WOx/TiO2 catalysts prepared by grafting of tungsten alkoxides: Morphological properties and catalytic behavior in the selective reduction of NO by NH3

Article abstract of DOI:10.1006/jcat.1996.0087

Tungsten oxide on titania catalysts with loadings of 3.5-11 μmol(W) m^-2_BET were prepared in a multiple grafting procedure using tungsten (VI)-oxo-methanolate and tungsten(V)-ethanolate as precursors. The efficiency of the deposition depended markedly on the combination of alkoxide precursor and solvent. Calcination at elevated temperature, 1023 K compared to 573 K normally used, resulted in the formation of additional Bronsted acid sites as indicated by TPD measurements. In line with this finding the activity for selective catalytic reduction of NO with NH₃ increased by almost 100% upon calcination at 1023 K and nitrogen remained the only product observed. The structural changes induced upon tungsten deposition and calcination in dry O₂/Ar at different temperatures were followed using XRD and laser Raman spectroscopy. The formation of a stabilized polytungstate layer, indicated by a prominent Raman band at 985 cm^-1, was found to be favored by increasing WO_x loading and calcination temperature. As a minor species, paracrystalline WO₃ was found in samples with loadings higher than 9 μmol(W) m^-2 after high temperature (1023 K) calcination in moist air. TPR investigations in combination with AAS indicated a stepwise reduction of W(VI), with W(IV) being stable up to ca. 1150 K.

Full text of DOI:10.1006/jcat.1996.0087

JOURNAL OF CATALYSIS 159, 259–269 (1996)
ARTICLE NO. 0087
WOx/TiO2 Catalysts Prepared by Grafting of Tungsten Alkoxides:
Morphological Properties and Catalytic Behavior in the
Selective Reduction of NO by NH3
J. Engweiler, J. Harf, and A. Baiker1
Department of Chemical Engineering and Industrial Chemistry, Swiss Federal Institute of Technology, ETH-Zentrum, CH-8092 Zu¨rich, Switzerland
Received September 22, 1994; revised November 25, 1995; accepted November 30, 1995
suppressed. Raman (7, 8) and diffuse reflectance FTIR
spectroscopic investigations (9) showed that tungsten oxide
did not interact with the vanadia species when both WOx
and VOx were present together on TiO2. More profound
investigations on the WOx/TiO2 system and on its catalytic
behavior in SCR are necessary for further understanding
its role in composite catalytic systems.
Tungsten oxide on titania catalysts with loadings of 3.5–11
ꢀmol(W) mꢀ2BET were prepared in a multiple grafting procedure
using tungsten (VI)–oxo–methanolate and tungsten(V)–ethanolate
as precursors. The efficiency of the deposition depended markedly
on the combination of alkoxide precursor and solvent. Calcination
at elevated temperature, 1023 K compared to 573 K normally used,
resulted in the formation of additional Brønsted acid sites as in-
dicated by TPD measurements. In line with this finding the activ-
ity for selective catalytic reduction of NO with NH3 increased by
almost 100% upon calcination at 1023 K and nitrogen remained
the only product observed. The structural changes induced upon
tungsten deposition and calcination in dry O2/Ar at different tem-
peratures were followed using XRD and laser Raman spectroscopy.
The formation of a stabilized polytungstate layer, indicated by a
prominent Raman band at 985 cmꢀ1, was found to be favored by
increasing WOx loading and calcination temperature. As a minor
species, paracrystalline WO3 was found in samples with loadings
higher than 9 ꢀmol(W) mꢀ2 after high temperature (1023 K) cal-
cination in moist air. TPR investigations in combination with AAS
indicated a stepwise reduction of W(VI), with W(IV) being stable
The morphological properties as well as the acidity of
WOx/TiO2 catalysts have been investigated for submono-
layer and monolayer samples. The structure of the sup-
ported WOx species has been described mainly by Raman
spectroscopy (8, 10, 11) and EXAFS (12). Tungsten oxide
on titania was reported to form isolated tetrahedrally co-
ordinated surface species as well as two-dimensional layers
of octahedrally coordinated species at low concentrations
(8, 12). Upon dehydration all tungsten oxide species were
converted into a distorted octahedrally coordinated struc-
ture (8, 12). No crystalline WO3 was detected with XRD at
surface coverages up to 10 ꢀmol(W) mꢀ2BET (13), but based
on Raman spectroscopy the formation of a paracrystalline
WO3 phase at >6 ꢀmol(W) mꢀ2 was stated to occur (8,
11). Recently, the acidic properties of tungsten oxide/titania
catalysts have been studied by several authors using vibra-
tional spectroscopy (10, 14–16). It was found that Brønsted
acid centers are formed upon deposition of WOx on TiO2
and that strong Lewis acid centers are present as well.
Different preparation methods have been reported for
WOx/TiO2 catalysts. Impregnation methods with aqueous
solutions of tungstate salts are widely used (3, 9, 10) and
Hilbrig et al. (12, 16) reported the formation of WOx layers
on TiO2 and Al2O3 by solid–solid wetting. To our know-
ledge the preparation of tungsten oxide/titania catalysts by
grafting of tungsten alkoxides has not been reported so far.
Grafting of vanadium alkoxides on titania and mixed silica–
titania support materials resulted in highly dispersed vana-
dia species active for SCR and allowed good control of the
vanadia loading in the submonolayer region (17, 18, 19).
The aim of this work was to gain information on the in-
fluence of preparation parameters on both the structural
c
up to ca. 1150 K. ꢁ 1996 Academic Press, Inc.
INTRODUCTION
Supported tungsten oxide catalysts are used for hy-
drotreating and for alkene metathesis reactions because of
their solid acid properties (1) and they are active for the
selective catalytic reduction (SCR) of nitric oxide with am-
monia (2, 3, 4).
Commercial SCR catalysts are mostly based on vana-
dia/titania and often contain considerable amounts of tung-
sten oxide and/or molybdenum oxide (5). Chen and Yang
(6)describedtheroleoftungstenoxideinvanadia–tungsten
oxide/titania catalysts. Tungsten oxide broadens the tem-
perature range for SCR and better resistance toward poi-
soning by alkali salts and arsenic is observed. Higher
temperatures can be used because ammonia oxidation is
1
To whom correspondence should be addressed.
259
0021-9517/96 $18.00
c
Copyright ꢁ 1996 by Academic Press, Inc.
All rights of reproduction in any form reserved.

260
ENGWEILER, HARF, AND BAIKER
TABLE 1
Morphological Properties of WOx/TiO2 Catalysts Calcined at Different Temperatures (O2/Ar, 7.2%)
Tcalcination
(K)
SBET
W loading
W loading
H2 consumed in
a
b
Sample
(m2 gꢀ1
)
(ꢀmol gꢀ1
)
(ꢀmol mꢀ2
)
TPR (ꢀmol gꢀ1
)
XRD: Ia Irꢀ1c
BET
TiO2(573)
TiO2(1023)
573
1023
47
13
0
0
0
0
—
—
2.28
0.00
Wm3.5(573)
Wm3.5(1023)
573
1023
47
32
155
155
3.3
4.8
171
—
—
1.65
Wm7(573)
Wm7(1023)
573
1023
45
41.5
314
314
7.0
7.6
323
—
—
2.00
Wm9(573)
Wm9(1023)
573
1023
42.5
40
384
384
9.0
9.6
406
351
—
2.12
Wm11(573)
Wm11(698)
Wm11(823)
Wm11(948)
Wm11(1023)
573
698
823
948
1023
41.5
41
40.5
39
453
453
453
453
453
10.9
11.0
11.2
11.6
11.9
484
—
—
—
—
2.25
—
2.26
2.26
2.25d
38
We3.5(1023)
We4.5(1023)
We5.5(823)
We5.5(948)
We5.5(1023)
1023
1023
823
948
1023
29
38
44
40.5
39
103
172
229
229
229
3.6
4.5
5.2
5.7
5.9
95
170
220
—
—
—
—
—
—
210
a The maximum error is estimated to ꢂ3%.
b Integrated signal from 550 to ca. 1140 K.
c XRD: Ia Irꢀ1: ratio of peak intensities (areas) of anatase (101) and rutile (110).
d Crystalline WO3 was detected by XRD only after calcination at 1023 K for 3 h in moist air (Fig. 2, Trace d).
properties of the grafted WOx/TiO2 catalysts and on their transferred into a dried glass reactor and degassed at 398 K
behavior in SCR of NO by NH3. The structural changes for 2 h under vacuum (10 Pa) to obtain a mildly dehydrated
induced upon tungsten deposition and calcination at differ- surface. Then 0.3 mmol tungsten precursor per gram TiO2,
ent temperatures were followed by means of XRD, TPR, dissolved in 3 ml solvent, were injected into the reactor at
and laser Raman spectroscopy. Temperature-programmed room temperature. The temperature was raised to 323 K
desorption after exposure of the samples to SCR reaction for 12 h. Afterward, the solution was removed, the sample
conditions (SCR-TPD) was used to investigate the changes was washed three times with fresh solvent, and it was
in the adsorption characteristics of the titania-based cata- dried in flowing nitrogen at 353 K for 3 h. Subsequently a
lysts resulting from the tungsten oxide deposition.
stream of 200 ml minꢀ1 N2, saturated with water at room
temperature, was passed through the sample bed at 573
K to hydrolyze the anchored alkoxides and to remove
the organic ligands from the catalyst. Calcination was
performed in pure O2 at 573 K for 3 h.
EXPERIMENTAL
Catalyst Preparation
The catalysts are referred to as “WpL(T )” where Wp
specifies the alkoxide precursor used for the grafting (Wm
for WO(OCH3)4, We for W(OC2H5)5), L gives the approx-
imate loading of tungsten in ꢀmol mꢀ2BET for the catalysts
after standard calcination at 573 K in oxygen, and (T ) in-
dicates the calcination temperature in K. Note that upon
calcination of one sample (Wm3.5) at 1023 K sintering of
the support material occurred leading to lower surface area
(Table 1) and consequently to higher tungsten loading per
The WOx /TiO2 catalysts were prepared by a multi-step
grafting procedure similar to the one described for VOx/
TiO2 (18). Titania (P25, Degussa) was stirred with doubly
distilled water, then successively dried in a rotary evapo-
rator for 12 h at 323 K/100 mbar and in an oven for 12 h
at 423 K. The TiO2 agglomerate was crushed and sieved
to particles with a diameter of 0.3–0.5 mm to give the
desired support material. Two different precursors were
used for the grafting procedure. A tungsten (VI)–oxo–
methanolate, WO(OCH3)4 (Wm), was prepared according
to the procedure described in Ref. (20), and commercial
W(OC2H5)5 (We) was obtained from Gelest. Wm and
We were dissolved in tetrahydrofurane and n-hexane
(both FLUKA, p.a.), respectively. The following grafting
procedure was then used. The preconditioned support was
SBET
.
Catalyst Characterization
Nitrogen physisorption. The specific surface areas,
SBET, and the mean pore diameters, hdpi, were determined

WOx GRAFTED ON TITANIA
261
from nitrogen physisorption at 77 K using a Micromeritics Catalytic Tests
ASAP 2000 instrument. Prior to the measurements sam-
Tests for the selective catalytic reduction (SCR) or NO
by NH3 were carried out in a continuous flow fixed-bed
microreactor made of a quartz glass tube with 4 mm inner
diameter. Volumes of 0.126 cm3, corresponding to 95–120
mg of catalyst granules, were used for the measurements.
Prior to SCR activity tests the catalysts were pretreated
in situ in flowing O2/Ar (7%) at the temperature speci-
fied (573–1023 K). The reaction gas mixture consisted of
900 ppm NO, 900 ppm NH3, and 1.8% O2 in an argon bal-
ance. This gas mixture, which is referred to as the SCR
feed, was mixed from argon (99.999%) and single compo-
nent gases in an argon balance (3600 ppm NO/Ar, NH3/Ar
certified by ꢂ2%, Union Carbide). Feed and product con-
centrations of NO, NO2, NH3, H2O, N2O, O2, and N2 were
quantitatively analyzed on line using a computer controlled
Balzers quadrupole mass spectrometer QMA 112A.
ples were degassed to 0.1 Pa at 423 K. BET surface areas
were calculated in a relative pressure range of 0.05–0.2 as-
suming a cross-sectional area of 0.162 nm2 for the nitrogen
molecule. The pore size distributions were calculated from
thedesorptionbranchesoftheisotherms(21)usingtheBJH
method (22).
X-ray diffraction, XRD. To check the coexistence of
different crystal phases X-ray powder diffraction patterns
were measured on a Siemens ꢁ/ꢁ D5000 powder X-ray
diffractometer. The diffractograms were recorded with
CuKꢂ radiation over a 2ꢁ range of 20ꢃ to 44ꢃ. A scintilla-
tion counter with monochromator was used. Ia/Ir, the ratio
of the intensities of the {101} reflection of anatase (23) and
the {110} reflection of rutile (24), respectively, was used as
a measure of the anatase fraction of the TiO2 phases.
Conversion measurements as a function of temperature
were carried out at a gas hourly space velocity, GHSV, of
ca. 24,000 hꢀ1 (flow rate per total bed volume; STP; εbed
ꢄ 0.5) after steady-state SCR activity had been established
at 423 K (typically after 2 h). The activity parameters, km
and kS, and the apparent activation energies, Ea, were calcu-
lated assuming first order kinetics in NO and zero order in
NH3 and O2 (3). km values were calculated according to the
performance equation of an isothermal plug-flow reactor,
Laser Raman spectroscopy, LRS. For the Raman spec-
troscopy 0.1 g of calcined catalyst was ground and the sam-
ples were kept in a glass cuvette. Spectra were recorded
under ambient conditions. The spectra were excited using
the 1064 nm line of a NdYAG laser and recorded with
a Perkin–Elmer 2000 NIR-FT-Raman spectrometer. The
laser power was 50–150 mW.
Temperature-programmed reduction, TPR. For TPR
measurements 1.5% H2 in He at 250 mlSTP minꢀ1 was passed
through a sample bed containing the equivalent of ca. 0.2
mmol W. Temperature was raised from ambient to 1350 K
at 10 K minꢀ1. The H2 consumption was determined from
the concentration measured by mass spectrometry. The ap-
paratus used and further details of the measurements are
given in Refs. (25, 26).
km = Fmꢀ1 ꢅ ln[(1 ꢀ XNO
)
ꢀ1],
where F is the feed flow rate in m3 sꢀ1 and m is the catalyst
mass in kg. The rate constants of the NO conversion will be
presented for specific temperatures as kS on a surface area
ꢀ1
basis: kS = km (SBET
)
.
The criterion of Weisz and Prater (28) was checked for
the most active catalyst, Wm11(1023), under the reaction
conditions of the SCR tests. Any influence of internal mass
transfer could be ruled out for this sample and all the other
catalysts showing virtually the same pore size distribution.
For all measurements, a nitrogen balance including feed
and product stream concentrations of all N-containing com-
pounds was calculated. Even for high conversions of NO
and NH3, the error in the N-balance did not exceed ꢂ5%.
Temperature-programmed desorption, TPD. SCR-
TPD experiments were performed in situ with the same
samples as used for the catalytic tests. SCR feed gas (50 ml
minꢀ1) was passed through the catalyst bed at 303 K for 1
h after the complete SCR testing (12 h, see below). After
purging for 1 h with argon (99.999%, Pan Gas, 20 ml minꢀ1
)
the temperature was raised at 10 K minꢀ1 up to 723 K and
the concentrations of the evolving species H2O, NH3, NO,
N2, and N2O were monitored by mass spectrometry (MS).
To assess the influences of mass transfer and readsorption
on the TPD spectra the criteria of Demmin and Gorte were
applied (27). For the working conditions used, intraparticle
concentration gradients seem to be negligible as the ratio
of the carrier-gas flow rate to the rate of diffusion was
calculated to be 0.07 (>0.05 is given as a limit for negligible
gradients). In contrast, readsorption phenomena cannot be
ruled out in TPD measurements carried out at atmospheric
pressure. The diffusion rate of the desorbed species as
well as the carrier gas flow rates are much lower than the
adsorption rate of, e.g., NH3, even under optimum flow
TPD conditions.
RESULTS AND DISCUSSION
Properties of Calcined Catalysts
The standard preparation procedure included calcina-
tion at 573 K for 3 h in a stream of dry oxygen. Prior
to the catalytic tests the samples were calcined in situ in
O2/Ar (7.2%) at a temperature ranging from 573 to 1023 K.
Table 1 summarizes the morphological properties of the
in situ calcined tungsten oxide on titania catalysts.
Temperature-programmed oxidation experiments (condi-
tions: 50 ml minꢀ1 of 7% O2/Ar, heating rate = 10 K minꢀ1
)

262
ENGWEILER, HARF, AND BAIKER
ing up to 9 ꢀmol(W) mꢀ2
by solid–solid wetting. Let
BET
us keep in mind for the further discussion of the morpho-
logical properties and the activity results that the loading
of the catalysts Wm11 exceeds the capacity of an experi-
mentally observed monolayer (11, 12) and their W surface
concentrations correspond approximately to a theoretical
monolayer. Sample Wm9 contains ca. the experimentally
observed monolayer of WOx on TiO2, whereas all the other
samples, including the catalysts of the WeL series, contain
lesstungstenthanwouldcorrespondtoamonolayerofWOx
on TiO2.
Sintering and crystallinity. Calcination of the pure tita-
nia support at 1023 K resulted in complete anatase to rutile
transformation and in a marked decrease of the BET sur-
face area from 47 to 13 m2gꢀ1 (Table 1). Upon deposition of
increasing amounts of tungsten, the extent of phase trans-
formation as well as sintering was suppressed, resulting in
a less pronounced loss in surface area for Wm11 (8%) and
an unchanged ratio of the two titania phases upon in situ
calcination at 1023 K (cf. Ia/Ir, Table 1). As already stated
by Ramis et al. (10) tungsten deposition on TiO2 stabilizes
the morphology of the titania. It inhibits sintering as well
as the anatase to rutile transformation and therefore leads
to a higher thermal stability of the catalysts.
FIG. 1. Influence of number of grafting steps on tungsten loading for
the two series of WOx/TiO2 catalysts prepared by multiple step grafting
with different precursors: WmL, ᭺ WeL.
performed with samples after preparation (dried and cal-
cined) indicated that the organic ligands had been removed
to >99% by the calcination in dry oxygen.
Tungsten deposition. The stepwise deposition of tung-
sten on the titania support by the multiple grafting proce-
dure is illustrated in Fig. 1. The first and the second grafting
steps with the methanolate precursor (WO(OCH3)4, Wm)
resulted in a deposition of (ꢆ160 ꢀmol(W) gꢀ1 each,
whereas the following two grafting steps increased the load-
ing only by ꢆ70 ꢀmol gꢀ1. A slight decrease in total surface
area, SBET, occurred with the multiple grafting steps for the
samples calcined at 573 K (47 to 41.5 m2gꢀ1). Using the
ethanolate precursor (W(OC2H5)5, We) the first deposition
was again the most effective, giving ꢆ105 ꢀmol(W)gꢀ1. The
subsequent two steps both resulted in additional ꢆ65 ꢀmol
The XRD patterns of the catalyst Wm11, depicted in
Fig. 2, show well-developed anatase and rutile crystallites.
g
ꢀ1. The WOx deposition is obviously more effective when
using WO(OCH3)4 in THF than W(OC2H5)5 in n-hexane.
Two effects possibly contribute to the higher efficiency of
the Wm grafting procedure: different wetting abilities of the
titania support by the solvent and a steric effect due to the
different size of the alkoxide ligands. THF results in better
wetting of the support material because it is more protic
than n-hexane.
Note that a saturation effect is observed at a loading cor-
responding to 7–9 ꢀmol mꢀ2BET for the preparation of the
Wm catalyst series, which is less hindered by both steric ef-
fects and incomplete wetting. A theoretical monolayer cov-
erage, deduced from a virtual spreading of a single layer of
the WO3 structure on the support material, would corres-
pond to 11.2 ꢀmol(W) mꢀ2 (11). Bond et al. (11) found
an experimental monolayer capacity based on XPS mea-
surements at 6.0 ꢀmol(W) mꢀ2 for P25-based samples. This
is slightly more than a ratio W : Tis of 0.5 (Tis = Ti4 + ions
exposed to the surface; 6.25 nmꢀ2 for {100} anatase sur-
face (29) and matches double anchoring on surface OH
groups on titania (ca. 6 nmꢀ2 (30)). Hilbrig et al. (12, 16)
prepared fully dispersed tungsten oxide on titania contain-
FIG. 2. XRD analysis of the WOx/TiO2 catalyst with the highest tung-
sten loading, Wm11, calcined (a) in O2 at 573 K for 3 h; (b) in O2/Ar (7.2%)
at 1023 K for 3 h; (c) in O2/Ar (7.2%) at 1023 K for 3 h, followed by ex-
posure to SCR reaction conditions at ca. 600 K for 24 h; (d) in moist air
at 1023 K for 3 h. Reflections due to crystalline phases are marked: ᭡
anatase, ♦ rutile, ଙ WO3.

WOx GRAFTED ON TITANIA
263
Crystalline WO3 was not detected after calcination up to
1023 K with all catalysts. XRD analysis indicated a crys-
talline tungsten oxide phase only after calcination in moist
air at 1023 K (cf. Fig. 2). The same picture emerges from the
LRS measurements presented in Fig. 3. The characteristic
peaks for WO3 observed at 807, 715, 324, 293, and 270 cmꢀ1
(31) are clearly present for the Wm11 sample only after cal-
cination in moist air. Chan et al. (32) have shown that the
ratio of the relatꢀiv1e Raman intensities of the major band
for WO3 (807 cm ) and of the surface tungsten oxide com-
pound (at 985 cmꢀ1, cf. discussion of the surface tungsten
species, below) is 160 : 1. An estimation on this basis indi-
cates an amount of ca. 1% paracrystalline WO3 of the total
tungsten content.
Hilbrig et al. (12, 16) reported on the spreading of WO3
on TiO2 in O2 at 723 K and found that this process took
place only in the presence of moisture, whereas the spread-
ing of vanadia on both TiO2 and ꢃ -Al2O3 took place under
dry conditions as well, but was accelerated by the presence
of water vapor (33). Although the microscopic mechanism
of the solid–solid wetting is still not known in detail, the
phenomenon is attributed to a minimization of the surface
free energy. The influence of water is related to the possi-
ble formation of oxyhydroxides as the mobile phase of the
layered oxide (33). The formation of crystalline WO3 at el-
evated temperature in the presence of water vapor (under
SCR reaction conditions and upon calcination in moist air)
FIG. 3. Raman spectra of the grafted WOx/TiO2 catalyst with the
is therefore attributed to the inverse process of the spread- highest tungsten loading, Wm11, recorded under ambient conditions. Dif-
ferent calcination procedures were applied: (a) in O2 at 573 K for 3 h; (b)
ing phenomenon. Still the minimization of surface free en-
ergy is the driving force of the process. Tungsten that is not
in O2/Ar (7.2%) at 1023 K for 3 h; (c) in moist air at 1023 K for 3 h. Most
intense peak of anatase at 635 cmꢀ1 was set 100%. For better visualization
stabilized in a surface layer agglomerates and forms crys-
tallites of WO3 as soon as its mobility is increased by ele-
vated temperature and by the presence of water vapor. We
conclude that a highly dispersed tungsten oxide overlayer
was produced by the multiple grafting procedure which ex-
ceeds the monolayer capacity in the case of Wm11. Upon
annealing under moist conditions the thermodynamically
most stable proportion of surface tungsten oxide species
and paracrystalline WO3 is established.
of scattering due to WOx species the region of 750–1100 cmꢀ1 is enlarged
(original magnification factor: 5).
440 and 612 cmꢀ1, respectively, and a second order feature
of TiO2 occurs at 783 cmꢀ1 (35).
The Raman spectra after calcination at 573 K in dry oxy-
gen, after calcination under dry conditions (in situ, O2/Ar
7.2 %) at 1023 K, and after calcination at 1023 K in moist air
are presented in Fig. 3. As discussed above (see sintering
Raman investigations. To identify structural changes of and crystallinity) the tungsten oxide exceeding the mono-
the supported tungsten oxide upon calcination under dif- layer capacity of the titania support forms crystalline WO3
ferent conditions, Raman spectra of the Wm catalyst series in the presence of moisture at elevated temperature. The
were recorded. For better visualization of the scattering layered oxide on the sample calcined under moist condi-
caused by the grafted tungsten oxide a magnification of tions is therefore regarded as a thermodynamically equi-
the 1100–750 cmꢀ1 region is shown in the figures (orig- librated monolayer and its Raman spectrum shows a sin-
inal magnification factor of 5) together with normalized gle maximum at 985 cmꢀ1 with some tailing toward lower
spectra in the 1100–220 cmꢀ1 range (most intense peak wavenumbers. Compared to the spectrum recorded after
of anatase at 635 cmꢀ1 was set 100%). Observable peaks dry calcination at 1023 K the shoulder at ꢆ930 cmꢀ1 is less
attributed to tungsten (surface) species are expected in prominent.
the 740–1060 cmꢀ1 region for W O W modes and W O
– –
The Raman spectra after calcination at 573 K are de-
==
stretching modes (34) and crystalline WO3 shows sharp picted in Fig. 4A for the Wm catalyst series. The spectrum
bands at 807 and 715 cmꢀ1 and in the range 240–350 cmꢀ1 of Wm3.5(573) shows a band at ꢆ940 cmꢀ1 together with
(31). The bands due to the anatase and rutile modifications the second order feature of TiO2 at 783 cmꢀ1 (35). With
of the support are detected at 635, 515, and 400 cmꢀ1 and at increased loading the band shifts upward to 985 cmꢀ1 for

264
ENGWEILER, HARF, AND BAIKER
FIG. 4. Raman spectra of the WmL catalyst series with different WOx loading recorded under ambient conditions: (A) samples after calcination
at 573 K: (a) TiO2, (b) Wm3.5, (c) Wm7, (d) Wm9, (e) Wm11; (B) after additional in situ calcination for 3 h at 1023 K in O2/Ar(7.2%): (a) Wm3.5, (b)
Wm7, (c) Wm9, (d) Wm11. Assignments as in Fig. 3.
Wm9(573) and it is centred at 970 cmꢀ1 for Wm11(573). increasing tungsten loading and the above assignments are
The difficulties to assign Raman bands of surface tungsten supported by other authors (8, 11). Note that in the studies
species between 915 and 980 cmꢀ1 have been discussed by cited, the samples were calcined at 723 K.
Horsley et al. (34). They suggest a characteristic frequency
The changes occurring upon calcination at 1023 K un-
range of 740–980 cmꢀ1 for octahedrally coordinated WOx der dry conditions become apparent when the spectra in
species and bands between 913 and 1060 cmꢀ1 to be indica- Figs. 4A and 4B are compared. The band observed at
tive for tetrahedral species. It has been reported by Deo 940 cmꢀ1 for the single loaded catalyst (Wm3.5) after cal-
and Wachs (36) that the molecular structure of various sup- cination at 573 K is shifted to 965 cmꢀ1. A similar shift to
ported transition metal oxides follows structures observed higher wavenumbers occurs for the doubly loaded sample
in aqueous medium and depends on the pH at the point (from 960 cmꢀ1 broadened to 960–980 cmꢀ1). The three to
of zero charge of the support. For WOx on TiO2 they pre- fourfold grafted samples show a sharp band at 985 cmꢀ1
dicted the presence of species with structures like WO4(aq) after in situ calcination at 1023 K. For Wm9(1023) a super-
position with a broad band centred around 900 cmꢀ1 is vis-
(at ꢆ930 cmꢀ1) and W12O39(aq) (at ꢆ960 cmꢀ1), the latter
ible, whereas for Wm11(1023) a shoulder around 930 cmꢀ1
being the major species at monolayer coverage. The band
is formed. The presence of the Raman band at 985 cmꢀ1 in-
dicates the formation of a two-dimensional surface layer of
polytungstate species, as also observed after the moist cal-
cination of Wm11 at 1023 K (cf. Fig. 3). The position of this
band coincides with the one reported for surface tungstate
species produced by solid–solid wetting (33).
observed in the 930 to 990 cmꢀ1 region will therefore be
interpreted as an overlap of two characteristic bands. At
low tungsten loadings mainly isolated fourfold-coordinated
tungsten species are formed (band at ꢆ935 cmꢀ1 dominat-
ing the spectrum). With increased loading, polytungstate
species with an octahedral environment become predomi-
nant (band at 985 cmꢀ1). The formation of a polytungstate
layer appears to be complete for Wm9(573) (maximum
band wavenumber). Upon a further grafting step surplus
tungsten is deposited and the band is shifted downward
Temperature-Programmed Reduction, TPR
The TPR profiles of the Wm series are presented in Fig. 5
to 970 cmꢀ1. The shift to higher wavelength observed with for samples calcined in O2 at 573 K. A maximum hydrogen

WOx GRAFTED ON TITANIA
265
imately to the total tungsten content determined by AAS
(cf. Table 1), assuming a reduction of W(VI) to W(IV). This
is a strong indication that W(IV) species are stable on the
titania surface up to ca. 1100 K. Further reduction to lower
oxidation states of W occurs at higher temperature, as indi-
cated by the TPR profile. Vermaire and van Berge (37) used
TPR (10% H2/Ar, 10 K minꢀ1) to investigate the reduction
behavior of WO3/TiO2 (P25) samples prepared by ion ex-
change. The stepwise reduction followed the stoichiometry
observed in this study. The authors proposed W(IV) to be
stabilized on TiO2 (anatase and rutile) because of the simi-
larity in the ion radius of W4 + and Ti4 + , the W–O and Ti–O
bond lengths, and the crystal structures of WO2 and titania.
The influence of calcination at 1023 K on the reduc-
tion behavior was checked for sample Wm9 (not shown
in Fig. 5). The low-temperature peak shifted by ca. 30 K to
higher temperature and decreased in intensity, indicating a
lower amount of easily reducible WOx species, which fur-
ther points to a stabilization of the WOx surface layer upon
calcination at elevated temperature.
Temperature-Programmed Desorption, TPD
Recently we have shown that SCR-TPD is a suitable
method to describe the changes in Brønsted and Lewis acid-
ity as well as the oxidizing abilities of vanadia on titania
(anatase) (38, 39).
FIG. 5. Temperature-programmed reduction profiles for the Wm cat-
alyst series. The samples were previously calcined in situ for 3 h in O2 at
Influence of tungsten loading. For the Wm catalysts the
desorption profiles after exposure of SCR feed gas at 303 K
are shown in Fig. 6. The samples were calcined at 1023 K and
the TPD was run after the complete SCR activity testing.
The desorption rates of both water (left) and ammonia
(right) are presented on a surface area (SBET) based scale.
Neither N2, O2, NO, nor N2O was detected in the argon
carrier gas stream during TPD, indicating that no other ad-
sorbed species apart from ammonia and water are present
in significant amounts under the conditions applied. On the
573 K. TPR in flowing H2/He (1.5%), heating rate: 10 K minꢀ1
.
consumption is observed at 1050–1065 K and a second peak
appears at 750–820 K for all samples, whereas the small
peak observed for Wm11 at 920 K corresponds to a shoul-
der for Wm9, and is not resolved for the one- and two fold
grafted Wm samples. It emerges that the reduction is not
complete at 1150 K. The hydrogen consumed in the range
ꢆ500 K ꢇ T ꢇ 1150 K corresponds for the samples approx-
FIG. 6. Temperature-programmed desorption of H2O (left) and NH3 (right) after exposure to SCR feed at 303 K (SCR-TPD) for the catalysts of
the WmL series, calcined at 1023 K and used for catalytic tests for 12 h. Numbers on curves indicate tungsten loading. The concentrations are referred
to the total surface area (BET) of the samples as given in Table 1. TPD in flowing Ar (20 ml minꢀ1, STP), sample bed volume: 0.126 ml, heating rate:
10 K minꢀ1
.

266
ENGWEILER, HARF, AND BAIKER
FIG. 7. Profiles of temperature-programmed desorption of H2O (left) and NH3 (right) for the triply loaded Wm7 catalysts (SCR-TPD). Same
experimental conditions as in Fig. 6, but the catalysts were calcined in situ at different temperatures (indicated on curves).
otherhand, itshowsthatnooxidationofadsorbedammonia effects the preferential formation of protonic ammonia co-
species occurred during the TPD experiments, which im- ordination sites at the expense of centers showing strong
plies that all water evolved in TPD originates from molec- NH3 coordination and likely to be of Lewis type.
ularly adsorbed H2O or from a recombination of surface
Influence of calcination temperature. Desorption pro-
hydroxyl groups.
files of the threefold loaded catalyst Wm9 calcined in situ
The desorption profile of H2O shows a peak maximum at
at different temperatures are shown in Fig. 7. After cal-
405 K for Wm3.5(1023) and Wm7(1023) with a long tailing
cination at 573 K the traces of ammonia and water show
to the highest temperature recorded (700 K). The maxi-
rather constant desorption rates at ca. 400–700 K. Weak
mum intensity of the signal decreases slightly with higher
tungsten loading (Wm7(1023)). Upon further loading of
the WOx/TiO2 catalysts the peak exhibits a broad max-
imum at 390–470 K (Wm9(1023)). Only half of the esti-
TABLE 2
Selective Catalytic Reduction of NO by NH3
mated total amount of water desorbing from the pure ti-
tania samples TiO2(573) (ca. 3.6 ꢀmol mꢀ2) or TiO2(1023)
Tcalcination kS ꢅ 108 (585 K)
Ea
Conversiona
TX (K)
(ca. 3.4 ꢀmol mꢀ2) is calculated for the Wm catalysts (1.5–
1.75 ꢀmol mꢀ2; 300–700 K) with Wm9(1023) showing the
lowest total amount of evolved water. A high degree of hy-
droxylation can be expected, because most of the desorbing
water is supposed to originate from the recombination of
surface OH groups (especially at T ꢈ 450 K). The evolution
of ammonia shows a more distinct change with increasing
tungsten concentration. A very broad feature is observed
at 400–700 K (maximum at ca. 600 K) for Wm3.5(1023).
Upon increasing the tungsten loading the high-temperature
ammonia evolution decreases in intensity and a steadily in-
creasing maximum NH3 concentration is observed at ca.
450 K. The estimation of the total amount of ammonia
desorbed during the TPD experiments ranges from 2.35
ꢀmol mꢀ2 (Wm7(1023)) to 3.25 ꢀmol mꢀ2 (Wm11(1023)).
This corresponds to approximately twice the amount of wa-
ter detected. WO3/TiO2 catalysts are known to show both
Brønsted and Lewis acidity (10, 16). An increasing den-
sity of acidic protons was described for increasing tungsten
loading by Hilbrig et al. (16). Ramis et al. (10) found that
coordinates ammonia is thermally more stable than proto-
nated ammonia. In line with these findings we assign the
ammonia desorbing at 400–500 K to be mainly bound to
Brønsted acid centers. Tungsten deposition on titania at
half to full monolayer coverage (ca. 5–9 ꢀmol mꢀ2) thus
Sample
(K)
(m3sꢀ1 mꢀ2
)
(kJ molꢀ1
)
BET
=0.5
NO
Wm3.5(573)
Wm3.5(1023)
573
1023
1.5
12.9
149
111
647
616
Wm7(573)
Wm7(1023)
573
1023
11.3
26
149
109
603
588
Wm9(573)
Wm9(1023)
573
1023
23
54
143
121
589
571
Wm11(573)
Wm11(698)
Wm11(823)
Wm11(948)
Wm11(1023)
573
698
823
948
1023
37
37
49
58
67
124
124
121
117
115
575
577
571
567
563
We3.5(1023)
We4.5(1023)
1023
1023
2.4
120
107
669
619
10.6
We5.5(823)
We5.5(948)
We5.5(1023)
We5.5(1023)
823
948
1023
1023
9.3
128
111
104
105
613
615
599
599
10.3
20.5
19.7
We5.5(1023)b
1023
12.7
117
608
Note. The kinetic data are given as surface based rate constants, kS,
determined at 585 K from the analysis of integral activity measurements,
and as apparent activation energies (Ea). The temperature required for
50% NO conversion (TX =0.5) reflects a measure of the overall activity.
a Determined at constNaOnt GHSV of 24,000 hꢀ1
b Feed containing ca. 3000 ppm H2O.
.

WOx GRAFTED ON TITANIA
267
Catalytic Behavior
In Table 2 the results of the SCR activity measurements
are summarized. For all measurements the selectivity to N2
was 100% and a 1 : 1 stoichiometry was observed for the
NH3–NO reaction. This indicates that no ammonia oxida-
tion occurred in the temperature range investigated (480–
700 K) (3). The overall activity is best represented by the
temperature to obtain 50% NO conversion, TX = 0.5. Cat-
NO
alyst Wm11(1023) converts 50% NO at the lowest temper-
ature, 563 K, whereas 575 K is needed after calcination at
573 K (Wm11(573)). A similar dependence on the calci-
nation temperature emerges for all catalysts. Comparing
all samples calcined at 1023 K, TX
decreases with in-
NO=0.5
creasing tungsten loading for both catalyst series prepared
(Wm and We), which indicates the increasing activity with
higher tungsten concentration.
NO reaction rates are used to illustrate the above men-
FIG. 8. Dependence of the SCR activity on the calcination temper-
ature kS, the rate constant of NO reaction referred to the surface area tioned dependences of the activity on the tungsten loading
(BET), is plotted versus the calcination temperature (calcination in situ,
O2/Ar 7.2%) for the two samples with highest tungsten loadings of both
preparation series, Wm11 and We5.5.
as well as on the calcination temperature. Because the spe-
cific surface areas (BET) of the samples varied between 32
and 47 m2gꢀ1 the NO reaction rate is referred to the to-
ꢀ1
tal surface area (BET) and represented as kS = km SBET
,
instead of km, the conventional rate constant, which is re-
ferred to the catalyst mass.
maximum desorptions are visible around 400 K for water
and at 660 K for the NH3 evolution. Such broad desorption
features are generally the result of a wide distribution of
adsorption sites of different strength. Upon in situ calcina-
tion at 1023 K ammonia desorbs with a distinct maximum
at 430–470 K and the amount of more strongly bound NH3
decreases. A less pronounced change is seen in the desorp-
tion behavior of water with a maximum at 390–470 K. The
increased low-temperature desorption of NH3 upon calci-
nation at elevated temperature indicates the formation of
additional Brønsted acid sites.
In Fig. 8, kS (at 585 K) is plotted versus the calcination
temperatureforthetwosampleswithhighesttungstenload-
ing of both preparation series, Wm11 and We5.5. The NO
reaction rate is doubled when Wm11 is calcined at 1023 K
instead of 573 K, and the same behavior is observed for
We5.5 upon increasing the calcination temperature from
823 to 1023 K. Figure 9 depicts the dependence of kS on
the catalyst loading. In the left panel (A) the Wm samples
are represented after calcination at 573 K and the right
panel (B) shows the reaction rates for the samples of both
FIG. 9. Dependence of the SCR activity on the tungsten loading of the catalysts. (A) shows kS values for the Wm samples after calcination at 573 K
and (B) the reaction rates for the samples of both series, calcined at 1023 K.

268
ENGWEILER, HARF, AND BAIKER
series, calcined at 1023 K. After calcination at 573 K very occurred during TPD. This reflects the 1 : 1 stoichiometry
low activity is determined for W loading <7 ꢀmol mꢀ2 and for NH3 and NO observed under SCR reaction conditions
moderate activity is found at 7–11 ꢀmol mꢀ2. Note that (absence of ammonia oxidation).
in situ calcination at 1023 K on the other hand leads to a
The SCR activity of the WOx/TiO2 catalysts is strongly
linear dependence of the NO reaction rate on the surface influenced by the tungsten loading and the calcination tem-
concentration of tungsten. At low loadings, however, very perature. Brønsted acid sites were found to be important
low SCR activity is measured. This hints toward uniform for the SCR reaction over WOx/TiO2 catalysts. The surface
intrinsic SCR activity of the tungsten surface species above density of protonic acid sites increases with the tungsten
a basic loading of ca. 3 ꢀmol mꢀ2. When 9A and 9B are loading and with calcination at elevated temperature.
compared the increase in activity with higher calcination
temperature emerges.
Addition of ca. 3000 ppm H2O to the SCR feed leads
ACKNOWLEDGMENTS
to a decrease in activity, as listed in Table 2 for We5.5 At
585 K the NO reaction rate was lowered to ca. 2/3 of the
value determined for reaction under dry conditions. If we
consider a reaction mechanism involving the reaction of
adsorbed NH3 with NO, either adsorbed or from the gas
phase (3), this would imply a H2O desorption step. At an
increased partial pressure of water the desorption of H2O
may influence the reaction rate because it determines the
site density for ammonia activation. Thus we may explain
the decrease in activity observed by concurrent adsorption
of NH3 and H2O on active sites.
Recapitulating the changes in activity and desorption be-
havior of the grafted WOx/TiO2 catalysts, the increase of
both Brønsted acidity and SCR activity at increasing tung-
sten loading and calcination at elevated temperatures has
to be stressed. The results illustrate the important role of
protonic acid centers in the SCR of NO with NH3 over
WOx/TiO2 catalysts.
Thanks are due to H. Viebrock for the preparation of the Wm pre-
cursor. Financial support by the “Nationaler Energie-Forschungs-Fond”
(NEFF) and by the Swiss National Science Foundation (NFP 24) is kindly
acknowledged.
REFERENCES
1. Tanabe, K., Misono, M., Ono, Y., and Hattori, H., Stud. Surf. Sci. Catal.
51, 5 (1989).
2. Bosch, H., and Janssen, F., Catal. Today 2, 369 (1988).
3. Imanari, M., and Watanabe, Y., Stud. Surf. Sci. Catal. 7, 841 (1981).
4. Morikawa, S., Takahashi, K., Mogi, J., and Kurita, S., Bull. Chem. Soc.
Jpn. 55, 2254 (1982).
5. Sloss, L. L. et al., “Nitrogen Oxide Control Technology Fact Book,”
Noyes Data Corporation, Park Ridge, 1992.
6. Chen, J. P. and Yang, R. T., Appl. Catal. 80, 135 (1992).
7. Deo, G., and Wachs, I. E., J. Catal. 146, 335 (1994).
8. Vuurman, M. A., Wachs, I. E., and Hirt, A. M., J. Phys. Chem. 95, 9928
(1991).
9. Ramis, G., Busca, G., and Bregani, F., Catal. Lett. 18, 299 (1993).
10. Ramis, G., Busca, G., Cristiani, C., Lietti, L., Forzatti, P., and Bregani,
F., Langmuir 8, 1744 (1992).
11. Bond, G., Flamerz, S., and van Wijk, L., Catal. Today 1, 229 (1987).
12. Hilbrig, F., Go¨bel, H. E., Kno¨zinger, H., Schmelz, H., and Lengeler,
B., J. Phys. Chem. 95, 6973 (1991).
13. Cristiani, C., Bellotto, M., Forzatti, P., and Bregani, F., J. Mater. Res. 8,
2019 (1993).
14. Lietti, L., Svachula, J., Forzatti, P., Busca, G., Ramis, G., and Bregani,
F., Catal. Today 17, 131 (1993).
15. Patrono, P., La Ginestra, A., Ramis, G., and Busca, G., Appl. Catal. A
107, 249 (1994).
16. Hilbrig, F., Schmelz, H., and Kno¨zinger, H., Stud. Surf. Sci. Catal. B
75, 1351 (1993).
17. Bond, G. C., and Tahir, S. F., Appl. Catal. 71, 1 (1991).
18. Baiker, A., Dollenmeier, P., and Glinski, M., Appl. Catal. 35, 351
(1987).
19. Handy, B. E., Baiker, A., Schraml-Marth, M., and Wokaun, A., J. Catal.
133, 1 (1992).
20. Kucheiko, S., Turova, N. Ya., Kozlova, N. I., and Zhadanov, B. V., Sov.
J. Coord. Chem. 12, 867 (1986).
21. Broekoff, J. P., In “Preparation of Catalysts, II” (B. Delmon, P.
Grange, P. Jacobs, and G. Poncelet, Eds.), p. 663. Elsevier, Amsterdam,
1979.
22. Barrett, E. P., Joyner, L. G., and Halenda, P. P., J. Am. Chem. Soc. 73,
373 (1951).
23. JCPDS Mineral Powder Diffraction Data File 21-1272, Park Lane,
Pennsylvania.
CONCLUSIONS
Grafting of tungsten alkoxides onto titania allows the im-
mobilization of tungsten oxide species in highly dispersed
form at loadings up to a monolayer coverage, which is esti-
mated to be ca. 9 ꢀmol (W) mꢀ2BET. Even higher loadings
(11 ꢀmol mꢀ2) are stabilized after calcination at 573 K, but
the surplus tungsten content is converted into crystalline
WO3 when the sample is treated under moist conditions
at elevated temperature (1023 K). Tungsten deposition on
TiO2 leads to a higher thermal stability of the catalysts due
to a concomitant suppression of sintering as well as rutiliza-
tion of the titania support.
The Raman results indicate the formation of a tungsten
oxide layer upon calcination at 1023 K, which is character-
ized by a band at 985 cmꢀ1 under ambient conditions.
SCR-TPD investigations revealed that both calcination
at elevated temperature (1023 K) and increasing of the
tungsten loading increased the ratio of Brønsted-to-Lewis
acid site density of the catalysts. As no N2, O2, NO, or N2O
are observed during TPD we conclude that no NOx or O-
containing reaction intermediates are adsorbed under the
conditions applied and that no oxidation of adsorbed NH3
24. JCPDS Mineral Powder Diffraction Data File 21-1276, Park Lane,
Pennsylvania.

WOx GRAFTED ON TITANIA
269
25. Koeppel, R. A., Nickl, J., and Baiker, A., Catal. Today 20, 45 (1994).
32. Chan, S. S., Wachs, I. E., and Murrell, L. L., J. Catal. 90, 150 (1984).
26. Hurst, N. W., Gentry, S. J., Jones, A., and McNicol, B. D., Catal. Rev. 33. Leyrer, J., Margraf, R., Taglauer, E., and Kno¨zinger, H., Surf. Sci. 201,
Sci. Eng. 24, 233 (1982).
603 (1988).
27. Demmin, R. A., and Gorte, R. J., J. Catal. 90, 32 (1984).
28. Weisz, P. B., and Prater, C. D., Adv. Catal. 6, 143 (1954).
34. Horsley, J. A., Wachs, I. E., Brown, J. M., Via, G. H., and Hardcastle,
F. D., J. Phys. Chem. 91, 4014 (1987).
29. Bond, G., Zurita, J. P., Flamerz, S., Gellings, P. J., Bosch, H., van Om- 35. Beattie, I. R., and Gilson, T. R., J. Chem. Soc. A 2322 (1969).
men, J., and Kip, B. J., Appl. Catal. 22, 361 (1986). 36. Deo, G., and Wachs, I. E., J. Phys. Chem. 95, 5889 (1991).
30. Iwasawa, Y., in “Tailored Metal Catalysts” (D. Reidel Ed.), P. I. Reidel, 37. Vermaire, D. C., and van Berge, P. C., J. Catal. 116, 309 (1989).
Dordrecht, 1986.
38. Schneider, H., Tschudin, S., Schneider, M., Wokaun, A., and Baiker,
A., J. Catal. 147, 5 (1994).
39. Engweiler, J., and Baiker, A., Appl. Catal. A 120, 187 (1994).
31. Daniel, M. F., Desbat, B., Lassegues, J., Gerand, B., and Figlarz, M., J.
Solid State Chem. 67, 235 (1987).