Insights into photoexcited electron scavenging processes on TiO2 obtained from studies of the reaction of O2 with OH groups adsorbed at electronic defects on TiO2(110)

Article abstract of DOI:10.1021/jp0262113

In this study we show that molecular oxygen reacts with bridging OH (OH_br) groups formed as a result of water dissociation at oxygen vacancy defects on the surface of rutile TiO₂(110). The electronic structure of an oxygen vacancy defect on TiO₂(110) is essentially the same as that of electron trap states detected on photoexcited or sensitized TiO₂ photocatalysts, being Ti³⁺ in nature. Electron energy loss spectroscopy (EELS) measurements, in agreement with valence band photoemission results in the literature, indicate that water dissociation at oxygen vacancy sites has little or no impact on the electronic structure of these sites. Temperature programmed desorption (TPD) measurements show that O₂ adsorbed at 120 K reacts with near unity reaction probability with OH_br groups on TiO₂(110) to form an unidentified intermediate that decomposes to generate terminal OH groups at nondefect sites. Commensurate with this process, the electronic defect associated with the original oxygen vacancy defect (Ti³⁺) is oxidized. Vibrational EELS results indicate that the reaction between O₂ and OH_br occurs at about 230 K, whereas electronics EELS results suggest that charge is transferred away from the vacancies at 90 K. Detailed TPD experiments in which the precoverage of water was varied indicate that chemisorption of O₂ at cation sites on the TiO₂(110) surface is not required in order for the reaction between O₂ and OH_br to occur, which implies a direct interaction between weakly bound (physisorbed) O₂ and the OH_br groups. In agreement with this conclusion, we find that second-layer water, which selectively hydrogen-bonds to bridging O^2- sites and bridging OH groups, blocks the reaction of O₂ with OH_br groups and prevents oxidation of the vacancy-related Ti³⁺ electronic state. These results suggest that the electron scavenging role of O₂ in photocatalysis may involve a direct reaction between O₂ and trapped electrons located at bridging OH groups. Our studies suggest that the negative influence of high water concentrations in gas-phase heterogeneous photocatalysis studies results from hydrogen-bonded water blocking access of O₂ to trapped electrons located at surface OH groups.

Full text of DOI:10.1021/jp0262113

534
J. Phys. Chem. B 2003, 107, 534-545
Insights into Photoexcited Electron Scavenging Processes on TiO2 Obtained from Studies of
the Reaction of O2 with OH Groups Adsorbed at Electronic Defects on TiO2(110)
Michael A. Henderson,* William S. Epling, Charles H. F. Peden, and Craig L. Perkins‡
†
EnVironmental Molecular Sciences Laboratory, Pacific Northwest National Laboratory,
P.O. Box 999, MS K8-93, Richland, Washington 99352
ReceiVed: May 30, 2002; In Final Form: September 30, 2002
In this study we show that molecular oxygen reacts with bridging OH (OHbr) groups formed as a result of
water dissociation at oxygen vacancy defects on the surface of rutile TiO
an oxygen vacancy defect on TiO (110) is essentially the same as that of electron trap states detected on
photoexcited or sensitized TiO photocatalysts, being Ti in nature. Electron energy loss spectroscopy (EELS)
2
(110). The electronic structure of
2
3+
2
measurements, in agreement with valence band photoemission results in the literature, indicate that water
dissociation at oxygen vacancy sites has little or no impact on the electronic structure of these sites. Temperature
programmed desorption (TPD) measurements show that O
probability with OHbr groups on TiO (110) to form an unidentified intermediate that decomposes to generate
terminal OH groups at nondefect sites. Commensurate with this process, the electronic defect associated with
2
adsorbed at 120 K reacts with near unity reaction
2
3
+
the original oxygen vacancy defect (Ti ) is oxidized. Vibrational EELS results indicate that the reaction
between O and OHbr occurs at about 230 K, whereas electronic EELS results suggest that charge is transferred
away from the vacancies at 90 K. Detailed TPD experiments in which the precoverage of water was varied
indicate that chemisorption of O at cation sites on the TiO (110) surface is not required in order for the
reaction between O and OHbr to occur, which implies a direct interaction between weakly bound (physisorbed)
and the OHbr groups. In agreement with this conclusion, we find that second-layer water, which selectively
2
2
2
2
O
2
2-
hydrogen-bonds to bridging O sites and bridging OH groups, blocks the reaction of O
2
with OHbr groups
and prevents oxidation of the vacancy-related Ti electronic state. These results suggest that the electron
scavenging role of O in photocatalysis may involve a direct reaction between O and trapped electrons located
at bridging OH groups. Our studies suggest that the negative influence of high water concentrations in gas-
3
+
2
2
phase heterogeneous photocatalysis studies results from hydrogen-bonded water blocking access of O
trapped electrons located at surface OH groups.
2
to
1
. Introduction
also be populated by photoexcited electron transfer from
9
,25-31
an adsorbed sensitizer, such as a dye molecule
or a
Several
The primary role of molecular oxygen in photooxidation of
3
2,33
narrow band gap semiconductor particle.
organic molecules over TiO2 photocatalysts is to scavenge
photoexcited electrons that would otherwise deactive the pho-
tocatalyst by negative charge accumulation.1 The species
resulting from this electron scavenging process are typically
7
,8,12,24,26,32,34-36
groups
have observed that electron trap states
susceptible to scavenging by O2 are located at the surface of
-5
3
+
TiO2 as Ti -related sites.
-
6-12
The ability of oxygen to scavenge photoexcited electrons has
been shown to resemble the reaction of oxygen with thermally
generated electronic defects on anatase and rutile. For example,
assigned to O2 or HO2.
These species may also play a direct
1-3,13-15
role in oxidation of organics.
Formation of HO2 requires
the presence of water/OH. Fu et al.16 recently reviewed the
6
-
Anpo et al. found that the EPR signature of O2 formed after
photoexcitation of O2 adsorbed on a polycrystalline TiO2 film
was the same as that observed in the dark after exposure of O2
literature on the role of water in photocatalysis over TiO2, and
there appears to be much speculation about the role of water/
OH in photochemical processes on TiO2.
-
to the thermally activated film. The thermal stability of O2 in
both cases was the same, with the O2 EPR signal absent after
heating to 423 K. Jenkins and Murphy recently observed that
the same O2 species (on the basis of EPR data) forms at 100
It is generally held that O2 scavenges photoexcited electrons
-
4
,17-20
from the TiO2 conduction band (CB); recent research
3
7
suggests that the photoexcited electrons are first trapped in
shallow band gap states on the subpicosecond time scale prior
to any O2-related chemistry. Measurements of the trapped
-
K on both thermally annealed and UV irradiated rutile TiO2
3
5
electron lifetimes have ranged from picoseconds to min-
surfaces. Kormann et al. also observed that the interactions
of O2 with anatase and rutile under UV illumination were similar
to the O2/TiO2 interactions observed for thermal activation in
the dark. In both cases (light and thermal activation), the authors
utes7
,17,20-24
depending, in part, on whether electron acceptors
(such as O2) are present. The electron trap states on TiO2 can
3
+
*
Corresponding author. E-mail: ma.henderson@pnl.gov.
Present address: EmeraChem (formerly GLET), 2375 Cherahala Blvd.,
proposed that O2 reacted with Ti surface sites to make HO2
†
3+
species. In the photoexcitation case, the Ti sites were formed
Knoxville, TN 37932.
‡
from photoexcited electron trapping, whereas in the thermal
activation case these sites were formed from reduction of the
Present address: National Renewable Energy Laboratory, 1617 Cole
Blvd, MS 3215, Golden, CO 80401-3393.
1
0.1021/jp0262113 CCC: $25.00 © 2003 American Chemical Society
Published on Web 12/17/2002

Photoexcited Electron Scavenging Processes on TiO2
J. Phys. Chem. B, Vol. 107, No. 2, 2003 535
surface. We recently showed that O2 adsorption at low tem-
perature on “thermally activated” rutile TiO2(110) results in
-
38,39
formation of O2 species at oxygen vacancy sites.
The
formation of this species at oxygen vacancies on TiO2(110) has
also been modeled by de Lara-Castells and Krause4
0,41
using
periodic ab initio Hartree-Fock calculations. Our studies show
-
that the O2 species is stable on the surface up to 410 K, at
which point it desorbs as O2. This is in general agreement with
-
6
the stability of O2 observed via EPR by Anpo et al.
In this study, we examine how water influences the interaction
of O2 with electrons trapped at oxygen vacancies on TiO2(110)
in an effort to better understand the electron scavenging role of
O2 in photocatalysis on TiO2. Our results show that O2 “titrates”
both charge and hydrogen atoms associated with bridging OH
(hereafter referred to as “OHbr”) groups associated with water
dissociation at oxygen vacancy sites on TiO2(110). However,
hydrogen-bonding of a water molecule to these bridging OH
blocks the ability of O2 to reaction with the OH groups and to
scavenge the electrons associated with defect.
2
. Experimental Section
Experiments in this study were performed in two separate
4
2,43
UHV systems,
using TiO2(110) crystals from several dif-
ferent vendors (First Reaction, Marketech International, Com-
mercial Crystal Growers, and Princeton Scientific). Data
obtained from this distribution of crystals were consistent with
each other, indicating that any vendor-specific variations in
crystal quality did not noticeably impact the surface chemistry
of water on TiO2(110). The as-received crystals were mounted
Figure 1. H
approximately 1.4 ML of H
vacancy sites. The “filled” spectrum results from the same conditions
except that the water adlayer was preheated to 370 K and recooled to
2
O TPD spectrum (m/e ) 18) obtained after exposure of
O at 120 K to TiO (110) with 14% oxygen
2
2
1
20 K prior to performing TPD.
in a manner described previously,4 cleaned by Ar ion
sputtering and annealing in UHV at 850 K. No contaminants
were detected by Auger electron spectroscopy (AES) or X-ray
photoelectron spectroscopy (XPS) after the sputter/anneal treat-
ments. Furthermore, a sharp (1 × 1) pattern was routinely
observed in low energy electron diffraction (LEED) after
annealing. Signatures of the presence of surface oxygen vacan-
cies were evident in temperature programmed desorption (TPD),
electron energy loss spectroscopy (EELS) and XPS after bulk
reduction of the crystal to a light blue color. The surface oxygen
vacancy population varied depending mostly on the level of
bulk reduction, which more-or-less tracked the intensity of the
crystal’s blue color and was used as a rough indicator of the
level of bulk reduction.45 The exact relationship between the
level of bulk reduction and the surface vacancy population on
TiO2(110) has not yet been addressed in the literature. However,
4
+
gases. Gases were typically exposed to the TiO2(110) surface
using a directional doser, which minimized background effects.
In some experiments, oxygen or water was dosed by backfilling
the chamber. H2O was purified using liquid nitrogen (LN2)
freeze-pump-thaw cycles, and O2 was passed through a LN2
trap prior to being dosed. TPD spectra from both chambers were
performed with a 2 K/s ramp. Vibrational (HREELS) and
electronic (EELS) electron energy loss measurements were
performed in the specular scattering geometry with primary
electron beam energies of 8 and 20-25 eV, respectively. No
electron-induced decomposition was observed during these
experiments. The flux of electrons at the sample in these
experiments was on the order of 0.1 nA, which is insignificant
to produce detectable levels of electron-induced decomposition
products during the course of the electron energy loss measure-
ments. (At an incident flux of 0.1 nA and an assumed DEA
37
Jenkins and Murphy have correlated the relative concentration
of Ti3 in the bulk and surface of TiO2 powders using EPR as
a function of annealing temperature in a vacuum. These authors
found that as the annealing temperature was increased, the ratio
of bulk-to-surface Ti3 increased from about 0.05 at 773 K to
about 0.5 at 1073 K. For TiO2(110), the progression toward
increasing levels of bulk reduction results in the surface
reconstructing to form a (1 × 2) structure that possesses Ti2O3-
like features (see refs 46 and 47 and references therein) instead
of forming additional vacancies. The level of bulk reduction at
which this occurs is not known; however the (1 × 2) surface
was not observed on any of the TiO2(110) crystals used in this
study.
+
-16
2
cross section of 10
cm for water, it would require ap-
proximately 20 days of continual electron exposure to ac-
cumulate 0.1 monolayer of electron-induced decomposition
products.) All electron energy loss spectra were recorded at 120
K.
+
3
. Results and Discussion
3.1. Interaction of Water with Oxygen Vacancies on TiO2-
(110). Figure 1 shows TPD data for a 1.4 ML coverage of water
adsorbed at 120 K on a TiO (110) surface possessing about 14%
2
14
oxygen vacancy sites. (One monolayer (ML) equals 5.2 × 10
2
The oxygen vacancy population for each crystal (at its
respective point in the bulk reduction “life cycle” of TiO2) was
sites per cm .) Features at 178 (off scale), 270 and 520 K are
assigned to desorption from the second layer, the first layer and
oxygen vacancy sites, respectively. The former two TPD states
result from molecularly adsorbed water hydrogen-bonded to
bridging oxygen sites and chemisorbed at nondefect cation sites,
respectively, whereas the 520 K TPD state is from water
easily calibrated using H2O TPD, as discussed in previous
studies.3
9,48-50
For each TPD experiment (unless otherwise
stated), the crystal was annealed in UHV at 850 K for 10 min
and subsequently exposed at 90-120 K to research-grade purity

536 J. Phys. Chem. B, Vol. 107, No. 2, 2003
Henderson et al.
vacancies is an occupied band gap state located at about 0.8
eV below the Fermi level (which is close to the CB edge).60
This energy position is in the range observed for the position
of electron trap states generated via direct photoexcitation of
TiO2 or via charge transfer to TiO2 from a photoexcited
sensitizer. Photoelectrochemical measurements place the po-
3
+
tential energy of these surface Ti -related electron trap states
between 0.5 and 0.8 eV below the TiO2 CB edge,2
1,27,31,61-63
whereas transient photoabsorption studies place the position at
about 2 eV below the CB edge on the basis of observations of
a broad absorption feature at about 600-650 nm.7
,24,32,34,64,65
The latter studies assume that excitation of trapped electrons is
to the TiO2 CB edge. However, given that the first maximum
in Ti 3d density of states (for rutile) is located at about 1 eV
6
6
above the CB edge, the position of the trapped electron states
as derived from these optical absorption measurements is more
likely located at about 1 eV below the CB edge, in general
agreement with photoelectrochemical measurements. In this
sense, the chemistry of oxygen vacancies on TiO2(110) can be
used to study electron scavenging processes that occur during
photocatalysis on TiO2.
The EELS spectrum of a thick ice layer on TiO2(110) is
shown in Figure 2b. All loss features of the underlying TiO2-
(110) surface are missing because the scattered electrons “see”
only the vacuum-ice interface and not the ice-TiO2(110)
interface. Instead, the spectrum reflects the band gap of ice
Figure 2. EELS spectra from the interaction of water and oxygen on
TiO (110) with 14% oxygen vacancy sites: (a) the clean surface; (b)
multilayer H O; (c) 1 ML of H O; (d) 1 ML of H O after preheating
to 370 K; (e) 1 ML of H O after preheating to 590 K; (f) 1 ML of H
after exposure to O at 700 K. All spectra were recorded in the specular
(about 7.5 eV) and the first electronic excitation process of
2
2
2
2
molecular water, namely the 1b1 f 4a1 transition at about 8.3
6
7-69
2
2
O
eV.
2
The EELS spectrum of approximately 1 monolayer of H2O
on TiO2(110) with 14% vacancies is shown in Figure 2c. This
water coverage was obtained by preheating a multilayer
exposure to 180 K, which desorbed all multilayer water and
virtually all second-layer water, leaving only the 270 and 520
K TPD states (see TPD data in Figure 1). As mentioned above,
the 270 K TPD state is associated with desorption of water
direction at 120 K with an electron beam energy of 25 eV. Spectra are
displaced vertically for clarity.
dissociatively adsorbed at vacancies.48-50 (Nondefect sites on
TiO2(110) are inactive for water dissociation under UHV
4
4,48,51-54
conditions.
) Water dissociation at vacancies results in
two bridging OHbr groups per each vacancy site. One of these
OHbr groups can be thought of as comprising oxygen from the
water molecule adsorbed at the vacancy, whereas the second
results from hydrogen transfer to a neighboring bridging oxygen
4
+
molecularly adsorbed at five-coordinate Ti sites, whereas the
20 K TPD peak arises from recombinative desorption of OHbr
5
groups. The EELS spectrum of 1 ML of H2O on TiO2(110)
(Figure 2c) reveals an intense loss feature at 6.2 eV and a weak
loss feature at about 8.5 eV. Both of these features are associated
with molecularly adsorbed water at five-coordinate Ti4 sites
because they are absent after heating above 300 K (see Figure
2
-
anion site (O br). In principle, these two OHbr groups are
indistinguishable. The coverage in the 520 K H2O peak,
compared to that in the 270 K state, provides an accurate
measure of the surface oxygen vacancy population. A surface
possessing only dissociative water at vacancies can be prepared
by preheating a multilayer exposure of water to 370 K, as shown
by the filled spectrum in Figure 1.
+
2
d). On the basis of the EELS spectrum of ice (Figure 2b), the
weak 8.5 eV feature may be assigned to the 1b1 f 4a1 transition
in molecularly adsorbed water. Because the 1b1 f 4a1 transition
is the lowest energy transition of molecular water and because
no transitions relating to gap states were detected in the EELS
spectrum of ice (Figure 2b), the loss feature at 6.2 eV for 1
ML of H2O probably results from an excitation between water
and the oxide. This might correspond to either excitation from
the highest occupied molecular orbital of water (1b1 state) into
the oxide’s conduction band (CB) or excitation from the oxide’s
valence band (VB) to the lowest unoccupied molecular orbital
of water (4a1). On the basis of published photoemission spectra
We examined the interaction of water with oxygen vacancies
using EELS to address the issue of whether the electronic
structure of vacancies is influenced by water/OHbr. Figure 2
shows EELS data for the interaction of water with a TiO2(110)
surface possessing about 14% oxygen vacancy sites. The EELS
spectrum for the clean surface is shown in Figure 2a. Loss
intensity above 3.4 eV, which peaks at about 5.2 eV, is
associated with the band-to-band transition in the TiO2(110)
surface, whereas the nearly as intense loss feature at about 0.9
eV is from excitation of the shallow band gap state associated
with oxygen vacancies. The intensity of the 0.9 eV loss, in
comparison with the band-to-band transition intensity, indicates
that the excitation probability of the defect state (with 25 eV
electrons) is considerably greater than that of the band-to-band
5
3,60,70
of monolayer water on TiO2(110),
which provide the
location of the 1b1 orbital relative to the VB edge, the 1b1 f
CB edge transition should be between 7 and 8 eV, which
precludes assignment of the 6.2 eV loss to a 1b1 f CB
transition. In fact, inverse photoemission results by See et al.66
indicate that the first maximum in unoccupied Ti 3d states
should be at about 1 eV above the CB edge, implying that the
maximum loss intensity for a 1b1 f CB transition should be at
about 8-9 eV. Assuming that the 8.5 eV loss may be assigned
transition. This is consistent with the 0.9 eV loss resulting from
excitation of a polaron state.5
5-58
These results are also
5
9
consistent with photoemission studies by numerous groups.
The valence band photoemission signature of surface oxygen

Photoexcited Electron Scavenging Processes on TiO2
J. Phys. Chem. B, Vol. 107, No. 2, 2003 537
to the water 1b1 f 4a1 molecular transition, then the 4a1 orbital
can be placed at about 0.5 to 1.5 eV above the TiO2 CB edge.
A transition from the O 2p-derived maximum density-of-states
nearest the VB edge, located at about 5 eV below the VB edge
on the basis of photoemission (see references in ref 59), to the
4
a1 orbital would give a loss feature at about 6 eV. This
transition essentially amounts to the formation of a temporary
negative ion state of water (albeit in the vicinity of a hole in
the oxide’s valence band).
Regardless of the origins of the 6.2 and 8.5 eV losses arising
from 1 ML of H2O on TiO2(110) in Figure 2c, it is important
to note that the 0.9 eV loss feature is only slightly attenuated
by adsorption of 1 ML of H2O at 115 K. This implies that
electronic structure of the vacancies is not strongly coupled to
electronic states of water or OHbr. Heating 1 ML of H2O on
TiO2(110) to 370 K desorbs all molecularly adsorbed water (see
TPD data of Figure 1), leaving only OHbr groups from H2O
dissociation at the vacancies. The EELS spectrum of this surface,
shown in Figure 2d, shows that the losses associated with
molecularly adsorbed water are absent and that the 0.9 eV loss
feature is essentially unchanged from that observed prior to
water adsorption (Figure 2a). Recombinative desorption of the
OHbr groups (as water) by heating to 590 K (Figure 2e) also
did not significantly change the 0.9 eV loss feature. On the basis
of these results, water adsorption and dissociation at vacancies
does not “oxidize” the oxygen vacancy-related electronic state
that is associated with the 0.9 eV EELS loss, in agreement with
Figure 3. H
OHbr groups (about 0.28 ML) on TiO
at 120 K. The surface with only OHbr groups (TPD of which is shown
in the lower trace) was prepared by preheating a multilayer H O
exposure to 370 K and recooling to 120 K prior to O
are displaced vertically for clarity.
2
O TPD spectra (m/e ) 18) obtained after exposure of
(110) to various amounts of O
2
2
48,50,51
60
previous TPD
and photoemission results. In other words,
little or no charge transfer occurs between the vacancy electronic
2
3
+
state (Ti ) and the vacancy-bound OHbr. In contrast, the 0.9
eV loss feature is absent after exposure of the surface to oxygen
at temperatures above about 150 K (see Figure 2f for spectrum
obtained after O2 exposure at 700 K). O2 adsorption below 150
K also results in depletion of the 0.9 eV loss feature; however,
a new loss is observed at 2.8 eV associated with an adsorbed
2
exposure. Spectra
in the “preheated” TPD trace of Figure 1. TPD of 0.28 ML of
OHbr on TiO2(110), reproduced from Figure 1, is shown in
Figure 3 as the “0 langmuir O2” condition. (One langmuir (L)
equals an exposure of 1 × 10 Torr s.) This surface was
regenerated by vacuum annealing prior to each O2 exposure
experiment shown in Figure 3. As a function of increasing O2
exposure (bottom to top in Figure 3), the H2O TPD state at 520
K due to recombination of OHbr groups is progressively
-
6
-
38
O2 species.
3
.2. Reaction of O2 with OHbr Groups at Oxygen Vacan-
cies on TiO2(110). Surface hydroxyls have been proposed to
be involved in electron scavenging processes during photoca-
talysis over TiO2. On the basis of pH dependent changes in
photooxidation reaction rates, Poznyak et al.71 proposed that
surface OH groups mediate the photoexcited electron transfer
“
titrated” by reaction with O2 and replaced by a sharper H2O
TPD state at about 300-320 K. On the basis of our previous
39,49
work,
this new H2O TPD peak can be assigned to combina-
10
from TiO2 to chemisorbed O2-related species. Liu et al. showed
that the amount of HO2 formed from O2 scavenging of dye-
photosensitized TiO2 increased and then decreased as the amount
of water in a methanol-based suspension was increased. Also,
two groups have suggested that electron traps are associated
with OH groups. Szczepankiewicz et al.22 proposed that
tion of two terminal OH (OHt) groups adsorbed at nondefect
five-coordinate Ti sites that typify the ideal TiO2(110) surface.
The combination of two OHt groups to form water leaves O
adatoms on the surface at the five-coordinate Ti sites. Oxygen
4+
4+
adatoms are also formed from O2 dissociation at oxygen
38,49
vacancies
and may be responsible for the “bright spots”
3
+
photoexcited electrons were trapped at the surface as Ti -
OH groups, and this trapping was facilitated by the ability of
OH groups to dissipate the resulting energy through O-H
related vibrations. Similarly, Corrent et al.72 suggested that
changes in the luminescence of trapped electrons brought about
by hydration/dehydration indicates that some electron traps are
4+
observed on the Ti rows in STM by Diebold and co-workers
73
after exposure of TiO2(110) to oxygen at room temperature
(RT). The new TPD peak due to OHt recombination remained
sharp until the O2 exposure exceeded about 1 L, at which point
virtually all OHbr groups were consumed and a 410 K O2 state
appeared in TPD (see Figure 4) similar to that seen for low-
3
+
associated with Ti -OH groups.
38,39
temperature adsorption of O2 on the clean surface.
(The
In this section we explore the interaction of O2 with OH
groups formed from water dissociation at oxygen vacancy sites.
Figure 3 shows H2O TPD results from the reaction of O2 with
OHbr groups on a TiO2(110) crystal with 14% oxygen vacancies.
In each experiment, a fresh TiO2(110) surface was prepared by
annealing in UHV at 850 K for 10 min. The TiO2(110) surface
was then exposed to a multilayer coverage of water and
preheated to about 350 K to remove all molecularly adsorbed
water leaving only roughly 0.28 ML of OHbr groups, as shown
broad O feature at about 200 K that increases with exposure is
2
3
9
mostly, but not entirely, due to O adsorption on the sample
2
holder, which arises from the fact that O2 was dosed by
backfilling. The amount of O desorbing in this region is
2
considerably less when O2 is dosed through a directional
3
8
doser. ) The coincident broadening of the 300-320 K H2O
TPD peak and the appearance of the 410 K O2 TPD peak implies
that the chemisorbed O2 perturbed the recombination of OHt
groups. The 410 K O2 TPD state arises in Figure 4 from an

538 J. Phys. Chem. B, Vol. 107, No. 2, 2003
Henderson et al.
Figure 5. H
triangles, from data in Figure 4) TPD peak area data from the reaction
of various amounts of O with 0.28 ML of OHbr groups on TiO (110).
Empty triangles: TPD peak area in the 410 K O TPD peak from
2 2
O (filled circles, from data in Figure 3) and O (filled
2
2
2
2
O on the clean surface possessing 10% oxygen vacancies (taken
from 3 ). The dashed lines are drawn to guide the eye.
8
coverage decreased steeply as a function of O2 exposure during
the first 0.4 L of dosed O2. Above an O2 exposure of 0.4 L, the
remaining OH coverage was associated with the 400-420 K
H2O TPD state, and not due to remaining OHbr groups, as can
be seen from the TPD spectra in Figure 3. Using the H2O TPD
peak area data from O2 exposures between 0 and 0.4 L, the
slope of a least-squares fit indicates that 0.66 OHbr groups were
titrated per incident O2 molecule in this exposure range. Given
that the initial sticking coefficient of O2 on the clean surface
Figure 4. O
performed in Figure 3. Note that for simplicity not all of the O
data corresponding to each experiment in Figure 3 are shown in Figure
(and vice versa).
2
TPD spectra (m/e ) 32) from the same experiments
2
TPD
4
O2- surface species that is formed by charge transfer from the
vacancy-related electronic state to an O2 molecule adsorbed at
(
with vacancies but without OH/H2O) at 120 K is about 0.5,38
38
these data indicate that the reaction probability of an adsorbed
O2 molecule with an OHbr group on TiO2(110) is near unity.
As will be shown below, the O2 + OHbr reaction does not occur
at 120 K, and may proceed through an OHbr-O2 complex that
does not involve a chemisorbed O2 species. The high reaction
probability between O2 and OHbr on TiO2(110) is remarkable
because it implies that OHbr groups formed at trapped electron
sites on TiO2 photocatalysts should have a short residence time
in the presence of O2. It should be noted, however, that the
high reaction probability seen from the data of Figure 5 is linked
to the O2 sticking coefficient, which decreases as the adsorption
cation sites next to (but not at) the vacancy sites. Theoretical
calculations by de Lara-Castells and Krause4
0,41
predict that very
little electron density is required to stabilize O2 at these sites,
leading to the conclusion that the responsible species is better
δ-
described as a O2 species, where δ < 1. The possibility that
δ-
this O2 species is responsible for oxidation of the vacancies
and reaction with the OHbr groups will be discussed below.
Figure 3 also shows that a second, considerably broader H2O
TPD state at 400-420 K results from reaction of O2 and OHbr.
This state is most apparent in TPD for O2 exposures above about
0
.4 L. Although the identity of the species is not known, it is
3
8
temperature is increased.
presumably OH-related and might be from OHt groups that have
7
4
diffused to or formed at step sites. Using STM, Suzuki et al.
Figure 5 also shows the change in the 410 K O2 TPD peak
area as a result of the reaction with OHbr. On the clean TiO2-
(110) surface (empty triangles), O2 adsorbed at low exposures
is destined for oxidization of vacancy sites. (These data were
taken from a previous study in which the surface vacancy
have recently observed features at step edges on TiO2(110) as
a result of water exposure at RT that they have proposed to be
OHt groups. Given that the species responsible for the H2O TPD
feature at 300-320 K (which we assign to OHt groups at
nondefect sites) should not be stable on the TiO2(110) surface
at RT in UHV, the 400-420 K H2O TPD feature can tentatively
be assigned to a combination of OHt groups at step edges. This
assignment is consistent with previous TPD results on sputter/
annealed TiO2(110).48
The TPD data in Figures 3 and 4 were used to estimate the
reaction probability of an incident O2 molecule with an OHbr
group on TiO2(110). Figure 5 shows the change in the OHbr
coverage as a function of O2 exposure at 120 K. The OHbr
coverage was estimated using the H2O TPD peak area above
3
8
coverage was 10%. ) Adsorbed O2 molecules in excess of the
vacancy population (in the clean surface case, occurring for
exposures above 0.8 L) are reversibly bound at five-coordinate
4
+
Ti sites next to the vacancies (along the [1 1h 0] direction) and
evolve in TPD at 410 K. These cation sites are not strictly 4+
but possess some degree of electron density associated with the
4
0,75-77
oxygen vacancies.
The onset of O2 desorption from O2
adsorbed in the presence of OHbr (solid triangles in Figure 5)
coincides with the point at which the OHbr coverage has
minimized (circles), and is consistent with that observed on the
clean surface, with the noted consideration that the vacancy
populations were different in the two sets of experiments (10%
4
00 K, which tends to overestimate the coverage of the OHbr
groups because this lower limit also captures a portion of the
00-420 K H2O TPD state. The initial OHbr coverage (in the
3
8
4
for the clean surface and 14% in this study). Assuming that
the O2 sticking probabilities at 120 K are similar in the two
cases (O2 on the surface with and without vacancies filled with
absence of postdosed O2) was 0.28 ML, or twice the water
coverage in the 520 K TPD peak. Figure 5 shows that the OHbr

Photoexcited Electron Scavenging Processes on TiO2
J. Phys. Chem. B, Vol. 107, No. 2, 2003 539
18
Figure 6. H
2
16O (m/e ) 18, solid traces) and H
2
O (m/e ) 20, dashed
with 0.28 ML
(110). Lower traces are TPD in the absence
exposure, whereas the middle and upper traces are TPD spectra
Figure 7. Fourier deconvoluted HREELS spectra from the reaction
of O with OH groups on TiO (110) possessing 14% oxygen vacancy
2 br 2
sites. The traces labeled (a) are for about 1.2 ML of H O adsorbed at
2
traces) TPD spectra obtained after reaction of 0.08 L O
2
1
6
of OHbr groups on TiO
of O
obtained after exposure of
2
2
1
6
18
120 K. Trace (b) is from the surface in (a) preheated to 375 K, which
isolates only the OHbr groups on the surface. Trace (c) is from exposure
2 2
O and O , respectively. The upper and
middle spectra are displaced vertically for clarity.
of the surface in (b) to 2 L O
in (c) preheated to 230 K. Trace (e), in contrast, is from a switch in
the order of O /H O adsorption, in which approximately 1 ML of H
is adsorbed after oxygen vacancy sites were first filled by exposure to
L O at 120 K (with a preheating step to 575 K to desorb molecularly
2
at 120 K. Trace (d) is from the surface
OHbr groups) and taking into account the different vacancy
populations of the two samples, the similar onsets for O2 TPD
uptakes imply that roughly the same amount of O2 is consumed
per vacancy in the low exposure regime for the two cases. In
other words, just as one O2 molecule is consumed per vacancy
on the clean surface,38 it appears that one O2 is consumed per
vacancy when the vacancies are filled with dissociated water
molecules. Because all of the OHbr groups are consumed, the
ratio of O2 consumed per OHbr group is 1:2.
2
2
2
O
4
2
adsorbed O ). Spectra were recorded at 120 K, and are displaced
2
vertically for clarity.
eventually decomposes to 18OHt groups. On the basis of studies
1
,10-12,14,35
on TiO2 photocatalysts,
it seems likely that this
intermediate is a HO or H O species. It should also be noted
2
2
2
1
8
18
Additional insights into the reaction of O2 with OHbr groups
were obtained using isotopic labeling studies. Figure 6 shows
2 2
that at higher O exposures (above 1 L), only O was
observed in the 410 K TPD peak (no O or O O). This
1
6
16 18
2
1
6
18
16
H2 O and H2 O TPD traces for OHbr groups alone (bottom),
indicates that the 410 K O2 TPD peak is due to molecularly
adsorbed O2, and not recombinative O2 desorption, in agreement
16
16
after exposure of 0.08 L O2 to OHbr groups (middle), and
after exposure of 0.08 L 18O2 to OHbr groups (top). In each
case, the clean surface oxygen vacancy population was 14%.
In these experiments, O2 was dosed by backfilling the chamber,
and as a result the walls of the mass spectrometer ionizer region
were unavoidably incorporated with O, resulting in small
amounts of spurious 16O- O scrambling in subsequent water
TPD experiments (see bottom and middle sets of traces). A
similar effect occurred in the opposite direction with desorption
16
38
18
with results from the clean surface. The origin of O2
exchange with polycrystalline TiO2 surfaces during UV expo-
18
78,79
80
sure, observed for example by Courbon et al.
and by Sato
16 18
16
to yield O O and O2, may not be linked to chemistry
occurring on (110) terminations or may not be observable in
UHV on TiO2(110) because of a “pressure gap”.
18
18
HREELS analysis also provides insights into the reaction of
O2 and OHbr on TiO2(110). Figure 7 shows Fourier deconvoluted
18
16
of H2 O registering H2 O signal due to oxygen scrambling
8
1
with 16O on the walls of the mass spectrometer. (The origin of
HREELS spectra for various combinations of H2O and O2 on
a TiO2(110) surface possessing about 14% oxygen vacancies.
The traces shown in Figure 7a correspond to the deconvoluted
1
6
18
this problem was easily diagnosed by observing the O: O
ratio in the desorption of multilayer water. ) Despite this
experimental artifact, it is clear that the reaction of O2 with
49
1
8
2 2
spectra for 1.2 ML of H O. The 1.2 ML of H O surface was
1
6
prepared by adsorption of a thick H O multilayer followed by
OHbr groups results a significant portion of the 300 K water
2
18
heating to 180 K, which desorbed all the multilayer and most
TPD peak containing O, but virtually none of the unreacted
OHbr groups that recombinatively desorb at 520 K being
incorporated with O (top set of Figure 6). This result indicates
52
of the second-layer water. The presence of some second-layer
18
water is evidenced by the broad ν(OH) feature centered at 3250
1
8
16
-1
-1
that O2 does not displace OHbr groups from vacancies onto
cm , whereas the more intense feature at 3500 cm is due to
five-coordinate Ti sites, but that 18O2 abstracts hydrogen from
4+
52
molecularly adsorbed water in the monolayer. (Note that there
16
the OHbr groups to form an unknown surface intermediate that
is intensity in the loss region of each spectrum in Figure 7a-f

540 J. Phys. Chem. B, Vol. 107, No. 2, 2003
Henderson et al.
-1
at about 2200 cm . This may be due to an artifact of the
deconvolution process or it may result from background CO
adsorption.)
Two significant differences are present in the HREELS
spectrum of water on TiO2(110) with 14% vacancies (Figure
7
a) versus what was reported earlier for water on the nearly
52
defect-free TiO2(110) surface. The first difference is that the
phonon remnants” (the intensity leftover in the first multiple
“
-1
loss features at 1205 and 1520 cm after Fourier deconvolution)
are considerably more intense than what was observed on the
nearly defect-free TiO2(110) surface. In this case, the δ(HOH)
mode of water, marked with an asterisk in the “x100” spectrum
of Figure 7a, is barely discernible from the 1520 cm- phonon
remnant, whereas the δ(HOH) loss was more intense than the
1
5
2
phonon remnant on the nearly defect-free surface. This
comparison suggests that the intensity in the multiple phonon
remnant increases as the surface oxygen vacancy population
increases. Wulser and Langell82 have proposed that the intensity
left in the first set of multiple phonons after Fourier deconvo-
lution results from non-Poisson behavior in the scattering of
electrons from oxides, likely as a result of combinations of
impact and dipole excitation events. In this case, the probability
of exciting a phonon by impact scattering increases as the
population of surface electronic defects increases. The second
difference between the HREELS spectrum in Figure 7a and that
52
observed earlier on nearly defect-free TiO2(110) is the presence
Figure 8. EELS spectra from the interaction of water and oxygen on
TiO (110) with 10% oxygen vacancy sites. The solid line trace
2
corresponds to the surface after annealing at 850 K. The dashed line
trace is after decomposition of water in the oxygen vacancies, which
generates about 0.2 ML of OHbr groups. The dotted line trace
corresponds to the spectrum obtained after reaction of the 0.2 ML of
-
1
of a more intense 3665 cm loss in the present case. As it
turns out, this difference is also linked to the presence of a much
higher population of oxygen vacancy defects in the present case
(
about 14%) versus the trace level present in the earlier study.
53
Recently, Brookes and co-workers reassigned the small 3690
cm loss observed in the previously published HREELS spectra
2
OHbr groups with O at 90 K. Spectra were recorded in the specular
direction at 90 K with an electron beam energy of 25 eV.
-1
52
by the present lead author to OHbr groups from water
dissociation at oxygen vacancies on the basis of their STM
results (as opposed to the previously proposed assignment of
OH groups at steps). Given the intensity of the 3665 cm loss
in Figure 7a, we concur with the reassignment of this loss by
Brookes and co-workers. Inspection of the relative ν(OH)
physisorbed state.38,39 Heating the surface of Figure 7c to 230
-
1
K resulted in almost complete attenuation of the 3665 cm
-
1
loss (see Figure 7d), indicating that almost all of the OHbr groups
were removed. The product of the reaction of O2 and OHbr,
however, is not evident in HREELS. As shown in Figure 7d,
-
1
intensities for molecularly adsorbed water (at 3500 cm ) and
-
1
the ν(OH) region (3200-3800 cm ) is essentially devoid of
-1
for OHbr (at 3665 cm ) indicates that the excitation probability
of the latter is roughly twice that of the former on the basis of
the relative coverages of the two species. This may indicate
that the O-H bond direction in the OHbr groups is oriented
normal of the surface, whereas the O-H bond directions of
molecularly adsorbed water are inclined from the surface normal.
The presence of the ν(OH) loss at 3665 cm- due to OHbr
groups in Figure 7a suggests that water dissociates at oxygen
vacancies at or below 180 K (the preheating temperature used
to desorb multilayer waterssee above). This is consistent with
the STM results of Schaub et al.54 who also observed water
dissociation at vacancies on TiO2(110) at 180 K. In contrast,
loss features despite the fact that TPD clearly shows that no
water has desorbed below 230 K (see top TPD trace of Figure
-1
3
). The absence of a ν(OH) loss at 3500 cm indicates that
the reaction of O2 and OHbr does not generate molecularly
adsorbed water. We therefore propose that the reaction product
of O2 and OHbr at 230 K possesses O-H bonds that are
significantly inclined parallel to the surface so as to render them
with little or no dynamic dipole moment normal to the surface
1
(a necessity for dipole scattering in HREELS). Such a condition
may arise from hydrogen-bonding interactions, for example,
4+
between neighboring OHt groups at five-coordinate Ti sites.
Alternatively, an intermediate species such as HO2/H2O2 may
be formed which then decomposes to OHt groups.
5
3
Brookes et al. proposed that water did not dissociate at
vacancies until 290 K on the basis of the temperature at which
OH features became apparent in their VB photoemission spectra.
Finally, Figure 7e provides additional support for the 3665
cm-1 loss being due to OHbr groups. In this HREELS spectrum,
The HREELS spectrum in Figure 7b shows that all molecu-
larly adsorbed water is desorbed after heating the surface to
the TiO2(110) surface was pretreated with 4 L O2 at 120 K,
followed by heating to 535 K prior to water exposure at 120
K. The O2 adsorption and preheating treatments resulted in all
-
1
375 K, leaving only the 3665 cm loss associated with the
3
8,49
OHbr groups. Exposure of these OHbr groups to 2 L O2 at 120
oxygen vacancies being oxidized.
The HREELS spectrum
-
1
K did not remove the 3665 cm loss (Figure 7c), although it
broadened the loss feature somewhat. This suggests that O2 and
OHbr groups interact sufficiently at 120 K to affect the latter’s
ν(OH) mode, but that a reaction does not take place. This
behavior is consistent with previous observations that at low-
temperature O2 is adsorbed weakly to the surface, possibly in a
of 1 ML of H2O on this surface (Figure 7e) shows the presence
-
1
of molecularly adsorbed water at 3500 cm , but no sign of
-
1
loss intensity at 3665 cm from OHbr groups, consistent with
the absence of oxygen vacancies.
The reaction between O2 and OHbr was also evident in EELS.
Figure 8 shows EELS spectra collected from a TiO2(110) crystal

Photoexcited Electron Scavenging Processes on TiO2
J. Phys. Chem. B, Vol. 107, No. 2, 2003 541
readily diffuse along the rows of bridging oxygen anions during
heating and recombine to abstract water and form new oxygen
vacancies randomly located along the bridging oxygen rows.48
This implies that either the hydrogen atoms diffuse along the
rows of bridging oxygen sites as H radicals or as protons with
the electron passed from underlying cation to cation as the
2
-
proton is transferred along the row of bridging O sites. It is
unclear whether the hydrogen atoms diffuse independently or
53,54
as pairs, although two recent STM
studies show no evidence
for pairing of the bright spots associated with OHbr groups.
Irrespective of this point, the lower model of Figure 9 conceptu-
ally illustrates the conversion of OHbr groups to OHt groups as
a result of reaction with O2. The model emphasizes three points
from the data: (1) that the oxygen vacancies are “oxidized”,
2
-
that is, the sites are filled with bridging O sites, (2) that each
reacted O2 molecule consumes two OHbr groups, and (3) that
the resulting O-H bonds in the products (presumably OHt
groups and/or HO2/H2O2 intermediates) are tilted from the
surface normal. After desorption of water in the 300-320 K
TPD peak (Figure 3) the surface then possesses a coverage of
4+
oxygen adatoms located at five-coordinate Ti sites equivalent
to the original oxygen vacancy population. These species are
not detected in photoemission or in EELS, but their presence
on the surface is evident in the chemistry of coadsorbates. We
have previously studied the reaction of these oxygen adatoms
4
9
49
83
Figure 9. Proposed model for the reaction of O
2
with OHbr groups on
species with coadsorbed water, ammonia and methanol,
TiO (110). The upper drawing represents a view of the surface with
2
and have shown that they abstract protons from the O-H and
N-H bonds of neighboring coadsorbates. In the case of water,
this reactions yields two OHt groups. Therefore, the final product
from the interaction of O2 with vacancies on the clean surface
is the same as that from O2 reacting with OHbr groups formed
by dissociation of water at vacancies, that being filled vacancies
and deposited oxygen adatoms.
an oxygen vacancy site in the middle the row of bridging O sites. The
middle drawing illustrates the dissociation of water in the vacancy to
form to OHbr groups. The lower drawing illustrates the end point in
2
the reaction between O and OHbr, wherein the vacancy is filled and
4+
two OH groups are formed at adjacent five-coordinate Ti sites.
t
that possessed about 10% oxygen vacancy sites. The EELS
spectrum for the clean surface with 10% oxygen vacancies
exhibits an intense loss at 0.82 eV due to oxygen vacancies.
As was the case of the crystal with 14% oxygen vacancies (see
Figure 2), exposure of this surface to water (followed by
desorption of all molecular water by heating to 380 K) did not
significantly deplete the vacancy-related loss feature (dashed
line trace in Figure 8). Subsequent exposure to O2 (of an
unknown exposure) at 90 K (dotted line trace in Figure 8)
resulted in a nearly complete attenuation of the vacancy loss
feature. Heating the surface to 230 K and above (data not shown)
did not change the EELS spectrum. In contrast to the vibrational
data shown in Figure 7, the electronic spectra presented in Figure
These results provide evidence for one mechanism by which
O scavenges trapped electrons on TiO photocatalysts. The
2
2
overall stoichiometry for the scavenging of electrons by reaction
of O with OH groups formed at oxygen vacancies sites is
2
br
illustrated in the following reactions:
-
2-
-
vac(2e ) + O
+ H O(g) f 2OH (e )
(e180 K)
br
2
br
-
2-
2
OH (e ) + O (g) f 2 O
br
+ 2OHt
(90-230 K)
br
2
2
OH f H O(g) + O
a
(300-320 K)
t
2
where “vac” and “O2-br” correspond to oxygen vacancy and
8
suggests that the charge-transfer process between O2 and OH-
2-
bridging O anion sites, respectively, and the OHt and Oa
filled vacancies takes place at 90 K.
The results in Figure 8 provide evidence that the reaction of
O2 with OHbr is accompanied by the equivalent of electron
4+
species are located at nondefect five-coordinate Ti sites. (The
temperature references are the points at which each of these
processes is observed.) Two electrons are localized at each
vacancy on the clean surface, and thus one per OHbr group after
dissociation of water at the vacancies. On the basis of these
charge counting arguments, the OHt and Oa species are charge
neutral as, for example, water would be if adsorbed at the same
sites. Although the resulting OHt groups are presumably less
thermodynamically stable than the OHbr groups (on the basis
of their respective recombinative desorption temperatures), the
thermodynamic basis for the conversion of OHbr groups to OHt
3+
scavenging of the trapped electrons at the Ti -OHbr sites.
Figure 9 presents a tentative model for the reaction of O2 with
OHbr groups on TiO2(110). The top model shows a space-filling
representation of the TiO2(110) surface with a single oxygen
vacancy site. Both cations at the vacancy can be thought of as
3
+
being Ti in nature, although it is becoming more apparent
that these electrons are spin polarized and possess some degree
4+
40,75-77
of delocalization to adjacent Ti cations.
Adsorption and
dissociation of water at the oxygen vacancy results in two
equivalent OHbr groups, as shown in the middle model of Figure
2-
groups arises from the greater stability of a O br site than that
of OHbr. The overall stoichiometry for this chemistry is
3
+
9
. As described above, this process does not oxidize the Ti
-
2-
cations at the vacancy. However, because the two OHbr groups
are, in principle, indistinguishable from each other, the electrons
associated with the original Ti3 cations should be delocalized
over both OHbr sites, yielding the equivalent of one electron
per OHbr group. Previous results show that the hydrogen atoms
vac(2e ) + O (g) f O
+ Oa
br
2
+
which, as mentioned above, is the same as that observed for
the oxidation of oxygen vacancies on the clean surface by
3
8,49
O2.

542 J. Phys. Chem. B, Vol. 107, No. 2, 2003
Henderson et al.
Figure 10. (A) D
case followed by a 10 L O
precoverages above 1 ML. The inset to (A) shows the temperature region between 450 and 550 K for one TPD spectrum from each case. (B) O
TPD spectra (m/e ) 32) from the precoverages of D O in the submonolayer regime shown in (A). Spectra are displaced vertically for clarity.
2
O TPD spectra (m/e ) 20) from varying precoverages of D
2
O on a TiO
2
(110) surface possessing 2.5% oxygen vacancies in each
2
2
exposure at 90 K. The lower portion of (A) is for submonolayer precoverages of D O, whereas the upper portion is for
2
2
3
.3. Blocking of the O2 + OHbr Reaction by Preadsorbed
Water. Numerous groups have observed that water has both a
positive and negative effect on photochemical rates at the
16
gaseous-TiO2 heterogeneous interface. Fu et al., who recently
reviewed the literature on this subject, pointed out that photo-
oxidation rates are generally higher in the gas phase compared
to the aqueous phase due in large part to the role of water. For
example, the rates for photooxidation of many gas-phase species
over TiO2 have been observed to increase as the relative
humidity (RH) is increased from 0% to about 30-60% but then
decrease as the RH is increased to saturation.8
4-94
The increase
rates are typically interpreted in terms of hole-related processes,
although some groups have found that water impacts how
photoexcited electrons are scavenged by O2.1
0,22,71,72,95
By and
large, the inhibition effect of high RH has been ascribed to
competition for adsorption sites between water and the reactant
being photooxidized.
In this section we examine how coadsorbed water influences
the reaction of O2 with bridging hydroxyl groups (in this case,
ODbr groups) formed from water (D2O) decomposition at
oxygen vacancies on TiO2(110). Figure 10 shows D2O (A) and
O2 (B) TPD spectra from the reaction of O2 with ODbr as a
function of increasing preadsorbed D2O coverage on a TiO2-
Figure 11. D
2
O (hollow triangles) and O
2
(filled circles) TPD peak
area data taken from the spectra in Figure 10. Only the D
area above 400 K was used to approximate the ODbr coverage, and the
TPD peak area is from the 410 K O TPD peak.
2
O TPD peak
O
2
2
the trailing edges of all the D2O TPD traces (located above 300
K) indicates that the extent of the O2 + ODbr reaction was not
influenced by submonolayer coverages of D2O. Consistent with
this observation, the 500 K D2O TPD peak from recombination
of ODbr groups was absent for D2O precoverages in the
submonolayer regime (see the inset of Figure 10A), indicating
complete conversion of ODbr to ODt. The corresponding O2 TPD
spectra as a function of D2O precoverage (Figure 10B) reveal
that preadsorbed D2O blocked O2 adsorption sites associated
with 410 K O2 TPD state to the extent that no O2 TPD signal
was detected after preadsorption of 0.8 ML of D2O. This effect
is better illustrated in Figure 11 using the peak area from the
410 K O2 TPD state as a function of the D2O precoverage. The
O2 TPD peak area (circles) linearly decreased for increasing
D2O precoverages up to 0.6 ML and then more abruptly
attenuated to zero by the 0.8 mL of D2O precoverage. How-
ever, the TPD peak area in the 500 K D2O state (triangles),
which reflects the ODbr coverage, was near zero and did not
(110) crystal possessing 2.5% oxygen vacancy sites. In these
experiments, D2O was adsorbed at 90 K at coverages in case
excess of that needed to form the ODbr groups. The excess D2O
was not removed from the surface before O2 exposure, as was
the case in the experiments discussed in the previous section.
A near saturation O2 exposure of 10 L at 90 K was used. Any
influence of coadsorbed water on the reaction of O2 with ODbr
groups should be evident by changes in the D2O TPD spectrum.
Starting with the bottom portion of the D2O panel (A), one can
see that the D2O TPD traces possess similar trailing edge profiles
for D2O precoverages between 0.20 and 0.90 ML, with an
increasing D2O peak area that shifts to lower temperature with
increasing coverage. Oxygen vacancies are filled with ODbr
groups for all of these precoverages of D2O, yielding in each
case an ODbr coverage of about 0.05 ML. The coincidence of

Photoexcited Electron Scavenging Processes on TiO2
J. Phys. Chem. B, Vol. 107, No. 2, 2003 543
increase, indicating that the extent of O2 + ODbr reaction was
not influenced by D2O precoverage in submonolayer regime.
In contrast to the absence of an effect for submonolayer
precoverages of D2O, data in the upper panel of Figure 10A
show that D2O precoverages in excess of 1 ML had a significant
effect on the O2 + ODbr reaction. The trailing edge of the 270
K D2O TPD peak shifted toward lower temperature as the
precoverage of D2O was increased above 1 ML to the extent
that the TPD spectra of 1.5 ML of D2O with and without
subsequent O2 exposure were nearly identical. Coincident with
the change in the trailing edge of the 270 K D2O TPD peak,
the inset of Figure 10A shows that the D2O TPD state due to
ODbr recombination reappeared as the D2O precoverage is
increased above the 1 ML point (see also the TPD peak area
data in Figure 11). In the D2O precoverage range between 1
and 1.5 ML, the first layer has filled with D2O and second layer
with D2O (175 K TPD peak) begins to fill. Second-layer water
is adsorbed to the surface through hydrogen-bonding interactions
with the rows of bridging oxygen sites.51,52
Figure 12. Schematic model illustrating the effect of second-layer
water on the reaction of O with OH groups on TiO (110).
2
br
2
Several observations can be gleaned from the data in Figures
0 and 11. First, preadsorbed water does not exhibit a strong
1
4. Implications for Electron Scavenging Processes on
TiO2
preference for five-coordinate cation sites in the vicinity of OHbr/
ODbr groups (the adsorption sites of O2 in the 410 K TPD
The reaction of O2 with OHbr groups formed from water
dissociation at electronic defects on TiO2(110) is similar in two
ways to proposed mechanisms in which O2 scavenges trapped
electrons during heterogeneous photocatalysis over TiO2. First,
3
8
peak ) over similar sites far from these groups. If this were
the case, preadsorption of 0.1 ML of D2O would have
completely blocked all O2 adsorption into the 410 K TPD state,
which was not observed. Second, when preadsorbed water
molecules occupy O2 adsorption sites, subsequent exposure to
O2 does not displace these water molecules. This is evident from
the fact that the O2 TPD peak area progressively decreased as
a function of D2O precoverage as opposed to staying nearly
constant at low D2O precoverages. Finally, and perhaps of
greater importance, the data in Figures 10 and 11 indicate that
the reaction of O2 and ODbr likely involves a direct O2-ODbr
interaction that does not involve the chemisorbed O2 state
associated with the 410 K O2 TPD state. Although first-layer
water had no effect on the O2 + ODbr reaction, second-layer
water prevented O2 from reacting with ODbr groups. The
immediate product of the reaction (DO2 and/or D2O2) may
displace monolayer water or adsorb at cation sites after water
desorption.
3
+
in both cases surface Ti -OH sites are oxidized by O2. In
heterogeneous photochemical studies, several groups have
4+
proposed these sites are formed from electron trapping at Ti -
10,22,71,72,95
OH groups,
whereas on TiO2(110) they result from
water dissociation at oxygen vacancies. Nevertheless, the
resulting surface species in both cases are similar. In the UHV
experiments, the species formed from the reaction of O2 + OHbr
involves hydrogen atom transfer to a physisorbed O2 molecule
and formation of an unidentified intermediate species (presum-
ably HO /H O ) which eventually decomposes to form OH
2
2
2
t
groups. Second, results in this study showing that second-layer
water blocks the reaction of O and OH are consistent with
2
br
the inhibiting influence of water on the rates of photooxidation
8
4-94
18
16
at the gas-TiO2 interface,
as well as the O2- O2
8
0
exchange process observed by Sato during UV illumination
of TiO2. These studies show that photooxidation rates increase
with increasing RH from 0 to 30-60%, but then decrease as
the RH is increased to saturation. The increased rates in the
low RH range have typically been ascribed to enhanced rates
of OH radical production arising from hole-related processes,
whereas overall rate degradation at higher RH levels has been
ascribed to blocking of reactant adsorption sites. Results in this
study open the possibility that second-layer water (which should
be more prevalent at high RH) inhibits the ability of O2 to
scavenge trapped electrons from TiO2 photocatalysts via reaction
The interaction between a second-layer water molecule and
a OHbr/ODbr group should be somewhat different from that of
2
-
the interaction of a second-layer water molecule with a O br
site because ODbr groups should function primarily as hydrogen-
bond donors whereas O2 br sites can only function as hydrogen-
bond acceptors. However, the fact that the extent of the reaction
decreases progressively for D2O precoverages between 1 and
-
1
.5 ML indicates that second-layer D2O has a weak preference
2-
for adsorption at ODbr versus O br sites.
The role that second-layer water plays in blocking the reaction
between O2 and ODbr groups at electronic defects on TiO2
surfaces is schematically shown in Figure 12. In the absence of
second-layer water (left side of Figure 12), incident O2
molecules react with OHbr/ODbr groups to form (ultimately) OHt/
ODt groups and to oxidize the electronic defects (“trapped”
electrons) associated with the original oxygen vacancy sites. In
contrast, an overlayer of water blocks access of O2 to these OHbr/
ODbr groups (right side of Figure 12), presumably by hydrogen-
bonding to the OHbr/ODbr groups, thus preventing charge
transfer to O2 and conversion of OHbr/ODbr groups to OHt/ODt
groups.
3+
with Ti -OH groups. Similar arguments have been applied
to the influence of water on the extent of photon-induced
96-100
adsorption of O2 on TiO2.
Finally, one might ask, if a second layer of water blocks
access of O2 to sites where electron scavenging can take place
then how is it that photooxidation can take place in solution or
at high RH? The answer to this question comes from considering
the dynamics of the two systems discussed here (UHV versus
solution/high RH). The UHV experiments in which second-layer
water is observed to block the O + OH reaction (see Figure
2
br
10) are similar to those of the solution and high RH cases in
that the “coverage” of second-layer water is near saturation.

544 J. Phys. Chem. B, Vol. 107, No. 2, 2003
Henderson et al.
However, in the UHV experiments the rates of adsorption and
desorption of water from the second layer are zero below 170
K, whereas the adsorption/desorption rates of water associated
with what could be termed the “second layer” in solution or at
high RH are nonzero and more-or-less in equilibrium with each
other. Therefore, the UHV situation provides a static “snapshot”
of the dynamic situation taking place in solution or at high RH.
In the latter, O2 can find its way to electron trap states on the
surface because the second-layer water molecules are adsorbing
and desorbing, which provides O2 and other species partial
access to the surface. Therefore, the electron scavenging reaction
does occur in solution and at high RH, although many attempts
by impinging O2 molecules are “turned away” from the surface
when they encounter water hydrogen-bonded to the OH groups
located at electron trap sites. This behavior, which is difficult
to detect in solution, is revealed by the “static” UHV experi-
ments performed here.
(23) Yamakata, A.; Ishibashi, T.-a.; Onishi, H. J. Phys. Chem. B 2001,
05, 7258.
1
2
9
(
24) Asahi, T.; Furube, A.; Masuhara, H. Chem. Phys. Lett. 1997, 275,
34.
(25) Kay, A.; Humphry-Baker, R.; Gr a¨ tzel, M. J. Phys. Chem. 1994,
8, 952.
26) Vinodgopal, K.; Hua, X.; Dahlgren, R. L.; Lappin, A. G.; Patterson,
L. K.; Kamat, P. V. J. Phys. Chem. 1995, 99, 10883.
27) Boschloo, G. K.; Goossens, A. J. Phys. Chem. 1996, 100, 19489.
(
(
(28) Schlichth o¨ rl, G.; Huang, S. Y.; Sprague, J.; Frank, A. J. J. Phys.
Chem. B 1997, 101, 8139.
(29) Haque, S. A.; Tachibana, Y.; Klug, D. R.; Durrant, J. R. J. Phys.
Chem. B 1998, 102, 1745.
(30) Haque, S. A.; Tachibana, Y.; Willis, R. L.; Moser, J. E.; Gr a¨ tzel,
M.; Klug, D. R.; Durrant, J. R. J. Phys. Chem. B 2000, 104, 538.
(
31) Franco, G.; Gehring, J.; Peter, L. M.; Ponomarev, E. A.; Uhlendorf,
I. J. Phys. Chem. B 1999, 103, 692.
32) Gopidas, K. R.; Bohorquez, M.; Kamat, P. V. J. Phys. Chem. 1990,
94, 6435.
(
(
(
(
33) Kamat, P. V. Chem. ReV. 1993, 93, 267.
34) Howe, R. F.; Gr a¨ tzel, M. J. Phys. Chem. 1985, 89, 4495.
35) Kormann, C.; Bahnemann, D. W.; Hoffmann, M. R. J. Phys. Chem.
1
988, 92, 5196.
Acknowledgment. We thank Geoff Thornton for insightful
(36) Wang, H.; He, J.; Boschloo, G.; Lindstroem, H.; Hagfeldt, A.;
-
1
Lindquist, S.-E. J. Phys. Chem. B 2001, 105, 2529.
discussions leading to reassignment of the 3665 cm loss in
Figure 7. This work was supported by the U.S. Department of
Energy Office of Basic Energy Sciences, Division of Materials
Sciences, and by the U.S. Department of Energy Environmental
Management Science Program. Pacific Northwest National
Laboratory is a multiprogram national laboratory operated for
the U.S. Department of Energy by the Battelle Memorial
Institute under Contract DE-AC06-76RLO 1830. The research
reported here was performed in the William R. Wiley Envi-
ronmental Molecular Science Laboratory, a U.S. Department
of Energy user facility funded by the Office of Biological and
Environmental Research.
(
(
37) Jenkins, C. A.; Murphy, D. M. J. Phys. Chem. B 1999, 103, 1019.
38) Henderson, M. A.; Epling, W. S.; Perkins, C. L.; Peden, C. H. F.;
Diebold, U. J. Phys. Chem. B 1999, 103, 5328.
39) Perkins, C. L.; Henderson, M. A. J. Phys. Chem. B 2001, 105,
856.
40) de Lara-Castells, M. P.; Krause, J. L. J. Chem. Phys. 2001, 115,
4798.
(41) de Lara-Castells, M. P.; Krause, J. L. Chem. Phys. Lett. 2002, 354,
83.
42) Henderson, M. A. Surf. Sci. 1994, 319, 315.
43) Herman, G. S.; Peden, C. H. F. J. Vac. Sci. Technol. A 1994, 12,
2087.
(
3
(
4
(
(
(44) Brinkley, D.; Dietrich, M.; Engel, T.; Farrall, P.; Gantner, G.;
Schafer, A.; Szuchmacher, A. Surf. Sci. 1998, 395, 292.
(45) Li, M.; Hebenstreit, W.; Diebold, U.; Tyryshkin, A. M.; Bowman,
M. K.; Dunham, G. G.; Henderson, M. A. J. Phys. Chem. B 2000, 104,
References and Notes
4944.
(
46) Ashino, M.; Sugawara, Y.; Morita, S.; Ishikawa, M. Phys. ReV.
(
(
(
1) Mills, A.; Le Hunte, S. J. Photochem. Photobiol., A 1997, 108, 1.
2) Fox, M. A.; Dulay, M. T. Chem. ReV. 1993, 93, 341.
3) Linsebigler, A. L.; Lu, G.; Yates, J. T., Jr. Chem. ReV. 1995, 95,
Lett. 2001, 86, 4334.
(47) Bennett, R. A.; Stone, P.; Price, N. J.; Bowker, M. Phys. ReV. Lett.
1999, 82, 3831.
(48) Henderson, M. A. Langmuir 1996, 12, 5093.
(49) Epling, W. S.; Peden, C. H. F.; Henderson, M. A.; Diebold, U.
Surf. Sci. 1998, 412/413, 333.
7
4
35.
(
17.
4) Serpone, N.; Khairutdinov, R. F. Stud. Surf. Sci. Catal. 1996, 103,
(5) Ollis, D. F. CATTECH 1998, 2, 149.
(50) Henderson, M. A. Surf. Sci. 1998, 400, 203.
(
51) Hugenschmidt, M. B.; Gamble, L.; Campbell, C. T. Surf. Sci. 1994,
(6) Anpo, M.; Aikawa, M.; Kubokawa, Y.; Che, M.; Louis, C.;
3
8
02, 329.
Giamello, E. J. Phys. Chem. 1985, 89, 5694.
7) Castellano, F. N.; Stipkala, J. M.; Friedman, L. A.; Meyer, G. J.
Chem. Mater. 1994, 6, 2123.
(
(
52) Henderson, M. A. Surf. Sci. 1996, 355, 151.
53) Brookes, I. M.; Muryn, C. A.; Thornton, G. Phys. ReV. Lett. 2001,
(
7, 266103/1.
54) Schaub, R.; Thostrup, P.; Lopez, N.; Laegsgaard, E.; Stensgaard,
I.; Norskov, J. K.; Besenbacher, F. Phys. ReV. Lett. 2001, 87, 266104/1.
55) Egdell, R. G.; Eriksen, S.; Flavell, W. R. Solid State Commun. 1986,
0, 835.
56) Egdell, R. G. Electronic structure of oxide surfaces. In Oxide
Surfaces; Woodruff, D. P., Ed.; Elsevier Science: Amsterdam, 2001; Vol.
, p 550.
(
99.
8) Nakaoka, Y.; Nosaka, Y. J. Photochem. Photobiol., A 1997, 110,
(
2
(9) Konovalova, T. A.; Kispert, L. D.; Konovalov, V. V. J. Phys. Chem.
(
B 1999, 103, 4672.
10) Liu, G.; Zhao, J.; Hidaka, H. J. Photochem. Photobiol., A 2000,
33, 83.
11) Liu, G.; Li, X.; Zhao, J.; Horikoshi, S.; Hidaka, H. J. Mol. Catal.
A: Chem. 2000, 153, 221.
12) Coronado, J. M.; Maira, A. J.; Conesa, J. C.; Yeung, K. L.;
Augugliaro, V.; Soria, J. Langmuir 2001, 17, 5368.
13) Schwitzgebel, J.; Ekerdt, J. G.; Gerischer, H.; Heller, A. J. Phys.
Chem. 1995, 99, 5633.
6
(
(
1
(
9
(
(
57) Eriksen, S.; Egdell, R. G. Surf. Sci. 1987, 180, 263.
58) Kodaira, S.; Sakisaka, Y.; Maruyama, T.; Haruyama, Y.; Aiura,
(
Y.; Kato, H. Solid State Commun. 1994, 89, 9.
59) Henrich, V. E.; Cox, P. A. The Surface Science of Metal Oxides;
Cambridge University Press: Cambridge, U.K., 1994.
60) Kurtz, R. L.; Stockbauer, R.; Madey, T. E.; Rom a´ n, E.; De Segovia,
J. L. Surf. Sci. 1989, 218, 178.
61) Redmond, G.; Fitzmaurice, D.; Gr a¨ tzel, M. J. Phys. Chem. 1993,
7, 6951.
(
(
(14) Wang, Y.; Hong, C.-S. Water Res. 2000, 34, 2791.
(
(15) Cermenati, L.; Pichat, P.; Guillard, C.; Albini, A. J. Phys. Chem.
B 1997, 101, 2650.
16) Fu, X.; Zeltner, W. A.; Anderson, M. A. Stud. Surf. Sci. Catal.
996, 103, 445.
17) Colombo, D. P., Jr.; Bowman, R. M. J. Phys. Chem. 1995, 99,
1752.
(
(
9
1
1
1
(
(
(
62) Boschloo, G.; Fitzmaurice, D. J. Phys. Chem. B 1999, 103, 2228.
63) Siripala, W.; Tomkiewicz, M. J. Electrochem. Soc. 1982, 129, 1240.
64) Rothenberger, G.; Moser, J.; Gr a¨ tzel, M.; Serpone, N.; Sharma, D.
(
(
18) Colombo, D. P., Jr.; Bowman, R. M. J. Phys. Chem. 1996, 100,
K. J. Am. Chem. Soc. 1985, 107, 8054.
65) Serpone, N.; Lawless, D.; Khairutdinov, R. F. J. Phys. Chem. 1995,
99, 16646.
(66) See, A. K.; Bartynski, R. A. J. Vac. Sci. Technol., A 1992, 10,
2591.
(67) Shibaguchi, T.; Onuki, H.; Onaka, R. J. Phys. Soc. Jpn. 1977, 42,
152.
(68) Michaud, M.; Cloutier, P.; Sanche, L. Phys. ReV. A 1991, 44, 5624.
(69) Kobayashi, K. J. Phys. Chem. 1983, 87, 4317.
8445.
(
(19) Furube, A.; Asahi, T.; Masuhara, H.; Yamashita, H.; Anpo, M. J.
Phys. Chem. B 1999, 103, 3120.
(
(
20) Leytner, S.; Hupp, J. T. Chem. Phys. Lett. 2000, 330, 231.
21) Hagfeldt, A.; Lindstroem, H.; Soedergren, S.; Lindquist, S.-E. J.
Electroanal. Chem. 1995, 381, 39.
22) Szczepankiewicz, S. H.; Moss, J. A.; Hoffmann, M. R. J. Phys.
Chem. B 2002, 106, 2922.
(

Photoexcited Electron Scavenging Processes on TiO2
J. Phys. Chem. B, Vol. 107, No. 2, 2003 545
(
70) Krischok, S.; Hofft, O.; Gunster, J.; Stultz, J.; Goodman, D. W.;
Kempter, V. Surf. Sci. 2001, 495, 8.
71) Poznyak, S. K.; Kokorin, A. I.; Kulak, A. I. J. Electroanal. Chem.
998, 442, 99.
72) Corrent, S.; Cosa, G.; Scaiano, J. C.; Galletero, M. S.; Alvaro, M.;
Garcia, H. Chem. Mater. 2001, 13, 715.
73) Diebold, U.; Lehman, J.; Mahmoud, T.; Kuhn, M.; Leonardelli,
G.; Hebenstreit, W.; Schmid, M.; Varga, P. Surf. Sci. 1998, 411, 137.
74) Suzuki, S.; Fukui, K.-i.; Onishi, H.; Sasaki, T.; Iwasawa, Y. Stud.
Surf. Sci. Catal. 2001, 132, 753.
75) Lindan, P. J. D.; Harrison, N. M.; Gillan, M. J.; White, J. A. Phys.
(86) Fu, X.; Clark, L. A.; Zeltner, W. A.; Anderson, M. A. J. Photochem.
Photobiol., A 1996, 97, 181.
(87) Lichtin, N. N.; Avudaithai, M. EnViron. Sci. Technol. 1996, 30,
2014.
(
1
(
(88) Luo, Y.; Ollis, D. F. J. Catal. 1996, 163, 1.
(
(
(
89) Hung, C.-H.; Marinas, B. J. EnViron. Sci. Technol. 1997, 31, 1440.
90) Obee, T. N.; Hay, S. O. EnViron. Sci. Technol. 1997, 31, 2034.
91) Cao, L.; Huang, A.; Spiess, F.-J.; Suib, S. L. J. Catal. 1999, 188,
(
(
4
1
1
8.
(92) Wang, K.-H.; Hsieh, Y.-H.; Lin, C.-H.; Chang, C.-Y. Chemosphere
(
999, 39, 1371.
93) Einaga, H.; Futamura, S.; Ibusuki, T. Phys. Chem. Chem. Phys.
999, 1, 4903.
94) Vorontsov, A. V.; Savinov, E. V.; Davydov, L.; Smirniotis, P. G.
Appl. Catal., B 2001, 32, 11.
95) Szczepankiewicz, S. H.; Colussi, A. J.; Hoffmann, M. R. J. Phys.
Chem. B 2000, 104, 9842.
ReV. B 1997, 55, 15919.
(
(76) Paxton, A. T.; Thien-Nga, L. Phys. ReV. B 1998, 57, 1579.
(77) Bredow, T.; Pacchioni, G. Chem. Phys. Lett. 2002, 355, 417.
(78) Courbon, H.; Herrmann, J. M.; Pichat, P. J. Phys. Chem. 1984, 88,
(
5
5
(
50.
(96) Bickley, R. I.; Stone, F. S. J. Catal. 1973, 31, 389.
(80) Sato, S. J. Phys. Chem. 1987, 91, 2895.
81) Cox, P. A.; Flavell, W. R.; Williams, A. A.; Egdell, R. G. Surf.
(97) Bickley, R. I.; Jayanty, R. K. M. Faraday Discuss 1974, 58, 194.
(
Sci. 1985, 152/153, 784.
(98) Boonstra, A. H.; Mutsaers, C. A. H. A. J. Phys. Chem. 1975, 79,
1694.
(99) Gonzalez-Elipe, A. R.; Munuera, G.; Soria, J. J. Chem. Soc.,
Faraday Trans. 1 1979, 75, 748.
82) Wulser, K. W.; Langell, M. A. Phys. ReV. B 1993, 48, 9006.
1
999, 114, 313.
(
(
84) Sampath, S.; Uchida, H.; Yoneyama, H. J. Catal. 1994, 149, 189.
85) Obee, T. N.; Brown, R. T. EnViron. Sci. Technol. 1995, 29, 1223
(100) Munuera, G.; Rives-Arnau, V.; Saucedo, A. J. Chem. Soc.,
Faraday Trans. 1 1979, 75, 736.