Titania±silica as catalysts: molecular structural characteristics and
physico-chemical properties
Xingtao Gao, Israel E. Wachs
*
Department of Chemistry and Chemical Engineering, Zettlemoyer Center for Surface Studies, Lehigh University, 7 Asa Drive,
Bethlehem, PA 18015, USA
Abstract
Recent results on characterization and applications of titania±silica materials as photocatalysts, acid catalysts and oxidation
catalysts are reviewed. The similarities and differences in structural characteristics and physico-chemical properties between
titania±silica mixed and supported oxides are emphasized. The generation of new catalytic active sites either on the silica
surface or in the silica matrix is discussed with respect to the formation of Ti±O±Si bonds and the local structure. The insights
obtained from these studies allow a fundamental understanding of the relationships between the structural characteristics and
the physico-chemical/reactivity properties of titania±silica catalysts. # 1999 Elsevier Science B.V. All rights reserved.
Keywords: TiO
2
±SiO
2
mixed oxides; TiO
2
/SiO
2
supported oxides; Catalyst; Glass; Sol±gel; Multilayers/thin ®lms; Surface structure;
Coordination geometry; Electronic property; Photocatalysis; Acidity; Epoxidation and oxidation reactions; Isomerization and dehydration
reactions; EXAFS/XANES; UV±Vis DRS; Raman; IR; XPS spectroscopy
1. Introduction
Titania±silica represents a novel class of materials
that have attracted much attention in recent years.
Titania±silica materials have been extensively used as
catalysts and supports for a wide variety of reactions
[1±48], as summarized in Table 1. In addition, titania±
silica can be utilized as protective coating on stainless
steel to resist oxidation and chemical attack [49,50],
antire¯ection coatings for optical glasses [51], and as
very interesting glass materials with ultralow thermal
expansion coef®cients [52±57] and high refraction
indices [58]. Such advanced titania±silica materials
not only take advantage of both TiO
2
(an n-type
semiconductor and an active catalytic support) and
SiO
2
(high thermal stability and excellent mechanical
strength), but also extends their applications through
the generation of new catalytic active sites due to the
interaction of TiO
2
with SiO
2
.
The applications of titania±silica materials as cat-
alysts and supports fall into three categories based on
their unique physico-chemical properties: (i) photo-
catalysis that is associated with the support effect and
the quantum-size effect; (ii) acid catalysis that is
related to the generation of new acid sites; and (iii)
excellent catalytic support materials that possess
enhanced thermal and mechanical stability due to
SiO
2
while preserving the catalytic performance of
TiO
2
. The understanding of the structural character-
istics of titania±silica and the relationships with the
physico-chemical/reactivity properties is also of great
importance in a wide range of applied sciences. This
Catalysis Today 51 (1999) 233±254
*Corresponding author. Tel.: +1-610-758-4274; fax: +1-610-
758-6555; e-mail: [email protected]
0920-5861/99/$ ± see front matter # 1999 Elsevier Science B.V. All rights reserved.
PII: S 0920-5861(99)00048-6
Table 1
Summary of various reactions for titania±silica as catalysts and supports
Catalysts Preparation
method
Reaction
temperature (K)
Reactions Reference
TiO
2
±SiO
2
Sol±gel RT Photodecomposition of chlorinated phenols [1,2]
TiO
2
±SiO
2
Sol±gel/CVD RT Photoreduction of CO
2
[3,4]
TiO
2
±SiO
2
Sol±gel RT Photodecomposition of rhodamine-6G and phenol [5,6]
TiO
2
±SiO
2
Sol±gel 380 Complete photocatalytic oxidation of C
2
H
4
[7]
TiO
2
/SiO
2
Impregnation RT Photoxidation of propane [8]
TiO
2
/SiO
2
Precipitation 673 Catalytic decomposition of 1,2-dichloroethane [9]
TiO
2
±SiO
2
± 823 Catalytic decomposition of Freon 12 [10]
TiO
2
/SiO
2
Precipitation 523±673 Catalytic decomposition of chloroform [11]
TiO
2
±SiO
2
Sol±gel/coprecipitation 423±523 Isomerization of 1-butene [12±18]
TiO
2
±SiO
2
Sol±gel 523 Isomerization of methyloxane to propanal [18]
TiO
2
±SiO
2
Sol±gel/coprecipitation ± Methanol dehydration [19]
TiO
2
±SiO
2
Coprecipitation 493 Ethene hydration [15]
TiO
2
±SiO
2
Coprecipitation 723 Phenol amination [15]
TiO
2
±SiO
2
Coprecipitation 673 Cumene dealkylation [20]
TiO
2
±SiO
2
Sol±gel ± Decane hydrocracking [21]
TiO
2
±SiO
2
Sol±gel/impregnation ± Propanol dehydration [12,20,36,37]
TiO
2
±SiO
2
Coprecipitation 303 Solvolysis of cis-2,3-epoxybutane [22]
TiO
2
±SiO
2
Sol±gel 493 Ammoxidation of cyclohexanone [23]
TiO
2
±SiO
2
Sol±gel 333 Epoxidation of a-isophorone by TBHP [24,25]
TiO
2
±SiO
2
Sol±gel/coprecipitation 323±363 Epoxidation of olefins by TBHP/NBHP/H
2
O
2
[21,22,26±31]
TiO
2
/SiO
2
Impregnation 363±383 Epoxidation of olefins by TBHP/EBHP [31±33]
TiO
2
±SiO
2
Sol±gel 353 Selective oxidation of cyclohexane by TBHP [34]
TiO
2
±SiO
2
Sol±gel 353 Hydroxylation of phenol by H
2
O
2
[35]
TiO
2
±SiO
2
Sol±gel 353 Oxidation of benzene and toluene by H
2
O
2
[35]
TiO
2
/SiO
2
Impregnation 503 Methanol oxidation [36,38]
Rh/TiO
2
±SiO
2
Sol±gel 303 Benzene hydrogenation [39]
Ni/TiO
2
±SiO
2
Coprecipitation 548 CO hydrogenation [14]
CrO
3
/TiO
2
±SiO
2
Coprecipitation/impregnation 373±383 Ethylene polymerization [40,41]
V
2
O
5
/TiO
2
±SiO
2
Sol±gel/impregnation 370±570 SCR of NO with NH
3
[42±44]
V
2
O
5
/TiO
2
±SiO
2
Coprecipitation 353 Synthesis of isobutyaldehyde from ethanolmethanol [45]
V
2
O
5
/TiO
2
/SiO
2
Precipitation 423±823 NO reduction with CO [46]
V
2
O
5
/TiO
2
/SiO
2
Precipitation 600±700 Selective oxidation of toluene [47]
V
2
O
5
/TiO
2
/SiO
2
Precipitation 533±673 Selective oxidation of o-xylene [48]
234 X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254
review focuses on the investigation and development
of amorphous titania±silica mixed and supported oxi-
des as catalysts, excluding Ti-silicalites, over the past
10 years with special emphasis on the structural
characterization and establishment of the relationships
between the structural characteristics and the physico-
chemical/reactivity properties.
Comparison of the various results in the literature
between different researchers is sometimes dif®cult
because of different preparation methods, sample
treatments, and characterization techniques used to
determine the structural and surface properties. In
spite of this situation, we will try to provide a con-
sistent picture to clarify the structure±property rela-
tionships for titania±silica materials based on the
combined results of different characterization techni-
ques and researchers.
2. Interaction of TiO
2
with SiO
2
There are two types of interaction between TiO
2
and
SiO
2
: physically mixed (with interaction forces being
nothing more than weak Van der Waals forces) and
chemically bonded (i.e., the formation of Ti±O±Si
linkages). When strong interaction results in chemical
bonding, the physico-chemical/reactivity properties of
titania±silica are very different from the simple com-
bination of the individual phases (mechanical mix-
tures). The degree of interaction, or in other words
homogeneity or dispersion, largely depends on pre-
paration methods and synthesis conditions (see
below). Many different preparation methods have
been employed to synthesize titania±silica. The most
widely used methods to prepare TiO
2
±SiO
2
mixed
oxides and glasses are sol±gel hydrolysis [1,3,5±
7,10,12,26±30,35,39,42,44,49,50,59±64], coprecipi-
tation [15,16,19,20,22,45,65], and ¯ame hydrolysis
[66±68]. The supported oxides of TiO
2
deposited on
the SiO
2
substrate, denoted as TiO
2
/SiO
2
supported
oxides, have been much less investigated and are
prepared by impregnation [8,36±38,69±76], chemical
vapor deposition [77±79] and precipitation
[11,71,72,76,80]. The interaction of TiO
2
with SiO
2
at the interface of TiO
2
/SiO
2
multilayers/thin ®lms as
one type of supported oxides will also be addressed in
order to have a clearer overview about the intrinsic
nature of interaction between TiO
2
and SiO
2
. The
general term titania±silica is used to include both
mixed and supported oxides.
2.1. TiO
2
±SiO
2
mixed oxides
2.1.1. Preparation of TiO
2
±SiO
2
mixed oxides
TiO
2
±SiO
2
mixed oxides are generally prepared by
sol±gel and coprecipitation methods. Two types of Ti
species are present in TiO
2
±SiO
2
mixed oxides: seg-
regated TiO
2
microdomains and isolated Ti species,
with the relative ratio depending on the chemical
composition, the preparation methods and synthesis
conditions (the hydrolysis route, Ti content, drying
method and calcination temperature) [13,31,81]. The
degree of homogeneity at the atomic level is com-
monly associated with the relative amount of Ti±O±Si
linkages in TiO
2
±SiO
2
mixed oxides
[12,13,19,26,31,64,82,83], and Ti±O±Si bonds are
more effectively formed as the homogeneity increases.
Among the various preparation methods, sol±gel
hydrolysis is most widely used due to its possible
capability in controlling the textural and surface prop-
erties of the mixed oxides. In sol±gel processes,
domain formation due to the differences in the hydro-
lysis and the condensation rates of Ti- and Si- alk-
oxides was identi®ed to be a major problem in the
preparation of atomically mixed TiO
2
±SiO
2
oxides
[82]. However, the two-stage hydrolysis procedure
recently developed, which is performed in acidic
conditions, seems to have overcome this problem
and results in the best Ti±O±Si connectivities and
the highest homogeneity [13,62,82].
Although the homogeneity of TiO
2
±SiO
2
mixed
oxides varies with the preparation methods and synth-
esis conditions, the atomically mixed TiO
2
±SiO
2
oxi-
des can only be obtained at low TiO
2
content, with the
maximum TiO
2
concentration less than 15 wt%
[65,68], or Si/Ti atomic ratio higher than 7.5. At
higher Ti contents, TiO
2
crystallites tend to form as
a separate phase, demonstrating that silica could not
favorably accommodate all the Ti atoms in the net-
work above a certain limit.
2.1.2. Experimental evidence for Ti±O±Si bonding
The presence of Ti±O±Si linkages in TiO
2
±SiO
2
mixed oxides has been detected directly or indirectly
by many techniques, such as XPS [65,67], EXAFS/
XANES spectroscopy [27±29,68,81,84,85], IR and
X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254 235
Raman spectroscopy [62,64,82,86±88],
29
Si and
17
O NMR spectroscopy [26,35,61±63,89].
The simplest way to examine the formation of
Ti±O±Si bonds is to use IR spectroscopy. The IR band
observed at 910±960 cm
ÿ1
is widely accepted as the
characteristic vibration due to the formation of
Ti±O±Si bonds [30,35,62,64,81,82,86±88], with the
exact band position depending on the chemical com-
position of the sample as well as calibration and
resolution of the instrument. The intensity of this
IR band has been used to evaluate the absolute amount
of Ti±O±Si linkages [62,64,82,86±88], which has been
shown to increase with increasing Ti content up to
20 wt% TiO
2
[64]. The dispersion of Ti in the SiO
2
matrix has been associated with the ratio of IR vibra-
tion due to Ti±O±Si bond at 930±960 cm
ÿ1
to that due
to Si±O±Si at ca. 1210 cm
ÿ1
[30,64]. However, IR
spectroscopy is not sensitive to the formation of TiO
2
crystallites.
Raman studies of TiO
2
±SiO
2
mixed oxides provide
additional evidence for the assignment of the vibra-
tional modes of the Ti±O±Si bond. The Raman spectra
of TiO
2
±SiO
2
mixed oxides show two bands at 935±
960 and 1100±1110 cm
ÿ1
[35,38,62,84,86,88], which
are associated with vibrational modes involving
Ti±O±Si bonding (the assignment of the vibrational
modes will be discussed in detail below in Sec-
tion 2.3). Raman spectroscopy, with especially a mul-
tichannel analyzer, is extremely sensitive to the
presence of TiO
2
crystallites. The formation of TiO
2
crystallites (anatase) can be recognized by a sharp
Raman band at 144 cm
ÿ1
with a minimum detect-
able amount of 0.05 wt% TiO
2
[85]. Interestingly, at Ti
content above 10 mol%, Best et. al [84] observed a
Raman band at 665 cm
ÿ1
, which was assigned to
Ti(IV) in ®vefold or sixfold coordination. An IR band
has been also observed at 665 cm
ÿ1
, and the authors
only assigned it to the Ti±O±Si vibration [87]. How-
ever, this particular vibrational band was never
observed by other authors [64,82,86,88].
XPS analysis also reveals the strong interaction
between TiO
2
and SiO
2
in the mixed oxides
[35,65]. The Ti 2p
3/2
binding energy (BE) at low
titania content (<10 wt%) is 460 eV, which is higher
than the value for pure TiO
2
(458.7 eV) [65]. This
upward shift has been explained by the increase in the
interatomic potentials due to the decrease of the
coordination number of Ti and the shortening of the
Ti±O bond, which suggests the insertion of Ti
4
cations into tetrahedral sites of the silica network
[65,67,90]. However, since the Ti atoms are less
electronegative and more polarizable than the Si
atoms, this shift could also be accounted for by the
increase of the effective positive charge on the Ti
atoms due to the Ti±O±Si bonds. This latter explana-
tion is supported by the observation that BE values of
both O1s and Si 2p shift downward, indicative of a
decrease in the effective positive charge on Si and an
increase in the negative charge on O due to the
formation of Ti±O±Si bonds [65]. As will be shown
later in Section 2.2, TiO
2
/SiO
2
supported oxides and
multilayers/thin ®lms also show similar shifts, sug-
gesting that the changes in BE values of Ti, Si and O
are associated with the formation of Ti±O±Si linkages
that are common in these systems.
29
Si NMR spectroscopy has been applied to study
the changes in the SiO
2
structure when Ti atoms are
incorporated into the silica network [26,35,61±63].
Broadband decoupled spectra yield the percentage of
Q
2
,Q
3
and Q
4
sites, where Q
n
denotes a
29
Si nucleu
with a Si(OSi)
n
(OX)
4ÿn
local environment (XHor
Ti) [62,63]. The contributions from Ti±O±Si bridges
for Q
2
and Q
3
sites were found to be dominant in the
sol±gel prepared mixed oxides [26,63]. However, this
technique cannot provide quantitative information
about the Ti±O±Si linkages because the
29
Si reso-
nances from Si±O±H and Si±O±Ti bonds are not
distinguishable. In addition, this technique is not
sensitive to the presence of crystalline TiO
2
[61±63].
Direct information about Ti±O±Si linkages in
TiO
2
±SiO
2
gels can be obtained with
17
O NMR spec-
troscopy [61,89]. A chemical shift at 280±260 is
observed for Ti±O±Si that is intermediate between the
chemical shifts of Si±O±Si and Ti±O±Ti. This reso-
nance is uniquely assigned to the oxygen atoms in
Ti±O±Si linkages, which offers the possibility to
determine the numbers of different oxide linkage
quantitatively [89].
All the above characterization results con®rm the for-
mation of Ti±O±Si linkages in TiO
2
±SiO
2
mixed oxides.
2.1.3. Characterization of the local structure of the
Ti atoms in the SiO
2
matrix
To determine the local structure of Ti(IV) in amor-
phous TiO
2
±SiO
2
mixed oxides (where TiO
2
crystal-
lites are not present to complicate the analysis), the
236 X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254
physical techniques used should possess the ability to
probe short-range order. Neutron diffraction and X-ray
diffraction [91,94], extended X-ray absorption ®ne
structure (EXAFS) and X-ray absorption near edge
spectroscopy (XANES) [27±29,68,81,84,85] and
UV±Vis [60,79,81,93,94] spectroscopy have been
employed to establish the coordination geometry of
the Ti atoms in the SiO
2
matrix.
Ti K-edge EXAFS/XANES has been shown to be
very powerful in studying the local structure of Ti in
amorphous TiO
2
±SiO
2
mixed oxides. From the
EXAFS experiment, the Ti±O, Ti±Si and Ti±Ti bond
lengths and the coordination number around Ti can be
obtained. The XANES pre-edge features provide addi-
tional formation about the coordination geometry of
Ti. Ti atoms in tetrahedral sites without inversion
symmetry exhibit strong pre-edge absorption features,
and Ti atoms in octahedral sites give rise to a very
small or no pre-edge absorption features. Early
EXAFS/XANES studies by Greegor et al. [68,95]
suggested that the Ti atoms in TiO
2
±SiO
2
glasses
possess predominantly tetrahedral coordination in
the range 0.05±9 wt% TiO
2
with a small amount of
Ti in octahedral coordination (less than 5% of the total
Ti atoms). Increasing concentration of TiO
2
up to
15 wt% increases the sixfold/fourfold ratio, and crys-
talline TiO
2
aggregates were found as a second phase
at ca. 15 wt% TiO
2
. They obtained an average bond
length of 1.81 A
Ê
for the tetrahedrally coordinated
Ti±O and 1.99 A
Ê
for the octahedrally coordinated
Ti±O [68]. Similar conclusions have been described
by Liu et al. [81] with an average Ti±O bond distance
of 1.82 A
Ê
at a low Ti content (Ti/Si1/8.2). Moreover,
a strong pre-edge peak intensity (58±75%) for these
TiO
2
±SiO
2
mixed oxides at low Ti contents further
con®rms that Ti atoms are located predominantly in
tetrahedral sites [68,81].
It should be pointed out that the pre-edge peak
intensity alone is not suf®cient to make the unambig-
uous assignment of the Ti coordination since similar
peak intensity can arise from Ti in fourfold and
®vefold coordinations as well as sixfold coordination
with Ti in highly distorted oxygen octahedra [96]. As
will be discussed in detail later (Section 2.3.2), both
the pre-edge position and the normalized height
should be used to correctly determine the Ti coordina-
tion. However, the combined results from both
EXAFS and XANES experiments, which reveal the
average short Ti±O distance of 1.80 A
Ê
and a strong
pre-edge peak, strongly suggest that the Ti atoms
mostly reside in the tetrahedral sites at low Ti content
(<15 wt% TiO
2
).
Rosenthal et al. [97] have performed a molecular
dynamics computer simulation study of TiO
2
±SiO
2
glasses. They obtained an average Ti±O bond length of
1.77 A
Ê
for the tetrahedral coordination and 1.96 A
Ê
for
the octahedral coordination, which is in agreement
with the above EXAFS results. Interestingly, the
results from the computer simulation also suggests
the presence of ®vefold coordination of Ti with an
average Ti±O bond length of 1.87 A
Ê
. However, a small
amount of ®vefold coordination of Ti is very dif®cult
to identify by EXAFS/XANES or other characteriza-
tion techniques because a mixture of fourfold and
sixfold coordination may give rise to similar features.
Walters et al. [91] applied neutron diffraction
experiment to study the local environment of Ti in
the silica matrix as well as the in¯uence of Ti on the
silica network. They found that at low Ti contents, Ti
is in fourfold coordination with an average Ti±O bond
length of 1.8 A
Ê
, consistent with the EXAFS results.
Moreover, the silica network is signi®cantly affected
with the Si±O bond-length distributions being nar-
rowed and a high level of network ±OH. The authors
argued that the effect of the larger Ti atom as com-
pared to the Si atom will produce some distortion of
the SiO
4
tetrahedral network, therefore increasing the
amount of network strain which may give rise to a high
±OH content when the strain is relieved. The authors
suggest that the most important effect of Ti on the
macroscopic structure of these materials is the effect
of the network-terminating/modifying of O±H bond.
Rigden et al. [92] have tried to use X-ray diffraction
to get some structural information about TiO
2
±SiO
2
mixed oxides. However, the bond length obtained for
the fourfold coordinated Ti atom is averaged over 1.4±
1.9 A
Ê
. Due to its low sensitivity, X-ray diffraction
cannot distinguish between ®rst-neighbor Si±O and
Ti±O distances and cannot measure quantitatively the
small differences in structure between the pure silica
and titania±silica [92].
UV±Vis spectroscopy is a very useful tool to inves-
tigate the band structure of TiO
2
±SiO
2
mixed oxides at
the molecular energy level. Blue shifts of bandgap
absorption edges are always observed at low Ti con-
tents. The contributions to the increase of the bandgap
X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254 237
energy may result from (i) the quantum size effect and
(ii) the matrix/support effect. The well known quan-
tum size effect is expected to be detectable for semi-
conductor titania at the nano-sized level [98]. In order
to discriminate between the two effects, it is important
to ®rst use other techniques such as Raman spectro-
scopy to identify the types of titanium oxide species,
otherwise, this technique may not provide very valu-
able information about the molecular structure of Ti
species due to the complexity introduced by simulta-
neously having multiple Ti species.
The UV±Vis DRS spectra may provide some infor-
mation on the ®rst and the second coordination
spheres of Ti. Following the literature [99], the charge
transfer transitions (LMCT) between the ligand
(XH±O
ÿ
, Si±O
ÿ
, Ti±O
ÿ
, etc.) and the empty d-
orbital of Ti
4
can be estimated from the optical
electronegativities of ligand X and Ti
4
by the
following equation:
cm
ÿ1
30 000
opt
Xÿ
opt
Ti: (1)
On one hand, the increase of the coordination of Ti
from tetrahedral to octahedral increases the
opt
(Ti)
value from 1.85 to 2.05 [99]. The LMCT band for Ti in
octahedral sites is supposedly at a lower wavenumber
compared to that of Ti in tetrahedral sites. For exam-
ple, the LMCT transitions of the well-documented
TS-1 are observed at 50 000±48 000 cm
ÿ1
that are
assigned to Ti in isolated tetrahedral sites [90], while
the LMCT transitions of Ti in octahedral sites of
anatase are often observed at 30 000 cm
ÿ1
. Therefore,
in some cases, the above equation could be used to
justify the coordination of Ti. On the other hand, the
coordination change is usually accompanied by a
change in the second coordination sphere (ligands).
For example, the coordination change due to hydration
of tetrahedral Ti sites is only partially responsible for
the downward shift of the LMCT band [81,93,100].
Unfortunately, the electronegativity values (
opt
(X))
of oxygenated ligands are quite arbitrary in the litera-
ture: Si±O
ÿ
(
opt
3.17) and H±O
ÿ
ligands (
opt
2.9)
[101], X±O
ÿ
ligands (
opt
3.45, XH±O
ÿ
and O
2ÿ
)
[93,100], H
2
O ligand (
opt
3.5) [101]. The ligand
change in the coordination sphere of tetrahedral Ti
from Si±O
ÿ
to H±O
ÿ
due to hydration is estimated to
redshift the LMCT band by 8000 cm
ÿ1
, according to
Eq. (1)[60,99]. However, the redshift of the LMCT
transitions of Ti observed for Ti-silicalite and TiO
2
±
SiO
2
mixed oxides [60,93,100] upon hydration is
usually in the range 1000±6000 cm
ÿ1
, and is a com-
bined effect of both the increased coordination num-
ber and the change in ligand. Thus, it is dif®cult for a
single value of the optical electronegativity to properly
represent the Ti atoms in different coordination and
ligand environments.
Furthermore, it has been shown that the ligand
environments of Ti can play a major role in deter-
mining the LMCT transitions [102]. Three Ti-refer-
ence compounds with different coordination and
ligands are compared. Ba
2
TiO
4
possesses isolated
TiO
4
tetrahedra linked by Ba atoms, JDF-L1
(Na
4
Ti
2
Si
8
O
22
.4H
2
O) and Ti-umbite (K
2
TiSi
3
O
9
.H
2
O)
contain isolated TiO
5
square pyramids and isolated
TiO
6
octahedra with Ti±O±Si linkages, respectively.
The LMCT transitions of Ti cations decrease in the
order: JDF-L1>Ti-umbite>Ba
2
TiO
4
, which is in con-
trast to the expectation that Ba
2
TiO
4
with fourfold
coordination would display the highest LMCT transi-
tions. The actual lowest LMCT transitions of the Ti
atoms in Ba
2
TiO
4
can be accounted for by Ba atoms as
the second coordination sphere that are much less
electronegative than Si atoms as the second coordina-
tion sphere of Ti atoms in JDF-L1 and Ti-umbite.
Therefore, it is practically not possible to apply
Eq. (1) to quantitatively calculate the shift of the
LMCT transitions upon the change in coordination
and ligand environments.
A comparative study on the molecular structures of
1.5 wt% TiO
2
±SiO
2
mixed oxide and TS-1 by UV±Vis
spectroscopy has been performed by On et al. [60] (in
both samples, Ti is in tetrahedral sites with the same
low Ti content). The LMCT transition of tetrahedral Ti
in the dehydrated TS-1 was deconvoluated into three
components centered at 50 200, 44 060 and
40 290 cm
ÿ1
[60,94]. Whereas the LMCT band for
the dehydrated TiO
2
±SiO
2
mixed oxide was centered
at 40 000 cm
ÿ1
, which is 10 000 cm
ÿ1
lower than
that of TS-1. Since the Ti atoms in both samples reside
in tetrahedral sites, the authors argue that a large
average Ti±O±Si bond angle of 1638 in TS-1 as
compared to 1598 in the mixed oxides [68,94] may
result in a higher donation, contributing to the higher
LMCT transition energy. Thus, they suggest that Ti in
TiO
2
±SiO
2
mixed oxides is located more in closed
sites Ti(OSi)
4
with a smaller bond angle, while Ti in
TS-1 is located more in open sites Ti(OH)(OSi)
3
.
238 X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254
However, Klein et al. [93] observed an absorption at
45 000 cm
ÿ1
for dehydrated TiO
2
±SiO
2
mixed oxides
at a low Ti concentration, and this band shifts slightly
downward to ca. 41 000 cm
ÿ1
with increasing Ti
concentration up to 9 mol%. This LMCT band was
simply assigned to the isolated tetrahedral Ti species.
The LMCT transition assigned to tetrahedral Ti spe-
cies in TiO
2
±SiO
2
mixed oxides [26,60,93] is often
observed at a lower wavenumber than in TS-1. It is
very unlikely that this difference is due to the so-called
open and closed sites, since ammonia adsorption
results show that a large fraction of tetrahedral Ti
atoms in TiO
2
±SiO
2
mixed oxides at low Ti contents
are located around the surface [12] which is the right
location for the open sites. However, the presence of a
small fraction of octahedral Ti species in the mixed
oxides even at very low Ti contents [12,68], together
with the possible presence of some Ti±O±Ti linkages
(no experimental results have shown that all Ti atoms
in TiO
2
±SiO
2
mixed oxides are isolated), might
account for this difference.
All results point toward the substitution of Ti as a
tetrahedral species for Si in the silica network at low Ti
content, and that the fraction of tetrahedral Ti atoms
decreases as the Ti content increases [26]. Further-
more, the degree of disorder of the Ti±O bond lengths
and Ti±O±Si angles seems substantially higher com-
pared to the framework Ti atoms in crystalline TS-1
[85].
The effect of water vapor on the coordination
geometry of the Ti atoms in atomically mixed
TiO
2
±SiO
2
oxides has been examined by XANES
and UV±Vis spectroscopies [60,81]. XANES experi-
ment demonstrated that the coordination environment
of Ti is dramatically affected by the coordination of
water vapor, since the pre-edge feature doubled after
dehydration [81]. UV±Vis spectroscopy also shows
the coordination change upon hydration since the
LMCT band for TiO
2
±SiO
2
mixed oxides shifts down-
ward for 1000 cm
ÿ1
[60]. In these cases the coordi-
nation of Ti is supposed to change from fourfold to
®ve or sixfold upon water addition.
2.2. TiO
2
/SiO
2
supported oxides
2.2.1. Preparation of TiO
2
/SiO
2
supported oxides
In order to improve the mechanical strength, ther-
mal stability and surface area of TiO
2
,TiO
2
/SiO
2
supported oxides have been considered as an
advanced support material to replace pure TiO
2
[36,37,46±48,71,72,75,76]. Unlike the substitution
of Ti for Si in the silica network in the mixed oxides,
the interaction of TiO
2
with the silica support is
limited to the surface, which is one of surface mod-
i®cations of silica by chemical reactions
[36,37,77,78,102].
It is well known that the silica surface is quite inert,
and it is very dif®cult to synthesize highly dispersed
metal oxides on the silica surface. Of all the chemical
modi®cation reactions on silica, the surface hydroxyls
generally act as the adsorptive/reactive sites because
of their hydrophilic character. Thus, the preparation of
highly dispersed metal oxides on silica by either
impregnation or the chemical vapor deposition
method often involves a highly reactive precursor,
such as TiCl
4
or Ti-alkoxides, to react with the surface
hydroxyls on silica [36,37,74,75,77±79,102]. The
titration of the surface hydroxyls with Ti-precursors
is either monofunctional (one Ti-alkoxide molecule
titrating one OH group) or bifunctional (one Ti-alk-
oxide molecule titrating two OH group) depending on
the pretreatment temperature, the reaction tempera-
ture, and the size and reactivity of the precursor
[77,78,102]. Thus, the Ti atoms bind to the silica
surface via oxygen as a bridge. Two types of Ti
species, highly dispersed surface TiO
x
species and
TiO
2
crystallites, are possibly present on the silica
surface, depending on the preparation conditions and
chemical compositions [37,75,77,102].
The dispersion capacity is closely related to the
concentration of hydroxyls on the silica surface and
the preparation conditions (e.g., the pretreatment tem-
perature, the reaction/impregnation time, the reactiv-
ity and molecular size of the precursor)
[37,75,77,78,102]. A maximum dispersion of TiO
2
was reached at 3.0 Ti/nm
2
on a non-porous SiO
2
(Sto
È
ber silica spheres) due to its highest concentration
of surface hydroxyls and highest accessibility to the
reagent [37]. However, a recent study [102] shows that
a maximum capacity of highly dispersed TiO
x
species
on silica can be reached at 4.0 Ti atom/nm
2
by
carefully controlling the above preparation variables.
This maximum capacity seems correlated with the
surface concentration of OH groups on silica, suggest-
ing a maximum coverage of surface TiO
x
species on
the silica surface.
X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254 239
2.2.2. Experimental evidence for Ti±O±Si bonding
In contrast to the large number of research papers on
TiO
2
±SiO
2
mixed oxides, the characterization studies
on TiO
2
/SiO
2
supported oxides are very limited. Fer-
nandez et al. [69] performed a comparative study of
small colloidal TiO
2
particles and 12 wt% TiO
2
/SiO
2
supported oxides. The TEM micrographs show that
both samples consist of very small particles (d<3 nm).
A similar blueshift in the bandgap absorption edge of
both samples was observed by UV±Vis spectroscopy,
which is attributed to the quantum size effect. How-
ever, XPS analysis shows that a signi®cant difference
of ÿ1.1 eV in the relaxation energy (the Auger para-
meter) with respect to the TiO
2
bulk oxide was
observed for the TiO
2
/SiO
2
supported oxides as com-
pared to the difference of only ÿ0.3 eV created by
colloidal TiO
2
due to the particle size effect. The
authors proposed that the SiO
2
support plays a more
important role than the particle size and/or water
content in decreasing the extra-atomic relaxation
energy of the photohole, in agreement with the for-
mation of Ti±O±Si bonds.
Raman studies of the dehydrated, dispersed TiO
2
/
SiO
2
supported oxides (TiO
2
loading <15 wt%) exhi-
bit two bands at 950 and 1080 cm
ÿ1
[102], similar to
that of TiO
2
±SiO
2
mixed oxides, which is indicative of
the formation of Ti±O±Si bonds. Furthermore, XPS
analysis shows that the BE value of Ti 2p
3/2
electrons
for the dispersed TiO
2
/SiO
2
supported oxides
increases by about 0.8±1.9 eV [75] or 0.5±1.0 [102]
as compared to pure TiO
2
, while the BE value of O 1s
electrons in Ti±O±Si is an intermediate value between
Si±O±Si and Ti±O±Ti. These results strongly suggest
the formation of Ti±O±Si bonds in TiO
2
/SiO
2
sup-
ported oxides.
2.2.3. TiO
2
/SiO
2
multilayers/thin films
TiO
2
/SiO
2
multilayers/thin ®lms can be regarded as
one type of TiO
2
/SiO
2
supported oxides. When TiO
2
layers are thin enough on the silica substrate, a highly
dispersed titanium oxide species may be present. The
physico-chemical nature of this extremely thin layer
should be the same as that of the highly dispersed
TiO
2
/SiO
2
supported oxides. TiO
2
/SiO
2
multilayers/
thin ®lms are often prepared by the deposition of
evaporated titanium metal [103,104] or oxides
[105,106] on the silica substrate, or by precipitation
of hydrolyzed titanium alkoxide precursors [71,72].
Titania thin ®lms with different morphology have
been obtained by controlling preparation parameters,
which is con®rmed by small angle X-ray scattering
(SAXS) and transmission electron microscopy (TEM)
[71].
IR studies of TiO
2
/SiO
2
multilayers/®lms show a
band at 930 cm
ÿ1
, which was assigned to the vibration
of Ti±O±Si bonds [105]. This band increases with
decreasing the thickness of the TiO
2
multilayer ®lms
(down to 0.3 nm), while the crystallization of TiO
2
is
lost. Also, with decreasing TiO
2
®lm thickness, an
increase in surface hydration was observed [105],
indicating an increase in the number of bonding
defects and a similar surface behavior as the TiO
2
/
SiO
2
supported oxides.
Studies of the interface region of TiO
2
/SiO
2
multi-
layers by electron energy loss spectroscopy (EELS)
reveal that some EELS ®ne structures cannot be ®tted
to a combination of reference spectra of pure TiO
2
and
SiO
2
, but are representative of hybrid environments of
Ti±O±Si. A zone of about 5 nm width at the interface
was estimated to contribute to the presence of a TiO
2
/
SiO
2
solid solution with Ti±O±Si bonds [106].
The formation of Ti±O±Si bonds strongly modi®es
the electronic structure of the oxygen and titanium
atoms at the interface of TiO
2
/SiO
2
multilayers. At
low Ti coverage, XPS analysis detected a new O 1s
peak at 531.2 eV, which is between Si±O±Si
(532.9 eV) and Ti±O±Ti (530.7 eV), suggesting the
formation of Si±O±Ti bonds [103]. In addition, a
signi®cant blueshift in the bandgap edge of the sup-
ported TiO
2
thin ®lms was observed by UV±Vis
spectroscopy, which has been attributed to a combina-
tion of the size quantization effect and the substrate
effect [103], with the substrate effect being more
important [104].
2.2.4. Characterization of the molecular structure
of the surface Ti atoms on SiO
2
Since the Ti atoms in TiO
2
/SiO
2
supported oxides
are located at the surface of silica, a higher Ti±OH
concentration is expected due to a higher degree of
coordinative unsaturation at the surface. The Ti±OH
groups due to highly dispersed TiO
x
species on the
silica surface have been observed by FTIR and
1
H MAS NMR spectroscopy [77]. However, these
hydroxyls are not stable and begin to dehydroxylate
at a temperature of 1808C [77]. The molecular struc-
240 X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254
ture of the dispersed TiO
x
species on silica in the
hydrated state has been examined by EXAFS/XANES
[8]. The pre-edge feature and the EXFAS data suggest
that the Ti atoms are ®vefold coordinated mononuc-
lear and binuclear species with 2±3 Ti±OH hydroxyls.
However, this assignment is quite questionable, and
will be discussed later in Section 2.3.2.
Systematic investigation of the surface structure of
dispersed TiO
2
/SiO
2
supported oxides under hydrated
and dehydrated conditions has recently been con-
ducted by Gao et al. [102]. They employed in situ
UV±Vis DRS and XANES spectroscopies to provide
more reliable structural information. It was found that
the surface structure of TiO
x
species on silica is a
strong function of environmental conditions as well as
the TiO
2
loading. In the dehydrated state, 1.05 wt%
TiO
2
/SiO
2
is predominantly composed of isolated
TiO
4
units (the LMCT band 47 600 cm
ÿ1
),
6.58 wt% TiO
2
/SiO
2
of dimeric or one-dimensional
polymerized TiO
4
units (the LMCT band
40 600 cm
ÿ1
), and 14.75 wt% TiO
2
/SiO
2
of two-
dimensional, polymerized TiO
5
units (the LMCT band
39 000 cm
ÿ1
).
Pronounced structural changes were observed upon
hydration [102]. A Raman band appears at 940±
960 cm
ÿ1
, which is most likely due to the Ti perturbed
Si±OH. The adsorption of water molecules breaks the
Ti±O±Si bridging bonds, resulting in the formation of
Ti±OH hydroxyls as well as the Ti perturbed Si±OH.
XANES analysis demonstrated that hydration appears
to increase the average coordination number of the
surface Ti cations by 1 for the dispersed TiO
2
/SiO
2
supported oxides [102].
2.3. Structural difference and similarity between
titania±silica mixed and supported oxides
2.3.1. Comparison of various characterization data
The most important spectroscopic features asso-
ciated with the atomically mixed TiO
2
±SiO
2
oxides and the highly dispersed TiO
2
/SiO
2
sup-
ported oxides are summarized in Table 2.
Characterization data on the relatively well-docu-
mented TS-1 (Ti-silicate) are shown for reference.
The comparison between the characterization data
on these materials could shed some light on under-
standing the behavior of Ti atoms in different local
environments.
As shown in Table 2 XPS analysis of the three types
of materials indicates that the BE value of the Ti 2p
3/2
electrons signi®cantly increases when connected with
different silica matrices. This result can be explained
in terms of the increase of the effective positive charge
on the Ti atoms due to the Ti±O±Si linkages, as
discussed previously. The large shift of the BE value
of Ti 2p
3/2
electrons (1 eV) is usually associated
with the isolated TiO
4
sites that possess maximum
number of Ti±O±Si bonds per Ti atom.
In agreement with the XPS data, similar vibrational
bands are observed on mixed and supported titania±
silica oxides as well as TS-1, suggesting that these
vibrations can only be safely associated in the pre-
sence of Ti±O±Si linkages. The assignment of the IR
band at 960 cm
ÿ1
and its Raman counterpart at
960 cm
ÿ1
in Ti-silicalites has been discussed in
detail in the literature. These are due to silica vibra-
tions perturbed by the presence of Ti and are indicative
of the formation of Ti
d
±O
dÿ
±Si bonds [38,107]. The
Raman band at 1080±1200 cm
ÿ1
has been assigned to
the stretching modes of other Si±O
dÿ
bonds indirectly
perturbed by the presence of Ti [107]. A more detailed
assignment has been given by Deo et al. [38]. They
assigned this band to the SiO
4
unit containing two
non-bridging oxygens as in the case of some silicate
glasses. Since EXAFS analysis shows that there are no
[TiO
x
] units sharing edges with [SiO
4
] units [96] and
Ti is connected to four [SiO
4
] units in Ti-silicalite
[108], the only possible presence of a SiO
4
unit
with two non-bridging oxygens is in the structural
unit of Ti±O±Si±O±Ti. However, there are no avail-
able techniques capable of detecting the presence of
Ti±O±Si±O±Ti linkages. In fact, the Ti atoms that
perturb the vibrational modes of the [SiO
4
] units are
not necessarily located in a tetrahedral environment,
since the coordination changes due to the addition of
NH
3
and H
2
O cause only slight changes in the vibra-
tional spectra [38,107].
The comparison of the UV±Vis results of the three
types of materials indicates that titania±silica mixed
and supported oxides at low Ti contents (<5%) pre-
dominantly possess isolated TiO
4
species (the LMCT
band at 45 000±50 000 cm
ÿ1
). Nevertheless, the
LMCT band of TS-1 is usually higher than these of
titania±silica mixed and supported oxides, suggesting
that isolated TiO
4
sites in titania±silica mixed and
supported oxides even at low Ti contents are not
X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254 241
unique, and a small amount of polymerized Ti species
may also be present. The maximum LMCT band of
titania±silica mixed and supported oxides redshifts
with increasing Ti content [93,102], suggesting an
increase in the polymerization degree of Ti atoms
[102]. The maximum polymerization degree of Ti
atoms is observed on the monolayerly dispersed
TiO
2
/SiO
2
supported oxides as extended two-dimen-
sional, polymerized TiO
5
units. Moreover, UV±Vis
results indicated that hydration increases the total
number of Ti±O±Ti bonds (polymerization degree)
of TiO
2
/SiO
2
supported oxides at high loadings
(>5 wt%TiO
2
) [102]. Similarly, Klein et al. [93] also
observed an increasing tendency for the formation of
Ti±O±Ti bonds through hydration at higher Ti-con-
tents in the mixed TiO
2
±SiO
2
oxides.
2.3.2. Are the structural assignments reasonable?
Most of the structural information about titania±
silica comes from Ti K-edge EXAFS/XANES studies.
However, the ®tted value of the coordination number
from EXAFS data often shows a relatively large
variation, even in the case of TS-1 (0.5±0.6)
[85,96], which gives rise to uncertainty in the predic-
tion of the coordination number of the Ti atoms in the
titania±silica system. Therefore, it is necessary to
develop a reliable approach to verify the feasibility
of structures proposed by EXAFS/XANES studies.
A general empirical relationship for relating the
Ti
4
±O bond strength s
i
to bond length R (A
Ê
) has been
developed by Brown and Wu [109] based on the data
from many Ti±O-containing compounds with Ti
4
cations in different environments:
s
i
vuR=1:806
ÿ5:2
vu valence unit: (2)
According to the valence sum rule, the sum of each
individual Ti±O bond valence relating to a Ti
4
site is:
X
s
i
vu4:00 0:20 vu: (3)
This bond valence±bond length correlation can
provide a more stringent constraint on the local envir-
onment around Ti in Ti-silicates, titanates and Ti±O-
containing compounds. By using the Ti±O bond
lengths derived from Ti K-edge EXAFS data, the
valence states of Ti
4
cations in some compounds/
Table 2
Comparison of various characterization data on TS-1, TiO
2
±SiO
2
mixed oxides and TiO
2
/SiO
2
supported oxides
Techniques TS-1 TiO
2
±SiO
2
mixed oxides TiO
2
/SiO
2
supported oxides Reference
XPS
BE (Ti 2p
3/2
)
a
1.5 1.2±1.3 eV 0.8±1.9 eV [65,67,75,126]
IR
Ti±O±Si vibration 960 cm
ÿ1
910±960 cm
ÿ1
930 cm
ÿ1
[85,105,107]
Raman 960 cm
ÿ1
935±960 cm
ÿ1
950 cm
ÿ1
Ti±O±Si vibration 1125 cm
ÿ1
1100±1110 cm
ÿ1
1080 cm
ÿ1
[38,85,102]
UV±Vis
LMCT peak (dehydrated) 45 000±50 000 cm
ÿ1
40 900±45 000 cm
ÿ1
39 000±47 900 cm
ÿ1
[85,88,102]
XANES (dehydrated)
b
Pre-edge peak intensity 75% 58% 65% [85,102]
Pre-edge peak position
c
4969.7 eV 4969.7 eV 4969.5 eV
EXAFS
Ti±O bond length (average) 1.80±1.81 A
Ê
1.81±1.82 A
Ê
1.81 A
Ê
[68,81,85,96]
Ti±O±Si bond angle 1638 1598 ± [108,115]
a
BE (Ti2p
3/2
) corresponds to the increase of the BE in Ti±O±Si bonds as compared to Ti±O±Ti bonds.
b
Samples with low Ti contents (<2 wt% TiO
2
) are used for comparison.
c
Pre-edge peak position given using first inflection point of Ti foil at 4966.0 eV.
242 X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254
oxides with different coordinations are calculated and
shown in Table 3. The Ti
4
cations in the compounds/
oxides listed there, possess a total valence state
between 3.68 and 4.18 vu. The average Ti±O bond
lengths for the different coordinations are 1.80±1.84 A
Ê
for fourfold, 1.89±1.91 A
Ê
for ®vefold, and 1.95±
1.96 A
Ê
for sixfold, as listed in Table 3. It should be
mentioned that for ®vefold coordinated compounds/
oxides, it does not matter whether Ti is in square
pyramidal [110,111±113] or in distorted trigonal
bipyramidal coordinations [114], the valence state
of Ti and the average Ti±O distance always ®t into
the expected range. Thus, it appears that the total
valence state of a Ti
4
cation obtained from the
Ti±O bond lengths and the average Ti±O bond length
can be used to justify the local structure of the Ti
atoms in various environments.
The valence states and the average bond lengths for
TS-1 and the atomically mixed TiO
2
±SiO
2
oxides are
consistent with the standard set for a tetrahedral
coordination. The Ti atoms in TS-1 are isolated as
demonstrated by EXAFS experiment [96,108]. For
atomically mixed TiO
2
±SiO
2
oxides, although it is
believed that the Ti atoms are also isolated, the
experimental evidence seems insuf®cient since the
Ti±Ti distances have not been reported in the litera-
ture.
The only exception to the valence sum rule in
Table 3 is the result reported by Yoshida et al. [8]
for the dispersed TiO
2
/SiO
2
supported oxides in
hydrated conditions. The valence state for a Ti
4
cation is calculated to be 4.98 vu, which is signi®-
cantly out of the acceptable range. Moreover, the
average Ti±O bond length is very short (1.82 A
Ê
),
which is within the range of fourfold coordination
instead of ®vefold coordination. This discrepancy
might be accounted for by the poor quality of the
®tting values and incorrect assignment of the Ti
coordination. In contrast, a well-performed EXAFS
study by Maschmeyer et al. [115] on the TiO
2
/MCM-
41 (mesoporous silica) supported oxides indicates that
the valence state and the average bond-length of Ti
4
cations are in excellent agreement with the standard
set for a tetrahedral coordination. Their EXAFS
experiment shows that the fourfold coordinated Ti
atoms are located at the wall of mesopores, and act
as the active centers for the epoxidation reaction.
The pre-edge features in XANES spectra have been
regarded as a supplementary evidence for the presence
of certain coordination geometry based on the com-
parison with the model compounds. Furthermore, the
absolute position and intensity of pre-edge peak have
been considered to be the key factors in determining
the coordination number of Ti [111,112,113]. The pre-
Table 3
Valence state of Ti
4
cation in various Ti±containing materials obtained from Ti±O bond lengths derived from Ti K-edge EXAFS data
Ti±O bond length (A
Ê
) Average Ti±O (A
Ê
) Valence state
P
s
i
(vu) Reference
6-Coordinated
TiO
2
(anatase) 1.934, 1.982 1.95 4.04 [96]
TiO
2
(rutile) 1.954, 1.982 1.96 3.93 [96]
5-Coordinated
Na
4
Ti
2
Si
8
O
22
.4H
2
O 1.701, 1.964 1.91 3.98 [110]
Rb
2
Ti
4
O
9
1.651, 1.984 1.91 4.08 [111,112]
Ti(OEt)
4
and Ti(OBu
n
)
4
1.803, 2.052 1.90 4.09 [114]
Ti±silicate/aluminosilicates (1.67±1.70)1, (1.94±1.95)4 1.89 4.13±4.18 [113]
TiO
2
/SiO
2
supported oxide (hydrated) 1.753, 1.861, 2.001 1.82 4.98(?) [8]
4-Coordinated
Ni
2.6
Ti
0.7
O
4
1.844 1.84 3.68 [111,112]
Ti(OAm
t
)
4
and Ti(OPr
i
)
4
1.804 1.80 4.07 [114]
TS-1 (1.80±1.81)4 1.80±1.81 4.07±3.95 [94,96]
TiO
2
/MCM-41 (dehydrated) 1.814 1.81 3.95 [115]
TiO
2
±SiO
2
mixed oxide/glasses (1.81±1.82)4 1.81±1.82 3.95±3.84 [68,81]
X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254 243
edge features with the highest intensity at the lowest
energy correspond to Ti in tetrahedral sites and the
opposite pattern holds for Ti in octahedral sites. As
listed in Table 2, the pre-edge position is about the
same for the three different types of materials with
similar low Ti contents, however, the pre-edge inten-
sity is the highest for TS-1. The slightly lower pre-
edge intensity for titania±silica mixed and supported
oxides implies that their average coordination number
and/or the distortion of coordination geometry are
slightly higher than that of TS-1.
It is concluded for both atomically mixed
TiO
2
±SiO
2
oxides and molecularly dispersed
TiO
2
/SiO
2
supported oxides that at low Ti contents,
the Ti atoms predominantly reside in isolated tetra-
hedral sites with possibly a small amount of higher
coordination Ti sites. The isolated TiO
4
sites are
coordinatively unsaturated and are subject to change
upon hydration. Polymerization of Ti atoms occurs at
high Ti contents, which results in an increase in the
average coordination number of Ti atoms.
2.3.3. Location of the Ti atoms
The location of the Ti atoms is a very important
issue because catalysis only takes place at the surface.
For TiO
2
/SiO
2
supported oxides, there is no doubt that
the Ti atoms are mostly located on the outer surface of
the silica support. For TiO
2
±SiO
2
mixed oxides, how-
ever, not all the Ti sites are accessible [12]. As will be
shown later, TiO
2
±SiO
2
mixed oxides not only possess
Lewis acidity but also generate new Brùnsted acid
sites. The location of these Ti atoms should be on the
surface of micro, meso- or macro-pores that are
accessible to the reactants or probe molecules. Since
the surface Ti sites are potential adsorption sites for
some reactants such as alcohols [19,116] and for probe
molecules such as NH
3
and pyridine [12,19], it is
possible to employ some selective adsorbing mole-
cules to measure the number of surface Ti atoms.
Evidently, the size and adsorption selectivity of the
probe molecules are crucial for determining the right
number of surface Ti atoms. However, up to now, no
suitable method has been established due to the
incomplete understanding of the structure±adsorption
relationships between the mixed oxides and the probe
molecules.
As discussed previously, the maximum Si/Ti atomic
ratio for the atomically mixed TiO
2
±SiO
2
is ca. 7.5,
which corresponds to 3±4 Ti atoms/nm
2
if we consider
that all the Ti atoms are located around the surface
with an average surface area of 300±400 m
2
/g. It is
possible that most of the Ti atoms in the atomically
mixed TiO
2
±SiO
2
oxides reside near the surface, since
the maximum loading for the molecularly dispersed
TiO
2
/SiO
2
supported oxides is ca. 4 Ti atoms/nm
2
. Liu
et al. [12] reported that a large fraction of Ti atoms in
the TiO
2
±SiO
2
mixed oxides with Si/Ti molar ratio of
8.2 is located around the surface (corresponding to ca.
2.6 Ti atoms/nm
2
).
Mukhopadhyay and Garofalini [67] have pointed
out that a surface rich in Ti is energetically less
favorable than rich in Si. XPS and XANES studies
of TiO
2
±SiO
2
mixed oxide prepared by coprecipita-
tion reveal that the surface is enriched in Si and the
degree of enrichment increases with Ti content [65].
As will be discussed later (Section 3.2.2), the acidity
studies by ammonia and pyridine adsorption demon-
strate that the surface enrichment of Ti or Si is a strong
function of preparation procedures and chemical com-
positions. In general, TiO
2
±SiO
2
mixed oxides with
a high Ti content exhibit surface enrichment of Si.
In fact, the signi®cantly higher surface area of
TiO
2
±SiO
2
mixed oxides (even with a very low
SiO
2
content) than the surface area of pure TiO
2
may also be accounted for by the surface enrichment
of Si atoms that locate near the surface to stabilize the
material from sintering.
It is interesting to note that no matter what the
preparation methods (mixed oxides or supported
oxides) or chemical compositions are employed, the
Ti/Si atomic ratios obtained by XPS experiments
are often lower than the corresponding calculated
bulk ratios [28,65,74,75,102]. It is hard to believe
that the surface enrichment of Si that might occur
in mixed oxides could also occur on the supported
oxides where the highly dispersed TiO
x
species are
supposed to anchor on the silica surface via its
outermost layer of hydroxyls. This implies that either
the Ti atoms are located on the surface of inside
channels or pores of silica that is out of the XPS
detection range, or there might be some type of
experimental error that underestimates the concentra-
tion of Ti versus Si. Therefore, XPS analysis may not
be a suitable technique for providing reliable quanti-
tative information about the surface enrichment of
Ti or Si.
244 X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254
3. Structure±property relationships for
titania±silica as catalysts
3.1. Electronic properties and photocatalysis
3.1.1. Quantum-size effect and support effect
Silica has been widely used as a substrate or
matrix to disperse and stabilize nano-sized TiO
2
par-
ticles [1,3,5,6,7,103,117]. The UVabsorption edge, or
optical band-gap, is a strong function of the TiO
2
particle size with diameters less than 4 nm. These
small particles consist of a few hundred TiO
2
mole-
cules [98], which are believed to exhibit the well-
known quantum-size effect. The quantum-size effect
of TiO
2
on/in SiO
2
has been repeatedly reported
[1,74,103,117]. The study of the TiO
2
/SiO
2
multilayer
®lms by Nakayama et al. [117] showed that the TiO
2
size change even in one dimension also shows the
quantum-size effect. Nakayama et al. found that the
optical energy gap E
g
remains constant until the
thickness of TiO
2
®lms on SiO
2
are reduced to less
than 2 nm, then E
g
increases as the ®lm thickness
further decreases.
As discussed previously, the Ti±O±Si bonds
strongly modify the electronic structure of the Ti
atoms in both titania±silica mixed and supported
oxides, and increase the effective positive charge on
the Ti atoms. As pointed out by Lassaletta et al.
[103,104], the increase in the band-gap energy could
be attributed to a combination of the quantum-size
effect and the interface interaction due to the support
effect, with the support effect probably being the most
important.
TiO
2
(anatase) has a band-gap of ca. 3.3 eV, while
the band-gap for titania±silica can be increased up to
4.1 eV due to the above two effects [103]. An
increase in the band-gap of the nano-sized TiO
2
particles has been ascribed to an elevation of the
conduction band edge as well as the lowering of the
valence band edge [98,118]. Thus, the electronic
properties are changed due to the increased band-
gap energy. The oxidizing potential of the photon
generated holes (h
) and the reducing potential of
the photon generated electrons (e
ÿ
) will increase with
increasing band-gap, as shown in Scheme 1.
Therefore, the nano-sized TiO
2
particles in or on
SiO
2
possess high photo-oxidation as well as photo-
reduction capabilities.
3.1.2. Photocatalysis
Since the nanosized TiO
2
particles possess a larger
band-gap with a higher photo-oxidation capability, the
strong oxidizing potential of the photogenerated holes
has made titania±silica one of the most attractive
photocatalysts for oxidation reactions. For example,
TiO
2
±SiO
2
gel has been used for decontamination of
aquatic environment [1]. The rate of photodecomposi-
tion of organic pollutants was reported to increase as
the TiO
2
particle size decreased, which is in agreement
with the increasing oxidation potential of the oxidiz-
ing holes.
In addition to the improved oxidizing abilities,
some synergism between the TiO
2
and SiO
2
phases
has been noticed by Anderson and Bard [5,6]. They
found that TiO
2
±SiO
2
gel shows a higher photode-
composition rate of rhodamine-6G (R-6G) as com-
pared to TiO
2
(P-25) [5]. Since R-6G adsorbs on SiO
2
sites, but not on TiO
2
, the presence of an adsorbent
(SiO
2
) was considered to promote the ef®ciency by
increasing the concentration of R-6G near the TiO
2
sites relative to the solution concentration of R-6G.
They suggest that the photogenerated intermediate
oxidants on the TiO
2
sites, such as hydroxyl radical
(HO
) or species from reduction of O
2
(HO
2
), must
diffuse and react with R-6G on the SiO
2
sites. Similar
synergy mechanism, but with the organic mole-
cules being activated in the TiO
2
±SiO
2
region, has
been suggested to explain the increased activity of
the photocatalytic decomposition of phenol on
TiO
2
±SiO
2
gel [6].
The strong reduction potential of nano-sized TiO
2
in sol±gel prepared TiO
2
±SiO
2
mixed oxides has been
observed by Inoue et al. [3]. TiO
2
±SiO
2
mixed oxides
shows higher activities for photocatalytic reduction of
CO
2
to formate, methane and ethylene compared to
pure TiO
2
. The authors demonstrated that the activities
(i.e., quantum ef®ciencies) of TiO
2
±SiO
2
mixed
Scheme 1.
X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254 245
oxides increase with decreasing size of the TiO
2
microcrystals.
However, the above results would have made more
sense if the total effective surface area of the nano-
sized TiO
2
particles could be known. A method based
on the selective adsorption of isopropanol for measur-
ing the effective surface area of titania on silica, has
been suggested by Biaglow et al. [116]. They found
that the total effective titania surface area measured on
titania±silica could be three times higher than that of
pure TiO
2
. This fact implies that the increased surface
area due to the decreased particle size may also
contribute to the increased photocatalytic activity.
Therefore, it is not quite clear from the above results
whether the enhanced photocatalytic activity is due to
the increased effective TiO
2
surface area or due to the
increased oxidizing or reducing potentials.
3.2. Acidic properties and acid catalysis
3.2.1. Coordination geometry and generation of
new Brùnsted acid sites
Titania±silica has been employed for a number of
acid catalyzed reactions such as isomerization and
dehydration, see Table 1. The acidic properties of
titania±silica are quite different from that of either
pure titania or pure silica, since pure titania only
possesses Lewis acidity while silica has neither
Brùnsted nor Lewis acidity. However, new Brùnsted
acid sites are created when titania and silica form
Ti±O±Si chemical bonds [12,18,19,119,120].
Several models have been proposed to explain the
generation of the new acid sites [118,119,121]. A
charge imbalance localized at the Ti±O±Si bond has
been attributed to the generation of new acid sites
[119,121]. Tanabe et al. [121] ®rst proposed a model
for mixed binary oxides based on two hypotheses.
According to their hypotheses, the coordination num-
bers of Ti and Si are maintained at 6 and 4, respec-
tively, as in the pure oxides. In addition, oxygen
assumes the coordination number of the major com-
ponent oxide, i.e., in Ti-rich mixed oxides, the coor-
dination number for all oxygens is 3, while in Si-rich
mixed oxides, the coordination number is 2. There-
fore, in Ti-rich mixed oxides, Si is the acid site and an
excess of the positive charge of 4/3 distributed on
four Si±O bonds is produced and assumed to generate
Lewis acidity. While in Ti-rich mixed oxides, Ti is the
acid site and an excess of the negative charge of -2
distributed on six Ti±O bonds is produced and
assumed to generate Brùnsted acidity. However, Tana-
be's model is not consistent with more recent struc-
tural characterization data. As described previously, in
Si-rich mixed oxides, the coordination number of the
Ti atoms that substitute into the silica network is 4, not
6. Moreover, the Ti-rich mixed oxides show a sub-
stantial amount of new Brùnsted acid sites that cannot
be explained by this model [12,19].
Nakabayashi et al. [118] explained the generation of
new and strong Lewis acid sites on ®nely divided pure
TiO
2
by the quantum-size effect. They demonstrated
that the acid strength of TiO
2
particles increases as the
particle size decreases, which is explained in terms of
the lowering of the valence band edge associated with
the enlarged bandgap due to the quantum-size effect.
However, this effect could not account for the forma-
tion of new Brùnsted acid sites in TiO
2
±SiO
2
mixed
oxides [120], and Nakabayashi has to attribute the new
Brùnsted acidity also to the presence of Ti±O±Si
bonds.
Kataoka and Dumesic [119] proposed a model
based on the studies of silica-supported metal oxides
by pyridine adsorption with infrared spectroscopy.
They found that the strength of Brùnsted acidity on
1 wt% TiO
2
/SiO
2
is rather weak. When the sample
was evacuated at 420 K, only Lewis acid sites were
detected. Brùnsted acid sites were generated when
water vapor was introduced at room temperature.
When TiO
2
/SiO
2
was reduced, no Brùnsted acidity
was observed even in the presence of water vapor.
According to Pauling's electrostatic valence rule, they
stated that the degree of undersaturation for surface
oxygens is dependent only on the valence and the
coordination of the cations, and the terminal hydroxyls
can never be acidic. The Brùnsted acid sites are
generated on M±O±Si bridges to compensate for the
undersaturated bridging oxygen as in the case of
silica±alumina. According to this model, Ti in tetra-
hedral coordination is not expected to generate
Brùnsted acidity because the undersaturation of oxy-
gen in Ti±O±Si bond is zero. However, when octahe-
dral coordinated Ti is produced in the presence of
water vapor, an undersaturation of 0.33 valence unit
(vu) for Ti±O±Si bridging oxygen is resulted, and a
proton is needed to saturate such oxygen, therefore,
Brùnsted acidity is generated. With this model, the
246 X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254
undersaturation for Ti±O±Si bridging oxygen with Ti
in pentahedral coordination is estimated to be 0.2 v.u.,
and Brùnsted acidity should also be observed. Evi-
dently, the Brùnsted acidity on titania±silica is
expected to be rather weak due to the change of
coordination geometry of Ti upon dehydration as well
as the poor stability of hydroxyls associated with Ti.
In both Tanabe's and Kataoka and Dumesic's
model, the difference in coordination geometry of
Ti from Si in Ti±O±Si bond plays a vital role in
causing a charge imbalance to generate Brùnsted acid
sites. Kataoka and Dumesic's model seems more
consistent with the experimental results. This model
predicts that:
1. Brùnsted acid sites are associated with Ti±O±Si
bridges where the Ti atoms reside not in
tetrahedral sites but in pentahedral or octahedral
sites, regardless of composition;
2. the coordination change of the Ti atoms upon
hydration will generate weak Brùnsted acid sites.
These predictions are quite consistent with recent
results by several research groups [12,59,80]. Liu et
al. [12] reported that the Brùnsted acidity in atom-
ically mixed TiO
2
±SiO
2
oxides is correlated with the
non-tetrahedral sites. In addition, surface hydroxyla-
tion is crucial to the presence of Brùnsted acidity
[19,59]. In this model, the bridging Ti±OH±Si hydro-
xyls are supposed to be the Brùnsted acid sites, in line
with the generation of Brùnsted acid sites in alumina±
silica zeolites.
More recently, Liu et al. [12] modi®ed Tanabe's
model to explain the generation of Brùnsted acid sites
in Ti-rich mixed oxides. They proposed that the
positive charge (4/3) on Si is balanced by a hydroxyl
group on Si, thus producing Brùnsted acidity. In
contrast, Contescu et al. [59] proposed that the acidic
hydroxyl might be located on a titanium surface ion.
Although it is widely accepted that the Brùnsted acid
sites are associated with the Ti±O±Si bridges, how-
ever, the exact location of the proton is still open to
question.
3.2.2. Acidic properties and isomerization/
dehydration reactions
To understand the acidity characteristics of titania±
silica better, the relationship between the acidic prop-
erties (e.g., the strength, the amount and the type of
acid sites) and the acid catalyzed isomerization and
dehydration reactions will be discussed below.
Because of the great concern about the possible
residual contaminants of sulfate and/or alkali (Na
)
that might signi®cantly modify the acidity of the
catalyst, the acidic characterization results obtained
from the samples prepared from precursors such as
titanyl oxysulfate and sodium/potassium silicates will
not be considered in this paper.
The acidity (in terms of both density and strength)
of solid catalysts is usually measured by TPD of
ammonia or pyridine, IR studies of ammonia or
pyridine adsorption, and amine titration with Hammett
color indicators. The experimental determination of
acidity is a complex issue since the coordination
geometry of Ti that determines the acidity of mixed
oxides is different under different experimental con-
ditions. The amine titration with Hammett indicators
in non-aqueous solvents at room temperature mea-
sures the ability of oxides to protonate various basic
indicators. The acidity measured in this way cannot
re¯ect well the acidity under speci®c reaction condi-
tions, therefore, this type of acidity measurement will
not be cited here. TPD and IR studies of ammonia/
pyridine adsorption can measure acidity under condi-
tions closer to reaction conditions and can provide
valuable information about the acidity pro®le of
mixed oxides when related to the acid catalyzed
reactions, and these results will be discussed here.
The 1-butene isomerization and alcohol dehydration
reactions occur at mild temperatures and are fre-
quently employed to probe the acidic properties of
solid surface, see Table 1. These reactions can be
regarded as ``catalytic acidity'' measures and can
provide better insight for understanding titania±silica
as a solid acid catalyst. Moreover, it was found that the
isomerization activity is greatly enhanced due to the
presence of Brùnsted acid sites. Thus, isomerization is
considered as a selective chemical probe reaction for
Brùnsted acidity [12,14,17,19,59,120]. On the other
hand, the alcohol dehydration is catalyzed by both
Lewis and Brùnsted acid sites and is proportional to
the total number of acid sites on the catalyst, and
therefore, it is considered as a measure for total acidity
[19,114].
The acidic properties of TiO
2
±SiO
2
mixed oxides
depend largely on the preparation method, synthesis
condition and chemical composition, which are in turn
X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254 247
related to the Ti±O±Si connectivities, the degree of
surface hydroxylation and TiO
2
particle size as dis-
cussed above [12,13,19,59]. The acid strength, when
evaluated by TPD and IR experiments of ammonia
adsorption, shows a decreasing trend with increasing
Si content [12]. The surface density (per surface area)
of Brùnsted acid sites is generally seen to be the
highest in Ti-rich mixed oxides [12,19,28]. The sur-
face density of total acid sites on TiO
2
±SiO
2
mixed
oxides, both the Lewis and Brùnsted acid sites, when
measured with ammonia or pyridine adsorption, gen-
erally decreases with increasing Si content over the
whole composition range [12,13,19,28], as shown in
Fig. 1. The decreased total acid sites with increasing
Si content can be accounted for by the decreased
surface density of the exposed Ti atoms. It is worth-
while to point out that the two curves of the total
acidity (per surface area) shown in Fig. 1 are different,
which may result from the different preparation pro-
cedures and acidity measurement procedures. The ®rst
curve obtained from [19] shows a quite linear relation-
ship of the total acid sites with the Ti mol% up to high
Ti content of 92 mol% Ti. While the second curve
obtained from [12] shows a sharp increase of the total
acid sites with increasing Ti concentration up to
18 mol% TiO
2
, and the total acid sites keeps almost
constant with only a slight increase up to TiO
2
content
of 85 mol%. The sharp increase at low Ti contents on
the second curve might be explained by the high
fraction of Ti atoms located on the surface [12]. As
the Ti content increases, the fraction of the exposed Ti
atoms appears to decrease while keeping the surface
density of the exposed Ti atoms almost constant. In
addition, both data sets show that the density of acid
sites on pure TiO
2
is higher than the end point of the
curve when extrapolated from the data of TiO
2
±SiO
2
mixed oxides with different chemical compositions,
suggesting that the surface of the mixed oxides at a
high Ti content is rich in Si. These acidity studies show
that the surface of Si-rich mixed oxides might be rich
in Ti depending on the preparation procedures,
whereas the surface of Ti±rich mixed oxides is most
likely rich in Si. In conclusion, the surface enrichment
of Ti or Si is a strong function of preparation proce-
dure and chemical composition.
Since butene isomerization and alcohol dehydration
occur at moderate temperatures (423±523 K), it is
expected that the catalyst surface is partially hydrated.
This is supported by FTIR experiments that demon-
strate that a large number of surface hydroxyls are
present under reaction conditions [17]. Consequently,
the reaction activity will be determined by the type,
density and strength of surface acid sites present under
the speci®c reaction condition.
The isomerization of 1-butene on TiO
2
±SiO
2
mixed
oxides has been reported to be acid-catalyzed, pro-
ceeding via butyl-carbonium ion as the intermediate
[12,15]. The enhanced isomerization activity is attrib-
uted to the Brùnsted acidity generated in the mixed
oxides [12,14,59]. Contescu et al. [17,59] found a
Brùnsted-type linear correlation between the speci®c
isomerization rates and the surface density of one
particular type of site identi®ed in the pK spectrum
of the hydrous oxide surface. They concluded that
both quantity and quality of Brùnsted acid sites affect
the overall catalytic performance. This conclusion is
also supported by results obtained by Liu et al. [12].
Their results show that although the density of acid
sites in Ti-rich mixed oxides is about two times higher
Fig. 1. Acidity and alcohol dehydration reactivities as a function of
Ti concentration for TiO
2
±SiO
2
mixed oxides. (Acidity (10
6
mol/
m
2
) data are adapted from [12,19], methanol dehydration reactivity
(10
ÿ2
mmol/min m
2
) from [19], and 2-propanol reactivity (10
ÿ1
percentage of 2-propanol reacted) from [12].)
248 X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254
than in Si-rich mixed oxides (more than 80% of the
acid sites on either Ti-rich or Si-rich mixed oxides is of
Brùnsted type), the speci®c rate of butene isomeriza-
tion in Ti-rich mixed oxides is about six times higher
than in Si-rich mixed oxides. Thus, the lower acid
strength in silica-rich mixed oxides is responsible for
the much lower isomerization acidity as compared to
Ti-rich mixed oxides. The fact that the highest speci®c
isomerization rate is generally located around 70±
90 mol% TiO
2
±SiO
2
mixed oxides [12±15] also
demonstrates that in butene isomerization reaction
the quality of Brùnsted acid sites (acid strength)
may also play a very important role in addition to
the quantity of Brùnsted acid sites. Unfortunately, the
isomerization reaction has not yet been applied for the
investigation of the acidic properties of TiO
2
/SiO
2
supported oxides, which could provide some interest-
ing comparative implications.
Alcohol dehydration provides additional informa-
tion about the acidic properties of titania±silica.
Results re-interpreted from methanol dehydration
on TiO
2
±SiO
2
mixed oxides [19] show that the speci®c
reactivity generally increases with the surface density
of acid sites, however, the increment is higher than the
increment of the density of acid sites. Moreover, a
maximum is observed in the Ti-rich mixed oxides, see
Fig. 1. Results from isopropanol dehydration [12] also
show a similar trend when the isopropanol dehydra-
tion reactivity is interpreted as the percentage of
isopropanol molecules converted to propene. As dis-
cussed previously, the increased acid strength with
increasing Ti content might partially contribute to the
increased reactivity. However, this interpretation can-
not explain the highest speci®c reactivity located
around 80±90 mol% Ti, which is similar to the iso-
merization reaction on TiO
2
±SiO
2
mixed oxides. This
result might also suggest that the Brùnsted acid sites
are more effective than Lewis acid sites for alcohol
dehydration since the surface density of Brùnsted acid
sites is generally seen to be the highest in Ti-rich
mixed oxides [12,19,28].
Isopropanol dehydration has been applied to dis-
persed TiO
2
/SiO
2
supported oxides by Biaglow et al.
[116]. They tried to develop a method for measuring
the titania surface area on TiO
2
/SiO
2
supported oxides
based on the selective adsorption of isopropanol. The
percentage of isopropanol reacted on TiO
2
/SiO
2
sup-
ported oxides is notably higher than pure TiO
2
, which
suggests that some Brùnsted acid sites are present on
the supported oxides during isopropanol dehydration.
Unfortunately, the acidity studies for TiO
2
/SiO
2
sup-
ported oxides are far less than the corresponding
studies on TiO
2
±SiO
2
mixed oxides, which prevent
us from fully understanding the surface properties of
dispersed TiO
2
/SiO
2
supported oxides.
3.3. Local structure of Ti and epoxidation/oxidation
reactions
3.3.1. Epoxidation/oxidation with hydrogen
peroxide/alkyl hydroperoxides
Titania±silica mixed and supported oxides are
effective catalysts for epoxidation and selective oxi-
dation reactions using peroxides as oxidants, as shown
in Table 1. For crystalline Ti-silicalites such as TS±1,
the active sites are assumed to be the isolated tetra-
hedral TiO
4
units substituted in the framework [122],
proceeding via titanium peroxocompounds as the
reaction intermediate. The epoxidation mechanism
involves no change in the oxidation state of Ti
4
cations [32]. The selective oxidation of saturated
hydrocarbons with hydrogen peroxide on TS-1 also
involves no change in the oxidation state of Ti cations,
but possibly undergoes a radical type mechanism
[22,122].
Titanium silicalites are superior to titania±silica
mixed and supported oxides in epoxidation of lower
ole®ns by hydrogen peroxide, but limited sterically to
relatively small reactants that are capable of penetrat-
ing into the narrow channels where most active sites
are located [22,31,122]. Recent studies by Hutter et al.
[25,31] show that mesoporous TiO
2
±SiO
2
mixed oxi-
des are promising catalysts for epoxidation of bulky
reactants such as cyclododecane, norbornene and a-
isophorone. Their results indicate that when Ti is well-
dispersed in the silica matrix with mesoporous struc-
ture, a high epoxidation activity is possible with some
bulky reactants. Meanwhile Klein et al. [30,34]
reported that amorphous microporous TiO
2
±SiO
2
mixed oxides used for epoxidation of ole®ns with
TBHP exhibit catalytic properties comparable to those
of the crystalline Ti-silicalites. Moreover, the highly
dispersed titanium oxide on amorphous silica and
MCM-41 also demonstrate excellent catalytic activity
in epoxidation of ole®ns by alkyl hydroperoxides
[25,31,115].
X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254 249
Since the epoxidation reactions are usually carried
out at low temperatures (323±363 K), polar solvents,
particularly alcohols and water, greatly retard the
reaction by competing for the active coordination sites
[33]. The epoxidation of a-isophorone by TBHP with
mesoporous TiO
2
±SiO
2
aerogels [25] shows that the
selectivity and reactivity of a-isophorone are mark-
edly in¯uenced by the solvent polarity. Polar solvents
cause a signi®cant drop in the activity by hindering the
formation of the titanium±peroxide complex and the
advance of the hydrophobic ole®n to the active sites,
and no activity was observed in water [25].
Extensive investigation has reached the conclusion
that the major difference between crystalline Ti-sili-
calites and amorphous TiO
2
±SiO
2
mixed oxides is that
Ti-silicalites can utilize hydrogen peroxide as oxidant
while the mixed oxides are only effective when using
organic hydroperoxides, which is attributed to the
much higher hydrophilicity as a result of a large
number of surface hydroxyl groups on TiO
2
±SiO
2
mixed oxides [22,26,30,38]. For example, due to
the hydrophobic nature of TS-1 micropores
(0.6 nm in diameter), H
2
O is believed to be screened
out from the cavities thus protecting the Ti sites from
deactivation by H
2
O [26,123].
The catalytic activity of ole®ns epoxidation with
alkyl hydroperoxide on TiO
2
±SiO
2
mixed oxides has
been strongly correlated with the fraction of the Ti
atoms in tetrahedral sites [26,28,30], or with the
Ti±O±Si connectivity that is characteristic of Ti
dispersion in the silica matrix [24]. The key factors
determining the activity and selectivity of TiO
2
±SiO
2
mixed oxides prepared by sol±gel method are assumed
to be the morphology (surface area and pore size) and
the relative proportions of Ti±O±Si to Ti±O±Ti bonds
[31].
The highly dispersed TiO
2
/SiO
2
supported oxides
(<4%Ti) have been reported early by Sheldon [32,33]
to be very effective for ole®ns epoxidation with alkyl
hydroperoxides. The formation of Ti±O±Si bonds is
crucial as demonstrated by the much lower activities
of TiO
2
supported on other oxides and the physical
mixture of TiO
2
and SiO
2
[33]. Sheldon suggested that
the active sites are the isolated monomeric titanyl
(SiO)
2
Ti=O groups. However, there is no experimen-
tal evidence to support the presence of Ti=O bonds on
the silica surface. A recent report [115] shows that
TiO
2
supported on mesoporous silica MCM-41 mole-
cular sieve is highly active and selective for epoxida-
tion of alkenes by TBPH. XANES/EXAFS studies
demonstrated that all Ti atoms are isolated and located
on the wall of MCM-41 mesopores. Interestingly, the
coordination of these Ti atoms is fourfold after calci-
nation, but changed to sixfold during the epoxidation
reaction. This study strongly suggests that the surface
active Ti centers on TiO
2
/SiO
2
supported oxides for
epoxidation reactions might also be isolated tetrahe-
dral sites with four similar Ti±O bonds instead of
(SiO)
2
Ti=O.
Despite different reactants used in epoxidation
reactions, which is largely dependent on the morphol-
ogy (surface area and pore size) and surface properties
(hydrophobicity) of catalysts, the high epoxidation
reactivities on Ti-silicalites, titania±silica mixed and
supported oxides strongly suggest that a structural
similarity exists between the active sites on these
different types of catalysts, i.e., isolated TiO
4
sites.
3.3.2. Redox ability and oxidation reactions
The redox properties of titania±silica are much less
studied due to the low oxidation potential of Ti(IV)
cations relative to some other transitional metal
cations such as V(V), Mo(VI), Cr(VI), etc. The reduc-
tion of pure TiO
2
is dif®cult, since only a very small
amount of surface Ti
4
cations can be reduced to Ti
3
cations as detected by CO adsorption but not by XPS
experiment [124]. The reducibility of 7 wt% TiO
2
/
SiO
2
supported oxides has also been examined by CO
adsorption and XPS experiments [74]. A small amount
of the surface reduced Ti
3
cations was detected by
both CO adsorption and XPS measurement after
hydrogen reduction. However, the authors did not
provide further information about whether the oxida-
tion potential of the Ti(IV) cations on TiO
2
/SiO
2
supported oxides is different from that of pure TiO
2
.
An interesting phenomenon was observed with
dispersed TiO
2
/SiO
2
supported oxides during metha-
nol oxidation [36,38,102]. Methanol oxidation is
known as a chemical probe to distinguish between
acid sites and redox sites [38]. For pure TiO
2
, the
surface Lewis acid sites result in almost complete
production of dimethyl ether (the high coverage of
the methoxy intermediate on TiO
2
may also contribute
to the high selectivity of dimethyl ether). However, for
the dispersed TiO
2
/SiO
2
supported oxides, a high
selectivity (94%) to the redox products (formalde-
250 X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254
hyde and methyl formate) was obtained, demonstrat-
ing that the redox property of surface TiO
x
species on
silica plays a key role. The decreased dehydration
ability (formation of dimethyl ether) relative to the
redox ability (formation of formaldehyde and methyl
formate) suggests an increased oxidizing potential and
a decreased acidity of the Ti cations when dispersed on
the silica support. The increased oxidizing potential of
the Ti(IV) cations, due to the formation of Ti±O±Si
bonds, is also re¯ected by the increased BE value of Ti
2p
3/2
and the higher LMCT transitions of the Ti atoms
in the dispersed TiO
2
/SiO
2
catalysts [102]. Similarly,
methanol oxidation on Ti-silicalites also gives rise to a
high selectivity of the redox products [38], indicating
that the Ti
4
cations with Ti±O±Si linkages serve as
the redox sites.
Furthermore, the highest speci®c catalytic activity
of the dispersed TiO
2
/SiO
2
supported oxides for
methanol oxidation was observed at the lowest loading
of 1.05 wt% TiO
2
/SiO
2
, which is dominated by the
isolated TiO
4
sites [102]. The observation of the sur-
face Ti-methoxy species resulting from the breaking
of the Ti±O±Si bridging bonds by Raman and UV±
Vis±NIR spectroscopy strongly supports the conclu-
sion that methanol oxidation involves participation of
Ti±O±Si bonds. The isolated TiO
4
sites provide the
maximum number of Ti±O±Si bonds and, conse-
quently, exhibit the highest speci®c activity for metha-
nol oxidation. Polymerization of the surface Ti atoms
on SiO
2
decreases the fraction of Ti±O±Si bonds and,
thus, signi®cantly decreases the activity of the Ti sites.
This is in agreement with the activity pattern of the
TiO
2
±SiO
2
mixed oxides for liquid phase ole®n epox-
idation with alkyl hydroperoxide where the activity
has been correlated with the Ti±O±Si connectivity or
the relative proportions of Ti±O±Si to Ti±O±Ti bonds
[24,31].
Contradictory results have been obtained for CO
oxidation on TiO
2
±SiO
2
mixed oxides [125]. The CO
oxidation activity decreases with increasing Ti con-
tent, and pure TiO
2
shows the highest activity. This
contradiction to the above reactions may be due to the
different active sites needed for CO oxidation. How-
ever, on the basis of the quantum-size effect, the
authors proposed that the increased energy gap
between the lowest unoccupied molecular orbital
(LUMO) and the highest occupied molecular orbital
(HOMO) is responsible for the decreased redox abil-
ity. Unfortunately, when Ti is substituted into the silica
network as an isolated TiO
4
unit, no corresponding
theoretical study has been done to determine how the
molecular orbital changes. It seems inappropriate to
apply the band-gap theory to the isolated TiO
4
unit
with localized Ti±O±Si bonds. As indicated by UV±
Vis spectra, the charge transfer O!Ti(IV) occurs at
the highest energy for the isolated TiO
4
unit in the
silica network, which indicates some contributions
from surrounding Si(IV) with a higher electronega-
tivity. The electron-accepting ability (af®nity) of
Ti(IV) in Ti±O±Si bonds should be higher than in
Ti±O±Ti bonds, in agreement with the increased BE
value of Ti 2p as compared to pure TiO
2
.
It is known that during reactions the coordination
geometry of Ti may change from fourfold to ®ve or
sixfold coordination as in the case of the epoxidation
reactions [115]. Consequently, the electron af®nity of
Ti is subject to change upon the coordination geome-
try generated during reactions, similar to the genera-
tion of new Brùnsted acid sites upon hydration [119].
It is suggested that the oxidation potential of Ti(IV) is
more likely related to its reaction intermediates
formed during reactions than its initial states. In other
words, the redox ability of Ti(IV) is dependent on the
speci®c reaction environment. Therefore, the in situ
investigation of the coordination geometry and the
oxidation potential of Ti(IV) atoms during reactions is
crucial for fully understanding the redox properties as
well as the reactivity properties of titania±silica in
oxidation reactions.
4. Summary
Titania±silica materials represent a novel class of
catalysts and have been widely applied in photocata-
lysis, acid catalysis and oxidation catalysis. The inti-
mate interaction of TiO
2
and SiO
2
has been shown to
result in new structural characteristics and physico-
chemical/reactivity properties. The degree of interac-
tion, in other words, the homogeneity or dispersion
when TiO
2
is mixed with or supported on SiO
2
, largely
depends on the preparation methods and synthesis
conditions. Also, the surface enrichment of either Ti
or Si in TiO
2
±SiO
2
mixed oxides has been shown to
depend on the preparation conditions and chemical
compositions.
X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254 251
Signi®cant improvement in the fundamental under-
standing of the structural characteristics of titania±
silica has been realized in recent years through the use
of many advanced techniques. Some characterization
techniques such as XPS, IR and Raman spectroscopies
can provide information about the formation of che-
mically bonded Ti±O±Si linkages, which might be
indirectly associated with the local structure of the Ti
atoms. EXAFS/XANES spectroscopy is a powerful
tool to investigate the local structure of Ti in titania±
silica materials, however, the uncertainty in the ®tting
parameters is of great concern. The combined results
from all these characterization techniques strongly
suggest that at low Ti contents, both the atomically
mixed TiO
2
±SiO
2
oxides and molecularly dispersed
TiO
2
/SiO
2
oxides possess predominantly isolated
TiO
4
sites. The difference in structural characteristics
of titania±silica mixed and supported oxides may just
be a matter of the degree of exposure, with TiO
2
±SiO
2
mixed oxides having some of the Ti atoms located in
the silica matrix.
The physico-chemical/reactivity properties of tita-
nia±silica are a strong function of the structural char-
acteristics. New Brùnsted acid sites are generated by
the charge imbalance on Ti±O±Si bond due to the
difference in coordination geometries of Ti and Si.
The new Brùnsted acid sites seem more effective for
isomerization and dehydration reactions. The reactiv-
ities of epoxidation/oxidation reactions have been
related to the exposed isolated TiO
4
sites as well as
the fraction of Ti±O±Si bonds. The structural char-
acteristics under speci®c conditions are crucial for
understanding the catalytic behavior of titania±silica
in a reaction, since the coordination geometry and the
physico-chemical properties of Ti changes in the
reaction. Therefore, future experiments should focus
on the in situ investigation of the coordination and
oxidation states of the Ti atoms in titania±silica so as
to develop a fundamental understanding about the
catalysis and chemistry of this interesting and impor-
tant catalytic material.
Acknowledgements
The authors would like to thank Dr. Simon R. Bare
for his helpful suggestions and comments on the
original paper. This work was ®nancially supported
by the US National Science Foundation Grant CTS-
9417981.
References
[1] G. Dagan, S. Sampath, O. Lev, Chem. Mater. 7 (1995) 446.
[2] R.W. Matthews, J. Catal. 113 (1988) 549.
[3] H. Inoue, T. Matsuyama, B. Liu, T. Sakata, H. Mori, H.
Yoneyama, Chem. Lett. (1994) 653.
[4] M. Anpo, K. Chiba, J. Mol. Catal. 74 (1992) 207.
[5] C. Anderson, A.J. Bard, J. Phys. Chem. 99 (1995) 9882.
[6] C. Anderson, A.J. Bard, J. Phys. Chem. B 101 (1997) 2611.
[7] X. Fu, L.A. Clark, Q. Yang, M.A. Anderson, Environ. Sci.
Technol. 30 (1996) 647.
[8] S. Yoshida, S. Takenaka, T. Tanaka, H. Hirano, H. Hayashi,
Eleventh International Congress on Catalysis, Stud. Sur. Sci.
Catal. 101 (1996) 871.
[9] S. Imamura, H. Tarumoto, S. Ishida, Ind. Eng. Chem. Res.
28 (1989) 1449.
[10] S. Imamura, T. Higashihara, H. Jindai, Chem. Lett. (1993)
1667.
[11] T. Liu, T. Cheng, Catal. Today 26 (1995) 71.
[12] Z. Liu, J. Tabora, R.J. Davis, J. Catal. 149 (1994) 117.
[13] J.B. Miller, S.T. Johnston, E.I. Ko, J. Catal. 150 (1994) 311.
[14] E.I. Ko, J.P. Chen, J.G. Weissman, J. Catal. 105 (1987) 511.
[15] M. Itoh, H. Hattori, K. Tanabe, J. Catal. 35 (1974) 225.
[16] H. Nakabayashi, Bull. Chem. Soc. Jpn. 65 (1992) 914.
[17] C. Contescu, V.T. Popa, J.B. Miller, E.I. Ko, J.A. Schwarz,
Chem. Eng. J. 64 (1996) 265.
[18] A. Molnar, M. Bartok, M. Schneider, A. Baiker, Catal. Lett.
43 (1997) 123.
[19] P.K. Doolin, S. Alerasool, D.J. Zalewski, J.F. Hoffman,
Catal. Lett. 25 (1994) 209.
[20] J.R. Sohn, J.H. Jang, J. Catal. 132 (1991) 563.
[21] W.F. Maier, J.A. Martens, S. Klein, J. Heilmann, R. Parton,
K. Vercruysse, P.A. Jacobs, Angew. Chem. 108 (1996) 222.
[22] C.B. Khouw, C.B. Dartt, J.A. Labinger, M.E. Davis, J. Catal.
149 (1994) 195.
[23] A. Bendandi, G. Fornasari, M. Guidoreni, L. Kubelkova, M.
Lucarini, F. Trifiro, Topic. Catal. 3 (1996) 337.
[24] R. Hutter, T. Mallat, A. Baiker, J. Catal. 153 (1995) 665.
[25] R. Hutter, T. Mallat, A. Baiker, J. Chem. Soc., Chem.
Commun. (1995) 2487.
[26] Z. Liu, G.M. Crumbaugh, R.J. Davis, J. Catal. 159 (1996)
83.
[27] S. Imamura, T. Nakai, H. Kanai, T. Ito, Catal. Lett. 28
(1994) 277.
[28] S. Imamura, T. Nakai, H. Kanai, T. Ito, J. Chem. Soc.,
Faraday Trans. 91 (1995) 1261.
[29] S. Imamura, T. Nakai, H. Kanai, T. Shiono, K. Utani, Catal.
Lett. 39 (1996) 79.
[30] S. Klein, S. Thorimbert, W.F. Maier, J. Catal. 163 (1996)
476.
[31] R. Hutter, T. Mallat, A. Baiker, J. Catal. 153 (1995) 177.
[32] R.A. Sheldon, J.A. Van Doorn, J. Catal. 31 (1973) 427.
252 X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254
[33] R.A. Sheldon, J. Mol. Catal. 7 (1980) 107.
[34] S. Klein, J.A. Martens, R. Parton, K. Vercruysse, P. Jacobs,
W.F. Maier, Catal. Lett. 38 (1996) 209.
[35] A. Keshavaraja, V. Ramaswamy, H.S. Soni, A.V. Ramaswa-
my, P. Ratnasamy, J. Catal. 157 (1995) 501.
[36] S. Srinivasan, A.K. Datye, M. Hampden-Smith, I.E. Wachs,
G. Deo, J.M. Jehng, A.M. Turek, C.H.F. Peden, J. Catal. 131
(1991) 260.
[37] S. Srinivasan, A.K. Datye, M.H. Smith, C.H.F. Peden, J.
Catal. 145 (1994) 565.
[38] G. Deo, A.M. Turek, I.E. Wachs, D.R.C. Huybrechts, P.A.
Jacobs, Zeolites 13 (1993) 365.
[39] M.A. Cauqui, J.J. Calvino, G. Cifredo, L. Esquivias, J.M.
Rodriguez-Izquierdo, J. Noncryst. Solids 147 148 (1992)
758.
[40] M.P. McDaniel, M.B. Welsh, M.J. Dreiling, J. Catal. 82
(1983) 98.
[41] S.J. Conway, J.W. Falconer, C.H. Rochester, J. Chem. Soc.,
Faraday Trans. 1 85 (1989) 71.
[42] J.J. Calvino, M.A. Cauqui, G. Cifredo, L. Esquivias, J.A.
Perez, J. Mater. Sci. 28 (1993) 2191.
[43] R. Mariscal, M. Galan-Fereres, J.A. Anderson, L.J.
Alemany, J.M. Palacios, J.L.G. Fierro, in: G. Centi et al.
(Eds.), Environmental Catalysis, SCI, Rome, 1995.
[44] B.E. Handy, A. Baiker, M. Schraml-Marth, A. Wokaun, J.
Catal. 133 (1992) 1.
[45] B.M. Reddy, E.P. Reddy, B. Manohar, Appl. Catal. 96
(1993) L1.
[46] M. Galan-Fereres, R. Mariscal, L.J. Alemany, J.L.G. Fierro,
J. Chem. Soc., Faraday Trans. 90 (1994) 3711.
[47] A.A. Elguezabal, V.C. Corberan, Catal. Today 32 (1996) 265.
[48] C.R. Dias, M.F. Portela, M. Galan-Fereres, M.A. Banares,
M.L. Granados, M.A. Pena, J.L.G. Fierro, Catal. Lett. 43
(1997) 117.
[49] M. Atik, J. Zarzycki, J. Mater. Lett. 13 (1994) 1301.
[50] M. Atik, P.D.L. Neto, M.A. Aegerter, L.A. Avaca, J. Appl.
Electrochem. 25 (1995) 142.
[51] K. Yu-Zhang, G. Boisjolly, J. Rivory, L. Kilian, C. Colliex,
Thin Solid Films 253 (1994) 299.
[52] D. Zhu, T. Kosugi, J. Noncryst. Solids 202 (1996) 88.
[53] C.J. Brinker, G.W. Scherer, Sol±gel Science: The Physics
and Chemistry of Sol±gel Processing, Academic Press, San
Diego, CA, 1990.
[54] Z. Deng, E. Breval, C.G. Pantano, J. Noncryst. Solids 100
(1988) 364.
[55] E. Breval, Z. Deng, C.G. Pantano, J. Noncryst. Solids 125
(1990) 50.
[56] M. Aizawa, Y. Nosaka, N. Fujii, J. Noncryst. Solids 168
(1994) 49.
[57] T. Hayashi, T. Yamada, H. Saito, J. Mater. Sci. 18 (1983)
3137.
[58] S. Satoch, K. Susa, I. Matsuyama, J. Noncryst. Solids 146
(1992) 121.
[59] C. Contescu, V.T. Popa, J.B. Miller, E.I. Ko, J.A. Schwarz, J.
Catal. 157 (1995) 244.
[60] D.T. On, L.L. Noe, L. Bonneviot, Chem. Commun. (1996)
299.
[61] P.J. Dirken, M.E. Smith, H.J. Whitfield, J. Phys. Chem. 99
(1995) 395.
[62] M. Schraml-Marth, K.L. Walther, A. Wokaun, B.E. Handy,
A.J. Baiker, J. Noncryst. Solids 143 (1992) 93.
[63] K.L. Walther, A. Wokaun, B.E. Handy, A.J. Baiker, J.
Noncryst. Solids. 134 (1992) 47.
[64] D.C.M. Dutoit, M. Schneider, A. Baiker, J. Catal. 153
(1995) 165.
[65] A.Y. Stakheev, E.S. Shpiro, J. Apijok, J. Phys. Chem. 97
(1993) 5668.
[66] C.H. Hung, J.L. Katz, J. Mater. Res. 7 (1992) 1861.
[67] S.M. Mukhopadhyay, S.H. Garofalini, J. Noncryst. Solids
126 (1990) 202.
[68] R.B. Greegor, F.W. Lytle, D.R. Sandstrom, J. Wong, P.
Schultz, J. Noncryst. Solids 55 (1983) 27.
[69] A. Fernandez, A. Caballero, A.R. Gonzalez-Elipe, Surf.
Interface Anal. 18 (1992) 392.
[70] A. Fernandez, A.R. Gonzalez-Elipe, C. Real, A. Caballero,
G. Munuera, Langmuir 9 (1993) 121.
[71] A. Hanprasopwattana, T. Rieker, A.G. Sault, A.K. Datye,
Catal. Lett. 45 (1997) 165.
[72] A. Hanprasopwattana, S. Srinivasan, A.G. Sault, A.K.
Datye, Langmuir 12 (1996) 3173.
[73] M.G. Reichmann, A.T. Bell, Appl. Catal. 32 (1987) 315.
[74] A. Fernandez, J. Leyrer, A.R. Gonzalez-Elipe, G. Munuera,
H. Knozinger, J. Catal. 112 (1988) 489.
[75] R. Castillo, B. Koch, P. Ruiz, B. Delmon, J. Catal. 161
(1996) 524.
[76] R. Mariscal, J.M. Palacios, M. Galan-Fereres, J.L.G. Fierro,
Appl. Catal. A 116 (1994) 205.
[77] S. Haukka, E. Lakomaa, A. Root, J. Phys. Chem. 97 (1993)
5085.
[78] B.A. Morrow, A.J. Mcfarlan, J. Noncryst. Solids 120 (1990)
61.
[79] J. Klaas, G. Schulz-Ekloff, N.I. Jaeger, J. Phys. Chem. B
101 (1997) 1305.
[80] M. Galan-Fereres, L.J. Alemany, R. Mariscal, M.A.
Banares, J.A. Anderson, J.L.G. Fierro, Chem. Mater. 7
(1995) 1342.
[81] Z. Liu, R.J. Davis, J. Phys. Chem. 98 (1994) 1253.
[82] M. Aizawa, Y. Nosaka, N. Fujii, J. Noncryst. Solids 128
(1991) 77.
[83] M. Toba, F. Mizukami, S. Niwa, T. Sano, K. Maeda, A.
Annila, V. Komppa, J. Mol. Catal. 91 (1994) 227.
[84] M.F. Best, R.A. Condrate, J. Mater. Sci. Lett. 4 (1985)
994.
[85] S. Bordiga, S. Coluccia, C. Lamberti, L. Marchese, A.
Zecchina, F. Boscherini, F. Buffa, F. Genoni, G. Leofanti, G.
Petrini, G. Vlaic, J. Phys. Chem. 98 (1994) 4125.
[86] A. Chmel, G.M. Eranosyan, A.A. Kharshak, J. Noncryst.
Solids 146 (1992) 213.
[87] I.M.M. Salvado, J.M.F. Navarro, J. Noncryst. Solids 147 148
(1992) 256.
[88] C.C. Perry, X. Li, D.N. Waters, Spectrochim. Acta 47A
(1991) 1487.
[89] M.E. Smith, H.J. Whitfield, J. Chem. Soc., Chem. Commun.
6 (1994) 723.
X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254 253
[90] I. Grohmann, W. Pilz, G. Walther, H. Kosslick, V.A. Tuan,
Surf. Interface Anal. 22 (1994) 403.
[91] J.K. Walters, J.S. Rigden, P.J. Dirken, M.E. Smith, W.S.
Howells, R.J. Newport, Chem. Phys. Lett. 264 (1997)
539.
[92] J.S. Rigden, R.J. Newport, M.E. Smith, P.J. Dirken, G.
Bushnell-Wye, J. Mater. Chem. 6 (1996) 337.
[93] S. Klein, B.M. Weckhuysen, J.A. Martens, W.F. Maier, P.A.
Jacobs, J. Catal. 163 (1996) 489.
[94] L.L. Noc, D.T. On, S. Solomykina, B. Echchahed, F.
Beland, C.C.D. Moulin, L. Bonneviot, Eleventh Interna-
tional Congress on Catalysis, Stud. Sur. Sci. Catal. 101
(1996) 611.
[95] D.R. Sandstrom, F.W. Lytle, P.S.P. Wei, R.B. Greegor, J.
Wong, P. Schultz, J. Noncryst. Solids 41 (1980) 201.
[96] S. Pei, G.W. Zajac, J.A. Kaduk, J. Faber, B.I. Boyanov, D.
Duck, D. Fazzini, T.I. Morrison, D.S. Yang, Catal. Lett. 21
(1993) 333.
[97] A.B. Rosenthal, S.H. Garofalini, J. Noncryst. Solids 107
(1988) 65.
[98] C. Kormann, D.W. Bahnemann, M.R. Hoffmann, J. Phys.
Chem. 92 (1988) 5196.
[99] C.K. Jùrgensen, Morden Aspects of Ligand Field Theory,
North Holland, Amsterdam, 1971.
[100] M.R. Boccuti, K.M. Rao, A. Zecchina, G. Leofanti, G.
Petrini, Stud. Surf. Sci. Catal. 48 (1989) 133.
[101] J.A. Duffy, Struct. Bonding (Berlin) 32 (1977) 147.
[102] X. Gao, I.E. Wachs, J. Phys. Chem. B 102 (1998) 5653.
[103] G. Lassaletta, A. Fernandez, J.P. Espinos, A.R. Gonzalez, J.
Phys. Chem. 99 (1995) 1484.
[104] J.A. Mejias, V.M. Jimenez, G. Lassaletta, A. Fernandez,
J.P. Espinos, A.R. Gonzalez, J. Phys. Chem. 100 (1996)
16255.
[105] T. Nakayama, J. Electrochem. Soc. 141 (1994) 237.
[106] N. Brun, C. Colliex, J. Rivory, K. Yu-Zhang, Microsc.
Microanal. Microstruct. 7 (1996) 161.
[107] D. Scarano, A. Zecchina, S. Bordiga, F. Geobaldo, G. Spoto,
G. Petrini, G. Leofanti, M. Padovan, G. Tozzola, J. Chem.
Soc., Faraday Trans. 89 (1993) 4123.
[108] C. Cartier, C. Lortie, D. Trong On, H. Dexpert, L.
Bonneviot, Physica B 208 209 (1995) 653.
[109] I.D. Brown, K.K. Wu, Acta Cryst. B 32 (1976) 1957.
[110] M.A. Roberts, G. Sankar, J.M. Thomas, R.H. Jones, H. Du,
J. Chen, W. Pang, R. Xu, Nature 381 (1996) 401.
[111] F. Farges, G.E. Brown Jr., J.J. Rehr, Geochim. Cosmochim.
Acta 60 (1996) 3023.
[112] F. Farges, G.E. Brown Jr., A. Navrotsky, H. Gan, J.J. Rehr,
Geochim. Cosmochim. Acta 60 (1996) 3039.
[113] F. Farges, G.E. Brown Jr., J.J. Rehr, Phys. Rev. B 56 (1997)
1809.
[114] K.S. Kim, M.A. Barteau, W.E. Farneth, Langmuir 4 (1988)
533.
[115] T. Maschmeyer, F. Rey, G. Sanker, J.M. Thomas, Nature 378
(1995) 159.
[116] A.I. Biaglow, R.J. Gorte, S. Srinivasan, A.K. Datye, Catal.
Lett. 13 (1992) 313.
[117] T. Nakayama, K. Onisawa, M. Fuyama, M. Hanazono, J.
Electrochem. Soc. 139 (1992) 1204.
[118] H. Nakabayashi, N. Kakuta, A. Ueno, Bull. Chem. Soc. Jpn.
64 (1992) 2428.
[119] T. Kataoka, J.A. Dumesic, J. Catal. 112 (1988) 66.
[120] H. Nakabayashi, Bull. Chem. Soc. Jpn. 65 (1992) 914.
[121] K. Tanabe, T. Sumiyoshi, K. Shibata, T. Kiyoura, J.
Kitagawa, Bull. Chem. Soc. Jpn. 47 (1974) 1064.
[122] B. Notari, Catal. Today 18 (1993) 163.
[123] R.A. Sheldon, J. Dakka, Catal. Today 19 (1994) 215.
[124] M.I. Zaki, H. Knozinger, Spectrochim. Acta A 43 (1987)
1455.
[125] S. Imamura, S. Ishida, H. Tarumoto, Y. Saito, T. Ito, J.
Chem. Soc., Faraday Trans. 89 (1995) 757.
[126] D. Trong On, L. Bonneviot, A. Bittar, A. Sayari, S.
Kaliaguine, J. Mol. Catal. 74 (1992) 233.
254 X. Gao, I.E. Wachs / Catalysis Today 51 (1999) 233±254