J. Phys. Chem. B 2000, 104, 9851-9858
9851
Surface and Electronic Structure of Titanium Dioxide Photocatalysts
S. A. Bilmes*,† and P. Mandelbaum
INQUIMAE-Departamento de Quı́mica Inorgánica, Analı́tica y Quı́mica-fı́sica, Facultad de Ciencias Exactas y
Naturales, UniVersidad de Buenos Aires, Ciudad UniVersitaria Pab. II, C1428EHA Buenos Aires, Argentina
F. Alvarez*,‡ and N. M. Victoria
Instituto de Fı́sica “Gleb Wataghin”, UniVersidade Estadual de Campinas Unicamp,
13083-970 Campinas, São Paulo, Brazil
ReceiVed: March 16, 2000; In Final Form: July 21, 2000
TiO2 films prepared by sol-gel route are active photocatalysts for the oxidation of organics in photoelectrochemical cells. The as-grown films for photocatalysis applications and those exposed to Ar+ or H2++Ar+
ion bombardment are characterized by different spectroscopic methods, such as X-ray diffraction (XRD),
atomic force microscopy (AFM), UV-vis transmittance, photothermal deflection spectroscopy (PDS) and
X-ray photoelectron spectroscopy (XPS), as well as by conductance. This material has defects associated
with oxygen vacancies produced during the sample preparation which support nondissociative adsorption of
O2 when films are exposed to air. Charge transfer from reduced Ti species to adsorbed dioxygen leads to
Ti-O2- surface complexes that are partially removed by heating at 200 °C, and fully removed after 30 min
ion bombardment. By comparison with the relatively well-understood structural defects of bombarded TiO2
we arise to a quite complete structural model of the as grown material which corresponds to an amorphous
semiconductor possessing relative low disorder and density of states as compared with a pure amorphous
material. These TiO2 films are modeled as low size crystalline domain embedded in an amorphous matrix
whose electronic structure exhibit exponential band tails and a narrow band close to the conduction band.
The latter is fully or partially occupied depending on the presence of adsorbed electron scavengers such as
dioxygen.
Introduction
The need of new and effective methods for cleaning polluted
air or water has prompted considerable efforts in the research
for complete mineralization of a wide range of dissolved or
dispersed organic compounds. Since the pioneering work of
Carey et al.,1 heterogeneous photocatalysis has been proven for
decomposition of many organic compounds, conversion of
inorganic anions, and metal removal by photochemical reduction.2-5 Several commercial devices have been developed
for decontamination of air,6 but the low quantum yield of most
photocatalytic oxidation reactions in aqueous systems translates
into comparatively high power costs. The efficiency of the
overall process decreases as the catalyst is immobilized,7 and
it is not clear whether the interfacial structure or mass-transfer
limitations can account for these differences. However, the fact
that the photodegradation rate can be improved with TiO2 films
under bias in a photoelectrochemical cell8-11 provides evidence
that interfacial charge-transfer processes play an important role.
Thus, one fundamental question that is still open12 concerns to
the structure of the TiO2 film/electrolyte interface which in turn,
is determined by the electrolyte composition, and the semiconductor surface composition and electronic structure.
The interest of TiO2 films exceeds photocatalytic applications
and they are widely used as optical coatings and sensors.13-15
Their synthesis by sol-gel route has been proved to be effective
* Corresponding authors.
† E-mail: sarabil@q3.fcen.uba.ar.
‡ E-mail: alvarez@ifi.unicamp.br.
for producing photocatalysts operating under bias,10,11 for use
in solar cells,16,17 self-cleaning windows,18 and electrochromic
devices.19 The TiO2 surface is covered by adsorbates, mainly
dioxygen and water from the surrounding environment, which
play an important role in interfacial charge transfer: water
adsorbs dissociatively giving rise to efficient hole traps, such
as surface -OH, and dioxygen is an efficient electron scavenger
in nonaqueous media.14,20-25 From the structural point of view,
these materials are usually nanocrystalline structured in a
network of mesoscopic oxide particles,26-29 implying their
electronic structure description need to include localized states
from defects and grain boundaries, as well as band tails in the
gap.
The ideal TiO2 (110) surface contains 5-fold Ti4+ coordinated
to O2-, but even at room temperature a real surface has a non
negligible density of oxygen vacancies which give rise to
localized defects of Ti2+ and Ti3+.20,30,31 The presence of reduced
Ti species on the surface and underneath plays an important
role in adsorption.20 In particular, for the adsorption of dioxygen
it has been demonstrated that both molecular and dissociative
channels occur at defects with charge transfer from reduced
Ti.20,25,32-35 On the other hand, little is known about the structure
of TiO2 films and their interfaces in air or in aqueous
electrolytes;12,37,38 it is expected that crystallinity, density of
defects, and electronic structure depend on the interaction with
the substrate, synthesis pathway, and thermal treatment variables.
Then, to develop an efficient photocatalysts for air or water
10.1021/jp0010132 CCC: $19.00 © 2000 American Chemical Society
Published on Web 10/03/2000
Bilmes et al.
9852 J. Phys. Chem. B, Vol. 104, No. 42, 2000
detoxification it is mandatory an exhaustive structural characterization of the TiO2 film forming the semiconductor/electrolyte
interface.
In this work we present a comprehensive study containing
results on the surface structure, chemical composition, optical
absorption and electric transport for TiO2 films prepared by solgel route that are active photocatalysts for the oxidation of
organics in photoelectrochemical cells.11,39 The as-grown films
for photocatalysis applications are characterized by different
spectroscopies methods, such as X-ray diffraction (XRD), atomic
force microscopy (AFM), UV-vis transmittance, photothermal
deflection spectroscopy (PDS), and X-ray photoelectron spectroscopy (XPS). The conductance of the samples was also
measured. These results are compared with those obtained for
the same films exposed to Ar+ or H2++Ar+ ion bombardment.
Moreover, by a comparison with the relatively well understood
structural defects of bombarded TiO2 we arise to a quite
complete structural model of the as-grown material.33,40-42
Materials and Methods
Synthesis of Films. Titanium dioxide films were prepared
by the sol-gel route. Sols prepared by hydrolysis of Ti(IV)
butoxide at pH < 1 were dialyzed against water up to pH 3.4.
The resulting gel was spin coated either onto quartz or onto
conducting indium-tin oxide (ITO) glass and heated at 90 °C
for 5 min. This procedure was repeated six times; after the sixth
deposit, the assembly was sintered 6 h at 400 °C in air. The
thickness of each layer was 50 nm, as determined by AFM
images of the partially covered substrate in which a defined
step was formed by removing a strip of Scotch tape previously
adhered. The average thickness of the films was (300 ( 50)
nm as determined by profilometry and by atomic force
microscopy (AFM).
Experimental Methods. X-ray diffractograms were obtained
with a Siemens D5000 operating with Cu KR radiation source
filtered with a graphite monochromator (λ ) 1.5406 Å). The
incidence angle was 3°, and both emission and detection slits
were 1 mm. The length of the crystalline domain was calculated
by Scherrer equation43 with an instrumental factor of 0.03°.
AFM images were obtained with a PSI Autoprobe CP with
a 5 µm Si contact ultralever scanner. All scans were run in
contact regime at 10 pN and 2 Hz. The RMS roughness of the
surface was calculated as a standard deviation on the z-scale
with the PSI program.
All the X-ray photoemission spectra were taken at room
temperature with a Vacuum Generator (VG) CLAMP 3 (0.3
eV resolution) operating in the constant-analyzer-energy mode
in a UHV chamber (<10-9 mb) using the Al KR line (VG XR3
X-ray source, hν ) 1486.6, width 0.85 eV). The mirror analyzer
pass energy was 10 eV. The binding energies were calibrated
with reference to sputtered cleaned Ag standard. To confirm
that prolonged evacuation does not modify the surface chemical
composition, several test spectra were taken over 14 h. The XPS
chamber is attached in-line with that for ion treatment in such
a way of allowing in situ bombardment experiments. The
bombardment experiments were performed by using a Kaufman
cell feeding with Ar or a H2 + Ar mixture with energies 100
eV and a current density of 10 mA.cm-2. More experimental
details of the treatment chamber are published elsewhere.44
UV-vis transmittance spectra of films deposited onto quartz
were taken with a single beam spectrophotometer (Perkin-Elmer
λ9). High sensitive spectra in the 310-850 nm range were taken
in a standard photothermal deflection spectroscopy (PDS)
system. In PDS experiments, the sample was irradiated with
Figure 1. (a) DRX spectrum of TiO2 films 300 nm thick deposited
onto ITO. (b) AFM image for the same film.
chopped monochromatic light from a 1000 W Xe lamp. CCl4
was employed as deflecting medium, and the in phase deflection
of the He-Ne laser was detected with a photodiode.45 The
complete spectra in the 310-850 nm range were obtained by
matching the PDS and transmittance spectra in the 3.5-3.7 eV
region. Since the PDS is an optical absorption relative measurement, scaling the PDS with the visible spectra is a standard
procedure.46
Conductance measurements of films deposited onto quartz
were performed with 1 mm gap and 10 mm wide coplanar
evaporated Al contacts. The conductivity was calculated assuming uniform slab of material between the two aluminum
strip contacts.
Results
TiO2 Films as-Grown for Photocatalysis. Figure 1.a shows
the X-ray difractogram of TiO2 film deposited onto ITO
substrate. The broad peak at 2θ ) 25.3 is assigned to anatase
structure, and those at 21.5 and 30.5 to the indium oxide from
present in the substrate. The size of crystalline domains is (10
( 2) nm, as determined from the half width anatase peak by
Scherrer equation. AFM images (Figure 1b) show a surface
structure composed by grains with (300 ( 50) nm average size,
twice the average grain size of the ITO substrate. AFM images
recorded after deposition of each layer show that roughness
decreases from that of ITO on increasing fim thickness.
Parameters from AFM for the naked substrate, one-layer and
six-layer deposit is shown in Table 1.
Titanium Dioxide Photocatalysts
J. Phys. Chem. B, Vol. 104, No. 42, 2000 9853
TABLE 1: Parameters from AFM Image Analysis for TiO2
Films Grown for Photocatalysis
I TO substrate
one-layer TiO2
six-layer TiO2
rms roughness (Å)
real area/
geometric area
grain
size(nm)
50 ( 10
40 ( 10
20 ( 5
1.2 ( 0.1
1.1 ( 0.1
1.02 ( 0.02
150 ( 30
200 ( 30
300 ( 50
Figure 3. UV-vis spectrum of TiO2 films grown for (0) photocatalysis. Inset: plots for (left) indirect and (right) Tauc functions. For
the sake of comparison, spectra for a rutile crystal at (+) 10 K53 and
for an amorphous ()) TiO2 film54 are also plotted.
Figure 2. XPS spectrum of TiO2 films grown for photocatalysis in
the Ti 2p (a) and O 1s (b) binding energy (BE) regions. Dashed line in
(b) is the spectrum recorded after 30 min at 200 °C.
Figure 2 shows characteristic XPS spectra for a photocatalytic
TiO2 film in the Ti2p (Figure 2a) and O 1s (Figure 2b) electron
binding energy regions. Peaks at 458.7 and 464.5 eV correspond
to 2p3/2 and 2p1/2 core-levels of Ti(IV) bound to oxygen,
respectively.34,41,47 Both peaks are symmetric, and FWHM of
Ti2p3/2 is 1.25 eV, as in data reported either for “defect-free”TiO2 (110) or for TiO2 films prepared by sol-gel.34,48 In the O
1s binding energy region, the peak at 530.9 eV corresponds to
O 1s core-level of oxygen atoms bound to Ti(IV),34,48 whereas
that at 533.7 eV can be tentatively assigned to adsorbed
dioxygen. The latter can be partially removed by heating at 200
°C (dashed line in Figure 2.b), procedure that has no influence
on the height, width and location of other XPS peaks. The O/Ti
ratio calculated by integration of O 1s (530.9 eV) and Ti2p
bands, taking into account the sensitivity factor of each element
is 1.9.49 It should be noted that no carbon from atmospheric or
other organic contamination was detected after introducing the
samples in the UHV chamber for XPS analysis and reaching
the working vacuum (<1 × 10-9 mb), and no signals from In
or Sn were measured ruling out the possibility of pinholes.
The UV-vis spectrum of TiO2 films deposited on quartz
obtained by matching the PDS and transmittance spectra in the
3.5-3.7 eV region is shown in Figure 3 (square). In a
semiconductor containing gap-states, the absorption spectrum
has three well-defined regions referred to the energy gap, Eg:50
(a) E > Eg, corresponding to deep transitions between valence
and conduction bands states; (b) an intermediate, and in some
cases, near exponential region at E ≈ Eg, with absorption due
to the convolution of optical transitions from valence band states
to localized conduction band-tail states (or from localized
valence band-tail states to conduction band states); and (c) E
< Eg due to transitions from valence band states to localized
mid-gap states (or from localized mid-gap states to conduction
band states).
The elucidation of what type of mechanism is governing the
different parts of the absorption spectrum in TiO2 (Figure 3,
square symbols) is not an easy task. The existence of mid-gap
states and a fairly well defined exponential tail strongly suggests
characteristics commonly found in amorphous semiconductors.50
As it is well-known, region (a) of the absorption spectra
determine the semiconductor optical gap. In the attempt to
identify the most probable type of optical transitions we have
plotted R1/2 vs hν and (Rhν)1/2 vs hν. These plots are adequate
to describe the spectra of crystalline51 and amorphous52 materials, respectively. The two procedures lead to the similar gap
energy value (Eg ) 3.5 eV, inset Figure 3) close to that reported
for amorphous TiO1.92 films.54 However, the curve representing
indirect transitions expand for a shorter range of values
suggesting that this mechanism is less probable to represent the
actual transitions. Assuming that the intermediate region (3.2
< E < 3.7 eV) corresponds to region (b) discussed above, the
associated slope of the absorption curve defines the Urbach
energy, EU ) [d ln R/d(hν)]-1 ) 120 meV.50 This parameter is
generally accepted as a measure of the material disorder. For
the sake of comparison, Figure 3 includes data for a rutile crystal
at 10 K reported by Tang,53 and for an amorphous TiO2 film
grown by reactive sputtering.54 It is interesting to note that TiO2
films active in photocatalysis are mid way between these two
limits; the Urbach energy is approximately six-fold larger than
that of the crystal, and the absorption values of the amorphous
film in the subgap region is roughly 2 orders of magnitude
larger.
Room-temperature conductivity of TiO2 films as grown for
photo- and photoelectrocatalysis is relatively low in air, σ )
10-8 Ω-1cm-1. This value is similar to those found for
amorphous films prepared by sol-gel,26,48 and approximately
3 orders of magnitude lower than those of crystalline films.55
Figure 4 shows log σ vs 1/T plots measured in the 298K - 410
K temperature range in air (circles) and at 10-3 Torr (squares).
Heating in air up to 410 K does not improve conductivity, but
it increases after the chamber evacuation to 10-3 Torr. Conductance becomes thermally activated at the lower pressure
(Figure 4, branch I), and heating 40 min at 410 K at 10-3 Torr
raises the conductivity up to 3 orders of magnitude. Cooling
Bilmes et al.
9854 J. Phys. Chem. B, Vol. 104, No. 42, 2000
Figure 4. Conductivity change with temperature for as grown TiO2
films in (b) air and at (0) 10-3 Torr. Branches labeled I and II in the
plot at 10-3 Torr correspond to the temperature program inset in the
figure.
down the sample up to 295 K reduces the conductivity in an
activated fashion from 2 × 10-4 to 2 × 10-5 cm Ω-1 (Figure
4, branch II). The abrupt increase in conductivity upon heating
at 10-3 Torr is reversed in few seconds (below our measuring
time window) by further exposure to air.
Ion Bombarded TiO2 Films. Low energy ion bombardments
of TiO2 are widely used for producing compositional changes
by preferential sputtering of oxygen, inducing damage, disorder,
and reducing species.33,40-42 TiO2 films as grown for photoelectrocatalysis also exhibit drastic changes upon Ar+ or Ar+H2+ ion bombardment. The effect of bombardment on the X-ray
photoelectron spectra is the broadening of the bands corresponding to Ti 2p core levels (Figure 5a) and to a lower extent
that corresponding to O 1s electrons (Figure 5b). The broadening
of Ti 2p peaks is also found in ion bombarded crystalline TiO2
(110), and has been attributed to the formation of Ti species
with higher density of negative charge, i.e., Ti3+ and Ti2+.34,56
The energy of the O 1s core-level of oxygen atoms bound to
Ti remain at the same energy as the sample is bombarded. On
the other hand, the peak assigned to adsorbed O2 diminishes
and shifts to higher energies on ion bombardment (Figure 5b).
The fraction of each Ti species XTi was determined from the
ratio of areas obtained by a deconvolution of their contribution
to the overall spectra (Figure 5a, inset). This procedure was
done assuming that all bands are Gaussian, have the same fwhm,
and that the area under the peak assigned to 2p3/2 (458.7 eV)
core level is twice that of corresponding to 2p1/2 (464.5 eV).47
The location found for Ti2+ (2p3/2, 460.8 eV; 2p1/2, 455.4 eV)
and Ti3+ (2p3/2, 462.4 eV; 2p1/2, 457.4 eV) core levels is
coincident with that found for other reduced TiO2 materials.34,41,47 The evolution of XTi upon ion bombardment exposure
is shown in Figure 6.a. After 3 min ion bombardment the
composition of the (reduced) surface saturates to 40% Ti4+ and
60% reduced Ti (Ti3+ and Ti2+). The spectral deconvolution in
the O 1s binding energy region was done relaxing the condition
of equal FWHM Gaussians (Figure 5b, inset). In this process
was not possible to distinguish between oxygen bound to Ti3+
or Ti2+, and the peak located at 531.5 eV corresponds to oxygen
bound to both reduced Ti species. The fraction of oxygen atoms
bound to different Ti species XO follows the same trend found
for Ti species (Figure 6.b), whereas adsorbed dioxygen is
completely removed after 30 min of ion bombardment. It is
worth to remark that the same results are obtained with different
bombardment ion mixtures (Ar+ or Ar++H2+), confirming that
the main effect of ion bombardment is preferential oxygen
sputtering by Ar+.
The increasing density of defects and disorder upon ion
bombardment was probed by the evolution of absorption
spectrum. It was found that the film thickness decreases linearly
with the time of exposure to the ion beam (ca. 15% for 30 min)
and consequently the absorption intensity does decrease.
Therefore, to compare the spectra of samples with different
thickness, the curves are scaled in the region of highest
absorption values (E > 3.75 eV). The excellent spectra
coincidence in this energy region stems from the fact that ion
bombardment does not affect the contribution from band-toband transitions (Figure 7a). Spectral shape and intensity is
independent of the ion beam composition for the same exposure
time, in agreement with XPS results. For any beam exposure
time Eg has the same value found for the as grown material,
indicating that ion bombardment does not produce degeneracy
(Burstein effect).
Even thought the Urbach energy cannot be accurately
calculated due to the narrow range of linearity for ion bombarded
films, it increases upon exposure to ions. This behavior confirms
the increasing structural disorder indicated by the broadening
of XPS spectra. On the other hand, the relative increasing defect
density (ND) induced by ion bombardment can be estimated by
integration of the subband absorption part of the spectra (1.5
eV < E < 3.5 eV). We note, however, that a considerable
amount of band-tail states contribute to the subgap absorption
curve and must be properly subtracted.45,57-59 By following the
standard procedures applied in amorphous semiconductors, one
can write:
ND ) B1.5eV
∫3.2eVRe dE
(1)
where B is a parameter related to the oscillator strength of the
optical transition and Rex(E) is the absorption coefficient related
to transitions to or from defect states with energy in the
forbbiden gap. Then
Rex(E) ) R(E) - R(3.2 eV) exp(hν/Eu)
(2)
The integration shows that the defect density increases first
by a factor ∼4 after 3 min of ion bombardment, remaining
practically constant for longer exposures (Figure 7b). Unfortunately, to obtain absolute ND values this empirical method
requires the knowledge of the parameter B. This coefficient must
be independently determined by other techniques, such as
electron-spin resonance (ESR) in hydrogenated amorphous
silicon.45 In this study because of the lack of knowledge of B
values for TiO2 absolute measure of the film density of states
cannot be obtained, and a relative comparison is only possible.
Figure 8 shows the σ - 1/T logaritmic plots for the as grown
film for photoelectrocatalysis, for different bombardment times.
The conductivity in air appreciably increases upon bombardment; independently of ion beam composition, a saturation value
is achieved after ∼3 min (inset in Figure 8). Although the
absolute conductivity increases, its temperature dependence does
not change appreciably, suggesting that the transport mechanism
is the same. It is worth noting that bombarded material is
unstable and shows an important conductivity hysteresis that
was not observed for the as grown material. This unstability
accounts for maxima and positive slopes in the curves depicted
in Figure 8.
Discussion
Results presented in previous section allow the description
of morphology and electronic properties of TiO2 films, its
chemical composition, and its interface in air or under vacuum.
Titanium Dioxide Photocatalysts
J. Phys. Chem. B, Vol. 104, No. 42, 2000 9855
Figure 5. XPS spectra recorded after tb ) 3, 15, and 30 min Ar+-H2+ beam exposure in the (a) Ti 2p and (b) O 1s binding energy (BE) regions.
Arrows indicate the evolution as the bombardment time tb increases. Spectral deconvolution inset in each figure. Numbers in the spectral deconvolution
of Ti 2p express the oxidation number of Ti.
Figure 7. (a) UV-vis spectra recorded after Ar+-H2+ beam exposure
for (b) 3 min; (∇) 15 min and (4) 30 min. (0) Spectrum recorded
before exposure to ion beam. (b) Dependence of defect density (scaled
to the parameter B, see text) with (0) Ar+-H2+ bombardment time tb.
(x) Value obtained from experiments with Ar+ beam.
Figure 6. Dependence of surface composition (expresed as atomic
fraction, xi) on bombardment time. (a) Ti species, (b) O species.
Crystallinity and Morphology. Films as grown for photoelectrocatalysis are composed by anatase nanocrystals embedded
in an amorphous matrix, as concluded from XRD and AFM
results shown in Figure 1. This picture is related with the
structure of the precursor gel which is composed by a macromolecular network resulting from polymerization of Ti-alkoxide
following hydrolysis and condensation reactions60-63 with
crystalline nanoparticles formed in the first steps of hydrolysis
and amorphous titania formed by random cross-linking of highly
hidroxylated particles and titania oligomers. The final structure
of the precursor gel results from the competition between
polymerization and particle growth and has large cavities filled
with water whose size depends on temperature.64 Thermal
treatment at 90 °C between successive deposits lead to the
formation of a more rigid network that can be thought as a
“sponge” with cavities bigger than 500-600 Å3, even water
inside the cavities is partially expelled as the cross-linking
proceed.64 The final thermal treatment of the material (electrode)
at 450 °C increases water expulsion from cavities and promotes
a higher degree of cross-linking, above the percolation threshold,
with the resulting assembly fully connected. This temperature
is certainly below that corresponding to the amorphous-anatase
transition, which lies between 450 and 700 °C,28,66 and the solid
has an important degree of disorder (amorphicity). Within this
framework, it can be concluded that grains in AFM images are
formed by an amorphous solid which contains small crystallites
dispersed in this amorphous matrix. The low crystallinity is
further reflected in the penetration (width) of band tails in the
gap given by the value of the Urbach energy obtained from
Figure 3, which is six times that calculated for anatase single
crystals. Although one should expect the presence of pores in
the material that would result from the loss of water inside the
cavities, none of the techniques employed in this work allow
their direct detection.
Chemical Composition. From XPS results for the as grown
material (Figure 2) the film surface is composed by Ti4+
coordinated to O2- in a 1.9 O:Ti ratio. When considering in
the calculation the contribution from the peak at 533.7 eV
(Figure 2b) this ratio rises up to 4.5, a value higher than that of
a stoichiometric TiO2 and can be attributed to a surface excess
of oxygen. On defective TiO2 (110) it is generally believed that
molecular oxygen is adsorbed as a negative ion.32,33 Proposed
species include superoxide ion O2-, peroxide ion O22-, and
O-.20,34 Superoxide and peroxide ions have been detected in
photochemical studies involving oxygen and TiO2 powder in
9856 J. Phys. Chem. B, Vol. 104, No. 42, 2000
Bilmes et al.
Figure 9. Schematized density of states for TiO2 films in air and optical
transitions described in the text. Dashed lines indicate band tails for a
crystal.
Figure 8. Conductivity dependence on temperature in air for as (0)
grown TiO2 film and for the same film exposed to (4) 3 min, (∇) 15
min, and (O) 30 min Ar+-H2+ beam. In all cases runs start at the
lowest temperature, as indicated by arrows. Inset: dependence of
conductivity at 295 K with (×) Ar+-H2+ beam bombardment time.
(b) Value obtained from experiments with Ar+ beam.
air or in an aqueous suspension of the oxide,2 and the presence
of O2- species for adsorbed dioxygen on several n-semiconductor oxides has been confirmed byESR.66,67 Surface O2- has
similar unpaired electron spin density in both oxygen atoms
giving an electron density uniformly distributed over the entire
molecular ion. Then, the electron density on each oxygen atom
in Ti-O2- must be lower than that of those coordinated to two
Ti neighbors giving a higher binding energy for the O 1s electron
core level, as found.
In a recent paper, Henderson et al.33 identify O2- species
adsorbed on Ti(110) at 120 K, desorbing at 410 K with a
characteristic stretching frequency and electronic transitions
consistent with those of inorganic Ti-O2- complexes. In this
case, the formation of surface superoxide is associated with the
presence of oxygen vacancy sites with enough electron density
for transferring the necessary charge to form Ti-O2- bonds.
Despite the differences between both TiO2 systems, films grown
by sol-gel in air contain O2 either occluded or adsorbed, then
on this defective material there is a considerable density of
surface Ti-O2- complexes that allow their detection by XPS.
Oxygen binding should be stronger than that for O2 adsorbed
onto a clean oriented surface; in agreement, O2- is released
slowly even at higher temperatures (Figure 5 b). From the area
of XPS peaks and taking into account the O2- contribution, the
O/Ti ratio is 4.5. This value is higher than that expected for the
fully covered surface containing each Ti atom bonded to one
O2-. Although it was suggested that each surface oxygen
vacancy site is responsible for binding up to three O2 molecules,33 this value may also be assigned to the presence of pores
in the material. However, it is not possible to draw a definite
conclusion on the presence of pores with diameter enough to
support the inclusion of molecular oxygen in the bulk material
from XPS experiments sensing 3 nm from the surface.
O2- modifies the density of reduced Ti species and strongly
influences charge transport through the film. O2- is an efficient
electron scavenger,25 and its presence prevents the detection by
XPS of reduced Ti species in the surface layer analyzed.
Moreover, even the O(1s)/Ti(2p) ratio of 1.9 indicates the
presence of oxygen vacancies, reduced Ti species are found in
XPS experiments only after exposure to ion beams that increase
the density of defects and structural disorder (Figure 5). Ion
bombardment of films used in photoelectrocatalysis removes
O2-, and the preferential sputtering of O atoms gives rise to
oxygen vacancies and to more abundant reduced Ti species, as
shown in Figures 5 and 6. On the other hand, partial removal
of O2- by heating at 200 °C does not increase noticeably the
amount of reduced Ti species at the surface, that remain below
the detection limit. The fact that XPS results for ion bombarded
films and TiO2 single crystals are coincident reflects the lack
of sensitivity of this technique to the long-range environment
for each atom.
Electronic Structure. The noncrystallinity of the material
employed for photoelectrocatalysis implies that electron distribution exceeds the simplest semiconductor picture of conduction
and valence bands separated by an energy gap. The experimental
evidences demonstrate the existence of localized states, bands
of defects and band-tails penetrating into the forbidden gap, that
give rise to the optical transitions schematized in Figure 9.
Transition (1) produces optical absorption in region (a) above
3.75 eV; from a chemical point of view this is a O2p f Ti3d
charge-transfer transition. Absorption between 3.25 and 3.7 eV
(region (b)) is assigned to transitions 2 and 3, i.e., between band
tails. Exponential or parabolic band tails are typical of amorphous solids, and the degree of amorphicity can be parametrized
by the value of Urbach energy EU. These band tails result from
the distortion of octhaedral units that weakens Ti-O bonds
reducing the splitting between bonding and nonbonding levels
which give rise to a distribution of states near the band edges.
The increase of disorder induced by ion bombardment, among
other effects, increases the number of distorted cells, spreading
band tails more deeply into the gap and increasing EU. Although
EU could not be calculated with accuracy for ion bombarded
films in Figure 7 its value is close to that for amorphous TiO2-x
films produced by reactive sputtering54 shown in Figure 3.
Another consequence of intrinsic disorder in TiO2 is the
presence of oxygen vacancies and dangling bonds (generalized
as defects) that reduce the cation charge and coordination
number leading to localized states in the gap. Depending on
their density and energy distribution, these states can merge into
a midgap band, and allow low energy excitation pathways, as
indicated by transitions 4 and 5 in Figure 9. Although localized
Titanium Dioxide Photocatalysts
states also arise from grain boundaries, their density ratio should
be in the order of the crystallite/grain volume ratio (see Figure
1). Absorption below 3.2 eV increases with ion bombardment
exposure (Figure 7), following the same trend as the density of
reduced Ti species (Figure 5); thus, these bands can be assigned
to transitions between occupied and unoccupied Ti 3d orbital
(d-d transitions). Occupied Ti3d orbital are Ti3+ species
associated with adjacent oxygen vacancies VO, leading to a
2Ti3+-VO complex with poorly screened electrons. Electronic
transitions involving these defects on ion bombarded TiO2(110)
surfaces lead to a broad feature in electron energy loss spectra
over a range between 0.5 and 2.5 eV,20,34 in agreement with
the absorbance measured in PDS experiments in the subgap
region.
In the as grown material Ti3+ species are detected by PDS,
but not by XPS. The higher penetration depth of visible light
makes PDS less sensitive to the removal of surface Ti3+ by
association with O2 that leads to Ti4+-O2- surface complex.
Electrical Properties. Electron transport in either TiO2 films
active in photocatalysis or modified by ion bombardment can
be analyzed on the basis of the electronic structure sketched in
Figure 9. The material is a nonstoichiometric n-type semiconductor (TiO2-x), and its conductance depends on the density of
donor Ti3+ states located close to the conduction band, i.e.,
ionizable impurities. Assuming full donors ionization, the
density of carriers at the conduction band is given by the bulk
and/or surface oxygen vacancies. The low slope of log(σ) vs
1/T plots is typical of amorphous materials. It markedly differs
from the slope observed in polycrystalline, microcrystalline and
nanocrystalline TiO2 films where transport is mainly controlled
by the height of intergrain barriers.55,68 The conductivity
dependence on temperature does not change significantly by
ion bombardment (Figure 8). This result suggests that bombardment does not alter the transport mechanism by hopping between
localized states. Therefore, it is important to elucidate if the
observed conductivity changes upon heating or ion bombardment of the as grown TiO2 film are due to bulk and/or surfaces
effects. The penetration depth in our experimental conditions
is beyond 200 Å;69 it can be thought therefore that the oxygen
vacancies shown by XPS in Figure 5 (an essentially surface
technique) are true bulk features. However, from PDS of ion
bombarded films (Figure 7) the increase of defect density is
derived to be only about 2 orders of magnitude. To explain the
increase in conductivity of over 5 orders of magnitude, it is
necessary to consider the conductivity changes due to the
depletion region formed beneath the surface by electron
scavenging.25 As discussed above, adsorbed dioxygen is fully
removed after 30 min ion bombardment, and half of the initial
O2- present on the as grown material is removed by heating
the sample at 200 °C in a vacuum (Figure 2b). Removing O2under these conditions increases the conductivity by 3 orders
of magnitude, as shown in Figure 4. Therefore, one can conclude
that for TiO2 films active in photocatalysis the conductivity is
mainly controlled by the space charge layer formed by chemisorption of dioxygen.
As shown in Figure 7b, the density of defects increases first
and remains essentially constant after few minutes of ion
bombardment. The correlation between density of states and
conductivity confirms that an occupied Ti3d narrow band of
states is located below the bottom of the conduction band, whose
density increases on removing lattice O atoms. Indeed, this is
similar to the “tailing” effect found in heavily doped crystalline
semiconductors.51 Therefore, the increasing conductivity on ion
doses is associated with two effects: (i) the increasing density
J. Phys. Chem. B, Vol. 104, No. 42, 2000 9857
of states near to the bottom of the conduction band, and (ii) the
increasing population of this band at the surface by O2
desorption.
Conclusions
The behavior of TiO2 films prepared by sol-gel is similar
to an amorphous semiconductor with low disorder and density
of states as compared with a pure amorphous material. The
electronic structure, schematized in Figure 9, differs from that
of crystalline TiO2 by the exponential band tails and the
existence of a narrow band close to the conduction band. The
latter is fully or partially occupied depending on the presence
of adsorbed electron scavengers such as dioxygen. The main
effect of bombardment is to increase the penetration of band
tails into the gap (increase of distorted cells) and the number
of states in the narrow band (increase of defects).
Surface oxygen vacancies support non dissociate oxygen
adsorption that remove the excess of electrons associated with
the vacancies leading to a poorly conducting material. This
makes this material as very suitable for air photocatalysis in
which the hole transfer to any organic molecule must be
accompanied by electron transfer to an appropriate acceptor.
However, one expects that in aqueous environment water
adsorption removes surface dioxygen, and in consequence
efficient photocatalysis for water remediation needs the presence
of other electron acceptors or its drift from the surface by an
external electric field. The question of the role of gap states
and defect band in the efficiency of photocatalysis under bias
remains open.
Acknowledgment. Work supported by Grants UBACyT TX02, CONICET-PIP- 387, and ANPCyT PICT 3844. Exchange
between University of Campinas, SP, Brazil, and University of
Buenos Aires was partially founded by the network CYTED
VIIIG. FAPESP and CNPq supported part of the work in
Campinas. The authors thank Marcelo Otero for AFM measurements, Galo Soler Illia and Matı́as Jobbaggy for XRD measurements, and P. Hammer for assistance with the IBAD deposition
system, and Miguel A. Blesa for critical reading of the
manuscript.
References and Notes
(1) Carey, J. H.; Lawrence, J.; Tosine, H. M. Bull. EnViron. Contam.
Toxicol. 1976, 16, 697.
(2) Serpone, N.; Pellizzetti, E. Eds. Photocatalysis Fundametals and
Applications; Wiley Interscience: New York, 1989.
(3) Legrini, O.; Braun, A. M.; Oliveros, E. Chem. ReV. 1993, 93, 671.
(4) Bahnemann, D.; Cunningham, J.; Fox, M. A.; Pelizzetti, E.; Pichat,
P.; Serpone, N. In Aquatic and Surface Photochemistry; Helz, G., Zepp,
R., Crosby, D., Eds.; CRC Press: Boca Raton, 1994.
(5) Hoffmann, M. R.; Martin, S. T.; Choi, W.; Bahnemann, D. W.
Chem. ReV. 1995, 95, 69.
(6) Heller, A. Acc. Chem. Res. 1995, 28, 503.
(7) Pozzo, R. L.; Baltanás M. A.; Cassano, A. E. Catal. Today 1997,
39, 219.
(8) Kim, D. H.; Anderson, M. A. EnViron. Sci. Technol. 1994, 28, 479.
(9) Kamat, P. V. Prog. Inorg. Chem. 1997, 44.
(10) Candal, R. J.; Zeltner, W. A.; Anderson, M. A. J. AdV. Oxidat.
Technol., in press.
(11) Mandelbaum, P.; Bilmes, S. A.; Regazzoni, A. E.; Blesa, M. A.
Solar Energy 1998, 65, 75.
(12) Shiga, A.; Tshujiko, A.; Ide, T.; Yae, S.; Nakato, Y. J. Phys. Chem.
B 1998, 102, 6049-6055.
(13) Judin, V. P. Chemistry in Britain 1993, 5, 503.
(14) Brown, G. E.; Henrich, V. E.; Casey, W. H.; Clark, D. L.; Eggleston,
C.; Felmy, A.; Goodman, D. W.; Grätzel, M.; Maciel, G.; McCarthy, M.
I.; Nealson, K. H.; Sverjensky, D. A.; Toney, M. F.; Zachara, J. M. Chem.
ReV. 1999, 99, 77.
(15) Göpel, W.; Schierbaum, K. D. In Sensors Vol. 2, 1991, Göpel, W.;
Hesse, J.; Zemel, J. N. Eds.; VCH: Weinheim; p 430.
9858 J. Phys. Chem. B, Vol. 104, No. 42, 2000
(16) O’Regan; Grätzel, M. Nature 1991, 353, 737.
(17) Bach, U.; Lupe, D.; Grätzel, M. Nature 1998, 395, 183.
(18) Paz, Y.; Luo, Z.; Rabenberg, L.; Heller, A. J. Mater. Res. 1995,
10, 2842.
(19) Bechinger, C.; Ferrere, S.; Zaban, A.; Sprague, J.; Gregg, B. A.
Nature 1996, 383, 608-610.
(20) Henrich, V. E.; Cox, P. A. The Surface Science of Metal Oxides
Cambridge University Press: 1996, New York.
(21) Hugenshmidt, M. B.; Gamble, L.; Campbell, C. T. Surf. Sci. 1994,
302, 329.
(22) Henderson, M. A. Surf. Sci. 1994, 319, 315.
(23) Brinkley, D.; Dietrich, M.; Engel, T.; Farrall, P.; Gantner, G.; Shafer,
A.; Szuchmacher, A. Surf. Sci. 1998, 395, 292.
(24) Mao, Y.; Schöneich, Ch.; Asmus, K. J. Phys. Chem. 1991, 95,
10080.
(25) Göpel, W.; Rocker, G.; Feierabend, R. Phys. ReView 1983, 28, 3427.
(26) Yoshimura, N.; Itoi, M.; Sato, S.; Taguchi, H. Electrical Eng. Japan
1992, 112, 10.
(27) Könenkamp, R.; Wahi, A.; Hoyer; P. Thin Solid Films 1994, 246,
13.
(28) Xu, W. X.; Zhu, S.; Zhou, J.; Fu, X. C.; Zhao, X. N. J. Chem. Soc.
Faraday Trans. 1997, 93, 4187.
(29) Gerfin, T.; Grätzel, M.; Walder, L. Prog. Inorg. Chem. 1996, 44,
346.
(30) Casarin, M.; Maccato, C.; Vittadini, A. J. Phys. Chem. B 1998,
102, 10745.
(31) Bates, S. P.; Kresse, G.; Gillan, M. J. Surf. Sci. 1997, 385, 386.
(32) Lu, G.; Linsenbigler, A.; Yates J. T. J. Chem. Phys. 1995, 102,
3005.
(33) Henderson, M. A.; Epling, W. S.; Perkins, C. L.; Peden, H. F.;
Diebold, U. J. Phys. Chem. B 1999, 103, 5328.
(34) Göpel, W.; Anderson, J. A.; Frankel, D.; Phillips, K.; Schäfer, J.
A.; Rocker, G. Surf. Sci. 1984, 139, 333.
(35) Rusu, C. N.; Yates, J. T. Langmuir 1997, 13, 4311.
(36) Zaban, A.; Meier, A.; Gregg, B. A. J. Phys. Chem. B. 1997, 101,
7985.
(37) Cao, F.; Oskam, G.; Searson, P. C.; Stipkala, J. M.; Heimer, T. A.;
Farzad, F.; Meyer, G. J. J. Phys. Chem. 1995, 99, 11974.
(38) Boschloo, G.; Fitzmaurice, D. J. Phys. Chem. B. 1999, 103, 2228.
(39) Mandelbaum, P. A.; Bilmes, S. A.; Regazzoni, A. E.; Blesa, M. A.
J. Phys. Chem. B 1999, 103, 5505.
(40) Mohamed, M. H.; Sadeghi, H. R.; Henrich, V. E. J. Phys. Chem.
B 1988, 37, 8417.
(41) Pétigny, S.; Mostéfa-Sba, H.; Domenechini, B.; Lesniewska, E.;
Steinbrunn, A.; Burgeois, S. Surf. Sci. 1998, 410, 250.
(42) Epling, W. S.; Peden, C. H. F.; Henderson, M. A.; Diebold, U.
Surf. Sci. 1998, 412/413, 333.
Bilmes et al.
(43) Warren, D. 1954, X-ray Diffraction Procedures for Polycrystalline
and amorphous Materials, Chapmann Hall, NY.
(44) Hammer, P.; Vitoria, N. M.; Alvarez, F. J. Vac. Technol. A 1998,
16, 2941.
(45) Jackson, W. B.; Amer, N. M. Phys. ReV. 1982, 25, 5560.
(46) Alvarez, F.; Sebastiani, M. J. Appl. Phys. 1992, 71.
(47) Sullivan, J. L.; Saied, S. O.; Bertoti, I. Vacuum 1991, 42, 18.
(48) Atashbar, H. T.; Sun, B.; Gong, W.; Wlodarski, R.; Lamb Thin
solid Films 1998, 326, 238.
(49) Moulder, J. F.; Stickle W. F.; Sobol, P. E.; Bomben, K. D.
Handbook of X-ray Photoelectron Spectroscopy; Chastain, J., Ed.; PerkinElmer Corp. : Prairie, Minnesota, 1992.
(50) Street, R. Hydrogenated Amorphous Silicon; Cambridge University
Press: Cambridge, 1991.
(51) Pankove, J. I. Optical Processes in Semiconductor; Dover Publications, Inc.: New York, 1971.
(52) Mott, N.; Davis, E. A. Electronic Processes in Noncrystalline Solids,
2nd ed.; Clarendon Press: Oxford, 1979.
(53) Tang, H.; Berger, H.; Schmid, P. E.; Lévy, F. Solid State Comm.
1994, 92, 267.
(54) Radecka, M.; Zakrzewska, K.; Czternastek, H.; Stapinski Appl. Surf.
Sci. 1993, 65/66, 227.
(55) Demetry, C.; Shi, X. Solid State Ionics 1999, 118, 271.
(56) Pan, J. M.; Maschoff, B. L.; Diebolt, U.; Madey, T. E. J. Vac. Sci.
Technol. A 1992, 10, 2470.
(57) Cuerins, H.; Favre, M. Amorphous Silicon and Related Materials;
Fritzsche, H., Ed.; World Scientific: Singapore, 1988.
(58) Jackson, W. B.; Amer, N. M.; Boccara, A. C.; Fournier, D. Appl.
Opt. 1981, 20, 1333.
(59) Amato, G.; Fizzoti, F. Phys. ReV. B 1992, 45, 44.
(60) Brinker, C. J.; Scherer, G. W. Sol-Gel Science; Academic Press:
New York, 1990.
(61) Sanchez, C.; Ribot, F. New J. Chem. 1994, 18, 1007.
(62) Livage, J.; Sanchez, C. J. Non-Cryst. Solids 1992, 145, 11.
(63) Bischoff, B. L.; Anderson, M. A. Chem. Mater. 1995, 7, 1772.
(64) Marchi, M. C.; Bilmes, S. A.; Negri, R. M. Langmuir 1997, 13,
3665.
(65) Martin, N.; Rousselot, C.; Rondot, D.; Palmino, F.; Mercier, R.
Thin Solid Films 1997, 300, 113.
(66) Anpo, M.; Tomonari, M.; Fox M. A. J. Phys. Chem. 1989, 93,
7300.
(67) Schvets V. A.; Kazanski V. B. J. Catal. 1972, 25, 123.
(68) Balachandran, U.; Eror N. G. J. Mater. Sci. 1988, 23, 2676.
(69) Biersack, J. B.; Haggmark, G. L. Nucl. Instrum. Methods 1980,
174, 257.