Journal of Catalysis 239 (2006) 200211

www.elsevier.com/locate/jcat

Synthesis, characterization, and catalytic function of novel highly dispersed

tungsten oxide catalysts on mesoporous silica
Jose E. Herrera a , Ja Hun Kwak a , Jian Zhi Hu a , Yong Wang a , Charles H.F. Peden a, ,
Josef Macht b , Enrique Iglesia b
a Institute for Interfacial Catalysis, Pacific Northwest National Laboratory, PO Box 999, MS K8-98, Richland, WA 99352, USA
b Department of Chemical Engineering, University of California, 103 Gilman Hall, Berkeley, CA 94720-1462, USA

Received 13 September 2005; revised 24 January 2006; accepted 25 January 2006

Available online 28 February 2006

Abstract
The physical and chemical properties of tungsten oxide supported on SBA-15 mesoporous silica prepared by a controlled grafting process
through atomic layer deposition (ALD) were studied using complementary characterization methods. X-ray diffraction, optical absorption, and
transmission electron microscopy showed that tungsten oxide species are highly dispersed on SBA-15 surfaces, even at 30 wt% WOx content
(surface density, 1.33 WOx /nm2 ). ALD methods led to samples with much better thermal stability than those prepared via impregnation. Dehy-
dration reactions of 2-butanol and methanol dehydration were used as probe reactions. Differences in reaction rates between the samples prepared
by ALD and conventional impregnation may reflect the sintering resistance of catalysts prepared by ALD. Notably, temperature-programmed
oxidation of spent catalysts showed that carbon formation was not responsible for the different dehydration rates in samples prepared by ALD and
impregnation.
2006 Elsevier Inc. All rights reserved.

Keywords: Tungsten oxide; SBA15; Methanol dehydration; 2-Butanol dehydration; DRS UVvis; TEM; 1 H NMR; XRD; TPO

1. Introduction Transition metal oxides with hexavalent and pentavalent

cations have shown the highest activity for the aforementioned
Supported metal oxides are an important family of cata- transformations. For some of these metal oxides, catalytic
lysts used in the petrochemical and refining industry. Catalytic rates appear to be linked to the formation of strong Brnsted
reactions, such as ethylene polymerization, alkane oxidative acid sites, which are stabilized by the effective delocalization
dehydrogenation, alkene metathesis, and selective catalytic re- of electron density in the conjugated base [1113]. Among
duction of NOx with ammonia, occur on supported metal ox- the transition metals with these characteristics, tungsten oxide
ides [17]. Many studies have addressed the structure and func- gives the strongest Brnsted acid sites, either as bulk or sup-
tion of these catalytic materials, but the origins of their high ported oxides [1418], but their structure and catalytic proper-
activity remain the subject of active study. There is general con- ties are strongly influenced by the support. Several studies have
sensus that active sites are of an acidic nature [8,9], but some explored how interactions between support and tungsten oxide
studies suggest the possible involvement of redox cycles in- domains affect catalytic activity. Indeed, even for a given WOx
volving the reduction of metal oxide centers and radical-like support system, WOx domains exhibit a range of surface struc-
intermediates [10]. tures that depend strongly on synthetic protocols, such as the
WOx precursor and content, as well as on the thermal history
of these materials [1923].
Several reports have addressed the characterization of metal
* Corresponding author. oxide-supported tungsten oxide species [16,2427], and the
E-mail address: [email protected] (C.H.F. Peden). specific nature of tungsten supported on ZrO2 as acid cata-
0021-9517/$ see front matter 2006 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcat.2006.01.034
 J.E. Herrera et al. / Journal of Catalysis 239 (2006) 200211 201

lysts for o-xylene isomerization [1113] and butanol dehydra- Samples with 25% (0.33 WOx /nm2 ), 50% (0.66 WOx /nm2 ),
tion [28,29]. There are very few reports describing the use of 75% (1.00 WOx /nm2 ), and 100% (1.33 WOx /nm2 ) monolayer
mesoporous silica as a support for tungsten oxide species, how- coverage were prepared. As-synthesized WOx /SBA-15 silica
ever [30,31]. Zhang et al. described the synthesis of tungsten- samples were treated in flowing dry air (Airgas, zero grade,
containing MCM-41 with good dispersion, but segregated 1.7 cm3 s1 g1
cat ) at 673 and 773 K for 2 h. An additional sam-
crystalline WO3 was detected even after mild thermal treat- ple with 200% (2.67 WOx /nm2 ) monolayer coverage was pre-
ments [32]. Briot et al. attempted to avoid crystalline phases by pared by repeating the ALD process described above on a 1.33-
using oxoperoxometalate precursors that form at low pH in the WOx /nm2 sample after calcination. Two other samples were
presence of H2 O2 . The poor stability of these materials was ev- prepared by incipient wetness impregnation (RI) of SBA-15
ident from the extensive leaching of the tungsten species [33]. with aqueous solutions of ammonium metatungstate (Aldrich,
Here we report a novel method for grafting tungsten ox- 99.99%) to give 1.00 and 1.33 WOx /nm2 surface densities.
ide species onto mesoporous silica (SBA-15) surfaces. This These samples were also treated for 3 h at 673 K in flowing
atomic layer deposition (ALD) method uses organic solvents dry air. After calcination, all catalysts obtained by ALD were
and anhydrous conditions to control the deposition of tungsten analyzed by Fourier transform infrared spectroscopy to ensure
oxide species onto the silica surface and to avoid the forma- that no organic residues remained on the samples.
tion of WOx oligomers prevalent in aqueous solutions at nearly
neutral pH. The resulting improvements in dispersion and ther- 2.2. XRD
mal stability of tungsten oxide species were confirmed using
UVvis light diffuse reflectance spectroscopy (DRS), X-ray XRD data were collected with a Philips PW3040/00 XPert
diffraction (XRD), transmission electron microscopy (TEM), MPD system equipped with a Cu source ( = 1.5406 ), a ver-
and magic-angle spinning (MAS) 1 H nuclear magnetic reso- tical goniometer (220 mm radius), and focusing optics
nance (NMR). Methanol and 2-butanol dehydration were used (BraggBrentano geometry). The in situ experiments were per-
as probe reactions. Temperature-programmed oxidation (TPO) formed using a heated stage equipped with an Anton Paar
and UVvis spectra of spent samples were used to explore cat- HTK 1200 oven. The samples were mounted within an alu-
alyst deactivation mechanisms and the effects of synthesis pro- mina cavity-type holder (18 mm diameter) for analysis. Dif-
tocols on catalytic function. fraction data were analyzed using JADE (Materials Data, Liv-
ermore, CA) and the Powder Diffraction File database (2003
2. Experimental Release, International Center for Diffraction Data, Newtown
Square, PA).
2.1. Catalyst synthesis
2.3. TEM
Mesoporous SBA-15 silica was prepared using a previously
reported protocol [34]. The resulting BET surface area after TEM micrographs were obtained using a JEOL 2010 high-
treatment in air at 773 K for 4 h was 860 m2 /g, and the resolution analytical electron microscope operating at 200 kV.
average pore size was 7 nm using the N2 adsorption method. An energy-dispersive X-ray (EDX) analyzer was used for el-
This sample was suspended in anhydrous toluene (Aldrich, emental analysis. Samples were dispersed using 2-propanol
99.8%) and refluxed for 3 h in a N2 atmosphere to remove ph- (Aldrich, 99.5%) onto a carbon-coated copper grid.
ysisorbed water. Tungsten precursor solutions were prepared
by dissolving a given amount of WCl6 (Aldrich, 99.9%) in 2.4. 1 H NMR
150 cm3 of toluene at ambient temperature and then adding
20 cm3 ethanol (Aldrich, anhydrous). The solution was then Solid-state 1 H NMR spectra were measured using a Varian/
refluxed in N2 until a wet pH paper strip indicated that HCl Chemagnetics CMX Infinity 300 MHz NMR device equipped
was no longer present in the nitrogen effluent. The suspen- with a Varian/Chemagnetics 7.5 mm HX MAS probe, operating
sion containing dehydrated mesoporous silica in toluene was at a spectral frequency of 299.98 MHz. All spectra were exter-
added to the tungsten precursor solution after the latter was nally referenced to tetramethylorthosilicate at 0 ppm and were
cooled to ambient temperature, and the mixture was refluxed obtained using a 1-s recycle delay and a 5-kHz spinning rate.
overnight in N2 . The reaction mixture was cooled, filtered,
and washed with toluene several times until optical absorp- 2.5. UVvis spectroscopy
tion analysis of the washing solvent indicated that tungsten
precursors were no longer present. The solids were then dried UVvis spectra were measured in diffuse reflectance mode
in ambient air at 393 K for 0.5 h. This sample is desig- using a Varian Cary 5G UVvisNIR spectrophotometer with
nated as as-synthesized. The amount of WCl6 added was an internal integration sphere. Powder samples were placed
varied to give 0.332.66 WOx /nm2 surface density on SBA- within a quartz cell, and spectra were measured in the region
15. We assume that three SiOH groups on SBA-15 react with of 200800 nm after drying ex situ at 343 K. A halon white
each WCl6 , which corresponds to a full monolayer at 30 wt% (PTFE) reflectance standard was used for background sub-
WO3 , or 1.33 WOx /nm2 . The density of SiOH groups on the traction unless otherwise indicated. Spent catalysts were first
mesoporous silica has been estimated as 4.0 SiOH/nm2 [35]. treated in flowing air (Airgas, zero grade) at 673 K to remove
202 J.E. Herrera et al. / Journal of Catalysis 239 (2006) 200211

carbon deposits, and then directly transferred from the reac- Products were analyzed by on-line GC using an Agilent 3000
tion chamber to a He glove bag, from which they were loaded micro gas chromatograph equipped with a thermal conductiv-
into sealed quartz cells to avoid exposure to ambient moisture. ity detector and a Plot Q column. All dehydration rates are
In situ measurements were performed on a Varian (Cary 4) reported on the basis of the number of W atoms in each sam-
spectrometer with a Harrick Scientific diffuse reflectance at- ple.
tachment (DRP-XXX) and a controlled environment chamber
(DRA-2CR). Samples were treated in 20% O2 He (Praxair, 2.7. TPO
99.999%, 0.83 cm3 s1 ) at 673 K and cooled in the same mix-
ture to ambient temperature before spectra were measured. UV TPO measurements were carried out by passing a continu-
vis spectra were also recorded on similarly treated samples dur- ous flow (0.25 cm3 s1 ) of 2% O2 /He (Airgas 99.99%) over
ing steady-state 2-butanol dehydration [0.5 kPa 2-butanol, He samples used for methanol dehydration as the temperature was
(Praxair, UHP grade, 1.33 cm3 s1 )] at 423 K; MgO was used increased linearly at 0.2 K s1 to 1073 K. The CO2 and CO thus
as a reflectance standard. formed were quantitatively converted to methane in a methana-
tor by mixing the effluent with 0.83 cm3 s1 H2 and passing
2.6. Catalytic rate measurements over a 15% Ni/Al2 O3 catalyst held at 673 K. CH4 forma-
tion rates were measured by a flame ionization detector (SRI
Catalytic rates and selectivities were measured to evaluate model 110), calibrated with 100-l pulses of CO2 and by com-
dehydration of 2-butanol and methanol reactants. 2-Butanol de- bustion of known amounts of graphite.
hydration rates were measured at 373 K in a quartz flow re-
actor (1.0 cm i.d.) using samples (0.0200.050 g) dispersed 3. Results and discussion
onto a quartz frit. All samples were treated in flowing dry
air (Airgas, zero grade, 0.8 cm3 s1 ) at 673 K for 1 h before 3.1. XRD
catalytic measurements. Liquid 2-butanol (Aldrich, 99.5%, an-
hydrous) was vaporized into flowing He (Praxair, UHP, 0.3 The XRD patterns of WOx /SBA-15 (1.33 W/nm2 ) prepared
5.3 cm3 s1 ) at ambient pressure and 373 K using a syringe by ALD and treated in dry flowing air at 673 and 773 K
pump to give 0.5 kPa 2-butanol. Reactant conversions were var- are shown in Fig. 1, together with those for SBA-15 and for
ied by changing the He flow rate (0.35.3 cm3 s1 ) at constant WOx /SBA-15 (1.33 W/nm2 ) prepared by conventional impreg-
2-butanol pressures (0.5 kPa). Reactant and product concen- nation (RI). The (100), (110), and (200) low-angle diffraction
trations were measured by gas chromatography (GC), using a lines, corresponding to hexagonal mesoporous silica, are evi-
Hewlett-Packard 6890 gas chromatograph with a 50-m HP-1 dent in all samples, indicating that the nearly amorphous mate-
methyl silicone capillary column and a flame ionization de- rials are characterized by a short-range order. These data show
tector. Methanol dehydration rates were measured at 573 K that the SBA-15 mesoporous structure was retained after WOx
using the same protocols as for 2-butanol. Liquid methanol deposition and treatments at temperatures up to 773 K. No
(Aldrich, 99.9%, anhydrous) was vaporized into flowing He other diffraction lines were detected, indicating that crystalline
(Airgas, UHP) using a syringe pump to give 1 kPa of methanol. WO3 was not formed even after the samples were treated with
Reactant conversion was varied by changing the He flow rate high WOx content at 773 K. In contrast, impregnated samples
(0.68.3 cm3 s1 ) at constant methanol pressure (1.0 kPa). (Fig. 1b, top pattern) showed some diffraction lines at higher

Fig. 1. X-Ray diffraction patterns obtained for SBA-15, ALD-WOx /SBA-15 (1.33 WOx /nm2 ) calcined at 673 K, ALD-WOx /SBA-15 (1.33 WOx /nm2 ) calcined
at 773 K, and impregnated WOx /SBA-15 (1.33 WOx /nm2 ) calcined at 773 K.
 J.E. Herrera et al. / Journal of Catalysis 239 (2006) 200211 203

angles, corresponding to crystalline WO3 (2 = 23.1 , 23.6 , 3.3. In situ XRD studies
and 24.4 ) [36].
The structural stability of WOx /SBA-15 materials was also
3.2. TEM examined by in situ XRD during thermal treatment of a sam-
ple prepared by ALD (1.33 WOx /nm2 ) in air (Fig. 5). No
TEM micrographs (at various sample orientations) of WOx / crystalline WO3 phases were detected at temperatures up to
SBA-15 prepared via different methods and with a range of 823 K. At higher temperatures, some diffraction lines appeared,
WOx contents and thermal treatments are shown in Figs. 2 the intensity of which increased with increasing treatment tem-
4. These micrographs show well-ordered hexagonal arrays of perature. Comparison with previously published patterns (Joint
one-dimensional mesopores typical of SBA-15 (Fig. 2a) for all Committee on Powder Diffraction Standards JCPDS 83-0951)
samples [37]. WO3 crystallites are not present at external SBA- shows that these lines correspond to crystalline WO3 species,
15 surfaces in samples prepared by atomic layer deposition a conclusion confirmed by TEM analysis and UVvis spec-
(Figs. 2 and 3), indicating that WOx species are well dispersed troscopy. It is noteworthy that even at the highest temperature
in these samples even at 30 wt% WO3 (1.33 WOx /nm2 ). (1023 K), we still observed the low-angle diffraction lines (not
Fig. 2 shows TEM micrographs for WOx /SBA-15 with a shown), implying that the mesoporous structure of the hexago-
1.33 W/nm2 surface density prepared by atomic layer depo- nal mesoporous silica was retained.
sition after thermal treatments at 393, 673, and 773 K. Samples
treated at 673 K are indistinguishable from the SBA-15 support 3.4. Solid-state 1 H NMR
or the untreated sample. Even after treatment at 773 K (Fig. 2d),
no WO3 clusters are detected. Samples prepared by impregna- High-resolution 1 H NMR spectroscopy (combining rotation
tion methods (Fig. 4a) show WO3 crystallites larger than 10 nm and multiple-pulse spectroscopy [CRAMPS] [38] and MAS
throughout the mesoporous silica structure after impregnation techniques) has been extensively used to characterize silica,
and drying and even before thermal treatment. In these sam- alumina, and silicaalumina structures [3942]. MAS-only
ples, thermal treatment causes further crystallite growth; WO3 spectra obtained for silica and zeolites are very similar to
structures of submicron size are evident in these samples after those obtained by more technically challenging CRAMPS tech-
treatment at 673 or 773 K, consistent with the crystalline WO3 niques, because of the rather weak 1 H1 H dipolar interactions
phases detected by XRD (Fig. 1b, top diffraction pattern). in these systems [43]. This makes MAS-only techniques at-

Fig. 2. TEM micrographs obtained for (a) SBA-15, (b) WOx /SBA-15 (1.33 WOx /nm2 ) as-synthesized, (c) calcined at 673 K, and (d) calcined at 773 K. The last
three samples were prepared by atomic layer deposition.
204 J.E. Herrera et al. / Journal of Catalysis 239 (2006) 200211

Fig. 3. TEM micrographs obtained for WOx /SBA-15 samples with different Fig. 4. TEM micrographs obtained for a WOx /SBA-15 (1.33 WOx /nm2 ) cata-
tungsten oxide loadings: (a) 0.33 WOx /nm2 , (b) 0.66 WOx /nm2 , and (c) 1.00 lyst prepared by conventional impregnation, (a) and (b) calcined at 673 K and
WOx /nm2 . All three samples were prepared by atomic layer deposition and (c) calcined at 773 K.
were calcined at 673 K before analysis.
arise from methyl and aromatic protons, respectively, in resid-
tractive for characterizing surface protons in high-surface area ual toluene. The line at 3.5 ppm arises from hydrogen-bonded
materials, including heterogeneous catalysts. hydroxyl groups and/or physisorbed water, as in pure SBA-15.
1 H NMRMAS spectra of ALD-synthesized WO /SBA are
x In contrast with pure SBA-15, the line at 1.6 ppm (SiOH) was
shown in Figs. 6 and 7. Fig. 6 shows spectra for pure SBA-15 not observed in the spectra of 1.33 W/nm2 WOx -SBA-15. In-
mesoporous silica (a), for as synthesized 1.33 W/nm2 WOx - stead, a sharp line was detected at 0.75 ppm, which is assigned
SBA-15 (b), and for the latter after treatment in air at 673 K (c) to water molecules that are hydrogen-bonded to WOH groups
and 773 K (d). The NMR spectrum for pure SBA-15 shows based on quantum chemistry calculations at the density func-
a broad line centered at 3.5 ppm and a very sharp line at tion level. The details of these calculations will be published in
1.6 ppm, the latter reflecting the presence of isolated silanols a separate report. This assignment was confirmed by increasing
(SiOH) [44]. The broad line at 3.5 ppm (Fig. 6a) is assigned the tungsten content, which led to the concurrent disappearance
to protons in H2 O and SiOH interacting with each other via of the line for SiOH groups and appearance of the 0.75-ppm
hydrogen bonding [45]. line, demonstrating the titration of silanols in SBA-15 by tung-
Four 1 H NMR lines were observed in the as synthe- sten species.
sized WOx -SBA-15 sample with 1.33 W/nm2 surface density Fig. 6c shows NMR spectra after treatment at 673 K in dry
(Fig. 6b), with chemical shifts at 0.75, 2.0, 3.5, and 6.9 ppm. air. Three sharp lines are observed at 0.75, 1.61, and 3.5 ppm.
Toluene was used as a solvent for synthesis; thus, a 1 H MAS The reappearance of SiOH protons (at 1.61 ppm) indicates that
NMR spectrum was measured for pure SBA-15 impregnated thermal treatment causes the coalescence of some grafted WOx
with toluene. This demonstrates that the 2.0 and 6.9-ppm lines species to form WO3 clusters, which, however, remain highly
 J.E. Herrera et al. / Journal of Catalysis 239 (2006) 200211 205

Fig. 7 shows 1 H NMRMAS spectra for four WOx -SBA-15

samples of varying W content treated in air at 673 K. The in-
tensity of the line at 0.75 ppm (assigned to water molecules that
are hydrogen-bonded to WOH protons) increases with increas-
ing W content, and the intensity ratio of the line at 0.75 ppm
relative to the 1.61-ppm line (SiOH protons) increases con-
currently. This confirms the consumption of the silanol groups
via reaction with tungsten oxide species with increasing WOx
content. In addition, the relatively high concentration of tung-
sten hydroxyl groups is consistent with highly dispersed tung-
sten oxide species. Indeed, if a significant amount of crystalline
tungsten oxide clusters were to form via condensation processes
that yield WOW linkages, then the value of the ratio of the
intensity of the peak at 0.75 ppm to that at 1.6 ppm would dra-
matically decrease; however, this is not observed even for the
samples with surface density of 1.33 W/nm2 (30 wt% WO3 ).
As for WOH protons, the intensity of the sharp water peak at
3.5 ppm increases in parallel with that of molecules that are
hydrogen-bonded to tungsten hydroxyl protons at 0.75 ppm,
Fig. 5. X-Ray diffraction patterns obtained on a ALD-WOx /SBA-15 (1.33
WOx /nm2 ) catalyst prepared by atomic layer deposition. The diffractograms and the ratio of the peak intensity at 3.5 ppm relative to that at
were obtained in situ during thermal treatment at the indicated temperatures. 0.75 ppm is nearly constant (0.57 0.02), indicating that this
water peak also may be associated with the tungsten hydroxyl
dispersed (see the discussion on UVvis spectra below). This protons rather than with the silanol groups. Further investiga-
change in WOx morphology leads in turn to domains of bare sil- tion is needed to understand the exact nature of this peak.
ica with isolated SiOH at the surface. In contrast, treatment in
air at 773 K (Fig. 6d) leads to the disappearance of the 0.75-ppm 3.5. UVvis spectroscopy of fresh samples
line with only the silanol line remaining in the NMR spectrum.
This dehydroxylation process of the WOx species reflects, at The absorption edge energy in the UVvis spectra is a use-
least in part, the formation of WO W bonds, as discussed be- ful indicator of the dispersion and local structure of transition
low. metal cations in oxide domains [13,4648]. Several methods

Fig. 6. 1 H NMR MAS spectra of (a) SBA-15 and of ALD-WOx /SBA-15 Fig. 7. Variation of 1 H NMR MAS spectra with tungsten oxide loading on
(1.33 WOx /nm2 ) after subsequent treatments: (b) as-synthesized, (c) calcined WOx /SBA-15 catalysts prepared by atomic layer deposition: (a) 0.33 WOx /
at 673 K, and (d) calcined at 773 K. The peaks at 2.0 and 6.9 ppm in spectrum nm2 , (b) 0.66 WOx /nm2 , (c) 1.00 WOx /nm2 , and (d) 1.33 WOx /nm2 . All
(b) correspond to methyl and aromatic protons of residual toluene respectively. samples were calcined at 673 K before analysis.
206 J.E. Herrera et al. / Journal of Catalysis 239 (2006) 200211

for estimating the band gap energy of tungsten oxide com- Table 1
pounds using optical absorption spectroscopy have been pro- Edge energy values for different WOx /SBA-15 catalysts obtained from the op-
posed. Davis and Mott [49] proposed an expression to relate tical absorption spectra. The values obtained for samples of known domain size
are also included for reference
the absorption coefficient and the photon energy for various
types of electronic transitions prevalent in semiconductors. For Sample Thermal Edge energy
treatment (K) value (eV)
dispersed WOx domains, Barton et al. [13] used the square
root of the KubelkaMunk function multiplied by the photon WO3 2.4 eV
(NH4 )10 W12 O41 3.0 eV
energy, which allows the edge energy to be obtained by extrap- Na2 WO4 4.6 eV
olation to zero absorbance for amorphous semiconductors. The WOx /SBA-15 0.33 WOx /nm2 673 K 3.8 eV
values thus obtained carry information about the average do- WOx /SBA-15 0.66 WOx /nm2 673 K 3.5 eV
main size of the oxide nanoparticles [50], although the values WOx /SBA-15 1.00 WOx /nm2 673 K 3.6 eV
also depend on local symmetry [51] and support electronega- WOx /SBA-15 1.33 WOx /nm2 673 K 3.2 eV
tivity [29]. WOx /SBA-15 2.66 WOx /nm2 673 K 3.0 eV
WOx /SBA-15 1.33 WOx /nm2 773 K 3.1 eV
Diffuse-reflectance UVvis spectra are shown in Fig. 8 for
WOx /SBA-15 1.33 WOx /nm2 1073 K 3.0 and 2.6 eV
the WOx -SBA-15 samples. The absorption edge arises from RI WOx /SBA-15 1.00 WOx /nm2 673 K 3.5 eV
ligand-to-metal charge transitions [5254]. The position of the RI WOx /SBA-15 1.33 WOx /nm2 673 K 3.1 eV
absorption edge allows comparisons between the samples re-
ported here and reference compounds with known domain size,
although such comparisons are not always straightforward. For cate highly dispersed WOx species. The shift to lower energies
instance, based on previously published X-ray absorption stud- for the sample treated at 773 K is consistent with the growth of
ies [55] and earlier extended Hckel calculations [56], Barton et WOx domains, probably via condensation reactions of WOH
al. [13] showed that for zirconia-supported WOx catalysts, edge to form WOW linkages. Indeed, 1 H NMR data show that
energy values >3.5 eV indicate the presence of isolated WOx WOH groups disappear when the catalyst is treated above
species in distorted octahedral symmetry and without bridging 773 K. The dispersion of this sample is relatively high, but
WOW bonds. treatment at 1073 K leads to the appearance of a low-energy
Table 1 shows absorption edge energies for several WOx shoulder (at 2.6 eV) near the absorption edge, corresponding
species of known domain size and coordination, together to the formation of large WO3 crystallites, also detected in this
with those for SBA-15-supported WOx samples prepared via sample by XRD. The residual edge at higher energies indicates
ALD and impregnation methods with various tungsten con- that some WOx species remain reasonably well dispersed even
tents and thermal treatments. Edge energies for all samples after treatment at 1073 K.
(3.03.8 eV) except the sample treated at 1073 K resemble
those for (NH4 )10 W12 O41 , indicating that the WOx domains 3.6. Catalytic dehydration of 2-butanol and methanol
are very small. The edge energy value decreases systematically
with increasing WO3 content and thermal treatment (Fig. 8), The dehydration of 2-butanol and methanol reactions were
consistent with a parallel increase in WOx domain size. used to probe the catalytic properties of WOx -SBA-15 samples
Fig. 8b shows the effect of thermal treatment temperature prepared via ALD and impregnation methods. Only dehydra-
on the UVvis spectra of 1.33 W/nm2 WO3 -SBA-15. As for tion products were detected for both alcohols (1-butene, cis-
the samples in Fig. 8a, edge energy values of 3.1 eV indi- and trans-2-butenes, dimethyl-ether). Bimolecular dehydration

Fig. 8. (a) UVvis DRS spectra for three WOx /SBA-15 samples calcined at 673 K and with different tungsten loadings. (b) UVvis DRS spectra for a WOx /SBA-15
(1.33 WOx /nm2 ) sample after subsequent thermal treatments. All samples were prepared by atomic layer deposition. [F (R )h] represents the KubelkaMunk
function multiplied by the photon energy.
 J.E. Herrera et al. / Journal of Catalysis 239 (2006) 200211 207

Fig. 9. (a) 2-Butanol dehydration rates (per W-atom) at 373 K as a function of time on stream and WOx surface density: (F) ALD 0.33 WOx /nm2 , (Q) ALD
1.33 WOx /nm2 , and (") conventional impregnated 1.33 WOx /nm2 (0.5 kPa 2-butanol, 100.8 kPa He, 80 sccm total flow rate). (b) Methanol dehydration rates
(per W-atom) at 573 K as a function of time on stream and WOx surface density: (F) ALD 0.33 WOx /nm2 , (Q) ALD 1.33 WOx /nm2 , and (") conventional
impregnated 1.33 WOx /nm2 (1.0 kPa methanol, 100.3 kPa He, 80 sccm total flow rate).

of 2-butanol to ethers was not observed, because such reactions

are significantly slower than monomolecular pathways leading
to alkenes, especially at low alcohol pressures [28,57,58].
Fig. 9a shows steady-state 2-butanol dehydration rates
(per W atom) as a function of time on stream on two sam-
ples prepared by ALD with surface densities of 0.33 and
1.33 WOx /nm2 and on a sample prepared by impregnation
with a surface density of 1.33 WOx /nm2 . These data show that
2-butanol dehydration rates are strongly influenced by WO3
content and that optimum rates (per W atom) are achieved at
intermediate WO3 contents, as also shown previously for WOx
species on ZrO2 [11,28] and Al2 O3 [29] supports. Rates are Fig. 10. (Q) 2-Butanol dehydration rates (per W-atom) at 373 K as a function of
similar on the samples with a surface density of 1.33 WOx /nm2 conversion for an ALD sample with surface density 0.66 WOx /nm2 [0.5 kPa
prepared by either ALD or impregnation methods and treated 2-butanol, 100.8 kPa He, 6.2111.3 mol 2-BuOH/(mol W h)]. (") Methanol
at 673 K. dehydration rates (per W-atom) at 573 K as a function of conversion for an ALD
sample with surface density 0.66 WOx /nm2 [1.0 kPa methanol, 100.3 kPa He,
Fig. 9b shows that methanol dehydration rates (per W atom)
24.6394 mol MeOH/(mol W h)].
also depend on WO3 content, but in this case, the rate is higher
on the sample prepared by ALD than on the sample with similar
composition but prepared via impregnation. Also, deactivation to contain predominately monotungstate species and show very
is more rapid on samples with lower tungsten oxide loadings low catalytic dehydration rates (Fig. 11). Dehydration rates in-
than on samples with higher densities of WOx species. crease as isolated WOx species condense into polytungstate
Fig. 10 shows that 2-butanol and methanol dehydration species with increasing WOx content. At even higher WOx con-
rates (373 and 573 K, respectively) decrease with increas- tent (200% monolayer coverage), dehydration rates decrease
ing reactant conversion on a sample with a surface density of as three-dimensional WO3 crystallites form and WOx species
0.66 WOx /nm2 prepared by ALD; the conversion can be var- become inaccessible to reactants. These trends are also seen
ied by changing the space velocity. Similar trends are found on for methanol dehydration rates, which appear to require simi-
all of the other samples. These trends reflect kinetic inhibition lar Brnsted acid sites.
by water formed during 2-butanol dehydration [28]. A more Methanol dehydration rates decrease with time-on-stream
rigorous comparison of intrinsic reactivity requires that we ex- more rapidly than 2-butanol dehydration rates (Fig. 9), possibly
trapolate measured rates to zero reactant conversion; these rates because of the higher temperatures required for methanol reac-
are shown in Fig. 11 as a function of WOx surface density. The tions. The presence of water at these higher reaction tempera-
trends observed for 2-butanol dehydration are similar to those tures (573 K) may cause more rapid sintering of WOx species
reported on WOx ZrO2 [28] and WOx Al2 O3 [29] as a func- or carbon deposition at Brnsted acid sites formed via reduc-
tion of WOx surface density. Samples with low WOx surface tion of W centers [28]. These two factors may account for the
densities and high absorption edge energies (Table 1) appear different activities observed for methanol dehydration (Figs. 9
208 J.E. Herrera et al. / Journal of Catalysis 239 (2006) 200211

Fig. 11. (Circles) Initial 2-butanol dehydration rates (per W-atom) at 373 K as a function of tungsten oxide surface density in the catalyst. (Triangles) Initial methanol
dehydration rates (per W-atom) at 573 K as a function of tungsten oxide surface density in the catalyst. The filled symbols correspond to catalyst obtained by ALD
and the open ones to conventionally impregnated samples.

and 11) on samples with identical tungsten loadings but pre- ferences observed in the two TPO profiles shown in Fig. 12.
pared using different methods. The rationale for the observed It is also possible that the crystalline WO3 phase present on
differences in catalytic behavior between the ALD and impreg- the sample prepared by conventional impregnation overreduces
nated samples and the origin of the deactivation are discussed during methanol dehydration, leading to the formation of more
next based on characterization of spent catalysts. refractory carbon. In any case, the total amount of carbon is al-
most the same in both samples, which seems to indicate that
3.7. TPO the observed differences in catalytic activity for methanol de-
hydration between ALD and conventional impregnated samples
TPO can be used to probe carbonaceous deposits on catalyst (Fig. 11b) are not due to the blockage of the catalyst surface by
surfaces. Analysis of the evolved CO2 /CO allows both qualita- carbonaceous species.
tive and quantitative assessments of the type and reactivity of
these deposits [59,60]. TPO measurements detected significant 3.8. UVvis spectra of spent catalysts
amounts of carbon after methanol dehydration, but no carbon
deposits after 2-butanol dehydration. This indicates that car- Fig. 13 shows the in situ UVvis spectra of two WO3 /SBA-
bonaceous deposits on the catalysts may well be one reason for 15 samples (1.33 W/nm2 ) prepared by ALD and impregnation
catalyst deactivation during methanol dehydration. as a function of time on stream during 2-butanol dehydration
Fig. 12 compares TPO profiles for two samples with a sur- at 423 K. Absorption intensity appears in the pre-edge region
face density of 1.33 W/nm2 , prepared by ALD and impregna- as reaction proceeds, reflecting the formation of reduced color
tion, after methanol dehydration at 573 K for 2 h. Both samples centers with concurrent generation of Brnsted acid sites via
show maximum oxidation rates at 783 K, but the impregnated dehydrogenation of 2-butanol during the initial stages of the
samples show an additional strong feature at 843 K. These reaction [28]. Interestingly, in the RI sample (Fig. 13a), the
more refractory carbonaceous deposits may reflect an intrinsic pre-edge appears to plateau as the reaction proceeds, whereas
difference in acid site strength between these materials. Indeed, in the ALD sample, a plateau in the intensity of the pre-edge
MeOH-TPD experiments [61] indicated that samples prepared is not observed within the range of time on stream studied.
by ALD and conventional impregnation with the same WO3 These differences may indicate differing degrees of reduction
loading (i.e., 1.00 WOx /nm2 ) show a small but clear difference of the tungsten species as the reaction progresses, with a larger
in the temperature at which methoxide species recombinatively amount of reducible tungsten domains in the case of the ALD
desorb as methanol (332 K vs. 346 K for desorption peaks), sample. However, as mentioned above, differences in activity
suggesting that methoxide species are more strongly adsorbed between these two samples are not observed; this apparent dis-
on the impregnated sample. These stronger sites are likely to crepancy can be rationalized in terms of the stability of the
deactivate more rapidly during the initial reaction stages of catalysts, as explained below.
reaction and affect the interaction of the dehydrogenation prod- The results obtained for the edge energy values on these
ucts (formed at the initial stages of the reaction as well) with samples merit some additional discussion. The position of the
the surface, thus leading to the formation of more refractory primary absorption edge remains unchanged as the reaction oc-
carbon on the catalyst surface. This may account for the dif- curs and pre-edge features appear. Thus, the WOx domain size
 J.E. Herrera et al. / Journal of Catalysis 239 (2006) 200211 209

Fig. 12. Temperature-programmed oxidation (TPO) profiles of the carbona-

ceous species generated after methanol dehydration over (a) a conventionally
impregnated sample with surface density 1.33 WOx /nm2 and (b) an ALD sam-
ple with surface density 1.33 WOx /nm2 . The profiles were obtained using a
mixture of 5% of O2 in He at a heating rate of 0.2 K s1 .

remains unchanged by the reaction for both the ALD and con-
ventionally impregnated samples, consistent with the similar
2-butanol dehydration rates on these two samples, which do not
sinter during reaction. Nonetheless, studies of these samples be-
fore reaction show the presence of crystalline WO3 only in the
impregnated sample. Thus, it appears that impregnated samples
Fig. 13. Diffuse reflectance UVvis absorption spectra obtained during 2-bu-
have a broader WOx domain size distribution but an average ac- tanol dehydration on a sample with surface density 1.33 WOx /nm2 prepared by
tivity similar to that of the ALD samples, which have a narrower (a) conventional impregnation and (b) atomic layer deposition (423 K, 0.5 kPa
distribution of domain sizes. The similar steady-state activities 2-butanol, 100.8 kPa He, 80 sccm total flow rate). [F (R )h] represents the
observed are also related to the mild reaction conditions used KubelkaMunk function multiplied by the photon energy.
to perform 2-butanol dehydration, under which both type of
catalysts are stable. Under the more severe reaction conditions (Fig. 14), indicating that WOx domains grow during reaction.
needed for methanol dehydration (573 K), differences in the Pre-edge features for reduced centers are not detectable, be-
stability and performance between the ALD and conventionally cause the samples were treated in air at 673 K for 2 h to remove
impregnated materials are indeed observed. carbon deposits before measuring spectra. WOx growth occurs
Methanol dehydration reactions lead to a decrease in the ad- rapidly during methanol dehydration, because spectral changes
sorption edge energy for both ALD and impregnated samples are evident even after only 300 s on stream. It is likely that sin-
210 J.E. Herrera et al. / Journal of Catalysis 239 (2006) 200211

Fig. 14. Diffuse reflectance UVvis absorption spectra obtained after methanol dehydration on a sample with surface density 1.33 WOx /nm2 prepared by (a) con-
ventional impregnation and (b) atomic layer deposition (573 K, 1.0 kPa 2-butanol, 100.3 kPa He, 80 sccm total flow rate). [F (R )h] represents the KubelkaMunk
function multiplied by the photon energy.

tering accompanies the reduction of WOx species during initial improved resistance toward sintering of the catalyst prepared by
contact with methanol at 573 K. Moreover, these changes in ALD, which in turn is rationalized in terms of the higher density
edge energies are much more pronounced for impregnated sam- of SiOW moieties due to a better interaction of WOx with
ples than for ALD samples (0.2 vs. 0.5 eV shifts; Fig. 14), con- the support surface. The TPO results reveal that the samples
sistent with the higher apparent methanol dehydration rates on prepared by ALD have a more uniform acid strength compared
the latter samples (Fig. 11). However, this difference in activ- with the samples prepared by conventional impregnation and
ity between the ALD and conventionally impregnated samples that coke formation is not a major factor causing the different
is not present in the case of 2-butanol dehydration, where the steady-state activities in the ALD and impregnated samples for
UVvis DRS results indicate a lack of WOx sintering. As men- methanol dehydration.
tioned above, this difference in catalytic behavior appears to be
related to the different reaction conditions used for methanol Acknowledgments
and 2-butanol reactants. At the higher reaction temperatures re-
quired for methanol dehydration, the presence of water vapor This work was supported by US Department of Energy
(a byproduct of dehydration) is likely to enhance sintering by (DOE), Office of Basic Energy Sciences, Division of Chemical
hydrolyzing the SiOW bonds between WOx and the support Sciences under contract DE-AC06-76RLO-1830 #45385. The
and mobilizing tungsten oxide species. The higher resistance research was performed in the Environmental Molecular Sci-
to this sintering process of the catalyst prepared by ALD may ences Laboratory, a national scientific user facility sponsored
reflect either a higher density of SiOW moieties or a more by the DOE Office of Biological and Environmental Research
effective use of the available surfaces area of silica by ALD rel- and located at the Pacific Northwest National Laboratory. Some
ative to conventional impregnation methods. of the research was carried out at the Berkeley Catalysis Cen-
ter facilities of the University of California at Berkeley (UCB).
4. Conclusion The UCB research was supported by a grant from the DOE, Of-
fice of Basic Energy Sciences, Division of Chemical Sciences.
Supported tungsten oxide materials synthesized by ALD are The authors thank Professor Daniel Resasco, University of Ok-
very promising catalysts. High dispersions are achieved even lahoma for his help with the TPO profile measurements.
at a WO3 loading of 30 wt%, and enhanced thermal stability
is observed, leading to superior catalytic activity for reactions References
requiring severe conditions, such as methanol dehydration. The
differences in activity observed between the samples prepared [1] L.S. Wang, L.X. Tao, M.S. Xie, G.F. Xu, J.S. Huang, Y.D. Xu, Catal.
by ALD and conventional impregnation are attributed mainly to Lett. 21 (1993) 35.
 J.E. Herrera et al. / Journal of Catalysis 239 (2006) 200211 211

[2] C.L. Thomas, Catalytic Processes and Proven Catalysts, Academic Press, [32] Z. Zhang, J. Suo, X. Zhang, S. Li, Appl. Catal. A: Gen. 179 (1999) 11.
New York, 1970. [33] E. Briot, J.-Y. Piquemal, M. Vennat, J.-M. Bregeault, G. Chottard, J.-M.
[3] J. Pasel, P. Kassner, B. Montanari, M. Gazzano, A. Vaccari, W. Makowski, Manoli, J. Mater. Chem. 10 (2000) 953.
T. Lojewski, R. Dziembaj, H. Papp, Appl. Catal. B: Environ. 18 (1998) [34] D. Zhao, J. Feng, Q. Huo, N. Melosh, G.H. Fredrickson, B.F. Chmelka,
199. G.D. Stucky, Science 279 (1998) 548.
[4] A. Butler, C. Nicolaides, Catal. Today 18 (1993) 443. [35] I.G. Shenderovich, G. Buntkowsky, A. Schreiber, E. Gedat, S. Sharif, J.
[5] W.Z. Cheng, V. Ponec, Catal. Lett. 25 (1994) 337. Albrech, N.S. Golobev, G.H. Findenegg, H. Limbach, J. Phys. Chem.
[6] M.A. Alvarez-Merino, F. Carrasco-Marn, C. Moreno-Castilla, J. Ca- B 107 (2003) 11924.
tal. 192 (2000) 374. [36] Y. Wang, Q. Cheng, W. Yang, Z. Xie, W. Xu, D. Huang, Appl. Catal. A:
[7] C. Moreno-Castilla, M.A. Alvarez-Merino, F. Carrasco-Marn, React. Gen. 250 (2003) 25.
Kinet. Catal. Lett. 71 (2000) 137. [37] M.H. Lim, A. Stein, Chem. Mater. 11 (1999) 3285.
[8] L.L. Murrell, N.C. Dispenziere Jr., Catal. Lett. 4 (1990) 235. [38] B.C. Gerstein, C.R. Dybowsky, Transient Techniques in NMR of Solids,
[9] A.F. Perez-Cadenas, C. Moreno-Castilla, F.J. Maldonado-Hodar, J.L.G. Academic Press, San Diego, 1985.
Fierro, J. Catal. 217 (2003) 30. [39] M. Hunger, D. Freude, H. Pfeifer, H. Bremer, M. Jank, K.P. Wendland,
[10] J.C. Vartuli, J.G. Santiesteban, P. Traverso, N. Cardona Martinez, C.D. Chem. Phys. Letts. 100 (1983) 29.
Chang, S.A. Stevenson, J. Catal. 187 (1999) 131. [40] C.E. Bronnimann, I.S. Chuang, B.L. Hawkins, G.E. Maciel, J. Am. Chem.
[11] D.G. Barton, S.L. Soled, E. Iglesia, Top. Catal. 6 (1998) 87. Soc. 109 (1987) 1562.
[12] D.G. Barton, S.L. Soled, G.D. Meitzner, G.A. Fuentes, E. Iglesia, J. Ca- [41] J.M. Miller, L.J. Lakshmi, Appl. Catal. A: Gen. 190 (2000) 197.
tal. 181 (1999) 52. [42] C.E. Bronnimann, R.C. Zeigler, G.E. Maciel, J. Am. Chem. Soc. 110
[13] D.G. Barton, M. Shtein, R.D. Wilson, S.L. Soled, E. Iglesia, J. Phys. (1988) 2023.
Chem. B 103 (1999) 630. [43] S.F. Dec, C.E. Bronnimann, R.A. Wind, G.E. Maciel, J. Magn. Res. 82
[14] G. Busca, Phys. Chem. Chem. Phys. 1 (1999) 273. (1989) 454.
[15] P. Carniti, A. Gervasini, A. Auroux, J. Catal. 150 (1994) 274.
[44] G.E. Maciel, P.D. Ellis, in: A.T. Bell, A. Pines (Eds.), NMR Techniques in
[16] J. Bernholc, J.A. Horsley, L.L. Murrell, L.G. Sherman, S. Soled, J. Phys.
Catalysis, Marcel Dekker, New York, 1994, pp. 231309.
Chem. 91 (1987) 1526.
[45] J.Z. Hu, J.H. Kwak, J.E. Herrera, Y. Wang, C.H.F. Peden, Solid State Nucl.
[17] V.M. Benitez, C.A. Querini, N.S. Figoli, R.A. Comelli, Appl. Catal. A 178
Magn. Reson. 27 (2005) 200.
(1999) 205.
[46] A. Gutirrez-Alejandre, P. Castillo, J. Ramirez, G. Ramis, G. Busca, Appl.
[18] L. Karakonstantis, H. Matralis, Ch. Kordulis, A. Lycourghiotis, J. Ca-
Catal. A: Gen. 216 (2001) 181.
tal. 162 (1996).
[47] A. Khodakov, J. Yang, S. Su, E. Iglesia, A.T. Bell, J. Catal. 177 (1998)
[19] D.S. Kim, M. Ostromecki, I.E. Wachs, J. Molec. Catal. A: Chem. 106
343.
(1996) 93.
[48] D. Wei, G.L. Haller, Proceedings of the 2nd Memorial G.K. Boreskov
[20] I.E. Wachs, Catal. Today 27 (1996) 437.
International Conference, Novosibirsk, July, 1997, p. 110.
[21] J.C. Mol, J.A. Moulijin, J. Mol. Catal. 15 (1982) 157.
[22] C. Martin, O. Malet, G. Solana, V. Rives, J. Phys. Chem. B 102 (1999) [49] E.A. Davis, N.F. Mott, Philos. Mag. 22 (1970) 903.
2759. [50] R.S. Weber, J. Catal. 151 (1995) 470.
[23] R. Thomas, E.M. Van Oers, V.H.J. De Beer, J. Medema, J.A. Moulijin, [51] M. Fournier, C. Louis, M. Che, P. Chaquin, D. Masure, J. Catal. 119 (1989)
J. Catal. 76 (1982) 241. 400.
[24] S. Soled, L. Murrell, I. Wachs, G. McVicker, Prep.-Am. Chem. Soc., Div. [52] H. So, M.T. Pope, Inorg. Chem. 11 (1972) 1441.
Pet. Chem. 28 (1983) 1310. [53] M.L. Good, Spectrochim. Acta A 29 (1973) 707.
[25] S.L. Soled, G.B. McVicker, L.L. Murrell, L.G. Sherman, N.C. Dispenziere [54] M. Iwamoto, H. Furukawa, K. Matsukami, T. Takenaka, S. Kagawa,
Jr., S.L. Hsu, D. Waldman, J. Catal. 111 (1988) 286. J. Amer. Chem. Soc. 105 (1983) 3719.
[26] S. Soled, L.L. Murrell, I.E. Wachs, G.B. McVicker, L.G. Sherman, S. [55] E. Iglesia, D.G. Barton, S.L. Soled, S. Miseo, J.E. Baumgartner, W.E.
Chan, N.C. Dispenziere, R.T.K. Baker, ACS Symposium Series 279 Gates, G.A. Fuentes, G.D. Meitzner, Stud. Surf. Sci. Catal. 101 (1996)
(1985) 165. 533.
[27] W.V. Knowles, E.I. Ross, I.E. Wachs, M.S. Wong, Abstracts of Papers, [56] D. Masure, P. Chaquin, C. Louis, M. Che, M. Fournier J. Catal. 119 (1989)
228th ACS National Meeting, Philadelphia, PA, August 2226, 2004 415.
(2004). [57] H. Knozinger, R. Kohne, J. Catal. 5 (1966) 264.
[28] C.D. Baertsch, K.T. Komala, Y.-H. Chua, E. Iglesia, J. Catal. 205 (2002) [58] H. Knozinger, H. Buhl, K. Kochloefl, J. Catal. 24 (1972) 57.
44. [59] D.E. Resasco, B.K. Marcus, C.S. Huang, V.A. Durante, J. Catal. 146
[29] J. Macht, C.D. Baertsch, M. May-Lozano, S.L. Soled, Y. Wang, E. Iglesia, (1994) 40.
J. Catal. 227 (2004) 479. [60] S.M. Stagg, C.A. Querini, W.E. Alvarez, D.E. Resasco, J. Catal. 168
[30] X.-L. Yang, W.-L. Dai, H. Chen, Y. Cao, H. Li, H. He, K. Fan, J. Catal. 229 (1997) 75.
(2005) 259. [61] J.H. Kwak, J.E. Herrera, Y. Wang, C.H.F. Peden, manuscript in prepara-
[31] J. Jarupatrakorn, M.P. Coles, T.D. Tilley, Chem. Mater. 17 (2005) 1818. tion.