a catalyst. Then, a detailed investigation concerning these groups was undertaken to understand their formation, their localization, and their reactivity in order to propose the pathway of the NO_x SCR into N₂. Three elemental sequences are suggested explaining, first, the formation of silver cyanide and its transformation into Al³⁺NCO, then isocyanates hydrolysis into ammonia, and finally the reaction of the latter species with NO in the presence of oxygen giving rise to nitrogen.

Full text of DOI:10.1016/S0021-9517(03)00035-6

Journal of Catalysis 217 (2003) 47–58
www.elsevier.com/locate/jcat
Study by in situ FTIR spectroscopy of the SCR of NO by ethanol
x
on Ag/Al O —Evidence of the role of isocyanate species
2
3
a
a
b
a,∗
Nicolas Bion, Jacques Saussey, Masaaki Haneda, and Marco Daturi
a
Laboratoire de Catalyse et Spectrochimie, UMR 6506, ISMRA, ENSICaen, 6, bd du Maréchal Juin, 14050, Caen cedex, France
b
National Institute of Advanced Industrial Science and Technology (AIST), AIST Tsukuba Central 5, 1-1-1 Higashi, Tsukuba, Ibaraki 305-8565, Japan
Received 7 August 2002; revised 9 January 2003; accepted 16 January 2003
Abstract
The selective catalytic reduction (SCR) of NOx in excess of oxygen using ethanol as reducing agent on silver/alumina was studied by
coupling in situ FTIR spectroscopy and mass spectrometry. A comparison with the same reaction on bare alumina permitted us to show
the connection between the presence of silver and isocyanate species on a catalyst. Then, a detailed investigation concerning these groups
was undertaken to understand their formation, their localization, and their reactivity in order to propose the pathway of the NOx SCR
3
+
into N . Three elemental sequences are suggested explaining, first, the formation of silver cyanide and its transformation into Al NCO,
2
then isocyanates hydrolysis into ammonia, and finally the reaction of the latter species with NO in the presence of oxygen giving rise to
nitrogen.

2003 Elsevier Science (USA). All rights reserved.
Keywords: Selective reduction of NO; Ag–Al O ; In situ diffuse reflectance FT-IR; Time-resolved analysis; Transient conditions; Reaction mechanism;
2
3
Reaction intermediates; Isocyanates
1
. Introduction
Since the elimination of soot particulates emitted from
reaction commonly used is the SCR of NOx by propene. Re-
sults obtained from this reaction studied by in situ FTIR have
been verified in those species [15]. But the weak concentra-
tion of isocyanates obtained and detected by IR in the reac-
tion has caused some controversy.
diesel has been resolved using a filter associated to a cerium-
based additive [1], diesel engines are considered to be less
polluting than gasoline engines thanks to their lower fuel
consumption, which involves lower CO2 rejects. However,
a new generation of cars using gasoline, with a large excess
of oxygen (called lean-burn motorization) is commercialized
because of its weak CO2 pollution as well. Nevertheless,
the drawback of these two kinds of engines is the large
excess of oxygen making the NOx removal difficult. This
problem is still the subject of much research on the selective
catalytic reduction (SCR) reaction, particularly on alumina-
based oxide catalysts, which seem to be the best candidates
for applications under real conditions [2–14].
The goal of the present report, using ideal conditions,
is to definitively demonstrate that isocyanates are really
intermediate species and then to understand how these
species react. Thus we decided to undertake the analysis of
the C H OH + NO + O reaction on Ag/Al O and Al O
2
5
2
2
3
2
3
in a steady state and pulse regime, the choice of ethanol as
reducing agent being dictated by the necessity to increase the
NCO concentration. A specific study of isocyanate species
has allowed us to determine a possible mechanism for the
DeNOx reaction.
Among the mechanisms proposed by the different au-
thors, isocyanate species have often been proposed as reac-
tion intermediates. However, in many cases, the demonstra-
tion of their role is often made without real proof. The model
2
. Experimental
Al2O3 and Ag/Al2O3 catalysts, prepared by RHODIA,
*
present respectively a specific surface area of 112 and
103 m /g. The silver catalyst contains 2wt% of metal that
Corresponding author.
E-mail address: daturi@ismra.fr (M. Daturi).
2
0021-9517/03/$ – see front matter  2003 Elsevier Science (USA). All rights reserved.
doi:10.1016/S0021-9517(03)00035-6

48
N. Bion et al. / Journal of Catalysis 217 (2003) 47–58
is usually proposed as the optimum percentage for the SCR
of NO using ethanol as reducing agent [16].
Our in situ system allows the analysis of the reactives
and of the reaction products by mass spectrometry and
FTIR spectroscopy. The catalyst is placed in a DRIFTS
reactor cell commercialized by Spectratech, which permits
the observation of adsorbed species present on its surface
under reaction conditions. IR spectra are collected with
a Nicolet 550 FT-IR spectrometer, accumulating 64 scans
−1
at a resolution of 4 cm . Gas products are analyzed by
quadrupole mass spectrometer QMS, Balzers TCP 121, and
3
by FTIR via a gas microcell of 0.088 cm volume.
Twenty milligrams of catalyst is pretreated for 2 h at
8
73 K under a flow (25 ml/min) of 10% O2 in Ar. The
sample is then taken to the reaction temperature. For steady-
state analysis, the gas mixture is composed of 1000 ppm
C2H5OH, 500 ppm NO, and 10% O2 with Ar balance
Fig. 1. Conversion of NOx versus reaction temperature. Reaction flow: NO
−3
(
500 ppm), C H OH (1000 ppm), O (10%); W/F = 0.048 g s cm
.
2
5
2
(
total flow 25 ml/min). For the pulse regime, the catalyst
is exposed to NO + O2 mixtures in Ar flow and then pulses
of C2H5OH are introduced into the stationary stream. After
optimization of the experiments, the following parameters
were chosen. Ethanol containing saturator placed at ambient
silver/silica catalysts has not shown any activity. Thus, we
think that support as well as silver content plays an important
role in the SCR of NOx by ethanol in excess of oxygen.
3
temperature, purged by 10 cm /min argon flow (saturation
vapor pressure of ethanol at 298 K: about 56 Torr). Use of a
3
0
.02 cm injection loop. One-tenth of the carrier flow purges
3.1.2. Gas-phase analysis
the injection loop circuit.
Thanks to the coupling of a IR gas microcell with our
analysis system, gas spectra presented in Fig. 2 are obtained
for each temperature corresponding to the adsorbed phase
spectra. In a first experience, the C2H5OH + NO + O2 reac-
tion was realized by replacing the catalyst with carburondum
to determine the specific products of the thermal reaction
To achieve the aim of this study, we have undertaken a
panel of different experiments. Thus, gas compositions are
given in detail along with the results. However, all these
experiments were realized at atmospheric pressure and the
−3
W/F ratio has been kept at 0.048 g s cm
.
−1
(
Fig. 2A). The mixture NO (1875 cm ) + O2 causes the
−1
formation of small amounts of NO2 (1617 cm ). Ethanol
addition (1394, 1241, 1066, and 879 cm ) produces (al-
ready starting from 623 K) acetaldehyde (1745 cm ), CO2
(667 cm ), and, above 673 K, CO (2143 cm ) and eth-
ylene (949 cm ). The characteristic band of acetaldehyde
is situated at the same wavenumber as formaldehyde, but
it is possible to differentiate the two species observing the
distance between the P and R rotation-vibration compo-
−1
3
. Results
−1
− −1
1
3
.1. Study of the C2H5OH + NO + O2 steady-state reaction
−1
Our in situ technique gives us the possibility of detecting,
during the reaction, the surface adsorbed species and the
gas products simultaneously. In the C2H5OH + NO + O2
reaction study spectra were collected on the catalyst surface
by the DRIFTS cell and in the gas phase by the FTIR gas cell
for each temperature after 20 min of steady-state reaction
−1
−1
nents maxima: 35 cm for CH3CHO and about 50 cm
for HCHO, which has a lighter mass. When the flow goes
through the Ag/Al2O3 catalyst, NO as well as ethanol con-
versions are much more important than during the thermal
reaction. Other differences appear, such as CO and C2H4
formation, at lower temperatures; we also detect new cat-
(
under isotherm conditions).
3
.1.1. DeNOx activity: comparison between Al2O3
−
1
and Ag/Al2O3 versus temperature
alytic conversion products at 623 K (HCN, 713 cm ) and
−1
Fig. 1 compares the DeNOx activity of alumina and
silver/alumina catalysts versus reaction temperature. The
gas flow consists of 1000 ppm of C2H5OH, 500 ppm of
NO, and 10% of O2. From 623 to 773 K, silver-containing
catalysts present higher activity than Al2O3: almost 30%
of NOx conversion under our conditions for Ag/Al2O3
compared to 10% for bare alumina. In the C2H5OH +
NO + O2 reaction, the support does not seem to be a
major element. However the same experiment realized on
between 623 and 723 K (NH3, 932 and 964 cm ). The last
compound is the only gas product not present on bare alu-
mina, because either it is not produced or its reaction is too
fast to allow detection.
However, the presence of this molecule near the temper-
ature of maximal conversion and on silver-containing cat-
alyst leads us to consider a mechanism of selective reduc-
tion of NOx detected on Cu–ZSM-5, in which isocyanate
groups were proposed as possible reaction intermediates. In

N. Bion et al. / Journal of Catalysis 217 (2003) 47–58
49
3
.1.3. Catalyst surface adsorbed species
In Fig. 3, spectra of adsorbed species at 623 K are re-
ported when ethanol is added into the initial flow com-
posed by Ar, NO, and O2. Spectra corresponding to silver-
containing catalysts (Fig. 3b) are remarkable by bands situ-
−
1
ated in the 2260–2125 cm region, as well as for features at
−
1
1
2
630, 1412, and 1340 cm . We attribute bands at 2255 and
228 cm to AlVINCO and to AlIVNCO species, respec-
−1
tively. However, controversial opinions exist about these as-
signments, and the authors have already clarified this point in
a previous report [19]. Therefore isocyanate species will be
assigned as in the above-noted reference. We also find in the
literature attributions concerning the other less intense and
−1
well-defined bands at 2155 and 2127 cm : they seem be
due to the presence of cyanide groups on alumina and/or sil-
−
1
ver (ν(C≡N) of AgCN complex absorbs at 2146 cm ) [20].
−1
Bands observed in the region below 2000 cm contain vi-
brations characteristic of ethyl carbonate species [21] on alu-
mina (1630, 1412, and 1340 cm ) and whose formation
is favored by the presence of the metal. Other features ex-
ist and are common for both catalysts. Four minutes after
C2H5OH introduction, the first spectrum (full line) presents
the bands of vibration of formate species (ν(COO ) at 1593
and 1375 cm ; ν(CH) at about 3000 cm —zone not re-
ported here; δ(CH) at 1390 cm ). Two intense bands at
578 and 1466 cm are also visible on all the spectra: they
−1
−
−1
−1
−1
−1
1
are due to acetate species. A study of surface species formed
Fig. 2. FTIR spectra of gas products after NO (500 ppm) + O (10%) flow
2
during the C2H5OH + O2 reaction showed that all theses
introduction on (A) SiC and (B) Ag/Al O at (a) 623 K and then C H OH
2
3
2
5
−1
bands detected in the 1700–1100 cm range appear, which
(1000 ppm) addition at (b) 623, (c) 673, (d) 723, (e) 773, (f) 823, and (g)
confirms that they are not nitrogen-containing species. Dot-
ted lines spectra were collected about 15 min later. A com-
parison between these spectra and the others previously in-
terpreted shows that species characterized by bands in the
1650–1200 cm area have the tendency to accumulate on
catalyst surfaces (bands intensities increased with time). So,
acetate and carbonate species produced by the decomposi-
tion or by the oxidation of ethanol do not seem to act as in-
termediate compounds in the SCR of NOx. On the contrary,
8
73 K.
that mechanism NCO species hydrolyze leading to N2 pro-
duction [17,18]. Under our conditions silver would play a
role similar as copper and would explain the good activity
of Ag/Al2O3; the presence of ammonia would confirm the
specific reaction for isocyanate species on these catalysts.
We will reconsider this possibility later.
−1
Fig. 3. FTIR spectra of surface adsorbed species on (a) Al O , and (b) Ag/Al O under C H OH (1000 ppm) + NO (500 ppm) + O (10%) flow at 623 K;
2
3
2
3
2
5
2
(full line) 4 min and (dotted line) 20 min after ethanol introduction.

50
N. Bion et al. / Journal of Catalysis 217 (2003) 47–58
Fig. 4. FTIR spectra of surface adsorbed species on Ag/Al O under C H OH (1000 ppm) + NO (500 ppm) + O (10%) flow at (a) 623, (b) 673, (c) 723,
2
3
2
5
2
(d) 773, (e) 823, and (f) 873 K.
isocyanate and cyanide species are in major part consumed:
indeed, the intensities of the bands supposed to characterize
them decrease.
Fig. 4 shows the evolution with temperature of Ag/Al2O3
catalyst surface adsorbed species. The reaction is studied be-
tween 623 and 873 K; spectra are collected every 50 K af-
ter about 20 min of steady state. The intensities of NCO
bands decrease quite strongly with a temperature increase up
to 773 K, a temperature at which these bands almost com-
pletely disappear. Features at lower wavenumbers (around
−1
1
500 cm ) follow the same evolution. The spectrum col-
lected at 823 K (Fig. 4e) clearly shows two bands at 1550
−1
and 1450 cm that characterize acetate species, which are
more stable with respect to other compounds. Similar phe-
nomena are observed on Al2O3 (spectra not shown). We
have observed that bare alumina catalysts do not present
NCO groups on their surface. Thus, increasing the tem-
perature, we observe only the decrease of acetate and car-
bonate corresponding bands intensities. The simultaneous
analysis of catalytic surface and catalytic activity curves in-
forms us about the eventual role of isocyanate species in the
SCR of NOx. These species do not accumulate at 823 K,
at temperature at which the NOx conversion is still rela-
tively important, but follow the activity evolution, disappear-
ing above 773 K, a temperature where the conversion begins
to heavily decrease. In this temperature domain, –CN and
Fig. 5. Correlation of FTIR analysis of gas and adsorbed species during the
introduction of 4 ethanol pulses at 698 K on Ag/Al O catalysts under NO
2
3
(
1000 ppm) + O (10%) flow.
2
Fig. 5 shows the trend of the integrated area of the
characteristic bands for each gaseous compound noted
above. These curves called chemigrams make the analysis
of the results obtained during these pulse experiments
easier and can be directly correlated to those presented in
Fig. 6, which correspond to the evolutions of the different
gases detected by mass spectrometry. Comparing the data
belonging to the two analysis systems is very important:
mass spectrometry alone would not have permitted us to
distinguish CO2 from N2O if the last had been present in the
gas phase, to detect ethylene, acetaldehyde, and hydrogen
cyanide; on the other hand IR spectroscopy is not sensitive
to centro-symmetric molecules, such as O2 and N2.
Alcohol detection is concomitant with the start of CO (in
a small amount), CO2, C2H4 (weak), CH3CHO (not shown),
and HCN (very weak and not represented) production and
with NO and NO2 consumption. This double consumption
shows either NOx reduction into N2 (because N2O and
NH3 are not detected) or NOx transformation into adsorbed
surface species, or both phenomena. If C2H5OH, C2H4,
–
NCO groups are mainly present on silver-containing cata-
lysts, which possess the most important DeNOx activity.
3
.2. Study of pulse reaction: NO reduction by ethanol
pulses in oxygen excess
3
.2.1. Gas-phase observations
We have analyzed the gas phase by FT-IR, collecting a
spectrum every 4 s during the introduction of four successive
ethanol pulses on Ag/Al2O3 at 698 K. The resulting
waterfall of spectra (not presented in the figures) shows
the same gas products observed during the study of the
stationary reaction: (from higher to lower wavenumbers)
CO2, CO, NO, CH3CHO, NO2, C2H5OH, C2H4, and HCN.
Only NH3 is absent or it is in too small amount to be
detected.

N. Bion et al. / Journal of Catalysis 217 (2003) 47–58
51
(a)
(b)
Fig. 7. FTIR spectra of adsorbed species during the introduction of 4 ethanol
pulses at 698 K on Ag/Al O catalysts under NO (1000 ppm) + O (10%)
2
3
2
−1
flow. Part (b) represents a magnification of part (a) in the 2500–1900 cm
spectral region. The third axis reports time in seconds.
collecting time as well as to CO2 detection, which can
be made in both phases) shows that formate and acetate
species are detected on the catalyst surface before
ethanol appears in the gas phase. This observation
shows that, first of all, ethanol is oxidized into these
compounds.
Fig. 6. Analysis by mass spectrometry of the gas phase during the
introduction of 4 ethanol pulses at 698 K on Ag/Al O catalysts under
2
3
NO (1000 ppm) + O (10%) flow.
2
–
–
Almost at the same time cyanides characteristic bands
increase in intensity, followed by isocyanates (Fig. 7b).
The integrations of the different bands versus time show
that the evolution of formate, cyanide, and isocyanate
species is similar to ethanol evolution (presence of
a maximum) whereas acetate species (integration of
and CH3CHO present the same evolution, their maxima
shortly precedes those of CO and CO2. These features are
reproducible for each pulse.
3
.2.2. Catalytic surface observation
The species detected in the adsorbed phase (Fig. 7) are
−1
1
460 cm band) accumulate.
in major part due to the presence of gas-phase species
as formates, acetates, carbonates, cyanide, and isocyanates.
However, a scrupulous observation of the different spectra
3.2.3. Correlation between surface and gas phases
The juxtaposition, on a same time scale, of gas and
adsorbed phases analysis (Fig. 5) is a very interesting
approach to the intermediary role of cyanide and isocyanate
species:
(
1 spectrum/4 seconds) gives the following added informa-
tion:
–
–
Before the appearance of the first ethanol pulse, the
catalyst is under NO + O2 flow and nitrate species are
−
present on its surface (ν3(NO3 ) at 1650–1500 and
– First we note the importance of NO consumption dur-
ing ethanol pulse sending. Each pulse induces the disap-
pearance of some NO amount, but in particular the NO
consumption persists continuously between 2 minima.
Indeed, for 300 s (corresponding to the efficiency of the
4 ethanol pulses), the average conversion is 20%.
−1
1
300–1170 cm [22]). They disappear when ethanol
arrives, while a weak peak appears in the NO2 gas
evolution.
The synchronism between gas and adsorbed species
evolution (possibly thanks to the knowledge of the

52
N. Bion et al. / Journal of Catalysis 217 (2003) 47–58
–
Second, we note that time resolution in the formation
3.3. Investigation about isocyanate species
and disappearance of the different gas and adsorbed
species (measured in FTIR as well as in mass spectrom-
etry) gives information on the NOx SCR pathway on
Ag/Al2O3 catalysts. We are able to establish the follow-
ing time scale according to the observations of maxima
and minima for the evolution of each species:
The species characterized by the bands localized at 2258
−1
and 2235 cm
seem to play an important role in the
C2H5OH + NO + O2 reaction on Ag/Al2O3 catalysts. Thus
accurate information about them is necessary. To eliminate
the ambiguity caused by the large amount of literature and
to give a reliable interpretation of our results, a complete
characterization has been undertaken on these groups [19,
and refs. therein]. The EtNCO adsorption as well as the iso-
topic substitution experiments and other experiments have
allowed us to conclude on the nature and the coordination of
the species produced in the C2H5OH + NO + O2 reaction.
t0: C2H5OHgas maximum,
t1(t0 + 7 s): –NCOads maximum,
t2(t0 + 15 s): NOgas minimum and CO2gas maximum.
t0 could correspond to the maximum ethanol concentra-
tion, followed by the fast formation of cyanide species prob-
ably by the reaction of NO with ethylene, nitroso species
formation, and then isomerization into intermediate oximes,
which transform themselves into nitrile (acetonitrile). The
last one would be cut to give silver cyanide as it has been
shown on Cu–ZSM-5 [17]. In this first step, experimen-
tal proof is lacking to support all these elemental reactions
−
1
The band observed at 2255 cm characterizes the adsorp-
3
+
tion of the –NCO group on octahedrally coordinated Al
−
1
ions, while the band at 2228 cm corresponds to the same
3
+
species coordinated on tetrahedral Al ions. We think that
silver enhances the formation of isocyanate species but that
this molecule rapidly spills over Al2O3 support [19].
To understand this mechanism further, we propose a NCO
formation pathway and we will investigate the reactivity of
these groups toward molecules that may be present among
the reaction compounds.
(
probably because of the short lifetime of intermediate com-
pounds), the only evidence being the presence of ethylene in
the gas phase.
3.3.1. NCO formation
t1 would match with the slower transformation of cyani-
des into isocyanates, which probably needs an intermedi-
ate isomerisation –CN → –NC before oxidation into iso-
cyanate. The isomerization step is also much slower than
oxidation on Cu–ZSM-5 [17]; on the contrary, here, the ox-
idation is much faster since we do not detect the interme-
diate isocyanide. In this second sequence, the experimen-
tal evidence supports our statements, although we have not
been able to show (as in the case of Cu–ZSM-5) isocyanide
Studies in the literature for the past 30 years ago ex-
plain the formation schemes of isocyanate species by the
CO + NO reaction on different metals supported on oxides
[23–31]. Several authors agree that on oxide-based catalysts
such as SiO2, Al2O3, MgO, and TiO2 supporting noble met-
als as Pt, Pd, Ru, Rh, isocyanate species are never observed
on the metal. These elements would play an indirect role
dissociating NO molecules. Then, adsorbed nitrogen atoms
would migrate on the support by spillover. All these authors
propose or show that after migration, the nitrogen atom pref-
erentially reacts by insertion of a CO molecule in the gas
phase more than with adsorbed CO to form the isocyanate
group on the support. Almost all these results have been ob-
tained using FTIR spectroscopy with or without isotopic ex-
change reactions.
−1
−1
species expected at 2060 cm (about 100 cm lower than
the cyanide species wavenumber).
t2 would correspond to the isocyanate hydrolysis using
water due to the alcohol combustion (Fig. 2) to give,
as usually seen, ammonia and carbon dioxide. Ammonia
has been detected in the gas phase but its concentration
is never high enough to allow integration: its activity as
reducing agent in NOx SCR is too important. NH3 formed
by isocyanate hydrolysis classically reacts with another NO
molecule, that could partially explain the concomitance
between NO minimum and CO2 maximum observed in
Fig. 5.
The presence of carbon monoxide and ethanol in the gas
phase does not allow us to follow the nitrogen formation
by mass spectrometry analysis, these 3 compounds giving
a response in 28 mass; it would be interesting to localize the
N2 maximum in our relative time scale.
In an additional isotopic substitution experiment using
C 18O + N O, we have observed C O formed by reac-
16
16
1
8
tion of C O with surface oxygen atoms or with oxygen
atoms from NO molecules. This phenomenon explains why
when the reaction is performed on oxidized Ag/Al2O3 we
observed the same band as for the reaction with unlabeled
molecules. To obtain the expected isotopic shift [19], we
must undertake the same experiment on reduced catalysts
to inhibit the transformation C 18O → C O due to surface
16
1
8
oxygen atoms. Therefore, we have introduced a C O probe
alone on the catalyst. After heating at 873 K, the tempera-
ture is decreased to 473 K and CO is evacuated. At this tem-
perature NO (15 Torr) is introduced. Very quickly, we have
detected N 18O in the gas phase. That means that NO is dis-
In the interpretation of this experiment, several surface
intermediate species are evoked; among them, only cyanide
and isocyanate species have been detected on the catalyst
surface. Their intermediate nature in the SCR of NOx seems
very reasonable, so that a deeper investigation is necessary.
1
8
sociated very easily on the catalyst surface and that C
O
was dissociated previously, leaving 18O and carbon on the

N. Bion et al. / Journal of Catalysis 217 (2003) 47–58
53
It is certainly the signature of aluminum cyanide or sil-
ver cyanide (although ν(CN) on Ag is usually given at
−
1
2
165 cm ).
3
.3.2. Reactivity of NCO species
The first step in the study of isocyanate reactivity consists
in isolating it on Ag/Al2O3 at a temperature in the domain
of activity of this catalyst. However, to prevent a too fast
decomposition of the adsorbed species under argon flow,
we have undertaken these experiments at 573 K. Indeed, at
−1
this temperature, the intensities of the 2255 and 2228 cm
bands do not decrease very quickly.
Fig. 8. Magnification on FTIR spectrum of the adsorbed phase recorded
12
18
14 16
during the
C
O (4 Torr) +
N
O (2.2 Torr) reaction on Ag/Al O
2 3
prereduced by H at room temperature for 10 min and then increasing the
2
temperature to 873 K for 90 min.
3
.3.2.1. Experimental data. Both CO + NO reaction and
adsorption of EtNCO have been used to isolate NCO groups
on the silver/alumina catalyst surface. The sample is placed
in the crucible of the DRIFTS cell and is first activated at
873 K under Ar + O2 (10%). Then, NCO species have been
isolated by:
surface (explaining the black color of the pellet at this mo-
ment). Nevertheless no isocyanate species are observed on
the surface at this temperature. This also shows that it is
necessary that both reactive molecules be present to form
the isocyanate group. Thus, it seems that contrarily to previ-
ous results reported on noble metal supported catalysts, on
Ag/Al2O3 isocyanate species are not formed by linkage of
CO molecules to Nads due to the dissociation of NO. Ac-
cording to our results, the following reactions occur during
the C 18O + N O experiment:
– Sending NO + CO reactive molecule, heating at 873 K,
then decreasing the temperature to 573 K before stop-
ping the NO + CO flow and purging the catalyst under
argon;
– Adsorbing ethyl isocyanate at 573 K; this technique
allows us to obtain much more adsorbed isocyanate
species but it involves the appearance on the surface of
secondary species as carbonates, formates, urethane, etc.
16
C18O → Cads + 18Oads,
N16O → Nads + 16Oads,
Concerning the first technique, the reactive molecules
Nads + Oads → N18O,
18
(
O2, NO, H2O) are introduced 2 min after NO + CO flows
cutting because they are very quickly evacuated, whereas for
the second experimental protocol we let more time for the
evacuation of EtNCO, to be sure that this molecule is absent
from the system lines. The different molecules sent to the
catalyst so impregnated of isocyanates are:
Cads + Oads → C16O.
16
Fig. 8 confirms the above-reported reactions: a spectrum
1
6
18
with both NC O and NC O species is recorded after the
1
8
16
C
O + N O reaction experiment. If we follow the idea
developed in the literature that CO reacts with adsorbed
(i) O , because this molecule is in large excess in the
2
1
8
nitrogen atoms, the band attributed to NC O should be
much more important than the NC 16O corresponding band.
On the contrary, according to our dissociation hypothesis,
the statistic presence of O from CO and NO explains the
equivalent mixture of both isotope groups. A fortiori, when
reactive SCR flow;
(ii) NO, because it is the molecule we want to reduce;
(iii) H O, because this molecule seems to play a role in the
2
mechanism proposed after the preliminary study. All
these compounds are introduced by pulses.
1
6
CO reduced the catalyst, the presence of C O was favored
and so NC 16O became predominant.
The isocyanate species are formed at temperatures higher
than those of the NO dissociation reaction, which is thus the
initial step in the NCO formation mechanism. That explains,
on the one hand, the need to work at high temperatures
3.3.2.2. Reactivity toward O2. In Fig. 9 are grouped the
series of spectra collected into the gas cell and the cor-
responding series collected on the adsorbed phase. NCO
groups are produced by NO + CO reaction; 2 min after the
(
isocyanates species do not appear before 773 K under these
Ar purge, O pulses are sent every minute (30 s for injecting;
2
conditions) and, on the other hand, the predominant role of
silver. Bare alumina is unable to break the CO bond.
We can suppose that cyanide species is one of the recom-
bination products. On the different adsorbed phase spectra,
a broad and weak band is detected at 2150 cm and is
only shifted for nitrogen 15 and carbon 13 substitutions [19].
30 s for recharging the loop). On adsorbed phase spectra, we
discover that each O pulse corresponds to a period when in-
2
−1
tensities of 2255 and 2228 cm bands decrease, a sign of
decomposition or consumption.
−1
From IR gas spectra, we detect that the decrease of NCO
bands is concomitant with the production of a small amount

54
N. Bion et al. / Journal of Catalysis 217 (2003) 47–58
(a)
(a)
(b)
Fig. 10. (a) FTIR chemigrams of adsorbed species (NCO) correlated to
gas species produced during the introduction of O pulses on Ag/Al O
2
2 3
(
b)
previously treated under CO + NO at 873 K. (b) Analysis by mass
spectrometry of O (mass 32) and N (mass 28) during this experiment.
2
2
Fig. 9. FTIR spectra of (a) adsorbed phase and (b) gas phase, collected at
5
73 K during the introduction of 10 µl O pulses on Ag/Al O previously
2 2 3
treated under CO + NO at 873 K.
of N2O as well as with a more important production of CO2.
−1
The NO band localized at 1875 cm , which disappears
after the argon purge and does not reappear later. NO2 is
no longer detected on these spectra. Thus NCO groups are
not decomposed into NO or NO2 and O2. However, the
amount of N2O being weak (taking into account its molar
absorption coefficient), other nitrogen species, which are
not detectable by FTIR, must be produced. To confirm this
deduction, the mass spectrometry analysis is very useful.
Indeed, we have represented it in Fig. 10b, where we
observe by this technique that for each O2 pulse (32 mass
peak), some 28 mass peaks appear. Given the absence
of CO in the IR gas spectra and after subtraction of the
contribution due to CO2, these peaks can only be due to
N2 formation. If we correlate the adsorbed species and gas
product evolution (Fig. 10a), we clearly note that sending
O2 pulses involves the transformation of NCO groups into
N2 and N2O according to the following reactions:
Fig. 11. Analysis by mass spectrometry of NO (mass 30), O (mass 32),
2
and N (mass 28) during the introduction of 12 NO (10 µl) pulses followed
2
by 1 O (10 µl) pulse on Ag/Al O previously treated under CO + NO at
2
2 3
873 K.
3
.3.2.3. Reactivity toward NO. Under the same conditions
as those of previous experiment, we have analyzed the
reactivity of NCO groups when NO pulses are introduced on
Ag/Al2O3 catalysts. The result of this experiment obtained
by mass spectrometry is shown in Fig. 11 where we could
2
2
NCO + O2 → N2 + 2CO2,
3
2
NCO + O2 → N2O + 2CO2.

N. Bion et al. / Journal of Catalysis 217 (2003) 47–58
55
Fig. 12. FTIR spectra of adsorbed species, collected at 573 K during the introduction of 1.5 µl H O pulses on Ag/Al O previously treated under CO + NO
2
2 3
at 873 K.
not detect N2 production peaks concomitant with the NO
peaks. This observation matches with the FTIR spectra
corresponding to the adsorbed phase (not shown), where
we have only remarked that the introduction of NO pulses
does not modify the NCO bands at all. Indeed a weak and
continuous decrease in the NCO evolution curve has been
verified to be caused by the small decomposition of NCO
groups under argon flow at the temperature of the study.
The rate of this decrease is not increased when NO is sent
through the catalytic bed. To support these results we have
sent O2 pulses after the NO pulses. We have observed the
same N2 production as reported above, due to the reactivity
of isocyanates (Fig. 10b).
Fig. 13. FTIR chemigrams of NCO consumption and of CO and NH
2
3
3
gas species produced during the introduction of H O pulses on Ag/Al O
2
2
3
.3.2.4. Reactivity toward H2O. To conclude this study,
previously treated under CO + NO at 873 K.
we have analyzed the behavior of adsorbed isocyanates
species on Ag/Al2O3 catalyst toward H2O. Because of the
technique, we have introduced 1.5 µl H2O pulses instead of
4
. Discussion and conclusion
1
0 µl for the other reacting agents.
Isocyanate species were adsorbed introducing EtNCO.
First, to draw a mechanism for DeNOx reaction via
isocyanate intermediates, it appears useful to summarize
the reactivity experiments reported above. We can establish
a classification of the molecules reacting with isocyanate
species:
As in the case of reactivity toward oxygen, we can clearly
distinguish several decreasing levels in the Al–NCO evolu-
tion curve, while no other new species appear on the catalyst
surface (Fig. 12). However by observing mass spectrome-
try results (not shown), we have noted that H2O pulses are
well consumed by reaction with –NCO groups and that the
product formed by this reaction is neither N2 nor N2O. Mass
results are also confirmed by IR spectra:
– H O is the most reactive molecule because, despite the
2
low amount sent in the injection loop, it is with this
molecule that the isocyanates have disappeared the most
rapidly. This reaction is a hydrolysis, whose products are
NH3 and CO2.
–
–
production of CO2,
production of neither N2O nor NO2.
– O is also reactive and completely selective into N .
Nevertheless a large proportion of the oxygen pulses is
not consumed. And we can suppose that if isocyanates
2
2
Furthermore, the characteristic bending bands δs(N–H)
were very reactive toward O , the first pulses would be
completely consumed.
– NO alone does not react with isocyanates.
2
−1
of gaseous NH3 are detected at 964 and 932 cm (the 17
mass perturbed by the H2O presence cannot permit us to
observe this NH3 production). For each pulse, the evolution
of the gas products curves (CO2 and NH3), reported in
Fig. 13, is concomitant with the consumption of isocyanate
species. Thus, we can conclude that this experimental result
corresponds to the hydrolysis of –NCO groups into NH3.
The reactivity of isocyanate species toward oxygen must
be minimized in the case of competitiveness with water, as
shown by the small amount of oxygen pulses consumed.
Moreover, the strong reactivity of isocyanates species toward

56
N. Bion et al. / Journal of Catalysis 217 (2003) 47–58
water as well as the formation of ammonia is in good agree-
ment with the results obtained during the C2H5OH + NO +
O2 reaction studied in steady state (Figs. 1, 2, and 4). Indeed,
we had noted at the maximum conversion temperatures (623
and 673 K) a decrease of NCO characteristic bands intensi-
ties and simultaneously appearance of NH3 in the gas phase.
Thus, we think that this reaction way may be privileged
rather than the direct reactivity toward O2, even if the
presence of this molecule is necessary to inhibit the NOx
reduction to NH3.
visible on Ag/Al2O3 catalysts as well as on bare alumina,
3
+
could be attributed to aluminum cyanide (Al CN). Thus
−
1
+
the 2127 cm feature is likely assigned to Ag CN. Silver
as well as copper can complex from one to four –CN radi-
−1
cals, which present stretches appearing between 2146 cm
for one radical to 2092 cm for four radicals [20].
−1
This conclusion on the formation of cyanide by CO+NO
reaction does not disagree with the hypothesis proposed in
Section 3.2, because the conditions are completely different:
oxidizing conditions in the case of ethanol pulses, and
reducing conditions in the CO + NO reaction case. In
the SCR of NOx by ethanol in steady state, the gas
phase contains CO from 623 K and the amount of this
molecule increases with the temperature. Then, CO and
In Section 3.2 the time-resolved analysis permitted us
to hypothesize a NOx SCR mechanism on silver/alumina
catalyst developed in three sequences (the asterisk indicates
an adsorption site):
First sequence (hypothetical),
2 5
NO being present, we cannot exclude that in a C H OH
molecule decomposition environment the atmosphere could
be temporarily and locally reducing and that the cyanide
formation by NO + CO reaction might occur.
C2H5OHgas → C2H4gas → nitroso → oxime
→
CH3CN∗ → CN∗;
The study, performed on cyanide and isocyanate localiza-
Second sequence,
+
3+
tion, shows that if –CN groups are linked to Ag and Al
ions, –NCO groups are only detected on Al3 ions. The
+
CN∗ → NC∗,
–
CN → –NCO transformation, which seems to be the slow
1
2
step of the mechanism, is more complex than expected since
we observe Ag CN to finally obtain Al NCO. Therefore
two hypotheses can be formulated:
NC∗ + O2 → NCO∗;
+
3+
Third sequence,
NCO∗ + 2H2O → OH∗ + NH3 + CO2,
+
+
(
a) Ag CN can isomerize into isocyanide species (Ag NC),
+
and can oxidize into silver isocyanate (Ag NCO), and
4
NH3 + 4NO + O2 → 4N2 + 6H2O,
then would give Al3 NCO by spillover;
+
or alternatively
(b) (a more interesting hypothesis that needs less elemen-
tary steps) a transformation by intermediate complexa-
tion between silver and aluminum ions can take place:
2
NCO∗ + O2 → N2 + 2CO2 + 2∗.
The experiments on the nature, the localization, and the
Ag CN Al → Ag CN → Al → Ag+ CNAl
+
3+
+
3+
3+
reactivity of cyanide and isocyanate species have given
us complementary information, which allow us to define
several steps of the pathway.
The study undertaken by isotopic substitution experi-
ments on the nature of the species detected during the
C2H5OH + NO + O2 reaction supports their assignment
to cyanides and isocyanates. This study also reveals the
probable formation of cyanide and isocyanate species from
(
I)
O2
→ Ag
+
3+
.
OCNAl
This bridged complex (I) has already been identified in
the case of nickel cyanide linked to BF3 [31]. Nakamoto [20]
explains that the cyanide group M–C≡N can be bridged to
ꢀ
form M–C≡N–M involving a blue shift of the vibration
1
6
18
NO+CO reactions. In particular, the use of N O and C
O
ν(CN). For example, this vibration in K2[Ni(CN)4] complex
−1
−1
on Ag/Al2O3 catalysts showed that NO dissociation would
occur at 474 K and CO dissociation at about 673 K. The first
intermediate formed would be the adsorbed cyanide species
contrary to what was expected (the bond of a CO molecule
with an adsorbed nitrogen atom). The cyanide due to the re-
action between N and C atoms is probably initially formed
absorbs at 2130 cm and the band is shifted to 2250 cm
in K2[Ni(CN)4] · 4BF3 due to the complexation Ni–C≡N–
BF3 [32]. The existence of this bridged species let us con-
sider a reversal of the CN group: a cyanide linked to a sil-
ver atom or to an aluminum atom could give an isocyanide
linked to a neighboring aluminum ion acceptor of the ni-
trogen free doublet. The absence of bridged species charac-
teristic vibration bands on our spectra can be explained by
their instability (that would involve the reversal as soon as
formed) and by the quick oxidation of isocyanide species
+
on Ag . Cyanide species are not directly detected in the
NO + CO reaction but the characteristic wavenumbers of
these groups have been observed under transitory conditions
−1
during ethanol pulses experiments (2155 and 2127 cm at
73 K under dynamic conditions) or in the EtNCO decom-
as described in the Scheme 1. Al3 ions are also, as well
as BF3, strong Lewis acid sites and we can imagine that an
+
6
−1
position experiments (2165 cm at 573 K under static con-
ditions) [19], proving that their evanescence is likely due to
+
3+
unstable bridged complex Ag CN→Al could be formed
−1
+
3+
their high reactivity. The latter bands (2165 and 2155 cm ),
and provoke a Ag → Al reversal of the group. In this

N. Bion et al. / Journal of Catalysis 217 (2003) 47–58
57
Scheme 1. Proposed formation mechanism of NCO groups in a CO + NO reaction on Ag/Al O catalysts.
2
3
case, the oxidation would happen on the aluminum iso-
cyanide. One argument, which matches with this hypothesis
is the fact that the formation and the distribution of –NCO
groups are different between EtNCO decomposition and
CO+NO or C2H5OH+NO+O2 reactions; in the first case,
Acknowledgments
The authors are grateful to Rhodia Electronics and
Catalysis researchers for sample preparation and for fruitful
discussions.
the main Al3 NCO band is situated at about 2260 cm
+
−1
,
3
+
whereas in the other cases the most intense Al NCO band
−1
is at 2230 cm . The formation of bridged species can only
be possible in the CO + NO and C3H6 + NO + O2 reactions
where cyanide species are intermediate. To form the bridged
complex, the nitrogen atom can choose between two alu-
References
[
[
1] http://www.rhodia-ec.com/site_ec_us/catalysis/index_automotive.htm.
2] A. Ueno, H. Suzuki, Y. Kotera, J. Chem. Soc., Faraday Trans. I 79
(1983) 127, and references therein.
minum sites in γ -Al2O3. The tetrahedrally coordinated Al3
+
site (more acidic by Lewis considerations than the octahe-
[3] Y. Kintaichi, H. Hamada, M. Tabata, M. Sasaki, T. Ito, Catal. Lett. 6
(1990) 239.
[4] Y. Torikai, H. Yahiro, N. Mizuno, M. Iwamoto, Catal. Lett. 9 (1991)
3
+
drally coordinated Al site) is privileged to host the free
−1
doublet of nitrogen, which explains the band at 2230 cm
.
9
1.
5] H. Hamada, Y. Kintaichi, M. Sasaki, T. Ito, M. Tabata, Appl. Catal. 75
1991) L1.
To conclude, all these results allow us to propose the
following mechanism for SCR of NOx by ethanol on
silver/alumina catalysts:
[
(
[6] T. Miyadera, Appl. Catal. B 2 (1993) 199.
First sequence (2 hypotheses),
[7] R. Burch, P.J. Millington, A.P. Walker, Appl. Catal. B 4 (1994) 65.
[8] K.A. Bethke, M.C. Kung, B. Yang, M. Slah, D. Alt, C. Li, H.H. Kung,
(
(
a) C2H5OHgas → C2H4gas → nitroso → oxime
Catal. Today 26 (1995) 169.
[9] H. Hamada, Y. Kintaichi, M. Inaba, M. Tabata, T. Yoshinari, H.
Tsuchida, Catal. Today 29 (1996) 53.
+
→
CH3CNads → Ag CNads,
[
10] M. Haneda, Y. Kintaichi, H. Shimada, H. Hamada, Chem. Lett. (1998)
+
b) C2H5OH + NO + O2 → CO + NO → Ag CNads;
181.
Second sequence,
[11] T. Maunula, Y. Kintaichi, M. Inaba, M. Haneda, K. Sato, H. Hamada,
Appl. Catal. B 15 (1998) 291.
[12] T. Nanba, A. Uemura, A. Ueno, M. Haneda, H. Hamada, N. Kakuta,
H. Miura, H. Ohfune, Y. Udagawa, Bull. Chem. Soc. Jpn. 71 (1998)
2331.
13] K. Takagi, T. Kobayashi, H. Ohkita, T. Mizushima, A. Kakuta, N. Abe,
K. Yoshida, Catal. Today 45 (1998) 123.
14] T. Maunula, Y. Kintaichi, M. Haneda, H. Hamada, Catal. Lett. 61
+
3+
+
3+
Ag CN Al
→ Ag CN → Al
→
tetra
tetra
O2
+
3+
3+
tetra
Ag , CNAl
→ Al NCO;
tetra
[
[
[
Third sequence,
Al NCO + 2H2O → Al3+ –OH + NH3 + CO2,
3
+
tetra
tetra
(1999) 121.
15] M. Haneda, N. Bion, M. Daturi, J. Saussey, J.-C. Lavalley, D. Duprez,
H. Hamada, J. Catal. 206 (2002) 114.
[16] K.A. Bethke, H.H. Kung, J. Catal. 172 (1997) 93.
[17] F. Poignant, PhD thesis, Caen, 1997.
18] F. Poignant, J.-L. Freysz, M. Daturi, J. Saussey, J.-C. Lavalley, in:
A. Corma, F.V. Melo, S. Mendioroz, J.L.G. Fierro (Eds.), 12th Inter-
national Congress on Catalysis, in: Stud. Surf. Sci. Catal., Vol. 130,
Elsevier, 2000, pp. 1487–1492.
19] N. Bion, J. Saussey, C. Hedouin, T. Seguelong, M. Daturi, Phys.
Chem. Chem. Phys. 3 (2001) 4811.
4
NH3 + 4NO + O2 → 4N2 + 6H2O.
This mechanism is proposed on Ag/Al2O3 catalysts
and the reducing agent used is C2H5OH. Nevertheless, the
results obtained on Ga2O3/Al2O3 catalysts with propene
as reducing agent allow us to consider that the mechanism
could be generalized to other alumina-based catalysts and
to different hydrocarbons [15]. Indeed, Haneda et al. [15]
have shown that in the C3H6 +NO+O2 reaction, isocyanate
species are formed on Ga2O3/Al2O3 surfaces. Moreover, on
the same catalyst, the NH3 + NO + O2 reaction has shown
that ammonia selectively reduces NO into N2, in oxygen
excess. In this case, gallium will operate as silver does. For
these reasons we can hypothesize that the mechanism of
NOx SCR by hydrocarbon on alumina-based catalysts can
follow the same steps as those for Ag/Al2O3.
[
[
[20] K. Nakamoto, Infrared and Raman Spectra of Inorganic and Coordi-
nation Compounds, Wiley–Interscience, New York, 1986.
[
[
[
[
[
21] C. Morterra, G. Magnacca, Catal. Today 27 (1996) 497.
22] K.I. Hadjiivanov, Catal. Rev.-Sci. Eng. 42 (2000) 71.
23] J. Raskó, F. Solymosi, J. Chem. Ser. Faraday Trans. 1 76 (1980) 2383.
24] Y. Ukisu, T. Miyadera, A. Abe, K. Yoshida, Catal. Lett. 39 (1996) 265.
25] S. Sumiya, A. Abe, N. Takesawa, K. Yoshida, J. Chem. Soc. 94 (1998)
2217.

58
N. Bion et al. / Journal of Catalysis 217 (2003) 47–58
[
[
[
26] S. Kameoka, T. Chafik, Y. Ukisu, T. Miyadera, Catal. Lett. 55 (1998)
11.
27] S. Kameoka, Y. Ukisu, T. Miyadera, Phys. Chem. Chem. Phys. 2
2000) 367.
28] F. Solymosi, L. Völgyesi, J. Sárkány, J. Catal. 54 (1978) 336.
[29] F. Solymosi, L. Völgyesi, J. Raskó, Z. Phys. Chem. Neue Folge 120
(1980) 79.
[30] W.C. Hecker, A.T. Bell, J. Catal. 85 (1984) 389.
[31] F. Solymosi, J. Raskó, Appl. Catal. 10 (1984) 19.
[32] D.F. Shriver, J. Am. Chem. Soc. 84 (1962) 4610.
2
(