Vanadia-based SCR catalysts supported on tungstated and sulfated zirconia: Influence of doping with potassium

Article abstract of DOI:10.1016/j.jcat.2007.07.016

A series of vanadium-based SCR catalysts supported on sulfated or tungstated ZrO₂ were synthesized and characterized by means of N₂-BET, XRD, NH₃-TPD and in situ Raman spectroscopy. The effect of potassium doping on the properties of vanadia species was studied in detail. A number of catalyst preparation parameters were examined, including the choice of precipitant, variation of carrier surface area, potassium poisoning, crystallinity, and ZrO₂-phase composition. The results show that the catalyst structure and SCR activity are affected from the synthesis route by the support crystallinity and morphology, the surface composition, and the molecular configuration of the dispersed vanadates. It was observed that poisoning with potassium had a negligible effect on the surface vanadate species (especially the V{double bond, long}O stretching frequency observed by in situ Raman spectroscopy) if supported on the sulfated zirconia. Conversely, a more pronounced influence on the structure of surface vanadate was observed for the corresponding unsulfated samples, in which a significant red shift in the V{double bond, long}O stretching frequency was observed on potassium doping. Computational studies suggested that potassium was responsible for both the observed decrease in V{double bond, long}O stretching frequency and the higher proportion of dimers and higher polymers through coordination between K⁺ and two neighbouring V{double bond, long}O. The results suggest an increased resistance toward potassium doping for the vanadia-based catalysts supported on sulfated zirconia.

Full text of DOI:10.1016/j.jcat.2007.07.016

Journal of Catalysis 251 (2007) 459–473
www.elsevier.com/locate/jcat
Vanadia-based SCR catalysts supported on tungstated and sulfated zirconia:
Influence of doping with potassium
a,∗
b
a
a
Johannes Due-Hansen , Soghomon Boghosian , Arkady Kustov , Peter Fristrup ,
b
a
a
a,∗
George Tsilomelekis , Kenny Ståhl , Claus Hviid Christensen , Rasmus Fehrmann
a
Department of Chemistry and Center for Sustainable and Green Chemistry, Technical University of Denmark, 2800 Lyngby, Denmark
b
Department of Chemical Engineering, University of Patras and FORTH/ICE-HT, GR-26500 Patras, Greece
Received 19 April 2007; revised 13 July 2007; accepted 13 July 2007
Available online 5 September 2007
Abstract
A series of vanadium-based SCR catalysts supported on sulfated or tungstated ZrO were synthesized and characterized by means of N -BET,
2
2
XRD, NH -TPD and in situ Raman spectroscopy. The effect of potassium doping on the properties of vanadia species was studied in detail.
3
A number of catalyst preparation parameters were examined, including the choice of precipitant, variation of carrier surface area, potassium
poisoning, crystallinity, and ZrO -phase composition. The results show that the catalyst structure and SCR activity are affected from the synthesis
2
route by the support crystallinity and morphology, the surface composition, and the molecular configuration of the dispersed vanadates. It was
observed that poisoning with potassium had a negligible effect on the surface vanadate species (especially the V=O stretching frequency observed
by in situ Raman spectroscopy) if supported on the sulfated zirconia. Conversely, a more pronounced influence on the structure of surface
vanadate was observed for the corresponding unsulfated samples, in which a significant red shift in the V=O stretching frequency was observed
on potassium doping. Computational studies suggested that potassium was responsible for both the observed decrease in V=O stretching frequency
+
and the higher proportion of dimers and higher polymers through coordination between K and two neighbouring V=O. The results suggest an
increased resistance toward potassium doping for the vanadia-based catalysts supported on sulfated zirconia.
©
2007 Published by Elsevier Inc.
Keywords: NO SCR with ammonia; Potassium poisoning; Biomass; Deactivation; Tungstated zirconia; Sulfated zirconia; Vanadium oxide; In situ Raman
spectroscopy; DFT calculations
1
. Introduction
other pertinent topics; these have been summarized in several
recent reviews [1,2]. Concerning the reaction mechanism, most
research findings suggest the dual-site Eley–Rideal mechanism
involving a surface vanadia redox site and an adjacent nonre-
ducible acid–base site [3–5]. The key and rate-determining step
in this mechanism is assumed to be the activated adsorption of
The selective catalytic reduction (SCR) of NOx by NH3 in
the presence of O2 remains among the state-of-the-art technolo-
gies for controlling NOx emissions from stationary sources,
even though alternative strategies are under intensive develop-
ment. Transition metal oxides (most commonly vanadia) sup-
ported on anatase and promoted with tungsten or molybdenum
oxides are the most active SCR catalysts used on an indus-
5+
ammonia on the V –OH Brøndsted acid site, which reduces
5
+
4+
a nearby V =O redox site to V . This activated interme-
diate V4 –OH, is very reactive toward gaseous NO, leading
+
◦
trial scale at 300–450 C. Numerous studies have been pub-
to the products, N2 and H O, which desorb from the sur-
2
lished related to catalytic activity, reaction mechanism, effects
of vanadia loading, effects of active phase composition, and
face [3–7].
The use of biomass as an alternative to fossil fuels has at-
tracted increased interest because it is considered CO2-neutral
with regard to the human impact on the CO2 content in the at-
mosphere. However, a major disadvantage of using biomass is
the resulting fast deactivation of the SCR catalyst. Commercial
*
Corresponding authors.
E-mail addresses: jdh@kemi.dtu.dk (J. Due-Hansen), rf@kemi.dtu.dk
(R. Fehrmann).
0021-9517/$ – see front matter © 2007 Published by Elsevier Inc.
doi:10.1016/j.jcat.2007.07.016

460
J. Due-Hansen et al. / Journal of Catalysis 251 (2007) 459–473
SCR catalysts, consisting of TiO2 as a high-surface area sup-
port and V2O5–WO3 as the active component, deactivates about
three to four times faster than what is observed in conventional
coal-fired installations. The main reason for this deactivation
is the presence of high amounts of potassium (up to 2 wt% in
straw), which acts as a poison for the catalyst [8]. Therefore,
there is a need for new SCR catalysts that are more resistant
to deactivation by alkali metals, especially potassium. There
exist only few studies dealing with the effect of alkali oxide
additives on the activity and physicochemical properties of the
vanadium-based catalysts. The influence of alkali metals on the
activity of V O /TiO catalysts has been reported by several
authors [9–12], most of whom concluded that poisonous addi-
tives are affecting the Brønsted acid sites, which are responsible
for ammonia adsorption, thus decreasing both their number and
activity in NO reduction. Different strategies are available to
regenerate deactivated catalysts; for example, washing with a
solution of diluted sulfuric acid enables complete restoration of
the initial activity of the catalyst [8,13]. The major disadvantage
of such procedures is that in most cases, the power plant must
be closed down to regenerate the deactivated catalyst. Another
efficient way to overcome the deactivation problems could be
to use an alternative catalyst with greater resistance toward poi-
soning with potassium. In this context, the use of highly acidic
sulfated or tungstated zirconia as a support for vanadia-based
SCR catalysts has been reported to enhance catalyst resistance
toward alkali poisoning to some extent [14,15], although these
early studies performed no comprehensive characterization of
the surface and the influence of potassium on the vanadium
species.
needed for the creation of a highly active catalysts. Anatase is
known to fulfill this requirement, whereas silica is not.
The present work, in contrast to our previous work on this
subject [14,15], is concerned with the synthesis and multidisci-
plinary characterization of vanadium-based SCR catalysts sup-
ported on sulfate- or tungstate-promoted ZrO2, particularly the
effect of potassium doping on the properties of vanadia species
and catalytic activity. In addition, it examines the influence of
various catalyst preparation parameters, including the choice of
precipitant, variation of carrier surface area, potassium poison-
ing, crystallinity, and ZrO2-phase composition. To gain insight
into the structure and deactivation of the catalytically active
species by potassium, density functional calculations were ap-
plied and possible surface configurations investigated.
2
5
2
2. Experimental
2.1. Catalyst preparation
The zirconium oxide was prepared using two procedures
adapted from previous work [18,19]: a one-step synthesis, in
which tungsten oxide and hydrous zirconia are coprecipitated,
and a two-step method, in which the precipitated hydrous zirco-
nia is subsequently immersed into a diluted solution of sulfuric
acid. The resulting tungstated or sulfated hydrous zirconia was
◦
then dried and calcined at 700 C. The overall synthesis route
is summarized in Scheme 1.
Zirconyl nitrate hydrate, (ZrO(NO ) ·nH O, 27% Zr, Fluka),
3
2
2
was dissolved in water at room temperature with ammonium
metatungstate (Purum >97%) to obtain 20 wt% WO /ZrO af-
3
2
The surface composition under operating conditions is a sub-
ject of discussion, and the molecular structure and configura-
tions of the catalytically active components are under debate.
In this respect, the reported influence of specific oxide supports,
along with the observed stability of terminal V=O bonds dur-
ing the SCR reaction [5], suggests that the V–O–support bond
is involved in the rate-determining step. It has been shown that
the SCR activity of a series of transition metal oxides supported
ter calcination. This WO content was selected based on the
studies of Sohn et al. [20] and Santiesteban et al. [18], who
3
found the maximum surface acidity at 20 wt% loading of WO3
on zirconia. In previous work [15], the calcination temperature
◦
of the tungstated zirconia support was optimized to be 700 C,
which gives optimal results in terms of surface area, surface
acidity, and catalytic performance in the SCR reaction. Zirco-
nia support promoted with the tungsten oxide is further labelled
as WZ_AH, with AH corresponding to the precipitating agent
ammonium hydroxide.
on TiO correlates well with the extent of interactions between
2
the active phase and the support [16]. Finally, structure–activity
relationships based on in situ FTIR and Raman spectra, in com-
bination with data from catalytic experiments, suggest that the
active sites are most probably pairs of surface vanadate species
in close proximity to each other [4,17]. Thus, it seems that
the combination of an electronic effect of the support on the
V=O redox site and the presence of adjacent Brønsted sites are
The same zirconyl precursor was applied for the synthesis
of SO2 /ZrO , but ethylene diamine was used as a precipita-
−
2
4
tor instead of aqueous ammonia, resulting in zirconia with a
higher surface area. According to D’Souza et al. [19], ethylene
diamine most likely acts as a bifunctional agent, both precipi-
tating the zirconyl to hydrous zirconia and also functioning as
Scheme 1.

J. Due-Hansen et al. / Journal of Catalysis 251 (2007) 459–473
461
a colloidal protecting agent for the Zr(OH)4 domains. Several
2.4. X-ray powder diffraction
samples were also prepared with ammonium hydroxide as pre-
cipitating agent, to compare the effect of precipitants.
Room-temperature X-ray powder diffraction (XRPD) mea-
surements were performed on a Philips PW 1820/3711 powder
diffractometer using CuK radiation (λ = 1.5406 Å/1.5443 Å)
For the synthesis of SO2 /ZrO2, 66.9 g ZrO(NO3)2·nH2O
−
4
was dissolved in 700 mL of water at room temperature. When
the zirconyl precursor was dissolved, 75 mL of ethylene di-
amine was added slowly during stirring, followed by precipita-
tion of hydrous zirconia. The stirring was continued overnight
α
1
◦
◦
with a range of 10–65 in steps of 0.02 in 2θ. In situ high-
temperature XRPD was carried out at beamline I711 at the
MAXII synchrotron, MAX-lab, Lund, Sweden, on a Huber
G670 imaging strip Guinier camera, using a wavelength of
◦
◦
at 50 C, followed by heating for 1 week at ∼65 C, to avoid de-
◦
composition of ethylene diamine (boiling point 117 C). After
λ = 1.154424(8) Å. The data were accumulated for 60 s in
2
◦
◦
this, two phases were present: a clear slightly yellow liquid at
the top indicating an excess of ethylene diamine and the precip-
itated hydrous zirconia in the bottom fraction. The slurry was
filtered and washed with about 1000 mL of water in a Büch-
the range of 4–100 in steps of 0.005 in 2θ. For easy compari-
son, the angle of the latter is converted by applying Braggs law:
θ = arcsin((λ /λ )sinθ ), where θ is the angle correspond-
1
1
2
2
2
ing to the data obtained with the wavelength 1.154424 Å.
◦
ner funnel and dried in vacuum oven overnight at 100 C and
1
0 mmHg. The resulting xerogel was sulfated, based on an op-
2
.5. In situ Raman spectra
timized procedure from Gonzalez et al. [21]. Freshly prepared
Zr(OH)4 was immersed overnight in a 0.25 N H2SO4 solution,
1
a small amount of 0.25 N H2SO4 (0.125 M) and dried in a
vacuum oven at 100 C overnight. All promoted supports were
A homemade Raman cell was used for recording the in situ
5 mL/g. Then the mixture was suction filtered, washed with
Raman spectra of the studied catalysts consisting of a double-
walled quartz-glass transparent tube furnace mounted on a xyz
plate, allowing it to be positioned on the optical table. The in-
ner furnace tube (23 mm o.d., 20 mm i.d., and 10 cm long) is
Kanthal wire-wound for heating the cell. The cell has a gas in-
let and a gas outlet, as well as a thermocouple sheath with a
sample holder at its tip. Approximately 180 mg of each catalyst
was pressed into a wafer and mounted on the holder that could
be vertically adjusted in the in situ cell.
The gases used were O2 (L’Air Liquide 99.995%), 10,000
ppm NH3/N2 and 10,000 ppm NO/N2 mixtures (L’Air Liquide)
and N2 (L’Air Liquide 99.999%) as balance gas. The gas feed
consisted of 2000 ppm NO, 2200 ppm NH3 (NH3/NO = 1.1),
and 7% O2 balanced in N2 or various mixtures of these at a
total feed flow rate of up to 50 mL/min. To study the effect
of H2O(g) on the in situ Raman spectra, the dry feed gas was
enriched by 3% H2O(g) as described earlier [22].
◦
◦
calcined at 700 C for 4 h in air. The short calcination time min-
imized the decomposition of sulfate at elevated temperatures.
The supported V-based catalysts were prepared by impreg-
nating the support with an aqueous solution of ammonium
vanadate to obtain a vanadium loading of 1.7 wt% (3.0 wt%
V2O5). The incipient wetness method was used for impreg-
nation. After impregnation, the samples were dried for 2 h at
◦
◦
1
00 C and calcined for 3 h at 400 C in air.
The potassium-poisoned catalysts were prepared by impreg-
nation with a solution of KNO3 to obtain a potassium loading
of 134 µmol/g, corresponding to a K/V molar ratio of 0.4. Sam-
◦
◦
ples were then dried in air at 100 C and calcined at 300 C. The
K-doped, vanadium-impregnated sulfated zirconia samples are
designated KVSZ_AH or KVSZ_ED, where AH and ED indi-
cate the precipitating agent ammonium hydroxide and ethylene
diamine, respectively.
+
The 488.0-nm line of a Spectra Physics Stabilite 2017 Ar
laser was used for recording of the Raman spectra. The laser
beam, operated at 40 mW, was focused on the sample by a
cylindrical lens, thus dispersing it to reduce sample irradiance.
2
.2. Nitrogen adsorption measurements
◦
The scattered light was collected at 90 , analyzed with a 0.85-m
Nitrogen adsorption measurements were performed at liquid
Spex 1403 double monochromator, and detected by an RCA
◦
nitrogen temperature using a Micromeritics Gemini analyzer.
PMT cooled to −20 C equipped with EG&G photon-counting
◦
The samples were evacuated at 200 C for 1 h before the mea-
electronics. The entire setup is illustrated in Fig. 1.
◦
surements. The total surface area was calculated according to
the BET method.
Recording of spectra started in O at 400 C (dehydrated
2
◦
conditions) after the sample was exposed for 1 h at 400 C in
pure O2. Raman spectra were then recorded sequentially after
stabilization for 1 h in NH3/H2O/N2 (reductive), NH3/NO/O2/
H2O/N2 (NO-SCR operational), and finally in O2 (reoxidation)
2
.3. Thermogravimetric analysis coupled with mass
spectroscopy
◦
at 400 C.
Simultaneous thermogravimetric analysis (TGA) and mass
spectroscopy (MS) of the gases evolved from the sulfated zir-
conia sample were performed using a TGA–MS system that
includes a NETZSCH STA 409 PC/PG and a NETZSCH QMS
2.6. NO SCR with ammonia
Before the activity measurements, samples where pressed
into tablets, crushed, and sieved to obtain fractions of 0.18–
0.295 mm. The SCR activity measurements were carried out in
a fixed-bed reactor, with 75 mg of the catalyst loaded between
4
03 C mass spectrometer. The samples were purged with N2
◦
and scanned from 20 to 1200 C, with a temperature ramp rate
◦
of 20 C/min.

462
J. Due-Hansen et al. / Journal of Catalysis 251 (2007) 459–473
Fig. 1. Schematic illustration of the in situ Raman setup.
two layers of inert quartz wool. The typical reactant gas com-
position was 1000 ppm NO, 1100 ppm NH3, 3.5% O2, 2.3%
H2O, and balance N2. The total flow rate was maintained at
LACVP* basis set as incorporated in the Jaguar program pack-
age [26]. The LACVP* basis set uses the Los Alamos effec-
tive core potential (ECP) and basis set for the description of
heavy atoms (V, K) [27], whereas all lighter atoms (O, C, H)
are described using the 6-31G* basis set [28]. The ECP was
augmented with an explicit treatment of the outermost core
electrons and valence electrons, resulting in the electron config-
3
00 mL/min (ambient conditions). The NO concentration was
continuously monitored by a Thermo Electron model 10A rack-
mounted chemiluminescent NO–NOx gas analyzer.
The catalytic activity, represented as a first-order rate con-
stant (k), can be calculated from the NO conversion, X, as
2
6
4
1
2
6
1
urations [core]3s 3p 3d 4s for vanadium and [core]3s 3p 4s
for potassium. For the obtained minimum energy structures,
Mulliken charges [29] and charges fitted from the electrosta-
tic potential (ESP) [30–32] were calculated to characterize the
F0
k = −
ln(1 − X),
[
NO]0 · Vcat
+
interactions among V, O, and K .
where F0 is the molar inflow of NO, [NO]0 is the initial molar
concentration of NO, and Vcat is the volume of the catalyst bed.
Analytical frequencies were calculated for all structures,
which allowed derivation of thermodynamic properties at T =
2
98 K. The reported frequencies were scaled using a scaling
2
.7. NH3 temperature-programmed desorption
factor of 0.9614, which is the appropriate scaling factor for the
B3LYP/6-31G* method [33].
Temperature-programmed desorption of ammonia (NH3-
TPD) was performed as follows. First, 100 mg of the sample
was loaded into a quartz tube reactor and saturated with ammo-
nia (100 mL/min, 1% NH3/N2) for 45 min at room tempera-
ture. Before the NH3 desorption measurement, the sample was
heated to 100 C in a dry nitrogen flow (100 mL/min) and kept
at this temperature for 1 h to remove physisorbed ammonia.
3
. Results and discussion
3
.1. Surface area and vanadium surface coverage
◦
The results of the BET surface area measurements for sam-
ples prepared using different precipitating agents are summa-
rized in Table 1. It is noticeable that the use of ethylene di-
amine as precipitant resulted in a roughly twofold higher spe-
cific surface area for the ZrO2 carrier compared with the sam-
ples prepared with ammonium hydroxide as a precipitant. In
addition, the use of different precipitating agents during prepa-
◦
Then the sample was cooled to 50 C, after which the tempera-
◦
◦
ture was raised at a rate of 5 C/min up to 650 C. The rate of
NH3 desorption was monitored by a computer-interfaced Jasco
V-570 UV/VIS/NIR spectrophotometer using the characteristic
ammonia band at 201 nm.
ration of the sulfated carrier (SO2 /ZrO2) resulted in materials
−
4
2
.8. Computational methods
of 2- to 3-fold greater surface areas compared with the corre-
sponding nonsulfated carrier. Incorporation of tungsta in zir-
conia (tungstated zirconia supports) resulted in materials with
nearly 2-fold higher specific surface areas compared with the
Energy minimization was performed using the hybrid den-
sity functional B3LYP [23–25], in combination with the

J. Due-Hansen et al. / Journal of Catalysis 251 (2007) 459–473
463
Table 1
Sample characteristics
Table 2
a
Atomic fractional coordinates for the phases used in the Rietveld refinements.
The space group P2 /n is only applied for the samples containing tungsten ox-
ide
1
Catalyst
sample
SBET
(m /g)
ns
ns
NH des.
3
(µmol/g)
2
2
2
(VOx/nm )
(WOx/nm )
Atom
Fractional coordinates
SZ_ED
VZ_ED
KVZ_ED
VSZ_ED
KVSZ_ED
118.0
51.6
50.2
102.9
97.0
375
3.9
4.0
2.0
2.1
x
y
z
tet-ZrO (P4 /nmc)
Zr
O
2
2
551
363
0.750
0.250
0.250
0.250
0.750
0.800
VZ_AH
WZ_AH
VWZ_AH
KVWZ_AH
23.4
94.7
58.4
57.9
8.6
mon-ZrO (P2 /c)
Zr
O1
O2
2
1
5.5
8.9
9.0
253
351
191
0.276
0.072
0.449
0.040
0.333
0.758
0.209
0.347
0.476
3.4
3.5
SZ_AH
VZ_AH
KVZ_AH
VSZ_AH
KVSZ_AH
84.8
23.4
21.6
71.9
66.6
354
cubic-ZrO (Fm3m)
2
Zr
O1
8.6
9.3
2.8
3.0
0.000
0.250
0.000
0.250
0.000
0.250
516
292
mon-WO3 (P21/n)
W1
W2
O1
O2
O3
O4
O5
O6
0.252
0.247
0.000
0.000
0.290
0.210
0.280
0.280
0.026
0.033
0.025
0.475
0.290
0.290
0.040
0.240
0.285
0.781
0.220
0.220
0.265
0.735
0.030
0.000
VWT_ref
KVWT_ref
68.3
62.7
2.9
3.2
2.7
2.9
a
All vanadia-containing samples are loaded with 1.7 wt% vanadium. For
samples doped with potassium, K/V = 0.4 (molar ratio).
corresponding nontungstated samples (Table 1). The increased
surface area would be expected to lead to better dispersion of
the active components and lower values for the vanadium sur-
face coverage, ns (VOx/nm ), where the saturation value for
the monolayer surface coverage of vanadium was reported to
2
Integration of the SO -desorption signals for the AH- and
2
ED-precipitated samples (Fig. 2) revealed an approximately
33% higher SO -desorption area for the ED-precipitated sam-
ples. VSZ_ED (102.9 m /g) exhibited a 43% greater surface
area than VSZ_AH (71.9 m /g). This suggests that the amount
2
be approximately 8 VOx/nm [34].
2
2
2
3
.2. Thermogravimetric analysis and mass spectroscopy
of sulfate on the samples is, to some extent, proportional to the
surface area.
The results of TGA–MS for the sulfated Zr(OH)4 xerogel
samples precipitated with ED and AH and dried at 100 C are
presented in Fig. 2. The TGA curve of the sulfated zirconia xe-
◦
3.3. XRD and phase compositions
rogel contains at least two weight loss features. The first, below
◦
2
00 C with m/e = 18 response in the MS signal, is attributed
The temperature used for calcination of the sulfated zirconia
samples was based on results from temperature-programmed
XRD, depicted in Fig. 3. At 700 C, a maximum in the amount
of tetragonal ZrO2-phase was observed, which has been sug-
gested to be necessary for the generation of high acidity in sul-
fated zirconia [35,36]. This phase remains stable during cool-
ing, as can be seen in Fig. 3 (right). The periodic variations in
the intensities are due to instabilities in the beamline.
The XRD patterns of selected samples were refined with
the Rietveld refinement method, as shown in Fig. 4. All of the
samples were composed of three crystal phases with different
positions of zirconium and oxygen atoms and bond lengths.
The unit cell symmetries can be described by the space groups
P42/nmc, P21/c, and Fm3m, corresponding to the tetragonal,
monoclinic, and cubic ZrO2 phases (P21/n for WO3-containing
samples), with atoms in the positions specified in Table 2. This
unit cell was used in the model of the XRD pattern for refining
the crystalline structures with the Rietveld method. The results
of the refinements of the crystalline phases are summarized in
Table 3. Through the refined XRD patterns, it is possible to
comment on the effect of the sulfation and the precipitant on
the crystal form and crystallinity of the zirconia matrix.
to desorption of physisorbed water. The second, between 400
◦
◦
and 1000 C, with MS signals of m/e = 64 and 32, can be at-
tributed to desorption of sulfate species in the form of SO2 and
O2, respectively.
A particular difference between the two samples was ob-
served regarding the temperature of surface sulfate species de-
composition, indicating that the preparation method influenced
the stability of these species. In the case of VSZ_AH, the
◦
m/e = 64 signal was observed at temperatures above 700 C,
whereas for the VSZ_ED, decomposition of the sulfate species
◦
proceed in two steps at temperatures around 400–500 C and
5
◦
00–900 C. Desorption of a small amount of water and SO2
was observed for the ED-precipitated sample at around 400–
◦
00 C, which could be connected to the decomposition of
5
H2SO4 or hydrated sulfate species.
The stability of the sulfate species was somewhat higher on
◦
the AH-precipitated samples, reflected by the ∼200 C higher
temperature of SO2 desorption for the AH-precipitated sample
compared with the corresponding ED-precipitated sample. This
could be due to differences in the textural properties (i.e., sur-
face acidity, porosity) related to the precipitating agent.

464
J. Due-Hansen et al. / Journal of Catalysis 251 (2007) 459–473
Fig. 2. Thermogravimetric and mass spectral analysis of the vanadium-impregnated sulfated zirconia samples, precipitated with two different agents, AH (ammonium
hydroxide) and ED (ethylene diamine).
2
−
◦
◦
Fig. 3. X-ray diffraction of SO /Zr(OH) heating from 450–700 C with 10 C steps (left), and cooling (right). The zirconia reflex of the (111)t and (111)m is
indicated.
4
4
The difference in SO2-desorption temperature observed in
the TGA–MS spectra (Fig. 2) also can be connected with the
distinction in crystallographic composition of ZrO2 and the
resulting difference in coordination strength of sulfate to the
support. Rietveld refinements suggested a higher amount of
monoclinic zirconia of the AH-precipitated samples (Table 3).

J. Due-Hansen et al. / Journal of Catalysis 251 (2007) 459–473
465
Table 3
Crystallinity and the calculated volumetric distribution of ZrO (and WO )
2
3
phases from Rietveld refinement
a
Sample
Crystallinity
V
V
Vmon
(vol%)
WO
tet+cubic
3
(%)
(vol%)
(vol%)
KVWZ_AH
KVSZ_AH
KVSZ_ED
KVZ_AH
KVZ_ED
SZ_AH
68.8
76.1
71.4
72.3
74.7
72.2
76.5
9.2 (0.2)
83.5 (1.6)
51.4 (1.8)
59.3 (1.8)
48.4 (2.1)
75.2 (2.6)
54.8 (2.2)
61.4 (2.3)
7.3 (0.4)
–
–
–
–
–
–
48.6 (0.9)
40.7 (0.9)
51.6 (1.1)
24.8 (1.2)
45.2 (1.0)
38.6 (1.3)
SZ_ED
a
Integration over (observed pattern − refined background pattern)/observed
pattern.
stabilizing effect of the ED in the zirconia matrix, as proposed
by Richards et al. [19]. These authors suggested a colloidal pro-
tecting effect of ethylene diamine, which hinders the hydrous
zirconia particles from sintering. This stabilizing effect of ED
could thus retard the phase transformation from tetragonal to
monoclinic zirconia. Comparing the effect of AH or ED precip-
itants on the SZ, KVZ, and KVSZ samples reveals that in all
cases, precipitating with AH resulted in a significantly higher
amount of monoclinic ZrO2 phase.
A similar comparison could not be performed for the copre-
cipitated WZ samples, due to the necessary washing step when
precipitating with ED. The washing would remove a significant
part of the coprecipitated tungsten oxide. However, tungsten
oxide is known to have a similar retarding effect on the transfor-
mation from tetragonal to monoclinic zirconia [20]. Therefore,
tungstated samples are assumed to consist exclusively of tetrag-
onal ZrO2 phase.
Fig. 4. Rietveld refinement of the XRD data of sulfated and tungstated zirconia
samples, precipitated with AH or ED, all calcined at 700 C in air. The dif-
◦
ference between the observed and calculated intensities is shown at the lower
curve below the corresponding profile.
3.3.2. Effect of sulfate
It is generally agreed that sulfation retards crystallization of
the zirconia matrix and thus the transformation from tetragonal
to monoclinic form [37]. The incorporation of sulfate species
into the Zr(OH)4 structure presumably stabilizes the tetrago-
nal zirconia phase by shifting the equilibrium phase distrib-
ution from the thermodynamically more stable monoclinic to
the metastable tetragonal phase. However, in our studies, we
observed that doping the zirconia matrix with sulfate seemed
to favor the formation of monoclinic ZrO2 after calcination
Although not studied further in this work, this implies that the
m-ZrO2 may be responsible for the relatively stronger binding
of sulfate species.
3
.3.1. Precipitant effects
Comparison of different ED-precipitated samples (i.e.,
SZ_ED, KVZ_ED, and KVSZ_ED in Table 3) indicates that
the presence of sulfate, vanadia, and potassium have a negligi-
ble effect on the crystallinity of the ZrO2 matrix, which in all
cases is >70%. The phase composition of the crystalline frac-
tion consists mostly of tetragonal ZrO2. Several studies have
proposed that the ZrO2 tetragonal phase is more active in catal-
ysis, and it also has been reported to be responsible for the
generation of high acidity [35,36].
The samples precipitated with AH also exhibited high crys-
tallinity (>70%), whereas about 50 vol% (of the crystalline
fraction) was calculated to exist as tetragonal or cubic zirconia.
The higher amount of tetragonal phase in the ED-precipitated
samples can be explained if we suggest that the phase transfor-
mation from tetragonal to the thermodynamically more stable
monoclinic ZrO2 is retarded to a higher degree for the ED-
precipitating agent than for AH. This can be explained by the
◦
at 700 C. Comparing the fraction of monoclinic ZrO2 for
KVZ_ED and two sulfated samples KVSZ_ED and SZ_ED
(Table 3) showed that the share of monoclinic phase increased
from 24.8 to 40.7 and 38.6 vol%, respectively, after sulfa-
tion.
Sulfation of AH-precipitated samples seemed to have less
impact on the generation of a specific crystal phase. Sul-
fating the KVZ_AH samples resulted in an insignificant in-
crease in tetragonal zirconia phase from 48.4 to 51.4 vol% for
KVSZ_AH.
3.3.3. Tungstated zirconia
The XRD pattern of KVWZ_AH displayed three crystalline
phases, two zirconia phases, and one monoclinic WO3 phase
(Fig. 4), with no crystalline vanadium or potassium oxides ob-

466
J. Due-Hansen et al. / Journal of Catalysis 251 (2007) 459–473
served due to their low loadings. The incorporation of tungsten
oxide into the zirconia matrix is known to stabilize the tetrag-
onal crystal phase over the monoclinic form [15,20]. This ten-
dency can be clearly seen when calculating the contribution of
different ZrO2-phases, with 83.5 vol% of the crystalline zirco-
nia existing as tetragonal and only 7.3 vol% existing as mono-
clinic. (Note that the values are not normalized to the presence
of crystalline WO3.) The appearance of crystalline tungsten ox-
◦
ide for the sample calcined at 700 C can be attributed to the
low surface area and thus a high surface concentration of WO3
species, far exceeding monolayer coverage.
3
.4. In situ Raman spectra and structural characteristics of
catalysts
3
.4.1. Effects of sulfation of the zirconia support and K
poisoning of catalysts
◦
Fig. 5 shows the in situ Raman spectra obtained at 400 C
under a flow of oxygen for representative samples prepared us-
ing ethylene diamine as the precipitant during the zirconia sup-
port synthesis step. It should be pointed out that the preparation
procedure resulted in mixtures of tetragonal and monoclinic zir-
conia phases and that the Raman spectra were dominated by
the spectral characteristics of the phase formed in excess. Spec-
−
1
trum 5(a) has a characteristic band at 636 cm , corresponding
to tetragonal ZrO2 [20] (ED-precipitated, calcined at 700 C).
◦
In agreement with the results of XRD analysis, the addition
of sulfuric acid in the precursor solution during the prepara-
tion of the sulfated zirconia (SO2 /ZrO2; spectrum Fig. 5b)
−
◦
Fig. 5. In situ Raman spectra of catalyst samples recorded under O at 400 C:
2
4
2
−
a) ZrO ; (b) SO /ZrO ; (c) 1.7V O /ZrO ; (d) K O–1.7V O /ZrO ;
4
2 2 2 5 2 2 2 5 2
resulted in the formation of a higher concentration of mono-
(
−
1
clinic ZrO2 (band at 622 vs 636 cm for the tetragonal phase).
The formation of surface sulfate species also was visible in the
Raman spectra. The Raman bands due to the surface sulfates
probably originated from two kinds of surface sulfate species.
2−
2−
(e) 1.7V O –SO /ZrO ; (f) K O–1.7V O –SO /ZrO . Laser wavelength,
2
5
4
2
2
2
5
4
2
−1 −1
λ = 488.0 nm; laser power = 40 mW, scan rate = 0.07–0.3 cm
s
; spectral
0
−
1
slit width = 8 cm
.
−1
−1
The bands at 994 cm and 1035 cm2 in Fig. 5b observed in
the V2O5/ZrO2 sample (Fig. 5c) leads to the following obser-
vations:
−
the expected range of ν1 and ν3 SO4 modes probably corre-
spond to the sulfate groups connected in a bidentate coordina-
−
1
−1
2 5
(i) The band at 993 cm due to V O particles was not
tion to the carrier. The band at 1394 cm is assigned to the
S=O stretching mode of a surface tridentate sulfate with a (Zr–
O)3–S=O configuration [38,39]. The Raman spectrum of the
V2O5/ZrO2 (1.7VZ_ED) sample (Fig. 5c) exhibited the known
2−
observed in the spectrum of the V2O5–SO4 /ZrO2 cat-
alyst (Fig. 5e). This is due to the fact that sulfation of
the support resulted in a significantly greater surface area
[
40] features of surface amorphous vanadates, that is, isolated
2
(
S
= 102.9 m /g) and correspondingly lower vana-
BET
tetrahedral-like monomeric and polymeric vanadates (Zr–O)3–
2
dium surface coverage (ns = 2.0 VOx/nm ) compared
−
1
V=O. The band at ∼1030 cm is due to V=O stretching and
2
with V2O5/ZrO2 (ns = 3.9 VOx/nm ); therefore, vanadia
−
1
the broad band in the 700–940 cm region, centered around
60 cm , is due to V–O–V structures. The weak narrow band
at 993 cm indicates the presence of traces of V2O5 bulk par-
ticles. Such V2O5 particles with distorted VO5 coordination
was dispersed only in the form of amorphous vanadates.
Decreasing the V surface coverage also resulted in a higher
signal-to-noise ratio, due to lower absorption of the inci-
dent laser light by the sample surface.
−
1
7
−
1
can be present even in catalysts with submonolayer coverage
2
(ii) The peak related to the V–O–V mode was blue-shifted for
(
4 VOx/nm in this sample, slightly above half monolayer) and
2
−
the V O –SO /ZrO catalyst (Fig. 5e) and appeared to
2
5
2
are not detectable by XRD. These particles can be dispersed
into amorphous state by subjecting the catalyst to the reaction
mixture [34]. Indeed, after the V2O5/ZrO2 catalyst was exposed
to a flow of NH3/H2O/N2 and subsequent reoxidation, the band
4
−
1
−1
be centered at ca. 900 cm , compared with ca. 760 cm
in the spectrum of V2O5/ZrO2 catalyst (Fig. 5c). Accord-
ing to a recent study [17], such spectral behavior has sig-
nificant structural implications and is compatible with a
small amount of V–O–V chains consisting of polyvana-
dates. We discuss this last point in more detail later.
−
1
at 993 cm nearly disappeared.
We now focus on the effect of sulfation. Comparing the spec-
tra of the V2O5–SO2 /ZrO2 (VSZ_ED) catalyst (Fig. 5e) and
−
4

J. Due-Hansen et al. / Journal of Catalysis 251 (2007) 459–473
467
−
1
(
iii) The position of the V=O mode at 1030 cm remained
unaffected.
The effect of K poisoning on the Raman spectra of the
catalysts can be examined by comparing the spectrum of the
K–V2O5/ZrO2 (K–1.7VZ_ED) catalyst (Fig. 5d) to the spec-
trum of the V2O5/ZrO2 sample (Fig. 5c) and by comparing the
2−
spectra of the K–V2O5–SO4 /ZrO2 (K–1.7VSZ_ED) catalyst
Fig. 5f) and the V2O5–SO2 /ZrO2 sample (Fig. 5e). The fol-
−
(
4
lowing observations can be made:
(
i) The band of the tridentate surface sulfate, (Zr–O)3–S=O,
1
−
at 1394 cm S=O was not present in the spectrum of the
K-poisoned sulfated catalyst (Fig. 5f).
(
ii) The V=O band was red-shifted from ∼1030 to ∼990
−
1
cm
1005 cm
Fig. 5f).
for K–V2O5/ZrO2 (spectrum in Fig. 5d) and to
−1
2−
for K–V2O5–SO4 /ZrO2 (spectrum in
∼
(
iii) The peak related to the V–O–V functionalities was red-
shifted, and its intensity was enhanced relative to that at
−1
6
30 cm in both K-doped samples. In particular, the high-
frequency wing almost disappeared on going from spec-
trum c to spectrum d, and the center of the broad peak was
−
1
−1
red-shifted by ∼80 cm (from ∼900 to ∼820 cm ) on
going from spectrum e to spectrum f.
It is evident that the simultaneous presence of K2O and sul-
fate groups dispersed on the surface can result in the formation
of potassium sulfate, thereby contributing also to enlargement
◦
Fig. 6. In situ Raman spectra of catalyst samples recorded under O at 400 C:
2
−
1
of the band at ∼1000 cm (where the main stretching band for
(
a) ZrO ; (b) WO /ZrO ; (c) 1.7V O /ZrO ; (d) K O–1.7V O /ZrO ; (e)
.7V O –WO /ZrO ; (f) K O–1.7V O –WO /ZrO . Recording parameters:
2 3 2 2 5 2 2 2 5 2
the sulfate group is expected) in the spectrum of the K–V2O5–
1
2 5 3 2 2 2 5 3 2
2
−
SO /ZrO2 sample (Fig. 5f). Moreover, as has been suggested
See Fig. 5 caption.
4
by Bulushev et al. [12], potassium could, through an electro-
δ−
δ−
static interaction, (Zr–O)3–V=O ···K , cause an elongation
of the V=O bond, thereby explaining the red shift observed
for this band for the K-poisoned samples. Another explanation
of the decreased red shift of the V=O band on K-doping the
phous polytungstate species: a strong sharp band due to W=O
−
1
stretching at 1019 cm originating from mono-oxo species
−
1
[
42] and broad bands at ∼910 and 825 cm occurring in the
asymmetric W–O–W stretching region. No bands due to crys-
2
−
−1
V O –SO /ZrO sample (∼1006 cm ; Fig. 5f) compared
talline WO can be observed, indicating very good dispersion
2
5
2
3
4
−
1
with K–V2O5/ZrO2 (988 cm ; Fig. 5d) could be attributed to
the presence of sulfate species, leaving the V=O sites less af-
fected by the interaction with potassium. The red shift of the
V–O–V band indicates a weakening of the bridging V–O bonds
of surface tungstates and nearly complete coverage of zirconia.
2
The surface coverage of the sample, 5.5 WO /nm , slightly ex-
x
2
ceeded the experimentally reported monolayer (4 WO /nm
x
[42]). It is possible that tungstates remained trapped in the
inner pores of the support as a result of the applied synthe-
sis route. The Raman spectrum shown in Fig. 6b is in agree-
ment with previously reported spectra for WO /ZrO samples
and could be a result of the formation of (VO ) units with a
x n
higher degree of polymerization, as we discuss later. Finally, it
has been proposed that some of the O=V–O–support anchor-
ing bridges can be substituted by O=V–O–K [41], which also
could contribute to elongation of the V=O bond.
x
2
[43,44] and is characteristic of a sample with high tungsta load-
ing not exceeding the monolayer coverage. The structure of
surface tungsta species on metal oxide supports is a function
3
.4.2. In situ Raman spectra of V2O5–WO3/ZrO2 catalysts
Fig. 6 shows the in situ Raman spectra obtained at 400 C
of WO coverage [42], and multiple geometrical configura-
x
◦
tions for polytungstates with varying bond orders for bridging
−
1
under a flow of O2 for representative samples prepared using
ammonium hydroxide as precipitant in the zirconia support syn-
thesis step. Fig. 6a corresponds to monoclinic ZrO2 and has
a characteristic band at 637 cm [20]. In situ Raman spec-
trum of the mixed WO3/ZrO2 support material, synthesized by
coprecipitation as described earlier, is shown in Fig. 6b. The
bands observed are characteristic of dispersed surface amor-
W–O–W bonds (spectral region 800–950 cm ) occur at the
surface for the samples with high tungsten loadings [44].
The spectrum obtained for the V O /ZrO (VZ_AH) sample
2
5
2
−
1
is shown in Fig. 6c. Due to the lower BET area of this sam-
2
ple (23.4 m /g) compared with the sample prepared using ED
2
as precipitant (51.6 m /g), and its correspondingly higher sur-
2
face VOx coverage (8.6 VOx/nm ), the saturation value for the

468
J. Due-Hansen et al. / Journal of Catalysis 251 (2007) 459–473
2
surface VOx monolayer (∼8 VOx/nm ) was exceeded, and the
−1
band at 993 cm due to crystalline V2O5 can be clearly seen in
spectrum in Fig. 6c. The remaining features observed in spec-
−
1
trum in Fig. 6c (i.e., the band at ∼1030 cm due to V=O and
−
1
the very broad V–O–V band at 800–930 cm ) are character-
istic of surface amorphous vanadates in the catalysts with high
vanadium loading.
The effect of tungstation can be revealed by comparing
spectrum in Fig. 6e obtained for the V2O5–WO3/ZrO2 sam-
ple (VWZ_AH) with the spectrum in Fig. 6c obtained for the
corresponding V2O5/ZrO2 (VZ_AH) sample. Two very intense
−
1
bands at 714 and 801 cm dominate in spectrum in Fig. 6e and
are characteristic of crystalline WO3 [45]. The positions and
shapes for these bands are known to be dependent on the crys-
tallite size and temperature [45]. It is evident that deposition
Fig. 7. Molecular models/configurations for monomeric (a), dimeric (b, c) and
polymeric (d, e) dispersed surface vanadates.
strength in valence units (vu). According to the Pauling’s va-
lence sum rule, the sum of the individual V–O bond strengths
for a viable configuration with four-coordinated vanadium for
a vanadium oxide species should be equal to 5 vu [22,40,49].
This approach is based on the diatomic approximation, which is
necessary for justifying the direct relationship between metal–
oxygen bond lengths and Raman stretching frequencies [48].
As an example, we can calculate the bond orders for the surface
isolated (Zr–O)3–V=O species (configuration a in Fig. 7). The
of vanadia on WO /ZrO resulted in the “extraction” of amor-
3
2
phous (WOx)n and a redistribution of a significant portion of
surface tungsta in the form of crystalline WO3. The rest of the
features in Fig. 6e are largely obscured by the envelopes of the
above bands and appear to consist of overlapping contributions
from bands due to noninteracting surface amorphous vanadates
−
1
−1
and tungstates at ∼1020 cm (M=O region) and ∼940 cm
(
M–O–M region). Earlier work [46] has demonstrated that non-
interacting surface vanadates and tungstates are also formed on
TiO2.
The effect of K poisoning is similar to that discussed in the
context of Fig. 5 for sulfated supports, that is, elongation of the
−1
V=O band at 1029 cm (Fig. 5c) corresponds to a bond order
of 1.89 vu (1.581 Å), and the remaining 3.11 vu for the three
V–O–Zr give rise to a BO close to unity (1.04 vu) for each V–O
along the anchors.
−
1
M=O bonds (red shift of the band at ∼1020 cm in the spec-
Likewise, we can address the case of surface polyvanadates,
for which the V–O–V functionalities exhibit a very broad band
−
1
trum in Fig. 6e to ∼990 cm in the spectrum in Fig. 6f) and
reorganization of polyvanadates and polytungstates to species
−1
character (740–920 cm ) in the Raman spectra (Fig. 5c) due to
−
1
with weaker M–O–M bridging bonds. The band at 940 cm in
the spectrum in Fig. 6e was red shifted and presumably covered
under the envelope of the 801 cm WO band in spectrum in
the variation of V–O bond orders within geometrically related
configurations for the (VOx)n species. Band positions at ∼900
−
1
3
−1
and ∼750 cm imply bond orders of 1.51 and 1.14 vu, respec-
Fig. 6f.
tively. Thus, checking the viability of configuration b with one
−
1
−1
,
short V=O (1029 cm , 1.89 vu) and one V–O–V (900 cm
3
.4.3. Raman band assignments and structural implications
The molecular configurations for surface vanadates in sup-
ported vanadia catalysts under dehydrated conditions evolve
from isolated monovanadates and larger polyvanadates with
tetrahedral coordinated vanadium [47] at low vanadium cover-
1
.51 vu), the valence units remaining for the two V–O(–Zr)
bonds (determined from the constraint on the total bond order
for the V atom) correspond to 0.8 vu for each V–O(–Zr). Con-
−
1
figuration d also can be viable with one short V=O (1029 cm
,
−
1
1
0
.89 vu) and two V–O–V (750 cm , 1.14 vu), resulting in ca.
age to bulk V O when the monolayer coverage of the support
2
5
.8 vu for the V–O(–Zr) bond (5 − 1.89 − 2 × 1.14 = 0.83).
is exceeded. Fig. 7 shows molecular models for possible con-
figurations of the isolated and polymeric amorphous surface
vanadates. Here we attempt to correlate the effect of the vari-
ous preparation steps (e.g., choice of precipitant, sulfation, and
K-doping) as expressed in the Raman spectra with the response
of the particular structure of the dispersed vanadia phase.
The consistency of the Raman assignments can be checked
using the following empirical correlations between V–O Raman
stretching frequencies and the corresponding V–O bond lengths
and Pauling bond strengths [48]:
It is evident that a V–O–V in the high-frequency edge (e.g.,
−
1
9
00 cm ) is compatible with a small size for the (VOx)n unit
(
n = 2, 3); if there were two V–O–V bridges per V (configura-
tions c and d in Fig. 7) with the BO for each V–O around 1.5 vu,
then there would remain no valence residue for the fourth (an-
choring) V–O. Therefore, a red shift of the V–O–V functionali-
ties indicates a reorganization of the surface polyvanadates in a
form with longer chains and fewer anchoring V–O–Zr bridges
per V. In contrast, a blue shift of the broad band is compatible
with formation of smaller units with less V–O–V bridges per V.
In the K-doped V O /ZrO sample (KVZ_ED; Fig. 5d)
ν = 21349exp(−1.9176R),
(1)
(2)
2
5
2
ꢀ
ꢁ
−1
−
5.1
the V=O band was observed at 988 cm , red-shifted from
BO = 0.2912ln(21349/ν)
,
−
1
−1
1
029 cm (VZ_ED, Fig. 5c). The V=O band at 988 cm
where ν is the V–O Raman stretching frequency (cm 1), R is
−
corresponds to a bond strength of 1.79 vu. This suggests a
the corresponding bond length (Å), and BO is the Pauling bond
weakening of the V=O bond order from 1.89 to 1.79 vu, length-

J. Due-Hansen et al. / Journal of Catalysis 251 (2007) 459–473
469
nonsulfated zirconia-based catalysts have a much lower alkali
resistance [14], activity data for V2O5/ZrO2 catalysts are not
shown here.
The 1.7VSZ_AH and the 1.7VSZ_ED catalysts had very
similar surface acidities but different surface areas (71.9 and
2
1
02.9 m /g, respectively) (Table 1). Thus, the relative proxim-
ity of the Brønsted and the V=O redox sites seems to be higher
in the AH-prepared catalyst compared with the ED catalyst, ex-
plaining the somewhat higher activity of the former.
In all cases, doping the catalysts with potassium resulted in
an overall decrease in catalytic activity combined with a shift
in the maximum catalytic activity toward lower temperatures
(Fig. 8, top right). A possible explanation for such a temper-
ature shift is that the potassium loading reduced the activity
of the main NO-SCR reaction while the rate of the side reac-
tion of ammonia oxidation remained constant or even increased.
The sulfate-promoted catalyst based on zirconia precipitated
with ED displayed increased resistance to the alkali poisoning
◦
and was deactivated by only 58% at 350 C, whereas the cor-
responding AH-precipitated, reference, and KVWZ catalysts
were deactivated by 70%, 77%, and 88%, respectively. The cat-
alyst resistance toward potassium poisoning seems to correlate
well with the surface area and surface acidity (listed in Table 1),
with VSZ_ED and VSZ_AH having the highest surface acid-
ity and surface area after potassium doping. We discuss the
effect of potassium on the relative SCR activity of VSZ_AH
and VSZ_ED, leading to a higher activity of the latter, in more
detail later.
Fig. 8. The temperature dependency of the NO-SCR activity for the sulfate-
and tungstate-promoted catalysts impregnated with 1.7 wt% vanadium (top
left) and the corresponding potassium doped samples with a K/V molar ratio
of 0.4 (top right). The effect of potassium on the SCR activity in the tempera-
◦
ture range 250–500 C is shown in the bottom as the remaining activity, defined
by: 1 − [kfresh − kK-doped]/kfresh
.
Comparing the impact of K-doping on the SCR activities
of the sulfate- or tungsta-promoted catalysts, it is evident from
the remaining SCR activity versus temperature plot in Fig. 8
ened 0.02 Å (1.581 to 1.603 Å) when K-doping the V2O5/ZrO2
sample.
(
bottom) that the sulfated catalysts (especially the ED-prepared
3
.5. NO-SCR activity measurements
sample) had enhanced resistance to potassium. The resistance
to alkali doping and the vanadium surface coverage presented
in Table 1 suggest that low vanadium coverage, high surface
area, and high surface acidity are critical parameters for in-
creased resistance to alkali. KVSZ_ED, with a surface cover-
age of 2.1 VO /nm , displayed the greatest remaining activity
The catalytic activity of the vanadia-impregnated samples
promoted with tungsten oxide or sulfate were measured in the
temperature range of 25–500 C. The temperature dependency
of the SCR activity is depicted in Fig. 8 for the supported vana-
dia catalysts before (Fig. 8, top left) and after doping with
potassium oxide (Fig. 8, top right). The catalytic activity is
represented as the first-order rate constant k. The activity of a
◦
2
x
(Fig. 8, bottom) whereas the severely deactivated KVWZ_AH
2
sample had a surface coverage of 3.5 VOx/nm —almost half a
monolayer. For all of the promoted catalysts, the resistance to-
ward potassium poisoning seemed to follow the order VSZ_ED
commercial 3%V O –7%WO /TiO catalyst (VWT_ref) cal-
cined at 450 C is shown as a reference.
2
5
x
2
◦
2
2
(2.0 VO /nm ) > VSZ_AH (2.8 VO /nm ) > VWT_ref (2.9
The vanadia-impregnated catalysts promoted with sulfate
x x
2
2
VOx/nm ) > VWZ_AH (3.4 VOx/nm ).
(
1.7VSZ_AH and 1.7VSZ_ED) showed higher SCR activ-
Higher alkali resistance toward potassium of sulfate-promot-
ed catalysts also correlated with the relative strength of the
V=O bond of the surface vanadates, detected via in situ Raman
ity compared with the corresponding tungstated catalyst (1.7
VWZ_AH, Fig. 8, top left). This is in accordance with the sur-
face acidity given in Table 1, where the two sulfated catalysts
had a significant higher surface acidity (516–551 µmol/g) than
the tungstated catalyst (351 µmol/g), thus increasing the prob-
ability of formation of V–OH or Zr–OH Brønsted sites adjacent
to V=O redox sites.
spectroscopy. Alkali doping of V O /ZrO has been reported
2
5
2
to result in severe deactivation [14], whereas the corresponding
sulfate-promoted catalyst shows considerably better alkali re-
sistance. As discussed above, the V=O stretching frequency is
−
1
Previous studies of the effect of sulfation on the SCR activ-
ity of zirconia-based vanadia catalysts [14], also indicated that
sulfation had a significant effect on the catalyst performance,
increasing the activity up to 50% compared with that of the
red-shifted from 1029 to 988 cm when V2O5/ZrO2 is doped
with potassium, whereas the V=O band in the corresponding
−1
sulfated catalyst is red-shifted to only 1006 cm . The pro-
tecting effect of sulfation on the V=O redox site is thus ob-
vious.
nonsulfated V O /ZrO catalyst. Due to this and the fact that
2
5
2

470
J. Due-Hansen et al. / Journal of Catalysis 251 (2007) 459–473
Fig. 9. The three cationic and one neutral monomeric vanadium complexes included in the computational investigation.
3
.6. Computational studies
To investigate the possible structural changes responsible for
poisoning with potassium, the exact location of the band is not
crucial.
To probe the effect of potassium poisoning on the structural
features of the V=O moiety, a single potassium ion was in-
cluded, and two different positions were investigated for the
monomer: coordination to potassium with the double-bonded
oxygen (Fig. 9b) and coordination between two oxygen atoms
the differences observed in the in situ Raman spectra (Fig. 5)
on doping with potassium, we performed a modeling study us-
ing density functional theory (DFT) as described in Section 2.8.
The SCR reaction has been subject of several theoretical studies
(Fig. 9c). Finally, a deprotonated model structure in which the
[
50–53], and a recent quantum mechanical study of the entire
negative charge became delocalized between two equivalent
oxygen atoms was considered (Fig. 9d).
SCR reaction on a V2O9H8 cluster suggested that the active
catalyst consists of a two neighboring vanadium oxo sites [54].
4
+
Coordination of the potassium cation to the double-bonded
oxygen atom resulted in the expected elongation of the V=O
bond to 1.600 Å while simultaneously contracting the three V–
O single bonds by 0.20–0.25 Å (Fig. 9b). The increased bond
length for the V=O bond was accompanied by a red shift of
The Brønsted acidic V site protonates the incoming NH3
+
molecule to form NH , which is then susceptible to an at-
4
tack by an NO molecule to form an unstable NH3NHO adduct,
which subsequently decomposes into nitrogen and water. Here
we focus on the molecular basis for the detrimental effect ob-
served in the presence of potassium cations (“poisoning”). The
main observable change in the in situ Raman spectra on in-
troduction of potassium on the vanadium-based catalyst is the
red-shifting of the V=O stretching frequency by approximately
−
1
the V=O stretching of 32 cm , in good agreement with the
value derived from the experimental spectrum, where a red shift
−
1
−1
from 1029 cm (1.581 Å) to 988 cm (1.603 Å) was ob-
served when V2O5/ZrO2 was doped with potassium (Fig. 5).
When the potassium ion was instead coordinated by two oxygen
atoms as in model c (Fig. 9), then the calculated change in the
−1
4
0 cm , and the primary objective of our computational study
was to gain insight into this phenomenon. The structural effect
of alkali metals in general (Li, Na, K, and Rb) have previ-
ously been studied by Witko et al. [55] using a relatively large
V10O31H12 cluster, but in the current study we attempted to also
elucidate the changes in the vibrational spectrum.
−1
V=O stretching frequency was decreased by 21 cm , and the
concomitant increase in bond length was significantly smaller
(0.015 Å vs 0.028 Å for b). The energy for the structural iso-
mers b and c can be compared directly; with the methods used
here, we found them to be virtually equienergetic (with model
c favored by only 2 kJ/mol in Gibbs free energy).
The structures were all assumed to be singlets (closed-shell
species), and only fully oxidized vanadium species were con-
sidered (V5 ). In a related study by van Dam et al. [56], a series
of model systems of increasing size were investigated using
similar methods, and it was found that the structure of the vana-
dium oxide unit did not depend significantly on the size of the
computational model system. Thus, in the present work we lim-
ited ourselves to relatively simple model systems in which the
carrier surface is not explicitly included.
+
On deprotonation, there is a strong tendency to form a struc-
ture in which the potassium atom is coordinated between two
equivalent oxygen atoms, leaving only two anchoring points to
the carrier (Fig. 9d). The two V–O bonds now have identical
+
lengths of 1.641 Å, considerably longer than for the other K -
doped structures, indicating additional weakening of the former
V=O bond, which is also observable in the large shift in fre-
−
1
Initially, we considered the isolated V=O moiety with three
hydroxyl moieties. To emulate the fixation to the bulk ZrO2
surface, the capping hydrogen atoms were oriented in the op-
posite direction of the V=O bond and given a mass of 1000
amu [57]. With this model system, the equilibrium structure
has a V=O bond length of 1.572 Å with a stretching frequency
quency of 137 cm . A shift of this magnitude is more dramatic
than that observed in the experimental spectra, indicating that
structures of this type are not dominating under experimental
conditions. We calculated atom-centered point charges on V, O,
and K using both Mulliken and ESP schemes; these are given
in Table 4. On introduction of potassium (cf. Figs. 9a and 9b),
the polarization of the V=O bond was increased, resulting in
an overall weakening of the bond (Vδ+–Oδ−). There was little
difference between the two coordination modes for the cationic
vanadium complexes (Figs. 9b and 9c); however, the interaction
between potassium and the oxygen atoms was stronger for the
deprotonated complex (Fig. 9d), leading to the largest decrease
in the positive charge on potassium.
−1
of 1101 cm , and three identical V–O bonds with a length
of 1.780 Å (Fig. 9a). Compared with the experimental in situ
Raman spectra, the calculated vibrational frequency appears at
a slightly higher wavenumber, which could be related to the
fact that under experimental conditions, a fraction of the vana-
dia atoms are in a lower oxidation state. But because we are
concerned mainly with the change in stretching frequency on

J. Due-Hansen et al. / Journal of Catalysis 251 (2007) 459–473
471
Table 4
Table 5
Calculated Mulliken and ESP atomic charges for V, O, and K in monomeric
vanadium complexes listed in Figs. 9a–9d
Calculated Mulliken and ESP atomic charges for V, O, and K in dimeric vana-
dium complexes listed in Figs. 10a–10d
Model
system
Mulliken
charge
ESP
charge
(V)
Mulliken
charge
(=O)
−0.36
−0.70
−0.69
−0.65
ESP
Mulliken
charge
(K )
ESP
Model
system
Mulliken
charge
(V)
ESP
charge
(V)
Mulliken
charge
(=O)
−0.35
−0.51
−0.36
−0.47
ESP
Mulliken
charge
(K )
ESP
charge
(=O)
−0.35
−0.64
−0.64
−0.72
charge
charge
(=O)
−0.38
−0.54
−0.42
−0.53
charge
+
+
+
+
(V)
(K )
(K )
9a
9b
9c
9d
+1.22
+1.33
+1.33
+1.17
+1.20
+1.23
+1.25
+1.28
–
–
10a
10b
10c
10d
+1.28
+1.38
+1.32
+1.40
+1.32
+1.38
+1.34
+1.38
–
–
+0.94
+0.92
+0.87
+0.98
+0.98
+0.94
+0.90
+0.97
–
–
+0.91
+0.96
Fig. 11. Effect of potassium on the equilibrium between monomeric and
dimeric vanadia complexes.
Fig. 10. The two forms of dimeric vanadia-oxo moieties used to examine the in-
fluence of potassium poisoning. Above: The monooxo-bridged vanadia dimer.
Below: The dioxo-bridged vanadia dimer.
With the energies for both monomeric and dimeric vanadium
complexes available, it is possible to examine the condensa-
tion (dimerization) equilibrium as illustrated in Fig. 11. In the
absence of potassium ions, the enthalpy of formation of the
dimer with a mono-oxygen bridge was almost thermoneutral
It is well known that the vanadium oxo moieties can con-
dense to form polymers under the reaction conditions with re-
sulting loss of catalytic activity. To investigate this, we chose
the monooxo-bridged (a, Fig. 10) and dioxo-bridged dimers
(
ꢀH = 1 kJ/mol); however, the liberation of water at high tem-
peratures still may have allowed such formation to a certain
extent. In the presence of potassium, formation of the dimer
was clearly favored in enthalpy (ꢀH = −37 kJ/mol), which
(c, Fig. 10) as models. For both undoped dimers, the equilib-
rium V=O distance was approximately 1.570 Å, which in both
readily explains the increase of the V–O–V bands at around
+
cases was elongated by 0.023 Å on complexation with a K ion
−1
8
00 cm observed experimentally in Figs. 5 and 6 with K-
(
b and d, Fig. 10). These calculated increases in bond lengths
doping of V2O5/ZrO2. This also correlates well with the find-
ings of Chen and Yang [9], who measured the SCR activity
of V2O5/TiO2 versus the K/V molar ratio and found that the
catalyst deactivated almost completely at K/V ≈ 0.5, because
one potassium unit was coordinated to two VOx units. For the
dioxo-bridged dimer in Fig. 9, the dimerization was very unfa-
vorable (ꢀH = 135 kJ/mol), and there was only a negligible
effect of potassium on the enthalpy of dimerization. Presum-
ably, this was due to the additional strain in the dioxo-bridged
species, which forced the two coordinating oxygen atoms fur-
ther away from the potassium by about 0.2 Å, thus preventing
efficient coordination.
The position of the monomeric–dimeric equilibrium of the
vanadate species in the presence of potassium can provide a
possible explanation for the observed potassium-induced de-
activation of catalysts with relatively high surface coverage.
Comparing the vanadium surface coverage and deactivation
on potassium doping of the two iso-acidic sulfated catalysts,
VSZ_AH and VSZ_ED (Table 1) illustrates an interesting
point. The main difference between the samples is the surface
area and, consequently, the acid site concentration and vana-
correlate well with that derived from the in situ Raman spectra
(
0.02 Å; spectra c and d in Fig. 5).
The increased strain in the dioxo dimer in Fig. 10c resulted
−
1
in a slightly higher V=O stretching frequency (1114 cm vs
−1
1
100 cm for a in Fig. 10), and in both cases a red shift
of the V=O stretching frequency on complexation with the
−
1
−1
;
potassium ion to ∼1070 cm was observed (b: 1074 cm
−1
d: 1072 cm ). Note that the calculated red shift for the
−
1
monooxo dimer of 36 cm is in excellent agreement with the
experimental value, whereas the calculated shift for the dioxo
dimer appears to be too high. The calculated Mulliken and
ESP charges for the dimeric vanadium complexes are shown
in Table 5. The coordination of potassium also resulted in po-
larization of the V=O bond, although to a lesser extent than
that for the monomers, as demonstrated by the partial neg-
ative charge on oxygen. The magnitude of this polarization
was slightly larger for the monooxo-bridged dimer than for the
dioxo-bridged dimer (cf. Figs. 10b and 10d), possibly due to the
larger distance between the oxygen atoms and potassium in the
more strained dioxo-bridged dimer.

472
J. Due-Hansen et al. / Journal of Catalysis 251 (2007) 459–473
dium coverage. From the computational studies, we found that
dimerization (resulting in loss of active sites) of the monomeric
vanadate units was favored in enthalpy when potassium was
present. Considering the impact of surface coverage, the dimer-
of potassium-induced dimerization with resulting loss of active
sites.
In the computational investigation, we analyzed the struc-
tural effects of potassium poisoning in detail and found good
qualitative correlation with the in situ Raman measurements.
This supports the proposition that a coordination interaction
ization of vanadate monomers should be preferred for the cat-
2
alyst with the highest coverage [i.e., VSZ_AH (2.8 VO /nm )
x
2
+
over VSZ_ED (2.0 VOx/nm )], which correlates well with the
between K and two neighboring V=O is responsible for the
observed catalytic measurements. The relatively high loss of
acidity of the VSZ_AH catalyst by K-doping is thus under-
standable, because dimerization occurs by destruction of two
observable decrease in V=O stretching frequency, and also pro-
vides a rationale for the observed higher proportion of dimers
and higher polymers in the presence of potassium cations. The
+
V–OH Brønsted sites (liberating a H O molecule).
scenario proposed here of K joining two VOx units under the
2
liberation of H2O agrees well with the experimental observa-
tions of almost complete deactivation of the V2O5/TiO2 SCR
catalyst by potassium salts at a K/V molar ratio of ∼0.5.
These results provide new insight into the previously re-
ported improved resistance of sulfate- and tungstate-promoted
SCR catalysts to poisoning with potassium compared with tra-
ditional V2O5/TiO2 catalyst. In the case of acidic supports, the
potassium preferentially interacts with promoter surface species
4
. Conclusion
The characterization of vanadia-based SCR catalysts, pro-
moted by sulfate or tungsten oxide supported on zirconia pre-
pared from precipitated hydrous zirconia, showed that the cat-
alyst structure and SCR activity were affected by the synthe-
sis route, especially by the choice of the Zr(OH) -precipitant.
Using ethylene diamine in the precipitation of hydrous zirco-
nia resulted in overall higher surface areas and a higher frac-
tion of crystalline tetragonal ZrO compared with using am-
monium hydroxide as the precipitant. Conversely, TGA–MS
data showed that sulfate-impregnated samples obtained after
precipitation with ammonium hydroxide formed more stable
sulfate species, decomposing to SO at around 700–1000 C,
4
(
sulfate or tungstate), leaving a part of the vanadium-based
sites, which are responsible for the catalytic activity, largely un-
affected.
We believe that the detailed information reported herein may
prove valuable in the development of new catalysts with im-
proved resistance toward alkali poisoning.
2
◦
2
Acknowledgments
whereas the corresponding ethylene diamine-precipitated sam-
ples started liberating SO2 already at 500 C.
◦
The Center for Sustainable and Green Chemistry is spon-
sored by the Danish National Research Foundation. This work
was supported by Energinet.dk (project PSO-FU5201). The
authors thank MAX-lab, Sweden, for facilitating the high-
temperature XRD measurements.
Doping the vanadia-based SCR catalysts with potassium to
a K/V molar ratio of 0.4 was proved by in situ Raman spec-
troscopy to affect the stretching frequency (i.e., strength and
length) of the V=O bond in surface vanadates. This indicates
that potassium altered the active vanadium surface species in
the SCR reaction, lengthening the V=O bond. Catalyst pro-
motion by sulfate resulted in a lower degree of V=O bond
lengthening on potassium doping, which can account for the
higher resistance to alkali doping of sulfate-promoted vanadia-
Supporting information
XYZ coordinates, SCF energies and additional thermody-
namic data are available for all structures discussed in the man-
uscript.
based SCR catalysts. This suggests that the active VO units are
x
less affected by potassium if the catalyst is sulfated (i.e., have
strong surface acid sites). Here the acidic support functioned
as a host for the potassium, which was bound preferentially to
the strongest acid sites, the sulfated zirconia, over the catalytic
active vanadate sites.
The effect of potassium on the V=O bond could not be
studied in the in situ Raman spectra of the tungstate-promoted
catalysts, where the Raman bands due to formation of crys-
Please visit DOI: 10.1016/j.jcat.2007.07.016.
References
[
[
[
[
1] G. Busca, L. Lietti, G. Ramis, F. Berti, Appl. Catal. B 18 (1998) 1.
2] V.I. Pârvulescu, P. Grange, B. Delmon, Catal. Today 46 (1998) 233.
3] N.-Y. Topsøe, H. Topsøe, J.A. Dumesic, J. Catal. 151 (1995) 226.
4] N.-Y. Topsøe, J.A. Dumesic, H. Topsøe, J. Catal. 151 (1995) 241;
N.-Y. Topsøe, Science 265 (1994) 1217.
talline WO obscured the V=O stretching frequency range.
[5] I.E. Wachs, G. Deo, B.M. Weckhuysen, A. Andreini, M.A. Vuurman,
M. de Boer, M.D. Amiridis, J. Catal. 161 (1996) 211.
3
The degree of deactivation due to potassium doping seems
to be proportional to the vanadium coverage for the samples
in focus. The highest alkali resistance decreased in the or-
[
6] J.A. Dumesic, N. Topsøe, H. Topsøe, Y. Chen, T. Slabiak, J. Catal. 163
1996) 409.
7] M. Anstrom, J.A. Dumesic, N.-Y. Topsøe, Catal. Lett. 78 (2002) 281.
(
[
2
2
der VSZ_ED (2.0 VO /nm ) > VSZ_AH (2.8 VO /nm ) >
[8] Y. Zheng, A.D. Jensen, J.E. Johnsson, Appl. Catal. B 60 (2005) 253.
[9] J.P. Chen, R.T. Yang, J. Catal. 125 (1990) 411.
10] H. Kamata, K. Takahashi, C.U.I. Odenbrand, J. Mol. Catal. A 139 (1999)
x
x
2
2
VWT_ref (2.9 VOx/nm ) > VWZ_AH (3.4 VOx/nm ). This
correlates well with the results from the computational studies,
which predicted that the dimerization of vanadate monomers
would be favored in enthalpy if potassium were present. In other
words; low surface vanadium coverage reduces the possibility
[
189.
[
[
11] I.E. Wachs, B.M. Weckhuysen, Appl. Catal. A 137 (1997) 67.
12] D. Bulushev, F. Rainone, L. Kiwi-Minsker, A. Renken, Langmuir 17
(2001) 5276.

J. Due-Hansen et al. / Journal of Catalysis 251 (2007) 459–473
473
[
13] R. Khodayari, C.U.I. Odenbrand, Appl. Catal. B 30 (2001) 87.
[35] X. Yang, F.C. Jentoft, R.E. Jentoft, F. Girgsdies, T. Ressler, Catal. Lett. 81
(2002) 25.
[14] A.L. Kustov, M.Y. Kustova, R. Fehrmann, P. Simonsen, Appl. Catal. B 58
(
2005) 97;
A.L. Kustov, S.B. Rasmussen, P. Simonsen, R. Fehrmann, Appl. Catal. B
2007), doi:10.1016/j.apcatb.2007.05.004.
[36] N. Yi, Y. Cao, W. Feng, W. Dai, K. Fan, Catal. Lett. 99 (2005) 73.
[37] K. Arata, M. Hino, Mater. Chem. Phys. 26 (1990) 213.
[38] J.P. Chen, R.T. Yang, J. Catal. 139 (1993) 277.
(
[
15] J. Due-Hansen, A.L. Kustov, S.B. Rasmussen, R. Fehrmann, C.H. Chris-
[39] J.P. Dunn, J.-M. Jehng, D.S. Kim, L.E. Briand, H.G. Stenger, I.E. Wachs,
J. Phys. Chem. B 102 (1998) 6212.
tensen, Appl. Catal. B 66 (2006) 160.
[
[
16] K. Bourikas, Ch. Fountzoula, Ch. Kordulis, Appl. Catal. B 52 (2004) 145.
17] I. Giakoumelou, Ch. Fountzoula, Ch. Kordulis, S. Boghosian, J. Catal. 239
[40] A. Christodoulakis, M. Machli, A.A. Lemonidou, S. Boghosian, J. Catal.
222 (2004) 293.
(
2006) 1.
18] J.G. Santiesteban, J.C. Vartuli, S. Han, R.D. Bastian, C.D. Chang, J. Catal.
68 (1997) 431.
[41] G. Ramis, G. Busca, F. Bregani, Catal. Lett. 18 (1993) 299.
[42] D.S. Kim, M. Ostromecki, I.E. Wachs, J. Mol. Catal. A 106 (1996) 93.
[43] F. Figueras, J. Palomeque, S. Loridant, C. Feche, N. Essayem, G. Gelbard,
J. Catal. 226 (2004) 25.
[44] M. Scheithauer, R.K. Grasselli, H. Knözinger, Langmuir 14 (1998) 3019.
[45] M. Boulova, G. Lucazeau, J. Solid State Chem. 167 (2002) 425.
[46] M.A. Vuurman, I.E. Wachs, A.M. Hirt, J. Phys. Chem. 95 (1991) 9928.
[47] I.E. Wachs, Catal. Today 100 (2005) 79.
[
[
1
19] L. D’Souza, A. Suchopar, K. Zhu, D. Balyozova, M. Devadas, R.M.
Richards, Microporous Mesoporous Mater. 88 (2006) 22.
20] J.R. Sohn, M.Y. Park, Langmuir 14 (1998) 6140.
21] M.G. Cutrufello, U. Diebold, R.D. Gonzalez, Catal. Lett. 101 (2005) 5.
22] A. Christodoulakis, S. Boghosian, J. Catal. 215 (2003) 139.
23] C. Lee, W. Yang, R.G. Parr, Phys. Rev. B 37 (1988) 785.
24] A.D. Becke, J. Chem. Phys. 98 (1993) 5648.
[
[
[
[
[
[
[
[
[
[
[
[
[48] F.D. Hardcastle, I.E. Wachs, J. Phys. Chem. 95 (1991) 5031.
[49] A. Christodoulakis, E. Heracleous, A.A. Lemonidou, S. Boghosian, J. Ca-
tal. 242 (2006) 16.
25] A.D. Becke, J. Chem. Phys. 98 (1993) 1372.
26] Jaguar, version 6.5, release 106, Schrodinger, LLC, New York, NY, 2005.
27] P.J. Hay, R.W. Wadt, J. Chem. Phys. 82 (1985) 299.
28] R. Ditchfield, W.J. Hehre, J.A. Pople, J. Chem. Phys. 54 (1971) 724.
29] R.S. Mulliken, J. Chem. Phys. 23 (1955) 1833.
30] L.E. Chirlian, M.M. Francl, J. Comput. Chem. 8 (1987) 894.
31] R.J. Woods, M. Khalil, W. Pell, S.H. Moffat, V.H. Smith Jr., J. Comput.
Chem. 11 (1990) 297.
[50] F. Gilardoni, J. Weber, A. Baiker, J. Phys. Chem. A 101 (1997) 6069.
[51] M. Anstrom, N.-Y. Topsøe, J.A. Dumesic, J. Catal. 213 (2003) 115.
[52] K. Jug, T. Homann, T. Bredow, J. Phys. Chem. A 108 (2004) 2966.
[53] A. Vittadini, M. Casarin, A. Selloni, J. Phys. Chem. B 109 (2005) 1652.
[54] S. Soyer, A. Uzun, S. Senkan, I. Onal, Catal. Today 118 (2006) 268.
[55] M. Witko, R. Grybos, R. Tokarz-Sobieraj, Top. Catal. 38 (2006) 105.
[56] H.J.J. van Dam, M.F. Guest, P. Sherwood, J.M.H. Thomas, J.H. van
Lenthe, J.N.J. van Lingen, C.L. Bailey, I.J. Bush, J. Mol. Struct.
(THEOCHEM) 771 (2006) 33.
[32] C.M. Breneman, K.B. Wiberg, J. Comput. Chem. 11 (1990) 361.
[33] A.P. Scott, L. Radom, J. Phys. Chem. 100 (1996) 16502.
[34] H. Tian, E.I. Ross, I.E. Wachs, J. Phys. Chem. B 110 (2006) 9593.
[57] V.I. Avdeev, G.M. Zhidomirov, J. Struct. Chem. 46 (2005) 577.