XAFS studies of surface structures of TiO2 nanoparticles and photocatalytic reduction of metal ions

Article abstract of DOI:10.1021/jp971930g

To probe the origin of the unique functions of titanium dioxide (TiO₂) nanoparticles observed in photocatalytic reactions, structures of Ti atom sites in titanium dioxide (TiO₂) nanoparticles with different sizes were studied by Ti K-edge XAFS (X-ray absorption fine structure). Compared to the bulk TiO₂ structure, a shorter Ti-O distance from surface TiO₂ resulting from Ti-OH bonding was observed. The XAFS spectra also revealed an increasing disorder of the lattice with decreasing sizes of the nanoparticles based on a coordination number decrease for the third-shell O atoms as well as changes in relative intensities of preedge peaks A₁, A₂, and A₃. However, the Ti sites largely remain octahedral even in the 30 ? diameter particles. These results imply that the increasing number of surface Ti sites as well as possible corner defects in small nanoparticles may be the main cause of the unique surface chemistry exhibited by nanoparticles of TiO₂. XAFS was also used in monitoring the photoreduction reaction products of Cu²⁺ and Hg²⁺ on TiO₂ nanoparticle surfaces, with or without surface adsorbers, alanine (Ala) and thiolactic acid (TLA). Ala dramatically enhanced photoreduction of Cu²⁺ on TiO₂ nanoparticle surfaces, whereas thiolactic acid did not affect or even hindered Hg²⁺ photoreduction. Although both surface adsorbers chelated with the metal ions in the absence of TiO₂ nanoparticles, this chelation was drastically changed in the Cu-Ala complex but was largely retained in the Hg-TLA complex when TiO₂ was present. This may correlate with the different effects of the adsorbers on the photoreduction of the metal. Our experimental results suggest that a proper balance between the affinities of the adsorber to the metal ions and to the surface Ti atoms of TiO₂ may be one of the keys in selecting a surface adsorber for enhanced photoreduction efficiency.

Full text of DOI:10.1021/jp971930g

10688
J. Phys. Chem. B 1997, 101, 10688-10697
XAFS Studies of Surface Structures of TiO2 Nanoparticles and Photocatalytic Reduction of
Metal Ions
Lin X. Chen,* Tijana Rajh, Zhiyu Wang, and Marion C. Thurnauer
Chemistry DiVision, Argonne National Laboratory, 9700 South Cass AVenue, Argonne, Illinois 60439
ReceiVed: June 13, 1997; In Final Form: September 26, 1997X
To probe the origin of the unique functions of titanium dioxide (TiO2) nanoparticles observed in photocatalytic
reactions, structures of Ti atom sites in titanium dioxide (TiO2) nanoparticles with different sizes were studied
by Ti K-edge XAFS (X-ray absorption fine structure). Compared to the bulk TiO2 structure, a shorter Ti-O
distance from surface TiO2 resulting from Ti-OH bonding was observed. The XAFS spectra also revealed
an increasing disorder of the lattice with decreasing sizes of the nanoparticles based on a coordination number
decrease for the third-shell O atoms as well as changes in relative intensities of preedge peaks A1, A2, and
A3. However, the Ti sites largely remain octahedral even in the 30 Å diameter particles. These results
imply that the increasing number of surface Ti sites as well as possible corner defects in small nanoparticles
may be the main cause of the unique surface chemistry exhibited by nanoparticles of TiO2. XAFS was also
used in monitoring the photoreduction reaction products of Cu2+ and Hg2+ on TiO2 nanoparticle surfaces,
with or without surface adsorbers, alanine (Ala) and thiolactic acid (TLA). Ala dramatically enhanced
photoreduction of Cu2+ on TiO2 nanoparticle surfaces, whereas thiolactic acid did not affect or even hindered
Hg2+ photoreduction. Although both surface adsorbers chelated with the metal ions in the absence of TiO2
nanoparticles, this chelation was drastically changed in the Cu-Ala complex but was largely retained in the
Hg-TLA complex when TiO2 was present. This may correlate with the different effects of the adsorbers on
the photoreduction of the metal. Our experimental results suggest that a proper balance between the affinities
of the adsorber to the metal ions and to the surface Ti atoms of TiO2 may be one of the keys in selecting a
surface adsorber for enhanced photoreduction efficiency.
I. Introduction
Although many observations led to the same conclusions that
photocatalytic activities were higher in TiO2 nanoparticles
compared to the bulk and sometimes higher in modified
nanoparticles compared to the bare particles, the mechanisms
for the enhancement were not completely clear. To further our
understanding of the reaction mechanisms, we examine in this
work (1) the surface and lattice structures of TiO2 nanoparticles
and (2) oxidation states of the metal photoreduction products.
In previous work, metallic states were identified from the optical
absorption spectra that had an onset of absorption below 800
nm. However, these spectra can also be characteristic of binary
semiconductors with small bandgaps. Thus, positive identifica-
tion of the metal oxidation states following photoinduced
reaction is crucial to understanding the reaction mechanisms.
Photocatalytic reactions on TiO2 surfaces are very important
in environmental cleanup and remediation,1-6 such as oxidation
of organic material and reduction of heavy metal ions in
industrial waste streams.7-11 Differing from the bulk material,
TiO2 nanoparticles are expected to have unique surface chem-
istry due to their larger surface area.12 Photoirradiation of TiO2
nanoparticles, with photon energies larger than the bandgap (3.2
eV), creates electron-hole (e-/h+) pairs,
TiO2 9
hν8 ec-b(TiO2) + hv+b(TiO2)
Consequently, following irradiation, the TiO2 nanoparticles act
as either electron or hole donors to reduce or oxidize materials
in the surrounding media. However, the photoinduced charge
separation in bare TiO2 nanoparticles has a very short lifetime
because of charge recombination. Therefore, it is important to
prevent the hole-electron recombination before a designated
redox reaction occurs. One way to accomplish this is to
scavenge photogenerated holes or electrons with molecules
strongly adsorbed onto TiO2 surfaces.11,13 In this way, scav-
enged charge carriers can be transported away from the surface
of the TiO2 particles, providing a larger spatial separation
between the photogenerated holes and electrons. Consequently,
the lifetime of charge separation is prolonged, and the photo-
catalytic activity of TiO2 particles may be increased. Recent
EPR studies indicated enhanced redox activities of surface-
modified TiO2 particles as a result of transferring the photoge-
nerated holes away from the TiO2 surface through the surface
adsorbed molecules.11
The techniques of X-ray absorption spectroscopy (XAS),
namely X-ray absorption near edge structure (XANES) and
X-ray absorption fine structure (XAFS), probe local structural
details around specific metal atoms and are sensitive to the
oxidation states of the metal.14 At the Ti K-edge, structural
changes around Ti atoms as a function of the particle sizes can
be probed, and this could be correlated with the unique surface
reactivities of TiO2 nanoparticles. At the metal absorption edges
(K or L), XAFS can be used to examine the surroundings of
the metal species adsorbed on the TiO2 nanoparticle. Com-
parison of spectra from TiO2 nanoparticles with and without
surface-adsorbed small molecules as obtained with and without
photoillumination provides evidence for metal photoreduction
and interactions between metal ions and TiO2 surface or
adsorbed species. Correlations between the structures of the
metal complexes and their photoreduction reactivities can be
established through such studies. This correlation, if any, will
help in the selection of TiO2 nanoparticle surface modifiers that
will enhance reaction efficiencies.
* To whom correspondence should be addressed: FAX (630) 252-9289,
e-mail lchen@anlchm.chm.anl.gov.
X Abstract published in AdVance ACS Abstracts, November 15, 1997.
S1089-5647(97)01930-5 CCC: $14.00 © 1997 American Chemical Society

Structures of TiO2 Nanoparticles
J. Phys. Chem. B, Vol. 101, No. 50, 1997 10689
In this paper, we use XAFS to investigate the effect of the
size on the lattice structure of TiO2 nanoparticles. In the size
regime of nanoparticles (e.g., particles with diameters of 20-
100 Å), a significant fraction of the atoms are located on the
surface (e.g., ∼40% for 30 Å diameter TiO2 particles; see
below). Therefore, it is feasible to probe by XAFS the unique
structural features around surface Ti atoms in TiO2 nanoparticles.
Both structures for the interior and surface TiO2 can be
characterized by comparing the XAFS spectra of TiO2 particles
of different sizes. In addition, we also used XAFS to monitor
metal ion photoreduction of Cu2+ and Hg2+ on TiO2 colloid
surfaces with and without surface modifiers. The oxidation state
of the metal ions in these processes can be identified directly
by XANES at respective metal absorption edges. The correla-
tion of the reaction efficiency and the structures of the metal
complexes with TiO2 or surface adsorbers will be discussed.
For filtered TiO2 gel samples with metal ions, X-ray fluores-
cence detection was carried out using a Lytle detector with a
proper filter. No filter was used for measurements at the Ti
K-edge. A nickel or germanium filter was used for the
measurements at the Cu K-edge or the Hg LIII-edge, respectively.
At the Ti K-edge, helium gas was used to purge the sample
chamber of the Lytle detector. For most of the solid TiO2
nanoparticle and reference compound powders, the sample was
sandwiched between two pieces of Kapton tape. For the filtered
TiO2 or TiO2 gel samples with or without adsorbed metal ions,
the sample was placed in a sample cell with a rubber spacer
and Mylar windows.
3. Data Analysis. XAFS data analysis was carried out using
standard procedures. Considering only 3% of the TiO2 is on
the surface (see below), TiO2 nanoparticles with 400 Å diameters
were used as a reference for the structure of anatase TiO2 lattice,
where the first-shell Ti-O distance is 1.96 Å with six O atoms,
the second-shell Ti-Ti distance is 3.04 Å with four Ti atoms,
and the third-shell Ti-O distance is 3.86 Å with 12 O atoms.16
Five other reference compounds were used to extract structural
parameters for the metal complexes. For Cu ions, the structural
reference parameters were extracted from the Cu(glycine)2
complex (Cu-Gly2) with a known X-ray structure.17 Cupric
oxide (CuO) was also used as a reference based on its crystal
structure.18 Cu foil was used for characterizing Cu-Cu
distances. HgS (black) and HgO were used as references19,20
for Hg complexes. In most cases, k3 weighting was used, and
II. Experimental Section
1. Sample Preparation. All chemicals were reagent grade
from Aldrich or Baker Chemical Companies and used without
further purification. The TiO2 particles with 30, 50, and 200
Å diameters were prepared by dropwise addition of titanium-
(IV) chloride to cooled water. The details of preparation are
presented elsewhere.15 The TiO2 nanoparticles with 400 Å
diameters were gifts from Nanophase Inc. The sizes of the
colloid particles were determined by electron microscopy. Ala
or TLA/metal ion mixtures were made with concentrations of
0.02 M/0.01 M, respectively. The metal ion concentration in
all starting solution samples was 0.01 M. Methanol was added
as a hole scavenger to all TiO2-containing samples with mercury
ions and without surface adsorber. Copper and mercury acetate
salts were dissolved in aqueous solutions as respective metal
ion sources.
The photoillumination was provided by an ILC UV-enhanced
xenon lamp (150 W). The sample with a mixture of TiO2
nanoparticles and metal ion solution (0.01 M) was sealed in an
airtight glass cell and stirred constantly during the illumination.
The sample container was placed in a continuous flow water
bath with a glass wall to maintain the sample at constant
temperature during irradiation. After a certain time of illumina-
tion the sample was put into a bag which had a continuous
nitrogen gas flow. An aliquot of the mixture was filtered and
transferred under nitrogen atmosphere from the glass cell to a
sealed XAFS cell for the measurement. The remaining mixture
was kept in the dark under nitrogen atmosphere until the next
illumination. This procedure was repeated for samples with
different illumination times. The illumination time was the total
time that the sample was exposed to the lamp.
2. XAFS Measurements. XAFS experiments were carried
out at Beamlines X6B, X18B, and X19A of the National
Synchrotron Light Source (NSLS) at Brookhaven National
Laboratory. In Beamlines X6B and X19A, the X-ray beam was
focused, and the spot size at the sample was 1 mm × 0.5 mm.
In X18B, the beam size at the sample was 15 mm × 1 mm. A
Si(111) double-crystal monochromator was used for Ti K-edge
(4.966 keV) experiments, and a Si(220) monochromator crystal
was used for metal K- or L-edge experiments. The monochro-
mator was detuned by 20-50% in order to reduce the harmonic
component in the incident X-ray at the sample. The scans were
collected from -100 to 1000 eV relative to the transition edge
energies which were calibrated using TiO2 powder and metal
foils.
Fourier transforms were taken between k ) 3 and 11 Å-1
.
To demonstrate the effects of the size of the TiO2 lattice and
the locations of absorbing Ti atom on the XAFS spectra, we
used the program FEFF6.0 (University of Washington) to
simulate the XAFS spectra of TiO2 anatase lattice. Simulations
were carried out with different sizes of the TiO2 lattice: (1)
TiO6, (2) Ti9O38, and (3) Ti45O106 as well as different locations
of absorbing Ti atom (on the surface vs in the middle of the
lattice). Multiple scatterings were included in the calculations.
The maximum scattering path length was set at 9.5, and the
maximum number of scattering paths in each route was 8.
Because the goal of these simulations is to find the origin of
the spectral characteristics of the Ti at different locations in a
TiO2 lattice cluster and the effect of cluster sizes, an ideal
anatase lattice structure in the cluster was used, and the Debye-
Waller factors were assumed to be zero for simplification.
III. Results and Discussion
1. Particle Size Dependence of TiO2 Nanoparticle Struc-
tures. Ti Site Structures in TiO2 Nanoparticles: Surface Vs Bulk.
The origin of the unique photocatalytic activities exhibited by
nanoparticles of TiO2 is crucial to understanding the reaction
mechanisms, whether it is purely due to the increased surface
area or due to distorted sites on the surface or the whole lattice
of the particle. Because the XAFS technique detects the local
structural details around the X-ray-absorbing atoms, this tech-
nique is particularly powerful in determining short-range
structures of surface Ti sites in TiO2 nanoparticles. In studying
the surface structure of TiO2 nanoparticles using XAFS, one
faces a challenge of the coexistence of the interior and the
surface structures. One way to solve this problem is to
systematically measure XAFS spectra of a series of TiO2
particles with different sizes. As the particle size gets smaller,
the fraction of TiO2 on the surface increases, and the XAFS
spectral features that reflect the surface TiO2 structure become
more prominent. In this way, we could distinguish the structural
features that arise from the surface layer of TiO2 from those of
the structures associated with bulk TiO2.
For TiO2 and reference compound powder samples, the
transmission detection by ion chambers (helium gas for Ti
K-edge, nitrogen gas for Cu K-edge and Hg LIII-edge) was used.

10690 J. Phys. Chem. B, Vol. 101, No. 50, 1997
Chen et al.
Figure 2. Relative intensities of preedge peaks (see text) A1, A2, and
A3 as functions of TiO2 particle diameters.
tion of the preedge for rutile TiO2 is rather successful, that for
the anatase TiO2 still lacks agreement with the spectral detail.26
We have observed a systematic increase in the intensity of the
A2 peak with a decrease in particle size (Figure 1), indicating
an increasing distortion from the octahedral TiO6 unit. As the
particle size becomes smaller, more Ti atoms are exposed on
the surface of the particle experiencing an anisotropic environ-
ment where a Ti atom bonds to inner and outer O atoms
simultaneously. This anisotropic environment for the surface
Ti atoms may be the cause of the possible distortion around Ti
atoms from an octahedral TiO6 unit. Meanwhile, the intensity
of the A1 peak decreases as the particle size decreases.
However, the fractions of integrated intensities of A1, A2, and
A3 as functions of the sizes of the particles seem to correlate
with each other (see Figure 2). The most obvious changes occur
for particles with diameters less than 50 Å. In other words,
the distortion of the lattice consistently increases as the particle
size gets smaller. Nevertheless, the preedge of all three
transitions is still clearly visible in the spectrum of 30 Å diameter
particles even though the A2 peak dominates. However, this
preedge spectrum is significantly different from that for Ti atoms
in tetrahedral TiO4 sites, where only one sharp peak, about one-
half of the white line peak height, was observed.23 Although
the observed A2 peak enhancement in particles with 30 Å
diameter could result from a very small number of tetrahedral
sites, the Ti sites in TiO2 particles studied here are mostly
octahedral. The distortion around Ti is never severe enough
that the Ti site becomes tetrahedral as for TiO2 in SiO2 glass.23,25
XANES of Ti K-Edge for TiO2 Nanoparticles of Different
Diameters. Other spectral features in the XANES region (see
Figure 1) include a peak around 13.5 eV above the A1 peak, in
the middle of the rising edge. The first derivatives of the
XANES spectra for different particle sizes show that the peak
diminishes as the particle size is reduced. However, the peak
is still visible even for particles with 30 Å diameters. This peak
was assigned to the 1s f 4p transition.22 Other peaks in this
region, along with those in the XAFS spectra, also significantly
broaden as the particle size gets smaller. Two effects could
contribute to the observed spectral changes: (1) an increase of
the fraction of surface TiO2 and (2) the inhomogeneity in the
lattice structures in small nanoparticles.
Figure 1. Preedge and XANES spectra of TiO2 nanoparticles at Ti
K-edge (E0 ) 4.966 keV for Ti0). Top panel: the spectra of TiO2
particles with different sizes. Bottom panel: first derivatives of the
spectra in the top panel.
Fractions of Surface TiO2 in Nanoparticles. To analyze the
TiO2 nanoparticle spectra, we need to estimate the fraction of
surface TiO2 in a nanoparticle with a specific size. This estimate
was made based on the anatase lattice structure,16 which is the
structure determined for TiO2 nanoparticles synthesized by
controlled hydrolysis of TiCl4 at low temperature.21 The average
volume and surface areas of each TiO2 were calculated using
the dimensions of the anatase unit cell. The approximate
average surface area for each TiO2 was calculated based on the
number density of Ti on different lattice planes and an assumed
equal probability of surface exposure for (110), (101), and (011)
planes. In other words, the surface area occupied by each TiO2
is averaged over these three planes. Based on these parameters,
the fraction of the surface TiO2 is proportional to 12.5/d, where
d is the diameter of the TiO2 nanoparticles in angstroms. Thus,
the fraction of exposed TiO2 is 3% for 400 Å, 6% for 200 Å,
25% for 50 Å, and 41% for 30 Å diameter particles. However,
the surface distortion could be more than one layer deep. Since
the 400 Å has only 3% on the surface, we used this as the
standard for the bulk anatase.
Preedge Spectra of Ti K-Edge for TiO2 Nanoparticles of
Different Diameters. The preedge structure at the Ti K-edge
for TiO2 features three peaks: A1, A2, and A3 (see Figure 1).
According to previous studies, the origins for A2 and A3 are
attributed to 1s f 3d transitions and designated to 1s f 2t2g
and 1s f 3eg transitions in an octahedral field, respectively,
and the origin for A1 was assigned to an exciton band22 or a
transition from 1s f 1t1g resulting from possible perturbation
due to shake-up and shake-off processes.23 An increase in
relative intensity of A2 has been observed in the TiO2 materials
with increasing distortion from the octahedral TiO6 unit, and
an intense single preedge peak has been observed for TiO2 with
a tetrahedral geometry around Ti.23-25 Extensive efforts have
been made to theoretically calculate the preedge structure of
TiO2 with a limited number of atoms.26-28 While the reproduc-

Structures of TiO2 Nanoparticles
J. Phys. Chem. B, Vol. 101, No. 50, 1997 10691
Figure 4. XAFS spectra of TiO2 nanoparticles with different diameters.
Figure 3. Calculated XAFS spectra using FEFF6.0 for TiO2 lattice
with different sizes and different locations of absorbing Ti atom (on
the surface or in the center of the cluster).
The first effect above is simulated by calculated XAFS spectra
of Ti in TiO2 anatase clusters of different sizes and at different
locations (center or surface) using FEFF 6.0. The results are
depicted in Figure 3. When a Ti atom is in an isolated TiO6
unit surrounded only by a single shell of six O atoms, its XAFS
spectrum has a single frequency sinusoidal oscillation without
fine features. When an X-ray-absorbing Ti is surrounded by
several shells of backscattering atoms, superpositioning of the
backscattered photoelectron waves from different shells and the
multiscattering produces many fine features in an XAFS
spectrum. Therefore, for the Ti atom situated in the center of
TiO6, Ti9O38, and Ti45O106, an increasing complexity in spectral
features is observed. When the absorbing Ti atom is moved
from the center to the surface in a TiO2 cluster, part of the
multiscattering pathways are truncated, resulting in a spectrum
with fewer fine features. Therefore, an increase in the surface
fraction of the Ti atom alone could cause the diminishing fine
features in XAFS spectra as the TiO2 particle size decreases.
However, the fine spectral features for the Ti atom could also
be compromised by heterogeneous broadening when there is
an inhomogeneity in the local structure around Ti atoms. The
static disorder in the Ti-to-other atom distances and the bond
angles will cause the heterogeneous broadening of the spectra.
Because the XAFS effect is due to short-range structure around
the X-ray-absorbing atoms, the broadening itself suggests
inhomogeneity of the local structure, rather than loss of the
coherence due to a size effect observed in X-ray diffraction.
XAFS results alone are not sufficient to distinguish between
the two possible causes of the broadening of spectral features,
but the analysis of XAFS spectra presented in the following
section could provide the structural information that is uniquely
present in TiO2 nanoparticles.
Figure 5. FT-XAFS spectra of TiO2 nanoparticles with different
diameters.
400 Å diameter particles is significantly broadened in 200 Å
diameter particles, although the fraction of surface TiO2 changes
from 3% to 6%. This observation indicates that the observed
spectral changes are not entirely due to the increase of the
surface TiO2, and a global disorder throughout the entire particle
may also occur when the particle size reaches a certain limit.
In addition to the damping of the oscillation amplitudes, the
oscillation patterns of the spectra in the high-k region also show
substantial differences as a function of particle size. This
implies that there are structural differences other than the
differences in the Debye-Waller factor alone between the
particles with different sizes. This can be seen more clearly in
the FT-XAFS spectra in Figure 5. The bond distance of Ti to
the first-shell O atom varies with the size of the particles. The
FT window is taken from k ) 3-11 for all the spectra. The
first-shell fit for TiO2 particles with 200, 50, and 30 Å diameters
XAFS Spectra of TiO2 Colloidal Particles with Different
Diameters. The XAFS spectra for TiO2 colloid particles of
different diameters are shown in Figure 4. As a general trend,
the oscillation amplitude in the spectra reduces as the particle
size decreases. Meanwhile, the oscillations also broaden, and
fine structure is lacking in the smaller particles, similar to the
results in the XANES region. The fine structure observed in

10692 J. Phys. Chem. B, Vol. 101, No. 50, 1997
Chen et al.
TABLE 1: XAFS Fitting Results for TiO2 Colloids
Particlesa
diameter (Å)
400 (ref)
shell
N
R (Å)
1.96
3.04
σ2 (Å2)b
1
2
3
1
2
3
1
1
2
3
1
1
2
3
6.0
4.0
12.0
3.86
200
50
6.4 ( 0.5
4.6 ( 0.5
12 ( 2
5.0 ( 0.5
1.0 ( 0.2
2.5 ( 0.5
4 ( 1
4.0 ( 0.5
0.8 ( 0.2
3.2 ( 0.5
5 ( 1
1.92 ( 0.02
3.09 ( 0.04
3.87 ( 0.04
1.94 ( 0.02
1.77 ( 0.02
3.09 ( 0.04
3.77 ( 0.04
1.99 ( 0.02
1.84 ( 0.02
3.13 ( 0.04
3.83 ( 0.04
0.010
0.006
0.010
0.003
-0.006
0.005
0.006
0.003
-0.006
0.008
0.007
30
a TiO2 particles of 400 Å diameters were used as the reference
compound, with the assumption that they have the same lattice structure
as anatase. The first shell has Ti-O distance with N ) 6, R ) 1.95 Å,
the second shell has Ti-Ti distance with N ) 4, R ) 3.04 Å, and the
third shell has Ti-O distance with N ) 12, R ) 3.86 Å. b Relative to
400 Å diameter TiO2.
The first-shell Ti-O peak in the FT-XAFS spectra for 30
and 50 Å diameter particles fits a two-shell model, where one
Ti-O distance is close to 1.95 Å and the other 0.1-0.2 Å
shorter (see Table 1). The necessity of introducing the second
Ti-O distance in the smaller particles suggests that the shorter
Ti-O distance is from the exposed Ti-O. The fact that the
second-shell Ti-Ti peak and the third-shell Ti-O peak do not
need the second distance in the fitting supports the idea that
the shorter first Ti-O distance is from the surface Ti-OH,
rather than from the disorder of the lattice.
Figure 6. First-shell Ti-O distance fitting for TiO2 particles with
different sizes. One shell model fits well for 200 Å particles (solid
line in top panel), but not for 50 and 30 Å particles (middle and bottom
panels, dotted lines, respectively). However, two-shell model fits
improved significantly the fits (solid lines).
In summary of this section, our XAFS measurements at the
Ti K-edge show shorter Ti-O distances for surface TiO2,
resulting from Ti-OH bonding. Minor disorder of the lattice
in smaller nanoparticles was observed based on the additional
loss of the third-shell O atom and the relative intensity changes
as a function of particle size for the A1, A2, and A3 peaks.
However, the Ti sites largely remain octahedral even in the 30
Å diameter particles. Because the interfacial electron/hole
transfer occurs via surface Ti or O atoms, the observed structural
changes around the surface Ti atoms in small TiO2 particles
could be responsible for the unique photochemistry displayed
in nanoparticle-assisted reactions.
2. Photoreduction of Metal Ions on Modified TiO2
Surfaces. There have been a number of studies of photore-
duction of metal ions on TiO2 surfaces.7-11 The modification
of the TiO2 nanoparticle surfaces with adsorbents has made some
photoreductions of metal ions feasible or improved the efficiency
of the reduction.13 However, several questions arise regarding
these photoreactions: (1) are metal ions reduced to the metal
on the TiO2 surface, (2) does the surface adsorber affect the
efficiency of the putative photoreduction, (3) what are the
structures of the metal complexes involved in the reaction, and
(4) what roles do the metal complexes play in the putative
photoreduction? The surface modifiers were chosen based on
their capability of binding both metal ions and the TiO2 surface,
acting only as a bridging agent. We present XAFS results for
photoreactions of Cu2+ and Hg2+ ions in TiO2 surfaces with
and without modification. The results provide information on
the metal oxidation states and geometry of the metal complexes.
Photoreduction of Cu2+ on TiO2 Surface. A series of XAFS
spectra were taken at the Cu K-edge for a mixture of 0.01 M
Cu2+ ions (from copper acetate) and TiO2 nanoparticles (50 Å
diameters) with and without the surface modifier Ala after
different lengths of time of xenon lamp illumination. Since the
is presented in Figure 6. Even for 200 Å TiO2 colloid particles,
the width of the first peak is broader than that of 400 Å particles,
and the average distance of the first-shell Ti-O bonds is 1.92
( 0.02 Å with a relatively large σ, instead of 1.95 ( 0.02 Å as
in the 400 Å particles. A single-shell model is not adequate to
fit the first-shell Ti-O distance in 50 Å particles. However, a
two-shell model with 17% of the first-shell Ti-O distance of
1.77 ( 0.04 Å and a longer Ti-O distance of 1.94 ( 0.02 Å
for the remaining 83% provides a reasonable fit. The first shell
of TiO2 particles with 30 Å diameters has a slightly higher
fraction (20%) of Ti-O with a distance of 1.84 ( 0.04 Å, and
the remaining Ti-O has a distance of 1.99 ( 0.02 Å. The fits
for the second-shell Ti-Ti distances and the third-shell Ti-O
distances are shown in Table 1. The coordination numbers are
still within the experimental error for the 200 Å particles, but
the second- and third-shell coordination numbers are reduced
for 30 and 50 Å diameter particles. Simultaneous reduction of
the second-shell coordination number and the size of the
particles suggests a correlation between the two trends. If the
coordination number reduction in small nanoparticles is due to
the surface TiO2 alone, the reduced fractions for the second
and third shells should be the same, because the surface Ti loses
the two shells simultaneously. However, disorder in the lattice
in small particles would cause only the reduction in the third
and higher shells because of the heterogeneous broadening. Our
results show that the lost fraction of the third shell is 58% and
67%, whereas that for the second shell is 20% and 38% for the
particles with 30 and 50 Å diameters, respectively. This extra
30-40% loss of the third-shell coordination numbers compared
to those of the second shell could be from the disorder of the
lattice structures, such as the “corner effect” where the distortion
is due to the large curvature of the particle surface.

Structures of TiO2 Nanoparticles
J. Phys. Chem. B, Vol. 101, No. 50, 1997 10693
TABLE 2: XAFS Fitting Results for Cu-Alanine
Complexes during Photoreduction on TiO2 Colloid Surfacesa
illumination
time (min)
shellb
N
R (Å)
rel σ2 (Å)
0
20
20
40
40
60
60
Cu2+
Cu2+
Cu0
4.2 ( 0.5
3.8 ( 0.5
0.8 ( 0.2
3.3 ( 0.5
1.7 ( 0.5
2.9 ( 0.5
2.6 ( 0.5
1.96 ( 0.02
1.96 ( 0.02
2.58 ( 0.02
1.96 ( 0.02
2.57 ( 0.02
1.97 ( 0.02
2.56 ( 0.02
0.001
0.002
-0.002
0.003
-0.001
0.004
-0.001
Cu2+
Cu0
Cu2+
Cu0
a Reference compounds, Cu-(Gly)2, Cu-O(N) distance 1.96 Å, N
) 4, and Cu foil, Cu-Cu distance 2.56 Å, N ) 12. b The shells here
are defined based on Figure 8, where the first major peak is from Cu2+
and the second peak is from Cu0.
spectrum (top trace, Figure 8). As the illumination proceeded,
this newly appearing peak grew at the expense of the first
Cu2+-O (or Cu2+-N) peak, and three other peaks appeared
between R ) 3 and 5 Å, also at the same position as those
observed in the copper foil spectrum (top trace, Figure 8).
Appearance of these four peaks during the illumination period
clearly confirmed the photoreduction of Cu2+ as well as the
formation of Cu0 clusters. After illuminating for 60 min, ∼50%
of Cu2+ adsorbed on the TiO2 nanoparticle surfaces was
photoreduced. Small metal copper particles became visible at
Figure 7. Cu K-edge (8.979 keV) XANES spectra for copper metal
and Cu2+/Ala/TiO2 at different illumination times.
that time. The XAFS data analyses for the first peak (Cu2+
)
and the second peak (Cu0) are listed in Table 2. A Cu-glycine
(Cu-Gly2) complex17 with an average Cu-O and Cu-N
distance of 1.96 ( 0.02 Å, and a copper foil with a nearest
Cu-Cu distance of 2.56 ( 0.02 Å,29 were used as reference
structures. Although the Cu-Cu distances beyond the first shell
are clearly shown through the cluster of the peaks between R
) 3 and 5 Å in the spectra after 40 min illumination, they could
appear after shorter illumination periods at a concentration that
is too low to detect. The growth of the Cu0 with illumination
time can also be monitored by XANES and XAFS spectra in
Figures 7 and 8. Photoreduction efficiency of Cu2+ on TiO2
colloid surfaces in the presence of copper acetate only was very
low, and little convincing evidence of Cu0 from XAFS spectra
could be observed after prolonged illumination (see the second
spectrum from the top in Figure 9).
The reoxidation of Cu0 in the mixture with TiO2/Ala was
observed after the photoreduced sample was exposed to air and
was recorded by XAFS (not shown) in a similar fashion as the
photoreduction. As the photoreduced sample was exposed to
air, the Cu0 peaks in FT-XAFS spectra gradually diminished
while the Cu2+ peaks appeared, a reverse process of the
photoreduction observed earlier.
Figure 8. FT-XAFS spectra for copper metal and Cu2+/Ala/TiO2 at
To know why the formation of the Cu-adsorber complex
affects the photoreduction efficiency and how the structures
surrounding Cu in the cases with and without Ala differ, FT-
XAFS spectra for various Cu2+ complexes (Figure 9) were
analyzed. The spectrum for the Cu-Ala complex (Cu-
[NHCHCH3COO]2, Cu-Ala2, second from the bottom) shows
not only the first shell but also two other shells at R ∼ 2.4 and
3.5 Å. This spectrum is virtually identical with that previously
observed for the Cu-Gly2 complex (Lytle, private communica-
tion). Thus, the Cu-Gly2 complex was used as the reference
compound for Cu-Ala2 structural determination. The X-ray
structure of the Cu-Gly2 complex shows that the Cu2+ atom
chelated directly with one of the O atoms of the carboxyl group
and the N atom of the amino group from each of the two Gly
molecules, forming square-planar, two five-membered rings.17
The mean Cu-O distance is 1.97 ( 0.02 Å, and the mean
Cu-N distance is 1.96 ( 0.02 Å. According to the X-ray
structure, the second shell in the spectrum of the Cu2+-Ala
different illumination times.
lifetimes of photogenerated charge carriers are very short due
to rapid recombination, only reactions with adsorbed species
lead to an efficient charge separation. We have found that with
the presence of Ala a significant amount of adsorbed Cu2+ was
reduced and the amount of Cu0 increased with the time of
illumination, whereas little photoreduction was observed for the
same sample without Ala. These results are shown in both
XANES (Figure 7) and FT-XAFS (Figure 8) spectra. Before
illumination, the FT-XAFS spectrum (not corrected for phases)
for the Cu2+/TiO2/Ala mixture has a dominant peak between R
) 1 and 2 Å and a minor peak at 3.3 Å. These peaks are
assigned to backscattering from O or N atoms from H2O or
Ala molecules directly chelated with Cu atoms and to the O
atoms of the second layer of solvated H2O, respectively.17 After
illuminating for 20 min, a new peak appeared between R ) 2.0
and 2.5 Å at the same position as the first shell in a copper foil

10694 J. Phys. Chem. B, Vol. 101, No. 50, 1997
Chen et al.
Figure 9. FT-XAFS spectra for different Cu2+ complexes.
Figure 10. Top: XAFS spectra of Cu2+/Ala/TiO2 before and after 60
min illumination. Bottom: XAFS spectrum of copper foil (dotted) and
a difference spectrum from subtracting 60% Cu2+/Ala/TiO2 spectrum
without illumination from that of a mixture after 60 min illumination.
The bottom two spectra overlap well, indicating the remaining spectrum
complex in Figure 9 is from the R-carbon and carboxyl carbon
atoms. The third shell in the spectrum is from a combination
of the methyl carbon and oxygen (not chelated to the Cu atom)
as well as the second layer of solvated H2O.17 The third-shell
peak in the vicinity of R ) 3.5 Å appears in other spectra in
Figure 9 as well. In most of the hydrated transition-metal ions,
this peak is due to backscattering from the second O atom in a
configuration of metal-O-H‚‚‚O-H, indicating the solvated
is from the Cu0 photoreduced from initial Cu2+
.
region if Cu+ is present. No extra peak was observed in addition
to the peaks for Cu0 and Cu2+. To see whether Cu+ is present,
we examined spectra from the following: (1) Cu foil, (2) Cu2+
/
H2O layer around Cu2+ 30
.
TiO2/Ala with no illumination, and (3) Cu2+/TiO2/Ala with 60
min illumination. If there is no Cu+ species present, the
spectrum from (3) should be a mixture of the spectra from (1)
and (2). Otherwise, there could be extra spectral features
representing the third Cu species. Within the error of the XAFS,
we did not find the third species. The remaining spectrum after
we subtracted 60% of Cu2+ + TiO2 + Ala without light from
the spectrum of Cu2+/TiO2/Ala with 60 min illumination
resembles the Cu foil spectrum well (see Figure 10). Our results
show that 40-50% Cu2+ adsorbed has been reduced after 1 h
illumination. Although a small amount of Cu+ could be present
in the mixture, this amount is beyond our detection limit.
Therefore, we can conclude that no observable amount of Cu+
is present in the system during the light-induced reduction on
TiO2 colloid surfaces.
The second-shell peak (at R ∼ 2.4 Å without phase correction
in Figure 9) is only present in the Cu-Ala2 complex, which
indicates that the structure of the complex is well-defined by
the chelation with two five-membered rings. When TiO2 was
added, this second-shell peak disappeared, suggesting that the
original five-membered ring structure in the Cu2+-Ala2 complex
has been destroyed. If one of the two five-membered rings
survives, the second shell should still be present. Two possible
structures of the Cu surroundings in the Cu2+/TiO2/Ala mixture
are suggested based on the experimental observation of the
similarity between the FT-XAFS spectra of Cu2+/H2O and
Cu2+/TiO2/Ala: (1) Cu2+ binds to only one end of each Ala
and to O atoms of surface Ti; (2) Cu2+ binds only on TiO2
surfaces via dynamic electrostatic interactions. Both ways of
binding will cause disappearance of the second-shell structure
in the FT-XAFS spectrum. However, the former is considered
more plausible because only when Cu2+ binds to Ala can the
photoelectron that is transported away from the TiO2 surface
Photoreduction of Hg2+ on TiO2 Surface with and without
Thiolactic Acid. Photoreduction of Hg2+ on TiO2 surfaces was
examined by XAFS in a similar fashion as for Cu2+ photo-
reduction. The formation of Hg0 on TiO2 colloid surfaces was
monitored at the Hg LIII-edge. Mixtures of 0.01 M Hg2+/TiO2
with and without 0.02 M thiolactic acid (TLA, CH3CHSHCOOH)
were illuminated for 1 h. We added 0.1 M methanol as a hole
scavenger to the sample without TLA, because its oxidation
potential is the same as that of TLA (E°(TLA/TLA+) ) E°(CH3-
OH/CH2O) ≈ 1.2 V vs NHE) and it does not chelate with Hg2+
ions. Figure 11 shows XANES spectra of the filtered mixture
before and after illumination, along with the spectra of other
relevant Hg complexes. The XANES spectra of Hg2+/TLA with
and without TiO2 before the illumination are very similar and
distinctively different from that of Hg2+/TiO2, indicating direct
chelation of TLA with Hg2+. The presence of Hg2+ in aqueous
solution is characterized by a peak in the middle of the rising
edge at 4.2 eV, and that peak is absent in the Hg0 spectrum.
be effectively transferred to Cu2+
. Alanine has the same
electrochemical properties as the acetate used in both solutions
(E°(Ala/Ala+) ) E°(CH3COO-/CH2CO) ) 2.2 V vs NHE).31
Therefore, only the difference in binding between Cu2+ and the
TiO2 nanoparticle surface leads to the observed difference in
photoreduction efficiency when TiO2 is present.
The other important question is whether Cu+ is present during
the photoreduction. It has been reported that photoreduction
of Cu2+ at the TiO2 surface produces a mixture of Cu2O and
Cu0.8 The copper K-edge is sensitive to the oxidation states of
the metal ions, and Cu+ has a preedge structure distinct from
that of Cu2+. The Cu+ K-edge is between that for Cu0 and
Cu2+ at about 5 eV above the Cu0 K-edge.32 Therefore, we
expect that an edge structure would show up at this energy

Structures of TiO2 Nanoparticles
J. Phys. Chem. B, Vol. 101, No. 50, 1997 10695
TABLE 3: Structural Parameters of Hg Complexesa
complex
shell (atom)
N
R (Å)
rel σ2 (Å2)
Hg2+/H2O
1 (O)
1 (S)
2.9 ( 0.5 2.02 ( 0.02
0.004
Hg2+/TLA
3.8 ( 0.5 2.35 ( 0.02 -0.004
Hg2+/TiO2 (light)
Hg2+/TiO2 (dark)
Hg2+/TiO2/TLA
(light)
1 (O)
1 (S)
2.9 ( 0.5 2.05 ( 0.02 0.004
1.2 ( 0.4 2.39 ( 0.03 -0.005
0.4 ( 0.2 2.60 ( 0.03 -0.008
2.1 ( 0.5 2.39 ( 0.02 -0.009
Hg2+/TiO2/TLA
(light)
2 (S)
1 (S)
2 (O)
Hg2+/TiO2/TLA
(dark)
Hg2+/TiO2/TLA
(dark)
0.7 ( 0.3 2.04 ( 0.04
0.005
a The reference compounds used here are HgO for Hg-O shell and
HgS(black) for Hg-S shell. Hg-O distance in HgO is 2.03 Å with a
coordination number of 2, and Hg-S distance in HgS(black) is 2.53
Å with a coordination number of 4.
solvated H2O or surface O atoms in TiO2 nanoparticles. After
1 h photoillumination, this peak almost completely disappeared,
while a group of peaks that resemble those in Hg0 appeared,
indicating an almost 100% conversion from Hg2+ to Hg0. In
the Hg2+/TiO2/TLA system without methanol, the first-shell
peak representing Hg2+ is shifted to 1.93 Å (no phase correc-
tion), indicating a different ligation structure of Hg2+. Over
50% of this peak remained after 1 h photoillumination, while
the Hg0 peaks grew. The FT-XAFS spectra for the Hg2+/TiO2/
TLA system after illumination indicated that less than 50% Hg2+
was converted to Hg0. This is consistent with current doubling
mechanisms in the presence of methanol, previously described
in the literature.32,33 As a current-doubling reagent, methanol
is expected to double the efficiency of photoreduction compared
to other reagents with similar redox potentials. However, less
than 50% Hg2+ photoreduction with TLA suggests that TLA
does not enhance the reaction efficiency as Ala in the Cu2+
photoreduction case.
Figure 11. Hg LIII-edge (12.284 keV) XANES spectra of different
Hg-containing samples (indicated when photoillumination is applied).
Figure 12 also shows the FT-XAFS spectra (without phase
correction) of various Hg complexes along with the Hg0
spectrum. The positions of the first-shell atoms for Hg2+/TLA
and Hg2+/H2O are drastically different. The shorter first-shell
distance in the Hg2+/H2O spectrum is from solvation of Hg2+
in water and chelation with O atoms, whereas the longer first
shell in the Hg2+/TLA spectrum is an indication of the Hg-S
bond in this complex. When TiO2 was added, the spectrum of
the Hg2+/TLA complex was largely retained. This situation
differs from that in the Cu2+-Ala complex where the structure
of the complex is significantly altered when TiO2 is added.
To find out how TLA affects the photoreduction efficiency,
we need to examine various structures of Hg complexes. We
used HgO and HgS (black) as the reference compounds. The
former has a zigzag chain structure, where Hg and O alternate.20
Each Hg has two O atoms as the nearest neighbors, and the
Hg-O distance is 2.03 Å. Each Hg in black HgS has four
tetrahedral neighboring S atoms, and the Hg-S distance is 2.53
Å.19
Using these reference compounds, Table 3 lists the structural
parameters for several Hg complexes. Since the Hg-O distance
is ∼2.0 Å and the Hg-S distance varies from 2.3 to 2.5 Å, and
the phases of O and S backscattering differ significantly, the
atomic species chelated to Hg can be distinguished. Hg2+ in
H2O has an average first-shell Hg-O distance of 2.02 Å. In
the presence of TiO2, this Hg-O distance became slightly
longer, 2.05 Å. This result suggests a direct binding of Hg on
TiO2 surfaces. Hg2+ in Hg2+/TLA has a coordination number
of 3-4 and a Hg-S distance of 2.35 Å, shorter than the Hg-S
distance in HgS. The data cannot be fit using the Hg-O
Figure 12. FT-XAFS spectra of XANES spectra of different Hg-
containing samples (indicated when photoillumination is applied). The
spectra were rescaled to align them at the most intense peak of each
spectrum.
On the basis of this observation, we found that the XANES
spectrum of adsorbed Hg2+ on TiO2 (methanol was added) after
1 h illumination had essentially the same spectral features as
Hg0. Thus, almost all the Hg2+ adsorbed was reduced.
However, the photoreduction of the adsorbed Hg2+ is incomplete
when TLA is present (without methanol) under similar experi-
mental conditions, because the peak at ∼4.2 eV is still visible.
This difference in reaction efficiencies for the two systems
can also be demonstrated by FT-XAFS spectra in Figure 12.
Before the 1 h photoillumination in Hg2+/TiO2 with methanol,
the first-shell peak was at R ) 1.72 Å (no phase correction),
which is characteristic of Hg2+ surrounded by O atoms in

10696 J. Phys. Chem. B, Vol. 101, No. 50, 1997
Chen et al.
for the observed XAFS spectra, it cannot readily explain the
difference in photoreduction enhancement between Ala and TLA
in these two systems, because they both should have one
functional group attached to the metal and the other functional
group to Ti on the nanoparticle surface. The difference in the
photoreduction efficiency deserves further explanation and
study. Two possible causes are proposed based on the
experimental results. First, the affinity of S for Hg may be
higher than Cu for N (in NH2), because the Cu to first-shell
atom distance in Cu2+/Ala/TiO2 is the same as Cu in H2O,
indicating that Cu atoms may be associated with H2O or
carboxyl instead of N atoms of Ala. This higher affinity
between the S-containing ligand and the metal ion may affect
the reaction efficiency. Second, the possible formation of HgS
could reduce the efficiency of photoreduction to Hg0. At this
point, our observations imply that a proper balance between
the affinities of the metal to the adsorber and surface O atom
of TiO2 may be one of the keys in selecting a proper surface
adsorber for an enhanced photoreduction efficiency of the metal
ions on nanoparticle surfaces.
Acknowledgment. This work is supported by the Chemical
Science Division, Office of Basic Energy Science, U.S. Depart-
ment of Energy, under Contract W-31-109-Eng-38. We thank
Dr. Olga Mic´ic´ for her generous help in providing TiO2 particles
with 30 and 200 Å diameters, Dr. David M. Tiede for his
inspiring discussions, and Dr. Farrel Lytle for his Hg0 spectra
and helpful suggestions. We also appreciate the support from
Dr. Pedro A. Montano and the personnel at Beamlines X6B,
X18B, and X19A, National Synchrotron Light Source,
Brookhaven National Laboratory.
Figure 13. Fits for the first shell in Hg2+/TiO2/TLA complex.
distance and the O backscattering phase from HgO as reference
parameters. Thus, the chelated atoms in the Hg-TLA complex
are most likely S atoms from TLA molecules, and only the S
end of TLA binds to Hg, unlike the Cu-Ala complex where
both functional groups from each alanine participate in the
chelation.17 This agrees with Hg-S distances observed in other
Hg complexes with small organic molecules.34,35
The first peak in the FT-XAFS spectrum of the Hg2+/TiO2/
TLA complex fits much better with a two-shell model, Hg-S
at 2.39 Å and Hg-O at 2.04 Å (see Figure 13). The
coordination number ratio of Hg-S and Hg-O is 2.1/0.7,
approximately 3:1, indicating that one of the TLAs may be
replaced by a surface O atom from TiO2. After 1 h illumination,
the Hg nearest neighbors in Hg2+/TiO2/TLA can be fit to two
S shells with a relative ratio of 3:1 and Hg-S distances of 2.39
and 2.60 Å, respectively. Thus, these two distances may be
from Hg-S that bonds to TLA and possible formation of HgS,
where Hg-S is 2.54 Å. The next-nearest neighbor of Hg in
Hg2+/TiO2/TLA after photoillumination is Hg having an Hg-
Hg distance as in Hg0 (see Figure 12).
In summary of this section, the products of the photoillumi-
nation of Cu2+ or Hg2+ on TiO2 nanoparticle surfaces with and
without surface adsorbers were analyzed with XAFS. Alanine
dramatically enhanced photoreduction of Cu on TiO2 nanopar-
ticle surfaces, whereas thiolactic acid did not enhance or even
hindered Hg photoreduction. Although both surface adsorbers
chelate with the metal ions in the absence of TiO2 nanoparticles,
this chelation is drastically changed in the Cu-Ala complex
but is largely retained in the Hg-TLA complex when TiO2 is
present. This difference could be explained as follows. First,
in the CuAla2 complex each of the two Ala chelates with the
Cu atom, forming a stable five-membered ring where both
carboxyl and amino groups are involved. When TiO2 is present,
one of the two functional groups of Ala, most likely the carboxyl
group, chelates with a Ti atom on the nanoparticle surface,
resulting in the collapse of the five-membered ring structure.
However, in the Hg2+/TLA complex, the sulfhydryl group from
three or four TLA molecules is the sole functional group that
chelates with Hg2+, leaving the carboxyl group free at the other
end of the molecule. When TiO2 is added, the carboxyl group
will chelate with TiO2 while retaining the first-shell S atoms
around Hg. Although this explanation is completely reasonable
References and Notes
(1) Homogeneous and Heterogeneous Photocatalysis; Pelizzetti, E.,
Serpone, N., Eds.; NATO ASI Ser. C 1986, 174.
(2) PhotocatalysissFundamentals and Applications; Serpone, N.,
Pelizzetti, E., Eds.; John Wiley and Sons: New York, 1989.
(3) Stafford, U.; Gray, K. A.; Kamat, P. V. Heterog. Chem. ReV. 1996,
3, 77.
(4) Serpone, N. Res. Chem. Intermed. 1994, 20, 953.
(5) Hoffmann, M, R.; Martin, S. T.; Choi, W. Y.; Bahnemann, D. W.
Chem. ReV. 1995, 95, 69.
(6) Linsebigler, A. L.; Lu, G. Q.; Yates, J. T. Chem. ReV. 1995, 95,
735.
(7) Serpone, N.; Lawless, D.; Terzian, R.; Minero, C.; Pelizzetti, E. In
Photochemical ConVersion and Storage of Solar Energy; Kluwer Academic
Publishers: Dordrecht, The Netherlands, 1991; pp 451-475.
(8) (a) Foster, N. S.; Noble, R. D.; Koval, C. A. EnViron. Eng. 1988,
27, 651. (b) Jacobs, J. W.; Kampers, F. W. H.; Rikken, J. M. G.; Bulle-
Lieuwma, C. W. T.; Koningsberger, D. C. J. Electrochem. Soc. 1989, 136,
2914.
(9) Foster, N. S.; Lancaster, A. N.; Noble, R. D.; Koval, C. A. Ind.
Eng. Chem. Res. 1995, 34, 3865.
(10) Forouzan, F.; Richards, T. C.; Bard, A. J. J. Phys. Chem. 1996,
100, 18123.
(11) (a) Rajh, T.; Ostafin, A. E.; Micic, O. I.; Tiede, D. M.; Thurnauer,
M. C. J. Phys. Chem. 1996, 100, 4538-4545. (b) Rajh, T.; Tiede, D. M.;
Thurnauer, M. C. J. Non-Cryst. Solids 1996, 207, 815-820.
(12) Klabunde, K. J.; Stark, J.; Koper, O.; Mohs, C.; Park, D. G.; Decker,
S.; Jiang, Y.; Lagadic, I.; Zhang, D. J. Phys. Chem. 1996, 100, 12142.
(13) Thurnauer, M. C.; Rajh, T.; Tiede, D. M. Acta Scand. 1997, 51, 5.
(14) X-ray Absorption: Principles, Applications, Techniques of EXAFS,
SEXAFS, and XANES; Koningsberger, D. C., Prins., R., Eds.; Wiley: New
York, 1988.
(15) Rajh, T.; Saponjic, Z. V.; Micic, O. I. Langmuir 1992, 8, 1265.
(16) Bragg, L.; Claringbull, G. F. Crystal Structures of Minerals; G.
Bell & Sons Ltd.: London, 1965.
(17) Tomita, K.; Nitta, I. Bull. Chem. Soc. Jpn. 1961, 34, 286.
(18) Åsbrink, S.; Norrby, L. J. Acta Crystallogr. 1970, B26, 8.
(19) Aurivillius, K. Acta Crystallogr. 1956, 9, 685.
(20) Aurivillius, K. Acta Chem. Scand. 1964, 18, 1552.
(21) Serpone, N.; Lawless, D.; Khairutdinov, R. J. Phys. Chem. 1995,
99, 16646.
(22) Grunes, L. A. Phys. ReV. B 1983, 27, 2111.

Structures of TiO2 Nanoparticles
J. Phys. Chem. B, Vol. 101, No. 50, 1997 10697
(23) Greegor, R. B.; Lytle, F. W.; Sandstrom, D. R.; Wong, J.; Schultz,
P. J. Non-Cryst. Solids 1983, 55, 27.
(24) Pei, S.; Zajac, G. W.; Kaduk, J. A.; Faber, J.; Boyanov, B. I.; Duck,
D.; Fazzini, D.; Morrison, T. I.; Yang, D. S. Catal. Lett. 1993, 21, 333.
(25) Yamashita, H.; Ichihashi, Y.; Anpo, M.; Hashimoto, M.; Louis,
C.; Che, M., J. Phys. Chem. 1996, 100, 16041.
(26) Ruiz-Lo´pez, M. F.; Mun˜oz-Pa´ez, A. J. Phys: Condens. Matter
1991, 3, 8981.
(27) Bohr, F.; Ruiz-Lo´pez, M. F.; Mun˜oz-Pa´ez, A. Catal. Lett. 1993,
20, 59.
(28) Poumellec, B.; Marucco, J. F.; Touzelin, B. Phys. ReV. B 1987,
35, 2284.
(29) Oxtoby, D. W.; Nachtrieb, N. H. Principles of Modern Chemistry;
Saunders College Publishing: Philadelphia, 1987.
(30) Miyanaga, T.; Watanabe, I.; Ikeda, S.; Tashiro, K.; Fujikawa, T.
Bull. Chem. Soc. Jpn. 1988, 61, 3199.
(31) (a) Hasegawa, K.; Neta, P. J. Phys. Chem. 1978, 82, 859. (b) Tamba,
M.; Badiello, R. Ind. J. Radiat. Biol. Relat. Stud. Phys. Chem. Med. 1985,
48, 259. (c) Encyclopedia of Electrochemistry of the Elements, Organic
Section; Bard, A. J., Lund, H., Eds.; Marcel Dekker: New York, 1978;
Vol. XII, p 269.
(32) Nogami, G.; Kennedy, J. H. J. Electrochem. Soc. 1989, 136, 2583.
(33) Micic, O. I.; Zhang, Y.; Cromack, K. R.; Trifunac, A. D.; Thurnauer,
M. C. J. Phys. Chem. 1993, 97, 13284-13288.
(34) Kau, L.-S.; Spira-Solomon D. J.; Penner-Hahn, J. E.; Hodgson, K.
O.; Solomon, E. I. J. Am. Chem. Soc. 1987, 109, 6433.
(35) Wright, J. G.; Tsang, H.-T.; Penner-Hahn, J. E.; O’Halloran, T. V.
J. Am. Chem. Soc. 1990, 112, 2434.