The role of AgOAl species in silver-alumina catalysts for the selective catalytic reduction of NOx with methane

Article abstract of DOI:10.1016/j.jcat.2005.09.036

We examined the role of silver and alumina in Ag-alumina catalysts for the selective catalytic reduction (SCR) of NO_x by methane in gas streams containing excess oxygen. A cogelation technique was used to prepare Ag-alumina materials with high dispersion of silver even at high metal loadings (>10 wt%) and after air calcination at 650 °C. Typically, a part of silver is present as fine nanoparticles on the alumina, whereas another part is ionic, bound with the alumina as [AgOAl] species. Dilute nitric acid leaching was used to remove the silver particles and all weakly bound silver from the surface of these materials. Complementary structural characterization was performed by HRTEM, XPS, XRD, and UV-vis DRS. We found that the higher the initial silver content, the higher the amount of the residual [AgOAl] species after leaching. NO-O₂-TPD tests identified that silver does not modify the surface properties of the alumina. The SCR reaction-relevant NO_x adsorption takes place on alumina. Temperature-programmed surface reaction (TPSR) and kinetic measurements at steady state were used to check the reactivity of the adsorbed NO_x species with methane and oxygen to form dinitrogen. Only the alumina-adsorbed nitrates react with CH₄ to produce N ₂ in the presence of oxygen, beginning at ～300 °C as found by TPSR. Moreover, the SCR reaction rates and apparent activation energies are the same for the leached and parent Ag-alumina catalysts. Thus, metallic silver nanoparticles are spectator species in CH₄-SCR of NO_x. These catalyze the direct oxidation of methane at temperatures as low as 300 °C, which explains the lower methane selectivity for the SCR reaction measured over the parent samples.

Full text of DOI:10.1016/j.jcat.2005.09.036

Journal of Catalysis 237 (2006) 79–93
www.elsevier.com/locate/jcat
The role of Ag–O–Al species in silver–alumina catalysts for
the selective catalytic reduction of NOx with methane
X. She, M. Flytzani-Stephanopoulos ∗
Department of Chemical and Biological Engineering, Tufts University, Medford, MA, USA
Received 5 July 2005; revised 21 September 2005; accepted 26 September 2005
Available online 23 November 2005
Abstract
We examined the role of silver and alumina in Ag–alumina catalysts for the selective catalytic reduction (SCR) of NO by methane in gas
x
streams containing excess oxygen. A cogelation technique was used to prepare Ag–alumina materials with high dispersion of silver even at high
◦
metal loadings (>10 wt%) and after air calcination at 650 C. Typically, a part of silver is present as fine nanoparticles on the alumina, whereas
another part is ionic, bound with the alumina as [Ag–O–Al] species. Dilute nitric acid leaching was used to remove the silver particles and all
weakly bound silver from the surface of these materials. Complementary structural characterization was performed by HRTEM, XPS, XRD,
and UV–vis DRS. We found that the higher the initial silver content, the higher the amount of the residual [Ag–O–Al] species after leaching.
NO–O -TPD tests identified that silver does not modify the surface properties of the alumina. The SCR reaction-relevant NO adsorption takes
x
2
place on alumina. Temperature-programmed surface reaction (TPSR) and kinetic measurements at steady state were used to check the reactivity
of the adsorbed NO species with methane and oxygen to form dinitrogen. Only the alumina-adsorbed nitrates react with CH to produce N in
x
4
2
◦
the presence of oxygen, beginning at ∼300 C as found by TPSR. Moreover, the SCR reaction rates and apparent activation energies are the same
for the leached and parent Ag–alumina catalysts. Thus, metallic silver nanoparticles are spectator species in CH -SCR of NO . These catalyze
x
4
◦
the direct oxidation of methane at temperatures as low as 300 C, which explains the lower methane selectivity for the SCR reaction measured
over the parent samples.
2005 Elsevier Inc. All rights reserved.
Keywords: Silver; Alumina; Leaching; Silver aluminate; Selective catalytic reduction of NO; Methane oxidation; Selectivity; Nitrogen oxides; TPD;
Temperature-programmed surface reaction
1. Introduction
review by Burch et al. [13] discusses these types of catalysts,
which include rare earth oxides (REO) and alumina- (or other
oxides) supported Pt, Ga, In and Pd catalysts. The platinum
group catalysts are active at low temperatures, but are limited by
a narrow operating temperature window in which they display
good selectivity. They are also more selective to undesired N2O,
a potent greenhouse by-product. Ga and In catalysts are limited
by loss of activity in the presence of water; and Pd catalysts
are inhibited by excess of oxygen in the exhaust gas stream.
Other reports in the literature point to the enhancement of both
activity and stability using sulfated alumina or zirconia as sup-
ports for Pd [14–16], Mn [17] and Co [18,19] for the SCR of
NO with methane. The acidity of these supports was considered
key to stabilizing the active state of the metal species.
The selective catalytic reduction (SCR) of NO to N2 with hy-
drocarbons is a promising technology for NOx removal, having
attracted much attention for nearly two decades. Initial stud-
ies were focused on zeolites loaded with metals, such as Cu,
Co, Ga, In, and Pd [1–12], since zeolites are well known to sta-
bilize metal ions, the presence of which is deemed necessary
for the SCR reaction. However, zeolite-based catalysts suffer
from deactivation in water vapor- and sulfur dioxide-containing
exhaust gas streams, which renders them less attractive for prac-
tical applications. Metal oxide catalysts have been examined as
alternatives due to their high hydrothermal stability. A recent
*
Since Miyadera et al. [20] reported that silver/alumina cata-
lysts exhibited relatively good activity and selectivity for NO
Corresponding author.
E-mail address: mflytzan@tufts.edu (M. Flytzani-Stephanopoulos).
0021-9517/$ – see front matter
doi:10.1016/j.jcat.2005.09.036
2005 Elsevier Inc. All rights reserved.

80
X. She, M. Flytzani-Stephanopoulos / Journal of Catalysis 237 (2006) 79–93
reduction to N2 and moderate resistance to water and sul-
fur dioxide, many studies have been performed on this cata-
lyst system [21–41]. The most commonly used reductants are
propane/propene [20–27], but there have also been reports on
using higher hydrocarbons [28–33]; oxygenated hydrocarbons
[34–40]; and even methane [41] as reducing agents. Higher
hydrocarbons were found to shift the active temperature win-
dow to lower values and increase the tolerance to water [28].
In particular, using octane as the reductant has attracted much
interest in the recent literature [30–33], with reported light-off
temperature as low as 250 ◦C and optimal C:N ratios of 4–6.
Interestingly, it has been reported that gas-phase reactions past
the catalyst bed can contribute to the reduction of nitric oxide
to dinitrogen [30,31]. However, a significant drawback of using
octane as a reductant is the significant formation of CO accom-
panying the NO reduction reaction. With oxygenated hydro-
carbons, high NO conversions can be obtained at 250–400 ◦C;
however, the main problem is the formation of a large amount
of harmful nitrogen-containing byproducts [34–40]. An impor-
tant finding for the Ag/alumina system is that the activity for
NO reduction is strongly correlated to the silver loading, and the
most active Ag–alumina catalysts have been reported to contain
1.2–3 wt% Ag [23–26,31,33,41]. Structural analysis identified
oxidized silver in the SCR-active Ag/alumina catalysts of inter-
found to dominate in the SCR-active (3.2 wt% Ag) catalysts by
UV–vis DRS [41].
Thus, evidence points to oxidized silver on alumina as the
NO-SCR sites and attributes the loss of selectivity to the pres-
ence of silver nanoparticles, which catalyze the direct oxidation
of the hydrocarbon, including methane [23,41,42]. However,
this general picture cannot explain how methane is activated
on silver aluminate structures. Is the presence of silver particles
needed for this? How does the presence of silver modify the
alumina surface? In a recent paper [47], the importance of Ag+
ions for CH4-SCR of NO was shown using Ag-ZSM-5. Acti-
vation of CH4 was suggested to occur on isolated Ag ions in
the zeolite. It is of interest to investigate the role of silver ions
for CH4-SCR of NO in Ag–alumina catalysts with no silver
particles present. To our knowledge, no such particle-free Ag–
alumina catalysts have been examined in the literature. Even
with the most active, low-content silver–aluminas, some sil-
ver particles were always present, depending on the preparation
method and calcination temperature.
In this work, dilute nitric acid leaching was used to suc-
cessfully remove metallic silver particles from the as-prepared
Ag–alumina catalysts. The resulting Ag–O–Al surfaces were
characterized and compared with unmodified aluminas, and in
kinetic experiments the role of silver particles for the CH4-SCR
of NO was clarified by comparing the leached and parent cata-
lysts.
0
mediate silver loading, whereas metallic silver particles or Agn
clusters were dominant in the high-silver content alumina cata-
lysts, which were less selective for NO reduction and were good
for the direct combustion of hydrocarbons [23–26,29,41]. This
also holds true for methane. In their study of methane combus-
tion over Ag/ZrO2, Kundakovic et al. [42] found that metallic
Ag nanoparticles (oxygen-covered) are excellent catalysts for
the direct oxidation of methane.
2. Experimental
2.1. Catalyst preparation
2.1.1. One-pot co-gelation method
Ag–alumina catalysts were prepared by a one-pot co-
precipitation-gelation method as described previously [41]. Ac-
cordingly, aluminum nitrate (Fluka, 99% or Alfa, 98%) and
silver nitrate (Aldrich, 99.995%) were used as precursors. Af-
ter the desired amount of these two nitrate salts was dissolved
in deionized water, the precipitation agent, tetramethylammo-
nium hydroxide solution (25%, Fluka), was added dropwise at
room temperature until a yellow/gray-colored precipitant ap-
peared (pH = 10–11). This was left to gel at this condition
for 12 h, then filtered, washed with deionized water several
times, and dried in a vacuum oven (25 in Hg, 60 ◦C) for 10–
12 h. The dried solid was crushed thoroughly, then heated at
a rate of 2 ◦C/min up to 650 ◦C and kept at this temperature
for 5 h. The elemental compositions of the final catalysts were
analyzed by inductively coupled plasma (ICP) using a Leeman
Labs PS-1000 instrument. The as-prepared Ag–alumina cata-
lysts are denoted as AlAg(x,CG), where x is the wt% of Ag
and CG denotes the coprecipitation-gelation method applied.
In addition to the change in the silver oxidation states, Wang
et al. [43] explained the loading effect of silver on alumina as
relating to acidity changes of the alumina surface. Finally, an-
other type of structure, silver–aluminate, has been suggested
as important for NO SCR [22,23]. Nakatsuji et al. [44] ob-
tained a AgAlO2/Al2O3 catalyst by hydrothermal treatment at
high temperature that was even more active than the respec-
tive Ag/Al2O3 catalyst. However, it has also been observed that
silver phases exhibit significant mobility under SCR reaction
conditions [26,27], thus contradicting the proposal of stable sil-
ver aluminate.
The proper synthesis method is crucial to the preparation of
catalysts with desirable silver structures. Impregnation methods
are limited by silver particle aggregation at relatively high silver
loadings [23–26], and the catalysts thus prepared have a very
narrow active temperature window. In contrast, coprecipitation-
gelation methods have been found to better disperse and stabi-
lize oxidized silver, possibly due to better interaction of silver
species with the alumina during the gelation process [41,45,
46]. A broader active temperature window was reported for a
5 wt% Ag–Al2O3 prepared by a sol–gel technique [46]. In our
previous work, Ag–alumina catalysts prepared by a single step
co-gelation method [41] were active for the SCR of NO with
methane in excess O2. Silver ions and silver oxide clusters were
2.1.2. Nitric acid leaching
Leaching by dilute nitric acid solutions was the technique
chosen to remove weakly bound silver from the Ag–Al2O3
samples by immersing each sample (∼1.5 g) in 100 ml of 10%
HNO3 at room temperature for 8 h. After leaching, the recov-
ered solids were washed with deionized water several times,

X. She, M. Flytzani-Stephanopoulos / Journal of Catalysis 237 (2006) 79–93
81
dried in a vacuum oven (25 in Hg, 60 ◦C) for 10–12 h, and
heated at a rate of 2 ◦C/min to 650 ◦C and kept there for 3 h.
Again, ICP analysis measured the amount of residual silver
in the leached catalysts. The leached samples are denoted as
AlAg(x, L), where x is the wt% of Ag, and L indicates a
leached sample.
rate of 100 ml/min were used at this space velocity. The sam-
ples were tested in the temperature range of 300–700 ◦C in 50◦-
steps. At each temperature, data were collected for 1–1.5 h after
steady state was established. After the ascending-temperature
tests, descending-temperature tests were run to check for possi-
ble catalyst deactivation. The product gas stream was analyzed
by a gas chromatograph (HP 5890) equipped with a 10-ft×1/8-
inch diameter 5A molecular sieve column and a thermal con-
ductivity detector (TCD). NO, CH4, O2, N2, and CO species
could be separated by this column. Catalysts were ranked by
comparing the values of xNO, xCH , and SCH , representing con-
2.2. Catalyst characterization
The BET specific surface area of each sample was measured
by single-point N2 adsorption/desorption on a Micromerit-
ics Pulse Chemisorb 2705 instrument. HRTEM/EDS was per-
formed on a JEOL 2010 instrument equipped with a LaBr6 elec-
tron gun source with a resolution of 0.14 nm. The microscope
was operated at 200 kV, and had an attachment for X-ray dis-
persive spectroscopy (EDS) for elemental analysis of selected
areas. The catalyst preparation involved suspending the sample
in isopropyl alcohol in an ultrasonic bath and then depositing it
on a carbon-coated 200-mesh Cu grid.
X-Ray powder diffraction (XRD) analysis was performed on
a Rigaku 300 X-ray diffractometer. Copper-Kα radiation was
used. The tube voltage was 60 kV, and the current was 300 mA.
The oxidation states of silver and the atomic ratio of silver
to aluminum in the surface region of the Ag–alumina catalysts
were examined on a Perkin–Elmer Model 5200C X-ray photo-
electron spectrometer (XPS), with an Al-Kα anode used as the
X-ray source. All samples tested by XPS were in as-prepared
form. Each powder sample was pressed onto a copper foil, then
placed into the vacuum chamber without any pretreatment. The
Al 2p signal of Al2O3 was used as internal reference to correct
the XP spectra [25]. The atomic ratio of Ag/Al was based on
the core level spectra of Ag 3d and Al 2p.
4
4
version of NO to N2, conversion of CH4 to COx, and selectivity
of CH4 for NO reduction, respectively, defined as follows:
2[amount of N2 produced]
xNO
xCH
=
,
(1)
(2)
(3)
[initial amount of NO]
[amount of reacted CH4]
=
,
4
[initial amount of CH4]
2[amount of N2 produced]
[amount of reacted CH4]
SCH
=
.
4
2.4. Reaction rate measurements
Reaction rate measurements were made using the same re-
actor system as described above. Samples with a particle size of
53–150 µm were used for these tests. To keep the reactions in
the kinetic regime, the conversion of NO was kept at <10% and
the conversion of CH4 was <30%. Typically, the reaction rates
were measured in a feed gas mixture of 0.5% NO–0.5% CH4–
5% O2/He at a temperature range of 425–525 ◦C. Steady-state
rates were recorded at each temperature. After measurements in
the ascending mode, temperatures were reduced and measure-
ments repeated in the descending mode, which also checks for
any catalyst deactivation. A blank test was also performed with
empty reactor at the above conditions to check for gas-phase
reactions.
UV–vis diffuse reflectance spectra (DRS) were collected
with a Hewlett Packard 8052A diode array spectrophotometer
equipped with a diffuse reflectance attachment (DRA) (Har-
rick). The UV–vis spectra of alumina and Ag–alumina sam-
ples were collected at room temperature in air in the range of
190–820 nm with a resolution of 2 nm. MgO powder (99.99%,
Aldrich) was used as a reference.
The rate of NO reduction to N2 (−rNO) and the overall CH4
oxidation rate (−rCH ) were calculated as follows:
4
−rNO = FNOXNO/W (mol/(gs))
and
(4)
2.3. Microreactor activity tests
−rCH = FCH XCH /W (mol/(gs)),
(5)
4
4
4
Catalyst activity tests were performed in a quartz tube (1 cm
i.d.) packed-bed flow microreactor. The catalyst powder (finely
ground) was held with a layer of quartz wool on a quartz frit
at the center of the reactor. The reactor was heated inside an
800-W Lindberg furnace. A K-type thermocouple embedded in
the catalyst bed from the top of the reactor and connected to an
Omega temperature controller was used to monitor the bed tem-
perature. Four gas mixtures of grade 5 purity (3% NO/He, 10%
CH4/He, 20% O2/He, and pure He) were blended to achieve
the desired reaction gas stream composition. Each feed gas flow
rate was measured and controlled independently by a mass flow
controller. All tests were performed at atmospheric pressure.
Most tests were conducted under a reactant gas composi-
tion of 0.25% NO–2% CH4–5% O2/He and a space velocity of
9000 h−1 (NTP). Typically, a 0.4 g sample load and a gas flow
where W is the catalyst loading (in g), Fi is the molar flow rate
(i = NO, CH4, in mol/s), and Xi is the conversion of NO to
N2 or CH4 to COx (i = NO, CH4). Reaction rates, normalized
by the surface area, were calculated by dividing these rates by
the initial specific surface area of each sample. Apparent activa-
tion energies were calculated from Arrhenius-type plots of the
reaction rate.
2.5. Temperature-programmed desorption (TPD) tests
NO–O2-TPD was performed to study the surface adsorption
of NO on the various types of catalysts in the presence of O2,
as is the case under SCR conditions. NO and O2 were used
as adsorbates. TPD tests were performed in a quartz fixed-bed

82
X. She, M. Flytzani-Stephanopoulos / Journal of Catalysis 237 (2006) 79–93
flow reactor (1 cm i.d.) coupled to a quadrupole mass spec-
trometer (MKS-PPT-200EM). Generally, 0.1 g of catalyst was
held between two layers of quartz wool in the reactor. The tem-
perature was controlled by a thermocouple in contact with the
upper layer of quartz wool. The sample was pretreated at 500 ◦C
for 1.5 h in 50 ml/min of pure helium, then cooled to room
temperature in helium. A gas stream of 1.5% NO–5% O2 at
50 ml/min was then flowed over the sample until the NO and
O2 signals monitored by the mass spectrometer leveled off, nor-
mally around 2.5 h. Then the catalyst was purged with helium
for ∼10 h to remove any weakly adsorbed species. Desorp-
tion was performed by flowing 50 ml/min He or 5% O2/He (to
simulate the lean-burn condition with excess O2) and ramping
the temperature from ambient temperature to 700 ◦C at a rate
of 10 ◦C/min. The outlet gas composition was continuously
monitored by mass spectrometry; typically, NO (m/e = 30),
NO2 (m/e = 46), N2O (m/e = 44), O2 (m/e = 32), and N2
(m/e = 28) were monitored. The mass spectrometer was cal-
ibrated for NO, and therefore quantitative analysis of the des-
orbed NO was possible. Whenever desorption of NO and NO2
appeared together, the NO signal was corrected to account for
cracking of NO2 to NO in the chamber. Accordingly, 2.7 times
the NO2 signal was subtracted from the NO signal, in accor-
dance with the literature [48]. Only the corrected NO signals
are presented here.
of catalyst was used, after pretreatment in He at 500 ◦C for
1.5 h. Consumption of CH4 (m/e = 15) and O2 (m/e = 32)
was monitored by mass spectrometry, along with production
of CO (m/e = 28) (after subtracting the amount of CO due to
cracking of CO2 in the chamber), CO2 (m/e = 44), and H2O
(m/e = 18). To check the contribution of gas-phase combus-
tion of CH4, blank tests were performed with an empty reactor
at the same conditions.
3. Results and discussion
3.1. Catalyst characterization
Table 1 lists the parent and nitric acid-leached Ag–alumina
catalysts examined in this work. Leaching effectively re-
moved all weakly bound silver from the sample. For exam-
ple, AlAg(0.8,L), a leached sample with 0.8 wt% Ag, was
obtained from the parent catalyst AlAg(4.9,CG). Thus, only
∼16% of the original silver remained in the sample after
leaching. Accompanying the removal of silver, a small de-
crease in surface area (from 260 to 223 m2/g) was observed
in the leached sample, which could be due to the further
calcination of the leached catalyst at 650 ◦C for 3 h. Sim-
ilarly, AlAg(1.1,L) was obtained from the parent catalyst
AlAg(5.8,CG). For the parent catalysts with higher silver load-
ings, such as AlAg(9.7,CG) and AlAg(21.5,CG), the corre-
sponding leached samples, AlAg(2.6,L) and AlAg(6,L), respec-
tively, retained higher amounts of bound silver. Interestingly,
the surface areas of these two leached samples increased com-
pared with the parent catalysts, for example, from 191.4 m2/g
in AlAg(21.5,CG) to 222.5 m2/g in AlAg(6,L). Surface area
increase may be due to the removal of silver particles block-
ing pores in the as-prepared Ag–alumina catalysts with high
Ag content. A small loss of alumina (3–5 wt%) also occurred
on leaching. For comparison, the as-prepared silver-free alu-
mina, Al2O3(CG), was treated similarly in nitric acid, and the
sample thus obtained is denoted as Al2O3(L). The surface area
2.6. Temperature-programmed surface reaction (TPSR) tests
TPSR experiments were conducted in the same apparatus as
the TPD tests by first adsorbing NO and O2 on the catalysts
and then ramping up the temperature in a flowing gas stream
containing CH4 and O2. A total of 0.1 g of catalyst was used
for TPSR. The adsorption and purging procedure was similar to
that for the NO–O2-TPD described above. The surface reaction
was carried out by flowing a mixture of 2% CH4 and 5% O2
over the catalyst and increasing the temperature to 700 ◦C at a
rate of 10 ◦C/min. The following reactants and products were
monitored by mass spectrometry: CH4 (m/e = 15), O2 (m/e =
32), NO (m/e = 30), NO2 (m/e = 46), N2/CO (m/e = 28),
CO2/N2O (m/e = 44), and H2O (m/e = 18). The m/e ratio
of 15 was used for CH4 to distinguish it from atomic O. Sim-
ilar to the TPD data, the NO signal was corrected (2.7 × NO2
value was subtracted from the NO recorded) when both NO
and NO2 were eluted during TPSR. FTIR gas analysis was
used to check for CO production under these conditions, be-
cause CO and N2 are indistinguishable by mass spectrometry
(both having m/e = 28), and also for distinguishing CO2 from
N2O. The FTIR apparatus (Mattson, Research Series 1) was
equipped with a 0.75 L/5.6 m cell operating at 150 ◦C. Gen-
erally, 100 scans were used to collect the FTIR spectra with
a resolution of 4.0 cm−1 and a gain of 4.0. Both the mass
spectrometer and FTIR device were calibrated for selected gas
components, to quantify the N2, CO, and CO2 produced.
Table 1
Properties of Ag–alumina catalysts
a
b
c
d
Sample
BET SA Sample
BET SA
Surface Ag
e
2
e
2
(Ag, bulk wt%) (m /g
)
(Ag, bulk wt%) (m /g
)
content (wt%)
cat
cat
AlAg(4.9,CG)
AlAg(5.8,CG)
AlAg(9.7,CG)
260.9
248.6
220.7
AlAg(0.8,L)
AlAg(1.1,L)
AlAg(2.6,L)
AlAg(6,L)
223 (171.5) 0.9
229.3 (171.3) 0.9
230 (175.8)
222.5 (200.6) 4.7
205.5
–
AlAg(21.5,CG) 191.4
Al O (CG) 220.6
Al O (L)
2 3
–
2
3
a
Samples in this column are as prepared by cogelation, after air calcination
◦
at 650 C for 5 h.
b
Samples in this column are the corresponding nitric acid leached materials,
for example, AlAg(0.8,L) was leached from AlAg(4.9,CG), etc. All leached
◦
samples were calcined at 650 C for 3 h.
c
The value in the parenthesis is the BET surface area of used sample at the
conditions of Fig. 5 at 300–700 C for ∼30 h, except the AlAg(6,L) run at
◦
Reaction light-off in methane combustion was examined
over several catalysts by flowing a mixture of 2% CH4–5%
O2–He at 50 ml/min and ramping the temperature from room
temperature 700 ◦C at a rate of 10 ◦C/min. Generally, ∼0.1 g
◦
600 C for ∼14 h.
d
Surface content of silver in the leached samples was measured by XPS.
ICP analysis was used to measure the total amount of silver in the parent
e
and leached samples.

X. She, M. Flytzani-Stephanopoulos / Journal of Catalysis 237 (2006) 79–93
83
(a)
Fig. 2. XRD of alumina and Ag–alumina catalysts after calcination in air at
650 C.
◦
was also used to analyze the other high-silver-content catalyst,
AlAg(21.5,CG-850), which was calcined at 850 ◦C for 5 h in an
attempt to form silver aluminate [44]. As can be seen in Fig. 2,
strong diffraction lines due to metallic silver (2θ ≈ 38.1◦,
44.3◦, 64.5◦, 77.6◦) and silver aluminate (2θ ≈ 33.7◦) were
observed in AlAg(21.5,CG-850C) (line 6), and the silver parti-
cle size was calculated as 30.5 nm. Thus, the high-temperature
calcination (850 ◦C) and high silver loading causes agglomer-
ation of silver particles, in comparison with the <10 nm silver
nanoparticles shown in Fig. 1a for AlAg(4.9,CG). The crystal-
lite size of alumina was calculated as ∼5 nm based on peaks
at 2θ ≈ 45.8◦ and 66.8◦ [49]. For the two leached Ag–alumina
samples, and even for AlAg(11,CG), the diffraction peaks were
due only to the alumina phase (indicated by the dotted lines),
and no metallic Ag was identified in these samples. However,
a weak new peak appeared at 2θ ≈ 33◦ for AlAg(7.7,L) which
is assigned to silver aluminate [44,49]. Some new features also
appear around the peaks of alumina at 20◦ and 60◦, where lines
due to silver aluminate also exist. Because the carrier alumina
is of small crystallite size, it exhibits broad diffraction peaks at
these positions, making it difficult to discern peaks due to silver
aluminate. In the case of the AlAg(0.8,L) sample, even though
no silver aluminate lines were detected, this may be due to the
small amount of silver and high dispersion. Silver aluminate
has been proposed as an active species for SCR in the literature
[22,23,44].
To further examine the oxidation states of silver in leached
Ag–alumina catalysts, XPS was carried out with three leached
samples, AlAg(0.8,L), AlAg(1.1,L), and AlAg(2.6,L). For
comparison, XPS was also performed on an as-prepared Ag–
alumina sample, AlAg(4.9,CG). The XP spectra are shown in
Fig. 3, which also lists the binding energies of Ag 3d5/2 on
these samples. The binding energy of Ag 3d5/2 (at ∼368.0 eV)
does not shift much among the leached catalysts and the par-
ent catalyst. The binding energies of Ag0 and Ag2O are 368.3
and 367.5 eV, respectively [25,50]. It is hard to assign the BE
values shown in Fig. 3 to either metallic silver or oxidized sil-
ver, because a more complex electronic environment may be
considered a result of the co-gelation preparation (i.e., silver
(b)
Fig. 1. HRTEM of (a) AlAg(4.9,CG); (b) nitric acid–leached AlAg(0.8,L) after
◦
calcination in air at 650 C.
of Al2O3(L) was 205.5 m2/g, compared with 220.6 m2/g for
Al2O3(CG).
The surface composition of the leached catalysts was exam-
ined by XPS, and results are compared with the bulk composi-
tion obtained by ICP, as shown in Table 1. Clearly, the surface
concentration of Ag is very close to that in the bulk for the
leached samples, AlAg(0.8,L), AlAg(1.1,L), and AlAg(6,L).
Bulk analysis was done by ICP. These findings suggest that sil-
ver is well dispersed in Ag–alumina catalysts, prepared by the
coprecipitation-gelation method used in this work.
A HRTEM picture of a pair of parent and leached Ag–
alumina catalysts is shown in Fig. 1. Fig. 1a shows the parent
sample AlAg(4.9,CG). Silver particles <10 nm in size are ob-
served in good contact with the alumina matrix. After nitric
acid leaching, no particles remained on the leached sample,
AlAg(0.8,L), as shown in Fig. 1b.
XRD was performed to identify the crystalline phases
present in the Ag–alumina catalysts. Fig. 2 shows the results for
the alumina (both as-prepared and nitric acid-treated), leached
AlAg(0.8,L), and AlAg(7.7,L) samples, and one as-prepared
high-Ag-content sample AlAg(11,CG). For comparison, XRD

84
X. She, M. Flytzani-Stephanopoulos / Journal of Catalysis 237 (2006) 79–93
(a)
Fig. 3. XPS of leached and as prepared Ag–alumina catalysts; binding en-
ergy of Ag 3d : AlAg(0.8,L) 368.1 eV; AlAg(1.1,L) 368.0 eV; AlAg(2.6,L)
5/2
367.9 eV; AlAg(4.9,CG) 368.0 eV.
(b)
Fig. 5. Selective catalytic reduction of NO with CH over alumina and
4
Fig. 4. UV–vis DRS of as prepared alumina and Ag–alumina catalysts; at RT in
air.
Ag–alumina catalysts. (a) Conversion of NO to N ; (b) conversion of CH to
2
4
−1
CO . Feed gas: 0.25% NO–2% CH –5% O /He; SV = 9000 h
.
x
4
2
aluminate species). Seker et al. [46] reported two main peaks
of Ag 3d5/2 at 370.6 and 372.0 eV (C 1s at 284 eV) for their
Ag–alumina samples prepared by a sol–gel technique. The type
of silver species was not identified in that work either, due to
the relatively high BE values, which indicated strong electronic
interaction with the alumina.
δ+
hence the band due to Agn
grows stronger, but not the
band due to Ag0. It is also noteworthy that even for the high
silver-content sample, AlAg(21.5,CG), the color is surprisingly
light and still off-white. Hence, in the Ag–alumina catalysts
as prepared in this work, silver is highly dispersed and ex-
ists mainly as isolated silver ions, possibly embedded in the
alumina, [Ag–O–Al], and oxidized silver nanoparticles. Fur-
thermore, the stability of the [Ag–O–Al] species in the SCR
reaction conditions is excellent, as can be seen by the spectra of
the used AlAg(0.8,L).
We further explored the oxidation state of silver in our sam-
ples using UV–vis DRS. As shown in Fig. 4, UV–vis analysis
demonstrates the existence of different oxidation states of sil-
ver. All of the spectra are shown just as obtained. AlAg(0.8,L)
had been used under reaction conditions before the analysis,
whereas the other samples were as-prepared. Compared with
alumina, a significant new peak appears at 220 nm for both
the leached AlAg(0.8,L) and parent AlAg(4.9,CG), which is
assigned to Ag+ [23,41]. For the latter sample, there is also
a small peak at 350 nm, assigned to oxidized silver clusters
(Agnδ+), and this feature is still apparent after subtraction of
the alumina baseline. With higher silver amounts, such as in
AlAg(21.5,CG), besides Ag+ (220 nm), a broad peak appears
at 290–350 nm (Agnδ+), as well as a weak peak at ∼420 nm
(Ag0) [23,41]. This can be explained by the presence of sil-
ver nanoparticles and clusters covered by oxygen (in air), and
3.2. SCR of NO with CH4 over Ag–alumina
Fig. 5 shows the conversion of NO to N2 and CH4 to COx
over the leached Ag–alumina catalysts at 300–700 ◦C. The fig-
ure also shows the results from tests with silver-free Al2O3 and
one parent catalyst, AlAg(4.9,CG). Over Al2O3(CG), NO re-
duction begins at 450 ◦C and reaches a maximum conversion to
N2 of 70% at 650 ◦C. This alumina is a poor CH4-SCR cat-
alyst. High surface area alumina has been reported to be an
active catalyst for the SCR of NO with propene/propane or oxy-
genates [48,51]. Burch et al. [48] reported a peak NO to N2

X. She, M. Flytzani-Stephanopoulos / Journal of Catalysis 237 (2006) 79–93
85
conversion of ∼80% at 550 ◦C for C3H8-SCR of NO over γ -
alumina at C:N ratio of 6 and space velocity of 18,000 h−1
.
Meunier et al. [51] found that the maximum NO conversion
(60%) was reached at 600 ◦C in propene SCR, and that adding
a small amount of silver (1.2 wt%) caused no significant change
in activity, even though it shifted the active temperature window
to slightly lower values. Kung et al. [23] also pointed out the
important role of alumina sites for SCR of NO in a proposed
bifunctional mechanism, with metal ions as sites for NO2 or
adsorbed nitrate formation and alumina providing sites for dini-
trogen formation.
With the addition of 0.8 wt% Ag in alumina, as in the sam-
ple AlAg(0.8,L), the conversion maximum of NO to N2 shifts
to 600 ◦C and reaches 95%. AlAg(0.8,L) and AlAg(1.1,L) show
very similar conversions of NO and CH4. When the silver con-
tent increases, such as in AlAg(2.6,L), a higher conversion of
NO and a wider active temperature window are observed, with
NO to N2 conversion reaching ∼100% at 550–650 ◦C. The con-
version of NO to N2 is similar over the pair of leached and
parent catalysts, AlAg(0.8,L) and AlAg(4.9,CG), as shown in
Fig. 5a. However, the parent catalyst shows much higher con-
version of CH4 (Fig. 5b), i.e. 78% compared to 36% for the
leached material at 550 ◦C. The lower conversion of CH4 is at-
tributed to the absence of silver particles on AlAg(0.8,L), which
is consistent with literature reports of metallic silver nanoparti-
cles catalyzing the direct CH4 combustion reaction [22–26,29,
41,42]. Hence, leaching enhances the catalyst selectivity toward
CH4 oxidation by NOx.
Fig. 6. NO–O -TPD in He from AlAg(0.8,L).
2
Fig. 6 shows the TPD profiles from the leached AlAg(0.8,L)
sample. Desorption of O2 starts at around 300 ◦C, peaks at
497 ◦C, and levels off at ∼620 ◦C. No production of N2O or
N2 is found over the whole temperature range. Two desorption
peaks of NO are observed, with the first beginning at 170 ◦C
and peaking at 300 ◦C, followed by a second peak with a maxi-
mum at 539 ◦C. A NO2 peak is also observed at 539 ◦C. There-
fore, the NO peak at 539 ◦C is accompanied by NO2 and O2
desorption. Kameoka et al. [54] reported similar TPD features
over a 2 wt% Ag/alumina. In their case, two NO desorption
peaks were observed, one at low temperature (250 ◦C) and the
other at high temperature (480 ◦C). The latter peak was accom-
panied by O2 and NO2 desorption. Based on their FTIR results
(NO–O2 adsorption and then heating in He), these authors as-
signed three types of surface nitrates on Ag/alumina with ther-
mal stability in the order monodentate > bidentate > bridging-
nitrates (NO3−). Hence, the desorption peaks at 539 ◦C in Fig. 6
may be assigned to decomposition of nitrate species, whereas
the NO desorbed at low temperature (<300 ◦C) must be due
to weakly bound NO, because neither O2 nor NO2 is observed
accompanying this peak.
To elucidate the role of silver sites and alumina sites in
adsorption of NO/O2, similar TPD tests were also conducted
over bare alumina (both as-prepared and acid-treated) and the
silver-containing alumina samples. The desorption patterns are
shown in Figs. 7a–7c for O2, NO, and NO2, respectively. As
can be seen from these figures, the TPD profiles from alumina
and AlAg(4.9,CG) are in good correspondence with those over
AlAg(0.8,L), as described above for Fig. 6. Briefly, a similar
main desorption peak for both NO and NO2 is seen at 539 ◦C in
all three samples. The results indicate that silver does not mod-
ify the alumina surface, and that the suface nitrate species are
bound on the alumina sites rather than on silver.
3.3. Surface adsorption of NO/O2 and reactivity with CH4
3.3.1. NO–O2-TPD
The surface interaction of NO and O2 with alumina or
Ag/alumina catalysts has been reported to play an important
role in the mechanism of SCR of NO by hydrocarbons through
the formation of adsorbed species, such as nitrites or nitrates
[27,29,48,51–55]. Formation of surface nitrates has been found
by FTIR spectra over both alumina [51–53] and Ag/alumina
[27,29,51,54,55] catalysts after adsorption of NO and O2. TPD
is also useful because it probes different surface structures by
monitoring the species desorbed over the range of temperature
of interest to catalysis [48,54]. Therefore, in this work NO–O2-
TPD was performed over alumina and leached and as-prepared
Ag–alumina samples to identify the types of surface adsorption
sites of different structures and elucidate their relevance for the
catalysis of the NO reduction by CH4.
3.3.1.1. NO–O2-TPD in He NO–O2-TPD results are shown
in Figs. 6 and 7 for desorption in flowing He gas. Analysis of
the desorbed species was done by mass spectrometry. Signals
due to O2 (m/e = 32), NO (m/e = 30), and NO2 (m/e = 46)
were detected. No N2O (m/e = 44) or N2 (m/e = 28) was ob-
served, in agreement with previous reports for NO–O2-TPD
over alumina [48] and Ag–alumina [54]. For most of the data
reported here, replicate tests were performed. The TPD-mass
spectrometry technique is very stable, and reproducible results
were obtained.
To probe the effect of adding more silver onto the alumina,
a sample with higher silver content (11 wt%), AlAg(11,CG),
was also examined; the results are shown in Fig. 7. For this
sample, NO begins to desorb from 170 ◦C and continues up
to 400 ◦C, while the second peak is still present, with a max-
imum slightly shifted to 529 ◦C (vs. 539 ◦C). Thus, the TPD
of NO is similar to that of the other samples, except that an en-

86
X. She, M. Flytzani-Stephanopoulos / Journal of Catalysis 237 (2006) 79–93
(a)
(a)
(b)
(c)
(b)
Fig. 8. NO–O -TPD in 5% O /He: (a) NO; (b) NO .
2
2
2
loading (e.g., for the AlAg(11,CG) sample), the percentage of
NO desorbed at low temperature increases to 50%, indicating
a greater contribution of the Ag nanoparticles. Quantitatively,
the total amount of desorbed NO over the whole temperature
range was close among all these catalysts, 0.9–1.1 µmol/m2.
Thus, no significant effect of silver loading is observed. Ear-
lier, Kameoka et al. [54] found no major NO adsorption dif-
ference between a 2 wt% Ag/Al2O3 and Al2O3 by TPD and
DRIFTS measurements, but did not examine the potential effect
of higher silver loadings. Meunier et al. [51] proposed that in-
creasing the loading of silver (2–10 wt%) on alumina promotes
NO/O2 adsorption based on in situ DRIFTS and thermogravi-
metric analyses.
Fig. 7. NO–O -TPD in He from alumina and silver–alumina catalysts: (a) O ;
2
2
(b) NO; (c) NO .
2
3.3.1.2. NO–O2-TPD in O2 Although TPD is generally run in
inert gas streams, such as helium, it is also important to consider
the effect of a reactive sweep gas. Thus, in this work TPD was
also performed in a 5% O2/He gas mixture to simulate the lean-
burn conditions in practical SCR systems. The results, shown
in Fig. 8, indicate that, similar to TPD in He, no N2 or N2O was
identified in the desorbed gases. However, in contrast to TPD in
He, no O2 desorption occurred. Figs. 8a and 8b show the TPD
profiles of NO and NO2 from alumina, leached AlAg(0.8,L)
and its parent AlAg(4.9,CG), another leached sample with
hanced NO adsorption is shown over the low-temperature range
over AlAg(11,CG). This is attributed to NO adsorbed on the
silver nanoparticles, which are present in this sample but ab-
sent from alumina or AlAg(0.8,L). Using peak deconvolution
(PeakFit V4.12; Seasolve Software), we found that the low-
temperature peak comprises 21.9–30% of the total desorbed
NO from Al2O3(L or CG) and AlAg(0.8,L), whereas it is 37.6%
of the NO desorbed from AlAg (4.9,CG). With even higher Ag

X. She, M. Flytzani-Stephanopoulos / Journal of Catalysis 237 (2006) 79–93
87
Table 2
mina and silver–alumina samples after they were exposed to co-
adsorption of NO/O2. The TPSR results are compared among
catalysts with different structures (i.e., blank alumina, silver
particle—free leached Ag–alumina, and silver particle—laden
Ag–alumina with high silver content).
a
a,b
N , CO , and NO eluted during TPSR and TPD
2
2
Sample
Amount of
produced
Amount of
NO (TPSR) NO (TPD)
2
2
N
CO produced (µmol/m
)
(µmol/m
)
2
2
cat
cat
2
2
(µmol/m
)
(µmol/m
)
cat
cat
Al O
2
–
–
–
53.6
152.5
819
1.0
0.8
1.2
1.0
1.2
1.3
1.5
1.7
3
Figs. 9a–9e presents the TPSR results over Al2O3(CG),
AlAg(0.8,L), AlAg(4.9,CG), and AlAg(21.5,CG). The spectra
were repeated twice for each sample, and reproducible TPSR
results were obtained. In Fig. 9a, the top part shows the pro-
files of O2 signal for the above four samples, and the bottom
part shows the CH4 profiles. As indicated by the dotted line
a, over AlAg(21.5,CG), consumption of O2 and CH4 first ap-
pears at 430 ◦C, followed by a much faster consumption af-
ter 460 ◦C. For AlAg(4.9,CG), these two temperatures shift to
higher values, as shown by the dotted line b at 510 and 578 ◦C,
respectively. Similar profiles are observed over AlAg(4.9,CG)
and AlAg(0.8,L), although the consumption of O2 and CH4 is
weaker for the latter, which could be due to the absence of sil-
ver particles. Alumina exhibits an even weaker consumption of
the two reactants.
Accompanying the consumption of O2 and CH4, m/e sig-
nals of 28 for N2/CO and 44 for CO2 were produced, as well as
desorption of NO/NO2, as shown in Figs. 9b and 9c. It should
be pointed out that no N2O was produced over these samples,
as verified by FTIR gas analysis. Hence the mass signal of 44
was assigned to CO2 only. The m/e = 28 signal is more diffi-
cult, however, because it can be due to both N2 and CO. Two
sources of CO are considered: (1) CO from the cracking of CO2
in the mass spectrometer (at 12% of total CO2 [57]) and (2) CO
produced during TPSR. For the first effect, the raw data were
corrected by subtracting this part CO from signal m/e = 28
in Fig. 9b. To check for the second effect, CO was monitored
by on-line FTIR; the results are presented in Fig. 9d. As this
figure shows, CO is produced from Al2O3(CG) and leached
AlAg(0.8,L) in the temperature range 345–700 ◦C, with the
amount of CO increasing with increasing temperature, reach-
ing 333 ppm for Al2O3(CG) and 300 ppm for AlAg(0.8,L) at
700 ◦C. Thus, partial oxidation of CH4 occurs on these two cat-
alysts. Over AlAg(4.9,CG), however, much less CO is produced
compared with the former two catalysts. This means that the
presence of silver particles is beneficial to burning off the CO
produced. In view of Fig. 9d, in Fig. 9b, the m/e = 28 sig-
nal is denoted as N2/CO for Al2O3(CG), leached AlAg(0.8,L),
and AlAg(4.9,CG). For AlAg(21.5,CG), the m/e = 28 signal
is assigned to N2 only. Also note that no reduced nitrogen-
containing products, such as NH3 and HCN, accompany the
CO production during TPSR, as checked by IR.
AlAg(0.8,L)
AlAg(4.9,CG)
AlAg(21.5,CG) 0.34
2982.6
a
◦
In the temperature range 25–700 C; measured by mass spectrometry.
In O /He gas mixture.
b
2
6 wt% Ag, AlAg(6,L), and two Ag–alumina samples with rel-
atively high Ag loadings, AlAg(11,CG) and AlAg(21.5,CG).
For alumina, NO desorption starts at 200 ◦C, peaks at 460 ◦C,
and then levels off at 600 ◦C. A similar trend was observed
for AlAg(0.8,L), with the desorption peak also positioned at
460 ◦C. For the parent catalyst AlAg(4.9,CG), desorption of
NO also begins at 200 ◦C and levels off at 600 ◦C, even though
the peak is slightly shifted to 450 ◦C and the peak intensity
is slightly increased. The amount of NO desorbed from the
above three samples was 1.2–1.5 µmol/m2, as shown in Table 2.
A similar desorption pattern is also observed for AlAg(6,L)
with a major peak at 460 ◦C and desorbed NO of 1.6 µmol/m2.
Comparing these catalysts, we assign the desorption feature
with a peak at ∼460 ◦C to NOx species adsorbed on alumina
sites. As before, these TPD data also show that the alumina-
bound ionic silver species [Ag–O–Al] do not modify the ad-
sorption properties of alumina surface.
As shown in Fig. 8, the high-silver-content samples, AlAg
(11,CG) and AlAg(21.5,CG), have exaggerated low-temperature
desorption. Peak deconvolution performed for all of the TPD
profiles of Fig. 8a shows a small NO desorption peak with
maximum at 320–350 ◦C accounting for 11–18% of the to-
tal desorbed NO from the alumina, leached Ag–alumina, and
AlAg(4.9,CG) samples. However, this value increases to 30–
45% of the total NO desorbed for the samples AlAg(11,CG)
and AlAg(21.5,CG), which corresponds very well with the data
of Fig. 7. Relatively high total NO desorption (1.7 µmol/m2)
is reported in Table 2 for AlAg(21.5,CG), attributed to the ad-
ditional low-temperature adsorption of NOx on the oxidized
surface of silver nanoparticles.
3.3.2. TPSR tests
In previous work, the reactivity of surface species formed
after adsorption of NO/O2 with various hydrocarbons or oxy-
genates, such as propene/propane [27,48,51,53–56], higher hy-
drocarbons (C4–C8) [29,56], and ethanol/methanol [48,54], has
been examined over alumina [48,51,53] and silver/alumina [27,
29,51,54–56] catalysts by FTIR or TPR measurements. Con-
sistently, different research groups have reported high reac-
tivity of adsorbed NOx species to these reductants, confirm-
ing these species importance as intermediates for the SCR of
NO. However, to our knowledge, no such information is avail-
able for the reactivity of adsorbed NOx species with CH4 over
alumina or Ag/alumina catalysts. Consequently, in this work
TPSR tests were performed by flowing CH4/O2 over the alu-
Similar to the NO–O2-TPD results (Fig. 8), NO starts to de-
sorb at ∼200 ◦C during TPSR (Fig. 9b). Along with the NO
desorption, NO2 desorption was also observed; the NO2 lev-
els produced were measured separately by FTIR, as shown in
Fig. 9e. The peak positions of desorbed NO2 by both mass
spectrometry and FTIR are consistent between Figs. 9b and
9e. Although the NO2 level decreases with temperature above
460 ◦C, Fig. 9e shows that NO2 (500 ppm) is present in the
gas even at 700 ◦C. Over AlAg(21.5,CG), the m/e = 28 signal
starts to increase from 300 to 400 ◦C as a small peak, then in-

88
X. She, M. Flytzani-Stephanopoulos / Journal of Catalysis 237 (2006) 79–93
(a)
(b)
(c)
(d)
(e)
Fig. 9. TPSR in CH /O , over alumina and Ag–alumina catalysts pre-exposed to NO–O : (a) O and CH ; (b) NO, NO , N /CO and CO(IR); (c) CO ; (d) CO and
4
2
2
2
4
2
2
2
(e) NO profile measured by FTIR. In (b), the NO signal was scaled up 2× for alumina, AlAg(0.8,L) and AlAg(4.9,CG) due to the weak signal. The inset in (c)
2
2
highlights the CO signal at low temperatures.
2
creases sharply and peaks at 430 ◦C and then decreases slowly;
it is still measurable at 700 ◦C. For the m/e = 28 signal over
AlAg(4.9,CG), there is also a broad peak at 300–450 ◦C, and
a peak at 510 ◦C. The profiles for m/e = 28 are similar for
Al2O3(CG) and AlAg(0.8,L). For AlAg(0.8,L), the m/e = 28
signal begins to increase at 320 ◦C, peaks at 523 ◦C and then
reaches a valley at 570 ◦C, however, above 570 ◦C, the sig-
nal keeps going up with temperature. A similar two-stage pat-
tern is also found for the Al2O3(CG) sample, with the peak at
470 ◦C and the valley at 540 ◦C. For the latter three samples, the

X. She, M. Flytzani-Stephanopoulos / Journal of Catalysis 237 (2006) 79–93
89
m/e = 28 signal may comprise both N2 and CO, as revealed by
independent FTIR gas analysis for CO in Fig. 9d. Accordingly,
the mass signals of CO corresponding to Fig. 9d are also shown
in Fig. 9b, denoted as CO(IR), over AlAg(4.9,CG), Al2O3(CG),
and AlAg(0.8,L). The net N2 production can be obtained by
subtracting the CO(IR) from the m/e = 28 signal; this subtrac-
tion has no significant effect on the m/e = 28 profile. Hence,
N2 production continues at high temperatures (up to 700 ◦C),
although the extent is strongly affected by the methane combus-
tion activity of each catalyst and the corresponding availability
of methane. (As shown in Fig. 9a, very little methane remains
over the AlAg(21.5,CG) material at 700 ◦C.) On all samples,
temperatures above ∼300 ◦C are necessary to initiate the CH4
activation via adsorbed NOx species for N2 production.
TPSR profiles of CO2 are shown in Fig. 9c. The light-off
temperature is 300 ◦C for CO2 production over all four sam-
ples, the same as the light-off temperature for N2/CO shown in
Fig. 9b. Apparently, much stronger CO2 production occurs over
AlAg(21.5,CG) than over the other samples with lower silver
loading or without silver. Comparing the leached AlAg(0.8,L)
and parent AlAg(4.9,CG), it is clear that removing silver parti-
cles suppresses the direct CH4 oxidation to CO2, whereas the
CO2 production over AlAg(0.8,L) is slightly higher than that
over Al2O3(CG).
Furthermore, the TPSR profiles of m/e = 28 and 44 (CO2)
for each catalyst are in total correspondence with the profiles
of CH4 and O2. For example, a peak and a valley are seen
for N2/CO and CO2 in Figs. 9b and 9c at exactly the same
temperatures as for CH4 and O2 in Fig. 9a, 430 and 460 ◦C
for AlAg(21.5,CG) and 510 and 578 ◦C for AlAg(4.9,CG), re-
spectively, as represented by the dotted lines in Figs. 9b and
9c. Shi et al. [58] observed a very similar two-stage profile of
CH4/O2 consumption and accompanying N2/CO2 production
during TPSR in a CH4–O2 stream after coadsorption of NO/O2
over an Ag-ZSM-5 catalyst. The light-off temperature for CH4
consumption and N2/CO2 production was reported to be around
350 ◦C, and Shi et al. proposed that CH4 reacted with surface
nitrates by SCR before reaction with oxygen by direct combus-
tion.
taining the amount of N2 from the m/e = 28 signal is dif-
ficult, and so the numbers are not shown. The desorbed NO
was also calculated and compared with the NO desorbed dur-
ing TPD in O2 (Fig. 8). The difference between these two NO
values should indicate the part of NO consumed during TPSR;
for example, over AlAg(21.5,CG), the difference in NO is
0.7 µmol/m2, corresponding to 0.35 µmol/m2 of N2 produced
(based on stoichiometry), very close to the measured amount of
N2 (0.34 µmol/m2) produced during TPSR. The trend of CO2
values show that the presence of silver particles enhances CO2
production by the direct methane oxidation reaction.
3.3.2.1. Transient methane combustion tests Separate tests
with CH4 and O2 were conducted to examine the surface activa-
tion of CH4 without first adsorbing NO–O2 (nitrates). The CO2
profiles produced are shown in Fig. 10. To check for the possi-
bility of gas-phase reaction, a blank test was run with an empty
reactor (as shown in Fig. 10) in which only a trace amount of
CO2 (44) was detected above 600 ◦C. Therefore, CH4 is not
thermally activated at temperature below 600 ◦C in the pres-
ence of O2.
Fig. 10 also shows the CO2 profiles obtained over alumina
and three leached samples, AlAg(0.8,L), AlAg(2.6,L), and
AlAg(6,L). CO2 production starts at ∼500 ◦C over alumina and
AlAg(0.8,L) and shifts slightly to ∼480 ◦C for AlAg(2.6,L)
and AlAg(6,L). The presence of silver particles promotes the
light-off for CO2 at 400 ◦C in AlAg(4.9,CG) and 300 ◦C in
Ag(21.5,CG). Kundakovic et al. [42] reported light-off values
of 300, 350, and 500 ◦C for CH4 oxidation over Ag particles
(7.9 wt% Ag/ZrO2), Ag clusters (Ag(83)-ZSM-5) and isolated
silver ions (Ag(16)-ZSM-5), respectively, in steady-state tests.
Table 3 gives the amounts of CO2 produced for alumina and
Ag–alumina catalysts. Comparing AlAg(0.8,L) and alumina in-
dicates that the presence of Ag+ enhances CO2 production.
Further, the amount of CO2 produced also increases with an
increasing amount of silver ions in the leached catalysts, follow-
ing the order AlAg(6,L) > AlAg(2.6,L) > AlAg(0.8,L). Much
more CO2 was produced for the two as-prepared catalysts,
AlAg(4.9,CG) and AlAg(21.5,CG), apparently due to the pres-
ence of silver nanoparticles.
Combining the transient reaction results (Fig. 10) and TPSR
(Fig. 9) demonstrates that adsorbed nitrates shift the CH4 acti-
vation temperature to 300 ◦C during TPSR, especially for alu-
mina and leached Ag–alumina catalysts. We postulate that for
all catalysts, adsorbed surface nitrates are essential for CH4
activation at temperatures as low as 300 ◦C with concomitant
production of N2. In contrast, for Ag–alumina catalysts covered
with silver nanoparticles, methane can be activated directly by
adsorbed oxygen, so that methane combustion begins at 300 ◦C.
However, preadsorbed NOx on the silver particles does not ap-
pear to contribute to N2 formation at this temperature. As will
become clear in what follows, all N2 produced in the CH4-SCR
of NO comes from the reaction of NOx adsorbed on the alumina
sites only.
The TPSR results of Fig. 9 also reflect the structural effect
of silver in Ag–alumina catalysts. From the CO2 production
shown in Fig. 9c, the CH4 combustion-dominated range shifted
by ∼120 ◦C to lower temperature over AlAg(21.5,CG) com-
pared with the other samples. The high combustion activity of
this sample is attributed to the presence of oxygen-covered sil-
ver particles [42] as shown on UV–vis (Fig. 4). Further, compar-
ing the leached AlAg(0.8,L) and parent AlAg(4.9,CG) samples,
with similar TPSR, the similarity can be attributed to the exis-
tence of [Ag–O–Al] species dominant in both samples, as also
shown by UV–vis (Fig. 4). However, Fig. 9c shows much lower
CO2 on AlAg(0.8,L) than on AlAg(4.9,CG), due to the absence
of silver particles on the leached catalyst, verified by TEM in
Fig. 1b.
Table 2 compares the amounts of N2 and CO2 eluted in
TPSR. For all of the data presented in this table, the val-
ues were obtained by integrating from room temperature to
700 ◦C. For the samples that produce CO during TPSR, ob-
Table 3 also gives the amounts of CO produced for the above
samples during transient CH4–O2 reaction. For alumina and
all three leached samples, CO accompanies CO2 production.

90
X. She, M. Flytzani-Stephanopoulos / Journal of Catalysis 237 (2006) 79–93
◦
Fig. 10. CO produced during transient CH –O reaction over alumina and Ag–alumina samples from RT to 700 C. The inset highlights the CO signal over
2
4
2
2
AlAg(21.5,CG) (line g).
Table 3
a
Gaseous products of the CH –O reaction
4
2
Sample
Amount of CO
produced (µmol/m
Amount of CO
produced (µmol/m
2
2
2
)
)
cat
cat
Al O
37.9
49.7
149
196
363
48.3
31.4
51.1
10.2
0
2
3
AlAg(0.8,L)
AlAg(2.6,L)
AlAg(6,L)
AlAg(4.9,CG)
AlAg(21.5,CG)
1840
0
a
Measured by mass spectrometry during transient tests at the conditions of
Fig. 10.
However, for the two as-prepared catalysts, AlAg(4.9,CG) and
AlAg(21.5,CG), no CO production is observed over the whole
temperature range, again attributed to the presence of silver par-
ticles in these samples.
(a)
3.3.3. Kinetics of CH4-SCR
Kinetics measurements were made over alumina (CG), two
pairs of leached and parent Ag–alumina catalysts [AlAg(0.8,L)
vs. AlAg(4.9,CG) and AlAg(6,L) vs. AlAg(21.5,CG)], and
one leached sample from an 850 ◦C-calcined parent, AlAg(7.5,
L-850P). Rates of NO to N2 and overall rates of CH4 to COx
were typically measured at 425–525 ◦C. The gas composition
used was of C:N = 1 (0.5% NO–0.5% CH4–5% O2). Some
data points were measured at temperatures of 550–600 ◦C and
gas composition of C:N = 5 (0.2% NO–1% CH4–7.5% O2) for
AlAg(0.8,L) and AlAg(6,L). Blank tests were also run with the
empty reactor at the above temperature range and gas composi-
tion conditions. No reaction occurred in the gas phase alone at
the above conditions.
(b)
Fig. 11. Rates of CH -SCR of NO over alumina and Ag–alumina (leached
Arrhenius-type plots of the rates of CH4-SCR of NO are il-
lustrated in Fig. 11. Fig. 11a shows the rates of reduction of NO
to N2, and Fig. 11b shows the overall rates of CH4 oxidation to
x
4
or as prepared); (a) rate of NO to N ; (b) rate of the overall CH oxidation
2
4
reaction.

X. She, M. Flytzani-Stephanopoulos / Journal of Catalysis 237 (2006) 79–93
91
Table 4
Apparent activation energies of CH -SCR of NO over various catalysts
x
4
Sample
E -NO
(kJ/mol)
E -CH
(kJ/mol)
Reductant
Reference
a
a
4
a
Al O (CG)
AlAg(0.8,L)
123.8
94.7
106.4
113.0
97.2
119.7
98
111.4
118.8
93.9
128.0
63.2
105.5
87
102
–
–
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
This work
This work
This work
This work
This work
This work
[59]
[5]
[60]
[61]
[62]
2
3
4
4
4
4
4
4
4
4
4
4
4
AlAg(4.9,CG)
AlAg(6,L)
AlAg(21.5,CG)
AlAg(7.5,L-850P)
Ce–Ag-ZSM-5
Co-ZSM-5
b
91
La O , Sr/La O
111.8
103.2
147
2
3
2 3
Sr/La O /Al O
2
3
2
3
Li/MgO
–
Fig. 12. TOF of NO to N over leached Ag–alumina catalysts.
2
a
Apparent activation energy of the overall methane oxidation reaction.
b
◦
This sample is leached from an 850 C-calcined parent; after leaching, it
◦
COx. Blank alumina exhibits the lowest rates of NO to N2 and
overall CH4 oxidation. Doping with silver enhances the above
rates, as shown for the rates over AlAg(0.8,L), AlAg(4.9,CG),
and AlAg(7.5,L-850P). Interestingly, the rates for these sam-
ples are very close to one another. The similar rates of NO to
N2 in the leached sample, AlAg(0.8,L), and the parent sample,
AlAg(4.9,CG), indicate that oxidized silver species are the sites
responsible for the SCR, and that silver particles do not partici-
pate in CH4-SCR of NO. AlAg(4.9,CG) displays higher overall
rates of CH4 oxidation than AlAg(0.8,L), which is consistent
with the literature [42] that silver nanoparticles contribute to
methane combustion and also with the CO2 elution profiles dur-
ing TPSR in Fig. 9c.
was calcined again at 650 C for 3 h.
For HC-SCR of NO, the prevailing mechanisms found in the
literature involve adsorbed nitrates and partially oxidized hy-
drocarbons [63], for which extensive in situ DRIFTS evidence
exists. For Ag/alumina, nitrates and acetates are generally ac-
cepted as reaction intermediates in SCR of NO with reductants
other than CH4 [29,51,53,56]. However, the pathway for dini-
trogen formation is still unclear, with species such as R–NOx,
isocyanate, cyanide, amines, and NH3 proposed as possibili-
ties [51]. For a reductant such as methane, an important mech-
anistic aspect is activation via abstraction of hydrogen from the
C–H bond, which has been proposed to be the rate-determining
step during CH4-SCR of NO [64]. For Co-ZSM-5, adsorbed
NO2 on Co sites is reported to activate CH4 [64], whereas a Pd
ion can break the C–H bond directly, as has been proposed for
Pd/H-ZSM-5 [65]. Direct activation of methane on Ag+ ions in
Ag-ZSM-5 has been proposed by Shi et al. [47].
For the Ag–alumina system, our results show that the
SCR-active NOx species are adsorbed on alumina sites. How
methane activation occurs is also indicated by the data of
Figs. 9–12 taken collectively. First, methane is activated by
adsorbed oxygen on silver nanoparticles at 300 ◦C, but this
pathway leads to CO2 only. No N2 is produced from the nitrate
species adsorbed on silver nanoparticles (see Figs. 8 and 9). The
rate of N2 production is the same for the parent and the leached
catalysts, AlAg(4.9,CG) and AlAg(0.8,L), and AlAg(21.5,CG)
and AlAg(6,L). But because the amount of strongly bound Ag+
is increased in the leached samples (e.g., from 0.8 to 6 wt%),
the rate increases proportionately (see Fig. 12). We can sur-
mise from this that the additional [Ag–O–Al] sites serve to
activate more methane; thus, the reaction is limited by the acti-
vation of methane on Ag–alumina catalysts. Production of N2
begins at ∼300 ◦C from the interaction of adsorbed NOx on
alumina and CHx species on Ag–O–Al groups (see Fig. 9).
This scheme explains all of the data presented in this work.
Fig. 13 shows a schematic of the mechanistic pathway. Al-
though this pathway is plausible, future work should involve
spectroscopic verification of the proposed reaction mecha-
nism.
In Fig. 11a, for AlAg(0.8,L), the rates measured at C:N = 5
are comparable with the rates measured at the C:N = 1 condi-
tion after extension to lower temperatures. Hence, the C:N ratio
does not affect the SCR reaction pathway.
For samples with higher silver amounts, such as AlAg(21.5,
CG), significantly enhanced rates of NO to N2 are observed.
The rates over the corresponding leached sample AlAg(6,L)
were similarly enhanced, and the activation energies were the
same. For the leached AlAg(0.8,L) and AlAg(6,L), using the
amount of silver present on the surface of these two catalysts
to scale the rate, we can produce a TOF plot. The TOF val-
ues for NO to N2 were calculated from the rates in Fig. 11a
and the number of surface silver sites based on the XPS results
(Table 1). As shown in Fig. 12, the TOF plots are in excellent
agreement.
Table 4 gives the apparent activation energies, Ea, obtained
from Fig. 11. For NO reduction to N2, the Ea values for the
Ag–alumina catalysts are in the range of 94.7–119.7 kJ/mol,
whereas for Al2O3(CG), a slightly higher value of 123.8 kJ/mol
is calculated. The Ea values for overall methane oxidation are
in the range of 93.9–128 kJ/mol for alumina and low-silver
content catalysts and 63.2 kJ/mol for AlAg(21.5,CG). Table 4
also lists the activation energies for CH4-SCR of NO reported
for other catalysts, including Co-ZSM-5, Sr/La2O3, and others.
Interestingly, the Ea values for Ag–alumina catalysts reported
here are very close to the activation energies reported for other
very different CH4-SCR catalysts. Whether this is indicative of
a common methane activation step over all these catalysts re-
mains unclear.

92
X. She, M. Flytzani-Stephanopoulos / Journal of Catalysis 237 (2006) 79–93
References
[1] H. Hamada, Y. Kintaichi, M. Sasaki, T. Ito, Appl. Catal. 64 (1990) L1.
[2] Y. Li, J.N. Armor, Appl. Catal. B 1 (1992) L31.
[3] Y. Li, J.N. Armor, Appl. Catal. B 2 (1993) 239.
[4] Y. Li, J.N. Armor, Appl. Catal. B 3 (1993) L1.
[5] Y. Li, P.J. Battavio, J.N. Armor, J. Catal. 142 (1993) 561.
[6] Y. Li, J.N. Armor, J. Catal. 150 (1994) 376.
[7] Y. Li, J.N. Armor, Appl. Catal. B 5 (1995) L257.
[8] P. Budi, R.F. Howe, Catal. Today 38 (1997) 175.
[9] E. Kikuchi, K. Yogo, Catal. Today 22 (1994) 73.
[10] K. Yogo, M. Ihara, I. Terasaki, E. Kikuchi, Chem. Lett. (1993) 229.
[11] K. Yogo, M. Ihara, I. Terasaki, E. Kikuchi, Appl. Catal. B 2 (1993) L1.
[12] B.J. Adelmana, W.M.H. Sachtler, Appl. Catal. B 14 (1997) 1.
[13] R. Burch, J.P. Breen, F.C. Meunier, Appl. Catal. B 39 (2002) 283.
[14] N. Li, A. Wang, L. Li, X. Wang, L. Ren, T. Zhang, Appl. Catal. B 50
(2004) 1.
[15] K. Okumura, T. Kusakabe, M. Niwa, Appl. Catal. B 41 (2003) 137.
[16] A. Bahamonde, S. Campuzano, M. Yates, P. Salerno, S. Mendioroz, Appl.
Catal. B 44 (2003) 333.
[17] N. Li, A. Wang, X. Wang, M. Zheng, R. Cheng, T. Zhang, Appl. Catal.
B 48 (2003) 259.
Fig. 13. Proposed reaction scheme for the selective catalytic reduction of NO
with CH over Ag–alumina.
4
4. Conclusion
In this work we investigated the role of silver nanoparticles,
silver ions embedded in alumina, and alumina itself in the SCR
of NO with methane under excess oxygen. Nitric acid leaching
was used to remove silver particles and other weakly bound sil-
ver from Ag–alumina materials prepared by a coprecipitation-
gelation method. For the leached catalysts, the surface silver
content was found by XPS to be similar to the bulk value. UV–
vis DRS identified silver cations and oxidized silver clusters
and nanoparticles in the parent Ag–alumina catalysts. However,
only ionic silver, bound with alumina, [Ag–O–Al], remained in
the leached samples. The amount of ionic silver increases with
increasing silver loading; that is, parent materials with high sil-
ver content have both more silver nanoparticles and more bound
Ag–O–Al species. The SCR reaction rate scales with the num-
ber of [Ag–O–Al] sites.
NO–O2-TPD tests revealed that the leached samples and
bare alumina have the same NOx adsorption features. Thus,
bound silver ions cause no modification of the alumina surface
sites. The presence of silver nanoparticles enhances the low-
temperature NOx adsorption peak. This species does not react
with methane, which follows the direct oxidation route with ad-
sorbed oxygen on the silver surface. N2 production on alumina
begins at 300 ◦C, as determined by TPSR. In contrast, transient
reaction tests in CH4–O2 show that the light-off temperature for
CH4 combustion is above 470 ◦C for alumina, AlAg(0.8,L), and
AlAg(6,L).
CH4 combustion is much more facile on silver nanoparti-
cles, with a light-off temperature of ∼300 ◦C. This explains the
lower selectivity of CH4-SCR of NO over silver particle—laden
Ag–alumina catalysts. Kinetic measurements demonstrated that
the SCR reaction rates and activation energies are the same for
the leached and parent Ag–alumina catalysts. The only advan-
tage of the silver nanoparticles is the burnoff of CO produced
over the Ag–O–Al structures.
On the basis of our findings, a plausible reaction pathway
for SCR of NO with methane over Ag–alumina catalysts has
emerged. Accordingly, CHx species on [Ag–O–Al] react with
NOx adsorbed on alumina to produce dinitrogen.
[18] N. Li, A. Wang, L.J. Tang, X. Wang, D. Liang, T. Zhang, Appl. Catal. B 43
(2003) 195.
[19] M. Kantcheva, A.S. Vakkasoglu, J. Catal. 223 (2004) 364.
[20] T. Miyadera, K. Yoshida, Chem. Lett. (1993) 1483.
[21] N. Aoyama, K. Yoshida, A. Abe, T. Miyadera, Catal. Lett. 43 (1997) 249.
[22] T.E. Hoost, R.J. Kudla, K.M. Collins, M.S. Chattha, Appl. Catal. B 13
(1997) 59.
[23] K.A. Bethke, H.H. Kung, J. Catal. 172 (1997) 93.
[24] T. Furusawa, K. Seshan, J.A. Lercher, L. Lefferts, K. Aika, Appl. Catal.
B 37 (2002) 205.
[25] M. Richter, M. Langpape, S. Kolf, G. Grubert, R. Eckelt, J. Radnik,
M. Schneider, M.M. Pohl, R. Fricke, Appl. Catal. B 36 (2002) 261.
[26] N. Bogdanchikova, F.C. Meunier, M. Avalos-Borja, J.P. Breen, A. Pes-
tryakov, Appl. Catal. B 36 (2002) 287.
[27] A. Martinez-Arias, M. Fernández-Garcia, A. Iglesias-Juez, J.A. Anderson,
J.C. Conesa, J. Soria, Appl. Catal. B 28 (2000) 29.
[28] K. Shimizu, A. Satsuma, T. Hattori, Appl. Catal. B 25 (2000) 239.
[29] K. Shimizu, J. Shibata, H. Yoshida, A. Satsuma, T. Hattori, Appl. Catal.
B 30 (2001) 151.
[30] L.-E. Lindfors, K. Eranen, F. Klingstedt, D.Yu. Murzin, Top. Catal. 28
(2004) 185.
[31] F. Klingstedt, K. Eranen, L.-E. Lindfors, K. Eranen, S. Andersson, L. Ci-
der, C. Landberg, E. Jobson, L. Eriksson, T. Ilkenhans, D. Webster, Top.
Catal. 30/31 (2004) 27.
[32] B. Wichterlova, Top. Catal. 28 (2004) 131.
[33] K. Arve, L. Capek, F. Klingstedt, K. Eranen, L.-E. Lindfors, D.Yu. Murzin,
J. Dedecek, Z. Sobalik, B. Wichterlova, Top. Catal. 30/31 (2004) 91.
[34] T. Miyadera, Appl. Catal. B 13 (1997) 157.
[35] T. Miyadera, Appl. Catal. B 16 (1998) 155.
[36] T. Chafik, S. Kameoka, Y. Ukisu, T. Miyadera, J. Mol. Catal. A 136 (1998)
203.
[37] S. Kameoka, T. Chafik, Y. Ukisu, T. Miyadera, Catal. Lett. 55 (1998) 211.
[38] S. Kameoka, T. Chafik, Y. Ukisu, T. Miyadera, Catal. Lett. 51 (1998) 11.
[39] S. Symiya, M. Saito, H. He, C. Qing, N. Takezawa, K. Yoshida, Catal.
Lett. 50 (1998) 87.
[40] A. Abe, N. Aoyama, S. Sumiya, N. Kakuta, K. Yoshida, Catal. Lett. (1998)
5.
[41] A. Keshavaraja, X. She, M. Flytzani-Stephanopoulos, Appl. Catal. B 27
(2000) L1.
[42] Lj. Kundakovic, M. Flytzani-Stephanopoulos, Appl. Catal. A 183 (1999)
35.
[43] Z. Wang, M. Yamaguchi, I. Goto, M. Kumagai, Phys. Chem. Chem.
Phys. 2 (13) (2000) 3007.
[44] T. Nakatsuji, R. Yasukawa, K. Tabata, M. Niwa, Appl. Catal. B 17 (1998)
333.
[45] K. Takagi, T. Kobayashi, H. Ohkita, T. Miztshima, N. Kakuta, A. Abe,
K. Yoshida, Catal. Today 45 (1998) 123.
Acknowledgments
Financial support of this work by a grant from the National
Science Foundation (NIRT grant 0304515) is gratefully ac-
knowledged. We thank Dr. Ed Neister for his assistance with
the FTIR experiments.

X. She, M. Flytzani-Stephanopoulos / Journal of Catalysis 237 (2006) 79–93
93
[46] E. Seker, J. Cavataio, E. Gulari, P. Lorpongpaiboon, S. Osuwan, Appl.
Catal. A 183 (1999) 121.
[55] F.C. Meunier, V. Zuzaniuk, J.P. Breen, M. Olsson, J.R.H. Ross, Catal. To-
day 59 (2000) 287.
[47] C. Shi, M. Cheng, Z. Qu, X. Bao, Appl. Catal. B 51 (2004) 171.
[48] R. Burch, E. Halpin, J.A. Sullivan, Appl. Catal. B 17 (1998) 115.
[49] A commercial software JADE was used for phase identification and lattice
parameters calculation, which was coupled to ICDD Database.
[50] G.B. Hoflund, Z.F. Hazos, Phys. Rev. B 62 (2000) 11126.
[51] F.C. Meunier, J.P. Breen, V. Zuzaniuk, M. Olsson, J.R.H. Ross, J. Catal.
187 (1999) 493.
[52] G.M. Underwood, T.M. Miller, V.H. Grassian, J. Phys. Chem. A 103
(1999) 6184.
[53] K. Shimizu, H. Kawabata, A. Satsuma, T. Hattori, J. Phys. Chem. B 103
(1999) 5240.
[56] K. Shimizu, J. Shibata, A. Satsuma, T. Hattori, Phys. Chem. Chem. Phys. 3
(2001) 880.
[57] Mass spectrometer-library (MKS-PPT v4.24).
[58] C. Shi, M. Cheng, Z. Qu, X. Yang, X. Bao, Appl. Catal. B 36 (2002)
173.
[59] Z. Li, Ph.D. Thesis, Tufts University (1998).
[60] X. Zhang, A.B. Walters, M.A. Vannice, Appl. Catal. B 7 (1996) 321.
[61] C. Shi, A.B. Walters, M.A. Vannice, Appl. Catal. B 14 (1997) 175.
[62] X. Zhang, A.B. Walters, M.A. Vannice, J. Catal. 146 (1994) 568.
[63] O. Gorce, F. Baudin, C. Thomas, P. Da Costa, G. Djéga-Mariadassou,
Appl. Catal. B 54 (2004) 69.
[64] A.D. Cowan, R. Dumpelmann, N.W. Cant, J. Catal. 151 (1995) 356.
[65] H. Kato, C. Yokoyama, M. Misono, Catal. Today 45 (1998) 93.
[54] S. Kameoka, Y. Ukisu, T. Miyadera, Phys. Chem. Chem. Phys. 2 (2000)
367.